6th International Symposium On Deformation Characteristics of Geomaterials Argetina 2015

DEFORMATION CHARACTERISTICS
OF GEOMATERIALS
Advances in Soil Mechanics and
Geotechnical Engineering
Advances in Soil Mechanics and Geotechnical Engineering (ASMGE) is a peer-reviewed book series
covering the developments in the key application areas of geotechnical engineering. ASMGE will focus on
theoretical, experimental and case history-based research, and its application in engineering practice. The
series will include proceedings and edited volumes of interest to researchers in academia, as well as industry.
The series is published by IOS Press under the imprint Millpress.
Volume 6
Recently published in this series
Vol. 5. A.O. Sfriso, D. Manzanal and R.J. Rocca (Eds.), Geotechnical Synergy in Buenos
Aires 2015 Invited Lectures of the 15th Pan-American Conference on Soil
Mechanics and Geotechnical Engineering and the 8th South American Congress on
Rock Mechanics, 1518 November 2015, Buenos Aires, Argentina
Vol. 4. V.M. Ulitsky, Michael B. Lisyuk and Alexey G. Shashkin (Eds.), Soil-Structure
Interaction, Underground Structures and Retaining Walls Proceedings of the
ISSMGE Technical Committee 207 International Conference on Geotechnical
Engineering
Vol. 3. D.G. Toll, H. Zhu, A. Osman, W. Coombs, X. Li and M. Rouainia (Eds.), Information
Technology in Geo-Engineering Proceedings of the 2nd International Conference
(ICITG), Durham, UK
Vol. 2. Y.-J. Cui, F. Emeriault, F. Cuira, S. Ghabezloo, J.-M. Pereira, M. Reboul, H. Ravel
and A.M. Tang (Eds.), Proceedings of the 5th International Young Geotechnical
Engineers Conference 5th iYGEC 2013
Vol. 1. P. Arnold, G.A. Fenton, M.A. Hicks, T. Schweckendiek and B. Simpson (Eds.),
Modern Geotechnical Design Codes of Practice Implementation, Application and
Development
ISSN 2212-781X (print)

ISSN 2212-7828 (online)
Deforrmatio
on Chaaracterristics
of Geeomateerials
Proceeedings of th
he 6th International Symposium
S m on Deforrmation
C
Characterist tics of Geo
omaterials, IS-Buenoss Aires 20115,
1518 November
N 2
2015, Buennos Aires, Argentina
A
Edited by
y
Victtor A. Rin
naldi
Dep
partment of Civil
C Engineeering, Univerrsidad Nacio
onal de Crddoba,
Argentina
Marceelo E. Zeeballos
Dep
partment of Civil
onal de Crddoba,
Argentina
and
n Jose Cllari
Juan
Dep
partment of Civil
onal de Crddoba,
Argentina
A
Amsterdam Berlin Wa
ashington, DC
C
2015 The authors and IOS Press.
All rights reserved. No part of this book may be reproduced, stored in a retrieval system,
or transmitted, in any form or by any means, without prior written permission from the publisher.
ISBN 978-1-61499-600-2 (print)

ISBN 978-1-61499-601-9 (online)
Library of Congress Control Number: 2015955277
Publisher
IOS Press BV
Nieuwe Hemweg 6B
1013 BG Amsterdam
Netherlands
fax: +31 20 687 0019
e-mail: [email protected]
Distributor in the USA and Canada

IOS Press, Inc.
4502 Rachael Manor Drive
Fairfax, VA 22032
USA
fax: +1 703 323 3668
e-mail: [email protected]
LEGAL NOTICE
The publisher is not responsible for the use which might be made of the following information.
PRINTED IN THE NETHERLANDS

v
Preface
We are pleased to present herein the Proceedings of the 6th International
Symposium on Deformation Characteristics of Geomaterials, held in the city of
Buenos Aires (Argentina) from 15 to 18 November 2015. IS-Buenos Aires
2015 was carried out in parallel with the 8th South-American Conference of
Rock Mechanics (VIII SCRM), the 15th Pan-American Conference on Soil
Mechanics and Geotechnical Engineering (XV PCSMGE), and the 22nd
Argentinean Congress of Soil Mechanics and Geotechnical Engineering
(CAMSIG XXII).
IS-Buenos Aires 2015 was organized under the support of the Argentinean
Geotechnical Society (SAIG) and the International Society for Soil Mechanics
and Geotechnical Engineering (ISSMGE) Technical Committee TC 101,
after the successful previous Symposia first set in Hokkaido 1994, and
thereafter in Torino 1999, Lyon 2003, Atlanta 2008, and most recently Seoul
2011.
This proceeding includes 7 lectures delivered by invited keynote speakers
and the Third Bishop Lecture delivered by Professor Herve Di Benedetto from
the University of Lyon, France, who presented a reference work on the
advanced testing and modeling of bituminous bounded and unbounded granular
materials. A total of 128 articles were presented and discussed during the
symposium. The 118 articles selected for publication in this proceeding were
peer-reviewed by an international review board and 7 additional articles were
published in a special number of the Journal of Soils and Foundation.
The Symposium brought together practitioners, researchers and educators
from all around the world who are engaged in the understanding of the
deformation properties of geomaterials before failure, and the small strain
parameters as fundamental characteristics of geomaterials. The main topics
covered by the symposium include:
1. Experimental investigations from very small strains to beyond failure
including multiphysical approach.
2. HTCM coupling behavior, characterization and modeling of various
geomaterials and Interfaces.
3. Practical prediction and interpretation of ground response: field
observation and case histories.
Buenos Aires is a charming metropolis where history and modernity come
together. It has a wide cultural and tourist life. IS-Buenos Aires-2015 was held
at the Hilton Hotel, in one of the most fashionable and exclusive areas of the
city, Puerto Madero.
vi
We would like to express our gratitude to the International Society of Soil

Mechanics and Geotechnical Engineering (ISSMGE) Technical Committee
TC 101 , the Sociedad Argentina de Ingeniera Geotcnica (SAIG), and all the
sponsors and organizations who generously contributed and supported to the
success of IS-Buenos Aires 2015. We also acknowledge the great effort,
compromise and enthusiasm in cooperation of all members of the local
organizing committee, the international advisory board and the local reviewer
board. We finally thank the editorial board of the Journal of Soils and
Foundations for providing and preparing a special number for this event.
Victor A. Rinaldi
Marcelo E. Zeballos
Juan Jose Clari
Buenos Aires, 1518 November 2015
vii
Conference Organization
General Organizing Committee Conferences Buenos Aires 2015
President:
Alejo O. Sfriso
General Secretariats:
Diego Manzanal
Ricardo Jose Rocca
Vctor Rinaldi
Pablo Torres
Finance:
Mauro Codevilla
Organizing Committee TC101
Conference Chairman
Vctor Rinaldi
General Secretary:
Juan Jos Clari (Jr.)
Administration:
Germn Rodrigo Molina
Julio Capdevila
Pedro Arra
Mara Pa Cruz
Publication:
Marcelo Zeballos
Marcos Montoro
Franco Francisca
Gonzalo Aiassa
Noem Maldonado
Public Relations:
Federico Pinto
Marcelo Eberhardt
Diego Turello
Ignacio Maldonado
Gustavo Bogado
viii
Finance:
Pedro Covassi
Agustn Adami
Paula Vettorelo
International Advisory Committee
David Airey (Australia)

Beatrice Baudet (Hong Kong)
Nilo Consoli (Brazil)
Matthew Coop (Hong Kong)
Antonio Gomes Correia (Portugal)
Federica Cotecchia (Italy)
Herve Di Benedetto (France)
Antonio Viana da Fonseca (Portugal)
Antonio Gens (Spain)
Masayuki Hyodo (Japan)
Erdin Ibrahim (UK)
Richard Jardine (UK)
Dong-Soo Kim (Korea)
Junichi Koseki (Japan)
Lyesse Laloui (Switzerland)
Carlos Santamarina (USA)
Satoru Shibuya (Japan)
Fumio Tatsuoka (Japan)
TC 101 Members
TC Chair:
Junichi Koseki (Japan)
TC Vice Chair:
Lyesse Laloui (Switzerland)
TC Secretary:
Erdin Ibrahim (United Kingdom)
Members:
David Airey (Australia)
An Baertsoen (Belgium)
Beatrice Baudet (Hong Kong)
Jan Boh (Czech & Slovak Republics)
Thomas Brandon (USA)
Tcio Mauro Pereira de Campos (Brazil)
Tim Carrington (United Kingdom)
Choong-Ki Chung (Korea)
Matthew Richard Coop (Hong Kong)
ix
Herv DiBenedetto (France)

Rune Dyvik (Norway)
Jos Estaire (Spain)
Cristiana Ferreira (Portugal)
Apollonia Gasparre (United Kingdom)
Vasiliki Georgiannou (Greece)
Gert Greeuw (Netherlands)
Seyed Mohsen Haeri (Iran)
Martin Holmn (Sweden)
Maosong Huang (China)
Fardin Jafarzadeh (Iran)
Richard Jardine (United Kingdom)
Dong-Soo Kim (Korea)
Vladislava Kostkanov (Czech & Slovak Republics)
Taeseo Ku (Singapore)
Reiko Kuwano (Japan)
Lus Leal Lemos (Portugal)
Hoe Ling (USA)
Mirosaw Lipiski (Poland)
Antonio LLoret (Spain)
Satoshi Nishimura (Japan)
Sadik Oztoprak (Turkey)
Chung Philip (Hong Kong)
Robinson R G (India)
Victor Alejandro Rinaldi (Argentina)
Erza Rismantojo (Indonesia)
J.T. Shahu (India)
Alojzy Szymaski (Poland)
Daniel R. Verastegui (Belgium)
Yangping Yao (China)
Corresponding Members
Ozer Cinicioglu (Turkey)
Antnio Gomes Correia (Portugal)
Kartal Toker (Turkey)
Gauthier Van Alboom (Belgium)
Antonio Viana da Fonseca (Portugal)
x
Peer-Reviewers of Buenos Aires Conferences 2015
Abdelkareem Alzo'ubi Celso Romanel

Abel Jacinto Cesar Pasten
Adam Bezuijen Christian Ledezma
Adinele Guimaraes Christian Mendoza
Adrin Rodrguez-Marek Christophe Dano
Agustin Adami Claudia Medina
Alan Lutenegger Claudio Fernando Mahler
Alejandro Kerguelen Cristina de Santiago
Alejandro Pujol Crysthian Purcino
Alejandro Verri Dante Fratta
Alejo O. Sfriso David Airey
Alessandra Conde De Freitas Deb Mishra
Alexandra Ossa Lpez Diego Lo Presti
Alexandre Gomes Diego Manzanal
Alfonso Mariano Ramos Diego Turello
Alfredo Lpez Dipanjan Basu
Alvaro Gutierrez Edgar Bard
Alvaro Pereira Edgar Rodriguez Rincn
Amrico Fernandez Edmundo Rogerio Esquivel
Andrew Cho Eduardo Botero Jaramillo
Andrzej Niemunis Eduardo Capdevila
Anna dOnofrio Eduardo Luis Martin
Anna Peixoto Eduardo Nuez
Antonio Bobet Emilio Redolfi
Antonio Gomes Correia Enmanuel Carvajal Diaz
Antonio Lloret Ennio Marques Palmeira
Antnio Thom Enrique Asanza
Apollonia Gasparre Erinaldo Hilrio Cavalcante
Aravind Pedarla Ernesto Abril
Arcesio Lizcano Erol Tutumuer
Ariel Abad Esteban Saez
Ariel di Rado Estefana Muoz
Arnaldo Barchiesi Eugenio Mendiguren
Arsenio Negro Ezequiel Zielonka
Asal Bidarmaghz Federico Paz
Augusto Jose Leoni Federico Pinto
Aurea Perucho Federico Pisano
Beatrice Baudet Fernando Lopez
Belayneh Desta Fernando Mantaras
Bernardo Caicedo Fernando Marinho
Bernardo Castellanos Fernando Martinez
Briardo Pacheco Fernando Schnaid
Carlos Carranza-Torres Francisco Arroyo
Carlos Grandas Francisco Mingorance
Carlos Jaramillo Francisco Ruz
Carol Murillo Franco M. Francisca
Carola Sanhueza Freddy Lopez
xi
Gabriele Della Vecchia Juan Jos Clari

Gastn Quaglia Juan Jos Goldemberg
Gerardo Botasso Juan Manuel Fernandez Vincent
Germn Van de Velde Juan Muoz
Gilson Gitirana Jr. Juan Pablo Ibaez
Gonzalo Martn Aiassa Juliane Marques
Gonzalo Montalva Julio Alfredo Capdevila
Goran Vukotic Katia Vanessa Bicalho
Gregg Fiegel Kevin Gaspard
Guillermo Narsilio Ku Taeseo
Guillermo Zavala Kyle Rollins
Haitao Yu Laureano Hoyos
Haitham Dawood Leonardo Guimaraes
Hasan Kazmee Lucas Festugato
Hebe Barber Luciano Oldecop
Heinrich Heinz Jr. Lcio Flvio de Souza Villar
Heraldo Giacheti Lucio Gerardo Cruz
Herve Di Benedetto Luis Alberto Torres Cruz
Hesham El Naggar Luis Felipe Prada
Hugo Bianchetto Luis Lemos
Ignacio Ezama Luis Valenzuela
Ignacio Maldonado Luis Vallejo
Ignacio Vernengo Luiz Sozio
Iaki Garca Mendive Macarena Ayarza
Jaeho Lee Mahdi Disfani
James B. Nevels, Jr. Mahdi Taiebat
Jan Bohac Mahsa Hedayati
Jarbas Milititsky Manuel Ocampo
Javier Buraschi Manuel Pastor
Javier Gonzlez Marcela Gonzalez
Jayhyun Kwon Marcello Teixeira
Jean Michel Pereira Marcelo Eberhardt
Jernimo Caffaro Marcelo Sanchez
Joaqun Cnchero Marcelo Zeballos
John Mccartney Mrcia M. Dos Anjos Mascarenh
Jorge Juan Marcuzzi Marcio Muniz de Farias
Jorge Lain Marco Terzariol
Jorge Ramoneda Marcos Barreto de Mendona
Jos Ale Vliz Marcos Carniglia
Jos Alvarellos Marcos Emanuel Calvente
Jos Antonio Schiavon Marcos Massao Futai
Jos Estaire Marcos Montoro
Jos Fernandez Merodo Maria Claudia Barbosa
Jos Fernando Thom Juc Mara Cristina Herrera
Jose Segovia Maria Do Socorro So Mateus
Juan Carlos Rojas Vidovic Mara Elizabeth Pardini
Juan De Dios Alemn Velsquez Maria Gardoni
Juan Felix Rodriguez Mara Pa Cruz
xii
Maria Teresa Zandarin Paulo Albuquerque

Mariana Mendiguren Paulo Orostegui
Mariano Balbi Paulo Teixeira da Cruz
Marina Casaux Pedro Arduino
Marlisio O. Cecilio Pedro Arra
Martn Boefer Pedro Covassi
Martn Sanchez Pedro D. Marques Priett
Mary Ellen Large Priscilla Paniagua Lopez
Matas Calvente Rafael Baltodano
Matthew Coop Ramiro Verastegui Flores
Mauricio Abramento Ramn Verdugo
Mauricio Ehrlich Raydel Lorenzo
Mauricio Giambastiani Renato Cabral Guimares
Mauro Codevilla Renato Cunha
Mauro Poblete Ricardo Barletta
Mehran Mazari Ricardo Bergan Born
Michle Dal To Casagrande Ricardo Nicolau
Michelle Akin Ricardo Rocca
Miguel Jaramillo Ricardo Schiava
Miguel Martins Stickle Riccardo Corti
Miguel Pando Richards Regueiro
Miguel Soto Robb Moss
Monica Prezzi Roberto Flores
Muhammad Safdar Roberto Morandi
Murad Abu-Farsakh Roberto Terzariol
Nancy Whiting Rodney Collins
Natalio Figueroa Rodrigo Molina
Negar Tavafzadeh Rodrigo Murillo
Nicols Espinoza Rodrigo Salgado
Nicolas Estrada Sabatino Cuomo
Nilo Consoli Sadaf Khosravifar
Noem Graciela Maldonado Sandra Orlandi
Norberto Jorge Cerutti Sandro Machado
Norma Patricia Lopez Acosta Sebastin S. Fernandez
Orencio Vilar Sedat Cetin
Oscar Cordo Sergio A. Reyes
Oscar Giovanon Sergio Diaz
Oscar Moreno Sergio Fontoura
Oscar Taiba Sergio Seplveda
Osvaldo Ledesma Shunyi Christopher Chen
Otavo Santos Silvia Angelone
Pablo Abbona Slvio Romero de Melo Ferreira
Pablo Cuellar Stephen Lane
Pablo Mira McWilliams Stuart Colls
Pablo Torres Sudip Basack
Patricia Hennig Osmari Svetlana Melentijevic
Paul Mayne Tacio de Campos
Paula Vettorelo Tae Sup Yun
xiii
Tatsuya Ishikawa Werner Bilfinger

Thomas Brandon William Fuentes
Tiago Dias William Lawson
Toms Laria Willy Alvarenga Lacerda
Vctor Rinaldi Xavier Vera Grunauer
Virginia Sosa Xinbao Yu
Vitor Pereira Faro Xiong Zhang
Waldemar Hachich Yazen Khasawneh
Walter Paniagua Yolanda Alberto Hernandez
This page intentionally left blank
xv
Contents
Preface v
Victor A. Rinaldi, Marcelo E. Zeballos and Juan Jose Clari
Conference Organization vii
Bishop Lecture and Plenary Lecture
Advanced Testing and Modelling of Granular Materials with and Without Viscous
Glue: Research and Practical Implication The Third Bishop Lecture 3
Herv Di Benedetto
Compaction Characteristics and Physical Properties of Compacted Soil Controlled
by the Degree of Saturation 40
Fumio Tatsuoka
Keynote Lectures
Rockfill Mechanics. Experimental Observations and DEM Modelling 79

E.E. Alonso and M. Tapias
Limitations of a Critical State Framework Applied to the Behaviour of Natural
and Transitional Soils 115
Matthew Coop
Shear Strength and Stiffness Anisotropy of Geologically Aged Stiff Clays 156
Richard Jardine, Amandine Brosse, Matthew Coop and
Ramtin Hosseini Kamal
Measurement and Application of Shear Wave Velocity to Various Geotechnical
Problems 192
Dong-Soo Kim, Heon-Joon Park and Hyung-Ik Cho
Hydro-Mechanical Behaviour of Shales 222
Lyesse Laloui, Alessio Ferrari and Valentina Favero
Improved Laboratory Techniques for Advanced Geotechnical Characterization
Towards Matching in Situ Properties 231
Antnio Viana da Fonseca, Cristiana Ferreira, Marisa Soares and
Assaf Klar
Reviewed Papers
Behavior of Compacted Unsaturated Soil in Isotropic Compression, Cyclic

and Monotonic Shear Loading Sequences in Undrained Condition 267
Ali Murtaza Rasool, Jiro Kuwano and Shinya Tachibana
xvi
Influence of Initial Stress/Strain State on the Coefficient of Earth Pressure

at Rest 275
Ali Akbar Eliadorani and Yoginder Vaid
Small Strain Modulus of Bio-Cemented Sand 283
Youventharan Duraisamy and David Airey
Disk Transducer for Stiffness Measurement on Granular Materials 291
Laxmi P. Suwal and Reiko Kuwano
Experimental Evaluation of Liquefaction Resistance of Unsaturated Sandy
Soils 299
Hailong Wang, Junichi Koseki, Takeshi Sato and Jaylord Tan Tian
Development of Stacked-Ring Shear Apparatus for Multiple Liquefactions
Tests 307
Seto Wahyudi, Sato Takeshi, Junichi Koseki and Yudai Aoyagi
Effects of Inherent Anisotropy on Deformation and Strength Characteristics
of a Reconstituted Sand 315
Hirofumi Toyota and Susumu Takada
Element Tests on Lumpy Inhomogeneous Soil and Their Interpretation 323
X.S. Shi and I. Herle
Dynamic Shear Modulus of Kaolin-Silt Clay Using a Novel Technique 331
Saeed Ahmad, Ahmed M. Alnuaim and M. Hesham El Naggar
Study of the Mechanical Behavior of Unsaturated Argillaceous Rocks 342
Jairo Espitia and Bernardo Caicedo
Effect of Cement Type on the Mechanical Behavior of Fiber Reinforced Sands 350
Navid Nikzad and Amir Hamidi
Relationship Between Undrained Shear Strength and Shear Wave Velocity
for Clays 358
Shehab S. Agaiby and Paul W. Mayne
Development of Large Size Disk Transducer to Evaluate Elastic Properties
of Coarse Granular Materials 366
Abilash Pokhrel, Reiko Kuwano and Laxmi P. Suwal
Assessment of Shear Modulus by Different Seismic Wave-Based Techniques 374
Jaime Santos, Joo Santos, Cristiana Ferreira, Cludio Pereira and
Antnio Gomes Correia
In Situ and Laboratory Mechanical Characterization Using High-Resolution
Fiber Optic Distributed Sensing 382
Assaf Klar, Shun Uchida and Eyal Levenberg
Experimental Investigation of Wave Propagation in Three Dimensions in
Unbounded Particulate Assemblies 390
Simon Hamlin, Erdin Ibraim, Martin Lings, David Muir Wood,
Ignazio Cavaretta and Jean Francois Camenen
xvii
Frequency Domain Method in Bender Element Testing Experimental

Observations 398
Javier Camacho-Tauta, Hassan Ali, Giovanni Cascante and
Antnio Viana da Fonseca
An Alternative Shear Strength Test for Saturated Fine-Grained Soils:
Preliminary Results 407
Kamil Kayabali, Nihan Pinar Yilmaz and Murat Mollamahmutoglu
Influence of Grading and Mineralogy on the Behaviour of Saprolites 415
Irene Rocchi, M. Cristina Todisco and Matthew R. Coop
Towards the Measurement of Fabric in Granular Materials With X-Ray
Tomography 423
Max Wiebicke, Edward And, Gioacchino Viggiani and Ivo Herle
Observing Breakage in Sand Under Triaxial and Oedometric Loading in 3D 431
Zeynep Karatza, Edward And, Stefanos-Aldo Papanicolopulos,
Jin Y. Ooi and Gioacchino Viggiani
Early Age Cemented Paste Backfill Stiffness Development 441
Lucas Festugato, Nilo Cesar Consoli and Andy Fourie
Influence of Volumetric and Shear Strains on the Destructuration of Saprolitic
Soils 447
Irene Rocchi and Matthew R. Coop
Description and Calibration of Triaxial Tests with Internal Measurement of
Displacement on Artificially Cemented Lateritic Soil 454
Roberto Q. Coutinho, Olavo F. Santos Jr. and Ricardo N.F. Severo
Studying Collapse Behaviour of Sandy Silt Under Generalised Stress Conditions 462
Octavio E. Crdenas, Rodrigo C. Weber, Enrique Romero,
Antonio Lloret and Josep Suriol
Travel Time Measurements of Shear Waves in Bo Bo Sand Using Bender
Elements 470
Juan Ayala, Felipe Villalobos and Giovanny Alvarado
3D Experimental Characterization of Particle Rotation and Local Dilatancy
in Angular Sand 478
Khalid Alshibli, Andrew M. Druckrey, Maha Jarrar and
Riyadh I. Al-Raoush
1D Creep and Delayed Rebound During Unloading and Reloading of Clay
and Its Model Simulation 486
Shohei Kawabe and Fumio Tatsuoka
Creep and Stress Relaxation Envelopes of Granular Materials in Direct Shear 494
Antoine Duttine, Alice Salotti, Fumio Tatsuoka and Alan Ezaoui
Creep and Stress Relaxation of Granular Materials Simulated by Non-Linear
Three-Component Model 503
Fumio Tatsuoka, Antoine Duttine, Alice Salotti and Alan Ezaoui
xviii
Environmentally Friendly Transparent Soil 511

Jun Kang Chow, Ghee Leng Ooi and Yu-Hsing Wang
Local Deformation Measurement Using Image Analysis Technique in Hollow
Cylindrical Torsional Shear Test 519
Usama Juniansyah Fauzi, Junichi Koseki, Yukika Miyashita,
Ryoichiro Hoshino and Hasbullah Nawir
Strain-Dependent Damping Ratio of Compacted Gravel 527
Tianfei Liao, Michael McHood, Nasser Massoudi, Kenneth H. Stokoe
and MinJae Jung
Experimental Assessment of the Influence of Load-Induced Deformation on
Interparticle Contacts 535
Masahide Otsubo, Way Way Sim and Catherine OSullivan
Comparison Between Conventional and Large Scale Triaxial Compression
Tests on Peat 543
Cor Zwanenburg and M.A. Van
Three-Dimensional Quantification of the Morphology and Intragranular Void
Ratio of a Shelly Carbonate Sand 551
Joana Fonseca, Constantino Carlos Reyes-Aldasoro and Laure Wils
Evaluation of Stationarity and Selection of Appropriate Transformation for
Geostatistical Modeling of Geotechnical Projects 559
Tejo V. Bheemasetti, Anand J. Puppala, Aravind Pedarla,
Raju Acharya and Santiago R. Caballero
A Study of the Micro-Mechanics of Sand Particle Breakage Using
a High-Speed Microscope Camera 567
Wanying Wang and Matthew Coop
Study on Viscous Property of Sedimentary Soft Rock in Drained Triaxial and
Unconfined Compression Tests 575
Yukika Miyashita, Junichi Koseki, Kenji Namikawa and Masashi Matsumoto
Assessment of Parameters of Compressibility and Strength of Sedimentary
Clays Found in the Northeast Region of Venezuela 583
Mariajos Guevara C.
The Effect of Grain Size Distribution on the Shear Strength- Dilation Relation
of Granular Material 591
Samaneh Amirpour Harehdasht, Mourad Karray, Mahmoud N. Hussien,
Varvara Roubtsova and Mohamed Chekired
Soil-Pile Interaction During Pile Installation. Observations from X-Ray
Tomography and 3D-DIC 600
Matias Silva and Gal Combe
Investigation of Soil Nonlinearity at Very Small Strains Using Ground Buried
Fiber Optic Sensors 608
Balz Friedli and Alexander M. Puzrin
xix
Behavioural Features of Fissured Clays: Experimental Evidence and Modelling 615

Federica Cotecchia and Claudia Vitone
Selection of Design Friction Angle Based on Soil and Project Characteristics 623
Emirhan Sancak and Ozer Cinicioglu
Stress Strain Behavior of a Desaturated Loessian Lightly Cemented Soil Under
Triaxial Compression Test 631
Julio A. Capdevila and Vctor A. Rinaldi
Analysis of the Hydro-Mechanical Behaviour of Volcanic Ash Slopes
Submitted to Rainfall 639
Alessio Ferrari, Lyesse Laloui, Chao Li and John Eichenberger
Granular Temperature Measurements of Uniform Granular Flows 647
Devis Gollin, Elisabeth T. Bowman and Paul Shepley
Micromechanical Insight into the Undrained Instability of Granular Materials 655
Daniel Barreto, Luis Felipe Prada-Sarmiento and
Alfonso Mariano Ramos-Can
Sampling and Sample Preparation on Structure Effects 663
Carlos Rodrigues and Lus Lemos
Validation of Contact Model Between Two Cylindrical Rods and DEM
Simulations of Biaxial Tests on a Rod Packing 671
Zhaofeng Li, Yu-Hsing Wang and Xia Li
Laboratory Test and Numerical Study of Cyclic Behavior of Unsaturated
Liquefiable Sandy Soil 679
Takaki Matsumaru, Ryosuke Uzuoka, Mineo Kiguchi and
Tomoyoshi Nishimura
Effects of Entrapped Air Bubbles on the Mechanical Behaviour of
Quasi-Saturated Soils 687
B.T. Lai, A. Fabbri, H. Wong and D. Branque
Discrete Element Investigation of Rate Effects on the Asymptotic Behaviour
of Granular Materials 695
David Man and Jan Jerman
Study of the Effect of Drainage Conditions on Cone Penetration with
the Material Point Method 703
Francesca Ceccato, Lars Beuth and Paolo Simonini
Deformation of Structured Soil with Cementation 711
Suksun Horpibulsuk and Martin D. Liu
Aging of Sand Grain Contacts at Microscale: Numerical Study 719
Zhijie Wang and Radoslaw Michalowski
A Non-Linear Poroelastic Approach to Rate Effects on Piezocone Tests in
Tailings Materials 727
Gracieli Dienstmann, Samir Maghous and Fernando Schnaid
xx
Simulation of Traffic Loading on an Embankment by the Finite Element

Method with Different Soil Models 737
Javier Camacho-Tauta, Fausto Molina and Oscar Reyes-Ortiz
Characterization of Consolidation Stress-Strain-Time Histories on Pre-Failure
Behavior of Natural Clayey Geomaterials 745
John N. Mukabi
Modeling of the Small Strain Shear Modulus on a Fiber Reinforced Sand 753
Jun Jos Clari and Paula Vettorelo
Calibration of Rotational Hardening Model for Clays Based on Dilatancy 761
Georgios Belokas and Yannis F. Dafalias
The Theory of Granular Packings and the Strength of Coarse Soils 768
Calixtro Yanqui
Understanding the Disintegration of Sensitive Clays Using Remolding Energy 776
Vikas Thakur and Samson Abate Degago
Integration Scheme for a Thermo-Elasto-Plastic Model with Unconventional
Yield Surfaces 784
Annan Zhou and Yue Zhang
Improvement of the Dynamic Behavior of Soil Structures Underlain
by Liquefiable Soil Using the Geosynthetics-Encased Columns 792
Hassan Alkayyal and Talal Awwad
Observation on Trace of Sand Boiling Erupted from Liquefied Sandy Ground
and Its Simulation by Laboratory Model Test 800
Yuki Horiuchi, Ngo Ngoc Anh, Reiko Kuwano and Jiro Kuwano
Visualization of Shear Bands in Cohesionless Soils 808
Behzad Soltanbeigi, Adlen Altunbas and Ozer Cinicioglu
A Microstructure-Based Finite Element Analysis of the Response of Sand 816
Sadegh Nadimi, Joana Fonseca and Neil Taylor
Stiffness Anisotropy Characteristics of Natural Fined-Grained Seabed Sediments 824
Delvin Abdiel Martnez and Satoshi Nishimura
Strain Localization Characteristics of Liquefied Sands in Undrained Cyclic
Torsional Shear Tests 832
Gabriele Chiaro, Takashi Kiyota and Junichi Koseki
Evaluation of Heterogeneities in Reconstituted Samples Using X-Ray
Tomography 840
Luis Carlos Leguizamn Barreto, Leonardo Flrez Valencia,
Alfonso Ramos Can, Luis Felipe Prada Sarmiento and
Daniel Oswaldo Tique Alviar
Inherent Anisotropy of an Undisturbed and Compacted Loess Soil 850
Pedro A. Covassi and Victor A. Rinaldi
Influence of Shearing Rate on Residual Strength of Clays 858
Kenny Kataoka Sorensen, Simon Steen Gadegaard and Jonas Kragh Nielsen
xxi
Advanced Laboratory Investigation of Axial Cyclic Loading in Silica Sands 866

Cristina de H.C. Tsuha, Richard J. Jardine, Zhongxuan Yang, Matias Silva
and Siya Rimoy
Influence of Volcanic Fly Ash on the Stress-Strain-Strength Behaviour of
Loessial Soils 874
Agustin Adami and Victor A. Rinaldi
Consolidation of Shales and Porosity-Permeability Relationship 882
Valentina Favero, Alessio Ferrari and Lyesse Laloui
Soil Behavior Under Unsaturated and Long Term Vertical Cyclic Loading 890
Binod Kafle and Frank Wuttke
Cyclic Behavior of an Over-Consolidated Remoulded Clay 898
Christophe Dano, Jian Han and Pierre-Yves Hicher
Evaluation of Consolidation Behavior Under Horizontal Drainage Condition
Using Digital Image Analysis 906
Joon-Young Kim and Choong-Ki Chung
Creep in Clay During the First Few Years After Construction 915
Gustav Grimstad, Magne Mehli and Samson Abate Degago
Comparison of Two Sample Quality Assessment Methods Applied to
Oedometer Test Results 923
Helene Alexandra Amundsen, Vikas Thakur and Arnfinn Emdal
Model for Predicting and Controlling Creep Settlements with Surcharge
Loading 931
Yixing Yuan, Andrew J. Whittle and David F.T. Nash
Pore Pressure Generation in Clays Soils Normally Consolidated Under
Undrained Cyclic Load Process 939
Julin Chaves-Agudelo and Flix Hernndez-Rodrguez
A Rheological Model for Peat That Accounts for Creep 947
Djamalddine Boumezerane and Gustav Grimstad
Effects of Initial Conditions on Strength of Acrylamide Grouted Sands 955
L. Sebastian Bryson and Ryan Ortiz
Stress-Strain Behavior of a Cement-Based Stabilized Soft Soil 963
Antnio A.S. Correia, Paulo J. Venda Oliveira and Lus J.L. Lemos
Effect of Treatment on the Microstructural Characteristics of Bio-Improved
Sand 970
Dimitrios Terzis and Lyesse Laloui
Deformation of Microbial Induced Calcite Bonded Sands: A Micro-Scale
Investigation 978
Brina Montoya and Kai Feng
Transitional Behaviour in Asphalt Derived Aggregate Specimens 986
Katherine Ailiang Kwa and David Airey
xxii
Study of the Hot Mix Asphalt Used in Middle West Region of Brazil 994
Carolina Q. Arantes, Lilian R. Rezende and Liedi L.B. Bernucci
Hydraulic and Mechanical Behaviour of Cement-Bentonite Mixtures
Containing HYPER Clay: Impact of Sulfate Attack 1002
Gemmina Di Emidio, R. Daniel Verastegui Flores, Claudia Scipioni,
Evelina Fratalocchi and Adam Bezuijen
Monotonic and Cyclic Behaviour of Weakly Cemented Sands 1010
Antigoni D. Vranna and Theodora Tika
Deformation Characteristics of Grain-Displacing GH-Bearing Sediments 1018
Joo Yong Lee, Hyoung Lee, Taewoong Ahn and SeJoon Kim
A Thermo-Hydro-Mechanical Constitutive Model for Saturated Frozen Soils 1024
S.A. Ghoreishian Amiri, M. Kadivar and G. Grimstad
Numerical Analysis of Geosynthetic Reinforced Soil Models for Shallow
Foundations 1032
Gonzalo Aiassa Martinez, Danny Useche Infante, Pedro Arrua and
Marcelo Eberhardt
Behavior of a Sand-Polyacrilate Mixture for Seismic Isolation 1040
A. Flora, E. Bilotta, S. Lirer, D. Lombardi and V. Nappa
Undrained Failure of Fine-Grained Soils with Particular Reference to Shallow
Slope Instability 1049
Anand Panta and Satoshi Nishimura
Influence of Smectite Content on the One-Dimensional Deformation Behavior
of Reconstituted High Plasticity Clays 1057
Michael Rosenlund Lodahl and Kenny Kataoka Sorensen
Experimental Study on the Cementation Level in an Artificial Rock with
Crushable Grains 1065
Erika Tudisco, Francesca Casini and Giulia M.B. Viggiani
Laboratory Investigation on the Mechanics of Soft-Rigid Soil Mixtures 1073
Salman Rouhanifar and Erdin Ibraim
The Bearing Capacity of Debris Flows: Laboratory and Centrifuge Analyses 1081
Luis E. Vallejo, Hankyu Yoo and Bernardo Caicedo
The Effect of Fine Content and History of Hydraulic Loading on
the Characterization of Suffusion Sensibility of Cohesionless Soil 1089
Abdul Rochim, Didier Marot, Luc Sibille and Fateh Bendahmane
Testing Tube Specimens from Soft Clay Deposits Containing Variable
Amounts of Shells 1097
Guan T. Lim, Nathalie Boukpeti, J. Antonio H. Carraro and
Jubert A. Pineda
Simple vs Advanced Interface Model: A Comparison Using a Deterministic
Quality Approach 1105
H. Stutz, G. Mortara and F. Wuttke
xxiii
Capillary Rise in Corners of Rough Pores: Laboratory and Theoretical

Analyses 1113
Shirley Tang, Luis E. Vallejo and Bernardo Caicedo
Geotechnical Characterization of a Pyroclastic Sand and Tuff 1121
Pedro A. Covassi, Marcelo E. Zeballos and Sofa I. Gorosito
Numerical Assessment of Shear-Induced Particle Interactions Under Different
Loading Conditions by Means of Axial and Torsional Interface Shear Tests 1129
Alejandro Martinez and J. David Frost
Stress-Deformation Modeling of Geosynthetic-Reinforced Pavement
Structures at Different Water Contents 1137
Rafael Baltodano Goulding and Paulina Leiva Padilla
Constitutive Model Input Parameters for Numerical Analyses of Geotechnical
Problems: An In-Situ Testing Case Study 1145
Crystal Cox and Paul Mayne
Geotechnical Characterization of Calcareous Sands in the Northeastern Coast
of Venezuela 1153
Nelly Vieira Faria and Diana De Ponte Fernandes
Laboratory Determination of Parameters for Transversely Anisotropic Model
of Stiff Clay 1161
Monika ernkov, Martin Krupika, Tom Mohyla, Jan Boh,
Josef Rott and David Man
Interpretations of Measurements for Both Rigid and Flexible Retaining
System in Cohesive Cemented Sandy Soils in Lebanon 1169
Youssam Kazan
Primary Consolidation Settlement of Embankments Constructed on Young
Bay Mud of San Francisco Bay 1177
Abhijit Bathe, W. Martin McCabe, L. Sebastian Bryson,
Frank Lobedan and Samuel Won
Seismic Wave Velocities of Soft Clays Measured with Piezoelectric Crystals
in an Oedometer 1186
Miguel Diaz Pardave and Efran Ovando-Shelley
Determination of the Maximum Density of Carbonate Silty Sands 1194
Naemeh Naghavi Alhosseini, M. Hesham El Naggar and
Abouzar Sadrekarimi
Subject Index 1201

Author Index 1207
Bishop Lecture and Plenary Lecture
Deformation Characteristics of Geomaterials 3
V.A. Rinaldi et al. (Eds.)
IOS Press, 2015
2015 The authors and IOS Press. All rights reserved.
doi:10.3233/978-1-61499-601-9-3
Advanced Testing and modelling of

Granular materials with and without
viscous glue:
Research and practical implication
-The third Bishop Lecture-
-La troisime Confrence Bishop-

Essais et Modlisation avancs pour les
matriaux granulaires avec et sans colle
visqueuse : recherche et implication dans la
pratique
Herv Di Benedetto
University of Lyon, ENTPE, LTDS (UMR CNRS 5513),Lyon, France
Abstract. This lecture presents an overview of some of the results obtained by the
authors team on the mechanical behaviour of unbound granular materials (UGM)
and bituminous mixtures (BM). Experimental advanced devices and obtained
results, rheological modelling and calculation of practical cases are proposed.
Linear and non-linear domains of behaviour are considered including viscous and
therrmal effects. A unified framework allowing describing complex behaviour of
geomaterials is proposed.
Keywords. unbound granular materials, geomaterials, bituminous mixtures,

laboratory tests, rheological modelling, linearity, non-linearity, viscous behaviour
1. Introduction
This paper gives a rapid overview of some of the results obtained by the authors team
on the mechanical behaviour of unbound granular materials (UGM), mainly sands, and
bituminous mixtures (BM) used in road and railway construction. Bituminous mixtures
are composed of a granular skeleton (80% to 85% in volume) and bituminous binder
(12% to 20% in volume). Void content is generally small (0% to 10% in volume). The
binder is a viscous glue that brings to the mixtures very significant viscous properties.
4 H. Di Benedetto / Advanced Testing and Modelling of Granular Materials
Most of the presented qualitative observations and modelling can be extended to other
geomaterials (soils, rocks, and hydraulic concretes), considering very different
amplitudes for the observed phenomena, thus needing very different values for
constants used for rheological simulations. A more extended bibliography can be
obtained in the papers from the author given as references. More experimental results
and practical examples are given in the Bishop Lecture presentation, which is
associated with this paper.
It is a real challenge to propose a unified framework allowing describing complex
behaviour of geomaterials. Considering experimental results from literature and the
large amount of data obtained during research works performed at the ENTPE
laboratory during the last decades, a general framework based on the 3 component
model (Figure 1) could be proposed and extended. Most of the ENTPE results were
obtained during PhD theses. I take the opportunity of this paper to thank all my former
and present PhD students. The following references consist in PhD dissertations, where
detailed developments and results on the topic considered in this paper are presented
[[1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18],
[19], [20], [21]]. This unified framework and asymptotic expression within the small
strain loading domain, are Presented in chapter 2.
Many experimental results ([22], [23], [24], [25], [26], [27], [28], among others)
show that behaviour of materials can be considered as linear in the small strain domain.
The strain amplitude limiting this linear domain is about some 10-5 m/m for unbound
granular materials and also for bituminous mixtures if the number of applied cycles is
not too high. Some of the recent outputs from ENTPE team for both types of
materials within this domain are presented in chapter 3. It includes experimental data
obtained thanks to developed advanced accurate testing devices and modelling.
Chapter 4 deals with the behaviour of both materials (UGM and BM) outside the
linear domain.
2. General framework for rheological modelling: 3 component model and

extension
It is assumed that the strain increment d (or strain rate) is the sum of a non-viscous
dnv (or instantaneous) part and a viscous dv (or deferred) part, Eq. (1). Based on the
experimental observations, the framework given by the general three-component model
can be successfully applied for UGM such as sand and sand clay mixture ([24], [29],
[30], [31], [32]) or for bituminous materials ([33], [34], [35], [36]). Figure 1 gives an
analogical presentation in the 1 dimensional case of general considered model and
asymptotic obtained expression in the linear case (ie. in the small strain domain) for
both UGM and BM.
ddnvdv (1)
The n bodies EPi of Figure 1, loaded by stress fi , have a non-viscous (sometimes
called elastoplastic) behaviour. A great number of constitutive laws have been
proposed to describe non viscous (or elastoplastic) behaviour, such as: elasticity,
plasticity, elastoplasticity, hypoplasticity, interpolation type, among many others (cf.
H. Di Benedetto / Advanced Testing and Modelling of Granular Materials 5
Figure 1). It can be shown that the general form of the strain increment given by the
body of EP type, is given Eq. (2).
d M {h, dir (d )} d (2)
Where these quantities are tensors and dir (d ) d / d is the direction of the
stress increment (d) whose norm is 1. The letter h represents the whole history
parameters, also called memory, hardening or state parameters. M is the constitutive (or
compliance) tensor, which depends on h and dir(d).
The introduction of dir(d) expresses the irreversibility and parameters h, which
may be scalars, vectors or tensors, describe stress history dependence.
Bodies Vi of Figure 1, loaded by stress vi , create the viscous property
dependency of the material. It represents a specific time-dependent behaviour, which is
expressed by Eq. 3.
F (h , ) (3)
Where F is the viscous tensor, which depends on h and (objective strain rate)
applied to the V body.
EP1 1 EPn n
f f

EP0

V1 v1 Vn vn
UnboundGranular v
nv
Bituminousmixtures(BM)
Materials(UGM)
DBNmodel
EP1 EP1 EPn 2S2P1Dmodel
EP0 EP0

V 1 n Parabolic
creep
elements h
Asymptoticbehaviour inthesmallstraindomain
E1 E1 En k
GKVmodel
E0E00
E0 E0 E00
V 1 n
Figure 1. Mono dimensional analogical representation of considered models for UGM and BM and
asymptotic expression in the small strain domain. Non viscous EP body type can take into account non-
linearity, irreversibility, stress path history (examples are Elasticity, plasticity, elastoplasticity, hypoelasticity,
hypoplasticity, interpolation type). Viscous V body type can take into account viscous effect, non-linearity,
irreversibility, stress path history) (examples are Newtonian linear, Newtonian non-linear, parabolic creep,
viscous evanescent)
As schematized in Figure 1, EPi bodies have an asymptotic behaviour, which is
linear in small strain amplitude loading. In this domain of loading they can be
represented by elastic bodies, i.e. spring of stiffness Ei in the 1-dimensional case or 3-

dimensional hypoelastic bodies respecting equation 4.
dM(h) d
A specific procedure was developed to quantify the asymptotic hypoelastic tensor

Mi(h) for each of the EPi bodies. Similar approaches can be used for UGM and BM.
The process needs good experimental data that can only be obtained with accurate
testing devices.
3. Behaviour in the small strain domain
3.1. Unbound Granular Materials (UGM)
When small strain amplitude loading is considered, the behaviour is quasi linear. It is
either hypoelastic or viscous, depending on the ability of the material to exhibit or not
viscous behaviour, Figure 1. As only small strain domain is treated in this section, only
the asymptotic behaviour of bodies of the 3 component model is introduced.
3.1.1. Experimental devices: T4C StaDy & Triaxial StaDy

To investigate small strain behaviour of UGM, specifically accurate devices must
be used. A hollow cylinder experimental prototype has been designed and developed at
ENTPE since 1991. It is a precision device able to apply Torsion, Compression,
Confinement on Hollow Cylindrical sample for Static and Dynamic loading (T4C
StaDy). A precision triaxial prototype (Triaxial StaDy) has also been developed
following the same requirements in terms of local strain measurements and dynamic
loadings.
The T4C StaDy (Figure 2) sample has a 12 cm height, an outer diameter of 20
cm and an inner diameter of 16 cm. Two Neoprene membranes (0.5 mm thickness)
constitute the lateral side while two rigid platens close the sample at the top and at the
bottom. The Triaxial StaDy (Figure 3) specimen has a 14 cm height and a 7 cm
diameter. Investigation of soil response from very small to large strain domains is
possible with these devices thanks to local strain measurement systems designed in the
same way. Vertical (and angular) displacements are measured on two levels thanks to
two light rings (made of duralumin) hanged by 3 points on the outer membrane and
carrying targets (made of aluminum) aimed by non-contact transducers. Radial (outer
and/or inner) displacements are also measured by non-contact transducers pointing
towards sheets of aluminum paper placed on the inner side of the membrane(s). All the
transducers are fixed on (electronically or manually) movable supports. The
displacements of the transducers, monitored from outside the cell, during a test allow
covering a large range of deformation, while keeping a good accuracy. The number of
non-contact transducers used for T4C StaDy and Triaxial StaDy devices is 14 and
6, respectively.
Scheme of "T4CStady" apparatus System of strains measurement

Piezoelectric sensors
Piston axis Suspension of ring (x3)
/
Confining cell Rings at two levels
/
// "high" and "low"

//
Internal barrels Aluminium paper target

//
(radial strain)
//
Top platen
Non-contact sensors (x14)
//
Piezoelectric sensors
//
Aluminium target
screwed to ring
//
Sample
//
Targets for vertical

//
Bottom platen
strain measurement (x4)
//
6i 1
Support of shear strain
//
Ring hanging by 3 strips

measurement sensor
//
6i
//
Support of radial strain Pair (x3) of sensors

3
//
measurement sensor ("inner" and "outer")

6i
//
measuring radial strain

//
Targets for shear

strain measurement (x4)
Figure 2. Schematic view of the T4C StaDy hollow cylinder apparatus and its system of strains
measurement
Figure 3. Triaxial StaDy apparatus: sample and axial strains measurement (a), radial strain measurement
using micro motors (b)
Moreover, both devices are equipped with piezoelectric sensors (compression

elements and bender elements) located in each platen (Figure 4). Concerning the T4C
StaDy apparatus (Figure 4a), two pairs of bender elements give characteristics of
waves propagating in the axial direction. The first pair concerns waves polarized in
radial (Sr) direction and the other in orthoradial (S) direction. The two pairs of
compression elements are identical. They are noted Pr and P and are close to the
respective sensors S. Regarding the Triaxial StaDy (Figure 4b), one single pair of
compression element and one single pair of bender elements are used in the platens.
Three couples of piezoelectric transducers were added to measure wave propagation in
horizontal direction (Figure 4c). These bender type transducers are all identical and
placed in different ways along the membrane in order to generate shear or compressive
radial waves. These arrangements are described in Figure 4c.
(c)
Figure 4. Location of piezoelectric transducers in the hollow cylinder device T4C StaDy (a), and platen of
the triaxial device Triaxial StaDy (1. compression element, 2. porous stone, 3. bender element) (b), and
along the lateral surface of the sample (3 couples of bender elements) (c)
Back analysis of wave travel times provides dynamic elastic parameters.
For more details, the T4C StaDy device has been more extensively presented for
example in [6], [9], [31], [37], the triaxial StaDy is described in [8], [38].
3.1.2. Some noticeable results obtained with the 2 devices

3.1.2.1. Viscous effects
Viscous effects or loading-rate effects on UGM are presented in different publications
([29], [31], [39], [40], [41], [42], [43], [44], among others) and are not developed in
this paper. A long term cooperation with Professor Tatsuoka and his team was also
devoted to this topic. The main conclusion is that four viscosity types could be
identified and modeled: Isotach, Combined, Viscous evanescent (or TESRA
Temporary effect of strain rate acceleration) and Positive & Negative (P & N) as
illustrated in Figure 5. Each viscosity type is dominant following the type of materials
(i.e. cohesion, grading curve, particles shape and mineralogy, compaction) and shear
level. A new type of model was introduced to express V behaviour (Figure 1).
Amplitudes of viscous phenomena remain low. However, these phenomena may not be
ignored and may exhibit non negligible effects at an engineering scale. Modelling of
Ageing Effects and viscous properties are treated in [45].
Figure 5. Illustration of Positive and Negative (P & N) viscosity (from [43])

3.1.2.2. Hypoelastic tensor

For UGM, during small quasi static cycles or wave propagation tests, experiments
reveal that no viscous deformation appears. Then only body EP0 (Figure 1) is activated
for the considered granular materials (the body V locks the viscous deformation in the
considered domain of loading). As the experimental behaviour is (quasi) elastic, a
hypoelastic formulation, the DBGS (Di Benedetto, Geoffroy, Sauzat) model, has been
developed. This model is introduced in the next section. More details on DBGS model
and its extension DBGSP, which takes into account the inherent and strain-induced
anisotropy, can be found in ([29], [37], [46], [47]). During creep or relaxation periods
or stepwise changes, bodies EP1 and V play a key role and have to be considered. The
hypoelastic body E0 of Figure 1 respecting Eq. (4) was investigated.
Symmetry of the hypoelastic tensor M0

The term Mz obtained by performing small axial cycles (only zz varies) with the
T4CStaDy device, and the term Mz resulting from pure torsion small cycles (only z
varies), appear as very close, and could be considered as equal for the different tested
geomaterials. An example of results is given in Figure 6 for Toyoura sand. These quite
unique results lead to the conclusion that tensor M0 can be considered as symmetric.
4
R=0,7702
Toyoura
1,09
Mrreq Mreq Mrzeq Mreq
eq
Mr Meq Meqz Meq
Mzreq Mzeq Mzzeq Mzeq
eq
M r Meq Meqz Meq
Mz (GPa )
-1
0
eq
Mrreq Mreq Mrzeq Mreq

eq
Mr Meq Meqz Meq
Mzreq Mzeq Mzzeq Mzeq
eq
M r Meq Meqz Meq
-4
-4 0 4
eq -1
Mz (GPa )
Figure 6. Experimental values for terms Mz obtained from small axial cycles (only zz varies) and term Mz
resulting from pure torsion small cycles (only z varies) on Toyoura sand (from [9])
DBGS and DBGSP formulation

Di Benedetto et al. [29] suggest the following anisotropic (orthotropic) and
symmetrical expression of hypoelastic tensor M (eq. (4) for body E0 of Figure 1) :
1 S .p p . S
t t
M DBGS
f (e) 2 (5)
With f(e) being a function of void ratio (e), approximating void ratio effects. t
denotes the function transpose. {Sv ; p} are the tensors originally defined by Hardin &
Blandford [48]:
1 0 0 0 0 0
1 0 0 0 0
0
0 1 0 0 0
S 0
0 0 0 1 0 0 0
0 0 0 0 1 0 0

0 0 0 0 0 1 0
(6)
1
m 0 0 0 0 0
1
1
0 m2
0 0 0 0

1
0 0
3m
0 0 0

p 1
0 0 0 m m
0 0
2 3 2
2

1
0 0 0 0 0

m m
1 2 3 2

0 0 0 0 0
1
m m
1 2 2 2
(7)
Where {m ; 0} are two constants and stand for respectively the power coefficient and
the isotropic Poissons ratio value at an isotropic stress state. {1, 2 and 3} are the
principal stress values.
Note that the expression of p tensor (Eq. (7)) is valid only in the stress principal
axes. For instance, as far as pure torsional shear tests (from an initial isotropic or
anisotropic stress state) are concerned, (sudden or continuous) rotation of principle
axes from fixed sample axes are involved and rotation tensors are introduced. The
MDBGS expression becomes therefore more complicated ([3], [9]).
The DBGS model does not introduce eventual anisotropy due to strain and
considers that the behaviour is isotropic when stress tensor is isotropic. A modified
version of the model called DBGSP has been proposed ([8], [47]). It introduces
fabric and strain induced anisotropy.
Similar results between wave propagation and quasi static small cycle analyses
Many experimental campaigns on UGM with the two devices T4CStaDy and triaxial
StaDy, confirmed that back analysis with P and S waves and direct analysis of quasi
static cycles of small amplitudes give close results. Combining these two methods
allowed obtaining the 5 parameters (3 moduli and 2 Poissons ratios) of the transverse
isotropic hypoelastic tensor M0 during triaxal compression or triaxial extension tests
([38]). Figure 7 shows obtained values for Hostun sand prepared using pluviation
method.
700 dyn
Ez a)
dyn
600 Er
500

Induced anisotropy
: MPa
400
dyn

300
Simple Pluviation method

200
Deviatoric stress state r= 400kPa)
TE test TC test
TC_H400.82p / TE_H400.80p
100
200 400 600 800 1000 1200
z: kPa
0,4 dyn dyn
rz rr b)
dyn
Linear fit (
rr ) (slope = 0)
0,3
dyn
0,2 y = 0.191

TE test TC test
0,1
Simple Pluviation method
Deviatoric stress state r= 400kPa)
TC_H400.82p / TE_H400.80p
0,0
200 400 600 800 1000 1200 1400
z: kPa
Figure 7. Evolution of Youngs moduli Ez, Er and Poissons ratios rz and rr, during triaxal compression (in
fact, contraction) and extension tests for sand samples prepared using pluviation method (From [38])
Anisotropy of hypoelastic tensor

It is therefore important to consider anisotropic behaviour when analyzing wave
propagation tests as illustrated in Figure 8. In this figure comparison between dynamic
Youngs modulus obtained from back analysis of P and S waves Ezdyn and statically
determined Youngs modulus Ezstat from small cyclic loadings is presented considering
two hypotheses: isotropic behaviour and transverse isotropic behaviour. Data of Figure
8 clearly show isotropy is only a coarse approximation.
3.1.3. Example of site back calculation

The St Alban (France) power plant site was investigated. A series of cross-hole tests
were performed (Figure 9a). The DBGS model was used to back-calculate horizontal
and vertical Youngs moduli, shear modulus and Poissons ratios in the different
directions from the measurements of wave propagation velocities (P and S) ([40]).
Comparison between isotropic behaviour and anisotropy (transverse isotropy)
hypotheses could be obtained (Figure 9c). From Figure 9, it can be seen that 0*
(constant corresponding to the Poissons ratio at initial isotropic stress state of the
DBGS model) is around 0.42 for sands and gravels, 0.37 for sandy marls and 0.32 for
sandy marls with limestone. These values, although slightly higher than expected, are
more in accordance with soil skeleton properties than the ones deduced from the
isotropic hypothesis. Regarding Youngs moduli, the vertical drained modulus is found
between 15% (for sands and gravels) and 20% (grey sandy marls) higher than the
obtained isotropic drained modulus, reaching 40% for sandy marls with limestone.
The above results singularly enhance the differences brought on the values of
Poissons ratios or Youngs moduli between isotropic and transverse isotropic
assumptions. Thus, these results reveal the importance of considering anisotropic
behaviour for soils.
1200
"T4C StaDy" & "Triaxial Stady" apparatus 1
(MPa)
Edyn
1000 Hostun (29 tests)
Toyoura (7 tests)
dyn
M15 mixture (4 tests)

Dynamic Young's modulus E
800
0.75
600
400
Ezstat
200 z
z 10-5
0
0 200 400 600 800 1000 1200
stat
Static vertical Young's modulus Ez (MPa)
1200
"T4C StaDy" & "Triaxial Stady" apparatus
(MPa) (transv. iso. elasticity + DBGS)
1
1000
ANISO
800
600
400
Ezstat
200 Hostun (m=0.50) (29 tests) z

dyn
Toyoura (m=0.40) (7 tests)

Ez
M15 mixture (m=0.74) (4 tests) z 10-5
0
0 200 400 600 800 1000 1200
stat
Ez (MPa) (static small cyclic loading)
Figure 8. Comparison between dynamically determined Youngs modulus Ezdyn and statically determined
Youngs modulus Ezstat (small cyclic loadings) for 36 Triaxial Compression (TC) tests on dry sands and 4 TC
tests on sand-clay mixtures (from [46]): isotropy hypothesis (above) and transverse isotropy + DBGS model
(below)
Figure 9. Results on St Alban site: Schematic view of the cross-hole testing facility (a); Measured
compression and shear wave velocities (b); Youngs and shear moduli and Poissons ratio for overall soils : E,
G and (isotropic elasticity) and for soil skeleton moduli : Ghv*, Eh* and Ev* in the case of transverse
isotropic elasticity with the DBGS model ( 0* is the constant corresponding to the Poissons ratio at initial
isotropic stress state of the DBGS model) (c)
3.2. Bituminous Mixtures (BM)
Mechanical behaviour of bituminous mixtures used in road construction is very

complex. It is characterized by large viscous effects and great thermal sensitivity. It is
possible to qualify bituminous mixture behaviour by considering the applied cyclic
strain amplitude and number of cycles (Figure 10). When loading up to a few hundred
cycles with "small" strain amplitudes (<some 10-5m/m) behaviour can be considered, as
a first approximation, as linear viscoelastic ([49]). With loading of several hundreds of
thousands of cycles and small strain amplitudes, damage phenomena appear, and the
material fails due to fatigue. For strain levels of a few per cent, the observed behaviour
becomes highly non-linear. Non-linearities can induce macro cracks or permanent
deformations depending on temperature and stress path history.
When small strain amplitude loading and small number of cycles are considered,
the behaviour of BM is linear viscoelastic and parameters of bodies from the
asymptotic model presented in Figure 1 can be determined. It should be underlined that
viscous effects are of paramount importance for BM when comparing with other
geomaterials. Figure 11 from [30] illustrates this affirmation. It can be seen in this
figure that the modulus of BM varies by a ratio of about 1000 when considering very
small strain rates and/or high temperatures and very high strain rates and/or low
temperatures, ranging between values observed for NC clays up to values for cement
concrete.
Figure 10. Typical domains of behaviour for bituminous mixtures, Logarithm of strain amplitude versus
Logarithm of applied cycles (from [50])
3.2.1. Experimental devices

Two types of testing device (Figure 12) were developed at the ENTPE laboratory in
Lyon to investigate thermo-mechanical properties of BM, including mastics which
aggregates are only fine particles (size below around 0.1 mm). These devices cover
small and large strain domains. For both types, the specimen is placed within a thermal
chamber that allows regulating temperature within the range from -30C to 70C.
Tension-compression type devices (Figure 12 above) give access to Youngs
moduli and Poissons ratios. Annular Shear Rheometer (ASR) device (Figure 12
below) allows measuring sinusoidal shear stress and sinusoidal shear strain (distortion)
applied on a hollow cylinder of bitumen or mastic, at different temperatures and
frequencies. The shear modulus (G*) can be obtained.
Figure 11. Vertical modulus, at a strain amplitude close to 10-5 m/m, as a function of strain rate (or
equivalent strain rate for cyclic tests) for different geomaterials, the very important viscous sensitivity for
bituminous mix is clearly visible (from [30])
3.2.2. Some obtained new results

Within linear domain, many researches were performed in the one-dimensional case for
bituminous materials. They confirmed that the behaviour of bituminous materials can
be modelled using linear viscoelastic (LVE) theory. In addition, Time Temperature
Superposition Principle (TTSP) can be successfully applied.
Some new results were obtained from researches performed at ENTPE laboratory
mainly concerning the generalization in the tri-dim case of one dim results, including
time-temperature principle validation and Poissons ratio values determination ([34],
[35], [36], [51], [52], [53], [54], among others).
All the developments could only be made thanks to data obtained with the accurate
testing devices introduced previously.
Figure 12. Tension-compression test device used for BM (above), Annular Shear Rheometer developed for
bitumens and mastics mixture of fines and bitumen- (below)
Theory of viscoelasticity shows that linear viscoelasticity (LVE) properties can be
obtained when considering cyclic sinusoidal loadings. All data obtained (from quasi
static analysis) by Uni. of Lyon/ENTPE team consider cyclic loading at strain
amplitude smaller than 10-4 m/m, ranging from 0.003 Hz to 10 Hz and performed at
different constant temperatures within the range from -30C to 75C. An example of
signal obtained during T/C test at 1Hz and 15C is presented in Figure 13. LVE
rheological parameters are complex Youngs modulus (E*) and complex Poissons
ratio(s) ( *) expressed as functions of frequency (f) or pulsation (=2f) and
temperature (T). Eqs. 8 to 14 give notation and determined parameters when
measurements of radial strain are made in 2 directions (II and III). In case of isotropy
(or transvers isotropy) only one Poissons ratio is considered.
ax (t) = Aax sin(t)

(8)
ax (t) = Aax sin(t + E )

(9)
rad II (t) = - Arad II sin(t + II )

(10)
rad III (t) = - Arad III sin(t + III ) (11)
where E is the phase angle between axial strain and axial stress; II and III are
the phase angles between axial strain and radial strains in directions II and III,
respectively; Aax, Arad II, Arad III and Aax are the axial strain amplitude, the radial strain
amplitudes in directions II and III, and the axial stress amplitude, respectively.
From previous determined parameters (Eqs. 8 to 11), complex modulus (E*) and
complex Poissons ratios in directions II and III (*II-I and *III-I) are calculated as
follows:
* j * j
E = Aax e E = | E | e E
Aax
(12)
* Arad II j II * j II
II-I = - e = | II-I | e
Aax
(13)
* Arad III j III * j III

III-I = - e = | III-I | e
Aax
(14)
Where |E*| is the norm of complex modulus; |*II-I |, |*III-I | are the norms of
Poissons ratio in directions II and III, respectively, and j is the complex number
defined by j2 = -1.
6,E-05 axial strain 0,6

ax
4,E-05 radial strain 0,4
Stress (MPa)
ax II
2,E-05 0,2
Strain
0,E+00 radII radIII 0

-2,E-05 -0,2
-4,E-05 -0,4
-6,E-05 -0,6
0 0,2 0,4 0,6 0,8 1
Time (s)
Figure 13. Example of cyclic T/C test results at 1Hz, 15C for BM : axial strain, radial strains in two
perpendicular directions (II and III) and axial stress (from [55])
3.2.2.1. Complex Youngs modulus and complex Poissons ratio

Some results obtained during the PhD of Pham NH [21] are presented as an example of
results obtained for nine different types of BM containing or not Recycled Asphalt
Pavement (RAP). The BMs differ in manufacturing process, production temperature,
percentage of Reclaimed Asphalt Pavement (RAP), use of additives and air void
content. More details are given in [55], [56]. The new information coming from these
results is the evolution of complex Poissons ratio, which is an essential data for 3D
characterization.
Specimens were loaded at 6 frequencies (0.03Hz, 0.1Hz, 0.3Hz, 1Hz, 3Hz, 10Hz)
and 10 temperatures (from -25C to 45C, with 3 repetitions at 15C : 15C(1), -25C,
-15C, -5C, 5C, 15C(2), 25C, 35C, 45C, 15C(3)). The amplitude of axial strain
during cyclic loading was about 50 m/m.
Figure 14. Complex modulus in the Cole-Cole (above) and in the Black (below) spaces. Experimental data
and simulations with 2S2P1D model for 9 different types of bituminous mixtures (from [55])
Figure 14 presents the obtained complex Youngs modulus for the 9 different types
of BM (each test is duplicated) in the Black (norm of E* as a function of phase angle)
and Cole-Cole (imaginary part of E* as a function of real part) axes. When changing
frequency and temperature, data for each specimen fall on a unique curve, which
indicates that the Time Temperature Superposition Principle (TTSP) is valid for all
the materials, in the linear viscoelastic (LVE) domain. Curves simulated with 2S2P1D
model (for two Springs, two Parabolic elements, one Dashpot; cf. next section) are also
plotted in Figure 14.
Master curves for the norm of complex modulus (|E*|) and for its phase angle (E)
are plotted in Figure 15 as functions of equivalent frequency, which is equal to
frequency (f) multiplied by shift factors aT. Obtained values of aT are presented in
Figure 17 as a function of temperature. The chosen reference temperature is 15C for
master curves. As a consequence, aT is equal to 1 for 15C.
100000
10000
|E*| (MPa)
1000
Tref = 15C
100
10
-7 -5 -3 -1 1 3 5 7 9
10 10 10 10 10 10 10 10 10
f.aT (Hz)
60
50
40 Tref = 15C
E ()
30
20
10
0
-7 -5 -3 -1 1 3 5 7 9
10 10 10 10 10 10 10 10 10
f.aT (Hz)
Figure 15. Master curves of the complex modulus, at 15C reference temperature: norm (above) and phase
angle (below). Experimental data and simulations with 2S2P1D model same tests as Figure 14 (from [55])
Obtained results for Poissons ratio *av ( *av = ( *II-I+ * III-I)/2) are presented in
Figure 16. Master curves of norm and phase angle ( av ) of *av are plotted. The
same values of shift factors aT determined from complex modulus E* data are used and
the same reference temperature 15C is chosen. The data confirms that, i) Poissons
ratio is a complex number, and ii) TTSP can be generalized to the 3-dimensional
behaviour as shown previously by other studies from ENTPE team ([35], [54], [57],
among others). A similar evolution of *av with equivalent frequency could be
observed for all the materials. Some differences appear only for the asymptotic values
(at high or low equivalent frequency). The asymptotic value 0 at high frequency/low
temperature is always smaller than asymptotic value 00 at low frequency/high
temperature. The ratio between these two limit values is between 2 to 3, which is quite
noticeable. Poissons ratio is considered as constant for all structure design calculations.
Our results show it should not be the case for a good simulation of road structures.
Shift factors from data presented in Figures 14 to 16 are plotted in Figure 17. It
should be underlined that the same shift factor values are obtained for E* and *.
0.5
0.4
0.3
| *av|
0.2
Tref = 15C
0.1
0.0
-7 -5 -3 -1 1 3 5 7 9
10 10 10 10 10 10 10 10 10
f.aT (Hz)
Tref = 15C
2
0
av ()
-2
-4
-6
-8
-7 -5 -3 -1 1 3 5 7 9
10 10 10 10 10 10 10 10 10
f.aT (Hz)
Figure 16. Master curves of norm (above) and phase angle (below) of Poissons ratio *av (average of
Poissons ratios measured in directions II and III) plotted at 15C reference temperature: experimental data
and simulations with 2S2P1D mode same tests as Figure 14 (from [55])
3.2.2.2. 2S2P1D and Generalised Kelvin Voigt models

2S2P1D -two Springs, two Parabolic elements, one Dashpot- model
Extensive works on the modelling of complex modulus of bituminous material ([34],
[35], [58], [59], [60], [61]) allowed formulating a general 1D and 3D linear viscoelastic
model with a continuum spectrum, called 2S2P1D (two Springs, two Parabolic
elements, one Dashpot) (cf. Figure 1 for the 1 dim case). It was shown that the 2S2P1D
model is powerful to simulate 1D and 3D LVE behaviour (i.e. in the small strain
domain) of bituminous binders, mastics and mixes, over a very wide range of
frequencies and temperatures. Complex modulus and complex Poissons ratio
expressions of 2S2P1D model for isotropic case are given in Eqs. 15 and 16.
8
10
6
10
4
10
Shift factor aT
2
10
0
10
Tref = 15C
-2
10
-4
10
-30 -20 -10 0 10 20 30 40 50
Temperature (C)
Figure 17. Shift factors aT for data presented in Figures 15 and 16 and WLF law fitting (Eq. 17). Identical
values for E* and *
Anisotropic (orthotropic) case is treated in [13], [55], [56] and [62]. It is shown that
BM are slightly orthotropic (less than 15% relative difference in modulus). Modelling
of orthotropic properties using 2S2P1D is made simply from asymptotic values of
Youngs moduli and Poissons ratios obtained at high frequency/low temperature and
low frequency/high temperature. Other parameters, which introduce viscous effects,
remain the same. The thermodynamic stability condition should remain verified on the
whole frequency-temperature range.
E 0 - E 00
E () E 00 +
*
(15)
-k -h -1
1+(j) +(j) +(j )
*
E () - E 00
() = 00 + 0 - 00
(16)
*
E 0 - E 00
: the pulsation, = 2
f (f is the frequency)
k, h: exponents such as 0 < k < h < 1, : constant
E00: the static modulus as 0,
E0: the glassy modulus as

: the Newtonian viscosity, = (E0 E00) .
: characteristic time, which value varies only with temperature T, (T) = aT(T)0
where 0 = (Tref) is determined at the reference temperature Tref
00: the static Poissons ratio when 0
0: the glassy Poissons ratio when
At a given temperature, 7 constants (E00, E0, , k, h, , ) are required to
completely characterize the 1D LVE properties of the considered material. When 3D
LVE properties are considered, 9 constants (7 constants of 1D case plus 00 and 0) are
required. evolution is approximated by a WLF type law (Eq. 17 & Figure 17). When
temperature effect is considered, the number of constants amounts to 11, including the
2 WLF constants (C1 and C2 calculated at the reference temperature).
C1 T - Tref
log a T
= - (17)
C2 + T - Tref
Simulation curves obtained with 2S2P1D model are plotted together with data in
Figures 14 to 17 for the materials presented in [55], [56], which correspond to very
different types of BM containing or not recycled asphalt pavement (RAP). These
figures confirm that 2S2P1D model can well simulate behaviour of BM on a wide
range of temperatures and frequencies. Many other examples for very different types of
BM can be found in the authors papers given in the list of references.
When considering normalized values of complex modulus E*nor and Poissons
ratio *nor (Eqs. 18 and 19), curves in complex plane become independent of
parameters E0, E00, 00 and 0. In addition if parameter is fixed by adjusting reference
temperature Tref, a unique master curve is obtained for E*nor and *nor if only
parameters , k, h and are fixed. Another output from ENTPE team is that parameters
, k, h and (and also C1 and C2) are fixed by the origin (crude oil) of binder used in
the mastic or in the mix. For different granular skeletons (mastics or mixes) for
different void contents a unique master curves should then be obtained for E*nor and
*nor, if these materials are made with a bitumen from the same origin. This affirmation
is confirmed by Figure 18 where master curves of the norm of normalized values of
data presented in Figures 14 to 16 are plotted. Even if scattering is higher for Poissons
ratio values, all modulus and Poissons ratio normalized data are close to the unique
2S2P1D simulated curve for the 9 different BMs made with bitumens from same origin.
As predicted by Eqs. 18 and 19, all data range between 0 (at low frequency/high
temperature) and 1 (at high frequency/low temperature).

(18)

*av
* (19)

C4P1M1 (9.8%)-Tref=15.5C
C2P1M1 (10.2%)-Tref=15.7C
1.0 C5P2M1-2 (6.9%)-Tref=17.8C Lines : 2S2P1D modelling
C6P2M1-2 (6.8%)-Tref=17.7C
C6P1M2 (11.7%)-Tref=15.2C
0.8 C2P2M2 (11.4%)-Tref=15.1C
C2P2M2-2 (6.8%)-Tref=15.2C
C5P2M2-2 (6.9%)-Tref=15.4C
|(E*-E00)|/(E0-E00)
0.6
C2P2M3 (7.2%)-Tref=15.4C
C7P2M3 (7.1%)-Tref=15.2C
C2P2M4 (7.6%)-Tref=14.1C
0.4 = 0.2 C3P2M4 (7.4%)-Tref=14.0C
C3P2M5 (7.4%)-Tref=12.9C
C5P2M5 (7.3%)-Tref=13.1C
0.2 C2P2M6 (7.5%)-Tref=14.7C
C5P2M6 (7.3%)-Tref=14.8C
C2P2M7 (7.1%)-Tref=13.5C
C4P2M7 (7.4%)-Tref=13.5C
0.0
-7 -5 -3 -1 1 3 5 7 9
10 10 10 10 10 10 10 10 10
f.aT (Hz)
Figure 18. Master curves of norm of normalized complex modulus (above) and norm of normalized complex
Poissons ratio (below) with 0 = 0.2 (same data as Figures 14 to 17): 2S2P1D simulation (which gives a
unique curve) and data
GKV -Generalized Kelvin Voigt- model
The 2S2P1D model developed at ENTPE laboratory is well adapted to simulate the
linear viscoelastic behaviour of bituminous materials (mixes, mastics and bitumen) as
previously shown. Besides, this model is easy to calibrate due to a rather few numbers
of constant to set up. However the 2S2P1D model has a continuous spectrum and can
hardly be used in finite element calculation of structures. This explains why the global
strategy developed at ENTPE is to use 2S2P1D model for the calibration of
Generalized Kelvin Voigt model (GKV) (Figure 1) having n elements. The number
of elements n is the choice of the user. As LVE theory shows the fitting becomes
perfect when number of elements n becomes infinity, it is clear that the quality of
fitting improves with n. An optimization procedure allowing to obtain the 3n + 2
constants (consisting of Ei, i and i of each body of GKV model) from the 9 constants
of 2S2P1D model, is presented in [59]. Figure 19 presents the relative difference
obtained for the norm of complex modulus between 2S2P1D model and three GKV
models having 5, 10 and 15 elements, respectively. It can be seen in this figure that 5
elements can give up to 50% difference depending on frequency and temperature. Then
it appears that n=5 is an inferior limit to be considered in order to obtain a correct
approximation of BM behaviour to be used for calculation of structures. Considering 2

or 3 elements, as often proposed in literature, is obviously a rough approximation.
D EGKV
E2*S 2 P150 *
(%) GKV (n=5)

E2*S 2 P1D GKV (n=15)
GKV (n=25)
40
30
20
10
0
-10 -5 0 5 10 15
10 10 10 10 10 10
Frequency (Hz)
Figure 19. Relative difference of norm of complex modulus between GKV models (5, 15, 25 elements) and
2S2P1D model calibrated for a BM made with 6% of 50/70 pen bitumen (from[59])
Tr=10C
3
10
E* (MPa)
1
10
2S2P1D model
GKV model (n=15)
experimental data-B5070
experimental data-M507032
-1 experimental data-Mix6%5070
10
-6 -4 -2 0 2 4 6
10 10 10 10 10 10 10
Frequency (Hz)
2S2P1D model
Tr=10C
GKV model (n=15)
Experimental data-B5070
0.5 Experimental data-M507032
Experimental data-Mix6%5070
0.4
*
B5070 and M507032

0.3
Mix6%5070
0.2
-6 -2 2 6 10 14
10 10 10 10 10 10
Frequency (Hz)
Figure 20. Master curves of norm of complex modulus (above) and norm of complex Poissons ratio (below).
Experimental data, 2S2P1D and 15-element GKV simulations plotted for 3 types of bituminous materials:
bitumen (50/70 pen), mastic (32 % of filler in 50/70 pen bitumen) and mixture (6% of 50/70 pen bitumen )
((from[59])
Experimental data, 2S2P1D and 15-element GKV simulations are considered in

Figure 20 presenting master curves of norm of complex modulus and norm of complex
Poissons ratio. In this figure, 3 types of bituminous materials are presented: bitumen
(50/70 pen), mastic (32 % of filler in 50/70 pen bitumen) and mixture (dosed at 6% of
50/70 pen bitumen) ([59]). 15 elements for GKV model appears as a good compromise
between correct modelling of BM behaviour and limited number of elements assuring
not too long computation time for finite element calculation convergence.
3.2.2.3. Wave propagation analysis

Different wave propagation or seismic tests were also proposed for BM ([52], [63],
[64], [65], [66]). Due to the great importance of viscous effects, the analysis of
dynamic tests should be done considering LVE behaviour.
Figure 21. Illustration of the dynamic test set-up used in [65]
Figure 22. Comparison of the complex modulus, for a BM labelled GB3 with French standard, including
absolute value (a), phase angle (b), and loss and storage modulus (c): data points from cyclic T/C test and
simulated curves, with 2S2P1D and NH models calibrated from dynamic (seismic) test result back analysis
using set-up presented in Figure 21. (from [65])
Figures 22 and 23 present some results from Gudmarsson et al. 2014. The applied
seismic test method (schematized in Figure 21) includes measurements of frequency
response functions (FRFs) at different temperatures and optimization of calculated
FRFs (using FEM calculation) to characterize the material properties. Back analysis of
seismic tests provides the constants of the 2 considered models: 2S2P1D and HN
(Havriliak-Negami). Complex modulus and complex Poissons ratio can both be
obtained. A good approximation is found for complex modulus results (Figure 22),
with respect to T/C test results. Although complex Poissons ratio results (Figure 23)
show higher differences, this method appears overall quite promising.
Figure 23. Comparison of complex Poissons ratio, for a BM labelled GB3 with French standard, including
absolute value (a), phase angle (b), and imaginary and real part of complex Poissons ratio (c): data points
from cyclic T/C test and simulated curves, with 2S2P1D and NH models calibrated from dynamic (seismic)
test result back analysis using set-up presented in Figure 21. (from [65])
3.2.3. Examples of structure simulations

This section focuses on the linear behaviour that is considered for road design. Two
examples show the importance of choosing a correct behaviour for stress-strain field
calculation in structures including BM. The first example considers the French five
point bending test. The second example presents numerical simulations for the Millau
Viaduct, known as one of the highest (240m high) and longest multiple cable stayed
bridge (2460m long) in the world. The presented results were obtained during the PHD
thesis of S. Pouget [14]. Most of the work was performed during French national
project called Orthoplus: Advanced engineering of orthotropic decks and their
wearing courses for a global optimization of their life-cycle
3.2.3.1. French five point bending test: LVE versus LE

The five point bending test (Figure 24) is the French normalized test (French standard
NF P 98-286) which is used to study fatigue of BM surfacing of orthotropic steel
bridges. Finite element calculation results considering linear elastic behaviour (always
considered in practical cases) and linear viscoelastic behaviour (19-element GKV
model optimized from 2S2P1D model having frequency (or time) dependent Poissons
ratio) for this test are compared. Geometry, boundary conditions and FEM mesh of the
specimen, are given in Figure 24. The specimen is made of: a 12-mm thick steel plate
(classical thickness for fast lanes on orthotropic steel bridges with a bituminous
surfacing), a 3-mm thick sealing which provides an impermeable barrier and absorbs
differential movements between the surface course and the steel plate and a 60-mm
thick bituminous mix. The bottom central support is considered as clamped while the
two others allow free horizontal movement. The sinusoidal compression loading P is
ranging between 0 and -0.72MPa at a 4Hz frequency. The simulation is made for
10000 seconds, which correspond to 40000 cycles. Steel and bituminous sealing sheet
are assumed to be isotropic linear elastic materials. The bituminous mix behaviour is
LVE. The FEM calculations are performed at two temperatures (10C and 30C) in
order to underline the influence of temperature and viscous effects. Calculations
considering linear elastic properties for the bituminous surfacing are also proposed to
compare both cases: LE and LVE. The reader can find more details about five point
bending test and the numerical simulations in [67].
Figure 24. Five point bending test, picture (above) and geometry, boundary conditions and FEM mesh
(below) (from [67]).
Stress-strain curves (component xx) at point A (Figures 24 and 25) in the
bituminous mix for 40000 cycles are plotted in Figure 25 for two LE and LVE cases at
10C and 30C. In the LE cases, only tension occurs during the considered cycles. In
the LVE cases, the very first cycles create mainly tension but rapidly both tension and
compression appear. This observation reveals that fatigue of bituminous mix during the
test is not due to stress cycles only in tension (which would probably induce large
permanent strain ([68], [69]), but is created by successive tension and compression
phases. The result is at odds with the classical interpretation of the French Five Point
Bending Test, which considers LE analysis giving only tension. It is clear that LVE
calculations, which fit more correctly the real BM behaviour, give correct stress-strain
evolution.
2.8
2.4 LVE1 (10C) t=0s
2.0 LVE2 (30C)
1.6 T=30C T=10C
1.2 t=10000s t=10000s
0.8 cycle n40000 cycle n40000
(MPa)
0.4 10 C
0.0
-0.4
-0.8 30 C
T=10C; t=41s
-1.2 LVE T=30C; t=0.1s
-1.6
-4000 -3000 -2000 -1000 0 1000
2.8
2.4 LE1 (10C)
2.0 LE2 (30C)
1.6 LE 10 C
1.2
0.8
(MPa)
0.4 30 C
0.0
-0.4 No compression in the
-0.8 Linear Elastic (LE) case
-1.2
-1.6
-4000 -3000 -2000 -1000 0 1000
-6
(10 )
Figure 25. Calculated stress-strain curve at point A (Figure 24) for two temperatures (10C and 30C) and
for two cases, LE (linear elastic) and LVE (linear viscoelastic). For clarity of the figure, only the first four
cycles and the last cycle (40000) are plotted (from [67])
3.2.3.2. Linear viscoelastic calculation of Millau bridge response

The Millau viaduct and the section considered for FEM calculation are presented in
Figure 26. FEM calculation is made on a 10.5-m long, 1.2-m wide and 0.595-m high
part of the orthotropic structure. The deck is composed of a 60-mm thick bituminous
mix, a 3-mm thick sealing sheet and a 12-mm thick steel plate. The goal of the
investigation is to evaluate the influence of the viscous behaviour of the bituminous
mix and sealing sheet on the response of orthotropic steel deck bridges loaded by a
moving wheel. Perfect bond is assumed between the different layers. Loading consists
in a truck wheel load moving at constant rate along the longitudinal direction. Isotropic
Linear Viscoelastic behaviour (ILVE) is considered for the bituminous materials
(wearing course and sealing sheet). A 5-element GKV model, optimized from 2S2P1D
model using advanced viscoelastic characterization of bituminous materials developed
at ENTPE laboratory (see previous section) is chosen for finite element calculation.
Steel is assumed to be an Isotropic Linear Elastic (ILE) material. In this paper,
implementation of the model in the finite element software, numerical calculations
etc. are not detailed. This information is given in [70], [71].
To characterize the influence of temperature and viscous effects, calculations are
carried out at 3 temperatures (0C, 30C and 60C) and for 2 wheel speeds (50km/h
and 100km/h) (Figure 27). Another calculation considering steel plate only is also
performed. The deflection and the longitudinal stress yy at point B (Figure 26), below
the steel plate, are plotted in Figure 27 as functions of wheel position. The results are
plotted in the mid-length of the bridge, which corresponds to the main critical point for
this kind of structure. It can be noticed that the deflection and the longitudinal stress yy
in the steel at point B increase significantly with temperature. The calculations in the
steel structure display more severe stress when neglecting the contribution of the
bituminous layers. Wheel speed has a greater influence at 30C compared to the two
other temperatures (0C and 60C). In particular, viscous effects at 30C are clearly
shown by the non-symmetry of the yy curve (no compression after the wheel pass) and
induce a maximum of yy values, at point B, when the wheel is about 10 cm from the
vertical position of point A. At high temperature the influence of the surfacing is rather
negligible. As a general observation, temperature and viscous effects have a great
importance on stress level in the steel (from 5MPa to 55MPa). Taking them into
account for bridge design would definitely improve the optimization of the structure.
a)
b)
c)

Figure 26. View of the Millau Viaduct (France) (a) and section considered for FEM calculation including
dimension (b) and finite element mesh, boundary conditions and vertical displacement field (c) ([70], [71])
70
steel plate only
60
longitudinal stress yy (MPa)
60C
50
50 km/h
40 100 km/h 30C
50 km/h
30 100 km/h
20 0C
50 km/h
10 100 km/h
0
-10
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
wheel position (m)

Figure 27. Results at point B below the steel plate (Figure 26), for the simulated scenarios: vertical
displacement (above) and longitudinal stress yy (below) (from [70])
4. Behaviour outside the linear domain
Due to the lack of available space, this section only presents a rapid overview of
ENTPE team outcomes.
4.1. Unbound Granular Materials (UGM)
A new elastoplastic model called LMS (Loading Memory Surface) based on the critical
state concept and the multi-surface framework is proposed for geomaterials ([32]). The
model uses the hypo-elastic formulation presented in section 3.1 and two plastic
mechanisms. This model was developed taking into account the following constraints:
i) having a limited number of constants, ii) considering constants and mechanisms
linked with classical soil mechanics parameters, iii) obtaining cyclic behaviour from
monotonic loading and, iv) remaining within a simple but rigorous framework. The
LMS model has no elastic domain.
The most innovative part of the model is the Cyclic Hardening (CH) mechanism
that can be created, activated, inactivated and erased inside the first Envelope
Hardening (EH) mechanism. The newly introduced formalism gives a 3-dimension
generalization of the well known Masing rule. One great interest is that cyclic response
is directly obtained from the first monotonic loading. Figure 28 gives an example of
creation and evolution rules for CH and EH mechanisms.
The model has been validated against test results for Hostun sand under several
conditions: monotonic and cyclic, drained and undrained, tests in contraction and in
extension, at different confining pressures and different densities. Figure 29 presents an
example (from [32]) of comparison between simulation with LMS model and
experimental data for drained cyclic triaxial test on Hostun sand at e0 = 0.77 and at PC
= 200kPa. The fitting is not perfect but can be considered as satisfactory when
considering that LMS model was calibrated with triaxial monotonic loadings.
It should be underlined that although modelling with LMS model is rather
satisfactory when considering loadings without rotation of axes, it fails in capturing
some of the features observed when rotation of axes is predominant [37]. To the
authors knowledge, this remark is probably valid for all proposed incremental models.
More reliable experiment data are still missing for this type of complex loading paths,
which exists within all civil engineering structures or infrastructures.
4.2. Bituminous Mixtures (BM)
As shown in figure 10 different domains of behaviour can be identified outside the

linear domain for bituminous mixtures. Fatigue, non-linearity, permanent deformation
can appear. Phenomena such as healing, thixotropy, crack propagation, brittle failure or
viscoplastic flow, can also exist depending on temperature, loading path and loading
rate. In addition, thermo-mechanical coupling is always predominant for these
materials. Due to limited available space, no more development is proposed in this
paper. Interested readers could refer to the following references from ENTPE team,
where different aspects of the BM behaviour outside the linear domain are investigated:
[36], [57], [68], [69], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83],
[84], [85], [86], [87], [88].
EH surface fd s1 s1
active
fd in memory
s 1st CH surface fd(1) s sinv

1
active
s01
s3 s2 s3 s2
(a) first loading to s : EH (b) unloading from sinv

1
to s : 1st CH
s1 s1
fd in memory
fd in memory
fd(1)in memory
s inv
2
s inv
sinv
1 1st CH surface fd(1) 1

active s
2st CH surface fd(2)

s 0
2
active
s 0
1
s 0
1
s
s3 s2 s3 s2
sinv
to s : 2nd CH sinv
2 1
(d) unloading from (c) unloading from to s : 1st CH
s1 s1
fd in memory
fd = fd(1)
2
1
sinv sinv
1
s inv
fd(1) = fd(2)
s 0
s0
1 2
s01
s3 s s2 s3 s2
s
sinv
to s : 2nd CH
2 s1
(e) unloading from (f) reloading to s : 1st CH
fd active
s3 s2
s
(g) loading to s : EH

Figure 28. Mechanisms of appearance, evolution and vanishing of the yield surfaces in the deviatoric
normalized plane during a complex loading path involving envelope (EH) and cyclic (CH) hardenings
(from [32])
700
Hostun sand - Pc = 200kPa - e0 = 0.77
600
500
400
q [kPa]
300
200
100
0 simul simul
exp exp
-100
0 2 4 6 8 10 12 14 9 10 11
5 [%] 1 [%]
Hostun sand - Pc = 200kPa 1- e0 = 0.77
4
3
v [%]
0 simul simul
exp exp
-1
0 2 4 6 8 10 12 14 9 10 11
1 [%] 1 [%]

Figure 29. Simulation and experimental data: drained cyclic triaxial test on Hostun sand at e0 = 0.77 and at
PC = 200kPa - right zoom on the last cycles (from Di Benedetto et al. 2014). LMS model calibrated with
triaxial monotonic loadings (from [32])
5. Conclusion
This paper gives an overview of the results obtained on unbound granular materials
(UGM) and bituminous mixtures (BM) at the University of Lyon/ENTPE. All the
results and developments could only be obtained thank to Advanced Laboratory
Testing as promoted by Bishop and Technical Committee 101 of ISSMGE.
Mechanical behaviour of UGM and BM is very complex but some similarities are
observed. Considering this remark, which is also valid for all geomaterials, a unique
general framework for rheological modelling is proposed. This framework, based on 3
component model and extension, is versatile. It is used in this paper for both types of
materials. Behaviours inside (where strain amplitude is smaller than some 10-5 m/m and
number of cycles limited) and outside the linear domain are considered. Focus is made
on viscous and anisotropic effects for monotonic, cyclic and dynamic types of loading.
Some of the presented results and modelling are quite new and reveal surprising
characteristics of material behaviour. For example, Poissons ratio values, which are
time (or frequency) and temperature dependent for BM.
Within the linear behaviour domain, examples of structure calculations or back
calculations show the importance of considering correctly viscous and anisotropic
properties. In particular for BM having viscous effects of paramount importance, linear
viscoelastic stress-strain fields obtained with generalized Kelvin Voigt model,
optimized from continuous spectrum 2S2P1D model (developed by the ENTPE team),
are very different from elastic results. This observation is true for the two presented
examples of finite elements calculations: the French five point bending test and the
Millau Viaduct. Presented examples also reveal that dynamic back analysis can give
good results if anisotropy and viscous effects are correctly introduced.
Outside the linear domain, non-linearities and other phenomena appear and may
become predominant. For UGM an elastoplastic type model, called LMS, was
developed. For BM some studies deal with fatigue, non-linearity and permanent
deformation involving or not phenomena such as healing, thixotropy, crack
propagation, brittle failure or viscoplastic flow, which can also exist depending on
temperature, loading path and loading rate.
Even if a huge literature is devoted to material behaviour some issues still remain
to dig or to discover. One of the key ones may be the influence of the rotation of
principal axes, systematically observed in engineering structures.
As last and main conclusion, it should be underlined that a correct description of
material behaviour is of paramount importance for engineering construction design.
Some important outputs have been obtained up to now but materials are far from
having delivered all their mystery.
6. Acknowledgements
I wish to thank all my previous and present PhD students. They allowed obtaining nice
experimental results following long and painful hours in the lab. Without these data
and scientific exchanges with them, all the developments presented in this paper would
not have been possible. Special thanks are devoted to Cdric Sauzat, with whom I
have been closely working on these topics since more than 17 years.
I also thank Professors Flix Darve and Fumio Tatsuoka for a long friendship
sowed of animated and interesting discussions, with one more on the theoretical side
and with the other on the experimental and practical aspects.
References
[1] Hameury O., "Quelques aspects du comportement des sables avec ou sans rotation des axes principaux
des petites aux grandes dformations", PhD thesis, ENTPE, 1995 [in French]
[2] Boubbaneh E., Comportement mcanique des enrobs bitumineux des "petites" aux "grandes"
dformations, PhD thesis, ENTPE, 1995 [in French]
[3] Cazacliu B., Etude du comportement des sables en petites et moyennes dformations: prototype
dessais de torsion, compression, confinement sur cylindre creux, PhD thesis, ENTPE, 1996 [in French]
[4] Neifar M., Comportement thermomcanique des enrobs bitumineux: exprimentation et modlisation,
PhD thesis, ENTPE, 1997 [in French]
[5] Ibraim E., Diffrents aspects du comportement des sables partir d'essais triaxiaux: des petites
dformations la liqufaction cyclique, PhD thesis, ENTPE, 1998 [in French]
[6] Sauzat C., Comportement des matriaux granulaires, influences de l'amplitude et du chemin de
sollicitation", PhD thesis, ENTPE, 2003 [in French]
[7] Olard F., Comportement thermomcanique des enrobes bitumineux basses tempratures : Relations
entre les proprits du liant et de lenrob, ENTPE, 2003 [in French]
[8] Pham Van Bang D., Comportement instantan et diffr des sables des petites aux moyennes
dformation : exprimentation et modlisation, PhD thesis, ENTPE, 2004 [in French]
[9] Duttine Antoine, Comportement des sables et des mlanges sable/argile sous sollicitations statiques et
dynamiques avec et sans rotations daxes , PhD thesis, ENTPE, 2005 [in French]
[10] Delaporte Brice, Etude de la rhologie des mastics bitumineux laide dun rhomtre cisaillement
annulaire , PhD thesis, ENTPE, 2007 [in French]
[11] Ezaoui Alan, Comportement exprimental statique et dynamique des sables et modlisation lasto-
visco-plastique anisotrope en petites et moyennes dformations, PhD thesis, ENTPE, 2008 [in
French]
[12] Nguyen Hoang Minh, Comportement cyclique et dformations permanentes des enrobs bitumineux,
Thse de doctorat, PhD thesis, ENTPE, 2010 [in French]
[13] Clech Pauline, Comportement des enrobs bitumineux sous sollicitation multidirectionnelle, PhD
thesis, ENTPE, 2010 [in French]
[14] Pouget Simon, Influence des proprits lastiques ou viscolastiques des revtements sur le
comportement des ponts dalle orthotrope, PhD thesis, ENTPE, 2011 [in French]
[15] Blanc Matthieu, Etude exprimentale et modlisation du comportement des sables avec rotation
daxes principaux de contrainte , PhD thesis, ENTPE, 2011 [in French]
[16] Nguyen Quang Tuan, Comportement thermomcanique des enrobs bitumineux sous sollicitations
cycliques dans les domaines linaire et non-linaire , PhD thesis, ENTPE, 2011 [in French]
[17] Tapsoba Nouffou, Comportement des enrobs bitumineux base de matriaux recycls et/ou
fabriqus temprature rduite , PhD thesis, ENTPE, 2012, [in French]
[18] Tiouajni Samir, Etude du comportement des sables avec analyse de la coaxialite : exprimentation et
modlisation , PhD thesis, ENTPE, 2013 [in French]
[19] Mangiafico Salvatore, Linear viscoelastic properties and fatigue of bituminous mixtures produced with
reclaimed asphalt pavement and corresponding binder blends, PhD thesis, ENTPE, 2014
[20] Lamothe Sbastien, Endommagement dun enrob bitumineux partiellement satur en eau ou en
saumure soumis des sollicitations cycliques de gel-dgel et mcanique, thse en cotutelle ETS &
ENTPE, July 2014, [in French]
[21] Pham Nguyen Hoang, Etude thermomcanique avance de diffrents types denrobs recycls tides
avec additifs, Thse de Doctorat, ENTPE, December 2014, [in French]
[22] Tatsuoka,F. and Shibuya,S. (1991a), Deformation characteristics of soils and rocks from field and
laboratory tests, Keynote Lecture for Session No.1, Proc. of the 9th Asian Regional Conf. on SMFE,
Bangkok, Vol.II, pp.101-170
[23] Tatsuoka,F., Jardine,R.J., Lo Presti,D., Di Benedetto,H. and Kodaka,T. (1999a): Characterising the
Pre-Failure Deformation Properties of Geomaterials, Theme Lecture for the Plenary Session No.1,
Proc. of XIV IC on SMFE, Hamburg, September 1997, Volume 4, pp.2129-2164.
[24] Di Benedetto,H. and Tatsuoka,F. (1997): Small strain behaviour of geomaterials: Modelling of strain
rate effects, Soils and Foundations, Vol.37, No.2, pp.127-138
[25] Jardine,R.J. (1992): Some observations on the kinematic nature of soil stiffness, Soils and
Foundations, 1992, Vol. 32, No 2, pp. 111124
[26] Lo Presti, D., Shibuya,S. and Rix,J.,G. (2001): Innovation in soil testing, Key Note Lecture, Proc. of
the Second International Conference on Pre-failure Deformation Characteristics of Geomaterials,
Torino, 1999, Balkema (Jamiolkowski et al., eds.), Vol. 2, pp1027-1076.
[27] Tatsuoka F., Laboratory stress-strain tests for developments in geotechnical engineering research and
practice, first Bishop Lecture, International Symposium on Deformation Characteristics of
Geomaterials, September 1~3, Seoul, Korea, 2011
[28] Jardine R.J., Advanced laboratory testing in research and practice: the Second Bishop Lecture, La
deuxime confrence Bishop : Utilisation des essais en laboratoire de pointe dans la recherche et
dans l'industrie , Second Bishop Lecture, Proceedings of the 18th International Conference on Soil
Mechanics and Geotechnical Engineering, Paris 2013
[29] Di Benedetto,H., Geoffroy,H. and Sauzeat,C. (2001b): Viscous and non viscous behaviour of sand
obtained from hollow cylinder tests, Advanced Laboratory Stress-Strain Testing of Geomaterials
(Tatsuoka et al. eds.), Balkema, pp.217-226
[30] Di Benedetto, H., Tatsuoka, F., Lo Presti, D., Sauzeat C., Geoffroy H., (2005a). Time effects on the
behaviour of geomaterials. Keynote lecture in : Di Benedetto H. et al. Eds. Proc. of the 3rd Int. symp.
on Deformation Characteristics of Geomaterials, sept. 2003, Lyon. Rotterdam, Pays-Bas : Balkema,
vol.2, pp59-124
[31] Di Benedetto, Small Strain Behaviour and Viscous Effects on Sands and Sand-Clay Mixtures, Keynote
lecture and paper, Geotechnical Symposium Roma 2006, Ed.H. ling et al., Springer, SOIL STRESS-
STAIN BEHAVIOR: MEASUREMENT, MODELING AND ANALYSIS Book Series: SOLID
MECHANICS AND ITS APPLICATIONS, Volume 146, pp. 159-190, 2007
[32] Di Benedetto H., Blanc M. , Tiouajni S., Ezaoui A., Elasto-plastic model with loading memory surfaces
(LMS) for monotonic and cyclic behaviour of geomaterials, Int. Jl for Numerical and Analytical
Methods in Geomechanics, Vol. 38, pp.1477-1502, [DOI: 10.1002/nag.2265], 2014
[33] Di Benedetto H. (1987). Modlisation du comportement des gomatriaux : application aux enrobs
bitumineux et aux bitumes. Thse de docteur dEtat. Grenoble : USTMG, 1987. [in French]
[34] Di Benedetto H., F. Olard, C. Sauzat, B. Delaporte (2004a), Linear viscoelastic behavior of
bituminous materials: from binders to mixes, Road Materials and Pavement Design, Vol.5, Special
Issue, pp. 163-202.
[35] Di Benedetto H., B. Delaporte, C. Sauzat, (2007a) Three-dimensionnal linear behavior of bituminous
materials: experiments and modelling, International Journal of Geomechanics (ASCE), Volume 7
(issue 2), pp. 149-157
[36] Di Benedetto H., Neifar M., Sauzeat C, Olard F., (2007b) Three-dimensional thermo-viscoplastic
behaviour of bituminous materials: the DBN model, Road Materials and Pavement Design, Vol 8,
Issue 2, pp. 285-316.
[37] Blanc M, Di Benedetto H, Tiouajni S. Deformation characteristics of Hostun sand under loadings with
principal stress axes rotation. Soils and foundations. 2011;51(4):749-60
[38] Ezaoui A, Di Benedetto H. Experimental measurements of the global anisotropic elastic behaviour of
dry Hostun sand during triaxial tests, and effect of sample preparation. Gotechnique. 2009;59(7):621-
35
[39] Di Benedetto,H., Tatsuoka,F. and Ishihara,M. (2002): Time-dependent shear deformation
characteristics of sand and their constitutive modelling, Soils and Foundations, 42-2, pp.1-22
[40] H. Di Benedetto, H. Geoffroy, A. Duttine, C. Sauzat, B.Chau, Anisotropic behaviour of soils and site
investigation based on wave propagation tests , Proceedings of the 16th International Conference on
Soil Mechanics and Geotechnical Engineering, Vols 1-5 - GEOTECHNOLOGY IN HARMONY
WITH THE GLOBAL ENVIRONMENT, pp. 675-680, 2005b
[41] Di Benedetto H., Blanc M., Ezaoui A., Geoffroy H., Viscous behavior of dry sand: different
experimental loading paths and general modelling , Proc. 4th Int. Symp. on Deformation
Characteristics of Geomaterials, IS Atlanta, Ed. Burns et al., IOS Press, pp. 495-502, 2008
[42] Tatsuoka F, Ishihara M, Di Benedetto H, Kuwano R.(2002). Time dependent shear deformation
characteristics of geomaterials and their simulation. Soils and Foundations, vol.42, n2, pp.103-129
[43] Tatsuoka, F., Di Benedetto, H., Enomoto, T., Kawabe, S. and Kongkitkul, W., Various viscosity types
of geomaterial in shear and their mathematical expression, Soils & Foundations,Vol. 48-1, pp. 41-60,
February 2008
[44] Pham Van Bang D, Di Benedetto H, Duttine A, Ezaoui A. Viscous behaviour of dry sand. International
Journal for Numerical and Analytical Methods in Geomechanics. 2007;31(15):1631-58
[45] Tatsuoka, F., Di Benedetto, H., Kongkitkul, W., Kongsukprasert, L., Nishi, T. and Sano, Y., Modelling
of Ageing Effects on the Elasto-viscoplastic Behaviour of Geomaterial, Soils & Foundations,Vol. 48-2,
155-174, April 2008
[46] Duttine A, Di Benedetto H, Pham Van Bang D, Ezaoui A. Anisotropic small strain elastic properties of
sands and mixture of sand-clay measured by dynamic and static methods. Soils and foundations.
2007;47(3):457-72
[47] Ezaoui A., Di Benedetto H., Measurement and modelling of anisotropic elastic behaviour of unboud
granular materials , 1st Int. Conf. on Transportation Geotechnics, advances in transportation
geotechnics, Eds. Ellis et al., CRC press, Nottingham, pp. 125-134, August 2008
[48] Hardin, B.O. and Blandford, G. E. (1989). Elasticity of particulate materials. Journal of Geotechnical
Engineering, ASCE, 115(6), 788805.
[49] Di Benedetto H., M.N. Partl, L. Francken, C. De la Roche, (2001a), Stiffness testing for bituminous
Mixtures, Journal Materials and Structures, Vol 34, pp. 66-70.
[50] DI BENEDETTO H., CORTE J.F., Matriaux Routiers Bitumineux 2: constitution et proprits
thermomcaniques des mlanges, Lavoisier, p. 288, 2005 [In French]
[51] Delaporte B., Di Benedetto H., Chaverot P., Gauthier G., Linear Viscoelastic Properties of Bituminous
Materials: from Binders to Mastics, Presentation at the annual meeting of the Association of Asphalt
Paving Technologists (AAPT), San Antonio (Texas), March 2007, Publication in the Journal of the
AAPT, 34 p., 2007
[52] Di Benedetto H., Sauzat C., Sohm J., (2009) Stiffness of bituminous mixtures using ultrasonic
waves propagation, Road Materials and Pavement Design, Vol. 10, Issue 4, pp. 789-814
[53] Olard F. and Di Benedetto H. (2005), Experimental characterization and constitutive modeling of the
thermo-visco-elasto-plastic behavior of bituminous mixes : the DBN law, Jl of the Association of
Asphalt Paving Technologists, Vol 74, pp. 791-827.
[54] Nguyen H. M., S. Pouget, H. Di Benedetto, C. Sauzat, (2009) Time-Temperature Superposition
Principle for bituminous mixtures, European Journal of Environmental and Civil Engineering, pp.
1095-1107
[55] Pham Nguyen Hoang, Sauzat C., Di Benedetto H., Gonzalez-Leon J.A., Barreto G., Nocolai A.,
Jajubowski M., RAP and additives influence on 3D linear behaviour of warm mix asphalts, Road
Materials and Pavement Design, [10.1080/14680629.2015.1021108], on line first, March 2015a
[56] Pham Nguyen Hoang, Sauzat C., Di Benedetto H., Gonzalez-Leon J.A., Barreto G., Nocolai A.,
Jajubowski M., Analysis and modeling of 3D complex modulus tests on hot and warm bituminous
mixtures, Mechanics of Time Dependent Materials, [DOI: 10.1007/s11043-015-9258-8], published on
line first, March 2015b
[57] Nguyen, Q. T., Di Benedetto, H., Sauzat, C., & Tapsoba, N. (2013). Time temperature superposition
principle validation for bituminous mixes in the linear and nonlinear domains. Journal of Materials in
Civil Engineering, 25(9), 1181-1188
[58] Olard F. and Di Benedetto H. (2003), General 2S2P1D model and relation between the linear
viscoelastic behaviors of bituminous binders and mixes, Journal of Road Materials and Pavement
Design, Vol. 4 (Issue 2), pp. 185224.
[59] Tiouajni S., Di Benedetto H., Sauzat C., Pouget S., (2011), Approximation of linear viscoelastic
model by generalized Kelvin Voigt or generalized Maxwell Models: Application to bituminous
materials in the 3 dimensional case, Road Materials and Pavement Design, Issue 4, Vol 12, pp. 897-
930
[60] Cannone Falchetto A., Marasteanu M. and Di Benedetto H., (2011), Analogical Based Approach to
Forward and Inverse Problems for Asphalt Materials Characterization at Low Temperatures, Journal
the Association of Asphalt Paving Technologists, AAPT, vol 80, pp. 549-582
[61] Mangiafico S., Di Benedetto H., Sauzat C., Olard F., Pouget S., Planque L., New method to obtain
viscoelastic properties of bitumen blends from pure and RAP binder constituents", Road Materials and
Pavement Design, Vol 15, N 2, pp. 312-329, [DOI:10.1080/14680629.2013.870639], 2014
[62] Clech P., Sauzat C. and Di Benedetto H., Linear viscoelastic behavior and anisotropy of bituminous
mixture compacted with a French Wheel compactor. Second GeoShanghai International Conference,
ASCE, p. 10, 2010
[63] Mounier D., Di Benedetto H., Sauzat C., Determination of bituminous mixtures linear properties using
ultrasonic wave propagation, Construction and Building Materials, Volume 36, pp. 638-647,
[DOI:10.1016/j.conbuildmat.2012.04.136] , November 2012
[64] Mounier D., Di Benedetto H., Sauzat C, Bilodeau K., Observation of fatigue of bituminous mixtures
using wave propagation, Journal of Materials in Civil Engineering, to be published, accepted April
2015
[65] Gudmarsson A., Ryden N., Di Benedetto H., Sauzat C.,Tapsoba N., Birgisson B., Comparing linear
viscoelastic properties of asphalt concrete measured by laboratory seismic and tension-compression
tests, Jl of Nondestructive Evaluation, Volume 33, Issue 4, Page 571-582, [DOI 10.1007/s10921-014-
0253-9], 2014
[66] Gudmarsson A., Ryden N., Di Benedetto H,. Sauzat C., Complex modulus and Complex Poisson's
ratio from cyclic and dynamic modal testing of asphalt concrete, Construction and Building Materials,
Vol 88, pp. 20-31, [doi 10.1016/j.conbuildmat.2015.04.007], 2015
[67] Pouget, S., Sauzat, C., Di Benedetto, H., Olard, F., (2010b), Numerical simulation of the five-point
bending test designed to study bituminous wearing courses on orthotropic steel bridge, Materials and
Structures, Vol 43/3, pp. 319-330
[68] Gabet T., Di Benedetto H., Perraton D., De Visscher J., Gallet T., Bankovski W., Olard F., Grenfell J.,
Bodin D., Sauzat C., (2011) French wheel tracking round robin test on a polymer modified bitumen
mixture, Materials et Structures, Volume 44, Number 6, pp. 1031-1046
[69] Perraton D., Di Benedetto H., Sauzat C., De La Roche C., Bankowski W., Partl M. & Grenfell J.,
(2011) Rutting of Bituminous Mixtures: Wheel Tracking Tests Campaign Analysis, Materials &
Structures, Vol. 44, N5, pp. 969-986.
[70] Pouget S., Sauzat C., Di Benedetto H., Olard F., (2010a) From the behaviour of constituent materials
to the calculation and design of orthotropic steel bridge structures, Road Materials and Pavement
Design, Vol 11 SI EATA 2010, pp 111-144.
[71] Pouget S., Sauzat C., Di Benedetto H., Olard F., "Modeling of viscous bituminous wearing course
materials on orthotropic steel deck", Materials and Structures, volume 45, Issue 7, pp. 1115-1125,
[DOI 10.1617/s11527-011-9820-z], 2012
[72] Di Benedetto H., De la Roche C., Baaj H., Pronk A., & Lundstrom R, Fatigue of bituminous Mixtures,
Materials and Structures, p. 15, Vol 37, pp. 202-216, 2004.
[73] Lundstrom R., Di Benedetto H., & Isacsson U., Influence of asphalt mixture stiffness on fatigue failure,
Journal of Materials in Civil Engineering (ASCE), vol 16, issue 6, pp. 516-525, 2004
[74] BAAJ H., DI BENEDETTO H., CHAVEROT P., Effect of binder characteristic on fatigue of asphalt
pavement using an intrinsic damage approach , International Journal Road Materials and Pavement
Design, Vol. 6, Issue 2, pp. 147_174, 2005
[75] Di Benedetto H., Nguyen Quang Tuan, Sauzat C., Nonlinearity, Heating, Fatigue and Thixotropy
during cyclic loading of Asphalt Mixtures, Road Materials and Pavement Design, Issue 1, Vol 12, pp.
129-158, [DOI: 10.3166/RMPD.12.129-158], 2011
[76] Nguyen Quang Tuan, Di Benedetto H., Sauzat C., Determination of Thermal properties of Asphalt
mixtures as another output from cyclic Tension-Compression Test, Road Materials and Pavement
Design, Issue 1, Vol 13, pp. 85-103, (DOI:10.1080/14680629.2011.644082), 2012
[77] Sohm J., Gabet T., Hornych P., Piau J.M., Di Benedetto H., Creep tests on bituminous mixtures and
modeling, Road Materials and Pavement Design, Issue 4, Vol 13, pp. 832-849,
[DOI:10.1080/14680629.2012.735795], 2012
[78] Van Rompu J., Di Benedetto H., Buannic, M., Gallet T., Ruot C., New fatigue test on bituminous
binders: experimental results and modeling, Construction and Building Materials, 37, pp. 197-208,
[doi.org/10.1016/j.conbuildmat.2012.02.099], 2012
[79] Tapsoba N., Sauzat C., Di Benedetto H., Analysis of fatigue test for bituminous mixtures, Jl Mat. in
Civil Eng, Vol. 15, N 6, pp.701-710, [doi: 10.1061/(ASCE)MT.1943-5533.0000636], 2013
[80] Nguyen Mai Lan, Sauzat C., Di Benedetto H., Tapsoba N., Validation of the Time-Temperature
superposition principle for crack propagation in bituminous mixtures, Materials & Structures, volume
46, Issue 7, pp. 1075-87, [DOI: 10.1617/s11527-012-9954-7], 2013
[81] Baaj H., Ech M., Tapsoba N., Sauzat C., Di Benedetto H., Thermomechanical Characterisation of
asphalt mixtures modified with high content of asphalt shingle modifier (ASM) and reclaimed asphalt
pavement (RAP), Materials & Structures, Vol. 46, Issue 10, pp. 1747-67, [DOI 10.1617/s11527-013-
0015-7], 2013
[82] Tapsoba, N., Sauzat, C., Di Benedetto, H., Baaj, H. and Ech, M.. Behavior of asphalt mixtures
containing RAP and shingles, Road Materials and Pavement Design, Vol 15, N 2, pp. 330-347,
[DOI:10.1080/14680629.2013.871091], 2014
[83] Mangiafico S., Sauzat C., Di Benedetto H., Pouget S., Olard F., Planque L., van Rooijen R., Statistical
Analysis of Influence of Mix Design Parameters on Mechanical Properties of Mixes with Reclaimed
Asphalt Pavement, Transportation Research Record: Journal of the Transportation Research Board, No
2445, pp. 29-38, [DOI:10.3141/2445-04], 2014
[84] Gayte P., Di Benedetto H., Sauzat C., Three dimensional behaviour of bituminous mixtures in the
linear viscoelastic and viscoplastic domains: the DBN model, 12th Int. Conf. of Int. Soc. Asphalt
Pavements, Railey NC, p. 12, USA, June 2014
[85] Blanc J., Gabet T., Hornich P., Piau J.M., Di Benedetto H., Cyclic triaxial tests on bituminous mixtures,
Road Materials & Pavement Design, Vol. 16, N1, pp. 46-69, [DOI: 10.1080/14680629.2014.964293],
2015
[86] Mangiafico S., Di Benedetto H., Sauzeat C., Olard F., Pouget S., Dupriet S., Planque L., Van Rooijen,
Statistical analysis of the influence of RAP and Mix Composition on Viscoelastic and Fatigue
Properties of Asphalt Mixes, Materials & Structures, Vol 48-4, pp. 1187-1205, [DOI 10.1617/s11527-
013-0225-z], 2015
[87] Nguyen Quan Tuan, Di Benedetto H., Sauzat C., Linear and nonlinear viscoelastic behaviour of
bituminous mixtures, Materials & Structures, [DOI: 10.1617/s11527-014-0316-5], published on line
first May 2014
[88] Tapsoba N., Sauzat C., Di Benedetto H., Baaj H. and Ech M., Three dimensional analysis of fatigue
tests on bituminous mixtures, Jl of Fatigue & Fracture of Eng Mat. & Struc., Vol 38-6, pp. 730-741,
[DOI: 10.1111/ffe.12278],2015
40 Deformation Characteristics of Geomaterials
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-40
Compaction Characteristics and Physical

Properties of Compacted Soil Controlled
by the Degree of Saturation
Fumio TATSUOKA
Tokyo University of Science, Chiba, Japan
Abstract. The conventional fill compaction procedure usually controls the dry
density d and the water content w referring to the maximum dry density (d)max and
the optimum water content wopt determined by laboratory compaction tests per-
formed at a certain compaction energy level (CEL). However, (d)max increases and
wopt decreases with CEL, while, since Proctor (1933), CEL practically available in
the field has been increasing and the required d value has generally been becom-
ing higher for more satisfactory performance of soil structures. Besides, the values
of (d)max and wopt change with soil type. In a single earthwork project, the actual
CEL and soil type may vary and it is very difficult to accurately estimate the field
CEL and identify the actual soil type at a given moment at a given place. There-
fore, the actual values of (d)max and wopt are usually unknown. On the other hand,
the optimum degree of saturation (Sr)opt defined as Sr when (d)max is obtained for a
given CEL and the d/(d)max vs. Sr (Sr)opt relation are rather independent of CEL
and soil type. The unsoaked and soaked strength and stiffness and the saturated
hydraulic conductivity are a function of d and compacted Sr not including CEL as
a variable. It is proposed to control Sr of compacted soil to be close to (Sr)opt and d
to be large enough to achieve the physical properties required in design, together
with pre-compaction control of water content.
Keywords. CBR, compaction energy, dry density, hydraulic conductivity, soil

compaction, the degree of compaction, the degree of saturation, water content
1. Introduction
The conventional soil compaction procedure usually controls the dry density, d, and
the water content, w, referring to the maximum dry density, (d)max, and the optimum
water content, wopt, obtained by laboratory compaction tests on a representative sample
at a certain compaction energy level, CEL (Fig. 1). However, since Proctor (1933) in-
troduced the above, the CEL practically available in the field has been increasing and
the value of d required to ensure satisfactory performance of soil structures has gener-
ally been becoming higher. Therefore, compaction at the water content, w, equal to wopt
evaluated at a CEL lower than the field CEL may be inefficient and even over-
compaction may take place. Besides, even when the soil type and the values of w and
CEL are nominally fixed at a given site, they inevitably vary. As the soil becomes
coarser and better graded and as CEL increases, (d)max increases and wopt decreases,
while it is very difficult to evaluate these variations in the field. Therefore, the accurate
values of (d)max and wopt at each moment and each location in the field are usually un-
known.
F. Tatsuoka / Compaction Characteristics and Physical Properties of Compacted Soil 41
Figure 1. Definition of the degree of compaction, Dc.
In this paper, it is shown that the optimum degree of saturation (Sr)opt defined as Sr
when (d)max is obtained and the d/(d)max vs. Sr (Sr)opt relation, both for a given CEL
and a given soil type, are essentially independent of CEL and soil type. Besides, the
CBR values before and after soaking and the coefficient of hydraulic conductivity un-
der saturated conditions of compacted soil are a function of d and Sr at the end of
compaction, not including CEL as a variable. Based on these findings, it is proposed
to control the value of Sr of compacted soil to become equal to (Sr)opt and d to be large
enough to achieve the soil properties required in the design of a given soil structure,
together with pre-compaction control of the water content of fill material.
2. Compaction Characteristics
2.1. Several Issues with Conventional Soil Compaction Control
A series of full-scale compaction tests using a wide variety of compaction machines

used in field earthworks was performed in a large concrete pit (24 m-long, 3.5 m-wide
& 1 m-deep, Fig. 2a during a period from 1965 to 1990 (Nemoto & Sasaki, 1994).
Overlying a 60 cm-thick fixed base soil layer, a 30 cm-thick surface layer of a single
soil type, sandy loam (Fig. 2b), was prepared repeatedly at different water contents for
respective compaction machines. After spreading the soil with preliminary compaction
by eight passings of a light compaction machine, the respective compaction machines
were passed over the test soil layer. The d values in the 10 cm-thick upper soil layer
were measured four times by the sand-replacement method after the number of passing,
N, became 0, 2, 4, 8 and 16. The d values of the 10 cm-thick lower soil layer was
measured only after N became 16. The data points , , , and in Fig. 2c denote
the average values of d and w from a test series using a 1.6 ton-vibratory pneumatic
tire roller. The values of CBR were measured three times after respective N values. The
broken curves denote the contours of CBR depicted based on an empirical equation
derived from the measurements (i.e., Eq. (3) explained later). The data points and +
denote the results of laboratory compaction tests (1Ec & 4.5Ec) performed at Tokyo
University of Science. Several issues with the conventional soil compaction control in
field earthwork are discussed below referring to these data.
Effects of CEL on compaction curves: In Fig. 2c, the compaction curve by the full-
scale tests moves toward upper left with an increase in N: i.e., with an increase in CEL.
The CEL in the upper part of the compacted soil layer when N = 4 already exceeds the
standard Proctor (1Ec: ASTM-698-78) and the value when N = 16 is much higher than
42 F. Tatsuoka / Compaction Characteristics and Physical Properties of Compacted Soil
Figure 2. Full-scale and laboratory compaction tests on sandy loam (the data by Nemoto & Sasaki, 1994):
a) test pit; b) sandy loam used in the tests; and c) compaction curves and contours of CBR.
the modified Proctor (4.5Ec: ASTM D-1557-78). Ishii et al. (1987) performed a series
of similar full-scale compaction tests using several different compaction machines to-
gether with laboratory compaction tests at CEL = 1Ec and 4.5Ec on clayey sand
(Fig. 3b) and compared the compaction curves (Fig. 3a). In Fig. 3a, the CEL when
compacting a 35 cm-thick soil layer by 16 passings of a flat surface 9.8 ton-vibratory
steel roller (i.e., machine No. 1), which is typical of modern compaction machines for
highway embankment, exceeds 4.5Ec (i.e. compaction curve C). The in-situ wopt in that
case is much lower than wopt by the laboratory compaction tests at 1Ec, while the in-
situ (d)max largely exceeds (d)max by 1Ec. On the other hand, the CEL when 16
passings of pneumatic tire rollers (machine Nos. 4, 5 & 6) is around 1Ec (i.e. compac-
tion cures A and B).
Compaction at w higher than wopt for 1Ec is often recommended in practice to
avoid large collapse deformation and a large decrease in the strength and stiffness upon
wetting. However, w higher than wopt for 1Ec becomes considerably higher than in-
situ wopt if the field CEL is much higher than 1Ec. In that case, compaction would
become inefficient and even over-compaction may take place. In the case presented in
Fig. 3a, the wopt value for N = 16 of machine No. 4 is about 12%, which is similar to
wopt by laboratory compaction tests at 1Ec. It may be seen from Fig. 3c that, in compac-
tion using machine No. 4 at w = 14%, which is higher only by 2% than the in-situ
wopt = about 12%, the d value consistently decreases with an increase in N (i.e. over-
compaction). The CEL at N = 16 when using machine No. 1 or 2 exceeds 4.5Ec and the
Figure 3. Full-scale and laboratory compaction tests (Ishii et al., 1987): a) compaction curves; b) tested soil;
c) over-compaction; and d) d values in the top and bottom layers in a 30 cm-thick compacted soil layer.
in-situ wopt is below 10% (Fig. 3a). Therefore, if w is higher than wopt for 1Ec
(=around 12%) in compaction using machine No. 1 or 2, over-compaction will definite-
ly take place.
Effects of lift: In Fig. 2c, when N = 16, the d values in the lower soil layer (denoted
as 16L) are much lower than those in the upper soil layer. Figure 3d compares the d
values in the 15 cm-thick top and bottom layers in a single lift with an initial thickness
equal to 35 cm. d in the top 15 cm is higher on average by about 0.2 g/cm3 than d
in the bottom 15 cm. Correspondingly, the degree of compaction Dc by 1Ec, (Dc)1Ec, in
the top 15 cm is mostly higher than 90% and the average is around 100%, whereas the
average (Dc)1Ec in the bottom 15 cm is only around 90%. These results shown in Figs 2c
and 3d indicate that a lift thicker than around 30 cm may result in a too large decrease
in d with depth in a single soil layer and should be avoided even when using a heavy
compaction machine.
Figure 4. Distributions of measured values of Dc of sandy and gravelly soils for highway embankments
(Yokota & Nakamura 2009).
Figure 5. Distributions of measured values of Dc of well-graded round gravelly soil (RGSZ in Fig. 11).
Control of the degree of compaction: Even at a site where the soil type and the
values of CEL and w are nominally fixed, the measured values of Dc scatter due to an
inevitable scatter in the actual soil type and the actual values of CEL and w. Therefore,
all the measured Dc values are usually required to be at least a specified lower bound
(often called the control value). For ordinary road embankment in Japan, for example,
the control value of (Dc)1Ec is 90%. In Fig. 2c, the d value for (Dc)1Ec = 90% is about
1.8 g/cm3 and this can be achieved only by soil spreading work if w is around wopt for
1Ec. It is understood that the control value of Dc is not the compaction target Dc value,
but it is expected that the average of Dc becomes much higher than the control value.
For highway embankment in Japan, the control value of Dc is 92% for 1Ec with the
subsoil layers and 4.5Ec with the roadbed. Figure 4 shows the distributions of 19,245
data of Dc from 94 highway embankment construction sites for a period from Novem-
ber 2004 to June 2008. In these projects, 200 kN-level vibratory steel rollers (flat sur-
face) were used and the backfill had Dmax 40 mm and Fc 20%. The average Dc of all
the measurements is larger by about 56% than the control value. Figure 5 shows an-
other similar example. According to these and many other data sets available to the
author, we can expect that the average Dc becomes higher by about 5% than the control
value in ordinary earthworks. It is proposed to determine the target of Dc so that the
strength/stiffness and/or other physical quantities at the target Dc satisfy design re-
quirements and to specify the control Dc value to be smaller by 5% than the target value
postulating that the target Dc is equal to the average Dc value. On the other hand, in the
conventional soil compaction control, the link of the control Dc value to the design of a
given soil structure is usually not clearly defined, so not clearly specifying the target
Figure 6. a) Compaction curves and CBR contours on d Sr plane of the data presented in Fig. 2b; and
b) d/(d)max vs. Sr (Sr)opt relations of the data presented in Fig. 6a.
compacted state. Correspondingly, in many design/construction codes/manuals, it is not

well explained how the control Dc value has been determined or how the control value
is to be determined linked to design requirements.
2.2. Normalized Compaction Curves
The CEL and soil type could be largely different at different earthwork projects, while
their actual values vary even in the same project using nominally the same CEL and
soil type. So, it is desirable if the compaction curves with the values of ( d)max and wopt
at a given moment at a given location can be readily inferred without knowing CEL
and soil type from the one obtained by laboratory compaction tests using a certain CEL
on the soil sample representative of the site. In Fig. 6a, the data presented in Fig. 2c
have been re-plotted changing the abscissa from w to Sr. In this figure, another set of
data obtained for different CELs by laboratory compaction tests using the same soil
type (Murata et al., 2011) has been added. The compaction curves from the laboratory
and full-scale compaction tests at different CELs exhibit nearly the same shape. Be-
sides, the values of Sr where (d)max is obtained are nearly the same (i.e. 81.3% on aver-
age) irrespective of CEL and compaction method. This Sr is defined as the optimum
degree of saturation, (Sr)opt. Besides, the d/(d)max vs. Sr (Sr)opt relation is essentially
independent of CEL and nearly the same between the laboratory and full-scale compac-
tion tests (Fig. 6b).
The same trends as seen from Fig. 6 are observed with other soil types compacted
at different CELs (Figs 7, 8 and 9). Figure 9 shows the compaction characteristics of
the core material for a rockfill dam with a crest height of 131 m completed in 1961
(Miboro dam, Japan). It may be seen from Fig. 9c that the d/(d)max vs. Sr (Sr)opt rela-
tion is independent of CEL in both of the field full-scale compaction tests on the actual
core material and the laboratory compaction tests on the material sieved from the core
material. Besides, the relations from the field and laboratory compaction tests are simi-
lar, as the case shown in Fig. 6b. These facts imply that the d/(d)max vs. Sr (Sr)opt
relation obtained by laboratory compaction tests at a certain CEL using the representa-
tive sample can be applied to the field.
Figure 7. Compaction characteristics of Inagi sand (batch 8): a) grading curve; b) d w relations for differ-
ent CELs; c) d Sr relations; and d) d/(d)max vs Sr (Sr)opt relations.
Figure 8. Sieved core material (SCM) for Miboro dam (Mikuni, 1962): a) grading curve; b) d w and logk
(saturated) w relations at different CELs; c) d Sr relations; and d) d/(d)max vs Sr (Sr)opt relations.
Figure 9. Compaction curves by laboratory and field compaction tests at different CELs of core material for
Miboro dam site (Mikuni, 1962; Asao, 1964): a) core material used in the field compaction test (see Fig. 8a
for the grading curve of the sieved core material used in the laboratory tests); b) d w relations; and
c) d/(d)max vs Sr (Sr)opt relations.
In addition, the effects of soil type on the d/(d)max vs. Sr (Sr)opt relation are very
small with eight soil types from clayey to gravelly soils (Fig. 10) and a wide range of
sandy and gravelly soils (Fig. 11). Besides, Joslin (1959) reported 26 average compac-
tion curves from laboratory compaction tests (1Ec) of 10,000 soil types (Fig. 12b).
These standard compaction curves were prepared to infer the compaction curve for a
measured data set of d and w by plotting these data set on this figure The reported av-
erage values of the particle specific gravity, Gs, and (Sr)opt for each group were plotted
in Fig. 12a and their average values for the whole data were obtained by linear fitting
shown in this figure. The average Gs value was used to analyze the data. Figure 12d
shows the d/(d)max vs. Sr (Sr)opt relations obtained by using the (Sr)opt value of the re-
spective groups. It may be seen that the 26 relations are very similar to each other
without showing a specific trend of scatter according to soil type.
Figure 13 summarizes the d/(d)max vs. Sr (Sr)opt relations shown above. The re-
lation can be represented by the following equation:
d/(d)max = fd [Sr (Sr)opt] (1)
where both d and (d)max are those obtained for the same soil type and the same CEL.
Therefore, d/(d)max may be defined as the true degree of compaction, (Dc)t, for a giv-
en soil type and a given CEL. It is important to note that CEL is not the variable of the
Figure 10. Eight soil types around Miboro dam site (JGS, 1979): a) d w and logk (saturated) w relations
for 1Ec; b) grading curves; and c) d/(d)max vs Sr (Sr)opt relations.
Figure 11. Sandy and gravelly soils tested at TUS (Tatsuoka, 2011): a) grading curves; b) d w relations for
4.5Ec; and d/(d)max vs Sr (Sr)opt relations : c) sandy soils and d) gravelly soils and Inagi sand 8.
Figure 12. 26 groups of 1,000 compaction test data by Ohio State (Joslin, 1959): a) estimation of average
values of Gs and (Sr)opt; b) average d w relations for 1.0Ec of 26 soil groups; c) d vs Sr relations; and
d) d/(d)max vs Sr (Sr)opt relations.
Figure 13. Summary of all the d/(d)max vs Sr (Sr)opt relations of the data presented in this paper.
function fd. It may be seen from Fig. 13 that the variation of fd is not very small, due
likely to a very wide range of soil type from soft clay to well-graded gravelly soil and a
wide variety of data source. Besides, as shown above and also below, the value of
(Sr)opt is rather independent of CEL while it is insensitive to variations in soil type.
Therefore, in a given earthwork project using nominally the same soil type and the
same CEL with a limited variation of actual soil type and CEL, it is reasonable to as-
sume that the function fd is kept essentially constant.
Figure 14. Inference of: a) (Dc)t = d/(d)max Sr relation; and b) d Sr relation in the field.
Figure 15. (Sr)opt (d)max relations from a wide variety of laboratory compaction test data (the data of Kobe
University: Tatsuoka & Shibuya, 2014).
Suppose that the value of (Sr)opt and the function fd have been obtained for a given
earthwork project. Then, as illustrated in Figs 14a and b, without knowing the value of
CEL, by substituting the measured in-situ values of Sr and d into Eq. (1), the true de-
gree of compaction, (Dc)t, and then the value of (d)max at a given moment during a
compaction process is readily obtained. Besides, as seen from Fig. 14a, by controlling
the value of Sr to become (Sr)opt, the value of (Dc)t becomes 100% irrespective of CEL
and soil type: i.e., the (d)max state for current compaction efforts is achieved. This indi-
cates the paramount importance of controlling the Sr value in field soil compaction.
2.3. Optimum Degree of Saturation and Effects of CEL on the Compacted Dry Density
Figure 15 shows the (Sr)opt vs. (d)max relations from many laboratory compaction tests
of a wide variety of soil type available to the author, mostly those presented in this pa-
per. The meanings of the two broken curves for constant air void ratios are explained
later related to Eq. (11). For the whole data, the scatter of (Sr)opt is not small, of the or-
der of 30%. The cause for this scatter of (Sr)opt has not be found. Despite the above, in a
given earthwork project in which the soil type is nominally fixed, we can expect that
Figure 16. Compaction coefficient C vs. [(d)max]1Ec relations from many laboratory compaction tests.
the variation of actual soil type is not significant and the variation of ( d)max is small,
therefore, the variation of (Sr)opt is not significant.
Sometimes it becomes necessary to infer the value of (d)max for a given CEL or the
value of CEL for a given (d)max value based on results from laboratory compaction
tests performed at a certain CEL. Figure 16 shows the relationships between the com-
paction coefficient, C, defined by Eq. (2) and (d)max for 1Ec, [(d)max]1Ec, from labora-
tory compaction tests using different CELs:
(d)max = [(d)max]1Ec [C log(CEL) + 1.0] (2)
where CEL is the compaction energy level defined as the ratio of compaction energy to
that for the Standard Proctor (i.e. 1.0Ec). So, CEL = 4.5 for the Modified Proctor. It
may be seen from Fig. 16 that the value of C for the same [(d)max]1Ec value is noticea-
bly larger with clayey soils than with sandy and gravelly soils and tends to decrease
with an increase in [(d)max]1Ec for the respective soil types. Then, the value of C for a
given value of [(d)max]1Ec and a given soil type can be inferred from the plot presented
in Fig. 16. By substituting the value of C thus obtained together with the value of
[(d)max]1Ec into Eq. (2), an approximate value of (d)max for a given CEL or CEL for
a given value of (d)max can be obtained.
3. Engineering Properties of Compacted Soil
3.1. CBR as a Function of d and Sr at the End of Compaction
Figures 17a and b show the CBR vs. d relations for different constant w values ob-
tained after the number of passing of a compaction machine N became 8 and 16 (the
last passing) from the all full-scale compaction tests explained related to Fig. 2. Similar
figures were obtained for other N values. The solid curves denote an empirical equation
(i.e., Eq. (3) explained later). Associated with an increase in CEL during compaction at
a fixed w value, CBR increases with an increase in d until d becomes a certain value
and then decreases as d further increases. This peculiar trend can be explained as fol-
Figure 17. CBR vs. d relations when N = 8 and 16 from the all full-scale compaction tests (Fig. 2) for:
a) different constant w values; and b) different constant dSr values.
lows. The CBR value of unsaturated compacted soil is a function of at least: 1) d;

2) the matric suction during the measurement of CBR; and 3) the microstructure pro-
duced by compaction, which becomes more stable with an increase in the matric suc-
tion. The matric suction increases with a decrease in Sr. Due to factors 2 and 3, the
CBR value, thus the strength and stiffness, of unsaturated compacted soil decreases
with an increase in Sr for a fixed value of d. On the other hand, the Sr value increases
with an increase in d during compaction at a fixed w. So, with an increase in d associ-
ated with an increase in CEL during compaction at a fixed w, once the negative effect
of Sr on CBR (i.e. factors 2 & 3 at a fixed d value) becomes more dominant than the
positive effect of the increase in d (i.e. factor 1 at a fixed Sr value), the CBR value
decreases despite an increase in d.
With respect to factor 3, Fig. 18 shows the likely mechansim of the effect of Sr at
compacted state on the compacted micro-strucutre. When compacted to point C,
where Sr < (Sr)opt, a coherent micro-structure is formed with fine particles sticking to
coarse particles due to high matric suction resulting from a low Sr. This micro-structure
is relatively stable with fine particles reinforcing the contact points among coarse parti-
cles while forming relatively large voids, resulting in strong and stiff properties with a
high saturated hydraulic conductivity (as discussed later). The terminology flocculat-
ed, which has been used to describe this micro-structure of clay, is not used in this
paper dealing with a wide range of soil type including sandy and gravelly soils. On the
Figure 18. Effects of Sr at compacted state on the micro-structure of compacted soil.
Figure 19. Unsoaked CBR vs. d relations when Sr = 3040% and 7080%.
other hand, when compacted to point D, where Sr > (Sr)opt, a dispersive micro-structure
is formed with fine particles filling voids of the fabrics of coarse particles in a disper-
sive manner due to low matric suction resulting from a high Sr. This micro-structure is
relatively unstable resulting in relatively weak and soft properties because fine particles
do not contribute to the stability of the fabrics of coarse particles, while the voids be-
come relatively small resulting in a low saturated hydraulic conductivity (as discussed
later). The difference in the strength and stiffness between the masses of the same soil
type compacted to points C and D is equivalent to the one between relatively well- and
poorly-graded soils compacted by using the same CEL.
The analysis shown above indicates that the Sr value at the end of compaction is a
much more relevant parameter than w during compaction to describe the CBR value
(i.e., the strength and stiffness) of compacted soil. Then, the data shown in Fig. 17a
have been re-plotted in Fig. 17b changing the parameter from w to Sr. For any fixed Sr,
CBR consistently increases with an increase in d. This means that, if d increases with
an increase in CEL during a compaction process in which the Sr value is maintained
constant, the CBR value consistently increases without exhibiting its peak at an inter-
mediate d value. It may also be seen that the CBR vs. d curves for different Sr values
exhibit nearly the same trend. It may also be seen that CBR largely decreases with an
increase in Sr from around 50%. Figures 19a and b show the CBR vs. d relations when
Sr is in a range of 3040% and 7080% obtained from the data plotted in Fig. 17. Simi-
lar plots are obtained in other ranges of Sr. It may be seen that the CBR vs. d relation
for a fixed Sr value is nearly the same for different N values and different compaction
Figure 20. Function expressing the effects of Sr on unsoaked CBR.
Figure 21. CBR vs. d relations for different w values and for Sr = (Sr)opt and 100% obtained from Eq. (3).
machine types (i.e., for different CELs). Besides, the same d function is relevant to
these relations for different Sr values while the coefficient decreases with an increase in
Sr. Accordingly, the test data were fitted by:
CBR = fCBR (Sr) (d /w b)c (3)
where w is the density of water; b = 0.4 and c = 0.9 (the positive material constants);
and fCBR is a function of Sr. In Fig. 20, the values of fCBR obtained from the relations
shown in Fig. 19 and other similar ones are plotted against Sr. Importantly, the in-situ
unsoaked CBR can be predicted only by substituting measured in-situ values of Sr and
d into Eq. 3 without referring to the in-situ CEL value, which is usually very difficult
to know. In Figs 17a and b, Eq. (3) fits the data very well. In Figs 17b and 19, the scat-
ter of the data relative to Eq. (3) increases with an increase in Sr, in particular when Sr >
(Sr)opt. This trend is due likely to that, when Sr > (Sr)opt, over-compaction tends to take
place more easily and the compacted value becomes more unstable as shown in Fig. 3c.
Figure 21 shows the CBR vs. d relations for several constant w values and those
for Sr = (Sr)opt and 100% obtained from Eq. (3). With an increase in d associated with
an increase in CEL during compaction at a fixed w, the CBR value increases until CBR
becomes the maximum at a certain d value. At this stage, Sr is still lower than (Sr)opt,
where d becomes (d)max for a given value of CEL (i.e. for a given compaction effort).
The (Sr)opt state is reached only after CBR has started decreasing despite an increase in
Figure 22. a) Contours of unsoaked and soaked CBR values on the d w plane and CBR vs. w relations
along two compaction curves, A & B; and b) contours of unsoaked and soaked CBR values on the d Sr at
end of compaction plane and CBR vs. Sr relations along two compaction curves, A & B.
d associated with an increase in CEL. This means that the (Sr)opt state cannot be detect-
ed only by measuring the values of CBR or the strength/stiffness of compacted soil. In
other words, the (Sr)opt state cannot be reached by maximizing the value of CBR or the
strength/stiffness during a given compaction process.
The contours of unsoaked CBR obtained from Eq. (3) are depicted on the d w
plane in Fig. 2c and the top figure of Fig. 22a. On the other hand, the CBR after soak-
ing of the compacted soil is the relevant index of the strength and stiffness of highly or
fully saturated soil usually referred to in the design of soil structures that would be
submerged by flood/impounding or would become wet by heavy/prolonged rainfalls
sometime during their lifetime span (Tatsuoka et al., 2014b). Figure 22a also shows the
contours of soaked CBR obtained by incorporating the effects of soaking on the func-
tion fCBR in Eq. (3). The effects of soaking were evaluated by laboratory CBR tests on
unsoaked and soaked specimens compacted to the different d and w states shown in
Fig. 23a (Tatsuoka et al., 2014a). In these tests, the same soil type as the full-scale
compaction tests was used. Figure 23b shows the unsoaked and soaked CBR values
plotted against d. When compacted to Sr lower than (Sr)opt, CBR drops largely upon
soaking and the drop becomes larger with a decrease in Sr. On the other hand, when
compacted to Sr around, or higher than (Sr)opt, the drop in CBR upon soaking is very
small or negligible. Figure 23c shows only the data when compacted to Sr of around
83.5%. As these unsoaked and soaked CBR values exhibit nearly the same trend, they
were all together fitted by Eq. (4) assuming that the d function of Eq. (3) is also valid
with these data:
CBR = 0.08 (d /w 0.4)9.5 (4)

Figure 23. a) Compacted states of sandy loam to evaluate CBR values before and after soaking; b) all data;
c) data when compacted to Sr of around 83.5%; and d) CBR/CBR when Sr = 83.5% vs. Sr at the end of
compaction relations of unsoaked and soaked specimens (Tatsuoka et al., 2014a).
To extract the effect of Sr at the end of compaction on CBR from the data pre-
sented in Fig. 23b, the ratio of the respective measured CBR values to the one when
Sr = 83.5%, R(Sr), was obtained by dividing the respective measured CBR values by the
value obtained by substituting the d value into Eq. (4). The values of R(Sr) thus ob-
tained have been plotted against Sr at the end of compaction in Fig. 23d. It may be
seen that, with a decrease in Sr at the end of compaction from (Sr)opt, the R(Sr) value
of the unsoaked specimens from the laboratory CBR tests becomes larger than the val-
ue for the CBR measured in the full-scale compaction tests (i.e., a broken curve shown
in Fig. 23d, obtained from the curve shown in Fig. 20). The reason for this difference is
not known. It may also be seen that, with a decrease in Sr from 83.5%, the R(Sr) value
of the soaked specimens, [R(Sr)]soaked, becomes smaller than [R(Sr)]unsoaked to a larger
extent. [R(Sr)]soaked can be represented by:
[R(Sr)]soaked = 2.0 + 0.0151 Sr 3.224 104 Sr2 (5)
where Sr is in %. The fact that [R(Sr)]soaked decreases with an increase in Sr means that,
for the same d value, the soaked CBR value decreases with an increase in Sr at the end
of compaction. This trend is due likely to that the effect of Sr on the micro-structure of
compacted soil, as illustrated in Fig. 18, is not totally erased by soaking. It was as-
sumed that the soaked CBR values under the full-scale compaction test condition can
be obtained based on Eq. (5), as follows:
When Sr 83.5%, [CBR]soaked = {[CBR]unsoaked when Sr = 83.5%} [R(Sr)]soaked (6a)

When Sr 83.5%, [CBR]soaked = [CBR]unsoaked (6b)
where [CBR]unsoaked is the CBR value obtained by Eq. (3). The soaked CBR value ac-
cording to Eq. (6) is also a function of d and Sr at the end of compaction, not includ-
ing CEL as a variable. As w during compaction is not the basic controlling parameter
for CBR (i.e., not for the stress-strain properties of soil), the CBR contours on the d vs.
w plane shown in Figs 2b and 22a are very complicated.
Curve A depicted in Fig. 2b and the top figure of Fig. 22a is the compaction curve
fitted to the data when N = 16 in the full-scale compaction tests referring to compaction
curve B from the laboratory compaction test of 1Ec. The bottom figure of Fig. 22a
shows the relationships between the unsoaked and soaked CBR values and the water
content, w, along compaction curves A and B. The following trends may be seen along
both compaction curves A and B (i.e., irrespective of CEL):
1. When compacted to Sr < (Sr)opt, CBR is relatively large before soaking, where-
as it drops largely upon soaking. Collapse deformation upon soaking is also
significant when Sr < (Sr)opt and d is low (Tatsuoka & Shibuya, 2014).
2. When compacted to Sr = (Sr)opt, CBR drops only slightly upon soaking, while
the soaked CBR is only slightly smaller than its peak value along the respective
compression curves. Collapse deformation upon soaking is insignificant
(Tatsuoka & Shibuya, 2014).
3. When compacted to Sr > (Sr)opt, the drop of CBR upon soaking is negligible.
Collapse deformation upon soaking is also negligible. However, as seen from
Fig. 22a, with an increase in CEL from 1Ec to 4.5Ec, d increases only slightly
while Sr increases toward 100%, resulting in a decrease in CBR (i.e., a de-
crease in the strength and stiffness) from an already low value. Then, the dan-
ger of over-compaction, as shown in Fig. 3c, increases.
It may also be seen from Figs 2c and 22a that, when compacted by CEL higher
than 1Ec at w = the wopt for that CEL, which is lower than wopt for 1Ec, the d val-
ue becomes higher than (d)max for 1Ec and the CBR value (i.e. the strength & stiff-
ness) becomes higher than the one obtained by compaction using CEL = 1Ec at w =
wopt for 1Ec. As a result, the collapse deformation and the decrease in the strength
and stiffness upon saturation are maintained negligible. This means that compaction at
w = field wopt is the relevant target of compaction. However, it is usually very diffi-
cult, if not impossible, to know the actual value of CEL and the actual soil type (i.e.,
the actual field wopt) at a given moment at a given location. On the other hand, the
value of (Sr)opt is essentially independent of limited variations in CEL and soil type at a
given site where CEL and soil type are nominally fixed, while the (Sr)opt value can be
evaluated by laboratory compaction tests at a certain CEL using a sample representa-
tive of the site. So, the (Sr)opt state is the relevant compaction target free from varia-
tions in CEL and soil type.
Figure 22b shows the compaction curves and the contours of CBR on the d Sr
plane and the CBR vs. Sr relations along compaction curves A and B for the data
plotted in Fig. 22a. The effects of Sr at the end of compaction on the compaction
characteristic and the stress-strain behavior of compacted soil can be understood more
easily from Fig. 22b than from Fig. 22a. For example, it is obvious in Fig. 22b that,
irrespective of CEL, [CBR]soaked when Sr = (Sr)opt is close to its peak value along the
respective compaction curves and [CBR]soaked decreases very fast as Sr increases from
(Sr)opt.
3.2. Stress-Strain Behaviour of Compacted Soil as a Function of d and Sr

at Compaction
Tatsuoka (2011) showed that, in drained triaxial compression tests (c = 50 kPa) on

saturated specimens of a wide variety of sandy and gravelly soils compacted to differ-
ent values of Dc and w, the strength and deformation characteristics are controlled by
Dc and Sr at the end of compaction, not by Dc and w during compaction. Figure 24
shows the behaviours of Dokigawa gravel, typical of the above. In Fig. 24d, the angle
of internal friction peak at the same d decreases with an increase in w during compac-
tion, whereas, with an increase in Dc from around 95%, the peak value at the same w
increases only slightly when w = 5.45% or 6.41% and even decreases when w = 7.05%.
Besides, in Fig. 24e, the secant modulus at a deviator stress equal to a half of the com-
pressive strength, E50, at the same Dc decreases with an increase in w, while the E50 val-
ue at a constant w generally decreases with an increase in Dc. Moreover, in Fig. 24f, the
creep axial strain 1 at the same Dc increases with an increase in w, while the 1 val-
ue at a fixed w starts increasing as Dc increases from about 95%. Such a peculiar trend
that the strength and stiffness decreases as the material becomes denser at a constant w
can be explained by that negative effects of the increase in Sr associated with an in-
crease in d (or Dc) at a constant w become larger than the positive effect of the increase
in d. That is, as seen from Figs 24d, e and f that, with an increase in Sr at a fixed value
of d (or Dc), the values of peak and E50 decrease and 1 increases systematically,
while, with an increase in d (or Dc) at a fixed value of Sr, the values of peak and E50
increase and 1 decreases consistently.
Similar trends as Dogigawa gravel can be seen from the drained TC test results of
Inagi sand No. 8 presented in Fig. 25. It is shown in Figs 25e and f that the strength and
stiffness can be represented by empirical relations having Dc and Sr at the end of vari-
ables, not including CEL, in a similar way as Eq. (3) that the strength and stiffness of
granular materials are a function of d (or Dc) and Sr, not d and w. Compaction control
based on this fact is explained later in this paper.
3.3. Saturated Hydraulic Conductivity as a Function of d and Sr at Compaction
Figure 8b shows the relationship between the coefficient of hydraulic conductivity un-
der saturated condition, k, and w during compaction of the sieved core material
(SCM) for Miboro dam compacted at different water contents by five different CELs.
When compacted by 1Ec, the k value becomes the minimum at point A, where w is
noticeably higher than wopt for 1Ec. Compaction at w around the value at point A
(i.e., wetter than the optimum for 1Ec) is often recommended for dam cores and river
dykes. However, if in-situ CEL is 4Ec, for example, compaction at w around the value
at point A results in only a small increase in d and a small decrease in k, whereas a
higher d value and a lower k value is obtained by compaction at w at point B (i.e.,
wopt for 4Ec). Besides, the minimum k value for 4Ec is obtained by compaction by
Figure 24. Well-graded gravelly soil from a riverbed (Dokigawa gravel) (see Figs 11a and b for the grading
curve and compaction properties): a) particles of different sizes; b) compacted states of the specimens (30 cm
D & 60 cm H); and results from drained TC tests (c= 50 kPa); c) stress-strain relations; d) the angle of
internal friction; e) E50; and f) creep axial strain plotted againt Dc for 4.5Ec.
4Ec at w at point C, which is much lower than wopt for 1Ec. However, as the in-situ
CEL is usually unknown, it is difficult to predict the value of w for which the mini-
mum k value is obtained at a given location by referring to such a plot as shown in
Fig. 8b.
Figure 26a shows the logk vs. Sr relation obtained by changing the abscissa from w
to Sr of the data plotted in Fig. 8b. A much more systematic and simpler trend may be
seen in Fig. 26a. A scatter in the k values for a fixed Sr is due mostly to a variation in
the d value among these data. To evaluate the effects of d on k, the k values in differ-
ent ranges of Sr are plotted against d in Fig. 26b. As the slope of the linear relation is
essentially the same for the different ranges of Sr, the data was fitted by Eq. (7):
log k = log fk (Sr) + 5.02(1.872 d /w) (7)

Figure 25. Well-graded sandy soil from an inland area in Tokyo (Inagi sand batch No. 8; see Figs 7a and b
for the grading curve and the compaction curves); a) compacted states of the specimens for drained TC tests
(c = 50 kPa); b) the angle of internal friction; c) E50; d) creep axial strain, plotted againt Dc for 4.5Ec; and
fuctions for e) E50; and f) creep axial strain fitted to the data plotted in Figs c) and d).
where k is in cm/sec; and fk(Sr) is the k value when d is equal to [(d)max]1Ec =

1.872 g/cm3. Figure 26c shows the values of fk(Sr) plotted against Sr at the end of
compaction obtained by substituting the values of k and d of each data point shown in
Fig. 26a into Eq. (7). It may be seen from Fig. 26c that the fk(Sr) Sr relation is rather
unique among the specimens compacted using different CELs and this relation can be
rather accurately representated by a single relation represented by a broken curve that is
independent of CEL. This means that, without knowing the CEL value, the in-situ satu-
rated k value at a given compacted state can be obtained by substituting the measured
values of d and Sr into such an empirical equation as Eq. (7) that has been obtained by
relevant laboratory and/or field tests (as discussed later).
Figure 26. Coefficient of saturated hydraulic conductivity k plotted against Sr at the end of compaction of
sieved core material (SCM) for Miboro dam (Fig. 8b); b) log k vs. d relations for different ranges of Sr; and
c) logk when d = [(d)max]1Ec = 1.872 g/cm3, fk(Sr), vs. Sr relations (Tatsuoka, 2014a).
It may also be seen from Fig. 26c that fk(Sr) is nearly constant as long as Sr < 60%
and starts decreasing as Sr increases from about 60%. The decreasing rate becomes very
high as Sr becomes larger than about 70%. This trend is consistent with the effects of
Sr at the end of compaction on the micro-strucutre of compacted soil illustrateted in
Fig. 18. As suggested by Alonso et al. (2013), it is likely that, when the global Sr value
increases toward a certain limit (i.e., 60% in this case), the clusters of fine particles
absorb the increased volume of pore water maintaining the coherent micro-structure.
As the global Sr value exceeds this limit, the clusters of fine particles stops fully absorb-
ing the increased volume of pore water and, due to a decrease in the matric suction, the
clusters start collapsing dispersing the fine particle into the voids of the fabrics of
coarse partiles. Then, the saturated k value decreases significantly.
The logk w relations of the eight soil types compacted by 1Ec shown in Fig. 10b
are very complicated and the law governing the data cannot be readily found. However,
Eq. (7) also fits very well these data when taking into account the effect of particle size
on fk(Sr) as shown below. The values of fk(Sr) obtained by substituting the values of k
and d of these data into Eq. (7), together with those of SCM plotted in Fig. 26c, have
been plotted against Sr at the end of compaction in Fig. 27a. Despite a very wide
range of soil type (Fig. 10b), the fk(Sr) vs. Sr relations exhibit a similar simple
shape while the fk(Sr) value (i.e., the k value for the same d) at a given Sr becomes
larger systematically as the soil becomes coaser. So, Eqs (8a) and (8b) were fitted to
Figure 27. a) Coefficient of saturated hydraulic conductivity k when d = [(d)max]1Ec = 1.872 g/cm3, fk(Sr),
plotted against Sr at the end of compaction of eight soil types around Miboro dam site and SCM (Fig. 26c);
and b) particle size coefficient P plotted against the particle size (Tatsuoka, 2014a).
these data:
log fk (Sr) = P + log[log fk (Sr)]SCM (8a)
log k = P + log[log fk (Sr)]SCM + 5.02(1.872 d /w) (8b)
where [fk(Sr)]SCM is the function fk(Sr) for SCM (Eq. (7)), which is presented by the bro-
ken curve in Figs 26c and 27a; and P is the parameter representing the effects of parti-
cle size on k with P = 0 for SCM. Equation (8b) is obtained by substituting Eq. (7) into
Eq. (8a). The P value of the respective soil types was obtained by substituting the value
of fk(Sr) when Sr = 84% (i.e., (Sr)opt of SCM) of the respective soil types together with
[fk(Sr)]SCM when Sr = 84% into Eq. (8a). These P values have been plotted against the
ratios of D50 to 0.854 mm (i.e., D50 of SCM) and D30 to 0.356 mm (i.e., D30 of SCM) in
Fig. 27b. Although D10 is usually considered to be the relevant particle size parameter
controlling the value of k, this value was not available with the two finest soil types
among the eight soil types (Fig. 10b). Therefore, D30 was alternatively chosen as the
particle size parameter. D50 was also tried. As seen from Fig. 27b, the P value increases
consistently with an increase in the particle size. Within the range of these data, the
scatter is similar when using D50 and D30. It is to be noted that Eqs (8a) and (8b) are
independent of CEL in the same way as Eq. (7).
Figure 28. Effects of Sr at the end of compaction on the coefficient of saturated hydraulic conductivity k of a
wide variety of soil types (Tatsuoka, 2014b).
To examine whether the functional form of fk(Sr) is similar among a wide range of
soil type, Eq. (7) was re-written to Eq. (9):
log k = log k0 + Fk (Sr) + {[(d)max]1Ec d}/w (9)
where Fk(Sr) is the function representing the effects of Sr on the k value with Fk (Sr =
0) = 0; and k0 is the value of k when Sr = 0 and d = [(d)max]1Ec. The k0 value is different
among different soil types controlled mainly by particle size. The values of Fk(Sr) of six
soil types among the nine soil types of which the data are presented in Fig. 27a were
obtained by substituting the values of k0 and [(d)max]1Ec of the respective soil types into
Eq. (9) and plotted against Sr in Fig. 28. In so doing, it is assumed that = 5.02 for
SCM is applicable to the other five soil types. With soil type Nos. 6, 7 and 8, the k0
values could not be obtained, as seen from Fig. 27a, so their data are not plotted. The
data of soil type A (a mixture of sand, silt and clay; wL = 55%; PI= 27) and type B
(a more sandy soil than type A; wL = 34%; PI = 18) reported by Daniel and Benson
(1990) are also plotted in Fig. 28. Note again that Eq. (9) does not include CEL as a
variable.
It may be seen from Fig. 28 that the Fk(Sr) vs. Sr relations all exhibit a similar trend
in that, with an increase in Sr from 0%, Fk(Sr) is kept nearly zero until Sr becomes about
70%, then it starts decreasing at a similar rate. This result suggests that Eqs (8) and (9)
are applicable to general cases if the parameters can be determined. Several practical
methods to estimate k values at given locations and moments based on different sets of
available data are discussed below.
When only the grading characteristics are available but no data from compaction
and permeability tests are available: The value of k for given values of d and Sr is ob-
tained as follows: 1) the value of [fk(Sr)]SCM is obtained by substituting the value of Sr
into the [fk(Sr)]SCM vs. Sr relation (i.e., the broken curve in Fig. 26c); 2) the value of P is
obtained by substituting the value of D50 or D30 into the relation shown in Fig. 27b;
and 3) these values of [fk(Sr)]SCM and P together with the d value are substituted into
Eq. (8b). As the applicability of the empirical relations used in this estimate to general
cases is not known, the value of k thus estimated should be considered to be only an
approximate value.
When the values of [(d)max]1Ec and (Sr)opt from compaction tests and a data set
from a limited number of permeability tests are available: The k values when Sr >
around 70% are necessary in usual practice. In that case, the k value for given values of
d and Sr (>70%) can be estimated from Eq. (10), which is a simplified form of Eq. (9)
valid only for Sr > 70%:
log k = log kopt.1Ec + [Sr (Sr)opt] + {[(d)max]1Ec d}/w (10)
The linear Fk(Sr) vs. Sr (%) relation depicted in Fig. 28 represents Eq. (10) for SCM.
Here, is the negative slope of this linear relation when Sr is around (Sr)opt. The value
of SCM is equal to 0.128. The value is similar among the different soil types de-
scribed in Fig. 28. is the positive constant representing the effect of d on k. The value
of of SCM is equal to 5.02 with. The value of may be different among different soil
types, as seen from the table inset in Fig. 28. kopt.1Ec is the value of k when Sr = (Sr)opt
and d = [(d)max]1E. When the measured value of kopt.1Ec is not available but a set of
measured values of (k, Sr, d) and the values of and are available, the value of kopt.1Ec
can be obtained by substituting these values into Eq. (10). If three or more sets of
measured values of (k, Sr, d) are available, the values of kopt.1Ec, and can be ob-
tained by substituting these values into Eq. (10). Then, the k value for another set of
measured values of (Sr, d) is obtained by substituting these values into Eq. (10) with-
out referring to CEL.
In summary, the coefficient of saturated hydraulic conductivity, k, of compacted
soil is controlled by d and Sr at the end of compaction, similarly as the CBR values
and the strength and deformation characteristics. Consequently, in field soil compac-
tion, it is necessary and efficient to control the values of d and Sr of compacted soil so
that the strength/stiffness and/or k value required in design are realized, as discussed in
details below. The control of water content w of the backfill material to be compacted
is also necessary but it is sub-ordinate to the control of d and Sr.
4. Unified Soil Compaction Control
4.1. Several Basic Limitations of Conventional Soil Compaction Control
Many of the existing design/construction codes/manuals specify the method to control

the values of Dc and w and/or the value of Sr or the air void ratio, va, defined as:
Figure 29. Compaction curves and contours of k of SCM for Miboro dam.
(11)
where va and Sr are in %. However, the objectives of these specifications and their rela-
tions are usually poorly explained or not at all. In the following, this issue is discussed
and a more rational approach is proposed based on the analysis presented in the preced-
ing sections.
In Fig. 29, several contours of k obtained from Eq. (7) together with compaction
curves for 1Ec and 4Ec (shown in Fig. 8b), both the data of SCM for Miboro dam, are
depicted. First of all, to control the values of Sr and w of compacted soil, such an ac-
ceptable zone as zone a shown in Fig. 29 is specified, where zone a is determined based
on the results of laboratory compaction tests using CEL equal to 1Ec (Standard Proc-
tor). Such soil compaction control based on the compaction characteristics for 1Ec as
above does not encourage the compaction to higher d (or Dc) by higher CEL that is
practically achievable under ordinary modern earthwork conditios. Moreover, in some
codes/manuals in Japan, the area between the compaction curve by 1Ec and the zero air
void curve (where Sr = 100%) is peculiarly elimiated from zone a. This specification
discourages compaction to d values higher than (d)max by 1Ec.
Secondly, in many codes/manuals, the link of such acceptable zones as zone a the
soil properties required in design, such as the soaked CBR values (as shown in Fig. 22)
and/or the values of k (as shown in Fig. 29) is not clearly described or not at all. Then,
when aiming at a sufficiently high dry density to achieve the physical properties of
compacted soil required in design, we often encounter several problems with the com-
paction control based on such an acceptable zone as zone a shown in Fig. 29. In the
area where Sr is larger than about 70%, the contours of k are rather in parallel to con-
stant Sr curves, implying that the control of Sr at the end of compaction is much more
important than that of d (or Dc) to realize required k values. If the compacted state is
located around point b in zone a due to insufficient CEL and a too low water content, it
may be judged based on the countors of k depicted in this figure that the k value could
be too high to be accepted in design. Besides, if the compacted state is located around
point c due to insufficient CEL and a too high water content, the strength/stiffness is
relatively low, although this strength/stiffness is often employed in design. This design
Figure 30. Implications of va control on: a) d w plane; b) CBR d plane; and unsoaked and soaked CBR
during compaction at constant w plotted against: c) d; d) Sr; and e) va for the data of sandy loam presented in
Figs 2 and 17.
practice could be too conservative, because the average of measured values of Dc is

usually higher at least by 5% than the allowable lower limit of Dc (the value along line
b c), (Dc)DL, as typically seen from Figs 4 and 5. It seems that (Dc)DL plus 5% rea-
sonably represents the actual field compacted condition.
Thirdly, in the construction of long embankment for roads and railways or long
river/coastal dykes, the soil type may change often and largely. In that case, soil com-
paction control based on compaction curves by laboratory compaction tests on encoun-
tered different soil types becomes extremely time-consuming. It is in particular the case
in recent speeded-up earthworks using heavier construction machines (e.g., Mokwa and
Fridleifsson, 2005, 2007). Besides, some soil types do not exhibit a clear (d)max state
along a compaction curve. For these reasons, in many construction codes/manuals, it is
specified that the air void ratio, va, (Eq. (11)) becomes lower than a certain value (e.g.,
10%) with or without controlling the value of d (or Dc). As va is not uniquely related to
Sr unless when fully saturated (i.e., when va = 0% & Sr = 100%), va is not well related to
matric suction as Sr. Furthermore, as seen from Fig. 15, the va value when Sr = (Sr)opt
tends to decrease with an increase in (d)max. Therefore, for soil compaction as well as
the strength/stiffness and saturated hydraulic conductivity, va is not a better parameter
than Sr. Yet, for a limited range of d encountered in respective earthwork projects us-
ing nominally the same soil type, the variation pattern of va on the d w plane is simi-
lar as Sr, as typically seen in Fig. 30a. In such a case, the control of va is practically
equivalent to the control of Sr.
There are several confusions and mixing-ups with respect to the implications and
objectives of the va control, so the va control is often not properly practiced. Firstly, in
some codes/manuals, only the allowable upper bound of va is specified without specify-
ing the allowable lower bound. Then, the compaction target is often misunderstood to
be va = 0% (i.e., Sr = 100%). However, both soaked and unsoaked CBR values decease
largely when approaching the state of va = 0% (i.e., Sr = 100%) for a given CEL (see
Fig. 22b). Obviously, the allowable lower bound of va should also be specified indicat-
ing that the compaction target is the state of Sr = (Sr)opt.
Secondly, the va value cannot be decreased unconditionally in the va control. For
example, if va decreases at a constant d by adding water such as along path d c f
in Fig. 30a, the unsoaked CBR decreases drastically as shown in Fig. 30b. Along path
c f, the soaked CBR decreases similarly as the unsoaked CBR. The va control should
be applied to only a compaction path at constant w, such as path a b c at w =
12% (=wopt for 1Ec) shown in Fig. 30a. Like any general path, such as g h, path
a b c comprises two components: (I) path a d, along which the strength/
stiffness increases consistently and significantly due to an increase in d at a constant Sr
(Fig. 30b); and (II) path d c, along which the strength/stiffness decreases significant-
ly due to an increase in Sr at a constant d (Fig. 30b). Component (II) comprises two
sub-components: (II-1) the decrease in the strength/stiffness by the formation of weaker
micro-structure due to an increase in Sr at the end of compaction (as represented by
Eq. (5)); and (II-2) the decrease in the strength/stiffness with a decrease in the matric
suction due to an increase in Sr during shearing (as seen from a large drop in CBR upon
soaking in Fig. 22). In Fig. 30c, the unsoaked and soaked CBR values observed along
path a b c are plotted against d. The unsoaked CBR increases with an increase in
d until point b, which is due to component (I) overwhelming component (II). Subse-
quently, the unsoaked CBR decreases despite an increase in d, which is due to compo-
nent (II) overwhelming component (I). The soaked CBR, which lacks component
(II-2), consistently increases with an increase in d due to component (1) overwhelming
component (II-1). It may be seen from Figs 30d and e that, when the unsoaked CBR
increases along path a b and the soaked CBR increases along path a b c, Sr
increases and va decreases. However, this increase in CBR cannot be attributed to an
increase in Sr nor a decrease in va, but the negative effect of component (II) (i.e., the
decrease in CBR due to an increase in Sr or a decrease in va) is masked by the contribu-
tion of component (I) (i.e., the increase in CBR due to an increase in d). That is, only
for a compaction process at constant w, because Sr increases and va decreases as a result
of an increase in d, we can use Sr or va as the compaction index despite a decrease in
the strength/stiffness with an increase in Sr or a decrease in va. Indeed, it is to be under-
stood that the original objective of the va control is to control the Sr value of compacted
soil to approach the (Sr)opt value irrespective of CEL and soil type.
Thirdly, to realize the soil properties required in design, it is usually not sufficient
to control only the value of va (or Sr), but the control of d (or Dc) is also necessary.
Otherwise, the va control will allow and may encourage the use of soft soil at high wa-
ter content of which the compacted d (or Dc) cannot become high enough even when Sr
becomes (Sr)opt.
4.2. Compaction Target
To achieve the soil properties required in design, it is proposed to define a relevant

compaction target in terms of Sr and d (or Dc), as illustrated in Fig. 31a. It is proposed
to set the target value of Sr to be equal to (Sr)opt irrespective of CEL and soil type for
Figure 31. Implications of (Sr)opt: a) the compaction target on the d w plane; and b) changing rates of
soaked CBR (Fig. 22a) and saturated k (Fig. 26c) with Sr along compaction curves for constant CEL of SCM.
several reasons. Firstly, (d)max is obtained for a given compaction effort. Secondly, as
shown in Fig. 31b, the soaked CBR when Sr = (Sr)opt is only slightly lower than its peak
value along the compaction curve for a given CEL, while CBR decreases very fast if Sr
becomes higher than (Sr)opt. Thirdly, for soil structures that need sufficiently low k val-
ues in addition to sufficiently high strength and stiffness, the k value when Sr = (Sr)opt is
only slightly higher than its minimum along the compaction curve for a given CEL,
while the k value increases very fast if Sr becomes lower than (Sr)opt.
Then, it becomes relevant to determine the target value of d, (d)target, under the
condition that Sr = (Sr)opt in such that the soil properties required in design are achieved.
The above can be done based on results of laboratory compaction tests on a representa-
tive sample of a given compaction project using a certain CEL, together with a) empiri-
cal equations for the strength and stiffness (e.g., Eq. (6) for soaked CBR) or results
from relevant laboratory stress-strain tests; and/or b) empirical equations for saturated
hydraulic conductivity (e.g., Eq. (8) or (10)). For example, Fig. 32 shows the relation-
ships between the angle of internal friction and the Dc values of the granular materials
described in Fig. 11. The specimens were compacted at w equal to the respective wopt
values (mostly those for 4.5Ec). Drained triaxial compression tests ( c = 50 kPa) were
performed on these samples that were either moist as compacted or those that had been
made saturated. Large effects of Dc on the drained strength may be seen. Figures 33 and
34 show effects of Dc on the drained and undrained strengths obtained from triaxial
compression tests on sandy loam used in the full-scale compaction tests (Fig. 2) and
sandy soil retrieved from an irrigation dam. It may be seen that the effects of Dc are
much larger on the undrained strength than on the drained strength. In seismic design
of embankments, the undrained behavior of saturated soil in the saturated zone subject-
ed to cyclic loading becomes one of the key design parameters. Although this issue is
very important, it is beyond the scope of this report.
4.3. Unified Control of Water Content, Dry Density and the Degree of Saturation
Figure 35 shows the unified method of soil compaction control proposed based on the
analysis presented in the preceding sections. The proposed method does not contradict
the conventional approaches of soil compaction control but it unifies them into a single
framework comprising steps 19 shown in Fig. 35 and described below. For the clarity
Figure 32. Relationship between peak by drained triaxial compression tests (c = 50 kPa) and a) Dc (4.5Ec)
and b) Dc (1.0Ec) of sandy and gravelly soils of which the grading and compaction characteristics are de-
scribed in Fig. 11 (Tatsuoka, 2011).
of explanation, the following two differently defined degrees of compaction are herein
used: the true degree of compaction, (Dc)t, defined as the ratio of d to (d)max, both ob-
tained along a single compaction curve for the same soil type and the same CEL; and
the apparent degree of compaction, (Dc)a, defined as the ratio of a given measured d
value of actual soil type at a given location to a fixed (d)max value from laboratory
compaction tests at a certain CEL using the soil sample representative of a given site
(as defined in Fig. 1). Field soil compaction is usually controlled based on (Dc)a, but
(Dc)t is usually unknown.
Step 1: Target T is determined, where Sr = (Sr)opt irrespective of CEL and soil type,
and d = (d)target, which can achieve the soil properties required in design of a given soil
structure, as described in Section 4.2.
Step 2: The target compaction curve (i.e., the target d vs. Sr relation) passing
target point T is obtained by substituting the values of (Sr)opt and (d)max = (d)target in-
to the (Dc)t = d /(d)max vs. Sr (Sr)opt relation (i.e., Eq. (1)) obtained by laborato-
ry compaction tests on the representative sample at a certain CEL, as illustrated in
Fig. 14.
Figure 33. Triaxial compression tests (c = 50 kPa) on sandy loam used in full-scale compaction tests
(Fig. 2; Kasuya, 2014); a) compacted states of specimens; and b) compressive strength against (Dc)1Ec.
Figure 34. TC tests (c = 50 kPa) on Hokota sand retrieved from an irrigation dam (Kasuya et al., 2015);
a) grading; b) compaction curves (1Ec & 4.5Ec) and compacted states of TC specimens; c) drained & un-
drained stress-strain relations of saturated specimens; and d) compressive strength against (Dc)1Ec.
Figure 35. Soil compaction control to achieve the soil properties required in design of a given soil structure.
Step 3: Even in a given earthwork project where the soil type and the values of
CEL and water content, w, are nominally fixed, the actual soil type and the actual val-
ues of CEL and w scatter inevitably. For this reason, the measured values of (Dc)a and
w of compacted soil are allowed to deviate from their target values as long as they re-
main within specified ranges. The allowable lower bound for (Dc)a, [(Dc)a]DL, is speci-
fied in such that [(Dc)a]DL = 95%, for example, to avoid too low d values due to a too
low CEL and/or the use of too low- compactable soil. The meanings and purposes of
this and other compaction bounds are listed in the table inset in Fig. 35 and discussed
below.
Step 4: Along DL, point B where w = the water content at target T is obtained.
Step 5: If it is not required to take into account the saturated hydraulic conductivi-
ty, k, the constant Sr curve that passes point B is specified as the allowable lower bound
for Sr, SL. SL crosses the target compaction curve at point C. If the actual CEL is al-
ways kept the same as the one by which the target compaction curve is obtained while
the actual soil type may deviate from the nominal one, the condition that Sr (Sr)SL re-
sults in (Dc)t (Dc)t at point C (=d at point C/(d)target). The value of (Dc)t at point
C is obtained by substituting (Sr)SL into Eq. (1) and this (Dc)t value is much higher
than the allowable lower bound [(Dc)a]DL = (Dc)a at point B (=d at point B/(d)target).
This issue is discussed in details in the next section. On the other hand, if an allowable
maximum value of k is specified in addition to the required minimum strength/stiffness,
based on the contours of k plotted on the d w plane as shown in Fig. 29, (Sr)SL is de-
termined in such that the k value is always lower than the allowable maximum: i.e., if
the point along DL where the k value is equal to the allowable maximum value is locat-
ed right of point B, SL is determined to pass that point where Sr = (Sr)SL.
Step 6: The allowable lower bound for w, wWL, is represented by a vertical line WL
that passes point C. When the CEL by which the target compaction curve is obtained is
used in compaction at w < wWL, the Sr value may become too low to prevent large col-
lapse deformation and a large strength drop upon wetting.
Step 7: The allowable upper bound for w, WU, where w = wWU = wtarget + x, is spec-
ified to avoid such over-compaction as shown in Fig. 3c. Trial field compaction tests
may be necessary to obtain the reliable value of x. Only the backfill having w between
wWL and wWU is allowed to be compacted.
Step 8: The allowable upper bound for Sr, SU, is specified to pass point D where
WU crosses the target compaction curve. It is assumed that the major parameter con-
trolling the over-compaction phenomenon is Sr.
Figure 36. Roles of DL together with SL & WU in keeping high the actual degree of compaction.
Step 9: The acceptable zone for compacted soil comprises bounds SL, DL, WU
and SU (i.e., the dotted zone in Fig. 35). Bound WL is not used to define this zone,
because compacted soil with w < wWL could satisfy the required soil proprieties if the
compacted state is located inside this acceptable zone. If w < wWL, however, it is very
difficult to reach this acceptable zone even by compaction using a very high CEL.
Therefore, the use of soil with w < wWL in compaction is not allowed.
4.4. Roles of DL Together with SL and WU
The roles of DL together with SL and WU in dealing with inevitable scatters in CEL,
soil type and w in a given earthwork project are herein discussed referring to Fig. 36.
Just for the simplicity of explanation, suppose that the apparent degree of compaction,
(Dc)a, at a given location is defined as the ratio of d measured at that location to
(d)target at target point T obtained from laboratory compaction tests on the representa-
tive sample using CEL equal to (CEL)target. That is, the value of (Dc)a at point T,
[(Dc)a]target, is defined as 100%. (Dc)a at DL, defined as the ratio of d on DL to
(d)target, is denoted as [(Dc)a]DL. Suppose that [(Dc)a]DL = 95% is specified. Then, sup-
pose that the compacted state is located at point B in Fig. 35, where (Dc)a = [(Dc)a]DL =
95%. If the soil type and CEL are the same as those for which the target compaction
curve is obtained, the value of (Dc)t at point B is equal to [(Dc)a]DL. In that case, the
strength/stiffness may be significantly lower and the saturated hydraulic conductivity
may be significantly higher than the respective target values (i.e., the values at point T,
where (Dc)t = (Dc)a = 100%). However, in ordinary modern earthworks, the actual CEL
is well maintained equal to (CEL)target by specifying the lift, the compaction machine
type and the number of machine passing, whereas the actual soil type inevitably scat-
ters. Then, by keeping control bounds DL as well as SL and WU, most of the (Dc)t val-
ues in a given earthwork project using the nominally the same soil type can be main-
tained to be much higher than [(Dc)a]DL, as explained below.
Referring to Fig. 36, the value of (Dc)t when the compacted state is located on DL
(i.e., when (Dc)a = [(Dc)a]DL) is denoted as [(Dc)t]DL. The value of [(Dc)t]DL is not neces-
sarily equal to [(Dc)a]DL, but this value is variable depending on the actual soil type and
actual CEL and could be higher than [(Dc)a]DL. There are two extreme cases:
Case 1: The compacted state is located on DL due to that the actual CEL is lower
than (CEL)target while the actual soil type is the same as the one used to obtain the target
compression curve. In this case, the value of (d)max of this soil is equal to (d)target.
Therefore, the value of [(Dc)t]DL of this soil is equal to [(Dc)a]DL = (d)DL/(d)target
100% (=95% in this case). When the compacted state moves as B F along DL in
Fig. 36a, the [(Dc)t]DL state moves as B1 F1 in Fig. 36b. The role of DL in this case
is to prevent (Dc)t to become lower than [(Dc)a]DL.
Case 2: The compaction state is located on DL due to the use of soil type that is
less compactable than the one used to obtain the target compaction curve while CEL is
maintained to be equal to (CEL)target. This is the usual case in modern earthworks.
Compaction curve passing point B shown in Fig. 36a denotes the one obtained by la-
boratory compaction test using (CEL)target on the actual soil type at a given location. In
this case, the value of (Dc)t at point B is equal to the ratio of d at point B to (d)max
at point P along this compaction curve passing point B x 100%. In an earthwork pro-
ject where the soil type as well as CEL are nominally fixed, we can consider that the
value of (Dc)t is a unique function of Sr (Sr)opt irrespective of soil type and CEL, as
illustrated in Fig. 14a, and the value of (Sr)opt is essentially constant. Besides, the value
of Sr is the same at points B and C. Therefore, the value of (Dc)t at point B is the same
as the value at point C (=d at point C/(d)target 100%). This (Dc)t value is much
higher than [(Dc)a]DL (=95% in this case). When the compacted state moves as E H
along DL in Fig. 36a, the [(Dc)t]DL state moves as E B2 F2 G2 H in
Fig. 36b. When the compacted state approaches point E or H, the value of [(Dc)t]DL ap-
proaches [(Dc)a]DL. However, by keeping bounds SL and WU are kept, the actual com-
pacted state on DL is kept in a range between points B and F. Then, in Fig. 36b, the
value of [(Dc)t]DL is restricted to the values between those at B2 and F2, which are
much higher than [(Dc)a]DL. Then, if the compacted state is located in the acceptable
zone, the value of (Dc)t is maintained much higher than [(Dc)a]DL. When located on the
line of Sr = (Sr)opt, the value of (Dc)t becomes the target value, [(Dc)a]target = 100%. It is
seen from the above that, if the actual CEL is kept equal to (CEL)target, DL together
with SL and WU contribute to maintaining high (Dc)t values. This indicates the para-
mount importance of maintaining the constant CEL by specifying the actual compac-
tion procedure in terms of compaction machine type, lift and the number of machine
passing in addition to specifying such the control bounds and acceptable zone of d, Sr
and w as illustrated in Figs 35 and 36 for the cost-effective construction of soil struc-
tures that can exhibit satisfactory performance.
If the compacted state is located on DL due to the use of actually less compactable
soil type and reduced CEL, the value of [(Dc)t]DL becomes in between those in cases I
and II (i.e., case III in Fig. 36b).
5. Conclusions
The following conclusions can be derived from the analysis of the results from labora-
tory and field compaction tests, CBR tests and laboratory stress-strain and permeability
tests of compacted soil presented in this report:
1) The optimum degree of saturation (Sr)opt, defined as the degree of saturation, Sr,
where the maximum dry density, (d)max, is obtained, and the relationship be-
tween the true degree of compaction (Dc)t = d/(d)max and Sr (Sr)opt obtained by
compaction tests using a certain soil type at a certain compaction energy level
(CEL) are essentially independent of CEL and insensitive to soil type variations.
2) The strength and stiffness before and after soaking and the hydraulic conductivi-
ty after saturation of compacted soil are a function of the dry density, d, and Sr
at the end of compaction, not d and water content, w, in addition to soil type in
terms of particle size, grading etc.
3) Even at a given earthwork project where the soil type and the value of CEL are
nominally fixed, the actual soil type and the actual CEL inevitably vary, thus the
actual values of (d)max and the optimum water content, wopt, at each location are
variable and usually unknown. Despite the above, according to the facts 1) and
2), the value of (Dc)t and the physical properties of compacted soil at a given lo-
cation can be estimated from measured values of d and Sr, without referring to
the actual value of CEL and the actual soil type, if the value of (Sr)opt and the
(Dc)t vs. Sr (Sr)opt relation have been obtained by laboratory compaction tests on
the representative sample using a certain CEL.
4) For efficient compaction control, it is proposed to ensure whether the Sr value of
compacted soil has become close to (Sr)opt while d has become large enough to
achieve the soil properties required in design, together with pre-compaction con-
trol of water content. The proposed method does not contradict the conventional
soil compaction control, but it unifies several approaches in the conventional
practice that are generally not well related to each other and/or apparently incon-
sistent with each other. The proposed method is characterized by the following
features:
a) The compaction target is specified at which the soil properties satisfy the re-
quirements set in design, establishing a link between the soil compaction con-
trol and the design of soil structure.
b) Not only d (or Dc) and w but also Sr are controlled in a unified manner.
c) The results from laboratory compaction tests performed at a specific CEL is
the basis for the proposed compaction control. However, the proposed method
is applicable to any other CEL values. Yet, it is not required to evaluate the
field CEL values, which are variable and very difficult to evaluate.
d) The control bounds of d, Sr and w are specified to deal with their inevitable
scatters in actual earthwork projects. They are related to each other in a con-
sistent manner.
It is believed that the proposed method is more realistic and therefore more practicable
than the conventional method, so it makes the compaction control smoother. In particu-
lar, it encourages compaction to higher dry densities by using higher CELs at water
contents lower than the optimum water content defined for CEL lower than field CEL.
Acknowledgements
The main part of this report was derived from a number of research programs in Ge-
otechnical Engineering Laboratory of Tokyo University of Science (TUS) for the last
ten years. I sincerely thank very much all my previous and present colleagues including
Prof. Kikuchi, Y., Prof. Tsukamoto, Y., Dr. Hirakawa, D., Dr. Kiyota, T., Dr. Dut-
tine, A., Dr. Ezaoui, Dr. Kawabe, S., as well as many TUS students, over-sea graduate
students and visiting researchers. The information and suggestions provided by Prof.
Tateyama, K. (Ritsumeikan University, Japan), Dr. Matsumoto, N. (Japan Dam Engi-
neering Center), Prof. Shibuya, S. (Kobe University, Japan), Prof. Koseki, J. (Universi-
ty of Tokyo, Japan), Prof. Ling, H.-I. (Colombia University, USA), Dr. Yoshida, T.
(Kajima Technical Research Institute, Japan) and Prof. Correia, A.G. (University of
Minho, Portugal) and Prof. Lo Presti, D. (University of Pisa, Italy) are highly appreci-
ated.
References
Alonso, E.E., Pinyol, N.M. and Gens, A. (2013): Compacted soil behaviour: initial state, structure and consti-
tutive modeling, Gotechnique 63, No. 6, 463478.
Asao, I. (1964): About Miboro dam (paper related to No. 31 issue in 8th ICOLD), Large Dam, Journal of
Japanese Commission on Large Dams, No. 27, 3344 (in Japanese).
Blotz, L.R., Benson, C.H. and Boutwell, G.P. (1998): Estimating optimum water content and maximum dry
unit weight for compacted clays, Journal of Geotechnical and Geoenvironmental Engineering,
Vol. 124, No. 9, September, 907912.
Daniel. D.E. and Benson, C.H. (1990): Water content-density criteria for compacted soil liners, Journal of
Geotechnical Engineering, Vol. 116, No. 12, 18111830.
Ishii, T., Mishima, N. and Takashina, K. (1987): Characteristics of soil compaction using vibratory rollers
and pneumatic-tire rollers, Proc. 22th Japan National Conference on Geotechnical Engineering, JGS,
Niigata, 16831684 (in Japanese).
Japanese Geotechnical Society (previously Japanese Society for SMFE) (1979): Interpretation of the results
of ground investigation and soil testing and their application cases, Chapter 8 Permeability tests, 336
337 (in Japanese).
Joslin, J.G. (1959): Ohios typical moisture-density curves, ASTM STP239, Proc. of Symposium on Applica-
tion of Soil Testing in Highway Design and Construction, 111118.
Kasuya, Y. (2014): Shear strength of compacted sandy loam and effects of dry density and the degree of
saturation, Graduation Thesis, Geotechnical Laboratory, Tokyo University of Science (in Japanese).
Kasuya, Y., Otani, Y., Masuda, Y., Hyodo, T., Kikuchi, Y. and Tatsuoka, F. (2015): Effects of drainage
condition on shear properties of sandy soil compacted differently, Proc. 50th Japan National Confer-
ence on Geotechnical Engineering, Sapporo, JGS (in Japanese).
Mikuni, E. (1962): Study on the properties of fill dam core material and its compaction (No. 1), Monthly
Journal of JGS, Tsuchi-to-Kiso, 10(1): 412 (in Japanese).
Mokwa, R.L. and Fridleifsson (2005): Soil air voids method for compaction control, Research report
(FHWA/MT-05-010/8117-23) prepared for The State of Montana, Department of Transportation.
Mokwa, R.L. and Fridleifsson (2007): Earthwork compaction evaluation using soil air voids, Canadian Ge-
otechnical Journal, 44: 151159.
Murata, T., Tateyama, K., Yamamoto, T., Yoshida, T. and Kawano, K. (2011): Experimental method of
predicting the compaction effect in fill work practice, Geotechnical Engineering Journal, JGS, 5(4):
589601 (in Japanese).
Nemoto, T. and Sasaki, T. (1994): Soil compaction characteristics, Celebrating 30th Anniversary Volume,
Construction Method & Machinery Research Institute, 4558 (in Japanese).
Proctor, R.R. (1933): Fundamental principles of soil compaction, Engineering News-Record, New York,
Vol. 111, No. 9, August 31, 5558.
Tatsuoka, F. (2011): Laboratory stress-strain tests for the development of geotechnical theories and practice,
Bishop Lecture, Proc. 5th International Conference on Deformation Characteristics of Geomaterials,
Seoul, Korea, Sept., 350.
Tatsuoka, F. (2014a): Soil compaction No.11, Geotechnical Engineering Note, Monthly Journal, the Founda-
tion Engineering & Equipment (Kisoko), 42(5): 8184 (in Japanese).
Tatsuoka, F. (2014b): Soil compaction No.12, Geotechnical Engineering Note, Monthly Journal, the Founda-
tion Engineering & Equipment (Kisoko), 42(6): 9698 (in Japanese).
Tatsuoka, F. and Shibuya, S. (2014): Several issues in the compaction of residential embankments, Monthly
Journal, the Foundation Engineering & Equipment (Kisoko), 42(9): 1723 (in Japanese).
Tatsuoka, F., Fujishiro, K., Kasuya, Y., Kikuchi, Y. and Tateyama, K. (2014a): Soil compaction No. 13,
Geotechnical Engineering Note, Monthly Journal, the Foundation Engineering & Equipment (Kisoko),
42(7): 97100 (in Japanese).
Tatsuoka, F., Koseki, J. and Kuwano, J. (2014b): Natural disasters mitigation by using construction methods
with geosynthetic s (earthquakes), Keynote Lecture, Proc. 10th Int. Conf. on Geosynthetics (10ICG),
Berlin, September.
Yokota, S. and Nakamura, M. (2009): Design and control of fill compaction for highway embankments,
Monthly Journal Kiso-Ko, July, 37(7): 4750 (in Japanese).
Keynote Lectures
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-79
Rockfill Mechanics. Experimental

Observations and DEM Modelling
E.E. ALONSOa,1 and M. TAPIAS a
a
Department of Geotechnical Engineering and Geosciences
UPC, Barcelona, Spain
Abstract. Particle breakage explains the specific features of rockfill behaviour.

Relevant features concern the dilatant behaviour under increasing confining stress,
the long term deformations (creep) and the response under changes in relative
humidity (RH). RH effects on deformation are well known in practice and result in
marked collapse behaviour of rockfill structures as a consequence of full or partial
saturation. The lecture will first present an ordered set of experimental results. The
following aspects will be introduced: Suction effects on compressibility and
yielding of compacted gravel specimens of different nature and grain size
distribution, delayed creep deformations under isotropic conditions, deviatoric
behaviour, dilatancy and the evolution of grain size distribution. These results sets
the ground for the presentation and discussion of a Distinct Element model
developed in an attempt to create a numerical laboratory capable of predicting
real behaviour. The model includes particle shapes which reproduce, in a
reasonable manner, the irregular shape of real rock fragments. Particle breakage is
introduced as a fundamental deformation mechanism. Unlike other approximations
reported in the literature, particle breakage was approached from the perspective
offered by fracture mechanics. Particle breakage is the consequence of the
propagation of pre-existing cracks. Crack propagation velocity of each individual
particle is approximated by means of analytical solutions. Delayed deformations
are a natural consequence of the model and the effect of RH can also be introduced
in a simple manner. The process of parameter determination will be discussed with
specific reference to large diameter oedometer and triaxial tests.
Keywords. Rockfill, Particle breakage, Distinct Element Method, Suction effect,

Time dependence, size effect, crack propagation
1. Introduction
Field observations on rockfill structures, namely rockfill and zoned earthdams indicate
two distinct features of behaviour: the delayed deformations, which extend for decades
and the sensitivity to water action which results in settlements when rockfill is
saturated [1-4]. The second effect is well noticed in upstream shoulders when they are
subjected to reservoir impounding, a behaviour which cannot be explained in terms of
the concept of effective stress. In addition, records of long term settlements reveal also
the direct effect of rainfall on embankment settlement [5,6,2,3]. These observations
suggest that rockfill behaviour is not a result of the material being dry or saturated.
1
Corresponding Author. UPC-Campus Nord. Department of Geotechnical Engineering and
Geosciences. c/ Jordi Girona 1-3, Building D2. Barcelona 08034, SPAIN.
80 E.E. Alonso and M. Tapias / Rockll Mechanics. Experimental Observations and DEM Modelling
The effect of humidity, understood as a continuous variable, must also control rockfill
behaviour.
Laboratory tests have been performed on reduced scale samples since the mid-
sixties of the 20th century. We refer here to large scale tests even if the grain size
distribution of the tested materials corresponds to gravel. Oedometer [7-20] and triaxial
[9-11, 21, 22, 14-16, 23-28, 20] large scale tests have been reported. They were
performed in connection with earth and rockfill dam studies but, also, in the
investigation of railway ballast. Very few of these laboratory test programs introduced
on a systematic manner the effect of humidity despite its relevance. This variable was
present in the tests performed at UPC.
Laboratory tests provided a wealth of information which is summarized as follows:
Particle breakage is the basic phenomenon which explains the singularity of
rockfill behaviour. The evolution of grain size distribution during tests appears to
be fundamental information which helps to build conceptual models [29,30] but
also becomes relevant data to validate theories of rockfill behaviour, an aspect
developed in more detail in this paper.
The strength envelope is markedly nonlinear. The tangent friction angle decreases
continuously with confining stress. A phenomenon which is largely explained by
the evolution of the grain size distribution.
Despite the granular character of rockfill, dilatancy (volume increase) is very
much counteracted by particle breakage. Volume change, either compression or
expansion, is present during a much extended range of deviatoric deformations.
In fact, current laboratory tests (triaxial in particular) are not suitable to
investigate the evolution of dilatancy at large strains. A practical result is the
difficulty to define critical state concepts in these materials. In addition, flow
rules investigated in a relatively narrow range of imposed shearing strains are not
associated.
In terms of plasticity concepts, observed yield surfaces have a distinct cap. In
other words, irreversible compressive deformations are ever present in
conventional stress paths. This is again a consequence of stress- induced grain
crushing. From a modelling point of view the implication is that purely Coulomb
or frictional type of models are unable to represent rockfill behaviour.
Water effects are well interpreted in terms of the Relative Humidity (RH)
prevailing at the rockfill pores. RH controlled oedometer and triaxial tests
indicate that RH is a necessary state parameter to explain water effects. Two
relevant phenomena depend on RH: the increase in stiffness as RH decreases and
the development of collapse strains as RH increases. Both phenomena are closely
linked, very much in the manner that they are linked in the BBM elastoplastic
model for unsaturated soils [31, 18 and 32] even if the deformation mechanisms
of a (fine grained) unsaturated soil and rockfill are widely different. In fact every
aspect of the constitutive behaviour of rockfill is affected by RH. This is, for
instance, the case of dilatancy (flow rules) or the strength envelope.
The effect of RH is physically explained by the propagation of cracks in loaded
rock particles (which eventually result in particle splitting) and the corrosive
effect of water in controlling the velocity of crack propagation [33-40,4].
Crack propagation within particles, which is a phenomenon intrinsically
dependent on time, explains the long term behaviour of rockfill and, also, water
effects.
E.E. Alonso and M. Tapias / Rockll Mechanics. Experimental Observations and DEM Modelling 81
Rockfill, unlike soils, exhibit a strong scale effect. Large particle lead to large
compressibility and reduced strength if compared with smaller particle size (at
equivalent structure and density). This phenomenon is a direct consequence of the
breakage phenomenon and may be physically approached from basic fracture
mechanics concepts [41-44].
The set of features outlined above suggest that building constitutive models in the sense
of relating stress, strain, their rates and time is not a simple task. The alternative
explored in this paper and related references is to accommodate the most relevant
features of rockfill behaviour, at particle scale, into a particulate model. If the basic
mechanisms of particle breakage are well captured, chances are that the integrated
particulate model would be able to reproduce the macroscopic engineering behaviour in
a satisfactory way. It is also likely that once geometry is defined in a reasonable
accurate way, the remaining physical properties would be limited in number. In
addition, large scale properties observed in granular media (nonlinear stress-strain
behaviour, effect of rotating principal stresses, strain-induced anisotropy, etc.) would
be automatically represented.
The paper starts by presenting a summary of crushing mechanisms of granular
aggregates. Then, the concept of fracture propagation in grains as a convenient basic
process to explain particle breakage is described. It introduces in a natural way the
effect of relative humidity and time. Then, some relevant experimental results
performed in large scale suction controlled oedometer and triaxial tests performed at
the UPC Geotechnical laboratory will be described. The discussion will concentrate on
three singular aspects of rockfill behaviour: scale, suction and time effects.
The second part of the paper concentrates in describing the particulate model
developed which is believed to represent a step forward over pre-existing DEM models
for crushing aggregates. Particle attention will be given to the important question of
parameter identification.
Finally, some model predictions, essentially based on basic particle tests and
validated by a backanalysis of a large scale oedometer tests will be compared with the
triaxial and delayed response of the granular aggregate.
2. Crushing Mechanisms
A natural procedure to investigate the evolution of grain breakage during loading is to

determine the grain size distribution (gsd) at increased levels of loading. This
information is often reported and it provides good validation information for modelling
procedures which introduce explicitly grain breakage. Some contributions [41,45-48]
highlight that the evolution of gsd has a limit an attractor which marks the limit of
this evolution. The particle breakage mechanics developed by Einav [29,30,49] uses
this concept in his formulation.
Experiments also indicate that it is possible to define two classes of breakage: One
results in small new particles and it is explained as a local failure taking place in the
vicinity of concentrated loads among particles. The second class which may be
described as a splitting mechanism transforms a given particle into two large sub
particles having, roughly, half the size of the original particle. We refer to a
comminution mechanism when describing the local crushing.
Examining the evolution of the gsd provides limited information of the relative
importance of comminution and splitting mechanisms during deformation. However,
this is important information specially when developing a granular model capable of

reproducing realistically the actual breakage mechanisms.
In a recent and rather unique experimental investigation [48, 50] it was decided to
explore the nature of breakage mechanisms at a limited cost and effort. The idea was to
load arrangements of parallelepipedic sugar cubes. The initial regular geometry of the
cubes was useful to
Test widely different particle arrangements and initial void ratios.
Examine the broken/damaged particles and to identify the type of breakage.
Figure 1 shows a disordered arrangement of high porosity (e 0.78). An ordered
arrangement and low porosity (e 0.18) aggregate is achieved when particles are set in
a face to face contact.
Figure 1. Disordered arrangement of sugar cubes (e 0.78) inside the oedometer cell (diameter 152mm)
[48,50].
A remarkable finding of the oedometer tests performed was to realize that a common
gsd attractor could be established for the low and high porosity samples (Fig. 2) whose
initial granular structure was widely different.
The relative significance of comminution and splitting breakage is shown in Figure
3 which shows the evolution of the particle divisions for each category as the vertical
stress increases. Remarkably, the results do not depend on the initial sample
arrangement.
Beyond an initial stage, at low stress, dominated by contact breakage,
comminution describes roughly 70% of the particle divisions observed and splitting the
remaining 30%. These proportions may change for different grain geometries and
particle strength. For instance Nakata et al. [51] report a relative weight of 50% and
50% for comminution and splitting breakage in their tests on uniform silica sand.
Tested samples were found to yield plastically, following the standard procedure
of identifying the yield stress in compression plots, at stresses which depend on the
initial pore structure (y = 60 kPa for e0= 0.78 and y = 250 kPa for e0= 0.18). The
interesting finding was that yielding could be associated with a rapid increase in the
rate of particle breakage.
Figure 2. Comparison of gsd curves for disordered (M8, eo=0.78) and ordered (M9, eo=0.18) arrangements of
sugar cubes after an oedometer test [48, 50].
Figure 3. Comparison of the evolution of the percentage of divisions according to breakage mechanisms in
terms of applied vertical load for ordered (e0 =0.18) and disordered (e0 =0.78) arrangements [48, 50].
This is shown in Figure 4 for the two initial granular structures. Splitting and
comminution follow a similar evolution. Breakage indices (Hardin, Marsal) follow a
similar pattern.
Loaded samples exhibited clear creep behaviour. Deformations were linearly
related with time in logarithmic scale. This is a common observation. What is not
usually reported is the evolution of grain size distribution in time. This is shown in
Figure 5 for a sample of disordered particles (e0= 0.78) under a constant vertical stress
at 240 kPa. The figure shows the gsd for the unloaded sample (a set of equal grains)
and its progressive breakage in time for a period of 150 days. The double s-shape
reflects the bimodal initial distribution of grain sizes: the large size is associated with
the half (average) size of the sugar cubes (mean initial size: 27.4x17.6x12.2mm) and
the small size is associated with the saccharose crystals (mean diameter: 0.45mm),
lightly welded, which integrate the large cube. gsd curves in Figure 5 are remarkably
similar to an equivalent set of curves measured at increasing vertical stress and short
testing times (Fig. 6). These results indicate that short and long term behaviour have a
common underlying mechanism. The subcritical crack propagation in grains, analyzed
in [18] provides a unifying breakage mechanism which is consistent with the set of
observations outlined in the previous paragraphs.
Figure 4. Comparison of the evolution of particle divisions according to breakage mechanism in terms of
vertical stress normalized with respect to divisions at maximum stress, for ordered (e0 =0.18) and disordered
(e0 =0.78) arrangements [48,50].
Figure 5. gsd curves at the end of oedometer tests under an applied vertical stress of 240 kPa at different
elapsed time (disordered samples) [50].
Summarizing the experiments performed with the purpose of guiding the numerical
model, it is concluded that comminution and splitting mechanisms contribute, in
variable proportions which seem to remain approximately constant during the loading
path to particle breakage and therefore to the overall (macro) deformation. Breakage,
of any type, seems to be the consequence of the propagation of cracks presumably
starting at some initial defect or microcrack. This explanation is consistent with the
evolution of grain size distribution as a result of stress changes and waiting time.
Figure 6. gsd curves at the end of oedometer tests under different applied vertical stresses (disordered
samples) [48,50].
3. Singular Aspects of Rockfill Behaviour
These singular aspects explain the behaviour of rockfill as observed in large structures
and in the laboratory: Scale effects, time effects (delayed deformations) and
environmental effects, notably water effects. All of them receive a reasonable physical
explanation from fracture mechanics.
Consider first scale effects. Laboratory experiments show that increasing grain size
leads to an increase in compressibility and a reduction in strength. These are important
considerations when translating laboratory-determined parameters to field scale.
Scale effects on grain breakage may be approached from the Weibull [52, 53]
theory of strength of materials. It is also known as a weak-link approach. Weibull
proposed the following equation for the probability of survival of a sample of size d
subjected to a tensile stress
d
nd
m
Psurvival d , exp (1)
d0 0
where d0, 0 are reference states and nd, m are experimental parameters.
For a given probability of survival,
f Cd nd m
(2)
which explains that the stress inducing failure depends on sample size.
Fracture mechanics [54] (see Fig. 7) explains also this dependency. The figure
shows the relationship between the stress at failure and size of a given sample. When
sample size increases, linear elastic and nonlinear fracture mechanics predict that
sample strength is proportional to the inverse of the square root of sample size. In fact,
linear elastic fracture mechanics for a mode I type of failure (failure in extension)
predicts failure when the stress intensity factor of a given sample (say particle i of a
loaded granular aggregate),
Ki i i ai (3)
where i depends on particle geometry, position and size of a potentially active

microcrack, and ai is the half length of the crack, becomes equal to a material property
(toughness Kc). Failure is understood as a rapid propagation of a critical fracture
leading to a particle breakage.
Therefore, at failure
1
i a 1 2 d 1 2 (4)
d
(Note that in statistical terms particle size and the length of a critical crack are directly
correlated).
Point load tests on rock specimens reported by many authors (Fig. 8) support Eq.
(4) (and Eq. 2).
Figure 7. Size effect in the strength under different failure criteria [54].
Consider now time effects. They are conveniently approached from the concept of
subcritical crack propagation. It states that even for a stress intensity factor Ki < Kc,
fracture propagates at a given speed V. Charles law provides the following empirical
relationship between V and K:
n
K
V V0 c (5)
K
where n is a model parameter and V0 is a reference velocity (Fig. 9 provides a set of

experimental results). The figure indicates also a strong dependency between V and the
relative humidity (RH) of the atmosphere in equilibrium with the specimen being tested.
Figure 8. Results of Point Load tests obtained for different authors: Tensile strength vs. particle size.
Figure 9. Charles Model and stress corrosion experimental data for different materials: rocks, quartz and
glass, under different environmental conditions (immersed in liquid water, relative humidity conditions (%),
and vacuum) [4].
The data in Figure 9 suggest that n, V0 or both depend on RH (or, in an equivalent
manner, on total suction s). For instance, n (s).
In a loaded granular arrangement (Fig. 10) individual particles tend to be heavily

loaded in a given direction (which describes internal loading chains). In such a scenario,
a stress intensity factor may be defined for each particle defining an internal loaded
chain. Therefore, combining Eqs. (3) and (5),
n
da K
n
a
V i V0 c V0 i i c i (6)
dt K K

which can be integrated if i is known. This is the approach followed by Oldecop &
Alonso [4] to derive crack length-time relationships for a disk-shaped particle loaded
diametrically. Figure 11 taken from this reference indicates that fractures remain
dormant for a given period of time and become suddenly active at a given critical time
which strongly depends on crack size.
(a)
(b)
Figure 10. (a) Loaded granular arrangement containing cracks inside of rock particles; (b) Detail of rock
particle subjected to a heavy contact load due to loading chains inside the particle arrangement [4].
The third significant aspect (water effects) is essentially explained by the same
framework. Crack propagation depends on RH because water acts as a corrosive agent
controlling the chemical reactions taking place at the tip of the propagating crack. This
effect was already shown in Figure 9. In a synthetic representation (Fig. 12), the crack
propagation velocity is shown to depend on K and RH. Two limiting thresholds define
the dependence of V on K: Ko and Kc. Ko is a stress corrosion limit. If K Ko cracks do
not propagated. At K= Kc the propagation is essentially instantaneous.
These concepts will help to define the main hypothesis of the DEM model
described below. Before this is done, some results of large scale tests on gravel samples
will be presented.
20
18 Initial crack
length
Crack half-length, a [mm]
16 8 mm
14 10 mm
12 mm
12 14 mm
10
8
6
4
2 1 10 100 1 10 100 1000
days years
0
6 8 10
0.01 1 100 10000 1x10 1x10 1x10
Time [minutes]
Figure 11. Crack length propagation in time considering different initial crack length (4, 5, 6 and 7mm):
Crack inside of a disk subjected of a stress of 5MPa ( P=2kN); Disk diameter=40mm; Disk
thickness=10mm; Kc= 1 MPa.m0.5; V0=0.1m/s; n=60 [4].
Crack propagation
velocity (Log scale)
100%RH
or liquid water
>R
H "very dry"
(vacuum)
K0 KKCC
Stress intensity factor, K
Figure 12. Conceptual model for crack propagation velocity in terms stress intensity factor [4].
4. Compressibility and Yielding in Laboratory Experiments
Suction controlled oedometer and triaxial large diameter testing devices have been
described by Oldecop & Alonso [18, 19 and 32] and Chavez & Alonso [28]. Figure 13
shows the significant effect of RH on the compressibility of Pancrudo slate, a rockfill
used in the construction of Lechago dam [55]. Compressibility coefficient depends on

RH. Also yielding is shown to the dependent on RH.
15%
96%
50%
Figure 13. Compresibilty curves for Pancrudo slate gravels tested in a suction controlled oedometer.
Maximum particle diameter of 40mm [32].
Figure 14. Grain size distribution curves after oedometer tests of uniform samples of Garraf hard limestone
gravels having different initial particle sizes. Black lines indicate initial gsd before testing [20].
Collapse phenomenon is explained by an increase in RH. Collapse strains depend on

the change in RH. It is easy to explain the overall aggregate behaviour in Figure 13 if
the behaviour of a single particle is interpreted as shown in Figure 12.
Ortega [20] tested limestone ballast a material significantly harder than the
Pancrudo slate. Some of these experimental observations were used to calibrate the
DEM described in the next section. Despite the hard nature of limestone gravels,
oedometer and triaxial loading at stresses of engineering significance (thinking in
dams) a significant grain breakage was observed. This is shown in Figure 14 for
oedometer tests on four uniform samples of different initial size. Compression was
carried out in an oedometer cell under a high suction (s = 320 MPa). Samples were
loaded to v = 2.7 MPa, they were fully wetted at this confining stress and unloaded.
The figure shows the significant development of sandy particles in all cases.
Figure 15. Deviatoric behaviour for well-graded aggregates of Garraf limestone gravels (Diameter size:
1.440mm). Results of suction controlled triaxial tests for 3=1MPa and different Relative Humidity
conditions: 10%, 50% and 100%. One of the samples at RH=10% was flooded at a 12% (10%-Soaking)
and the other one was unloaded at a 11% and then flooded, before resuming the test (10%-Soaking-UR).
[20].
Time effects were also recorded in all oedometer tests performed. The time dependent
compressibility index (t = d/ d(ln t)) was shown to depend on current stress and
suction.
Ortega [20] tested also well graded and poorly graded ballast samples in a suction
controlled triaxial cell. Figure 15 shows a characteristic result. In this case a well
graded sample (in the range coarse sand medium gravel) was initially confined to a
cell pressure of 1 MPa and then sheared by increasing the vertical stress. Three levels
of humidity are compared in Figure 15: RH=10%, 50% and 100%.
The figure shows that the effect of RH on strength is quite moderate. The sample
shows a ductile behaviour despite the relatively high density (e0=0.54). In fact, none of
the samples tested, irrespective of their initial density and grading, exhibited peak
behaviour.
For a wide range of applied vertical strains (010%) the behaviour is contractant.
Note the relevant effect of RH to control volume change, especially when the dilatant
behaviour manifests. Also shown in the figure is the effect of wetting the sample
loaded under RH=10% at a particular value of deformation, when the limiting strength
was reached. The test was performed at a constant rate of vertical deformation. Wetting
the dry sample results in a rapid reduction of measured deviatoric stress (this is a
consequence of an internal collapse of the granular structure) and the reactivation of a
compressive volumetric strain, even if dilatancy (expansion) was already under way. It
will be shown below that all of these features are reproduced by the particulate model
developed.
Figure 16. Yield locus and Incremental plastic strain vectors. Results of suction controlled triaxial tests on
well-graded aggregates of Garraf limestone gravels (Diameter size: 1.440mm): (a) RH=100%; (b) RH=50%.
[20].
Figure 16 shows the approximate shape of the yield locus determined by means of
stress paths, at decreasing confining stress, performed in a triaxial cell. The figure
compares the yield locus for two RHs: 50% and 100%. Also shown in the figure are
the incremental plastic strain vectors determined at particular values of the total
deviatoric strain applied. The figure shows the irreversible compressive strains on the
wet side of the yield locus and the dilatancy observed on the dry side. The effect of
RH is also clear: increasing RH reinforces the compressive strains. Critical state
conditions could not be found in the tests performed. Dilatancy was increasing at a
significant rate even if deviatoric strains reached 20%. These observations stress also
the difficulty of building continuum constitutive models.
5. A Particle Model
Distinct Element models and their predicting capabilities for granular soils have often
been reported in the Geotechnical literature ever since the pioneering contribution of
Cundall & Strack [56] was published. However, even if the computer power has
increased dramatically, an accurate geometrical representation of a sand specimen is
not yet a realistic possibility. This is especially evident when soil-structure
interactions are examined: For instance, the analysis of cone penetration in sands.
The alternative is to modify the scale of the grains. However, rockfill (including
gravels) offer the possibility of a more accurate representation of prototype samples.
This is a positive and necessary feature because of scale effects mentioned before. Even
large scale structures (a rockfill embankment for instance) may perhaps be realistically
simulated in a not so distant feature. The main purpose of the research conducted,
however, is to improve our current understanding of coarse soil behaviour under
mechanical and environmental actions. One may think also on the possibility of
generating a numerical constitutive model which may be plugged into a general
purpose continuum model for boundary value problems.
In view of the previous representation of rockfill features, a realistic simulation of
the granular aggregate, via DEM must:
Reproduce particle shape to a certain degree of accuracy
Introduce particle breakage incorporating basic concepts of fracture mechanics. In
particular, subcritical crack propagation offers the possibility of accounting for, in
an automatic manner, the effects of time, relative humidity and particle size.
Base the parameter determination on independent tests of grain-to-grain contacts
and, if necessary, on the back-analysis of simple tests. This is a desirable feature
to increase the confidence and consistency of the model.
5.1. A DEM Model for Gravel Particles
Consider in Figure 17a a photograph of the Garraf limestone gravel used in Ortega [20]
oedometer and triaxial experiments. Particles exhibit sharp edges because of its origin
(crushing at the quarry). Particle geometries are complex but pyramidal shapes (Fig.
17b) tend to dominate.
The approach followed to reproduce particle geometry is indicated in Figure 17c.
A reference macroparticle is defined by aggregating 14 spheres in a pyramidal shape.
Successive breakage transforms the initial geometry in a variety of shapes (truncated
pyramids, parallellepipedic shapes and, finally, a single sphere.
The simulation of particle breakage requires three successive steps: a) Calculation
of stresses within particles; b) Definition of a failure criterion for macroparticles and c)
Division of macroparticles.
5.2. Stresses in Macroparticles and Division Criteria
The development of strong force chains inside loaded granular media suggests that
particles prone to break are diametrically loaded as indicated in Figure 18. It will be
further assumed that, with the purpose of determining the tensile stresses developing
inside macroparticles, a spherical equivalent macroparticle, diametrically loaded will
be adopted. Tapias et al. [57] performed analyses of irregular polyhedron shapes and
alternative loading conditions simulating the effect of additional concentrated loads

with the purpose of examining the limitations of the model hypotheses. The conclusion
was that the dominant concentrated force acting on opposite points dominates the
tensile stress field inside particles.
(a)
(b) (c)
Figure 17. (a) Garraf hard limestone gravels; (b) Detail of a hard limestone rock particle; (c) DEM
macroparticle model [57].
Figure 18. Stresses in a sphere loaded under concentrated, diametrically opposed loads [57].
The elastic solution for the problem represented in Figure 18, when the concentrated
load acts on a (small) circular area on the sphere surface, defined by an angle 2 0 (Fig.
19) was given by Russell & Muir Wood [58]. The circumferential tensile stress is given
by

1 (1 ) 1
0.5
1.5
*p (7)
2 w
2
w
2

1 2* 1
z
z
where
F
p (8)
R sin 2 0
2
where R is the sphere radius.

This equation is represented in Figure 20. Note how sensitive is 0 to the
concentration of the load F. The smaller the angle 0, the higher the tensile stresses.
Maximum tensile stresses develop at distances varying between 0.05R and 0.2R from
the point of application of the concentrated load. In the following calculations a contact
angle 20 = /36 was adopted.
Figure 19. Contact load applied to a small circular area [57, 58].
Figure 20. Dimensionless horizontal stress ( /(F/R2) ) in particle axis, for =0.25 and different 0 values:
/200; /50 and /36 [57].
Cracks are randomly distributed in particles. Their length is chosen to be uniformly
distributed between 0 and R. This assumption introduces a direct relationship between
particle size and crack size.
The second analytical solution embodied in the model is the velocity of crack
propagation. The solution given by Oldecop & Alonso [4] was selected. It describes the
propagation of a crack of half length a under a mode I failure. The particle shape in this
case is a thick cylinder.
The two analytical equations are obviously approximated but this is part of the
trade-off inherent in DEM. The advantages of introducing two analytical solutions to
describe particle behaviour are significant because they solve in an approximate way
the problem of calculating stresses and fracture propagation inside particles. Solving
these problems numerically would lead to an intractable problem.
The protocol to decide a particle breakage is synthesized in Figure 21. Calculation
progresses in time. For each new time increment all particles in the arrangement are
checked and the stress intensity factor is calculated using solution (7)(8) and the
current value of the crack length. Whenever KKc, the particle is divided into two new
ones. If K<Kc and the crack size exceeds R, the particle is also divided into two
particles. Otherwise the particle is not broken at the current time step.
Two types of breakage were described in section 2: comminution (or local)
breakage and splitting (or equal volume) breakage. Their relative proportions which are
approximately maintained during a loading process have been quantified in some
experiments. In this work they were assigned the weights of 60% (comminution) and
40% (splitting). These proportions are close to the values reported by Nakata et al. [51].
The breakage protocol is shown in Figure 22. The limiting minimum size is one
spherical particle. A further division criterion, following the same protocol, was
applied to single particle spheres coming from the first set of divisions. In this way the
isolated sphere becomes a macroparticle which is assumed to be made up by 13 new
smaller spheres. This second division stage was found necessary to simulate properly
the development of the finer fractions of tested specimens. The ratio of sizes of the
biggest macroparticle and smallest sphere is about 7. For instance, if the mean diameter
of macroparticle is 2.8cm, the diameter for the spherical particles integrating the initial
macroparticles and the finest spherical subparticles would be 1.185 cm and 0.395 cm
respectively.
Figure 21. Particle breakage Protocol.
Figure 22. Comminution and Splitting (equal volume). Breakage Protocol.

Figure 23 shows the macroparticle model for the oedometer and triaxial samples tested.
The initial size corresponds to the uniform grain size distribution in the range 2530
mm tested by Ortega [20].
6. Model Performance and Parameter Determination
Once the geometry of the particles and the initial arrangement (overall void ratio,
essentially) are fixed, the model is characterized by three parameters: Friction
coefficient, , at particle contacts; grain toughness, Kc and a contact stiffness kn
(normal and tangential contact stiffness were made equal for simplicity).
Contact friction is amenable to direct determination by simple laboratory tests.
However, it is highly dependent on the roughness of surfaces in contact. Bruce et al.
[59] related roughness with a particular friction angle: Mineral friction required
polishing the surfaces to a #1000 sandpaper roughness whereas a basic friction angle
could be determined by polishing at a #80 sandpaper roughness.
(b)
(a)
Figure 23. Sample DEM model of macroparticles for the following tests: (a) Oedometer test: Cylindrical
sample (0.25x0.25m) made of 471 macroparticles; (b) Triaxial test: Cylindrical sample (0.25x0.50m) made
of 971 macroparticles [57].
The friction of limestone rock of the ballast tested by Ortega [20] was investigated
by tilt table tests performed on rock to rock surfaces polished as suggested by Bruce et
al. [59] (Fig. 24).
A basic friction angle of 17o (B 0.3; B = 17o) was determined. Mineral friction
dropped to M = 0.23. The roughness left by saw-cutting the rock samples led to a
rough friction M = 0.5 (R = 29o). A contact friction coefficient B = 0.3 was adopted
in the simulations performed.
Rock toughness values (for mode I-Tensile) have been reported by several authors
[60-65]: 0.41 MPa.m0.5 for sedimentary rocks (shale, sandstone, limestone) and 1.54
MPa.m0.5 for intrusive rocks (granite, diorite).
Figure 24. Limestone friction angle determined in tilt table tests on polished planar joints [57].
Sedimentary rocks were found to have Kc values in the range 0.4 1 MPa.m0.5. Contact
stiffness may be approximated by matching large scale tests (say, an oedometric
compression).
Some DEM simulations of triaxial response were performed for the following set
of reference particle-related parameters: B = 0.3; Kc = 1 MPa.m0.5; kn = 4 MN/m. The
idea was to check the sensitivity of the DEM model against parameter variations within
a reasonable range. In all simulations parameters were kept equal to the basic set except
for the property being analyzed.
Consider first the effect of particle friction (Fig. 25). It has a very significant effect,
controlling overall strength, initial stiffness and dilatancy. The numerical test was
performed at a confining stress of 1.0 MPa and an initial porosity of 0.51. A brittle
type of failure was never observed in laboratory tests and therefore Figure 25a indicates
that a basic friction angle close to 17o is more likely than a high contact friction. A
similar conclusion is reached when observing the calculated dilatancy (Fig. 25b).
Results are not so sensitive to toughness, especially because it is not so variable
when considering a given type of rock. This is shown in Figure 26. A major reduction
in Kc (compared with the reference value Kc = 1 MPa.m0.5) is required to affect the
stress-strain behaviour. The high friction adopted in this set of results explains the peak
response observed in the simulated tests.
The effect of contact stiffness (kn = ks) is shown in Figure 27: It has a definite
effect on the initial stiffness of the sample but a more limited effect on strength.
An oedometer test on the real sample provides two additional sources of
information: the sample compressibility and the change in grain size distribution. A
comparison of experimental and model compression curves is given in Figure 28 for
the set of reference values ( = 0.3; Kc = 1 MPa.m0.5; kn = 4 MN/m). The agreement is
good.
A further check is provided by comparing the grain size distributions. The
oedometer test simulated was performed at very dry conditions (RH=10%). The sample
was loaded in steps up to v = 2.8 MPa, fully wetted and unloaded. The grain size
distribution is given in Figure 29 for the final stage once the test was dismantled.
The agreement between the model and the test is also good and provides an
additional support to parameters determined or approximated by other procedures. The
numerical sample evolved from an initial number of 471 equal size macroparticles to
12146 particles at the end of the test. However, a significant number of particles (183)
did not break during the entire loading and wetting process.
This is a consequence of the protecting effect of smaller particles, which helps to
reduce the intensity of concentrated loads experiencing the large original particles. This
phenomenon, already mentioned in previous contributions [41,51,67] could be
identified by examining the progressive change in the internal resisting structure of the
granular aggregate.
A final point concerns the effect of interlocking among particles, a question
associated with the particle shape. Runs were made by changing the shape of the initial
set of particles defining the triaxial test. The initial particles were defined by
aggregations of 14, 13, 5, 4 and 1 particle. The transition from 14 to 13 implies loosing
the apex of the pyramid, which becomes a truncated pyramid. The 5 and 4 aggregations
of spheres reproduce plate-like shapes. The fifth case corresponds to a set of equal size
spheres. In all cases the initial porosity was similar (n 0.5). The differences in
calculated deviatoric stress-strain and dilatancy response are very significant (Fig. 30).
This example highlights the relevance of a proper representation of particle geometry.
Figure 25. Contact friction angle effect in triaxial DEM tests: (a) Deviatoric behaviour ; (b) Volumetric
behaviour. Comparison among three friction coefficients (0.93 (43o), 0.50 (27o), 0.30 (17o)). Sample of 1000
macroparticles using clumps of 14 microparticles. Confining stress: 1.0MPa. Initial porosity: 51% [66].
Figure 26. Toughness effect in triaxial DEM tests: Deviatoric behaviour. Comparison among four different
toughness values (106, 105, 104, 102 Pa.m0.5). Sample of 1000 macroparticles using clumps of 14
microparticles. Confining stress: 1.0MPa. Initial porosity: 51%. =43o [66].
Figure 27. Contact stiffness effect in triaxial DEM tests: Deviatoric behaviour. Comparison among three
different values of normal stiffness (2e6, 2e7, 2e8 N/m). Sample of 1000 macroparticles using clumps of 14
microparticles. Confining stress: 1.0MPa. Initial porosity: 51% [66].
Figure 28. Measured and calculated oedometer compressibility of the limestone gravel sample. DEM
parameters: =0.3; Kc=1 MPa*m0.5; kn=4 MN/m [57].
Figure 29. Evolution of particle size distribution in oedometer test. Experimental results reported by Ortega
[20] are compared with DEM simulations [57].
7. Predicting Triaxial Tests
The set of model parameters was determined by combining a direct measurement of

particle friction, an estimation of rock toughness mainly based of published data as
well as a good matching of grain size distribution after oedometric loading and an
estimation of contact stiffness by reproducing a compression curve of an oedometer
test.
If the DEM model developed is able to reproduce to a reasonable accuracy the
oedometer test, including the evolution of breakage, it could be expected that it should
be able to reproduce also any other type of test because the model only understands
what happens at particle level and the application of any other stress path does not
introduce, a priori, a significantly different fundamental phenomena concerning particle
contacts or particle breakage: If this is a correct way of thinking, triaxial tests
performed on the same material should be reproduced by the DEM model without any
change in model parameters.
Figure 30. Shape effects. Results of numerical simulation of triaxial test using macroparticles of 1, 4, 5, 13
and 14 microparticles. (a) Deviatoric behaviour ; (b) Volumetric behaviour [66].
The response to this argument is given in Figure 31 for a triaxial test performed on
gravel having the initial uniform grain size distribution (sizes in the range 25 30 mm)
also used in the back-analyzed oedometer test. DEM model parameters were: = 0.3;
Kc = 1 MPa.m0.5; kn = 4 MN/m. The sample was loaded to a limiting state under a
RH=10% and it was flooded when the vertical strain reached 10%. This is shown by
the transient drop of deviatoric stress which was later recovered as deformation
continued.
The volumetric deformation of the sample is well captured by the model. The
predicted deviatoric stresses underestimate by 30% the actual measurements.
Figure 31. Comparison of triaxial test results reported by Ortega [20] and model predictions. DEM
parameters: =0.3; Kc=1 MPa*m0.5; kn=4 MN/m [57].
Figure 32. Measured and calculated grain size distributions of triaxial test [57].
The grain size distribution after dismantling the wetted triaxial sample is compared
with the model predictions in Figure 32. The agreement is very good.
8. Scale Effects
Consider in Table 1 some examples of large diameter testing cells, their overall
dimensions and the largest acceptable particle size. The largest particle sizes
(0.150.18m) could be tested in huge triaxial cells developed in the 60s in connection
with rockfill dams being built in the U.S. and Mxico. In recently built equipment the
largest particle seldom exceeds 10 cm.
In contrast, the range of particle sizes in rockfill embankments and dams is 0.20-
1.50m. A procedure to overcome the change in scale, often reported when justifying
testing of gravel size in an attempt to determine parameters for larger particle sizes is to
test a scaled grain size distribution (Fig. 33). However, the intrinsic scale effects
associated with particle breakage make this hypothesis wrong.
Figure 33. Grain size distributions of prototype and scaled sample.

If the DEM model described is correct it offers the possibility of investigating scale
effects at a limited cost. If the objective is the variation of compressibility with grain
size, one possibility is to run a series of numerical oedometric tests changing only the
particle size. The performed exercise is defined in Table 2. Four samples of
increasing size were tested. The soil had initially a uniform particle size ranging from
2.8mm to 560mm (there was a scaling factor of 20 when comparing maximum and
minimum sizes). Also indicated in the Table 2 is the necessary size of a testing cell to
accommodate the four materials. They are, except for the smaller particles outside a
reasonable cell size. The initial porosity was essentially constant. The following
common set of material parameters was used. kn = ks = 4 MN/m: = 0.3; Kc = 5
MPa.m0.5. Tests were performed at a constant RH=10%. Calculated compression
curves taken to a maximum vertical stress of 2.8 MPa are given in Figure 34.
The calculated average compressibility indices for vertical stresses in excess of 1
MPa are given in Figure 35. Also indicated in the figure are the experimental values
reported by Ortega [20] for three specimen sizes (1.52cm; 2.53cm; 34cm). The
experimental data is concentrated around a particle size D = 3cm and no test data on
significantly lower or higher sizes is available. Nevertheless, the numerical value for D
= 2.8cm is reasonably close to measurements. The plot shows also the significant
increase in compressibility with particle size.
Table 1. Examples of large diameter testing cell [57].
Sample Dimension Maximum

Test Reference grain size
Diameter (m) Height (m) (m)
UNAM CFE (Mexico) [9,10] 1.13 2.5 0.18
ISMES (Italy) [11] 0.35 0.70 -
Univ. California Berkeley (USA) 0.07; 0.31; 0.15; 0.68;
0.152
[21,22] 0.91 2.29
Monash U. (Australia) [14] 0.38; 0.57 - -
Triaxial LNEC (Portugal) [15,23] 0.30 0.76 0.100.05

AIT (Bangkok) [24] 0.30 0.60 0.0381
[25] 0.30 0.60 0.053
Japan [26,27] 0.30 0.60 0.0635
UPC (Spain) [28] 0.25 0.50 0.04
UPC (Spain) [20] 0.25 0.50 0.04
Norwegian Geotechnical Institute
0.50 0.25 0.064
(Norway) [7]
Georgia Inst. of Technoloy (USA) [8] 0.19 0.10 0.038
UNAM CFE (Mexico) [9,10] 1.13 1.13 0.20
0.10; 0.50; 0.20; 1.00;
ISMES (Italy) [11] 0.26
1.30 2.00
Build. Research Establishment (UK)
0.60; 1.00 0.50 0.125
[12]
Oedometer
Build. Research Establishment (UK)
0.45 0.225 0.076
[13]
Monash U. (Australia) [14] 0.635 0.61 0.09
LNEC (Portugal) [15,16] 0.50 0.50 0.10
Brasil [17] 1.00 1.00 0.20
UPC (Spain) [18] 0.30 0.20 0.04
UPC (Spain) [20] 0.30 0.20 0.04
Table 2. Dimensions of samples and macroparticles in DEM simulation of oedometer tests: Scale effect [43].
Sample Size Equivalent

Macroparticle
Sample Relative scale Height Diameter Diameter Initial Porosity
(mm) (mm) (mm)
S25 0.1 25 25 2.8 0.5314

S250 1 250 250 28 0.5314
S1250 5 1250 1250 140 0.5337
S5000 20 5000 5000 560 0.5314
Figure 34. Calculated compressibility of samples of increasing initial equivalent particle size D: Scale effect
[43].
The capability of the model to simulate scale effects was further checked by
performing numerical triaxial tests on samples of uniform particle sizes (3040mm;
1520mm) which could be compared with actual laboratory tests performed by Ortega
[20]. The gravel in this case was Garraf limestone and the model parameters = 0.3; Kc
= 1 MPa.m0.5; kn = ks = 4 MN/m were already discussed in Section 6.
The experiments were performed under dry conditions (RH=10%). A full wetting
was applied at a large vertical strain (12%) when shear strength was already very close.
The collapse of the real samples can be seen in Figure 36.
Figure 35. Calculated compressibility indices for different initial particle diameter at v = 2.8 MPa: Scale
effect. Comparison with results of laboratory tests of Ortega [20, 43].
Figure 36. Experimental and simulated triaxial tests on uniform limestone gravels of different initial grain
size. Experimental data reported by Ortega [20]. DEM parameters: =0.3; Kc=1 MPa*m0.5; kn=4 MN/m [57].
The DEM simulation was run under a RH = 10%. The calculated deviatoric stress-
strain curves compare well with actual measurements. Volumetric compression
(enhanced immediately after wetting) was more pronounced in the real experiments but
nevertheless the overall performance of the model is quite satisfactory.
Frossard et al. [42] describe an analytical procedure to estimate size effects for
uniform granular materials of different sizes. Their approach is limited to strength
envelopes or linear compressibility parameters. The advantage of the DEM approach is
that it may refer to any aspect of behaviour and it may handle arbitrary grain size
distributions.
9. Time and Relative Humidity Effects
Model formulation, outlined in Section 5, accommodates in a natural way time and

suction effects. Reported experimental observations may then be compared with model
predictions. Suction effects are included in Charles law (Eq. 5) if exponent n is made
dependent on suction. Data in Figure 9 was interpreted with the objective of finding a
relationship n(s), which is shown in Figure 37. This relationship is preliminary
accepted as an approximation for a variety of materials. Oldecop & Alonso [4]
suggested a constant value for V0 (V0=0.1m/s). Time effects require only a specification
of the parameter entering in the stress intensity factor (Eq. 3). The expression for a
disk of a brittle material having a central defect of length 2 a in the direction of two
opposite loads [4] was used. It does not require any additional material parameter.
Figure 37. Relation between exponent n of Charles Law (1958) and Relative Humidity.
Summarizing, including time and suction effects may be done without any specific
additional basic test or back calculation of large scale tests on aggregates.
Figure 38 shows a comparison of the long term oedometric behaviour of Garraf
limestone samples (initial particle size: 28mm) subjected to a set of incremental vertical
stresses. New loading steps were applied when the previous one was maintained for
around 17 h (103 min). The sample was dry (RH=10%) during the entire step-loading
test. However, it was fully wetted (RH=100%) when the vertical stress reached 2.8
MPa. The sample experienced a collapse, which developed in time, in the manner
indicated in Figure 38b. Computer results are shown in Figure 38a. The simulation
reproduces well the strain development in time and the effect of suction changes.
The effect or wetting a dry sample at some stage of a strain-controlled triaxial path
was also simulated. One example, taken from the experiments of Ortega [20] for Garraf
limestone is given in Figure 39. The model reacts in a correct way: the deviatoric stress
reduces immediately after wetting but it recovers progressively the strength expected
for the new imposed suction (s = 0 in this case) as vertical deformation progresses.
The major discrepancy was experienced when comparing the dilatant behaviour.
The real sample compressed substantially under deviatoric loading a result which was
not reproduced to such an extent by the model. In fact, at the time of wetting the model
already was predicting some expansive dilatancy rate which was temporarily lost due
to the full wetting (this is the actual behaviour observed in other samples). Eventually
the model predicted expansive rates for large vertical deformations. The experiment
could not be taken to such a high sample straining.
(a)
(b)
Figure 38. Long term oedometric behaviour of Garraf limestone samples (initial particle size: 28mm)
(RH=10%): (a) DEM simulations; (b) Experimental test [20].
Figure 39. Wetting effect on deviatoric behaviour of triaxial tests for limestone gravels: Comparison
between experimental [20] and DEM tests. (3 = 1MPa).
10. Conclusions
Rockfill behaviour is dominated by particle breakage which is a consequence of the

high concentrated loads acting at particle contacts. This phenomenon is significant at
stresses prevailing in common engineering structures (embankments, dams).
Rock particles are brittle materials and breakage is better approached by fracture
mechanics concepts. If a stress intensity factor, K, could be estimated at particle level,
failure can be predicted when K reaches the material toughness, Kc. In addition, time
and suction effects could be introduced in a natural manner by estimating the rate of
crack propagation in particles whenever K<Kc. The consequences of this approach have
been discussed and the results of some large scale tests have been presented to support
the underlying basic mechanisms of rockfill behaviour.
The paper describes also a promising particulate model which includes particle
breakage as a consequence of propagation of internal cracks or defects. The main idea
was to solve the problem at particle level by means of two analytical solutions: the first
one provides the tensile stress distribution inside a spherical particle, the second one
provides an expression for the propagation of a critical crack randomly introduced in
particles.
In this way the numerical calculations are affordable despite the heavy
computational effort necessary to reproduce experiments on gravels performed in large
diameter (25cm in the cases presented) cells.
The DEM model developed requires the definition of particle shapes which could
compare with real geometries and a protocol for particle breakage. The relevance of
particle shape, because of the notorious interlocking effect of particle geometry, has
been highlighted. The model requires a few material parameters: friction and stiffness
at particle contacts and the material toughness. They are amenable to direct
determination by specific tests and they may also be back-interpreted from real tests
which provide not only a macro stress-strain behaviour but, also, the evolution of
grain size distribution, which is also calculated in the model.
At essentially no extra cost in terms of material parameters the model provides
the effect of suction, the long term creeping behaviour and the identification of scale
effects, which is relevant in practice.
The paper includes a comparison of model predictions and a few laboratory tests
performed at the UPC Geotechnical Laboratory on large scale suction controlled
oedometer and triaxial tests. The comparison is, in general, quite satisfactory. This
validation exercise confirms the possibilities of the approach to constitute an accurate
virtual laboratory and it opens also the possibility of modelling rockfill structures in the
near future.
References
[1] Nobari, E.S. & Duncan, J.M. (1972). Effect of reservoir filling on stresses and movements in earth and
rockfill dams. Report No. TE-72-1. Department of Civil Engineering, University of California.
[2] Oldecop, L. (2000). Compresibilidad de escolleras. Influencia de la humedad, PhD Thesis, Department
of Geotechnical Engineering and Geosciences, Technical University of Catalonia. UPC, Barcelona,
Spain.
[3] Alonso, E.E., Olivella, S. & Pinyol, N.M. (2005). A review of Beliche Dam. Gotechnique 55, No. 4,
267285.
[4] Oldecop, L. & Alonso, E.E. (2007). Theoretical investigation of the time-dependent behaviour of rockfill.
Gotechnique 57, No. 3, 289301.
[5] Soriano, A. & Snchez, F.J. (1999a). Settlements of railroad high embankments. Proc. 12th European
Conf. on Soil Mechanics and Geotechnical Engineering, Netherlands, 1885-1890.
[6] Soriano, A. & Snchez, F.J. (1999b). Deformaciones en presas de materiales sueltos. VI Jornadas
Espaolas de Presas, Mlaga, Spain.
[7] Kjaernsli, B. & Sande, A. (1963). Compressibility of some coarse-grained materials. Proc. of the First
European Conference on Soil Mechanics and Foundation Engineering, 1, Weisbaden, Germany, 1518
October 1963, 245251.
[8] Sowers, G.F., Williams, R.C. & Wallace, T.S. (1965). Compressibility of broken rock and settlement of
rockfills. Proc. 6th ICSMFE, Montreal: Vol. 2, 561565.
[9] Marsal, R.J. (1967). Large-scale testing of rockfills materials. Journal of the Soil Mechanics and
Foundation Division 93, No. 2, 2743.
[10] Marsal, R.J. (1973). Mechanical properties of rockfill. In Embankment dam engineering. Casagrande
Volume I (eds R. C. Hirschfeld & S. J. Poulos). New York: John Wiley & Sons, 110200.
[11] Fumagalli, E. (1969). Tests on cohesionless materials for rockfill dams. Journal of the Soil Mechanics
and Foundation Division 95, No. SM1, 313330.
[12] Penman, A. & Charles, J. (1976). The quality and suitability of rockfill used in dam construction. Dams
and Embankments. Practical Studies from the BRE, The Construction Press, 6, 7285.
[13] Clements, R.P. (1981). The deformation of rockfill: inter-particle behaviour, bulk properties and
behaviour in dams. PhD. Thesis, Faculty of Engineering, Kings College, London University.
[14] Parkin, A.K. & Adikari, G.S.N. (1981). Rockfill deformation from large-scale tests. Proceedings, 10th
Int. Conf. Soil Mech. And Found. Engg. (ICSMFE), Stockholm, 4, 727-731.
[15] Veiga Pinto, A.A. (1983). Previsao do comportamento estrutural de barragens de enrocamento. PhD
thesis, Laboratorio Nacional de Engenharia Civil, Lisbon.
[16] Naylor, D.J., Maranha das Neves, E., Mattar, Jr.D. & Veiga Pinto, A.A. (1986). Prediction of
construction performance of Beliche Dam. Gotechnique 36, No. 3, 359376.
[17] Caproni, N. & Armelin, J.L. (1998). Deformabilidade de um enrocamento obtida em ensaios de
compresso unidimensional. XI Congreso Brasileiro de Mecnica dos Solos e Engenharia Geotcnica,
Brasilia, Brasil, 2, 705712
[18] Oldecop, L. & Alonso, E.E. (2001). A model for rockfill compressibility. Gotechnique, 51, No. 2,
127139.
[19] Oldecop, L. & Alonso, E.E. (2004). Testing rockfill under relative humidity control. Geotech. Test. J.
27, 269278.
[20] Ortega, E. (2008). Comportamiento de materiales granulares gruesos - Efecto de la succin. PhD
Thesis, Department of Geotechnical Engineering and Geosciences, Technical University of Catalonia.
UPC, Barcelona, Spain.
[21] Marachi, N.D., Chan, C.K., Seed, H.B. & Duncan, J.M. (1969). Strength and deformation
characteristics of rockfill materials, Report No. TE-69-5. Department of Civil Engineering, University
of California, Berkeley, Sept., 139 pp.
[22] Marachi, N.D., Chan, C.K. & Seed, H.B. (1972). Evaluation of properties of rockfill materials. Journal
of the Soil Mechanics and Foundation Division 98, No. SM1, 95114.
[23] Naylor, D.J., Maranha, J.R., Maranha das Neves, E. & Veiga Pinto, A.A. (1997). A back-analysis of
Beliche Dam. Gotechnique 47, No. 2, 221233.
[24] Indraratna, B., Wijewardena, L.S.S., & Balasubramaniam, A.S. (1993). Large-scale triaxial testing of
greywacke rockfill. Gotechnique 43, No. 1, 3751.
[25] Indraratna, B., Ionescu, D. & Christie, H.D. (1998). Shear behaviour of railway ballast based on large-
scale triaxial tests. Journal of Geotechnical and Geoenvironmental Engineering 124, No. 5, 439449.
[26] Yasuda, N. & Matsumoto, N. (1994). Comparisons of deformation characteristics of rockfill materials
using monotonic and cyclic loading laboratory tests and in situ tests. Canadian Geotechnical Journal
31, No. 2, 162174.
[27] Yasuda, N., Matsumoto, N., Yoshioka, R. & Takahashi, M. (1997). Undrained monotonic and cyclic
strength of compacted rockfill material from triaxial and torsional simple shear tests. Canadian
Geotechnical Journal 34, No. 3, 357367.
[28] Chvez, C. & Alonso, E.E. (2003). A constitutive model for crushed granular aggregates which
includes suction effects. Soils and Foundations 43, No. 4, 215227.
[29] Einav, I. (2007a). Breakage mechanics. Part I: Theory. J. Mech. Phys. Solids 55, No. 6, 12741297.
[30] Einav, I. (2007b). Breakage mechanics. Part II: Modelling granular materials. J. Mech. Phys. Solids 55,
No. 6, 12981320.
[31] Alonso, E.E., Gens, A. & Josa, A. (1990). A constitutive model for partially saturated soils.
[32] Oldecop, L. & Alonso, E.E. (2003). Suction effects on rockfill compressibility. Gotechnique 53, 289
292.
[33] Charles, R.J. (1958a). Static fatigue of glass I. Journal of Applied Physics 29, No. 11, 15491553.
[34] Charles, R.J. (1958b). Static fatigue of glass II. Journal of Applied Physics 29, No. 11, 15541560.
[35] Wiederhorn, S.M. & Boltz, L.H. (1970). Stress corrosion and static fatigue of glass. J. Am. Ceram. Soc.
53, 543548.
[36] Atkinson, B.K. (1979). A fracture mechanics study of subcritical tensile cracking of quartz in wet
environments. Pure Appl. Geophys. 117, 10111024.
[37] Atkinson, B.K. (1984). Subcritical crack growth in geological materials. J. Geophys. Res. 89, 4077
4114.
[38] Freiman, S.W. (1984). Effects of Chemical Environments on Slow Crack Growth in Glasses and
Ceramics. Geophys. Res. 89, 40724076.
[39] Atkinson, B.K. & Meredith, P.G. (1987). The theory of subcritical crack growth with applications to
minerals and rocks, in: Atkinson, B.K. (Ed.), Fracture Mechanics of Rock. Academic Press, London
(U.K), 111166.
[40] Wiederhorn, S.M., Freiman, S.W., Fuller, E.R. & Simmons, C.J. (1982). Effects of water and other
dielectrics on crack growth. Mater. Sci. 17, 34603478.
[41] McDowell, G.R. & Bolton, M.D. (1998). On the micromechanics of crushable aggregates.
[42] Frossard, E., Hu, W., Dano, C. & Hicher, P.Y. (2012). Rockfill shear strength evaluation: a rational
method based on size effects. Gotechnique 62, No. 5, 415427.
[43] Alonso, E.E., Tapias, M. & Gili, J. (2012). Scale effects in rockfill behaviour. Gotechnique Letters 2,
No. 3, 155160.
[44] Ovalle, C., Frossard, E., Dano, C., Hu, W., Maiolino, S. & Hicher, P. (2014). The effect of size on the
strength of coarse rock aggregates and large rockfill samples through experimental data. Acta
Mechanica 225, No. 8, 21992216.
[45] Turcotte, D.L. (1986). Fractals and fragmentation. Journal of Geophysical Research- Solid Earth 91,
No. B2, 19211926.
[46] Coop, M.R., Sorensen, K.K., Freitas, T.B. & Georgoutsos, G. (2004). Particle breakage during shearing
of a carbonate sand. Gotechnique 54, No. 3, 157163.
[47] Zhang, X. & Baudet, B. (2013). Particle breakage in gap-graded soil. Gotechnique Letters 3, No. 2,
7277.
[48] Tapias, M., Alonso, E.E. & Gili, J. (2014). Compressibility and grain breakage of a gap graded material.
International Symposium on Geomechanics from Micro to Macro (IS-Cambridge 2014), University of
Cambridge, Cambridge, UK. September, 2014, 10751080.
[49] Einav, I. (2007c). Fracture propagation in brittle granular matter. Proceedings of the Royal Society A
463, 30213035.
[50] Tapias, M., Alonso, E.E. & Gili, J. (2015b). Compressibility, grain breakage and time dependent of a
gap graded material. Soils and Foundations. Under review.
[51] Nakata, Y., Hyodo, M., Hyde, A., Kato, Y. & Murata, H. (2001). Microscopic particle crushing of sand
subjected to high pressure one-dimensional compression. Soils and Foundations, Vol. 41(1), 6982.
[52] Weibull, W. (1939). A statistical theory of the strength of materials. Proc. Roy. Swedish Inst. Eng. Res.
151.
[53] Weibull, W. (1951). A statistical distribution function of wide applicability. J. Appl. Mech. 18,
293297.
[54] Bazant, Z.P. (1985). Size effect in blunt fracture: Concrete, rock, metal. Journal of Engineering
Mechanics 110, No. 4, 518535.
[55] Alonso, E.E., Olivella, S., Soriano, A., Pinyol, N.M. & Esteban, F. (2011). Modelling the response of
Lechago earth and rockfill dam. Gotechnique, 61, No. 5, 387407.
[56] Cundall, P.A. & Strack, O.D.L. (1979). A discrete numerical model for granular assemblies.
[57] Tapias, M., Alonso, E.E. & Gili, J. (2015a). A particle model for rockfill behaviour. Gotechnique.
Accepted paper- In eddition.
[58] Russell, A. & Muir Wood, D. (2009). Point load tests and strength measurements for brittle spheres. Int.
J. Rock Mechanics & Mining Sciences 46, No. 2, 272280.
[59] Bruce, I.G., Cruden, D.M. & Eaton, T.M. (1989). Use a tilting table to determinate the basic friction
angle of hard rock samples. Canadian Geotechnical Journal 26, No. 3, 474479.
[60] Zhixi, C., Mian, C., Yan, J. & Rongzun, H. (1997). Determination of rock fracture toughness and its
relationship with acoustic velocity. Int. J. Rock Mech. & Min. Sci. 34, No. 3-4, 49.e1-49.e11.
[61] Chang, S.H , Chung-In L. & Jeon S. (2002). Measurement of rock fracture toughness under modes I
and II and mixed-mode conditions by using disc-type specimens. Engineering Geology 66, No. 1-2,
7997.
[62] Backers, T. (2004). Fracture toughness determination and micromechanics of rock under mode I and
mode II loading. Ph.D. Thesis. Institut fr Gravissenschaften. University of Postdam, Postdam,
Germany.
[63] Alehossein, H. & Boland, J.N. (2004). Strength, toughness, damage and fatigue of rock. Proc.
Structural Integrity and Fracture International Conference (SIF04). Brisbane (Australia)
2629/09/2004.
[64] Ayatollahi, M.R. & Aliha, M.R.M. (2008). Mixed mode fracture analysis of polycrystalline graphite-a
hite-a modified MTS criterion. Carbon, 4, No. 10, 13021308.
[65] Saouma, V.E. (2007). Lecture Notes in Fracture Mechanics, Department of Geotechnical Engineering
and Geosciences, Technical University of Catalonia. UPC, Barcelona, Spain. (2007).
[66] Tapias, M., Alonso, E.E. & Gili, J.A. (2011). Analysis of micro-properties for triaxial behaviour on
coarse aggregates using DEM. Proceedings of the 2nd International Conference on Particle-Based
Methods: Fundamentals and Applications. Barcelona (Spain), October 2628th, 2011. 12p.
[67] Tsoungui, O., Vallet D. & Charmet, J.C. (1999). Numerical model of crushing of grains inside two-
dimensional granular materials. Powder Technology 105, No. 1-3, 190198.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-115
Limitations of a Critical State Framework

Applied to the Behaviour of Natural and
Transitional Soils.
Matthew COOPa
a
City University of Hong Kong, Kowloon, Hong Kong
Abstract. Critical state soil mechanics has provided an invaluable framework,

essential for a proper understanding of the mechanics of many soils, both
reconstituted and natural, at both small and larger strains. A limitation in its
applicability results from robust forms of fabric in natural and reconstituted soils
that mean that the critical states that can be defined in conventional tests do not
correspond to a unique fabric. Transitional behaviour, in which the initial soil
density plays a major role, seems also to result from robust fabrics. A number of
examples of transitional behaviour are explored, emphasising that this type of
behaviour does not threaten the applicability of critical state soil mechanics, but
only the definition of intrinsic behaviour, such as we might require to examine the
effects of structure. A range of behaviour appears from the examples, from very
little convergence of specific volumes during compression or shear to a slow but
gradual convergence. Even slow convergence, however, may still preclude the
definition of normal compression and critical state lines that are independent of
initial density within a useful stress range.
Keywords. Critical state soil mechanics, structure, laboratory testing.
1. Introduction
As the research discipline of soil mechanics becomes ever larger, introducing a

plethora of influences that we must account for in understanding the mechanics of our
soils, it is perhaps only with reluctance that we should admit yet another new factor and
at the very least we must be clear about whether or when it is absolutely necessary or
not. As will be discussed in this paper one factor it may be necessary to consider is the
possible influence of the initial specific volume of a soil, but as will also be emphasised,
for many practical applications the effect may simply be ignored. This mode of
behaviour has been given the somewhat unsatisfactory name of transitional, perhaps
because it was initially observed for soils with gradings intermediate to clays and clean,
single graded sands.
Transitional soils are defined as those soils for which the initial specific volume
continues to affect its compression behaviour and shearing behaviour to large strains, to
the extent that unique normal compression lines and critical state lines cannot be found
by means of conventional laboratory testing. It is a mode of behaviour that has only
been emphasised in relatively recent years and has been found typically to apply to
soils with mixed grading and/or mixed mineralogy (e.g. Shipton & Coop [1]). Perhaps
many of us have been making the assumption that if clays followed a critical state
116 M. Coop / Limitations of a Critical State Framework
framework (Schofield & Wroth [2]) and that if sands had also been found to follow
very similar patterns of behaviour, with some allowance for the effects of particle
breakage, (e.g. Coop & Lee [3]), then intermediate soils also should. Instead, work on
soils of mixed gradings has tended to concentrate on the quest to normalise out the
effects of added fines by means of various skeletal or granular void ratios (e.g.
Thevanayagam & Mohan [4]), for example by finding a unique critical state line for all
gradings, but it is possible that this may miss the more important point that the lines
may sometimes not be unique even for one mixture.
This influence of initial specific volume could only be maintained through the
fabric of the soil and while there is general agreement that sample preparation methods
can have strong effects on the behaviour of reconstituted soils or that intact soils can
possess fabrics that are rather robust and difficult to break down, there is a reluctance to
accept that the initial specific volume, per se, represents at least one element of fabric,
independently of whether the same preparation procedure was used or not. As will be
discussed in this paper, one of the difficulties to overcome in this respect may be the
emphasis in soil mechanics of the influence of fabric on anisotropy. While anisotropic
fabrics may be easily quantified by various means, such as the orientations of particles,
voids or contacts [5, 6, 7, 8], this is not the only type of fabric and heterogeneous forms
may also be important. The papers starts with a brief review of examples of soils with
natural structures that are robust in nature and not easily broken down under strain,
several of which are heterogeneous rather than specifically anisotropic.
One of the principle problems caused by a transitional mode of behaviour is that,
as pointed out by Ferreira & Bica [9], for such soils the effects of the natural structure
cannot be quantified by the means proposed by Burland [10] of comparing the
behaviour of a natural soil with the intrinsic properties defined by reconstituted
samples because there are no unique intrinsic properties. Burland suggested that unique
intrinsic properties could be defined provided that a clay was reconstituted at water
contents significantly above the liquid limit and that it was compressed to at least
100kPa, but the data he examined in making this proposal were for a limited range of
clays and silty clays, and, as will be discussed in this paper, for many other soils this
may not be appropriate.
2. Examples of Robust Fabrics in Natural Soils
In research on natural soils there has been an acceptance for some time that not all
forms of fabric may necessarily be completely broken down either in compression or
shearing. Figure 1 gives a recent example in which the oedometer compression curves
of four stiff UK clays have been compared to their intrinsic compression lines, ICL,
using Burlands [10] normalised void index, Iv:

(1)

where e*100 and e*1000 are the void ratios on the intrinsic normal compression line at 100
and 1000kPa. In each case the natural soils yields slowly, after which any convergence
with the ICL due to destructuration is at best slow, and in some cases the compression
curves actually diverge from the ICL.
M. Coop / Limitations of a Critical State Framework 117
Figure 1. Oedometer compression paths for four UK stiff clays (Hosseini-Kamal et al. [11]).
This lack of convergence of compression paths, or sometimes simply very slow

convergence, was related by Cotecchia & Chandler [12, 13] to the soil fabric through
SEM analyses of samples compressed to very high stress levels. They observed that the
natural soil maintained a distinctly different fabric to the reconstituted throughout all
stages of compression and that there was little tendency towards convergence of the
fabric. In simpler reconstituted kaolin, Hattab & Fleureau [7] have been able to relate
particle orientation to direction of loading. But in these natural soils, Cotecchia &
Chandler observed that the fabrics of both the intact and reconstituted samples were
complex in nature with some elements that showed orientation while others were of a
more heterogeneous nature. Compression of the natural soil was not found simply to
induce greater clay particle orientation, but there were much more complex and
heterogeneous changes, with alternating particle stacks and flocculated honeycomb
structures with some horizontal bands of well-orientated domains.
While Hosseini-Kamal et al. [11] could not identify the critical state lines in the
volumetric plane (e:lnp' or v:lnp') due to strain localisation, for other stiff geologically
old clays, Rampello & Silvestri [14] and also Cotecchia & Chandler [12] have found
that the critical state lines of the intact and reconstituted soils have a similar offset to
the compression paths.
A much simpler and larger scale example of robust structure can be found through
the effects of layering in younger soils. Figure 2a shows one-dimensional compression
paths from oedometer and triaxial tests for normally consolidated Holocene sediments
of coastal and alluvial origin from the archaeological site of Sibari, Italy. The paths
have little tendency to converge with the ICL post-yield. In shearing the same pattern
of lack of convergence is seen (Fig.2b) and the end of test states of the intact samples
lie well outside the intrinsic state boundary surface defined by the reconstituted
samples. The shearing data have been normalised by an equivalent pressure taken on
the one-dimensional compression ICL:

(2)

where N0 and are the intercept at 1kPa and gradient of the ICL. One of the intact
samples was one-dimensionally overconsolidated back to an isotropic initial state. If
the normally consolidated samples had not reached the intrinsic critical state line
(ICSL) through any incompleteness of shearing then this should have allowed the path
to converge towards the intrinsic ICSL more easily, but in fact the path diverges from it,
confirming that the lack of convergence is the result of a particularly robust fabric. The
nature of that fabric can be seen in Fig.3. The soil is distinctly heterogeneous in nature,
with a layered fabric that has resulted from periodic changes in the sediments deposited.
This thin section was cut from a triaxial sample after shearing to a critical state, with
constant stress and volume, emphasising that this type of fabric cannot be removed by
triaxial shearing. Similar robust effect of layered fabrics can be seen in varved clays
(DeGroot & Lutenegger [15]).
(a) Oedometer compression paths (b) Normalised shearing paths

Figure 2. Behaviour of normally consolidated sediments from Sibari (Coop & Cotecchia [16]).
Figure 3. Thin section of Sibari sediments.
As a simplistic model to demonstrate the effects of layering, Coop & Cotecchia

[16] created triaxial samples with two layers, a lower one of kaolin and an upper one of
quartz sand, comparing the behaviour with that of reconstituted samples mixed in the
same proportion. The volumetric offset of the isotropic normal compression lines
(NCL) was large and when normalised for volume with a p'e taken on the isotropic ICL
the state boundary surface of the intact soil is very much larger, emphasising the very
significant effects that layering may have (Fig.4). In the case of a crude layering like
this the different soils could of course be tested individually, but at the finer scale, as in
Fig.3, it becomes part of the structure.
Figure 4. State boundary surface of layered model structured soil (modified from Coop & Cotecchia [16]).
Figure 5. Oedometer data for structurally complex clays (Fearon & Coop [17]; a-d, DElia [18]; e Bertuccioli
& Lanzo [19]; f, g Picarelli [20]; h Guerriero et al. [21]; i-k Santaloia et al. [22]).
To allow for elements of structure that could not easily be broken down by
compression or shearing, in their sensitivity framework Cotecchia & Chandler [23]
allowed different isotropic NCLs and CSLs to exist for the intact and reconstituted soil
in the v:lnp' plane, and this feature was included in the constitutive model of Baudet &
Stallebrass [24]. However, the volumetric offsets due to such robust effects of fabric
are not always positive and in Fig.5 the compression curves of Italian structurally
complex clays have negative offsets and many of them show little sign of convergence
even at high stress levels. This is again because of a heterogeneous, often scaly fabric
and Cotecchia et al. [25] have shown that these soils also have offset CSLs for their
intact and reconstituted states.
Figure 6. Maximum rates of dilation (dmax) for a cemented Greensand (Cuccovillo & Coop [26]).
Figure 7. SEM image of an uncemented Greensand (Fonseca et al. [27]).
Examples of robust fabrics in natural sands are harder to establish, because very
high pressures may be needed to reach states outside the ICL in compression and at
lower stress levels critical states may be difficult to identify because of strain
localisation. For a cemented Greensand Cuccovillo & Coop [26] showed that the rates
of dilation at peak (dmax) were very much higher for the intact than the reconstituted
soil due to the cementing combined with a locked fabric. When plotted against a
quantification of state using an equivalent pressure on the intrinsic CSL (p'cs) (Fig.6)
the data also indicated that the CSL for the intact soil must be offset compared to the
reconstituted since the values of dmax for the intact soil were still large at p'/ p'cs=1. This
was unlikely to result from the effects of the locked fabric or cementing, which were
both broken during shearing and testing an uncemented Greensand Fonseca et al. [27]
found that the particles were frequently fractured (Fig.7) and similar particle fissuring
can be seen in the cemented Greensand. This is the most likely cause of the CSL
movement, because the reconstitution changed the grading. This example highlights the
pitfalls of our simple distinction between soil particles and the soil fabric.
3. Examples of Robust Fabrics in Reconstituted Soils
There is a tendency to believe that reconstituted clays are structureless and have no
defined fabric, but as Cotecchia & Chandler [12] stressed, they are not unstructured
but simply have a fabric that is different to that of the natural soil and only appropriate
to the particular means of reconstitution chosen. This was emphasised by the work of
Masin et al. [28] who found that London clay that was sedimented in the laboratory had
a NCL and CSL that lay above and practically parallel to the same soil reconstituted in
the usual fashion. Another interesting example of the variability of fabric of
reconstituted clays is that of the structurally complex clays. The volumetric offsets seen
on Fig.5 are caused by the scaly meso-fabric that can be removed by conventional
reconstitution. However these soils also have a micro-fabric of aggregated or composite
clay particles (Fig.8), which survives reconstitution. Fearon & Coop [29] broke down
these aggregates by passing the soil repeatedly through an industrial food mincer,
increasing its plasticity and moving the isotropic NCL in the v:lnp' plane and increasing
its gradient (Fig.9). Even undrained shearing to a constant stress state could not erase
the effects of the different fabrics since the clay aggregates are very robust. This is
therefore another example of heterogeneity of fabric controlling behaviour. Fearon &
Coop concluded that the appropriate reconstituted soil for the assessment of the effects
of structure in their structurally complex clay was that reconstituted by hand, not
disaggregating the particles, since in most applications that would not occur. However,
care must be taken about when violent remoulding may take place in practice and in the
Mercia Mudstone Atkinson et al. [30] attributed the blocking of a tunnel boring
machine to the disaggregation of similar composite clay particles.
Figure 8. SEM micrograph of a structurally complex clay (Fearon & Coop [29]).
The effects of different sample preparation techniques on the behaviour of sands

are well documented, so that the definition of an intrinsic behaviour is again
problematic. Chu et al. [31] and Jefferies & Been [32] have seen differences at small to
intermediate strains between moist tamped and water pluviated samples, and in cases
there have also been effects at larger strains, affecting the location of the CSL [31, 33,
34, 35]. Yang et al. [36] quantified the anisotropy of fabric that they found was
responsible for differences of behaviour between dry pluviated and wet compacted
samples up the large strains. Using DEM of elongated particles Nougier-Lehon et al.
[37] showed that while reasonably constant volumes and stresses could be reached at
strain levels that would be possible in common laboratory tests, reaching a unique
critical state fabric could require strains of up to 100%, depending on the initial fabric
orientation. The effects of fabric on the stress-strain behaviour and even critical states
are particularly pronounced in sands with added fines [38, 39] and especially well
graded soils, as recently shown by Madhusudhan & Baudet [40].
Figure 9. Triaxial data for reconstituted and minced samples of a structurally complex clay (Fearon & Coop
[29]).
Earlier research on the effects of particle breakage on the mechanics of sands (e.g.
Coop & Lee [3]) suggested a unique relationship between particle breakage and the
critical state so that the CSL was thought to be unique in the v:lnp' plane. However, as
Muir Wood [41] emphasised, this was because the tests used to construct the CSL were
all normally consolidated, in the sense that the samples were at the maximum stresses
they had experienced. Cheng et al. [42] simulated particle breakage in DEM analyses in
which prior breakage was created by loading the samples to high stresses before
shearing at lower stresses. This led to movement of the critical state line as a result of
the prior breakage. Experimental conformation of this was provided by Bandini &
Coop [43]. The Dogs Bay sand, a biogenic carbonate with especially weak particles,
was sheared to critical state in a triaxial at three higher stress levels following constant
p' paths (500, 1000 and 4000kPa) to create large amounts of breakage. The samples
were then unloaded and sheared a second time at lower stresses, identifying three new
CSLs for the three degrees of relative breakage (Hardin [44]) that had been created by
first shearing (Br=0.14-0.23, 0.31-0.35 and 0.42-0.47). This confirmed Muir Woods
proposal that the CSL would move due to breakage, although large amounts of
breakage were needed to cause a significant movement compared to the CSL of the
original grading determined by Coop [45] (Fig.10).
2
from estimated initial void ratio
1.6 CSL original grading

CSL for second shearing
Br=0.14-0.23
Br=0.31-0.35
Br=0.42-0.47
1.2
void ratio
Second shearing stage

0.8 CS points
OR1
NCL
OR3
OR9
OR10
OR11
0.4 Extrapolated CS points
OR2
OR6
OR7 CSL DBS
OR8 (Coop, 1990)
0
10 100 1000 10000 100000
mean effective stress (kPa)
Figure 10. Movement of the CSL of a carbonate sand due to prior breakage (Bandini & Coop [43]).
2
from estimated initial void ratio
1.6
CSL original grading

CSL for second shearing
1.2 Br=0.42-0.47
void ratio
CS points
RC5
0.8
RC13
RC14
RC15 NCL
0.4 CSL of reconstituted DBS with Br=0.42-0.47
CSL DBS
(Coop, 1990)
0
10 100 1000 10000 100000
mean effective stress (kPa)
Figure 11. Comparison of effects of prior breakage and grading change with reconstitution on the movement
of the CSL of a carbonate sand (Bandini & Coop [43]).
What was perhaps overlooked about the work of Bandini & Coop [43] was that if
samples of the same grading as those with Br=0.42-0.47 were created without prior
breakage in the triaxial, then these reconstituted samples gave a completely different
CSL in the v:lnp' plane to those that had undergone prior breakage without
reconstitution (Fig.11). Analysis of thin sections of the samples that had been sheared
to critical state revealed that the reason for this discrepancy lay in the fabrics of the
samples. In Fig.12a the breakage that many particles have undergone is evident, but the
broken pieces remain in close proximity to each other, and a number of examples have
been highlighted. In contrast, for a sample reconstituted with the same grading the
broken pieces will be uniformly distributed throughout the sample, as can be seen in
Fig.12b. It is therefore a robust heterogeneous fabric in the samples that had undergone
prior breakage without reconstitution that affects the CSL location for triaxial testing.
In these tests the samples had been sheared in compression until constant stress and
volume states had been reached, and it might be argued that perhaps other stress paths
would cause greater disruption to the fabric, but it is difficult to imagine that any
monotonic loading path would be able to cause the broken particle pieces to
redistribute throughout the sample to give the same fabric as the reconstituted soil.
(a) sample with prior breakage (b) reconstituted sample

Figure 12. Thin sections of carbonate sands tested by Bandini & Coop [43].
4. Examples of Transitional Behaviour
Regardless of its arguably unsatisfactory name, anyone who is in any doubt that
transitional describes aspects of behaviour not easily otherwise categorised or
described should carry out a simple series of oedometer tests on 75% quartz sand 25%
kaolin mixtures. As Martins et al. [46] and Shipton & Coop [1, 47] have highlighted,
the compression curves will show little convergence towards a unique NCL, no matter
what sample preparation method is used. Typical data are shown in Fig.13 and in
Fig.14 data are shown for similar tests on a 75% quartz sand 25% non-plastic fines
(crushed quartz) mixture, which demonstrate that the phenomenon is not confined to
sand-plastic fines mixes. In both cases the oedometer compression curves, which reach
higher stress levels than in isotropic compression, fail to reach a unique NCL even at
the highest stresses. The compression curves do not resemble those of clean sands, for
which the compression would be initially flat until the onset of particle breakage gave a
well-defined yield and a NCL [3, 48], but instead are gently and gradually yielding and
steepening from low stress levels. At extremely high stresses, beyond those used, the
compression curves must all tend to a horizontal asymptote at v=1 but there is little
evidence that there is a likelihood that they could possibly define a unique NCL before
they do that.
1.8
one-dimensional compression
isotropic compression
1.6
specific volume, v
1.4
1.2
1
10 100 1000 10000 100000
p' (kPa)
Figure 13. Compression of 75% quartz sand-25% kaolin mixture (Shipton & Coop [47]).
1.6 one-dimensional compresison

isotropic compression
1.4
specific volume, v
1.2
1
1 10 100 1000 10000 100000
p' (kPa)
Figure 14. Compression of 75% quartz sand-25% crushed quartz silt mixture (Shipton & Coop [47]).
The tests in Fig.13 essentially repeated the work of Martins et al. [46] (Fig.15).
Martins et al. had used the simple quartz sand - kaolin mixture to simulate the
behaviour of reconstituted samples of the Botucatu residual sandstone (BRS) from
Brazil, for which they had first identified the lack of convergence. The 75/25 mix
reflected the gap grading of the BRS and observing similar behaviour in the
sand/kaolin mix enabled Martins et al. to conclude that this mode of behaviour did not
result from the complex mineralogy arising from the weathered origin of the BRS but
simply from its grading.
Figure 15. Oedometer data for Botucatu residual sandstone and a sand-kaolin mixture (Martins et al. [46]).
Attention then turned to well graded rather than gap graded soils. Nocilla et al.
[49] created well graded samples of silts from the Po River, Italy, controlling the clay
fraction through a sedimentation technique. Four gradings were created, with clay fines
contents of 45%, 25%, 8% and 3.5%. In oedometer testing the compression curves for
the 45% and 25% fines converged towards unique NCLs (Fig.16a) but at 8% and 3.5%
they did not (Fig.16b). In shearing the 45% and 25% soils defined unique CSLs that
were parallel to the isotropic NCLs in a conventional manner. The triaxial data for the
8% clay content are shown in Fig.17. In isotropic compression the paths again show no
tendency to converge, and the critical states that were defined by shearing were widely
dispersed, seemingly dependent on the initial specific volume. This work showed that
transitional modes of behaviour were not confined to gap graded soils.
3 3
v Slurry v
Dry Compaction
Wet Compaction
2.5 Natural Sample 2.5
2 2
1.5 1.5
1 1
1 10 100 1000 10000 1 10 100 1000 10000
'v [kPa] 'v [kPa]

(a) 25% clay content (b) 8% clay content
Figure 16. Oedometer compression data for Po River silts (Nocilla et al. [49]).
Ferreira & Bica [9] went on to show that the BRS in a reconstituted state had two
different CSLs for two different initial specific volumes, although the CSL of the
natural samples was apparently unique. This work was repeated for a greater range of
initial specific volumes and for the simpler 75% sand- 25% kaolin mix by Shipton &
Coop [47], the results of which are shown in Fig.18. Samples were made with five
different initial specific volumes and the critical states in triaxial compression of each
group defined separate and parallel lines; it would be impossible on Fig.18a to find any
unique line that could describe the data. The differences between the various CSLs are
large, and as will be discussed later, are far too large to be dismissed as incomplete
testing or inaccuracy of the data. The strains during isotropic compression were
isotropic for all but the initial loading stages, and so these differences of CSL are
unlikely to arise from an anisotropic fabric. A similar influence of the initial specific
volume on CSL location was also found for the sand with crushed quartz fines and
Yamamuro & Lade [50] had also seen an effect of initial density on the location of the
CSL in the v:lnp' plane for reconstituted silty sands. In neither the quartz sand with
kaolin nor the quartz sand with crushed quartz silt was there any evidence of particle
breakage in any test, either in oedometer loading to high stresses or in shearing to
critical state at lower stress levels.
2.4
Isotropic Compression
v Shearing Path
Critical State
2.2
2
T14
T13
1.8 T15
T20 T16
T18
1.6 T17
1.4
10 100 p' [kPa] 1000
Figure 17. Isotropic compression and shearing of Po River silts with 8% clay content (Nocilla et al. [49]).
1.8 1.8
isotropic compression isotropic compression
shearing shearing
critical state lines PS41 critical state lines
A
1.6 1.6
B PS40
PS41
C PD31 PS40
Group A
1.4 D 1.4
V
Group B
E
Group C
1.2 group initial v critical states 1.2

A 1.690-1.691
B 1.571-1.578 critical states
C 1.472-1.481 PD31 (overconsolidated)
D 1.416-1.418 PS40 (sheared twice)
E 1.314-1.316 PS41 (sheared twice)
1 1
10 100 1000 10000 10 100 1000 10000
p' (kPa) p' (kPa)
(a) five groups of specific volume (b) overconsolidation and second shearing
Figure 18. Triaxial data for for 75% sand- 25% kaolin samples (Shipton & Coop [47]).
Because of the possibility that a simple triaxial compression test might be thought
insufficient to reach a true critical state, Shipton & Coop [47] investigated the
robustness of the critical states they had determined by two means. The first was to
recompress the samples to a higher stress level after reaching a critical state and then
shear again. Samples PS40 and PS41 on Fig.18b show that the samples simply reach
the same CSLs on second shearing, those for Groups B and A respectively. Sample
PD31 was heavily overconsolidated prior to undrained shearing. Similar to the tests on
the Sibari natural clays (Fig.2b), if it were assumed that for these soils, which are
mostly compressive during shearing, that the true CSL lay below those identified
because of incompletes of shearing, then this might have allowed the sample to reach
that lower state. However, it did not and the critical state was on the same CSL as other
samples of specific volume Group C.
drained
undrained
D dry compacted
2 w wet compacted
s slurry
q'/p'cs
CSL
SBS?
1
w
s
s
s
w w D
w
s s s s s s
s s s s s s
0 Dw s s s s s
0 1 2 3 4 5
p'/p'cs
Figure 19. State boundary surface for 75% sand 25% clay mix (Shipton & Coop [47]).
1000
s = 0.01% Undrained
s = 0.1% Undrained
s = 0.1% Drained
100
A
10
1.2 1.3 1.4 1.5 1.6 1.7
Vcurrent
Figure 20. Values of stiffness parameter A for 75% sand 25% clay mix (Shipton & Coop [47]).
If instead of assuming one unique CSL for all initial specific volumes, it is
assumed that the CSL location is linked to the initial specific volume, so that there are
an infinite number, then the shearing paths may be normalised with respect to an
equivalent pressure, p'cs taken on the respective CSL for each initial specific volume,
and this does lead to a unique state boundary surface being identified (Fig.19). The
effects of volume on the stiffness of transitional soils are not clear. Tangent shear
moduli calculated at two strain levels in Fig.20 indicate little clear effect of the specific
volume on the stiffness through the value of A in G=Ap'n.
Oedometer tests are especially convenient in identifying when non-convergent
behaviour occurs, particularly if they are continued to high stress levels so that large
strains are imposed, and since the strains are both volumetric and shear, they can also
give an indication of whether a unique CSL can be expected or not. With oedometer
testing three or four accurately conducted tests with different initial specific volumes
can be sufficient to identify this mode of behaviour, while to verify that the CSLs are
also non-unique requires a series of triaxial tests at each initial specific volume, which
is very much more onerous.
Shipton & Coop [1] made an extensive investigation of when non-convergence
occurs in oedometer tests for a wide variety of mixed soils, concluding that the factors
involved were complex, and could not be related, for example, simply to grading. As
seen from the examples of Nocilla et al. [49] and Shipton & Coop [47], it may occur in
well graded or gap graded soils. Shipton & Coop [1] tested other ratios of quartz sand-
kaolin in the oedometer (Fig.21). For 10% kaolin content the data are presented in
terms of a granular specific volume, vg (=1+eg=1+[e+fc]/[1-fc], where fc is fines
content) and at 50% content a fines specific volume, vf, has been used (=1+ef=1+[e/fc]).
The use of vg or vf only serves to translate the compression curves for comparison with
the 100% sand or 100% kaolin data, but will not affect convergence. At 50% kaolin
content the compression curves define a unique straight NCL that is only slightly
displaced from that of the 100% kaolin when using vf, while at 10% content the data
resemble those for 100% sand, with perhaps slightly less tendency to converge, but
both show a typical sand mode of behaviour of yield before approaching a unique NCL
at high stresses [3, 48]. Neither show the lack of convergence evident at 25% clay
content (Fig.13) emphasising the sensitivity of the mode of behaviour to the grading.
(a) 0% and 50% sand (b) 100% and 90% sand

Figure 21. Oedometer test data for quartz sand: kaolin and mixtures (Shipton & Coop [1]).
To investigate the effect of sand mineralogy on transitional behaviour, Shipton &

Coop [1] changed the quartz sand fraction for Dogs Bay carbonate sand (Fig.22).
While there was no particle breakage of the sand fraction for the quartz sand kaolin
mix, there was substantial breakage for the carbonate sand, even at a 10% sand content,
for which the values of granular void ratio, eg, would be in excess of 20 and so it is
likely that few sand particles were in direct contact with each other. At 10% sand
content there was a non-convergent behaviour (Fig.22), which was also seen at 40%
and even at 5%, emphasising that very small changes in grain content may provoke a
transitional mode of behaviour. At 70% sand content the behaviour reverted to a typical
sand mode with well-defined yield and NCL at higher stress levels. The range of fines
contents for which non-convergent behaviour was seen was therefore much greater for
the carbonate sand kaolin mixtures than when quartz sand was used for the coarse
fraction, emphasising that grading alone could not be used as an indicator to predict
transitional behaviour. The data also contrast with that of the well-graded samples of
Nocilla et al. [49] for which 25% and 40% clay contents gave convergent behaviour
and non-convergence was only seen at low clay contents (8% and 3.5%). However, in
contrast to a quartz sand (Fig.14) when the fines were changed to crushed quartz silt no
mixture with the carbonate sand was found to be transitional and all defined NCLs,
emphasising that the plasticity of the fines can be important, but not in all soils.
In an attempt to model soils more like the well graded Po River silts of Nocilla et
al. [49], rather than gap graded soils, Shipton & Coop [1] created reconstituted samples
of 25% quartz sand, 25% crushed quartz silt and 50% kaolin, which gave non-
convergent compression curves for an artificial well graded soil, but at rather higher
clay contents than for the Po River silts, for which 25% clay content was convergent
with a unique NCL.
Figure 22. Oedometer data for a 10% carbonate sand-kaolin mixture (Shipton & Coop [1]).
Shipton & Coop [1] assembled the gradings curves of all the soils that had been
tested up to that time, either by themselves or co-workers or in the literature, which had
shown transitional behaviour. The gradings in Fig.23 are not especially helpful as they
cover a very wide range of well graded and gap graded soils, and as they had
emphasised grading alone is insufficient on its own to predict when transitional

behaviour will be seen or not and the mineralogies, strengths and natures of the
particles are likely also to be of importance.
Figure 23. Grading curves of soils which show transitional and non-convergent compression behaviour
(Shipton & Coop [1]; data from Shipton [51]).
With the exception of the Botucatu residual sandstone (BRS) most of the soils so
far described as transitional have been artificial soils recreated in the laboratory. Even
the Po River silts of Nocilla et al. [49], which would have had a natural mineralogy,
had their gradings controlled. The Thanet sand, tested extensively by Ventouras &
Coop [52] provides one example where non-convergent compression behaviour can be
seen in a soil of natural origin. Thanet sand occurs in London in two principal strata, a
shallower one of lower fines content and a deeper one with a higher fines content
(<63m) of 15-30%, which was found to have non-convergent compression behaviour
(Fig.29) even up to stress levels of 30MPa in the oedometer. These samples had a fines
content of 27%.
2.6
2.4
Specific Volume
2.2
1.8
1.6
1.4
1.2
1
10 100 1,000 10,000 100,000
'v (kPa)
Figure 24. Oedometer compression curves for high silt content Thanet sand (adapted from Shipton & Coop
[1]; data from Ventouras [53] and Ventouras & Coop [52]).
The samples of the Thanet sand for which data are given in Fig.24 were of course
reconstituted to give the range of initial specific volumes needed to check if the
behaviour was convergent or not. The work of Ponzoni et al. [54] provides, however,
an example of transitional behaviour seen in intact samples of natural soils. They
analysed a very large number of commercially conducted oedometer and triaxial tests
carried out on sediments from the locations of two of the MOSES barriers for the
protection of Venice, Malamocco and Lido, selecting only the more silty and silty clay
samples for which transitional behaviour might be expected. The soils are of upper
Pleistocene and Holocene origin and have roughly equal amounts of siliceous and
carbonate minerals, along with clay minerals. Oedometer tests were carried out with
both incremental load (IL) and constant rate of strain (CRS). Figure 25 shows examples
of intact soils with very similar gradings, but with different initial specific volumes and
in each case there is little convergence towards a unique NCL up to the maximum
stresses reached of 6MPa. This was confirmed by tests on reconstituted samples. While
there is definite convergence of the compression data, in contrast to the almost parallel
compression curves seen in some transitional soils (e.g. Fig. 24), that convergence is so
slow that it would barely occur before the asymptote at v=1 and many tens of MPa
would be required, so that a usable NCL cannot be identified.
Figure 25. Example oedometer data for groups of tests on similar soils from Venice lagoon sediments
(Ponzoni et al. [54]).
Given the heterogeneity of the soils at both sites, the approach adopted by Ponzoni
et al. [54] was to group the soils according to their clay fraction; Fig.26 shows
oedometer data for a typical group, Malamocco soils with CF=20-30%. The same
pattern of lack of convergence can be seen as in smaller groups of tests with more
tightly controlled similarity of grading on Fig.25, but as will be discussed later these
larger groups enabled a quantification of the lack of convergence to be made. Data for a
small number of tests that are indicated were discarded in the subsequent analyses as
they had unusual compression curves. Triaxial test data from another group are shown
in Fig.27. They indicate very widely dispersed critical states, but the simple model of
parallel CSLs, the location of which depends on the initial specific volume, v0, adopted
by Shipton & Coop [47], seems to fit the data well.
Figure 26. Oedometer test data for IL tests on Malamocco soils CF=20-30% (Ponzoni et al. [54]).
Figure 27. Typical CSL groups for intact samples of Venice lagoon sediments; Lido CF=10-20% (Ponzoni et
al. [54]).
Altuhafi et al. [55, 56] tested a sub-glacial basaltic till from Iceland. The natural
grading was fractal as a result of the intense shearing the sediment had experienced
under the glacier and this led to a compression behaviour that was non-convergent with
no measurable particle breakage, although shearing did bring the soil to a unique CSL.
This is not inconsistent with non-convergence in compression since the fabrics that
may be responsible for non-convergence in different transitional soils may have
different sensitivities to shear and volumetric strains, just as in structured natural soils
where different constitutive models for different soils often require different
destructuration dependence on shear and volumetric strains [24, 57]. If the fines were
removed from the basaltic till, creating a poorly graded soil that might have resembled
that in-situ before grinding under the glacier, the behaviour reverted to a simple sand
pattern with well-defined yield in compression, a unique NCL and large amounts of
particle breakage.
D1-26
NCL, Cc= 0.32 D1-27 CD-16
2
2 D1-28 CD-17
CD-19
1.6 CD-20
1.6
CD-21
Void Ratio, e
CD-22
Void Ratio, e
1.2 1.2
0.8 0.8
0.4 0.4
0 0
10 100 1000 10000 100000 10 100 1000 10000 100000

Vertical stress, kPa Vertical stress, kPa
(a) uniform D1 grading (b) fractal CD grading

Figure 28. Oedometer compression of Dogs Bay carbonate sand (Altuhafi & Coop [58]).
100
D1 grading
D1-26
80 D1-27
D1-28
D2
60 D2-23
%Finer
D2-24
D2-25
40
20
1 10 100 1000
Particle size, m
100
CD
CD-16
80 CD-17
CD-19
CD-20
60 CD-21
%Finer
CD-22
D4
40 D3
D3-4
D3-5
D3-6
20 D3-7
D3-8
D3-9
0
1 10 100 1000
Particle size, m
Figure 29. Gradings curves for Dogs Bay carbonate sand after oedometric compression (Altuhafi & Coop
[58]).
Altuhafi & Coop [58] repeated similar tests for a wider variety of gradings and a
wider range of sands of different mineralogies and typical data are shown in Figs.28
and 29 for the Dogs Bay carbonate sand. For each sand single gradings gave
oedometer compression curves that showed a typical sand mode [3, 48] with initially
flat compression curves, a well-defined yield corresponding to the onset of significant
breakage and then a unique NCL. As the initial grading changed from poorly graded
(D1) through progressively flatter gradings (D2 and D3), the amount of breakage
reduced and the NCLs became less steep and less distinct. Grading CD was fractal with
a fractal dimension of 2.57, being the final grading that Coop et al. [59] had determined
from higher stress level ring shear tests on the same sand. This grading, as well as an
even flatter fractal grading (D4) gave transitional behaviour with little convergence of
the compression paths. The final gradings of the CD samples on Fig.29 are
indistinguishable from the initial. This pattern of transitional behaviour with no
measurable breakage was also seen for a quartz sand.
One key difficulty in studying transitional soils has been the identification of any
clear element of fabric that gives rise to the lack of a convergent behaviour. Figure 30
shows scanning electron micrographs of two transitional soils. The 75% quartz sand -
25% kaolin soil of Shipton & Coop [47] (Fig.30a) indicates that the kaolin fills the void
spaces between the clay. The larger cavities are probably voids created by sand
particles being pulled out of the fabric when the broken sample surface was created for
the SEM imaging. Neither this soil nor the silty Po River soil in Fig.30(b) showed any
clear element of fabric in SEMs that could be quantified and correlated with the
transitional behaviour. One reason for this difficulty is that transitional behaviour tends
to be found in gap graded or well graded soils, with soil particle sizes that cover a very
wide range, and so finding the correct scale range at which to identify the fabric is not
straightforward.
(a) 75% quartz sand 25% kaolin (b) Po River silt with 8% clay content
Figure 30. Scanning electron micrographs two transitional soils (Shipton & Coop [47], Nocilla et al. [49]).
5. Common Objections to Transitional Behaviour
The existence of transitional behaviour strikes at the heart of many long held beliefs in
soil mechanics, notably the immutability of the CSL and that it must correspond to a
unique fabric as well as stresses and volumes. The dependence of the CSL location on
v0 must indicate that the final fabric is not unique and so the concept of critical state
becomes more of a pragmatic and empirical choice rather than a necessary consequence.
This has meant that the introduction of transitional modes of behaviour has met with
some intransigence, but some key objections must be also be considered carefully if the
non-convergent behaviour is to be validated.
There are many cases in the literature where soils have been investigated with
similar gradings and/or mineralogies to those discussed but it cannot be identified
whether they have transitional behaviour or not because the authors have only created
their samples at one initial specific volume, often concentrating more on sample
preparation effects. The temptation must also be resisted to expect transitional
behaviour in more soils than it actually occurs in. The above examples have
emphasised the sensitivity of whether this mode of behaviour is seen or not to
differences of grading and mineralogy and also how it is currently difficult to predict.
Any investigation of a new soil in a reconstituted state should include a range of
samples at different initial specific volumes to check the uniqueness of the NCL and
CSL, compressing the samples to large stress levels and shearing to large strains if
necessary. One example of where this was done was the Botanj silty sand from
Slovenia, investigated by Vilhar et al. [60] (Fig.31). This is a soil of alluvial origin that
has 30% non-plastic fines with a mineralogy dominated by limestone/dolomite. The in-
situ silty sand grading led the authors to check for transitional behaviour by creating
samples at a wide range of v0, but isotropic compression gave a well-defined yield and
a clear NCL, while shearing gave a unique CSL, within some small scatter of data.
Both compression and shearing were accompanied by significant particle breakage,
mainly in the coarser fractions. The modes of behaviour for this sand with non-plastic
fines were therefore in complete contrast to the 75% quartz sand with 25% non-plastic
fines tested by Shipton & Coop [1, 47].
0.9
0.8 5 4
16 19
22 T
18 1
0.8 3 2
0.7 17
20 23 3
24 4
8 21
7 5
9 ISO-NCL
CSL curved part: 14 15 6
10 7
e = 0.799 - 0.12 (p'/100)0.363 13
0.7 0.6 6
8
11
9
e (-)
ISO-NCL 1
e (-)
12 1
1
1
0.5 1
0.6 Starting point of 2 1
1
straight CSL
CSL 1
1
0.4 1
0.5 1
2
2
2
2
0.3 2
0.4
10 100 1000 10000 1 10 100 1000 10000
p' (kPa) p' (kPa)
(a) isotropic compression (b) shearing and critical states

Figure 31. An example of convergent behaviour in a Slovenian silty sand (Vilhar et al. [60]).
5.1. Sample Preparation Effects
As discussed above, different sample preparation methods can have a significant

effect on soil behaviour at any range of strain, from small to large. Nocilla et al. [49]
included an investigation of sample preparation effects, and some of these data can be
seen in Fig.16. The samples were made with a range of techniques, compression from
slurries, dry compaction and wet compaction, yet within the overall pattern of non-
convergence any effect of the sample preparation method is not discernible. Similarly
in the 10% carbonate sand - 90% kaolin mixture tested by Shipton & Coop [1] (Fig.22)
samples were both made as slurries and using dry compaction, with no visible effect on
the compression data.
1.9
slurry consolidated in cake
slurry in floating ring consolidometer PS41
slurry consolidated on pedestal (2nd shear)
1.8 wet compaction on pedestal PS41
dry compacted cake (1st shear)
PS40
(2nd shear)
1.7 PS40
PW29 (1st shear)
PS32
PD31
1.6
PW33
1.5
1.2 1.3 1.4 1.5 1.6 1.7 1.8
V20
Figure 32. Intercept of CSLs for 75% quartz sand - 25% kaolin mixture (Shipton & Coop [47]).
Shipton & Coop [47] included a more extensive investigation of the effects of
sample preparation for a transitional soil. The normalised shearing stress paths on
Fig.19 are distinguished between those samples created by wet compaction, dry
compaction and compression from a slurry. There is no significant difference between
them. If it is accepted that a family of parallel CSLs exits, as in Fig.18, the influence of
the initial specific volume of the samples on the CSL location can be established, as in
Fig.32. The initial specific volume here has been taken as that at 20kPa, v20, the start of
isotropic compression in the triaxial tests, and the location of the CSLs is quantified by
the intercept at 1kPa. A different value of has been calculated for every critical
state on Fig.18 and other tests not shown. The resulting relationship indicates a direct
and linear influence of v20 on , with a gradient of 0.52, so that some of the effect of
the initial specific volume and hence fabric must have been erased by compression and
shearing, as can also be seen by the slow convergence of the compression curves of
Fig.13. However, for the 75% quartz sand - 25% crushed quartz silt mixture the
equivalent gradient is 0.97, indicating that the effect of the initial specific volume is
much less easily removed in this soil. The different sample preparation methods do
allow different ranges of v20 to be achieved, and it is perfectly possible that for other
soils sample preparation might affect whether transitional behaviour were seen or not,
but there is little evidence sample preparation in any way affects the general trend of
Fig.32.
5.2. Incomplete Shearing
If different CSLs are to be firmly established then it must be certain that they do
not simply result from incomplete shearing, and a key drawback of triaxial testing are
the limited shear strains that may be applied either before the apparatus reaches a
physical limit or the strains become too non-uniform. Figure 33 shows data from
undrained compression tests on the 75% quartz sand - 25% kaolin soil, for different
initial specific volumes. If there is a unique CSL then they must reach unique values of
p' at the ends of the tests for each initial specific volume at the start of shearing. It is
clear that they do not even up to large strains and the paths tend to be parallel at the
ends of the tests. In some cases hyperbolic extrapolations have been fitted to extend the
data even further, but they do not give any greater convergence. Similarly, drained tests
at similar values of p' should reach unique values of specific volume at the ends of the
tests, but again they did not. The generally compressive nature of the soil should have
limited any effects of strain localisation.
100000
hyperbolic
extrapolations
10000 PW26
PS19 PW25
1000
p' (kPa)
PS40 PS29
PS18
100 PS14
v at start shear
PS08 1.23-1.24
10
1.35-1.36
1.41
1.44-1.46
1
0 10 20 30 40 50
Axial Strain (%)
Figure 33. Lack of convergence in undrained tests with strain for the 75% quartz sand - 25% kaolin soil
(Shipton & Coop [47]).
The use of lubricated end platens can make a significant difference to volumetric
strains in triaxial testing (e.g. [61]). Tests PS29 and PW33 were fitted with lubricated
end platens, and on Fig.33 there is some difference between the behaviour of similar
samples with lubricated end platens (Test PS29) and without (Test PS18), but within
the context of the much larger degree of differences seen between tests at critical state,
the effect is strictly secondary, so that on Fig.32 the tests plot within the general scatter
of data. The difference between the densest and loosest samples at critical state on
Fig.18 would represent a very large volumetric strain of about 14%, but what is more
unusual about the data is that there is very little tendency to converge during shearing.
If it were a case of incomplete testing then at least it might be expected that the tests
started to approach each other, especially in the early stages of shearing.
5.3. Accuracy of Specific Volumes
In a conventional soil with a unique CSL the accuracy of the specific volumes
for the tests is evident from the scatter of the data around the chosen CSL. For
example, from a comparison of specific volumes based on measurements of the final
water contents and the initial dry density, Vilhar et al. [60] estimated that the accuracy
of their specific volume measurements was about 0.01 for most of their tests rising to
0.03 for some of the looser samples, and this is consistent with the scatter of data
around the CSL on Fig.31(b). However, in order to prove a negative, that is that there is
no unique CSL, special care needs to be taken to prove that the different CSLs
identified do not simply correspond to errors in the initial specific volume. In many
cases authors have estimated the accuracy of their specific volumes by an analysis of
the precision and repeatability of individual measurements, such as sample dimensions

[62, 63, 64] and when special precautions are taken in the measurements this type of
approach tends to lead to estimates of precision that are quite tight, typically less than
0.01, although often consistent with the data scatter.
Because of the need for transitional soils of proof that the specific volumes are
correct rather than just making an estimate of their accuracy, the approach adopted has
been to require redundancy in the measurements made, so that as many measurements
of sample weights and dimensions are made as are possible both before and after
testing (Rocchi & Coop [65]), relating measurements before and after testing through
the strains measured during the test. Which may be useful or even possible to measure
will depend on the test type, soil type and sample preparation method. For example,
final dimensions may be easily taken for some sample types in oedometer tests, but are
always difficult to measure accurately in triaxial tests. Final measurements of water
content may be possible in more plastic soils, but less easy in free draining soils.
The approach then adopted has been to calculate as many values of the initial
specific volume as are possible with the greatest degree of independence possible.
These may typically be based on the initial or final values of dry unit weight, bulk unit
weight and water content, combined with the strains measured during the test. In some
cases assumptions may be needed, so it may, in some cases, be reasonable to assume
saturation so that specific volume may be calculated directly from the water content,
but in many other cases that may not be appropriate. There is often a trade-off to be
made in making an assumption, for example about saturation, which will lead to a
greater number of measurements of specific volume, perhaps of lower accuracy, and
not making the assumption and having fewer values of greater accuracy.
The degree of independence of the various measurements is also variable and
needs to be assessed for each test series. For example, the initial and final dimensions
should be directly linked though the strains measured during the test and the extent to
which they are considered independent of each other will depend on confidence in that
calculation and so the reliability of the strain measurements, so that the initial and final
heights of an oedometer sample may be considered more directly linked than the
dimensions of a triaxial sample.
The accuracy and repeatability have sometimes been improved by changing the
testing techniques, so Rocchi & Coop [65] removed the filter papers from their
oedometer tests on decomposed granites and constructed an oedometer without
drainage from the base so that the final water contents could be measured with greater
accuracy.
Different calculation methods and combinations or unit weights and water contents
have therefore been used in different soils, apparatus and preparation techniques, but by
calculating the specific volume by at least two ways, an assessment of the accuracy
may be made, for example by calculating the differences between individual value of
specific volume and the mean. This type of technique has been used for most
transitional soils. The advantage is that there is a positive proof of the accuracy of the
specific volumes that may be compared to differences in NCL or CSL, but the
disadvantage is that the accuracy tends to be worse than when it is simply estimated
from individual measurements. This may be because of inaccurate assumptions or an
inaccurately measured value, for example specific gravity, that give errors that
propagate throughout the calculations and bias the comparisons. This type of error
might not be known if only one method is used, since all the values calculated would
simply shift in one direction.
The resulting maximum estimated errors were, for example, about 0.02 for the
triaxial tests of Shipton & Coop [47], tests with greater errors being discarded. The
mean estimated error was about 0.005. Both values are small relative to the very large
differences of specific volume at critical state on Fig.18. The accuracy achieved will
depend on the test and soil type, so Ponzoni et al. [54] estimated an accuracy of only
0.05 for the oedometer tests they analysed, but a better value of less that 0.01 for the
triaxial tests. However, neither value affects significantly the reliability of the
conclusions about the degree of convergence observed.
5.4. The Influence of Anisotropic Fabrics
It might be suggested that the transitional behaviour could result from an anistropic
fabric and that given that Nougier-Lehon et al. [37] had shown with DEM that
extremely large strains may be required to reach a unique critical state fabric, that
perhaps the same might be the case for transitional soils. However, as shown in Fig.33,
there is no evidence that continued shearing would give any significantly improved
convergence of ultimate states. On Fig.30 there is no clear anisotropic fabric and the
strains during isotropic compression of the 75% quartz sand - 25% kaolin soil were
predominantly isotropic, beyond the initial loading stages. While it may be important in
some soils, it therefore seems unlikely that difficulty in reorienting an anisotropic
fabric is the cause of all transitional behaviour. As discussed above, heterogeneous
fabrics can be particularly robust in natural soils, and this may also be the case in
transitional soils.
5.5. Terminology
While there may be discussions about mechanics of transitional soils and whether
they do constitute a separate category of behaviour, there is fairly universal agreement
about the unsatisfactory nature of the name. The question that arises is transitional
between what? Since this type of behaviour has now been seen for such a wide range of
soils that are gap graded, well graded, mixed mineralogy, with and without particle
breakage, with plasticity and with none, the answer now seems unclear. If the mode of
behaviour is considered rather than the soil type then it might be said that they are
intermediate or transitional between a conventional clean sand mode of behaviour with
yield and particle breakage at higher pressures leading to a unique NCL and a
steepening CSL and plastic clays for which usually parallel NCLs and CSLs may be
defined from very low stress levels. Instead, transitional soils tend to give gradually
steepening compression paths, with indistinct yield, so that large volumetric strains
occur without much convergence of the compression paths.
The term transitional also risks confusion with the Transitional or Limiting Fines
content that is often used when fines are added to sands to denote a change from coarse
to fine dominated behaviour (e.g. [66, 67]) or an intermediate behaviour identified in
the ring shear apparatus by Lupini et al. [68], where there was a transition from well-
defined residual states for plastic clays towards no residual states and continued
shearing at a critical state for sands. It is possible that these three uses of the term
transitional are related, but as yet it is not known how and the duplicate term just
tends to cause confusion.
6. Quantification of Transitional Behaviour
The 75% quartz sand - 25% kaolin mixture gave a gradient of 0.52 on Fig.32, which
investigated the influence of the initial specific volume and the CSL intercept . In
comparison the 75% quartz sand - 25% crushed quartz silt mixture gave a gradient of
0.97. Similar differences can be seen in the degrees of convergence during compression.
For example, some give almost parallel compression paths, such as the Po River
sediments (Nocilla et al. [49]; Figs.16 & 17), the carbonate sand kaolin mixes
(Shipton & Coop [1]; Fig.22), the Thanet sand (Shipton & Coop [1]; Fig.24) and the
fractal gradings of the sands (Altuhafi & Coop [58]; Fig.28b). Others give compression
paths that show much more convergence, even if this often occurs at rather too high
pressures and too low specific volumes for there to be the possibility to define a NCL,
for example the Venice Lagoon sediments (Ponzoni et al. [54]; Figs.25, 26), or the
quartz sand - crushed quartz silt mixture at extremely high pressures (Shipton & Coop
[47]; Fig.14). The latter case of the quartz sand - crushed quartz silt mixture indicates
that there may be more convergence in the oedometer tests at extremely high stress
levels than is evident in triaxial tests at more modest stresses.
Figure 34. Examples of calculation of p for intact samples of Venice Lagoon sediments (Ponzoni et al. [54]).
To create a useful framework these differences need to be quantified. One means

of doing this is to use the gradient of the initial specific volume : graph (e.g. Fig.32),
and Ponzoni et al. [54] have denoted this gradient the parameter p. Three examples are
shown in Fig.34, which include the data from Fig.27. For each triaxial test a separate
value of has been calculated, as in Fig.32, by assuming that the CSL for each passes
through the test end point and only the gradient, is taken from Fig.27. The resulting
plots give well-defined straight lines, with only slight data scatter. The values of p are
then plotted against the clay fraction on Fig.35. Higher values of p represent less
convergence and less disruption to the effect of the initial specific volume through the
isotropic compression followed by shearing that the samples have undergone. A unique
CSL for all initial specific volumes would give p=0. The values of p for the Venice
Lagoon sediments therefore represent substantial degrees of transitional behaviour,
which tend in this case to increase with plasticity, in contrast to comparison between
the quartz sand kaolin and quartz sand crushed quartz silt mixtures, which
suggested less transitional behaviour for greater plasticity.
Figure 35. Variation of p with clay fraction for Venice Lagoon sediments (Ponzoni et al. [54]).
An alternative means of quantifying convergence is from the oedometer test data

as illustrated schematically in Fig.36. The initial specific volumes, here taken as those
at 20kPa vertical stress, v20, are plotted against the values at the maximum stress level,
which is 6000kPa for the Venice Lagoon sediments, v6000. The resulting plot, an
example of which is given in Fig.37 again gives a clear linear relationship, the gradient
of which Ponzoni et al. [54] defined to be m. The value of m would be 0 if there were a
unique NCL and 1 if there were completely parallel compression curves. The value of
0.40 on Fig.37 therefore represents a moderate degree of transitional behaviour. In
Fig.37 the oedometer test data for intact samples of Malamocco soils with CF=20-30%
from Fig.26 have been used, again discarding some anomalous tests.
Figure 36. Schematic diagram showing calculation of m from oedometer tests (Ponzoni et al. [54]).
Figure 37. Calculation of m value for oedometer tests on intact samples of Venice Lagoon sediments in
Fig.26 (Malamocco CF=20-30%) (Ponzoni et al. [54]).
Figure 38. Comparison of m values for intact and reconstituted specimens of Venice Lagoon sediments
(Ponzoni et al. [54]).
On Fig.38 the values of m have been correlated with the clay fraction, but a
problem is evident in that the values for the reconstituted samples are lower than those
for the intact and also show a different trend. While for the intact samples m is
approximately constant with CF, for the reconstituted it has a maximum at about a CF
of 15-25%. The reason for the different values is evident from Fig.36. The intact
samples, although mostly normally consolidated in situ, had all been loaded from low
initial effective stresses in the oedometer tests, and so were overconsolidated, which
will have the effect of reducing the range of v20 values and so increasing m, compared
to the reconstituted samples which were all normally consolidated. To compensate for
this the reconstituted samples were then taken through the same stress cycle and each
was loaded to an estimate of the in-situ vertical effective stress in the oedometer, before
being unloaded back to 20kPa, now taking the v20 values at this new starting point to
calculate new values of m referred to as munload. The resulting comparison of the values
of m for the intact and reconstituted samples on Fig.39 show much better agreement,
both being constant with CF and having similar values for the incremental load, IL, and
constant rate of strain tests, CRS, and also for the two sites, Malamocco and Lido.
Figure 39. Comparison of m values for intact and reconstituted specimens of Venice Lagoon sediments
accounting for overconsolidation (Ponzoni et al. [54]).
The comparison of Figs.35 and 39 reveals another apparent discrepancy, which is

that the value of p on Fig.35 are generally larger than those of m on Fig.39. It is well
known for natural soils that different forms of structure may require different
destructuration sensitivities to shear and volumetric strains (e.g. [24, 57]), and the same
could be expected from transitional soils. However, it might be expected that the values
of m should be larger than the values of p since both the shear and volumetric strains
should be greater in shearing to the CSL for a normally consolidated soil than in
compression along the one-dimensional NCL. The problem arises from the different
stress levels for the oedometer and triaxial tests and when the values of m are
recalculated at an effective stress of 300kPa, which is similar to the mid-range for the
CSLs, the values of m for the oedometer tests are much higher (Fig.40). Provided that
munload is used for the reconstituted samples the agreement with the intact samples is
again good.
Figure 40. Values of m for intact and reconstituted specimens of Venice Lagoon sediments recalculated for a
stress level of 300kPa.
The change of m value with stress level then offers another means by which the
degree of transitional behaviour may be assessed. This is shown for each of the
categories of the Venice Lagoon sediments in Fig.41. An average of the data points for
the two sites and for the IL and CRS tests is shown for clarity. In this case the different
CF show quite similar decay rates for m. That reduction of m emphasises that increased
stress level is causing greater convergence of the compression curves through the
strains applied to the soil, but for all CF values that convergence is too slow for a
usable NCL to be defined before approaching v=1.
Figure 41. Variation of m with stress level for the Venice Lagoon sediments.
It is probable that the values of p for the CSLs will also change with stress level, as
the CSLs are likely to show some convergence at very high stress levels, but these data
are not available for the Venice Lagoon sediments. Clearly the assessment of
convergence by this means is much more easily done with oedometer tests than with
triaxial tests, since only a few oedometer tests that are loaded to high stress levels are
needed for each soil, while to make the same assessment as a change of p with stress
level would require series of tests for different initial specific volumes and different
stress magnitudes, requiring also triaxial testing at very high stresses.
7. Transitional Behaviour in Tailings
Numerous tailings disasters have been provoked by the liquefaction of tailings, for
example at Mufulira, Zambia, in 1970, Stava, Italy in 1985 and Marriespruit, South
Africa, in 1994 [69, 70, 71]. The typical gradings of tailings would cause suspicion
about whether they might be subject to transitional behaviour, and for this reason when
Carrera et al. [72] studied the static liquefaction behaviour of the Stava tailings they
made an examination for all gradings of whether any transitional behaviour could be
found.
Figure 42. Oedometer compression data for Stava tailings of two controlled gradings (Carrera et al. [72]).
(a) stress paths for undrained shearing,
(b) critical state line and zones of behaviour

Figure 43. Static liquefaction of pure sand of Stava tailings (Carrera et al. [72]).
The Stava tailings are tailings of fluorite and have a predominantly siliceous
mineralogy. To investigate the effects of grading Carrera et al. [72] created samples
with various ratios of fine sand to silt, and to check for transitional behaviour a wide
range of initial specific volumes was created, but for both gradings shown in Fig.42
and others not shown, the oedometer compression curves converged to unique NCLs.
Also the CSLs for all gradings were unique, an example of which is shown in Fig.43.
Carrera et al. found that the CSLs all had clear horizontal asymptotes in the v:lnp' plane
(or here void ratio e: lnp') and that static liquefaction occurred for undrained shearing if
the initial void ratio lay above the asymptote, which could only be achieved at low
confining stresses. At higher stress levels the compression of the sediments gave lower
initial void ratios, bringing the initial state below the horizontal asymptote and there
was a strongly strain softening behaviour but no complete liquefaction. At even higher
stress levels the behaviour was predominantly strain hardening. Bedin et al. [73]
describe a very similar framework for Brazilian gold tailings. Although neither the
Stava tailings nor the gold tailings showed transitional behaviour for any grading, this
framework does illustrate the possible importance of transitional behaviour in tailings
liquefaction, because if the CSL were not unique then it would be very important at
what density the tailings were placed as the initial specific volume might control the
location of the CSL and hence this horizontal asymptote.
100
Percentage Passing (%)
80
60
40
20 clayey silt
silt (Carrera et al. [72])
sand (Carrera et al. [72])
0
0.0001 0.001 0.01 0.1 1 10 100
Particle Size (mm)
Figure 44. Gradings of some of the Stava Tailings.
In some cases clear, unique CSLs or steady state lines have been identified for
tailings. For example Fourie & Papageorgiou [64] found unique lines for all the
gradings they tested, but for Chang et al. [74] and Al-Tarhouni et al. [75] the evidence
of a unique line CSLs and NCLs was less clear. The gradings of the Stava tailings
sands and silts that were combined in various proportions by Carrera et al. [72] are
given in Fig.44. Also shown are gradings for some finer clayey silt Stava tailings, the
oedometer compression data from which are given in Fig.45, and which indicate that
for the finest of these fluorite tailings the convergence of the compression curves is
poor.
Specific Volume, v
1.8
1.6
1.4
1.2
1 10 100 1000 10000
v' (kPa)
Figure 45. Oedometer tests on clayey silt from Stava.
2.4
2.2
Specific volume, v
USA, isotropic
compression
1.8
1.6
1.4
1.2 Zambia, oedometer

Democratic Republic of Congo, oedometer
1
1 10 100 1000 10000 100000
v' or p' (kPa)
Figure 46. Isotropic compression and oedometer tests on copper tailings
In Fig.46 a selection of oedometer and triaxial isotropic compression data are

given for copper tailings. In this case the USA copper tailings, also with a clayey silt
grading gave the highest initial specific volumes and a clear convergence in isotropic
compression, while the Zambian and Democratic Republic of Congo (DRC) tailings
gave no clear NCL in one-dimensional compression. In triaxial tests the Zambian
copper tailings again gave no clear unique isotropic NCL and the critical states were
scattered, indicating no unique CSL (Fig.47).
Specific volume, v
1.8
1.6
1.4
Isotropic Compression
Undrained Shearing
end of shearing
1.2
1 10 100 1000 10000
p' (kPa)
Figure 47. Isotropic compression and undrained shearing of Zambian copper tailings.
2.2
Specifiv Volume, v
1.8
1.6
1.4
1.2
10 100 1000 10000
v' (kPa)
Figure 48. Oedometer tests on Canadian gold tailings.
Figure 48 shows the oedometric compression curves for Canadian gold tailings,
the gradings of which are given in Fig.49. Again the convergence of the curves is very
slow, with large volume changes occurring without indicating complete convergence,
and indicating some degree of transitional behaviour. It therefore seems that some
tailings may indeed give transitional behaviour, while others do not and whether they
do or not is not related to grading.
100
80
Percentage Passing (%)

60
40
20 USA Copper Tailings

DRC Copper Tailings
Canada Gold Tailings
0
0.0001 0.001 0.01 0.1 1 10 100
Particle Size (mm)
Figure 49. Gradings of various tailings.
8. Conclusions
After a seminar I gave on transitional soils, a practicing engineer asked the question
But I dont really need to worry about this do I? As long as I test intact samples Ill
still get the right answer, wont I?. And of course he was absolutely right, for many
engineering applications there are few implications of transitional behaviour. The key
area where it does impact our understanding of soil behaviour is in quantifying the
effects of structure through comparisons of intact and reconstituted samples highlighted
by Ferreira & Bica [9] since it implies that for many soils there may be no true intrinsic
behaviour with which to compare. One possibility would be to make the reconstituted
samples with the same initial specific volume as the intact soils. But how could that be
determined, unravelling the combined effects of structure and overconsolidation on the
current state?
Transitional soil behaviour is often seen as a threat to critical state soil mechanics,
but that need not be so. As has been shown in this paper, many soils, natural and
reconstituted, have fabrics that are robust in nature, which mean that the although they
may reach reasonable states of constant stress and volume at large shear strains, they
will not do so at a unique fabric, and it is probably robust fabrics of this nature that
cause a dependence of the final state on the initial volume for a transitional soil, since
different specific volumes can only be sustained through different fabrics. But as long
as we are prepared to suspend our requirement to have a unique fabric, each initial state
will give unique normal compression lines, critical state lines and state boundary
surfaces, and we can apply our conventional soil mechanics without difficultly. As
Shipton & Coop [47] demonstrated the critical state lines for each initial density are
really rather difficult to move with simple monotonic paths, but we also must keep in
mind that complex or vigorous loading may bring about more radical changes to a soil
fabric, such as experienced by Atkinson et al. [30] in the disaggregation of their Mercia
Mudstone when cut with a TBM. While it is attractive to believe that a critical state
with unique volume, stress and fabric can be reached for all soils, if it is so difficult to
achieve through simple monotonic loading in some soils, it ceases to be a useful

assumption.
It is frequently the case that researchers have only made reconstituted samples at
too small a range of initial specific volumes to be aware if transitional behaviour occurs
in their soil or not, so while it is important not to start seeing this mode of behaviour
everywhere, it is possible that its frequency is currently being underestimated. Since it
is so difficult to predict from simple factors such as grading and can be induced with
relatively small changes to the components of a soil, we should all make the simple
check for each new soil we investigate of whether oedometer tests on samples of
different initial specific volumes do converge or not. Transitional behaviour seems to
occur in a very wide range of soils of mixed grading and/or mixed mineralogy and
tailings seem particularly susceptible.
It is certainly simplistic that some soils should be transitional and others not, and
as has been shown in the paper there is a spectrum of behaviour within what might be
called transitional, from completely parallel compression and shearing behaviour to
convergent behaviour that occurs too slowly for a useful normal compression line to be
found. This underlines the need to quantify this type of behaviour and two methods
have been presented.
Acknowledgements
The work described in this paper was partially supported by a grant from the
Research Grants Council of the Hong Kong Special Administrative Region, China
(Project No. CityU 112813). The author is indebted to the following for their help in
preparing this paper; Dr Batrice Baudet, Ms Elisa Ponzoni, Ms Cristina Todisco and
Mr Li Wei.
Nomenclature
A intercept of log G : log p' graph at 1kPa

Br relative breakage
Cc compression index
CF clay fraction
CSL critical state line
dmax maximum rate of dilation
e void ratio
eg granular void ratio
ef fines void ratio
e*100/1000 void ratio at 100 or 1000kPa on intrinsic one-dimensional ICL
fc fines content
G shear modulus
ICL intrinsic compression line
ICSL intrinsic critical state line
Iv void index
m quantification of convergence of oedometer compression curves
M critical state line gradient (q'/p')
n gradient of log G : log p' graph

N intercept of isotropic NCL at 1kPa in v:lnp' plane
N0 intercept of one-dimensional NCL at 1kPa in v:lnp' plane
NCL normal compression line
OC overconsolidated
p quantification of convergence of CSLs
p' mean normal effective stress = ('a + 2'r)/3
p'e equivalent value of p' on ICL
p'cs equivalent value of p' on CSL
q, q' deviatoric stress = 'a - 'r
SBS state boundary surface
SCC sedimentation compression curve
SCL sedimentation compression line
v specific volume
vf fines specific volume
vg granular specific volume
v0 initial specific volume
v20, v6000 specific volume at 20kPa or 6000kPa
s shear strain
intercept of CSL at 1kPa in v:lnp' plane
NCL and CSL gradient in v:lnp' plane
'a axial effective stress
'r radial effective stress
'v vertical effective stress
References
[1] Shipton, B. & Coop, M.R. On the compression behaviour of reconstituted soils. Soils & Foundations, 52,
(2012) No.4, 668-681.
[2] Schofield, A. N. & Wroth, C. P. Critical State Soil Mechanics. McGraw-Hill, (1968), p. 310.
[3] Coop, M.R. & Lee, I.K. The behaviour of granular soils at elevated stresses. Predictive Soil Mechanics,
Thomas Telford, London, (1993), 186-198.
[4] Thevanayagam, S. & Mohan, S. Intergranular state variables and stress-strain behaviour of silty sands.
Gotechnique, 50, (2000), No.1, 1-23.
[5] Chen, Y.C., Ishibashi, I. & Jenkins, J.T. Dynamic shear modulus and fabric: Part 1 Depositional and
induced anisotropy. Gotechnique, 38, (1988), No.1, 25-32.
[6] Oda, M. & Iwashita, K. (eds) Mechanics of Granular Materials: An Introduction. Balkema, Rotterdam,
(1999).
[7] Hattab, M. & Fleureau, J-M. Experimental study of kaolin particle orientation mechanism. Gotechnique,
60, (2010), No.5, 323-331.
[8] Fonseca, J., OSullivan, C., Coop, M.R. & Lee, P.D. Quantifying the Evolution of Soil Fabric during
Shearing using Directional Parameters. Gotechnique, 63, (2013), No.6, 487-499.
[9] Ferreira, P.M.V. & Bica, A.V.D. Problems on the identification of structure in a soil with a transitional
behaviour. Gotechnique, 56, (2006), No.7, 445454.
[10] Burland, J.B. On the compressibility and shear strength of natural clays. Thirtieth Rankine Lecture.
Gotechnique, 40, (1990), No.3, 329-378.
[11] Hosseini-Kamal, R., Brosse, A., Coop, M.R. & Jardine, R.J. The post-yield behaviour of four Eocene-
to-Jurassic UK stiff clays. Gotechnique, 64, (2014), No.8, 620 634.
[12] Cotecchia, F. & Chandler, R. J. The influence of structure on the pre-failure behaviour of a natural clay.
Gotechnique, 47, (1997), No.3, 523-544.
[13] Cotecchia, F. & Chandler, R. J. One-dimensional compression of natural clay: Structural changes and
mechanical effects. The Geotechnics of Hard Soils Soft Rocks. Evangelista & Picarelli (eds), Balkema,
Rotterdam, (1998), 103-114.
[14] Rampello, S. & Silvestri, F. The stress-strain behaviour of natural and reconstituted samples of two
overconsolidated clays. Proc. Int. Symp. Geotech. Engng Hard Soils- Soft Rocks, Athens, (1993), 769-
778.
[15] DeGroot, D.J. & Lutenegger, A.J. Geology and engineering properties of Connecticut Valley Varved
Clay. Characterisation and Engineering Properties of Natural Soils. Vol.1, Swets & Zeitlinger, Lisse:
(2003), 695-724.
[16] Coop, M.R. & Cotecchia, F. The geotechnical settlements of the archaeological site of Sibari.
Geotechnical Engineering for the Preservation of Monuments and Historic Sites. Viggiani, C. ed.,
Balkema, Rotterdam, (1997), 138-149.
[17] Fearon, R.E. & Coop, M.R. The influence of landsliding on the behaviour of a structurally complex
clay. Quart. J. Eng. Geol. & Hydrogeol., 35, (2002), No.1., 25-32.
[18] DElia, B. Deformation problems in the Italian structurally complex clays. Proc. X ECSMFE, Firenze,
Vol.4., (1991), 11159-110.
[19] Bertuccioli, P. & Lanzo, G. Mechanical properties of two Italian structurally complex clay soils. Proc.
Int. Symp. Geotech. Engng Hard Soils- Soft Rocks. Athens, Vol.1, (1993), 383-389.
[20] Picarelli, L. Discussion of: Leroueil, S. & Vaughan, P.R. (1990). The general and congruent effects of
structure in natural soils and weak rocks. Gotechnique, 41, (1991), No.2, 281-284.
[21] Guerriero, G., Olivares, L. & Picarelli, L. Modelling the mechanical behaviour of clay shales: some
experimental remarks. Colloquium on Chalk and Shales. Brussels, Vol.2, (1995), 20-30.
[22] Santaloia, F., Cotecchia, F. & Polemio, M. Mechanics of a tectonized soil slope: influence of boundary
conditions and rainfall. Quart. J. Eng. Geol. & Hydrogeol., 34, (2001), 165-185.
[23] Cotecchia, F. & Chandler, R. J. A general framework for the mechanical behaviour of clays.
Gotechnique, 50, (2000), No.4, 431447.
[24] Baudet, B.A. & Stallebrass, S.E. A constitutive model for structured clays. Gotechnique, 54, (2004),
No.4, 269-278.
[25] Cotecchia, F., Mitaritonna, G. & Vitone, C. Investigating the influence of microstructure, loading
history and fissuring on the clay response. Int. Symp. on Deformation Characteristics of Geomaterials,
Seoul, Korea, (2011), 53-83.
[26] Cuccovillo, T. & Coop, M.R. On the mechanics of structured sands. Gotechnique, 49, (1999), No.6,
741-760.
[27] Fonseca, J., OSullivan, C., Coop, M.R. & Lee, P.D. Non-invasive characterisation of particle
morphology of natural sands. Soils & Foundations, 52, (2012), No.4, 712-722.
[28] Masin, D., Stallebrass, S. E. & Atkinson, J. H. Laboratory modelling of natural structured clays. Proc
int. workshop on geotechnics of soft soils: theory and practice. VGE, The Netherlands, (2003), 491
496.
[29] Fearon, R.E. & Coop, M.R. Reconstitution - what makes an appropriate reference material?
Gotechnique, 50, (2000), 4, 471-477.
[30] Atkinson, J.H., Fookes, P.G., Miglio, B.F. & Pettifer, G.S. Destructuring and disaggregation of Mercia
Mudstone during full-face tunnelling. Quart. J. Eng. Geol. & Hydrogeol., 36, (2003), No.4, 293-303.
[31] Chu, J., Leong, W.K. & Loke, W.L. Discussion of Defining an appropriate steady state line for
Marriespruit gold tailings. Can. Geotech. J., 40, (2003), 484-486.
[32] Jefferies, M.G. & Been, K. Soil liquefaction a critical state approach. Taylor & Francis, London &
New York, (2006).
[33] Riemer, M.F. & Seed, R.B. Factors affecting apparent position of steady-state line. J. Geotech. &
Geoenviron. Eng, 123, (1997), No.3, 281-288.
[34] Chen, Y.C. & Chuang, J.C. Effects of fabric on steady state and liquefaction resistance. Proc. 11th Int.
Offshore and Polar Engng. Conf., Stavanger, (2001), 524-529.
[35] Vaid, Y.P. & Sivathayalan, S. Fundamental factors affecting liquefaction susceptibility of sands. Can.
Geotech, J., 37, (2007), 592-606.
[36] Yang, Z.X., Li, X.S. & Yang, J. Quantifying and modelling fabric anisotropy of granular soils.
Gotechnique, 58, (2008), No.4, 237-248.
[37] Nougier-Lehon, C., Vincens, E. & Cambou, B. Structural changes in granular materials: the case of
irregular polygonal particles. International J. Solids & Structures, 42, (2005), No.2425, 63566375.
[38] Wood, F.M., Yamamuro, J.A. & Lade, P.V. Effect of depositional method on the undrained response of
silty sand. Can. Geotech J., 45, (2008), 1525-1537.
[39] Yamamuro, J.A., Wood, F.M. & Lade, P.V. Effect of depositional method on the microstructure of silty
sand. Can. Geotech. J., 45, (2008), 1538-1555.
[40] Madhusudhan, B.N. & Baudet, B.A. Influence of reconstitution method on the behaviour of completely
decomposed granite. Gotechnique, 64, (2014), No.7, 540 550.
[41] Muir Wood, D. Critical states and soil modelling. Deformational Characteristics of Geomaterials. Vol.
1, IOS Amsterdam, (2008), 51-72.
[42] Cheng, Y.P., Nakata, Y. & Bolton, M.D. Grain crushing and critical states observed in DEM
simulations. Powders and Grains. Taylor and Francis, London, Vol.2, (2005), 1393-1397.
[43] Bandini, V. & Coop, M.R. The influence of particle breakage on the location of the critical state line of
sands. Soils & Foundations, 51, (2011), No.4, 591-600.
[44] Hardin, B.O. Crushing of soil particles. J. Geotech. Eng., 111, (1985), No.10, 1177-1192.
[45] Coop, M.R. The mechanics of uncemented carbonate sands. Gotechnique, 40, (1990), No.4, 607-626.
[46] Martins, B.F., Bressani, L.A., Coop, M.R. & Bica, V.D. Some aspects of the compressibility behaviour
of a clayey sand. Can. Geotech. J., 38, (2001), No6, 1177-1186.
[47] Shipton, B. & Coop, M.R. Transitional Behaviour in Sands with Plastic and Non-plastic Fines. Soils &
Foundations, 55, (2015), No.1, 1-16.
[48] Pestana, J.M. & Whittle, A.J. A compression model for cohesionless soils. Gotechnique, 45, (1995),
No.4, 611-631.
[49] Nocilla, A., Coop, M.R. & Colleselli, F. The mechanics of an Italian silt: an example of transitional
behaviour. Gotechnique, 56, (2006), No.4, 261-271.
[50] Yamamuro, J. A. & Lade, P. L. Steady-state concepts and static liquefaction of silty sands. J. Geotech.
Geoenviron. Eng., 124, (1998), No.9, 868877.
[51] Shipton, B. The Mechanics of Transitional Soils. PhD Thesis, Imperial College, London, (2010).
[52] Ventouras, K. & Coop, M.R. On the behaviour of Thanet sand: an example of an uncemented natural
sand. Gotechnique, 59, (2009), No.9, 727-738.
[53] Ventouras, K. Engineering Behaviour of Thanet Sand. PhD Thesis, Imperial College, London, (2009).
[54] Ponzoni, E., Nocilla, A., Coop, M.R & Colleselli, F. On the Identification and Quantification of
Transitional Modes of Behaviour in the Sediments of the Venice Lagoon. Gotechnique, 64, (2014),
No.9, 694 708.
[55] Altuhafi, F., Baudet, B.A. & Sammonds, P. The mechanics of subglacial sediment: an example of new
transitional behaviour. Can. Geo. J., 47, (2010), No.7, 775790.
[56] Altuhafi, F., Baudet, B.A. & Sammonds, P. On the particle size distribution of a basaltic till. Soils &
Foundations. 51, (2011), No.1, 113-121.
[57] Baudet, B.A. & Ho, E.W.L. On the behaviour of deep ocean sediments. Gotechnique, 54, (2004), No.9,
571-580.
[58] Altuhafi, F. & Coop, M.R. Changes to particle characteristics associated with the compression of sands.
Gotechnique, 61, (2011), No.6, 459-471.
[59] Coop, M.R., Sorensen, K.K., Bodas Freitas, T. & Georgoutsos, G. Particle breakage during shearing of
a carbonate sand. Gotechnique, 54, (2004), No.3, 157163.
[60] Vilhar, G., Jovicic, V. & Coop, M.R. The role of particle breakage in the mechanics of a non-plastic
silty sand. Soils & Foundations, 53, (2013), No.1, 91-104.
[61] Klotz, E.U. & Coop, M.R. On the Identification of Critical State Lines for Sands. Geotech. Testing J.,
25, (2002), No.3, 289-302.
[62] Garga, V.K. & Zhang, H. Volume Changes in Undrained Triaxial Tests on Sands. Can. Geotech. J., 34,
(1997), 762772.
[63] Jefferies, M.G. & Been, K. Implications for Critical State Theory From Isotropic Compression of Sand.
Gotechnique, 50, (2000), No.4, 419429.
[64] Fourie, A.B. & Papageorgiou, G. Defining an appropriate steady state line for Merriespruit gold tailings.
Can. Geotech. J., 38, (2001), No.4, 695706.
[65] Rocchi, I. & Coop, M.R. Experimental accuracy of the initial specific volume. Geotech. Testing J., 37,
(2014), No.1, 169175.
[66] Pitman, T.D., Robertson, P.K. & Sego, D.C. Influence of fines on the collapse of loose sands. Can.
Geotech. J., 31, (1994), 728-739.
[67] Polito, C.P. & Martin, J.R.II. Effects of Nonplastic Fines on the Liquefaction Resistance of Sands. J.
Geotech. & Geoenvironmental Eng., 127, (2001), No.5, 408-415.
[68] Lupini, J.F., Skinner, A.E. & Vaughan, P.R. The drained residual strength of cohesive soils.
Gotechnique, 31, (1981), 181-213.
[69] Blight, G. E. & Fourie, A. B. A review of catastrophic flow failures of deposits of mine waste and
municipal refuse. Proceedings of the international conference FSM2003, Naples, (2003), 117.
[70] Chandler, R.J. & Tosatti, G. The Stava tailings dams failure, Italy, July 1985. Proc. Instn Civ. Engrs,
Geotech. Eng., 113, (1995), No.2, 6779.
[71] Colombo, P. & Colleselli, F. The collapse of the Stava tailings basins analysis of the event and
observations. Int. Workshop on Occurrence and Mechanisms of Flows in Natural Slopes and Earthfills.
Sorrento, Italy (ed. L. Picarelli), Bologna: Patron, (2003), 8796.
[72] Carrera, A., Coop, M.R. & Lancellotta, R. The influence of grading on the mechanical behaviour of
Stava tailings. Gotechnique, 61, (2011), No.11, 935946.
[73] Bedin, J., Schnaid, F., Da Fonseca, A.V., & Costa Filho, L. De M. Gold tailings liquefaction under
critical state soil mechanics. Gotechnique, 62, (2012), No.3, 263267.
[74] Chang, N, Heymann, G. & Clayton, C. The effect of fabric on the behaviour of gold tailings.
Gotechnique, 61, (2011), No.3, 187197.
[75] Al-Tarhouni, M., Simms, P. & Sivathayalan, S. Cyclic behaviour of reconstituted and desiccatedrewet
thickened gold tailings in simple shear. Can. Geotech. J., 48, (2011), 10441060.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-156
Shear strength and stiffness anisotropy of

geologically aged stiff clays
Richard JARDINEa1, Amandine BROSSEb, Matthew COOPc and Ramtin HOSSEINI
KAMALd
a
Imperial College London, UK
b
Geotechnical Consulting Group, London, formerly Imperial College London, UK
c
Hong Kong City University, Hong Kong
4
DNV.GL, London, formerly Imperial College London, UK
Abstract. This paper considers the deformation behaviour of four geologically

aged, medium-plasticity, heavily overconsolidated stiff clays that affect a broad
swathe of infrastructure projects in the SE of the United Kingdom. Static triaxial
and hollow cylinder stress path experiments on high quality samples are examined
along with dynamic multi-axial bender element and resonant-column
measurements. Patterns of undrained shear strength anisotropy are revealed that
are governed by the clays meso and micro-structures. The clays are brittle in shear
and their stiffness characteristics are shown to be markedly anisotropic, highly
non-linear and pressure dependent. The results obtained have many implications
for practical geotechnical engineering.
Keywords. Clays, anisotropy, stress path tests, shear strength, stiffness
1. Introduction
A large proportion of the southern United Kingdom (UK) is underlain by Triassic to

Eocene stiff clays and mudrocks. The planning of nationally important construction
projects, including a new high-speed rail line and other economically important
facilities that traverse the full succession of these strata, prompted a characterization
study on the stiff, medium-plasticity heavily overconsolidated, Jurassic-to-Eocene,
London, Gault and Kimmeridge and Oxford clays. The research at Imperial College
encompassed six doctoral studies and led to ten primary references, [1] to [10], that we
draw on to summarize the anisotropic and pressure-dependent shear strength and non-
linear stiffness characteristics revealed by advanced laboratory testing.
The research focused on geological units that had been deposited in broadly
similar depositional environments. The site locations identified in Figure 1 were chosen
to reduce the potential complications of coastal action and tectonic disturbance. High
quality samples were taken by block and Geobor-S rotary techniques at the Heathrow
(London clay), High Cross (Gault clay), Willowbrook Farm (Kimmeridge) and Elstow
(Oxford clay) geotechnical sites, where in-situ penetration and seismic testing was also
performed. The profiles reported for London clay in [3] and summarized for the other
three clays in Figures 2 to 4 show in-situ shear strengths and shear wave velocities that
1
Corresponding Author. Professor of Geomechanics at Imperial College, London.
R. Jardine et al. / Shear Strength and Stiffness Anisotropy of Geologically Aged Stiff Clays 157
varied significantly with depth, due to effective stress increases, weathering and
occasional cemented hard bands. The experiments reported here concentrated on
samples taken from around 10m below the clays upper surfaces, where the effects of
post-depositional weathering, glacial and tree action were considered to be less
significant.
Figure 1. London Heathrow Airport Terminal 5 (London clay), High Cross (Gault clay), Willowbrook Farm
(Kimmeridge) and Elstow (Oxford clay) sampling locations in South East United Kingdom (UK), after [7].
Tables 1 and 2 summarise the index properties, in-situ stresses and soil fabrics
applying over the depth range of present interest at each site, listing water contents that
fell below the clays respective plastic limits. The London clay specimens from
Heathrow were taken and tested at higher effective stresses than the other clays because
they had been covered by 6m of River Terrace gravels and came from a greater total
depth below ground level. An earlier paper [7] reported parallel suites of oedometer
and triaxial compressive shear tests on reconstituted and high quality natural samples
158 R. Jardine et al. / Shear Strength and Stiffness Anisotropy of Geologically Aged Stiff Clays
and showed that that meso-structure has the most important influence on the natural
clays yielding behaviour. The Oxford clay samples had clear bedding features, but no
fissures, and developed markedly higher triaxial compression shear strengths and post-
peak brittleness. All four soils were brittle in shear and developed low residual
angles in ring shear tests. The peak triaxial shear strengths of natural samples of the
Gault, Kimmeridge and London clays were strongly affected by their greater fissure
intensity. The Su values listed in Table 1 are taken from tests on 100mm high quality
specimens that had been sheared relatively slowly after reconsolidation to in-situ
stresses. The listed K0 values, which were assessed from geological history and suction
measurements could not be applied in laboratory testing without causing extension
failures, so the test K values were reduced to 1.75 0.05 for all four clays.
Water Plastic Liquid Plasticity Triaxial Residual

Soil content Limit Limit Index SU '
[%] [%] [%] [%] [kPa] [Deg]
Oxford clay 24-27 32-35 64-67 30-34 236 10
Kimmeridge clay 16-21 18-22 46-55 28-33 85 7
Gault clay 28-30 28-31 71-77 40-46 87 10
London clay
23-25 26-27 68-70 41-44 108-175* 12
(Unit B2c)
Table 1. Index properties for four clays investigated at sites shown in Figure 1. *Note London clay range
reflects whether the specimen contained an optimally aligned pre-existing fissure.
In-situ stresses
Clay Age Micro-structure Meso-structure
K0 p'0 'v0-'h0
Mya kPa kPa

Upper Strong horizontal No significant fissuring but
Oxford
Jurassic 3.2 250 -223 orientation of particles highly bedded. Horizontal
Clay
161-156 and shells shell beds
Upper No strong preferred Close spaced fissuring (2-

Kimmeridge
Jurassic 1.8 186 -97 orientation due to 5cm spacing) sub-horizontal
clay
156-151 presence of larger silt & 40- 60 to vertical
particles
Close spaced fissuring

No strong preferred
Lower (spacing of 2-5cm) sub-
orientation due to
Gault clay Cretaceous 1.8 163 -85 horizontal and sub-vertical
presence of larger
112-99 with frequent zones of
micro-fossils
intense fissuring. Presence
of nodules
Highly fissured (spacing
London
Eocene Moderate particle typically 10-20cm) in sub-
clay 1.77 323 -165
56-49 orientation vertical and sub-horizontal
(Unit B2c)
directions
Table 2. Geological ages, estimated in-situ stresses, micro and meso-structures of the four stiff clays.
Figure 2: Profile for Gault Clay at High Cross, near Cambridge, showing CPT profile for tip (qc) and sleeve
friction (fs) and shear modulus Gvh traces obtained for the current research by seismic CPT measurements.
Also shown are Ghh, Ghv and Gvh profiles obtained in earlier down-hole and cross-hole seismic testing at the
same site. Nominal undrained shear strength (Su) profiles are also shown from earlier Unconsolidated
Undrained (UU) triaxial tests on 100mm and 38mm diameter samples taken by thin walled tube sampling and
a further Self Boring Pressuremeter SBPM Su profile. Note that the Su trends depend on the sampling testing
technique and do not match exactly those from the triaxial and HCA tests described herein. After [10]; see
[7] and [8] for full references to earlier research at this site.
Figure 3: Profile for Kimmeridge Clay at Willowbrook Farm, near Abingdon, showing CPT profile for tip
(qc) and sleeve friction (fs) and shear modulus Gvh traces obtained for the current research by seismic CPT
wave velocity measurements.
Also shown is a nominal undrained shear strength (Su) profile from Marchetti Dilatometer (DMT) test carried
out at the same site for the present research. Note that the Su trends may not match exactly those from the
laboratory tests described herein. After [10]; see [7] and [8] for full references to earlier research at this site.
Figure 4: Profile for Oxford Clay near Elstow, showing CPT profile for tip (qc) and sleeve friction (fs) and
shear modulus Gvh traces obtained for the current research by seismic CPT wave velocity measurements.
Also shown is a Ghh profile obtained in earlier Self Boring Pressuremeter SBPM testing. Nominal undrained
shear strength (Su) profiles are also shown from earlier Unconsolidated Undrained (UU) triaxial tests on
100mm diameter samples in a nearby commercial site investigation. Note that the Su trends do not match
exactly those from the triaxial and HCA tests described herein. After [10]; see [7] and [8] for full references.
2. Apparatus and Procedures
2.1 Hollow Cylinder Apparatus (HCA) tests
The four clays anisotropic shear strength and non-linear stiffness behaviours
were explored in undrained experiments with two HCA sets that tested specimens with
outside diameters of 72 and 100mm and heights of 190 and 200mm respectively [2],
[5], [9]. The specimens were prepared from high-quality samples by careful trimming
and reaming and were all reconsolidated to the in-situ p levels expected in-situ at
10m depth and K 1.75 0.05. Most tests employed controlled major principal stress
increment direction (d) shearing paths (see [2] and [5] for further explanation),
although Simple Shear (SS) HCA tests were also conducted. The d path tests
involved changing b = (2 3)/(1 3) prior to shearing under undrained conditions
to 0.5 while keeping p and q constant, as indicated in Figure 5. After allowing 12 hours
for creep, undrained shearing was initiated with constant p and b while imposing for
each test a specified major principal stress axis increment direction, d that led to the
desired final failure value of f the inclination of the 1 axis from the vertical. The SS
tests were conducted under undrained conditions with controls that kept the specimens
axial, radial and circumferential strains equal to zero while applying shear straining in
the vertical plane. Non-destructive small strain resonant column tests were conducted
in one of the HCA equipment throughout testing to monitor how Gz varied during
reconsolidation and shearing.
Figure 5. Scheme of stress paths followed during shearing from in-situ stresses in HCA tests, after [5].
2.2 Triaxial testing
Most triaxial tests involved hydraulic Imperial College stress path cells equipped
to take either 38 or 100mm diameter specimens with 2:1 height to diameter ratios.
Recognising the importance of bedding and fissuring, the undrained tests that
investigated the shearing resistance of natural samples were conducted principally on
100mm diameter specimens, [1], [7]. Consideration of the extended test durations led
to the very small strain drained stiffness anisotropy probing tests being concentrated on
the more rapidly draining 38mm diameter samples, as were the reconstituted clay
triaxial experiments [1], [6]. All the apparatus employed local axial strain (mostly with
LVDT) sensors. As described in [1], [6] and [8], most were fitted with either platen-
mounted or with a lateral T-configuration bender element systems, along with mid-
height pore pressure transducers to track local pore pressure generation and dissipation.
The natural and reconstituted samples were prepared and tested following a variety
of effective stress paths that included anisotropic reconsolidation to K = 1.750.05 and
isotropic compression (or swelling) followed by undrained shearing or drained probing
tests. The critical importance of allowing samples to rest under drained conditions after
completing all pore pressure dissipation during reconsolidation was recognised and all
shearing stages were delayed until drained axial creep rates had fallen to rates lower
than 5 x 10-5 %/hr. The shearing rates in undrained tests were 0.02 %/hr, while drained
tests were conducted sufficiently slowly to dissipate excess pore pressures.
High sensor resolution and stability are required to track soils stress-strain
responses from their (very limited) pseudo-elastic ranges through to ultimate failure.
Even when the standard deviations in strain measurements fall below 10 -6 and those for
stresses below 0.05kPa, multiple readings and averaging are required to establish initial
stiffness trends. Highly flexible stress-path control systems are also essential. Multiple
bender element measurements were made throughout testing of the Vhh and Vhv shear
wave velocities developed by horizontally and vertically polarised horizontally
travelling waves, allowing the Ghh and Ghv shear stiffnesses to be tracked continuously.
3. Experimental characterization of shear strength anisotropy
We discuss below the shear strength anisotropy manifested in the undrained HCA tests
before considering the pre-failure deformation behaviours of the four clays.
3.1 Shear strength anisotropy revealed by undrained HCA tests
Undrained shear strength anisotropy was investigated in 32 HCA tests that brought
specimens to failure from K = 1.75 in-situ conditions following stress paths that
maintained nominally plane strain conditions with b = 0.5, while rotating towards
failure as indicated in Figure 5. The shear strengths can be compared with the ( = 0, b
= 0) triaxial compression strengths listed in Table 1 determined in undrained tests
involving the same sample quality, reconsolidation approach and shear strain rates.
The four brittle plastic overconsolidated clays all reached their peak HCA shear
strengths after developing shear discontinuities [2], [5] and [9] and did not manifest
any extended period of dilative behaviour. Their Su values depended principally on the
effective stress state under which the envelope is first engaged, and was not uniquely
related to water content or void ratio, [13]. Imposing the correct effective stress and K 0
conditions is therefore critical when attempting to measure the undrained shear strength
anisotropy that would apply to such clays in the field. It is generally not possible to
impose such conditions in triaxial compression tests on inclined samples.
Figure 6a. Variation with of Su at first rupture and ultimate peak strength states for Gault clay, after [5].
Figure 6b. Variation with of Su at first rupture and ultimate peak strengths for Kimmeridge clay, after [5].
Figure 6c. Variation with of Su at first rupture and ultimate peak strength states for Oxford clay, after [5]
We examine the outcomes first in total stress terms, considering the clays as if they
were Tresca materials whose shear strength is independent of p. Figures 6a to c detail
the variations with of peak undrained shear strength, Su, in the Gault, Kimmeridge
and Oxford clay tests, covering both the sets of five d tests (failing at f angles of 0,
23, 45, 67 and 90 degrees) and the single Simple Shear (SS) test conducted for each
clay from in-situ effective stress conditions. Similar plots are given for the London clay
in [2]. The figures all highlight the peak shear strengths developed. Also shown fitted
are curves relating Su and that follow the empirical equations proposed by Bishop,
where Su0 is the undrained shear strength applying when = 0:
Su () = Su0 (1 a sin2)(1-b sin2(2)) (1)
A summary is made in Figure 7 of the four clays trends with of peak Su = q/2. It
is evident that the Gault and Kimmeridge specimens developed generally lower
undrained shear strengths than the Oxford and London clays, although the latter were
tested at significantly higher effective stresses: see Table 2. The key points are:
The Gault clay showed the least variation of Su with , with its nominally
plane strain (b = 0.5) peak shear strengths all exceeding the triaxial
compression and extension shear strengths. The minimum b = 0.5 shear
strength applied at 40o from the vertical and its maximum when = 90o.
The Kimmeridge clay developed a similar = 0 o, b = 0.5 Su value to the Gault,
but manifested its minimum at lower values (between 0 o and 20) which fell
close to the triaxial compression peak Su and its maximum, which was almost
twice as high, at = 90o.
The Oxford clay developed its b = 0.5 Su minimum with around 45o and a
maximum that was again around twice as high at = 90o. In this case the
triaxial compression Su was substantially higher than the b = 0.5 shear
strengths developed at < 90o while the Oxford clay triaxial extension
strength was notably low.
As shown on Figure 7 the London clay trend was broadly similar to that for
Oxford clay. Although the London clays greater intensity of fissuring led to a
wider range of triaxial Su values (108 to 175 kPa) and greater overlap with the
b= 0.5 HCA Su range.
The four clays showed b = 0.5 Su minima with failure mechanisms and 1
orientations with respect to the vertical (with 10o < f < 60o) that could best exploit the
specimens pre-existing meso-fabrics (fissures and bedding features). They also all
developed their Su maxima at = 90o. The most intensively fissured Gault showed the
least variation between these limits, while the London and Oxford clays showed the
most marked Su anisotropy.
Another key factor that contributes to the significant variations of Su with is the
systematic trend for the p' and s' values applying at failure to increase with because
the pre-failure effective stress paths have negatively inclined d(z-)/d(2p') gradients.
The latter reflect the clays oriented micro-structures and the pronounced stiffness
anisotropy that we report later in this paper (see [2], [5] and [9]).
It is also interesting that the Simple Shear (SS) HCA tests, which failed at around
55o and b 0.6, gave relatively low Su values that fit well with the d experiments
anisotropic Su trends and fell close to the b = 0.5 Su minima.
Figure 7. Variation with of Su at ultimate peak strength states for all four clays, after [5].
3.2 Peak shear strength anisotropy in terms of mobilized t/s ratios
The conditions identified in undrained tests from in-situ stress conditions are
reported for Gault, Kimmeridge and Oxford clays in Figures 8a to c by plotting t/s'
against and showing at the right hand side equivalent values of ' = sin-1 t/s'. The t/s'
minima develop with 45 < < 70o and the maxima at = 0, rather than the 90o
orientation applying to Su, reflecting the higher effective stresses developed by samples
taken to failure at larger values, as discussed above. The figures also show the ratios
applying in undrained triaxial compression and extension tests.
The mobilised HCA ' angles found with b = 0.5 are all relatively high. The
specimens peak t/s' = sin ' ratios vary with stress level because the failure envelopes
associated with each value are curved, with t/s' dipping as s increases.
Detailed analysis in [5] and [9] indicate that the two = 0 Oxford clay tests
developed final circumferential '3 values that were just below zero (tensile). Although
impossible under ordinary triaxial conditions, this can apply in HCA tests conducted
with b > 0. The Oxford clay = 22.5o test also failed close to the no tension line. As
summarized in Table 2, the Oxford clay samples meso-fabrics were unique in
displaying primarily bedding and no fissuring; these features enhanced its shear
strength at low values.
Figure 9 presents an overall summary of the peak t/s' trends, including the equivalent
data from the earlier reported London clay HCA tests [2]. As noted in the Su plots, the
SS experiments led to peak t/s' points that were compatible with the trends given by the
d experiments conducted over the same range.
Figure 8a. Variation with of t/s at ultimate peak strength states for Gault clay, after [5].
Figure 8b. Variation with of t/s at ultimate peak strengths for Kimmeridge clay, after [5].
Figure 8c. Variation with of t/s at ultimate peak strength states for Oxford clay, after [5].
Figure 9. Variation with of t/s at ultimate peak strength states for all four clays, after [5].
4. Experimental investigation of the four clays stiffness behaviour
We explore below the four clays pre-failure deformation behaviour, focusing on

stiffness anisotropy, pressure dependency and non-linearity over the main range of
engineering interest, reviewing first the general framework within which the tests were
planned and the results interpreted before considering the key experimental outcomes.
4.1 Framework for interpretation of stiffness characteristics

Soil deformation behaviour can usually only be considered elastic within a very
small area of the available effective stress space. This region is limited by a kinematic
hardening (Y1) true yield surface; see [11], [12], [13] or [14]. The Y 1 yield surface is
dragged with the current effective stress point, growing and shrinking with p and
changing in shape with proximity to the outer Y 3 surface, which corresponds to the
yield surface recognised in classical critical state soil mechanics. Behaviour within the
true Y1 yield surface can be highly anisotropic. The anisotropy seen in sands and soft
clays evolves as K changes, although such changes appear to be less significant with
stiff high OCR natural clays.
Plastic straining commences once the Y1 surface is engaged and becomes
progressively more important as straining continues along any monotonic path. An
intermediate kinematic Y2 surface was identified that marks: (i) potential changes in
strain increment directions, (ii) the onset of marked strain-rate or time dependency and
(iii) a threshold condition in cyclic tests beyond which permanent strains (or p
reductions in constant volume tests) accumulate significantly. The Y 1 and Y2 surfaces
of London clay specimens were tracked in experiments conducted over the 50m deep
Heathrow Terminal 5 profile, confirming that both surfaces scale linearly with
increasing p, that behaviour was anisotropic within Y 1 and that the Y2 surface marked
also the limit to the region within which recent incoming stress excursions had no
effect on the behaviour developed on common outgoing stress paths; see [1].
It is recognised that strain rate affects stiffness and that the Y 1 and Y2 surfaces
expand when strain rates increase: see [14]. The Y 3 surface is generally anisotropic and
can be located relatively easily with structured soft clays [15]; it is harder to identify in
stiff natural clays: see [1].
In reporting the experimentally observed stiffness trends, we assume that the clays
behaviour is transversely anisotropic, so allowing Cartesian and cylindrical coordinate
axis systems to be used interchangeably. While we focus mainly on the behaviour of
natural clay samples, we also consider parallel experiments on reconstituted specimens.
4.2 Stiffness anisotropy within the stiff clays elastic Y1 ranges

The first programme of research on London clay involved suites of slow small-
strain drained triaxial probing tests [1] in which pairs of tests were conducted that (i)
raised a single specified effective stress component and (ii) reduced the same effective
stress.
These slow experiments required very high instrument resolution and stability.
Special measures were taken to reduce temperature fluctuations to below 0.1 oC within
the triaxial cells. Similar experiments were conducted on Gault, Kimmeridge and
Oxford clay samples reconsolidated to the nominal in-situ (10m depth, K0 = 1.75
0.05) in-situ stress conditions.
Figures 10a and b illustrate the typical response observed, showing a pair of axial
(a) loading and (b) unloading probing tests conducted on a Kimmeridge rotary core
sample. Figures 11a and b present the equivalent pair of small-strain radial loading and
unloading tests.
The key features observed are:
The strain sensors showed a scatter of around 0.0001%, while the inherently
more difficult radial strain measurements scattered by around 0.0003%.
The loading and unloading tests all show approximately linear initial
behaviour over their very small ranges.
The axial tests show clear non-linearity after (local) axial strains of 0.0015%
and 0.0025% in compression and extension respectively. The radial probing
tests remained approximately linear within the same ranges of radial strains.
The initial da/da gradients are approximately equal under loading and
extension under axial probing, but the radial probing tests showed softer
gradients, A = dr/dr on unloading than under loading.
Consistent ratios develop between the strain components; dr/da -0.22 under
purely axial loading, and -0.55 under radial loading.
Figure 10. Drained axial stress probing test on Kimmeridge Clay; axial response above (a), radial strain
response below (b); from [6] and [10].
Figure 11. Drained radial stress probing test on Kimmeridge Clay; radial response above (a), axial strain
response below (b), from [6], [10].
Provided that behaviour within the Y1 loci can be considered as being elastic, the
effective stress cross-anisotropic compliances can be expressed as Eq. 2:
' ) x $ 1 Eh( ( Eh(

hh ( Ev(
vh 0 0 0 ') (x $
! ) ! ( E (
! y ! hh h 1 Eh( ( Ev(
vh 0 0 0 !!) (y !!
( Eh(
!! ) z !! hv ( Eh(
hv 1 Ev( 0 0 0 !!) (z !!
& # & #
!)+ xy ! 0 0 0 1 Ghv 0 0 !)* xy !
!)+ yz ! 0 0 0 0 1 Gvh 0 !)* yz !
! ! ! !
!%)+ zx !" 0 0 0 0 0 1 Ghh !%)* zx !"
(2)
Axial probing tests performed under constant radial stress probes give:
1
) v ) v' (3)
(
Ev
(
vvh
) h ) v' (4)
Ev (
Noting that the axial strain data show less scatter than radial, the Poissons ratio is
measured less reliably than Ev'.
Radial probing tests performed under constant axial stress (v'=0), lead to
equations which can be manipulated [see 16] by identifying the relationship between
radial effective stress change and radial strain, A = dr/dr and combining this with
dynamic Ghh Bender Element shear wave velocity measurements to give:
(5)
(6)
The remaining effective stress Poissons ratio is then hv' = vh' Eh' / Ev' and the
compliance matrix completed by substituting Bender Elements or Resonant Column
Ghv shear moduli. Noting that the denominator of Equation (6) involves subtracting two
quantities of similar sizes, the effects of measurement errors on the hh evaluation are
naturally more severe than those applying to hv. In the same way the Eh outcomes are
liable to greater error margins that those for Ghv or Ev' because Equation (5) requires: an
additional stiffness measurement; input from (more scattered) radial strain data rather
than axial; and an additional assumption that stiffness is strain rate independent. The
latter is critical to combining the static A and dynamic Ghh measurements in Equations
(5) and (6). In principle, undrained elastic vertical and horizontal stiffnesses can be
found through Equations (7) and (8) from [18].
(7)
(8)
However, the outcomes are highly sensitive to any errors in stresses or strains,
particularly in the case of the Ehu Equation (8). Clearer, more direct, measurements of
Ehv, Evu and Ghv were obtained by manipulating the measurements made in undrained
HCA tests on all four clays: see [5] and [8]. Most of the HCA experiments relied on
external or platen-to-platen strain measurements, so every feasible step was taken to
allow for effects of apparatus compliance, eliminate bedding errors and reduce
instrument scatter. The experiments could clearly be improved by applying higher
resolution local strain sensor systems.
Ev and vh Eh and hv, hh Evu Ehu
Clay MPa MPa MPa MPa
Drained axial Drained axial & radial Undrained Drained axial, radial
triaxial probes probes and Eqs. 5 & 6 triaxial probes & Eq. 8
Gault 75, 0.20 234, 0.62, 0.06 132 690
Kimmeridge 90, 0.22 219, 0.53, -0.10 147 472
Oxford 113, 0.22 321, 0.62, -0.34 270 862
London 142, 0.07 232, 0.50, 0.01 182 330
Table 3. Vertical and horizontal drained and undrained Youngs moduli and drained Poissons ratios from
within elastic ranges from samples at 10m depth at four clay sites
Summaries are given in Tables 3 and 4 of the compliance terms found for all four
clays in drained triaxial probing and resonant column tests conducted under the in-situ
conditions expected 10m below each clay stratums upper surface. Also listed are
ranges for the field Gvh and Ghh values interpreted from in-situ shear wave velocity
measurements summarized in Figures 2 to 4 and in [1] for the London clay. The
laboratory Poissons ratio measurements are presented in Table 3.
The vertical Youngs moduli represent averages from both loading and extension
tests. Best estimate vertical elastic undrained triaxial Evu stiffness and Bender Element
Ghh and Ghv = Gvh moduli are also listed, which were derived from undrained triaxial
tests. The values tabulated for Gault, Kimmeridge and Oxford clays were derived by
substituting the in-situ p values into best fitting relationships established between Evu,
Ghh and Ghv and p that were derived from multiple individual direct measurements.
Estimates for Ehu derived from Equation (8) are also shown that are acknowledged to
be less reliable, as argued above. Where multiple measurements were made of the same
quantity (for example resonant column Gz values) the averages are given along with
standard deviations.
Ghv Ghh Gz, Gvh Ghh
Clay MPa MPa MPa MPa MPa
Bender Bender Resonant Field Field
Test Test Column Shear waves Shear waves
Gault 57 110 74 (5) 80 to 110 150-165
Kimmeridge: 70 121 100 50 to 80 NA
Oxford 105 243 85 (6) 40 to 70 NA
London 60 115 86 ( 4) 75 to 105 150-165
Table 4. Horizontal and vertical shear stiffness within elastic range from samples at 10m depth
Overall, the laboratory tests confirm marked initial stiffness anisotropy in all the
four cases that appears to be related primarily to the clays micro-structures. The spread
of Poissons ratios shows another remarkable aspect of the clays anisotropic small
strain deformation behaviour. While the vh values fall in the commonly expected (0.07
to 0.22) range, the hv values are far larger (0.50 to 0.62) the hh ratios are either lower
or negative. As outlined above, the hv and hh ratios cannot be measured with the
same accuracy as vh, but the results obtained by applying hv' = vh' Eh' / Ev' and
Equation (6) are consistent between the four clays and therefore convincing.
4.3 Relationships between elastic stiffnesses and mean effective stress level
Multiple sets of probing tests were conducted on samples from the 50m deep
Heathrow T5 London clay profile [1] that revealed how the elastic stiffnesses and their
anisotropy varied with depth. One key finding was that the (in-situ and laboratory)
stiffnesses, which had developed over geological time, correlated linearly with the in-
situ mean effective stress p0. However, when power-law functions with the form given
in Equation 9 were fitted to laboratory measurements made at effective stress states
imposed to either side of the in-situ p values the resulting stiffness-pressure
relationships were far flatter and gave exponents N well below unity.
G0 = f (e) M Pa (p/Pa) N (9)

In this case the void ratio normalising function f(e) = e-1.3 after [17] and Pa is the
atmospheric pressure, which is taken nominally as 100 kPa. The material coefficient M
is independent of the system of units applied. The limited depth ranges of the
investigated Gault, Kimmeridge and Oxford clay profiles (see Figures 2 to 4) make it
hard to evaluate equivalent relationships between the in-situ stiffnesses and effective
stresses. The geophysical data tend to scatter due to occasional hard inclusions and also
show near surface variations due to weathering, seasonal variations and other factors.
However, the multiple undrained triaxial and non-destructive (Bender Element and
Resonant Column) tests allow the relationships between elastic stiffness and p to be
investigated over ranges above and below the natural samples in-situ p0 values.
Comparable trends are also available from experiments on reconstituted samples and
Figs. 12, 13 and 14 present summary plots for the dynamic Ghh, Ghv and Gz values
found for Gault, Kimmeridge and Oxford clays respectively. The void ratio function
has been applied and power law functions (Equation 9) have been fitted. Points to note
include:
All four clays natural samples show markedly higher Bender Element Ghh
values (from horizontally travelling and polarised waves) than in equivalent
Gvh (or Ghv) measurements, with ratios greater than 2 applying to the M values
from the two wave orientations.
The independent Resonant Column HCA Gz measurements, which involve
the same shearing mode as the Gvh Bender Element tests, but different
geometries and frequencies, give broadly comparable although not identical,
trends.
The power law exponents N applying after normalisation for void ratio
changes generally range between 0.2 to 0.6 and scatter within 0.05 of 0.5 in
six out of the nine cases considered.
After normalisation for void ratio, the vertical platen-to-platen Gvh
measurements made on reconstituted samples show fair agreement with the
natural samples Ghv Bender Element trends.
However, the isotropically consolidated reconstituted samples showed little or
no difference between Ghh and Gvh Bender Element trends. It appears that with
reconstituted samples, shear wave anisotropy develops primarily as a result of
imposing anisotropic consolidation stress conditions.
Although scattered, the Gvh moduli from in-situ shear wave tests (see [5], [6] and
[7]) are broadly compatible with the laboratory trends summarized later in Table 4.
However, the field cross-hole Ghh ranges tend to exceed the laboratory Bender Element
values significantly, possibly due to the presence of hard bands that provide short cuts
for shear waves travelling horizontally in the field, leading to possibly unrepresentative
overall indications of the soil mass stiffness anisotropy.
Figure 12. Dynamic elastic shear stiffness-mean effective stress relationships

from triaxial Ghh and Ghv BE and HCA RC Gz tests on Gault clay, after [10].

from triaxial Ghh and Ghv BE and HCA RC Gz tests on Kimmeridge clay, after [10].

from triaxial Ghh and Ghv BE and HCA RC Gz tests on Oxford clay, after [10].
4.4 Undrained triaxial compression stiffness over the non-linear range
The non-linear stiffness response was investigated in multiple undrained triaxial

tests taken through to failure. Figures 15 to 17 show how non-linear vertical undrained
secant stiffness Euv varied with axial strain (where a = invariant shear strain s under
undrained traixial conditions) in multiple triaxial compression tests conducted from a
range of isotropic initial stress states. Each curve degrades from an initial linear plateau
that extends to less than 0.002% strain in most cases; some tests clearly manifest more
scatter than others.
The undrained stiffness trends for the Gault, Kimmeridge and Oxford clays are
explored further in Figures 18 to 20 by plotting against p the secant stiffness values
developed at a = s levels of 0.001, 0.01, 0.1 and 1%, ranging from the initial Y 1
response up to failure. The equivalent London clay data are presented in [1]. Noting
that the void ratio function applied in Equation 9 may not apply over the non-linear
range, the moduli are not normalized by f(e). The secant stiffness-pressure relationships
are therefore presented in terms of the simpler expression given in Equation 10.
Euv = m Pa (p/Pa) n (10)

Figure 15. Non-linear stiffness-strain relationships from undrained triaxial tests on Gault Clay, after [6].
Figure 16. Non-linear stiffness-strain relationships; undrained triaxial tests on Kimmeridge Clay, after [6].
Figure 17. Non-linear stiffness-strain relationships from undrained triaxial tests on Oxford Clay, after [6].
Figure 18. Secant undrained stiffness-mean effective stress relationships

at four strain levels from triaxial tests on Gault clay, after [10].

at four strain levels from triaxial tests on Kimmeridge clay, after [10].

at four strain levels from triaxial tests on Oxford clay, after [10].
It is interesting that the effective stress level exponents (n) assessed for the initial
Euv maxima range from 0.6 to 0.88 and exceed the equivalent N values from the Bender
Element and Resonant Column shear stiffness tests, particularly for the Gault and
Kimmeridge clays. It is also interesting that the exponents decline rather than increase
with strain level as has been noted with sands and sandy clays [11], [13].
4.5 Undrained stiffness anisotropy over the full non-linear range from HCA tests
Triaxial tests are limited to exploring how the Evu or Ev compliance components
vary with strain beyond the clays respective elastic Y1 limits. Undrained triaxial
compression and extension tests on identical specimens should manifest the same
initial (vertical) elastic stiffnesses. However, vertical loading and radial loading
tests often show different Y1 strain limits and subsequent non-linear responses as their
stress paths traverse different regions of effective stress space towards potentially
anisotropic failure states.
HCA experiments can establish how other compliance components vary with
strain beyond the elastic limits. The Evu, Ehu and Gvh trends of all four clays were
tracked in HCA experiments that led to failures with f values of 0, 22.5, 45 and 90 o.
As noted earlier, the tests maintained b values equal to 0.5 and started from the
estimated in-situ p and K = 1.75 0.05 conditions. Simple shear tests were also
undertaken and Figures 21 to 32 present summary plots for all four clays. Initial secant
values of Evu, Ehu and Ghv were found for each test by fitting linear regressions over the
trends applying up to 0.005% strain. These initial HCA stiffnesses might differ from
the elastic maxima applying at still smaller strains. The triaxial, bender element and
resonant column test outcomes are listed in Tables 3 and 4 and are also noted for
reference in the captions for each figure.
The experiments reveal interesting trends. They also expose the limitations of the
elastic interpretive framework. The stiffness parameters of any truly elastic material
should be stress path independent. While the Gvh traces of each clay fall within
relatively narrow ranges, the Ehu traces vary significantly, depending on the HCA stress
path followed. The Evu characteristics show intermediate degrees of stress path
dependence. The main points to note are:
The initial HCA Evu values fall close to the triaxial measurements for all cases
except the Kimmeridge, where their average is considerably higher.
The equivalent HCA Gvh maxima generally match the Bender Element data
for all cases apart from the Oxford, but fall below the Resonant Column range
for all four clays.
The initial HCA Ehu maxima fall well below the values computed by
substituting triaxial and Bender Element data into Equation (8) for all cases
except the London clay. Given the difficulties of applying Equation (8)
reliably, we conclude that the HCA data are more representative.
The HCA traces show continuous reductions in secant stiffness from their
initial maxima with increasing strain. All four clays manifest clear anisotropy
over their full non-linear ranges with Gvh < Evu < Ehu.
Figure 21. Variation of vertical undrained Youngs modulus with vertical strain. HCA tests on Gault clay;
after [8]. Note maximum triaxial compression Evu = 132 MPa; see Table 3.
Figure 22. Variation of horizontal undrained Youngs modulus with mean of radial and circumferential strains.
HCA tests on Gault clay; after [8]. Note maximum Ehu from Equation 8 = 690 MPa; see Table 3.
Figure 23. Variation of torsional shear modulus with horizontal shear strains. HCA tests on Gault clay; after
[8]. Note Triaxial Bender Element and field seismic Gvh are 57 and 80 to 110 MPa respectively; see Table 3.
Figure 24. Variation of vertical undrained Youngs modulus with vertical strain. HCA tests on Kimmeridge
clay; after [8]. Note maximum triaxial compression Evu = 147 MPa; see Table 3.
HCA tests on Kimmeridge clay; after [8]. Note maximum Ehu from Equation (8) = 472 MPa; see Table 3.
Figure 26. Variation of horizontal shear modulus with horizontal shear strains. HCA tests on Kimmeridge
clay; after [8]. Note Triaxial Bender Element and field seismic Gvh are 70 and 50 to 80 MPa respectively; see
Table 3.
Figure 27. Variation of vertical undrained Youngs modulus with mean of radial and circumferential strains.
HCA tests on Oxford clay; after [8]. Note maximum triaxial compression Evu = 270 MPa; see Table 3.
HCA tests on Oxford clay; after [8]. Note maximum Ehu from Equation (8) = 862 MPa; see Table 3.
Figure 29. Variation of horizontal shear modulus with horizontal shear strains. HCA tests on Oxford clay;
after [8]. Triaxial Bender Element and field seismic Gvh are 105 and 40 to 70 MPa respectively; see Table 3.
Figure 30. Variation of vertical undrained Youngs modulus with vertical strains. HCA tests on London clay;
after [8]. Note maximum triaxial compression Evu = 182 MPa; see Table 3.
HCA tests on London clay; after [8]. Note maximum Ehu from Equation (8) = 330 MPa; see Table 3.
Figure 32. Variation of horizontal shear modulus with horizontal shear strains. HCA tests on London clay;
after [8]. Triaxial Bender Element and field seismic Gvh are 60 and 75 to 105 MPa respectively; see Table 3.
5. Summary and conclusions
This paper has drawn on ten primary sources (references [1] to [10]) to summarize the
outcomes of integrated studies into the anisotropic shear strength and stiffness
behaviour of four high OCR stiff-to-hard medium plasticity UK strata: the Gault,
Kimmeridge, Oxford and London clays. High quality rotary and block samples were
tested in triaxial and hollow cylinder stress path experiments, with local strain, multi-
axial bender element and resonant-column techniques that were interpreted in
combination with independent field seismic test data. Overall, it has been shown that:
1. The field and laboratory experiments conducted on the stiff clays offer
overlapping capabilities and data that can be integrated to develop continuous
profiles of key quantities, such as cone resistance and shear wave velocities as
well as intensive descriptions of the stiff-to-hard clays shear strength and
stiffness behaviours.
2. The laboratory tests confirm that the clays mechanical behaviours are
markedly brittle, anisotropic, pressure dependent and highly non-linear.
3. The b = 0.5 HCA experiments reveal broadly compatible hierarchies of shear
strength anisotropy, as expressed by variations of peak Su with , the 1 axis
orientation. Three of the four clays develop their minimum shear strengths
when falls in the mid-range of 50o 10o. The exception, Kimmeridge Clay,
develops its minimum Su at ~ 10o.
4. The b = 0.5 nominally plane strain shear strengths of Gault and Kimmeridge
clays exceed at all values the values seen in triaxial compression (b = 0, =
0o) and extension (b = 1, = 90o) tests. However, the b = 0.5 Su values fall
significantly below the triaxial compression values when < 85o in Oxford
clay and overlap with the London clays triaxial compression spread, which
depends critically on fissure orientation and spacing.
5. The four clays peak effective stress t/s ratios vary with , showing anisotropy
and with the b value showing variation also in the deviatoric plane. Far lower
shear strengths develop post-peak in all four clays once oriented shear bands
form and allow low residual r shear strengths to develop.
6. The suites of laboratory Bender Element, Resonant Column and static HCA
experiments led to a range of small strain elastic Gvh measurements that were
broadly compatible with the field data, although the latter showed significant
scatter with depth due to occasional hard bands and other features [7], [9]. The
Bender Element Ghh values fell more significantly below cross-hole field
measurements. While this may be due to an influence of hard bands in the
field, Ghh values gauged from HCA tests match to the field trends better.
7. Direct drained and undrained measurements of elastic Ev, Evu, vvh, Gvh and
Ghh values are reported from bender-equipped triaxial experiments. These
measurements are more reliable than the estimates for Ehu, vhh and vhv derived
through applying expressions derived from elastic theory. Values of Eh
derived from the triaxial static and bender element measurements tests have an
intermediate reliability.
8. The experiments showed clear and marked elastic stiffness anisotropy, with
horizontal Youngs and shear moduli showing notably higher values than their
vertical equivalents under all conditions. The cross anisotropic Poissons ratios
were also strongly anisotropic.
9. Drained HCA tests can also provide direct measurements of all the effective
stress cross anisotropic compliance parameters over the full range of strain;
see [7] and [19].
10. Experiments performed over a range of mean effective stress levels were able
to show how the various stiffness components varied with strain and mean
effective stress levels.
11. Power law expressions that related initial (elastic) stiffness to p in cases
where the stress states varied to either side of the in-situ mean effective stress
showed exponents that ranged from 0.6 < n < 0.9 for undrained vertical
Youngs modulus and from 0.2 < N < 0.6 for the Gvh, Ghv and Ghh shear moduli.
The secant stiffness curves developed in undrained triaxial tests indicated
exponents n that declined with increasing strain levels.
12. While undrained triaxial tests can show how Evu decays with strain, the
undrained HCA tests were also able to show how Ehu and Gvh vary over the
non-linear range. When treated and interpreted as described in [8], they can
also provide reasonable estimates for the initial elastic Evu, Ehu and Gvh moduli.
Detailed information on the anisotropy of shear strength and stiffness of the four
stiff clays considered has not been available previously, so limiting the accuracy with
which geotechnical predictions could be made of a wide range of problems; see for
example [20]. The new findings will help to improve the characterization and
modelling of practical problems involving similar stiff clays. More specifically, the
results should find application in major new UK infrastructure projects that involve the
four strata investigated specifically in the study described.
Acknowledgements
The Authors acknowledge with thanks EPSRCs funding of the research described and
to the contributions made by many current and former colleagues at Imperial College,
particularly Drs Anh-Minh, Gasparre and Nishimura who undertook the earlier
Heathrow Terminal Five London clay study and Dr Wilkinson who carried out the
geological research at the new Gault, Kimmeridge and Oxford clay sites. The Authors
are grateful to In Situ Site Investigation for CPT profiling at the new sampling sites, to
Cambridge University for access to the High Cross site, to Arup for access to the
Elstow site and earlier data and to Mr Walker who permitted sampling at his
Willowbrook Farm.
References
[1] Gasparre, A., Nishimura, S., Anh-Minh, N., Coop, M.R. & Jardine, R.J.
(2007). The stiffness of natural London clay. Gotechnique, 57(1), pp 33-
48.Gasparre, A.,
[2] Nishimura, S., Minh, N. A. & Jardine, R. J. (2007) Shear strength anisotropy
of natural London Clay. Gotechnique 57, No. 1, 4962.
[3] Hight, D.W., Gasparre, A., Nishimura, S., Anh-Minh, N., Jardine, R.J. &
Coop, M.R. (2007). The London clay at T5: An overview of the
characterisation study. Gotechnique, 57(1), pp 3-18.
[4] Wilkinson, S. (2011) The microstructure of UK mudrocks. Ph.D Thesis,

Imperial College London.
[5] Brosse, A. (2012). Study on the anisotropy of British stiff clays using a
Hollow Cylinder Apparatus. Ph. D. thesis, Imperial College London.
[6] Hosseini Kamal, R. (2012). Experimental study of the geotechnical properties
of UK mudrocks. Ph. D. thesis, Imperial College London.
[7] Hosseini Kamal, R., Coop, M.R., Jardine, R.J. and Brosse, A. (2014). The
post-yield behaviour of four Eocene-to-Jurassic UK stiff clays. Geotechnique.
64, No. 8, 620634.
[8] Brosse A., Jardine R.J. and Nishimura, S. (2015). Undrained stiffness
anisotropy from Hollow Cylinder testing of four Eocene-to-Jurassic UK stiff
clays. Under review, Canadian Geotechnican Journal.
[9] Brosse A., Jardine R.J. and Nishimura, S. (2015). The undrained shear
strength anisotropy of four Eocene-to-Jurassic stiff clays. Under review,
Geotechnique.
[10] Brosse, A., Hosseini Kamal, R., Jardine, R.J. and Coop, M.R. (2015). The
shear stiffness characteristics of four Eocene-to-Jurassic UK stiff clays. Under
review, Geotechnique.
[11] Jardine, R.J, Gens, A., Hight, D.W. and Coop, M.R. (2004). Developments in
understanding soil behaviour. Keynote paper. Advances in Geotechnical
Engineering. Proc Skempton Memorial Conference. Pub Thomas Telford,
London, pp. 103-207.
[12] Jardine, R.J. (1992). Observations on the kinematic nature of soil stiffness at
small strains. Soils and Foundations, 32 (2): 111-124.
[13] Jardine, R.J. (2013). Advanced laboratory testing in research and practice. 2nd
Bishop Lecture. Proc. ICSMGE, Paris, Presse des Ponts. Vol 1, p. 35-55.
[14] Tatsuoka, F. (2011). Laboratory stress-strain tests for developments in
geotechnical engineering. 1st Bishop Lecture, Deformation Characteristics of
Geomaterials. Proc. IS-Seoul, Hanrimwon, Vol. 1, p 3-53.
[15] Smith P.R., Jardine R.J. and Hight, D.W. (1992). On the yielding of
Bothkennar clay. Geotechnique, Vol 41, No 2, pp 257-274.
[16] Kuwano, R. (1999). The stiffness and yielding anisotropy of sand. PhD Thesis,
Imperial College, London.
[17] Jamiolkowski, M., Leroeuil, S. and Lo Presti, D. C. F. (1991). Design
parameters from theory to practice. Geo-coast '91 Int. Conf., Yokohama, Port
and Harbour Research Institute, Yokosuka, Vol. 2, pp 877-917.
[18] Lings, M. L. (2001). Drained and undrained anisotropic elastic stiffness
parameters. Geotechnique 51, No. 6, 555565.
[19] Anh-Minh, N. (2006). An investigation of the anisotropic stress-strain-
strength characteristics of an Eocene clay. PhD Thesis, Imperial College
London.
[20] Addenbrooke, T.I., Potts, D.M. and Puzrin, A.M. (1997). The influence of pre-
failure stiffness on the numerical analysis of tunnel construction.
Geotechnique, Vol 47, No 3, pp 693-712.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-192
Measurement and Application of

Shear Wave Velocity to
Various Geotechnical Problems
Dong-Soo KIMa,1, Heon-Joon PARK a and Hyung-Ik CHO a
a
Department of Civil and Environmental Engineering, KAIST
Abstract. The shear wave velocity (Vs), which is directly related to the small strain
shear modulus, Gmax, can be measured by both field and laboratory tests and has a
great potential in applications to various geotechnical problems. Soil exhibit
nonlinear stress-strain behavior from very small strains and the reliable strain
dependent modulus is usually determined by combining Gmax obtained from field
seismic test and G/Gmax curve from the laboratory test. Therefore, the Vs is a key soil
parameter for the deformation analysis of geotechnical structures. In this paper, the
advantages of using Vs in geotechnical applications are discussed. The main features
of various intrusive and non-intrusive Vs measuring techniques in the field, resonant
column, torsional shear and bender element tests in the laboratory, and bender
tomography in centrifuge model are summarized. Test results obtained by various
field tests, laboratory element tests, and centrifuge models are compared and
critically discussed considering parameters affecting soil stiffness as well as
reliability of the test results. The Vs has been traditionally applied to earthquake
related problems but its use has expanded to cover even the static deformation
analysis problems. In this study, three cases of using Vs to geotechnical problems
are introduced: (1) Evaluation of Ko value using Vs in centrifuge model, (2)
settlement prediction of shallow foundation using Vs profile, and (3) evaluation of
ground improvement and densification using Vs. The background theory, procedure,
laboratory and field tests, physical modeling, and case studies for each application
are discussed.
Keywords. Shear wave velocity, deformation behavior, measurement, application
1. Introduction
The design paradigm of civil engineering structures have moved toward the performance
based design (PBD) from the traditional safety factor based design. In the PBD, the great
emphasis is placed on control of structural deformation and not on safety factor in order
to assure the serviceability and durability. The crucial factor controlling the deformation
of ground supported structure is stiffness not strength. Therefore, the evaluation of
deformation characteristics of geomaterials is very important in site investigations for
PBD [1], [2].
It has been well known that the soil behaves non-linearly from the small strains [3],
[4]. The maximum shear modulus (Gmax) below the elastic threshold strain, which is a
fundamental stiffness in design, is usually inferred from the shear wave velocity obtained
1
Corresponding Author: Professor at KAIST; E-mail: [email protected]
D.-S. Kim et al. / Measurement and Application of Shear Wave Velocity 193
by seismic wave propagation tests. The variation in shear modulus with strain can be
determined at small to intermediate strains by resonant column, torsional shear and
triaxial testing equipment, and the normalized modulus reduction curve (G/Gmax vs. log
!$[5], [6]. Because the modulus value obtained by laboratory
test is affected by sampling disturbance and is difficult to be representative of the site,
the reliable nonlinear stiffness variation is usually determined by combining Gmax
obtained by field seismic test and G/Gmax curves from laboratory test, in which Gmax
converted from the shear wave velocity can be used as key soil property for the
deformation analysis of soil-foundation systems.
The strain amplitude is understood to be a key variable in predicting soil behavior,
whether the strain comes from static or dynamic phenomena [3], [4], [5]. Therefore,
reliable nonlinear stiffness variations can be determined by combining Gmax converted
from Vs in the field and G/Gmax measured by laboratory resonant column/torsional shear
tests. The Vs can be obtained by both field and laboratory tests and the in-situ nonlinear
stiffness variation can be reliably estimated by combining field and laboratory results.
As the Vs is dependent on the effective stresses, the field stress conditions can be also
evaluated by measuring Vs.
In this paper, the advantages of using Vs in geotechnical applications are discussed
and various field and laboratory measuring techniques of Vs are summarized. And in-
flight Vs measurement and tomography within centrifuge model are introduced. The
applications of Vs in the analysis of various geotechnical problems are discussed: (1)
evaluation of Ko value using Vs in the centrifuge model, (2) settlement prediction of
shallow foundation, and (3) evaluation of ground improvement and densification.
2. Advantages of Vs in the Geotechnical Applications
The shear wave velocity (Vs), which is directly related to the small strain shear modulus,
Gmax, has several advantages in the geotechnical applications and a great potential in the
evaluation of deformation characteristics. The Vs has several advantages in the
geotechnical applications as follows.
(1) The Vs is directly related to the deformation property of soil, the elastic shear
modulus (G), i.e. soil stiffness
(2) The Vs can be obtained by both field and laboratory tests and the in-situ
nonlinear stiffness variation can be reliably estimated by combining the test
results.
(3) The Vs represents a large volume of soil, and hence characterizes the average
properties, and not a point information.
(4) The Vs can be measured non-intrusively on the gravel and cobble layers which
are hard to be characterized.
(5) The Vs on particulate materials is dependent on the effective stresses in the
direction of wave propagation and particle motion [7], [8]. The field stress
conditions and degree of consolidation can be evaluated by measuring Vs.
(6) The Vs is an essential parameter in the geotechnical earthquake engineering
because the propagation characteristics are similar to the real seismic wave. The
small-strain, elastic shear modulus (Gmax) of soils is an important parameter in
the site response analysis of earthquakes, the design of machine foundations,
and soil dynamics problems. Measurement of the Vs near the subsurface is also
194 D.-S. Kim et al. / Measurement and Application of Shear Wave Velocity
important in geotechnical engineering practice, since it can provide input to

seismic design methods such as site response analysis and the evaluation of
liquefaction potential.
(7) As much interest has been developed in the area of low amplitude problems
associated with human-made vibrations, the Vs of the site can be used for the
characterization of ground vibrations, and for the study of propagation and
attenuation of ground vibrations [9]. Nondestructive evaluation of integrity of
geotechnical structures can be performed using primary and/or shear wave
propagation [10].
(8) The perceived difference between the static and dynamic moduli is decreasing
as the accuracy in static measurement is improving, and so the Vs can be applied
in static deformation analysis as well as in traditional dynamic problems. The
importance of Gmax has been emphasized even in static deformation problems
such as those induced by excavation and settlement [11], [12].
(9) As the Vs represents the material and structural conditions of the site, it can be
applied for the evaluation of layer structures, degree of compaction or
consolidation of soft soil and weak zones found within a site [13], [14], [15].
For PBD, the deformation behavior of geotechnical structures governs the design
while current design methods usually rely on the strength of geomaterial and the
corresponding factor of safety. It is important to characterize the soil conditions of
geotechnical systems accurately to predict deformation behavior. The Vs can be used as
a key parameter for evaluating deformation of geotechnical structures because it is highly
relevant to the deformational characteristics of soil; maximum shear modulus, Gmax.
3. The Seismic Testing Methods for the Vs Measurements
3.1. Vs measurements in field tests
An elastic half-space is appropriate as a model of the earth. Upon hitting the ground using
hammer or some other input source, the energy transmitted into the ground is propagated
away from the source in a combined form of P-, S-, and R-waves. Body waves (P and S
waves) propagate outward from the source along the hemispherical wave front, whereas
the surface (R) wave propagates along a cylindrical wave front. S waves often divide
into two types, as SH and SV waves, depending on the plane of particle motion. Surface
wave energy exists mostly within a depth of one wavelength, and the dispersive
characteristic in which waves of different wavelengths propagate at different velocities,
is shown in layered media. Wave propagation theory and the observed phenomena
should be further refined by considering a layered half-space model, where the waves
are reflected and refracted at the layer boundary. In the site investigation of soil layers,
S and R waves are mostly used because the stiffness of the soil structure cannot be
distinguished using P wave below the ground water table and the shearing mechanism
during earthquake is similar to shear wave propagation. Compared to other geophysical
methods, seismic method has a certain advantage of obtaining wave velocity profiles
which are engineering properties directly related to the soil stiffness [16], [17].
The Vs that is determined in the field is measured below the elastic threshold strain,
which generally means the strain at or lower than 10-3 %. The shear wave velocity
measurement techniques in the field can be classified into surface investigation (non-
intrusive method) and borehole investigation (intrusive method). Various field

investigation techniques have been researched and developed so as to determine the
distribution of the Vs of the ground. The test techniques can be classified in accordance
with the sensor arrangement or the propagation characteristics of the seismic wave that
is used. It is important to select and apply a reasonable test method by considering
various things including the ground characteristics of the site, test expenses and time,
and the accessibility.
The testing borehole is utilized in the intrusive method, whereas the source and
receivers are on the ground surface in the non-intrusive methods. Typical intrusive
methods include crosshole [18], downhole [19], [20], SPT based uphole [21] and
suspension PS-logging [22] methods. The special features of each intrusive method were
summarized in [1], [12].
Some of the popular surface site investigation methods are reflection method,
refraction method and surface wave method. While the surface investigation methods
can be applied even when there is no borehole, the vertical resolution is weaker than in
the borehole investigation. The surface investigation methods that is mostly utilized in
measuring the Vs of the ground is the inspection by surface wave (Rayleigh wave mostly
used). For the surface wave tests, the spectral analysis of surface waves (SASW) [23],
multichannel analysis of surface waves (MASW) [24], frequency-wavenumber (f-k)
spectrum [25], continuous surface wave (CSW) [26], harmonic wavelet analysis of
waves (HWAW) [27], and refraction microtremor (ReMi) [28] methods are frequently
utilized.
At present, the inspection by surface wave is mainly used for the measurement of Vs
of the ground from the earth surface. Lately, diverse techniques are developed and
evolved to overcome the limitations of the conventional surface wave methods. The
special features of surface wave methods in each step were summarized in in [1], [12].
Most of the methods use the active ground sources, but some methods use the passive
ambient vibrations. The surface wave method which uses the dispersion characteristics
in the layered media consists of three steps: field testing, evaluation of dispersion curve,
and determination of Vs profile using an inversion. The schematic diagram for field
seismic tests and the regions of site sampled during each testing are shown in Fig. 1 [2].
Figure 1. Schematic diagram for field seismic tests [2]

The round robin testing (RRT) program was carried to assess the reliability and
limitation of various intrusive and non-intrusive methods [29]. After a suitable site for
RRT was determined, volunteer groups were invited to perform the various field testing
methods. Five site investigation companies, three universities, and one research institute
participated for this RRT. Typical intrusive borehole methods such as crosshole,
downhole, inhole, uphole seismic tests, and suspension PS logging test were applied. For
non-intrusive methods, surface wave methods such as SASW, MASW, HWAW, ReMi
methods were applied. Main points discussed in this comparative study include (1) the
reliability and limitation of each method with focus on the Vs profile evaluation, (2) the
cause of difference in results, (3) the suggestions for better Vs profile evaluation, and (4)
the feasibility of 2-D subsurface image construction based on surface wave methods.
The obtained Vs profiles determined from crosshole, inhole and uphole tests and
various surface wave methods are displayed in Fig. 2. The reasons of difference in results
can be explained by the difference of employed shear wave and its coverage and
sampling area from source to receiver. Downhole test and SPS logging test is more stable
and economic to obtain Vs profile of a site compared with other borehole seismic tests.
If one can clearly pick the reference point for obtaining travel-time information in SPS
logging test, Vs profile produced by using SPS logging test will be reliable. However, it
is considered that Vs value is less reliable in the zone of above water level, low velocity
medium and cased-hole condition. Therefore, the combination survey of downhole
seismic and SPS logging tests was recommended in order to get the whole Vs profile.
Four different surface wave methods were applied and all eight data sets were
acquired. In this study, testing depth of interest was about 30 m, and it is desirable to
obtain experimental dispersion curves at least to the wavelength of 60 m. Only four
groups could get enough range of dispersion curves. SASW A and HWAW have very
wide dispersion curves and it could produce Vs profile to the over 25 m. Below depths
of 10m, MASW produced small Vs values compared with other Vs profiles because it
was determined from narrow dispersion curves having few low frequency information.
MASW C which has relatively wide dispersion curve produced deep Vs profile and
reliable Vs values compared with other MASW results.
Figure 2. Vs profiles determined by various borehole and surface wave methods as a RRT results
Through the comparison study of RRT for field seismic tests, it was concluded that
(1) the verified testing equipment and interpretation method should be employed in order
to obtain reliable Vs distribution of a site, (2) the quality of Vs profile depends on the
quality of the intermediate products such as travel-time measurement and/or dispersion
curve, (3) the Vs profiles determined from various field seismic test show good
agreements and same trend with depth, and (4) the reasons of small difference in results
can be explained by the difference of employed shear wave and its coverage and
sampling area from source to receiver.
3.2. Vs measurements in laboratory tests
Several techniques such as resonant column (RC) and torsional shear (TS) tests have
been widely used to measure Gmax in the laboratory [30], [31], [5]. In addition to these
techniques, the bender element (BE) method, a simple technique involving measurement
of the velocity of a shear wave through a soil specimen, has been widely used since its
introduction by Shirley and Hampton [32].
A schematic configuration of a fixed-free RC and TS testing equipment with the BE
is shown in Fig. 3. The bottom of the specimen is fixed to the base and the top (free end)
is connected to a drive system used to excite and monitor torsional motion. The basic
operational principle is to vibrate the cylindrical specimen in a torsional motion and
detect the first-mode resonance. Once first mode is established, measurements of the
resonant frequency and amplitude of vibration are made. Those measurements are then
combined with equipment characteristics and specimen size to calculate shear wave
velocity, shear modulus, and shearing strain amplitude [4].
The TS test is another method for determining the shear modulus using the same RC
equipment but operating it in a different fashion. In this test a cyclic torsional force with
a given frequency, generally below 10 Hz, is applied at the top of the specimen while the
bottom is held fixed, as shown in Fig. 3. Instead of determining a resonant frequency,
the torque-twist response of the specimen is determined. Proximitors are used to measure
twist, and the current applied to the coils is calibrated to yield torque. Shear modulus
corresponds to the slope of a line through the end points of the stress-strain hysteresis
loop.
Figure 3. Schematic diagram of the modified RC, TS and BE testing equipment [33]
The RC/TS apparatus has several advantages in the investigation of small strain
deformational characteristics. Both RC and TS tests can be performed with the same
setup simply by changing the frequency of the forcing function. Small torque can be
applied using a coil-magnet system without introducing mechanical compliances. Thus,
the results of RC and TS tests can easily be compared over a wide range of strains. The
effect of loading frequency on modulus can be investigated effectively because the
loading frequency in the TS test can be varied from 0.01 to 10 Hz.
A BE is a piezo-electrical transducer that either bends as voltage is applied or
generates voltage as it bends. Two BEs are set up at both ends of a soil specimen, acting
as a transmitter and a receiver. A function generator feeds a transmitter element with an
input waveform voltage to bend it, and a shear wave is thereby propagated through the
specimen. When the receiver at the other end is bent by the arrival of the shear wave, a
small electrical signal is generated as a waveform. The transmitted and received waves
are captured and displayed by a digital oscilloscope, and Gmax is determined from the Vs.
The BE technique offers several advantages: (1) it is nondestructive, and thus there
is virtually no limit to the number of repetitive tests that can be performed; and (2) it can
be set up in most laboratory apparatus, thus facilitating comparison with other test results.
For verification of the BE test, Dyvik and Madshus [34] performed both RC and BE tests
during consolidation of five different clays, and a good agreement between the Gmax from
the two tests was obtained. However, the subjectivity of determination of the arrival time
is a critical problem in BE test. Viggiani and Atkinson [35] and many other researchers
have attempted various methods to determine the arrival time to reduce the degree of
subjectivity. Lee and Santamarina [36] and Leong et al. [37] reported a method of travel
time determination in BE tests. However, there are some disagreements on the
determination of first arrival of the shear wave and the preferred input wave. Therefore,
there is still a strong need for verification of BE method with respect to reliable
determination of arrival time.
In the previous study [33], BEs were mounted in a Stokoe type RC and TS
equipment. Toyoura and silica sands were collected. The BE, RC, and TS tests were
performed in a small-strain range (<103 %) at various relative densities, drainage
conditions (dry and saturated), and effective confining pressure. Consequently, the
values of Gmax obtained by three different tests were compared.
The BEs have the following dimensions; 12.7 mm long, 8 mm wide, and 0.6 mm
thick. The BEs were electrically connected in series and coated with waterproofing
material (polyurethane). The BEs were then placed in slots of a top cap and a bottom
pedestal with a protrusion length of about 4 mm, and each gap in the slots was filled with
epoxy. Lastly, the BE receiver was covered with a conductive coating to prevent
electromagnetic coupling and cross-talk [38].
The values of Vs determined from the BE and RC tests in the dry conditions on
Toyoura and silica sands are compared at various effective confining pressures shown in
Fig. 4(a), respectively. The values of Vs obtained from BE tests are in good agreement
with those obtained from RC tests in all cases for two sands in the dry conditions. The
Vs values obtained from the BE and RC tests were converted to Gmax, and the results are
compared with the Gmax values obtained from TS tests shown in Fig. 4(b). The Gmax values
of BE, RC, and TS tests at confining pressures of 50, 100, and 200 kPa show a maximum
difference of about 3 % relative to the mean value at each confining pressure stage.
Figure 4. Comparison RC, TS and BE test results in (a) Vs and (b) Gmax with effective confining stress
[39]
Methods to determine the travel time in BE tests were investigated through the
comparison of analytical and experimental signals, and confirmed by comparing Vs of
BE tests with those of RC and TS tests. In dry conditions, the values of Vs between BE
and RC tests correspond well and the Gmax values were in good agreement with those
obtained from TS tests without the influence of loading frequency. In saturated
conditions, the values of Vs obtained from BE tests were somewhat greater than those
obtained from RC and TS tests, thus reflecting the dependence of Vs on frequency.
3.3. In-flight Vs measurements within centrifuge
There have been several methods suggested for evaluating the performance of
geotechnical structures. Physical modeling using geotechnical centrifuge can offer
unique capabilities in wide range of performance evaluation from small-strain soil
behavior to structural failure condition.
There have been a series of methods for measuring Vs in centrifuge models.
Installation of piezoelectric oscillator [40], a vertical array of bender elements [41], and
a mini-air hammer [42] were utilized for direct measurement of shear wave velocities.
Vs measurement using bender element array at various depths was adopted for centrifuge
tests before earthquake simulation to obtain the Vs profile [43], and the resulting
tomography measurement system using bender elements have been reported [44]. Vs
measurement using these methods can offer important information for the deformation
analysis.
An experimental setup to evaluate shear wave velocity distribution in model soil has
been developed for centrifuge model tests [45]. Piezoelectric BEs are used for both
generating and receiving shear wave signals and Vs distribution can be visualized by
tomography inversion. The change in Vs distribution before and after applying surface
load is monitored during in-flight condition and it has been found that the developed
system is useful for characterizing the soil condition and monitoring the soil behavior.
The purpose of Vs tomography measurement in centrifuge model is to monitor the
change of soil condition in each performance stage of geotechnical structures. This
program could be used for monitoring soil condition under foundation load in order to
assess its applicability for measuring Vs increase in soil mass due to additional stress
developed by footing. At the first trial, 100 mm diameter aluminum plate is prepared to
simulate footing load at the central part of the model surface and it is attached to the
vertical loading equipment to measure load-settlement behavior of the foundation.
The same soil model described previously is prepared and the centrifugal
acceleration of 50 g is applied. To compare the effect of footing load in the Vs distribution
of soil layer, shear wave travel time is first measured without any surface load to obtain
original tomography image. While controlling the vertical loading amplitude, load and
settlement of the foundation was recorded and the cross sectional area was also
monitored by tomography method using the BE arrays.
Fig. 5(a) shows the Vs tomography images for homogeneous soil sample before
loaded with footing. This is the reference information for assessing the changes of Vs in
soil mass due to loading condition. As the footing load increases, Vs increases at the
middle part beneath the footing as shown in Fig. 5(b). These phenomena can be clearly
displayed by comparing the tomography image before and after loading stage.
Subtracting the Vs value of original condition from the Vs value when the load is applied,
increment of Vs can be obtained as shown in Fig. 5(c). This increase in Vs beneath footing
is localized within 1.5 times of diameter of loading plate in depth, and 2 times of that in
the horizontal direction. This Vs increase can be interpreted as additional effective stress
by the footing load. Even if the distribution shape of increasing Vs is a little different
from analytical solution suggested by Boussinesq (1885), it is still useful for
understanding how the footing load affects the soil behavior beneath footing.
Figure 5. Vs increment due to footing load; (a) original condition, (b) VS distribution under 254kPa
footing load, (c) VS increment in percent [45]
An experimental method to evaluate the Vs distribution in a soil model has been

introduced for centrifuge testing in this study. Since Vs can be used as a key soil property
for the deformation analysis of geotechnical systems, it is concluded that the Vs
measurement in a centrifuge model can provide basic information of deformational
characteristics of the model.
4. Applications of Vs to Various Geotechnical Problems
The Vs has been used as an essential input parameter for dynamic studies, such as
geotechnical earthquake engineering or vibration problems, but recently its application
has widened to an engineering property to solve static deformational problems. The Vs
has been applied for various geotechnical problems, such as ground response analysis
for seismic load, evaluation of potential for site amplification, seismic design of
geotechnical structures, static deformational problems of ground and earth structures,
geotechnics in pavement engineering, ground improvement assessment, and so on. Three
cases of application of Vs to geotechnical problems are discussed: (1) Evaluation of Ko
value using Vs in centrifuge model, (2) Settlement prediction of shallow footings using
Vs, (3) Evaluation of ground improvement using Vs.
4.1. Evaluation of Ko value using Vs in centrifuge model
The coefficient of the earth pressures at rest (Ko, hereafter) is an important parameter in
analysis and design of geotechnical structures because the behavior of soil is governed
by stress conditions. The stress conditions are expressed by horizontal and vertical
stresses. The horizontal stress in soil media is highly dependent on horizontal direction
constraint unlike the behavior of liquid. Ko is defined as a ratio of horizontal effective
stress to vertical effective stress without horizontal displacement of ground called as at-
rest state.
Centrifuge modeling has been widely used to understand the prototype behavior of
geotechnical structures, and to calibrate numerical models in both working stress
conditions and conditions near failure. Ko in the centrifuge model may vary depending
on the distance from the boundary because of the friction between the wall and soil
particles. Thus, the desired simulations and measurements are usually carried out in the
central parts of the model in order to avoid boundary effects. However, because of the
relatively small size of the model box and the capacity limit of the centrifuge facility,
centrifuge simulations are usually performed within the zone influenced by the boundary
effect. Therefore, it is important to evaluate the size of the affected zone and its effect on
the centrifuge model. In addition, most of the centrifuge tests are performed as staged
tests by changing acceleration levels, thus, the stress states in the model may vary during
loading-unloading-reloading stages. Hence, it is important to identify the variation in
stress states at different locations during loading-unloading-reloading cycles.
Vs measurements could provide an alternative way for evaluating Ko in centrifuge
model because the Vs is directly related to effective stress conditions in the direction of
wave propagation and particle motion [7], [8]. Accordingly, shear waves can be
represented by Vsij, where the subscripts i and j represent the directions of wave
propagation and particle motion, respectively. For example, VsHV indicates the Vs
propagating horizontally while the direction of particle motion is vertical direction. The
relationship between Vs and effective stress can be expressed as follows:

V = C (1)

Where, C = material constant, = effective confining stress, n = stress exponent,

and P = atmosphere pressure.
Several proposed VsKo relationships that were used to determine Ko are derived
from Eq. (1) using stress variables related to the directions of wave propagation and the
particle motion. Depending on the stress variables and assumptions used, the
relationships take different forms. Among various proposed VsKo relationships,
Hatanaka & Uchida [46] suggested the Vs-equal method by combining Eq. (1) and the
mean effective stress as a stress variable as shown in Eq. (2). The significant benefit of
the Vs-equal method is that only one Vs value and vertical effective stress are required in
order to determine Ko.
/
K = 1 (2)

In this study, above VsKo relationship was used to determine Ko in the centrifuge
model. The centrifuge test was designed to assess the feasibility of using the VsKo
relationship, and to study the dependence of the Ko value on locations (in the center
versus near the boundary) and loading stages (loading-unloading-reloading).
In Fig. 6, a schematic diagram of the test is shown. The Vs was measured by bender
elements in a cross-hole configuration. In order to install bender elements, two sets of
four aluminum rods were used, with one set installed in the center and the other installed
near the boundary of the rigid box. One set of rods were fixed to the bottom of the rigid
box. Two rods were for bender elements to measure VsHH, and two others were for bender
elements to measure VsHV in order to check the suitability of each Vs for determination
of Ko. This cross-hole-type bender element testing system was developed by Kim & Kim
[45]. Bender elements were installed on the rods at a depth of 37.7 cm from the top of
the rigid box, and the tip to tip distance between a pair of bender elements at same
elevation was 15 cm. To obtain the vertical effective stresses at each location within the
rod sets, earth pressure transducers were installed on the bottom pedestals.
The model ground was prepared via air pluviation method using dry silica sand. Its
basic properties are shown in Table 1. The relative density of the prepared model ground
was 70 %, and the height was 47.7 cm, as shown in Fig. 6.
After the model soil was prepared, the rigid box was accelerated to 70 g, then
decelerated to 10 g using 10 g steps. The model was accelerated up to the 70 g level and
unloaded again, thereby completing two loading-unloading cycles. At every 10 g
increment, the shear wave velocity was measured using the bender elements. A 100 Hz
square wave was adopted as an input signal, and the received signals were stacked 64
times to increase the signal-to-noise ratio.
Table 1. Physical properties of silica sand
Material USCS !" #$% (mm) PI
Silica sand SP 2.63 1.114 0.599 0.162 NP

Figure 6. Schematic diagram of large rigid box [47]
During the test, one pair of bender elements in the center did not work properly.
Therefore, the signals that were obtained were VsHH at the center, and VsHV and VsHH near
the boundary.
The vertical effective stress is required when using Eq. (2), and it was measured at
the center and near the boundary. The measured vertical effective stresses were also
linearly interpolated to obtain the vertical effective stress at the corresponding depths of
the bender elements.
Fig. 7 shows interpolated vertical effective stresses for each location. There were
some discrepancies between the measured vertical stresses, even though the depth of the
sand was the same. The stress measured near the container boundary was less than that
measured at the center for both loading and the slightly overconsolidated unloading
stages, and this trend remained the same in both cycles. This could be explained by an
arching effect caused by side friction between the rigid box wall and the sand particles,
leading to comparatively less influence of vertical stress in the region near the container
boundary than in the center. The significance of the arching effect was shown in the
patterns of decreasing stress during unloading. In particular, the stress measured near the
container boundary decreased more slowly than the stress measured at the center during
unloading, likely because side friction affected particle interlocking and the residual
stress remained near the container boundary during the unloading stage. This
phenomenon is verified in Fig. 7(b). In this figure, the measured earth pressures are
presented as hysteresis loops for each location. The slope of the near boundary was less
than that at the center because of the influence of the arching effect. As a result, the
measured stress near the boundary was not merged with the original stress state when the
model was subjected to its final unloading stages (i.e., earth pressure at the 10 g level).
This was contrary to the stress measured at the center. It is interesting to note that the
measured stress near the boundary exceeded that at the center in the heavily
overconsolidated stage. Therefore, the OCR value at the center was greater than that near
the boundary. By measuring the vertical effective stress, it is found that stress states were
different with respect to the location within the model ground.
(a) Measured pressures with testing time (b) Measured pressures with g-level
Figure 7. Measured vertical effective stresses
In Fig. 8, the measured Vs values are presented for each location. In both locations,
Vs increased with increasing centrifugal acceleration level (g-level), and vice versa.
When the model was subjected to the reloading stage, Vs gradually went back to its
original position as the OCR approached 1. Compared to the stress measured at the center
in Fig. 7(b), the corresponding Vs did not form a closed hysteresis loop even though the
measured vertical effective stress formed a closed loop for the first loading cycle. This
indicated that Vs represented not only the vertical stress but also the horizontal stress,
and the variation in the horizontal stress was more dominant at the center. This might
have been due to a smaller arching effect caused by side friction.
(a) Center (b) Boundary

Figure 8. Measured Vs for each g-level (VsHV, VsHH) [47]
The VsHH values obtained at both locations are compared in Fig. 9. VsHH obtained at
the center was greater than VsHH obtained near the boundary for all stages. Even though
the measured stress at the center was less than that near the boundary after the slightly
overconsolidated state (Fig. 7), it is interesting to note that Vs values measured at the
center were greater than those near the boundary for all loading-unloading stages. At the
center, the variation in horizontal effective stress was more significant than that near the
boundary because of the smaller arching effect. Therefore, Vs appeared to be more
affected by horizontal effective stress, but it also showed different stress states at
different locations depending on the model conditions (i.e., loading, unloading, or
reloading).
Figure 9. Comparison of VsHH obtained at center and near boundary [47]
Fig. 10 shows Ko values determined via Eq. (2) at different locations in the model.
The measured vertical effective stresses and VsHH were utilized. The material constant
Cs and exponent n for Eq. (2) were determined using the corresponding vertical effective
stress and regression analysis. The calculated Ko values are compared with the values
obtained using empirical equations. The calculated Ko values using Eq. (2) were similar
with those found using the empirical equations for two cycles. This can be attributed to
the use of an accurate vertical effective stress to determine Ko, resulting in Vs values that
more accurately describe the change in horizontal effective stress.
(a) Center (b) Boundary

Figure 10. Ko using Eq. (2) [47]
The calculated Ko at the center was greater than the Ko near the boundary for the
unloading stages. During unloading, the measured vertical effective stress at the center
decreased more rapidly than that near the boundary because the arching effect was
significant near the boundary resulting in different OCR values at the two locations. As
Ko is mainly affected by the OCR, the Ko values at the two locations in the centrifuge
model were different depending on the stress conditions.
The VsHH and VsHV resulted in very small difference (maximum deviation of about
3%) during every loading stage as presented in Fig. 8(b). Even though the measurements
of VsHV was failed at center, it is expected that Ko values using VsHV result in almost same
value and trends with Ko values using VsHH.
In summary, Eq. (2), which required one horizontally propagating Vs (VsHV or VsHH)
and the corresponding vertical effective stress, could be adopted to determine Ko in the
centrifuge model. If the measured vertical effective stress is employed in Eq. (2),
accurate evaluations of Ko are possible regardless of the location in the model because
the measured Vs represents both horizontal effective stress and measured vertical
effective stress. Consequently, it may be concluded that Ko in the model could be
evaluated using the VsKo relationship based on both measurements of horizontally
propagating Vs (VsHV or VsHH) and the corresponding vertical effective stress.
4.2. Settlement prediction of shallow footings using Vs
Schmertmanns method [48], [49] is widely used for estimation of footing settlement in
practice. The settlement of footing as suggested by Schmertmann et al. [49] is expressed
as:
.02 3)2
s& = C C (q' ),* ) - (3)
42
Where, C1 and C2 are depth and time factors. qb is footin\!^`z,d is initial vertical
effective stress at footing base level. Izi is vertical strain influence factor, 5zi is thickness
of each sub-layer and Ei is elastic modulus of each layer.
One of the most important parameters in Schmertmanns method is the Youngs
modulus, E, of each layer. The Schmertmanns method utilizes CPT tip resistance, qc, to
evaluate the Youngs modulus using soil type based empirical correlations. The
transmitted stress into soil media caused by footing load can be calculated by strain
influence factor, Iz, with respect to depth. This strain influence factor is strongly related
with the amount of applied load spreading to the subsoil layer. Then, the settlement of
each sublayer can be calculated using the converted modulus and transmitted stress based
on elasticity theory, and finally the total settlement of footing over the soil profile is
calculated.
In the settlement prediction, the crucial factor is stiffness not the strength. However,
qc is strongly related to soil strength rather than the stiffness. Vs is directly related to
deformation properties (i.e. small-strain shear modulus, Gmaxs2), hence, Vs can
replace the role of CPT tip resistance in determining deformation modulus for settlement
predictions. The detailed procedures for adopting the Vs in Schmertmanns method is
explained below.
Fahey [50] reported that hyperbolic equations can be rewritten as a function of shear
stress rather than shear strain to describe the nonlinear stress-strain relationship. Based
on this approach, Fahey & Carter [51] proposed two additional model parameters f and
g to have the flexibility of changing the shape of the stress-strain curve, and especially
to adjust the model to force failure at finite strain. The proposed equation is as follow:
<
G8 9
G6 7 = 1 f 9:; (4)
^max is the shear strength. It is convenient to adopt this modified hyperbolic

model for the estimation of footings settlement, because transmitted stress into soil
media caused by footing load (or design load) can be derived by strain influence factor.
Modulus reduction can, thus, be considered as a mobilized stress level (e.g., G/Gmax
max) and it is more advantageous than using strain.
While the footing is loaded, the confining stress increased in the underlying soil and
this effect can be considered as a change of Gmax as:
@
G6 7 ( ) = > G6 7 ( ) (5)
?
^1 !\2 is the confining stress with

the footing load. This stress state dependent shear modulus in Eq. (5) can be used in the
settlement prediction procedure to reflect reasonable stress increment in the soil.
\`^^\!!!^$!
model in Eq. (4) and the effect of confining stress increment in Eq. (5) can be combined
and rewritten in axial stress-strain formula as:
B3
3 <
E = E6 7 1 f (6)
:;
Where, Emax !!!\`!^!!^

!$\!
Schmertmann et al. [49] proposed simple triangular strain influence factor. Instead
of using the simple triangular strain influence factor, new strain influence factors derived
from elastic theory can be utilized to improve accuracy of prediction. The new strain
influence factors can be calculated from the elastic solution using constitutive
relationship of Hookes law and integrating the Boussinesqs point loads over a
distributed area.
The transmitted stress into each sublayer by surcharge load is estimated by the strain
influence factor from the elastic solution and the modulus could then be determined
considering stress level at each depth. The deformation modulus, Eij, considering both
stress-dependent stress-strain relationship and nonlinear stress-strain hyperbolic
equation can be expressed as:

H,IBJ2 .0,I J2 .0,I <

E&D = EF,D 1 f (7)
L,I :;
^0,j is the initial vertical stress at depth zj^max is the axial strength, and E0,j
is the initial small strain Youngs modulus equal to Emax=2Gmax$\.
(3) and (7) without considerations on the depth or time factors, total settlement of each
sublayer corresponding to applied stress qi can be calculated using this nonlinear stress-
dependent modulus.
.02 3)2
s = q& - M NOQ02 R OQ U (8)
4H,2 2 ST 02
M2 M:;
Finally, this settlement prediction requires two parameters Emax from Vs max
from internal friction angle ().
Three-dimensional geotechnical centrifuge tests were designed in order to validate
the settlement prediction method stated above. Model footings having the L/B ratio of
1.0, 2.5 and 5.0 were modeled to represent square footing and rectangular footing,
respectively. Fig. 11 shows the model footing constructed in the model box. Before
filling dry silica sands in a model box, bender elements were installed in the model box
for the measurement of Vs of the model ground while accelerating the model in the
centrifuge [45]. Target g-level was 40 g and the model footings represent prototype
footing lengths of 2.4 m, 6 m and 12 m, respectively, according to scaling law. After the
1st loading, the load was completely removed and the 2nd loading was applied to a bigger
load than that of the 1st loading. In case of the last test for L/B = 5, an additional 3rd
loading was applied.
Fig. 12, shows the measured Vs profiles and interpolated Vs profiles for three
experiments using a regression analysis. As shown in the figure, three Vs profiles before
1st loading are almost identical, which means the model grounds were constructed
homogeneously and uniformly. It is interesting to note that Vs measured before 2nd
loading becomes greater than that measured before 1st loading and the depth of increased
Vs was deeper with increasing L/B ratio. This could be attributed to residual stress in
subsurface after 1st loading and the area influenced by the residual stress are deeper with
increasing L/B ratio. Therefore, the Vs profile before each loading should be used in each
settlement prediction.
Figure 11. Model configuration (cross section)
Figure 12. Vs profiles; (a) L/B = 1.0 (b) L/B = 2.5 (c) L/B = 5.0
Table 2. Soil parameters for settlement predictions

f g n 3RLVVRQUDWLR Friction angle, %HDULQJFDSDFLW\max
L/B=1.0, 2842 kPa
1.0 0.2 0.5 0.2 42 L/B=2.5, 2249 kPa
L/B=5.0, 2063 kPa
Settlement predictions were made based on Eq. (8) using soil parameters tabulated
in Table 2. Two different strain influence factor profiles were utilized to compare with
each other (Fig. 13). Centrifuge test results are compared with the predictions as shown
in Fig. 14 for only the final loading stage because the results for 1st loading showed very
soft deformable behavior regardless of the L/B ratio. It could be reasoned that particle
rearrangement, caused by interparticle slip and rotation, governs the deformation
behavior rather than the compression of the soil mass related to modulus for the 1st
loading [52]. Moreover, the model grounds were constructed by using an air pluviation
method, which results in regular packing of particle structure that is susceptible to
movements.
Figure 13. Strain influence factor profiles used in predictions
Settlement predictions using the proposed strain influence factor from elastic
solution gave good agreement with each centrifuge test result, especially for normalized
settlement of less than 2%. Based on these comparisons between load-settlement curves,
it is concluded that the proposed settlement prediction method represents the deformation
behavior reliably, according to precise modulus determination from the Vs (small strain
modulus), and modulus reduction except settlements on the virgin sand deposit. In order
to check the effect of two different strain influence factors for the developed method,
predictions using the strain influence factor from Schmertmanns triangle also are
presented in the figure.
(a) L/B = 1.0
(b) L/B = 2.5 (c) L/B = 5.0

Figure 14. Load settlement curves for final loading
In order to check and complement the reliability of the developed method, the
modulus profiles mobilized in the settlement predictions are shown with respect to the
L/B ratio in Fig. 15. Each profile is evaluated by using applied pressure (q) starting from
200 kPa. In addition, the normalized settlements corresponding to the applied pressure,
are presented as a percentage. The applied pressures were selected to induce settlements
less than normalized settlement of 4 %. General transition trends in modulus profile were
observed, where the modulus increased for low applied pressures and decreased near the
surface after applied pressure of 800 kPa due to nonlinear property of soil. The threshold
depth where the modulus becomes smaller than the previous value was deeper with
increasing L/B ratio. This phenomenon could result from shapes of the strain influence
factor derived from the elastic solution. These trends in modulus profiles show the
increase of modulus resulting from confining pressure as well as a modulus reduction by
nonlinear stress strain behavior.
(a) L/B = 1.0
(b) L/B = 2.5 (c) L/B = 5.0

Figure 15. Youngs modulus profiles
To examine the applicability of the proposed settlement prediction method in the

field, in-situ measurements from large scale plate load test (PLT) was compared with the
prediction. The test has been conducted in a highway construction site in southern
province in Korea.
In-situ VS profile in the testing site was evaluated by HWAW (harmonic wavelet
analysis of waves) method [27], [53] before preparing large scale plate load testing setup.
Fig. 16 shows the VS profile derived at a large scale plate load test site. The top layer of
the site has been confirmed as very dense silty sand (SM) and the layer is laid on a very
dense completely weathered rock. SPT was not available below 10m depth. It is difficult
to obtain the deformation characteristics of this kind of soil condition without using the
seismic method or loading test. Hence, it is strongly recommended to use the deformation
modulus determined using these methods for settlement prediction of footings.
After conducting seismic test and evaluating VS profile, 2.4m2.4m large steel plate
and loading frame was installed for the plate load test. Fig. 17 shows the testing setup.
The loading to the plate was applied using hydraulic jack and the reaction force was
supported by steel frame structures anchored in the ground. Magnetic extensometer was
installed at the center borehole and the settlement of footing was determined from the
top prove. The load was applied up to 1,500 ton so the maximum bearing pressure on the
plate was about 3,000kPa.
Figure 16. Vs profiles from two test sites Figure 17. Large scale plate load test
Figure 18. Comparison of load settlement from PLT and predictions
Fig. 18 shows the results of load settlement measurements from the test site and the
results are compared with predictions using Vs based method described in this section.
To apply the Vs based prediction method, model parameters to describe soil deformation
and the strain influence factor profile, Iz, should be selected. The model parameters used
for centrifuge tests were equally utilized in the predictions while the friction angle was
converted from SPT-N value using an empirical correlation. In case of Iz profile, Iz profile
for square footings from elastic solution was adopted as mentioned before.
The resolution of settlement measurement is quite poor and it was difficult to
generate full load settlement curve from the measurement data due to alignment or
bedding error. However, it was still possible to see the global load settlement behavior
from the measurement. The prediction does not perfectly match at initial stages, however,
the trend follows the measurement data well.
In this section, the settlement prediction method based on Vs was discussed.
Complete load settlement behavior of footing could be predicted using Vs based
Schmertmanns approach considering nonlinearity of soil. Geotechnical centrifuge tests
were conducted to simulate vertical load tests of footing with different L/B ratio. Load
settlement curves from the tests were compared with Vs based prediction method. The
predictions show reasonable agreement with test results for final loading stages. Field
experimental results also showed that the proposed method can predict the load
settlement curve of loading plate using Vs profile, even where CPT is not available.
Therefore, the use of Vs has expanded the applicability of the settlement prediction
method. Further study is required to improve the applicability of the proposed method.
The model parameters in deformation model need to be updated considering the soil
types, so refinement of model parameters of the stress strain model by conducting a series
of element tests are also required.
4.3. Evaluation of ground improvement using Vs
To assess the quality and depth of ground improvement of sand and clay layers, SPT and
CPT are often performed before and after applying countermeasures. Both methods are
intrusive and cover a small area, and hence they require substantial amount of time to
evaluate a larger area. The quality and extent of ground improvement can be evaluated
by effectively combining the Vs profile obtained from the field with the correlations from
corresponding laboratory tests. The Vs represents a large volume of soil having spatial
variation and can be also measured non-intrusively on the gravel and cobble layers which
are hard to be characterized by penetration tests. Furthermore, the Vs on particulate
materials is dependent on the effective stresses, and so the problems related to ground
improvement (i.e. increment of effective stress) including ground densification and
consolidation can be evaluated by Vs measurement; degree of consolidation can be
estimated by measuring the increment of the in-situ Vs and applying the relationship
between Vs and effective stress obtained from the laboratory tests. Vs can be used
reasonably as a relationship between Vs and effective stress for the improving clays soils
as follows:
^ ^ ^
Z Z B_ (Bà )Z
WX = Y = Y = Y (9)
[\] [\] [\]
Where, av \!!âv xyx

and y \!
!!$ \
!!!!$âv in Eq. (9) can be rewritten
\ v h, which are the vertical and the horizontal effective stresses. After
adopting coefficient of earth pressure at rest, Ko, Vs can be expressed as a function of
vertical effective stress.
As the effective stress of clays increases with consolidation, Vs can be formulated
as follows:
^ ^?
(Bà )(Z B3Z Z B3Z
WX = Y = b (10)
[\] [\]
Assuming same Ko values for laboratory and field state, the increased Vs can be
obtained using the right-hand side in Eq. (10). It is possible to evaluate the increment of
effective stress at a given time by measuring Vs from field seismic tests with Eq. (10).
Ground densification can be evaluated by combining the Vs profile obtained by in-situ
SASW tests with the correlation between the normalized Vs and density determined by
laboratory resonant column tests. In the natural soil deposit, mean effective stress
increases with increasing depth, and Vs measured at a deep layer would be larger than
that measured at a shallow layer, even if the density of the site is uniform. To assess the
density profile based on the Vs profile determined by SASW tests, the effect of
confinement should be considered in the data reduction process. In order to develop the
correlation between Vs and density while minimizing the effect of confinement, the
normalized shear wave velocity (Vs,n) was used as follows:
\ @
WX,@ = WX,c (11)
d
eg
Where, Vs,m is the Vs at a mean effective stress of m; Pa is the reference stress,
typically atmospheric pressure; and n is the stiffness parameter for Vs which is
determined as 0.25 by various laboratory tests for sands.
Kim & Park [13] and Kim et al. [54] studied the relationship between normalized
shear wave velocity (Vs,n) and density by RC and FF-RC test with respect to various
confining stress increments, and showed that the correlation between Vs,n and density is
unique and this well-defined correlation appears to support the use of the normalized Vs
in the density evaluation.
Typical examples of using Vs in the evaluation of degree of consolidation and
ground densification during ground improvement are introduced below.
4.3.1 Evaluation of degree of consolidation on soft clay deposit using Vs

In the project involving soft ground improvement, the evaluation of field degree of
consolidation on soft clays is very important for determining construction schedule, and
stability evaluation for the next construction process. To evaluate the field degree of
consolidation, Vs can be an alternative in place of monitoring settlements and pore water
pressures.
The idea using Vs for evaluation of the consolidation state on soft clays has been
investigated by previous researchers [32], [55], [56]. However, the developed evaluation
procedure has not been actively applied to the field ground improvement site.
In this regard, a study for the application of Vs for evaluating the degree of
consolidation on soft clays at Busan Newport construction area in Korea is introduced
[39]. The evaluation procedure using Vsobtained by both field and laboratory test results
were applied to the field case study for evaluating the feasibility of the proposed method.
Procedures for evaluation of soft clay improvement on the basis of Vs measurement
is explained. The subsurface condition of the site is divided into several layers based on
the subsoil exploration data. Undisturbed sample (UD) is then taken from each layer
before the site improvement. By using the consolidometer integrated with the bender
elements, Vs is measured simultaneously while performing the consolidation test. The
increment and range of stress applied to the sample are determined based on the field
loading conditions which the soil experiences during the surcharge load increment at the
site. For each stress increment, the relationship between Vs and is obtained at the end
of primary consolidation, and the relation is displayed in the log-log scale. Because even
a UD sample experiences the stress release during the sampling process, the relation is
generally divided into two regions, normally consolidated and overconsolidated regions,
as shown in Fig. 191 1, which determine the relationship between
Vs and effective stress for normally consolidated region, can be estimated from the result
obtained by laboratory test.
Figure 19. Evaluation procedure of the degree of consolidation using Vs [2]
\!1 1for each soil layer where samples were taken,
the increase of Vsowing to the increase of vertical effective stress ('v) in the field can
be calculated for each individual layer by using Eq. (10). Fig. 20 displays an example for
the evaluation of soil improvement when the surcharging load is completely transferred
to the effective stress (end of primary consolidation). If the surcharging stress is yet to
be completely transferred to the effective stressi.e., the consolidation is on-going, the
increase of effective stress (') is in relation to the average degree of consolidation (U).
The degree of consolidation can be evaluated using Eq. (10) and following equation:
3i 3Z
U= 1 = (12)
ja 3Z,k
where 'v,f is final effective stress increment by the surcharge. The degree of
consolidation of improved clays can be evaluated by applying the Vs profiles obtained
from the field to these relationships.
To evaluate the variation of Vs during ground improvement, the downhole seismic
tests were performed 5 times as consolidation progressed. Fig. 20 shows the history of
surcharge loading and downhole tests, and the settlements measured using the multi
depth extensometer near the borehole H1. All Vs profiles obtained from downhole test
are shown in Fig. 21. The boundary between sand mat and original clay is clearly shown
at all test results and they matched well with those obtained from settlement records. Fig.
21 shows that the Vs of clays increases significantly as consolidation progresses by
surcharge loads and the Vs decrease slightly after the removal of surcharge.
Figure 20. Surcharge and downhole test history with embankment settlements [2]
Figure 21. Comparison of Vs profiles obtained Figure 22. Comparison of laboratory and field
from a series of downhole tests [2] seismic test results [2]
In order to effectively compare the field measurement with the estimation based on
laboratory test results, Vs profiles of clays are marked except those of sands in Fig. 22.
After a layer of sand mat has been sprayed, a borehole was drilled to obtain two
undisturbed samples from upper highly plastic clay layer at DL-7.4 m(H1-1) and DL-
18.4 m(H1-2) using thin-walled tubes. Those samples were used to determine the
1 1 in laboratory tests. Upon using the results of H1-1 sample, the Vs
profiles obtained by the downhole tests are similar to those predicted using the laboratory
results at two surcharging conditions. It can be judged that the primary consolidations
were almost finished at the time when the downhole tests were performed. In case of H1-
2 sample, it can be judged that the primary consolidation was ended from the results of
downhole tests conducting after surcharging DL +8 m and the consolidation was still
progressing at the time when the Test No. 2 were performed. For the more accurate
evaluation, it is desirable to construct the database of the relation between the Vs and
effective stress by performing laboratory tests with significant number of specimens.
The evaluation method of the degree of consolidation of soft clays using Vs was
explained and the applicability of the method was investigated from a case study. The
field Vs profiles obtained from a series of downhole tests during consolidation of clays
increased significantly. The increased Vs profiles by field surcharging loads were
determined using Vs-' relationship obtained by laboratory tests and the degree of
consolidation of clay layers were evaluated by comparing the Vs from the laboratory and
the field seismic tests. The degree of consolidation evaluated using Vs matched well with
that obtained from field settlement records.
4.3.2 Evaluation of ground densification using Vs

The Vs determined in the field can be related to the maximum shear modulus of granular
soils at small strains which is mostly influenced by density (d) and confinement (m)
[57]. The Vsdm relationship has been frequently determined by laboratory tests such
as resonant column (RC) tests [4], [13], [30] and free-free resonant column (FF-RC) tests
[54], [58], [59]. The Vs can be measured both in the laboratory and in the field, and the
possible correlation can be determined in the laboratory and then readily applied to the
field problem. Therefore, the quality of ground densification due to compaction can be
evaluated by effectively combining field and laboratory test results.
Kim & Park [13] and Kim et al. [54] studied the relationship between normalized
shear wave velocity (Vs,n) and density by RC and FF-RC tests with respect to various
confining stress increments, and showed that the correlation between Vs,n and density is
unique and this well-defined correlation appears to support the use of the normalized
shear wave velocity in the density evaluation.
To assess the applicability of the correlation between Vs,n and density in the
evaluation of ground densification, field test was performed. The site was at Yongjong
Island, where the Incheon International Airport was being constructed (Fig. 23).
The test site consists of a reclaimed soil layer of about 4 m, a weak alluvial clay
layer of about 10 m, an alluvial sand layer, and other soil profiles below which are not
presented here. The water table was located at a depth of 2 m. After placing hydraulic
fill to form the reclaimed layer, the underlying soft clay layer was improved by vertical
drains and preloading. The layer of preload fill about 6 m thick is currently being
removed (Fig. 23). The reclaimed layer, classified as SM, required improvements to
build pavement structures for the runway, taxiway, and apron. Hydraulic hammer
compaction with a 10 t tamper and a drop height of 1.2 m was used to densify the
reclaimed layer. One test section was prepared and compacted using three passes and
2325 drops of the hammer in a grid pattern.
Figure 23. Soil profile of Yongjong Island site [13]

To evaluate the effectiveness of ground densification, SASW tests and laboratory

tests were utilized. The RC tests were performed at various densities and confining
pressures using reconstituted specimens. The Vsdm and Vs,nd relationships of the
site were determined from the RC tests (Fig. 24). The SASW tests were performed before
and after compaction. Vertical velocity transducers with a natural frequency of 2 Hz were
used as receivers and a bulldozer was used as the source. Four measurement setups with
receiver spacing of 2, 4, 8, and 12 m were used for each site with the common receivers
midpoint geometry [60]. The Vs profiles of the site were determined by iterative forward
modelling using the WinSASW computer program [61]. The Vs,n profiles of the test site
before and after compaction are shown in Fig. 25.
Combining the Vs,n profiles determined by SASW tests (Fig. 25) with the Vs,nd
relationship determined by RC tests, in-situ density profiles before and after compactions
were evaluated as shown in Fig. 25. Because the reclaimed soil layer exists at a depth of
approximately 45 m, the density variation below that level (in the alluvial clay layer)
has no meaning. The dry unit weights of reclaimed soil at the test site were increased by
about 1.96 kN/m3 at a depth of 15 m due to hydraulic compaction. Sand cone tests were
performed after hydraulic compaction at various depths and the test results are plotted in
Fig. 26 for comparison. Both results matched well, thus highlighting the great potential
of applying the proposed method for the evaluation of ground densification.
(a) Vs-lm -m relationship of the specimen (b) Vs,n-lm relationship for the specimen
Figure 24. Laboratory test (RC test) results [13]
Figure 25. Vs,n profiles before and after Figure 26. In-situ density profile before and
compaction [13] after compaction [13]
5. Conclusions
In this study, field and laboratory measuring techniques of Vs were summarized and
the applications of Vs in the analysis of various geotechnical problems were discussed
using typical examples: the evaluation of Ko value in centrifuge model, settlement
prediction of shallow foundation, and evaluation of ground improvement and
densification using Vs.
Accurate evaluations of Ko using Vs were possible regardless of the location in the
centrifuge model during various loading stages. Complete load settlement behavior of
footing could be predicted using Vs based Schmertmanns approach considering
nonlinearity of soil. The procedure for evaluation of ground improvement of soft clay
deposit using Vs measurements was introduced and a field case study was performed at
Busan Newport construction site. The degree of consolidation evaluated using the Vs
matched well with that obtained from field settlement record, showing the potential of
using Vs in the evaluation of degree of consolidation.
In summary, the Vs could be considered a key engineering property to evaluate
deformation properties and stress states of ground not only for dynamic problems but
also for conventional static problems. This means that the Vs has a great potential because
of its wide-ranging applications for solving various geotechnical problems.
Acknowledgement
This research was supported by a grant (030401) from Infrastructure and

transportation technology promotion research program funded by Ministry of Land,
Infrastructure and Transport of Korean government. This research was also supported by
a Basic Science Research Program through the National Research Foundation of Korea
(NRF) funded by the Ministry of Education, Science and Technology (Grant Number:
2009-0080575). The authors acknowledge the KREONET service provided by Korea
Institute of Science and Technology Information.
References
[1] D.S. Kim, H.J. Park, J.T. Kim, N.R. Kim, & E.S. Bang, Evaluation of spatial variations in soil stiffness
using stress wave propagations, Performance-Based Design in Earthquake Geotechnical Engineering -
From Case History to Practice (IS-Tokyo 2009), Tokyo, Japan (2009) 163-178.
[2] D.S. Kim, J.U. Youn, & H.J. Park, Session report: Applications of shear wave velocity on various
geotechnical problems, ISC4, Geotechnical and Geophysical Site Characterization, (2012) 661-673.
[3] J.B. Burland, Ninth Lauritis Bjerrum Memorial Lecture: 'Small is beautiful' - the stiffness of soils at amall
strains, Canadian Geotechnical Journal, 26 (1989), 499-516.
[4] D.S. Kim, Deformational characteristics of soils at small to intermediate strains from cyclic tests. Ph.D.
Dissertation, The University of Texas at Austin, 1991.
[5] F. Tatsuoka & S. Shibuya, Deformation properties of soils and rocks from field and laboratory tests,
Keynote Lecture, Session 1, Proceedings: 9th Asian Regional Conf. on SMFE, Bangkok, 2 (1992) 101-
170.
[6] D.S. Kim & K.H. Stokoe, Torsional motion monitoring system for small-strain (10-5 to 10-3%) soil testing,
Geotechnical Testing Journal, 17(1) (1994), 17-26.
[7] S.K. Roesler, Anistropic shear modulus due to stress-anisotropy, Journal of Geotechnical Engineering
Division, ASCE, 105(7) (1979) 871-880.
[8] D.P. Knox, K.H. Stokoe, & S.E. Kopperman, Effect of state of stress on velocity of low-amplitude shear
wave propagating along principal stress directions in dry sand, University of Texas at Austin, Report
GR82-23, 1982.
[9] D.S. Kim & J.S. Lee, Propagation and attenuation characteristics of various ground vibrations, Soil
Dynamics and Earthquake Engineering, 19(2) (2000), 115-126.
[10] D.S. Kim & H.W. Kim, Evaluation of the base condition of drilled shafts by the impact-echo method,
Geotechnical Testing Journal, 27(5) (2004) 496-503.
[11] J.B. Burland, Behavior of foundations and structures on soft ground, Proceedings: 9th ICSMFE, 2 (1977)
495-546.
[12] K.H. Stokoe, S. Joh, & R.D. Woods, Some contributions of in situ geophysical measurements to solving
geotechnical engineering problems, Proceedings: International site characterization (ISC-2) on
Geotechical and Geophysical Site Characterization, Porto, Portugal, Rotterdam, Millpress, (2004) 97-
132.
[13] D.S. Kim & H.C. Park, Evaluation of ground densification using SASW method and resonant column
tests, Canadian Geotechnical Journal, 36(2) (1999) 291-299.
[14] I.H. Chang, G.C. Cho, J.G. Lee, & L.H. Kim, Characterization of clay sedimentation using piezoelectric
bender elements, Key Engineering Materials, 321 (2006) 1415-1420.
[15] B.C. Lee & D.S. Kim, Cylindrical diaphram wall movement during deep excavation for an inground
LNG storage tank at a coastal area, International Journal of Offshore and Polar Engineering, 13(4)
(2003) 285-292.
[16] F.E. Richart, J.R. Hall, & R.D. Woods, Vibrations of soils and foundations, Prentice-Hall, Inc., 1970.
[17] S.L. Kramer, Geotechnical earthquake engineering, Prentice-Hall, Englewood Cliffs, N.J., 1996.
[18] Y.J. Mok, I. Sanchez-Salinero, K.H. Stokoe, & J.M. Roesset, In situ damping measurements by crosshole
seismic method, Earthquake Engineering and Soil Dynamics IIRecent Advances in Ground-Motion
Evaluation, ASCE (1988) 305-320.
[19] Y.J. Mok, K.H. Stokoe, & C.R. Wilson, Analysis of downhole seismic data using inverse theory,
Proceedings, Ninth World Conference on Earthquake Engineering, Tokyo, Japan, (1989) 65-70.
[20] D.S. Kim, E.S. Bang, & W.C. Kim, Evaluation of various downhole data reduction methods for obtaining
reliable Vs profiles, Geotechnical Testing Journal, 27(6) (2004) 585-597.
[21] E.S. Bang & D.S. Kim, Evaluation of shear wave velocity profile using SPT based uphole method, Soil
Dynamics and Earthquake Engineering, 27 (2007) 741-758.
[22] S. Ohya, K. Ogura, & T. Imai, The suspension PS velocity logging system, Offshore Technology
Conference, (1984).
[23] K.H. Stokoe, S.G. Wright, J.A. Bay, & J.M. Roesset, Characterization of geotechnical sites by SASW
method, Geophysical characterization of sites, (1994) 15-25.
[24] C.B. Park, R.D. Miller, & J. Xia, Multi-channel analysis of surface waves, Geophysics, 64(3) (1999) 800-
808.
[25] P. Gabriels, R. Snieder, & G. Nolet, In situ measurements of shear-wave velocity in sediments with
higher-mode Rayleigh waves, Geophysical prospecting, 35(2) (1987) 187-196.
[26] M.C. Matthews, V.S. Hope, & C.R.I. Clayton, The use of surface waves in the determination of ground
stiffness profiles, Proceedings of the ICE-Geotechnical Engineering, 119(2) (1996) 84-95.
[27] H.C. Park & D.S. Kim, Development of seismic site characterization method using harmonic wavelet
analysis of wave (HWAW) method, International Site Characterization, ISC-2 Porto, Portugal, (2004)
767-774.
[28] J.N. Louie, Faster, better: shear-wave velocity to 100 meters depth from refraction microtremor arrays,
Bulletin of the Seismological Society of America, 91(2) (2001) 347-364.
[29] D.S. Kim, E.S. Bang, & H.J. Park, Round robin test for comparative study of in-situ seismic tests, 4th
International Conference on Geotechnical and Geophysical Site Characterization (ISC'4), Porto de
Galinhas, Brazil, (2012).
[30] V.P. Drnevich, Recent developments in resonant column testing, Richart Commemoration Lectures,
ASCE Annual Meeting, (1985) 79-107.
[31] W.M. Isenhower, Torsional simple shear/resonant column properties of San Francisco Bay mud, M.S.
thesis, The University of Texas at Austin, 1979.
[32] D.J. Shirley & L.D. Hampton, Shear wave measurements in laboratory sediments, Journal of the
Acoustical Society of America, 63(2) (1978) 607-613.
[33] J.U. Youn, Y.W. Choo, & D.S. Kim, Measurement of small-strain shear modulus Gmax of dry and
saturated sands by bender element, resonant column, and torsional shear tests, Canadian Geotechnical
Journal, 45(10) (2008) 1426-1438.
[34] R. Dyvik & C. Madshus, Lab measurements of Gmax using bender elements, Advances in the art of testing
soils under cyclic conditions, ASCE, (1985) 186-196.
[35] G. Viggiani & J.H. Atkinson, Interpretation of bender element tests, International Journal of Rock
Mechanics and Mining Sciences and Geomechanics Abstracts, 8(32) (1995) 373A.
[36] J.S. Lee & J.C. Santamarina. Bender elements: performance and signal interpretation, Journal of
Geotechnical and Geoenvironmental Engineering, 131(9) (2005) 1063-1070.
[37] E.C. Leong, S.H. Yeo, & H. Rahardjo, Measuring shear wave velocity using bender elements,
Geotechnical Testing Journal, 28(5) (2005) 488.
[38] J.C. Santamarina, K.A. Klein, & M.A. Fam, Soils and waves. John Wiley & Sons Ltd., New York. (2001)
238-282.
[39] J.U. Youn, J.T. Kim, Y.W. Choo, & D.S. Kim, Evaluation of the degree of consolidation using laboratory
and field seismic tests for soft clay deposits, Proceedings: IS-Atlanta, (2008) 215-220.
[40] T. Shibata, K. Kita, S. Kobayashi, & A. Yashima, Performance of shaking table tests and measurement
of shear wave velocities in a centrifuge, Centrifuge, 91 (1991) 391-398.
[41] W.B. Gohl & W.D.L. Finn, Use of piezoceramic bender elements in soil dynamics testing, Recent
Advances in Instrumentation, Data Acquisition and Testing in Soil Dynamics, ASCE, (1991) 118-133.
[42] R. Arulnathan, R.W. Boulanger, B.L. Kutter, & W.K. Sluis, New tool for shear wave velocity
measurements in model tests, Geotechnical Testing Journal, 23(4) (2000) 444-453.
[43] L. Fu, Application of piezoelectric sensors in soil property determination, Ph.D. Dissertation, Case
Western Reserve University, 2004.
[44] K.I. Rammah, M.A. Ismail, & M. Fahey, Development of a centrifuge seismic tomography system at
UWA, Physical Modelling in Geotechnics: ICPMG'06, (2006) 229-234.
[45] N.R. Kim & D.S. Kim, A shear wave velocity tomography system for geotechnical centrifuge testing,
Geotechnical Testing Journal, 33(6) (2010) 434-444.
[46] M. Hatanaka & A. Uchida, A simple method for the determination of Ko value in sandy soils, Soils and
foundations, 36(2) (1996) 93-99.
[47] H.I. Cho, H.J. Park, D.S. Kim, & Y.W. Choo, Evaluation of Ko in centrifuge model using shear wave
velocity, Geotechnical Testing Journal, 37(2) (2014) 255-267.
[48] J.H. Schmertmann, Static cone to compute static settlement over sand, Journal of Soil Mechanics &
Foundations Division, (1970).
[49] J.H. Schmertmann, P.R. Brown, & J.P. Hartman, Improved strain influence factor diagrams, Journal of
the Geotechnical Engineering Division, 104(8) (1978) 1131-1135.
[50] M. Fahey, Shear modulus of cohesionless soil: variation with stress and strain level, Canadian
Geotechnical Journal, 29(1) (1992) 157-161.
[51] M. Fahey & J.P. Carter, A finite element study of the pressuremeter test in sand using a nonlinear elastic
plastic model, Canadian Geotechnical Journal, 30(2), (1993) 348-362.
[52] G. Mesri & B. Vardhanabhuti, Compression of granular materials, Canadian Geotechnical Journal, 46(4)
(2009) 369-392.
[53] H.C. Park & D.S. Kim, Evaluation of the dispersive phase and group velocities using harmonic wavelet
transform, NDT & E International, 34(7) (2001) 457-467.
[54] D.S. Kim, M.K. Shin, & H.C. Park, Evaluation of density in layer compaction using SASW method, Soil
Dynamics and Earthquake Engineering, 21(1) (2001) 39-46.
[55] M. Fam & J.C. Santamarina, A study of consolidation using mechanical and electromagnetic waves,
Geotechnique, 47(2) (1997) 203-219.
[56] I.H. Chang & G.C. Cho, A laboratory procedure to characterize reclaimed clay deposits using shear wave,
Geotechinical special publication, 164 (2007) 29-37.
[57] B.O. Hardin, The nature of stressstrain behavior of soils, Proceedings of the Earthquake Engineering
and Soil Dynamics Conference, 1 (1978) 3-90.
[58] D.S. Kim, G.C. Kweon, & K.H. Lee, Alternative method of determining resilient modulus of compacted
subgrade soils using free-free resonant column test, Transportation Research Record, 1557 (1997) 62-
69.
[59] J.G. Vaghela & K.H. Stokoe, Small-strain dynamic properties of dry sand from the free-free resonant
column, Geotechnical Engineering Report GT95-1, Geotechnical Engineering Center, University of
Texas at Austin, 1995.
[60] S. Nazarian & K.H. Stokoe, Use of surface waves in pavement evaluation, Transportation Research
Record 1070, National Research Council, Washington, D.C., (1984) 132-144.
[61] S.H. Joh, Users guide to WinSASW, a program for data reduction and analysis of SASW measurements,
The University of Texas at Austin, 1992.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-222
Hydro-Mechanical Behaviour of Shales

Lyesse LALOUI 1, Alessio FERRARI and Valentina FAVERO
Swiss federal Institute of Technology, Lausanne, EPFL, Switzerland
Abstract. The geomechanical behaviour of shales is quickly becoming one of the

most important issues in modern geomechanics, largely driven by the geological
storage of nuclear waste, the extraction of shale gas and the sequestration of CO 2 .
In this context, fundamental issues come along with the complex multiphysical
conditions and a deep understanding of the hydro-mechanical behaviour of shales
becomes of primary significance. In this paper selected results on the high pressure
oedometric compression behaviour and on the water retention behaviour of shales
are presented and discussed.
Keywords. Shales, hydro-mechanical behaviour, water retention properties.
1. Introduction
The involvement of shales in new energy-related fields such as the extraction of shale
gas and shale oil, the deep geothermal energy capture, the sequestration of CO 2 and the
nuclear waste geological storage, has raised a new and growing interest in the
geomechanical behaviour of this material. In this context, fundamental issues come
along with the complex multiphysical conditions and a deep understanding of the
hydro-mechanical behaviour of shales becomes of primary significance. In fact, the
mechanical properties and the permeability are fundamental aspects in determining the
production capacity of shale gas reservoirs or the sealing potential when the shale
formation is considered as cap rock (e.g. in the case of CO 2 sequestration) or host rock
(e.g. in the case of nuclear waste geological storage).
The quantification of the hydro-mechanical properties of the shale formation is
fundamental to assess the engineering performance of the material during tunnels
excavation, drilling operations, gas extraction, CO 2 injection or nuclear waste
emplacement. The water retention mechanism also plays a primary role in determining
the hydro-mechanical response of the involved material. Phenomena such as the fluid
trapping due to the capillary forces present in low permeability formations and the
resaturation of the shale formation after ventilation (as in the case of deep geological
repositories) contribute to develop hydro-mechanical couplings that cannot be
neglected. In fact, wetting and drying episodes have significant impacts on shale
volumetric behaviour: swelling/shrinkage of shales may be caused by changes in
suction or variations in the degree of saturation; additionally, significant swelling
pressures may develop when volumetric expansion upon wetting is prevented.
In this paper selected results on the high pressure oedometric compression
behaviour and on the water retention behaviour of shales are presented and discussed.
1
Corresponding Author.
L. Laloui et al. / Hydro-Mechanical Behaviour of Shales 223
2. Tested Shales
The experimental results presented in this paper are obtained on three Swiss shales.
Two of the considered shales come from a deep geothermal well near the village of
Schlattingen in the Molasse Basin in Northern Switzerland; one belongs to the
Opalinus Clay formation and one to the so-called Brown Dogger formation. The
third shale considered is the Opalinus Clay which comes from the Mont Terri
Underground Rock Laboratory (URL) in Northern Switzerland. The Opalinus Clay
from Schlattingen (OPA-deep) is the deepest shale formation considered in this study
(depth of 837.44 - 891.25 m); it consists mainly of claystone with a carbonate content
of 10-20% and a quartz content in the range of 15-32%. The Brown Dogger from
Schlattingen (BD-deep) was extracted at a depth of 766.67-807.44 m and consists of
silty to clay-rich marls with 30-50% carbonates and 20-25% quartz. The Opalinus Clay
from the Mont Terri URL (OPA-shallow) from the shaly facies consists of dark grey
silty, calcarerous shales with a typical content of 25-30% carbonate, 5-10% quartz and
10-15% biodetritus. Index properties for the tested core samples are summarized in
Table 1, which reports the average particle density ( s ), the bulk density (), the
gravimetric water content (w), the void ratio (e), the degree of saturation (S r ) and the
Atterberg limits (the liquid limit w L and the plastic limit w P ). The OPA-deep cores
present two different densities; the most superficial cores, indicated as OPA-deep,
have a higher void ratio and water content compared to the deepest samples (OPA-
deep). Mineralogical analysis of the cores showed that the OPA-deep with higher
porosity presents reduced quartz content (approximately 17%) with respect to the
denser cores (typical value about 28%); also they have slightly higher Atterberg limits.
The porosity network has been analysed, in terms of pore size density (PSD)
functions, by means of Mercury Intrusion Porosimetry (MIP) technique. Unimodal
PSD functions were found for all the tested shales, showing pore modes at
approximately 20 nm for the OPA-shallow and BD-deep and at 8 nm for the OPA-
deep.
Synthetic waters were used in the experiments in order to recreate the in-situ pore
water compositions for the cores that were extracted at Mont Terri [1] and at the
Schlattingen site [2]. The osmotic suction of these synthetic waters was measured by a
chilled mirror dew-point psychrometer and resulted in values of 1.2 MPa for the Mont
Terri synthetic water (OPA-shallow) and 0.98 MPa for the Schlattingen synthetic water
(BD-deep and OPA-deep) [3].
Results on the high pressure oedometric compression behaviour of OPA-shallow
and BD-deep are reported here, while the water retention behaviour of all the presented
shales was investigated and the results are presented in the paper.
Table 1.Geotechnical characteristics of the tested shales.
Shale Depth s w e Sr wL wP
m Mg/m3 Mg/m3 % - % % %
OPA-shallow ~300 2.74 2.46 6.9 0.21 92 38 23
BD-deep 767-807 2.75 2.5-2.56 2.7-4.4 0.09-0.12 69-99 25-33 10-23
OPA-deep 837-838 2.74 2.49 4.3-4.9 0.14-0.15 80-94 36-39 22-25
OPA-deep 855-891 2.71 2.54-2.55 3.3-3.6 0.09-0.11 93-96 29-33 19-22
224 L. Laloui et al. / Hydro-Mechanical Behaviour of Shales
3. Experimental procedure for high pressure oedometric testing
High pressure oedometric testing allows deriving significant information on the hydro-
mechanical behaviour of shales at in-situ conditions. To this aim, a multi-purpose
oedometric cell was developed [4]. The layout of the apparatus is shown in Figure
1.The cell is designed to hold cylindrical specimens (12.5 mm in height and 35 mm in
diameter). The oedometer cell is inserted into a rigid stainless steel frame. The loading
ram is positioned in the lower part of the system to prevent the specimen from being
loaded before the test starts. The vertical load is applied by a hydraulic jack connected
to a pressure/ volume controller. The relationship between the controller pressure and
the vertical stress on the specimen was assessed through a calibration step, which took
into account the friction developed within the system [5]. The maximum vertical stress
that can be imposed on the specimen is 100 MPa.
The volumetric strains are measured by three LVDTs (with a resolution of 1 m),
which are fixed to the frame and are in contact with the loading ram. Two
pressure/volume controllers are used to control the pore water pressure at the bottom
and top bases (up to 2 MPa) and to measure the volume changes of the pore water. The
preparation of the specimen is carried out in a controlled temperature and controlled
relative humidity room in order to avoid the disturbance of the material.
Figure 1.The high-pressure oedometric cell used to perform tests on shales in saturated conditions [4].
4. Experimental methodology for studying the water retention behaviour of shales
The experimental technique presented here for investigating the water retention
behaviour of shales is based on the direct control of the water content and on the
subsequent measurement of the total suction [3]. Different hydraulic paths are
considered in order to analyse the response of the material in wetting and drying
episodes. The following sections describe the specimen preparation technique, the
procedure for the control of the water content, the suction measurement and the volume
change detection.
4.1. Specimen preparation technique and procedure for water content control
The specimens for water retention investigation are obtained from a shale core: slices
of approximately 7 8 mm in height are cut with a diamond saw without using water
(to preserve the initial condition), and they are immediately divided into smaller
specimens with lateral sizes of approximately 20 mm. Three specimens are tested to
obtain the initial condition in terms of water content, void ratio and total suction.
Firstly, the main wetting and drying paths are sought; to this aim the material is
initially brought to a dry state or to a zero-matric-suction state, respectively. The dry
state is obtained by placing the specimens in silica gel desiccators and leaving them to
dry for approximately three weeks (applied total suction of approximately 300 MPa)
until no significant weight changes can be detected. The zero-matric-suction
condition is achieved by placing the specimens in a sealed-glass jar, within which a
relative humidity of 100% is imposed; the specimens are wrapped in filter paper and
rest on porous stones immersed in water. The specimens were hence wetted by
capillarity action, allowing for volume changes.
The weight evolution over time is monitored to assess the equilibrium, and the
process is stopped once no further evolution is observed (usually in three weeks). Once
equilibration is achieved, the main wetting and drying paths can be determined. To
obtain the main drying path, the equalized specimens are air-dried under laboratory-
controlled conditions (T = 22 C, RH 45%, equivalent total suction of about 110
MPa) or placed in a desiccator with silica gel for the driest states. During the drying,
the weight of the specimens is monitored using a precision balance (0.001 g), and the
process is stopped once the target water content is reached. The specimens are then
packed hermetically for three days to allow for internal redistribution of the water
content. The procedure to obtain the main wetting path consists on the addition of a
target amount of water on the surface of the specimen with a syringe. Each specimen is
placed in a hermetic container for three days to allow for the internal redistribution of
the water.
4.2. Suction and volume measurements
The total suction of each specimen is measured using the chilled mirror dewpoint
psychrometer. The psychrometer gives an indirect measure of the total suction of a
sample by reading the relative humidity established by the sample in a closed
environment. The total suction is then obtained by the psychrometric law [6] which
relates the relative humidity (RH) and the absolute temperature (T) to the total suction:
w RT
! ln RH
(1)
Mw
!\ w and M w are the density and the molecular

mass of water, respectively. When the retention properties are sought in terms of
degree of saturation, the measurement of the volume change upon water content
evolution is required. This determination allows for the quantification of shale
swelling/shrinkage behaviour related to suction changes under unstressed conditions. In
this case, the specimen volume is measured just after the measurement of the total
suction with a destructive technique. The method adopted for the volume measurement
is based on a fluid displacement technique using a pycnometer filled with kerdane.

Kerdane was selected for its immiscibility with water and its ability to invade the air-
filled pore spaces of the surface without affecting the soil structure [7].
5. Results on the high pressure oedometric compression behaviour of shales
The high pressure oedometric tests were performed by incremental loading and attained
a maximum vertical stress equal to 30 MPa and 100 MPa for the OPA-shallow and the
BD-deep, respectively. The high level of applied vertical stress was required to observe
the transition from the over- to the normally-consolidated state of each material. The
specimens were initially saturated in isochoric conditions, by applying a back pressure
in the range of 2050 kPa, and the developed swelling pressure was measured.
Afterward, two loading-unloading cycles were performed in steps and in drained
condition, allowing for the complete dissipation of the excess pore water pressure
generated by the oedometric compression.
Figure 2 depicts the total settlements of the OPA-shallow and the BD-deep,
respectively, which are measured at the end of each loading step and corrected in light
of the deformation of the apparatus. The results are presented in terms of volumetric
strain versus vertical effective stress.
Figure 2 also depicts the volumetric strain relative to the end of the primary
consolidation. The difference between the total and the end-of-primary strains is
associated with the secondary consolidation component. The average computed
coefficients of secondary compression were 0.0006 for the OPA-shallow and 0.0004
for the BD-deep. The end-of-primary strains are used to compute the compression and
the swelling indexes and to assess the vertical effective yield stresses.
0.00 initial void ' ~ 2.5 MPa 0.000 Cc = 0.0016

ratio = 0.25 vc
initial void ' ~ 15 MPa
vc
ratio = 0.11
Cs = 0.007 0.005
Volumetric strain
Volumetric strain
Cc = 0.038 Cc = 0.011
0.02
Cs = 0.016 Cc = 0.005
0.010
0.04
0.015
end-of-primary end-of-primary
total total
0.06 0.020
0.1 1 10 100 0.1 1 10 100
Vertical effective stress (MPa) Vertical effective stress (MPa)
Figure 2.Results of the oedometric tests for OPA-shallow (left) and BD-deep (right) [4].
6. Results on the water retention behaviour of shales
6.1. Main drying and wetting paths
The main wetting and drying paths obtained for the OPA-shallow are shown in Figure
3. The evolution of the water content with the total suction is shown in Figure 3a; each
point of the curve represents an average value of three specimens, which were brought
to the same target water content value. The initial condition corresponds to a water
content of 6.7%. For the specimens equalized to the "zero matric suction" state, the
measured total suction was 1.4 MPa, which is in good agreement with the osmotic
suction of the synthetic water used for this initial wetting. The points obtained by the
controlled drying procedure describe a progressive increase in total suction along with
a reduction in water content; the driest state obtained with this procedure corresponds
to a water content of 1.9% and a measured total suction of 134 MPa. This drying path
tends to the residual condition that was achieved in the desiccator (w = 1.0%, suction =
213 MPa), when the initial state for the main wetting path was sought. The retention
curves in terms of the degree of saturation were obtained by combining the volume
evolution and the water content along the wetting and drying paths. The results have
been fitted with a Van Genuchtens type [8] expression (Figure 3b).
Figure 3.Water retention behaviour of the OPA-shallow shale in terms of water content (a) and degree of
saturation (b) as a function of the total suction [3].
The initial state for the OPA-shallow corresponds to a degree of saturation of 95%;
this value seems to confirm the fact that the material is reported to be close to
saturation in situ; in addition, the initial state is found to be located along the main
drying path as a consequence of the coring process and the exposure to the atmosphere
before preservation. The main drying path allows for the identification of the maximum
total suction value that the material can sustain without significant changes in the
degree of saturation (air entry value, ! e ). For the OPA-shallow, this value is 11 MPa
(taken as the total suction corresponding to 95% of the degree of saturation), and the
corresponding void ratio is 0.19. The material shows a hysteretic behaviour which is
significant for total suction values lower than 100 MPa.
Similar behaviours were obtained for OPA-deep and BD-deep. The initial states
correspond to a water content of 3.2% for both the BD-deep and OPA-deep shale
samples. For the BD-deep and OPA-deep specimens that were equalized to the "zero
matric suction" state, the measured total suction was 1.2 MPa, which is in good
agreement with the osmotic suction of the synthetic water. The initial conditions
correspond to a degree of saturation of 80% for BD-deep and 54% for OPA-deep. The
registered low initial degree of saturation for OPA-deep is attributed to a non-perfect
preservation of the core used in the investigation. The tested shales reveal that very
limited changes in the gravimetric water content are required to undergo complete
cycles of wetting and drying, especially for the deeper shales.
6.2. Void ratio changes
The volume determination was carried out after the total suction measurement. The
obtained results are depicted in Figure 4 in terms of void ratio versus total suction.
Significant swelling/shrinkage behaviour was observed with suction variation. The
values of the void ratio obtained for the lowest measured suction are in very good
agreement with the results of free swelling tests (also shown in Figure 4) carried out
with the same synthetic water used for determining the water retention curves. The
shrinkage limits (here defined when e/! < 0.1 %) were identified at suction value of
approximately 10 MPa for OPA-shallow and 14 MPa for BD-deep and OPA-deep, and
they were found to be approximately 4.1% for the BD-deep and 6.4% and for the OPA-
shallow and OPA-deep. The void ratio evolution with suction was fitted with the
following incremental expression:
# " ref $ m
"
e % & 1 e0
(2)
K ref ' " (
in which K ref , ! ref and m are the fitting parameters and e 0 is a reference void ratio. The
values of the parameters obtained by the least square method are reported in Table 2.
Only few results were collected on OPA-deep, due to the limited quantity of available
material, and they are shown in Figure 4. The collected points align well with the trend
registered for the BD-deep shale.
Figure 4.Void ratio evolution as a function of the total suction [3].
The analysis of the volumetric response to suction changes allows for the quantification
of the swelling potential of the tested shales. The measured volumetric deformations
associated with the complete wetting of the material, starting from the driest condition,
were 12.3%, 11.4% and 8.6%, for OPA-shallow, BD-deep and OPA-deep, respectively.
The highest swelling potential was observed for OPA-shallow; this behaviour can be
explained by the higher clay content of this shale compared to the deep shales; in
addition, greater depth and diagenetic processes, which result in a higher degree of
cementation, might be responsible for the lower swelling potential of the deepest shales.
Table 2. Values of fitting parameters for volumetric behaviour.
Shale K ref ! ref m e0
MPa MPa - -
OPA-shallow 14.66 1.62 2.63 0.31
BD-deep 29.00 1.03 1.50 0.20
OPA-deep 45.00 0.98 1.30 0.22
7. Conclusions
This paper presented selected results on the high-pressure oedometric compression

behaviour and on the water retention behaviour of shales. Results from samples of
different shales, such as Opalinus Clay and Brown Dogger, have been presented.
Testing different shales coming from different depth, allowed to capture the important
features of their mechanical and retention behaviour. This information is highly
relevant when considering that these materials are host rock candidates for the deep
geological disposal of high-level nuclear waste. As a consequence, the hydro-
mechanical performance and the retention capacity will be fundamental aspects during
the construction of the tunnel, the ventilation (drying) and resaturation (wetting) phases
of the repository, as well as in the long term storage of the radioactive waste.
Results from high-pressure oedometric tests on shales were summarized.
Attainment of a vertical effective stress in the range of 30 100 MPa was needed to
measure the vertical yield stress. The presented technique for the water retention
analysis of shales is capable of capturing the important features of the retention
behaviour, such as the hysteresis domain and the air entry value. The assessment of the
volume change due to suction variation has allowed (i) expressing the retention
behaviour in terms of degree of saturation, (ii) capturing the main drying and wetting
paths, and (iii) highlighting the significant changes in porosity undergone by the shales
along wetting and drying hydraulic episodes.
Acknowledgement
The support of the Swiss National Cooperative for the Disposal of Radioactive
Waste (NAGRA) for this research is acknowledged.
References
[1] F.J. Pearson. Opalinus Clay experimental water: A1 Type, Version 980318. PSI Internal report TM-44-
98-07, Paul Scherrer Institut, Villigen PSI, Switzerland. 1998.
[2] U. Mder. Recipe and preparation of a simplified artificial pore water for Opalinus Clay and Brown
Dogger. NAGRA AN (1998), 11-159.
[3] A. Ferrari, V. Favero, P. Marschall and L. Laloui. Experimental analysis of the water retention behaviour
of shales. International Journal of Rock Mechanics and Mining Sciences 72 (2014), 61-70.
[4] A. Ferrari & L. Laloui. Advances in the testing of the hydro-mechanical behaviour of shales. In
Multiphysical Testing of Soils and Shales: 57-68. Springer Berlin Heidelberg. 2012.
[5] S. Salager, A. Ferrari and L. Laloui. New experimental tools for the characterization of highly
overconsolidated clayey materials in unsaturated conditions. In L. Laloui, editor, Mechanics of
unsaturated geomaterials: 113-126. John Wiley & sons. 2010.
[6] D.G. Fredlund and H. Rahardjo. Soil Mechanics for Unsaturated Soils. John Wiley and Sons. 1993.
[7] H. Peron, T. Hueckel, and L. Laloui. An improved volume measurement for determining soil water
retention curves. Geotechnical Testing Journal 30(1) (2007), 18.
[8] M.T. Van Genuchten. A closed-form equation for predicting the hydraulic conductivity of unsaturated
soils. Soil science society of America journal 44(5) (1980), 892-898.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-231
Improved laboratory techniques for

advanced geotechnical characterization
towards matching in situ properties
Antnio VIANA DA FONSECA1, Cristiana FERREIRA1, Marisa SOARES1 and Assaf
KLAR2
1
CONSTRUCT-GEO, Faculty of Engineering (FEUP), University of Porto, Porto,
Portugal
2
Fac. Civil & Env. Eng., Technion - Israel Institute of Technology, Haifa, Israel
Abstract. Laboratory tests are well recognized as highly appropriate for

defining the engineering properties of geomaterials, in terms of
constitutive law parameters for modeling geotechnical engineering
problems. The strong development of advanced techniques, both in
equipment and in data interpretation, has increased the confidence in
laboratory testing, while on the other hand the limitations due to the
quality of soil sampling with depth and the spatial representativeness of
the samples are less consensual. Still, the development of new methods
for assuring high quality samples is increasing, together with sampling
quality assessment by non-destructive methods using vibration wave
velocities. Interpretation methods of in situ tests for ground
characterization has also evolved significantly, increasing the reliability
of these methods. Their versatility to cover large areas on site and the
fact that these tests are, in principle, performed at the actual state
(physical and stress) conditions, as well as the improvements in the
correlations between field tests and hydraulic and geomechanical
parameters, allows joining the quality of data and theoretical
approaches, namely through critical state soil mechanics. Current
techniques are usually associated either with very low stress-strain
levels, such as in geophysical surveys, or with very high stress-strain
levels, near failure, as in dynamic penetration tests. This practice means
that the complete range of stress-strain response is rarely covered in the
investigation. Exceptions can be made when using the pressuremeter
test, especially the self-boring technique, although time-consuming and
expensive. New research trends are making use of a single technology
for characterization at different scales (e.g. element, layer and global
characteristics), which is the case of the use of high-resolution fiber
optic distributed sensing technology for in situ moduli profiling and in
laboratory element testing.
Keywords. In situ and laboratory test, sampling quality, non-destructive methods.

232 A. Viana da Fonseca et al. / Improved Laboratory Techniques
1. Introduction
This paper discusses some of the aspects that can and should enable the association
of ground characterization from in situ testing and from advanced laboratory testing, as
a key approach for calibrating both field and laboratory derived parameters, particularly
when one or the other cannot be simultaneously performed, at least extensively.
A laboratory-based characterization of soil behavior is directly dependent on the
samples selected for testing, and therefore dependent on their representativeness and
quality in relation to the in situ conditions. The industrial investment in new high
performance samplers has been significant, proving its importance, not only for granular
soils silty, sandy and gravelly materials where the preservation of the natural structure
is challenging, but also for clayey soils which are quite sensitive to the sampling
operations. Recent studies (e.g. Gasparre et al., 2014) demonstrated that high-quality
sampling, preservation and specimen preparation are primary key factors affecting
reliable stiffness measurements on stiff clays at small strains.
In granular soils, Gel-push sampling (GP-S) has been adopted as an economical
approach to obtain high quality undisturbed samples without resorting to expensive
ground freezing. The technology has been developed over the last decade in Japan
(Kazuo & Kaneko, 2006), and also trialed in Taiwan (Huang et al., 2008) and more
recently, since 2013, introduced at KGHM Zelazny Most (ZM) copper tailings disposal
located in South-West Poland, as described by Jamiolkowski & Masella (2015). After
the Mw 6.2 earthquake of February 22nd 2011 that struck beneath the city of
Christchurch, New Zealand, a large research project was led by the University of
Canterbury to characterize the engineering behavior of the soils in the region comprising
in situ tests (cone penetration test, CPT, borehole drilling, shear wave velocity profiling)
and gel-push sampling followed by a program of laboratory testing including monotonic
and cyclic testing of the soils (Cubrinovski et al., 2011). This, still on-going work, has
made use of the gel-push piston sampler at two trial locations in the CBD with great
success (Taylor et al. 2012), using a 70 mm internal diameter Osterberg-type fixed piston
sampler, modified for use with the polymer gel.
Recently, Pineda et al. (2014a) presented the development of an innovative direct-
push sampler specifically designed to reduce sample disturbance and maximize sample
retention. The sampler performance for testing soft clays was proved, using laboratory
tests as well as CAT scans to evaluate the reliability of the sampler and also to estimate
sample disturbance.
It is also very important to be able to assess the quality of the samples, in order to
guarantee that only high-quality samples are used for extensive laboratory
characterization. Different methods for the assessment of sampling quality have been
proposed over the years relying on: fabric inspection; measurement of initial mean
effective stress, p (Ladd & DeGroot, 2003); measurement of strains during
reconsolidation (Lunne et al., 1997a,b); comparison of in situ and laboratory
measurements of seismic wave velocities (Shibuya, 2000; Nash et al., 2006; Landon et
al., 2007; Sukolrat et al., 2008, Ferreira et al., 2011). It is worth highlighting that among
the commonly used methods for assessing sample quality, the last method is the only one
capable of effectively considering the effects of destructuration in soils with low
reconsolidation strains and conditions ahead of the advancing borehole during drilling
operations; penetration of the sampling tube and sample retrieval to ground surface;
water content redistribution in the tube; extrusion of the sample from the tube; drying
and/or changes in water pressures; stress relief due to the removal of the sample from the
A. Viana da Fonseca et al. / Improved Laboratory Techniques 233
ground to zero total stress state in the laboratory; and trimming and other processes
required to prepare specimens for laboratory testing (Viana da Fonseca et al., 2011).
Another subject to be addressed is the natural soil variability along the tube, e.g.
presence of shells or inclusions with similar size as soil particles, which may affect
laboratory tests and lead to important discrepancies between in situ and laboratory data
(Pineda et al., 2014b). Visual inspection of the sample quality is only possible after
complete soil extrusion which is not practical since the time required for visual inspection,
sample selection and preparation (or sealing) is too long and may affect the initial state
of the soft clay (e.g., due to drying). Non-conventional techniques such as X-Ray or
Computer Axial Tomography (CAT) analyses are becoming popular in geotechnical
engineering due to their non-destructive nature and simple procedure. The main
drawback of the X-Ray technique is that all features of the specimen are superimposed
into a 2D image. On the other hand, a 3D reconstruction is obtained from CAT analysis.
This will be described in what follows, together with other important subjects, such as
the chemical interaction between the pore fluid constituents and the solid phase in natural
soft clays, from which the determination of index properties in previously dried soils is
strongly affected by this preparation process. Another issue is the advantage of
geochemical analysis (electrical conductivity, EC, and resistivity, ER, pH, cation/anion
analysis, XRD analysis, organic matter content, specific surface) as well as
microstructural studies (scanning electron microscopy -SEM or ESEM-; mercury
intrusion porosimetry MIP-) to give important clues for the expected macroscopic
response of the soil (Pineda et al., 2014b). Soil electrical conductivity (EC) is relevant
in case of natural clays coming from estuaries and river systems where the pore water is
saline as proved by Pineda et al. (2014b).
The effects of laboratory specimen preparation have also been investigated by
measuring seismic wave velocities in high-quality block samples before and after
trimming cylindrical specimens (Besenzon, 2013). An average drop of 10% in shear
wave velocity was measured in the cylindrical specimens, compared to that measured in
the original block sample, evidencing minimal disturbance. However, the last tested
specimens, over two months after trimming from the block, showed a 30% reduction of
shear wave velocity, indicating that ageing during storage is an influential parameter in
the shear stiffness of a soil, as also pointed out by Ferreira (2009).
Some soils, however, have proved to be extremely difficult to effectively sample in
undisturbed conditions, namely tailings, especially under the water table. For this reason,
laboratory testing of these geomaterials is generally conducted on reconstituted
specimens, for which the selection of the most adequate sample preparation method is
crucial to capture and replicate the mechanical behavior of the undisturbed soil. Mulilis
et al. (1977) showed that laboratory specimen preparation has a significant effect on the
measured cyclic resistance, on account of the arrangement of particles or soil fabric
created by moist tamping or air pluviation. Layering within a specimen induced by the
preparation method has also been observed to affect the cyclic resistance (Yoshimine
and Koike, 2005). Chang et al. (2011) compared the structure and fabric of gold tailings
in undisturbed and reconstituted moist tamped and slurry deposited specimens, reporting
evidence of clear differences in behavior. Time since deposition has also a significant
effect on soil response. This influence has been collectively attributed to ageing effects,
and is thought to be due to a possible combination of stress history, mechanical creep,
and the development of physico-chemical bonding between particles (Mitchell, 1976).
Therefore, in order to characterize the mechanical behavior of a natural sandy deposit, it
is essential to perform laboratory testing on high quality undisturbed samples where the
natural fabric and any associated ageing effect are preserved (Taylor et al., 2013).
Accurate control of the soil state conditions is also fundamental in advanced
laboratory characterization. Full saturation is one of the most important issues,
particularly for undrained laboratory testing, namely for static instability and liquefaction
assessment. Checking of the degree of saturation (Sr) during triaxial testing has been
traditionally made by measurements of the Skempton parameter value, which
corresponds to the ratio between the change in pore pressure, u, in undrained conditions
caused by an increment in mean normal confining pressure 3 (Bishop, 1973). On the
other hand, it has been analytically and experimentally demonstrated that the
compression wave velocity (VP) is very sensitive to the degree of saturation, especially
as Sr increases from about 99% to 100% (Biot, 1956a; 1956b; Ishihara et al., 1998; Yang
and Sato, 2000; Kokusho, 2000; Yang, 2002; Ferreira, 2009). Valle-Molina and Stokoe
II (2012) reported that, for example, VP increases from a relatively low value for sand, in
the range of 400 m/s, to about 1500 m/s as the degree of saturation increases from
approximately 99.5% to 100%. Therefore, it has been advocated by many authors that
the measurement of P-wave velocity should be used in detriment to the Skempton B value
for the assessment and verification of full saturation (Ferreira, 2009; Soares, 2014).
Moreover, Naesgaard et al. (2007) have shown that this technique is also applicable to
soils with viscous pore fluid, widening its applicability.
Another crucial state parameter is the void ratio. The final critical void ratio has been
found to be most accurately determined by recurring to the freezing technique of the
sheared specimen, as proposed by Sladen and Handford (1987). This procedure is
particularly relevant in the laboratory definition of the critical state line (Soares 2014).
Laboratory testing of undisturbed soil samples always involve some violation of the
in situ condition which may affect test results. In addition, they are costly and limited to
local material characterization. In situ test methods have been developed to complement
conventional laboratory testing. These methods range from conventional seismic
profiling to hybrid methods, such as the seismic piezocone penetration test (Campanella
et al., 1986) and the seismic at dilatometer test (Martin and Mayne, 1997), which allow
both stiffness evaluation from shear wave velocity and penetration test parameters. While
these methods allow evaluation of the entire depth of investigation, the profiling
resolution may not always be high, as it relates to the manner in which the seismic waves
travel and spread in the soil.
In the last decade, a number of advanced fiber optic sensing technologies have
matured and developed into commercial analyzers for temperature and strain
measurement. The spatial resolution of these technologies has recently increased
significantly by improved algorithms and they now facilitate accurate strain
measurements with a spatial resolution of less than 1 cm. The great flexibility of such
technology, has been explored for on-specimen strain measurements using high spatial
resolution fiber optic distributed sensing (Uchida et al., 2015). The approach provides
the means to evaluate specimen surface deformation through a novel conjuncture helical
envelope configuration of a single optical fiber. This has shown that the approach
provides a full-field view of surface strains with a resolution and accuracy level that is
comparable with traditional deformation sensors. This leads to evaluation of small-strain
mechanical properties as well as visualization and quantification of any indication of
non-uniform test conditions. Because of the relative ease and low-cost of
instrumentation, the suggested novel approach has a great potential to become a routine
application for on-specimen strain measurement.
The possibility of utilizing high resolution fiber optic distributed sensing for in situ
evaluation of the soil properties, especially stiffness moduli, has been recently
demonstrated by Klar et al., (2015a,b). The evaluation procedure involves a recursive
calculation of an elastic problem together in an input of a measured strain profile from a
vertically installed optical fiber under surface loading. Results from such a field trial will
be presented and discussed.
2. Sampling in granular soils and representativeness of in situ state conditions
Natural sands and non-plastic silty sands: focus on cyclic instability earthquake
engineering
Tube sampling in sands is drained, so volumetric as well as shear strains will occur.
It is highly probable that the magnitude of the shear strains will be sufficient to
destructure the sand, particularly since yield strains in granular materials are low (Hight,
2000). The volumetric shear strains caused by sampling will depend on the initial density
of the sand being sampled: initially loose sand will contract and densify, while initially
dense sand will dilate and reduce in density (Marcuson III & Franklin, 1979).
Truly undisturbed samples of sand can only be obtained if the in situ sand structure
is preserved before sampling. Singh, Semente & Chan (1982) demonstrated that the
characteristics of sandy soils (including in situ stresses) could be effectively preserved
by freezing the ground. Despite being an excellent technique for obtaining undisturbed
samples, the freezing process is complex and expensive, and only justifiable in specific,
high level projects. This method is however limited to shallow depths, besides
introducing volumetric changes in the water surrounding soil particles. Moreover,
ground freezing may cause drifting of fines content in silty sands and disturbance on the
sensitive microstructure of these soils would likely occur during freezing and thawing
process. The difficulty in obtaining undisturbed samples in saturated clean to silty sands,
namely due to the excessive friction generated during penetration of conventional tube
samplers is known to cause serious disturbance to the specimens. As a result, subsequent
advanced laboratory characterization would be severely compromised, particularly for
studies involving cyclic instability and the assessment of liquefaction potential. Recently,
the gel-push sampling technique has been developed and successfully employed to
obtain undisturbed samples on non-plastic silty sands, namely in liquefaction sites in
southern Taiwan, Christchurch in New Zealand, or Tokyo Bay in Japan (Lee and Ishihara,
2006; Huang et al., 2008; Lee et al., 2012), as well as in very difficult non-plastic tailings
in Zelasny Most in Poland (Jamiolkowski & Masella, 2015).
The gel-push sampler (GP-S) was developed based on modifications to the
conventional Osterberg-type fixed piston sampler to accommodate the thin wall tube
inside the sampler to become a triple tube system. The new sampler was designed to
allow polymer lubricant to seep into the tube wall while the tube penetrates into the soil.
The polymer gel would contaminate only a limited superficial portion of the sample,
since a very small amount of polymer gel is required to effectively minimize wall friction,
therefore allowing sensitive silty sand samples to be recovered with high quality (Lee et
al., 2012). A scheme of the sampler is presented in Figure 1.
Figure 1. Gel-push sampler (Apud Jamiolkowski & Masella, 2015)
A combination of shearing and loss of confining stress during sampling may cause
irreversible changes to the fabric of the sample (destructuring) and result in a loss of
ageing effects, which can be measured by a reduction in shear wave velocity, VS (lab
versus field measurements). Huang et al. (2008) showed a comparison of field (sCPTu)
and laboratory (bender elements) measurements of VS, with good agreement, suggesting
minimal disturbance using this technique. Also, Lee et al. (2012) reported a well-
preserved soil fabric from GP-S collected samples of silty sands of fluvial origin in
Taiwan. Taylor et al. (2012) briefly reviewed the GP-S procedure including comparison
to other soil samplers in the literature and a preliminary evaluation of its sampling quality
indicated that high quality samples of silty sands had been obtained.
Tailings of silts and silty sands focus on static instability
Undisturbed sampling of some soils, especially when saturated and below the
groundwater level, has proved to be extremely difficult or even impossible. That is the
case of tailings of silts and silty sands, for which laboratory testing is generally conducted
on reconstituted material. There are a number of laboratory specimen preparation
methods, namely moist tamping, air pluviation, slurry deposition and static compaction;
however, the structure and soil fabric induced by each method is different, and may not
be representative of the soil deposit under study. For this reason, the selection of the most
representative reconstitution method is fundamental to capture and replicate the
mechanical behavior of the undisturbed soil. Various authors have studied the influence
of the laboratory specimen preparation technique, particularly for the characterization of
the cyclic behavior of the soil. The influence of laboratory specimen preparation on the
measured cyclic resistance of specimens created by moist tamping or air pluviation has
been demonstrated by Mulilis et al. (1977) and attributed to the distinct soil structure or
fabric induced by its preparation. Yoshimi et al. (1984, 1989) reported significant
differences in the liquefaction strength of high quality undisturbed samples obtained by

the in situ freezing technique and of reconstituted specimens, by moist tamping or air-
pluviation. The liquefaction strength of the reconstituted specimens also differed,
depending on the method of remolding, and this fact confirmed the importance of the
effect of particle arrangement on the cyclic liquefaction of the sand.
Castro et al. (1989) aimed at defining the undrained steady state strength of
undisturbed samples of silty soil from Lower San Fernando Dam containing high
percentage of fines. The deposits were highly stratified with extensive layers on the order
of 1 mm thickness. Apparently the undrained steady state strength of undisturbed
samples were higher than that of remolded homogeneous samples. This discrepancy of
steady state strength was attributed to the graded and layered structure of the undisturbed
samples that had not been reproduced by laboratory reconstitution methods, which tend
to generate homogeneous and uniform specimens. Later, Baziar & Dobry (1995)
duplicated the original layered structure of grading on remolded specimens of the same
soil by depositing the material through water to create graded bedding, having found that
the steady state of the remolded but layered specimens were positioned near to the steady
state of undisturbed samples. Yoshimine & Koike (2005) also identified the effects on
liquefaction characteristics of stratified structured well-graded sands, due to segregation
of particle size and graded bedding.
Chang et al. (2011) compared the structure and fabric of gold tailings in undisturbed
and reconstituted moist tamped and slurry deposited specimens, reporting evidence of
clear differences in behavior. Time since deposition was also found to have a significant
effect on soil response. This influence has been collectively attributed to ageing effects,
and is thought to be due to a possible combination of stress history, mechanical creep,
and the development of physico-chemical bonding between particles (Mitchell, 1976).
Therefore, in order to characterize the mechanical behavior of a natural sandy deposit, it
is essential to perform laboratory testing on high quality undisturbed samples where the
natural fabric and any associated ageing effect are preserved (Taylor et al., 2013).
RESIDUAL SOILS:
Focus on stress-strain response, from the very small stiffness
Residual soils are abundant in many parts of the world. These geomaterials result
from in situ weathering of igneous, sedimentary and metamorphic rocks. The degree and
extent of weathering varies considerably with depth, hence weathered rock profiles may
contain material grades from fresh rock to completely weathered material, usually
classified as residual soil. When sampling these profiles, it is therefore inevitably to
penetrate through several different grades of geomaterials (Brand and Phillipson, 1985).
Due to their specific genesis, these soils present complex characteristics, which are a
consequence of the overall variability and heterogeneity of the parent rock, as well as of
the spatial arrangement and distribution of the particles and pore spaces. The resulting
residual soil is characterized by the presence of a bonded structure and fabric, which has
significant influence on its engineering behavior, particularly in its small-strain stiffness
properties (Viana da Fonseca, 2003). Sampling problems are due to the difficulty in
preserving its relict structure and its partial saturation, as well as the variability of the
soil fabric, namely the presence of hard weathered rock fragments in a soft soil matrix.
For these reasons, conventional tube sampling is expected to introduce some degree of
disturbance, and only block sampling can be reliably considered undisturbed.
Results obtained by Ferreira et al. (2011) on two experimental sites on Porto residual
soil enabled to identify considerable differences in the sampling quality of block and
tube samples recovered by different samplers (e.g. driven samplers with various cutting
edges, Mazier, Osterberg, Shelby) by the comparison of in situ and laboratory shear wave
velocities. The loss of shear wave velocity from in situ to the laboratory ranged between
14% (block samples) to nearly 50% for some of the tube samples. Ng & Wang (2001),
Ng et al. (2004) and Ng & Leung (2007) also measured shear wave velocities on block
specimens of completely decomposed granite and tuff, reporting values around 30%
higher than those of the Mazier specimens. More recently, Rocchi & Coop (2015)
reported small disturbance on Mazier collected specimens on Hong Kong residual soil,
which may be a result of the grading characteristics of the soil and to a higher weathering
degree.
According to Ferreira et al. (2011), for residual soils, considering the stages of a
sample (from sampling, to storage, preparation and finally to laboratory testing) shear
velocity losses below 15% appear to be minimal and therefore acceptable as an indicator
of an excellent quality sample. A gradual scale can then be empirically and
experimentally established: below 30% for a very good quality sample; below 40% for
a good sample; below 50% for a fair quality sample. For a loss in VS above 50%, the
quality of the sample is poor and the sample should be considered disturbed, therefore
unsuitable for careful laboratory testing and characterization.
Lunne et al. (1997) proposed a criterion for the assessment of sampling quality for
fine soils, based on the e/e0 ratio, a measure of the volumetric changes due to the
reconsolidation to in situ stresses, as discussed in the following section. The applicability
of this method to more well-graded soils, such as residual soils, was evaluated by Ferreira
et al. (2011).
Given the low compressibility of these soils, resulting from its natural fabric and soil
skeleton, this method alone was found to be inappropriate for assessing the sampling
quality, since all samples would be classified as very good to excellent quality samples.
In order to establish a more suitable sample quality classification, different category
ranges were suggested, by scaling down by a factor of 7 the values of the original
criterion. The comparison between the two methods of assessing sample quality,
corrected for residual soils, is shown in Figure 2.
The effects of laboratory specimen preparation on changes in natural residual soil
specimens have also been investigated by measuring seismic wave velocities in high-
quality block samples before and after trimming cylindrical specimens (Besenzon, 2013).
An average drop of 10% in shear wave velocity was measured in the cylindrical
specimens, compared to that measured in the original block sample, evidencing minimal
disturbance. However, the last tested specimens, over two months after trimming from
the block, showed a 30% reduction of shear wave velocity, indicating that ageing during
storage is an influential parameter in the shear stiffness of a soil, as also pointed out by
Ferreira (2009).
100%
Vs* lab/
Vs* in situ 90% A
80%
B
70%
C
60%
D
50%
40% E
30%
20%
10%
0%
0.000 0.005 0.010 0.015 0.020 0.025
,e/e0
Figure 2. Comparison of sampling quality classifications: normalized VS lab/VS in situ versus e/e0
(Ferreira et al. 2011)
3. Sampling in fine soils and representativeness of in situ state conditions
Two key effects of tube sampling on soft clays have been identified: a reduction in
mean effective stress and shrinking of the bounding surface of the soil. It is therefore
possible to anticipate the consequences of these effects in testing tube samples. One of
the most significant advances in understanding disturbance by tube sampling was the
application of the strain path method (Baligh, 1985; Baligh et al., 1987) to the penetration
of a sampling tube into the ground, which was further extended by Clayton et al. (1998)
to examine the effects of different features of the sampling tube geometry.
In case of homogeneous low permeability clays, quality undisturbed samples can be
evaluated in the laboratory by measuring the sample suction immediately after its
retrieval from the ground (Skempton, 1961; Chandler et al., 2011), although this
approach is quite complex and time consuming, thus not routinely employed. A more
widely used criterion has been proposed by Lunne et al. (1997, 2006) for fine grained
soils in terms of e/e0 ratio, where e is the void ratio reduction during one-dimensional
recompression of an undisturbed specimen to in situ vertical effective stress v0 of the
sample and e0 corresponds to the in situ void ratio. These authors established ranges of
variation of this ratio, from which the quality of each tested sample can be classified.
Another criterion, applicable to both coarse and fine grained soils is based on the
comparison of normalized shear wave velocity Vs1(L) measured on laboratory specimens
with that measured in the field Vs1(F) (Landon et al., 2007; Sukolrat et al., 2008; De
Groot et al., 2011; Fioravante et al. 2012). Landon et al. (2007) and Sukolrat et al. (2008)
compared laboratory and in situ shear wave velocities in Boston and Bothkennar clays,
respectively and used the reconsolidation strains method by Lunne et al. (1997) to define
categories of sample quality.
A large research program is currently ongoing on Ballina clay (New South Wales,
Australia), as Australias first National Soft Soil Field Testing Facility (NFTF). The
research aims to provide valuable information regarding the effects of sample
disturbance on soft soils due to tube sampling, devising more appropriate laboratory
procedures to handle samples and to develop new sampling tools to minimize sample
disturbance (Pineda et al., 2014a). An important outcome has already been reported: the
development of an innovative direct-push sampler specifically designed to reduce sample
disturbance on these soils and maximize sample retention (Pineda et al., 2014b). The
sampler performance for testing soft clays was proved, using laboratory tests as well as
CAT scans to evaluate the reliability of the sampler and also to estimate sample
disturbance.
4. Representative state conditions in triaxial specimens: moist and void ratio
4.1. The importance of measuring precise systems or alternative procedures even for
large strain states
The triaxial test is perhaps the most popular and used equipment to study soil
properties. The direct control of the variables involved, tests feasibility and easiness of
performance makes triaxial cell a winner among all other devices. Therefore the triaxial
cell is, perhaps, among other sophisticated devices in geotech labs, such as Hollow
Cylinders and Simple Shear, the more interesting compromise, despite not allowing
principal stress rotation and principal stresses control other than two of them, the vertical
and the horizontal ones.
Research in FEUP, for studying the major factors and limit boundaries in soil
liquefaction potential based on laboratory data has evolved generalize the use of triaxial
cells supplied with internal instrumentation (locally or directly pivoted to specimens
Figure 3).
These are extremely important, mainly for materials that do not exhibit a structure
collapse, thus with a medium/low volumetric strain variation during the saturation phases,
since some of those devices will surpass their range. Internal instrumentation allows
valuable information about the volumetric strains during the percolation and saturation
phases, but it can be erroneous due the non-homogeneity of specimens shape during
testing, particularly in the development of localization shearing discontinuities and
barreling. When the precise evaluation of void ratio is necessary, for instance, being the
the definition of ultimate state decisive to critical frameworks in instability models for
tailings and slopes masses studies (Viana da Fonseca, 2012; Bedin et al., 2013; Viana da
Fonseca & Soares, 2015) post-failure freezing technique, proposed long ago by Sladen
& Handford (1987), should be attained more commonly in geotechnical labs.
Figure 3. Hall-Effects transducers in local instrumentation and post-failure freezing technique for
void ratio determination in critical state studies in sandy soils (Rocha et al., 2012)
Extensive studies developed in FEUP for the evaluation of the performance of

experimental tool-based design approaches in liquefiable deposits, and reported in the
PhD thesis of Soares (2014), have proved the strong influence of this complementary
procedure for the identification of critical and steady locus in the volume (void ratio)
effective stress space in silts, sandy silts and silts (from natural deposits and tailings),
having a relevant influence in the definition of either stable, metastable or unstable states
(Soares & Viana da Fonseca, 2015).
The majority of the soil specimens were prepared by moist tamping with a certain
water content, depending on the soil type, while a few set of tests were prepared with a
dry pluviation technique. Among other techniques, moist tamping was chosen as it is
able to produce the widest range in void ratio with satisfactory uniformity (Ishihara,
1996), allowing for a clear definition of the CSL. Although being a popular laboratory
technique, it is commonly associated with the creation of a honeycomb fabric which
leads to a higher susceptibility to liquefaction. Notwithstanding it is an easier, low time-
consuming technique to perform, which does not segregate soil particles (such as dry
funnel pluviation technique), but mainly allows the operator to control the final density
much more readily (adapted from Chang et al. 2011). The moist tamped reconstitution
technique the soil specimens are described in Viana da Fonseca et al. (2011).
Common phases were developed before shearing in monotonic and cyclic triaxial
tests. After percolation, a complete saturation was initially accepted for values for
Skemptons B parameter of 0.92-0.95, which was revaluated after measuring seismic P-
wave velocities with bender/extender elements (to be discussed below), which proved
that the full saturation was only attained for B0.97. This was accepted increasing the
back pressure up to 600kPa to 700kPa, keeping the effective confinement at the level of
10kPa. During the first stages, specimen dimensions were re-measured with the help of
internal/local transducers (inductive hall-effect calipers) measuring axial and radial
deformation (Soares, 2014). Isotropic and anisotropic consolidation, for K0=0.5, were
also accompanied by redundant volume measurement with a volume change gauge. Any
correction for the void ratio after the saturation and consolidation phases was assured by
using the ultimate void ratio deduced with the freezing technique. This consists on
freezing the sample in the end of the test (Figure 3). With the freezing technique, the
volume was accurately determined by keeping the volume of water inside the specimen,
which is completely saturated (G x w = S x e). With this final void ratio and the precise
measurements of volume changes from the beginning of the consolidation, void ratio
was easily determined at the end of each stage. As previously referred, this freezing
technique was initially proposed by Sladen & Handford (1987) in order to correct the
potential systematic error in laboratory testing of very loose sands.
In the beginning of this research, percolation and saturation phases were time-
consuming, taking 5 to 6 days to be completed. Then the generalization of the recourse
to CO2 percolation prior to de-aired water percolation allowed B parameters of higher
than 0.97 to be reached in less than 1 day. This procedure had already been used at FEUP
some years ago, when a series of studies were made for the characterization of the
geomechanical behavior of residual soils (Viana da Fonseca, 2003; Viana da Fonseca &
Coutinho, 2008). Finally, monotonic or cyclic shearing was applied under strain-control
or stress-control, respectively, being the monotonic performed in constant strain rate,
increasing the deformation up to around 20% or more, until the critical state was reached,
while the cyclic liquefaction deemed to have liquefied when a single amplitude shear
strain exceeded 2.5% of single amplitude axial strain (Ishihara, 1996). Drained tests were
performed with a velocity of 0.01 mm/min, while for undrained tests the velocity of 0.05
mm/min was adopted (an axial strain velocity clearly compatible with the rate of excess
pore pressure dissipation).
4.2. The importance of using the freezing technique as a beneficiary complement to

local strain measuring systems for the definition of ultimate states
One of the materials chosen to exemplify the importance of the freezing technique
is a silt, produced by artificially grinding Osorio sand, which was initially studied in a
PhD research by Bedin (2010). During such study, which purpose was to assess its
liquefaction susceptibility, the researcher performed a set of both drained and undrained
compression triaxial tests relying exclusively on local strain registers.
In the scope of a more recent PhD study (Soares, 2014), some of the compression
triaxial tests aiming to evaluate some discrepancies then observed, were reevaluated by
using the freezing technique. This resulted in the correction of the position of CSL and
SSL. A large set of anisotropic consolidated triaxial tests was also conducted to failure
by increasing pore water pressure and compression triaxial tests with decrease of lateral
confinement effective stress, completing the onset of this new study (Soares, 2014).
The more compressible silty materials, the clearer and more surprising results were
obtained in what respects to the instability associated to the microstructural fabric
rearrangement, generated under drained as in undrained conditions. This reevaluation of
the previous results arose from the correction of CSL positioning, when complementary
tests results were in disagreement with the previous, resulting from the implementation
of freezing technique in the testing practice. While the initial void ratio changes were
obtained by monitoring the internal/local strain transducers, by the end it was clear their
inability to fully capture the volumetric strain developed by the soil specimens during
the percolation and the saturation stages. Therefore, as originally proposed by Sladen &
Handford (1987), the specimens were frozen at the end of the shearing stage, which
allowed an accurate determination of the final void ratio. The corrections to the previous
results were implemented based on the differences between the older and new NCL
positioning.
The drained behavior of the silt can be seen in Figure 4. The stress-strain and the
volumetric-strain curves present a clear contractive tendency. This was already expected
as all silt specimens were prepared with very high void ratios. Tests over the silty soils
were essentially in compression starting from the NCL, exhibiting similar volume
contraction, being NCL and CSL very much parallel, as expected. These differences were
not so significant in the tested sandy soils, since these are not as contractile during
percolation and saturation stages, the phases that are determinant in the errors detected
in ulterior analyses. On the other hand, silty and sandy-silt mixtures, which are prepared
with very high void ratios, do have very high volume changes when prepared in high
initial void ratios before being consolidated.
Figure 4. Stress-strain and volumetric-strain curves of drained triaxial tests on silt
From Figure 4 it can be observed that in the shearing phase, volumetric strain and
shear stress only stabilize, for large axial strain, respectively of about 10% to 15%.
The high volumetric strains prevent rearrangement between grains due to large
amount of small particles between the larger ones, significantly higher than in sandy
soils specimens. Figure 5 shows separately the behavior of one high-pressure triaxial
test performed on a silty soil specimen. The evolution of the grain size distribution
during the high-pressure triaxial test is seen in Figure 6. Three distinct grain size
distribution curves are plotted: the continuous line is the original material, and two
distinct curves of the material at the end of the high-pressure triaxial cell. The dashed
line was obtained by the sedimentation test and the dash-dot line was determined
through a particle size analyzer laser equipment. Despite the first sections of the lines
not being coincident, both curves display a slight increase of the finer particles. So
when comparing the two lines obtained by sedimentation it appears that the curve has
been shifted to the left, which can be explained by the finer fraction increase. In reality,
the test displayed a maximum shear strength located well below the failure envelope.
It is believed, despite not being proved, that the uneven grains got more rounded
during the shear stage, explaining both the increasing finer fraction and the lower
critical friction angle, more similar to the one obtained for Osorio sand, another well
studied Brazilian sand (Viana da Fonseca et al., 2009) or other morphologies (Consoli
et al., 2007). The undrained behavior of the silt can be seen in Figure 7.
Figure 5. Stress-strain and strain-volumetric curves of the high-pressure triaxial test on a silt
0.074
0.105
0.180
0.250
0.841
2.000
ASTM standard sieve seizes (mm) 4.760

100
90
80
70
60
% Percent finer by weight
50
40
30
Silt 2010
20 Silt High-pressure Tx Granul.
Silt High-pressure Tx Sed.
10
0
0.001 0.01 0.1 1 10
Particle diameter (mm)
Figure 6. Evolution of the grain size distribution in high-pressure triaxial test (pi=7.4 MPa)
According to the stress-strain and the pore-pressure-strain curves, a decrease of

contractiveness is evident with increasing initial confining stress and decreasing void
ratio (as the tests start from the NCL). Figure 7 shows that tests conducted with low
confining stresses liquefied, evolving to a strain softening behavior with medium
confining stresses, while those confined with higher confining stresses clearly exhibit
a much more stable behavior (with strain hardening). This is a typical behavior of
cohesionless finer soils, exhibiting a similar trend of the other soils (silty-sands,
dandy-silts, from natural deposits and tailings) studied by Soares (2014).
Figure 7. Stress-strain and strain-pore water pressure behavior of undrained triaxial tests of Silt at
low to high confining pressures (Soares, 2014)
All specimens were prepared with very high initial void ratios, e 1.2, which,
after percolation and saturation, reached distinct void ratios after the consolidation
stage, typically on the NCL. The drop in contractiveness is a result of increasing initial
confining stress and decreasing void ratio, similarly to what was observed in
monogranular sands, e.g. Algeria sand, Coimbra sand, Osorio sand (Soares, 2014).
This contractiveness was different in the two test series, due to an insufficient
saturation degree in the tests of the first series. The inclusion of CO2 percolation prior
to the water percolation and increasing back-pressuring, resulted in distinct degrees of
saturation, thus explaining the remarkable differences of the two curves in Figure 8,
mainly for the triaxial tests performed with high values of confining pressures. Despite
not knowing the degree of saturation of the first series of tests, P-wave velocities were
measured in first series of tests while the second series only allowed the measurement
of the Skempton parameter, B, as the current bender-extender elements used in
LabGeo are not prepared to be frozen. Using Yang (2002) correlation, between the
Skempton parameter, B, and the P and S-wave velocities, it was concluded that most
specimens tested in the first series had only achieved a Skemptons B parameter
between 0.89 and 0.95 (Soares, 2014; Soares & Viana da Fonseca, 2015).
On the other hand and also relying on Yang (2002) correlation, the second series
of tests revealed, for B>0.97, P-wave velocities near 1600 m/s (see Figure 9).
Figure 8. Stress-strain and strain-pore water pressure curves comparison with distinct saturation
degrees
Figure 9. Relation between P-wave velocity and B for Silt (adapted from Yang, 2002)
Aware of the insufficient degree of saturation, as determined by post-failure

freezing of the tested specimens, it became clear that the differences shown in Figure
8, between the strengths and the volumetric variations during loading in the two tests
was significantly associated to that.
The soil ultimate states were separated according to its final behavior. Thus, while
the CSL was built upon the ultimate states achieved by the drained triaxial tests and
the undrained triaxial tests which exhibited strain hardening, the SSL was delineated
by the ultimate states of the strain-softened undrained triaxial tests (Figure 10).
Nevertheless, some of the undrained triaxial tests which converged with the CSL
were verified to not have a guaranteed full saturation.
Figure 11 emphasizes the distinct ultimate states in constant volume loading
conditions, defined in the clear distinct Steady State Lines (SSL) expression,
resulting from the high sensitivity of undrained responses, when there is air
compressibility due to insufficient degrees of saturation. Therefore, with a decreasing
degree of saturation, the ultimate state moves upwards and to the right in the log p-e
space, converging with the ultimate state reached by the drained triaxial tests.
Figure 10. Construction of the critical state and steady state lines for silt
Figure 11. Critical state line and steady state line of silt
4.3. Influence of isotropic / anisotropic consolidation in the definition of ultimate states
Jointly with the degree of saturation, other factors such as the induced anisotropy
were identified to also influence the ultimate state achieved by silty-sand specimens
tested under undrained conditions (Soares, 2014).
The study of a silty-sand soil (entitled Mixture in Soares, 2014), involved the
performance of a large set of compression triaxial tests. The determination of ultimate
states allowed outlining both CSL and SSL, by the triaxial tests sheared under drained or
undrained conditions, respectively (as represented in Figures 10 and 11). Due to its high
brittleness, the state conditions of this soil evidence no signs of strain hardening.
Therefore, none of the undrained triaxial tests converged with the ultimate states
achieved by the drained triaxial tests. It is also perceptible that the high-pressure triaxial
test seems to follow the same trend line as the low and medium confined tests which is
likely to be due to the non-occurrence of grain crushing.
During the research of the silty-sand, detailed in Soares (2014), two anisotropic
consolidated specimens were also carried out under undrained conditions. It was
concluded that the effect of the induced anisotropy is very pronounced on the SSL
positioning, unlike what was found for Algeria sand (a monogranular sand also studied
in detail by Soares, 2014). This effect is partially associated to the high range of void
ratios covered by the CSL slope, which implies that only by assessing the brittleness
index jointly with the state (or steady state) parameter, major conclusions about the effect
of the induced anisotropy on the soil liquefaction potential could be anticipated. As can
be visualized in Figures 12 and 13, the SSL positioning denotes low stability of the
anisotropically consolidated specimens. Nevertheless, despite the very high drop of the
shear strength, it is not as high as the one exhibited by the CIU tests. Both equations for
SSL and CSL found for Mixture, are defined in those figures.
Figure 12. Construction of the critical state and steady state line for the Mixture
Figure 13. Critical state line and steady state line of the Mixture
5. Laboratory and in situ strain evaluation using fiber optic sensing
Strain is a fundamental component of civil engineering. It directly provides

information on the stress levels in elastic systems and on the cumulative damage and
fatigue in elasto-plastic systems. Most engineering design processes of civil engineering
structures (e.g. buildings, foundation systems, embankments, tunnels, pavements,
retaining walls, pipelines, etc.) involve either directly or indirectly (through stress
analysis) limits on the allowable strain levels in the systems. It is therefore no wonder
why measuring of the developed strain in civil engineering infrastructure is of the utmost
importance for its effective design, construction and maintenance.
Traditionally, strains were (and still are) evaluated locally using changes in electrical
resistance of foil (bonded wire) strain gauges, or by more advanced local devices such
as vibrating wires and MEMS. The interpretation of a local strain measurement for the
purpose of understanding and analyzing geotechnical structures is, however, rather
limited (considering the continuum nature of the soil). The local measurement neither
can be integrated (to result in deformation) nor differentiated (to evaluate soil-structure
interaction forces). The development of distributed fiber sensor technology, and its
capabilities to provide spatial profiles of strains along conventional telecommunication
fibers, has led to a re-evaluation of the manner in which strains can be used in
geotechnical engineering. This section presents various laboratory and field applications
aiming at characterizing both fundamental (element scale) and global parameters.
5.1 Advanced and high precision systems for specimen testing
This subsection points out the possibility of the using high precision systems in
specimens tested in laboratory with a global geometrical spectra, which will respond to
the limitations of precise local strain measurements, identified previously, when diverse
non-homogeneities often incur in errors which do not relate to the material
characterization, relevant to its in situ behavior. Such testing systems should involve
multiple or distributed sensitive sensors that do not interfere with the specimen behavior,
yet provide a better insight into the local and global sample response. Potential
candidates are image analysis methods and on-specimen fiber optic strain measurements.
Image analysis methods are capable of providing full-field deformation
measurements, and have been gaining popularity in recent years for civil engineering
applications (e.g. Sutton et al., 2009; Brigisson et al., 2008; Dijkstra & Broere, 2010). It
has even been used in triaxial testing systems for pressure core analysis of hydrate
bearing sediments (Yoenda et al., 2013). Figure 14a shows the triaxial testing apparatus
of Yoneda et al. (2013), which include 8 digital cameras surrounding the soil sample.
Figure 14b shows surface contour plot of the axial strain obtained from the image
analysis (the sample is under averaged vertical strain of 18%). The accuracy of strain
evaluation by image processing depends both on the digital image resolution (number of
pixels) and the field-of-view. Consequently, the field of view must be a few millimeter
in size in order to match the strain accuracy of conventional on sample devices. In
Yoneda et al. (2013)'s experiments the field-of-view includes the whole sample, leading
to a measurement accuracy of 0.1% (=1000). This accuracy is quite low compared to
that which is possible by fiber optic sensors (order of a few microstrains). Moreover,
stringent experimental requirements, such as vibration-free optical platform and
sufficient availability of a line-of-sight between the specimen and the imaging
equipment, are imposed.
(a) (b)
Figure 14. (a) Camera arrangement around of Yoneda et al. (2013) apparatus, (b) image analysis based
vertical strain field of loaded sample.
Throughout the last decade significant advancement in distributed fiber optic strain
sensing has been achieved in the optical fiber sensors research community. These include
the long distance, continuous monitoring of static strain up to 300 km (Gyger et al.,
2014); the enhancement of spatial resolution towards cm and even mm-scale (Song et
al., 2006; Elooz et al., 2014; Cohen et al., 2014; London et al., 2014); and dynamic
measurements at hundreds of Hz rates over hundreds of meters (Peled et al., 2011,
2012,2013; Motil et al., 2014; Sovran et al., 2015). The most widely employed
measurement technique is based on the optical interaction of stimulated Brillouin
scattering (Kurashima et al., 1990; Nikles et al., 1996). Other principles such as coherent
Rayleigh back-scatter analysis (Lu et al., 2010) and cascaded fiber Bragg gratings
(Kersey et al., 1993) are being pursued as well, with much success. Many of these
developments had matured into commercial products, and have been introduced in the
geotechnical engineering (Vorster et al., 2006; Klar et al., 2006; Iten and Puzrin, 2009;
Klar and Linker, 2010; Mohamad et al., 2011; Klar et al., 2014), other (especially the
high resolution dynamic capabilities) will become commercially available in the next 2
to 5 years. All of these technologies facilitate new possibilities of improving strain
measurement techniques of geomaterials. Lee et al. (2011) described the principles of
using FBG optic fiber sensors in triaxial test set-ups, demonstrating their advantage over
electric sensors. While the electric sensors are currently subject to short circuit when
submerged under water and electromagnetic interference (EMI), optic fiber Bragg
grating (FBG) are waterproof and independent of EMI noise. FBG sensors have the
additional advantage of being partially distributive where multiple sensors can share the
same signal transmission line. Taking advantage of these unique capabilities, Lee et al.
(2011) explored the possibility of converting all pressure/force and linear displacement
transducers in a triaxial testing device into FBG based sensors (Figure 15 illustrates a
schematic view of their system and a picture of the fiber optic sensored trixial cell). A
series of shearing tests on unsaturated and saturated soil specimens were carried out using
the new FBG sensored triaxial testing device. In most cases, the measurement of physical
quantities were paired with electric sensors so that the results compared well (Figure 15).
Figure 15. Schematic view of the testing system and fiber optic sensored triaxial cell, comparison
between deviator stress-axial strain and pore water pressure-axial strain relationships from constant water
content triaxial tests (adapted from Lee et al., 2011)
While FBG sensors may be distributed along the same fiber, they still constitute
local measurement devices. Truly distributed sensing is achievable by investigating the
Brillouin scattering or Rayleigh back-scattering from conventional telecommunication
fibers, allowing evaluation of strains in spatially continuous manner. One may utilize
these capabilities together with very flexible fiber optic cables (such as bare fibers or
flexible ribbon fibers) to evaluate the overall on-specimen surface deformation with high
resolution and accuracy. First steps towards such utilization of high resolution distributed
fiber optic sensing have been made at the Technion Israel Institute of Technology
(Uchida, Levenberg & Klar, 2015). It has been suggested that a conjuncture helical
envelope configuration of a single optical fiber be used to evaluate the surface strain
field, in both small and high strain level. As detailed earlier, much work attention has
been placed on material characterization under small-strains to better capture the
behavior in typical service conditions. These deformations can be reliably monitored
with local (on-specimen) measurement instrumentations (Clayton, Khatrush, 1986, Goto
et al., 1991; Viana da Fonseca, 1998; Ibraim, Benedetto, 2005; Tutumluer et al., (1998),
Yimsiri et al., 2005; Viana da Fonseca et al., 2014). As referred by Uchida et al. (2015),
all these sensors, and other, are accurate and reliable. However, in order to ensure the
validity of the measurement, complicated and delicate procedures are required such as
drilling and gluing, and, as already referred in the previous section, they have the
disadvantage of offering only a limited view of the deformation field at locations, which
must be decided prior to the test due to physical size and cost restriction. The use of
wrapped-around fiber optics together with their spatially continuous, high resolution
(<1cm) and high accuracy (small strains <5) measurement capabilities may allow
evaluation of the complete on-specimen strain profile as demonstrated by Uchida et al.
(2015) for uniaxial loading of acrylic glass. Figure 16 shows the conjuncture helical
configuration as was applied by Uchida et al. (2015) to the acrylic glass specimen, where
three different angles were considered (2.28, 6.81 and 13.43 degrees). Essentially, only
2 pitch angles are required, and the third (or more) may be used for validation. The helical
envelop may add to the confining stress, however this is usually very small and can be
incorporated into the interpretation of the results (for example in the presented case 0.004
kPa of every 1of vertical strain). High resolution strain may be obtained along each
fiber, and can be used either to evaluate the strains in the pitch direction through
extrapolation, or integrated for on-specimen deformation (which can also be extrapolated
and used for further analysis). Clearly, further applicative research is needed for
fabrication of special membranes, embedded with bare fiber optics, which can be used
for triaxial test of soil samples.
piston
(EP10000)
ball & socket
joint D

0.01m
1
2
LVDT4
LVDT1
LVDT1
LVDT3
LVDT1
LVDT2
LVDT4
0.06m
LVDT2
0.03m
acrylic glass
fiber ribbon =0 = =2
(a) (b)
Figure 16. (a) Instrumented specimen; (b) Net view of the cylindrical specimen (after Uchida et al., 2015)
5. Fiber optic sensing for in situ testing
The versatility of fiber optics is not limited to laboratory on-specimen

measurements. Fiber optics may be used to both enhance accuracy and precision of
transducers involved in conventional devices, and to offer alternative measurement
techniques. Fiber Bragg Grating is the most commonly used fiber optic technology for
strain measurement. It has been extensively used for different geotechnical applications.
Huang et al. (2013) described various field and laboratory applications of FBG, together
with brief background of the FBG technology and the concepts of design. FBG can be
used either as local strain gauges or as a sensor array for distributive (profile)
measurements. They have been implemented in various devices for in situ testing, such
as the pre-bored pressuremeter (PMT), flat dilatometer (DMT) and the piezo-
penetrometer.
Huang et al. (2013) presented the design and application of a FBG sensored in-place
inclinometer (FBG-IPI) capable of monitoring absolute inclinations (Figure 17). A pair
of FBGs spaced from the center line of the pendulum and stemmed from the top end are
fixed to the sides of the mass. The FBG pair was used to sense the amount of rotation of
the mass and offset the temperature effects. Details of data interpretation are included in
Huang et al. (2013). Figure 17 shows the FBG-IPIs, and their assembly and insertion
into an inclinometer casing. The FBG-IPIs were used to monitor ground movement of
a slope in the up-down slope plane, at Hua-Fan University in Shihding District, New
Taipei City. As described by Huang at al. (2013), the monitored area was underlain by
colluvial material, with occasional steep slope in excess of 25. Each FBG-IPI unit had
a length of 1.25m, capable of measuring inclination with a full range of 10 and a
resolution of 0.0016, in a single plane. The accuracy of FBG-IPI is 0.116% full scale.
The FBG-IPI installation consisted of an array of twenty, 1.25m units extending to 25m
below ground surface. Figure 17 shows a comparison of ground displacement profile
according to the FBG-IPI and those from manual IP readings. Both measurements
showed a shear plane at 20m deep and similar pattern of ground movements, although
the magnitudes were different. The difference in ground displacement magnitude may
well be associated with spatial material property variations.
hinge
linkage
FBG
mass
Figure 17. FBG In-Place Inclinometer (FBG-IPI) and comparison of in ground displacement magnitudes
on site (adapted from Huang et al. 2013)
Another fully automated slope stability monitoring with FBG in place presented by
Huang et al.(2013) are pore pressure arrays. A Five Turn Point located in a sloping area
of approximately 1200m by 1000m, with a topographical relief of up to 400m, was
studied by the authors. The slope has an inclination angle of 23 and the shear planes
associated with earlier ground failures according to available investigations. The area is
covered by 0-26m of colluvial material, a mixture of soil and rock pieces, interlayered
by sandstone, with shale underlying the colluvium, which extends up to 200m below the
ground surface (details in Huang et al. 2013). Figure 18 presents a cross sectional view
and a photo of the FBG pressure transducer. The FBG was used to sense the deflection
of a metallic diaphragm inside of the transducer. The diaphragm separates the reference
and input pressure chambers. When used as a gauge pressure transducer, the reference
chamber can be sealed or exposed to the atmospheric pressure. The reference chamber
is connected to a controlled reference pressure when used as a differential pressure
transducer. The amount of deflection at center of the diaphragm is linearly related to the
pressure difference between the reference and input pressure chambers. A separate FBG
was placed inside the transducer to monitor temperature fluctuations and compensate
thermal effects. The range of pressure transducer was controlled by the stiffness of the
diaphragm.
Figure 18. The FBG pressure transducer and the FBG piezometer array installation compared with a
styanpipe (adapted from Huang et al. 2013)
For field measurements, the pressure transducer was converted into a piezometer by
surrounding the drains with filter material (Figure 18), with a PVC pipe to prevent
seepage between piezometers from within the PVC pipe. The PVC pipe serves as a spacer
and housing for the piezometers and optical fiber (Huang et al. 2013).
Figure 19 shows a set of representative pressure head (hp) profiles based on the FBG
piezometer readings recorded from the beginning of typhoon Morakotto the time when
hp reached the maximum measurement values.
Figure 19. Profile of hp during typhoon Morakot
Case histories that include: PMT and triaxial tests on unsaturated soils, field
differential pressure piezometer and DMT to characterize a reservoir sediment under
40m of water, and fully automated slope stability monitoring with FBG in place
inclinometer and pore pressure arrays were presented and their implications in
geotechnical testing and monitoring discussed.
Minardo et al. (2014) have developed a Brillouin scattering based fiber optic
inclinometer, for evaluation of the 3D deformation of the soil. Their inclinometer was
composed of PVC pipe instrumented by four fiber optic cables (separated by 90 degrees
in the pipe cross section). The four fibers, together with the strain evaluation from the
Brillouin distributed sensing, allow evaluation of the curvature and the average axial
strain in any cross section (assuming the PVC pipe answers the assumptions of Euler-
Bernoulli Beam). By (single and double) integration of the curvatures and axial strain,
the 3D movement of each cross section can be estimated. Minardo et al. (2014) tested
their inclinometer in an unstable slope in Basilicata region in Italy. While the comparison
of the new inclinometer did not yield a very good match with the conventional
inclinometer (because the installed length was only to depth of 7.5m, while the sliding
had occurred at depth of 12m, and the boundary conditions for the integration are not
known in this case), the concept of such fiber optic inclinometer appears to be solid. In
fact, the same approach for evaluation of inclination and deformation using distributed
fiber optics, has been used in secant pile walls by Mohammed et al. (2007; 2011). Figure
20 shows schematics of the instrumented pile and a comparison of curvature, inclination
and displacements obtained by the fiber optic measurements and by a conventional
inclinometer. As can be seen, an excellent agreement exists between the measurement
methods.
Figure 20. (a) fiber optic instrumented pile; (b) location of instrumented piles by fiber optic and
inclinometer; (c) protection tube and (d) fiber optic access point; (g) comparison of BOTDR fiber optic
measurements to conventional inclinometers results for excavation stage between two support levels in the
project reported by Mohammed et al. (2007,2011).
Distributed fiber optic measurements may also be used to evaluate tunneling induced
(greenfield) ground displacement parameters, such as the volume loss and the inflection
point (associated with the tunneling settlement trough). Since many design processes are
based on a greenfield displacement input, such measurement could be used as a safety
measure to ensure that the tunnel excavation process lies within the design bounds,
simply by monitoring a fiber at a greenfield area in front of the structure of interest (e.g.
pipeline, building, railway, etc.). Klar et al. (2014) have demonstrated how distributed
strain measurements along buried fibers provide valuable information for determination
of the greenfield parameters. Figure 21 shows the outcome of such a fiber optic
measurement based analysis aiming at predicting the parameters for a 3D ground
displacement model. The fiber optic cable at the considered case was buried at a shallow
depth of 0.5m and measured the soil deformation (i.e. horizontal strains) due to an
excavation of a 3m diameter tunnel at depth of 18m. The tunnel excavation front
advanced in a non-perpendicular angle ( ) of 20 degrees relatively to the fiber, and
measurements were taken throughout the process. Fig. 21b shows the strain
measurements along the fiber, taken when the front was positioned before the fiber
(negative L' values) and ending when the tunnel face was beyond the fiber (positive L'
values), together with the optimal model prediction of the expected strains in the fibers.
The analysis allowed evaluation of classical ground displacement parameters such as the
settlement trough width (ix), the trough length (iy), the volume loss (VL), and the
maximum settlement (smax). As can be seen, a shift in the location of the maximum strain
value occurs when the tunnel advances in a non-perpendicular direction to the fiber. In
the paper, the authors demonstrate how the change of peak strain location with tunnel
face advancement (,xp/,L) can be used to evaluate directly the settlement trough length
(iy) without any global optimization scheme. A similar application of fiber optic sensing
for ground displacement evaluation was considered by Hauswirth et al. (2014). These
approaches are now standardized within ASTM F3079 14 "Standard Practice for Use
of Distributed Optical Fiber Sensing Systems for Monitoring the Impact of Ground
Movements During Tunnel and Utility Construction on Existing Underground Utilities".
100
50

Optical fiber xf 0
Fiber strain ()
-50
L=20.4 m
L
Tunnel face -100 L=7.4 m
-150
L=-16 m
-200
-250
-40 -20 0 20 40
Figure 21. Fiber optic surface measurement of a tunneling process (a) non-perpendicular, (b) measured
strain and optimized model prediction (Klar et al. 2014)
Klar et al. (2015) further explored the idea of utilizing fiber optic sensing for in situ
material characterization, by utilizing recent advancements in resolution capabilities.
The authors' objective was to evaluate the in situ stiffness moduli profile with depth. For
this aim they have suggested a simple, rapidly converging, elastic continuum recursive
calculation which involves a measured strain profile from a vertically positioned optical
fiber under surface loading. Essentially, the recursive analysis updates the calculated
moduli profile to result in an elastic continuum solution that provides analytical vertical
strains equal to those measured in the field (details about the recursive analysis are
provided in the current conference proceeding, in Klar et al. 2015b). The authors have
evaluated the sensitive of the approach to imperfect fiber optic installation and concluded
that near fiber disturbance marginally affect the resultant moduli profile. The approach
was then evaluated in the field. The fiber optic cables were installed in a similar manner
to that of wick drains (Figure 22), being the fiber thread through the mandrel and fixed
to the anchor plate and then the mandrel driven to the soil. Then some tension was applied
to the fiber while the mandrel was withdrawn. Finally the fiber was pulled, fixed at
ground level, and left to rest over a significant period to allow the soil to contract around
the fiber. The resultant converged shear stiffness (modulus) values are also plotted in
Figure 22. The values agreed well with those obtained by a geophysical survey
(conducted exactly a year later) and known correlation with SPT and shear strength
values.
Shear stiffness, G, (MPa)

0 20 40 60 80 100
0
4 CH
Depth (m)
Soil profile as observed in

nearby site investigation
6 boreholes
SP
8
CH
10
Figure 2. Fiber optic installation using a wick drain machine and an example of two measured
vertical strain profiles (Klar et al. 2015b)
The above review demonstrates the potential use of fiber optic measurements to
bridge between different scales of geomaterial characterization and behavior, from small
scale on-specimen measurements to in situ stiffness moduli profile evaluation and global
ground response (such as the evaluation of tunneling induced settlement trough).
Ongoing research in fiber optic sensing will most certainly facilitate in the coming years
more exciting techniques for geomaterial characterization. For example, the application
of fiber optics measurements for identification of `below ground consolidation front' (and
the associated parameter coefficient of consolidation, etc.); characterization of in situ
anisotropic behavior by multiple loading positions and multiple fiber optics cables; and
the complete dynamic analysis of pile load testing (using the newly developed slope-
assisted Brillouin sensing technologies of Peled et al., 2013).
6. Final considerations
In this keynote address the authors had the intention to give an insight on how can
laboratory tests over representative soils samples can be considered reference methods,
if some ameliorations are incorporated, relocating them in the front line of ground
property characterization. This was done at three levels: sampling quality, careful
reconditioning and preparation of specimens to guarantee the operational states, and
integrated measuring instrumentation that can give a complete view of the soil specimen
behaviour by using precise local strain systems that measure spatially those strains in a
continuous manner.
At the level of the sampling quality new equipments and techniques have improved
a lot the capacity of retrieving very good undisturbed samples, with emphasis to the
recent Gel-Push Sampler with an utmost successful performance in sandy and other
difficult granular soils under water table (such as silty soils in tailing ponds), by one side,
and high precision sample quality evaluation when recurring to shear wave velocities
measured in situ and in lab on specimens after sampling and/or their preparation prior
testing, by the other side. These steps ahead become crucial to capture and replicate the
mechanical behavior of the undisturbed soil.
In the preparation of soil specimens for specific demanding purposes such as the
correct definition of instability stress-state locus or critical conditions, so important in
specific studies, such as the study of liquefiability in granular soils under water table,
new approaches were described that can guarantee perfect and well controlled degrees
of saturation, or simply the precise definition of void ratios pre and post failure. The
different results obtained in considering these techniques that include compression waves
velocities and post testing freezing of the specimens, have shown hereby how significant
this can become in laboratory practice.
Both Optic Fiber Bragg Grating (FBG) and Brillouin based distribute fiber optic
sensing techniques have been proving an immense potentiality for both in situ and lab
applications of geometrical characterization, from small scale on-specimen
measurements to in situ stiffness moduli profiling and global ground response
characterization. They facilitate convenient ways of evaluating strains at different levels
from the very small strain levels to ultimate failure conditions, in a spatially continuous
manner.
Acknowledgements
This paper is a result of the financial support due to the research project
POCI/ECM/55589/2004 Dynamic properties in residual soils from granite: strain-rate,
frequency & time effects and developed under the research activities of ICIST of IST
and CEC from FEUP, supported by pluri-annual funding from FCT (Portuguese Science
and Technology Foundation). Javier Camacho-Tauta acknowledges the support from
Nueva Granada Military University, Research Project INV-ING-1765. This research is
part of a study on non-destructive testing of infrastructure. Support is also provided by
the Natural Sciences and Engineering Research Council of Canada (NSERC).
References
[1] Birgisson, B., Montepara, A., Romeo, E., Roncella, R., Roque, R., Tebaldi, G.: An optical strain
measurement system for asphalt mixtures, Materials and Structures, 2008, 42 (4): 427441.
[2] Campanella, R. Robertson, P. and Gillespie, D.: Seismic cone penetration test, ASCE Geotechnical
Special Publication No. 6, Use of in situ tests in geotechnical engineering, 1986: 116-130.
[3] Clayton, C. R. I., Khatrush, S. A.: A new device for measuring local axial strains on triaxial specimens,
Gotechnique, 1986, 36 (4): 593597.
[4] Cohen, R., London, Y., Antman, Y., and Zadok, A.: Brillouin optical correlation domain analysis with
4 millimeter resolution based on amplified spontaneous emission. Opt. Express, 2014, 22(10), 12070-
12078.
[5] Cubrinovski M., Bradley B.A., Wotherspoon L., Green R.A., Bray J.D., Wood C., Pender M., Allen J.,
Bradshaw A.S., Rix G. and others: Geotechnical Aspects of the 22 February 2011 Christchurch
Earthquake. Bulletin of the New Zealand Society for Earthquake Engineering, 2011. 44(4): 205-226.
[6] Dijkstra, J., Broere, W.: Combined full field stress and strain measurement methods for granular
materials, EPJ Web of Conferences, 2010, 6 (1): 22020.
[7] Elooz, D., Antman, Y., Levanon, N., and Zadok, A.: "High-resolution long-reach distributed Brillouin
sensing based on combined time-domain and correlation-domain analysis," Opt. Express, 2014, 22(6),
6453-6463.
[8] Ferreira, C.: The use of seismic wave velocities in the measurement of stiffness of a residual soil. PhD
Thesis, University of Porto, 2009
[9] Gasparre, A., Hight, D.W., Coop, M.R. and Jardine, R.J.: The laboratory measurement and interpretation
of the small-strain stiffness of stiff clays, Gotechnique, 2014, Volume 64 (12), pp. 942-953. DOI:
10.1680/geot.13.P.227.Bishop, A. W. (1973). The influence of an undrained change in stress on the pore
pressure in porous media of low compressibility. Geotechnique 23, No. 3, 435442.
[10] Goto, S., Tatsuoka, F., Shibuya, S., Kim, Y., Sato, T.: A simple gauge for local small strain
measurements in the laboratory, Soils and Foundations, 1991, 31 (1): 169180.
[11] Gyger, F., Rochat, E., Chin, S., Nikls, M., and Thvenaz, L.: Extending the sensing range of Brillouin
optical time-domain analysis up to 325 km combining four optical repeaters, Proc. SPIE 9157, 23rd
International Conference on Optical Fibre Sensors, 2014, 91576Q.
[12] Hauswirth, D., Puzrin, A. M., Carrera, A., Standing, J. R., & Wan, M. S. P.: Use of fibre-optic sensors
for simple assessment of ground surface displacements during tunnelling. Geotechnique, 2014, 64(10),
837842.
[13] Huang AB, Tai YY, Lee WF, Ishihara K: Sampling and field characterization of the silty sand in Central
and Southern Taiwan. 3rd International Conference on Site Characterization (ISC-3), 2008, Taipei,
Taylor & Francis:1457-1463
[14] Huang, A., Ho, Y., Lee, J., Wang, C.: Fiber Optic Sensored Geotechnical Testing and Field Monitoring.
Sound Geotechnical Research to Practice, Proc. Sound Geotechnical Research to Practice: Honoring
Robert D. Holtz II, 2013, ASCE GSP 230: 382-404.
[15] Ibraim, E., Di Benedetto, H.: New local system of measurement of axial strains for triaxial apparatus
using LVDT, Geotechnical Testing Journal, 2005, 28 (5): 11630
[16] Iten, M., Puzrin, A. M., and Schmid, A.: Landslide monitoring using a road-embedded optical fiber
sensor," In W. Ecke, K. J. Peters, & N. G. Meyendorf (Eds.), Proc. SPIE Voi. 6933: Smart Sensor
Phenomena, Technology, Networks and Systems, 2008, 6933, 6933151.
[17] Jamiolkowski, M. and Masella, A.: Geotechnical characterization of copper tailings at Zelasny Most
Site. Keynote Lecture, DMT'15 The 3rd Int. Conf. on the Flat Dilatometer Rome, Italy (Marchetti,
Monaco & Viana da Fonseca, eds ISBN 979-12-200-0116-8), 2015: 25-42.
[18] Kazuo T., Kaneko S.: Undisturbed sampling method using thick water-soluble polymer solution Tsuchi-
to-Kiso, Journal of the Japanese Geotechnical Society [In Japanese], 2006, 54(4): 145-148.3.
[19] Kersey, A. D., Berkoff, T. A., and. Morey, W. W.: Multiplexed fiber Bragg grating strain-sensor system
with a fiber FabryPerot wavelength filter. Optics letters, 1993, 18.16: 1370-1372.
[20] Klar, A. and Linker, R.: Feasibility study of automated detection of tunnel excavation by Brillouin
optical time domain reflectometry, Tunnelling and Underground Space Technology, 2010, 25(5), 575
586.
[21] Klar, A. Uchida, S. and Levenberg E.: In situ stiffness profiling using high resolution fiber optic
distributed sensing. Journal of Geotechnical and Geoenvironmental Engineering, 2015a, Under review.
[22] Klar, A. Ucihda, S. and Levenberg, E.: In situ and laboratory mechanical characterization using high-
resolution fiber optic distributed sensing. 6 th International Symposium on Deformation Characteristics
of Geomaterials, 2015b, 15-18 Nov, Buenos Aires 2015.
[23] Klar, A., Bennett, P. J., Soga, K., Mair, R. J., Tester, P., Fernie, R., St John, H.D. and Torp-Peterson, G.
Distributed strain measurement for pile foundations, Proceedings of the Institution of Civil Engineers-
Geotechnical Engineering, 2006, 159(3): 135144.
[24] Klar, A., Dromy, I., and Linker, R.: Monitoring tunneling induced ground displacements using
distributed fiber-optic sensing. Tunnelling and Underground Space Technology, 2014, 40, 141150.
[25] Kurashima, T., Horiguchi, T., and Tateda, M.: Distributed-temperature sensing using stimulated
Brillouin scattering in optical silica fibers, Opt. Lett. 1990, 15(18): 1038-1040.
[26] Lee J., Tien K., Te Ho Y., Huang A.: A Fiber Optic Sensored Triaxial Testing Device. Geotechnical
Testing Journal, 2011, 34 (2): GTJ102825
[27] London, Y., Antman, Y., Cohen, R., Kimelfeld, N., Levanon, N., and Zadok, A.: High-resolution long-
range distributed Brillouin analysis using dual-layer phase and amplitude coding. Opt. Express, 2014.
22, 27144-27158.
[28] Lu, Y., Zhu, T., Chen, L., and Bao, X.: Distributed vibration sensor based on coherent detection of
phase-OTDR. Lightwave Technology, 2010, 28(22), 3243-3249.
[29] Martin, G. and Mayne, P.: Seismic flat dilatometer tests in Connecticut valley varved clay, ASTM
Geotechnical Tesing Journal, 1997, 20(3):357-361.
[30] Minardo A, Picarelli L, Avolio B, Coscetta A, Papa R, Zeni G, Di Maio C, Vassallo R, Zeni L.: Fiber
optic based inclinometer for remote monitoring of landslides: on site comparison with traditional
inclinometers. In: Proceedings of International Geoscience and Remote Sensing Symposium (IGARSS
2014) / 35th Canadian Symposium on Remote Sensing (35th CSRS), Qubec City, Qubec, Canada,
2014, 4078 - 4081
[31] Mohamad, H., Bennett, P. J., Soga, K., Klar, A., and Pellow, A.: Distributed optical fiber strain sensing
in a secant piled wall. In 7th International Symposium on Field Measurements in Geomechanics, 2007,
ASCE GSP 175, 112, Boston.
[32] Mohamad, H., Soga, K., Pellew, A., and Bennett, P. J.: Performance Monitoring of a Secant-Piled Wall
Using Distributed Fiber Optic Strain Sensing. Journal of Geotechnical and Geoenvironmental
Engineering, 2011, 137(12), 12361243.
[33] Naesgaard, E., Byrne, P., and Wijewickreme, D.: Is P-Wave Velocity an Indicator of Saturation in Sand
with Viscous Pore Fluid?, Int. J. Geomechanics, 2007, 7(6), 437443.
[34] Ng, C. W. W., Leung, E. H. Y., and Lau, C. K.: Inherent anisotropic stiffness of weathered geomaterial
and its influence on ground deformations around deep excavations. Can. Geotech. J., 2004, 41 (1): 12
24.
[35] Ng, C.W.W. & Wang, Y.: Field and laboratory measurements of small strain stiffness 503 of
decomposed granites. Soils and Foundations, 2001, 41, No. 3, 57-71.
[36] Nikls, M., Thvenaz, L., and Robert, P. A.: Simple distributed fiber sensor based on Brillouin gain
spectrum analysis. Opt. Lett., 1996, 21(10), 758-760.
[37] Peled, Y., Motil, A. and Tur, M.: Fast Brillouin optical time domain analysis for dynamic sensing. Optics
Express, 2012, 20, 8584-8591.
[38] Peled, Y., Motil, A., Kressel, I, and Tur M.: Monitoring the propagation of mechanical waves using an
optical fiber distributed and dynamic strain sensor based on BOTDA. Optics Express, 2013, 21, 10697-
10705.
[39] Peled, Y., Motil, A., Yaron, L. and Tur, M.: Slope-assisted fast distributed sensing in optical fibers with
arbitrary Brillouin profile. Optics Express, 2011, 19, 19845-19854.
[40] Peled, Y., Yaron, L., Motil, A. and Tur, M.: Distributed and Dynamic Monitoring of Stain Waves
Propagating at a Velocity of 4km/sec Using a Brillouin Optical Fiber Strain Sensor. Proc. SPIE 8794,
Fifth European Workshop on Optical Fibre Sensors, 2013, Paper No. 879434. doi:10.1117/12.2025640.
[41] Pineda, J.A., McConnell, A., and Kelly, R.B.: Performance of an innovative direct-push piston sampler
in soft clay. In. Proc. 3rd Symposium on Cone Penetration Testing, CPT14, Las Vegas, USA, CPT14
Press, 2014a, 279-288.
[42] Pineda, J.A, Suwal, L.P. and Kelly, R.: Sampling and laboratory testing of Ballina clay. Australian
Geomechanics, 2014b, 49(4), 27-38.
[43] Rocchi, I. & Coop, M.R.: The effects of weathering on the physical and mechanical properties along a
profile of a granitic saprolite. Gotechnique, 2015, In Press.
[44] Song, K., He, Z., and Hotate, K.: Distributed strain measurement with millimeter-order spatial resolution
based on Brillouin optical correlation domain analysis. Opt. Lett., 2006, 31(17), 2526-2528.
[45] Sovran, I., Motil, A., Danon, O., and Tur, M.: An Ultimately Fast Frequency-scanning Brillouin Optical
Time Domain Analyzer. Optical Fiber Communication Conference and Exposition, OFC, 2015, paper
W2A.44, Los Angeles, California, USA.
[46] Sutton, M. A., Orteu, J.-J., Schreier, H.: Image correlation for shape, motion and deformation
measurements: basic concepts, theory and applications, 1st Edition, Springer Publishing Company,
Incorporated, 2009
[47] Taylor M.L., Cubrinovski M., Haycock I.: Application of new 'Gel push' sampling procedure to obtain
high quality laboratory test data for advanced geotechnical analyses. New Zealand Society for
Earthquake Engineering Conference. 2012, Christchurch, New Zealand, NZSEE.
[48] Tutumluer, E. , Garg, N., Thompson, M.: Granular material radial deformation measurements with a
circumferential extensometer in repeated load triaxial testing, Transportation Research Record n. 1614,
1998, 1: 6169.
[49] Uchida, S., Levenberg, E., and Klar, A.: On-specimen strain measurement with fiber optic distributed
sensing. Measurement, 2015, 60, 104113.
[50] Valle Molina, C. & Stokoe II, K.H.: Seismic measurements in sand specimens with varying degrees of
saturation using piezoelectric transducers. Can. Geotech. J., 2012, 49(6): 671685.
[51] Viana da Fonseca, A., Amaral, M.F., Panico, F., Rios, S.: Indexation of dynamic and static
geomechanical properties of a cemented aggregate for transportation engineering. Journal of
Transportation Geotechnics, 2014, 1: 31- 44. Elsevier.
[52] Viana da Fonseca, A., Ferreira, C., and Fahey, M.: A framework interpreting bender element tests,
combining time-domain and frequency-domain methods, Geotechnical Testing Journal, 2009, 32 (2): 1
17
[53] Viana da Fonseca, A.: Identifying the reserve of strength and stiffness characteristics due to cemented
structure of a saprolitic soil from granite. 2nd International Symposium on Hard Soils Soft Rocks,
Naples, 1998, 1: 361- 372. A. A. Balkema, Rotterdam.
[54] Vorster, T. E. B., Soga, K., Mair, R. J., Bennett, P. J., Klar, A. and Choy, C. K. (2006). The use of fibre
optic sensors to monitor pipeline response to tunneling. In ASCE Geocongress, 2006, 16.
[55] Yang, J.: Liquefaction resistance of sand in relation to p-wave velocity. Gotechnique, 2002, 52 (4):
295298.
[56] Yimsiri, S., Soga, K., Chandler, S.: Cantilever-type local deformation transducer for local axial strain
measurement in triaxial test, Geotechnical Testing Journal, 2005, 28 (5): 11432.
[57] Yoneda, J., Masui, A., Tenma, N., & Nagao, J.: Triaxial testing system for pressure core analysis using
image processing technique. The Review of Scientific Instruments, 2013, 84(11), 114503.
Reviewed Papers
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-267
Behavior of compacted unsaturated soil in

isotropic compression, cyclic and
monotonic shear loading sequences in
undrained condition
Ali MurtazaRASOOLa,1,JiroKUWANOa and Shinya TACHIBANAa
a
Department of Civil & Environmental Engineering, Saitama University
Abstract. Unsaturated soils are most common in geotechnical practice. In order to

study mechanical behavior of unsaturated soil a series of constant water content
tests were performed in this study. Soil specimens were tested under isotropic
loading-unloading, monotonic and cyclic loading sequences to study the effects of
stress history on change in void ratio and suction, and on shear characteristics of
unsaturated soil.It was concluded that the mechanical behavior of silty soil named
as DL clay is not affected significantly by loading sequences incorporated.
Keywords.Cyclic consolidation, Cyclic loading, Triaxial test, Undrained
1. Introduction
In geotechnical practice most soils are often in unsaturated condition and shear testing
of soil in unsaturated conditions is not common in soil mechanics laboratories. The
constant water content test corresponds to a field condition in which the rate of loading
is much higher than the rate at which the pore water is able to drain out of the soil. The
mechanical behavior of unsaturated soil has been discussed by several researchers, e.g.
Fredlund et al. (1978); Cui and Delage(1996); Geiser et al.(2000). Many of studies
were conducted under drained conditions. In practice, drained conditions for water and
air cannot always be attained. Therefore, it is necessary to investigate the unsaturated
soil behavior under fully undrained conditions by measuring both the pore air pressure
and the pore water pressure. Yabuki et al. (2007) conducted cyclic triaxial tests of
unsaturated silt under fully undrained conditions, and measured the changes in pore air
pressure and pore water pressure separately. Unno et al. (2008) performed cyclic tests
on unsaturated sand with low levels of suction, in which the pore air pressure as well as
the pore water pressure were measured. Oka et al. (2010) conducted monotonic triaxial
tests with different strain rates for unsaturated compacted silt, and showed that the
strain rate affected the stress-strain relations and the stress paths for the fully undrained
tests. In this study, a series of monotonic and cyclic load tests in undrained conditions
has been conducted on silty soil. The effect of stress history on change in void ratio and
suction and on shear characteristics of unsaturated soil was studied in this research.
1
Corresponding Author: Department of Civil & Environmental Engineering Saitama University, 255
Shimo-okubo, Sakura-ku, Saitama-shi, Saitama, Japan; Email:[email protected].
268 A.M. Rasool et al. / Behavior of Compacted Unsaturated Soil
2. Description of the test device
Figure 1 gives a schematic diagram of the testing device for cylindrical specimens
50mm in diameter and 100mm in height. In order to separate the routes for the
measurement and the control of the pore air pressure and the pore water pressure, a
membrane and a PTFE (poly tetra fluoro ethylene, known as Teflon) sheet were used.
The PTFE sheet was placed on the top of the specimen to cut off the flow of water and
to control the air pressure. The membrane filter was installed in the lower pedestal to
cut off the flow of air and to measure the water pressure. The thin membrane filter with
pores of 0.45mm has air entry value of 420kPa. For undrained tests, no pore water was
allowed to flow out in both the isotropic consolidation and monotonic shearing process.
The salient feature of this triaxial apparatus is that both pore air and water pressures
can be measured separately. Pore air pressure transducer was installed in the top cap
and connected to air regulator in order to give continuously supply of the air
throughout the test. In addition to this, solenoid valve was also installed in the air
drainage line inside the top cap. The change in volume of specimen during the triaxial
tests was measured as the change in the water level in the inner cell by LCDPT (Low
Capacity Differential Pressure Transducer). Both monotonic loading tests and cyclic
loading tests were conducted under strain-controlled conditions. The vertical load of
the soil specimen was measured by the inner load cell. All instruments were connected
with dynamic strain amplifiers (used to amplify data obtained from strain
gauges/transducers) which were then connected to A/D board and finally to PC.
Displacement Cell Pressure

LCDPT
Personal cable
Computer tube
Load cell
Upper
Tank
Cap
Dynamic
PTFE
pore pressure
displacement
cell pressure
air pressure
strain
load cell
balance
amplifiers Balance
Membrane filter
Inner cell
Lower
Outer cell Tank
Pedestal
Transducers
Pore Pressure
Figure 1.Schematic diagram of triaxial cell
3. Physical properties of soil
The silty soil known as DL clay in Japan was tested in this study. The grain size
distribution curve for the soil tested is shown in Figure 2.The soil density is 2.635g/cm3
and the liquid limit is non-plastic. The reason of using this clay is that it has lower
initial suction than that of kaolin clay under same degree of saturation.
A.M. Rasool et al. / Behavior of Compacted Unsaturated Soil 269
Figure 2.Grain size distribution curve
4. Sample preparation
In this research static compaction has been used for preparation of samples. The
purpose of using static compaction as opposed to dynamic compaction is to obtain
more homogenous specimen in terms of density. The specimens were mixed well with
water to make up the water content of approximately 25.30%, and statically compacted
in layer of 2cm (each) in a special apparatus with a hydraulic jack in order to obtain
homogeneity. The specimens were compacted to achieve a dry density of about
1.30g/cm3. The initial properties of all specimens are listed in Table 1. As the
procedure of static compaction was applied, it can be seen that all specimen have
almost same properties.
Table 1. Initial properties of specimens

Specimen 100 200/100 400/100 100/CS 200/100/CC
Ht. (cm) 9.96 9.96 9.96 9.96 9.96
Dia. (cm) 4.94 4.94 4.94 4.94 4.94
- (%) 25.30 25.28 25.20 25.00 25.20
(g/cm3) 1.29 1.29 1.30 1.30 1.30
Sr (%) 64.30 64.26 64.18 64.03 64.37
5. Test procedure
In this research five compacted unsaturated soil specimens with almost same properties
were prepared using DL clay and sheared monotonically to failure under net confining
stress of 100kPa after different stress histories. Specimen 100 was consolidated
isotropically under net pressure of 100kPa and shared monotonically. Specimen
200/100 was consolidated under net pressure of 200kPa then unloaded to 100kPa,
and sheared monotonically. Specimen 400/100 was consolidated under net pressure
of 400kPa then unloaded to 100kPa and sheared monotonically. Specimen 100/CS
was consolidated under net pressure of 100kPa then a small cyclic deviatoric stress of
,q=+20kPa was applied just before monotonic shearing. Specimen 200/100/CC was
consolidated under net pressure of 200kPa and sheared monotonically after isotropic
unloading/reloading cycles of 200/100kPa. The net pressure was increased in 5, 10,
15kPa steps during consolidation with interval of one hour and left overnight before
shearing. Schematic illustration and procedure of test series is shown in Figure 3(a,b).
Figure 3.(a)Schematic illustration of test series (b) Testing Procedure
6. Experimental results
6.1. Initial Suction
It is clear that initial suction influences the behavior of unsaturated soil. However, the
suction depends on pore-water distribution and degree of saturation. Figure 4 shows the
variation of the monitored values of initial suction with time of statically compacted
specimens with degree of saturation approximately 64% when they were set in the
triaxial apparatus. Due to almost same values of degree of saturation for all specimens,
it generally look that initial suction stabilizes after 20min and all curves converged to
an initial suction of about 15 17kPa. The less time in stabilizing is due to use of thin
membrane filter having air entry value of 420kPa.
Time (min)
0 5 10 15 20
0
-2 100
Pore water pressure (kPa)
200/100
-4 400/100
100/CS
-6 200/100/CC
-8
-10
-12
-14
-16
-18
-20
Figure 4.Initial suction of specimens
6.2. Effect of stress history on deformation prior to triaxial compression
In isotropic loading stage, stresses in three directions are equal i.e. (dx=dy=dz=d3)
and no shear stress is developed. It is supposed that soil will undergo equal
deformations in all direction when subjected to an equal all around pressure or isotropic
pressure. The pre-consolidation pressure during the static compaction was kept
approximately 100kPa on all specimens. During isotropic consolidation the isotropic
load applied on specimens were 100, 200 and 400kPa which was increased in steps of 5,
10, 15kPa with a time interval of one hour for each loading and unloading stage.
1.8
400/100-Vertical strain
1.6 400/100-Horizontal strain
100/CS-Vertical strain
1.4
Consolidation strains (%)
100/CS-Horizontal strain
200/100/CC-Vertical strain
1.2 200/100/CC-Horizontal strain
1
0.8
0.6
0.4
0.2
0
0 50 100 150 200 250 300 350 400
log (3 - ua)
Figure 5.(a)Effect of isotropic loading on specimen, (b)loaded up to 35kPa
It was observed that the vertical strain of specimen was almost equal to horizontal
strain i.e. v = h up to 100kPa. Therefore, specimens behave almost isotropically when
loaded up to 100kPa. However, anisotropy was observed when loaded beyond 100kPa
due to over-consolidation load as shown in Figure 5(a). Beyond 100kPa, the vertical
strain of specimens was less than horizontal strain i.e. v<h. Isotropic behavior of soil
was also observed during loading and unloading phase as shown in Figures5(b) and
6(a,b). In order to check the isotropic behavior during loading and unloading phase
vol = 3h is plotted against vol = v + 2h. It was observed that specimen behave
isotropically when loaded up to 100kPa i.e. up to pre-consolidation pressure, whereas
anisotropy was observed when specimen was loaded upto 200kPa, unloaded to 100kPa.
Figure 6.Effect of isotropic loading on specimen (a) loaded up to 200kPa, (b) unloaded to 100kPa
6.3. Change of void ratio with change of net stress
Unsaturated soils are characterized by the presence of an air phase, a water phase and
an air-water interface in the voids. The void changes with increase or decrease in load
resulting in change in characteristics of the unsaturated soil. The change of void ratio
with change of net pressure in undrained condition was observed in this research. It
was noticed that void ratio decreases with increase in net pressure for all the specimens
during the loading phase as shown in Figure7 (a). No significant change in void ratio
was observed in specimens when simply unloaded from 200kPa to 100kPa and 400kPa
to 100kPa. A little more decrease in void ratio was observed in specimen 200/100/CC
when isotropic unloading/reloading cycles were applied as shown in Fig. 7(b). It was
also observed that due to application of deviatoric stress cycles of ,q=+20kPa in the
case of 100/CS, there was just a little decrease in void ratio because of small value of
cyclic deviatoric stress.
Figure 7.(a) Effect of net stress on void ratio, (b) Effect of unloading/reloading on void ratio
6.4. Effect of net stress change on degree of saturation
As change of void ratio changes degree of saturation, the water-retention curve will be
affected during loading and unloading of net stresses. Figure 8 shows the effect of
change in degree of saturation on suction. With the increase in net stress, the void ratio
decrease, degree of saturation increases and pore water pressure increases. Due to
increase in pore water pressure, suction decreases. Therefore it is noted that suction
decreases with increase in degree of saturation during the loading phase for all the tests
and increase during unloading phase for test 400/100. In 200/100/CC, as a result of
decrease in void ratio due to cyclic consolidation, the degree of saturation increased
during unloading/reloading cycles. In 100/CS test, suction and degree of saturation
remained almost constant for cyclic deviatoric stress application.
73
200/100/100
72 70.40
Degree of saturation (%)
71 70.30
100 70.20
70
200/100
Sr (%)
70.10
69 200/100-unloading
400/100 70.00 Increase in degree
68 400/100-unloading of saturation due to
69.90 isotropic
100/CS ,q =+20kPa unloading/reloading
67 100/CS-cyclic loading 69.80 cycles
200/100/CC
66 200/100/CC-cyclic loading 69.70
65 13 13.5 14 14.5 15
0 5 10 15 20 Suction (kPa)
Suction (kPa)
Figure 8.(a) Effect of degree of saturation on suction, (b) Effect of load/reloading on degree of saturation
6.5. Effect of axial strain on suction
Effect of axial strain on suction was studied during monotonic shearing. It was
observed that suction increased with increase in axial strain for almost all the
specimens as shown in Figure9. The suction increased with decrease in pore water
pressure during monotonic shearing.
25
400/100
20 200/100/CC
200/100
100/CS
15
Suction (kPa)
100
10 100
200/100
400/100
5 100/CS
200/100/CC
0
0 5 10 15
Axial strain (%)
Figure 9.Effect of axial strain on suction
6.6. Stress strain relationships in monotonic and cyclic loadings
Most of the experimental investigations conducted on unsaturated soil have been

performed under a constant air pressure. In the present study, silty soil was sheared
without drainage of both air and water. A small cyclic deviatoric stress of +20kPa was
given to specimen 100/CS just before monotonic shearing, whereas all other specimens
were subjected to monotonic shear right after isotropic consolidation. Change in axial
strain with cyclic deviator stress application in 100/CS is shown in Figure 10 (b) and
other monotonic shear test results in Figure 10 (a). 100/CS specimen showed gradual
increase in the axial strain up to 0.05% after 20 cycles. As suction and net pressure was
almost the same for all specimens, they showed almost same behavior. However, the
400/100 specimen showed the highest stiffness among others probably due to the
highest maximum past pressure of400kPa. The 200/100/CC specimen also showed
higher stiffness owing to the less void ratio caused by isotropic unloading/reloading
cycles. It was also noted that, for all the specimens, suction increased due to decrease in
pore water and dilatant behavior with increase in axial strain.
250
200/100/CC
25
100/CS 100
200
400/100 20
200/100
Deviatoric stress (kPa)
15
Deviatoric stress (kPa)
10
150
5
0
100 -0.04 -0.02 0.00 0.02 0.04 0.06 0.08 0.10
100 -5
200/100 -10
400/100
50 100/CS -15
100/CS
200/100/CC -20
-25
0 Axial strain (%)
0 2 4 6 8 10 12 14
Axial strain (%)
Figure 10.Effect of (a) monotonic loading, (b) cyclic deviatoric stress
6.7. Effective stress path
A change in the stress state of a soil can be described using stress paths. A stress path is
a curve drawn through the stress points for successive stress states. As we have
performed test under undrained conditions, matric suction was not constant during the
monotonic shearing. Therefore, the stress paths in Figure 11 are for undrained
conditions. The stress path was drawn by taking deviatoric stress (q) and p. p is
combination of net and suction stress. It was observed that as net stress during
monotonic shearing, deviatoric stress and matric after monotonic shearing was same
for all specimen, therefore they showed same stress path.
250
200
150
q
100 100
200/100
400/100
50 100/CS
200/100/CC
0
0 50 100 150 200 250
p'
Figure 11.Effective stress path
7. Conclusions
A series of tests on DL clay shows that suction decreases with increase in degree of
saturation as void ratio decreases with increase in net pressure. Unloading does not
affect void ratio or degree of saturation whereas cyclic consolidation results in a little
decrease in void ratio and increase in degree of saturation. Unsaturated soils behave
isotropically under isotropic loading at low net pressure. Axial strain increases as a
result of cyclic deviatoric stress. Suction increases with increase in axial strain during
monotonic shearing. Finally mechanical behavior of DL clay is not affected
significantly by loading sequences incorporated in this study.
References
[1] Kimoto, S. and Oka, F. (2011), Monotonic and cyclic behavior of unsaturated sandy soil under drained
and fully undrained conditions, Soils and foundation 51, 663-681.
[2] Oka, F., Kodaka, T., Suzuki, H., Kim, Y., Nishimatsu, N. and Kimoto, S. (2010), Experimental study on
the behavior of unsaturated compacted silt under triaxial compression, Soil and foundation 50, 695 - 700.
[3] Jonggil Chae, Byeongsu Kim, Seong Park, and Shoji Kato (2009), Effect of suction on unconfined
compressive strength in partly saturated soils, KSCE Journal of Civil Engineering 4, 281-290.
[4] Unno, T., Kazama, M., Uzuoka, R. and Sento, N. (2008), Liquefaction of unsaturated sand considering
the pore air pressure and volume compressibility of soil particle skelton, Soil and foundation 48, 87 - 99.
[5] Yabuki, T., Oka, F., Kimoto, S., Higo, Y. and Watanabe, T. (2007), Mechanical behavior of unsaturated
silt under cyclic loading, Proc. 42nd Annual meeting of JGS, Nagoya, 771 772
[6] Geiser F., Laloui L. &Vulliet L. (2000),On the volume measurement in unsaturated triaxial test, Proc.
Conf. Unsaturated Soils for Asia, Unsat-Asia,Balkema, Rotterdam, 669-674.
[7] Cui Y.J. &Delage P. (1996), Yielding and plastic behaviour of an unsaturated compacted silt,
Gotechnique 46, 291- 311.
[8] Fredlund, D. G. &Rahardjo, H. (1993), Soil Mechanics for Unsaturated Soils, John Wiley and Sons Inc.
[9] Fredlund, D. G., Morgenstern, N. R., and Widger, R. A. (1978), The shear strength of unsaturated soils,
Canadian Geotechnical Journal, 15, 313321.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-275
Influence of Initial Stress/Strain State on

the Coefficient of Earth Pressure at Rest
Ali Akbar ELIADORANIa,1 and Yoginder VAID b
a
Professor, South Carolina State University, Orangeburg, SC, USA
b
Professor Emeritus, The University of British Columbia, Vancouver, BC, Canada
Abstract. An experimental technique for the K0, the coefficient of earth pressure at
rest, is presented. The technique consists of drained vertical compression of a triaxial
specimen using the strain path control, in which a positive control of the strain
increments rather than the conventional stress increments is exercised. The method
does not require the use of any lateral strain sensors as used in previous studies and
is free from any side friction affects, typical of oedometer testing. A detailed series
of tests on three (3) sands are presented to illustrate the capabilities of the proposed
technique. Possible influence of stress/strain history the soil inevitably experiences
during sampling or reconstitution together with the initial state of stress prior to
initiating zero lateral strain loading on the measured K0 are critically reviewed and
evaluated. It is shown that the requirement of absolute zero lateral strain may not be
essential for obtaining credible estimates of K0 as long as the ratio of radial to axial
strain during vertical compression does not exceed about 3 to 5 %.
Keywords. strain path control, membrane penetration, K0, triaxial test, sand
1. Introduction
The compression history of many soils that are deposited over areas of large lateral extent
dictates increasing overburden stress v must have been entirely one-dimensional in the
vertical direction and zero strain in the lateral direction. The horizontal effective stress
h in these one-dimensionally compressed soil deposits is often expressed as v=K0 v
in which K0 is called the coefficient of earth pressure at rest. In geotechnical engineering
problems, particularly those which requires determining ground movements and
distribution of stresses and pore pressures induced on loading, a credible estimate of in-
situ K0 is thus of central practical importance.
In the laboratory methods of K0 determination, the constraint of zero lateral strain is
enforced during drained (u = 0, where u is the excess pore pressure) monotonic vertical
compression of a cylindrical soil specimen, and the effective stresses that develop in the
lateral direction are recorded. Instrumented oedometers capable of measuring the hoop
stress (e.g. [1-2]), and the triaxial test devices with radial strain sensors have been used
for this purpose (e.g. [3, 4]). These techniques when used with undisturbed specimens
must commence loading from an initial condition where the soil is no longer in a state of
absolute zero lateral strain, as it would ideally be under the undisturbed in-situ condition.
It would have experienced a small, but, an unknown amount of compressive lateral strain
(assuming in-situ K0 <1) due to the inevitable release of in-situ shear stresses on
1
276 A.A. Eliadorani and Y. Vaid / Inuence of Initial Stress/Strain State
undisturbed sampling [5]. Moreover, after sampling the residual stress state in the soil
becomes hydrostatic (since a=r=0 where a and r are axial and radial stresses,
respectively, in the triaxial test). Similar situation exists when testing reconstituted sand
specimens. It is because of the small vacuum needed to confine specimens after their
reconstitution. The oedometer technique, in addition, suffers from the unavoidable side
friction effects. As a result of these limitations it is not known, to what extent the release
of in-situ shear stress and now the associated residual hydrostatic effective stress
conditions following undisturbed sampling, and the side friction in oedometer
compression, influence the measured values of K0 on subsequent loading [6]. It must be
emphasized that the laboratory procedures can impose the constraint of zero lateral strain
only with respect to the specimen configuration that ensues after sampling, and any
further unknown strains associated with sample extrusion, trimming and set up in the
apparatus.
This paper describes a study of the measurement of K0 of reconstituted sand. The
experimental technique utilizes strain incremental instead of the usual stress incremental
technique of loading the specimen in a fully drained (=, u = 0) triaxial test. No
radial strain sensors are used. Instead, the condition of zero lateral strain, from the start
of vertical loading, is enforced by the strain path loading control, such that the axial and
volumetric strain increments (da and dv) in the soil skeleton are equal [7, 8]. It
automatically guarantees zero radial strain increment. Clearly, the loading rate in these
tests must be sufficiently slow so that negligible excess pore pressures develop in the
specimen. Such a rate is determined by trial. A number of identical specimens are
subjected to successively slower rates of loading. When further reduction in the loading
rate does not influence the results (i.e. r/a) at any mobilized a, this is taken as the
satisfactory rate for the test. Alternatively, pore pressure measurements at both instead
of only one end of the specimen can be used as a confirmation of this criterion. The
effects of deliberately allowing small lateral strain increments to occur during loading,
beyond the 0 state, instead of maintaining them strictly zero, on the measured K0 are
also explored.
2. Test materials
Three different granular soils (sands) were used for experimentation. All were relatively
uniformly graded, but differed in their mean grain size D50 and grain angularity. Their
physical properties are listed in Table 1. In triaxial tests on granular soils the observed
volume change under increasing r exceeds the volume change of the soil skeleton by
an amount due to the membrane penetration into the peripheral voids. For a given
increment in r, the magnitude of this membrane penetration volume change depends
primarily on the D50 of the sand [9], and for a given sand, it should conceptually, further
depends on its relative density, Dr. It is because both parameters influence the size of the
peripheral voids into which the membrane penetrates. The determination of a credible
value of membrane penetration volume changes, specific to each sand at the test density
is therefore very important. The credibility of the measured K0 by the proposed method
will clearly depend on the confidence with which membrane penetration values have
been evaluated.
A.A. Eliadorani and Y. Vaid / Inuence of Initial Stress/Strain State 277
Table 1. Physical properties of sands used.

Sand emax emin D50 Cu
(mm)
Fraser River sand 0.900 0.594 0.35 1.72
(subangular)
Ottawa sand 0.820 0.500 0.40 1.63

(rounded)
Silica sand 0.600 0.420 0.90 1.20

(subangular)
Membrane penetration values specific to each sand were separately determined over
a range of its relative density Dr. This was done by the method suggested by Vaid and
Negussey [10]. The results from these measurements are shown in Figure 1. Figure 1(a)
illustrates the relationship of unit membrane penetration (membrane penetration volume
per unit surface area of the specimen) versus log radial effective stress r at four relative
density states. Linear relations (slope m) may be noted at each density. The magnitude
of incremental volumetric change due to membrane penetration, for an incremental
change dr in the radial stress is then given by
)r d )r
dVm m log As (1)
)r
in which As = the surface area of the triaxial specimen covered by the membrane and m
is specific to the density of the test specimen. The data illustrated in Figure 1(a) confirms
the expected dependence of m on Dr for a given sand, whereas, Figure 1(b) illustrates
the dependence of m on D50 at a given Dr among different sands. The former fact, can
be very important, but has often been ignored by previous researchers. For Fraser River
sand, for example, m at Dr = 80% is only half of the value at Dr = 20%.
3. Experimental procedure
Triaxial specimens of sand, nominally 62 mm diameter x 125 mm high, were

reconstituted by water pluviation (WP). This reconstitution technique yields repeatable
6
0.012
Unit membrane penetration in cm (volume
per unit surface area of membrane) x 1000
e (Dr)
0.836 (21%)
0.803 (32%) Silica

0.733 (55%) m
4
0.650 (82%)
0.008
D50 = 0.9 mm
1
m
Ottawa
2 0.004 D50 = 0.4 mm
D50 = 0.35 mm
(a) Fraser River
0 (b)
0.000
10 100 1000 0 40 80
Radial effective stress, 'r (kPa) Dri (%)
Figure 1. Membrane penetration correction curves (a) unit membrane penetration (b) dependence of unit
membrane penetration on relative density for three sands.
uniform specimens, whose fabric has been shown to mimic that of natural fluvial and
hydraulic fill sands [11]. Therefore, laboratory studies on WP sands provide a convenient
means of assessing their K0 in-situ. All test specimens were deposited in the loosest
conditions and then densified to target the desired density.
The details of the strain path control capable triaxial apparatus and experimental
procedures are described elsewhere [8]. All test specimens, following their reconstitution
by WP, were confined by a vacuum of about 20 kPa. After setting up the triaxial cell in
the loading frame, this confining vacuum was replaced by the difference in the applied
cell and back pressures. The configuration of the test specimen under this finite
hydrostatic stress state, 0, was taken as representing zero strain - both axial and radial.
An assumption of this kind is necessary as the precise zero reference of lateral strain
cannot be known. This is true for both undisturbed as well as reconstituted specimens.
A detailed series of tests was carried out on Fraser River sand. In the first series, the
effective stress state ri prior to initiating K0 loading using the dr=0 strain path was
selected as hydrostatic, at the set up value 0 of 20 kPa. In other series, this initial
hydrostatic stress of 20 kPa was first increased to a target value ri before dr =0 loading
was initiated. This was intended to determine what influence does the level of ri, and
hence small radial compressive strains, exert on the measured K0 value. Tests were also
carried out using identical ri = 20 kPa, but the specimen was first loaded conventionally
drained (dr=0) to a targeted R=a/r, prior to initiating dr=0 loading. Thus the initial
state of stress was now non-hydrostatic instead of the hydrostatic used in earlier series.
This phase of pre dr=0 vertical compression loading will cause some additional
compressive axial and compressive or even extensional lateral strain, depending on the
selected magnitude of targeted stress ratio, R. Clearly this will result in a further
departure from the state of zero lateral strain over and above what is inevitable during
sampling or reconstitution of specimen in the laboratory. The effects of this additional
departure from a state of ideal zero lateral strain prior to the commencement of dr=0
loading should become apparent in this series of tests.
For the initial conditions starting from ri values in excess of the selected reference
hydrostatic ri=20 kPa, additional axial and lateral compressive or extensional strains
will occur in the soil depending upon the magnitude of ri. In the reckoning of the total
strain ratios r/a later on, for addressing to what degree the deliberately induced lateral
strain affects the mobilized value of K0, these pre- dr=0 loading strains have been
included to compute the total strains experienced during what has been labeled herein as
dr=0 loading.
4. Test results
4.1. Vertical dr=0 Compression Commencing from Initial Hydrostatic Stress State
The results of a series of tests, namely, vertical dr=0 compression commencing from
initial hydrostatic stress state on Fraser River sand at a loose relative density (Drc) of
about 20% are illustrated in Figure 2. Several levels of ri (or rc) were selected after
the specimens were set up under the reference stress state of 0= a= r= 20 kPa. In
the r versus a plot (Figure 2a), the data from tests with ri 50 kPa may be noted to
soon align themselves along a straight line passing through the origin. The slope of this
line (~0.442) should be the desired K0-value. Since the magnitude of lateral effective
stress ratio K= a/r at the commencement of dr =0 loading is 1.0, it will be expected
that some threshold value of a must be exceeded before this K reduces to approach K0.
In fact, as seen in Figure 2(b), the value of K reduces rapidly from its initial value 1.0 to
within less than about 5% of the eventual K0 state, by the time a has exceeded about
400 kPa. The fact that K0 remains constant over a considerable range of a in excess of
this threshold value may be seen as a confirmation of the typical feature of normally
consolidated soils. It appears that there is a tendency for the threshold a to reduce if
ri selected is smaller (Figure 2b).
It would be of interest to examine to what extent the mobilized a and r are
affected in tests with ri in excess of the reference 20 kPa. Figure 2(c) illustrates the
variation of a/r with a. It may be noted that as soon as this strain ratio r/a falls below
about 3%, the approach to the K0 state appears virtually complete. It should be
remembered that in this series of tests with ri =20 kPa, there is no increase in r, which
stays at the pre- dr=0 loading value (assumed zero). Only a increases, and consequently
the ratio r/a decreases with increase in a (or a). The magnitude of r/a may be noted
to be approximately 2.0 at the start of loading, regardless the level of ri from where
dr=0 loading was initiated. This is a typical characteristic of sands during hydrostatic
loading, and is considered a reflection of its inherently anisotropic nature [12]. The
magnitude of this ratio depends on the relative density of sand. It increases as the density
decreases. The results presented in Figure 2(c) thus show that as long as the ratio dr/da
does not exceed about 3%, a strict adherence to the requirement of zero lateral strain may
not be necessary for obtaining credible estimate of K0. In this series of tests reported,
the small lateral strain prior to dr=0 loading was compressive, which occurred during
hydrostatic compression from ri =20 kPa to the value of ri prior to initiating dr=0
loading. It is seen that for initial hydrostatic ri values in excess of 50 kPa, r/a does
not reduce to a low of about 3%, and therefore no approach to a credible K0 appears
likely even at a=1200 kPa, when the tests were terminated.
4.2. Vertical dr=0 Compression Commencing from Initial Non-Hydrostatic Stress

State
This series of tests was carried out using a value of ri=30 kPa but with initial effective
stress ratios Ri>1 prior to dr=0 loading. It was felt that commencing dr=0 loading from
an estimated anisotropic stress state close to the real K0 stress state in-situ, might prove
more reliable for obtaining K0. The values of Ri= ai/ ri selected in the test series
ranged from 1.0 to 3.4.
The range selected would then surely have straddled the real R corresponding
to K0. The results presented in Figure 3 show that with Ri = 2.28 the approach to the
approximate K0 state occurred at a threshold a, which strangely is greater rather than
that with Ri=1.0. Possible reason for this increased a will be explained later.
Nevertheless, the K0 values from the two series of tests are essentially identical. As noted
earlier Ri=2.28 was specifically chosen because it represented commencement of dr=0
loading from an initially approximate K0 (estimated by K0 1- sin) rather than the
hydrostatic effective stress state, and it was expected that this starting condition might
cause the real K0 state to be reached at a lower than that observed for the initial state
with Ri=(1/Ki)=1. It may be argued that such a behavior might have occurred because
this specimen had experienced (during loading from the zero strain reference hydrostatic
ri=20 kPa, Ri=1.0 to non-hydrostatic ri=30 kPa, Ri=2.28) an extensional lateral strain
of about -0.33% prior to the initiation of dr =0 loading. It is interesting to note that the
measured K first increase above the initial selected value of 0.442, but finally did
stabilize back to an equilibrium K0, when the strain ratio |r/a| dropped below about 3 to
5%. It, therefore, appears that no benefit was achieved by starting dr=0 loading
commencing from an approximately K0 rather than the K=1 state. In fact any extensional
lateral strain history prior to dr=0 loading, if it did occur, may delay rather than
accelerate an ultimate approach to K0 state with a lower threshold a.
It is also of interest to point out that the approach to a |r/a| level of 3 to 5% for
initial Ri=1 occurred from the positive, whereas that for Ri 2.26 it did from the
800 800
(a) (a)
a a
600
' r (kPa)
' r (kPa)
r r
400 400
Drc = 22%
200
Line with K0 = 0.442 Line with K0 = 0.442
passing through origin passing through origin
0 0
0 400 800 1200
0 400 800 1200 1.0
1.0 Ri
Drc = 20%
'rc Drc ' ri = 30 kPa
20 kPa 18%
1.00
2.28
35 kPa 17%
K = ' r/' a
2.80
0.8
K = ' r/' a
50 kPa 17%
3.40
100 kPa 21% 0.6
200 kPa 24%
K = K0
0.6 (b)
(b)
K = Ko
0.2
0.4 0 400 800 1200
0 400 800 1200 'a (kPa)
' a (kPa) 10
c (ii)
5
c (ii)
0
0 20 -10
200 10
200 400 800 1200 0 400 800 1200
0
c (i)
-10
r / a (%)
r / a (%)
-20
0 -30
c (i)
400 800 1200 -40
0 400 800 1200 r/a=0
0
0
r/a=0 (c)
(c)
0 400 800 1200
0 400 800 1200 ' a (kPa)
' a (kPa)
Figure 2 (left). Vertical compression of loose Fraser River sand under zero lateral strain increment initial
state of stress is hydrostatic.
Figure 3 (right). Vertical compression of loose Fraser River sand under zero lateral strain increment initial
state of stress is non-hydrostatic.
negative values. An undisturbed sample after its retrieval from the ground, as pointed
out earlier, has already experienced some compressive lateral strain (for the case if K0
<1). Therefore, the net strain ratio r/a would have never ever been extensional if Ri =1.
On the other hand an extensional lateral strain history may occur for dr=0 loading tests
which commence from some initial states with Ri perceived close to corresponding to
estimated K0, or larger. For in-situ K0>1, extensional lateral strain on perfect sampling
is inevitable. Any negative initial r may be responsible for a slower approach to the
final K0-state.
4.3 Empirical Correlations with Jakys Expression

For each sand conventional drained triaxial tests were carried out to determine the
friction angles. The tests were performed using a range of confining pressure from 100
to 400 kPa and over a range of relative density. K0 was then computed by the Jaky [13]
empirical expression, K0 = 1- sin. The relationship between K0 thus computed versus
relative density for each sand is shown in Figure 4. Superimposed on these curves are
the actual K0 determined by the test method proposed (test with dr=0). The Jaky
empirical expression seems an excellent approximation for estimating K0 value of
Ottawa sand. This expression, however results in somewhat underestimate of K0 for
coarser Silica sand but overestimate of K0 for Fraser River sand. The reason for such
differences may lie in both the average grain size of sand as well as its angularity, the
0.60
Silica Sand
0.50
K0
0.40
K0 test
1- sin*'
0.30
0 20 40 60 80 100
0.6
Ottawa Sand
0.5
K0
0.4
K0 test
1- sin*'
0.3
20 40 60 80 100
0.60
Fraser River Sand
K0 test
0.50 1- sin*'
K0
0.40
0.30
0 20 40 60 80 100
Drc (%)
Figure 4. Comparison of measured K0 with Jakys empirical expression
characteristic which are not embodied in Jaky expression. In that case, the observed
correspondence of observed K0 with Jaky expression may just be fortuitous.
5. Conclusions
An experimental study of the K0 of a loose reconstituted sand has been carried out.
Triaxial specimens were used which were loaded fully drained using the strain path
instead of the commonly used stress path control. An important advantage of the method
is that it dispenses with the use of any lateral strain sensing device or a continual check
on the quality of accumulated axial and volumetric strains.
Based on the tests carried out on reconstituted specimens of three different sands,
the following conclusions can be arrived at:
+ Credible estimate of K0 were obtained by vertical compression of specimens
initially hydrostatically consolidated using the dr =0 strain path control.
+ K0 state arrived at by initiating loading from initially hydrostatic effective stress
state was essentially identical to that using an initially non-hydrostatically state
close to the estimated K0 value.
+ The K0 state was reached sooner, i.e. at a lower mobilized a if the initial
hydrostatic stress prior to strain path loading was lesser.
+ The common belief that a strict condition of zero lateral strain must be adhered
to for credible estimates of K0 was found not to be necessary. As long as the
ratio r/a stayed below about 3 to 5%, no difference in K0 was noted compared
to when it was held strictly zero.
+ The measured K0 for several sands show that both the average grain size of sand
as well as its angularity are not embodied in Jaky expression and the observed
correspondence of observed K0 with Jaky expression may just be fortuitous.
References
[1] C. Mejia, C. et al., Time dependent behavior of sand, In Keedwell, M.J. (Ed.) (1993), 312-326.
[2] G. Mesri and T.M. Hayat, The coefficient of earth pressure at rest, Can. Geotech. J. 30 (1993), 647-666.
[3] A.W. Bishop and D. Henkel, The Measurement of Soil Properties in the Triaxial Test, Arnold Pub, 1962.
[4] V. Sivakumar et al, An assessment of the earth pressure coefficient in overconsolidated clays,
Gotechnique 59 (2009), 825-828.
[5] A.W. Skempton & V.A. Sowa, The behavior of saturated clays during sampling and testing, Geotechniuqe
13, (1963), 269-290.
[6] R. Fukuyawa & H. Ohta, Effect of some factors on K0-value of a sand, Soils & Fndns 28 (1988), 93-106.
[7] B.K. Menzies, A computer controlled hydraulic triaxial testing system, In Advanced Triaxial Testing of
Soil and Rock, ASTM STP 977, 1988.
[8] Y.P. Vaid and A. Eliadorani, Instability and liquefaction of granular soils under undrained and partially
drained states, Can. Geotech. J. 35 (1998), 1053-1062.
[9] S. Frydman et al., The membrane effect in triaxial testing of granular soils, J. of Testing and Evaluation
11(1) (1973), 37-41.
[10] Y.P. Vaid and D. Negussey, A critical assessment of membrane penetration in the triaxial test.
Geotechnical Testing J. 7: (1984), 7076.
[11] M. Oda et al, Experimental study of anisotropic shear strength of sand by plane strain test. Soils & Fndns,
18(1) (1978), 2538.
[12] D. Negussey & Y. Vaid, Sand deformation under proportional loading, Can Geo J. 23 (1986), 155-163.
[13] J. Jaky, The coefficient of earth pressure at rest, J. Society of Hungarian Arch & Eng (1944), 355-358.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-283
Small Strain Modulus of Bio-Cemented

Sand
Youventharan DURAISAMYa and David AIREY b,1
a
Faculty of Engineering, University of Malaysia Pahang
b
School of Civil Engineering, University of Sydney
Abstract. There is growing interest in using microbially induced calcite

precipitation (MICP) for ground improvement. As part of a study to assess the
performance of ex-situ mixed bio-cemented sand columns, a series of triaxial tests
has been conducted to quantify the effects of the bio-cementation. Bender
elements mounted in the end platens of a triaxial cell have been used to monitor
the shear wave velocity throughout the cementation process and then during
application of stress in the triaxial shearing tests. The results of tests on the bio-
cemented sand are compared with tests on gypsum cemented and uncemented
specimens. To assess the bio-cemented specimens a series of standard drained
triaxial tests with bender elements were performed on Sydney sand which was
mixed with urea contents in the range of 5 to 20% by weight, urea to calcium
chloride in a 1:1 molar ratio, and a bacterial broth containing Bacillus Megaterium.
Bacillus Megaterium is a ureolytic bacterium which hydrolyses urea to precipitate
calcium carbonate (calcite). After curing the samples were subjected to various
levels of mean effective pressure. The results show a good correlation between the
amounts of urea and calcite precipitated and between the calcite precipitated and
the degree of cementation achieved. For a given amount of cementation higher
moduli and strengths were measured for the bio-cemented specimens than when
using gypsum. Creating the bio-cementation by mixing produces homogeneous
specimens with similar strengths and stiffnesses to the commonly used flushing
technique.
Keywords. MICP; Shear wave velocity; Small-strain shear modulus; Bender

elements; Triaxial Tests
1. Introduction
Bio-cementation is a new technology with potential applications in ground stabilization,

which makes use of biologically induced bacteria-driven cementation. This approach
promises to be far less costly, more effective and more environmentally sustainable
than current cement based methods of ground improvement. However, the technology
remains in its infancy and requires further experimental and theoretical support and
optimization to reach its full potential. Predicting the behaviour of bio-cemented soil
presents a unique set of challenges because, unlike conventional processes where the
level of cementation depends simply on the amount of binder used, the bio-cementation
process depends on the ability of bacteria to produce enzymes (urease activity), the
amount of nutrients and on the chemical conditions (pH, Eh, temperature) [4, 8, 15].
These factors have been reported to contribute to the variance in the mechanical
properties of bio-cemented samples [3, 6].
1
Corresponding author: School of Civil Engineering, The University of Sydney,
NSW, 2006, Australia: email: [email protected]
284 Y. Duraisamy and D. Airey / Small Strain Modulus of Bio-Cemented Sand
Many natural soil deposits have some degree of cementation, which can occur due
to the precipitation of minerals such as salts, iron oxides and calcite [13]. One of the
characteristics of these deposits is their variability, and this has led to research into the
mechanical behavior of cemented soils using predominantly artificial cementing agents
such as lime, ordinary Portland cement and gypsum. The behavior of cemented soils
has been extensively investigated, and it has been shown that all cemented soils show
similar patterns of behaviour, but nevertheless, there are significant differences in the
effectiveness of different cements [11]. It is well understood that cementing agents act
to increase strength, stiffness and brittleness and that the effects of cementation
increase as the cementation level increases and as the confining stress decreases [9, 11,
13]. Studies using calcite as the cementing agent have been reported using the CIPs
technique [12], and more recently using MICP [3, 4, 6]. The available research into the
behaviour of calcite cemented sands suggests that effective cementation only occurs at
the points of contact between sand grains and that the filtering effect, produced when
bacterial solutions are flushed through the soil, enhances this effect.
To monitor the cementation process and assess its effectiveness many studies have
used shear wave velocity (or small strain stiffness to which it is simply related) as this
may be measured non-destructively, and small amounts of cement have been shown to
cause large increases in the small strain stiffness of naturally and artificially cemented
sands [2, 13, 14]. In this paper the results of shear wave velocity and small strain
stiffness from tests on bio-cemented sand, produced simply by mixing all the
ingredients, are reported. It is shown how using bender elements the response
throughout the tests, during curing, compression and shearing can be monitored, and
the effectiveness of the bio-cementation determined.
2. Experimental Procedure
Cylindrical soil specimens of cemented sand have been prepared for triaxial and UCS
tests. These have been comprised of Sydney sand with cement produced by either
gypsum or microbially induced calcite precipitation (MICP). Sydney sand is a uniform
clean quartz sand with no fines, a mean particle size, d50 = 0.3 mm, uniformity
coefficient, Cu = 3, coefficient of curvature, Cc = 1.2, and specific gravity, Gs = 2.65.
The maximum and minimum void ratios are 0.78 and 0.57 respectively.
Specimens with diameter of 55 mm and height to diameter ratios of about 2 were
prepared by pouring the sand and sand-cement mixtures into a split mould mounted on
the base pedestal of the triaxial apparatus. In the majority of the tests the triaxial
pedestal and top platen were fitted with bender elements to monitor the shear wave
velocity. For specimens subjected to triaxial tests the soil mixture was placed into a
rubber membrane, whereas for the UCS tests no membrane was used.
The gypsum cemented samples were prepared by mixing gypsum, sand, and water
with amounts of gypsum in the range of 5% to 20% of the dry weight of the sand. The
dry Sydney sand and gypsum powder were thoroughly mixed together and then
distilled water, weighing around 10% of the total weight of sand and cement, was
added and the mixture stirred thoroughly for about a minute before it was placed into
the mould. In order to prepare repeatable and uniform specimens the mixture was
placed in five layers, each of which were lightly tamped. As soon as the top platen was
placed the bender element readings were acquired and monitored continuously
throughout the curing process. After 24 hours the split mould was removed.
Y. Duraisamy and D. Airey / Small Strain Modulus of Bio-Cemented Sand 285
A similar procedure was used for the bio-cemented specimens, however in this case
additional steps were required to prepare the bacterial solution with an appropriate
biomass before mixing with the sand [7]. To prepare the bacterial solution a KWIK
STIK containing Bacillus Megaterium was placed into a solution of about 100 ml
containing nutrient broth, urea, calcium chloride, sodium bicarbonate and ammonium
chloride. This was then placed in an incubator at a temperature of 30 C for 24 hours.
Before mixing with the sand, a sample of the solution was placed in a
spectrophotometer to confirm that the bacterial concentration was greater than 9 x 109
CFU/mL. When the bacterial solution was ready, sand, urea powder and calcium
chloride powder were thoroughly mixed together. The mass of the urea used varied in
the range of 5% to 20% of the dry weight of the sand, and the amounts of urea and
calcium chloride were always in a 1:1 molar ratio. The highly enriched bacterial
solution, and some extra water to make the total weight of water about 10% of the dry
ingredients, was then mixed thoroughly into the sand and nutrients for about a minute
before being placed into the mould in five layers lightly tamping between each layer.
Similarly to the gypsum specimens the shear wave velocity was then measured
continuously for 24 hours during curing after which the split mould was removed.
After curing UCS specimens were wrapped in plastic film and left to cure for 7
days before subjecting them to UCS tests. They were loaded to failure in typically 15
minutes. After failure the dry mass was recorded and the soil placed in hydrochloric
acid to determine the amount of calcite precipitated. The gypsum cemented triaxial
specimens were saturated by flushing with water and then using an elevated back
pressure of 500 kPa. For the bio-cemented specimens flushing resulted in some calcite
precipitation in the drainage tubes and to avoid this saturation was achieved by only
pumping water into the specimens. Following this conventional CID triaxial tests were
performed with continuous monitoring of the shear wave velocity. The shear wave
velocity was calculated using the instantaneous sample length and the travel time,
which was estimated by the cross correlation between input and output waveforms [1].
After the triaxial tests the amount of calcite was determined using the same method as
for the UCS tests.
3. Results and Discussion
3.1. Cementation
Figure 1 shows typical variations of shear wave velocity during curing for specimens of
sand prepared with different amounts of gypsum and bio-cementing materials. For the
gypsum cement there is a rapid increase in the shear wave velocity for the first hour,
after which the shear wave velocity is essentially constant. This is consistent with the
setting time of gypsum, which is reported by the supplier as 55 minutes, and is
attributed to hydration. As the amount of gypsum is increased, there is a trend for
higher values of shear wave velocity to be recorded. A similar trend is also observed in
the bio-cemented samples as the nutrient (urea and calcium chloride) content is
increased. However, there is a significant difference in the rate of cementation as
indicated by the increase of shear wave velocity. The bio-cemented samples required
approximately 30,000 seconds (8 hours) to achieve 90% of the velocity increase, after
which the shear wave velocity increased slowly for a further 24 hours.
800
700
600
Shear Wave Velocity (m/s)
500
400
300
2.5% Calcite (5% Urea)
4.8% Calcite (10% Urea)
200 5.8% Calcite (15% Urea)
5% Gypsum
10% Gypsum
100
15% Gypsum
0
0 10000 20000 30000 40000 50000 60000 70000 80000 90000
Time (sec)
Figure 1. Variation of shear wave velocity during curing.
Figure 1also shows the shear wave velocity of the loose uncemented sand mixtures
is approximately 100 m/s before curing commences. This corresponds to a stress level
of 2-3 kPa, which is provided by the weight of the top platen. Other tests on
uncemented sand indicated a shear wave velocity range of 120 m/s to 160 m/s at a
mean effective stress of 10 kPa. Significant increases in the shear wave velocity occur
as a result of the cementation process with values reaching 700 m/s with 15% gypsum,
and 650 m/s with the precipitation of 5.8% calcite (by dry mass), which occurred when
15% urea was added. Figure 1 shows the increase in the shear wave velocity depends
on the amount of gypsum or the amount of calcite precipitated, as expected. It has also
been found that the amount of calcite produced is approximately linearly related to the
amount of nutrients added [7].
The values of shear wave velocity, measured after curing, for all the triaxial
specimens are shown in Figure 2. This shows a fairly consistent relation between the
amount of cement and the resulting shear wave velocity, and that calcite is a much
more effective cementing agent than gypsum. The greater effectiveness of the calcite
cement is consistent with observations made by other studies [2, 4, 11]. These have
also shown the calcite tends to be precipitated preferentially at the points of contact,
whereas the particulate nature of the gypsum means that a significant amount is present
in the void spaces where it may not contribute to the bonding of the sand particles. It
should also be noted that increases in the amount of cement are associated with
increases in dry unit weight [7], which have been shown in many studies [5, 9] to lead
to increases in strength and stiffness at constant cement content.
Another significant difference between the current study and most previous studies
of bio-cementation is that specimens were prepared by mixing and cured with no
further addition of nutrients or bacteria. In contrast, most prior studies have produced
cementation by pumping solutions of nutrients and bacteria through uncemented sand.
1200 2000
Calcite
1800
Gypsum
1000
1600
1400
800
Shear Wave Velocity (m/s)
1200
Gmax (MPa)
600 1000
800
400
600
Calcite
400
200
Gypsum
200
Al Qabany et al [2]
0 0
0 4 8 12 16 0 4 8 12 16
% Calcite or Gypsum % Calcite or Gypsum
Figure 2. Variation of (a) shear wave velocity and (b) Gmax after curing with % cementing agent.
The results of one of these studies [2] are included on Figure 2a for comparison.
This shows that there is no significant difference between the two data sets, in this
study produced by mixing, and in the earlier studies by pumping in solutions of
nutrients and bacteria. The data from [2] were limited to calcite contents of 4.5%,
although higher calcite contents and stiffnesses have been reported in other pumping
studies. Our results suggest that further increases in calcite content deviate from the
linear relation suggested by [2]. This is believed to be a consequence of additional
calcite filling the pore space and not being as effective in cementing the grains.
Figure 2a shows that the rate of increase of shear wave velocity reduces as the
cement content increases. However, if the results are presented as shear modulus, G max
versus cement content an approximately linear relationship is obtained as shown in
Figure 2b. Similar linear relationships have been observed for other artificially
cemented sands. From an earlier study of gypsum cemented carbonate sand [14] it was
found that the shear modulus, Gmax could be predicted using Equations 1 and 2, where
the contribution to Gmax of the cementation is provided by equation 2.
Gmax
G
G
max max (1)

p p p
r cemented r uncemented r *
0.01 0.02
Gmax
1.08 p (
q
p A GC e 2.42 1

(2)
p p(
r * r
In these equations pr is a reference stress taken as 1 kPa, GC is the cement content as a
percentage of the dry mass, p( is mean effective stress, e void ratio, q deviator stress,
and A is a dimensionless constant
The value of the constant A for the gypsum cemented carbonate sand was found to
be 16407 [14]. To obtain reasonable fits for the data in Figure 2b, the value of A would
need to be two and six times bigger for the gypsum cemented quartz sand and calcite
cemented quartz sand respectively. These results indicate both the importance of the
sand particles and the greater effectiveness of the calcite cement.
3.2. Shearing
Typical stress, strain responses of the bio-cemented specimens from drained triaxial
tests conducted with an effective confining stress of 50 kPa are shown in Figure 3. The
results are consistent with studies on a wide range of artificially cemented soils. They
show a significant increase in strength for the cemented samples, increasing brittleness,
a suppression of dilation until after the peak followed by much greater rates of dilation.
It can again be seen that the calcite is more effective than gypsum, with the strength of
calcite cemented specimens about double that using gypsum for a similar mass of
cementing agent. Figure 3a also shows that the gypsum cemented samples are stiffer
than the bio-cemented specimens, which is the reverse of the trend in the stiffness at
very small strains shown in Figure 2b. This is believed to be related to the different
modes of cementation created by calcite (crystals nucleating on particle surface) and
gypsum (gypsum particles filling the void space).
Figure 4 shows the influence of the amount of cement and confining stress on the
peak strength. This figure shows the significant effect of the calcite cement on the
strength at low effective confining stresses and also that small amounts of cement still
have an effect on the strength at confining stresses up to 500 kPa. The figure also
shows comparison with UCS test data reported [3] for specimens produced by flushing
bacteria and nutrient solutions through a similar sand with similar relative densities.
These data lie slightly below the current study even though similar shear wave
velocities have been reported as shown in Figure 2. Previous studies of bio-cementation
have suggested there is a lower limit of about 2% calcite below which the cement is not
effective. However, this is not evident in this study where thorough mixing of soil and
cement has occurred. This suggests that inhomogeneity in cementation occurring with
the flushing technique is responsible for both low apparent strengths with weak
cementation and the scatter in the data reported in previous studies [12, 16].
The variation of Gmax as estimated from the shear wave velocity for triaxial
2000
Uncemented
-0.050
Uncemented
2.79% Calcite 2.79% Calcite
4.4% Calcite
4.4% Calcite
1600 -0.040 7.3% Calcite
7.3% Calcite
Deviator stress (kPa)
5% Gypsum
5% Gypsum
15% Gypsum
15% Gypsum -0.030
1200
Volume strain
-0.020
800
-0.010
400
0 0.02 0.04 0.06

0.000
Axial strain
0
0.00 0.02 0.04 0.06
Axial strain 0.010
Figure 3. Typical deviator stress (a), and volume strain (b) responses from triaxial tests ((3 = 50 kPa)
2500
Unconfined Compression
UCS Al Qabany & Soga (2013)
2000 Effective Confining Stress = 50 kPa
Effective Confining Stress = 200 kPa
Peak Deviator Stress (kPa)
Effective Confining Stress = 500 kPa

1500 UCS Gypsum
1000
500
0
0 2 4 6 8 10 12 14 16
Calcite or Gypsum (%)
Figure 4. Influence of calcite content on peak strength for different confining stresses.
specimens subjected to an effective confining stress of 500 kPa and then sheared
drained to failure is shown in Figure 5. The figure also shows the variation of G max with
mean effective stress during compression for an uncemented sand specimen. As p
increases to 500 kPa the shear modulus of all cemented specimens increased at a rate
only slightly less than the uncemented sand. However, the effect of the cement is
significant with Gmax for 4% calcite approximately three times that of the uncemented
sand. Increases in p beyond 500 kPa in Figure 5 are associated with shearing. It can be
seen that initially the stiffness continues to increase, but with shearing the shear
modulus gradually reduces until it reaches close to the value of the uncemented sand at
the peak. For this relatively high confining stress it can be seen that the decrease in
800
700
600
500
Gmax (MPa)
400
300
200 Uncemented
1.88% calcite
3.02% calcite
100
3.26% calcite
4.16% calcite
0
0 200 400 600 800 1000 1200
p' (kPa)
Figure 5 Variation of Gmax during compression and shear

Gmax towards the uncemented sand is gradual, and associated with a gradual breakage
of the cementation bonds. At lower confining stresses failure is more sudden and based
on values of Gmax there is still significant cementation at the peak.
4. Conclusions
Bio-cemented specimens have been produced by mixing sand, bacteria and nutrients
and these have been compared with specimens cemented with gypsum. It has been
found that mixing produces more uniform specimens than when bio-cement is created
by pumping and that the strength and stiffness do not appear to be greatly affected by
the method of sample preparation. Preparation by mixing is recommended to
investigate the response of weakly cemented material.
As noted in several studies calcite is a highly effective cementing agent, and for a
given amount of cement gives higher strength and stiffness than other cementing agents.
Use of automated shear wave velocity measurement has enabled variations in
stiffness, and hence degree of cementation, to be monitored throughout the processes of
curing, stress application and shearing.
References
[1] Airey D.W and Mohsin A.K.M. (2013). "Evaluation of Shear Wave Velocity from Bender Elements
Using Cross-Correlation." Geotech Testing J, 36(4), 506-514.
[2] Al Qabany A, Mortensen B.M, Martinez B.C, Soga K and DeJong J.T, (2011) Microbial Carbonate
Precipitation: Correlation of S-Wave Velocity with Calcite Precipitation, Proc, Geofrontiers 2011,
ASCE GSP 211, 3993-4001.
[3] Al Qabany, A., and Soga, K. (2013). "Effect of chemical treatment used in MICP on engineering
properties of cemented soils." Geotechnique, 63(4), 331-339.
[4] Cheng, L., Cord-Ruwisch, R., and Shahin, M. A. (2013). "Cementation of sand soil by microbially
induced calcite precipitation at various degrees of saturation." Canadian Geotechnical Journal, 50(1),
81-90.
[5] Consoli, NC (2009) "Fundamental Parameters for the Stiffness and Strength Control of Artificially
Cemented Sand," J. Geotech. Geoenviron. Eng., 135, 1347.
[6] DeJong, J.T, Mortensen, B.M, Martinez, B.C, and Nelson, D.C, (2010). "Bio-mediated soil
improvement." Ecological Eng, 36(1), 197-210.
[7] Duraisamy Y, Airey DW and Maggi F (2014) Urea Hydrolysis Rate of B. Megaterium for Typical Soil
Bio-cementation Application, Proc 7th Int Conf Environmental Geotechnics, 1143-1150.
[8] Hammes, F., and Verstraete, W. (2002). "Key roles of pH and calcium metabolism in microbial
carbonate precipitation." Reviews in Environmental Science and Biotechnology, 1(1), 3-7.
[9] Huang, J. T., and Airey, D. W. (1998). "Properties of artificially cemented carbonate sand." Journal of
Geotechnical and Geoenvironmental Engineering, ASCE, 124(6), 492-499.
[10] Ismail, M.A, Joer, H. A., Randolph, M. F. and Meritt, A. (2002). "Cementation of porous materials
using calcite." Geotechnique, 52(5), 313-324.
[11] Ismail, M.A, Joer, H. A., Sim, W.H. and Randolph, M. F. (2002) Effect of Cement Type on Shear
Behavior of Cemented Calcareous Soil, J. of Geotech. Geoenviron. Eng., ASCE, 128(6), 520-529
[12] Kucharski, E., Price, G., Li, H., and Joer, H. A. (1996). "Laboratory evaluation of CIPS cemented
calcareous and silica sands." Proc., 7th Australia New Zealand Conference on Geomechanics, 102-107.
[13] Mitchell J.K and Soga K (2005) Fundamentals of Soil Behaviour, Wiley.
[14] Mohsin, A., and Airey, D. W. (2008). "Using Gmax measurement to monitor degradation of an
artificially cemented sand." Proc. 4th Int. Symp Deformation Characteristics of Geomaterials, 305-310.
[15] Mortensen B.M, Haber M.J, DeJong J.T, Caslake L.F and Nelson D.C (2011),"Effects of environmental
factors on microbial induced calcium carbonate precipitation," J. of Applied Microbiology, 111,338-349.
[16] Whiffin V.S, van Paassen L.A and Harkes M.P, (2007) "Microbial Carbonate Precipitation as a Soil
Improvement Technique," Geomicrobiology Journal, 24, 5, 417-423.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-291
Disk Transducer for stiffness measurement

on granular materials
Laxmi P. SUWALa, 1 and Reiko KUWANO b
a
ARC Center of Excellence for Geotechnical Science and Engineering, University of
Newcastle, Australia
b
Institute of Industrial Science, The University of Tokyo, Japan
Abstract. Small-strain stiffness is one of the prominent characteristics of geo-

materials on analysis of deformation behavior. Elastic wave measurement
technique is becoming stronger non-destructive tool than other technique. In this
paper, authors have focused on the application of the flat surfaced transducer 'Disk
transducer', enabling to measure both compression and shear wave on triaxial
specimen. Disk transducer was built the University of Tokyo using piezoceramic
element and encapsulated in the top cap and pedestal of small scaled triaxial
apparatus then performed the tests. This paper presents the results of three sorts of
granular materials, fine, medium and coarse sand, were investigated statically and
dynamically at isotropic and anisotropic stress states in the laboratory. The elastic
parameters obtained on the granular materials; Young`s modulus (E) and shear
modulus (G), obtained by statically and by means of disk transducer method were
examined and compared. It was observed, the statically and dynamically obtained
results are mutually comparable and in agreement with previous studies. It is also
confirmed the elastic parameters are to be a unique function of stress states with
experimental evidence.
Keywords. compression wave, shear wave, small strain stiffness, Poissons ratio,
laboratory test
1. Introduction
Elastic waves associated with the elastic property of material and is quite informative
to reliable prediction of deformation behavior, since investigation of elastic properties
propagating elastic wave in the laboratory specimen is important. Numerous
transducers are being used in laboratory based geotechnical engineering practice;
Lawrence [1], [2] was pioneer of using the flat piezo-electric transducers in shear wave
measurement in soil specimens. Bimorph element (piezo ceramic element) based
transducer Bender element has gained the popularity amongst the researchers after its
investigation by Shirley and Hampton [3]. Flat surfaced transducer, Disk transducer',
enabling to measure both compression and shear wave on triaxial specimen built in the
University of Tokyo [4], has main additional benefit of its non-intrusive nature as
bender element has to be inserted in the specimen and not always suited to undisturbed
and cemented sample. Furthermore, protrusive nature of transducer increases the
bedding errors around the transducer too. In this paper, authors have focused on the
application of the flat surfaced transducer 'Disk transducer', made encapsulating piezo-
1
Corresponding Author
292 L.P. Suwal and R. Kuwano / Disk Transducer for Stiffness Measurement on Granular Materials
ceramic element in the top cap and pedestal of small scaled triaxial apparatus and
presents the results of three sorts of granular materials, fine, medium and coarse sands,
were investigated statically and dynamically at isotropic and anisotropic stress states in
the laboratory. Statically, axial and lateral deformation were locally evaluated by Local
Deformation Transducer (LDT) [5] and clip gauges. Small strain stiffness was
!!$\!!$!!\!
2. Transducer and Apparatus
2.1. Disk transducer and elastic wave measurement system
Disk transducer, developed in the University of Tokyo using piezoceramic elements
of 20 mm in diameter made by Lead Zinronate Titanate Ceramic [Pb(Ti.Zr)O3] as
described in [4], [6] was employed to measure elastic wave. Disk transducer was
placed at the top cap and the pedestal of the small scale triaxial apparatus. With aid to
elastic wave measurement system the signal was excited from transmitter and retrieved
in the receiver. Beside transducers, wave measurement system consists of chain
composed instruments. It includes three electrical devices; Signal generating device,
an amplifier and signal acquiring device (oscilloscope). A digital function generator,
which can produce a maximum peak to peak voltage of 10 V and is capable of
producing twelve kinds of different waveforms at frequency ranges of 0.001Hz to 25
MHz was utilized. An amplifier, with maximum 100 times amplification capacity, was
used to amplify the input signal generated by the function generator before feeding it
into the transducer. The multi-channel oscilloscope was used to record and display
waveforms of the transmitted and received signals.
2.2. Interpretation of elastic wave signals
The first deflection of the signal in time domain is dealt as the arrival of the
compression wave. The travel time of the propagated shear wave was determined as
recommended by Jovocic [7]. The zero baseline correction was imposed to minimize
the disturbances arose by the near field effect and reflected compression waves. The
typical waveforms indicating the point of arriving of the received wave are shown in
Figure 1.
3. Material and methodology
3.1. Material
Granular geo-materials ranging from fine to coarse sand; Toyoura sand, Silica sand and
Hime gravel were tested in this work. Toyoura sand is well-known sand that used in
Japan for research purposes. It fine, uniformly graded sand originating from the
Toyoura beach of Yamaguchi Prefecture, Japan. Silica sand is one of the varieties of
sand in the world. In this study, silica sand with mean diameter D50 of 0.45mm was
used. Hime gravel having mean diameter D50 of 1.72 mm is poorly sorted gravel with
angular grain shapes. It was originated from the Hime River, Nagano Prefecture, Japan.
L.P. Suwal and R. Kuwano / Disk Transducer for Stiffness Measurement on Granular Materials 293
The photographs of used materials are shown in Figure 2 and the physical properties
are given in Table 1.
Figure 1 Typical waveforms indicating the arrival Time
[a] [b] [c]
Figure 2. [a] Toyoura sand, [b] Silica sand, and [c] Hime gravel
Table 1 Physical properties of tested materials
Parameters Toyoura sand Silica sand Hime gravel
Specific gravity, Gs 2.62 2.64 2.65
Max. void ratio, emax 0.946 0.787 0.709
Min. void ratio, emin 0.637 0.538 0.480
Mean diameter, d50,mm 0.19 0.45 1.72
3.2. Specimen preparation
The tested materials were granular so the pluviation technique was adopted for
specimen preparation. Densities were maintained controlling pouring height. While
pouring the materials, the vertical distance between the top surface of the sand and the
tip of the funnel opening was kept constant and the pouring direction was changed
from clockwise to anti clockwise regularly .The specimens of 75mm in diameter and
height 150 mm were prepared at stress state of 25kPa in all cases and the unit weight of
tested specimens are tabulated in Table 2.
Table 2 Test list
3.3. Stress path
As illustrated in Figure 3 [I: isotropic; II and III: anisotropic], both isotropic and
anistropic stress path were employed in this study. As prior mentioned, the samples
were made at an isotropic stress state of 25 kPa. Then, the stress level was increased to
50, 100, 200 and 400 kPa respectively in isotropic cases. Two types of the anisotropic
loading sequences were employed. The first type of anisotropic stress path was
executed maintaining the constant stress ratio between the vertical and horizontal
stresses. The vertical and horizontal stresses were increased maintaining the stress ratio
vh) equal to 2. In another anisotropic stress path, t!h) was kept
!$!v) was raised to 50, 75, 100, 125 and 150 kPa.
11 cyclic loadings with peak to peak strain amplitude of 0.001% were applied in
vertical direction in each stress state.
3.4. Analysis
Small-strain cyclic loadings in vertical direction were employed in each stress state.
h = 0), the elastic parameters; Vertical Youngs modulus
(Ev) and Poissons ratio applying cyclic loading along vertical direction (vvh) are
determined as
Ev , v , v (1)
vh , h , v (2)
The elastic waves were propagated along the vertical direction therefore the elastic
parameters by wave velocity pertain;
Gvh - Vs2 (3)
M v - V p2 (4)
Elastic moduli values obtained on the specimen of various densities were
normalized by the function proposed by Hardin and Richart, [9]
( 2.17 e) 2
f (e) (5)
1 e
The constrained modulus, Youngs modulus and shear modulus are dealt with the
elasticity theory; based on stress-strain relationship following to past studies [9], [10].
4. Test results and Discussion
4.1. Waveforms
Each stage of loading sequences, elastic wave measurement was conducted. P and
S waveforms obtained on Toyoura sand specimen at isotropic stress states of 50 kPa to
400 kPa are plotted in Figure 4. Those waveforms are achieved exciting single
sinusoidal wave of 25 kHz through transmitter in Top cap. In these plots, the input
signal is presented in volt and received signals are in millivolt. Visible cross-talk
signals are achieved in both P and S waveforms. In both P and S waves, the received
signals possess the higher amplitude in the higher stress states than lower stress states.
The arrival of signal was estimated as shown in the corresponding plot. The arrival
point of the signals on those plots is visually picked. While estimating arrival point of
signal, the first deflection is assumed as arrival point in P waves and disturbances due
to possible noises, reflected P waves, so called near-filed effects are considered in S
waves. Travelling time was calculated considering the time gap between rise of input
signal to rise of the received signal in each case and the velocities were calculated.
In same way, the waveforms obtained on the other specimens including silica sand
and Hime gravel specimen in both isotropic and anisotropic stress states are dealt and
elastic parameters are calculated. The input signal having frequency of 25 kHz and 15
kHz was employed to evaluate the propagated wave velocity in Silica sand and Hime
gravel respectively.
4.2. Elastic moduli
As explained above the specimen was subjected to either isotropic or anisotropic

stress path. The elastic moduli, Youngs moduli and shear moduli are evaluated for
three sort of granular materials; Toyoura sand, silica sand and Hime gravel by static
and disk transducer method then normalized by the function (5) for each cases.
Figures 5, 6 and 7 are the youngs moduli obtained on the Toyoura sand, Silica sand
and Hime gravel respectively. In each plot, the results obtained by static method
applying cyclic loadings are represented by hollow symbols and results derived by disk
transducer method are depicted in filled symbols. The normalized elastic moduli are
plotted against the effective vertical stress. The tests were performed in dry and drained
condition so the total stress is equivalent to the effective stress. In all cases, stiffness
were obtained by wave measurement were found to be laid above the results by small-
strain cyclic loading. It can be seen in the figures; 5, 6 and 7, the Youngs moduli are
being fitted in a line with respect to vertical stress;
Ev A1. f (e).( v ) n
In aggregate, two fitted lines are drawn in each plot; one representing the results
obtained by static method and another line is representing the result of disk transducer
method. The obtained results show a good agreement with fitting lines which indicate
that the Youngs modulus can be represented as a function of normal stress state. In the
figures, the fitting parameters; A1, derived on static results are found to be lower than
the dynamic results where the value of n is found higher in static results and lower in
dynamic results.
Figure 5 Youngs moduli obtained on Toyoura sand Figure 6 Youngs moduli obtained on Silica sand
Figure 7 Youngs moduli obtained on Hime gravel Figure 8 Shear moduli obtained on Toyoura sand
Figure9 Shear moduli obtained on Silica sand Figure 10 Shear moduli obtained on Hime gravel
Similarly, the shear moduli by elastic wave measurement were computed from the
shear wave velocity measured by disk transducer. Like as the Young`s moduli, the
shear moduli were also normalized by same void ratio function. The normalized shear
moduli obtained in three sorts of granular materials; Toyoura sand, Silica sand and
Hime gravel were plotted against the stress parameter constisting of both vertical and
horizontal stresses in Figures 8, 9 and 10 respectively. Same as the Youngs moduli,
two fitting lines; one representing the shear moduli by static method and another
representing the shear moduli by elastic wave measurement with the disk transducer
method. The obtained results are fitted in the function;
n
Gv h A2 . f (e).( v . h ) 2
The results shows an agreement with fitting lines which reflect that the shear
moduli is the unique function of both vertical and horizontal stress states
The fitting parameters; A2, derived on static results are found to be lower than the
dynamic results where the value of n is found higher in static results and lower in
dynamic results. Those fitted lines reflect that the statically derived shear stiffness is
lower than the derived by the disk transducer method at lower stress states. At higher
stress states, the static results seem to be higher.
With regard to stiffness; Young`s modulus and shear modulus, analyzed by elastic
wave measurement were seen to be higher by around 10% to 20% as compared to
static results at stress states of 50 kPa. Another speculated matter is the apartness
between the statically and dynamically measured values. The higher apartness at the
lower stress states and merging trend at higher stresses were noticed. The causes of
being the lower values of static results at the lower stress states were thought the effect
of the heterogeneity as pointed out by Tastuoka and Shibuya [11], the elastic wave
tends to pass faster through the stiffer parts where average stiffness might be lesser as
explained in AnhDan and Koseki, [12].
5. Conclusion
The experimental investigation on the three sorts of granular materials; fine,

medium and coarse grained sand were executed on the cylindrical specimen of triaxal
test. The elastic parameters obtained on the granular materials; Young`s modulus (E)
and shear modulus (G), obtained by statically and by means of disk transducer method
were examined and compared. It was observed, the statically and dynamically obtained
results are mutually comparable. It is confirmed, regardless to loading sequences, the
stiffness is a unique function of stress with experimental evidence.
References
[1] F. V. Lawrence, Propagation of Ultrasonic Waves Through Sand`, Research report R63-8 (1963),
Massachusetts Institute of Technology, Cambridge, MA, March
[2] F. V. Lawrence, Ultrasonic Shear Wave Velocity in Sand and Clay, Research report R65-05, Soil
Publication No. 175, (1965), Massachusetts Institute of Technology, Cambridge, MA, January
[3] D. J. Shirley and L. D. Hampton, Shear-wave Measurements in Laboratory sediments, Journal of
Acoustical Society of America, 63 (2) (1978).
[4] L. P. Suwal, and R. Kuwano, Disk shaped piezo-ceramic transducer for P and S wave measurement in a
laboratory soil specimen, Soils and Foundations, 53(4) (2013), 510-524.
[5] S. Goto, F. Tatsuoka, S. Shibuya, Y.- S. Kim, T. Sato, A simple gauge for local small strain
measurements in the laboratory, Soils and Foundations, 31 (1) (1991), 169 180.
[6] L. P. Suwal and R. Kuwano, Statically and Dyanamically Measured Poissons Ratio of Granular Soils on
Triaxial Laboratory Specimens, Geotechnical testing Journal 36(4) (2013), 493- 505
[7] V. Jovicic, M. R. Coop, and M. Simic, Objective Criteria for Determining Gmax from Bender Element
Tests, Geotechnique, 46(22), (1996), 357-362.
[8] B. O. Hardin, and F. E. Richart, Elastic Wave Velocities in Granular Soils, Journal of Soil Mechanics
and Foundation, 89(1), (1963), 33 -65.
[9] E. Hoque, F. Tatsuoka, T. Sato, 1998, Measuring anisotropic elastic properties of sand using a large
triaxial specimen, Geotechnical Testing Journal, 19(4), (1998), 411-420. Paper ID: GTJ10718J,
DOI: 10.1520/GTJ10718J
[10] R. Kuwano, J. R. Jardine, On the applicability of cross-anisotropic elasticity to granular materials at
very small strains, Geotechnique, 52(10), (2002), 727 749.
[11] F. Tatsuoka, and S. Shibuya, Deformation Characteristics of Soils and Rocks from Field and
Laboratory Tests, Proceedings of 9th Asian Regional Conference of SMFE, Bangkok, 2 (1992), 101-
170.
[12] L. Q. AnhDan, J. Koseki, and T. Sato, Comparison of Young`s Moduli of Dense Sand and Gravel
Measured by Dynamic and Static Method`, Geotechnical Testing Journal, 25(4), (2002), 349 368.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-299
Experimental evaluation of liquefaction

resistance of unsaturated sandy soils
Hailong WANGa,1, Junichi KOSEKIa, Takeshi SATOb and Jaylord TAN TIAN a
a
The University of Tokyo, Tokyo, Japan
b
Integrated Geotechnology Institute Ltd., Tokyo, Japan
Abstract: A series of undrained cyclic loading tests of three types of saturated and
unsaturated soils were conducted on a triaxial apparatus. To overcome the
difficulties of this type of test, cell pressure control system to maintain constant p,
membrane filter technique to measure pore water pressure and double cell system
to measure volumetric strain (v) were introduced to the apparatus. Test results
shown that the cell pressure control system reduced more than 60% variation of p
which would be induced by applying vertical cyclic loading. The membrane filter
technique could measure negative pore water pressure in a prompt manner and as a
result suction of the unsaturated specimens was well monitored under the 0.1 Hz
cyclic loading. The time delay problem of electric transducer employed in the v
measurement system, which may bring significant system error, was recognized
and the measurement error of v was discussed. Regarding the resistance against
liquefaction of the unsaturated soils, the test results shown increase in resistance
with reduction in degree of saturation and the p-constant condition was necessary
to correctly evaluate the resistance for the unsaturated soils.
Keywords: Unsaturated soils, liquefaction resistance, undrained cyclic loading

test, membrane filter technique, cell pressure control system, double cell
1. Introduction
Liquefaction problem has long been a big concern for countries, such as Japan,
suffering from many earthquakes. Severe damages were caused by liquefaction in
kanto areas during the Great East Japan Earthquake Disaster in 2011 [1]. Studies have
been launched for the countermeasures of liquefaction that can be applied to the
residential areas with low cost. The techniques of ground desaturation by either
dewatering the ground or injecting the micro air bobbles have been paid attention in
recent years. Experimental tests and field applications in Japan [2-3] show that these
methods can be cheap and efficient ways to increase the resistance against liquefaction
(RL) of the ground. On the other hand, however, only few studies can be found to
investigate the RL of the unsaturated soils qualitatively through the laboratory tests
probably because of the technical difficulties of the test. In this study, some new testing
techniques were introduced to a triaxial apparatus and a series of undrained cyclic
loading tests on both saturated and unsaturated specimens were conducted.
1
Institute of Industrial Science, The University of Tokyo, 4-6-1 Komaba, Meguro-ku, Tokyo 153-8505,
Japan. Email: [email protected]
300 H. Wang et al. / Experimental Evaluation of Liquefaction Resistance
Fig. 1 Illustration of the apparatus
2. Apparatus
Fig. 1 schematically shows the configuration of the apparatus for the unsaturated
specimens. It has the following three main features:
1. Cell pressure (h) control system [4] to maintain total mean principle stress (p)
constant (p-constant condition) as shown in Fig. 1. This system can adjust h
simultaneously following the vertical cyclic loading applied by the vertical stress
control system. The magnitude of increment of h (,h) was set as one third of deviator
stress (q) as shown schematically in Fig. 2, through which p can be kept constant.
According to the concepts of pore pressure coefficients A and B [5], the increment
of pore water pressure (,uw) of the specimens under the undrained condition can be
expressed as:

(Eq. 1)

If we define increment of effective mean principle stress (,p) as:

(Eq. 2)

it is clear that ,p will be rarely affected by ,p for the saturated specimens which are
normally of B values of 0.95. However, it is not valid for the unsaturated specimens,
of which B values may reduce sharply with slight reduction in degree of saturation (Sr),
e.g., it was 0.22 for a Toyoura sand specimen with Sr of 99% [6]. To remove such an
effect, the p-constant condition was applied to the unsaturated specimens in this study.
2. Suction measurement system. It may become difficult to measure the pore water
pressure (uw) of the unsaturated specimens under relatively high loading frequency (0.1
Hz with stress control in this study) by using the ceramic disk [7]. Thus, the membrane
filter technique [8] was introduced to measure uw. The thin membrane filter was fixed by
a stainless cover with six screws as shown the details in Fig. 1. Pore air pressure (ua)
was measured by a pressure transducer being connected with the drainage path of the
H. Wang et al. / Experimental Evaluation of Liquefaction Resistance 301
top cap. A hydrophobic filter was glued on the surface of the top cap to reduce the
invasion of pore water during the tests. Suction (S) was calculated by S=ua-uw.
3. Volumetric strain (v) measurement system. The double cell system with the
same concept as that in reference [ 9 ] was used to measure v of the unsaturated
specimens. The equation of v is:

(Eq. 3)

where, ,Vspecimen: volume change of the specimen (set compression as positive);

,VLVDT: volume change of the top cap submerged by water in the inner cell; ,VDPT:
total volume change under the water level in the inner cell; V0: initial volume of the
specimen.

,

,
,

Fig. 2 Principle of cell pressure control Fig. 3 Gradations of tested sands
3. Testing program
In the testing program, three types of sands having the gradation curves as shown in
Fig. 3 were used. Toyoura sand is a clean sand with fines content (Fc) of less than 0.5%.
Inagi sand and iron ore fines type B (IOF-B) possess Fc of 30% and 24%, respectively.
Iron ore fines (IOF) containing 10% or more of particles less than 1mm and 50% or
more particles less than 10mm [10], is processed from the iron ore. The liquefaction
property of IOF raised people attention because of two accidents in 2009 [11-12].
All the specimens are of 50mm in diameter and 100mm in height. The test
conditions are described in Table 1. Both saturated and unsaturated specimens were
tested. For Toyoura sand, the saturated specimens were made by air pluviation method
and saturated by flushing CO2 and de-aired water; the unsaturated specimens were
reconstituted by moisture tamping and then flushed by water. For Inagi sand and IOF-B,
the specimens were molded by one dimensional compression and saturated by the
double vacuuming method[13]. For the unsaturated specimens of Inagi sand, extra water
was added from the top of the specimens and curing was treated for approximately 10
hours for water infiltration, while water flushing was applied to the unsaturated IOF-B.
Table 1 Test conditions
Sands type Properties Density and 0

Toyoura sand emax: 0.989; emin: 0.611; Gs: 2.652 Dr: 64-70%; 0:60kPa
Inagi sand emax: 1.650; emin: 0.910; Gs: 2.656 Dr: 64-78%; 0:60kPa
IOF-B .dmax:2.79 g/cm ; Gs: 4.444
3
Dc: 93%; 0:100kPa
emax, emin, Gs, Dr, 0,.dmax and Dc are maximum void ratio, minimum void ratio, specific gravity,
*
relative density, initial confining pressure before cyclic loading, maximum dry density and compaction
degree (=.dtest/.dmax100%).
,
,
,
,

,
,

Fig. 4 Typical measurements of q/3 and ,h Fig. 5 Normalized relationship between average
during cyclic loading amplitudes of q/3 and ,p during cyclic loading
4. Test results
4.1. Measurements of p for the unsaturated specimens with the p-constant condition
Fig. 4 describes typical time histories of q/3, -,h and ,p/0 of tests with the p-
constant condition during undrained cyclic loading. It shows that there are some
differences between q/3 and -,h in the magnitude and in the time axis, and
consequently variation in ,p/0 was observed. The average amplitudes of ,p and q/3
normalized by 0 of all the tests in this study were summarized in Fig. 5. It can be seen
that the h control system can remove more than 60% of variation of p which would be
induced by cyclic vertical loading (q).
4.2. Suction measurement
4.2.1. Suction before cyclic loading

Fig. 6 shows the typical time histories of uw, ua and suction of the unsaturated specimen
of Inagi sand (Sr=74%) before applying cyclic loading. uw was measured as soon as the
specimen was placed on the pedestal. It can be seen that the measurement of u w
stabilized in a very short period (for some specimens with higher Sr, it took less than
1min to get stabilized measurement). The pore air drainage was only allowed during
the consolidation process to maintain the desired Sr and thus ua kept almost zero. It
implies that suction may not be affected significantly by consolidation for Inagi sand.
The suction values of the unsaturated Inagi sand specimens measured on the
triaxial apparatus were compared with soil water characteristic curve (SWCC) of Inagi
sand in Fig. 7. It can be seen that the measurements are consistent with the wetting
curve of SWCC, which may be caused by the process of the specimen preparation
(illustrated in Section 3). Together with Fig. 6, the test results imply that the suction
measurement system performed very well under the static condition.

uw

ua

Fig. 6 Measurements of uw, ua and suction Fig. 7 Comparison of suction measured on the
before cyclic loading for Inagi sand (Sr=74%) triaxial apparatus with SWCC of Inagi sand

uw uw

ua ua

Fig. 8 Measurements of uw, ua and suction Fig. 9 Measurements of uw, ua and suction
during cyclic loading for Inagi sand with the during cyclic loading for Inagi sand without the
p-constant condition p-constant condition
4.2.2. Suction during cyclic loading

Figs. 8 and 9 show the typical time histories of pore pressure measurements of two
Inagi sand specimens with and without the p-constant condition, respectively. The uw
and ua increase gradually with the similar patterns for both cases, and the suction
values keep almost constant until the peak values of ua exceed about half of 0, after
which suction start to fluctuate increasingly and drop to negative side transiently. Since
the deformation of the specimens was rather small in the early stage of the tests,
suction may not be affected prominently by loading, while, as the increase in ua and
deformation of the specimens, suction may vanish when, for instance, the pore air
network was separated into air bubbles. Regarding the p-constant condition, it shows
that in the early stage of the tests, the pore pressure curves of the specimen with the p-
constant condition are smoother (with less fluctuation) than that without the p-constant
condition, which implies that the control system efficiently reduced the effect of ,p.
Due to the page limitation, the characters of effective stress paths of specimens with the
p-constant condition will be presented elsewhere.
4.3. Volumetric strain measurement
Before showing the measurements of v, the time delay issue of the differential pressure
transducer (DPT), measuring ,VDPT in Eq. 3, has to be explained. The configuration of
a series of trial tests is illustrated in Fig. 10. A spring was placed between the top cap
and the pedestal, the water level in the inner cell was monitored by a pressure
transducer (PT) and the DPT, and the vertical displacement was measured by the
LVDT. Fig. 10 also shows the time histories of the three transducers (with the unit in
voltage) while applying 0.1 Hz sinusoidal loading to the spring. It can be seen that the
responses of the LVDT and the PT are consistent with each other in the time axis,
while, there is a clear delay in the time axis of the DPT (red line). The response of the
DPT could well match the other two transducers if it was shifted in the time axis
slightly by 0.4s (green line). The square loading was also applied to the spring to
simulate the characters of axial strain when the specimens approach liquefied condition
as shown in Fig. 11. It shows that except the time delay, the increase/decrease rate of
the DPT response is also slower than those by the other two transducers. As a result,
the response could not well match the others even after shifting the time axis of the
DPT (green line).

Fig. 10 Responses of transducers under the Fig. 11 Responses of transducers under the
sinusoidal loading square loading

Fig. 12 Measurement of v of the saturated Fig. 13 Measurement of v of the unsaturated
specimen of Inagi sand specimen of Inagi sand
The time delay issue of the DPT was confirmed by many trial tests applied on two
types of DPTs. It seems that slightly shifting the time axis of DPT is necessary to
reduce the system error, while it may still cause significant error even after shifting
when the specimens experienced large vertical deformation.
Fig. 12 shows the time histories of v and axial stain of the saturated Inagi sand
specimen. It can be seen that the error of v measurement system is within 0.5%
(rather reasonable) before the axial stain exceeds around 7.5%. The v varies
significantly thereafter following the large axial strain. Fig. 12 implies that the current
measurement system can measure v reasonably when the axial strain of the specimens
is relatively small, while it is difficult to obtain acceptable results when the amplitude
of axial strain becomes larger (when the specimen is near to the liquefied state).
Fig. 13 shows the typical measurement of v for the unsaturated Inagi sand
specimens without the p-constant condition. For comparison purpose, the calculated v
by using Boyles law and ua measurement was also plotted. It can be seen the general
trend of two curves are similar, while the calculated v is obviously smaller that the
measured v. The maximum difference between two curves in each cycle up to 400s
(the time point when axial strain reached -5%) is less than 1.5%. As compared with [14],
1.5% difference may still be regarded as acceptable for this type of test. For the
measurement after 400s, it shows similar variation as that in Fig. 12, which is difficult
to use for the evaluation of v.

Fig. 14 Relationships between CSR and number of cycles for a) Toyoura sand, b) Inagi sand and c) IOF-B
4.4. Liquefaction resistance
The relationships between cyclic stress ratio (CSR) and number of cycles which cause
double amplitude axial stain of 5% (DA=5%) were plotted for the tested sands in Fig.
14. It shows that the resistance against cyclic loading (or liquefaction) of the
unsaturated specimens is higher than that of the saturated specimens for the same sand.
It is noted that the Inagi sand specimens with the p-constant condition resisted more
cycles than those without the p-constant condition. It is also shown that the liquefaction
resistance curves are not parallel with each other for IOF-B.
5. Conclusions
Some testing techniques and liquefaction resistance of the unsaturated sandy soils were
discussed in this paper. The following conclusions can be derived:
1. Cell pressure control system could reduce variation of p by more than 60% while
applying undrained vertical cyclic loading on the unsaturated specimens, through
which the effect of variation of p on the effective stress was well limited.
2. Suction measurement system could efficiently monitor uw, ua as well as suction
under both the static and cyclic loading conditions. The measurements show that
unsaturated Inagi sand could maintain its suction until it experienced severe
deformation.
3. The v measurement system could measure v of the unsaturated specimens
subjected to 0.1 Hz cyclic loading with acceptable system error when the
deformation of the specimens was relatively small, however, the system needs to
be improved to measure v when the specimens approach liquefied condition.
4. Liquefaction resistance of the unsaturated specimens is higher than that of
saturated specimens under otherwise similar conditions and the resistance may
increase under the p-constant condition for the unsaturated specimens.
6. Reference
[1] Towhata, I., Maruyama, S., Kasuda, K., Koseki, J., Wakamatsu, K., Kiku, H., Kiyota, T., Yasuda, S.,
Taguchi, Y., Aoyama, S. and Hayashida, T. (2014). Liquefaction in the kanto region during the 2011
off the pacific coast of Tohoku earthquake. Soils and Foundations, 54(4): 859-873.
[2] NILIM (2011) Report of continuous monitoring of an air injection ground: regarding the method of the
micro-bubble water injection as a countermeasure of liquefaction of residential land (in Japanese).
http://www.nilim.go.jp/lab/jbg/takuti/ekijyoka/20110513sato.pdf
[3] NILIM (2013) Examination and investigation on the ground dewatering method applied on liquefied
areas (in Japanese). http://www.mlit.go.jp/common/000986855.pdf
[4] Tsukamoto, Y., Ishihara, K., Nakazawa, H., Kamada, K. and Huang, Y. (2002). Resistance of partly
saturated sand to liquefaction with reference to longitudinal and shear wave velocities. Soils and
Foundations, Vol. 42, No. 6, 93-104.
[5] Skempton, A. W. (1954). The pore-pressure coefficients A and B. Geotechnique, vol. 4, No. 4, 143-147.
[6] Yoshimi, Y., Tanaka, K. and Tokimatsu, K. (1989). Liquefaction resistance of a partially saturated sand.
Soils and Foundations, Vol. 29, No. 3, 157-162.
[7] Unno, T., Kazama, M., Uzuoka, R. and Sento, N. (2008). Liquefaction of unsaturated sand considering
the pore air pressure and volume compressibility of the soil particle skeleton. Soils and Foundations,
Vol. 48, No. 1, 87-99.
[8] Nishimura, T., Koseki, J., Fredlund, D.G. and Rahardjo, H. (2012). Microporous membrane technology
for measurement of soil-water characteristic curve. Geotechnical Testing Journal, Vol.35, 201-208.
[9] Ng, C.W.W., Zhan, L.T., Cui, Y.J. (2002). A new simple system for measuring volume changes in
unsaturated soils. Can. Geotech. J., Vol. 39, 757-764.
[10] DSC.1/Circ.71. (2013). Early implementation of draft amendments to the IMSBC Code related to the
carriage and testing of iron ore fines. International Maritime Organization. https://webaccounts.imo.org/.
[11] Isacson, C. (2010a). Re: India- safe shipment of iron ore fines from Indian ports. International Group
Member Circular No.16.
http://www.gard.no/ikbViewer/Content/10129118/Member%20Circular%2016%202010%20India%20
Safe%20shipment%20of%20iron%20ore%20fines%20from%20Indianports.pdf .
[12] Isacson, C. (2010b). Re: Indonesia and the Philippines Safe Carriage of Nickel Ore Cargoes.
International Group Member Circular No.23.
http://www.gard.no/ikbViewer/Content/12143419/Member%20Circular%2023%202010%20Indonesia
%20and%20the%20Philippines%20%BF%20Safe%20Carriage%20of%20Nickel%20Ore%20Cargoes.
pdf.
[13] Ampadu, S.K. and Tatsuoka, F. (1993) Effect of setting method on the behavior of clays in triaxial
compression from saturated to undrained shear. Soils and Foundations, Vol. 33, No. 2, 14-34..
[14] Kimoto, S., Oka, F., Fukutani, J., Yabuki, T. and Nakashima, K. (2011) Monotonic and cyclic behavior
of unsaturated sandy soil under drained and fully undrained conditions. Soils and Foundations, Vol. 51,
No. 4, 663-681.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-307
Development of Stacked-Ring Shear

Apparatus for Multiple Liquefactions Tests
Seto WAHYUDIa , Sato TAKESHI b, Junichi KOSEKIc, and Yudai AOYAGId,1
a
Formerly PhD student, Institute of Industrial Science, University of Tokyo
b
General Manger of Integrated Geotechnology Institute Ltd., Tokyo
c
Professor, Dept. of Civil Engineering, University of Tokyo
d,1
Master student, Dept. of Civil Engineering, University of Tokyo
Abstract. Multiple liquefactions phenomenon gained much attention in the

Geotechnical Engineering field after the 2011 Off the Pacific Coast of Tohoku
Earthquake in Japan and Christchurch Earthquakes in New Zealand. It was found
that not just liquefaction can appear twice, but also can appear multiple times at
the same sites. In addition, the damage caused by the multiple liquefactions were
often more severe than the damage caused by the single liquefaction. This issue
raises great concern on the re-occurrence of liquefaction in the future great
earthquakes.
So far, several pioneer works have been able to investigate soil behaviors
during re-liquefaction. However, most of those studies were limited mostly up to 2
stages of liquefaction due to the limitation of the apparatus itself. In this paper, a
newly developed apparatus, so-called stacked-ring shear apparatus, is introduced.
This apparatus aims to investigate the soil behaviors not just during 2 stages of
liquefaction, but also during multiple liquefactions, by applying cyclic shear
loading under constant volume condition.
The stacked-ring shear apparatus has a unique feature as compared to the
common type of ring shear apparatus. However, both of these types of apparatuses
still have common drawback, which is excessive friction between the soil particles
and the metal rings. In the current study, several attempts have been employed to
reduce the amount of excessive friction within the specimen. First attempt was
done by reducing the number of rings composing the stacked-ring from 31 rings to
11, 8, and 5 rings. The second attempt was done by layering the surface of the ring
using frictionless coating material. It is shown that reducing number of rings
(reducing specimens height) is more effective than the use of frictionless coating.
However, there exists a certain limit, at which the number of rings can be reduced.
Below this limit, fair comparison cannot be made since fewer stacked rings tend to
generate un-equal amounts of friction on each stage of multiple liquefactions.
Keywords. Stacked-ring, re-liquefaction, multiple liquefactions, and friction.
1. Introduction
Multiple-liquefaction phenomenon has been emerged to be a major concern among

researchers after the recent cases during Christchurch Earthquake in New Zealand
(2010 - 2011) and Great East Japan Earthquake (2011).
The occurrences of multiple liquefactions have been studied and documented in
several parts of the world. In Japan, Kuribayashi and Tatsuoka (1975) [8], Yasuda and
Tohno (1988) [16] and among others investigated the case histories of liquefaction.
Recently, Wakamatsu (2012) [14] found 90 cases of multiple liquefactions in 2011
Great East Japan Earthquake alone.
308 S. Wahyudi et al. / Development of Stacked-Ring Shear Apparatus
Several pioneer laboratory works have been conducted to investigate this

phenomenon. Among those are the studies conducted by Finn et al. (1970) [4], Seed et
al. (1977) [11], Ishihara and Okada (1978 and 1982) [6][7], Towhata and Ishihara
(1985) [12], Oda et al. (2011) [9], Yamada et al. (2010) [15]. These studies found that
soil behaviors subjected with multiple liquefactions were affected by several factors
such as level of strain amplitude, anisotropy, among others.
Most of those mentioned experimental studies were conducted only up to two-
stage of liquefaction (re-liquefaction) test due to the limitation of their apparatuses
(triaxial and torsional shear apparatuses). Null effective stress (v= 0 kPa) combined
with large shear deformation during full liquefaction state made it difficult for these
apparatus to maintain the shape of the specimen to remain constant. Chiaro (2010) [2]
showed the twisted specimen under very large shear deformation during single stage
liquefaction test using hollow cylinder torsional shear apparatus.
To solve most of these shortcomings in conducting multiple-liquefaction test,
Institute of Industrial Science University of Tokyo has developed a new apparatus,
so-called stacked-ring shear apparatus. Using this apparatus, a single specimen can be
subjected to virtually unlimited number of successive liquefactions while maintaining
the geometry of the specimen to remain constant. Therefore, the soil behaviors
subjected with multiple liquefactions can be revealed.
2. Testing Apparatus and Procedures
The outline of the stacked-rings shear apparatus is shown in Figure 1(a). The axial
stress is applied through a pneumatic system using bellofram cylinders with the
capacity of 1000 kPa (1.0 MPa), while the torque is controlled by direct motor system
with the maximum and minimum rotation speeds of 0.015 deg./min and 64.8 deg./min,
respectively. Initially, the apparatus is equipped with a single load cell located at the
top cap of the specimen. Later, an additional load cell was installed at the bottom cap of
the specimen. The capacity of axial force and torque on each load cell are 15 kN and 3
kN.m, respectively. The applied stresses in the specimen are solely controlled by the
top load cell, while the bottom load cell measures the amounts of transferred stresses.
An annular specimen is placed in between inner and outer parts of the stacked-ring
as shown in Figure 1(b). Initially, both parts of the stacked-ring are composed by 31
pieces of vertically stacked individual rings, having a thickness of 5 mm as shown in
Figure 1(c).
The inner and outer diameters of the specimen are 90 mm and 150 mm,
respectively, while the height of the specimen is 155 mm. There is no direct contact
between the rings. Six pieces of metal bearings with a thickness of 0.1 mm were
inserted in between the rings, so the side wall friction in circumferential direction
between the soil particles and rings can be reduced as minimum as possible. In addition,
the 0.1 mm gap is also small enough to ensure the sand particles with a mean diameter
larger than 0.1 mm (D50> 0.1mm) will not extrude during shearing. Each of the rings is
free to move individually in the circumferential direction, while it fixed in the vertical
direction.
Toyoura sand was used as the test material. Its particles have an angular or sub-
angular shape with the following physical properties: specific gravity, Gs= 2.656; mean
diameter, D50= 0.193mm; fines content, Fc= 0.1%; max. void ratio, emax= 0.992; min.
S. Wahyudi et al. / Development of Stacked-Ring Shear Apparatus 309
o, emin= 0.632.. The specimeens were prep

void ratio pared by pluv viation of air--dried sand
particles in
nto a mold thrrough air.
In thiis paper, two types of test were conduccted. The firstt one consistss of 5 tests
(Mf1 to Mf5
M tests) aim med to measu ure the side wall
w friction in n the stacked--ring shear
apparatus during multiple liquefactions. The seco ond one consists of four teests (ML1,
ML2, ML L3, and ML4 tests)
t that aimed to investig
gate the soil beehaviors durinng multiple
liquefactioons. All speciimens were prepared
p to haave initial relative density within the
range of about
a 51-54%%. Later, four single-stage liquefaction tests
t (SL1 to SL4 tests)
were cond ducted additionally with different
d initial relative deensities to coompare the
results on multiple-liqu
uefaction tests..
(b)
(
(a) (c)
Fig
gure 1. Overall picture of stacked--ring shear apparaatus
The procedures
p to conduct multiiple-liquefactiion test can bee seen in Figur
ure 2(a) and
Figure 2(b b). First, the specimen
s wass consolidated d one-dimensionally up to 2200 kPa of
vertical sttress (v.Top= 200
2 kPa) from m point A to point
p B. Thenn, each of the specimens
was subjeected with cyclicc shear stress
s of 25 kPa (v= 2 25 kPa) undeer constant
volume co ondition by locking the top cap position.
To innvestigate thee effects of strain amplitude on the beehaviors of saand during
multiple liquefactions, each of the sp pecimens wass sheared up to t a pre-fixed maximum
shear straiin double amp plitude (DA.maax) at state C. The Mf1 to Mf5
M tests weree subjected
with DA.mmax up to 10.0 0% on each liiquefaction sttage. On the other
o hand, thhe ML1 to
ML4 testss were subjectted with adiffeerent shear strrain amplitudee up to 2%, 5% %, 7%, and
10%, resppectively.
A staage of liquefa faction was completed by adding anoth her half cycle of shear
loading from state C to state C, wheere the shear sttrain () is equ ual to zero (Fiigure 2(b)).
States C and D in thiss study were set s to be the ending
e and th
he starting stattes of each
stage during multiple-liquefaction test, respectively. Next liquefaction stage was started
by re-consolidating the liquefied specimen into their original vertical stress (v) of 200
kPa at state D. The next subsequent liquefaction stages follow the same procedures as
the one described in the first liquefaction stage.
On each stage, the liquefaction resistance was evaluated by calculating the number
of cycles needed to induce the double amplitude of shear strain 2.0% (NDA(2.0%)),
40
Initial liquefaction stage
A-B: Initial consolidation
A B-C: Initial liquefaction DAmax
Average shear stress, avg: kPa

C-D: Re-consolidation 20
D-E: Re-liquefaction
C, C
Void ratio (e)
B 0
C'
D
E -20
C
Dr0= 51.6%
-40
-8 -4 0 4 8
Effective vertical stress (v)
Shear strain, : %
(a) (b)
Figure 2. Procedures to conduct multiple liquefaction test and schematic test results
Similar to the other types of ring shear apparatus, liquefaction test in stacked-ring
shear apparatus is conducted with dry specimen under constant volume condition (Finn
and Vaid, 1977 [5], Sadrekarimi and Olsen, 2011 [10]). It is based on the assumption
that the decrease in applied vertical stress (v) during shearing in constant volume test
is equal to the increase in shear-induced pore water pressure (u) that would occur in an
undrained saturated test. Bjerrum and Landva (1966) [1] proposed this assumption for
DSS test (Direct Simple Shear) while Dyvik et al. (1987) [3] confirmed it again later.
3. Test Results
3.1. Development of the Apparatus
Typical results on a stage of multiple-liquefaction test using 31 stacked-rings are

shown in Figure 3. For 200 kPa of vertical stress and 25 kPa of cyclic shear stress, the
number of NDA(2.0%) are too few. There was sudden decrease of effective stress after
several few cycles. This result may suggest that large amount of side wall friction is
generated between the stacked-ring and the soil particles.
To investigate such behavior, an additional load cell was installed at the bottom of
the stacked-ring to measure the difference of vertical stress at the top and bottom of the
specimen. Due to the side wall friction, the shear stress and vertical stress may not be
distributed uniformly throughout the specimen.
The difference of vertical stress at the top and bottom of the specimen is shown in
Figure 4(a). The corresponding friction angles in vertical and circumferential directions
during consolidation are shown in Figure 4(b). The soil-ring friction angle in the
vertical direction was about 22, while in the circumferential direction was 0-0.2. The
friction in the circumferential direction appears to be negligible. This happens due to

the free circumferential movements of each ring.
Non-uniform distribution of vertical stress could potentially underestimate the
liquefaction resistance of soil. In addition, liquefaction may not appear simultaneously,
but rather progressively from bottom to the upper part of the specimen. This is not the
case in the reality. Therefore, the stacked-ring shear apparatus needs to be modified in
order to reduce the extent of friction.
In the current study, the modification was conducted by the combination of two
means, which are reducing the number of rings and layering the rings with frictionless
coating. The so called Diamond-Like-Coating (DLC) material was used to coat the ring.
According to the specification provided by the manufacturer, the DLC has three-quarter
less friction as compared to the original stainless steel ring (DLC = 0.1).
20 20
st st
Initial liquefaction (1 stage) Initial liquefaction (1 stage)
10 10
Shear stress, : kPa
Shear stress, : kPa
0 0
-10 -10
Dr0= 55.5% Dr0= 55.5%

NL= 2.0 DAmax= 10% DAmax= 10%
-20 -20
-10 -8 -6 -4 -2 0 2 4 6 8 10 0 50 100 150 200 250
Shear strain, : % Effective vertical stress, v': kPa
(a) (b)
Figure 3. (a) Shear stress and shear strain relationship; (b) shear stress and vertical stress relationship
50
Virgin consolidation
Virgin consolidation
0.0
Soil-wall friction angle: degree
Volumetric strain, Vol. (%)
Top load cell 40

Bottom load cell
Vertical friction angle (z)
30 Circumferential friction angle ()
-0.1
20 z
Generated friction
-0.2 10
0
Dr0= 54.9%
-0.3 Dr0= 54.9%
0 50 100 150 200 250 -10
Vertical stress, v (kPa) 0 5 10 15 20 25
Elapsed time: min.
(a)
(b)
Figure 4. (a) Typical volumetric strain and vertical stress; (b) soil-wall friction angles during consolidation
The numbers of ring composing the 31 non-coated stacked rings were reduced
subsequently to 11, 8, and 5 rings (Mf1, Mf2, Mf3, and Mf4 tests). In addition, these
rings were also coated with DLC. Then, each of these types of stacked-ring was use to
investigate soil behaviors during multiple liquefactions, while at the same time
measuring the generated soil-ring friction.
The measurements of soil-ring friction of all types of stacked-ring are shown in

Figure 5. The 31 non-coated stacked-ring has the smallest vertical stress at the bottom
of the specimen on each liquefaction stage, followed by the 11, 8, and 5 coated stacked-
rings. Naturally, the shorter the stacked-ring, the smaller the generated soil-ring friction.
The impact of DLC coating can be seen in 11 non-coated and 11 coated stacked-
ring shear tests. The DLC coated stacked-ring shear apparatus has about 20 kPa more
than the non-coated one. In term of friction angle, the coated rings about 2-3 less than
the non-coated one. These results shows the use of DLC coating has smaller impact
than reducing the stacked-rings height.
Shorter stacked-ring generates less friction, but the amount of friction may not be
constant at each stage of liquefaction (e.g. 8 and 5 stacked-ring shear). This causes
another issue, where direct comparison of soil behaviors at each liquefaction stage
cannot be made. Finally, the 11 stacked-ring combined with DLC coating was used for
the current setting of the apparatus.
Percentage of stress lost in vertical direction, (%)

Vertical stress measured at bottom load cell,
200 100
5 coated rings
8 coated rings
11 coated rings
11 non-coated rings
80
150 31 non-coated rings
60
v.Bo (kPa)
100
40
50 5 coated rings
8 coated rings
20
11 coated rings
11 non-coated rings
31 non-coated rings
0 0
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
Liquefaction stage Liquefaction stage
(a) (b)
Figure 5. (a) Vertical stress measured at bottom load cell; (b) Soil-wall friction angle vertical direction
during multiple liquefaction tests
3.2. Soil Behaviors During Multiple Liquefactions
Typical results from 1st and 2nd stages of multiple liquefaction test using 11 stacked-
ring shear apparatus can be seen in Figure 6. The change in specimens density during
consolidation/re-consolidation as well as the number of cycle to cause 2% shear
deformation at each liquefaction stage were evaluated.
Relationships between the relative density and liquefaction resistance during
multiple-liquefaction tests are plotted in Figure 7. For the comparison purpose, the test
results from single stage liquefaction tests (SL1 to SL4) were overlapped. The
liquefaction resistance of specimen that had been sheared with DA.max of 2% in the
previous liquefaction stage was always higher than those of single liquefaction tests.
Specimen sheared with DA.max of 5% has smaller liquefaction resistance at the 2nd
liquefaction stage, then getting higher at the following stages. The specimens sheared
with DA.max of 7% and 10% always have smaller resistances than those of the single
liquefaction tests.
The reason for such behaviors might be closely related to the change of soils
fabric during liquefaction. Wahyudi (2014) [13] traced the movements of each ring
using image analysis technique to investigate the local deformation behaviors of
specimen during liquefaction. The results showed that the specimen sheared with
smaller strain amplitude has more uniform strain distribution at each liquefaction stage,
while the specimen sheared with larger strain amplitudes showed more non-uniform
strain distributions. Smaller deformation induces re-arrangement of soil particles
during cyclic loading, whereas larger deformation destroys the soils fabric.
40 40
st st
Initial liquefaction (1 stage) Initial liquefaction (1 stage)
20 20
Shear stress, : kPa
Shear stress, : kPa

0 0
Top load cell
Bottom load cell
-20 -20
Dr0= 51.6% Dr0= 51.6%

Top load cell
NL= 12.7 DAmax= 10% DAmax= 10%
Bottom load cell
-40 -40
-8 -6 -4 -2 0 2 4 6 8 0 50 100 150 200 250
Shear strain, : % Effective vertical stress, v': kPa
(a) (b)
Figure 6. (a) Shear stress and shear strain relationship; (b) Shear stress and vertical stress relationship on 11
stacked-ring shear apparatus during multiple-liquefaction test
1000
DAmax
Number of cycle to liqueacftion, NL
4 2%
5%
6 7%
3
10%
100 5
4 Virgin
2 liquefaction
3
curve
5
1
6
4
10 2
3
4
2 5
3
2 Liquefaction stage shown at each data point

1
50 60 70 80 90
Relative density, Dr: %
Figure 7. Change in relative density and liquefaction resistances during multiple liquefaction tests
4. Conclusions
The investigation of soil behaviors during multiple-liquefaction test using stacked-ring

shear apparatus has several observations:
1. Stacked-ring shear apparatus is one of the few apparatuses that are capable to
simulate multiple-liquefaction phenomenon. The apparatus is capable to maintain
the geometry of the specimen to remain constant during multiple-liquefaction test.
2. The apparatus has its own drawback, which is the excess of friction due to the
contact between the rings and soil particles.
- The side wall soil-ring friction in the stacked-ring shear was reduced by the
combination of two means, which are reducing the numbers of ring and
layering the rings with frictionless coating.
- Short stacked-ring may significantly reduce the excess of soil-ring friction.

However, the generated side wall friction may not be constant on each
liquefaction stage. Thus, fair comparison cannot be made. Finally, the 11
stacked-ring shear with DLC coating is used for the current setting of the
apparatus.
3. The soil behaviors during multiple liquefactions reveal several observations:
- The specimens liquefaction resistances during multiple-liquefaction test are
highly affected to the applied strain amplitude. Specimen sheared with larger
strain amplitudes has smaller resistances in the future liquefactions, while the
opposite impact happens if the specimen subjected with smaller amplitude.
- Soil that had undergone multiple liquefactions could have smaller liquefaction
resistance compared with the soil that was subjected with only single
liquefaction. The reason is due to the change in soils fabric to the soil during
liquefactions. Larger strain amplitude means larger changes in soils fabric on
each liquefaction stage.
References
[1] Bjerrum, L. and Landva, A. (1966): Direct simple shear tests on a Norwegian quick clay.
Geotechnique, 16(1), pp. 1-20.
[2] Chiaro, G. (2010) Deformation properties of sand with initial static shear in undrained cyclic torsional
shear tests and their modelling, Doctor of Engineering Dissertation, Department of Civil Engineering,
The University of Tokyo, Japan.
[3] Dyvik, R., Berre, T., Lacasse, S., and Raadim, B. (1987) Comparison of truly undrained and constant
volume cyclic simple shear tests, Geotechnique, 37(1), pp. 3-10
[4] Finn, W. D. L., Bransby, P. L., and Pickering, D. J. (1970) Effects of strain history on liquefaction of
sand. J. Soil Mech. and Found. Div., ASCE, 96(6), pp. 1917-1934.
[5] Finn, W. D. L. and Vaid, Y. P. (1977): Liquefaction potential from drained constant volume cyclic
simple shear test. Proc. Of the 6th World Conference on Earthquake Engineering, New Delhi, India.
[6] Ishihara, K. and Okada, S. (1978) Effects of stress history on cyclic behavior of sand. Soils and
Foundations, Tokyo, 18(4), pp. 31-35.
[7] Ishihara, K. and Okada, S. (1982) Effects of large pre-shearing on cyclic behavior of sand. Soils and
Foundations, Tokyo, 22(3), pp. 109-125.
[8] Kuribayashi, E. and Tatsuoka, F. (1975) Brief review of soil liquefaction during earthquakes in Japan.
Soils and Foundations, 15(4), pp. 81-92.
[9] Oda, M., Kawamoto, K., Suzuki, K, Fujimori, H., and Sato, M. (2001) Microstructural interpretation
on reliquefaction of saturated granular soils under cyclic loading. J. Geotech. and Geoenvironmental
Engrg, 127(5), pp. 416-423.
[10] Sadrekarimi, A. and Olsen, S. M. (2011) A new ring shear device to measure the large displacement
shearing behavior of sands, Geotech. Testing J., 32(3).
[11] Seed, H. B., Chan, C. K., and Mori, K. (1977) Influence of seismic history of liquefactions of sands, J.
Geotech. Engrg. Div., ASCE, 103(4), pp. 257 270.
[12] Towhata, I. and Ishihara, K. (1985) Undrained strength of sand undergoing cyclic rotation of principal
stress axes. Soils and Foundations, 25(2), pp. 135-147.
[13] Wahyudi, S. (2014) Cyclic simple shear tests using stacked-ring on multiple liquefactions properties of
sands, PhD Thesis, University of Tokyo Japan.
[14] Wakamatsu, K. (2012) Recurrence of liquefaction at the same site induced by 2011 Great East Japan
Earthquake Compared with Previous Earthquakes. Proc. of 15th Intl Conf. of Earthquake. Engrg.
[15] Yamada, S., Takamori, T., and Sato, K. (2010) Effects on reliquefaction resistance produced by
changes in isotropy during liquefaction. Soils and Foundations, 50(1), pp. 19-25.
[16] Yasuda, S. and Tohno, I (1988) Sites of reliquefaction caused by the 1983 Nihonkai-Chubu
earthquake. Soils and Foundations, Tokyo, 28(2), 1-34, 61-72.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-315
Effects of inherent anisotropy on

deformation and strength characteristics of
a reconstituted sand
Hirofumi TOYOTAa,1 and Susumu TAKADA a
a
Department of Civil and Environmental Engineering, Nagaoka University of
Technology, Japan
Abstract. In order to design and construct the underground structures rationally,

accurate elastic shear modulus of soils is an important parameter. The real ground
formed by sands and/or clay has usually anisotropic mechanical behaviour.
Although many researchers have discussed the anisotropy of strength, anisotropy
concerning deformation modulus under small strain level have not been deeply
understood. Anisotropy can be classified as an inherent anisotropy and an induced
anisotropy, and the former anisotropy is treated in this study. Specimens are made
using Toyoura sand with changing the angle of sedimentation. Local small strain
(LSS) tests, by which precise deformation modulus can be measured, and bender
element (BE) tests were simultaneously conducted in the same specimens to
examine the influences of inherent anisotropy on deformation modulus and strength
under triaxial state of stresses. The results show that drained shear strength decreases
with the angle of sedimentation direction, which is defined as the angle of the
depositional plane from the vertical. On the contrary, elastic shear modulus
increases with the sedimentation angle.
Keywords. anisotropy, bender element test, deformation modulus, oriented

structure, sand, shear modulus, shear strength, triaxial compression test
1. Introduction
Natural sand particles have not perfect round shape but flat shape. Hence, two types of
anisotropy in soils were first defined by Casagrande and Carillo [1] from a formative
perspective, even though both anisotropy have close relation each other. According to
their definition, inherent anisotropy is likely to be formed during sedimentation of soil
particles. Induced anisotropy can be created when stresses or strains are applied to soils.
Inherent anisotropy, which is sometimes referred to as fabric anisotropy, is mainly
treated in the study. It has been demonstrated that the inherent anisotropy related to the
depositional (sedimentation) plane affects both deformation and strength characteristics
of sands (e.g., [2], [3], [4], [5], and [6]). However, the accuracy of the measuring of small
strain was insufficient even now.
Anisotropy of initial shear modulus was recently estimated under different
sedimentation plain and anisotropic stress conditions (Jamiolkowsk et al., [7]; Lo Presti
1
Corresponding Author: Associate Professor, Department of Civil and Environmental Engineering,
Nagaoka University of Technology, 1603-1 Kamitomioka, Nagaoka, Niigata, Japan; E-mail:
[email protected].
316 H. Toyota and S. Takada / Effects of Inherent Anisotropy
and ONeil, [8]; Yamashita et al., [9]). However, in those studies, the shear modulus was
estimated in only limited directions, which are vertical and horizontal.
Therefore, in the study, the specimens having various angle of sedimentation plain
were made. Then the shear modulus of the specimen was measured using local strain
measurement and bender element methods to discuss the shear modulus obtained from
the different methods. Moreover, shear strength also measured in the same specimen to
examine in detail the mutual relationship between the shear modulus and the shear
strength.
2. Experiments
2.1. Soil and Specimen Preparation
Toyoura sand was used to investigate inherent anisotropic behaviour in this study.
Toyoura sand is Japanese standard sand whose mineralogy is almost entirely quartz. 50%
diameter on the grain size diagram D50 is 0.2 mm and the uniformity coefficient Uc is
1.48. The particles have an angular to sub-angular shape. A maximum void ratio, emax,
and a minimum void ratio, emin, are, respectively, 0.990 and 0.597.
A inclined container whose side walls can be change their positions as shown in the
Fig. 1 was made to easily prepare a specimen that have inclined depositional plane. The
specimen size was d=50 mm diameter and h=120 mm height. Toyoura sand was
horizontally deposited in the container through following two methods. The first is air
pluviation method through a sieve with a mesh size of 850 m. As a preliminary tests,
the relationship between falling height of sand (height of the sieve) and relative density
was examined using this method (Fig. 2). Although the large change of density appears
in the falling height between 10 cm and 20 cm, density keeps almost the same value in
the falling height larger than 20 cm. Therefore the falling height was set larger than 20
cm to make a specimen.
The second is a sticking method where deposited sand in each 3 cm depth was stuck
using a rod having 2 mm diameter to disturb the sand particle orientation (fabric). The
same procedure continued in ten layers.
Those dry sand in the container was gradually submerged. Then the water in the
sand sample was removed by vacuuming of about 20 kPa. The unsaturated sand block
made in this way can stand by itself without outer confining pressure. The specimen was
formed by cutting and carefully trimmed from the soil block. The definition of
depositional plane in the specimen is illustrated in Fig. 3.
2.2. Apparatus
Figure 4 shows the schematic diagram of triaxial apparatus for local small strain
measurement by proximity transducers whose capacity is 2 mm. A pair of the proximity
transducers was employed to measure the small axial strain. The position or the
transducers can be changed from the outside of the cell. On the other hand the radial
strain was estimated using the proximity transduce set on a device which holds the radial
direction of the specimen. The targets were glued and connected on the membrane after
saturation process of the specimen. The targets for axial strain measurement were fixed
H. Toyota and S. Takada / Effects of Inherent Anisotropy 317
over the central part of specimen about 80 mm. For the radial strain measurement, the
holding device was placed on the centre of the specimen as shown in Fig. 4.
Figure 1. Inclined container to make Figure 2. Effect of falling height on

specimens. relative density of sand.
Figure 3. Depositional plane. Figure 4. Local small strain measurement.
2.3. Testing Methods
The test procedure is outlined as follows:

The specimen that set on the triaxial apparatus was saturated by the vacuum
saturation procedure [10] and a back pressure of 200 kPa was applied to produce
a B-value greater than 0.97.
The specimen was consolidated isotropically under a confining pressure c of

150 kPa.
The shear wave velocity was measured using a bender element method, which
installed in the top cap and the pedestal.
The drained triaxial compression tests were performed under constant cell
pressure of 150 kPa with axial strain rates of 0.0025%/min up to axial strain of
0.1%. The reason of such small axial strain rate is to achieve the precise
measurement in the small strain level. Then the axial strain rate change to
0.025%/min until failure occurred.
3. Test Results
3.1. Shear Strength
Figures 5 (a) and (b) respectively show stress strain relationships and volumetric strain
shear strain relationships for the specimens prepared by the air pluviation method. The
shearing was finished at about the shear strain of 10% because of the interference of LSS
measuring devices. According to continuous dilation in volumetric strain (Fig. 5 (b)),
although the specimen did not reach the critical state, the peak strength was attained.
There are apparent effects of depositional angle on shear behaviour. When is larger,
the shear strength becomes smaller. Although the dilative behaviour is exhibited in Fig.
5 (b) because of very dense specimens, contractive tendency increases with an increase
of . Sand particles will deposit with their major axis reclined horizontally by the air
pluviation. It is clear that larger shear strength appears when sand particles recline
horizontally in the triaxial compression tests.
Figure 6 shows shear behaviour for the specimens prepared by the sticking method.
The results of srress strain relationships are a good agreement in each (Fig. 6 (a)).
The volumetric strain behaviour also has a small difference between different values of
(Fig. 6 (b)). Those results indicate that fabric of sand particles will well disturbed by
the sticking and have random orientation of the particles.
The density of sand has a large effect on its shear strength. Therefore, the density of
each specimen were carefully examined. Figure 7 depict the specimens densities
prepared by each depositional method. The relative densities are roughly distributed at
90% and 80% in the air pluviation and the sticking methods, respectively. However the
densities are approximately consistent in the same method independent of the angle of
depositional surface. Therefore the mechanical properties of sand can be compared in
each method without effect of density.
Relationships between shear strength and are depicted in Fig. 8. The largest
orientation effect, where shear strength reduced with an increase of , appears in the
specimens prepared by the air pluviation method. However, almost the same strengths
were obtained from the specimens prepared by the sticking method. Thus, the shear
strength of sand is easily influenced by the degree of orientation effects. Concerning the
value of the shear strength, the shear strength of the specimen prepared by the air
pluviation method is larger than that prepared by the sticking method because of bigger
density obtained from the air pluviation method (Fig. 7). Next, secant shear modulus at
small strain level was examined using those kinds of the specimens.
Figure 5. Effect of on shear Figure 6. Effect of on shear

behaviour in air pluviation method. behaviour in sticking method.
Figure 7. Density of the specimens Figure 8. Effect of on shear strength

made by each method. at different specimen preparation
methods.
3.2. Deformation Characteristics
Figure 9 shows the variation of secant modulus with respect to shear strain level on the
specimens prepared by the air pluviation method. Although large difference appears in
the shear strain between 0.01% and 0.1%, small difference is exhibited in extremely
small strain of 0.001%. From the detailed examination, there is a tendency that larger
secant shear modulus appears when is larger. This is interesting because the tendency
is opposite with that of shear strength, in which smaller shear strength appears when
is larger.
In the same scale of graph with Fig. 9, the secant shear modulus on the specimens
prepared by the sticking method is shown in Fig. 10. There is no obvious difference in
the variation of secant shear modulus with the difference in . This is consistent tendency
with that of shear strength (Fig. 6). Therefore, the orientation of sand particles is well
disturbed by the sticking the thin rod into the sand many times.
Figure 9. Effect of on secant shear Figure 10. Effect of on secant shear

modulus in air pluviation method. modulus in sticking method.
The initial (maximum) shear modulus was estimated from the local small strain
(LSS) measurement and the bender element (BE) test. Figures 11 and 12 demonstrate the
initial shear modulus obtained from LSS and BE in the specimens prepared by the air
pluviation and sticking methods, respectively. Although inclined straight lines were used
for fitting to the plotted data in Fig. 11, horizontal line was used in Figs. 12 from the
tendency of the results.
The reason is not clear but in all specimens the initial shear stiffness obtained from
LSS is slightly larger than that estimated from BE. The one reason is considered to be
the estimation of arrival time of the shear wave in BE test. Although the arrival time has
been estimated from the wave that have a significant amplitude to exclude waves besides
the shear wave, the shear wave might arrive faster than the estimated. However, the
tendency, which indicates effect of , is a good agreement in the both method (LSS and
BE).
The initial shear modulus slightly increases when is large on the specimen
prepared by the air pluviation method (Fig. 11). It is interesting that this behaviour is
opposite with the tendency where shear strength decreases when is large (Fig. 8).
However this tendency becomes unclear on the specimen prepared by the sticking
method (Fig. 12). This result is the same tendency as that of shear strength (Fig. 5). The
specimens prepared by the sticking method exhibit the isotropic shear behaviour.
Concerning the value of the initial modulus, the initial shear modulus of the specimen
prepared by the air pluviation method is larger than that prepared by the sticking method
because of bigger density obtained from the air pluviation method (Fig. 7).
The one reason of this anisotropic behaviour can be considered as follow.
Horizontally lied particles have good engagement of particle structure. Then this particle
structure increases the shear strength during the triaxial compression test. In contrast,
concerning the initial shear modulus smaller number of contact points might be more
important than the firm engagement of particle structure. Because shear wave is
transmitted slower in =0 specimen than =90 specimen because the number of
contact points through transmitted direction is greater in =0 specimen and part of rigid
body formed from sand particles decreases as shown in Fig. 13.
Figure 11. Effect of on initial shear Figure 12. Effect of on initial shear
modulus in air pluviation method. modulus in sticking method.
Figure 13. Schematic diagram of orientation of sand particles.

4. Conclusions
The specimens having different angle of depositional plane were prepared using two
depositional methods which are the air pluviation and the sticking. Then the shear
properties of those specimens were carefully examined to understand the inherent
anisotropy of soils mechanical properties. The knowledge obtained by this study is
shown as follows:
Orientation of sand particles affects not only the shear strength but also the shear
modulus of the sand, although the effect on the shear modulus is smaller than
that on the shear strength.
When the angle of the depositional plane, which possess particle orientation,
becomes large, the shear strength decreases. On the contrary the initial shear
modulus increases.
When the particle orientation of the sand is well disturbed, anisotropic shear
behaviour disappeared. Therefore, in addition to the density, particle orientation
is also important to discuss the mechanical properties of sand in consideration
of anisotropic mechanical properties.
5. Acknowledgements
Experiments presented herein were conducted with the help of former graduate students
at the Department of Civil and Environmental Engineering, Nagaoka University of
Technology. The authors express their sincere thanks for their experimental assistance
and helpful cooperation.
References
[1] A. Casagrande and N. Carillo, Shear failure of anisotropic materials, J. Boston Soc. Civil Engineers 31,
7487, 1944.
[2] J.R.F. Arthur and B.K. Menzies, Inherent anisotropy in a sand, Gotechnique, 22(1), 115128, 1972.
[3] M. Oda, Initial fabrics and their relations to mechanical properties of granular materials, Soils and
Foundations, 12(2), 1736, 1972.
[4] M. Oda, and J. Konishi, Microscopic deformation mechanism of granular material in simple shear, Soils
and Foundations, 14(4), 2538, 1974.
[5] M. Oda, I. Koishikawa and T. Higuchi, Experimental study of anisotropic shear strength of sand by plane
strain test, Soils and Foundations, 18(1), 2538, 1978.
[6] Y. Yamada and K. Ishihara, Anisotropic deformation characteristics of sand under three dimensional stress
conditions, Soils and Foundations, 19(2), 7994, 1979.
[7] M. Jamiolkowski, R. Lancellotta and D.C.F. Lo Presti, Remarks on stiffness at small strains of six Italian
clays, Pre-failure Deformation of Geomaterials; Shibuya, Mitachi and Miura eds., Balkema, Netherlands,
pp.817836, 1995.
[8] D.C.F. Lo Presti and D.A. O'Neil, Laboratory investigation of small strain modulus anisotropy in sands,
Proceedings of ISOCCT1, Huang ed., Elsevier, New York, pp.213224, 1991.
[9] S. Yamashita, S. Kitagawa, T. Hori and T. Suzuki, Effects of fabric and induced anisotropy on initial shear
modulus of sands, Journal of JSCE, JSCE, No. 694(3-57), pp. 319329, 2001. (in Japanese)
[10] N. S. Rad and G. W. Clough, New procedure for saturating sand specimens, Journal of Geotechnical
Engineering, ASCE, 110(9), pp. 12051218, 1984.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-323
Element Tests on Lumpy

Inhomogeneous Soil and Their
Interpretation
X.S. Shi and I. Herle
Institute of Geotechnical Engineering, Technische Universitt Dresden, D-01062

Dresden, Germany
Abstract. Lumpy soils are inhomogeneous materials which can be encountered in

open pit mining. The initial structure of the clayfills is similar to that of the
granular material, which is stable due to the absence of a reconstituted soil.
However, it might be transformed into a lumpy-composite structure due to the
influence of climate. The lumps are randomly distributed in the reconstituted soil,
which can be bounded by two basic configurations. For this purpose, two series of
triaxial shear tests were performed on artificially prepared samples with parallel
and series configurations of the different constituents. The test results show that
the shear strength of the series samples is only slightly higher than that of the
weaker constituent and the strength of the parallel samples lies between those of
the constituents. For the parallel samples, there is a significant difference in the
stress concentration ratio at different consolidation stress; while, the strain
concentration ratio of the series samples is not significantly affected by the
consolidation stress.
Keywords. Clayfills, Element tests, Inhomogeneity, Stress (strain) distribution
1. Introduction
In many cases, soil can not be considered to be a homogeneous material at the level of
samples tested in the geomechanical laboratories. The lumpy soil may be an example
of such a material which is encountered e.g. in land reclamation and open pit mining
practice. A landfill of clayey overburden is a topical soil with a double porosity system
(Hollinderbaeumer and Kraemer, 1994; Boh et al., 2003; Shi and Herle, 2014). The
large-scale open-cast coal mining results in depositing the clayey overburden in non-
engineered landfills (Herbstov et al., 2009; Najser et al., 2012). These kinds of
landfills usually reach depths of several tens of meters. The clayey overburden soils are
being dumped without any compaction in the form of lumps and blocks ranging in size
from millimeters up to about tens of centimeters (Fig. 1) (Man et al., 2005). The
initial structure of the freshly deposited clayfills is stable due to the absence of a
reconstituted soil in the inter-lump voids.
324 X.S. Shi and I. Herle / Element Tests on Lumpy Inhomogeneous Soil and Their Interpretation
Figure 1. Fresh landfills and its simplified structure (Man et al., 2005; Shi and Herle, 2014)
Figure 2. Old landfills and its simplified structure (Man et al., 2005; Shi and Herle, 2014)
After a reasonable time from filling, mainly due to the influence of climate,
gravitational effect and precipitation, the originally clays in landfills with a high inter-
lump porosity tend to become relatively homogeneous soils at the macro scale.
However, the double porosity still remains in this lumpy-composite structure (Fig. 2)
(Man et al., 2005; Najser et al., 2010; Najser et al. 2012). The matrix and inclusions
of the material are the reconstituted soil and stiff lumps, respectively. Hartlen and
Ingers (1981) conducted field investigation in land reclamation. Large instantaneous
vertical and horizontal movements indicated that the stiffness and strength of the
clayfills mainly depend on the reconstituted soil in the inter-lump voids. Shi and Herle
(2014) conducted series of oedometer and triaxial tests on the artificially prepared
lumpy material. They found that the reconstituted soil, which exists in the inter-lump
voids, plays a crucial role in the behaviour of lumpy materials.
X.S. Shi and I. Herle / Element Tests on Lumpy Inhomogeneous Soil and Their Interpretation 325
Figure 3. Simplified structure of a clayfill material after long time of filling
In the homogenization framework, two basic models have been used to estimate the
mechanical behaviour of inhomogeneous materials: the parallel structure (Fig. 3 right
bottom) which gives the upper bound of the stiffness (Voigt, 1928), and the series
structure (Fig. 3 right top) which corresponds to the lower bound (Reuss, 1929). For the
lumpy materials, the lumps are randomly distributed in the reconstituted soil. The
structure is composed by both series and parallel structures at different orientations.
Therefore, the understanding of these two basic configurations can help as a reference
for the assessment of the mechanical behaviour of lumpy material.
Two inhomogeneous artificial samples of the series and parallel configurations

respectively, are tested in this study. The stress strain relationship and the failure mode
for these two configurations are compared. Based on the homogenization model, the
stress and strain concentration ratios are analyzed.
2. Materials and sample preparation
The soils used in this study are a high plasticity clay (Clay1) and a silty clay (Clay2).
Clay1 was taken from Guttau city (east of Germany). Clay2 was taken from a highway
foundation in Dresden (southeast of Germany). The basic physical properties are given
in Table 1. The organic contents of Clay1 and Clay2 are 10.6% and 2.0%, respectively.
The materials were first mixed with water and then sieved through a mesh with
diameter of 0.425 mm to exclude coarse grains from the natural soil.

Table 1. Basic physical properties of the reconstituted clays
Soil Density of particles Liquid limit Plastic limit Clay Silt Sand Gravel
g/cm3 % % % % % %
ClayA 2.63 93.5 40.9 91 6 3 0
ClayB 2.70 33.8 19.6 18 70 10 2

After mixed with water, the slurry for the homogeneous samples was poured into a
consolidometer (with the diameter of 3.8 cm), then consolidated to 92 kPa by gradually
adding slotted weights on the weight hanger. The duration of each load increment was
24-48 hours. After fully consolidated, the samples were extruded out from the
consolidometer and trimmed to 7-8 cm in height.
The preparation of the series samples was similar to that of the homogeneous ones. The
slurry of Clay1 and Clay2 were poured successively into a consolidometer (with the
diameter of 3.8 cm) and then consolidated stepwise.
The preparation of the parallel samples started by consolidation in a large

consolidometer (diameter = 14 cm). Similarly to the preparation of the series samples,
the slurry of Clay1 was poured there and initially consolidated separately to 20-25 kPa.
Afterwards, the slurry of Clay2 was placed above Clay1. The consolidation stresses
increased then to 80-120 kPa. After fully consolidated, the sample was extruded out
from the consolidometer and cut with a thin-wall tube in the parallel direction to the
interface.
Figure 4. Parallel and series samples
The samples (see Fig. 4) were then placed into a triaxial cell and isotropically
consolidated to specified effective mean pressures. The compression tests were
performed in drained conditions. The sample dimensions and the initial void ratios are
listed in Table 2.
Table 2. Initial void ratio and size of the specimens
Test number Height Diameter Initial void ratio Consolidation pressure

/cm /cm Clay1 Clay2 Total / kPa
Clay1-1 7.59 3.80 1.72 --- 1.72 150
Clay1-2 7.19 3.76 1.78 --- 1.78 300
Clay2-1 7.79 3.82 --- 0.71 0.71 150
Clay2-2 7.46 3.75 --- 0.78 0.78 300
Series-1 7.58 3.80 1.99 0.75 1.16 150

Series-2 7.57 3.80 1.99 0.75 1.14 300
Parallel-1 7.41 3.80 1.75 0.74 1.07 150
Parallel-2 7.52 3.80 1.75 0.74 1.12 300
The initial void ratios of the inhomogeneous samples in Table 2 were calculated
from the overall water content and particle densities of Clay1 and Clay2 using the
following equation:
. s ,1 . s ,2
.s (1)
. s ,1 (1 f m ) . s ,2 f m
where . s ,1 and . s ,2 are the particle densities of Clay1 and Clay2, respectively, f m is
the ratio between the dry weight of Clay1 and the whole inhomogeneous sample.
3. Comparison between inhomogeneous and homogeneous samples
As shown in Table 2, two series of drained triaxial tests were performed, including
both the homogeneous and inhomogeneous samples. In the evaluation of the results,
the stresses and strains within the samples were calculated regardless the
inhomogeneity. Thus, the samples were considered as soil elements. Based on test
results of Clay1 and Clay2, the properties of the inhomogeneous samples can be
analyzed. Figs 5 and 6 show the comparisons of the homogeneous and the
inhomogeneous samples in form of stress-strain curves at 150 kPa and 300 kPa,
respectively.
Figure 5. Comparison of the stress-strain and volumetric curves of the homogeneous and of the
inhomogeneous samples at 150 kPa
Figure 6. Comparison of the stress-strain and volumetric curves of the homogeneous and of the
inhomogeneous samples at 300 kPa
It can be seen in Figs 5 and 6 that the stiffness and strength of Clay2 are significantly
higher than those of Clay1 at the same consolidation pressure. The stress-strain curves
of Clay1 show softening behaviour at deviatoric strain larger than 10% to 12%, while
Clay2 still shows strain hardening beyond 12% of deviatoric strain. The stress-strain
curves of the series samples show softening behaviour at relatively small deviatoric
strains (6% to 8%). The parallel samples show also a slight softening, although at
higher strains. As expected, the curves of inhomogeneous samples are located between
those of homogeneous ones. The stiffness of the series samples is higher than that of
Clay1, however, their limit stress values are relatively close to each other. The stiffness
and limit stress of the parallel samples lie between those of the constituents.
Figure 7. Comparisons of the stress ratio between the homogeneous and inhomogeneous samples
The stress ratios of both the homogeneous and the inhomogeneous samples are shown
in Fig. 7 for two different cell pressures. The limit stress ratio of the series sample is
only slightly higher than that of Clay1 sample, while the values of the parallel sample
lies between those two constituents. It indicates that the shear strength of the series
samples is controlled by only Clay1, while, the mechanical behaviour of the parallel
samples depend on both constituents.
4. Stress (strain) concentration based on homogenization method
After Hill (1963), the stress concentration ratio is defined as the ratio between the
deviatoric stress of Clay2 and that of Clay1. Similarly, the strain concentration ratio
is defined as the ratio between the axial strain of Clay2 and that of Clay1. Both of them
can be calculated from the stiffnesses of its constituents according to the
homogenization theory.
Figure 8. Stress concentration ratio in parallel samples at different consolidation pressures and axial strains
Fig. 8 shows the calculated stress concentration ratio in the parallel samples at different
consolidation stress. The stress concentration ratio is approximately constant with the
increase of the axial strain, however, it is significantly affected by the consolidation
stress in the tested stress range (from 150 kPa to 300 kPa). An interaction on the
interface or the penetration between the constituents may contribute to this effect.
Fig. 9 shows the strain concentration ratio as a function of the deviatoric stress within
the tested stress range. It is not significantly affected by the deviatoric stress.
Furthermore, there is no significant dependence of the strain concentration ratio on
consolidation stress.
Figure 9. Strain concentration ratio between Clay1 and Clay2 in series samples at different consolidation
pressures and axial stresses
5. Conclusions
This paper investigates the behaviour of two types of inhomogeneous samples. The
stress strain behaviour and the ratios of stresses and strains within the constituents are
discussed. The following conclusions can be done:
(1) The initial structure of the clayfills is similar to that of the granular material, which
is stable due to the absence of a reconstituted soil. However, it might be transformed
into a lumpy-composite structure due to the influence of climate. The structure is
composed by both series and parallel structures at different orientations, which can be
bounded by two basic configurations.
(2) The limit stress in the samples with series configuration of the constituents is
comparable to that of the Clay1 samples, which means that the shear strength of the
series samples is fully controlled by Clay1. The limit stress of the parallel samples is
located between those of the constituents, indicating that their mechanical behaviour
depend on both constituents.
(3) The stress concentration ratio of parallel samples is almost constant at different
axial strains, however, it is significantly affected by the consolidation stress. The strain
concentration ratio of the series samples is not significantly affected by the
consolidation stress and the deviatoric stress.
References
[1] E.W. Hollinderbaeumer and U. Kraemer, Waste disposal and backfilling technology in the German hard
coal mining industry, Bulk solids handling 14 (1994), 795-798.
[2] J. Boh, V. Herbstov, P. Koran et al., Properties of unsaturated colliery clayfills in North-Western
Bohemia, Proc. 13ECSMGE, Prague, Vol. 1 (2003), 33-36.
[3] X. S. Shi and I. Herle, Laboratory investigation of artificial lumpy materials, Engineer Geology 183
(2014), 303-314.
[4] V. Herbstov and I. Herle Structure transition of clay fills in North-western Bohemia. Engineering
Geology 104 (2009), 157-166.
[5] J. Najser, E. Pooley, Springman et al., Mechanisms controlling the behaviour of double porosity clayfills
- insitu and centrifuge study, Quarterly Journal of Engineering Geology and Hydrogeology 43 (2010),
1-14.
[6] D. Man, V. Herbstov and J. Boh, Properties of double porosity clay fills and suitable constitutive
models. Proc.16th Int. Conference ICSMGE, Osaka, Japan, Vol. 2 (2005), 827-830.
[7] J. Najser, D. Man, and J. Boh, Numerical modeling of double porosity clay landfill. International
Journal of Numerical and Analytical Methods in Geomechanics, 36 (2012), 17-35.
[8] J. Harden, and C. Ingers, Land reclamation using fine-grained dredged material. Proceedings of the
International Conference on Soil Mechanics and Foundation Engineering, Stockholm, Swed, A. A.
Balkema, Rotterdam 1 (1981), 145-148.
[9] W. Voigt, Lehrbuch Der Kristallphysik, Teubner Verlag, Berlin-Leipzig, (1928).
[10] A. Reuss, Berechnung der fliessgrenzen von mischkristallen aurf grund der plastiizitatsbedingug fuer
einkristalle, Zeitschrift fur angewandte mathematik und mechnik 9 (1929), 49-58.
[11] R. Hill, Elastic properties of reinforced solids: some theoretical principles. Journal of the Mechanics
and Physics of Solids 11 (1963), 357-372.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-331
Dynamic Shear Modulus of Kaolin-Silt

Clay Using a Novel Technique
Saeed AHMAD a, Ahmed M. ALNUAIM b, M. Hesham EL NAGGAR c.1
a
PhD candidate and research assistant at the Department of Civil and
Environmental Engineering, The University of Western Ontario, London, ON, Canada,
N6A 5B9. ([email protected])
b
Assistant Professor at Civil Engineering Department, King Saud University,
Riyadh, Saudi Arabia and affiliated with The University of Western Ontario, London,
Ontario, Canada. ([email protected])
c
Professor and Research Director, Geotechnical Research Centre, Department of
Civil and Environmental Engineering, The University of Western Ontario, London,
Ontario, Canada ([email protected])
Abstract. An artificial/reconstituted clay was prepared by mixing kaolin (K) and

silt (S) in the proportion of 50% kaolin and 50% silt by weight employing a slurry
deposition technique. The K-S clay has been used extensively to simulate natural
clays for geotechnical centrifuge testing, often involving dynamic loading. It is
necessary to accurately measure the soil shear wave velocity (Vs) or the low-strain
shear modulus (Gmax) for the analysis of geotechnical problems involving dynamic
loading. In this study, Vs and Gmax of K-S clay were measured using a novel
device, utilizing piezoelectric transducer ring actuators, incorporated in an
oedometer. The new device eliminates some of the shortcomings of the bender
element test. For example, the piezoelectric transducer ring actuators do not
penetrate the soil and hence do not cause any sample disturbance.
The value of Vs of the K-S clay was measured at different effective vertical
pressure increments ranging between 25 and 500 kPa. The obtained values of Vs
and Gmax were independently compared with the values predicted using empirical
correlations available in the literature for clay soil. An excellent agreement was
found between the measured and the predicted values. The results of the present
study are presented as Vs - m and Gmax - m models. The proposed models will be
helpful in designing a K-S clay testbed with a targeted stiffness for dynamic
centrifuge applications.
Keywords. Kaolin-Silt clay, centrifuge, shear wave velocity, low-strain shear

modulus, piezoelectric transducer.
1. Introduction
The use of pure kaolin clay to simulate the behavior of natural clay in geotechnical
physical modeling has been adopted in many research studies. Using kaolin clay
facilitates simulating the behavior of different soils with different stress histories and
undrained shear strengths (Su). In addition, its high permeability compared to natural
1
332 S. Ahmad et al. / Dynamic Shear Modulus of Kaolin-Silt Clay Using a Novel Technique
clays reduces the laboratory and centrifuge consolidation times. Due to the high
plasticity of pure kaolin clay itself, it is difficult to achieve soil profile with high
strength and stiffness to replicate the true behavior of natural soils. Generally, a soil
with high clay content tends to develop a residual shearing surface with low strength. A
typical residual friction angle, r of about 12/ smaller than the critical friction angle,
c, has been reported by [1], [2] and [3]. Therefore, different attempts have been made
to increase the strength and stiffness of kaolin by adding granular materials.
References [4] and [5] studied the performance of the 50% speswhite kaolin clay and
50% Sil-Co-Sil silt (by weight) as the modeling clay. The clay was preconsolidated to
400 kPa with an overconsolidation ratio (OCR) of 4.2 under centripetal acceleration of
50g. The Su of K-S clay was about 50 kPa at vertical effective stress (v) = 95 kPa. The
K-S specimens tested in the current study were prepared for an ongoing centrifuge
testing program.
The use of K-S for dynamic centrifuge testing to evaluate the soil-structure
interaction, e.g. dynamic performance of foundations and pipelines, requires evaluating
its dynamic properties, most importantly the low-strain shear modulus [6].
A novel device utilizing piezoelectric rings was incorporated in an oedometer cell
and was used to measure the shear wave velocity (Vs) of the test specimens. The device
was rigorously tested using well documented Ottawa sand and was found to yield
excellent results. The current study focuses on measuring Vs and hence finding Gmax of
K-S clay using this device.
2. Tested material and preparation procedure
A brief description of the kaolin clay and silt used in this study, which includes;
the physical properties, partial size distribution and Atterberg limits, is presented. The
Speswhite kaolin clay was manufactured by IMERYS in England. The silt used in the
K-S mixture was Sil-Co-Sil 75, supplied by US SILICA. The specific gravity for the
kaolin and K-S clay are 2.63 and 2.62, respectively [4]. A number of tests were
conducted in order to obtain the physical properties of the pure kaolin and K-S mixture.
The hydrometer method was used according to [7] to evaluate the soil particles
distribution. It was found that the mean grain size, D 50, for the kaolin, silt and K-S
mixture to be about 0.00052 mm, 0.017 mm and 0.0025 mm, respectively (Figure 1).
The results are in agreement with those reported by [4] and [5]. Figure 1 presents the
particles size distribution curves for K-S clay as well as kaolin and silt alone. The
Atterberg limits shown in Table 1 for the K-S mixture were determined according to
[8].
Table 1. Properties of K-S mixture.

Soil Type 50% kaolin and 50% silt
Liquid Limit, wL 38%
Plastic Limit, wP 22.3%
Plasticity Index, Ip 15.72%
Mean Particle size, D50 0.0025 mm
Clay Fraction (<2m) 50%
S. Ahmad et al. / Dynamic Shear Modulus of Kaolin-Silt Clay Using a Novel Technique 333
Figure 1. (a) Soil particles distribution for different materials; and (b)
The preparation of the K-S clay bed involved a number of steps, including:
estimating the laboratory consolidation pressure in which the required OCR in the
centrifuge was attained; mixing the 50% kaolin clay with 50% silt; 1g-laboratory
consolidation of slurry; and finally, the consolidation in the centrifuge. The initial
moisture content and void ratio for the slurry were 80% and 2.1, respectively. The
following section describes these steps in details.
A 5HP Bower horizontal paste mixer with 200 L capacity was used to mix a total
of 180 kg of kaolin clay, 180 kg of silt and 288 kg of water. The total weight of
material was divided into four batches, each weighing about 162 kg. The ingredients
were mixed for about 30 minutes at 14 RPM and were subjected to vacuum pressure of
65 kPa for minimum 3 hours. The purpose of using the vacuum pressure was to reduce
the air trapped in the slurry. This procedure ensured homogenous slurry with high
degree of saturation towards the final consolidated K-S mix.
A 904 mm diameter and 817 mm extended height container was used to
consolidate the slurry. The extended container consisted of two parts: (I) the primary
round tub, 500 mm high (Figure 2a); and (II) 317 mm high extension was connected to
the primary tub to contain the slurry before the consolidation process (Figure 2b). A
steel extrusion plate was placed on the bottom of the tub to facilitate post-test
extraction of the clay test bed. A 40 mm saturated sand layer was placed at the bottom
of the tub to accelerate the consolidation process (Figure 2c). A layer of geotextile and
filter paper were placed on top of the sand layer, which allowed the water to be
squeezed out without squeezing any clay particles (Figure 2d). To reduce the friction
between clay mixture and the containers walls during the consolidation process, the
container wall was coated using a thin grease layer (Figure 2d). The slurry was then
poured into the container with an extreme caution not to allow air to be trapped within
the slurry (Figure 2e). A layer of filter paper was provided at the top of the mix,
overlain by the layer of geotextile (Figure 2f).
After placing the slurry in the container (Figure 2f), it was allowed to consolidate
under its self-weight for one day. A 233 kg steel rigid piston was then carefully placed
level on top of the slurry (Figure 3a), which produced a pressure 3.6 kPa. The slurry
was subjected to the piston pressure until 90% consolidation was achieved.
The settlement was measured using two string pots. Next, the tub was loaded under
the consolidation frame to consolidate the slurry under higher pressures using hydraulic
press (Figure 3b), which can apply required pressure. The slurry was drained from the
top and the bottom of the tub during the consolidation process. A series of stress
increments were applied to consolidate the slurry till the maximum consolidation stress
of 300 kPa. The applied stress was doubled after each successive load increment. Each
load increment was kept until at least 90% consolidation was achieved. The entire
process took about five weeks to complete with a total consolidation settlement 200
mm. Figure 4 shows the e-log v of K-S clay sample for the centrifuge testing program.
a b
c d
e f
Figure 2. Different steps involved in preparing the steel tub and pouring the slurry
After the slurry was consolidated to a maximum stress level, the unloading process
was started by reducing the pressure by half until it was removed completely.
a b
Figure 3. (a) The rigid steel piston: (b) The tub under the consolidation frame
2.50
2.00
1.50
e
1.00
0.50
0.00
0.1 1 10 100 1000
log v (kPa)
Figure 4. The e-log v of K-S clay during preparing clay sample for centrifuge tests.
3. ID-consolidation test on the prepared K-S Samples
Four conventional oedometer tests were performed on the prepared K-S soil
samples. Initial water content of the tested samples was found to be 30% and an
average initial void ratio (eo) of 0.85, which are consistent with the values obtained
during preparing the samples from the slurry and pre-consolidated to a pressure of 300
kPa. Samples were tested under v (including the saturation load) of 26.25, 51.25,
101.25, 201.25, 276.25, 326.25, 401.25, 451.25, and 501.25 kPa
Figure 5 shows the e-log v of K-S clay. Compression index (Cc) of the tested soil
was found to be around 0.21. An average recompression index (Ccr) was around 0.15
and the swell index (Cs) was about 0.04. The value of Cc was found to be consistent
with those obtained from the empirical equations ([9] - [13]) using different
geotechnical properties of K-S soil, e.g., liquid limit, plasticity index, and specific
gravity.
Figure 5. Results of 1D consolidation tests on K-S clay
4. Methodology, Testing Details and Interpretation Techniques for Vs

Measurements
A novel device utilizing piezoelectric rings was incorporated into an oedometer to

measure the shear wave velocity (Vs) of K-S clay. High voltage input sinusoidal signals
were used to measure Vs. Input signals of different frequencies between 15-45 kHz
were used and both the input and output signals were captured and analyzed using a
program written in the LabVIEW environment. Different frequencies were used to
evaluate the effects of L/ (L = height of soil sample, = wavelength of input wave).
Previous tests results on sand and clay specimens demonstrated reliable measurements
of Vs when L/ 3 - 4. Similar conclusions were drawn by other researchers ([14], [15]
and [16]). On the other hand, very high values of L/ cause significant attenuation of
output wave amplitudes due to soil damping [17, 15 and 18].
A typical signal obtained during the present study is shown in Figure 6. Using the
direct arrival method, the first zero crossing before the major peak i.e. pt. A is taken as
the arrival of the shear wave. A detailed analysis of output signals was performed in
both time and frequency domain. The measured Vs values established using pt. A as the
shear wave arrival time for the well documented Ottawa sand were in good match with
values reported in the literature. The group velocity method (Figure 7), which involves
processing the output signals in the frequency domain [19] was found to be most
reliable for interpreting the test data and it predicted the shear wave arrival point that
matched very well with the direct arrival method (pt. A). Other methods of analysis e.g.
characteristic peaks and cross-correlation were considered in this study but yielded
highly erroneous results specially the cross-correlation.
Figure 6. Typical output signal obtained during the current study showing very clear shear wave
The cross-correlation method (time domain analysis) was found to significantly

overestimate the arrival time and hence underestimate Vs. Figure 7 shows the results of
time and frequency domain analyses on the output signal highlighting the discussed
details.
Figure 7. Result of the arrival times through time and frequency domain analyses
5. Measured Shear wave velocity and comparison with literature
Figure 8 shows the effect of m on Vs of K-S clay, presented as Vs - m

relationship. The relationship is shown in Eq. (1)

Where factor (m/s) is Vs at 1 kPa and the exponent shows sensitivity of Vs to

mean in the polarization plane. For isotropic stress condition, [20] presented an inverse
relationship between and i.e.

Figure 8. Power curve fitting to obtain and parameters
Figure 8 represents the average values for Vs for all the four tests. For the tested
soil, 59.8 and = 0.256 as can be seen in Figure 8, i.e.

The measured parameters and shows an excellent agreement with the inverse
relationship given in Eq. (2).

Reference [21] presented an empirical correlation i.e. Eq. (4) for Vs of

overconsolidated cohesive soils. Where m is the mean effective stress in kPa and e is
the void ratio. Measured values are compared with the predicted values in Figure 9.
Average values of e for all the four tests were used in Eq. (4). As shown in Figure 9, an
excellent agreement was found between the measured and predicted values. At all
pressures, the arithmetic averages of the measured values of all four tests are slightly
higher than the predicted values using Eq. (4).
6. Results of obtained Gmax and comparison with literature
Using the measured Vs values and the densities () from the 1D consolidation data,
the low-strain shear modulus (Gmax) was calculated using the simple relationship given
in Eq. (5)

Figure 9. Comparison of measured Vs values with [21]
Figure 10 shows the variation of the measured Gmax with m for all the fours tests.
A power curve was fitted into the measured values. The figure shows Gmax varies with
m with a power of 0.52 which is twice the power of m in Vs- m model. This is
same trend as documented in a lot of published literature.
Figure 10. Measured Gmax values with power curve fitting
Finally, m calculated from v using (ko)NC for normally consolidated soil [4]. To
account for the overconsolidation ratio (OCR) effect, (ko)OC was calculated for each
load increment along the recompression curve. The effective friction angle, = 25
was measured from CIU triaxial tests and was use to evaluate (ko)NC and (ko)OC.
There is no published data on measurements of Vs and Gmax for K-S clay; hence
comparisons are made with values relevant to soils K-S clay is intended to simulate,
which are predicted using empirical correlations [22, 23]. The value of Gmax for several
types of soils can be related to the mean effective stress m and the soil void ratio, e,
[22], i.e.:

For clayey-silts, Gmax can be given by [23]:

In Eq. (7), Pa ( 100 kPa). Figure 11 compares the measured and the predicted
Gmax values. The predicted Gmax using both the equations matched very well with the
measured values at low pressures. At moderate pressures or pressures up to the
preconsolidation pressure, the predicted values using Eq. (6) are slightly higher than
the measured values, whereas, at high pressures or in the normally consolidated range
the measured ones are fairly lower as compared to the predicted values using Eq. (6).
On the contrary, an excellent agreement was found between the measured and the
predicted values using Eq. (7) at all pressures before, at, and after the preconsolidation
pressure.
Figure 11. Measured and the predicted Gmax using Eqs. (6) and (7)
7. Conclusions
K-S clay was prepared by consolidating the mixture of kaolin and silt in the 50-50
proportion by weight. A conventional oedometer was modified to measure Vs using a
novel technique. Four oedometer tests were performed and Vs was measured
simultaneously. A Vs - m model is presented for K-S clay. Comparing the measured Vs
values with the available literature showed a very good agreement. Finally, the small
strain shear modulus values Gmax were calculated and compared with predicted values
using relevant available empirical correlations in the literature. Excellent agreement
with available empirical correlations that evaluate Gmax as a function of confining
pressure and void ratio was observed.
References
[1] N. K. Tovey, Electron Microscopy of Clays, University of Cambridge PhD dissertation, (1970).
[2] G. Rossato, R. J. Jardine, and N. L Ninis. Properties of Some Kaolin Based Model Clay Soils.
ASTM Geotechnical Testing Journal, 15, 2 (1992), 66-179.
[3] S. M. Springman, Centrifuge Modelling in Clay: Marine Application. Proceedings of the 4th
Canadian Conference on Marine Geotechnical Engineering, St. Johns, Newfoundland. (1993) 397-403.
[4] L. Lin, Strength Characteristics of a Modelling Silty Clay M.Eng. Thesis, Memorial University of
Newfoundland. St. Johns, Newfoundland. Canada, (1995).
[5] M. J. Paulin, An Investigation into Pipelines Subjected to Lateral Soil Loading. PhD Thesis,
Memorial University of Newfoundland. St. Johns, Newfoundland, Canada, (1998).
[6] S.L. Kramer, Geotechnical Earthquake Engineering, Prentice Hall, Inc., Upper Saddle River, New
Jersey, 1996.
[7] ASTM D422-63,"Standard test method for particle-size analysis of soils," ASTM International, West
Conshohocken, PA, (2007).
[8] ASTM D4318," Standard Test Methods for Liquid Limit, Plastic Limit, and Plasticity Index of Soils"
ASTM International, West Conshohocken, PA, (2010).
[9] T. Nagaraj, and B. S. R. Murty, Prediction of the Preconsolidation Pressure and Recompression Index
of Soils, Geotechnical Testing Journal, 8, 4, (1985), 199202.
[10] C.P. Wroth, and D. M. Wood, The Correlation of Index Properties with Some Basic Engineering
Properties of Soils, Canadian Geotechnical Journal., 15, 2 (1978), 137145.
[11] A.W. Skempton, Notes on the Compressibility of Clays. Q. J. Geol. Soc. London, 100, 1-4, (1944),
119-135.
[12] O. Rendon-Herrero, Universal Compression Index, J. Geotech. Eng., 109, 10, (1983), 1349.
[13] F.H. Kulhawy, and P. W. Mayne, Manual on Estimating Soil Properties in Foundation Design, Elect.
Power Res. Inst., Palo Alto, Calif., (1990).
[14] E.C. Leong, S.H. Yeo, and H. Rahardjo, Measuring Shear Wave Velocity Using Bender Elements.
ASTM Geotechnical Testing Journal 28, 5 (2005), 111.
[15] E.C. Leong, J. Cahyadi, and H. Rahardjo, Measuring Shear and Compression Wave Velocities of Soil
Using Benderextender Elements. Canadian Geotechnical Journal, 46, 7, (2009), 792812.
[16] C. Mancuso, and F. Vinale, Propagazione delle onde sismiche: teoria e misura insito. Atti del Convegno
del Gruppo Nazionale di Coordinamento per gli Studi di Ingegneria Geotecnica, Monselice, Rome:
Consiglio Nazionale delle Ricerche, (1988), 115-138.
[17] C. M. Chan, On the Interpretation of Shear Wave Velocity from Bender Element Tests. ACTA
TECHNICA CORVINIENSIS-Bulletin of Engineering: (2012), 2934
[18] E.G.M. Brignoli, M Gotti, and K.H. Stokoe. Measurement of Shear Waves in Laboratory Specimens
by Means of Piezoelectric Transducers. ASTM Geotechnical Testing Journal 19, 4, (1996), 384397.
[19] G. Viggiani, and J.H. Atkinson, Interpretation of Bender Element Tests (Technical Note).
Geotechnique, 45, 1, (1995), 149154.
[20] M. Cha, J.C. Santamarina, H-S. Kim and G-C. Cho, Small-Strain Stiffness, Shear-Wave Velocity and
Soil Compressibility Technical Note. ASCE Journal of Geotechnical and Geoenvironmental
Engineering. (2014). ISSN 1090-0241/06014011 (4)
[21] K. Markowska-Lech, The determination of shear modulus in overconsolidated cohesive soils.
Foundations of Civil and Environmental Engineering, Poznan University of Technology, 12, (2008),
61-71.
[22] J. Biarez, and P.Y. Hicher, Elementary Mechanics of Soil Behaviour-Saturated Remoulded Soils,
Balkema, 1994.
[23] B. D Elia, and G. Lanzo, Laboratory and field determinations of small-strain shear modulus of
natural soil deposits, Proc. of the 11th World Conference on Earthquake Engineering. (1996).
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-342
Study of the mechanical behavior of

unsaturated argillaceous rocks
Jairo ESPITIAa,b and Bernardo CAICEDO a
a
Civil and environmental department, Universidad de los Andes, Bogot, Colombia
b
Mining engineering department, Universidad Pedaggica y Tecnolgica de Colombia
Abstract. This paper study the influence of water content on the mechanical
behavior and failure process of an unsaturated argillaceous rock. For this, uniaxial
compression tests with acoustic emission recording were conducted on specimens
at different levels of relative humidity. Results indicate that uniaxial compressive
strength and the elastic modulus increases with a reduction in the water content.
Additionally, rock moisture seems to influence the volumetric deformation.
Analyzing the volumetric stress-strain curves in conjunction with acoustic
emission technique was possible to identify the crack initiation and damage stress
thresholds at different values of relative humidity.
Keywords. Unsaturated rocks, acoustic emission, argillaceous rocks.
1. Introduction
The mechanical properties of argillaceous rocks are the great interest in different areas
of rock engineering. These are strongly dependent on the variation of the water content
and therefore on the environmental conditions. Also, the mineralogical composition has
a significant influence on their behavior because of the sensitivity of clay minerals to
rock moisture [1]. Moreover, its mechanical behavior is the result of the combination of
sedimentation, gravitational compaction (consolidation), uplift/unloading
(overconsolidation) and cementation/bonding [2].
Some experimental studies have been carried out recently on argillaceous rocks in
unsaturated conditions [3-5] These investigations agree on some features of the hydro-
mechanical behavior of argillaceous rocks. First, during desorption adsorption cycles
without confinement are observed hysteresis, anisotropic deformations and permanent
volumetric strains in the material. Second, elastic properties and strength are controlled
by the hydric state and the hydric history of the material. Finally, the material is
degraded as a consequence of hydric and mechanical loading cycles.
An additional feature of interest in the behavior of argillaceous rocks corresponds
to its brittle behavior. Indeed, some experimental findings demonstrate that
argillaceous rocks may fail in a brittle manner under low-confinement conditions such
as those characterizing the near-field of the excavation [6].
The process of failure and deformation of brittle rocks have been studied by
numerous researchers over the last decades [7, 8]. In general, this process can be
divided into five stages based on the stress-strain curves obtained from uniaxial and
triaxial compression tests (Fig. 1). Bieniawski [7] defined these phases as being: (I)
crack closure, (II) linear elastic deformation, (III) crack initiation and stable crack
J. Espitia and B. Caicedo / The Mechanical Behavior of Unsaturated Argillaceous Rocks 343
growth, (IV) crack damage and unstable crack growth, and (V) failure and post-peak
behavior. Also, each stage is limited by a stress threshold: (a) cc stress at crack closure,
(b) ci crack initiation stress threshold, (c) ci crack damage stress threshold and (d)
peak peak stress.
Figure 1. Stages and stress thresholds for the brittle behavior of rocks
Considerable research has been devoted to study the influence of water content and
relative humidity cycles on some physical, mechanical properties and deformation
characteristics of argillaceous rocks. However, only a few studies have considered the
study of the brittle behavior of argillaceous rocks in unsaturated conditions. For this, an
experimental study was carried out to evaluate the influence of hydric states on the
mechanical behavior of this material.
2. Materials and methods
2.1. Material
The material studied is a lower Hauterivian Barremian argillaceous rock taken from the
quarry of limestone of the Holcim Company located in the eastern Andes mountain
range in Nobsa (Colombia). The average mineralogical composition is: 53.2% kaolinite
and illite, 29.0% calcite, 11.0% quartz, 4.0% opaque minerals 4.0% and 1.0%
Muscovite. Table 1 summarizes some properties of the rock.
Table 1. Index properties of argillaceous rock

Property Value
Bulk density 2570 kg/m3
Specific gravity 2.74
Water content 1.51%
Liquid limit 25
Plastic limit 15
The irregular rock blocks sampled from the quarry were covered with plastic
stretch film and placed inside an EPS foam Box until preparation of samples in the
laboratory. Samples, of 20 mm in diameter and 40 mm in height, were bored with air
344 J. Espitia and B. Caicedo / The Mechanical Behavior of Unsaturated Argillaceous Rocks
pressure to minimize alteration caused by liquid water. Then, the end faces were
polished with a lathe. The dimensions of the specimens were chosen to reduce the time
needed to achieve moisture equilibrium within a controlled humidity atmosphere.
Samples tested were obtained from five rock blocks (rock block 6-B6, rock block 15-
B15, rock block 1-B1, rock block 10-B10 and rock block 17-B17).
Mercury intrusion porosimetry was performed on freeze-dried specimens using an
AutoPore IV 9500-Micrometrics Instrument Corp. porosimeter. Figure 2 shows the
pore size distribution of the undisturbed rock. The dominant pores of the material vary
between 20 and 80 nm.
Figure 2. Pore size distribution for undisturbed argillaceous rock
2.2. Specimen conditioning
The experimental program consisted of setting the cylindrical specimens in a tight box
in which relative humidity (RH) was imposed by saline solutions. For this purpose, two
experiments were conducted. First, a group of 15 samples from B6 and B15 were
divided and set in five hermetic containers at different relative humidity (RH). Second,
a group of 6 specimens from B1, B10 and B17 were first conditioned at 36% of RH.
Then, half of the specimens were mechanically tested. The remaining were conditioned
at 46% of RH (Fig. 3). Table 2 reports the adopted saline solutions for this study with
its corresponding suction value. Relative humidity and temperature measurement were
continuously monitored using a digital hygrometer placed inside the hermetic
containers. During the measuring period, the temperature stayed approximately
constant at 20.80.9C.
Figure 3. Specimen conditioning experiments

The moisture equilibrium of the rock specimens with the controlled humidity
atmosphere were assumed when weight was stabilized. In general, the equilibrium was
reached after eight days. Once, specimens reached the equilibrium mechanical tests
were performed.
Table 2. Relative humidity for saline solutions used.

Saline solution Relative Humidity (%) Suction (MPa) Water content (%)
Potassium sulfate 94 8 2.71
Sodium chloride 70 48 2.01
Sodium nitrite 63 64 1.86
Potassium carbonate 46 106 1.44
Calcium chloride 36 147 1.22
2.3. Mechanical tests
Uniaxial compression tests were conducted on conditioned specimens in a controlled

displacement load frame. The axial displacement rate was 0.1 mm/min. Strain gages
were fixed to measure the local strains in the axial and radial axis directions. The rock
moisture was assumed to remain constant during the test due to their short duration (20
min).
During the mechanical tests, Acoustic Emission (AE) activity was monitored to
track the process of failure of the specimens. These measurements were stored with
AE equipment from the Physical Acoustics Corporation that consisted of two channels.
For all the samples, uniaxial stress was applied perpendicular to the bedding plane. In
Figure 4, we report the stress-strain curves obtained at different levels of RH for the B6
samples of the first experiment. From this, we determined the Elastic Modulus and the
compressive strength. The elastic modulus was estimated as the tangent slope of the
axial strain-stress curve at one-half of the peak stress.
Figure 4. Stress-Strain Curves of argillaceous rock for different values of RH for B6 specimens
In general, compressive strength and elastic modulus of the material is sensitive to

water content. As shown in Figure 5, the compressive strength and the elastic modulus
are highly affected by the RH imposed to the specimens. For B6 specimens, the
compressive strength increases from 22.4 MPa at high levels of RH to 45.1 MPa at low
levels of RH, an increment of 2.01. B15 specimens present an increment of the peak
stress of 2.1.
In the two groups of specimens, elastic modulus illustrates a downward trend
according as the RH increases. For B6 specimens, the elastic modulus increases from
1754 MPa (RH=36%) to 3216 MPa (RH=94%). The ratio E(RH=36%)/E(RH=94%) is
1.83 and 1.53 respectively for B6 and B15 specimens. These results confirm the
influence of water content on the mechanical behavior of argillaceous rocks.
Figure 5. Compressive strength and elastic modulus versus relative humidity (RH)
The Monitoring of Acoustic Emission (AE) has been proven as a powerful tool to
track the process of failure of materials subjected to mechanical load [9]. It was
considered to study some features of the brittle behavior of the material. Specifically,
we use the cumulative hits to identify the crack initiation stress threshold (CI) and
crack damage stress threshold (CD). For this, Diederichs et al. [10] state that these
stresses can be determined from cumulative hits versus axial stress log-log plots. CI is
the first point where the rate of crack emissions suddenly increases with a small change
in load. Besides, CD is where the second sudden increase in the slope of the AE curve
occurs (Figure 6.a.). On the other hand, CI and CD can be also obtained from the axial
stress versus volumetric strain plot. Bieniawski [7] defines CI as the stress level that
limits the linear axial stress versus volumetric strain behavior and CD as the stress
level where the volumetric strain is reversed (Figure 6.b.).
Figure 6. Identification of stress thresholds a.) from AE hits b.) from volumetric strains [10]
Figure 7 presents the identification of the stress thresholds according to the

methodology described in Figure 6 from one sample conditioned at RH=36%. Figure 8
reports values of crack initiation threshold (CI) and crack damage threshold (CD)
determined from volumetric strains and cumulative AE activity for different RH values.
As illustrated in Figure 8, stress thresholds fall with RH and present a satisfactory fit
(Fig. 8.a.). Normalized data with the peak stress of each specimen reveal less influence
of RH. For the different hydric states, CI appears approximately at 20% of peak stress
while CD show large variability and occur between 60% and 80% of the rupture stress
(Fig. 8.b.).
Figure 7. Identification of stress thresholds a.) from AE hits b.) from volumetric strains
For the second experiment, Figure 9.a displays a reduction in the compressive
strength of specimens conditioned at high relative humidity (46%) compared with
specimens at low relative humidity (36%). In this case, a reduction between 5% and
15% is observed. Regarding volumetric strains (Figure 9.b), these preliminary results
show higher volumetric strains at higher levels of RH as an evidence of stiffness
reduction. Also, the reversal point of volumetric strains appears at the lower level of
stress in the case of specimens at higher values of RH. This feature indicates that crack
damage and unstable cracking appears before in samples with higher water content.
Figure 8. Crack initiation threshold (CI) and crack damage threshold (CD) for different levels of RH.
Figure 9. Stress-strain curves (second experiment).
Figure 10. Moving point regression analysis of axial stress-strain data for different levels of RH
We use the moving point regression technique to improve the analysis of strain
gage data. This due to the high degree of error and subjectivity in selecting particular
points or slopes in the stress-strain curves. This technique uses a sliding window
approach to move through an x, y data set, fitting a straight line over a user-defined
interval. The slope at each point is calculated over the interval and recorded, the
process being repeated at successive points [8]. Figure 10 illustrates the application of
the technique to obtain the axial average stiffness. In all cases, a reduction in the
average axial stiffness with the increment of relative humidity can be noticed.
4. Conclusions
In this paper, we analyze the influence of relative humidity on the mechanical behavior
of an unsaturated argillaceous rock. Preliminary results confirm the role of water
content in some mechanical and elastic properties of unsaturated rocks. The elastic
modulus and compressive strength increase as relative humidity decreases.
Regarding volumetric strains in specimens with higher water content, they exhibit
higher contractancy until the reversal point of volumetric strains. On the other hand, the
average axial stiffness shows a reduction in samples at higher values of relative
humidity.
Acoustic emission activity is used to identify the crack initiation and the damage
stress thresholds. Despite the limited results, this feature is of great interest for linking
the influence of rock moisture on the mechanical properties of argillaceous rocks and
their process of failure. However, further tests are needed to confirm the influence of
relative humidity on the pattern of acoustic emission.
References
[1] J. A. Pineda, "Swelling and degradation of argillaceous rocks induced by relative humidity
effects: an experimental study.," Ph.D., Universitat Politcnica de Catalunya, Spain, 2012.
[2] A. Gens, "On the hydromechanical behaviour of argillaceous hard soils-weak rocks," presented at
the 15th European Conference on Soil Mechanics and Geotechnical Engineering, Athens, Greece,
2013.
[3] Q. T. Pham, F. Vales, L. Malinsky, D. Nguyen Minh, and H. Gharbi, "Effects of desaturation
resaturation on mudstone," Physics and Chemistry of the Earth, Parts A/B/C, vol. 32, pp. 646-655,
2007.
[4] F. Vals, D. Nguyen Minh, H. Gharbi, and A. Rejeb, "Experimental study of the influence of the
degree of saturation on physical and mechanical properties in Tournemire shale (France),"
Applied Clay Science, vol. 26, pp. 197-207, 8// 2004.
[5] D. Yang, M. Bornert, S. Chanchole, L. Wang, P. Valli, and B. Gatmiri, "Experimental
investigation of the delayed behavior of unsaturated argillaceous rocks by means of Digital Image
Correlation techniques," Applied Clay Science, vol. 54, pp. 53-62, 11// 2011.
[6] A. Lisjak, "Investigating the Influence of Mechanical anisotropy on the Fracturing Behaviour of
Brittle Clay Shales with Application to Deep Geological Repositories," PhD Thesis, Graduate
Department of Civil Engineering, University of Toronto, Toronto, Canada, 2014.
[7] Z. T. Bieniawski, "Mechanism of brittle rock fracture. Part I. Theory of the fracture process.,"
International Journal of Rock Mechanics and Mining Sciences & Geomechanical Abstracts, vol. 4,
pp. 395-406., 1967.
[8] E. Eberhardt, D. Stead, B. Stimpson, and R. S. Read, "Identifying crack initiation and propagation
thresholds in brittle rock," Canadian Geotechnical Journal., vol. 35, pp. 222-233., 1998.
[9] C. A. Tang and J. A. Hudson, Rock Failure Mechanisms: Illustrated and Explained: CRC Press,
2010.
[10] M. S. Diederichs, P. K. Kaiser, and E. Eberhardt, "Damage initiation and propagation in hard rock
tunnelling and the influence of near-face stress rotation," International Journal of Rock
Mechanics & Mining Sciences, vol. 41, pp. 785-812, 2004.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-350
Effect of Cement Type on the Mechanical

Behavior of Fiber Reinforced sands
Navid NIKZADa and Amir HAMIDI a,1
a
School of Engineering, Kharazmi University, Tehran, Iran
Abstract. Early attempts to increase the soil strength by mixing it with other
materials date back 3000 years ago in construction of ziggurats when the soil was
mixed with plant roots. Coupled usage of cement and fiber in soil, however, refers
to more recent years. Reinforcement with cement and fiber is one of the soil
improvement methods for increase in bearing capacity and reduction of
settlements. Cemented soils are usually found naturally, however, artificially
cemented samples are widely used in the laboratory to investigate deformation and
shear strength characteristics due to the limitations in undisturbed sampling. In
present study, series of conventional triaxial tests were carried out to investigate
the effects of cement type on the mechanical behavior of fiber reinforced sand.
Accordingly, Portland cement type (II), hydrated lime and gypsum plaster were
used as the cementing agents. The cement content was 3% (dry wt.) of the base
soil. Fibers 0.012 m in length and 23 m thicknesses were added at 0.0%, 0.5%
and 1.0% (dry wt.) of the sand-cement mixture. Samples were prepared at 70%
relative density and triaxial tests were performed under confining pressures of 100,
200 and 300 kPa. Samples cemented with Portland cement were cured for 7 days
in humid room, however, samples with hydrated lime cured for 7 days in a
constant temperature about 40C and samples with gypsum cured for 24 hr in a
constant temperature about 50C. Samples with Portland cement and hydrated
lime were tested in saturated condition but gypsum cemented ones were tested in
dry condition. The results of experiments showed that fiber inclusion increases
peak and residual strengths, axial strain at failure and energy absorption for all
cement types. For the mentioned curing condition, samples cemented with
Portland cement indicated the most shear strength and samples with hydrated lime
showed the least value. Also, fiber addition was more effective on the peak shear
strength of samples cemented with hydrated lime and had the least influences on
the strength of gypsum cemented samples.
Keywords. Cemented sand, Fiber reinforcement, Shear strength, Triaxial test,

Volume change
1. Introduction
Despite the complexity and diversity of soil, it is used in engineering works worldwide.
Although usually found in soils provided for the basic needs of a project that are
technically or economically are not affected. The construction of buildings and other
structures on poor soil is very dangerous because its settlement is too due to low shear
strength and high compressibility. Thus, soil improvement is a basic work in
engineering- special civil-projects. There are many methods to improve soil
1
Corresponding author. [email protected]
N. Nikzad and A. Hamidi / Effect of Cement on Fiber Reinforced Sands 351
characteristics; one of them is mixing the soil with cement and fiber. Early attempts to
increase the soil strength by mixing it with other materials date back 3000 years ago in
construction of ziggurats when the soil was mixed with plant roots. Coupled usage of
cement and fiber in soil, however, refers to more recent years. Reinforcement with
cement and fiber is one of the soil improvement methods for increase in bearing
capacity and reduction of settlements. Cemented soils are found in nature also can be
made artificially, widely used in to improve deformation and shear strength
characteristics of soil. Several researchers have reported the effect of coupled usage of
cement and fiber on the mechanical behavior of sandy soils. Maher and Ho (1993)
conducted triaxial tests and showed the increase in peak shear strength, energy
absorption and brittleness index by fiber inclusion and increase of fiber length. Consoli
et al. (1998) showed that increase in the peak strength of sand and cemented sand with
fiber inclusion, although the effect of fiber reinforcement was more visible in sand
compared to cemented sand. Consoli et al. (2009) reported that fiber inclusion was
more effective on the peak strength of cemented sand for small cement contents while
it was more effective on the residual strength of sand cemented with higher cement
contents. Park (2009, 2011) examined unconfined compressive properties of fiber
reinforced sand and showed that increase in fiber content and the layers reinforced with
fiber result in the increase of unconfined compressive strength of cemented sand. Silva
dos Santos et al. (2010) proved more extension and breakage of the fibers recovered
from the cement-fiber reinforced samples under isotropic compression loading which
confirmed more anchorage of fibers into the cement during tensile failure mode. Liu
and Starcher (2012) studied the effect of curing conditions on the strength of cement-
fiber reinforced soft soil. Hamidi and Hooresfand (2013) examined the effect of fiber
reinforcement on triaxial shear behavior of cement treated sand and represented the
increase in peak and residual strength, energy absorption and principal stress ratio and
decrease in initial stiffness and brittleness index by fiber inclusion. Also it was showed
that the effects of fiber inclusion reduce with increase in confining pressure or decrease
in relative density.
As stated above, both fiber and cement inclusion increase the shear strength of
sandy soils, however, review of the technical literature shows that impact of fiber
inclusion is not clear for different cement types. In present study, conventional triaxial
compressive tests is used to investigate the mechanical behavior of cement-fiber
reinforced sand using Portland cement, hydrated lime and gypsum as the cementing
agent. It is tried to illustrate the differences in behavior and their mechanisms with
special attention to the cement type effects.
2. Materials and Test Procedure
Clean and uniform Babolsar sand was used as the base material in this study. Portland
cement, hydrated lime and gypsum was used as cement agents in this study. Sample
cemented with different cement types were cured in different methods. Samples
cemented with Portland cement were cured for 7 days in a humid room. Samples using
hydrated lime cured 7 days in plastic bag and a constant temperature of 40 C and
samples using gypsum cured 24 hr in a constant temperature of 50 C to completely dry,
All samples cemented with gypsum were tested under complete dry condition duo to
the reduction of gypsum strength with increase in water content (Haeri et al. 2005).
352 N. Nikzad and A. Hamidi / Effect of Cement on Fiber Reinforced Sands
White monofilament polypropylene fibers with circular cross section, average length of
0.012 m, diameter of 23 m and aspect ratio of about 500.
In this study conventional consolidated-drained triaxial tests were used. Samples
cemented with Portland cement and hydrated lime tested under saturate condition, but
Samples cemented with gypsum were tested under dry conditions. As a result, the
change in volume during consolidation and axial loading was measured using a sensor
on the cell pressure line which was controlled at the back pressure considered for each
test. It should be said that in this method all effective factors on the volume change,
except the volume change during consolidation or axial loading were calculated and
subtracted from measured volume change values. Haeri et al. (2005) reported that
results of saturated and dry tests were compatible with maximum difference of 5% in
stress-strain and volume change behaviour. The water drained from samples in
saturated tests is the same as water flowed from triaxial cell in dry tests. As a result,
this method was considered to calculate volume changes in dry samples and data could
be used for comparison of the volume change behaviour of cemented sand under
saturated and completely dry conditions
3. Results
3.1. Parameter definition
In order to analyze the results of triaxial tests, a number of parameters like , , , e,

, q, were used. The effective major and minor stresses are named as and ,
respectively, are the axial strains, e is the void ratio, specific volume (), deviatoric
stress (q) and mean effective stress ( ) are defined as below:
=1+e (1)
q= - (2)
=( +2 )/3 (3)
The net impact of fiber inclusion on the shear strength of cemented sand is also
investigated using the following new-defined index:
IFI=(q(c+f)-qc)/qc 100% (4)
Where, q(c+f) is the peak deviatoric stress of cement-fiber reinforced sand and q c is
the peak deviatoric stress for cemented treated sand. This index shows the net impact of
fiber inclusion (IFI) on the shear strength of a cement treated sand. Using this
parameter allows us to investigate the net impact of fiber reinforcement for different
cementing agents involved in present study.
Increase in deformability of the samples due to the fiber inclusion was also
determined using deformability index as follows (Park, 2011):
D=fiber/no-fiber (5)
Wherre fiber is the axial strain associated

a to the maximum m shear strenggth of fiber
reinforced
d cemented san and and no-fibeer is the axial strain
s at peak deviatoric streess of non-
fiber rein
nforced cemeented sand. This index was also ussed to invesstigate the
deformabiility of fibers anchored in different
d cemennt types.
3.2. Stresss strain and vo

olume changee behavior
In presentt study, samp ples cemented d with differeent cement tyypes experiennced brittle
failure moode. For all fib
ber contents and
a confining pressures, thee samples cem mented with
Portland cement
c showeed the most innclination. Thee inclination of
o the shear bband varied
in a narroow range of 454 to 67 whicch confirms little influence of fiber conttent on the
friction anngle of cemeented soil. Th he width of shear band was w minimum m for lime
cementatio on and variedd between 2.2 to 5.8 mm as fiber contentt increased froom 0.0% to
1.0% as shown
s in Figuure 1. Howev ver, it was maaximum for gyypsum cemennting agent
and chang ged from 3.8 to 7.7 with increase
i in fib
ber content frrom 0.0% to 1.0%. The
lower shear band width h associated to
o the lime cemmenting agentt can be attribbuted to the
more pulllout resistancee of fibers annchored in lim me cement com mpared to othher cement
types. Sh hear band width
w was thhe average values
v of twwo methods, laboratory
measurem ments and geo ometrical calcculations for each
e sample. Shear band width was
measured after each tesst using a dig gital ruler in several points of shear areaa of sample
(laboratorry measurement). After modification of volume
v changge data based on ASTM
instructionn, volume change at the critical
c state was
w divided to t the estimatted area of
shear planne to achieve mean shear band b width (ggeometrical calculation).
c Innformation
presented in figure 1 are the aveerage of the mean laboraatory shear bband width
measurem ments and meaan shear band width
w from geeometrical callculations.
Figure 1. Variation
V of shear band width with confining pressu
ure for different ceement type and ddifferent fiber
content.
In ordder to consideer the effects of fiber inclu usion on stresss-strain and volumetric
behavior of samples with w differentt cement typees, deviatoricc stress-axial strain and
specific volume-
v cuurves of cem mented and cement-fiber
c r
reinforced saamples are
depicted ini Figure 2 for fo a confining g pressure off 100 kPa. Alll specimens sshowed an
apparent peak
p point (qmax
m ) at an axiall strain level lower
l than 10% %. After that,, deviatoric
stress dropps until its ulttimate value inn large axial strain
s about 200% (qres). Thee change in
specific volume
v () veersus mean efffective stresss (p') in conso olidated drainned triaxial
tests are considered
c in Figure 2 for a confining prressure of 100 0 kPa. All sam mples show
dilative behavior after a little initiaal contraction n. It is indicaated that fiberr inclusion
increases the value of associated
a meean effective stress
s to the end
e of contracction. Also,
samples cemented
c witth gypsum haave the leastt initial contrraction and thhe highest
dilation and
a samples with w Portland cement own n the highest initial contracction. It is
indicated that
t samples cemented
c withh Portland cem ment has the highest
h p' at faailure point.
Figure 2. Effect of cemeent type and fiberr inclusion on the stress-strain and volume change bbehavior.
According g to Figure 22, samples cem mented with Portland cem ment showed tthe highest
peak sheaar strength vallues. Sampless cemented with w gypsum showed the loowest shear
strength and
a the highest volume chan nge which is presented
p heree by the specific volume
() at critiical state. Sam
mples cementeed with lime indicated
i the lowest volum me changes.
The same trend was observed for all fiber contents and confinin ng pressures thhat used in
present stu udy.
In ordder to investiggate the net im
mpact of fiberr inclusion forr different cem ment types,
IFI index that was intro oduced in sectiion 3.1 is calcculated for diffferent fiber coontents and
confining pressure as shown in Figure 3. Figurre 3 (a), (b),, (c) indicate that fiber
reinforcemment has the mostm influencce on the peak k shear strenggth of sampless cemented
with lime and the minim mum effect on n the peak sheear strength off samples cem mented with
gypsum. TheT shear streength of lime treated sand increased
i betwween 50% to 125% with
0.5% fibeer inclusion an nd between 100% to 200% % with 1.0% fiberf reinforceement. For
gypsum cementation,
c however, 40 0% to 60% increase
i in the
t shear strength was
observed with 0.5% fiber f inclusion while it was w 70% to 120% with 11.0% fiber
reinforcemment. This is in i agreement with the loweest shear band d width measuured for the
lime cem menting agent in Figure 1. 1 The strong ger pull out mechanism and better
anchoragee of fibers forr this cement type result in n the more in ncrease in shear strength
compared d to the other cementing agents.
a It is evident
e from the figures thhat the net
impact off fiber reinforcement on the shear streng gth of cementt treated sandd decreases
with the inncrease in con nfining pressu
ure for all cement types.
Figure 3. Net im
mpact of fiber incllusion on the peak
k shear strength of
o cemented sandd.
3.3. Energ
gy absorption and deformab
bility
Energ
gy absorption
n is identified
d as area below the stresss-strain curvee and it is
related to the amount of energy reequired to ind duce deformaation in fiber reinforced
orption at 10%
materials. Energy abso % axial strain for 100 kPa confinement calculated
and reporrted in Table 1. A comparrison among resultsr shows that sampless cemented
with Porttland cement and reinforcced with fiberr have the most m energy aabsorption,
although samples cem mented with hydrated
h limee and gypsum m are almost the same.
Energy ab bsorption is reelated to the am
mount of enerrgy required to
t induce defoormation in
fiber reinfforced materiaals. So, more energy
e requireed to begin to deformation oof fibers in
samples cemented with h Portland cem ment.
( m 3)
Table 1. Eneergy Absorption (kJ/
Fib
ber content (%)
Cement ty
ype
0.0 0.5 1.0
Portland ceement 50 112 163
Hydrated lime 37 88 119
Gypsum 40 82 128
Deforrmability indeex as defined by

b Park (2011) was used to describe the dductility of
fiber reinfforced cementted sand. Figuure 4 shows the
t variation of D-index foor different
cement tyypes. Results show that saamples cemen nted with Porrtland cemennt have the
maximum m D-value and d samples cem mented with gypsum
g have the least onee, this is in
agreementt with results of energy ab bsorption. Theese results aree indicated thhat samples
cemented with Portland d cement are more
m ductile and
a deformablle than two othher cement
types.
Figure 4. Vaariation of D-indeex with fiber conttent for different cement types.
usions
4. Conclu
The highest sh
T hear band widdth occurred in
n samples cem
mented with gyypsum and
th
he lowest shear band width
h in samples ceemented with lime.
Samples cemented with Portland cement showed the highest peak shear
strength. Samples cemented with gypsum showed the lowest shear strength
and the highest volume change at critical state and samples cemented with
lime indicated the lowest volume changes at critical state.
The greatest net impact of fiber reinforcement on peak shear strength was
found for samples cemented with lime and the lowest effect of fiber on the
peak shear strength was found for samples cemented with gypsum.
Fiber reinforced samples cemented with Portland cement showed the highest
energy absorption values. This trend was observed for all confining pressures
that used in this study.
The results showed maximum deformability index for the samples cemented
with Portland cement and minimum values for the samples cemented with
gypsum. It is in agreement with results of energy absorption.
Abbreviations
dry wt. Dry weight

hr Hour
FC Fiber content
CP Confining pressure
References
[1] M.H. Maher, Y.C. Ho, Behavior of fiber-reinforced cemented sand under static and cyclic loads,
Geotechnical Testing Journal 16 (1993), 330-338.
[2] N.C. Consoli, P.D.M. Prietto, L.A. Ulbrich, Influence of fiber and cement addition on behavior of sandy
soils, Journal of Geotechnical and Geoenvironmental Engineering 124 (1998), 1211-1214.
[3] S.M. Haeri, A. Hamidi, N. Tabatabaee, The effect of gypsum cementation on the mechanical behavior of
gravely sands, Geotechnical Testing Journal, 28 (2005), 380-390.
[4] N.C. Consoli, M.A. Vendruscolo, A. Fonini, F. Dalla Rosa, Fiber reinforcement effects on sand
considering a wide cementation range, Geotextiles and Geomembranes, 27 (2009), 196-203.
[5] S.-S. Park, Effect of fiber reinforcement and distribution on unconfined compressive strength of fiber-
reinforced cemented sand, Geotextiles and Geomembranes, 27 (2009), 162-166.
[6] S.-S., Park, Unconfined compressive strength and ductility of fiber-reinforced cemented sand,
Construction and Building Materials, 25 (2011), 1134-1138.
[7] A.P. Silva dos Santos, N.C. Consoli, K.S., Heineck, M.R. Coop, High-pressure isotropic compression
tests on fiber-reinforced cemented sand, Journal of Geotechnical and Geoenvironmental Engineering,
136 (2009), 885-890.
[8] C. Liu, R. Starcher, The effect of curing conditions on unconfined compressive strength of cement- and
cement-fiber-improved soft soils. Journal of Materials in Civil Engineering, 25 (2012), 1134-1141.
[9] Hamidi, M. Hooresfand, M. Effect of fiber reinforcement on triaxial shear behavior of cement treated
sand, Geotextiles and Geomembranes 36 (2013), 1-9.
[10] ASTM, 1998. Annual Book of ASTM Standards: Soil and Rock Division. West Conshohocken,
Philadelphia.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-358
Relationship Between
Undrained Shear Strength and
Shear Wave Velocity for Clays
Shehab S. AGAIBYa,1 and Paul W. MAYNE a
a
Georgia Institute of Technology, Atlanta, United States
Abstract. The interrelationship between undrained shear strength (s u) and

downhole shear wave velocity (VsVH) of normally consolidated (NC) and lightly
overconsolidated (LOC: OCR < 2) to overconsolidated (OC) to highly
overconsolidated (HOC: OCR > 10) clays is investigated in the presented study.
The main objective of this research program is to develop a worldwide database of
high quality in-situ geophysical and laboratory strength data from thirty seven
well-documented geotechnical sites from locations in Australia, Brazil, Canada,
China, Italy, Japan, South Korea, North Sea, Norway, Singapore, Sweden,
Thailand, United Kingdom, USA, and Vietnam. The study includes undrained
shear strength measurements on undisturbed samples of normally to lightly
overconsolidated intact to overconsolidated and fissured clays using
anisotropically-consolidated triaxial compression tests (CAUC). Shear wave
velocities were measured in the field by downhole tests (DHT), in many cases via
seismic piezocones (SCPTu). Analyses of the compiled database found
approximate trends between undrained shear strength and shear wave velocity.
Tentative correlations are explored by including other various parameters such as
Atterberg limits, void ratio, overconsolidation ratio (OCR), and effective vertical
stresses. The correlative trends may aid geotechnical engineers in helping to assess
su profiles in clay deposits in preliminary investigations and as an independent
method in collaboration with sampling, lab testing, and other field data.
Keywords. Clays, Downhole Test, OCR, Shear Wave Velocity, Triaxial Tests,
Undrained Shear Strength
1. Introduction
The profiling of in-situ shear wave velocity provides a means to evaluate several soil
parameters such as stiffness, unit weight and stress history [1]. For the purpose of the
current study, the focus is on the estimation of undrained shear strength (su) from shear
wave velocity (Vs) that is measured in the field using either in-situ tests or geophysical
methods. Accurate and detailed in-situ Vs measurements are critical and essential for
the fields of geotechnics and geophysics and can be utilized to provide an acceptable
estimate of the undrained shear strength without the need to use low quality disturbed
1
Corresponding Author: Geosystems Engineering Group, School of Civil and Environmental
Engineering, Georgia Institute of Technology, 790 Atlantic Drive, Atlanta, Georgia, USA 30332-0355
Email: [email protected]
S.S. Agaiby and P.W. Mayne / Undrained Shear Strength and Shear Wave Velocity for Clays 359
samples and/or rely on uncertain results from laboratory tests affected by sample
preparation or reconsolidation procedures.
2. Undrained Shear Strength Measurement
Undrained shear strength (su) is considered one of the most critical parameters in
geotechnical engineering practice and design. The measured undrained shear strength is
affected by many variables, including: initial stress state (isotropic or K 0 consolidation),
strength anisotropy, loading direction (compression or extension), and boundary
conditions (plane strain or triaxial) defining the shearing mode in the test. Other
influencing factors include: strain rate, sensitivity, ageing, inherent fabric anisotropy,
strain compatibility, thixotropy, and specimen quality associated with sampling
disturbance [2]. A comparison of different testing methods in the lab: Ko consolidated
triaxial compression & extension tests (CK0UC & CK0UE), direct simple shear test
(DSS) and in the field: field vane test (FV) for measuring su is given in Figure 1(a) that
shows the significance of shearing mode for soft clay at the national geotechnical
experimentation site of Bothkennar, UK [3]. At a depth of 6 m, for instance, su varies
by a factor of over 3 for these four modes.
3. Shear Wave Velocity Measurement
A key parameter for static and dynamic geotechnical analyses is the small-strain shear
modulus (Gmax) which represents the initial fundamental stiffness of soils and rocks.
The shear modulus is obtained from shear wave velocity (Vs) and the total mass density
(), where: Gmax = G0 = .Vs2. The value of Gmax can be determined using either in-situ
and/or laboratory methods. Laboratory methods include testing undisturbed soil
specimens using resonant column, torsional shear, bender elements, shear plates, non-
resonance methods, and/or triaxial tests with local strain measurements. Lab-based
techniques have several shortcomings; including sampling disturbance that may break
weak inter-particle bonds, stress relief, in addition to ageing, diagenesis, and
depositional process effects; that could affect small-strain stiffness measurements.
Hence, Vs is most reliably measured in-situ rather than in the laboratory. The shear
wave has directional and polarized characteristics, multiple types of V s can be
measured: Vsij where i is the direction of propagation and j is the direction of
polarization (VsVH, VsHV, and VsHH) where subscripts V is for vertical and H is for
horizontal. For in-situ measurements, the most common shear wave mode is obtained
by downhole testing DHT (ASTM D 7400) measuring the VsVH mode (the focus of
current study), which is carried out in either drilled-cased-grouted boreholes or by
seismic CPT or seismic DMT. By comparing different in-situ methods for Vs, it is
obvious that the specific methods provide different resolutions for Vs measurements at
different depths. To illustrate, Figure 1(b) shows profiles of measured VsVH, VsHV, VsHH,
and Vs measured using Rayleigh waves (VsRW) in soft clay at Bothkennar, UK [3].
By examining prior correlations reported by researchers, it is evident that there is

no unique relation between undrained shear strength and shear wave velocity. Hence, a
global study was undertaken that includes data from well documented clays.
360 S.S. Agaiby and P.W. Mayne / Undrained Shear Strength and Shear Wave Velocity for Clays
(a) (b)
Figure 1. (a) Profiles of four different undrained shear strength modes (b) Profiles of four
different shear wave velocity modes in Bothkennar soft clay, UK [3]
4. Compiled suTC and VsVH Database
For the current comprehensive study, a special database has been carefully collected
from a total of 37 well-documented worldwide geotechnical test sites. The sites mostly
include soft to firm young to aged clays, but also consider stiff to hard clays, and
fissured fine-grained soils. For the undrained shear strength, only high-end laboratory
tests were included with a focus on the triaxial compression mode, designated suTC.
Regarding the shear wave velocity, only in-situ data were compiled using downhole
shear wave velocity, VsVH which is primarily measured using DHT, SCPT, and SDMT.
At each of the studied sites, complementary data included in-situ void ratio (eo) and
one-dimension consolidation tests for stress history profiles that were obtained from
published sources with detailed understanding of the site-specific geologic conditions.
In addition, laboratory data such as water content, Atterberg limits, and unit weight
were collected. The compiled database can be classified into two main categories: (1)
soft to firm NC to LOC clays that have 1 < OCRs < 2.5. These are presented in Table 1
and include 31 clay sites with a total 360 CAUC or CKoUC tests with their
corresponding VsVH measurements; (2) overconsolidated (OC) clays that have OCRs >
2.5. The latter are listed in Table 2 with 6 clay sites with a total of 60 CAUC or CK0UC
tests along with their corresponding VsVH values at same elevations.
5. Relationship between Undrained Shear Strength and Shear Wave Velocity
By considering the geotechnical sites from Table 1, Figure 2 shows the mean trend plus
or minus 2 standard deviations between suTC (measured in kPa) and VsVH (measured in
m/sec) for NC to LOC clays to be:
suTC 0.152 0 VsVH

1.14
(1)
Table 1. Summary of Compiled suTC and VsVH Database for Intact Normally Consolidated (NC) to
Lightly Overconsolidated (LOC) Clays
Site + Location Soil Type Reference Sources + Index

Amherst, MA, USA Soft Varved DeGroot & Lutenegger (2003) [4]; Hegazy (1998) [5]
Ariake, Japan Soft Clay Tanaka, et al. (2001) [6]
Backebol, Sweden Postglacial Larsson(1980) [7]; Larsson& Mulabdic (1991) [8,9]
Ballina, Australia Soft Estuarine Pineda, et al. (2014) [10]
Belfast, Ireland Soft Clay Silt Lehane (2003) [11]
Bothkennar, UK Soft Silty Clay Hight, et al. (2003) [3]
Burswood, Australia Soft Clay Low, et al. (2011) [12]
Busan, Korea Soft Clay Chung, et al. (2011 & 2012) [13 & 14]
Chicago, IL, USA Glacial Clay Chung & Finno (1992) [15]; Finno et al (2000) [16]
Fiumicino, Italy Soft Clay Cavalera & Scarpelli (1980) [17]; Monaco (2007)[18]
Hai-Phong, Vietnam Soft Alluvial Watabe, et al. (2004) [19]
Islais Creek, CA, USA Soft Bay Mud Pestana, et al. (2002) [20]
Kurihama, Japan Alluvial Tanaka (1995) [21]; Shibuya & Tanaka (1996) [22]
Lake Bonneville, UT, USA Sensitive Clay Garner (2007) [23]& Cole (2003) [24]
Lianyungang, China Soft Clay Liu, et al. (2008) [25]
Lierstranda, Norway Sensitive Clay Lunne & Lacasse (1999) [26]
Lilla Mellosa, Sweden Soft Organic Larsson & Mulabdic (1991 a, b) [8, 9]
Louiseville, Quebec, Canada Champlain Sea Leroueil (2003)[27]; Yafrate & DeJong (2006) [28]
Newbury, MA, USA Soft Silty Clay Landon (2007) [29]
Nong Ngu Hao, Thailand Soft Bangkok Shibuya & Tamrakar (2002)[30]
Norrkoping, Sweden Varved Clay Kennet (1994) [31]; Larsson & Mulabdic (1991)[8, 9]
Northwestern Univ. IL, USA Soft Clay Finno, et al (2000) [16]
Onsoy, Norway Soft Marine Lunne, et al. (2003) [32]
Pentre, UK Silt Deposit Lambson, et al. (1993) [33]
Sarapui, Brazil Very Soft Clay Ortigao (1983) [34]; Almeida & Marques (2003) [35]
Saro Road 6/900, Sweden Soft Organic Larsson & Mulabdic (1991 a, b) [8, 9]
Saro Road 7/600, Sweden Organic Clay Larsson & Mulabdic (1991 a, b) [8, 9]
Singapore Clay, Singapore Soft Marine Tanaka, et al. (2001) [6]
South Gloucester, Canada Sensitive Leda Bozozuk (1972)[36]; Yafrate & DeJong (2006) [28]
Sutthisan, Thailand Soft Bangkok Shibuya & Tamrakar (2002) [30]
Troll, North Sea Soft Clay Lunne, et al. (2007) [37]
Table 2. Summary of Compiled suTC and VsVH Database for Fissured and Overconsolidated (OC) Clays
Site + Location Soil Type Reference Sources + Index

Beaumont Clay, TX, USA Stiff Fissured O'Neill &Yoon(1995) [38]; Mahar&O'Neill (1983)[39]
Brent Cross, UK Fissured-HOC Hight, et al. (2003) [40]
Heathrow - T5, UK London Clay Hight, et al. (2003) [40]
Montalto di Castro, Italy Stiff OC Clay Jamiolkowski, et al. (1982 & 1994) [41 & 42]
Pisa Clay, Italy Pancone Clay LoPresti, et al. (2003) [43]
Port of Anchorage, AK, USA Stiff Clay Zapata-Medina(2012)[44]; Mayne & Pearce(2005)[45]
Similarly, by investigating the additional geotechnical sites listed in Table 2,

Figure 3 shows the studied relation for all the clays under study, ranging from NC to
LOC obtained from Table 1 to OC and/or fissured HOC clays, the relationship between
suTC (measured in kPa) and VsVH (measured in m/sec) can be expressed as:
suTC 0.0672 0 VsVH

1.33
(2)
Figure 2. Triaxial compression undrained shear strength vs. downhole shear wave velocity for
NC to LOC intact clays in arithmetic scale.
1000
Fissured - OC Clays
Undrained Shear Strength, su TC (kPa)
with su (kPa) and Vs (m/s)

R2= 0.795
N = 37
100 n = 420
Intact Clay - Table 1

10 Beaumont Clay
Brent Cross
Montalto di Castro
Pisa
Port of Anchorage
T5 - London Clay
1
10 100 1000
Shear Wave Velocity, Vs VH (m/s)
Figure 3. Triaxial compression undrained shear strength vs. downhole shear wave velocity for
intact NC, LOC, to OC and HOC fissured clays in logarithmic scale.
6. Generalized Expression for suTC and VsVH
To gain a more comprehensive understanding for the studied relationship, a number of

additional parameters were investigated such as stress history, void ratio, Atterberg
limits, and effective vertical overburden stresses. In-situ shear wave velocity can be
used as an index for the magnitude of preconsolidation stress of clays, where void ratio
decreases when the clay deposit is subjected to higher pre-stressing leading to an
increase in the magnitude of shear wave velocity [46]. According to basic expressions
calculating the small strain shear modulus and undrained shear strength from the
literature, it is evident that these parameters are both dependent on the stress history
profile, expressed in the form of OCR.
Multiple regression analyses were used to combine different parameters and
investigate their significance for the suTC versus VsVH relations that produced a
generalized expression. Figure 4 compares the actual measurements of su (in kPa) to the
estimated values using VsVH (in m/sec), e0, OCR, PI (in %), and effective vertical stress
(in kPa). The generalized expression obtained can be written as:
suTC 0.038VsVH
1.083
PI 0.14 OCR 0.31e0
0.07

' 0.23
vo (3)
300
Best Fit Line (b=0)
Lab Measured Undrained Shear Strength, su (kPa)
y = 1.098x
n = 362
250
R = 0.865
200
150
100
Least Square Regression:

50 y = 1.172x - 6.08
n = 362
R = 0.870
0
0 50 100 150 200 250 300
Predicted Undrained Shear Strength, su* (kPa)
Figure 4. Measured suTC vs. predicted value from VsVH for NC and OC clays using generalized
expression.
By considering the results of multiple regression analyses, it can be noted that the
predicted values for fissured and HOC clays from Table 2 do not agree well with the
actual measurements, as in the case of intact and LOC clays. It can be also concluded
that using more than three variables at a time (effective stresses, e, OCR, and PI) did
not provide sufficiently good enough statistical results for correlations where low
significance level for the void ratio was observed, which is also expressed by the small
exponent value. This can be attributed to the redundant effects between void ratio and
OCR (or vo') such that a single variable dominates. When less variables (PI, effective
stress and either e0 or OCR) are involved, the overall correlations are improved albeit
some outliers, discrepancies, and uncertainties still remain.
7. Conclusions
A careful compiled database study was analyzed to produce a general expression for
evaluating undrained shear strength (su) from downhole shear wave velocity (VsVH).
The data were collected from 37 worldwide well-documented natural soils primarily
covering NC to LOC intact soft to firm clays, but also included results from a few OC
to HOC fissured clays. The study focused primarily on benchmark su values obtained
from high-level laboratory triaxial compression tests (CAUC) and in-situ downhole
shear wave velocity measurements from downhole tests in cased boreholes or seismic
piezocone tests.
References
[1] P.W. Mayne, Integrated ground behavior: In-situ and lab tests, Deformation Characteristics of
Geomaterials, 2 (Proc. Lyon, France), Taylor & Francis, (2005): 155-177.
[2] P.W. Mayne, M.R. Coop, S. Springman, A-B. Huang, and J. Zornberg, State-of-the-Art Paper (SOA-1):
GeoMaterial Behavior and Testing. Proc. 17th Intl. Conf. Soil Mechanics & Geotechnical Engineering,
4 (ICSMGE, Alexandria, Egypt), Millpress/IOS Press Rotterdam: (2009): 2777-2872.
[3] D.W. Hight, M.A. Paul, B.F. Barras, J.J.M. Powell, D.F.T. Nash, P.R. Smith, R.J. Jardine, and D.H.
Edwards. The characterisation of the Bothkennar clay. Characterisation & Engineering Properties of
Natural Soils, 1, Balkema, Rotterdam, (2003): 543-597.
[4] D.J. DeGroot & A.J. Lutenegger, Geology and engineering properties of Connecticut Valley varved
clay, Characterization & Eng. Proprieties of Natural Soils 1, Balkema, Rotterdam, (2003): 695724.
[5] Y.A. Hegazy, Delineating Geostratigraphy by Cluster Analysis of Piezocone Data, Doctoral
Dissertation, School of Civil & Environmental Engineering, Georgia Institute of Technology, 1998.
[6] H. Tanaka, J. Locat, S. Shibuya, T.S. Tan, and R.S. Dinesh, Characterization of Singapore, Bangkok,
and Ariake clays, Canadian Geotechnical Journal 38.2 (2001): 378-400.
[7] R. Larsson, Undrained shear strength in stability calculation of embankments and foundations on soft
clays, Canadian Geotechnical Journal 17.4 (1980): 591-602.
[8] R. Larsson & M. Mulabdi, Shear moduli in Scandinavian clays: Report No.40, Swedish Geotechnical
Institute, Linkping, (1991a): 127 p.
[9] R. Larsson & M. Mulabdi, Piezocone tests in clays: Report No.42, Swedish Geotechnical Institute,
Linkping, (1991b): 240 p.
[10] J.A. Pineda, A. McConnell, and R.B. Kelly, Performance of an innovative direct-push piston sampler in
soft clay. Proc. CPT'14, Las Vegas, Nevada (2014): 279 288.
[11] B.M. Lehane, Vertically loaded shallow foundation on soft clayey silt, Proc. The Institution of Civil
Engineers - Geotechnical Engineering 156.1 (2003): 17-26.
[12] H.E. Low, M.L. Maynard, M.F. Randolph, and D.J. DeGroot. Geotechnical characterisation and
engineering properties of Burswood clay. Gotechnique 61.7 (2011): 575-591.
[13] S.G. Chung, Y.P. Hong, J.M. Lee, and S.C. Min, Evaluation of the undrained shear strength of Busan
clay, KSCE Journal of Civil Engineering 16.5 (2012): 733-41.
[14] S.G. Chung, C.K. Ryu, S.C. Min, J.M. Lee, Y.P. Hong, and E. Odgerel, Geotechnical characterization
of Busan clay, KSCE Journal of Civil Engineering 16.3 (2012): 341-50.
[15] C.K. Chung & R. Finno, Influence of depositional processes on the geotechnical parameters of Chicago
glacial clays, Engineering Geology 32.4 (1992): 225-242.
[16] R.J. Finno, S.L. Gassman, and M. Cavello, The NGES at Northwestern Univ., National Geotechnical
Experimentation Sites, GSP 93, ASCE, Reston/Virginia, (2000): 130-159.
[17] L. Cavalera & G. Scarpelli, Anisotropy of the undrained strength of normally consolidated clays (in
Italian), G Riv Ital Geotec 14, N3, JulySept (1980): 181-190.
[18] P. Monaco, S. Marchetti, and G. Totani, The flat dilatometer Test (DMT): design applications and
recent developments, Common Ground, Proc. 10th Australia-New Zealand Conference on
Geomechanics, Brisbane, Australia, 21-24 October, Vol. 2, (2007): 516-521.
[19] Y. Watabe, M. Tanaka, and J. Takemura, Evaluation of in-situ Ko for Ariake, Bangkok and Hai-Phong
clays, Proc. The 2nd Int. Conf. on Site Characterization, Porto, Portugal, Vol. 2 (2004): 1765-1772.
[20] J. Pestana, C. Hunt, and J. Bray, Soil deformation and excess pore pressure field around a closed-ended
pile, Journal of Geotechnical and Geoenvironmental Engineering 128.1 (2002): 1-12.
[21] H. Tanaka, National Report-the Current State of CPT in Japan, International Symposium on Cone
Penetration Testing: Linkping, Sweden, Sweden: Swedish Geotechnical Society, (1995): 115-124.
[22] S. Shibuya & H. Tanaka. Estimate of clastic shear modulus in Holocene soil deposits. Soils and
Foundations 36.4 (1996): 45-55.
[23] M.P. Garner, Loading rate effects on axial pile capacity in clays. Doctoral Dissertation, Brigham
Young Univ. Department of Civil and Environmental Engineering, Provo, Utah, 2007.
[24] R. Cole, Full-Scale effects of passive earth pressure on the lateral resistance of pile caps, Doctoral
Dissertation, Dept. Civil and Environmental Eng, Brigham Young University, Provo, UT, 2003.
[25] S.Y. Liu, G.J. Cai, L.Y. Tong, and G.Y. Du, Approach on the engineering properties of Lianyungang
marine clay from piezocone penetration tests. Marine Georesour Geotechnol 26 (3), (2008): 189-210.
[26] T. Lunne & S. Lacasse, Geotechnical characteristics of low plasticity clays, Proc. International
Symposium on Characterization of Soft Marine Clays, Japan 26-28 February (1997): 33-56.
[27] S. Leroueil & K. Hamouche, Geotechnical characterization & properties of sensitive clay from Qubec,
Characterization & Eng. Properties of Natural Soils 2. Lisse: Swets & Zeitlinger, (2003): 363-393.
[28] N.J. Yafrate & J.T. DeJong, Interpretation of sensitivity and remolded undrained shear strength with
full flow penetrometers, The Sixteenth International Offshore and Polar Engineering Conference,
International Society of Offshore and Polar Engineers, San Francisco, CA, USA, (2006): 572-577.
[29] M.M. Landon, Development of a non-destructive sample quality assessment method for soft clays,
Doctoral Dissertation, University of Massachusetts, Amherst, 2007.
[30] S. Shibuya & S.B. Tamrakar, Engineering properties of Bangkok clay. Characterization and
Engineering Properties of Natural Soils. Balkema, Rotterdam, Netherlands (2002): 645-692.
[31] A. Kennet, Y. Yu, and B. Westerberg, Behavior and modelling of Swedish natural soft clays, Proc. 13th
ICSMFE, New Delhi, India (1994): 57-60.
[32] T. Lunne, M. Long, and C. Forsberg, Characterization and engineering properties of Onsy
clay. Characterization & Eng. Properties of Natural Soils 1, Lisse: Swets & Zeitlinger (2003): 395-428.
[33] M.D. Lambson, D.G. Clare, and R.M. Semple, Investigation and interpretation of Pentre and Tilbrook
Grange soil conditions, Large-scale pile tests in clay: Proc. Recent Large-scale Fully Instrumented Pile
Tests in Clay, Institution of Civil Engineers, London, 23-24 June 1992.
[34] J.A. Ramalho-Ortigao, M.L. Werneck, and A.L. Willy, Embankment failure on clay near Rio de
Janeiro, Journal of Geotechnical Engineering 109.11 (1983): 1460-1479.
[35] M.S.S. Almeida, and M.E.S. Marques, The behavior of Sarapu soft clay, Characterization and
Engineering Properties of Natural Soils. Vol. 1, Lisse: Swets & Zeitlinger (2003): 477-504.
[36] M. Bozozuk, The Gloucester test fill, Doctoral Dissertation, Purdue University, 1972.
[37] T. Lunne, M. Long, and M. Uzielli, Characterization and engineering properties of Troll clay,
Characterization & Eng. Properties of Natural Soils, Lisse: Swets & Zeitlinger 4 (2007):1939-1972.
[38] M.W. O'Neill & G. Yoon, Engineering properties of overconsolidated Pleistocene soils of Texas Gulf
coast, Transportation Research Record 1479, National Res. Council, Washington, DC (1995): 81-88.
[39] L.J. Mahar & M.W. O'Neill, Geotechnical characterization of desiccated clay, Journal of Geotechnical
Engineering 109.1 (1983): 56 -71.
[40] D.W. Hight, F. McMillan, J.J.M. Powell, R.J. Jardine , & C.P. Allenou, Some characteristics of London
clay, Characterization & Eng. Properties of Natural Soils 2, Balkema, Rotterdam, (2003): 851907.
[41] M. Jamiolkowski, R. Lancellotta, M.L. Tordella, and M. Battaglio, Undrained strength from CPT, Proc.
European Symposium on Penetration Testing, Amsterdam (1982): 599-606.
[42] M. Jamiolkowski & D.C.F. LoPresti, Validity of in-situ tests related to real behavior, Proc. 13th
ICSMGE, New Delhi Vol. 5 (1994): 51-55.
[43] D.C.F. LoPresti, M. Jamiolkowski, and M. Pepe, Geotechnical characterization of the subsoil of Pisa
Tower, Characterization & Eng. Properties of Natural Soils 2, Balkema, Rotterdam, (2003): 909-946.
[44] D.G. Zapata-Medina, Evaluation of dynamic soil parameter change due to construction-induced
stresses, Ph.D. dissertation, Northwestern University, Civil Engineering, 260p, 2012.
[45] P.W. Mayne & R.A. Pearce, Site characterization of Bootlegger cove formation clay for Port of
Anchorage, Frontiers in Offshore Geotechnics (Proc. ISFOG, Perth), (2005): 951-955.
[46] P.W. Mayne, P.K. Robertson, and T. Lunne, Clay stress history evaluated from seismic piezocone
tests. Geotechnical Site Characterization 2 Balkema, Rotterdam, Netherlands, (1998): 1113-1118.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-366
'HYHORSPHQWRIODUJHVL]H'LVN7UDQVGXFHUWRHYDOXDWH
HODVWLFSURSHUWLHVRIFRDUVHJUDQXODUPDWHULDOV
Abilash POKHRELa,1 Reiko KUWANOa Laxmi P. SUWALb
a
Institute of Industrial Science, The University of Tokyo, Japan
b
ARC Center of Excellence for Geotechnical Science and Engineering, University of
Newcastle, Australia
Abstract.A new type of Disk transducer with the size of 80 mm diameter has been
developed by assembling p-type piezo ceramic elements and s-type piezo ceramic
elements, in order to carry out the elastic wave study of small to large grain size
geomaterials. It has been proved that by using multiple numbers of piezo ceramic
elements, development of required size of wave measurement device (i.e. disk
transducer) is possible. To clarify the workability of such disk transducer, elastic
properties of Toyoura sand (D50 = 0.20mm) were evaluated by three sorts of
method (i.e. static method, disk transducer method and Trigger accelerometer
method) by using the large triaxial apparatus with the rectangular prismatic
specimen of 50cm*23cm*23cm. All tested sample were in completely dry
condition. Applying 11 number of very small strain (i.e. strain less than 0.001%)
cyclic loading youngs modulus and Poissons ratios were derived and shear
modulus were calculated. Both the axial and radial strain was locally measured by
local deformation transducers. Newly developed large size disk transducer has
been used to evaluate the elastic properties by wave measurement method. Trigger
Accelerometer method was also used to compare the elastic wave properties
obtained by elastic wave propagation method. To prove the reliability, consistency
and further application on the geotechnical engineering study elastic properties
measured by newly developed disk transducer method were compared with
previous research and the stiffness found to be fallen in similar range within
allowable scatters.
Keywords.Large size Disk Transducer, Small strain stiffness, shear wave,

laboratory test
1. Introduction
Bender element is the most popular means of wave measurement in research on

geotechnical engineering field because of not only their smaller size, and lower voltage
required and easier operation but also cost-effectiveness and provides the realistic
design parameter. By using bender element in the various testing apparatus such as
odometer [1] triaxial apparatus [2], [3] and torsional shear apparatus [4] carried out the
dynamic methods of investigations of clayey and sandy soils. But bender element
method is not suited for the coarse grain material and undisturbed or cemented material.
The application of wave measurement to coarse grained material is not easy, because of
the tested material does not behave as continuous media.
This study focused to the development of large size wave measuring device which
must be applicable to study the elastic wave properties of geomaterials having small to
large grain on single apparatus. Recently developed DT method [5] showed better
result than bender element in laboratory soil testing because of its completely flat
1
A. Pokhrel et al. / Development of Large Size Disk Transducer to Evaluate Elastic Properties 367
surface. Undisturbed and cemented material can also be tested easily. However, it is not
useful for large triaxial apparatus with the specimen size of 50cm*23cm*23cm triaxial
cell available in IIS, The University in Tokyo because of the limitation of size and
working capacity. In addition the amplitude of Received signal decrease with the
increase in specimen length and more than 20mm size Piezo ceramic disk is not
available in market. Hence this study fulfills the long felt need of geotechnical
laboratory testing device.
2. Transducer for Wave measurement
2.1. Piezo-ceramic elements
The direction of the deformation of piezo-ceramic depends on the shape and

composition of ceramic disk, direction of poling axis and applied electric field. A flat
shaped P-type piezo-ceramic element is polarized in the direction perpendicular to the
electrodes whereas S-type element is polarized in the direction parallel to the electrodes.
As shown in figure 1 upon the application of electric voltage the direction of
polarization is the direction of thickness of the element, so it generates compression
wave. Flat disk having the properties as shown in table 1 were used in this study.
2.2. Development of large size disk transducer
As shown in figure 2, a large size i.e. 80mm dia. Disk had been developed as a single
unit having the properties to the piezo ceramic disk. Total eight numbers (4 p-types and
4 s-types) of piezo elements were used in this study.
A combination of p-type and s-type element called PS type element having the property
to transmit and receive both the compression and shear wave was made by merging the
P-type and S-type element by applying the rapid araldite and to protect from being
cross connection a small plastic plate having a thickness of 2mm was placed between
them before merged together. Other two metallic plates were glued on the both side of
previously made single unit of p-type and s-type element (PS type element) for the
same. A circular metallic plate of diameter 80mm was screwed with 4 sets of PS
element on the P side. S side of PS element was glued with another metallic plate to
ensure the free movement of S element in the horizontal direction. Figure 3 explains
the steps followed during the development procedure. During this procedure the
polarization direction of s-type element should be same orientation to protect from
being independent movement of piezo element on the application of electric field.
Figure 1Schematic figures of piezo element; S-type piezo element (left) and P-type piezo element (right)
368 A. Pokhrel et al. / Development of Large Size Disk Transducer to Evaluate Elastic Properties
Table 1.Properties of piezo ceramic elements

Items k15* C** Length Width Thickness
Unit (%) pF mm mm mm
P-type, Maximum value 54.7 4310 19.99 19.99 2.012
Z2T2020S- Minimum value 53.1 4060 19.87 19.87 2.005
LLYXN(C-6) Average 54.32 4147.5 19.957 19.94 2.0101
Standard deviation 0.677 67.27 0.0341 0.0247 0.00186
S-type, Maximum value 69.7 4540 20.07 19.99 2.007
SZ2T2020S- Minimum value 69.1 4450 19.96 19.89 1.995
LLYX (C-6) Average 69.39 4502.5 20.036 19.944 2.0001
*coupling factor, ** electro static capacity
Figure 2 Setting of peizo ceramic element in large size disk transducer
Figure 3Development steps of large size disk transducer

After the assembling procedure, proper connection of plates and ceramic element
must be ensured. The connection between the piezo elements and plastic element, piezo
elements to metallic plates should be checked carefully to make the noise free disk
transducer. The wire connection in the transmitter was parallel because to give same
voltage input from the entire transmitting piezo-ceramic element. On the other hand,
for the receiver wire connection was made in series to gain the voltage as much as
possible.
2.3. Verification of workability
Since, newly developed disk transducer is assemblage of multiple number of piezo

ceramic elements, the proof of receiving signals amplitude with increasing number of
receiving/transmitting elements is necessary. Another thing to be considered is the time
required to travel from the piezo element to the top surface of the Disk Transducer
because the distance of s-type and p-type element from the top surface are 8mm and
12mm respectively. The test (calibration) of disk was carried out by facing the
transmitter and receiver of disk each other as shown in figure 4 [A]. This test was
conducted without applying any load. The signals were transmitted through the
transmitting disk and received by the receiver. Both input and output signal were
recorded by the HIOKI 8860-50 MEMORY HICORDER Oscilloscope on various
input frequencies. Figure 4[B] shows the typical S-wave form recorded on the pulse 5
Hz, Sine 10 kHz, Sine 20 kHz and Sine 30 kHz in time domain series at input voltage
of 80 Volt. Considering the peak of the input and output signal as well as rising of the
input and output signal, no time difference between the input and output signal was
recorded in all frequencies. Similar phenomenon was recorded on the p-waveform too.
In figure 5 increasing pattern of amplitude of receiving waves at increasing number of
receiving/transmitting element in 20 kHz input frequencies is shown. Similar types of
compressional waveforms were also recorded.
Figure 4 [A] Disk transducer, [B] S-waveform obtained at different frequency input
Figure 5S- waveform obtained on sine 20 kHz input frequency at various combination of piezo element
3. Experimental Program
3.1. Tested materials
Toyoura sand was taken for the verification of workability, reliability and consistency
of newly developed disk transducer. This sand is fine grained, uniformly graded and
yellowish brown in color. The physical properties of Toyoura sand and its photograph
are shown in figure 6.
Figure 6Toyour sand, Photograph (left), gradation curve (right)
3.2. Triaxial apparatus and wave measurement equipment
The large scale tri-axial apparatus has been used in this study. A personal computer is
connected with the tri-axial apparatus to control, measure the data and feed back to the
system which is capable of various operations like setting the test conditions, control of
machines, recording of data and processing of data.
In order to transmit the signal, function generator Textronix Co. Ltd., Model:
AFG3021 was used. The trigger rate range of this function generator is from 1ms to
500 sec. An amplifier is used to magnify the input signal before transmitting into the
disk transducer. An Oscilloscope HIOKI 8860-50 MEMORY HICORDER record
and displays the waveform of transmitted and received signals.
3.3. Specimen preparation and testing procedure
Considering the properties of testing material, air pluviation technique was adopted to
prepare the specimen. In this method approximately homogenous density was maintain
by controlling the pluviation height and changing the pouring direction from clockwise
to anticlockwise or vice versa. All the tested specimen were prepared in completely dry
condition with dimension of approximately 23cm*23cm*50cm at 40 kPa negative
stress state.
During the experiment, the stress state level was increased to 50, 100, 200 and
400 kPa. The specimen was subjected in creep condition for at least 10 minutes to
dissipate the stress thoroughly in the specimen. Then 11 numbers of cyclic loading was
applied ensuring that strain is in the range if 0.001%. Again creep stage was applied for
the elastic wave measurement in each stress state. Local deformation transducer was
used to measure the radial and axial strain. Disk transducer and Accelerometer methods
were carried out simultaneously to measure the elastic wave study of the sand. In
accelerometer method (see also AnhDan and Koseki[6]), disk transducer was worked
as a trigger source instead of conventional trigger.
3.4. Interpretation of signals in this study
Kawaguchi [7] suggested that the first major deflection point of waveform on zero base
line is the first arrival point. This method is call the rise to rise method and used for the
Accelerometer method in this study. But the most reliable estimation of travel time
from visually picking is determined from the first major peak position (see also
Viggani and Atkinson, [2]) therefore this approach has been adopted in the disk
transducer method in this study.
4. Measurement of Elastic properties
Two numbers of tests were carried out by static as well as dynamic method of study.
4.1. Elastic properties by static and dynamic method
Obtained stiffness value of Toyoura sand by static measurement method is plotted in

figure 7. Since two tests have a different density, the result shows the different value
but consistent.
Disk transducer and Accelerometer method of elastic wave propagation method were
used to evaluate the elastic properties of Toyoura sand. A typical shear waveform by
Disk Transducer and accelerometer method obtained on various input frequency at 200
kPa stress state is shown in figure 8. The input voltage is 100V in each case. As
explained above, the arrival point was visually picked as a peak to peak basis in disk
transducer method and rise to rise in the accelerometer method.
Velocity of the propagated signal is the major indicator of stiffness characteristics of
the materials. The travel time for both P and S waves are achieved by analyzing
waveforms. In disk transducer method height of specimen is considered as a travel
distance and distance between the accelerometer in the accelerometer method. By using
the velocity of the wave and density of the materials, small strain stiffness was
calculated.
Figure7Young's modulus obtained by static method (left), shear modulus by static method (right)
Figure 8 S-waveform obtained by the Disk transducer method (left), S-waveform obatined by the
Accelerometer method (right)
4.2. Comparision of static and wave measurement method
The normalized youngs and shear modulus calculated from all the methods are plotted
in figure 9. The difference of stiffness derived from the static and disk transducer
method are found to be varied by less than 10%. The accelerometer method shows
higher result in both cases it may because of the effect of refracted wave since wave
were measured on sideways.
Figure 9Normalized Young's modulus obtained by three test methods (left), Normalized Shear modulus
obtained by three test methods (right)
Finally, the results obtained in this research by static as well as disk transducer method
are compared with the previous research (Suwal [5] [8] [9]). The previous research
methodology, materials and type of piezo ceramic disk are same with this research. The
only difference is the size of testing device. So author took the 5 specimen results
conducted in small tri-axial cell having the specimen dimension 15cm height and 75
mm dia.
The small strain stiffness of Toyoura sand obtained in author and previous research
are shown in figure 10. The stiffness shows the consistency in both the research.
Figure 10comparison of Young's modulus (left) and Shear modulus (right) obtained with past research work
5. Conclusions
By using multiple numbers of piezo ceramic elements, development of required size of

wave measurement device (i.e. disk transducer) is possible. Therefore by increasing the
size of testing device larger particle size geomaterials can be easily tested.
The elastic properties on Toyoura sand measured by large size disk transducer fell in
similar range with past research results and it was confirmed that the newly developed
transducer is one of the applicable tool in the laboratory experiment.
References
[1] Jamiolkowski, M., Lancellotta, R. and Lo Presti, D.C. F. (1995): Remarks on the stiffness at small strains
of six Italian clays, Pre-failure Deformation of Geomaterials (edited by: Shibuya S. et al.) Balkema,
817-836
[2] Viggiani, G. and Atkinson, J. H. (1995): Interpretation of Bender Element Tests, Geotechnique, 45(1),
149-154
[3] Jovicic, V., Coop, M. R., and Simic, M. (1996): Objective Criteria for Determining Gmax from Bender
Element Tests, Geotechnique, 46(22), 357-362Suwal, Laxmi Prasad and Kuwano, Reiko, Statically
and Dynamically Measured Poissons Ratio of Granular Soils on TriaxialLaboratory
Specimens,Geotechnical Testing Journal, Vol. 36, No. 4, 2013, pp. 113,
[4] Duttine, A., Di Benedetto, H., Pham Van Bang, D. and Ezaoui, A. (2007): Anisotropic small strain elastic
properties of sands and mixture of sand-clay measured by dynamic and static methods, Soils and
Foundations, 47, 457-472
[5] Suwal, Laxmi Prasad and Kuwano, Reiko, Disk shaped piezo-ceramic transducer for P and S wave
measurement in a laboratory soil specimen, Soils and Foundations, Vol. 53, No. 4, 2013, pp. 510524
[6] AnhDan, L. Q., Koseki, J., and Sato, T. (2002): Comparison of Youngs Moduli of Dense Sand and
Gravel Measured by Dynamic and Static Methods, Geotechnical Testing Journal, 25(4), 349-368.
[7] Kawaguchi, T., Mitachi, T., & Shibuya, S., (2001): Evaluation of shear wave travel time in laboratory
bender element test, Proc. 15 th International Conference Soil Mechanics and Foundation Engineering,
Istanbul 1, 155-158.
[8] Suwal, Laxmi Prasad and Kuwano, Reiko, Statically and Dynamically Measured Poissons Ratio of
Granular Soils on TriaxialLaboratory Specimens,Geotechnical Testing Journal, Vol. 36, No. 4, 2013,
pp. 113
[9] Suwal, P. Laxmi, 2013, Disk transducer for elastic wave measurement and its application to unsaturated
sandy soils, Doctors degree Dissertation, The University of Tokyo.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-374
Assessment of shear modulus by different

seismic wave-based techniques
Jaime SANTOSa,1, Joo SANTOSa, Cristiana FERREIRA b, Cludio PEREIRA c and
Antnio GOMES CORREIA c
a
University of Lisboa, Lisboa, Portugal
b
University of Porto, Porto, Portugal
c
University of Minho, ISISE, Guimares, Portugal
Abstract. Using combined setup of bender elements and accelerometers, tests

were conducted on Coimbra sand specimens in order to measure and interpret
seismic wave velocities to assess initial shear modulus. For these tests both time
and frequency domain analyses were performed. Resonant column tests were also
performed on the same sand to validate the results obtained with the bender
elements and accelerometers setup. As is well known, in the last decades the
development of new laboratory techniques to assess soil stiffness through the use
of seismic wave-based techniques, has received significant attention due to its
simplicity and versatility of the equipment setup. One of these techniques is the
bender elements test which have been one of the most widely used, although some
limitations concerning its usage. In this context, the combined use of bender
elements with other seismic wave-based testing techniques, such as accelerometers
or the resonant column, is quite important to compare and validate the testing
techniques. Given its miniature size, the installation of accelerometers on the side
of the sample is considered feasible without significant disturbance on the other
measuring techniques. The resonant column is a widely used and accurate testing
technique due to its reliability and repeatability. Finally, the results of this
combined tests allow a critical discussion on the advantages and limitations of the
use of bender elements and accelerometers, in contrast with the resonant-column
for the assessment of the shear modulus in sand.
Keywords. Accelerometers, Bender elements, Resonant-column, Shear modulus
1. Introduction
The importance of soil characterization on the very small to small strains domain (i.e.
shear strain from 10-6 to 10-4) for engineering design purposes is well established. At
this deformation domain, the response of soil can be considered quasi-elastic, being the
corresponding shear modulus designated by maximum or initial shear modulus, Gmax or
G0). This shear modulus is not affected by the nature of the loadings (monotonic or
cyclic) since there is no stiffness degradation in load-unload cycles [1]. Previous
studies [2] and [3] using different testing techniques showed that the initial shear
modulus is not affected by the type of test. The source of excitation frequency and the
strain rate do not seem to affect the response of the soil, and the major constraint for
1
J. Santos et al. / Assessment of Shear Modulus by Different Seismic Wave-Based Techniques 375
determining the initial shear modulus is usually attributed to the precision of the
measuring instruments.
The effect of different factors on the initial shear modulus of sands was studied by
[4] and [5] with resonant column tests. The research revealed that the initial shear
modulus can be well expressed by:
2
(B-e) n n
G0 =A ('0 ) =A.Fe).('0 ) (1)
1+e
where A, B and n are empirical constants experimentally determined, e is the void ratio,
is the mean effective stress and F(e) is a function of the void ratio. Generally, G0
increases with '0 and decreases with e, which means that it increases with the increase
of the relative density of the material:
emax -e
ID= e (2)
max -emin
According to data collected by [6], dense sand and gravels adjust well with Eq. (1),
the exponent n varies between 0.38 and 0.85 mostly depending of the grain size
distribution and contact conditions between particles and B is equal to 2.17 for the
majority of soils. If one divides G0 by Fe the ratio gives the normalized value of the
G0 allowing comparing test results of the same material but with different void ratio.
2. Shear modulus assessment
There are two main ways to determine G0 : one through the theory of elasticity, using
stress-strain measurements under small cycles, and another through the theory of wave
propagation, using the measurement of shear wave velocities. The standard test for G0
assessment is the resonant column test (RC), which uses the shear wave propagation
velocity theory. Bender elements test (BE) is one of the most spread techniques to also
assess shear modulus due to its simplicity. Both techniques apply a shear strain level
near of 10-6 to the material [1].
In a conventional RC test, a cylindrical specimen of soil is subjected to a steady-
state harmonic excitation, and the response of the system in terms of vibration is
measured. The frequency of the input signal is shifted until resonance is achieved. It is
possible to compute the dynamic properties of the soil (stiffness and damping) as
derived from the dynamic equilibrium of the specimen [7]. In other hand, in a BE test a
voltage signal is applied to a piezoceramic element (transmitter) which transmits a
small shearing movement over one end of the cylindrical soil specimen. This
disturbance travels across the specimen length until the other end is reached, where a
similar piezoceramic element (receiver) receives the mechanical perturbation and
generates a voltage. The time interval between the emitted and received signals enables
to compute the shear velocity. BE tests have however some limitations which influence
its accuracy [7]. Table 1 lists the main limitations identified and the possible
alternatives to overcome these difficulties.
376 J. Santos et al. / Assessment of Shear Modulus by Different Seismic Wave-Based Techniques
Table 1 - Main limitations of RC and BE tests and the possible alternatives (from [6])
Method Limitations Alternative considered

RC test Only harmonic excitation Random noise, ambient noise
Shear strain > 10-6 Random noise, combined
Determination of G and only at resonant frequency methods
Time consuming procedure Sweep sine
Too many cycles (soil disturbance) Controlled source
BE test Excess of human judgment Frequency domain and statistical

Uncertainties about the actual behavior of BE transmitter methods
Great amount of data to process and analyze External controlled source
Miniature accelerometers
Automation
The use of more than one type of test, complementary between them, can improve
the testing reliability. Thus the results of one type of test can be compared with another
one and therefore the interpretation can be more consistent.
2.1. Interpretation of Bender Elements and Accelerometers results
Once the interpretation of BE results involves some uncertainty different approaches

have been proposed to deal with the interpretation issues and they are usually based on
the time or on the frequency domain analysis. Generally the frequency domain method
produces an estimate of shear wave velocity, which is lower than that from traditional
time domain readings [8].
Known as the most simple, common and usual procedure for interpreting BE
measurements, the first direct arrival method consists on the identification of the first
instant of arrival of the wave in the output signal, similarly to the techniques used in
geophysical tests. The usage of this method is sometimes source of error once some
factors can interfere with the wave arrival identification. These factors, reported in [9],
can be overcome using some alternatives considered in Table 1. The use of
piezoelectric accelerometers (AC) takes advantage of the accuracy on using calibrated
equipment and the possibility of reading acceleration determining the arrival of the
shear wave in a particular direction [10]. It has been demonstrated that accelerometers
used in a coupled system with benders work as receivers in a more accurate way [11].
The use of continuous signals is also an alternative to be considered to reduce the
error on BE tests. This technique requires the shear wave velocity to be decoded from
measurements of relative phase of transmitted and received signals. These called
frequency domain (FD) methods have a number of advantages over traditional time-
based (TD) measurements, namely the possibility of creating an algorithm to determine
travel time by establishing the gradient of a graph of phase difference against frequency
[8]. Generally, a continuous harmonic sinusoid is used as input signal, though a generic
input signal can also be used to evaluate the phase delay by decomposing the signal
into its harmonics, using the Fourier transform. Also in this technique, accelerometers
can be used to acquire the response of the system as a receiver.
3. Method
ds
3.1. Equip
pment and testted material
The RC equipment frrom the Univ versity of Lissbon (Figure 1) is a Drnnevich-type

manufactuured by Seiken Inc. in 19 992. It consissts of three subsystems: ppneumatic,
electro-meechanic and electronic.
e The pneumatic subsystem
s pro
ovides the connditions to
the controol of cell prressure, backppressure and axial force; the electro-m mechanical
subsystemm allows the torsional vibrration and th he electronic subsystem prrovides the
input sign
nal and measurres the respon
nse of the systeem.
Figure 1. RC details: a) Prepared sample; b) Pneumatic sy

ystem control; c) Electronic system
m control
The system
s at Unniversity of Minho
M for BE
E and AC testts is a 100mm m Bishop-
Wesley sttress-path chaamber, adapted to accomm modate BE. Two AC weere applied
directly onn the side of the
t sample, in n order to validate the BE signals and miinimize the
subjectivitty in interprettation [10] and
d [11] (Figuree 1a). The ACC used are from m Bruel &
Kjr, these are piezoeelectric senso ors (type 4513 3-001, 100 mV/g
m sensitivvity, 50 g
measuring g range, 1 Hz to 10 kHz freequency rangee, 12.7 mm in n diameter, 155.65 mm in
height, 9.0 g in weigh ht). In order to ensure adeequate coupliing and stabillity during
he AC were fix
testing, th xed to the sides of the speccimens at speccific points, byy means of
pins screwwed to the baack of the AC C, which involved puncturring the mem mbrane and
carefully insulating
i the hole (Figure 1b).
a) b) c)
Figure 1. a)) Schematic view
w of the 100mm stress-path
s chamb
ber system integrrating the combinned use of BE
and AC; b) AC
A pins and its issolation; c) Setup
p overview
The first
f AC (AC1 1) was placed at 30 mm fro om the base off the specimenn and AC2
was placeed 100 mm above a the firrst. The AC axis was plaaced in the same plane
direction of the moveement of soil particles. TD D and FD teechniques werre used in
combinatiion and a miniimum of fourr input and outtput signals werew recorded in order to
eliminate problems succh as random noisen and to get
g a clear respponse signal. Regarding
to the FDD techniques, a sinusoidal signal
s of lineaar sweep of frrequencies froom 1 to 50
kHz for a total period of 20 ms and d amplitude of o 20 peak-to--peak voltagee (Vpp) was
used.
A porrtuguese riverr sand called Coimbra
C Sand d was used forr this study. Thhe samples
were prep pared using the dry depositiion technique,, by means off deposing thee sand with
a funnel to achieve a certain relaative density.. The tests were w med in dry
perform
conditionss. The sand used
u for the tests exhibit a D50 = 0.28 8 mm and a uniformity
coefficien
nt (Cu) around d 1.22. The particle
p size distribution
d obbtained for this material
[12], is prresented in Fiigure 2. The initial
i physicaal properties of
o the sampless tested on
the RC (RRC1 and RC2) and BE/AC equipment
e (P11) are shown in
i Table 2.
Table 2 - Initiaal physical properrties of

Coimbraa sand specimens
Height Diameter
Test ID (%)
(mm) ((mm)
RC1 83 101.0 69.8
RC2 65 101.0 70.4
P1 61 200,8 98,9
Figure 2. Grain
G size distrib
bution of Coimbraa sand
4. Resultss
Figure 3 shows
s the initiial shear modu
ulus results off the RC tests for the two saamples and
for differeent mean effecctive stresses. The regressioon equations obtained
o for eaach test are
also showwn in the samee figure. Figu ure 4 shows th he normalized d measured innitial shear
modulus in i bi-logarithmmic scale.
ure 4. Initial shearr modulus normaalized by the

Figu
Figuree 3. Initial shear modulus.
m RC testts
void ratio function.
f RC tests
As ex xpected, the in nitial shear modulus

m decreaases with the decrease of tthe relative
density annd the values ofo G0 are conssistent with th
he type of tested material [113]. As can
be seen ini Figure 4, the normalized values off the initial shear moduluss are very
consistentt, with A=8.9 95, B=2.17 and n=0.42 from the Eq q. (1). The rresults are
consistentt with [6] and the previouslyy referred valuues of n betweeen 0.38 and 00.85.
Figurre 5 shows the overlapp ping signals (transmitter and receiverr) and the
identificattion of the trav
vel time for th
he P1 BE test, by the first diirect arrival m
method.
F
Figure 5. Examplle of overlapping
g signals and identification the trav
vel time for BE teest
Figurre 6 shows thee initial shear modulus meaasured on the BE

B setup and on the AC
setup, for TD and FD an nalysis, for th
he P1 test (ID=
=61%).
a) b)
Figure 6. G0: a) measured on the BE setup, for TD and FD; b)
b measured on th
he AC setup, for TD and FD
As ex
xpected, the reesults obtained by the FD method
m gave an
a estimate off the initial
shear mod dulus lower th han time dommain readings [8]. Comparing the BE reesults with
those fromm AC, it can be said that the results arre similar. Ho owever the FD analysis
results show lower vallues on the AC A compared to the BE an nd higher valuues on TD
analysis on
o the AC for f higher mean
m effectivee stresses. This differences can be
explained by the fact off the AC alloww to measure the shear wav ve velocity beetween two
cross secttions that aree not the bou undaries and are not influenced by thee boundary
effects and coupling. Figure 7 comppares the resullts obtained inn BE and AC setups, for
time and frequency dom main, for the P1 test (ID=6 61%) (simultaaneous measuurements at
the same stress
s conditioons).
Figure 7. Reelation between initial shear modu

ulus measured on
n the BE and AC setups, for time aand frequency
domain
The correlation
c beetween the reesults of the two types off tests indicattes a good
agreementt between theem (less than 10% differen nce) even showing higher ddifferences
for the FD
D analysis, ass previously referred. Figurre 8a shows the
t comparisoon between
BE (TD and
a FD) and RC setups fo or the same sttress conditioons and relativve density.
Figure 8bb compares thhe results betwween AC and d RC setups, for time and frequency
domain. In
I both figurees the values plotted
p on thee horizontal axis
a are the innitial shear
modulus G0 for RC2 teest (ID=65%) and on the vertical
v axis arre plotted the G0 results
for P1 test (ID=60%, BE on Figu ure 8a and ACA on Figuree 8b both forr time and
frequencyy domain.
a) b)
Figure 8. Reelation between initial
i shear modu
ulus measured onn the: a) BE (P1) and
a RC (RC2) seetups, for time
and frequenccy domain; b) ACC (P1) and RC (RRC2) setups, for tiime and frequency domain.
As caan be seen thee correlation between

b the reesults is goodd since the maajor part of
the results fall inside the +/-10% difference
d ran
nge. Only forr higher meann effective
stresses a ratio higher than
t 1.1 has been
b observedd. Besides the results from tthe AC are
closer to the
t RC than thet BE, one can c conclude thatt the referrred boundary conditions
influence the results an
nd the stress leevel increase that
t effect. Ass shown in Figure 6, the
differencee between freqquency and tim me domain an nalyses is larg ger in the BE ttest than in
the AC tessts.
5. Conclusions
Using combined setup of BE and AC, tests were conducted on Coimbra sand in order
to assess initial shear modulus. RC tests were also performed on the same sand to
validate the results obtained with the BE and AC setup.
The results obtained by the frequency domain method gave an estimation of the
initial shear modulus lower than time domain readings both in BE and AC tests. The
correlation between the results of BE and AC tests indicates a good agreement between
them (differences less than 10%) even showing higher differences for the frequency
domain analysis.
The BE and AC test results also agrees well with RC. Only for higher mean
effective stresses a ratio higher than 1.1 has been observed. The accelerometers reveal
a better accuracy than the BE being less influenced by the boundary and stress
conditions.
Acknowledgements
This work was developed with the financial support provided by FCT (Portuguese
Foundation for Science and Technology) under the research project WaveSoil -
PTDC/ECM/122751/2010 from FCOMOP-01-0124-FEDER-020365 project.
References
[1] J. A. Santos, Caracterizao de solos atravs de ensaios dinmicos e cclicos de toro - Aplicao ao
estudo do comportamento de estacas sob aces horizontais estticas e dinmicas, Lisbon, PhD Thesis,
1999
[2] F. Tatsuoka, R. Jardine, D. Lo Presti, H. Di Benedetto and T. Kodaka, Characterising the Pre-Failure
Deformation Properties of Geomaterials, Hamburg, 1999
[3] J. Martins and A. Gomes Correia, Comparison between Laboratory and Field Stiffness by Wave
Measurements, Journal of Testing and Evaluation, , (2015) Vol. 43, No. 2, pp. 113
[4] B. O. Hardin and F. E. J. Richart, Elastic Wave Velocities in Granular Soils, Journal of the Soil
Mechanics and Foundations Division, ASCE, (1963) 89 (1), pp. 33-65
[5] F. E. Richart, J. R. Hall and R. D. Woods, Vibrations of soils and foundations, Englewood Cliffs:
Prentice-Hall, 1970
[6] K. Ishihara, Soil Behaviour in Earthquake Geotechnics, Clarendon Press, 1996
[7] J. F. Camacho-Tauta, Evaluation of the small-strain stiffness of soil by non-conventional dynamic testing
methods, Lisbon: PhD Thesis, 2011
[8] P. Greening, D. Nash, N. Benahmed, A. Viana da Fonseca and C. Ferreira, Comparison of shear wave
velocity measurements in different materials using time and frequency domain techniques, in
Proceedings of Deformation Characteristics of Geomaterials, Lyon, 2003
[9] C. Ferreira, A. Viana da Fonseca and J. A. Santos, Comparison of simultaneous bender elements and
resonant column tests on Porto residual soil, in Soil Stress-Strain Behavior: Measurement, Modeling
and Analysis - Geotechnical Symposium in Rome, Rome, 2006
[10] J. P. Martins, Compaction and its influence on the structural behaviour of high-speed railways. Ph.D.
thesis, University of Minho, 2011
[11] C. Ferreira, J. P. Martins, and A. Gomes Correia, Determination of the small-strain stiffness of hard
soils by means of bender elements and accelerometers, Geotechnical and Geological Engineering
(2013), 6, 17
[12] P. I. of Quality, NP EN 933-1 Tests for geometrical properties of aggregates, 2000
[13] J. A. Santos, R. Gomes, J. Loureno, F. Marques, P. Coelho, R. Azeiteiro, L. Santos, V. Marques, A.
Viana da Fonseca, M. Soares and . Abreu, Coimbra Sand - Round Robin Tests to Evaluate
Liquefaction Resistance, 15 WCEE, Lisbon, 2012
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-382
In Situ and Laboratory Mechanical

Characterization Using High-Resolution
Fiber Optic Distributed Sensing
Assaf KLAR a,1, Shun UCHIDA b, and Eyal LEVENBERG a
a
Technion Israel Institute of Technology, Haifa, Israel.
b
Rensselaer Polytechnic Institute, Troy, United States.
Abstract. This paper explores the potential use of high-resolution fiber optic
distributed sensing technology for in situ moduli profiling and in laboratory
element testing. In recent times, strain measurement using fiber optics has been
employed in innovative civil engineering applications such as in the health
monitoring of ageing infrastructures. Through recent developments, in particular
Rayleigh backscatter optical frequency domain reflectometry technique, the fiber
optic sensing technology is nowadays capable of providing continuous distributed
strain measurement with a higher spatial resolution of the order of millimeters. As
a result, the technology can potentially serve as a viable alternative to conventional
strain gauges (i.e. high-spatial resolution yet localized measurement devices) or
seismic geophysical measurement (i.e. distributed yet low-spatial resolution). This
paper provides two examples of its applicability to both in situ and laboratory
mechanical characterization.
Keywords. Fiber optics sensing, Mechanical characterization, Load tests, Small

strain modulus, Stiffness profile, Strain Measurements, Element testing.
1. Introduction
Distributed fiber optic sensing has received a considerable focus over the last decade
from both the fiber optic and civil engineering communities, resulting in substantial
advances in both fundamental developments and field integration and interpretation.
These include structural health monitoring [1,2], pipeline integrity evaluation [3,4,5],
and tunneling and tunneling induced ground displacement evaluation [6,7,8]. Recent
fiber optic developments have increased significantly the spatial resolution of
distributed sensing, resulting in sub-centimeter strain measurements capabilities [9,10].
This paper explores the potential of distributed sensing for both in situ and
laboratory characterization of soils. In specific, the paper covers recent trials to
evaluate the in situ stiffness profile using vertically installed fiber optic cable, together
with surface loading, and the full-field view of strain profile in laboratory uniaxial tests.
The demonstrated cases in the paper utilized the Rayleigh backscatter optical frequency
domain reflectometry technique.
1
Corresponding Author. Email: [email protected].
A. Klar et al. / In Situ and Laboratory Mechanical Characterization 383
2. In situ stiffness profile evaluation using fiber optics
Determination of soil stiffness moduli and their spatial distribution is a crucial

component of engineering design aiming at deformation evaluation. Laboratory testing
of undisturbed soil samples usually involve some violation of the in situ condition
which may affect the test results and the accurate determination of stiffness moduli.
Furthermore, no matter how comprehensive the investigation may be, it will
characterize the material locally without truly revealing the spatial distribution of
properties. In situ test methods have been developed throughout the years to
compliment conventional laboratory testing. These range from conventional seismic
profiling (e.g. refraction, crosshole or downhole tests, and spectral analysis of surface
waves) to hybrid methods, such as the seismic piezocone penetration test [11] and the
seismic flat dilatometer test [12]. Although these methods provide evaluation of the
entire depth of investigation, the profiling resolution may not always be high, due to
the manner in which seismic waves travel and spread in the soil.
This section briefly explores the possibility of utilizing high resolution fiber optic
distributed sensing technologies for in situ determination of the elastic shear moduli
profile with depth. A more detailed explanation of both the analytical approach and the
experimental validation is provided in [15].
2.1. Methodology
Let us consider a stratified layered soil loaded at the surface. A sufficiently flexible and
sensitive fiber that is pre-positioned vertically within the ground can potentially sense
the resulting vertical strains. It is assumed that under small loads, the response of the
soil continuum will be governed by its linear elastic stiffness (i.e. small strain modulus).
We may identify the stiffness profile by an inverse problem, in which we seek the
modulus distribution with depth, in a well-defined boundary value problem, that yields
a vertical strain distribution that best agrees with the measured strain distribution.
Linear elasticity is governed by two parameters (for example, Youngs modulus and
Poissons ratio, shear and bulk moduli, or any combination of them). The engineering
determination of the Poissons ratio (when considering total elastic behavior) is rather
straight forward (either 0.2-0.3, for drained soils or nearly 0.5 for undrained soils). Let
us assume that the Poissons ratio is known, in which case, we may be interested in
recovering the profile of Youngs modulus. Mathematically, the inverse problem for
determination of the Youngs modulus profile can be written as an optimization
problem as follows:
4 L
E * ( z ) arg min 3 (4 ) Fv 14 , E ( z ), ( z )2 d4

2
m
(1)
E(z)
4 0
where E*(z) is the inferred Youngs modulus profile from the inverse problem, (z) is
the assumed Poissons ratio, m(z) is the measured strain along the vertical optical fiber
(having a length of L), and Fv[4,E(z),(z)] is an operator which provides the vertical
strain at depth 4for a given distribution of E and , under a prescribed surface load.
E(z) is the argument of minimization. Fv can be based on closed form analytical
solutions of the potential function of multi-layered elastic domain [13] as described in
[14,15], or on finite element simulation of the problem. The optimization problem
384 A. Klar et al. / In Situ and Laboratory Mechanical Characterization
presented in Eq. 1can be solved using different minimization techniques. However, a

much simpler algorithm may be proposed by recognizing that if a local agreement is
achieved (that is, m(z) - Fv[z,E(z),(z)]=0) so is the general minimization. Since Fv
answers the elastic constitutive law at any point, the condition of local agreement can
be written as:
) z ( z )) x ( z )) y
m ( z ) Fv 1z, E ( z ), ( z )2 (2)
E( z)
where )z,)x and )y are the stress changes with depth due to surface loading. Note
that these stress changes depend on the boundary value problem involving a specific
E(z) and(z). Eq. can be rewritten as:
) z ( z )) x ( z )) y
E( z) (3)
m ( z)
It is suggested that this equation be solved iteratively together with the measured strain,
m(z), to result in the distribution of E with depth. The iterative procedure involves four
stages: [1] consider a uniform Youngs modulus with depth as an initial guess; [2]
solve the linear elastic problem under the considered modulus profile to establish the
stress change; [3] update the modulus profile using Eq. 3 under the derived stresses
from stage 2; [4] return to stage 2 for a new evaluation of stresses until convergence.
The solution of stage 3 is mathematically identical to that of updating the modulus
profile with the ratio of analytical strains to measured strains; that is
E(z):=E(z)a(z)/m(z) (where `:=' means replace or update). Note that the modulus value
associated with the initial iteration is of no importance, since the stress distribution for
the case of a homogenous half space is independent of the modulus value. Examination
of the above procedure under the input of ideally synthetic cases has shown accurate
convergence after only a few iterations [15].
2.2. Field case
A field test of the suggested approach was conducted at the city of Nesher, located
at the foot of Mount Carmel (Israel). The site is characterized by a 7 meter clay layer
overlying a 1 m clean sand layer, below which a stiff clay layer exists. The average
void ratio in the upper and lower clay layers is 1.6 and 0.95, respectively. The area is
known to be slightly over-consolidated (OCR from 1 to 3). The water table is found at
depth of 2m to 3m. In this field test a vertical drains installation machines (stitchers)
was used for fiber placement. The fiber optic cables were installed in a similar manner
to that of wick drains. Fig. 1a shows how this process was applied in the field. First, the
fiber was thread through the mandrel and fixed to the anchor plate. Secondly, the
mandrel was driven to the soil. Thirdly, some tension was applied to the fiber while the
mandrel was withdrawn. Fourthly, the fiber was pulled, fixed at ground level, and left
to rest over a significant period to allow the soil to contract around the fiber. This
installation was preformed 2 months prior to the surface loading.
The loading experiment was performed as follows. A 16 tonne truck loaded with 4
concrete blocks (each weighing 23.2kN, altogether 93kN) arrived at distance of
approximately 7 meters from the fiber. The blocks were then lifted from the truck and
placed around the fiber, roughly creating a 4 m2 loading area. The truck then brought 4
more concrete blocks which were placed above the previous ones, bringing the total
load to 186kN. Remaining in place, the truck then began unloading, transferring each
block at a time to the greatest distance possible on its other side (rather than loading the
blocks back on cargo cab). The duration of this whole process was 50 minutes, out of
which 20 minutes for the unloading procedure. Fig. 1b above shows the developed
vertical strain profile at (A) 93 kN loading, (B) 186 kN loading, (C) 93 kN unloading,
and (D) 0 kN unloading. The strain readings above depth of 2.5 meters exhibited
significant noise, possibly due to bad contact between the soil and the fiber at depth
above water table, and are therefore disregarded. Specific load-strain curves are shown
in Fig. 1b below for depth of 3 and 6 meters. Note that the conducted measurement
interval was every block loading/unloading. As can be seen, irrecoverable strains were
developed during the course of the loading in the clay layer, but not in the sand layer
(shown along line D).
The unloading section from the strain profile presented in Fig 1b was used as an
input for the suggested iterative calculation. It was assumed that the sand layer and the
clay above water table are governed by drained behavior, while the saturated clays by
undrained (zero volumetric strain) behavior. Note that when Poisson's ratio tends to be
0.5 the inverse problem solution becomes sensitive to errors associated with both the
conversion process and the optical fiber measurements. This is because any local strain
change cannot be obtained numerically without a significant modulus change due to the
requirement for zero volumetric strain. Analyses of synthetic cases have shown that
when a 0.5 Poisson's ratio case is back analyzed with a smaller Poisson's ratio, the
resulting Young's modulus agrees well with the accurate value, with an average over
estimation of roughly 1.5/(1+) where is the Poisson's ratio used in the inverse
problem solution. This approach was adopted in the analysis of the field data, using a
Possion's ratio of 0.4 instead of 0.5 for the saturated clay (together with an error
compensation of 7%, i.e. a factor of 1.4/1.5) and a Poissons ratio of 0.25 for the sand
layer.
Fig. 2 shows the resultant modulus profile (expressed as shear modulus). The shear
modulus of the top clay layers slightly increases with depth, from values of 20 MPa at
depth of 3 m to values of 30 MPa at depth of 6 m. A significantly higher shear
modulus, of 75 MPa, is attributed to the depth associated with the sand layer. The
aforementioned moduli values correspond fairly well to those obtained by correlations
using the in situ data from the nearby site investigation boreholes. Specifically, using
Seeds [16] SPT correlation for small strain shear stiffness (i.e. Gmax=100020(N1,60)
1/3
'm1/2) results in a value of 76 MPa for the sand layer. For the upper clay layer a
stiffness value in the range of 20 to 35 MPa is obtained when using the correlations
with void ratio and plastic index presented in [17] (page 232). The use of the same
correlations for the lower clay layer results in stiffness values in the range of 60 to 80
MPa. This overall good agreement suggests that the proposed approach has a potential
of being developed into a routine field investigation method.
Figure 1. (a) Field installation process; (b) strain measurements results.
Shear stiffness, G, (MPa)

0 20 40 60 80 100
0
4 CH
Depth (m)
Soil profile as observed in

nearby site investigation
6 boreholes
SP
8
CH
10
Figure 2. Shear modulus profile based on fiber optic measurements and iterative calculation.
3. Laboratory on-specimen strain measurements using fiber optics
In this section the potential of using high-spatial resolution fiber optic distributed
sensing for better laboratory soil characterization is briefly discussed. It follows recent
first step taken by the authors of application to on-specimen strain measurement for
uniaxial testing [18].
3.1. Demonstration on Acrylic glass sample
It was suggested in [18] to wrap a single fiber ribbon around the specimen twice (or
more) with different pitch values. This configuration ensures that the fiber ribbon
remains in tension throughout a compression test, allowing continuous on-specimen
strain measurements. Fig. 3 shows the conjuncture helical configuration as was
performed in [18] on acrylic glass specimen, where three different angles were
considered (2.28, 6.81 and 13.43 degrees). Essentially, only 2 pitch angles are required,
and the third (or more) may be used for validation. The helical envelop may add to the
confining stress, however this is usually very small and can be incorporated into the
interpretation of the results (for example in the presented case 0.004 kPa of every
1of vertical strain).
Figure 3. (a) Instrumented specimen; (b) Net view of the cylindrical specimen
High resolution strain may be obtained along each fiber (5, and can be used
either to evaluate the strains in the pitch direction through extrapolation, or integrated
for on-specimen deformation (which can also be extrapolated and used for further
analysis). For ideal sample deformation field, where the radial, tangential and vertical
displacement follow the relations of u r 6 r , u 0 and u z 6 z , respectively, the radial,
tangential and vertical strains (, rr, zz) can be given by:
sin 2 cos 2

sin 2 cos 2
rr ; zz (4)
sin
2
cos 2
1 2 1
sin cos 2
For uneven sample deformation, which often inherently occurs even in element
testing [19], the above expression may only be used with averaged values of and
to result in the averaged radial, tangential and vertical strains. Nonetheless, the
obtained spatially continuous strain profile facilitates, together with a more elaborate
interpretation procedure, the study of the full-field on-specimen strain profile, as
described and demonstrated in [18].
3.2. Future developments
The task of applying the conjuncture helical configuration presented in the previous
section to triaxial testing of soil elements is not as straight forward and requires some
applicative research. There are several candidate approaches for instrumentation of
fiber optics to the specimens. One is the fabrication of a special membrane embedded
with bare fiber optics, preferably in helical configuration. Another approach, which is
currently being considered and developed by the authors, is to 3D print a thin
interlocked double helix plastic spring, through which bare fiber could be be threaded.
The double helix spring can be vertically compressed to expand radially, creating a
space for a soil specimen to be inserted. After insertion, the spring would close on the
specimen as the vertical compression is released. Since the elaborate procedure of full-
field on-specimen strain evaluation requires knowledge of the exact position of the
fiber, it is believed that the 3D printed model may hold certain advantage over a fiber
optic embedded membrane.
4. Summary and conclusions
The paper briefly presented some of the recent advances made by the authors with
respect to applying high resolution fiber optic distributed sensing techniques to in situ
and laboratory material characterization.
It was demonstrated that the elastic stiffness moduli profile with depth can be
estimated efficiently by a rapidly converging recursive calculation which involves a
measured strain profile from a vertically positioned optical fiber under surface loading
or unloading. Soil stratification (clay-sand) was clearly evident from the resulting
stiffness profile, with higher stiffness values for the sand layer. A fair agreement was
found between the obtained stiffness values and those established by correlations with
standard in situ testing. While the suggest approach is promising, it requires further
research and improvements in order to be suitable for routine use. For example, many
improvements could be achieved with respect to the fiber installation process: [1] the
use of a scaled down (suitable for fibers) stitchers machine may facilitate an easier
installation especially in stiff soils, [2] the use of flexible filling materials may allow
immediate testing without the need to wait for contraction of the induced cavity. Fiber
optic measurements, resulting from load-unload sequences, include a wealth of
mechanical information. Interpretation of more advanced properties will be the topic of
future research. For example, it may be of interest to examine whether the suggested
approach could be used for evaluation of anisotropic elastic properties by combining
non-axial surface loadings.
Following the application to uniaxial testing by the authors [18], the applicability
of high resolution distributed fiber optic sensing for laboratory soil testing was also
discussed with a few possible future developments that would allow instrumentation of
fiber optics for triaxial soil testing. The suggested devices such as a 3D printed double
helix spring are currently under development.
One of disadvantages of distributed fiber optic sensing is the high cost of the
optical analyzer, albeit the low cost of the fiber optic cables. The current cost of a
distributed strain analyzer ranges between 70,000 to 150,000 USD. It is believed that
with decreasing price of optical analyzers, distributed fiber optic sensing will become
routine practice.
References
[1] R. Regier, and N.A. Hoult, Concrete deterioration detection using distributed sensors, Proceedings of
the ICE - Structures and Buildings (2014): 10.1680/stbu.13.00070.
[2] Y. Goldfeld, and A. Klar, Damage identification in reinforced concrete beams using spatially
distributed strain measurements, Journal of Structural Engineering 139 (2013): 04013013.
[3] T.E.B. Vorster, K. Soga, R.J. Mair, P.J. Bennett, A. Klar, and C. Choy, The use of fibre optic sensors to
monitor pipeline response to tunneling, GeoCongress (2006): pp. 1-6. doi: 10.1061/40803(187)33.
[4] M. Nikles, Long-distance fiber optic sensing solutions for pipeline leakage, intrusion and ground
movement detection, Proc. of SPIE 7316 (2009):731602-1.
[5] B. Simpson, N. Hoult, N., and I. Moore, Distributed Sensing of Circumferential Strain Using Fiber
Optics during Full-Scale Buried Pipe Experiments, Journal of Pipeline Systems Engineering Practice
(2015) , doi: 10.1061/(ASCE)PS.1949-1204.0000197:04015002.
[6] H. Mohamad, K. Soga, P.J. Bennett, R.J. Mair, and C. Lim, Monitoring Twin Tunnel Interaction Using
Distributed Optical Fiber Strain Measurements, Journal of Geotechnical and Geoenvironmental
Engineering 138 (2012):957967.
[7] A. Klar, I. Dromy, and R. Linker, Monitoring tunneling induced ground displacements using distributed
fiber-optic sensing, Tunnelling and Underground Space Technology 40 (2014):141-150.
[8] D. Hauswirth, A.M. Puzrin, A. Carrera, J.R. Standing and M.S.P Wan, Use of fibre-optic sensors for
simple assessment of ground surface displacements during tunnelling, Geotechnique 64 (2014):837-842
[9] X. Bao, and L. Chen, Recent progress in distributed fiber optic sensors. Sensors 12 (2012):8601-8639.
[10] A. Zadok, Y. Antman, N. Primerov, A. Denisov, J. Sancho, and L. Thevenaz, Random access
distributed fiber sensing. Laser & Photonics Reviews 6 (2012):L1-L5.
[11] R. Campanella, P. Robertson, and D. Gillespie, Seismic cone penetration tests. ASCE Geotechnical
Special Publication - Use of In-Situ Tests in Geotechnical Engineering (1986): 116-130.
[12] G. Martin and P. Mayne, Seismic at dilatometer test in Connecticut Valley varved clay. ASTM
Geotechnical Testing Journal 20 (1997): 357-361.
[13] D. Burmister, The general theory of stresses and displacements in layered systems. Part I. Journal of
Applied Physics 16 (1945):89-94.
[14] E. Levenberg, R.S. McDaniel, and J. Olek, Validation of NCAT structural test track experiment using
INDOT APT facility: final report, Joint transportation research program, Report FHWA/IN/JTRP-
2008/26, Purdue University.
[15] A. Klar, S. Uchida, E. Levenberg, In situ stiffness moduli profiling using high resolution fiber optic
distributed sending. Under review.
[16] H. Seed, R. Wong, I. Idriss, I., and K. Tokimatsu, Moduli and damping factors for dynamic analyses of
cohesionless soils, Journal of Geotechnical Engineering 112 (1986):1016-1032.
[17] Kramer, S. Geotechnical Earthquake Engineering. Prentice Hall, London. 1996.
[18] S. Uchida, E. Levenberg, and A. Klar, On-specimen strain measurement with fiber optic distributed
sensing, Measurement 60 (2015), 104113.
[19] W. Kim, J. Labuz, B. Chadbourn, M. Loken, Uniformity of axial displacement in element testing,
Geotechnical Testing Journal 31 (2008): 100722.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-390
Experimental investigation of wave

propagation in three dimensions in
unbounded particulate assemblies
Simon Hamlina, Erdin Ibraima, Martin Lings a, David Muir Wood b, Ignazio Cavarettac,
Jean Francois Camenend
a
University of Bristol, UK
b
University of Dundee, UK
c
University of Surrey, UK (formerly at University of Bristol)
d
University of Rijeka, Croatia (formerly at University of Bristol)
Abstract. Understanding wave propagation through soils is essential for site

response analysis in earthquake engineering, interpretation of geophysical surveys
and SASW (Spectral Analysis of Surface Waves), interpretation of laboratory
bender element tests, etc. Analysis of wave propagation has largely been based on
continuum descriptions and two dimensional analyses. This study presents recent
developments in multiaxial testing that permit the combination of laboratory
seismic testing with exploration of three-dimensional principal stress space. A
Cubical Cell Apparatus with bender-extender piezoceramic elements fitted in all
six faces are used so that wave propagation velocities of an analogue granular
material can be determined. The results of a first series of wave propagation tests
for a sample under isotropic confinement are presented.
Keywords. Laboratory, cubical cell apparatus, wave propagation
1. Introduction
Predictions of soil movements during construction and operation of many geotechnical

systems rely on accurate description of the response of soil at small strain levels. In the
laboratory the measurement of stiffness of soils in the small strain domain can be
achieved from static or dynamic laboratory tests. Static tests require precise control
and measurement of very small stress and strain increments while the dynamic tests
like resonant column [1], and the measurements of the body wave velocities within the
soil element [2-8] present real challenges in data interpretation ([9], among others)
which are based on the theory of wave propagation and require either prior knowledge
of a constitutive model for soil or manipulation of continuum medium theory
assumptions.
The measurement of body wave velocities is based on the generation at one end of
a confined soil sample of an elastic wave produced by piezoceramic bender/extender
(B/E) elements [10] which is received at the other end of the sample by other
piezoceramic elements, so that shear and compression wave velocities (and
corresponding stiffnesses) can be determined non-destructively and relatively fast.
Recent studies [11], including numerical analysis of wave transmission [9], leave
S. Hamlin et al. / Experimental Investigation of Wave Propagation in Three Dimensions 391
questions regarding the optimum nature of the input signal, identification of travel time,
effect of boundary reflections.
This paper explores some recent developments in multiaxial testing that permit the
combination of laboratory geophysics with exploration of principal stress space. A
Cubical Cell Apparatus (CCA) with piezoceramic elements fitted in the faces are used
so that wave propagation velocities of an analogue granular material made of uniform
spherical glass beads can be determined. The results of a first series of wave
propagation tests for a sample under isotropic confinement are presented.
1.1. Cubical Cell Apparatus
Incremental stiffness of soil depends on the nature of the particles, state of packing,
stress state and strain history. Soil fabric is usually anisotropic gravitational
deposition leads to initial anisotropy, and additional anisotropy is induced by stress
history. Most studies of soil stiffness have worked with the axisymmetric triaxial cell
which has two degrees of freedom and provides very limited possibilities of
exploration of stress space. In practice, elements of soil in the ground or around
geotechnical systems will experience variations of six independent stress variables. The
torsional hollow cylinder apparatus provides the possibility of controlling four of the
six degrees of freedom but at the expense of radial variation of stress and strain
components through the thickness of the wall of the sample. True triaxial apparatus
(Figure 1a) provide the possibility of controlling the magnitude of the three principal
stresses and strains under relatively well preserved uniformity of stresses and strains
within cuboidal sample, thus approaching a little closer to the six degrees of freedom of
a general stress state. The results presented here come from experiments conducted in
a flexible boundary cubical cell (Cubical Cell Apparatus - CCA, [12]) which has been
built at the University of Bristol [13]. A general view of the CCA is shown in Figure 1b,
while the sample (10x10x10 cm3 volume) contained in a latex rubber bag is shown in
Figure 2.
(a) (b)
Figure 1. (a) True triaxial apparatus: independent control of three principal stresses. (b) Cubical Cell
Apparatus.
392 S. Hamlin et al. / Experimental Investigation of Wave Propagation in Three Dimensions
Figure 2. Installation of the cubical sample into the apparatus.
The soil sample is supported by flexible top-hat shaped air-filled cushions

clamped into the sides of a cubical frame. Opposite pairs of cushions are connected so
that changes in the applied stresses are synchronized. Three LVDTs are used on each of
the six faces for the measurement of the displacements, while the pressures in the
cushions are measured with pressure transducers located immediately behind cushions.
1.2. Material and sample preparation
Stress controlled isotropic compression tests have been performed on dry ballotini glass
particles. The borosilicate glass ballotini with a solid density of 2.23 g/cm3 and with
nominal diameters between 2.4mm and 2.7mm were tested. This material is almost
mono-disperse and the coefficient of uniformity U = d60/d10 is 1.02. Detailed
characterisation of the shape of the particles as well as particle-scale mechanical
measurements have been conducted by [14].
The samples are prepared by pluviation into a membrane held by vacuum against
the sides of a cubical mould. A special pluviation device for fabrication of cubical
samples of granular materials designed also to allow the control of the height of fall has
been developed. A schematic view of its functioning principle is shown in Figure 3,
while detailed analysis of sample quality produced by the pluviation process can be
found in [15]. Once full, the top of the sample was carefully sealed and a thin tube was
subsequently used to establish a vacuum of about 50 kPa in the sample. This provided
sufficient effective stress within the glass material for it to be stiff enough to be lifted
in its containing membrane from the mould and locate it in the cubical cell (Figure 2).
Figure 3. Schematic functioning principle of the pluviation device.
1.3. Wave propagation technique
A piezoelectric material generates electrical output when subjected to mechanical

deformation or vice versa and changes its shape when an electrical field is applied to it.
Based on the technique developed by [10], pairs of Bender/Extender (B/E) elements
have been manufactured and mounted in all six faces of the CCA sample for body
wave propagation velocity assessment of the testing material. The B/E element consists
of two bimorph piezo-ceramic plates. Each B/E element is coated with epoxy and a pair
of them encapsulated in a T-shaped form into a cylindrical plug using resin (Figure 4a).
The dimension of the cantilever part of the B/E element, which penetrates the soil, is
approximately 5x5x0.51mm. A bender transmitter requires a three-wire parallel
connection while the bender receivers have a two-wire series connection and opposite-
sense polarised plates. By changing the parallel wired connection to series one, the
bender transmitter becomes an extender receiver and the same principle applies for the
series connection, when it is changed to parallel connection while keeping the opposite-
sense polarised plates, then the bender receiver is switched to an extender transmitter
and the resulting motion changes from cantilever bending to longitudinal tension and
contraction [10]. Electronic amplification signals, a wave form function generator and a
high resolution oscilloscope complete the measurement chain.
The installation of the B/E elements represents a critical and delicate step as
penetration of the latex membrane in the center of the faces through a T-shaped cut
having exactly the same size and shape as the B/E element should avoid any loss of the
vacuum. In the process, a sealing grommet guides the cylindrical body that houses the
T-shaped pair of B/E elements (Figure 4b), while several latex layers of adhesive
solution are then applied to the grommet-membrane boundary to ensure the sealing of
the sample.
In a normal wave propagation investigation all types of body and shear waves can be
transmitted along the principal sample directions.
(a) (b)
Figure 4. (a) Picture of the T-shaped B/E system. (b) Installation of the B/E element into the sample.
Experimental results are presented for a test on a pluviated assemblage of 2.4-2.7 mm

borosilicate glass beads loaded isotropically to 500 kPa in the CCA, with
bender/extender elements on all three axes of the CCA. The axial stress-strain results
are shown in Figure 5 and show that the pluviated cross-anisotropic fabric has slightly
rotated so that the Y horizontal - and Z vertical axes are broadly symmetrical
while the X - horizontal axis is different showing a marginally stiffer stress-strain
response; this is presumed to be caused by some disturbances in the sample during
bender/extender installation processes amplified by the coarse size of the tested
material. Further investigation on this issue, which hasnt been observed for fine
graded granular particles, is in progress.
Figure 5: Axial stress-strain curves during isotropic loading; direction of X, Y, and Z axes as shown in
Figure 1a.
Wave transmission tests were carried out during isotropic loading at 50 kPa (the
pressure of the internal vacuum applied to the sample during fabrication) and 100 kPa
followed by 100 kPa increments up to 500 kPa. The transmitted signal was a single
sinusoid with a function frequency of 15 kHz and a 270o phase angle, giving us a single
peak only (Figure 6). The wave travel time is measured as the time offset between the
transmitted peak and the first positive peak in the received signal, termed peak-to-peak
(P2P). A nomenclature for velocity of the form Vxy is adopted, where V denotes
velocity, the first subscript indicates the axis of wave propagation and the second
denotes the direction of polarisation (so both subscripts indicate the plane of wave
polarisation). Vxy therefore denotes an S-wave propagating in the direction of the X
axis and polarised in the XY plane. For P-waves the directions of propagation and
oscillation are the same, so the subscript is repeated (e.g. V xx).
P2P
Figure 6: Example transmitted and received S-wave propagating on the Y-axis and polarised in the XY
horizontal plane.
The calculated wave velocities as a function of mean effective pressure p are shown in
Figure 7 (P-waves) and Figure 8 (S-waves). As expected, the velocities increase with
the effective pressure following power laws. The effective medium theory (EMT) used
to describe sound propagation in granular media predicts that the ultrasound velocity V s
should depend on the applied consolidation stress according to the equation: Vs (p)n.
In the case of Hertzian contacts between elastic spherical particles, n = 1/6. However,
in disordered assemblies of spheres, the experimentally obtained power law exponent is
higher between 1/4 and 1/3, while [16] showed that for assemblies of particles with
conical contacts instead of spherical, n = 1/4. The P-wave results show Vyy and Vzz are
similar while Vxx is slightly greater, which conforms to the stress-strain results in
Figure 5 where the X-axis shows a stiffer response. However, the power coefficients
are similar, around 0.20. The results for S-waves, however, are more complex most
probably due to some dependence of the S-wave velocity on the sample fabric
properties in the propagation and polarisation directions. It appears that the fabric in the
propagation direction dominates over the polarisation direction, as Vxz and Vxy are
fastest while Vyx is slowest. The velocities Vyz and Vzy, for waves polarised in the
symmetrical YZ plane, are nearly identical and fall in the middle of the range of the
others. An electrical fault prevented the Vzx velocity from being obtained. The power
coefficients are around 0.25 for V xy, Vyz and Vzy, and 0.21 for Vxz and Vyx (Figure 8).
800
700
Vxx
600 y = 219,94x0,1968
500 R = 0,9991
Vp (m/s)
Vyy
400
y = 212,16x0,1918
300 R = 0,9986
200
Vzz
y = 213,37x0,1931
100 R = 0,9985
0
0 100 200 300 400 500
p' (kPa)
Figure 7: P-wave velocities with p during isotropic loading.
500
450 Vxy
y = 88,639x0,26
400 R = 0,9985
350 Vxz
y = 118,35x0,2132
300
Vs (m/s)
R = 0,9985
250 Vyx
y = 105,6x0,2102
200
R = 0,9976
150 Vyz
100 y = 90,752x0,2461
R = 0,9974
50 Vzy
0 y = 92,698x0,2444
0 100 200 300 400 500 R = 0,9962
p' (kPa)
Figure 8: S-wave velocities with p during isotropic loading.
3. Conclusions
This study presents recent developments in multiaxial testing that permit the
combination of laboratory seismic testing with exploration of three-dimensional
principal stress space. A Cubical Cell Apparatus with bender-extender piezoceramic
elements fitted in all six faces are used so that wave propagation velocities of an
analogue granular material can be determined. Understanding wave propagation
through soils is essential for site response analysis in earthquake engineering,

interpretation of geophysical surveys and SASW (Spectral Analysis of Surface Waves),
interpretation of laboratory bender element tests, etc. The results of a first series of
wave propagation tests for a sample of coarse, very poorly graded material under
isotropic confinement showed some correlation between wave velocities and the
sample stress-strain response. While P-wave velocity shows a direct relationship, with
the stiffer axial stress-strain response corresponding to higher wave velocity, the S-
wave behaviour appears to be controlled more by the fabric properties in the
propagation direction than those in the polarization direction. The power law that links
wave velocity to pressure shows power coefficients of 0.20 for P-waves and between
0.21 and 0.25 for S-waves.
Acknowledgements
Funding for this research was provided via EPSRC UK grant EP/G064180/1.
References
[1] B.O. Hardin and F.E. Richart. Elastic wave velocities in granular soils, Journal of Soil Mechanics &
Foundations Div. ASCE, 89 (1963)
[2] D. J. Shirley and L. D. Hampton. Shear-Wave Measurementsin Laboratory Sediments. Journal of the
Acoustical Society of America, Vol. 63, No. 2, (1978), 607613
[3] P.J. Schultheiss. Simultaneous measurement of P and S wave velocities during conventional laboratory
soil testing procedures. Marine Geotechnology, Vol 4, no 4, (1980), 343-367
[4] R. Dyvik and C. Madshus. Laboratory measurements of Gmax using bender elements. Proc. ASCE Annual
Convention: Advances in the art of testing soils under cyclic conditions, Detroit, Michigan, (1985),
186-197
[5] G. Viggiani and J.H. Atkinson. Interpretation of Bender Element Tests". Gotechnique, Vol. 45, No. 1,
(1995), 149154
[6] E.G.M. Brignoli, M. Gotti and J.H. Stokoe, Measurement of shear waves in laboratory specimens by
means of piezoelectric transducers. Geotechnical Testing Journal, 19, (1996), 284-397
[7] D.S. Pennington, D.F.T. Nash, and M.L. Lings. Anistropy of G0 Shear Stiffness in Gault Clay.
Gotechnique, Vol. 47, No. 3, (1997), 391398.
[8] S. Yamashita, T. Kawaguchi, Y. Nakata, T. Mikami, T. Fujiwara and S. Shibuya. Interpretation of
international parallel test on the measurement of Gmax using bender elements" Soils and foundations,
49(4), (2009), 631-650.
[9] M. Arroyo, D. Muir Wood, P. Greening, L. Medina and J. Rio. Effects of sample size on bender-based
axial G0 measurements". Gotechnique, 56(1), (2006), 39-52.
[10] M.L. Lings and P.D. Greening. A novel bender/extender element for soil testing. Gotechnique, 51(8),
(2001), 713-717
[11] J.-S. Lee and J.C. Santamarina. Bender Elements: Performance and Signal Interpretation. Journal of
Geotechnical and Geoenvironmental Engineering, 131(9), (2005),1063-1070
[12] Ko, H.-Y. & Scott, R.F., 1967. Deformation of sand in shear. Journal of Soil Mechanics and
Foundations Div ASCE
[13] Sadek, T., 2006. The Multiaxial Behaviour and Elastic Stiffness of Hostun Sand, PhD thesis. University
of Bristol
[14] Cavarretta, I., OSullivan, C., Ibraim, E., Lings, M., Hamlin, S., & Muir Wood, D., 2012.
Characterization of artificial spherical particles for DEM validation studies. Particuology, 10(2),
pp.209220
[15] J.F. Camenen, I. Cavarretta, S. Hamlin and E. Ibraim. Experimental and numerical assessment of a
cubical sample produced by pluviation. Gotechnique Letters, 3(April-June), (2013), 44-51
[16] J.D. Goddard. Nonlinear elasticity and pressure-dependent wave speeds in granular media. Proceedings
Royal Society of London A: Mathematical and Physical Sciences 430, (1990), 105131
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-398
Frequency domain method in bender

element testing experimental
observations
Javier CAMACHO-TAUTAa, Hassan ALIb, Giovanni CASCANTE1,b, and Antnio
VIANA DA FONSECAc
a
Faculty of Engineering, Nueva Granada Military University, Bogota D.C., Colombia
b
Department of Civil and Environmental Engineering, University of Waterloo
c
Faculty of Engineering, University of Porto, Porto, Portugal
Abstract. In bender element (BE) testing, shear-wave velocities are measured

using two main methods: time (pulse excitation) and frequency (sine sweep
excitation) domain analysis. The time domain analysis is simple but requires
subjective judgment for the non-automatic determination of the arrival time.
Conversely, the frequency domain analysis can be performed automatically.
However, its results show high variability because they are affected by the
selection of different variables such as the frequency content of the excitation, the
resonant frequency of the BE system, and the resolution of the frequency domain
analysis. The reliability of the frequency domain method is studied in this paper by
using experimental results from bender element and resonant column tests.
Laboratory results show that amplification peaks in the transfer function of the BE
system can significantly affect the shear wave velocity measurements (up to
300%). In addition, the high correlations coefficients typically observed at high
frequencies are finally explained. A modified frequency domain analysis is
proposed to enhance the reliability of the method using a high frequency
narrowband excitation outside the resonant peaks of the BE system. Results from
the modified frequency domain analysis tests show less than 10% error in the
measurement of the shear wave velocity with respect to the RC results.
Keywords. Bender element test, frequency domain analysis, dynamic soil

properties.
1. Introduction
Bender element (BE) testing is a laboratory method to measure the shear wave velocity
of soils. The technique was proposed for geotechnical applications by Shirley and
Shirley and Hampton [1, 2]. Due to its nondestructive nature, the method has been
implemented in a variety of geotechnical devices [3]. The operation principle,
experimental procedure and interpretation of results are described in detail by Dyvik
and Madshus [4]. Factors as quality of manufacturing and installation of the bender
elements, coupling and alignment in the soil, near-field effect [5], geometric and size of
specimen [6, 7], and boundary effects [8], affect the reliability of results. Given the
1
J. Camacho-Tauta et al. / Frequency Domain Method in Bender Element Testing 399
small size of the laboratory specimen, the travel time of shear waves is a crucial issue.
Viggiani and Atkinson [9] found that the main source of error is related to the
estimation of arrival time in BE testing.
Due to the variability in the results, mainly depending on the criterion to identify
the arrival time in the bender element receiver, Jovicic et al. [10] recommends the use
of various methods. Arulnathan et al. [8] analyzed different methods for the evaluation
of the travel time using different characteristic peaks of input and output signals, cross-
correlation of input and output signals, and phase analysis of the cross-power spectrum.
They showed that the first arrival of the received signal are more consistent than the
results from the cross-correlation or phase analysis of the cross-power spectrum.
Greening and Nash [11], developed a methodology in which the evaluation of the
travel time is made in frequency domain. They used a broadband sinusoidal sweep as
excitation signal. The travel time is computed from the slope of the unwrapped phase
angle of the cross-power spectrum. An analysis framework that combines time-domain
and frequency-domain methods was proposed by Viana da Fonseca et al. [12]. First,
sine-wave pulses at different frequencies are used to evaluate the first arrival. Then,
sinusoidal sweeps are applied and the results are analyzed for different frequency
windows. However, the method is still dependent on the user criteria to choose the
most probable travel time.
The international parallel test carried out by 23 laboratories shows that the more
used testing method is the sine pulse test, Yamashita et al. [13]. In addition, the more
consistent interpretation is made in time domain using the start-to-start method, in
which the travel time is the time difference between the starting of the input sine pulse
and the corresponding point in the received wave. In contrast, frequency domain
method based on the phase of the cross spectrum was the less confident of the two
methods evaluated. Two main causes observed dispersion in the estimation of shear
wave velocity by the frequency domain method are: the assumption that transformation
between input and output signals depends only on the soil properties [14] and the
undesirable effect of resonant frequencies on the slope of the phase angle function [15].
The frequency domain method assumes that transmitter-soil-receiver is a system in
which the input is the electric waveform used to excite the transmitter, while the output
is the electrical signal generated by the receiver. However, there are different sequential
complex transformations between these two signals [14]. Pallara et al. [16] show a
comparison between the pulse signal used to excite the bender element transmitter and
its actual movement measured by a laser vibrometer. They found a time delay between
them and the displacement of the transmitter is non-symmetric. The same technique
was used by Rio [17], finding that under free conditions, the response of the transmitter
is dominated by its own resonant frequency; while under embedded conditions the
natural frequency increases and the magnitude of the oscillation decreases due the
stiffening of the medium. As result, there are suggestions of using bender elements
only as receivers in combination with other sources of perturbation [16, 18, 19].
The transfer function of the BE system depends on the confining stress, not only
due to the change in the soil stiffness, but also because changes in resonant frequencies
of bender elements and other components of the system. The peaks in the amplitude of
the transfer function move to higher frequencies as the soil stiffness increases [15]. On
the other hand, in frequencies near resonant peaks, the slope of the phase function is
not constant, causing high variations in the estimated shear wave velocity [20, 21].
Miniature accelerometers located inside the soil specimen have shown the complex
patterns of wave propagation in bender element testing [22].
400 J. Camacho-Tauta et al. / Frequency Domain Method in Bender Element Testing
In this paper, experimental evidence of the performance of a BE system in a dry

sand specimen is analyzed. The soil specimen is installed in a resonant column device
and therefore reference shear wave velocities are available for comparisons.
Experimental results identify the main causes affecting the frequency domain method.
The modified frequency domain method is used to evaluate the shear wave velocities
from high frequency excitations and results are compared with resonant column tests.
2. Estimation of the shear wave velocity in the frequency domain analysis
The shear wave velocity by bender element testing is not always clear. Different
interpretation methods have been proposed. Currently, there is no single and accurate
technique that can be adopted as standard. In the frequency domain analysis, a sine
sweep signal is used as excitation of the transmitter and the shear wave velocity is
computed through the transfer function of the system. Based on the phase delay method
[23], Greening and Nash [11] proposed an automatic method by transformation of time
signals using Fourier analysis. The group wave velocity (Vgr) is given as the derivative
of the frequency (f) with respect to the wave number (k) as presented by Eq. (1)

(1)
!
For a constant distance between the bender elements (Ltt), the wave number can be
expressed in terms of the phase lag () between the input and output signals (k = /Ltt)
and Eq. 1 can be expressed in Eq. (2),

#$%% (2)
"
The phase angle as a function of frequency is obtained by the transfer function or

the cross-spectrum between the two signals. The transfer function H(f) between input
signal x(t) and output signal r(t) is given in terms of the average cross power spectral
density Grx(f) and the auto spectral density Gxx(f) as presented in Eq. (3). Averaging
spectral densities reduces the effect of noise on the measurements. The coherence
function (f) shown in Eq.(4) is a measure of the linearity of the system
,-
..........
+ ( *
&' ()* .......... (3)
+-- ( *
.......... 2
1+
,- ( *1
/ 0 ()* .......... ......... (4)
+-- ( *3+,, ( *
Where Grr(f) is the auto spectral density of the output signal. The slope of the
unwrapped phase function is used to evaluate the group wave velocity by mean of Eq.
2. In general, the unwrapped phase angle function does not have perfectly constant
slope even for a coherence close to unity. Viana da Fonseca et al. [12] estimated the
travel time by plotting the travel time as a function of frequency for different moving
frequency windows. For each frequency window, the slope and its correlation
coefficient are obtained by linear regression.
3. Methods and Materials
Figure 1 shows a schematic of the Stokoe-type resonant-column system [24]. The RC

system is composed of a dynamic signal analyzer (HP-35670A) connected to a power
amplifier (Bogen, GS-250) that produces a torsional excitation to the specimen by four
pairs of coils. An accelerometer (PCB 352A78) attached to the driving plate measures
the response of the system. The accelerometer is fed by a power source unit (Dytran
4121). Both signals in the coils and the accelerometer are filtered and amplified by a
signal conditioning system (Krohn-Hite 3384) and monitored by a digital oscilloscope
(HP-54645A). The Fourier spectra are computed in real time by the dynamic signal
analyzer. Base and top cap of the RC system are equipped with bender elements. The
transmitter excited by an arbitrary wave generator (Physical Acoustics Corp) is located
at the base pedestal. The output signal of the receiver located at the top cap is
conditioned using a filter-amplifier (Krohn-Hite 3384).
At the specified confining stress of 50, 100, 200, and 400 kPa, RC and BE test are
done sequentially. Frequency domain test is done by exciting a sine sweep from 0 to 20
kHz in traditional method and from 20 kHz up to 50 kHz with an interval of 3.2 kHz in
the modified frequency domain method. Results from sine sweep tests are analyzed and
the shear wave velocity is calculated using the moving frequency window method. In
the frequency window method, different frequency windows are selected to find the
best coefficient of correlation that matches with the shear wave velocity from RC test.
A soil specimen of Toyoura sand (Gs=2.635; Dmax=0.425 mm; Cu=1.6; Cz =0.97;
emax=0.968; emin=0.628; e=0.732) was prepared by the dry compaction method. The
initial specimen dimensions were verified by using a digital caliper. The tip-to-tip
distance is calculated by subtracting the height of the bender elements from the current
specimen length. For the modified frequency domain method, a sand specimen, Barco
71 was prepared using the dry compaction method. The parameters for the Barco 71
sand are, Gs = 2.65, Cu = 1.78, Cz = 1.20, emin = 0.53, and emax = 0.75, respectively. The
index and mechanical properties of the specimen are: height, H = 14.975 cm; mass, M
= 928.0 gm, void ratio, e = 0.53, density, = 1.6 g/cm3, and Elastic modulus, E = 89
MPa for = 0.2 respectively.
Figure 1. Schematic description of the resonant-column device with bender elements.

4. Experimental Results
Figure 2a show the typical signals obtained by the BE receiver in the bender
element test for different confining stresses. The characteristics of the input sine sweep
signal are: 10 V amplitude, 12 ms duration and 1 to 20 kHz frequency content. Signals
were processed in frequency domain and transfer functions between transducers were
computed. Figure 2b shows the transfer functions between the electrical signal sent to
the transmitter and the output of the receiver for the different stress conditions. In
general, the shape of each transfer function is approximately preserved through the
change in confinement. The main peaks move to the right as the confinement increases.
400 kPa
Normalized Transfer Function, H xr [.]

400 kPa
200 kPa
Normalized Amplitude [.]
100 kPa 200 kPa
50 kPa 100 kPa
Input
50 kPa
0 5 10 15 0 5 10 15 20
Time [ms] Frequency [kHz]
Figure 2. a) Time signals obtained from frequency sweep (0 20 kHz) and b) Transfer function between
electrical input and receiver BE.
The moving frequency window method is used to obtain a best estimation of the
shear wave velocity. These values were compared against the values obtained by the
resonant-column test. The error varied from -6.7% for 50 kPa to -15.9 for 400 kPa. All
bender element estimations were lower than the resonant-column measurements. The
causes of this discrepancy are studied in follows. The estimation of the shear wave
velocity exhibits high oscillations around the peak frequencies indicated in Figure 2b.
This evidence suggests that the resonant frequency of the bender elements affects the
consistency of the frequency domain method.
The zone with the best correlation coefficient is located in a plateau located far to
the resonant frequency. The resonance has small influence and therefore the estimation
is acceptable. As the confinement increases, the resonant frequency of the bender
element increases affecting a wider range and the error in the shear velocity estimation
increases.
Figure 3 shows that in general, estimations of the shear wave velocities computed
between different pairs of sensors are lower than the resonant column value. The
difference increases as confining pressure increases. The estimations were more
consistent until 200 kPa of confining stress.
4.1. Results from modified frequency domain method
Results of Figure 3 show the shear wave velocity estimation for excitation
frequency up to 20 kHz. In this range, only the lower confinement (50 kPa) has good
agreement with resonant column test results. Therefore, this section presents the results
obtained from high frequency excitation (20 to 50 kHz).
400
1:1 line
BE
350
VS by BE test [m/s]
300
250
200
200 250 300 350 400
VS by RC test [m/s]
Figure 3. Comparison of shear wave velocity from RC and BE (frequency domain method) tests.
Frequency sweep results for the test are shown in Figure 4a; while the transfer
function between the electrical input and output signals for each confinement are
presented in Figure 4b. At low confinement, the main energy in the frequency spectrum
is up to 20 kHz; while for higher confinement, the frequency peaks shifts to higher
frequencies.
400 kPa 400 kPa

Normalized Transfer Function, H xr [.]
200 kPa
Normalized Amplitude [.]
100 kPa 200 kPa
50 kPa
100 kPa
50 kPa
0 5 10 15 0 10 20 30 40 50
Time [ms] Frequency [kHz]
Figure 4. a) Time signals obtained from frequency sweep (0 50 kHz) and b) Transfer function between
electrical input and receiver BE.
Figure 5 present results of shear wave velocity from RC and BE test. The under-
estimation of results in case of 0-52 kHz window can be observed. The values are 28%
lower for low confinement and reduce to 8% for higher confinement. All three spans
show good match of shear wave velocity with RC values while results for the 29.6
32.8 kHz window show the best match. The best match obtained in these cases is due
to higher coherence, in the range of 0.8 to 1. Although, a coherence of one was
obtained for the 32.8 36 kHz window at 400 kPa; the shear wave velocity is lower
than the RC value. This could be attributed to change in transfer function amplitude,
which is -4dB peak value as compared to -30dB for all other spans. Also, as the peaks
in the transfer function move to higher frequencies for higher confinement, the shear
wave velocity should be searched at the flat portion of the transfer function. Due to the
limitation of the instrument, frequency excitation higher than 50 kHz could not be
generated. Further, investigation is required in this case.
350
1:1 Line
0 kHz - 52 kHz
300 23.2 kHz -26.4 kHz
29.6 kHz - 32.8 kHz
32.8 kHz- 36 kHz
VS by BE test [m/s]
250
200
150
100
100 150 200 250 300 350
VS by RC test [m/s]
Figure 5 Comparison of shear wave velocity from RC and BE (modified frequency domain method) tests
The variation in the mean shear wave velocities for values obtained in Figure 5are
presented in Figure 6. The variation is less than 10% for 50, 100, and 200 confinements,
respectively. The variation is more than 20% for effective confinement of 400. If the
value shown in circular marker is removed from the mean value, the error reduces to
11% showing the validity of the method at higher confinement.
350
400 kPa
300
VS by BE test [m/s]
250 200 kPa
100 kPa
200
50 kPa
150
100
100 150 200 250 300 350
VS by RC test [m/s]
Figure 6 Variation of shear wave velocity from the mean values obtained in Figure 5.
Conclusions
The transfer function between transmitter and receiver has a number of peaks. These
peaks are due three types: resonant frequencies of the BE, mode vibrations of the soil
specimen and wave reflections against ends.
As the soil stiffness increases, the transfer function gradually changes, being the
main characteristic of this change, the displacement to the right (it moves to higher
frequencies).
The travel time estimation should be done in a frequency range where no peaks
appear. This ideal condition is not possible in actual testing. In consequence, an
approximate estimation could be obtained if the frequency range is located as far as
possible to the main peaks. For example, good estimations were obtained for low
confining pressures, where the main peaks left a sufficient frequency range to the right.
In contrast, for the higher confinement, the estimation fails because these peaks are
close to the maximum experimental frequency.
The selection of the frequency range should take into account the geometry of the
specimen. The vibration modes of the cylinder depend on the inverse square of the
height. Therefore, short specimens require the use of higher frequency in order to allow
enough frequency range free of the influence of the peaks.
The shear wave velocity obtained from modified frequency method matches well
with the values from RC and BE test. The error in the values is less than 10 % for
confinement up to 200 kPa. Further, investigation is needed for confinement greater
than 400 kPa.
Acknowledgements
This paper was developed under the research activities of ICIST of IST and
CONSTRUCT from FEUP, supported by funding from FCT (Portuguese Science and
Technology Foundation). Javier Camacho-Tauta acknowledges the support from Nueva
Granada Military University, Research Project INV-ING-1765. This research is part of
a study on non-destructive testing of infrastructure. Support is also provided by the
Natural Sciences and Engineering Research Council of Canada (NSERC).
References
[1] D. J. Shirley, An improved shear wave transducer, J. Acoust. Soc. Am. 63-5 (1978), 16431645.
[2] D. J. Shirley and L. D. Hampton, Shear-wave measurement in laboratory sediments, J. Acoust. Soc.
Am. 63-2 (1978), 607613.
[3] C. Ferreira, The use of seismic wave velocities in the measurement of stiffness of a residual soil, PhD
Thesis, Faculty of Engineering, University of Porto, Porto, 2008.
[4] R. Dyvik and C. Madshus, Lab measurements of Gmax using bender elements, in Proceedings of the
ASCE Annual Convention on Advances in the Art of Testing Soils under Cyclic Conditions. (1985),
186-196.
[5] E. G. Brignoli, M. Gotti and K. H. Stokoe, Measurement of shear waves in laboratory specimens by
means of piezoelectric transducers, Geotech. Testing J. 19-4 (1996), 384397.
[6] J. Rio, P. Greening and L. Medina, Influence of sample geometry on shear wave propagation using
bender elements, in Deformation Characteristics of Geomaterials. (2003), 963967.
[7] M. Arroyo, D. Muir Wood and P. D. Greening, Source near-field effects and pulse tests in soils
samples, Gotechnique 53-3 (2003), 337345.
[8] R. Arulnathan, R. W. Boulanger and M. F. Riemer, Analysis of bender element tests, Geotech. Testing
J. 21-2 (1998), 120131.
[9] G. Viggiani and J. H. Atkinson, Interpretation of bender element tests, Gotechnique 45-1 (1995),
149154.
[10] V. Jovii, M. R. Coop and M. Simi, Objective criteria for determining Gmax from bender element
tests, Gotechnique 46-2 (1996), 357362.
[11] P. D. Greening and D. F. T. Nash, Frequency domain determination of G0 using bender elements,
Geotech. Testing J. 27-3 (2004), 17.
[12] A. Viana da Fonseca, C. Ferreira and M. Fahey, A framework interpreting bender element tests,
combining time-domain and frequency-domain methods, Geotech. Testing J. 32-2 (2009), 117.
[13] S. Yamashita, T. Kawaguchi, Y. Nakata, T. Mikami, T. Fujiwara and S. Shibuya, Interpretation of
international parallel test on measurement of Gmax using bender elements, Soils and Foundations 49-4
(2009), 631650.
[14] J.-S. Lee and J. C. Santamarina, Bender elements: performance and signal interpretation, Journal of
Geotechnical and Geoenvironmental Engineering, ASCE 131-9 (2005), 10631070.
[15] G. Alvarado and M. R. Coop, On the performance of bender elements in triaxial tests, Gotechnique
62-1 (2012), 117.
[16] O. Pallara, M. Mattone and D. C. F. Lo Presti, Bender elements: bad source - good receiver, in
Deformational Characteristics of Geomaterials. (2008), 697702.
[17] J. Rio, Advances in laboratory geophysics using bender elements, PhD Thesis, University College
London, University of London, London, 2006.
[18] J. Camacho-Tauta, J. A. Santos and A. Viana da Fonseca, Two-Bender Receivers Frequency Domain
Analysis in Resonant-Column Tests, in GeoShanghai, Geotechnical Special Publication 240
Advances in Soil Dynamics and Foundation Engineering. (2014), 72-82, 2014.
[19] J. Camacho-Tauta, J. A. Santos and A. Viana da Fonseca, Evaluation of the small-strain shear modulus
using two bender-receivers and a controlled external source, in Deformational Characteristics of
Geomaterials. (2008), 703-709.
[20] J. Camacho-Tauta, G. Cascante, J. A. Santos and A. Viana da Fonseca, Measurements of shear wave
velocity by resonant-column test, bender element test and miniature accelerometers, in Proceedings of
the 2011 Pan-Am Geotechnical Conference. (2011).
[21] M. A. Styler and J. A. Howie, Combined time and frequency domain approach to the interpretation of
Bender-Element tests, Geotech. Testing J. 36-5 (2013), 111.
[22] J. Camacho-Tauta, G. Cascante, A. Viana da Fonseca and J. A. Santos, Time and frequency domain
evaluation of bender element systems, Gotechnique In Print-(2015),
[23] E. A. Kaarsberg, Elasic-wave velocity measurements in rocks and other materials by phase-delay
methods, Geophysics 40-6 (1975), 955960.
[24] G. Cascante, J. Vanderkooy and W. Chung, A new mathematical model for resonant-column
measurements including eddy-current effects, Can. Geotech. J. 43-4 (2005), 121135.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-407
An Alternative Shear Strength Test for

Saturated Fine-Grained Soils:
Preliminary Results
Kamil KAYABALIa,1, Nihan Pinar YILMAZb, Murat MOLLAMAHMUTOGLUc
a
Geological Engineering Department, Ankara University, Turkey
b
General Directorate of Mineral Research and Exploration, Ankara, Turkey
c
Civil Engineering Department, Gazi University, Ankara, Turkey
Abstract. A proper definition of the effective shear strength parameters is a major

concern for long-term evaluation of loads on fine-grained soils. The most common
technique to measure such entities is the conventional triaxial shear (3X) test.
Numerous researchers outlined the limitations, uncertainties and operator effects
pertinent to this test. Furthermore, the consolidated 3X tests take extremely long
time to complete. Recently introduced in soil mechanics, the reverse extrusion
(RX) test was shown to be capable of measuring the undrained shear strength of
fine-grained soils with a reasonable degree of success. The hypothesis of our
investigation is that the RX test should also be able to provide drained (or
effective) strength parameters if it is facilitated to measure the pore-water pressure
(PWP). A new instrument was developed for this purpose. Three remolded soil
samples were tested using both the conventional 3X and RX tests. A comparison
between the effective internal frictional angles from 3X-CU tests and the effective
extrusion pressures from the RX test test reveals a remarkably good match. The
investigation is in progress to increase the number of remolded soil samples and to
extend the research to include natural soil samples.
Keywords. Reverse extrusion test, shear strength, pore-water pressure, fine-

grained soils, consolidated-undrained triaxial test, effective friction angle.
Introduction
The measurement of the shear strength of fine-grained soils is crucial. Depending on

the specific design requirements, effective shear strength parameters are often needed.
The most common laboratory methods for determining the effective shear strength
parameters are the consolidated-undrained triaxial shear (3X-CU), consolidated-
drained triaxial shear (3X-CD), the direct shear (DS) and the direct simple shear test
(DSS). Nevertheless, direct shear tests do not include PWP measurements.
Any of the 3X tests require many corrections and are prone not only to the sample
disturbance and operator effects, but also a number of uncertainties. In summary, they
are highly complex tests that require long time to complete. The major constraints of
the triaxial shear tests can pose difficulties in regard to achieving full saturation,
membrane effects, filters, drainage complexities, rate effects, and the uncontrolled
post-failure behavior of the specimen due to the sudden failure of the test specimen [1].
Corresponding author. Geological Engineering Department, Ankara University, Ankara, Turkey, 06100,
[email protected]
408 K. Kayabali et al. / An Alternative Shear Strength Test for Saturated Fine-Grained Soils
Triaxial shear tests involve instrumental uncertainties and difficulty with sample
preparation and saturation, and they require expertise for evaluating the test results;
these effects bring about some limitations with the test and can introduce erroneous
results [2]. The soil type, stress history and the selected test procedure are also part of
the constraints of the test system. In addition, the effective stress parameters may
involve errors owing to the selection of inappropriate loading rates [3]. Concerning the
length of time required to run a complete set of tests, OKelly and Zhang [4] showed
that some consolidated 3X tests conducted on peats took as long as 43 days.
The extrusion tests were used in soil mechanics to determine the Atterberg limits
and the undrained shear strength. The extrusion test was first brought into attention by
Timar [5] as direct extrusion in relation to soil consistency. Later, Whyte [6] and
Medhat and Whyte [7] employed the reverse extrusion (RX) test in relation to the
undrained shear strength of fine-grained soils. The RX test can be utilized in a
reasonable success to determine the Atterberg limits [8]. A recent study by Kayabali
and Ozdemir [9] showed that the RX test provides reliable undrained shear strength
(su) values for fine-grained soils. We hypothesize that if an RX test is capable of
providing shear strength values for undrained conditions, it should also be able to
provide a shear strength to simulate drained conditions just by attaching a PWP sensor.
The scope of this investigation for the moment is to test the usability of the RX
test as a shear strength test for fine-grained soils to obtain the effective stress strength
parameter for remolded soils. The ultimate goal is to match the results obtained from
the conventional 3X-CU tests with those from the RX-PWP test, which was first
introduced in this investigation. The applicability of the newly developed technique on
natural soils will depend on the degree of success obtained in this investigation.
Materials
Three types of soils were used in this investigation. They were all remolded samples
that were sieved through a #40 mesh prior to conducting laboratory experiments. Their
Atterberg limits are provided in Table 1.
Table 1. Atterberg limits of soil samples employed for the study.
No. LL PL PI -#40 (%) -#200 (%)

AK 74.1 35.6 38.5 100 86
37A 92.5 26.5 66.0 100 78
73A 40.3 19.1 21.2 100 41
One of the major instruments utilized for the purpose included a set of the
conventional triaxial shear apparatus that was capable of handling the consolidated-
undrained (CU) tests. The second one was the RX test apparatus equipped with a PWP-
measurement unit (Figure 1). The details of the RX-PWP setup are illustrated in
Figure 2.
K. Kayabali et al. / An Alternative Shear Strength Test for Saturated Fine-Grained Soils 409
Figure 1. Left: The RX-PWP unit is on the loadframe. Right: Pop-up view of the RX-PWP unit.
Methods
The present investigation includes two major tests. The first one is the reverse
extrusion test. The sample preparation and testing procedures are as follow. The major
concern for this test is the determination of the water content (w) of the test specimen.
The range of the water content is determined such that it falls into the range between
the plastic limit (PL) and a point of PI/2, where PI is the plasticity index. It should be
noted that the water contents between the liquid limit (LL) and PI/2 are considerably
high, which do not permit the construction of artificial specimens for the 3X-CU tests.
Therefore, four to five predetermined water contents are selected for the RX-PWP tests
in the interval mentioned above. For this, approximately 120 grams of -#40 dry soil
material is taken and wetted at the first predetermined water content. The test specimen
is mixed by hand until a homogeneous mixture is attained. Before running the RX-
PWP test, the porous stone is saturated and vacuumed to release the air bubbles inside
it. The porous stone is placed inside the top cap and a filter paper is placed the between
the porous stone and the specimen. Then, the top cap is attached to the container and is
tightly rotated to prevent possible water leaks. An O-ring is placed between the
container and the top cap. A wet soil chunk about the size of a small walnut is dropped
into the container while the top cap is in place. The soil chunk is tapped gently by a
rammer to ensure that the large cavities in the specimen are eliminated. This procedure
is repeated several times until the whole wet specimen is placed in the container part of
the RX-PWP assembly. At the end of this procedure, a gently compacted soil specimen
is obtained.
Figure 2. a) Schematical illustration of the general setup for the RX-PWP instrument. b) Pop-up view of the
dashed rectangle in (a). c) View of the water channels inside the top cap. d) Top view of the top cap.
Explanations for the numbers: 1. Loadcell 2. Container 3. Rammer 4. Die orifice (diameter = 6 mm) 5.
Alignment unit for the rammer 6. Soil specimen [diameter (D) = 38 mm, height (H) = 1D to 1.5D] 7. Top
cap 8. Porous stone 9. Valve 10. Transducer for PWP measurement 11. Water channels
The diameter of the container is 38 mm; the height of the specimen H is kept
between 1D < H < 1.5D to obtain a reliable test result. The water channels at the
bottom surface of the top cap are filled with water and carefully vacuumed to remove
air bubbles. Following the specimen preparation, the RX-PWP assembly is placed into
a loadframe as shown in Figure 1, which can be found in any of the common soil-
mechanics laboratories. The specimen is compressed by rammer at a rate of 0.5
mm/min. The rammer hosts a die orifice of 6 mm at the center (Figure 2a). As the
compression progresses, the force applied to the specimen (P, in Newtons) and the
PWP (u, in kPa) increases. The force and pore-water pressure are recorded
simultaneously. The compression continues until a soil worm extrudes through the die
of the rammer. This is the stage of the plastic failure of the compressed specimen
inside the container. The results are plotted as shown in Figure 3. The onset of the flat
portion of the load curve is determined as the extrusion force at failure (P f). The PWP
at failure is also recorded. It should be borne in mind that the shear strength parameters
are defined in terms of stresses; therefore, the extrusion force is divided by the cross-
sectional area of the container and the extrusion pressure at failure is obtained. For the
moment, this parameter is termed as the total extrusion pressure at failure (P E, in
kPas). The next step includes the determination of the effective extrusion pressure at
failure (PE), which is obtained by subtracting the PWP at failure from the total
extrusion pressure at failure. In this way, a set of three parameters is obtained: w, P E,
and PE. The procedure explained up to this point is repeated four to five times at
different water contents so that a sufficient data set is obtained to plot on a semi-
logarithmic diagram as shown in Figure 4, where the PE versus w and PE versus w data
are fitted by straight lines. These lines may be thought of failure envelopes and are
characteristic for soil of a certain plasticity.
Figure 3.
3 A typical plot for the RX-PWP tests. Figure 4. Plot of thee extrusion pressuures at failure
verssus the water co ontent (solid linne: PE vs w,
dashhed line: PE vs. w).
w
The second
s test typ
pe employed for f this investtigation is the consolidated--undrained
triaxial shhear test (3X X-CU). The specimen
s preeparation for this test is somewhat
different from
f the RX--PWP test. For the beginn ning water co ontent, any vaalue in the
range for the RX-PWP P test is seleccted. From th he dry soil sam mple of -#400 size, 120
grams off is weighed d and the ap ppropriate am mount of waater is addedd for the
predeterm mined water co ontent. It is miixed until a ho
omogeneous wet w mixture iss obtained.
The next step
s is the dettermination off the static compression forrce to be appllied for the
reconstrucction of a cylin ndrical 3X-CU U specimen. ItI is determineed from the soolid line on
Figure 4 by finding th he appropriatee extrusion pressure for th he predetermiined water
content. Then,
T this totaal extrusion prressure is divid ded by the cro oss-sectional aarea of the
container. The wet soil sample is plaaced into the container
c in chhunks and genntly tapped
by a ramm mer as it was done for the RX-PWP test. The contain ner, which haas different
top cap without
w water channels and d a transducerr along with the specimenn inside, is
placed into a loadframee and is comprressed by a raammer, this tim me without a ddie orifice.
The comp pression is co ontinued at a rate of 0.5 mm/min
m until the compresssion force
reaches th he total extru usion force deetermined from Figure 4 for f the particular water
content. When
W this is done, the speecimen is loaaded again to the same coompression
force, butt this time it is from the opposite
o end. The specimeen is removedd from the
container. Because thee amount off the dry soiil material fo or the test iss adjusted
according gly, the height of the specim men remains within
w the rangge described inn common
standards..
The cylindrical
c remmolded/reconsstructed soil specimen is pllaced in a celll of the 3X
test apparratus. For the saturation pro ocess, one thiirds of the tarrgeted first cell pressure
(i.e., 3 = 50 kPa) is taaken and a bacck-pressure sllightly less th han this pressuure level is
applied. Back
B pressure is continuou usly applied un ntil the B parrameter is at least 0.95.
Then, the back-pressurre is doubled and the procedure is repeated. Finally, the back-
pressure is tripled and it is ensured that t the specim men is fully saturated,
s provviding that
the ratio of
o back pressu ure to the targgeted cell presssure is equal to or greater than 0.95.
The next stage is the co onsolidation of
o the test speecimen. At thiis stage, the sppecimen is
allowed to t drain from m both endss, and the deformation
d readings aree recorded
automatically. From a plot of this data, the prim mary consolid dation of 90%% and the
correspon nding time aree obtained, an nd the strain rate is determ mined accordiingly. The
final stagge is the sheearing of thee consolidateed specimen. During sheearing, the
deformatio on, pore-presssure and the axiala loading parameters
p aree recorded.
The procedure
p of 3X-CU test outlined abo ove is repeateed for two m more water
contents. In this way, three
t sets of data
d are obtaiined to constrruct a Mohr ddiagram as
shown in Figure 5, froom which the effective ang gle of friction
n is read. At this stage,
because thhe remolded samples
s are noot expected to
o have cohesio on, we are nott interested
in small amount of co ohesion valuees and only need
n the effeective angle oof internal
).
friction (
Figure 5. Moh
hr diagram of thee test results for th
he sample 73A with
w w = 23.0%.
Prelimina
ary Results
Three remmolded soil sammples were tested using botth the RX-PW WP and 3X-CU U methods.
The RX-P PWP tests werre carried out using four to five differentt water contennts, and the
plots from
m each of wh hich were obtained similar to the one showns in Figuure 3. The
extrusion pressures and pore-water pressures fro om each grap ph were plotted against
water conntents, similar to the graph shown in Fig gure 4. Three plots of this type were
obtained for
f three soil samples.
s
The same
s three so
oil samples were
w subjected
d to 3X-CU experiments,
e eeach being
tested at three
t differentt water conten
nts. For instan
nce, the plot given
g in Figuure 5 is for
the samplee 73A at the water
w content of 23.0%. Twwo other waterr contents werre tried for
the soil sample
s 73A. In this way a total 9 Mo ohr diagrams was obtainedd for three
different soil
s samples. Owing to spaace limitationss, all of those diagrams aree not given
here. Onlyy a summary plot of the reesults is preseented in Figurre 6. Note thaat the data
points for each soil sam
mple display a certain patternn.
he moment, 9 sets of effecttive angles off friction and water contentt from 3X-
For th
CU tests are
a available. For instance,, the water co ontents of the sample AK aare 34,0%,
38,0% and d 42.2% and the respectivee effective an on are 27o, 244o and 14o.
ngles of frictio
The respeective effectivve extrusion pressures
p for the very samme water conttents from
Figure 4 are
a 1550 kPa,, 1200 kPa an nd 750 kPa. The
T summary of experimenntal results
from the 3X-CU
3 and RXX-PWP tests are
a provided in i Table 2.
Figure 6. Plot illustrating th

he brief summaryy of the experim
mental results (cirrcles: sample AK
K, rectangles:
sample 37A,, triangles: samplle 73A).
Table 2. Summ
mary of the experrimental results.
No
o LL PL
L w (%) PE (kPa) (o)
AK
K 74.1 5.6
35 34.0 1550 2
27
AK
K 74.1 35
5.6 38.0 1200 2
24
AK
K 74.1 5.6
35 42.2 750 1
14
37A 92.5 6.5
26 30.0 540 2
24
37A 92.5 26
6.5 40.0 315 1
19
37A 92.5 6.5
26 50.0 80 1
13
73A 40.3 9.1
19 19.0 2250 4
40
73A 40.3 19
9.1 23.0 780 2
27
73A 40.3 9.1
19 27.0 260 2
20

The ultimate
u goal of the investtigation is to establish a reelationship beetween the
effective angle
a of interrnal friction and
a the effectiive extrusion pressure. Thee effective
extrusion pressure and effective anglle of internal friction are bo oth affected nnot only by
the water content of the tested samp ples but also by
b the plasticiity of soils. It should be
appreciateed that highlyy plastic soilss tend to yield d lower internnal friction aangles, and
vice versaa. Therefore, any attemptt to correlate the two sheear strength pparameters
(namely and PE) sh hould includee the LL and PL values as a well. Thats why the
Atterberg limits were incorporated into the regrression analyssis. Thus, thee effective
internal frriction angle should
s be a function
fu of thee water contennt, Atterberg limits and
the effectiive extrusion pressure.
p
A reg gression analyysis was con nducted using g the data in Table 2. Thhe ultimate
equation isi to predict thhe effective innternal friction
n angle whenn the liquid limmit, plastic
limit, watter content an nd the effecttive extrusion n pressure aree given. The following
equation is
i the result off the regressioon analysis:
= 0.112 L 0.392w + 0.0088PE + 30.6

2LL 0.353PL 3 (R2 = 0.9
98) (1)
The regression coefficient (R2) turns out to be 0.98, which indicates a very strong
correlation between the effective internal friction angle and the other four independent
variables. While the resulting equation is highly promising, it was establihed upon only
the three soil samples. Further justification is needed by including more soils into the
analysis.
Conclusion
Based upon a limited set of data, the present investigation shows that the newly
introduced shear strength test--namely the reverse extrusion equipped with a porewater
pressure measurement assembly (RX-PWP)--appears to predict the effective friction
angle for remolded soils with a great degree of success. Considering that the total
duration for the RX-PWP test for a single soil sample is only around 1 hour, it can be
asserted that this newly developed technique is quick, simple and less operator-
dependent. Further analyses on both remolded and natural soils are required to validate
the findings presented herein.
References
[1] G. Baldi, D.W. Hight, G.E. Thomas, A Reevaluation of conventional triaxial test methods. In: Advanced
Triaxial Testing of Soil and Rock, ASTM STP 977. T. Robert, R. Donaghe, C. Chaney and L.S.
Marshall, Eds., American Society for Testing and Materials, Philadelphia, 219-263 (1988).
[2] J.T. Germaine, C.C. Ladd, Triaxial testing of saturated cohesive soils: In: Advanced Triaxial Testing of
Soil and Rock, ASTM STP 977. T. Robert, R. Donaghe, C. Chaney and L.S. Marshall, Eds., American
Society for Testing and Materials, Philadelphia, 421-459 (1988).
[3] S. Lacasse, T. Berre, Triaxial testing Medhods for soils, In: Advanced Triaxial Testing of Soil and Rock,
ASTM STP 977. T. Robert, R. Donaghe, C. Chaney and L.S. Marshall, Eds., American Society for
Testing and Materials, Philadelphia, 264-289 (1988).
[4] B.C. OKelly, L. Zang, Consolidated- drained triaxial compression testing of peat, Geotechnical Testing
journal, 36(3), (2013), 1-12.
[5] A. Timar, Testing the plastic properties of cohesive and intermediate-type soils by extrusion, Acta Tech.
Ac. Sci. Hungary, 76(3-4), (1974), 355-370.
[6] I.L. Whyte, 1982. Soil plasticity and strength a new approach for using extrusion, Ground Engineering,
15(1), (1982), 16-24.
[7] F. Medhat, I.L. Whyte, An appraisal of soil index tests, Geological Society, Engineering Geology Special
Publication, 2, (1986), 317-323.
[8] K. Kayabali, O.O. Tufenkci, Determination of plastic and liquid limits using the reverse extrusion
technique, Geotechnical Testing Journal 33(1), (2010), 14-22.
[9] K. Kayabali, A. Ozdemir, Use of reverse extrusion method to determine undrained shear strength of
Clayey Soils, Geotechnical end Geological Engineering, 31, (2013), 719-727.

IOS Press, 2015
doi:10.3233/978-1-61499-601-9-415
Influence of grading and mineralogy on

the behaviour of saprolites
Irene ROCCHIa, 1 , M. Cristina Todiscob and Matthew R. COOPb
a
Department DICAM, University of Bologna, formerly City University of Hong Kong
b
ACE Department, City University of Hong Kong
Abstract. Soils originating from chemical decomposition do not undergo any

sorting in their grading resulting in a well-graded material. Their composition may
be influenced by weathering in one or more aspects, e.g. particle size, particle size
distribution, mineralogy or particle morphology. Each of these aspects and their
effects on the mechanical behaviour have been studied separately in the past. To
identify the dominant parameters a series of standard oedometer tests was carried
out to investigate the influences of the grading and mineralogy on the Normal
Compression Line (NCL) separately. This was achieved taking the most and least
decomposed granite sampled at a same location and reconstituting them to the
same grading to study the influence of mineralogy or reconstituting one of them to
both the gradings to study that of grading. In the first case, the NCLs were
practically coincident, while in the second they were further apart, showing that
the grading has greater importance in determining the mechanical behaviour in
compression.
Keywords. Residual soil, one-dimensional compression, laboratory tests
1. Introduction
Soils originating from chemical decomposition do not undergo any sorting in their
grading and are therefore well graded. Their composition may be influenced by
weathering in one or more aspects, e.g. particle size, particle size distribution,
mineralogy or particle morphology, resulting in a considerable range of values for
engineering parameters along a vertical profile. Each of these aspects and their effects
on the mechanical behaviour has received attention in the past. Coop & Lee [1] studied
the behaviour of three reconstituted uniform sands having different geological origins
in one-dimensional compression. They showed that their behaviour was similar despite
the different mineralogies and particle morphologies. The compressibility was rather
small until the yield stress was reached, which corresponded to the start of major
particle splitting, and all the tests followed the same path afterwards regardless of the
initial density, determining the Normal Compression Line (NCL). They found higher
yielding stresses for particles having rounder shapes and stronger minerals. These and
other experimental works formed the bases for numerical modelling [2].
Binary mixtures of soils have also been studied rather extensively [3], while in
recent years Altuhafi & Coop [4] investigated the behaviour of a number of well graded
soils. In particular, Altuhafi & Coop [4] took three sands having different geological
origins and observed their behaviour in one-dimensional compression while changing
the particle size distribution from uniform to well-graded. They found that the NCL
416 I. Rocchi et al. / Inuence of Grading and Mineralogy on the Behaviour of Saprolites
progressively became flatter and the breakage reduced. Despite significant compression
even for the most well-graded specimens tested, they did not measure any breakage for
these samples. Minh & Cheng [5] were able to reproduce this behaviour qualitatively
using DEM, confirming that the grading is well linked to the coefficient of
compressibility. Coop & Lee [1] as well as Altuhafi & Coop [4] used soils having
different mineralogies, but the soils usually each had a homogeneous mineralogy. On
the other hand, Leleu & Valdes [6] studied quite extensively the behaviour of sand
mixtures having two different minerals, but they used uniformly graded samples having
the same grain size and particle morphology. Leleu & Valdes [6] found that the NCLs
of the mixtures are positioned between those of the two pure sands, but their yielding
was more gradual and the particles from both the mineralogies suffered less breakage
when a balanced mixture was formed.
In this work, the separate influence on the mechanical behaviour of particle
mineralogy and morphology and particle size distribution are investigated, as part of a
larger study on a naturally occurring well-graded soil having mixed mineralogy, i.e. a
decomposed granite saprolite from Hong Kong. In addition, artificial mixtures having
greater differences in their gradings and mineralogies were tested to corroborate the
conclusions.
2. Soil Properties and Methodology
This work is part of a wider research on the evolution of properties induced by

weathering on a granitic saprolite, the details of which can be found in [7]. In this work,
an extensive study of a decomposed granite saprolite was carried out, testing Mazier
samples cored along a weathered profile. Generally, this soil is a naturally occurring
well-graded sandy gravel having mixed mineralogy, whose characteristics change with
depth due to different amounts of weathering. However, in this work the focus was
only on the most and least weathered soils, although some soil particles belonging to an
intermediate weathering degree were also used as a base material for some of the tests.
The shallowest and deepest soils sampled, respectively at depths of 6.5-12m and 24-
27m, correspond to the most and least weathered soils and belong respectively to
grades V and IV, i.e. Completely and Highly Decomposed Granite [8]. The soil having
an intermediate weathering degree was sampled at 12-20m depth and also belong to
grade V. Rocchi & Coop [7] referred to these soils as the sh ewCDG, HDG and
evwCDG, respectively. The sh ewCDG and HDG were at the opposite extremes of the
ranges encountered for grading, mineralogy and morphology.
Table 1. Sample composition in the testing programme.

Particles
sh ewCDG evwCDG HDG
sh ewCDG mwp mwg lwp
Grading
HDG lwg
I. Rocchi et al. / Inuence of Grading and Mineralogy on the Behaviour of Saprolites 417
Table 1 explains how the grading and soil particles were combined in the testing
programme. The particle size distribution and particle mineralogies and morphologies
were characterised for the sh ewCDG and HDG and these results will be discussed later.
The particles of the sh ewCDG were reconstituted, as will be explained later, using its
own initial grading (mwp sample) and similarly the HDG particles were also
reconstituted to the sh ewCDG grading to obtain the lwp sample. Particles from an
intermediate weathering degree (evwCDG) were instead used to form the mwg and lwg
samples, so that an adequate amount of fine and coarse particles were available, where
mwg and lwg are based on the gradings of the sh ewCDG and HDG, respectively.
In Table 2 the XRD results are presented illustrating the sample mineralogy. If the
components are divided so as to have only three groups, i.e. quartz, feldspars and
others, then the sh ewCDG consists approximately of a mixture having 40% quartz
15% feldspars and 45% remainders, while the percentages are 45, 30 and 25% for the
HDG. The XRD analyses were carried out on a powder obtained from intact pieces of
the Mazier samples and therefore might include particle aggregates larger than the
maximum size used for preparing the reconstituted samples (6.3mm). However, from
visual inspection it did not appear that there was a correlation between the size of the
particle aggregates and their mineralogy as the smaller soil particles also consisted of
aggregates. These mineralogies should therefore correspond to those of the mwp and
lwp samples.
The morphologies of the soil particles were investigated by means of a Qicpic laser
scanner [9]. The Qicpic was used in the GRADIS mode, which captures the silhouette
of the soil particles as they fall under gravity, and can investigate a range of sizes
between 0.1 and 3mm. The resolution of the equipment is about 10m, which is
sufficient to define shape parameters for the smallest particles captured. In Figure 2 the
cumulative distribution curves of the aspect ratio (AR), sphericity (S) and convexity
(C) are presented for the mwp and lwp. The aspect ratio (AR) is the ratio of the
minimum to the maximum Feret diameters, where the Feret diameter is the distance
between any pair of parallel lines that encompass the particle silhouette. The convexity
(C) is the ratio between the area of the particle silhouette and its convex hull, while the
sphericty (S) is the ratio between the perimeter of a circle with an area equivalent to
that of the scanned silhouette and the real perimeter of the particle.
Table 2. Mineralogical composition by XRD for the mwp and lwp.

mwp lwp
% %
Amorphous content 22 21
Illite/Muscovite 6 3
Kaolinite 19 -
Potassium feldspar 15 15
Sodium plagioclase 1 16
Quartz 38 45
Figure 1. Cumulative distributions of shape Figure 2. Shape characterisation.

descriptors.
The cumulative distributions of the shape descriptors shown in Figure 1 are based
on volume and therefore give greater weight to particles having larger sizes, being
more consistent with the particle size distribution that would be obtained by sieving.
Although the average values are quite close to each other for the two soils studied, it is
clear that the lwp have consistently lower values for all the shape descriptors, especially
aspect ratio.
In Figure 2, the soils are classified based on the chart proposed by Altuhafi et al.
[10]. Both samples have sub-angular particles, the lwp being closer to angular. It can be
seen from the scale of this chart that relatively small changes in the values of the
descriptors are sufficient to account for the change from angular to round particles.
Figure 3 presents the average grading curves of the whole set of the sh ewCDG
and HDG samples that were sieved. These curves were used as the basis for preparing
the reconstituted samples of the mwg and the lwg as already mentioned, but terminating
them at 6.3mm to ensure a reasonable ratio between the maximum grain size and the
oedometer ring diameter (50mm). The gradings were obtained according to BS1377
[11], keeping the preparation of the soil before sieving to a minimum, i.e. only air
drying the core obtained by Mazier sampling. After removing the particles larger than
6.3mm, the ratios were 10:56:34% for the mwg and 46:49:5% for the lwg. After
separating the soil by dry sieving, the samples were reconstituted to the target average
grading and prepared by moist tamping.
Figure 3. Particle size distributions.
Additional compression tests on both uniform and well graded fine sands of
different mineralogies were carried out for comparison, where the gradings and
mineralogy differed considerably more than for the natural soils considered. Two types
of sands were used: Leighton Buzzard Sand (LBS), which is a quartz sand from the UK,
and a crushed Limestone from China (LMS), which is a calcium carbonate sand. The
uniform grading (-UG) of both sands comprised particles ranging from 0.3 and 0.6mm,
while the well-graded sand (-WG), which was composed of the crushed limestone, had
particles ranging from approximately 0.005 to 0.6mm.
The oedometers used were equipped with either 50mm diameter fixed rings or
floating rings having either 20 or 30mm diameter. The maximum vertical stresses
reached (v) were approximately 7MPa for the fixed rings and 20 or 50MPa for the
floating rings. The tests were corrected for the equipment compliance, which was
measured separately carrying out a test on a steel disk using the same set up as for the
soil specimens. In addition, precautions were taken to determine accurately the initial
specific volume (v) of the specimens. These consisted of taking measurements of the
specimen height and mass both at the beginning and at the end of the test.
3. One-dimensional Compression
Figure 4 shows the compression tests on the mwp and the lwp samples and the NCLs
chosen based on these tests. These are practically the same for the two soils, although
the range of initial specific volumes and the initial compressibilities of the tests are
different for the mwp and the lwp. On the other hand, the NCLs for the mwg and the
lwg samples are further apart in Figure 5. This would suggest that for these soils the
initial grading of the soils and the resulting initial fabrics are more important than their
mineralogies in determining the behaviour in compression.
Figure 4. Influence of mineralogy on Figure 5. Influence of grading on

weathered soils. weathered soils.
When comparing tests from the artificial mixtures soils in Figures 6 and 7, similar
trends of behaviour can be observed. In Figure 6, the compression paths of the LMS-
UG and the LBS-UG show that the slope of their NCLs is almost identical, while their
N values are only slightly different. When comparing the compression tests on the
well-graded limestone (LMS-WG) with those of the uniform sand (LMS-UG) in Figure
7, it is evident that the slope of the NCLs is significantly different, that of the LMS-
WG being much more gentle than that of the LMS-UG, as might be expected. These
results confirm what was observed for the granitic saprolite.
The breakage of the samples was analysed to shed light on the mechanisms of
compression. Most of the tests carried out on the granitic saprolite reached a similar
maximum stress, corresponding to 7MPa, and only these tests are presented in Figures
8 and 9.

artificial soils. artificial soils.

breakage. breakage.
The breakage was analysed with respect to the initial specific volume (v0), as
Altuhafi & Coop [4] showed that v0 does have an influence on breakage. The relative
breakage (Br) presented here is as defined by Hardin [12]. However, despite comparing
only those tests starting at a similar v0 and reaching the same pressure, within the data
scatter it was not possible to find a significant difference in the trends between those
samples whose grading was the same, but the particles making up the specimen
changed (Figure 8) and those tests where the grading was changed while keeping the
same soil particles (Figure 9), despite there being a difference in their NCLs.
4. Conclusions
Soils formed from the weathering of decomposed granite represent an example of a

naturally occurring soil having mixed mineralogy, particle morphology, absolute
particle size and grading. As part of wider study on decomposed granite saprolites, one-
dimensional compression tests were carried out to investigate which of these aspects
dominates the mechanical behaviour. It was found that, for the soil tested, weathering
influenced the particle mineralogy and morphology less than the grading and absolute
particle. As a result when using the same soil particles to reconstituted two samples
having different gradings, the difference in the NCLs obtained was greater than when
particles taken from samples at two opposite extremes of weathering were reconstituted
to a same grading, when the NCL was virtually the same. However, different trends of
behaviour might be expected in shear, as it was shown by Coop et al. [13] that the
angle of shearing resistance at critical states is not strongly relate to the initial grading.
Acknowledgments
The work described in this paper was fully supported by a grant from the Research
Grants Council of the Hong Kong Special Administrative Region, China (Project No.
CityU 112911). The authors would like to express their gratitude to Dr. John Endicott
from AECOM for kindly providing the samples tested and to Dr. Batrice Baudet
(Hong Kong University) for providing access to the Qicpic.
References
[1] M.R. Coop & I.K. Lee, The behavior of granular soils at elevated stresses. Predictive soil mechanics,
Thomas Telford, London, 186198, 1993.
[2] J.M. Pestana & A.J. Whittle, Compression model for cohesionless soils, Gotechnique 45 (1995), 611-
631.
[3] S. Thevanayagam & S. Mohan, Intergranular state variables and stressstrain behaviour of silty sands.
Gotechnique 50 (2000), 1-23.
[4] F. Altuhafi & M.R. Coop, Changes to particle characteristics associated with the compression of sands.
Gotechnique 61 (2011), 459-471.
[5] N.H. Minh & Y.P. Cheng, A DEM investigation of the effect of particle-size distribution on one-
dimensional compression, Gotechnique 63 (2013), 44-53.
[6] S.L. Leleu & J.R. Valdes, Experimental study of the influence of mineral composition on sand crushing.
Gotechnique 57 (2007), 313-317.
[7] I. Rocchi & M.R. Coop, The effects of weathering on the physical and mechanical properties of a granitic
saprolite, Gotechnique, provisionally accepted.
[8] Geotechnical Engineering Office. Guide to rock and soil descriptions. Geoguide 3. Geotechnical
Engineering Office, Hong Kong, 1988.
[9] Windox-operating instructions release 5.4.1.0. Sympatec, Clausthal-Zellerfeld, Germany, 2008.
[10] F. Altuhafi, C. OSullivan, & I. Cavarretta, Analysis of an image-based method to quantify the size and
shape of sand particles. ASCE Journal of Geotechnical and Geoenvironmental Engineering 139 (2013),
1290-1307.
[11] BS 1377:1990: Methods of test for soils for civil engineering purposes. British Standard Institution,
London, UK, 1990.
[12] B.O. Hardin, Crushing of soil particles. ASCE Journal of Geotechnical and Geoenvironmental
Engineering 111 (1985), 1177-1192.
[13] M.R. Coop, K.K. Sorensen, T. Bodas Freitas & G. Georgoutsos, Particle breakage during shearing of a
carbonate sand. Gotechnique 54 (2004), 157-163.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-423
Towards the measurement of fabric in

granular materials with x-ray tomography
Max Wiebickea,b,c,1, Edward Anda,c, Gioacchino Viggiania,c, Ivo Herleb
a
Univ. Grenoble Alpes, 3SR, F-38000, Grenoble, France
b
Technische Universitt Dresden, Institute of Geotechnical Engineering, Germany
c
CNRS, 3SR, F-38000, Grenoble, France
Abstract. Recent developments in experimental geomechanics have enabled the

characterization of grain-scale measurements with x-ray computed tomography
(CT) and thus paved the way for the micromechanical characterization of granular
materials. However, the images coming from x-ray CT are limited with respect to
their resolution if a representative specimen is to be investigated [1]. When trying
to extract the fabric of such images the accuracy of the methods is crucial. This
study investigates the accuracy of some standard and advanced approaches to
determine different fabric entities with synthetic as well as real images.
Keywords. Fabric, microstructure, x-ray CT, granular materials, image analysis
1. Introduction
In geotechnical engineering, soil behavior is treated by continuum constitutive models.

These are usually based on phenomenological approaches and therefore suffer from
different drawbacks. In order to overcome these phenomenological approaches and to
increase our understanding of soil behavior, it is crucial to start from a
micromechanical scale and relate the interactions of the discrete particles that define a
granular medium with the macroscopic observations and phenomena that can be and
already have been made from standard soil testing. The continuum based constitutive
models can then be enhanced using these grain-scale relations.
Extracting the granular fabric as one of the characteristics of granular materials has
been the topic of numerous studies since a long time. In [2] fabric is defined by the
spatial arrangement of the solid particles and the associated voids. It can be described
either by scalar quantities, such as the local void ratio or the coordination number, or
by higher order (vectorial or tensorial) quantities [3]. Possible orientational fabric
entities can be the principal particle orientation, the contact normal orientation [4] or a
measurement of void fabric [5,6].
Historically, one way of experimentally measuring the fabric has been to inject a
polyester resin into a soil specimen and to cut it into thin sections allowing for example
the determination of the particle and the contact normal orientation [4]. As this
approach is a post-mortem analysis, different but similar specimens are loaded to
different states and the fabric is then measured from each of them. A non-destructive
1
Corresponding Author. Email: [email protected]
424 M. Wiebicke et al. / Towards the Measurement of Fabric in Granular Materials
way to investigate the relation between fabric and macroscopic behavior is to use two-
dimensional rods in special apparatus. Here, photos are taken in defined intervals
allowing the micromechanical behavior to be compared to the macroscopic one [7].
Furthermore, discrete methods are able to simulate the interactions between sets of
particles caused by a macroscopic loading [8]. The determination of fabric is fairly easy
in the simulations since the positions and geometric relations between the particles are
perfectly known. However, all of these approaches are limited either to the nature of
post-mortem analyses or the two-dimensional and regular shape of the particles
investigated.
This limitation, however, can be overcome with the application of x-ray
tomography on granular materials, which was mainly developed in the last two decades
[9,10]. However, when trying to image specimen of a representative amount of grains,
the resolution of individual grains and grain-to-grain contacts is limited and thus poses
problems when trying to extract different fabric entities [11], see Figure 1. Applying
standard approaches to determine, for example, the contact normal orientation leads to
strongly biased results [10,12]. Therefore, quantifying the accuracy of different
standard and advanced approaches is crucial in order to deal with the complexity of
granular materials. This paper presents a study on the accuracy of the determination of
the principal particle orientations, the detection of contacts and the orientation of
contact normals using either high resolution x-ray images or synthetic ones.
Figure 1: X-ray tomographies of Hostun Sand at different resolutions. a): Standard resolution for a triaxial
test -- trade-off between a sufficient number of grains in the specimen for a mechanically relevant test, and
enough information per grain for tracking. (pixel size=15.56m) - Left: Slice through a x-ray CT image.
Right: 3D rendering of a single grain. b) High resolution scan (pixel size=0.35m) Left: Orthogonal slices
through a high resolution image obtained with nano x-ray CT. Right: 3D rendering of this single grain.
M. Wiebicke et al. / Towards the Measurement of Fabric in Granular Materials 425
2. Particle orientation
One of the above mentioned fabric entities is the principal particle orientation, which
can be expressed as an orthonormal base. One way to compute such a base from a 3D
object is with the eigenvectors of the moment of inertia tensor. The tensor is calculated
on a segmented binary image of an individual particle. The eigenvectors can be subject
to non-physical changes of orientations, such as snapping [13]. It was guessed that the
reason for such snapping is either due to the shape of the particles or to the resolution,
meaning that for a grain with a low number of voxels defining it, a rotation might cause
a loss of a few voxels at a certain side and thus change the principal axis significantly
[14].
Figure 2: Particle Orientation. a) 3D rendering of the high resolution scan of a single Hostun sand grain. The
coloured arrows represent the eigenvectors of the moment of inertia tensor. b) Relation between the accuracy
of the eigenvectors of the moment of inertia tensor and the scale of the image. The colours in the plot
represent the eigenvectors in the 3D rendering.
To answer the question of whether the snapping of principal particle orientations is

caused by the resolution of the image or the shape of the grain, a single Hostun sand
grain was scanned using a nano x-ray CT, see Figure 2a.To assess the accuracy of the
determination of the principal particle orientation, a high resolution greyscale image of
this grain was rotated via 4 different angles around a defined axis and these different
configurations were scaled down to a minimum with a grain resolution corresponding
to a typical size measured in an experiment in the x-ray CT. The orientations
determined on the original high resolution image were defined as the reference and
rotated in the same way as the images. The error was then defined as the angle between
the orientations of the high resolution image and the scaled down images.
The results of this study can be seen in Figure 2b. The mean error was calculated
for each scaling using the three eigenvectors and all investigated rotations. As it was
expected, the mean error increases with decreasing resolution of the image. The
eigenvector corresponding to the smallest eigenvalue (Eig3, blue vector) of the moment
of inertia tensor was found to be the most accurate one for this grain. This is caused by
the shape of the grain, as it has a clearly defined long axis which corresponds to this
vector. As can be seen in Figure 2, the mean error at a reasonable scale, which is the
lowest one investigated here, amounts 1.8 for a highly angular grain and the
investigated rotations. It can be expected that the accuracy of this approach for rounded
grains is lower as the principal axes are not as clearly defined. Surprisingly, no particle
snapping was encountered. In a further study more rotations around different axes will
be considered to get a statistically sound estimate of the accuracy of this method.
3. Contact detection
The most widely used fabric entity is the orientation of the contact normal between two
grains. Apart from the accuracy of the contact orientation, the detection of the contacts
is a crucial point itself.
Artificial spheres were chosen in order to investigate the performance of the
techniques currently used to detect contacts, since the contact orientation between two
spheres is identical to their branch vector. The artificial spheres were created using
Kalisphera [15] which takes into account the partial volume effect and is able to blur
the spheres. In Figure 3 a slice through the center of two sample spheres and a 3D
rendering of the two spheres are presented. The spheres are put just in contact without
any overlap, but due to the partial volume effect of both spheres in the contact region
the contact area is artificially increased, which can clearly be seen in the 3D rendering.
Figure 3: Image of two touching partial volume spheres created using Kalisphera.
3.1. Limitations of the common (state-of-the-art) approach
To investigate the contact detection, two partial volume spheres were created at a
specific distance with a random branch vector in between them. The image is binarised
using a global threshold (volumetrically correct threshold is 0.5) and it is checked
whether the spheres are in contact or not. If the distance is for example 1 voxel and no
contact is found, the image will be blurred until a contact can be found. Blurring
represents the noise inherent in x-ray reconstruction, which can vary from machine to
machine, and for different imaging condition. This analysis was conducted for 4000
pairs of spheres connected with randomly orientated branch vectors.
The result of this approach varying the distance and the blur as described above for
different orientations is plotted in Figure 4a. The points in this graph indicate when a
contact is found. Lets take for example the blue continuous line (Mean): for a distance
of 1 voxel no contact can be found for the clean image. The blur has to be introduced
and increased to a value of 0.8 until a contact is found again. Having a closer look at
the 3D rendering in Figure 3, the reason of this over-detection becomes obvious. The
values of the voxels in the contact region are influenced by the partial volume effect of
both spheres. Even if the spheres are not touching, a voxel close to both spheres might
get a value higher than the global threshold as both spheres take part of the volume of
the considered voxel.
a) 2.5 b) 2.5
2 2
Gaussian blur [vx]
Gaussian blur [vx]

1.5 1.5
1 1
threshold=0.5
0.5 Mean 0.5 threshold=0.7
Max threshold=0.8
Min threshold=0.9
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
Distance between the spheres [vx] Distance between the spheres [vx]
Figure 4: Investigation on the accuracy of contact detection. A point indicates when a contact is found. The
black line marks a system-inherent level of blur [13]. a) Evaluation of the analysis with 4000 randomly
distributed spheres. Global threshold = 0.5. b) Application of a local threshold on the contact area.
In [15] the authors found that a reasonable blur in images taken with x-ray CT with
a laboratory micro-focus x-ray CT machine at a pixel size of around 10m/px is about
0.8 voxels. Thus, the mean error of the contact detection at this system-inherent blur
lies in between 0.9 and 1.2 voxels, meaning at a distance of up to 1 voxel the state-of-
the-art approach will still detect contacts although there are none. In this way, contacts
are systematically over-detected, implying a higher number of contacts and making an
impact on the statistics of the orientation measurements.
3.2. Overcoming the limitations
As contacts should only be detected where they exist, a refinement of the current
approach has to be done. After applying the procedure described above, solid voxels in
between grains are identified as contacts. These voxels will be taken into account again
and a local threshold, which has to be higher than the global one, will be applied on
them. Assuming such a local threshold is known, the contacts, that were identified
beforehand as contacts but which are none, can be erased from the list of contacts. The
image which will be worked on in the following steps, to determine the contact
orientation for example, will still be the one binarised using the global threshold. The
local re-thresholding is used only to clean the list of detected contacts.
The results of this refined approach are plotted in Figure 4b in the same way as
before, but only mean values are included. It can clearly be pointed out that the error of
contact detection decreases significantly introducing a local threshold. Nevertheless,
the local threshold has to be chosen with care in order not to lose contacts. As the
images used in this study were purely artificial ones, no recommendation for
experiments on real granular materials in x-ray CT can be given so far. Applying the
refined approach on images of real granular materials will be subject of a further study.
4. Contact Orientations
To determine contact normal orientations usually a standard watershed approach is

used [16]. It was found in [12] that the random walker watershed does not introduce the
strong bias that is usually obtained by applying standard watersheds.
4.1. Watershed methods
In order to assess the performance of the different approaches, they were applied on
realistic partial volume spheres, created using Kalisphera [15], as the contact
orientation for a pair of spheres is identical to the branch vector, which is imposed. In
order to obtain statistically exploitable results, a number of 4000 pairs of spheres with a
randomly orientated branch vector were created. Principal Component Analysis (PCA)
is used to fit a plane, whose normal represents the contact orientation, onto the
positions defining a contact.
In Figure 5a the results of such an investigation for spheres of different diameters
are plotted. The error is defined as the angle between the randomly imposed branch
vector and the contact normal determined by the evaluation of the watershed results.
The random walker approach yields far better results in terms of accuracy. The
application of the segmentation approaches on the blurred image gave a lower mean
error, indicating that the size of the contact region influences the accuracy as it grows
with an increase of blur.
a) 30 b) 7
Standard Watershed gb=0.0
25 Random Walker 6 gb=0.8
mean error [degree]
mean error [degree]
gb = 0.0
gb = 0.8 5
20
4
15
3
10
2
5 1
0 0
10 15 20 25 30 5 10 15 20 25 30 35 40 45
diameter [vx] number of pixels representing the contact [-]
Figure 5: Accuracy of watershed methods to determine the contact orientation. a) Relation of mean error and
the size of the spheres. b) Relation of mean error to the number of pixels representing the contact used for
PCA for spheres of a diameter of 16 voxels and the random walker approach. gb Gaussian blur.
Furthermore, the geometrical properties seem to have an impact on the accuracy of

the contact orientation as the mean error decreases with an increasing size of the
spheres. The number of pixels representing the contact used in the fitting of the contact
plane was chosen as the property to investigate as the number of data in PCA should
have a strong impact on the results. The comparison of the number of pixels
representing a contact and the error for spheres of a diameter = 16 voxels is plotted in
Figure 5b. As assumed, the number of pixels representing a contact influences the
accuracy of the measurement. The higher the number the lower is the actual error. In
this case, 14 pixels seem to mark a border after which the error oscillates around and
below 3. However, this analysis was done using synthetic spheres and will have to be
applied on real images coming from x-ray CT.
4.2. Level set method
The level set method describes an individual particle as a continuous distance function
called the level set. As it is a continuous function it can be arbitrarily located on a
computational grid and thus does not suffer resolution effects [17]. A similar analysis
on contact orientations as before was carried out and is shown in Figure 6. The random
walker yields a higher mean error for either perfect or blurred images, whereas the
level set method yields a mean error that is lower than 2 in the region of interest.
6
Random Walker
Mean error [degree] 5 Level Set
gb = 0.0
4 gb = 0.8
0
10 12 14 16 18 20
Diameter [vx]
Figure 6: Contact orientation. Relation of the mean error of 4000 pairs of spheres with their diameter.
Comparison of the performance of the level set method and the random walker approach.
In a study on the sensitivity it was found, that the accuracy of the contact
orientation of a single pair of spheres was improved remarkably by primarily adapting
the number of iterations and varying the parameters and . The accuracy depends
highly on the calibration of the parameters and the chosen number of iterations.
However, as in this study the number of iterations was prescribed without any
measure of convergence, further studies have to be conducted in order to define a
measure of convergence and optimize the method.
5. Conclusion
In this study different approaches to extract the fabric of a granular material were
investigated with respect to their accuracy. The principal particle orientation was
determined using the moment of inertia tensor and it was found to yield relatively low
errors for the investigated scales and orientations. Surprising results were obtained in
the investigation of the contact detection: using the regular approaches contacts are
systematically over-detected due to a combined partial volume effect, which
necessarily leads to noise in the statistical measurements when analysing real
specimens with thousands of grains and contacts between them. Applying a local
threshold on contact areas that were identified with the usual approach improves the
accuracy substantially. A statistical analysis was conducted on different methods to
determine the contact orientation of two touching spheres. The size of the contact area,
namely the number of intervoxel positions, appears to be a possible indicator of the
accuracy of a measurement. Level sets smooth the way to increase the accuracy of such
measurements, but must be further improved.
We have chosen to start from synthetic spheres, where the geometry and thus the
contact orientations are well known. This enabled us to precisely quantify the error of
the different approaches that were used to determine the contact orientation. The next
step will be to carry out a similar analysis on realistic images of materials such as glass
ballotini, for which the ground truth of contact orientation is only approximately known
due to manufacturing errors. The investigation of particle orientation will be applied on

a large number of random orientations at the different scales in order to obtain
statistically exploitable results. These advanced techniques and the knowledge of their
accuracy and their weaknesses allows a reliable extraction of the fabric of real soils
consisting of thousands of grains. The fabric can then be statistically described in
different kinds of fabric tensors [18]. Using the tensorial description of fabric, the
statistical micromechanics can be related to the macroscopic behavior, which finally
leads to the possibility of improving existing phenomenological continuum models.
Acknowledgements
The work reported in this paper is part of a project funded by the DFG (HE 2933/8-1).
The financial support is gratefully acknowledged. Laboratoire 3SR is part of the LabEx
Tec 21 (Investissements dAvenir - grant agreement nANR-11-LABX-0030)
References
[1] E. And, G. Viggiani, S. A. Hall, J. Desrues, Experimental micro-mechanics of granular media studied
by x-ray tomography: recent results and challenges, Gotechnique Letters 3 (2013), 142-146.
[2] R. Brewer, Fabric and mineral analysis of soils, J. Wiley, New York, 1964.
[3] M. Oda, J. Konishi, S. Nemat-Nasser, Some experimentally based fundamental results on the mechanical
behavior of granular materials, Gotechnique 30 (1980), 479495.
[4] M. Oda, Initial fabrics and their relations to mechanical properties of granular material, Soils and
Foundations 12 (1972), 17-36.
[5] M. Oda, S. Nemat-Nasser , J. Konishi, Stress-induced anisotropy in granular masses, Soils and
Foundations 25 (1985), 85-97.
[6] X. Li, X. Li, Micro-macro quantification of the internal structure of granular materials, Journal of
engineering mechanics 135 (2009), 641-656.
[7] F. Calvetti, G. Combe, J. Lanier, Experimental micromechanical analysis of a 2D granular material:
relation between structure evolution and loading path, Mechanics of Cohesivefrictional Materials 2
(1997), 121-163.
[8] P. Fu, Y. Dafalias, Fabric evolution within shear bands of granular materials and its relation to critical
state theory, International Journal for Numerical and Analytical Methods in Geomechanics (2010).
[9] S.A. Hall, M. Bornert, J. Desrues, Y. Pannier, N. Lenoir, G. Viggiani, P. Besuelle, Discrete and
continuum analysis of localised deformation in sand using X-ray CT and volumetric digital image
correlation, Gotechnique 60 (2010), 315-322.
[10] G. Viggiani, E. And, C. Jaquet, H. Talbot, Identifying and following particle-to-particle contacts in
real granular media: An experimental challenge, Powders and Grains, AIP Conf. Proc. 1542 (2013),
60-65.
[11] E. And, G. Viggiani, S. A. Hall, J. Desrues, Experimental micro-mechanics of granular media studied
by x-ray tomography: recent results and challenges, Gotechnique Letters 3 (2013), 142-146.
[12] C. Jaquet, E. And, G. Viggiani, H. Talbot, Estimation of Separating Planes between Touching 3D
Objects Using Power Watershed, International Symposium on Mathematical Morphology 11 (2013).
[13] E. And, S. A. Hall, G. Viggiani, J. Desrues, P. Besuelle, Experimental micromechanics: grain-scale
observation of sand deformation, Gotechnique Letters 2 (2012), 107-112.
[14] E. And, S. A. Hall, G. Viggiani, J. Desrues, P. Besuelle, Grain-scale experimental investigation of
localised deformation in sand: a discrete particle tracking approach, Acta Geotechnica (2011).
[15] A. Tengattini and E. And, Kalisphera: an analytical tool to reproduce the partial volume effect of
spheres imaged in 3D (Submitted to Measurement Science and Technology).
[16] R. C. Gonzalez, R. E. Woods, Digital Image Processing, Prentice-Hall, 3rd edition, 2007.
[17] I. Vlahinic, E. And, G. Viggiani, J. E. Andrade, Towards a more accurate characterization of granular
media: extracting quantitative descriptors from tomographic images, Granular Matter 16 (2013), 9-21.
[18] K. Kanatani, Distribution of directional data and fabric tensors, International Journal of Engineering
Science 22 (1984), 149-164.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-431
Observing breakage in sand under triaxial

and oedometric loading in 3D
Zeynep KARATZAa,b,1, Edward ANDb,c, Stefanos-Aldo PAPANICOLOPULOSa, Jin
Y. OOIa, Gioacchino VIGGIANI b,c
a
School of Engineering, The University of Edinburgh,
Edinburgh EH9 3JL, United Kingdom
b
Universit Grenoble Alpes, 3SR, F-38000, Grenoble, France
c
Abstract. The mechanisms involved in grain crushing/particle breakage are a

common research topic in several fields including geomechanics, geoscience and
particle technology. This topic is of interest to disciplines including material
processing, minerals and mining engineering, geology and geophysics, since
changes in the microstructure can lead to significant changes in the macroscopic
bulk behaviour of a material. The main goal of this research is to develop a new
understanding of the particle breakage mechanisms in a particulate system under
specific loading regimes. This is achieved by studying the microscopic changes
that an assembly of grains undergoes while being loaded/deformed. The possibility
to study and visualise the 3D volume of a specimen during loading is given by x-
ray computed tomography. The material under study is a sand (Caicos ooids)
which consists of very rounded particles that grow from fragments of corals
(Aragonite) with a D50 of approximately 350m. This material has been subjected
to triaxial and oedometric loading, in order to compare the response of the material
under different loading regimes. The acquisition of 3D images enables the study of
the deformation of the microstructure and especially the evolution of grain
breakage at different loading increments. The high resolution of the images enable
the authors to clearly observe and perform algorithmic measurements of broken
grains in small fragments. The image processing involves the creation of
algorithms to track each individual grain, match the fragments to the intact labeled
grains, in order to precisely describe the deformation throughout the loading
process. The comparison between the oedometric and triaxial loading shows a
clear difference in the distribution of damage. In the first case a dispersed pattern
of breakage can be observed, indicating that breakage originates predominantly
from the boundary conditions (especially the moving boundary loading platten)
and transmission of normal forces between particles, whereas in the second case a
very clear strain localisation has been observed, which indicates breakage is
mainly due to the deviatoric loading and observed mostly in a narrow shear band.
Keywords. 3D imaging, Grain breakage, Image processing, Oedometric test, Sand,

Triaxial test, X-ray computed micro-tomography (XCT).
1. Introduction
The mechanisms involved in grain crushing/particle breakage are of interest to many

disciplines including material processing, minerals and mining engineering, geology,
1
Corresponding Author; E-mail: [email protected]
432 Z. Karatza et al. / Observing Breakage in Sand Under Triaxial and Oedometric Loading in 3D
geophysics, geomechanics, geoscience and particle technology [e.g., 1-5] and therefore
has been a common research topic in these disciplines. Particle breakage in
geomechanics is important because it can produce changes in the microstructure of a
material, that will greatly affect the macroscopic behaviour of the material (e.g.,
strength, permeability, volumetric behaviour under shear); be it a specimen in a
laboratory or a whole soil formation in nature or in industry. It is understood that
breakage can occur under high stresses, however breakage has been observed even at
relatively low stresses [4] and in both cases this can lead to significant variations in the
particle size distribution and/or cause discontinuities; an undeniable reason as to why it
has been extensively studied experimentally, theoretically and numerically [5].
The possibility to study and visualise a 3D volume of a specimen during loading is
given by techniques such as x-ray computed tomography (hereinafter XCT) [e.g., 6-11].
Even though XCT has been used since the early nineties [6] it has only recently started
to advance as a powerful tool to investigate in 3D the response of several geomaterials
undergoing deformation. The analysis of the images will be based on full-field
measurements [12] that will enable the study and give a deeper insight into grain
crushing.
A sand was studied, where both its compressibility and shear strength were
evaluated with respect to grain breakage. Oedometric and triaxial tests were carried out,
during which XCT was performed. The purpose of the image acquisition is to observe
phenomena that take part in the whole specimen, which will be followed by
algorithmic measurements of variable quantities, which could be generalised into
constitutive models.
2. Experimental Campaign
2.1. Material Description
The sand used for this study is Caicos ooids with a D50 of 350m; a material with very
rounded particles that grow from fragments of corals. Ooids are spherical or ellipsoid
concretions of calcium carbonate and aragonite crystals (here more than 95%
aragonite) arranged around a nucleus. The ooids form in marine environments (rich in
carbonate calcium), by gaining successive layers (smooth, uniform carbonate coatings)
around a nucleus; a process followed by a resting period before the addition of each
new layer [13]. Depending on the successive growth and resting periods, variation in
the thickness of the coatings and the appearance of the nuclei can be expected, altering
the internal porosity of each grain and potentially its strength and stiffness.
2.2. X-ray Computed Tomography
An x-ray source emits a polychromatic beam which interacts with the material that is
being scanned. The number of photos that cross the material (not attenuated by the
material) are recorded by a detector as a projection of the density of the scanned object
at a certain orientation. The possibility to get a 3D image of the whole specimen, is
dictated by the need to acquire multiple projections at different orientations, therefore
the specimen is being rotated. The number of times the specimen is rotated will define
the number of acquisitions, that will be later on reconstructed into a 32-bit grey scale
stack of images (3D volume). The magnification is such that images of a pixel size
Z. Karatza et al. / Observing Breakage in Sand Under Triaxial and Oedometric Loading in 3D 433
(resolution) of 15.57m/px or 8.83m/px for the triaxial tests and 12.25m/px for the
oedometer are obtained. Consequently, the size of each voxel (3D pixel) is such that
each intact grain comprises of 20 to 40 voxels at its diameter, which provides enough
information to visualise breakage of individual grains.
Figure 1. The x-ray scanner configuration, along with a set of radiographs and a 3D reconstructed volume.
2.3. Triaxial & Oedometric tests
Two types of tests were performed on the same material; four triaxial compression tests
and three oedometric compression test, of which three and one were scanned
respectively. In order to acquire the radiographs, the specimen must not be straining or
changing in any way, and therefore the loading was paused for the duration of the
image acquisition (indicated by the stress relaxations that can be observed in Figure 2).
These increments will be referred to as loading stages. The loading direction at all tests
was ascending with a displacement rate of 21m/min (a=0.001%) and 50m/min
(a=0.003%) at the triaxial and oedometrer respectively. Another condition is that the
loading apparatus have to be transparent to x-rays and therefore they were specially
designed for XCT. The triaxial cell is polycarbonate (described in [11]) and the
oedometer is made of polyether ether ketone (PEEK), due to its high tensile strength
and transparency to x-rays. Table 1 provides some basic details of the performed
experiments and Table 2 some image based information about the tests were XCT was
performed. From the images, the number of grains of each specimen has been
accurately measured and the grain size distribution has been plotted from where the D 50
and the coefficient of uniformity (Cu) have been calculated (Table 2).
Table 1. Basic information regarding the performed tests.

Test Reference1 Resolution Specimen Specimen Number of
[m] Height (initial) Diameter Scans
[mm] (initial) [mm]
TXC01 No XCT 22.00 11.00 N/A
TXC02 15.57 23.48 9.36 12
TXC03 15.57 24.46 10.28 7
TXC04 8.83 22.00 10.00 2
1DC01 No XCT 15.00 15.00 N/A
1DC02 No XCT 15.00 15.00 N/A

1DC03 12.25 16.72 15.00 7
1
TXC indicates triaxial compression tests, 1DC indicates oedometric compression tests.
Table 2. Image based information about tests where XCT was performed
Test Reference Void Ratio D50 [mm] Cu Number of
(initial - e0) Grains
TXC02 0.66 0.32 1.58 32464
TXC03 0.69 0.28 2.65 103135
TXC04 0.77 0.37 1.60 (estimated) 33000
1DC03 0.58 0.37 1.55 54456
3. Results & Analysis
3.1. Macroscopic response
Figure 2. Response of all the performed tests on Caicos ooids (please note that the colours of the lines
correspond to the same specimens in all the graphs)
From the macroscopic response (Figure 2) of the specimens it is clear that both
experimental procedures are fairly reproducible. However the response of the triaxial
tests largely depends on the grain size and the initial density of the specimens. From
the triaxial tests, the peak stress indicates that TXC01 is the denser specimen, TXC02
and TXC03 have similar densities and TXC04 is the most loose specimen. This can be
verified from the initial void ratio (Table 1). The plot of the volumetric strain has been
created by measuring from the images the volumes of the grains and the bulk volume.
For the specimens that were not scanned, volumetric strain data are not available. It
should be taken into account that while the specimens are strained and more grain
breakage occurs, there is a threshold to the size of the fines that can be visualised,
based on the resolution. The lower the resolution the more fines can not be identified in
the images resulting in a loss of solid volume. This could result to a higher reduction in
volumetric strain for the loading stages after peak stress, as fragments start to fall
below the detectable size; TXC03 has the smallest grain size and the lowest resolution.
Needles to say though, as the initial grain size of TXC03 is smaller than at the other
specimens, the potential of breakage of its particles is also reduced comparatively due
to the decreased normal contact forces and the lower probability of the grains having
defects [14]. The compactive behaviour during the triaxial and oedometric tests can be
also verified from the volumetric strain measured from the images, which is also in
accordance with the initial densities of the specimens.
The preparation of the specimens was identical, and that explains why specimens
TXC02-04 have similar initial density. However, TXC01 was rather well distributed
compared to the other uniform specimens; therefore the finer particles would fill up the
pore space, resulting in a much denser specimen. From the grading curve it is evident
that TXC02 and 1DC03 have a similar grain size distribution; the higher percentage of
fines for 1DC03 is a direct result of the higher resolution, producing potentially more
noise throughout the image during segmentation.
Comparing the scanned oedometric test with the triaxials, it is concluded that less
breakage occurred however, as it can be seen from the first plot in Figure 2, 1DC03
was not strained enough due to limitations of the loading system on the applied force
and therefore regardless of the type of loading, less breakage was anticipated.
However, regarding the curvature and the slope of the oedometric compression curves,
it can be initially concluded that the yielding characteristics will be similar (as
expected) and that little if any breakage will occur.
3.2. Breakage patterns
An observation of the 3D volumes shows that during the triaxial tests consistently some
grains start to break close to the moving boundary before the peak stress is reached
(around 5% of axial strain; after the second scan but before the third one). Apart from
the grains close to the boundary, breakage starts being evident at the scan immediately
before the peak, but it can not be described as being concentrated in a band. However,
localised distribution of breakage becomes important after the peak stress, which
indicates that some time after 10% of axial strain the grains start to collapse due to the
intense shearing around the shear band.
Unfortunately, for the oedometer there is XCT data only for one test, which
refrains us from drawing generalised conclusions. However, the repeatability of the
tests and the geometry of the loading curve allow us to assume that the other specimens
would have exhibited similar breakage behaviour. There is data from two scans
performed close to the breaking point, as it would be defined by the change in the
curvature of the macroscopic loading response. In the first one, hardly any breakage
can be identified in the specimen, but in the second one there is clearly some that does
not progress vigorously as the loading increases. Once again the breakage is
concentrated initially close to the moving boundary, but there is no clear breakage
patterning as it can be seen for the triaxial tests.
Figure 3 shows a representative example of the observed distribution of damage in
the specimens as seen in a horizontal section (middle) and a vertical section (right)
during oedometric and triaxial loading (top and bottom respectively). The black dashed
lines show the position of the cross-section, the red colour (or darker grey areas in
black and white print) indicates the position where more grains have broken and its
intensity relates to a higher percentage in breakage. The selection of the red portions
is only based on visual, yet accurate, inspection. However, quantitative Digital Image
Correlation (DIC [11, 12]) analysis (shown on Figure 4 and detailed in section 3.3)
confirms these observations.
Figure 3. Breakage patterns on oedometer (top) and triaxial (bottom) specimens.
To summarise, it appears that during the triaxial more localised damage occurred,
which forms a shear band and also while we move away from the shear band the
damage becomes less and less, creating different damage zones of breakage and/or
reorientation and/or simple compaction [15]. In addition to the damage being less
intense, there is no evidence of spacial organisation (patterning) at the oedometer test.
In both cases however, there seems to be more breakage on the lower part of the
specimens (where the loading is applied) than in the upper part, as indicated both by
visual inspection and DIC results. This gives a clear impression of the effect of the
moving boundary. Nevertheless, the rest of the boundaries seem not to have an
important effect on the breakage, since hardly any broken grains have been identified
close to them.
Figure 4. Digital Image Correlation results for certain TXC02 increments.
Based on the above observations, it was decided to performed test TXC04, where a
higher resolution was applied. Using a higher resolution in the x-ray scans, more details
about breakage patterns are provided to the price of not acquiring images of the whole
specimen and so bulk measurements could not be performed (Figure 5).
Figure 5. An image of the shear band in TXC04.

3.3. Image processing
After making several observations regarding the deformation of the sand algorithmic
measurements can be performed to confirm and quantify what has been observed. The
DIC results (Figure 4) on TXC02 clearly indicate a localisation of shear strain forming
a shear band (Figures 3, 4, 5) along with evident compactive and dilative volumetric
strains within the band. Also, it is clear from the analysis of increments 2-3 (pre-peak)
that there is more breakage close to the moving boundary, where from the shear band
will start to form as described in the previous section. On the contrary, no localisation
patterns were identified after the DIC was performed on 1DC03, however higher
compactive strains were developed close to the moving boundary, where more
breakage was evident due to boundary conditions.
In order to describe the evolution of breakage the images have to be segmented
into two phases (solid and void) and unnecessary objects have to be removed (e.g.
membrane surrounding the specimen [e.g. 11]). Both these processes have been
carefully performed and multiple methods have been evaluated until an optimal was
selected where the least volume (therefore information) was lost [16]. After segmenting
the images, a watershed algorithm was used to separate the particles in contact, as
described by [7, 18-22], in order to label each individual particle and perform
measurements on each grain. This analysis helps follow the evolution of deformation of
each grain and gain information (e.g. contact network, type of breakage, rotation of
each grain) that will help get a deeper insight as to when a grain breaks; similar
techniques have also been used in literature [7-11, 18-22]. In the following figure an
example of the aforementioned process is presented for a small region of the images,
where grains with the same colour represent the same grain that has been identified.
The ones without colouring are left grey in order to make easier the distinction of the
matching grains.
Figure 6. Segmented, labeled and matched grains. Starting from upper left is the initial (unloaded) loading
stage and the lower right is the last loading stage.
It appears as if some grains break after the peak stress, however it should be
mentioned that this could either be a result (error) of the segmentation method or the
grain might seem broken on one cross section just because it has been sliced through a
slightly (one or two pixels) different plane. The 3D information from the breakage of
the individual grains has been assessed macroscopically by creating a grading curve
and following the evolution of the shape of the curve and therefore, the segmentation
errors have been accounted for. To the best of authors' knowledge, the evolution of
breakage has not been investigated throughout an experimental procedure, on the
contrary most constitutive continuum breakage models [e.g. 3, 14] are fundamentally
based on measurements made before and after the test.
4. Conclusions
The deformation patterns and mechanisms that lead to grain breakage of Caicos ooids
under triaxial and oedometric loading have been studied. The analysis was enabled by
the acquisition of 3D images with XCT. By coupling the macroscopic response to the
XCT observations and thereinafter performing algorithmic measurements of several
properties, valuable information about grain breakage has been gained. The main
objective of this paper is to give a demonstration of the capabilities that XCT offers and
to present some initial results about image processing in regards to particle breakage.
Four triaxial tests and three oedometric tests were performed in total, of which
three and one respectively were done in parallel with XCT. The high resolution of the
images contributed to the precise measurement of parameters such as the density, the
homogeneity and the grain size distribution of the specimens, the position of the shear
band, the distribution of damage and the volumetric strain. The tomographs were also
analysed to determine the size, position and translation of each individual grain, along
with the breakage of the particles. This information not only helps understand in depth
the mechanisms that govern grain crushing, but can also be used as a more detailed and
improved benchmark test for the calibration and comparison of DEM particle scale
simulations.
Acknowledgements
The authors would like to thank Pascal Charrier for his valuable help during the XCT and
Laurent Debove for helping create the oedometric apparatus. Laboratoire 3SR is part of the
LabEx Tec 21 (Investissements dAvenir - grant agreement nANR-11-LABX-0030). The first
author would like to thank the UK Engineering and Physical Sciences Research Council
(EPSRC) DTP PhD studentship and the International Fine Particles Research Institute (IFPRI)
for funding her PhD research. Dr Papanicolopulos acknowledges the support received from the
People Programme (Marie Curie Actions) of the European Union's Seventh Framework
Programme (FP7/2007-2013) under REA grant agreement n 618096.
References
[1] Rudnicki, J.W. 2000. Geomechanics. International Journal of Solids and Structures 37:349-358.
[2] Davis, R.O., Selvadurai, A.P.S. 2002. Plasticity and geomechanics. Cambridge University Press.
[3] Einav, I. 2007.Breakage mechanicspart I: theory. J Mech Phys Solids 55(6):1274-1297.
[4] Luzzani, L., Coop, M.R. 2002. On the relationship between particle breakage and the critical state line of
sands. Soils and Foundations 42(2):71-82.
[5] Muir Wood, D., Maeda, K. 2007. Changing grading of soil: effect on critical states. Acta Geotechnica
3:3-14.
[6] Desrues, J., Chambon, R., Mokni, M., Mazerolle, F. 1996. Void ratio evolution inside shear bands in
triaxial sand specimens studied by computed tomography. Gotechnique 46(3): 529-546.
[7] Hasan, A., Alshibli, K.A. 2008. Spatial variation of void ratio and shear band thickness in sand using x-
ray computed tomography. Gotechnique 58(4):249-257.
[8] Hasan, A., Alshibli, K.A. 2010. Experimental assessment of 3D particle-to-particle interaction within
sheared sand using synchrotron microtomography. Gotechnique 60(5):369-379.
[9] And, E., Hall, S.A., Viggiani, G., Desrues, J., Bsuelle, P. 2012. Experimental micromechanics: grain-
scale observation of sand deformation. Gotechnique 2:107-112.
[10] And, E., Viggiani, G., Hall, S.A., Desrues, J. 2013. Experimental micro-mechanics of granular media
studied by x-ray tomography: recent results and challenges. Gotechnique 3(3):142-146.
[11] Alikarami, R., And, E., Gkiousas-Kapnisis, M., Torabi, A., Viggiani, G. 2015. Strain localisation and
grain breakage in sand under shearing at high mean stress: insights from in situ x-ray tomography. Acta
Geotechnica 10:15-30.
[12] Viggiani, G., Hall, S.A. 2012. Full-field measurements, a new tool for laboratory experimental
geomechanics. ALERT Doctoral school.
[13] Lloyd, R.M., Perkins, R.D., Kerr, S.D. 1987. Beach and shoreface ooid deposition on shallow interior
banks, Turks and Caicos islands, British West Indies. Journal of Sedimentary Petrology 57(6):976-982.
[14] Hardin, B. O. 1985. Crushing of soil particles. J. Geotech. Engineering 111(10): 11771192.
[15] Gabrielsen, R.H., Aarland, R.-K., Alsaker, E. 1998. Identification and spatial distribution of fracture in
porous, siliciclastic sediments. Structural Geology in Reservoir Characterization. Geological Society,
London, Special Publications 127: 4964.
[16] Glasbey, C.A. 1993. An analysis of histogram-based thresolding algorithms. CVGIP. Graphical Models
and Image Processing. 55(6):532-537.
[17] Marak, L., Couprie, M., Talbot, H. 2011. The pink image processing library. Proceedings of EuroSciPy.
[18] Andrew, M., Druckrey, S.M., Khalid, A., Alshibli, M. 2014. 3D behavior of sand particles using x-ray
synchrotron micro-tomography. Proceedings of Geo-Congress.
[19] And, E., Hall, S.A., Viggiani, G., Desrues, J., Bsuelle, P. 2011. Grain-scale experimental
investigation of localised deformation in sand: a discrete particle tracking approach. Acta Geotechnica
7(1):1-13.
[20] Garboczi, E.J. 2002. Three-dimensional mathematical analysis of particle shape using x-ray
tomography and spherical harmonics: application to aggregates used in concrete. Cement and concrete
research 32:1621-1638.
[21] Masad, E., Saadeh, S., Al-Rousan, T., Garboczi, E., Little, D. 2005. Computations of particle surface
characteristics using optical and x-ray CT images. Computational materials science 34:406-426.
[22] Fonseca, J., Sim, W.W., Shire, T., OSullivan, C. 2014. Microstructural analysis of sands with varying
degrees of internal stability. Gotechnique 64(5):405-411.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-441
Early age cemented paste backfill stiffness

development
Lucas FESTUGATOa,1, Nilo Cesar CONSOLIa and Andy FOURIE b
a
Federal University of Rio Grande do Sul
b
The University of Western Australia
Abstract. Paste filling is a process utilized in many underground mining

operations. This involves the filling of tall underground voids with a free flowing
combination of mine tailings, water and cement. In order to evaluate early age
stiffness development, hydration tests were carried out on cemented mine tailings
mixtures. The hydration test involved preparing a sample in a triaxial cell and may
be used to gather the relevant properties relating to stiffness development,
permeability and self-desiccation. This study was focused on the stiffness
development. The mixtures were prepared in a slurry form and therefore to support
the material in this state a perforated mould was utilized. The samples were poured
into a membrane that was supported by this mould and after sealing the sample the
mould was left in the instrument for the duration of the test. Bender elements were
attached to the end platens. These are used to excite small shear waves through the
sample that may be recovered by the receiver element at the opposite end of the
sample. The speed of shear wave could then be used for an estimative of the
materials small strain shear stiffness over time. The tailings, produced from gold
mining, are classified as sandy silt with traces of clay and were cemented with
different amounts of Portland cement, ranging from 3% to 10% by dry weigh of
soil. From the moment of the materials mixture up to about 170 hours, the results
showed the evolution of initial stiffness values raise logarithmic over time and
increase with the increase of the cementation level.
Keywords. Cemented paste backfill, bender elements, small deformation stiffness
1. Introduction
Tailings are essentially comprised by crushed rock waste particles derived from ore
processing [1]. Paste filling is a process utilized in many underground mining
operations. This involves the filling of tall underground voids with a free flowing
combination of mine tailings, water and cement. These voids are known as stopes. In
order to contain the material in a stope (stopping it from flowing into nearby mine
workings) containment barricades are constructed in the stope drawpoint [2]. The
mining process usually includes exposure of a face of these stopes, which can reach
heights greater than 80m in many cases.
In order to evaluate early age stiffness development, hydration tests were carried
out on cemented paste backfill mixtures. The hydration test involved preparing a
sample in a triaxial cell and may be used to gather the relevant properties relating to
1
442 L. Festugato et al. / Early Age Cemented Paste Backll Stiffness Development
stiffness development, permeability and self-desiccation. This study was focused on the
stiffness development.
2. Materials and Methods
2.1. Mine tailings
The studied tailings, produced from gold mining, are classified as sandy silt with traces
of clay. The particles size distribution of the material is presented in Figure 1
(d10=3m, d50=17m, Cu =10, Cc=0.63). The specific gravity (Gs) of the material is 2.72
and its mineralogical composition is presented in Table 1. This material has been
previously studied by [4].
Figure 1. Particles size distribution.
Table 1. Chemical composition of the studied mine tailings

Mineral Chemical composition Amount (%)
Quartz SiO2 29
Muscovite (K,Na)Al2(Si,Al)4O10 (OH)2 24
Albite NaAlSi3O8 25
Calcite CaCO3 1
Dolomite CaMg(CO3)2 17
Orthoclase KAlSi3 O8 4
2.2. Cement
For the samples preparation, general Portland cement (Type II) was used as cementing
agent. The cement content, defined as the mass of dry cement divided by the dry mass
of soil, ranged from 3% to 10%.
L. Festugato et al. / Early Age Cemented Paste Backll Stiffness Development 443
2.3. Sample preparation and test procedure
The weights of moist mine tailings and cement necessary for the samples preparation
were determined with an accuracy of 0.01g. The two materials were mixed in a
mechanical blender for about 10 minutes. The mixtures were prepared in such amount
to produce the samples and allow two measurements of moisture content. The studied
cemented contents were 3%, 5%, 7.5% and 10% by dry weight of mine tailings. The
mixtures moisture content was similar, approximately 28.5%.
The mixtures were prepared in a slurry form and therefore to support the material
in this state a perforated mould was utilized. The samples were poured into a
membrane that was supported by this mould and after sealing the sample the mould
was left in the instrument for the duration of the test. After saturation in the triaxial cell,
confining pressure and back pressure were raised to 500kPa. The Skempton B value
was checked and found to be higher than 0.97. The back pressure valve was then closed
and pore pressure was measured over time. The effective pressure could be determined
by the difference between the confining pressure and the pore pressure.
Bender elements were attached to the end platens. These were used to excite small
shear waves through the sample that may be recovered by the receiver element at the
opposite end of the sample. The speed of shear wave could then be used for an
estimative of the materials small strain shear stiffness over time. Figure 2 schematically
presents the test set up (a) and an image of an actual test (b). Figure 3 presents an
example of the bender element tests carried out in this study. The transmitter bender
element was excited by a single sine-wave pulse, nominally of 14V amplitude, and the
shear wave arrival the other end of the sample was read by the receiver bender element.
The shear wave velocity can be obtained based on the transmission time and the length
of the sample. With the shear wave velocity and the material density, the value of
initial stiffness G0 can be determined.
(a)
(b)
Figure 2. Test set up: (a) schematics (adapted from [3]) and (b) actual test.
Figure 3. Typical bender element test result.
3. Results
Figure 4 gathers the results of hydration tests with initial stiffness measurement, G0,
carried out on cemented mine tailings mixtures.
L. Festugato et al. / Early Age Cemented Paste Backll Stiffness Development 445
Figure 4. Cemented gold mine tailings small strain stiffness against time.
Due to the growth and strengthening of the cement hydrates, the mine tailings
matrix undergoes an increase in stiffness during hydration. According to Figure 4, the
logarithmic evolution of G0 over time is observed for all samples. Furthermore, it is
possible to observe, as expected, the increase of stiffness values with the cement
content increase.
In Figure 5, the evolution of effective pressure increase over time of the mixtures
can be analyzed.
Figure 5. Cemented gold mine tailings effective pressure development against time.
According to Figure 5, the hydration process results in an apparent reduction in

water volume, which leads to a reduction in pore pressure and a corresponding increase
in effective stress. The greater is the cement content, the higher is the water
consumption. It is observed that the effective pressure of the cement mixtures, similarly
to the initial stiffness values, increases over time and tends to a plateau. For the studied
mixtures, the higher the cement content the higher the increase of effective stress.
4. Conclusions
Based on hydration tests carried out on cemented gold mine paste backfill, a few
conclusions can be draw.
From the moment of the materials mixture up to about 170 hours, due to the
growth and strengthening of the cement hydrates, the results showed the evolution of
initial stiffness values raises logarithmic over time and increase with the increase of the
cementation level.
Similarly, as a result of water consumption during hydration, the effective pressure
increase of the cemented mixtures increases over time and tends to a plateau.
The greater is the cement content, the higher is the initial stiffness and the effective
pressure of the cemented mixtures.
References
[1] D. Wijewickreme, M.V. Sanin & R.G. Greenaway. Cyclic shear response of fine-grained mine tailings.
Canadian Geotechnical Journal 42 (2005), 1408-1421.
[2] M. Helinski, A. Fourie, M. Fahey. & M. Ismail. Assessment of the self-desiccation process in cemented
mine backfills. Canadian Geotechnical Journal 44 (2007), 1148-1156.
[3] M. Helinsky. Mechanics of mine backfill. Thesis (Ph.D.), The University of Western Australia (2007).
[4] L. Festugato, A. Fourie & N.C. Consoli. Cyclic shear response of fibre-reinforced cemented paste
backfill. Gotechnique Letters 3 (2013), 5-12.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-447
Influence of volumetric and shear strains

on the destructuration of saprolitic soils
Irene ROCCHIa, 1 and Matthew R. COOPb
a
Department DICAM, University of Bologna, formerly City University of Hong Kong
b
ACE Department, City University of Hong Kong
Abstract. Based on a review of numerical models it would appear that debate is

still open as to which combination of volumetric and shear strains is responsible
for destructuration of intact soils. Direct experimental evidence is rather limited
and mostly related to sedimentary clays. A saprolitic soil from Hong Kong that
had been block sampled was tested to investigate the effects of sample disturbance
and more in general of destructuration. The amount of destructuration was
quantified by measuring the elastic shear stiffness by means of bender elements
mounted axially in triaxial cells. The tests were designed to investigate separately
the effects of volumetric and shear strains, by way of loading cycles reaching
increasingly larger strains. The loading cycles were carried out under isotropic
conditions and at constant mean effective stress for the volumetric and shear
strains, respectively. In addition, the results obtained were compared with those
from a test on a reconstituted specimen that was carried out in a similar fashion. In
general, a reduction in shear stiffness was observed for the intact specimens, while
a slight increase was measured for the reconstituted specimen at the same strain
levels. The same held true also after normalising for void ratio, which was
necessary as this changed as a result of straining hence having an effect on the
shear stiffness. The reduction in stiffness observed for the intact specimens, which
indicates loss of structure, was observed for both test types. However, at a given
strain threshold, the reduction was larger for the specimen that underwent cycles of
volumetric strains.
Keywords. Residual soils, laboratory testing, structure
1. Introduction
The importance of destructuration is nowadays widely acknowledged and a number of

numerical models exist that try to take into account its effects to simulate the
progressive reduction in stiffness and/or size of the elastic zone, caused for example by
the breakage of the interparticle bonding. Some of these numerical models take into
account only volumetric strains [1] to calculate the amount of destructuration, while
others a combination of volumetric and shear strains, which may consider them equally
[2] or not [3]. However, the experimental data that support one choice over the other
are limited to a number of sedimentary clays [4] and generally are based on a
comparison of the behaviour in shear of specimens that were either compressed
isotropically or along a k0 path.
To contribute to the understanding of the destructuration mechanisms in non-
sedimentary soils, a granitic saprolite from Hong Kong was tested in a triaxial system
equipped with bender elements. Initially, the elastic shear stiffnesses of an intact and a
448 I. Rocchi and M.R. Coop / Inuence of Volumetric and Shear Strains on the Destructuration
reconstituted sample were compared while applying isotropic cycles of loading to

identify any effect of destructuration. A second intact sample was then tested applying
deviatoric cycles of loading, i.e. at constant mean effective stress, so that the reduction
in stiffness could be compared with that observed during the isotropic loading of the
other intact sample and it could be assessed directly whether shear or volumetric strains
result in greater destructuration.
2. Soil Properties and Methodology
The soil tested was a saprolite from the Kowloon Granite belonging to grade V, i.e.
Completely Decomposed Granite [5], which was block sampled on Hong Kong island
at a shallow depth (1.7m). Its particle size distribution is presented in Figure 1. It can
be observed that the soil consists of gravel and sand in approximately equal parts and
the average and maximum sizes are 2.2 and 10mm, respectively. The soil is very well
graded, as is typical of soils originating from chemical weathering, which do not
undergo any transportation. Its coefficient of uniformity is equal to 23 and that of
curvature is 2.
To prepare the intact specimens the block sample was cut into sub-samples, while
keeping it confined in its wooden sampling box, until its size was only slightly larger
than that of the dimensions required for the triaxial specimen, i.e. 60mm diameter and
120mm height. The sample was then placed in a hand lathe and manually trimmed into
a cylindrical shape using a hack saw. The trimming was carried out with great care so
as not to disturb the sample. As the sample had a low water content (3-6%), it was not
necessary to control the humidity while trimming. The soil at the depth sampled would
have experienced many cycles of wetting and drying, as a result of the rainy and dry
seasons.
Figure 1. Particle size distribution.

I. Rocchi and M.R. Coop / Inuence of Volumetric and Shear Strains on the Destructuration 449
After completing the trimming process the specimen was weighed and its
dimensions measured so as to calculate its initial void ratio. The specimen was then
placed onto the pedestal of the triaxial cell. A small inset was excavated in each of the
two faces of the specimen with the hack saw to house the bender elements. These were
backfilled with the fine fraction of the trimmings to ensure a better contact between the
elements and the soil. The reconstituted specimen was prepared by moist tamping the
soil into a membrane inside a mould directly on the pedestal and applying a small
suction afterwards until the cell was filled with water and a confining pressure could be
applied. Both intact and reconstituted specimens were saturated by back pressure until
a B value in excess of 0.97 was achieved. The triaxial cell was equipped with local
radial and axial LVDTs. These were used to measure the strains for both the intact and
reconstituted specimens as the membrane penetration was significant due to the large
grain size. The connection between the specimen and the load cell was achieved
through a half ball so that any slight tilt of the top platen did not cause strain
localization. However, this meant that it was not possible to bring the specimen into
extension.
The tests were devised to start from an isotropic state close to the estimate of the
minimum in-situ mean effective stress (pin situ). Since neither in-situ measurements of
k0 nor suction were available, pin situ was calculated based on the overburden pressure.
In addition, it was considered that at some point in its history, the soil would have
experienced full saturation, with the water table at the surface, which represents a lower
bound for the in-situ stress. Due to the shallow depth, this meant that the initial stress
was so low that it was not possible to study the behaviour in swelling.
The testing programme consisted of two tests on intact specimens and one on a
reconstituted sample, the details of which can be found in Table 1. Figure 2 shows
schematically the stress paths adopted. Starting from pin situ (A), the specimen was put
through cycles of either isotropic or deviatoric compression reaching threshold strains
(B) at which the shear wave arrival time (ta) from the bender elements was measured.
Subsequently, the specimen was brought back to the in-situ stress (C) and ta was
measured again. A new probe was then started reaching a greater strain (D), based on a
logarithmic increase of strain. However, the readings shown are not exactly at the
values of strain set as thresholds, as part of the strains were recovered during the
unloading or because the strains did not cease immediately once the stress was kept
constant and the analysis has been made in terms of total strains.
The arrival time (ta) was identified comparing the received signals at the
frequencies 4, 5 and 6 kHz. This range was used based on previous tests on other
saprolitic soils from Hong Kong, where a wider range of frequencies was investigated
(1-10kHz) and ta was found to be identified best at these intermediate frequencies. The
single shot sine wave sent to the transmitter had an amplitude of 10V. In Figure 3, the
signals for these three frequencies can be observed and it is clear that they show the
same arrival time.
Table 1. Details of the testing programme.

Test code e0 (-) w0 (%) Probe direction
RV 0.91 15 Isotropic
IV 0.91 6 Isotropic
IS 1.11 3 Deviatoric
Figure 2. Scheme followed during the tests.
Figures 4 and 5 show the evolution of the elastic shear stiffness (G vh) with volumetric
strain (vol) for an intact and a reconstituted specimen as progressive cycles of isotropic
stress were applied. These are the tests RV and IV, respectively, in Table 1. Although
the reconstituted specimen was prepared at a similar initial void ratio, the stiffness
presented is normalized for the void ratio (e) according to Jamiolkowski et al. [6], as it
changes by different amounts for the intact and reconstituted specimens.
Figure 3. Example of the bender elements signal at difference frequencies.

Figure 4. Effects of volumetric change on the elastic shear stiffness for a reconstituted sample.
For both tests, as the stress reached increases so does Gvh. However, not only does
the reconstituted specimen have a lower initial value (Gvh,0,i), but its increase is less
than for the intact sample. As the stress is reduced, the strains are not fully recovered,
indicating that the soil is already in the plastic region. However, while the reduction in
Gvh and the recovery in strain proceed similarly for RV, for IV the stiffness initially
reduces while almost no strain is recovered and then its reduction slows down
considerably while some of the strain is recovered. For Gvh at the in situ stress, the
value is almost constant for the reconstituted specimen, with a slight increase as the
stresses reached become significant, while there is a slow but constant reduction for the
intact specimen.
Figure 5. Effects of volumetric changes on the elastic shear stiffness for an intact sample.
In Figure 6 the values of the stiffness at pin situ (Gvh,0) for the effect of volume
change are compared with those of a similar test that had probes in the deviatoric
direction, test ID in Table 1. The initial void ratios were slightly different for the two
specimens as shown in Table 1 and so the data presented are again normalized for the
void ratio. The initial stiffness (Gvh,0,i) for ID was greater than for IV, despite its initial
void ratio being slightly higher. However, given its magnitude the difference might be
explained by heterogeneity, which is common for saprolitic soils. As the loading cycles
are applied, the change with the plastic shear strains after unloading to the in situ stress
(s,pl) for ID is rather slow, but constant, while Gvh,0 remains approximately constant
initially for IV and then reduces slowly but constantly until there is a sudden drop
before vol,pl=2%. When observing the development of vol with p, a change in slope
can be observed at about this point.
Figure 6. Effects of volumetric and shear strains on the elastic shear stiffness of intact samples.
4. Conclusions
To add to the experimental dataset related to the effects of destructuration, a saprolitic

soil from Hong Kong was tested imposing either progressively larger volumetric or
shear strains and measuring the elastic shear stiffness by means of bender elements in a
triaxial cell. It was clear that volumetric strains progressively caused loss of stiffness in
the intact sample, while an increase was observed for the reconstituted specimen, which
indicates that saprolitic soils are subject to destructuration. When comparing the effects
of shear and volumetric strains on intact samples, Whe loss in stiffness was slightly less
for the sample subjected to isotropic loading, for which a significant reduction was
observed only at much larger volumetric strains.
Acknowledgments
The work described in this paper was fully supported by a grant from the Research
CityU 112911). The authors would like to thank Ir Ken Ho from GEO for kindly
providing the sample tested.
References
[1] R. Lagioia & R. Nova, An experimental and theoretical study of the behaviour of a calcarenite in triaxial
compression, Gotechnique 45 (1995), 633648.
[2] B. Baudet & S. Stallebrass, A constitutive model for structured clays, Gotechnique 54 (2004), 269-278.
[3] M. Kavvadas & A. Amorosi, A constitutive model for structured soils, Gotechnique 50 (2000), 263-273.
[4] L. Callisto & S. Rampello, An interpretation of structural degradation for three natural clays, Canadian
Geotechnical Journal 41 (2004), 392-407.
[5] Geotechnical Engineering Office. Guide to rock and soil descriptions. Geoguide 3. Geotechnical
Engineering Office, Hong Kong, 1988.
[6] M. Jamiolkowski, S. Leroueil, & D. Lo Presti, Design parameters from theory to practice. Geo-Coast
1991: Proceedings International Conference on Geotechnical Engineering for Coastal Development.
Yokohama (1991), Japan, 877-917.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-454
Description and Calibration of Triaxial

tests with internal measurement of
displacement on artificially cemented
lateritic soil
Coutinho, Roberto Q. a,1, Santos Jr, Olavo F. b, Severo, Ricardo N. F. c
a
Federal University of Pernambuco, Recife PE, Brazil, [email protected]
b
Federal University of Rio Grande do Norte, Natal - RN, Brazil, [email protected]
c
Federal Institute of Education, Science & Technology of Rio Grande do Norte, Natal -
RN, Brazil, [email protected]
Abstract. This work is part of a larger study that was conducted in order to identify
the instability mechanisms of a cliff in northeastern Brazil. Previous studies
indicated that the stability of cliffs in this region is guaranteed by natural
cementation between soil particles. A comprehensive study of the effects of
cementation on the mechanical behavior of the soil was carried out in the doctoral
thesis of Severo (2011) in a joint research UFPE-UFRN. Drained triaxial tests on
artificially cemented samples were performed. The purpose of this article is to
describe the equipment, the methodology developed to analyze the influence of
cementation on soil behavior, and also to demonstrate the effects of using internal
sensors to measure displacements. Pressure cells were used at the base and at the
confining pressure line of the triaxial cell. Expansion and contraction of the samples
were recorded by means of an electronic device to measure the volume variation. A
load cell was used to measure the axial load on the sample. Displacement external
sensors were coupled on the top of a triaxial cell, and internal sensors were attached
to the sample. Internal instrumentation consisted of Hall Effect sensors, which
allowed us to measure radial and axial displacement of the samples. Test results are
presented in the form of deviator stress versus axial strain and volumetric strain
versus axial strain curves. The results showed that the use of external sensors do not
reflect the initial behavior of the soil under small strains. This initial behavior can
only be observed with the use of equipment for measuring small displacements. This
aspect is particularly important to measure the initial stiffness. For large strain
conditions, the external measure proves to be more appropriate.
Keywords. Keyword, keyword triaxial tests, Hall Effect sensors, measuring small
displacements
1. Introduction
This paper is part of a wider study to identify the instability mechanisms of a cliff in the
Barreiras Formation at Ponta do Pirambu in the town of Tibau do Sul, northeastern Brazil,
considering the contribution of soil cementation in its behavior. Earlier studies indicate
that the stability of cliffs in this region is guaranteed by the natural cementation between
soil particles. The purpose of this article is to present the equipment and methodology
1
Coutinho, Roberto Quental.
R.Q. Coutinho et al. / Description and Calibration of Triaxial Tests 455
adopted to analyze the influence of soil cementation on the cliffs behavior.

It is an experimental study on the mechanical behavior of lateritic sedimentary soils.
Twenty-seven (27) isotropically consolidated drained triaxial tests (CID) were
performed in a laboratory on larger samples (height 200mm and diameter 100mm), with
measurements of small axial and radial strains (Hall effect sensors). All the triaxial tests
were performed using electronic instrumentation on load, pressures, volume and
displacement.
2. Equipment and Methodology
2.1. Planning, Procurement and Implementation of New Equipment
After detailed planning an exchangeable triaxial cell was made and installed for samples
with diameters of 5.0 cm or 10.0 cm and height twice the diameter, with an internal and
external load cell. The material used to make the cell was acrylic, anodized naval
aluminum and stainless steel.
The triaxial cell is hybrid to operate with both the mercury column system and
through a hydraulic/pneumatic system. The fully assembled cell has external dimensions
of 46.0 cm in height and 32.0 cm in diameter.
This triaxial cell has a measurement system for small strains, two axial and one radial
Hall Effect sensors, two pressure cells (confining stress and backrpressure) and LVDTs
(Linear Variation Displacement Transducers) for external measurement of strains, and
the ELE International meter of electric volume variation.
For the systems operation a data concentrator box was made, connected to an
Agilent HP data acquisition box and with installed HP software, compatible with the new
equipment.
The system also has an S-40 external electronic load cell, a HBM traction and
compression of 3.0 tons, and internal load cells (compression) of 300 and 500 kg, to be
used on request.
Metal molds were made and procured measuring 50 mm x 100 mm and 100 mm x
200 mm, and other necessary equipment to adapt the metal molds to the 10-ton load
frame of Wykeham-Farrance to mold the static compaction samples.
Figure 1 shows clockwise some items of equipment installed in the UFPE
Geotechnics and Instrumentation Laboratory, namely: Hall effect sensors (a) and (b);
triaxial cell (c), placed in the 10-ton load frame, with the external load cell, external
vertical extensometer, pipes coming from the bladers, and mercury columns; converters
of pneumatic pressure into hydraulic pressure (bladers) together pressure valves and air
filter with pneumatic pipes in galvanized steel (d), Agilent HP data acquisition system
and the source (e); 10-channel data concentrator box made for compatible running of the
LVDTs and sensors with the HP system; and lastly (f).
2.2. Calibration Procedures
2.2.1. Hall Effect Sensors

The Hall Effect is the production of a voltage difference (the Hall voltage) across an
electric conductor, transverse to an electric current in the conductor and a magnetic field
456 R.Q. Coutinho et al. / Description and Calibration of Triaxial Tests
perpendicular to the current. Edwin Hall discovered it in 1879. The Hall coefficient is
defined as the ratio between the induced electric field to the product of the current density
and the applied magnetic field. It is a characteristic of the material comprising the
conductor, since its value depends on the type, number and properties of the load carriers
forming the current.
A size gauge 160 x 50 x 30 mm and micrometer were used to calibrate the Hall
Effect sensors. One radial and two axial sensors were used. For every hundredth of a
millimeter displacement on the micrometer the response was found on its Hall effect
sensor. To use these sensors work is done only on their linear stretch; that is, achieved
during the setting up of the test. The calibration curves of the Hall Effect sensors are
shown in Figure 2 (a) Axial 1, (b) Axial 2 and (c) Radial.
(a) (b) (c)
(d) (e) (f)

Figure 1. Equipment installed in the UFPE Geotechnics and Instrumentation Laboratory
2.3. Drained Triaxial Tests
Drained triaxial tests were run, with measurements of small strains, using the Hall effect
sensors on soil samples, prepared in the laboratory with artificial cementation, using
Portland cement. The chosen soil was from the top layer of the cliff since it was the most
homogeneous, both in color and chemical and mineral composition.
(a) Axial 1 (b) Axial 2
(c) Radial
Figure 2. Calibration of Hall effect sensors
2.3.1. Preparing the Samples

The samples were molded with the necessary quantities of soil, distilled water and
cement to reach the cement percentage, moisture content and dry apparent specific mass,
provided for each sample. A quantity of fines equal to the weight of the cement added to
the mix was removed from the soil for each sample, in order not to alter its particle size.
They were placed in a cylinder with internal dimensions of 200 mm in height and 100
mm in diameter in four statically compacted layers in order to reach the dry unit weight
determined for each sample.
Before placing the next layer, the previous layer was scarified. The mold in every
molding was lubricated with Vaseline and filter paper was placed on both its base and
top to prevent soil adhesions. The mixing and compacting procedure, after pouring in
water for kneading, lasted no more than 30 minutes, to prevent excessive losses of
moisture.
After mixing and homogenizing the material, the quantities referring to each layer
were packed in plastic zip bags to reduce moisture loss. The natural moisture of the soil
crumbled in the laboratory is around 0.5%, and this initial moisture was found to
compensate the losses during molding. The samples were weighed immediately after
molding, discounting the weight of the mold. The weight and dimensions of each sample
are controlled and carefully assessed using a strict acceptance criterion. The few samples
that did not achieve the desired specifications were discarded.
Tolerance in relation to the acceptance criteria is as follows: variation of up to 1%
in the dry unit weight, a variation of 1.0 mm for the diameter and 2.0 mm for the height
of the sample and molding moisture content varying between 9.5% and 10.0%.
After 12 hours the samples were removed from the molds. Throughout this time they
remained in plastic zip bags in the wet room, and on the sixth day were placed in a plastic
bucket with distilled water in the wet room to facilitate saturation.
On the seventh day the test was set up and the sample left to percolate from one day
to another, from the bottom upwards with an effective stress of 20 kPa. On the eighth
day, that is, after seven days of molding, they were saturated, consolidated and sheared.
The sequence of molding activities can be seen in Figure 3.
2.3.2. Assemblage, Percolation, Saturation and Consolidation of the Tested Samples

On the sixth day of curing, the cured artificially cemented samples in the wet room
are immersed in a bucket of distilled water to facilitate saturation. On the seventh day of
curing, the sample is taken from the bucket, inside the wet room, and placed in the trixial
cell for setting up the CID test, as described in the following procedure: The metal porous
stone with filter paper both at the top and bottom of the sample and the rubber membrane
is immediately put in place, adapting the top cap to circulate the percolation water from
the bottom up in the sample; next, the Hall effect sensors are glued and regulated to the
rubber membrane, so that its response signal is on the linear stretch of the calibration and
with room to move in the linear stretch to prevent readings in the non-linear stretch.
After assembling and filling the triaxial cell with distilled water, water began to
circulate through the sample, and the water percolates from the bottom to the top, with a
head of 10 kPa and confining stress of 30 kPa. After 24 hours of water circulation and a
quantity of water at least equal to half the sample volume or more has percolated,
saturation begins.
The CID triaxial tests were saturated by back pressure to 400 kPa, based on the pore-
pressure parameter B (Skempton, 1954). The technique used was to apply increases of
50 kPa of total stress at half-hour intervals, keeping the effective stresses constant, in the
case of 20 kPa. After applying total stress of 230 kPa the Skempton parameter B was
checked at each increase of 50 kPa. The last increase is the only one of 40 kPa, and the
resulting confining stress is 420 kPa, with back pressure of 380 kPa; after stabilizing the
pressures the final value of parameter B was obtained and lastly, the back pressure of
400 kPa is applied.
After saturation, the samples with effective stress of 20 kPa were considered
consolidated for this stress and, in the case of the samples with 100 kPa and 300 kPa of
effective stress, consolidation occurred until the volumetric meter reading stabilized. For
effective stresses of 100 kPa and 300 kPa the confinement pressures of 500 kPa and 700
kPa are applied through 50 kPa stress increases until reaching the desired stress. Figure
4 shows the installation sequence of the sample in the triaxial cell.
(a) Mixing soil with cement (b) Homogenizing and adding (c) Weighing the four layers
distilled water
d) Static compaction by layer, (e) Unmolding, weighing and (f) Curing in wet room (last day
with scarification acceptance of sample under water)
Figure 3. Sample molding procedure
a) Placing filter paper and b) Placing the membrane and c) Placing the top-cap for
porous stone rubber circlips bottom-to-top water circulation
d) Placing and regulating the e) Sample with a radial and two f) Triaxial cell installed ready for
radial Hall effect sensor axial Hall effect sensors percolation
Figure 4. Sequence of installing the sample in the triaxial cell
2.3.3. Shear, Dismantling of Triaxial Test and Extraction of Tested Samples

After consolidation, the sample begins to shear at an axial strain rate of 0.012 mm/min,
slow enough to prevent generation of pore-pressure, which was measured and monitored
while testing. The velocity was defined based on the drainage conditions and through
triaxial tests performed before defining the experimental program.
After concluding the test, the sample was photographed, the equipment dismantled,
the sample diameter measured and the water content was determined, to check the
moisture and post-rupture saturation. Figure 5 shows a view of the dismantling of the
test with an already ruptured artificially cemented sample.
(a) Sample ruptured with membrane

(b) Rupture plane side view (c) Rupture plane front view
and sensors
Figure 5. Sample CID(2)100A3 rupture in the triaxial cell and consequent dismantling of the test
3. Effects of using internal strain sensors to measure stiffness of the samples
The graph in Figure 6 shows on a semi-logarithmic scale the effect of adjusting the
sensors for the sample CID(3.5)300A2 (drained triaxial test with 3.5% cement content,
confining shear stress of 300 kPa, A2 corresponding to a dry unit weight of 18.0 kN/m)
until the axial strain of approximately 6%, representating various effective stresses,
cement contents and densities used in this study; in other words, for any tested density
or cementation content the effect is always noted of using the sensors in the samples
stiffness.
When we analyze the use of the internal and external axial strain sensors, we see the
effect of adjusting the equipment absorbed by the external sensor and completely
eliminated by the internal sensors at the start of shearing. In any case, it is found that,
with or without internal strain sensors, the deviator stress remains the same, at the peak
and in large strains. The adjustment seems to only affect the measurement of the sample
stiffness, causing a translation of the axial strain measurement. It is important to note
that the initial levels appearing in the Figure 6 graphs correspond to an extremely small
strain of less than 0.01% and are highlighted due to the effect of the scale on the semi-
logarithmic graph. Nevertheless, the measurements made by the Hall effect sensors are
already noticeable, while in the case of the green curve, the external sensors have only
been ajdusted in relation to the equipment.
In this case, the internal sensors during the initial phase more accurately read the
strains, since they eliminate the effects of adjusting the press to the cell and the sample
inside it. Next, from a certain point, specific for each sample, depending on the internal
strain sensors behavior, the external strain sensors measurements now become more
accurate, since the field of action of the internal strain sensors is restricted by its linear
response stretch, as shown in item 2. This transition point, depending on the sensors
behavior, may occur before, during or after the rupture and is a single one for each test.
Figure 6. Effect of using the Hall ingternal strain sensors to measure the stiffness of the samples
4. Conclusions
The results showed that the use of external sensors do not reflect the initial behavior
of the soil under small strains. This initial behavior can only be observed when using
equipment for measuring small displacements. This aspect is particularly important to
measure the initial stiffness. For the conditions of large strain, the external measure
shows to be more appropriate.
It was also concluded that the eight sensors duly calibrated and checked throughout
the testing had an excellent performance, together with all triaxial equipment and fittings.
References
[1] BISHOP, A. W.; HENKEL, D. J. The measurements of soil properties in the triaxial test. 2. ed. London:
Edward Arnold, 1962. 227p.
[2] SEVERO, R.N.F. (2011). Caracterizao Geotcnica da Falsia da Ponta do Pirambu em Tibau do Sul -
RN Considerando a Influncia do Comportamento dos Solos nos Estados Indeformado e Cimentado
Artificialmente. Doctoral thesis, UFPE, 280p.
[3] SKEMPTON, A. W. The Pore-pressure Coefficients A and B. Gotechnique, London, v. 4, p. 143-147.
1954.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-462
Studying collapse behaviour of sandy silt

under generalised stress conditions
Octavio E. CRDENASa, Rodrigo C. WEBERb
Enrique ROMEROb,1, Antonio LLORET b and Josep SURIOLb
a
Universidad Autnoma de Coahuila, Torren, Mxico
b
Universitat Politcnica de Catalunya, Barcelona, Spain
Abstract. This paper presents preliminary results on the collapse response of an

artificially prepared sand-silt mixture in a hollow cylinder apparatus. The wetting
stage is performed under constant mean and deviatoric stresses but at different
intermediate principal stresses (controlled through the principal stress parameter
b = (2-3)/(1-3)). Properties of the mixture are first described to complement its
hydro-mechanical information. The preliminary results seem to indicate that
collapse is larger when the intermediate stress coincides with the minor one, i.e.
under conventional axi-symmetric triaxial state.
Keywords. Unsaturated soil, collapse, hollow cylinder apparatus
1. Introduction
There are few experimental results on the influence of the intermediate principal stress
on the behaviour of unsaturated soils. Hoyos [1] and Matsuoka et al. [2] used a true
triaxial apparatus, whereas [3-4] a hollow cylinder device, to perform tests under
constant suction to study the influence of this constitutive stress variable on soil
behaviour under three-dimensional stress conditions. Nevertheless and to the authors
knowledge, the effect of intermediate stress on the collapse response on soaking has
not been studied in depth. In fact, this effect is by no means an academic issue, since it
has been taken into account in modelling earthwork constructions (see for instance [5]).
The authors of the present paper consider that this limitation is related to experimental
difficulties, mainly associated with the homogenisation of the hydraulic field (this is
particularly important when working with large samples as in the hollow cylinder
apparatus and when liquid contact/pressure is applied at the lower boundary) and to the
large deformations undergone on soaking. Important deformation non-uniformities
occur on soaking in the specimen as a result of its curvature and end-restraint effects,
and this is important when selecting the type of transducers and the way to measure the
displacement field and thus precisely determine the volume change response.
To get further experimental insight into the behaviour of unsaturated soils under
generalised stress conditions, a preliminary experimental program has been carried out
to explore the deformational response of a lightly compacted and artificially prepared
mixture upon soaking using a hollow cylinder apparatus. A mixture of 30% sand and
1 Corresponding Author: [email protected]

O.E. Crdenas et al. / Studying Collapse Behaviour of Sandy Silt 463
70% clayey silt by dry mass has been statically compacted at low dry density and low
water content to induce high collapsibility on suction reduction.
2. Artificially prepared mixture used in the investigation
The soil used is the result of the mixture of two materials: fine sand of Castelldefels
(Barcelona) and Barcelona (BCN) clayey silt, a reference material widely used in the
Geotechnical Laboratory [6]. The particle size distribution of the sand, gs, is plotted in
Figure 1. Regarding BCN clayey silt, it presents a liquid limit wL = 36% and a plastic
index PI = 19%, and Figure 1 shows its original particle size distribution curve. This
curve has been modified, gc, to remove particle sizes larger than 1.18 mm that are not
present in the sand (modified BCN clayey silt in Fig. 1).
100
80
Percent finer (%)
60 Barcelona
clayey silt
Castelldefels
40 sand
Modified BCN
clayey silt
20
Mixture
0
10 1 0.1 0.01 0.001
Diameter (mm)
Figure 1. Particle size distribution curve for constituents (sand and clayey silt) and mixture.
The mixing ratio was studied to find a suitable combination of shear strength,
water permeability and volume change properties on soaking. Finally, a mixture of
30% sand and fc = 70% clayey silt by dry mass was selected. Figure 1 shows the
particle size distribution curve of the mixture: gm = 0.3gs+0.7gc. The solid density of
the mixture is sm = 2.66 Mg/m3 ( ss = 2.65 Mg/m3 for the sand and sc = 2.67 Mg/m3
for the clayey silt). The mixture with a water content of 4.0 % (suction around 20 MPa)
was statically compacted up to a dry density of 1.55 Mg/m3 (void ratio e = 0.716,
degree of saturation Sr = 0.15). The vertical stress applied to compact the sample was
slightly below 10 kPa.
Since the sand grains are immersed in the dominant clayey/silty matrix, it is
important to know the real void ratio of the clayey silt. This real void ratio will control
the collapse behaviour, the permeability and water retention properties of the mixture.
The real dry density of the clayey silt can be estimated by the following expression (in
the expression, e = 0.716 represents the overall void ratio of the mixture)
464 O.E. Crdenas et al. / Studying Collapse Behaviour of Sandy Silt
6 789
45 ; < =>?#@ABCD (1)
: 6 89
<86
This real dry density is associated with a real void ratio of the clayey silt ec = 1.02,
which is large enough to induce an important collapse on wetting (refer to the next
section), to present suitable water permeability k = 4.610-8 m/s and to display suction
values that are measurable by tensiometer readings, as will be shown below.
Figure 2. Above: retention curve (wetting) for the different constituents and mixture. Below: retention curve
(drying) for the different constituents and mixture.
Figure 2 displays the water retention curves on wetting and drying for the different
constituents (sand and BCN clayey silt) and mixture. Data have been obtained with
different techniques depending on the suction range, namely axis translation,
tensiometer measurements and dew-point psychrometer readings. The water retention
information of the BCN clayey silt that was measured with axis translation technique
corresponds to a slightly lower void ratio (e = 0.75 to 0.82). The retention curve of the
mixture will tend to w = 26.9% at saturation, which matches data of the BCN clayey
silt at low suctions. It is anticipated that the wetting and drying curves of the mixture
will approximately follow the curves of the filling BCN clayey silt in the low suction
range, where void ratio effects are dominant in the storage mechanism. In the high
suction range (above 1 MPa), dominated by adsorption mechanism, the mixture will
display a somewhat lower water content (approx. 70%) of the corresponding water
content of the BCN clayey silt. The retention curve will be used to estimate suction
during the soaking process, since no tensiometers were installed in the preliminary
stage of the hollow cylinder tests.
3. Preliminary insight into the collapse response of the mixture
A first insight into the volume change behaviour of the mixture during soaking has
been performed using oedometer cells (50 mm in diameter and 20 mm high). Different
stress paths have been followed, and the volume change results are plotted in Figure 3
The first path involved loading the mixture at constant water content (w = 4.0% and
indicated as loading constant w/c in the figure) from 10 kPa to 400 kPa. The water
undrained path approximately corresponded to a constant suction path at s = 20 MPa,
since it developed in the water retention regime not affected by void ratio changes. A
maximum Sr = 0.22 was attained at the maximum vertical stress. Another
loading/unloading path was performed under saturated conditions (saturation was
induced at a very low vertical stress around 2 kPa). A last type of path was performed,
which involved fast loading at constant water content followed by soaking at constant
vertical stresses (10, 20, 50, 100, 200 and 400 kPa). Soaking was performed after
10 min of the water undrained loading step. As observed, the material underwent
collapse even at the small vertical stress of 10 kPa. Maximum collapse deformation
(around 10.3%) was attained at a vertical stress of 200 kPa. This stress level will kept
when performing the hollow cylinder tests, which will be performed under a mean
stress of p = 210 kPa.
0
Loading/unloading
saturated
Loading constant
5 w/c
Volumetric strain (%)
Soaking under
load
10
15
20
25
10 100 1000
Vertical stress (kPa)
Figure 3. Compressibility on loading (at constant as-compacted water content and saturated states) and
results of soaking under load tests.
4. Collapse under constant mean and deviatoric stresses but at different

intermediate principal stresses
4.1. Hollow cylinder apparatus
The sample installed in the hollow cylinder apparatus features 50 mm of external radius
r0, 30 mm in inner radius ri, and 200 mm in height Ho. The capacity of the axial load
cell and the torque cell are 10 kN and 100 Nm, respectively. The principal stress
parameter b = (2-3)/(1-3) [7] can be automatically controlled during the tests (i
are principal stresses, being 1, 2 and 3 the major, intermediate and minor ones,
respectively). Both the base and top caps have conventional porous discs. Three
pressure / volume controllers were used for the application and control of the inner and
outer cell pressures (P0 and Pi), as well as for the measurement of the water volume
changes. Figure 4 shows a scheme of the geometry, force (torque) and pressures
applied, as well as the stress state of the soil with the definition of the deviatoric stress
(q), mean stress (p) and Lode angle () invariants. The procedure to obtain the principal
stress values from the applied force and pressures has been described in [8].
Figure 4. Stress state in the hollow cylinder device.
4.2. Stress paths followed
The specimen was compacted inside a set of two concentric cylindrical moulds: an
inner mould with diameter 60 mm and split up into four parts and an external split
mould of 100 mm in diameter. The static compaction was carried out by a hydraulically
driven press at a displacement rate of 0.2 mm/min up to a maximum vertical stress
around 200 kPa (in this case lateral friction affected the value of the maximum stress
applied to fabricate the mixture). The assembly of the specimen in the cell was
carefully carried out due to the high compressibility of the mixture. Once the sample
was installed, the inner and outer chambers were filled with water. Once the filling was
finished, the external pressure was fixed at 10 kPa in the pressure controller to maintain
certain confinement on the sample (this confinement is expected to induce some
volumetric compression as detected in the oedometer test results).
To determine the effect of the intermediate principal stress 2 on collapse, three

preliminary tests were performed by changing the value of parameter b, which was set
to 0, 0.5 and 0.8. The tests were carried out by maintaining constant the mean p and
deviatoric stresses q during the saturation stage. The tests started with an isotropic
compression stage by applying approximately the same inner and outer chamber
pressures up to a maximum mean net stress p = 200 kPa under constant water content
conditions (w = 4.0%) and atmospheric air pressure. A rate of 2.5 kPa/hour was
selected for the application of the external Po and internal Pi pressures during isotropic
compression. After this initial stage, a small increase of deviatoric stress q = 30 kPa
was applied before setting parameter b. Afterwards, the value of q was increased at a
rate of 15 kPa/hour up to approximately 200 kPa by maintaining p . 200 kPa and
approximately constant b parameter. Once the target stresses were reached, the samples
were soaked under constant stress state by setting a small water pressure at the base
(< 10 kPa). Water volume changes on soaking were monitored by an automatic
pressure / volume controller. The evolution of suction during this process can be
approximately estimated by the retention curves, since no tensiometers were installed
in the hollow cylinder apparatus. Table 1 presents the principal stresses and stress
invariants at different b values, which were kept constant along the hydraulic process.
Table 1. Principal stresses (and stress invariants) applied during the three soaking tests performed in the
hollow cylinder apparatus.
b (-) 1 (kPa) 2 (kPa) 3 (kPa) q (kPa) p (kPa) ()

-30 (triaxial
0 340 125 125 215 197
compression)
0.5 320 205 90 200 205 0
0.8 290 245 60 211 198 19.1
The axial strain z was evaluated directly from the vertical displacements controlled
by the stepper motor and the initial height of the sample. Nevertheless, the most
important limitation was the lack of local instrumentation in the sample for volumetric
strain v measurement in these preliminary results. It was necessary to obtain this strain
from alternative recorded values (change in sample height H and the volume changes
of inner Vi and outer V0 chambers). The measurement of the final water content and
the initial and final dry densities of the sample allowed better assessing void ratio
changes volumetric strain measurement.
The evaluation of volumetric v, radial r, and circumferential strains needed the
knowledge of the magnitude of the changes in the internal and external radius wi = ri
and wo = ro respectively (see for instance [8] and refer to Fig. 4). These changes can
be evaluated from inner and outer chamber volume changes, which were measured by
automatic pressure/volume controllers, together with height changes that were
registered with stepper motor counts. Outer chamber volume changes were corrected to
take into account the deformability of the cell with pressure changes.
ri 2 H o Vi ro 2 H o Vi Vo

wi ri ri ; wo ro ro (2)
H o H
H o H
5. Collapse results at different intermediate principal stresses
The paths in terms of principal strains 1, 2 and 3 followed during the shearing stages
are shown in Figure 5. It should be remarked that values of 2 changed from negative
(expansion) for the case of b = 0, to positive (compression) for b = 0.8. This behaviour
is equivalent to the one observed by [1] in tests performed using a true triaxial
apparatus. During the shearing stage, volumetric strains attained values around 2.0%.
The maximum volumetric strain during this shearing stage was measured at b = 0.5.
250

b = 0.0 b = 0.5 b = 0.8
Deviatoric Stress, q [kPa]
200
150
100
1
50 2
3
0
-1 0 1 -2 -1 0 1 2 -1 -0.5 0 0.5 1
Principal Strain, i [%] Principal Strain, i [kPa] Principal Strain, i [kPa]
Figure 5. Principal strain paths followed during the shearing stages for different values of b.
Figure 6 shows the time evolution of the volumetric strain v (associated with
collapse) for different values of b during the saturation stage at constant q . 200 kPa
and p . 200 kPa. Important collapse strains were recorded along the three soaking tests.
Final v values around 20%, 9.6% and 8.4% were measured at the end of these soaking
stages for values of b = 0, b = 0.5 and b = 0.8, respectively. A picture of the final state
after soaking at b = 0.5 is also included in the figure.
Figure 6. Time evolution of volumetric strains during soaking at different b. Picture of sample after test.
6. Summary and concluding remarks
A mixture of 30% sand and 70% clayey silt by dry mass has been prepared statically
compacted at low dry density (1.50 Mg/m3) and low water content (4%) to induce high
collapsibility on suction reduction. The filling void ratio of the clayey silt is large
enough to present relatively high water permeability and to display suction values that
are measurable by tensiometer readings. A previous laboratory program was also
carried out under oedometer conditions to explore the stress level associated with the
maximum collapse zone of the mixture. With this information, a laboratory test
program was started aimed at studying the influence of intermediate principal stress on
the collapse behaviour of the mixture at an equivalent mean stress to that of the
maximum collapse under oedometer conditions. A conventional hollow cylinder device
can be a suitable equipment to perform these tests, if changes in inner and outer
diameters of the specimen are correctly evaluated from volume changes of inner and
outer chambers and by taking into account the initial and final dry densities of the
material. Water retention results will be useful to estimate suction changes during the
soaking process, since water content changes were measured along the hydraulic
process with an automatic pressure/volume controller (in this preliminary stage
tensiometers were not used in the hollow cylinder apparatus).
Preliminary results on the hollow cylinder apparatus on soaking at constant mean
and deviatoric stress (p . 200 kPa and q . 200 kPa) have shown to be influenced by the
intermediate principal stress. Test results indicated that collapse was larger when the
intermediate principal stress was equal to the minor one (b = 0), as in conventional axi-
symmetric triaxial compression conditions. A systematic decrease of collapse strain
was detected as the intermediate principal stress tended to the major principal one
(increasing b). Further research is currently being undertaken with an improved cell.
Acknowledgements
The second author acknowledges the financial support of Capes Scholarship Proc.
Bex13299/13-1 (Brazil).
References
[1] L. R. Hoyos, Experimental and computational modeling of unsaturated soil behavior under true triaxial
stress state. PhD Thesis, Georgia Institute of Technology, Atlanta, 1998.
[2] H. Matsuoka, D. Sun, A. Kogane, N. Fukuzawa and W. Ichihara, Stress-strain behaviour of unsaturated
soil in true triaxial test, Canadian Geotechnical Journal 39 (2002), 608-619.
[3] H. Toyota, K. Nakamura, N. Sakai and T. Nakamura, Effect of the stress history due to unsaturation and
drainage condition on shear properties of unsaturated cohesive soil, Soils and Foundations 41(1)
(2001), 13-24.
[4] H. Toyota, K. Nakamura and W. Sramoon, Failure criterion of unsaturated soil considering tensile stress
under three-dimensional stress conditions, Soils and Foundations 44 (5) (2004), 1-14.
[5] F. Z. Zerfa and B. Loret, Coupled dynamic elasticplastic analysis of earth structures, Soil Dynamics and
Earthquake Engineering 23 (2003), 435454.
[6] E. Romero, A microstructural insight into compacted clayey soils and their hydraulic properties,
Engineering Geology 165 (2013), 3-19.
[7] A. W. Bishop, The strength of soils as engineering materials, Gotechnique 16 (2) (1966), 91-128.
[8] D. W. Hight, A. Gens and M. J. Symes, The development of a new hollow cylinder apparatus for
investigating the effects of principal stress rotation in soils, Gotechnique 33 (4) (1983), 355-383.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-470
Travel time measurements of shear waves

in Bo Bo sand using bender elements
Juan AYALAa, Felipe VILLALOBOS b,1and Giovanny ALVARADOc
a
Golder Associates Chile, Santiago, Chile
b
Laboratory of Geomaterials, Catholic University of Concepcin, Chile
c
AECOM Perth, Australia
Abstract. The design and set up of a system to measure the travel time of shear waves
in Bo Bo sand samples using bender elements, is described. An oedometer device
was adapted to install bender elements. Measurements were carried out for sand
samples with different relative density, pressure/deformation, diameter/height,
dry/saturated and varying the frequency, amplitude and type of the triggered electric
signals. Repetitiveness and calibration were evaluated. Travel times are evaluated
using criteria in the time and frequency domain. From the theory of elasticity, results
are interpreted as shear wave velocity and elastic shear modulus as a function of
vertical effective stress, void ratio and vertical deformation. The shear wave velocity
and shear modulus increased with relative density and effective vertical stress as found
previously by other authors. Finally, well known formulae for estimating the elastic
shear modulus are used to compare with the experimental results. The estimations
were good for effective vertical stresses around 100 kPa, but underestimation and
overestimation of the shear modulus occurred for stresses below and above that value,
respectively. Soil stiffness during unloading/reloading cycles was underestimated with
the expressions used for loading.
Keywords. bender element, travel time, shear wave velocity, elastic shear modulus,
vertical deformation
1. Introduction
The use of bender elements BE for the assessment of shear wave velocities Vs and
elastic shear moduli for small deformations G0 has been more often adopted owing
meanly to the non disturbing nature and low cost of the testing. BE are piezoelectric
elements which were first used to measure compression waves velocities in sand and
glass crystals [1] and then to measure shear waves velocities in sedimentary soil
deposits [2]. BEs have been mostly adapted to triaxial equipments [3-9]. BEs can be
also adapted to other laboratory devices such as direct shear box, resonant column and
oedometer [8]. Despite the simple principle of measuring the travel time through a
known soil sample distance to in this form determine the shear wave velocity Vs, there
are issues related to for instance, the interpretation of arrival signals and appropriate
frequency ranges. Yamashita et al. [8] reported differences in G0 in Toyoura sand
determined by 23 different laboratories in the world using BE under similar conditions
of testing. They indicate the main reasons for some of the most important differences
and propose a standard testing methodology. Yang and Gu [10] reported similarities
when comparing G0 results from BE and resonant columns RC using glass beads,
which agree with previous results in soils [2]. In addition, using three uniformly graded
glass bead test specimens with mean particle size d50 equal to 0.195, 0920 and 1.750
1
J. Ayala et al. / Travel Time Measurements of Shear Waves in Bo Bo Sand 471
mm, Yang and Gu [10] found that G0 is almost independent of size particles, but BE
output signals can be significantly affected by d50. For that reason, proper interpretation
of signals is vital for obtaining adequate G0 values. Moreover, Alvarado and Coop [9]
have systematically used transfer functions to analyse input/output signals, concluding
that the response of the system (sample, BE, triaxial cell) is linear. This means that
output signals should be independent of the input signal, which makes possible the use
of elasticity for the determination of G0.
A particular motivation for this research has been the study of Bo Bo sand
dynamic properties. Alluvial deposits of this sand form part of the soil foundation
around the city of Concepcin in Chile, where a Mw = 8.8 earthquake took place in
2010. Since then, the shear wave velocity in the first 30 m, Vs30, has been compulsory
required for the seismic building code. As a consequence, geophysical in situ testing
has been profusely employed for obtaining profiles of Vs in 30 m depth. Results from
this experimental study aim to offer insight in the analyses of geophysical results not
only for soil seismic classification, but also for other studies such as seismic site
response.
In this work an oedometer apparatus is used to perform compressibility tests in
dry and saturated Bo Bo sand samples. BE were mounted at the top and bottom of a
specially designed cell to measure travel times of signals sent from one BE to the other.
The BE system and soil used are first described. Then, interpretation of travel times is
analysed for instance in terms of time and frequency domain, frequency range, input
wave shape, voltage and gain. Finally, results are transformed into Vs and G0 for
engineering interpretation in terms of relative density and vertical stress.
2. Measurement system
The bender element measurement system is made up of five parts as shown in Figure 1.
The soil sample is inside an adapted perspex oedometer cell with an internal diameter
of 44.8 mm and 88 mm height. A transmitter BE is installed on top of the soil sample
and it is connected to the function generator and the oscilloscope. The function
generator allows the emission of signals with different frequencies (1 Hz to 25 MHz),
voltages and signal types (sinusoidal, square, ramp, burst); the maximum signal
amplitude is 10 V or 20 Vp-p peak-to-peak.
The receiver BE is at the bottom of the sample and it is connected to an amplifier
which deliver signals to an oscilloscope connected to a computer through an USB port.
Since the receiver BE can detect shear waves propagated in the soil, in this case by the
transmitter BE, this movement is converted into electric signals. A Tektronix
oscilloscope was used to record the onset time in the transmitter BE and the arrival
time in the receiver BE, and because the distance between BE tips is known, 56 mm
(initially and updated after loading), it is possible to determine the signal travel velocity.
Moreover, the system has a bandwidth of 100 MHz, a sample rate of 1 G samples/s and
its data acquisition system allows the identification of maxima and average of 4, 16, 64
and 128 samples. The use of 128 samples reduces significantly the noise of the signals,
which is very important considering the low voltage in the receiver (< 1 mV). The
software OpenChoice Desktop (downloadable from Tektronix webpage www.tek.com)
was used to transfer 2500 data per channel from the oscilloscope to the Windows
clipboard in a laptop computer and later on to an Excel macro spreadsheet. In this
spreadsheet information of the sample tested as well as the type of test is stored. The
472 J. Ayala et al. / Travel Time Measurements of Shear Waves in Bo Bo Sand
data was also analysed in a Matlab graphical user interface to determine the travel time
from the transmitter and receiver signals. An in house power amplifier was used since
the received signals were too low for the oscilloscope voltage range.
Figure 1. a) Bender element measurement system and b) BE adaptor into the oedometer,
1) double protection cable, 2) welding, 3) transparent poxipol 10 min glue, 4) piece of copper tube and 5) 16
mm BE in contact with soil)
Dimensions of BE were 31.8 mm long, 12.7 mm wide and 0.66 mm thick, a weight of
2.1 g and maximum voltage of 500 V and 250 V for in series and parallel connexions,
respectively. Only half of the BE was in contact with the soil sample (16 mm) and a
protected cable inside a mesh and plastic tube was welded to the BE. To protect the BE
against the friction of the soil and short circuit in saturated tests, transparent nail polish
was applied to the BE surface to isolate electrically the transducers. The BE were
inserted in poxipol 10 min glue inside a piece of tube of 24 mm diameter, 12 mm long
and 2 mm thick, which allowed the BE attachment into the oedometer as shown in
Figure 1b. A ground cable was also welded to the BE.
2.1. Soil sample

The soil tested corresponds to Bo Bo sands from alluvial deposits of the city of
Concepcin, Chile. This clean sand has a uniform grain size distribution which
classifies as SP according to the Unified Soil Classification System USCS. This sand
has been transported by the Bo Bo River from past eruptions of the Antuco volcano
200 km upstream in the Andes. It was deposited in a deltaic form around the mouth of
the Bo Bo River. It is composed mainly of basalt and feldspar and has a grey colour
although there are also brown and white particles. Representative size grains d10, d50
and dmax of the samples used in this research are 0.4, 0.8 and 2.2 mm, respectively. For
the samples tested, specific gravity Gs is 2.84, maximum and minimum void ratios are
1.0 and 0.65, respectively. Samples were prepared using the pluviation method, which
consists of raining sand grains through a funnel with an opening of 2.2 mm diameter
from a height of 10 cm. Thus, values of relative density RD of 61, 71 and 85%, were
obtained. Dry pluviation was selected because relatively homogeneous samples can be
obtained.
3. Analysis of the arrival time

Figure 2 shows two criteria to estimate the arrival time in the time domain, where td
represents the travel time of the maximum amplitude of the first arrival wave whereas
td max corresponds to the travel time of the maximum amplitude wave. The latter can be
also associated with the method of cross correlation in the frequency domain [4,5]. In
this work the peak-to-peak criteria in the time domain td max has been used [11-13].
1.0 Transmitter
t d max
0.8 td Receiver
0.6
0.4
Voltage, mV
0.2
0.0
-0.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-0.4 Time, ms
-0.6
-0.8
-1.0
Figure 2. Selection of arrival time

A comparison between the time domain method td max and the frequency domain method
tcc (cross correlation), as a function of the vertical effective stress v, shown in Figure
3a, indicates that they have practically the same values (for the conditions tested: 20 V,
RD = 85%, dry sand, 9 kHz, sine wave). This similar travel time represents a linear non
dispersive system, i.e., the shear wave velocity Vs is independent of the frequency.
0.6 td max td, 9 kHz td, 15 kHz
0.6
tcc tcc, 9 kHz tcc, 15 kHz
0.5 0.5
Travel time, ms
Travel time, ms
0.4 0.4
0.3 0.3
0.2 0.2
0 100 200 300 400 0 100 200 300 400
Effective vertical stress 'v, kPa Effective vertical stress 'v, kPa
(a) (b)
Figure 3. a) Travel time for the time domain method td max compared with the cross correlation method tcc and
b) travel times versus vertical effective stress and frequency for time td and frequency domain methods tcc
Conversely, the comparison between the time domain method td with the cross
correlation method tcc, results in clearly different travel times as shown in Figure 3b for
the frequencies of 9 and 15 kHz. As expected tcc travel times are longer than td for the
frequency tested under the same conditions (20 V, RD = 85%, dry sand, sine wave).
This difference is because the first arrival wave does not have the largest amplitude.
For v = 25 kPa and for the largest amplitude methods, Figures 3a and 3b show that
the travel time value does not follow the trend. This is caused by spikes that appear in
waves before the largest amplitude, which the Matlab program is not able to recognise;
a manual correction of tdmax is possible to follow the trend. The longer travel times from
the largest amplitude methods represent a condition of non-linear and dispersive
systems (media where the waves travel), which is in contradiction with what it was
pointed out above.
The effect of the frequency is shown in Figure 4a for frequency values from 1
to 15 kHz and for v from 25 to 400 kPa. It can be clearly observed that for
frequencies below 9 kHz travel times do not stabilise around a single value, but they
fluctuate. For these lower frequencies wavelengths are too long for a wave to travel
completely the distance through the soil between the BE tips. For that reason [8]
recommend at least two wavelengths to reduce this near field effect. Since the distance
between BE tips is around 55 mm and assuming a shear wave velocity of 250 m/s
results in a frequency f = 4.5 kHz, a minimum f 9 kHz is necessary to obtain stable
travel time values. As also observed in Figures 3 and 4, the travel time diminishes with
the increase of v, since the soil settles becoming denser with the increase of v.
0.45 0.45
25 kPa 50 kPa 100 kPa
200 kPa 400 kPa sine ramp square
0.40 0.40
0.35 0.35
Travel time, ms
Travel time, ms
0.30 0.30
0.25 0.25
0.20 0.20
0.15 0.15
1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15
Frequency, kHz
Frequency, kHz
(a) (b)
Figure 4. a) Travel times versus frequency and vertical effective stress for sine waves and b) travel time
versus frequency for different wave shapes (for v = 400 kPa)
The effect of the wave shape on the travel time was also investigated in terms of the
frequency from 1 to 15 kHz and for v = 400 kPa. It can be observed in Figure 4b that
the travel time stabilises for frequencies higher than 9 kHz and the same occur for ramp
waves. The travel time does not show much variation for square waves and it is slightly
higher than for the other wave shapes because the maximum in the square wave is
taken almost at time zero as shown in Figure 5. Sine waves have been used in the
majority of the experiments as a trigger signal because the receiving waves are also
sinusoidal.
0.05
transmitter-sine receiver-sine
Voltage, mV
0
0 0.1 0.2 0.3 0.4 0.5
Time, ms
-0.05
0.05
transmitter-ramp received-ramp
Voltage, mV
0
0 0.1 0.2 0.3 0.4 0.5
-0.05 Time, ms
0.05 transmitter-square receiver-square

Voltage, mV
0
0 0.1 0.2 0.3 0.4 0.5
Time, ms
-0.05
Figure 5. Sine, ramp and square waves showing maximum point as initial time
3.1 Gain
The ratio between the power measured in the receiver and the transmitted is referred to
as gain. The evaluation of the gain can assist in identifying ranges of better signal
quality, i.e. clearer signals less affected by noise. Figure 6a shows the gain variation
with frequency and voltage, where the gain increases with the frequency in dry samples
with RD = 85% and v = 400 kPa. It can be observed that the gain variation with the
frequency is independent of the voltage.
The gain was also evaluated in terms of vertical effective stress as well as in
terms of frequency under the same voltage of 20 V. Figure 6b shows that the gain has
the tendency to increase with frequency, but it does not occur steadily. In fact, it is
possible to observe peak values for frequencies 9 kHz (25 kPa), 13 (50, 100), 11 (200)
and 15 (400). The complexity of the gain variation with frequency includes the
combination of increasing soil packing (density) as v increases in conjunction with
boundary effects. For each set of points representing a constant v local maximum and
minimum points can be identified. These gain values represent resonant and anti-
resonant frequencies. The different frequency found for maximum gain values reveals
that the system changes the first mode of vibration, which may imply interferences of
the sample and BE vibration with the perspex oedometer cell [9].
5.5 5.5
25 kPa 50 kPa
5.0 5.0
Gain receiver/transmitter (10 )

Gain receiver/transmitter (10 )
6
6
100 kPa 200 kPa

4.5 4.5 400 kPa
4.0 4.0
3.5 3.5
3.0 3.0
2.5 2.5
2.0 2.0
5 Vp-p 10 Vp-p
1.5 1.5
15 Vp-p 20 Vp-p 1.0
1.0
0.5 0.5
5 10 15 20 1 3 5 7 9 11 13 15
Frequency, kHz Frequency, kHz
(a) (b)
Figure 6. Gain versus frequency as a function of a) and voltage and b) vertical effective stress
4. Results interpreted as Vs and G0

Determination of the shear wave velocity Vs has been included as a compulsory
parameter for medium to important projects for the seismic classification of soils in
Chile after the 2010 earthquake. For that reason there has been an increasing interest in
studying Vs, but mainly from in situ geophysical measurements using active and
passive techniques. However, these techniques need to be verified since the results
depend on assumptions such as statistical averages, soil properties (density, Poisson
ratio), layer thickness and depth, among others. The better controlled conditions in the
laboratory reduce the uncertainties of Vs measurements. The results presented
previously will be interpreted in terms of Vs and G0 assuming that the medium is non
dispersive, hence the expression from the elasticity G0 = Vs2 can be used.
Figure 7a shows the increase of Vs with the vertical effective stress independently
of whether the sample is dry or saturated or whether the frequency is 11 or 15 kHz. It
can be observed that saturated samples show values of Vs slightly lower than for dry
samples. Moreover, the Vs increase rate is around 60 to 70 m/s in the first 200 kPa and
from 200 kPa to 400 kPa that rate reduces to around 20 to 30 m/s.
The effect of the initial relative density RD on Vs is shown in Figure 7b, where it is
clear the increase of Vs with RD, however, this increase is not that significant because
the RD tested are not too low. Loose samples are difficult to prepare because they
densify easily with handling, setting up and loading. Further attempts with adequate
techniques are required to test looser samples.
290 290
270 270
Shear wave velocity V s , m/s
S hear wav e v eloc ity V s , m /s

250 250
230 230
210
210
190
190
dry, 11 kHz dry, 15 kHz
170
sat., 11 kHz sat., 15 kHz 170
RD = 61% RD = 71% RD = 85%
150
150
0 100 200 300 400
0 100 200 300 400
Effective vertical stress 'v , kPa Effective vertical stress 'v , kPa
(a) (b)
Figure 7. Shear wave velocity versus vertical effective stress for a) dry and saturated samples with RD =
85% and b) for dry samples with RD = 61, 71 and 85%
Figure 8a shows the increase of G0 with v, confirming what has been found before by
other authors [8,10]. In Figure 8a are also shown in dotted lines estimations of G0 using
the expressions for sands with round and angular grains [14].
6.908(2.17 e) 2
G0 = '0 round grains (1)
1 e
3.23( 2.97 e)2
G0 = '0 angular grains (2)
1 e
Where e is the void ratio and 0 is the effective mean stress which was calculated as
0 = v/3(3-2sin*), for Bo Bo sand * varies from 40 for low stresses to 36 for
high stresses.
160
130
150 a) round
b)
140 angular 120
130 110
120
100
110
G 0, MPa
G 0, M P a
100 90
90
80
80
70 70
60 60
50
RD = 61% 71% 85% 50 RD = 61% 71% 85%
40
30 40
0 100 200 300 400 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Effective vertical stress 'v , kPa Vertical strain v , %
Figure 8. Elastic shear modulus versus a) effective vertical stress, showing estimations in dotted lines and b)
vertical stress
It can be observed that for the data for RD = 71% the trend of the estimations
underestimates at low stresses and overestimates G0 at high stresses, having almost a
perfect match for stresses around 100 kPa. Therefore, attention should be paid when
using expressions (1) and (2), especially for stresses which are far from 100 kPa.
Moreover, Figure 8b shows the increase of G0 with the vertical stress v, which is the
result of the sample compaction owing to the increased effective vertical stress.
5. Conclusions
A series of BE measurements have been carried out during compression tests in an
adapted oedometer using samples of Bo Bo sand. A range of frequency for the input
wave was found to start from 9 kHz to avoid inconsistency of the travel times measured.
This confirmed previous results suggesting a minimum of two wave lengths between
benders tips. Important differences were found in the cross correlation method when
the first wave had not the maximum amplitude in comparison with the time domain
method. Sine waves were used although other wave shapes were checked such as the
square, however this type of wave has an important effect on the results because its
maximum starts from time zero. The wave amplitude (input voltage) is less important
than the frequency according to the gain (ratio between the received and transmitted
voltages). The gain is also influenced by the effective vertical stress. Elasticity was
assumed to calculate shear wave velocities Vs from the measured travel times and
sample density. It was confirmed that Vs increases with relative density and effective
vertical stress. Finally, the well known Hardin and Black [14] formulae for the elastic
shear modulus were used to compare with the experimental results. The estimations
were good for effective vertical stresses around 100 kPa, below that value there is
underestimation and beyond that value there is overestimation of G0. Further research
is needed to study loose samples of sand and the effect on G0 for unloading/reloading
cycles.
References
[1] F.V. Lawrence, Propagation of ultrasonic waves through sand, Report MIT, Boston, 1963
[2] D.J. Shirley, L.D. Hampton, Shear-wave measurements in laboratory sediments, The Journal of the
Acoustical Society of America 63, No.2 (1978), 607-613
[3] C.R. Bates, Dynamic soil property measurements during triaxial testing, Gotechnique 39, No. 4 (1989),
721-726
[4] G. Viggiani, J.H. Atkinson, Interpretation of bender elements tests, Gotechnique 45, No. 1 (1995), 149
154
[5] G. Viggiani, J.H. Atkinson, Stiffness of fine-grained soil at very small strains, Gotechnique 45, No. 2
(1995), 249-265
[6] E.G.M. Brignoli, M. Gotti, K.H. Stokoe, Measurement of shear waves in laboratory specimens by means
of piezoelectric transducers. Geotechnical Testing Journal 19, No. 4 (1996), 384-397
[7] D.S. Pennington, D.F.T. Nash, M.L. Lings, Horizontally mounted bender elements for measuring
anisotropic shear moduli in triaxial clay specimens, Geotechnical Testing Journal 24, No. 2 (2001),
133-144
[8] S. Yamashita, T. Kawaguchi, Y. Nakata, T. Mikami, T. Fujiwara, S. Shibuya, Interpretation of
international parallel test on the measurement of Gmax using bender elements, Soils and Foundations 49,
No. 4 (2009), 631650
[9] G. Alvarado, M.R. Coop, On the performance of bender elements in triaxial tests, Gotechnique 62, No. 1
(2012), 117
[10] J. Yang, X.Q. Gu, Shear stiffness of granular material at small strains: does it depend on grain size?,
Gotechnique 63, No2 (2013), 165-179
[11] V. Fioravante, R. Capoferri, On the use of multi-directional piezoelectric transducers in triaxial testing.
Geotechnical Testing Journal 24, No. 3 (2001), 243-255
[12] J.-S. Lee, J.C. Santamarina, Bender elements: performance and signal interpretation. Journal of
Geotechnical and Geoenvironmental Engineering 131, No.9, (2005), 1063-1070
[13] A. Viana da Fonseca, C. Ferreira, M. Fahey, A framework interpreting bender element tests, combining
time-domain and frequency-domain methods. Geotechnical Testing Journal 32, No. 2, (2009), 91-107
[14] B.O. Hardin, W.L. Black, Vibration modulus of normally consolidated clays, Journal of the Soil
Mechanics and Foundation Division 94, No. SM2 (1968), 353-369
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-478
3D Experimental Characterization of
Particle Rotation and Local Dilatancy in
Angular Sand
Khalid Alshiblia,1, Andrew M. Druckreya, Maha Jarrara, and Riyadh I Al-Raoush b
a
University of Tennessee-Knoxville, Tennessee, USA
b
Qatar University, Doha, Qatar
Abstract. Particle-to-particle interaction plays a major role in characterizing shear

strength and deformation characteristics of granular materials. This paper presents
an experimental measurement of particle kinematic behavior for very angular sand
tested under axisymmetric triaxial loading condition. Synchrotron micro-computed
tomography (SMT) technique was used to acquire 3D images of the specimen at
0%, 4%, 8%, and 12% axial strain levels. The results of the analysis showed that
some particles rotated more than 20o from the loading direction with a higher
rotation about the short diameter of the particle than its long diameter. A definition
of local strain and dilatancy based on particle displacement field is adopted from
[1] and applied in this paper. There is a high variation in local dilatancy
distribution where some particle groups exhibited high dilatancy angles while
others showed negligible dilation or contractive behavior.
Keywords. Shear, sand, synchrotron computed tomography, kinematic behavior
1. Introduction
Granular materials consist of discrete particles that interact with each other when the
mass of the material is sheared/compressed. The shearing resistance of granular
materials is derived from particle-to-particle interactions that include interlocking,
translation, rotation, and fracture of particles if the material is tested under high
compressive stresses. Most experimental studies on sands focused on investigating the
behavior of laboratory-size specimens and measuring stresses and strains at the
boundaries of the specimen and use the measurements to characterize the constitutive
behavior of sand. Peak and/or critical state friction angles and a dilatancy angle are
commonly used as parameters to measure the shear strength of uncemented sands. The
dilatancy angle is calculated using on the highest rate of dilation (volume increase) of
the specimen from the global volume change measurement of the specimen.
The literature is rich with studies aimed at measuring particle kinematic behavior
and fabric evolution using discrete element method (DEM). For example, [2] used
DEM to emphasis the importance of particle rotation during shearing of granular
materials and found that particles rotation plays a major role in controlling dilatancy
and failure of granular materials. [3] used 2D DEM as a tool to investigate dilative
1
K. Alshibli et al. / 3D Experimental Characterization 479
behavior of particles within force chains and their relationship to buckling of force
chains. [4] used 2D DEM to investigate the influence of changes in topology (flips) and
elastic relaxation on dilatancy and critical state of granular materials. [5] compared 3D
DEM prediction of particle kinematic behavior with experimental measurements and
found that DEM method over-predicted particle translation and rotation in all directions
when compared to experimental measurements.
Few studies investigated sand particle-to-particle interaction in 3D experimentally.
For example, [1] used computed tomography (CT) to track the translation and rotation
of 6.5 mm smooth spherical plastic particles of a triaxial specimen and proposed a
framework to measure local strain distribution based on particle kinematic properties.
They also measured particle rotation angles in 3D. [6] used SMT to investigate the
behavior of a specimen composed of F-75 sand and tracked the fabric evolution and
kinematic behavior of few sand particles within and outside the shear band. [7,8] used
digital volume correlation technique to track the kinematic behavior of sand particles
within triaxial specimens that were scanned using CT technique and quantified the
spatial distribution of particle translation and rotation. Two sands (one with angular
particles and the other with rounded particles) were used in [7,8] experiments.
The objective of this paper is to investigate particle translation and rotation in 3D
for angular sand and to present a quantification of local dilatancy angle within a triaxial
specimen. [1] investigated kinematic behavior of spherical beads with relatively
smooth surface. The literature lacks experimental measurements of particle kinematic
behavior for angular sand in 3D. Therefore, this paper focuses on capturing particle-to-
particle interaction at the micro-level with emphasis on particle rotation (due to page
limitation).
2. Experiment
SMT is a non-destructive high-resolution imaging technique that can produce 3D

images of the specimen using x-ray transmission. The specimen is placed between the
x-ray source and the detector, and a high-energy monochromatic x-ray beam produced
by synchrotron sources is attenuated as it travels through the specimen. A scintillator
converts the transmitted beam to visible light, which is projected onto a detector and
recorded by a high-resolution camera system. The captured projection represents the
attenuation data of the specimen acquired at a certain projection angle. To generate a
3D map of the specimen, a set of 2D images needs to be recorded by detectors while
the specimen is rotated on a high-precision stage at small rotation increments from 0
to 180. The acquired projections are processed with the help of computer code to
reconstruct the complete 3D attenuation data of the specimen. SMT enables acquiring
images with enhanced spatial resolution, which makes it favorable among the 3D
imaging techniques and is sufficient for quantitative analysis of images of granular
media under various loading conditions.
A miniature triaxial cell was especially fabricated to conduct an in situ
axisymmetric triaxial experiment. The triaxial cell has capabilities similar to the
conventional triaxial cell. A stepper motor was used to apply the axial load at a
constant displacement rate, and a load cell located inside the test cell was used to
record the load. A data acquisition system acquired the signal from the load cell with
an interface to a computer. A very angular poorly-graded sand was used in the
experiment. It has a mean particle size of 1.6 mm, minimum and maximum void ratio
480 K. Alshibli et al. / 3D Experimental Characterization
vales of 0 .677 and 0.843, respectively and specific gravity of solids of 2.65. The
specimen is cylindrical in shape and measures 9.46 mm in diameter by 20.2 mm in
height. It had a dry bulk density of 1.45 g/cm3 and was loaded in the axial direction at a
constant axial displacement rate of 0.1 mm/min while maintaining a constant confining
pressure (o ) at 57 kPa.
The SMT scans were conducted at GeoSoilEnviroCARS (GSECARS) Sector 13 of
the Advanced Photon Source (APS), Argonne National Laboratory (ANL), Illinois,
USA. The specimen was scanned using an energy of 30 keV, which provided a spatial
resolution of 21.71 m/pixel. Compression was paused at 0%, 4%, 8%, and 12%,
nominal axial strain ( ) to acquire the SMT scans. It took approximately 1 hour to
acquire each scan.
3. Image processing and particle tracking
The processing of SMT images was initially performed using Avizo Fire software,
which uses Visilog algorithms for image processing and quantification. Raw SMT data
for each loading stage were loaded into the software according to the step size and
merged together, forming a complete scan volume (Figure 1). Next, each of the images
were segmented using the same thresholding value for all SMT scan sets to consistently
separate the solid material from the surrounding air. In the binary images, however, the
particles remain in contact with each other, making it impossible to identify individual
particle morphology. Visilog has an algorithm based on the watershed principle to
separate particles from each other and remove the small area of contact between them.
After the separation algorithm was applied, another algorithm was applied to delete any
object touching the borders of the image. Each of the particles was then individually
labeled so that each could be identified and discrete quantification of each particle
would be possible (Figure 1e). The particle labels were then expanded and then masked
with the binary image to fill in the small contact area removed during separation and
provide a labeled image with particles in contact.
The 3D images were then saved for further processing and analysis using a special
code that was developed by [9]. Particle centroids in the x, y, and z directions were
calculated based on the center of mass of the particle. Particle volume was calculated
by summing the voxels that belong to the particle. Similarly, particle surface area was
computed as the summation of exterior faces of boundary voxels belonging to the
particle. The shortest (dS), intermediate (dI), and longest (dL) diameters of the particle
that pass through the center of mass were calculated using the concept of principal
component analysis (PCA) on the particle voxels. The three orthogonal directions were
statistically calculated using singular value decomposition and the distances between
the extreme points along those directions represent the three particle lengths. The
orthogonal directions of the particles principal axes (particle diameters) as calculated
from PCA were used to define particle orientation. The direction from the global
vertical axis was determined for dS, dI, and dL.
Particle tracking was conducted on successive SMT images. It begins with
isolating a sub-volume around each of the particle centroids in the first image. A cubic
sub-volume with a size range of 1 to 3.5 mm centered on the particle centroid was
chosen. The sub-volume was searched in the second image for particles within 5% to
10% of the volume, surface area, length, and width of the particle in the first image. If
more than one particle in the second image matches the criteria for the particle in the
first image, the most closely morphologically matching particle is chosen. This process
is repeated for all successive images. After particles are matched, their displacements
and rotations can be calculated and analyzed.
Figure 1. Axial sections of SMT images of the specimen and the labeled Avizo image.
(a) At 4% axial strain (b) At 8% axial strain (c) At 12% axial strain
Figure 2. Cumulative values of orientation of particle long diameter measured from the direction of global
major principal stress (z-axis) for the experiment.
4. Particle kinematic and local dilatancy behavior
Cartesian coordinates of the centroid of the particles were calculated for the four scans.
The translation in any direction can be calculated as the difference between the location
of the particle centroid in one scan, and its location in the following scans. The
cumulative rotation of particle long diameter with respect to the direction of major
principal stress (z or vertical direction) is shown in Figure 2. Initially particles exhibit
small rotations (mostly less than 5o) between 0% and 4% axial strain with fairly
uniform spatial distribution throughout the specimen (Figure 2a). It represents the
hardening regime of the loading. Then some particles began to exhibit a higher rotation
(yellow and red particles in Figure 2b & c) as compression progressed with few
particles exhibited rotation angles higher than 20o and many particles rotated in the
range of 10o to 15o. Figure 3 displays the cumulative rotation of particle short diameter
with respect to the direction of major principal stress (z or vertical direction). It is
interesting to notice particle exhibited higher rotation about their short diameter than
their long diameters because it takes more energy to force a particle rotate about their
long diameter.
(a) At 4% axial strain (b) At 8% axial strain (c) At 12% axial strain
Figure 3. Cumulative values of orientation of particle short diameter measured from the direction of global
major principal stress (z-axis) for the experiment.
The displacement field in 3D space is described by Equations 1 through 3

assuming a non-linear displacement field [1]:
p (1)
p (2)

p (3)
where r are displacements in the x, y, and z directions, respectively; x,y,z are
the coordinates of the particles centroid. Ten coefficients are present in each of Eq. 1
to 3; therefore, the particle assembly is divided into groups of 10 particles using the
data of the first SMT scan. Boundary particles are shared with adjacent groups to
ensure continuity of strain calculations. The number of shared particles between any
two groups varies from 2 to 4 depending on the group location. A total of 30 groups
were identified. Having the coordinates and displacements of the particle group, one
can write the following system of equations:
F

F F
G

F
F
F

F F F
F

F F F H

F F F I
= (4)
F F F J

F F F

F
F
F K

F F L
F

F F F M

E
Where, x,y and z denote the coordinates of the centroid of the group and i ,j, k, .
denote the individual particles in each group. This system was solved to obtain the a
coefficients. Then, the same procedure was repeated to calculate the b and c
coefficients using the v and w displacement fields, respectively. Strains are calculated
using Lagrangian description defined by Eq. (5). They were calculated with reference
to the initial state (the first scan) by using displacement fields defined by Eq. 1 through
3 and centroid of particle group. The procedure was repeated for all particle groups in
the specimen. As deformations progress, strains were calculated for the same particle
groups that were chosen from the first scan.

= + + (5)

E
Local dilatancy angle ( is defined as:

(6)

where and are the changes in volumetric and deviatoric strains for the
general 3D strain condition since particles are not loaded under the axisymmetric
triaxial condition at the micro-level (the global loading condition on the specimen).
They are defined as:
(7)

(8)

The results of the analysis are listed in Table 1. Values of were calculated with
reference to the initial state and in incremental form. has a wide range of variation
ranging from 67.1o to -37.9o. A negative represent a contractive behavior of the
particle group (i.e. volume decrease). The highest cumulative were recorded for 0% -
4% strain level with a mean value of 9.1o and a standard deviation (SD) of 21.3o.
Table 1. Dilatancy angle for particle groups using both incremental and cumulative strains
Group # 0% to 4% 0% to 8% 0% to 12% 4% to 8% 8% to 12%
(deg.) (deg.) (deg.) (deg.) (deg.)
1 19.4 16.5 9.9 27.2 0.3
2 15.2 34.8 26.0 26.2 21.6
3 20.5 8.1 0.4 67.1 0.9
4 55.4 38.0 0.025 31.6 8.9E-04
5 0.3 0.1 1.5E-05 0.1 4.0E-06
6 0.3 9.1E-02 1.7E-02 2.0 2.1
7 22.2 2.1E-04 2.1E-04 2.7E-04 7.7E-02
8 5.6E-05 1.1E-03 8.8E-05 3.1E-02 0.0
9 23.7 27.7 20.4 -0.9 27.4
10 23.0 13.9 2.3E-02 23.5 6.4E-03
11 6.0 5.6 1.7 25.6 5.1
12 21.0 5.5E-03 0.6 2.7E-03 0.0
13 21.5 17.4 0.1 26.5 0.1
14 -37.9 -18.7 0.5 5.2 0.4
15 17.5 22.0 8.3 35.2 14.1
16 16.7 10.6 1.2E-02 18.2 1.4E-02
17 18.7 21.2 1.6E-02 22.9 0.1
18 26.3 22.8 13.6 38.8 22.8
19 0.5 2.5 4.6E-02 3.4 15.3
20 4.0E-04 4.6E-02 0.1 9.4E-03 0.0
21 2.7E-02 5.8E-03 5.7E-03 2.1E-02 0.0
22 10.1 2.9 2.7 12.4 21.9
23 4.0 2.6E-02 2.7E-02 3.4E-02 10.2
24 9.3 5.7E-03 5.3E-04 1.1E-02 0.0
25 -54.5 4.2E-04 4.3E-04 4.5E-04 3.2E-02
26 16.5 18.2 20.0 0.8 34.1
27 15.8 11.8 2.4 0.8 2.2
28 1.0E-02 1.1E-06 9.3E-06 1.6E-04 4.3E-02
29 -36.3 5.3 2.0 31.3 2.1
30 37.9 1.5E-02 5.1E-04 1.6E-02 1.1E-03
Mean 9.1 8.7 3.6 13.3 6.0
As loading progressed, the values of decreased along with their SD values.

values for 0% - 12% and 8%-12% strains and their SD are the lowest. It appears that
the specimen reached a critical state before 8% based on stress-strain response (not
shown in this paper due to space limitation). Global volume change measurements
were not acquired for the specimen because water was not used as confining medium;
therefore, a global dilatancy angle is not calculated for the specimen. Very often some
particles groups (e.g., groups 4, 14, 29, etc.) show a drastic change from one strain
increment to another which is caused by neighbor particle interactions that very often
force a collapse of a large void between particles which results in a generation of a high
volume change in another particle group. Some particle groups did not dilate whereas
others have dilative behavior. The results of the analysis present an interesting insight
into particle-to-particle interaction at the micro-level where very angular sand particles
exhibit a heterogeneous spatial distribution of particle translation, rotation and
dilatancy angle. These experimental measurements are valuable to calibrate continuum
and discrete models aimed at capturing the constitutive behavior and strain localization
in granular materials.
5. Conclusion
The paper investigated rotation behavior of very angular sand focusing on particle-to-
particle rotation for a specimen loaded under axisymmetric triaxial loading condition.
Particles exhibited higher rotations about their short diameter when compared to their
long diameter with rotation angles higher than 20o for some particles. Some particles
continued to show little or no rotation is spite of the high strain suggesting a high
degree of interlocking between these particles. A definition of local dilatancy is
revisited and applied to very angular particles. has a wide range of variation ranging
from -37.9o to 67.1o. Some particle groups did not dilate whereas others have dilative
behavior. The results of this paper can serve as key experimental measurements to
calibrate constitutive models to describe shear strength and strain localization in sand.
6. Acknowledgments
This material is based on work supported by the National Science Foundation

under Grant No. CMMI-1266230 and the Office of Naval Research (ONR) grant No.
N00014-11-1-0691.. The SMT images presented in this paper were collected using the
x-ray Operations and Research Beamline Station 13-BMD at Argonne Photon Source
(APS), Argonne National Laboratory. We thank Dr. Mark Rivers of APS for help in
performing the SMT scans. We also acknowledge the support of GeoSoilEnviroCARS
(Sector 13), which is supported by the National Science Foundation, Earth Sciences
(EAR-1128799), and the US Department of Energy (DOE), Geosciences (DE-FG02-
94ER14466). Use of the Advanced Photon Source, an Office of Science User Facility
operated for the DOE Office of Science by Argonne National Laboratory, was
supported by DOE under Contract No. DE-AC02-06CH11357.
References
[1] K. A. Alshibli and B. Alramahi, Microscopic Evaluation of Strain Distribution in Granular Materials
during Shear, ASCE, Journal of Geotechnical & Geoenvironmental Engineering, 132 (2006), 80-91.
[2] M. Oda, K. Iwashita, and T. Kakiuchi, Importance of Particle Rotation in the Mechanics of Granular
Materials, Powders & Grains 97, Behringer and Jenkins (eds), 207-210, 1997.
[3] A. Tordesillas, J. Shi, and T. Tshaikiwsky, StressDilatancy and Force Chain Evolution, International
Journal for Numerical and Analytical Methods in Geomechanics, 25(2011), 264-292.
[4] S. Mesarovic, J.M. Padbidri, and B. Muhunthan, B., Micromechanics of Dilatancy and Critical State in
Granular Materials, Gotechnique Letters, 2(2012), 61-66.
[5] M. B. Cil and K. A. Alshibli, K. A., 3D Analysis of Particle Interaction of Granular Materials in Triaxial
Testing Using DEM with Flexible Membrane Boundary, Acta Geotechnica, 9(2014), 287-298, DOI:
10.1007/s11440-013-0273-0
[6] A. Hasan and K. A. Alshibli, Three Dimensional Fabric Evolution of Sheared Sand, Granular Matter,
14(2012), 469-482.
[7] E. And, S. A. Hall, G. Viggiani, J. Desrues, and P. Bsuelle, Grain-scale experimental investigation of
localised deformation in sand: a discrete particle tracking approach. Acta Geotechnica, 7(2012), 1-13
[8] E. And, S. A. Hall, G. Viggiani, J. Desrues, and P. Bsuelle, Experimental micromechanics: grain-scale
observation of sand deformation. Gotechnique Letters, 2(2002), 107-112.
[9] R. Al-Raoush, Microstructure characterization of granular materials. Physica A, A377 (2007), 545-558.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-486
1D creep and delayed rebound during

unloading and reloading of clay and its
model simulation
Shohei KAWABEa,1 and Fumio TATSUOKA b
a
National Institute for Rural Engineering, Ibaraki, Japan
b
Tokyo University of Science, Chiba, Japan
Abstract. Most of the previous studies on the creep behaviour of soft clay focused
on the one during otherwise primary loading under normally consolidated
conditions associated with fill construction, pumping of ground water and so on.
Despite strong geotechnical engineering needs, the study on the creep behaiour
under general loading conditions including not only primary loading but also
unloading and reloading has been very limited. In this study, a series of one-
dimensional compression tests were performed along complicated loading
histories and the creep strain during general loading history was studied. The test
results are simulated successfully by an elasto-viscoplatic non-linear three-
component model.
Keywords. Soft clay, creep, one-dimensional compression, three-component

model, unload/reload cycles, viscous properties
1. Introduction
Most of the previous studies on the creep behaviour of soft clay focus on the one
during otherwise primary loading under normally consolidated (NC) conditions,
associated with the design of fill construction, pumping of ground water and so on.
Despite strong geotechnical engineering needs, the study on long-term creep
deformation under general loading conditions of not only primary loading (PL) but also
unloading (UL) and reloading (RL) is very limited (e.g., Acosta-Martnez et al. 2005;
Kawabe et al., 2009). Kawabe et al. (2011) performed a series of one-dimensional (1D)
compression tests, including multiple UL/RL cycles with many sustained loading (SL)
stages, on reconstituted Fujinomori clay. The test method is explained in the next
section. Fig. 1 summarises the relationship between the creep strain observed at SL
stage for three hours and the ratio of the stress at the start of UL or RL at an axial strain
rate equal to 0.005 %/min, UL or RL, to the stress at SL stage, SL. The creep strain
rates during otherwise UR and RL are smaller than those during otherwise primary
loading becoming very small or even negative at SL stages after larger unloading. It
may be seen from Fig. 1 that positive and negative creep strains are a unique function
of, respectively, the loading stress ratio, UL/ SL (i.e., over-consolidation ratio, OCR)
and RL/ SL, not the stress differences, |UL SL| and |RL SL|. This relation is also a
1
Shohei KAWABE, [email protected], National Institute for Rural Engineering, Ibaraki, Japan.
S. Kawabe and F. Tatsuoka / 1D Creep and Delayed Rebound During Unloading and Reloading 487
function of other factors. For example, the creep strain increases with an increase in the
initial strain rate (i.e., the strain rate immediately before the start of concerned SL
stage) and the elapsed time during the SL stage. In this study, similar tests as this
previous study were performed along more complicated loading histories. The drained
creep behaviour observed in these and previously performed tests were analysed and
simulated by an elasto-viscoplatic non-linear three-component model (Fig. 2). In
particular, it is shown that the trend seen in Fig. 1 can be theoretically predicted.
The creep deformation of clay is due to the viscous properties as well as delayed
dissipation of excess pore pressure (e.g., Adachi et al. 1996). Most of the previous
models of the viscous properties of clay under one-dimensional (1D) NC is of Isotach
type: i.e., the current effective stress under the loading condition is a unique function of
instantaneous irreversible strain and its rate irrespectively of previous loading history
(e.g., Imai 1995; Niemunis & Krieg 1996; Leroueil et al. 1996; Leroueil & Marques
1996). Kawabe et al. (2009, 2011) showed that the viscous properties of some soft clay
types are of such non-Isotach type that the current effective stress is controlled by not
only the instantaneous irreversible strain and its rate but also previous loading history.
In this study, however, only the observed rate-dependent stress-strain behaviour that is
essentially of Isotach type is presented and analized.
log a
0.2 a(creep)-RL/SL
0.1
SL
Plastic
0.0
a UL Hypo- af
P
a(creep) (%)
log a
elastic
a : axial
stress
-0.1 E
RL
SL a(creep)-UL/SL
V a : axial
-0.2 a av strain rate
TEST 1 (Loop1) Viscous
-0.3 TEST 1 (Loop2) ae air
TEST 1 (Loop3)
TEST 2 a
-0.4 TEST 3
0.01 0.1 1 10 100

UL/SL or RL/SL
Figure 1. Relationship between the creep strain and loading Figure 2. Non-linear three-component
stress level, Fujinomori clay (Kawabe et al. 2011). model (Di Benedetto et al. 2002;
Tatsuoka et al. 2002, 2008) applied to 1D
compression of soft clay.
2. Test method
The test method is basically the same as Kawabe et al. (2011). Large reconstituted soft
clay cakes were produced by consolidating slurry of Fujinomori clay (wL 42 % & PI
15) prepared at a water content twice of wL at a = 100 kPa in a 20 cm-diameter
cylinder for a week. The clay cakes were cut into disc-shaped specimens ( 60 mm h
20 mm) for oedometer tests (i.e., one-dimensional compression tests).
Two cases of test were performed. In the first case, a specimen drained from the
top and bottom ends was set in a newly designed oedometer (Fig. 3). To achieve
uniform deformation in the horizontal cross-sections, the specimen cap was fixed to a
loading capsule guided against rotation in vertical planes. Axial strains of the specimen
were evaluated from vertical displacements of the loading piston. The specimens were
vertically loaded using a stress-controlled apparatus automatically controlled the air
488 S. Kawabe and F. Tatsuoka / 1D Creep and Delayed Rebound During Unloading and Reloading
pressure in an air-cylinder by using a feed-back system. Constant-Rate of Strain (CRS)

tests including sustained loading (SL) and multiple unloading/reloading (UL/RL)
cycles with relatively large stress amplitudes were performed. In the second test case,
another apparatus strain-controlled using a precision-gear system was used. The strain
rate was suddenly changed many times during otherwise CRS loading to evaluate the
viscous properties. The specimen was back-pressurised measuring the pore water
pressure at the undrained bottom end. In this case, the measured excess pore water
pressure was always lower than 1 % of the total axial stress at the same loading strain
rate. The excess pore water pressure inside the specimen drained at the top and bottom
ends in the first test case must be even lower. So, in both test cases, the measured total
axial stress, a, can be regarded as the effective axial stress, so the observed stress-
strain behaviour can be deemed to be the drained one.
Figure 3. Oedometer used in the present study (Kongkitkul et al. 2011).
In both test cases, the inner face of the ring was smeared with a grease layer to
minimize the wall friction. Before applying respective specified loading histories, such
as the one shown in Fig. 4a, to minimize effects of disturbance during specimen
preparation, each specimen was subjected to a pre-load scheme comprising monotonic
recompression at a = 0.05 %/min to a = 100 kPa; SL for one day; UL at a =
0.05 %/min to a = 5 kPa; and SL for one day. Logarithmic axial strains are used
throughout this paper.
3. Test result
Figs. 4a-c show the results of test FJM1008, in which a number of SL stages for
three hours were applied during multiple unloading/reloading (UL/RL) cycles at
constant strain rate equal to 0.005 %/min changing the stress amplitude. Figs. 4a and c
show the time histories of axial stress, a, and creep axial strain at each sustained
loading (SL) stage, a(creep). In the Fig. 4c, the data for the similar values of (UL and
SL) or (RL and SL) after different loading histories are denoted by the same symbol.
For example, SL stage 6 and 10 were arranged during unloading at same values of UL
and SL. The time histories of creep at by similar values of (UL and SL) or (RL and
SL) are similar despite different loading histories. Another similar test, FJM1016, was
performed, in which the stress amplitude was wider than test FJM1008.
In Fig 5, the results from these two tests have been added to those plotted in Fig. 1.
It may be seen that the two sets of data exhibit nearly the same trend. This result
indicates that, as far as the void ratio range is small as in these tests, a(creep) is
independent of loading history and axial stress at SL stage, SL, but depends on the
stress ratio, UL/SL or RL/SL. Note again that this relation changes with changes in
the material type, the initial creep strain rate and the elapsed time at SL stages.
1500
Axial stress, a (kPa)
1000
500

0
0 24 48 72 96 120
a) Elapsed time, t (hour)
15 0.10
Test Name: FJM1008

Creep axial strain, a(creep) (%)
0.05

16
Axial strain, a (%)

0.00

17

-0.05

18 -0.10
Test Name: FJM1008
w0 = 37.5 %, e0 = 1.02
-0.15
19 0 1 2 3 4
100 500 1000
Elapsed time, t (hour)
c)
b)
Figure 4. Test FJM1008: a) stress history; b) aloga relation; and c) time histories of creep strain.

Figure 5. a(creep)UL/SL or RL/SL relations from tests FJM1008 & FJM1016 and those shown in Fig. 1
(Kawabe et al. 2011).
4. Model and Simulation
4.1. Non-linear three component model
According to this model, the axial strain rate, a is divided in to the elastic and
irreversible (visco-plastic) components, ae and air . The stress is divided into the
inviscid and viscous stresses, af and av . The a (or air ) af relations are the
relations obtained by imaginary primary loading (PL), unloading (UL) or reloading
(RL) at air = 0, called the reference relations. The reference relations for PL, UL and
RL are different and should be defined separately. When the current ( a , a ) state is
located on a reference curve, the creep does not take place, whereas it takes place when
the ( a , a ) state moves toward the reference curve at a constant a (i.e., sustained
loading, SL). In Fig. 5, the creep strain rate is zero when UL/SL or RL/SL becomes a
certain value, so these stress states should be located on the reference relations. In
addition, it may be seen from Fig. 5 that the values of UL/SL or RL/SL at these
different inviscid stress states in the UL/RL cycles are all the same for the different
combinations of UL and SL or RL and SL. This fact indicates that the UL and RL
reference curves can be obtained by proportionally changing the stress values of the
stress-strain curve obtained by UL and RL at a constant strain rate. This means that, for
a given constant irreversible strain rate, air , the viscous stress, av , is proportional to
the inviscid stress, af , and can be represented as:
av = af g v (air ) (1)
g v (air ) = [1 exp{1 - (| air | / rir + 1) m }] (0) (2)
where g v (air ) is the viscosity function, which is a highly non-linear of rir (Di
Benedetto et al., 2002; Tatsuoka et al., 2002). Its three parameters were determined
based on the rate-sensitivity coefficient, 1D, defined as the changing rate of the stress
upon a step change in air by a factor of 10. The data from a typical test is presented in
Figs. 6 and 7. A linear relation was fitted to the data in Fig. 7. The relation is
independent of the stress where the strain rate was changed stepwise and the step
change rate, as Eq. 2 assumes. The 1D value is equal to the slope of this linear relation
and was determined by the equation inset in Fig. 7.
2000 900
Fujinomori clay n
e0 = 1.134 Elastic relation
w0 = 42.3 %
1500 m
800 i
wL = 43.6 %
PI = 20.0
a
1000 b-c;f-g;h-i;l-m: 0.05 %/min l 700
j
a-b;g-h;m-n: 0.005 %/min i k
c-d;e-f;i-j;k-l: 0.0005 %/min
d-e;j-k: SL for one day
h
a
500 g 600 a
h
d f
c
e
a
a
b
0 500
0 5 10 15 20 25 30 16 17 18 19
a
Axial strain, (%) a
Axial strain, (%)
a) b)
Figure 6. Stress- strain relation of Fujinomori clay with many step changes in the strain rate during CRS
primary loading.

)XMLQRPRULFOD\

1D = 0.0568

= 1D log
( )
a
after

( )
a before

Figure 7. Definition of the rate sensitivity coefficient, 1D.
4.2. Simulation
As the observed drained stress-strain behaviour presented in this paper only slightly
deviates from the Isotach one (Kawabe et al. 2011), the data was analyzed assuming the
Isotach type to evaluate the applicability of the model under general loading condition
in a simple framework. The reference curves for UL and RL starting from given stress
levels were obtained by the following proportional rule defined in the arithmetic scales
of stress and strain (Fig. 8):
y yiUL x x iUL ; y yiRL x x iRL

= g = h (3a & b)
n m n m
yiUL yiRL k y yiRL k

n= , m = 0 iUL = 0 n (4a & b)
y0UL y0RL ki y0UL y0RL ki
( )
ki = Ev yiUL (5)
Eqs. 3, 4 and 5 are formulated in terms of the total stresses, not in the inviscid stresses,
for a simplicity in the analysis, because the analysis becomes very complicated when
based on the inviscid stresses that were not measured. y and x represent the inviscid
stress, af and the strain, air . Referring to Fig. 9, y = f(x) is the reference curve for PL;
yUL and xUL are the stress and strain at the start of UL, while yRL and xRL are those at the
start of RL of a given loop; Eqs. 3a and b mean that the UL and RL curves of any
arbitrary loop, for which the stress and strain are denoted by the subscript i in Fig. 9,
are obtained by scaling and shifting the stress and strain of the basic UL and RL curves,
y = g(x) and y = h(x), for which the stress and strain are denoted by the subscript 0 in
Fig. 9. The scaling factors, n and m, are obtained by Eq. 4. In the modelling of the
stress-strain behaviour of soil under cyclic loading conditions, it is often assumed that
m = n and both are kept constant. This restriction is valid only if the k value is constant
among different loops. In the present modeling, the k value is different for different UL
branches, therefore the restriction described above was not used.
Elastic relation y
k0
Reference curve I (x0UL, y0UL)
8/
Creep
y=f(x)
y=g(x)
H LU
ki

y=h(x)
Y (xiUL, yiUL)
I
I 5/
(x0RL, y0RL)
(xiRL, yiRL) x
Figure 8. Reference curves for primary loading, Figure 9. Scaling rule used in the present study.
unloading and reloading.
The simulation of test FJM1008 is presented in Fig. 10, where the largest
(outermost) loop was chosen to define the basic curves 0. The stress-strain relation
and representative time histories of creep strain, together with their simulations, are
shown in Figs. 10a and b. In Fig. 10b, the experimental data and simulated ones are
denoted by solid and hollow symbols, respectively, and the numbers shown next to the
data are the same as those in Fig. 4. It may be seen that the all details of the test results
are simulated very well. Fig. 11 compares the measured and simulated creep strain
rates at different SL states in terms of the a(creep)UL/SL or RL/SL relation. It may
be seen that both positive and negative creep behaviours at arbitrary states during
arbitrary UL and RL are well simulated. This indicates that the proposed model is able
to predict the rate-dependent drained stress-strain behaviour under general
loading/unloading/reloading conditions.

([SHULPHQW )-0

6LPXODWLRQ

5HIHUHQFHFXUYH

a) b)
Figure 10. Simulation of test FJM1008: a) stressstrain relations; and b) time histories of creep strain.

([SHULPHQW
6LPXODWLRQ

0.5 5

Figure 11. a(creep)UL/SL or RL/SL relation of test FJM1008 (experiment and simulation).
5. Summary
The conclusions that can be derived from the test results on the drained creep behaviour
of soft clay and their model simulations presented in this paper can be summarised as
following:
The drained creep strain during otherwise unloading and reloading is
controlled by the ratio of the stress at the latest load reversal to the stress at
concerned sustained loading.
The stress-strain curves that are obtained by arbitrary imaginary unloading and
reloading at zero strain rate (i.e., the reference relations) can be obtained by
properly scaling and shifting the basic curves obtained by unloading/reloading
tests with a relatively large stress amplitude.
The test results, including the time histories of creep strain at an arbitrary
fixed stress state during otherwise unloading and reloading as well as those
during otherwise primary loading, can be simulated very well by the non-
linear three-component model incorporating the viscous properties and the
reference curves determined experimentally.
References
[1] Acosta-Martnez H., Tatsuoka, F., and Li, J.-Z., Viscous property of clay in 1-D compression: evaluation
and modelling, Proceedings of the 16th ICSMGE, Osaka, (2005) 779-783.
[2] Adachi,T., Oka,F., and Mimura,M., Modeling aspects associated with time dependent behavior of soils,
S-O-A Report, Measuring and Modeling Time Dependent Soil Behavior, ASCE Geotech. Special
Publication 61 (1996), 61-95.
[3] Di Benedetto, H., Tatsuoka, F., and Ishihara, M., Time-dependent deformation characteristics of sand and
their constitutive modeling, Soils and Foundations, 42(2) (2002), 1-22.
[4] Imai, G., Analytical examination of the foundations to formulate consolidation phenomena with inherent
time-dependence, Keynote Lecture, Proceedings of International Symposium on Compression and
Consolidation of Clayey Soils, IS Hiroshima 2 (1995), 891-935.
[5] Kawabe, S., Kongkitkul, W., Hirakawa, D., and Tatsuoka, F., Rate-dependent behaviour of clay during
cyclic 1D compression, Proceedings of the 17th International Conference on SMGE, Alexandria, Egypt
(2009), 303-307.
[6] Kawabe, S., Kongkitkul, W. and Tatsuoka, F., 1D compression with unload/reload cycles on soft clay
and its simulation, Proceedings of the 14th Asian Regional Conference on SMGE, Hong Kong, China
(2011), 6 pages.
[7] Kongkitkul, W., Kawabe, S., Tatsuoka, F. and Hirakawa, D., A simple pneumatic loading system
controlling stress and strain rates for one-dimensional compression of clay. Soils and Foundations
51(1) (2011), 11-30.
[8] Leroueil, S., and Marques, M.E.S., Importance of strain rate and temperature effects in geotechnical
engineering, S-O-A Report, Measuring and Modeling Time Dependent Soil Behavior, ASCE Geotech.
Special Publication 61 (1996), 1-60.
[9] Leroueil, S., Perret, D., and Locat J., Strain rate and structuring effects on the compressibility of a young
clay, S-O-A Report, Measuring and Modeling Time Dependent Soil Behavior, ASCE Geotech. Special
Publication 61 (1996), 137-150.
[10] Niemunis, A., and Krieg, S., Viscous behaviour of soil under oedometric conditions, Canadian
[11] Tatsuoka, F., Ishihara, M., Di Benedetto, H., and Kuwano, R., Time-dependent deformation
characteristics of ge-omaterials and their simulation, Soils and Foundations, 42(2) (2002), 103-129.
[12] Tatsuoka, F., Di Benedetto, H., Enomoto, T., Kawabe, S., and Kongkitkul, W., Various viscosity types
of geomaterials in shear and their mathematical expression, Soils and Foundations, 48(1) (2008), 41-60.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-494
Creep and Stress Relaxation Envelopes of

Granular Materials in Direct Shear
Antoine DUTTINE a, Alice SALOTTI b, Fumio TATSUOKA c,1 and Alan EZAOUI d
a
Integrated Geotechnology Institute Limited, Tokyo, Japan
b
TPC Progretti, Padova, Italy
c
Department of Civil Engineering, Tokyo University of Science, Japan
d
Centre dEtudes Techniques de lEquipement CETE Lyon-DLA, France
Abstract. The behaviours of creep and stress relaxation (SR) during otherwise
monotonic loading (ML) drained direct shear (DS) at a fixed shear displacement
rate of air-dried specimens of three types of clean sand and a gravel having
different particle shapes were evaluated and compared. As the particles become
rounder, the grading becomes more uniform, as the specimen becomes looser and
as the shear displacement becomes larger, the viscous property type deviates more
from Isotach type. The SR envelope comprising multiple stress-shear displacement
states reached after stress relaxation is always located considerably below the
creep envelope reached by a creep process for the same duration of time. The
difference becomes larger as the viscous property becomes less Isotach.
Keywords. creep, direct shear, granular material, stress relaxation
1. Introduction
A number of case histories demonstrated paramount importance of better

understanding of the viscous properties of geomaterial for realistic prediction of ground
deformation and structural displacements. The Isotach model [1, 2] can describe rather
properly the rate-dependent stress-strain behaviour of many types of geomaterial
comprising particles well bound or interlocked: e.g., undisturbed and reconstituted
plastic clays; sedimentary soft rocks and cement-mixed soils (e.g., [3 - 8).
Correspondingly, many elasto-viscoplastic constitutive models assuming the Isotach
property have been proposed for geomaterials.
However, recent studies have revealed that the Isotach viscous property is only a
specific type among others [8 - 16]. The following four viscous property types have
been identified with geomaterials in drained shear: Isotach, Combined, TESRA
(Temporary Effects of Strain Rate and strain Acceleration) and P&N (Positive and
Negative). Isotach or combined type is observed with well-graded angular granular
materials (GMs), TESRA type with poorly graded angular GMs and P&M wit rly
with poorly
graded round GMs. Upon a step increase in the shear strain rate (e.g., from 0 to 10 0
at point A in Fig. 1a) during otherwise monotonic loading (ML) at a constant strain rate,
1
Corresponding Author. Professor, Department of Civil Engineering, Tokyo University of Science, 2641,
Yamazaki, Noda City, Chiba Prefecture, Japan; E-mail: [email protected]
A. Duttine et al. / Creep and Stress Relaxation Envelopes of Granular Materials 495
the Isotach type material exhibits a persistent stress increase. Under loading conditions
along a fixed stress path,, a unique stress is defined for given instantaneous irreversible
strain, ir, and its rate, ir . The stress-strain relation by infinitely slow loading (i.e., ir =
0), which does not include any viscous effects, is called the reference relation (Fig. 1b).
In Fig. 1b, a common reference relation is assumed for the four viscous property types.
With Isotach type, after an infinitive long period of creep process at a given fixed stress
and stress relaxation (SR) at a given fixed strain, the stress-strain state reaches the
reference relation.
With the other viscous property types (Combined, TESRA and P&N), the increase
in the stress upon a step increase in the strain rate is temporary (or transient), which
decays towards different residual values during subsequent straining depending on the
viscous property type. In Fig. 1a, a common stress-strain relation by continuous ML at
the same strain rate (= 0 ) is assumed for the four viscous property types. Then, the
current stress state is determined not only by given instantaneous strain and its rate but
also by strain history. The test results indicated that, with Combined, TESRA and P&N
types, during creep and SR stages, the stress-strain state ultimately passes the reference
relation as illustrated in Fig. 1b [12 - 16].
,VRWDFK ,VRWDFK
,VRWDFK
&RPELQHG
&RPELQHG
6WDUWRIFUHHS
&RPELQHG
VWUHVVUHOD[DWLRQ
7(65$ 7(65$
7(65$
3 1 3 1
%URNHQFXUYHV %URNHQFXUYHV
ir UHODWLRQV ir UHODWLRQV
$ E\FRQWLQXRXV0/ E\FRQWLQXRXV0/DW
3 1
DWDFRQVWDQWVWUDLQ
. DFRQVWDQWVWUDLQ
.
6WHSLQFUHDVHLQWKHVWUDLQ UDWHHTXDOWR UDWHHTXDOWR
UDWHE\DIDFWRURI
ir
&RPPRQUHODWLRQE\FRQWLQXRXV0/
. f ir
&RPPRQUHODWLRQUHIHUHQFHFXUYH
DWDFRQVWDQWVWUDLQUDWH
ir
ir
a) b)
Figure 1. Four different viscous property types of geomaterial in shear: different behaviors: a) upon a step
increase in the strain rate from a common curve at the same strain rate; and b) during creep and stress
relaxation during otherwise monotonic loading at a certain strain rate for a common reference curve.
Di Benedetto et al. [17] and Tatsuoka et al. [9 & 10] showed that the three-
component model is relevant to simulate a wide variety of rate-dependent stress-strain
behaviour caused by these viscous property types observed in triaxial and plane strain
compression tests of GMs and this model can simulate both creep and SR behaviours.
Fig. 2 shows this model modified to the DS test condition. Despite the above,
systematic tests performing both creep and SR tests on geomaterials exhibiting
different viscous property types are very limited (e.g., [18 & 19]). When using a
displacement-control apparatus developed to perform ML tests at fixed strain rates, SR
tests by simply fixing the displacement are much easier to perform than creep tests.
Besides, the time needed reach the same stress-strain state relative to the ML stress-
strain relation is much shorter in a SR test than in a creep test. For these reasons, it is
very convenient if the long-term creep behaviour (and other rate-dependent
phenomena) can be predicted from SR tests performed for a much shorter duration.
In the present study, to understand the processes of creep and stress relaxation (SR)
and their relationship, effects of the viscous property type on the amount and rate of
496 A. Duttine et al. / Creep and Stress Relaxation Envelopes of Granular Materials
creep deformation and SR were evaluated by performing a series of drained direct

shear tests on four types of granular materials (GMs). In the companion paper [20], it is
examined whether the three-component model can simulate these test results in the
consistent manner and shown that the trends of viscous behaviour observed in the
experiments of GMs having different viscous properties can be simulated very well by
the three component model.

Inviscid component
RDSf(sir)
GeDS P
' RDS=*vh/v
&.
%s
Elastic
component V .
RDSv(sir,sir,hs)
Viscous component
. .
se sir
Figure 2. Non-linear three-component model adapted for the DS test conditions [14 & 16].
100
Toyoura sand
1 Chiba gravel-a
80 4
Percent finer than D [%]
60
Albany sand
Ticino sand
28
40 34
20
0
0.01 0.1 1 10
Particle diameter, D [mm]
Material Gs D50 (mm) Uc emin(1) emax(1) A*(2)

Chiba gravel-a 2.74 1.93 3.05 0.602*) 0.907*) 1967
Toyoura sand 2.648 0.180 1.625 0.592 0.978 896
Ticino sand 2.680 0.527 1.521 0.590 0.960 449
Albany sand 2.671 0.300 2.200 0.505 0.804 209
(1)
Determined following JIS A 1224:2009 (the Japanese Geotechnical Society).
(2)
Total degree of angularity [21]). *) H 0.016 (scatter due to relatively large particle size).
Figure 3. Granular materials used in this study.

2. Apparatus and test materials
A fully automated precise direct shear (DS) apparatus [14 & 16] was used. The shear
displacement history was controlled to an accuracy of less than 1 m to perform in a
single test: i) arbitrary smooth switching between displacement and load control phases
and among sustained loading (i.e., creep), stress relaxation and constant displacement
rate loading; and ii) stepwise or gradual changes in the displacement rate by a factor of
up to 100. The specimens (12 cm x 12 cm x 12 cm) were produced using the granular
materials described in Fig. 3. They are three poorly-graded natural sands: a sub-angular
to angular sand (Toyoura sand, Japan) and a round sand (Albany sand, Australia), both
quartz-rich sands; and a sub-round/sub-angular sand (Ticino sand, Italy), comprising
quartz (28 % by volume), feldspar (30 %) and mica (5 %) [21]; and an angular gravelly
soil (Chiba gravel-a, Japan), comprising crushed angular sandstone particles from a
quarry in Japan. The specimens were divided into several sub-layers (typically 7 or 8)
and each sub-layer was tamped with a square steel rod very carefully so that the
DST Air dried Toyoura sand

1.1
Drc=96.28% v=50kPa
.
s=0.08mm/min
creep
Simulation
Stress ratio RDS=*vh / v
1.0
SR SR
5 hour creep & SR envelopes
0.9 Experiment
SR
CL Parameters for simulation

creep . ir -9
0.8 =0.25 ; m=0.025 ; sref=10 mm/s
-15 -25
Reference curve i=0 ; f=-0.54 ; ri=10 ; rf=10
ir
so=2.4 mm n =1.556
e
0.7 GDS=30 /mm
SR
Creep & SR: for 5 hours
CL: cyclic loading (5 cycles)
0.6 creep
0.0 0.5 1.0 1.5 2.0

a) Shear displacement, s [mm]
-0.75
-0.60 Creep & SR (stress relaxation): for 5 hours

Vertical displacement, d [mm]
CL: cyclic loading (5 cycles) SR
-0.45
SR
I=18.7/
-0.30
Dilatation
Creep
-0.15 SR
CL
Creep
Creep
0.00
SR
0.15
0.0 0.5 1.0 1.5 2.0
b) Shear displacement, s [mm]
Figure 4. a) Stress ratio - shear displacement relation and its simulation; and b) volume change shear
displacement relation, dense Toyoura sand.
Figure 5. Typical time histories of shear displacement and shear stress during: a) sustained loading; and
b) stress relaxation and their simulations, dense Toyoura sand (Fig. 4).
prescribed dry density was precisely achieved. The initial opening between the top and
bottom shear boxes (before consolidation) was set equal to 10 0 D50 for dense specimens
and 20 0 D50 for loose specimens of the respective materials, except Chiba gravel-a,
with which the opening was set equal to 10 mm. The drained DS tests were performed
at a constant normal pressure v= 50 kPa.
3. Test results
Fig. 4a presents the relationship between the ratio of RDS (= average shear stress *vh"
to "average vertical stress v) and the shear displacement (s) of dense air-dried
Toyoura sand. The simulations shown in this and other similar figures are explained in
the companion paper (Tatsuoka et al., 2015). At different shear stress levels during
otherwise ML at a constant shear displacement rate, the specimen was subjected to a
series of alternative 5-hour sustained loading for creep and 5-hour stress relaxation
(SR). A set of five small unload/reload cycles was also applied to evaluate the elastic
properties at different stress states. Significantly rate-dependent stress-strain behaviours
may be seen. Fig. 4b shows the relationship between the vertical displacement, d (i.e.,
the volume change, positive in compression) and s. The d - s relation exhibits a small
Figure 6. Stress ratio - shear displacement relations and their simulations from DS tests on dense
specimens of: a) Chiba gravel-a; b) Ticino sand; and c) Albany sand.
deviation during SR stages from the relation during ML at a constant displacement rate.
This trend is due essentially to elastic rebound associated with a decrease in the shear
stress and its recovery during reloading. On the other hand, the volume change during
creep stages is due essentially to 1) creep shear deformation-induced volume changes,
which is dilative except for the first creep stage; and b) contractive volumetric creep.
Duttine et al. [14] showed that the shear deformation-induced volume change
characteristics (i.e., the flow characteristics) are essentially rate-independent, while the
volumetric creep becomes more significant as the specimen becomes looser. Typical
time histories of s and *vh during creep and SR are shown in Figs. 5a & b. The shear
stress was kept nearly constant during creep (Fig. 5a). During SR, an inevitable very
small elastic rebound of the loading frame ,s of around 15m took place immediately
after the start of SR stage (Fig. 5b). This instantaneous small displacement increment
was taken into account in the simulations. After a period of 5 hours, the rates of creep
and stress relaxation have become very small (although they have not become perfectly
null).
In this paper, only the test results of dense specimens are presented. Figs. 6a, b and
c show the similar test results of dense specimens of Chiba gravel-a, Ticino sand and
Albany sand. The same remarks as Toyoura sand are relevant to these test results, other
than effects of different viscous property types, as described below. In the companion
paper, based on the results from other test series in which the displacement rate was
step-wise changed many times and their analysis by the model (Fig. 2), it is shown that,
for the pre-peak behaviors, the viscous property of Toyoura and Ticino sands is Isotach,
that of Chiba-gravel-a is Combined and that of Albany sand is P &N. Referring to Fig.
1b, these different viscous property types can be identified by comparing each
measured RDS - s relation at a constant displacement rate with the respective reference
curves inferred by the simulation presented in Figs. 4 and 6.
In Figs. 4a and 6, the 5-hour creep and stress relaxation (SR) envelopes
experimentally obtained by connecting multiple end states (RDS, s) of creep and SR
processes are depicted. With all these granular materials, the SR envelope is
systematically located significantly lower than the creep envelope for the same period
as the SR (i.e., five hours), regardless of viscous property type. The difference between
the creep and SR envelopes is noticeably larger with Albany sand P&N than with
Chiba gravel-a (Combined) and TESRA (Toyoura and Ticino sands). This trend
corresponds to the fact that the effects of viscous property type on the creep
deformation are larger than those on the stress drop during SR. In the companion paper,
these trends are explained by detailed numerical analysis based on the three-component
model.
4. Conclusions
From the test results, the following conclusions can be derived:

1) Different unbound natural sands and gravel having largely different particle shapes
exhibited different viscous property types in drained direct shear, while showing
significant creep and stress relaxation (SR) during otherwise monotonic loading
(ML) at a constant displacement rate.
2) With these granular materials, the creep and SR envelopes representing the shear
stress-shear displacement states reached after a series of creep and SR for five
hours starting from different stress/displacement states along the same ML shear
stress-shear displacement curve are very different from each other. Consistently,
the SR envelope is located largely below the creep envelope relative to the ML
curve.
3) The difference between the creep and SR envelopes for five hours increases as the
viscous property type becomes less Isotach, corresponding to the fact that the
effects of viscous property type on the creep deformation are larger than those on
the stress drop during SR.
References
[1] Suklje, L., The analysis of the consolidation process by the isotache method, Proc. 4th Int. Conf. Soil
Mech. and Found. Engineering, London, 1(1957), 200-206.
[2] Suklje, L., Rheological aspects of soil mechanics, Wiley-Interscience (1969).
[3] Hayano, K., Matsumoto, M., Tatsuoka, F. and Koseki, J., Evaluation of time-dependent deformation
property of sedimentary soft rock and its constitutive modelling, Soils and Foundations, 41(2) (2001),
21-38.
[4] Leroueil, S., The isotach approach: where are we 50 years after its development by Prof. Suklje?, Proc.
XIII Danube-European Conf. on Geotech. Engineering, Slovenia, 1 (2006), 55-88.
[5] Sorensen, K., Baudet, B. A. and Simpson, B., Influence of structure on the time-dependent behaviour of
a stiff sedimentary clay, Gotechnique, 57(1) (2007), 113-124.
[6] Kongsukprasert, L. and Tatsuoka, F. (2005): Ageing and viscous effects on the deformation and strength
characteristics of cement-mixed gravely soil in triaxial compression, Soils and Foundations, 45(6): 55-
74.
[7] Ezaoui, A., Tatsuoka, F., Sano, Y., Iguchi, Y., Maeda, Y., Sasaki, Y. and Duttine, A., Ageing effects on
the yielding characteristics of cement-mixed granular materials, Soils and Foundations, 50(5) (2010),
685-704.
[8] Ezaoui, A., Tatsuoka, F., Duttine, A. and Di Benedetto, H., Creep failure of geomaterials and its
numerical simulation, Geotechnique Letters (accepted, published online July 2011, doi:
10.1680/geolett.11.00009) (2011).
[9] Tatsuoka, F., Ishihara, M., Di Benedetto, H. and Kuwano, R., Time-dependent shear deformation
characteristics of geomaterials and their simulation, Soils and Foundations, 42(2) (2002), 103-129.
[10] Tatsuoka, F., Di Benedetto, H., Enomoto, T., Kawabe, S. and Kongkitkul, W., Various viscosity types
of geomaterial in shear and their mathematical expression, Soils and Foundations, 48(1) (2008), 41-60.
[11] Tatsuoka, F., Laboratory stress-strain tests for developments in geotechnical engineering research and
practice, Bishop Lecture, Proc. 5th Int. Symp. on Deformation Characteristics of Geomaterials, IS-
Seoul 2011 (Chung et al. eds.) (2011), 3-50.
[12] Kongkitkul, W., Tatsuoka, F., Duttine, A., Kawabe, S., Enomoto, T. and Di Benedetto, H., Modelling
and simulation of rate-dependent stress-strain behaviour of geomaterial, Soils and Foundations, 48(2)
(2008), 175-194.
[13] Enomoto, T., Kawabe, S., Tatsuoka, F., Di Benedetto, H., Hayashi, T. and Duttine, A., Effects of
particle characteristics on the viscous properties of granular materials in shear, Soils and Foundations,
49(1) (2009), 25-49.
[14] Duttine, A., Tatsuoka, F., Kongkitkul, W. and Hirakawa, D., Viscous behaviour of unbound granular
materials in direct shear, Soils and Foundations, 48(3) (2008), 297-318.
[15] Duttine, A., Tatsuoka, F., Lee, J. and Kongkitkul, W., Viscous properties of Toyoura sand over a wide
range of strain rate and its model simulation, Soils and Foundations, 49(2) (2009), 221-247.
[16] Duttine, A. and Tatsuoka, F., Viscous properties of granular materials having different particle shape in
direct shear, Soils and Foundations, 49(5) (2009), 777-796.
[17] Di Benedetto, H., Tatsuoka, F. and Ishihara, M., Time-dependent shear deformation characteristics of
sand and their constitutive modelling, Soils and Foundations, 42(2) (2002), 1-22.
[18] Lade, P.V., Nam, J. and Liggio Jr., C.D., Effects of particle crushing in stress drop-relaxation
experiments on crushed coral sand, Journal of Geotechnical and Geoenvironmental Engineering,
ASTM, 136(3) (2010), 500-509.
[19] Lade, P.V. and Karimpour, H., Static fatigue controls particle crushing and time effects in granular
materials, Soils and Foundations, 50(5) (2010), 573-583.
[20] Tatsuoka, F., Duttine, A., Salotti, A. and Ezaoui, A., Creep and stress relaxation envelopes of granular
materials simulated by non-linear three-component model, Proc. of 15th Pan-American Conference on
Soil Mechanics and Geotechnical Engineering and 6th International Conference on Deformation
Characteristics of Geomaterials, Buenos Aires (2015), this conference.
[21] Lees, G., A new method for determining the angularity of particles, Sedimentology, 3(1964), 2-21.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-503
Creep and Stress Relaxation of Granular

Materials Simulated by Non-linear Three-
Component Model
Fumio TATSUOKA a,1 Antoine DUTTINE b, Alice SALOTTI c, and Alan EZAOUI d
a
Department of Civil Engineering, Tokyo University of Science, Japan
b
Integrated Geotechnology Institute Limited, Tokyo, Japan
c
TPC Progretti, Padova, Italy
d
Centre dEtudes Techniques de lEquipement CETE Lyon-DLA, France
Abstract. Creep deformation and stress relaxation during otherwise monotonic

loading drained direct shear (DS) at a fixed shear displacement rate of three types
of clean sand and a gravel having different particle shapes and exhibiting different
viscous property types are simulated by the non-linear three component model.
Elastic body represents the hypo-elastic property, plastic body represents the
plastic properties exhibiting a non-linear inviscid stress - irreversible strain relation
and viscous body represents the viscous properties as a function of irreversible
strain and its rate with or without control of loading history. It is shown that the
trends of rate-dependent behavior observed in the DS tests, particularly creep and
stress relaxation and their relationship, can be successfully simulated in the
consistent manner by the non-linear three-component model.
Keywords. creep, direct shear, granular material, simulation, stress relaxation,

three-component model
1. Introduction
It is shown in the companion paper [1] that, in drained direct (DS) tests, unbound
granular materials (GMs) (i.e., sands & gravels) exhibit creep and stress relaxation
(SR) envelopes that represent the shear stress - shear displacement states reached by a
series of creep and SR for a certain duration starting from different stress/displacement
states along the same monotonic loading (ML) shear stress - shear displacement curve.
The SR envelope is located largely below the creep envelope. Moreover, GMs having
different particle shapes and grading characteristics exhibit different viscous property
types (i.e., Isotach, Combined, TESRA and P&N), as illustrated in Fig. 1 of the
companion paper. As the viscous property type changes from Isotach toward P&N,
both creep deformation and stress drop during SR decrease while the decrease in the
creep deformation is more significant, therefore the SR envelope is located below the
creep envelope more significantly. In this paper, it is shown that these trends of rate-
dependent behavior can be simulated very well by the three-component model.
1
Corresponding Author. Professor, Department of Civil Engineering, Tokyo University of Science,
2641, Yamazaki, Noda City, Chiba Prefecture, Japan; E-mail: [email protected]
504 F. Tatsuoka et al. / Creep and Stress Relaxation of Granular Materials
Inviscid component
RDSf(sir)
Ge
DS
P
' RDS=*vh/v
&.
%s
Elastic
component V .
RDSv(sir,sir,hs)
Viscous component
. .
se sir
.
.
RDSv(sir,sir,hs) DS(sir).[RDSv]iso(sir,sir)
Viso
V =
VTESRA .
1JDS(sir)].[RDSv]TESRA(sir,sir,hs)
a)
b)
Figure 1. a) Non-linear three-component model adapted to the DS test conditions (hs: loading history
parameter); and b) relevant stress parameters on *vh- v plane [2 & 3].
2. Non-linear three-component model and model parameters
It has been shown that the three-component model can simulate a wide variety of rate-
dependent stress-strain behaviour including creep and SR of geomaterials observed in
the triaxial and plane strain compression tests [4 - 7] and in the drained DS tests [2, 3 &
6 - 9]. Fig. 1 shows this model adapted to the DS test condition. The shear
p
displacement rate ( s ) comprises elastic part ( s e ) and irreversible (or visco-plastic) part
ir
( s ). In this study, it is assumed that the tangent elastic modulus GeDS of the RDS - se
relation of Elastic component is constant in the respective DS tests. The value of GeDS
in each test was determined by performing small unload/reload tests during otherwise
ML. The shear stress ratio (RDS) comprises inviscid part (RDSf) in P body and viscous
part (RDSv) in V body. RDSv is expressed as:
v
RDS (sir ,siri ,hs ) DS (siri ) 0 RDS
v
(s ,s ) 1 DSS (s ) 0 RDSS
S
ir
i iir ir
i v
(siri ,siri ,h
hs ) (1)
iso TESRA
where hs is the parameter representing the previous strain history; and DS is the DS
viscous property type parameter characterizing the different viscous property types:
i.e., DS 1.0 (Isotach; RDSv becomes Isotach viscous stress, [RDSv]iso); 0 <DS< 1.0
(Combined); DS= 0 (TESRA, RDSv becomes TESRA viscous stress, [RDSv]TESTA); and
DS< 0 (P&N). [RDSv]iso and [RDSv]TESTA are obtained as:
F. Tatsuoka et al. / Creep and Stress Relaxation of Granular Materials 505
'
! s ir $ !
RDS
v
(s ir , s iri ) RDS
f
(s ir ) 0 gv (s ir ) RDS
f
(s ir ) 0 0 &1 exp 1 ( ir 1) m #
(2a)
iso
!
% sref "
!
ir ir
(2b)
3 3 dR
* )
0 gdecay (sir * )
ir
RDS
v
(sir , siir ,hs ) dRDS
v
v
DS
0 r( s
TESRA iso( * ) iso( * ) 1
* a * a
where RDSf is the inviscid stress ratio. The reference relation in terms of the RDSf - s
(=se+sir) relation in the respective DS tests was determined by fitting the test data to:
2
sir s p

sr

Rr R p Rr .e
a
2 s .s ir
1 1

( s ir )
p
.Rp . (3)
ir
f
RDS
s p s ir 1 e 2. k . s 2.k . s s p
ir
2.k . sir s p
1 e 1 e
where Rp and Rr are the peak and residual values of RDSf; sir= s - se (= RDS/GDSe); sp and
sr are the values of sir at the peak stress state and the start of residual state; a (> 0) is a
constant; and k is another constant controlling the transition from the first term in the
pre-peak regime toward the second terms in the post-peak regimes. The value of k is
equal to 100 for dense specimens and 50 to 60 for loose to medium-dense specimens.
Figure 2. a) gv; and b) DS & r1 as a function of sir used in the simulation in this study.
gv (sir ) in Eq. 2a is the viscosity function that is a non-linear function of s ir (Fig. 2a).
The method to determine the parameters {; m; s } of this function is explained later.
ir
ref
gdecay= r1( s * ) in Eq. 2b is the decay function, where r1 is the decay parameter (< 1.0). r1
ir
controls the decay rate of [dRDSv]iso that takes place when sir = with an increase in sir
during subsequent loading. When r1= 1.0 therefore, r1( s * ) = 1.0, [dRDSv]iso does not
ir
decay and [RDS ]TESRA (Eq. 2b) becomes [RDS ]iso (Eq. 2a). When 0 <r1< 1.0, [dRDSv]iso
v v
decays at a rate that increases with a decrease in r1. In this case, [RDSv]TESRA (Eq. 2b)
could be either positive or zero or negative depending on loading history. The direct
integration of Eq. 2b for a given full previous loading history in a numerical simulation
(e.g., FEM analysis) of a boundary value problem is extremely time-consuming, thus
not practical. This integration becomes practical by using an equation in an incremental
form derived from Eq. 2b by taking advantage of the nature of the power law of the
decay function [5, 6, 7 & 8]. DS and r1 decrease with sir as illustrated in Fig. 2b, where
{iKri} and {fKrf} are respectively the initial and final values of DS and r1; and {n;
s0ir } are constants. There exists a strong correlation between DS and r1 with a large
variety of poorly-graded granular materials [8].
RDS=*P
vh/v
.
safter
,(RDS)r
,RDS
.
sbefore (RDS)r
. RDS
a step
A stepchange
.
sbefore in the shear
change in s
displacement rate s
Figure 3. Definitions of DS & (DS)r [1 & 2].
Table 1. Parameters for viscosity function, gv, and decay parameters (D: dense, L: loose)
Table 2. Viscous property type parameter (D: dense, L: loose)
ir
The parameters {; m; sref } of the viscosity function, gv (Fig. 2a), were determined
as follows. As the present study does not deal with the behaviour at extremely high
shear displacement rates, it is not necessary to accurately determine the value of . So,
=LMNOwas assumed. On the other hand, the value of m was determined so that the
slope of the linear part of the gv log( s ir ) relation (Fig. 2a), which is theoretically
equal to m, becomes equal to DS/ln(10), where DS is the rate-sensitivity coefficient

defined as [,RDS RDS ]/ log sirafter sirbefore [3 & 8]. As illustrated in Fig. 3, RDS is the
change in RDS= vh/v upon a step change in s ir from sirbefore to s irafter during otherwise
ML at a constant shear displacement rate. The parameters of gv used in the simulation
presented here are listed in Table 1.
0.86
DST Air-dried Chiba gravel-a
0.80 Dr0=63.86% , v=50kPa
0.80 reference curve
.

Stress ratio RDS=*vh/v
simulation
0.74
0.72 simulation
experiment
0.68
experiment
0.62
DST Air-dried Chiba gravel-a 0.64
reference curve Dr0=63.86% , v=50kPa
0.56 .

0.56
0.50
1 2 3 4 5 6 6 7 8 9 10 11 12 13
Shear displacement, s [mm] Shear displacement, s [mm]
a)
0.74 DST Air-dried Ticino sand
0.72 Drc=58.42% , v=50kPa
0.72 reference curve

reference curve
0.68
simulation
0.70
0.64
DST Air-dried Ticino sand 0.68 experiment
experiment Drc=58.42% , v=50kPa
simulation
.
0.60
0.66
0.56 0.64
1 2 3 4 5 6 5 6 7 8 9
Shear displacement, s [mm]
b) Shear displacement, s [mm]
0.75
DST Air-dried Albany sand DST Air-dried Albany sand
reference curve
Drc=67.30% , v=50kPa Drc=67.30% , v=50kPa
0.70 0.72
simulation .

reference curve
0.65 simulation
0.64
0.60
SD
SD SD
0.55 SD
experiment
SD experiment
0.56
0.50 SD SD
SD (stress drops due to stick-slip phenomenon,
ignored in the simulation) SD (stress drops due to stick-slip
phenomenon, ignored in the simulation)
0.45 .

0.48
0 1 2 3 4 5 6 7 8 9
c) Shear displacement, s [mm] Shear displacement, s [mm]
Figure 4. DS tests including step changes in the shear displacement rate in the pre-peak and post-peak
regimes and their simulations: a) loose Chiba gravel-a; b) loose Ticino sand; and c) loose Albany sand.
The viscous property type parameter, DS, is defined as (DS)r/DS, where (DS)r is
the residual value of DS, defined as (RDS)r/(RDS)r/ log sirafter sirbefore (Fig. 3), where
(RDS)r is the value of RDS that has fully decayed during the subsequent ML after a
step change in s ir ; and (RDS)r is the RDS value at the shear displacement at which
(RDS)r is defined along the stress-displacement curve obtained if ML had continued at

the previous s ir , sirbefore . (DS)r shows how RDS that has taken place by a step change in
s ir survives subsequent shearing. Fig. 4 show typical measured RDS - s relations from
DS tests in which s was suddenly changed by a factor of 10 or 100 at every shear
displacement increment of 0.5 mm or 1.0 mm on loose specimens of three GM types
[7]. In the respective tests, the value of DS decreases with shear displacement (Fig. 2b).
The values of DS, (DS)r and DS, hence the values of the parameters (m; iKfKn; s0ir ),
for the tests presented in Fig. 4 and others are listed in Table 2.
It was confirmed by sieving the test materials before and after the respective tests
presented in this paper and the companion paper [1] that crushing of particles in these
DS tests was not noticeable. Therefore, the trends of rate-dependent behaviour
observed in the present study are not dominantly associated with crushing of particles.
0.5
Toyoura Ticino Chiba Albany v=50 kPa

sand sand gravel-a sand .
Creep displacement increment, ,s [mm]
s=0.08 mm/min
experiment
0.4
simulation
0.3
after 5hours
Chiba (Dr0=64.03%)
0.2
Ticino (Drc=59.14%) Toyoura

0.1 (Drc=60.37%)
Albany
Drc=67.74%
0.0
a) 0.5 0.6 0.7 0.8 0.9 1.0
Toyoura Ticino Chiba Albany

0.5 sand sand gravel-a sand
Creep displacement increment, ,s [mm]
Experiment
Simulation
0.4
v=50 kPa
.
after 5hours
s=0.08 mm/min
0.3
Chiba (Drc=88.66%)
Toyoura
0.2 (Drc=96.28%)
Albany
Ticino (Drc=98.42%) (Drc=94.67%)
0.1
0.0
0.5 0.6 0.7 0.8 0.9 1.0
Stress level, RDS/RDS.peak
b)
Figure 5. Creep shear displacement after five hours as a function of stress level during otherwise ML (ds/dt=
0.08 mm/min) and their simulations: a) loose; and b) dense granular materials.
3. Simulation of DS tests
From the different trends of behaviour upon a step change in s seen in Fig. 4 while
referring to Fig. 1a of the companion paper [1], it may be seen that these GMs exhibit
different viscous property types. That is, in the pre-peak regime, the viscous property
type is Combined with Chiba gravela (Fig. 4a); TESRA with Ticino sand (Fig. 4b)
(and Toyoura sand [8]); and P&N with Albany sand (Fig. 4c). In the post-peak regime,
these viscous property types gradually change with an increase in s toward,
respectively, TESRA with Chiba gravel-a; P&N with Ticino sand (and Toyoura sand);
and P&N coupled with a stick-slip phenomenon with Albany sand. It may also be seen
from Fig. 4 that these trends are simulated very well by the three-component model.
Stress relaxation increment, ,*vh [kPa] after 5hours
-6
Toyoura Ticino Chiba Albany v=50 kPa
-5 experiment s=0.08 mm/min
simulation
-4
Chiba (Dr0=64.03%)
-3
Ticino (Drc=59.14%)
-2
Toyoura
(Drc=60.37%)
Albany
-1 Drc=67.74%
0
a) 0.7 0.8 0.9 1.0
-7
Toyoura Ticino Chiba Albany
v=50 kPa
-6 s=0.08 mm/min
Stress relaxation increment, ,*vh [kPa]
Experiment
Simulation
-5 Chiba
(Drc=88.66%)
after 5hours
-4
Toyoura (Drc=96.28%)
-3
-2
Albany (Drc=94.67%)
-1 Ticino
(Drc=98.42%)
0
0.7 0.8 0.9 1.0
Stress level, RDS/RDS.peak
b)
Figure 6. Stress relaxation after five hours as a function of stress level during otherwise ML (ds/dt= 0.08
mm/min) and their simulations: a) loose; and b) dense granular materials.
It may also be seen from Figs. 3, 4 and 5 of the companion paper [1] that the creep
and SR behaviours are simulated very well by the model. Figs. 5a & b compare the
experimentally obtained and simulated creep shear displacement after 5 hours as a
function of the shear stress level, RDS/RDS.peak, where RDS.peak is the measured or
simulated peak stress ratio, of loose and dense specimens of four GM types exhibiting
different viscous property types. Figs. 6a & b show similar comparisons of SR after 5
hours as a function of RDS/RDS.peak, where RDS is the value at the start of SR. As seen
from Fig. 5, the effects of shear stress level and specimen density on the creep
deformation are significant. On the other hand, the effects of the initial shear stress
level on SR are much smaller. This trend is due to that the shear stress level
consistently drops during a given SR process. In addition, the effects of particle shape
and the coefficient of uniformity (i.e., the effects of viscosity property type) on the
creep behaviour are significant: i.e., the creep deformation noticeably decreases as the
viscous property type becomes less Isotach. On the other hand, the effects of viscous
property type on SR are smaller. These observed trends of behaviour are well simulated
by the three-component model.
4. Conclusions
The following features of the viscous responses observed in drained direct shear tests
can be very well simulated by the three-component model using the same model
parameters for the same GM type: a) a stress jump upon a step increase or decrease in
the shear displacement rate and their decay at different rates during subsequent
monotonic loading (ML) at a constant shear displacement rate; b) a transition of
viscous property type with shear displacement; c) development of a high stiffness zone
upon the restart of ML at a constant strain rate from the end of creep or stress
relaxation (SR) stage; and d) a strong increase in the creep deformation with an
increase in the shear stress level with a weak increase in SR. Moreover, irrespective of
viscous property type (i.e., Isotach, Combined, TESRA and P&N), the SR envelope
reached by a series of SR for a certain period that have started from different shear
stresses along the same ML shear stress shear displacement curve is located
consistently lower than the corresponding creep envelope obtained under otherwise the
same condition. This trend and the effects of viscous property type on the amount and
rate of creep deformation and SR can be simulated very well.
These results indicate that, when the model parameters are determined
appropriately, the three-component model can predict the histories of stresses and/or
deformation along arbitrary stress or strain paths, including the creep and SR paths.
References
[1] Duttine, A., Tatsuoka, F., Salotti, A. and Ezaoui, A., Creep and stress relaxation envelopes of granular
materials in direct shear, Proc. of 15th Pan-American Conference on Soil Mechanics and Geotechnical
Engineering and 6th International Conference on Deformation Characteristics of Geomaterials, Buenos
Aires (2015), this conference.
[2] Duttine, A., Tatsuoka, F., Kongkitkul, W. and Hirakawa, D., Viscous behaviour of unbound granular
materials in direct shear, Soils and Foundations, 48(3) (2008), 297-318.
[3] Duttine, A., Tatsuoka, F., Lee, J. and Kongkitkul, W., Viscous properties of Toyoura sand over a wide
range of strain rate and its model simulation, Soils and Foundations, 49(2) (2009), 221-247.
[4] Di Benedetto, H., Tatsuoka, F. and Ishihara, M., Time-dependent shear deformation characteristics of
sand and their constitutive modelling, Soils and Foundations, 42(2) (2002), 1-22.
[5] Tatsuoka, F., Ishihara, M., Di Benedetto, H. and Kuwano, R., Time-dependent shear deformation
characteristics of geomaterials and their simulation, Soils and Foundations, 42(2) (2002), 103-129.
[6] Tatsuoka, F., Di Benedetto, H., Enomoto, T., Kawabe, S. and Kongkitkul, W., Various viscosity types of
geomaterial in shear and their mathematical expression, Soils and Foundations, 48(1) (2008), 41-60.
[7] Kongkitkul, W., Tatsuoka, F., Duttine, A., Kawabe, S., Enomoto, T. and Di Benedetto, H., Modelling
and simulation of rate-dependent stress-strain behaviour of geomaterial, Soils and Foundations, 48(2)
(2008), 175-194.
[8] Duttine, A. and Tatsuoka, F., Viscous properties of granular materials having different particle shape in
direct shear, Soils and Foundations, 49(5) (2009), 777-796.
[9] Tatsuoka, F., Laboratory stress-strain tests for developments in geotechnical engineering research and
practice, Bishop Lecture, Proc. 5th Int. Symp. on Deformation Characteristics of Geomaterials, IS-
Seoul 2011 (Chung et al. eds) (2011), 3-50.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-511
Environmentally Friendly Transparent Soil

Jun Kang CHOWa, Ghee Leng OOI a, Yu-Hsing WANGa1
a
The Hong Kong University of Science and Technology, Kowloon, Hong Kong
Abstract. Much of the work performed in geotechnical engineering

cannot be directly observed and measured as everything beneath
the Earths surface is unobservable. Transparent soil modeling, a
non-intrusive optical visualization technique, has been introduced
to visualize the soil behavior. Limitations are found in using
transparent soils developed, such as non-environmentally friendly,
corrosive to latex membranes of geotechnical apparatus and
hazardous to human health. Hence, safe, colorless, stable and
odorless sugar solution (pore fluid) was selected, together with
fused quartz (granular material), to develop transparent soil of
environmentally friendly, compatible to geotechnical apparatus,
safe to human health and feasible in geotechnical engineering
modeling. To demonstrate the feasibility of this transparent soil,
compaction grouting and pile installation processes were modeled.
Based on the preliminary results, the environmentally friendly
transparent soil developed was applicable to geotechnical
engineering modeling to visualize the soil behavior from a different
perspective.
Keywords. Transparent soil, geotechnical modeling, environmentally friendly
1. Introduction
Since time immemorial, all geotechnical engineering works performed under the ground,
e.g. pile driving, grouting and etc., cannot be directly observed and measured. For a
workaround, the accuracy of computed results never exceeds that of a crude estimate as
mentioned by Karl Terzaghi in 1936 [1]. An educated judgment based on direct
observations is more convincing than computational results as the real observations show
direct insights to visualize the soil behavior. This motivated geotechnical engineers [2-
4] to develop transparent soils, a non-intrusive optical visualization technique, to observe
and measure the three-dimensional deformation and flow characteristics within a soil
continuum. Transparent soils consist of a granular material and a pore fluid with matched
refractive index which exhibit transparency characteristic as a whole. The combinations
of granular material and pore fluid that have been used to model transparent soils are
summarized in Table 1.
1
512 J.K. Chow et al. / Environmentally Friendly Transparent Soil
Table 1. Summary of combinations of granular material and pore fluid used to model transparent soils. (Data
are compiled based on [3, 4, 5, 6, 8,10, 11, 12]
Granular Material Pore Fluid Refractive Index
Amorphous silica powder Mineral oil and paraffinic solvent 1.448
blend; calcium bromide brine
Silica gel Mineral oil and paraffinic solvent 1.448
blend; calcium bromide brine
Aquabeads Water 1.333
Nafion Sorbitol Solution 1.340
Fused quartz Mineral oil and paraffinic solvent 1.459
blend; sucrose solution
As reported in the literature, few types of granular materials have been selected to
develop transparent soils, i.e. (1) amorphous silica powder [5-6] to model clay in
studying pile penetration [7]; (2) silica gel [8] to model sand in studying soil deformation
due to tunneling [9]; (3) Aquabeads [10] to model weak marine sediments; (4) Nafion
[11] to image the living plants and root-associated microorganisms and; (5) fused quartz
[4, 12] to investigate the soil-geosynthetic interaction during pullout [13].
However, limitations are found in the transparent soil modeling done previously,
such as non-environmentally friendly, corrosive to latex membranes of geotechnical
apparatus and hazardous to human health. Hence, the objective of this paper is to develop
a transparent soil of environmentally friendly, compatible to geotechnical apparatus, safe
to human health and feasible in geotechnical engineering modeling.
The paper begins with a short introduction to the material selected to model
environmentally-friendly transparent soil. The feasibility of this transparent soil in
modeling compaction grouting and pile installation processes was then demonstrated.
2. Preparation of Transparent Soil
Inspired by Guzman and his team [12], to develop environmentally friendly transparent
soil, sugar solution was selected as pore fluid due its properties of safe (low toxicity),
colorless, stable and odorless contrary to other pore fluids used. According to [4], mineral
oil was reported to be corrosive as it deteriorated the rubber membranes of geotechnical
apparatus; paraffin oil is flammable and toxic to human health under overexposure [14].
Fused quartz with size ranged 1 to 10 mm as shown in Figure 1a (for grain size
distribution, see Figure 1b), manufactured by Lianyungang Fenqiang Trading Co. Ltd,
was used as the granular material to develop transparent soils. Fused quartz was selected
because it is a hard material with good optimal transmission, refractive index of 1.459
and has geotechnical properties similar to those of granular soils with angular particles
[4]. The combination of sugar solution and fused quartz aimed to develop a transparent
soil of environmentally friendly, safe and feasible to geotechnical engineering modeling.
J.K. Chow et al. / Environmentally Friendly Transparent Soil 513
100
10.00 mm 6.30 mm 5.00 mm 3.35 mm

80
Percentage passing (%)

60
40
20
0
2.00 mm 1.18 mm 0.60 mm
0.01 0.1 1 10
Particle Size (mm)
(a) (b)
Figure 1 a. Different sizes of fused quartz supplied; b. Particle size distributions of fused quartz
Sugar was first added into water to increase the refractive index of sugar solution up
to 1.459 which is same as that of fused quartz. To dissolve the sugar evenly, mixer was
used to stir the sugar solution. The refractive index of sugar solution was measured using
Refractometer R5000. Refractive index of sugar solution reached 1.459 when the
concentration of sugar solution prepared was approximately 185 g/100ml of water as
shown in Figure 2.
1.46
1.459
1.44
1.42
Index
acti Ind
1.40
Refractive
1.38
Re
1.36
1.34
185
1.32
0 20 40 60 80 100 1200 140 160 180 200
Concentration of sugar solution (g/100 ml of water)
Figure 2. Graph of refractive index versus concentration of sugar solution
Next, wet pluviation method was used to place fused quartz into the solution. Fused
quartz was added slowly into solution layer by layer. According to [15], this technique
is useful to place soil at 100% degree of saturation as it creates consistent results since
terminal velocity of falling particles are achieved quickly in a denser fluid compared to
air. In the meanwhile, the air bubbles trapped within the voids were displaced as the
presence of air bubbles deteriorated the transparency of transparent soils due to
differences in refractive index value [3, 4, 7].
3. Geotechnical Engineering Modeling with Transparent Soil
3.1. Compaction Grouting
To demonstrate the feasibility of transparent soils in geotechnical modeling, compaction

grouting was modeled to observe the mechanism of grout formation. In compaction
grouting, low-mobility grouts are used to fill the cavities, sinkholes and large voids in
order to achieve the purpose of densification of soils at depth. The reduced confidence
in the reliability of compaction grouting is mainly due to the limited understanding of its
mechanism as no one actually observes the complete product of grouting as well as the
process of grout injection and formation beneath the soil surface [16]. By using
transparent soils modeling, researchers are able to visualize the grout formation inside
soil, consequently re-examine the factors affecting the effectiveness of compaction
grouting.
In our experiment, compaction grouting was modeled using a syringe as a grout tube
in drilling grouting hole and pumping grout material modeled by dyed gypsum (ratio of
water:gypsum powder = 3:10) into an acrylic glass container of 8 8 12 cm3 (length
width height).
The process of injecting grout material was depicted in Figure 3a-d and pulling out
grout tube was shown in Figure 3e-d. The grout material was initially injected
downwards (see Figure 3a-b). Parts of the growing grout segregated and dispersed into
the solution (see Figure 3c-d). Inappropriate mix of grouting material may cause the
grout to be unduly mobile or behaves as a fluid in ground, which in turn induce hydraulic
fracturing of soil as mentioned by Warner [17]; hence, the observation in Figure 3c-d
may reflect the inappropriate mixing ratio of gypsum and water and unsatisfactory
mixing method.
Slight contraction of the grout was observed when the grout tube was being pulled
out (see Figure 3e-h). The shrinkage of grout may be caused by the suction force created
during pull-out. In the field, the shrinkage of grout may cause unwanted change in
permeability of soils, which in turn influences the stability of structures as observed by
Mackey and Haskins [18]. Without visualizing the grouting mechanisms, all judgments
are merely unfounded guesses as the seemingly unimportant steps may defeat the original
purpose of compaction grouting. Here transparent soil allows researchers to understand
and re-examine the mechanism of compaction grouting.
t = 0.00 s t = 11.95 s t = 23.91 s t = 39.75 s
(a) (b) (c) (d)

t = 0.60 s t = 1.80 s t = 2.40 s t = 3.00 s
(e) (f) (g) (h)

Figure 3 a-d. Observation of grout injection at 0.00s, 11.95s, 23.91s, and 39.75s respectively; e-h.
Observation of pulling out grout tube at 0.60s, 1.80s, 2.40s, and (h) 3.00s respectively
3.2. Pile Installation
When driving pile into soil layers, soil particles are displaced resulting in ground
movements. As described by Poulos [19], these movements tend to cause destructive
consequences such as structural damage of structures nearby arising from induced
bending moments and excessive foundation settlement in post construction. The exact
nature of this mechanism is still subject to considerable debate as everything at
underground is unobservable. Transparent soil provides insights on the displacement of
soil particles in order to understand the movement of soil surrounding the pile during
installation.
An acrylic glass container of 8 8 12 cm3 (length width height) filled with
transparent soil and dyed particles were used. A brass tube of 8 mm diameter was used
to model a pile with penetration rate of 5 mm/minute for 17 minutes. A series of digital
images was taken for further analysis. Open source software Tracker [20] was used to
trace the displacement of dyed particles from front view and top view as shown in Figure
4.
FRONT VIEW TOP VIEW
Pile tube
Surface of
Pile tube
transparent soil
Traced Particle B
*Initial depth
~35.44 mm
from surface =
z Traced Particle A y ~19.41 mm
x x
(a) (b)
Figure 4 a. The extracted front view of pile installation experiment; b. The extracted top view of pile
installation experiment
Figure 5 shows the displacement of Particle A analyzed by Tracker from the front
view. Before penetrating pile into transparent soil, the pile was adjusted to the surface of
transparent soil. The initial depth of traced Particle A was about 35.44 mm from the
surface (see Figure 4a) and approximately 15 mm away from the center of pile. No
significant movement was observed before pile penetrated through the initial depth of
particle. A sudden displacement was observed when the depth of pile was 38.6 mm ~
44.2 mm, which was slightly below the initial depth of Particle A (see Figure 5c). This
sudden displacement lasted for about 80 s, which was around 520 s ~ 600 s. Referring to
Figure 5a, d, Particle A was dragged down ~ 1.24 mm (52.72 mm ~ 54.16 mm) in terms
of z-coordinate and displaced ~1.44 mm (41.48 mm ~ 42.72 mm) away from the pile in
terms of x-coordinate. After ~600 s (pile penetration of ~44.2 mm), no significant
movement of Particle A was observed.
Figure 6 shows the displacement of Particle B analyzed by Tracker from the top
view. Particle B was located at about 19.41 mm from the surface (see Figure 4b) and
about 10 mm away from the center of pile. No significant movement of Particle B was
observed before the pile depth of ~20.3 mm and after the pile depth of ~25.5 mm. Around
280 s ~ 360 s (depth of pile penetration of 20.3 mm ~ 25.5 mm), Particle B was pushed
further from the center of pile, ~0.25 mm (16.85 mm ~ 17.10 mm) in terms of y-
coordinate (see Figure 6a, c) and ~1.04 mm (46.40 mm ~ 47.44 mm) in terms of x-
coordinate (see Figure 6c, d). Surprisingly, the influenced time was about 80 s too,
similar to that of Particle A.
Figure 5. Analysis of movement of traced Particle A: a. movement of Particle A in z-coordinate over the
duration of pile installation; b. Movement of Particle A in x-coordinate over the duration of pile installation;
c. Graph showing spatial movement of Particle A in terms of x- and z-coordinate
Figure 6. Analysis of movement of traced Particle B: a. movement of Particle B in y-coordinate over the
duration of pile installation; b. Movement of Particle B in x-coordinate over the duration of pile installation;
c. Graph showing spatial movement of Particle B in terms of x- and y-coordinate
This experiment shows that transparent soil modeling is possible to quantify the 3-
dimensional spatial deformation of soil, and this quantification in turn aids in identifying
the associated displacement field developed in soil surrounding the pile during
installation and understanding the mechanism of friction fatigue in jacked piles.
4. Conclusion
Physical experimental results provide researchers direct insights in understanding the

soil behavior. As demonstrated in this project, visual monitoring of compaction grouting
helps identifying the previously unobserved steps leading to grouting failure. Also,
observing the particle movement during pile installation aids researchers in
understanding the displacement field in soil during pile installation.
By using sugar solution as pore fluid, the transparent soil becomes environmentally
friendly, non-hazardous to human health, compatible to geotechnical apparatus and
feasible in geotechnical engineering modeling. Hence, transparent soil made up of fused
quartz and sugar solution should be well quantified as this geotechnical modeling turns
out to be a timely research to allow geotechnical researches to study, observe and re-
examine the principles and mechanisms of soil behavior.
5. Acknowledgements
This research was supported by the Hong Kong Research Grants Council
(HKUST9/CRF/09 and HKUST6/CRF/12R).
References
[1] R.E. Goodman, Karl Terzaghi: The engineer as artist, American Society of Civil Engineers, 1999.
[2] R. Mannheimer, Slurries you can see through, Technology Today 3.2 (1990).
[3] M. Iskander, Modelling with transparent soils, Springer, 2010.
[4] F.M. Ezzein, R.J. Bathurst, A transparent sand for geotechnical laboratory modeling, ASTM Geotechnical
Testing Journal 34.6 (2011), 590-601.
[5] M.G. Iskander, J. Lai, C.J. Oswald, R.J. Mannheimer, Development of a transparent material to model the
geotechnical properties of soils, ASTM Geotechnical Testing Journal 17.4 (1994), 425-433.
[6] M.G. Iskander, J. Liu, S. Sadek, Transparent amorphous silica to model clay, Journal of Geotechnical and
Geoenvironmental Engineering 128.3 (2002), 262-273.
[7] Q. Ni, C.C. Hird, I. Guymer, Physical modelling of pile penetration in clay using transparent soil and
particle image velocimetry, Gotechnique 60 (2010), 121-132.
[8] S. Sadek, M.G. Iskander, J. Liu, Geotechnical properties of transparent silica, Canadian Geotechnical
Journal 39.1 (2002), 111-124.
[9] M. Ahmed, M. Iskander, Evaluation of tunnel face stability by transparent soil models, Tunneling and
Underground Space Technology 27.1 (2012), 101-110.
[10] H.C. Lo, K. Tabe, M. Iskander, S.H. Yoon, A transparent water-based polymer for simulating multiphase
flow, ASTM Geotechnical Testing Journal 33.1 (2010), 1-13.
[11] H. Downie, N. Holden, W. Otten, A.J. Spiers, T.A. Valentine, L.X. Dupuy, Transparent soil for imaging
the rhizosphere, PloS One 7.9 (2012), e44276.
[12] I.L. Guzman, M. Iskander, E. Suescun-Florez, M. Omidvar, A transparent aqueous-saturated sand
surrogate for use in physical modeling, Acta Geotechnica 9.2 (2014), 187-206.
[13] F.M. Ezzein, R.J. Bathurst, Development of a geosynthetic pullout test apparatus with transparent
granular soil, Proceedings Pan-Am CGS Geotechnical Conference, Ontario, Canada, 2011.
[14] Exxon Company, U.S.A, Norpar 12 [Material Safety Data Sheet]. Retrieved from
http://www.msdshazcom.com/MSDS/E/exxon/wcd00351.htm
[15] S.B. Peter, G. Siemens, W.A. Take, Characterization of transparent soil for unsaturated applications,
ASTM Geotechnical Testing Journal 34.5 (2011), 445-456.
[16] R.W. Boulanger, R.F. Hayden, Aspects of compaction grouting of liquefiable soil, Journal of
Geotechnical Engineering (1995), 844-855.
[17] J. Warmer, Practical handbook of grouting: soil, rock, and structures, John Wiley & Sons, 2004.
[18] S.G. Mackey, T.A. Haskins, The initial grouting in the embankment foundation at Wolf Creek Dm near
Jamestown, Kentucky, Grouting and Deep Mixing 2012 (2012), 1324-1334.
[19] H.G. Poulos, Effect of pile driving on adjacent piles in clay, Canadian Geotechnical Journal 31.6 (1994),
856-867.
[20] D. Brown, Tracker video analysis and modeling tool, (2014), Retrieved from
https://www.cabrillo.edu/~dbrown/tracker/
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-519
Local deformation measurement using

image analysis technique in hollow
cylindrical torsional shear test
Usama Juniansyah FAUZIa,1, Junichi KOSEKIb, Yukika MIYASHITAb, Ryoichiro
HOSHINOa, and Hasbullah NAWIRc
a
Department of Civil Engineering, The University of Tokyo
b
Institute of Industrial Science, The University of Tokyo
c
Department of Civil Engineering, Bandung Institute of Technology
Abstract. A special image analysis technique is presented to improve local

deformation measurement of sand specimen in hollow cylindrical torsional shear
test. This technique involves the use of a latex membrane with dots marked in a grid
pattern with approximately 5 mm spacing. These dots are tracked using a digital
camera that synchronously shoots the membrane surface at desired time intervals.
The dots on the membrane surface are tracked in terms of their center of mass using
an image processing software with the accuracy of about 0.01 mm. By comparing
image data at two different time intervals, the displacement of each dot during that
time interval is obtained. The coordinate correction is performed on the pictures to
cancel the distortion of an image by the effects of cylindrical shape of specimen, the
camera lenses, and refraction from lighting through water in the cell. The high
degree of polynomial equations with two variables are used to transform apparent
coordinates to real coordinates. Strains are calculated based on the displacement
vectors defined by the shape function matrix and the deformation matrix of the four
nodes using two-dimensional linear isoparametric rectangular elements. In order to
study liquefaction behavior, undrained cyclic loadings are conducted on Toyoura
sand prepared by air pluviation and Katori sand prepared by moist tamping and
water sedimentation. It is found that this image analysis technique and coordinate
correction are powerful tool in measuring the local deformation. It is also found that
the local deformations of specimens are different with the global one for both moist
tamping and water sedimentation specimens.
Keywords. hollow cylindrical torsional shear, image analysis, liquefaction, local

deformation
1. Introduction
It is well known in classical soil mechanics that soil specimens subjected to laboratory
testing in triaxial, plane strain, torsional, or ring shear test, are assumed as homogeneous
soil element. Thus, most simulation approaches using numerical method use the
homogenized strains on their constitutive model. However, this approach is not suitable
since localization (i.e. shear banding) is an important phenomenon in soil associated with
the micro-structural behavior. It has been known that the local soil deformation has
unique characteristics which are different from the global deformation.
1
Corresponding Author. Usama Juniansyah Fauzi; Email: [email protected]
520 U.J. Fauzi et al. / Local Deformation Measurement Using Image Analysis Technique
General principles of strain localization into development of shear band in granular

material are well understood in drained monotonic or cyclic loading tests in triaxial or
plane strain apparatus. However, little attention has been made in the study of strain
localization in hollow cylindrical torsional apparatus especially liquefaction during
undrained cyclic loading tests. One of the previous studies in hollow cylindrical torsional
apparatus is a shear band characteristic study during drained monotonic and cyclic
loading test [1]. In undrained loading test, the behavior of granular specimen during
liquefaction has been derived by assuming that the global stress-strain measurements by
the potentiometer and LVDT are representative of the deformation process and at all
points within the soil specimen.
Several experimental studies focused on localized deformation have used advanced
tools, such as stereophotogrammetry (i.e. [2] and [3]) and X-ray computed tomography
(i.e. [4]). However, digital cameras become popular nowadays due to lower price and
easy setup, for instance, image analysis techniques (i.e. [5]), digital correlation
techniques (i.e. [6]), and particle tracking velocimetry technique (PTV) (i.e. [7]), among
others. PTV method by tracking dots movement attached on membrane surface is used
in this study. PTV is faster than other image analysis methods and the membrane surface
does not have to be textured.
2. Materials and methods
2.1. Test samples and specimen preparations
Toyoura sand and Katori sand were used as the test materials. Toyoura sand particles
have an angular or sub-angular shape with the following physical properties: specific
gravity, Gs = 2.65; mean diameter, D50=0.16 mm, fines content, Fc=0.1%, maximum void
ratio, (emax) = 0.99, and minimum void ratio (emin) =0.63. Katori sand particles have mean
grain size (D50) = 0.16 mm, coefficient of uniformity (Cu) = 1.81, coefficient of gradation
(Cc) = 0.97, specific gravity (Gs) = 2.65, maximum void ratio (emax) = 1.35, minimum
void ratio (emin) = 0.92, and fines content of about 5%.
Specimens for Toyoura sand were reconstituted by using air pluviation method.
Katori sand were reconstituted by using two different techniques: water sedimentation
and moist tamping. The water sedimentation specimen was prepared initially by filling
the acrylic pipe with de-aired water. The height of water was 50 cm from the bottom. A
mixture of soil and water at a mass ratio of 1:2 was poured into the acrylic pipe using a
funnel. By repeating these procedures three times, a segregated specimen was prepared,
where the top part of fine layer in the third layer was cut so that the specimen consists of
three sand layers segregated by two fine layers. The moist tamping method was used to
reconstitute uniform specimen by ten layers.
2.2. Testing procedure for torsional test
Torsional shear tests were conducted on hollow cylindrical specimens having an initial
height of 30 cm, inner and outer diameters of 12 cm and 20 cm, respectively. Double
vacuum method [8] was applied during the saturation process, considering the large
specimen size. As a result, Skemptons B-values could be achieved to be greater than
0.96 in all the specimens. After completion of saturation, the specimens were
U.J. Fauzi et al. / Local Deformation Measurement Using Image Analysis Technique 521
consolidated isotropically to effective stress, c = 100 kPa with relative density after
consolidation, Drc = 49.5% (air pluviation) and Drc = 26.5-27.3% (moist tamping and
water sedimentation). The undrained cyclic loadings were applied with constant shear
strain rate of 1%/min with single amplitude shear stress of 15 kPa (CSR =0.15) while
maintaining the specimen height constant.
3. Image analysis technique
In this current study, the image analysis technique involves the use of a latex membrane
with dots marked in a grid pattern with approximately 5 mm spacing. During the loading
test, several photos of membrane surface were taken by using a digital camera with the
resolution of 7360x4912 pixels at desired time intervals such as 0.1 minutes to 1 minutes.
The coordinates of dots in photos were tracked by a commercial program called Move-
Tr/2D, manufactured by Library Inc.
This program reads the (m, n) coordinates of each dot, with respect to a fixed
reference system, for each photo. Coordinates of each dot are tracked by centers of
gravity movement. Each dot is identified as a moving object if the dot has contrasting
color when compared to their surroundings in the photos. Even though the dot shape
changes during the loading test, for instance from circle shapes to oval or any kind of
shapes, the dot position can be tracked because the software is able to calculate the
position by centers of mass. Fixed points had been drawn on stainless steel frame of
pedestal to make correction for relative movement changes in the position of the camera
and specimen in each photo. One of fixed points in the pedestal was also set as reference
coordinate of (0, 0). In most experiment, 28 dots in each column and 59 dots in each row
or in total 1652 dots are used.
3.1. Coordinate corrections
It should be noted that the dot coordinates are obtained from distorted image. The
distortion of an image is caused by cylindrical shape of the specimen where three-
dimensional objects are transformed into two-dimensional image. In addition, the
refraction of lighting through water, lens effect of both the camera and the cell plexy
glass, the unsymmetrical setting of camera, and inclination of camera can also cause
distortion of photos. Therefore, corrections were applied to the data obtained from digital
images to determine true specimen dimensions and dot point positions.
The first correction is inclination correction by rotating the dots position into vertical
position. Inclination of camera is corrected by
m cos sin m' (1)

n sin cos n'

where m and n are apparent coordinates after rotation and is gradient of inclination.
Then, the second correction is transforming the cylindrical shape surface of specimen to
square shape surface. In current study, a mathematical solution to restore the distorted
coordinates into 2 dimensional (2D) coordinate system is proposed. The coordinates are
corrected by fitting the dots in apparent coordinates with parabolic function. The real
coordinates are determined by transforming the measured parabolas into straight line.
Initial After correction

15 15
10 10
5 5
n' (cm)
y (cm)
0 0
-5 -5
-10 -10
-15 -15
-8 -6 -4 -2 0 2 4 6 8 -8 -6 -4 -2 0 2 4 6 8
m' (cm) x (cm)
Figure 1. Coordinates correction
The coordinate correction by polynomial nine degree by i and j indices is developed in

this study. The reasons are to increase the accuracy and the computer is capable and fast
enough to conduct the calculation. The horizontal coordinates are represented by
xi , j a0 a1.m a2 .n a3 .m 2 a4 .m.n a5 .n 2 a6 .m 3 a7 .m 2 .n a8 .m.n 2 (2)

a9 .n 3 ... a45.m 9 .... a54 .n 9
and the vertical by
yi , j b0 b1.m b2 .n b3 .m 2 b4 .m.n b5 .n 2 b6 .m3 b7 .m 2 .n b8 .m.n 2 (3)

m
b9 .n 3 ... b45.m9 .... b54.n 9
where m and x are the apparent and real coordinates, respectively, in horizontal gridline,
n and y are the apparent and real coordinates in vertical gridline, i and j are gridline
indices. The value of a0 through a54 and b0 through b54 are the real and apparent
coordinate parameters that are fitted by polynomial function. The coordinate corrections
are illustrated in Figure 1.
3.2. Strains formulation
In order to understand the local deformation of specimens during loading tests, the small
strain theory was employed, in which the displacements of any part in membrane surface
is assumed to be much smaller than global specimen. Strains are calculated at the center
of each element. The element is defined by fours nodes which are the four dots that were
imprinted on the membrane. Having determined the displacement components in both x
and y directions at each measurement point, two-dimensional linear isoparametric
rectangular elements that have been developed [9] are used. It consists of local
coordinates (Q, R) of any points within an element and global Cartesian coordinates
R
y, v
v4
v
3
u 4
u
3
(x ,y )
3 3
(-1,1) (1,1)
4 (x ,y )
4 4 3
R
Q
Q
u11 (x ,y )
1 1
2
v 1
u 2
(x ,y )
2 2
v 2
(-1,-1) (1,-1)
x, u
Figure 2. Global Cartesian coordinates and local natural coordinates for general quadrilateral element
(x, y) as shown in Figure 2.

For an element within a soil mass that deforms, the displacement at any node i are
calculated as ui = xi (xi)initial and vi = yi - (yi)initial where (xi, yi)initial are the coordinates of
node i before the beginning of loading (initial condition). The natural coordinates of the
nodes (Qi, Ri) are fixed for all elements where u and v represent horizontal and vertical
movements of any point within the element. Based on elasticity theory,
Su Sv Su Sv
x y + xy
Sx ; Sy ; Sy Sx (4)
In this research, strains were calculated at the center of each element and thus Q = 0
and R = 0. The strain-displacement matrix is shown as
' Su Sv $ SN1 SN 2 SN 3 SN 4 'u1 v1 $

' x a$ ! ! !
Sx !

1 SQ SQ SQ
!
SQ !u2 v2 !
!
Sx 1 2 & #
&
y #" &! Su #
Sv J
%b ! SN1 SN 2 SN 3 SN 4 !u3 v3 !
(5)
! Sy
% !
Sy " SR
SR SR SR !%u4 v4 !
"
+xy = (a + b) (6)
where x is horizontal strain, y is vertical strain, +xy is shear strain, u1 to u4 are

displacement of each node in horizontal direction, v1 to v4 are displacement of each node
in vertical direction, N1 to N4 are shape function, and J is the Jacobian matrix.
3.3. Accuracy of displacements and strains
In order to describe the displacements and strains with reliable results, their accuracies
have to be quantified. The errors are divided into two categories as 1) error due to
tracking of dots coordinates and 2) error due to the least mean method during coordinate
correction. To reduce the errors during dots tracking, many LED strips are installed
outside of the triaxial cell. These can improve the accuracy in observing the movements
Global and local shear strain, + (%)
12 Potentiometer (global) horizontal direction
vertical direction
8 Observation of dots 1E-3
1-R (x) and 1-R (y)

displacement (local)
4
2
1E-4
0
-4
1E-5
2
1593
-8 Average of T+ loc
i 0
+Loc (local) 1593
-12 1E-6
0 2000 4000 6000 8000 10000 2 4 6 8 10
Time (s) Degree
(a) (b)
Figure 3. Verification of image analysis (a) and accuracy of coordinate correction (b)
of the dots on the membrane surface. However, it should be noted that when the
membrane is wrinkled, the dots can be invisible so that displacements and strains at
disappeared dots are not reliable anymore.
To verify the accuracy of local displacements and strains based on dots movement
during loading tests, two method are used (Figure 3 (a)). The first involved comparing
the shear strain measured by potentiometer (global) with local shear strain measured by
a dot movement attached in top cap. In the second method, the global shear strains are
compared with average local shear strain of 1593 elements. It is found that average
standard deviation is about less than 1% and 9% for first and second method, respectively.
The accuracy of coordinate correction is defined by coefficient of determination (R2),
which is a value of how well corrected coordinate fits the target coordinate. The accuracy
of coordinate corrections in horizontal and vertical direction are shown in Figure 3 (b).
4. Results and discussions
The stress-strain relationships and local shear strain contours in case of CSR = 0.15 were
shown in Figure 4(a) for air pluviation, Figure 4(b) for moist tamping, and Figure 4(c)
for water sedimentation. The number 1, 2, 3, and 4 in these figures corresponded to the
stages where the contour plots of local shear strain were made. In air pluviation specimen,
it was found that the specimen deformed uniformly until the specimen liquefied or
effective stress was equal to zero. However, when global shear strain double amplitude,
+GDA, reached 4% (Figure 4 (a) Stage 4), several strain localizations with diagonal pattern
were observed. Similar observations were found in moist tamping and segregated
specimens. This high concentration of local shear strains in these figures was affected by
the membrane wrinkled. However, it should be noted that although strain localizations
were observed on membrane surface, those strain localizations were not shear bands.
Therefore, the deformation in air pluviation specimen was assumed uniform.
Liquefaction resistance of Katori sand specimens prepared by water sedimentation
was higher than specimens prepared by moist tamping at similar relative density, Drc,
after consolidation. Local deformation for moist tamping specimen was not uniform as
expected. As shown in Figure 4 (a,b), local deformation for Toyoura sand specimen
prepared by air pluviation was more uniform than Katori sand prepared by moist tamping.
U+LocU (%)
30
2 4
Shear stress, * (kPa)
15 10 10 10 10
0
5 5 5 5
-15
y (cm)
y (cm)
y (cm)
y (cm)
3 1 0 0 0 0
-30
-8 -4 0 4 8
Shear strain, +G (%) -5 -5 -5 -5
-10 -10 -10 -10
-5 0 5 -5 0 5 -5 0 5 -5 0 5
x (cm) x (cm) x (cm) x (cm)
Stage 1 Stage 2 Stage 3 Stage 4
(a) Toyoura sand by air pluviation (Drc = 49.5%)
U+LocU (%)
30
15 2 3 4 10 10 10 10
0
5 5 5 5
-15
y (cm)
y (cm)
y (cm)
y (cm)
1 0 0 0 0
-30
-8 -4 0 4 8 12 16
Shear strain, +G (%) -5 -5 -5 -5
non-uniform
-10 -10 -10 -10
-5 0 5 -5 0 5 -5 0 5 -5 0 5
(b) Katori sand by moist tamping (Drc = 26.5%)
U+LocU (%)
30
2 10 10 10 10
15 4
0 5 5 5 5
y (cm)
y (cm)
y (cm)
y (cm)
-15
3 1 0 0 0 Fine0layer
-30
-16 -12 -8 -4 0 4 8 12 16
-5 -5 -5 -5
Shear strain, +G (%)
-10 -10 -10 -10
-5 0 5 -5 0 5 -5 0 5 -5 0 5
(c) Katori sand by water sedimentation (Drc = 27.3%)
Figure 4. Local shear strain plot for (a) air pluviation, (b) moist tamping, and (c) water sedimentation
At stage 2, 3, and 4, the local shear deformation in some part of the upper half was larger
than other parts as shown in Figure 4(b). It should be noted that specimen prepared by
moist tamping was prepared in ten layers. Therefore, relative densities at some layers
were smaller than average global relative density.
For the test with specimen prepared by water sedimentation, the local deformation
of segregated specimen was also not uniform. However, localization appeared clearly
around the segregated zone. At stages 2, 3, and 4, the local shear strain around fine layer
was observed higher than those at other parts. This condition was due possibly to the
effect of presence of fine layer. The vertical expansion was also observed near fine layer.
This behavior may be caused by at least three factors, first due to horizontal membrane
wrinkle (local drainage), second due to water film formation, and third due to pore water
migration. It should be noted that in this study, fine has higher liquefaction resistance
than sand. Therefore, excess pore water would flow from sand layer to fine layer. It
would result in loosening of fine layer due to increase of pore water pressure in fine layer.
5. Conclusions
We have developed an image analysis technique that uses a latex membrane with dots
marked in a grid pattern with approximately 5 mm spacing and equations for distortion
coordinates. The technique can be employed to measure the local deformation. It is found
that air pluviation specimen can reconstitute specimen in more homogenous compare to
moist tamping and water sedimentation specimens. The water sedimentation specimen
shows higher liquefaction resistance than moist tamping specimen. This fact indicates
that the in-situ liquefaction resistance of sand deposits could be underestimated by
laboratory testing if uniform specimens are used. In water sedimentation, accumulation
of local shear strains was observed at fine layers. This behavior might be caused by local
drainage, a water film formation, and excess pore water migration.
References
[1] Wahyudi, S., Miyashita, Y., and Koseki, J. (2013). Observation of Shear Banding Characteristics of Sand
in Torsional Shear Test Using Image Analysis Technique. In Multiphysical Testing of Soils and Shales (pp.
201-206). Springer Berlin Heidelberg.
[2] Yoshida, T. (1994). Shear banding in sands observed in plane strain compression. In Proc. Symp. on
Localization and bifurcation Theory for Soils and Rocks (pp. 165-179). Balkema.
[3] Desrues, J., and Viggiani, G. (2004). Strain localization in sand: an overview of the experimental results
obtained in Grenoble using stereophotogrammetry. International Journal for Numerical and Analytical
Methods in Geomechanics, 28(4), 279-321.
[4] Han, C., and Vardoulakis, I. G. (1991). Plane-strain compression experiments on water-saturated fine-
grained sand. Geotechnique, 41(1), 49-78.
[5] Lin, H., and Penumadu, D. (2006). Strain localization in combined axial-torsional testing on kaolin clay.
Journal of engineering mechanics, 132(5), 555-564.
[6] Rechenmacher, A. (2005). Imaging-based experimental soil mechanics. In Geomechanicss Testing,
Modeling, and Simulation (pp. 653-663). ASCE.
[7] Sadrekarimi, A., and Olson, S. M. (2009). Shear band formation observed in ring shear tests on sandy soils.
Journal of geotechnical and geoenvironmental engineering, 136(2), 366-375.
[8] Ampadu, S.K. and Tatsuoka, F. (1993). Effect of setting method on the behaviour of clays in triaxial
compression from saturation to undrained shear. Soils and Foundations, 33(2), 14-34.
[9] Salas-Monge, R. (2002): Effects of large amplitude cyclic loading on deformation and strength properties
of cement treated sand, Master of Engineering Thesis, Department of Civil Engineering, The University
of Tokyo, Japan
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-527
Strain-Dependent Damping Ratio of

Compacted Gravel
Tianfei LIAO a,1 , Michael McHOOD a, Nasser MASSOUDI a,
Kenneth H. STOKOE b and MinJae JUNG b
a
Bechtel Corporation, Frederick, MD, USA
b
University of Texas, Austin, TX, USA
Abstract. Compacted gravel is often used as engineered fill to provide required

bearing capacity for structures. The dynamic properties of the gravel fill, such as
strain-dependent shear modulus and material damping ratio, are required in
seismic analyses of these systems. Strain-dependent damping ratio as a function of
cyclic shear strain were obtained from a series of combined Resonant Column and
Torsional Shear (RCTS) tests on two types of crushed gravel fill. The material
damping ratio curves are presented and compared to empirical relationships in the
literature that have been proposed for gravelly soils.
Keywords. Material Damping Ratio, Gravel, Engineered Fill, Resonant Column,

Torsional Shear, Strain-Dependent
1. Introduction
Compacted gravel is frequently used as engineered fill beneath the foundations of

critical and/or important structures, such as nuclear power facilities and high-rise
buildings. To evaluate the seismic response of the ground supporting these structures,
the shear modulus (G) and material damping ratio (D) need to be determined as a
function of shear strain ().
It is very difficult to obtain the strain-dependent G and D directly from in-situ tests
[1]. In current engineering practice, the effects of effective confining pressure ('0) and
shear strain () on G and D are primarily evaluated through laboratory tests, such as
cyclic triaxial (CTX), cyclic simple shear (CSS), cyclic torsional shear (TS), and
resonant column (RC) tests. These tests not only give the values of G and D at small
strains, but also yield the variation of G and D with and '0. However, such tests are
rarely performed on gravels, due to the large size of the testing apparatus required to
test representative specimens. Additionally, because it is difficult to obtain undisturbed
samples of gravelly soils, tests on natural gravel deposits have typically been limited to
high-quality undisturbed gravel samples obtained by in-situ ground freezing, a very
expensive sampling method (e.g., [2]; [3]; [4]; [5]). The strain-dependent D of two
types of crushed gravel are discussed in this paper. The strain-dependent G of these two
types of crushed gravel has been addressed previously [6].
1
Corresponding Author: Bechtel Infrastructure Corporation, Frederick, MD 21703, USA; E-mail:
[email protected]
528 T. Liao et al. / Strain-Dependent Damping Ratio of Compacted Gravel
2. Literature Review
2.1. Nonlinear Material Damping Ratio
After removing (or scalping) particles greater than 51 mm in diameter, a series of

cyclic triaxial tests were performed on 305-mm diameter specimens of four different
types of well-graded gravels [7], which were isotropically-consolidated and tested
under undrained cyclic loading. Based on the test results, the D - log() curves of
gravelly soils are in the range shown in Figure 1. It was suggested by Seed et al. [7]
that the D - log() curves for gravels fall in the same range as that of sands. It was also
suggested that dense material have higher values of D than loose material.
By analyzing the test results for gravelly soils mainly found in literature, a slightly
different range for the D - log() curves was suggested by Rollins et al. [8] for gravelly
soils (Figure 1). Most of the data used by Rollins et al. [8] came from cyclic triaxial
tests (CTS) typically performed on specimens that were 300 mm in diameter and 600
mm in height, with a small number of cyclic torsional simple shear tests performed on
specimens of larger diameters. All tests reviewed by Rollins et al. [8] were performed
in various literatures around the world and were carried out at a loading frequency in
the range of about 0.01 Hz to 0.2 Hz.
30
Average for Sand/Gravel
(Seed et al., 1986)
25
Variation Range for Gravel
Damping Ratio, D (%)
(Rollins et al., 1998)

20
15 Variation Range for

Sand/Gravel (Seed et al., 1986)
10
5
Average for Gravel
(Rollins et al., 1998)
0
0.0001 0.001 0.01 0.1 1
Cyclic Shear Strain, (% )
Figure 1. Typical D - log() relationships for gravels suggested by Seed et al. [7] and Rollins et al. [8].
2.2. Relationship between G/Gmax and D
Material damping ratio (D) increases with the decrease of G. The value of D was
assumed to be proportional to (1-G/Gmax) by Hardin & Drnevich [9], but it was
suggested by Tatsuoka et al. [10] that although D increases with (1-G/Gmax), this
relationship is slightly nonlinear. The relationship between D and G/Gmax is unique for
a sand type under a specific number of loading cycles, and is not affected by void ratio,
confining pressure, or shear strain amplitude [10].
The value of D is associated with G/Gmax in Ishibashi & Zhang [11] and Borden et
al. [12] using a polynomial expression, mainly based on torsional shear test results on
sandy soils. Based on test data available in the literature, the following best-fit
relationship is suggested for sandy soils [11]:
D = 0.333[0.586 (G / G max ) 2 1.547 (G / G max ) + 1] (1)

T. Liao et al. / Strain-Dependent Damping Ratio of Compacted Gravel 529
Modeling the hysteretic damping assuming a modified Masing behavior, the

damping equation of D=f(G/Gmax)+ Dmin was suggested by Darendeli [13], where
f(G/Gmax) is a function of G/Gmax. By fitting this equation to test data compiled from
the literature for both cohesive and cohesionless soils, the suggested best-fit curve
between D-Dmin and G/Gmax is as follows [14]:
For TS data, D Dmin = f (G / G max ) = 10.6(G / G max ) 2 31.6(G / G max ) + 21.0 (2)
For RC data, D Dmin = f (G / G max ) = 9.4(G / G max ) 2 26.5(G / G max ) + 17.1 (3)
3. Tested Material
Two types of gravel were tested for potential use as engineered fill in this study. Both
gravel types were composed of aggregates derived from processing crushed stone
mined from a rock quarry, and consist of hard angular particles, with one of the
materials (Uniformity coefficient, Cu = 2.1; Grain diameter corresponding to 50%
passing, D50 = 11.8 mm) being poorly graded and designated as PA and the other (Cu =
150 - 170; D50 = 3.2 - 3.4 mm) being relatively well graded and designated as WA.
Three batches of the WA material (WA-1, WA-2, and WA-3) and one batch of the PA
material (PA-1) were tested. Further information about the tested material can be found
in Liao et al. [6].
4. Testing Procedure
During RCTS testing, the specimen is sealed in a membrane, and the pore pressure in
the specimen is vented to atmosphere pressure. The results of cyclic triaxial tests on
Toyoura sand show that the drained tests and the undrained tests give almost identical
strain-dependent variation of the damping ratio within strain level from 10 -4% to 0.5%
[16]. Since the gravel specimens have larger permeability due to the larger grain sizes
and the maximum shear strain reached in the RCTS tests was less than 0.5%, the effect
of the drainage condition did not affect the measured dynamic properties.
For each gravel specimen, combined Resonant Column and Torsional Shear
(RCTS) testing was performed at four effective isotropic confining pressures (0) (i.e.,
52, 207, 414, and 827 kPa). Two PA specimens (i.e., PA-1-A and PA-1-B) were tested.
The membranes around these specimens were punctured by the test material when the
confining pressure was increased to 414 kPa, and thus no further testing was carried out.
Six WA specimens (WA-1-A, WA-1-B, WA-2-A, WA-2-B, WA-3-A, and WA-3-B,)
were tested. The RCTS tests were also performed on some of the WA specimens when
the confining pressure was unloaded to lower values (i.e. 207 kPa and then 52 kPa).
The TS test is performed at a loading frequency of 0.5 Hz, while the RC test is
carried out around the resonant frequency of the specimens (typically in the range of 30
to 200 Hz). With the RC test, the material damping is derived from the decaying
pattern of the free-vibration curve for the first three cycles of free vibration [15], and
the corresponding shear strain is calculated from the average amplitude of these three
cycles of vibration. Further information about the testing procedure can be found in
Liao et al. [6].
5. Test Results
5.1. Measured Nonlinear Damping Ratio of Compacted Gravel
The measured D values at different confining pressures are presented in Figure 2 for
the PA specimens and in Figure 3 for the WA specimens. The circular and triangular
symbols represent the measured data points from RC and TS tests, respectively. In
addition, the symbols with -UL in the figure legend represent the results of tests
performed when the confining pressure is unloaded to lower values. For the PA
material, the RC test results and the TS test results under the same confining pressure
are presented in the same figure [i.e., Figures 2(a) and (b)]. While for the WA samples,
the D measured from the RC tests and the TS tests are presented separately in Figure 3.
14 14
12 Confining Pressure = 52 kPa 12 Confining Pressure = 207 kPa

10 10
w = 0.8 - 1.0% PA-1-B PA-1-B
w = 0.8 - 1.0%
RC RC
8 PA-1-A 8 PA-1-A
PA-1-B PA-1-B
6 Dr=80% TS 6 TS
PA-1-A
Dr=80% PA-1-A
4 4
2 2
Dr=100% Dr=100%
0 0
0.0001 0.001 0.01 0.1 1 0.0001 0.001 0.01 0.1 1
Cyclic Shearing Strain, (%) Cyclic Shearing Strain, (%)
(a) PA specimens, 52 kPa, RC&TS tests (b) PA specimens, 207 kPa, RC&TS tests
Figure 2. D - log() curves for PA specimens determined by RCTS testing.
Figures 2 and 3 show, for both the PA and the WA material, the D - log() curve
shifts to lower values of D with increasing confining pressure, which is typical for
granular soil. At the same confining pressure (i.e., 52 kPa or 207 kPa), both the RC test
and TS test give similar D - log() curves for the PA specimens (Figures 2 (a) and (b)),
showing that the effect of frequency is not significant on the D - log() curves of PA
specimens. The primary reason for this close comparison is believed to be due to the
low water content ( 1%) of the material.
The D - log() curves from RC and TS tests are quite different for the WA material
as seen by comparing Figures 2 and 3. Due to the moisture in the WA samples (and not
the PA samples), the D - log() curves of the WA specimens from the RC tests are
higher than those from the TS test at the same confining pressure. Differences in
moisture content (w) seems to lead to noticeable differences in D - log() curves
measured from RC tests. Based on only the RC test results shown in Figure 3, the D of
sample WA-2 is generally slightly lower than that of samples WA-1 and WA-3 at the
same . This difference is likely attributed to the relatively lower moisture content in
WA-2. This difference is consistent with the conclusion from Menq [15] that D of
gravelly soil increases with increasing moisture content. However, this difference is not
apparent in the TS results, confirming that water content is less significant on D as the
loading frequency decreases [15]. As with the G/Gmax - log() curves described in Liao
et al. [6], the density of the specimens has little effect on the D - log() curves, which is
consistent with the observation of Ishihara [1] on the relationship between the D and
shear strain for sand.
14 14
Confining Pressure = 52 kPa
12 12 Confining Pressure = 52 kPa
WA-1-A
WA-1 & WA-3
WA-1-A-UL

10 10 WA-1-A
(w = 5.8-6.4% ) WA-1-B
WA-1-B
WA-1-B-UL
8 8 WA-2-A
RC WA-2-A TS
WA-2-B
WA-2-B
6 6
WA-2-B-UL WA-3-A
WA-3-A WA-3-B
4 4
WA-3-B
2 WA-2 2
(w = 4.4-5.5% )
0 0
0.0001 0.001 0.01 0.1 1 0.0001 0.001 0.01 0.1 1
Cyclic Shearing Strain, (% ) Cyclic Shearing Strain, (% )
(a) WA specimens, 52 kPa, RC tests (b) WA specimens, 52 kPa, TS tests

14 14
12 WA-1-A 12 Confining Pressure = 207 kPa
WA-1-A-UL
10 10 WA-1-A
WA-1-B
WA-1-B
WA-1 & WA-3 WA-1-B-UL
8 8 WA-2-A
(w = 5.8-6.4% ) RC WA-2-A TS
WA-2-B
WA-2-B
6 6
WA-2-B-UL WA-3-A
WA-3-A WA-3-B
4 4
WA-3-B
2 WA-2 2
(w = 4.4-5.5% )
0 0
0.0001 0.001 0.01 0.1 1 0.0001 0.001 0.01 0.1 1
(c) WA specimens, 207 kPa, RC tests (d) WA specimens, 207 kPa, TS tests
14
12
10
WA-1-A
8
WA-1 & WA-3
WA-2-A
(w = 5.8-6.4% ) RC
6 WA-3-A
WA-3-B
4
2 WA-2
(w = 4.4-5.5% )
0
0.0001 0.001 0.01 0.1 1
Cyclic Shearing Strain, (% )
(e) WA specimens, 414 kPa, RC tests

14 14
12 12 Confining Pressure = 827 kPa
10 10 WA-1-A
WA-1-A
WA-1-B
WA-1-B
8 8 WA-2-A
WA-1 & WA-3 WA-2-A TS
(w = 5.8-6.4% ) RC WA-2-B
WA-2-B
6 6
WA-3-A WA-3-A
WA-3-B WA-3-B
4 4
2 WA-2 2
(w = 4.4-5.5% )
0 0
0.0001 0.001 0.01 0.1 1 0.0001 0.001 0.01 0.1 1
(f) WA specimens, 827 kPa, RC tests (g) WA specimens, 827 kPa, TS tests
Figure 3. D - log() curves for WA specimens determined by RCTS testing.
Figure 4 (a) presents the average D measured from TS tests for PA specimens and
WA specimens. Under the same confining pressure, the D of the PA material is lower
than that of the WA material, which is consistent with a key observation of Menq [15]
that at a given shear strain, D - log() curves shifts to lower D values as the uniformity
coefficient Cu decreases.
The average D measured from TS tests is also compared to the variation range
proposed by Seed et al. [7] and Rollins et al. [8] in Figure 4. The comparison in Figure
4 (a) shows that the D - log() curves for both WA and PA fall in the range proposed by
Seed et al. [7], while only the D - log() curves at strains above 0.003% for the PA
specimens are within the upper range proposed by Rollins et al. [8].
In Figure 5(b), the average D measured from RC tests for sample WA-2 is
compared to the average D of the other two WA samples (WA-1 and WA-3). The
average D for sample WA-2 is lower than that of samples WA-1 and WA-3 under the
same confining pressure. As discussed previously, this difference is attributed to the
lower moisture content of the WA-2 sample. As expected, comparison shows the D -
log() curves from RC tests are higher than the variation range proposed by Seed et al.
[7] and Rollins et al. [8], due to the moisture content in the test specimens as well as
the fact that the RC tests are performed at much higher frequencies than the laboratory
cyclic tests that are the basis for the proposed ranges. Comparison also shows that the
difference between the measured D and the values suggested by Seed et al. [7] or
Rollins et al. [8] becomes less significant as shear strain increases, because the viscous
portion of D, which can be significant at very small strains, becomes a small part of D
at larger strains where hysteretic damping dominates.
To further examine the difference between D measured by RC and TS tests, the
average D measured from TS tests for the WA material are compared to the average D
of WA-2 sample measured from RC tests in Figure 4(c). At the same confining
pressure, the average D of WA-2 from RC tests is higher than that of the average WA
material measured by TS tests. This difference is more significant at lower confining
pressure and at lower shear strain.
5.2. Relationship Between D and G/Gmax for Compacted Gravel
The measured D values can be plotted against the corresponding G/Gmax values
measured at the same strain level and at the same confining pressure, which are not
shown here due to limits set for paper length. By extrapolating the D - G/G max curves to
G/Gmax =1, the Dmin value can be derived for each set of D - G/Gmax data.
Then the (D-Dmin) value can be plotted against G/Gmax. Figure 5(a) and Figure 5(b)
present the RC data and TS data, respectively, for all the performed tests. The (D-D min)
- G/Gmax curves based on TS tests are generally higher than those based on RC tests,
but they each fall in a relatively narrow band. The best fit curves suggested by Zhang et
al. [14] [i.e., Equation (2) and (3)] are also plotted on Figure 5. It can be seen that the
suggested curves fit well the test data from this study.
25 25
Average (Seed et al., 1986)
Variation Range (Seed et al., 1986)
20 20
Average (Rollins et al., 1998)
Variation Range (Rollins et al., 1998)
Damping, D (%)
Damping, D (%)
15 15
WA-1 & WA-3 (RC, 52 kPa)
WA-1 & WA-3 (RC, 207 kPa)
WA-1 & WA-3 (RC, 414 kPa)
10 WA (TS, 52 kPa) 10 WA-1 & WA-3 (RC, 827 kPa)
WA (TS, 207 kPa) WA-2 (RC, 52 kPa)

5 5
PA (TS, 52 kPa) WA-2 (RC, 414 kPa)
WA-2 (RC, 827 kPa)
PA (TS, 207 kPa)
0 0
0.0001 0.001 0.01 0.1 1 0.0001 0.001 0.01 0.1 1
Cyclic Shearing Strain, (% ) Shearing Strain, (% )
(a) Average curves from TS tests (b) Average curves from RC tests
of PA and WA samples of WA samples
25
20

Damping, D (%)
15
WA (TS, 52 kPa)
WA (TS, 207 kPa)

10
WA-2 (RC, 52 kPa)

5
WA-2 (RC, 414 kPa)
WA-2 (RC, 827 kPa)
0
0.0001 0.001 0.01 0.1 1
Cyclic Shearing Strain, (% )
(c) Average curves for WA based on TS tests compared to average curves for WA-2 based on RC tests.
Figure 4. Comparison between D - log() curves for compacted gravel in this study and gravel curves
in the literature.
8 8
Zhang et al (2005) Zhang et al (2005)
7 7
Damping, (D - Dmin ) (%)
6
Damping, (D - D min ) (%)
5 RC Tests 5 TS Tests
4 4
3 3
2 2
1 1
0 0
0.4 0.5 0.6 0.7 0.8 0.9 1 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized Shear Modulus, G/Gm ax Normalized Shear Modulus, G/Gm ax
(a) RC tests (b) TS tests

Figure 5. (D Dmin) - G/Gmax curves for the specimens subject to RCTS tests.
6. Conclusions
Combined RCTS tests were performed on two types of compacted, crushed gravel
produced in a rock quarry, with one gravel (Cu = 2.1; D50 = 11.8 mm) being poorly-
graded and the other (Cu = 150 - 170; D50 = 3.2 - 3.4 mm) being well graded.
The density of each type of specimen has little to no effect on the D - log() curves.
The loading frequency has no significant effect on the D - log() curves of the poorly-
graded gravel due to the low water content. For the well-graded gravel, the D - log()
curves from the RC tests are higher than those from the TS tests at the same confining
pressure due to the higher water content. This upward shifting of the D - log() curves
of the well graded gravel in the RC tests relative to the TS tests has to do with water
movement in the void space at higher frequencies.
The D measured by RCTS tests for the poorly-graded gravel is lower than that for
the well-graded gravel under the same confining pressure. Based on the TS results, the
D - log() curves for both well-graded gravel and poorly-graded gravel agree well with
the typical range suggested by Seed et al. [7], and the D - log() curves for the poorly-
graded gravel are within the range recommended by Rollins et al. [8]. However, it is
important to note that the effects of confining pressure and grain size distribution are
neglected in each set of the published curves ([7] and [8]).
The (D-Dmin) - G/Gmax curves based on TS tests are generally higher than those
based on RC tests, and the best fit curves suggested by Zhang et al. [14] [i.e., Equations
(2) and (3)] fit well the test data from this study.
References
[1] K. Ishihara, Soil Behavior in Earthquake Geotechnics, Oxford Science Publications, 350, 1996.
[2] S. Goto, Y. Suzuki, S. Nishio and H. Oh-oka, Mechanical properties of undisturbed tone-river gravel
obtained by in-situ freezing method, Soils and Foundations, 32 (1992), 1525.
[3] S. Goto, S. Nishio and Y. Yoshimi, Dynamic properties of gravels sampled by ground freezing, Ground
Failures Under Seismic Conditions, GSP No. 44, ASCE, 141157, 1994.
[4] M. Hatanaka and A. Uchida, Effects of test methods on the cyclic deformation characteristics of high
quality undisturbed gravel samples, Static and Dynamic Properties Of Gravel Soils, GSP No. 56,
ASCE, 136151, 1994.
[5] T. Kokusho and Y. Tanaka, Dynamic properties of gravel layers investigated by in-situ freezing sampling,
Ground Failures under Seismic Conditions, GSP No. 44, ASCE, 121140, 1994.
[6] T. Liao, N. Massoudi, M. McHood, K.H. Stokoe, M.J. Jung, and F.-Y. Menq, Normalized shear modulus
of compacted gravel, Proceedings, 18th International Conference on Soil Mechanics and Geotechnical
Engineering, Paris, France, 2013.
[7] H.B. Seed, R.T. Wong, I.M. Idriss, and K. Tokimatsu, Moduli and damping factors for dynamic analyses
of cohesionless soils, Journal of Geotechnical Engineering, ASCE, 112 (1986), 1016-1032.
[8] K.M. Rollins, M. Evans, N. Diehl and W. Daily, Shear modulus and damping relationships for gravels,
Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 124 (1998), 396-405.
[9] B.O. Hardin, and V.P. Drnevich, Shear modulus and damping in soils: design equation and curves,
Journal of the Soil Mechanics and Foundation Division, ASCE, 98 (1972), 667 692.
[10] F. Tatsuoka, T. Iwasaki, and Y. Takagi, Hysteretic damping of sands and its relation to shear modulus,
Soils and Foundations, 18(1978), 25-40.
[11] I. Ishibashi, and X. Zhang, Unified dynamic shear moduli and damping ratios of sand and clay, Soils
and Foundations, JSSMFE, 33(1993), 182-191.
[12] R. Borden, L. Shao, and A. Gupta, Dynamic properties of Piedmont residual soils, Journal of
Geotechnical Engineering, ASCE, 122(1996), 813821.
[13] B.M. Darendeli, Development Of A New Family Of Normalized Modulus Reduction And Material
Damping Curves, Ph. D. Dissertation, Univ. of Texas at Austin., TX, USA, 362, 2001.
[14] J. Zhang, R.D. Andrus, and C.H. Juang, Normalized shear modulus and material damping relationships,
Journal of Geotechnical and Geoenvironmental Engineering, 131 (2005), 453-464.
[15] F.Y. Menq, Dynamic Properties Of Sandy And Gravelly Soils, Ph.D. Dissertation, University of Texas
at Austin, TX, USA, 364, 2003.
[16] T. Kokusho, Cyclic triaxial test of dynamic soil properties for wide strain range, Soils and Foundations,
20 (1980), 45-60.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-535
Experimental assessment of the influence

of load-induced deformation on
interparticle contacts
Masahide OTSUBO,1 Way Way SIM and Catherine OSULLIVAN
Imperial College London, UK
Abstract. A comprehensive understanding of soil response to applied loads and

deformations requires an understanding of the particle-scale interactions. The
overall stiffness can be related to the mean number of particle-particle contacts
(mean coordination number, NC), the void ratio (e), the particle characteristics and
the confining pressure. While most of these parameters can be measured, it is
difficult to determine the mean coordination number directly from a laboratory test.
The influence of the confining pressure on particle-particle contacts and therefore
the e-NC relationship is poorly understood. This paper presents experiments using
ink to visualize particle contact points and to relate the mean coordination number
to the void ratio (e-NC relationship) under different vertical stresses.
Keywords. Coordination number, oedometer, micromechanics, DEM
1. Introduction
Full, fundamental understanding of the mechanical response of soil requires an

understanding of the particle-scale interactions. While discrete element method (DEM)
simulations are very useful to develop an understanding of the precise nature of the
relationship between particle-particle interactions and the overall response, DEM
models inevitably involve simplifications. Direct measurements of the contact points
and the evolution of the contact surfaces are needed to confirm and supplement
conclusions developed from these idealized models.
Analysis of 3D images of soil from micro-computed tomography (CT) can be
used to relate the coordination number and the void ratio or packing density (e.g.
Fonseca et al. [1]). The interpretation of CT data is non-trivial as the quality of the
data is influenced by the scan resolution and the ability to accurately threshold the void
and particle phases in the material. CT scan data cannot provide information on the
contact surfaces with the nm level of resolution needed to provide meaningful insight.
This gap in available information can be filled by creating dry samples in the
laboratory, impregnating the samples with a tracer ink that can mark the contact points
and allow subsequent detailed examination, e.g. Oda [2].
This study aims to examine the influence of confining pressure on contacts by
extending prior research where contact points have been directly measured using a
tracer ink. The first part of this paper highlights how inter-particle contacts are related
1
536 M. Otsubo et al. / Experimental Assessment of the Inuence of Load-Induced Deformation
to the behavior of soil and the second part addresses experimental assessment of ink
tests using an oedometer apparatus.
2. Inter-particle contacts and soil stiffness
2.1. Soil stiffness and coordination number
The load-deformation relationship at the particle scale is key in evaluation of the small
strain stiffness of soil. The small strain stiffness (e.g. sample shear modulus, G0) of an
isotropically loaded, random assembly of homogeneous spheres can be estimated using
the static assumption of Chang & Liao [3]. Yimsiri & Soga [4] derived the following
expression using Hertz & Mindlin contact model (Mindlin [5]), where the partial slip
due to tangential contact forces was considered negligible:
1 2
1 1
5 NC 3 3
G0 12G P2 1 P 3 ' (1)
210 7 P 1 e
where P = particle Poissons ratio, GP = particle shear modulus, e = void ratio, NC =

mean coordination number (mean number of contacts per particle) and = isotropic
stress. Eq. (1) assumes that e and NC are independent parameters which can be
determined from state of packing. While the void ratio can be obtained from laboratory
tests, the determination of mean coordination number is more difficult.
2.2. Experimental study to visualize contact points
Establishing a relationship between the mean coordination number and the void ratio
enables the soil stiffness at small strains to be estimated using available information
from laboratory tests. Prior experimental attempts to develop such a relationship
include Smith et al. [6] who visualized the contact points between lead shot spheres
with diameters of 7.56 mm at different packing densities using 20% solution of acetic
acid and found that densely packed samples have a higher NC. Bernal & Mason [7]
used black Japanese ink to identify contact points on ball bearings with diameters of
6.35 mm, they submerged the particles in ink, then allowed the ink to drain; as
discussed below this process leaves circular marks on the particles at the contact points.
Ink menisci formed between particles that were close to each other but not actually
contacting, these can be considered as virtual contacts (termed near by Bernal &
Mason), Bernal & Mason identified these as cases where the menisci radii were less
than 5% of the particle radius. Engaged contacts are those contacts that actually
transmit force (termed close by Bernal & Mason) and Bernal & Mason hypothesize
that in this case the radii of the menisci exceeded 5% of the particle radius. Oda [2] also
used the Japanese ink method to study contact points on glass ballotini and concluded
that the coordination number frequencies in a random assembly of monosized spheres
can be represented by a Gaussian distribution. Pinson et al. [8] considered mixtures of
two distinct sizes of steel spheres and found that the mean coordination number
depends on the void ratio rather than the particle size distribution.
M. Otsubo et al. / Experimental Assessment of the Inuence of Load-Induced Deformation 537
While the experiments in the literature are useful to investigate the state of packing
in a sample, no study has yet been carried out to investigate the influence of confining
pressure on the relationship between e and NC. This study aims to develop our
understanding of how the e-NC relationship evolves due to K0 compression.
3. Experimental method to visualize inter-particle contact points
3.1. Sample preparation and oedometer apparatus
Borosilicate glass beads (ballotini) were used to prepare a cylindrical sample of 38 mm

in diameter and 19 mm in height. The mean RMS roughness of the ballotini surfaces
was 73 nm based on 40 measurements considering an area of 7070 m2 using an
optical interferometer (Otsubo et al. [9]). The ballotini diameters ranged from 1.00 mm
to 1.18 mm, with an average diameter of 1.09 mm. These particles are smaller than
those used in the earlier studies mentioned above (where the diameters were between
6.35 and 12.7 mm). The mean values of shape parameters of the ballotini were obtained
using an automatic shape analyzer (QicPic): sphericity = 0.946, aspect ratio = 0.981
and convexity = 0.984 as described by Altuhafi & Coop [10] and Cavarretta et al. [11],
which indicate the highly spherical nature of the test material. Dry dense samples were
prepared by pouring the ballotini in 10 separate layers with each layer densified by
tapping the side walls of the mold with a rubber hammer. The initial void ratios were
obtained at vertical stresses of 11 kPa, 44 kPa and 58 kPa for test cases 1, 2 and 3,
respectively. Further vertical load was applied to the sample and creep was allowed to
take place before impregnating the ink. The initial void ratio and the void ratio at the
impregnation of the ink are listed in Table 1 together with their stress levels.
Figure 1. Ink in oedometer apparatus. Figure 2. Void ratio and vertical stress using oedometer.
Table 1. Tested cases using oedometer apparatus

Initial void ratio Void ratio at test Mean coordination
Test case
(initial vertical stress) (vertical stress at test) number, NC
1 0.593 (11 kPa) 0.593 (11 kPa) 6.87
2 0.594 (44 kPa) 0.585 (703 kPa) 7.29
3 0.596 (58 kPa) 0.582 (2010 kPa) 7.39
3.2. Impregnation and draining of ink
The experimental approach followed Oda [2]. A Japanese ink (Bokuju) was used as a
tracer ink. The ink solution was empirically diluted with distilled water to reduce
viscosity and enable flow through the sample voids. The sample was slowly
impregnated with ink through the base porous stone while maintaining the vertical
stress constant. To ensure high degrees of saturation, the ink was left in the sample for
at least 5 hours prior to being drained out. Figure 1 shows the oedometer apparatus
with the sample saturated in the ink solution, awaiting drainage. The ink was drained
from the base of the apparatus while maintaining the vertical load applied during
impregnation. Once drained, the sample was unloaded in a single increment, leaving
the mass of the top cap as the only applied load acting on the sample. As the sample
remained confined by the top-cap, is was assumed that the fabric did not differ
significantly from the fabric under the previously applied load. The sample was then
dried in an oven at 105 C for 24 hours together with the mold and the porous stones so
that the sample packing was not disturbed. After the sample was dried, random
ballotini were selected for imaging from the central section of the sample in order to
avoid wall effects. 150 particles were considered to get reliable data on the
coordination number frequencies; Pinson et al. [8] recommended 100 for monosized
spheres.
4. Experimental observation
The relationship between void ratio (e) and the vertical stress (v) observed for the
tested ballotini using an oedometer apparatus is plotted in Figure 2. The point of ink
impregnation and drainage for each test are indicated with an arrow. Representative
images of contact points printed on tested ballotini are shown in Figures 3, where the
virtual contacts were indicated by a broken circle and the center of the engaged
contacts were indicated by an arrow. When two particles were in contact, an annulus of
ink was observed around the point of contact due to the formation of a capillary barrier.
When two or more contacts were located close to each other, surface of the capillary
barriers at neighboring contacts connected forming a bridge (Figure 3). These
bridges between neighboring contact points have not been previously reported in the
literature and they may be related with the small void space between the tested spheres
as the size of ballotini were smaller than those used in prior research. Contacts were
defined to be virtual if a junction of bridges were formed without showing a clear
shape of annulus of contact at the junction. The virtual contacts found in test case 1 (v
= 11 kPa), e.g. contacts denoted 1 and 2 in Figure 3a, were rarely observed in the
other cases as the vertical stress increased. The contact denoted 3 in Figure 3b was
considered to be engaged. Bridges were also clearly observed inside the mold (test case
3) where wall-particle contact points were traced (Figure 4). It is also clear that the
contact points were not equally distributed on the mold due to the nature of random
packing.
(a) (b) (c)

Figure 3. Print of contact points on ballotini: (a, b) visual and engaged contacts from test case 1 (11 kPa) and
(c) engaged contacts from test case 2 (703 kPa) (Black arrows indicate the center of engaged contacts).
(a) (b)
Figure 4. Print of contact points on wall after test case 3 (2010 kPa): (a) magnified and (b) view of the mold.
The coordination number frequency obtained from 150 arbitrary particles for each
test case is summarized in Figure 5. In agreement with Oda [2], the observed
frequencies of coordination number for this random packing approximately follow a
Gaussian distribution, particularly for the lowest stress level (test case 1). For test case
1 the mean coordination number was 6.87 at a void ratio of 0.593. As the vertical stress
increased to 703 kPa and 2010 kPa, the mean coordination numbers increased to 7.29
and 7.39, respectively at void ratios of 0.585 and 0.582. Considering the fact that
virtual contacts were rarely observed at the higher stress levels, the change of
coordination number probably resulted from previously virtual contacts becoming
engaged (Figure 3a). As the vertical stress increased, the distribution lost its symmetric,
Gaussian shape and the frequency of contacts with higher coordination numbers
increased. For example, the frequency of the number of particles with 6 contact points
decreased by around 10% as the vertical stress increased from 11 kPa to 2010 kPa,
while that with 9 contact points increased by around 10%. In contrast, the lowest and
highest coordination numbers of 4 and 10, respectively, maintained their low
frequencies, i.e. an increase of the vertical stress is unlikely to change the range of
coordination number significantly. The most likely coordination number remained 7.
Figure 5. Frequency of coordination number (data from 150 particles each test case).
5. Discussion
The main aim of the study is to investigate the relationship between void ratio (e) and
the mean coordination number (NC) as the applied vertical load increased under K0
conditions. The mean coordination number obtained for test case 1 (11 kPa) is plotted
against the initial void ratio (e = 0.593, NC = 6.87) in Figure 6 and compared
experimental data for monosized spheres available in the literature. The empirical e-NC
relationship suggested by Smith et al. [6] is also given. In general for the data
considered, the mean coordination number tends to increase linearly with decreasing
void ratio. The results from the current study are close to the data (e = 0.612, NC =
7.26) from Oda [2] who used monosized ballotini with diameters of 12.4 mm. The
result of Pinson et al. [8] for an assembly of uniform spheres with diameters of 12.7
mm also agrees with the trend from the data in Oda [2]. Bernal & Mason [7] also
shows a similar range. Fonseca et al. [1] proposed an e-NC relationship based results
obtained using micro-computed tomography (CT) on sand. This curve appears in a
good agreement with prior experiments except Smith et al. [6]. While the magnitude of
NC at a given void ratio is approximately 20% larger in Smith et al. [6], the trend of e-
NC curve is similar to the Fonseca et al. [1] and other experimental data.
Figure 6. e-NC relationship at initial packing using monosized spheres and data from Fonseca et al. [1].
Figure 7. Evolution of e-NC relationship due to loading.
The evolution of e-NC relationship during K0 compression observed in this study is

plotted in Figure 7. The mean coordination number increased and the void ratio
decreased with increasing vertical stress. Bernal & Mason [7] confined their sample
with thick rubber bands, which can be considered as an isotropic compression, and
their data are also included on Figure 7. Discrete element method (DEM) results by
ODonovan [12], who used realistic properties of ballotini based on the previous
experimental study (Cavarretta et al. [11]) are also plotted in Figure 7. For the DEM
data, the variation in NC with e during isotropic compression from 100 to 1000 kPa
followed a similar trend to that observed for engaged contacts in Bernal & Masons
results. The slope of the e-NC curve for the K0 experiments data exceeds the slopes
observed for either the isotropic experiments or DEM simulations. This may indicate
that the K0 compression slightly alters the e-NC relationship by gaining new engaged
contacts from the virtual contacts due to shearing at the contact points.
6. Conclusion
This paper has investigated the influence of load-induced deformation on the frequency
of inter-particle contacts. The relationship between the mean coordination number and
the void ratio is needed to analytically estimate the small strain stiffness of assemblies
of spheres. While numerical simulations using DEM can provide this information,
physical verification of the trends observed in such simulations is important. This study
has investigated the stress-dependency of the e-NC relationship. To quantify the
coordination number of each ballotini experimentally, samples inside the oedometer
apparatus were impregnated with a black ink. The data showed that the mean
coordination number increased with increasing vertical stress as would be expected,
however it was interesting to note that the frequency of coordination numbers in the
range of 7 - 9 increased, while there was a decrease in the frequency of coordination
numbers in the range of 4 - 6. There was no noticeable increase in the number of
contacts with a high coordination number (10) or a low coordination number (3). The
experimental results generally agreed with prior experimental research. A comparison
of the experimental data and DEM results indicates that the nature of loading (K0
versus isotropic) may influence the nature of the e-NC relationship.
References
[1] J. Fonseca, C. O'Sullivan, M.R. Coop, P.D. Lee, Quantifying the evolution of soil fabric during shearing
using scalar parameters, Gotechnique, 63 (2013), 818829.
[2] M. Oda, Co-ordination number and its relation to shear strength of granular material. Soils and
Foundations 17 (1977), 2942.
[3] C. Chang, C. Liao, Estimates of elastic modulus for media of randomly packed granules, Appl. Mech. Rev
47 (1994), (1-2) S197S206.
[4] S. Yimsiri, K. Soga, Micromechanics-based stressstrain behaviour of soils at small strains.
Gotechnique 50 (2000), 559571
[5] R.D. Mindlin, Compliance of elastic bodies in contact Trans. ASME 71 (1949), A259268
[6] W.O. Smith, P.D. Foote, P.F. Busanng, Packing of homogeneous spheres, Phys. Rev. 34 (1929), 1271
1274.
[7] J.D. Bernal, J. Mason, Co-ordination of randomly packed spheres, Nature 188 (1960), 910911.
[8] D. Pinson, R.P. Zou, A.B. Yu, P. Zulli, M.J. McCarthy, Coordination number of binary mixtures of
spheres, J. Phys. D: Appl. Phys. 31 (1998). 457462.
[9] M. Otsubo, C. OSullivan, W. Sim, A methodology for accurate roughness measurements of soils using
optical interferometry, Geomechanics from Micro to Macro Proceedings IS-Cambridge 2014, Soga, K.,
Kumar, K., Biscontin, G. and Kuo, M. (Eds.) 11171122, CRC Press, 2014.
[10] F.N. Altuhafi, M.R. Coop, Changes to particle characteristics associated with the compression of sands,
Gotechnique 61 (2011), 459471.
[11] I. Cavarretta, C. OSullivan, E. Ibraim, M. Lings, S. Hamlin, D.M. Wood, Characterization of artificial
spherical particles for DEM validation studies, Particuology 10 (2012), 209220.
[12] J. O'Donovan, Micromechanics of Wave Propagation through Granular Material, PhD Thesis Imperial
College London, 2013.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-543
Comparison between conventional and

large scale triaxial compression tests on
peat
Cor Zwanenburga,1, M.A. Van a
a
Deltares, Delft, The Netherlands
Abstract. Large sized triaxial compression tests, diameter 40 cm and height 80 cm,
are conducted on peat samples. The aim is twofold. First, heterogeneity of peat
causes variation in parameters obtained from conventional laboratory testing.
However, heterogeneity in peat is found at any level from microscopic to
macroscopic. When considering stability of embankments on a peat foundation,
variation in parameters at a scale which is smaller than the dimensions of the
potential sliding plane will be (partly) averaged. So, the variation found in
conventional laboratory tests might overestimate the variation in parameters
relevant for the stability of embankments. Second, conventional tests on peat
samples show some peculiar behaviour compared to tests on clay and sands, which
might be explained by presence of fibres with a length equivalent to sample
dimensions. Results of the large triaxial tests show that variation in parameters
obtained from large tests is significantly smaller than found from conventional
sized tests. The large tests on peat samples show the same behaviour as found for
conventional tests, except that failure is found in large tests, which is usually not
found in conventional tests.
Keywords. Peat, heterogeneity, scale effects, triaxial testing
1. Introduction
In geotechnical engineering, peat is often considered as a complex material to deal with,

(among others [2] and [4]). One aspect that contributes to its complexity is the strong
heterogeneity in stiffness and strength behaviour. The strong heterogeneity causes
scatter in conventional geo-technical test results which introduces the use of high
(partial) safety factors to obtain design parameters.
In many lowland areas dikes have been built on organic subsoil. Assessing the
stability of these embankments requires information about the operational shear
strength of peat and the safety factors that should be applied. To study scale effects
large size triaxial compression tests are conducted on samples with a diameter of 40 cm
and a height of 80 cm. This paper discusses scale effects and its potential influence in
section 2. Next section 3 gives a description of the tested peat. Section 4 discusses the
results of conventional sized tests, followed by a description of the large triaxial test
set-up, test results and a comparison of the results of the conventional and large triaxial
tests, in section 5 and 6. Section 7 finalizes with conclusions.
1
Corresponding Author: [email protected]
544 C. Zwanenburg and M.A. Van / Conventional and Large Scale Triaxial Compression Tests on Peat
2. Scale effects
The study on scale effects contains two aspects. The first aspect deals with
heterogeneity. The second aspect considers peculiar behaviour found in conventional
triaxial compression tests on peat. Both aspects are discussed below.
Aspect 1; the heterogeneity of peat is found at any level. Figure 1a and b show
heterogeneity at microscopic level due to the cell structure in the organic origin of the
peat. Figure 1C shows the fiber structure found in one of the tested samples, indicating
the heterogeneity at fibre, cm, level. Next, due to the geological processes that took
place during deposition of a peat deposit weak and strong spots can be identified in
depth, [3], providing heterogeneity in dm or m level. Finally along a dike section,
which might have a length of 10 100 km, stronger and weaker spots in the subsoil,
can be found giving heterogeneity at km-scale.
Figure 1 a), b) microscopic photos of Sphagnum austinii (Sphagnum) photo G. van Wierdum, c) close-up of
sample tested in large triaxial test T9-1
Regarding a typical failure plane for dikes, local weak and strong locations might
compensate for each other and strength parameters will be (partly) averaged. In
conventional sized triaxial tests this averaging will not take place. As a consequence
the variation in strength of peat might be overestimated leading to unnecessary
conservative design values and partial safety factors.
This observation leads to the hypothesis that test results based on conventional
sized samples lead to a larger deviation in strength parameters than relevant for a
potential sliding plane. To validate this hypothesis a series of laboratory tests is
conducted on large sized samples and results are compared to conventional sized
laboratory tests.
C. Zwanenburg and M.A. Van / Conventional and Large Scale Triaxial Compression Tests on Peat 545
Aspect 2, conventional triaxial compression tests on peat show behaviour that

differs from tests on clay or sand. In tests on peat, the excess pore pressure in the
undrained shearing phase increases such that lateral effective stresses reduce to 0 and
stress paths reach the Tension Cut Off, TCO, line. ([2], [5]), see Figure 3. When
shearing continues the stress paths follow the TCO-line. Peak strength is rarely found,
instead strength increases for increasing axial strain. It should be noted that for fibrous
peat the fibres might have a length in the order of sample dimensions. Therefore it can
be questioned if conventional sized triaxial tests on peat are relevant for field
conditions. The large triaxial tests provide the option to check if this behaviour is an
artifact of conventional laboratory testing.
3. Characterisation of tested peat
An extensive description of the site at which the tested samples are retrieved is given in
[5] and [6]. A short summary is given here. The test site is located approximately 15
km north of Amsterdam. Figure 2 shows the sub soil profile. The groundwater table is
10 to 20 cm below ground level. Table 1 shows some characterization of the peat layer.
It should be noted that the density is low and falls below the density of water, which
might be explained by the presence of gas. The peat comprises mainly Phragmites with
sedge and sphagnum inclusions with minor decomposition, H2-H3 in the von Post
classification.
Figure 2, Typical CPTu, ball penetrometer test and soil profile at the test site, from [5]
Figure 2 gives a typical CPTu and ball-probe measurement. The pre-consolidation

stress vy, in the peat layer ranges from 15.2 kPa to 8.4 kPa, found from oedometer
tests using the Becker et al method, [1]. The vy depth profile shows a minor decline in
depth, which is explained by the small density found below the dry crust.
Table 1. Characterization of the peat at the test site

parameter unity average range
density, [t/m3] 0.95 7%
density of solids, s [t/m3] 1.53 1.6%
water content, w [%] 912 12%
loss on ignition, N [%] 85.7 5%
4. Conventional laboratory tests
An extended field and laboratory investigation was conducted, see [5]. Table 2 and
Figure 3 show the results of isotropically (CIUC) and an-istropically (CAUC)
consolidated conventionally sized triaxial compression tests.
Table 2, Summary of conventional triaxial compression tests, CIUC = isotropically consolidated undrained
compression tests, CAUC = an-isotropically consolidated undrained compression tests, vc = the vertical
consolidation stress at start shearing phase
parameter CIUC CAUC
su/vc [ - ] 0.59 0.54
VAR su/vc [ - ] 0.13 0.13
number of tests 23 8
su
120
extrapolation
q
120
100
100 p'
80
80
q [kPa]
q [kPa]
60 60
40
40
20
20
0
0 20 40 60 80 100
p' [kPa]
0
0 5 10 15 20 25 30
[%]
a
Figure 3 Results of CIUC tests, from [5]
Figure 3 shows the stress paths and stress strain curves of the CIUC tests. As discussed
in section 2 the stress paths reach the TCO line. Tests are stopped at the maximum
plunger displacement, which is for most tests close to reaching the TCO-line. The
stress strain curve, see Figure 3, shows the strain hardening found after 0.02 0.03
axial strain and no peak strength can be distinguished. This behaviour deviates
significantly from tests on clay or sands and makes it difficult to define failure.
Following [2], the central figure shows the definition for undrained shear strength, su,
which has been applied in this paper. The intersection of the extrapolation of the central
part of the stress path and the TCO-line is taken as value for su. For comparing the
different tests and test series the ratio su/vc is used. Table 2 gives a summary of the
results for the CIUC and CAUC tests. It should be noted that only normally
consolidated samples are discussed in this section.
5. Large Triaxial Tests
Figure 4 shows the set-up of the large triaxial device. The tested samples have a
diameter of 0.40 m and a height of 0.80 m. The large triaxial apparatus resembles the
set-up of a conventional device as closely as possible. A membrane with a thickness of
1 mm is placed around the sample. Drainage is allowed by porous stones placed on top
and bottom in combination to side drains. The vertical force is applied by a plunger,
which could induce a maximum displacement of 0.50 m, a, max = 62%. A load cell is
placed directly below the plunger. The pore pressures are measured at the upper porous
stone. The cell pressure is induced by applying compressed air at the top of the cell. A
back pressure of 300 kPa, is applied to eliminate gas bubbles. Measuring the weight the
volume of the expelled pore water gives the volume change of the sample, under the
assumption that the solid parts of the peat are incompressible and gas bubbles are
sufficiently dissolved. The device was placed in a temperature controlled room, at a
temperature of 18/C.
Figure 4 a); photo of test set-up, b); rupture plane found in test T9-1, c) sketch of set up.
In total 12 tests are conducted. In 6 tests the sample is consolidated at

approximately field stress level, these tests are referred to as field stress tests. The
remaining 6 tests are consolidated at a stress level beyond the pre-consolidation stress,
referred to as NC-tests. The effective stress level in the field and the K0-value are low
for peats. Therefore, the initial stress conditions are difficult to reproduce in triaxial test
set up. Instead, for the field stress tests, the samples are isotropically consolidated at a
stress level of approximately the vertical effective field stress. For the NC- tests an an-
isotropically stress condition is applied during the consolidation phase using a K0 = 0.3,
see [5]. Tests were conducted at vertical effective stress of 33.33 kPa, 50 kPa and 66.67
kPa respectively. Each NC-test was conducted in duplicate. In the analysis, equivalent
to conventional tests, the large triaxial tests are considered as element tests.
6. Analysis of test results
Figure 5 and 6 give a summary of the test results. In Figure 5a, the 6 field stress tests
are represented by the group of stress paths that start at p= 5 kPa. The stress paths of
the NC- tests start at higher stress levels. All tests reach the TCO-line, after which the
NC-tests follow the TCO-line. The field stress tests fail shortly after reaching the TCO
line. Figure 5b shows the development of horizontal, h and vertical effective stress,
v, during the test. Due to excess pore pressure, induced in the shearing phase, h
reduces to 0. In contrast to the NC- tests, the field stress tests fail when this condition is
reached. For the NC- tests v increase further, while h = 0 until failure. For two field
stress tests, a negative h is found, which corresponds to an overshoot of the TCO-line
in Figure 5a. This might be explained by a combination of measurement accuracy, the
low stress levels for these tests and a reinforcing influence of the membrane.
a
120 T3-1
T3
T4-1
T4-2
T4
100
T5-1
T6-1
T6-3
T7-1
80 T7
T8-1
b
25
T8 T3-1
q [kN/m2 ]
T9-1 T3
60 T9 20 T4-1
TCO - line T4-2
T4
15 T5-1
T6-1
h' [kN/m ]
2
40 T6-3
10 T7-1
T7
T8-1
5 T8
20 T9-1
T9
0
0 -5
0 10 20 30 40 0 20 40 60 80 100 120
p' [kN/m2 ] ' [kN/m2]
v
Figure 5 Large triaxial test results, a) stress paths in isotropic effective stress, p and deviator stress, q,
b) horizontal effective stress, h versus vertical effective stress, v
Figure 6 shows the stress strain curves. The strains are related to the sample size
at the start of the shearing phase. All tests show strain hardening, which starts after a =
0.02 to 0.03. In contrast to the conventional tests, Figure 3, NC-tests show failure at the
end of the strain hardening phase, which is found by an abrupt decline in deviator stress
q. For the tests at field stress level failure was only found in two tests. It should be
noted that the post failure behaviour is not shown by Figure 5, for clarity reasons.
150
T3-1
T3
T4-1
T4-2
T4
100 T5-1
T6-1
q [kN/m2]
T6-3
T7-1
T7
T8-1
50 T8
T9-1
T9
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
[-]
a
Figure 6 Results large triaxial tests, deviator stress, q vs axial strian, a
Table 3, Summary of NC- large triaxial tests.

parameter value
su/vc [ - ] 0.63
VAR su/vc [ - ] 0.05
number of tests 6
Table 3 gives the su /vc ratio and VAR su/vc for the NC tests, using the criterion for
su given by Figure 3. Figure 4b shows the nearly vertical rupture plane found for NC-
consolidated test T9-1, indicating a failure in tension, which is to be expected when
TCO-conditions are reached failure mechanism found for the.
The occurrence of a clear failure for the large triaxial tests is a remarkable
difference to the conventional sized tests and can be explained by scale effects. In [4]
and [7] the strain hardening is explained by fibre action. Initially at the start of the
shearing phase, horizontally aligned fibres are stretched. In Figure 6 this occurs for
a < 0.02-0.03. When the fibers are stretched, the fibers act as reinforcement. This
explains the continuing increase in v in Figure 5, while h = 0. At large deformations
the fibers will break or slip leading to a rupture plane shown in Figure 4. For
conventional tests, displacements remain small although sample strain might be large.
Due to small displacements in conventional tests the strain in the fibers remain small
and failure due to rupture of the fibres is not reached. For large samples larger
deformations occur and failure of the fibres is reached.
7. Conclusions
Comparison of large and conventional triaxial test results leads to the following
conclusions:
+ Stress paths of both, conventional and large, triaxial tests reach the TCO line and
follow it. For both, large and conventional triaxial tests on peat strain hardening is
found. This behaviour is not an artifact of small sized samples.
+ In large NC-tests, clear failure at a . 0.30 is often found while in conventional tests
barreling of the sample is usually found. This difference can be explained by scale
effects. In conventional testing displacements remain small, although sample strain

might be large. In large tests, large displacements are applied, causing more
straining of the fibres and finally failure.
+ The failure planes found in the large NC-tests are steep, indicating failure in tension
which corresponds to TCO conditions.
+ Using the strength at maximum strain as a proxy for su from conventional test data,
might result in an unsafe approach. Since, due to scale effects failure is found in
large tests, which is not found in conventional tests.
+ Comparison of table 2 and 4 shows that for large triaxial tests a larger su/vc ratio is
found.
+ The comparison also shows that VAR su/vc found for tests on large samples is
considerably smaller than for conventional sized samples.
+ This indicates that heterogeneity at small level averages for large samples, leading
to a smaller variation in operational shear strength compared to conventional tests.
Tests on conventional sized samples will overestimate the variation in strength
parameters that is relevant for stability analysis for dikes on a peat foundation.
Therefore the use of conventional sized samples might lead to conservative design
parameters.
Acknowledgement
The authors would like to acknowledge Rijkswaterstaat-WVL and the HHNK Water
Authority for initiating and financing this research
References
[1] Becker D.E., Crooks J. H. A., Been K., Jefferies M.G. Work as a criterion for determining in situ and
yield stress in clays Canadian Geotechnical Journal 24 (1987), 549-564
[2] Den Haan, E.J. & Kruse, G.A.M. Characterisation and engineering properties of Dutch peats in:
Characterisation and Engineering Properties of Natural Soils, Tan, Phoon, Hight & Leroueil (eds)
Taylor& Francis Group London ISBN 978-0-415-42691-6, 2007
[3] Hobbs N.B. Mire morphology and the properties and behaviour of some British and foreign peats
Quaretly Journal of Engineering Geology 19 1986, 7-80
[4] Landva A.O. Characterization of Escuminac peat and construction on peatland in: Characterisation and
Engineering Properties of Natural Soils, Tan, Phoon, Hight & Leroueil (eds) Taylor& Fransic Group
London ISBN 978-0-415-42691-6, 2007
[5] Zwanenburg C., Jardine R.J. Laboratory, in-situ and full-scale load tests to assess flood embankment
stability on peat Gotechnique to be published in 2015.
[6] Zwanenburg C., Van M.A. Full scale field tests for strength assessment of peat in: proceedings of the
18th International Conference on Soil Mechanics and Geotechnical Engineering Paris, 2013
[7] Cola S., Cortellazzo G. The shear strength behaviour of two peaty soils, Geotechnical and Geological
Engineering 23 (2005) 679-695
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-551
Three-dimensional quantification of the

morphology and intragranular void ratio of
a shelly carbonate sand
Joana FONSECAa,1Constantino Carlos REYES-ALDASOROa, Laure WILSb
a
City University London, London, United Kingdom
b
Ghent University, Ghent, Belgium
Abstract. Shelly carbonate sands represent an extreme soil type in terms of their
mechanical behavior which derives from the bioclastic nature of the constituent
grains. In their uncemented form, these deposits exhibit very high compressibility,
which has posed a number of geotechnical engineering problems; in most cases
related to the reduction in the bearing capacities of both shallow and deep
foundations. Remarkable features of these carbonate sands include the complex
shape and the structural weakness of the grains and the high inter and intra
granular porosity. Previous studies, have quoted the interlocking of the angular
shelly particles to be at the origin of their high friction angles and high initial void
ratio, however, up until now, no scientific micro-scale examination has been
carried out. This paper presents a non-invasive image based investigation into the
grain morphology of a carbonate sand from the Persian Gulf. This sand has a
median grain size of 570m and a high CaCO3 content in the form of aragonite
and calcite. Three-dimensional images from x-ray computed tomography
(3DXRCT) with a size of 6m were used. The presence of various skeletal bodies
such as shells of small organisms with distinct densities and composition poses
real challenges for an accurate segmentation. Image processing algorithms were
developed in order to identify the individual sand grains and quantify their
properties. Earlier work on silica sands has highlighted the importance of 3D non-
invasive techniques in providing an accurate distribution of the grain sizes when
compared to more traditional techniques such as sieving analysis and 2D
microscopy. The methodology here proposed allows an accurate quantification of
grain shape and size and the assessment of grain damage following mechanical
deformation. This study, contributes towards improving our understanding of the
engineering properties of carbonate sands and thus, predicting their response under
loading.
Keywords. 3D imaging, carbonate sand, microstructure, shape, particle breakage
1. Introduction
Shelly carbonate sands are widely spread throughout the oceans seabed and have
represented a major challenge for geotechnical engineering since the late 60s. In
particular, major interest on this material has developed from problems encountered
during offshore platform installation. Numerous cases have been reported of bearing
capacities in shelly carbonate sands being much less than predicted using conventional
1
Corresponding Author: City University London, London EC1V 0HB, UK. E-mail:
[email protected]
552 J. Fonseca et al. / 3D Quantication of the Morphology and Intragranular Void Ratio
theories essentially based on Quartz sand behavior [1, 2]. There is a general recognition
that carbonate sands warrant special consideration and their behavior remains an
engineering enigma [3].
Shelly carbonate sands are bioclastic origin deposits comprising essentially the
remains of marine organisms, such as shell and skeletal materials and therefore,
possess very distinct characteristics and mechanical behavior when compared to more
common terrigenous soils. Outstanding features of the shelly grains include the
predominance of plate-like and angular shapes, structurally weak and thin-walled
bodies and the presence of intragranular voids. Shelly carbonate sands tend to form
loose soil packing, void ratio greater than 1 are commonly measured and this is
believed to be a result of the interlocking of the angular grains. Moreover, the high
compressibility of this sand is likely to derive from the collapsible nature of the soil
fabric also exacerbated by the relative softness of the grains (calcite has half the
hardness of quartz in the Mohs scale).
The loose internal structure developed in these sands results in few inter particle
contact points, which according to previous studies [4] leads to high inter granular
stresses at relatively low applied stress and hence, lower mean stress at which crushing
occurs. Semple suggested that this behavior approaches that of 'quick' sands and clays
[5]. The contribution of the structure of the individual grains on the physico-
mechanical characteristics of the soil has also been highlighted by Levacher in [6]. In
particular, the author has pointed out i) an intragranular porosity as high as 20%, ii) a
high compressibility loading to a collapsible fabric and iii) the change in granulometry
due to crushing which affects the mechanical behavior of the soil.
Despite the recognition of the effect of the micro scale properties on the
engineering properties of these sands, previous studies have been essentially qualitative
and based on two dimensional (2D) image analysis of thin sections. With the increase
of offshore activity in temperate and tropical regions of the world where shelly
carbonate sands proliferate, a more scientific explanation for the mechanical behavior
accounting for their microstructure is needed. This paper presents a methodology
developed to quantitatively describe the overall grain morphology and intragranular
porosity of the grains of a shelly carbonate sand using three dimensional tomographic
images.
2. Grain shape and the mechanical response
The influence of grain shape on the engineering properties of sand is widely reported in
the literature. Previous studies, mainly on quartz sands, have shown that the maximum
void ratio and minimum void ratio are closely related to the shape of the soil particles.
The range of void ratios (e max - emin) tends to decrease with increasing sphericity and
roundness, e.g. [7]. When bulky particles are mixed with platy ones, the effect is a
significant increase in the void ratio accompanied by an increase in compressibility and
reduced shear strength, e.g. [8]. Therefore, given the diversity in biogenic detritus
forming shelly carbonate sands, it is suggested that as important as particle size
distribution, particle shape distribution constitutes a fundamental characteristic of this
soil. A key characteristic of shelly carbonate sands is the friable nature of the grains
that tend to break and get damaged more easily when compared to silica rich sands.
Mesri in [9] has defined three levels of particle damage including i) abrasion or
polishing of particle surface asperities, ii) chipping, breaking or crushing of particle
J. Fonseca et al. / 3D Quantication of the Morphology and Intragranular Void Ratio 553
surface protrusions and sharp particle corners and iii) fracturing or splitting of particles.
The relative softness of the carbonate grains together with the angular nature of the
grains makes polishing of surface asperities and crushing of sharp corners particularly
significant. Experimental evaluation of grain fracture is generally obtained in post-test
analyses by investigating the change in grain size distribution using sieve analysis.
However, as highlighted by [10], sieving analysis presents major limitations when
compared to non-invasive analysis such as 3D-XRCT. While these findings were
related to silica sands, it is suggested here that for shelly carbonate, sieving limitations
will be aggravated. In fact, the predominance of plate-shaped grains with c<<a,b (for a,
b and c defining the length of the major, intermediate and minor axes of the grain,
respectively) and needle-shaped grains with a>>b,c; sieve analysis cannot capture grain
splitting since a given grain with an initial major axis length of a0 and after splitting
af=a0/2 with no change of the minor axis length, i.e. c 0=cf, is likely to fall through the
same sieve aperture [11]. Moreover, subtle grain damage can only be captured by a
detailed 3D description of the grain shape pre and post testing.
3. Soil description and image acquisition
The primary material used for this study is an uncemented shelly carbonate sand from
the Persian Gulf, termed S2 sand with a high CaCO3 content in the form of aragonite
and calcite. This sand is a well graded material with a median grain size of 570m and
the coefficient of uniformity of approximately 3.67. Additional physical properties of
this sand include an emin=0.832 and emax =1.382. Recent oedometer and triaxial
shearing tests on S2 sand have shown that the stress-strain and strength characteristics
of this sand are significantly affected by the presence of water. During 1D compression
the wet sand tends to compress more that the dry material, as shown in Figure 1.
Moreover, during shearing, saturated samples tend to exhibit less dilation and a lower
shear strength is reached when compared to the dry samples. As presented in [12],
granulometry analysis obtained from sieveing has indicated a clear shift in the PSD
following oedometer compression, but resulted of limited use to clearly identify
changes in grading between wet and dry samples. The methodology proposed in this
paper will provide a more scientific approach to assess the damage mechanism,
including breakage, caused by the effect of water on shelly carbonate sands.
Figure 1. Void ratio changes for wet and dry samples of S2 sand.
3.1. Image acquisition
Three dimensional images of the internal structure of S2 sand were obtained using high
resolution, x-ray computed tomography (XRCT). X-ray micro-CT is a non-destructive,
3D high resolution imaging method that allows for the internal structure of objects to
be investigated (Figure 2). The sand was scanned with a nanotom m (phoenix|x-ray,
GE) and the images obtained have a spatial resolution of 6m. Micro-CT scans
generate 3D images that map the variation of x-ray attenuation within objects,
represented by the intensity value (I) of each voxel in the image. This attenuation,
resulting from the x-ray interaction with matter, is a function of the composition and
density of the object and the beam energy. Dense materials (i.e., materials with high
atomic numbers) attenuate more the x-ray beam energy than less dense materials and
this difference in attenuation provides the contrast that forms the image. The capacity
of x-rays to penetrate an object is, therefore, highly material dependent. The presence
of various skeletal bodies such as shells of small organisms with distinct densities and
composition leads to images with a wide range of I values, which poses real challenges
for an accurate segmentation, as discussed in the following Section.
Figure 2. Slice through a 3D tomographic image of the shelly carbonate sand investigated in this study.
4. Methodology
4.1. Image processing
Tomographic images are rich in information. A key step required prior to extract
measurements from an image is the segmentation process where the voxels in the
image are classified into features of interest, in this case, the individual grains and the
void space. The image segmentation in this study was performed using an in-house
built algorithm based on Matlab (The Mathworks). The algorithm exploited the fact
that the grains exhibit a significant difference in the intensity values compared with the
background or void space. This allowed using an intensity-based algorithm to separate
the two phases, i.e. solid grains and void space. However, the significant difference of
intensity levels exhibited by the grains, as shown in Figures 3a, 3b, 3c and discussed
previously, introduces major challenges when it comes to identify the individual grains
within the solid phase. A simple binarisation with one threshold value followed by a
watershed segmentation technique, to separate grains touching [13, 14] would imply
that in some cases, dark objects could be losing some of their detail whilst bright
objects could be merged. Therefore, a multilevel intensity segmentation was
implemented. Multilevel intensity segmentation has proved to be successful at
segmenting complex shapes with varied intensity levels such as fluorescently-labelled
immune cells observed in biological experiments [15]. This intensity-based approach
was followed by the application of morphological operators (closing and opening) with
small structural element to enhance the result. Once the individual particles were
segmented, the data were stored as three-dimensional arrays, where each voxel in the
three-dimensional image was mapped to an array cell. Each particle-phase voxel was
assigned an integer identification number (id) to associate it with a specific grain.
(a)
(b) (c)
Figure 3. (a) Slice through a 3D image with the different intensity levels (i.e. brightness) of the grains
highlighted. (b) Histograms describing the total 3D tomographic data set (blue dashed line) and the histogram
of the slice shown in (a). Notice how some peaks that correspond to different regions are apparent in the
selected slice as compared with the total data set. (c) One profile line in which particles of three distinctive
intensity levels can be distinguished from the background.
4.2. Morphology and intra granular void ratio characterization
Having the 3D microstructure stored in MATLAB matrices where each grain is defined
by a cloud of points or elements with coordinates (x, y, z, I), it is possible to obtain
various metrics related to the morphology of the grains. The grain size is one example,
as shown in Figure 4, the grain size can be described using the global volume of the
grain, i.e. by counting the total number of voxels with a specific id. In addition, the size
can be described by the major, intermediate and minor dimensions of the particle. This
involves using Principal Component Analysis to obtain the principal axis orientations
and applying an orthogonal rotation to the voxel coordinates so that its principal axes
of inertia are parallel to the Cartesian axes, more details can be found in [10].
(a) (b)
(c)
Figure 4. Segmented grains of S2 sand, grains id identified by color and grouped by size: (a) large grains,
(b) medium grains and (c) small grains.
The overall shape characterization for bulky-shaped grains has been done primarily in
terms of sphericity, convexity, and the aspect ratios elongation and flatness, e.g. [10].
Characterizing the shape of biogenic sediments is of utmost importance as it allows
obtaining the distribution of these shapes. In particular, the predominance of plate-like
of bulky-like shapes using the flatness index defined as the ratio between the minimum
and the maximum principal axes lengths. Convexity and sphericity indices are more
relevant for bulky-shaped grains. These indices are however of limited use to describe
the intricate internal structures commonly found in some shells and often associated to
intra granular porosity. The quantification of the complex morphologies can help
evaluating the susceptibility of the grain to crush. To this end, the identification of
structural weaknesses such as thin walls, as illustrated in Figure 6, and measurement of
their thicknesses and orientation in relation to major principal stress, can provide
valuable information. Also important is the measurement of the intragranular void
ratio, which can be obtained by segmenting the void space contained in the overall
volume as depicted in Figures 5b and 6b. This investigation has shown that intra
granular voids can be internally as well as externally connected, the latter is commonly
found for the grain type shown in Figure 5. For bulky-type grains, most of the internal
voids are likely to be unconnected as illustrated in Figure 6. Aside from enhancing the
potential of grain to crush, the intricate internal topology of the grains and their
connected/unconnected internal voids is also responsible for difficulties in achieving a
desired B value during sample saturation, as reported by Coop in [16].
(a) (b)
Figure 5. (a) Individual grain showing intra granular voids and thin walls. (b) Segmented interconnected
intra granular voids.
(a) (b)
Figure 6. (a) Intra granular voids on a bulky-shaped grain. (b) Segmented intra granular voids.
5. Conclusions
This paper presents a significant first step towards the full characterization of the
microstructure of uncemented shelly carbonate sands. While the distribution of particle
sizes and the grain shapes are fundamental characteristics of any soil, traditional
experimental techniques such as sieving or thin section analysis fails to provide
adequate quantification in the case of the complex shapes with intricate internal
structures. Of particular interest is the need to obtain the distribution of plate-like and
bulky-type grains, for which a full 3D analysis is required. The methodology here
proposed enables a more scientific quantification of the overall grain morphology and
of the internal topology found in many shelly grains. Complex internal topologies are
associated with the presence of intragranular voids and can be directly linked to the
structural weakness of the grain and its susceptibility to crush. In addition, when
applied prior and after loading, this grain scale approach can provide a more precise
assessment of grain damage, including breakage, following mechanical deformation.
References
[1] N. Yasufuku, A.F.L. Hyde, Pile end-bearing capacity in crushable sands. Gotechnique 45, No. 4 (1995),
663676
[2] H.G. Poulos, M. Uesugi, G.S. Young, Strength and deformation properties of Bass Strait carbonate sands.
Geotechnical Engineering 13 (1982), 189-211.
[3] B. McClelland, (1988). "Calcareous Sediments: an Engineering Enigma,
[4] D.W. Airey, M.F. Randolph, A.M. Hyden, The strength and stiffness of two calcareous sands.
Engineering for Calcareous Sediments; Proc. Int. Conf., Perth, R.J. Jewell & D.C. Andrews, 43-50,
1988
[5] R.M. Semple, The mechanical properties of carbonate soils, International Conference on Calcareous
Sediments, Jewell and Andrews, eds., Balkema, Rotterdam, Perth, W.A., 807-836, 1988.
[6] D. Levacher. Static, tension or cyclic testing of model piles driven in calcareous formations 1988
[7] G.C. Cho, J. Dodds, J. Santamarina, Particle Shape Effects on Packing Density, Stiffness, and Strength:
Natural and Crushed Sands. J. Geotech. and Geoenvir. Engrg. 132, No. 5 (2006), 591-602.
[8] C. Clayton, M. Xu, A. Bloodworth, A laboratory study of the development of earth pressures behind
integral bridge abutments. Gotechnique 56, No. 8 (2006), 561572.
[9] G. Mesri, B. Vardhanabhuti, Compression of Granular Materials. Canadian Geotechnical Journal 46
(2009), 369-392.
[10] J. Fonseca, C. OSullivan, M. Coop, P.D. Lee, Non-invasive characterisation of particle morphology of
natural sands. Soils and Foundations 52, No. 4 (2012), 712-722,
[11] G. Miao, D. Airey. Breakage and ultimate states for a carbonate sand. Gotechnique 63, No. 14 (2013),
1221 1229.
[12] L. Wils, P. Van Impe, W. Haegeman, Geomechanics from Micro to Macro, 2 (2014), 1403-1408,
[13] J. Fonseca, C. OSullivan, M. Coop, P.D. Lee, Quantifying the evolution of soil fabric during shearing
using scalar parameters. Gotechnique, 63, No. 10 (2013), 818-829.
[14] J. Fonseca, C. OSullivan, M. Coop, P.D. Lee, Quantifying evolution soil fabric during shearing using
directional parameters, Gotechnique 63, No. 6 (2013), 487499.
[15] K.M. Henry, L. Pase, C.F. Ramos-Lopez, G.J. Lieschke, S.A. Renshaw, C.C. Reyes-Aldasoro,
PhagoSight: An Open-Source MATLAB Package for the Analysis of Fluorescent Neutrophil and
Macrophage Migration in a Zebrafish Model PloS one 8 (8) (2013), e72636
[16] M.R. Coop, The mechanics of uncemented carbonate sands. Gotechnique, 40, No. 4 (1990), 607-626.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-559
Evaluation of stationarity and selection

of appropriate transformation for
geostatistical modeling of geotechnical
projects
Tejo V. BHEEMASETTI1, Anand J. PUPPALA2, Aravind PEDARLA3,

Raju ACHARYA4, Santiago R. CABALLERO5
1, 3
Post-Doctoral Fellow, 2Professor, 4,5Doctoral Student
Department of Civil Engineering, The University of Texas at Arlington, Arlington,
Texas 76019
Abstract. Spatial variability of the soil properties is one of the major concerns for
the design and analysis of geotechnical engineering projects. Kriging is a
stochastic interpretation tool of geostatistics that considers the spatial variability to
predict soil properties at unsampled locations. Often conventional geostatistics rely
on second order stationarity, where mean and variances are assumed to be
constant. Geostatistics being widely adopted tool for prediction analysis requires
the data to be stationary and select appropriate transformation in the case of non-
stationary data. In this research study, a formal approach to determine the
stationarity and selecting the appropriate transformation using conventional
statistics is discussed, where geostatistics is applied for predicting the stiffness of
Controlled Low Strength Material (CLSM) which is used as a bedding material for
a pipeline. CLSM is initially tested using nondestructive testing for a 500feet
pipeline test section comprising of 10 different segments of 50 feet each. In this
paper, random factor analysis of variance (ANOVA) and bartletts test is used to
check the stationarity in the stiffness data at 10 different sections measured using
Spectral Analysis of Surface Wave (SASW) technique. Also, in case of non-
stationary data, different empirical methods, including box-cox method were
discussed to find out appropriate transformation for stabilizing the variance. The
test results and geoestatistical analysis depicted that the stiffness measurements
obtained at different sections are stationary and this conclusion is validated using
the semivariogram analysis. This analysis is successful in evaluating the
stationarity of the data for geostatistical modeling and also assessed the
performance of CLSM as a bedding material using geostatistical analysis.
Keywords: Stationarity, Controlled Low Strength Material (CLSM), Random

Factor Analysis of Variance (ANOVA), Bartlett Test, Box-Cox Transformation.
1. Introduction
The variability and uncertainties present in soil properties are often complex to assess
and implement in the Geotechnical engineering design and analysis. Previous
560 T.V. Bheemasetti et al. / Evaluation of Stationarity and Selection of Transformation
researchers have implemented various statistical and reliability analysis to quantify the
uncertainty and variability present in soil properties [1, 2, 3, 4, 5]. The existing
statistical analysis makes use of univariate and bivariate analysis, which doesnt
account for the spatial variability present in the soil properties. This lack of
incorporating spatial variability of the soil properties into the analysis brought
inevitable uncertainties to the geotechnical designs [6, 7].
Geostatistics is one strongly emerging field in the recent decade for the evaluation
of spatial variability present in the material properties. The use of geostatistics gained
prominence in the 1960s with the work of French professor Georges Matheron on
regionalized random variables, which originated from the work of Danie G. Krige in
the 1950s [8]. Since then, geostatistics have been applied to various disciplines related
to earth sciences [9]. One primary essential condition for applying geostatistics is to
suffice the stationarity condition.
Stationarity refers to the data having same joint probability distribution with
constant mean and variance [10]. The implication of strict stationarity condition
constraints the geostatistical application in Geotechnical engineering as the soils are
inherently heterogeneous in nature. Especially, satisfying the stationarity condition with
limited observations in Geotechnical projects is often not possible and it can only be
achieved with an infinite domain of observations [11]. In geotechnical engineering, the
stationarity condition is referred to the statistical homogeneity in soil parameters [12].
This paper focuses on evaluation of stationarity using a statistical approach,
random factor analysis of variance and bartletts method. The conclusions obtained
from these two methods are validated with a conventional geostatistical approach using
a variogram analysis. In case of non-stationarity in the data, the appropriate method to
convert the non-stationary data to stationary data is discussed. The current analysis is
performed to assess stiffness data of the CLSM pipeline bedding material.
2. Background of Test Procedures
The random factor analysis of variance is used in this study for evaluating the constant
mean in the data. The bartletts test is used to evaluate the constant variance in the data.
The Box- Cox transformation is implemented in this study, to select the appropriate
transformation function for converting the non-stationary data to stationary.
2.1. Random factor analysis of variance
Analysis of variance (ANOVA) is a statistical test procedure developed by R.A. Fisher

to analyze the difference in group means. The variability in the data is analyzed by
determining the variability between the groups (SSTreatment) and within the groups
(SSerror). The detailed theory of this test procedure can be found in statistical text books
[13].
In this research study, ANOVA test was used to evaluate the Stationarity present in
the data. This was accomplished by developing a statistical hypothesis. The test statistic
(F0) is used to test the hypothesis which is compared to the critical value obtained from
F-distribution table with (a - 1) and (N a) degrees of freedom.
T.V. Bheemasetti et al. / Evaluation of Stationarity and Selection of Transformation 561
2.2. Bartletts Test
The bartletts test is a statistical test procedure named after Maurice Stevenson Bartlett
to analyze the variance in k samples obtained from a population. As a part of the
stationary requirement, in this research study, this test was used to evaluate the constant
variance present in the data in all the treatments/sections under consideration. This was
accomplished by developing a statistical hypothesis.
The basic procedure involved in Bartletts test is computing a test statistic (
whose sample distribution can be approximated using a chi - square distribution with a-
1 degrees of freedom. The detailed procedure in developing the test statistic is given in
statistical text books [13].
2.3. Box-Cox Transformation
The non-stationarity in the data often brings complexity in the geostatistical analysis.
The predictions that are obtained through non-stationary data mislead and bring
inevitable problems. Transforming the data from non-stationarity to stationarity is one
way to approach this problem. Past researchers have provided the guidelines for
selection of appropriate transformation methods based on the skewness present in the
data [14]. However, the skewness obtained from histogram is sensitive to the bin size
selected.
The Box-Cox method is a more formal approach to choose the appropriate
transformation function instead of using empirical methods to select the transformation
function [15]. Of all methods, the power family of transformations (i.e., y* = y) are
extremely useful in stabilizing the variance [16]. In this research study, Box-Cox
procedure is utilized to select the appropriate transformation. The selected
transformation is adopted for stabilizing the variance present in the data and to convert
the non-stationary data into stationary.
3. Geotechnical Data for Analysis
In this study, the stiffness of the bedding material is used for analyzing the stationarity
in the data. The controlled low strength material (CLSM) prepared using native
plasticity clay is used as a bedding material for a 150 mile water pipeline project. Using
the existing native soil reduces the project cost and provides a sustainable engineering
solution by reducing the negative impacts on the environment [17, 18].
The water pipeline addressed in this study is intended to collect water from
different lakes and bring additional water supplies to the Dallas Metroplex in Texas,
United States. In order to test a novel bedding material prepared with in situ clayey soil,
the CLSM mixture is initially tested for 500ft pipeline to evaluate its performance.
The stiffness of the bedding material is determined using spectral analysis of
surface waves technique (SASW). SASW is a seismic non-destructive method used to
determine the small strain shear modulus of the materials. This method uses the
dispersive characteristics of surface waves to determine the variation of the surface
wave velocity of layered systems with depth. Since, this paper is focused on evaluating
the stationarity of the data, the detailed procedures for the seismic analysis using
SASW for this project can be found in [19]. For the ease in analyzing the data, the
pipeline of 500ft length is divided into ten sections with each 50ft length. Table 1
below provides the stiffness data of the bedding material determined after 28 days
curing period.
Table 1. Stiffness Measurements of bedding material after 28days Curing period
Secti Stiffness Observations in MPa
on 1 2 3 4 5 6 7 8 9 10
1 432.3 465.9 458.7 447.5 432
2 417.7 463.1 414.1 460.1 409 473.7 455.9 436.8 450.7 470.1
3 455.5 407.6 437.3 410.2 446
4 442.2 421.0 440.6 417.5 440 420.0 455.9 443.0 454.3 427.6
5 434.7 444.8 452.4 439.2 434 446.5 471.5 453.1 440.8 441.7
6 424.6 414.2 453.2 440.8 421 424.6 414.2 413.2 440.8 420.9
7 461.9 457.7 419.8 432.7 461
8 419.4 438.5 428.8 429.9 418 456.4 458.7 442.5 470.5 465.2
9 449.5 478.2 420.2 468.1 450
10 456.4 458.7 442.5 470.5 465
4. Evaluation of Stationarity
In this study, the stiffness of the CLSM material after 28 days curing period is used for
evaluation of stationarity. The constant mean and variance of the data is analyzed using
ANOVA and Bartlett test procedures.
4.1. Check for Constant Mean
Constant mean refers to all the sections having an equal mean value without local
variations in the data. This condition suffices, one part of the stationary definition. The
constant mean of the data values in different sections is determined using the random
factor analysis of variance (ANOVA) method. In this case, a null hypothesis is
developed for conducting ANOVA test as shown below.
H0 : 1 = 2 = 3 = 4 = 5 = 6 = 7 = 8 = 9 = 10
H1 : At least one mean is different
The above hypothesis is evaluated using the test statistic (F0). This statistic is compared
to the critical value at a significance level of 0.05. Table 2 shown below represents the
ANOVA test results for the stiffness values after a curing period of 28 days.
Table 2. ANOVA test results for Stiffness measurements of bedding material after 28days Curing period
Analysis of Variance
Source of Sum of Squares DOF Mean Square F0

Variation
Treatment 5758.3 9 639.8 2.11
Error 19679.0 65 302.7
Total 25437.3 74
*DOF = Degrees of freedom
From the f-distribution table [13], the fcritical value for a significance level of 0.05 is
obtained as 2.16. Since the F0 value calculated is smaller than the critical value, the null
hypothesis cannot be rejected, concluding the mean values are constant between the 10
sections after 28 days curing period.
The conclusion also signifies the uniform development of stiffness values along
the 500ft pipeline. However, in order to validate the conclusion obtained from the
ANOVA tests, the basic assumption in developing the ANOVA test has to be satisfied.
The ANOVA test assumes that the residual error values (difference between actual
value and mean value) are normally distributed and the residual variances in all the
sections are structure less.
The residual values calculated were used to check for the model adequacy in the
ANOVA model. The normal distribution of the residual values was evaluated using the
normality plots, and the variance was evaluated by plotting the residual values against
the fitted values for all the sections. Figure 1 represents the model adequacy plots for
stiffness values after 28 days curing period. From the plots, it can be inferred that the
residual error values follow a normal distribution with a coefficient of determination of
0.98. Also, the residual variance plots do not depict any locally varying variance values
satisfying the assumptions of the ANOVA test. Thus, through this test, the mean can be
considered constant in all the sections.
Figure 1. Model Adequacy plots for residual values of CLSM after 28 days curing period
4.2. Check for Constant Variance
In this section, the second part of the stationarity assumption is determined, which is
evaluating constant variance in the data. This condition is evaluated using bartletts
test. A null hypothesis is constructed as shown below:
Null hypothesis: H0: 12 = 22 = 32 = 42 = 52= 62= 72 = 82 = 92= 102
Alternative hypothesis: H1: At least one variance is different
The above hypothesis tests for a critical value obtained at a significance value of 0.05.
The chi-square distribution table [13] is used to determine the critical value, which is
16.9. Table 3 below presents the summary of results obtained from the bartletts test:
Table 3. Summary of Bartletts test for stiffness measurements after 28 days curing period
Parameter Results
Q 4.2
C 1.06

Test Statistic ( ) 9.10

Critical Region 16.9
From the above table, it can be inferred that the statistical value for the developed
hypothesis is smaller than the critical region. Therefore, the null hypothesis cannot be
rejected, concluding the variance of stiffness values are constant after 28 days curing
period.
This concludes the stationarity assumption, i.e., constant mean and variance in the
stiffness values of the pipeline bedding material after 28 days of curing period.
5. Validation of test results with variogram analysis
The conventional approach to determine stationarity in the data is performed by

plotting various variogram values with the lag distance. The variogram or semi-
variogram (h) is defined as the average of the squared difference of the random
variable for different lag distance. In other words, the variogram plots, presents the
similarity or correlation in the measurements with the distance. In this study, the
variogram analysis is conducted for the stiffness values obtained after 28 days curing
period to validate the conclusions obtained from earlier statistical tests. Figure 2
presents the variogram plot after 28 days curing period.
(h) = 200 + 400 [ 1- exp (-h/24)]
Figure 2. Variogram plot and model for stiffness values of CLSM after 28 days curing period
The Stationarity condition is evaluated in the variogram plots, by identifying the trend
in the variogram values. In an ideal variogram plot, the variogram values initially will
increase with an increase in lag distance and reaches sill value or constant value. If the
variogram values continues to increase with an increase in lag distance beyond the
global variance, which is indicated by dotted line in the figure 2, it depicts non-
stationary condition.
From the figure 2, it can be inferred that the variogram value increased initially
with an increase in the lag distance and reached a constant value. The global variance
of the stiffness values is indicated using the dotted line, where the variogram values are
below the global variance indicating the stationarity in the data. This validates the
conclusion obtained from the ANOVA test and Bartletts test that the mean and
variance in the data values is constant. The variability in the stiffness values is modeled
using an exponential function which is plotted in blue line.
6. Non- Stationarity data
In the above study, the stiffness of the pipeline bedding material is evaluated for
stationarity in the data. The ANOVA and Bartletts test conducted has concluded that
the mean and variance in the data is constant, reflecting the data is stationary. However,
in case of non-stationarity condition that the data is not normally distributed and the
data values contain locally varying mean and variance, the geostatistics modeling
cannot be performed. In this scenario, the data needs to be transformed using
appropriate transformations. By transforming the data, the variance in the data can be
reduced without detrending the actual information. The right transformation function is
determined using Box-Cox approach as discussed in section 2.3. Since, in this study the
data has sufficed the stationary condition; the box-cox approach is not performed.
7. Summary and Conclusion
Geostatistical modeling aids the geotechnical engineers in interpreting the soil

properties at unsampled locations. However, one primary condition to perform
geostatistics is to satisfy the stationarity condition in the data, which is constant mean
and variance. Earlier researchers have assumed the data was stationary or used
conventional variograms to analyze the stationary condition. With the limited available
information and soil being a heterogeneous material, strict stationarity condition cannot
be achieved. Therefore, in this study, the stationarity in the data is evaluated using new
statistical approaches. Below are findings from this study:
The ANOVA test conducted on stiffness measurements concluded the
constant mean in the data. This conclusion signifies the uniform stiffness
development of CLSM in all the sections at the end of 28 day curing period.
This test also confirms the normal distribution in the data which is important
for most of the statistical analysis.
The Bartletts test conducted on the stiffness measurements confirmed the
constant variance in the data satisfying the stationarity condition.
The results obtained from the ANOVA and Bartlett test are validated using the
variogram analysis, where the both these tests can be successfully used in
evaluating the stationarity in the data.
An exponential variability model is developed for the stiffness of the CLSM,

which can be used for interpreting the stiffness value for geostatistical
mapping.
The power family of transformations in the Box-Cox method can reduce the
variance in the data, however due to stationarity in the data this method is not
deployed in this study.
References
[1] Spry, M.J., Kulhawy, F.H., and Grigoriu, M.D. (1988) Reliability-based Foundation Design for
Transmission Line Structures: Geotechnical Site Characterization Strategy. Electric Power Research
Institute, Palo Alto, Calif., Report EL-5507(1).
[2] Soulie, M., Montes, P. and Silvestri, V. (1990), Modelling of Spatial Variability of Soil Parameters.
Canadian Geotechnical Journal, Vol. 27, pp. 617-630.
[3] Phoon, K.K., Kulhawy, F.H. and Grigoriu, M.D. (1995), Reliability-based Design of Foundation for
Transmission Line Structure, Report TR-105000, Electric Power Research Institute, Palo Alto.
[4] Chittoori, B. S., Puppala, A.J. (2011). Reliability Based Approach to Quantify Mineral Montmorillonite
in Expansive Clays GeoRisk, June 2011, Atlanta, GSP No. 224, June 26-28 2011, pp. 201-208.
[5] Cai, G., Liu, S., Puppala, A.J. Reliability Assessment of CPTU-Based Pile Capacity Predictions in Soft
Clay Deposits, Engineering Geology, Vol 141-142, 2012, pp. 84-91.
[6] Einstein, H.H. and Baecher, G.B. (1982), Probabilistic and Statistical Methods Engineering Geology, I.
Problem Statement and Introduction to Solution. Rock Mechanics and Rock Engineering, Vol. 12, pp.
47-61.
[7] Lacasse, S. and Nadim, F. (1996), Uncertainties in Characterizing Soil Properties. in Uncertainty in the
Geologic Environment (GSP 58), edited by Shackelford, C.D., Nelson, P.P., and Roth, M.J.S., ASCE,
New York, pp. 49-75
[8] Davidovi . N and Prolovi. V (2010). Modeling of Soil Parameters Spatial Uncertainty by
Geostatistics. Facta Univ. Ser. Archit. Civil Eng., 8 (2010), pp. 111118.
[9] Holdaway, K., Harness Oil and Gas Big Data with Analytics. Optimize Exploration and Production with
Data-Driven Models. John Wiley & Sons, ISBN: 9781118910955.
[10] Isaaks, E. H., and Srivastava, R. M. (1989). An Introduction to Applied Geostatistics, Oxford
University Press, New York.
[11] Phoon, K. K., Quek, S. T., and An, P. (2003). Identification of statistically homogeneous soil layers
using modified Bartlett statistics. Journal of Geotechnical and Geoenvironmental Engineering, 129(7),
649659.
[12] Baecher, G. B., Christian, J. T. (2003). Reliability and Statistics in Geotechnical Engineering[M]. John
Wiley & Sons Ltd., Chichester, England.
[13] Montgomery., Runger and Hubele (2010). Engineering Statistics, 4th edition.
[14] Tabachnick, B.G., & Fidell, L. S. (2007), Using multivariate statistics, 5th edition.
[15] Box, G. E. P., and Cox, D. R. (1964). An analysis of transformations. Journal of the Royal Statistical
Society, Series B 26 (2), pp. 211252.
[16] Douglas, C., and Montgomery. (2009). Design and Analysis of Experiment, 7th edition
[17] Chittoori, B., Puppala, A. J., Reddy, R., and Marshall, D. (2012). Sustainable Reutilization of
Excavated Trench Material. Geocongress 2012, ASCE, Reston, VA, 4280-4289.
[18] Puppala, A. J., Saride, S., and Williammee, R. (2012a). Sustainable Reuse of Limestone Quarry Fines
and RAP in Pavement Base/Subbase Layers. J. Mater. Civ. Eng., 10.1061/(ASCE)MT.1943-
5533.0000404, 418-429.
[19] Mothukori. P. R., (2014) Non-Destructive Sasw Evaluation of Controlled Low Strength Material as a
Pipeline Bedding Material. M.S. thesis, Univ. of Texas, Arlington, TX, 110.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-567
A study of the micro-mechanics of sand

particle breakage using a high-speed
microscope camera
Wanying WANGa, Matthew COOPa
a
City University of Hong Kong, Kowloon, Hong Kong
Abstract. Single particle uniaxial compression tests are widely used to study the
micro-mechanical properties of particles. In recent years, much research studying
the micro-mechanisms of sand particle breakage has been carried out using this
method. However, the breakage mechanisms are still not well known, partly
because of the rapid brittle failure of the particles. In this study, a high-speed
camera was used in single particle uniaxial compression tests to capture the
breakage process. The equipment enabled a maximum of about 2000 frames to be
obtained per second, which could identify the crushing process in detail. In this
paper, two modes of single particle breakage are proposed for Leighton Buzzard
sand. Firstly splitting caused by cracks slowly propagating from one contact
surface to the opposite one and secondly catastrophic explosion associated with
very rapid crack propagation from the interior of the particle outwards. Single
particles immersed in water were also compressed uniaxially to find if the
breakage mode and characteristic strength would be affected by water. Immersion
in water reduced the compression strength slightly and affected the single particle
crushing mode.
Keywords. high-speed microscope camera, sand particles, uniaxial compression

test, micro-mechanics
1. Introduction
In the past few decades, a great deal of research on the macro-mechanical properties of
sands has been carried out by means of laboratory tests [1-3]. In recent years, the
importance of micro-mechanisms of single particles on the macroscopic behaviour has
been realized and recent studies on the particle mechanics have focused on the soil
behaviour at small strains. Single particle uniaxial compression tests are widely used to
study the micro-mechanical properties of particles. For example, the relationship
between particle strength and compression behaviour has been investigated in detail by
[4-5]. Nakata et al. [6] carried out both one-dimensional compression tests and single
particle crushing tests on silica sand to illustrate the relationship between the macro-
scale mechanical behaviour of particles and the micro-scale mechanical behaviour of
individual ones. Cavarretta and OSullivan [7] conducted a series uniaxial compression
tests on Leighton Buzzard sand to study the effect of particle shape and geometry on
the particle strength and stiffness. Brzerowsky et al. [8] carried out crushing tests on
single sand grains and found a positive correlation between grain size and critical flaw
size.
568 W. Wang and M. Coop / A Study of the Micro-Mechanics of Sand Particle Breakage
Cundall and Strack [9] proposed the Discrete Element Method (DEM), which
achieved a more realistic simulation of the response of granular assemblies and a better
understanding about relationship between the response of soils and the complex
internal mechanisms.
However, the breakage mechanisms are still not well known, partly because of the
rapid brittle failure of the particles and partly because the breakage mechanisms of
natural particles may not correspond with that in ideal particles. This study aims at a
better understanding of the microscopic behaviour of natural sands under compression
by using a high-speed microscope camera.
2. Apparatus and materials
The system consists of a high-speed camera, a microscope lens, a loading frame, a load
cell, an LVDT, a data logger and a high intensity focused lighting system. It enables a
single particle to be tested under uniaxial compression between two rigid platens with
displacement control.
Details of the testing system are shown in Figure 1. The high-speed camera (1)
used in this study was a GigaView camera from SVSi. This allowed a maximum of
2130 frames to be captured per second, which could freeze the particle crushing
process. The frame rate used in this study was 1000 frames per second. The size of
particle was quite small, hence a microscopic lens (2) was needed to magnify the image
by 16 times. The selected frame rate determined the maximum exposure of the high-
speed camera. If more frames were captured per second, a lower exposure would be
obtained, which meant less light would go into the camera lens. A high intensity
focused light (3) was therefore necessary. Some image processing software, e.g. Sony
Vegas, was also needed to improve the image quality. The loading frame (4) had an
upper cross-head that was fixed while the lower platen moved up. The machine was
displacement-controlled and the speed of the platen was 0.1mm per minute. A high
resolution load cell (5) with capacity of 1000N and an LVDT (6) with a resolution of
0.2m were used to record the force and displacement.
Figure 1. The system used for the tests.

W. Wang and M. Coop / A Study of the Micro-Mechanics of Sand Particle Breakage 569
The upper platen and connection to the load cell (Figure 2) were made from a
single piece of stainless steel and kept as short as possible to improve stiffness and so
repeatability. The LVDT was mounted below the load cell to avoid the load cell
flexibility affecting the accuracy of the displacements. The lower platen was simply a
stainless steel plate. A glass box was used to contain water in order to conduct tests on
sand particles immersed in water.
a
Load cell
Connection
LVDT
Upper platen
b
Upper platen
Glass box
Water
Lower platen Particle
Figure 2. a Loading apparatus. b Schematic diagram showing the platen loading a particle immersed in water.
Leighton Buzzard sand was tested in this study, which is quartz. The particles are
sub-rounded, translucent and brown or white in colour with a greasy lustre. The
fraction used here was from 1.18mm to 2.36mm.
3. Experiments and results
A total of 96 tests were carried out. Among them, 49 particles were loaded in dry
conditions and 47 particles were immersed in distilled water. The whole loading
process was recorded by the high-speed microscope camera while the force and
displacement of the particle were recorded.
3.1 Crushing type classification by failure mode
There were two types of failure for Leighton Buzzard sand particles in the crushing
process, splitting and explosion. The relationship between the force and displacement
for a typical splitting type is shown in Figure 3. All data were corrected for the
compliance of the apparatus. In the initial loading stages, the response shows a small
concave curvature. This may be caused by local crushing of asperities at the contact
surface due to stress concentrations. Afterwards, it tends to be linear but dropped
suddenly at point 2. After a trough, there is a saw-toothed curve, resulting from a main
crack propagating.
200
2
150
force (N)
4
100
3
50 1
0
0 0.05 0.1
displacement (mm)
Figure 3. Force-displacement curve for a typical splitting mode.
Corresponding to the Figure 3, Figure 4 shows the crushing process of the particle
in this mode. At point 2, a front part broke away from the particle, corresponding to a
sharp drop in the force-displacement curve. After that, a main crack initiated and
propagated gradually, leading to the particle splitting into two large parts. It can be
seen from frame 3 to frame 4 that the width of the crack became larger. In frame 3, it is
clear that the width of the crack at the bottom is larger than that at the top. That implies
the crack may have propagated from bottom to top. The local crushing of asperities
before point 1 could not be observed because the deformation is only around 5m
according to Figure 3. During the loading process, for this type of failure few
fragments were created.
Figure 4. Crushing process for a typical splitting mode.
The second type of failure is quite different. Figure 5 illustrates a typical

relationship between the force and displacement for the explosion mode. It can be
observed that when the force is less than 40N, the curve is highly non-linear and more
concave to horizontal axis than that for splitting mode. When the force is higher, the
relationship tends to be linear up to the peak then reduces rapidly to zero. The
unloading response indicated by a dashed line is not measured accurately due to release
of stored energy in the system.
Figure 5. Force-displacement curves for a typical explosion mode.

From start to point 2 in Figure 6, no obvious changes can be observed until a crack
implied by a shadow occurred at point 2. Once the central crack occurred, it propagated
quickly before point 3 causing a sudden break within 0.04s. During the crushing
process, numerous fragments exploded (3-1and 3-2).
1 11.54 2 23.98
1mm 1mm
A shadow occurred
3-1 24.02 3-2 24.04

1mm 11mm
Numerous fragments
Figure 6. Crushing process for a typical explosion mode.
3.2 The relationship between the compression stress, crushing mode and the shape of
the particle
Wadell [10] proposed a parameter, roundness, which took all the corners of the
particles into consideration to describe the sharpness of the shape. However, later
studies showed that some parts of a particle under compression far away from the
loading axis had little impact on the crushing behaviour [11-12]. Hence, local
roundness is introduced here, defined as:
where is the average radius of curvature near either contact surface and is
the radius of the maximum inscribed circle of the particle outline. If the value of local
roundness is higher, the contact surface is flatter. In this study, the outlines of particles
were fitted by polynomial functions and as well as were calculated by a

custom-made code using MATLAB.
Figure 7 illustrates the relationship between the compression peak stress, crushing
mode and particle shape. For particles in a dry condition, the average local roundness is
1.1 and the average peak stress is 39.8MPa for the splitting mode while the average
local roundness is 1.8 and the average peak stress is 52.2MPa for the explosion mode.
For particles immersed in water, the average local roundness is 0.6 and the average
peak stress is 38.4MPa for the splitting mode while the average local roundness is 2.6
and the average peak stress is 47.2MPa for the explosion mode. It can be observed that
particles with a higher local roundness tend to explode while those with a lower local
roundness tend to split both in air or immersed in water. The particles failing in the
splitting mode tended to be weaker than those failing in an explosion mode. Those
particles that did not fail at lower loads as a result of splitting due to stress
concentrations or internal flaws would have a high stored energy. Water immersion had
little impact on the peak stress of the single LBS particles. One reason is that the
Leighton Buzzard sand has a compact structure and few initial flaws on the surface,
which leads to the matric suction effect of water not being significant [13].
a 100
80
peak stress (MPa)
60
splitting
40 explosion
20
0
0 1 2 3 4
local roundness (Rin)
b 100
80
peak stress (MPa)
60
splitting
40 explosion
20
0
0 1 2 3 4
local roundness (Rin)
Figure 7. The relationship between the local roundness, peak stress and crushing mode in different testing
conditions: a. in air. b. immersed in water.
Conclusion
A series of uniaxial compression tests on single Leighton Buzzard sand particles using
a high-speed microscope camera was carried out to capture the single particle crushing
process and find the relationship with the particle geometry. Particles in both dry
conditions and immersed in water were tested to explore the effect of water on the
particle strength. There were two types of failure, splitting and explosion. The former
mode described particles that broke into several large main pieces, some of them
remaining between the two platens after failure. For the latter, the particle explodes
fiercely into a large number of fragments. The study underlined the importance of
particle morphology. Local roundness was introduced here, which was used to describe
the sharpness near the contact surface. Local roundness affected both the particle
crushing mode and the characteristic strength. No significant difference of either
particle crushing mode or strength was observed for dry and immersed conditions in
this study.
Acknowledgement
This work described in this paper was fully supported by a grant from the Research
CityU 112712).
References
[1] Burland, J. B., On the Compressibility and Shear Strength of Natural Clays, Gotechnique 40 (1990), No.
3, 329-378.
[2] Coop, M. R., The Mechanics of Uncemented Carbonate Sands, Gotechnique 40 (1990), No. 4, 607-626.
[3] Coop, M. R. and Lee, I. K. The Behaviour of Granular Soils at Elevated Stresses, Predictive soil
mechanics (1993), 186198.
[4] McDowell, G. R. and Bolton, M. D. On the Micro Mechanics of Crushable Aggregates. Gotechnique 48
(1998), No. 5, 667-679
[5] Nakata, Y., Hyde, A. F. L., Hyodo, M., and Murata, H. A Probabilistic Approach to Sand Particle
Crushing in the Triaxial Test. Gotechnique 49 (1999), No. 5, 567-583
[6] Nakata, Y., Hyodo, M., Hyde, A. F. L., Kato, Y. and Murata, H. Microscopic Particle Crushing of Sand
Subjected to High Pressure One-Dimensional Compression. Soils and Foundations 41 (2001), 69-82
[7] Cavarretta, I. and OSullivan, C., The Mechanics of Rigid Irregular Particles Subject to Uniaxial
Compression. Gotechnique 62 (2012), No. 8, 681-692
[8] Brzerowsky, R. H., Spiers, C. J., Peach, C. J. and Hangx, S. J. T. Failure Behaviour of Single Sand
Grains: Theory versus Experiment. Journal of Geophysical Research 116 (2011)
[9] Cundall. P. A.; Strack, O. D. L., A Discrete Numerical Model for Granular Assemblies, Gotechnique 29
(1979), Issue 1, 47 65.
[10] Wadell, H. Volume, Shape, and Roundness of Rock Particles. Journal of Geology 40 (1932). 443-451
[11] Gallagher, J.J., Jr. Herzian Theory Applied to Tensile Fracture of Natural Sand Grains (Abs.). American
Geophysical Union Transactions 56 (1974), 1195
[12] Marshall, J. R. Clastic Particles-Scanning Electron Microscopy and Shape Analysis of Sedimentary and
Volcanic Clasts. Van Nostrand Reinhold Company Inc., 1987.
[13] Oldecop, L. A. and Alonso, E. E., A Model for Rockfill Compressibility. Gotechnique 51 (2001), No.
2, 127-139
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-575
Study on Viscous Property of Sedimentary

Soft Rock in Drained Triaxial and
Unconfined Compression Tests
Yukika MIYASHITAa,1, Junichi KOSEKIa, Kenji NAMIKAWAb
and Masashi MATSUMOTOc
a
The University of Tokyo, b Metropolitan Expressway Company Limited
and c NOM Company Limited, Japan
Abstract. To study the characteristics of creep deformation in sedimentary soft

rock under the construction of tunnels, rate-dependency characteristics due to the
viscous property of the sedimentary soft rock was investigated. As the results of
the series of drained triaxial compression tests and the unconfined compression
test, it was found that the type of viscous property in the pre-peak region was
Isotach. Creep failure due to the Isotach viscosity was observed during sustained
loading in one of the triaxial tests. The type of viscous property was changed from
Isotach to TESRA when the deviator stress was decreased to residual stress state.
Keywords. Creep deformation, drained triaxial compression test, sedimentary soft

rock, unconfined compression test, viscous property
1. Introduction
Evaluation of rate-dependency characteristics due to the viscous property of a

sedimentary soft rock is required in developing a proper constitutive model. It was
revealed by the previous experimental studies that the type of viscous property of the
bounded materials, such as rocks with very high strength [1], sedimentary soft rocks [2],
cement treated gravel [3] [4], and undisturbed stiff clays [5] [6], exhibited generally
Isotach in the pre-peak region. In contrast, less studies among them were conducted on
the viscous property in the post-peak region, the effect of confining stress on the
viscous property and the behavior of creep deformation until failure.
In this study, therefore, the viscous property of the sedimentary soft rock was
investigated by conducting a series of consolidated drained triaxial compression (herein
denoted as CD) tests and an unconfined compression (herein denoted as UC) test. Not
only verifying the viscous property in the pre-peak region, viscous property in the post-
peak region and the behavior of creep deformation until failure were focused on. To
evaluate the effect of confining stress, different levels of confining stress were applied
in the series of CD tests.
While several types of viscous property were found [7], two of them called as
Isotach and TESRA were observed in this study. It was classified as Isotach that the
1
Technical Support Specialist, Institute of Industrial Science, the University of Tokyo, 4-6-1 Komaba,
Meguro-ku, Tokyo, Japan; E-mail: [email protected]
576 Y. Miyashita et al. / Study on Viscous Property of Sedimentary Soft Rock
level of stress-strain relationship at a certain strain rate became higher/lower than that
at the smaller/larger strain rate. When the stress-strain relationship at a constant strain
rate is essentially independent of the value of strain rate, its viscosity type was
classified as TESRA.
2. Materials, Apparatus and Testing Procedure
The sedimentary soft rock used in this study was taken from Kazusa group formation
which was widely deposited in Kanto region of Japan. Undisturbed block samples in a
cubic shape having dimensions of 270 mm on each side were retrieved from a
construction site of an evacuation tunnel, which horizontally connects two main tunnels
of an expressway in the Tokyo metropolitan area. Their sampling depth from the
ground surface was approximately 45 meters. The samples after artificial crushing
consist of sand (12 % by weight), silt (77 %) and clay (11 %). The mean particle
diameter was 0.035 mm. A rectangular specimen with the dimensions of 58 mm in
width, 76 mm in depth and 160 mm in height was trimmed from the block sample.
To apply axial monotonic and sustained loadings accurately, a gear-type loading
apparatus was used, which could change stepwise the strain rate by a factor of 3000 at
maximum and could reverse the loading direction without a backlash [2].
A picture of a triaxial cell and a schematic figure of a specimen and deformation
transducers used in this study are shown in Figure 1. Axial load was measured by using
an inner load cell to avoid the influence of loading shaft friction. Global axial strain
was measured by an external displacement transducer (herein denoted as EDT). The
effect of bedding error was reduced by introducing thin layers of dental gypsum at the
top and bottom ends of the specimen. In addition, local measurements of axial and
horizontal strains were also conducted using vertical and horizontal local deformation
transducers (herein denoted as LDTs, [8] for the details). Two vertical and two
horizontal LDTs were placed on opposite sides of the specimen as shown in Figure 1 b.
To evaluate global volumetric strain during testing by measuring the volume of drained
water from the specimen, low capacity differential pressure transducer (herein denoted
as LCDPT) was employed. The summation of vertical and horizontal strains obtained
by LDTs, such as 1_LDTave + 23LDTave, was also evaluated as volumetric strain without
the effect of bedding error and a compliance associated with drainage system from the
specimen to the LCDPT.
Wet density and testing conditions of all the specimens are listed in Table 1. In the
series of CD tests, a specimen was saturated by a combination of vacuuming and back
pressurizing to 200 kPa in the pressure cell filled with water [9]. It was subjected to
isotropic consolidation (herein denoted as IC) from an initial effective confining stress
of 30 kPa towards each value of confining stress as presented in Table 1. After keeping
the isotropic stress state for approximately 40 hours under drained condition,
volumetric strain obtained by LDTs during the IC on each test was confirmed to be
small as summarized in Table 1. The process of saturation was not applied to the
specimen subjected to the UC test, while covering it by the same rubber membrane
used in the CD tests to protect against desiccation.
During axial loading, the axial strain rate was changed stepwise many times by a
factor of 10 in all the tests. The maximum value of the strain rate was 1000 times larger
than the minimum value which was equal to 0.0015 %/min. In the test CD_400, two
stages of sustained loading were applied during otherwise monotonic loading.
Y. Miyashita et al. / Study on Viscous Property of Sedimentary Soft Rock 577
b. External displacement
transducer, 1_EDT
76 mm LCDPT, vol_LCDPT
3
2
160 mm
4
1
58 mm
13 Vertical LDTs, 1_LDTave vol_LDTs
24 Horizontal LDTs, 3_LDTave = 1_LDTave + 23_LDTave
Figure 1. a. Picture of triaxial apparatus and b. schematic figure of specimen and deformation transducers.
Table 1. Wet density and testing conditions of specimens.
Test Wet density B value Confining Volumetric strain Sustained loading

name (g/cm3) stress (kPa) during IC by LDTs (%)
UC 1.884 - - - Not applied
CD_100 1.896 0.95 100 0.021 Not applied
CD_400 1.894 0.90 400 0.087 Applied
CD_800 1.896 0.86 800 0.219 Not applied
3. Test Results and Discussions
3.1. Overall Stress and Strain Relationships
Relationships between axial stress of the UC test, deviator stress of the series of CD
tests, volumetric strain measured by LCDPT and axial strain measured by EDT are
compared in Figure 2. The values of the maximum deviator stress in the CD tests
increased with the increase in the confining stress. These values were larger than the
value of maximum axial stress in the UC test. Reduction of stress to residual state was
observed in the tests UC and CD_100. Creep failure occurred in the test CD_400
during the second stage of sustained loading at the deviator stress of 2545 kPa.
Contractive volumetric strain at the same strain level accumulated in almost the
same manner in the pre-peak region of all the CD tests. After the maximum stress state
and the state in which creep failure was initiated, the trend of its accumulation was
changed to dilation in the tests CD_100 and 400, while contractive behavior continued
until the end of the test CD_800.
Figure 3 shows the status of deformation in the four side surfaces of all the
specimens after the tests. Numbers 1 to 4 attached to the left ends of the pictures
indicate the orientation of the specimen sides and coincide with the numbers shown in
Figure 1 b. Formation of multiple shear bands and large deformation along them were
observed in the tests CD_100 and 400 as shown in Figures 3 b and c. Dilative behavior
in the global volumetric strain as shown in Figure 2 may have been linked to the
progressive failure in the shear bands. On the other hand, deformation along the strain
localization zones denoted with broken lines in Figure 3 d was not significant in the test
CD_800.
4000
a) qmax
CD_800
3500 qmax
Deviator stress q, axial stress (kPa)

3000
max
2500 CD_400
2000 Monotonic loading at 0.15 %/min.
1500
CD_100
1000
500 UC
0 = 0.0015 1.5 %/min.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Axial strain by external displacement transducer 1_EDT (%)
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Volumetric strain vol (%)
1.0
b)
0.5
CD_800
0.0
Monotonic loading at 0.15 %/min.
-0.5 CD_400
CD_100
-1.0
-1.5
Figure 2. Relationships between axial stress/deviator stress in UC/CD tests, axial strain and volumetric strain.
Figure 3. Deformation of specimens after tests.
3.2. Viscous Property in Pre-peak Region
Typical relationships between the deviator stress and the axial strain measured by the
pair of vertical LDTs in the pre-peak region are shown in Figures 4 and 5 for the tests
CD_100 and 800, respectively. Figures 4 b and 5 b are the magnified views of Figures
4 a and 5 a up to the axial strain of 0.35 %. Vertical lines on the stress-strain
relationships indicate the application of stepwise changes in the strain rate.
Instantaneous strain rates in all the steps are shown in these figures, while the reference
strain rate, , was set as 0.015 %/min. Hypothetical stress-strain relationships at the
constant strain rates of 1/10 , , 10 and 100 were added with four broken curves
in each figure. One of the curves was drawn by considering the locations of the vertical
lines. The other curves were drawn by magnifying or shrinking this curve in the
direction of the deviator stress axis, while assuming that a value of axial strain
exhibiting the maximum stress state is unique in all the curves. When the strain rates
4000 3000
a. b. 100&0 &0 &0

3500 qmax
100&0
2500 1
&0
Deviator stress q= (1'- 3') (kPa)

10&0 10&0 10
3000 100&0 10&0 100&0
& &0
10&0 0 2000
2500 1 1
100&0 &0 &0 &0
10
1
&0 10&0 10
2000 &0
10&0 1500
(qmax )& =100& = 1.11 (qmax )& =1/10&0 10
&0 1
&0 0
&0
10
1500
&0 1 1 100&0
10&0 &0 10 &0 &0 1000 10&0
10 10&0 100&0
1000 1 1
&0 100&0 &0
10 10
500
500 10&0
CD_100 &0 10&0 CD_100
&0 &0 = 0.015% / min . &0 = 0.015% / min .
0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Axial strain by LDTs 1_LDTave (%) Axial strain by LDTs 1_LDTave (%)
Figure 4. Stress-strain relationship in pre-peak region of test CD_100.

4000 3000
a. 100&0
qmax b. &0
3500 &0
100&0 2500
&0
100&0 10&0 10&0 1

&0
3000 &0 10& 100&0 10
0
1
&0 10& &0 100&0
&0 0
10 2000
2500
10&0 1
100&0 &0 &0 10&0 1
10 &0
2000 1500 10
&0 10&0
(qmax )& =100& = 1.21 (qmax )& =1/10&0 &0
1
1 0
&0
1500 &0
&0 10
1 10 10&0
&0 1000
10 100&0
1000 1 100&0 10&0
&0 1
10 &0
500 10
500
10&0 CD_800 10&0 CD_800
&0 &0 = 0.015% / min . &0 &0 = 0.015% / min .
0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
Axial strain by LDTs 1_LDTave (%) Axial strain by LDTs 1_LDTave (%)
Figure 5. Stress-strain relationship in pre-peak region of test CD_800.
suddenly increased or decreased, the stress-strain relationship shifted to that of higher

or lower position after indicating high stiffness and subsequent yielding as shown in
these figures. The same trend appeared in the UC test as well. That is, Isotach viscosity
was observed from the beginning of loading until the peak stress state, irrespective of
the different levels of confining stress. The type of viscosity was the same as the other
bounded materials reported in the previous studies.
According to the hypothetical curves in Figure 4 a, the value of maximum deviator
stress obtained at the largest strain rate, 100 , was 1.11 times as large as that at the
lowest strain rate, 1/10 , in the test CD_100. Corresponding multiplication factors in
the tests CD_800 and UC were 1.21 and 1.07, respectively. It could be estimated from
these factors that the difference of axial/deviator stress at the same strain level
mobilized under the different strain rates due to the Isotach viscosity became larger
under the higher levels of confining stress.
3.3. Viscous Property in Post-peak Region
Relationships between the deviator stress and the axial strain measured by EDT after
the maximum stress states of the tests CD_100 and 800 are shown in Figures 6 and 7.
1750 700
a. CD_100 b. 100&0
CD_100
1500 675 100&0

1250 650 10&0 &0
&0 10&0
10&0
1000 625 10&0
100&0
Magnified in b. Drastic decrease in q
750 600 due possibly to
progressive failure in
1 shear bands and cracks
&0
500 10 575
&0 = 0.015% / min . &0 = 0.015% / min .

250 550
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5.0
Axialby
Axial strain strain by external
external transducer
displacement transducer 1_EDT (%)
1_EXT (%) Axialby
external transducer
1_EXT (%)
Figure 6. Stress-strain relationship in post-peak region of test CD_100.

3750
CD_800
3500
100&0
3250
3000
2750
1
&0
2500 10
&0 = 0.015% / min .

2250
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Axialby
external transducer
1_EXT (%)
Figure 7. Stress-strain relationship in post-peak region of test CD_800.
Figure 6 b is the magnified view of the area surrounded by a broken rectangle in Figure
6 a. In these two tests, difference of deviator stress due to the Isotach viscosity became
smaller in the post-peak strain softening region with the increase in the axial strain. In
the test CD_100, the trend of viscosity changed gradually to TESRA type [7] with the
progress of loading. When the deviator stress was decreased to residual stress state, the
temporary increment or decrement of the deviator stress which took place upon a step
change in the strain rate tended to fully decay with the increase in the axial strain. The
hypothetical curves for the different strain rates appeared to join a unique curve which
was independent of the strain rate, as schematically added with a broken curve in the
range of axial strain from 3 % to 5 % in Figure 6. Similar change of viscosity type from
Isotach to TESRA also appeared in the UC test.
By the visual inspection of the specimens after the tests, it was confirmed that local
deformation in the shear band, as typically exhibited in Figure 3 b by connecting the
circle, triangle and cross marks with the broken lines, was significant in the test
CD_100. Large deformation in the vertical cracks and the shear bands occasionally
developed was observed in the UC test as shown in Figure 3 a. In contrast to these,
strain localization zones distinguished by the several marks in Figure 3 d were not fully
developed as any shear bands in the test CD_800, while the degree of local deformation
in these zones was smaller than that in the tests CD_100 and UC. It was estimated from
the observation that the change in the type of viscous property in the tests CD_100 and
UC was caused by the damage to the structures of the soil particles due to the extensive
deformation along the shear bands and cracks. That is, the type of viscous property of
the original sediment which consisted mainly of silt was TESRA. It was inferred that
lithification of the sediment into the soft rock changed the type of viscosity to Isotach.
3.4. Behavior of Creep Deformation
A relationship between the deviator stress and the axial strain measured by LDTs in the
test CD_400 is shown in Figure 8 a. Two stages of sustained loading at the deviator
stresses of 1778 and 2545 kPa were applied. The instantaneous strain rates indicated by
arrows during the sustained loadings were calculated from the sequential increment of
axial strain. In addition to the four hypothetical curves at the given strain rates, another
curve at the minimum strain rate obtained during the second stage of sustained loading,
1/45 , was plotted with a thicker broken curve. Figure 8 b shows time histories of the
instantaneous strain rate during the stages of sustained loading.
Both stages of sustained loading started just after the steps of monotonic loading at
the strain rate of 100 . In the beginning of both stages, accumulation of the axial strain
and decrease in the strain rate were observed. The degree of strain increment was not
large in the first stage, while the strain rate was decreased to the level of 1/100 after
20 minutes from the beginning of sustained loading. Creep failure occurred during the
second stage of sustained loading. As compared to the first stage, the amount of strain
increment was more significant, while the rate of decline in the strain rate was smaller.
Strain rate was linearly decreased during the primary creep, followed by the
convergence of its decrement in the secondary creep and the rapid increase from the
minimum value in the tertiary creep, on full logarithmic scale in Figure 8 b. Minimum
value of strain rate in the secondary creep was calculated as 1/45 (= 3.3 x 10-
4
%/min) at the elapsed time of 255 minutes. It was obtained at the axial strain of
0.43 % as shown in Figure 8 a.
As shown with the hypothetical curves at the given strain rates in Figures 4, 5 and
8 in the beginning of loading, stress-strain relationship in the pre-peak region depends
on the strain rate due to the Isotach viscosity. Under a constant stress state such as
sustained loading, accumulation of the axial strain continues until the stress-strain
relationship contacts with the hypothetical curves at the strain rates subsequently
obtained. When a point on the stress-strain relationship during sustained loading
4000 10
a. b. Primary creep Tertiary
Instantaneous strain rate by LDTs (%/min)
3500 creep
100&0 1
&0 1 Secondary 100&0
10 creep
3000 Sustained loading 10&0 &0
at q= 2545 kPa Overscaled 10&0
until end of test 100&0
0.1
2500
Sustained loading &0
10&0 &0 10& &0
2000 at q= 1778 kPa 0 1
&0 1
10 &0 0.01
for 30 minutes 1
&0 45
100&0
10 1 1
1500 &0 &0
1 45 10
&0 10&0 &0 1E-3
1 100
1000 &0 1 &0
10 1
&0 &0
10 CD_400 100
100&0 10&0
500 1E-4 &0 = 0.015% / min .
1 &0 10& CD_400 q= 1778 kPa 1
&0 &0 1
0
&0 = 0.015% / min . q= 2545 kPa 45 &0
0 10 1000
1E-5
0.0 0.1 0.2 0.3 0.4 0.5 1E-3 0.01 0.1 1 10 100 1000 10000
Axial strain by LDTs 1_LDTave (%) Elapsed time during sustained loading (min)
Figure 8. Behavior of creep deformation in test CD_400, a. stress-strain relationship and b. time histories of
strain rate during two stages of sustained loading.
coincides with the maximum stress state of the hypothetical curve at a certain value of
strain rate, unstable creep failure occurs as described more in detail in [4]. After the
coincidence, accumulation of the axial strain is accelerated by the increase in the strain
rate.
Based on this concept of Isotach viscosity, behavior of creep deformation in the
test CD_400 during the stages of sustained loading could be explained as follows;
Creep deformation was not largely accumulated during the first stage, as the stress-
strain relationship was quite far from the maximum stress states of the hypothetical
curves at the strain rates subsequently obtained. In contrast to this, the stress-strain
relationship at the axial strain of 0.43 % coincided with the maximum stress state of the
hypothetical curve at the minimum strain rate (= 1/45 ), during the second stage. It
was suggested that occurrence of the unstable creep failure due to the coincidence
caused the progress of creep deformation until failure as shown in Figure 3 c.
4. Conclusion
As the results of the series of drained triaxial compression tests and the unconfined
compression test on the sedimentary soft rock in this study, it was found that the
type of viscous property in the pre-peak region was Isotach.
Difference of the stresses at the same strain level mobilized under the different strain
rates due to the Isotach viscosity became larger under the higher levels of
confining stress.
Creep failure due to the Isotach viscosity was observed during the sustained loading in
one of the triaxial tests.
The type of viscous property was changed from Isotach to TESRA when the stress was
decreased to residual stress state under the low levels of confining stress.
References
[1] K. Hashiba, S. Okubo and K. Fukui: A new testing method for investigating the loading rate dependency
of peak and residual rock strength, International Journal of Rock Mechanics & Mining Sciences 43
(2006), 894904.
[2] K. Hayano, M. Matsumoto, F. Tatsuoka and J. Koseki: Evaluation of time-dependent deformation
properties of sedimentary soft rock and their constitutive modeling, Soils and Foundations (herein
denoted as S&F) 41 (2) (2001), 2138.
[3] L. Kongsukprasert and F. Tatsuoka: Ageing and viscous effects on the deformation and strength
characteristics of cement-mixed gravelly soil in triaxial compression, S&F 45 (6) (2005), 5574.
[4] A. Ezaoui, F. Tatsuoka, A. Duttine and H. Di Benedetto: Creep failure of geomaterials and its numerical
simulation, Geotechnique Letters 1 (2011), 4145.
[5] N. Komoto, F. Tatsuoka and T. Nishi: Viscous stress-strain properties of undisturbed Pleistocene clay
and its constitutive modelling, Proceedings of 3rd International Symposium in Deformation
Characteristics of Geomaterials, IS Lyon (2003), 579587.
[6] K. K. Sorensen, B. A. Baudet and B. Simpson: Influence of structure on the time-dependent behavior of a
stiff sedimentary clay, Geotechnique 57 (1) (2007), 113124.
[7] F. Tatsuoka, H. Di Benedetto, T. Enomoto, S. Kawabe and W. Kongkitkul: Various viscosity types of
geomaterials in shear and their mathematical expression, S&F 48 (1) (2008), 4160.
[8] S. Goto, F. Tatsuoka, S. Shibuya, Y. S. Kim and T. Sato: A simple gauge for local small strain
measurements in the laboratory, S&F 31 (1) (1991), 169180.
[9] S. K. Ampadu and F. Tatsuoka: Effects of setting method on the behavior of clays in triaxial compression
from saturation to undrained shear, S&F 33 (2) (1993), 1434.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-583
Assessment of parameters of
compressibility and strength of sedimentary
clays found in the northeast region of
Venezuela
Mariajos GUEVARA C.a
a
Geohidra Consultores
Abstract. The objective of the present study is to assess strength and

compressibility properties of sedimentary clays found offshore in the Gulf of Paria
in Venezuela. The clays subject of study are normally consolidated to lightly over
consolidated and are located in the Orinoco Delta, in sites where structures bore
by piles will be placed. The soil samples were obtained using 3" thin-walled
Shelby tubes in three boreholes of 120 meters depth. The parameters obtained are
those corresponding to the reconstituted clay at 1.2-1.5 wL (intrinsic parameters).
These were assessed by interpreting the results of one-dimensional consolidation
tests. The intrinsic compression line and the compression curves were plotted for
the reconstituted and the undisturbed clay, respectively. From these curves the Ivy
parameter was obtained. This parameter is used in the pile design method of the
Imperial College, to determine the value of the Kc ratio. This ratio affects the local
radial effective stress calculated. A different approach is to determine Kc using the
sensitivity of the clay obtained by undisturbed and remoulded triaxial UU tests.
The value of Kc obtained by the compressibility curves method was compared to
the correlation obtained by the UU sensitivities, for the clay between 0 and 64m
depth.
Two types of behaviour were observed, one from 0 to 64m (upper clay) and
another one from 64 to 120m (lower clay). It was observed that the clays do not
have a brittle behaviour. For the upper clay, the beginning of the compression
curves is below the sedimentary compression line and after the yield stress the
curves lie close to the intrinsic compression line. For the lower clay, the
compression curves lie below the intrinsic compression line. For the upper clay the
values of Kc obtained by the compression curves method do not differ significantly
from those obtained by the correlation with UU sensitivities. This can be attributed
to the low St because of the rapid deposition of the soil, due to the fact that it is
located near the Orinoco Delta.
Keywords. Clays, compression behaviour, pile foundation, sensitivity, intrinsic

parameters.
1. Introduction
The objective of the present study is to assess strength and compressibility properties of
sedimentary clays found offshore in the Gulf of Paria in Venezuela. This study is based
on work by [1] and [2] on the response of natural and reconstituted materials, and the
effects of structure on soil behaviour.
The term structure will be used here to define the combination of fabric, the
arrangement of the component particles, and bonding, the interparticle forces, which
584 M. Guevara C. / Assessment of Parameters of Compressibility and Strength
are not of a purely frictional nature [3], [2]. According to [4] clayey soils can be
encountered or produced in four different states of structure:
The intact state, as it occurs in natural deposits, the destructured state [5], the
remoulded state and the resedimented state.
There are two basic classes of structure [2]: The first is the sedimentation structure,
which is the one that develops as a result solely of one-dimensional consolidation, and
is present only in normally consolidated clays. The fabric formed during deposition
(the sedimentation structure) can be modified by post-depositional phenomena [6]. This
leads to the post-sedimentation structure, which develops when some geological
process subsequent to normal consolidation intervenes.
2. Soil profile
The soil samples were obtained using 3" and 2 1/4" thin walled Shelby tubes in three
boreholes of 120 meters depth. The clays subject of study are normally consolidated to
lightly over consolidated and are located in the Gulf of Paria, in a region near the
Orinoco Delta, which implies the existence of currents and active sedimentation. From
approximately 9000 years, the dominating sedimentation in the gulf consists on
packages of clays and silts from the Orinoco and Amazonas rivers.
Figure 1. Typical soil profile for the area of study

The generalised profile (Figure 1) consists on fine-grained soil with occasional
sand strata, occasional presence of organic matter and shell fragments. The fine grained
soil comprises medium to high plasticity clays [7].
From the mud line to a depth of 64m the clay is dark grey coloured with an YSR
around 1. This clay will be referred as upper clay. The graph (fig. 1) shows that the
clay, in its intact state, has a water content close to the liquid limit near the mudline
decreasing to a middle point between wP and wL at 64m.
M. Guevara C. / Assessment of Parameters of Compressibility and Strength 585
From 64 m until the final depth explored, the clay presents a light brown colour
and the Yield Strength Ratio YSR [1] obtained by one-dimensional consolidation tests
is higher than 1. This suggests a possible desiccation followed by a subsequent rise in
the groundwater table. This clay will be referred as lower clay. Figure 1 shows that
the clay, in its intact state, has a water content close to the plastic limit and that it
increases with depth.
Sample Depth (m) e0 'v0 (kPa)

12.7 1.130 96.1
38.2 1.102 279.2
53.5 1.151 403,0
71.7 0.809 558.9
81.7 0.880 648,0
85.5 0.762 689.3
103.37 0.919 849,0
107.5 0.713 889.2
111.7 0.788 924.9
Table 1. In situ void ratio and stress for the tested samples.
3. Intrinsic properties of the soil
According to [1], the intrinsic properties are those inherent to the clays that have been
reconstituted at water content between wL and 1.5 wL, without air drying or oven
drying, and then consolidated under one-dimensional conditions. The properties of
reconstituted clays are termed intrinsic properties since they are inherent to the soil
and independent of the natural state, and are denoted with an asterisk.
[1] introduced a normalising parameter based on mechanical properties of the soil,
the Void Index Iv:
e-e*100 e-e*100
Iv = = (1)
e100 -e*1000
*
C*c
e*100 , e*1000: Intrinsic void ratios for one-dimensional compression corresponding

to vertical effective stresses v of 100kPa and 1000kPa, respectively.
C*c : Intrinsic compression index.
In the normalised graph Iv-v, the compression curves of reconstituted materials
plot on a unique line called the Intrinsic Compression Line (ICL).
For the studied clays, one-dimensional consolidation tests at water content between
wL and 1.5 wL were performed. The parameters e*100, e*1000 and C*c are shown in Table 2.
Table 2. Average intrinsic constants of compressibility for the reconstituted natural clay.
wL wP Gs e100* e1000* Cc*
49.8 23.7 2.72 1,09 0,74 0,35
Figure 2 shows the intrinsic compression lines for the studied clay. From v= 100
kPa to 1000 kPa the experimentally obtained IC Lines correlate well with the ICL
proposed by [1]. However, below v= 100 kPa, the IC Lines tend to locate below the
correlation. For the clays subject of study, a new ICL correlation is proposed, which
fits better the behaviour of the soil at vertical stresses below 100 kPa:
3 2
I*v =0.048log10 ' v -0.320log10 ' v -0.315log10 ' v +1.527 (2)
ICL (correlation Burland 1990)

ICL Regression Curve
1 ICL 1.39 wL 71.7m
ICL 1.20 wL 81.7m
ICL 1.22 wL 107.5m
ICL 1.26 wL 111.7m
Intrinsic void index I*v
-1
-2
10 100 'v (kPa) 1000 10000
Figure 2. Normalised intrinsic compression curves and intrinsic compression line for the area of study.
4. Sedimentation compression curve
In situ, all normally consolidated clays from deposition follow a Sedimentation

Compression Curve in the e-v plane during one-dimensional compression [8]. In the
normalised plane Iv-v, [1] fitted a regression line through the natural SCC determined
by [9], and identified a unique Sedimentation Compression Line (SCL). This SCL is
recognised as the average curve for clays (with sedimentation structure) of sensitivities
between 2-9 [2].
According to [1], the intrinsic properties and their comparison to the properties of
natural clay can be used to assess the influence of in situ structure on its properties. In
figure 3, the SCC for the upper clay (in the Iv-v plane) is shown. [1] also states that in
an over consolidated clay the location and slope of the natural SCC are unknown, so
the SCC is not shown for the lower clay.
The shape of the SCC curve is sawtoothed, and is located above the intrinsic
compression line. However, the curve is too scattered to draw any sensible conclusions.
2.0
Sedimentation Compression
Curve
1.5
ICL Regression Curve
1.0
Void index Ivo
0.5
0.0
-0.5
-1.0
-1.5
-2.0
10 100 'v (kPa) 1000 10000
Figure 3. Normalised intrinsic compression curves and intrinsic compression line for the area of study.
5. Shear strength and compressibility of the clay
The clays do not show a sensitive behaviour since the compression curves lie below the
SCL (Figures 4 and 5). For the upper clay, the beginning of the compression curves is
below the sedimentary compression line. After the yield stress, the curves are steeper
than the SCL and lie close to the intrinsic compression line. For the lower clay, the
compression curves lie below the ICL, and after the yield stress, they tend to be parallel
to the ICL.
2.0 ICL Regression Curve
1.5 Sedimentation Compression Line

(correlation Burland 1990)
1.0 12.7m
'vo='vy 38.2m
0.5
'vo='vy
0.0
Void index Iv
-0.5
'vo='vy
-1.0
-1.5
-2.0
-2.5
10 100 1000 10000
'v (kPa)
Figure 4. Compression curves for the area of study 0-64m.
From the comparison between the compression curves and the ICL it can be
obtained the stress sensitivity. [2] suggested that strength sensitivity St is numerically
equal to the ratio of vy to the equivalent stress ey on the ICL, and it is denoted stress
sensitivity S:
St =S = 'vy *ey (3)
2.0 ICL Regression Curve
1.5 Sedimentation Compression Line

(correlation Burland 1990)
85.5m
1.0
103.37M
0.5
Void index Iv
0.0
-0.5 'vo
'vy
-1.0
-1.5 'vo 'vy
-2.0
10 100 1000 10000
'v (kPa)
Figure 5. Compression curves for the area of study 64-120m.
Table 3 shows the stress sensitivity S and the strength sensitivity St for the
samples shown in Figures 4 and 5. It can be observed that stress and strength
sensitivities are very similar for the upper clay. For the lower clay, S t values were not
obtained because UU tests were not performed below 64m.
Table 3. Stress sensitivity and strength sensitivity values.
Prof (kPa) 'vy (kPa) *ey (kPa) St S
12,7 99 42 2,0 2,4
38,2 265 140 2,0 1,9
53,5 392 164 2,0 2,4
85,5 1085 812 - 1,3
103,4 1488 1274 - 1,2
6. Comparison of Kc calculations
Since the site of study was planned to have structures bore by piles, estimations of axial
capacity were needed. One of the methods of assessment of pile capacity used was the
one proposed by [10]. In this method the capacity of piles in clays depends on the local
radial stress after equalisation 'rc and the operational interface angle of friction at
failure f. The radial stress 'rc depends on the vertical effective stress and a ratio
named Kc (as shown in equation 4). This ratio is obtained by parameters assessed from
comparison of the ICL and the oedometer compression curves, by the yield stress ratio
YSR, the relative pile tip depth h and the pile radius R. From the comparison of ICL
and the compression curves, two parameters can be obtained, Ivo and Ivy , which are
used in equation (5) or in equation (6), respectively, to calculate K c. Figure 6 shows the
procedure to obtain Ivo and Ivy .
'rc =Kc 'vo (4)
Kc =[2+0.625 Ivo ]YSR0.42 h/R -0.20 (5)
Kc =[2.2+0.016 YSR-0.870 Ivy ]YSR0.42 h/R -0.20 (6)
Alternatively, if such curves are not available, Ivy can be obtained using the
following expression, [10]:
Ivy =log10 St (7)
Where St= Strength sensitivity. However, [10] states that St values from UU tests
are subject to uncertainty due to disturbance on soil samples, and that they can be as
variable as the peak UU Su values are. Their recommendation is to check design
parameters by both strength and oedometer procedures, whereas possible.
Figure 6. Definitions of intrinsic properties of clay relating to oedometer behaviour [1].
For the upper clay, both procedures where applied and Kc values (divided by
h/R -0.20 for the purpose of comparison) were obtained. Table 4 shows the results of
these calculations. It can be observed that Kc values for the upper clay do not differ
significantly between strength and oedometer procedures. This can be attributed to the
fact that the clay has low strength sensitivity and low YSR. For the lower clay, K c
values were not obtained because UU tests were not performed below 64 metres and
sensitivity values were not assessed.
Table 4. Results of calculations of Kc.

'vo

Prof (m) Ivy St YSR
(kPa) /. /.
12,7 96 0,380 2 1,0 1,976 1,954
38,2 279 0,317 2 1,0 1,980 1,954
53,5 403 0,402 2 1,0 1,975 1,954
7. Conclusions
The present study assessed strength and compressibility properties of sedimentary clays
found offshore in the Gulf of Paria in Venezuela. Values for the intrinsic parameters
e*100, e*1000, C*c were obtained, as well as a correlation for the ICL for these clays. The
sedimentation compression curve was plotted and its behaviour compared to the ICL
obtained.
The oedometer compression curves for the upper and lower clays were plotted and
compared to the ICL and the SCL. Two types of behaviour were observed: one from 0
to 64m (upper clay) and another one from 64 to 120m (lower clay). It was observed
that the clays do not show a sensitive behaviour, since the compression curves lie
below the SCL. For the upper clay, the beginning of the compression curves is below
the sedimentary compression line and after the yield stress the curves lie close to the
intrinsic compression line. For the lower clay, the compression curves lie close to the
intrinsic compression line. Values of stress sensitivity S were obtained and compared
to strength sensitivity St values.
From the oedometer compression curves the Ivy parameter was obtained. The
value of Kc obtained by oedometer procedure was compared to the correlation obtained
by the strength procedure. Significant differences were not observed for the clay
between 0 and 64m depth. This can be attributed to the fact that the clay has low S t.
The low St might be due to its location in the Orinoco Delta, where rapid deposition
and non-stillness of the water are the regular conditions of sedimentation.
References
[1] J. B. Burland, On the compressibility and shear strength of natural clays, Geotechnique 40 (1990), No. 3,
329-378.
[2] F. Cotecchia & R. J. Chandler, A general framework for the mechanical behaviour of clays,
Geotechnique 50 (2000), No.4, 431-447.
[3] T. W. Lambe & R. V. Whitman, Soil Mechanics, Wiley, New York, 1969.
[4] F. Cotecchia & R. J. Chandler, The influence of structure on the pre-failure behaviour of a natural clay,
Geotechnique 47 (1997), No. 3, 523-544.
[5] S. Leroueil, F. Tavenas & J. Locat, Discussion: Correlations between index tests and the properties of
remoulded clays. W. D. Carrier III and J. F. Beckman. Geotechnique 34 (1984), No. 2, 211-228.
[6] S. Leroueil, F. Tavenas, F. Brucy & P. La Rochelle, Behaviour of destructured natural clays, Journal of
Soil Mech. Fdns Division American Society of Civil Engineers 105 (1979), GT6, 759-778.
[7] A. Gasparre, Advanced laboratory characterisation of London clay, (Doctoral Thesis), Imperial College
London, London, 2005.
[8] British Standard Institution, Code of Practice for Site Investigations BS 5930, London, 1999.
[9] A. W. Skempton, The consolidation of clays by gravitational compaction. Quaterly Journal of the
Geological Society 125 (1970), 373-411.
[10] R. Jardine, F. Chow, R. Overy & J. Standing, ICP design methods for driven piles in sands and clays,
Thomas Telford, London, 2005.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-591
The effect of Grain size distribution on the

Shear strength- dilation relation of
Granular material
Samaneh AMIRPOUR HAREHDASHT a,1 , Mourad KARRAY a,

Mahmoud N. HUSSIENa, Varvara ROUBTSOVAb, Mohamed CHEKIREDb
a
Department of Civil Engineering, Universit de Sherbrooke (Qubec), Canada
b
Institut de Recherche d'Hydro-Qubec (Qubec), Canada
Abstract. Over the past decades, investigating the dependence of mechanical

response of granular material on the discrete constituent particle characteristics such
as particle shape, particle surface, particle size and gradation, has attracted
considerable attention from both experimental and numerical points of view. In the
present paper, a framework is suggested to obtain initial insight into the contribution
of some parameters, corresponding to the nature of the sand particles (size, shape)
while evaluating the dilation response of granular material and developing a
correlation between dilatancy and shear strength in further steps of this study.
Extensive experimental results obtained from direct shear tests at different confining
pressures and initial relative densities are analyzed. These results show that the
particle size distribution and angularity have a substantial effect on the dilatancy
characteristics of the granular assembly. Analyzing the results calls into question the
accuracy of classical dilatancy theories to reflect the role of the grain size
distribution on the shear response and dilation of granular material.
Keywords. Granular material, Dilatancy, Particle-size distribution, Particle shape
1. Introduction
The shear resistance of the granular material can be characterized by a single parameter,
namely the angle of internal friction. This parameter stems from the inter-particle friction
angle, particle rearrangement, and dilation [1].
The inter-particle friction angle, , regarded conceptually as a material constant,
depends chiefly on particle surface roughness, and is essentially independent of
confining stress and packing density [1-2]. The term dilatancy was first used by Renolds
[3] to describe the different behavior of volumetric change observed in dense and loose
granular assemblies subjected to shear, while the loose sand contracts and the dense sand
dilates when subjected to shearing. By deducting the dilatancy components of the shear
resistance from the total stresses, the shear resistance due to other strength mobilizing
factors may thereby be determined. Rowe [1] proposed the 'stress-dilatancy' theory and
concluded that throughout most of the range of void ratios there is a third component of
1
Corresponding Author: Samaneh Amirpour Harehdasht, Universit de Sherbrooke, Sherbrooke
(Qubec) J1K 2R1, Canada; Email: [email protected]
592 S. Amirpour Harehdasht et al. / The Effect of Grain Size Distribution
shear strength other than frictional resistance and dilation, which is developed by energy
required to rearrange and reorient soils particles.
Over the past decades, a variety of experiments and theories have been put forward
to explain the relationship between friction angle and dilation, some of which have found
favor when describing the behavior of sands [1-4-5-6-7-8-9-10]. In spite of the
significant progress that has been made to indicate correlation between the shear strength
and dilation [11-12-13-14], the existing equations were developed without the explicit
consideration of the grain characteristics such as particle shape, roughness, and gradation.
The present paper aims to investigate the deviation in dilatancy predictions made
with the existing relationships that lack of grain characteristics parameter through
experimental study. For this purpose, a series of direct shear box tests on a full range of
cohesionless materials were performed. In addition, a framework is suggested for a better
fundamental understanding of grain size distribution effect together with grain shape on
dilation and its correlation with the shear strength of a granular material made up of
rounded and angular particles. The results obtained from this study provide initial insight
into the effect of parameters corresponding to the nature of the sand particles (mineralogy,
size, shape) on dilatation which should be considered while developing a correlation
between the particle size distribution and dilatancy in further steps of this study. For
geotechnical applications, improving the accuracy of dilatancy predictions based on
particle size distribution would be significant in cases where dilation has further effect
on soil behavior by increasing its confinement, besides the way that it controls the
appropriate angle of friction.
2. Methodology
The goal of this paper is to discover the main factors corresponding to particle
characteristics such as particle size and shape that may affect the dilatancy response of
granular material.
For the initial approach, section 3.1 aims to evaluate existing dilatancy theories to
understand the contribution of particle size distribution, to the dilation response of
granular material. For this purpose, first the existing stress-dilatancy relationships are
addressed in detail. Then, the disparities among dilatancy estimations of the existing
stress-dilatancy relationships, using the direct shear test results on basalt beads at
different confining pressures and initial relative densities, for different distributions are
discussed. In section 3.2, the potential impacts associated with shape quantification on
dilation behavior of granular material are investigated. The direct shear test results of
three different material of basalts beads, Pribonka and Eastmain sand that are
respectively made up of round, sub-rounded to sub-angular, and angular particles are
presented and compared.
2.1. Material
Three different types of granular material were used in this study: (1) Basalt
microspheres produced by Whitehouse Scientific Ltd, England, which is considered as
the reference material to investigate the particle size effect on shear strength and
dilatancy. For this purpose, 20 narrow gradations of beads between 63 and 2000 m were
mixed to produce the 6 particle-size distribution curves presented in Figure 1 (2)
Pribonka sand, obtained at Northern Quebec, Canada, consists of sub-angular to sub-
S. Amirpour Harehdasht et al. / The Effect of Grain Size Distribution 593
round grains;and (3) Eastmain sand which has angular grains. The same procedure as
described for basalt beads were used on Pribonka and Eastmain sands to produce
particle size distributions C1, C41 and C5. The physical properties of these three granular
materials are presented in Table 1. The values of the minimum and maximum void ratios
were determined in the dry state, according to the method specified by Muszynski [15]
(Table 1).
100
D50 Cu
90 C1 106 1.5
C2 212 1.5
80 C3 350 1.5
C4 600 1.5
70
C45 350 5
Finer by Weight (%)
C5 106 1.5 C1 C2 C4 C3 C5
60
50
C45
40
30
20
10
0
10
10 60 100 200 600 1000 2000 10000
Mean Grain Size (m)
Figure 1. Tested grain- size distribution curves
Table 1. General description of materials
Gs Void Ratio, e
Material Mineralogy Angularity
(gr/cm3) emin emax
(C1,B) 0,56 0,77
(C41,B) 0,54 0,72
Microspheres 2.9 Basalt Rounded
(C45,B) 0,41 0,59
(C5,B) 0,51 0,69
(C1,Pr) 0,55 0,88
Quartzo-feldspathic Subrounded (C41,Pr) 0,56 0,82
Pribonka sand 2.7 with ferromagnesium to subangular (C45,Pr) 0,46 0,64
(C5,Pr) 0,60 0,85
(C1,Em) 0,63 0,91
Eastmain sand 2.7 Quartzo-feldspathic Angular (C41,Em) 0,62 0,88
(C5,Em) 0,64 0,86
2.2. Shear box testing device, specimen preparation and testing procedure
The direct shear apparatus is used to measure shear strength parameters (peak and large
displacement) and dilation characteristics, which are related to the average conditions
existing on the mid-height horizontal plane. Although the direct shear test, with its rigid
confinement parallel to the shear direction, includes some limitations such as non-
uniformities of stresses and strains within the specimen, there is both numerical and
experimental evidence [16-17] that at peak state the direct shear test can produce
meaningful data, and that within the shear zone the deformation approximates to simple
shear.
The direct shear test apparatus used in this study consists of a metal box with a
square cross-section (55 mm wide), divided horizontally into two halves. During the tests,
the lower section of the box was moved forward at a constant velocity of 0.0051 mm/s
while the upper part was restrained against horizontal movement. In order to prevent the
unwanted effects causes the collapse of the material at the edges, which assists
progressive failure mechanism, a supplementary membrane, was laid over the interior
walls of the frames. The samples were prepared and sheared in dry conditions. Each test
specimen was prepared in the shear box in three layers of equal thickness. Dry mixtures
were carefully laid down in the shear box using a spoon to prevent particle segregation.
To obtain the different desired densities, mechanical compaction of each layer was
required. The state of the compactions was monitored by measuring the sample height,
until the required density was achieved.
3. Results
3.1. Evaluating the existing dilatancy theories to reflect the effect of particle size
distribution
The shear strength response and dilatancy of granular material is known to be affected
by the stress state and density. Extensive laboratory and theoretical researches have been
made to investigate the influence of density and pressure on shear strength and dilatancy
of sands [1-5-8-9-10-19-20-21]. Comparing the existing stress-dilatancy relationships
presented in Table 2 shows that the disparate understanding in the fundamentals of soil
dilatancy such that the inclusion of important factors differs from theory to theory.
In general, these relationships can be applied in two distinct senses. First, the can be
used to express the relationship between the mobilized angle of friction and the current
dilation rate as the test progresses. Secondly they may be used to express the relationship
between peak friction angle and the maximum dilation rate for several tests on the same
material. Here the focus is on the empirical relationships which are based on the
observation of latter type where equations are expressed in terms of the shearing
resistance and the incremental strain on a plane along which there is zero extension.
The most widely used expression is that of Bolton [8] who conducted a detailed
analysis of the strength and dilatancy of 17 sands in axisymmetric and plane strain
conditions at different densities and various confining pressures. He proposed the relative
dilatancy index (IR) as a measure of dilatancy potential of sands with reasonable accuracy
for practical purposes. This index is related to the relative density and effective confining
stress level through Equation (7).
The objective of this section is to examine the existing stress-dilatancy relationships
in predicting the effect of grain-size distribution on dilatancy angles of material
consisting essentially of spherical particles. The probable variations in dilatancy angles
of these materials can only be attributed to the particle size effect, among other particle
characteristics, such as shape and surface roughness, which have also an impact on the
dilatancy behavior of granular material.
For this purpose, 12 direct shear box tests were performed at four different normal
pressures (50, 100, 200 and 400 kPa) and three relative densities (Dr= 50%, 70% and
90%) for each distribution shown in Figure 1, giving a total of 72 tests. The dilation
angles derived from the existing equations are plotted versus relative dilatancy index to
account also for the effect of the relative density and normal stress on dilatancy. As
shown in Figure 2a, applying different stress-dilation relationships for distributions C1
leads to the different dilatancy predictions coming from the quite different approaches
taken by the authors in the development of their theories. The average of max values
from the different dilatancy theories for distributions C1 and C5 are shown in Figure 2b.
Comparing the size of error bars shows the increase of the uncertainty in the average and
deviation in dilatancy predictions between the various relationships by the increase of
D50 value. In addition, the average max values for distributions C1 to C5 are plotted
against their corresponding IR values in Figure 3a. It is clear from this Figure that the D50
has an effect on max while the average max decreased as D50 of samples increased at a
given IR.
28 C1 (a) 28 (b)
26 Eq. (1) 26 C1
Eq. (2)
24
Eq. (4)
24 C5
22 Eq. (6) 22
20 Eq. (8) 20
Eq. (12)
18 18
max ()
max ()
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
R R
Figure 2.(a) max values derived from the existing equations for C1 (b) The average of max values for C1-C5
The second reasonable interpretation for direct shear test results is to estimate the
dilation angle from Equation (13). From the observation in Figure 3b, it can be concluded
that the conventional interpretation for direct shear test also shows the tendency of max
to decrease as D50 increased at a given IR. However, the average dilation angles obtained
from the different dilatancy theories underestimate the dilation angles for given D50
values, compared to the conventional ones, which is of the order of 26%.
d yy dv
tan = = (13)
d yx du
28 28
(b)
(a)
26 (C1-B) 26 (C1-B)
24 (C2-B) 24 (C2-B)
22 (C4-B) 22 (C4-B)
max () tan-1(dv/du)
20 (C45-B) (C45-B)
max () Average
20
18 (C3-B) 18 (C3-B)
16 (C5-B) 16 (C5-B)
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
R R
Figure 3.(a) the average max values (b) conventional max values for distributions C1 to C5
From Figures 2 and 3, it can be concluded that the normal effective stress, relative
density, and D50 are the significant factors in the amount of dilatancy exhibited by
granular material, while it seems that the effect of coefficient of uniformity, Cu, is
negligible. At a given external load and relative density, the maximum differences in
dilatancy estimates using the equations presented in Table 2 fell within a range of 11-
21, depending on their particle size distribution. Therefore, for a more accurate estimate
it is essential to consider other factors to fully describe dilatancy behavior of granular
material. As the particle size is not included in the existing theories and it is hypothesized
that the inclusion of this parameters will help in better predicting granular dilatancy.
Table 2. Existing stress-dilatancy relationships

Authors Dilatancy Relationships Note
Taylor tanp = sincv
+ tan (1) Direct shear
(1948)
Rowe sin cv
+ sin Plane Strain
(1969) tan p = (2)
1 + sin cv
sin
De Josselin de Jong 1 k = 1 Plane Strain
(1976) [19] +1
k 3 k = 3 Triaxial Loading
sin =
(3)
1
1
k 3
Atkinson Sawtooth Model
and Bransby (1978) p = cv
+ (4)
[24]
Vermeer and de Borst

vp v = 1 + 2 + 3 Plane
(1984) sin = (5) strain and triaxial compression
21p + vp
Bolton Plane Strain, I R = D R (Q ln p ) R (7)

(1986) p cv
= 5I R (6) Q = 10 R = 1
p cv
= 0.8 ... (8) Quartz, Feldspar
Triaxial,
p cv
= 3I R (9)
Wood p simple shear loading
(1990) [20] p = sin 1 v (10) conditions

Vaid and Sasitharan triaxial compression
(1992) conditions

= sin1 (11)
2
3
+ 1
vp 1p

Schanz and Vermeer IR plane strain and
(1996) [21] = sin 1
(12) axisymmetric loading conditions
6 .7 + IR
* If the plane strain relationships are to be used for direct shear, they must be converted to a common
form, which can be done using Daviss (1968) relation.[18]
* p : Peak Friction Angle, : Critical State Friction Angle, : Dilation Angle
cv
* 1 , 2 , 3 : Principal Strains, v : Volumetric Strain, : Shear Strain (p denotes plastic quantity)
3.2. Grain shape effect together with grain size distribution on dilatancy prediction
The above section shows how the particle size distribution of sand has been
identified as one of the important factors affecting the dilation behavior of granular
material. In addition to the particle size distribution which is controlling the dilation
behavior of granular material, particle shape also emerges as a significant parameter that
needs to be properly identified [12-14-22-23]. These studies proved an increase in peak
shear strength and dilation with particle angularity through numerical and experimental
works. However, few studies separate the effect of particle angularity from that of other
factors such as grain size distribution. In addition, no systematic assessment of particle
shape have been done to evaluate the amount of dilatancy variation for different given
type of sand (rounded, sub-angular, or angular grains) with respect to the particle size
distribution
In this section an attempt is made to investigate 1) if the dilation is influenced by the
angularity of particles for three materials with identical grading curves and different
particle angularities. 2) if for a given soil consist of angular or sub-angular particles, the
dilation is affected by particle size distribution as much as what was observed for
idealized granular material (basalts beads).
28 (a) 28 (b)
26 (C1-B) 26 (C4-B)
24 (C1-Pr) 24 (C4-Pr)
22 (C1-Em) 22
(C4-Em)
20 20
18 18
max ()
max ()
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
R R
28
(c)
26 (C5-B)
24 (C5-Pr)
22 (C5-Em)
20
18
max ()
16
14
12
10
8
6
4
2
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
R
Figure 4. The max values increased by the increase of the angularity of particles for particle size
distributions of (a) C1,(b) C4 and (c) C5
The answer to these two questions may help in finding a comprehensive approach
to deal with particle angularity while developing the strength- dilation correlation for
different particle size distributions. For these purpose, the max values of Pribonka sand,
Eastmain sand and basalt beads, were compared in Figure 4 for a given particle size
distributions of C1, C4 and C5.The details of direct shear tests on Pribonka and Eastmain
sands (Dr and normal stresses values) are summarized in Table 3.
These results are in agreement with previous studies, showed the higher average
max values for these sands compared to that of Basalt beads, which is attributed to the
shape and angularity of their particles. Angularity of particles may cause an interlocked
fabric in addition to providing extra restraint to particle rotation and hence enhancing
dilation [12]. In addition, Figure 5 shows that for a given material made up of angular
particles, the max values are still affected by the D50, while the average max decreased
as D50 of samples increased at a given IR for both Pribonka and Eastmain sands.
However, it seems that the effect of D50 on dilation of granular material is lower than
that of samples made up of basalts beads. These results indicate that both particle shape
and size can affect the arrangement of the particles, the ability for one particle to ride up
over its surrounding neighbors, and thus the dilation behavior of granular materials.
Table 3. The details of direct shear tests on Pribonka and Eastmain sands
Curve No. Pribonka sand (Dr%, n kPa) Eastmain sand (Dr%, n)
C1 (90,50),(90,400),(70,100), (70,400), (90,50),(90,400),(70,100), (70,400),
(50,50), (50,100),(50,200) (50,50), (50,100),(50,200)
C4 (90,50),(90,200),(70,50),(50,50),(50,400) (90,50),(90,200),(70,50),(50,50),(50,400)
C5 (90,50),(90,100),(90,200),(90,400),(70,200), (90,50),(90,200),(90,400),(50,50),
(70,400),(50,50) (50,100),(50,200
28 28 (b)
(a)
26 (C1-Pr) 26 (C1-Em)
24 (C4-Pr) 24 (C4-Em)
22 (C5-Pr) 22
(C5-Em)
20 20
18 18
max ()
max ()
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
R R
Figure 5.(a) the average max values of (a) Pribonka and (b) Eastmain sand for distributions C1, C4 and C5
4. Conclusions and Recommendations
The study investigated the influence of particle size distribution curves on dilatancy and
its contribution to the peak friction angle of three different granular materials in direct
shear tests. The following conclusions emerged from the experimental results at different
confining pressures and initial relative densities:
Dilation angles obtained from the existing stress-dilatancy relationships and

conventional analysis show the effect of D50 on max. So that the average max decreased
as D50 of samples increased at a given IR for the three tested materials made up of rounded
to angular particles. The dilation is increased by the angularity of particles for the three
materials with identical grading curves, but different particle angularities. In addition,
for a given soil, the dilation is affected by D50. However, the effect of D50 decreased as
the angularity increased.
References
[1] Rowe, P. W. 1962. The stress-dilatancy relation for static equilibrium of an assembly of particles in contact.
Proceedings of the Royal Society, London, A269: 500-527
[2] Lee, K. L. and Seed, H. B. 1967 . Drained strength characteristics of sands. Journal of Soil Mechanics and
Foundation Divisions. ASCE 93, No. SM6, 117 141
[3] Reynolds, O. 1885. On the dilatancy of media composed of rigid particles in contact. Philos. Mag. 5(2):
469482
[4] Casagrande, A. 1936. Characteristics of cohesionless soils affecting the stability of earth fills. Journal of
Boston Society of Civil Engineers, 23(1), 13-32
[5] Taylor, D. W. 1948. Fundamentals of soil mechanics. John Wiley and Sons Inc., New York, N.Y.: 406-479
[6] Skempton A.W., Bishop A.W. 1954. Soils, in building materials, their elasticity and inelasticity. North
Holland Publishing Company, Amsterdam: 417482
[7] Stroud, M. A. (1971). The behaviour of sand at low stress levels in the simple shear apparatus. PhD thesis,
University of Cambridge
[8] Bolton, M. 1986. The strength and dilatancy of sands. Gotechnique, 36(1): 65-78
[9] Vaid, Y. P., Sasitharan, S. 1991. The strength and dilatancy of sand. Canadian Geotechnical Journal 29:
522526
[10] Vermeer. P. A. & de Borst, R. (1984). Non-associaled plasticity for soils, concrete and rock. Heron
29(3):164
[11] Alshibli, K.A., and Alsaleh, M.I. 2004. Characterizing Surface Roughness and Shape of Sands Using
Digital Microscopy. Journal of Computing in Civil Engineering, 18 (1), January, 36-45
[12] Guo, P., Su, X. 2007. Shear strength, interparticle locking, and dilatancy of granular materials. Canadian
Geotechnical Journal 44: 579
[13] Cox, M.R., and Budhu, M., 2008. A Practical Approach to Grain Shape Quantification. Engineering
Geology, 96 (1), Jan., 1-16
[14] Cox, M.R. 2008. The Influence of Grain Shape on Dilatancy. PhD thesis. The University of Arizona
[15] Muszynski, M. R. 2006. Determination of maximum and minimum densities of poorly graded sands
using a simplified method. Geotechnical Testing Journal, ASTM 29(3): pp.263-272.
[16] Potts, D. M., Dounias, G. T., Vaughan, P. R. 1987. Finite element analysis of the direct shear box test.
Gotechnique 37(1): 11-23
[17] Jewell, R. A., Wroth, C. P. 1987. Direct shear tests on reinforced sand. Gotechnique 37(1): 53-68
[18] Davis, E. H. 1968. Theories of plasticity and failures of soil masses. In Soil Mechanics, Selected Topics
(editeb by I.K. Lee), Butterworth, London
[19] De Josselin de Jong, G. 1976. Rowes Stress-Dilatancy Relation based on Friction. Geotechnique,
London, 26 (3), 527534
[20] Wood, D.M. 1990. Soil behaviour and critical state soil mechanics. Cambridge, Cambridge University
Press
[21] Schanz, T., and Vermeer, P.A. 1996. Angles of Friction and Dilatancy of Sand. Geotechnique, 46 (1),
145-151
[22] Alshibli, K.A., and Alsaleh, M.I. 2004. Characterizing Surface Roughness and Shape of Sands Using
Digital Microscopy. Journal of Computing in Civil Engineering, 18 (1), January, 36-45
[23] Cho, G.C., Dodds, J., and Santamarina, J.C. 2006. Particle Shape Effects on Packing Density, Stiffness,
and Strength: Natural and Crushed Sands. Journal of Geotechnical and Geoenvironmental Engineering,
132 (5), May, 591-602
[24] Atkinson, J.H. and Bransby, P.L. 1978. The Mechanics of Soils. McGraw-Hill, London.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-600
Soil-pile interaction during pile installation.

Observations from x-ray tomography and
3D-DIC
Matias SILVAa,b,1 and Gal COMBEa,b
a
Universit Grenoble Alpes, 3SR, F-38000, Grenoble, France
b
Abstract. The mechanisms behind the installation of a driven pile in sands are
complex and difficult to identify from field tests. Phenomenon such as grain
breakage, local variations in sand density, pile abrasion and arching effects, are
commonly studied through indirect methods or by using small laboratory models.
This article presents the results of an experimental laboratory research devoted to
the analysis of the soil kinematics during the installation of an uninstrumented
close-ended model pile in a specimen of siliceous sand. The research has focused
on understanding the micromechanical aspects of the installation process by
focusing on sand-pile interaction around the piles tip. A three dimensional
displacement field was obtained with the combined use of x-ray tomography and
three dimensional digital image correlations. The results show two well-identified
zones where the displacement and therefore strains concentrate. The interaction of
these zones, which are linked by consecutive phases of sand contraction and
extension as the pile embeds into the sand, gives new ideas about the mechanisms
controlling the pile shaft friction.
Keywords. X-ray tomography; digital image correlation, pile installation,

kinematics.
1. Introduction
The mechanisms behind the installation of driven piles in sands are complex and
difficult to identify from field tests. Phenomenon such as grain breakage, local
variations in sand density, pile abrasion and sand arching effects, are commonly
studied through indirect methods or by using small laboratory models [1], [2], [3], [4].
Robinsky & Morrison [1] were among the first to address the problem of the
response of the surrounding soil during the installation of a small model pile. Using x-
ray radiographies, following lead shot deployed in the sand mass, Robinsky &
Morrison observed the formation of a zone of loose sand around the pile during its
installation. When a pile is pushed into a sand mass, the sand near the piles surface is
dragged down with the pile movement under the constant confinement of the
surrounding soil, resulting in the formation of an arc of sand around the pile. These
conclusions were later confirmed by other authors from density measurements on large
calibration chamber tests [5], photoelastic observations on a plain strain model [3] and
1
Corresponding Author. Laboratoire 3SR, Domaine Universitaire, BP53, 38041 Grenoble Cedex 9,
France E-mail: [email protected]
M. Silva and G. Combe / Soil-Pile Interaction During Pile Installation 601
from x-ray absorption analyses [6], [7], [8]. In addition, results from Particle Image
Velocimetry (PIV) on a plain strain model [4] suggested that a zone of high density
will develop around the pile as the latter embeds deeper into the sand mass. These
results show irrecoverable volume reduction beneath the pile tip as a result of particle
breakage and soil compaction due to the high stresses developed at the advancing pile
tip [3]. As the pile installation progresses into this highly overconsolidated sand [4], a
zone of high density will be created around the pile shaft. The cylindrical cavity
collapse of this highly compacted sand layer may lead to reductions in radial stress and
to the development of an arching effect similar to that observed earlier by Robinsky &
Morrison [1]. Plain strain analyses [4] also showed a series of strain path reversals with
the sand mass around the pile tip experiencing consecutives stages of local compressive
and dilatant phases.
This paper presents the first results of an experimental programme devoted to the
analysis of the kinematics developed in the soil mass during the installation of a model
pile into a siliceous sand specimen using micro tomography (X-ray micro CT) and
three dimensional (3D) digital image correlation (DIC).
2. Experimental arrangements
2.1. General Arrangements
The tests were conducted in the micro computed (CT) tomograph of the laboratory
3SR in Grenoble, France. A pseudo-triaxial cell capable of applying several lateral
confining pressures to a sand specimen during a tomography scan was designed and
built, inside which an uninstrumented 5 mm diameter cone-ended smooth aluminium
(pile roughness of 0.7 m) model pile was installed by monotonic loading.
x-ray source sand specimen
model pile
electric jack
Figure 1: Test arrangements during x-ray micro tomography.

The sample consists of a cylinder 70 mm in diameter and 100 mm in height of dry
Fontainebleau NE34 siliceous sand (see Table 1). Samples were prepared by four
consecutive layers 25 mm high and a density of 1.60 , corresponding to a
relative density, Dr, of approximately 70%. A constant isotropic confining pressure of
100 kPa was applied to the specimen during pile installation and X-ray scanning
602 M. Silva and G. Combe / Soil-Pile Interaction During Pile Installation
through a water-air interface similar to a triaxial test. The pile installation took place
within an hour after pressurization. Important scale and boundary effects are expected
to occur due to the reduced size of the model pile with respect to the confining
boundary and with respect to the mean particle size, d50. If we consider our experiment
as a small pressurised laboratory calibration chamber, a ratio of pile diameter, D, to the
chamber diameter, B, of at least 60 should be considered to avoid boundary effects, as
reported by several authors from experimental [10], [11] and numerical analyses, [12].
In addition, a size effect is also expected from the small ratio of the pile diameter, D, to
mean particle size, d50, [13].
Table 1: Index properties of Fontainebleau NE34 sand
emin emax d10 (mm) d50 (mm) d60 (mm)
0.51 0.90 0.15 0.21 0.23
2.2. X-ray micro CT
Due to the predefined sand sample and pile diameter, scans were set to a tension of
200 kV and a current of 150 A. This configuration allows for a spatial resolution of
approximately 23, and 43 m/pixel for global and local tomographies respectively
(scans of the entire sample volume, or scans focusing on the pile tip). There were 1024
radiographies made of the sand sample for the reconstruction of the entire volume.
Using micro CT, a 3D digital image of the x-ray attenuation in the soil sample and the
pile was obtained. An example of the reconstructed volume during a local tomography
is presented in Figure 2.
Figure 2: Vertical slice across the sand sample during local tomography.
Due to the configuration of the 3SR tomograph's cabin, the pile was installed into
the sand mass from the bottom section of the sample (in an upward movement). The
pile embedded length was limited to 50 mm in order for the pile tip to be positioned
approximately at the middle of the sample. This preparation aimed at avoiding
boundary problems with the bottom part of the sample, with an allowance for
subsequent pile displacement for Digital Image Correlation analysis.
The pile was installed using an electric jack into the sand mass at a rate of 0.1
mm/s, where a first tomogram was taken. The pile head load was kept constant during
the different tests. Several tomograms were then obtained after incremental pile
displacements, ranging from 0.5 to 3.0 mm, to perform further digital image correlation
analysis in order to evaluate the incremental displacement field between consecutive

loading steps.
2.3. 3D-Digital Image Correlation
In order to perform an image correlation between two images, it is first required to

define a regular grid of analysis points (Node Spacing, NS) over the reference image;
then, a group of pixels or correlation window (CW) is defined for each node of the grid.
Image correlation is performed, for each node, by identifying the most similar CW in
the deformed image based on statistical measures of correlation of the images within
each subset, and mapping functions between the subsets. Different measures can be
used to assess the degree of similarity of the subset images. In this particular case,
correlation between images is estimated based on equation 1:

(1)

where x and y are the (2D) spatial coordinates, u and v the displacements of a pixel
from image 1 (I1) to a new position in image 2 (I2). In the case of rigid body
displacements of a subset, u and v, will be constant for all (x, y) positions in that CW.
An analogy of equation 1 is used for the estimation of the correlation coefficient in
three dimensions. Sub-pixel refinement techniques are required, by interpolating the
correlation coefficient between two images by a polynomial or surface function, to
estimate the maximum value which will correspond to a non integer value (lower than
one pixel). DIC analysis is completed with strain calculations, which are based on the
gradients of the derived displacements and a continuum assumption.
The results presented in the following sections were derived using the 3D-DIC
code TomoWarp [14], [15]. The full strain tensor field in the form of volumetric strain
and maximum shear (distortional) strains was derived. This code calculates the strains
from the derived displacements under continuum assumptions. The experimental
programme and the parameters considered for DIC calculations are outlined in Table 2.
Table 2: Summary of experimental programme
Test ID Pile head displacement (mm) NS (pixels) CW(pixels x pixels) Resolution (m)
T1 0.5 10 30 23.57
T2 1.0 10 30 23.57
T3 2.0 10 30 23.57
T4 2.0 7 15 43.55
T5 3.0 7 15 43.55
3. Results
Displacement contours following pile installation are plotted in the vertical and
radial axes with respect to pile position at normalised distances from the pile axis, r/R,
and by the normalised distance to the pile tip, h/R, r and h being the distance to the
pile's axis and to the pile radius, R, respectively.
Since DIC results are fully three-dimensional, the results presented in the
following section represent the mean value of those displacement/strain points at a
specific spatial distance to the pile/tip axis. As a sign convention, displacements are
considered to be positive when moving away from the pile axis. Positive strains
represent extension.
3.1. Soil displacement
Vertical and horizontal displacements were combined to represent the radial

displacements due to the (almost) axisymmetric condition of the experiments. Pile
alignment was considered for the h/R and r/R normalised distances. Figures 3 a) and b)
presents contour plots of the soil radial and vertical displacements respectively during
test T2. The horizontal displacements reach maxima approximately when the pile tip
passes the soil horizon. On the contrary, maximum vertical displacements are attained
before the tip has reached the soil horizon, in this case at a distance of one pile diameter
above the soil level.
In both figures, displacements concentrate in two well-identified zones around the
pile tip. The first one is located beneath the tip and extends downwards 4 times the pile
radius. The sand mass follows the pile movements, spreading away radially as the pile
advances. Within this zone, a bulb in the vicinity of the tip concentrates most of the
vertical displacements. A second zone, in some cases almost completely isolated from
the surrounding sand mass, is observed behind the base of the cone tip at a distance of
1 to 1.5 h/R.
3 -3,0 -1,0
3
-10
-10
-9,0
2 0,50 2
0,0 1,5
2,0
-8,0
5,09,0
7,0 1,0
1,0
1,5
1 5,0 9,0
7,0 1,5
8,0
10
1
8,0
0,50
10
h/R
-2,0
-3,0
h/R
0 0
-10
-6,0 5,0 0,0
7,0
0,0
6,0 -0,50
-1 4,0 -8,0 -1
-7,0
-5,0 3,0 -6,0
-1,5 -1,0
-7,0
-2,0
0,50
0,0
-4,0
-2 -5,0
-3,0 -2
-9,0
-1,0
2,0
-0,50
1,0
-3 1,5 1,0 -4,0 -3
1,0
-1,5
-2,0 1,0
1,0 10
-4 -4
0 1 2 3 4 0 1 2 3 4
r/R r/R Zero displacement
a) b)
Figure 3: Incremental displacement after a 2 mm pile head displacement (values in pixels, 1 pixel = 45 m):
a) radial and b) vertical displacement considering a node spacing of 10 pixels during T3 test.
The soil in this region experiences an upward displacement opposite to the pile
movement. The extension and maximum values are significantly lower compared to
those observed beneath the tip and this zone is not in direct contact with the pile
surface. The extension of these zones is directly linked to the imposed pile head
displacement, in this case of 1 mm (Table 2). During tests T4 and 5, radial and vertical
displacements are observed at distances greater than 6 times the pile radius.
In addition to the identification of displacement contour, it is also possible to

observe an area in both Figures 3a) and b) where no displacements are observed. These
limit lines (segmented green lines in Figures 3a) and b)) mark out the boundary
limits. We note that radial displacements are observed in a wider area radially from the
central pile axis, but to a limited depth. On the contrary, vertical displacements are
visible to a greater depth, but to a lesser extent from the pile axis. Figure 4 presents a
summary for tests T1-T3 of the spatial position of these boundary lines.
2,5
2,0
1,5
Radial displacement
1,0
h/R
0,5
Test:
T1
0,0
T2 Vertical displacement
T3
-0,5
1,0 1,5 2,0 2,5 3,0
r/R
Figure 4: Position of zero radial and vertical displacement contours with respect to h/R and r/R after Figures
3a) and b)
3.2. Strain field
The analysis of strain patterns coming from the measurements of gradients on the
kinematic field during a loading step can be characterised by Figures 5a) and b). A
zone of high vertical contraction and radial extension develops beneath the pile tip as
the pile is continuously pushed into the sand. A zone of high compaction will develop
in front of the advancing tip. Some of the points were removed from the graphs, due to
the large variations observed. Between two consecutives images, the characteristic
pattern required to perform DIC is highly modified due to the irrecoverable changes in
the sand microstructure.
Above this zone and oriented almost perpendicularly to the cone tip edge, the
vertical extension and radial compression develop. These results agree with those
obtained using plain strain models [4].
Figure 6 presents the volumetric strain after a 2 mm pile head displacement during
test T4. The interaction between the different phases of vertical and horizontal
contraction and extension presented in Figure 5, suggests that the soil beneath the tip
undergoes an increase in volumetric strain leading to a lower relative density. It is not
yet completely clear whether similar observations also apply for the sand mass in the
vicinity of the cone edge where high degree of disturbance renders DIC analysis
difficult. Results from Figure 6 are consistent with the analyses from X-ray absorption
on similar specimens [16] which show a reduction in sand density around the pile.
Unlike what is observed in Figure 3 (and summarized in Figure 4 for the soil
displacements), the boundary between the phases of radial/vertical contraction and
extension where no strain develops presents similar orientations with respect to the pile
position. These results confirm what was observed using PIV techniques on a plain
strain model [4].
3 3 -5,0%
0,0% 0,2%
0,0%
-0,5%
2 2
0,0%
-3,0% 5,0%
-3,0% 2,0%
1 -2,0% 1
4,0%
0,2%
-5,0%
4,8% -2,0%
-1,5% 5,0%
-2,0% -0,5%
-1,5%
-0,5%
-0,2%
h/R
h/R
0 3,0% 2,0% 2,0% 0

5,0%
3,8% 0,5% -0,5%
-1,0% -5,0% 1,0%

2,8%
-1 5,0%
-0,2%
-1 0,0%
1,8%
-4,0% 0,3%
0,0%
-2 -4,0% -0,2% -2
-2,0% 0,5%
0,5%
-1,0%
-3 -1,0% -0,5% -3
-3,0%
-0,5%
-2,0%
-4 -3,0%
0,0%
-1,0% -4
0,3%
-1,0% -1,0%
0,3%
-1,5% 5,0%
-5 -0 2% -5
0 1 2 3 4 5 0 1 2 3 4 5
r/R r/R Zero strain
a) b)
Figure 5: Radial (a) and vertical (b) strains after a 2 mm pile head displacement during test T4. Values in %.
3 -0,2% 0,0%
0,5% 0,0% 0,0%
-5,0%
0,0% 0,0%
0,0%
2
-0,5% -1,0%
-0,2%
-5,0% 3,0%
-1,0% 0,0%
-4,0%
1 -1,0%
0,0
-0,5% 0,0%
-1,0%
0,0%
-0,2%
0,0%
3,0% 0,5%
0 3,0% 0,0%
h/R
-3,0% 1,0%
0,5%
-0,2%
4,0%
-1 5,0%
2,0%
1,0%
-1,5%
-2
-1,0% 0,5% 5,0%
-3 -2 0% 0 0% 0,0%
0 1 2 3 4
r/R Zero strain
Figure 6: Volumetric strain after a 2 mm pile head displacement during test T4. Values in %.
Strains tend to develop mostly in the vicinity of the pile tip, in this case, to a
maximum distance of five times the pile radius both vertically and radially. Taking into
account an imposed pile head displacement of almost 20% pile diameter, it is possible
that the flexible-constant pressure boundary condition influences the results, leading to
mainly displacements in the soil without high degree of strains.
4. Conclusions & Perspectives
The combined use of X-ray micro CT with 3D-DIC helps to understand the
kinematics behind the installation of a model pile in a sand specimen. The results show
two clearly distinct zones where vertical and radial displacements concentrate inside
the sand mass. The sand particles on the shaft side of the cone tip seem to undergo
vertical movements opposite to the direction in which the pile is driven.
The analysis of vertical and horizontal strain distribution shows clear trends of
vertical compaction and radial extension for sand particles beneath the pile tip. On the
contrary, the particles aligned perpendicularly to the tip edge behave in the opposite
way, presenting radial compaction and vertical extension.
The results from DIC showed a reduction in density beneath the pile tip and
around the pile shaft. The effect of particle breakage should be investigated within this
zone. Further work should also analyse the effect of lateral pressure and boundary
effects in the observed sand kinematics.
References
[1] Robinsky, E. I., & Morrison, C. (1964). Sand displacement and compaction around model friction piles.
Canadian Geotechnical Journal, 1(2), 8193.
[2] Biarez, J., & Gresillon, J. (1972). Essais et suggestions pour le calcul de la force portante des pieux en
milieu pulvrulent. Geotechnique, 22(2), 433450.
[3] Allersma, H. G. B. (1988). Optical analysis of stress and strain around the tip of a penetrating probe.
Penetration Testing 1988, Vol. 2, 615620, Balkema, Rotterdam.
[4] White, D. J., & Bolton, M. D. (2004). Displacement and strain paths during plane-strain model pile
installation in sand. Gotechnique, 54(6), 375397.
[5] Chong, M. K. (1988). Density changes of sand on cone penetration resistance. Penetration Testing 1988,
Vol. 2, 707714, Balkema, Rotterdam.
[6] Van Nes, J. H. . (2004). Application of computerized tomography to investigate strain fields caused by
cone penetration in sand. (Master's thesis). Delft University of Technology.
[7] Ngan-Tillard, D., Cheng, X. H., Van Ness, J., & Zitha, P. L. J. (2005). Application of x-ray computed
tomography to cone penetration tests in sands. Proceedings of the Geo-Frontiers, Site Characterization
and Modeling, 112.
[8] Morita, K., Otani, J., Mukunoki, T., Hironaka, J., & Pham, K. D. (2007). Evaluation of vertical and
lateral bearing capacity mechanisms of pile foundations using X-ray CT. Advances in Deep
Foundations: International Workshop on Recent Advances of Deep Foundations, 217223.
[9] Jardine, R., Zhu, B., Foray, P., & Yang, Z. (2013). Interpretation of stress measurements made around
closed-ended displacement piles in sand. Geotechnique, 63(8), 613627.
[10] Lunne, T., P. Robertson, and J. Powell (1997). Cone Penetration Testing in Geotechnical Practice.
Blackie Academic, EF Spon/Routledge Publishers, New York.
[11] Bolton, M. D., Gui, M. W., Garnier, J., Corte, J. F., Bagge, G., Laue, J., & Renzi, R. (1999). Centrifuge
cone penetration tests in sand. Gotechnique, 49(4), 543552.
[12] Salgado, R., J. Mitchell, and M. Jamiolkowski (1998). Calibration chamber size effects on penetration
resistance in sand. Journal of Geotechnical and Geoenvironmental Engineering, 124 (9), 878-888.
[13] Garnier, J. (2002). Size effects in shear interfaces. In Constitutive and Centrifuge Modelling: Two
extremes, 334-345.
[14] Hall, S. A. (2006). A methodology for 7D warping and deformation monitoring using time-lapse
seismic data. Geophysics, 71(4), O21-O31.
[15] Hall, S. A., Bornert, M., Desrues, J., Pannier, Y., Lenoir, N., Viggiani, G., & Besuelle, P. (2010).
Discrete and continuum analysis of localised deformation in sand using X-ray uCT and volumetric
digital image correlation. Geotechnique, 60(5), 315322.
[16] Silva, M., (2014). Experimental study of ageing and axial cyclic loading effect on shaft friction along
driven piles in sand. PhD Thesis University of Grenoble.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-608
Investigation of Soil Nonlinearity at Very

Small Strains Using Ground Buried Fiber
Optic Sensors
Balz FRIEDLIa,1 and Alexander M. PUZRIN a
a
Institute for Geotechnical Engineering, ETH Zurich, Zurich, Switzerland
Abstract: Effects of small strain nonlinearity on deformation behaviour of

geotechnical structures have been extensively studied in the literature. This paper
offers a slightly different perspective, looking at effects of the small strain non-
linearity on the identification of surface loads caused by pedestrians and vehicles.
Due to the small load magnitudes caused by these objects, the corresponding
strains in the ground are extremely small, making them difficult to detect using
conventional strain measurements. An experimental field study as been carried out
to investigate deformation behaviour of soil using ground buried distributed fibre
optic sensors (FOS). FOS has been proven an adequate tool for measuring
extremely small strains in soils and capturing accurately the small strain
nonlinearity of soil behaviour. In contrast to conventional discrete strain
measurements at the surface, FOS provides a large array of strain data at higher
accuracy. The deformation of soils at very low stresses and strains has been shown
to be highly nonlinear, affecting the surface load identification.
Keywords: Fiber Optic, Field Tests, Small Strain Nonlinearity
1. Introduction
An object moving on the ground surface can be considered as a mechanical contact

problem. The resulting contact forces cause deformations of the ground. If these
deformations are known accurately and with high spatial resolution the contact forces
acting at the surface can be calculated using mechanical inverse analysis. These contact
forces are correlated to certain objects moving on the surface (e.g., pedestrians, animals,
vehicles, etc.) this procedure will be further referred to as object identification
(Figure 1). It may be used in various applications spreading from perimeter security
with intrusion detection, to weigh-in-motion for airport and road traffic as well as
biomechanics and sports medicine, where contact forces play an important role in
understanding and optimizing movement patterns. Recent advances in the field of
distributed fiber optic sensors (FOS) provide the possibility of measuring the
deformations along of standard fibers with very high precision and spatial resolution
allowing gaining the necessary information for highly accurate object identification.
1
Corresponding Author: Balz Friedli, Institute for Geotechnical Engineering, ETH Zurich, Stefano-
Franscini-Platz 5, 8093 Zurich, Switzerland, E-mail: [email protected].
B. Friedli and A.M. Puzrin / Investigation of Soil Nonlinearity at Very Small Strains 609
contact force acting

on ground surface
ground surface
x y
z distributed strain
measureme nt along
fibre optic cable
buried fibre optic cables
strain field induced by

the force on the surface
Figure 1. Schematic sketch of the idea of object identification.
This concept and the corresponding analysis used as a perimeter security system have
been filed at the European patent office [1].
Effects of small strain nonlinearity on deformation behaviour of geotechnical
structures have been extensively studied in the literature, e.g., [2,3]. This paper offers a
slightly different perspective, looking at the effects of small strain non-linearity on the
identification of surface loads caused by pedestrians and vehicles. Within this context,
accounting for small strain soil behaviour becomes even more challenging and
important. Firstly, compared to conventional geotechnical problems, the load
magnitudes caused by these objects are extremely small, making the corresponding
strains difficult to detect using conventional strain measurements. Secondly, the surface
load identification uses inverse analysis, which requires even higher accuracy of the
mechanical ground model for convergence of its optimization algorithms than the
forward analysis. Finally, in contrast to conventional geotechnical problems where
small strains contribute only to a certain fraction of the total displacement, in the object
identification problem the strains are very small within the entire soil domain, so that a
wrong estimation of the small strain stiffness has a much higher impact on the solution.
This paper presents results of an experimental field study attempting to quantify
the small strain non-linearity within the context of object identification.
2. Field Experiment
2.1. Experimental Setup
The field test site was constructed to investigate the influence of the small strain non-
linearity on the problem of object identification. The FOS are buried in the natural soil
610 B. Friedli and A.M. Puzrin / Investigation of Soil Nonlinearity at Very Small Strains
Figure 2. Layout of fiber optic cables in the field experiment. (a) Ground view. (b) Cross section. (c) Photo
of built in cables. (d) Photo of the refilled trench with grass.
classified as sandy silt with a small amount of clayey fines. The site has several FOS
buried in a shallow trench excavated in a lawn placed at the depth of 0.2 m in five
parallel lines at the distance of 25 cm to each other (Figure 2). The FOS were
prestrained and protected by a thin sand layer before the trench was refilled with the
excavated material. One end of the FOS was placed in a shaft, where it was connected
to the strain sensing interrogator; the other end remained buried in the trench. The
backfill was then re-compacted to its original density using a vibrating plate. After a
few weeks grass grew over the excavated trench making it indistinguishable from the
surrounding lawn.
The FOS used in the present study consist of a standard single mode fiber
protected by plastic coating. The coating is tight buffered to ensure good transfer of
strain from the coating to the fiber core.
2.2. Strain Sensing Device
Distributed strain measurements along of the FOS were taken using swept wavelength
interferometry (SWI) measuring Rayleigh scattering [4]. A change of strain or
temperature causes a spectral shift in the Rayleigh backscatter, which can be measured
and correlated to the change of strain or temperature along the cable. This interrogation
technique allows measuring strain along standard fibers with very high precision
(N1 ) and high spatial resolution (5 mm) on sensor lengths of up to 70 m.
Figure 3. (a) Ground view of the loading assembly. (b) Photo of the loading assembly.
In measurements using distributed FOS, strain and temperature effects cannot be

separated. Furthermore, the measurements can be influenced by environmental effects
such as changes of humidity which cause swelling of the plastic coating of the sensor.
However, in the present study only short term measurements are used (i.e., the time
between the reference measurement and the measurement under load is short), resulting
in negligible environmental influences and the temperature can be regarded as constant.
2.3. Loading Assembly
The ground surface was loaded using a plastic barrel filled with water and placed on a
circular aluminium plate of a 10 cm diameter, which can be regarded as a point load.
The plate was placed directly over the cable 1 and the plastic barrel was horizontally
stabilized with wooden girders (see Figure 3) supported more than 3 m away to ensure
that there is no interference with the load of the barrel. Between the horizontal
stabilization girders and the barrel was a small gap to ensure that the whole load of the
filled barrel pushed on the aluminum plate and is not transferred via the wooden
girders. During the experiment the barrel was loaded with water in steps of 5 kg from
the barrels self-weight (8.2 kg) up to 100 kg. The water was brought into the barrel
using a wooden bridge located next to the horizontal stabilization girders. To ensure the
accuracy of the loading steps, the weight of the water was taken before filling it into the
barrel. After each loading step a strain measurement was taken.
3. Results
In Figure 4 the strain measurements for 10 loads between 10 kg and 100 kg are shown
for cable 1, located at a depth of 0.2 m directly under the load. The position of the load
is at the cable coordinate x = 0.0 m. Only a half of the measured practically symmetric
strain profiles is presented. The strain profile along the cable has a distinct shape with a
tension zone directly under the load and a compression zone further away. The zero
strain is reached at about 0.1 m from the center of the load and remains constant for all
the loads. The maximum measured strain ranges from N11 to N305 at loads of
10 kg to 100 kg, respectively. The corresponding maximum horizontal displacements,
300 100kg
90kg
250
80kg
200
measuredstrain ()
70kg
150
60kg
100 50kg
40kg
50 30kg
20kg
10kg
0
-50
-100
-1 -0.8 -0.6 -0.4 -0.2 0
cable coordinate(m)
Figure 4. Measured Strain in cable 1 (directly under the load at the cable coordinate x=0.0m) for 10
measurements between 10kg and 100kg load, only one half.
calculated by integration of the measured strains along the cable are 7.2O10-4 mm and
0.024 mm, respectively. It appears that the strain profile is simply scaled with
increasing load, but not proportionally to this increase. Indeed, looking at the maximum
measured strain at the coordinate 0.0 m it can be seen that the distance between the
consecutive strain measurements becomes larger with increasing strain even though the
load increments remain constant. This confirms that the stiffness of the soil decreases
with increasing load.
To illustrate the decreasing stiffness with increasing load, the maximum measured
strain for each load step is plotted against the load magnitude in Figure 5 a. It can be
seen that for smaller loads the stiffness seems to decrease very fast with increasing
load. At higher loads the stiffness appears to approach a constant value. Proper
mechanical modelling requires a better understanding of the stress-strain behavior of
soil. For this purpose an equivalent Youngs modulus E and an equivalent Poisson ratio
were back calculated for each load step using as input the strain measurements of
cables 1, 2 and 3 for each load step. The strains calculated using the Boussinesq
solution [5] for the linear elastic half space loaded by a point load at the surface were
calibrated against the measured strains using a non-linear least squares optimization
algorithm. The resulting Youngs moduli represent the secant stiffness. The calculated
equivalent elastic parameters for each load step show a strong decrease in stiffness with
increasing load whereas the Poisson ratio remains more or less constant over the load
steps considered (Figure 5 b). Although the approach of calculating equivalent linear
elastic stiffness has obvious theoretical limitations (i.e., it assumes the same stiffness in
the whole half space) it is useful in illustrating the influence of the soil nonlinearity on
the evolution of strains caused by small surface loads.
(a) 1 (b) 60 0.5
Youngs modulus (MPa)

0.8 50 0.4
load magnitude (kN)
Poisson ratio (-)

0.6 40 0.3
0.4 30 0.2
20 E 0.1
0.2

0 10 0
0 100 200 300 0 0.25 0.5 0.75 1
max. measured strain () load magnitude (kN)
(c) 150 (d)

300
measured c1 100kg
90kg
calculated strain ()
100 calculated
200 80kg
strain ()
70kg
50 60kg
100 50kg
40kg
c2 30kg
20kg
0 c3 0 10kg
-50 -100
-1 -0.8 -0.6 -0.4 -0.2 0 -1 -0.8 -0.6 -0.4 -0.2 0
cable coordinate (m) cable coordinate (m)
Figure 5. (a) Load magnitude vs. max. measured strain in cable 1 directly under the load. (b) Back calculated
equivalent Youngs moduli and Poisson ratios for varying load magnitude. (c) Comparison between
measured and calculated strains for cable 1, 2 and 3 (denoted as c1, c2 and c3) for a load of 60kg. (d)
Calculated strain profiles with the elastic parameters from (b) for the same 10 loads as in Figure 4.
In Figure 5 c the comparison of the calculated and the measured strains are shown
for a load of 60kg, where the linear elastic model proves to capture accurately the strain
profile for this particular load magnitude. Figure 5 d shows the calculated strains with
the elastic parameters from Figure 5 b for the same loads as in Figure 4. While the
resemblance of the strain profiles is striking, it is only possible to achieve by using the
strongly decreasing stiffness shown in Figure 5 b. This finding confirms that the soil
nonlinearity at the very small strains occurring in the ground caused by small surface
loads is an important phenomenon for the identification of surface loads. The objective
of the ongoing research is to use nonlinear elastic and elastic-plastic kinematic
hardening constitutive models (e.g., Puzrin and Burland model [3]), which are capable
of capturing accurately the observed small strain nonlinearity of soil behavior, in order
to achieve a more accurate identification of surface loads.
4. Conclusions
Ground buried fiber optic sensors (FOS) have been used in a field test to measure the
strains in the ground caused by small surface loads within a context of the problem of
identification of surface loads caused by pedestrians and vehicles. Within this context,
accounting for small strain soil behaviour becomes even more challenging and
important than for conventional geotechnical problems due to the following reasons:
+ Load magnitudes caused by these objects are extremely small, making the
corresponding strains difficult to detect;
+ Inverse analysis, requires higher accuracy of the mechanical ground model for
convergence of its optimization algorithms than the forward analysis;
+ The strains are very small in the entire soil domain, with the small strain
stiffness having a much higher impact on the solution.
The high precision and spatial resolution of distributed FOS allowed for accurate
investigation of the strain evolution caused by changing surface loads. While a linear
elastic half space model was capable of capturing the shape of the measured strain
profile, it could not reproduce measured strains for a range of surface loads using the
same set of elastic constants. For the inverse analysis of surface loads using strain
measurements of buried FOS this highly nonlinear soil behaviour proves to be critical.
The test results presented in this paper will serve for the calibration of nonlinear elastic
and elastic-plastic kinematic hardening constitutive models (e.g., Puzrin and Burland
model [4]) simulating small strain nonlinearity of soil behavior. This will lead to a
more accurate surface load identification which will find application in a wide field
spreading from perimeter security with intrusion detection, to weigh-in-motion for
airport and road traffic as well as biomechanics and sports medicine, where contact
forces play an important role in understanding and optimizing movement patterns.
5. Acknowledgement
The authors would like to thank Simon Vogler for his support on the experimental
study during his Master Thesis and the technical support of Dominik Hauswirth
concerning fiber optic sensors.
References
[1] A. M. Puzrin, D. Hauswirth, B. Friedli, Fibre optic based intrusion sensing system, European Patent
PCT/EP2014/001006, filed October 17, (2014)
[2] J. B. Burland, Small is beautiful - the stiffness of soils at small strains. Canadian Geotechnical Journal
26 (4), 499-516 (1989).
[3] A. M. Puzrin, J. B. Burland, Non-linear model of small-strain behaviour of soils. Gotechnique 48,
(1998), 217213.
[4] D. K. Gifford et al., Distributed Fiber-Optic Temperature Sensing using Rayleigh Backscatter, 31st
European Conf. and Exhibition Optical Communications, Glasgow, Scotland, 3, (2005), 511-512.
[5] J. Boussinesq, Application des potentiels l'tude de l'quilibre et du mouvement des solides
lastiques :, Gauthier-Villars, (1885).
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-615
Behavioural features of fissured clays:

experimental evidence and modelling
Federica COTECCHIAa and Claudia VITONE a,1
a
Technical University of Bari, ITALY
Abstract. Fissured clays are widespread all over the world and they are often
catalysts of several and unexpected in-situ problems. In the last decade, a research
started at the Technical University of Bari aiming to interpret the influence of
fissuring on the clay behaviour. According to the approach adopted, a Fissuring
IDentity (F-ID) has been associated to each clay being studied, as extracted from a
general fissuring characterisation chart. Hence, the analysis of the fissured clay
behaviour has benefited from systematically coupling the micro- to meso- features
(i.e., F-ID charts) and processes, and the observations of the macro-response. The
comprehensive experimental programme carried out on several Italian fissured
clays has allowed to recognise that they can be still modelled as single
geotechnical class which follows an extended sensitivity framework of behaviour.
Moreover, for clays of specific F-IDs, most recent developments of the research
have put in evidence discrepant behavioural facets induced by scale effects.
Keywords. fissured clay behaviour, sensitivity framework, scale effect, REV.
1. Introduction
In the last century, intensive experimental geotechnical studies were addressed to

characterise the decayed strength of fissured clays. They prompted the development of
two main streams of research: the behaviour very highly fissured clays on one side [e.g.
1-3] and that of stiff clays and weak rock masses of low fissuring intensity on the other
[e.g. 4-8]. Nonetheless, no definite general assessment of the mechanical behaviour of
fissured clays has been pursued to date, probably for the apparent necessity of using
different methodological approaches to the interpretation and modelling of soils
depending on their fissuring features [8-10]. The present paper represents a further step
of an on-going experimental research that aims to interpret and model the general
influence of fissuring on the mechanical behaviour of clays [11]. According to the
approach adopted, a Fissuring Identity (F-ID) has been associated to each clay being
studied, as extracted from a general fissuring characterisation chart. The analysis of the
influence of the F-IDs on the clay behaviour has been carried out through element
testing in the laboratory, analysis of the results in the framework of Critical State Soil
Mechanics (CSSM, hereafter) and modelling according to traditional elastic-plastic
theory. In this paper, the behaviour of four fissured clays outcropping within the
Apennine chain in the South of Italy (Figure 1a) is analysed in the light of the
framework of behaviour of unfissured clays, either natural or reconstituted [12-13]. The
experimental results show that, despite their different F-IDs, fissured clays can be still
modelled as single geotechnical class through an extended sensitivity framework of
behaviour. However, the most recent studies have allowed to both recognise discrepant
behavioural facets induced by size effects and to identify the triggering F-IDs.
1
Corresponding author.
616 F. Cotecchia and C. Vitone / Behavioural Features of Fissured Clays
2. Origin, physical properties and fissuring characterisation

Fissuring is the result of the localisation of intense deformation th at develops
within the soil during its loading history as a consequence of several geological
processes [14]. This is also the case of the fissured clays located within cha in areas, as
those of reference in this study. The most intensely fissured clays being i nvestigated
are the scaly clays from Santa Croce di Magliano (SCM hereafter; Figures 1a-b) and
Senerchia (SEN hereafter, Figures 1a-c) which are part respec ively of the Red Flysch
Formation (Oligocene-Miocene) and the Variegated Clay Formatio n (Upper
Cretaceous- Lower Miocene). The less intensely fissured clays are th e Chattian-
Langhian bentonite clay (BENT clay hereafter; Figures 1a-d) also sampled a t SCM and
the stiff fissured clay f om San Giuliano di Puglia (SGP hereafter; Figures 1 a-e) which
belong to the Toppo Capuana Formation (Upper Miocene).The physical prroperties of
the clays were determined in the laboratory according to ASTM standards . The clay
fraction, CF, of SEN and SCM scaly clays, is on average 59% and 91% r espectively,
whereas CF values of 66% and 47% have been measured for the BENT cl ay and the
SGP clay. The BENT clay has the highest activity, A=1.3 (A=0.57-0.60 f or the other
three clays). Almost all the samples can be classified as CH, according to t he Unified
Soil Classification System. The mineralogical composition, measured by x-ray
diffraction, is mainly characterised by high contents of illite-smectite mixe d-layer clay
minerals with high percentage of smectite in both the scaly c ays and the SGP clays,
whereas pure smectite is predominant in the BENT clays. The first model r equirement
accomplished with in this research has been the definition of a rational chara cterisation
of fissuring, that is shown in Figure 1f in its most recent version.
e)
Figure 1. a-e): Sampling sites and photographs and F-IDs of the clays. f): General characterisattion chart of
fissured clays. F-IDs: SEN scaly clay (), SCM scaly clay (*), BENT (**), SGP marly cl ay (+).
F. Cotecchia and C. Vitone / Behavioural Features of Fissured Clays 617
This chart allows to convert the soil meso-fabric in a set of parameters, i.e. the F-
ID of the material, to which specific behavioural features can be related. It accounts for
both the lithology and consistency of the soil matrix (categories A and B) and the
discontinuity features: fissuring nature (C, D, E), orientation (F) and geometry (G-H-I).
Fissuring intensity can be characterised by either the volume of the inter-fissure clayey
elements (i.e. the inter-fissure volume, VI), or by their specific surface. The research
entailed the investigation of clays of different fissuring features whose F-ID envelope
is represented in grey in Figure 1f. Moreover, in the Figure, the F-IDs of the four clays
under study have been distinguished by specific keys. Experimental results showed that
the categories that influence most the fissured clay behaviour are both the fissuring
orientation (category F) and the intensity (category I; Figure 1f). Both SEN and SCM
clays (i.e. the most intensely fissured clays being studied) are I6 clays and the fissures
split the clay into millimetre-sized lens-shaped elements. They are known as scales,
which is the origin of their being referred to as scaly clays in the literature. Differently,
BENT and the SGP clays are of I5 and I4 intensity respectively, and fissuring confines
elements of centimetre size and prismatic shape (Figures 1d-e). In addition, reference is
made solely to either random, F3 (SGP, SEN and BENT clays), or single, F1 (SCM
scaly clay), fissuring orientation. The mechanical characterisation of the SCM scaly
clay, discussed in the following, has been carried out with fissures oriented normally to
the direction of the maximum principal stress.
3. Mechanical characterisation
Figure 2a shows the results of the restrained-swelling oedometer tests carried out
on the natural specimens of the four clays up to medium-high vertical effective
pressures. The Figure also shows the oedometer tests on the corresponding samples
when reconstituted in the laboratory following Burland [6]. Despite their different F-ID
charts, all the clays being studied follow a single framework of behaviour in
compression. In particular, the Intrinsic Compression Lines (ICLs; [6]) of the
reconstituted clay specimens, are on the right of the compression curves of the natural
clay up to high pressures. In Figure 2a an arrow indicates the gross yield state of the
fissured clays under study. It has been identified by making reference to the highest
gradient in compression index (Cc) increase. Figure 2b shows the variation in specific
volume, , and mean effective stress, p, experienced by 38 mm diameter specimens of
BENT clay during consolidated isotropically undrained shearing (CIU) triaxial tests.
The data show that, during isotropic compression, BENT clay still follows the same
pattern of behaviour recorded in one-dimensional compression. The contractive nature
of the shear response of the specimens isotropically consolidated to -p states to the
left of the isotropic normal consolidation line of the reconstituted clay, INCL*,
confirms that wet shearing behaviour and gross yielding occur on the left of INCL*. It
follows that, as in oedometer compression, the INCL of the natural clay is located on
the left of the INCL* up to medium-high pressures also in the -p plot. The undrained
shear state paths of the BENT clay are shown in the q-p plane in Figure 3, where q is
the deviator stress and both q and p are normalised for by means of the equivalent
pressure pe* taken on the INCL* (see Figure 2b). The data confirm that, according to
CSSM, dry behaviour is exhibited by specimens swelled to overconsolidation ratios R
(= py/p) higher than 2. Moreover, the q-p states possible for the BENT clay are
located inside the state boundary surface of the reconstituted clay, SBS*, and the size
of the SBS of the natural clay is smaller than SBS*. The normalised stress paths of the
specimens consolidated beyond gross yield before shearing (e.g. specimens BENT-B
and BENT-C in Figures 2b and 3), confirm that the wet side of the SBS of the fissured
clay increases in size with compression (i.e. the size of SBC1 is smaller than SBC2 and
SBC3).
3.1 The extended sensitivity framework
The results in Figures 2a-b and 3 have been found to be common not only to the
four fissured clays here of reference, but also to the clays of F-IDs that cover the grey
envelope in Figure 1f. This means that the general effect of fissuring on the clay
behaviour can be derived as the effect of a new internal state variable. To this purpose,
Figures 4a-b compare the sketches of the behavioural trends recognized in compression
and shear for the fissured clays, with those applying to intact clays, either natural or
reconstituted. Figure 4a shows that fissured clays reach either one-dimensional or
isotropic compression states on the left of the normal compression line of the
reconstituted clay (i.e. ICL or INCL*, respectively) up to high pressures, differently
from sensitive unfissured soils. This reveals that fissuring bars out the space on the
right of the ICL, that is the so-called structure permitted space [12], regularly stated as
attainable by all structured soils. From what above, this space should be renamed as
unfissured structure permitted space. Cotecchia & Chandler [13] quantified the
influence of microstructure on the clay behaviour through the stress sensitivity ratio,
S = y/*e (where *e is the equivalent pressure on the ICL). S is equal to one by
definition for the reconstituted clay.
The Authors found that S values higher than one characterise unfissured sensitive
clays. The results here presented suggest that fissuring generates the common
geotechnical effect of making the clay under-sensitive, that is characterised by S
values below one, i.e. even lower than those of the same clay when reconstituted.
The sketch in Figure 4b shows that fissuring is also detrimental to strength, even
with respect to the reconstituted clay. Consistently with S<1, the SBS of the fissured
clays is smaller than SBS*, irrespective of any other F-ID feature. The increase in size
of the SBS of the natural fissured clay, reveals that structure strengthening is on-going
with post-gross yield compression, giving rise to a positive hardening contribution.
Furthermore, for fissured clays, S St and, as such, the sensitivity framework can be
extended to represent the influence of the fissured clay structure on the gross yield
surface (or SBS) of the clay.
3.2 Discrepancies from the framework: REV and size effect
New recent experimental investigations have been carried out in order to focus on
the effect of the specimen size on the clay behaviour, in the light of its F-ID. Figures 5a
shows the same data reported in Figure 2b together with new data of CIU triaxial tests
carried out on three specimens of larger size, i.e. 50 mm diameter - 100 mm high
(BENT-F50) and 70 mm diameter- 140 mm high specimens (BENT-G70 and BENT-
H70). Figure 5a clearly shows the overlapping up to high pressures of the compression
curves of standard triaxial specimens (BENT-B and BENT-C) and of the larger
specimen BENT-H70.
ICL F3-I5 F1-I6

Y
1.8 F3-I6
F3-I5
1.6 F3-I4
1.4
1.2 ICL F3-I4
e
Contractive (wet)
1 ICL F1-I6 Dilative (dry)
ICL F3-I6 Y
0.8
0.6
Y
0.4
0.2
0
10 100 1000 10000 100000
'v (kPa) p*e
Figure 2. Oedometer compression curves of natural and reconstituted (dashed lines) I6 to I 4 clays (left);
BENT clay (38 mm diameter specim
p engs): isoytropic compression and undrained shear state paths (right).
SBS*
0.4
/p* e
p
/
q
0.3
0.2
0.1 SB C3
SB C2
SB C1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
p* e
p' /p
Figure 3. Natural and reconstituted BENT clay (38 mm diameter specimen ): normalised st ress paths.
Un ssureduncl ayssu ed
e Inftiact clay
Sensitive fi r
I6-I4
I6-
I6-I4 lcay
lay
5cclay q/p*e Intact clay
ICL
ICL tuted clay
Reconstituted
Recontsitituted
Recons clay
clay I6-I4 clay
Y Reconstituted clay
post -gross yield

S > 1
STRUCTURE
R PERMITTED SPACE
Y (Leroueil & Vaughan, 1990)
S < 1
SBS*
pre-gross yield

ee** yy
yy
ee**
v (kPa)
[pe*] [py] [py] [pe*] p/p*e
[p(kPa)]
Figure 4. Framework of compression (left) and shearing (right) behaviour of fissured and in tact clays.
These results are representative of similar others carried out on SCM scaly clay
and SGP clay specimens of different size. In the limit of the grey envelop e of F-IDs
represented in Figure 1f, it may be concluded that the framework of c ompression
behaviour in Figure 4a is not size dependent. In Figure 5b the q-p stress pat h followed
by the larger specimens in Figure 5a are shown together with those of t he standard
specimens isotropically consolidated up to the same p values. The figure also shows
the pictures of the specimens at the end of the tests. Contrarily to what observed in
compression, the specimen size affects the peak strength of the BENT clay. The larger
specimens, either sheared pre-gross yield (BENT-F50, BENT-G70) or post-gross yield
compression (BENT-H70), exhibit higher strength than expected if compared to the
corresponding data from standard size specimens (Figure 5b). Recording these
discrepancies suggests that the standard specimen size, 38 diameter by 76 mm high, is
not the Representative Element Volume (REV, hereafter) of the BENT clay. However,
although the bigger is the specimen size, the higher are the peak strength values of the
natural clay, the behaviour in undrained shear of larger specimens is similar to that of
the standard ones having the same R (=py/p). Therefore, the specimen size seems to
influence just the value of the strength parameters, because no modifications are
recorded on the pattern of the fissured clay behaviour. Moreover, the fissure surfaces in
the photographs became visible to the naked eye on the natural clay specimens since
shear strains, s, about 3.0 6.0%, i.e. near q/p peak ratio for almost all the clay
specimens. This allowed the BENT clay specimens to exhibit predictable behavioural
patterns that are consistent with their overconsolidation ratios although an effect of the
specimen size on the clay peak strength is recorded. Similar results have been found for
the other F-IDs (I5 to I4 and F1 to F3) on 38 mm diameter specimens, whereas no size
effect even on the peak strength has been recorded on I6 clay specimens.
If the REV is the minimum specimen size above which no variation of mechanical
behaviour is measured for increasing specimen size, it can reproduce the in-situ soil
behaviour both quantitatively and qualitatively. It follows that 38 mm diameter
specimen is the REV only for I6 clay specimens. For I5 to I4 and F1 to F3, the
experimental results provide indication that this specimen size, although smaller than
the REV, is still useful to qualitatively analyse the main patterns of mechanical
behaviour of the fissured clay. One might conclude that for these F-IDs, 38 mm
diameter specimens represent a Qualitative REV (Q-REV hereafter). It is worth noting
that Q-REV still complies with the definition of representative sample provided by [15],
as homogeneously heterogeneous material that should contain a sufficient number of
the patterned cells that the variations within the cells are smeared out. Figure 3 also
shows that the normalised effective stress path of the standard specimen BENT-D does
not join the boundary curve SBC1 of the I5 clay pre-gross yield. Moreover, the friction
angle at peak stress of specimen BENT-D equals the residual one, that is r = 5
(Bromhead ring shear measurement [16]). Only for this specimen a single failure
surface inclined 40 to the horizontal was already visible since s much lower than 3%.
This behaviour has to be imputed to the presence of a pre-existing single fissure which
crossed the specimen, making change its F-ID. In particular, the maximum inter-fissure
volume (VI) increases, so that the specimen becomes representative of a I3 clay. It
follows that, in this case, the specimen size becomes too small to be even considered a
Q-REV. Similar results have been found by Marsland [5] during triaxial tests on 38
mm diameter specimens of London clay (i.e. I5-I3 clay). In particular, similarly to what
observed for BENT-D specimen (Figures 2b and 3), lowest strengths are measured on
specimens containing a single fissure (i.e. I4-I3 specimens) which made the specimen
far smaller than REV. Also Ward et al. [4], when testing London clay specimens at
Ashford Common, found that, in general, weaker specimens failed at smaller strains,
suggesting that the lower strengths are associated with planes of weakness, probably
fissures too fine to be seen (and that are making vary the F-IDs of these clay
specimens). The Authors consider that the scatter in the strength results is much more
likely to be associated with the degree of fissuring of the clay i.e. category I in Figure
1f). It follows that, the strength of any specimen depends on the propor tion of the
rupture plane which coincided with a fissure and varies between a lower l imit where
the rupture occurred entirely along a fissure (the fissure strength), and an upper limit
corresponding to the intact clay strength. In this respect, full field analyses allowed to
recognize the physical processes in the background of distinctive charac ters of the
fissured clay response. In particular, Vitone et al. [16-17] applied Dig ital Image
Correlation (DIC) during biaxial tests on several fissured clay specimen s of F-IDs
included into the grey area in Figure 1. When the specimen is much small er than Q-
REV (Figure 6a), strain localisation is limited solely to few regions, which evolve little
and control the overall response up to large strains, bringing about a loss of strength
and reduced softening. Since an early stage of shearing, strain localisati on regions
become the location of a slip surface where the onset of sliding occurs an d interrupts
the mobilisation of strength in the specimen. It can be inferred that similar l ocalisation
phenomena occurred in the triaxial specimen BENT-D, making its stress p ath stop far
inside the SBS. On the contrary, when loading either Q-REV or REV speci mens of the
fissured material, strain fields are similar to those of REV specimens of unfi ssured soils
and include both diffuse straining and the activation of various region s of strain
localisation before peak strength (Figure 6b). It is only fter one of the strain
localisation regions has won the competition and governs the global res ponse, that
significant softening ensues and a complete shear band develops across the s pecimen.
Figure 5. BENT clay specimens (standard and larger size): isotropic compression (left ), stress paths
and photographs of the specimens at the end of the tests (right).
I4- I3 clays
(38 mm diameter specimens )
I6- I4 clays
m diameter spe cimens)
(38 mm
a) b)
Figure 6. DIC-derived strain fields and sketch of global behaviour: a) specimen smaller than Q-REV;
b) Q-REV and REV specimens.
4 Conclusions
Despite their different geological histories and F-IDs, fissured clays represent a single
geotechnical class of materials whose behaviour at the macro-scale still follows the
sensitivity framework in compression and shear. However, if for unfissured clays the
natural (micro-)structure adds bonding to the reconstituted clay as evolving internal
factor and provides the natural clay of a temporary additional strength, the natural
(micro- to meso-)structure of fissured clays makes their stress sensitivity become even
lower than that of the reconstituted. Moreover, new developments of the research have
demonstrated that this framework of behaviour holds not only when the REV is tested,
but also for specimens smaller than the REV, but large enough to qualitatively
represent the soil behaviour (i.e. the Q-REV). Within the Q-REV, strain localisation
processes are still similar to those of the REV up to the onset of the sliding mechanism,
i.e. they are diffuse pre-peak and become more confined about peak. It follows that the
specimen still maintains its load-carrying capacity and significant softening occurs only
after one of the shear bands takes control. It has been observed that 38 mm diameter
specimens are REVs for I6 clays and Q-REVs for I5-I4 clays. They are too small to be
even the Q-REV for I4-I3 specimens, so that, in these cases, early sliding occurs the
more frequently the lower is their specimen/inter-fissure volume ratio (V S/VI).
References
[1] A.G.I., Some Italian experiences on the mechanical characterisation of structurally complex clay soils.
Proc. 4th Int. Cong. Int. Soc. Rock Mechanics, Montreux, Switzerland, 1, 827846, 1979.
[2] R. Fearon, M.R. Coop, Reconstitution - what makes an appropriate reference material?, Gotechnique
50(4) (2000), 471-477
[3] F. Cotecchia, F. Santaloia, Compression behaviour of structurally complex marine clays. Proc. Nakase
Memorial Symp. on Soft Ground Engineering in Coastal Areas, Japan, 63-72, 2003.
[4] W.H. Ward, A. Marsland, S.G. Samuels, Properties of the London clay at the Ashford Common shaft: in-
situ and undrained strength tests, Geotchnique 15(4) (1965), 321-344
[5] A. Marsland, The shear strength of stiff fissured clays. Roscoe Mem. Symp., Cambridge Un, 5968, 1971.
[6] J.B. Burland, On the compressibility and the shear strength of natural clays, Gotechnique 40(3) (1990),
329378.
[7] D.W. Hight, A. Gasparre, S. Nishimura, N.A. Minh, R.J. Jardine, M.R. Coop, Characteristics of the
London Clay from the Terminal 5 site at Heathrow airport, Gotechnique 57(1) (2007), 318.
[8] M. Kavvadas, Hard soils soft rocks: modelling the behaviour and selection of model parameters,
General report, 2nd Int. Symp. Geotech. Hard Soils Soft Rocks, Naples (Italy), 14411482, 1998.
[9] D.W. Hight, S. Leroueil, Characterisation of soils for engineering purposes, Proc. Int. Workshop on
characterisation and engineering properties of natural soils, 1, 255362, 2003.
[10] A. Gens, On the hydromechanical behaviour of argillaceous hard soils-weak rocks, Keynote Lecture.
Proc. XV European Conf. soil mechanics and geotechnical engineering, Athens (Greece), 71-118, 2011.
[11] C. Vitone, F. Cotecchia, The influence of intense fissuring on the mechanical behaviour of clays,
Geotchnique 61(12) (2011), 1003-1018
[12] S. Leroueil, P. Vaughan, The general and congruent effect of structure in natural soils and weak rocks,
Gothecnique 40(3) (1990), 467488
[13] F. Cotecchia, R.J. Chandler, A general framework for the mechanical behaviour of clays, Gotechnique
50(4) (2000), 431-447
[14] F Cotecchia, P. Lollino, F. Santaloia, C. Vitone, G. Mitaritonna, Deterministic landslide hazard
assessment at regional scale. Geotechnical Special Publication, Vol. 199, Geoflorida (2010), 3130-3139
[15] D.M. Wood, Heterogeneity and soil element testing, Gotechnique Letters 2 (2012), 101-106
[16] C. Vitone, F. Cotecchia, G. Viggiani, S.A. Hall, Strain fields and mechanical response of a highly
fissured bentonite clay, Int. J. Numer. Anal. Meth. Geomech. 37(11) (2013), 1510-1534
[17] C. Vitone, G. Viggiani, F. Cotecchia, S.A. Hall, Localized deformation in intensely fissured clays
studied by 2D digital image correlation, Acta Geotechnica 8(3) (2013), 247-263
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-623
Selection of Design Friction Angle

Based on Soil and Project
Characteristics
Emirhan SANCAK and Ozer CINICIOGLU1
Bogazici University, Istanbul, Turkey
Abstract. The aim of this study is to develop a method that would aid design
engineers in the selection of project based design friction angles for cohesionless
soils. In this context, design friction angle is defined as either one of peak or
critical state friction angles. In line with this objective, shear strain at which peak
friction angle is mobilized, is quantified as a function of soil properties. Then, it is
necessary to relate allowable structural deformations to magnitudes of possible
shear strains through numerical modelling and make an assessment on which
friction angle to be used for the design involved. To develop this approach, data
from triaxial compression tests under drained conditions conducted on 8 different
sands are investigated. These data were used to construct relationships between
peak dilatancy angle and shear strain at which these peak dilatancy angles are
measured. Obtained relationships are combined with a novel equation that relates
dilatancy angle to soils state. Then the results were evaluated using the results
from 1 g small scale retaining wall model tests that were conducted on a
completely different sand.
Keywords. Retaining wall, friction angle, Particle Image Velocimetry, dilatancy

angle, strain.
1. Introduction
Friction angle is regarded as the main shear strength parameter for cohesionless soils.
Owing to the granular nature of cohesionless soils, two different limiting values of
friction angles are used by geotechnical engineers: peak friction angle and critical state
friction angle. Selection of the design friction angle between these two options
constitutes a compromise between safety and economy. However, there is no
quantitative way for making that selection, and the decision solely rests on engineers
sound judgment. Several researchers have clearly shown that the difference between
peak and critical state friction angles result from the dilatant properties of soils [2, 3, 4,
5, 6, 7]. Powrie (2004) noticed that peak strength state where the peak friction angle
(p) is mobilized differs by the density and the stress state throughout the cohesionless
soil mass. Angle of dilatancy embodies the influences of both soil density and stress
state [9]. As it is very well known, as the soil deforms by shearing, dilatancy is
mobilized and the maximum value of dilatancy rate is measured at the instance of peak
failure. The authors experimentally observed that the magnitude of shear strain that
corresponds to the mobilization of peak friction angle (q-f) is influenced by density,
1
Corresponding Author: Ozer Cinicioglu, Civil Engineering Department, Bogazici University, 34342
Istanbul, Turkey, E-mail: [email protected]
624 E. Sancak and O. Cinicioglu / Selection of Design Friction Angle
stress state and gradation characteristics. Therefore, this study starts with the premise
of linking peak dilatancy angle (p) and gradation characteristics to shear strain at
which p mobilizes. Once q-f can be estimated, it would be possible to compare it to
the possible shear strains that would develop at the permissible deformation level of the
proposed structure through numerical modelling.
To achieve this goal, triaxial compression test results of several different sand
samples were investigated. Each sand was tested in numerous different combinations of
confining pressure and density. It is noticed that there is a semi-logarithmic relationship
between p and q-f for each soil sample and the characteristics of these relationships
are dependent on samples gradations. Resulting functions are presented in graphical
form. Combining the resulting functions with a novel equation proposed by Cinicioglu
and Abadkon (2014) that calculates p as a function of mean effective stress prior to
shearing (pi) and relative density (ID), it becomes possible to estimate q-f for problems
under consideration. Finally, proposed relationships are evaluated through physical
model tests and the results are discussed.
2. Peak dilatancy angle shear strain at failure relationship
Relationship between p and q-f was investigated by examining the triaxial test results
of 8 different sands [10, 11, 12]. Examples of obtained p-q-f relationships are shown
in Figure 1.
Figure 1. p-q-f relationships resulting from triaxial compression tests on different sands; (a) Erksak
sand (data obtained from Sasitharan (1989)), (b) Bafra sand (data obtained from Erzin (2004)), (c)
Ceyhan sand (data obtained from Erzin (2004)), and (d) Sinop sand (data obtained from Erzin (2004)).
From the results in Figure 1, it is deduced that p-q-f relationship can be quantified
using a semi-logarithmic relationship:
E. Sancak and O. Cinicioglu / Selection of Design Friction Angle 625
p
ln q
f (1)
The empirical fitting parameters in Eq. 1, and , vary from soil to soil. Thus
their dependency on the gradation characteristics should be investigated.
The goal in this study is to estimate q-f based on p which embodies the combined
influences of pi and ID. Therefore Eq. 1 is rearranged as follows:
p

q
f e
(2)
3. Influence of grain size on the fitting parameters of p-q-f relationship
Dilatant behavior is the volume change that results from the movement of grains
around and over each other during shearing. Since the magnitude of volume change
during dilation is directly dependent on the average grain size, it can be anticipated that
q-f (the shear strain at which p is mobilized) is dependent on mean particle size (D50).
Accordingly, variations of and line-fitting parameters in Eq. (2) with respect to D50
are plotted in Figure 2.
Figure 2. Variation of and with respect to D50
It is decided to express -D50 and -D50 relationships using a linear form since
linear function captures the relationship with the highest accuracy. Resulting linear
relationships are given below:
(3)
(4)
Combining Eqs. (2), (3), and (4), following empirical equation is obtained:
(5)
Using Eq. (5), q-f can be estimated from the values of p and D50. D50 can be obtained
from the results of sieve analysis. On the other hand, for the appropriate value of p, a
novel equation proposed by Cinicioglu and Abadkon (2014) can be used (Eq. (6)).
p
tan p i m I D (6)
pa
Here, is the mean effective stress prior to shearing and is the atmospheric
pressure; corresponds to the decrease in dilatancy angle per unit increase in
and m corresponds to the increase in dilatancy angle per unit increase in ID. Both
and m are empirical, unit-independent, line-fitting parameters that can be obtained by
conducting a few triaxial or plane strain tests. Thus, using Eq. (6), it is possible to
estimate p based on the stress state and density of the soil considered.
Eq. (5) is used to develop a design chart shown in Figure 3. Figure 3 can be used
to estimate q-f as a function of p and D50.
Figure 3. Design chart that allows the estimation of q-f based on the D50 of sand and the p value
based on the stress state and density in the field
4. Evaluation of the proposed method by physical modelling
In order to evaluate the validity of the proposed empirical equation, results of a

physical model study by Altunbas et al. (2014) were used. The model was a small scale
retaining wall model that was tested to active failure under 1g conditions. Deformation
of the model backfill was tracked using particle image velocimetry (PIV) method.
Photograph of the test setup is given in Figure 4 and the details of the tests are given in
Soltanbeigi et al. (2014). Model tests were conducted on Akpinar Sand, which was not
considered for the development of the chart in Figure 3. Properties of this sand are
given in Table 1.
Figure 4. Photograph of the test setup
Table 1. Engineering properties of Akpinar sand (Cinicioglu et al., 2014)
D50 'c
Soil Type emax emin Cu m r ID
(mm) (degree)
Akpinar Sand 0.87 0.52 0.27 1.23 -0.066 0.64 0.39 33.8 0.8
p of Akpinar sand is calculated as 27 by using the pi and ID values of the model
test in Eq. (6). Following, using the D50 of Akpinar sand and the calculated p in Figure
3, q-f is calculated to be 0.76 % for the model test.
Analyzing the model test results using PIV at different stages of model wall
translation, evolution of the shear band can be identified. Figure 5 shows the evolution
of shear band at different stages of model wall translation. Once a shear band emerges,
friction angle drops from the peak to the critical state value (c). It is clear in Figure 5
that 0.8mm of wall translation correspond to the stage just prior to the emergence of the
shear band. Thus, it is expected that the q-f that is obtained using the chart in Figure 3
should be close to the maximum shear strain that is calculated from the PIV analysis at
0.8mm of wall translation.
Figure 6 shows the distribution of shear strains along the profiles selected in Figure
5. Maximum shear strain at 0.8mm of wall translation is obtained to be 0.65% which is
very close to the calculated q-f, 0.76%. Therefore, it is possible to argue that the results
of the design chart given in Figure 3 are satisfactory.
Figure 5. Evolution of shear band at different stages of model wall translation
Figure 6. Distribution of shear strain along the profiles selected in Figure 5

5. Method for the selection of the design friction angle
The suggested method is applied step by step as follows:
i. p is determined either by calculation using Bolton (1986) or Cinicioglu

and Abadkon (2014) equations or testing.
ii. D50 value is obtained from sieve analysis.
iii. Using D50 and p values in the proposed design chart (Figure 3), q-f is
estimated.
iv. In order to estimate the maximum q that would develop in the soil if the
proposed structure is allowed to deform to its permissible limits (q-max), a
numerical model of the proposed design is constructed. Permissible levels
of deformation are imposed on the numerical model of the proposed
structure via prescribed displacements and the resulting q-max is obtained.
v. If q-f>q-max, then p is used in design. However, if q-fq-max, then c
should be used.
The condition of q-maxq-f does not mean failure for the soil body, it just signals
the emergence of a shear band. Therefore, for these levels of shear strains, it is
appropriate to design the system by adopting c as the design friction angle.
6. Conclusion
Totally 146 triaxial compression tests which are conducted on Silivri, Erksak and 8
different Anatolian sands were examined within this study. The results revealed a
relationship between peak dilatancy angle (p) and shear strain that corresponds to the
mobilization of peak friction angle (q-f). It was further observed that p-q-f
relationships are dependent on median grain size (D50). Accordingly, an empirical
design chart was developed. This chart allows the estimation of q-f from the values of
p and D50. Combining the chart with a novel equation proposed by Cinicioglu and
Abadkon (2014) that allows the calculation of p, it is possible to estimate q-f from the
gradation characteristics, stress state and density information. Validity of the proposed
chart was validated through physical modelling. Finally, a step-by-step method is
proposed for the selection of a suitable friction based on project and soil characteristics.
Acknowledgements
The authors would like to thank The Scientific and Research Council of Turkey
(TUBITAK) for supporting this study with Project number 110M595. The financial
support of Bogazici University Scientific Research Projects (BAP7600) is also
acknowledged.
References
[1] E. Sancak, Strain based selection of friction angle for geotechnical design, Master of Engineering
Thesis, Bogazici University, (2014).
[2] D. W. Taylor, Fundamentals of Soil Mechanics, John Wiley, New York, 1948.
[3] A. W. Bishop, Discussion to Measurement of the shear strength of soils by Skempton and Bishop,
Geotechnique 2(1) (1950), 113116.
[4] P. W. Rowe, The relation between the shear strength of sands in triaxial compression, plane strain and
direct shear, Geotechnique 19(1) (1969), 75-86.
[5] M. D. Bolton, Strength and Dilatancy of Sands, Geotechnique 36(1) (1986), 65-78.
[6] Y. P. Vaid, S. Sasitharan, The strength and dilatancy of sand, Canadian Geotechnical Journal 29 (3)
(1992), 522-526.
[7] T. Schanz, P.A. Vermeer, Angle of friction and dilatancy of sand, Geotechnique 46(1) (1996), 145-151.
[8] W. Powrie, Soil Mechanics: concepts and applications, 2nd edition, Spon Press, London and New York,
2004.
[9] O. Cinicioglu and A. Abadkon, Dilatancy and friction angles based on in situ soil conditions, J.
Geotech. Geoenviron. Eng. (DOI) 10.1061/(ASCE) GT.1943-5606.0001272 , 06014019 (2014).
[10] Y. Erzin, Strength of different Anatolian sands in wedge shear, triaxial shear, and shear box tests, Ph.D.
Dissertation, The Middle East Technical University, (2004).
[11] S. Sasitharan, Stress path dependency of dilatancy and stress-strain response of sand, Master of
Engineering Thesis, the University of British Columbia, (1989).
[12] A. Abadkon, Strength and dilatancy of anisotropic cohesionless soils, Ph.D. Dissertation, Bogazici
University, (2012).
[13] Cinicioglu, O., Abadkon, A., Altunbas, A., Abzal, M., Variation of friction angle and dilatancy for
anisotropic cohesionless soils, Proceedings, 18th International Conference on Soil Mechanics &
Geotechnical Engineering (2013), Paris.
[14] D. J. White, W. A. Take, M. D. Bolton, Soil deformation measurements using particle image
velocimetry PIV and photogrammetry, Geotechnique 53 (2003), 619 631.
[15] B. Soltanbeigi, A. T. Gezgin, A. Altunbas and O. Cinicioglu, Backfill failure surfaces and failure
displacements during active failure states, 11th International Congress on Advances in Civil
Engineering (ACE), Istanbul, Turkey (2014).
[16] A. Altunbas, B. Soltanbeigi, A.T. Gezgin, O. Cinicioglu, Shear band localization behind retaining walls,
23rd European Young Geot. Eng. Conf., Barcelona, Spain, (2014), 201-204.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-631
Stress strain behavior of a desaturated

loessian lightly cemented soil under triaxial
compression test
Julio A. CAPDEVILAa and Vctor A. RINALDIb,1
a
Professor, Universidad Nacional de Crdoba, Argentina
b
Head Professor, Universidad Nacional de Crdoba, Argentina.
[email protected]
Abstract. The upper most modern loess deposits of the central area of Argentina
are characterized by an open structure made of fine sand and volcanic silt particles
weakly bonded and usually in unsaturated conditions. Heterogeneity in loess is
usually present and arises mainly from the non-homogeneous cementation of the
soil mass. As saturation increases, the structure of loess collapses even under
geostatic pressures. In this work, some results of triaxial compression test
performed in loess are presented. Specimens were tested in saturated condition.
Testing results allows analyzing the stress-strain behavior of loess in a wide range
of strain levels, taking into consideration the influence of structure and confining
pressure. Below the yielding stress (collapse), the soil behaves almost as linear
elastic and stiffness is governed by degree of cementation. At higher stresses, the
collapse potential of the soil skeleton is governed by a complex interplay between
applied external pressure and internal forces due to cementation, and probably
others attraction-repulsion forces developed at particles of colloidal size.
Cementation causes an increase in stiffness, yielding stress and shear strength.
Keywords. Loess,stress,strain, soil modulus
1. Introduction
Loess deposits cover a wide area of the central region of Argentina. The thickness of
the deposits ranges from 20 to 60 meters. The word loess is a German term that
refers to a windblown deposit of silty soil characterized by an open structure. The
origin of Argentinean loess and its main physical properties has been extensively
described by Rinaldi et al. [1]. Loess of Argentina is composed mainly of platy shape
silt particles (40% and 60 %) and sand (5 % to 20 %) from volcanic origin. The clay
fraction (20 % to 35 %) is usually illite and montmorillonite. The combined effect of
particle shape, origin and particle gradation, render poorly accommodated and open
fabric. Usually loess presents some degree of cementation given by clay bridges,
soluble salts, silica amorphous, calcium carbonate, gypsum and iron oxide. The high
spatial variability of soil properties is originated from a non-homogeneus distribution
of cementation during the different postdepositional processes. Then, cementation in
1
632 J.A. Capdevila and V.A. Rinaldi / Stress Strain Behavior
loess may be found distributed at particle contacts or localized forming nodules and
particles aggregates. Additionally, when large amounts of cement (usually carbonates
or iron oxide) precipitated at some soil levels, may create a very stiff structure that
behaves as a true sedimentary rock-like material which is locally known as tosca.
Cemented soils can exhibit properties that are very distinctive from those of the
original uncemented soil or the freshly remolded one. There is an increasing body of
evidence that shows that cementation can have an important effect on the mechanical
parameters of the soil. Small strain stiffness increases with the amount of cementing
agent respect to the uncemented soil [2][3][4]. Airey and Fahey [5] showed also that
cemented soils display very small threshold strain and experience small volumetric
strains until decementation begins. Thereafter, at large strains they are prone to exhibit
dilative behavior making the effect of initial density less significant on volume changes
of the soil. While cemented sands can also liquefy, they exhibit a lower rate of pore
pressure generation and an increased resistance to cyclic liquefaction [5][6][7]. Shear
strength of soil increases with the amount of cement [8]. The increment of shear
strength becomes more significant at low confining pressures [9][10] and is due mainly
to an increase in cohesion while the increment in the value of the friction angle still
remain controversial [11][12][13]. The load-deformation behavior of cemented soils
ranges from brittle at low confinement to ductile at high confinement. The strong
dilative tendency, brittleness and post-peak strain softening behavior of cemented soils
at low confinement is associated to progressive failure and strain localization
[14][15][16]. At the microscale, loading under stress-controlled boundary conditions
allows the oriented propagation of contact decementation and the formation of shear
bands [17]. In agreement with trends for dilatancy, the severity of strain localization
increases with the degree of cementation and with decreasing confinement. While
uncemented contractive specimens do not localize under deviatoric drained loading, the
same specimen may exhibit strong localization after diagenetic cementation.
This work presents some fundamental experimental results which highlight the
effect of confining pressure and soil structure on the stress-strain behavior of
argentinean loess characterized by an open and slightly cemented fabric. Undisturbed
and remolded specimens of loess were tested in the triaxial cell in drained conditions.
Local displacement transducers were used to measure vertical and horizontal strains.
2. Soil Description and Testing Procedures
Block samples of loess were obtained at the campus of the National University of
Crdoba from a 10 meters deep open trench. The evaluation of sampling disturbance
by the method used here has been discussed extensively in the work of Rinaldi and
Capdevila [18]. Table 1 shows the most significant physical parameters of the soil
tested. Figure 1.a displays the grain size distribution curves of a soil specimen obtained
from a sample recovered at the site following two different test procedures. Curve (I)
was performed by fully remoulding the soil and then following the conventional
sieving test [19] and curve (II) was determined by placing the saturated and structured
sample on the coarsest sieve of the series and gently washed until the water extracted
from last sieve N 200 (0.075 mm) become clear indicating the absence of fine particles
detached from the aggregates retained in the upper sieves. No energy (eg. vibration or
shaking) was applied to the soil as sieving was performed. Figure 1.b shows some
pictures of the loess aggregates retained in different sieves. The differences observed
J.A. Capdevila and V.A. Rinaldi / Stress Strain Behavior 633
between both
b curves is considered heere as a good indicator to evaluate the strructuration
of the soiil. Furthermore, results inddicate that ceemented nodu ules remain sstable after
saturationn.
Table 1.Maiin physical param
meters of the soil ttested in this worrk.
Natural
Plasticity Passing Sieve
Dry Density water Liquid Limit Plastic Limit
Index #200 USCS
[kN/m3] content [%] [%]
[%] [%]
[%]
12.7 17.7 24.5 21.1 3.4 82.5 ML
Figure 1.
1 a) Sieve analyssis performed on structured and fu ully remolded loesss specimens. b) ccemented
aggreegates obtained frrom sieving analyysis of a loess sam
mple without desttructuration of thhe soil.
The triaxial
t comprression tests w were performed in the auto omatic cell EL LE Digital
Tritest. Vertical
V strain
ns were meassured externaally by mean n of a Linear ar Variable
Differential Transformeer (LVDT), an and locally, ussing a Local Deformation
D TTransducer
(LDT). Eaach LDT senssor consisted oof a thin, flex xible, strip of phosphor broonze with a
strain gagge glued at th he central secction of the specimen undeer test, similaar to those
originally proposed by Goto et al. [220]. Two LDT Ts were placeed around thee sample to
measure vertical
v strain
ns and a thirdd sensor was placed
p to meaasure horizonntal strains.
LDTs hinnges were glu ued to the sam mple membraane. Testing specimens
s of 50 mm in
diameter and
a 110 mm in height weree trimmed from m the block saamples. Tablee 2 displays
the initial testing condiitions of the uundisturbed sp pecimens prep pared here. U
Undisturbed
specimenss (U) were satturated in the ttriaxial cell unnder the initiaal confining prressure and
with very low gradientss (less than 1)). Remoulded (D) specimen ns were prepaared from a
fully destrructured soil by
b compactioon in a mould in 5 successiive layers to tthe desired
void ratioo. Table 2 alsso shows the initial testing g conditions of the compaacted loess
specimenss. The confiniing pressure uused here werre 10 kPa, 20 kPa, 40kPa aand 80 kPa
and it wass applied by mean
m of air too avoid the daamage of elecctronic compoonents. The
Skemptonn coefficient B was determiined from 0.9 95 and 0.98. The T deviatoricc load was
applied ata a velocity of 0.02 mm m/min which was calculateed to obtain a drained
condition..
Table 2: Initial testing conditions of undisturbed specimens of loess tested in triaxial compression.
U: Undisturbed. D: Remoulded and compacted
Dry Density Test Moisture Test Matrix Confining Pressure

Name
[kN/m3] Content [%] Suction [kPa] [kPa]
12,2 44,2 10
12,3 43,6 20
U 0
12,6 41,6 40
12,5 42,3 80
12,2 44,5 10
12,3 43,8 20
D 0
12,6 41,9 40
12,5 42,5 80
3. Test Results
3.1. Stress-Strain
Figure 2 displays the triaxial test results obtained for the saturated undisturbed
specimens tested at confining pressures of 10 kPa, 20 kPa, 40 kPa and 80 kPa. Curves
show strain hardening and ductile behaviour at most confining pressures. The higher
the confining pressure, the higher is the increment in the rate of hardening. Here
specimens bulge and no stress localization could be observed up to the 6% strain level.
In general, higher deviatoric stress curves are obtained with increasing confining
pressure. The effect of cementation may be dominant as is observed in Figure 2 where
the specimen tested at 10 kPa of confining pressure develops higher values of
deviatoric load than that obtained for the specimen tested at 20 kPa. Spatial variability
of cement distribution may be responsible for the anomalous behaviour described. It is
believed here that the above mentioned specimens tested at 10 kPa are comparatively
more cemented than the others. As strain level increases the effect becomes less
noticeably.
The influence of soil structure is more pronounced at low strain levels on the
secant modulus displayed on Figure 3 which were obtained from the stress-strain
curves of Figure 2. Notice that soil stiffness is clearly not related to confining pressure.
Opposite as could be expected, some specimens tested at the lower confining pressures
display higher modulus. Additionally, modulus degradation does not vary smoothly as
a function of strains but jumping. The irregular degradation is observed up to a vertical
strain of 0,1% approximately is attributed here to the effect of cementation as elastic
energy accumulates at cemented contacts and suddenly is released as stresses overcome
shear strength of the binder. At small strains, stiffness is controlled initially by the
cemented bonds. At higher strain levels than that corresponding to the elastic threshold,
the secant modulus increase with confining pressure and cementation has not
significant influence. Thus, soil structure becomes progressively decemented and
finally, at very large strains, the influence of confining pressure becomes dominant.
90
3 = 80 Kpa
Deviatoric Stress, 1-3 [kPa]

80 3 = 40 Kpa
70 3 = 20 Kpa
3 = 10 KPa
60
50
40
30
20
10
0
0,00 0,01 0,02 0,03 0,04 0,05 0,06
Vertical Strain, v
Figure 2. Stress-strain curves of undisturbed specimens tested at different confining pressures and saturated.
w = 42.9%
100000
3 = 80 Kpa
3 = 40 Kpa
Secant Modulus, E [kPa]
3 = 20 Kpa
10000 3 = 10 KPa
1000
100
0,001 0,01 0,1 1 10
Vertical Strain, v [x10-2]
Figure 3. Modulus degradation curves of undisturbed specimens tested at different confining pressures and
saturated. w = 42.9%
Figure 4 compares the stress-strain curves obtained for undisturbed and remoulded
specimens at two different confining pressures and saturated. No remarkable
difference is observed among curves obtained for remoulded and undisturbed
specimens tested at similar water content, confining pressures and density. Curves for
undisturbed specimens seem to be stiffer at medium strains and more brittle with a
clear locus of yielding pressure. Remolded specimens develop a more ductile behavior
and the yielding locus is not clearly identified. Figure 5 compares the secant Young
modulus for the same specimens of Figure 4. Here, the modulus degradation curves of
remolded specimens decay smoothly as would be expected for most unstructured soils.
In general, secant moduli of undisturbed specimens are higher than that of the
corresponding remoulded specimens at strain levels higher than 0.1%. The authors
believe here, that at small strain levels, strains distribute not uniformly throughout the
specimen and may be larger or lower in the central portion of the sample registered by
the LDT depending on the distribution of cement in the soil mass. Other explanation
may be the effect of confining pressure that cause a densification of the remolded
specimens, and has no the same influence on the structured specimens. Additional tests
are required here to find a more definitive explanation.
25 100

a) 90 b)
20 80
70
15 60
50
10 40
30
5 Undisturbed 20 Undisturbed
Remoulded 10 Remoulded
0 0
0 0,01 0,02 0,03 0,04 0,05 0,06 0 0,01 0,02 0,03 0,04 0,05 0,06
Vertical Strain, v Vertical Strain, v
Figure 4.Stress-strain curves of undisturbed and remoulded specimens tested with different confining
pressures and saturated. a) 3 = 10 KPa and saturated. b) 3 = 80 KPa and saturated.
100000 100000
a) Undisturbed b) Undisturbed
Remoulded Remoulded
10000
10000
1000
100 1000
0,001 0,01 0,1 1 10 0,001 0,01 0,1 1 10
Vertical Strain, v [x10-2] Vertical Strain, v [x10-2]
Figure 5. Modulus degradation curves of undisturbed and remoulded specimens tested with different
confining pressures and saturated. a) 3 = 10 KPa and saturated. b) 3 = 80 KPa and saturated.
3.2. Shear Strength
Figure 6 compares the failure envelops corresponding to the CD triaxial test for the
undisturbed and the remolded specimens tested in the saturate condition and at similar
dry unit weight. The Mohr circles plotted here correspond to a deviatoric stress
determined at the 6 % strain level and to the yielding point, determined in the point of
maximum curvature. Figure 6 shows that shear strength values of the undisturbed
specimen are slightly higher than that of the remolded specimens. The difference
between both envelopes reduces as the confining pressure increases. Thus, it is
believed here that in saturated conditions, soil structure is broken as confining pressure
increases and the behavior of the structured soil previous shearing tends to be that of
the remolded. The envelop corresponding to the undisturbed specimen shows a small
but true cohesion intercept at zero confining pressure. The friction angle corresponding
to the remolded soil is slightly higher than that of the undisturbed soil. The same Figure
6 compares the failure envelops corresponding to the shear strength determined at 6 %
strain level and at that corresponding to the point of yielding for the saturated samples.
The results show that the difference between both envelops increases with confining
pressure. The yielding envelop corresponding to the yielding stress seems to be less
influenced by confining pressure. Both envelopes develop cohesion intercept. Thus, it
seems that yielding stress, in saturated condition shear strength, is more affected by
cementation and less influenced by confining pressure. After yielding, confining
pressure increases shear strength significantly.
70
Remolded
60
0 = 24
Undisturbed
Shear Stress [KPa]
50
0 = 22 Yielding
40 envelope
30
20
10
0
0 20 40 60 80 100 120 140 160 180 200
Normal Stress [KPa]
Figure 6.Mohr-Coulomb failure and yielding envelopes corresponding to the undisturbed and remolded
specimens tested in drained (CD) and saturated conditions. Shear strength corresponds to the 6 % strain level.
4. Conclusions
A battery of tests was performed in this work to evaluate the influence of soils structure
on the stress-strain behaviour of loess soil. The following conclusions can be drawn
from this study:
+ Cementation of loess has significant effect on the secant modulus at small
strain levels (less than 1%). The influence of cementation tends to vanish at
higher strain levels. The degradation curve of secant modulus is distinctly
jumpy as compared with uncemented soils.
+ In general, secant modulus and shear strength of loess increase with increasing
confining pressure. However, non-homogeneous spatial distribution
aggregates could modify this tendency since the structure of the trimmed
specimens could be markedly different due to spatial variability of
cementation.
+ In saturated loess, the failure envelope of undisturbed specimen presents a
small value of true cohesion, that shows the effect of cementation on shear
strength but it is reduced significantly with confining pressure.
Acknowledgement
Authors acknowledge National Research Council of Argentina (CONICET) for the

financial support of this research.
References
[1] V. A. Rinaldi, R. J. Rocca and M. E. Zeballos, Geotechnical characterization and behaviour of

argentinean collapsible loess, Characterization and engineering properties of natural soils, Tan, Phoon,
Hight&Leroueil (eds), Taylor& Francis Group, London, 2007.
[2] S. Leroueil and D. W. Hight, Behaviour and properties of natural soils and soft rocks. Characterization
and Engineering Properties of Natural Soils, Tan et al. (eds), Swets&Zeitlinger, Lisse, 2003.
[3] C. S. Chang, A. Misra and S. S. Sundaram, Micromechanical modelling of cemented sands under low
amplitude oscilations, Gotechnique, Vol. 40(2) (1990), 251-263.
[4] Y. B. Acar and E. A. El-Tahir, Low strain dynamic properties of artificially cemented sand, Journal of
Geotechnical Engineering, ASCE, Vol. 112(11) (1986), 1001-1015.
[5] D. W. Airey and M. Fahey, Cyclic response of calcareous soil from the North-West shelf of Australia.
Geotechnique, 41 (1) (1991), 101 121.
[6] P. K. Robertson, D. J. Woeler and W. D. I. Finn, Seismic cone penetration test for evaluating liquefaction
potential under cyclic loading, Canadian Geotechnical Journal, Vol. 29 (1992), 686-695.
[7] G. W. Clough, J. Iwabuchi, N. S. Rad and T. Kuppusamy, Cemented sands under static loading, Journal
of Geotechnical Engineering, Vol. 115(8) (1989), 1102-1117.
[8] K. D. Walsh, Shear strength components attributable to cementation versus soil suction in arid
environments, Unsaturated Soil Engineering Practice, Geotechnical Special Publication N60, ASCE,
Houston and Fredlund (Eds), (1997), 169 197.
[9] S. Leroueil and P. R. Vaughan, The general and congruent effects of structure in natural soils and weak
rocks, Geotechnique, Vol. 40(3) (1990), 467-488.
[10] S. C. R. Lo, P. V. Lade and S. P. R. Wardani, An experimental study of the mechanics of two weakly
cemented soils, Geotechnical Testing Journal, Vol. 26(3) (2003), 1-13.
[11] G. W. Clough, N. Sitar, R. C. Bachus and N. S. Rad, Cemented sands under static loading, Journal of
the Geotechnical Engineering Division, Vol. 107(GT6) (1981), 799 - 817.
[12] F. Schnaid, P. D. M. Prietto and N. C. Consoli, Characterization of cemented sand in triaxial
compression, Journal of Geotechnical and Geoenviromental Engineering, ASCE, Vol. 127(10) (2001),
857-868.
[13] F. Saihi, S. Leroueil, P. La Rochelle and I. French, Behaviour of the stiff and sensitive saint-jean-
vianney clay in intact, destructured and remoulded conditions. Canadian Geotechnical Journal, Vol. 39
(2002), 1075-1087.
[14] R. J. Hill and J. W. Hutchinson, Bifurcation phenomena in the plane tension test, Journal of Mechanics,
Physics and Solids, Vol. 23 (1975), 239-264.
[15] J. Rice and J. W. Rudnicki, A note on some features on the theory of localization of deformations.
International Journal of Solids Structures, Vol. 16 (1980), 597-605.
[16] T. Schanz, A constitutive model for cemented sands, Localization and bifurcation theory for soils and
rocks, Balkema, Rotterdam, 1998, 165-172.
[17] G. C. Cho and J. C. Santamarina, The Omnipresence of Localization in Geomaterials. 3rd Deformation
Characteristics of Geomaterials, Di Benedetto et al. Eds., Lyon, 2003, 465-473.
[18] V. A. Rinaldi and J. A. Capdevila, Effect of sampling disturbance on the stress-strain behavior of
structured collapsible soil, Proc. 17th International Conference on Soil Mechanics and Foundation
Engineering, Alexandria, Egypt, 2009, 837-840.
[19] S. Goto, F. Tatsuoka, S. Shibuya, Y. S. Kim and T. Sato, A simple gauge for local small strain
measurements in the laboratory, Soils and Foundations, Vol. 31(1) (1991), 169 180.
[20] ASTM, Anual Book of ASTM Standards, 1991.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-639
Analysis of the hydro-mechanical

behaviour of volcanic ash slopes submitted
to rainfall
Alessio FERRARI, Lyesse LALOUI, Chao LI, John EICHENBERGER
Swiss Federal Institute of Technology, Lausanne, EPFL, Switzerland
Abstract. Rainfall-induced landslides in volcanic ashes represent a major natural

hazard in many densely populated regions around the world. Volcanic ashes in
steep slopes are most of the time in a partially saturated state. Capillary effects
greatly influence the soil behaviour. During rain infiltration, volcanic ashes display
strong volume changes and shear deformations associated with the development of
the failure mechanism. Environmental fluctuations and the consequent wetting and
drying cycles in the slope are rarely considered. This paper presents a hydro-
mechanical analysis of the behaviour of a volcanic ash slope subjected to daily
wetting and drying cycles. A transient finite element simulation is performed with
a 2D slope model that is representative for a landslide-prone volcanic area in Costa
Rica. The ash behaviour is reproduced by means of the bounding surface and
multi-mechanism elasto-plastic model ACMEG-s for unsaturated soils; the model
was previously calibrated on the basis of a laboratory testing programme designed
to characterise the hydromechanical behaviour of the volcanic ash. The results of
the finite element simulation demonstrate the important role of wetting and drying
cycles, slope geometry and bedrock on the evolution of matric suctions, soil
volume changes and shear deformations.
Keywords. Hydromechanical modelling, Unsaturated soil, Volcanic ash, Wetting

pore-collapse.
1. Introduction
Rainfall-induced landslides in steep slopes of volcanic origin are common around the
world and have caused massive destruction and the loss of lives (Italy: [1, 2]; Hong
Kong: [3]; Central America: [4-6]). In Central America, the populated areas surrounded
by steep volcano slopes are highly vulnerable to rainfall- and earthquake-induced
landslides that are triggered in loose volcanic soils. In 2001, the Las Colinas landslide
in El Salvador claimed > 500 casualties [5], and in 1998, thousands of landslides were
caused by Hurricane Mitch throughout all of Central America with > 9000 casualties
[7]. Due to the loose structure from the depositional process, narrow grain-size
distribution, degree of weathering and collapsibility, failed soil masses from volcanic
deposits are often reported to turn into rapid flow slides or debris flows with
catastrophic consequences in the runout zones [8]. At the origin, the involved materials
are mostly in a state of partial saturation.
At the end of the wet season of 2005, two rainfall-induced landslides in volcanic
ashes caused material damage and production losses in a quarry situated along the
hillslopes of Irazu volcano in Costa Rica. In this paper, a physically-based, hydro-
mechanically coupled continuum modelling approach is applied performing a 2D finite
640 A. Ferrari et al. / Analysis of the Hydro-Mechanical Behaviour of Volcanic Ash Slopes
element slope model, aimed at reproducing the behavior of the landslide-prone site on
the foothills of Irazu volcano in Costa Rica. Realistic rainfall conditions are considered,
which reproduce in particular the intense, short-duration rainfall events during a typical
wet season. Particular attention is paid to the analysis of the spatial-temporal evolution
of matric suctions, degree of saturation and unsaturated flow, as well as reversible and
irreversible volumetric and deviatoric deformations induced by rainfall infiltration.
Emphasis is placed on a realistic representation of the hydro-mechanical behaviour of
volcanic ashes by means of an elasto-plastic constitutive model for partially saturated
soils. An extensive laboratory testing programme was carried out [9] to characterise the
hydro-mechanical behaviour of the involved volcanic ash and to calibrate the various
model components, which are used in the modelling approach and the numerical
analysis presented in this paper.
2. The hydro-mechanically coupled modeling framework
The ACMEG-s elasto-plastic constitutive model [10, 11] is used to reproduce the
hydro-mechanical behavior of the volcanic ash. Along with the model for the soil water
retention behaviour [12, 13], the constitutive models define the stress-strain and
suction-degree of saturation relationship, respectively. Both these mechanical and
hydraulic constitutive components are embedded in a generalized effective stress
framework that is suitable to describe hydro-mechanically coupled processes in
unsaturated and saturated conditions [14, 15]:
= pa I + Sr ( pa pw ) I = Sr pw I (1)
Where is the effective stress tensor, is the total stress tensor, pa (equal to zero)
and pw are the pore air and pore water pressure, respectively, and Sr is the degree of
saturation.
In the ACMEG-s model the elastic volumetric and deviatoric strains are given by:
dp ' dq
d ve = n
and d de = n
p' p' (2)
K ref
p '
3Gref
p '
ref ref
Where K ref , Gref and ne are, respectively, the bulk elastic modulus at a reference
mean stress p 'ref , the reference shear elastic modulus at p 'ref , and the non-linearity
exponent.
Plasticity is characterised by two different yield criteria, the isotropic yield

function with an associated flow rule and the deviatoric yield function with a
non-associative flow rule, each of which corresponds to a plastic mechanism associated
to either the volumetric or distortional deformation mode (Equations 3a, 3b):
fiso = p pc riso = 0 (3a)

A. Ferrari et al. / Analysis of the Hydro-Mechanical Behaviour of Volcanic Ash Slopes 641
d p
f dev = q Mp 1 b ln rdev = 0 (3b)
pc
Where is the preconsolidation mean effective stress, is a soil parameter which
defines the shape of the deviatoric yield function,
is the ratio between the

preconsolidation pressure and the corresponding critical state pressure , and
are internal hardening variables and is the slope of the critical state line in plane
. The formulation proposed by Van Eekelen (1980) is adopted to account for the
third stress invariant Lode angle as shown below [16]:
nL
M = 3 3 L 1 + bL sin 3 (4)
Where is the Lode angle and aL, bL and nL are material parameters. The plastic
potential is given by [17]:
1
1 d p'
g dev = q Mp' 1 (5)
1
p'c
Where is a soil parameter. The ACMEG-s model uses an isotropic hardening

rule. The size of the elastic domain is controlled by the (apparent) preconsolidation
pressure [18]:
pc = pc0 exp ( vp ) (6)

Where is the plastic stiffness coefficient. The volumetric plastic strain is the
internal hardening variable for both plastic mechanisms.
The ACMEG-s model presents a bounding surface plasticity formulation [19, 20].
The dependency of apparent preconsolidation pressure on matric suction is given
by:
pc0 if s sE

pc = s
nLC
(7)
p 'c = p 'c 0 1 + s log s if s sE
E

Where is the preconsolidation pressure at zero suction and reference void ratio
eref, s is a material parameter and sE is the air entry suction. The apparent
preconsolidation pressure remains constant up to the air entry suction since capillary
effects only occur as air enters into the pore space and inter-particle menisci start to
form.
Experimental data on the volcanic ashes suggest marginal hysteretic effects and
dependency on void ratio [9]. In this sense, a Van Genuchten type soil water retention
curve [21] is used in this analysis:
1 Sres
Sr = Sres +
1 + ( s)n
m (8)

Ferrari et al. (2013) showed a clear dependency for a volcanic ash of the
permeability on degree of saturation and void ratio. To account for that, the intrinsic

isotropic permeability is expressed as a function of initial void ratio and
volumetric strain :
kwint = kwint,0 e0 v (1+ e0 )

ck
(9)

where is the intrinsic isotropic permeability for a reference void ratio
and the exponent is a fitting parameter.
All the constitutive models parameters for the volcanic ash involved in this study
were obtained through extensive laboratory tests [9, 22]. The ACMEG-s model and the
soil water retention model are implemented into the nonlinear finite element code
LAGAMINE which solves mass balance equation and the momentum equation [23, 24].
3. The finite element model
Hydro-mechanically coupled slope simulation is run for a two-dimensional slope

section in plain strain conditions, representing the steepest slope section above a
Pozzolana quarry on the South-West slope of Irazu Volcano at 2500 m above sea level
(Figure 1). The finite element model is built on the basis of topographical data of the
landslide-prone area situated above the quarry. The volcanic ash mantling the
consistent and relatively impermeable Pozzolana bedrock is assumed to present
generally a constant thickness of 3 m. The volcanic ash is characterised as fairly
homogeneous, cohesionless silty sand with a shearing resistance angle = 35.5. The
natural water content is in the range of 0.22-0.24 and the average dry unit weight d =
9.9 kN/m3, which yields an average void ratio of 1.58 [9].
The soil cover is separated into two material layers with preconsolidation pressure
increasing from the top to the bottom layer aiming at reproducing normally
consolidated conditions prior to rain infiltration. The Pozzolana bedrock is modelled as
an elastic, impermeable body. Uniform matric suctions equal to 20 kPa are defined in
the slope as initial conditions. After an initial gravity initialisation phase, a hydro-
mechanical simulation is run over a period of 64 days of the wet season of 2005.
Hydro-mechanically coupled simulations were run in order to analyse the slope
response to realistic rainfall scenarios, with the aim to gather information on the
evolution of significant geomechanical variables such as matric suction, volumetric and
deviatoric deformations. Real rainfall conditions are simulated here on the basis of
daily rainfall data from Juan Santamaria Int. airport close to the capital San Jos
(NOAA 2011) [25].
Figure 1. Two-dimensional slope section, representing the steepest slope section above a Pozzolana quarry
on the South-West slope of Irazu Volcano at 2500 m above sea level.
4. Selected results and process understanding
As an example of the outcomes of the modelling results, the evolution in time of

matric suction is plotted in Figures 4a and for the vertical section A, indicated in Figure
2. Matric suctions within the first meter of soil depth fluctuate in accordance with daily
rainfall input. At the toe of the slope, the suction response is more progressive as a
consequence of lateral subsurface flow moving water from the top to the bottom of the
slope. Between 2 and 3 m soil depth, matric suctions remain low even during longer
dry periods (e.g. day 25-35). In the 0.5 m above the bedrock, suctions are
monotonically decreasing, which is an indicator for worsening slope stability
conditions throughout the wet season. This indicates also the presence of continuous,
lateral groundwater flow in the quasi-saturated zone above the impermeable substratum.
As a consequence of variable rainfall boundary conditions and the complex unsaturated
hillslope flow regime, the volcanic ash undergoes stress paths involving cyclic changes
in effective stresses. The capillary stress changes lead together with total stress changes
from rain infiltration and stress redistributions within the slope to cyclic variations in
mean effective stresses and deformations. Volumetric plastic strains occur primarily in
locations with strong losses in matric suctions due to the normally consolidated state of
the volcanic ashes. The temporal evolution of volumetric plastic strains is displayed for
section B and three different depths (1, 2 and 3 m) in Figure 3. Noticeable increases in
contractive volumetric strains are observed after 15, resp. 39 days, both of which
correspond to series of 3 to 4 days of consecutive rainfall.
The deviatoric plastic strain contours indicate for the steep part of the slope section
shear localisation along the interface between bedrock and soil cover. Based on the
distinctive magnitude of plastic deviatoric strains in the lower part of the slope towards
the surface, it can be stated that the most probable failure mechanism would surface at
the toe of the slope. The location of the bedrock has clearly a strong influence on the
probable failure mechanism. In analogy to the volumetric behaviour, plastic deviatoric
strains increase significantly after particularly rainy periods. Deviatoric stresses
undergo cyclic variations correlated to wetting and drying periods as shown in Figure 4.
Figure 2. Evolution of matric suction at different depths for the vertical section (slope angle =51).
Figure 3. evolution of volumetric strains in section B

Figure 4. variations in mean effective stresses p = p-pt=0 and capillary stresses (Sr s) = (Sr s) -
(Sr s) t=0 at 1, 2 and 3 m of soil depth in section B.
5. Conclusions
A continuum modelling framework was presented for the analysis of instability

processes at the onset of failure in loose volcanic ash slopes. Unlike slope stability
methods representing failure with a single plastic limit condition, the bounding surface
and multi-mechanism plasticity model with a hydro-mechanically coupled formulation
is suitable for representing instability processes in unsaturated volcanic ash slopes
subjected to rain infiltration.
The analysis of the results from the transient 2D slope simulation revealed the
complex flow regime in the volcanic ash cover with unsaturated conditions. Wetting
pore-collapse occurs in locations with strong losses in matric suctions and steep slope
sections with strongly anisotropic stress states. Besides capillary stress changes, total
stress changes from rain infiltration and stress redistributions were observed to
influence the stress paths and the overall stability of the slope.
6. Acknowledgements
The following projects are acknowledge: SafeLand, Living with landslide risk in
Europe: Assessment, effects of global change, and risk management strategies, under
Grant Agreement No. 226479, Wandland, Effects of wetting and drying cycles on
landslide activity, under Grant Agreement No. PERG06-GA-2009-256426 in the 7th
Framework Programme of the European Commission, and TRAMM 2 Swiss
Competence Center Environment and Sustainability. The authors gratefully
acknowledge Mr. James Fern for the rainfall data collection.
References
[1] Crosta G.B., Dal Negro P. Observations and modeling of soil slip-debris flow initiation processes in
pyroclastic deposits: the Sarno 1998 event. Natural Hazards and Earth System Sciences (2003); 3: 53-
69.
[2] Cascini L., Cuomo S., Guida D. Typical source areas of May 1998 flow-like mass movements in the
Campania region, Southern Italy. Engineering Geology (2008); 96: 107-25.
[3] Fuchu D., Lee C.F., Sijing W. Analysis of rainstorm-induced slide-debris flows on a natural terrain of
Lantau Island, Hong Kong. Engineering Geology (1999); 51: 279-90.
[4] Capra L., Lugo-Hubp J., Borselli L. Mass movements in tropical volcanic terrains: the case of Tezitlan
(Mexico). Engineering Geology (2003); 69: 359-79.
[5] Crosta G.B., Imposimato S., Roddeman D., Chiesa S., Moia F. Small fast-moving flow-like landslides in
volcanic deposits: The 2001 Las Colinas Landslide (El Salvador). Engineering Geology (2005); 79:
185-214.
[6] Harp E.L., Reid M.E., McKenna J.P., Michael J.A. Mapping of hazard from rainfall-triggered landslides
in developing countries: Examples from Honduras and Micronesia. Engineering Geology (2009); 104:
295-311.
[7] U.S. Agency for International Development Fact Sheet #17, 1998.
[8] Picarelli L., Olivares L., Avolio B. Zoning for flowslide and debris flow in pyroclastic soils of Campania
Region based on infinite slope analysis. Engineering Geology (2008); 102: 132-41.
[9] Ferrari A., Eichenberger J. and Laloui L. Hydro-mechanical behaviour of a volcanic ash. Geotechnique
(2013); 63(11): 1433-1446.
[10] Nuth M. Constitutive modelling of unsaturated soils with hydro-geomechanical couplings. Ecole
Polytechnique Fdrale de Lausanne, Switzerland. Ph.D. thesis (2009).
[11] Nuth M., Laloui L. New insight into the unified hydro-mechanical constitutive modelling of
unsaturated soils. Unsat Asia, Nanjing (2007); 109-125.
[12] Nuth M., Laloui L. Advances in modelling hysteretic water retention curve in deformable soils.
Computers and Geotechnics (2008); 35(6): 835-844.
[13] Salager S., Nuth M, Ferrari A, Laloui L. Investigation into water retention behaviour of deformable
soils. Canadian Geotechnical Journal (2013); 50(2): 200-208.
[14] Nuth M., Laloui L. Effective stress concept in unsaturated soils: Clarification and validation of a
unified framework. International Journal for Numerical and Analytical Methods in Geomechanics
(2008); 32: 771-801.
[15] Laloui L., Nuth M. On the use of the generalised effective stress in the constitutive modelling of
unsaturated soils. Computers and Geotechnics (2009); 36: 20-23.
[16] Di Donna, A., Laloui L. Numerical analysis of the geotechnical behaviour of energy piles. International
Journal for Numerical and Analytical Methods in Geomechanics (2014). DOI: 10.1002/nag.2341.
[17] Nova, R. & Wood, D.M. A constitutive model for sand in triaxial compression. International Journal
for Numerical and Analytical Methods in Geomechanics (1979); 3: 225-278.
[18] Hujeux J. Une loi de comportement pour le chargement cyclique des sols. Gnie Parasismique. Paris,
Les dition de lE.N.P.C. (1985) ; 287-353.
[19] Hujeux J. Calcul numrique de problmes de consolidation lastoplastique. Ecole Centrale Paris,
France. Ph.D. thesis (1979).
[20] Dafalias Y., Hermann L. A bounding surface soil plasticity model. International Symposium on Soils
under Cyclic and Transient Loading, Swansea (1980); 335-45.
[21] Van Genuchten M.Th. A closed-form equation for predicting the hydraulic conductivity of unsaturated
soil. Soil Science Society of America Journal (1980); 44(5): 892-898.
[22] Ferrari A., Eichenberger J., Fern J., Ebeling P.and Laloui L. Experimental and numerical analysis of an
unsaturated volcanic ash deposit for the establishment of an early warning system in a quarry in Costa
Rica. Geocongress 2012: State of the Art and Practice in Geotechnical Engineering, Oakland,
Geotechnical Special Publications No. 225. ASCE, 2012, 2512-2521.
[23] Charlier R. Approche unifie de quelques problmes non linaires de mcanique des milieux continus
par la mthode des lments finis. University of Lige, Belgium. Ph.D. thesis (1987).
[24] Collin F. Couplages thermo-hydro-mcaniques dans les sols et les roches tendres partiellement saturs.
University of Lige, Belgium. Ph.D. thesis (2003).
[25] Eichenberger J., Ferrari A., Laloui L. Early warning thresholds for partially saturated slopes in volcanic
ashes. Computers and Geotechnics (2012); 49, 79-89.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-647
Granular temperature measurements of

uniform granular flows
Devis GOLLINa,1, Elisabeth T. BOWMANa and Paul SHEPLEY a
a
Department of Civil and Structural Engineering, University of Sheffield, Sheffield,
United Kingdom
Abstract. The investigation described here consists of a study aimed at validating

a reliable and straightforward method to estimate granular temperature.
Experiments on dry granular flows of sand and spherical beads were performed on
a simple inclined chute geometry. Flow properties at the sidewall were obtained
from images recorded at high frame rate using particle image velocimetry. Direct
measurements of granular temperature were estimated by means of statistical
analysis. Low values of granular temperature were measured indicating that a
collisional state was not fully achieved. Hence, we deemed the flows to fall into
the dense flow regime. In spite of this, the dynamics of the flow exhibited a highly
inhomogeneous behaviour evidenced by our results, which showed the variable
nature of velocity fields and granular temperature observed.
Keywords. Granular flow, granular temperature, particle image velocimetry
1. Introduction
Granular materials are encountered in a diverse range of geophysical contexts such as

granular chute flows, landslides and debris avalanches. The majority of these processes
undergo rapid rates of deformation in which momentum transfer is mainly carried by
frictional and collisional stresses, although shear stresses in the interstitial fluid can
represent an additional element of momentum exchange. More specifically, collisions
between neighbouring particles will cause random fluctuations in the particle velocity,
reminiscent of the thermal fluctuation motion of molecules [1] in kinetic theory. The
random components that generate particle stresses through collisions can be related to
the concept of granular temperature. This entity represents the basic concept
underpinning the kinetic theory of granular gases. Granular temperature provides a
measure of the energy associated with the fluctuating nature of the granular motion, i.e.,
the agitation within the medium. Thus, granular temperature may be used to
characterize the flow regime and infer the ability of a granular material to flow.
Generally, granular systems are easier to treat numerically than they are
experimentally [2,3]. Kinetic theory and its extensions have been idealised by making
strong assumptions about the inelasticity, shape, and rotational motion of the grains [4].
Moreover, the introduction of kinetic theory and similar microstructural theories has
created new challenges to experimentalists. Many parameters introduced in the
simulations are hard to measure in laboratory tests and direct quantitative comparison
1
Corresponding Author. E-mail address: [email protected]
648 D. Gollin et al. / Granular Temperature Measurements of Uniform Granular Flows
with experiments is difficult to make. Yet verification of the prediction of models

should always be tested and validated to results taken from experimental investigations.
As a consequence, it is extremely important to provide experimental data against which
theories and simulation can be judged.
Direct estimations of granular temperature are rarely found in the literature and in
any case are difficult to interpret. However, the last few decades have seen the
development of new optical techniques along with high-speed recording systems,
which can now enable the acquisition of robust measurements of particle motion. A
variety of instrumentation and mechanical devices are utilized to obtain such
measurements. Reynolds et al. [5] analysed the bulk motion of granular mixtures in a
high shear granulator. Owing to the experimental configuration, measurements were
limited to surface velocity obtained using particle image velocimetry (PIV). The
analysis of the temporal velocity was set at a single localized position in the centre of
the measured surface velocity. The analysis was then extended to extract velocity
correlation and granular temperature at a length scale of a single particle. Inclined chute
geometries have also been used. Azanza et al. [6] measured granular temperature in dry,
highly energetic and diluted flows. A two-dimensional analysis was performed by
constraining a single layer of particles in a narrow channel. Armanini et al. [7,8]
studied saturated granular flows in a recirculating flume. In their investigations, flows
were observed at the near wall and, to some extent, at a certain distance from the
sidewall [8]. Granular temperature profiles were also extracted for different flow
regimes by using Vorono imaging. Hanes and Walton [9] carried out numerical
simulations in order to interpret flow properties interior to the flow down a bumpy
incline and compared these with experimental results taken at the sidewall.
Interestingly, they noted a discrepancy in the variation of granular temperature in
comparison to the sidewall. They deemed the boundary conditions responsible for
strong three-dimensional structure in the flow, strongly influencing the measured
granular temperature.
Physical modelling is usually constrained by the impracticality of obtaining
measurements of the internal character of the flow. Indeed, measurements of velocity
fields in both dilute and dense granular systems are often limited by the opaque nature
of such flows. For this reason, flowing materials are probed at the boundaries where the
conditions control the flow properties. Alternative non-intrusive techniques can then be
used to overcome these limitations. In the context of geophysical flows, Sanvitale and
Bowman [10] performed physical experiments on an unsteady non-uniform saturated
granular medium. Their technique (see [11]) allowed for the acquisition of flow
properties internally to the bulk granular mass. They used the same approach as [5] to
estimate granular temperature. The results were encouraging even for an agitated flow
composed of different particle sizes. However, further validations and supporting
evidence are required to examine the validity of this technique to correctly capture
granular temperature.
In this paper we attempt to measure granular temperature for flows of dry nearly
monodisperse sand and spherical ceramic beads. Our approach to estimate flow fields
is similar to previous works by Reynolds et al. [5] and Sanvitale and Bowman [10].
They have suggested that a combination high-speed high resolution image acquisition
with PIV provides a possible solution to determine granular temperature. We present
our results for two flows down a flat, frictional incline and discuss the significance of
the granular temperature profiles we have extracted.
D. Gollin et al. / Granular Temperature Measurements of Uniform Granular Flows 649
2. Physical experiments
2.1. The apparatus
A small inclined chute has been developed in order to reproduce dry granular flows.
The flume is a sloping rectangular 10 cm wide and 150 cm long channel that can tilt
from horizontal up to 45. The choice of such a narrow channel was made to constrain
the flows between well-defined boundaries, thus allowing measurements to be taken at
the sidewall. At one end, the material is held inside a hopper, which can contain
approximately 10 kg of material. A double-gate opening system is used. Before
releasing the material, the first gate is set at a certain height from the flume bottom,
which in turn reflects the thickness of the granular flowing layer. The thickness is well
controlled by the gate, although it is unlikely to be equal to the opening.
Interchangeable sheets, onto which different particle sizes can be glued to control basal
roughness, cover the bottom of the chute. The walls are made of transparent material;
therefore observation may be obtained throughout the entire length of the chute.
2.2. Data acquisition analysis
The optical equipment consists of a Phantom Miro 310 high-speed camera, which is
supplemented with two light sources. Images were captured at a frame rate of 10,000
frames per second. Camera resolution used was 448 x 448, which roughly corresponds
to an image size of 32 x 32 mm (Fig. 1)
A particle image velocimetry (PIV) algorithm [12] was employed directly to the
captured images and was found sufficient to determine the velocity fields. The software
allows a static mesh to be kept in the same location while the granular mass is flowing.
This PIV is designed for non-transparent material and does not require any tracer
particles. The surface structure of the flow is sufficient to determine the vector field via
the cross-correlation analysis between interrogation regions in the first and second
images. Different square interrogation sizes were used, specifically 16, 24, 32, 40 and
48 pixels [10,12]. The corresponding spatial regions are approximately 1.15, 1.7, 2.5,
2.8 and 3.4 mm. The size of the interrogation area has an important effect on the range
of velocity fields that can be determined [5]. Notably, patch sizes smaller than the
particle diameter does not produce sufficient textural information. Conversely, larger
interrogation area implies that the measure will refer to an ensemble of particles.
2.3. Experimental procedure
Granular materials used in this study are sand and spherical ceramic beads. Two
granular flows are presented here. For clarity we will refer to SaT1 for the flow of sand
and BeT2 for the flow of beads. Beads were selected because of their spherical shape
which are typically assumed in the majority of the granular flow theories. For the flow
of beads the flume bottom roughness was made by gluing the same spherical material.
In case of sand, after some trials the roughness was substituted with a finer sand
fraction in order to allow analyses at less steep inclinations. SaT1 consists of a batch of
sand that was sieved to produce a nearly monodisperse material with particle sizes
between 1.4 and 1.7 mm. In comparison, ceramic beads are nominally 1.4 1.6 mm in
diameter. The static friction angles, determined from a tilt test, were 34.5 for sand and
24 for beads, and dry bulk densities of 1.52 g/cm3 (loose density) and 2.43 g/cm3,
respectively.
SaT1 was released from the hopper at an inclination of 38. The flume was tilted at
32 for BeT2. The camera was positioned 30 cm before the outlet where the flowing
material is more likely to reach an approximately steady regime. The analysis
performed here is valid if the flows are uniform in the direction parallel to the main
plain of sliding. The uniformity was checked by visual inspection. We selected
temporal regions where flow exhibited no detectable variations in the depth profile.
The nominal flow depth was 20 particle diameters for SaT1 and 18 for BeT2. Videos
were recorded at the initial frame rate of 10,000 fps over 0.5 s, producing 5,000 frames.
At a later stage, we reduced the temporal resolution to 5,000 fps and 2,000 fps by
removing images from the original recording. In doing so, we kept the same flow
dynamics but established different frame rates at which the flow could be probed.
Granular temperature requires the determination of kinematic properties, e.g.
fluctuation velocity. Reynolds et al. [5] showed that granular temperature might be
measured by statistical analysis. If the fluctuations are assumed to be normally
distributed, granular temperature is given by the variance of the fluctuation velocity.
Similarly, the standard deviation of velocity can be used as a good indicator of the
intensity of vibrational kinetic energy in flowing material [9]. If the mean velocity of
all particles is constant, the variance of the velocity may characterise the particle
granular temperature. Sanvitale and Bowman [10] adopted this approach; the same
analysis is also followed in this study.
(a) (b)
32 32
24 24
y (mm)
16 16
8 8
0 0
0 8 16 24 32 0 8 16 24 32
u x (mm) x (mm)
Figure 1. Example of images extracted from sand (SaT1 (a)) and beads (BeT2 (b)) flows.
3. Results and analysis
We used particle image velocimetry (PIV) to measure the velocity fields in our
flows. A careful judgement with regard to the interrogation size for the cross-
correlation analysis has to be made in order to estimate the correct flow properties.
Notably, a patch size of the order of the particles generates the most accurate values of
granular temperature [5]. Larger interrogation regions cause a reduction in the

measured fluctuation velocity that in turn reflects a reduction in measured granular
temperature. As we employed material uniform in diameter, two fixed interrogation
areas were considered. 32 x 32 and 24 x 24 pixels patches were chosen to analyse SaT1
and BeT2, respectively. Note that a 24 pixel patch corresponds to 1.7 mm, hence it
covers the entire surface of a singular bead. A slightly larger size was required for SaT1
to take into account the diverse shape of the particles. All the results reported here are
shown for these two interrogation sizes only. We evaluated the influence of patch size
and found what was previously stated in [10]. That is, very small interrogation areas
produce high levels of noise whereas larger areas progressively reduce the value of
fluctuation velocity and hence, granular temperature.
Velocity components for the two flows are shown in Fig. 2 (u and v coordinates
refer to the perpendicular and parallel flow directions, respectively). It can be seen that
a static layer toward the bottom of the flow characterizes SaT1 (Fig. 2(a)). In
comparison, BeT2 (Fig. 2(b)) presents an increasing linear velocity gradient from base
to the top; in this flow the static layer was absent. Flow properties depend on the shape
anisotropy of the material and the velocity profile is strongly controlled by the
angularity of the particles [13]. Thus, it is normal to expect that these granular flows
show different avalanching behaviour. Velocity fluctuations are shown as the standard
deviation of the velocity near the sidewall over the entire duration of the flow (0.5 s).
The fluctuations clearly increase toward the free surface, and decrease toward the base.
A less clear trend developed for BeT2 (Fig. 2(b)), which may reflect the more
collisional behaviour of this material. Nonetheless, the bumpier base may create a slip
zone induced by the particles interacting with the base.
In many studies a uniform stationary flow state is assumed for the simplicity it
provides in the calculations. However, granular flows always present some degree of
unsteadiness. Through our imaging measurements and physical interpretation we
assumed that the flow was steady over time, although this is a weak assumption. If
plotted in the same figure, velocities calculated between pair of images fluctuate with a
clear degree of correlation within a range that has the order of the velocity fluctuations
depicted in Fig. 2. Thus, velocities averaged over shorter periods may deviate
considerably from the long-term average, suggestive of a mean velocity (relative to
local fluctuations) that is essentially unsteady. Many theoretical considerations of
(a) (b)
30 30
20 20
y (mm)
10 10
fluctuation velocity fluctuation velocity

0 mean velocity 0 mean velocity
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
u, v (m s-1) u, v (m s-1)
Figure 2. SaT1 (a) and BeT2 (b). Mean and fluctuation velocity profiles.
granular flows derive from fluid mechanical approaches. Careful considerations should
be made as to whether the measured temperature derives from real (i.e. random and
uncorrelated) fluctuations as defined by the kinetic theory or just turbulences
generated by the mechanics of the flow. We did not find any suitable explanation; a
further investigation is required to ascertain the nature of the fluctuating motion.
This consideration aside, granular temperature was extracted as the variance of the
velocities for the entire duration of the flow (0.5 s); temperature was then taken as the
non-stationary component of the local mean velocity. The fluctuation velocity gradient
perpendicular to the flow direction (u) showed non-negligible effects. Thus, in the
calculation we have taken both u and v components into account.
Granular temperature profiles are shown in Fig. 3. The most striking result regards
the magnitude of granular temperature. Our values ranged from 0.0 up to
approximately 0.005, which are not in agreement with measurements for collisional
state motion found in the literature. Drake [14] reported granular temperature between
0.2 and 0.4 for very disperse and diluted dry flows. Lower values, varying around 0.1,
were found in much denser flows. From numerical simulation down a bumpy incline,
Hanes and Walton [9] estimated temperature at the sidewall up to 0.01. In saturated
granular flows for material sliding over a dense layer of particles Armanini et al. [7]
measured granular temperature of 0.05 at the surface while decaying to 0.0 at the
bottom. It seems that our flows did not reach a fully collisional state, falling into the
dense flow regime [15]. In this regime particles remain closely packed and interact by
enduring contacts and few inter-particle collisions. A contact network and unevenly
distributed structures evolve throughout the flow. The length scale of the correlated
motion (i.e. the degree to which particles act as a cluster rather than individually) has
an impact on the measured granular temperature: internal momentum transport is
carried by the coherent motion of correlated particles with a low level of stress
generation; collisions are less frequent which in turn reduces the calculated granular
temperature. Nevertheless, the boundary condition at the sidewall may also act to damp
the generation of fluctuating motion [16]. Additionally, uncertainties might also be
connected with the PIV algorithm we used to estimate granular temperature. Granular
temperature profiles in Fig. 3 do not show a linearly increasing gradient, but different
peaks appear. The peaks in curves shown in Fig. 3 are representative of the highly
variable nature of the flow. Similar profiles taken on the top of flows from numerical
(a) (b)
30 30
20 20
y (mm)
10 10
0 0
0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.000 0.001 0.002 0.003 0.004 0.005 0.006
T (m2/s2) T (m2/s2)
Figure 3. SaT1 (a) and BeT2 (b). Granular temperature profiles.

simulation and through the sidewall of saturated granular flow in experimental work
have been previously observed [9,10]. While it is possible that fluctuation velocity and
therefore, granular temperature, averaged over longer a period of time may produce
smoother curves, interestingly, a system transition forms in the middle of the flow in
both tests. Granular temperature drops at around 20 mm for SaT1 and 15 mm for BeT2.
Subsequently, the gradient returns to be sloped in the flow direction. This may be
indicative of a change of stress transfer mechanism to one which is more frictional. The
discrepancy towards the free surface may be linked to the loss of textural information
in PIV when the flow becomes more diluted.
Gopalan and Shaffer [17] argued that many experimental techniques are unable to
accurately measure the random components of velocity. Before decomposing the
velocity field into its mean and random quantities, the necessary sample rate for full
temporal resolution variations should be established. Our results are based on images
captured at 10,000 fps. The same profiles plotted for 5,000 fps only shows slight
reductions of granular temperature. In contrast, at 2,000 fps the system transition
disappears indicating a loss of recorded fluctuations and consequently of granular
temperature, hence it would appear that 5,000 fps is the limit here.
4. Conclusions and Future work
Granular temperature is a variable upon which constitutive behaviour of granular

material can depend. Thus, direct measurements of this entity are very useful to
elucidate the extremely variable nature of granular flows. A simple inclined chute
geometry has been used to produce granular flows in a laboratory-scale environment.
We have extracted granular temperature from two different flows by means of
statistical analysis and PIV. The analysis was carried out at high temporal resolution
due to the unsteady nature of the motion and showed that while velocity measurements
taken over a long time-scale may suggest steady flow, over a shorter period, significant
correlated velocity fluctuations may also occur which could be due to granular
turbulence rather than granular temperature via collisions, as usually understood.
Despite the few studies mentioned in this paper, the authors do not know of any
other study that gives a direct measure of granular temperature for flows down inclines
by means of PIV. Hence, we need to ascertain the reliability of our method. It would be
interesting to compare measurements provided by PIV against alternative techniques
such as particle tracking (PTV). Moreover, additional work has to be conducted to
verify the influence of the interrogation region and temporal resolution.
Our results are in disagreement with values of granular temperature for collisional
state motions found in the literature. It appears that our flows fall into the dense flow
regime. While has been suggested that kinetic theory may also be valid for relatively
dense flow [18,19] a more careful judgment on the laboratory conditions has to be
made. Perhaps, creating a more collisional flow can provide further insights on the
validity of this method and how far it can be pushed.
We have proposed a method that might provide a means to estimate the true
granular temperature and which may prove particularly useful for well-graded
(polydisperse) systems [10]. The mechanics presented here are strongly influenced by
the material characteristics of the flow particles, friction and additional effects induced
by the boundary conditions. We are aware of the implications induced by the sidewall
and the material we have chosen. However, this is a preliminary study devoted to better
understanding how far this approach could provide valid and explicit results. Our final
goal is to verify the applicability of this method into three-dimensional cross-sections
to elucidate the highly inhomogeneous rheological behaviour of natural granular flows.
References
[1] C.S. Campbell, Granular material flows An overview, Powder Technology 162 (2006), 208229.
[2] T. Brzsnyi and R. Stannarius, Granular materials composed of shape-anisotropic grains, Soft Matter 9
(2013), 74017418.
[3] I. Goldhirsch, Rapid granular flows, Annual Review of Fluid Mechanics 35 (2003), 267293.
[4] R.D. Wildman and J.M. Huntley, Experimental measurements and modelling of rapid granular flows,
Powder Technology 182 (2008), 182191.
[5] G.K. Reynolds, A.M. Nilpawar, A.D. Salman, M.J. Hounslow, Direct measurement of surface granular
temperature in a high granulator, Powder Technology 182 (2008), 211217.
[6] E. Azanza, F. Chevoir, P. Moucheront, Experimental study of collisional granular flows down an inclined
plane, Journal of Fluid Mechanics 400 (1999), 199227.
[7] A. Armanini, H. Capart, L. Fraccarollo, M. Larcher, Rheological stratification in experimental free-
surface flows of granular-liquid mixtures, Journal of Fluid Mechanics 532 (2005), 269319.
[8] A. Armanini, M. Larcher, E. Nucci, M. Dumbser, Submerged granular channel flows driven by gravity,
Advances in Water Resources 63 (2014), 110.
[9] D.M Hanes and O.R. Walton, Simulations and physical measurements of glass spheres flowing down a
bumpy incline, Powder Technology 109 (2000), 133144.
[10] N. Sanvitale, E.T. Bowman, Towards measurements of granular temperature inside a flowing
unsteady saturated granular medium, Geomechanics from Micro to Macro 2 (2014), 15531558.
[11] N. Sanvitale and E.T. Bowman, Internal imaging of saturated granular free-surface flows, International
Journal of Physical Modelling in Geotechnics 12 (2012), 129142.
[12] D.J. White, W.A. Take, M.D. Bolton, Soil deformation measurement using particle image velocimetry
(PIV) and photogrammetry, Geotechnique 53 (2003), 619631.
[13] T. Brzsnyi, T.C. Halsey, R.E. Ecke, Avalanche dynamics on a rough inclined plane, Physical Review
E 78 (2008), 011306.
[14] T.G. Drake, Granular flows: physical experiments and their implications for microstructural theories,
Journal of Fluid Mechanics 225 (1991), 121152.
[15] F. da Cruz, S. Emam, M. Prochnow, N.J. Roux, F. Chevoir, Rheophysics of dense granular materials:
discrete simulation of plane shear flows, Phys. Rev. E 72 (2005), 021309.
[16] H. Ahn, C.E. Brennen, R.H. Sabersky, Measurement of velocity, velocity fluctuation, density, and
stresses in chute flows of granular materials, Journal of Applied Mechanics 58 (1991), 792.
[17] B. Gopalan, F. Shaffer, A new method for decomposition of high speed particle image velocimetry data,
Powder Technology 220 (2012), 164171.
[18] K. A. Reddy, V. Kumaran, Applicability of constitutive relations from kinetic theory for dense granular
flows, Phys. Rev. E 76 (2007), 061305.
[19] L.E. Silbert, G.S. Grest, R.E. Brewster, A.J. Levine, Rheology and contact lifetimes in dense granular
flows, Phys. Rev. E 99 (2007), 068002.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-655
Micromechanical insight into the undrained

instability of granular materials
Daniel BARRETOa,1, Luis Felipe PRADA-SARMIENTO b and Alfonso Mariano
RAMOS-CAN b
a
Edinburgh Napier University, United Kingdom
b
Pontificia Universidad Javeriana, Colombia
Abstract. There is no agreement between experimental researchers whether the

point where a granular material responds with a large change of stresses, strains or
excess pore water pressure given a prescribed small input of some of the same
variables defines a straight line or a curve in the stress space. This line, known as
the instability line, may also vary in shape and position if the onset of instability is
measured from drained or undrained triaxial tests.
Failure of granular materials, which might be preceded by the onset of
instability, is a subject that the geotechnical engineers have to deal with in the
daily practice, and generally speaking it is associated to different phenomena
observed not only in laboratory tests but also in the field. Examples of this are the
liquefaction of loose sands subjected to undrained loading conditions and the
diffuse instability under drained loading conditions.
This research presents results of DEM simulations of undrained triaxial tests
with the aim of studying the influence of stress history and relative density on the
onset of instability in granular materials. Micro-mechanical analysis including the
evolution of coordination numbers and fabric tensors is performed aiming to gain
further insight on the particle-scale interactions that underlie the occurrence of
this instability. In addition to provide a greater understanding, the results presented
here may be useful as input for macro-scale constitutive models that enable the
prediction of the onset of instability in boundary value problems.
Keywords. Undrained instability, DEM, instability line
1. Introduction
In the design of conventional geotechnical structures (footings, slopes, retaining walls,

excavations, etc.), two limit states are usually considered: an ultimate limit state
associated with conditions of failure, and a serviceability limit state related to large
displacements in the structure. However, geotechnical engineers must also guarantee
that their structures are stable under working conditions, long before reaching the point
of failure [1].
Instability can be easily defined from an engineering standpoint: a geotechnical
system (slopes, excavations, retaining walls, etc.) is stable if it is insensitive to small
perturbations. Given that all designs are fraught with many uncertainties, not only in
the applied loads but also in the parameters of the constitutive materials, it is necessary
1
656 D. Barreto et al. / Micromechanical Insight into the Undrained Instability
that small changes in the prescribed input do not induce harmful consequences for the
structure under design[2].
Geotechnical engineers deal on a daily basis with instability processes in granular
materials, both in the laboratory and in the field. Three modes of failure of granular
materials have been identified in the literature, with the common feature that in all
cases, failure is preceded by a sudden increase of strain upon a small increment of load
(instability) [3]:
+ Liquefaction of loose sands subjected to undrained loading conditions (diffuse
undrained instability).
+ Shear bands on dense sands subjected to drained loading conditions.
+ Diffuse instability under drained loading conditions in loose and dense sand.
This mechanism is associated with debris flow resulting from an increase of
the water table level or with a decrease of the confining pressure.
This article will focus on the instability related to the first mode of failure, i.e.
static liquefaction of loose sands that undergo undrained shearing. Static liquefaction is
characterized by a sudden build-up of excess of pore pressure under monotonic
undrained loading conditions, associated with large increments of the shear strain rate.
Alarcn-Guzmn et al. [4] described flow liquefaction as the collapse of the structure
of the sand. Flow liquefaction is thought to be caused because the structure of the
contractive sands is metastable. Since the skeleton is collapsible, small shear strains
produce a sudden arrange of the grains including loss of contacts between them. As the
shear process is undrained, the volumetric load is transferred to the pore water pressure
and the resistance decreases. The sand undergoes large deformations in a small
increment of time, and its final strength is controlled by the steady state condition. It
follows that flow liquefaction is commonly associated with large strains within the soil
mass. However, the onset of instability occurs in the small-strain region under both,
monotonic and cyclic conditions of loading. This is the reason why there is a
connection between the fabric, which mainly affects the behavior under small strains,
and the possibility of flow liquefaction of sand which occurs when the soil reaches
large levels of strains [3].
The diffuse undrained mode of failure has been analyzed both from analytical and
experimental perspectives. Many authors identify the diffuse undrained instability as
the peak point of the p-q stress-path curve of an undrained triaxial test on loose sand
samples [5]. A line joining the peak points in the p-q plane [6] helps to define the
region of instability, comprised between the line of instability and the failure surface
(Mohr-Coulomb, Lade-Duncan, Matsuoka-Nakai, etc).
Borja [7] and Andrade [8] derived analytical expressions to predict the onset of
static liquefaction in granular materials, based upon Nova's concept of loss of
controllability [9]. Ramos [3] extended the approach developed by Andrade [8] to
detect the onset of liquefaction, using the advanced elasto-plastic model for granular
materials with kinematic hardening formulated by Dafalias and Manzari [10].
However, the analytical approach to understand the onset of diffuse undrained
instabilities has focused exclusively on concepts derived from the continuum
mechanics, like second order work, loss of uniqueness, etc. This article presents one of
the first insights into the instability of undrained triaxial tests on loose samples from a
micromechanical perspective. Authors like Gong [11] and Gu [12] have also analyzed
the evolution of some micromechanical variables during shearing, but have
D. Barreto et al. / Micromechanical Insight into the Undrained Instability 657
concentrated their attention mainly on the critical state and the conditions for
liquefaction, with little discussion on the conditions for instability. Simulations of
isotropically and anisotropically (K0) compressed samples brought to failure under
undrained triaxial shearing were performed using the discrete element method (DEM)
using the open source platform YADE [13]. Evolution of stress paths and of some
micro-mechanical state variables like the coordination number, and the principal
components of the fabric tensor are monitored throughout the shear phase to gain
further insight into the particle scale interactions that underlie the phenomenon of
undrained stability.
2. DEM Simulations
For the two three-dimensional DEM simulations presented here polydisperse

assemblies of 10000 spheres were used with a particle size distribution that replicates
real Leighton Buzzard sand. A non-linear Hertzian contact model was used to calculate
inter-particle forces in the normal direction and a frictional Coulomb law for those in
the tangential direction. A summary of the input parameters used in the simulations is
included in Table 1. For comparison purposes, two specimens were generated with
isotropic and anisotropic (K0) initial conditions as described in section 2.1.
Table 1. Input parameters for DEM simualtions

Parameter Value
Youngs modulus, E 70 GPa
Poissons ratio, 0.22
Coefficient of inter-particle friction, 0.57
Specific gravity of particles, Gs 2.65
2.1. Specimen Preparation
For the generation of the isotropic specimen, particles were generated randomly within
a periodic cell with dimensions of 3030 30 mm. The particles were then compressed
with an inter-particle friction coefficient equal to 0.57 in order to achieve a specimen
with very loose initial density. Isotropic compression of the periodic cell was
performed using a servo-control algorithm that altered the strain rate and the position of
the boundaries and particles in order to achieve an isotropic confining stress equivalent
to 250 kPa. During isotropic compression, the maximum strain rate used was 0.1/s with
a global damping of 0.001 applied to the accelerations and velocities of all particles at
every time step. The compression stage was continued until the unbalanced force
(defined as the ratio between the mean particle force and the maximum particle force)
was less than 0.03. At this point the average number of contacts per particle was
constant and the assembly was considered to be under static equilibrium. The resulting
initial void ratio for this assembly was 0.650 with an average of 3.56 contacts per
particle.
Based on the final size of the isotropically compressed sample, an initial periodic
cell with size of 61 21 21 mm was used to randomly locate the sand particles and
create the initially anisotropic specimen. In order to reproduce anisotropic (K0)

compression the smallest dimensions of the specimen were kept unchanged and the
periodic cell compressed using a servo-control algorithm in order to achieve a normal
stress in the longest specimen direction equal to 315 kPa. This stress was required in
order to achieve a specimen which being anisotropically consolidated had the same
initial void ratio and mean effective stress (p'=(1+ '2+ '3)/3=250 kPa) as the
isotropically compressed specimen. A trial and error process was then used by varying
both the normal stress and the inter-particle friction coefficient which was finally set
(for the anisotropic specimen) equal to 0.46 during the compression stage. Damping
conditions and equilibrium conditions were kept the same as for the isotropic specimen.
The average number of contacts per particle for the anisotropic specimen was 3.66,
already illustrating the difference in fabric anisotropy with respect to the isotropic
specimen. Hence, as a result of these generation procedures two specimens with
identical void ratio and mean effective stress, but different initial fabric anisotropy
were created. Fabric is normally quantified with the fabric tensor, ij defined by Satake
[14] and given as
S
PQR T6 ! !
!VS UQ UR (1)
6
where UQ! is the unit vector describing the contact normal orientation and Nc is the
number of contacts. Such a tensor has eigenvalues which quantify the magnitude of
anisotropy in three (principal) orthogonal directions given by their corresponding
eigenvectors. The eigenvalues for the isotropic specimen were 0.34, 0.33 and 0.33
indicating an isotropic fabric (equal concentration of contacts in every direction). For
the anisotropic specimen these were 0.36, 032 and 0.32 evidencing a higher
concentration of contacts in the direction coinciding with the direction of (K0)
deposition.
2.2. Undrained shearing procedures
Both the isotropic and anisotropic specimens were sheared under constant volume
(undrained) conditions with a coefficient of inter-particle friction, =0.57 and a global
damping equal to 0.001. Each specimen was sheared under undrained conditions by
keeping constant volume conditions. As such, the length of each specimen in the
(vertical) x-direction was reduced with a strain rate equal to 0.1/s and the lengths in the
y- and z- directions simultaneously expanded with a strain rate of 0.05/s. The choice of
damping and strain rate values is the result of a preliminary study in which micro-
mechanical parameters and energy measures such as kinetic energy, unbalanced force
and energy dissipation were monitored. To further guarantee the constant volume
condition the allowed inter-particle overlap was monitored and limited, resulting in
void ratio changes on the eleventh significant figure only at large axial strains (a>40%).
2.3. Macro-scale observations
Figure 1 illustrates the stress paths for both the isotropically and anisotropically
compressed specimens. It can be observed that although the initial stiffness of the
isotropic specimen is high, the anisotropic specimen experienced an initial hardening
response producing and initial increase in the mean effective stress p. A peak
condition in the p-q diagram as noted by Vaid and Chem [5] is easily distinguished for
both specimens. It can be observed that condition of instability is less stable for the K0
condition. It must be highlighted that the peak of both graphs is well below the strength
envelop calculated from DEM simulations of drained triaxial tests on the isotropic
specimen at different stress levels (with a resulting angle of shearing resistance of 21o).
Another point to note from Figure 1 is the existence of static liquefaction as evidenced
by the convergence of the stress paths at the origin of the graph. This is of course the
result of an increase of pore pressure which is equal in magnitude to the initial mean
effective stress (p).
Figure 1. Stress paths on the p-q space for the (a) isotropic and (b) anisotropic specimens.
In order to define the point of instability, the evolution of the deviatoric stress (q)
against axial strain (a) is illustrated in Figure 2. The dotted lines show that both
specimens reached the point of instability at very small strain levels, as also observed
by [3]. An interesting observation is that in addition to the higher stiffness and less
stable response at the point of instability observed in Figure 1, the anisotropic specimen
requires a relatively larger strain level to reach static liquefaction.
Figure 2. Evolution of the deviatoric strain against the axial strain for the (a) isotropic and (b) anisotropic
specimens.
2.4. Micro-scale observations
It is widely recognised that stress transmission in granular materials is not

homogeneous. It is therefore common to perform micro-mechanical analyses in terms
of the contacts that are effectively transmitting strong forces and hence contributing
greatly to the current stress state. The simplest micro-mechanical parameter that can be
considered is the mechanical coordination number, Zm defined by Thornton [15] as:

6 Y
WX : [ , (2)
Y ZY Y
where Nc is the number of contacts, Np is the number of particles and \]^ and \]S are the
numbers of particles with a single or no contact(s), respectively. This definition hence
assumes that stress transmission can only be carried out with a minimum of 2 contacts.
Figure 3 illustrates that for both specimens the value of Zm reduces sharply as the
liquefied state is approached, but that such parameter does not change significantly in
the region of the undrained stability.
Figure 3. Evolution of the mechanical coordination number, Zm for (a) isotropic and (b) anisotropic
specimens.
Previous studies by Thornton and Anthony [16] and Radjai et al [17] defined
strong contacts to be those carrying a normal force which is equal or greater than the
average normal force at a given instant. In this paper we consider a similar
decomposition to study the principal strong components of the fabric tensor defined
in Equation (1). Figure 4 shows that at the beginning of shearing the major principal
component of the strong fabric increases rapidly up to a point relatively close to the
occurrence of instability. A similar trend can be observed in the opposite direction for
the intermediate and minor principal components of the strong fabric tensor. These
changes indicate that the fabric is becoming increasingly anisotropic as the instability is
approached. When the instability occurs the concentration of contacts in the major
principal stress direction is close to being the highest while it is close to be the lowest
in the intermediate and minor directions. According to previous research by Barreto
and OSullivan [18] this would seem to indicate that as the instability happens the
orthogonal support to the strong force chains is close to be the minimum (and this
analysis was also supported when the evolution of weak fabric tensor forces lower
than average- was evaluated). It must be noted that this loss of support in the
perpendicular direction with respect to the principal stress direction happens with a
relatively constant number of contacts, as elucidated by Figure 3.
Figure 4. Evolution of the principal strong fabric components for the (a) isotropic and (b) anisotropic
specimens.
The study of Barreto and OSullivan [18] also suggests the use of a deviatoric
invariant of the fabric tensor. For triaxial conditions this deviatoric fabric, also in terms
of the strong fabric tensor is defined as dS = 1S 3S, where iS is the corresponding
eigenvalue (major or minor) of the strong fabric tensor. Figure 5 illustrates the
evolution of such parameter for both simulations and demonstrates that in the
proximity of the occurrence of the instability the rate of change of the deviatoric fabric
is close to the maximum for the isotropic specimen, while for the anisotropic specimen
this invariant is close to its maximum on the vicinity of the point of instability.
Figure 5. Evolution of the deviatoric fabric for the (a) isotropic and (b) anisotropic specimens.
3. Conclusions
In the present study two three-dimensional simulations of assemblies of spheres with

identical initial void ratio and mean effective stress, but different initial fabric were
performed, providing further insight into the influence of anisotropy and stress history
on undrained stability. The assemblies were sheared under undrained (constant volume)
conditions and macro- and micro-scale analyses were performed.
The point of instability was identified from the peak of the stress-strain response
and micro-scale analyses demonstrated that at the point of instability, the anisotropy of
the specimen is (close to) the highest for the anisotropic specimen, and that for the
isotropic specimen the rate of change of the fabric is also close to the maximum.
These observations show that it is possible to use alternative measures in addition
to those normally assessed from physical experiments such as deformations, pore water
pressure, second order work, loss of uniqueness, etc. in order to identify the undrained
instability and the reasons for its occurrence. This study also showed that there is
profound influence of the initial fabric on the occurrence of this mechanical instability.
Further simulations are required to observe the influence of different mean effective
stresses and initial densities of granular samples on the onset of the instability.
4. Acknowledgements
The authors would like to express their gratitude to the British Council as it provided
funds for both Dr Barreto and Dr Prada-Sarmiento to stay in Colombia and the United
Kingdom, respectively, with two individual grants from its Researcher Links
programme.
References
[1] Di Prisco, C., Imposimato, S., and Vardoulakis, I., Mechanical modelling of drained creep triaxial tests
on loose sand, Geotchnique 50 (2000), 7382.
[2] Chambon, R., Caillerie, D., and Viggiani, G., Loss of uniqueness and bifurcation vs instability: some
remarks, Revue Francaise de Genie Civil 8 (2004), 517535.
[3] Ramos, A., Instabilities in sands, PhD Thesis. Universidad de Los Andes, Colombia (2011).
[4] Alarcn-Guzmn, A., Leonards, A., Chameau, L., Undrained monotonic and cyclic strength of sands,
Journal of Geotechnical Engineering 114 (1988), 10891108.
[5] Vaid, Y., Chern, J., Cyclic and monotonic undrained response of saturated sands, Advances in the Art of
Testing Soils under Cyclic Conditions. ASCE. (1985), 7382.
[6] Lade, P. V., Instability and liquefaction of granular materials, Computers and Geotechnics 16 (1994),
123151.
[7] Borja, R., Condition for liquefaction instability in fluid-saturated granular soils, Acta Geotechnica 1
(2006), 211224.
[8] Andrade, J., A predictive framework for liquefaction instability, Geotchnique 59 (2009), 673682.
[9] Nova, R., Controllability of the incremental response of soil specimens subjected to arbitrary loading
programmes, Journal of the Mechanical Behavior of Materials 5 (1994), 193201
[10] Dafalias, Y., Manzari, M., Simple plasticity sand model accounting for fabric change effects, Journal of
Engineering Mechanics 130 (2004), 622633.
[11] Gong, G., Lin, P., Qin, Y., Wei, J., Dem simulation of liquefaction for granular media under undrained
axisymmetric compression and plane strain conditions, Acta Mechanica Solida Sinica 25 (2006), 562
570.
[12] Gu, X., Huang, M., Qian, J., DEM investigation on the evolution of microstructure in granular soils
under shearing, Granular Matter 16 (2014), 91106.
[13] milauer, V., Catalano, E., Chareyre, B., Dorofeenko, S., Duriez, J., Gladky, A., Kozicki, J., Modenese,
C., Scholts, L., Sibille, L., Strnsk, J. and Thoeni, K. Yade Documentation (V. milauer, ed.), The
Yade Project, 1st ed., (2010). http://yade-dem.org/doc/.
[14] Satake, M., Fabric tensor in granular materials. In P. Vermeer and H. Luger (eds), Proceedings of
IUTAM Symposium on Deformation and Failure of Granular Materials (1982), 63-68. A.A. Balkema
[15] Thornton, C., Numerical simulations of deviatoric shear deformation of granular media, Geotchnique
50 (2000), 43-53.
[16] Thornton, C., Anthony, S.J., Quasi-static deformation of particulate media. Philos. Trans. R. Soc.
London A Math. Phys. Eng. Sci. 356 (1998), 2763-2782.
[17] Radjai, F., Wolf, D.E., Jean, M., Moreau, J.-J., Bimodal character of stress transmission in granular
packings. Phys. Rev. Lett. 80 (1998), 61-64.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-663
Sampling and sample preparation on

structure effects
Carlos Rodrigues a, Lus Lemos b1
a
Polytechnic Institute of Guarda, Portugal
b
Coimbra University, Portugal
Abstract. This paper evaluates the disturbance during the process of sampling and
specimen preparation using natural and artificial samples, of a residual soils from a
granite saprolite located in Guarda NE region of Portugal. This soil profile is
described and the soil characterized and classified. The intrinsic mechanical
behaviour of these materials is evaluated on samples compacted to different
densities. Also, this behaviour is analysed on natural and artificial structured
materials.
Keywords. Artificial soils, disturbance effects, residual soil, sampling.
1. Introduction
The granitic residual soils yielded by in situ changes of the bedrock usually exhibit
considerable heterogeneity, both laterally and vertically, and a complex structural
layout. These soils have different fabrics, depending on the degree of weathering and
leaching they underwent.
As weathering progresses, the primary inter-particle bonds between the grains are
weakened and the inter-granular void is increased by leaching. Afterwards, weathering
makes the feldspars and the micas unstable, allowing leaching to occur, with the
creation of a network of intra-granular voids. In addition, the more stable minerals,
mostly quartz grains, are weakly bonded by highly weathered grains of feldspars and
micas to form a soil skeleton that is unstable and sometimes quite open.
2. Characterization of the soil profile
The climate in the study area is temperate and wet, which favours weathering, turning
the granite masses into a permeable sandy frame. Being submerged in water in the wet
season and dry during the summer months, up to depths of 5.0 m to 6.5 m, it creates the
conditions to favour the constant weathering of the rocky substrate.
The saprolite exposed in some of the slopes into the study area is generally light in
colour, with shades of greyish-white, with a yellower shade on the surface of the
profile. There are some zones where the yellow colour becomes dark orange.
1
664 C. Rodrigues and L. Lemos / Sampling and Sample Preparation on Structure Effects
Figure 1 shows the particle distribution of the soils found throughout the profile,
which reveals appreciable particle size homogeneity. The results show, in general,
well-graded soils. Applying the Unified Classification System [1] to the set of samples
mentioned above, it was found that they could all be classified as silty sands (SM). The
fines could be classified as ML, which means these soils have low plasticity. The
position of the water table was assessed, using a piezometric tube installed in the
ground, and it was found that this fluctuated during the work, at a depth between 5.0 m
and 6.5 m, over two years.
Grain size (%) w(%) Void ratio 7 (kN/m3)
0 25 50 75 100 10 12 14 16 18 0,4 0,6 0,8 16 18 20 22

0
Horizon I clay
1
Saprolitic
2 soil, with
opened
3 matrix
Depth (m)
4
Horizon II silt sand gravel
5
Saprolitic
6 soil, with
closed
7 matrix
Figure 1. Soil classification and index properties with depth.
3. Intrinsic behaviour
The intrinsic behaviour of the granite residual soil was characterized from the samples
collected at the surface. The samples were collected by means of block sampling. The
soil structure, derived from the bedrock, was destructured manually, keeping its
original composition and particle size. The intrinsic mechanical behaviour of the
granitic residual soil was characterized by means of 20 samples, obtained by dynamic
compaction of the samples in several layers, with different initial void ratios and
sheared in the triaxial equipment [2]. Figure 2 shows the intrinsic strength envelop and
the critical state void ratios as a function of p.
Analysis of the results clearly shows a strength envelope corresponding to the
intrinsic/critical state behaviour. The critical state shear resistance angle of cs=35.5o,
which corresponds to a slope M=1.44 in the (p:q) space. The intrinsic void ratio shows
some dispersion. Vaid et al. [3], Coop, M. [4] and Toll et al. [5] stated the same, noting
that the sands in the stable state, like the clays in the critical state, exhibit a single line
in stress space ( is independent of the stress path), while the same does not happen in
space 6:lnp. Figure 2b) shows a band involving the results obtained for the ultimate
values (6:p), as in [5].
C. Rodrigues and L. Lemos / Sampling and Sample Preparation on Structure Effects 665
800 1,8
Critical
700 q = 1.44 p' State Area
Deviatoric stress, q (kPa)
'= 35.5o CSL
Specific volume, 6 = 1+e

600 1,7
500 NCL
400 1,6
300
200 1,5
100 6 = -0.085ln(p') + 1.974

(CSL)
0 1,4
0 200 400 600 10 100 1000
Mean effective stress, p' (kPa) Mean effective stress, p' (kPa)
a) b)
Figure 2. Intrinsic shear resistance envelope and critical state void ratio function of mean effective stress p.
4. Sampling effects
Since residual soils usually exhibit high sensibility to sampling procedures, artificial
cemented samples, prepared in a controlled way, were used to investigate sampling and
sample preparations effects. Specimens constituted by the same in situ material,
thereby preserving their composition, were artificially bonded in order to simulate the
natural material that exhibits interparticle bonds inherited from the bedrock.
4.1. Sampling effects artificial cemented soils
A set of test specimens was prepared. It was made up of residual soil that had been
previously dried and destructured and to which Portland cement and water was added.
The proportions used were those which, matched the results of simple compression
tests of residual soils and artificially cemented specimens. The relative amount of
cement and water and the curing time used are as follows:
dry, destructured residual soil (84 %);
Portland cement (2.5 %);
water (13.5 %);
curing time (14 days).
An appropriate curing chamber with temperature and humidity controlled
respectively at 20o and 95 %, was used. The mixture with the same composition as that
mentioned above was used to prepare several samples, which were taken to be
identical, since they had the same composition and were subjected to the same curing
conditions.
The intact sample was prepared using a bipartite mould of the same size as that
intended for the final specimen (140 mm in height and 70 mm in diameter). The
surfaces were lubricated with mould oil to facilitate the removal of the sample. After
curing, the sample was immediately placed in the triaxial cell, so that it was a test
specimen that had not undergone any sampling or sample preparation procedures, it
was therefore considered intact.
In addition, two more samples were prepared from the same batching. They
were placed in big cylindrical containers of 700 mm in diameter and 800 mm in height.
After 14 days curing the mould was carefully axial sectioned using a rotating trimmer
to avoid confinement during driving of the samplers of 100 mm (A100) and 70 mm
diameter (A70). The 100 mm diameter sample was then worked on and trimmed off to
the diameter of 70 mm.
The artificially cemented specimens underwent triaxial compression tests with
isotropic consolidation and drained shear phase. A pair of internal Hall-effect
transducers was attached to all the specimens to measure locally the radial and axial
strain. The specimens were fully saturated. After saturation, the specimens underwent
controlled isotropic consolidation (30 kPa/h) to low effective stress (8c=15 kPa), to
ensure the preservation of their artificial cemented structure. Shearing was drained,
keeping 83 constant, at a constant strain rate of 0.05 mm/min.
Figure 3 displays the stress ratio and volumetric strain as a function of axial strain.
The maximum strength of all the samples is more or less the same and appears to
precede the point of maximum expansion (d max) as shown in Figure 3. The axial strain
to the peak increases and the initial stiffness decreases considerably, as the sample
disturbance increases, due to sampling and sample preparation, Fig 3a) and 4a).
3 -6
2,5
Volumetric strain, ev (%)
-4
Stress ratio, q/p'
1,5 -2
Undisturbed Undisturbed
1
A100
0 A100
A70
0,5 A70
dmax
dmax
0 2
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Axial strain, a (%) Axial strain, a (%)
a) b)
Figure 3. Triaxial tests on artificially cemented specimens consolidated at low pressures, showing the points
of maximum dilatancy (dmax): a) q/p:a relations; b) v:a relations.
Figure 4a) shows the secant stiffness (Esec) variation with axial strain (a). There is
a significant decrease of the maximum stiffness values, in the specimens that
underwent sampling and sample preparation procedures, when compared with the
intact sample. The sample with larger diameter 100 mm yielded only a slightly better
result than the sampler with the same diameter as the specimen, exposing the influence
of sample preparation procedures in reducing the diameter from 100 mm to 70 mm.
In Figure 4a) are identified three discontinuities, which may be considered as yield
points. The first yield was established in logEsec:loga space and relates to the point
where the initial constant stiffness ends. This yield corresponds to the end of the initial
linear elastic behaviour, reached for very small strains (less than a=0.005%). The next
discontinuity, at about 0.07% of axial strain, corresponds to the second yield, where it
is observed an increase in the rate volumetric deformation v in relation to the triaxial
distortion s, as was considered by Alvarado et al. [6]. The first two yields correspond
to Y1 and Y2 defined by Jardine [7]. The last yield (a=2 to 3 %), established in
logEsec:loga space, corresponds to a point in which the structure is dramatically
destroyed at about the maximum shear resistance.
1000
3 q /p'
Cemented trend
2,5
100
2 Undisturbed
A100
10
Esec (MPa)
1,5 A70
1st Yield
Frictional trend
1 of
intrinsic behaviour 2nd Yield
1
Undisturbed A100 Gross Yield
A70 1st Yield 0,5
CS
2nd Yield Gross Yield
0,1 0
0,001 0,01 0,1 1 10 -1 0 1 2
Axial strain, a (%) Dilatancy, d=9v/9s
a) b)
Figure 4. Triaxial tests on artificially cemented specimens consolidated at low pressures, showing the points
of maximum dilatancy (dmax): a) q/p:a relations; b) v:a relations.
Observing Figure 4b) we can see that the intact sample, which was not submitted
to sampling or sample preparation, there is an initial increase in the dilatancy d=9v/9s
up to values of 1.17 (a=0.007%). Next the stress ratio (q/p) increases from 0.08 up to
1.97 keeping the dilatancy nearly constant, with only a slightly decrease from 1.17 to
1.07. It may be postulated, that at this point there was a breakage of the cement bonds,
which corresponds to axial deformations within the range of (0.007%<a<0.02%). It
should be noted that the final point of this interval is almost coincident with the second
yield Y2 (a=0.06%). Next, the dilation tends to change rapidly from compression
(d=1.07) to expansion (d= -0,89), for axial extensions respectively equal to 0.02% and
2.43%, with only a small increase in the stress ratio (q/p), from 1.97 to 2.50. The
maximum shear resistance (q/p=2.5) is coincident with the point of maximum
expansion. After the peak the sample softens and tends to the critical state. In general,
the minimum stress ratio does not equal the critical state value observed for the
reconstituted samples, due to deformation concentration and localized failure.
For the samples A100 and A70 that underwent sampling and preparation
procedures, the behaviour, in Figure 4b), the dilatancy increases to a maximum
compression value (:1.2), at small stress ratios (q/p=0.6), much smaller than the value
of 1.97 obtained for the intact sample, and next steadily decreases the dilation toward
negative values (expansion). The stress ratio attained values close to the maximum
(q/p:2.4) when the dilatancy was zero. The maximum stress ratio was observed at the
point of maximum expansion, as for the intact sample. In these tests a nearly constant
value of dilation with an increase in the stress ratio values was not observed. The
cement bonds were seriously affected during sampling and sample preparation.
However, the stress ratio (q/p) is above the frictional stress ratio, which may indicate
some cement bond contribution. This clear indicates that the process of sampling and
sample preparation affects the structure and weakens cement bonds.
4.2. Sampling effects in natural granitic residual soils
The extent of the disturbing effects of this sampling methodology was further evaluated
by carrying out a series of triaxial tests on natural samples collected at the surface, with
open tube samplers, and also on test specimens collected by block sampling, using
identical test conditions. The disturbing effects have been widely recognized [8], [9],
[10], [11].
Triaxial samples of 70 mm diameter and 140 mm height were prepared through the
following procedures:
- Sampling with A100 sampler samples were taken in situ with a 100 mm
internal diameter and then trimmed to 70 mm diameter.
- Sampling with A70 sampler samples were obtained in situ with a 70 mm
internal diameter sampler.
Block sampling two methods were used to collect block samples:

- Block B1 one block (25;25;25 cm) was worked using the ordinary tools
normally used for trimming the sides and cutting the base of the block. The
final test specimen was obtained using a saw and wire to scrape the sides of a
pre-test specimen with an initial diameter of about 100 mm, until the desired
dimensions were achieved (70 mm diameter, 140 mm height).
- Block B2 another block was prepared using a long-toothed saw, used for
working the sides of the block, and notched wire to cut the base, more or less
in line with the method described by Cresswell [12]. The test specimen was
initially prepared by wearing down the block with a disc brush attached to a
drill, which was moved over the block. When it was nearly the size of the
specimen, a steel brush was used to carefully wear down the side of the
sample until the desired size was achieved (<=70 mm, H=140 mm).
Triaxial compression tests were carried out with and isotropic consolidation phase
and a drained shear phase, as already described above for the artificial cemented
samples.
The plot of the stress-strain and secant modulus-strain relations of the artificially
cemented granitic soils and natural granitic soils is shown on Figure 5. Shear stiffness
obtained through in situ Cross-hole tests is also represented on the same figure. The
initial stiffness of the intact artificially cemented soil show a value of the same order of
magnitude as the in situ Cross-hole tests, that is just convenient for the following
discussion on the influence of sampling and the sample preparation. The peak of shear
resistance and stiffness showed a decrease when the undisturbed artificial bonded
samples are compared with samples which were carefully prepared. The natural
samples Block B2 carefully prepared from a block sample showed an equivalent
notable decrease in stiffness when compared with the Cross-hole value. The volumetric
expansion in natural soil (block B2) is about twice than all the other samples of residual
soils or artificial cemented. The fabric of natural and artificial samples should be quite
different. It is clear, from the results, that even carefully prepared natural specimens
(block B2) and artificial bonded specimens (A100) yield results considerably lower
than those achieved by the truly intact specimen.
Sample preparation, which goes through a process of reduction of its dimension
(70 mm diameter and 140 mm height), is sufficient to destroy the bond strength and
decrease the peak strength and stiffness. Alvarado at al. [6] observed bond degradation
due to unloading, showing a considerable degradation of the shear modulus.
The sampling disturbance which occurs when we use tube samplers of 70 mm
diameter, whose dimension is the same of the triaxial test, did not add a significant
reduction to the peak shear resistance or to the shear modulus, when compared with the
sample collected by block, using classical techniques to trim the sample, see Figure 5
The results from block B1 and B2 clearly demonstrate that investment on sampling
resources to reduce disturbance in structured soils is annulated if careful techniques are
not used to prepare the specimens.
1000 Natural soils

2,5 Block B2
CH-tests A100
A70
2 Artificial soils
100 Undisturbed
Stress ratio, q/p'
A100
Esec (MPa)
1,5 Natural soils A70

Block B2
A100
1 10
A70
Artificial soils
0,5 Undisturbed
A100
A70 1
0
0 5 10 15 0,001 0,01 0,1 1 10
Axial strain, a (%) Axial strain, a (%)
a) b)
Figure 5. q:a and Esec:a relations for triaxial tests carried out on natural and artificially cemented specimens.
Analysis of the results for the initial Youngs modulus (E 0) obtained in situ with
the seismic cross-hole tests and those from the triaxial tests, for small strains (E0.001%)
reveal very significant differences (of about 10 times), even in tests performed in a
theoretically better quality specimen (block B2). This substantial difference is due both
to the fact that the shear modulus in the two cases is determined for quite distinct
deformation levels, and to the unavoidable disturbance of the samples that caused
considerable structural damage to the materials. This resulted in significant loss of
mechanical characteristics, even when painstaking sampling techniques were used,
which shows the existence of extremely fragile inter-particle bonds that are susceptible
to extreme disturbance due to total stress change.
5. Concluding remarks
The main conclusions to be drawn from this are as follows:

a) The points corresponding to the ultimate strain states for the triaxial tests
carried out on compacted specimens allow the definition of a single line in the stress
space q:p (critical state). These points are related with the intrinsic behaviour of the
granitic residual soil. The points corresponding to the ultimate state conditions exhibit
greater dispersion in the definition of this line in the space v:lnp.
b) The effects of disturbance due to the sampling technique and sample
preparation are felt in terms of sample quality. Sampling techniques and sample
preparation clearly influence both the stiffness parameters and the peak shear resistance.
The deformation needed to mobilize the peak state increases with poorer sample quality.
c) The laboratory shearing resistance shows there was no evidence of influence
of inter-particle bonding on the natural samples. The cement bond is weak and will be
destroyed during collection and sample preparation.
d) Block sampling, usually accepted as the least disturbing technique may turn
out to be in vain if gentle procedures are not used to work the specimen. The total
stress change effect during sampling will certainly play an important role in soils with
weak bonds.
Acknowledgement
The research presented in this paper was supported by the Polytechnic Institute of
Guarda, by the University of Coimbra and by the Fundao para a Cincia e
Tecnologia through Research Centre of Civil Engineering.
References
[1] ASTM D2487-98, Standard Classification of soils for Engineering Purposes (Unified Soil Classification
System), American Society for Testing and Materials (1998).
[2] Rodrigues, C.M.G., Geotechnical characterization and geomechanic behaviour study on saprolitic granite
soils from Guarda. PhD thesis (in Portuguese), Coimbra University, Portugal (2003).
[3] Vaid, Y.P., Chung, E.K.F., Kuerbis, R.H., Stress path and steady state. Canadian Geotech. J., 27[1]
(1990), 1-7.
[4] Coop, M.R., The mechanics of uncemented carbonate sands. Gotechnique 40[4] (1990), 607-626.
[5] Toll, D.G., Malandraki, Ali Rahman, Z. and Gallipoli, D., Bonded Soils Problematic or Predictable? Proc.
2nd Int. Conf. on Problematic Soils, Malaysia, Singapore: CI Premier, (2006), 55-62.
[6] Alvarado, G., Coop M.R., Willson, S.M., On the role of bond breakage due to unloading in the behaviour
of weak sandstones. Gotechnique, 62[4] (2012), 303316.
[7] Jardine, R.J., Observations on the kinematic nature of soil stiffness at small strains. Soils and
Foundations, 32 (1992), 111-124
[8] Clayton, C.R.I., Siddique, A., Tube sampling disturbanceforgotten truths and new perspectives. Proc.
Inst. Civ. Engrs. Geotech. Engng, 137, (1999), 127-135.
[9] Siddique, A., Farooq, S.M., and Clayton, C.R.I. (2000). Disturbances due to tube sampling in coastal
soils. J. Geot. Geoenv. Eng., 126(6), pp. 568-575. ASCE, New York.
[10] Hight, D.W. Sampling methods: evaluation of disturbance and new practical techniques for high quality
sampling in soils. Keynote Lect. Proc. 7th Portuguese Nat. Cong. of Geotechnics, (2000), Porto.
[11] Rodrigues, C.M.G., Lemos, L.J.L., SPT, CPT and CH tests results on saprolitic granite soils from
Guarda, Portugal. 2nd Int. Site Characterization, ISC2, Porto, 2, (2004), 1345-1351.
[12] Cresswell, A., Block sampling and test sample preparation of locked sands. Gotechnique 51(6), (2001),
567-570.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-671
Validation of Contact Model between Two

Cylindrical Rods and DEM Simulations of
Biaxial Tests on a Rod Packing
a,1 a b
Zhaofeng LI , Yu-Hsing WANG and Xia LI
a
Department of Civil and Environmental Engineering, The Hong Kong University of
Science and Technology, Hong Kong, China
b
Materials, Mechanics and Structures Research Division, Process and Environmental
Research Division, Faculty of Engineering, The University of Nottingham, NG7 2RD,
U.K.
Abstract. Elastic theory has demonstrated that the contact behavior between
cylindrical rods with parallel axes is a shape, size and supporting-way dependent
problem. In the 2D discrete element method (DEM) simulation, the most
commonly used contact models (i.e., the Hertz-Mindlin model and linear model)
do not accurately describe the force-displacement relationship of cylindrical
contact. Hence, the main objective of this paper is to obtain a more precise
solution of cylindrical contact model based on the elastic theory and finite element
analysis (FEA), and then to incorporate the model into a 2D DEM simulation. In
this study, the normal force-displacement (NFD) model follows the classic theory
where a cylinder is subjected to a diametrically opposite force, while the
associated tangential force-displacement (TFD) model is derived from the
relationship between normal and tangential compliances under plane-strain
condition. FEA verified the proposed model by utilizing the commercial FE
software, ABAQUS. This FEA-validated contact model is implemented in the
open-source DEM software OVAL. Two DEM simulations of biaxial tests on a
rod packing using the FEA-validated contact model and linearly approximated
contact model are carried out, respectively. Due to the similarity at the contact
level, both samples have comparable stress-strain and volumetric responses.
Keywords. Contact model, discrete element method, finite element analysis
1. Introduction
To better visualize the behavior of granular material, many research works have been
carried out using biaxial tests with long cylindrical rods [1, 2, 3]. These biaxial tests
brought benefits to interpreting micro-mechanics and being compared with DEM
simulation intuitively [4].
In order to reproduce such biaxial test by DEM simulation, the cylindrical contact
model should be well studied and calibrated. However, two most commonly used
models (i.e., the Hertz-Mindlin model and linear model) are not designed for the
cylindrical contact. The Hertz-Mindlin model is only applied to the 3D contacts, while
1
Corresponding Author: Zhaofeng LI, Department of Civil and Environmental Engineering, The Hong
Kong University of Science and Technology, Hong Kong, China; E-mail: [email protected].
672 Z. Li et al. / Validation of Contact Model Between Two Cylindrical Rods and DEM Simulations
the linear model is sometimes mistaken as the analytical solution [5]. In fact, the
cylindrical contact behavior depends on particle shape, size and even the supporting
way [6]. Hence, the first objective of this study is to propose a more precise contact
model between cylindrical bodies with parallel axes.
The contact model sometimes does significantly affect the sample response in
DEM simulations. For instance, in the case of Iwashita & Oda [7], the shear band can
only be well simulated under the consideration of rolling resistance model. Therefore,
to introduce a good contact model, the criterion mainly lies on its clear physical
meaning and its capability of better reproducing sample responses. Hence, it is of
necessity to examine the validity of the proposed model through DEM simulation,
which is the second objective of this study.
This paper begins with the discussion of cylindrical contact model based on elastic
theory and then conducts the FEA for model validations. To examine the proposed
model using DEM simulation, the open-source code OVAL provided by Kuhn [8] is
utilized here. Two DEM simulations with the proposed model and the linearly
approximated model were carried out, respectively. The comparison ultimately tells the
influence of the proposed models from the viewpoint of macroscopic stress-strain and
volumetric responses.
2. Contact Models of Cylindrical Contact
2.1. Normal Force-displacement (NFD) Model
The contact models discussed here are only applied to the contact between long
cylindrical rods with parallel axes, namely the plane-strain cylindrical contact. Figure
1(a) shows that a circular rod is subjected to a diametrically opposed force P. For this
case, the NFD model between the contact point C1 and the center point O1 within the
elastic scope has been given as [6, 9]
where z1 is the indentation between C1 and O1; E and are the Youngs modulus and
Poissons ratio of the material, respectively; P is the normal force; and R and a are the
rod radius and the contact radius, respectively.
For a more general condition, when two rods made of the identical material are
normally compressed, by writing the composite modulus as E * = 2E 1 2 , then we( )
have the NFD model between the center points given as
where z is the contact overlapping; and R1 and R2 are the radii of two rods.
Z. Li et al. / Validation of Contact Model Between Two Cylindrical Rods and DEM Simulations 673
2.2. Tangential Force-displacement (TFD) Model
Figure 1(b) illustrates two rods subjected to compression and shearing simultaneously.
Assuming a no-slip condition at the whole contact area, the ratio of the tangential
compliance over the normal compliance can be expressed as
where Cx and Cz are the tangential compliance and normal compliance, respectively;
x is the tangential displacement between the center points O1 and O2; and Q is the
tangential force.
Under the assumption of elastic half-space for each rod, Mindlin [10] has
suggested that this ratio is only the function of the Poisons ratio and shape of contact
area. For a cylindrical contact, when the contact radius tends to be 0.0, the assumption
of half-space is satisfied and the compliance ratio (Eq. 3) tends to be 1.0. Hence,
according to dimensional homogeneity, when two rods are tangentially sheared, the no-
slip TFD model should be given as
where f() is a function of the Poissons ratio of the material, which can be calibrated
using the finite element method or experiment.
Figure 1. Cylindrical contact: (a) a circular rod subjects to diametrically opposite forces P and (b) two rods
are normally compressed and tangentially sheared simultaneously.
2.3. Finite Element Analysis (FEA) on Cylindrical Contact
The finite element analysis (FEA) has become an important tool for contact model
verification and calibration [11, 12]. To validate the aforementioned contact model in
Eqs. (2) and (4), we carry out the finite element analysis using the commercial FE
software, ABAQUS. Figure 2 shows the finite element model and discretization of a
typical case. Two halved rods with a radius of 2.2 mm are compressed and sheared.
Four-node, plane-strain elements are employed (Figure 2(b)) and the edge length of
element gradually decreases to 10-6 m around the contact area (Figure 2(c)). The
contact area is assigned as a no-slip condition.
Both the normal force P and tangential force Q are applied to the node RP-1. For
the NFD model verification, the normal force P linearly increases from 0.0 to 3.0 kN/m,
while for the TFD model calibration, the tangential force Q linearly increases from 0.0
to 0.5 kN/m under a constant normal force, P = 1.5 kN/m. The relative displacement
between the reference nodes RP-1 and RP-2 tells the force-displacement relationship
between two rods. More detailed setting in ABAQUS can be referred to Ding et al. [13].
Figure 3(a) compares the NFD curves obtained from the FEA and Eq. (2). The
FEA curve closely follows the theoretical model. The good agreement between these
two curves demonstrates the reliability of the FEA, as well as the validity of the NFD
model. Figure 3(b) shows the TFD curve obtained from FEA and by changing elastic
properties in FEA, Eq. (4) can be calibrated and then be further given as
In summary, the proposed NFD model in Eq. (2) has been first validated by FEA.
Based on the NFD validation, the proposed TFD model in Eq. (4) is then calibrated by
FEA data and then is further given as Eq. (5).
Figure 2. FEA model for cylindrical contact: (a) FEA model domain; (b) discretization of the domain; and
(c) the fine elements around the contact area.
Figure 3. Model validations: (a) normal force-displacement relationship obtained from the FEA and Eq. (3)
and (b) tangential force-displacement relationship obtained from the FEA result and the TFD model can be
further given as Eq. (5) after calibration. The linearly approximated model is also added for comparison.
3. Details in DEM Simulations
To examine the influence of this FEA-validated contact model, two DEM simulations
are carried out with similar fabric but different contact models. The first simulation
runs with the FEA-validated model, while the second one with a linearly approximated
model. Figure 4 shows the rod packing of the first simulation under the confining
pressure of 50 kPa. The packing composes of 972 mono-sized elliptical particles and
448 circular membrane particles. The top wall shears both packing at a constant
shearing rate until the axial strain comes to 20%. During the shearing, it is supposed
that the contact deformation is small enough and that the plasticity only comes from the
structure deformation and contact sliding. The simulations are conducted in a modified
open-source DEM code, OVAL provided by Kuhn [8]. Parameters assigned to these
two simulations are summarized in Table 1.
Figure 4. Packing of the first simulation using the FEA-validated contact model. Note that the second
simulation using the linearly approximated model has similar fabric and the same particle shapes and sizes.
Table 1. Parameters used in two DEM simulations

Parameters Values
Sample Particle parameters
Particle number 972
Long axis, (mm) 2.60
Short axis, (mm) 1.73
Density, (kg/m3) 1025
Young's modulus, (Pa) 1.78 108
Poisson's ratio 0.12
Approximated normal stiffness, (N/m2) 8.6 107
Approximated tangential stiffness, (N/m2) 5.4 107
Inter-particle friction angel, () 28
Membrane Particle parameters
Radius, (mm) 0.4
Density, (kg/m3) 1025
Normal stiffness, (N/m2) 1.8 106
Tangential stiffness, (N/m2) 1.8 106
Inter-particle friction angel, () 28
The proposed NFD and no-slip TFD models have been implemented in OVAL. In
the coding of the NFD model, two problems raised by Pereira et al. [14] should be
noticed, namely the validity domain and the implicit function form. For the validity
domain, the maximum normal force Plim using Eq. (6) has to be applied. Otherwise,
when particles penetrate each other (i.e. unrealistic overlapping), the NFD model would
produce unacceptable error. For the second problem, to efficiently calculate normal
force under a given displacement, we can employ the iteration method such as the
secant method and Newton-Raphson method to find the corresponding normal force.
4. Simulation Results and Discussions
Figure 5 compares the stress-strain and volumetric responses of two simulations. It can
be seen that the FEA-validated contact model can reproduce the dilatancy behavior for
the dense sample, as well as the critical state for the granular materials at large axial
strain. When it comes to the comparison of two samples, before the axial strain of 6%,
the stress ratio and volumetric strain for both samples are close. Also, only minor
difference can be observed at the peak state and critical state.
The similarity between the two sample responses is probably due to that the
proposed contact model is approximated linear to a certain extent (Figure. 3) and the
fabric of both samples are controlled to be similar as well. That is, before shearing to
large strains, the change at the contact level should be similar, which in turn leads to
the similarity at the macroscopic responses. Hence, it should be noted that behind the
numerical simulation, the contact model plays an important role and the close contact
behaviors may fundamentally give the close results.
Figure 5. Comparison of two simulation results: (a) stress ratio, = (1 3) / (1 + 3) and (b) volumetric
strain, a.
5. Conclusion Remarks
In this study, a more precise cylindrical contact model has been proposed and then is
incorporated into DEM simulations for further examinations. It has been found that the
proposed contact model can well describe the force-displacement relationship for a
plane-strain, cylindrical contact. This contact model is validated using the finite
element analysis and can capture the nonlinearity of the contact behavior. In the DEM
simulations of biaxial tests on a rod packing using OVAL, due to the similarity
between the linearly approximated contact model and FEA-validated one (for the case
studied herein), the differences in the corresponding stress-strain and volumetric
responses are not pronounced.
Acknowledgements
This research was supported by the Hong Kong Research Grants Council
(HKUST9/CRF/09 and HKUST6/CRF/12R).
References
[1] F. Calvetti, G. Combe, and J. Lanier. Experimental micromechanical analysis of a 2D granular material:
relation between structure evolution and loading path, Mechanics of Cohesive frictional Materials 2
(1997), 121-163.
[2] A. Misra and H. Jiang. Measured kinematic fields in the biaxial shear of granular materials, Computers
and Geotechnics 20 (1997), 267-285.
[3] C. O'Sullivan, J.D. Bray, and M.F. Riemer. Influence of particle shape and surface friction variability on
response of rod-shaped particulate media, Journal of Engineering Mechanics 128 (2002), 1182-1192.
[4] M. Ltzel, S. Luding, H.J. Hermann, and D.W. Howell. Comparing simulation and experiment of a 2d
granular couette shear device, The European Physical Journal E: Soft Matter and Biological Physics 11
(2003), 325-333.
[5] L. Rothenburg and R.J. Bathurst. Numerical simulation of idealized granular assemblies with plane
elliptical particles, Computers and Geotechnics 11 (1991), 315-329.
[6] K.L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge, 1985.
[7] K. Iwashita and M. Oda. Rolling resistance at contacts in simulation of shear band development by DEM,
Journal of Engineering Mechanics 124 (1998), 285-292.
[8] M.R. Kuhn, OVAL and OVALPLOT: Programs for analyzing dense particle assemblies with the discrete
element method (2002). Available from: http://faculty.up.edu/kuhn/oval/oval.html
[9] G. Lundberg and E. Yhland. Cylinder compressed between two planes bodies, SKF Reg. 4134, 1949.
[10] R.D. Mindlin. Compliance of elastic bodies in contact, Journal of Applied Mechanics, 16 (1949), 259
268.
[11] C.T. Lim and W.J. Stronge. Oblique elasticplastic impact between rough cylinders in plane strain,
International Journal of Engineering Science 37 (1999), 97-122.
[12] L. Vu-Quoc, X. Zhang, and L. Lesburg. Normal and tangential forcedisplacement relations for
frictional elasto-plastic contact of spheres, International Journal of Solids and Structures 38 (2001),
6455-6489.
[13] J. Ding, S.B. Leen, and I.R. McColl. The effect of slip regime on fretting wear-induced stress evolution,
International Journal of Fatigue 26 (2004), 521-531.
[14] C.M. Pereira, A.L. Amlcar, and J.A. Ambrsio. A critical overview of internal and external cylinder
contact force models, Nonlinear Dynamics 63 (2011), 681-697.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-679
Laboratory test and numerical study of

cyclic behavior of unsaturated liquefiable
sandy soil
Takaki MATSUMARU a, Ryosuke UZUOKA b, Mineo KIGUCHI c
and Tomoyoshi NISHIMURA d
a
University of Tokyo
b
University of Tokushima
c
Integrated Geotechnology Institute Limited
d
Ashikaga Institute of Technology
Abstract. To investigate the cyclic behavior of unsaturated soil, a series of cyclic

triaxial tests and numerical simulations were performed. The tests results indicated
that the liquefaction would occur if the suction value before cyclic loading was
small. This tendency was well reproduced by the numerical simulations using the
conventional elasto-plastic constitutive model proposed for unsaturated soil.
Keywords. Unsaturated soil, liquefaction, cyclic triaxial test, constitutive model
1. Introduction
In 2004, Niigata-ken Chuetsu earthquake with the magnitude of 6.8 occurred in Japan,
and induced enormous disaster widely in Niigata Prefecture [1]. The liquefaction
caused the failure of natural and structure slopes, large settlement and ground
deformation. The embankments consisted of a wide variety of sands with large fine
fraction contents which were not compacted sufficiently, and had commonly low
densities compared to the maximum dry densities obtained from the laboratory
compaction tests. The geotechnical engineers were concerned that both the seepage and
the rainfall events before earthquake have influenced on potential for liquefaction of
earth-structures. It is possible to predict the aspect that the slope subjected to rainfall
was not in fully saturated condition, and to unsaturated soils with high degree of
saturation capability were applied cyclic loading. The aspect can reveal that effective
stress in unsaturated soils prior to earthquake decreases significantly. Therefore, it is
important to understand how the behavior of unsaturated sandy soil having small initial
suction is affected by the cyclic loading.
In this paper, a series of cyclic loading were conducted on the unsaturated soil with
different value of the initial suction. Furthermore, numerical simulations of the tests
were conducted using the conventional constitutive model for describing the behavior
of the cyclic loading of unsaturated soils in order to investigate the applicability of the
model.
680 T. Matsumaru et al. / Laboratory Test and Numerical Study of Cyclic Behavior
2. Test conditions
2.1. Materials properties
Table 1 shows the properties of two different materials used in cyclic triaxial tests. One
is Inagi sand (Material A), and the other is the embankment material which was
obtained from the damaged railway embankment in 2004 Niigata-ken Chuetsu
earthquake [1]. Both materials show similar fine fraction contents and are categorized
into sand with fine fraction. The dry densities of specimens prepared by Material A and
B are different as shown in Table 1. The density of Material A is the density for model
embankment used in the shaking table tests [2, 3], and that of Material B is the density
for the real damaged embankment.
Table 1. Properties of materials used in undrained cyclic triaxial tests.

Soil particle 50% diameter Uniformity Fine fraction Dry density
density Gs (g/cm3) D50 (mm) coefficient Uc content Fc (%) d (g/cm3)
Material A 2.723 0.134 9.29 23.6 1.108
Material B 2.629 0.134 11.65 23.3 1.290
2.2. Test apparatus
The apparatus consists of vibrator, control equipment, triaxial inner and outer cell,
supply system for cell pressure, pore water pressure and pore-air pressure. The cell
pressure, pore water pressure and pore air pressure can be controlled independently.
Inner cell was used for measuring volume change of the soil specimen. The micro
porous membrane [4] was installed into the modified pedestal. The identified air entry
value of the membrane was 250 kPa.
2.3. Procedure of unsaturated cyclic triaxial tests
Figure 1 shows the procedure of cyclic undrained triaxial tests. Cyclic undrained
triaxial testing consists of three cases for both materials in this study. Net normal
stresses of 25 kPa and 98 kPa were applied to all compacted, unsaturated specimens
prepared by Material A and those prepared by Material B, respectively. After
consolidation equilibrium, back pressure was applied from the bottom portion through
pedestal. It was assumed that the specimen reached to saturation even if degree of
saturation is not equal to 100 %. Subsequently, back pressure reduces to zero and
effective stress was remained in the specimen. After saturation, a matric suction of 20
kPa was applied to saturated specimen using the pressure membrane technique. After
the specimen matric suction of 20 kPa was reduced to the corresponding value of each
case, the cyclic triaxial test was performed on unsaturated specimen. Although initial
matric suction, pore air pressure and pore water pressure were different in each case as
shown in Table 2, the density of specimens was maintained constant before cyclic
loading. Both the excess pore water pressure and pore air pressure were measured
during the tests.
The cyclic shear was applied to the specimen under undrained air and water
conditions. Figure 2 shows the time history of the input strain. The input axial strains of
eleven cycles of the triangular waves were applied with multi step amplitudes whose
T. Matsumaru et al. / Laboratory Test and Numerical Study of Cyclic Behavior 681
respective amplitudes were 0.1, 0.2, 0.5, 1.0, 1.5, and 2.0. The frequency of the
sinusoidal wave was 0.001 Hz. Small loading rate was required to measure both pore
air and water pressure accurately.
Figure 1. Procedure of unsaturated cyclic triaxial tests.
Table 2. Test cases of undrained cyclic triaxial tests

(a) Material A
Pore air pressure Pore water pressure Suction
Case1 18kPa 5kPa 13kPa
Case3 1.5kPa 0kPa 1.5kPa
(b) Material B
Pore air pressure Pore water pressure Suction
Case3 0.3kPa 0kPa 0.3kPa

3
Axial strain [%]
2 1.5% 2%
1%
1 0.1% 0.2% 0.5%
0

-1
-2
-3
0 200 400 600 800 1000
Time [min.]
Figure 2. Time history of inputted axial strain.
3. Method of numerical simulation and analytical conditions
3.1. Basic equations
The overall density of the mixture can be expressed as

@ @ s @ w @ a 1
n@ sR n>s w @ wR 1
s w @ aR ? (1)
s w a
where , and are the partial densities of soil skeleton, pore water and air
respectively. sR, wR and aR are the real densities of each phase, n is the porosity and
sw is the degree of water saturation.
The skeleton stress which means the effective stress of unsaturated soil is expressed
as
s w a
p a I s w p a
p w I
(2)
where s, w and a are the partial stress of soil skeleton, pore water and air. ' is the
skeleton stress tensor [5] defined as positive in extension. pw is the pore water pressure
and pa is the pore air pressure. These pressures are defined as positive in compression.
The SWCC (soil water characteristic curve) is assumed as a logistic function [6],
derived by

s w s sw
s rw s ew s rw >
s ew 1 exp a lg p c blg ?

clg
(3)
w
where ss is the saturated (maximum) degree of saturation, is the residualsrw
(minimum) degree of saturation and sew is the effective water saturation. The
relationship between sew and suction pc is assumed as a logistic function with the
material parameters alg , blg and clg. The logistic SWCC is a continuous function at pc =
0; therefore the convergence in the iterative numerical scheme can be achieved.
3.2. Constitutive model of unsaturated soil
A simplified constitutive model for saturated sandy soil is used for unsaturated soil
with using a skeleton stress in place of an effective stress of saturated soil [7].
Assuming that plastic deformation occurs only when the deviatoric stress ratio changes,
the yield function is assumed as
3
f

k 0 (4)
2
where is the stress ratio and k is the material parameter which defines the elastic
region. The kinematic hardening parameter (back stress) , and its nonlinear evolution
rule [8] is assumed as
2
a be p
dp dp e p (5)
3
where a and b are the material parameters, and e p is the plastic deviatoric strain rate
tensor.
Using the following equation, the hardening parameter is reduced according to
the value of the equivalent plastic strain dp n which is reset when the direction of
loading changes, assumed as
a0
a1
a a0
(6)
1 a 0
a1 exp
C f dp ( n )
where a0 is the initial value of the material parameter a, a1 is the lower limit value of a
and Cf is the parameter which controls the amount of the reduction of a.
In order to describe plastic strain rate more precisely, the non-associated flow rule
[9] was adopted, and the following equations were obtained.
Ag Ag
e p 7 vp D7 (7)
As Ap
where 7 is the hardening coefficient and D is the coefficient of dilatancy.
In order to change the volumetric strain rate depending on the suction pc of
unsaturated soil, the relationship between the coefficient of dilatancy and the suction pc
was assumed as

D D1 D0
D1 exp
p c / pref
c
p c
C0 (8)
D D0 p c
B0
where D0 is the initial value of the coefficient of dilatancy D, D1 is the lower limit
value of D and the pcref is the parameter which controls the amount of the reduction of
D. The equation (8) means that the plastic volumetric strain tends not to occur if the
suction pc is large.
With the non-associated flow rule, the plastic potential function is assumed as
3 p
g
M m ln (9)
2 pa
where Mm is the critical state stress ratio, and pa is p when
0 .
Finally the elastic bulk modules are assumed as
K e
K * p G e
G* p (10)
e
where K is the elastic bulk modulus, G is the elastic shear modulus, K * and G *
e
are the dimensionless elastic modules, respectively.
3.3. Governing equations for simulation of triaxial test
A right-handed coordinate system is adopted and the vertical axial direction of triaxial
specimen is adopted as the z-direction. Taking material time derivative of the equation
(2) under the boundary condition that the lateral total stresses are constant, and under
the undrained conditions for pore water and air [10], the governing equation is obtained
as
P ns w K w
nc nc 2s w 0M J p w G J
s wz G
N KD D D D
N nc ns w K w
nc 2s a 0K D p a D D
s az D (11)

F I F
N s w
cp w cp a s w cp w
cp a
D11
D12 0K I D

D 13 z D
N w KD D
x
ON s
cp cp s cp
cp
D31
D32 1LK D H8 z DE DH D33z D
w a w w a
E
where Kw is the bulk modulus of the pore water, c is the specific water capacity, Dij is
the tangential module. The unknown variables, namely the lateral strain rate x , the
vertical total stress rate 8 z , the pore water pressure rate p w and the pore air pressure
rate p a can be calculated from the input axial strain rate z .
3.4. Material parameters for simulations
Table 3 shows the calibrated material parameters of the constitutive model. For the
constitutive model, the bulk modulus, shear modulus, and critical state stress parameter
were determined from the monotonic and cyclic triaxial tests for saturated soils. The
other parameters were determined by try and error.
Table 4 shows the calibrated material parameters of the SWCC. The LG model
parameters were determined so as to reproduce the wetting curve for large suction and
drying curve for small suction.
Table 3. Parameters for elasto-plastic constitutive model

Material A Material B
Bulk modulus and shear modulus (K*, G*) 1012.0, 467.0 1349.0, 622.0
Nonlinear hardening parameter (a0, a1, Cf) 1300, 30.0, 5500.0 1500, 30, 5000.0
Nonlinear hardening parameter (b) 1.187 1.331
Critical state stress ratio (Mm) 1.187 1.331
Coefficient of dilatancy (D0, D1, pcref) 0.25, 0.05, 0.6kPa 0.05, 0.05, --
Yield function parameter (k) 0.05 0.05
Table 4. Parameters for SWCC

Material A Material B
Maximum and minimum degree of saturation (ssw, srw) 0.985, 0.170 0.962, 0.265
LG model parameter (alg, blg, clg ) 2.2, -2.7, 0.16 10.0, -2.0, 0.02
4. Results and discussions of triaxial tests and its numerical simulations
4.1. Material A
Figure shows the experimental results of Material A. These results contain time
histories of pore water pressure and stress path for mean skeleton stress for each case.
The pore water pressure increased during cyclic undrained shear. As shown in the
graphs of the stress path, the mean skeleton stress attained zero in Case 3, which means
that the specimen was liquefied completely. On the other hand, the graphs of Case 1
and Case 2 showed different results, which mean that the strength and stiffness were
maintained during the cyclic shear.

50 50
[kPa]
[kPa][kPa]
c
Case1(p =13kPa)
[kPa]
c
40 Case2(p =5kPa) 40
skeleton stress
c
pressure
Case3(p =1.5kPa)
30 30
Mean

20 20
Pore water
10 10
0 0
0 200 400 600 800 1000 0 200 400 600 800 1000

Time[min.]
[sec.] Time[min.]
[sec.]
Figure . Experimental results of Material A
Figure shows the simulated results in which the constant dilatancy parameter
D=0.25, not depending on the value of suction, was used. The simulations of Case 1
and 2 overestimated the increase of pore water pressure and liquefaction occurred. In
Case 3, the increase of the pore water pressure is earlier than the experiments. The
simulations could not explain the experimental results with constant dilatancy
parameter.

50 50
[kPa]
[kPa][kPa]
Case1
[kPa]
40 Case2 40
skeleton stress
pressure
30 Case3 30
Mean

20 20
Pore water
10 10
0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Time[min.]
[sec.] Time[min.]
[sec.]
Figure . Simulated results of Material A with constant dilatancy parameter
Figure shows the simulated results with the changeable dilatancy parameter
depending on the value of suction as shown in Table 3. These results show that the
increase of pore water pressure and the decrease of mean skeleton stress became more
inactive than those by the simulations with constant dilatancy parameters. The results
show good agreement with the experimental results for all the cases.

50 50
[kPa] [kPa]
[kPa]
c
Case1 =13kPa)
Case1(p
[kPa]
40 Case2 c=5kPa)
Case2(p 40
skeleton stress
pressure
Case3 c=1.5kPa)
Case3(p
30 30
Mean

20 20
Pore water
10 10
0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Time[min.]
[sec.] Time
[sec.]
[min.]
Figure . Simulated results of Material A with changeable dilatancy parameter
4.2. Material B
Figure shows the time history of the pore water pressure of the experiments and the
simulations using Material B. In the same way as the tests using Material A,
liquefaction would occur in Case 3, though the mean skeleton stress was maintained
due to the small increment of pore water pressure in Case 1 and 2. The simulated
results could explain the tendency of experiments for all cases. For this material, the
constant dilatancy parameters were used as shown in Table 2. The density of specimen
of Material B was larger than that of Material A, so the cyclic behavior would not be
affected by the value of suction for Material B. The conventional constitutive model
would be suitable for describing the behavior of unsaturated soil under cyclic loading
with any value of suction, depending on the value of the dilatancy parameters.

100 100
c
Case3(p =0.3kPa)
Pore water pressure [kPa]
Pore water pressure [kPa]
80 80
c
60 Case3(p =0.3kPa) 60
40 40
Case1 Case1
20 c 20 c
(p =14kPa) (p =14kPa)
c c
Case2(p =4kPa) Case2(p =4kPa)
0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Time [min.] Time [min.]
Figure . Time histories of pore water pressure of Material B (left: experiment, right: simulation)
5. Conclusions
In this paper, a series of the unsaturated cyclic loading tests were conducted with
different values of the initial suctions. Furthermore, their numerical simulations were
conducted by using the conventional constitutive model. The following results were
obtained.
1. In the triaxial tests, the saturated specimens were dried and rewetted before cyclic
loading, and all the specimens kept the constant density for different initial
suctions. The test results showed that the mean skeleton stress was reduced with
the increase of the pore water in the case of small initial suction. The specimens
would be liquefied by the cyclic loading. On the other hand, the mean skeleton
stress was not reduced with a certain value of reduction in the case of the large
initial suction.
2. The unsaturated cyclic triaxial test results were used for verification of the
conventional constitutive model. Constitutive relationships described that the
dilatancy parameter was changeable for the value of the suction. The proposed
constitutive model reproduced unsaturated soil cyclic behaviors, and the model
was validated to be suitable for understanding both the liquefiable and non-
liquefiable sand in unsaturated conditions.
References
[1] H. Morishima, K. Saruya and F. Aizawa, Damage and reconstruction of the old railway at the section of
structures using soil. Foundation Engineering and Equipment, 33(10) (2005), 7883 (in Japanese).
[2] T. Matsumaru, K. Kojima, M. Tateyama, K. Watanabe and H. Watanabe, Calculation method for
residual displacement during earthquake for embankment affected by seepage water, Proceedings of
17th International Conference on Soil Mechanics and Geotechnical Engineering (2009), 1706-1709.
[3] T. Matsumaru, M. Suga and R.Uzuoka, Shaking table test of embankment on inclined ground affected by
rainfall, Proceedings of International Symposium on Earthquake-induced Landslides (2012), 437-444.
[4] T. Nishimura, J. Koseki, D.G. Fredlund and H. Rahardjo, Micro-porous membrane technology for
measurement of soil-water characteristic curve, Geotechnical Testing Journal 35(1) (2012), 201208.
[5] D. Gallipoli, A. Gens, R. Sharma and J. Vaunat, An elasto-plastic model for unsaturated soil
incorporating the effects of suction and degree of saturation on mechanical behavior, Geotechnique 53(1)
(2003), 123-135.
[6] T. Sugii, K. Yamada and T. Kondou, Relationship between soil-water characteristic curve and void ratio,
Proc.of3rd International. Conference on Unsaturated Soils 1 (2002), 209-214.
[7] T. Matsumaru and R. Uzuoka, Three-phase coupled analysis of unsaturated embankment subjected to
rainfall infiltration and seismic motion, Unsaturated Soils: Research & Applications (2014), 597-604.
[8] P.J. Armstrong and C.O. Frederick, A mathematical representation of the multiaxial Bauschinger effect,
C.E.G.B. Report RD/B/N731 (1966).
[9] F. Oka, A. Yashima, A. Tateishi, Y. Taguchi and S. Yamashita, A cyclic elasto-plastic constitutive
model for sand considering a plastic-strain dependence of the shear modulus, Geotechnique 49(5) (1999),
661-680.
[10] R. Uzuoka, T. Unno, N. Sento and M. Kazama, Effect of pore air pressure on cyclic behavior of
unsaturated sandy soil. Unsaturated Soils: Research & Applications (2014), 783-789.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-687
Effects of entrapped air bubbles on the

mechanical behaviour of quasi-saturated
soils
B.T. Laia , A. Fabbria, H. Wonga*, D. Branquea
a
Universit de Lyon/ENTPE/LGCB-LTDS (UMR CNRS 5513) 3 rue Maurice Audin,
69120 Vaulx en Velin, France.
*
Corresponding author
Abstract. This paper presents a model on the hydromechanical behavior of quasi-

saturated soils containing entrapped air bubbles, constituting a discontinuous
gaseous phase. These air bubbles reduce the compressibility of pore fluid hence
impact on the mechanical behavior, while the liquid-gas interface gives rise to
menisci and hence possibly suction effects on the soil skeleton. In parallel, air can
dissolve in water and then migrate through the porous space by water flow and by
diffusion. In practical situations, different parts of a soil mass can be at different
degrees of saturation; transition through different saturation regimes relative both to
space and time is possibly. The presentation of a theoretical model recently
developed by the authors, capable of simulating such continuous transition relative
to space and time, is the objective of this paper. A limited numerical study will also
be presented to show the consistency of this model.
Keywords. Phase transformation, air dissolution, Henrys Law, coupled

hydromechanical behavior, percolation suction
1. Introduction
The presence of a discontinuous air phase in quasi-saturated soils has been known
to impact significantly on the hydromechanical behaviour of geotechnical structures built
on or inside such soils. Despite a few past studies on effects of entrapped air ([8]; [13];
[2]), a theoretically consistent model capable of treating the simultaneous presence of
different saturation regimes including the above quasi-saturation regime remains to be
built. The central difficulty is the modelling of the kinematics of air molecules under
gaseous and dissolved form, which interacts with the hydromechanical behaviour of the
solid skeleton. This paper presents a theoretical model built upon a set of consistently
chosen assumptions which has shown to be able to produce results consistent with
physical intuition and available experimental data.
2. Physical mechanism involved in a quasi-saturated soil
An unsaturated soil is conceptually a tri-phasic media, comprised of a solid skeleton

(S), a porous network saturated in part by a liquid phase (L) and in part by a gas phase
688 B.T. Lai et al. / Effects of Entrapped Air Bubbles on the Mechanical Behaviour
(G). Relative to a representative elementary volume (REV) of the solid skeleton, of initial
and actual volumestuF and tv, the actual liquid and gas volumes are denoted by tuw
and tux . The porosity y and the saturation degree Sr retain their usual meanings: y z
*{ Bm{} *{|
| ;~w z . Here, the liquid phase is only composed of water and dissolved
*{L *{| B*{
air, represented symbolically by z , where the first (small) letter represents
the species and the second (capital) letter represents the phase containing the species.
Similarly, two species are present in the gas phase: dry air and water vapor. To simplify,
water vapor is supposed absent, hence: z . Clay particles constitute the first
component of the solid phase~. The entrapped air sharing the same kinematics as the
solid skeleton constitutes the second component: ~ z ~ ~. The capillary suction,
defined as the difference between gas pressure x and liquid pressure w is related to the
radius of the interface wx hence to the pore size via Laplaces law:
z x w z wx z ~
(1)
with is the interfacial tension and ~ a monotonic decreasing function defining

water retention behavior. The transformation between dissolved air and gaseous air is
assumed instantaneous, so that thermodynamic equilibrium always holds, governed by
Henrys law:
w (2)
x z
x
where x can be the gas pressure within the trapped air pocket or the continuous air phase,
z ]w w is the mass of dissolved air per unit of liquid water mass and the
exponent means at thermodynamic equilibrium, w and x are respectively the
molar mass of water and air, and is Henrys constant. At 20C, and assuming that air
is composed of 20 % O2 and 80% N2, . Note that when gaseous air is
not at proximity, the mass concentration of dissolved air can be less than that given by
thermodynamic equilibrium. This leads us to the following conditions:
~w ~w z (3)
We also suppose that the air bubbles follow closely the shape of the porous space in order
to maximize its volume-to-surface ratio, thereby acquiring the highest stability. The
radius of curvature of the gas-liquid interface will therefore be conditioned by the pore
size and geometry. As discussed in [11] and [5], due to Laplaces law, the air pressure
increases with decreasing curvature. The smaller air bubbles will tend to dissolve and
supply their air content in favor of larger bubbles. Since the characteristic time of
diffusion at the pore scale is much smaller than other characteristic times at the structural
scale [1], the coalescence of small air pockets into bigger ones is assumed to be
instantaneous. The trapped air will therefore be located within the bigger pores and the
air pockets would take a shape that maximizes the radius of its interface with the
surrounding liquid phase. In order to be in line with the above observation, we assume
in this study a simplified geometry of the air pockets, as reported in Figure 2. Namely,
all the air pockets are of identical size and are trapped within the biggest pores of
radius z . In consequence, the bubbling/dissolution processes induces an
B.T. Lai et al. / Effects of Entrapped Air Bubbles on the Mechanical Behaviour 689
increase/decrease of the volume of the pocket, but the radius of the liquid-gas interface
remains equal to .
2.1. Definition of the three different domains of saturation
We will account for three different domains of saturation: the unsaturated domain
with continuous liquid and gas phases (US); the quasi-saturated domain where the gas
phase in the form of bubbles is discontinuous (QS); and the fully saturated domain where
air is only present in dissolved form (FS). To fix ideas consider the case of wetting
starting from the unsaturated state (US). In order to arrive at a consistent description of
a continuous transition across different saturation domains, the porous space occupied
by the gaseous phase will conceptually be divided into two sets that can freely exchange
within each other. The first, named displaceable air is the part that will be invaded by
water during the wetting process before the loss of connectivity of the gaseous phase;
this volume is noted aG. The remaining part, called non-displaceable air,
corresponds to locations where the air is trapped when the saturation attains the
threshold~ . The air within this volume is noted aS, the capital "S" being used to
emphasize the fact that the trapped air has the same kinematics as the solid matrix. Under
this notation, the saturation ratio can be expressed in the form:
~w z ~]x ~] (4)
where ~]x is the fraction of pore space presently occupied by the gaseous phase that
will be invaded by liquid phase when ~w z ~ , while ~] represents the part that will
remain to be occupied by the trapper air when ~w z ~ . A simplified description of the
saturation domains and the continuous transition assumed here is summarized in Figure
1. Note that the above description is far from unique; more complicated descriptions have
also been discussed elsewhere (see for example [7], [12], [14], [2], [10]). Consistently
with the simplified description adopted above and the range of saturation levels in the
QS-domain, it is assumed that the positive suction effects in this domain can be neglected.
In the US-domain, the gaseous phase is continuous and connected to outside; the
difference between the gas pressure (equal to atmospheric pressure here) and the liquid
pressure, defined as matric suction, is higher than the air-entry suction. The two sets of
gaseous air, aS and aG, are simultaneously present, physically connected to each
other and share a common pressure equal to the atmospheric pressure:
F w x z ] z ]x z F (5)
This implicitly implies that the flow of gas towards the external boundary surface is
instantaneous. Note that, by hypothesis, the degree of saturation corresponding to the air
set aS remains constant in this phase: ~]X z ~ . Suction and degree of saturation
are linked to each other via the relation proposed by [3] for saturation ratios below ~ :
X S
~w z ~ .
X
Figure 1. Transition between different domains of saturation during wetting.

In the QS-domain, suction is below the air-entry value and the gaseous phase is
discontinuous; air is only present in the form of bubbles. Being non-connected to the
outside, pressure in the air bubbles can now exceed the atmospheric pressure. Note that
the gas pressure is continuous at the transition US/QS, equal toF . In summary.
F w ~]x z ~] x z ] z w F (6)
The FS-domain is reached when all the trapped air is dissolved, the material becomes
fully saturated. The mathematical conditions is~] z .
3. Poromechanical modelling
3.1. Constitutive equations and momentum balance
Elastic behaviour for the solid skeleton is assumed, where the strains are entirely
governed by an effective stress z , with the equivalent pore pressure.
Consistently with classic results, is taken to be the average pore pressure in the US-
domain, and to be the liquid pressure in both the QS and FS domains, neglecting suction
effects due to air bubbles in the QS-domain. Following [4], the incremental relation
linking stress to strain variation writes:
(7)
t z t t

Where K and G are the bulk and shear moduli, the strain tensor and the volumetric
strain. The momentum balance reads:
z % (8)
with z y y~ w y ~ x the overall density of the porous medium.
3.2. Mass balance equations
There are three independent mass balance equations, on liquid water, gaseous air and
dissolved air. According to [5], the mass balance of liquid water writes:
w (9)
z w w
wF
with w the liquid density, w the liquid water mass per unit of overall initial volume,
w z F w the permeability coefficient, w the dynamic viscosity of pore liquid and
F the intrinsic and relative permeabilities. The latter is a function of the saturation
ratio, that varies from 1 when ~ z to 0 when~ z . The dependence of on the
degree of saturation is well-known and is taken into accounted here using the relation
z ~ from [10]. For the pore liquid, where pressure and density are linked by the
* *
elastic relation | z | . The mass fraction of dissolved air being much lower than that
| `|
of liquid water, we can writew z w y~w . Recalling that w wF , the above
equations combine to give:
~w y~w w y (10)
~w z w w w wF
w
The mass balance of dissolved air is due to three main contributions: (a) the advective
transport of dissolved air within the liquid phase; (b) the diffusion of the dissolved air
within the liquid phase; and (c) exchange with gaseous air via the dissolution/bubbling
process. After some detailed developments, we get:
y~w w ~w y
y~w y ~w (11)
w
x]w

z w w w
w
where is the effective diffusion coefficient of dissolved air, x]w the mass rate of
dissolution/bubbling of air within the liquid phase, the effective diffusion coefficient
of dissolved air. To derive the equation of mass balance for gaseous air, we assume that
the in-pore air phase follows the law of perfect gas:
y~]x ]x x y~] ] x (12)

]x z ] z

with the universal gas constant and the absolute temperature. Neglecting
evaporation and condensation phenomena, the general form of the mass conservation of
gaseous air is:
]x ] (13)
z
x]w

where stands for the mass in-flux due to the Darcy flow of gas through the porous
network. To advance further, we need to substitute the expressions of various quantities
into equations (10), (11) and (12) and simplify. After a lengthy manipulation, we obtain
the following three governing equations:
w ~] (14)
z w w

w x ~]
y~w y
w (15)
z w w
x ~] w
y z w w (16)
w
with z F . to are coefficients that depend on , w and ~] .
The coefficients and take up respectively the values of and in the
regimes~ ~ ~.
4. Numerical example
In order to check the consistency of the model and to illustrate its applicability, consider
a 1D configuration of a homogeneous soil sample of 2 cm thick in an oedometric cell.
Initially, the sample is in thermodynamic and mechanical equilibrium and is in an
unsaturated state (US), with an initial negative liquid pressure implying the presence of
suction effects. The sample is then subjected to a monotonic increase of liquid pressure
at both top and bottom boundary surfaces, from to in 24 hours:

w z w z @] z
@]
The total stress is kept constant equal to its initial value F given in table1. This means:
z z F z z
whereas radial displacement and radial flow are null. This simulation corresponds to a
progressive wetting of a sample. The parameters used for the simulation are summarized
in Table 1. Finally, we assume that the mass concentration of dissolved air in the
incoming liquid water is at any time equal to the concentration of dissolved air within
the pores of the sample which are directly in contact with the external surfaces:
z z z z
From figs 2 and 3, it can be seen that consistently with fig 1, ~] remains constant equal
to ~ while ~w increases. The transition US/QS takes place at around t=8h when all
displaceable air is invaded by liquid water and the liquid pressure rises above the critical
valueF . ~] decreases afterwards. Logically when air dissolution is accounted for,
~] decreases more rapidly, and full saturation can be attained. Otherwise the volume of
entrapped air remains positive. Fig 4 shows the variation with time of the mass contents
of air under 3 different forms: displaceable (]x ), non-displaceable (] ) gaseous air
and dissolved air (]w ), as well as the total mass content x to facilitate consistency
check. The quantity of gaseous air decreases quickly in the US-domain during which
gaseous air forms a continuous network (~]x ), so that air-outflow encounters little
resistance whereas ] stays constant as mentioned above. The QS-domain starts when
~]x drops to zero. When dissolution is accounted for (fig 4a), the quantity of dissolved
air (]w ) changes little in the US-domain, and then rises quickly in the QS-domain, until
all the isolated gaseous air is dissolved (] z ) and the FS-domain is encountered.
Figure 2 Variation of degree of saturation and degree Figure 3 Variation of degree of liquid saturation
of saturation of trapped air. vs liquid pressure.
(a) (b)
Figure 4 Variation of the air mass contents depending on time in both cases: (a) accounting for air dissolution
(b) not accounting for air dissolution.
5. Conclusions
An original model was developed to simulate couplings between physical and

hydromechanical phenomena in quasi-saturated soils, where the key issue is the
consistent description of the kinematics of air in its different forms. Developed based on
well-proved thermodynamic principles ([3], [6]), this model gives a consistent
description of the simultaneous presence of different saturation states, as well as their
evolution relative to space and time. In particular, air molecules can be present in
different forms: mobile gaseous form, entrapped gaseous air, or dissolved in liquid water,
and can transform from one form to another. The numerical examples illustrate the
consistency of the model. On-going work consists of introducing a plastic constitutive
law for the solid skeleton so that the model can be applied to realistic cases.
Table 1. Parameters used in the simulation

Symbol Description Value (unit)
Henrys constant for air 7326 [MPa]
ideal gas constant 8.314 [J/(mol.K)]
w Molar mass of water 18 [g/mol]
x Molar mass of air 29 [g/mol]
F Atmospheric air pressure 100 [kPa]
Absolute temperature 293 [K]
w Density of water 1000 [kg/m3]
w Bulk modulus of water 1970 [MPa]
Dynamic viscosity of water 10-3 [kg/(m.s)]
w Permeability 8.10-9 [m/s]
Effective diffusivity 1.10-9 [m/s]
F Initial void ratio 1.97 [-]
~ Degree of saturation at air-entry 0.915 [-]
F Initial air-entry suction 110 [kPa]
b Coefficient of retention curve 0.4 [-]
wF Initial liquid pressure -250 [kPa]
F Initial total stress 1000 [kPa]
~wF Initial degree of saturation 0.566
F Initial mass fraction of dissolved air 22.10-6
References
[1] L. S. Barden, Engineering Behavior and Structure of Compacted Clay, Journal of the Soil Mechanics
and Foundations Division, ASCE 96(4) (1970), 1171-1200.
[2] L. Boutonnier, (2007), Comportement hydromcanique des sols fins proches de saturation cas des
ouvrages en terre: coefficient B, dformations instantanes et diffres, retrait/gonflement. Grenoble:
Thse l'Institut National Polytechnique de Grenoble.
[3] R. Brooks,A. Corey, (1964), Hydraulic properties of Porous Media. Colorado State University.
[4] O. Coussy, (2004), Poromechanics. John Wiley & Sons.
[5] O. Coussy, Deformation and stress from in-pore drying induced crystallization of salt, Journal of the
Mechanics and Physics of Solids 54 (2006), 1517-1547.
[6] O. Coussy (2010). Mechanics and Physics of Porous Solids. United Kingdom: John Wiley & Sons, Ltd.
[7] A. Denoth, The pendular-funicular liquid transition and snow metamorphism, Journal of Glaciology
28(99) (1980), 357-364.
[8] B. A. Faybishenko (1984). Influence of entrapped on soils permeability: theory and experiment. Vodn.
Resur., 4, 48-60.
[9] P.L. Martin,E.E. Alonso and J. Alcoverro, Catsius clay. BM 3.2: Febex mock-up test, Benchmark definition
(1998).
[10] J. Monnet and L. Boutonnier L, Calibration of an unsaturated airwatersoil model. Archives of civil and
mechanical enggineering (2012), 493-499.
[11] F. Osselin, A. Fabbri, T. Fen-Chong, P. Dangla, J-M. Pereira, A. Lassin, Stress from NaCl crystallisation
by carbon dioxide injection in aquifers, Environmental Geotechnics (2013)
[12] S. Salage, Influence de la temprature sur le comportement hydrique de milieux poreux partiellement
saturs. Montpellier: XXIVemes Rencontres Universitaires de Gnie Civil (2006).
[13] Sakaguchi, A., T. Nishimura, and M. Kato. (2005). The effect of entrapped air on the quasi-saturated soil
hydraulic conductivity and comparison with the unsaturated hydraulic conductivity. Vadose Zone Journal,
4, 139-144.
[14] S. K. Vanapalli, D.G. Fredlund, and D.E. Pufahl, Influence of soil structure and stress history on the soil-
water characteristics of a compacted till, Geotechnique 49(2) (1999), 143-159.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-695
Discrete element investigation of rate

effects on the asymptotic behaviour of
granular materials
David MAN a,1 and Jan JERMANa
a
Faculty of Science, Charles University in Prague, Czech Republic
Abstract. The effect of loading rate on the asymptotic behaviour of a granular

material with permanent particles is investigated using discrete element method. The
asymptotic states could uniquely be represented in terms of dimensionless inertial
number I. For low inertial numbers (I lower than 0.01 to 0.001), the asymptotic stress
ratio is independent of I (and thus of loading rate and of mean stress). This agrees
with observation on soils, showing critical state friction angle independent of mean
stress. For low inertial numbers, the critical state line as well as the isotropic normal
compression line have similar slope in the plane of void ratio vs. mean stress to the
isotropic rebound line. This agrees with conclusions of other scientists, who suggest
that particle crushing needs to be involved to represent more realistically the
volumetric asymptotic response of granular material.
Keywords. Asymptotic states, rate effects, discrete element method, granular flow
1. Introduction
In this paper, we present a continuation of the work by Man [1], who studied asymptotic
properties of granular materials using discrete element method (software Yade).
Asymptotic states can be defined as states reached after sufficiently long proportional
loading (loading with constant direction of stretching rate). They have been known from
both experiments on granular materials and from numerical simulations. In fact,
asymptotic states are incorporated in most of the soil constitutive models based on the
critical state soil mechanics, including both the elasto-plastic and hypoplastic models
(explicit incorporation of asymptotic states into hypoplasticity is described in [2],[3]).
To graphically illustrate the asymptotic states, we first define two stress and strain
obliquity measures and (Gudehus and Man [4]), see Fig. 1. The asymptotic state
concept is then introduced in Fig. 2. Each of the strain rate directions is associated
with a single stress obliquity direction (Fig. 2a). In addition, each of the strain rate
directions is also associated with a unique asymptotic trace in the mean stress p vs.
void ratio e plane, denoted as normal compression line (Fig. 2b). The two graphs can be
combined into a three-dimensional object denoted as asymptotic state boundary surface,
whose constant volume section is shown in Fig. 2c. Man [1] also defined asymptotic
states in extension, in addition to the asymptotic states in compression shown in Fig. 2.
The most important of all the asymptotic states, representing the state when the granular
1
696 D. Man and J. Jerman / Discrete Element Investigation of Rate Effects
material flows at constant volume (thus associated with ). It is traditionally

denoted as critical state in soil mechanics.
(a) (b)
Figure 1. Definition of angles and
(a) (b) (c)
Figure 2. Graphical representation of compression asymptotic states
The main conclusions of the work by Man [1] were as follows. First of all, he confirmed
existence of asymptotic states, both in compression and in extension. He noted that the
asymptotic states were predicted irrespectively of the fact the particles were elastic and
spherical (thus non-crushable). He also observed that the asymptotic stress ratio
(including the critical state friction angle) depended on the mean effective stress. This
conclusion, differing from the known soil behaviour, was attributed to the loading rate,
but no further investigation was done in this respect.
In the present paper, we continue the work by Man [1] by investigating of the rate
effects on the asymptotic response of granular materials.
2. Dicrete element model
The charateristics of the discrete element model were identical to the model from
[1]. It has been setup within dicrete element code Yade [5]. The particles were spherical
and permanent with granulometry of a real sand (particle sizes ranging from 0.2 mm to
1 mm). To eliminate the influence of the model boundaries, periodic boundary conditions
have been adopted, so that the modelled unit cell (as well as all its particles and all their
D. Man and J. Jerman / Discrete Element Investigation of Rate Effects 697
properties) is surrounded by identical cells shifted along the cell edges. A simple elastic-
frictional contact law has been adopted with parameters E = 500 MPa, = 0.3 and =
0.5 Rad (coefficient of friction contact = 0.546). The rolling resistance (rotational spring)
was not considered. The sample was generated using an algorithm ensuring that they
followed the prescribed particle size distribution, were randomly distributed, and were
initially not in contact. The prescribed particle density was s = 2650 kg/m3 and
acceleration due to gravity was zero. The sample has first been compressed under
isotropic conditions with constant value of the Euclidean norm of the Euler stretching
tensor ||D|| up to different isotropic stresses. As in this work we focused on constant
volume loading only, the sample was subsequently sheared (with the same value of ||D||
as used in compression) at a condition of . The influence of ||D|| was
investigated.
The analyses by Man [1] have been performed with specimen consisting of 150000
spherical particles. This high number of particles ensured accuracy of the obtained
response; it however also led to computationally demanding simulation, not feasible at
much lower stretching rates investigated in this work. We thus first investigated the
influence of the specimen size on the simulation results (Fig. 3; note that abbreviation
[GM12] indicates the results from Man [1] throughout this paper). We observed that
the response is not significantly affected by the specimen size for specimens with more
than approx. 15000 particles. In this work, a sample consisting of 16700 particles has
been used in the simulations.
(a) (b)
Figure 3. The influence of the specimen size on the simulation results ( ). Specimen with 16700
particles selected for rate-effects investigation. (a): rate 334 s-1 [GM12], (b) rate 33.4 s-1.
3. Rate-effects on the asymptotic response
3.1. Rate effects for a single consolidation stress
In the first set of analyses, the specimens have been consolidated isotropically up the
mean stress of 1000 kPa and sheared subsequently at up to the asymptotic state
(critical state) at various values of ||D||. Results are shown in Fig. 4, showing graphs for
the loading phase (Fig. 4a,c,d), and also for the isotropic compression stage
(Fig. 4a, b).
Figures 4c,d show that decreasing the loading rate by almost 4 orders of magnitude
decreases the asymptotic stress ratio and, concurrently, it also changes the stress path so
that the asymptotic states are with decreasing loading rate reached at progressively lower
mean stresses. Interestingly, the curves at low loading rate show intermittent peak in the
stress path (Fig. 4d), followed by an immediate mean stress decrease and subsequent
increase of the stress ratio up to the asymptotic state. This behaviour will be discussed
later in Sec. 4.
(a) (b)
(c) (d)
Figure 4. The influence of loading rate ||D|| on the asymptotic behavior for consolidation stress of 1000 kPa
Figure 4b shows that the loading rate affects not only the asymptotic stress ratio at
constant volume, but it also affects the isotropic loading phase of the test, such that the
isotropic normal compression lines are progressively lower in the effective stress vs. void
ratio plane. In fact, for low loading rates the normal compression lines have small slope,
such that there is not a significant difference between the compression and rebound line
(rebound response not shown in Fig. 4 for clarity, however). This is consistent with
findings by [6,7] and others, who suggested that to realistically model normal
compression of soil-like material, particle breakage must be considered. Finally, Fig. 4a
demonstrated that the final state of the constant volume shearing depends on shearing
rate not only in terms of stress ratio, but also in terms of the asymptotic void ratio.
3.2. Asymptotic states for different consolidation stresses
In the subsequent set of analyses, we selected five different loading rates (||D||=334s-1,
33.4s-1, 16.7s-1, 2.79s-1 and 0.33s-1) and studied the asymptotic response for different
consolidation stresses (Fig. 5). Consistently with Fig. 4d, decreasing the loading rate
increases a tendency to mean effective stress decrease in proportional loading. For fast
loading, the stress ratio increases continuously and the asymptotic states align evenly
along a well-defined asymptotic state line. For slower loading rates, however, different
behaviour is observed. The samples compressed to lower stresses evince a sharp stress
ratio decrease before reaching the asymptotic state. The asymptotic state is still well-
defined in the vs. mean stress plane, we can however distinguish two different data
clusters. One cluster corresponds to samples compressed isotropically to high stresses.
These samples reach the asymptotic state before the stress ratio decrease could have
occurred and the asymptotic state is associated with high mean stresses. The second
group of samples show a stress ratio decrease before they could have reached the
asymptotic state, and consequently the mean stress at the asymptotic state is remarkably
low.
(a) (b)
(c) (d)
Figure 5. The influence of loading rate ||D|| on the asymptotic behaviour for different consolidation stresses
The asymptotic states from Fig. 5 are re-plotted in Fig. 6 for better demonstration of
the loading rate effects. A decreased loading rate decreases the dependency of the
asymptotic stress ratio on mean stress (Fig. 6a). This leads us to hypothesis that for low
loading rates the asymptotic stress ratio would be independent of the mean effective
stress: consistently with the known behaviour of soils, where the critical state friction
angle is generally assumed to be independent of the stress level. More discussion on this
fact is given in Sec. 4. Figure 6b shows that not only the stress ratio, but also the void
ratio at the asymptotic state depends on the loading rate (decreasing loading rate implies
decreasing asymptotic void ratio). This has already been observed in Fig. 4a.
(a) (b)
Figure 6. Asymptotic states for different loading rates plotted in p vs. (a) and p vs. e (b) planes
4. Dimensionless analysis
Dimensionless quantities proved to be a powerful unifying reference for representation

of granular flows. It has been observed by various authors [8][9], that the coefficient of
friction * of a granular assembly can be uniquely represented in terms of so-called
inertial number I, which has been for general loading case defined by Jop et al. [10] as

(1)

The coefficient of friction * was found to be independent of I for values of inertial

number lower than approx. 0.01 [8][9]. For higher inertial numbers, the coefficient of
friction increased with increasing I.
Figure 7 shows values of the inertial number of the three loading rates investigated.
Two lines are plotted for each loading rate, the higher one corresponds to the largest
particle within the system (d=1 mm) and the lower one for the smallest particle within
the system (d=0.2 mm). The threshold value of I=0.01 is also plotted for indication. If
the asymptotic stress ratio was to depend on I with the threshold value of I=0.01, it would
be expected from Fig. 7 that the specimen loaded at ||D||=0.33s-1 would evince the
asymptotic stress ratio practically independent of mean effective stress. The other two
loading rates would lead to the asymptotic stress ratio dependent on p.
Figure 7. Inertial number I plotted for the three loading rates investigated. The higher line corresponds to the
largest particle size (d=1 mm), whereas the lower line to the smallest particle size (d=0.2 mm).
A graph showing directly the dependency of the asymptotic states on the inertial
number is in Fig. 8 (plotted for I calculated for the largest particle within the system, that
is for d=1 mm). The dependency of the asymptotic stress ratio on I is shown in Fig.
8a. All the available data align along a curve in the vs. I plane. Still, however, the
curve is not unique and some rate-dependency is observed. This indicates that the
dimensionless analysis using the inertial number is valuable not only for the
representation of friction coefficient in the simple shear test, but it is representing the
asymptotic behaviour in general. The limiting inertial number separating the quasi-static
and the dynamic regimes is between approximately I=0.001 and I=0.01. Figure 8c shows
the same data re-plotted in terms of coefficient of friction *.
(a) (b)
(c) (d)
Figure 8. Asymptotic state for represented in terms of the dimensionless inertial number I.

Similarly to the stress ratio, also the void ratio vs. I relationship depends on the
inertial number (Fig. 7b). For low inertial numbers the slope of the e vs. I curve is very
low, similar to the isotropic rebound line (not shown here). Note that the apparent steep
tails of the e vs. I dependencies at low inertial numbers are caused by non-negligible
particle overlaps, which is a side-effect of the discrete element method simulation
procedure (see [1] for the particle overlap quantification). The same data are re-plotted
in Fig. 7d in terms of volume fraction , defined as a ratio of volume of particles and the
sample volume.
It is to be pointed out that while the void ratio vs. I at asymptotic state relationship
is remarkably flat at low inertial numbers (the slope not being substantially different from
the isotropic rebound modulus, see [1]). This explains the tendency of the system to the
sharp decrease of the stress ratio before reaching the asymptotic state observed above
(Fig. 5). loading is, by definition, characterized by constant void ratio. The
samples consolidated to higher void ratios must show remarkable mean stress decrease
during asymptotic loading to reach the asymptotic state; contrary, the samples
consolidated to lower void ratios (caused mostly by particle overlap) would remain at
high mean stresses at the asymptotic state. As the void ratio vs. I curve is flat for low
inertial numbers, the gap between the two modes of behaviour is very narrow and finally
the data show two distinctive clusters (Fig. 5).
In the paper, we investigated the effect of loading rate on the asymptotic behaviour
of granular material using discrete element method. Among different asymptotic states,
we selected constant volume proportional loading for our investigation. We have shown
that the asymptotic stress ratio and void ratio can be represented in terms of
dimensionless inertial number I. For I lower than 0.001 to 0.01, quasi-static regime with
constant asymptotic stress ratio has been observed, which agrees with the known
behaviour of soils (critical state friction angle independent of mean stress). For low
inertial numbers, the normal compression lines are remarkably flat. This agrees with
conclusions of other scientists who suggest that particle crushing needs to be involved to
represent more realistically volumetric asymptotic response. For higher inertial numbers,
both the stress ratio and asymptotic void ratio depend on I.
Acknowledgement
Financial support by the research grant 15-05935S of the Czech Science Foundation
is greatly appreciated.
References
[1] Man D., Asymptotic behaviour of granular materials. Granular Matter 14 (2012), 759774.
[2] Man, D. Hypoplastic Cam-clay model. Gotechnique 62 (2012), 549553
[3] Man, D., Clay hypoplasticity with explicitly defined asymptotic states. Acta Geotechnica 8 (2013), 481
496.
[4] Gudehus, G. and Man, D., Graphical representation of constitutive equations. Gotechnique 59 (2009)
147151.
[5] milauer, V., Catalano, E., Chareyre, B., Dorofenko, S., Duriez, J., Gladky, A., Kozicki, J., Modenese, C.,
Scholts, L., Sibille, L., Strnsk, J., Thoeni, K.: Yade Documentation, 1st edn. The Yade Project (2010).
http://yade-dem.org/doc
[6] Cheng,Y.P.,Nakata,Y.,Bolton, M.D, Discrete element simulation of crushable soil. Gotechnique 53
(2003), 633641.
[7] McDowell, G.R., Bolton, M.D., On micromechanics of crushable aggregates. Gotechnique 48 (1998),
667679.
[8] da Cruz, F., Emam, S., Prochnow, M., Roux, J.N., Chevoir, F, Rheophysics of dense granular materials:
discrete simulation of plane shear flows. Phys. Rev. E 72 (2005), 021309.
[9] Hatano, T., Constitutive law of dense granular matter. J. Phys.: Conf. Ser. 258 (2010), 012006.
[10] Jop, P., Forterre, Y., Pouliquen, O., A constitutive law for dense granular flows. Nature 441 (2006), 727
730.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-703
Study of the effect of drainage conditions

on cone penetration with the Material Point
Method
Francesca CECCATOa,1, Lars BEUTHb and Paolo SIMONINIa
a
DICEA University of Padua, Italy
b
Deltares, Delft, The Netherlands
Abstract. Cone Penetration Testing (CPT) is a widely used in situ soil testing
technique which allows to estimate soil stratigraphy and various soil parameters.
Depending on the soils consolidation coefficient, undrained, partially drained or
drained conditions might occur. Partially drained conditions are commonly
encountered in soils such as silt and sand-clay mixtures. Correlations between CPT
measurements and soil properties are usually only valid for fully drained or fully
undrained conditions which may lead to inaccurate estimates of the properties of
silty soils. This paper aims at improving the understanding of the penetration process
in different drainage conditions through advanced numerical analyses. A two-phase
Material Point Method is applied to simulate the large soil deformations and the
generation and dissipation of excess pore pressures that occur during penetration.
The constitutive behavior of the soil is modelled with the Modified Cam Clay model.
The implemented MPM formulation is validated by comparing numerical results
with the results of centrifuge tests under different drainage conditions.
Keywords. CPT, MPM, site investigation, numerical modelling, partially

drained conditions, multiphase formulation.
1. Introduction
Standard Cone Penetration Testing (CPT) consists in pushing a measuring device with a
conical tip into the ground with a constant velocity of 2 cm/s. From the measurements of
tip resistance (qc) and sleeve friction (fs) various soil properties, such as the friction angle,
the undrained shear strength and the compression index, can be estimated; see for
example Ref. [1].
When performing CPT, different drainage conditions might be encountered
depending on the soils consolidation coefficient. In highly permeable soils, such as sand,
fully drained conditions occur: negligible excess pore pressures build up around the tip
and quickly dissipate. In nearly impermeable soils, such as clay, fully undrained
conditions occur: considerable excess pore pressures are generated that dissipate at a
much slower rate than the penetration velocity. In soils such as silt, partially drained
conditions are encountered: excess pore pressures dissipate to some extend in the vicinity
of the cone during its penetration.
1
704 F. Ceccato et al. / Study of the Effect of Drainage Conditions on Cone Penetration
If the pressure dissipation rate is relatively high compared to the penetration rate,
the soil in the vicinity of the advancing cone consolidates during penetration. Effective
stresses increase, resulting in larger shear strength and stiffness compared to undrained
conditions and subsequently higher tip resistance and sleeve friction [2].
Most of the empirical and theoretical correlations commonly used to estimate soil
properties are not valid for partially drained conditions. This may lead to an inaccurate
estimate of soil properties.
This paper investigates the effect of the drainage condition on the cone resistance in
soft soils through numerical analyses. For this purpose, a fully coupled two-phase
Material Point Method (MPM) featuring a contact formulation for modeling soil-
structure interaction is applied. The method is described in Section 2. The soil response
is simulated with the Modified Cam Clay (MCC) model [3].
To the authors knowledge, a numerical study of CPT which considers the three-
dimensional large deformations, the roughness of the penetrometer surface as well as
non-linear soil behaviour in a wide range of drainage conditions is a novelty. Indeed,
previous numerical studies [4, 5] could not consider all these important aspects.
In Section 3 the numerical simulation of cone penetration is presented, followed by
the results in Section 4. The paper ends with concluding remarks and suggestions for
future research (Sec. 5).
2. The two-phase Material Point Method
The MPM has been developed for large deformation problems in solid mechanics [6]
and was first applied to granular materials by Wickowski [7, 8] and Coetzee et al. [9].
It has been used successfully in the study of numerous geomechanical problems; amongst
others anchor pull-out [9], dam failure [10], landslides [11], cone penetration [12], and
pile installation [13].
With the MPM, arbitrary large deformations of a body are simulated by material
points (MP) moving through a finite element mesh. The MP carry all the information of
the continuum such as velocity, acceleration, stress, strain, material parameters as well
as external loads. The finite element mesh is used to solve the equations of motion for
each time step, but does not store any permanent information.
At the beginning of each time increment, information is mapped from the MP to the
nodes of the mesh. The governing equations are solved identically to the classical
Updated Lagrangian Finite Element methods. Strains, stresses and other state variables
Figure 1. (left) configuration at the beginning of a time step in which the red dots are the MP; (center)
incrementally deformed mesh; (right) reset mesh at the end of a time step
F. Ceccato et al. / Study of the Effect of Drainage Conditions on Cone Penetration 705
at the MP as well as the updated position of the MP are computed from the obtained
nodal accelerations. Once information is mapped from the nodes to the MP, the mesh is
usually reset to its original state, though it could also be adjusted or entirely redefined.
Figure 1 illustrates the computation steps of a time step with the MPM.
The finite element grid must extend across the entire region of space through which
the material is expected to move. However, only nodes attached to finite elements that
contain material points (active elements) are considered when setting up the equilibrium
equations.
An available 3D MPM with explicit time integration is used in this study. Despite
specifically developed for dynamic problems, it is also well suited for the analysis of the
considered quasi-static CPT problem as described below.
The generation and dissipation of excess pore pressure is simulated with a fully
coupled two-phase formulation [14]. The equilibrium equations are solved for the
accelerations of soil skeleton and water phase as primary unknown variables as presented
in [15].
The soil-cone interaction is modelled with a contact formulation specifically
developed for the MPM [16], which is based on the Coulombs friction law. The
advantage of this algorithm is that it detects the contact surface automatically and does
not require any special interface elements. It is proved to be efficient in modeling
interaction between solid bodies as well as shearing in granular materials [9, 17].
(a) (b)
Figure 2 Geometry and discretization of the CPT problem.

Table 1 Material parameters of numerical analyses

Parameter Symbol Value
Virgin compression index [-] QR 0.205
Recompression index [-] SR 0.04
Effective Poissons ratio [-] 6 0.25
Slope of CSL on p-q plane [-] M 0.92
Bulk modulus of the water [kPa] Kw 36600
Initial void ratio [-] e0 1.41
Saturated density [kg/m3] @sat 1700
3. The numerical model
Taking advantage of the rotational symmetry of the cone penetration problem, only a 20
slice is considered. The cone is slightly rounded in order to circumvent numerical
problems induced by a discontinuous edge at the base of the cone. Apart from this
modification, the dimensions of the penetrometer correspond to those of a standard
penetrometer: the apex angle is 60 and the cone diameter (D) is 0.036m.
The size and the refinement of the mesh have been determined through preliminary
calculations as a compromise between computational cost and accuracy. It extends 8D
in radial direction and initially 14D below the tip. It counts 13 221 tetrahedral elements.
105 634 material points are located in the initially active elements. Figure 2 shows the
geometry and discretization of the CPT problem.
At the lateral mesh surfaces displacements are constrained in normal direction while
the bottom of the mesh is fully fixed. For the water phase, the radial boundaries are
impermeable since they correspond to symmetry axes of the problem while the bottom
and the circumferential boundaries are permeable.
In order to maintain a refined mesh always around the cone, a special procedure,
called moving mesh, is adopted [18]. It consists in adjusting the part of the mesh adjacent
to the penetrometer (Fig. 2b) to the movement of the cone after each time step. This
ensures that the penetrometer surface coincides with element boundaries. The elements
of this zone keep the same shape throughout the computation, while the elements in the
compressed zone below the cone (Fig. 2b) reduce their vertical length. The discretization
of the compressed zone is defined in such a way that the elements keep a reasonable
aspect ratio throughout the analysis.
The material weight is neglected because the gradient of the vertical effective stress
is negligible compared to the stress level developed during the penetration. The initial
vertical and horizontal effective stresses are 50kPa and 34kPa respectively. A vertical
stress of 50kPa is applied at the top surface of the soil simulating an initial position of
the cone at about 5m depth. The clay is considered normally consolidated. A further
penetration by 10D is simulated.
The input parameters for the MCC model are summarized in Table 1; they are typical
of kaolin clay, which is a material often used in laboratory tests, thus allowing a
comparison with experimental evidence.
The variation of drainage conditions is usually achieved in experimental

investigations by varying the penetration velocity v. In this study the variation of
drainage is obtained by changing the permeability k, while keeping v = 0.02m/s.
Although the complete range of drainage conditions can be simulated with the two-
phase formulation, fully drained and fully undrained conditions are simulated in a
simplified way considering the soil as a one-phase material, thus reducing the
computational effort. In fully drained conditions the presence of the water is neglected.
In fully undrained conditions there is negligible relative movement between solid and
water; therefore the equilibrium of the soil-water mixture is considered rather than the
equilibrium of soil and water as separate phases. The excess pore pressure increment is
computed multiplying the volumetric strain by the bulk modulus of the water [19].
In this study, the soil-cone interface is characterized by a friction coefficient of 0.3.
Reasonable values of the friction coefficient for low plasticity clay in contact with steel
lie between 0.2 and 0.35 [20].
4. Results
The dependence upon the drainage conditions can be expressed through the
normalized penetration rate introduced by Finnie and Randolph [21]:

(1)

where v is the penetration velocity, D the cone diameter and cv the soils consolidation
coefficient, which can be estimated according to [22] with:

(2)

where k is Darcys permeability, 8v0 is the initial vertical effective stress, 7w=10kN/m3
is the water unit weight.
Figure 3 shows how the tip stress increases with the vertical displacement of the
cone for different drainage conditions. The steady state tip stress, which corresponds to
the tip resistance qc, is reached after a displacement of about 6D. As expected the tip
resistance increases with decreasing V, i.e. moving from undrained to drained conditions.
In case of V = 1.2 the tip resistance is 7% lower than for drained conditions, and in case
of V = 12 the tip resistance is 5% higher than for undrained conditions.
Figure 3 Tip stress over normalized cone displacement for different drainage conditions.
The excess pore pressure distributions are shown in Figure 4. Approximately

undrained behavior is observed for V = 12 at which the pore pressure is about 150kPa.
On the contrary, for V = 1.2 the behavior is nearly drained and the pressure is about
30kPa.
Figure 5 shows the deviatoric stress distributions for four of the performed analyses.
The shear strength and the stiffness of the soil increase with decreasing the normalized
penetration rate because of pore pressure dissipation and subsequent increase of effective
stress. The lowest deviatoric stress around the cone is observed in undrained conditions
(Fig. 5a), which corresponds to the undrained shear strength of the soil. In partially
drained conditions, it is about 45kPa in case of V=12 (Fig. 5b) and 90kPa in case of V=1.2
(Fig. 5c). The highest deviatoric stress, about 120kPa, is encountered in drained
conditions (Fig. 5d).
Figure 4 Excess pore pressure in case of undrained conditions (left) and partially drained conditions for
V=12 (center) and V=1.2 (right).
(a) (b)
(c) (d)
Figure 5 Deviatoric stress in case of undrained conditions (a), partially drained conditions with V=12
(b), partially drained conditions with V=1.2 (c) and drained conditions (d)
Figure 6 Normalized resistance as function of normalized velocity
The net resistance for a generic value of the normalized velocity V, qnet = qc - 8v0,
can be normalized by dividing it by the undrained net tip resistance, qref = qc,und - 8v0
where qc,und is the undrained net tip resistance. Figure 6 shows the normalized resistance
plotted over the normalized velocity and compares the numerical results with
experimental data on kaolin [22-24]. Numerical results agree well with experimental data;
differences can be attributed to uncertainties in the estimation of the input parameters for
the soil model. The normalized resistance is constant and equal to 2.1 for drained
conditions (V<0.1). Undrained conditions are observed for V>200.
5. Conclusions
This paper shows the capability of the two-phase MPM using a contact formulation
and the MCC model to simulate the complex problem of cone penetration under different
drainage conditions. The effect of drainage conditions has been studied through the
variation of the normalized penetration rate.
The cone resistance increases with decreasing normalized penetration velocity
because the soil near the advancing cone consolidates, thereby developing larger shear
strength and stiffness. Indeed, as the penetration velocity decreases, the deviatoric stress
increases and the pore pressure decreases. Numerical results are in good agreement with
experimental evidence.
To further improve the understanding of CPT, the MPM will be used to analyze the
effect of cone roughness, material parameters, and initial stress state on the cone tip
resistance.
References
[1] T. Lunne, P. Robertson, and J. Powell, Cone penetration testing in geotechnical practice. EF
Spon/Routledge Publ., New York, 1997
[2] K. M. Kim, M. Prezzi, R. Salgado, and W. Lee. Effect of penetration rate on cone penetration resistance in
saturated clayey soils. Journal of geotechnical and Geoenvironmental engineering 134.8 (2008), 1142-
1153.
[3] Schofield, A. and P. Wroth. Critical state soil mechanics. McGraw-Hill London, 1968.
[4] Silva, M. F., D. J. White, and M. D. Bolton. An analytical study of the effect of penetration rate on
piezocone tests in clay. International journal for numerical and analytical methods in geomechanics 30(6)
(2006), 501527.
[5] J. Yi, M. Randolph, S. Goh, and F. Lee. A numerical study of cone penetration in fine-grained soils
allowing for consolidation effects. Geotechnique 62(8) (2012), 707719.
[6] Sulsky, D., Z. Chen, and H. Schreyer, A particle method for hystory-dependent materials. Computer
Methods in Applied Mechanics and Engineering 118 (1-2) (1994), 179196.
[7] Z. Wieckowski, S.-K. Youn, and J.-h. Yeon . A particle -in-cell solution to the silo discharging problem.
International Journal for Numerical Methods in Engineering 45 (1999), 1203-1225.
[8] Z. Wieckowski, The material point method in large strain engineering problems. Computer Methods in
Applied Mechanics and Engineering 193 (39-41) (2008), 44174438.
[9] C. J. Coetzee, The material point method, Technical Report November, University of Stellenbosh,
Stellenbosh, South Africa, 2005.
[10] E. Alonso, F. Zabala.,Progressive failure of Aznalcollar dam using the material point method.
Geotechnique 61(9) (2011), 795808.
[11] S. Andersen and L. Andersen, Modelling of landslides with the material-point method. Computational
Geosciences 14 (1) (2010), 137-147.
[12] L. Beuth and P. A. Vermeer, Large deformation analysis of cone penetration testing in undrained clay.
Installation Effects in Geotechnical Engineering, Hicks et al. (eds), Taylor & Francis Group, London,
2013
[13] P. Nuygen, A. F. van Tol, A. S. K. Elkadi, and A. Rohe, Modelling of pile installation using the material
point method (MPM). Numerical Methods in geotechnical Engineering, 271 (2014).
[14] Van Esch, J., D. Stolle and I. Jassim. Finite element method for coupled dynamic flow-deformation
simulation. In 2nd International Symposium on Computational Gaomechanics (COMGEO II), Cavat-
Dubrovnik, Croatia (2011).
[15] I. K. J. Al-Kafaji, Formulation of a Dynamic Material Point Method (MPM) for Geomechanical
Problems, Ph. D. thesis, University of Struttgart, Germany, 2013
[16] S. Bardenhagen, J. E. Guilkey, K. Roessig, J. Brackbill, W. Witzel, & J. Foster, Improved contact
algorithm for the material point method and application to stress propagation in granular material.
Computer Modeling in Engineering and Sciences 2 (2001), 509522.
[17] H. W. Zhang, K. P. Wang, and Z. Chen. Material point method for dynamic analysis of saturated porous
media under external contact/impact of solid bodies. Computer Methods in Applied Mechanics and
Engineering 198.17 (2009): 1456-1472.
[18] L. Beuth, Formulation and Application of a Quasi-Static Material Point Method. Ph. D. thesis, University
of Struttgart, 2013
[19] P. A. Vermeer, PLAXIS 2D Reference Manual Version 5, Balkema, Rotterdam / Brookfield, 1993
[20] Lemos, L. J. L., and P. R. Vaughan. Clayinterface shear resistance. Geotechnique 50.1 (2000): 55-64.
[21] I. M. S. Finnie and M. F. Randolph, Punch-through and liquefaction induced failure of shallow
foundations on calcareous sediments. In Int. Conf. on Behavior of Offshore Structures, (1994) Volume
187, Zielona Gora, Poland, pp. 217-230.
[22] J. Schneider, B. M. Lehane, and F. Schnaid, Velocity effects on piezocone measurements in normally
and over consolidated clays. International Journal of Physical Modelling in Geotechnics 7(2), (2007).
2334.
[23] M. Randolph and S. Hope, Effect of cone velocity on cone resistance and excess pore pressures. In Proc.,
Int. Symp. on Engineering Practice and Performance of Soft Deposits, (2004) pp. 147-152. Yodagawa
Kogisha Co., Ltd.
[24] Mahmoodzadeh, Hamed, and Mark F. Randolph. Penetrometer Testing: Effect of Partial Consolidation
on Subsequent Dissipation Response. Journal of Geotechnical and Geoenvironmental Engineering 140.6
(2014).
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-711
Deformation of Structured Soil with

Cementation
Suksun HORPIBULSUK a,1 and Martin D LIU b
a
School of Civil Engineering, Suranaree University of Technology, Thailand
b
Faculty of Engineering and Information Sciences, The University of Wollongong,
Australia
Abstract. In this paper, a study on the deformation of structured soil with

cementation is made. The Structured Cam Clay model with cementation effect is
introduced. The Structured Cam Clay model was generated for the cementation
effect via the concept of the operative mean effective stress in which the
cementation effect is represented as an isotropic increment of the effective stress.
The removal of cementation structure is also simplified and described in the model.
The Structured Cam Clay model thus formed has the capacity to describe the
behaviour of soils with a wide range of structures as reconstituted state, natural
state, and cemented state. Model simulations are made and validated against
experimental data. Discussions on modeling soil behaviour with various
cementation structures are also presented.
Keywords. Structure Cam Clay model, deformation, cemented soil
1. Introduction
Soft soil is frequently encountered in geotechnical engineering practice such as in

coastal cities where the space for engineering work is constrained and as in
infrastructure constructions. The soil normally possesses very low strength and
stiffness, and thus constitutes a great challenge to geotechnical engineers. Soils
improvement techniques are generally required to increase both the strength and
deformation stiffness of the soils to support the engineering structure. Cement
stabilisation is now increasingly popular in geotechnical engineering because of its
relatively low cost and high efficiency. During the past five decades, a large amount of
laboratory and site investigations have been performed in order to understand the
mechanical properties of the cemented soft soil and to characterise its behaviour [1-5].
There are also some important developments on modelling the behaviour of structured
and cemented soil [6-9]. However, these constitutive models are generally complicated
and more work is still needed in order to provide models suitable for numerical
analyses for practical problems.
In this paper, a study on modelling the behaviour of structured clays with strong
cementation is made. The Structured Cam Clay model with cementation effect is
introduced. The Structured Cam Clay model proposed by Liu and Carter [10] and
Carter and Liu [11] was generated for the cementation effect, based on the concept of
1
712 S. Horpibulsuk and M.D. Liu / Deformation of Structured Soil with Cementation
the operative mean effective stress in which the cementation effect is represented as
an isotropic increment of the effective stress. Then, the model is used to simulate the
behaviour of soil with various structures including cement and fibre reinforced soil.
2. Structured Cam Clay Model with Cementation Effect
The Structured Cam Clay model for cemented clay is formulated based on the
theoretical framework of the Structured Cam Clay Model proposed by Liu and Carter
[10-11]. A brief introduction of the model is introduced here and details of the model
can be found in papers by Horpibulsuk et al. [12] and Horpibulsuk and Liu [13].
In the model, the behaviour of the parent clay without cement in reconstituted state
is used as reference state to measure the influence of cementation structure.
Consequently, the influence of soil structure including that of cementation effect is
measured as the difference in the mechanical response between a cemented soil and the
parent soil without cement in reconstituted states.
2.1. Operative Effective Mean Stress for Cemented Clay
Based on an analysis of experimental data on the yielding, and the yielding surface as
well as the peak strength and tensile strength of cemented soils, an operative mean
stress parameter for cemented soil is proposed as follows [14-17],
p p C (1)
T
where p is the mean effective stress, T is the critical state shear stress ratio of the
cemented clay, and C is a parameter related to the shear strength contributed by
cementation. The operative shear stress ratio is modified as,
q
U (2)
p
2.2. Material idealisation
In the SCC model for cementation effect, cemented clay is idealised as an isotropic
material with elastic and virgin yielding behaviour. The yield surface varies
isotropically with plastic volumetric deformation. Soil behaviour is assumed to be
elastic for any stress excursion inside the current structural yield surface. Virgin
yielding occurs for a stress variation originating on the structural yield surface and
causing it to change. During virgin yielding and softening, the current stress stays on
the structural yield surface. The removal of cementation is assumed to occur mainly
during the rearrangement of soil particles to form a critical state of deformation and
therefore is represented only after soil reaches its peak strength.
The idealisation of the mechanical behaviour of cemented clays is illustrated in
Figure 1. The Isotropic Compression Line (ICL) forms the boundary of the soil states
in the e lnp space. The virgin compression behaviour of a cemented soil is related to
that of its parent clay in reconstitued state by the following equation,
S. Horpibulsuk and M.D. Liu / Deformation of Structured Soil with Cementation 713
e e * Ve (3)
where e* is the voids ratio of the reconstituted clay at same stress state with the same
yield surface. Ve is the additional voids ratio sustained by soil structure.
Elastic behaviour Virgin yielding
Structured soil:
Vei
e = e* + Ve
Voids ratio e
e
Ve
e*
Reconstituted soil: e*
p'y,i p'
Mean effective stress lnp'
(a) Compression behaviour of structural soils
q CSL
M
M Structural yield
surface
Equivalent
yield surface
p'
p' o
C/M p' s
(b) Structural and equivalent yield surfaces
Figure 1. Structured Cam Clay model with cementation effect
Following the work by Liu and Carter on the compression behaviour of structured
soils and considering the consistency with the introduction of operative mean effective
stress p [18-19], the following compression equation for cemented clay is proposed,
b
py ,i C
e e * Vei
c T c (4)
p

Parameter py,i is the mean effective stress at which virgin yielding of the structured
soil begins, b and c are soil parameters describing the additional voids ratio sustained
by cementation. Vei is the value of the additional voids ratio at the start of virgin
yielding (Figure 1a). Parameter c is the part of cementation structure that cannot be
removed by loading.
Following the tradition of the Modified Cam Clay model, the yield surface of a
cemented clay in p-q space is also assumed to be elliptical and is described as (Figure
1b),
f q 2
Tp ps
p 0 (5)
where ps is the size of the yield surface, which is effectively shifted to the left along
the p axis by a distance of C / T . The size of the initial yield surface, ps,i is thus
linked to the initial mean effective yield stress py,i by the following equation,
ps,i py ,i C (6)

T
2.3. Virgin yielding behaviour
For stress states on the yield surface and with dps > 0, virgin yielding occurs. The
deformation of the soil is made up of elastic part and plastic part. A detailed study on
constitutive modeling of geomaterials based on plastic volumetric hardening and
destructuring can be found in papers by Liu and Carter [20] and Horpibulsuk and Liu
[13]. The plastic volumetric strain increment obtained is
J P 7U M G dps
d v d ve I Q *
S b Ve
c N1 T
U K F 1 e p (7)
H O LE s
where Q* and S are the compression and the swelling indices of reconstituted clay,
respectively, 7 is a soil parameter describing the destructuring associated with shearing,
and
JVe
c, if Ve
c W 0
Ve
c I (8)
H 0, if Ve
c X 0
The deviatoric strain increment is worked out based on the plastic flow rule. In the
formulation of the flow rule suitable for cemented clays, the following conditions have
been considered
(1) The flow rule should be identical to that of the MCC model if the soil has no
cementation or if the cementation structure has been completely removed, i.e., Ve = 0.
(2) A final failure state has yet to be reached if all the part of cementation
structure destructible has not been completely removed, i.e., (Ve - c) > 0.
d dp
(3) Soil reaches the critical state, i.e., Y when U U M , C = 0, and
d vp
(Ve - c) = 0.
d dp 2U
(9)
d vp po
T
U Z 1
2 2
ps
where Z is a model parameter. po is the size of the equivalent yield surface, and its
value is defined by
eIC *
e c

Q *
S
e
po S
(10)

p Q *
S
where e*IC is the voids ratio of the reconstituted isotropic compression line at p = 1
kPa. When stress states reach the yield surface and with U C T , softening occurs.
During the softening process, the yield surface shrinks, i.e., dps < 0. The volumetric
deformation of soil is described by the same equation as that for virgin yielding, i.e.,
Eq.(10). However, a modification to the deviatoric strain increment is made to ensure
that the deviatoric deformation contributed by destructuring is always positive, i.e.,
2U P 7U b Ve
c M dps
d dp
; N Q *
S
K (11)
T
U Z 1
2 2 po O T
U L 1 e ps
ps
2.4. Removal of cementation structure
The soil usually has two strengths: the peak strength and the final failure strength,
irrespective of the initial stress state or the values of OCR. Cementation contributes the
peak strength of soil. After peak strength state, the strength of soil will decrease with
the breakdown of cementation. The strength of soil after the removal of cementation is
inevitably lower than that at the peak. Thus, two strengths are usually observed. Final
failure strength is reached when the cementation strength is completely broken.
The post-peak deformation of soil is generally much larger than the pre-peak
deformation. For artificially strongly cemented soil, the breakdown of cementation
mainly takes place during post-peak strength deformation. It is therefore assumed that
the breakage of cementation only occurs during the post peak stage. This assumption is
an approximation and for the purpose of simplicity. The following assumptions are
made in order to work out the stress and strain relationship during this process.
(1) the operative effective stress state stays on the line defined by M but may
travel along the line either upwards or downwards, depending on hardening or
softening, respectively. Therefore
q
T (12)
p
(2) Based on trial and error, the function for the crushing of soil-cementation
structure is dependent on the size change of the structural yield surface
C dps
dC
(13)
Cin q

T
p
where Cin is the value of the initial cementation strength. At the end of the process, soil
reaches the final critical failure state with C = 0 and Ve = c.
2.5. Model parameters
There are in total twelve parameters in the SCC model with cementation effect. They
are e*IC, Q*, 6*, T, S, b, c, 7, Z, py,i, Vei, and C as listed in Table 1. The first three
parameters, denoted by the symbol *, are intrinsic soil properties. They are
independent of soil structure and thus are assumed to be the same as the parent clay.
Parameters T and S are dependent on cement content. These five parameters are the
same as those adopted in the Modified Cam Clay model, i.e., their physical meaning
and their identification [21].
Parameters, b, c, 7, Z, py,i and C are strongly dependent on soil cementation
structure. Their values are dependent on the magnitudes of cementation. The physical
meanings of these parameters are basically the same as those given in the work by Liu
and Carter [10], Carter and Liu [11], and Horpibulsuk et al. [12].
Table 1. Parameters of the Structured Cam Clay with cementation effect

Symbol Description
Intrinsic soil properties
e*IC void ratio at the p = 1 on the ICL in e-lnp space
Q* gradient of the normal compression line in e-lnp space
6* Poissons ratio
Parameters defining soil structure
M critical state stress ratio for cemented clay
S gradient of the unloading and reloading line in e-lnp space
b destructuring index
Vei additional voids ratio sustained by soil structure at the start of virgin yielding
c additional voids ratio sustained by soil structure at very large confining pressures
py,i initial (structural) yield stress for isotropic compression
7 parameter defining the plastic potential
Z flow rule parameter
C cementation strength
3. Simulation of the behaviour of cemented soils
The proposed SCC model for cementation effect is applied to simulate the behaviour of
soil treated by cement and fibre reinforcement in this section. The parent soil is a silty
sand. The tests were performed by Consoli et al [22]. The behaviour of the parent soil
(with no cement and no fibre) was simulated first, and the SCC model for cementation
effect is identical to the Modified Cam Clay (MCC) model under this situation, i.e.,
Ve = 0 and C = 0. The values of MCC model parameters obtained are shown in Table
2. The comparison between model simulations and experimental data are shown in
Figure 2a. The behaviour of the treated soil with 1% cement and 3% fibre was also
simulated. The values of structural parameters obtained are also shown in Table 2. The
comparison between model simulations and experimental data are shown in Figure 2b.
The untreated soil behaviour is represented by the Modified Cam Clay model,
because the soil is assumed to possess no structure or in reconstituted state. The silty
sand is compacted and has a large yield surface. It is seen that the model captures the
behaviour of the parent soil in reconstituted states only qualitatively. All the three
specimens are in highly over-consolidated states and the behaviour of the soil is
typically dry behaviour. The final strength and the general volumetric behaviour are
reasonably represented. However, the peak strengths of the soil are not reliably
predicted. This may be attributed to the fact that the MCC model generally over-
predicts the peak strength of over-consolidated soil, especially compacted silty sand.
Table 2 Values of SCC model parameters for a silty sand

Soil Q* S T eIC* 6* b Vei c psi Z 7 C
(kPa) (kPa)
Parent soil 0.1 0.012 1.37 2 0.3 -- -- -- 440 -- -- --
1% cement & 3% fibre 0.1 0.012 1.77 2 0.3 1 0.04 0 900 0.8 2 30
Figure 2. Stress and strain behaviour of a) the untreated and compacted silty sand the cement and b) fibre
reinforced silty sand (Data after Consoli et al. [22])
The behaviour of the cement (1%) and fibre (3%) treated soil was simulated with
the proposed model. In the model, the effect of the cement and fibre reinforcement is
modelled by three major features: (1) the cementation effect by parameter C, (2) the
increased final failure strength by M, and (3) the enlarged yield surface by psi. The
values of parameters C and M identified by Consoli et al [22] are adopted. Parameter
psi is determined by the peak strength of the test with 83 = 100 kPa. All the three soil
specimens exhibit softening behaviour and breakdown of cementation. For the test with
83 = 100 kPa, volumetric compression is observed during softening. As compared with
untreated parent soil, the treated soil has a higher final failure shear stress ratio and
larger drop in the strength during the softening process. All these features have been
captured well by the proposed model qualitatively. For the same reason as that for the
untreated soil, the main reason for the quantitative discrepancy between the model
simulation and experimental data is that the peak strength of the soil in over-
consolidated states is not predicted reliably.
4. Conclusion
The Structured Cam Clay Model is extended for strong cementation effect.
Modifications to the Structured Cam Clay model are proposed so that the influence of
cementation on soil behaviour can be incorporated. The model is used to simulate the
shearing behaviour of a compacted silty sand with cement and fibre reinforcement. It is
seen that the model has the capacity to describe consistently the behaviour of soils with
a wide range of structures from natural states, artificially cemented states, to cement
and fibre reinforced states.
5. References
[1] J.T. Huang, and D.W. Airey, Properties of an artificially cemented carbonate sand, J. Geotechnical
and Geoenvironmental Engineering 124 (1998), 492-499.
[2] S. Horpibulsuk, D.T. Bergado, and G.A. Lorenzo, Compressibility of cement admixed clays at high
water content, Geotechnique 54 (2004), pp.151-154.
[3] G.A. Lorenzo, and D.T. Bergado, Fundamental parameters of cement-admixed clay: new approach,
Journal of Geotechnical and Geoenvironmental Engineering 130 (2004), 1042-1050
[4] F.H. Lee, Y. Lee, S.H. Chew, and K.Y. Yong, Strength and Modulus of marine Clay-cement mixes, J.
of Geotechnical and Geoenvironmental Engineering, 131 (2005), 178186.
[5] N.C. Consoli, G.V. Rotta, and P.D.M. Prietto, Yielding compressibility strength relationship for an
artificially cemented soil cured under stress, Gotechnique 56 (2006), 6972.
[6] J.C. Chai, N. Miura, and H.H. Zhu, Compression and consolidation characteristics of structured
natural clays, Canadian Geotechnical Journal 41 (2004), 1250-1258.
[7] D. Masn, A hypoplastic constitutive model for clays with meta-stable structure, Canadian
[8] W.M. Yan and X.S. Li, A model for natural soil with bonds, Gotechnique 61 (2011), 95-106.
[9] J. Suebsuk, S. Horpibulsuk, and M.D. Liu A critical sate model for overconsolidated structured clays,
Computers and Geotechnics 38 (2011), 648-658.
[10] M.D. Liu, and J.P Carter, Structured Cam Clay Model, Canadian Geotechnical Journal 39 (2002),
1313-1332.
[11] C.P. Carter and Liu M.D., Review of the Structured Cam Clay model, Soil constitutive models:
evaluation, selection, and calibration, ASCE, Geotechnical special publication No. 128 (2005), 99-
132.
[12] S. Horpibulsuk, M.D. Liu, S. Liyanapathirana, and J. Suebsuk, Behaviour of cemented clay
simulated via the theoretical framework of the SCC model, Computers and Geotechnics 37 (2010), 1-9.
[13] S. Horpibulsuk, and M.D. Liu, Structured Cam Clay Model with Cementation Effect, accepted by
Geotechnical Engineering Journal of the SEAGS & AGSSEA (2015).
[14] M.R. Coop and J.H. Atkinson, The mechanics of cemented carbonate sands, Gotechnique 43 (1993),
53-67.
[15] A. Gens, and R. Nova, Conceptual bases for constitutive model for bonded soil and weak rocks,
Geotechnical Engineering of Hard Soil-Soft Rocks, Balkema (1993).
[16] S. Horpibulsuk, N. Miura, and D.T. Bergado, Undrained shear behaviour of cement admixed clay at
high water content, Journal of Geotechnical and Geoenvironmental Engineering 130 (2004), 1096-
1105.
[17] K. Kasama, H. Ochiai, and N. Yasufuku, On the stress-strain behaviour of lightly cemented clay based
on an extended critical state concept, Soils and Foundations 40 (2000), 37-47.
[18] M.D. Liu, and J.P Carter, Virgin compression of structured soils, Gotechnique 49 (1999), 43-57.
[19] M.D. Liu, and J.P. Carter, Modelling the destructuring of soils during virgin compression,
Gotechnique 50 (2000), 479-483.
[20] M.D. Liu, and J.P. Carter, The volumetric deformation of natural clays, International Journal of
Geomechanics 3 (2003), 236-252.
[21] D. Muir-Wood, Soil Behaviour and Critical State Soil Mechanics, Cambridge University Press (1990).
[22] N.C. Consoli, P.D.M. Prietto, and L.A. Ulbrich, Influence of fibre and cement addition on behaviour
of sandy soil, J. Geotechnical and Geoenvironmental Engineering 124 (1998), 1211-1214.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-719
Aging of Sand Grain Contacts at

Microscale: Numerical Study
Zhijie WANGa,1 and Radoslaw MICHALOWSKI a
a
Department of Civil and Environmental Engineering, University of Michigan
Abstract. Sands change their properties over time even under constant loads: they
age. The changes include increase in small strain stiffness and the resistance to
cone penetration. Mechanisms behind aging have been proposed in the past, but
none has gained wide acceptance. Previous research did not produce sufficient
information about the behavior of individual contacts between grains. A static
fatigue hypothesis is adopted here, which attributes aging of sand to evolution of
micromechanical properties at grain contacts, caused by delayed fracturing of
textural features at grain surfaces. A distinct element model of a small portion of
an inter-grain contact region was constructed; the solid material of the grain was
modeled with an assembly of sub-particles bonded together with the parallel-bond
stress corrosion model (PSC), which is able to mimic the stress corrosion process
at the near-surface grain region. The surface at the contact was generated with a
random topography to simulate the rough sand grain surface. Numerical static
fatigue tests were conducted on this model, where normal load was applied by a
rigid plate and was kept constant throughout the test. The true contact area was
monitored in term of the number of contact points, as stress corrosion developed.
The results provided useful information regarding evolution of grain contacts,
which in turn, shed light on the mechanisms of sand aging.
Keywords. Sand aging, static fatigue, grain contacts, numerical modeling
1. Introduction
Sands experience changes over time: they age. Reported examples of sand aging
include changes in small strain stiffness, cone penetration resistance, liquefaction
resistance, and pile setup [1-4]. Previous researchers have proposed several
mechanisms to explain this phenomenon, but no consensus has been reached as to what
mechanism plays a predominant role. Gradual formation of inter-grain bonds was
proposed as one of the first mechanisms behind sand aging [2], but evidence did not
support this hypothesis. Another proposed mechanism is rearrangement of sand grains
which causes interlocking of particle roughness and more efficient particle packing
[5,6]. This rearrangement hypothesis may be undermined by the experimental
observation that penetration resistance often decreases immediately after
disturbance/vibro-compaction [2]. Vibro-compaction is an efficient way to rearrange
the particles into a denser packing, thus an increase in cone penetration resistance
should have been observed (immediately after dissipation of excess pore water
pressure) if grain rearrangement was the key mechanism. In reference [15] a different
hypothesis was proposed: a cause of time effects in geomaterials was attributed to
1
Corresponding Author (E-mail: [email protected])
720 Z. Wang and R. Michalowski / Aging of Sand Grain Contacts at Microscale: Numerical Study
viscous flow. Viscosity, however, is a phenomenological concept, and it does not

indicate a specific mechanism that is responsible for rate effects and aging; thus, it
cannot be used as explanation of the nature of the process.
A static fatigue hypothesis is adopted here [7], which attributes aging of sand to
evolution of micromechanical properties at grain contacts, caused by delayed fracturing
of topographic features at the grain surface. After disturbance (such as vibro-
compaction) there are many fresh contacts formed, which have not been subjected to
aging before. The surfaces in contact are rich in intact micro-topographic features
(asperities). Under loading, micro-fractures develop and they continue to develop in
time even under constant load. Consequently, surface topography and micromechanical
properties of a contact are altered. As a result of micro-fracturing at a contact, the true
contact area increases, leading to an increase in contact stiffness and friction. A discrete
element model of a contact was developed. The model was able to simulate stress
corrosion of surface material under loading. Numerical static fatigue tests were carried
out on this model and the change in the number of contact points was monitored. The
results of simulations are presented, and the changes in micromechanical properties are
discussed.
t=0 t = 0.99tf t = tf
(a) (b)
Figure 1. Simulation of static fatigue of rock with PSC model: (a) 2D model of a rock specimen (31.7 by
63.4 mm) modeled with 5003 bonded particles; (b) time-dependent appearance of cracks in the model, tf is
the time-to-failure (recalculated after Potyondy 2007 [9]).
2. Discrete Element Modeling
Discrete element modeling (DEM) was used to simulate a grain contact. DEM has been
reported to be capable of simulating not only elastic but also fracture behavior of
geomaterials.
Potyondy and Cundall (2004) [8] proposed a bonded-particle model (BPM) to
simulate rocks, with fracturing simulated with breakage of inter-particle bonds. This
model reproduced most of mechanical behavior of rock [8]. To simulate stress
corrosion cracking in rocks, Potyondy (2007) [9] modified BPM to include a decaying
mechanism in the parallel bonds between particles under tensile stress and the modified
model was named parallel-bonded stress corrosion model (PSC). Figure 1 shows an
example of the application of PSC model. A 2D model of a rock specimen (31.7 by
63.4 mm) was formed by 5003 particles bonded through PSC model, Figure 1(a), and
the stress corrosion microcracks developed over time in the specimen are shown in
Figure 1(b). The model was loaded with a constant vertical stress of 144 MPa and a
Z. Wang and R. Michalowski / Aging of Sand Grain Contacts at Microscale: Numerical Study 721
constant confining stress of 0.1 MPa. Figure 1(b) show the micro-cracks developed at
the initial stage, at 0.99tf and tf, where tf is the time for the model to fail (here: ~59.4
hours); the micro-cracks shown as light gray were caused by tension and dark gray
were caused by shear.
3. Model Implementation
In order to study the evolution of grain surfaces due to stress corrosion, a DEM model
of a small region of grain surface material was developed using the PSC model.
Different tests were carried out on this model and the evolution of the surface at contact
was analyzed.
3.1. Dimensions of the Model
Scanning electron microscope images indicate that the size of the areas of inter-
granular contacts is of the order of tens of microns. This is illustrated in Figure 2.
Figure 2. Scanning electron microscope images of a contact on an Ottawa 20-30 sand grain before and after
loading.
Consequently, the size of the contact was estimated as 30 30 m, and the 3D region
taken in simulations was 30 30 14 m, as indicated in Figure 3.
Figure 3 DEM model of a small region of grain in the neighborhood of the contact, with the bottom surface
being the nominal area of contact.
3.2. Random Surface Generation
The surface at contact in this model was a random surface generated as a surface with a
random distribution in the heights of asperities (spheres on the surface). This was
achieved by first generating a random surface from Diamond-Square Algorithm (DSA)
and then approximating this random surface by carving the bottom surface of the DEM
model [12]. DSA has been used widely to generate random and fractal terrain. Two
random surfaces with different resolutions, but the same roughness, were generated
with a DSA [13]. The two examples are shown in Figure 4: both surfaces have an area
of 30 30 m; the coarse surface was divided into 289 unit squares with the Hurst
roughness exponent H equal to 0.7; the fine surface was divided into 1089 unit squares
with the same value of H. The random surface at contact in the DEM model was
generated by carving the initial nominally flat surface formed by bonded particles to
approximate surfaces generated by DSA. In an attempt to approximate the DSA
surfaces with DEM accurately, a refinement region with smaller particles was
generated. The diameters of the particles within the refined region were kept smaller
than the edge length of a unit square in the DSA surface.
((a)) 5 iterations,
it ti 1717
17 17 = 289 squares, (b) 6 iterations, 3333 = 1089 squares,
Hurst roughness exponent H = 0.7 Hurst roughness exponent H = 0.7
Figure 4. Random surfaces generated by DSA: (a) coarse (low-resolution) surface; (b) fine (high-resolution)
surface.
3.3. DEM Model Parameters
The values of parameters used in this DEM model were chosen to be in reasonable
ranges. More elaborate procedure for parameters selection will be carried out in future
studies, which will involve the calibration of the model. The parameters used in this
model are listed in Table 1. The material within the refined region was defined as softer
and weaker to account for weathering at grain surface. For the meaning of each
parameter, refer to [8,9,14].
Table 1. Parameters used in the model of grain surface material
Parameter Value
Wall stiffness 2 107 N/m
Sub-particle properties
Mass density, 4109 kg/m3
Youngs modulus, Ec 72 GPa (7.2 GPa in RR)
Minimum radius, Rmin 0.75 10-6 m
3 (for low-resolution surface)
Refinement level
4 (for high-resolution surface)
Rmax/Rmin 1.66
Friction coefficient, 0.5
Normal stiffness/shear stiffness ratio 2.5
Parallel bond
Bond radius multiplier, 1
Youngs modulus, E c 72 GPa (7.2 GPa in RR)
Normal stiffness/shear stiffness 2.5
Mean normal strength, 8 175 MPa (17.5 MPa in RR)
Mean shear strength, [ 175 MPa (17.5 MPa in RR)
Normal and shear strength standard deviation 40 MPa (4 MPa in RR)
PSC parameters
Threshold stress, 8 a 70 MPa
Material damage-rate constant, 1 5 10-17 m/s
Material damage-rate constant, 2 30
Note: RR refinement region
3.4. Tracking number of contact points
Figure 5. Tracking number of contact points.
The number of contact points was determined by counting all spheres that had a contact
point with the rigid plate surface; these are shown as dark spheres in Figure 5.
4. Simulation Results
The contact region in Figure 6 was confined with constant force Pc equal to 4.2 mN in
all simulations. This region was then loaded with different vertical forces, and the
evolution of the contact (the change in the number of contact points) was monitored
during the fatigue time. The maximum vertical force used in simulations was 22.1 mN,
which is the force causing a 1% vertical strain of the region modeled (Figure 6). The
fatigue time was chosen as the time needed for the contact region to reach the total
strain of 1% (this total strain includes the immediate strain after loading).
Figure 6. Loading of the contact region (303014 \m).
4.1. Numerical Static Fatigue Tests
In static fatigue tests, both vertical and confining loads, shown in Figure 6, were kept
constant. The change in the number of contact points was monitored over time. An
increase in the number of contact points owed to static fatigue is illustrated in Figure 7.
The particles marked dark gray represent surface asperities being in contact with the
rigid plate. The change in the number of contact points is demonstrated for two
surfaces modeled with different resolutions.
Selected results from the simulations are illustrated in Figure 8, where the increase
in the number of contact points is shown as a function of the load applied. The increase
in the contact points is represented by the ratio of the number of contact points at the
end of the fatigue period (time needed to produce 1% of total strain) to the number of
the contact points immediately after load application. The load in the simulations is
normalized in Figure 8 by the load causing 1% of the immediate strain. It is interesting
to notice that the relative change in the number of contact points can be larger at
smaller loads.
The area of contact (measured here by the number of contact points) increases as a
result of static fatigue; therefore, both stiffness and friction are expected to increase
[11]. Because the stiffness and friction at grain contacts evolve in time as the result of
the process, contact fatigue needs to be considered as a factor in time-dependent
behavior of sand. The simulation results are then consistent with the static fatigue
hypothesis.
Beginning of static fatigue: 147 contact points End of static fatigue: 158 contact points
(a)
Beginning of static fatigue: 511 contact points End of static fatigue: 589 contact points
(b)
Figure 7. Increase in the number of contacts points (dark particles) during static fatigue: (a) the coarse
surface, and (b) the fine surface.
Figure 8. Change in the number of contact points as a function of normal load. The number of points is
described as the ratio of the contact points after the fatigue time to the number of points immediately after
load application; the normal force is described as a ratio of the constant force used in the fatigue simulation
to the force producing 1% immediate strain in the contact region.
5. Conclusions
The evolution of a contact between grains was demonstrated using the distinct element
method (DEM). The resolution of the surface texture affects the simulated behavior of
the contact model. The number of contact points increases in time as the result of static
fatigue, hence an expectation of a time-dependent increase in contact stiffness and
friction. The lower the load the larger the relative increase in the number of contact
points. This initially surprising result can be explained by pointing out that the contact
loaded to a large force has fewer sensitive textural features left, because of initial
crushing. It is hoped that further research will allow quantitative predictions of
increases in stiffness and friction at contact. The simulations presented are consistent
with the static fatigue hypothesis.
Acknowledgements
The work presented was supported by the National Science Foundation through grants
from the Civil, Mechanical and Manufacturing Innovation Program, No. 1129009 and
No. 1537222. The authors also would like to thank Itasca Consulting Group for
providing the PFC2D/3D codes, and Dr. David O. Potyondy for his suggestions
regarding the use of the Parallel-Bonded Stress Corrosion model.
References
[1] Afifi, S.S. and Woods, R.D. (1971). Long-term pressure effects on shear modulus of soils, Journal of
the Soil Mechanics and Foundations Division, 97(10), 1445-1460.
[2] Mitchell, J.K. and Solymar, Z.V. (1984). Time-dependent strength gain in freshly deposited or
densified sand, Journal of Geotechnical Engineering, 110(11), 1559-1576.
[3] Chow, F.C., Jardine, R.J., Brucy, F. and Nauroy, J.F. (1998). Effects of time of capacity of pipe piles in
dense marine sand, Journal of Geotechnical and Geoenvironmental Engineering. 124(3), 254-264.
[4] Seed, H. B. (1979). Soil liquefaction and cyclic mobility evaluation for level ground during
earthquakes, Journal of the Geotechnical Engineering Division, 105(2), 201-255.
[5] Mesri, G., Feng, T.W. and Benak, J.M. (1990). Postdensification penetration resistance of clean sands,
Journal of Geotechnical Engineering, 116(7), 1095-1115.
[6] Suarez, N.R. (2012). Micromechanical aspects of aging in granular soils, Ph.D. thesis, Virginia
Polytechnic Institute and State University, Blacksburg, VA.
[7] Michalowski, R.L. and Nadukuru, S.S. (2012). Static fatigue, time effects, and delayed increase in
penetration resistance after dynamic compaction of sand. Journal of Geotech. Geoenv. Engineering,
138(5), 564-574 (DOI: 10.1061/(ASCE)GT.1943-5606.0000611).
[8] Potyondy, D.O. and Cundall, P.A. (2004). A bonded-particle model for rock, International journal of
rock mechanics and mining sciences, 41(8), 1329-1364.
[9] Potyondy, D.O. (2007). Simulating stress corrosion with a bonded-particle model for rock,
International Journal of Rock Mechanics and Mining Sciences, 44(5), 677-691.
[10] Hertz, H. (1881). On the contact of elastic solids, J. Reine Angew. Math. 92, 156.
[11] Johnson, K.L. (1985). Contact Mechanics, Cambridge University Press, Cambridge, UK.
[12] Voss, R.F. Random fractal forgeries. Springer, Berlin 1985.
[13] Kaya, H. (2013). Fractal landscape generation with diamond-square algorithm
(http://cn.mathworks.com/matlabcentral/fileexchange/44714-fractal-landscape-generation-with-
diamond-square-algorithm). MATLAB Central File Exchange. Retrieved Dec 17, 2013.
[14] PFCTM (2014). Particle Flow Code, version 4. Itasca Consulting Group, Minneapolis, MN.
[15] Tatsuoka, F., Ishihara, M., Di Benedetto, H. and Kuwano, R. (2002). Time dependent deformation
characteristics of geomaterials and their simulation, Soils and Foundations, 42(2), 103-129.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-727
A non-linear poroelastic approach to rate

effects on piezocone tests in tailings
materials
Gracieli DIENSTMANNa,1, Samir MAGHOUSa, and Fernando SCHNAID a
a
Dept. of Civil Engineering, Federal University of Rio Grande do Sul, Brazil
Abstract. An analytical poroelastic model is briefly presented and applied in the

present paper to analyze the rate effects during the execution of in situ tests (e.g.
piezocone test) in silty materials. The study is based upon a simplified
poromechanics analysis for the expansion and subsequent consolidation of a rigid
cylinder deeply embedded within an isotropic non-linear elastic medium of infinite
extent. The pore pressure characterization is defined according the Biot
generalized theory of effective stresses. Typical results are compared to numerical
solutions showing good agreement in terms of radial distributions of stresses and
pore pressure. The suitability of employing the proposed poroelastic equations to
the interpretation of piezocone tests executed in gold tailings is considered in some
detail.
Keywords. Poromechanics, Transient flow, Cavity expansion, In-situ tests
1. Introduction
In the present paper, the theory of poroelasticity formulated by Biot [1] is applied to
investigate the problem of an infinite long solid cylinder expanding in a porous
isotropic medium. The understanding of the transient flow in this simplified model was
defined to provide insights into the interpretation of in-situ tests (e.g. piezocone, vane-
test pressumeter), pile foundation, and other geotechnical works executed in transient
geo-materials. The latter term refers to materials whose permeability is ranging from
10-5 to 10-8 m/s (silts) [2]. In these soils, the idealised drained (gravels and sand) or
undrained (clay) flow regime used for the in situ tests interpretation cannot be directly
assumed since the test response can be affected by a partially drained condition.
In this context, evaluating the drainage effects into silty materials (transient geo-
materials) is been a challenge to the geotechnical engineer. The actual knowledge is
based on the definition that the drainage condition is dependent on permeability,
compressibility, shear strength and loading rate [3, 4, 5 and 6]. Usually the
interpretation of these effects is done by experimental observation [7, 8, 9, 10, 11 and
12]. Yet few attempts have been made to develop a structured theoretical framework to
anticipate drained conditions in intermediate permeability soils (e.g. [3]), which is the
aim of the present research.
1
Corresponding Author: Gracieli Dienstmann Dept. of Civil Engineering, Federal University of Rio
Grande do Sul, Brazil, e-mail: [email protected]
728 G. Dienstmann et al. / A Non-Linear Poroelastic Approach to Rate Effects on Piezocone Tests
2. Problem statement and simplified framework of analysis
The problem is structured as a consolidation analysis of a rigid cylinder deeply

embedded within an isotropic fully saturated poroelastic medium of infinite extent. The
cylinder is subjected to a prescribed radial expansion (R) in a time (t). Figure 1 shows
the conceptual geometry and its simplification to a plane strain condition. Due to
problem symmetry, the displacement distribution can be characterized according to
] f ( r )e r (1)
Figure 1. Geometry definition.
The radial function f ( r ) that defines the displacement solution is subjected to the
following boundary conditions
f ( R) R at r=R for all t W 0 (2)
f (a) 0 at r a R for all t W 0 (3)
where condition (3) expresses that the displacement induced by the cylinder expansion
vanishes at a distance a R , which stands for the radius of the zone of influence.
Similarly, the pore fluid pressure p p ( r ) depends only on the radial coordinate
and should comply with hydraulic boundary and initial conditions:
Ap
0 at r=R for all t W 0 (4)
Ar
p p0 at t 0 for all r W R (5)
p p0 at r a R for all t W 0 (6)

G. Dienstmann et al. / A Non-Linear Poroelastic Approach to Rate Effects on Piezocone Tests 729
where conditions (4), (5) and (6) refer to the impermeability condition at the cylinder
wall, to initial pore pressure distribution and to non-disturbance of initial pore pressure
by cylinder expansion, respectively.
2.1. Poroelastic constitutive equations
Denoting 8 0 and p0 the initial field of stress and fluid pressure, the poroleastic state
equations for an isotropic material take form of
V8 Q tr 1 2G
b Vp 1 (7)
Vp
V b tr b (8)
M
where Q and G are the Lames constants, b is the Biot Coefficient, M the Biot Modulus,
while is the strain tensor, V8 8
8 0 defines the stress change, Vp p
p0 is the
pore pressure change, and RR V
0 is the Lagrangian porosity change^
Adopting the principle of infinitesimal strains the deformation tensor and the
stresses variation can be derived from (1):
f (r )
f '(r )er ` er e_ ` e_ (9)
r
f (r )
8 rr
8 rr0 Q f '(r ) 2Gf '(r )
bVp (10)
r
f (r ) f (r ) (11)
8 __
8 __0 Q f '(r ) 2G
bVp
r r
To formulate the non-linear poroelastic properties the Drucker-Prager yield
condition is considered:
F (8 ) 8 d T (8 m '
h) (12)
where 8 d is the equivalent shear stress and 8 m ' is the mean Terzaghi effective stress.
Parameters h and T respectively characterize the tensile strength and the friction
coefficient. Denoting by d 2
3
v f ' f / r
( f '2 f 2 / r 2
f f '/ r ) and
respectively the equivalent deviatoric strain and volumetric strain, the method consists
in defining a non-linear elastic material such as the stress 8 8 0 V8 derived from
the state Equation (7) meets asymptotically the above yield condition (12):
lim F (8 ) 0 (13)
d
aY
ref
where ref 1 represents a reference strain. Following the idea developed in [13], a
simple choice consists in adopting
T 1/ ref
G( v , d ) (h
K v
9 ) ; with 9 8 0 m ' (1
b) Vp (14)
2 1 d / ref
where the bulk modulus K is considered as constant.
2.2. The fluid flow problem
Combining the fluid mass balance together with Darcy law and poroelastic state
Equation (8) yield
A tr 1 A Vp
b k b 2u (15)
At M At
where b 2 stands for the Laplacian operator, and k denotes the permeability.
Seeking to obtain an analytical solution for the flow problem in terms of pore
pressure distribution, we approximate the shear modulus G of the medium by a
constant mean value G : Geq into (7). Definition of Geq is discussed later in section
2.3. Using the balance of momentum and state Equation (7) the above Equation (16)
can be reformulated as
4
M . K Geq
Au k 3
c f b 2u with c f f (16)
At @o P 4 M
N M .b K 3 Geq K
2
O L
The distribution of excess pore pressure u p
p0 is solution to the hydraulic
problem defined by the diffusion Equation (16) and associated boundary and initial
conditions (4-6). The general solution of the latter boundary value problem reaches
Y
u d
Cn* [ J 0 ( n r )
c .Y0 ( n r )].e.

c f . n2 .t
(17)
n 1
where
a
Cn *

e R
u0 (r ).[c nY0 ( r )
J 0 ( r )]rdr
(18)
a
eR
[c nY0 ( r )
J 0 ( r )]2 rdr
Functions J0 and Y0 are zero-order Bessel functions of the first and second kind
respectively. Scalar n is the n-th root of the following algebraic equation with
respect to variable x
Y1 ( x R) J 0 ( x a )
J1 ( x R) Y0 ( x a) 0 (19)
where J1 and Y1 are first-order Bessel functions referring respectively to the first and
second kind . Scalar Zn is computed from n as
Zn
Y1 ( n R) / J1 ( n R ) (20)
To complete the determination of pore pressure field, one should prescribe the
initial pore pressure distribution. Based on a literature review (e.g. [14, 15,16]), the
following expression is proposed for initial excess pore pressure:
F (r ) a a a
u0 (r ) u0,max with F (r ) 1
ln for R r a (21)
F ( R) r R r
where u0,max denotes the initial maximum value of pore pressure generated by cylinder
insertion. Its value is estimated by the pore pressure generated during consolidation
phase in a triaxial test. In the latter, the excess of pore pressure is proportional to
applied loading along the stress path. Hence,
p
u0,max c 0 1 M cs (22)
2
where pc0 denotes a reference initial consolidation pressure, and Mcs is the critical state
line inclination.
Once the excess pore pressure u is computed from Equations (16) to (20), the fluid
pressure increment can hence be determined as p Vu p0 .
2.3. Further considerations
As mentioned before, a constant elastic shear modulus was adopted in the flow
problem characterization to make possible the derivation of an analytical solution for
the problem defined by Equations (16-20). We actually consider the average shear
modulus
a
1
a
R eR
Geq G( d , v )dr (23)
This average approximation allows to mathematically uncouple pore pressure

distribution from skeleton strains in the diffusion Equation (16). Once the pore
pressure distributions u and p are determined, the displacement field is therefore
computed by using the stress expressions (10-11) into the local equilibrium equation.
This procedure yields a differential equation that governs the displacement field f ( r )
and whose solution is computed numerically.
It should be emphasized that the model described in the present paper is still under
development to incorporate more physical and theoretical aspects. The numerical
application shall thus be seen as preliminary results of the model.
3. Numerical application and experimental validation
With the aim of verifying the validity of the proposed analytical model a finite element
analysis was done using ABAQUS@ [17] finite element software. The conceptual
geometry was defined according an axisymmetric model shown in Figure 2. An initial
field of excess pore pressure u0 is adopted according Equation (21). As regard the
initial stress distribution, an isotropic and uniform (in the plane of analysis) stress field
8fR8fRg will be considered along the subsequent simulation. The normally
consolidated modied Cam Clay model [18] has been used and is typically presented as
a function of ve parameters: slope of the isotropic normal compression line QhRslope of
the unload reload line S, slope of the critical state line in p: q plane Mcs, initial voids
ratio e0 and Poissons ratio 6. Biots model for consolidation adopted in Abaqus@
follows the same principles established in the analytical solution.
8 0 , p0
Figure 2. Numerical conceptual geometry

The numerical validation was done using the average properties of a gold tailings
deposit, which is been studied at the Federal University of Rio Grande Sul, and can be
characterized as a silty material. A better characterization and a review of the set of
tests that had been done in this material can be viewed at [19, 20, 21]. For the
numerical validation the tailings properties are considered as: Qf^fihRSf^fghRe0=1.2,
Ks=0.1GPa, Kw=2.2GPa, Mcs=0.8, 6=0.3, ref=0.1, R=2.5cm, a=100R, pc0=100kPa,
p0=50kPa. Applied displacements were limited to 15% of the equipment radius (R).
Figure 3 shows the variation of radial stresses and pore pressures which were
normalized by p0 and plotted as a function of radial distance (r/R). Analytical and
numerical solutions are direct compared for simulations performed at several expansion
rates showing a reasonable agreement over the entire loaded zone. The ratio of 8rr/p0
varies logarithmically with the reciprocal of the radius, showing sensitivity to
expansion rate.
Figure 4 shows the comparison between analytical and numerical predictions of
stress strain relation, for drained and undrained expansions of an element close to the
cylinder wall. In this case it is also possible to characterize a reasonable agreement
between models. A slightly more rigid response is characterized by the analytical
results in both drained and undrained regimes.
After the previous verification according the stresses and strain prediction (Figure
3 and 4) analytical and numerical results were compared to field data into a normalized
V ;U plot (where U is the degree of drainage and V is a normalized velocity V=v.d/ch,
where v is the test velocity, d is the test diameter and ch is the coefficient of
consolidations) shown in Figure 5. This space of normalized velocity is used for
assessing the transition from drained, undrained and partially drained condition.
It can be seen from Figure 5 that the transition from drained to partially drained
arises at a normalized velocity of the order of 0.0001 for analytical and numerical
results, whereas measured data yields a value of about 1. The onset of undrained
response takes place at normalized velocity of the order of 1 for analytical and 100 for
the numerical data, whereas measurements indicate the onset of undrained expansion at
V=100 (agreement with the numerical). The results indicate that the undrained
behavior is partially captured by the models, whereas the transition to drained regime is
underestimated.
Given the approximate nature of solutions to model piezocone testing data,

comparisons between measured and predicted values are considered acceptable in
providing insights on how to assess drained conditions in the interpretation of in situ
tests.
(a) (b)
Figure 3. Radial stresses (a) and pore pressure (b) distribution according the normalized radial distance r/R
Figure 4. Stress strain relation
Figure 5. Comparison between Gold Tailings simulations and field data

4. Conclusions
A poroelastic analytical solution, specifically devised to investigate rate effects

induced by a rigid cylinder expanding on a soil mass, has been presented and compared
to numerical values. A reasonable agreement between analytical and numerical
predictions of stress/ strain behavior was observed, thus demonstrating the model
capabilities and reliability to future applications. The direct comparison between
analytical and numerical predictions to measurements data was done in a normalized
space of V ;U providing insights to the interpretation of the drainage condition in field
tests.
However, the discrepancy observed between model predictions and finite element
solution emphasize the need to improve the model capabilities. This is actually the
object of on-going research.
List of symbols
a radius that defines the limit of the influence zone
b Biot coefficient
cf coefficient of diffusion
cv coefficient of vertical consolidation
d cylinder diameter
e0 initial void ratio
G elastic shear modulus
Geq equivalent elastic shear modulus
h tensile strength Drucker Praguer
Jo Bessel function of the first kind and order zero
J1 Bessel function of the first kind and order one
K Bulk modulus
k permeability
Ks Bulk modulus of the soil grains
Kw Bulk modulus of the water
M Biot Modulus
Mcs slope of the critical state line in p: q plane
p,p mean total and effective stress, respectively
pc0 initial consolidation pressure
p pore pressure
r radial distance
R cylinder radius
t time
T friction coefficient Drucker Praguer
U degree of drainage
u excess pore pressure
u0, u0,max initial excess pore pressure and maximum initial pore pressure, respectively
V normalized velocity
v expansion velocity / test velocity
Y0 Bessel function of the second kind and order zero
Y1 Bessel function of the second kind and order one
RRRRRRRRRRRRRRdisplacement rate
Rn n-th root of the Bessel function
strain tensor
v volumetric strain
d equivalent deviatoric strain
ref reference strain
SRRRRRRRRRRRRRRslope of the unload reload line
QRRRRRRRRRRRRRRslope of the isotropic normal compression line
QRRRRRRRRRRRRRRLams parameter
]RRRRRRRRRRRRRRdisplacement
8,8 total and effective stress tensor
8d equivalent shear stress
8m mean Terzaghi effective stress
8rr radial stress
, 0 Eulerian porosity and initial Eulerian porosity, respectively
b2 Laplace operator
References
[1] M.A. Biot (1941). General theory of three-dimensional consolidation, Journal of Applied Physics, Vol.
12, pp. 155-164.
[2] F. Schnaid (2009). In Situ Testing in Geomechanics. 1. ed. Oxon: Taylor & Francis. v. 1. 329 p
[3] M.F. Silva (2005). Numerical and physical models of rate effects in soil penetration. Doctorate
dissertation, University of Cambridge. p. 220
[4] K. Kim, M. Prezzi, R. Salgado, and W. Lee (2008). Effect of penetration rate on cone penetration
resistance in saturated clayey soils.J. Geotech. Geoenviron. Eng., 134(8), 11421153.
[5] J.A. Schneider, B.M. Lehane and F. Schnaid (2007). Velocity effect of piezocone measurements in
normally and over consolidated clays. International Journal of Physical Modelling in Geotechnics. V
2, p. 23-34.
[6] B.M. Lehane, C.D. OLoughlin, C. Gaudin, & M.F. Randoplh (2009). Rate effects on penetrometer
resistance in kaolin. Gotechnique, 59(1): 4152.
[7] A.R. House, J.R.M.S. Oliveira and M.F. Randolph (2001). Evaluating the coefficient of consolidation
using penetration tests. International Journal of Physical Modelling in Geotechnics, 1(3): 17-25.
[8] M.F. Randolph and S.N. Hope (2004). Effect of cone velocity on cone resistance and excess pore
pressure. In: Proc. Engineering practice and performance of soft deposits, Osaka, pp. 147-152.
[9] F. Schnaid, B.M. Lehane and M. Fahey (2004). In situ test characterization of unusual geomaterials.
Proc., 2nd Int. Conf. on Site Characterization, Vol. 1, Millpress, Rotterdam, Netherlands, 4974.
[10] S.F. Chung, M.F. Randolph, and J.A. Schneider, (2006). Effect of penetration rate on penetrometer
resistance in clay. J. Geotech. Geoenviron. Eng., 132(9), 11881196.
[11] J.T. DeJong, R.A. Jaeger, R.W. Boulanger, M.F. Randolph and D.A.J. Wahl (2012). Variable
penetration rate cone testing for characterization of intermediate soils. Geotech. and Geophy. Site
Characterization4.
[12] F. Schaind, J. Bedin, and L.M. Costa Filho (2007). Piezocone in silty tailing materials. Studia
Geotechnica et Mechanica, 29(1-2),151162
[13] S. Maghous, L. Dormieux, and J. Barthelemy (2009). Micromechanical approach to the strength
properties of frictional geomaterials. European Journal of Mechanics. A, Solids, v. 28, p. 179-188.
[14] M.F. Randolph and C.P. Wroth (1979). An analytical solution for the consolidation around a driven
pile, Intl. J. Numer. and Anal. Methods in Geomech., 3(2): 217-229.
[15] H.G. Poulos and E.H. Davis (1980). Pile Foundations Analysis and Design, John Wiley and Sons, Inc.,
USA.
[16] P.H. Morrins and D.J. Willians (2000) A revision of Blights model of field vane testing. Canadian
Geotechnical Journal. 37: 10891098
[17] ABAQUS. ABAQUS Users Manual. Dassault Systmes, Simulia Corp. ABAQUS vs. 6.9,
Providence, Rohde Island, USA. 2009
[18] K.H. Roscoe and J.B. Burland (1968). On the generalized stress-strain behavior of wet clay.
Engineering Plasticity, Cambridge University Press, p 535-609.
[19] J. Bedin, (2010). Characterization of Tailings by Laboratory and in situ Tests, PhD, Federal University
of Rio Grande do Sul, Brazil. (In Portuguese).
[20] F. Schnaid, A.J.P.V. Fonseca, and L.M. Costa Filho (2012). Gold tailings liquefaction under critical
state soil mechanics. Geotechnique, v. 62, p. 263-26.
[21] F. Schnaid ; J. Bedin ; A.J.P. Viana da Fonseca and L. De Moura Costa Filho, L. (2013) . Stiffness and
Strength Governing the Static Liquefaction of Tailings. J. Geotech. Geoenviron. Eng, v. 139, p. 2136-
2144.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-737
Simulation of traffic loading on an

embankment by the finite element method
with different soil models
Javier Camacho-Tautaa,1, Fausto Molinaa, Oscar Reyes-Ortiza
a
Nueva Granada Military University, Bogota D.C., Colombia
Abstract. Large road embankments when vertical loaded, behave under

oedometric conditions. However, the measurement of the soil stiffness by triaxial
tests usually is done by imposing a deviatoric stress on a specimen previously
consolidated under isotropic conditions. Then, an elastic model is used to estimate
the vertical deformations due to traffic loadings. Moreover, a static model is not
sufficient to simulate the real condition of the traffic loading. The difference
between simple assumptions and more realistic conditions could result in a
misestimating of the strains produced by traffic. The purpose of this work is to
evaluate the effect of the dynamic traffic loading on a road embankment using the
finite element method with different constitutive soil models. A triaxial cell
equipped by local transducers, capable of manage different stress paths, was used
to measure the soil stiffness and strength properties of a granular material under
isotropic conditions. In addition, the dynamic properties of the soil were obtained
by cyclic triaxial test with axial and radial measurements by mean of local
transducers and resonant-column tests. A 2D Dynamic Finite Element Model in
PLAXIS was created to estimate the cumulative deformations produced by a
repetitive moving load on a road segment. Mohr-Coulomb, Hardening Soil and
Hardening Soil with Small Strain Stiffness constitutive models were used to
evaluate the dynamic behaviour of a road embankment.
Keywords. FEM, Hardening-Soil, PLAXIS, Small strain stiffness, Traffic Loading
1. Introduction
The traffic action on pavement structures produces elastic and permanents

deformations, the permanent deformation of the subsoil due to traffic load is one of the
important factors which controls the design life as well as the maintenance cost of the
road [1]. These stress-strains conditions can be applied in laboratory by using a triaxial
apparatus, simulating the cyclic traffic loading under different stress paths and
confining pressures.
The strains in the base, sub base or subgrade can produce settlements which can
affect the quality service in any pavement structure; even the small strain which lead
significant settlements after several years of operation [2]. The existing methods to
estimate the settlements on the roads can be divided into three groups: 1) numerical
methods; 2) equivalent static loading methods; and 3) empirical equations [1].
Niemunis [3] proposed the model of High-Cycle Accumulation Model based in the
1
738 J. Camacho-Tauta et al. / Simulation of Trafc Loading on an Embankment by the FEM
finite-element method (FEM), this model was developed originally to estimate the
accumulated deformations due to cyclic loads in sands.
The purpose of this paper is to compare the plastic deformations obtained in three
sceneries that include three different constitutive models, simulating the traffic
conditions upon of a granular base. Each model was developed in the geotechnical
software PLAXIS using the FEM. Numerical results obtained through this
investigation will be validated in a near future using a full-scale model, currently under
construction at the University Campus.
2. The plasticity and permanent strain in granular bases
In many materials the Hookes law can describe the strain stress response. For a small
strain range, known as the elastic range, the strain is recoverable. However in upper
ranges, the material experiments energy dissipation and eventually the deformation is
permanent. Figure 1 shows the process of deformation during a loading and unloading
process; which produces a plastic deformation without generating a failure on the
specimen.
Figure 1. The elastic plastic response.
The irrecoverable or permanent deformation that remains under zero loading is the
plastic strain and can be regarded as a new reference state, from which the elastic
response can be measured, provided the failure has not been exceeded [4]. This
situation usually occurs in pavement structures.
3. Constitutive models
Different constitutive models can model the soil behaviour; each one adapts better
according to soil type, geotechnical structure, phenomena and loading conditions. A
constitutive model describes the stress strain behaviour by a set a mathematical
equations. Material models for soil and rock are generally expressed as a relationship
between infinitesimal increments of effective stress and infinitesimal increments of
strain [5].
J. Camacho-Tauta et al. / Simulation of Trafc Loading on an Embankment by the FEM 739
3.1. Mohr-Coulomb Model
Mohr-Coulomb is an elastic plastic-perfect model, which was developed from elasticity

law of Hooke. This model describes the behaviour of loose sands and normally
consolidated clays, involving a unique and fixed yield surface associated with
irreversible plastic deformations [6]. This model does not have an associated flow rule,
therefore not included the plasticizing phenomena as is showing in the Figure 2.
Figure 2. The Mohr-Coulomb yield surface in principal stress [5]
In PLAXIS the basic formulation requires five basic parameters: Youngs

Modulus E, Poisson ratio , cohesion c, friction angle , and dilatance angle . Mohr-
Coulomb is widely used in geotechnical engineering to determinate the shear strength
although the model does not consider effects under different stress paths.
3.2. Hardening Soil Model
Hardening Soil is an advanced model that involves the stress-strain dependence as a

function of stiffness, due plasticizing of the material. A non-associated flow rule is
adopted when related to frictional hardening and an associated flow rule is assumed for
the cap hardening. [7] [8] and [9] explain in detail the formulation and verification of
the Hardening Soil Model. The hardening is presented due to a stiffness increasing
caused by the nonlinear elastic behavior of the soil. This model is presented in Figure 3.
Figure 3. The Mohr-Coulomb yield surface in principal stress [5]
This model can simulate the loading and unloading behaviour in granular and
cohesive soils [6]. In PLAXIS, the basic formulation requires the same parameters
used in Mohr-Coulomb. In addition, it is necessary to provide the stress dependent
stiffness m, the plastic straining due to primary deviatoric loading E 50, the plastic
straining due to primary oedometric loading Eoed, and the elastic unload-reloading
modulus Eur. Figure 4 explains how to obtain these parameters. The Hardening Soil is
used in geotechnical engineering to assess the deformations in retaining walls, pile

sheets, and process of load and reload.
Figure 4. Stress-strain relation in primary loading and unloading [10].
3.3. Hardening Soil Small Strain Model
Hardening Soil Small Strain considers the response during load, unload and
reloading stages in granular and fine materials [5]. Also, the model works the real
elastic behaviour, because it considers the small strain range where the deformations in
the soil are not permanents. The analysis of the stiffness of this model in PLAXIS is
based on the Hardening Soil model. The software uses all the parameters of the
Hardening Soil model and additionally includes the secant shear modulus G S and the
variation of the stiffness regarding the shear strain.
GS is a reduction of G0 to about 70%; this reduction can be obtained by the modulus
reduction curve of the material. The modulus reduction curve is a representation of the
secant shear modulus divided by the initial shear modulus (i.e the shear modulus at
very small strain) as a function of the strain level, G/G 0 vs [11]. The normalized curve
is an advantageous option because it can be simply multiplied by an appropriate
experimentally determined G0 to obtain the shear modulus for any strain level. This
curve synthesizes the nonlinear behaviour of the soil (Figure 5).
Figure 5. Characteristic modulus reduction curve and typical strain ranges[12]
4. Materials
Five specimens of a granular base from Soacha, Colombia were compacted at optimal
water content by the standard effort test method (ASTM D 698 [13]). These specimens
were tested to determinate the shear strength parameters and the small strain behaviour
during loading and unloading process. The physical properties of each specimen are
showed on Table 1, and the grain size distribution is presented in Figure 6.
Table 1. Physical properties of the material
Parameter Value
Gs 2.76
w [%] 9.6
[g/cm3] 1.98
LL NP
IP NP
Figure 6. Grain size distribution curve.
5. Results
Three specimens were used to perform the triaxial unconsolidated undrained test
(ASTM D 2880 [14]) to obtain the shear strength parameters and the elastic moduli.
The confining pressures were 50, 100 and 150kPa (Figure 7). The fourth specimen was
mounted on a Bishop-type triaxial cell equipped with axial and radial local transducers.
Loading and unloading cycles were applied to this specimen, subjected to three
different confinements (50, 100 and 150kPa). Figure 8 shows typical curves obtained
under a confinement of 100 kPa. The fifth specimen was installed in a resonant-column
apparatus and the modulus reduction curve was obtained for three different
confinements (50, 100 and 200 kPa) by using the ASTM D-4015 test method [15].
Figure 9 exhibits the normalized modulus reduction curves.
W
f

t W

s W
Figure 7. Mohr-Coulomb failure envelope and shear strength parameters.

^

^
Figure 8. Deviatoric Stress vs Axial Strain of granular Base, p0=100kPa.
Figure 9. Degradation curves of shear modulus.
To evaluate all the model parameters, it was calculated a relationship between the
moduli and the confinement by the criterion of confining stress dependent stiffness
modulus, that is presented by Equation 1. The amount of stresses dependency is given
by the power m, which is 0.62 to the granular base studied in this work. The
calculations results are presented in Table 2.
ccos f - s 3 sin f m
Ex = Exref (1)
ccos f + p sin f
ref
Table 2. Elastic moduli of the granular base
p0
Parameter pref=100kPa
50kPa 100kPa 150kPa
E (MPa) 472.1 555.0 700.0 577.6
E50 (MPa) 226.7 269.2 311.2 270.4
Eoed (MPa) 181.4 215.4 249.0 216.3
Eur (MPa) 465.1 523.5 672.4 555.5
Gur (MPa) 465.1 523.5 672.4 555.5
A finite element model in PLAXIS of 12m of length and 5m in depth (Figure 10)
was developed to simulate the deformations on a granular base after 14 cycles of
loading applied by a wheel that generates a stress of 120kPa on the surface.
Figure 10. Vertical stresses representation of the Mohr-Coulomb model.
The comparison between the three constitutive models shows that Hardening Soil
presents the higher deformations and Hardening Soil Small Strain presents
deformations in a range of 0 to 0.26mm. The Mohr-Coulomb model has deformations
amplitudes less than the obtained by the Hardening Soil model. The results of the three
models are in Figure 11. This figure shows the vertical deformation in a point located
on the surface at the middle of the structure.
Figure 11. Vertical deformations in the surface middle point by the three constitutive models.
6. Conclusions
An experimental program to measure the elastic properties and shear strength of a

granular base was developed by using conventional triaxial, stress path with local
sensors and resonant-column tests. All the parameters used in numerical simulations
were completely obtained experimentally.
A granular base of a simple pavement structure was simulated in PLAXIS to
estimate the deformations induced by a dynamic traffic loading. Three different
constitutive models were used: Mohr-Coulomb, Hardening Soil and Hardening Soil
Small Strain. However, to decide which model best represent the actual behaviour of
this structure is necessary compare the numerical results with the results obtained by a
full-scale model.
The magnitudes of plastic strains are different for each model. The Hardening Soil
Small Strain presented deformations in the range of very small strain. The Mohr-
Coulomb model showed a deformation asymptotic with the cycles, due to the rigid
plastic behaviour of the model. The Hardening Soil had the largest deformations and
these permanent deformations present a lineal tendency with the number of cycles.
Acknowledgments
This work is part of the Research Project INV-ING-1765 (2015) supported by the
Research Direction of the Nueva Granada Military University. The second author
acknowledges to COLCIENCIAS, Call 617/2013.
References
[1] J-C. Chai1 and N. Miura, Traffic-Load-Induced Permanent Deformation of Road on Soft Subsoil,
Journal of Geotechnical and Geoenvironmental Engineering 128 (2002), 907-916.
[2] T. Wichtmann, H.A. Rondn, A. Niemunis, T. Triantafyllidis, A. Lizcano, Prediction of Permanent
Deformation, Journal of Geotechnical and Geoenvironmental Engineering 136 (2010), 728-740.
[3] A. Niemunis, T. Wichtmann, T. Triantafyllidis, A high cycle accumulation model for sand. Journal of
Computers and Geotechnics 32, (2005), 245263.
[4] D.M. Wood, Soil Behavior and Critical State Soil Mechanics, Cambridge University Press, Cambridge,
1999.
[5] PLAXIS, Materials models manual, Amsterdam, 2014.
[6] A. Nieto, J. Camacho, E. Ruiz, Determinacin de parmetros para los modelos elastoplsticos Mohr-
Coulomb y Hardening Soil en suelos arcillosos, Revista Ingenieras Univerrsidad de Medelln 8, (2009),
75-91.
[7] C. Suraraka, S. Likitlersuangb, D. Wanatowskic , A. Balasubramaniama , E. Oha , H. Guana, Stiffness
and strength parameters for hardening soil model of soft and stiff Bangkok clays, Soils and
Foundations (2012), 52(4), 682697.
[8] T. Schanz, P.A. Vermeer, P.G. Bonnier, The hardening soil model: formulation and verification,
Beyond 2000 in Computational Geotechnics (1999). Balkema, Rotterdam.
[9] R.B.J Brinkgreve, PLAXIS Finite Element Code for Soil and Rock Analysisversion 8. Balkema,
Rotterdam, 2002.
[10] R. Obrzud, On the use of the Hardening Soil Small Strain model in geotechnical practice. Numeric in
Geotechnics and structures. Poland, 2010.
[11] J.F. Camacho-Tauta, Evaluation of the small-strain stiffness of soil by non-conventional dynamic
testing methods, Ph.D Thesis, Lisbon, 2011.
[12] J. Atkinson, G. Sallfors, (1991). Experimental determination of soil properties. In Proc. 10th ECSMFE
(1991), 3, 915-956.
[13] ASTM International, "Laboratory Compaction Characteristics of Soil Using Standard Effort (12400 ft-
lbf/ft3 (600 kN-m/m3)) (D 698)". Annual Book of ASTM Standards (2002d), West Conshohocken.
[14] ASTM International, "Standard Test Method for Unconsolidated-Undrained Triaxial Compression Test
on Cohesive Soils (D 2850)". Annual Book of ASTM Standards (2002d), West Conshohocken.
[15] ASTM International, "Modulus and Damping of Soils by the Resonant-Column Method (D 4015)".
Annual Book of ASTM Standards (2002d), West Conshohocken.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-745
Characterization of Consolidation Stress-

strain-time Histories on Pre-failure
Behavior of Natural Clayey Geomaterials
John N. MUKABIa,1
a
Engineerintel - R&D, Design & Engineering, Kensetsu Kaihatsu Ltd, Nairobi, Kenya
Abstract. Moderate to strong vibrations and ground motions, environmental factors and human
activities may cause progressive and/or abrupt changes in the recent consolidation stress-strain-time history
transforming it to a new current state. It is therefore important to evaluate such changes in order to capture a
more realistic picture of the field behavior for purposes of determining the relevant and most appropriate design
parameters.
In this study, fundamental theories related to the recently introduced Consolidation and Shear Stress
Ratio (CSSR) and Yield Strain Limit (YSL) concepts are applied, within the framework of kinematic
hardening sub-yield surface limits, in carrying out retrospective analysis in simulating, predicting and
characterizing the variation in the behavior of clayey composite ground and geomaterials subjected to various
consolidation stress-strain-time histories. In order to characterize the effects of any abrupt changes that may
occur as a result, a geomathematical module within the GECPROM (Geo-Changes Probing Model) is adopted
and quantitative analyses, simulation and modelling are undertaken in predicting the transformed ground and
material geo-characteristics. It is partially demonstrated and proposed that the characterization and
quantitative determination of sub-yield strain and multiple kinematic hardening sub-yield surface limits can be
versatile in the monitoring of geo-structural behavior, sophisticated geotechnical engineering analysis of
ground subjected to seismic and other modes of loading, ground improvement, multi-stage construction control
as well as enhanced modelling and prediction simulation.
The CSSR and YSL concepts are effectively applied within the TACH-GECPRO Model to modify the
Hardening Soil Small Strain (HSSS) hyperbolic model to partially account for the effects of consolidation
stress-strain-time history and loading rate.
Keywords. Consolidation, elastic limit strain, characterize, geo-changes
1. Introduction
The recently preferred PerformanceBased Design (PBD) fundamentally entails that,

deformation in ground and foundation soils, along with the reciprocal and/or
retrospective structural deformation and stress states, are comprehensively analyzed by
adopting sophisticated methods, particularly for structures within high exposure seismic
action regions. On the other hand, when subjected to various modes of loading and
environmental conditions, most ground and geomaterials undergo geotechnical changes
that are, as far as the geotechnical engineer is concerned, of great exigency in terms of
project costs, particularly in developing countries which suffer, more predominantly, in
lack of adequate project funding. It is therefore important to evaluate such changes in
order to capture a more realistic picture of the field behavior for purposes of determining
the relevant and most appropriate design approach and corresponding parameters. Some
1
John N. Mukabi.
746 J.N. Mukabi / Characterization of Consolidation Stress-Strain-Time Histories
relevant conceptual and quantitative models that may be useful in the correlation of
consolidation and shear behavior as well as for the characterization of related influences
such as anisotropic stress history (re-consolidation stress path) Over-Consolidation Ratio
(OCR), Long-Term Consolidation (LTC) and reconsolidation regimes are also passively
introduced.
2. Influence of Reconsolidation Stress-strain-time Histories
2.1. Stress States under Varying Reconsolidation Regimes
In order to investigate the influence of reconsolidation stress path and regime on current
stress states, undisturbed specimens extruded from samples of natural relatively stiff
Osaka Bay and stiff OAP Pleistocene clays were reconsolidated to various stress states
and axial stresses equivalent to the field overburden pressure ( ) tracing different stress
paths using a fully automated triaxial testing apparatus capable of measuring very small
to large strains. The geomaterial properties, testing regimes and
equipment can be referenced from [1].
Figure 1 shows both the varying stress states at which the small strain deformation
characteristics were investigated by performing small cyclic Triaxial Compression (TC)
probing tests and the corresponding Elastic Limit Strains {(a)ELS(YI)} determined at the
respective stress states for both normally and overconsolidated specimens.
The results in this Study show that the initial small strain stiffness as well as the YI
zone are stress state dependent notwithstanding the consolidation stress ratio, when
compared at the same stress state and magnitude. It can further be observed that the size
of the initial yield surface YI: i) increases with progressive loading up to the in-situ
overburden state of stress, subsequent to which it reduces with further loading; ii)
increases with Ko rebounding (as OCR increases); iii) exhibits interesting characteristics,
whereby upon virgin loading at constant p' (stress points 12 3), YI expands in size.
However, once it is unloaded and offset (stress points 678) and reloaded at constant
p' (stress points 89), YI is compressed, contracting in size. This may be attributed to
the memory of the most recent history, which seems to persist in Ko compression (stress
points 911), only to recover through Ko rebound indicating that the initial yield surface
is significantly influenced by the current stress state-induced anisotropy [2].
Figure 1. Effect of reconsolidation stress states on Elastic Limit Strain {(a)ELS(YI)}.

J.N. Mukabi / Characterization of Consolidation Stress-Strain-Time Histories 747
2.2. Overconsolidation Ratio (OCR)
Figure 2, which simulates cyclic loading unloading reloading in Short Term and
Long Term Consolidation (STC and LTC) shows that: i) the initial stiffness decreases
exponentially with increasing Over-Consolidation Ratio (OCR) and is practically
insensitive to the load unload reload cycles; ii) the initial modulus is virtually the
same at constant OCR notwithstanding creep effects (secondary consolidation time); iii)
long term secondary consolidation has practically no effect on the initial stiffness.
However, analogous to OCR, it can be seen to significantly increase the size of the initial
yield surface {(a)ELS(YI)}; iv) loading unloading reloading cycles in consolidation
have minimum impact on YI when compared at a constant OCR within the same
consolidation time history; and, v) LTC has irrecoverable effects on YI.
Figure 2. Effect of OCR and LTC on Elastic (initial) modulus and (a)ELS(YI)
2.3. Over-densification by Reconsolidation
Destructuration of the structure of natural geomaterials caused by heavy over-

densification was simulated by adopting the SHANSEP concept whereby the specimen

is reconsolidated well beyond its field overburden stress, . It is appreciated that
densification within the natural boundary limits of a geomaterial enhances its properties.
However, the findings from this study confirm that excessive densification beyond the
in-situ yield stress leads to large scale straining (softening) and destruction of
cementation, bonding and thixotropic components as well as causing change in the
preferred particle orientation and inherent/induced anisotropic properties of natural
clayey geomaterials. This characteristic can be modelled using Eqs. (1), (2) and (3).
(1)
where, is the resulting initial modulus and is the pseudo-yield initial

modulus determined at the stress level (pseudo-yield stress) which is higher than the
yield stress and from which the specimen is rebound, defined as;

(2)
On the other hand, deterioration leading to the reduction of the initial yield strain
{(a)ELS(YI)} is modelled based on Eq. (3).

(3)

where, is the resulting size of the initial yield strain, is the initial yield
strain determined at the pseudo-yield stress level and is a
constant that is dependent on the stiffness and nature of the geomaterial.
3. Yield Strain Limit (YSL) Concepts and Proposed Model Functions
3.1. Fundamental Postulates
Methods of quantitatively determining the YSL, which fundamentally define the

bounding strain limits of the yield surfaces considered within the framework of the
kinematic hardening multiple-yield surface concepts, were proposed by [2] and [5].
Based on experimental data and analysis, as had been postulated by various
researchers including [3] and [4], [2] confirmed that the size and shape of the yield
surfaces, particularly the initial YI, which definitively defines the range of linear elastic
and recoverable behavior circumscribing very small strains, , and, the
secondary YS defined within the small strain region, , are mostly influenced
by: i) recent stress and time history as the stress point transcends to the current stress-
state; ii) structuration due to ageing (Long Term Consolidation(LTC)) and cementation
occurring at the current stress-state as particles agglomerate; iii) loading rate and path of
perturbing stresses or strains; iv) drainage conditions; and, v) mode of loading static
(monotonic, cyclic) or dynamic. Models to characterize these aspects were also proposed
by [2],
3.2. Proposed Universal YSL Model Function for Clayey Geomaterials
The proposed model function that can be applied in delineating multiple YSLs is defined
in Eq. (4).

(4)

where, is the current stress state, accounts for the effects of consolidation
stress-strain history, is the secondary consolidation time (ageing),
is the overconsolidation ratio factor, defines the drainage conditions and
is the strain rate effect.
The initial very small strain YSL, is determined by inputting
and , the secondary small strain YSL, by considering and
and the tertiary pre-failure YSL, is derived from and
. It is important to note that the YSLs are computed cumulatively, i.e.,
, while .
4. Application of the GECPRO Model
An appreciably versatile mechanistic-empirical geo-mathematical model (GECPROM),

is introduced. GECPROM is designed to probe and estimate changes in vital geo-
properties for clayey geomaterials and ground. The significant advantage of this model
is that various geotechnical changes and geo-structural behavior can be modeled from a
single sophisticated experimental test, whilst simultaneously catering for the effects of
drainage conditions, loading rate, and consolidation stress-strain-time history.
The fundamental model equation for the elastic modulus developed from CSSR
concepts and elastic theory is expressed as:

(5)

where is the initial elastic modulus at a variable stress point , = is

the arbitrary or designated consolidation stress ratio traced to , is the initial
shear modulus determined at in-situ overburden pressure, =0.95 and =0.35 are
geomaterial constants, the values of which are applicable for most natural stiff to hard
clayey geomaterials, while =1.16 and =0.4 for stress states in the 1st quadrant and =-
1 for stress states in the 4th quadrant accordingly.
The shear modulus is then computed from Eq. (6), which adopts a new model
equation that correlates with elastic modulus proposed in [6].

(6)

On the other hand, the basic model equation defining the impact of stress states on
the elastic limit strain that defines the initial yield surface is expressed as:

(7)

where, constants =0.98, =0.32, =0.4, for stress states in the 1st quadrant
and, =-1, in the 4th quadrant, while =1.16.
Models that define the kinematic hardening sub-yield strain and surface limits and
characterize the impacted geo-changes pertaining to consolidation stress-strain-time
histories, drainage conditions, cyclic prestraining and strain rate are discussed by [1] and
[4] in detail.
The significant influence of varying stress states and overburden pressure on the
magnitude of both the stiffness and elastic limit strain can be clearly observed from the
GECPRO modelled results of the Osaka Bay Clay depicted in Figures 3a and 3b,

respectively. It can be seen that

increases with increasing which is logical
considering the effects of progressive Long-Term Consolidation (LTC) within deeper
layers of the strata (refer to Figures 4 and 5)
9000 (a) LTC-LD Osaka Bay Clay 0.018 (b) LTC-LD Osaka Bay Clay
Kc = 0.2 Kc = 0.2
Modelled Elastic Limit Strain
Elastic Modulus, E0 (MPa)
8000 0.016
Elastic Limit Strain, (a)ELS (%)

Modelled Elastic Modulus Kc = 0.3 Kc = 0.3
for Varying Stress States for Varying Overburden
7000 Kc = 0.4
0.014 Kc = 0.4
Stress States
6000 Kc = 0.5 0.012 Kc = 0.5
5000 Kc = 0.6 0.01 Kc = 0.6
4000 Kc = 0.7 0.008 Kc = 0.7
3000 Kc = 0.8 0.006 Kc = 0.8
2000 Kc = 0.9 0.004 Kc = 0.9
1000 Kc = 1.0 0.002 Kc = 1.0
0 0
0 1 2 3 4 5 0 1 2 3 4 5
Mean Effective Stress, p' (MPa) Effective Overburden Pressure, a0' (MPa)

Figure 3. GECPRO Modelled at varying stress states; a) Elastic modulus; b)

5. Proposed Modification of the HSSS Hyperbolic Model
5.1. Proposed Model Functions Considering Consolidation History and Loading Rate
Modification of the Hardening Soil Small Strain (HSSS) hyperbolic model is proposed
to account for the effects of consolidation stress-strain-time history, drainage conditions,
cyclic prestraining and loading rate. In particular, simulation of Long-Term
Consolidation (LTC), in-situ state of stress, OCR and loading rate are considered vital.
The modified secant stiffness model that accounts for these effects is expressed as:

(8)

where, = strain rate, = axial strain, and = secondary consolidation time.

Figure 4 is a comparison of the measured and modelled results considering loading rate
and LTC effects. An appreciably good agreement can be observed.
Figure 4. Comparison of modelled and measured results for: a) loading rate; and, LTC effects
A comparison of results using the modified and conventional models is presented in

Figure 5. The fact that the conventional model does not account for LTC is quite clear.
Figure 5. Comparison of: a) conventional and modified models; and, b) modified model LTC results
6. Application of the YSL and CSSR Models
6.1. Application in Characterization of Construction and Ground Geomaterials
Figure 3 is a graphical depiction of the comparison of modeled and experimentally

derived multiple kinematic sub-yield surfaces; hypothetic-empirically bound around the
respective computed Yield Strain Limit [{YSL {(a)YLS(YLS)}] related congruent p~ q
stresses. In order to generate the modelled characteristic curves depicted in Figure 3, the
numerator in Eq. (4) was applied in determining at each YSL. TACH-CSSR
functions [1] were then employed in deriving the corresponding, . This procedure
was iteratively applied for varying reconsolidation stress paths and yield planes. The
reconstituted specimens were studied to characterize the degree and detrimental impact
of destructuration in the simulation of critical geo-changes.
Modeled and experimentally derived multiple to kinematic subyield surfaces
hypothetic-empirically bound around the yield strain
1.2 = 469
1.1 INTACT p'=0.611
YI: Initial sub-yield surface q=0.185
1 YS: Secondary sub-yield surface INTACT
Deviator stress, q (Mpa)
0.9 YT: Tertiary sub-yield surface YI=1% (1.02)

YU: Ultimate yield surface YS =2.5% (1.025)
0.8 p'=0.562 YT= 4.0% (1.04)
q=0.638 =30.3 YU=3.0% (1.03
0.7
Experimental Reconstituted
0.6 Modeled curves
= 1% (1.01) for
p'=0.571 all
q=0.691 (Ko )NC Consolidation Line
0.5
0.4
p'=0.563
0.3 q=0.345
p'=0.599 p'=0.60
q=0.228
p'=0.604
q=0.146 q=0.140
0.2 RECONSTITUTED

YI: Initial sub-yield surface (Ko )OC Rebound
0.1 YS: Secondary sub-yield surface
YT: Tertiary sub-yield surface
0.09
0 p'=0.603
YU: Ultimate yield surface q=0.122
0.08
-0.1
0.56 0.57 0.58 0.59 0.6 0.61 0.62 0.63
Mean effective stress, p' (Mpa)
Figure 6. Comparison of modeled and experimentally derived multiple kinematic sub-yield surfaces
The following derivations can be made from this figure: i) the intact specimen
exhibits much larger sub-yield as well as yield surfaces compared to the reconstituted
one; a manifestation of destructuration effects; ii) the orientations of reconstituted
specimen surfaces are different from those of intact surfaces in all cases; a manifestation
of the differences in structure; iii) a link exists between the magnitude of stiffness and
the configuration of the multiple kinematic sub-yield surfaces; iv) within the Yu surface,
the magnitude and stress path rotation increases as the size of the sub-yield surfaces
becomes larger; and, v) the farther the distance of the current stress point from the failure
line, the closer the range of engagement of the neigboring surfaces.
6.2. Application in Stage Construction, Monitoring and Performance Evaluation
The concepts and models introduced in this paper have been successfully applied in the
stage construction of embankments on soft ground and swampy areas and, most recently,
for the monitoring and performance evaluation of the Geosynthetics Reinforced Soil
Retaining Walls (GRS-RWs) along the Nairobi ~ Thika Highway (A2) in Kenya. During
stage construction, for example, the YSL models are employed in ensuring that the
surcharge loadings are controlled within the predetermined small strain limits, whilst it
is monitored that the designated consolidation periods and loading rates are such that
growth of the elastic limit range can be appreciably achieved.
7. Conclusions
1) The importance of quantitatively determining the multiple kinematic hardening sub-

yield surface limits and the congruent loading limits, their application in monitoring
of geo-structural behavior, sophisticated geotechnical engineering analysis of ground,
ground improvement, multi-stage construction and control as well as enhanced
modeling and prediction, has been partially demonstrated.
2) Sophisticated analyses based on new developments in geosciences, soil mechanics
and geotechnical engineering are essential in the effective simulation, modeling and
prediction of geotechnical changes in ground and geomaterial properties.
3) The GECPROM is versatile and appreciably effective in probing, simulating,
modeling and predicting geotechnical changes in ground and geomaterial properties
and may be useful in the modification of some existing clay models.
References
[1] J.N. Mukabi, F. Tatsuoka, Y. Kohata & N. Akino, Small strain stiffness of Pleistocene clays. Procs. Int.
Symp. On Pre-failure Deformation Characteristics of Geomaterials, IS-Hokkaido 94 Balkema, Vol. 1,
pp. 189-195.
[2] J.N. Mukabi, Recent development in the quantitative determination of kinematic hardening sub-yield
surface limits. In C. Atalar, F. Cinicioglu, B.M. Das, A. Sgglamer & E. Togrol (ed), New Developments
in Soil Mechanics and Geotechnical Engineering Procs. Int. Conf. North Cyprus (2012): 321-329.
[3] J.H. Atkison, & S.E Stallebrass, Experimental determination of stress-strain-time characteristics in
laboratory and in-situ tests. Proc., 10th European Conference on SMGE. Florence (1991). pp. 915-956.
[4] F. Tatsuoka, R.J. Jardine, D.L. Presti, H.D. Benedetto, & T. Kodaka, Characterizing the pre-failure
deformation of Geomaterials. XIV IC on SMFE, Theme Lecture, Hamburg (1999), pp. 2129-216.
[5] J.N. Mukabi, & Z. Hossain, Characterization and modeling of various aspects of pre-failure deformation
of clayey Geomaterials Application in Modeling. In Z. Hossain & T. Sakai (ed), KEYNOTE LECTURE,
Proc., 1st International Conference on Geotechnique, Construction Materials & Environment,
Mie(2011b): Vol. 2. 1-10.
[6] J.N. Mukabi, Proposed Versatile Model for Determining Poissons Ratio for Civil Engineering
Applications based on Elastic Modulus, Electronic Pre-print on Academia.edu Website (2014g).
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-753
Modeling of the Small Strain Shear

Modulus on a Fiber Reinforced Sand
Jun Jos Claria,1 and Paula Vettorelob
a
Universidad Nacional de Crdoba, Crdoba, Argentina.
b
Institute for Advanced Studies in Engineering and Technology, UNC-CONICET,
Crdoba, Argentina.
Abstract. The mechanical behavior of fiber reinforced soils has been extensively
studied in the last decades. Previous studies have shown that inclusion of fibers
increases the shear strength of the reinforced soil. However in some cases the
presence of fibers can reduce the stiffness of the composite material. In this paper,
we study the change on the initial stiffness in an alluvial sand reinforced with
polypropylene fibers. A model based on Hertz elastic contact theory is developed
in order to explain the trends of the maximum shear modulus in the fiber
reinforced sand as the fiber content is varied. The model assumes that the shear
wave is transmitted through elastic distortions at the contacts, so the stiffness of
the contacts governs the initial shear modulus, which in turn is affected due to
fibers addition. Furthermore, the ratio between the amount of grain to fiber
contacts and the total of contacts on the shear wave path influence the maximum
shear modulus. An experimental testing program involving confined compression
tests with shear wave velocity measurements of unreinforced and fiber-reinforced
sand specimens was undertaken to validate the proposed model trends. The model
predictions were found to agree well with the experimental results.
Keywords. Sand with fibers, shear modulus, shear wave velocity.
1. Introduction
Soil reinforcement by means of fiber addition has been reported in the last few decades
by several investigators ([1], [2], [3], [4], [5], and [6]). In general, studies show that
addition of fibers increase the shear strength at large strains of the reinforced soil.
However, as far as we know, only a few studies dealing with the effect of fiber
inclusion on the stiffness of reinforced soil at low strain levels have been published.
Among these, Heineck et al. ([7]) observed that inclusion of fibers do not change the
initial stiffness at low strain levels (10-5) of the reinforced soil, when the fiber content is
up to 0.5% by weight of dry soil. Furthermore, Diambra et al. ([8]) observed that shear
modulus at medium strain levels (10-3) is not affected by fibers incorporation.
However, other investigators suggest that when fiber content is higher than 0.5% by
weight, the stiffness of the reinforced soil at low strain levels is reduced ([9]).
1
[email protected].
754 J.J. Clari and P. Vettorelo / Modeling of the Small Strain Shear Modulus
This work presents a model based on Hertz theory (elastic contact theory)
developed with the aim to explain the trends of the maximum shear modulus (G max) in a
fiber reinforced sand as the fiber content is varied. The model predictions are validated
by means of an experimental testing program.
2. Physical Model of Contacts
2.1. Introduction to the Model
At low strain levels (10-5 or less), it can be assumed that soils behave elastically, so
there is a unique and direct relation between shear wave velocity (V s) and initial shear
modulus or maximum shear modulus (Gmax) given by:
Where is the soil density. The shear wave velocity is calculated by measuring the
time that a mechanical shear wave needs to travel a certain distance along a soil
specimen. This shear wave is transmitted through elastic distortions at the contacts of
the soil grains and the contacts of fibers to soil grains (Figure 1).
Figure 1. (a): The shear wave is transmitted through elastic distortions at the grain-to-grain contacts; (b): The
shear wave is transmitted through elastic distortions at the grain-to-grain and fiber-to-grain contacts.
Thus, the initial shear modulus of the fiber reinforced soil (G max) is a function of
the grain-to-grain contact stiffness (gg), the fiber-to-grain contact stiffness (gf), the
number of grain-to-grain contacts (Ngg) and the number of fiber-to-grain contacts (N gf).
We define the parameters as the ratio between the fiber-to-grain contact stiffness and
the grain-to-grain contact stiffness (Eq. (2)), and as the ratio of the number of fiber-
to-grain contacts to the total of contacts through which the shear wave is transmitted
(Eq. (3)).
When fiber content is zero, the initial shear modulus is the shear modulus of the
soil without reinforcement (Gmax) and is equal to zero. Contrarily, if all contacts are
of the fiber-to-grain type, the shear modulus will take a value G max directly related to
J.J. Clari and P. Vettorelo / Modeling of the Small Strain Shear Modulus 755
the grain-to-fiber stiffness, and will be equal to one. Finally, the shear modulus G max
of the fiber reinforced soil is given as:
For grains with a stiffness very high compared with the stiffness of the contacts, it
is reasonable to assume the relationship showed in Eq. (5).
And then, the initial shear modulus of the fiber reinforced soil is obtained as:
2.2. Determination of
The parameter relates the stiffness between fiber-to-grain contacts and grain-to-grain
contacts. In order to quantify this parameter, Hertz Theory of Contacts Mechanics is
used. Particles of soils are idealized as spheres, while fibers are assumed to be
cylinders (Figure 2). Both materials are considered elastic.
According to the Hertz Theory, the shear stiffness of a contact () between two
elastic bodies is:
Where rc is the contact radius between the two bodies, and G* is the effective shear
modulus given by a combination of the elastic properties of the two bodies under
consideration (Eq. (8)).
Figure 2. Hertz theory of contact mechanics, a) particles of soils are idealized as spheres and, b) fibers are
considered cylinders..
Introducing (8) into (7) we obtain:

In Eq. (9), Gg is the shear modulus and g is the Poisson ratio of the material of the
grains of soil, and rc is the contact radius between two grains of soils of diameter d g,
given by Eq. (10):
being N the contact force.

On the other hand, the fiber-to-grain stiffness is given by the following equation:
Where Gf is the shear modulus and f is the Poisson ratio of the material of the
fiber, respectively, and rc is the contact radius between the fiber and the soil grain,
which can be approximated by:
Combining Eqs. (9), (10), (11) and (12), we obtain the parameter as follow:
2.3. Determination of
In order to evaluate , it is assumed that fibers are uniformly distributed and randomly
oriented in the soil mass. To quantify the number of contacts inside a soil cube, only
contacts between particles of soil and contacts of soil particles and fibers are
considered, but not contacts at the boundaries of the soil mass. In addition, in order to
minimize the influence of the boundaries, a volume of soil large enough is considered
to calculate .
Figure 3. Simple cubic packing (left) and face-centered cubic packing (right).
Now, we analyze the influence of the packing on parameter. For this, we studied
two cases: a simple cubic packing for a soil in a loose state, and a face-centered cubic
packing for a soil in a dense state (Figure 3).
2.3.1. Simple Cubic Packing

In the simple cubic packing, particles of soils are idealized as spheres of equal
diameter, each of one in contact with other 6 spheres, so the coordination number
(CN) is 6. Taking a cubic region of n3 particles, the number of contacts will be:
Figure 4a shows the variation of the ratio number of contacts (N c) to the

number of spheres (Ns) with the sample weight, considering particles with specific
gravity (s) equal to 2,67. It can be seen from Figure 4a that for weights of soil over 250
g the relation Nc/Ns is practically constant and approximately equal to 3.
Thus, considering that the number of contacts (Nc) is equal to three times the
number of spheres (Ns), and the number of spheres is equal to the weight of soil
specimen divided the weight of one sphere, the number of contacts can be described as
a function of the soil sample weight (Ws), particle diameter (dg) and specific gravity of
soil (s):
In order to evaluate the number of fiber-to-grain contacts, the number of fibers for
certain fiber content (CF) is calculated (Eq. (16)). Then, the number of contacts
between one single fiber and grains of soil is approximated by Eq. (17).
In Eq. (16) f is the specific gravity of the fiber, and in Eq. (17) Lf is the fiber
length. The number of fiber-to-grain contacts is given by the combination of Eqs. (16)
and (17):
Therefore, the parameter will be:
2.3.2. Face-centered Cubic Packing

The coordination number for the face-centered cubic packing is 12. Taking a cubic
region of n spheres by side, the total number of spheres is in this case:
and the number of contacts is given by Eq. (21):
Figure 4b shows the variation of the number of contacts to the number of spheres
ratio with the weight of the soil mass. As it was highlighted for the simple cubic
packing, for weights of soil over 250 g. the relation Nc/Ns is nearly constant and, in this
case, approximately equal to 6.
3,00 6,00
a) b)
5,90
2,95
5,80
2,90
Nc /N s
Nc /Ns 5,70
2,85
5,60
Simple cubic packing Face-centered cubic packing

2,80 5,50
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
Weight of soil mass [g] Weight of soil mass [g]
Figure 4. Number of contacts to number of spheres ratio versus weight of soil specimen: a) simple cubic
packing, b) face-centered cubic packing.
Now we calculate the number of contacts as a function of soil weight:
The number of fiber-to grain contacts is calculated in the same way than for the
simple cubic packing, but considering that each fiber has two times the contacts that the
same fiber in a simple cubic packing, obtaining the following result:
The parameter is obtained from combination of Eqs. (22) and (23):
It can be seen that does not depend on the packing, or the void ratio of the soil
mass, but it is a function of fiber content, the specific gravity of each material, and the
diameter of fiber to diameter of grain ratio.
3. Validation of the Model
3.1. Experimental Program
The soil used in the present study was an alluvial siliceous well-graded sand. Main
geotechnical properties of the sand are listed in Table 1. Polypropylene fibers of 10 mm
length and 0.16 mm in diameter were used throughout this work.
Table 1. Geotechnical properties of the soil used in this work.
U.S.C.S.: Unified Soil Classification System; CU: coefficient of uniformity; Cg: coefficient of gradation;
%PT#200: percentage of passing weight through the sieve IRAM N 200 (75um); s: specific gravity; d50:
diameter of 50% passing weight.
To quantify the maximum shear modulus of the reinforced soil at low strain levels
( 10-5), confined compression with shear wave velocity measurement tests were
performed. These tests were conducted in a modified oedometer with the incorporation
of bender elements in its upper and bottom caps. A detailed description of this
equipment is given in [10].
3.2. Quantification of the Model Parameters
Table 2 shows the data used to calculate parameters and . Properties of fibers were
obtained from the manufacturer. The grains diameter was assumed to be the d 50.
Finally, elastic parameters of the soil grains were assumed to be equal to the granite
properties, because of the nature of the sand.
Table 2. Soil and fibers data used in the calculation of the model's parameters.
3.3. Results and Discussion

35 40
Loo se Sand Dense Sand
30 35
25 ' v = 440 kPa 30 ' v = 440 kPa
' v = 220 kPa

Gmax 25
Gmax 20 [MPa] ' v = 220 kPa
[MPa]
' v = 110 kPa
20 ' v = 110 kPa
15
' v = 55 kPa
' v = 55 kPa
10 15
' v = 28 kPa
' v = 28 kPa
5 10
0,0 0,5 1,0 1,5 2,0 0,0 0,5 1,0 1,5 2,0
Fiber content [%] Fiber content [%]
Figure 5. Maximum shear modulus versus fiber content, for different vertical pressures. Comparison
between test results and predictive model; a) sand in a loose state; b) sand in a dense state.
Figure 5 shows the effect of fiber content on the maximum shear modulus, for vertical
stresses ranging from 28 kPa to 440 kPa, corresponding to the sand in a loose state (5a)
and the sand in a dense state (5b).
From these figures it can be seen that inclusion of fibers tends to reduce the initial
stiffness of the reinforced soil at low strain levels. Also, a good agreement is observed
between the trends of the experimental data and the model predictions.
4. Conclusions
A physical model based on Hertz theory is presented in order to explain and justify the
maximum shear modulus drop as synthetic fibers are added to a fiber reinforced sand.
The mathematical model assumes that shear wave velocity and maximum shear
modulus of the reinforced sand depends mainly on the stiffness of the grain to grain
and fiber to grain contacts.
The model predictions are compared to experimental results obtained by means of
bender element measurements in an alluvial clean silica sand reinforced with
polypropylene fibers tested in confined compression state. The model predictions fit
very well with the laboratory measurements.
The proposed model allow concluding that the maximum shear modulus of the
reinforced sand decreases as the fiber content increases because of the drop of stiffness
at particle contact level when fibers are added to the soil mass.
The ratio between the amount of grain to fiber contacts and the total of contacts on
the shear wave path controls the maximum shear modulus value.
References
[1] D. H. Gray, H. Ohashi, Mechanics of Fiber Reinforcement in Sand, Journal of Geotechnical Engineering
109, No. 3 (1983), 335-353.
[2] M. H. Maher, D. H. Gray, Static Response of Sands Reinforced with Randomly Distributed Fibers,
Journal of Geotechnical Engineering 116, No. 11 (1990), 1661-1677.
[3] R. L. Michalowski, A. Zhao, Failure of Fiber-Reinforced Granular Soils, Journal of Geotechnical
Engineering 122, No. 3 (1996), 226-234.
[4] J. G. Zornberg, Discrete framework for limit equilibrium analysis of fibre-reinforced soil, Gotechnique
52, No. 8 (2002), 593-604.
[5] N. C. Consoli, K. S. Heineck, M. D. T. Casagrande, M. R. Coop, Shear Strength Behavior of Fiber-
Reinforced Sand Considering Triaxial Tests under Distinct Stress Paths, Journal of Geotechnical and
Geoenvironmental Enginnering 133, No. 11 (2007), 1466-1469.
[6] E. Ibraim, A. Diambra, D. Muir Wood, A. R. Russell, Static Liquefaction of fibre reinforced sand under
monotonic loading, Geotextiles and Geomembranes 28 (2010), 374-385.
[7] K. S. Heineck, M. R. Coop, N. C. Consoli, Effect of Microreinforcement of Soils from Very Small to
Large Shear Strains, Journal of Geotechnical and Geoenvironmental Enginnering 131, No. 8 (2005),
1024-1033.
[8] A. Diambra, E. Ibraim, D. Muir Wood, A. R. Russell, Fibre reinforced sands: Experiments and modeling,
Geotextiles and Geomembranes 28 (2010), 238-250.
[9] R. L. Michalowski, J. Cermk, Triaxial Compression of Sand Reinforced with Fibers, Journal of
Geotechnical and Geoenvironmental Enginnering 129, No. 2 (2003), 125-136.
[10] V. A. Rinaldi, J. J. Clari, Low strain dynamic behavior of a collapsible soil, Proceedings of the XI
Panamerican Congress of Soil Mechanics and Geotechnical Engineering, Foz de Iguaz, Brasil (1999),
Vol. 2, 835-841.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-761
Calibration of rotational hardening model

for clays based on dilatancy
Georgios BELOKASa, 1 and Yannis F. DAFALIASb
a
Technological Educational Institute of Athens
b
National Technical University of Athens and
University of California at Davis
Abstract. Rotational hardening (RH) is a specific characteristic of clay plasticity

constitutive modeling describing the rotation of the yield surface (YS) in stress
space during plastic loading. RH reflects macroscopically the anisotropic fabric
evolution at the microstructural level of particle mechanics. For calibration
purposes of RH the determination of the evolving YS in stress space under fixed
stress ratio loading, is a tedious and elusive process. On the other hand, the
execution of radial stress path tests (e.g. isotropic and 1D compression) while
measuring the corresponding stressdilatancy relationship, is feasible even in
common commercial laboratory practice. This paper presents a methodology for
the calibration of the RH based on such simple laboratory tests and on an empirical
stress dilatancy relationship.
Keywords. Clay, dilatancy, plasticity, rotational hardening, yield surface
1. Introduction
Rotational hardening (RH) is a specific characteristic of clay plasticity constitutive

modeling describing the rotation of the yield surface (YS) in stress space during plastic
loading. RH reflects macroscopically the anisotropic fabric evolution at the
microstructural level of particle mechanics. The analytical description of RH and the
calibration of the corresponding parameters must satisfy several necessary conditions
within critical state soil mechanics [1], [2]. In particular, the calibration of the RH
parameters is not a trivial procedure and the scope of this work is to present an
effective way to accomplish it.
Previously, Wheeler et al ([3]) observed from their experimental data that under
fixed stress ratio =q/p stress path loading, the normalized orientation /M of the YS in
triaxial p-q space ( measures the orientation and M the critical slope) is related to the
normalized stress ratio /, based on which they proposed a specific RH rule. To
accomplish this it was necessary to fit the analytical expression of YS proposed by
Dafalias ([4]) to measured elusive yield points under various stress ratio tests. Dafalias
& Taiebat ([1], [2]), referring to these data, presented other RH rules that were
modifications of the original RH rule by Dafalias ([4]) with the additional property that
the maximum /M was restricted for very large /, as opposed to the unbounded
values of /M in previous RH propositions. Within this concept an attractor/bounding
1
762 G. Belokas and Y.F. Dafalias / Calibration of Rotational Hardening Model for Clays
quantity b()=/x(), function of , of the yield surface orientation, , is incorporated

for radial stress paths, which
varies from b(=0)=0 to b(=M)=c=
0[=M).
The foregoing experimental procedure of Wheeler et al, while insightful, is also
tedious, as it requires the determination of elusive YS points in stress space after fixed
stress-ratio loading. On the other hand, measuring the stress-dilatancy relationship
during the execution of fixed stress-ratio path tests (e.g. isotropic and 1D compression)
is feasible even in common commercial laboratory practice.
This work explores the application of such a stressdilatancy relationship to the
calibration of the RH, provided the RH rule and elasticity are known. Based on such
knowledge, the total dilatancy is back calculated and compared to experimental results
and to an existing empirical relationship of Belokas & Kavvadas ([5]).
2. Radial Stress Path Behavior
It is well established that the consolidation line on the v p plane, the so called
Intrinsic Compression Line (ICL, [6]), from triaxial radial stress paths depends on the
consolidation stress ratio (=q/p). The ICL may be described by Eq. (1), while the
elastic rebound lines by Eq. (2) (Figure 1).
v = N lnp Intrinsic Compression Line) (1)
v = v lnp/p (rebound line) (2)
where v=1+e is the specific volume, and are compressibility constants, N a

constant dependent on (e.g. [5]) that controls the position of the ICL in v-lnp plane
and the p, v the consolidation pressure and the corresponding specific volume on the
current ICL. Usually, the Critical State Line (CSL) is taken to be a lower limit of ICL
lines.
q
ICC1: Isotropic Intrinsic
N1=Niso
Critical State, =M Compression Line (ICLiso)
5 N2
4
3 v N3

2 4 N4 1
Isotropic ICL2
N5=
compression 2
3 ICL3
3 ICL4
2 4
ICL5: Critical
5
1 State Line (CSL)
p lnp
1kPa p p
Figure 1. Intrinsic compressibility of structureless soils ([5]).
Any random intrinsic (or structureless) state (v, p) may be linked to a reference
state on the ICL (vo, po) on the ICLiso along a - rebound line (Figure 1). By setting for
the ICLs v = N lnp and vo = Niso lnpo and for the rebound line v = vo lnp/po
we get Eq. (3), which describes the distance between the various Intrinsic Compression
G. Belokas and Y.F. Dafalias / Calibration of Rotational Hardening Model for Clays 763
Lines. Concerning N Belokas & Kavvadas ([5]) have proposed an empirical

relationship for N, which takes into account the normalized ratio /.
p # N N iso $
exp % & (3)
po ' 0 1 (
3. Basic model formulation
The basic formulation of the Dafalias and Taiebat ([1], [2]) model includes the yield
surface (YS, Eq. (4), Figure 2), the plastic potential surface (PPS, Eq. (5), Figure 2), the
rotational hardening ( 2 , Eqs (6) and (7)), the poro-elasticity (volumetric strain
increment, Eq. (8)), and the deviatoric strain increment (Eq. (10)).
Figure 2. Yield and plastic potential surfaces.
F q p2
N 2 2 2
p p3 p
0
2
(4)
g q p2
: q p2
M 2 2 2
p p3 p
0 (5)
p
2 L cpat 2 b
2 (6)
po
2 b
/ x
(7)
ve p / K 1 vin p
p , qe q / 3G (8)
where L is the plastic loading index, represents the rotation of the YS and PS
major axis, po represents the size of the YS and PPS, M is the critical slope, N is a
constant of the YS, b() is the attractor/bounding value of for a given with which
converges under a fixed stress ratio path =ct, K the bulk modulus and G the shear
modulus.
The YS may be rewritten in the form of Eq. (9), which, by setting p=p and =b
for states on the ICL of a random consolidation stress ratio (), gives the distance of
ICL from the isotropic ICLiso. Solving with respect to N we get Eq. (10). By knowing
any random ratio (p/po), and the corresponding anisotropy b() we can determine
constant N.
p po N 2 2 2
N 2
22 2
(9)
N 2 2 p po
2 22
1 p po
(10)
4. Total and plastic dilatancy for radial stress paths
The PPS g=0 of Eq.(5) gives the plastic dilatancy . Under a normally consolidated
fixed stress ratio (=ct) path loading, where b=, is given by Eq. (11). Eq. (11) can
be solved for b as shown in Eq. (12), or based on Eq.(7) for variable x() as in Eq. (13).
" vp qp M 2 2

2 2 b
(11)
2 b M 2 2
2"
4 2 b M M M 1 M
2"
2
(12)

x( ) 2" 2" M 2

2
(13)
Eq. (12) can be used to calibrate any given function b() by measuring and
examine its applicability. Herein, this will be examined for the simplified case where
the contribution of elastic strains in ignored, i.e. total dilatancy t.. For the Ko
conditions, where =, b()=b=/x() and "=3/2, Eq. (12) gives
b=(32+2)/3. For an expression of b accounting for elastic volumetric
strains, the reader is referred to [4] and [1]. Concerning the calibration of b, the simple
linear evolution law of Eq. (14) from Dafalias ([4]) that assumes a fixed x not function
of , will be examined.
2 b
/ x( ) / x (14)
Eq.(14) implies that the various x calculated from Eq.(13) by measuring at

various constant stress path, must be same, which can be only approximately true.
5. Calibration with respect to the total dilatancy data
Eq. (12) exhibits that, under a fixed stress ratio path, the attractor b of the rotational
hardening plasticity rule (Eq. (7)) depends on and the dilatancy . Therefore, the
experimentally measured may serve the purpose of calibrating the b function
constants and the constant N, which will be portrayed in the following for the Lower
Cromer Till experimental results as presented by Gens ([7]).
The constant and variable parameters that describe LCT behavior under radial
stress path behavior are summarized in Tables 1 and 2. Concerning Table 2,
parameters , /, t and N (or p/po) have been computed/interpreted directly from
the experimental results and are used next on the calibration procedure (column b/M is
not determined experimentally and is explained later). For the following analyses,
Ko=0.754 and NKo=1.775 is adopted, which were determined by a linear interpolation
to match Ko=1.5.
Table 1. Reworked constants for LCT according to Gens ([7]) data.
Niso v
0.066 0.009 1.20 1.788 1.736 0.24
Table 2. ICL for LCT according to Gens ([7]) data..
/M t N p/po b/M *
0 0 5 1.788 (=Niso) 1.000 -
0.21 0.175 6.199 1.786 0.965 0.081
0.42 0.350 2.951 1.783 0.916 0.172
0.74 0.620 1.557 1.775 0.796 0.379
1.00 0.833 0.813 1.765 0.668 0.608
1.20 1.000 0.000 0.400 -
*
This column is determined indirectly by application of Eq. (14)
Belokas & Kavvadas ([5]) have proposed an empirical equation for the total
dilatancy given by Eq. (15), in which for Ko=0.754, M=1.2 and Ko=1.5 we get
z=1.4846 and Figure 3a. Eq. (15) fits reasonably well the experimental data, by only
fitting the Ko conditions. For the case that there is very few available experimental data
for 0 < / < 1, Eq. (15) could be used for the calibration of b.
Having measured experimentally the vs (e.g. Table 2), the corresponding b/M
ratios are indirectly calculated (i.e. not experimentally calibrated) by applying Eq. (12),
which for the LCT result in the values of Table 2 and the b/M vs /M plot of Figure 3b.
Note that each b/M, /M pair corresponds to a specific experimentally determined ICL
described by Eqs (1) and (3) and presented in Table 2. In Figure 3b, Eq. (16) is plotted,
which results from application of the empirical Eq. (15) into Eq. (12). Eq (16) fails to
predict a saturation of b/0/M for /M=1 as observed by the experimental data of
Wheeler et al ([3]), as Eq. 15 does not inherently include the evolution of b.
M 1 e z /" 4 " 1 z
ln 1 M
(15)
2 1 z
ln 1 M
M 2 2
/ 2 (16)
(a) (b)
Figure 3. Application of Eq. (17) to: a) fit the LCT directly from experimental data and b) to predict b/M
t
within the rotational hardening plasticity.

So far discrete b() values have been determined to fit the experimentally
determined /M values (one could also use the empirical Eq. (17)). Constant x can
be chosen to give the best fit of the / data or curve by means of Eq.(15)
assuming x is not function of .
It is convenient to calibrate x for the Ko (or =) conditions, where =Ko\=1.5,
which gives x=1.627 that for = gives Ko=0.464 or Ko/M=0.387 and c=0.738 or
c/M=0.615. This solution is plotted in Figure 4.
(a) (b)
Figure 4. Application of: a) Eq. (25) for x=1.552 and b) Eq. (14) for given by Eq. (25).
Having computed c, constant N can then be calibrated from the critical state,
where pcs/po=0.400 (see Table 2) and Eq. (12) leads to N=0.829 or N/M=0.691. Eq. (11)
results in p.R/po=0.8843, greater than p/po=0.754 (or equivalently =2.795,
greater than =2.74), which means that there is not a complete match for the Intrinsic
Compression Lines. The N=0.829 is very close to the value used by Dafalias & Taiebat
(2013), which was at that time calibrated to match undrained stress path behavior. It is

!$?\^
_!?`!
$???!!_t) data
and the computed from these /M, /M pairs, it fails to predict the critical state value
c/M, as this empirical equation was derived based only on t and without taking into
account the shape and orientation of the yield surface.
6. Conclusions
An alternative procedure on the RH calibration was presented, by exploiting the

experimentally recorded total dilatancy (t) stress ratio () relationship and the
corresponding normal compression lines, which come from conventional radial stress
path consolidation tests that are typical to perform. It is based on the fitting of the
resulting from the RH stress dilatancy relationship into the experimental (discrete
points) or the empirical (continuous equation) stress dilatancy, under the
approximation that total dilatancy equals plastic dilatancy. This procedure was directly
demonstrated on the Dafalias ([4]) b() function for the LCT experimental data. The
least possible information required is the isotropic compression, one random radial
compression (e.g. 1D compression) and the subsequent undrained shearing.
The constants of the b function are calibrated based solely on t /M p/po data.
This procedure includes an indirect determination of b/ versus /M values by the
application of Eq. (12). The proposed methodology requires the execution of common
commercial laboratory tests. Therefore, it is advantageous over the direct calibration of
the RH rules, such the one presented by Wheeler et al ([3]), which requires the
performance of an elaborate set of tests, a procedure difficult to follow for conventional
laboratory practice.
Acknowledgements
The research leading to these results has received funding from the European
Research Council under the European Union's Seventh Framework Program FP7-ERC-
IDEAS Advanced Grant Agreement n 290963 (SOMEF).
References
[1] Y. F Dafalias and M. Taiebat, Anatomy of rotational hardening in clay plasticity, Gotechnique 63.16
(2013): 1406-1418.
[2] Y. F Dafalias and M. Taiebat, Rotational hardening with and without anisotropic fabric at critical state,
Gotechnique, 64.6 (2014): 507 511.
[3] S.J. Wheeler, A. Ntnen, M. Karstunen & M. Lojander, An anisotropic elastoplastic model for soft
clays, Canadian Geotechnical Journal 40.2 (2003): 403-418.
[4] Y.F. Dafalias, An anisotropic critical state soil plasticity model, Mechanics Research Communications
13.6 (1986), 341-347
[5] G. Belokas and M. Kavvadas, An intrinsic compressibility framework for clayey soils, Geotechnical and
Geological Engineering 29.5 (2011): 855-871.
[6] J.B. Burland, On the compressibility and shear strength of natural clays, Geotechnique 40.3 (1990): 329-
378.
[7] A. Gens, Stress strain characteristics of a low plasticity clay, PhD, Imperial College of Science,
Medicine and Technology, University of London, 1982.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-768
The theory of granular packings and the

strength of coarse soils
CalixtroYANQUI1
San Agustin National University of Arequipa
Abstract.The various and careful experiments performed to find the shear

strength have shown that there is a value called "critical void ratio" for each
granular soil, toward which it tends, being dense or loose. In this paper, the soil
assemblage is represented by an ideal packing of spheres, which inherently admits
two states, one dilative and other contractive. This duality and the Casagrandes
diagram that relates the void ratio with the shear strength are the basis to establish
the parameters of the packing at the critical state, as well as the physically possible
packings for a dense and a loose soil. Furthermore, in regards to the Trollopes
hypothesis of the centroidal reactions, and recent experiments, chains of forces of
contact between grains are used to describe the transmission of stresses in soils at
dense state. Under this hypothesis, the kinematics and the statics of the dilative
packings yield the fundamental relationship between the initial void ratio and the
peak angle of friction. On the other hand, in soils at loose state, the deformation is
contractive, and occurs by a sliding of the grains, and stresses are transmitted
through the Rowes shear mechanism, following the laws of the slip lines theory.
So that, for this state, the angle of friction is constant and equals the critical
friction angle. A good correspondence between the results of the theory and the
reported experimental data is shown.
Keywords.Packings, critical state, shear strength, dilatancy, contractancy.
1. Introduction
Granular matter does not meet the traditional classification of matter, because, under
certain conditions, it may behave as a solid, in others, as a liquid, and even in others, as
a gas [1]. This is due to the fact that this substance is made up of a large number of
individual grains, whose mainfeature is the contact of each of them with the
neighboring grains. This feature and the spheroidal shape of the grainsgiveto this
substance a set of special characteristics such as: duality, or the existence of two
configurations for the same volume;dilatancy, or the volume change during the
distortion of the system; multiplicity, or the change of the configuration of the contacts
by the change ofthe kind of stress; and hyperbolicity, or transmission of the normal
stress along chains of grains. Since ordered packings of spherical grains exhibit also
these properties, they are used in this paper to describe the especial features of the
behavior of sands and other coarse soils at failure; for instance, the Casagrandes
critical state, the dilatancy under shear stress, and the constancy of the critical shear
strength, among others.This model has the following advantages over the continuum
1
Cooperativa Universitaria, B-11, Arequipa, Per. E-mail: [email protected]
C. Yanqui / The Theory of Granular Packings and the Strength of Coarse Soils 769
model: a) the actual soil grains are better represented by spheres, b) the analysis is
simple and straightforward, and c) the results fit well with the experimental data.
2. Behavior of sands under shear stresses
The numerous experimental evidence provided since the beginning of modern soil
mechanics has led to the conclusion that the stress-strain diagrams of shear tests made
in granular soils, particularly in the sands, radically change according to the pore
volume. Loose sands are related to a monotonic curve, in which a gradual increase in
shear stresses, , gives rise to the increment of the shear strain, !, until an
approximately constant shear stress is attained,cs (Fig. 1a). They become denser until
a constant void ratio is reached, ecs (Fig.1b); so that, they are called contractive. Dense
sands show a convex curve, with an initial branch characterized by a small increase in
shear strain, !, generated by a rapid growth of shear stress, and a final concave upward
branch, which tends towards a constant value of shear stress, cs. They become looser
until a constant void ratio is attained, ecs; so that, they are called dilative (Fig.1b). The
quantities csand ecs, common to loose and dense sands, are the shear strength and the
void ratio at the critical state. This becomes clear in the diagram e-, (Fig. 3.c2),
highlighted by Casagrande[2] and Hirschfeld[3].
Likewise, Taylor [4] noted that, with the void ratio, not only changes the stress-
strain curve but also the mode of the specimen deformation. Loose sand samples bulges
symmetrically at large deformations. No localized failure planes are observed. Dense
sand samples, tested under similar conditions, show an inclined plane of failure, called
shear band and associated to the decrease of the deviatory stress.
0.8 / -0.6
Shortenig of sample height (mm) .
0.7
cs/ -0.4 Expansion
0.6 Loose sand
0.5 -0.2
Dense sand
0.4 0.0
Loose sand
0.3 Dense sand 0.2
0.2
Compression
0.1 0.4
0.0 0.6
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
Horizontal displacement (mm) Horizontal displacement (mm)
Figure. 1. Strain controlled test in the direct shear apparatus on sand. Initial void ratios were 0.63 for the
dense state, and 0.74 forthe loose state[5].
3. Ordered granular packings
An ordered granular packing is a collection of spheres whose centers are the nodes of a
crystallographic cell. Among the fourteen Bravais cells, the rhombohedron is the best
suited basic cell for describing a deforming granular packing, because the twelve edges
of this polyhedron are of the same length, equal to the diameter of contact between the
770 C. Yanqui / The Theory of Granular Packings and the Strength of Coarse Soils
spheres, D (Fig. 2). Regarding the principle of multiplicity of granular matter, the
spatial orientation adopted by the rhombohedron depends on the kinematic conditions
imposed by the applied stress. However, as long as the geometric description of this
polyhedron may be cumbersome, the cell of the triclinic system is better used. In this
case, as is known, the cell is a parallelepiped with different edge lengths, a, b and c, and
two characteristic angles: the angle of the base,, and the slanting angle,", the values of
which are chosen according to the kind of test performed to determine the soil strength.
Moreover, the plane on which is measured the angle " defines the packing dimension:
the coordinate plane x-z, and the diagonal plane, #-z, for the two and three-dimensional
packings, respectively.
Figure 2.Packing of spheres and its polyhedral representation: a) The rhombohedron cell, and b) The
parallelepiped cell.
3.1. Vertical compression cell
The kinematics of a parallelepiped that undergoes a three-dimensional vertical

compression demands that each sphere of the upper layer be incrusted in the pore
formed by three spheres of the lower layer. Statically, the upper sphere should be in
contact with the three lower spheres, along the edges of a tetrahedron, which, together
with the inverted similar tetrahedron, forms the basic parallelepiped. If the angle of the
slanting edge is , then =, and c = D. The other geometric elements are determined
by knowing that non-horizontal edges of the tetrahedron have a length D. With them,
the void ratio of the packing is found to be [6]:
12
e= sin (1 + cos ) sin 2 cos 1 (1)

where, by symmetry, the angle can only take values of 60 and 90, resulting in the
so called tetrahedral and octahedral packings, respectively. The parameter is the grain
factor that represents the connection between the packing of spheres and the actual soil
assemblage.
3.2. Horizontal shearing cell
The kinematics of a cell that deforms by horizontal shearing requires that one face of
the basic rhombohedron be horizontal. This means that the rhombohedron cell coincide
with the parallelepiped cell. In this case, the angle of the edge respect to the vertical has
different meaning, so that, it will be denoted by. The edges of the parallelepiped are
defined by a=b=c=D, and the angles, by = , and . Then, the void ratio is obtained
from the expression, valid for two- or three-dimensional packings:
6
e= sin cos 1 (2)

4. Granular kinematics and critical state mapping
For illustration, the two-dimensional prismatic packing submitted to a horizontal shear

stress is described[7]. Figure 3 shows that, for the same void ratio, two cells exist: one,
whose tangential displacement due to a horizontal shearinggoes upwards, causing the
dilation of the packing (Fig.3a); and another, in which, for the same shearing, the
tangential displacement goes downwards, resulting in a contraction of the packing
(Fig.3b). Obviously, the value that separates one kind of cell from the other is $ = 0
(Fig.3.c3). Therefore, if the angle $ is positive, the packing is dilative, and, if the angle
$ is negative, the packing is contractive. Several authors have tried to use this
simplistic model to describe the soil behavior during the shear test assuming
reasonably that friction at critical statecorresponds to the loosest state,when $= 0. But
the conclusions drawn from this assumption are in clear contradiction with the
experimental data. This is due to two shortcomings: first, the critical condition is not
related to the loosest packing, in which $=0, and second, the equilibrium used to
determine the peak friction angle is hypostatic, because, in this model, the vertical
shear stress that comes from the law of reciprocity is not taken into account.

Figure 3.A two-dimensional model for the horizontal shearing ofgranular soils: a) dilative packing, b)
contractive packing, c) mapping of diagrams: c1)stress-strain curves, c2) Casagrandes stress-void ratio
diagram, [2] and [3], c3) angle of packing-void ratio diagram, and c4) arcs of contact in a spheroidal grain.
5. Granular statics of the drained triaxial compression test
By virtue of their inherent multiplicity, packings change the configuration of contacts

between grains according to the kind of the applied stress. This fact allows the ordered
packing to be chosen in such a way that it is the best representation of the granular
kinematics of the test. Obviously, the three-dimensional vertical compression cell is
well suited for describing the axisymmetric compression test.
5.1. Dilative packing
Based on studies with photoelastic discs, it has been well established that stresses are
transmitted as chains of contact forces in granular media [8]. This other property of
random packings has originated the theory of the network of contact forces, which do
not necessarily coincide with the packing of grains. Particularly, if they do coincide, as
occurs in the at rest state, the Trollopes principle of the centroidal reactions [9] can
be applied to calculate the stresses in the soil mass. Under this assumption and
considering an axisymmetric packing, the application of the theory of the network of
contact forces to the problem of an infinite soil slope submitted to the soil self-weight
yields the following relationship between the angle of packing, , and the angle of
drained internal friction, [10]:
n 1
tan 2 = (3)
1 + n tan 2
wherenstands for the dimension of the packing.Particularly, for a three-dimensional

packing, eliminating from equations (1) and (3),the relationship between the initial
void ratio, e, and the angle of peak internal friction,, shows up. For the integral
average of the void ratio over the angle , and the three-dimensional packing, n=3, it is
obtained [7]:
10 3
e= cos 2 3 2 cos 2 1 (4)
2
5.2. Contractive packing
Differentiating the equation (1) with respect to and equating to zero the
discriminative value is found to be: 0 = 54.74. The packing is dilative if <0, and it
is contractive if >0. In the latter case, due to the assumption that the network of
contact forces and the packing of grains coincide, the lateral stress is greater than the
vertical stress, which does not agree with the experimental data. This observation leads
to the conclusion that, in a contractive packing, the tangential component of the contact
force plays an important role, and therefore, the deformation mechanism occurs by
sliding in each plane of contact between two grains, without reaching to develop a
single plane of failure; in agreement with the pioneering experimental observations of
Taylor [4]. As such, the force of contact acts in the tangent plane between two spheres,
which makes an angle with the horizontal, rather than with the vertical. Statics, under
these considerations, leads to the conclusion that stresses in a contractive packing do
not depend on the number of contacts of the grain; on the kind of granular packing, on
the number of coordinates, and on the intermediate principal stress. This may be
considered as the proof of one of the hypothesis on which the Mohr-Coulomb law is
based. Moreover, if the deviation of the contact force is denoted by, the active
coefficient of lateral pressure is K=cottan(-) and decreases as increases. As the
test goes further, reaches the value of cs at the moment of the failure. It is easy to
show that, for =45+cs/2, the value of K is a minimum, in agreement with the
derivation of Rowe [11], is independent of the initial void ratio, and is given by:
3
= K = tan 2 (45 cs ) (5)
1 2
6. Granular statics of the drained horizontal shearing test
6.1. Dilative packing
The statics of a dilative shearing packing is based on three conditions: first, stresses are
transmitted as chains of contact forces, second, they induce a plane strain state, third,
the imposed kinematics by the shear box necessitates of a horizontal failure plane.
Therefore, in the basic prismatic cell, one force chain is horizontal and the other force
chain is inclined at an angle with respect to the vertical. Then, the equilibrium with
the external applied forces is accomplished by transforming the resulting stresses acting
on each face of the prismatic cell into the principal stresses acting along the
corresponding diagonals of the equivalent rhombic cell, the angle of which is BAC in
the figure 3a. Denoting this angle by 2, the interesting relationship = 45-/2 shows
up, which.Consequently, the proper combination of this relationship with equations (2)
and (3) for n=2, yields the relationship between the initial void ratio and the peak
friction angle, , for =60:
3 3
e= cos 2 cos 2 1 (6)

6.2. Contractive packing
In recognition of the relationship between and, all arguments described for triaxial
compression testing are valid for direct shear test, settling down the relationship:
cs=cs, and concluding that the angle of friction at the critical state is independent of
the initial void ratio.
7. Experimental verification
One of the most important parameters of the granular soils mechanics is the angle of
friction. So that, engineers have put considerable effort in the study of the global
factors influencing this parameter, such as porosity, grain shape and angularity,
mineralogical composition, and history of sedimentation, among others. However, both
the experiments and the theory of granular packings have shown that most of these
factors are includedwithin the void ratio. Because of this, a way to confirm the present
theory of the critical state is the experimental verification of the relationship between
the angle of internal friction and the initial void ratio. Fortunately, there exists a great
deal of experimental data, reported by several authors everywhere.
44 44
Steel balls Limassol sand [17]
Glass balls-dry-3mm 42 Guinea sand [17]
Glass balls-dry-1.0 mm Dilatant packing
Angle of friction .
40 Glass balls-wet-3.0mm 40 Contractive packing
Friction angle ()
Glass balls-wet-1.0mm
Dilatant packing
Contractive packing 38
36
36
32 34
32
28
30
0.45 0.5 0.55 0.6 0.65 0.7 0.3 0.4 0.5 0.6 0.7
Scaled initial void ratio Scaled initial void ratio
44 46
Brasted Sand [17] Portland sand [17]
Mersey river sand [15] Southport sand [17]
42 Medium fine sand [11] 44 Olivine sand [19]
Ottawa sand [19] Chattahoochee sand [16]
40 Chattahoochee sand [20]
Friction angle ()
42 Franklin Falls Dam [13]

Friction angle ()
Sandy gravel [21] Germany sands [21]

Dilatant packing
38 Contractive packing 40 Monterey sand #20 [18]
Dilatant packing
Contractive packing
36 38
34 36
32 34
30 32
0.3 0.4 0.5 0.6 0.7 0.2 0.3 0.4 0.5 0.6 0.7
Scaled initial void ratio Scaled initial void ratio
Figure 4.Relationship between the initial void ratio and the peak friction angle: a) direct shear teston
disordered packings of uniform spheres [12]; b), c), and d) triaxial compression test on granular soils.
7.1. Disordered packings of uniform spheres
Skinner [12] reported to Geotechnique the experimental results on the influence of the
interparticle friction on the shear strength of random assemblages of spherical particles.
The author describes a series of strength tests in direct shear apparatus on spheres of
different diameter made of glass, steel and lead, to determine the interparticle friction
angle as well as the angle of friction of the random packing of these spheres under dry
and wet conditions. In figure 4a, the theoretical curve (6) and the experimental points,
except for the lead, that exhibits a strong dispersion, are shown.
7.2. Granular soils
By using differentkind of equipment for triaxial compression testing on granular soil

samples, several authors have reported the experimental results relating the peak
friction angle and the initial void ratio. In figures 4b, 4c and 4d, the graphs of equations
(4) and (5) are compared with these experimental data, grouped according to the
angularity and particle size distribution of the grains. In general, a good correlation can
be seen between the theory and the experiments.
8. Conclusion
The granular mechanics of ordered systems is a powerful tool in studying the shear
strength of coarse soils. The critical condition is related to a particular value of the
angle of packing,whichdoes not correspond to its loosest state. In a dense soil,
inherently dilative, the stresses are transmitted as chains of contact forces. In a loose
soil, inherently contractive, the transmission of stress occurs as it were a continuous
medium. The relationship between the initial void ratio and friction angle, obtained by
means of the granular mechanics, fits well with the experimental data reported by
several authors.
References
[1] J.H. Snoeijer, T.J.H. Vlugt, M. van Hecke&W. van Saarloos:Force network ensemble: a new approach to
static granular matter. Physics Review Letters. 92, 054302, (2004).
[2] A. Casagrande:Characteristics of cohesionless soils affecting the stability of slopes and earth fills. J.
Boston Soc. Civil Engs. Vol. 23: 13-32. 1936.
[3] R.C. Hirschfeld:Stress-deformation and strength characteristics of soils. PhD. Thesis. Harvard
University. 1963.
[4] D.W. Taylor:Fundamentals of Soil Mechanics. John Wiley & Sons: New York. 1948.
[5] J.A. Jimenez&J.L.de Justo:Geotecnia y Cimientos, tomo I. Editorial Rueda. Madrid. 1975.
[6] C. Yanqui:The generalizedtheory of granular packings as a model for studying the Portland cement and
concrete behaviour. 6th powders and grains. AIP. Vol. 1145, (2009), 215-218.
[7] C. Yanqui: Granular mechanics of the critical state of coarse soils. 7th Powders and Grains, AIP.
1542,(2013), 197-200.
[8] T.S. Majmudar& R.P. Behringer. Contact forces measurements end stress-induced anisotropy in granular
materials. Nature 435, (2005), 1079-1082.
[9] D.H. Trollope. The Stability of Wedges of Granular Materials. Ph. D. Thesis. University of Melbourne.
1956.
[10] C. Yanqui. Statics of gravitating discontinua. M.Sc. thesis, University of South Carolina. 1982.
[11] P.W. Rowe. The stress-dilatancy relation for static equilibrium of an assembly of particles in contact.
Proc. Royal Soc. A269. (1962), 500-527.
[12] A.E. Skinner. A note on the Influence of interparticle friction on the shearing strength of a random
assembly of spherical particles. Geotechnique, No. 19, (1969), 150-157.
[13] J.A. Jimnez Salas. Soil Mechanics and its Applications to Engineering. Editorial Dossat. Madrid. 1951.
[14] H. Leussink.& W. Wittke, Difference in triaxial and plain strain shear strength. ASTM, STP 361,
(1973), 77-89.
[15] P.W. Rowe. Technical paper. Geotechnique, vol. 19, No.1, (1969), 75-86. 1969.
[16] A.S. Vsic. Bearing capacity of deep foundations in sand. Highway Research Record No. 39. National
Academy of Sciences, (1963), 112-154.
[17] D.H. Cornforth. Prediction of drained strength of sands from relative density measurements. ASTM
STP 523, (1973), 281-303.
[18] N.D. Marachi, C.K. Chan, H.B. Seed, and J.M. Duncan.Strength and deformation characteristics of
rockfill materials. University of California. Report TE-69-5, 1969.
[19] I. Holubec& E. DAppolonia. Effect of particle shape on the engineering properties of granular
soils.ASTM STP 523, (1973), 304-318.
[20] M.M. Al-Hussaini. Influence of relative density on the strength and deformation of sand under plane
strain conditions. ASTM STP 523, (1973), 332-3478.
[21] K.J. Meltzer. Relative density: three examples of its use in research and practice. ASTM STP 523:
(1973), 463-477.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-776
Understanding the Disintegration of

Sensitive Clays using Remolding Energy
Vikas THAKURa,1 and Samson Abate DEGAGO b
a
Norwegian University of Science and Technology
b
Norwegian Public Roads Administration
Abstract. Sensitive clay materials are found in several areas of the world including
Alaska, Canada, Norway and Sweden. Studies by various researchers suggest that a
knowledge about the complete stress-strain curves help in the assessment of the flow
slide potential of sensitive clays. The post-peak stress-stain behaviour of sensitive
clays particularly indicates the disintegration process in the material as well as helps
in the estimation of the energy involved in the disintegration process. The energy
concept is a subject of current study by several researchers working on investigation
of flow slides on sensitive clays. Several terms such as degradation energy, strain
energy or remolding energy have been used to indicate the energy available for
disintegration of sensitive clays; this is referred to as remolding energy (RE) in this
work and is simply defined as the strain energy involved in the disintegration or
remolding of a material. A closer examination of the concept of RE provides an
understanding of the overall mechanical behavior of sensitive clays during flow
slides. In this paper, concept of RE analytically proposed by the authors is elaborated
in light of laboratory tests conducted to determine RE of sensitive clays.
Keywords. Remolding energy, sensitive clays
1. Introduction
Sensitive clays are often characterized using term sensitivity (St) which is a ratio between
the undrained shear strength (cu) measured on the intact (cui) and the remolded (cur)
sensitive clay using the fall cone method. Sensitive clays exhibit strain softening i.e. a
decrease in shear strength of the materials with increasing strain once the peak shear
strength is attained. Sensitive clays may disintegrate from their intact state to highly
viscous fluid when subjected to large strain. This is schematically illustrated in Figure 1
using a Norwegian sensitive clay sample. Standard triaxial tests provide stress-
deformation behavior up to an axial strain level of 10 to 20%. However, a complete
disintegration of sensitive clays may require very large strain. Ring shear tests, fall cone
test or reversal shear box test have been applied to attain a fully disintegrated state. Given
the simplicity, the remolded shear strength of sensitive clays is often measured using the
fall cone test. Sensitive clays are known for their potential for large landslides such as
flow slides which poses a serious risk to human lives, infrastructure, and surrounding
ecosystems within their reach. Several factors such as; erosion along rivers or canals,
human activities have been responsible in triggering such slides. For flow slides to occur
1
Corresponding Author: Professor Vikas Thakur, Geotechnical Engineering Division, Department of
Civil and Transport Engineering, Norwegian University of Science and Technology, Hogskoleringen 7A,
Trondheim, 7491, Norway; E-mail: [email protected].
V. Thakur and S.A. Degago / Understanding the Disintegration of Sensitive Clays 777
after an initial slide, it is important that at least the following two criteria are fulfilled [1-
8];
+ The slide debris should be sufficiently disintegrated or remolded.
+ The disintegrated slide debris should be able to flow out of the slide area.
There may be additional factors, such as the topography and the stability of the area
behind the initial slide zone. However, if the two criteria mentioned above are not
fulfilled, then vast landslides, such as those mentioned in the literature are less likely to
occur. In short, it is fair to say that the disintegration process and the behavior of sensitive
clays at fully disintegrated state are the key aspects of flow slides in such materials.
Figure 1. Schematic representation of a Norwegian sensitive clay sample subjected to strain from the intact to
the fully disintegrated state [9]
Several indicators of flow slide potential of sensitive clay slopes have been
suggested e.g. remolded shear strength (cur), the liquidity index (IL), the sensitivity (St),
the quickness (Q) or the stability number (Nc). Although these criteria are useful
indicators of potential for flow slides, whether a flow slide will actually occur or not
cannot be assessed using these geotechnical parameters individually. Another possibility
is to study the disintegration process of sensitive clays which involves an understanding
of complete stress-strain behavior. This can be done using the remolding energy concept
as explained in the following section.
2. On Evaluation of Remolding Energy
Since an early work by Bishop [10], the energy concept is a subject of study in relation
to the investigation of flow slides on sensitive clays e.g. [1-4], [6-8], and [10-15]. This
is referred to as remolding energy (RE) in this paper and is simply defined as the strain
energy involved in the disintegration or remolding of a sensitive clay.
Tavenas et al. [1] did some pioneering work to estimate the remolding energy of
Canadian sensitive clays using a different approach. They performed several laboratory
tests to show that there can be different processes by which a sensitive clay mass can be
disintegrated. Keeping the landslide challenges in mind, Tavenas et al. [1] investigated
778 V. Thakur and S.A. Degago / Understanding the Disintegration of Sensitive Clays
the disintegration process of seven different Canadian sensitive clays. In landslide the
main processes may be idealized to be shearing along with continuous straining and
displacement along a failure surface, squeezing and extrusion between relatively intact
clay blocks, impact of clay block on the bottom of the slide bowl or impact on clay blocks
from falling objects or soil. Tavenas et al. [1] attempted to reproduce all these processes
in the laboratory to study the disintegration of the undisturbed and high quality sensitive
clays collected from seven different historical landslide sites in Canada. A brief
description regarding the engineering characteristic of the tested clays is shown in
Table1.
Table 1. Engineering properties of the sensitive clays tested by Tavenas et al. [1]
Sample <2m w w L Ip IL Sali pc' cu cu cur St
Nr Site depth nity vane cone cone
m % % % % [-] g/l kPa kPa kPa kPa [-]
1 St.-Lon 9.3 70 70 68 42 1 - 180 43 49.5 2.1 24
2 St.Lon 4.8 76 75 73 46 1.1 2.1 100 31 30 1.1 27
3 Louiseville 6 77 75 73 46 1.1 7.6 125 37 39 1.3 30
4 St-Hilaire 5.6 85 73 60 35 1.3 0.4 115 35 35 0.8 44
5 St.-Hilaire 11.4 79 70 49 26 2 1 115 37 32 0.3 106
6 St.-Thuribe 6 43 50 42 20 1.6 0.5 175 39 55 0.4 137
7 St.-Thuribe 12 30 43 25 5 4 1.1 195 21 42 0.07 600
8 Mascouche 9 65 52 54 30 1.2 4.1 400 105 135 1.3 104
9 St.-Alban 6.6 58 66 41 19 2.4 0.2 85 20 21 0.2 105
10 St. Jean 30 56 36 35 12 1.1 1 1000 190 320 1.2 260
Vianney1
1
The tested sample was stored for 10 years and therefore not necessarily representative to the material involved
in the Sain-Jean-Vianney landslide. The fresh samples had higher water content, IL between 1.5 to 2.0 as well
as the lower values of cur. The results for this particular sample are highlighted using circles in Figures 2 and
5.
Tavenas et al. [1] performed series of tests to estimate the remolding energy and the
resulting degree of disintegration. The remolding energy was estimated based on the
mechanical energy imparted to the sample during the disintegration process. The degree
of disintegration was denoted by a term remolding index (Ir) which refers to the intact
and fully disintegrated (remolded) strength of the clays measured using the fall cone test.
Accordingly, the remolding index (Ir) was defined as;
wxQ y wx'

wxQ y wx
(1)
where cui and cur are the intact and remolded strength, respectively. cux is the strength
of partly disintegrated specimen. Tavenas et al. [1] also introduced a term called
normalized energy per unit volume (wN) in their study. This term refers to a ratio between
the remolding energy per unit volume and the energy required to achieve the limit state
which [16] estimated it for Champlain sea clay to be given by 0.013pc, where pc is the
pre-consolidation stress. Based on the earlier observation by Lebuis and Rissmann [6],
Tavenas et al. [1] also suggested that wN = 40 (or Ir 70%) as the threshold for the
occurrence of the large landslides in Canadian sensitive clays.
3. Discussion on Estimation of Remolding Energy
It is interesting to investigate how the index properties of the tested clay may play a role
in the disintegration process. Accordingly, Figure 2 indicates that increasing clay
content, salt content, liquid limit and the plasticity index is negatively correlated with the
measured remolding index (at wN = 40) measured using the simple shear tests on the
tested clays.
100 100
Remolding Index (Ir) at WN= 40 [%]

80 80
60 60
40 40
20 20
0 0
0 20 40 60 80 100 0 2 4 6 8
Clay content (<2m) [%] Salt content (IL) [g/l]
100 100
Remolding Index (Ir ) at W N= 40 [%]
80 80
60 60
40 40
20 20
0 0
0 20 40 60 80 0 20 40 60 80
Liquid limit (WL) [%] Plasticity Index (Ip) [%]
Figure 2. A relationship between the remolding Index (at wN =40) measured using the simple shear tests and
the index properties of the tested clays. The sample obtained from St. Jean Vianney site was stored for 10 years
and therefore not necessarily representative to the material involved in the landslide. The results for this
particular sample are highlighted using circles in this Figure and in Figure 5
Such behavior could be attributed to the fact that sensitive clays having wL < 40 %
and Ip < 10 % may have a little amount of clay content which attributes to a rather brittle
inter-particle bonds between the clay minerals. In return, it is much easier to remold such
low plastic brittle clays compare to high plastic clays which are usually rich in clay
content. Similarly, the tested clays with low salt content showed much larger degree of
disintegration because metastable behaviors of sensitive clays are closely linked with the
leaching of salt [e.g. 17].
The concept of remolding energy (RE) offers a valuable insight over the overall
mechanical behavior of sensitive clays during disintegration process. RE of a material is
estimated simply by calculating the area under the shear stress-strain curve of the
material, as shown in Figure 3. Tavenas et al. [1] adopted four different testing techniques
to study the disintegration. Accordingly, they were extrusion, free fall, impact and simple
shear. Even though one can guess the testing methodology simply by the name each of
these testing techniques, the authors refer to Tavenas et al. [1] for a detailed description.
Based on the laboratory testing by Tavenas et al. [1], a typical test results for Saint-
Thuribe (sample depth 12 m) is shown in Figure 3 (Left). Despite some scatter in the
measured data, the four different disintegration processes indicates the same trend i.e.
larger remolding energy is required to attain a higher degree of disintegration (higher Ir).
The disintegration process induced by the extrusion method required highest amount of
energy whereas the free fall method disintegrated the tested sample on a much lower
energy level. Tavenas et al. [1] suggest that the extrusion method may overestimate the
remolding energy due to the friction between the tested specimen and the apparatus;
while the free fall method may underestimate the same due to a non-uniform
disintegration of specimen. Tavenas et al. [1] recommend that the simple shear test was
best suited to investigate the disintegration process of sensitive clay samples.
Accordingly, simple shear tests were conducted to on samples collected from the all
seven locations. The simple shear test results also showed a great variation in their ability
to disintegrate and the amount of energy dissipated in the process of disintegration.
Figure 3(Right), illustrates this for the remolding energy and remolding index to achieve
wN = 40 for the tested clay samples.
Figure 3. Estimation of remolding energy [3]
Figure 4. (Left): Remolding Energy at different level of disintegration of the Saint-Thuribe specimen tested
using different techniques based on the laboratory observation by Tavenas et al. [1]. (Right): Remolding energy
and the remolding index to achieve wN = 40 for the tested specimen as shown on Table 1
4. A Pragmatic Approach for Estimation of Remolding Energy (RE)
An analytical solution is suggested to estimate RE for sensitive clays having linear elastic
hardening and plastic softening behavior as illustrated in Figure 4 [2-4]. Here, G is the
secant shear modulus, and is the shear strain, where the subscripts i and r represent the
peak and the residual strain levels, respectively. The average softening modulus is
represented by S. The analytical solution for the calculation of the required RE, ER, for
sensitive clays as proposed is as follows:

w|~ = = =
z{ w|~ /~ y ( y =* w|~

(2)
Thus, it is interesting to note that the RE for a material can be derived as a function of
cur, St, shear strain at the residual r and stiffness parameters (G and S). The equation also
implies that the required RE is less for sensitive clays with a lower cur and lower r.
100 100
80 80
60 60
40 40
20 20
0 0
0 1 2 3 0 1 2 3 4
Fall cone based intact shear strengh (cui ) [kPa] Liquidity Index (IL) [-]
100 100
Remolding Index (Ir) at WN = 40 [%]
80 80
60 60
40 40
20 20
0 0
10 100 1000 0 1 2 3 4
Sensitivity (St) [-] Remolded shear strength (cur) [kPa]
Figure 5. A relationship between the remolding Index at wN =40 and the relevant input soil parameters for the
proposed analytical solution
To appreciate the implication of RE estimation as presented in Equation 2 with respect

to different soil parameters involved, a detailed look at results of Table 1 is presented
and discussed here below. A relationship between the remolding Index (at wN = 40) and
the relevant input soil parameters for the proposed analytical solution is illustrated in
Figure 5. The trend shown by the laboratory testing confirm the analytical solution i.e.
larger sensitivity, lower intact and remolded shear strength, and higher liquidity index
were subjected to higher degree of disintegration level at wN= 40. It is worth mentioning
that the liquidity index and the remolded shear strength are inversely correlated with each
other [18, 19]. It must be noted that, due to the lack of information regarding the peak
and the residual shear strain of the tested clays it was not possible to calculate the
remolding energy using the analytical Eq. 2. Determination of the RE of sensitive clay
in the laboratory requires some special arrangements so that specimen can be deformed
to their residual strain level (r). This is not a straight forward task; thus, there is a lack
of a systematic laboratory and field studies on establishing this parameter. In order to
obtain these, sensitive clays samples must be subjected to very large strains. This has
been shown to be the case when performed a constant volume ring shear test was
performed on low sensitive Drammen plastic clay [20]. Strak and Eid [20] further
suggested that a complete residual (fully disintegrated) state may occur when a specimen
is sheared to several hundred millimeters corresponding to several hundred percent strain
inside the specimen. On other hand, based on the laboratory results by Tavenas et al. [1],
Leroueil [21] proposed the following pragmatic solution which requires only two
parameters;
ER = 16 cui Ip (3)
Earlier studies suggests that Eq. 2 and 3 are comparable and give reasonably well
approximation of RE despite several simplifications [2-4]. It is worth mentioning that
sensitive clays are notorious to sample disturbance. Therefore, laboratory derived
parameters are not always representative to in-situ condition. Keeping this in mind,
Thakur et al. [22] have recently proposed a methodology for an in-situ measurement of
the remolding energy.
5. Conclusion
This work advocates that a complete stress-strain behavior of soft sensitive clays
must be accounted in the assessment of disintegration of sensitive clays. Thus, a
simplified criterion based on the energy concept to assess the ease at which sensitive clay
may disintegration has been proposed. The laboratory observations on the Canadian
sensitive clays provide a good understanding over how the disintegration of these clays
were influenced by the index properties such as the liquid limit, plasticity index, the
intact and the remolded shear strength. These observations indicate that highly plastic
sensitive clays may need more energy to disintegration by the virtue of their ability to
sustain larger strain before they can be fully disintegrated. The proposed concept is in
line with the laboratory observations. This is an ongoing study and the authors will
continue to refine the proposed approach based on rigorous test measurements and other
methods to estimate the remolding energy.
6. Acknowledgement
The authors wish to acknowledge the support from the Norwegian national research
program Natural hazards: Infrastructure, Floods and Slides (NIFS). This program
(2012-2015) is initiated by the Norwegian Public Roads Administration, Norwegian
Water Resources and Energy Directorate and Norwegian National Railways
Administration.
References
[1] Tavenas F, Flon P, Lerouil S, J. Lebuis, Remolding energy and risk of slide retrogression in sensitive clays.
Symp. Slopes on Soft Clays, Linkping, Sweden, 1983.
[2] Thakur, V. & Degago, A. Disintegration energy of sensitive clays. Gotechnique Letters 3(2013), 2125.
[3] Thakur, V., Degago S, Oset, F., et al. Characterization of post-failure movements of landslides in soft
sensitive clays. Natural Hazards book: Advances in Natural and Technological Hazards Research, 91
104, 2014.
[4] Thakur, V., Degago S, Oset, F., Dolva B K, Identification of flow slide susceptible sensitive clay using the
remolding energy concept. The 14th International Conference on Advances in Computer Methods and
Advances in Geomechanics, Kytoto 1, (2014).
[5] Mitchell RJ, Markell AR, Flow slides in sensitive soils. Can. Geot. J. 11(1974),1131.
[6] Lebuis J, Rissmann P, Les coules argileuses dans le rgion de Qubec et de Shawinigan. In: Argiles
sensibles, pentes instables, mesures correctives et coules des rgions de Qubec et Shawinigan, Geo.
Association of Canada Guidebook, Canada, 1979.
[7] Karlsrud K, Aas G, Gregersen O, Can we predict landslide hazards in soft sensitive clays? Summary of
Norwegian practice and experience. NGI Publication 158, Oslo, 1985.
[8] Leroueil S, Locat J, Vaunat J et al, Geotechnical characterization of slope movements. Proceedings of the
7th International Symposium on Landslides (Senneset, K. (ed.)) 1(1996), 5374.
[9] Thakur V and Degago S, Quickness test approach for assessment of flow slide potentials Geotechnical
Engineering Journal of the SEAGS and AGSSEA: Physical Modelling in Geotechnical Engineering,
March Volume 45(2012), 4555.
[10] Bishop AW. Progressive failure - with special reference to the mechanism causing it. Proc. Geot. Conf.
on Shear Strength Properties of Natural Soils and Rocks, Oslo, 1(1967), 142150.
[11] Flon P, nergie de remaniement et rgression des coules dargiles. MSc thesis, Department
of Civil Engineering, Laval University, Qubec, 1982.
[12] Yong RN, Tang KY, 1983, Soil Remolding and Sensitivity Measurements. Geot.T.J. 6(1983), 2129.
[13] DElia B, Picarelli L, Leroueil S et al. Geotechnical characterization of slope movements in structurally
complex clay soils and stiff jointed clays. Rivista Italiana di Geotecnica XXXIII (1998),532.
[14] Hutchinson J N, Chalk flows from the coastal cliffs of northwest Europe, Catastrophic landslides: Effects,
occurrence, and mechanisms. Geol. Soc. of Ame. Reviews in Eng. Geo. 1(2002), 257302.
[15] Locat A, Leroueil S, Bernander S et al, Study of a lateral spread failure in an eastern Canada clay deposit
in relation with progressive failure: the Saint-Barnab-Nord slide. In: Proceedings of the 4th Canadian
conference on geohazards: from causes to management, 8996, 2008.
[16] Rosenqvist I (1955) Investigations in the clay-electrolyte-water system, Publication 9. Norwegian
Geotechnical Institute, Oslo, pp 107130.
[17] Leroueil S, Tavenas F, Le Bihan JP, Proprits caracteristiqus des argiles de Iest du Canada. Canadian
Geotechnical Journal 20(1983), 681705.
[18] Locat, J. & Demers, D, Viscosity, yield stress, remolded strength, and liquidity index relationships for
sensitive clays. Canadian Geotechnical Journal 25(1988), 799806.
[19] Tavenas F, Leroueil S, Creep and failure in slopes in clays. Can. Geo. J. 18(1981),06120.
[20] Strak TD, Eid HT, Drained residual strength of cohesive soils. J Geol Eng 120(1994), 856871.
[21] Leroueil S, Natural slopes and cuts: mov. and fail. mechanisms. Gotechnique 51(2001),197243.
[22] Thakur, V, Degago S A, Oset F, Gylland S A, and Sandven R, In-situ measurement of remolding energy
of sensitive clay. Submitted to Canadian Geotechnical Conference, GeoQuebec, 2015.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-784
Integration Scheme For A Thermo-Elasto-

Plastic Model With Unconventional Yield
Surfaces
Annan ZHOU a,1 and Yue ZHANGa
a
School of Civil, Environmental and Chemical Engineering, Royal Melbourne Institute
of Technology (RMIT), Melbourne, Vic 3001, Australia
Abstract. An adaptive substepping explicit integration scheme with a novel

loading-unloading decision method is developed here for the non-isothermal
unified hardening (UH) model. The non-isothermal UH model includes a convex
subloading surface in the p-q plane and a nonconvex subloading surface in the p-T
plane. Because of the convex/nonconvex subloading surfaces, the conventional
loading-unloading decision method used in stress integration schemes may lead to
incorrect elasticity/elastoplasticity judgments. In addition, the conventional
loading-unloading decision method is unable to determine the division point that
separates the elastic segment from the elastoplastic segment. A simple but robust
method, the Double Cosine (DC) method, is proposed in this paper to solve
loading-unloading decision problems. The proposed DC method is then embedded
into an adaptive substepping explicit integration scheme to implement the non-
isothermal UH model. The accuracy and efficiency of the DC method are
discussed by comparing the method with the conventional loading-unloading
decision method (the CV method) and the root-finding loading-unloading decision
method (the RF method).
Keywords. Explicit integration, non-isothermal, constitutive model, subloading

surface elastoplasticity, nonconvex surface, adaptive substepping
1. Introduction
An explicit integration scheme for the non-isothermal UH model [1] is developed in

this study. An efficient loading-unloading decision procedure (referred to as the
Double Cosine method, DC method) is proposed to address the numerical problems
associated with the unconventional elastoplastic model consisting of
convex/nonconvex subloading surfaces. Although this explicit integration scheme with
the DC method is designed based on the non-isothermal UH model [1], it can also be
applied to other subloading surface models. This paper is organized as follows. First,
for the purpose of FEM implementation, the constitutive equations of the non-
isothermal UH model are rewritten using generalized strain variables. The temperature
is treated as a strain variable here. Then, the DC method is introduced, which consists
of two steps, (i) the tentative decision and (ii) the rechecking procedure. If the
rechecking fails, it is able to find the division point that divides the strain subincrement
into a purely elastic part (unloading) and an elastoplastic part (loading). The DC
1
Corresponding Author. Dr Annan Zhou, Tel: +61-3-9925-0407, Email: [email protected]
A. Zhou and Y. Zhang / Integration Scheme for a Thermo-Elasto-Plastic Model 785
method is integrated into the adaptive substepping method, which was initially
proposed for conventional elastoplasticity models. Finally, the performance (both
accuracy and efficiency) of the proposed integration algorithm with the DC method is
investigated via a numerical example and compared with the existing loading-
unloading decision schemes (the conventional method, CV method and the root-finding
method, RF method [2]).
2. Constitutive equation
In the non-isothermal UH model [1], temperature is treated as an additional

constitutive variable, and the constitutive relationship can generally be expressed as:
, where is the
effective stress vector, is the elastic stiffness matrix (tangential) related to
mechanical loads, is the elastic strain vector related to mechanical loads, is the
total strain vector, is the plastic strain vector due to thermo-mechanical loads, d is
the temperature increment, is the elastic strain vector related to thermal loads, and
is the elastic stiffness matrix (tangential) related to thermal loads, which is equal to
.
In general, both the plastic potential surface and the subloading surface
can be defined by the effective stresses , temperature and hardening term ,
i.e., and . For the non-isothermal UH model, the
hardening term of the subloading surface is [1]. The plastic strain increment vector
( ) is assumed normal to the plastic potential surface, i.e., , where
d is the non-negative plastic multiplier that determines the plastic strain increments

owing to thermo-mechanical loads. Particularly, we have .
To keep the stress point on the subloading surface (i.e., consistency condition), the

total differential of the subloading surface can be calculated as

, where . The plastic

multiplier can be solved: , where , ,

and . In general, the
constitutive equations of the non-isothermal UH model can be written as follows:

= + = +

=

=

+ + (1)

= = + =

+

+

By treating as an additional strain increment, the constitutive equations of the

non-isothermal UH model can be simplified as: and , where

, and .
786 A. Zhou and Y. Zhang / Integration Scheme for a Thermo-Elasto-Plastic Model
3. Explicit numerical integration
3.1. Loading-unloading decision
A new loading-unloading decision method is proposed here for subloading surface

models, referred to as the Double Cosine (DC) method. Due to the non-linearity of
constitutive relationships, increment usually needs to be divided into several
substeps (i.e., )to ensure the accuracy of the stress integration (see the
next section, Stress integration with substepping). The loading-unloading decision
should be made in each subincrement (e.g., , which is the nth substep to impose ).
In general, the loading-unloading decision can be made by deciding whether the trial
vector for the subincrement directs towards the inside (unloading) or outside
(loading) of the current subloading surface. The tentative decision on whether the trial
vector directs outside or inside of the current subloading surface can be made by the
cosine of the angle . As shown in Figure 1, is the angle between the outward
normal vector (ONV) to the subloading surface at the current stress point (updated
from the (n-1)th substep) and the trial vector for the subincrement (i.e., the
nth substep) to be imposed. The cosine of angle can be calculated by

=

( )

(2)

where and are evaluated based on the updated stress and temperature
from the (n-1)th substep and . is the elastic thermo-mechanical
stiffness matrix updated with ( ) and = { }. In the condition of
, which means that angle is greater than , the trial
increment directs towards the inside of the initial subloading surface, which leads to the
tentative decision of unloading for . The non-linear elastic thermo-mechanical
stiffness matrix ( ) will be used in the stress updating ( ) for the nth substep.
is a suitable tolerance whose value can be set to 10-8~10-11. Otherwise (i.e.,
), the decision of loading is made tentatively, and the increments of stresses
( ) are calculated by the non-linear elastoplastic thermo-mechanical stiffness matrix
( ).
However, employing only the cosine of angle is not sufficient to complete the
loading-unloading decision for the nth substep. Figure 2a shows a case with a tentative
unloading decision involving a loading process and, Figure 2b shows a case with a
tentative loading decision involving an unloading process. The subloading surface in
the plane in Figure 2a is convex, and the subloading surface in the plane in
Figure 2b is nonconvex. Therefore, such a decision error can occur for both the convex
and nonconvex subloading surfaces. Thus, an additional checking procedure after
updating the stress and yield surface is necessary to avoid the problem shown in Figure
2. The rechecking procedure can be conducted using the cosine of angle , which is
the angle between the ONV of the subloading surface at the updated stress point
(updated from the nth substep) and the trial vector . The cosine of angle
can be expressed as

= ( )

(3)

where and are computed by the updated stress ( ) and the

updated temperature ( ).
Figure 1 Angle between trial vector and outward normal vector (ONV)
If has the same sign as , the tentative loading-unloading decision

can be confirmed and finally accepted. Otherwise, a division point ( , ) needs to be
found to separate the trial vector into a loading part and an unloading part. The division
point stays where the angle between the trial vector and ONV is equal to (i.e.,
, see Figure 2). The following equation can be used for determining the
division scalar :

= ( )

(4)

where and are evaluated at ( , ). Q- , and

. is a number between 0 and 1, that indicates the location of the division point in
the current subincrement. Once the local division point has been determined, the initial
subincrement for the nth substep should be reduced to . can be employed to
indicate the location of the division point towards the entire increment (), which can
be defined as
. Because and are one-to-one
corresponding to each other, the local division point can also be expressed as a
function of the entire division point by
. Therefore,

Q- and,

. The Pegasus method [3] is used in this paper to determine the value
of as it is unconditionally convergent, does not require the computation of derivatives,
and typically converges in four or five iterations.
Figure 2 Rechecking scheme for the convex/nonconvex subloading surface and the division point
3.2. Numerical integration with adaptive substepping
A modified Euler scheme with adaptive substepping is used for both the non-linear
elastic unloading and the elastoplastic loading to conduct the numerical integration for
the non-isothermal UH model. Because a loading-unloading decision is required for
each substep, achieving the required precision with a small number of substeps is key
to enhancing the efficiency of the numerical integration. Following the adaptive
substepping scheme [4], an appropriate size for each substep is determined by
controlling the local errors of the stresses and the hardening term.
A pseudo time t, lying in the range , is introduced in the adaptive
substepping scheme. At the start of the increment, where , the initial stresses, the
temperature and the hardening term are known as , and . A local error
measurement is used to automatically divide the imposed strain increment , which is
found by taking the difference between a second-order modified Euler solution and a
first-order forward Euler solution. The specific integration scheme for the non-
isothermal UH model has been described as follows.
Considering a pseudo time subincrement in the range of , the
subincrements of the generalised strain vector ( ), strain vector ( ) and
temperature ( ) are: , , . The values of , ,
and T at the end of a pseudo time step can be found using the forward Euler
method from: , , , where and
are the increment of the stresses and the increment of the hardening term according
to the forward Euler method, respectively. The superscript ~ indicates that its
variables are found by the forward Euler method. To determine the stresses and
hardening term at the pseudo time , the cosine of angle needs to
first be calculated by equation (2). According to the DC method discussed in the
section of Loading-unloading decision, the tentative unloading occurs in this pseudo
time subincrement if the angle is greater than ( ). We then
have . The thermo-mechanical state point at the end of the pseudo time
) always stays on the subloading surface due to the concept of the
(i.e.,
subloading surface elastoplasticity. Therefore, the hardening term increment can
be calculated by solving the subloading surface function ( ).
If angle is less than ( ), the tentative loading happens in the
imposed pseudo time interval. The elastoplastic constitutive law needs to be employed
to evaluate the values of the stresses and the hardening term at the end of interval
by the forward Euler method, which gives: and

.
A more accurate estimate of the stresses and the hardening term at the end of the
pseudo time increment can be computed using the modified Euler procedure. We

have , , and , where

the superscript ^ indicates that its variables are found by the modified Euler method.
For the tentative unloading, and the hardening term increment
can be calculated by solving the subloading surface function (
). For the tentative loading, and can be calculated as
follows: and .
The local error in the Euler and the modified Euler solutions can be estimated
using the stresses and the hardening term separately as:

{
= } (5)
If is not greater than a prescribed tolerance , the current strain subincrement
could be tentatively accepted. The increment of the stresses ( ) is equal to
, the stress vector ( ) is updated to , and the hardening term ( ) is
tentatively set to . Then, we have to recheck the tentative loading-unloading decision
by equation (3). If the rechecking is successful, the tentatively updated stresses, and the
tentatively updated hardening term are finally accepted and the current substep with a
pseudo time ( ) has been completed. Otherwise, the tentatively updated stresses and
tentatively updated hardening term should be rejected. A division point ( or ) should
be calculated according to equation (4), and the pseudo time for the current substep
should be reduced to . The entire substep with the reduced
pseudo time ( ) should be recalculated. If is greater than a fixed tolerance
, this substep should be rejected, and should be reduced to where
until the recalculated is not greater than
the . Then, the loading-unloading rechecking procedure will be applied until the
rechecking is successful. The next pseudo time step is set to . To limit the
maximum number of pseudo time increments in a given loading increment and prevent
the increments from growing immediately after a failed subincrement, minimum and
maximum pseudo time increments are needed. To allow for cases where no
subincrements are needed, the initial pseudo time step is typically set to unity. The end
of the integration procedure for this step is reached when all the substeps are complete
( ). At the end of each subincrement, the drift correction should be
imposed to ensure that the current stress-temperature state lies on the subloading
surface within a prescribed tolerance .
4. Performance evaluation
The accuracy (error) and efficiency (CUP time) of the DC method are assessed by a
numerical example in this section, which is designed to including both loading and
unloading processes within the given increments. The soil parameters used in this
example are extracted from [1]: Critical state, M=0.87; Hvorslev slope, MH=0.5;
elastoplastic compression index, =0.12; elastic unloading index, =0.02; thermo-
elastoplastic compression index, =410-4; thermo-elastic compression index, = -

110-4; Poisson's ratio, =0.3. Initial states: e0=1.2 and OCR = 4.
In this numerical example, the temperature remains unchanged ( ) during
the mechanical loading (drained). T is set to , which is higher than the reference
temperature ( ). The initial stresses at T are . The
total given strain increments are and , to ensure
that the stress path is initially toward the inside of the initial subloading surface on the
q-p plane and penetrates the initial subloading surface (see Figure 3). The LTOL for
this numerical example is set to 10-11. To evaluate the performance of the proposed
loading-unloading decision scheme, three different loading-unloading decision
methods (i.e., CV method, RF method and DC method) are employed in this example.
In this numerical example, all three loading-unloading decision methods are integrated
into the adaptive substepping scheme with automatic error control. The numerical
integration results are shown in Table 1.
Figure 3 Loading-unloading decision with a given strain-temperature increment in the p-q plane
The initial subloading surface is treated as the conventional yield surface by the
CV method, and the loading-unloading decision is made by only the trial stress
calculated at the beginning of the increment. Because the trial stress stays outside the
initial subloading surface, the whole stress path
is computed as elastoplastic. For
the CV method, the loading-unloading decision is made only once at the increment
level but not at the subincrement level.
Both the root finding method (RF) and the proposed method (DC) are able to find
division point B. Therefore,
is computed as purely elastic, and
is treated as
elastoplastic. The subloading surface shrinks before the state point reaches point B and
expands after that. The RF method needs to calculate the number of roots for each
substep, for example, by the Kronecker-Picard (KP) method, which involves a process
of numerical integration (integration range 0~1). For the KP numerical integration, we
use the forward Euler method with 1000 fixed steps. Because the KP numerical
integration will be repeated for each substep, applying the RF method will significantly
increase the computational time. In addition, the accuracy of the RF method for
determining the division point depends on the size of the substep. With a small number
of substeps, which satisfies a STOL of 10-4, the estimation error of the division point
(the RF method) can be 42 with a CPU time of 0.0306 seconds if the result of the DC
method is used as the benchmark. If we reduce the STOL to 10-8, the estimation error of
the division point (the RF method) can be reduced to 5, but the CPU time increases
to 1.3788 seconds. It is noted that along with the increase in the number of substeps in
the RF method with a huge sacrifice of computational efficiency, the predicted location
of the division point approaches the result of the DC method. Compared with the RF
method, the DC method is capable of determining the division point with more
accuracy and much less CPU time.
Table 1 Different loading-unloading decision methods for a given strain-temperature increment
Total CPU
Division
Method Stresses after the total given increment (kPa) stress ratio STOL success time
point
substeps (s)
1 2 3 p q (=q/p)
CV 100.6358, 47.6282, 47.6282, 65.2974, 53.0076 0.812 10-4 65 - 0.00624
98.2672, 46.5576, 46.5576, 63.7941, 51.7096 0.811 10-4 78 0.4309 0.0306
RF 98.2633, 46.5554, 46.5554, 63.7914, 51.7079 0.811 10-6 547 0.4154 0.1479
98.2632, 46.5554, 46.5554, 63.7913, 51.7078 0.811 10-8 5438 0.4138 1.3788
98.2628, 46.5551, 46.5551, 63.7910, 51.7077 0.811 10-4 56 0.4136 0.00941
DC 98.2632, 46.5554, 46.5554, 63.7913, 51.7078 0.811 10-6 545 0.4136 0.0130
98.2632, 46.5554, 46.5554, 63.7913, 51.7078 0.811 10-8 5439 0.4136 0.0428
The forward Euler method with 1000 fixed steps is used for KP numerical integration from 0 to 1. The
Division point is indicated by the value. An Intel Core i7-3770 CPU with a 3.4 GHz processor is used
for the numerical study
5. Conclusion
An adaptive substepping explicit integration scheme with a novel loading-unloading

decision method is developed for the non-isothermal UH model [1] by treating
temperature as an additional strain-like variable. The non-isothermal UH model
involves a convex subloading surface in the p-q plane and a nonconvex subloading
surface in the p-T plane, which may result in an incorrect elasticity/elastoplasticity
judgment if the conventional loading-unloading decision method is employed directly.
In this paper, a simple but robust loading-unloading decision method, the Double
Cosine (DC) method, is proposed to solve the problems associated with
convex/nonconvex subloading surface models. Compared with the root-finding (RF)
method [2], the proposed DC method can provide a more accurate loading-unloading
decision with higher computational efficiency.
References
[1] Yao, Y.P. and A.N. Zhou, Non-isothermal unified hardening model: a thermo-elastoplastic model for
clays. Geotechnique 2013. 63(15): p. 1328-1345.
[2] Pedroso, D.M., D. Sheng, and S.W. Sloan, Stress update algorithm for elastoplastic models with
nonconvex yield surfaces. International Journal for Numerical Methods in Engineering, 2008.
76(13): p. 2029-2062.
[3] Dowell, M. and P. Jarratt, The Pegasus method for computing the root of an equation. BIT Numerical
Mathematics, 1972. 12(4): p. 503-508.
[4] Sloan, S.W., A.J. Abbo, and D. Sheng, Refined explicit integration of elastoplsatic models with
automatic error control. Engineering Computations, 2001. 18(1/2): p. 121-154. Erratum:
Engineering Computations, 2002,19 (5/6): 594-594.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-792
Improvement of the dynamic behavior of

soil structures underlain by liquefiable soil
using the Geosynthetics-encased columns
M. Sc. Hassan ALKAYYALa and Dr. Eng. Talal AWWAD b
a
Institute for geotechnical engineering, Technical University of Freiberg, Germany
b
Higher institute for earthquake studies and researches, Damascus University, Syria
Abstract: From the simple separating layers up to the high strength tension
elements, the geosynthetics and their wide-range applications has innovated the
geotechnical engineering. After their successful application as reinforcement
elements, the concept of the geosynthetics-encased columns has developed to
benefit their mechanisms so they could improve the behavior of soil structures
constructed on weak soils. Due to the successive developments in the numerical
simulation and the constitutive modeling, very complex geotechnical problems
have modeled and the realistic behavior of the studied structure has obtained. In
this paper, a 3D finite-elements-analysis has been conducted to investigate the
enhancement, which was resulted by using the geosynthetics-encased columns on
the dynamic behavior of soil embankments that underlain by liquefiable sandy
soil. The results of the study analyzed and some recommendations, concerning the
improvement of the dynamic behavior of soil structures using the geosynthetics-
encased columns against liquefaction, have derived.
Keywords: Earthquakes, Liquefaction, geosynthetics-encased columns

(abbreviated as GECs), excess pore-water pressure.
1. Introduction
Geosynthetics have sustainably innovated the geotechnical engineering in the last 20

years as no other comparable building product. From the simple separating layers up to
the high strength tension elements, geosynthetics became calculable and manageable
due to their use in very complex problems and they led consequently to new
considerations in the design of the reinforced geotechnical structures. Depending on
their applications, geosynthetics have various functions, which are applicable in
different fields in geotechnical engineering. The main use of geosynthetics is to
reinforce weak soils in order to increase their shear strengths and to reduce their
deformations. Many applications of geosynthetics in the geotechnical engineering such
as the reinforced soil walls, reinforced embankments and foundation systems on the
geosynthetics-encased granular columns (abbreviated as GECs) were developed.
In the GECs, columns of non-cohesion material have installed to transfer the static
and dynamic loads from the geotechnical structures, which constructed on weak soils,
into deep soil layers of high bearing capacity. Encasing these columns by means of
H. Alkayyal and T. Awwad / Improvement of the Dynamic Behavior of Soil Structures Underlain 793
geosynthetics, the geosynthetics-encasement ensures the filter stability between the soil
inside and outside the columns. Because of the ring tension forces generated in the
encasement, the geosynthetics-encasement braces the columns in the radial direction,
which ensures the stability of the columns and enables the application of the GECs in
very weak soils. The granular columns (without geosynthetics-encasement) are for
many years state of the art in the geotechnical earthquake engineering and they are
widely employed to mitigate the hazard of soil liquefaction. By adding the
geosynthetics-encasement to the granular columns, the well-known functions of the
granular columns in mitigating the soil liquefaction hazards can be highly improved
and the shortcomings of the granular columns effectively excluded.
In the literatures, Adalier, Elgamal, Meneses and Baez 2003 carried out a series of
centrifuge tests to study the impact of the stone columns on the mitigation of
liquefaction hazard in non-plastic silty soils. They approved that the stone columns
significantly delayed the decay of soil strength loss and stiffened the overall behavior
of the reinforced system. Additionally, the composite ground had a considerably higher
liquefaction resistance than the uniform silt ground, which means that the soil treated
by stone columns required a higher number of cycles to liquefy. On the other hand, the
stone columns stiffened obviously the whole system, provided higher overall soil shear
strength and bearing capacity and prevented the excessive settlements. In addition, the
development of the excess pore-water pressures was slower and they dissipated faster
after reaching their ultimate values. Further, the stone columns retrofit process reduced
the permanent settlements by about 50%. In 2008, Madhav and Murali Krishna
discussed the functions of the granular columns and their mechanisms. They found out
that the principle mechanism of the granular stone columns in mitigating the soil
liquefaction is their drainage function because of their high permeability. In addition,
they studied the effect of the soil dilation on its liquefaction potential and concluded
that the densification provided by the stone columns increases the dilation of the
surrounding soil. They approved that this soil dilation generated negative pore-water
pressures in the granular columns, which led to permit faster rates of excess pore-water
pressure dissipation and enhanced therefore the liquefaction mitigation. On the other
hand, the dilation effect can offset the negative effect of the densification (reducing the
permeability). Moreover, the high deformation modulus and stiffness of the granular
columns provided reinforcement for the in-situ soil and thus offered another
mechanism to mitigate liquefaction. Zahn, Jiang and Yao 2014 carried out shaking
table model tests using a large-scale laminar shear box to study the dynamic
characteristics of saturated silty soil reinforced by means of stone columns composite
foundation triggered by seismic excitation. They found out that the stone columns
composite foundation could effectively restrain the increase of the excess pore-water
pressure to improve the anti-liquefaction ability of the ground. In addition, the stone
columns in the composite foundation played the most important role by carrying and
transporting the loads induced from the superstructure after the silty soil between the
stone columns had liquefied.
In this paper, a well-known case of study of liquefaction in the USA has
introduced and employed to validate the recently developed constitutive law UBC3D-
PLM to model soil liquefaction. The GECs have then benefited as a retrofit procedure
for mitigating soil liquefaction in the same case of study. The GECs applied for a case
of free field response and for a case of a geotechnical structure constructed on
liquefiable soil layer.
794 H. Alkayyal and T. Awwad / Improvement of the Dynamic Behavior of Soil Structures Underlain
2. The liquefaction phenomenon at the Wildlife Array
Hushman, Scott and Crouse (1992) in addition to Holzer and Youd (2007) mentioned
that the Wildlife liquefaction array established in 1982 on the flood plain of the Alamo
River in the Imperial Valley desert of southern California. The stratigraphy at the array
consists of saturated Holocene floodplain sediments, which are about 7.0 thick and
overlie denser sedimentary deposits (Figure 1). The floodplain sediments consist of a
2.5--thick layer of lean clay to silt (CL-ML) and overlie 4.3 of liquefiable silty
sand. The upper 1.0 of the liquefiable layer is sandy silt (ML) while the lower 3.3
is silty sand (SM). These two layers referred collectively to as the liquefiable sand
layer. The ground water table was at a depth of 1.2 below the ground upper surface.
Figure 1. Schematic cross section through the Wildlife Liquefaction Array (Bennett et al., 1984).
In November 1987, the main shock of the Superstitions Hills earthquake with a
magnitude of M = 6.6, a peak ground acceleration of PGA = -202 / and a
period of about 100 triggered the wildlife liquefaction array and liquefied the 4.3-
-thick sand layer. The liquefaction caused eruptions of water and muddy sediments,
which covered the land surface after the earthquake. Furthermore, sand boils, ground
cracks and ground lateral spreading accompanied the liquefaction of the sand layer.
Figure 2 depicts the north-south surface accelerations component of the strong
motion from the Superstitions Hills earthquake and the in-situ measured pore-water
pressure measurements. The excess pore-water pressures in all piezometers installed
inside the liquefied layer started to develop at 13.6 approximately and increased
slowly until the static effective overburden stresses reached and the silty sand layer
liquefied.
3. The UBC3D-PLM constitutive model
Introduced by Petalas and Galavi 2013, the UBC3D-PLM constitutive model, which is
an extension of the UBCSAND model, is an effective stress model based on the
classical theory of plasticity with a hyperbolic hardening role. It uses the Mohr-
Coulomb yield function in 3D instead of 2D principal stress space. In addition, it has a
modified non-associated plastic potential function based on Drucker-Pragers criterion.
Further, it includes a soil densification rule to predict the evolution of pore-water
pressure during cycling loading. This densification rule implemented by means of a
secondary yield surface, for which a simplified kinematic hardening used.
Figure 2. Time history of the northsouth surface accelerations and the generated excess pore-water pressure
recordings from all piezometers inside the liquefied layer.
The UBC3D-PLM is a descriptive model so its parameters derive by curve fitting
from laboratory tests on the studied soil. The most suitable test for this purpose is the
drained cyclic direct simple shear test (DSS). Moreover, there are several correlations
for the acquisition of the input parameters depending on the triaxial tests (CD-test) or
the SPT-results. Table 1 explains the parameters employed in the UBC3D-PLM model.
4. The validation of the constitutive model
A numerical 3D model which simulates the case of study at the Wildlife liquefaction
array was created using the finite elements code PLAXIS 3D to validate the
constitutive model UBC3D-PLM which will be later used to analyze more complicated
case throughout this study. Although the problem of soil liquefaction can be modeled
using the 2D modeling, the 3D modeling was chosen because the later intended
modeling of the GECs cannot be achieved in the 2D modeling. Figure 3 left shows the
created model and table 1 represents the soil parameters used to create this model. The
parameters employed throughout this study calculated using the correlations included
in the works of Makra 2013 and Galavi, Petalas, Brinkgreve 2013 depending on the
SPT-values published by Daftari and Kudla 2014.
The size of the model has chosen so that the boundary conditions did not have any
influence of the model behavior during the dynamic mode of calculation according to
Towhata 2008. The model had a length of 100 m and a depth of 50 m. Due to the
application of the dynamic load in the X-direction only, the model thickness (in the Y-
direction) would not affect the results. Therefore, the considered thickness of the model
was 3.0 m only. After modeling the soil layers and assigning the soil parameters, the
mesh has generated with fine coarseness and the static initial conditions were
calculated. In the dynamic mode of calculations, the record of the Superstitions Hills
earthquake has applied as a prescribed displacement on the base of the model. In
addition, absorbent (viscous) boundary conditions have assigned to the two lateral sides
of the model so that the earthquake-induced waves would be absorbed at the model
boundaries and would not reflect back in the model.
5. The geosynthetics-encased columns as liquefaction countermeasure
To analyze the effects of the GECs on the liquefaction phenomenon at the Wildlife
liquefaction array, a 3D numerical simulation with the aforementioned configurations
has carried out. In this model, a group of 20 GECs with a diameter of 1.0 m, a central
distance of 2.0 m and a length of 9.0 m were added as shown in figure 3 right. The
geosynthetics-encasements have modeled as an elastic material with a stiffness of 1000
kNm and the soil inside the columns has modeled as dense sand with a relative
density of 80%. The last column of table 1 shows the parameters of the dense sand
used to model the GECs.
To study the effectiveness of the GECs in mitigating the liquefaction hazards on
soil structures founded on liquefiable soil, another 3D model of an embankment
constructed on the previously mentioned soil layers has created. The previously
mentioned soil configurations, GECs and soil constitutive model have employed in this
model. In addition, an embankment of 5.0-m-height, which constructed using the same
dense sand, added on the top surface of the model as shown in figure 4. The
embankment was additionally reinforced by means of geosynthetics-layers of a
stiffness of 1000 kNm which was distributed every 0.5 m height of the embankment.
After modeling the GECs and constructing the embankment in static phases of
calculations, the record of the Superstitions Hills earthquake has applied as a prescribed
displacement on the base of the model and the two vertical sides of the model were set
as absorbent boundary conditions.
Figure 3. The 3D finite-elements-models (left) to validate the used constitutive model and (right) to assess
the effect of the GECs on the liquefaction.
Figure 4. The 3D model of the embankment constructed on the Wildlife liquefaction array.
Table1. The UBC-PLM3D parameters used in the numerical modeling and their explanation.
Parameter Unit I II III IV V VI
Classification - CL ML ML SM CH Dense sand
0 1.2 1.2 2.5 2.5 3.5 3.5 6.8 > 6.8 -
j@X] 16.0 19.4 19.7 19.7 20.0 22
X] 17.0 21.6 21.8 21.8 22.0 22.8
@ - 0.6799 0.7955 0.7400 0.7400 0.6878 0.8400
(o) 21.3 20 22 22 35 33
(o) 21.9 20.625 2.765 23.065 36 38.9
2.00 2.00 0.1 0.1 0.1 0.1
x - 788.2 798.9 854.6 954.1 934.3 1339.5

x - 185.1 193.6 250 424.7 380.3 3597.1
- 551.7 559.3 598.2 667.9 654 937.6
- 0.5 0.5 0.5 0.5 0.5 0.5
- 0.5 0.5 0.5 0.5 0.5 0.5
- 0.5 0.5 0.5 0.5 0.5 0.5
- 0.841 0.836 0.811 0.771 0.779 0.66
100 100 100 100 100 100
0 0 0 0 0 0
]m - 0.2 0.2 0.2 0.2 0.2 0.45
1F - 6 6.25 7.65 10.65 10 29.5
X - 0.02 0.02 0.02 0.02 0.02 0.02
Where:
j@X] is the unit weight above the phreatic level and X] is the unit weight below the phreatic level.
@ is the initial soil void ratio.
is the constant volume friction angle, is the peak friction angle and is the soil cohesion.
x , are the elastic shear modulus at the ref. level = 100 and the elastic shear modulus index.

x , are the drained plastic shear modulus and the plastic shear modulus index.
, are the elastic bulk modulus at the ref. level = 100 and the elastic bulk modulus index.
is the failure ratio.
is the reference stress which is equal to the atmospheric pressure.
is the tension cut-off stress.
]m is the densification factor.
1F is the corrected SPT value.
X is the post-liquefaction factor to determine the min. value of the shear modulus during stiffness
degradation.
6. Discussion of the results
Figure 5 left shows the numerically calculated excess pore-water pressures-time-

histories at different points inside the liquefiable layer. By comparing the diagrams of
the calculated (figure 5 left) and the measured (figure 2) excess pore-water pressures, it
is clear that there is a very good coincidence between the numerical simulation and the
reality. This outcome reveals the ability of the UBC3D-PLM to represent the main
mechanism regarding the liquefaction which is the increase of the excess pore-water
pressure and the decrease of the effective overburden pressure which lead to the loss of
the soil strength and consequently to soil liquefaction.
Moreover, Figure 5 right represents the development of the excess pore-water
pressures with time for the model improved using the GECs. It is clear that the
presence of the GECs inside the liquefiable soil layer has led to an explicit reduction of
the generated excess pore-water pressures in some parts of the model in compare with
the excess pore-water pressures diagrams measured in-situ or numerically calculated.
This reduction in the excess pore-water pressures was actually only noticed in the
lower part of the liquefiable soil layer (points A and C) while the excess pore-water
pressures in the upper parts of the model (points B and D) was increased until the
effective overburden stress totally vanished and soil liquefaction took place.
These results implicate on one hand that the GECs has enhanced the liquefaction
resistance of the sandy soil layer in compare with the behavior of the unimproved
model. On the other hand, this improvement could not prevent the occurrence of soil
liquefaction in the upper part of the model. This outcome has attributed to the
restriction effect provided on the lower parts of the GECs due to the columns
embedment in the underlain soil while the upper parts of the GECs were free to move
because of the absence of sufficient restriction. The aforementioned fact suggests that a
restricting mechanism, which decreases the available degrees of freedom on the upper
parts of the columns, is required to engage the reinforcing effect of the GECs.
100 100
A at -5,00m
90 Excess pore-water Pressure (kN/m2) 90 B at -3,00m
Excess pore-water Pressure (kN/m2)
80 80 C at -6,60m
D at -2,90m
70 70
60 60
50 50
40 40
30 30
20 A at -5,00m 20
B at -3,00m
10 C at -6,60m
10
0 D at -2,90m 0
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
Dynamic Time (S) Dynamic Time (S)
Figure 5. The results of the 3D modeling (left) before and (right) after using the GECs.
Figure 6 illustrates the results of the embankment model. It is obvious that the
presence of the GECs under the embankment body enabled the construction of the
embankment on the liquefiable soil layer and decreased the liquefaction hazard by
preventing the high increase of the excess pore-water pressures and by preserving the
soil effective overburden stress. By comparing the model response (acceleration-time-
history) in different points inside the liquefiable soil layers and the embankment body it
was found out that the stiffer behavior of the GECs increased the whole stiffness of the
model and provided higher overall shear strength and bearing capacity during the
dynamic mode of calculation. Furthermore, the GECs prevented the excessive
reduction of the embankment factor of safety, which had a value of 2.3 before applying
the seismic excitation and a value of 1.5 after executing the dynamic calculations.
35
30
Excess pore-water Pressure (kN/m2)
25
20
15
10
5
A at -5,00m B at -3,00m
C at -6,60m D at -2,90m
0
0 5 10 15 20 25 30 35 40 45
Dynamic Time (S)
Figure 6. The results of the embankment model constructed on the liquefiable soil layer.
7. Conclusion
A series of numerical simulations using the recently developed constitutive model

UBC3D-PLM has conducted throughout this paper to simulate a historical liquefaction
case of study at the Wildlife liquefaction array in California, USA. In addition, the
effect of the GECs on the liquefaction hazard mitigation has studied. This soil
improvement method has applied also to enhance the behavior of an embankment
underlain by liquefiable soil layer. The improvement of liquefiable soils by means of
the GECs has shown to be an effective method to mitigate the liquefaction hazards by
reducing the earthquakes-induced excess pore-water pressures and preserving the soil
effective overburden stress. Moreover, the embedment of the GECs in the underlain
and/or overlain soil layers has played a very important role in this type of soil
improvement by enhancing the restrictions on the tip and/or bottom of the GECs and
consequently improving the effectiveness of this method to mitigate liquefaction
hazards. More work in this subject is necessary to understand the enhancement
achieved by the GECs in mitigating the soil liquefaction hazards on geotechnical
structures.
References
[1] Hushmand, B., Scott, R.F., Crouse, C.B., In-place calibration of USGS pore pressure transducers at
Wildlife Liquefaction Site, California, USA, Earthquake engineering, 10th world conference (1992),
1263 1268.
[2] Adalier, K., Elgamal, A., Menses, J., Baez, J.I., Stone columns as liquefaction countermeasure in non-
plastic silty soils, Soil dynamics and Earthquake Engineering 23 (2003), 571 584.
[3] Holzer, T.L., Youd, T.L., Liquefaction, ground oscillation and soil deformation at the Wildlife array,
California, Bulletin of the seismological society of America, 97-3 (2007), 961 976.
[4] Madhav, M.R., Krishna, A., Liquefaction mitigation of sand deposits by granular piles An overview,
Geotechnical engineering for disaster mitigation and rehabilitation, Science press Beijing and Springer-
Verlag GmbH Berlin Heidelberg (2008), 66 79.
[5] Towhata, I., Geotechnical earthquake engineering, Sprivger-Verlag, Berlin (2008).
[6] Petalas, A., Galavi, V., Plaxis liquefaction model UBC3D-PLM, research report, PLAXIS, June 7,
(2013).
[7] Makra, A., Evaluation of the UBC3D-PLM constitutive model for prediction of earthquake induced
liquefaction on embankment dams, M.Sc. Graduation Thesis, Delft University of Technology (2013).
[8] Galavi, V., Petalas, A., and Brinkgreve, R.B.J., Finite element modelling of seismic liquefaction in soils,
Geotechnical engineering Journal of the SEAGS & AGSSEA 44-3 (2013), 55 - 64.
[9] Zahn, Y., Jiang, G., Yao, H., Dynamic characteristics of saturated silty soil ground treated by stone
column composite foundation, Advances in materials science and engineering (2014), Hindawi
publishing corporation, Article ID 745386, 7 pages.
[10] Daftari, A., Kudla, W., Prediction of soil liquefaction by using UBC3D-PLM model in PLAXIS, World
academy of Science, Engineering and Technology, International Journal of Environmental, Ecological,
Geological and Mining Engineering, 8-2 (2014), 108 113.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-800
Observation on Trace of Sand Boiling

Erupted from Liquefied Sandy Ground and
its Simulation by Laboratory Model Test
Yuki HORIUCHI a, Ngo Ngoc ANH b, Reiko KUWANO c1 and Jiro KUWANO d
a
Graduate student, Faculty of Engineering, The University of Tokyo
b
Student, Faculty of Engineering, Saitama University
c
Professor, Institute of Industrial Science, The University of Tokyo
d
Professor, Faculty of Engineering, Saitama University
Abstract. A fine sand ground below filled soil and silt layers appeared to liquefy
at Great East Japan earthquake. Undisturbed soil samples were taken from the sand
and silt layers of about 2 to 3 m deep. There were some traces of boiled sand
which elevated along cracks in the silt layer. X ray CT scanning was conducted for
the undisturbed specimens to observe the micro structure of liquefied soil. The
image obtained from X ray CT indicated the path of boiled sand erupted from
liquefied sand layer. The overall structure of liquefied layer is mostly disturbed in
contrast with laminar structures observed in a non-liquefied layer. A series of
model tests was conducted to simulate the phenomenon of sand eruption. A model
sand layer was prepared in the small soil chamber and the surface was covered by
an acrylic lid with a slit opening. Water penetrated into the model sand ground
from bottom to top. When the hydraulic gradient well exceeded critical hydraulic
gradient, sand eruption has started. Horizontal movement of sand grains at the
surface of liquefied layer was observed in the early stage of sand boiling. When
larger hydraulic gradient was applied, vertical movement of sand occurred. Similar
phenomenon was observed when the model sand ground was liquefied by the
vibration.
Keywords. Liquefaction, boiled sand, X-ray CT, model test
1. Introduction
The occurrence of sand liquefaction due to earthquake is often identified by the erupted
boiled sand on the ground surface. A trace of sand boiling is one of the most important
evidences of the presence of liquefied layer. On the other hand, the erupted sand often
causes trouble after the earthquake such as delay in the road restoration, generation and
scattering of dust, and etc. However, although liquefaction phenomena have been
investigated in detail, sand boiling and eruption are not well understood. In this study,
field observation and a series of laboratory model tests were carried out to investigate
the phenomenon of sand eruption.
1
Y. Horiuchi et al. / Observation on Trace of Sand Boiling 801
2. Traces of boiled sand elevated in a crack in a silt layer
Open trench was excavated to observe a liquefied sand layer at Itako city, Ibaraki,
Japan, where erupted sand heaps were found after Great East Japan earthquake in 2011.
The liquefied sand layer is covered by a silt layer and a filled sand layer as shown in
Figure 1. Some cracks filled with fine sand particles were found in the silt layer. It was
assumed that liquefied sand traveled up along the cracks in silt and eventually reached
the ground surface and the presence of liquefaction was identified at this location. The
lower boundary of the silt layer that is upper boundary of the sand layer was undulated
and sand grains come deep into the silt layer.
Figure 1. Ground profile
Samples were taken from each layer by block sampling and soil structure was
observed using X-ray CT scanning. Figure 2 shows images obtained from X-ray CT.
Although some disturbance during soil sampling was recognized, structure in soil can
be well identified. The filled sand layer which is considered not to have been liquefied
shows horizontal lamination in the structure. In the silt layer, the sand path was found
to extend in narrow cracks. Specimen taken from liquefied sand layer shows disturbed
structure to some extent.

Filled sand Liquefied sand

Silt
Figure 2. Structure of soil observed using X-ray CT scanning
802 Y. Horiuchi et al. / Observation on Trace of Sand Boiling
3. Model test for the simulation of sand eruption
In order to investigate how the sand grains are transported from the liquefied layer, a
series of laboratory model tests was conducted.
3.1. Apparatus
A model ground was prepared in a soil chamber of 30cm long, 8cm wide and 20cm
high, as shown in Figure 3. Water was supplied from the bottom of the model ground.
Hydraulic gradient of the water supply could be adjusted by the elevatable water tank
connected to the bottom of the soil chamber. In order to simulate the boundary of
relatively impermeable layer, surface of the model ground was covered by an acrylic
lid having 2mm wide opening in the center, from which boiling sand could be erupted.
Water table was adjusted by the drainage at the side wall of the chamber.

Figure 3. Photo and schematic figure of apparatus.
3.2. Model ground
Silica sand No.5, 6, 7 and 8 were used for the material of the model ground. It has
mean diameter of 0.36, 0.23, 0.13 and 0.08mm respectively. The particle size
distributions of the used materials are shown in Figure 4. For silica sand No.7,
maximum and minimum void ratios are 1.24 and 0.74. Loose sand ground, relative
density of approximately 50%, was prepared by the air-pluviation. Colored sand was
put on the surface and in front of the ground, as shown in Figure 5, for the observation
of sand grains movement.
Figure 4. Particle size distribution of tested materials


Figure 5. Example the model ground setup.
3.3. Test procedure
Water was slowly penetrated and the model ground was saturated in advance. Then the
water tank was elevated to apply additional hydraulic gradient to generate liquefaction
in the ground. Sand grains lost effective stresses and upward seepage flow caused sand
eruption from the opening in the lid. Three tests were conducted by applying vibration
to generate liquefaction instead of applying upward seepage flow. The soil chamber
was put on the shaking table and sinusoidal wave of 20Hz with maximum amplitude of
300gal was applied. Test case and the condition were presented in Table 1.
Table 1. Test condition

Case Sand type D50 Lid with Water head Water flow rate
No. (mm) opening difference*(cm) in the opening*(mm/s)
5w1 Silica sand No.5 0.36 with 55 26.3
6w1 No.6 0.23 with 40 9.6
7w1 No.7 0.13 with 30 5.8
7w2 No.7 0.13 with 25 6.6
8w1 No.8 0.08 with 25 3.6
8w2 No.8 0.08 with 15 5.4
7w0 No.7 0.13 without 70 (0.09)**
7v1 No.7 0.13 with 0 (vibration) 21.9
7v2 No.7 0.13 with 0 (vibration) 11.0
7v0 No.7 0.13 without 0 (vibration) (0.02)***
* when the sand eruption (or boiling) began
** estimated from the value of water head difference
*** estimated from the settlement of model ground surface
4. Test results and interpretation
4.1. Sand eruption and boiling due to upward seepage flow
Sand eruption was observed in all the test cases when the model ground was covered
with the lid. The water head difference at the beginning of sand eruption is shown in
Table 1. Although the drain lines were located at or below the model ground surface,
the drainage could not keep up with the supply of water. Therefore the water table at
the beginning of sand eruption came to be above the ground surface. When the water
head difference became larger, the sand eruption started. Considering that the height of
the model ground was 20cm, hydraulic gradient at the beginning of sand eruption was
between 0.75 and 2.75, which, in most cases, exceeded critical hydraulic gradient. The
flow rate of water at the opening was also measured and presented in Table 1.
At the beginning of sand eruption, it was observed that sand grains moved
horizontally toward the opening in the gap between the lid and the model ground
surface as shown in Figure 6. Further increasing hydraulic gradient caused vertical
movement of sand grains and a large amount of boiled sand flowed up from the
opening. The ground under the opening seemed to be highly disturbed as shown in
Figure 7. In the cases without the lid covering the model ground surface, whole ground
was heaved up at h=60cm and finally significant sand boiling occurred at the center
of the ground at h=70cm, as shown in Figure 8.
Figure 6. Movement of sand grains at the early stage of eruption
Figure 7. A large amount of boiled sand on the lid and deformation of the ground (Test 7w2, h=100cm)
a) h=60cm b) h=70cm
Figure 8. Movement of sand grains at the sand boiling (Test 7w0)
4.2. Sand eruption from the liquefied layer due to vibration
The sinusoidal wave of 20Hz was applied to the model ground for about 30 seconds.
Soon after the vibration has started, the model ground was liquefied and started to settle
as the shaking progressed. Sand eruption through the opening was only observed when
the lid can move down following the settlement of liquefied soil. The movement of
sand grains was mainly observed to be horizontal as shown in Figure 9 and Figure 10.
Even after the multiple shaking, the large scale sand boiling did not occur.
Figure 9. Movement of sand grains on the ground surface in a shaking table test (Test 7v2)
Figure 10. Observation of movement of sand grains in a shaking table test (Test 7v2)
4.3. Transportation of sand grains
The flow rate at the opening to cause sand eruption was obtained by measuring the
amount of drained water per unit time. They are shown in Table 1. Pidwirny (2006) and
Yee (2012) showed the relationship between stream flow velocity and particle erosion,
transport, and deposition as shown in Figure 11. Measured flow rates to cause eruption
of four sands with different grain sizes are also plotted in Figure 11. It is seen that the
data are located at the boundary between Transport and Deposition. Flow rate at
the opening should be high enough to bring out sand grains from the liquefied layer
along a crack in silt, i.e. to cause sand eruption on the ground surface.
Figure 11. Flow rate at the opening when the sand eruption started
It was found that the flow rate at the opening seems to be the most important
factors for the sand eruption. For the transportation of larger grains, higher flow
velocities are required. When relatively impermeable soil layer covers liquefied soil,
and if there are narrow cracks in the impermeable layer, it is assumed that water flow
concentrates on the cracks and the flow rate would be high. Sand particles can be
transported up along the cracks and reach the ground surface.
5. Conclusions
In order to understand the phenomenon of sand eruption from a liquefied sand layer
due to earthquake, the field observation and a series of model tests were conducted.
From the field investigation, it was observed that there were some traces of boiled sand
which elevated along cracks in the silt layer which is above the liquefied sand layer.
The X-ray CT scanning image of samples taken from the site indicated that the path of
boiled sand erupted from liquefied sand layer. The overall structure of liquefied layer is
mostly disturbed in contrast with laminar structures observed in a non-liquefied layer.
In the model tests, it was found that the sand eruption occurred when the flow rate at
the opening exceeds the threshold value which gives boundary between particle
transportation and deposition. Horizontal movement of sand grains at the surface of
liquefied layer was observed in the early stage of sand boiling. When larger hydraulic
gradient was applied, vertical movement of sand occurred. Similar phenomenon was
observed when the model sand ground was liquefied by the vibration.
Acknowledgements
The authors wish to thank Mr. S. Taira and Mr. T. Hofuku, former students of Saitama
University, for their effort of laboratory tests.
References
[1] Pidwirny, M. Erosion and Deposition. Fundamentals of Physical Geography, 2nd Edition. Viewed on Sep
11, 2014, http://www.physicalgeography.net/fundamentals/10w.html, 2006.
[2] Yee, T.W. Geosynthetics for erosion control in hydraulic environment, Proceedings of the 5th Asian
Regional Conference on Geosynthetics (Geosynthetics Asia 2012), Bangkok, Thailand, 119-134, 2012.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-808
Visualization of Shear bands in

cohesionless soils
Behzad SOLTANBEIGIa , Adlen ALTUNBASa and Ozer CINICIOGLU a1
a
Bogazici University, Istanbul, Turkey
Abstract. This study attempts to link the dilatancy angle to shear band thickness
and inclination. For this purpose, a small scale physical retaining wall model was
manufactured. Model wall is capable of translating horizontally to facilitate active
states in the backfill. On the other hand, transparent Plexiglas sidewalls of the
physical model test set-up both constrains the backfill in plain strain condition and
also allows observation of the progressive failure mechanism by employing
Particle Image velocimetry (PIV) method. Model backfills can be prepared at
various relative densities. Using the knowledge of density and stress state in the
backfill, dilatancy angles are estimated using a novel equation. As a result, shear
strain profiles of the deformed backfills, obtained from post-processing of the
displacement data captured using PIV technique, are presented and analyzed as
functions of the backfill dilatancy angles. The visualized intense shearing zones,
widely known as shear bands, were characterized with respect to dilative behavior
and mean grain size of the granular assembly. In this respect, the shear band
thickness and inclination along the retaining wall were assessed. Eventually,
geometric properties of generated shear bands were compared with the available
experimental and numerical data in the literature.
Keywords. Retaining wall, shear band, Particle Image Velocimetry (PIV),

localization, dilatancy
1. Introduction
Nonhomogeneous distribution of strain is a typical feature of granular soils during

shearing. The zones that deformations are localized are known as shear bands. Since
initiation of localization is coupled with diminution of strength in particulate media, it
is highly crucial to have an accurate estimation of shear band formation. For more than
thirty years, researchers investigated deformation localization in geomaterials [1-17].
One common principal objective of these studies was developing precise constitutive
models capable of determining the behavior of geomaterials during the initial
deformations and associated shear bands at pre-failure states. According to
experimental and numerical studies, some factors such as void ratio, grain size
distribution, grain shape, confining stress and direction of deformation influence the
shear band pattern and occurrence.
There are two different attitudes towards the evaluation of shear band formation
and its characteristics. The first is to monitor the voids inside the shear bands and the
neighboring soil. In this method, the density of the particles is considered as the main
1
Corresponding Author: Ozer Cinicioglu, Civil Engineering Department, Bogazici University, 34342
Istanbul, Turkey, E-mail: [email protected]
B. Soltanbeigi et al. / Visualization of Shear Bands in Cohesionless Soils 809
controlling parameter and tracked with utilizing several non-invasive techniques.

Leading approaches are gamma rays [18, 2, 19], and X-ray Computed Tomography
(CT) [20, 21, 23, 24, 25, 26, 28]. In this category of testing methods, the void ratio is
obtained indirectly through the correlation with attenuation measurements [11]. The
second group of testing techniques for the assessment of shear band evolution
concentrates on the physical movement of the particles, and provides an understanding
of strain distribution within the specimen. Several monitoring approaches were used to
catch the deforming pattern of soil, such as: spaced markers [27], grid points painted on
the specimen membrane [22]. Recently, digital imaging techniques (e.g. Particle Image
Velocimetry (PIV) and Digital Image correlation (DIC)) are employed for tracking the
displacement of the individual grains next to the transparent side of the sample [22, 29,
30, 11, 25, 31, 32].
In this paper, a detailed study was conducted to capture the initiation and
orientation of shear bands in the cohesionless backfill. In this respect, a small scale 1g
physical model, which allows the monitoring of the progressive failure mechanisms
through transparent Plexiglas sidewalls, was set up. Combined with the PIV method,
the model setup allows the investigation of failure surface geometries that emerge in
the retained backfill. The intensity profile of the backfills with various densities were
extracted and compared. Moreover, utilizing a novel equation that allows the
calculation of peak dilatancy and friction angle from the knowledge of pre-shearing
relative density and mean effective stress, the influences of dilatant behavior of
granular backfill on shear bands along the wall were investigated. The results supported
the dependency of the shear band properties on dilatancy angle of the tested backfills.
2. Physical Model Set-up and the Properties of the Testing Material
In the present study, in order to investigate the dependency of localized deformations

on dilatant properties of granular assemblies and to observe the failure mechanism
within the granular soil mass, 1g small scale retaining wall model tests were conducted.
The backfill soils in these tests were prepared with different relative densities (ID) to
achieve different dilatancy angles (p). Physical model set-up used for this purpose
consists of a testing box, a model retaining wall that is capable of translating laterally, a
sand pluviation system, a storage tank, a crane, and a data acquisition system. The
testing box is 140cm in length, 60cm in depth, and 50cm wide and simulates plane
strain condition, as shown in Figure 1. The dimensions of the model box are selected to
minimize the adverse effects of the boundaries. Width and height of the model are
chosen to achieve a model as large as possible without losing the practicality of
conducting numerous tests. Length (140cm) is selected to allow undisturbed formation
of passive failure planes, and thus provides much greater distance than the longest
passive failure plane ca cover. Sides of the testing box are made of 50mm thick
Plexiglas, allowing the observation and monitoring of the soil deformations. As a result,
photographic images of the backfill at different stages of wall deformation can be
captured for later analysis. Captured images are analyzed using particle image
velocimetry (PIV) method for identifying the geometry of failure surfaces. The material
used in model tests is a local sand called Akpinar Sand. Table 1 gives the principal
physical characteristics of Akpinar Sand.
810 B. Soltanbeigi et al. / Visualization of Shear Bands in Cohesionless Soils
3. Dilatancy Equation Based on In-Situ Properties
For obtaining (p) for the model backfills, a novel equation proposed by Cinicioglu and
Abadkon [33] is used (Eq. (1)). Eq. (1) allows the calculation of dilatancy angle using
pre-shearing soil properties:
pic
tan\ p D\ m\ I D (1)
pa
here, is the mean effective stress at the rest state prior to shearing and is the
atmospheric pressure; ID is the relative density of the sample; corresponds to the
decrease in dilatancy angle per unit increase in and m corresponds to the
increase in dilatancy angle per unit increase in ID. Both and m are empirical, unit-
independent, line-fitting parameters that can be obtained by conducting a few triaxial or
plane strain tests. Values of and m for Akpinar sand are -0.066 and 0.64,
respectively.
(a) (b)
Figure 1. Retaining Wall Model Set-Up (a) Plan view c) (b) A-A Cross-Section.
Table 1. Engineering Properties of Akpinar Sand.
Property Value
Classification Poorly Graded (SP)
Max. void ratio (emax) 0.87 (ASTM)
Min. void ratio (emin) 0.58 (ASTM)
D50 0.27
Uniformity coefficient (Cu) 1.23
Coefficient of gradation (Cc) 0.97
Average sphericity, Save 0.7
Average roundness, Rave 0.5
Critical state friction angle 'c () 33

4. Visualization of failure using PIV method
Particle Image Velocimetry (PIV) is a technique initially used for fluid mechanics
problems to identify the velocity field [34]. This approach was introduced into
geotechnical testing applications by White et al, 2003. The main objective of PIV is to
track the displacement fields over the surface of deforming soil by comparatively
analyzing two successive images captured at different stages of deformation.
In this study GeoPIV, a MATLAB module developed especially for geotechnical
applications [35] for detecting realistic deformations in granular media without any
physical contact, was used. While using GeoPIV, initially the area of interest (AOI) is
cut out of the digital image and divided into a grid of square patches, as shown in
Figure 2. These patches are distinguished by their unique pixel intensity variation
signatures. Afterwards, GeoPIV algorithm selects a specified search zone within the
deformed image. By using cross-correlation function, which tries to find an area that
has maximum overlap to the initial patchs signature and calculates the relevant
displacement, the local displacement vector is obtained. The correlation operations are
most successfully performed in the frequency domain by taking the Fast Fourier
Transform (FFT) of each patch [35]. This operation is repeated for all patches in the
domain and then for the next consequent image. This way, the evolution of any
deformation in the granular body can be captured.
(a) (b) (c)

Figure 2. Deformation determination in PIV a) AOI selection b) Initial image divided to square patches
c) displacement vector calculation.
GeoPIV provides detailed cumulative shear (s) and volumetric strain (v) maps for
every stage of the test. It defines a triangular mesh of elements connecting the center of
square patches with its neighboring patches. Relative patch displacement causes
extension or compression of these elements; this is used to calculate various
components of strain [36]. Through the analyses of these strain maps, shear strain
localization can be monitored and the evolution of slip plane geometry with wall
displacement can be observed, as presented in Figure 3.
(a) (b) (c)

Figure 3. Soil deformation observation a) Vector Field of Deformation b) Distribution of Shear Strains
(s) c) Distribution of Volumetric Strains (v).
Getting the initial idea from [16, 31], default color bar was changed to gray-scale
mode to interpret strain maps by intensity profiles. Accordingly, for gray-scale images,
which are based on 0255 scale, the lowest possible strain intensity corresponds to zero
(black), and the highest is distinguished with 255 (white), as illustrated in Figure 4(a).
Resulting intensity profile is used for identifying the quantified distribution of strain
values along a defined line segment, as shown in Figure 4(b). In this respect, the
visually defined strain maps of each test were quantified to identify the shear band
behavior with respect to dilative properties of the cohesionless backfill.
(a) (b) (c)

Figure 4. Shear strain distribution beyond the retaining wall a) Failure surface evolution of backfill
with (\p=0.4) with 3mm wall displacement (RGB image) b) gray scale shear strain profile c) Shear strain
intensity profile along the cross-section in section (1-1).
Determination of the shear band thickness is difficult and requires a reliable

method of measurement. Present study follows a convenient assumption proposed by
[37]. Accordingly, it was assumed that 50% of the maximum shear strain ( ) on
the intensity profile is the outer wedge of the shear band. This measurement allows
investigation of the relationship between shear band thickness (ts) and dilatancy angle.
Collecting data from all the model tests conducted, resulting variation of shear band
thickness (ts) as a function of peak dilatancy angle (\p) is presented in Figure 5. Based
on the results, it is clear that ts value increases with decreasing \p. To put it in other
words, thickness of the generated shear band is inversely proportional to peak dilatancy
angle of the backfill.
Additionally, in order to investigate the influence of grain size distribution on
shear band thickness, both the data from this study and the data available in literature
were collected [16, 29, 37, 39, 40, 41, 42, 43]. Resulting ts-D50 relationships are shown
in Figure 6 (several methods were used in the studies for measuring shear band
thickness such as: PIV, DEM, FEM). The results were categorized by density of the
samples since the dilatancy angles of the samples were not reported in other studies.
For dense specimens, the thickness of the shear band increases as D50 increases,
whereas for loose samples, an inverse relationship is observed.
The inclination of shear band () for various density states were measured and its
variation with peak dilatancy and friction angles (\p - p) is presented in Figure 7.
According to the results, it is revealed that, varies from 58 to 72 degrees with respect
to horizontal for the region starting from the bottom of the model wall, which provides
a good idea of how density and stress states of grains in particulate media influences
design criteria for retaining structures. It is seen that shear band inclination ()
decreases almost linearly with decreasing dilatancy and friction angles (\p - p ). So, it
is seen that initial inclination of shear band can be a function of either \p or p.
30
20
ts , mm
10
ts = -27.949tan (p)+ 25.577
R = 0.8532
0
0 0.2 0.4 0.6
tan ( p )
Figure 5. Influence of dilation angle on shear band thickness at active failure state.
(a) (b)
Figure 6. Influence of mean grain size diameter on shear band thickness at active failure state a) Dense
specimens b) Loose specimens.
Figure 7. Variation of Measured Shear Band Inclination and tan (45+p) with tan (p) at Active Failure
State.
5. Conclusions
A detailed experimental study was conducted for visualizing the localization of

shearing deformations in cohesionless soils. Model retaining wall tests were conducted
to active failure and PIV technique has been employed for capturing the resulting
deformations. Besides, a comprehensive literature survey was conducted to collect all

data regarding the geometric characteristics of shear bands corresponding to active
states. Consequently, following conclusions are drawn:
x Characteristics of shear bands, such as inclination and thickness are highly
dependent on dilatancy.
x As the dilatancy angle of the backfill increases, the shear band thickness
decreases.
x As the dilatancy angle of the backfill increases, the inclination of the shear
band with respect to horizontal increases.
x Shear band thickness is also dependent on the mean grain size of the sand
particles in the backfill; in the dense case, the shear band thickness increases
with increasing mean grain size, however, in the loose backfill the shear band
thickness decreases with increasing mean grain size.
Acknowledgements
The authors would like to thank The Scientific and Research Council of Turkey
(TUBITAK) for supporting this study with Project number 114M329. The financial
support of Bogazici University Scientific Research Projects (BAP7600) is also
acknowledged.
References
[1] J. R. F. Arthur, K. S. Chua and T. Dunstan, Induced anisotropy in a sand, Geotechnique, 27(1) (1977),
613622.
[2] J. Desrues, J. Lanier and P. Stutz, Localization of the deformation in tests on sand sample, Eng. Fract.
Mech., 214 (1985), 909921.
[3] F. Tatsuoka, S. Nakamura, C.C. Huang and K. Tani, Strength anisotropy and shear band direction in
plane strain test of sand. Soils and Foundations 30(1) (1990), 3554.
[4] W. Hammad, Modlisation non linaire et tude exprimentale des bandes de cisaillement dans les sables,
Thse de Doctorat, lUniv. J. Fourier, Grenoble, France (1991).
[5] C. Han and A. Drescher, Shear bands in biaxial tests on dry coarse sand, Soils Found. 33 (1) (1993) 118
132.
[6] J. Chu, S. C. R. Lo and I.K. Lee, Strain softening and shear band formation of sand in multi-axial testing,
Geotechnique, 46 (1) (1996), 6382.
[7] M. A. Mooney, R. J. Finno and M. G. Viggiani, A unique critical state for sand, J. Geotech. Geoenviron.
Eng., 124(11) (1998), 11001108.
[8] Q. Wang and P.V. Lade, Shear banding in true triaxial tests and its effect on failure in sand. J. Eng.
Mech., 127(8) (2001), 754761.
[9] J. A. Yamamuro and S. Shapiro, Failure and shear band in three-dimensional experiments on loose silty
sands, Proc., 15th Engineering Mechanics Conf., ASCE, Columbia University, New York, NY. (2002).
[10] K. A. Alshibli, S. N. Batiste and S. Sture, Strain localization in sand: plane strain versus triaxial
compression. J. Geotech. Geoenviron. Eng., 129(6) (2003), 483494.
[11] A.L. Rechenmacher, R.J. Finno, Digital image correlation to evaluate shear banding in dilative sands.
Geotech. Test. J. 27(1) (2003), 1322.
[12] G. Gudehus and K. Nbel, Evolution of shear bands in sand, Geotechnique 54(3) (2004), 187201.
[13] A.L. Rechenmacher, Grain-scale processes governing shear band initiation and evolution in sands. J.
Mech. Physics Solids 54 (2006), 2245.
[14] A. Ord, B. Hobbs, K. Regenauer-Lieb, Shear band emergence in granular materials a numerical study.
Int. J. Num. Anal. Meth. Geomech, 31 (2007), 373393.
[15] M.J. Jiang , H.B. Yan , H.H. Zhu , S. Utili, Modeling shear behavior and strain localization in cemented
sands by two-dimensional distinct element method analyses, Comp. and Geotechnics 38 (2011) 1429
[16] D. Lesniewska, M. Niedostatkiewicz, and J. Tejchman, Experimental study on shear localisation in

granular materials within combined strain and stress field, Strain, 48 (2012), 430444.
[17] A. Altunbas, B. Soltanbeigi, A.T. Gezgin, O. Cinicioglu, Shear Band Localization behind Retaining
Walls, 23rd European Young Geot. Eng. Conf., Barcelona, Spain, (2014), 201-204.
[18] D.G. Coumoulos, A Radiographic Study of Soils, Ph.D. Thesis, Cambridge University, UK (1968).
[19] S. Tan and T. Fwa, Influence of voids on density measurements of granular materials using gamma
radiation techniques, Geotech. Test J., 14(3) (1991), 257- 265.
[20] M. Mokni, Relations Entre Deformations en Masse et Deformations Localisees dans les Materiaux
Granulaires, These de Doctorat de lUniversite J. Fourier de Grenoble, France (1992).
[21] J. Desrues, R. Chambon, M. Mokni, and F. Mazerolle, Void Ratio Evolution Inside Shear Bands in
Triaxial Sand Specimens Studied by Computed Tomography, Geotechnique, 46(3) (1996), 529546.
[22] K.A. Alshibli and S. Sture, Shear Band Formation in Plane Strain Compression, Journal of
Geotechnical and Geoenvironmental Engineering, 126(6) (2000), 495503.
[23] K.A. Alshibli, S. Sture, N.C. Costes, M.L. Frank, F.R. Lankton, S.N. Batiste and R.A. Swanson,
Assessment of Localized Deformations in Sand Using X-Ray Computed Tomography, Geotechnical
Testing Journal, 23(3) (2000), 274299.
[24] A. Jaworski and T. Dyakowski, Application of electrical capacitance tomography for measurement of
gas-solids flow characteristics in a pneumatic conveying system. Measurement Science and Technology
12 (2001), 11091119.
[25] M. Niedostatkiewicz and J. Tejchman, Investigations of porosity changes during granular silo flow
using Electrical Capacitance Tomography (ECT) and Particle Image Velocimetry (PIV). Particle &
Particle Systems Charact. 24(4-5) (2007), 304312.
[26] Q. Marashdeh, W. Warsito, L.S. Fan and F. Teixeira, Dual imaging modality of granular flow based on
ECT sensors, Granular Matter, 10 (2008), 7580.
[27] M. Stroud, The behavior of sand at low stress levels in the simple shear apparatus. PhD thesis,
Cambridge University (1971).
[28] K.A. Alshibli and A. Hasan, Spatial variation of void ratio and shear band thickness in sand using X-
ray computed tomography, Geotechnique, 58 (4) (2008), 249-257.
[29] K. Nbel, Experimental and numerical investigation of shear localization in granular materials.
Publication Series of the Institute of Soil and Rock Mechanics, University Karlsruhe, 62.
[30] R.L. Michalowski, L. Shi, Strain localization and periodic fluctuations in granular flow processes from
hoppers, J. Geotech. Geoenviron. Engng., 129(6) (2003), 439449.
[31] D. Lesniewska, and D. Muir Wood, Photoelastic and photographic study of a granular material.
Geotechnique 61(7) (2011), 605611.
[32] B. Soltanbeigi, A. T. Gezgin, A. Altunbas and O. Cinicioglu, backfill failure surfaces and failure
displacements during active failure states, Int. Cong. on Advances in Civil Eng. (ACE), Turkey (2014).
[33] O. Cinicioglu and A. Abadkon, Dilatancy and Friction Angles Based on In Situ Soil Conditions." J.
Geotech. Geoenviron. Eng., 141(4), 06014019 (2015).
[34] R.J. Adrian, Particle Imaging Techniques for Experimental Fluid Mechanics", Ann. Rev. Fluid Mech,
23(1) (1991), 261-304.
[35] D. J. White, W. A. Take and M.D. Bolton, Soil deformation measurements using particle image
velocimetry PIV and photogrammetry, Geotechnique 53 (2003), 619 631.
[36] P. Jackson, an investigation into the deformation behavior of geosynthetic reinforced soil walls under
seismic loading, Master of Engineering Thesis, University of Canterbury, (2010).
[37] M. Niedostatkiewicz, D. Lesniewska and J. Tejchman, Experimental analysis of shear zone patterns in
sand during earth pressure problems using particle image velocimetry", Strain, 47 (2011), 218- 231.
[38] American Society for Testing and Materials, Standard Test Method for Particle Size Analysis of Soils,
ASTM Standards: D422-63 (2007).
[39] J. Tejchman, FE-Analysis of Patterning of Shear Zones in Granular Bodies for Earth Pressure Problems
of a Retaining Wall, Archives of Hydro-Eng. and Environ. Mech., 51(4) (2004), 317348.
[40] Niedostatkiewicz, M. (2009). Application of x-ray technique for monitoring deformation of granular
materials. Internal Report of Gdansk University of Technology.
[41] L. Widulinski, J. Tejchman, J. Kozicki and D. Lesniewska, Discrete simulations of shear zone
patterning in sand in earth pressure problems of a retaining wall, Int. J. of Solids and Structures, 48(7
8) (2011), 11911209.
[42] M. Jiang, J. He, J. Wang, F. Liu and W. Zhang, Distinct simulation of earth pressure against a rigid
retaining wall considering inter-particle rolling resistance in sandy backfill, Granular Matter Journal,
16(5) (2014), 797-814
[43] S.F. Tehrani, M.I. Arshad, M. Prezzi and R. Salgado, Visualization of Active Mode of Failure behind
Flexible Walls under Pure Rotation Using Digital Image Correlation, Geo-Congress Technical Papers,
ASCE (2014), 3393-3402.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-816
A microstructure-based finite element

analysis of the response of sand
Sadegh NADIMIa,1, Joana FONSECAa and Neil TAYLOR a
a
City University London, UK
Abstract. This paper presents a novel contribution towards understanding the stress
distribution amongst the constituent grains of an intact sand under loading.
Photoelasticity using birefringent materials has shown that forces in granular media
are transmitted from particle-to-particle via their contacts and the mode of load
propagation forms a complex force network. Particles carrying above average load
appear to form a network with special characteristics where stronger forces are
carried through chain-like particle groups, often referred as force chains. Fonseca et
al. (2013) showed that for a sand under shearing, the contact normals tend to be
orientated along the direction of the major principal stress, which suggests the
formation of force chains. Moreover, these quasi-vertically oriented vectors were
shown to be associated with contacts having large surface areas, contributing to the
formation of solid columnar structures of stress transmitting grains. This early study
demonstrates that a full characterization of force chains for real soils requires
accounting for the effects of the soil microstructure, including grain morphology
and contact topology, which the idealized nature of the particles used for discrete
element method simulations and photoelasticity studies cannot capture. In the
present work, high resolution x-ray tomographic data of an intact sand is converted
into a two dimensional finite element mesh, so that the microstructural details, such
as the geometrical arrangement of the grains and pores, as well as grain shape and
contact topology are incorporated in the model. In other words, the soil
microstructure is modelled using a computation approach that considers all available
geometrical data. The results suggested that the ability of the grains to transmit stress
via their contacts is directly associated to the degrees of freedom they have to move
and rearrange, which in turn is controlled by the topology of the contacts. The
insights into the effects of microstructure on the stress transmission mechanisms
provided in this study are fundamental to better understand and predict the macro
scale response of soil.
Keywords. Fabric/microstructure of soil, image, sand, FE, finite element

modelling
1. Introduction
The bulk behavior of granular materials subjected to external loading is determined by

the mode of propagation of the forces transmitted from particle-to-particle via their
contacts, the force networks [1]. The experimental measurement of these forces is non-
trivial, therefore, numerical modelling that account for the interaction of the constituent
particles of a sample under loading has been developed. For example, the Discrete
Element Method (DEM) has been extensively used for numerical simulation at the grain
1
Corresponding Author: Research Centre for Multi-Scale Geotechnical Engineering, City University
London, London EC1V 0HB, UK. E-mail: [email protected].
S. Nadimi et al. / A Microstructure-Based Finite Element Analysis of the Response of Sand 817
scale and has contributed to improve our understanding of the mechanical behavior of
granular materials. However, the idealized nature of the DEM simulations, i.e. spherical
shapes that are commonly used, remains an important drawback of the methodology.
While the influence of particle shape in the stress distribution amongst discrete grains
has been demonstrated in the literature [2, 3, 4], previous experimental studies have
shown that the influence of particle shape can be more significant than inter-particle
friction and surface roughness on the macro scale response of soil [5]. Grain morphology,
including size and shape, and contact topology properties are a consequence of the
geological history and the depositional settings of the soil. Consequently, intact soils
possess unique fabric features, in particular in terms of contact topology and grain
orientation, which are often lost in reconstituted soils and in addition cannot be
reproduced in the laboratory.
This paper presents a two dimensional microstructure-based finite element (FE)
model, which takes into account the natural depositional grain scale characteristics of an
intact sand. The model employs an image-based meshing method using x-ray
tomographic containing all the geometric information regarding the fabric of the sand.
FE methods have been widely used for solving problems in continuum mechanics [6]. In
order to account for the discrete nature of particle-to-particle interaction on the
simulations, FE methods that address contact mechanics applications and its
effectiveness have recently been established [7]. In particular, a number of contact laws
have been developed for DEM analysis based on these FE methods [8, 9].
2. Methodology
The methodology proposed in this study is summarized in the flow chart illustrated in
Figure 1. This includes image acquisition, image segmentation and contact detection,
which serve as input data for the image-based meshing and the FE modelling. This
Section discusses each stage of the methodology and provides a brief description of the
properties of the intact sand used in this study.
2.1. Image acquisition and segmentation
The material used in this study is Reigate sand, part of the Lower Greensand formation
(England). Reigate sand is a quartz-rich sand, with a median grain diameter (d50) of
300m and characterized by very high densities and an interlocked fabric which enables
the use of block sampling to collect intact samples from an outcrop of this material [10].
A minimal cement content was observed in the samples retrieved. The shape of the grains
in this geologically old, once deep buried sand, varies from near-spherical to highly non-
spherical with embayment [11]. The most striking feature of this sand is the
predominance of flat, extended contacts, contrasting to the point contacts found in more
recent sand formations or in reconstituted samples of the same sand.
Non-invasive images of the internal microstructure of Reigate sand were acquired
using high resolution x-ray computed tomography (CT). These graphs are maps of x-
ray attenuation based on composition and density of the material. Therefore, each pixel
in the image has an intensity value (or color) associated with the material it represents.
As shown in Figure 2a the grains or solid phase have brighter color (denser material)
while the pore space is darker. The images have a spatial resolution of 5m, i.e.
0.018d50, which means that each grain is represented by a large number of pixels and
818 S. Nadimi et al. / A Microstructure-Based Finite Element Analysis of the Response of Sand
the overall grain shape can be well captured. While the CT scans carried out generated
three dimensional images (3D), only slides along the full volume, i.e. two dimensional
(2D) images were investigated in this study.
In order to identify the individual grains, image segmentation techniques were used
[10]. This consisted of binarizing the image using a pre-set threshold value of intensity
to separate the pore space from the solid phase, followed by a watershed technique to
isolate the grains touching; details can be found in [11]. Figure 2b shows the segmented
image where each individual grain is represented by a unique color, the particles id that
takes values from 1 to N (N being the total number of grains) and the pore space that
takes the value 0 and is represented in black.
Image
Acquisition
Segmentation
Initial contact
detection
Mesh
generation
FE modelling
Boundary Contact
interaction Constitutive
conditions behaviour at
between grains
grain scale
External level
loading and Contact
fixities behaviour
Figure 1. Flow chart showing the steps of the proposed FE model.
(a) (b)
Figure 2. (a) CT raw image and (b) segmented image.
2.2. Contact detection
In order to complete the initial microstructure configuration, the contact region between
two contacting grains was detected and directly introduced in the model. Considering
two contacting grains, each grain has been defined as a set of pixels with a unique
intensity value in the segmented image, the particles id. Pixels are classified as contact
pixels if they are connected to a pixel with a different and non-zero id. The pixel contact
classification used in this study was based on a 4-connectivity pixel neighborhood
relation, i.e. face-to-face pixel connection. This operation required a total of four
orthogonal passes through the data along the x and y directions (x being the horizontal
and y the vertical directions). The size of the contacts was calculated by multiplying the
number of pixels defining each contact by the area of each pixel (i.e. the image
resolution). The contact surfaces were introduced in the FE model as the initial contacts
and the contact properties were assigned to them, as discussed later.
2.3. Image-based meshing
The segmented image representing the internal microstructure, in terms of grain

arrangements and real grain morphology, is the starting point for the image-based mesh.
In this work, an open source software package OOF2 (Object Oriented Finite Elements
v2, NIST) has been employed to create the mesh directly from the image [12]. An easy
and straight-forward approach to generate the mesh is by using a direct pixel-to-element
conversion. However, the computational costs of running the model, in this case, will be
very high due to large number of elements. A more efficient approach is proposed here,
which assumes that the grains are homogenous and therefore each grain can be
represented by larger elements containing a few pixels. These elements can be
represented by various polygonal shapes, for which the respective size and shape depend
on the size and shape of the grains.
The meshing process begins with specifying the minimum and maximum size of the
features presented in the image, i.e. the grains. The image is subsequently gridded
according to the specified sizes. It was found that that this automated gridding can often
introduce errors in the generated mesh, in particular, by creating elements that contain
part of two different grains or part of a grain and void space, named the inhomogeneous
elements. Amendment of these inhomogeneous elements can carried out either by
subdividing the elements, which increase the number of elements, or running a routine
that changes the topology of element without increasing the number of elements, which
is preferable. In these routines, nodes can be added, removed or reconnected to increase
the homogeneity of mesh. The effectiveness of enhancing the quality of the mesh by
modifying its topological structures has been recognized in previous studies [6].
Moreover, in order to model the contact interaction between grains more accurately, the
elements along the boundary of each grain were locally refined.
The assessment of the quality and efficiency of the mesh was taken into account by
quantifying i) the ability of the mesh to represent the features in the images, using the
homogeneity index, and ii) the mesh convergence behavior, using the shape index [12].
When all the pixels in the image are associated to a mesh element, the homogeneity index
equals 1. The shape index of regular shaped elements, such as triangles and squares
equals 0, while the shape index of thin and elongated elements takes higher values. High
aspect ratio elements may lead to slow convergence of the FE solver [13] and should,
therefore, explain.
The 600 600 pixel segmented image shown in Figure 2b was used to generate the
mesh for the example here presented. Mesh enhancement and refinement, resulted in a
homogeneity index of 0.97 and a shape index of 0. In total, it comprised 61884 nodes and
48963 elements. As can be seen in Figure 3a and in more detail in Figure 3b, the number
of elements containing more than one pixel, and in some cases a large number of pixels,
is significant, and this is critical to reduce the processing time and consequently improve
the computational efficiency.
(a) (b)
Figure 3. (a) Image-based mesh showing the refined elements along the boundary of each grain (b)
Detailed image of the mesh.
3. The FE model
This FE model consists of a 2D array of approximately 250 grains subjected to strain

controlled isotropic compression. In the case of a 2D simulation, it can happen that a few
grains do not exhibit any contact in the plane of the image and appear to be not in contact
to any surrounding grain. This is a limitation of 2D analysis and in order to minimize its
impact in the simulation the model was set to be capable of simulating large displacement
without inducing strain.
The image-based mesh generated in OOF2 has been converted to *.inp format and
imported to the commercial finite element package, Abaqus v6.13 (Dassault Systms).
This software has the capability of modelling contact interaction between the individual
grains comprising the sample. The geometric nonlinearity of the system was activated to
allow large deformation analysis. The external lateral boundaries of the model were fixed
in the x direction and loading was applied by defining prescribed displacement at the top
and bottom of the model, up to 10% strain value. The incremental change of boundary
conditions is described by a quasi-static model. Contact interaction properties and
constitutive parameters were assigned to the model, which are now described.
3.1. Contact interaction
Each contact between two given grains was modelled as surface-to-surface with
properties of hard contact in the normal direction and Coulomb friction in the tangential
direction. The effectiveness of these conditions when compared to Classic Hertzian and
MD theories described in the literature for idealized circular shape [8, 9, 14].
An element-based contact approach was considered, where the contacts were
detected in the initial assembly, prior to loading (Figure 4a). In addition, a node-based
contact has been defined along the free part of the boundary of each grain, i.e. defined
by the pixels in the grain that are in contact with pixels of value 0, i.e. pixels defining the
edge of the grain in contact to the pore space. This node-based contact represents the
potential surface of contact after grain rearrangement, imparted by the external loading
conditions (Figure 4b). For grains with irregular shape, as it is the case of real sand, it is
more appropriate to allow the whole free surface to potentially become a contact, which
was done by modelling all grains as general contact.
A refined mesh of approximately one element per pixel was used for the contacts,
which avoids potential noise in the measurement of the contact forces and, thus, enabling
a more accurate modeling of the forces at the contact interface. The surface-to-surface
technique was used to discretize the contact interfaces, which each contact constraint was
formulated based on an integral over the region surrounding the contact node, and
improves the accuracy of measuring the contact stresses. The coefficient of inter-particle
friction for the silica grains was considered to be 0.23 taken from experiments at the
grain scale [15]. To solve the model, a scripting user interface or Abaqus command
was used as the element based contact is not supported in the graphical user interface
Abaqus/CAE.
Contact Elements Potential Contact Nodes

along boundary
(a) (b)
Figure 4. Schematic definition of contacts: (a) element-based contact and (b) node-based contact.
3.2. Constitutive parameters
Linear and isotropic material behavior was assumed for the grains in this study. It is
important to note that plasticity or viscosity can also be introduced to model, and their
influence on the behavior at a grain scale level, investigated. Averaged values of 100
GPa for Youngs modulus and 0.15 for Poissons ratio, corresponding to bulk modulus
of K=47.6GPa and shear modulus of G=43.5GPa [16], were used. Here, we are assuming
that the bulk behavior is controlled by grain rearrangement and not by plastic behavior
at the grain scale.
4. Result and discussion
The results were investigated in terms of grain displacement and stress distribution.
Figure 5a shows the arrows representing the displacement of the grains, in this case, the
node displacement. In the same way, the stress at the nodes is illustrated in Figure 5b. It
can be observed that the grains showing larger displacement, i.e. significant
rearrangements, are associated with very small or zero stresses, while the grains carrying
the majority of the stress through contacts have more stable positions. This suggests that
the contact topology, in particular, the large area contacts, is responsible for controlling
the ability, or in this case the inability, of the grain to rearrange and therefore to be able
to transmit stresses, i.e. to be part of the force chain. Another key point to highlight from
these results is the anisotropy of the contact network of stress transmitting grains, when
responding to an externally applied load. This anisotropic and heterogeneous network is
a result of the irregular shapes and sizes of the grains and the complexity of the associated
geometrical arrangement or fabric.
In order to interpret the results, an awareness of limitations of a 2D analysis is
required. As already discussed previously, the fact that the 2D slice of the volume image
may not cut thought the contact between two given grains makes these grains more
vulnerable to move and lead to misrepresentative results.
(a) (b)
Figure 5. (a) Grain displacement represented by black arrows indicating the direction and magnitude
(given by the arrows length). (b) Stress distribution at the nodes of each grain in black.
5. Conclusion
This paper presents a novel contribution towards the development of FE models able
to investigate the primary fabric of stress transmitting grains. The model here described
takes as input an image of the internal microstructure of the soil, in order to account for
all the relevant geometric data that influences the response of the material under external
loading, such as, grain morphology, contact topology and grain arrangements. The
results have shown the development of a highly anisotropic network of stress
transmission under the applied loading and boundary conditions. The ability of the grains
to transmit stress via their contacts seems to be directly associated to the degrees of
freedom they have to move and rearrange. It is suggested that grains forming large
contact areas are more stable and therefore more likely to transmit stress and be part of
a force chain. This observation highlights the need to account for fabric of real soil and
in particular the intact fabric features imparted during geological history. The inherent
limitations of a 2D analysis are discussed. On-going work includes the 3D extension of
this FE model.
Acknowledgments
Authors acknowledge the National Institute of Standards and Technology (NIST),

an agency of the U.S. government, as the source of the code OOF2 used in this research
and the first author would like to express his thanks to City University London for his
doctoral scholarship.
References
[1] T.S. Majmudar, R.P. Behringer, Contact force measurements and stress-induced anisotropy in granular
materials. Nature 435 (2005), 1079-1082.
[2] H.G. Matuttis, Simulation of the pressure distribution under a two dimensional heap of polygonal particles,
Granular Matter 1 (1998), 83-91.
[3] I. Zuriguel, T. Mullin, The role of particle shape on the stress distribution in a sand pile, Proc. R. Soc. A
464 (2008), 99-116.
[4] I. Zuriguel, T. Mullin, J.M. Rotter, The effect of particle shape on the stress dip under a sandpile, Phys.
Rev. Lett. 98, 028001 (2007).
[5] I. Cavarretta, M. Coop, C. OSullivan, The influence of particle characteristics on the behaviour of coarse
grained soils, Gotechnique 60 (2010), 413423.
[6] O.C. Zienkiewicz, R.L. Taylor, The Finite Element Method. 6th Edition, Elsevier, Oxford, 2005.
[7] P. Wriggers, Computational Contact Mechanics, Springer, 2006.
[8] Q.J. Zheng, H.P. Zhu, A.B. Yu, Finite element analysis of the contact forces between a viscoelastic sphere
and rigid plane, Powder Technology 226 (2012), 130142.
[9] L. Vu-Quoc, X. Zhang, L. Lesburg, Normal and tangential forcedisplacement relations for frictional
elasto-plastic contact of spheres, International Journal of Solids and Structures 38 (2001), 64556489.
[10] C.C. Reyes-Aldasoro, Biomedical Image Analysis Recipes in MATLAB: For Life Scientists and Engineers,
Wiley-Blackwell, London, 2015.
[11] J. Fonseca, The evolution of morphology and fabric of a sand during shearing, PhD Thesis, Imperial
College London, (2011).
[12] A.C.E. Reid, S.A. Langer, R.C. Lua, V.R. Coffman, S. Haan, R.E. Garca, Image-based finite element
mesh construction for material microstructures, Computational Materials Science 43 (2008), 989-999.
[13] J.R. Shewchuk, What is a Good Linear Element? Interpolation, Conditioning, and Quality Measures,
Invited Talk, 11th International Meshing Roundtable, Springer-Verlag, (2002), 115-126.
[14] E. Dintwa, E. Tijskens, H. Ramon, On the accuracy of the Hertz model to describe the normal contact of
soft elastic spheres, Granular Matter 10 (2008), 209221.
[15] K. Senetakis, M.R. Coop, M.C. Todisco, The inter-particle coefficient of friction at the contacts of
Leighton Buzzard sand quartz minerals, Soils and Foundations 53 (2013),746-755.
[16] R. Holtzman, D.B. Silin, T.W. Patzek, Mechanical properties of granular materials: A variational
approach to grain-scale simulations Int. J. Numer. Anal. Meth. Geomech; 33 (2009), 391404.
[17] J. Fonseca, C.C. Reyes-Aldasoro, C. O'Sullivan, M.R. Coop, Experimental investigation into the primary
fabric of stress transmitting particles, Geomechanics from Micro to Macro, Eds. Soga et al., Taylor &
Francis Group, London.
[18] K.J. Hanley, X. Huang, C. OSullivan, F.C. Kwok, Temporal variation of contact networks in granular
materials. Granular Matter 16 (2014), 4154.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-824
Stiffness Anisotropy Characteristics of

Natural Fined-Grained Seabed Sediments
Delvin Abdiel MARTNEZa,1 and Satoshi NISHIMURAb
a
Graduate student, Graduate School of Engineering, Hokkaido University, Japan
b
Faculty of Engineering, Graduate School of Engineering, Hokkaido University, Japan
Abstract. The anisotropy characteristics of natural fined-grained soils' small-

strain stiffness were experimentally studied and interpreted with an effective-stress
cross-anisotropic elasticity model. Use of highly precise local instrumentation
allowed determining all the five parameters necessary for describing the cross-
anisotropic behavior of soil at small-strain levels. Similar techniques and
interpretations have so far been applied to uniform sands and clays, but the
investigated gradation ranges were fairly limited. In this research, natural samples
of Holocene and Pleistocene seabed sediments with a wide range of fines content
were tested to study their anisotropic stiffness parameters at small-strain levels.
The inherent anisotropy characteristics were further discussed by looking at two
more commonly quoted inter-modulus ratios, Ghh/Gvh and Eh'/Ev'. They were found
to range 0.9-1.3 and 0.6-1.4, respectively. These ranges are closer to those reported
for reconstituted sands than those for clays. There is a possibility that larger
particles are more dominant in determining the stiffness anisotropy patterns in
non-uniform intermediate soils.
Keywords. Cross-anisotropy, deformation, stiffness, intermediate soils
1. Introduction
The soil behavior at small-strain levels is commonly described by linear elasticity

theory. In particular, several studies from different point of views have demonstrated
that the sedimentary soils' behavior at small strains can be described by a cross-
anisotropic elasticity model. Conceptually, different aspects of the stiffness anisotropy
have been labeled inherent and induced. The inherent anisotropy is intrinsically
related to the grain forms and the depositional processes, and independent of the
applied stresses and strains, while the induced anisotropy is caused by stress or strain
changes following the depositional processes. The knowledge of the anisotropy
characteristics of the soil, both inherent and induced, is an important prerequisite for
making an accurate ground deformation prediction.
Despite the relative simplicity of the theory and formulation of cross-anisotropic
elasticity, existing experimental investigations into its parameters are limited to those
on soils with relatively uniform gradation, such as reconstituted sands [1-3],
reconstituted silt [4], natural shale [5] and natural clays [6-11]. In this study, a so-called
intermediate non-uniform offshore soil samples were studied to characterize their soil
1
Corresponding Author. Delvin Abdiel Martinez, Graduate School of Engineering, Kita 13. Nishi 8,
Kita ku, Sapporo, 060-8628. Hokkaido, Japan. Email: [email protected]
D.A. Martnez and S. Nishimura / Stiffness Anisotropy Characteristics 825
stiffness at small strains. The anisotropy stiffness patterns were investigated

experimentally via a simple effective-stress-dependent cross-anisotropic elasticity
model, and their relation to factors such as the sample age and physical properties was
considered. Understanding the general picture of multi-dimensional stiffness
characteristics of offshore intermediate soils will be helpful in analysing dynamic
responses of marine structure such as breakwaters and wind turbine piles against wave
and earthquake loading.
2. Cross-Anisotropic Elasticity Model
The cross-anisotropic model is a direct deduction of the Hookes law under some
mechanical and thermodynamic considerations. This model has a main assumption that
a material presents isotropic behavior in one plane and a same pattern of anisotropy in
the orthogonal planes. Cross-anisotropic elasticity in which the plane of isotropy is
horizontal is mathematically represented by Eq. (1).
1 c
hh c
vh
Ec 0 0 0
E hc Evc
h
c
hh 1 c
vh
x 0 0 0 x Gvh Ghv
E hc E hc Evc
c c
hv y Q' vh Q' hv
y hv 1
0 0 0
z
(1)
z
E hc E hc Evc E' v E' h

1
xy 0 0 0 0 0 xy E' h
yz Ghh yz
Ghh
2( 1 Q' hh )
1
zx
0 0 0 0 0 zx
Gvh

0 1
0 0 0 0
Gvh
The characters E, G and stand for Youngs modulus, shear modulus and
Poissons ratios, respectively, and the subscripts v and h represent the vertical and
horizontal directions, respectively. A full description of a cross-anisotropic elastic
material requires five independent parameters, E'v, E'h, 'vh, 'hh and Gvh. The prime
indicates that the parameters are defined in term of the effective stress.
In order to take into account the effective stress dependency of the modulus values,
a set of equations, Eqs. (2) to (5), proposed for the moduli of sand [3], were adopted to
parameterize the stiffness anisotropy patterns.
Evc f (e)Cv ( (cv ) / pr ) 2 mv ; Ehc f (e)Ch ( (ch) / pr ) 2 mh (2)
Gvh f (e)Cvh ( vc ch / pr2 ) nvh (3)
Ghh f (e)Chh ( hc / pr ) 2nhh (4)
where pr is a reference pressure and the in-situ vertical effective stress (v0') estimated
for each sample have been adopted as its values in this research. The coefficients Cv, Ch,
Cvh and Chh are normalized model parameters obtained from a linear regression analysis
of the moduli values measured by a triaxial test with local deformation measurements
826 D.A. Martnez and S. Nishimura / Stiffness Anisotropy Characteristics
and bender element tests. The exponents, mv, mh, nvh and nhh, represent the effective
stress dependency of each modulus for all the tested samples. The function f(e) is a
void ratio function, to take into account the effect of the density of the soil on stiffness.
A void ratio function previously proposed [12] is used, as is given by Eq. (5)
f (e) e F (5)
where e is the void ratio and is a constant. A value of 1.3 is used in this study, as
was proposed [12] for clays. The choice of value for this parameter in fact had very
little influence on the deduced parameters that are to be discussed in this paper.
3.1. Tested Materials
The present study is still in progress, and the tests have so far been performed on six
saturated natural samples of fined-grained Holocene and Pleistocene seabed sediments
from different depths. The samples were retrieved from a breakwater construction site
in the Ishikari Bay New Port, Japan, by a pushed thin-wall sampler or a triple-barrel
rotary core sampler, depending on the resistance of each layer. According to the
physical properties of the soil samples, many of them can be classified as low plasticity
(with three of them non-plastic) and non-uniform materials and regarded as
intermediate soils (sand content of 50-80% and the plasticity index of less than 30),
with a wide range of physical characteristics, as shown in Figures 1a) and 1b) and
Table 1.
The age of the tested samples shown in Table 1 was estimated by interpolating the
radiocarbon dating measurements carried out at six different depths. Although the
depths of investigation at present are still discrete, it is considered broadly
representative of the stratigraphy at the site, from the surface (seabed at -19m a.s.l.) to
the lower bound of the survey (-40m a.s.l.). The sample T2-1 seemed to be a
transition layer between sand and clay. The sample was constituted by clay with pure
fine sand nodules encrusted in it. In general, however, the non uniformity in the
samples texture was such that it can be captured by the adopted specimen size.
Figures 1.a) Gradation curves; b) plasticity of the tested samples, respectively.

Table 1. Physical characteristics of tested samples

Elevation Depth below wn wL wP s Age
Sample Clay/Silt/Sand
(m a.s.l.) seabed (m) (%) (%) (%) (g/cm3) (yr BP)
T2-1 -22.5 3.6 53.5 52.6 20.2 2.66 56/30/14 7600
T1-3 -27.5 8.8 27.5 - - 2.66 24/41/35 16200
T1-4 -28.5 9.8 37.1 31.1 27.1 2.60 26/71/3 17500
T1-5 -30.0 11.3 25.4 - - 2.66 14/41/45 17700
S2-3 -33.5 14.6 41.9 49.1 24.2 2.60 27/57/16 25000
T2-6 -37.5 18.6 32.4 - - 2.62 35/33/32 31400
3.2. Testing Machine
Two identical triaxial testing machines equipped with bender elements and local
deformation devices, illustrated in Figure 2a), were used to determinate the necessary
five elastic parameters in the cross-anisotropic elasticity model. The two pairs of
bender element, one across the sample diameter and the other across the height, were
used to measure the shear moduli, Ghh and Gvh, respectively [7].
In this study, specimens with nominally 75-mm diameter and 100-110 mm height
were tested. These dimensions are meant to match the common core sizes in Japan. A
detailed description of the triaxial testing machine can be found in [13].
3.3. Testing Method
The stiffness parameters were evaluated at three different stress states for each tested
soil sample, as shown in Figure 2b), starting from isotropic stress conditions to
anisotropic (K=0.5) stress states. At each stress point, three different types of loading-
unloading probes were conducted, following the procedures explained by [13]. Each set
of probes consisted of (a) 6 cycles of drained axial loading-unloading with a constant
radial stress, (b) 6 cycles of drained radial loading-unloading with a constant axial
stress, both with strain amplitude of 0.001% at a very slow rate of 0.001%/hour, and (c)
6 cycles of undrained axial loading-unloading with a strain amplitude of 0.001% at a
rate of 0.04%/hour. The resultant small deformations in the axial and radial directions
were measured locally with a pair of LVDTs and a set of three non-contact gap sensors,
respectively. To measure the remaining elastic parameters, bender element tests were
carried out at each stress point.
Figure 2. a) Triaxial testing machine equipped with bender elements and deformation measurement
devices; b) Effective stress paths followed during the tests and stress state points at which the sets of
probes were performed
4. Results and Discussions
4.1. Model Parameterization and Indices of Inherent Stiffness Anisotropy
The values of the normalized coefficients, Cv, Ch, Cvh and Chh, were determined by a
linear regression analysis of the relationship between each elastic moduli normalized
by the void function, f(e), and a product of corresponding stress components in a log-
log scale, as shown in Figure 3a). These coefficients express the magnitude of the
moduli at a particular reference pressure, pr. The ratios between them represent the
anisotropy at isotropic stress stresses, or inherent anisotropy, that underlies their
variation during anisotropic stress changes. The stress-dependency of each elastic
modulus was evaluated by the exponents in Eqs. (2) to (4) (mv, mh, nvh and nhh). These
exponents are defined as the slope of the trend lines of the relationships shown in
Figure 3a). The linear regression analysis seems to be justifiable in most cases,
although the horizontal moduli (i.e. Eh' and Ghh) were less well represented by a linear
trend, especially during the transition from the isotropic stress states to anisotropic
stress conditions (i.e. from A to B in Figure 4b)). This feature was consistently seen in
all the samples, and consequently the determination of the corresponding model
parameters (i.e. Ch, Chh, mh and nhh) includes greater uncertainty than that of other
parameter.
The derived model parameters are shown in Table 2. The pr values used in this
study (Vv0') were different for each sample, as shown in Figure 3b), and a direct
comparison of the normalized model parameters values (Cv, Ch, Chh, Cvh and Cvu) is not
meaningful, and only the ratio between them will be discussed. The constant Cvu is a
normalized parameter corresponding to the undrained vertical Youngs modulus, Evu,
and calculated from the other model parameters [11]. Tentative interpretations are
carried out by grouping the samples into two. Group 1 (T1-3, T1-4 and T1-5) considers
different soil types at an almost same depth, consequently of an almost same age, and
Group 2 (T2-1, T2-6 and S2-3) considers a similar soil type at different depths.
T1-3: Clay:24% Silt:41% Sand:35% In situ Effective Vertical Stress

100 -20
a) b)
-22
Elastic moduli/f(e) (MPa)
-24
Elevation (m a.s.l.)
-26
-28
-30
Cvh nvh
Ev'/f(e)
Ev'/f(e) -32
Eh'/f(e)
E '/f(e) -34
h
Gvh/f(e) -36
G /f(e)
vh
-38
pr2='v0 Ghh/f(e)
Ghh/f(e)
10 -40
500 5000 0
50000 50 100 150 200
v'v', h'h', v'h'(kPa2) Effective Vertical Stress (kPa)
Figure 3. a) Example of elastic moduli normalized by the void ratio function plotted
against corresponding stress products; b) In situ vertical effective stress against depths
Table 2. Values for the normalized model coefficients and exponents

Eh' Ghh Ch Chh
Sample Cv Ch Chh Cvh Cvu mv mh nhh nvh
Ev' Gvh Cv Cvh
T1-3 47 48 22 28 57 0.29 0.23 0.19 0.29 1.2 1.0 1.0 0.8
G.1
T1-4 52 69 34 35 72 0.31 0.24 0.27 0.28 1.4 1.0 1.3 1.0

T1-5 95 75 21 22 120 0.22 0.32 0.05 0.03 0.6 0.9 0.8 1.0
T2-1 28 48 23 21 44 0.14 0.23 0.10 0.17 1.4 1.3 1.7 1.1
G.2
S2-3 88 166 45 58 106 0.28 0.50 0.23 0.30 1.2 1.0 1.4 1.1
T2-6 89 126 62 58 102 0.27 0.26 0.21 0.19 1.4 1.1 1.9 0.8
4.2. Inherent Anisotropy
The inherent anisotropy is commonly evaluated through two pairs of normalized

modulus parameters ratios (i.e. Chh/Cvh and Ch/Cv) or their corresponding elastic
modulus ratios (i.e. Ghh/Gvh and Eh'/Ev') at isotropic stress states. These ratios are
summarized in Table 2. The difference between these pairs are due to factors that are
not captured by the model equations. In addition to these ratios, this study considers the
anisotropy in full by using four ratios: Ch/Cv, Chh/Cvh, 2(1+vh')Cvh/Cv and 3Cvh/Cvu [11].
The vh' values were taken as the average of those obtained for different stress stages
(i.e., A, B and O, in Figure 2b)). The selected ratios are all bound to be unity for
isotropic materials, and any deviation from unity indicates anisotropy. These ratios are
graphically shown in Figure 4 for all the six samples, together with the results obtained
from previous research [11] for sedimentary clays (natural and reconstituted
conditions). It is noticed that Group 2 as a whole indicated higher degree of anisotropy
than Group 1 in three of the stiffness ratios, Ch/Cv, 2(1+vh')Cvh/Cv and 3Cvh/Cvu, which
were all larger than 1. In Group 1, the sandy silt sample T1-5 indicated a very different
anisotropy patterns than the other samples, making Group 1 more variable. Although
the data obtained so far are severely limited, this finding might indicate that lithology
or gradation is more dominant over the inherent anisotropy pattern than depth or age.
The ratio Ghh/Gvh, involving parameters directly obtained from the bender element
tests, is a ratio typically quoted in existing studies to evaluate the anisotropy. In case of
sand and clays, the values reported for this ratio have been normally larger than unity
(e.g., sand: [1, 3], clay: [6-8, 11]. Although some clays indicate small values for this
ratio, there is a general tendency that sands have smaller Ghh/Gvh ratios than clays.
Compared to these values, the values obtained in the present study, varying from 0.9
(i.e. T1-5) to 1.3 (i.e. T2-1) in both Group 1 and Group 2, are closer to those for sands.
The ratio Eh'/Ev' is less commonly studied for fine-grained soils than Ghh/Gvh.
Ch /Cv 2(1+'vh)Cvh /Cv Ch /Cv 2(1+'vh)Cvh /Cv

T1-3
T1-4
T1-5
T2-1
T2-6
S2-3
Uniform clays [11]

1 1
Natural
Reconstituted
2 2
3Cvh /Cvu Chh /Cvh 3Cvh /Cvu Chh /Cvh
Group 1 Group 2
Figure 4. Anisotropy patterns of small-strain stiffness expressed by four indices

The existing studies mainly focused on coarse-grained materials, since the

evaluation of the drained stiffness for fine-grained soils is technically difficult [11].
These studies have shown a range of Eh'/Ev' values for granular material between 0.8
[3] to 1.2 [1]. However, some existent studies carried out on clay suggested that this
ratio was much larger (e.g., around 4 for the Gault clay; [6], and around 2 for the
London clay; [8]. The results obtained in this study were between 1.2 to 1.4, with the
exception of sample T1-5, for which the value was 0.6 (Table 2). Again, this range
matches better that for sands than for clays.
4.3. Stress-Dependency of the Elastic Moduli
The changes of the stress states can modify the stiffness anisotropy patterns, incurring
induced anisotropy. The parameterization via the cross anisotropic model given by Eqs.
(2) to (5) allows evaluating the sensitivity of the individual moduli to a stress change
by considering the values for its exponents (i.e. mv, mh, nhh and nvh). The values are
shown in Table 2 and graphically in Figure 5. Most of them ranged between 0.10 to
0.35 and more typically between 0.2 and 0.3, except nhh and nvh for the sample T1-5,
which showed very small values, and mh for S2-3, which on the contrary showed a
value as large as 0.50. Due to the small number of tests and these outlier data, it is still
difficult to discern a clear trend of these exponents against either depth or lithology. It
is at least found, however, that the variability in these exponents at this site is much
larger than for uniform clays. The data obtained from a previous study [11] show that
these four exponents for sedimentary clays exhibited relatively narrow range, between
0.12 and 0.31 for 17 tests. In some existing studies on soft rock and clays [7, 11], these
exponents showed a vague tendency to be smaller for older materials, therefore
decreasing with depth. In this study, however such a trend was not observed; this may
be because the variability of lithology was so large that any subtle effect of ageing is
masked. This fact may indicate, again, that the influence of gradation and lithology is
the primary factor to determine the anisotropic characteristics of stiffness in soils.
Stiffness Model exponent (mi & ni(j)) Percentage (%)

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 10 20 30 40 50 60 70 80 90100
-20
mv
mv
-22
mh
mh
-24 nhh
nhh
nvh
Elevation (m a.s.l.)
-26 nvh
T1-5
-28
-30
S2-3
-32
-34
-36 Sand
-38 Clay
Silt
-40
Figure 5. Stiffness model exponents against sample depth, shown with gradation cariation
5. Conclusions
This study investigated the anisotropic characteristics of natural non-uniform offshore

soils which are mostly categorized as intermediate soils. The inherent anisotropy
patterns were evaluated by four appropriately chosen indices and graphically presented.
The comparison of the anisotropy patterns of soils of similar types at different depths,
and soils of different types at similar depths, showed that the depth (therefore ages) did
not pose a clear influence on the inherent anisotropy patterns. This observation
suggests that the primary factor governing the inherent anisotropy characteristics is the
soil lithology or gradation rather than the level of effective stresses and age.
The inherent anisotropy characteristics were further discussed by looking at two
more commonly quoted inter-modulus ratios, Ghh/Gvh and Eh'/Ev'. They were found to
range 0.9-1.3 and 0.6-1.4, respectively. These ranges suggest relatively modest inherent
anisotropy in comparison with the values from existing studies, which reported
particularly large values for clays. The ranges are closer to those reported for
reconstituted sands than those for clays. There is a possibility that larger particles are
more dominant in determining the stiffness anisotropy patterns in non-uniform
intermediate soils.
The exponents in the adopted model, indicative of the sensitivity of the moduli to
effective stress changes, showed large scatters, varying from 0.05 to 0.5 and no clear
trend against depth was detected. The reason could be due to the variability of the soils
at the studied site.
References
[1] R. Bellotti, M. Jamiolkowski, D.C.F. Lo Presti and D.A. ONeill. Anisotropy of small strain stiffness in
Ticino sand, Gotechnique 46 (1996) 115-131.
[2] R. Kuwano, T.M. Connolly and R.J. Jardine, Anisotropic stiffness measurements in a stress-path triaxial
cell, Geotechnical Testing Journal 23 (2000) 141-157.
[3] R. Kuwano and R.J. Jardine, On the applicability of cross-anisotropic elasticity to granular materials at
very small strains, Gotechnique 52 (2002) 727-749.
[4] L. Zdravkovi, The stress-strain-strength anisotropy of a granular medium under general stress
conditions, PhD Thesis, Imperial College, University of London, 1996
[5] R.C.K. Wong, D.R. Schmitt, D. Collis and R. Gautam, Inherent transversely isotropic elastic parameters
of over-consolidated shale measured by ultrasonic waves and their comparison with static and acoustic
in situ measurement, Journal of Geophysics and Engineering 5 (2008)103-117.
[6] M.L. Lings, D.S. Pennington and D.F.T. Nash, Anisotropic stiffness parameters and their measurement
in a stiff natural clay, Gotechnique 50 (2000) 109-125.
[7] K. Piriyakul, Anisotropic Stress-Strain Behaviour of Belgian Boom Clay in the Small Strain Region,
Ghent University, Belgium, 2006.
[8] A. Gasparre, S. Nishimura, N. Anh-Minh, M.R. Coop and R.J. Jardine, The stiffness of natural London
Clay, Gotechnique 57 (2007) 33-47.
[9] A. Brosse, R. Hosseini Kamal, R.J. Jardine and M.R. Coop, Measuring the static and dynamic small
strain stiffness of UK mudrocks, Proc. 15th ECSMGE, Athens, (2011) 137-142.
[10] R. Hosseini Kamal, Experimental study of the geotechnical properties of UK mudrocks, PhD Thesis,
Imperial College London, 2012
[11] S. Nishimura, Small-strain stiffness characteristics of natural sedimentary clays interpreted with cross-
anisotropic elasticity model, Gotechnique 64 (2014b) 981-996.
[12] M. Jamiolkowski, R. Lancellotta and D. C. F. Lo Presti, Remarks on the stiffness at small strains of six
Italian clays, Proc. 1st Int. Conf. Pre-failure Deformation Characteristics of Geomaterials, Sapporo,
Vol.1 (1995) 817-836.
[13] S. Nishimura, Assessment of anisotropic elastic parameters of saturated clay measured in triaxial
apparatus:Appraisal of techniques and derivation procedures, Soils and Foundations 54(2014a)364-376.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-832
Strain localization characteristics of

liquefied sands in undrained cyclic
torsional shear tests
Gabriele CHIARO a,1, Takashi KIYOTA a, Junichi KOSEKI b
a
Institute of Industrial Science, University of Tokyo, Japan
b
Department of Civil Engineering, University of Tokyo, Japan
Abstract. In this paper, results of undrained torsional shear tests on Toyoura sand,
Urayasu boiled sand and Christchurch boiled sand specimens were examined to
evaluate and clarify the combined effects of key factors (i.e. density, mean effective
stress, cyclic and static shear stresses) on the limiting value of shear strain at which
strain localization (i.e. shear band formation) appears in liquefied sandy soils. All
tests were conducted on fully-saturated medium-size hollow cylindrical specimens
(H = 300 mm, Dout = 150 mm and Din= 90 mm), which were prepared by air-
pluviation method, thus producing a sand fabric with horizontal bedding planes. A
state-of-the-art torsional shear apparatus was used to achieve a single amplitude
shear strain level of 50%. Non-uniform specimen deformation was clearly observed
in all tests and shear strain level to initiate strain localization could be methodically
determined from test results based on the change in deviator stress properties during
cyclic loading.
Keywords. Sandy soils, strain localization, torsional shear, liquefaction, density,

stress state, undrained shear strength
1. Introduction
Failure of cohesionless materials, such as sandy soils, is typically the result of strain
localization, or to be precise the concentration of shear deformation into a narrow zone
of intense shearing commonly referred to as a shear band. Generally, the shear banding
process begins at the state when the mobilized strength of soil reaches its peak and
gradually develops as shear deformation continues. Full understanding of strain
localization mechanisms (formation and evolution) is essential to most geotechnical
problems, including liquefaction-induced large deformation [1]. However, this is not an
easy task due to the great number of factors that need to be considered, such as density
state, mean effective stress level, cyclic and static shear stresses, soil type etc.
Recent major earthquakes, including the 2011 off the Pacific Coast of Tohoku
earthquake in Japan and the 2010-2011 Christchurch earthquakes sequence in New
Zealand, have led to severe damage as a result of liquefaction-induced ground failure. In
particular, extremely large horizontal ground deformation was observed in low-lying
areas (level ground to gentle slopes) near bodies of water such as rivers, lakes, bays and
1
Corresponding Author: Institute of Industrial Science, University of Tokyo, Be 206, 4-6-1 Komaba,
Meguro-ku, 153-8585 Tokyo, Japan; Email: [email protected].
G. Chiaro et al. / Strain Localization Characteristics of Liqueed Sands 833
oceans. The mechanisms behind such ground failure are not fully understood yet.
Nevertheless, a recent study by Kiyota et al. [2] suggested that the maximum amounts of
liquefaction-induced ground displacement observed in relevant model tests [3] are
consistent with the limiting value to initiate strain localization observed in torsional shear
tests [4]. These features are rational considering the reduction in the mobilized cyclic
shear stress in liquefied soils due to degradation of shear resistance [2].
In this paper, to provide new insights into this matter, results of a comprehensive
series of undrained torsional shear tests conducted by the Authors on Toyoura sand,
Urayasu boiled sand (Japan) and Christchurch boiled sand (New Zealand) specimens
were methodically examined to evaluate and clarify the combined effects of density,
mean effective stress, cyclic and static shear stresses on the limiting value of shear strain
at which strain localization (i.e. shear band formation) appears in liquefied sandy soils.
2. Test apparatus, material and procedure
A torsional shear apparatus using hollow cylindrical specimens is recognized to be a

good tool to properly evaluate liquefaction soil response [5]. In particular, it offers the
7UDQVGXFHUV possibility to reproduce simple shear
7ZRFRPSRQHQW ORDG FHOO condition, which is a close
'LVSODFHPHQW WUDQVGXFHU IRU ODUJH YHUWLFDO GLVSODFHPHQW
+LJK FDSDFLW\ GLIIHUHQWLDO SUHVVXUH WUDQVGXFHU representation of field stress state
IRU FRQILQLQJ VWUHVV
/RZ FDSDFLW\ GLIIHUHQWLDO SUHVVXUH WUDQVGXFHU during earthquakes.
IRU YROXPH FKDQJH
In this study, to achieve torsional
/RDGLQJ VKDIWPP
single amplitude shear strain (SA) level
%HDULQJ KRXVH exceeding 50%, the state-of-the-art
&HOO SUHVVXUH torsional shear apparatus on hollow
cylindrical specimens shown in Fig. 1
was employed. Extremely large

torsional deformations were measured
by using a potentiometer with a wire
3UHVVXUH FHOO and a pulley. The torque and axial load
%DFN SUHVVXUH
were detected by using a two-
7RS FDS
component load cell, which was
6SHFLPHQ
'R FP 3RURXV
VWRQH %XUHWWH
installed inside the pressure cell.
'L FP Several medium-size hollow
+ FP
3HGHVWDO
cylindrical specimens with dimension
of 150 mm in outer diameter (Dout), 90

mm in inner diameter (Din) and 300 mm

in height (H) were prepared by air
FP
pluviation method, thus producing a
sand fabric with horizontal bedding
/LQHDUPRWLRQ JXLGH UDLO 3RWHQWLRPHWHU ZLWK
ZLUH ZRXQG DURXQG SXOOH\
planes. To minimize the degree of
3UHSUHVVXUL]HG QXW
inherent anisotropy in radial direction
of hollow cylindrical sand specimens,
sample preparation was carried out
3XOOH\ WR 6WHHO EHOWV
WUDQVPLW WRUTXH
%DOO VFUHZ
/ FP
carefully by pouring the air-dried sand
FP
%DOO VSOLQH VKDIW
particles into a mold while moving
FP radially the nozzle of the pluviator and
Figure 1. Torsional shear apparatus on hollow at the same time circumferentially in
cylindrical specimens used in this study (after [2]) alternative directions i.e. first in clock-
834 G. Chiaro et al. / Strain Localization Characteristics of Liqueed Sands
wise and then anti clock-wise direction [6]. In addition, to obtain specimens with highly
uniform density, the falling height was kept constant throughout the pluviation process.
High degree of saturation (i.e. Skemptons B-values > 0.96) was achieved by the double
vacuum method [7] while circulating de-aired water into the specimens. The specimens
were isotropically consolidated by increasing the effective stress state up to a specific
value (i.e. p0 = 50 and 100 kPa), with a back pressure of 200 kPa. Subsequently, to
replicate seismic conditions, a constant-amplitude undrained cyclic torsional shear stress
(cyclic) was applied at a shear strain rate of 0.5%/min. The loading direction was reversed
when the amplitude of shear stress, which was corrected for the effect of membrane force
[8], reached the target value. During the process of undrained cyclic torsional loading
the vertical displacement of the top cap was prevented with the aim to simulate as much
as possible the simple shear condition that ground undergoes during horizontal excitation.
Note that, in some tests a drained monotonic shearing was applied before the undrained
one in order to achieve a specified value 100
of initial static shear (static),
representing sloped ground condition. 80
Percentage finer by weigth: %
Urayasu
The tests were performed on three boiled sand
sandy soils (Fig. 2), namely Toyoura 60

Christchurch
sand (Japan), Urayasu boiled sand boiled sand
(Japan) and Christchurch boiled sand 40
(New Zealand), as in Table 1. The list of

20
undrained torsional shear tests Toyoura sand
examined as well as details of testing
0
conditions are presented in Table 2. 0.01 0.1 1 10
Particle size: mm
Figure 2. Particle size distribution for tested materials
Table 1. Index properties of tested sandy soils
Material Mean Fines Max. void Min. void Specific Remarks

diameter content ratio ratio gravity
D50 (mm) Fc (%) emax emin Gs
Toyoura 0.19 0.1 0.992 0.632 2.656 Clean sand
Sand
Urayasu 0.11 30 1.261 0.683 2.683 Non-plastic fines
boiled sand
Christchurch 0.16 5.0 1.081 0.654 2.654 Non-plastic fines
boiled sand
Table 2. Cyclic undrained torsional shear tests examined in this study
Test Material Dr p0 cyclic static und L(SA) Ref.

(%) (kPa) (kPa) (kPa) (kPa) (%)
1 Toyoura 26.6 100 27 0 22.4 32.9 [2, 4]
2 Toyoura 25.4 100 17 0 21.8 27.8 [2, 4]
3 Toyoura 26.6 100 11 0 22.4 35.2 [2, 4]
4 Toyoura 26.7 100 25 0 22.4 31.9 [2, 4]
5 Toyoura 40.7 100 16 0 28.9 27.4 [2, 4]
6 Toyoura 47.7 100 25 0 32.1 19.3 [2, 4]
7 Toyoura 49.6 100 16 0 33.0 22.6 [2, 4]
8 Toyoura 60.7 100 37 0 38.1 11.4 [2, 4]
9 Toyoura 64.5 100 27 0 39.8 14.0 [2, 4]
10 Toyoura 88.3 100 40 0 50.8 2.25 [2, 4]
11 Toyoura 73.9 100 40 0 44.1 9.05 [2, 4]
12 Toyoura 16.8 100 43 0 17.9 45.8 [2, 4]
13 Toyoura 46.4 100 16 0 31.5 28.8 [1, 8]
Table 2. Cyclic undrained torsional shear tests examined in this study (continued)
Test Material Dr p 0 cyclic static und L(SA) Ref.

(%) (kPa) (kPa) (kPa) (kPa) (%)
14 Toyoura 45.5 100 16 5 32.0 24.5 [1, 8]
15 Toyoura 46.6 100 16 10 33.4 18.2 [1, 8]
16 Toyoura 44.2 100 16 15 29.2 16.1 [1, 8]
17 Toyoura 48.1 100 20 0 32.3 23.3 [1, 8]
18 Toyoura 48.0 100 20 5 33.1 22.4 [1, 8]
19 Toyoura 45.6 100 20 10 32.9 21.4 [1, 8]
20 Toyoura 44.4 100 20 15 26.3 20.8 [1, 8]
21 Toyoura 32.9 50 10 0 12.6 33.3 [9]
22 Urayasu 30.5 50 6 0 - 26.5 [9]
23 Urayasu 37.1 50 7 0 - 21.7 [9]
24 Urayasu 32.2 50 10 0 - 24.0 [9]
25 Christchurch 49.0 100 12 0 25.8 18.2 [10]
26 Christchurch 48.9 100 24 0 25.8 17.7 [10]
27 Christchurch 48.6 100 30 0 25.8 15.9 [10]
3. Limiting value of shear strain to initiate strain localization
In all tests listed in Table 2, irrespective of specimen densities and applied stress
condition, SA values progressively increased up to 50% during cyclic loading.
Significantly, non-uniform specimen deformation was observed at large level of SA.
However, the initiation of strain localization could not be defined merely on the basis of
visual observation. Hence, the shear strain limit to begin strain localization was
systematically evaluated based on the change in deviator stress properties during cyclic
loading [1, 2, 4, 8] as described in detail hereafter.
3.1. Specimen deformation and strain localization of sand in undrained cyclic torsional
shear tests
For the sake of completeness, in Fig. 3, typical specimen deformation observed at several
loading stages during undrained cyclic torsional shear tests is explained for the case of a
medium dense Toyoura sand specimen (photos refer to test 15 listed in Table 2). At J <
20%, although the outer membrane may appear slightly wrinkled, usually the
deformation is almost uniform (as shown by the red dotted line) except for the areas close
to the pedestal and the top cap that are affected by the end restraint. Then, at J = 20-30%,
the outer membrane is noticeably wrinkled and in the zone near the top cap the
deformation of the specimen started to localize (as shown by the blue dotted line) due
probably to water film formation [11]. Next, at J = 30-40%, the localization of the
specimen deformation is visibly developed in the upper part of the specimen.
Alternatively, in the bottom part the uniformity of the specimen deformation is generally
maintained even if many wrinkles appear. Finally, at J > 40%, the specimen is almost
twisted near the top cap.
3.2. Evaluation of limiting value of shear strain to initiate strain localization
By keeping the vertical (Vv) and horizontal (Vh) effective stress values constant,
Tatsuoka et al. [5] performed a series of drained monotonic torsional shear tests on
hollow cylindrical Toyoura sand specimens. They reported that the vertical strain
accumulated on the extension side due to the mobilization of positive dilatancy, and it
was reduced suddenly when the shear band was formed in the specimen.
< 20% = 20-30% = 30-40% > 40%

Figure 3. Typical torsional deformation of a medium dense Toyoura sand specimen during undrained cyclic
torsional shear tests (adopted from [1]).
Later, Kiyota et al. [2] found the change in deviator stress q (= Vv- Vh) response
observed in undrained torsional shear tests, in which any vertical displacement of the top
cap was prevented (i.e. to attain simple shear condition), to be consistent with the
behavior observed during drained monotonic torsional shear tests by Tatsuoka et al. [5].
Therefore, Kiyota et al. [2] considered the state at which the deviator stress amplitude
abruptly decreased (referred as state A henceforward) as the limiting state to initiate
formation of shear band and, thus, strain localization. Furthermore, Kiyota et al. [2]
found that the drop in q at state A is usually accompanied by an increase in the increment
of single amplitude shear strain ('JSA) at state B. More recently, looking into the cyclic
strain softening behaviour of Toyoura sand, Chiaro et al. [1] confirmed that the state A
is effectively the beginning of shear band(s) formation into specimen and the state B is
the beginning of residual stress state after full development of shear band(s).
It is important to mention that originally, Kiyota et al. [2] defined the limit shear
strain in terms of half of double amplitude shear strain L(DA)/2, while Chiaro et al. [1]
recommended that L(SA) is a more appropriate parameter when non-symmetric cyclic
shear stress conditions are considered (e.g. an initial static shear in applied). Nevertheless,
in the case of symmetric cyclic loading (i.e. zero static shear) L(DA)/2 and L(SA) are well
in accordance to each other [1].
For comprehensiveness, hereafter, typical experimental results are presented for the
case of Christchurch boiled sand (i.e. test 26 listed in Table 2). As shown in Fig. 4, during
undrained cyclic simple torsional shear loading (where the vertical displacement is
prevented) a sudden drop of q could be clearly observed at state A. According to Chiaro
et al. [1] and Kiyota et al. [2], the state A can be regarded as the limit shear strain to
initiate strain localization (L(SA)). As anticipated, state A was closely followed by the
increase of 'JSA after state B (Fig. 5). In view of that, these features imply that the stress-
strain characteristics of the specimen change as a result of the formation of shear band(s)
and the initiation of strain localization in the specimen.
8 48
JL(SA)=17.7% Christchurch boiled sand JL(SA)=19.1%
Dr= 48.9% B
A A
Torsional shear stress, W: kPa

p0'= 100 kPa
Deviator stress, q (=Vv-Vh): kPa
4
Wcyclic= 24 kPa 24
'J
Sudden
0 drop in q
0
Christchurch boiled sand

-4
Dr= 48.9%
-24 p0'= 100 kPa
Wcyclic= 24 kPa
-8
0 20 40 0 20 40
Torsional shear strain, J: % Torsional shear strain, J: %
Figure 4. Typical change in deviator stress Figure 5. Change in shear stress properties during
properties undrained cyclic torsional shear loading undrained cyclic torsional shear loading
4. Key factors affecting strain localization characteristics of liquefied sands
In this section, the test results listed in Table 2 are analyzed to establish possible
correlations between the limiting value L(SA) at which strain localization appears in
liquefied sandy soils and the extent of density, mean effective stress, cyclic and static
shear stresses.
4.1. Effects of density
Density is indeed one of the key parameters that governs the cyclic undrained behavior
of sand, including liquefaction resistance and large deformation development. Based on
Fig. 6, it clearly appears that density also greatly affects the strain localization
characteristics of liquefied sandy soils. In particular, the greater the relative density (Dr)
is, the smaller the limit shear strain to cause strain localization is. In other words, dense
sands tend to show strain localization at smaller SA levels compared to loose sands. This
is rational considering that dense sands usually show a dilative behavior during
undrained shear loading, while loose sands show a contractive behavior.
4.2. Combined effects of mean effective stress and cyclic shear stress
Mean effective stress and cyclic shear stress are other two important factors that greatly
affect the liquefaction behavior of sand. To reflect this aspect, the ratio between cyclic
stress ratio and mean effective stress, namely cyclic stress ratio (CSR), is commonly
employed to evaluate several features of the cyclic undrained soil behavior. In Fig. 7,
CSR is plotted against L(SA), and it can be seen that L(SA) is not affected by CSR itself
but is affected by the relative density. In fact, for a given CSR, L(SA) decreases if a denser
sand is considered.
4.3. Combined effects of cyclic and static shear stresses
In sloped ground, the presence of initial static shear may have major effects on the
behavior of liquefiable soil during earthquakes [1]. In particular, its superimposition with
cyclic shear stress generates a non-symmetric shear stress condition, which is usually
expressed in terms of degree of reversal stress (R) [12] as follows:
W cyclic W static
R (1)
W cyclic W static
As shown in Fig. 8, the combined effects of static and cyclic shear stresses on limit shear
strain to cause strain localization can be then conveniently evaluated in a plot R-L(SA).
For the case of Toyoura sand (only tests for Dr = 40-50% are considered), it seems that
the smaller the reversal stress extent, the smaller the limit shear strain to cause strain
localization.
5. Relationship between undrained shear strength and limiting value shear strain
The previous analysis revealed that different factors influence the strain localization
properties of liquefied sands. However, only density, mean effective stress and static
shear have a major effect on shear strain characteristics. In fact, contrarily to cyclic shear
stress, they directly affects the undrained shear behavior of sand [13]. Hereafter, with the
aim of describing the mutual effects of density, mean effective stress and static shear, a
more comprehensive assessment was attempted by plotting the experimental data in
terms of undrained shear strength ratio (und / p0) against L(SA) (Fig. 9). Note that
definition of und is schematically provided in Fig. 9. In this study, when feasible und was
experimentally evaluated, otherwise it was estimated by using the constitutive model for
liquefiable sands developed by Chiaro et al. [14].
Thus, from Fig. 9, it can be seen that L(SA) decreases with an increase in und. This
is rational since und increases with the density, mean effective stress and static shear [13].
Alternatively, it is clear that there is no correlation between L(SA) and cyclic shear stress.
100 0.5
Toyoura sand Dr= 16-40%

Po Urayasu boiled sand Dr= 41-60%
Cyclic stress ratio, CSR =Wcyclic/p0'
ss
ible 0.4 Dr= 61-88%
80 up Christchurch boiled sand
pe
r li
mit
Relative density, Dr: %
60 0.3
Dense Loose
sand sand
40 0.2
20 0.1
0 0.0
0 10 20 30 40 50 0 10 20 30 40 50
Limit shear strain to cause strain localization, JL(SA): % Limit shear strain to cause strain localization, JL(SA): %
Figure 6. Correlation between relative density and Figure 7. Correlation between CSR and limiting
limiting shear strain shear strain
1.2 0.6
Undrained shear strength ratio, Wund/p0': kPa
Toyoura sand W cyclic W static W

1.0 R 0.40
W cyclic W static 0.5 Po
ssib Wund
Degree of reversal loading, R
le u
ppe
0.40 r lim
0.8 it
0.4 0.27 0.20
0.37 0.20
0.16
0.6 0.16 p'
0.3 0.16 0.25 0.20
0.20 0.16
0.4 0.30 0.16 0.27
0.24 0.12 0.11

0.2 0.17
0.2 0.25
0.43
0.1
0.0
Dr = 40-50% CSR shown at each data point

-0.2 0.0
0 10 20 30 40 50 0 10 20 30 40 50
Limit shear strain to cause strain localization, JL(SA): % Limit shear strain to cause strain localization, JL(SA): %
Figure 8. Correlation between degree of reversal Figure 9. Correlation between undrained shear
loading and limiting shear strain strength ratio and limiting shear strain
Conclusion
In this paper, results of a wide series of undrained torsional shear tests conducted on
Toyoura sand, Urayasu boiled sand (Japan) and Christchurch boiled sand (New Zealand)
specimens were carefully examined to evaluate and clarify the effects of density, mean
effective stress, cyclic and static shear stresses on the limiting value of shear strain at
which strain localization begins in liquefied sandy soils. It was established that only
density, mean effective stress and static shear have a direct influence on shear strain
characteristics of sand. In fact, differently from cyclic shear stress, they directly affect
the undrained shear behavior of sand, which in this study was recommended as a
convenient factor to estimate the limiting value of shear strain to cause strain localization
in liquefied sands.
Acknowledgements
Laboratory assistance by Mr. Hirotoshi Miyamoto and Mr. Chiehyu Wu is greatly appreciated. The Japanese
Society for Promotion of Science is acknowledged for funding the first Authors research fellowship in Japan.
References
[1] G. Chiaro, T. Kiyota and J. Koseki, Strain localization characteristics of loose saturated Toyoura sand in
undrained cyclic torsional shear tests with initial static shear, Soils and Foundations 53 (2013), 23-34.
[2] T. Kiyota, J. Koseki and T. Sato, Comparison of liquefaction-induced ground deformation between results
from undrained cyclic torsional shear tests and observations form previous model tests and case studies,
Soils and Foundations 50 (2010), 421-429.
[3] M. Hamada, S. Yasuda, and K. Wakamatsu, Case studies on liquefaction-induced permanent ground
displacement, Proc. of 1st Japan-US Workshop on Liquefaction, Large Ground Deformation and their
Effects on Lifeline Facilities (1988), 3-21.
[4] T. Kiyota, T. Sato, J. Koseki and M. Abadimaranad, Behavior of liquefied sands under extremely large
strain levels in cyclic torsional shear tests, Soils and Foundations 48 (2008), 727-739.
[5] F. Tatsuoka, S. Sonoda, K. Hara, S. Fukushima and T.B.S. Pradhan, Failure and deformation of sand in
torsional shear, Soils and Foundations 26 (1986), 79-97.
[6] L.I.N. De Silva, J. Koseki and T. Sato, effects of different pluviation techniques on deformation property
of hollow cylinder sand specimen, Proc. of the International Symposium on Geomechanics and
Geotechnics of Particulate Media (2006), Ube, Japan 29-33.
[7] S.K. Ampadu and F. Tatstuoka, Effects of setting method on the behavior of clays in triaxial compression
from saturation to undrained shear, Soils and Foundations 33 (1993), 14-34.
[8] G. Chiaro, J. Koseki and T. Sato, Effects of initial static shear on liquefaction and large deformation
properties of loose saturated Toyoura sand in undrained cyclic torsional shear tests, Soils and
Foundations 52 (2012), 498-510.
[9] C. Wu, Y. Kondo and T. Kiyota, Effect of vertical cyclic loading history on liquefaction properties of sandy
soils in large strain torsional shear tests, ERS Bulletin, University of Tokyo, 46 (2013), 63-72.
[10] G. Chiaro, H. Miyamoto and T. Kiyota, Large deformation properties of reconstituted Christchurch silty
sand subjected to undrained cyclic torsional simple shear loadings, Proc. of the New Zealand Society of
Earthquake Engineering, Rotorua, New Zealand (2015), pp. 8 (in press).
[11] T. Kokusho, Formation of water film in liquefied sand and its effects on lateral spread, Journal of
Geotechnical and Geoenvironmental Engineering ASCE, 125 (1999), 817-826.
[12] Y. Yoshimi and H. Oh-oka, Influence of degree of shear stress reversal on the liquefaction potential of
saturated sand, Soils and Foundations 15 (1975), 27-40.
[13] G. Chiaro, J. Koseki and T. Kiyota, An investigation on the liquefaction-induced sloped ground failure
during the 1964 Niigata Earthquake, Geotechnical Hazards from Large Earthquakes and Heavy Rainfalls,
Springer Series in Geomechanics and Geoengineering (2015), pp. 8 (in press).
[14] G. Chiaro, J. Koseki and L.I.N De Silva, A density- and stress-dependent elasto-plastic model for sands
subjected to monotonic torsional shear loading, Geotechnical Engineering Journal SEAGS, 44 (2013):
18-26.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-840
Evaluation of Heterogeneities in
Reconstituted Samples using X-Ray
tomography
Luis Carlos Leguizamn Barretoa,b,1, Leonardo Flrez Valenciaa, Alfonso Ramos
Cana, Luis Felipe Prada Sarmientoa and Daniel Oswaldo Tique Alviara
a
Facultad de Ingeniera, Pontificia Universidad Javeriana
b
Facultad de Ingeniera, Universidad Pedaggica y Tecnolgica de Colombia
Abstract
In the process of characterization and modeling of the mechanical behavior of
geomaterials assumptions are adopted related to the theory of continuous medium,
such as homogeneity and isotropy condition derived from the stress-strain
response. Thus the development of element tests on undisturbed samples or
reconstituted in laboratory the actual evaluation of discretization and heterogeneity
of the environment are not considered, and their impact on the response of shear
strength, this fact extends to the numerical implementation of different constitutive
models through the application of finite element methods and discrete elements.
Among the methods for the reconstitution of sand samples in a laboratory we have
compaction, pluviation and moist tamping to determine the density levels achieved
in the manufacture of loose or dense homogeneous samples. These samples are
used in stregth tests such as the triaxial resistance without taking into account the
definition of an indicator to establish the real state of heterogeneity of the same.
The progress of prospecting techniques in medicine such as X-ray computed
tomography has transcended to other fields such as the quality assessment of
materials engineering. A proposal for assessing the condition of heterogeneity of
sand samples considering the main building techniques shown in this paper, based
on the development of tomographic images and the subsequent treatment of the
Hounsfield [HU] information density readings with the initial application of image
segmentation techniques that allows to obtain baseline data points and the
subsequent statistical analysis of variability considering different volumetric
groupings. So a reconstitution evaluation measure is reached of soil samples that
could be generalized to the analysis of undisturbed samples, allowing analysis of
relationship structure - mechanical response characterization tests and shear
strength. The results of this study show a high influence of the boundary
conditions in the heterogeneous spatial distribution of the density in reconstituted
samples with the methods considered, also allows to think about the use of this
methodology as a tool for reviewing and adjusting of such methods and its
application to the study of undisturbed samples.
This heterogeneity study could be incorporated into the constitutive
modeling through the definition of characteristic parameters of the model and a
coherent analysis of inherent and induced fabric tensors.
Keywords: Heterogeneity, image segmentation, inherent anisotropy, reconstituted

soil, sands, shear strength, tomography
L.C. Leguizamn, Facultad de Ingeniera - Pontificia Universidad Javeriana

Bogot, Colombia. e-mail: [email protected]

L.C. Leguizamn Barreto et al. / Evaluation of Heterogeneities in Reconstituted Samples 841
1. Introduction
In the study and analysis of the mechanical behavior of geomaterials different methods
are used for reconstitution of samples in the laboratory, the literature usually refers to
methods such as moist tamping, dry pluviation and water pluviation for reconstitution
of dense or loose sand samples, considering that lead to obtaining homogeneous
samples, with consistent mechanical response in stresses and deformations, represented
by a global state (void ratio, density or unit weight), in addition to assuming the
repeatability of the samples in a controlled environment given by the process of
reconstitution [15]. As well there are concerns about the accuracy of the procedures of
reconstitution in relation to the structure and the soil response modeling, further
complicated by the difficulties of the experimental techniques used.
The initial conditions of soil fabric, related to stress analysis in the called inherent
anisotropy, and their modification in the development of shear strength tests, related to
the induced anisotropy, are fundamental to the formulation and the performance of
constitutive models that attempt to replicate numerically the phenomena of instability
in laboratory samples representing scale problems that commonly occur in field [13].
During the modeling process, these conditions are represented in second-order tensors,
which are referred to as fabric tensors [5], [6], highlighting their influence on the
results of shear strength and the framework of Critical State Theory [7], [16].
Different adaptations geared to the constitution of the Anisotropic Critical State Theory
[12] and the study of anisotropic soil strength have led the proposal of constitutive
models among which may be mentioned more recent presented by Gao et al, 2014 [9],
derive their efficiency in the approach of the mentioned fabric tensor. These results
make it necessary to approach of an in-depth review of the assumptions of formulation
of the theoretical framework [4], and look for a rapprochement between the micro and
macro mechanics analysis used in the conceptual basis of DEM and FEM by means of
the fabric tensor [11].
Currently the constitutive models implemented consider stress tensor and deformation
at the elemental level that do not conform to the actual conditions of initial and
evolutionary behavior of the internal structure of undisturbed or reconstituted samples
of geomaterials that are subjected to monotonic or cyclic loading conditions. Some
previous studies have detected the spatial heterogeneity of the samples during the
development of elementary tests under different loading conditions, they have tried to
establish a correlation of the anisotropy present with the fabric tensor through shear
wave testing, that has been incorporated into constitutive modeling through the use of
random spatial distributions of density using statistical description and associated
probability [2].
The performance of simulations DEM has a high commitment in regard to their ability
to create the inherent anisotropy generated by laboratory methods for reconstitution,
providing a consistent simulation with the initial conditions and evolution of the
relationship stress-strain in soils [1]. It is indisputable the need to define an initial
fabric and its change or rotation to the principal axes of stress tensor during
development of elemental tests given its impact on the mechanical behavior of
geomaterials in dry, unsaturated or saturated conditions because of this are derived

842 L.C. Leguizamn Barreto et al. / Evaluation of Heterogeneities in Reconstituted Samples
from fabrics with different induced conditions of suction and/or generation of the
excess pore water pressure locally, that may give rise to different phenomena of
instability locally.
Points have to consider are the dependence of the boundary conditions present in the
formation of the samples and the inclusion of devices that aim to incorporate variables
such as the intensity of deposition and the drop height to generate conditions of
homogeneity in the resulting samples. In order to evaluate the homogeneity of the
initial conditions and the repeatability of the test conditions was necessary to develop a
series of tests for reconstitution of samples through moist tamping, dry pluviation and
pluviation dry flask through the incorporation of variation in height and intensity of
deposition of the material.
The validation of the initial internal conditions of the samples reconstituted was
conducted through the use of computerized axial tomography, given its potential as
non-destructive technique in the process of compilation and analysis of three
dimensional information and its growing use in different research fields of the internal
structure of materials of interest in some areas of Geosciences, that today also offers
the possibility of 4D monitoring by improving the conditions for the revision of the
state of the internal structure of the materials [3], [10]; this technique also has wide
acceptance in other areas that rely on technological advances such as the X-ray
microtomography increasing its potential exponentially. Then proceeded to the
statistical analysis through the application of tests and analysis of variance (ANOVA)
for the comparison of means and variances reaching results related to the classification
of the samples and the methods of reconstitution in regard to the degree of
heterogeneity in density.
2. Experimental Tests
For the review of the processes of reconstitution and the study of heterogeneity was
used Guamo Sand which contained sizes between the 0.075mm and 2mm, with average
diameter, D50, of 0.42 mm; taking into account the Unified Soil Classification System
it was found to be poorly graded sand with a uniformity coefficient (Cu) of 3.05 and
coefficient of curvature (Cc) of 0.63, with a specific gravity (Gs) of 2.629 0.03,
critical friction angle ( ) of 33,421,2; with predominantly sub-angular particles and
maximum void ratio (emax) and minimum (emin) of 0.966 and 0.601, respectively.
Samples were reconstituted in laboratory by method of air pluviation (8 samples), flask
pluviation (6 samples) and moist tamping (3 samples); in methods of pluvitaion was
used the device developed by Tique et al, 2014 [14], Figure 1, in order to normalize the
drop height (7 and 12.5) and the intensity of deposition for a considered range of void
ratios. For the reconstitution of the samples different spatial patterns were generated
through the use of perforated foils, as shown in Figure 1 (b), (c), (d) and (e), which
were placed in the position shown in Figure 1 (a).

Figure 1. (a) Device used for reconstitution by methods of pluviation, and perforated
sheets employed
(b) Sheet 1, A hole of diameter 1, (c) Sheet 2, A hole of diameter 0.5
(d) Sheet 3, Four holes of diameter 1, (e) Sheet 4, Five holes of diameter 0.5
By calculating global density of each of the samples, determining the initial, maximum
and minimum values of void ratio, relative density was estimated qualitatively with the
purpose of correlating qualitatively the degree of densification of the different samples,
in this way is shown in Figure 2 the relationship between reconstitution methods
employed and the relative density achieved is presented for each of the samples
considered. The results indicate that the method of pluviacion allowed to obtain
samples relatively very dense, the flask pluviation produced samples dense to loose,
while with moist tamping samples were achieved more loose, considering a global
perspective.
Figure 2. Relative density obtained in the process of reconstitution of samples in

laboratory

3. Tomographic tests and tridimensional reconstruction
Computed tomography is a noninvasive technique that allows the acquisition of

information from samples of different types of material, and that in a complementary
manner to its applications in medicine has found various applications in different fields
of Geosciences; at present it has computational tools for three-dimensional
reconstruction of CT scan results including micro levels. With the use of the software
3D Slicer is made the import of the files resulting from the study of tomography, after
generating the rendering in the volumes of the reconstituted samples, for which
established a scale of intensity values, Figure 3, and by the validation of previous
works [8] Eq (1a) was used to convert HU readings (H) to density () gr/cm3, or to unit
weight () kN/m3.
= 0,001 H + 1 = 0,0098 H + 9,8 (1)
In this way was possible to find patterns of transition at the local level of loose to dense
states arising from each process of reconstitution, as shown in Figure 3.
It was necessary the use of the color scale shown in Figure 3, which corresponds to
ranges of readings HU set at 3D Slicer and then interpolated to values of density, in
order to establish a high contrast device for detecting variations in density introduced in
samples in the implementation of the methods of reconstitution cited. Thus, it is
evident that the devices employed in the experiments of pluviation introduce patterns
that generate heterogeneity in the concentration of the material, in the same way the
method of tamping wet, despite having a loose almost uniform distribution shows that
the process of compaction by layers, also introduces localized heterogeneity through
the interfaces.
4. Image Segmentation
Based on the geometries found in the 3D Reconstruction, look Figure 3 the cluster
analysis through the heights, radios and sectors was raised, and the combination of
these factors for groups finest for subsequently perform hypothesis tests related to the
homogeneity in means and variances, thus it was considered the generation of groups
adjusted to 7 heights, 4 radio and 8 sectors, and radial ratio required was established,
Eq (2), compensating axial areas with the indication of n radius, ensuring an
approximate number of voxels in the count for each group. In this equation n indicates
the number of circular sections to consider, rn the value of the radio to the external
section, i the value of the number of evaluated section and ri the estimated value of the
radio for the section, look Figure 5 (c).
i
ri = rn (2)
n
With the information of the tomographic study was developed an applicative ITK with
the use of techniques of image segmentation to extract HU values of the pixels that are
available in the axial, coronal and sagittal slices available in each sample for the

Figure 3. Three dimensional reconstruction of the tomography in the reconstituted

samples using 3D Slicer
subsequent reconstruction of the corresponding voxels and then their adjustment to the
geometry shown in Figure 4, which allowed the necessary data for the quantitative
analysis and the assessment of the statistical own of the analysis of variance.

(a) (b) (c) (d)
(e) (f) (g) (h)

Figure 4. Generation groupings for analysis of heterogeneity in means and variances
5. Statistical Study of Heterogeneity
Thus the spatial distributions of density for each of the 7 geometries considered was
obtained, Figure 4 (b)-(h), on which the results of density variation in height, radio and
sector for each of the methods of reconstitution can be seen in Figure 5. In Figure 5 (a)
y (b), method of air pluviation with drop height 7 and 12.5, y (c) y (d), flask
pluviation for the same drop heights, in these graphs can be seen as boundary
conditions imposed in height and radio have high influence on the density distribution
by having areas more loose in these contours, while with the moist tamping is
appreciates density concentration at the interfaces of the different layers of samples 2
and 3, Figure 5(e), in accordance with 3D reconstruction shown in Figure 3(e); it can
also appreciate that the geometry associated with the sector has no significant effect on
the density distribution.
Additionally, it carried out the statistical evaluation of the heterogeneity of the samples
reconstructed by analysis of variance in means and variances (ANOVA); hypothesis
tests carried out are based on the comparison of group mean with respect to the global
sample mean of the data available, so the resulting Fisher statistical, conventionally
used for acceptance or rejection of hypothesis of homogeneity, has been used as
parameter categorizing the degree of heterogeneity of the samples, after being
subjected to normalization.
The ANOVA results for each independent factor can be seen in Figure 6 (a), (b) and
(c), and corresponding to the combination thereof in (d), the first are consistent with the
spatial distributions of density presented in Figure 5, while the latter can be considered
as representing the heterogeneity detected in each reconstituted sample, since it
involves three dimensional variability of density.

Figure 5. Spatial density distribution according to the method of reconstitution

Figure 6. Heterogeneity of mean densities to individual factors

(a) Height, (b) Radius, (c) Sector, (d) Height-Radius-Sector
The review of these results confirms the influence of the boundary conditions related to
the radius and height in the process of reconstitution of samples in the laboratory, the
use of such results together with the global relative density shown in Figure 2, allow to
conclude that with the method of air pluviation samples very dense but with high
variability are achieved, flask pluviation reconstituted samples with average and loose
densities with a intermediate variability and moist tamping produces samples more
loose with low heterogeneities; a comprehensive review of these results allows to
establish a connection between the physical process considered in the reconstitution
and spatial variability of the density.
6. Conclusions
It has been shown that the boundary conditions in the vertical and radial direction, are
sources of heterogeneity in the reconstitution of samples, the radial being the highest
incidence in the construction of the initial fabric. This paper has presented a
methodology based on a statistical and probabilistic framework for determining
conditions of spatial heterogeneity of density in soil samples reconstituted in
laboratory, that could be applied to undisturbed samples soil, that can be used for the
study of the anisotropy inherent, which can also serve for the study and the formulation
of a consistent fabric tensor for use in DEM and FEM, in addition to extend its use to
the study of evolution of the induced anisotropy. The results presented can serve as a
basis to review and adjust the reconstitution processes of samples in the laboratory so

as to reflect the natural conditions of construction of geomaterials, and allows use in a

coherent study the behavior of shear strength. The use of microtomography enables the
implementation of similar methodologies in the study of materials such as silt and clays
where the particle size requires a high degree of refinement to capture an appropriate
micro behavior.
References
[1] M.L. Bernhardt, C. OSullivan, and G. Biscontin, Effects of sample preparation methods in DEM,
Geomechanics from Micro to Macro, Taylor & Francis Group, London, (2015), 3-19.
[2] R.I. Borja, X. Song, A.L. Rechenmacher, S. Abedi and W. Wu, Shear band in sand with spatially varying
density, Journal of the Mechanics and Physics of Solids 61 (2013), 219-234.
[3] V. Cnudde and M. Boone, High-resolution X-ray computed tomography in geosciences: A review of the
current technology and applications, Earth-Science Reviews 123 (2013), 1-17.
[4] Y.F. Dafalias and X.S. Li, Revisiting the Paradigm of Critical State Soil Mechanics: Fabric Effects,
Constitutive Modeling of Geomaterials (2013), 13-26.
[5] Y.F. Dafalias and M.T. Manzari, Simple plasticity sand model accounting for fabric change effects,
Journal of Engineering Mechanics 130(6) (2004), 622-634.
[6] Y.F. Dafalias, A.G. Papadimitrioy and X.S. Li, Sand plasticity model accounting for inherent fabric
anisotropy, Journal of Engineering Mechanics 130(11) (2004), 1319-1333.
[7] N. Della, M. Belkhatir, A. Arab, J. Canou and J.C. Dupla, Effect of fabric method on instability behavior
of granular material, Acta Mechanica 225(7) (2014), 2043-2057.
[8] M.J. Duchesne, F. Moore, B.F. Long and J. Labrie, A rapid method for converting medical Computed
Tomography scanner topogram attenuation scale to Hounsfield Unit scale and to obtain relative density
values, Engineering Geology 103 (2009), 100-105.
[9] Z. Gao, J. Zhao, X.S. Li and Y.F. Dafalias, A critical state sand plasticity model accounting for fabric
evolution, International Journal for Numerical and Analytical Methods in Geomechanics 38 (2014),
370-390.
[10] S.A. Hall, M. Bornert, J. Desrues, Y. Pannier, N. Lenoir, G. Viggiani and P. Besuelle, Discrete and
continuum analysis of localised deformation in sand using X-ray CT and volumetric digital image
correlation, Gotechnique 60(5) (2010), 315-322.
[11] X. Li and X.S. Li, Micro-Macro Quantification of the Internal Structure of Granular Materials, Journal
of Engineering Mechanics 135(7) (2009), 641-656.
[12] X.S. Li and Y.F. Dafalias, Anisotropic Critical State Theory: Role of fabric, Journal of Engineering
Mechanics 138 (2012), 263-275.
[13] A.G. Papadimitriou, Y.F. Dafalias, M. Yoshimine, Plasticity modeling of the effect of sample
preparation method on sand response, Soils and Foundations 45(2) (2005), 109-123.
[14] D. Tique, A. Ramos, L.F. Prada and H.A. Vacca, Estudio experimental de la inestabilidad difusa para la
arena del Guamo Tolima. Tesis de Maestra. Pontificia Universidad Javeriana - Sede Bogot, (2014).
[15] D. Wanatowski and J. Chu, Effect of specimen preparation method on the stress-strain behavior of sand
in plane-strain compression tests, Geotechnical Testing Journal 31(4) (2008), 308-320.
[16] J. Zhao and N. Guo, Unique critical state characteristics in granular media considering fabric
anisotropy, Gotechnique 63(8) (2013), 695-704.

IOS Press, 2015
doi:10.3233/978-1-61499-601-9-850
Inherent Anisotropy of an Undisturbed and

Compacted Loess Soil
Pedro A. COVASSIa,1 and Victor A. RINALDIa
a
Facultad de Ciencias Exactas, Fsicas y Naturales. Universidad Nacional de Crdoba
Abstract. This work discusses the result of an experimental study on inherent

anisotropy of the loess soil found at Crdoba, Argentina. A cubical true triaxial
device provided with flexible boundaries was used and the samples were subjected
to an isotropic compression stress path. A brief description of the true triaxial
device as well as the procedure for its assemblage and testing procedure is
presented. Tests were performed on undisturbed and compacted samples.
Undisturbed block type samples were recovered from open trenches. Compacted
samples were made either by tamping or static compaction. The effect of
compaction method on structural anisotropy was evaluated. Samples were
prepared at dry of optimum, optimum and wet of optimum. Isotropic compression
tests show that inherent structural anisotropy is found in both undisturbed and
compacted specimens. Undisturbed samples are less compressible and develop
higher degree of strain anisotropy respect the compacted samples at the same dry
unit weight.
Keywords. Inherent Anisotropy, Loess, True Triaxial Test, Compaction method.
1. Introduction
Argentinean loess in one of the largest deposits on the world, with thickness varying
between 25 and 60 m ([1], [2], [3], [4]). Loess is an eolian low dense formation
(usually densities range between 11.5 kN/m3 and 13.5 kN/m3) composed by fine sand
and silt particles and a minor clay fraction (mainly illite and montmorillonite) weakly
bonded. Under natural conditions, loess is able to withstand vertical slopes and to
sustain moderate stresses without significant deformations. In this state, both stiffness
and strength of this soil is governed by matric suction and cementation provided by
precipitated salts at particle contacts. Upon saturation, soluble salts dissolve, capillary
forces decrease, the structure weakens and collapse takes place, sometimes under self-
weight of the soil. Actually, it is difficult to differentiate the individual role of
cementation and suction in the collapse process [3].
Due to the large extent of loess deposits of Argentina, this soil is used in most
earthworks including excavations, tunneling, liners for waste and wastewater
containment, embankments, backfills, core dams, and hydraulic barriers. Improvement
of mechanical behavior of loess is usually achieved by means of static and dynamic
compaction in the field. The compaction procedure destroys cemented bonds and
consequently the collapse potential is significantly reduced [3].
At present, there are several research works on Argentinean loess mainly focused
in understanding the collapse mechanism, the relationship between collapse and soil
structure, the stress-strain behavior in situ and laboratory, and the behavior of
maximum shear modulus (Gmax) ([5], [6], [7], [8], [9], [10]). Limited research has been
1
P.A. Covassi and V.A. Rinaldi / Inherent Anisotropy 851
done for saturated undisturbed samples in drained conditions using the conventional
triaxial test device [11], [12].
This work presents an experimental study performed to evaluate the inherent
anisotropic behavior on undisturbed and compacted loess samples. A series of isotropic
compression tests over natural undisturbed and compacted samples were done in a
flexible boundaries cubical true triaxial device.
2. Soil Description and Testing Program
The soil used here includes both undisturbed and compacted loessial samples obtained
from the loess formation in Crdoba city, Argentina. Undisturbed block samples were
retrieved from a vertical wall at approximately 6 m depth of an open trench located at a
highway construction project in the North-West of Crdoba City. The samples were
placed in plastic bags and stored in a conditioned room for preservation of initial water
content and temperature. An average dry unit weight of d = 13.4 kN/m3 for
undisturbed specimens was measured. Natural moisture content of undisturbed samples
was around wnat = 12.5 %. Compacted samples were made from air dried soil, sieved
through sieve No. 40 (425 m) and mixed thoroughly. The standard Proctor test
yielded a maximum dry unit weight of d = 16,3 kN/m3 at the optimum moisture
content of wopt = 18,7 %. Other relevant physical properties of the soils tested are
summarized in Table 1.
Table 1. Physical properties and classification of soils tested.
Passing Passing Passing
Sieve Sieve Sieve Clay
#40 #60 #200 Content
PL LL PI [%< 425 [%< 250 [%< 74 [% < 2
Soil [%] [%] [%] Gs m] m] m] m] SC
Compacted 18.8 24.3 5.5 2.67 100 92.2 81.3 15.1 ML
Natural NP 21.7 NP 2,66 100 98.5 90.3 9.6 ML
Note: PL: Plastic Limit; LL: Liquid Limit; PI: Plasticity Index; Gs: Specific Gravity; SC: Unified Soil
Classification; NP: Nonplastic.
Compacted specimens were prepared by two compaction methods: dynamic and

static. The soil used for compacted specimens was mixed with the desired amount of
water, placed in a hermetic recipient and allowed to reach equilibrium for 24 hr. Each
specimen was compacted in four layers of approximately the same height [13]. The
dynamically compacted specimen was prepared by wet tamping in a two-piece split
cubical steel mold of 80 mm in side dimension. Compaction energy of 1.51 kg.cm/cm3
(corresponding to a quarter of the energy of Standard Proctor Test, ASTM D 698) was
applied and the moisture content of 16.4% (dry side of the Standard Proctor curve) was
selected. Figure 1 shows the Standard Proctor curve for the tested soil and the initial
density and water content conditions of prepared specimens. Static compacted
specimens were prepared in a two-piece split cubical acrylic mold of 80 mm in side
dimension using the compaction procedure described by Cui and Delage [14].
Compaction was performed in four layers at a compression rate of 1 mm/min in a
triaxial loading frame until the average value of d = 13.4 kN/m3 was obtained for all
the compacted samples and at three different moisture contents (see Figure 2). Table 2
852 P.A. Covassi and V.A. Rinaldi / Inherent Anisotropy
summarizes the initial conditions of the different soil specimens prepared and tested in
the present work.
17,0
3 Saturation
16,5 d max = 16,25 kN/m
curve
16,0
Dry Unit Weight gd [g/cm3]
15,5
w% opt = 18,7 %
15,0
14,5 Dynamic Compaction

Static Compaction
14,0
Undisturbed
13,5
13,0
12,5
12 14 16 18 20 22 24 26
Moisture Content, w %
Figure 1. Compaction moisture content and dry unit weight of the different specimens prepared.
Table 2. Initial state conditions for the different specimens tested in this work.
Max. Stress Level

Measured in Static
d wcomp/nat Compaction
Group Sample Designation [kN/m3] [%] [kPa]
Dynamic Compaction D-D (Dry of optimum) 13.4 16.4 -
S-D (Dry of optimum) 13.3 16.5 74
Static Compaction S-O (optimum) 13.4 18.7 40
S-W (Wet of optimum) 13.4 20.1 24
S1 13.5 12.3 -
Undisturbed Loess
S2 13.4 12.7 -
A flexible boundaries type cubical true triaxial apparatus developed in the

Geotechnical Laboratory of National University of Crdoba was used for testing. The
design was based on similar devices described elsewhere ([15], [16] [17]). Figure 2
sketches the cubical cell and peripherals. The cubical triaxial apparatus allows for the
application of three mutually and independently perpendicular principal stresses. In this
device, the aluminum reaction frame holds an 80 mm cubical specimen. Drainage,
vacuum, and pore water pressure monitoring ports were drilled diagonally as shown on
the same Figure 2. Each side wall assembly contains a linear variable deformation
transducer (LVDT) for measuring deformations at the center of each face of the
specimen and a in. hole to apply the air pressure. Three of the wall assemblies also
provide in. outlets to connect a pressure transducer. All electronic transducers are
connected to a computer. A specific software was developed for tests control and data
storage. A Tri-Flex 2 Master Control Panel (ELE International) is connected to the
drainage ports for pressure control and saturation of the specimens.
The use of flexible membranes allows boundary stresses to be uniform over the
faces of a specimen. In this work, the non-uniform distribution of deformations at each
sample face was evaluated using a simple profilometer [16] and a high density
expanded polystyrene cubical specimen to check the displacements contours
experienced under isotropic compression at two stress levels, 50 and 100 kPa. The
results of this calibration allows to conclude that the displacements contours are
reasonably uniform at relative large strains and that the displacement values obtained in
the center point of the membrane using LVDTs are reliable and representative in a
similar fashion as obtained by other authors [16], [18], [19]. Strain corrections related
to membrane compression were incorporated in strain computations, although they
were found to be very small respect to the strain levels measured in this work.
Figure 2. Test setup of the Cubical True Triaxial Device used in this work.
The prepared specimens were installed in the cubical triaxial cell, and all the parts
of the true triaxial device were assembled. Then, the specimens were permeated for 24
hours with de-aired water from the lower drainage port. Saturation of the specimen was
increased using back pressure until de pore pressure parameter B was above 0.90 at the
applied back pressures of 100 kPa. After saturation, samples were tested under an
isotropic drained compression stress path. Initial effective stress level for all tests was
set at 20 kPa with stress increments of 20 kPa up to 100 kPa. After this value was
reached, stress increments were of 50 kPa. Each stress increment was maintained until
primary consolidation ended which, for most samples, took place after 30 min.
3. Experimental Results and Discussions
Figure 3 shows the measured isotropic compressibility curves for the principal
directions (x, y and z) for the two undisturbed loess specimens tested. Here, z direction
(Z) is always perpendicular to the bedding plane and coincident with respect the in-situ
vertical direction, while the strains reported in x and y directions (X and Y) are parallel
to the bedding plane and coincident with the in-situ horizontal directions.
Mean effective stress, p' [kPa] Mean effective stress, p' [kPa]
1 10 100 1000 1 10 100 1000

0,00 0,00
Undisturbed Loess (a) Undisturbed Loess (b)
S1 0,50 S2
0,50
Depth 6,0 m Depth 6,0 m
Principal strain [%]

3
d = 13.3 kN/m3 1,00 d = 13.5 kN/m
1,00
1,50 X 1,50
Y X
2,00 2,00
Z Y
Z
2,50 2,50
3,00 3,00
Figure 3. Compressibility curves of two undisturbed samples of loess under isotropic

test conditions.
Results of Figure 3 show that compressibility of loess is higher in the horizontal

directions respect to the vertical direction. Similarly, yielding is clearly observed in all
curves being yielding pressure in vertical direction much higher than that in the
horizontal directions. Also small differences are observed between the horizontal
compressibility curves. Although the differences between compressibility curves in the
horizontal direction are small, such differences exists and it may be possibly attributed
either to: a) the in-situ state of stress conditions experienced by the soil in the face of
the open trench after excavation (the y direction is parallel to the face of the trench) or
b) to the deposition direction of the disk shape silt particles which main axis is usually
oriented to S-N direction in coincidence with the dominant deposition winds at the site.
Figure 4 shows the measured isotropic compressibility curves in the principal
directions (x, y and z) for dynamic and static compacted specimens at similar initial
water content (dry of optimum) and unit weight. Here, compressibility is much higher
than that corresponding to the undisturbed specimens at the same dry unit weight and
moisture content. Additionally, compressibility in the horizontal directions x and y are
quite similar and higher than that corresponding to the vertical direction. The results in
this case clearly show that the behavior of compacted specimens is cross-anisotropic
with stiffer response in the vertical direction. Observe that yielding is not well defined
or may be very low in all cases.
Mean effective stress, p' [kPa] Mean effective stress, p' [kPa]
1 10 100 1000 1 10 100 1000
0,00 0,00
1,00 Tamping Compaction 1,00 Static Compaction

D-D (Dry of optimum) S-D (Dry of optimum)
2,00 w comp = 16.4% 2,00 w comp = 16.5%
3,00 d = 13.4 kN/m3 3,00 d = 13.3 kN/m3
4,00 4,00
X
5,00 5,00 X
Y
6,00 Z 6,00 Y
Z
7,00 7,00
8,00 8,00
Figure 4. Compressibility curves of two compacted specimens under isotropic

compressions stress path: (a) tamping compaction (dynamic); (b) Static compaction.
The deformation anisotropy can also be evaluated by comparing the strains in

different directions. The difference between strains on each direction was determined
using the average relative difference (ARD) [20], which for the x direction can be
written as:
<6 1 i n x ,i av ,i 96 (1)
ARD [%] ; = 8100
6: n i 1 x ,i 67
Where n is the number of stress loading steps, x,i is the strain in a specific
direction (x in this case) for a specific stress level n = i, and av,i is the average strain
for the same stress level i. Alternating the x index by y and z, ARD values for the other
two directions could be calculated. Notice that ARD varies between 0% for isotropic
behavior and any positive value for an anisotropic response. The higher the ARD value
the higher the anisotropy. The values of ARD calculated respect to the three directions
for all the tests performed in this work are displayed on Table 3 being the z direction
coincident with the vertical direction. From the results of Table 3, it is observed that the
higher ARD values are obtained respect to z directions while the values in the x and y
directions being similar and much lower than for the z direction. The highest ARD
corresponds to the undisturbed specimens. Also, undisturbed specimens of loess show
similar behavior.
Table 3. Summary of different specimens tested in this work.

ARD
Group Sample Designation Comparison between strains [%]
X to the mean* 18
D-D (Dry of
Dynamic Compaction
optimum) Y to the mean* 14
Z to the mean* 65
X to the mean* 18
S-D (Dry of optimum) Y to the mean* 15
Z to the mean* 71
X to the mean* 15
Static Compaction S-O (optimum) Y to the mean* 25
Z to the mean* 46
X to the mean* 10
S-W (Wet of
optimum)
Y to the mean* 15
Z to the mean* 68
X to the mean* 28
S1 Y to the mean* 11
Z to the mean* 106
Undisturbed Loess
X to the mean* 25
S2 Y to the mean* 12
Z to the mean* 105
Note: * mean = (X +Y +Z)/3
Lade and Abelev [21] used the angle of inclination of the total strain increment
vector to the hydrostatic axis () to characterize the anisotropic behavior under an
isotropic compression stress path, which is defined as:
2
tan 1 3
(2)
2
Where =dz/dvol for total strain increments. The angle is considered positive
when the strain increment is pointing downwards from hydrostatic axis. This
corresponds to the case for cross-anisotropic soil deposits that are stiffer in vertical
direction. A value of zero in corresponds to an isotropic material. The increasing
mean confining pressure tends to override the effect of anisotropy.
Figure 5 shows the variation of with confining effective stress for all test
performed on compacted specimens in this work. Curves that best approximate the
experiments for each test are also included (dashed lines). Results show decreasing
values of with increasing isotropic effective pressure. Also, for the pressure range
used in this work, no specimen developed isotropic behavior.
30
S-D (Dry of optimum)

25
Inclination angles of total strain increment
S-O (Optimum)
vector to the hydrostatic axis, [deg.]
S-W (Wet of optimum)

20 D-D (Dry of optimum)
15
10
0
0 50 100 150 200 250 300
Me an e ffe ctive stress, p [kPa]
Figure 5. Variations of inclination angle with mean confining pressure for

compacted specimens.
4. Conclusions
Inherent anisotropy of natural and compacted loess soil under an isotropic compression
stress path have been evaluated. Results of isotropic compression tests show that
compacted loess soil behaves as a cross-anisotropic material, presenting a higher
stiffness in vertical direction. Anisotropy was also analyzed by means of the relative
difference (ARD) of principal strains and the inclination angle of total strain increment
vector (
. These approaches showed that anisotropy seems to be unaffected by the two
compaction methods and sample moisture. Undisturbed loess specimens seem to
behave as orthotropic materials. However, differences between the horizontal strains
are much smaller than the difference between the vertical and the average horizontal
strains.
References
[1] R.J. Rocca, Review of Properties of Loess Soils, CE 299 Report presented at the University of California,
Berkeley, CA, 1985.
[2] L. Moll and R.J. Rocca, Properties of Loess in the Center of Argentina, Proceedings of the IX
Panamerican Conference on Soil Mechanics and Foundation Engineering, Chile (1991), Vol. 1, 1-13.
[3] V.A. Rinaldi, J.J. Clari and J.C. Santamarina, The Small-strain Shear Modulus (Gmax) of Argentinean
Loess, Proceedings of the 15 International Conference on Soil Mechanics and Geotechnical
Engineering, Istambul, Turkey (2001), 495-499.
[4] R.J. Rocca, E.R. Redolfi and R.E. Terzariol, Geotechnical Characteristics of Argentinean Loess, Rev. Int.
de Desastres Naturales, Accidentes e Infraestructura Civil, Vol. 6 (2) (2006), 149-166. (In Spanish).
[5] A. Reginatto, Standard Penetration Test in Collapsible Soils, Proceedings of the IV Panamerican
Conference on Soil Mechanics and Foundation Engineering, Puerto Rico, (1971), Vol. II, 77-84.
[6] V.A. Rinaldi, J.C. Santamarina and E.R. Redolfi, Characterization of Collapsible Soils With Combined
Geophysical and Penetration Testing. Symposium In-Situ Characterization of Soils, Atlanta USA
(1998). Vol. 1, 581-588.
[7] V.A. Rinaldi and J.J. Clari, Low Strain Dynamic Behavior of a Collapsible Soil, Proceedings of the XI
Panamerican Conference on Soil Mechanics and Foundation Engineering, Foz de Iguazu, Brazil (1999),
Vol. 2, 835 841.
[8] V.A. Rinaldi and J.A. Capdevila, Effect of Cement and Saturation on the Stress Strain Behavior of a Silty
Clay. Proceedings of Fourth International on Unsaturated Soils, G.A. Miller, C.E. Zapata, S.L. Houston,
and D.G. Fredlund eds. ASCE (2006), VA, Vol 2, 1157-1168.
[9] F.M. Francisca, Evaluating the Constrained Modulus and Collapsibility of Loess from Standard
Penetration Test, International Journal of Geomechanics (2007), 7(4), 307-310.
[10] V.A. Rinaldi, R.J. Rocca and M.E. Zeballos, Geotechnical Characterization and Behavior of
Argentinean Collapsible Loess, on: Characterization and Engineering Properties of Natural Soils, Tan T.
S. et al. eds. (2007), Balkema, London, Vol. 4, 2259-2286.
[11] J.A. Capdevila, Personal communication, 2014.
[12] J.A. Capdevila and V.A. Rinaldi, Local and Global Measurement of Strains in Triaxial Cell: Influence
of Soil Conditions, Proceedings of the XIII Panamerican Conference on Soil Mechanics and
Foundation Engineering, Isla Margarita, Venezuela (2007).
[13] V. Sivakumar and S.J. Wheeler, Influence of compaction procedure on the mechanical behaviour of an
unsaturated compacted clay. Part 1: Wetting and isotropic compression, Gotechnique 50 (2000), No. 4,
359368.
[14] Y.J. Cui and P. Delage, Yielding and Plastic Behaviour of an Unsaturated Compacted Silt,
Gotechnique 46 (1996), No. 2, 291-311.
[15] J.H. Park, Performance and Check-out Verification Testing of a New True Triaxial Apparatus using
Partially Saturated Silty Sand, M. Sc. Thesis The University of Texas at Arlington (2005).
[16] C.H. Choi, P. Arduino and M.D. Harney, Development of a True Triaxial Apparatus for Sands and
Gravels, Geotechnical Tesing Journal, Vol. 31, No. 1, 2008, 1 13.
[17] L.R. Hoyos, D.D. Prez-Ruiz and A.J. Puppala, A refined True Triaxial Apparatus for Testing
Unsaturated Soils under Suction-controlled Stress Paths, International Journal of Geomechanics, Vol.
12, No. 3, (2012), 281-291.
[18] D. Mandeville and D. Penumadu, True Triaxial System for Clay with Proportional-Integral-Differential
(PID) Control, Geotechnical Tesing Journal, Vol. 27, No. 2, 2004, 1 11.
[19] A. Prashant and D. Penumadu, Effect of Intermediate Principal Stress on Overconsolidated Kaolin Clay,
Journal of Geotechnical and Geoenvironmental Engineering, Vol. 130, No. 3, (2004), 284-292.
[20] F. Li and V.M. Puri, Measurement of anisotropic behavior of dry cohesive and cohesionless powders
using a cubical triaxial tester, Powder Technology 89 (1996), 197-207.
[21] P.V. Lade and A.V. Abelev, Characterization of cross-anisotropic soil deposits from isotropic
compression tests, Soils and Foundations 45 (2005), No. 5, 89-102.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-858
Influence of Shearing Rate on Residual

Strength of Clays
Kenny Kataoka SORENSENa,1, Simon Steen GADEGAARDa
and Jonas Kragh NIELSENa
a
Department of Engineering, Aarhus University
Abstract. In this study a series of multiple-reversal direct shear tests have been
conducted with the aim to characterize the shear strength response of various clays
to step-changes in rate of shearing at residual state. Three different natural
undisturbed specimens of clay and additionally a reconstituted specimen of kaolin
clay have been tested. The tested clays range from low plasticity to extremely high
plasticity clays with liquid limits ranging from approximately 30 % to above
100 %. The stress-strain response of low plasticity clay is found to be
characterized by a pronounced temporary effect of strain rate changes at residual
state. As the plasticity of the clay increases, the response to rate changes is seen to
change, so persistent positive effects of rate changes starts to dominate the
behaviour, while temporary effects become less evident. High plasticity clays are
characterised by a positive persistent response to rate changes with no temporary
effects of strain rate changes at horizontal displacement rates greater than 0.2
mm/hr. The results are discussed and compared to findings from the literature.
Keywords. Clay, direct-shear test, rate-effects, residual strength
1. Introduction
Knowledge of the residual strength properties of clays is relevant to geotechnical

problems in which localized very large shear displacements are likely to occur; e.g. in
relation to driven piles, or reactivation of pre-existing slide surfaces in natural slopes.
Ring shear tests or multiple-reversal direct shear tests on clays, in which the specimens
have been sheared to very large displacements, have been found to show comparable
results, and the measured residual shear strength has been observed to be a function of
the mineralogical composition, clay-size fraction and stress level, cf. [1], [2] and [3].
Several other studies have additionally investigated the influence of shearing rate
on the residual strength of clays. The issue of strain rates needs to be considered when
applying lab element test results to the analysis of field problems. The effect of
shearing rates are typically observed to be small or insignificant within the range of
slow shear displacement rates (0.12 mm/hr 0.6 mm/hr) as typically adopted in
drained lab testing of clays [1]. In contrast significant effects have been reported at fast
and very fast rates of shearing, cf. [4], [5], [6], [7] and [8].
Three characteristic responses to changing shear displacement rates at residual
state have generally been observed in the literature [4] and [5]. Upon a step increase in
shear displacement rate from a given reference rate , the immediate response to
1
Corresponding Author. [email protected]
K.K. Sorensen et al. / Inuence of Shearing Rate on Residual Strength of Clays 859

: shear displacement rate
(A)
}}PPositive rate effects
Shear strength
(B)
Neutral rate effects

Negative rate effects
Shear displacement
Figure 1. Illustration of characteristic rate effects at residual state.
the imposed temporary strain acceleration will be an increase in the shear strength.
Then with continued shear displacement at constant rate of displacement the shear
strength is observed to either; remain at a shear strength higher than the reference
strength (positive rate effect), return to the reference strength (neutral rate effect) or
drop to a lower strength than the reference strength (negative rate effect), as illustrated
in Figure 1. The positive rate effects may be sub-categorised as either (A) Fully
persistent or (B) Combined temporary and persistent effects.
The rate effects observed at residual state have been related to the shearing modes
of the soil [4] and [5]. High plasticity clays typically display positive rate effects
associated with a sliding mode of shear, granular soils show neutral rate effects
associated with a turbulent mode of shear, while low and intermediate plasticity clays
show a transitional mode of shear which in some cases may be associated with negative
rate effects. Similar rate effects have also been observed in the pre-peak region in
triaxial tests on various natural and artificially cemented clays [9] and [10]. However
the observed rate effects in the pre-peak region appear more complex for some soils
[10] as a likely result of greater micro-structural changes occurring in the pre-peak
region with increasing shear strains compared to the residual state [11].
The underlying mechanisms behind the observed rate effects both in the pre-peak
and post-peak regions including residual state are nevertheless not yet fully understood,
though several different postulates have been put forward in the literature.
In this study a series of multiple-reversal direct shear tests have been conducted
with the aim to characterize the shear strength response of different clays to step-
changes in the rate of shear displacement at the residual state. Clays ranging from low
plasticity to extremely high plasticity have been tested. The study has been carried out
to supplement the existing database of results and to contribute to a better
characterization and understanding of the rate dependent behaviour of clays at residual
state.
2. Test Procedure and Samples
2.1. Soils tested
Specimens of three different natural undisturbed clays and additionally reconstituted

kaolin clay have been tested. The tested clays have been selected to range from low
plasticity to extremely high plasticity (classification according to BS 5930:1999). The
natural clay samples consisted of three different types of Danish clays;
860 K.K. Sorensen et al. / Inuence of Shearing Rate on Residual Strength of Clays
x Little Belt clay (LB); an Eocene stiff fissured overconsolidated sedimentary

marine clay of extremely high plasticity.
x Mica clay (M); a Neogene stiff overconsolidated marine sedimentary clay
with organic content and of intermediate plasticity.
x Wash-down sandy clay (Wd) of Late Glacial age and low plasticity.
Kaolin clay (type 50) is a commercial powder product derived from weathered
granite. It can be classified as clay of high plasticity.
In Table 1 the classification parameters of the tested clays are listed. For the
natural undisturbed clays the Atterberg limits and void ratio were determined from a
neighbouring sample taken from the same core. All specimens, except Wd1, are
expected to be nearly fully saturated before start of shearing.
2.2. Test Procedures
The conducted tests series consist of six multiple-reversal multistage direct shear tests
performed using an automatic direct shear apparatus (Sheartrac II manufactured by
Geocomp). 25.4 mm x 63.5 mm (initial height x diameter) specimens were tested.
Undisturbed natural samples were obtained by means of either a push-in 70 mm
Shelby tube sampler (M1 and Wd1) or 100 mm Geobor-S wireline coring (LB1-LB3).
The reconstituted specimen of kaolin on the other hand was prepared from a slurry with
an initial water content of approximately two times the liquid limit. The slurry was
poured into an acrylic floating ring consolidometer and preconsolidated under a
nominal stress of approximately 140 kPa before extraction. Each specimen was
trimmed into a cutting ring and subsequently transferred to the shear box.
In the initial stage a vertical (normal) stress slightly greater than the estimated in-
situ vertical effective stress (or nominal preconsolidation stress for the reconstituted
specimen) was applied, and the specimens were allowed to consolidate under the
applied stress before proceeding to the first stage of shearing. Tap water was generally
used in the carriage during all stages of testing for all specimens.
After the initial consolidation stage the test specimens were subjected to rapid
multi-reversal shearing to create the shear surface and to degrade the strength of the
clay to the residual state. A rate of 360 mm/hr was used. The cumulative horizontal
displacement for the multi-reversal phase was between 120 and 140 mm, except in test
LB2 where the cumulative displacement was 270 mm.
Multiple-reversal direct shear testing of undisturbed specimens may in some cases
result in an overestimation of the residual strength compared to results obtained from
Table 1. Classification parameters for test specimens.

Soil Type Specimen Void Natural Liquid Plasticity Classification
no. ratio water limit2 index according to BS
content 5930:1999
[/] [%] [%] [%]
Little Belt Clay LB1 42 111 57 CE
LB2 1.2-1.3 45 117 72 CE
LB3 41 - - -
Mica Clay M1 0.7 27 47 20 CM
Wash down Clay Wd1 0.9 22 29 11 CL
Kaolin K1 1.2 441 57 22 CH
1
Water content determined after preconsolidation
2
Determined using fall cone test.
tests on pre-cut specimens. It has been suggested that this discrepancy may result from
the influence of structure e.g. fissures as found in tests on Little Belt clay [12].
Nonetheless, testing has in this study been carried out using undisturbed specimens in
order to allow for observations of the gradual strength reduction from peak to residual
state.
Once the rapid multi-reversal phase had been completed, the specimens were
allowed to rest for 24 hrs at constant vertical stress to eliminate any build-up of pore
water pressures. The specimens were subsequently sheared in slow forward shear to
6.35 mm horizontal deformation, corresponding to 10 % of the diameter. To investigate
the rate effects the shearing stage was carried out with step changes in the rate of
horizontal displacement between the following three rates: 0.02, 0.2 and 2 mm/hr. The
length of these constant rate intervals were between 0.5 and 1 mm, each step long
enough to bring forth the soil characteristics for that particular rate. At the end of
shearing the shear box was rapidly reversed to its starting point. The specimen was then
consolidated under a higher vertical stress, before the shear stage with step-changes in
rate was repeated. Two to three stages with different vertical stresses were generally
carried out for each specimen to investigate the influence of normal stress.
An overview of the obtained test results from the multi-reversal direct shear tests are
given in Table 2. The table lists the coefficient of consolidation cv as determined from
the consolidation stages, the estimated in-situ vertical effective stress Vc0 for the tested
samples, the normal stress VN applied in the shear stages, the peak shear strength Wpeak
determined during the initial rapid shear stage, and the drained residual shear strength
Wcr,0.2 determined at a reference rate of displacement of 0.2 mm/hr after the initial stages
of rapid shearing. The residual strength was taken as the minimum value recorded after
the initial peak. The average changes in the residual shear strength relative to the
reference strength Wcr,0.2 for rates of displacement of 0.02 and 2 mm/hr are also listed in
the table.
3.1. Post Peak Strength Degradation and Residual Strength Envelope
Figure 2 shows an example of the observed strength degradation during the initial rapid
multi-reversal shear stages in test on an undisturbed specimen of Little Belt clay. The
tendency line has been plotted going through points of horizontal displacement
corresponding to the zero position of the shear box. Very significant strength
degradation is observed in the post peak region in tests on undisturbed natural clays of
intermediate plasticity (M1) and extremely high plasticity (LB1-LB3). In comparison,
the strength degradation for the undisturbed specimen of wash down clay of low
plasticity is seen to be minor, cf. Table 2. The high peak strength seen in tests on Little
Belt clay and Mica clay are likely to represent an undrained shear strength, which is
expected to be somewhat higher than the drained shear strength due to the build-up of
negative excess pore water pressures at the shear surface. The measured residual
strengths from the slow stages of shearing are on the other hand expected to represent
drained values of residual shear strength, although no measurements of pore water
pressures have been taken to confirm this.
Table 2. Overview of results of multistage direct shear tests.

Soil type Test cv1 Vc0 VN Wpeak Wcr,0.2 Wcr,0.02 Wcr,2
no. [m2/s] [kPa] [kPa] [kPa] [kPa] /Wcr,0.2 /Wcr,0.2
[kPa] [kPa]
Little Belt Clay - LB160 95 32 10 0.99 1.03
- 300 64 1.00 1.01
LB2 - 140 200 190 38 0.99 1.03
110-8 400 62 0.99 1.04
210-8 600 108 1.00 1.02
LB3 - 50 97 95 20 0.98 1.03
210-8 200 48 0.99 1.03
210-8 400 75 0.99 1.03
Mica Clay M1 - 230 300 126 44 1.00 1.02
-8
210 588 106 0.98 1.06
Wash down Clay Wd1 > 110-6 30 50 57 43 1.00 1.00
Kaolin K1 > 110-6 1402 150 59 45 1.00 1.02
> 110-6 300 88 1.00 1.02
1
Coefficient of consolidation determined from consolidation stages prior to shearing based on t50 found by
means of Casagrandes log time fitting method.
2
Nominal preconsolidation pressure (likely to be overestimated due to friction in the consolidometer).
An estimated horizontal displacement of approximately 60 to 80 mm was required

in all the tests to reach residual state, at which point only minimal further reduction in
shear strength with continued shearing was observed for all clays. This agrees well
with the findings by [1].
Though significant soil loss was observed during the initial stages of multiple-
reversal rapid shear, it appears that this have had insignificant influence on the shear
surface and the measured residual shear strength, as seen from the constant residual
strength, cf. Figure 2.
Figure 3 shows the residual strength data points for extremely high plasticity Little
Belt clay as listed in Table 2. The lower bound residual strength envelope ( Icres = 8.5
deg.) from direct shear tests on pre-cut specimens of Little Belt Clay found by [11] is
also indicated on the plot for comparison. It can be seen that the measured residual
strength for undisturbed specimens of extremely high plasticity Little Belt clay tend
towards ^the lower bound strength envelope^ in^ the ^first^ one ôr^ two ^stages ôf^ normal
stress. In ^subsequent^ stages ât^ higher ^normal ^stress ^the^ measured^ residual ^strengths
appears^ to shift upwards ^from the lower bound strength envelope. This ^may^ possibly
indicate structural changes in the shear zone, as a result of increased normal stress.
120
Residual shear strength, r [kPa]
200
150 100
Shear stress, W [kPa]
100 80
50 60
0 40
-50 20
-100 0
LB2, VN = 200 kPa 0 100 200 300 400 500 600
-150 Normal stress, N [kPa]
0,1 1 10 100 1000
Cumulative horizontal displacement, G [mm] LB1 LB2 LB3
Figure 2. Strength degradation observed in multiple- Figure 3. Residual strength envelope for
reversal direct shear test (LB2, VN = 200 kPa). Little Belt clay.
In comparison with the findings by [2] the measured residual strengths for Mica
clay of intermediate plasticity, as given in Table 2, is found to be very low, while the
results for Little Belt clay and kaolin are found to be in reasonable agreement with the
reported results from the literature.
3.2. Observed Rate Dependent Shear Strength Behaviour at Residual State
Figures 4a-d shows representative plots of the influence of step changes in the
displacement rate on the residual strength for the tested clays. From the figures the
following general trends can be identified;
x High and extremely high plasticity clays (kaolin clay and Little Belt clay) are
characterised by a persistent positive response to rate changes with
insignificant temporary effects of rate changes at rates greater than 0.2 mm/hr,
cf. Figures 4a and 4b.
x Intermediate plasticity clay (Mica clay) shows a combination of persistent
positive effects and temporary effects of rate changes at a stress level of 300
kPa, cf. Figure 4c.
x Low plasticity clay (Wash-down sandy clay) is found to be characterised by a
pronounced temporary effect of strain rate changes (neutral rate effect), cf.
Figure 4d.
The relationship between the residual strength and the rate of deformation for
Little Belt clay (LB2) is illustrated in Figure 5.
70 0.2 120
0.02 0.2 2 0.2 2 2 2 Rate High plasticity
0.02 0.02 0.2 [mm/hr] 115
65 Kaolin clay
0.2
Shear stress, [kPa]
Shear stress, [kPa]
110 0.02
60
105
55 Extremely high plasticity 100
Little Belt clay
2
95 0.02 0.2 2 2 2 Rate
50 0.2 0.02 0.2 [mm/hr]
90
45
LB2, VN = 400 kPa 85 K1, VN = 300 kPa
(a) (b)
40 80
0 2 4 6 8 0 2 4 6 8
Horizontal displacement, G [mm] Horizontal displacement, G [mm]
65 0.2 55
Intermediate plasticity Low plasticity
60 Mica clay Washdown sandy clay
50 0.2 2
Shear stress, [kPa]
Shear stress, [kPa]
55 0.02 0.2 2 2
0.02 0.2 2
0.02 2 Rate
50 2 45
0.2 2 Rate [mm/hr]
0.02 0.02 0.2 0.02 [mm/hr] 0.2
45 40
40
0.02 35
35 0.02
M1, VN = 300 kPa (c) Wd1, VN = 50 kPa (d)
30 30
0 2 4 6 8 0 2 4 6 8
Horizontal displacement, G [mm] Horizontal displacement, G [mm]
Figure 4. Representative stress-displacement curves for (a) Extremely high plasticity clay, (b) High plasticity
clay, (c) Intermediate plasticity clay and (d) Low plasticity clay.
1,12
Relative residual shear strength,

1,10
1,08
1,06
r/r,0.2 [-]
1,04
1,02
1,00
0,98
0,02 0,2 2 20 200 2000
Shear displacement rate, ' [mm/hr]
LB2 (600 kPa) LB2 (400 kPa) LB2 (200 kPa)
Figure 5. Relationship between relative residual shear strength and rate of deformation (LB2).
In Figure 5 the residual strength has been normalised by the residual strength
(Wref,0.2) found at the reference rate of 0.2 mm/hr. The average normalised values of
residual strength from all the tests are listed in Table 2.
From Figure 4a, Figure 5 and Table 2 it can be seen that the persistent positive
influence of rate changes between 0.02 mm/hr and 0.2 mm/hr is insignificant or very
small in test on Little Belt clay, while the influence of rate changes between 0.2 mm/hr
and 2 mm/hr are much more pronounced with an increase in shear strength of
approximately 3 % per log change in rate. The test on kaolin and Mica clay show
similar behaviour, as can be seen from Figures 4b and 4c and Table 2, and it appears
that the magnitude of persistent positive rate effects is not significantly influenced by
the plasticity of the clay.
In contrast, the degradation of temporary effects with increasing displacement after
a rate change is seen to be very different when comparing the results from test on
intermediate plasticity Mica clay (Figure 4c) and low plasticity wash-down sandy clay
(Figure 4d). Test M1 on Mica clay show only a gradual degradation of the temporary
effects with increasing strains, while test Wd1 on wash-down sandy clay shows neutral
rate effects with an almost immediate return to the reference state after the initial peak.
Unpublished results from tests on Little Belt clay have shown the residual strength
to increase approximately linearly with a log increase in the rate of horizontal
deformation at rates higher than 0.2 mm/hr, as was also found by [7] and [8]. This is
also indicated in Figure 5 if the additional data point from the initial rapid shear stage
(360 mm/hr) is considered.
The tests on low plasticity wash-down sandy clay (Wd1) and extremely high
plasticity Little Belt clay (LB1-3) show behaviour deviating from the general trend.
Negative persistent rate effects appear to be displayed in test Wd1 (cf. Figure 4d),
while temporary effects of rate changes appear in tests LB1-3 (cf. Figure 4a), when the
rate of deformation is changed from a fast rate of 2 mm/hr to the slowest rate of 0.02
mm/hr. The observed temporary effects are generally found to be more pronounced
with increasing level of normal stress. However, since no measurements of pore water
pressures were taken, it is uncertain if a build-up of excess positive pore water
pressures may have occurred during stages of fast shearing and partly explain the
observed effects in the subsequent very slow stages of shearing. Further investigations
are needed to clarify this.
From Figure 4c the response of Mica clay of intermediate plasticity at normal
effective stress of 300 kPa is seen to be clearly dominated by temporary effects. In
contrast, the response was found to change and be characterised by pronounced
persistent positive effects, when the normal effective stress was increased to 600 kPa
(plot not shown). Hence, it appears that structural changes may occur as a result of
increasing normal stress level, which is reflected in the response to rate changes.
Similar findings have been made by e.g. [4] and [6] in fast shearing of clays of low to
intermediate plasticity.
4. Conclusions
A series of multiple-reversal direct shear tests have been conducted on different

natural undisturbed and reconstituted clays. The tests have highlighted that the residual
strength is influenced by the applied shear rate, and that the response to changing rates
is strongly influenced by normal stress, the clay content and mineralogical composition
of the clay. In contrast, it appears that the magnitude of the observed persistent positive
rate effects is not significantly influenced by the plasticity of the clay.
Low plasticity clay is found to be characterized by a pronounced temporary effect
of strain rate changes. Intermediate plasticity clays show a combination of persistent
positive effects and temporary effects, while high plasticity clays are characterised by a
positive persistent response to rate changes with insignificant temporary effects of rate
changes at horizontal displacement rates greater than 0.2 mm/hr. The behaviour at rates
slower than 0.2 mm/hr and the micro-mechanisms behind the observed behaviour need
further investigation.
References
[1] A.W. Skempton. Residual strength of clays in landslides, folded strata and the laboratory, Gotechnique
35, No. 1 (1985), 3-18
[2] T.D. Stark and H.T. Eid, Drained residual strength of cohesive soils, J. Geotech. Geoenviron. Eng. 120,
No. 5 (1994), 856-871.
[3] L.D. Wesley, Residual strength of clays and correlation using Atterberg limits. Gotechnique 53, No. 7
(2003), pp. 669672.
[4] L.J. Lemos, A.W. Skempton and P.R. Vaughan. Earthquake loading of shear surfaces in slopes, Proc.
11th I.C.S.F.E., San Francisco, Vol. 4 (1985) 1955-1962.
[5] T.E. Tika, P.R. Vaughan and L.J. Lemos. Fast shearing of pre-existing shear zones in soil, Gotechnique
46, No. 2 (1996), 197-233.
[6] P. Carrubba and P. Colonna. Monotonic fast residual strength of clay soils. Italian Geotechnical Journal
3 (2006), 32-51.
[7] M. Suzuki, T. Yamamoto and Y. Kai. Rate effect on residual state strength of clay related with fast
landslide. Proc. of Int. Symp. on Prediction and Simulation methods for Geohazard mitigation, (2009),
347-352.
[8] G. Wang, A. Suemine and W.H. Schulz. Shear-rate-dependent strength control on the dynamics of
rainfall-triggered landslides, Tokushima Prefecture, Japan. Earth Surface Processes and Landforms 35,
No. 4 (2010), 407416.
[9] F. Tatsuoka, M. Ishihara, H.Di Benedetto and R. Kuwano. Time-dependent shear deformation
characteristics of geomaterials and their simulation. Soils and Foundations 42, No. 2 (2002), 531540.
[10] K.K. Sorensen, B. Baudet and B. Simpson. Influence of structure on the time-dependent behaviour of a
stiff sedimentary clay. Gotechnique 57, No. 1 (2007), 113124.
[11] J.F. Lupini, A.E. Skinner, and P.R. Vaughan. The drained residual strength of cohesive soils.
Gotechnique 31, No. 2 (1981), 181-213.
[12] K.K. Sorensen. Drained residual strength of Little Belt clay. Proc. 67th Canadian Geotech. Conf.
Regina (2014).
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-866
Advanced laboratory investigation of axial

cyclic loading in silica sands
(Dedicated to late Professor Pierre Foray)
Cristina de H. C. TSUHAa,1, Richard J. JARDINEb, Zhongxuan YANGc, Matias

SILVAd and Siya RIMOYe
a
University of So Paulo, Department of Geotechnical Engineering, Brazil
b
Imperial College, Department of Civil and Environmental Engineering, London
c
Department of Civil Engineering, Zhejiang University, China
d
Institut National Polytechnique de Grenoble, Laboratoire 3S-R, France
e
University of Dar es Salaam, Department of Transportation and Geotechnical
Engineering, Tanzania
Abstract. This paper describes the global behaviour of and local stress conditions
applying to highly instrumented cyclic laboratory model displacement pile tests
conducted in a calibration chamber on medium dense Fontainebleau NE34 sand.
The instrumentation provided measurements of the contact stresses on the jacked
pile shaft and the local vertical, radial and circumferential stresses in the
surrounding soil mass, in tests involving a range of cyclic load amplitudes. The
patterns of effective stress developing on and around the pile shaft are described.
The data obtained are interpreted by reference to cyclic soil element tests
conducted on the same soil, as described in the parallel investigation by
Aghakouchak et al (2015), reported in the same symposium. The calibration
chamber experiments offer a comprehensive understanding of the cyclic
degradation processes, and provide key information for improving the design of
piled foundations under cyclic loading.
Keywords. Cyclic loading, pilesoil interface, calibration chamber, sand, local

stresses
1. Introduction
The piled foundations of offshore platforms, wind turbines, and other types of
structures are usually subject to significant environmental and operating axial load
cycles. Such loading cycles can lead to changes in stiffness, permanent displacement
growth and changes in axial capacity. Shaft resistance can either degrade markedly, or
potentially improve [1], [2]. Field tests on piles driven in sands have identified three
main styles of response to cyclic axial loading, which are dependent of the mean load,
1
Corresponding Author: Cristina de Hollanda Cavalcanti Tsuha, University of So Paulo, Department
of Geotechnical Engineering, Av. Trabalhador Socarlense, 400, So Carlos, SP 13566-590, Brazil; E-mail:
[email protected].
C. de H.C. Tsuha et al. / Advanced Laboratory Investigation of Axial Cyclic Loading 867
cyclic amplitude, and number of applied cycles. The responses thought to be linked to
changes in the shaft radial effective stress distributions [3],[4]. Although potentially
important, such effects are rarely addressed in routine offshore or onshore pile design.
The experimental program presented in this paper was developed to better
characterize the three different cyclic behaviour styles and provide mechanical
explanations for the processes involved, by measuring how stresses evolve at the
displacement pile surface and in the surrounding sand mass under axial shaft loading.
The current study is a part of a joint research programme between the Laboratoire
3S-R (Grenoble Institute of Technology) and Imperial College London about
displacement pile behaviour in medium dense sand. A highly instrumented steel pile
Mini-ICP was developed by Imperial College of London and the experiments were
conducted in a calibration chamber at the Grenoble Institute of Technology (INPG).
The key findings described in this text are based on the results obtained from four
test series Mini-ICP1, Mini-ICP2, Mini-ICP3 and Mini-ICP4. The Mini-ICP pile was
installed into reconstituted siliceous Fontainebleau NE34 sand in the chamber tank.
The vertical, radial and circumferential stresses ('z, 'r, and ') developed in the sand
mass were measured by suites of sensors placed at up to three different levels.
2. Testing programme
2.1. Grenoble calibration chamber
The Grenoble calibration chamber (Figure 1) is 1.5 m high and 1.2 m in internal
diameter. The chamber base and top cover consist of rigid plates with a 100 mm
thickness. An upper membrane, filled with water, applies the vertical stress conditions
('vo around 150 kPa for the present study), and a single latex rubber sheet smeared
with silicone grease was employed to provide near K0 conditions while reducing wall
friction between the sand and the tank. The chamber tank was filled with NE34
Fontainebleau sand by air pluviation technique to provide a relative density of 72%.
Table 1 shows the index properties of this sand. Independent standard CPT tests were
conducted on sand samples under the same 150 kPa 'vo, and the cone end resistance
(qc) traces presented quasi-constant 21 2 MPa sections. The mechanical properties of
the sand are described in detail elsewhere [5].
Table 1. Fontainebleau NE34 sand

Grain shape D10 (mm) D50 (mm) D60 (mm) emax emin
Sub-angular 0.15 0.21 0.23 0.90 0.51
2.2. Mini-ICP pile and soil sensors
The 36 mm-diameter closed-ended stainless steel Mini-ICP pile (Figure2) is composed

of three instrumented clusters (leading or A, following or B, and trailing or C), a solid
60 cone fitted at the pile tip, and a pile cap. Each cluster contains an axial load cell
(ALC), which can be used together with adjacent ALCs to derive the average pile skin
friction; a surface stress transducer (SST) to measure radial, total stress ('r) and shear
stress (rz); a temperature sensor; and inclinometers sensors. Further details of the Mini-
ICP pile are given in [6].
868 C. de H.C. Tsuha et al. / Advanced Laboratory Investigation of Axial Cyclic Loading
The pressure sensors to monitor the stresses in the sand mass consist of disk-
shaped cells (diameters of 6 to 6.5mm, and thicknesses of 0.6 to 1.4mm). The sensors
faces were oriented in the direction of the normal stresses ('z, 'r, and ': vertical,
radial and circumferential stresses, respectively). The special calibration procedures
that are essential to interpret the soil sensors complex behaviour are detailed in [7].
Figure 1 illustrates the sensor configuration for the Mini-ICP3 test. The relative radial
distance of the sensors from the pile axis r/R varies from 2 to 20, as described in [6],
where R is the pile radius and r is each sensors radial distance from pile axis.
Figure 1. Typical arrangement for Mini-ICP tests.
2.3. Pile installation and cyclic loading tests
The Mini-ICPs were installed into the chamber tank by cyclic jacking to simulate key
aspects of the driving process, by applying stroke lengths between 5 and 20 mm and
reducing the pile head loads to zero at the end of each stroke. The final tip depths
below sand surface were 0.92 m for Mini-ICP1, and 0.99 m for the other installations.
Static monotonic loading tests were carried out before and after each cyclic test to
verify the cyclic loading influence on pile shaft capacity. The monotonic loading tests
were performed under displacement control at rates around 0.01 mm/s. After these
initial tests, series of tensile one-way cyclic loading were performed on Mini-ICP pile,
under load-controlled conditions. Following the sequence of one-way cyclic tests, high-
level two-way cyclic loading (both compression and tension) were conducted under
either displacement or load-controlled conditions. Table 2 summarizes the cyclic
testing programme and the results of pile tension capacity (QT) measured prior to the
cyclic tests. The loading parameters in this table are normalised by QT.
Figure 2. Mini-ICP1 configuration [6].
3. Pile cyclic axial response
In the current study, the classification of cyclic response is based on the number of
cycles (N) necessary to cause either failure at the shaft-soil interface or specified rates
of permanent displacement growth. In this case, cyclic failure occurs when
accumulated displacement (s) reached 10% of the pile diameter or the rate ds/dN shows
a sharp increase. Rates are considered slow if < 1 mm/104 cycles, and fast if > 1
mm/100 cycles). The 3 styles of response, identified in Figure 3, are then defined as:
Stable if no failure occurs before 1000 cycles, with slow rates of permanent
displacements.
Meta-Stable if failure occurs with 100 < N < 1000, with rates of permanent
displacements fail to stabilise slow values.
Unstable if failure develops before N =100.
Figure 3 presents a cyclic shaft failure interaction chart in which the cyclic loads
are plotted as Qcyclic and Qmean normalised by the static failure load QT. The annotated
numbers signify Nf - the number of uniform cycles required to induce a cyclic failure.
Where no number is given, failure did not occur within at least 1000 cycles. Further
details of the experiments are listed in Table 2. The pattern observed in Figure 3 is
closely comparable to that reported from full-scale piles driven in sand [1], [2].
Three key aspects of each style of cyclic behaviour are illustrated in Figures 4, 5,
and 6, covering Stable, Meta-stable and Unstable test outcomes respectively. The
cumulative displacements under Stable (ICP4-OW1), Meta-stable (ICP2-OW3) and
Unstable (ICP3-OW1) one-way cyclic experiments are shown in Figures 4a, 5a, and 6a.
The corresponding pile surface effective stress paths, deduced from the surface stress
transducers, are given in Figures 4b, 5b, and 6b, which also indicate the estimated
extent and position of the sands initial Y2 kinematic yield surfaces, as defined in [9].
Table 2. Mini-ICP cyclic loading details.

Mean cyclic
Install QT Qcyclic/ Qmean/
Test period T
No of cycles
ation (kN) QT QT
(min)
Mini- ICP1-OW1(Meta-stable) 1000 0.58 9.2 0.22 0.22
ICP1 ICP1-TW1 (Unstable) 100 (Nf =4) 2.64 10.8 0.41 0.06
ICP2-OW1 (Stable) 1000 0.43 12.1 0.12 0.12
Mini- ICP2-OW2 (Stable) 1000 0.70 13.2 0.20 0.20
ICP2 ICP2-OW3(Meta-stable) 500 1.04 14.0 0.28 0.28
ICP2-TW1 (Unstable) 100 (Nf = 4) 2.17 13.7 0.48 0.15
ICP3-OW1 (Unstable) 100(Nf =66) 1.64 12.5 0.38 0.38
CP3-TW1 (Unstable) 1 (Nf = 1) 6 10.9 0.54 -0.08
Mini-
ICP3-TW2 (Unstable) 199(Nf=165) 0.73 10.9 0.40 0.06
ICP3
ICP3-TW2 (Unstable) 50 (Nf =10) 1.04 10.9 0.44 0.02
ICP3-TW4 (Unstable) 37 (Nf = 3) 2.16 10.9 0.44 0.02
ICP4-OW1 (Stable) 7000 0.43 11.5 0.15 0.15
Mini-
ICP4-TW1(Meta-stable) 600(Nf =580) 0.44 13.9 0.23 0.06
ICP4
ICP4-OW2 (Metastable) 50(incomplete) 0.18 5.5 0.21 0.63
*OW = one-way tension test; TW = two way compression/tension test.
** The control mode of these cyclic tests is described in [6].
Nf=
1.0 1
ay
w
o
Tw
0.8
ay
w
Stable
ne
O
Meta-Stable
Qcyclic/QT
0.6 Unstable Unstable

1 Nf = number of cycles to failure
5
10 4 Note: numbers signify number of cycles
10 4
100 developed in each test
0.4
165 66
500
580 Meta-Stable
1000
0.2
>1000
Stable
0.0
-0.2 0.0 0.2 0.4 0.6 0.8 1.0
Qmean/QT
Figure 3.Summary of average cyclic shaft loading and failure conditions in Mini-ICP tests, showing Stable,
Meta-Stable and Unstable Zone boundaries [6].
The Stable test presented in Figure 4b illustrates only slight top-down degradation
with radial stresses reducing slowly as cycling continues. The local effective stresses
remain principally within the sands initial Y2 yield surface. In contrast, the Meta-
stable effective stress paths plotted in Fig 5b approach the static interface shear failure
envelope, described by [5]. The Unstable response shown in Fig. 6b indicates effective
stress paths that moved sharply to the left, engaging and relocating their Y 2 surfaces
within the first cycle. Figures 4c, 5c, and 6c report the effective stress paths measured
at r/R=5 and h/R~15 in the sand mass by the sensors described in [8] under the same
three styles of cyclic loading. In these figures, the initial Y 2 surface is represented by a
circle with a radius of 0.245p, corresponding to the triaxial test range reported in [10].
0.05
Cumulative displacement (mm)

(a)
0.00
-0.05
-0.10
0 2000 4000 6000 8000
Number of cycles
200 ' o
(b) =27
Shear stress rz (kPa)
100
estimated initial
Y2 surface
0
estimated initial
'
0.28r Y2 surface
-100
Leading A
-200 Following B
Trailing C
0 100 200 300 400 500
'
Radial stress r (kPa)
Figure 4. Stable test ICP4-OW1: (a) cumulative displacements, (b) effective stress paths at all SSTs and (c)
effective stress path developed in sand mass at 5R, h/R~15.
4. Effect of cyclic loading on shaft capacity
Tension pile tests demonstrated that capacity (QT) grew by up to 20% after prolonged
Stable cycling (see Table 2). This is interpreted as being due to the densification in the
interface zone and fabric rearrangement that increases dilation under static loading.
In contrast, two-way Unstable tests led to shaft capacities degrading by up to 50%.
In the case of Meta-stable cycles, the shaft capacity reduction depended on the cyclic
loading levels imposed. These features may be gauged by comparing the initial QT
listed in Table 2 for tests ICP1-TW1, ICP2-TW1, ICP3-TW4 and ICP4-OW2 with their
post cycling QT values of 4.9, 8.7, 4.8, and 6.0 kN respectively.
The laboratory piles cyclic responses have been investigated by locally
instrumented cyclic triaxial tests designed to model conditions around the pile shaft
[11]. These special cyclic triaxial tests were carried out on specimens of NE34
Fontainebleau, prepared at similar relative densities (70%) to the calibration chamber
tests. The samples were conditioned through controlled stages of consolidation to
elevated pressures, prior drained triaxial cycling and extended ageing to simulate the
pile installation process. Undrained cycling was then applied under a range of Cyclic
Stress Ratios (CSR = qcyc /p0). As with the pile tests, Stable, Meta-stable and Unstable
responses were observed in terms of mean effective stress drift, cyclic stiffness and
permanent strain growth. Such experiments provide a way of predicting field behaviour
at sites where cyclic pile loading experiments are possible.
1.0

0.5
0.0
-0.5
-1.0
0 200 400 600
Number of cycles
200 (b) ' o

=27
estimated initial
100
Y2 surface
0 estimated initial
Y2 surface
-100
'
Leading A 0.28 r
-200 Following B
Trailing C
0 100 200 300 400 500
'
Radial stress r(kPa)
Figure 5. Meta-Stable test ICP2-OW3: (a) cumulative displacements, (b) effective stress paths at all SSTs
and (c) effective stress path developed in sand mass at 5R, h/R~15.
0
Nf=66
-3
3.6mm
-6 (a)
-9
-12
-15
0 20 40 60 80 100
Number of cycles
200 (b) ' o

=27
100 Leading A
Following B estimated initial
Trailing C Y2 surface
0
-100
'
0.28 r
-200 estimated initial

Y 2 surface
0 100 200 300 400 500
'
Radial stress r (kPa)
Figure 6. Unstable test ICP3-OW1: (a) cumulative displacements, (b) effective stress paths at all SSTs and
(c) effective stress path developed in sand mass at 5R, h/R~15.
5. Conclusions
Calibration chamber experiments on a model displacement pile jacked into a medium

dense sand mass subjected to cyclic axial loading identify:
The combinations of mean load, cyclic load, and number of cycles that lead to
three distinct styles of pile response: Stable, giving no loss of shaft friction
and possibly gains, Unstable, where shaft capacity falls sharply, and Meta-
stable, where the shaft can sustain hundreds of such cycles before failing.
Explanations for the responses seen in full scale field tests.
Cyclic behaviours patterns that can be reproduced in appropriately designed
special cyclic triaxial tests.
Ways forward for practical cyclic design, as outlined in [12].
6. Acknowledgements
The authors gratefully acknowledge the support provided by Shell U.K. Limited, the
UK Health and Safety Executive, the UK Engineering Physical Sciences Research
Council, Total France, Centre National de la Recherche Scientifique (CNRS), and the
French Program ANR SOLCYP.
References
[1] R.J. Jardine, & R.J. Standing. Pile Load Testing Performed for HSE Cyclic Loading Study at Dunkirk,
France. Two Volumes. Offshore Technology Report OTO2000 007. Health and Safety Executive,
London. 2000.
[2] R.J. Jardine R.J. and J.R. Standing, J.R. Field axial cyclic loading experiments on piles driven in sand.
Soils and Foundations. 52, (2012), 723-737.
[3] R. J. Jardine, R.J. The cyclic behaviour of offshore piles. Chapter in 'The Cyclic Loading of Soils', (1991)
Eds. Brown & O'Reilly, Blackie & Son, Glasgow
[4] J.T. DeJong, M.F. Randolph, D.J. White. Interface load transfer degradation during cyclic loading: a
microscale investigation. Soils and Foundations 43 (2003), 9194.
[5] Z.X. Yang, R.J. Jardine, B.T. Zhu, P.Y. Foray, C.H.C. Tsuha. Sand grain crushing and interface shearing
during displacement pile installation in sand. Gotechnique 60 (2010), 469482.
[6] R.J. Jardine, B.T. Zhu, P. Foray, & C.P. Dalton. Experimental arrangements for the investigation of soil
stresses developed around a displacement pile. Soil and Foundations 49 (2009), 661-673.
[7] B.T. Zhu, R.J. Jardine, P. Foray. The use of miniature soil stress measuring sensors in applications
involving stress reversals. Soils and Foundations 49 (2009), 675688.
[8] C.H.C. Tsuha, P.Y. Foray, R.J. Jardine, Z.X. Yang, M. Silva, & S. Rimoy. Behaviour of displacement
piles in sand under cyclic axial loading. Soil and Foundations 52 (2012), 393410.
[9] R.J. Jardine. Some observations on the kinematic nature of soil stiffness. Soils and Foundations 32
(1982), 111124.
[10] R. Kuwano, & R.J. Jardine. A triaxial investigation of kinematic yielding in sand. Gotechnique 57
(2007), 563579.
[11] A. Aghakouchak, W.W. Sim., & R.J. Jardine. Stress-path laboratory tests to characterise the cyclic
behaviour of piles driven in sands. Soils and Foundations companion special volume. In Press.
[12] R.J. Jardine, K. Andersen, and A. Puech. (2012). Cyclic loading of offshore piles: potential effects and
practical design. Keynote Paper. Proc 7th Int. Conf. on Offshore Site Investigations and Geotechnics,
SUT London, pp 59-100.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-874
,QIOXHQFHRIYROFDQLFIO\DVKRQWKH
VWUHVVVWUDLQVWUHQJWKEHKDYLRXURI
ORHVVLDOVRLOV
Agustin ADAMIa,1 and Victor A. RINALDIa
a
Facultad de Ciencias Exactas, Fsicas y Naturales, Universidad Nacional de
Crdoba, Argentina.
Abstract. There are in nature a large number of soils formations that
exhibit increments in stiffness and strength properties after deposition or
even with aging after remolding and compaction. Post depositional
diagenetic processes in the former case and various thyxotropic mechanisms
in the second are usually postulated to explain the observed behavior.
Argentinean loess at Cordoba city is normally a highly unstable soil which
collapse when moistened. However, some specific horizons, shows very high
shear strength properties in nature despite high water content, and even
showing aging effects after remolding. Samples of this second type of loess
were studied in this work to evaluate such behaviour, either in undisturbed
and remoulded conditions. A set of mechanical test show that the increment
in strength and stifness is time dependent and measured values reach almost
a 70 % of that corresponding to undisturbed specimens after 28 days of
aging. Effective stress tests and suction control during aging confirms that
the regain in strength is due to a real cementation process and not to
suction increasing as is usually assumed. Mineralogical and chemical tests
confirm the presence of amorphous silica which in a high alkaline medium
is showed here that generate a cementation reaction, bonding particles
together and developing a stiffer structure.
Keywords. Loess, cementation, thyxotropy, shear strength
Introduction
Loess is an eolian formation and is one of the most abundant soils on the continental surface of
the world. The Argentinean deposit is the largest deposit in the southern hemisphere and is
mainly composed of platy shape silt and sand particles and a minor fraction of clay which is
usually Illite and Montmorillonite. The combined effect of particle shape, origin and particle
gradation, render poorly accommodated and open structure. A more detailed review of
Argentinean loess origin, structure and behavior at small and large strains can be found
elsewhere (i.e [1] and [2]). The stability of the structure is governed by the combined effect of
water suction and cementation. As cementation is poorly developed, the soil may experience
high volume changes when loaded or wetted and then be considered as a collapsible unstable
soil. Research efforts in the past decades have focused on understanding collapse mechanisms
(i.e. [3] and [4]). Most natural soil deposits have some degree of cementation arising from post-
depositional processes such as chemical bonding and cement precipitation ([5] and [6]). In
specific layers of loess formation is possible to find some amount of cementing agents fully
disseminated in the soil and precipitated at particle contacts increasing significantly its shear
strength and generating a true sedimentary rock (i.e. siltcretes and calcretes) also locally known
as tosca.
1
A. Adami and V.A. Rinaldi / Inuence of Volcanic Fly Ash 875
The presence of amorphous silica in some layers of loess is of volcanic origin and
arises from the ash rains of the Andean volcanoes eruptions. It has been estimated
between 5 to 10 m of pyroclastic material has accumulated in the last one million years.
It has been also noticed in the engineering practice an important regain of the
stiffness and shear strength with aging in some samples of loess after remoulding and
compaction which was usually attributed to various thyxotropic mechanism including
particle re-accommodation, water homogenization, water-meniscus development and
salt precipitation at particle contacts, as it was observed also in many other soils
reported in [7], [8], [9], [10] or [11]. More recently soil hardening with aging was
evaluated among others in [12], [13], [14] and [15]. Oppositely to the behaviour
described above, other loess samples with similar origin and structure develop
negligible regain in strength after remoulding.
The different behaviour observed originate the need to study the role of the
different processes that may take place in the soil after compaction. The potential
stabilization effect due to various cementing processes and particularly that arising
from the presence of amorphous silica is of increasingly interest in order to evaluate
and predict the behaviour and stability of compacted embankments by using this soil
without the requirement of addition of other type of cements as may be lime or
Portland cement.
This main goal of this work is to describe some fundamental results and
conclusions of an experimental study developed to evaluate the processes and
mechanisms responsible for the increment in stiffness and shear strength of some of the
selected soil samples with aging after remolding and compaction. Two types of
samples of loess were selected for this study which showed different behavior either in-
situ and in the laboratory after compaction. The gain in stiffness and undrained strength
with time were evaluate in this work at small and large strains levels by means of
unconfined compression tests, odometer tests and measurement of shear wave velocity
with bender elements. The role of water suction and effective stresses was also studied
by means of measurements of suction during aging and some drained triaxial tests. The
presence of potential reactive minerals as silica was determined using the well known
X-Ray spectometry, X-Ray diffraction, specific surface test, grain size distribution and
scanning electron microscope (SEM). Finally, the pozzolanic chemical activity of silica
minerals was monitored by means of the measurement of the electrical conductivity of
prepared soil samples mixed with solutions of calcium hydroxide.
1. Selected Samples and Geotechnical Parameters
Table 1 describe the main geotechnical parameters of the selected samples for this
study. The samples labelled as M correspond to a loessial type obtained at
approximately 13 m depth in the city of Cordoba in Argentina. In natural state, this
sample is very stiff and the penetration resistance in terms of the blow number (N) of
the standard penetration test (SPT) is very high (usually N>30 blows/foot). The sample
labelled as L corresponds to a typical compressible loess from a 2.5 m depth, with a
penetration resistance according to the SPT of N < 5. Specific Surface was determined
here by using the absorption method of the blue of methylene [16]. The larger fines
content of the sample M reflects a higher value in specific surface (69.7 m2/gr),
whereas the values for the sample L (2.3 m2/gr) is a typical value for silts.
876 A. Adami and V.A. Rinaldi / Inuence of Volcanic Fly Ash
Table 1: Main geotechnical and physical properties of the samples selected in this work.
As
Sample Depth d w Soluble %Passing
pH P.L. P.I. U.S.C.S.
Group [m] [tn/m3] [%] Salts sieve 200 [m2/gr]
M 13.0 1.17 36.8 < 0.1% 8.35 92.4 28 12 CL 69.7
L 2.5 1.26 15.2 < 0.1% 8.85 79.3 19 5 ML-CL 2.3

Note: d: dry unit weight, w: natural moisture content, P.L.: plastic limit, P.I.: plasticity index, U.S.C.S.:
Unified Soil Classification System, As: Specific Surface.
2. Structure, Fabric and Mineral Composition
Particle shape and structure of both samples can be compared on the images obtained
from the Scanning Electron Microscope (SEM) displayed on Figure 1 for two
magnifications 500X and 2000X. It can be seen that both samples show elongated and
platy shape particles of fine sand and silt size, being the particle dimensions of samples
M in average 10 m while for samples L are determined to be 20 m. Additionally,
there is a larger amount of clays in the samples M on the surface of the silt particles
which also look like more diagenized. Particle arrangement shows a preferential
direction with the largest axis almost parallel oriented yielding an inherent anisotropic
fabric.
Mineral composition of particles for each sample was studied by means of x-ray
spectometry and x-ray diffraction tests. The results of x-ray spectrometry are displayed
on Table 2 and shows a higher content of silica, aluminates and iron oxide in sample M
respect to that determined in sample L. In the former case the content of the three
component rises over 89% of the total which is considered much higher than the limit
value of 70% for pozzolanic materials considered by the standard [17].
(a) (b)
10m 10m
(c) (d)
2m 2m
Figure 1: Scanning Electron Microscope images: (a) Sample M for 500X magnification, (b) Sample L for
500X magnification, (c) Sample M for 2000X magnification and (d) Sample L for 2000X magnification.
Table 2: X-Ray Spectrometry results for the samples M and L tested in this work.
Mineral
Sample Sample
Component
M L
[%]
SiO2 62.53 40.58
Al2O3 19.90 10.94
Fe2O3 7.55 5.39
CaO 1.40 16.11
MgO 2.75 1.31
K2O 2.64 2.41
Na2O 1.47 2.00
3. Undrained Stress-Strain Behaviour
A series of tests were conducted for characterization on the effect of aging in the
undrained stress-strain behaviour of the samples selected. The tests included:
unconfined compression tests and odometer tests. All tests were performed at the
natural water content. The undisturbed samples were trimmed from block specimens
while the reconstituted samples were prepared by compaction in a three - piece split
mould 50 mm of inner diameter and 100 mm of height. Compaction was performed
with a 0.795 kg weight hammer falling from a constant height of 39 cm. The number of
layers (N = 3) and the number of blows per layer (n = 4) were set to reach the close
value of density of undisturbed samples.
Figure 2 shows the results for unconfined compression test performed on samples
M and L for different time of aging. Stress-strain curve of the undisturbed sample M
samples show a peak resistance of 340 kPa. When the sample is remolded and
immediately tested (0 days), the resistance drops up to 40 kPa and large deformations
are observed as typical behaviour of a loose structure. As the remolded sample is aged
at constant temperature and moisture, without pressure applied, undrained shear
strength tends to increase up to original values of the soil after 28 days of aging (Figure
2a). Aging also increases stiffness, represented by the increment of deformation
modulus E. The results for L samples show opposite behaviour with no changes in
stiffness and strength with aging as displayed on Figure 2b.
350 50
Undisturbed
Undisturbed
Aged 0 days
300 Aged 0 days
Aged 3 days
40
Aged 7 days Aged 3 days
Vertical Stress [KPa]

250 Aged 14 days
Vertical Stress [kPa]
Aged 7 days
Aged 21 days
30 Aged 14 days
200 Aged 28 days
150
20
100
10
50
0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Vertical Strain [%] Vertical Strain [%]
(a) (b)
Figure 2: Unconfined compression test results. (a) Results for samples M; (b) Results for samples L.
Similar results are obtained for the Oedometer tests carried out on undisturbed and
remoulded samples prepared from M soil and L soil displayed on Figure 3. The
undisturbed samples from M soil exhibit a high preconsolidation pressure of 250 kPa.
When the sample is remoulded and tested immediately, yielding pressure decrease to
120 kPa. However, remoulded samples aged after 28 days partially recovers its initial
structure increasing stiffness and yielding pressure (Figure 3a). Oppositely,
compressibility curves for L are almost identical for undisturbed and remoulded
samples, evidencing no effects of aging on soil structure (see Figure 3b).
0 0
4 4
Vertical Strain [%]
Vertical Strain [%]
8 8
Undisturbed Undisturbed
12 Aged 0 days 12 Aged 0 days
16 16
Aged 21 days
20 Aged 28 days 20
1 10 100 1000 1 10 100 1000

Vertical Stress [kPa] Vertical Stress [kPa]
(a) (b)
Figure 3: Confined compression test results.(a) Results for samples M; (b) Results for samples L.
4. Effective Stresses and Suction Forces
In order to evaluate the possible influence of suction on the observed increase of

stiffness and strength in samples M, a set of samples were prepared and tested in
saturated drained condition in the triaxial cell. Figure 4 displays the results for samples
tested at 150 kPa of confining pressure and at 3 days and 28 day of aging. Notice that
even in drained condition, the effect of increasing stiffness and strength with aging is
noticeably, however, the increment is not as significant as that observed in the
unconfined compression test. This result confirms that most of the observed process
can be undoubtfully attributed to a real cementation mechanism that takes place with
aging of the compacted sample. Furthermore, as can be observed on Figure 4, matric
suction forces shows no variation with aging for the same sample M. The initial
variation is attributed here to the time required by the tensiometer to reach equilibrium.
After 24 hour equilibrium is reached and no further variation is observed that could be
endorsed to suction the increase in stiffness and strength as presented in previous
results.
140
3 = 150 kPa 100
d = 1.23 gr/cm3
w%= 35.2 % 28 days Sample M
120
eo = 1,16 Sample L
80
100
S u ctio n [kP a]
1 - 3 [kPa]
60
80
60
40
3 days
40
20
20
0 0
0 1 2 3 4 5 6 7 8 9 10 11 0 2 4 6 8 10 12 14 16
1[%] Time of Aging [days]
(a) (b)
Figure 4: (a) Drained triaxial test for samples M prepared and tested at 3 days and 28 days after compaction.
(b) Variation of matric suction for sample M and L with time of aging.
. Pozzolanic Activity of the Selected Samples
The word pozzolan was originated to describe a volcanic soil used by the early Romans
near town of Pozzuoli to make their hydraulic mortars. The standard [17] describes
pozzolanic materials as siliceous or alumino- siliceous materials which show little or
no cementitious properties, but when they are finely divided and in presence of water,
they chemically react with calcium hydroxide at ambient temperature to form
compounds having cementitious properties.
A method to evaluate pozzolanic activity was proposed by [18] and consists on the
measurement of electrical conductivity with time of a solution prepared with 5.00 g of
soil (oven dried at 100 C) in 200 ml of saturated solution of calcium hydroxide.
Electrical conductivity was measured by a Multi 340 WTW device which allows
measurement in the range of 0 to 19.99 mS/cm with a resolution of 0.01 mS/cm.
Changes in temperature is automatically compensated reducing measurement error
within 0.5%.
The change in conductivity measured at 120 second (C=C0-C120) is taken as a
parameter of the ongoing pozzolanic reaction. Table 3 illustrates the classification of
puzolanic activity according to this test. The lowering of conductivity reveals the
formation of hydrate aluminum silicates with the corresponding reduction of ion
concentration in the solution. The larger the amount of formed silicates, the larger is
the change in conductivity and the pozzolanic activity. Figure 5 shows the experimental
results obtained for samples L and M tested here. The conductivity of the sample L
remains constant with time, indicating no pozzolanic reaction of the soil. Oppositely,
for M sample the conductivity decreases in 120 s of 1.50 microSiemens/cm (i.e. High
Pozzolanic Activity according to Table 3).
Table 3: Pozzolanic classification of activity according to [18].

Classification C=C0-C120
No Puzzolanic Activity C < 0.4 mS/cm
Low Puzzolanic Activity 0.4 < C < 1.2 mS/cm
High Puzzolanic Activity 1.2 mS/cm < C
Note: C: variation of electrical conductivity, C0: initial electrical conductivity and C120: electrical
conductivity after 120 seconds.
7.5
Electrical Conductivity [mS/cm]
Sample M
7.0 Sample L
6.5
6.0
5.5
5.0
4.5
0 30 60 90 120 150 180 210 240 270 300
Time [seg]
Figure 5: Variation of electrical conductivity with time for samples M and L according to the method
proposed by [18].
. Conclusions
This work presents a fundamental experimental study based on a set of mechanical and
physical tests which objective is to evaluate the effect of aging observed on some
specific horizons of loess formation. Physical tests clearly show that the selected
samples have differences in mineral composition being the presence of a considerable
amount of amorphous silica the main mineral component on samples which increase
strength and stiffness with aging. Loess samples of the M type clearly show an
increasing in stiffness and shear strength for total stress tests and even in effective
stress tests. Suction forces seem not to be responsible for such effects. Results of [18]
seem to be conclusive respect to the higher pozzolanic potential of sample M respect to
sample L. These chemical tests are very simple to perform and yield good and reliable
results.
References
[1] Rinaldi V., Clara J., Santamarina J.C., (2001), The Small Strain Shear Modulus (Gmax) of Argentinean
Loess, International Conference of Soil Mechanics and Foundation Engineering ICSMFE, Estambul.
pp. 495-498.
[2] Rinaldi V., Rocca R. And Zeballos M., (2007) Geotechnical Characterization And Behavior Of
Argentinean Collapsible Loess, In Second Internationsl Workshop on Characterization and
Engineering Properties of Natural Soils, Singapore, Eds. Tan T.S., et al. Balkema, London, Vol. 4, pp.
2259-2286.
[3] Alonso E. E. and Gens A. (1994). Keynote lecture: On the mechanical behavior of arid soils. Proc. of
the 1st International Symposium on Engineering Characteristics of Arid Soils, P. G. Fookes & R. H. G.
Parry (eds.), London U.K., Balkema, Rotterdam, pp. 173 - 205.
[4] Barton M. E. (1994). An Alternative Approach to the Understanding of the Collapse Mechanism in
Desert Sands, Loess and other Collapsing Soils. Proceedings of the 1st International Symposium on
Engineering Characteristics of Arid Soils, London U.K., P. G. Fookes & R. H. G. Parry (eds.), pp. 215
220.
[5] Bennet, R.H., O'Brien, N.R., and Hulbert, H., (1991), Determinants of Clay and Shale Microfabric
Signatures: Processes and Mechanisms, In Microstructure of Fine-Grained Sediments; From Mud to
Shale, Eds. Bennet et al., Springer-Velag, NY, Berlin, Heidelberg, London, Paris, Tokyo, Hong Kong,
Barcelona, pp. 5-32.
[6] Mitchell J. K. (1993). Fundamentals of Soil Behavior. John Wiley & Sons, Inc.
[7] Moretto O. (1948). Effect of Natural Hardening on the Unconfined Compression Strength of Remolded
Clay. Second International Conference of Soil Mechanics, Vol. I.
[8] Seed, H. B., Mitchell, J. K. and Chan, C. K., (1960), The strength of compacted cohesive soils,
Research Conference on Shear Strength of Cohesive Soils, ASCE, University of Colorado, Boulder,
Colorado, pp. 877 - 964.
[9] Trollope D. H. y Chan C. K. (1960). Soil Structure and the Step Strain Phenomenon. ASCE, Vol. 86,
N SM2.
[10] Day P. H. (1955). Effect of Shear on Water Tension in Structurated Clay. I y II, Annual Reports,
Western Regional Research Project W 30. De Mitchell J. K. (1960). Fundamental Aspects of
Thixotropy in Soils. Journal of the Soil Mechanics and Foundations Division, ASCE, Vol. 86, SM 3,
pp. 19 52.
[11] Mitchell J. K. (1960). Fundamental Aspects of Thixotropy in Soils. Journal of the Soil Mechanics and
Foundations Division, ASCE, Vol. 86, SM 3, pp. 19 52.
[12] Mukabi J.N., Tatsuoka F., (1999), Effect of Stress Path and Aging in reconsolidation on Deformation
Characteristics of Stiff Natural Clays, Pre-failure Deformation Characteristics of Geomaterials,
Jamiolkowki, Lancelota and lopresti (Ed.), Balkema, Rotterdam, Vol. 1, pp. 131-140.
[13] Troncoso J. H. and Garcs E. (2000). Aeging effects in the shear modulus of soils. Soil Dynamics and
Earthquake Engineering, Vol. 19, pp. 595-601.
[14] Shibuya S. (2000), Assesing Structure of Aged Natural Sedimentary Clays, Soils and Foudations, Vol.
40, N 3, pp 1-16.
[15] Shibuya S., Mitachi T., Temma M., (2001), Evaluating Aging Effect on Natural Clays, Geotechnical
Engineering Journal, Vol. 32, No. 2, pp. 53-58.
[16] Santamarina J. C., Klein K. A., Wang Y. H. y Prencke E. (2002). Specific surface: determination and
relevante. Canadian Geotechnical Journal, Vol. 39, pp. 233-241.
[17] ASTM C618-12. (2002) Standard Specification for Coal Fly Ash and Raw or Calcined Natural
Pozzolan for Use in Concrete.
[18] Luxn M., Madruga F., Saavedra J. (1989) Rapid Evaluation of Pozzolanic Activity of Natural
Products by Conductivity Measurements. Cement and Concrete Research. Vol. 19, pp. 63-68.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-882
Consolidation of Shales and Porosity-

Permeability Relationship
Valentina FAVERO 1 , Alessio FERRARI and Lyesse LALOUI
Swiss Federal Institute of Technology, Lausanne, EPFL, Switzerland
Abstract. The involvement of shales in many energy-related fields in the last

decades has led to the need to deeply study their mechanical and hydraulic
behaviour. Engineering applications such as the extraction of gas shale and shales
oil, the sequestration of CO2 and the nuclear waste geological storage, require
particular attention to the analysis of the volumetric behaviour, porosity change
and permeability-porosity relationship of the involved material as these aspects are
fundamental for the performance of the shale formation as reservoir rock or cap
rock. In this paper, an advanced oedometric cell is introduced in order to
investigate the transition from the over- to the normally consolidated state of
shales; in addition an analytical method is presented in order to analyse the one-
dimensional consolidation process of high-rigidity shales. As high stresses are
involved, the compressibility of the solid particles plays a role in the volumetric
behaviour of the material. Therefore, the poroelastic response of shales has to be
taken into account to correctly describe the volumetric behaviour and porosity
change of the shales. The permeability of the involved material is computed by the
analysis of the consolidation process and its dependency on the porosity is
investigated in the paper. Selected test results are presented for a shale from the
northern region of Switzerland.
Keywords. Shales, Consolidation, Porosity, Permeability
1. Introduction
The involvement of shales in many energy-related fields in the last decades has led to
the need to deeply study their mechanical and hydraulic behaviour. Engineering
applications such as the extraction of gas shale and shales oil, the sequestration of CO2
and the nuclear waste geological storage, require particular attention to the analysis of
the volumetric behaviour, porosity change and permeability-porosity relationship of the
involved material as these aspects are fundamental for the performance of the shale
formation as reservoir rock or cap rock. Such analysis requires tackling important
issues since the stress history, diagenesis and cementation may cause these materials to
have high yield stresses. Therefore, testing under high confining stresses is necessary
as well as taking into account the poroelastic behaviour of the material when evaluating
porosity and permeability changes.
In this paper, an advanced oedometric cell is introduced to analyse the transition
from the over- to the normally-consolidated state of shales; the apparatus allows
1
V. Favero et al. / Consolidation of Shales and Porosity-Permeability Relationship 883
investigating the volume change behaviour of shales at high vertical stresses (up to 100
MPa). An analytical method is then presented to analyse the one-dimensional
consolidation process of high-rigidity shales. The method allows to identify the
primary consolidation processes associated to the vertical load changes and to gather
information on the creep behaviour. As high stresses are involved, the compressibility
of the solid particles plays a role in the volumetric behaviour of the material. As a
consequence, the poroelastic response of shales has to be taken into account to
correctly describe the volumetric behaviour and porosity change of shales. The
permeability of the involved material is computed by the analysis of the consolidation
process and its dependency on the porosity is investigated.
Selected test results are presented for a shale from the northern region of
Switzerland. The results highlight how the proposed methodology allows gathering
information on the compressibility, consolidation coefficient and permeability of shales
in wide ranges of confining stresses.
2. Tested material
A shale from the northern region of Switzerland is considered in this study. The
Opalinus Clay from the Mont Terri Underground Rock Laboratory (URL) from the
shaly facies, consists of dark grey silty, calcarerous shales with a clay content of 55-
60%, 25-30% carbonate, 5-10 % quartz and 10-15% biodetritus.
The results of the geotechnical identification of the core are reported in Table 1,
which presents the particle density (s), the bulk density (), the water content (w), the
void ratio (e), the degree of saturation (Sr) and the Atterberg limits (the liquid limit wL
and the plastic limit wP).
Table 1. Geotechnical characterization of the tested shale [1].
s w e Sr wL wP
Shale
(Mg/m3) (Mg/m3) (%) (-) (%) (%) (%)
OPA-MT 2.74 2.46 6.9 0.21 92 38 23
3. The advanced high-pressure oedometric cell
A multi-purpose oedometric cell is developed to analyse the transition from the over- to
the normally-consolidated state of shales; the apparatus allows investigating the
volume change behaviour of shales at high vertical stress (up to 100 MPa). The layout
of the apparatus used is depicted in Figure 1. The cell is designed to hold cylindrical
samples (12.5 mm in height and 35 mm in diameter). The oedometer cell is inserted
into a rigid stainless steel frame. The loading ram is positioned in the lower part of the
system to prevent the specimen from being loaded before the test starts. The vertical
load is applied by a hydraulic jack connected to a volume/pressure controller. The
maximum vertical stress that can be imposed on the specimen is 100 MPa. The
volumetric strains are measured by three LVDTs (with a resolution of 1 m), which are
fixed to the frame and are in contact with the loading ram. The tests can be performed
in incremental loads, in a constant loading rate mode or in a constant vertical strain rate
884 V. Favero et al. / Consolidation of Shales and Porosity-Permeability Relationship
mode. Tw wo pore water pressure conttrollers are ussed to control the pore wateer pressure
at the botttom and top pllatens and to m
measure the volume
v changees of the pore water.
Figure 1. The high presssure oedometric cell

c at the LMS-E
EPFL [2].
The experimental
e campaign
c pressented in this work, consistts in the perfoormance of
high pressure oedomettric tests on the presented d shales in saaturated and isothermal
conditionss. The shale specimens
s aree initially satu
urated at consstant volume applying a
50kPa waater pressure ata the two basees and the swelling pressurres developed during the
saturationn phase are reccorded. Synthhetic pore wateer is used in the experimennts with the
aim to repproduce the coomposition off the in-situ water [3]. Several days are reequired for
the saturattion phase to be
b completedd (i.e. constantt swelling presssure is observved and no
more wateer flowing into o the cell is reecorded). Afteerwards, loadiing-unloadingg cycles are
performedd in steps, alllowing compllete dissipatio on of the excess pore wateer pressure
(drained conditions)
c an
nd the time-seettlements currves for each loading-unlooading step
are obtainned.
4. Analytical method
An analyttical method is developedd here to anaalyse the consolidation behhaviour of

shales whhen subjected tot a time depeendent loading. In fact, the use of pressuure-volume
controller allows the application
a off very high prressures, but, on the otherr hand, the
applied veertical load is not instantanneous. The devveloped analyytical solutionn takes into
account the
t applicatio on of a time--dependent lo oading on the analysis off the one-
dimensionnal consolidattion process oof shales, togeether with the deformation undergone
by the apparatus, th he primary consolidation n settlements and the secondary
consolidattion components.
In thhe considered framework, thhe settlementss of a sample can be expresssed by the
following equation:
t
0
(t ) '
t
'' t
(1)
where: 4^ is settlement related to the device deformations, 4 (* is the part of the

settlement due to the consolidation of the sample under applied load (primary
consolidation) and 4(* is the settlement due to creep phenomena (secondary
consolidation).
The solution of the diffusion equation for a time dependent loading q f t a
is
obtained by [4] where ta is the time of application of the load and d z dq . The load
time curve is divided into differential time steps and the differential excess pore
pressure is obtained as follows:
5 Cv t ta
2Cdq # Mz $ M 2
du z , t
= % sin &e
H2 (2)
m0 M ' H (
where H is the drainage path length, cv is the coefficient of consolidation,

M 0.5 (2m 1) with m being an integer, z is the depth and C is a coefficient defining
the ratio of the increment of pore water pressure over the increment of vertical stress
when 1D compression is applied in undrained conditions and it is equivalent to the B
value defined in isotropic compression conditions. The time dependent loading in
considered as follows:
< qb
6 t , 0 t tb
tb
6
66 # q q $ (3)
q t
;qb % c b & t tb
, tb t tc
6 ' tc tb (
6 qc , t > tc
6
6:
where qb and qc are the loads at the time tb and tc respectively.
4.1. Primary consolidation settlements
Combining eq.2 and eq.3, three different equations of the excess pore pressure, one for
each load phase, are obtained. The vertical effective stress is defined as follows:
z' (t ) z (t ) 2 u (t , z ) q(t ) 2 u (t , z ) (4)
where 2 is the Biot coefficient. The primary consolidation settlements are computed:
q(t ) 2 u (t , z )
2H
'(t) ?
0
Eoed
dz (5)
The detailed development of the equations can be found in [5].

4.2. Secondary consolidation settlements
The expression of the secondary consolidation settlement is obtained considering that a

secondary consolidation component is developed only after the material reaches a
degree of consolidation of 95%. The following expression is consequently derived:
< 0, t @ t95
'' t
; (5)
:2 HC2 logt logt95
, t > t95
With:
T95 H 2 (6)
t95 ; T95 1,12
Cv
4.3. Settlements due to the deformation of the device
As the vertical load is applied in time, also the devices settlements develop in time. As
a consequence they can be expressed as:
<
6
b
t , 0 t tb
tb
6
66 # b$ (7)
0 t
; % c & t tb
, tb t tc
' t c tb (
b
6
6
c 0, MAX , t > tc
6
6:
4.4. Back calculation of the hydraulic conductivity
The best fitting of the time-settlement curves allows the estimation of the oedometric
modulus Eoed, of the coefficient of consolidation cv, of the secondary consolidation
coefficient c and of the poroelastic coefficients C and , for each loading step and thus
has a function of the vertical effective stress or void ratio.
The hydraulic conductivity of the material at each loading step is back calculated
from the obtained parameters according to the following expression [after 6]:
cv2 w
k (8)
CEoed
5. Results
The analytical solution proposed in the previous section is applied for the data analysis
of the high pressure oedometer tests on the presented Swiss shale. The fitting of the
model with the experimental data is performed through the least square method; in this
way the model parameters which define the time-settlement curve are optimized in
order to obtain the best fitting. As shown in Figure 2, the model reveals a good
reproduction of the sample settlements at each time during the loading and unloading
steps. The fitting of the time-settlement curves allows deriving the parameters
describing the consolidation behaviour of the shale: the oedometric modulus Eoed , the
coefficient of consolidation cv , and the secondary consolidation coefficient C2 for each
loading step, highlighting their stress-dependency. The poroelastic parameters 2 and C
are also derived from the fitting of the model: they are expected to be lower than 1 and
to present a dependency on the stress.
Figure 2. Time vs settlements curves for loading and unloading steps.
Figure 3 depicts the oedometric curves of the tested shale: the void ratio change
related to the end-of-primary strains (corrected from the deformation of the apparatus)
versus the vertical effective stress is represented. The vertical yield stress and the
swelling and compression coefficients are also reported. The progressive increase of
the swelling index with the yield stress is observed thus highlighting the mechanical
degradation of the material when subjected to increasing stress and to various loading
and unloading cycles.
Figure 3. Oedometric curves for the tested Shale.
Figure 4 presents the deformation of the apparatus obtained by the best fitting of the
experimental data with the analytical model. The results are compared with the
displacement of the apparatus measured during a calibration test. Good agreement
between the derived and measured displacements is observed.
Figure 4. Apparatus compliance: the predicted values are compared with the measured ones.
6. Conclusions
An experimental device developed at the Laboratory for Soil Mechanics (LMS) of the
Swiss Federal Institute of Technology in Lausanne (EPFL), which allows the studied of
the volumetric behaviour of shales under high vertical stress, has been presented. The
high-pressure oedometric cell is used to analyse the transition from the over- to the
normally consolidated state of the shale samples since the attainment of a vertical
effective stress in the range of 100 MPa is needed to determine the yield vertical stress.
An analytical method that takes into account the application of a time-dependent
loading during the testing of the one-dimensional consolidation of shales has been
elaborated. The analysis of the settlements versus time curves performed with the
developed analytical method yielded information on the consolidation parameters, on
the poroelastic properties and on the permeability at various stress levels and void
ratios.
Selected experimental results for a shale from the northern region of Switzerland
have been presented and discussed.
References
[1] A. Ferrari, V. Favero, P. Marschall and L. Laloui, Experimental analysis of the water retention behaviour
of shales, International Journal of Rock Mechanics and Mining Sciences 72 (2014), 61-70.
[2] A. Ferrari and L. Laloui, Advances in the testing of the hydro-mechanical behaviour of shales, In L.
Laloui and A. Ferrari editors. Multiphysical Testing of Soils and Shales, Springer, 2012, 57-68.
[3] F.J. Pearson, Opalinus Clay experimental water: A1 Type, Version 980318, PSI Internal report TM-44-
98-07, Paul Scherrer Institut, Villigen PSI, Switzerland (1998).
[4] R.E. Olson, Consolidation under time-dependent loading, Journal of the Geotechnical Engineering
Division ASCE 103 (1) (1977), 55-60.
[5] A. Ferrari, V. Favero, P. Marschall and L. Laloui, On the consolidation of shales, (2015, in preparation).
[6] O. Coussy, Poromechanics. John Wiley & Sons, 2004.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-890
Soil Behavior under Unsaturated and Long

term Vertical Cyclic Loading
Binod KAFLEa,1 and Frank WUTTKE a
a
Marine and Land Geomechanics and Geotechnics, Kiel University, Kiel
Abstract. The deformation of the unsaturated granular material under the action of
long term cyclic loading can be interpreted by unique intrinsic soil properties. The
various changes on the soil properties by the application of the cyclic loading creates
a unique response to the applied external forces. As the existence of unsaturated soil
is predominant, over the saturated or dry soil, the study on the cyclic response of
footing over unsaturated soil is an imminent need to understand the response of the
structure and to evaluate its performance in the long term.
In this paper, experimental study on the response of the shallow footing resting over
multiphase granular soil under the action of cyclic loading is presented. The
macroscopic analysis on the evolution of stiffness with the increasing cyclic loops
with the change in soil suction is studied. Furthermore the change on the soil
response with the increasing number of cycle is studied. From the loops of cyclic
loading obtained from experimental results, the stiffness values and the accumulated
displacement are computed for different soil suction. The analyzed result provides
a strong correlation between the theoretical framework based on small strain
principle and soil response in the load displacement domain for multiphase soil
under long term cyclic loading. Furthermore a boundary domain for the
accumulation of displacement in the suction and cyclic history space is generated.
Keywords. Cyclic loading, Small strain stiffness, Unsaturated Soil, Accumulation

of displacement, Long-term Cyclic Loading
1. Introduction
Geotechnical structures are frequently loaded with cyclic loads of various kinds. The
structures free of cyclic loads are rarely present. Structural response from these cyclic
loads can lead to a cyclic stress generation both on the structure and soil. This cyclic
stress which can be a small in magnitude will lead to a large accumulated settlement with
the increasing number of cycles. This may lead to the plastic failure even though the
applied force is within the elastic limit.
The deformation of the granular material under the action of long term cyclic loading
can be interpreted by several ways. The various changes on the soil properties by the
application of the cyclic loading creates a unique response to the applied external forces.
A change in any one of the several properties of the soil will create a response which will
be different from the responses created by altering other properties. One of the parameter
influencing the soil response under the application of cyclic load is soil suction. The
1
B. Kae and F. Wuttke / Soil Behavior Under Vertical Cyclic Loading 891
influence of the microscopic to macroscopic phenomenon of soil by the soil suction has
been under extensive study since decades [1,2]. As the existence of unsaturated soil is
predominant, over the saturated or dry soil, the study on the cyclic response of footing
over unsaturated soil is an imminent need to understand the response of the structure and
to evaluate its performance in the long term.
The study on the cyclic loading and unloading phenomenon, has been undertaken by
notable researchers [3,4] from the microscopic level for single phase soil. Furthermore,
the study on the response of single phase soil under long term quasi static loading is
undertaken by [5]. The response of the footing resting over multiphase soil by the action
of quasi static loading has been a least studied phenomenon. A Macroelement capable to
predict the response of footing resting over multiphase granular material under the action
of long term cyclic loading has been proposed lately by [6] The evolution of stiffness
with the increasing number of cycle with the change in soil suction provides the
understanding of the interrelation between the accumulated displacement and the change
in force exerted by the water meniscus on the geomaterial. As the soil suction varies
constantly within the lifetime of structure, the response of the multiphase soil will
provide a crucial understanding on the performance of the structure at the particular time
of study.
2. Accumulation of displacement under cyclic loading
As the experiment to understand the load displacement behavior is performed in the

displacement framework and its governing principle, rather than the strain framework, a
transformation of the domain is needed. It is assumed that the physical behavior of the
geomaterial will be proportional in the transformed domain.
2.1. Existing accumulation models
Several authors have proposed the accumulation model to predict the accumulation
behavior of the geomaterials. The Explicit accumulation model expressed in Eq. (1)
proposed by [7], is able to predict the plastic displacement for large number of cycle
based on the strain amplitude, number of cycle N, average applied stress, void ratio,
cyclic strain history and the shape of the strain loop through a set of functions fampl fN fp
fY fe f, which are dependent on the applied boundary condition and the soil properties.
(1)
m: tensiorial vector defining the folw rule

fampl : function of strain amplitude ampl
fN : function of historiotropy
fp : function of rate of accumulation (Independent of N)
fY : function of average stress ratio
fe : function dependent on void ratio
f: Function dependent on polarization changes
Similarly the explicit model for cohesive and fine grained soil depending on the
number of cycle, normalized static and dynamic shear stress ratio was proposed by [8]
892 B. Kae and F. Wuttke / Soil Behavior Under Vertical Cyclic Loading
and later improved by [9] and [10]. The accumulation model existing in literature are
proposed either for dry soil or the saturated soil. In the framework of unsaturated soil, a
macroelement Eq. (2), capable of predicting the accumulation behavior of granular
material was proposed by [6], based on the earlier model of [11] and [12].

AQ !
Q !
I
Ag !
dq !
p B QI
*c !
,, k
(2)
The subscripts I represents the values to be calculated at the image point. The
parameter influenced by the soil suction are represented along with (). The plastic
displacement for multiphase soil dq()p is the function of the plastic multiplier at the
image point located on the bounding surface. The cyclic constitutive parameter c is the
function of memory parameter k which is updated when a new memory surface is
reached and , which locates the distance of current state of stress from image point.
2.2. Small strain approach
The response of the soil during the loading and unloading cycle can be best explained
from the small strain approach. The concept of the existence of finite elastic and plastic
displacement during the loading, unloading and reloading cycle as forwarded by the [3,4]
provides the insight to explain the changes on several soil properties depending on small
strain soil stiffness. As the cyclic load is imposed on the soil, the stiffness of the soil
changes from one position to another and also from one level to another following the
cyclic stiffness degradation curve. Let us assume that the constant amplitude cyclic load
is imposed on the soil as shown in Figure 1(the superscript * denotes state of the soil just
after load reversal).

Figure 1. Co-relation between the loading histories with the stiffness history.
Monotonic loading OA1: During the monotonic loading, the stiffness of the soil
changes from the maximum value to A1 following the hyperbolic description.
Unloading A1B1: The process of unloading is the elastic phenomenon with increase
in tangential stiffness as the load reversal takes place. This leads to the abrupt jump in
the stiffness degradation curve from A1 to A*1, keeping the absolute displacement 1
constant. As the unloading continues the tangential stiffness decreases from A*1 and
reaches to the new value B1, which is higher than the stiffness corresponding to the load
reversal point A1. The increase in the final absolute stiffness in the stiffness degradation
curve can also be conceived by the phenomenon that the micro plastic displacement will
not be recovered at the end of the unloading cycle.
Reloading B1A2: The reloading starts with the load reversal at point B1 with the
reversal of the tangential stiffness. This corresponds to the higher stiffness value at the
load reversal point with the jump in the stiffness in the stiffness degradation curve from
B1 to B*1. As the reloading continues in the direction B*1A2 the tangential stiffness goes
on decreasing, which leads to the decrease in the stiffness value in the stiffness
degradation curve and attains the final stiffness at a new point A2 corresponding to the
displacement a2.
Start of the new cycle: When a new cycle starts or when the applied load exceeds
the maximum load from earlier cycle, the memory will be updated and the soil behaves
as if the earlier unloading and reloading cycle did not exists. Similarly the earlier stiffness
degradation curve will be forgotten and a new stiffness degradation curve will be updated
in the memory. As the number of cycle increases the jump between the two stiffness
levels in the degradation curve goes on decreasing and the degradation curve shifts in the
curvilinear path A1An which ultimately leads to the observed accumulation behavior.
According to [13,14] the shear wave velocity measured by the bender element
increases by the factor of two when the state of the sand is changed from saturated to
dry. From the above understanding of the accumulation behavior of soil under cyclic
loading, the soil response depends upon the evolution of stiffness on the stiffness
degradation curve. As the stiffness degradation curve is dependent on the initial soil
stiffness, the increase in the initial stiffness due to the increase in soil suction will lead
to the lower jump during load reversal and the increased rate of change of stiffness with
the increasing number of cycle. This increase in initial stiffness due to change in soil
suction directly affects the response of the soil under cyclic loading as the evolution of
the unloading and reloading modulus directly dependent on the initial stiffness.
3. Experimental studies of multiphase granular material under cyclic loading
3.1. Material used
The material used for the test is the poorly graded Hostun sand with the USCS
classification (SP). The specific gravity of the sand is 2.65 with D50 = 0.35mm. The
friction angle () of the sand for non-lubricated foundation is =41.8. The values of
maximum and minimum void ratio (emax and emin) are 1.041 and 0.648 respectively [15].
The detail dry and multiphase properties of the used material can be obtained from [6].
3.2. Cyclic Loading
A box with the internal dimension of 980mm length, 480mm width and 480mm height
was used to prepare the sample. The sample with the height of 352cm was prepared by
filling the Hostun sand in several layers and compacted with the standard compactor and
with the standardized procedure to ensure the repeatability of the sampling. Decreasing
compaction energy with the depth was applied in several sample to obtain the required
compaction energy to obtain a constant void ratio of 0.7 within the accuracy 0.02. After
the preparation of the sample, de-aired and distilled water was slowly filled from the
bottom to obtain the saturated sample. The saturated sample was drained after 24 hours
to maintain a required constant water level. The soil suction was then measured by the
use of tensiometer in different depth and in different location of the box. The average
suction measured in different tensiometer up to the depth of 1.5 times the breadth of
foundation was used as the representative soil suction. During the test, soil suction is
precisely controlled and kept constant in the soil sample by connecting the box with the
water reservoir. A loading device with the rigid loading frame as shown in Figure 2,
equipped with the precision cyclic loading system capable of applying smallest load of
0.25N and the displacement transducers with the accuracy of 0.001mm was used. A
displacement controlled vertical centric cyclic loading of constant force amplitude of
0.1kN, at the rate of 0.0025mm/sec, with the frequency of 0.1Hz, and with the static
force of 0.8kN was performed on the square rough and rigid footing of the size 10cm by
10cm. Minute care was taken to reduce the smallest possible disturbances during the
cyclic loading to have the consistent result. The saturated soil was applied with 60
number of cycles, the dry soil with 150 number of cycles and two unsaturated soil with
average soil suction of 5.6kPa and 1.7kPa with 500 number of cycles.
Figure 2. Experimental setup for cyclic loading (after[13]).
4. Analysis and discussion
The response of the footing over the imposed 0.1kN of constant amplitude cyclic loading
with varying degree of soil suction is presented in Figure 3. The accumulation of the
plastic displacement with the increasing number of cycle presented in Figure 4 reveals
that the saturated soil accumulates the large plastic displacement continuously for large
number of cycle, unlike the other soil where the first 10 to 20 cycle rapidly accumulates
the plastic displacement. The soil with highest suction value of 5.6kPa shows a very rigid
behavior by accumulating least displacement per cycle.
The plastic displacement (dp) with each number of cycle is presented in Figure
5. In the saturated and the dry soil sample we do not observe the negative displacement
values within a cycle. Whereas in the unsaturated soil we observe the negative value of
displacement in a cycle. This is observed with the higher number of cycle and it increases
with the increasing number of cycle. As the number of cycle increases the magnitude of
plastic part of displacement (dp) of the total displacement (d = dp+ de) goes on
decreasing with the increasing value of elastic part of the displacement (d e) within a
complete loop of a cycle. As the soil is continually loaded with the constant amplitude
the plastic displacement will be observed after a larger number of cycle. Due to this
reason we observe within some cycle a negative value of displacement corresponding to
the elastic rebound phenomenon within the soil skeleton. As the loading progress, the
relocation of the grain, grain crushing, and change in the inter-granular strain shifts the
state of the soil. This change in the state of the soil generates the plastic or micro plastic
displacement. For this reason the plastic or micro plastic displacement within a cyclic
loop will be observed continually with the increasing number of cycle, although after a
large number of cycle this can only be observed after several cycles with the elastic
response. The change in slope (df/ d) within a cycle with the increasing number of cycle
also reveals the similar phenomenon.
Figure 3. Cyclic response of the soil at different soil suction

A large magnitude of the slope is observed in Figure 6 with increasing number of
cycle. Dispersion of slope and the displacement within a cycle is large in unsaturated soil
compared to saturated and the dry. We can observe that the larger the suction the larger
is the dispersion of the values. Furthermore we can observe the change in the magnitude
of slope of the force displacement with the increasing value of suction. For the 500 cycle
the increase in the magnitude of the slope is just by 50% by increasing the suction from
1.7 to 5.6kPa. The Air entry value of used hostun sand for drainage is 1.9kPa and the
residual suction is 3.3kPa [12]. The two suction used for this experiment are chosen such
that one of them lies on the saturated zone (1.7kPa) and another lies on the residual zone
(5.6kPa) of SWCC.
Figure 4. Accumulation of displacement with the number of cycle

From the above analysis we can construct a boundary in a 3D space for the
accumulation of displacement with the inclusion of the suction field. A boundary of the
accumulation of displacement for the increasing number of cycle can be constructed from
Figure 4, as shown in Figure 7.
Figure 5. Plastic displacement within a cycle with the increasing number of cycle
Figure 6. Change in the slope (df/d) of each cycle with increasing number of cycle
Figure 7. 3D boundary of accumulation of displacement in the suction plane

Three boundaries are governed by the relationship of the accumulation of the
displacement with the number of cycle and the another boundary is governed by the
relationship between the accumulated displacement for the given number of cycle. This
will provide an extra platform to understand the physical phenomenon behind the
ongoing process during the cyclic loading and if needed predict the absolute value of
displacement for the given number of cycle.
5. Conclusion
The influence of soil suction on the elastic and micro-plastic behavior of soil under cyclic
loading is observed. Among the coupled elastic and micro-plastic part of displacement
within a cyclic loop, the elastic part of displacement plays a dominant role at larger cycle
for unsaturated soil compared to dry and saturated soil. Very high and some negative
values of slope of load-displacement for certain number of cycle at large cycles is
observed in unsaturated soil providing the evidence of complete elastic response of soil
within those cycles. The complete clastic and the coupled elastic and micro-plastic soil
behavior is observed with alternation at large cycle predominantly for unsaturated soil.
The strong influence on the deformation behavior of soil by the soil suction is observed.
A 3D boundary domain for the accumulation of displacement in the suction and cyclic
history space is generated. A strong domination of soil suction in this boundary domain
is observed.
References
[1] A.W. Bishop, The principle of effective stress, Teknisk Ukeblad I Samarbeide Med Teknikk. , 106(39)
(1959), 859-863.
[2] S.K. Vanapalli, D.G. Fredlund, and D.E. Pufahl, The relationship between the soil water characteristics
curve and the unsaturated shear strength of a compacted glacial till, Geotechnical testing journal. 1
(1996), 259-268.
[3] R. Nova and T. Hueckel, An engineering theory of soil behaviour in unloading and reloading, Meccanica,
16(2) (1981), 136-148,
[4] T. Hueckel and R. Nova , Some hysteresis effects of the behaviour of geologic media, International Journal
of Solids and Structures, 15(8) (1979), 625-642
[5] T. Wichtmann, A. Niemunis, Th. Triantafyllidis, Validation and calibration of a high-cycle accumulation
model based on cyclic triaxial tests on eight sands, Soils and Foundations, 49(5) (2009), 711-728
[6] B. Kafle, and F. Wuttke, Cyclic macroelement for shallow footing over unsaturated soil, 1st Pan-American
Conference on Unsaturated Soils, Cartagena, Colombia, (2013), 521-526
[7] A. Niemunis, T. Wichtmann & Th. Triantafyllidis, Explicit accumulation model for cyclic loading,
International Conference on Cyclic Behaviour of Soils and Liquefaction Phenomena, Bochum, Germany,
(2004), 65-76
[8] C.L. Monosmith, , N. Ogawa, C.R. Freeme, Permanent deformation characteristics of subgrade soil due to
repeated loading, Transp. Res. Rec., Transportation Research Board, Washington D.C. 537 (1975) 117
[9] J.C. Chai, N. Miura,Traffic load induced permanent deformation of road on soft subsoil, J. Geotech. and
Geoenv. Engrg., ASCE, 128 (11) (2002), 907916
[10] Li. D., E.T. Selig, Cumulative plastic deformation for fine-grained subgrade soils, Journal Geotechnical
Engineering, ASCE, 122 (12) (1996), 10051013
[11] C. di Prisco and F. Pisano, Seismic response of rigid shallow footings, European Journal of
Environmental and Civil Engineering: (2011),185-221
[12] F. Wuttke, B. Kafle, Y. Lins, and T. Schanz, A macro-element for statically loaded shallow strip
foundation resting on unsaturated soil. Int. J. Geomech. 13(5) (2013),
[13] G. Cho and J. Santamarina, Unsaturated particulate materialsparticle-level studies. Journal of
Geotechnical and Geoenvironmental Engineering, ASCE, 127(1) (2001), 8496.
[14] F. Wuttke, Wave based parameter identification and process monitoring in Geotechnics Kolloquim der
Deutsche Geophysikalische Gesellschaft ngewandte Gesteinsphysik, Hamburg (2012)
[15] T. Schanz and P.A.Vermeer Àngles of friction and dilatancy of sand', Gotechnique 46(1)
(1996),145/151
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-898
Cyclic behavior of an over-consolidated

remoulded clay
Christophe DANOa,1, Jian HANa and Pierre-Yves HICHER a
a
Research Institute in Civil and Mechanical Engineering, UMR CNRS 6183, Ecole
Centrale de Nantes, 1 rue de la No, 44321 NANTES (France)
Abstract. The cyclic behavior of highly over-consolidated stiff clays has not been
examined so much in the literature. Here, we experimentally studied the behavior
of a remoulded clay submitted to cyclic loadings. The remoulded samples were
prepared from natural specimens retrieved in the north of France. First, the shear
strength characteristics along monotonic triaxial stress paths were identified. Then
load control cyclic tests were conducted in fully saturated and undrained
conditions. Small cyclic deviatoric stresses were applied in order to investigate
more particularly the behavior under a very large number of cycles (more than one
million). It was shown than even after 106 cycles, failure did not necessarily
occurred in most of the samples unless the cyclic stress level was greater than a
critical value. Moreover, for those over-consolidated samples, an unusual pore
pressure evolution, compared to normally consolidated specimens, was observed.
Additional creep tests showed that two processes were simultaneously acting in a
competitive manner: increase of the pore pressure due to the cyclic loading,
decrease of the pore pressure because of creep.
Keywords. Over-consolidated clay, cyclic behavior, creep
1. Introduction
Various design rules are proposed to geotechnical engineers to calculate both shaft
capacity and base resistance of piles under axial or lateral loads. Most of these rules are
calibrated from in situ tests (cone penetration test, pressuremeter test) that allow
estimating the static capacity. However, there are more and more situations where
geotechnical engineers may have to consider the effects of repeated loads with variable
amplitude with time. Wind, waves, tides are examples of physical phenomena that
induce cyclic loadings on structure foundations. A key issue is therefore to know
whether these cyclic effects require performing an explicit but complex design, or, as
suggested in most of the standard rules, inclusively overestimating safety factors.
Laboratory testing on soil specimens can advantageously provide useful
information to detect a critical sensitivity to cyclic loadings. In this paper, the results of
cyclic isotropically consolidated undrained triaxial tests on over-consolidated stiff clay
specimens are presented. The specificity of the experimental program is to investigate
the response of soil specimens subjected to a very large number of cycles (about one
million), which is not common in the literature.
1
Corresponding author: [email protected]
C. Dano et al. / Cyclic Behavior of an Over-Consolidated Remoulded Clay 899
2. Geotechnical properties of the clay
Clay specimens were retrieved at Merville, north of France. The site, located in the
local airport area, was also the place where field pile tests were conducted, focusing
mainly on the pile response to cyclic axial loading. The results of pile tests are
presented in [1]. One of the main conclusions of this field test campaign is that as long
as the maximum axial load reached during the one-way cycling sequence remains
below a threshold value of about 90 % of the static ultimate capacity in compression
Quc = 900 kN for piles of 420 mm in diameter and 13 m in length, then the accumulated
axial displacements are negligible, even for a number of cycles greater than 1000.
2.1. History of the site
The clay encountered in Merville lies within the Flanders clay deposit located in the
northern Europe. It presents close similarities with London clay or Boom clay deposits.
The clay was deposited at the Eocene era in a marine gulf and subsequently covered by
sediments over about 200 meters. The overlying layers were later eroded and partially
substituted by quaternary alluvial deposits of relatively low thickness. It also knew
different geological periods during which chemical cementation and ageing processes
occurred. The Merville clay is consequently an over-consolidated and stiff clay [2].
2.2. Geotechnical profile
Three boreholes, referred as SC1, SC2 and SC3, were cored to a depth of 11 meters.
The three very similar cored specimens were used to establish the geotechnical profile
of the site and to prepare samples for lab testing.
The top cover is composed of a 3 meters sandy to clayey silt layer which behavior
has not been investigated. The boundary between this silty cover and the underlying
clay layer of interest here corresponds approximately to the position of the water table.
The clay layer extends to a depth probably more important than the base of the
boreholes.
Clay specimens exhibit cracking patterns, more particularly in the depth range
between 3 to 6 meters, which originate from the erosion process. The existing fissures
can be easily activated by sample preparation. It was therefore difficult to prepare
cylindrical samples for triaxial testing. Moreover, the random distribution of the
fissures, even oriented along a preferential direction, prevents any comparison of
results obtained from inherently different natural samples. So, in this study, it was
decided to investigate the behavior of reconstituted samples, as described in section 3.
2.3. Geotechnical properties of the clay
The mineralogy of the Merville clay, determined using X-ray diffraction, consists
mainly of quartz and illite as dominant elements, and smectite and kaolinite as
secondary elements. The clay fraction (< 2 m) is in the range [20 %; 40%]. The mean
specific unit weight s is 26.3 kN/m3. The saturated unit weight sat is 19.6 kN/m3. The
carbonate content from depths of 4 meters is between 4 to 6%.
The profile of water contents for the borehole SC1 is presented in Figure 1. It is
very similar for boreholes SC2 and SC3. The water contents for the clay layer, beneath
900 C. Dano et al. / Cyclic Behavior of an Over-Consolidated Remoulded Clay
3 meters (red circles in Fig. 1), are relatively constant, between 29 and 36 %. The clay
is fully saturated. Atterbergs limits, namely liquid limit wL (between 89 and 101%;
blue triangles in Fig. 1) and plastic limit wP (between 46 and 61 %; blue squares in Fig.
1), are also reported in Figure 1. Plasticity index PI is therefore in the range [30 %;
55 %], indicating that the Merville clay is a highly plastic material.
(%)
20 40 60 80 100
0
4
w wP wL
Depth (m)
10
12
Figure 1. Profiles of water contents, liquid and plastic limits for borehole SC1.
High pressure (up to 12 MPa) oedometer tests were also carried out on natural and
undisturbed samples (70 mm large, about 20 mm high; depths between 7 to 9 m) to
identify the effective pre-consolidation pressure P and therefore the over-
consolidation ratio (OCR). An apparent pre-consolidation pressure was thus
determined at a stress level of about 2750 kPa using Casagrandes method, which
induces an OCR value of about 27 consistent with the geological history of the site.
3. Mechanical behavior along monotonic and cyclic loadings
3.1. Specimens preparation
A first series of triaxial tests on natural and undisturbed samples was carried out. But,
because of the heterogeneity induced by the inner cracking, test results exhibited a
relatively large scattering, which prevented us from a subsequent comparative analysis.
Therefore, reconstituted samples were fabricated in the following way. The natural clay
was dried in an oven, then grinded in a powder state. A suitable mass of clay powder
was then mixed with water at a water content close to 1.5*wL. The slurry was poured
into a consolidometer and consolidated at a vertical stress of 100 kPa or 300 kPa, for
three weeks. Finally, cylindrical samples (35 mm in diameter, 70 mm in height) were

cut and placed into a triaxial cell.
3.2. CIUc and CIUe triaxial tests
The reconstituted samples were then isotropically consolidated and sheared in

undrained conditions at a shear rate of 2.57 %/h, in compression (CIUc tests) and in
extension (CIUe tests), in order to identify the critical state parameters. Some of the
samples were isotropically unloaded before shearing to test overconsolidated samples,
at an OCR value of 4. This value was not representative of the natural conditions but
allows comparison with published data [5, 6] obtained in different clays. Table 1 shows
the initial characteristics of the reconstituted samples, where r stands for the vertical
effective stress in the consolidometer, w is the water content, e0 the initial void ratio,
P the effective consolidation stress in the triaxial cell and c0 the initial isotropic
effective stress at the beginning of the shearing. Figure 2 represents the stress paths
followed during shearing sequences. The slope of the critical state line is Mc = 1.0 in
compression and Me = 0.90 in extension.
Table 1. Samples characteristics for monotonic triaxial tests.

Test r (kPa) w (%) e0 p (kPa) c0 (kPa) OCR
rCIUc1 300 38.8 1.04 400.0 99.4 4
rCIUe1 300 38.9 1.04 399.5 99.6 4
rCIUc2 300 38.1 1.02 399.9 1
rCIUe2 300 37.8 0.99 289.9 1
rCIUc3 100 50.3 1.28 199.2 1
rCIUc4 100 50.1 1.28 400.6 1
rCIUc5 100 50.1 1.29 699.9 1
400 800
rCIUC1 rCIUC3
300 rCIUE1 rCIUC4
Critical state line
M c=1.0 rCIUC2 600 rCIUC5
200 rCIUE2
Critical state line
M c=1.0
q (kPa)
100
q (kPa)
400
0
0 100 200 300 400 500
-100 200
p ' (kPa)
-200
0
M e=0.90
0 200 400 600 800
(c)-300 (c) p ' (kPa)
Figure 2. Stress paths during monotonic shearing in compression and extension.
3.3. Cyclic tests on reconstituted samples
Reconstituted samples for the cyclic tests series followed the same preparation
procedure than previously described in section 3.1. They were all isotropically
consolidated at an effective mean stress of 400 kPa and isotropically unloaded to an
initial effective mean stress of 100 kPa, so that the OCR value was 4. The undrained
cyclic shearing was load-controlled. The mean deviatoric stress was null for all the
tests. Therefore, the sine shape cycles are defined by the frequency set to 1 Hz and the
cyclic stress half-amplitude qcyc. Table 2 shows the initial characteristics of the five
tested samples. Rc is the ratio qcyc / qmax where qmax = 194.5 kPa is the maximum
deviatoric stress obtained in a monotonic CIUc test on a sample with OCR = 4. N is the
total number of cycles.
Test results are presented in Figure 3, in terms of cyclic and permanent (mean)
shear strains, as well as cyclic and permanent (mean) pore pressures vs. the number of
cycles. Except for Test 5 with Rc = 0.52, the excess pore pressure started to increase as
usually observed in cyclic tests but decreased until the end of the loading. This point
will be discussed later.
At low stress ratios (i.e. Rc = 0.11 and 0.22), the cyclic effective stress paths did
not reach the critical state lines, neither in compression nor in extension. For Rc = 0.44,
the stress path exceeded the critical state line in extension only. For Rc = 0.48 and 0.52,
the stress paths reached the two critical state lines. A precise analysis of the
experimental data for Tests 4 and 5 shows that the specimens failed before the end of
the cyclic sequences: it was obvious for Test 5 for which the test was stopped after
6990 cycles. As observed during field tests on piles, it seems that the over-consolidated
clay, even if the OCR values are different, exhibits a threshold value Rc of
approximately 0.45 below which cycles do not lessen significantly the soil resistance.
Table 2. Samples characteristics for cyclic triaxial tests.

Test w (%) e0 qcyc (kPa) Rc OCR N
1 37.2 0.98 21.4 0.11 4 1,026,575
2 37.4 1.00 41.7 0.22 4 1,060,200
3 37.0 0.97 84.7 0.44 4 1,051,480
4 37.7 0.95 92.8 0.48 4 1,050,520
5 38.5 1.01 101.0 0.52 4 6,990
For comparison (Figure 4), the test results on reconstituted Merville clay (main
mineral: illite, PI = 55 %, clay fraction CF = 26%, OCR = 4) were compared with data
obtained by [5] on Drammen clay (hydromica and feldspar, PI = 27 %, CF 50%,
OCR = 4) and by [6] on both Black clay (kaolinite and illite, PI = 30 %, CF = 54 %,
OCR = 4) and Bentonite clay (smectite, PI = 54 %, CF = 69 %, OCR = 4). As shown in
Figure 4, the sensitivity of clay specimens to cyclic loading, through the evolution of
the cyclic axial strains, depends on the mineralogy [6] related to the plasticity indexes
PI. Clays (Drammen, Black clays) with smaller PI are thus more sensitive than clays
with higher PI (Bentonite, Merville clay) because they present a weakest resistance to
the evolution of their microstructure.
3.4. Investigation about the decrease of the excess pore pressure
The duration of the cyclic tests, with more than one million cycles, is about two weeks.
After checking that consolidation was fully achieved before shearing and that no
unexpected water leakage occurred, the question of creep was addressed. An additional
undrained triaxial test was then carried out. A deviatoric stress of 50 kPa was applied in
about one minute. The evolutions of axial strain and pore pressure are presented in
Figure 5. It can be seen a continuous decrease of the pore pressure due to creep, which
can also explain the specific evolution of the excess pore pressure during cyclic tests
where cycles and creep can be considered as two opposed mechanisms over the test
period.
Figure 3. Cyclic tests results.
4 Rc =0.48 Merville clay -this study

Drammen clay -Andersen
Black clay -Hicher
0.75 0.520.440.56 0.52 Bentonite clay -Hicher
3
0.52
cy (%)
2 0.35
0.48
0.42
1 0.34
0.44
0.22
0 0.11
1E+0 1E+1 1E+2 1E+3 1E+4 1E+5 1E+6 1E+7

Number of cycles
Figure 4. Sensitivity to cyclic loads.
1.5 20
Excess pore pressure (kPa)

Axial strain (%)
0
1 0 2000 4000 6000 8000
Time (min)
-20
0.5
-40
0 -60
(b)
(a) 0 2000 4000 6000 8000
Time (min)
Figure 5. Creep test.
4. Conclusions
In this paper, the behavior of a highly plastic and over-consolidated clay under cyclic
loading is investigated. By comparison with previous data published in the literature, at
similar over-consolidation ratio of 4, we show that mineralogy and plasticity indexes
play a fundamental role on the resistance to cyclic loads. Moreover, cyclic undrained
triaxial tests, with an unusual number of cycles larger than one million, at a frequency
of 1 Hz, reveal that two mechanisms have to be considered: increase of the excess pore
pressure due to the cycles and simultaneous decrease due to creep. Finally, our
experiments on the Merville clay confirm the possible existence of a threshold stress
ratio below which cycles have only a small effect on the soil resistance.
References
[1] O. Benzaria, A. Puech, A. Le Kouby, Cyclic axial load tests on bored piles in overconsolidated Flanders
clay, Proceedings of TC209 Workshop: Design for cyclic loading piles and other foundations, 18th
ICSMGE, Paris (2013), 73-76.
[2] H. Josseaume, Proprits mcaniques de largile des Flandres Dunkerque et Calais, Revue Franaise de
Gotechnique 84 (1998), 3-26.
[3] D.W. Hight, A.J. Bond, J.D. Legge, Characterization of the Bothkennar clay: an overview, Geotechnique
42(2) (1992), 303-347.
[4] F. Cotecchia, R.J. Chandler, The influence of structure on the prefailure behaviour of a natural clay,
Geotechnique 47(3) (1997), 523-544.
[5] K.H. Andersen, J.H. Pool, S.F. Brown, W.F. Rosenbrand, Cyclic and static laboratory tests on Drammen
clay, Journal of the Geotechnical Engineering 106(GT5) (1980), 499-529.
[6] P-Y. Hicher, Contribution ltude de la fatigue des argiles, PhD thesis, Ecole Centrale de Paris (France)
(1979).
Acknowledgements
This experimental investigation was done in the framework of the SOLCYP National Project. The authors
would like to thank IREX (Institut pour la Recherche Applique et lexprimentation en gnie civil), Fugro
Goconsulting SA and all partners of the project. The PhD candidate was granted by Groupe des Ecoles
Centrales and the Chinese Scholarship Council.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-906
Evaluation of Consolidation Behavior

under Horizontal Drainage Condition
Using Digital Image Analysis
Joon-Young Kima,1 and Choong-Ki Chung b
a
Ph.D Candidate, Department of Civil and Environmental Engineering, Seoul National
University
b
Professor, Department of Civil and Environmental Engineering, Seoul National
University
Abstract. Analytical solutions to predict the consolidation behavior of soft soils

with vertical drains generally assume that soils are compressed only in the vertical
direction. During consolidation, however, soils deform not only vertically but also
radially in cases of radial drainage conditions, even though it is assumed that overall
radial deformation is not occurred. Radial deformation may influence the
consolidation response and also results in unexpected non-uniform soil conditions
during and after consolidation. A new consolidation apparatus, which enables the
deformation of a specimen to be visually observed, and the total vertical stress as
well as the pore water pressure to be measured at several locations, was developed.
Consolidation test using the developed apparatus was performed for reconstituted
kaolinite specimen under horizontal drainage condition. Digital image analysis
technique was adopted to analyze the consolidation deformation of the specimen.
Horizontal deformation as well as vertical deformation developed during the
consolidation process were obtained and the local consolidation behaviors were
evaluated. The results show that consolidation behaviors are distinctive according
to the distance from the drainage surface and the ground surface.
Keywords. consolidation, digital image analysis, horizontal deformation, vertical

drain
1. Introduction
Soft grounds under preloading improvement with vertical drains have been known to
experience mostly vertical deformation with radial drainage. However, even though no
radial displacement is expected in the present consolidation theory, the progressive
consolidation process from vertical drains induces void ratio variation in the radial
direction under no overall horizontal deformation, which is strong evidence of radial
inward displacement of soils toward the drains [1]. Radial deformation may induce
additional excess pore pressure and dissipation rates far from the theoretical solution
particularly near the drainage surface where radial deformation is concentrated.
Consequently, differential settlement, and an increase in total settlement and
consolidation time could be followed.
1
J.-Y. Kim and C.-K. Chung / Evaluation of Consolidation Behavior 907
These radial displacements during consolidation have already been recognized by

many researchers. Pyrah and Tanaka [2] and Atkinson et al. [3] investigated the
possibility of horizontal movements of soils under radial drainage by measuring
horizontal variations of water contents of soil specimens at the end of consolidation.
Baek and Moriwaki [4] measured the radial displacements at three locations by
monitoring the movement of magnets installed in the soil specimens during consolidation
test.
To obtain the precise and detailed information on the consolidation behavior under
vertical loading with horizontal drainage, a physical model test equipment under overall
one dimensional deformation condition with the capability of monitoring soil responses
of the entire zone under consideration throughout the whole consolidation process was
developed. Soil movements during consolidation were determined by digital image
analysis using photo images. Detailed strains and void ratios of the entire soil specimen
can be computed. Also, variations of the total stress and pore pressure can also be
measured in several locations in the specimen. Using a reconstituted kaolinite specimen,
a consolidation test with horizontal drainage under vertical loading was carried out and
the spatial deformation characteristics with time during consolidation were evaluated
using digital image analysis.
2. Experimental Program
2.1. Test Apparatus
The schematics of the consolidation testing apparatus developed for this study is
illustrated in Figure 1. The apparatus takes a rectangular parallelepiped specimen of 150
mm in height, 140 mm in width, and 40 mm in thickness, surrounded by four transparent
acrylic walls. The photo images taken through the front wall during the test were used
for digital image analysis to evaluate the displacements of the specimen during the entire
consolidation process. On the opposite side, six pore pressure transducers at two different
vertical locations and three different horizontal locations are installed to obtain the pore
pressure distributions from the drain boundary. Three earth pressure gauges are mounted
on the top rigid loading platen, which measure the total vertical stress variations. A plate
of porous plastics used for drain materials are placed on the top of the specimen for
vertical drainage and at the side for horizontal drainage. Through the pressure lines
connected to the top and side of the model box, the samples were saturated by applying
back pressure and the developed excess pore water pressures were dissipated.
908 J.-Y. Kim and C.-K. Chung / Evaluation of Consolidation Behavior
Figure 1. Schematics of the developed consolidation test apparatus
2.2. Test Specimen and Test Condition
Reconstituted EPK kaolinite was used as the test specimen. The index properties of the
reconstituted EPK kaolinite are summarized in Table 1. The clay specimens were created
using the slurry consolidation techniques described by Sheeran and Krizek [5]. Dried
EPK Kaolinite powder was mixed with de-aired water until the water content reached
twice the liquid limit of Kaolinite. The prepared slurry was poured into a large
consolidometer with a diameter of 0.3m. The slurry was then loaded with 10 kPa
increments until the applied vertical pressure reached 100 kPa. After reaching the target
pressure, the vertical stress was kept constant for 7 to 10 days for complete consolidation.
The maximum past pressures of the reconstituted clay samples were confirmed as 100
kPa from the oedometer test.
Table 1. Index properties of the reconstituted EPK kaolinite

LL (%) PI (%) Specific Gravity USCS e0 pc (kPa)
64.4 21.8 2.62 MH 1.60 100
After trimming the reconstituted kaolin sample with auxiliary trimming tools which
are devised to minimize the disturbance of the sample, the consolidation apparatus was
assembled. Vacuum grease was applied on the side walls to minimize undesired friction.
Oil-based black paint was sprayed on one of the board surfaces to obtain random image
patterns for digital image analysis as shown in Figure 2.
Figure 2. Assembled test apparatus and image pattern on the sample surface
A plate of porous plastic was placed on one of the narrow sides of the specimen to
induce horizontal drainage during the consolidation test. The specimen was vertically
loaded through the rigid loading plate by a stress controlled loading system. The vertical
load and displacement were measured by a load cell and LVDT, respectively. After
stabilized under a vertical consolidation stress of 50 kPa, the specimen was consolidated
at three vertical stresses of 100, 150, and 200 kPa. During the whole consolidation
process of each loading step, images were taken at regular time intervals; 10 seconds in
the early stage of consolidation and increasing time steps as consolidation proceeded. A
Nikon D90 digital camera which has an image resolution of 4288x2848 pixels was used
to take the images.
2.3. Digital Image Analysis
Kim et al. [6] suggested a procedure determining the optimum image analysis condition
by statistically analyzing the data of image analysis resulting from the original image of
test specimen with its manipulated one. Following this procedure, the lowest maximum
error, which contains the value of accuracy and precision [7], of approximately 0.002
mm in 90 % confidence level was obtained when using PIV with bi-square interpolation,
60 x 60 pixel subset size and a vertical strain interval of 0.28%. As shown in Figure 2,
the adopted image in the test specimen is divided into 1554 (3742) pixel subsets; thus
providing 1554 displacement vectors at the center points of each pixel subset. The center
points of each pixel subset are equally spaced at 3.5 mm. Then, the strain was calculated
from the displacement vectors to investigate deformation characteristics. To calculate
strain, the square unit element with four equally-spaced center points of the pixel subsets
was first defined. For each unit element, the strain can be obtained as follows:
(1) Using four displacement vectors within an element, a linear displacement field
is obtained via the least square data fit. The linear fields of the horizontal and vertical
displacements for the i-th unit element are defined as:
u i aui x bui y cui , v i avi x bvi y cvi (1)
where x and y are the local coordinates of the plane, u (v) is the displacement
increment in the horizontal (vertical) direction, and a, b and c are the coefficients of the
displacement fields.
(2) Using expressions of u and v, the local strains, ixx (horizontal strain
increment) and iyy (vertical strain increment) of the i-th unit element can be computed
by:
Au i Av i
xxi aui , iyy bvi
Ax Ay (2)
In this paper, consolidation data of the loading step from 100 kPa to 150 kPa in a
normally consolidated state were used for the analysis.
Figure 3 shows settlement and volume change with time for a vertical stress increment
of 100 kPa to 150 kPa during the consolidation test under horizontal drainage condition.
The time for completion of consolidation and total settlement were estimated as 2700
minutes and 3.46 mm according to the Casagrande method.
Figure 3. Settlement and volume change with time (for 100 kPa to 150 kPa vertical stress increment)
The time-settlement curve was divided into ten distinct sections to adjust the
optimum condition of the digital image analysis with a vertical strain interval of 0.28%.
Image analysis was performed using the images taken when the vertical displacement
reached every 10% increment of the total settlement. The resulted 1554 displacement
vectors were used to construct the contours of strain increment at each time section.
Interpretation of the strain contours revealed three distinct stages based on deformation
characteristics of the soil specimen. Figure 4 shows representative horizontal strain
increment contours under horizontal drainage for early, intermediate and late stages
of consolidation, corresponding to 0 to 30 %, 30 to 60 %, and 60 to 100 % of the average
degree of consolidation, Uave , respectively. The compressive horizontal strain was
evaluated to have a positive value and was expressed in blue, whereas the extensive
horizontal strain was evaluated to have a negative value and expressed in red. The
drainage surface was located on the right side of the contours as indicated in the figures.
Figure 4. Horizontal strain increment contours (%)

In addition, the whole specimen was divided into 9 areas as figure 5 and the
evaluated horizontal and vertical strain within the each area were averaged in order to
qualitatively analyze the consolidation behavior of each area.
Figure 5. Divided nine areas
(a) Cumulative horizontal strain (b) Cumulative vertical strain

Figure 6. Cumulative strain curve at the upper part of the specimen

Figure 7. Cumulative strain curve at the middle part of the specimen

Figure 8. Cumulative strain curve at the lower part of the specimen
From figure 4(a), a compressive horizontal strain was observed near the drainage,
whereas an extensive horizontal strain was observed at the far side of the drainage during
the early stages of consolidation (U=0-30%). This is due to the fact that in the horizontal
drainage condition, pore water near the drainage boundary is discharged rapidly with a
decreasing void ratio, and the decrease of void ratio is induced by horizontal soil
displacements (i.e. soils moved in from nearby zone) under the equal vertical strain
condition. This phenomenon can also be confirmed through the cumulative horizontal
strain with time (Fig. 6-8(a)), which indicates that results at the early stages of
consolidation show an increasing positive horizontal strain at the specimen near the
drainage, and an increasing negative horizontal strain at the middle and far drainage
zones, regardless of the distance from the loading plate. The horizontal strain is largest
at the initiation of consolidation, and as the consolidation proceeds, the increment
gradually decreases, passing on to the intermediate consolidation stage. Such
consolidation behaviors in the horizontal direction make the void ratio of the area close
to the drainage become lower than the average void ratio, possibly resulting in a longer
consolidation time than predicted.
In the intermediate stage of consolidation (U=30-60%), the deformation
characteristics which developed during the early stages of consolidation are less
observed, rather indicating a mixed state of both compressive and extensive strain in the
entire soil specimen as shown in Fig. 4(b).
During the late stage of consolidation (U=60-100%), deformation characteristics in
an opposite way from the early consolidation stage are observed, but much less in strain
level. That is, an extensive horizontal strain was observed near the drainage, whereas a
compressive horizontal strain was observed at the far side of the drainage. This is
anticipated to be due to the fact that the excess pore water pressure remaining at the far
side zone is dissipated, resulting in compression, while relatively dense soil at the
drainage side migrate under the equal strain condition. However, the migration toward
the far drainage side during the intermediate and late consolidation stages are relatively
very small strains, and resultantly, positive horizontal strain at near side of drainage and
negative at far side are observed at the end of test as shown in Fig. 6-8(a). The differences
of induced horizontal strain, or volumetric strain in the horizontal direction after the
consolidation may cause differences in strength or stiffness of soil under the equal strain
condition and differential settlement under the free strain condition, both of which should
be accounted in the design process.
(a) Early stage (b) Intermediate stage (c) late stage

(U=0-10%) (U=30-40%) (U=60-70%)
Figure 9. Vertical displacement increment contours
The vertical deformation also showed spatially heterogeneous distribution although

the consolidation test was performed under the equal strain condition. From figure 9
which displays vertical displacement contours, it can be seen that comparatively uniform
displacement according to depth is observed at the undrained side. However, the vertical
displacement is concentrated in the upper part of the drainage side where the pore water
is likely to be discharged most rapidly due to its close distance from the load and drainage
surface. For this reason, at the drainage side, the vertical strain is largest at the upper part
of the specimen, whereas it is lowest at the lower part as shown in figure 6-8(b). As a
result, the vertical strain is largest at the upper part of the specimen at the drainage side,
and at the undrained side it is largest at the lower part. The sum of all vertical strains at
the upper, middle, and lower part according to the drainage distance becomes almost
identical, satisfying the equal strain condition. The comparatively larger horizontal
deformation of the lower part of the specimen (Fig. 6-8(a)) can be explained by the
spatial distribution of vertical deformation, that is, relatively small vertical deformation
of the lower part.
4. Conclusions
In this study, a consolidation test was performed under horizontal drainage condition
using a newly developed consolidation apparatus. The developed equipment can visually
monitor inner soil deformations and measure the pore water pressure and total vertical
stress distributions of the specimen. Digital image analysis was adopted to evaluate the
displacement of the entire specimen and the strain distribution as well. The consolidation
behaviors derived from the results of image analysis showed complex patterns according
to the degree of consolidation and the locations, and are briefly summarized as follows.
A compressive horizontal strain was observed near the drainage, whereas an
extensive horizontal strain was observed at the far side of the drainage during the early
stages of consolidation due to horizontal migration of soil particles into the voids where
the discharged pore water was originally filling. Passing through the transitional
intermediate stage, the opposite deformation behavior to the early stage was gradually
observed during the late stage. However, the migration toward the opposite side of the
drainage during the intermediate and late consolidation stages are very small compared
to the migration toward the drainage during the early consolidation stage, resulting in a
lower void ratio near the drainage at the end of consolidation. The vertical deformation
even showed a spatially heterogeneous distribution although the consolidation test was
performed under the equal strain condition.
References
[1] Yune, C.Y. Influence of void ratio variation on consolidation behavior of clayey soil with vertical drains.
PhD thesis. Seoul National University. 1989.
[2] Pyrah, I.C., Smith, I.G.N., Hull, D., and Tanaka, Y. Non-uniform consolidation around vertical drains
installed in soft ground. Proc. Of 12th Europen Conference on Soil Mechanics and Geotechnical
Engineering, 1999. p.1563-1569.
[3] Atkinson, J.H., Evans, J.S., and Ho, E.W.L. Non-uniformity of triaxial samples due to consolidation with
radial drainage. Geotechnique (1985), Vol. 35, No. 3, p. 353-355.
[4] Baek, W. and Moriwaki, T. Internal behavior of clayey ground improved by vertical drains in 3D-
consolidation process. Soils and Foundations (2004), Vol. 44, No. 3, pp. 25-37.
[5] Sheeran, D.E. and Krizek, R.J. Preparation of homogeneous soil samples by slurry consolidometers.
Journal of Materials (1971), Vol. 6, No. 2, p. 356-373.
[6] Kim, J.Y., Jang, E.R., and Chung, C.K. Evaluation of accuracy and optimization of digital image analysis
technique for measuring deformation of soils. Journal of KGS (2011), Vol. 27, No. 7, p.5-16. [in Korean]
[7] Taylor, J.R. An introduction to error rnalysis: The study of uncertainties in physical measurements,
University Science Books, 1999. p.128-129.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-915
Creep in clay during the first few years

after construction
Gustav Grimstada, 1, Magne Mehli b and Samson Abate Degagoc
a
Norwegian University of Science and Technology, Trondheim, Norway
b
Norwegian Geotechnical Institute, Trondheim, Norway
c
Norwegian Public Road Administration, Trondheim, Norway
Abstract. Currently there exist different assumptions and approaches, along with
various parameter interpretation procedures, that are used in the deformation and
consolidation analysis of soils. In this article the importance of correct
implementation of creep formulation in settlement analysis, during the first few
years after construction, is discussed. The significance of using an objective
interpretation of secondary compression index is also illustrated. The Onsy test
fill, which was monitored for a three year period after construction, is used as an
example. Measurements include vertical and horizontal settlement along with pore
pressure profiles. The Onsy test fill is of particular interest because of the
availability of good quality tests on block sampled Onsy clay along with well-
documented field measurements. The analyses show that disregarding creep in the
first 3 years period can underestimate the settlement by as much as 70% or more,
when the block samples are used as a basis for the interpretation of soil parameters.
Keywords. Clay, creep, time dependency
1. Introduction
1.1. Background
Creep during consolidation is shown to be best captured by creep hypothesis B. The

discussion raised by Ladd et al. [1] is, in the authors opinion, now settled with
convincing proofs as shown in e.g. [2-5] where a review of different field and
laboratory data are used to show that the isotache approach convincingly captures the
rate dependent characteristics of clay. However, the importance of incorporating creep
calculation into a settlement analysis is dependent on several factors such as the actual
importance of accurate settlement predictions with time for a particular project, size of
the foundation/load/embankment, soil characteristics and quality of laboratory data
(sample quality) etc.. In this paper the Onsy test fill [6] is analyzed using different
constitutive models. The Onsy test fill was constructed in 1972 and monitored for 3
years. The test fill was heavily instrumented and there exist data consisting of
settlement profiles, horizontal deformation profiles and pore-pressure measurements.
The Onsly test fill is of particular interest due to the availability of block samples
from the Onsy test site. The idea behind analyzing the fill is to demonstrate how
1
Corresponding Author: Professor, NTNU, BAT, Hgskoleringen 7A, N-7491 Trondheim, Norway.
\!?!
\!
916 G. Grimstad et al. / Creep in Clay During the First Few Years After Construction
different assumptions regarding creep, during this three year period, can affect the
calculation results; and, how the resulting calculations compare to field measurements.
Several constitutive models for general stress states are developed on the basis of
the isotache approach [7], e,g. the SSC model [8] and the n-SAC model [9]. However,
in the everyday geotechnical engineering practice such creep models are not commonly
used. There might be several reasons for this:
+ Consensus: The researchers have not bridged the gap between
academia and practice when it comes to modelling of creep in clay.
However, creep models, for use in finite element analyses, have been
available for more than 15 years or so, and 1D creep models have been
available for several decades. This means that the ones in academia
working on creep of geomaterials should make an effort in bringing creep
models into practice. In order to do this, it would be preferable if the
academic could agree on one general framework for incorporating creep
in modelling of clays (i.e. focus on the similarities in different models);
and, distinguish only on advanced features of soil behavior (i.e. other
effects such as small strain behavior, anisotropy and destructuration ).
+ Reliability: Soil models from academia are often not readily available for
commercial use because this would require more intensive testing,
verification and validation work than normally done in e.g. a PhD work.
Practicing engineers need models that are reliable and robust because of
time and cost constraints that are imposed on their projects along with the
need for design consideration of resulting consequences and implications.
+ Evolution: Researchers are always seeking improvement of existing
models. The evolution in research is often faster than in the engineering
practice. If models should reach the practice then researchers should make
an effort in making their new development correlate to behaviors that the
engineer observes using familiar soil parameters.
+ Experience: Experience of an engineer along with the availability and
quality of soil data at disposal is crucial. Lack of sufficient soil data
makes complex analysis challenging. Further challenge lays in that an
engineer needs to understand the models and get experience on what type
of data/knowledge is needed to use a particular model and which
parameters/features are important for particular applications before he/she
can use a model with confidence on a project.
+ Practice: Everyday practice is calibrated against experience of each
engineer. This calibration is dependent on consistency of a certain
sample quality level, experience with similar field conditions and project
type. However, if the condition changes, e.g. better quality data (block
samples?), then the method used is no longer correlated to experience and
hence one would have to rethink the way one does the calculations.
1.2. The CREEP project
Creep of geomaterials (CREEP) is a four year, ERC founded, IAPP (Industry and
Academia Pathways and Partnership) research project. The main aim of the project is to
establish a consensus in creep modelling for geomaterials and endorse the applicability
of creep models in the practice. The Work package 4 (WP 4) of the project states:
G. Grimstad et al. / Creep in Clay During the First Few Years After Construction 917
WP4 relates to model application and recommendations so as to reach the

project aim to formulate creep models for engineering practice. Finite element models
will be defined assuming materials within the scope of CREEP. The sensitivity of
constitutive parameters on predictions at boundary value problem level is looked at in
detail. This will enable the publication of good recommendations for the usage of
proposed models, which is a necessary prerequisite for their use in practical
geotechnical engineering.
2. Modelling creep in clay
2.1. Time resistance concept and equivalents
PLAXIS (www.plaxis.nl) is currently the most used FE package in geotechnical

engineering practice; hence, its Soft Soil Creep (SSC) model [8] is widely used for
creep calculations in soft soils. Several user defined creep models are also available for
use in PLAXIS, e.g. n-SAC [10], ACM [11] and CREEP-SCLAY1S [12]. For a one
dimensional condition, most of the soil models for creep follow the concept of the time
resistance concept introduced by Janbu [13]. Eq. (1) gives the time resistance, R, as
inverse of the strain rate. When R is plotted versus time, one would for a constant
effective stress level, observe an increase in R with time. The time resistance number,
r s , is obtained when this increase in R is linearized with time (Figure 1). Internationally
different quantities are used to define a creep parameter for clay. Table 1 gives an
overview of the different parameters used internationally and their interrelations.
dt
R (1)
d
Time
R
1

R =t/
Pure creep
rs
1
Rref
Time

Figure 1. Graphical determination of time resistance number, r s
Table 1. Different parameters used and interrelations

Compression index Recompression index Secondary
or swelling index compression index or
creep index
International Cc C r or C s C
Cam-clay = C c / ln10 = 3/ln10 (1-)/ C s C / ln10
Plaxis SSC * = / (1+e 0 ) * = / (1+e 0 ) * = / (1+e 0 )
Norway m = 1 / * m OC = ln10 (1+e 0 ) / C s r s = 1 / *
The internationally used creep parameter C is unlike the time resistance number as it
is determined from a log(time) vs void ratio plot where time starts when a total load
increment is applied in an oedometer test. However, time is not an objective measure, a
C value determined from such a plot would be dependent on at which log(time) the
inclination is found. Nash and Ryde [14] presented an alternative way of plotting the
data in a ln() vs plot. In such a plot one will find a line with inclination * (Eq. (2)).
# * $
* C ln % & ln 0
' t0 (
(2)
where */ t 0 represent the strain rate for time = 0 i.e. initial strain rate
Figure 2 and Figure 3 shows the same oedometer data for constant effective stress level,
i.e. strain due to effective stress change is removed. The data is plotted in a strain vs
time plot (Figure 2) and a strain vs rate of strain plot (Figure 3). As seen in Figure 2 the
two load steps has different inclinations of the curves at Time of 100 hours, while in
Figure 3 the same creep parameter is found for both steps. The same would apply if one
plotted the data according to the way it is illustrated in Figure 1. This illustration shows
that one should seek an objective way of determining a creep parameter (Figure 1 or
Figure 3) rather than interpreting it from a log(time) plot. This illustration does not
mean that the creep parameter could not be a function of strain/stress or time itself; but,
that such a relation of strain/stress time dependency should not be derived from a strain
vs log(time) plot. Note also that for a load step that have a high initial creep strain rate
(see the lowest curve in Figure 2), the inclination in a strain vs log(time) plot will give
a good estimate for (i.e. =0.0047 as compared to 0.005)
Figure 2. Illustration of time(log) vs strain plot for two Figure 3. Data in Figure 2 plotted strain rate vs
different stress levels in an oedometer test strain plot
2.2. Extension of 1D creep concept to a 3D elasto-visco-plastic model
Extending the 1D concept of creep to a 3D model has been done by e.g. [8], [9], [11]
and [15]. There are mainly two different assumptions done when extending from 1D to
3D, i.e. how the vertical (volumetric) strain in the oedometer is converted to a plastic
multiplier, Figure 4. Stolle et al. [8] and Leoni et al. [11] assumed that the volumetric
strain rate was equal on a surface with a given size (equivalent stress) while others like
Grimstad et al. [9; 10] and Freitas et al. [15] followed the over-stress principle after
Perzyna [16], meaning that a surface with an equivalent stress measure in stress space
represents a surface with equal value of rate of the plastic multiplier, d/dt, see Figure 4.
The main shortcoming of the equal volumetric strain approach is for stress states close
to the Critical State line (i.e. critical state will not be reached) and that such a model
will not be able to predict creep swelling for over-consolidated clays i.e. the models do
not allow a state of soil to be on the dry side of the critical state line. Another
implication of this is that these models fail to capture effect of post peak strain
variations that can be observed in undrained triaxial tests. These shortcomings do not
exist in the formulations adopted by e.g Grimstad et al. [9; 10] and Freitas et al. [15].
2.3. Short on the different models used in this paper
In this paper five different models are used to simulate the Onsy test fill [6]. The
isotropic Soft Soil Creep (SSC) model [8], the non-associated Structured Anisotropic
creep model (n-SAC) [10] and the Critical State SSCG (CS-SSCG) model (with small
strain) [17] are all used in the FE program PLAXIS (www.plaxis.nl) along with a
model without creep (SS model). Finally the 1D creep model KRYKON [18; 19],
available in the Geosuite package (www.geosuite.se) is used to show effect of 1D vs
2D plane strain assumption.
Figure 4. Curves in p-q space for equal value of the plastic multiplier for the models in [11] and [9]
3. Back-calculation of the Onsy test fill
3.1. Laboratory data and material parameters
Detailed documentation of the Onsy test fill is referred to Berre [6]. Details on the
selection of material parameters, for the calculations presented in this paper, is found in
Mehli [20]. Oedometer and triaxial test from high quality block samples are used to
determine the soil parameters. The optimal parameters are established in the stress
range that is expected to occur in the field situation.
Figure 5. Results from the analyses with the different models, compared with measurements
a) Time deformation curve (original surface under center of embankment)
b) Horizontal deformation under toe at day 36, 147 and 1116
c) Strain vs depth at day 26 and 1116 (1120)
d) Settlement profiles at day 36 and 1108
e) Excess pore pressure under embankment at day 36 and 1116
f) Finite element model of the test fill
3.2. Results
Figure 5 shows results from FE analyses and how they compare with the measurements.
As seen in the figure including creep increases the settlement after 3 years by 70%
(After 20 years the prediction tells that the increase would be 120%). SSC, SS and n-
SAC distinctly over-predicted horizontal deformations directly after application of load
while the CS-SSCG model does a much better prediction of the short term deformation
(undrained deformation). For the long term horizontal deformation, the n-SAC model
seems to do the best prediction; this is due to the anisotropic formulation in n-SAC.
Most of the models tend to show higher pore pressures directly after load application,
while after three years SS model significantly under-predicted the pore pressures. The
main reason for the different pore pressure prediction under the centerline of the
embankment is due to differences in calculated load distribution. For the KRYKON
analysis, in particular, the 1D situation (Boussinesq distribution) combined with too
high initial strain rate is expected to overestimate pore pressure. The more advanced
model (n-SAC) has a more contractive behavior and will therefore also give some
higher pore pressure. Compared to the measurements SSC gives the best fit for pore
pressure after 36 days.
4. Conclusion and recommendations
Simulations of the Onsy test embankment show that accounting for creep during the
first three years after constructing gives a 70% increase in settlement (as in the
observed measurements) versus the calculation without creep. The analysis without
creep also significantly underestimated the pore pressures after this period. However,
the results are dependent on that parameters are determined from high quality block
samples since selection of a reasonable OCR (i.e. the initial strain rate) is crucial,
especially in sections where the stress increase is moderate. The OCR determined from
a laboratory sample is often influenced by sample disturbance; thus, lower than the
value that should be used with the constitutive model. If a reasonable value for OCR
cannot be established then an analysis disregarding creep might produce more
reasonable results for deformations than a creep model will. Since the accuracy of the
results are so dependent on quality of data and extrapolation into much longer time
scales beyond laboratory testing, one should in the authors opinion check calculation
results against hand calculation methods where creep is disregarded.
Acknowledgments
This work is carried out under the Marie Curie Action, European FP7, CREEP project,
Grant number 17 PIAG-GA-2011-286397. The work was done during Secondment of
Magne Mehli from NGI to NTNU and secondment of Gustav Grimstad from NTNU to
NGI. Norwegian Public Road Administration (NPRA) is acknowledged for supporting
the development of improved guidelines for settlement calculation which will be used
by the geotechnical society in Norway. This development is in part a joint effort
between NGI, NPRA and NTNU. Hans Petter Jostad (NGI) and Kjell Karlsrud (NGI)
are acknowledged for their valuable feedback in making this paper.
References
[1] C.C. Ladd, R. Foott, K. Ishihara, F. Schlosser, and H.G. Poulos, Stress-deformation and strength
characteristics. state-of-the-art report., in: Proc. 9th Int. Conf. Soil Mech. Found. Engng,, Tokyo,
1977, pp. 421-494.
[2] S.A. Degago, G. Grimstad, H.P. Jostad, S. Nordal, and M. Olsson, Use and misuse of the isotache
concept with respect to creep hypotheses A and B, Gotechnique 61 (2011), 897-908.
[3] S.A. Degago, S. Nordal, G. Grimstad, and H.P. Jostad, Analyses of Vsby test fill according to
creep hypothesis A and B, in: Proceedings of the 13th International Conference of the
International Association for Computer Methods and Advances in Geomechanics. , N. Khalili and
M. Oeser, eds., Sydney, Australia, 2011, pp. 307-312.
[4] S.A. Degago, G. Grimstad, H.P. Jostad, and S. Nordal, The non-uniqueness of the end-of-primary
( EOP ) void ratio-effective stress relationship, in: Proceedings of the 17th International
Conference on Soil Mechanics and Geotechnical Engineering, M. Hamza, M. Shahien, and Y.
El-Mossallamy, eds., Alexandria, Egypt 2009, pp. 324-327.
[5] S. Leroueil, Compressibility of Clays: Fundamental and Practical Aspects, Journal of
Geotechnical Engineering 122 (1996), 534-543.
[6] T. Berre, Test fill on soft plastic marine clay at Onsy, Norway, Canadian Geotechnical Journal
51 (2013), 30-50.
[7] L. uklje, The analysis of the consolidation process by the Isotaches method, in: 4th Int. Conf.
Soil Mech. Found. Engng, London, 1957, pp. 200-206.
[8] D.F.E. Stolle, P.A. Vermeer, and P.G. Bonnier, A consolidation model for a creeping clay,
Canadian Geotechnical Journal 36 (1999), 754-759.
[9] G. Grimstad, S.A. Degago, S. Nordal, and M. Karstunen, Modeling creep and rate effects in
structured anisotropic soft clays, Acta Geotechnica 5 (2010), 69-81.
[10] G. Grimstad and S.A. Degago, A non-associated creep model for structured anisotropic clay (n-
SAC), in: Numerical Methods in Geotechnical Engineering, CRC Press, 2010, pp. 3-8.
[11] M. Leoni, M. Karstunen, and P.A. Vermeer, Anisotropic creep model for soft soils, in:
Gotechnique, 2008, pp. 215-226.
[12] M. Karstunen, N. Sivasithamparam, R.B.J. Brinkgreve, and P.G. Bonnier, Modelling rate-
dependent behaviour of structured clays, in: Installation Effects in Geotechnical Engineering,
2013, p. 43.
[13] N. Janbu, The resistance concept applied to deformations of soils, in: 7th International
Conference Soil Mechanics Foundation Engineering, Mexico city, 1969, pp. 191196.
[14] D.F.T. Nash and S.J. Ryde, Modelling consolidation accelerated by vertical drains in soils subject
to creep, in: Gotechnique, 2001, pp. 257-273.
[15] F.T.M. Bodas, D.M. Potts, and L. Zdravkovic, A time dependent constitutive model for soils with
isotach viscosity, Computers and Geotechnics 38 (2011), 809-820.
[16] P. Perzyna, Constitutive equations for rate sensitive plastic materials, Quart Appl Math 20 (1963),
321-332.
[17] M.A. Haji Ashrafi, Implementation of a Critical State Soft Soil Creep Model with Shear Stiffness,
MSc MSc thesis, Norwegian University of Science and Technology, 2014.
[18] G. Svan, Program KRYKON, documentation and manual (The Soft clay deformation project.),
STF69 F86017, in, SINTEF, Trondheim, Norway, 1986.
[19] G. Svan, S.O. Christensen, and S. Nordal, A soil model for consolidation and creep, Institutt for
geoteknikk, Trondheim, 1991.
[20] M. Mehli, R3 - Creep analysis of the Onsy test fill, in: PIAG-GA-2011-286397, NGI, 2014, p. 75.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-923
Comparison of two sample quality

assessment methods applied to oedometer
test results
Helene Alexandra Amundsena,b,1, Vikas Thakur a and Arnfinn Emdal a
a
Norwegian University of Science and Technology, Trondheim, Norway
b
Norwegian Public Road Administration, Trondheim, Norway
Abstract. The issue of sample quality is important in regards to determining

reliable and representative soil parameters for fine-grained soils. In the literature,
there exist several assessment criteria that can be used to assign a degree of
confidence to laboratory parameters, such as the strength and stiffness. In this paper,
two commonly used methods for assessment of oedometer tests are compared in
light of laboratory results for a low plastic Norwegian sensitive clay. The results
show that the conclusions drawn from these two methods differ significantly from
each other. Possible reasons for discrepancies between the methods of the sample
quality assessment are discussed.
1. Introduction
The issues related to assessment of sample disturbance have been given much attention
for many years e.g., Bjerrum [1]; Berre et al. [2]; La Rochelle and Lefebvre [3]; Lunne
et al. [4]; Ladd and DeGroot [5]; Karlsrud and Hernandez-Martinez [6]. Laboratory tests
on undisturbed samples of good quality can provide reliable consolidation and strength
properties for design of geotechnical structures. A reliable characterization of fine-
grained soil samples demands an undisturbed sampling followed by careful handling of
the material. However, this is not a straightforward task because of the practical
constraints associated with the sampling techniques, transportation methods, storage and
handling procedures. Experience shows that low plastic fine-grained soils such as
Norwegian sensitive soft clays are prone to sample disturbance - especially when
sampled using tube samplers. On the contrary, the block sampling in such materials is
considered a relatively gentle approach, but is still challenging due to, for instance, the
stress relief that will occur in the sample [3, 5, 7-10]. In short, what exactly constitutes
an undisturbed sample is unknown, as no definitive method exists to obtain a perfect
sample. Therefore, sample quality assessment becomes essential to assign a confidence
level to laboratory test results regardless of the sampling methods and types.
In the literature, there exist several assessment for this. Each of these assessment
criteria are intrinsically linked to one specific type of laboratory test except for
1
Corresponding Author: Helene Alexandra Amundsen, MSc., Geotechnical Engineering Division,
Department of Civil and Transport Engineering, Norwegian University of Science and Technology,
Hogskoleringen 7A, Trondheim, 7491, Norway; E-mail: [email protected].
924 H.A. Amundsen et al. / Comparison of Two Sample Quality Assessment Methods
assessment of quality of oedometer tests. Oedometer test results can be assessed based
either on change in the void ratio (e/e0), a criteria proposed by Lunne et al. [4], or using
the oedometer stiffness ratio (M0/ML), recently proposed by Karlsrud and Hernandez-
Martinez [6]. The former method is a well-adopted method in the Norwegian practice
whereas the latter is relatively new and it is interesting to compare and evaluate the two
methods. This paper does this using laboratory data for low plastic and highly sensitive
soft clay samples obtained from central Norway, a material very sensitive to sampling
and handling. The possible reason for discrepancies between the methods are discussed.
2. Indicators of sample quality
Even the best sampling techniques cause some mechanical disturbance. Okumara [11],
suggests the following quantitative requirements to establish indicators of sample quality.
+ Easy to determine for the perfectly undisturbed conditions.
+ Regularly variable with disturbance, regardless of the depth of extraction, the
stress system experienced, and the soil type.
+ Sensitive to change due to disturbance.
+ Easily and accurately measured.
Table 1. Some indicators and methods of sample quality estimation from the literature.
Year Method Para- Very good to Very poor
meter excellent quality quality
Triaxial and oedometer tests:
1979-1988 Volumetric strain (v0) at in situ v0 <1% >10%
effective stress (v0) [12], [13]
1996 Specimen Quality Designation v0 <1% >8%
(SQD) [14]
Triaxial and oedometer tests
1997 Change in void ratio (e/e0) [4], e/e0 <0.04 (OCR 1-2) >0.14 (OCR 1-2)
which depends on the <0.03 (OCR 2-4) >0.10 (OCR 2-4)
overconsolidation ratio (OCR)
2013 Oedometer stiffness ratio [6] M0/ML >2.0 <1.0
Uniaxial compression tests:
1979 Strain at failure (vf) in an vf 3-5% 10%
unconsolidated and undrained (UU) (UU)
test on soft clay [12]
1980 Unconsolidated and undrained shear su (UU) Relative assessment based on
strength (su (UU)) measured in the information about stress history and
laboratory [15], [5] predicted strength using SHANSEP
Suction and shear wave velocity measurements:
1963-2002 Residual effective stress (s) and s/ps 0.25-0.50 (OCR>1.5)
the effective stress for a perfect 0.05-0.25 (OCR<1.5)
sample (ps) [9], [5], [10]
1996-2000 Soil suction (ur) [16], [17] ur/v0 1/5 to 1/6
2007 Shear wave velocity (V) [18], 0.60 <0.35
Vvh is measured in the field and Vvh/VSCPTU
VSCPTU is from SCPTU.
2010 Combination of normalized shear Lvs Lvs < 0.65 Lvs > 0.8
wave velocity (Lvs) and normalized Lu Lu < 0.4 Lu > 0.6
soil suction (Lu) [19]
1985-2014 Radiography [20], [5] Visual identification of sample disturbance.
Since no conclusive method exists to assess the absolute sample quality, disturbance
of the soil fabric, breaking of cementation and other interparticle bonds, one often uses
H.A. Amundsen et al. / Comparison of Two Sample Quality Assessment Methods 925
the void ratio change or volumetric strain as the indicator of the degree of damage to the
soil structure. Ensuring that the sample quality criteria are independent of sample depth
and soil type is the most difficult requirement to satisfy. Table 1 provides an overview
over different indicators that have been proposed in the literature. Of these, only two
methods are evaluated in this paper. They are briefly discussed in the following section.
2.1. Change in the void ratio (e/e0)
Based on volume change during reconsolidation, Lunne et al. [4] modified the v0-
criterion to a ratio of the change in void ratio (e) and the initial void ratio (e0) at the
start of reconsolidation. The criteria shown in Table 2 take OCR into account. Equation
1 shows that the initial water content (w) and grain density (s) of the clay are included
in the formula for e/e0, which assumes a 100% saturated soil sample.
e 1 e0 1 sw
v0 v0 (1)
e0 e0 sw
The basis for the e/e0-criterion was the influence of sample disturbance on
laboratory measurements, such as volume strain during reconsolidation (v0),
preconsolidation stress, constrained modulus (M0, where M=d/d) and coefficient of
consolidation from oedometer test results [4].
Table 2. Sample quality assessed on basis of e/e0 and M0/ML values from oedometer tests.
Quality Sample quality Ratio e/e0 Ratio e/e0 Ratio M0/ML [6]
rating for OCR 1-2 [4] for OCR 2-4 [4]
1 Very good to excellent <0.04 <0.03 >2
2 Good to fair 0.04-0.07 0.03-0.05 1.5-2
3 Poor 0.07-0.14 0.05-0.10 1-1.5
4 Very poor >0.14 >0.10 <1
2.2. The stiffness ratio (M0/ML)
Based on the shape of the oedometer curve, Karlsrud and Hernandez-Martinez [6] have
proposed a new criterion. It uses the oedometer stiffness ratio, M0/ML, where M0 is the
maximum constrained modulus in the overconsolidated stress range and ML is the
minimum constrained modulus after preconsolidation stress, see Figure 1. Table 2 shows
the criteria.
Figure 1. Definition of modulus relationships from oedometer tests.

3. Laboratory testing
Norwegian Public Road Administration (NPRA) initiated an extensive field

investigation to the northwest of Trondheim, at the Klett site, as part of a proposed
development of a new highway. The samples that were obtained and tested were 160 mm
diameter block samples, 75 mm and 54 mm diameter tube samples.
The site consists of a large thick marine deposit. The Klett clay is a normally to
lightly overconsolidated sensitive clay with high silt content, which makes the clay
extremely low plastic with a plasticity index (IP) of around 4%. The clay is leached with
salt content of about 1 g/l and water content of about 35%.
The tested samples were obtained from depths between 7 and 19 m. An overview
over the engineering characterization of the tested material is presented in Table 3.
Table 3. Material properties of Klett clay
Material properties
Depth interval, m 7-19
Water content w, % 32-37
Plasticity index IP, % 3.6-5.1
Liquidity index IL 2.3-4.0
Void ratio eo 0.86-0.95
Porosity n, % 46-49
Particles <2 m, % 30-35
Particles 2-50 m, % 65-70
Remoulded undrained shear strength, cr, kPa 0.1-0.2
Sensitivity St 120-380
Overconsolidation ratio, OCR 1.2-2.9
Storage time after sampling, days 0-3
3.1. Oedometer test results
9 odometers (Constant Rate of Strain, CRS) were performed with oedometer tests with
a rate of strain of 0.7-1.5%/hr. The tested samples were cut into an oedometer ring with
a diameter of 50 mm and height of 20 mm. Figure 2 shows typical results of oedometer
tests performed on the low plastic sensitive soft clay with different types of samplers.
The quality of both tube and block samples was assessed using the normalized void ratio
change and oedometer stiffness ratio criteria. The test results are discussed in the next
section.
Table 4. Results of CRS oedometer tests on Klett clay
Test Sample Depth w v0 IP c v0 at e/e0 M0 M0/ML
No. diameter v0
(mm) (m) (%) (kPa) (%) (kPa) (%) (-) (MPa) (-)
1 160 10.0 35.7 109.9 4.0 150 4.8 0.098 3.5 2.85
2 160 14.2 33.3 151.5 4.0 182 5.9 0.123 3.8 1.74
3 160 14.9 35.4 159.1 4.0 190 4.2 0.086 6.0 1.82
4 75 12.5 36.7 135.0 4.5 198 4.1 0.081 5.7 2.93
5 75 12.5 42.1 135.2 4.5 170 6.9 0.126 4.0 1.54
6 75 18.7 34.2 196.7 3.6 265 6.5 0.131 7.5 1.35
7 54 7.5 37.0 84.8 2.8 - 4.5 0.089 - <1.0
8 54 9.4 33.0 114.1 5.1 320 3.9 0.082 10.9 2.41
9 54 9.5 33.1 115.1 4.3 270 4.0 0.083 9.1 1.83
Figure 2. Oedometer tests on block (160 mm) and tube samples (54 and 75 mm) from Klett. For more details
on tests, see Table 4.
4. Discussion
The results from Table 4 are presented in Figure 3, which specifies the quality of the
samples based on the e/e0- and M0/ML-criterion. According to the e/e0-criterion, all of
the block and tube specimens are of quality rating 3; poor quality. The same specimens
were also evaluated using the M0/ML-criterion; the majority of the specimens change the
quality rating for the better. For example, a block sample (test no. 1) is classified as
poor, according to the proposal of Lunne et al. [4]. On the other hand, the same sample
gets the denomination very good to excellent when quality is evaluated according to
Karlsrud and Hernandez-Martinez [6].
A specimen that is heavily disturbed, as a 54 mm sample (test no. 7) in Figure 2,
classifies as of poor or very poor quality using both criteria. This is shown in Figure 3.
In fact, the quality of all the test results fall into the category labelled poor, according
to Lunne et al. [4], whereas only test no. 6 and 7 are classified as poor or very poor
according to Karlsrud and Hernandez-Martinez [6].
The test performed on the specimen obtained using block sampling shows less signs
of disturbance, with clearer preconsolidation stress than the tube samples; see Figure 2.
The tests performed on the specimens of better quality, based on both discussed criteria,
is seen to give a significantly higher interpreted preconsolidation stress, see Table 2, such
as tests no. 2 and 3, 4 and 5 or 8 and 9. Preconsolidation stress decreases with increasing
degree of sample disturbance. This observation is in line with the literature e.g. Berre et
al. [2], La Rochelle et al. [21], Lacasse et al. [22], Hight et al. [10], Ladd and DeGroot
[5].
Figure 3. Plot of (a) ratio e/e0 and (b) ratio M0/ML from CRS oedometer tests versus depth for Klett
Karlsrud and Hernandez-Martinez [6] presented e/e0 values versus depth from tests
carried out on high-quality block samples taken from 23 different sites. The data suggests
that the quality of the block samples is generally reduced with depth, and thus with
increasing level of unloading. It may be assumed that the volume change during
reloading is mainly caused by stress relief rather than physical sample disturbance [6]. It
is however difficult to distinguish between these two effects in absence of suction
measurements on the tested samples. The stress relief may have minimal effect on
consolidation properties, but it can be vital for the development of high volumetric
deformation during reconsolidation. This is used as an indicator of the degree of
disturbance of the soil structure in the e/e0-criterion. In practice, the e/e0 ratio is not
used as a soil parameter and does not matter as long as the preconsolidation stress and
stiffness parameter are reasonably well defined during the test.
Karlsrud and Hernandez-Martinez [6] also used the M0/ML-criterion as a basis for
comparison of oedometer test results on block samples versus 54 mm samples. Based on
the evaluated data, the M0/ML -ratio seem to be independent of sample depth, but is highly
sensitive to small changes in the ML value, which lends itself to personal judgment.
The M0/ML and e/e0 methods evaluates the soil disturbance in a different way. The
e/e0-criterion evaluates the soil behaviour during reconsolidation to the in situ stresses,
but the M0/ML-criterion assesses the quality by the shape of the oedometer curve. If one
should equate the M0/ML and e/e0 criteria, or accept the M0/ML as truly representative
of the sample quality yet to be verified.
Figure 4 shows a schematic sketch of a hypothetical swollen clay structure attributed
to stress relief and its reaction to reloading in an oedometer test. The effect of stress relief
is instant for block samples, than for the tube samples due to support from the tube walls.
The sketch in Figure 4 is an attempt to explain the large volumetric change during
reconsolidation of a low plastic sensitive soft clay in an oedometer cell, which may cause
the difference in the criteria. An idealized cardhouse structure illustrates a high void
ratio, in agreement with Rosenqvist [23]. In the middle of the oedometer sample in Figure
4, the clay structure is assumed swollen due to stress relief. During reconsolidation, the
clay structure releases pore water. This is partially because the water holding capacity of
a soil depends on capillary action and the size of the pores (e.g., Rosenqvist [23]),
meaning that a soil with high silt content, such as the Klett clay, does not have the same
ability as clay rich soils to retain pore water. In addition, the clay structure has a disturbed
zone on the sample surface because of the trimming and handling of the specimen in the
laboratory. It is anticipated that this disturbance would be enhanced in a more silty clay
due to the larger silt particles.
Figure 4. Schematic sketch of idealized swollen clay structure (from a block sample) in an oedometer
ring during reloading, assumed cardhouse structure.
5. Conclusion
This study indicates that the assessment of sample disturbance in low plastic sensitive
soft clays is an important topic of research. The comparison of test results obtained on
block and tube samples from the Klett clay shows that M0/ML and e/e0 methods indicate
a somewhat different assessment of sample quality. The purpose of oedometer tests is to
obtain knowledge of parameters such as the constrained modulus and the
preconsolidation stress representative to the in situ conditions. The M0/ML method
consider this aspect and therefore has an advantage over the e/e0 method. However, it
remains to find out whether the limiting values in the M0/ML-criterion, for evaluation of
degree of disturbance, needs to be revised.
6. Acknowledgement
Engineers J. Jnland, G. Winther and P. stensen at NTNU are gratefully acknowledged

for their skills and knowledge that made the experimental work possible. The author
would like to thank Professor S. Nordal from NTNU and Dr. S. Degago from NPRA for
constructive comments on the paper.
The authors wish to acknowledge the support from the Norwegian national research
program Natural hazards: Infrastructure, Floods and Slides (NIFS). This program
(2012-2015) is initiated by the NPRA, Norwegian, Water Resources and Energy
Directorate and Norwegian National Railways Administration. The OFFPHD, a program
by the Research Council of Norway, is gratefully acknowledged for their financial
support.
References
[1] Bjerrum, L., Problems of soil mechanics and construction on soft clays. State-of-the-art report., in Proc.,
8th ICSMFE: Moscow. 1973, 111-159.
[2] Berre, T., Schjetne, K., and Sollie, S., Sampling disturbance of soft marine clays, in Proc. of the 7th
ICSMFE, Special Session: Mexico. 1969, 21-24.
[3] La Rochelle , P. and Lefebvre, G., Sampling disturbance in Champlain clays, in Proc. of the Symp. on
Sampling of soil and rock, ASTM. 1970, 143-163.
[4] Lunne, T., Berre, T., and Strandvik, S., Sample disturbance effects in soft low plastic Norwegian clay, in
Conf. on Recent Developments in Soil and Pavement Mechanics: Rio de Janeiro. 1997, 81-102.
[5] Ladd, C.C. and DeGroot, D.J., Recommended practice for soft ground site characterization: Arthur
Casagrande Lecture, in Proc. of the 12th Pan American Conf. on Soil Mechanics and Geotechnical
Engineering: MIT, Cambridge, Massachusetts. 2003.
[6] Karlsrud, K. and Hernandez-Martinez, F.G., Strength and deformation properties of Norwegian clays
from laboratory tests on high-quality block samples. Can. Geotech. J. 50(12) (2013), 1273-1293.
[7] Hvorslev, M.J., Subsurface exploration and sampling of soils for civil engineering purposes. 1949.
[8] Skempton, A.W. and Sowa, V.A., The Behaviour of Saturated Clays During Sampling and Testing.
Gotechnique. 13(4) (1963), 269-290.
[9] Ladd, C.C. and Lambe, T.W., The strength of undisturbed clay determined from undrained tests, in Symp.
on Laboratory Shear Testing of Soils, ASTM. 1963, 342371.
[10] Hight, D.W., Bese, R., Butcher, A.P., Clayton, C.R.I., and Smith, P.R., Disturbance of the Bothkennar
clay prior to laboratory testing. Gotechnique. 42(2) (1992), 199-217.
[11] Okumara, T., The variation of mechanical properties of clay samples depending on its degree of
disturbance, in Proc. Spec. Session on Quality in Soil Sampling, 4th Asian ISSMFE: Bankok. 1971, 73-
81.
[12] Andresen, A. and Kolstad, P., The NGI 54-mm samplers for undisturbed sampling of clays and
representative sampling of coarser materials, in Int. Symp. on Soil Sampling: Singapore. 1979, 13-21.
[13] Lacasse, S. and Berre, T., State-of-the-Art: Triaxial testing methods for soils, in Proc. of the Symp. on
Advanced Triaxial Testing of Soil and Rock. STP 977, ASTM. 1988, 264-289.
[14] Terzaghi, K., Peck, R.B., and Mesri, G., Soil Mechanics in Engineering Practice, Wiley, 1996
[15] Ladd, C.C., Azzouz, A.S., Martin, R.T., Day, R.W., and Malek, A.M., Evaluation of the compositional
and engineering properties of offshore Venezuelan soils. M. Dept. of Civil Engr., 1980.
[16] Tanaka, H., Sharma, P., Tsuchida, T., and Tanaka, M., Comparative Study on Sample Quality Using
Several Types of Samplers. Soils and foundations. 36(1) (1996), 57-68.
[17] Tanaka, H., Sample quality of cohesive soils : Lessons from three sites, Ariake, Bothkennar and
Drammen. Soils and foundations. 20(4) (2000), 57-74
[18] Landon, M., DeGroot, D., and Sheahan, T., Nondestructive Sample Quality Assessment of a Soft Clay
Using Shear Wave Velocity. Journal of Geotechnical and Geoenvironmental Engineering. 133(4) (2007),
424-432.
[19] Donohue, S. and Long, M., Assessment of sample quality in soft clay using shear wave velocity and
suction measurements. Gotechnique. 60(11) (2010), 883-889.
[20] ASTM, Annual Book of ASTM Standards, in Standard Practice for X-Ray Radiography of Soil Samples,
ASTM. 2014.
[21] La Rochelle, P., Sarrailh, J., Tavenas, F., Roy, M., and Leroueil, S., Causes of sampling disturbance and
design of a new sampler for sensitive soils. Can. Geotech. J. 18(1) (1981), 52-66.
[22] Lacasse, S., Berre, T., and Lefevbre, G., Block sampling of sensitive clays, in Proc. of the Int. Conf. on
Soil Mechanics and Foundation Engineering, 11: San Francisco. 1985, 887-892.
[23] Rosenqvist, I.T., Physio-chemical properties of soils: soil water systems. Journal of the Soil Mechanics
and Foundations Division. 85(SM2) (1959), 31-53.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-931
Model for Predicting and Controlling

Creep Settlements with Surcharge Loading
Yixing YUANa, Andrew J. WHITTLEa and David F.T. NASHb

a
Dept. Civil & Environmental Engineering, MIT, Cambridge, MA 02139, USA.
b
Dept. Civil Engineering, University of Bristol, Bristol BS8 1TR, UK.
Abstract. It is well known that long-term, creep settlements of embankments on

clay can be reduced through prior surcharge loading. This behavior can be
explained by considering the relationship between creep properties and stress
history of the clay, and is broadly consistent with the isotache framework of soil
behavior, which assumes a unique relation between the effective stress, strain and
strain rate (Suklje, 1957). However, most existing isotache models tend to
overestimate the creep strains measured after surcharge loading, because these
models assume isotaches are equally spaced in terms of ratios of strain rates, and
imply a constant value for the secondary compression coefficient, C. This paper
presents a simple modification of the standard isotache concept using two material
input parameters that are readily obtained from creep measurements of lightly
overconsolidated clay specimens in 1-D laboratory oedometer tests. Data from
creep tests on a low plasticity clay from Salt Lake City (Ng, 1998) are then used to
develop relationships between the surcharge-induced overconsolidation ratio
(OCR) and the subsequent creep property, , and initial creep strain rate, . The
proposed relationship indicates a significant reduction in with surcharge
loading as compared to the normally consolidated value C, and hence, can
provide a more realistic basis for predicting and controlling long-term settlements
using surcharge loads. Comparisons are made with tests on reconstituted
Bothkennar clay, which exhibit very different creep behavior. This highlights the
need for future tests on high quality natural clay samples to clarify the creep
behavior for a wide range of OCRs and to validate the performance of the
proposed model.
Keywords. Creep, Isotache, OCR, One-dimensional consolidation, Stress history,

Surcharge
1. Introduction
Surcharge loading is a widely used technique to reduce/control the long-term

secondary settlement of embankments and reclamations on soft clay. The reduction of
creep deformations is attributed to the change of the stress history of the foundation
soil upon the removal of a temporary surcharge. Field and laboratory observations
have shown that the development of creep can be significantly reduced when the soil is
unloaded to a moderate overconsolidation ratio (OCR). Although graphical methods
932 Y. Yuan et al. / Model for Predicting and Controlling Creep Settlements
have been proposed to interpret post-surcharge creep behavior (e.g., [1] and [2]), they
are prone to subjective judgments in curve fitting and are difficult to relate to
theoretical frameworks of soil behavior. There is an obvious need for an analytical
model, which can describe accurately the creep behavior after unloading (i.e., removal
of surcharge load).
The one-dimensional isotache framework, first proposed by Suklje [3], is also well
established as the basis for modeling time-dependent behavior of normally
consolidated clay [4, 5, 6, 7, 8]. This concept assumes a unique relation between
effective stress, and compression strain, given strain rate, . Figure 1(a) gives a
typical illustration of this framework in the effective stress strain space, where
parallel isotache lines represent the - relations for different strain rates (equal
spacing between the three isotache lines indicates that ). Bjerrum [9]
proposed the possible use of such isotache models to predict the creep behavior after a
reduction of , and this approach has been applied by one of the authors in numerical
analyses [10,11]. However, there is a dearth of reliable data on the post-unloading
creep behavior that can be used to calibrate and validate the isotache models. As will
be shown later, standard isotache models tend to overestimate the post-unloading creep
strains, as they eventually revert to creep described by a unique secondary compression
coefficient C (i.e., the same value assumed for normally consolidated (NC) clay); it is
commonly observed that there is a constant ratio between C and compression ratio
CR [12].
This paper presents a simple 1D formulation of post-unloading creep behavior
derived from the isotache concept using two material input parameters (, ). The
proposed equation is then used to interpret the data from a number of carefully
executed creep tests on Salt-Lake-City (SLC) clay [13], and provides predictive
relationships between the surcharge-induced overconsolidation ratio (OCR) and the
subsequent creep property, , and initial creep strain rate, .
sur log' logt
0
Isotaches
1
2 EOP
Path1 C
Path2
0 > 1> 2 Path3
C
a) b)
Figure 1. Predictions of a conceptual isotache model on a) stress-strain response b) strain history under
different loading paths.
Y. Yuan et al. / Model for Predicting and Controlling Creep Settlements 933
2. Isotache Prediction of Post-surcharge Creep Behavior
Figure 1a considers isotache predictions of - curves for three different loading

scenarios. Path 1 shows the stress-strain behavior of an element of clay as it
consolidates under a basic load as well as an additional surcharge sur, and
continues to creep after the End of Primary consolidation (EOP). On the other hand if
the surcharge sur is removed near to the EOP state (emulating field procedures), the
subsequent behavior deviates from Path 1 to Path 2. If the clay is initially loaded with
no surcharge (i.e., only), its behavior will follow Path 3. All paths are predicted
with a constant secondary compression coefficient C.
Figure 1b shows the corresponding temporal evolution of strain ( - logt) curves
for these scenarios. Paths 1 and 3 result in linear creep curves (in - logt space) after
EOP with a unique slope governed by C (the normally consolidated value at large
times associated with a unique ratio of C/CR [12]). For Path 2, the clay initially
rebounds upon unloading but ultimately coincides with Path 3 as the creep reappears.
This result contradicts the majority of the field experience, as it suggests that even
significant temporary surcharging cannot reduce C at large times.
This problem can be traced to the isotache model used to describe the evolution of
creep strain at constant effective stress. The model incorporates a function that relates
the accumulation of creep strain to the logarithmic decay of strain rate , as expressed
by Eq. (1), where governs the isotache spacing in terms of the ratio of (where
is the initial creep rate). The conventional isotache model assumes that is a constant
and equal to C irrespective of the use of surcharging.
1
= 0.434C ln (1)

3. Modified Isotache Formulation
Despite the preceding limitation of the standard isotache framework, Eq. (1) does
provide an insightful and applicable formulation for the creep behavior. Its use in direct
interpretation of measured creep behavior is hindered by the fact that most laboratory
data on creep deformation are presented in the form of - logt. In the following we
propose a simple modification of Eq. (1) that allow a direct comparison with the creep
data in the relation of logt. It involves an integration of a rearranged form of Eq. (1)
with initial conditions = 0 at t= 0 and results in Eq. (2).
1
= 0.434C ln t + 1 (2)
0.434C
Equation (2) expresses the temporal evolution of creep strain under constant
effective stress in terms of two parameters and . This expression can be directly
used to interpret laboratory creep tests that emulate effects of surcharge loading. The
goal is then to establish the relationship between the creep properties and , and
the surcharge-induced stress history (OCR) of the clay.
4. Interpretation of Laboratory Creep Test with Surcharge
Ng [10] conducted a series of experiments to investigate the surcharging effects on the

creep behavior of natural Salt-Lake-City (SLC) clay. The SLC clay samples had an
average wP=22.3% and wL=43.6%, and were tested in conventional oedometer devices.
During each test unloading increments were performed, to imitate the surcharging of
the normally consolidated (NC) clay at a range of consolidation effective stress levels.
After each unloading the creep was monitored over a period of two to three days.
Figure 2 shows typical measurements of deformation for one unloading increment.
The specimen is unloaded at a time tr shortly after the EOP (as indicated by t100 on the
consolidation curve) and rebounds rapidly to a local peak strain state (peak rebound).
The deformation curve then levels off and compression resumes with an increasing
slope indicating a creep-dominant behavior (solid symbol). This unloading induces an
OCR = 1.25 (here OCR is taken as the ratio of the effective stresses immediately before
and after unloading).
Equation (2) has been used to fit the data of the creep deformation after unloading,
and obtain best fits for the parameters and . Figure 3(a) shows the results using
the typical data presented in Figure 2. The plot assumes zero strain at peak rebound and
tracks the elapsed time for the subsequent creep. As can be seen, the derived Eq. (2)
manages to fit the data remarkably well, and most significantly, the best fitted value
is less than half (46%) of the C value that was be measured from the secondary
compression behavior of the normally consolidated SLC clay, i.e., the conventional
isotache assumption (= C) overestimates the measured creep strains (dash line in
Figure 3(a)).
Figure 3(b) shows the same data and the fitted results in - log space, where is
obtained from the time derivative of . In this way, the fitted relationship now reverts to
the model of Eq. (1) showing a log-linear decrease of strain rate with strain level. The
two fit parameters are easily visualized in this space: corresponds to the strain rate
intercept at zero strain, and is the gradient of the - log and appears to be constant
within the 3-day period of the test. Figure 3(b) demonstrates that the creep behavior of
SLC clay after unloading complies with the relationship based on isotache theory.
0.0
3.10 1.0
2.0 t =1.13hr

3.20 100
t =2.08hr 3.0
r
Vertical strain [%]
3.30
4.0 -4
10 10 -2 100 10 2
t
3.40
3.50
3.60 Unload increment Oed. B9-9

Induced OCR=1.25
3.70
100 10
1
10
2
Time t [hr]
Figure 2. Rebound and creep strains from a typical laboratory oedometer test on natural SLC clay simulating
surcharge removal
5. Development of Predictive Relations
The same data-fitting approach has been successfully applied to interpret unloading
data from five multi-stage oedometer tests on SLC clay [13]. The best fit parameters
and are then normalized with respect to the reference secondary compression
coefficient of the normally consolidated clay, C, and the measured strain rate at
time tr just prior to unloading, respectively. Figures 4(a) and (b) show that both
parameters decrease with the level of surcharge-induced overconsolidation. For
example, at OCR=1.4, /C 0.20 indicating an 80% reduction in the creep property
compared to the normally consolidated clay. This finding is consistent with the
empirical relations proposed by Ladd [2], but is achieved with a more systematic
approach. The following relation is used to fit the result,
C 2
= (3)
C OCR + 1
where controls the rate of the decrease of /C with the induced OCR. Eq. (3)
implies that the isotache spacing decreases with the unloading-induced OCR. Other
researchers, notably Yin [14] and Watabe et al. [15], have proposed similar variations
in creep properties with stress history. However, both assume that creep ceases above a
threshold value of OCR. The SLC data show no such limit (for OCR 1.5).
If a conventional isotache model with equal to C is used to predict the creep
rate immediately after unloading, it may be shown [10] that
m
1 1
= (4)
r OCR
where m is equal to the ratio (CR-RR)/C in which CR is the compression ratio, and RR
is the recompression ratio. For the SLC clay, the experimental data imply a narrow
range of the parameter m equal to 19.1 1.3.
Figure 4(b) shows the normalized experimental data for plotted against OCR
on a double logarithmic plot together with the relationships implied by Eq. (4) for the
above range of m values. This shows that the initial creep rate does indeed decrease
with the induced OCR following a power-law fashion and actually may be predicted
well by the conventional isotache model as illustrated by the shaded zone in Figure
4(b).
6. Discussion and Conclusions
This paper proposes a simple 2-parameter model based on isotache theory to represent
the post-surcharge creep behavior of natural SLC clay. Eqs. (3) and (4) present
relationships between the normalized model parameters and the surcharge-induced
OCR. These equations provide a reliable framework for predicting and controlling the
long-term creep settlement of embankments and reclamations using surcharge loads.
=6.79E-3%/hr
1
0.00 0.00
a) 2
b)
R =0.9992
0.05
0.05
Vertical strain [%]

Vertical strain [%]
0.10
^ 0.10
Data B9-9 C=0.00415
0.15 Best fit
^
C = C 0.15
0.20
=6.79E-3%/hr 0.20
0.25 1
^
C=0.00415 < C =0.00895
0.30 0.25
-3 -2 -1
10 -1 100 10 1 102 10 10 10
Time t [hr] Strain rate [%/hr]
Figure 3. a) Curve fitting of post-surcharge data with derived formulation and an overestimation of creep
deformation for case with = C; b) Data and the fitted relation in - log space
1.00
Fitted relation
^ 2
0.80 C / C =
(OCR +1)
=7.20
2
R =0.757
C / C
0.60
< 0.40 Test #

B4
B7
0.20 B8
B9
B10
0.00
0
10
Fitted relation
(-m)
/ =OCR
1 r
(log scale)
m ~ (CR-RR)/C
-1
10
For SLC clay m=19.11.3
r
/
Test #
-2
1
10
B4
B7
B8 : strain rate before
r
B9
B10 removal of surcharge
-3
10
1.00 1.10 1.20 1.30 1.40 1.50
OCR (log scale)
Figure 4. Relations of a) normalized /C and b) normalized with surcharging induced OCR
The model has been mainly validated for the long-term creep behavior of SLC clay
with an induced OCR of less than 1.5. It was fortunate that in this case the swell
process was short and did not interfere with subsequent creep compression.
In an attempt to explore the applicability of the model to other clays, comparison
was made with some carefully executed incremental load oedometer tests on
reconstituted Bothkennar clay (wP 32.57.5%; wL 808%)[16]. Figure 5(a) presents
results from a series of long-term 1D unloading tests at much higher OCRs=1.5, 3 and
6.5. It can be seen that as the OCR increases the swelling deformation becomes
significant (empty symbols), and creep eventually reappears (solid symbols) for all
cases considered. In contrast to the behavior of intact SLC clay (Fig. 3b) where there is
a monotonic decrease of, the reconstituted Bothkennar clay samples consistently
show a slow increase of with considerable accumulation of before the creep slows
down (the creep rate at OCR = 6.5 has not decreased yet). This discrepancy might be
attributed to the coupling between viscous swelling and creep processes.
Another interesting observation in Figure 5(b) is that the creep rates in the
reconstituted Bothkennar clay after unloading to different OCRs are very similar and
bounded in a narrow range from 1x10-7%/hr to 8x10-7%/hr. This finding is very
different from that predicted by Eq. (4) and implies that the surcharging has little effect
on reducing the long-term creep of reconstituted Bothkennar clay, and in contrast to the
observation of natural SLC clay. Clearly the simple 1D isotache relation (Eq. 1) alone
is not adequate to predict the creep behavior of this reconstituted Bothkennar clay.
One possible explanation is that the reconstituted Bothkennar clay has much
weaker inter-particle bonding than a natural soft clay. This weak structure could be
further damaged by unloading to the point at which considerable creep might develop.
The significant differences in creep behavior of reconstituted Bothkennar and
natural SLC clay highlight the need for further studies (preferably using high quality
natural samples) to understand the long-term creep behavior of clay after unloading to
a wide range of OCRs and establish capabilities and limitations of the model proposed
in this paper.
Figure 5. Results of long-term unloading tests on reconstituted Bothkennar clay with induced OCRs =1.5, 3,
and 6.5: a) strain history data and b) data of - log (after [13]).
Acknowledgments
The Authors are grateful for support provided by Ferrovial-Agromn through the
Ferrovial-MIT program. They are also especially appreciative of the advice and
assistance provided by the late Professor Charles C. Ladd.
Reference
[1] Mesri, G.: Discussion of 'Post-construction settlement of an expressway built on peat by

precompression, Canadian Geotechnical Journal, 23(3), 1986, 403-407.
[2] Ladd, C. C.: Unpublished Class Notes for 1.322, Soil Behavior, Department of Civil and
Environmental Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, 1989.
[3]
EE
EEE
E E
EE
E
E
E

EE EE
[4] Leroueil, S. The isotache approach. Where are we 50 years after its development by Professor uklje?
Proc. European-Danube Conference on Soil Mechanics, Ljubljana, 2006, 1, 55-88
[5] Yuan Y. and Whittle A. J.: Examination on time-dependent soil models in one-dimensional
consolidation. Constitutive Modeling of Geomaterials: Advances and New Applications, Springer
Series in Geomechanics and Geoengineering, Springer, 2013, 159-166.
[6] Nash, D. and Brown, M.: A comparison of four elastic visco-plastic models for soft clay. Constitutive
Modeling of Geomaterials: Advances and New Applications, Springer Series in Geomechanics and
Geoengineering, Springer, 2013, 121-124
[7] Rowe, R. K., and Hinchberger, S. D. (1998). The significance of rate effects in modelling the Sackville
test embankment. Canadian Geotechnical Journal, 35(3), 500516.
[8] Watabe, Y., Udaka, K., Kobayashi, M., Tabata, T., and Emura, T. (2008). Strain rate effects on long-
term consolidation of Osaka Bay clay. Soils and Foundations, 48(4), 495509.
[9] Bjerrum, L.: Embankments on soft ground. State of the Art report. Proc. Specialty Conf. on
Performance of Earth and Earth-supported structures, Purdue University. ASCE. Vol. 1, 1972, 1-54.
[10] Nash, D.: Precompression design for secondary settlement reduction. Geotechnique, 51(9), 2001,822
826.
[11] Nash, D. F. T.: Modeling the effects of surcharge to reduce long term settlement of reclamations over
soft clays: A numerical case study. Soils and Foundations, 41(5), 2001, 113.
[12] Mesri, G., and Castro, A. (1987). C/Cc Concept and K0 during secondary compression. Journal of
the Soil Mechanics and Foundations Division, 113(21323), 230247.
[13] Ng, N.S.Y.: Characterization of consolidation and creep properties of Salt Lake City clays. Master
Thesis, Massachusetts Institute of Technology, Cambridge, Massachusetts, 1998.
[14] Yin, J. H.: Non-linear creep of soils in oedometer tests. Geotechnique, 49(5), 1999, 699707.
[15] Watabe, Y., Udaka, K., Nakatani, Y., and Leroueil, S.: Long-term consolidation behavior interpreted
with isotache concept for worldwide clays. Soils and Foundations, 52(3), 2012, 449464.
[16] Sukolrat, J.: Structure and destructuration of Bothkennar clay. Ph.D. Thesis. University of Bristol.
2007.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-939
Pore pressure generation in clays soils

normally consolidated under undrained
cyclic load process
Chaves-Agudelo, Julina,1 and Hernndez-Rodrguez, Flix b

a
ME Geotechnical, Professor Pedagogical and Technological University of Colombia.
b
MSc Geotechnical, Assistant Professor, National University of Colombia.
Abstract. This paper presents a novel approach to study the dynamic behaviour of
a reconstituted kaolinitic clay considering experimental results of static and
cyclic triaxial evaluated in the context of the Critical State Soil Mechanics
(CSSM) and the Cam-clay model. In order to define the mechanism of pore
pressure generation under cyclic loading and under which conditions induce a state
of failure by cyclic softening.
Keywords. Critical state soil mechanics, Cam-clay model, pore pressure, stiffness
degradation, damping ratio, cyclic softening.
1. Introduction
When soil supports dynamic loads, it suffer stress increments and can exert a
significant influence on soil stability due to generated strains or due to mobilization of
shear resistance. Cyclic loading process on saturated clayey soils can be understood as
undrained process, which generates strains and pore pressure increments due to volume
change tendency. When pore-pressure-increments dissipate in a period of time it could
have a negative effect on any infrastructure located on this clayey soils. Different
experimental researches [1-5] have evaluated the influence of amplitude, frequency of
load on stiffness, energy dissipated by damping and development of excesses of pore
pressure. Recent experiments have shown significant effect of cyclic load on strains.
This process is known as Cyclic Softening Mechanism in soil dynamics [6].
Using the well-known framework of Critical State Soil Mechanics -CSSM- [7-9]
and considering the concept of Limit State Surface -LSS-, the admissible stress and
elastoplastic strains are defined. Thereafter, it is possible to define when an applied
cyclic load cannot be supported by soil, since it would require stress states (p,q, v)
over LSS, producing dynamics effects that allow the understanding of pore pressure
behavior during the cyclic load process.
In this research a series of static and cyclic triaxial tests were carried out on
reconstituted and normally consolidated clay soil. With use of CSSM it was possible to
1
North Central Avenue #39-115. Administrative Building, Postgraduate Education Office. Pedagogical
and Technological University of Colombia. TunjaBoyacColombia. [email protected].
940 J. Chaves-Agudelo and F. Hernndez-Rodrguez / Pore Pressure Generation
understand degradation of stiffness, dissipation of energy by damping during

deformations, excess of pore pressure during cyclic load processes.
References [10-16] have exhaustive analysis about the dynamic response of clays
under the framework used in this experimental research. The results provide an insight
of the generation of pore pressure under dynamic load, and the conditions to reach a
failure state due to cyclic softening.
2. Experimental results in static load process
2.1. Physical and Index Properties of reconstituted clayed soil
In this study, samples of kaolin reconstituted soil samples (diameter 2" and height 4")
were obtained, and a summary of the physical and index properties of the reconstituted
kaolin soil are presented in Table 1:
Table 1. Physical and index properties of the reconstituted kaolin soil.
Unit weight, t0 18.4 kN/m3 Water Content, Wn 33%
Specific Gravity, Gs 2.66 Liquid Limit, Wl 47%
Void Ratio, e0 0.90 Plastic Limit, Wp 23%
2.2. Triaxial static tests results and definition of critical state model according to
CSSM
To understand the mechanical behaviour of soil to static loads and eventually in cyclic
loads, it is paramount to determine the properties of stiffness and strength of soil to
predicting its elastic, elastoplastic and failure responses under different initial stress
states, histories and paths stresses in the CSSM framework [17].
This response explains the pore pressure generation process under static loads.
With the experimental results the Isotropic Compression Line -ICL- , Isotropic
Expansion Lines IEL- and Critical State Line -CSL- are established. The definitions
of soil mechanical behavior are as follows:
(1)
(2)
(3)

in triaxial test triaxial compression mode

(4)

2.3. Mechanical behaviour under different confining pressures
In undrained tests on soil samples NC, as well as the deformations to develop, pore
pressure increases by the volume change trend. The results stress - strain - pore
J. Chaves-Agudelo and F. Hernndez-Rodrguez / Pore Pressure Generation 941
pressure normalized to the confining pressure (p'c) are shown in Figure 1. There are
similitudes in the mechanical response of soil at different confinement levels.
q/pc vs. q Uu/pc vs. q

0,8 0,7
0,7 0,6
0,6
0,5
Uu / pc
q / pc
0,5
0,4
0,4
0,3
0,3 sc= 600kPa sc= 600kPa
0,2
0,2 sc= 400kPa sc= 400kPa
0,1 0,1
sc= 200kPa sc= 200kPa
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
q (%) q (%)
Figure 1. Stress strain pore pressure behaviour normalized respect confining pressure.
3. Result analysis according to Cam-clay model
3.1. Mechanical behaviour in the Cam-clay model
In [17] a detailed analysis about how to obtain the elastic parameters of reconstituted
kaolin soil and the physic-mathematical modeling employed to evaluate the mechanical
behaviour (in accordance with the Cam-clay model for a material that harden by strain)
is shown. The mechanical behavior of the kaolin considered as a normally consolidated
soil and ideally isotropic using the same confining pressures, is shown in Figure 2.
Esf corte eq normalizado vs Def corte eq Pore pressure normalizated vs Shear Deformation eq.
0,6 0,8
0,7
0,5
0,6
0,4
0,5
q / pc
Uu / pc
0,3 0,4
Sc= 200 kPa Sc= 200 kPa
0,2 0,3
Sc= 400 kPa 0,2 Sc= 400 kPa
0,1
Sc= 600 kPa 0,1 Sc= 600 kPa
0 0
0 0,01 0,02 0,03 0,04 0 0,01 0,02 0,03 0,04
q q
Figure 2. Prediction of the stress strain pore pressure behaviour for different confining pressure in a soil
elastoplastic and isotropic in the Cam-clay model.
3.2. Cam-clay model predictions
For concordance purposes between the elastoplastic model results and the experimental
results, an overconsolidation effect in the soil is assumed. The overconsolidation
level assumed in the Clam-clay model to consider the anisotropic behaviour of the
reconstituted soil is shown in Table 2.
Table 2. Sobreconsolidation ratio defined in Cam-clay model for to take into consideration the anisotropic
behaviour of kaolin reconstituted soil
Experimental Results Value defined in Cam-clay model
pconsolidation test pinitial test pconsolidation model pinitial model
RSC test RSC model
(kPa) (kPa) (kPa) (kPa)
207.8 203.9 1.02 235.0 203.9 1.15
405.8 401.8 1.01 480.0 401.8 1.19
602.7 597.1 1.01 740.0 597.1 1.24
Figure 3 shows the experimental and the prediction model results considering the
overconsolidation effect for a soil sample NC with a confining pressure of 602.7 kPa.
In this figure the lines LSL-INITIAL and LSL-FAILURE are the limit state lines (LSL)
corresponding to beginning and the end of the failure stage.
Esf. ef. normal eq. vs. Esf. desviador eq. Presin de poros vs. Def. cortante eq.
700 Experimen tal Results Path
450
Cam-Clay Model Pa th 400
600 CSL
LSL Failur e 350
500 LSL Initial
300
q (kPa)
Tota l S tress P ath

Uu (kPa)
400 250
300 200
150
200
100
100 Experimental Results
50
Cam-Clay Model
0 0
0 100 200 300 400 500 600 700 800 900 0 0,02 0,04 0,06 0,08 0,1
p , p' (kPa) q
Figure 3. Cam-clay model predictions considering overconsolidation effect for sample with c= 602.7 kPa.
4. Experimental results in cyclic load process
During the cyclic triaxial tests, the effect of the amplitude (q= 30-60 kPa) and
frequency (0.10, 0.20 y 0.40 Hz) of the cyclic load on the mechanical behaviour of the
soil under a NC state were evaluated. Every sample was analyzed under an
approximate pressure of 200 kPa, a stress state p= 260 kPa and q= 180 kPa under a
drained stress path, in order to have an anisotropic state on the soil, before applying an
undrained cyclic load.
The cyclic stress ratio (CSR) is defined as the ratio between the static deviatoric
stress (qESP STATIC) and the cyclic deviatoric stress (q) of the cyclic load stage respect
to the static undrained shear strength (qf).
Table 3. CSRs determinate for the cyclic triaxial tests.
p ESP STATIC q ESP STATIC
q (kPa) qf (kPa) CSRl
(kPa) (kPa)
259 177 30 215 0.96
260 180 90 215 1.26
Figure 4 shows the results of the Cam-clay model of the shear strength and excess
pore pressure under static load. In this figure, the same stress state on the soil before
applying the cyclic load, and the soil failure under undrained static load were
considered.
Esf. ef. normal eq. vs. Esf. desviador eq. Presin de poros vs. Def. cortante eq.
300 90
Cam clay M odel Path
CSL 80
250 LSL INITIAL
LSL Drained Stress Path
70
LSL FAILURE 60
200
q (kPa)
Total Stress Path
Uu (kPa)
50
150
40
100 30
20 Uu (kPa)
50 Uuq (kPa)
10
Uup (kPa)
0 0
0 50 100 150 200 250 300 350 400 450 500 550 0 0,02 0,04 0,06 0,08 0,1
p , p' (kPa) q
Figure 4. Shear strength and excess pore pressure in the Cam-clay model under a drained-undrained path in
NC samples, qf = 215 kPa y ruf = 0.35.
4.1. Dynamic Response in NC soils -CSR= 0.96
Figure 5 shows the dynamic response of the soil for CSR= 0.96. The stiffness
degradation in controlled stress triaxial cyclic is expressed in terms of Degradation
Index (D) [10], where ( a)1 y ( a)N are the axial deformation for the first and the
umpteenth load cycle, respectively:

(5)

The damping ratio (D) was determined from the relationship between the loading-
unloading loop area and the area under the stress-strain curve in every charge cycle.
Due to the cyclic load amplitude, the deformations where low and there were no
noticeable changes in soil stiffness. The energy dissipation capacity by damping was
high from the beginning of the cyclic load. The excess pore pressure is lower in
comparison to the one produced under static load and in any tests the critical state was
reached.
ndice de degradacin vs. Deformacin Relacin de amortiguamiento vs.

cortante cclica residual Deformacin cortante cclica residual
1,2 1
0.10 Hz
1,0
0,8 0.20 Hz
0,8 0.40 Hz
0,6
0.10 Hz
D
D
0,6
0.20 Hz 0,4
0,4 0.40 Hz
0,2
0,2
0,0 0
0,001 0,01 0,1 1 0,001 0,01 0,1 1
q (%) q (%)
Figure 5. Dynamic response of reconstituted kaolin for CSR= 0.96.
4.2. Dynamic Response in NC soils -CSR= 1.26
The mechanical behaviour with CSR = 1.26 is illustrated in Figure 6. For the three
levels of cyclic loading frequency a state of failure is reached because deviatoric
stresses applied are greater than the shear strength of soil under static condition.
ndice de degradacin vs. Deformacin Relacin de amortiguamiento vs.

cortante cclica residual Deformacin cortante cclica residual
1,2 1
0.10 Hz
1 0.20 Hz
0,8
0,8 0.40 Hz
0,6
D
D
0,6
0.10 Hz 0,4
0,4
0.20 Hz
0,2
0,2 0.40 Hz
0 0
0,01 0,1 1 10 100 0,01 0,1 1 10 100
q (%) q (%)
Figure 6. Dynamic response of reconstituted kaolin for CSR= 1.26.
5. Pore process generation in cyclic load process
5.1. Result analysis according to CSSM
Figure 7 shows the undrained stress path used in triaxial test with CSR=0.96 - 1.26 and
f=0.10 Hz. Figure 13 also shows the simulation path of static undrained test on Cam
Clay model. The higher peak and lower peak are represented by black empty dots and
end node by solid dots, respectively. With CSR=1.26 v-p'-q states were reached
outside those defined by the LSS generating dynamic effects that influence the
mechanical behavior of the soil, analyzed below.
Esfuerzo efectivo normal equivalente vs. Esfuerzo desviador equivalente Esfuerzo efectivo normal equivalente vs. Esfuerzo desviador equivalente
350 350
ESP Static ESP Static
Cam clay Model Path Cam clay Model Path
CSL CSL
300 300
LSL INITIAL LSL INITIAL
LSL Drained Stress Path LSL Drained Stress Path
LSL Failure LSL Failure
250 250
ESP Top Cyclic ESP Top Cyclic
ESP Final Cyclic ESP Final Cyclic
ESP Bottom Cyclic ESP Bottom Cyclic
200 200
q (kPa)
q (kPa)
150 150
100 100
50 50
0 0
0 50 100 150 200 250 300 350 400 450 500 550 0 50 100 150 200 250 300 350 400 450 500 550
p (kPa) p (kPa)
Figure 7. Effective stress paths by CSR=0.96 (left), CSR=1.26 (right) y f=0.10 Hz.
When the CSR=0.96, the stress path was under LSS and elastoplastic behavior was
displayed by soil. Stiffness degradation was minimum while damping was high and
shear strains were low, in contrast to the behavior when CSR=1.26.
5.2. Mechanism of pore pressure generation in cyclic load process
Figure 8 shows the effect of the load frequency in the pore pressure generation and in
Figure 9 is presented the stress-strain-pore pressure typical behavior in the first cycles.
The speed of load application and soil response generates an effect of "pulse", causing
strain viscous character and, therefore greatest excess pore pressure.
Relacin de presin de poros vs. Relacin de presin de poros vs.

Deformacin cortante cclica residual Deformacin cortante cclica residual
0,7 0,7
0,6 CSR= 0.96 0,6 CSR=0.96
0,5 CSR=1.26 0,5 CSR=1.26

0,4 0,4
ru
ru
0,3 0,3
0,2 0,2
0,1 0,1
0 0
0 5 10 15 20 0 5 10 15 20
q (%) q (%)
Figure 8. Pore pressure generation for f=0.10 Hz (left) y f=0.40 Hz (right).
Esfuerzo cortante eq., q vs. Presin de poros, Uu vs. Deformacin

Deformacin cortante eq., q cortante eq., q
300 7,0
250 6,5
200 6,0
q (kPa)
Uu (kPa)
150 5,5
100 5,0
50 4,5
0 4,0
1,1 1,3 1,5 1,7 1,9 2,1 1,1 1,3 1,5 1,7 1,9 2,1
q (%) q (%)
Esfuerzo cortante eq., q vs. Presin de poros, Uu vs. Deformacin

Deformacin cortante eq., q cortante eq., q
300 3,5
250 3,0
200 2,5
q (kPa)
Uu (kPa)
150 2,0
100 1,5
50 1,0
0 0,5
0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1
q (%) q (%)
Figure 9. Stress Strain - Pore pressure behaviour for CSR=1.26 in the number of cycles 4 and 5 for f=0.10
Hz (up) y f=0.40 Hz (down).
If results from Figures 7, 8 and 9 are considered, the mechanism of generation of

pore pressure is defined as follows: in the first cycles the degradation of stiffness is
very high while damping is minimum. Elastoplastic and viscous strains were produced
during the process. Failure by cyclic softening is facilitated by excesses of pore
pressure, and is more notorious when frequency decreases and induces additional
viscous strains.
6. Summary and Conclusions
Experimental results have shown that pore pressure mechanism depends on initial
stress state, stiffness and resistance parameters, as well as of amplitude and application
speed of cyclic load.
Other finding was that under load with CSR>1.0 the failure by cyclic softening in
saturated clays soils can be induced. This is because cyclic stress path is beyond of
admissible elastoplastic stress states, which is associated with acceleration of soil
stiffness degradation, reduced energy dissipation by damping and it incremented excess
of pore pressure.
Results revealed that the softening cyclic mechanism on clayey soils are produced
by stiffness degradation and the reduction on the capacity of energy dissipation by
damping, when cyclic stress path is beyond of possible elastoplastic stress states.
Acknowledgements
This research was developed in the Geotechnical Laboratory of National University of

Colombia. The principal researcher acknowledge the contribution of teachers Flix
Hernndez-Rodrguez, Manuel Garca-Lpez and Juan Garcia-Leal for their
collaboration in the development of this research and to strengthen my academic
process.
References
[1] Matsui, T., Ito, T., & Ohara, H., Cyclic stressstrain history and shear characteristics of clay, Journal of
Geotechnical Engineering ASCE 106 Number 10 (1980), 11011120.
[2] Ansal, A., & Erken, A., Undrained behavior of clay under cyclic shear stresses, Journal of Geotechnical
Engineering ASCE 115 Number 7 (1989), 968-983.
[3] Azzouz, A. S., Malek, A. M., and Baligh, M. M., Cyclic behavior of clays in undrained simple shear,
Journal of Geotechnical Engineering ASCE 115 Number 5 (1989), 637-657.
[4] Dobry, R. & Vucetic, M., Effect of soil plasticity on cyclic response, Journal of Geotechnical
Engineering ASCE 117 Number 1 (1991).
[5] Vucetic, M., Cyclic threshold shear strains in soil, Journal of Geotechnical Engineering ASCE 120
(1994), 22082228.
[6] Idriss, I.M. & Boulanger, R.W., Soil Liquefaction During Earthquakes. Earthquake Engineering
Research Institute. Oakland, California, USA, 2008.
[7] Atkinson, Jhon., The Mechanics of Soils and Foundations. Taylor & Francis, 2 Park Square, Milton Park,
Abingdon. New York, USA, 2007.
[8] Atkinson, J.H. & Bransby P.L., The Mechanics of Soils, An Introduction to Critical State Soil Mechanics.
Mc Graw Hill Book Company. Great Bretain, 1978.
[9] Muir Wood, David., Soil Behaviour and Critical State Soil Mechanics. Cambridge University Press. New
York, USA, 1990.
[10] Ishihara, Kenji., Soil Behaviour in Earthquake Geotechnics, Oxford Science Publications, Clarendon
Press. Oxford, 1996.
[11] Kramer, Steven., Geotechnical Earthquake Engineering. PrenticeHall, Upper Saddle River, N.J., USA.
1996.
[12] Mitchell, J. & Soga, K., Fundamental of Soil Behavior, Jhon Wiley & Sons, New Jersey, USA, 2005.
[13] Bishop, A. & Henkel, D., The Measurement of Soil Properties in the Triaxial Test. Edward Arnold Ltd.
Londres, Inglaterra, 1962.
[14] British Standard , BS 1377 8 Shear Strength Tests (effective stress).
[15] Kuwano et al., 2001, A Review of Japanese Standards for Laboratory Shear Tests en Advanced
Laboratory Stress-Strain Testing Geomaterials, Editor Tatsuoka et al.
[16] Betancourt-Cardozo, Ferney. Diagrama de estado para una arcilla tpica de la Sabana de Bogot.
Universidad Nacional de Colombia. Bogot, 1996.
[17] Chaves-Agudelo, Julin. Generacin de presin de poros en procesos cclicos no drenados.
Universidad Nacional de Colombia. Bogot, 2011.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-947
A Rheological Model for Peat that accounts

for Creep
Djamalddine BOUMEZERANE1, Gustav GRIMSTAD2
1,2
Norwegian University of Science and Technology
Department of Civil and Transport Engineering -Geotechnical Division
Abstract: The rheological model of Gibson and Lo is revisited for peat behavior
including creep. A top linear spring is used to account for macro-compressibility
of the peat skeleton, combined with a linear spring and dashpot that simulate the
compressibility and viscous behavior of porous peat structure.
The equation of consolidation of Terzaghi for a layer of peat, drained on the
top, is combined with the rheological model. Finite difference method is used to
solve the system of equations, based on initial condition of uniform pore pressure
distribution equal to the total applied pressure on top and a variation of pore
pressure (U) equal to zero at the undrained bottom of the layer. A constant load is
applied to simulate both consolidation and creep. The results are compared with an
empirical creep equation, giving the strain as a function of log(t). The physical
mechanisms involved in the compression of peat are discussed with a focus on
creep effect. Peat provides both viscous and micro-pore mechanisms of
consolidation. Constant Load tests are performed on samples of saturated peat
(10cm diameter, 25cm height), drained on the top and undrained on the bottom.
Vertical and lateral pressures are measured as well as pore pressure. The results
are compared to the predictions of the rheological model and to the empirical
relation. Oxidation in peat due to variations of water level and temperature is taken
into account using a model proposed by Stephens et al [10]. Combining creep and
oxidation gives significant vertical deformations in time.
Keywords. Consolidation, Creep, Peat behavior, Rheological model, Subsidence

due to oxidation
1. Introduction
Peats originate as the effect of plant decomposition. The variety of materials which
compose peat influences its behavior under load. It is difficult to simulate the behavior
of organic soils using elasticity or plasticity theories since peat displays significant rate
dependent behavior.
Peats are characterized by large compressibility. The initial permeability in these
soft organic soils is large and decreases with the value of compression [4]. The
compression behavior of peat varies from the compression behavior of other types of
soils because it is much larger, and the creep portion of settlement plays a more
significant role in determining the total settlement. Decomposition of peat occurs
naturally, with plant matter and molecules broken down into finer detritus and simple
molecules (e.g. carbon dioxide and water). The process is accompanied by permanent
material changes, including progressive destruction of constituent fibers [7].
1
Corresponding author: [email protected]
948 D. Boumezerane and G. Grimstad / A Rheological Model for Peat That Accounts for Creep
In the study a rheological model and equation of consolidation are combined to

describe the behavior in 1D of a peat layer. The rheological model is based on Gibson
and Lo theory; it combines a top linear spring for macro-compressibility with a dashpot
and another spring to catch the compressibility and viscous behavior of the material.
When the water level in a peat layer varies it provokes decay of the material due to
oxidation which leads to subsidence. Subsidence due to oxidation is taken into account
using a model proposed by Stephens et al. [10].
2. Mathematical model
The equation that governs the consolidation of a saturated peat layer undergoing one
dimensional compression and drainage is given by:

. (1)

Where k is the permeability, w the weight of water, U the excess pore pressure and
the deformation. The equation was first presented by Terzaghi, it assumes that
Darcys law is valid and conservation of mass is maintained. When the total stress is
constant during time the variation of effective stress depends on the variation of pore

pressure ( ).

3. Rheological model
The rheological model is composed of a spring combined to parallel dashpot and

spring as shown in figure 1.
Figure 1. Rheological model
The peat layer is drained on top and based on an impervious layer. E1 and E2 are
the springs parameters and the viscosity parameter of the dashpot.
The equilibrium equation in the rheological model is based on the transmission of
equal effective stress between spring 1 and the combined spring 2 and dashpot. An
equation is obtained in the form:

. .
.
(2)

: is the effective stress and the deformation,
,

D. Boumezerane and G. Grimstad / A Rheological Model for Peat That Accounts for Creep 949
Combined with the equation of consolidation the behavior of a saturated peat layer
can be studied using appropriate parameters. The permeability of peat decreases with
increasing effective stress, a mean value is used. For clays Terzaghi considered that
permeability k is constant over a load increment [2].
3.1. Numerical solution
A finite difference scheme is used to resolve the combined system of equations

consolidation and rheological behavior [7]. Let (t, z) be the time and depth variables
respectively. The drainage boundary is situated at z=0, where , 0 0 . At the
,
bottom of the peat layer, z=H, the variation of U is zero ( 0 ) [5].

i t
z
j
Figure 2.The finite difference scheme
The second variation of excess pore pressure with respect to z is replaced by:
, .

(3)

And the variation of the deformation with respect to time t, is replaced by:
,

(4)

The effective stress and its variation with time are given by;

, and (5)

If the load is constant,

3.2. Koppejan equation
Koppejan (1948) used a semi empirical function to calculate settlements [3], valid for
normally consolidated soils;

. . log. log
(6)

is the effective vertical stress before loading, effective stress after loading.
is the soil layer thickness. , primary and secondary coefficients of compressibility
and is time (days).
By deriving equation (7) to obtain (8) and substituting time (t) it could be
combined with equation (1) in order to study the evolution of deformations from the
beginning of the process. The same previous scheme of finite difference method is used
here also.

log (7)

log log
(8)

. (9)

, and the variation of effective stress is given by

if the applied load is constant

4. Subsidence due to oxidation
Subsidence due to oxidation occurs when the peat layer is subjected to variations of
water level and temperature. Oxidation causes decay of organic material and thus
provokes settlements. The irreversible peat displacement was studied by different
authors [1, 6, 10]. Different rates of peat subsidence are mentioned in the literature [7,
8]. In [7] it is mentioned for example rates varying between 3.1 and 5.2 cm / year on
degraded peat land. Subsidence due to oxidation can be modeled by relating land
subsidence to water level and soil temperature [10, 11]. The model proposed by
Stephens et al. [10] is used here. The biochemical subsidence rate at temperature T and
depth of water table h is given by:
, . . !!

(10)
Where is a reaction constant, is the threshold soil temperature above which the
chemical reaction is active, and (a,b) are fitting parameters. h and T denote the annual
average depth to the water table and soil temperature at 0.10m depth, respectively.
5. Laboratory test
A Constant Load (CL) test was performed on a peat sample. A vertical compression is
applied on a cylindrical sample of saturated peat of 25cm height and 10cm diameter,
drained from top. Lateral deformations are not allowed. To avoid totally lateral friction
a special product could be used inside. The load was applied in three steps, 12,5kPa,
24kPa and 50kPa. The first step lasted 3.5 days, the second 2 days and the last step

took 11days. The mean value of =0.34 as shown in Fig.5. The aim of the test was to

follow the vertical deformations of the sample during time and measure how the earth
pressure coefficient varies with time and stress level. The results are summarized in
figures 3,4,5 .

%.VGUV

&GHQTOCVKQP

6KOG
OKP Z

Figure 3.Vertical deformation with time from CL test.

%.VGUV

*QTK\QPVCNUVTGUU5KI*
M2C

6KOG
OKP Z

Figure 4.Horizontal stress evolution with time from CL test.

%.VGUV

-Q5KI*5KI8

6KOG
OKP Z

Figure 5.Ko evolution with time from CL test. Results from Bodeux (2014) [12]
6. Applications
The model was applied on a saturated peat layer under a constant load of 50kPa,
(using: ! =280kPa, ! =550kPa, " =1,5e5kPa.s and a mean value of permeability
(affected by effective stress) is considered #=10e-7 m/s [5], the parameters are fitted
according to [12]. The parameters used for Koppejan formulation are =0.24 and
=0.015 [13]. Evolution of vertical deformations from the model, Koppejan and CL
test is shown in figure6; a constant pressure of 50kPa was applied. As one can notice
the evolution of vertical deformations predicted from the rheological model starts
slowly compared to the CL test and Koppejan approach. After certain period of time
the formulation based on Koppejan approach overestimates the evolution of creep
deformations, when rheological model gives lower creep deformations compared to CL
test. The inclination of the curve permits predicting creep deformations for long time.

%.VGUV
/QFGN
-QRRGLCP

&GHQTOCVKQP

6KOG
OKP
Figure 6. Vertical deformation - time under 50kPa (CL test, Koppejan, Model)
The normally experienced stress levels in peat are much lower than what is commonly
encountered for other types of soils (clay, sand, silt). The organic constituents of peat
and high water contentmake it a special material. Peat is very sensitive stress increase
and deformations are much larger than normally experienced in other geomaterials [4].
As can be noticed, from the performed CL test, the level of vertical deformations
reached is almost 30% under a pressure of 50 kPa. Peat is a material that exhibits
significant deformations in time, and it is therefore natural to use a rate dependent
model for modeling the behavior of peat.
The model is also applied for a case when subsidence due to oxidation happens
after a certain period of time. Parameters of peat given in [10], a=-0.15, b=0.006 and
=0.06931 is used. The temperature and water level vary periodically. A pressure of
200kPa is applied and ! =1000 kPa, ! =2000 kPa, "=6.5e5kPa.s and #=10e-7 m/s.
Figures7 and 8 illustrate the predicted vertical deformation when subsidence effect due
to oxidation is taken into account. The deformation depends on the rate of oxidation,
is varied and its influence on deformation shown in figure 8. The effect of
subsidence due to oxidation becomes important with time and can provoke big
damages as mentioned by Zanello [11].

ETGGR
ETGGR QZKFCVKQP
-U

&GHQTOCVKQP

6KOG
FC[U
Figure 7. Deformation - time when subsidence due to oxidation occurs

ETGGR
ETGGR QZKFCVKQP
-U

&GHQTOCVKQP

6KOG
FC[U
Figure 8. Deformation - time with subsidence due to oxidation (varying Ks)
7. Conclusion
The rheological model of Gibson and Lo is revisited for peat behavior including
creep with the possibility of simulating subsidence due to oxidation.The equation of
consolidation of Terzaghi for a layer of peat, drained on the top, is combined with the
rheological model. Finite difference method is used to solve the system of equations,
based on initial condition of uniform pore pressure distribution equal to the total
applied pressure on top and a variation of pore pressure (U) equal to zero at the
undrained bottom of the layer. An incremental load test is simulated including both
consolidation and creep.
An original attempt was made on Koppejan empirical equation for secondary
settlements which is revisited and combined with consolidation in the finite difference
scheme by substituting time with an equivalent time for the Koppejan equation in rate
form.
Constant Load (CL) tests were performed on samples of saturated peat (10cm
diameter, 25cm height), drained on the top. Vertical and lateral pressures are measured
with pore pressure. The predictions of the rheological model are compared to the
vertical deformations obtained from the CL test and to the deformations using
Koppejans formula. The results from the model showed a good agreement with
experimental results after certain period of time (creep deformations). The
compressibility and viscous behavior of porous peat structureare taken into account
using linear springs and dashpot. Mean values of the parameters are used and give
reasonable results.
An important aspect related to decay by oxidation in peat is considered. The model
permits taking into account subsidence in peat due to variations of water level and
temperature (oxidation) combined with consolidation and creep. A model proposed by
Stephens et al.(1984) [10] is coupled with the rheological model. Combining creep and
oxidation gives significant vertical deformations with time.
Acknowledgements
This work is carried out as part of EC/Marie-Curie IAPP project CREEP, Creep of
Geomaterials Grant number 17 (PIAP-GA-2011-286397). MSc student Sarah Bodeux is
acknowledged for providing the experiments used in this paper.
References
[1] S. van Asselen, E. Stouthamer and T.W.J. van Asch, Effects of peat compaction on delta evolution: A
review on processes, responses, measuring and modeling. Earth-Science Reviews92 (2009), 35-51.
[2] L. Barden, Consolidation of clay with non-linear viscosity. Gotechnique15 (1965), 345-362.
[3] B. Felix, Prsentation dun nouveau modle rhologique et de ses applications la thorie de la
consolidation. Bulletin de Liaison des Ponts et Chausses111 (1981), 92-103.
[4] M. Gunaratne, P. Stinnette, A.G. Mullins, C.L. Kuo and W.F. Echelberger, Jr., Compressibility Relations
for Peat and Organics Soil, Journal of Testing and Evaluation. JTEVA26 No. 1(1998), 19.
[5] J.E. Garlanger, The consolidation of soils exhibiting creep under constant effective stress.
Gotechnique22 No 1, 71-78.
[6] T. Hoogland, J.J.H. van den Akker and D.J. Brus, Modeling the subsidence of peat soils in the Dutch
coastal area. Geoderma171-172 (2012), 92-97.
[7] T. Nagano, K.Osawa, T. Ishida, K. Sakai, P. Vijarnsorn5, A. Jongskul, S. Phetsuk, S. Waijaroen, T.
Yamanoshita, M. Norisada and K. Kojima, Subsidence and soil CO2 efflux in tropical peatland in
southern Thailandunder various water table and management conditions. Mires and Peat, Volume 11
(2013), 120
[8] B.C. OKelly and S.P. Pichan, Effect of decomposition on physical properties of fibrous peat.
Environmental Geotechnics, Issue EG1, volume 1 (2013), 22-32.
[9] G.D. Smith, Numerical Solution of Partial Differential Equations: Finite Difference Methods. Oxford
University Press, 1984.
[10] J.C. Stephens, L. H. Allen Jr. and E. Chen, Organic soil subsidence. Reviews in Engineering Geology 6
(1984), 107-122.
[11] F. Zanello, Contributions to modeling of Venice coastal peatlands and wetlands, PhD thesis, University
of Padova Italy, 2011.
[12] S. Bodeux, MSc Thesis Project , NTNU, 2014.
[13] Norm NEN 6740: 2006 .Geotechniek - TGB 1990 - Basiseisen en belastingen.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-955
Effects of Initial Conditions on Strength of

Acrylamide Grouted Sands
L. Sebastian BRYSON a, 1 and Ryan ORTIZ a
a
Department of Civil Engineering, University of Kentucky, Lexington, KY 40506 USA
Abstract. Chemical grouts have long been used as soil stabilizers to reduce
groundwater infiltration into underground structures and to inhibit seepage through
earth dams and other impoundments. Among the chemical grouts, acrylamide
grouts have shown increased usage within the last decade. Although these grouts
are intended for water control rather than for strength, mechanical properties must
be considered if there is a potential for movement of the grouted soil mass, such as
deformations in the form of ground subsidence. In this study, the effects of initial
soil conditions, such as moisture content and unit weight, on the strength of
acrylamide grouted sands were investigated. This study found that the compressive
strength of grouted sand was significantly influenced by the initial moisture
content and dry unit weight of the sand material. The results show that linear
relationships can be directly developed between strength, moisture content, and
void ratio parameters.
Keywords. Chemical Grout, Acrylamide Grout, Grouted Sand, Shear Strength,

Injection Pressure
1. Introduction
Chemical grouts have long been used as soil stabilizers to reduce groundwater
infiltration into underground structures and to inhibit seepage through earth dams and
other impoundments. Several researchers [1] [2] [3] have shown that chemical grouts
are particularly effective in soils not easily penetrated by cement-based grouts such as
fine sand and sandy silts. These chemical grouts have low initial viscosities (1-2 cP at
21 deg C) and can penetrate into formations with hydraulic conductivities as low as 10-
4
cm/sec [4]. Among the chemical grouts, acrylamide grouts have shown increased
usage within the last decade. Acrylamide grouts have been used successfully for
sealing leaking sewer and water lines [5], stopping water infiltration into very large
tunnel structures [6], and controlling brine inflows into underground mines [7].
Acrylamide grout consists of an acrylamide monomer and a catalyst that are mixed
to form a polymer solution. When acrylamide polymerizes, it solidifies into a flexible
grout that is relatively resistant to water and chemical degradation. These types of
grouts maintain constant viscosity from the time that they a mixed until just before
gelation. Also, by varying the proportions of the monomer and catalyst, the set time of
the grout can be varied from seconds to hours. Acrylamide is stable and non-reversible
in cured form. Although these grouts are intended for water control rather than for
1
Corresponding author, Department of Civil Engineering, 161 Raymond Bldg., University of Kentucky,
Lexington, KY 40506 USA, email: [email protected].
956 L.S. Bryson and R. Ortiz / Effects of Initial Conditions on Strength
strength, mechanical properties must be considered if there is a potential for movement

of the grouted soil mass, such as deformations in the form of ground subsidence.
Very few studies of the mechanical properties of acrylamide grouted soils are
found in the literature. Previous mechanical property studies on acrylamide grouted
sand include an investigation of the influence of fines content and curing conditions on
the strength of acrylamide grouted sand [2]; an investigation of the shrinkage
characteristics of acrylamide grouts [8]; an evaluation of the effect of cationic
surfactant on the pull-out strength of grouted sand [9]; and an evaluation of the long-
term stability of acrylamide grouts [7]. Unfortunately, there is an absence of published
data showing the effects of variations in initial dry unit weight and moisture content of
the sand matrix on the strength of the acrylamide grouted sand.
This paper presents the results of a laboratory study that investigated effects of
initial soil conditions, such as moisture content and unit weight, on the strength of
acrylamide grouted sands. The intent of the initial moisture and unit weight
investigation was to provide guidance for use of this material under conditions likely to
be encountered in the field.
2. Testing Materials
2.1. Acrylamide Grout
For this study, acrylamide monomer and catalyst materials were provided by
Avanti Grout International of Webster, TX. The specific materials used were the
Avanti AV-100 acrylamide chemical monomer, the AV-101 Catalyst T+ activator
triethanolamine and the Avanti AV-102 catalyst ammonium persulfate. These three
chemicals are the minimum necessary for the polymerization to occur. The AV-100
actually consists of two separate organic monomers; acrylamide monomer, which will
polymerize into long molecular chains, and methylene-bis-acrylamide, which is a
cross-linking agent that binds the molecular chains together. The acrylamide monomer
and the cross-linking agent were mixed at a 95:5 ratio, respectively. The AV-101
activator acts as a buffer and sometimes appears to act as a catalyst in the acrylamide
grout solution. Herein, the mixture of the two organic monomers is referred to simply
as monomer. The catalyst, activator, and monomer were supplied by Avanti as two
separate pre-mixed liquids labeled AV-100 and AV-102. The container for the AV-100
solution consisted of a 10 percent concentration solution by weight of the AV-100 and
a one percent concentration solution by weight of AV-101. The second pre-mixed
liquid consisted of a one percent concentration solution of the AV-102 by weight. In
this context, percent concentration is defined as parts of chemical component per 100
parts of total grout solution. It is noted that the pre-mixed AV-100 consisted of
approximately 54 percent water by volume, whereas the pre-mixed AV-102 consisted
of roughly 98 percent water by volume and 4 percent catalyst. The density of the
monomer solution at room temperature (21 deg C) is 1.04 g/cm3. The viscosity of the
monomer at room temperature is 1.2 cP, as compared to the viscosity of water of 1 cP.
In this study, the mixture of the monomer and catalyst is referred to as acrylamide gel
and synonymous with acrylamide grout. The mass formed by mixing the acrylamide
grout with sand is referred to grouted sand herein.
L.S. Bryson and R. Ortiz / Effects of Initial Conditions on Strength 957
2.2. Gel Time
The process through which acrylamide grout gains its strength is called free radical
polymerization. This is a rapid, exothermic reaction. Two interrelated variables for
evaluating grout strength are gel time and gel temperature [1]. Gel time is the time it
takes for the acrylamide grout to change from a low viscosity liquid to a non-viscous
gel. For this study, gel time is defined as the time beginning at the point at which the
monomer comes into contact with the catalyst and ending at the point when the grout
will no longer flow when tilted 45 degrees. Gel time was determined when no excess
flow existed. The gel temperature is sometimes referred to as curing temperature and is
the excess heat that results from the exothermic reaction during the state change from
liquid to semi-solid. As was mentioned previously, the grout used for this study
consisted of a 10 percent concentration AV-100 solution, a 1 percent concentration
AV-102 solution, and a 1 percent concentration AV-102 solution. The measured gel
time at room temperature (approximately 70 deg F) was approximately 47.6 seconds.
3. Test Sand
The primary thrust for this research effort was to investigate influence of the initial
density and moisture conditions on the mechanical behavior of a grouted sand mass.
The test sand used for this study was classified as poorly graded sand (SP), as
determined by the unified soil classification system (USCS). Table 1 presents the index
test data for the test sand.
Table 1. Test sand properties
Gs Cu Cc % Pass #10 % Pass #40 % Pass #200
Sieve Sieve Sieve
2.65 2.88 0.75 80.77 24.48 0.44
In the table, G s specific gravity; C u coefficient of uniformity = D60 / D10 ;

C c coefficient of curvature = D30 2 ( D10 - D60 ) ; D10 , D30 , and D60 are the
diameters associated with 10 percent, 30 percent, and 60 percent passing, respectively.
4. Grouting Apparatus
Field chemical grouting equipment often will include a two-tank system to mix
and inject the grout; Tank A for mixing the monomer and Tank B for mixing the
catalyst. It was the intent of this study to develop a laboratory-scale model grouting
apparatus that resembled typical grouting equipment encountered in the field. Figure 1
shows a schematic of the model grouting apparatus.
Pressure
Flexible
Panel
Tubing
Brass Threaded Rod

Acrylamide Fitting Effluent
Monomer PVC Container
Clamping Hex Brass Wire Mesh

Nut Fitting Filter
Catalyst
Acrylic Sand and

Chamber Grout mix
Ball Valve
Tee-Fitting Rubber
O-Ring
Flexible
Tubing
Metal Stand
Figure 1. Sand grouting apparatus.

The apparatus consisted of two PVC holding containers for the monomer and
catalyst solutions. Each container was approximately 61 mm in diameter and
approximately 102 mm long. Each container held approximately 300 mL of solution.
Pressure was supplied to the containers through two plastic tubes connected to the tops
of each container, which were connected to a commercially-available pressure panel.
Ball valves were placed at the bottom of each container to allow the outflow of solution
to be roughly regulated. The two solutions were mixed in a Tee-fitting at the bottom of
the sample cell. The sample cell consisted of an acrylic cylinder, approximately 102
mm in length and 51 mm in diameter.
Samples were placed in the sample cell at targeted unit weights and moisture
contents. For loose samples, sand was simply poured into the cell with no rodding or
tamping. For dense samples, sand was placed in very small lifts and tamped to the
desire density. Filter paper and a very fine wire mesh was placed at the end of the
samples to keep the sand from clogging the inflow and outflow ports. During testing,
the grouting was considered successful when the grout passed completely through the
sample and gel was observed in the effluent tube.
5. Mechanical Behavior of Grouted Sand
5.1. Unconfined Compression Tests
Shear strength tests were performed on the grouted sand using unconfined
compression testing. While it is intuitive that the compressive strength of grouted sand
will increase with increasing density of the grouted sand mass, the intent of these series
of tests was to quantify the magnitude of the strength increase, with respect to
variations in moisture content.
The grout sand samples were prepared by placing the test sand at targeted dry unit
weights and targeted moisture contents. The targeted dry unit weights used for this
study were 16.2 kN/m3, 16.8 kN/m3, and 17.4 kN/m3. The gravimetric moisture
contents used were 0 percent, 4 percent, 6 percent and 8 percent. The samples were
grouted in the acrylic cylinders, using procedures described earlier. Commercially-
available baking paper was taped to the inside of the acrylic cylinders to facilitate
removal of the grouted sand. After the samples were fully permeated with grout, the
samples were allowed to gel in the test cylinders for approximately 1 hour. The test
samples were then pushed out of the test cylinders an allowed to air cure for 7 days.
The samples were tested in unconfined compression at a rate of 0.254 cm/min. The
testing machine was allowed to run until failure or until a significant decrease in force.
Failure typically included vertical tension cracking at the top or bottom of the grouted
sand sample and the formation of a diagonal failure surface starting at the bottom of the
sample.
It should be noted that minimal data were collected for the dense (i.e.
dry 17.44 kN m 3 ) sample. Unfortunately, it was observed that the penetrability
of the grout into the sand was extremely problematic at the high unit weight. As a result
of this difficulty, only one viable sample was produced.
5.2. Compressive Strength
The effect of grain size parameters on the strength of grouted sands has been
investigated in several studies [10] [11]. Particle size distribution, unit weight, and
grain shape of soil have been reported to be the most important soil parameters
affecting the silicate grouted sand strength [12]. Studies of a few researchers [10] [13]
Christopher et al. [11] showed that the strength of grouted sand increased with unit
weight and decreased with effective grain size (d10) or mean grain size (d50) of soil, and
well graded soils had higher strengths than uniform soils of the same effective grain
size [13].
This current study evaluated both the influences of grain size parameters and the
initial state on compressive strength. Figure 2 presents the unconfined compressive
strengths (i.e. the peak stress values) for sand with zero moisture content. The data was
supplemented by including the data presented by Ozgurel and Vipulanandan [2].
1000
qu(gel) = 7.85 kN/m2
100 y = e1.8082x
R = 0.9719
qu0/qu(gel)
10
1
This Study
Ozgurel and Vipulanandan (2005)
0.1
0 0.5 1 1.5 2 2.5 3
Cc (d/w)
Figure 2. Effects of initial dry unit weight on compressive strength.
In Figure 2, the influence of the grain size and the initial state is described with a
term that combines the coefficient of curvature, C c , with the normalized dry unit
weight, d w . The coefficient of curvature gives an indication of the distribution of
particles sizes in a sample. In particular, a C c value between 1 and 3 implies a well
graded soil and a well graded soil is able to be compacted more densely than a poorly
graded soil. Consequently, the strength of the grouted sand is shown to increase as the
initial dry unit weight increases and with an increasing distribution of the particle sizes.
It is recognized that the compressive strength of the grouted sand will be a function
of the compressive strength of the grout material. The strength of the grout material
will most likely depend on the concentrations of monomer, catalyst, and activator in the
grout solution. Thus, the unconfined compressive strength of the grouted dry sand, q u 0
is normalized by the unconfined compressive strength of the grout solution, q u ( gel ) . It
is noted that the grout solution was cured in the same manner as the grouted sand
samples.
From the results presented in Figure 2, an expression can be developed to predict
the unconfined compressive strength of grouted sand, given, the initial dry unit weight
of the sand, grain size distribution information and the compressive strength of the
grout solution. The expression is given as
# $
qu 0 qu gel
exp 1.8082C c %% d && (1)
' w (
The effect of initial moisture content was also investigated for this study. Figure 3
shows the unconfined compressive strength as a function of the initial volumetric
moisture content of the sand. The volumetric moisture content is defined as
dry
(2)
w
where = volumetric moisture content; = the gravimetric moisture content;
w = unit weight of water. The volumetric moisture content allows both the
gravimetric moisture content and the dry unit weight to be represented in one term.
2.5
2 y = 0.0691x + 1
R = 0.9021
1.5
qu/qu0
0.5
Grouted Sand
0
0 5 10 15
(%)
Figure 3. Effects of initial volumetric moisture content.
In Figure 3, the unconfined compressive strength of the grouted moist sand is
normalized with unconfined compressive strength of the grouted dry sand, q u 0 . In
general, the results show a linear relationship between initial volumetric moisture
content and strength. The relation is given by
D
qu qu 0 1 0.0691
E (3)
Increases in volumetric moisture content resulted in an increase of the compressive
strength. The increase in strength for the moist sand is most likely due to capillarity. In
this context, capillarity pertains to the surface tension that develops at the air-water
interface resulting in a negative pore water pressure. This negative pore water pressure
serves to increase the shear of soils at low moisture contents, as is shown in Figure 3.
Although it is acknowledged that Equation 3 was developed from one soil type. It is
reasonable to assume that for a soil similar to the one tested in this study; Equation 1
can be combined with Equation 3 to produce one equation capable of predicting the
unconfined compressive strength of a grouted soil mass, which considers varying initial
dry unit weights, grain size distributions, and volumetric moisture contents. However,
additional research to confirm the reasonableness of this assumption is warranted.
5.3. Elastic Modulus
The deformation behavior of the grouted sand was evaluated using the tangent
modulus, E t . The tangent modulus is the parameter that represents the slope of the
stress-strain curve from 0 to 2 percent strain. The tangent modulus parameter is
typically used to predict non-volume change related deformations of a soil mass. Figure
4 shows the relationship between volumetric moisture content of the samples and
normalized tangent modulus, E t q u 0 .
25
Grouted Sand
20
15
Et/qu0
10
y = -1.1774x + 20.169
R = 0.982
5
0
0 5 10 15
(%)
Figure 4. Effects of moisture content on normalized modulus.
Figure 4 shows that as the volumetric moisture content increases, the Et qu 0

ratio decreases. The implication of this observation is that increasing the moisture
content produces more ductile behavior in the grouted sand sample. In terms of
compressive loading, the grouted moist sand would most likely experience more
settlement that the grouted dry sand for the same load. However, the grouted moist
sand would have a higher bearing capacity than the grouted dry sand. The relationship
between the normalized tangent modulus and the volumetric moisture content is given
as
Et
20.169 1.1774
(4)
qu 0
As with Equation 3, although Equation 4 was developed for the sand and testing
conditions used in this study, it is assumed that Equation 4 can be combined with
Equation 1 to expand its range to other sand types. However, additional testing is
warranted.
6. Conclusion
This study investigated effects of initial soil conditions, such as moisture content
and unit weight, on the strength of acrylamide grouted sands. The conclusions drawn
from this study are as follows:
The compressive strength of grouted sand specimens increase linearly with
increasing initial volumetric moisture content and gradation characteristics associated
with well graded soils.
For low moisture content conditions, the normalized tangent modulus decreases
with increasing volumetric moisture content indicating that the grouted sand samples
become more ductile under compressive loads.
7. Acknowledgements
This research was funded in part by the University of Kentucky Appalachian and
Minority Science, Technology, Engineering, and Mathematics Majors (AMSTEMM)
program under NSF Award # 0431552. This support was greatly appreciated. Grout
materials for this research were provided by Mr. Jim Gentry of Avanti International.
References
[1] Anagnostopoulos, C., and Hadjispyrou, S. (2004). Laboratory Study of an Epoxy Resin Grouted Sand.
Ground Improvement, 8(1), 39-45.
[2] Ozgurel, H. G. and Vipulanandan, C. (2005). Effect of Grain Size and Distribution on Permeability and
Mechanical Behavior of Acrylamide Grouted Sand. Journal of Geotechnical and Geoenvironmental
Engineering, ASCE, 131 (12), 1457-1465.
[3] Lirer, S., Flora, A., Evangelista, A., Verdolotti, L., Lavorgna, M., Iannace, S. (2006). Permeation
Grouting of a Fine-Grained Pyroclastic Soil, Ground Improvement, 10 (4), 135-145.
[4] USACE (1995). Engineering and Design Chemical Grouting EM 1110-1-3500, US Army Corps of
Engineers, Washington, DC.
[5] Ozgurel, H. G., Gonzalez, H.A., and Vipulanandan, C. (2005). Two Dimensional Model Study on
Infiltration Control at a Lateral Pipe Joint Using Acrylamide Grout. Proceedings of the Pipeline
Division Specialty Conference, 58, 1218-1244.
[6] Narduzzo, L. (2003). The Toronto Transit Commissions Subway Tunnel and Station Leak Remediation
Grouting Program. Grouting and Gound Treatment, 120 (2), 1617-627.
[7] Haug, M. D., Al-Manaseer, A.A., and Coode, A.M. (1998). Impact of Confining Pressure on Long-
Term Performance of Chemical Grout in Salt Water Journal of Materials in Civil Engineering, ASCE,
10(2), 70-75.
[8] Jasti, V., Vipulanandan, C., Magill, D. and Mack, D. (1996), Shrinkage Control in Acrylamide Grouts
and Grouted Sands, Proceedings, Materials for the New Millennium, ASCE, Washington D.C.,
pp.840-850.
[9] Vipulanandan, C., and Sunder, S. (2011).Effect of a Cationic Surfactant on the behavior of Acrylamide
Grout and Grouted Sand. Proceeding of Geo-Frontiers 2011, Dallas, TX, 13-16 March 2011, 687-696.
[10] Schiffman, R.L., Wilson, C.R., (1958). The Mechanical Behavior of Chemically Treated Grouted
Soils. Conference for Grout and Drilling Method in Engineering Practice, Butterwoths, London, 29-
35.
[11] Christopher, B. R., Atmatzidis, D. K., and Krizek, R. J. (1989). Laboratory testing of chemically
grouted sand. Geotechnical Testing Journal, 12(2), 109-118.
[12] Ata, A. and Vipulanandan, C. (1999). Factors Affecting Mechanical and Creep Properties of Silicate-
Grouted Sands. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 125(10), 868-
876.
[13] Clough, G. W., Kasali, G., and Kuck, W. (1979). Silicate-stabilized sands. Journal of the Soil
Mechanics and Foundations Division, ASCE, 105(1), 65-82.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-963
Stress-strain behavior of a cement-based

stabilized soft soil
Antnio A.S. CORREIAa, Paulo J. VENDA OLIVEIRAa and Lus J.L. LEMOSa,1
a
Department of Civil Engineering University of Coimbra, Portugal
Abstract. This paper presents the results of a laboratory study carried out on
Coimbra soft soil specimens in order to evaluate the effect of the artificial
cementation on the mechanical properties. A series of triaxial tests following a
compression path with undrained conditions were conducted on natural remolded
specimens and on laboratory prepared cemented-based specimens. Based on the
experimental results, the effect of the cementation on the stress-strain behavior and
strength property are discussed. For lower consolidation stress states, the behavior
of the cement-based stabilized soft soil under triaxial shear is characterized by an
initial elastic section, limited by a structural yield point (defined as the point from
where the plastic deformations significantly increase, i.e., the start point of the
breakage of the cementitious bonds), followed by a brittle failure (characterized by
the sudden breakage of the cementitious bonds with the consequent decrease of the
shear stress).
Keywords. Cemented soils, Stress Strain behavior, Soft soils
1. Introduction
Over the last few decades, the growth of urban perimeter in conjunction of land
occupation policies has demanded construction on soils with poor geotechnical
properties (in particular, soft soils). These soils are usually characterized by low
strength and high compressibility, demanding from geotechnical engineers new and
challenging solutions to overcome these undesirable engineering characteristics. One of
the ground improvement techniques that have been used with success in practice is the
chemical stabilization, where the natural soil is mechanically mixed in situ with binders.
As a result of this mixture there are physico-chemical interactions that occur between
soil particles, the binders and water present in the soil, resulting in a new composite
material with a better mechanical behavior than the original one. This stabilizing effect
is a consequence of cementitious bonds between soil particles which promote the
formation of a new stronger and stiffer matrix.
At the actual stage of knowledge, given the complexity of the physico-chemical
interactions and the large number of parameters having influence on the soil-binder-
water mixture, it is not possible to predict the mechanical behavior of stabilized soil
without a laboratory and field trial test program, as stated in the European standard EN
14679 [1]. In this paper it is presented a series of laboratory triaxial tests with the goal
of study the effect of the artificial cementation on the mechanical properties of the
Coimbra stabilized soft soil. On natural remolded specimens and on laboratory
1
964 A.A.S. Correia et al. / Stress-Strain Behavior of a Cement-Based Stabilized Soft Soil
prepared cemented-based specimens, triaxial tests following a compression path with

undrained conditions were done for different confining stress levels. The material
properties and the main results are presented in the next sections.
2. Materials and experimental procedure
2.1. Materials
The present work is based on a Portuguese soft soil, taken from a location in the center
of Portugal near the city of Coimbra (region known as Baixo Mondego). The soft soil
deposit is geologically very recent, exhibiting a normally consolidated or lightly
overconsolidated state, except for the shallow layers ( 3.5m to 6.5m) where the OCR
could reach values of 6.0 due to desiccation amplified by seasonal changes in the water
table. At the sampling site, the soft soil deposit has a thickness of 23m, presenting a
more or less uniform grain size distribution, with silt as the dominant fraction. The high
organic matter content (9.3%) found has a strong influence on some characteristics of
the soil, namely, low bulk unit weight, high plasticity, high natural water content, high
void ratio, low strength and high compressibility behavior [2, 3, 4]. A big
representative sample (1.5m3) was collected at the depth of 2.5m exhibiting the
physical and chemical properties listed in Tables 1 and 2, respectively. The Coimbra
soft soil is a clayey-silt organic soil with a high plasticity, classified by USCS as OH
[5].
Table 1. Geotechnical properties of the Coimbra soft soil

wnat sat Grain size Plasticity OM cu
G enat Cc
(%) (kN/m3) (%) (%) (%) (kPa)
clay = 15-30 wL = 71
2.55 80.87 14.6 2.10 silt = 40-72 wP = 43 7.7 < 25 0.57
sand = 35 IP = 28
Table 2. Chemical properties of the Coimbra soft soil

CaO SiO2 Al2O3 Fe2O3 MgO K2O
(%) (%) (%) (%) (%) (%)
0.74 62.00 16.00 4.80 1.10 3.00
The binders used in the present study to produce stabilized Coimbra soft soil
specimens were a Type I Portland cement, designated CEM I 42.5 R [6], and a blast
furnace granulated slag, here simply designated as SLAG. These two binders, on a dry
weight proportion of 75% cement and 25% slag as proposed by Correia [4], were
thoroughly mixed to obtain a uniform binder. The binder was added to the soil in a
quantity equal to 175 kilos per cubic meter of soil. The composition and the specific
surface of the binders are presented in Table 3.
Table 3. Chemical composition and specific surface of the binders.

CaO SiO2 Al2O3 Fe2O3 MgO S
(%) (%) (%) (%) (%) (m2/kg)
CEM I 42.5R 63.02 19.70 5.23 2.99 2.38 321.50
SLAG 37.02 38.74 11.59 0.85 6.75 363.00
A.A.S. Correia et al. / Stress-Strain Behavior of a Cement-Based Stabilized Soft Soil 965
2.2. Experimental procedure
To attenuate the inherent variability of the natural soil, remolded samples were used in
this study, maintaining the mineralogical, organic and granulometric composition of
the natural soil. To prepare the natural remolded specimens, the natural soil was
previously passed through a 2.36 mm mesh sieve to remove shell residues; after that, it
was mixed with enough distilled water to obtain a 1.5 liquidity index. Then soil was
placed inside a big oedometer cell (230mm in diameter and 315mm in height) and was
submitted to a vertical consolidation pressure of 48kPa. Once the primary consolidation
was finished, the sample was removed and triaxial specimens were prepared.
The laboratorial procedure to produce stabilized specimens followed the
laboratory procedure presented in EuroSoilStab [7] with the modifications proposed by
Correia [4] described in the following steps: (i) the destructured soft soil and the binder
were thoroughly mixed using a mechanical mixer at a speed of 142 rpm for 3 min; (ii)
the paste was introduced and compacted directly into the PVC molds (38mm in
diameter) in 6 layers; (iii) each layer was lightly tapped by hand, subjected to
vibrations by the application of a hand drill with a steel bar near the inner surface of the
mold and compacted with 100kPa for 10 seconds; (iv) the surface was lightly scarified
and another layer was introduced; (v) the specimens were cured for 28 days under a
vertical pressure of 24 kPa (simulating the field conditions at a depth of 5 m) and
stayed submerged in a water tank at a controlled temperature (202C); (vi) after the
curing period, the specimens were removed from the mold and triaxial specimens were
prepared.
Triaxial tests were done on remolded and stabilized specimens with 76 mm
height and 38 mm diameter, placed in a stress-path cell. The electronic devices (load
cell, strain gauge and pressure transducers) were set up and adjusted; during the tests
the data were recorded by automatic data acquisition. All tests followed a compression
path with undrained conditions with different confining stress levels (50, 150 and
250kPa). The undrained shear was carried out under a constant strain rate of 1%/hour.
The results of the triaxial tests are presented and discussed in the next section.
3. Results
Figure 1 presents the results of the triaxial tests made on natural remolded specimens,
consolidated isotropically (CI) for confining stress levels of 100 kPa and 300 kPa and
following a compression path with undrained conditions (U). The results are typical for
normally consolidated clay specimens: there are large strains at failure (greater than
10%) combined with a relatively large positive excess pore pressure generation due to
undrained condition.
The same plots are presented in Figure 2 for the cement-based stabilized
specimens. It can be seen that the deviator stress (q) increases almost linearly to a peak
value (at small strains) and then reduces to a lower value of q. As the effective
confining pressure increases, the peak deviator stress tends to decrease slightly. It is
possible that the initial confining pressure in the consolidation stage breaks some of the
cemented bonds, affecting the microstructure. The peak deviator stress is almost the
same for zero confined effective stress, with the state of stress on the line q=3.p, (pore
water pressure equals the cell pressure). Although the q- a relations are essentially the
same, the development of excess pore pressure is different. This depends upon the level
of effective confining pressure: the excess pore pressure increases with increasing
effective confining pressure so as to equalize the initial confining pressure. After the
peak failure, there is a sudden structure collapse and the specimens show a tendency to
develop positive excess pore pressures.
250
200 natural soft soil:

natU (100)
natU (100)
150
natU (300)
q (kPa)
100
50
a (%)
0
2 4 6 8 10 12 14 16 18
50
100
u (kPa)
150
200
250
Figure 1. Deviator stress-axial strain and excess pore pressure-strain plots for CIU tests (natural remolded
specimens).
3000
2500
2000
q (kPa)
1500
cement-based stabilized soft soil:
1000 U(50)
U(150)
500 U(250)
qmax , umax a (%)
0
-400 2 4 6 8 10 12 14 16 18
-300
-200
u (kPa)
-100
100
200
Figure 2. Deviator stress-axial strain and excess pore pressure-strain plots for CIU tests (cement-based
stabilized specimens).
Figure 3 presents the undrained paths for CIU triaxial tests, both for natural
remolded and cement-based stabilized specimens. The effect of cementation is clear
with higher deviator stress levels for the stabilized specimens. The peak failure is
characterized by a brittle failure with sudden breakage of cementitious bonds [9, 10, 12,
13, 14, 15] followed by a decrease of the deviator stress and an increase in pore water
pressures. After peak the undrained path starts to be friction dependent, evolving to the
critical state line of the remolded natural cemented soil.
3000
U(50)
U(150)
2500 U(250)
qmax
2000
q (kPa)
p'

1500
3
=
q

7,0
'
=3
1000
,51
=1
oil
ls
ura
500 M nat natural soft soil:
natU (100)
natU (100)
natU (300)
0
0 200 400 600 800 1000 1200
p' (kPa)
Figure 3. Undrained stress-path for triaxial CIU tests (natural remolded and cement-based stabilized
specimens).
The breakage of cementitious bonds starts before the peak failure. This fact is
illustrated in the plot undrained tangential stiffness (Eu tan) - axial strain, plotted to a
log-log scale (Figure 4). The major drop in the Eu tan is defined as the structural yield
point, associated with the breakage beginning of cementitious bonds. The structural
yield point is also represented in the axial strain-mean effective stress plot, Figure 5.
Independently of the effective confining pressure, the undrained compression shear is
characterized by initial elastic section. The initial effective stress path in the p-q moves
vertically showing an elastic response keeping p constant, to ensure zero volumetric
strain during undrained shear. The elastic behavior is limited by a structural yield point
(defined as the point from where the plastic deformations significantly increase, i.e., the
start point of the breakage of the cementitious bonds), which fits very well with the one
evaluated from Eu tan- a plot.
structural yield
1000
Eu tan (MPa)
100
10
U(50) U(150) U(250)

1
0.01 0.1 1 0.01 0.1 1 10 0.01 0.1 1 10
a (%) a (%) a (%)
Figure 4. Undrained tangential stiffness-axial strain plots for triaxial CIU tests (cement-based stabilized
specimens).
10
U(50)
8
U(150)
U(250)
6 structural yield
a (%)
0
0 200 400 600 800 1000 1200
p' (kPa)
Figure 5. Axial strain-mean effective pressure plots for triaxial CIU tests (cement-based stabilized
specimens).
4. Conclusions
The experimental results, the effect of the cementation on the stress-strain behavior and
strength property are discussed. For lower consolidation stress states, the behavior of
the cement-based stabilized soft soil under triaxial shear is characterized by an initial
section elastic, limited by a structural yield point (defined as the point from where the
plastic deformations significantly increase, i.e., the start point of the breakage of the
cementitious bonds), followed by a brittle failure characterized by the sudden breakage
of the cementitious bonds with a structure collapse and the consequent decrease of the
shear stress and buildup of positive pore water pressure. The behavior pre-failure is
governed by the cementitious bonds while the post-failure behavior is essentially
dependent on friction characteristics.
Acknowledgement
The authors would like to express their thanks to CIMPOR for supplying the binder
used and to the entities that support financially the re-search: MAIS CENTRO InovC
Ignition Grant/2014, FCT through the project PTDC/ECM/101875/2008, CIEPQPF

and ACIV.
References
[1] EN 14679 (2005). Execution of special geotechnical works Deep mixing. CEN, English version,
April of 2005, p. 52.
[2] Coelho, P.A.L.F., 2000. Geotechnical characterization of soft soils. Study of the experimental site of
Quinta do Foja, MSc Dissertation, University of Coimbra (in portuguese).
[3] Venda Oliveira, P.J., Lemos, L.J.L., and Coelho, P.A.L.P., 2010. Behavior of an atypical embankment on
soft soil: field observations and numerical simulation. Journal of Geotechnical and Geoenvironmental
Engineering, 136(1), 35-47.
[4] Correia, A.A.S., 2011. Applicability of deep mixing technique to the soft soil of Baixo Mondego, Ph.D.
Dissertation, University of Coimbra (in Portuguese).
[5] ASTM D 2487, 1998. Standard classification of soils for engineering purposes (Unified Soil
Classification System). American Society for Testing and Materials.
[6] EN 197-1, 2000. Cement - Part 1: Composition, specifications and conformity criteria for common
cements. European Committee for Standardization, June 2010.
[7] Eurosoilstab, 2001. Development of design and construction methods to stabilise soft organic soils.
Design guide soft soil stabilization. CT97-0351, EC Project No. BE 96-3177, Industrial & Materials
Technologies Programme (BriteEuRam III), European Commission.
[8] Coop, M.R. and Atkinson, J.H. (1993). The mechanics of cemented carbonate sands. Gotechnique,
Vol. 43, No. 1, p. 53-67.
[9] Uddin, K.; Balasubramaniam, A.S. and Bergado, D.T. (1997). Engineering behavior of cement-treated
Bangkok soft clay. Geotechnical Engineering Journal, Southeast Asian Geotechnical Society, Vol. 28,
No. 1, p. 89-119.
[10] Cuccovillo, T. and Coop, M.R. (1999). On the mechanics of structured sands. Gotechnique, Vol. 49,
No. 6, p. 741-760.
[11] Malandraki, V. and Toll, D.G. (2000). Drained probing triaxial tests on a weakly bonded artificial
soil. Gotechnique, Vol. 50, No. 2, p. 141-151.
[12] Horpibulsuk, S. (2001). Analysis and assessment of engineering behavior of cement stabilized clays.
PhD Dissertation, Saga University, Saga, Japan.
[13] Lorenzo, G.A. and Bergado, D.T. (2006). Fundamental characteristics of cement-admixed clay in deep
mixing. Journal of Materials in Civil Engineering, ASCE, Vol. 18, No. 2, p. 161-174.
[14] Leroueil, S. and Vaughan, P.R. (1990). The general and congruent effects of structure in natural soils
and weak rocks. Gotechnique, Vol. 41, No. 3, p. 467488.
[15] Horpibulsuk, S.; Miura, N. and Nagaraj, T.S. (2005). Claywater/cement ratio identity for cement
admixed soft clays. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 131, No.
2, p. 187.192.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-970
Effect of treatment on the microstructural

characteristics of bio-improved sand
DIMITRIOS TERZIS a,1 and LYESSE LALOUI a
a
Laboratory for Soil Mechanics, Swiss Federal Institute of Technology (EPFL),
Lausanne, Switzerland
Abstract. Bio-chemical processes have recently become the core of several studies
in the geoengineering field, where the emphasis is placed on new and emerging
applications. Microbially Induced Calcite Precipitation has been introduced as an
alternative cementation mechanism for soils, and it is studied for its potential to
offer substantial and environment-friendly solutions for soil improvement. In this
paper, we focus on the microstructural characteristics of the new geo-material,
which is engineered after harnessing a treatment process including multiple
injections of bacteria and chemical solutions, which ultimately induce the
formulation of calcite particles inside the soil matrix. Obtaining detailed
knowledge of the local structural properties of the precipitated nuclei is considered
an essential factor because calcite particles affect the macroscopic mechanical
response with the foreseen increase in strength and stiffness, which are attributed
to the precipitated mass and its geometrical arrangement. The main objective of
this study is to elucidate the role of the calcium chloride concentration on the
yielded precipitation and clarify some of the predominant fabric characteristics of
the new structure. For this purpose, microstructural observations with Scanning
Electron Microscopy (SEM) were used. Here, calcite mesocrystals, i.e., aggregates
of single particles, are identified as a distinct form of precipitates that provide the
crucial grain-to-grain contact surfaces. The cemented samples were subsequently
subjected to undrained triaxial shear. The results show the increase in peak
strength at low strains, followed by a brittle fracture and post-yield softening.
Keywords. Soil improvement, strength, stiffness, Microbially Induced Calcite

Precipitation (MICP), microstructure
1. Introduction
Biologically induced calcite mineralization has been introduced [1],[2],[3] to provide

an alternative cementation mechanism for granular geomaterials. Although this
phenomenon was known to scientists from natural environments, where bacteria are
active in the subsurface, it recently became an area of interest for geotechnical
engineers, and its reproduction started in laboratory conditions for geotechnical
applications [4]. Certain unicellular microorganisms are responsible for the biologically
driven urea hydrolysis (eq. 1) through their metabolic activity, which initiates the
nucleation conditions for calcite particles in the presence of calcium (eq. 2) in a
supersaturated system:
CH4N2O + 2H2O 2NH4+ + CO32- (1)
Ca2+ + CO32- CaCO3 (s) (2)

D. Terzis and L. Laloui / Effect of Treatment on the Microstructural Characteristics 971
Calcite particles endow the material with cohesion and contribute their own
mechanical properties to its overall response. Van Paassen [2] reported the increase in
Unconfined Compressive Strength (UCS) and peak strength under drained and
undrained shear when the calcite content increases. Al Qabany and Soga [5] reported
an increasing UCS for an increasing dry density because of cementation with the
precipitated mass obtained after prescribing the patterns of different urea-calcium
chloride concentration. The results obtained in this latter work attempt to highlight the
role of chemical factors in the geometrical and spatial distribution of crystals and
associate this role with the engineering properties of the material. Somewhat
surprisingly, for bigger crystals that are formulated after injecting 1 M of CaCl 2, no
increase in UCS has been reported for any calcite content. Indeed, the tested samples
were reported not to exhibit homogenous cementation for 1 M CaCl2 concentration
with the cemented volume limited close to the injection point.
The cementation of large volumes of soil in large-scale applications [6],[7] allows
one to obtain samples that are homogenously cemented at the scale of conventional
geotechnical laboratory testing. Furthermore, using laboratory experimentation, other
studies [8], [9] have attempted to capture the effect of the provided treatment
conditions on the overall process efficiency. Cheng et al. [10] investigated the
distribution of calcite by considering the role of the degree of saturation in soils. They
provided evidence that precipitated calcite was more efficiently distributed for partially
saturated conditions because the menisci that formed around the grains maintained the
reactant solutions in the desired area of crystal deposition.
At the particle scale, normal and shear forces are transmitted. Hence, different
sizes and spatial distribution of the calcite particles are considered to result in different
mechanisms of strain localization, failure initiation and failure propagation. This
consideration offers new challenges in the modelling conception and numerical
simulation of this complex material. Fauriel and Laloui [11] implemented a theoretical
model to capture the performance of bio-improved soils by extending the constitutive
concept that was developed for aggregated soils by Koliji et al. [12] to account for the
effects of bonding and density change. Numerical studies [13],[14],[15] that
implemented the Discrete Element Method also aimed at predicting the behaviour of
cemented soils by reproducing the packing structure composed of soil and bonding
particles, where the position and arrangement of the latter ones ultimately affect the
distribution of the inter-particle forces.
The main objective of this study is to elucidate the role of calcium chloride
concentration on the yielded precipitation and clarify some predominant fabric
characteristics of the new structure. A hypothesis adopted herein is that higher
concentrations of CaCl2 and urea result in higher calcium and carbonate ion activity
product; thus, a higher supersaturation state () is more rapidly attained. Therefore,
CaCO3 precipitation occurs from a different initial value of . When calcium carbonate
precipitates, the available dissolved ions deplete until the solubility product is obtained,
and the reaction stops. Hence, for higher initial values of , more ions are available for
the crystals to develop their structure and planes and grow in size, and the overall
reaction has higher completion times.
972 D. Terzis and L. Laloui / Effect of Treatment on the Microstructural Characteristics
2. Experimental Details and Results
2.1. Bacteria and Sample Preparation
Sporosarcina pasteurii (ATCC 11859) was grown in sterile conditions according to the
related protocol, which was provided by the supplier company (ATCC medium: 1376),
whereas the entire growing phase was monitored using photomicroscope analyses.
Bacteria were grown at an optical density of 0.65, which was measured at a wavelength
of 600 nm (OD600). This biomass concentration corresponded to a cell concentration in
the range of 1x108 cell-forming units per ml (cfu/ml), which was estimated after plate
counting and found to hydrolyse urea up to 95%. This later parameter was measured
with ion chromatography analysis to determine the ammonium concentration (NH4+),
which is a by-product of urea hydrolysis. Prior to the injections, the bacteria were
centrifuged twice at 4000 g and suspended in a urea-nutrient broth medium to allow
hydrolysis to occur when the bacteria attach on the grains before the first batch of
calcium-rich solution was injected. Itterbeck (Smalls IKW, Netherlands) silica sand
(SiO2>98%) was selected for the treatment technique and had a mean grain size
diameter of 190 m. Specimens with a height of 110 mm and a diameter of 70 mm
were prepared in plastic tubes with dry pluviation at a target dry density of 1600 kg/m 3.
The urea-CaCl2 solutions were injected in batches using alternatively the top and
bottom of the tube to ensure an equal distribution of the reactant solutions along the
flow path. A retention time was allowed for the compounds to react. The volume of
injected batches was 1.2 times the volume of the pores, which was estimated at 170 ml.
By injecting an excess volume with respect to the total pore volume of the sample, both
the effluent that corresponded to the retained solution of the previous batch and this
excess volume, which directly flowed out after filtering the column, could be analysed.
This latter part carried crucial information of the precipitation evolution. The solid
precipitates that were observed in this excess volume were collected and dried at 52C
after filtering the solution. Observations obtained using Scanning Electron Microscopy
(SEM) show that the metastable spherical precipitates were initially formed and
subsequently transform to calcite cubic particles (Figure 1). Bacteria traces and
calcified cells are observed on the precipitated nuclei surfaces. The calcite nuclei grow
in size because of the rearrangement of Ca2+ and CO32- ions from their positions in the
metastable spherical particles to form their final lattice in calcite.
The precipitated cubic crystals grew in size, and their planes exhibited a prismatic
expansion. The retention times between injections were exactly selected to enable the
completion of this process. The provided treatment patterns (TP) for the cementation of
samples are presented in Table 1. All solutions included 10 g/L ammonium chloride as
an electron and nitrogen source for the bacteria and 3 g/L nutrient broth as an energy
source for the protein synthesis of S. pasteurii. TP3 refers to the condition in which the
CaCl2 concentration was doubled for the second half of the injections. The calcite
content was estimated after acid digestion, where 1 g of soil was washed with 1 N HCl,
and the pressure of the emitted CO2 was associated with a standard curve for pure
CaCO3 digestion. In total, 1.65 M of urea and 1.25 M CaCl2 were injected for TP1, and
the respective concentrations for TP3 were 0.85 M and 0.65 M. Higher concentrations
of urea were prescribed under the assumption that the hydrolysis efficiency decreased
throughout the treatment. The yielded sample after using TP2 was not homogenously
cemented with a CaCO3 content of 14% (w/w) in the first 30 mm of its top and bottom.
This latter sample was not subjected to mechanical loading.
Bacteria cells
Calcified
bacteria
Expanding
planes
Bacteria
trace
Figure 1. SEM analysis of the effluent crystals, which were collected after a fresh cementation solution
filtered the soil column.
Table 1. Prescribed treatment patterns (TP) following the injection of bacteria cells.
Urea CaCl2 injection retention CaCO3
solution cfu/ml M M rate time %
per injection per injection (ml/min) (hr) (w/w)
bacteria 108 0.5 - 5 8 -
TP1 - 0.33 0.25 5 12 7.5
TP2 - 0.1 0.05 1.5 continuous 14
TP3 - 0.5 0.25/0.5 5 24 4.5
2.2. Microstructural Fabric Characteristics
Here, some preliminary characteristics are obtained from the post-treatment

microstructural observations using SEM. An interesting finding is that two
predominant distinct forms of precipitates are observed: one of single particles and one
of mesocrystals, i.e., aggregates of single crystals. Mesocrystals are responsible for
filling the interstitial space among the grains, whereas single crystals primarily cover
the grain surfaces (Figure 2a). For TP3, the prescribed concentration is doubled from
0.25 M to 0.5 M between the first and second halves of the injections. In Figure 3a, the
crystals appear larger both as single particles and in mesocrystal formations compared
with the precipitation observed in Figure 2. Here, more cutting-edge orientated planes
are observed (3b and 3c). After the X-ray diffraction analysis, the rhombohedral 104
plane is identified as the predominant geometry plane.
(a)
(b)
Figure 2. Post-treatment SEM analysis for the sample cemented after prescribing the pattern TP1.
(a)
100m
(b) (c)
50m 20m
Figure 3. Post-treatment SEM analysis for the sample cemented according to TP3. The inlets refer to single
particles where bacteria traces can be observed in the grain-calcite adhesion plane. Red areas in (a) highlight
mesocrystals.
The observed mesocrystals are found to reproduce the identical geometry and
planes with their neighbouring single particles (inlets in Figure 3). A primary
conclusion is that there is a spatial variation of the supersaturation state and that the
crystals that precipitate in a considered subspace of the porous material tend to
reproduce identical crystallographic characteristics.
On the contrary, continuous flow did not yield the precipitation of single crystals.
This result may be attributed to the constant supersaturation state (). Continuous
injections imply that the already formulated crystals are provided with fresh ions,
which allows them to grow further in size. The hypothesis that the growth in size is
favored over the development of certain crystallographic orientations is enhanced by
observing that crystals without any specific crystallographic orientation fill the pore
space among the grains (4a and 4c). In Figure 4b, a detached grain allows us to observe
the surrounding cementing particles. Here, the observed particles have a spherical
shape and practically adhere to their neighbouring calcite particles after growing (4d).
(a) (b)
(d)
(c) (d)
Figure 4. Post-treatment SEM observation for the sample cemented according to TP2. The inlet in (b) is
presented in (d).
2.3. Mechanical response in undrained shear test
Two fully saturated cemented samples of 50 mm in diameter and 90 mm in height were

consolidated at 100 kPa and subsequently subjected to undrained shear under a
confinement pressure of 100 kPa. The treated samples were cored from the calcified
sample, as described in 2.1, according to the treatment patterns TP1 and TP3, and their
microstructural characteristics are demonstrated in Figures 2 and 3. An untreated
sample was prepared at a target void ratio of 0.56. The maximum and minimum values
of the void ratio for Itterbeck fine sand were 0.78 and 0.56, respectively, which were
estimated using a combination of vibration and compaction. The obtained elastic
moduli that corresponded to 50% of the peak strength, the peak and residual shear
strengths are presented in table 2.
Table 2. Strength and stiffness of the tested cemented samples with respect to the untreated material.
CaCO3 E50 peak residual
sample initial e0
% (w/w) (MPa) (MPa) (MPa)
untreated 0.56 0 36 0.20 0.11
TP1 0.48 7.5 240 1.10 0.24
TP3 0.52 4.5 279 0.50 0.16
Figure 5. Undrained triaxial shear tests for two biocemented samples.
(a) (b)
Figure 6. Samples of 4.5% (a) and 7.5% (b) CaCO3 content (w/w) after shearing.
3. Summary and Conclusions
This study presents the effect of chemical conditions on the fabric characteristics of
bio-improved sand for three different treatment patterns. Crystals are found to
precipitate in two forms: smaller single particles and mesocrystals. The size and
geometrical arrangement of the precipitated calcite is found to rely on the
supersaturation of the environment. This result is particularly important for
understanding the underlying mechanisms that govern the structure of the bio-
cemented geomaterial. The extensive knowledge of the microstructure will allow one to
account for these properties in constitutive modelling and numerical simulations, where
the nature and amount of cementing particles intimately affect the obtained response.
Acknowledgement:
We express our sincere thanks to the SNSF Swiss National Science Foundation (SNSF)
for funding the project (grant 513113 140246 - 200021_140246).
References
[1] J.K. Mitchell and J.C. Santamarina, Biological Considerations in Geotechnical Engineering. Journal of
Geotechnical and Geoenvironmental Engineering 131 (2005), 1222-1233.
[2] L.A. Van Paassen Biogrout, Ground Improvement by Microbially Induced Carbonate Precipitation. PhD
thesis Delft University of Technology, Netherlands, October 2009.
[3] V. Whiffin, L.A. Van Paassen, M.P. Harkes, Microbial Carbonate Precipitation as a Soil Improvement
Technique Geomicrobiology Journal 24 (2007), 417423.
[4] J.T. Dejong, K. Soga, E. Kavazanjian, et al., Biogeochemical processes and geotechnical applications:
progress, opportunities and challenges, Gotechnique 63 (2013), 287301.
[5] A. Al Qabany and K. Soga, Effect of chemical treatment used in MICP on engineering properties of
cemented soils. Gotechnique 63 (2013), 331339.
[6] L.A. Van Paassen, Bio-mediated ground improvement: From laboratory experiment to pilot applications
Geo-Frontiers (2011), 4099-4108.
[7] A. Esnault Filet, J.P. Gadret, M. Loygue, S. Borel, Biocalcis and its applications for the consolidation of
sands, Grouting and Deep Mixing (2012), 1767-1780.
[8] B. C. Martinez, J. T. DeJong, T. R. Ginn, B. M. Montoya, T. H. Barkouki, C. Hunt, B. Tanyu, and
D. Major. Experimental Optimization of Microbial-Induced Carbonate Precipitation for Soil
Improvement Journal of Geotechnical and Geoenvironmental Engineering 139 (2013),587-598.
[9] A. Al Qabany, K. Soga and C. Santamarina, Factors Affecting Efficiency of Microbially Induced Calcite
Precipitation Journal of Geotechnical and Geoenvironmental Engineering 138 (2012),992-1001.
[10] L. Cheng, L. Cord-Ruwisch, and A. M. Shahin, Cementation of sand soil by microbially induced calcite
precipitation at various degrees of saturation, Canadian Geotechnical Journal 50 (2013), 8190.
[11] Fauriel S., L. Laloui, A bio-chemo-hydro-mechanical model for microbially induced calcite
precipitation in soils, Computers & Geotechnics (2012) 46, 104120.
[12] A. Koliji, L. Laloui, L. Vulliet, Constitutive modeling of unsaturated aggregated soils, International
Journal of Numerical and Analytical Methods in Geomechanics, 34 (2010), 18461876.
[13] Y.-H. Wang and S.-C.Leung, A particulate-scale investigation of cemented sand behaviour, Canadian
Geotecnical Journal 45 (2008), 2944.
[14] M. Jiang, F. Zhang, Y. Sun, An evaluation on the degradation evolutions in three constitutive models
for bonded geomaterials by DEM analyses, Computers and Geotechnics 57 (2013), 116.
[15] T.M. Evans and A. Khoubani, Simulating mechanical response in bio-cemented sands
Computer Methods and Recent Advances in Geomechanics Oka, Murakami, Uzuoka & Kimoto (Eds.)
(2014), 15691574.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-978
Deformation of Microbial Induced

Calcite Bonded Sands: A Micro-scale
Investigation
Brina MONTOYAa,1 and Kai FENG a
a
Department of Civil, Construction, and Environmental Engineering, North Carolina
State University, Raleigh, North Carolina, USA
Abstract. Microbial induced calcite precipitation (MICP) is a novel ground

improvement method that increases the strength and stiffness of sand using natural
biogeochemical processes. This study investigates the mechanics of how
microbial induced calcite bonds deform during loading. Surface energy
measurements are conducted on silica and calcite substrates in order to assess
whether the bonds experience a cohesive or adhesive failure. The results from the
surface energy assessment indicate that calcite-calcite cohesion tends to be the
weakest within the system and where the MICP bonded sands will likely fail.
These findings are supported by particle contact loading tests, where two silica
particles are bonded together through the MICP process, and the deformation of
the silica particles and the calcite cement is monitored during shearing and axial
loading. These results indicate the bond fails within the calcite phase. The results
of the micro-scale study are then translated to the shear bands of a triaxial
specimen of MICP sand. A triaxial specimen is cemented to a moderate level of
MICP cementation, using shear wave velocity measurements as a process
monitoring technique. The cementation degradation is monitored using shear
wave velocity during drained shear and the thickness of the shear band is estimated.
Scanning electron microscopy (SEM) is also used to evaluate the MICP bonded
sand grains within and outside the shear band. The SEM results agree with the
micro-scale investigation that cohesive failure appears to dominate during shearing.
Keywords. Cemented sands, microbial induced calcite precipitation, shear

strength, cementation degradation
1. Introduction
Microbial induced calcite precipitation (MICP) is a novel ground improvement method

using natural biogeochemical processes. Through urea hydrolysis induced by
Sporosarcina pasteurii, a common alkaliphilic soil bacteria, calcite is precipitated at
soil particle contacts which increases the strength and stiffness of sand [1]. Micro-scale
investigations have demonstrated that the MICP process results in calcite concentrated
at particle contacts, with a thin layer of calcite coating the sand grains [2, 3, 4]. This
precipitation pattern has an important effect on the bulk properties, since the calcite
precipitated at the particle contacts has a greater effect on increasing the strength and
stiffness of the bio-cemented sands. Micro-CT scanning of a MICP treated specimen
1
Brina Montoya, Assistant Professor, Department of Civil, Construction, and Environmental
Engineering, North Carolina State University, Raleigh, NC, USA; E-mail: [email protected].
B. Montoya and K. Feng / Deformation of Microbial Induced Calcite Bonded Sands 979
indicates that not only does the calcite have a preference to precipitate at the particle
contacts, but also the pore structure of the bio-cemented sand remains open and
connected [5]. As bio-cemented soils are sheared, they experience a decrease in shear
wave velocity due to the degraded calcite [6]. Micro-scale evaluations have
qualitatively indicated that the calcite fails within the calcite phase and not at the
contact between the calcite and silica sand grain [3]. Tagliaferri et al. [7] performed
drained triaxial tests within an X-ray micro-CT scanner and observed that as the bio-
cemented specimen was sheared, calcite would disappear from view because the size of
the generated calcite fines were below the resolution of the CT scanner.
This study further investigates the mechanics of how microbial induced calcite
bonds deform during loading. Surface energy measurements are conducted on silica
and calcite substrates in order to assess whether the bonds will fail in a cohesive failure
or adhesive failure (Figure 1). These findings are compared to particle contact loading
tests, where two silica particles are bonded together through the MICP process, and the
deformation of the silica particles and the calcite cement is monitored during shearing
and axial loading. The results of the micro-scale study are then translated to the shear
band of a triaxial specimen of MICP sand. The triaxial specimen is cemented to a
moderate level of MICP cementation, using shear wave velocity measurements as a
process monitoring technique. The cementation degradation is monitored using shear
wave velocity during drained shear and the thickness of the shear band is estimated as a
function of cementation level. Scanning electron microscopy is also used to evaluate
the MICP bonded sand grains within and outside the shear band after shearing and
compared to the results from the micro-scale investigation.
2. Micro-scale Investigation of MICP Bonded Sands
The mechanics of how microbial induced calcite bonds deform during loading is
investigated using two techniques: surface energy measurements and particle-bond-
particle deformation testing. The results from the micro-scale investigation provide
valuable insight into the behavior of bio-cemented sands and which failure type,
cohesive or adhesive (Figure 1), can be expected at the particle contact.
Cohesive Failure Adhesive Failure
Figure 1. Calcite bond failure mechanism alternatives (modified from DeJong et al. 2010).
2.1. Characterization of MICP Bond Surface Energy
The failure mechanism of the microbial induced calcite bond was first evaluated
by determining the work of cohesion and adhesion of a typical calcite-silica system by
measuring the surface energy of the minerals. The work of cohesion and adhesion are
algebraically related to surface energy of the minerals. Surface energy is defined as the
980 B. Montoya and K. Feng / Deformation of Microbial Induced Calcite Bonded Sands
amount of work required to create a unit area of new surface of specific material.
Hence, the work of cohesion for the calcite cement can be defined as:
Wc = 2 Sc (1)
where Sc is the solid surface energy of calcite. Similarly, the work of adhesion between
calcite (C) and silica (Si), WSiC, can be expressed as a function of their surface
energy components:
LW LW + +
WSiC =2(Si C ) + 2(Si C ) + 2(Si C ) (2)
The surface energy components presented in Equation 2 can be determined directly

from the sessile drop method presented in [8]. Surface energy () measurements are
used to determine the cohesive and adhesive characteristics of asphalt binder and
asphaltaggregate systems (e.g. [9]), and similar methods were applied to bio-cemented
sands. The sessile drop method using a goniometer was used in the study presented
herein [10]. In the goniometer, a drop of probe fluid was released from a syringe onto
the polished mineral aggregate. The contact angle between probe fluids and the mineral
aggregate was measured and used to calculate surface energy. The corresponding
surface energy components were then determined following [8] (Table 1).
Table 1. Surface Energy Components of Silica and Calcite
Material and Surface Energy Components (mJ/m2)
Sample No. + - AB LW Total
Optical Calcite 0.0 11.0 0.0 37.5 37.5
Yellow Calcite 0.1 16.8 2.6 37.5 40.1
White Calcite 0.9 30.3 10.4 38.2 48.6
Silica 0.2 43.0 6.2 37.0 43.2
The surface energy measurements conducted on geologic specimens are used to

determine the likelihood of cohesive or adhesive failure of MICP-cemented sands. The
surface energy of three calcite specimens and one silica specimen revealed that the
surface energy of the calcite and silica are relatively similar to each other and similar to
values reported in literature [10]. The average surface energy of the geologic calcite
specimens tested varied from 37.5 to 48.6 mJ/m2 while the average silica specimen
surface energy was 43.2 mJ/m2 (Table 1). Using these surface energy results, the work
of cohesion and adhesion was determined (Table 2).
Table 2. Work of Cohesion and Work of Adhesion of Silica and Calcite

Material and Surface Energy and Work Parameters (mJ/m2)
Sample No. Total Wc Wa
Optical Calcite 37.5 75.0 77.5
Yellow Calcite 40.1 80.2 82.3
White Calcite 48.6 97.2 92.6
Silica 43.2 86.4 /
The work of cohesion and adhesion results indicate, based on the limited geologic
specimens tested, it is more likely to have a cohesive failure within the calcite bond
than an adhesive failure at the interface of the calcite cementation and silica particle.
However, a relatively large range in calcite surface energy was observed, thus further
testing directly with MICP minerals is desired.
2.2. Particle Bond Particle Deformation Behavior
The mechanical properties of a calcite bond were evaluated by loading two rounded
silica gravel particles (e.g., pea gravel) bonded by MICP. An electromagnetic testing
system under displacement control was used to achieve fine control of loading for the
particle-bond-particle behavior. The gravel particles were epoxied to the end caps
(Figure 2) and the end caps were placed into a treatment bath to induce calcite
precipitation, following the procedure outlined in [11]. After 20 bio-cementation
treatments, when the development of calcite was visually confirmed (Figure 2), the end
caps were connected to the testing apparatus and the side bracing removed. The
electromagnetic testing system was then used to perform either an extension or torsion
loading sequence. Upon loading, the load-deformation results were recorded and the
resulting failed bonds were evaluated.
Figure 2. Test set-up (left) and treated particles (right) for particle-bond-particle testing.
The mechanical tests were conducted on particles with average diameters of 10
mm and 5 mm. Multiple diameters were used because preliminary observations
suggested that the mechanical properties are influenced by the particle diameter. The
results from the particle-bond-particle tests indicate that the smaller particles exhibited
a slightly larger tensile strength than the larger particles; however, the torsional
strengths for both particle diameters were equivalent (Table 3). Furthermore, the failed
surface between the particles was evaluated after loading and illustrated that the
particles failed within the calcite phase of the bond and not at the calcite-silica
boundary.
Table 3. Results of Particle-Bond-Particle Deformation Tests

Diameter Tensile Strength
Torsional Strength (Nm)
(mm) (N)
10 3.20 0.35
5 3.75 0.35
3. Behavior of MICP Bonds in Triaxial Shear
The results of the micro-scale study are translated to the shear band of a triaxial
specimen of MICP treated sand.
3.1. Shearing Behavior of MICP Cemented Sands
One MICP and one untreated triaxial specimen were prepared by dry pluviation to a
target relative density, DR, of 40%. The specimens were seated under 100 kPa
confining pressure. For the MICP specimen, bio-cementation injections were
conducted as outlined in [12]. To monitor the cementation process during the treatment
period, shear wave velocity measurements were conducted using bender elements
installed in the triaxial end caps, following the procedures outlined in [13]. The
cementation injections were repeated until the target shear wave velocity, Vs, was reach.
The MICP-treated sand reached a Vs = 610 m/s during treatment, which corresponded
to a mass of calcite of 1.7%, based on post-test acid washing [11]. Once the
cementation injections were terminated, the specimens were saturated with deaired
water and back-pressure was applied to achieve a B-value of 0.95 or higher. The
specimens were sheared at a rate of 2.5% axial strain per hour to a strain of 10 15%.
The results of the triaxial testing indicates that the MICP treatments increased both
the peak and residual shear strength of the sand (Figure 3). This response is similar to
other related studies [6, 12]. The degradation of cementation is captured using the Vs
measurements. The Vs is a bulk measurement based on the collective stiffness of the
specimen; therefore, cementation degradation within a localized zone (compared to
global degradation) will have a limited effect on the Vs values. The Vs measurements
rapidly decrease until the peak strength is reached, and continue to decrease during
shear strength softening. Once the residual strength was reached, the Vs remains
relatively constant, indicating that the shearing behavior was constrained to the
localization zone established within the first 3% strain.
Figure 3. a) Stress-strain and b) shear wave velocity measurements for uncemented (0% calcite) and
moderately cemented (1.7% calcite) specimens under isotropically consolidated drained compression loading.
The shear band thickness of the specimen was estimated using the shear wave
velocity measurements and the final specimen length. The specimen length at the end
of shearing was 11.5 cm and the final Vs was about 540 m/s. Assuming that the sand
outside of the shear band remains intact (e.g., representative of the initial Vs = 610 m/s)
and the MICP sand within the shear band is degraded to a Vs = 190 m/s (e.g., the
uncemented initial conditions), the thickness of the shear band can be estimated by
using a weighted average of the shear wave velocities with respect to the specimen
length. Following this approach, the shear band thickness was estimated to be about 19
mm. The calculated shear band thickness is similar to the 20 mm zone of localization
visually observed in the specimen after shearing.
3.2. Microscopy Investigation of Sheared MICP Cemented Sands
Scanning electron microscopy (SEM) was used to evaluate the MICP bonded sand
grains within and outside the shear band after shearing and compared to the results
from the micro-scale investigation. Samples were carefully collected from the MICP
cemented specimen to be used for the SEM investigation, where a Hitachi S3200N
Variable Pressure Scanning Electron Microscope was used. Samples within the dead
zones of the triaxial specimen [14] were collected to represent the non-sheared MICP
sands. Samples were also collected within the zone of localization, as identified by the
deformed shape through the membrane. Once collected, the samples were prepared for
use within the SEM by coating the cemented sands with a gold/palladium alloy to
reduce charging while imaging.
The SEM investigation indicated the non-sheared cemented sands are dusted with
the MICP minerals along the surface of the grains with a preference of calcite minerals
at the particle contacts. This distribution of calcite can be seen in Figure 4. Small
granular calcite minerals cover the sand grains in Figure 4, and as the magnification of
the SEM images is increased, the focus of the images is at the contact between the two
particles. When the magnification of the images is increased to 1000x and 2000x the
calcite bridge connecting the two sand grains can clearly be seen.
Figure 4. SEM images of MICP cemented sands, focusing on calcite bond between sand grains at
magnifications of (from top left to bottom right) 200x, 500x, 1000x, and 2000x.
The SEM images of particles collected within the localization zone reveal a
different behavior. The sand grains are still dusted with MICP minerals; however, the
particles are no longer bonded. The majority of the particles do not show any
indication that they had previously been bonded to other particles, thus implying that
the bonded particles experienced a cohesive failure during shearing. However, there
were a few indications of adhesive failure (Figure 5). The void where a sand particle
was attached to the calcite bond can be seen in Figure 5. Whether the detached sand
particle was removed during shearing or during the SEM sample preparing is unknown;
however, this illustration of adhesive failure was in the minority during the SEM
investigation of the particles collected from the localization zone.
Figure 5. SEM images of MICP cemented sands showing indications of adhesive failure collected from the
shear zone of the specimen at magnifications of (from left to right) 200x, and 500x.
The SEM investigation also illustrated that the calcite minerals coating the sand
particles were induced by bacterial. The impressions of the rod-shaped bacteria in the
calcite minerals can be seen in Figure 6. Energy dispersive X-ray spectra were also
collected on the MICP-treated sands, and the spectra indicated the elements present
include, in decreasing order of abundance, calcium, silicon, oxygen, and carbon. These
are the elements expected for the silica sand grains and calcium carbonate minerals.
Gold and palladium were also present in the spectra in small amounts due to the
gold/palladium alloy coating.
Figure 6. SEM image of bacterial imprints within a calcite mineral.

4. Conclusions
The results from the micro-scale and triaxial experiments presented herein have
provided insight into the behavior of cementation bonds within the MICP treated sands,
specifically:
Surface energy measurements and particle-bond-particle deformation
experiments indicate that a cohesive failure within the calcite bond is the more
likely failure mode.
MICP bond mechanical properties are likely dependent on the soil particle
diameter.
Shear wave velocity measurements capture cementation degradation during
shearing and allow for an estimation of shear band thickness.
SEM images illustrate the calcite mineral precipitation patterns, with a
precipitation preference at particle contacts, and that the cohesive failure
mode appears to be the dominant failure mechanism within the zone of
localization.
References
[1] DeJong, J.D., and 33 others (2013). Biogeochemical Processes and Geotechnical Applications:
Progress, Opportunities, and Challenges. Geotechnique, Vol. 63., No. 4, 287-301.
[2] Chou, C.W., Seagren, E.A., Aydilek, A.H., and M. Lai. (2011). "Biocalcification of sand through
ureolysis." ASCE Journal of Geotechnical and Geoenvironmental Engineering, Vol. 127, Issue 12, pp.
1179-1189.
[3] DeJong, J.T., Mortensen, B.M., Martinez, B.C., Nelson, D.C. (2010). Bio-mediated soil
improvement. Ecological Engineering, Vol 36, 197-210.
[4] Martinez, B. C. & DeJong, J. T. (2009). Bio-mediated soil improvement: load transfer mechanisms at
micro- and macro-scales. Proc. 2009 ASCE US-China Workshop on Ground Improvement
Technologies, Orlando, FL, 242251.
[5] DeJong, J.T., Soga, K., Banwart, S.A., Whalley, W.R., Ginn, T., Nelson, D.C., Mortensen, B.M.,
Martinez, B.C., Barkouki, T. (2011). Soil Engineering In-vivo: Harnessing Natural Biogeochemical
Systems for Sustainable, Multi-Functional Engineering Solutions. Journal of the Royal Society
Interface, Vol 8, 1-15.
[6] Montoya, B.M., and DeJong, J.T. (2015). Stress-Strain Behavior of Sands Cemented by Microbially
Induced Calcite Precipitation. Journal of Geotechnical and Geoenvironmental Engineering, 141(6).
[7] Tagliaferri, F., Waller, J., Ando, E., Hall, S. A., Viggiani, G., Besuelle, P. & DeJong, J. T. (2011).
Observing strain localisation processes in bio-cemented sand using X-ray imaging,. Granular Matter,
Vol 13, No. (3), pp. 247250.
[8] Little, D.N., Bhasin, A. (2006). Using Surface Energy Measurements to Select Materials for Asphalt
Pavement. Final Report for NCHRP Project 9-37, Transportation Research Board.
[9] Bhasin, A. Little, D.N. (2007). Characterization of Aggregate Surface Energy Using the Universal
Sorption Device. Journal of Materials in Civil Engineering, Vol. 19, No. 8, 634-641.
[10] Koc, M., Bulut, R. (2013). Assessment of sessile drop device and new testing approach for measuring
contact angles on aggregates and asphalt binders. Journal of Materials in Civil Engineering, 26(3), pp.
391398.
[11] Mortensen B.M., Haber, M., DeJong, J.T., Caslake, L., Nelson, D.C. (2011). Effects of Environmental
Factors on Microbial Induced Calcite Precipitation. Applied Microbiology.
[12] Feng, K., and Montoya, B.M. (2015). Influence of Confinement and Cementation Level on the
Behavior of Microbial Induced Calcite Precipitated Sands under Monotonic Loading. Journal of
Geotechnical and Geoenvironmental Engineering, (in review).
[13] Montoya, B.M., Gerhard, R., DeJong, J.T., Weil, M., Martinez, B., Pederson, L. (2012). Fabrication,
Operation, and Health Monitoring of Bender Elements for Aggressive Environments. ASTM Journal
of Geotechnical Testing, Vol 35, No. 5.
[14] Head, K. H. Manual of soil laboratory testing. Vol 3: Effective stress tests. London, 1986, p. 7431238.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-986
Transitional Behaviour in Asphalt Derived

Aggregate Specimens
Katherine Ailiang KWAa,1 and David AIREYa
a
University of Sydney, Sydney, Australia
Abstract. A Critical State Soil Mechanics approach has been taken to understand
the interactions between the different aggregates and bitumen present in an asphalt
mix. A series of conventional drained triaxial tests have been performed on
samples of the aggregate used in Dense Asphaltic Concrete (DAC). These
aggregate samples contained no bitumen and were prepared at two densities. The
dense aggregate samples had a void ratio equivalent to the asphalt (DAC) samples
where the bitumen was assumed to be part of the void space. The loose aggregate
samples were prepared at a void ratio that was as loose as physically possible. The
behaviour of the aggregate samples was compared to the asphalt samples when
loaded continuously at a rate of 0.3% strain/hour. It has been found that at these
slow rates of loading, the aggregates control the behaviour of the asphalt and the
5% bitumen content had little effect. Transitional soil behaviour of the aggregates
was also observed as the dense and loose aggregate samples did not approach a
unique critical state line when sheared to failure. The microstructure of the particle
arrangements in the dense and loose aggregate samples was also observed under
optical microscopes. Different particle arrangements were characteristic of the
dense and loose samples. More bulky grain-to grain contacts were evident in the
dense aggregate samples as the fines in the loose aggregate samples tended to
flock together and form bridges between the bulky grains. As a result of this
transitional soil behaviour, it was difficult to define Critical State Soil Mechanics
parameters for DAC as required for advanced numerical modeling. This difficulty
was also a consequence of the challenges in performing triaxial tests on the dense
aggregate material.
Keywords. bitumen, transitional soils, triaxial tests
1. Introduction
Asphalt is widely used in the construction and surfacing of roads. It is a mixture of

various coarse and fine aggregates, filler and bitumen. The Roads and Maritime Service
(RMS) in Australia define asphalt mixes by their different aggregate gradings as shown
in Figure 1. Dense Asphaltic Concrete (DAC), also known as Dense Graded Asphalt, is
commonly used in Australia. It is a well graded asphalt mix with a relatively low
bitumen content of 5% [1].
1
Corresponding Author: School of Civil Engineering, The University of Sydney, Sydney, NSW 2006,
Australia. Email: [email protected]
K.A. Kwa and D. Airey / Transitional Behaviour in Asphalt Derived Aggregate Specimens 987
Figure 1: Particle Size Distributions and Relative Proportions of Constituent Materials in DAC
Traditional asphalt mix design has been oversimplified as the individual

components of asphalt and their influence on the asphalt mix have not been considered
[2]. Researchers have developed numerous models and approaches that attempt to
describe the behaviour of asphalt under a variety of loading conditions. However, there
is disagreement as to how the bitumen interacts with the aggregates and fines and
which component when combined with the bitumen controls the behaviour of the
asphalt. However, researchers who have used a Critical State Soil Mechanics approach
have developed a logical and consistent description of models that describe how the
separate aggregates, bitumen and air voids components in an asphalt mix interact with
one another. Researchers [3], [4], [5] found that the angle of internal friction depended
on the type and gradation of the aggregate and that the angle of internal friction
decreased with increasing bitumen content. Other researchers [6], [7], [8] developed a
logical qualitative description of how the different components in asphalt behaved
when asphalt underwent permanent deformation. Nijoboer (1948) [9] successfully
modeled the compaction behaviour of asphalt through developing a modified Soil
Mechanics Mohr-Coulomb theory.
There is now an interest in the behaviour of asphalt under slow, continuous loading
conditions which is relevant for example when asphalt is subject to land subsidence
from underground mining [10]. This has been investigated by performing conventional
drained triaxial tests at various confining pressures to obtain DACs Critical State Soil
Mechanics parameters and to investigate the role of bitumen in DAC under slow
continuous rates of loading. Triaxial tests have been performed on cored DAC samples
and DAC aggregate samples of two different densities. The DAC dense and loose
aggregate samples were also examined for transitional soil behaviour as their particle
arrangements were observed under a microscope.
1.1. Specimen Preparation
The Roads and Maritime Services (RMS) supplied DAC sample cores and their
aggregate composites to the University of Sydney which conformed to the Roads and
Traffic Authoritys (RTA) Roadworks specifications R116 (DAC). The particle size
988 K.A. Kwa and D. Airey / Transitional Behaviour in Asphalt Derived Aggregate Specimens
distributions and the material proportions of the aggregates required in DAC are
outlined in Tables 1 (a) and (b).
Table 1(a). Particle Size Distributions Table 1(b). Mix Proportions

Sieve Size (mm) DAC % Passing
19 100
13.2 99 Aggregate DAC (%)
9.5 83.38 Basalt 14mm 19.8
6.7 66.74 Basalt 10mm 19.4
4.75 56.81 Basalt 7mm 6.4
2.36 39.97 Basalt Dust 35.1
1.18 27.44 Benedicts Glass 13
0.6 20.74 Hydrated Lime 1.14
0.3 11.79 C450 Bitumen 5.2
0.15 7.69
0.075 5.92
Bitumen 5.17
To prepare a batch of cored DAC samples, 200kg of DAC aggregate mix was
taken from a plant production run and delivered to the Sydney Fulton Hogan (FH)
asphalt laboratory. Hot bitumen was added to the asphalt aggregate mixes and
compacted in a shear box compactor in the FH laboratory and 450mm x 150mm x
185mm prismatic specimens of asphalt were produced. Approximately 10 cylindrical
specimens of asphalt with a diameter of 70mm and a length of 140mm were cored from
each of the compacted rectangular prisms of asphalt. These cored specimens as shown
in Figures 2(a) and (b) had very little air with a void ratio, of 0.08. DAC
aggregate only samples were prepared with identical gradings to the cored samples
outlined in Table 1(b). They did not contain any bitumen and were prepared in the
Sydney University Civil Engineering Geomechanics Laboratory as shown in Figure 3.
Two kinds of aggregate samples were prepared, dense and loose. The dense aggregate
samples had a target void ratio equivalent to the cored asphalt sample except that
the bitumen present in the (cored) asphalt samples was counted as air as shown in
Figure 4. The dense aggregate samples that had a target void ratio of were
compared to the asphalt samples. This target aggregate sample void ratio, was
calculated to be 0.24. To achieve this target void ratio, the aggregates were compacted
with water in a split mould to their optimum moisture contents (OMC) of . This
value had been determined by a modified compaction test performed as per
AS1289.5.5.1 [11]. The loose aggregate samples were prepared as loose as was
physically possible resulting in a void ratio of 0.55.
a b
Figure 2 (a): DAC Cored Sample Side View, (b) DAC Cored Sample Plan View
Figure 3: DAC Aggregate Only Sample and Split Mould Figure 4: Definition of Voids
In the triaxial, the samples were saturated under elevated back pressures of 600 to
1000kPa and were considered fully saturated when the inflow rate dropped below
3mm3/min. The samples were then isotropically consolidated under confining pressures
of 50 to 1000kPa and subsequently sheared under drained conditions. The asphalt and
dense aggregate samples were sheared at a rate of 0.0092mm/min (0.3% strain/hour).
This also allowed enough time so that most of the pore pressures could dissipate
throughout the samples given that the coefficient of consolidation was found to be
0.102mm2/min. The more permeable loose aggregate samples were sheared at a rate of
0.092mm/min.
To observe and compare the microstructure of the dense and loose aggregate
samples particle arrangements, dense and loose epoxy bound DAC aggregate samples
were prepared using a method similar to that previously described except that the water
was replaced with Megapoxy, a two part A-B epoxy that had a similar density to water.
One inch diameter samples were cored out of the epoxy bound samples as shown in
Figures 5 (a) and (b). A Struers TegraPol-21 grinding and polishing machine in the
Australian Microscopy and Microanalysis Research Facility (AMMRF) Material
Specimen Preparation Laboratory was used to prepare the end surfaces of each cored
sample for observation under optical microscopes in the AMMRF.
Figure 5(a): Epoxy Bound Sample with Core Holes, (b) A Cored Sample
From looking at the asphalt samples and the comparable dense aggregate samples that
were isotropically consolidated at confining pressures of 50 to 1000kPa and sheared to
failure in Figure 6, it is evident that the bitumen had little effect on the critical state
behaviour of DAC. The aggregates controlled the behaviour of DAC as a unique
Critical State Line (CSL) can be drawn through the end failure points of the asphalt and
aggregate samples test results. There were difficulties when trying to obtain these end
failure points in the tests due to the membrane rupturing and the development of non-
homogenous conditions towards the end of the test. This may cause some of the end
points to shift a little particularly for the samples that dilated large amounts.
Figures 7 and 8 compare the stress strain behaviour for the asphalt and aggregate
samples. At a shear rate of 0.3%/hour, the bitumen present in the asphalt had an effect
on the stress strain behaviour of DAC as the asphalt and aggregate stress strain curves
are different for the same confining pressures. At lower strains before the asphalt
reached failure, the bitumen was still acting like viscous fluid, holding the aggregates
together. This is why the asphalt samples reached a higher peak deviator stress and
generally dilated at a slower rate than the aggregate samples. There was also a small
variation in the density of the dense aggregate samples. It was difficult to consistently
achieve the target void ratio in the dense aggregate samples due to the large
differences in the aggregate sizes present. The combination of the effects from the
bitumen, small developments of excess pore pressures and minor variation in densities
in the aggregate samples, although small, make it difficult to conclude exactly how the
bitumen is affecting the stress strain behaviour of asphalt from the tests that were
performed.
In Figure 6, the loose aggregate samples shear to a different and approximately
parallel CSL to the dense aggregate samples. This is transitional soil behaviour.
Transitional soil behaviour is not well understood.
Figure 6: Consolidation and Shear Data for all Samples on Plane

Figure 7: Deviator Stress vs Axial Strain for Dense Aggregate and Asphalt Samples
Figure 8: Volume Strain vs Axial Strain for Dense Aggregate and Asphalt Samples
Researchers [12], [13], [14], [15], [16] have suggested that the particle
arrangements present in soils and breakage of the soils particles can explain
transitional soil behaviour. Figures 9(a) and (b) compare the different particle
arrangements that were observed using optical microscopes in the dense and loose
aggregate epoxy bound samples. In the dense samples, there were uniform distributions
of fines filling the voids between the bulky basalt particles. Contacts between the bulky
basalt particles were also observed. These particle arrangements are labeled in Figure
10. In contrast, the loose samples particles had flocks of fines that caused a bridging
effect between the bulky particles. This particle arrangement is highlighted in Figure 11.
There was also no detectable breakage in the DAC aggregate samples. Theoretically,
the particles of non-transitional soils that are sheared to a unique CSL are re-moulded
into random homogenous arrangements by the soil particles having to move and roll
over one another. However, the particle arrangement at the critical state could not be
verified as it was difficult to experimentally preserve and observe the particle
arrangements of an aggregate sample after it had been sheared to failure. Chang et al
[15] found that different characteristic particle arrangements formed during sample
preparation affected their critical state behaviour as they sheared to non-unique parallel
CSLs. Chang et al [15] tested silty sands and silty clays, whereas the DAC contains
gravelly, angular basalt aggregates that are comparatively larger. It is more difficult for
these large angular particles to move and roll over one another when sheared to failure.
Therefore, the different characteristic particle arrangements observed prior to testing,
shown in Figures 10 and 11, are more likely to remain in place up until the samples
have been sheared to failure.
a bulky
b aggregate
particles
uniform
distribution of fines
flocks of
bridging fines
bulky
aggregates
Figure 9: Idealised Particle Arrangement in (a) a Dense Aggregate Sample, (b) a Loose Aggregate Sample
large
basalt
bulky
aggregate
basalt
uniform aggregate
distribution of
bulky
fines
grain-to-grain
contacts
Figure 10: Particle Arrangement Characteristics in Dense Aggregate Samples
large
basalt
flocking aggregate
fines that bridge
between bulky
particles bulky
basalt
aggregate
large
voids
Figure 11: Particle Arrangement Characteristics in Loose Aggregate Samples

Thus a random homogenous arrangement is unlikely to be obtained and non-unique

critical state conditions, as shown in Figure 6, can be explained by the expected
difference in the particle arrangements.
3. Conclusion
At slow continuous loading rates of 0.3% strain/hour, it has been found that the 5%
bitumen content present in DAC has little effect on the critical state behaviour of DAC.
The well graded aggregates control the critical state behaviour of DAC. It is believed
that DAC aggregates displayed transitional soil behaviour because characteristically
different particle arrangements were present in the dense and loose aggregate samples.
Due to the transitional soil nature of the DAC aggregates, it is difficult to define
Critical State Soil Mechanics parameters required for advanced numerical modeling
4. Acknowledgements
The authors are grateful to the RMS for supplying the asphalt specimens and materials
used and to the AMMRF for the use of their polishing equipment and microscopes.
References
[1] Austroads. (2007). Guide to Pavement Technology Part 4B: Asphalt. Sydney: Austroads Incorporated.
[2] School of Civil and Environmental Engineering from the University of New South Wales, Road and
Traffic Authority of New South Wales. (2013). An Elasto-Viscoplastic Constitutive Model for Asphaltic
Materials. Sydney.
[3] Goetz, W. H., & Chen, C. C. (1950). Vacuum triaxial technique applied to bituminous aggregate
mixtures. Journal of the Association of Asphalt Paving Technologists , 19, 55-81.
[4] Van der Poel, C. (1958). On the rheology of concentrated suspensions. Rheologica Acta , 1, 198-205.
[5] Heukelom, W., & Klomp, A. J. (1964). Road design and dynamic loading. Proceedings of the
Association of Asphalt Paving Technologists , 33, 92-125.
[6] Hills, J.F. (1973) The Creep of Asphalt Concrete Mixtures. Journal of the Institute of Petroleum,
November 1973
[7] Van de Loo, P.J. (1978) The Creep Test, a Key tool in Asphalt mix Design and in the Prediction of
Rutting. Journal of the Association of the Asphalt Paving Technologists, Vol. 47 pp 389-413
[8] Song, J., & Pellinen, T. (2007). Dilattion Behaviour of Hot Mix Asphalt Under Triaxial Loading. Road
materials and Pavement Design , 8 (1), 103-125.
[9] Nijboer, L. W. (1948). Plasticity as a factor in the design of dense bituminous road carpets. New York:
Elsevier Science.
[10] Airey, D., & Prathapa, R. (2013). Triaxial Testing of Asphalt. International Conference on Soil
Mechanics and Geotechnical Engineering. Paris.
[11] Australian Standard (2000). AS1289 Method of testing soils for engineering purposes Section 5.5.1.
Australia
[12] Martins, F.B., Bressani, L.A., Coop, M.R. & Bica, V.D. (2001) Some aspects of the compressibility
behaviour of a clayey sand.Canadian Geotechnical Journal, 38(6) pp. 1177-1186
[13] Nocilla, A., Coop, M.R. & Colleselli, F. (2006) The mechanics of an Italian silt: an example of
'transitional' behaviour.Geotechnique 56(4) pp. 261-271
[14] Ferreira, P.M.V.,& Bica, A.V.D. (2006). Problems in identifying the effects of structure and critical
state in a soil with a transitional behaviour.Geotechnque, 56(7) pp. 445-454
[15] Chang, A., Heymann, G. & Clayton, C. (2011) The effect on the behaviour of gold tailings.
Geotechnique 61 (3) pp. 187-197
[16] Shipton, B. & Coop, M.R. (2012) On the compression behaviour of reconstituted soils. Soils and
Foundations, 2012;52(4) pp.668-681
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-994
Study of the hot mix asphalt used in middle

west region of Brazil
Carolina Q. ARANTES a,1, Lilian R. REZENDE b and Liedi L. B. BERNUCCI c
a
School of Civil Engineering, Federal University of Gois, Brazil
[email protected]
b
School of Civil Engineering, Federal University of Gois, Brazil
[email protected]
c
School of Civil Engineering, University of So Paulo, Brazil
[email protected]
Abstract. Flexible or asphalt pavements compose the Brazilian highway network,

which have to promote security on traffic on the highways, to support the effects of
changes in climate and to resist the traffic of vehicles. By this reasons, the structure
of the pavement is particularly important and must be constituted by layers, which
can distribute the loads into the structure, reducing the vertical load produced by the
action of traffic and, consequently, providing protection to the subgrade. In the
middle western Brazil, in the state of Gois, there is not advanced studies on asphalt
mixtures used in highways and in urban pavements, and there is not a technical
scientific proofing that the local aggregates available in the region, is the best option
considering the local conditions. Forward better conditions of infrastructure, it is
proposed to develop more efficient asphalt mixtures for pavements, by submitting
the mixtures through testing and analysis of mechanical strength, analyzing if the
most used asphalts mixtures are satisfactory or not. The hot asphalt mixture are
composed by aggregates and asphalt binders that are design in order to attend the
parameters set in the projects according to demand and traffic types. In this research,
it was proposed the study of three asphalt mixture design with different types of
asphalt binders with the same aggregates origin in a specific grain size range. The
binders used in this study are the conventional cement asphalt in the region, CAP
30/45 e 50/70 (Brazilian penetration grade), with low and medium penetration,
respectively, and the polymer modified asphalt (AMP). The CAP 50/70 is the most
used in the region considering the conventional binders, and right after we have the
CAP 30/45. The characterization tests of the materials that composed the mixture
are presented and their selection and characterization attended the standardized
procedures and requirements by the brazilian department responsible for highway
structures (DNIT). In a second stage, the design tests were performed, combining
aggregates and binders in different rates, by the Marshall Mix Design Method using
manual compression, until the great design was obtained. The laboratory tests were
realized with this great design to obtain its mechanical properties, given by the
tensile strength. This study presents that CAP 30/45 showed the best stability and
tensile strength for the mixture with regional aggregates.
Keywords. Hot Asphalt Mixtures, Marshall Mix Design Method, Mechanical

Properties.
1
Carolina Queiroz Arantes, School of Civil Engineering, Federal University of Goias, Avenida
Universitria, n 1488, Qd 86, Lt rea, Setor Leste Universitrio, CEP 74605-220, Goinia-GO, Brazil;
[email protected]
C.Q. Arantes et al. / Study of the Hot Mix Asphalt Used in Middle West Region of Brazil 995
1. Introduction
Hot mix asphalt are composed of aggregates and asphalt cement that are dosed to provide
a compact and resistant structure, attending the parameters determined in the projects
according to demand and traffic types. There are different solutions of asphalt mixtures
that can be applied and its variation happens in the particle size, and in the type of asphalt
binder used, which influence directly the quality, safety and durability of the pavement.
The aggregates represent the major portion of asphalt concrete, corresponding to
more than 90% by weight of the group: asphalt and aggregate. Thus, the aggregate
properties are very important for the performance of the pavement. The physical
properties of aggregates are the main reason to determine its application in asphalt
mixtures [1]. The properties of an asphalt mixture, such as stability, durability,
permeability, fatigue resistance and permanent deformation, are linked to its particle size
[2]. The shape and the texture of the aggregate are important factors that affect the
development of the sinking of wheel tracks [3]. The bituminous mixtures deportment
depends on the characteristics of aggregates (plastic deformation characteristics) and the
asphalt binder (fatigue characteristics) [4]. The interlocking between the particles and
their friction ensure the stability of the layer [5]. The selection of aggregates for the
utilization in asphalt concrete depends on their availability, cost, and quality, as well as
the type of application.
Asphalt cements (CAP) are binders with behavior as semisolid, viscoelastic, or
liquid, depending on the temperatures to which they are subjected, being respectively
low, normal and high temperatures. Have consistency limits, as applicable, and are
characterized according to Brazilian standards (NBR) of the Brazilian Association of
Technical Standards (ABNT) and the National Petroleum Agency (ANP).
The dosage of an asphalt mixture is a choice, by experimental procedures, an
optimum content of binder from a pre-defined range of particle size. This optimum
content varies according to the evaluation criteria, so this parameter is usually called
project content. In the 80s, American highway heavy traffic showed premature
permanent deformation attributed to excess binder in the mixtures. It was assumed that
the impact compaction of mixtures during dosing produced specimens with inconsistent
densities. This issue was addressed in US study on asphalt material, called Strategic
Highway Research Program (SHRP), which resulted in a new dosing procedure called
SUper PERfoming Asphalt PAVEment (SUPERPAVE).
In the state of Gois, central-western Brazil, there is a lack of advanced studies on
the asphalt mixtures used, both on highways and in urban pavements, and no technical-
scientific evidence that, in all aggregates available in the region, the most used asphalt
binder is the best to be employed on local conditions. Thus, aiming to analyze the
behavior of asphalt mixtures in this region, were realized the characterization of local
aggregates, the study of the dosage of three different mixtures and tensile strength tests.
2. Methodology
The selected mixtures were designed in the laboratory according to the Marshall Method
and submitted to conventional tests of stability Marshall and tensile strength. Tests were
performed in the Asphalt Laboratory of Civil Engineering School, Federal University of
Gois.
996 C.Q. Arantes et al. / Study of the Hot Mix Asphalt Used in Middle West Region of Brazil
In this study, the value of 7.5 x 106 was considered for traffic volume for a 10 years
project, because it is the most common volume used in the highway projects for the state.
The rainfall is usual in this region, but it is not equally distributed throughout the year,
which has tropical characteristics, with maximum in summer and minimum in winter.
Three different types of asphalt cements were used, the conventional asphalt binders
CAP 30/45 and 50/70 and the polymer modified asphalt (AMP) 60/85 type SBS
(Styrene-Butadiene-Styrene). These binders were chosen because they are the available
alternatives at the distributors that operate in the study region. The aggregates used
comes from Gois, whose the geological survey has considered the group of
metasediments as the Arax Group, with metamorphic (mainly micaschists and
quartzites) [6].
In this research, it was proposed the study of asphalt mixture design with the same
aggregates source in a specific grain size range as specified in the standard ES 031 [7]
and three different types of asphalt binders that was already mentioned.
2.1. Aggregates
For satisfactory performance of asphalt mixture, it is essential a good selection of the

aggregates, with the features of shape, adhesion, and resistance, specified in the standard
ES 031 [7] for the production of asphalt mixtures. The aggregates used in this research
were obtained from a private quarry, located approximately 15 km from the city of
Goiania, the state capital of Gois (Figure 1).
FEDERAL
DISTRICT
GOINIA
Figure 1. Location of Goinia, Gois state capital.

The commonly particle size range adopted in the region is the range C, established
by the National Department of Infrastructure and Transport (DNIT), and has guided the
choice of particle size for the study. The composition of the aggregates was performed
by the method of attempts, reaching proportions of each added aggregate and getting the
final particle size of the mixture, according standard ME 083 [8]. It has also set the
maximum size of the nominal aggregate opening, which is above the first sieve that
retains more than 10% of the material. The tests that were performed are listed below,
Table 1.
Table 1. Laboratory Tests performed with aggregates.

Tests Standard Observations
Determination of absorption and the
NBR NM 53 [9]
apparent density of coarse aggregates
Actual density of aggregates ME 084 [10]
Density of the Kerosene was used inside the
NBR NM 23 [11]
pulverized material bottle Le Chatelier
It is recommended that this value
Sand equivalent ME 054 [12]
is above 55%
Los Angeles Abrasion test ME 035 [13]
The aggregates were submitted to

Form Index ME 086 [14] the screens with circular aperture
sieves and framed graduation C
3 samples were tested with 500 g
Adhesiveness ME 078 [15] of passing material in 19mm sieve
and retained on the sieve 12.7mm
2.2. Binders
To specify an asphalt cements as suitable for paving asphalt, the physical characteristics
of the binder used are: hardness, measured by penetration of a standard needle in the
asphalt binder sample and flow resistance, as measured by viscosity tests. The
characterization of asphalt cements used in this research was conducted and provided by
Petrobrs.
2.3. Marshall Mix Design
The Marshall mix design was performed according to standard ME 043 [16] and the
aggregates were prepared in the proportions for each percentage of CAP (4.0; 4.5; 5.0;
5.5 and 6.0%). For the molding of the specimen, the materials were heated and the
mixture process was passed on a plate, heated to a temperature determined from the
viscosity-temperature curves of the binders (Table 2). After well homogenizing the
mixture and ensure optimum temperature, the mixture was brought to the mold and the
specimen was compacted.
Table 2. Temperatures of materials, mixtures and compaction.
Binders Binders Aggregates Mixture Compaction
Temperature (C) Temperature (C) Temperature (C) Temperature (C)
CAP 30/45 115 125 157 146
CAP 50/70 115 125 152 140
AMP 155 140 175 163
The specimens were cooled and extracted from the mold, and their diameter and
thickness were measured to find the correction factor, depending of the thickness. Their
dry mass and submerged were also measured for determining volumetric parameters, to
then carry out the test of stability. From this information, was determined the project
content for binders.
2.4. Tensile Strength
This test follows the requirements of the standard ME 138 [17] and used the same
Marshall mix design procedure for the implementation of the specimen. With the defined
project content, were molded three specimens of each type of CAP, which were used for
the tensile strength test, and the sample was placed between metal strips and subjected
to a load in the vertical diametrical plane, and the breaking load value was obtained.
3. Results
In this section, the results for characterization of the aggregates, mixtures design and
tensile strength are presented. An analysis of the aggregates on the perspective of
SUPERPAVE methodology is also discussed.
3.1. Aggregates
The granulometric analysis of the aggregate was performed according to procedures

described in the standard ME 083 [8] and tests were performed to determine the
proportion of material to form the chosen mixture of aggregates. The obtained size
distribution curves are presented in Figure 2.
100
90
80
Percent Passing (%)
70
60 GRAVEL 1
50 GRAVEL 0
40 GRIT
30 SAND
20 FILLER
10
0
0.010 0.100 1.000 10.000 100.000
Sieve Size (mm)
Figure 2. Particle size distribution of each aggregate.
The test results for density determination are shown in Table 3. The aggregates have
the same source rock, micaschists, and for this reason the test resulted in close density
values. Considering the absorption, the coarse aggregates did not present problems. The
aggregates used attended the parameters defined by standard ES 031 [7], as presented in
Table 4.
Table 3. Density results and absorption of the aggregates.
Aggregates Density (g/cm3) Absorption (%)
Gravel 1 (9.5 to 19 mm) 2.701 0.5
Gravel 0 (4.0 to 9.5 mm) 2.665 1.1
Grit (lower than 4.0 mm) 2.787 -
Sand (lower than 4.0 mm) 2.801 -
Filler 2.822 -
Table 4. Main parameters obtained for the aggregates.
Aggregate Abrasion (%) Form Index Sand equivalent (%)
Coarse 29 1.0 -
Fine (Sand) - - 76
Fine (Grit) - - 69
Standard ES 031 [7] < 50 > 0.5 > 55
The adhesion of the aggregates at binders was verified by the not shifting of the film
that covers the mixture and the results obtained for the three samples tested was
satisfactory in all cases (Figure 3).
CAP 30/45 CAP 50/70 AMP

Figure 3. Adhesion Tests.
The selected particle size of aggregates were checked to satisfy the SUPERPVAVE
requirements, wherein the mixture approaches the control point and prevents the
restricted area. In Figure 4, it is observed that the studied particle size range is below the
maximum density line on the graph, with maximum nominal size (TMN) equal to 19mm.
To the mixture, the particle size of the aggregates, shown in Figure 5, is included in the
range C of DNIT. The fractionation of the material was controlled to obtain the same
particle size for the mixtures with the three CAP.
100
90
80
70
Max Control Points
Percent Passing (%)
60
50 Min Control Points
40
Restrict Zone
30
Particle Gradation
20
10 Maximum Density
Gradation
0
00 0.25
0.0750.50.30.75
0.6 11.18
1.25 2.36
1.5 1.75 4.75
2 2.25 2.5 2.75
9.5 3 12.5
3.25 3.5 3.75
19.04
Sieve Size, mm Raised to 0,45 Power
Figure 4. Control points SUPERPAVE.
110
100
90
Percent Passing (%)
80 RANGE C - LOW
70 RANGE C - HIGH
60 COMPOSITION
50 WORK RANGE - LOW
40 WORK RANGE - HIGH
30
20
10
0
0.010 0.100 1.000 10.000 100.000
Sieve Size (mm)
Figure 5. Granulometric composition of the mixture Range C.

3.2. Marshall Mix Design
The composition of the aggregates was adjusted for each CAP content, and to target a
specimen with 1200g, the mass of each component of the mixture was determinated.
After compression and mold release the diameter (D), height (H), dry matter (DM) and
submerged weight (MSsub) of each specimen were measured, and volumetric parameters
were calculated. Analysis of the asphalt concrete stability results (Table 5) showed that
the great content of the samples had reached the minimum requirement of the standard
ES 031 [7]. The obtained parameters showed that the 30/45 CAP has the smaller binder
content and the best stability.
Table 5. Marshall stability results and volumetric parameters for each binder.
Parameters CAP 30/45 CAP 50/70 AMP
Density of CAP (g/cm) 1.02 1.02 1.03
Great content (%) 5.06 5.32 5.69
DMT (g/cm) 2.55 2.54 2.52
VMA 16.15 16.51 16.49
Vv (%) 4.0 4.0 4.0
VFA (%) 75.34 77.34 79.64
Stability (kgf) 1353.1 1046.4 1130.7
Obs.: CAP = asphalt cement, AMP = polymer modified asphalt, DMT = maximum theoretical density, VMA
= voids in the mineral aggregate, Vv = air void volume, VFA = voids filled with asphalt.
The analysis by SUPERPAVE methodology, the volumetric requirements of the
mixture are specified in standard MP2 [18], and the parameters obtained reached the
specified, as shown in Table 6. The exception happened to the VFA parameter, which
were specified the range of 65-75% depending on traffic volum considered (7.5 x 106).
This exception can be related to the fact that the process of compaction did not use rotary
compressor.
Table 6. Analysis by SUPERPAVE methodology.
SUPERPAVE Binder
Parameters
Specifications 30/45 50/70 AMP
Voids in th mineral
13.0 % 16.15 16.51 16.49
Aggregate (VMA)
Air void volume (Vv) 4.0 % 4.0 4.0 4.0
Voids filled with asphalt
65 75 % 75.34 77.34 79.64
(VFA)
3.3. Tensile Strength
Specimens were prepared with three asphalt mixtures with different types of CAP in
selected great content. The results are shown in Table 7. It is observed that the three
mixtures showed higher values than 0.65MPa, determined as minimum by standard ES
031 [7]. The mixture with the CAP 30/45 was the one with the highest value of tensile
strength by diametrical compression.
Table 7. Results for Tensile Strength.
Tensile Strength (MPa, 25C)
Binder
Average Standard deviation
CAP 30/45 1.25 0.09
CAP 50/70 1.06 0.07
AMP 1.09 0.03
4. Conclusion
The aggregates had properties that reach the parameters established by Brazilian
standards used for laboratory tests. To reach SUPERPAVE requirements regarding the
particle size, it is necessary that the project mixture passes through the control points and
avoid the restricted zone. It was observed that the particle size curve passes very close to
the restricted area, which indicates the presence of a lot of fine sand in the total sand, and
this mixture may show fragile behavior, and offering low resistance to permanent
deformation during service life.
It was observed that the CAP 30/45 delivers superior performance when used with
the most common aggregates in Goiania and region. However, additional tests are needed
to verify the performance of the mixture by testing the complex modulus, the permanent
deformation and the fatigue life. These additional tests will confirm the indication of
CAP 30/45, in the study area, for mixtures in the range C.
References
[1] F.L.Roberts, et al. Hot Mix Asphalt Materials, Mixture Design, and Construction. NAPA, 1996. Maryland,
USA, 1996.
[2] M.B.Cunha. Avaliao do Mtodo Bailey de Seleo Granulomtrica de Agregados para Misturas
Asflticas. Masters dissertation Univ. de So Paulo. So Carlos, Brazil, 2004.
[3] J. Zhang, et al. Comparison of Fundamental and Simulative Test Methods for Evaluating Permanent
Deformation of Hot-Mix Asphalt. Transportation Research Record TRR 1789, paper 02-3107. 2002.
[4] L.P.Specht. Avaliao de misturas asflticas com incorporao de borracha reciclada de pneus. Doctors
dissertation Programa de Ps Graduao em Engenharia Civil da Univ. Federal do Rio Grande do Sul.
Porto Alegre, Brazil, 2004.
[5] M.M.Lopes. Programa Interlaboratorial de Agregados e Misturas Asflticas da REDE ASFALTO N/NE.
Masters dissertation Univ. Federal do Cear. Fortaleza, Brazil, 2010.
[6] A.R.Sampaio, et al. Programa Levantamentos Geolgicos Bsicos do Brasil - PLGB. 2001 Braslia,
Federal District, Brazil, CPRM/DIEDIG/DEPAT, 2001.
[7] Departamento Nacional de Infraestrutura de Transportes DNIT. ES 031 Pavimentos Flexveis
Concreto Asfltico. Rio de Janeiro, Brazil, 2006. [in Portuguese]
[8] Departamento Nacional de Estradas de Rodagem DNER. ME 083 Agregados Anlise Granulomtrica.
Rio de Janeiro, Brazil, 1998. [in Portuguese]
[9] Associao Brasileira de Normas Tcnicas ABNT. NBR NM 53 Agregado grado Determinao da
massa especfica, massa especfica aparente e absoro de gua. Rio de Janeiro, Brazil, 2003. [in
Portuguese]
[10] Departamento Nacional de Estradas de Rodagem DNER. ME 084 Agregado Mido Determinao
da densidade real. Rio de Janeiro, Brazil, 1995. [in Portuguese]
[11] Associao Brasileira de Normas Tcnicas ABNT. NBR NM 23 Material Pulverulento
Determinao da massa especfica. Rio de Janeiro, Brazil, 2001. [in Portuguese]
[12] Departamento Nacional de Estradas de Rodagem DNER. ME 054 Equivalente de Areia. Rio de
Janeiro, Brazil, 1997. [in Portuguese]
[13] Departamento Nacional de Estradas de Rodagem DNER. ME 035 Agregados Determinao da
abraso Los Angeles. Rio de Janeiro, Brazil, 1998. [in Portuguese]
[14] Departamento Nacional de Estradas de Rodagem DNER. ME 086 Agregado Determinao do ndice
de forma. Rio de Janeiro, Brazil, 1994. [in Portuguese]
[15] Departamento Nacional de Estradas de Rodagem DNER. ME 078 Agregado grado Adesividade a
ligante betuminoso. Rio de Janeiro, Brazil, 1994. [in Portuguese]
[16] Departamento Nacional de Estradas de Rodagem DNER. ME 043 Misturas Betuminosas quente
Ensaio Marshall. Rio de Janeiro, Brazil, 1995. [in Portuguese]
[17] Departamento Nacional de Estradas de Rodagem DNER. ME 138 Resistncia Trao Esttica. Rio
de Janeiro, Brazil, 1994. [in Portuguese]
[18] American Association of State Highway and Transportations Officials AASHTO. MP2 Standard Test.
USA, 2001.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1002
Hydraulic and mechanical behaviour of

cement-bentonite mixtures containing
HYPER clay: impact of sulfate attack
Gemmina DI EMIDIOa,1, R. Daniel VERASTEGUI FLORESb, Claudia SCIPIONIa,
Evelina FRATALOCCHIc, Adam BEZUIJENa
a
Laboratory of Geotechnics, Ghent University, Ghent, Belgium
b
iMMC, Universit catholique de Louvain, Louvain-la-Neuve, Belgium
c
SIMAU, Universit Politecnica delle Marche, Ancona, Italy
Abstract. Cemented clays are regularly employed as cut-off walls to isolate

polluted soils or in ground improvement technologies. The objective of this
research was to evaluate the performance of a polymer-treated bentonite (HYPER
clay) and its impact on hydraulic and mechanical properties of cement-bentonite
(CB) mixtures in contact with sulfates. In this research, bender elements were
installed in a hydraulic conductivity cell, to simultaneously monitor the small-
strain shear modulus (G0) and the hydraulic conductivity (k) of cement-bentonite
mixtures. Initially, during permeation with deionized water, an increase of G0 with
time was observed, due to cement hydration (hardening). Conversely, after
prolonged permeation with sulfates, a decrease of G0 and a gradual increase of k
are expected due to degradation of the cemented material. Preliminary results
show that k of mixtures containing untreated bentonite increases due to contact
with sulfates; conversely, k of mixtures containing HYPER clay remains
unaffected. The stiffness modulus G0 was not significantly affected after a sulfate
contact period of about 100 days in none of the cement-bentonite mixtures. More
extensive monitoring is required for evaluating long-term mechanical degradation.
These observations suggest that the use of polymer-treated bentonites could
improve the resistance of cement-bentonite mixtures to sulfate attack maintaining
a low hydraulic conductivity.
Keywords. Clay, cement, sulfate attack, hydraulic conductivity, stiffness
1. Introduction
Cement-clay grouts are often used in contaminated soils either as parts of barrier
systems (e.g. cut-off walls) or for general ground improvement applications [1].
Deterioration of concrete by sulfates of an external source is a commonly observed
durability problem in cement structures exposed to seawater, contaminated soils or
groundwater containing high concentrations of sulfate ions. This durability problem,
also known as sulfate attack, occurs after a series of chemical reactions between sulfate
ions, cement paste and moisture [2].
1
G. Di Emidio et al. / Hydraulic and Mechanical Behaviour 1003
The two manifestations of sulfate attack are expansion, caused by formation of

ettringite and gypsum, and loss of strength and mass, caused by the deterioration of the
cohesiveness of the cement matrix [3].
For the design and the quality control of cut-off walls, cement-clay samples are
studied to evaluate their strength, hydraulic conductivity and chemical compatibility [4].
The small-strain shear modulus (G0) could be used to correlate the strength of a
cement-clay mixture [5]. G0 may be obtained from shear-wave velocity measurements
using bender elements. This technique was pioneered by [6], improved and studied by
many subsequent researchers [7,8, 9, 10].
Bender elements have been installed in a flexible-wall hydraulic conductivity cell,
to monitor (in a nondestructive way) G0 as well as the hydraulic conductivity,
simultaneously. In a previous work, in a cemented sample permeated with deionized
water, an increase in G0 with time was observed with increasing interparticle
cementation [11]. Conversely, a decrease of G0 may be expected due to the impact of
aggressive solutions, such as sulfates.
In this study, experiments were carried out to analyze the impact of sulfate attack
on cement-bentonite mixtures containing untreated bentonite and polymer-treated
bentonite using the HYPER clay technology [12] to evaluate their performance.
2. Materials and sample preparation
For this research a sodium-activated bentonite clay was treated with HYPER clay
technology, by mixing the clay with a polymeric solution containing an anionic
polymer, CarboxyMethyl Cellulose (CMC), and dehydrating the slurry at 105C for
about 24 hours. The chosen polymer dosage was 8% by dry weight (8 g of polymer for
100 g of clay).
The samples studied in this research consist of a mixture of bentonite clay (treated
and untreated), cement and water. Blast furnace slag cement of the type CEM III/B
32.5 N-LH / SR (26% Portland clinker, 70% blast furnace slag, 4% limestone) and
purified water with an electrical conductivity EC 10 S/cm and a pH of about 7.6
were used.
Deionised water, produced using a water purification system, was used to prepare
the samples and as reference permeant solution. During hydraulic conductivity testing,
the cement bentonite sample was initially permeated with purified water for 1 month to
allow for further hydration of the cement products. After that period, the sample was
permeated with a 25 g/L solution of Na2SO4 to study the impact of sulfate attack. Such
high Na2SO4 concentration was chosen here to accelerate the degradation process;
however, it may be too high to represent common sulfate exposure levels in the field.
Some physical properties and chemical composition of the clays and cement used
here are summarized in Table 1 and Table 2.
The prepared samples consist of 80.4 % water, 16.1 % cement and 3.5% bentonite
(by weight). This composition is in agreement with other studies in the literature [1, 13,
14]. First, a slurry of bentonite and water was mixed with a high-speed shear mixer for
5 minutes. The slurry was poured in a closed container and allowed to hydrate for
about 24 hours. Subsequently, the cement was added to the 24-hour hydrated bentonite
slurry and mixed with a high-speed shear mixer for about 2 minutes.
Then, the fresh CB slurry was poured in stainless-steel molds to prepare cylindrical
specimens. The bottom and top ends of the mold were sealed with plastic foil and
1004 G. Di Emidio et al. / Hydraulic and Mechanical Behaviour
silicon grease to prevent moisture loss. Then, the samples were allowed to cure for
about 7 days in a conditioned room at 18C. After that period, when the samples
showed enough strength to be handled, they were carefully extruded out of the molds.
Three different sample sizes were produced with a diameter D=100mm and height
H=70mm (for hydraulic conductivity cells modified with bender elements), D=70mm
and H=70mm (for standard hydraulic conductivity tests), D=50mm and H=100mm (for
unconfined compression tests).
Table 1. Physical properties of the treated and untreated bentonite clays.

clay HYPER clay
Liquid Limit [%] 527.15 592.10
Plastic Limit [%] 90.10 171.90
Plasticity Index [%] 437.05 420.20
Swell Index [%] 22 (M) 28
CEC [meq/100g] 50.21 64.96
(M) = Macropores
Table 2. Chemical composition of the cement and of the untreated bentonite clay.
cement clay
SiO2 [%] 29.6 52.8
Al2O3 [%] 8.4 3.44
Fe2O3 [%] 1.2 1.16
MgO [%] 5.6 24.25
CaO [%] 47.8 1.95
Na2O [%] 0.16 3.75
Other 7.24 16.4
3. Methods and procedures
The hydraulic conductivity and small-strain shear modulus of CB samples were studied
in flexible-wall hydraulic conductivity cells provided with bender elements
(Permeabender cells). These cells were provided with a cathetometer to monitor sample
height changes (Fig. 1). In addition to that, replicate hydraulic conductivity tests were
also performed in standard flexible wall permeameters. Furthermore, the unconfined
compressive strength (UCS) of CB samples cured in pure water was monitored through
unconfined compressive tests.
Samples with a diameter of 100 mm and a height of 70 mm were used for
hydraulic conductivity tests in the permeabender cells; samples with a diameter of 70
mm and a height of 70 mm were used for replicate standard hydraulic conductivity
tests and for UCS tests after permeation; samples of 50 mm diameter and 100 mm
height were used for standard UCS tests cured in deionized water.
The hydraulic conductivity k was evaluated out of a falling-head test performed in
a conditioned room at constant temperature (18C) and at an isotropic effective stress
of 30 kPa. The sample was first permeated with deionized water for about 1 month
(1.93 and 0.26 pore volumes of flow for the untreated and treated clay respectively).
Next, the deionized water was replaced with the 25 g/L solution of Na2SO4 and the test
was continued for a period of about 4 months (about 2.70 and 0.30 pore volumes of
flow, respectively for the untreated and treated clay respectively).
Figure 1. Flexible-wall hydraulic conductivity cell provided with bender elements and a cathetometer.
The small-strain shear modulus of the CB samples (starting at 7 days of age) was
evaluated by bender element testing. The bender elements used here are of the type
T220-A4203Y (Piezo Systems, Inc.). The effective bender element length penetrating
in the sample was about 4.5 mm.
The bender element transmitter and receiver, fixed to top plate and base pedestal of
the hydraulic conductivity cell, are vertically aligned. Bender element measurements
were performed on a regular basis from the first day of permeation up to 5 months (the
tests are still ongoing to analyze the long-term behavior).
In bender element testing, G0 is determined out of the propagation velocity (Vs) of
shear waves generated and detected by the transmitter and receiver bender elements
installed on opposite sides of a sample. G0 is estimated as:
G0 = Vs2 (1)
where is the density of the sample. Vs is evaluated as follows:
V s = L / ts (2)
where L is the tip-to-tip distance between the transmitter and receiver bender
elements, and ts is the travel time of the shear waves from the transmitter to the
receiver; ts is evaluated out of the signal recordings. In this research, ts was evaluated
by means of two methods. The first one consists of visually identifying the first direct
arrival from the output signal [e.g. 7, 15, 16]. The second method used here was the
cross-correlation method, first introduced by [7]. The cross-correlation analysis
measures the level of correspondence or interrelationship between two signals of
similar nature and it produces the time shift between them. An example of the recorded
signals is given in figure 2. Both interpretation methods produced similar outcome.
Lastly, the unconfined compressive strength (UCS) of some specimens immersed
in deionized water for 7, 28, 60 and 90 days was determined.
4000
3000
Transmitter signal
2000
Signal amplitude
1000
0
Receiver signal
1000
2000
0 1 2 3 4 5 6 7 8
Time (ns) 5
x 10
Figure 2. Example of bender element signals recorded in the hydraulic conductivity cell
4.1. Hydraulic conductivity
Figure 3 shows the hydraulic conductivity of the cement-bentonite mixtures. As

expected, the hydraulic conductivity to water of both type of mixtures decreased with
time according to the correlation reported by [17]:

(3)
where kr is the hydraulic conductivity at an age of 28 days (tr), t is the age of the
sample in days and is a constant.
After about 30 days of curing, the permeant solution was substituted with a
solution containing sulfates (25 g/L Na2SO4). As shown in Fig. 3, the hydraulic
conductivity of the cement-untreated clay mix increased after permeation with sulfates.
Conversely, the hydraulic conductivity of the cement-bentonite mixtures containing
HYPER clay continued decreasing with time.
A decrease of the hydraulic conductivity of the cement-untreated clay mixture was
noted immediately after contact with sulfates. This behavior was likely due to clogging
of the pores due to gypsum and ettringite formation. When the formation of these
products becomes significant fissures could appear leading to an increase of the
hydraulic conductivity.
4.2. Small-strain shear modulus
Figure 4 shows the small-strain shear modulus of the cement-bentonite mixtures. As

expected, a gradual increase of G0 was observed during the permeation with deionized
water as a consequence of the cement hydration process. The two mixtures showed
similar values of G0.
(a) (b)
Figure 3. Impact of sulfate attack on the hydraulic conductivity of CB mixtures containing untreated clay or
polymer-treated bentonite (HYPER clay) permeated with a 25 g/L Na2SO4 solution. (a) Hydraulic
conductivity vs. time in the permeabender cell. (b) Replicate test results of hydraulic conductivity vs. time in
a flexible wall cell.
After contact with sulfates for a period of about 100 days, no significant changes
were registered even if a decrease of the small-strain shear modulus was expected as a
result of the cement degradation caused by the expansion of gypsum and ettringite, as
observed in a previous study [18]. These tests are still ongoing to monitor the long-term
behavior.
Figure 4. Impact of sulfate attack on the small-strain shear modulus of cement-bentonite mixtures containing
untreated clay or polymer-treated bentonite (HYPER clay) .
4.3. Unconfined compression tests
Figure 5 summarizes all UCS measurements carried out on samples immersed in

deionized water.
Figure 5. Compressive strength versus time of the CB samples (immersed in deionized water) with untreated
clay (Clay) and polymer-treated bentonite (HClay) with different height/diameter (H/D) ratios.
Both types of mixtures showed a similar increase of unconfined compressive

strength with time. These results suggest that the polymer treatment on the clay did not
have negative effects on the strength of CB mixtures yet maintaining a lower hydraulic
conductivity.
5. Conclusions
The hydraulic and mechanical properties of cement bentonite samples are usually
studied separately. In this research the hydraulic conductivity and the small-strain shear
modulus of samples prepared with untreated and polymer-treated clays were
simultaneously measured using a flexible wall permeability cell provided with bender
elements.
Test results showed that the cement-HYPER clay mixtures were less permeable
and less sensible to the microstructural changes caused by the interaction between
cement hydration products and sulfates.
Samples prepared with HYPER clay technology achieved a strength and stiffness
comparable to CB samples containing untreated clay, still maintaining a lower
permeability compared to the cement-untreated bentonite mixtures.
The results and all the considerations about the hydraulic conductivity, small-strain
shear modulus and unconfined compressive strength are referred to a short study period
of about 150 days. Longer monitoring periods are necessary to evaluate the long-term
behavior of CB samples.
References
[1] S.A. Jefferis, Contaminant-grout interaction, Proceedings of the specialty conference grouting, soil
improvement and geosynthetics, New Orleans, 2 (1992), 1393-1402
[2] J.P. Skalny, J. Marchand, I. Odler, Sulfate attack on concrete, Spon Press, Taylor & Francise-Library,
2003.
[3] P.K. Mehta & P.J.M. Monterio. Concrete, 2nd ed., McGraw Hill, 1993
[4] R.A. Bell & J.L. Sisley, Quality control of slurry cut-off wall installation, ASTM STP 1129 (1992), 225-
233.
[5] R.D. Verstegui Flores, G. Di Emidio, W. Van Impe, Small-strain shear modulus and strength increase of
cement-treated clay, Geot. Test. J. 37 (2010), 6271.
[6] D.J. Shirley & L.D. Hampton, Shear-wave measurements in laboratory sediments, J. Acoust. Soc. Am.,
63 (1978), 607613.
[7] G. Viggiani & J.H. Atkinson, Interpretation of bender element tests, Geotechnique 45 (1995), 149-154
[8] E.G. Brignoli, M. gotti, K.H. Stokoe, Measurement of shear waves in laboratory specimens by means of
piez oelectric transducers, Geotech. Testing J. 19 (1996), 384-397
[9] R. Arulnathan, R.W. Boulanger, M.F. Riemer, Analysis of bender element tests. Geotech. Test. J. 21
(1998), 120-131.
[10] J.S. Lee & J.C. Santamarina, Bender elements: Performance and signal interpretation, J . Geotech. Geo-
environ. Eng . 131(2005), 1063-1070.
[11] R.D. Verstegui Flores & G. Di Emidio, Alternative Laboratory Technique to Monitor the Effect of
Cementing Additives on Clay for Cut-off Walls, Geo-Frontiers (2011), Geotech. Special Publication N.
211, 967-975.
[12] G. Di Emidio (2010). Hydraulic and chemico-osmotic efficiency of polymer treated clays. PhD
Dissertation, Ghent University
[13] Ryan C. and Day S. 1986. Performance evaluation of cement-bentonite slurry wall mix design. Proc.
HMCRI Conference, Washington.
[14] Opdyke S.M. and Evans J.C. 2005. Slag-cement-bentonite slurry walls. Journal of Geotechnical and
Geoenvironmental Engineering 131(6), 673681.
[15] Dyvik R. and Madshus C. 1985. Lab measurements of Gmax using bender elements. Proc. ASCE
Annual Convention: Advances in the art of testing soils under cyclic conditions, Detroit, pp. 186196.
[16] Jovii V., Coop M.R. and Simic M. 1996. Objective criteria for determining Gmax from bender
element tests. Gotechnique 46(2), 357-362.
[17] Fratalocchi E., Manassero M., Pasqualini E., Roccato M. and Spanna C. 1996. Predicting hydraulic
conductivity of cement-bentonite slurries. Proc. 2nd Int. Conf. Environmental Geotechnics, Osaka,
Rotterdam, Nederland, Balkema, pp. 503-508.
[18] Verstegui-Flores R. D., Di Emidio G. and Bezuijen A. Hydraulic conductivity and small-strain
stiffness of a cement-bentonite sample exposed to sulphates, Proceedings of the 18th International
Conference on Soil Mechanics and Geotechnical Engineering, Paris.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1010
Monotonic and cyclic behaviour of weakly

cemented sands
Antigoni D. VRANNAa,1 and Theodora TIKA b
a
Aristotle University of Thessaloniki, Thessaloniki, Greece
b
Aristotle University of Thessaloniki, Thessaloniki, Greece
Abstract. This paper presents a laboratory investigation into the undrained

monotonic and cyclic response of weakly cemented sands. To this end, undrained
monotonic compression and cyclic triaxial tests were performed on a clean sand,
with cement contents ranging from 1% to 8% by weight of dry soil. The test
results show that cementation has a significant effect on both monotonic and cyclic
behaviour. At a given density and effective confining stress, increasing
cementation results in the increase of both the monotonic undrained shear strength
and cyclic resistance, as well as the transition of soil behaviour from contractive to
a more dilative one.
Keywords. Cementation, sand, ageing, monotonic, critical state, cyclic resistance
1. Introduction
It is well known that microstructure together with density and stress state are the
dominant factors influencing behaviour of granular soils [1]. Natural sands in many
cases possess a microstructure similar to that of structured clays due to geological
processes that took place during their deposition, such as cementation and subsequent
ageing. In particular, the structure of weakly cemented sands is characterized by the
formation of weak interparticle bonds, due to the presence of, among others, small
amounts of silica, hydrous iron oxides and carbonates in the soil skeleton, as a result of
weathering, chemical depositions, environmental changes, ageing and others.
The effect of cementation has been considered in the study of the behaviour of
naturally cemented sands, in assessing the effectiveness of artificial cementation,
employed in soil improvement and in the study of the behaviour of aged sands, too.
Due to difficulties in sampling and the variability in density and cementation in natural
cemented sands, most of the research on the behaviour of cemented sands has been
conducted by testing artificially cemented specimens, in which agents (cement, gypsum,
lime among others), acting like cementing bonds at the contact points of grains and also
filling the voids, are introduced.
Cementation, induced either by natural, or artificial processes, influences
significantly both monotonic [2, 3, 4, 5] and cyclic [5, 6, 7, 8] sand behaviour. The
present work investigates the monotonic and cyclic shear strength of a weakly
cemented sand by means of undrained monotonic and cyclic tests. The results of the
1
A.D. Vranna and T. Tika / Monotonic and Cyclic Behaviour of Weakly Cemented Sands 1011
investigation are presented and discussed. A comparison is also presented with the
corresponding behaviour of the uncemented sand.
2. Experimental Procedure
2.1. Tested Materials
The soil used in this study is a natural clean quartz sand (M31) with grains of variable
roundness and sphericity. It has a specific gravity of Gs =2.655, maximum and
minimum void ratios of emax = 0.805 and emin = 0.558, respectively, a mean diameter of
D50 = 0.31mm and a uniformity coefficient of Cu = 1.50. Its gradation curve lies within
the bound gradation curves, suggested for liquefiable soils.
White Portland cement (CEM II/A-LL 42.5N) with a specific gravity of 3.103 was
used as the cementing agent for the preparation of weakly cemented sand specimens, at
cement contents of c.c. = 1, 3, 5 and 8% of the weight of dry soil. Its fast gain of
strength allowed the adoption of a curing period of 7 days. It is noted that longer curing
times of 50 days and 1 year tested for specimens with c.c. = 1% showed no strength
improvement. The specific gravity of the cemented specimens was calculated as a
weighted average, based on the proportions of soil and cement in each specimen.
2.2. Testing procedure
Cylindrical specimens (50mm diameter 100mm height) of uncemented and cemented

sand were prepared at various unit weights, using the undercompaction method, as
proposed by Ladd (1978) [9] and described in Vranna (2015) [10]. Cemented
specimens were formed by thoroughly mixing the relative quantities of dry sand and
cement, in order to achieve uniform consistency, at a water content ranging from 5% to
12% and depending on the target density.
Saturation of both clean sand and cemented sand specimens was achieved through
percolation first of carbon dioxide gas (CO2) and afterwards de-aired water. Cemented
specimens were left to cure for 7 days prior to testing, at a controlled temperature and
humidity chamber. In order to achieve full saturation, a series of steps of consecutive
cell and back pressure increments was applied, while maintaining an effective
confining stress of 10kPa, until a Skempton B value greater than 0.95 was obtained.
After the completion of saturation, specimens were isotropically consolidated under an
effective isotropic stress, p0, of either 100kPa or 300kPa. A period of time equal to
double the consolidation time was allowed before shearing.
The testing program consisted of undrained monotonic compression and cyclic
triaxial tests for the study of the monotonic and cyclic response of the tested materials,
respectively. Both types of tests were performed using a closed-loop automatic cyclic
triaxial apparatus (M.T.S. Systems Corporation). In the monotonic tests, specimens
were subjected to undrained compression at a constant strain rate of 0.1%/min. In the
cyclic triaxial tests, a sinusoidally varying axial stress ( d) was applied at a frequency
of f = 0.1Hz, under undrained conditions. In this work, the occurrence of double
amplitude axial strain, DA = 5%, is used as a reference point for the onset of
liquefaction of both cemented and uncemented specimens. For this reason, a series of
cyclic triaxial tests with different cyclic stress ratios, CSR = d / 2p0, was carried out
in order to determine the number of load cycles, N, required for the development of DA
1012 A.D. Vranna and T. Tika / Monotonic and Cyclic Behaviour of Weakly Cemented Sands
= 5%, both for the cemented and uncemented specimens. In view of the typical number
of load cycles of actual earthquakes (10 to 20 for an earthquake of M = 7.5 magnitude),
in this work the onset of liquefaction and thus the cyclic resistance ratio, CRR15, is
considered as the CSR, required to produce DA = 5% in 15 load cycles.
3. Test Results and Discussion
3.1. Monotonic Response
Figure 1 presents plots showing the variation of the deviatoric stress, q = - r, and
the excess pore water pressure, u, with axial strain, and the variation of q with
mean effective stress, p = ( + 2r)/3, for the uncemented and cemented specimens.
It is shown thus that at a given p0, cementation increases undrained shear strength, as
well as the tendency of soil to dilate. In particular, at p0 = 100kPa, an increase in u is
initially observed, followed by a decrease due to the decreasing contractiveness. At p0
= 300kPa, both the uncemented and cemented specimens show contractive behaviour
up to the end of shearing. Thus, increasing p0 increases the contractiveness in the
behaviour of cemented specimens, implying that some yield of cementation bonds
might have taken place during isotropic consolidation.
Figure 2 presents the shear strength envelopes corresponding to the maximum
values of q / p, at two different ranges of density corresponding to a loose and a
medium dense state. At the loose state and for pranging from 25kPa to 280kPa, an
increase in strength of the cemented specimens over that of the uncemented is observed
even at c.c=1%, mainly as a cohesion intercept. This indicates that cementation induces
a cohesion factor by infilling the voids and creating bonds among the sand grains.
However, at the medium dense state and for p ranging from 125kPa to 850kPa, the
cohesion intercept at c.c. = 1% practically diminishes, indicating that cementing bond
structure might gradually have been destroyed at this particular c.c.
Figure 3 shows the critical state lines, CSLs, on the void ratio, e, versus mean
effective stress at critical state, pcs, as well as on the deviatoric stress at critical state,
qcs, versus pcs plane for uncemented and cemented specimens. For the undrained
monotonic triaxial compression tests, it was considered that the onset of critical state
conditions corresponds to the points where the shear and the mean effective stresses, as
well as the pore water pressure, remained practically constant with axial strain.
However the cemented specimens did not always reach well defined critical states,
even when shearing continued at strains of = 30%, in which case the end points of
the tests were considered and are shown in Figure 3. As shown in the above Figure, the
CSLs of weakly cemented sands lay above the CSL of the uncemented sand and they
converge at pcs above 600kPa approximately.
3.2. Cyclic Response
Figure 4 presents the cyclic response of the uncemented and cemented specimens and
in particular their stress paths and the evolution of u / p0 and DA with time, at a given
density and CSR. It is shown that a significantly larger number of N is required for the
onset of liquefaction of the weakly cemented sand. This difference in behaviour is also
depicted in the excess pore water pressure development. For the particular density and
CSR, it is indicated that cementation leads to a transition of soil behaviour from flow
type liquefaction into cyclic mobility.
In a similar manner, Figure 5 presents the evolution of u / p0 and DA with time,
for cemented specimens with c.c. = 3% and 5%, at a given density and CSR. Whereas
the N required for the cemented specimen with c.c. = 3% to reach DA = 5% was N =
10.8, the specimen with c.c. = 5% was considered non-liquefiable, since it reached a
maximum value of DA = 0.28% at N = 1033.
800 800
p0 = 100kPa (a) p0 = 300kPa (d)
d = 14.96 - 15.25kN/m3 d = 14.92 - 15.16kN/m3
600 600 e = 0.732 - 0.748
Dr = 23 - 30%
q (kPa)
400 q (kPa) 400
200 Dr = 24 - 34% 200

e = 0.722 - 0.746
(q / p)max (q / p)max
0 0
0 10 20 30 40 0 10 20 30 40
(%) (%)
300 300
c.c. = 0% (b) (e)
c.c. = 1%
200 200
c.c. = 3%
c.c. = 5%
u (kPa)
u (kPa)
100 100
0 0 c.c. = 0%
c.c. = 1%
-100 -100 c.c. = 3%
c.c. = 5%
-200 -200
0 10 20 30 40 0 10 20 30 40
(%) (%)
800 800
(c) (f)
600 600
q (kPa)
q (kPa)
(q / p)max
400 400
(q / p)max
200 200
0 0
0 100 200 300 400 500 0 100 200 300 400 500
p' (kPa) p' (kPa)
Figure 1. Effect of cement content on the undrained monotonic response of uncemetend and cemented
specimens, at (a, b, c) p0 = 100kPa and (d, e, f) p0 = 300kPa.
1000 1200
(a) (b)
c.c. = 0% e = 0.647 - 0.699
800 c.c. = 1% Dr 43 - 64%
c.c. = 3% d = 15.45-15.97kN/m3
800
c.c. = 5%
600
q (kPa)
q (kPa)
c.c. = 8%
c.c. = 0%
400
400 c.c. = 1%
c.c. = 3%
200 d = 14.86 - 15.37kN/m3 c.c. = 5%
e = 0.717 - 0.753
c.c. = 8%
Dr 21 - 36%
0 0
0 200 400 600 0 200 400 600 800 1000
p (kPa) p (kPa)
Figure 2. Variation of q with p at (q / p)max, for uncemented and cemented specimens, at (a) a loose state
and (b) a moderately dense state.
(a) (b)
emax = 0.805
0.8 1200
900
qcs (kPa)
ecs
0.7
600 c.c.
0% : M = 1.224
1% : M = 1.265
300 3% : M = 1.316
0.6 5% : M = 1.313
emin = 0.558 8% : M = 1.387
0
10 100 1000 10000 0 300 600 900 1200
p'cs (kPa) p'cs (kPa)
Figure 3. Critical state lines on (a) ecs - pcs and (b) qcs - pcs planes of uncemented and cemented sand.
Figure 6 presents the variation of CSR with N, at different DA levels and at a loose
and a moderately dense state. It is shown that at a given value of CSR, even at the
smallest c.c. value of 1%, the number of N that the weakly cemented sand can sustain
is considerably larger than the corresponding of uncemented sand. This difference in
the number of N is reduced with increasing density.
Figure 7 presents the effect of cementation on the variation of CRR 15 with e and
state parameter, = e - ecs, for uncemented and cemented specimens. It is shown that at
a given e and p0 level, cemented specimens with c.c. = 1% and 3% have a significantly
higher CRR15 than that of the uncemented specimens. The CRR15 for specimens with
c.c.= 3% is 1.8 to 3.2 times higher than that of the uncemented sand specimens. The
same observation also holds for the variation of CRR15 with .
200 200
(a) DA = 1% (b) DA = 1%
DA = 2.5% L DA = 2.5%
SL CS
C DA = 5% DA = 5%
100 L 100
PT
q (kPa)
q (kPa)
0 0
c.c. = 0% c.c. = 1%
-100 -100
CSR 0.29 CSR 0.27
e = 0.668 e = 0.673
Dr = 56% d = 15.62kN/m3 Dr = 54% d = 15.60kN/m3
-200 -200
0 40 80 120 160 200 0 40 80 120 160 200
p (kPa) p (kPa)
1
CSR 0.28 (c) CSR 0.28 (d)
2
d = 15.60-15.62kN/m3 d = 15.60-15.62kN/m3
0.8 Dr = 54 - 56%
c.c. = 1%
0
c.c. = 0%
0.6
u / p0
DA (%)
c.c. = 1%
-2
0.4
-4 c.c. = 0%
0.2 DA = 1%
DA = 2.5%
-6
DA = 5%
0
0 200 400 600 800 1000 0 200 400 600 800 1000
t (s) t (s)
Figure 4. Comparison of q - p plots for (a) uncemented and (b) cemented specimens with c.c. = 1%, at a
given density and CSR and (c) evolution of u / p0 and (d) DA with time for the above soils.
1
CSR 0.42 (a) CSR 0.42 (b)
2
d = 15.02-15.25kN/m3 d = 15.02-15.25kN/m3
0.8
c.c. = 5%
0
c.c. = 3%
0.6
u / p0
DA (%)
c.c. = 5%
-2
0.4
-4 c.c. = 3%
0.2 DA = 1%
e = 0.722 - 0.743
DA = 2.5% Dr = 25 - 34%
-6
DA = 5%
0
0 200 400 600 800 1000 0 200 400 600 800 1000
t (s) t (s)
Figure 5. Comparison of the evolution of (a) u / p0 and (b) DA with time, for cemented specimens with c.c.
= 3% and 5%, at a given density and CSR.
0.6 (a) p0 = 100kPa 0.6 (b) p0 = 100kPa

CSR = d / 2p0
CSR = d / 2p0
0.4 0.4
d = 14.87 15.09kN/m3
0.2 0.2
Dr 44 56%
Dr 22 28% e = 0.668 0.696
e = 0.735 0.750 d = 15.38 15.66kN/m3
0 0
0.1 1 10 100 1000 0.1 1 10 100 1000
number of cycles, N number of cycles, N
c.c. (%)
0 1 3
1
DA (%) 2.5
5
Figure 6. Variation of CSR with number of cycles, N at (a) a loose state and (b) a moderately dense state at
p0 = 100kPa for uncemented and cemented specimens.
0.6 0.6
p0 = 100kPa (a) p0 = 100kPa (b)
0.4 0.4
CRR15
CRR 15
0.2 0.2
c.c. = 0% c.c. = 0%
c.c. = 1% c.c. = 1%
c.c. = 3% c.c. = 3%
0.0 0.0
0.66 0.69 0.72 0.75 0.78 -0.12 -0.09 -0.06 -0.03 0.00
e = e - ecs
Figure 7. Variation of CRR15 with (a) void ratio, e and (b) state parameter, for uncemented and cemented
specimens at p0 = 100kPa.
4. CONCLUSIONS
Cementation has a significant influence on both the undrained monotonic and cyclic
response of the tested sand. The following conclusions can be drawn from the work
presented:
i. An increase in the cementation results also in an increase of the undrained shear

strength and leads to a decrease in the contractiveness in soil behaviour. For a given
c.c. level, the contractiveness is increased with increasing p0. Even very weak
cementation at c.c. = 1%, may induce a cohesion intercept in the shear strength
envelope at a loose state. Critical state lines for the uncemented and the weakly
cemented sand with c.c. ranging from 1% to 8%, coincide at p'cs greater than
600kPa, approximately. At smaller p'cs, the CSLs of the weakly cemented sands lay
above the CSL of the uncemented sand.
ii. At a given e, p0 and CSR level, weakly cemented sands can sustain a significantly
larger number of N before reaching liquefaction, as compared to the uncemented
sand. This difference in behaviour is also reflected in the excess pore water pressure
development. Cementation may lead to a transition of soil behaviour from flow type
liquefaction into cyclic mobility. At a given e and p0 level, increasing cementation
results in higher values of CRR15.
iii. The above presented results for the weakly cemented specimens at c.c. = 1% may
be considered as indicating the behaviour of naturally aged soils, in which ageing
may be partly attributed to the chemical bonding between soil particles. It is thus
shown that ageing of sand deposits has a pronounced effect on both their monotonic
and cyclic behaviour.
References
[1] S. Leroueil & P.R. Vaughan, The general and congruent effects of structure in natural soils and weak
rocks, Gotechnique 40 (1990), 3: 467488.
[2] D.W. Airey, Triaxial testing of naturally cemented carbonate soil, J. Geotech. Eng. 119 (1993), 9: 1379
1398.
[3] G. Clough, J. Sitar, R.C. Bachus & N.S. Rad, Cemented sands under static loading, J. Geotech. Eng. 107
(1981), 6: 799817.
[4] M.R. Coop & J.H. Atkinson, The mechanics of cemented carbonate sands, Gotechnique 43 (1993), 1:
5367.
[5] A.D. Vranna, Th. Tika & G. Koninis, The liquefaction resistance of weakly cemented sands,
Geomechanics from Micro to Macro (2015), 12251229.
[6] G. Clough, J. Iwabuchi, N.S. Rad & T. Kuppusamy, Influence of cementation on liquefaction of sands, J.
Geotech. Eng. 115 (1989), 8: 11021117.
[7] S.K. Saxena, K.R. Reddy & A.S. Avramidis, Liquefaction resistance of artificially cemented sand, J.
Geotech. Eng. 114 (1988), 12: 13951413.
[8] S. Sharma & M. Fahey, Evaluation of cyclic shear strength of two cemented calcareous soils, J. Geotech.
Geoenviron. Eng. 129 (2003), 7: 608618.
[9] R.S. Ladd, Preparing test specimens using undercompaction, Geotech. Test. J. 1 (1978), 1: 1623.
[10] A.D. Vranna, Monotonic and cyclic behaviour of improved liquefiable soils, PhD thesis Aristotle
University, Thessaloniki, Greece, 2015 ( in progress).
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1018
Deformation characteristics of grain-

displacing GH-bearing sediments
Joo Yong LEE a,1 Jae Hyoung LEE a, Taewoong AHN a and SeJoon KIM a
a
Petroleum and Marine Resources Division, KIGAM, Republic of Korea
Abstract. Gas hydrates are widespread, occurring in both permafrost sediments

and deep marine sediments on the worlds continental margins where the pressure
and temperature conditions are within the gas hydrate equilibrium conditions. The
large amounts of gas hydrate reserves suggest the potential of gas hydrates as an
energy resource if economically viable production methods were developed. The
proper characterization of the physical property and the behavior of gas hydrate-
bearing sediments during gas production from gas hydrate deposits are crucial for
the efficient and safe gas production from gas hydrate deposits. Pore-filling
hydrate-bearing sands are of particular interest as energy resources, so the
deformation characteristics of pore-filling hydrate-bearing sediments have
extensively studied. However, the deformation characteristics in grain-displacing
hydrate-bearing sediments have not been thoroughly explored so far. The
deformation characteristics in grain-displacing hydrate-bearing sediments are
explored in this study using the properties of natural gas hydrate-bearing sediments
from the Ulleung Basin, East Sea, Korea. The one-dimensional consolidation tests
on these sediments were performed and the soil index properties were measured
after the tests. The parametric study on the behavior revealed that the magnitude of
deformation during consolidation is mostly governed by the compression index of
sediments, gas hydrate saturation, and morphology of hydrates.
Keywords. hydrates, deformation, grain-displacing hydrates
1. Introduction
Gas hydrates are widespread, occurring in both permafrost sediments and deep marine
sediments on the worlds continental margins where the pressure and temperature
conditions meet the gas hydrate equilibrium conditions. The large amounts of gas
hydrate reserves suggest the potential of gas hydrates as an energy resource if
economically viable production methods were developed. The proper characterization
of physical property and behavior of gas hydrate-bearing sediments during gas
production from gas hydrate deposits are crucial for the efficient and safe gas
production from gas hydrate deposits. Gas hydrate dissociation in sediments affects the
well bore stability, causes the geohazards such as landslides and subsidences [1,2]
Gas production methods from gas hydrates deviate from the conventional gas
production since the gas hydrate production involves phase transformations from solid
phase to gas and liquid phases. These phase changes produce uncertainties such as
permeability changes due to vanishing hydrates and subsequent settlements and
1
Principal Researcher. Korea Institute of Geosciences and Mineral Resources, 92 Gwahang-no,
Yuseong-gu, Daejeon, Korea; E-mail: [email protected]
J.Y. Lee et al. / Deformation Characteristics of Grain-Displacing GH-Bearing Sediments 1019
sediment softening and volume contraction due to loss of hydrate bonding [3]. These
uncertainties are mostly related safety issues during production which should not be
ignored during establishing production strategies. Numerous geomechancal laboratory
tests using synthetic gas hydrate-bearing sediment have been reported so far [3-6], but
only a few laboratory tests on in-situ gas hydrate-bearing sediments have been reported
[7-8] since sampling hydrate-bearing specimen without hydrate dissociation requires
expensive and sophisticated sampling procedures.
Pore-filling hydrate-bearing sands are of particular interest as energy resources, so
the deformation characteristics of pore-filling hydrate-bearing sediments have
extensively studied. However, the deformation characteristics in grain-displacing
hydrate-bearing sediments have not been thoroughly explored so far. The deformation
characteristics in grain-displacing hydrate-bearing sediments are experimentally
explored in this study using natural gas hydrate-bearing sediments from the Ulleung
Basin, East Sea, Korea.
2. Experimental methods
A pressure coring system enables a core tube to remain sealed at near in-situ pressure
conditions during the entire recovery process. Both the FPC and FRPC pressure coring
systems were used in UBGH2 (Ulleung Basin Gas Hydrate expedition 2). FPC and
FRPC were designed to retain a pressure of up to 25 MPa. FPC is a hydraulically
driven push sampler while FRPC is a rotary corer using a water-driven rotary motor.
The core diameters are 56 mm for FPC and 51 mm for FRPC. While remaining at
pressure, successfully recovered pressure corer were immediately transferred to an ice
bath where they sat for a minimum of 30 min to equilibrate in temperature to
approximately 4~6C. The pressurized corer autoclaves were then transferred to the
PCATS laboratory for non-destructive property scans. After all non-destructive tests
were completed, pressure core specimens were cut and transferred to GHOBS (Gas
Hydrate Ocean Bottom Simulator) cells for onshore analyses.
The GHOBS system is the experimental device that can accommodate pressure
core samples sub-sampled from pressure cores under pressure. GHOBS can
characterize the physical properties of GH-bearing sediments and the behavior of GH-
bearing sediments during GH production under different stress conditions, hydrate
saturations, and sediment types and structures. The device can apply 20 MPa of fluid
pressure and 5MPa of vertical effective stress, which are the typical ranges of stress
condition in deep sea sediments. The cell body is built with aluminum alloy so that the
specimen can be scanned by X-ray CT system. The top and bottom plate of the cell
houses sensors for vertical strain, elastic-wave velocity, electrical resistivity, pressure,
and temperature (Figure 1). Cell temperature is manually controlled with cooler and
pressures including pore-fluid pressure and vertical effective stress are controlled by
programmable syringe pumps. Wet test meter is used for recovery rate measurement
during production test.
The pressure core samples have been stored in pressure storage chamber onboard
and cut and transferred to GHOBS cells under in-situ fluid pressure condition. The in-
situ vertical effective stress condition is retained by consolidating a specimen to in-situ
vertical effective stress in three incremental loading steps. After the specimen is fully
1020 J.Y. Lee et al. / Deformation Characteristics of Grain-Displacing GH-Bearing Sediments
consolidated in the in-situ vertical effective stress, depressurization is applied to a

specimen.
Figure 1. Schematic diagram of GHOBS system.
3. Theoretical Models
The volume change of GH-bearing sediments due to GH dissociation is a largely

unstudied phenomenon and affects borehole stability, foundation settlement and ground
subsidence around a production well. Four causes of volume change during hydrate
dissociation can be identified [3]: (1) volume change from pore GH dissociation at
constant effective stress, thaw, (2) volume change from the vanishing of the hydrate
lenses in fine-grained sediments, lens, (3) volume change associated with a production
strategy, ps, and (4) volume change if the sediment is subjected to marked deviatoric
loading, q. In this study, three causes, thaw, lens, and ps are explored. The
sediment compressibility is characterized by the compression index Cc.
The settlement due to GH dissociation is highly dependent on Cc of sediments and the
compression index, Cc can be utilized to assess the settlement due to these factors as
below [3].
s1.4
th a w
h yd
3
Cc
13Cc
(1)
lens vlens (2)

Cc ' v u
ps log( ) (3)
1 e0 'v
where is a constant, Vlens is the volume fraction of GH lens, u is the amount of
depressurization. The constant is the function of the morphology of lenses.
The experimental results from previous studies are utilized to apply the theoretical
models for further analyses and model verifications. Experimental data from natural
GH-bearing sediments are obtained from lab-scale production experiment on natural
GH-bearing sediments recovered from the Ulleung Basin, East Sea, Korea, during the
Ulleung Basin Gas Hydrate drilling expedition 2 (UBGH2), using GHOBS (Gas
Hydrate Ocean Bottom Simulator) apparatus. The natural samples include three GH
lens-bearing fine-grained sediments from the same borehole at different depths, that
have distinctively different angles of GH lenses (Figure 2). Lenses from the specimen
A are more or less horizontal, while those from sample C are mostly vertical. Specimen
A, B, and C are sampled from 45, 60, and 90 mbsf respectively, at the UBGH 2-3 site.
Detailed discussions on non-destructive scan results are in [9].
In estimating volume change for GH lens-bearing specimens, lens and ps were
taken into account. thaw was excluded from estimation, since the amount of pore-
filling hydrates are usually insignificant in fine-grained sediments, especially in the
presence of grain-displacing hydrates nearby [9].
In general, the modeled results successfully replicated the experimental data,
suggesting that the simple models suggested by Lee et al. (2010) can be used as a tool
for the first approximations before starting comprehensive and sophisticated numerical
simulation prediction. The underestimation of volume change for sample c is due to the
vertical lenses presented in the sediments. In equation (2), is the constant that is the
function of GH morphologies, as mentioned earlier. In this estimation, was assumed
to be one for all specimens. However, it is expected that 1 for horizontal lenses, and
< 1 for vertical lenses [9]. The range of constant could be obtained from extensive
experimental data set on grain-displacing GH-bearing sediments. Appropriate values of
also could be inferred from numerical modeling that takes into consideration the
constitutive parameters of the surrounding sediment.
1022 J.Y. Lee et al. / Deformation Characteristics of Grain-Displacing GH-Bearing Sediments
Cc [ ]
0
0.01 0.1 1 10
-0.05
-0.1
A
v [ ] -0.15
-0.2
-0.25 B
-0.3
-0.35
C
-0.4
-0.45
-0.5
Figure 2. Model fittings to experimental results.
5. Conclusion
The one-dimensional consolidation tests on these sediments were performed and the
soil index properties were measured after the tests. The parametric study on the
behavior revealed that the magnitude of deformation during consolidation is mostly
governed by the compression index of sediments, gas hydrate saturation, and
morphology of hydrates.
References
[1] Collett, T.S., Dallimore, S.R., Detailed analysis of gas hydrate induced drilling and production hazards.
In: Paper Presented at 4th International Conference on Gas Hydrates, 19-23 May 2002, Yokohama,
Japan, pp. 47-52.
[2] Fulong, N., Yibing, Y., Signe, K., Thijs, J.H.V., Kirill, G., Mechanical properties of clathrate hydrates:
status and perspectives. Energy Environ. Sci. 5, (2012), 6779-6795.
[3] Lee, J.Y., Santamarina, J.C., Ruppel, C., Volume change associated with formation and dissociation of
hydrate in sediment, Geochem. Geophy. Geosy. 11, 3 (2010), doi:10.1029/2009GC002667.
[4] Hyodo, M., Yoneda, J., Yoshimoto, N., Nakata, Y., Mechanical and dissociation properties of methane
hydrate-bearing sand in deep seabed. Soils Found. 53, 2, (2013), 299-314.
[5] Yun, T.S., Santamarina, J.C., Ruppel, C., Mechanical properties of sand, silt, and clay containing
tetrahydrofuran hydrate. J. Geophys. Res. 112, B04106 (2007) http://
dx.doi.org/10.1029/2006JB004484.
[6] Miyazaki, K., Masui, A., Sakamoto, Y., Aoki, K., Tenma, N., Yamaguchi, T., Triaxial compressive
properties of artificial methane-hydrate-bearing sediment. J. Geophys. Res. 116, B06102. (2011)
http://dx.doi.org/10.1029/2010JB008049.
[7] Masui, A., Haneda, H., Ogata, Y., Aoki, K., Mechanical properties of sandy sediment containing marine
gas hydrates in deep sea offshore Japan. In: Proc.7th ISOPE-OMS 2007, Lisbon. International Society
of Offshore and Polar Engineers, 53-56.
[8] Yoneda, J., Hyodo, M., Nakata, Y., Yoshimoto, N., Triaxial shear characteristics of methane hydrate-
bearing sediment in the deep seabed. J. Geotech. Eng. Jpn. C. 66, 4, (2010) 742-756.
[9] Lee, J.Y., Jung, J.W., Lee, M.H., Bahk, J.J., Choi, J.Y., Kim, J.H., Pressure core based study on gas
hydrate occurrences in the Ulleung Basin and implication to geomechanical controlling factors., Mar.
Pet. Geol. 47, (2013) 85-98.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1024
A thermo-hydro-mechanical constitutive
model for saturated frozen soils
S. A. Ghoreishian Amiri1; M. Kadivar and G. Grimstad
Norwegian University of Science and Technology (NTNU), Trondheim, Norway
Abstract. In general, a saturated frozen soil sample, as a composite material,

consists of soil grains, ice and unfrozen water. The mechanical behavior of this kind
of geomaterial is strongly affected by the amount of ice that fills the pore spaces.
On the other hand, the amount of ice in a frozen soil sample depends on temperature
and applied mechanical stresses. This coupled behavior is the main characteristic by
which frozen soils behave differently compared to unfrozen samples. In this paper,
by dividing the total stress into fluid pressure and solid phase stress, in addition to
considering the cryogenic suction as an independent stress-state variable, an
elastoplastic constitutive model for saturated frozen soils is proposed. When all
water is unfrozen, the model will become a conventional critical state model. The
coupling effect of ice content on the mechanical behavior, as well as the coupling
effect in the reverse direction are taken into account in the proposed framework. The
effect of temperature on the mechanical behavior is also considered in the model.
The proposed model is able to represent many fundamental features of frozen soils
such as frost heave and strength weakening due to pressure melting. The typical
predictions of the model for simulating the characteristic trends of the frozen soil
behavior is described qualitatively. Model predictions are also compared with the
available test results and reasonable agreement is achieved.
Keywords. Frozen soil, Elastoplastic modeling, Coupled behavior, Pressure melting,

Frost heave
1. Introduction
With the increase in engineering activities in permafrost and seasonally frozen regions,
it is essential to study the mechanical behavior of frozen soils, especially when
developing a reasonable constitutive model. Two different kinds of approaches are
commonly used to simulate the mechanical behavior of frozen soils: total stress based
models and two stress-state variables approach.
Total stress models have been widely used in the literature to describe the
mechanical behavior of frozen soils[e.g. 1-6]. However, it seems that they only tended
to accentuate the influence of confining pressure on the elastoplastic behavior, placing
less emphasis on the influence of other very important parameters such as temperature
and ice content. In other words, these models cannot simulate deformations under the
variation of ice content and/or temperature during a freezing or thawing period.
Moreover, working with total stress, description of soil behavior in the presence of
unfrozen water will face some significant difficulties which are not described in the
models.
1
Corresponding Author: [email protected]
S.A. Ghoreishian Amiri et al. / A Thermo-Hydro-Mechanical Constitutive Model 1025
Recently, Nishimura et al. [7] proposed a two stress-state variables model for
simulating the behavior of frozen soils. They actually tried to adopt the Barcelona Basic
Model (BBM) [8] for the frozen state of saturated soils. The cryogenic suction has been
employed instead of matric suction, and the excess of total stress over ice pressure as the
net stress variable of the BBM. Moreover, the Suction Increase (SI) yield curve has been
ignored in this model. This approach is also employed by Shastri and Sanchez [9] to
show its ability for considering the influence of cryogenic suction and temperature on
the stress-strain behavior of frozen soils. In this model, ice segregation phenomenon is
captured with a tensile failure mechanism. However, after an experience of tensile failure
during a segregation phenomenon, the model will always predict dilative behavior.
In this paper, we try to cover the shortcomings of the model proposed by Nishimura
et al. [7]. In this way, the solid phase stress is introduced and the Suction Increase (SI)
yield curve is added to the model. Moreover, coupled hardening rules are also introduced
for considering the coupling effects of deformations due to mechanical loading and
suction variation.
2. Model Formulation
In this framework, the soil is assumed to be a saturated medium within which water could
be frozen or unfrozen. So, the material is considered as a composite, consists of solid
grains, ice and unfrozen water. In another point of view, this composite could be
decomposed into the solid phase (contains soil particles and ice crystals) and water phase.
The part of total stress carried by the solid phase is considered as the part which is
responsible for the deformation of the soil due to any mechanical loading. The solid
phase stress, *, could be defines as:
* sw pw I (1)
where is the total stress, sw and pw are the water saturation and pressure, respectively.
The solid phase stress and cryogenic suction are proposed as the stress state variables in
this framework. The cryogenic suction, Sc, is defined as:
Sc pi pw (2)
where pi is the ice pressure.

In this framework, it is assumed that any strain increment, d, can be additively
decomposed into the following parts:
d d me d se d mp d sp (3)
where d# me and d#mp are the elastic and plastic parts of strain due to solid phase stress
variation and d# and d# are the elastic and plastic parts of strain due to suction
variation, respectively. The equivalent elastic modulus, E, and Poisson ratio, , of the
solid phase can be written as a function of ice content, i [10]:
1026 S.A. Ghoreishian Amiri et al. / A Thermo-Hydro-Mechanical Constitutive Model
E
D s Es (1 2 i ) i Ei (1 2 s )ED s Es (1 i ) i Ei (1 s )E (4)
s Es (1 i )(1 2 i ) i Ei (1 s )(1 2 s )
s E s s (1 i )(1 2 i ) i Ei i (1 s )(1 2 s )
(5)
s E s (1 i )(1 2 i ) i Ei (1 s )(1 2 s )
where the subscriptions s and i stand for the soil and ice phases, respectively. In addition,
elastic modulus of ice, Ei, is considered as a function of temperature, T:
Ei Eiref Eiinc T Tref

(6)
where Ei-ref is the value of elastic modulus at the reference temperature, Tref, and Ei-inc is
a material parameter.
The elastic part of strain due to suction variation is computed as:
1s dS c
d se - I (7)
3(1 e ) ( S c patm )
where s is the expansion coefficient for changes in suction within the elastic region, e
is the void ratio and patm is the atmospheric pressure. Noting that an increase in suction
(due to increase of ice content) causes the soil to expand, a negative sign is required in
equation (7).
Based on the BBM [8], the complete yield surfaces can be described with the
following equations:
2 2
# p* k S $ q* # p k S $
*
F1 p* % y t c & 2 % y t c & 0. (8)

% & % &
' 2 ( M ' 2 (
F2 Sc S seg 0. (9)
00 1
# p*y $ 0 1
p*y pc* % *0 & (10)
% pc &
' (
0 00 D(1 r ) exp( F Sc ) r E (11)
where p* is the solid phase mean stress, kt is the parameter that controls the increase in
cohesion with suction, M is the slope of the critical state line, Sseg is the initial position
of the SI yield curve, py0* is the preconsolidation stress for unfrozen condition, pc* is the
reference stress, 0 is the compressibility coefficient for the unfrozen state along virgin
loading, is the compressibility coefficient along elastic stress paths [= (1+e)Op*/K] ,
and r are the parameters that control the increase in stiffness with suction. Figure 1 shows
the complete yield surfaces of the model in p*-q*-Sc space.
Figure 2 shows the typical evolution of the LC and SI yield curves in p*-Sc plane.
Plastic compression due to solid phase stress variation (yielding on LC) causes the LC
yield curve to move outward. In addition, there is a coupling effect on the SI yield curve.
Plastic compression results in a decrease in the dimensions of voids, so lower value of
suction (or ice content) is needed for ice segregation phenomenon (Fig. 2a). Therefore,
the SI yield curve should move downward. Similarly, due to ice segregation during a
freezing period, yielding will happen on the SI curve and it should move upward. Plastic
expansion due to this movement causes the soil strength to decrease, therefore the LC
yield curve should move inward (Fig. 2b). Following this line of discussion, the coupled
hardening rules are proposed as:
dp*y0 1 e 1 e
d vmp d vsp (12)
p*y0 00 1 0s 1 s
dS seg 1 e 1 e
d vsp d vmp (13)
S seg pat 0s 1 s 00 1
where dvmp and dvsp stand for the plastic volumetric strain due to the variation of the
solid phase stress and cryogenic suction, respectively.
Figure 1. Complete yield surfaces in p*-q*-Sc space
Figure 2. Movement of LC and SI curves; a)due to mechanical plastic compaction; and b) due to plastic
expansion during freezing period
Applying the associated flow rule, the plastic parts of strain is computed as:
AF1
d mp d 01 (14)
A*
AF2
d sp d 02 I (15)
ASc
3. Model Results
In order to examine the model ability to simulate the behavior of frozen soils in an
acceptable way, some triaxial compression tests in different temperatures and confining
pressures are simulated and compared in this section. Moreover, since experimental data
is still limited, model predictions for ice segregation phenomenon and also for strength
weakening due to pressure melting are presented and discussed at a qualitative level.
3.1. Triaxial tests under different temperatures
A series of triaxial compression tests on frozen sands at constant confining pressure (1

MPa), constant strain rate (1.67E10-4 s-1) and different temperatures of -1, -2 and -5o C
have been presented by Xu [6]. Details of the tests can be found in the original reference
[6]. Table 1 shows the value of the parameters for these tests and the comparison results
are shown in figure 3. As shown in the figure, the proposed model predicts reasonable
results when comparing with the test data.
Table 1. Model parameters for triaxial tests under different temperatures
pc* (MPa) 00 F (MPa 1 ) r kt M 0s 1s

0.04 4.5 0.11 0.69 0.25 1.52 0.4 0.008
p *y0 (MPa) Sseg (MPa) Es (MPa) s Eiref (MPa) Eiinc (MPa/K) i

5.35 15 20 0.2 600 280 0.31
Figure 3. Comparison between the measured and predicted results for triaxial tests under different
temperatures
3.2. Triaxial tests under different confining pressures
Triaxial compression tests at different confining pressures of 0.3, 0.6, 0.8 and 1 MPa are
also reported by Xu [6]. These tests were conducted at constant temperature of -4oC and
constant strain rate of 1.67E10-4 s-1. Model parameters for this set of tests are introduced
in table 2. It should be noted, the type of soil used for these tests is different from the one
in the previous section. Figure 4 shows the results of the model in compare with the
experiments. As shown in the figure, the proposed model can successfully follow the
trend of the measurements.
3.3. Ice segregation
During a freezing period, the ice content will increase. At lower ice content, soil strength
rises with increasing ice content. When the ice content approaches the saturated
condition, the strength will reach its peak value. Then due to ice segregation phenomenon
the strength of the soil will decrease with increasing ice content and significant expansion
will occur in the sample. This behavior can be captured in the model by the use of SI
yield curve. In this section, model prediction along a freezing period with temperature
variation of 0 to -4oC is presented. The model parameters for this test are selected same
as those presented in table 2, except for the segregation suction, Sseg, which is considered
to be 1 MPa. Figure 5 shows the model results for this simulation. As shown in the figure,
in the earlier stage of freezing, the volume change in the sample is negligible, but after
that the cryogenic suction reaches the SI yield curve and ice segregation phenomenon is
occurred and significant amount of deformation is seen. Although this is a qualitative
result which is not compared with laboratory data, it can show that the framework is
potentially able to predict the frost heave phenomenon during a freezing period.
Table 2. Model parameters for triaxial tests under different confining pressures
pc* (MPa) 00 F (MPa 1 ) r kt M 0s 1s

0.07 0.7 0.17 0.49 0.09 1.22 0.4 0.008
p *y0 (MPa) Sseg (MPa) Es (MPa) s Eiref (MPa) Eiinc (MPa/K) i

1.55 5 5 0.3 200 20 0.31
Figure 4. Comparison between the measured and predicted results for triaxial tests under different confining
pressures
Figure 5. Volume change during a freezing period
3.4. Pressure melting
Behavior of frozen soils under increasing confining pressure is considered to be the result
of combined mechanical and thermodynamic effects. Mechanical effect results in an
increase in the soil strength, however, thermodynamic effect results in strength
weakening of the soil, which only appears at higher confining pressures. The main reason
of strength weakening of frozen soils under high confining pressure is the pressure
melting of porous ice. The proposed model is also able to capture this phenomenon. Due
to a pressure melting, the value of water saturation and pressure will rise up, so there is
a decrease in the value of solid phase stress and soil strength.
In order to show the model ability to simulate this behavior, an imaginary triaxial
test with 5 MPa hydrostatic pressure and at -1oC is considered. The model parameters
for this sample is also selected same as those presented in table 2, except for the strain
hardening parameter, py0*, which is considered to be 5.85 MPa. Figure 6 shows the model
results for this simulation. As shown in the figure, until a certain value of confining
pressure, shear strength increases with confining pressure, but after that, when the
pressure melting is occurred, soil strength will decrease with increasing confining
pressure.
Figure 6. Model results for pressure melting simulation

4. Conclusion
In this paper, an elastoplastic framework for describing the stress-strain behavior of

frozen soils is presented. The model is proposed in the framework of two stress-state
variables. The employed stress variables are solid phase stress and cryogenic suction.
The model is able to represent many of fundamental features of the behavior of frozen
soils such as strength weakening due to pressure melting and frost heave due to ice
segregation phenomena. Influence of temperature and ice content variations on the
behavior of the soil are implicitly considered in the model by the use of cryogenic suction
as an independent stress-state variable. Effect of unfrozen water is also considered in the
model by introducing the solid phase stress as the second stress-state variable. The
proposed model is applied for simulating some triaxial tests in different temperatures and
confining pressures and reasonable agreement is achieved. In order to show the ability
of the model to simulate the strength weakening due to pressure melting and frost heave
due to ice segregation, typical predictions of the model are described qualitatively.
Acknowledgments
This work is carried out under the Marie Curie Action, European FP7, CREEP project,
Grant number 17 PIAG-GA-2011-286397.
References
[1] L.U. Arenson, S.M. Springman, Mathematical descriptions for the behaviour of ice-rich frozen soils at
temperatures close to 00 C, Canadian Geotechnical Journal 42 (2005), 431442.
[2] Y. Lai, et al., Strength distributions of warm frozen clay and its stochastic damage constitutive model,
Cold Regions Science and Technology, 53 (2008), 200215.
[3] Y. Lai, L. Jin, X. Chang, Yield criterion and elasto-plastic damage constitutive model for frozen sandy
soil, International Journal of Plasticity, 25 (2009), 11771205.
[4] Y. Lai, et al., Strength criterion and elastoplastic constitutive model of frozen silt in generalized plastic
mechanics, International Journal of Plasticity, 26 (2010), 14611484.
[5] Z. Zhu, J. Ning, M. Ma, A constitutive model of frozen soil with damage and numerical simulation for
the coupled problem, Science China Physics, Mechanics and Astronomy, 53 (2010), 699-711.
[6] G. Xu, Hypoplastic constitutive models for frozen soil, PhD dissertation, University of Natural Resources
and Life Sciences Vienna, 2014.
[7] S. Nishimura, et al., THM-coupled finite element analysis of frozen soil: formulation and application,
Gotechnique, 59 (2009), 159-171.
[8] E. Alonso, A. Gens, A. Josa, A constitutive model for partially saturated soils, Gotechnique, 40 (1990),
405-430.
[9] A. Shastri, M. Sanchez, Mechanical modeling of frozen soils incorporating the effect of cryogenic
suction and temperature, GeoCongress 2012: State of the Art and Practice in Geotechnical Engineering,
Oakland, California, United States, 2012, 2492-2501.
[10] Z. Zhiwu, N. Jianguo, S. Shuncheng, Finite-element simulations of a road embankment based on a
constitutive model for frozen soil with the incorporation of damage, Cold Regions Science and
Technology, 62 (2010), 151-159.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1032
Numerical Analysis of Geosynthetic

Reinforced Soil Models for Shallow
Foundations
GONZALO AIASSA MARTINEZ1,a , DANNY USECHE INFANTEa,b, PEDRO
ARRUAa, MARCELO EBERHARDTa
a
Universidad Tecnolgica Nacional
b
Doctoral Fellow CONICET
Abstract. One of the most effective alternatives to improve the bearing capacity of
soft soils is to replace part of their materials with granular soils reinforced with
geosynthetics, which have much higher resistance parameters. The behavior of
granular soils including geosynthetics has been largely documented by various
investigations around the world and these have concluded that this type of
treatment decreases settlements and increases the bearing capacity of the soil to the
loads from shallow foundations. This is achieved by the tensile strength that the
geosynthetic provides to the soil-geosynthetic composite material. A bibliographic
review is presented about geosynthetic reinforced soil including experimental
results, numerical and analytical models for mechanical behavior prediction under
shallow foundations [1, 2 and 3]. The different parameters governing soil-
geosynthetic interaction are listed. Certain models have been selected and
implemented numerically. The parameters of these models were adopted according
available results for Crdoba sands and gravels. Through the different models
studied, the settlements and ultimate bearing capacity of the reinforced soil were
analyzed, depending on the type and geometry of the foundation as well as the
type, quantity and spacing of geosynthetic layers. The purpose of this study is to
perform the numerical implementation in MATLAB of mathematical models to
simulate the mechanical behavior of soil-geosynthetic system under shallow
foundations. The parameters of models were selected for typical granular soils
commonly used in compacted embankments for shallow foundations in Crdoba,
Argentina. The most commonly used material corresponds to river sand. The
geosynthetics considered refer to geotextiles and geogrids available commercially.
Keywords. Reinforced Soil, Geosynthetic, Settlement, Ultimate Bearing Capacity.
1. Introduction
Extensive investigations report experimental works to determine the behavior of the

geosynthetic reinforced soil with shallow foundations [1], these researchers conducted
tests on reinforced sands contained in tanks, applying loads with hydraulic jacks. Also,
analytical methodologies have been developed for estimating the ultimate bearing
capacity of reinforced soil foundations [2,3,4 and 5], which take into account the effect
of traction provided by the reinforcement and are based on limit equilibrium methods
or models of unreinforced soils shallow foundations.
1
Corresponding Author: Tel. +54 - (0)351 5986000 , E-mail address: [email protected]
G. Aiassa Martinez et al. / Numerical Analysis of Geosynthetic Reinforced Soil Models 1033
In this paper, a review is performed of the technical literature available about

reinforced soil and the models developed by Sharma et Al., [2] and Michalowski [4]
are analyzed with the local soil conditions of Cordoba Argentina. The results obtained
from the two models are compared, and their characteristics are evaluated using the
MATLAB environment.
2. Materials
The parameters used in the implementation of Sharma et al.s method are those for
sand commonly used as support of foundation for different types of structures in
Crdoba Argentina. Some of these sands were characterized in [6] and grouped
according to their size characteristics, distribution and morphology. This study is
performed with the parameters for the group called General Paz with unit weight
=16 kN/m3 and friction angle =35.2. Geosynthetic parameters used are those shown
in Table 1, which correspond to orthogonal geogrids fabrics composed of filaments of
polyvinyl alcohol (PVA) tissues and marketed by the company CORIPA S.A.
Table 1. Mechanical properties of geogrids [7].
Property Unit Value
Functional tensile strength (at yield 5%) (L) kN/m 20
Stiffness functional modulus (5% def.).(L) kN/m 400
Failure strain. (L) % 5
Nominal strength (at break) resistance. (L) kN/m 20
Mesh opening mm x mm 20 X 30
3. Result and Discussion
3.1. Implementation of Sharma et al.s Method
After making several literary reviews, the authors of the method identified four
possible failure mechanisms for reinforced soil with shallow foundations. These
correspond to failure above the top layer of reinforcement, failure between
reinforcement layers, failure similar to footings on a two-layer soil system and failure
within reinforced zone [2]. This paper studies the behavior of a specific type of sand
with geosynthetic included for the last two failure modes.
With MATLAB a modeling for own material conditions of Crdoba Argentina was
performed. Different graphs were made that show the variation of model parameters. In
the study strip and square footings were used, width (B=1.0 m), supported on the
ground surface with tree layers of geosynthetic and for the failure modes it is defined in
[2]. The recommended values of [8] for the variables of geometrical configuration were
taken to start the calculations. The different modelations are realized by varying the
parameter of interest and leaving the other values from local soil characteristics
constant while taking into account the recommendations made by the authors of the
method.
In order to determine the behavior of the model parameters the bearing capacity
ratio is used (BCR) defined in [9] as:
1034 G. Aiassa Martinez et al. / Numerical Analysis of Geosynthetic Reinforced Soil Models
(1)
where and are the ultimate load-bearing capacity values for the unreinforced
and reinforced foundations, respectively.
For the failure mode under the reinforced zone [2], shows the variation of the
bearing capacity ratio (BCR) with the friction angle of the soil under the reinforced
zone and in the reinforced zone as shown in Figure 1. As this parameter increases,
bearing capacity ratio (BCR) grows proportionally to strip and square foundations.
(a) (b)
Figure 1. Bearing capacity ratio (BCR) versus angle of friction of the soil , (a) Under the reinforced zone ,
(b) In the reinforced zone.
The behavior of the bearing capacity ratio (BCR) subsequently analyzed with
foundation depth for the two types of footings and two failure modes of the system
soil-geosynthetic (Figure 2). For failure mode below reinforced zone an increase occurs
in the bearing capacity ratio (BCR) with an increase in the foundation depth, these
values are similar for both types of footings. When failure occurs in the reinforced zone,
BCR is independent from the foundation depth, showing a constant value of 1.30.
Figure 2. Bearing capacity ratio (BCR) versus depth of the foundation (Df).
Also, the behavior of the bearing capacity ratio (BCR) was studied with variation
of different parameters of geometric configuration of geosynthetic as are: the distance
between the base of the foundation and the first layer of geosynthetic (u), the vertical
distance between reinforcement layers (h), and the number of layers of geosynthetic
(N) for the two modes of system failure.
(a) (b)
(c) (d)
Figure 3. Bearing capacity ratio (BCR) vs. ratio (u/B) for different width of foundation. (a) Strip foundation,
failure mode under the reinforced zone; (b) Strip foundation, failure mode in reinforced zone; (c) Square
foundation, failure mode under the reinforced zone; (d) Square foundation, failure mode in reinforced zone.
(a) (b)
(c) (d)
Figure 4. Bearing capacity ratio (BCR) vs. ratio (h/B) for different width of foundation. (a) Strip foundation,
failure mode under the reinforced zone; (b) Strip foundation, failure mode in reinforced zone; (c) Square
foundation, failure mode under the reinforced zone; (d) Square foundation, failure mode in reinforced zone.
In Figure 3 the variation the bearing capacity ratio (BCR) is presented with respect
to the vertical distance between the base of the foundation and the first layer of
geosynthetic for various values of foundation widths (B). By the strip foundation it can
be seen how the bearing capacity ratio (BCR) increases like the vertical distance (u) in
the two failure modes. For the square foundation in both types of failure, an inflection
point can be seen in the curve at the relationship u/B=0.33, also according to the
recommendations of [8].
The relationship between the bearing capacity ratio (BCR) and the vertical distance
between layers of geosynthetic (h) for various values of foundation widths (B) can be
observed in Figure 4, the curves show that the bearing capacity ratio (BCR) decreases
while increasing the vertical distance between layers of geosynthetic. In Figure 5 the
curves that show the variation of the bearing capacity ratio (BCR) with the number of
the geosynthetic soil layers included, are represented. It can be seen from these results
that it is possible to achieve a considerable increase of 50% in bearing capacity ratio
(BCR) with 3 or 4 layers of geosynthetic.
Figure 5. Bearing capacity ratio (BCR) versus number of geosynthetic layers (N).
3.2. Comparison of Sharma et al.s Method and Michalowskis Method
The curves for strip foundations shown in the preceding paragraph are compared
with the curves generated by Michalowskis method under the same conditions of soil
and geosynthetic evaluated with Sharma et al.s method. The author of the method
conducted stability analysis of reinforced soil foundation based on upper-bound
theorem [4]. Two failure mechanisms, i.e. pull-out failure and tension failure were
considered in his study. Graphs below are performed in order to compare the behavior
of the two models with the local conditions of Cordoba Argentina.
Curves for failure mode reinforcement are low compared with the curves for
failure mode by pull-out defined in [4], likewise the curves obtained by failure mode in
the reinforced zone for both methods are compared. In Figure 6 the variation of the
BCR with foundation depth (Df) can be seen. The BCR increases proportionally with
the foundation depth when failure occurs below the reinforced zone, foundation depth
does not affect failure when it occurs in the reinforced zone.
On the other hand, Figure 7 presents the performance curves of the BCR with the
increase of the vertical distance between the base of the foundation and the first layer
of geosynthetic (u), behavior is increased for both models when failure is presented
below the reinforced zone with higher values of BCR for Michalowskis method. For
failure in the reinforced zone this variable does not affect the BCR in Michalowskis
method but presents an increased behavior with Sharma et al.s method. A notable
decrease of BCR is observed when the foundation width (B) increases for the
modelation of Michalowskis method for the failure mode in the reinforced zone.
(a) (b)
Figure 6. Comparison of the bearing capacity ratio (BCR) versus depth of the foundation (D f) applying
Sharma et al.s method and Michalowskis method. (a) Failure mode under the reinforced zone, (b) Failure
mode in reinforced zone.
(a) (b)
Figure 7. Comparison of the bearing capacity ratio (BCR) versus ratio (u/B) applying Sharma et al.s method
and Michalowskis method. (a) Failure mode under the reinforced zone (b) Failure mode in reinforced zone.
When studying the BCR with increasing distance between the layers of
geosynthetic (h) opposite behavior is shown when failure occurs below the reinforced
zone, for the Sharma et al.s method the BCR shows a decrease in behavior while for
Michalowskis method behavior is increased. By failure in the reinforced zone the BCR
increases like the distance between the layers of geosynthetic for the Sharma et al.s
method and remains constant for Michalowskis method (Figure 8).
In Figure 9 the variation of the BCR with the friction angle of the soil () is
presented. For failure below the reinforced zone the BCR increases to values lower to
25 and begins to decay from this value using Michalowskis method and grows more
moderately with Sharma et al.s method. When failure is in the reinforced zone the
angle does not affect the BCR in Michalowskis method and this variable causes an
increase in BCR in Sharma et al.s method.
Figure 10 presents the variation of BCR with the number of geosynthetic layers
(N). For failure below the reinforced zone Michalowskis method exhibits greater
increase than Sharma et al.s method. For failure in the reinforced zone, both methods
show similar increase of BCR with the number of layers.
(a) (b)
Figure 8. Comparison of the bearing capacity ratio (BCR) versus ratio (h/B) applying Sharma et al.s method
and Michalowskis method. (a) Failure mode under the reinforced zone (b) Failure mode in reinforced zone.
(a) (b)
Figure 9. Comparison of the bearing capacity ratio (BCR) versus angle of friction of the soil applying
Sharma et al.s Method and Michalowskis method. (a) Failure mode under the reinforced zone (b) Failure
mode in reinforced zone.
(a) (b)
Figure 10. Comparison of the bearing capacity ratio (BCR) versus number of geosynthetic layers (N).
(a) Failure mode under the reinforced zone (b) Failure mode in reinforced zone.
4. Conclusions
In the present study the implementation of Sharma et al.s model, with mechanical
parameters of materials of Crdoba Argentina was performed, and the results were
compared with the Michalowskis method. Based on the studies carried out the
following conclusions are drawn:
The application of Sharma et al.s model with mechanical parameters of materials

of Crdoba has shown that it is possible to increase the ultimate load capacity more
than 20% at least, including three layers of geosynthetic and using strip foundations.
This conforms to the results reported in the different consulted studies and shows the
advantages of including these materials to improve the performance of a shallow
foundation; Michalowskis method shows much higher values with increases of nearly
60%.
It is necessary to conduct an experimental program to define the methodology that
best fits the conditions of the sands of Crdoba considered in this study.
Michalowskis method presents higher values for the bearing capacity ratio (BCR)
when failure is similar to footings on a two-layer soil system. In the case of failure
within the reinforced zone, this method depends of the reinforced tensile strength, the
number of layers of geosynthetic and the width of the foundation; therefore the model
does not take into account the geometrical configuration of the layers in this failure
mode.
Sharma et al.s model does not take into account the length of geosynthetic for the
numerical calculation of the ultimate load capacity, while Michalowskis method does
consider this variable.
The mayor improvement is achieved with an increase of the distance to the first
layer (u) because more replacement of material is made, and with the number of layers
(N) it causes more reinforced layers that are capable to resist external loads.
5. Acknowledgment
The authors would like to thank the National Scientific and Technical Research
Council (CONICET), the National Technological University (UTN), Regional Faculty
of Crdoba (Argentina), the National Agency of Scientific and Technological
Promotion of Argentina and the Ministry of Science Technology (Government of
Crdoba).
References
[1] G.M. Latha and A. Somwanshi, Bearing capacity of square footings on geosynthetic reinforced sand.
Geotextiles and Geomembranes. 27 . (2009), 281-294.
[2] R. Sharma, Q. Chen, M. Abu-Farsakh and S. Yoon, Analytical modeling of geogrid reinforced soil
foundation. Geotextiles and Geomembranes. 27 (2009), 63-72.
[3] K. Yamamoto and K Kusuda, Failure mechanisms and bearing capacities of reinforced foundations.
Geotextiles and Geomembranes. 19 (2001), 127-162.
[4] R.L. Michalowski, Limit loads on reinforced foundation soils. Journal of Geotechnical and
Geoenviromental Engineering, ASCE. 130(4) (2004), 381-390.
[5] M.H. Wayne, J. Han and K. Akins, The design of geosynthetic reinforced foundations. In: Proceedings
of ASCEs Annual Convention & Exposition. ASCE Geotechnical Special Publication 76 (1998), 118.
[6] G.M. Aiassa, P.A. Arra. and M.G Eberhardt, Caracterizacin morfolgica y granulomtrica de algunas
arenas de la ciudad de Crdoba. Revista de Geologa Aplicada a la Ingeniera y al Ambiente. 31
(2013), 125128.
[7] CORIPA (S.A.): www.coripa.com.ar/productos/4-geogrillas
[8] Q. Chen, An experimental study on characteristics and behavior of reinforced soil foundation. PhD
dissertation, Louisiana State University, Baton Rouge, USA. (2007).
[9] B.M. Das, Shallow Foundations: Bearing Capacity and Settlement. CRC Press, Boca Raton, USA. (1999).
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1040
Behavior of a sand-polyacrilate mixture for

seismic isolation
A. FLORAa,1, E.BILOTTAa, S.LIRERb, D. LOMBARDIa, V. NAPPAa
a
University of Napoli Federico II, Italy
b
University of Roma Guglielmo Marconi, Italy
Abstract. Seismic isolation with traditional grouting is hardly effective. Recent

results have shown that if a soft grout is used and a continuous barrier having a
very low dynamic impedance is created, seismic isolation can be very effective.
However, a soft, continuous barrier can cause undesired settlements of existing
structures. This paper shows that, considering the barrier made of a mixture of soil
and grout made using a Super Absorbing Polymer (SAP), desired values of
dynamic impedance can be reached without causing any damage to existing
structures. The paper reports also the results of some laboratory tests on SAP-sand
specimens at different percentages of SAP, difficult to carry out because of the
enormous deformability of the mixture.
Keywords. Seismic isolation, soft grouting, numerical analyses, laboratory testing.
1. Introduction
Recently, it has been shown that buried continuous barriers with a dynamic
impedance(=VS, where is the material density and VS the velocity of shear
waves) and a shear strength much lower than that of the surrounding soil may be very
effective in the seismic isolation of existing buildings 1, 2. More recently, Lombardi
et al.3 and Flora et al.4 have successfully tested this idea considering continuous
barriers with different shapes (Figure 1), proposing to create such barriers with partially
overlapped columns of soil treated with a soft grout, giving to this new technology the
name Soft Deep Mixing 4.Lombardi et al. 3and Flora et al. 4have shown that soft
grouted barriers can be very effective to mitigate at ground level the effects of a
seismic event, depending on a large number of geometrical, physical and mechanical
parameters of the soil, of the structure, and of the barrier. The soft caisson shifts to
lower values the fundamental frequencies of the soil volume it bounds, filtering out the
high frequency components close to the natural frequency of massive buildings.
1
A. Flora et al. / Behavior of a Sand-Polyacrilate Mixture for Seismic Isolation 1041
Figure 1. Possible geometrical schemes for the soft caisson: rectangular caisson (a), V-shaped caisson (b).
Hence, this technology is best suited to protect squat structures, as it may be the case of
old monuments. A soft caisson cuts the dynamic energy travelling into the soil mostly
depending on the impedance ratio , defined as the dimensionless ratio between the
dynamic impedance of the natural soil s and that of the grouted layer g: the higher ,
the better is the seismic isolation. Values of at least equal to 20 are needed.
The adoption of a soft grouted barrier in the ground poses two problems to be solved:
the first is related to the static settlements induced by the creation of the barrier, that
have to be kept as low as possible. The second refers to the possibility of having a real
grout with properties that can give to the soft grouted barrier the low desired value of
the dynamic impedance ratio . This paper deals with these two key issues.
2. Settlements induced by the creation of a soft grouted barrier
2D numerical analyses have been carried out with Plaxis to check if a rectangular soft
caisson (Figure 2) may cause undesired displacements of the structure to be protected
at ground surface. The analyses have been carried out conservatively, assuming for the
building an infinitely deformable shallow foundation having width Ls=20 m. Three
different values of the load (q=0, 50, 100 kPa) have been adopted. An elastic perfectly
plastic model with Mohr-Coulomb failure criterion has been assumed for both the soil
and the soft layer. The soil is a medium dense sand having: unit weight S=20 kN/m3,
friction angle !'S=30, angle of dilatancy "S=0, cohesion c'S=0kPa, Poisson ratio
#'S=0.3, and shear wave velocity VS,s=300 m/s. Therefore, soil shear stiffness GS (being
G=VS2) is 185 MPa, while the bulk stiffness (K=G(1+#)/(1-2#)) is KS,s=400 MPa.
For the soft grouted soil (SGS), the following parameters were kept constant: SGS=10
kN/m3, "SGS=0, c'SGS=0kPa, VS,SGS=30 m/s (and therefore GSGS=0.9 MPa). The
Poisson ratio and the friction angle of SGS were varied in the analyses as listed in
Table1. In all cases, the dynamic impedance ratio is =20, which is the minimum value
to be effective in seismic isolation3.
Table 2 briefly summarizes the results of the analyses in terms of maximum relative
rotations =2w/Ls for different values of the shear strength angle of the soft layer
(keeping the ratio KSGS/GSGS equal to 2.2) and for different values of the ratio
KSGS/GSGS, being w the difference between the settlement in the middle and on the
edge of the foundation (Figure 2).
1042 A. Flora et al. / Behavior of a Sand-Polyacrilate Mixture for Seismic Isolation
Figure 2. Numerical model for the scheme with a soft caisson with a rectangular section.
Table 1. Name of the analyses and properties of the soft grouted soil (SGS).
Analyses # 'SGS KSGS (MPa) !'SGS ()
SGS1 0.15 1 5
SGS2 5
SGS3 0.3 2 10
SGS4 15
SGS5 0.4775 20 5
In all considered cases the maximum settlements are of no more than few centimeters,
increasing as the friction angle of SGS decreases, while is always negligible and
depends on the mechanical properties of the soft layer (shear strength, shear and bulk
stiffness). By increasing the ratio KSGS/GSGS (i.e. by increasing the volumetric stiffness
of the grouted soil), the settlements sharply reduce, even for the very low friction angle
of the soft grouted soil (!'SGS=5). The distortion of the foundation increases, but
keeps in all cases very low. The results show that the settlements caused by barrier can
be reduced to very low values by tuning either the friction angle or the stiffness of the
grouted soil to the values considered in the numerical analyses.
Table 2. Maximum distortion and maximum settlement wmax in analyses SGS2 (!'=5), SGS3 (!'=10)
and SGS4 (!'=15); effect of the ratio KSGS/GSGS (analyses SGS1, SGS2, SGS5).
Analyses KSGS/GSGS wmax (m)

SGS1 1.1 0.170 0.0000050
SGS2 0.050 0.0000075
SGS3 2.2 0.029 0.0000120
SGS4 0.011 0.0000120
SGS5 22.2 0.001 0.0000110

3. SAP-Soil mixtures: experimental evidences
3.1 Materials and testing programme

The material considered in this research is a polyacrylic acid partial sodium salt, a
Super Absorbent Polymer (SAP). It is a powder that can absorb and retain extremely
large amounts of water due to the ionic concentration of the aqueous solution. In
deionized and distilled water, the SAP adopted in this research may absorb 500 times
its weight. If immersed into a 0.9% saline solution, the absorbency drops to about 50
times its weight. Using the tap water of the laboratory of Napoli, it adsorbs 140 times
its weight. Upon wetting, the particles become extremely deformable and with a
gelatinous aspect.SAP may be easily mixed in situ with the soil, to reach as a final
result a soft soil-SAP mixture (soft grouted soil SGS), having the desired low value of
the dynamic impedance SGS. The grain-size distributions of the adopted sand (specific
gravity Gs= 2.83, uniformity coefficient C u=1.8)and of dry SAP (dry specific gravity
GSAP,d=0.69, uniformity coefficient Cu=4) are reported in Figure 3a. Upon contact with
water, SAP hydrated specific gravity becomes G SAP,h=1.00, and the SAP grains largely
increase their volume (Figures 3b and 3c), assuming a gelatinous consistency: hence
the single grains show a large distortional deformability and keep a constant volume,
similarly to balloons filled with water. Because of this deformability of the hydrated
SAP grains, in the case of mixtures with a relevant amount of SAP any attempt to
prepare specimens to be tested in the triaxial or torsional shear devices failed. Then, the
characterization of sand-SAP mixtures (SGS) had to be based on simpler tests
(oedometer tests, direct and ring shear tests, Table 2), where the specimens could be
prepared under kinematically constrained horizontal conditions. The void ratio reported
in Table 2 has been calculated as:
GSAP,h Gs w V (1)
e0 &1
$1 & ' % GSAP,h Gs W
where V and W are respectively the volume and the weight of the specimen, and ' is
the percentage of hydrated SAP by weight in the mixture. This definition is in this case
arguable: since the hydrated SAP grains are gelatinous, they are not able to keep their
original shape in a mixture and tend to accommodate to the shape of the available void
volume, filling it completely. In other words, the interaction with the surrounding
grains (either of SAP or sand) is such that, immediately around a single hydrated SAP
grain, the local void ratio tends to zero. As a consequence, the void ratio of a sand-SAP
mixture traditionally calculated (eq. 1) is that pertaining to the sandy part, that is the
one related to the voids surrounded by sand grains only. This is certainly the case for a
low SAP content, where the mixture matrix is made of sand grains. In this
configuration, the SAP grains may be considered as an additional void ratio having the
peculiar characteristic of being able to change shape but not volume, like undrained
voids. Once the SAP content increases and the hydrated SAP grains are the ones
building up the governing network in the granular mixture (say for SAP percentage
equal or higher than 50%), the void ratio calculated using eq. 1 tends to reduce to very
low values, because the gelatinous grains leave no voids among them. In such a case, e0
is close to zero. Because of this, e0is not reported in Table 2 for '>0.33.
Table 3. Summary of tests carried out on mixtures at different percentage of hydrated SAP.
SAP-sand % SAP eo Oedometer tests Direct or ring VS,SGS VP,SGS
mixture (SGS) ')
by weight (' (eq. 1) shear tests tests tests
Sand 0 0.40-0.80 ( ( ( (
SGS_009 9 0.64 ( (
SGS_023 23 0.49 ( (
SGS_033 33 0.16 ( ( ( (
SGS_050 50 - ( (
SGS_070 70 - ( (
SGS_090 90 - (
SGS_100 100 - ( (
100
90
80
finer by weight %
70
60
50
Sand
40
30 SAP
20
10
0
0.01 0.1 d (mm) 1 10
(a)
(b) (c)
Figure 3.(a) Grain size distributions of sand and Figure 4. Results of the oedometer tests: vertical
SAP, SAP grains (b)before and (c) after hydration. strain ) v versus vertical effective stress v.
3.2 Results of Oedometer Tests

Since the tests were carried out on dry specimens of hydrated SAP and sand, no
consolidation had to take place. However, settlements kept slightly increasing in time
because of the large distortional deformability of the hydrated SAP grains. The vertical
displacement in each loading step was calculated after a time t=100 s. Figure 4 shows
the results of the four oedometer tests. As expected, SAP increases the compressibility
of the natural sand, The vertical deformations became relevant starting from the very
beginning of the test, under very low vertical loads.
3.3Results of Direct and Ring Shear Tests

Some direct shear tests have been carried out (vertical stress 'v=50kPa) in order to
have some information on the shear strength of the SAP/sand specimens (Figure 5). The
mechanical behavior is largely influenced by the amount of SAP: in fact, the dilatancy
of the sand is inhibited by the high deformability of the hydrated SAP grains. When the
SAP percentage is low (SGS_009, only 9%), the dilatancy is inhibited but the final
strength is very close to that of the sand specimen. This means that the soft SAP grains,
that are dispersed in the sand matrix, have the only effect of increasing the void ratio,
thus making the behavior of the specimen resembling that of a loose sand specimen.
When the percentage of SAP increases, the soft grains tend to arrange in a more
continuous pattern, thus starting to have an effect not only in terms of dilatancy but
also in terms of final strength. This phenomenological consideration is confirmed by
the tests carried out in the ring shear apparatus, which allows to apply the very large
strains necessary to reach the final shear strength for high SAP contents (up to 90%).
Figure 6 summarizes the results of both direct and ring shear tests in terms of friction
angle at constant volume!'cv: shear strength decreases as SAP content increases. Figure
` qualitatively sketches mixtures with different amounts of SAP grains: such grains
have the effect of additional voids at low percentages, while at large percentages they
constitute the governing skeleton. Seismic isolation needs the latter configuration.
40 1 Sand
SAP0.09
35 0.8 SAP0.23
30 0.6 SAP0.33
25 0.4
* (kPa)
20 0.2
sv (mm)
15 Sand 0
10 SAP0.09 -0.2 0 2 4 6 8 10 12 14
SAP0.23
sh (mm)
5 -0.4
SAP0.33
0 -0.6
0 2 4 6 8 10 12 14 -0.8
sh (mm) -1
Figure 5. Results of the direct shear tests: shear stress * versus horizontal displacements sh (a) and vertical
displacements sv versus horizontal displacement sh (b).The sand specimen has eo=0.4.
40 Direct Shear Test

35 Ring Shear Test
30
25
CV ( )
20
15
10
5
0
0 20 40 60 80 100
% SAP
Figure 6. Friction angle at constant volume (!'cv) Figure 7. Sketches of SAP-soil mixtures with
from shear tests on different SAP/sand mixtures. (a) low , (b) medium and (c) high percentage of
SAP.
3.4 VS and VP measurements

The measurement of VS was carried out in a large triaxial^cell^with an
unconventional setup. The triaxial specimen was prepared as depicted in Figure^{: two
layers of dense sand were placed at the two extremes of the specimen (having heights
h1 and h3), and the sand-SAP mixture (SGS) was placed (for a height of h2=8 cm) in the
central part. By measuring the total travel time t of shear waves in the layered specimen,
knowing VS,S in the dense sand (previously measured), the velocity VS,SGS is:
h2
VS,SGS (2)
h1 h 3
t&
VS,S
The confinement (p') was applied with controlled vacuum. The values of VS,SGS are
reported in Figure | for different mixtures,^ largely decreasing as the percentage of
SAP increases. VS,SGS increases with the confining pressure p' with the exception of the
specimens made only of SAP, whose very low shear wave velocity (VS,SGS25 m/s)is
not affected by the confining pressure. These tests confirm that by mixing SAP with
soil it is possible to have extremely low values of the shear waves velocity and, as a
consequence, of both the dynamic impedance and the shear stiffness G. The values of
VS considered for the grouted layer in the numerical analyses described in the previous
section are therefore possible using SAP-soil mixtures (SGS) with very high
percentages of SAP.
100
h1 sand
h2 SAP/sand (SGS)
layer
h3 sand
Figure 8. Experimental setup adopted to measure VS in the Figure 9. Shear waves velocity Vs of
SGS layer, using a large triaxial cell (diameter D=20 cm, SAP/sand specimens measured in the large
height H=40cm). triaxial cell.
The compression wave velocity (Vp) of the SGS specimens has been measured by
means of a very simple experimental setup: specimens of a height of 12 cm were
confined by vacuum, gently suspended at a frame and then tested using a P-waves
generation and recording device placed at the two ends of the specimen. The tests have
been carried out on different SAP/sand mixtures, at three levels of confining pressure
(p'=20-40-60 kPa). Ten evaluations were carried out for VP at each confining level. The
scattering of the measurements was extremely low (less than 1%).
The mean values of VP are plotted in Figure^1}. Since the dependency of VP on p'
was negligible, only the results pertaining to one of the confining pressures (p'=40 kPa)
are shown in the figure. In the same figure, the values of V S measured in the TX cell at
p'=40 kPa ^(Figure |) âre also reported, to have an overall understanding of the
behavior of SAP-sand mixtures. As previously shown, the shear wave velocity
decreases as the percentage of SAP increases. VP decreases as the SAP percentage
increases up to 30%, and then it shows an opposite trend, reaching its highest value
(VP=800 m/s) for a specimen made of only SAP. A possible explanation of this result
may be attempted at the granular scale, considering the sketch reported in Figure `: for
low percentages of SAP in the mixture, the SAP grains are dispersed within the soil
matrix, and the shear and compression waves travel only through the latter. As a
consequence, up to 30% of SAP both V P and VS decrease. When the percentage of SAP
increases, the SAP particles tend to create a continuous network directly involved in
waves propagation. This has the effect of further reducing VS, but VP starts to increase
because the SAP particles are made of water and the mixture tends to assume
mechanical properties closer and closer to those of water (decreasing shear stiffness,
increasing bulk stiffness). Consistently, for a specimen made of only SAP (100%), VP
reaches the highest value and VS becomes very low. Since the relationship between V P
and VS for an elastic material depends on the Poisson ratio #, it is also interesting to
look at the latter, calculated considering the VS and VP fitting curves of Figure 1}.
Figure 1~ shows that it varies consistently with the proposed interpretation, its value
tending to that of water (#=0.5) as the SAP percentage tends to 100%.
Figure 10. Values of shear and compression waves Figure 11. Poisson ratio # for SAP/sand mixtures.
velocity (Vs and VP) of SAP/sand specimens
measured at a confining pressure p'=40kPa.
Conclusions
The paper has taken an insight is some of the problems to be faced when using soft
grouted barriers as seismic isolation means, following some indications given in
literature 1, 2, 3, 4. In particular, it is of the outmost importance to see if the desired
value of the dynamic impedance for the treated soil can be obtained using a real grout.
It was shown that the Super Absorbing Polymer (SAP) proposed in this paper to make
a soft grout is able to reduce the shear stiffness of the treated soil to the very low values
needed to have an effective seismic isolation. This is possible because of the ability of
SAP to absorb enormous amounts of water, thus creating a network of gelatinous
grains into the grouted soil. The percentage of SAP to be used to have very low
dynamic impedances of the grouted barrier is large. In such a case, the mixture is
mostly composed of gelatinous grains, and the void ratio tends to become nil because
of the grains inability to keep their original shape. As a consequence, the sand-SAP
mixture tends to behave as a material having high bulk stiffness (almost a constant
volume) and very low shear stiffness. A positive outcome of using large percentages of
SAP, as observed in numerical simulations carried out on a rectangular grouted caisson,
is that settlements generated by grouting may be reduced to tolerable values because of
the high bulk stiffness. This cannot be considered but a preliminary result, and more
experimental results are needed. The activity is ongoing, and in the next future in situ
experiments will be carried out, too.
Acknowledgements
The research activity was carried out at the University of Napoli Federico II as part of
the national research project PON_03 METRICS. The financial support of the
consortium Stress s.c.a r.l. is kindly acknowledged.
References
[1] E. Kirtas E. Rovithis , K. Pitilakis, Subsoil Interventions Effect on Structural Seismic Response. Part I:
Validation of Numerical Simulations , Journal of Earthquake Engineering13(2009), 155-169.
[2] A E. Kirtasand K. Pitilakis, Subsoil Interventions Effect on Structural Seismic Response. Part II:
Parametric Investigation. , Journal of Earthquake Engineering13(2009), 328-344.
[3] D. Lombardi, A. Flora, S. Lirer and F.Silvestri, Soil grouting for the seismic protection of existing
buildings, Ground Improvement167(2014), 206-218.
[4] A. Flora, E. Bilotta , F.Foria , V. Nappa, Soft deep mixing for the mitigation of seismic risk.
Int .Conference on Deep Mixing (2015) June, San Francisco (USA).
[5] A. Flora and S. Lirer ,Small strain shear modulus of undisturbed gravelly soils during undrained cyclic
triaxial tests, Journal of Geotechnical and Geological Engineering 31 (2013), 1107-1222.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1049
Undrained Failure of Fine-Grained Soils

with Particular Reference to Shallow Slope
Instability
Anand PANTAa,1 and Satoshi NISHIMURA b
a
Graduate School of Engineering, Hokkaido University, Japan
b
Faculty of Engineering, Hokkaido University, Japan
Abstract. This paper discusses the undrained strength of heavily overconsolidated

Izumi clay and a clay-sand mixed (C-S) soil at different states (i.e. intact,
reconstituted, and compacted) in terms of the magnitude of strength and the degree
of non-linearity of failure envelopes at low effective stress levels (5-20kPa),
corresponding to the shallow slope failure expected at 1-2m of depth. A series of
constant-volume direct shear (DS) tests supplemented by hollow cylinder (HC)
simple shear tests was performed from low consolidation stresses. To express the
non-linearity of failure envelopes at low stress ranges, a non-linear power law
strength criterion was applied. The results suggest that the normalized strength and
the degree of its non-linearity of Izumi clay at intact states were much higher than
at equivalent reconstituted states, potentially due to true inter-particle cohesion.
The compacted states led to slightly lower normalized strength than the
reconstituted state in Izumi clay, while they did not result in significant difference
in the C-S soil. Inclusion of coarser particle in a clayey soil seems to reduce the
low stress level non-linearity of the strength envelopes at both states. All the above
observations suggest that the low stress level strength characteristics and their
dependence on states are significantly influenced by soil types; possibly by soil
gradation.
Keywords. Fine-grained soil, laboratory tests, shallow slope instability, undrained

failure
1. Introduction
Shallow slope failure phenomena involving a mechanism at 1-2m depth pose

challenging problems in geotechnical engineering, being particularly relevant to many
engineered slopes such as cut slopes, highway embankments, river dykes, etc. This
type of failure in fine-grained soils is considered to be governed by the undrained
failure mechanism, as the failure occurs in relatively short term following a more long-
term rise in pore water pressures along potential shallow failure surfaces during a rainy
season, as shown in Figure 1. A number of studies have revealed that a blind
extrapolation of a linear Mohr-Coulomb failure envelope, a most commonly adopted
description of soil strength, to low stress ranges greatly over-estimates the shear
strength and eventually the factor of safety [1-5].
1
Corresponding author: Anand Panta, Graduate School of Engineering, Hokkaido University, Kita-13,
Nishi-8, Kita-ku, Sapporo, 060-8628, Hokkaido, Japan; Email:[email protected]
1050 A. Panta and S. Nishimura / Undrained Failure of Fine-Grained Soils
Rainfall

Infiltration
GWT
Test element
Shallow slip surface
(Near-saturation) '
Non-linear failure envelope
Figure 1. Schematic illustration of shallow slope failure due to rainfall
Many researchers have tried to express the non-linear failure envelope by a non-
linear power law strength criterion, similar to that familiar in rock mechanics, of the
following form [5-7];
* A ,b (1)
in which, the parameter b gives the degree of non-linearity of a failure envelope; i.e. the
smaller the value of b, the greater the curvature. Applying the normalization procedure
proposed by Hvorslev [8] with respect to the equivalent vertical effective pressure (*ve)
to correct for the effect of over-consolidation ratio, Eq. (1) becomes:

* 2 , / * * A ,*
00 * -- or (2)
ve
*
1 ve .
where the parameter is equivalent to b. By using this equation, the influence of

microstructures on the strength and the non-linearity of failure envelopes of different
states of soil can be investigated in a single plot.
The soil microstructures (i.e. fabric and bonding), the void ratio, and the stress
history have a significant influence on the pre-failure and failure behavior of soils, with
dilation and destruction of bonding possibly leading to the non-linearity of failure
envelopes at low confining stress in the natural soils (i.e. Burland et al. [9] and Burland
[10]). The study used the reconstituted bounding surface as a reference framework to
investigate the influence of soil microstructures on the strength following the above
normalization procedures. This paper reports the experimental study carried out on
Izumi clay and clay-sand mixed (C-S) soil at different states (i.e. intact, reconstituted,
and compacted) and interpretation of their low-stress level strength characteristics
interpreted with the above concepts of state-normalization.
2. Materials Tested
Izumi clay is a stiff overconsolidated clay of Pleistocene origin, block-sampled from an

excavation site in Izumi city, Japan. The clay sand mixed (C-S) soil is an on-site
artificial mixture of clay and sand sampled from a recently constructed river dyke
surrounding the Maizuru flood control site in Naganuma Town, Japan. In this soil, a
locally occurring soft high-plasticity clay was mixed with a volcanic sand generated by
excavation at another flood control site to reduce the overall water content and make it
A. Panta and S. Nishimura / Undrained Failure of Fine-Grained Soils 1051
a suitable material for dyke construction. These two soils were chosen as they represent
very different soil characteristics and geotechnical settings, with Izumi clay
representing an aged natural outcrop and the C-S soil the man-made geostructure with
unnatural gradation characteristics. The grain size distribution curves for both soils
are shown in Figure 2 and the physical properties are listed in Table 1.
The reconstituted samples were prepared in a laboratory consolidometer from a
slurry of water content of about 1.7 times the liquid limit and compressed one-
dimensionally to a vertical effective stress of 400kPa. Although the C-S soil contained
non-uniform soil particles ranging from finer clay to coarser sand and gravel, the
reconstitution was made without any problems of segregation or inhomogenisation.
The compacted samples were prepared by compacting with the standard compaction
energy (598kJ/m3) in five different layers, adopting the water content 24.5% and 27.4%
(slightly wet of optimum, as shown in Figure 3) for Izumi clay and the C-S soil
respectively, to attain a well-compacted state (the degree of compaction of about 95%)
relevant to actual embankments at the same time as obtaining the degree of saturation
higher than 90% to facilitate saturation in the shear tests. Each layer was compacted
with a thickness of about 25mm so as to produce a DS test specimen (20mm in height
and 60mm in diameter) from each layer to assure as less inter-specimen variability as
possible.
Particle diameter [mm]

0.001 0.01 0.1 1 10
100
Percentage passing [%]
Izumi clay
80
60
40
Clay-sand mixed soil
20
Clay & Silt Sand
0
Figure 2. Grain size distribution curves for Izumi clay and clay-sand mixed (C-S) soil
Compaction curve Compaction curve

Dry unit weight,d [kN/m3]
Dry unit weight,d [kN/m3]
20 20
100% saturation 100% saturation
line line
18 Specimen for shear 18 Specimen for shear
tests tests
(d ) max =16.1 kN/m3
16 16 (d) max =15.4kN/m3
14 14
OMC=21% OMC=20%
12 12
0 10 20 30 40 0 10 20 30 40
Water content, w [%] Water content, w [%]
(a) (b)
Figure 3. Standard proctor compaction curves (a) Izumi clay (b) Clay-sand mixed (C-S) soil
Table 1. Summary of physical properties of Izumi clay and clay-sand mixed soil
Name wn wP wL Clay Sand / Gs vy d
of soils (%) (%)(1) (%)(1) fraction Gravel (kPa) (3) (kN/m3) (4)
(%)(2) (%)
Izumi 34.8 28 54 25/36 0/0
2.714 1690 13.62
clay
C-S 27.4 30 49 22/30 39/7 2.711 - 14.69
soil
(1)
Measured for constituents passing through a 425 micron sieve, (2) (Finer than 2m)/(Finer than 5m),
(3)
Effective vertical yield stress (4) Intact Izumi clay and compacted C-S soil before tests
3. Methods of Testing
The critical slip surface for shallow slope failure is often envisaged to be parallel to a
slope surface. Consequently, the constant-volume direct shear (DS) tests, which shear a
K0-consolidated soil specimen horizontally, can be more relevant laboratory tests than
triaxial tests in this research, as illustrated in Figure 1. Hence, the study was carried out
by performing several DS tests at low to medium consolidating pressure (i.e. 10, 20 and
50kPa), with additional 400kPa cases to complete the whole picture. The specimens
were firstly allowed to swell or consolidate fully at a desired consolidation stress. To
investigate the mechanical behavior after a pore water pressure rise and near-saturation,
all the tests were performed under fully submerged conditions, allowing sufficient time
to remove any suction due to under-saturation in compacted specimens. The specimens
were then sheared under constant-volume conditions at a horizontal displacement rate
of 0.01mm/min.
Hollow cylinder (HC) simple shear tests were performed to supplement the DS
tests and to explore the evidence of true cohesion in naturally structured soils, as
explained earlier. These tests allowed observation of the three normal effective stresses
and one shear stress during simple shear deformation, enabling constructing Mohrs
stress circle for the 1:3 plane at any moment (e.g. Nishimura et al. [11]). With the
help of these tests, it was observed how the strength mobilization on each plane
evolved when applying boundary conditions similar to those in the DS tests (i.e. no
normal strains on each direction), as shown in Figure 4. The intact Izumi sample was
first isotropically consolidated to an effective stress of 100kPa (this is still well within
the yield surface; see Table 1) and then anisotropically unloaded to 10kPa under quasi-
K0-condition. The K0 value was estimated to be 3, a value near the coefficient of
passive earth pressure, for this low effective stress level. The back pressure during
initial consolidation and swelling processes was maintained at 200kPa and the B value
was measured as 0.94.
Axial force
Size of specimen: Size of specimen:
Torque z 1
Height=20mm Height=105mm
Diameter=60mm Inner diameter=30mm
v' z
Boundary Outer cell z Outer diameter=70mm
conditions=v= h= 0 hv pressure Boundary
z conditions=z= r= = 0
Shearing force hv
(motor) Inner cell 3 r=2
pressure
v' r
(a) (b)
Figure 4. Test specimens and stress states (a) DS test (b) HC simple shear test
After the swelling stage, the specimen was sheared torsionally by applying a
shearing rate of 0.006%/minute, while keeping all the normal strains zero. A problem
of slipping along the specimen-platen interfaces during shearing at low stress was
alleviated, if not fully overcome as explained later, by using a thin layer of gypsum
paste as a binding material along the interfaces.
4. Non-linear Strength Envelopes for Different Soil Structures
4.1. Undrained Effective Stress Paths and Non-linear Failure Envelopes for Izumi Clay
It can be seen from Figure 5 that the stress paths for all the samples show dilative
behavior at low stresses. The normalized stress paths for the reconstituted samples
provide an intrinsic state boundary surface which is used as a reference surface to study
the effect of different microstructures on the strength of clay. This normalization maps
adopted low stress range (10-50kPa) against a common reference point (v /ve*=1) and
allows a fair comparison among the differently prepared samples, in which the apparent
overconsolidation ratio and void ratio were all different. The intact samples showed the
uniform and smooth behavior until reaching the peak states, followed by a rapid loss of
strength. The post-peak shear stress increased again in the tests starting from low
stresses, seemingly converging along the intrinsic strength envelope and heading
towards a potential critical state. This meta-stable behavior and convergence to the
intrinsic states may be due to shear-induced destructuring (e.g. Burland et al. [9]).
However, its abruptness appears to be pronounced in comparison to existing data. This
observation led to the investigation by HC tests to explore a possibility of tensile
failure, as discussed later. The normalized stress paths for compacted samples lie in a
lower stress region in comparison to other two states due to its high density, as seen in
Figure 5(b). Figure 5(a) also illustrates the curved normalized effective strength
envelopes for all the states of Izumi clay, which are expressed in the non-linear power
law form as given in Eq. (2). They fit the peaks very well in all the cases and the
values suggest greater non-linearity for the intact state.
The normalized peak strength envelope of Izumi clay lies well above the intrinsic
envelope, showing much higher strength as well as the significant non-linearity in
comparison to the other two states. The peak points were chosen on the basis of
maximum ratios of shear stress and vertical effective stress rather than of maximum
shear stress only. The difference between the intact and reconstituted strength envelope
at the same confining stress is attributed to natural microstructure, enhancing the
strength and its non-linearity. Interestingly, the normalized compacted strength
envelope shows marginally smaller strength than the intrinsic envelope. This difference
indicates that a compacted state can provide a strength even smaller than an intrinsic
strength when normalized. It should be noted that the two states represented very
different void ratios, as shown in Figure 5(b). The compacted specimen was markedly
denser than those of the other two states. The smaller the void ratio is, the higher the
equivalent vertical effective pressure (*ve) becomes, leading to a lower normalized
compacted strength unless the strength increases proportionally. On the other hand,
there might also be some elements which could not be normalized in terms of OCR by
the equivalent pressure concept but dependent on the absolute magnitude of stress. In
the non-normalized form, the compacted specimens exhibited relatively large strength.
Intact Intrinsic Compacted

0.6 Intact DS test series
Intrinsic DS test series
Intact Compacted DS test series
* = 0.64'*0.51 2 Intact
Specific volume, 1+e

0.4 Yield stress
Intrinsic-NCL =1690kPa
1.8 1+e = 3.01-0.42 log10 v'
*
Intrinsic
0.2 *
= 0.55'*0.71
Compacted 1.6 Compacted
* = 0.53'*0.88 Yield stress=Not Reconstituted
clearly observed Yield stress=400kPa
0 1.4
0 A 0.2 0.4 0.6 0.8 1 10 100 1000
'* v' [ kPa]
(b)
0.2
0.1
0
0 0.1 0.2
A
(a)
Figure 5. Direct shear test results of Izumi clay (a) Normalized stress paths (b) Compression curves from
CRS oedometer tests and volume-effective stress states before undrained shearing
4.2. Undrained Effective Stress Paths and Non-linear Failure Envelopes for C-S Soil
Figure 6(a) illustrates the effective tress paths for the compacted and reconstituted C-S
soil samples. The degree of dilation is similar at higher stresses, but the reconstituted
samples tended to exhibit more dilatant stress path at lower stress levels. A notable
features is the stress path starting from v/ve*=1 (i.e. upon the NCL) indicating small
dilatancy. This behavior is indicative of granular soils characteristics, which is also
noted in the gentle compression curve. The coefficient of compressibility was small
(~0.2), in the range that is usually exhibited by coarse-grained soils without a well
defined yield, as shown in Figure 6(b).
Figure 6(a) also depicts the normalized effective strength envelopes for both the
states of the C-S soil expressed in the power law form as given in Eq. (2). Comparing
the intrinsic and compacted states in the C-S soil, the differences in the normalized
strength and its non-linearity are not significant, although the pre-failure deformation
characteristics were markedly different at low stress levels as seen by the effective
stress path shapes. In the C-S soil, the non-uniformity of soil gradation ranging from
fine to coarse grains resulted in relatively lower void ratios even in the reconstituted
state in comparison to more uniform clays, leading to nearly the same void ratios as
that of the compacted state as shown in Figure 6(b). Therefore, the difference in the
normalized effective stress levels in the tests on both types of samples was less
significant than for the Izumi clay. Similar characteristics of void ratio were also
observed by Prakasha and Chandrasekaran [12] in Indian marine sand-clay mixtures.
They explained that the inclusion of sand grains in clay resulted in a decrease in void
ratio and increase in the friction angle and pore pressure variations, leading to reduction
in undrained shear strength.
1.9 Reconstituted Intrinsic DS test series

Intrinsic Compacted
0.6 Preconsolidated Compacted DS test series
1.8 at 400kPa Intrinsic-NCL
Specific volume, 1+e

1+e =2.20-0.21 log10 v'
0.4 Compacted 1.7

Intrinsic
* = 0.75'*0.96
*
* = 0.73'*0.90 1.6 Reconstituted

0.2 Preconsolidated
at 50kPa Compacted
1.5
Yield stress=Not
clearly observed
0 1.4
0 0.2 0.4 0.6 0.8 1 10 100 1000
'* v' [kPa]
(a) (b)
Figure 6. Direct shear test results of clay-sand mixed (C-S) soil (a) Normalized stress paths (b) Compression
curves from CRS oedometer tests and volume-effective stress states just before undrained shearing
5. Exploration of True Cohesion by Hollow Cylinder Simple Shear
Figure 7(a) depicts the evolution of Mohrs stress circles for the 1:3 plane observed
during the initial stage of a hollow cylinder simple shear test on an intact Izumi
specimen. The onset of the slippage at the soil-platen interface could be identified by
comparing the local and global strain measurement, and the data here are shown until
that point. It can be seen that the minor principal stress evolved towards the tensile
stress zone, showing a notable cementation or true cohesion of up to 34kPa, which
supports the foregoing discussion assuming cementation to enhance the strength and its
non-linearity. Note that this value is a lower bound due to the slippage problem; the
true cohesion may well have been greater than this value. Although it is difficult to
establish only from the data obtained so far, it is probable that failure could be
triggered by a shear-tensile combined mode when a bonded natural clay is sheared
undrained from low stresses. It seems likely that the power law failure envelope
obtained from the DS tests can be modified by taking into account for this true
cohesion. The modified equation can be derived by including horizontal intercept with
, as shown in Figure 7(a). Observing the overall shearing behavior of all three states
of Izumi clay, the state boundary surfaces for Izumi clay can be constructed by
including this true cohesion and extending to the tensile zone, as shown in Figure 7(b).
240 DS and HC test

= 9.23 ('+10)0.57
DS test Intact
180
/*ve
= 9.23'0.57 CSL
[kPa]
120
Intrinsic state
60 boundary surface
Compacted
0
-20 40 100 160 220 280 340 400 ' /*ve
-60 ' [kPa]
(a) (b)
Figure 7. Boundary surface with true cohesion (a) Evolution of Mohrs stress circles HC simple shear tests
of intact Izumi clay (b) Modified state boundary surface after including the true cohesion
6. Conclusions
The undrained strength of fine-grained soils was investigated at low stress levels by
performing several DS tests and HC tests, noting that appropriate choice of such stress
parameters and their values is crucial in assessing the slope stability against shallow
failure. Izumi clay, a uniform clay of high plasticity, and a clay-sand mixed (C-S) soil
sampled from a river dyke were studied for this purpose. The intact state of Izumi clay
showed much higher normalized strength (i.e. normalized for the density or apparent
over-consolidation ratio) and significant non-linearity due to the natural microstructure,
potentially including significant true cohesion in comparison to the reconstituted and
compacted states. Although the hollow cylinder simple shear tests were incomplete due
to slippage along the soil-platen interfaces, they at least demonstrated that the minor
effective principal stress could reach negative values before failure in simple shear. On
the other hand, the compacted state of Izumi clay showed relatively lower normalized
strength than the reconstituted state. This fact means that the undrained strength of the
compacted clay is not as large as expected for its very dense states, although the
absolute magnitude of the strength is not necessarily small. In the C-S soil, the non-
uniformity of gradation ranging from fine to coarse grains was found to lead to smaller
difference in the normalized strength of reconstituted and compacted states in
comparison to Izumi clay, which is uniform without sand particles. The degree of non-
linearity of the normalized failure envelope for the reconstituted and compacted states
of C-S soil was comparable, with the former being only slightly higher than the latter,
whereas it was notable in the Izumi clay.
References
[1] Maksimovic, M. (1989). Nonlinear failure envelope for soils. J. Geotech Eng., 115(4), 581-586.
[2] Day, R. W. (1992). Effective cohesion for compacted clay. J. Geotech. Eng., 118, 611-619.
[3] Mesri, G. (1993). Cohesion intercept in effective-stress stability analysis. J. Geotech. Eng., 119, 1229-
1249.
[4] Baker, R. (2004). Nonlinear mohr envelopes based on triaxial data. J. Geotech. Eng., 130(5), 498-586.
[5] Atkinson, J. (2007). Peak strength of overconsolidated clays. Gotechnique, 57(2), 127-135.
[6] de Mello, V. (1977). Reflections on design decisions of practical significance to embankment dams.
Gotechnique, 27(3), 279-355.
[7] Charles, J. (1982). An appraisal of the influence of the curved failure envelope on slope stability.
Gotechnique, 32(4), 389-392.
[8] Hvorslev, M. (1937). Uber die Festigkeitseigenschaften Gestorter Bindinger Bodn. Denmarks
Naturvidenskabelige Samfund. Ingeniorvidensk. Skr., A(45).
[9] Burland, J. B., Rampello, S., Georginnou, V.N., Calabrasi, G. (1996). A laboratory study of the strength
of four stiff clays. Gotechnique, 46(3), 491-514.
[10] Burland, J. B. (1990). On the compressibility and shear strength of natural clays:30th Rankine lecture.
Gotechnique, 40(3), 327-328.
[11] Nishimura, S., Jardin, R. J., Brosse, A. (2008). Simple shear testing of London clay in hollow cylinder
apparatus. 4th International Symposium on Deformation Characteristics of Geomaterials, Atlanta, 199-
206.
[12] Prakasha, K. V., Chandrasekaran, V. S. (2005). Behaviour of Marine sand-clay mixtures under static
and cyclic triaxial shear. J. of Geotech. and Geoenviron. Eng., 131(2), 213-222.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1057
Influence of Smectite Content on the One-

Dimensional Deformation Behavior of
Reconstituted High Plasticity Clays
Michael Rosenlund LODAHLa+b,1 and Kenny Kataoka SORENSEN b
a
COWI A/S, Aalborg, Denmark
b
Aarhus University Department of Engineering, Aarhus, Denmark
Abstract. Danish geotechnical engineers are challenged when designing buildings

and foundations on terrain-near Danish Palaeogene clays. The Palaeogene clays
does not behave as the typical Danish deposits which have been subjected to glaci-
ation events, which is why some of the characteristics of the Palaeogene clays are
presented in the present paper. It has been speculated that the behavior to a large
degree is due to the high content of the clay mineral smectite in the Palaeogene
deposits, which is why a series of test on artificial, reconstituted, preconsolidated
clay specimens has been conducted. From this series of tests it is seen that the con-
tent of smectite does significantly influence the deformation behavior of clay.
Keywords. Artificial Soils, Clays, Compression behavior, High plasticity, One-

dimensional Consolidation Test, Smectite, Swelling behavior
1. Introduction
Danish high plasticity Palaeogene clays have a number of common traits: High clay
content (60 - 80%), very high plasticity index, (Ip = 40 210 %), and a significant
smectite-content (15 75 % of bulk mass) cf. [1]. Usually the Palaeogene clays are
found close to ground level in Denmark along the Eastern shore of Jutland and in the
most southern parts of Funen and Zealand.
The Palaeogene clays are significantly different from the typical Danish glacial
and late glacial deposits in terms of deformation characteristics and strength. Geologi-
cal models predict that Palaeogene clays should be heavily overconsolidated because of
past glaciations and erosion events. Despite this, the apparent overconsolidation pres-
sure, 'pc determined from one-dimensional compression tests on natural undisturbed
specimens, using common interpretation methods, is found to be lower than what can
be expected from the assumed geological load. Moreover, the strength and stiffness of
the clays are generally low compared to what might be expected for a heavily overcon-
solidated clay. On top of this, the Palaeogene clays show a very high swelling potential
which may result in very pronounced heave during unloading.
The compression curve ( log 'v) of a Palaeogene clay is characterised by gradu-
al yield rather than showing a well-defined break around the expected preconsolidation
pressure cf. [2]. It has been suggested that the content of smectite in the clays may be
1
1058 M.R. Lodahl and K.K. Sorensen / Inuence of Smectite Content
responsible for the behaviour explained above, i.e. that the gradual yield reflects the
soil behaviour rather than the influence of sample disturbance which must otherwise be
expected. Common interpretation methods are based on a well-defined break but it is
unknown whether the yield point or the apparent preconsolidation pressure 'pc corre-
lates well with the strength parameters or if the maximal load sustained by the clay,
'max is better suited. It has been postulated that the pronounced swelling under unload-
ing of the Palaeogene clays may lead to degradation of the structure which in turn re-
sults in a reduction in the apparent overconsolidation ratio, cf. [3].
1.1. Observed 1D Compression Behavior of Danish Palaeogene Clays and Clay Tills
Typical behaviour of Danish Glacial Clay Tills with low plasticity is presented in Fig-
ure 1. As seen from the figure most of the Glacial Clay Tills show a very stiff response
on the unloading-reloading paths due to past glaciations, and typically a clear yield
point signifying the apparent preconsolidation pressure, 'pc can be identified. General-
ly a close correspondence between 'pc and the maximal load sustained by the clay,
'max is found for the glacial deposits.
Some of the Danish Glacial Clay Till deposits in the Eastern part of Jutland, Den-
mark are more prone to settlement and have been found to be involved in a relatative
high number of desiccation damages on buildings, cf. [4]. It is possible that the Palaeo-
gene clays usually identified below these deposits during the glaciations provided the
clay part of the Glacial Clay Tills which will introduce some of the unfortunate proper-
ties of Palaeogene clays to the Tills. Often these Glacial Clay Tills have higher plastici-
ty compared to Clay Tills from other locations in Denmark.
In Figure 2 the deformation behavior of Little Belt Clay, an Eocene marine sedi-
mentary clay, is illustrated as presented by [5]. As seen from the figure the Little Belt
Clay exhibit a significant volume change during compression and reloading. Moreover,
it is found that the swelling during unloading and the compression during reloading
0
eop
1 tot
Vertical strain, v [%]
3 Yield
point
6
1 2 3 4
10 10 10 10
Vertical stress, v [kPa]
Figure 1: Typical consolidation behavior of Danish Glacial Clay Till with a clay content of 12-15 % extract-
ed 15.7 m below seabed at Femern Belt. 12 % of the clay fraction consists of smectite minerals, in situ stress,
'v0 = 150 kPa, 'pc 2000 kPa, Compression index CC = 4.7 %. [1].
M.R. Lodahl and K.K. Sorensen / Inuence of Smectite Content 1059

5
10
15
2 3 4
10 10 10
Figure 2: Consolidation behavior of a Danish Palaeogene clay (Little Belt clay) with a clay content of 65
80 % as presented by [5]. Plot has been digitalized by the authors. Smectite content: 20 40 %, 'v0 415
kPa, 'pc 500 kPa, [5] and Cc = 13 %.
does not clearly signify that the specimen is overconsolidated, as the reloading and
unloading curves yield compression indices almost identical to the virgin curve. The
behavior presented in Figure 2 is also identified for other clays of Palaeogene origin, i.e.
for the Rsns clay, cf. [1] and [2].
As seen from Figure 2, the reloading branches of the stress-strain curve bear close
resemblance to the initial curve, causing unusually high recompression indices for a
Danish deposit which have been exposed to glaciations as compared to e.g. Glacial
Clay Tills. This behavior may be caused by the destructuration under unloading as
observed by [2] or by the excessive absorption of water during the unloading phases
which may alter the void ratio of the specimens enough to destroy the soil structure as
described by [3]. Several studies suggest that the high content of the clay mineral smec-
tite in the Palaeogene clays may cause the unusual behavior, cf. [2], [3] and [4].
Generally, an increasing content of smectite is known to increase the liquid limit,
wL of the clay [6] and thus typically also an increase in the plasticity index, Ip. Many
relations exists between Ip and the compression index CC usually predicting CC to
increase with increasing value of Ip. However, the Palaeogene clays does not follow the
well-known relations from literature typically derived for soils with Ip < 70 %, cf. [7].
Thus, it is expected that increasing content of smectite is the root cause for the behavior
observed for the Danish Palaeogene clays. A higher content may cause more swelling
in unloading which may lead to structural breakdown and fissuring. The fissures in the
clay may lead to the observed gradual yield and a longer transition between the over-
consolidated branches and the virgin curve of the stress-strains curves from Oedometer
tests, cf. [8]. Thus, the gradual yield is not believed to be due to sample disturbance.
A series of Oedometer tests on artificial reconstituted, preconsolidated clay specimens

with varying content of smectite was carried out to investigate the behavior discussed
above. The test series constitute part of a preliminary study of the effect of smectite on
the deformation behavior of clays. The preliminary study has focused on determining
the influence of smectite on the compression indices of clays, for primary loading,
swelling during unloading and for reloading. Moreover, it was investigated whether the
content of smectite influences the apparent preconsolidation pressure of the clay and
whether 'pc is related to 'max.
2.1. Preparation of Artificial, Reconstituted Specimens
A series of artificial reconstituted clay samples was prepared for testing in the Oedome-
ter apparatus. The clay samples were prepared as slurries from bentonite and kaolin
powders mixed in different ratios. The bentonite consisted of approximately 75 %
smectite [9], which is why the smectite content in the tested specimens is presented in
Table 1. In the following the specimens used in the tests are named after the following
convention KXXBYY where XX signifies the content of kaolin and YY the content of
bentonite, both determined as percent of total mass (dry weight).
Tab water was added to achieve a water content of the slurry of 125 % of the liquid
limit, wL, i.e. w = 1.25wL cf. [10]. The slurries were poured into 70 mm acrylic float-
ing ring consolidometers and pre-consolidated incrementally, starting at 'pre = 2.5 kPa,
doubling the vertical load in each step to 'pre = 163 kPa. The classification parameters
for the artificial, reconstituted samples are presented in Table 1. wint and wfin designates
the water content of the slurry prior to consolidation and the water content of the spec-
imen after complete consolidation, respectively. The index parameters listed in Table 1
are in reasonable agreement with values found in literature, cf. [6]. Discrepancies may
be explained by differences in smectite concentration of the used bentonite types.
After the preconsolidation phase the slurry samples were trimmed into a cutting
ring with diameter of d = 63.5 mm and height h = 19.9 mm and installed in the fixed-
ring Oedometer cell. The cell was installed in a Geocomp LoadTrac III automatic Oe-
dometer apparatus controlled by the software ICONP.
2.2. Test Method and Apparatus
The specimens were tested using 1D Oedometer tests without pore pressure measure-
ment. Porous disks and wetted filter paper were installed on both sides of the speci-
mens. The load was during the test controlled by a stepper motor based on readings
from a S-type load cell. The axial strain was measured by a LVDT transducer and was
corrected for self-deflection of the machinery measured using the planned load steps.
Table 1. Classification parameters for artificial reconstituted specimens prepared from slurries.
Mixture % Smectite wp [%] wL [%] Ip [%] wint [%] wfin [%]
K100B0 0.0 32 59 27 75 48
K90B10 7.5 31 84 53 99 48
K80B20 15.0 31 113 82 142 60
K100B0 K90B10 K80B20

0 1.3 0 0 1.6
eop eop eop
5
total 1.2

total total

5 1.2 1.4
Void ratio [-]

10
Void ratio [-]

Void ratio [-]
1.1
10 15
1.1
10 1 1.2
15 20
1 0.9
25 1
15
20 0.8
0.9 30
0.8
20 2 4 25 0.7 35
2 4 2 4
10 10 10 10 10 10
Vertical stress, v [kPa] Vertical stress, v [kPa] Vertical stress, v [kPa]
Figure 3: Stress-strain curves from artificial reconstituted, preconsolidated clay specimens.
The tested specimens were loaded incrementally, allowing 100 % consolidation for
each step before additional load was added. The specimens were initially loaded to 'v =
50 kPa where tab water were added to the carriage. The specimen were then loaded to
'v = 800 kPa, doubling the load from one step to the next. Next a phase of unloading to
'v = 50 kPa were performed using steps of 'v = 100 kPa (Final unloading step 'v =
50 kPa). The specimens were then reloaded to 'v = 1300 kPa in 7 steps, ensuring high
resolution of the recompression curve near the previous 'max. Finally the specimens
were unloaded in two steps to 'v = 10 kPa. The final load step was maintained until the
specimens were extracted from the Oedometer apparatus, where after the height was
quickly measured and the water content determined on the entire specimen.
3. Results
For each load step the vertical strain at end of primary consolidation was extracted
from the time curves using Taylor's method [11] and the conventional compres-
0
K100B0
5 K90B10
K80B20
10
15
20
25
30
35
1 2 3 4
10 10 10 10
Figure 4: Comparison of the laboratory tests on artificial, reconstituted clays.

Table 2: Consolidation characteristics from the conducted tests on artificial, reconstituted clay.
Mixture e0 CC [%] Cr [%] Cs [%] 'pc,1 [kPa] 'pc,2 [kPa]
K100B0 1.30 14.7 3.3 2.7 127 853
K90B10 1.30 22.5 4.7 3.7 183 872
K80B20 1.62 39.1 7.2 5.4 189 853
sion/swelling curve developed. The obtained stress strain curves are illustrated in Fig-
ure 3 and the three curves are directly compared by plotting on the same scale in Figure
4. As indicated by the horizontal arrow in Figure 4 for specimen K80B20 the consoli-
dation time for the load step of 800 kPa was insufficient to allow for full consolidation.
Thus, the stress-strain curve has been constructed using the value for 700 kPa where
full consolidation has been achieved.
From the stress-strain curves Casagrandes method [12] for estimating the apparent
preconsolidation pressures, 'pc has been applied. The results from the initial loading
branch are noted 'pc,1, cf. Table 2 and should for ideal conditions reflect the maximal
nominal pressure from the preconsolidation phase, 'pre = 163 kPa assuming no friction
in the consolidometers. The preconsolidation pressures signified 'pc,2 should ideally
represent the maximal nominal pressure applied during the first loading phase in the
Oedometer test, 'v = 800 kPa.
The deformation characteristics were determined on the stress-strain curves devel-
oped, and the derived parameters are presented in Table 2. In Table 2 CC denotes the
compression index determined on the virgin part of the curve, Cr denotes the compres-
sion index of the reloading branch and CS denotes the swelling index on the unloading
curves. Both indices are determined over the entire phase, e.g. from 'v = 800 kPa to 'v
= 50 kPa for the swelling index.
From the Oedometer tests the coefficients of consolidation, cv was determined
based on the time for 50 % consolidation, t50 and was calculated following Eq. (1), cf.
[13].
T50 H2
cv = (1)
t50
In Eq. (1) T50 = 0.197 designates the theoretical time factor at 50 % consolidation
[13]. H is the length of the path of drainage (half of the specimen height for double
sided drained Oedometer cells). The range of t50 and cv for each test is presented in
Table 3 where the highest values of t50 tends to be found for loading steps on the virgin
part of the compression curve, cf. Table 3. The time curves for the unloading are gener-
ally very ambiguous and has thus been left out from Table 3. Moreover, the coefficient
of consolidation during the initial step where water was added was also not calculated.
Table 3: Coefficients of consolidation for artificial, reconstituted, preconsolidated clays. The label NC
designates steps on the virgin part of the compression curve whereas OC designates the reloading part.
Mixture t50 [min] NC t50 [min] OC cv, NC [m/s] cv, OC [m/s]
K100B0 0.1 0.3 0.06 0.16 1.06-6 4.89-6 2.03-6 4.89-6
K90B10 0.2 18.6 0.61 4.41 2.69-6 4.47-8 2.03-6 4.89-8
K80B20 2.3 25.8 0.22 16.0 1.26-7 3.00-8 1.45-6 2.03-8
4. Discussion
As seen from Figures 3 and 4 and Table 2 CC increases with increasing smectite con-
tent. When comparing the stress-strain curves from the test on pure kaolin with the test
on the specimen with 20 % bentonite almost a doubling of the developed strain is seen
at the level of maximum applied pressure. Moreover, a tendency of increasing swelling
index, CS and recompression index, Cr, with increasing content of smectite minerals is
also observed in the tests. As seen from Table 3 a significant increase in coefficient of
consolidation is seen with increasing content of smectite. The results from the artificial
tests are in reasonable agreement with natural Palaeogene clays, which typically have
cv = 10-8 10-9 m/s, with a smectite content of in the range of 20 50 %, cf. [1].
As observed in Table 2, the 'pc values identified are quite close to the known pres-
sures sustained by the specimen. At initial compression 'pc,1 was found to be approxi-
mately 180 190 kPa for specimens K90B10 and K80B20 as compared to the nomi-
nal preconsolidation pressure 'pre = 163 kPa. The indentification of a higher preconsol-
idation pressure that the applied nominal pressure may be caused by secondary consol-
idation during preconsolidation and for the first load step or stem from the performed
curve fitting. For the specimen K100B0 the observed 'pc,1 is significantly lower than
'pre. The difference may be caused by friction in the acrylic consolidometers resulting
in lower consolidation loads than expected. Other tests carried out by the authors but
not published on pure kaolin have shown excellent correspondence between 'max dur-
ing preconsolidation of the slurry and 'pc identified from the Oedometer stress-strain
curve. For the recompression stage the estimated preconsolidation pressure 'pc,1 850
870 kPa is found to be slightly higher than but close to the applied vertical effective
pressure 'max = 800 kPa in all the tests. It should be noted that difference in pH be-
tween the mixtures might influence the diffuse double layer of the smectite and cause a
change in deformation behavior, cf. [14]. A difference in temperature may have a mi-
nor effect on the deformation behavior. However, as the three specimens were all pre-
pared and tested in the same environment the effect is assumed to be insignificant.
The observations presented in section 3 indicate that the content of smectite influ-
ences the deformation parameters of clays. Thus, it seems that the swelling and com-
pression behavior is linked to the content of smectite meaning that a higher content of
smectite results in a softer clay with more potential for swelling when unloaded. This is
in good correspondence with what has been published on the Palaeogene clays from
Denmark, cf. [1][2][4][15]. Based on this, it may be possible that the clay mineral
composition of a soil is of equal importance as the clay content, which is also, what has
previously been suggested [3]. Thus, a soil with a low content of smectite has less po-
tential for volume change during loading or unloading compared to a soil with equal
clay content but higher content of smectite.
Based on the stress-strain curves presented in Figure 4 a gradual change in the cur-
vature near 'pc,2 is observed in test on specimen K90B10 and K80B20. It seems that a
higher content of smectite causes the change in response to occur more gradually,
which is also seen from the natural samples with a high smectite content. Since the
tests were carried out on reconstituted preconsolidated clays, no fissures are believed to
influence the results. Moreover, it cannot be ruled out, that another unloading-reloading
loop may cause the artificial specimens with the highest content of smectite to behave
more like the natural Palaeogene clays as illustrated in Figure 2, yielding a very curved
reloading branch where 'pc max. Additional testing with multiple unloading-
reloading branches are planned to investigate this claim.
5. Conclusion
Based on the observed deformation behavior of Danish Palaeogene clays a series of

Oedometer tests was prepared and carried out. The focus in the test series was an inves-
tigation of the influence of smectite content on the compression and swelling behavior
of clays. Thus, artificial specimens with varying content of smectite were prepared as
slurries, preconsolidated and tested in the fixed-ring Oedometer setup. Based on the
tests it can be concluded that the content of smectite significantly influences the index
parameters and the compression and swelling behavior of clays. A minor influence on
the curvature of the recompression curve near the apparent preconsolidation pressure
was found, and further testing is necessary in order to evaluate these matters.
6. Acknowledgements
Femern A/S is acknowledged for allowing access to the geotechnical database, and for
allowing publication of the results. It is stressed that any conclusions in the paper above
is solely the authors responsibility and does not necessarily reflect the opinion of
Femern A/S.
Master students participating in the 'Experimental Geotechnics' course at Aarhus
University in autumn semester of 2014 is acknowledged for their assistance in carrying
out the test presented above.
References
[1] Rambll Arup Joint Venture, GDR 00.1-001 - Ground Investigation Report, Femern A/S, 2013.
[2] A. Krogsbll, O. Hededal and N. Foged, Deformation properties of highly plastic fissured Palaeogene
clay lack of stress memory?, Proceedings of the 16th Nordic Geotechnical Meeting (2012), 133-140.
[3] N. Mortensen, High plasticity Palaeogene clay; A soil type with very moderate long-term memory,
Proceedings of the 16th Nordic Geotechnical Meeting (2012), 133-140.
[4] H. Trankjr and L. Christensen, Har lermineralsammenstningen i fede, danske lerarter betydning for
funderingen? , Proceedings of the 16th Nordic Geotechnical Meeting (2012), 251-259.
[5] J.B. Hansen and T. Mise, An empirical evaluation of consolidation tests with Little Belt Clay, Danish
Geotechnical Institute Bulletin No. 17, 1964.
[6] J.K. Mitchell and K. Soga, Fundamentals of Soil Behavior, John Wiley & Sons, 2005.
[7] K.K. Sorensen and N. Okkels: Correlation between compression index and index parameters for high
plasticity Palaeogene clays, Proceedings for the 17th European Conference on Soil Mechanics and Ge-
otechnical Engineering (XVI ECSMGE), Edinburgh, 2015.
[8] F. Cotecchia, C. Vitrone, F. Cafaro and F. Santaloia, The mechanical behavior of intensely fissured high
plasticity clays from Daunia, Proc. 2nd int. workshop on characterization and engineering properties
of natural soils, 2006.
[9] Rotek A/S, Material safety data sheet for Cebogel Wyoming API, personal communication, 2014.
[10] J.B. Burland, On the compressibility and shear strength of natural clays, Gotechnique, 40(3), (1990),
329-378.
[11] D.W. Taylor, Fundamentals of Soil Mechanics, John Wiley & Sons, 1948.
[12] A. Casagrande. The determination of the pre-consolidation load and its practical significance. Proceed-
ings for 1st International Conference on Soil Mechanics, (1936), 60-64
[13] K Terzaghi, R.B. Peck and G. Mesri, Soil Mechanics in Engineering Practice, John Wiley & Sons,
1996.
[14] G. Mesri and R.R. Olson, Consolidation characteristics of Montmorillonite, Gotechnique, 21(4),
(1971), 341-352.
[15] N.H. Christensen and B. Hansen, Shear strength properties of Skive Septarian clay, Danish Geotech-
nical Institute Bulletin No. 7, 1959
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1065
Experimental study on
the cementation level in an artificial rock
with crushable grains
Erika TUDISCOa1, Francesca CASINIb and Giulia M.B. VIGGIANIb
a
Division of Solid Mechanics, Lund University, Sweden
b
Dipartimento di Ingegneria Civile e Ingegneria Informatica, Universita di Roma Tor
Vergata, Roma, Italy
Abstract The results of an experimental investigation of the effects of the degree

of cementation on the mechanical behaviour of a porous artificial rock with
crushable grains are presented. The studied material is an analogue of real
cemented granular materials, such as pyroclastic weak rocks, carbonate sands,
calcarenites and compacted decomposed granite. The present study aims to clarify
the role of the degree of cementation in the relative importance or the sequential
nature of mechanisms in natural material, i.e., granular rearrangement and
breakage are expected to appear after de-bonding of particles. Cemented samples
with different cement content were investigated. To understand the influence of
the cementation, uncemented samples were reconstituted from a mixture of
crushable grains and cement (not hydrated) in the same percentage by weight as
the cemented (hydrated) samples, in order to obtain the same fraction of fines.
Preliminary results show that, for the same confining stress, the cemented sample
are more compressible than uncemented ones during isotropic compression; while,
during axial loading, the cemented samples show a more rigid behaviour and a
lower resistance.
Keywords. Porous rock, Grain crushing, Cement breakage, De-bonding
1. Introduction
This study aims to investigate the loading effect on the degradation of particles bonding
and grain crushing of weak rocks and cemented soils and focuses on the effects of the
degree of cementation on the mechanical behaviour. To this purpose an experimental
campaign on a porous artificial rock with crushable grains is ongoing.
The stress-strain behaviour of the artificial soft rock with crushable grains under
examination results from the interplay of different mechanisms of deformation,
including: (a) granular rearrangement with relative frictional sliding and rotation, (b)
de-bonding, and (c) breakage of particles (grain crushing). These mechanisms have all
been observed to occur in geotechnical materials, such as pyroclastic weak rocks,
1
1066 E. Tudisco et al. / Experimental Study on the Cementation Level in an Articial Rock
carbonate sands, calcarenites and compacted decomposed granite, of which the

artificial rock tested may be considered a physical model [1].
To obtain a rock-like material, grains of Light Expanded Clay Aggregates (LECA),
which are crushable particles, have been cemented in the laboratory. The use of an
artificial material allows the control of the degree of cementation as well as the initial
porosity of the samples. Furthermore, the artificial material is weaker than most natural
soft rocks, which makes it possible to use standard geotechnical apparatus, normally
used for soils, instead of apparatus designed for rocks.
The behaviour of the uncemented material was fully characterized in a previous
study [2] where the changes of grading of the material after isotropic, one-dimensional
and constant mean effective stress triaxial compression were described using a single
parameter based on the ratio of the areas under the current and an ultimate cumulative
particle size distribution, which were both assumed to be consistent with self-similar
grading with varying fractal dimension.
In this work cemented LECA (CLECA) samples were tested in dry conditions.
Samples with a cement percentage of 25% and 50% by particle weight were considered.
In order to compare the behaviour of cemented and uncemented material, uncemented
samples (ULECA) were reconstituted from a mixture of crushed LECA and not
hydrated cement. To obtain the same grain size distribution and fraction of fines,
cement was added in the same percentage by weight as the cemented (hydrated)
samples. The samples were isotropically compressed to the target confining pressure
and then axially loaded at constant rate.
2. Experimental work
2.1. Material
The degradation processes associated with grain crushing play a significant role in the
mechanical behaviour of the natural materials. However, systematic experimental
investigation of grain crushing for natural materials is often difficult due to the
relatively high stress required to crush the grains and the variability and heterogeneity
of natural deposits, which makes it difficult to obtain repeatable results. For these
reasons the experimental work was carried out on an artificial granular material
consisting of crushed expanded clay pellets, whose grains break at relatively low stress.
The main physical characteristic of the material is the very low apparent unit
weight of the particles; this is due to the existence of a double order of porosity: "inter-
granular", i.e. voids existing between particles, and "intra-granular", i.e. closed voids
existing within individual particles [2]. Figure 1(a) show examples of Scanning
Electron Microscopy (SEM) micrograph of crushed LECA particles.
Figure 1(b) shows the "apparent" unit weight of the ^!
_?as, as a function of
their size, d, as determined in the laboratory on different fractions of crushed material.
T!? as increases significantly with decreasing grain size and tends to the unit
weight of the _ _?!s 26.5 kN/m3 [3].
E. Tudisco et al. / Experimental Study on the Cementation Level in an Articial Rock 1067
(a) (b) 27
24
s = 12.64 (d/d0)-0.268
21
s (kN/m3)
d0= 1 mm
18
15
12
9 . .
0 05 1 15 2
d (mm)
Figure 1 (a) Scanning Electron Microscopy (SEM) micrograph; (b) Apparent unit weight of crushed LECA
particles as a function of grain size (after [3])
2.2. Sample preparation
The samples are made of crushed grains of LECA mixed with standard Portland
cement. The grains are crushed and sifted to obtain a homogeneous sand (grain size
170- \ Cement is added to obtain a 25% or 50% cement/grain weight ratio.
ULECA samples are then prepared by pouring this mixture into a mould, which is
mounted on the base pedestal of the triaxial cell. To control the homogeneity and the
porosity a funnel is used to this purpose and it is raised during the formation of the
sample according to the current height. A membrane is placed between the mould and
the sample and it assures its stability after the mould is removed and before the
confining pressure is applied.
CLECA samples aim to represent porous artificial sandstone; the mixture is
hydrated with a water/cement ratio of 100% in order to obtain a quite dry and not fluid
paste. The material was then poured in layers of about 2 mm into a mould and
compacted, using a piston designed to cover the surface area of the sample section,
dropping the piston from an average height of 20 mm.
The micro-structure of a small number of samples, obtained using the preparation
technique described above, was analysed with the aid of X-ray tomography. As shown
in Figure 2, the obtained artificial rock has the micro-structure of a soft porous rock, as
intended. Local porosity has been chosen as a representative parameter of the structure
and can be computed by first performing a binarisation of X-ray images. The values of
mean porosity in a layer, plotted over the height of the sample, present only slight
oscillation around an almost constant value of porosity [4]. CLECA and ULECA
samples are cylindrical, with standard size for triaxial test (diameter of 38 mm and
height of 76 mm).
Figure 2 Example of a reconstructed slice of x-ray tomography with a blow-up of detail (after [4])
2.3. Experimental program
The experimental programme, which is still in progress, consists of 16 tests including

isotropic and triaxial compression, carried out at increasing confining pressures, on
cemented and uncemented samples with two cement contents.
To study the influence of confining stress on the transition from a structured
material with intergranular boning to a purely granular material, samples are tested at
100 kPa, 200 kPa, 400 kPa and 800 kPa confining pressure. At low confining pressure,
cement breakage is likely to occur during the shearing phase, while at higher confining
pressure a certain amount of de-bonding is expected during isotropic compression. In
this paper, a limited set of results, at confining pressure of 400 kPa, will be discussed.
3.1. Isotropic compression
Figure 3 shows the evolution of the specific volume, v, defined as Vtot/Vs where Vtot is
the total volume and Vs is the volume of the solid, as function of the mean effective
stress. The volume of the solid is calculated from as, reported in Figure 1b, integrating
the equation as(d) over the range of diameters (0.425-0.17 mm), adding the unit weight
of cement (s = 30.9 KN/m3) for the relative fraction (as =20.68 kN/m3 with 50%
cement; as = 19.35 kN/m3 with 25% cement). The cemented samples appear more
porous than the corresponding uncemented ones. This is probably due to sample
preparation. Up to an isotropic pressure of about 70 kPa all the samples present a
rather rigid behaviour. Subsequently, the uncemented samples show a log-linear
behaviour. The sample with 50% of cement/grain ratio contracts more, as expected due
to the higher content of fines. Cemented samples show a higher compressibility, which
is not linear with the logarithm of the applied pressure, and the curve seems to indicate
a sudden collapse that might be due to cement breakage and then a partial de-bonding.
.
35
v()
3
.
25
C25 UC25
C50 UC50
2
1 10 100 1000
p' (kPa)
Figure 3 Results from isotropic compression tests on cemented and uncemented samples.
3.2. Shear phase
The deviatoric stress, q

!

!s curves are shown in Figure 4(a) during
the axial compression phase at controlled displacement rate va=0.02 mm/min. The
_
^ _
?\
!v
!

!s, are shown in Figure
4(b). The shear strain is calculated starting from the axial displacement and volumetric
change measurements. The relation between the specific volume, v, and the mean
effective stress, p, is presented in Figure 5.
All samples present a ductile behaviour and the samples compress throughout the
test. Whilst both cemented samples seem to reach a yielding point during the isotropic
compression, their stress-strain curves are initially linear up to a relatively small shear
strain, less than 1.5%, with an abrupt change of slope that became rather horizontal.
The response of the uncemented samples is nonlinear with a smaller initial stiffness
than cemented samples; the deviator stress increases to reach a very gentle peak at
shear strain of about 20%.
Shear resistance is lower for cemented samples than for uncemented ones and it is
influenced by the percentage of cement in opposite directions in the two cases. Higher
cement content increase the resistance for cemented sample, indicating that the yielding
is linked to cement breakage and grains de-bonding. In the loose samples the cement
acts as fines reducing the shear resistance.
UCLECA samples compress more than CLECA specimens and the compressibility
increase with the cement content. Figure 5 indicates that the specific volume of the
cemented samples is nearly constant up to the rupture, after which it dramatically
decrease, while the specific volume of the uncemented samples decrease gradually.
1500
1000
q (kPa)
500
C25 C50
UC25 UC50
confining stress p'c=400 kPa
0
0 5 10 15 20 25 30
%
s ( )
0
0 5 10 15 20 25 30
10
v ( )
%
15
20
25
30
Figure 4 triaxial compression tests on cemented and uncemented samples: !`s v s .
.
25
v()
.
15
C25 C50
UC25 UC50
1
300 1000
p' (kPa)
Figure 5 triaxial compression tests on cemented and uncemented samples: v vs p

3.3. Grading evolution
Figure 6 shows the grain size distribution by weight obtained before and after triaxial
compression on uncemented samples (cementation do not allow the measurement of
the grain size distribution after the test). In both cases, the cumulative grain size
distribution at the end of the test is modified by loading but the changes in the grain
size distribution are more pronounced for the sample with 25% of cement/grain ratio.
Previous studies showed that isotropic compression causes only very limited changes to
the cumulative grain size distribution even at relatively large mean effective stress [1].
Therefore the grain crushing can be associated to the shear loading. The two curves
obtained after test seem to converge toward a unique state of grading.
1 1
UC25 before test UC50 before test
UC25 after test UC50 after test
. .
08 08
. .
06 06
P()
P()
. .
04 04
. .
02 02
(a) (b)
0 0
. . . . .
0 001 0 01 01 1 0 01 01 1
d (mm) d (mm)
Figure 6 Grain size distribution evolution for uncemented samples with (a) 25% and (b) 50% cement/grain
ratio in terms of cumulative weight percentage.
4. Conclusions
This work presented preliminary experimental results of a wide laboratory

investigation, still ongoing, on the mechanical behaviour of artificial cemented and
uncemented granular material with crushable grains.
During isotropic compression this results show that all samples exhibits a rigid
behaviour up to a confining pressure of 70 kPa. After that, cemented samples show a
high compressibility, which probably indicates a sudden structural collapse that may be
due to cement breakage.
During the shear phase, all samples present a ductile behaviour and compress
throughout the test. Uncemented samples show a smaller initial stiffness than cemented
samples, with a gentle increase of deviatoric stress and a light peak at around 20% of
shear strain. The cemented samples show an abrupt change on the stress-strain curve
for a shear strain of about 1.5%. The shear strength is lower for the cemented samples
than for the uncemented ones. Strenght increases with cement content for CLECA
samples, indicating that the yielding is linked to de-bonding, while in the loose samples
the cement acts as fines reducing the strength.
Analysis of the complete set of tests, with confining pressure from 100 kPa to 800
kPa, will provide more information on the stress condition delimiting the transition of
the material behaviour from structured to loose.
References
[1] P.V: Lade and N. Trads, The role of cementation in the behaviour of cemented soils, Geotechnical
Research 1.4 (2014), 111-132.
[2] F. Casini and G.M.B. Viggiani, Experimental investigation of the evolution of grading of an artificial
material with crushable grains under different loading conditions, Proceedings of the 5th International
Symposium on Deformation Characteristics of Geomaterials, Seoul, Korea. 2011
[3] F. Casini, G.M. Viggiani, and S.M. Springman, Breakage of an artificial crushable material under loading.
Granular matter, 15(5), (2013), 661-673.
[4] E. Tudisco, Development and application of time-lapse ultrasonic tomography for laboratory
characterization of localized deformation in hard soils/soft rocks, Ph.D. thesis, Universite de Grenoble
and Universita degli studi di Roma Tor Vergata, 2013
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1073
Laboratory investigation on the mechanics

of soft-rigid soil mixtures
Salman ROUHANIFARa,1 and Erdin IBRAIM a
a
Department of Civil Engineering, University of Bristol, United Kingdom
Abstract. Rubber particles derived from recycled tyres are used as a practical
materials in construction industry, including geotechnical systems. Drained triaxial
compression tests are conducted to investigate the mechanical behavior of Leighton
Buzzard and rubber particles mixtures, bot materials with identical particle size
distributions. Distribution of constituents throughout the sample is also explored
through x-ray CT testing.
Keywords: laboratory, sand, tyres, triaxial test
1. Introduction
The number of scrap tyres is increasing rapidly in both developed and developing
countries due to the steady rise in the number of vehicles. As a consequence, the
accumulation of used tyres is gradually becoming a real societal problem, equally from
an economic and environmental points of view. Recent research showed that this
material can be considered as an alternative for some conventional materials in
construction industries. As far as geotechnical systems are concerned, the use of the
scrap tyres or tyre chips for backfilling can be an attractive solution that would provide
lighter weights on the retaining structures compared to traditional backfilling materials
[1], [2]. The tyre shreds apparently produce less horizontal pressure than conventional
granular backfills [3]. The possibility for the inclusion of tyres or rubber chips derivatives
in other geotechnical applications like soil prevention erosion [4], slope stabilization and
highway embankments [5, 6], road constructions [6, 7] and seismic isolation of
foundations [8] is under attention. However, before extensive implementation, further
research is still required in order to understand the behaviour of the soil/tyre chip
composite mixtures, including internal interaction mechanisms resulted from the
combination of two particular materials, one soft, tyre rubber, and one rigid, granular
soil. Insight into the underlying particle-level mechanisms and their role on the
macroscale behaviour for either fine rubber particles mixed with coarse sand grains or
coarse soft particles with fine rigid sand grains has been explored [9, 10]. The quality of
soft-rigid soil mixtures produced by normal mixing and depositional procedures,
including the mitigation of particle segregation phenomena represent a challenge for
their implementation at real scale.
1
Corresponding Author: Doctoral Researcher, Department of Civil Engineering, University of Bristol,
Queens Building, Bristol, BS8 1TR, United Kingdom: E-mail: [email protected]
1074 S. Rouhanifar and E. Ibraim / Laboratory Investigation on the Mechanics
2. Aims and Objectives
In this paper, laboratory experimental results are presented on sand-tyre rubber chips
mixtures in view of exploring the role of the soft particles on their mechanical behaviour.
The main idea here is the use of rigid and soft particle constituents with identical particle
size distributions (PSD), avoiding the contrast effects of the particle size scales.
Therefore, the analysis of the rigid/soft mixture is mainly focused on the effect of the
proportions of the mixture constituents. When testing rubber/sand composites in the
laboratory, one of the main concerns is the segregation of particles that can occur in the
fabrication process of the samples. A non-destructive method for the assessment of the
uniformity of the distribution of rubber particles throughout the sample is explored. The
mechanical behaviour of the sand/rubber mixture has been investigated in triaxial
compression testing under drained conditions and various confining stresses.
3. Materials
Leighton Buzzard fraction A sand has been chosen in this research. This is a natural,
silica sand characterized as a coarse material. Following a long and tedious process that
involved market investigation, individual assessment of rubber chip samples sent by
various suppliers, sieving and sorting of the particles, an equivalent rubber type material
has been created to match the particle size distribution (PSD) of the Leighton Buzzard
sand (Figure 1). Rubber material resulted from the shredding process of used lorry tyres
and it consists of polymer, acetone, carbon black, ash and sulphur.
Various index properties of the materials are given in Table 1. Figure 2 shows a
series of individual sand and rubber particles for some visual appreciation of the particle
shapes.
Figure 1.Particle Size Distribution of Leighton Buzzard (Fraction A) Sand and Rubber Particles
S. Rouhanifar and E. Ibraim / Laboratory Investigation on the Mechanics 1075
Table 1.Index Properties of Materials

Material Properties Rubber Particles Leighton Buzzard Sand
Specific Gravity 1.04 2.65
Minimum Void Ratio (e min) - 0.55

Maximum Void Ratio (e max) - 0.83
Mean Grain Size (D50 (mm)) 1.8 1.8
Coefficient of uniformity (Cu)* 1.27 1.27
Coefficient of Curvature (Cg)** 1.12 1.12

*Cu=D60/D10 **Cg= (D30*D30)/ (D10*D60)
Figure 2.Individual particles of Leighton Buzzard Sand (top row) and Rubber Particles (bottom row)
4. Sample Fabrication
There are different ways to create rubber-sand samples in laboratory but, in general, the
fabrication process invariably consists of mixing, deposition and compaction stages.
Although water content is normally used for mixing sand and shredded rubber tyre chips
[11], the mixing process can also be completed in dry conditions [12]. The deposition of
the composite mixture is made either in one layer or several successive layers by using
a funnel or by spooning of the mixture in small quantities and zero drop deposition height.
The compaction is completed by tamping [12], tapping [13, 14] or vibration [3]. However,
laboratory samples of rubber-sand mixtures are most commonly prepared using a moist
tamping (MT) technique. While this technique could be much more effective in
discouraging segregation of the constituents of the composite, providing good control of
sample density and homogeneous distribution of rubber, it produces a soil-rubber fabric
which corresponds to that obtained in rolled-compacted construction fills.
4.1. Distribution of tyre chips
Different fabrication methods (dry deposition and layers compaction, dry deposition and
vibration, moist tamping) have been explored in this research. However, the results
presented here will be limited to dry deposition of the mixture followed by light
compaction by a circular tamper (half sample diameter) and vibration. The samples with
a rubber fraction FR= 30% were prepared in a transparent Perspex tube of 70mm diameter
and 70mm height as shown in figure 3. The rubber fraction, FR, is defined by the
following relationship:

(1)

where the letter V stands for volume.
Figure 3.Sample in Perspex tube
The assessment of tyre chips distribution was conducted through 3D x-ray CT testing.
X-ray tomography provide 2D high resolution images with a resolution of 50m/px. The
particles of sand with high density will appear in the images in a grey colour, lighter than
and clearly distinguished from the rubber particles. The darkest zones represent the voids
(Figure 4a).
As it is shown in figure 3, the sample was divided in three horizontal sections: HT,
top side of the sample, located at 2cm from the top; HM, middle; and HB, bottom side,
located at 2cm from the bottom. Each horizontal section was further divided in four
square areas, as displayed in Figure 4. At each HT, HM and HB level, 40 successive
horizontal images representing an overall slice of 2mm thickness (approximate D50) 20
above and 20 below each level were selected and each image was analysed separately
using Matlab software to estimate the sand area (AS) and rubber area (AR) with respect
to the total areas (AT) of the analysed sub-section. As an example, Figure 4b shows the
sand particles (white) in four horizontal sub-sections, while the black zones represent
rubber and voids. Similarly, images with rubber selected particles can be generated and
rubber area AR estimated. In a first approximation, for one section, we can assume that
the FR ratio can be given by n1/ (n1+n2), where n1= AR/AT and n2=AS/AT. The FR ratio for
each level and sub-section, is then obtained by averaging of FR ratios of all 40 sub-
sections, above and below that level. The obtained FR values for HT, HM, and HB levels
are given in the Tables 2, 3 and 4 respectively. It can be observed that the distribution of
the constituents at one level is not particularly homogenous, with FR values within 10%
range. The FR average of rubber fraction for each level reveals the existence of some
degree of segregation between the constituents with higher rubber content on top of the
sample than at the bottom. The FR average decreases from 34% on top of the sample to
24% in the middle and 20% at the bottom. The analysis of various images also shows
that near the vertical sample boundary, over two to three times D 50 thickness area, the
concentration of rubber particles is much higher than the sand. In addition, it also appears
that some rubber particles tend to agglomerate and create visible rubber clusters.
Figure 4. (a) Horizontal top section of the sample (HT); (b) Equivalent HT1 to HT4 sub-sections for image
analysis with sand particles (white) and the black zones representing rubber and voids
Table 2.Area Ratios for Top horizontal sections

Element No. AS/AT AR/AT FR FR(average)
HT1 0.37 0.23 0.38
HT2 0.46 0.17 0.27
0.34
HT3 0.41 0.19 0.32
HT4 0.35 0.24 0.40
Table 3.Area Ratios for Middle horizontal sections

HM1 0.42 0.15 0.26
HM2 0.46 0.11 0.20
0.24
HM3 0.49 0.14 0.22
HM4 0.42 0.16 0.28
Table 4.Area Ratios for horizontal sections

HB1 0.46 0.10 0.18
HB2 0.43 0.13 0.23
0.20
HB3 0.49 0.07 0.13
HB4 0.50 0.16 0.25
5. Triaxial Compression Tests on Sand-Rubber Samples
Cylindrical samples with 70mm in diameter and 70mm in height were made with moist
tamping method. Different FR values of 0%, 5%, 10%, 20%, 30%, 50% and 100% have
been considered, while the sample fabrication void ratio of 0.64 has been kept identical
for all samples.
Figure 5.Triaxial Compression Tests on Leighton Buzzard/Rubber Mixture: (a) Deviator Stress-Axial Strain;
(b) Volumetric Strain-Axial Strain
The triaxial tests were performed on saturated samples and that allowed the
measurement of the sample volume change. Extensive research has shown that a sample
aspect ratio of 1 preserves better the sample uniformity (cylindrical deformation) at very
large strains and delays the manifestation of localisation when enlarged lubricated top
and bottom sample platen ends are used. Therefore, enlarged top and bottom platens as
well as anti-frictional systems have been used. Samples with different rubber fraction
were tested in drained triaxial compression under different confining pressures of 50kPa,
100kPa and 200kPa. However, in this paper only the results at 100kPa confining pressure
are presented. Deviator stress and volumetric strain of the samples with axial strain are
shown in figure 5. The results present the well expected behaviour of dense sand which
are pronounced peak deviator stress followed by softening towards a critical state and
highly dilative volumetric behaviour. Linear stress-strain response and contractive
volumetric behaviour of rubber sample are also shown in this figure which is consistent
with previous works [3]. It is clearly shown that adding rubber to sand generates an
intermediate response. Although the strength of the rubber-sand mixtures decreases, the
strain corresponding to the peak deviator stress increases. The volumetric behaviour for
20%, 30% and 50% rubber fractions is also closer to the pure rubber sample behaviour.
The volumetric behavior of samples with F R>30% are fully contractive compared to the
samples with less rubber fractions. On the other hand the volumetric behaviour of 5%
and 10% mixtures is closer to pure sand sample. These results are also consistent with
some previous studies [12]. The variation of the mobilized angle of friction for all the
tests under different confining pressures with the rubber fraction is presented in Figure
6.
Figure 6.Variation of mobilized angle of friction under different confining pressures for all the tests
6. Conclusion
Leighton Buzzard sand and rubber particles mixtures were studied in this paper. The
rubber particles, resulted from shredding process of used scrap tyres, were chosen with
the same particle size distribution of Leighton buzzard sand. Different sample fabrication
methods were considered to analyse the homogeneity of the composite and possibly
detect signs of segregation between the constituents. Samples fabricated by dry
deposition and light compaction and vibration, showed some degree of non-uniformity
and apparent segregation with a higher concentration of rubber particles on the top of the
sample. Samples with different rubber fractions of 0 to 100% were tested in drained
triaxial condition under different confining pressures. As expected, rubber particles have
significant effects on both stress-strain and volumetric strain of the mixtures. Further
investigation using a finer soil and finer rubber particles is in progress.
References
[1] D.N. Humphrey and W.P. Manion. Properties of tire chips for lightweight fill.
in Grouting, Soil Improvement and Geosynthetics. 1992. ASCE.
[2] V.K. Garga and V. O'shaughnessy, Tire-reinforced earthfill. Part 1:
Construction of a test fill, performance, and retaining wall design. Canadian
Geotechnical Journal. 37(1) (2000): p. 75-96.
[3] J. Lee, R. Salgado, A. Bernal, and C. Lovell, Shredded tires and rubber-sand as
lightweight backfill. Journal of Geotechnical and Geoenvironmental
Engineering. 125(2) (1999): p. 132-141.
[4] P.S. Poh and B.B. Broms, Slope stabilization using old rubber tires and
geotextiles. Journal of performance of constructed facilities. 9(1) (1995): p. 76-
79.
[5] P.J. Bosscher, T.B. Edil, and S. Kuraoka, Design of highway embankments
using tire chips. Journal of geotechnical and geoenvironmental engineering.
123(4) (1997): p. 295-304.
[6] D.E. Nightingale and W.P. Green, An unresolved riddle: Tire chips, two
roadbeds, and spontaneous reactions. ASTM SPECIAL TECHNICAL
PUBLICATION. 1275 (1997): p. 265-285.
[7] T.C. Heimdahl and A. Drescher, Elastic anisotropy of tire shreds. Journal of
geotechnical and geoenvironmental engineering. 125(5) (1999): p. 383-389.
[8] Y. Tsompanakis. , P.N. Psarrapolous, and V. Drosos, Low-cost seismic base
isolation using recycled tires cush-ions. 2009, Technical University of Crete.
[9] J.S. Lee, J. Dodds, and J.C. Santamarina, Behavior of rigid-soft particle
mixtures. Journal of materials in civil engineering. 19(2) (2007): p. 179-184.
[10] H.K. Kim and J.C. Santamarina, Sand-rubber mixtures (large rubber chips).
Canadian Geotechnical Journal. 45(10) (2008): p. 1457-1466.
[11] S. Youwai and D.T. Bergado, Strength and deformation characteristics of
shredded rubber tire sand mixtures. Canadian Geotechnical Journal. 40(2)
(2003): p. 254-264.
[12] J.G. Zornberg, A.R. Cabral, and C. Viratjandr, Behaviour of tire shred sand
mixtures. Canadian Geotechnical Journal. 41(2) (2004): p. 227-241.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1081
The Bearing Capacity of Debris Flows:

Laboratory and Centrifuge Analyses
Luis E. VALLEJOa,1, Hankyu YOO b and Bernardo CAICEDOc
a
Professor, Department of Civil & Environmental Engineering, University of
Pittsburgh, Pittsburgh, USA
b
Professor, Department of Civil Engineering, Hanyang University, Korea
c
Profesor, Departamento de Ingenieria Civil y Ambiental, Universidad de los Andes,
Bogota, Colombia
Abstract. Debris flows are made of a cohesive muddy matrix in which large
dispersed particles are in suspension. In this study the bearing capacity of muds is
investigated using penetration and centrifuge tests. For the penetration tests, solid
wooden balls attached to a bar were used. The resistance offered by the mud to the
penetration by the balls was evaluated. This resistance was related to the
undrained shear strength of the mud, cu , the diameter (D) of the balls, and the
closeness of the balls. It was determined that the closer a set of balls were to each
other, the lower was the resistance offered by the mud. To evaluate how cohesive
muds developed their bearing capacity, centrifuge tests were carried out on
cylinders embedded in a soft clay matrix. The cylinders used had a diameter equal
to 7 mm and a length equal to 2.5 cm. The cylinders were embedded in a soft
mass of kaolinite clay having a water content of 30%, and a cu = 6 kPa. The soft
clay mass with the embedded cylinders were placed in a box with transparent walls
that allowed the recording of the deformation experienced by the clay surrounding
the cylinders during the centrifuge tests. The box with the clay and the cylinders
was subjected to acceleration equal to 200g. The centrifuge tests revealed that the
clay around the cylinders developed a zone of slip surfaces on which the cu of the
mud was mobilized. The cylinders with the slip zones moved as one piece through
the mud. The combined cylinder-slip line zone had a diameter equal to 1.9d, where
d is the diameter of the cylinders. When the cylinders were close to each other,
their individual slip line zones overlapped causing a decrease in the bearing
capacity of the mud.
Keywords. Debris flows, bearing capacity, rocks in suspension, undrained shear

strength, slip line theory, penetration and centrifuge tests.
1. Introduction
Many investigators have described the ability of debris flows to transport in suspension
large rock particles, some several meters in diameter [1, 2]. One of the theories that has
been suggested to explain the ability of debris flows to transport large particles is based
on the undrained cohesive strength, cu, of the cohesive muddy matrix that surrounds the
large particles. This undrained shear strength of the matrix gives to it bearing capacity
that helps to maintain the large particles in suspension [1, 2, and 3].
1
Corresponding Author. Department of Civil & Environmental Engineering, University of Pittsburgh,
Pittsburgh PA 15261. vallejo@pitt,edu
1082 L.E. Vallejo et al. / The Bearing Capacity of Debris Flows
The second mechanism that acts together with the cohesive strength is buoyancy.
According to Johnson [1] and Hampton [4], the buoyant force acting on a large particle
in a debris flow is equal to the weight of the muddy matrix displaced by the large rock
particle.
Mathematically, the competence of the muddy matrix to carry in suspension a rock
particle (of diameter D) immersed completely in the muddy matrix can be obtained
from the following relationship,
WT WB F = 0 (1)
Where WT is downward force due to the weight of the large rock particle, WB is the
uplift force due to buoyancy, and F is the uplift force due to the undrained cohesive
strength of the muddy matrix.
The value of WT can be obtained from the following relationship,
WT = (1/6) Js (SD3) (2)
where, Js is the unit weight of the rock particle, and D is its diameter.
The value of the buoyant force, WB, can be obtained from the following relationship,
WB = (1/6) Jf (SD3) (3)
where, Jf is the unit weight of the muddy matrix.

For the case of a rough sphere, the value of F can be obtained from the following
relationship [5],
F = 12.2 D2 cu (4)
If one replaces Eqs. 2, 3 and 4 into Eq. 1, one can obtain the diameter of the rock
particle that the debris flow can support in suspension in function of the unit weights of
the rock particle and muddy matrix, and the undrained shear strength of the mud. This
resulting equation for the debris flow competence is,
D = (23.3 cu) /(Js - Jf ) (5)
Debris flows contain many large rock particles; to date no research has been
conducted to evaluate how the proximity of rock particles affects the bearing capacity
of debris flows. The purpose of this study is to clarify this very important issue.
2. Slip Line Theory and the Bearing Capacity of Debris Flows
The a sphere penetrates a plastic material, the plastic material opposes the penetration
by developing a region around the sphere on which the undrained shear strength, cu, is
mobilized on curves called slip lines. Fig. 1 shows the mobilized region around the
sphere with the related slip lines. The material beyond this mobilized region remains
undeformed. The form of this region of mobilized undrained shear has the form of
L.E. Vallejo et al. / The Bearing Capacity of Debris Flows 1083
truncated toroid with its section centered on the surface of the sphere with a diameter
equal to (1.41) D, where D is the diameter of the sphere [6].
Figure 1. Slip lines around a penetrating sphere in a plastic material
Randolph, Martin and Hu [5] used the slip line approach shown in Fig. 1 to obtain the
relationship given by Eq. 4.
3. Penetration Tests
For the penetration tests, a mixture of 50% kaolinite clay by weight, 10% of calgon
(hexamethaphosphate), used as dispersed agent), and 40% distilled water was used as
mud. This mixture was placed in a rectangular container, 31 cm in length, 31 cm in
width, and 43 cm in depth. Wooden, rough spherical balls attached to bars were used
for the tests. The spheres used had diameters D1=7.68cm , D2= 3.81 cm, and D3=2.54
cm.
Figure 2. Tank and loading equipment used for the penetration tests
The sphere with D1=7.68 cm was used to obtain the undrained shear strength of the
mud [Fig. 3(A). The spheres with D2= 3.81 cm were placed on a bar and were arranged
vertically at regular intervals [Fig. 3(B)], and the spheres with D3=2.54 cm were placed
on bars as shown in Fig [3(C)]. These spheres were arranged at regular intervals
horizontally and vertically. The penetration of the mud by the spheres was
accomplished by the upward movement of the base of the compression machine. Also,
before the penetration resistance tests started, the single as well as the multiple spheres
were fully embedded in the mud (Fig. 2). For the single sphere, the top of the sphere
was embedded 19.2 cm from the surface of the mud. The rate of penetration of the
spheres was kept constant at 0.0508 cm/min.
Figure 3. (A) Single sphere, (B) Spheres arranged vertically on a bar, (C) spheres arranged horizontally and
vertically on the bars.
3.1. Tests with a Single Sphere
Fig. 4 shows the results using a single sphere. The value of the resisting force, F,
increases with the depth of penetration. The value of F is obtained using Eq. 4. The
value of F was equal to 8.3 Newtons. With this value of F, we obtain that the mud had
a cu= 0.012 Newtons/cm2= 0.12 kN/m2.
Resisting Force, F
10
(Newtons)
8
6
4 Sphere D1=7.68 cm
2
0
0 0.2 0.4 0.6 0.8
Depth of Penetration (cm)
Figure 4. Resisting force F in function of the depth of penetration for sphere 1 [Fig. 3(A)]
3.2. Tests with Multiple Spheres
Fig. 5 shows the results using the arrangement of spheres shown in Fig. 3(B). The
results of the tests indicate that the frictional resistance F increased as the separation a
between the spheres increased. The increase of the shear strength is the result of non-
overlapping of the region with the slip lines shown in Fig.1.
50
Resisting Force, F
40
(Newtons)
30
20
Sphere D2=3.81
10 cm
0
0 5 10 15
Separation Between Spheres, a (cm)
Figure 5. Resisting force F in function of the vertical separation between the spheres shown in Fig.
3(B)
Fig. 6 shows the results using the arrangement of spheres shown in Fig. 3 (C). The
results of the tests indicate that for a fixed value of the vertical separation between the
spheres (e=5.08 cm), the frictional resistance F increases with an increase in the
horizontal separation between the spheres, b [Figs. 3 (C) and 6]
84
Resisting Force, F
82
(Newtons)
80
78
76
74 D3-2.54 cm,
72 e=5.08 cm
70
0 2 4 6 8
Horizontal Separation of Spheres, b (cm)
Figure 6. Resisting force F in function of the horizontal separation between the spheres shown in
Fig. 3(C).
The increase of the shear strength is the result of non-overlapping of the region
containing the slip lines shown in Fig.1.
4. Centrifuge Tests
To evaluate how the cohesive muddy matrix develops its bearing capacity in debris
flows, centrifuge tests were performed using the Universidad de los Andes mini beam
geotechnical centrifuge. The main characteristics of the centrifuge are: 0.3 m radius,
500g maximum acceleration, model dimensions 12 cm x 7 cm x 2.5 cm (Fig. 7). The
machine dimensions and its acceleration capacity permit simulate prototypes measuring
60 m in length, by 35 m in height and 12.5 m in width.
Figure 7. Mini-centrifuge used for the laboratory tests
The tests were performed on simulated clay-rock mixtures. These mixtures were
done with kaolinite clay and metallic cylinders. The number of cylinders used in the
testing program varied between 1 and 4, all the cylinders had 7 mm diameter and
lengths of 2.5 cm. The cylinders simulated the large rigid particles present in soil-rock
mixtures in the field under conditions of plane deformations. The clay used in the
experiments was a kaolinite with an undrained shear strength, cu = 17 kN/m2.
The procedure to create the mixture samples was the following:
(a) First, kaolinite clay in slurry form was prepared in the laboratory.
(b) This slurry was consolidated under self-weight in the centrifuge at 200 g up to
36 % consolidation (around 30 minutes consolidation).
(c) At the end of the previous consolidation, a 7 mm diameter holes were made in
the clay sample right in the position selected for each cylinder.
(d) The cylinders were placed in the holes and a 5 mm grid was drawn on the soil
side of the transparent wall of the centrifuge to measure deformations during the
centrifuge tests.
(e) The mixture was subjected to further consolidation. This consolidation process
continued until no more deformation was measured in the samples. No deformation
represented 100 % consolidation and it took about 4.5 hours.
4.1. Deformation Patterns in the Clay Surrounding the Cylinders
Figs. 8, 9 and 10 shows the deformation of the clay around one cylinder, two vertically
arranged cylinders and two horizontally arranged cylinders. An analysis of these
figures indicates that the clay surrounded the cylinders developed regions of slip lines
as predicted by the theoretical analysis shown in Fig. 1. For the case of the vertically
arranged cylinders one can see that the slip line regions interacted (Fig. 9). This
interaction will produce a weakening of the material surrounding the cylinders that will
cause a decrease in the undrained shear strength of the clay. This decrease in shear
strength will cause the cylinders to settle in the clay. This settlement is clearly seen in
Fig. 9. The slip lines shown in the clay samples with the cylinders developed under an
acceleration equal to 200g, g being the acceleration of gravity.
Figure 8. Region of slip lines in the clay around Figure 9. Region of overlapping slip lines
one cylinder in the centrifuge tests in a clay with 2 vertically arranged
cylinders
Figure 10. Regions of slip lines in the clay surrounding two horizontally arranged cylinders
5. Conclusions
From this study the following conclusions can be reached:

(1) Debris flows carry in suspension large rock particles. The muddy cohesive
matrix that surrounds the particles supports them by developing: (a) a region of slip
lines where the undrained shear resistance is mobilized, and (b) by buoyancy.
(2) The cohesive resistance of the muddy matrix was evaluated by penetration tests
using one and a group of spheres separated at regular intervals horizontally and
vertically. The tests with the group of spheres indicated that the cohesive resistance
offered by the mud to the penetration of the group of spheres decreased in value when
the spheres were close to each other. The reason for this decrease in cohesive resistance
was the overlapping of the regions of slip lines where this resistance is mobilized.
Overlapping regions of cohesive resistance represents zones where the mud is
overstressed. Overstressed zones will cause a decrease in the cohesive resistance
offered by the mud.
(3) The theoretical regions of slip lines that a mud develops in order to support
large particles in suspension were visualized using centrifuge tests on a mixture of
saturated clay in which metallic cylinders were dispersed. The centrifuge tests also
showed the overlapping of the slip line regions that takes place when the cylinders are
close to each other.
References
[1] A. M. Johnson, Physical Processes in Geology, Freeman and Cooper, San Francisco, 1970.
[2] M. A. Hampton, Competence of fine grained debris flows, Journal of Sedimentary Petrology,
45(1975), 834-844.
[3] L. E. Vallejo, Evaluation of test methods designed to obtain the undrained shear strength of muds,
Marine Geotechnology, 7(1988), 173-188.
[4] M. A. Hampton, Buoyancy in debris flows, Journal of Sedimentary Petrology, 49(1979), 753-758.
[5] M.F. Randolph, C. M. Martin, & H. Hu, Limiting resistance of a spherical penetrometer in cohesive
material, Geotechnique, 50 (2000), 573-582.
[6] R. W. Ansley, and T. N. Smith, Motion of spherical particles in a Bingham plastic, Am. Inst. Chem. Eng.
J. 13 (1967), 1193-1195.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1089
The Effect of Fine Content and

History of Hydraulic Loading
on the Characterization of Suffusion
Sensibility of Cohesionless Soil
Abdul ROCHIMa,b,1 and Didier MAROT a, Luc SIBILLEc, Fateh BENDAHMANEa
a
Institut GeM, University of Nantes, France
b
Civil Engineering Department, Sultan Agung Islamic University, Indonesia
c
University of Grenoble Alpes, CNRS, 3 SR, F-38000, Grenoble, France
Abstract. Suffusion involves a selective erosion of fine particles in a soil by the

water seepage in the matrix formed by coarser particles. This phenomenon can
induce a change in particle size distribution, porosity and hydraulic conductivity of
the material. With the objective to characterize the soil sensibility and propose
suffusion sensibility classification, a series of one-dimensional downward seepage
flow tests was realized with an erodimeter. Tests were performed under controlled
hydraulic gradients in binary mixtures of sandy gravel soils. Two controlling
parameters were used: 1) fine content and 2) hydraulic loading history. An analysis
based on energy induced by the seepage flow is proposed to characterize the
hydraulic loading and the cumulative eroded dry mass to characterize the soil
response. The results show that for a given soil different hydraulic loading
histories result in different soil sensibilities. Soils with the lowest fine contents
tend to require larger flow energy for the development of erosion. These results
demonstrate that this approach is effective to characterize suffusion sensibility for
cohesionless soils.
Keywords. Cohesionless soils, erodimeter, suffusion, water seepage power
1. Introduction
During the cycle of seasons, hydraulic structures made of soil may be in contact with
variable interstitial flows, possibly generating the detachment and the transport of some
constituent particles of the structure or their foundations. Different categories can be
distinguished in this problem generally is called internal erosion: concentrated leak
erosion, suffusion, contact erosion and backward erosion [1]. Suffusion corresponds to
selective erosion of fine particles by seepage flow leaving behind the matrix of coarser
particles. However, only soil sensibility classifications for interface erosion were
proposed so far [2, 3, 4], and classification of soils with respect to suffusion sensibility
is still missing.
1
Corresponding Author: Institut GeM, University of Nantes-ECN-CNRS, 58 rue Michel-Ange BP 420,
44600 Saint Nazaire Cedex, France; E-mail: [email protected].
1090 A. Rochim et al. / The Effect of Fine Content and History of Hydraulic Loading
2. Literature Review
Suffusion is a complex matter as it depends on geometric, mechanical and hydraulic

condition. The size of fine fraction should be smaller than the size of constrictions, in
addition the volume of fine particles should be less than the volume of void formed by
the coarse particles. This can result in criteria on the shape of the grading curves [5, 6].
Hydraulic conditions should be studied to investigate the onset of suffusion even if the
transport of particles is geometrically possible. The hydraulic loading applied on a soil
is often described by the hydraulic gradient, the pore velocity or hydraulic shear stress.
Recently a new method has been proposed in [4] to describe the hydraulic loading from
the energy dissipated by the fluid seepage.
The stream power related to water seepage, and called here erosion power,
Perosion can be expressed by:
' ' (1)
where Jw is the unit weight of water, 'P = PA PB, the pressure drop; 'Z = ZA ZB, the
altitude change for a one dimensional flow between an inlet section A and an outlet B;
and Q the fluid flow rate. The erosion energy, Perosion, is defined as the time integration
over the test duration of Perosion.
When the erodibility, or erosion sensibility, is estimated with respect to the
hydraulic shear stress (assumed to be representative of the hydraulic loading), an
erodibility coefficient indicates the erosion rate for a unit excess of shear stress through
an erosion function [7, 3]:
(2)
where is the soil erosion rate; kd the erodibility coefficient; W the hydraulic shear
stress at the soil-water interface; and Wc the critical shear stress at onset of erosion.
It is worth stressing that Equation 2 has been developed only for interface erosion.
However in case of suffusion, the expression of the hydraulic shear stress within a soil
defined in [8] can be used. For a vertical downward flow the shear stress is given by:
P
(3)

where 'h is the hydraulic head drop, k the hydraulic conductivity, P the viscosity, and
n the porosity.
3. Downward Seepage Test
3.1. Tested Gradations and Specimens
Three grain size distributions corresponding to sandy gravel soils were tested (Figure
1). Properties of gradations are summarized in Table 1. The grain size distributions are
assessed Table 1 with respect to several gradation-based criteria [6, 9, 10, 11]. The
definitions of H/F and d15/d85 are given in [9] and [10] respectively.
A. Rochim et al. / The Effect of Fine Content and History of Hydraulic Loading 1091
100
90
80
Percent finer by weight, F (%)
70
60
50
R2
40
30
20 B
10 A
0
0,001 0,01 0,1 1 10
Particle diameter, d (mm)
Figure 1. The particle size distribution of tested specimens
Table 1. Tested gradations properties and assessment of soil internal stability by recent criteria
Tested Cu Gr P (H/F)min d (H/F=min) d15/d85 Assessment method
gradation (%) (mm) [9] [10] [11] [6]
A 17.06 2.14 1.227 0.038 0.400 8.761 U U U U
B 19.52 2.14 1.533 0.035 0.400 8.741 U U U S
R2 24.46 WG 1.200 0.165 0.212 9.653 U U U S
Cu: uniformity coefficient; Gr: gap ratio (see [12]); P: percentage of particle smaller than 0.063 mm
U: internally unstable gradation, S: internally stable gradation
[6, 9, 10, 11] are gradation criteria: Wan and Fell, Kezdi, Kenney and Lau, Li and Fannin respectively
3.2. Erodimeter
An erodimeter as shown in Figure 2 was used to characterize the soil sensibility. It

consists of an erosion cell, a demineralised water supply system, a soil collection
system, and a water collection system. The sample is saturated by flushing water in the
upward direction, and erosion test is then performed by injecting through the sample in
the downward direction from the injection tank. The funnel-shaped draining system is
connected to effluent tank by a glass pipe. The effluent tank is equipped with an
overflow outlet in order to control the downstream hydraulic head and a rotating
sampling system containing 8 beakers for catching eroded particles. Overflow water is
continuously weighed to determine flow rate.
A series of tests was conducted in two steps: specimen preparation (production of
the specimen, and saturation) and downward seepage test. The sand grains and gravel
are then first mixed with a fixed moisture content. The specimens are then prepared
using a single layer semi-static compaction technique. The mixture is placed in a mould
of 50 mm diameter and 50 mm height and subsequently compressed under the action of
two pistons until the initial fixed dry density is reached. In the erosion cell device, the
specimen is placed on a 4 mm pore opening grid and wrapped within membrane. This
pore opening allowed the migration of all particles of sand. For the saturation phase
carbon dioxide is first injected in the specimen for duration of 5 minutes, before
injecting demineralised water. The whole saturation phase required approximately one
night. Finally, the specimen is subjected to a hydraulic flow in a downward direction
using demineralized water.
A series of nine tests with two different hydraulic loadings was performed: multi-
stage (a and b) and a single-stage (c) hydraulic gradients. The first (a) hydraulic
loading consisted of increasing the hydraulic head by steps from 0.1 to 2, then by steps
of 0.5 between 2 and 4 and by steps of 1 beyond; whereas steps were equal to 1 for the
second kind of hydraulic loading (b). For each step the hydraulic gradient was kept
constant during 10 minutes. However c corresponds to a single-stage hydraulic gradient
equal to 4 m/m.
Figure 2. Schematic diagram of the erodimeter
4.1. Erosion rate and Hydraulic conductivity
Erosion rate and hydraulic conductivity can be interpreted as a soil response to

hydraulic loadings. Figure 3 shows the evolution of hydraulic conductivity of tested
specimen during test and Figure 4 depicts the variation of erosion rate versus hydraulic
shear stress. The arrow signs in both figures highlight the time when hydraulic
conductivity reaches constant and significant eroded mass is captured. The evolution of
hydraulic conductivity present a typical trend: a decrease is firstly observed before it
progressively increases and finally reaches a constant value from the arrow sign as
shown in Figure 3. We assume this decrease of the hydraulic conductivity is attributed
to some detached particles that were transported under the imposed water seepage and
then filtered further within the soil itself. This filtration leads to partial clogging
corresponding the first decreasing of the hydraulic conductivity. Hydraulic
conductivity increases only latter, for much larger hydraulic gradients able to detach
again these filtrated particles.
From the variation of hydraulic conductivity of specimens subjected to a multi-
stage hydraulic gradients (A-90a, A-90b, B-90a, B-97a, B-97b, R2-90a, R2-90b) three
predominant processes can be identified: filtration, process of erosion and finally
steady state (represented by constant value of hydraulic conductivity). However such
successive phases do not seem to occur when a single stage hydraulic gradient equal to
4. Considering test B-90a, by increasing progressively the applied hydraulic gradient
from 1.3 to 4 the hydraulic conductivity increases by a factor of 20. Whereas in the
case of test B-90c, performed with a single-stage hydraulic gradient, even by applying
hydraulic gradient of 4, the hydraulic conductivity continuously decreases. Thus for the
need to the classification of suffusion sensibility, the method by increasing hydraulic
gradients is preferred.
1,0E-03
constant hydraulic
hydraulic gradient, conductivity
i=4
Hydraulic conductivity (m/s)
increased multi-stage
hydraulic gradients (B-90a )
A-90a
1,0E-04 A-90b
A-90c
B-90a
B-90c
B-97a
single-stage hydraulic
B-97b
gradient, i = 4 (B-90c )
R2-90a
1,0E-05 R2-90b
hydraulic gradient,
i = 1.3
Filtration erosion process steady state
1,0E-06
0 60 120 180 240 300 360
Times (min)
Figure 3. Time series of hydraulic conductivity
4.2. Soil sensibility

In conformity with methods proposed for interface erosion classification, a first
approach to define a suffusion sensibility classification can consist to investigate the
variation of the erosion rate with the hydraulic shear stress. It is worth stressing that the
erosion function proposed by [3] and [7] has been developed only for interface erosion.
1,0E-05
hydraulic gradient,
i=4
1,0E-06
Erosion rate (kg/s/m )
A-90a
2
A-90b
Erosion coefficient, kd
1,0E-07 A-90c
B-90a
B-90c
hydraulic gradient, B-97a
i = 1.3 B-97b
1,0E-08
R2-90a
R2-90b
1,0E-09
A-90c
B-90c
1,0E-10
0,001 0,01 0,1 1
W (N/m2)
Figure 4. Erosion rate versus hydraulic shear stress

The hydraulic shear stress within the soil is computed from Equation 3 and the
erosion rate coefficient kd is assumed to be representative of the erosion sensibility. In
the necessity to evaluate the classification of suffusion sensibility of tested specimens a
first approach proposed by [3] for interface erosion can be used. This approach is based
on the determination of kd coefficient. In order to apply such approach for suffusion
process, the value of kd is defined from linear approximation of the plot of the erosion
rate with respect to the critical hydraulic shear stress. The critical hydraulic shear stress
is determined at the point where the hydraulic conductivity presents a minimum or
where the hydraulic conductivity starts to inflect. For instance, for specimen B-90a this
point corresponds to a hydraulic gradient equal to 1.3 as shown in Figure 3 and 4. The
difficulty to determine kd coefficient is encountered for a single-stage hydraulic
gradient (A-90c and B-90c) as shown in Figure 4. Thus these specimens are not taken
into account for proposition of suffusion sensibility classification.
The approach based on the erosion energy, as introduced in Section 2, and
corresponding to the energy expanded by the water to seep through the soil, is now
considered. Figure 5 shows the relationship between cumulative expanded energy and
cumulative eroded mass. The arrow signs indicate the point from which the hydraulic
conductivity is stabilized.
1,0E+03
Cumulative eroded mass (kg/m )
3
1,0E+02
b
b A-90a
a
1,0E+01 a A-90b
B-90a
B-97a
b B-97b
1,0E+00 R2-90a
R2-90b
1,0E-01
1,0E-02
1 10 100 1000 10000 100000 1000000
Energy (J/m3)
Figure 5. Cumulative eroded dry mass in terms of the cumulative energy expanded
by the water to seep through the soil.
4.3. Classification of suffusion sensibility
The proposed classification is built from the approach based on the erosion energy
(Figure 5). This classification is divided into five classes from resistant to highly
erodible as presented in Figure 6. The points highlighted by arrows correspond to the
end of development of suffusion (i.e. time from which hydraulic conductivity is
stabilized) in order to classify the soil sensibility. It is worth noting that the points
before and after the arrows do not represent the soil sensibility. The eroded mass in the
beginning may be attributed to loss mass during saturation phase and during filtration
process. The after-arrow points also do not represent the soil sensibility since they
correspond to the steady state. If the series of final points is taken into account, the soil
sensibility becomes less erodible.
1,0E+03
Cumulative eroded mass (kg/m )
3
1,0E+02
R2-90b
B-97a
B-97b R2-90a
Highly B-90a
erodible
1,0E+01
A-90b
Erodible A-90a
1,0E+00
Moderately Moderately
erodible resistant resistant
1,0E-01
10 100 1000 10000 100000 1000000
3
Energy (J/m )
Figure 6. Classification of suffusion sensibility based on the energy expanded
by the water seepage in the soil
4.4. Effect of hydraulic loading history and fine contents
Figure 5 shows that soils A, B and R2 imposed to hydraulic gradient a are less
erodible than those imposed to hydraulic gradient b. It can be also shown in Figure 6,
given B-90 and R2-97 different hydraulic loading histories give different classifications.
Thus hydraulic loading history can be an important parameter to soil sensibility. With
respect to fine content, it seems that a specimen with a low fine content soil, specimens
A tends to require a larger energy for the onset of erosion than a specimen with the
same initial density but a higher fine content, specimens B. Indeed, since the same
initial global density is obtained for specimens A and B, but with a larger content of
coarse particles A, density proper to the coarse fraction should be higher in A than in B.
Therefore constriction sizes of the coarse skeleton may be smaller for specimen A,
limiting the possibility of erosion of the fine fraction. This result is in good agreement
with the test results presented in [13]. Given three different fine contents: 16.7%, 20%,
and 25%, it is demonstrated that the samples with the lowest fine content (16.7%)
required a larger critical hydraulic gradients for the onset of internal erosion for relative
density of 0.2 and 0.6 respectively. These results show that the energy-based approach
should be effective to distinguish suffusion sensibility for cohesionless soils.
5. Conclusion
In this study, the effect of history of hydraulic loading and fine content on the
characterization of suffusion sensibility of cohesionless soil was investigated through a
series of tests using erodimeter. When a soil is susceptible to erosion, the soil
sensibility has to be characterized by imposing seepage flow in suffusion tests. The test
should be performed by progressively increasing the applied hydraulic gradient and it
should be carried on until the hydraulic conductivity reaches a steady state. With the
objective to characterize independently the hydraulic loading and the induced erosion,
the cumulative eroded dry mass and the cumulative energy expanded by the seepage
flow, Eerosion are computed. Finally with the aim to classify suffusion sensibility, the
first time hydraulic conductivity reaches constant value can be determined as the time
to suffusion sensibility classification of a soil on the diagram of the cumulative eroded
dry mass versus the cumulative expanded energy. For given soils, it demonstrates
different soil sensibility classification for different hydraulic loading history and fine
content. The gradient of erosion coefficient for hydraulic gradients with increment 1
(hydraulic gradient b) shows more erodible than ones with increment 0.1 (hydraulic
gradient a). This result shows how hydraulic loading is an important parameter for
characterization of suffusion.
Acknowledgement
The authors thank The Indonesian Directorate General of Higher Education (DIKTI),
Sultan Agung Islamic University Indonesia, GeM Institute - University of Nantes-
ECN-CNRS for providing financial support.
References
[1] R. Fell, J. J. Fry, The state of the art of assesing the likelihood of internal erosion of embankment dams,
water retaining structures and their foundations, Internal Erosion of Dams and their Foundations,
Taylor & Francis Publisher, New York, 2007.
[2] G.J. Hanson, A. Simon. Erodibility of cohesive streambeds in the loess area of the midwestern U.S.A.
Hydrological Processes, 15(1) (2001), 23-38.
[3] C.F. Wan, R. Fell. Investigation of rate of erosion of soils in embankment dams. Journal of
Geotechnical and Geoenvironmental Engineering (ASCE) 130(4) (2004), 373-380.
[4] D. Marot, P.L. Regazzoni, T. Wahl. Energy based method for providing soil surface erodibility
rankings, Journal of Geotechnical and Geoenvironmental Engineering (ASCE) 137(12) (2011), 1290-
1294.
[5] J. Lafleur, J. Mlynarek and A.L. Rollin, Filtration of broadly graded cohesionless soils. Journal of
Geotechnical Engineering 115(12) (1989), 1747-1768.
[6] C.F. Wan, R. Fell, Assessing the Potential of Internal Instability and Suffusion in Embankment Dams
and Their Foundations, Journal of Geotechnical and Geoenvironmental Engineering ASCE 134(3)
(2008), 401-407.
[7] G.J. Hanson, Channel erosion study of two compacted soils. Transactions of the ASAE, 32(2) (1989),
485-490.
[8] L.N. Reddi, I. Lee, M.V.S. Bonala. Comparison of internal and surface erosion using flow pump test on
a sand-kaolinite mixture, Geotechnical Testing Journal, 23(1) (2000), 116-122.
[9] A. Kezdi, Soil physics selected topics, Elsevier Scientific Publishing Company, Amsterdam, 1979.
[10] T.C. Kenney, D. Lau, Internal stability of granular filters, Canadian Geotechnical Journal 23 (1985),
420-423.
[11] M. Li, J. Fannin, Comparison of two criteria for internal stability of granular soil, Canadian
[12] D.S. Chang, L.M. Zhang, Extended internal stability criteria for soils under seepage. Soils and
Foundations 53(4) (2013), 569-583.
[13] L. Ke, A. Takahashi, Strength reduction of cohesionless soil due to internal erosion induced by one-
dimensional upward seepage flow, Soils and Foundations 52(4) (2012), 698-711.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1097
Testing Tube Specimens from Soft Clay

Deposits Containing Variable Amounts of
Shells
Guan T. LIM a,1, Nathalie BOUKPETI a, J. Antonio H. CARRARO a and Jubert A.
PINEDA b
a
Centre for Offshore Foundation Systems, The University of Western Australia
b
ARC Centre of Excellence for Geotechnical Science and Engineering, University of
Newcastle
Abstract. This paper presents the results of an experimental program aimed at

characterizing tube specimens from an estuarine soft clay deposit, which contains
variable amounts of shells. The characterization of the natural clay involved two
non-destructive techniques for assessing specimen quality, namely computed axial
tomography (CAT) and shear wave propagation using bender elements, as well as
one-dimensional consolidation tests for evaluating soil compressibility. Image
analysis of CT scans was used to quantify the volumetric shell fraction and its
influence on the compressibility parameters derived from one-dimensional
compression tests. The experimental results show that both the compressibility and
the small-strain stiffness of the natural soft clay are influenced by the volumetric
shell fraction.
Keywords. Soft clay, one-dimensional consolidation test, image analysis, natural

heterogeneity and shells
1. Introduction
In order to obtain reliable results from laboratory tests, an important aspect of

experimental studies is to test good quality specimens that represent soils natural state.
Natural soil variability along the tube, e.g. presence of shells or inclusions as well as
soil disturbance caused by sampling may influence the laboratory test results. The
inconsistency of test result may lead to failure of infrastructure and excessive
mitigation costs in projects. Furthermore, the presence of natural heterogeneities, such
as inclusions of larger particles, results in additional experimental challenges,
particularly in regard to soil extrusion from sampling tubes and specimen trimming,
which may lead to additional soil disturbance.
The behavior of clays mixed with granular materials has been studied by a number
of researchers (e.g., [1], [2], [3], [4]). One-dimensional compression tests on clay-sand
or clay-gravel mixtures prepared in the laboratory have shown that the compressibility
of the mixture decreases with an increase in granular volume fraction [3], [4]. In
addition, experiments have demonstrated that for a granular volume fraction below a
critical value (in the range 0.3 to 0.45) the behaviour is determined by the clay matrix
1
1098 G.T. Lim et al. / Testing Tube Specimens
[1], [3]. While substantial amount of research has been carried out on the effects of
sampling disturbance on the behavior of clay materials (e.g., [5] [6], [7], [8]), little has
been reported on the disturbance caused by sampling process in mixtures of clay and
granular materials. The aims of this study are twofold: (i) investigate the influence of
shell fragments on the compressibility parameters of natural soft clay, (ii) assess the
quality of samples of this natural clay obtained with sampling tubes having different
cutting toe angles. The paper presents and discusses the results of a series of laboratory
tests, namely CAT tests, bender elements tests (to measure shear wave velocity) and
incremental loading one-dimensional compression tests (IL) carried out on tube
specimens of natural soft clay containing variable amount of shells.
2. Material Tested
Tube specimens of soft estuarine Ballina clay were obtained from the Australias first
national facility for soft soils testing located near Ballina, New South Wales. The soil
profile comprises an upper alluvium layer ( 1 m thick) overlying a 10 m thick soft
clay layer (126 % <w< 68 %; organic content 5 %) whose index properties (Atterberg
limits) are strongly sensitive to previous drying before testing [9]. The clayey fraction
is mainly composed of smectite, kaolinite and mica/illite with variable percentages [10].
Three Shelby tubes (75 mm in diameter) with outer cutting edge angles (OCA) of 5, 15
and 90 degrees were retrieved at a depth between 3.5 and 4.1 m. The tubes are labelled
U75-5, U75-15 and U75-90, respectively.
Characterization tests were carried out to determine the basic properties of the soil.
These tests included electrical conductivity (EC), Atterbergs limits, specific gravity
and particle size distribution tests. The top and bottom ends of each tube were used for
this purpose. The squeezing technique [10] was used to squeeze out the pore fluid from
the soil to determine the EC of the pore fluid. Due to the strong influence of the pore
fluid salinity on the mechanical behaviour of Ballina clay [10], a synthetic solution
prepared at the same EC as that of the natural pore fluid was used to perform the
consistency limits tests as well as the IL tests. Atterberg Limits tests were performed
using soil that was gently pushed through the 425 m sieve to separate the fine fraction.
Liquid limit was determined using fall cone test. The first point was taken at natural
water content. For the two subsequent points, the synthetic solution described above
was added to increase the water content of the paste. The soil was mixed well and left
to cure for 24 hours before performing the subsequent point. Plastic limit was
determined using the thread rolling method according to ASTM D4318-10, while
specific gravity of the soil was determined using water pycnometer method according
to ASTM D854-10 [11, 12]. Particle size distribution was determined using
Micromeritics Particle Size Analyzer. Table 1 summarizes the main characteristics of
the tube specimens tested, whereas Figure 1 shows the related particle size distributions.
Table 1: Basic soil properties of Ballina clay in each tube
Soil Properties U75-5 U75-15 U75-90
Electrical conductivity, EC (mS/cm) 22.54 24.54 19.28
Initial water content, wo (%) 85-87 77-90 90-93
Plastic limit, wp (%) 35.8 36.5 31.3
Liquid limit, wL (%) 100.2 95.7 94.3
Plasticity index, PI 64.2 59.3 63
Specific gravity, Gs 2.63 2.72 2.73
G.T. Lim et al. / Testing Tube Specimens 1099
Figure 1: Particle size distribution of Ballina clay in each tube
3. Experimental Procedures
Each tube sample was X-ray scanned to determine the distribution of shell fragments in
the tube. Two sections for IL testing were identified in each tube: one with a relatively
low volumetric shell fraction, labelled as OD1, and one with a relatively large
volumetric shell fraction, labelled as OD2. The specimens are named according to the
tube they were obtained from, their OCA and shell fraction. For instance, U75-5OD1
represents a specimen from U75 tube with 5 degrees OCA and relatively low
volumetric shell fraction.
3.1. Determination of the volumetric shell fraction
Two methods were used to determine the volume of shell fragments in the soil
specimens, namely CT-scan image analysis and wet sieving. After completion of the IL
test, the specimen was scanned using Xradia Versa Micro-CT (XRM) to obtain three
dimensional images of the specimen. Due to failure of the equipment in the middle of
this study, Bruker Skyscan 1176 In-Vivo Micro CT (Skyscan) was used to scan the
remaining specimens (U75-15OD2, U75-90OD1 and U75-90OD2). The scanning
equipment has a maximum field of view of 50 mm, with a resolution of 49 m for
XRM and 18 m for Skyscan. Hence, the specimen was trimmed to 50 mm diameter
before scanning and the trimming was used for water content test. The scanned images
obtained with XRM and Skyscan were processed using the commercial data
visualization and analysis softwares AVIZO and Brucker CT-Analyser (CTan),
respectively. After scanning, the 50 mm diameter specimen was dried to determine its
water content.
Wet sieving through the 425 Pm sieve was performed separately on the dried 50
mm diameter specimen and on the soil trimmed from the original IL specimen. In this
way, the volume of shell fragments determined from wet sieving and from CTscan
image analysis could be compared (for the 50 mm diameter specimen). The volume of
shell fragments was computed by dividing the dry mass of shell fragments by the shell
density, determined as shell = 2.89 g/cm3.
3.2. Bender Elements and One-Dimensional Compression Tests
Bender elements (BE) tests were performed on tube specimens 55 mm in height, prior
to the IL tests. A special frame was manufactured to hold the bender elements, which
were inserted at the top and base of the specimen. The frame ensured accurate
alignment of the BE transducers, as well as correct determination of the travel length.
BE measurements were taken under two configurations: (i) soil specimen inside the
tube, and (ii) after specimen extrusion. A sine pulse excitation (20 V peak to peak
amplitude) was used with frequencies of 1, 3 and 5 kHz. Shear wave velocity was
determined as Vs = leff / ts, where leff is the effective length (measured between the tips
of the BE) and ts is the travel time, determined from the distance between the first peaks
in the transmitted and received signals [13].
Six IL tests were carried out using 71 mm diameter and 25 mm height fixed ring
consolidometer cells. The ring was pushed directly into the intact specimen after it was
extruded from the tube. The pore fluid in the specimen during incremental loading test
was controlled by using the synthetic pore fluid. This synthetic pore fluid was used to
inundate the consolidometer cells. Each specimen was loaded up to a maximum
vertical stress of 500 kPa, including an unloading-reloading cycle at 250 kPa [14]. A
consolidation time of 24 hours was allowed between each load increment.
4.1. Volumetric shell fraction
3D scans were used to quantify the volumetric shell fraction of the specimens. Image
analysis is an operator-dependent tool and is also subjective to the image quality and
post-processing steps. A consistent approach was used here to remove the artifacts
before segmentation of the images. Segmentation was done manually by thresholding
appropriate grey values to remove the clay from the 2D images. The segmented binary
images were then used to generate the 3D view of the shell fragments distribution.
Using the analysis function in Avizo and CTan, the volume of shell in the 50 mm
diameter central part of the IL specimen was computed. The values are reported in
Table 2 together with the values obtained by sieving. It is apparent that the shell
volume computed by CT-scan analysis tends to be slightly higher than the one
computed by sieving, with the difference remaining relatively small. The difference
may be due to the difficulty in identifying the shell material accurately in the images
through appropriate tresholding of grey values or inaccuracies of the wet sieving
method. Overall, these results indicate that the volume of shell fragments can be
obtained using these two methods. The volumetric shell fraction (V sf) of the
compression specimens (71 mm diameter) was determined according to the following
equation

(1)

where Vshell is the volume of shell in the IL specimen determined by sieving and V IL is
the initial volume of the IL specimen. The values reported in Table 2 indicate that V sf
was relatively small (< 2%) for all the specimens tested, and varied in the range 0.21 to
1.59%. The relative difference between the volumes computed with the two methods is
small if compared with the total volume of IL specimen.
Table 2: Volume of shell fragments and volumetric shell fraction

Volume of shell in 50 mm
Relative difference (%) Volumetric shell
diameter specimen
Sample fraction, Vsf, (%)
(mm3)

Sieving, Vs CT-Scan, VCT Sieving
U75-5OD1 152.25 167.71 10 0.2
U75-5OD2 737.72 652.57 12 0.9
U75-15OD1 369.90 447.65 19 0.7
U75-15OD2 253.98 356.84 18 0.9
U75-90OD1 591.70 602.60 2 0.9
U75-90OD2 1151.90 1037.10 11 1.6
4.2. Shear wave velocity
The shear wave velocity results are reported in Table 3, except for U75-90OD1, for
which no signal was detected by the receiver. The values obtained for the clay
specimens inside the tube and for the specimens after extrusion are very similar, and
are close to the shear wave velocity measured in-situ using seismic dilatometer
(average value of 53 m/s).
The relationship between shear wave velocity (Vs) and volumetric shell fraction
(Vsf) is examined in Figure 2. Despite some scatter, Figure 2 shows a trend of
increasing Vs with increasing Vsf, which can be explained by the higher stiffness of the
shell compared to the soft clay. Figure 2 also seems to indicate that Vs increases with
OCA (compare U75-5OD2 and U75-15OD2 with similar Vsf). However, additional
data is required to confirm this trend. Let us note that shear wave velocity measurement
has been proposed as a non-destructive method for evaluation of sampling disturbance
in homogeneous clay [15]. These results indicate that this method may not be suitable
for clay containing granular inclusions, as Vs is influenced by variations in granular
volume fraction.
Table 3: Summary of bender element test results
Shear wave velocity, Vs (m/s)
Specimens Specimen in the tube After extrusion
U75-5OD1 52.3 52.4
U75-5OD2 41.0 38.6
U75-15OD1 55.3 56.1
U75-15OD2 53.2 49.3
U75-90OD1 Nil Nil
U75-90OD2 65.3 66.4
Figure 2: Shear wave velocity against volumetric shell fraction

4.3. One-dimensional compression tests and sample quality
Compressibility parameters obtained for all specimens are shown in Table 4. Three
methods were adopted to estimate the pre-consolidation stress, cp: Casagrande [16]
based on empirical observation, Becker et al. [17] based on work per unit volume and
Boone [18] based on simple slope-intercept mathematics preventing subjective
interpretation. The first two methods [16], [17] yielded reasonably close values of cp,
whereas the third method [17] tended to give slightly higher values.
Table 4: Summary of compression test results
Pre-consolidation stress, cp Compression Swelling

(kPa) Index, Cc Index, Cs Sample Quality
Specimens
Becker
Casagrande Boone
et al. e/eo Rating
U75-5OD1 36 45 46 1.03 0.14 0.07 Poor
U75-5OD2 32 30 37 0.94 0.11 0.11 Poor
U75-15OD1 39 30 53 1.08 0.19 0.09 Poor
U75-15OD2 31 29 37 0.89 0.14 0.10 Poor
U75-90OD1 45 40 47 1.12 0.13 0.07 Poor
U75-90OD2 55 60 58 0.91 0.10 0.10 Poor
Figure 3: Pre-consolidation stress (Casagrande) against volumetric shell fraction

The values of cp obtained using the method by Casagrande are plotted in Figure 3
as a function of Vsf. Except for two data points with lower cp values, Figure 3 shows
that cp increases with Vsf. This may suggests that the structure of the specimens with
higher Vsf requires higher stress level to be disrupted. In addition, from the data points
with similar values of V sf (Vsf~0.85%), it can be observed that cp increases with OCA.
The increase of cp with OCA is in apparent contradiction with the fact that higher
OCA is expected to induce higher sampling disturbance and thus lower cp. A possible
explanation for this result might be that other factors have affected the quality of the
specimens, which override the effect of OCA such as disturbance from specimen
preparation.
The variations of the compression index (Cc) and the swelling index (Cs) with Vsf
are plotted in Figure 4. For both indices, the variations remain relatively small. No
clear trend can be observed for Cc, whereas there is a tendency for Cs to decrease with
an increase in Vsf. This is consistent with the higher stiffness of the shell compared to
the clay particles.
Figure 4: (L) Compression index against volumetric shell fraction and (R) swelling index against volumetric
shell fraction
The sample quality of the specimen tested was determined according to the
criteria for evaluation of sample disturbance proposed by Lunne et al. [7]. These
criteria are based on the measurement of e/e0, where e is the change in void ratio
taking place in the specimen during reconsolidation to the in-situ effective vertical
stress and e0 is the initial void ratio of the specimen. Values of e/e 0 for all specimens
are reported in Table 4 and plotted in Figure 5 as a function of Vsf. All specimens
classify as poor quality samples (0.07 < e/e0 < 0.14). Figure 5 indicates a trend for an
increase in sample disturbance with an increase in Vsf. However, when considering the
data points with similar value of Vsf (Vsf~0.85%), the results show a decrease in sample
disturbance with an increase in OCA. This observation contradicts previous findings
showing an increase in sampling disturbance with increasing OCA [19]. This may
suggests that other factors may have caused sampling disturbance (e.g., inclination of
tube during penetration) or the results reflect soil variability.
Figure 5: Variation of sample disturbance with volumetric shell fraction
This paper has shown that the presence of inclusions can be assessed quantitatively
using non-destructive methods such as CT-Scan. This method of determining the
volume of shell fragments agreed reasonably well with the conventional technique of
wet sieving. In addition, the results of this study have shown that an increase in the
volumetric shell fraction may cause the shear wave velocity and the pre-consolidation
stress to increase, which indicates that the presence of shells contributes to a higher
stiffness and a more stable soil structure. The effects of the amount of shells on the
compression and swelling index were less significant, with a tendency for the swelling
index to decrease (i.e., increased stiffness) with an increase in volumetric shell fraction.
Based on the limited amount of data, no clear relationship could be established between
the sampling tube outer cutting angle and the various measured parameters. Further
testing will be conducted on specimens obtained with different types of sampler (e.g.,
fixed piston sampler, Sherbrook block sampler) to investigate the effects of sampling
disturbance on the measured properties of the soft natural clay with shell inclusions.
References
[1] Lupini, J.F. Skinner, A. E. and Vaughan, P.R. The drained residual strength of cohesive soils.
Geotechnique, 1981.
[2] Graham, J. Saadat, F. Gray, M. N., Dixon, D. A. & Zhang, Q.-Y. Strength and volume change
behaviour of a sand-bentonite mixture. Canadian Geotechnical Journal, 1989.
[3] Kumar, GV, and D Muir Wood. "Fall Cone and Compression Tests on Clay-Gravel Mixtures."
Geotechnique, 1999.
[4] Yin, J.-H. Properties and behaviour of Hong Kong marine deposits with different clay
contents.,Canadian Geotechnical Journal, 1999.
[5] Clayton, C. R. I., and A. Siddique. "Tube Sampling Disturbance-Forgotten Truths and New
Perspectives." Paper presented at the Proceedings of The Institution of Civil Engineers Geootechnical
Engineering, 2001.
[6] DeGroot, Don J., Steven E. Poirier, and Melissa M. Landon. "Sample Disturbance--Soft Clays." Studia
Geotechnica et Mechanica, 2005.
[7] Lunne, Tom, Toralv Berre, Knut H. Andersen, Stein Strandvik, and Morten Sjursen. "Effects of Sample
Disturbance and Consolidation Procedures on Measured Shear Strength of Soft Marine Norwegian
Clays." Canadian Geotechnical Journal , 2006.
[8] Horng, Vuthy, Hiroyuki Tanaka, and Takashi Obara. "Effects of Sampling Tube Geometry on Soft
Clayey Sample Quality Evaluated by Non-destructive Methods." Soils and foundations, 2010.
[9] Bishop, DT. Engineering Geology of the Richmond River Estuary. PhD Thesis, University of
Newcastle, 2009.
[10] Pineda, JA, R Kelly, L Bates, D Sheng, and S Sloan. "Effects of Pore Fluid Salinity on the Shear
Strength of a Soft Clay. In. Proc. 5th Biot Conference in Poromechanics, Vienna, Austria, Hellmich,
Pichler & Adam (eds), ASCE, Reston, 2013.
[11] Standard, ASTM. "D4318-10. Standard Test Methods for Liquid Limit, Plastic Limit, and Plasticity
Index of Soils." Annual Book of ASTM Standards, 2010.
[12] Standard, ASTM. "D854-10." Standard Methods for Specific Gravity of Soils. ASTM International,
West Conshohocken, Pennsylvania, 2006.
[13] Viggiani G. and Atkinson J.B. "Interpretation of bender element tests." Geotechnique, 1995.
[14] Standard, ASTM. "D2435-11. Standard Test Methods for One-Dimensional Consolidation Properties of
Soils Using Incremental Loading." Annual Book of ASTM Standards, 2011.
[15] Donohue, Shane, and Michael Long. "Assessment of Sample Quality in Soft Clay Using Shear Wave
Velocity and Suction Measurements." Geotechnique, 2010.
[16] Casagrande, Arthur. "The Determination of the Pre-Consolidation Load and Its Practical Significance."
Paper presented at the Proceedings of the international conference on soil mechanics and foundation
engineering, 1936.
[17] Becker, DE, JHA Crooks, K Been, and MG Jefferies. "Work as a Criterion for Determining in Situ and
Yield Stresses in Clays." Canadian Geotechnical Journal, 1987.
[18] Boone, Storer J. "A Critical Reappraisal of Preconsolidation Pressure Interpretations Using the
Oedometer Test." Canadian Geotechnical Journal, 2010.
[19] Clayton, C.R.I., Siddique, A. and Hopper, R.J. "Effects of sampler design on tube sampling disturbance
numerical and analytical investigations." Geotechnique, 1998.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1105
Simple vs Advanced Interface Model:

A Comparison Using a
Deterministic Quality Approach
H. Stutza,1, G. Mortara b and F. Wuttke a
a
Marine and Land Geomechanics & Geotechnics, Kiel University, Germany
b
University of Reggio Calabria, Italy
Abstract. Various constitutive models for the contact between soils and structural
elements have been proposed in the past. These models are important for holistic
modelling geo-structures like piles, anchors, retaining walls etc. Different soil-
structure interface models using different mathematical formulation have been
defined and some of these are analyzed in this publication using a deterministic
model quality approach. Experimental data from the literature serves as validation
and verification for the models used in this paper.
The different experimental data are related to Constant-Normal-Load and
Constant-Normal-Stiffness interface tests carried out with the simple shear and
direct shear interface apparatuses. These are the most appropriate experimental set
up used for the realistic estimation of the displacement analysis of soil-structure
interfaces. The paper concludes with benefits and drawbacks identified after the
comparison.
Keywords. Soil-Structure Interface Testing, CNS tests, Constitutive Modeling,

Shear displacement behaviour
Introduction
Soil-structure interface behaviour is important for the displacement analysis of a

geotechnical structure. Examples of the great influence in such geo-structures are
retaining walls [1], reinforced soils [2], piles [3] etc. Starting from the experimental
fundamental research starting in the 1960's [4] a lot of different experimental tests were
conducted. This paper analyzes some of these results restricting the analysis to the
behaviour of granular materials. On the basis of different experimental results several
researchers developed numerical constitutive laws to model the frictional behaviour of
soil-structure interfaces. However, quite few literature exists (e.g. [3]) where some of
these different constitutive models are examined and validated using experimentally
data as a qualitative comparison. In this paper a deterministic methodology is presented
by using the experimental data as reference values. This can be used to asset the quality
of soil-structure interface models.
1
[email protected], Kiel University, Institut for Applied Geo-Science, Ludewig-Meyn-Str. 10, 24118
Kiel, Germany
1106 H. Stutz et al. / Simple vs Advanced Interface Model
1. Experimental Observations
A high variety of experimentally data of soil-structure interfaces exists where the

major influences on the experimental behaviour are shown. The most common can be
classified and explained from literature.
The focus here is on non-crushable sands. Crushability is another important feature
of soil-structure interfaces that is particularly important when cyclic loading conditions
are imposed. In this paper only results related to monotonic loading are analyzed.
1.1. Testing procedures
There exists three different testing methods that are used to determine the
deformation behaviour of soil-structure interfaces.
x Constant-Normal Load test (CNL): stress increment in the normal direction,
, confining stiffness ;
x Constant-Normal Stiffness test (CNS): confining stiffness
x Constant-Volume test (CV): confining stiffness , vertical deformation
.
For the classification, it can be stated that, at least under monotonic conditions, the
upper and lower bounds in terms of shear stress can be measured with the CV and CNL
tests, respectively. CNS condition can also be described as a confined dilatancy
condition and is recognized as the most realistic test condition for soil-structure
interfaces.
1.2. Deformation behaviour of soil-structure interfaces - Key aspects
Figure 1. Key aspects for the deformation behaviour of geo-structure interfaces (modified after [6])
In this section, the factors having a major influence are listed and discussed. The
volumetric behaviour is of major interest regarding the accurate modelling of the thin-
interface zone. Due to the aptitude of the soil to develop dilative behaviour, the
coupling between the shear and normal behaviour becomes important. Pullout tests of
cylindrical anchors in granular materials [7] and direct shear tests on sand [8, 9, 10]
demonstrates that dilatancy is the key factor for interface soil-structure interaction
H. Stutz et al. / Simple vs Advanced Interface Model 1107
(Figure 1). The volumetric behaviour is closely linked with the roughness of the
surface. Particle Image Velocity measurements (PIV) shows that the particles have two
different patterns, rolling and sliding [5, 11]. Smooth surfaces predominantly cause
sliding of the particles on the surface whereas rough surfaces cause both sliding and
rolling. Sliding and rolling on rough surfaces influence to a great extent the interface
dilative behaviour which is coupled to the peak shear stress of the soil-structure
interface. The volumetric behaviour of the interface is coupled with the variation of
void ratio during the different stages of displacement [11]. Figure 1 and 2 efficiently
show the important coupling between dilation and confining stiffness, which is
common for soil-structure interfaces.
Figure 2. Behaviour of a soil-structure interface (modified after [7])
Other key aspects of interface behaviour are:

x Grain size
The mean diameter d50 plays a significant role for the shear band thickness.
The shear band thickness is approximately 5-10 d50 as studied in [11]. The
shear band thickness is important due to the influence of the volumetric
displacement variations in the thin zone separating the surface from the
surrounding soil. Thus, the particle size is of great importance for the normal
stress variation observed in interface tests under CNS conditions.
x Normal stress
An increasing normal stress leads to a smaller dilation of the soil around the
structure.
x Normal stiffness
An increasing normal stiffness generally leads to an increasing normal stress
around the interface (Figure 1). This is generally observed under monotonic
conditions.
x Density of the soil
As it can be observed from triaxial test performed at low confining stress, also
for interfaces, the higher is the initial relative density, the higher will be the
peak stress value and the maximum rate of dilation.
x Surface roughness
As described above, two patterns of behaviour can be observed, sliding and
rolling of the particles [5, 11]. Therefore, the volume-change of the soil is
strongly linked to the surface roughness. Different results related to smooth
and rough interfaces [12,13,14] show that smooth interfaces almost behaves as
perfectly plastic, whereas rough interfaces tend to produce a peak stress state.
2. Numerical Modelling of Soil-Structure interfaces
The numerical modelling of soil-structure interfaces can be classified in two major

approaches according to whether volumetric-shear coupling occurs. The classification
of shear-volumetric coupled models can be done by inspecting of the mathematical
formulation of the elasto-plastic matrix

(1)

If the off-diagonal terms are zero (i.e. Kns=Knt=Kst=0), no shear-volumetric

coupling occurs.
In this paper 4 different interface models are used for the constitutive description of the
contact interface.
x Mohr-Coloumb friction model [15]
The Mohr-Coloumb Model (MC) is nowadays the most used model for the
representation of contact behaviour in geotechnical engineering. It is a linear-
elastic perfectly plastic model that needs 5 parameters. The model applied
here can be handled with a variable surface friction which is inserted by a
simple coefficient R as described in [16].
x Hyperbolic Interface model [17]
The hyperbolic model (HM) is a non-linear elastic model that needs 4
constitutive parameters to be defined. Differently from MC, it allows to model
non-linear behaviour.
x Elasto-plastic interface model by Mortara [19,9]
The elasto-plastic interface model by Mortara (EP1) is based on a
hardening/softening law and a double stress-dilatancy relationship. It needs 12
parameters that are determined from CNL test results.
x Elasto-plastic interface model from Lashkari [3]
This elasto-plastic model (EP2) needs 9 parameters and is based on the Sani-
Sand continuum soil model by Manzari and Dafalias [18].
3. Experimental Deformation behaviour versus Numerical Modelling
For an objective comparison and an objective measure of the quality of the models
used in this study, the experimental shear stresses at different stages are used as
reference. The model variance can be computed according to the following equation:

(2)

where CVmod, is the model variance, exp is the experimental observed shear stress
and i,mod is the numerical computed shear stress. The scalar 1.645 represents the 90%
confidence level between the computed data to the experimental results. The
experimental data serves as reference model. For this analysis, experimental interface
simple shear results on Ottawa sands [21] are used in Figure 3. The experimental
results reported by [22] are instead shown in Figures 4 and 5. These tests were
performed using dense (Figure 4) and loose (Figure 5) Hostun sand in a modified direct
shear test. Due to the fact that the shear stress development is related to the shear
displacement, different mobilized shear stresses at different stages are used for the
model variance quantification. These different stages are combined by:

(3)
This combination is chosen according to the most significant phenomenas which can be
observed under different test conditions and soil materials. Figure 3 introduces the
different shear stress points that are used for the different mobilized shear
displacements.
Figure 3. Mobilized shear stress W at different shear displacement levels used for the model variance
quantification
Due to space limitations, here only the results in terms of shear displacement u t and
shear stress are shown.
Figure 4. Experimental versus numerical results for CNL test from [21].
As regards the data related to Figure 4 and 5, the constitutive parameters of the
different models were calibrated using the test data at 100 kPa. In the calibration of
models, the input parameters of the constitutive model can be adjusted until a
satisfactory agreement between experiment and simulation has been established. These
parameters were used later to simulate the tests at a different initial normal stress. In
Figure 6, instead, a set of CNL tests where used for the calibration of the different
models and subsequently, the simulations related to CNS condition were predicted.
Figure 5. Experimental versus numerical results for CNL test dense Hostun sand (experimental data from
[22])
Figure 6. Experimental versus numerical results for CNS tests (k = 1000MPa/m) on dense Toyoura sand
(experimental data from [19])
Table 1. Coefficients of Variation CV for the different tests analyzed in the paper
Set-ups MC [15] HM [13] EP1 [9] EP2 [3]
100 kPa 0.52 0.28 0.04 0.08
Dense Hostun-Sand [22]
200 kPa 0.27 0.17 0.11 0.24
100 kPa 0.36 0.40 0.14 0.19
Dense Toyorua Sand[19]
200 kPa 0.31 0.17 0.03 0.13
100 kPa 0.19 0.20 0.07 0.15
Ottawa Sand [21]
300 kPa 0.57 0.15 0.12 0.22
The results shown in Table 1 highlight the variance obtained with the different
approaches. The overall analysis shows that MC and HM models provide an
approximate description of the observed behaviour, especially when CNS conditions
are involved. On the contrary, both the elasto-plastic advanced models EP1 [9] and EP2
[3] exhibit a better match with the experimental results.
4. Conclusion & Discussion
From the previous investigation it is observed that the choice of the constitutive model
can have a major influence on the result of numerical computations related to interface
behaviour. From a quantitative view point, the predictions obtained using the Mohr-
Coulomb and the Hyperbolic models can be very far from the observed results.
Elastoplastic advanced models, instead, as for example those by [3] and [9], allows a
better description of the interface behaviour thanks to the more accurate prediction of
volumetric-deviatoric coupling.
It is observed that the qualitative investigation could produce even worst quality
indexes for MC and HM if also the normal stress component had been used for the
investigation.
The above investigation has shown that the evaluation of different constitutive interface
models is of major importance for realistically predicting the performance of
geotechnical structures.
References
[1] Day, R., & Potts, D. The effect of interface properties on retaining wall behaviour. Int. J. Numer. Anal.
Meth. Geomech. 22 (1998), 10211033.
[2] Gens, A., Carol, I. & Alonso, E. E. (1988). An interface element formulation for the analysis of soil-
reinforcement interaction, Computers and Geotechnics 7 (1988), 133151.
[3] A. Lashkari, Prediction of the shaft resistance of non-displacement piles in sand, Int. J. Numer. Anal.
Meth. Geomech. 37 (2013), 904931.
[4] Potyondy, J. G. Skin Friction between Various Soils and Construction Materials, Gotechnique, 11
(1961), 339353.
[5] Uesugi, M. & Kishida, H., Frictional resistance at yield between dry sand and mild steel. Soils and
Foundations, 26 (1986), 139149.
[6] Nakayama, H. Modelling interfaces between sand and structural elements. Ph.D. Thesis, Department of
Civil Engineering, Faulty of Engineering, University of Bristol. 2006
[7] Wernick, E. (1978). Skin Friction of Cyclindrical Anchors in Non-Cohesive Soils. In Symposium on Soil
Reinforcing and Stabilising Techniques (pp. 201219).
[8] Tabucanon, J. T., Airey, D. W. & Poulos, H. G. (1995). Pile Skin Friction in Sand from Constant Normal
Stiffness Test. Geotechnical Testing Journal, Vol. 18, 350364.
[9] Ghionna, V. N. & Mortara, G. An elastoplastic model for sand - structure interface behaviour.
Gotechnique, 52(1) (2002), 4150.
[10] Lings, M. L. & Dietz, M. S. (2005). The peak strength of sand-steel interfaces and the role of dilation
Soils and Foundations, 45, 114.
[11] DeJong, J. T., White, D. J. & Randolph, M. (1989). Microscale oberservation and modeling of soil-
structure interface behaviour using particle image velocimetry. Soils and Foundations, 37, 32293235.
[12] Hu, L. & Pu, J. (2004). Testing and Modeling of Soil-Structure Interface. Journal of Geotechnical and
Geoenvironmental Engineering, 130(8), 851860. doi:10.1061/(ASCE)1090-0241(2004)130:8(851)
[13] Mortara G., Mangiola A. & Ghionna V.N. (2007). Cyclic shear stress degradation and post-cyclic
behaviour from sand-steel interface direct shear tests. Canadian Geotechnical Journal 44(7), 739-752
[14] Mortara G., Ferrara D. & Fotia G. (2010). Simple model for the cyclic behaviour of smooth sand-steel
interfaces. Journal of Geotechnical and Geoenvironmental Engineering, ASCE 136(7), 1004-1009.
[15] Van Langen H. (1991) Numerical Analysis of Soil-Structure Interaction. Ph.D. Thesis, TU Delft
[16] Wehnert M. (2006) Ein Beitrag zur drainierten und undrainierten Analyse in der Geotechnik. Ph.D.
Thesis, Institut fr Geotechnik, University Stuttgart
[17] Clough, G. W. & Duncan, J. M. (1971). Finite Element Analyses of Retaining Wall Behaviour. Journal
of Soil Mechanics and Foundation Division, (12), 16571673.
[18] Manzari, M. T. & Dafailias Y.F. A critical state two surface plasticity model for sands. Geotechniqu
47 (2), 255-272
[19] Mortara G. (2001). An elastoplastic model for sand-structure interface behaviour under monotonic and
cyclic loading. Ph.D. Thesis, Dipartimento di Ingegneria Strutturale e Geotecnica, Politecnico di Torino.
[20] Most, T. (2011). Assessment of structural simulation models by estimating uncertainties due to model
selection and model simplification. Computers & Structures, 89(17-18), 16641672.
doi:10.1016/j.compstruc.2011.04.012
[21] Evgin, E. & Fakharian, K. (1996). Effect of stress paths on the behaviour of sandsteel interfaces.
Canadian Geotechnical Journal, 33, 853865. doi:10.1139/t96-116-336
[22] Shahrour, I. & Rezaie, F. (1997). An elastoplastic constitutive relation for the soil-structure interface
under cyclic loading. Computers and Geotechnics, 21(1), 2139
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1113
Capillary Rise in Corners of Rough Pores:

Laboratory and Theoretical Analyses
Shirley TANGa, Luis E. VALLEJOb,1, and Bernardo CAICEDO c
a,b
Graduate Student and Professor, Department of Civil & Environmental Engineering,
University of Pittsburgh, USA
c
Professor, Departmento de Ingenieria Civil & Ambiental, Universidad de los Andes,
Bogota, Colombia
Abstract. The capillary rise in pores forming part of soils and rocks is usually
obtained assuming the pores are smooth circular capillary tubes. The water rises in
these tubes to a height called the height of capillary rise. This height is easily
calculated using the surface tension and the unit weight of the water, the contact
angle, and the diameter of the circular tube. However, pores in soils and rocks are
not circular nor smooth but rough. This study presents the results of laboratory and
theoretical analyses designed to calculate the height of capillary rise in tubes with
rough walls. To obtain this height, laboratory experiments were conducted on the
height of capillary rise experienced by two glass surfaces meeting each other at a
corner and making a wedge. The wedge represents the sharp corners of a rough
pore. The angle between the two surfaces making the wedge was varied between
1oandRThe height of capillary rise was obtained at various distances from the
corner of the wedge. It was found that the height of capillary rise increases as the
distance from the corner of the wedge decreases. Also, as the angle of the wedge
increases, the height of capillary rise decreases in the wedge. These findings were
used to interpret the way capillary water moves in the rough pore system of soils
and rocks (shales). The profile of the pore walls in soils and shales were obtained
from thin sections of soil and rock samples. Using the way capillary water moves
in the sharp corners of rock voids, the non-slaking of shales was explained.
Keywords. Unsaturated rocks, rough pores, capillary rise, laboratory tests,

theoretical analysis, fractals.
1. Introduction
When a smooth circular tube of small diameter, d, is immersed in a container with

water, the water rises in the tube to a height called the height of capillary rise, h (Fig. 1).
The height of capillary rise, h, is easily calculated and is equal to:
h = (4Ts cos TJw d) (1)
In this equation, Ts is the surface tension of the water, Tis the contact angle, and Jw is
the unit weight of the water. However, pores in soils and rocks are not circular nor
___________________
b,1
Corresponding Author, Department of Civil & Environmental Engineering, 726 Benedum Hall, University
of Pittsburgh, Pittsburgh PA 15261 USA. [email protected]
1114 S. Tang et al. / Capillary Rise in Corners of Rough Pores
Figure 1. Capillary rise in a smooth circular tube of diameter d.
smooth but rough. The present study presents a method to calculate the height of
capillary rise in tubes with rough walls. The implications of the results of this study are
extended to analyze the slaking of shales.
2. Capillary Rise in Rough Tubes
The pore system in soils and rocks have boundaries that are rough and do not resemble
the smooth circular profile of the tube shown in Fig. 1. An example of how the
boundaries of the pores in a sample of shale from Tennessee looks like is shown in
Fig.2. This figure (obtained from thin section analysis) shows that the profiles of the
pores in the shale are made of jagged irregularities of different dimensions and aperture
angles. Little is known about how these irregularities of the pore profiles influence the
movement of water as a result of capillarity forces and how these forces influence the
slaking of shales.
Figure 2. Shape of pores in a shale sample from Tennessee
2.1 The rise of capillary water in a sharp corners, theoreticaland laboratory analyses
When a capillary tube with either square or triangular cross section as shown in Fig. 3
is immersed in water, the water advance in the capillary tube up to a distance h, after
which the water advances to very large distances following the corners of the tube [1].
S. Tang et al. / Capillary Rise in Corners of Rough Pores 1115
Figure 3. Capillary rise of water in corners of square and triangular cross sectional areas
The distance h can be obtained from Eq. (1), however, d, for the case of Fig. 3, is the
radius of the inscribed circle in the square or triangular cross sectional areas forming
part of the rough tube.
In order to explain why the water reaches very large distances in the corners of the
rough tube, a laboratory investigation developed by Greenslade [2] is presented next.
He conducted a laboratory experiment that consisted in immersing in water a wedge
made of two glass plates as shown in Fig. 4. The angle of the wedge formed by the two
glass plates was varied in the experiments. According to Greenslade [2], the water
meniscus in the wedge formed a curve (Fig. 4). The height of the curve reached a
maximum value at the corner of the wedge and decreased in height as the meniscus
moved away from the corner of the wedge. The height of the capillary rise between the
glass plates depended of the angle between the plates.
Figure 4. Capillary rise in a wedge formed by two glass plates

Figure 5. A wedge with inscribed circles of decreasing diameters and filled with capillary water.
Lets assume the area ABO is filled with capillary water. The height of the
capillary water in the wedge in each of the inscribed circular tubes of varying diameter
d in Fig. 5 can be obtained from the following relationship,
h = (4Ts cos TJw d) (2)
The arch s in Fig. 5 is approximately equal to the diameter of the circles tangent to
the sides of the wedge , and can be obtained from,
s d = Dx (3)
After replacing the value of the variable d (diameter of circles) given by Eq. (3)
into Eq., (2), we obtain,
h = (4TscosTJw Dx ) (4)
Assuming the value of the contact angle, T, between soils or glass with water to be
equal to zero, Eq. (4) becomes,
h = (4TsJwDx
If one analyzes Eq. 5, it is determined that the value of h in the cylindrical tubes
between the two glass plates shown in Fig. 4 depends of Dand x (Fig. 5). The value of
h in the cylinder located in the corner of the wedge reaches extremely high values
because either Dor x have very small values. In fact, as the value of Dor x tends to a
zero value, the value of h tends to an infinite value as shown in Fig. 3.
If one considers the space between three equal grains, an evolution of the inscribed
circles belonging to capillary tubes is shown in Fig. 6. This figure shows that when the
pore space has been filled with circles of smaller and smaller diameters, the inscribed
circles in the corners of the void space are very small when compared with the ones
that were inscribed away from the corners. The distribution of the inscribed circles is
fractal in nature [6]. If capillary water moves in the inscribed capillary tubes shown in
Fig. 6, the water will rise to higher elevations in the tubes in the corners than in the
tubes located in the center of the void. Thus, the distribution of the water levels in the
capillary tubes shown in the last figure forming part of Fig. 6 resembles the elevation
acquired by the water in Fig. 4. In this figure, the elevation of the water is the highest at
the corners of the wedge and decreases as the distance X increases from the corners of
the wedge (Figs. 4 and 5).
Figure 6. The evolution of inscribed circles of different diameters in the void formed by three large circles
of equal diameter.
In order to further determine the validity of Eq. (5), experiments similar those
reported in the literature conducted were carried in the laboratory [2]. Two square glass
plates of dimensions equal to 25 cm were used. The glass plates were joined at their
corners and were immersed in water. The opening angle, D of the glass plates were
varied between 1o and 30o. Fig. 7 shown the actual values of the height of the capillary
water (h) in between the glass plates corresponding to values of Dequal to 2o and 6o.
Also in Fig. 7 are shown the theoretical or calculated values of h obtained using Eq.
(5). This figure shows that the laboratory and calculated values of h were somewhat
different in the corners of the wedge formed by the glass plates. However, the
calculated and actual values coincided at locations away from the corner of the wedge.
It should be noted that the calculated value of h at a distance x= 0 is infinite if one
follows Eq. [5] (Fig. 7). The calculated values of h shown in Fig. 7 are effective at
distances very close to the corner of the wedge (Fig. 5].
The previous findings are used to interpret the mechanisms involved with the
slaking or non-slaking of shales when in contact with water.
Figure 7. Theoretical and laboratory values of the height of capillary rise in the glass plates for two different
values of Dand distance X from the corner of the wedge made by the plates.
3. Mechanics of the Slaking and Non-slaking of Shales
3.1 Slaking of Shales by Air Compression
There are different mechanisms discussed in the geotechnical literature which explain
the slaking of shales when immersed in water [3]. One slaking phenomenon is
attributed to the compression of entrapped air in the pores of the shales when water
enters them as a result of capillary suction [4].This entrapped air in the pores exerts
tension on the solid skeleton, causing the material to fail in tension. Pore-air
compression is the predominant slaking mechanism in shales composed primarily made
of kaolinite [5].
The mechanism for pore-air compression that breaks the shales has been
previously discussed in the geotechnical literature and is shown in Fig. 8 [3]. Fig. 8
shows a shale sample with a system of cylindrical macro-pores that run continuously
through it [Fig 8(A). These macro-pores which are assumed to be not connected,
resemble small cylindrical tubes inside the shale.
Figure 8. Mechanics of slaking of shales by air compression
When the sample is immersed in water, water will be pulled into the individual
macro-pores as a result of capillary forces, and the air that originally filled the macro-
pores will be subjected to compression [Fig 6(B)]. The system of forces acting at the
inter-face between the air and the water in a macro-pore are presented in Fig. 6(C). At
equilibrium conditions the following relationship applies [3]:
p = u + (4Ts/Jw d) (6)
where, d is the diameter of the cylindrical macropore, Ts is the surface tension of water,
p is the air pressure, and u is the pore water pressure at the level of the macropore.
An analysis of Eq. (6) indicates that the pore air pressure, p, in the portion of the
macro-pore filled with air [Fig. 8(B)] increases as the diameter, d , of the cylindrical
macro-pore decreases. Thus, the smaller the diameter of the macro-pore, the larger is
the pore air pressure, p. Since pore-air compression is favored by small pore radii,
slaking of the shales by air compression will be more pronounced in those shales
containing macro-pores with small diameters. Thus, the diameter of the macro-pores in
shales has a marked influence on their slaking with water.
To check the validity of Eq. (6), sixty eight shale samples from Appalachia were
tested for their durability when immersed in water. Of the 68 shale samples, 14 samples
had degraded into soil, and fifty four of the samples remained unchanged. The diameter,
d, of the shale samples that slaked was smaller than 0.06 mm. However, among the
samples that did not slake, several shale samples had pores with diameters of less than
0.06 mm. Thus, the size of the pores is not sufficient to discriminate between slaking
and non-slaking shales.
An example of a shale sample that did not slake and had an average diameter of the
pores of less than 0.06 mm is shown in Fig. 2. This figure shows a shale sample from
Tennessee (TN-7). The average diameter of the pores of this sample was equal to 0.056
mm and the boundaries of the pores were very rough. The degree of roughness of the
pore boundaries was measure using the fractal dimension concept [6]. The fractal
dimension of the pores of sample TN-7 was equal to 1.7767. This high value of the
fractal dimension represents a pore boundary that is very rough (Fig. 2).
Based on the findings of how capillary water moves through corners of the void
system of rocks, an explanation of why shales with small (d < 0.06 mm) but rough
pores (D > 1.425) do not slake is shown in Fig. 9 [3].
Figure 9. The advancement of capillary water in a rough pore
When a capillary tube with either square or triangular cross section as shown in Fig.
3 is immersed in water, the water does not completely advance in the capillary tube
following the whole cross sectional area, but advances to large distances in the tube
following the corners and crannies of the tube [1, 3]. This partial filling of the tube
cross sectional area will prevent the development of air pressure that is necessary to
cause the slaking of the shales [Fig 8(B)]. Fig 9 shows how water advances in a rough
pore. The water at the extreme portions of the macro-pore covers its whole cross
sectional area, however, after a short distance from the ends of the macro-pore, the
water follows the corners and crannies of the pore walls (Figs. 2) This capillary water
movement through the corners of the pore walls will prevent the development of the air
pressure that is required to cause the breakage of the shale [Fig. 8(B), Eq. (6)]. Thus,
the roughness of the pore boundaries has a significant influence on the slaking or non-
slaking of shales.
For the case of three dimensional water flow in shale pores, the total roughness of
the tortuous surface of the pore channels can be obtained by adding 1 to the fractal
dimension of the pore boundaries [7]. Thus, pore channels with small pore diameter (d
< 0.06 mm) and surface fractal dimension, D < 2.425, will slake as a result of capillary
suction [3, 7]. Pore channels with diameter of any size but values of D > 2.425 will not
slake as a result of water suction [3, 7].
. Conclusions
From this study the following conclusions can be made:

(1) The degree of roughness of the boundaries in capillary tubes influences the way
capillary water advance in the tubes.
(2) Shales with small pore diameters (d<0.06 mm) and smooth pore boundaries (D
< 1.425) were found to slake as a result of capillary suction. However, shales with
rough pore boundaries (D >1.425) and diameters of any size were found not to slake.
The rough pores in the shales allowed capillary water to move through the corners and
crannies of the pores. Thus, the capillary water did not move through the whole cross
sectional area of the rough pores. This type of water movement prevented the
development of pore air compression that is needed for the slaking of shales.
References
[1] T.C. Ransoff, & C.J. Radke, Laminar flow of a wetting liquid along corners of predominantly gas
occupied non circular pores, Journal of Colloidal and Interface Sciences, 121 (1988), 391-401.
[2] T.B. Greenslade, Capillary phenomenon, The Physics Teacher, 30 (1992), 300-301.
[3] L.E. Vallejo, Mechanics of the slaking of shales, Geomechanics and Engineering, 3 (2011), 219-231.
[4] K.Terzaghi, and R.B. Peck, Soil Mechanics in Engineering Practice. John Wiley & Sons, New
York, 1968
[5] Y. Moriwaki, Causes of Slaking of Argillaceous Materials. Ph.D. Dissertation, Dept. of Civil
Engineering, University of California at Berkeley, 1974.
[6] B.B. Mandelbrot, Fractals: Form, Chance and Dimension. W.H. Freeman, San Francisco, 1977.
[7] G. Korvin, Fractal Models in the Earth Sciences, Elsevier, Amsterdam, 1992.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1121
Geotechnical Characterization of a
Pyroclastic Sand and Tuff
Pedro A. COVASSIa,1, Marcelo E. ZEBALLOS a and Sofa I. GOROSITO b
a
Facultad de Ciencias Exactas, Fsicas y Naturales. Universidad Nacional de Crdoba
b
Undergraduate Student F. C. E. F. y Nat. Universidad Nacional de Crdoba
Abstract. This paper reports the geotechnical characterization of pyroclastic sand

and a tuff formation, both finding at the Valley of Pichin Leuf River, Ro Negro
province, Argentina. The geological origin of these materials is the precipitation of
volcanic ash. The materials studied present singular mechanical characteristics.
Tests results show that both materials present a low unit weight and low plasticity,
but high liquid limit, high compressibility and high strength parameters. The aim
of the geotechnical characterization is to determine an exhaustive set of both
stiffness and strength parameters. A series of conventional laboratory techniques,
such as one-dimensional compression, triaxial compression and direct shear tests
were performed in order to determine both deformational and strength parameters
of the pyroclastic sand. On the other hand, unconfined compression tests were
performed on specimens of the tuff rock in order to characterize this material. Also,
a series of direct shear tests in concrete-tuff interfaces were performed in order to
evaluate the skin friction for different of normal stress levels and concrete
roughness. A comparison with other tests results finding at the literature also are
presented.
Keywords. Volcanic tuff, pyroclastic sand, compressibility, shear strength.
1. Introduction
Site under study is located in the North-West sector of irihuau basin, most exactly
about 40 kilometers to the East from San Carlos de Bariloche city, at the intersection of
National Route 23 with Pichi Leuf River, in Ro Negro province. That basin is located
at the oriental subandinean flank of northern Patagonian Andes, extending from Nahuel
Huapi Lake (Ro Negro Province) at the North to Esquel city at the South (Chubut
Province), covering an area of approximately 5360 km2.
Two kinds of materials were studied in this work: a volcanic tuff and pyroclastic
sand formed by the tuff weathering. Both materials belong to the Huitrera Formation
[1]. These materials were formed during the Tertiary period of the Cenozoic era in a
range between 60 5 Ma and 42 5 Ma (i.e. in the upper Paleocene to middle Eocene)
associated with the evolution of the Andean continental margin. The principal outcrops
of this formation are presented in various sectors of Pichi Leuf River where both
materials are presented in alternated tuff and sand layers. Figure 1 shows a schematic
map presenting the site under study location and major formations and geological
structures of the region.
1
1122 P.A. Covassi et al. / Geotechnical Characterization of a Pyroclastic Sand and Tuff
Limay Route 23
River
Site under
Nahuel
Study
Huapi Lake
Pilcaniyeu
San Carlos de
River
Bariloche City
Patagonian Andes
irihuau
River Pichi Leuf
River
Route 40
Pilcaniyeu
Fault References:
Colohuincul Complex
Patagonian Batholith
Huitrera Formation
Catedral Ventana Formation

Fault irihuau Formation
Colln Cura Formation
Figure 1. Schematic map of site under study location.
2. Experimental Work
2.1. Description of Tested Materials
Materials used in the investigation include both natural tuff rock and pyroclastic sand
formed by the tuff weathering.
Physical properties and Unified Soil Classification of pyroclastic sand are
summarized in Table 1. The standard Proctor test for this soil yielded a maximum dry
unit weight of d = 12,6 kN/m3 at the optimum moisture content of w% = 30,3 %.
Table 1. Physical properties and classification of pyroclastic sand tested.
Specific
Passing Passing Passing Clay Surface
(Fraction
Sieve #4 Sieve #10 Sieve #200 Content Passing
LL PL PI [%< 4.75 [%< 2.00 [%< 74 [% < 2 Sieve #200)
[% [%] [%] Gs mm] mm] m] m] [m2/gr] SC
SC-
48.2 46.6 1.6 2.56 99.9 99.1 48.6 8.5 ~ 40
SM
Note: PL: LL: Liquid Limit; Plastic Limit; PI: Plasticity Index; Gs: Specific Gravity; SC: Unified Soil
Classification; NP: Nonplastic.
On the other hand, several 46 mm diameter and 92 mm length cylindrical

specimens of volcanic tuff were obtained by rotary coring from a natural rock block.
Intact rock core specimens were prepared following the ASTM D4543-01 [2] standard.
The ends of specimens were then squared on a band-saw. After that, all the specimens
were placed into a humid room during seven days for moisture equalization. A mean
dry unit weight of d = 15.0 kN/m3 for volcanic tuff was measured from all specimens.
Both high porosity as high silt-clay content makes these rocks possess low dry unit
weight and natural gravimetric moisture content of around 20 %.
P.A. Covassi et al. / Geotechnical Characterization of a Pyroclastic Sand and Tuff 1123
2.2. Laboratory Tests
For pyroclastic sand, consolidated drained triaxial tests, one-dimensional compression

and direct shear tests were performed. For volcanic tuff a series of unconfined
compression and direct shear tests were performed.
The data on compresibility of pyroclastic sand presented in this paper were
obtained from one-directional compression test performed in a 63.5 mm diameter
oedometer. Three specimen dry unit weights were adopted to perform the 1-D
compression tests: 11.3 kN/m3, 12.0 kN/m3 and 12.5 kN/m3. All the tests were carried
out submerged into tap water.
The data on strength of pyroclastic sand presented in this paper were obtained from
triaxial tests and direct shear tests. Two specimen dry unit weights were adopted to
perform CD triaxial compression tests for loose and dense specimens. All samples were
prepared by tamping compaction at optimum moisture content in a three piece split
mold. Isotropic stress consolidation levels of 50 kPa, 100 kPa and 150 kPa were
selected to perform the triaxial compression tests. All tests were carried out under fully
saturation (B values above 0.91 were obtained for all the tests) and drained conditions.
On the other hand, direct shear tests on samples of pyroclastic sand were carried out.
Two specimen dry unit weights were adopted. Also, direct shear tests were carried out
on both fractions passing through and retained on sieve #200 respectively. All these
tests were performed under saturated conditions.
A total of five unconfined compression tests were performed on volcanic tuff
specimens according to ASTM D2938 [3] standard. Also, direct shear tests on
concrete-tuff interfaces were performed in order to evaluate the rock behavior at its
discontinuities. The unconfined compression strength of concrete after seven days of
curing was about 20 MPa. Smooth surfaces were achieved by simple leveling flush of
mortar at the pouring moment. In the other hand, rough surfaces were achieved by
grooving the mortar surface with a thin cutting disc. Grooves were 1.0 mm width and
0.5 mm depth. Offset between two consecutive grooves was 4 mm. In the other hand,
rock specimen was installed in the upper half of shear box following the ASTM D5607
[4] standard. All direct shear tests were performed under saturated conditions.
Table 2 summarizes the characteristics of all tests performed and specimens tested.
Table 2. Summary of specimens tested.
Sample d
Test Material Type Designation [kN/m3]
Pyroclastic Sand TR-Dense 12.5
CD Triaxial Compresion
Pyroclastic Sand TR-Loose 11.5
Pyroclastic Sand OD-Dense 12.5
One-Directional Compression Pyroclastic Sand OD-Medium 12.0
Pyroclastic Sand OD-Loose 11.3
Pyroclastic Sand DS-Dense 12.4
Pyroclastic Sand DS-Loose 11.7
Pyroclastic Sand Pass. Sieve #200 DS-P#200 11.7
Direct Shear Tests
Pyroclastic Sand Ret. Sieve #200 DS-R#200 11.7
Volcanic Tuff -Smoot Concrete DSVT-SC 15.0
Volcanic Tuff-Rough Concrete DSVT-RC 15.0
Volcanic Tuff UC1 15.0
Unconfined Compression Test Volcanic Tuff UC3 15.0
3. Analysis of Main Results
3.1. One-Dimensional Compression Tests
Figure 2 shows the results of oedometer tests carried out on three samples of
pyroclastic sand. Data in Figure 2.a are plotted in terms of void ratio (e) vs. total
vertical stress (v). Relative strains experienced by loose specimen are almost twice the
strain response dense specimen. It is also evident in the plotted values of compression
indexes plotted against the dry unit weight in Figure 2.b. Compression index of loose
specimen is almost twice than compression index of dense specimen. Also, differences
between compression indexes of loose and medium specimen are much larger than
differences between compression indexes of medium and dense specimens. It is
because the higher compressibility of loose specimen is governed by the fine fraction
of soil, instead the lower compressibility of medium of dense specimens is governed by
coarse fraction. Swelling and reloading indexes for the three specimens are also
presented in Figure 2.b. It seems that dry unit weight of specimens has no significant
influence on both type of index.
1,30 0,250
1,25 0,225 Cr
Cc
Compressibility Index (Cr, Cc, Cs)
0,200
1,20 Cs
0,175
1,15
Void Ratio, e
0,150
1,10 0,125
1,05 0,100
1,00 0,075
OD-Dense
OD-Medium 0,050
0,95
OD-Loose
0,025
0,90
0,000
0,1 1 10 100 1000
11,2 11,4 11,6 11,8 12,0 12,2 12,4 12,6
Vertical stress, v [kPa] Dry Unit Weight, d [kN/m3]
Figure 2. 1-D compression tests results. (a) Compressibility curves; (b) compressibility indexes as function
of dry unit weight.
3.2. CD Triaxial Compression Tests
Figure 3a shows the stress-strain curves of CD triaxial tests performed on loose and
dense specimens of pyroclastic sand. Loose specimen had a contractive volumetric
response instead volumetric response of dense specimen was dilative. Figure 3.b shows
the variation of initial moduli (Ei) vs. effective confining stress level. Values of Ei were
computed following the methodology proposed by Duncan and Chang [5]. Also,
effective peak strength parameters obtained from CD triaxial tests are summarized in
Table 3.
Table 3. Summary of effective shear strength parameters of pyroclastic sand obtained from triaxial tests.
Effective Peak Friction Angle Effective Peak Cohession
Specimen *TR [] cTR [kPa]
TR-Dense 39 7.0
TR-Loose 39 0.0
600 70
TR-Dense_50 kPa
TR-Loose_50 kPa
TR-Dense
500 TR-Dense_100 kPa 60
TR-Loose
Deviatoric Stress, 1- 3 [kPa]
TR-Loose_100 kPa
Initial Moduli, E i [MPa]

TR-Dense_150 kPa 50
400 TR-Loose_150 kPa
40
300
30
200
20
100
10
0 0
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0 8,0 9,0 10,0 25 50 75 100 125 150 175
Axial Strain, a [%] Effective confining stress, c [kPa]
Figure 3. Results of CD triaxial compression test on loose and dense specimens of pyroclastic sand. (a)
Stress-strain curves; (b) Variation of Initial Moduli vs. effective confining stress level .
3.3. Direct Shear Tests on Pyroclastic Sand
Table 4 summarizes the strength parameters of piroclastic sand obtained from direct
shear tests. Results indicate that shear strength parameters of pyroclastic sand are
strongly influenced by the finer fraction (silts and clay sized particles). Values of
effective peak friction angles of loose and dense sand specimens are very similar to the
effective friction angle of fraction passing sieve #200. Also, CD triaxial tests show
similar (but lesser) values of peak friction angles.
Table 4. Summary of effective shear strength parameters of pyroclastic sand from direct shear tests.
Specimen *DS [] cDS [kPa]
DS-Dense 41 0.0
DS-Loose 41 0.0
DS-P#200 40 0.0
DS-R#200 49 0.0
Figure 4 shows the shear stress vs. horizontal deformation curves obtained from
direct shear tests over DS-Loose, DS-P#200 and DS-R#200 specimens.
180
3
d = 11.7 kN/m
160
v=150kPa
140
Shear stress [kPa]
120
100
v=100kPa
80
60
v=50kPa
40
DS-Loose
20 DS-R#200
DS-P#200
0
0,0 2,5 5,0 7,5 10,0 12,5 15,0 17,5 20,0 22,5 25,0 27,5
Horizontal strain, h [%]
Figure 4. Stress-strain curves of direct shear tests on pyroclastic sand at d = 11.7 kN/m3.
3.4. Direct Shear Tests on Volcanic Tuff-Concrete Interfaces
Figure 5 shows the shear stress vs. horizontal displacement curves of volcanic tuff-
concrete interfaces (both smooth and rough). There is no significant difference between
responses of rock-smooth and rock-rough interfaces for vertical stress level of 188 kPa.
For vertical stress level of 286 kPa, responses of tuff-smooth concrete and tuff-rough
concrete differ in a considerably manner. Table 5 summarizes the strength parameters
for direct shear tests performed in volcanic tuff-concrete interfaces. Peak strength
parameters for rock-rough interfaces were obtained using only the stress-strain curves
for vertical stresses of 92 kPa and 286 kPa. The stress-strain curve for rock-rough
interface with 188 kPa vertical stress was omitted. There is no appreciable influence of
surface roughness on peak cohesion obtained in both cases, but influence of roughness
in the peak friction angle is most significant. Finally, influence of vertical stresses on
initial slopes of stress-strain curves has no influence for range of stresses applied in this
work.
450
DSVT-SC_92 kPa
400 Direct Shear Tests
DSVT-RC_92 kPa
Volcanic Tuff-Concrete Interfaces
DSVT-SC_188 kPa 3
350 d = 15.0 kN/m
DSVT-RC_188 kPa
300 DSVT-SC_286 kPa

Shear Stress [kPa]
DSVT-SC_286 kPa
250
200
150
100
50
0
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0 8,0 9,0 10,0 11,0 12,0 13,0 14,0 15,0 16,0
Horizontal displacement [mm]
Figure 5. Stress-displacement curves of direct shear tests on volcanic tuff-concrete interfaces.
Table 5. Summary of effective shear strength parameters of volcanic tuff-concrete interfaces from direct
shear tests.
Specimen *DS [] cDS [kPa]
DSVT-SC 32 14.0
DSVT-RC 46 7.0
3.5. UCS Test on Volcanic Tuff
An average value of unconfined compression strength of 4.2 MPa was obtained from
the set of specimens tested.
A number of empirical relationships have been published for estimating the side
resistance of the rock-concrete interface in rock-socketed piles. All are based on studies
of field load tests and laboratory tests and relate socket friction capacity to the UCS of
rock or concrete, generally whichever is weakest. Table 6 summarizes several
relationships between UCS of rock r (MPa) and the ultimate side resistance qs (MPa).
The values of qs reported in Table 6 were obtained from the mean UCS value
determined for the volcanic tuff analyzed in this work. Also allowable side resistance
qs_allow, was obtained using the shaft resistance mobilization factor (partial factor)
fs = 1.5 [6].
Table 6. Relationships between UCS of rock r (MPa) and the ultimate side resistance qs (MPa) reported by
several authors.
qs qs_allow
Author Equation [MPa] [MPa Observations
Rosemberg and Journeux [7] q s 0.375( r ) 0.515 0.79 0.52 -
Horvath and Kenney [8] qs b r 0.51 0.34 0.2 < b < 0.3 (0.25 adopted)
Rowe and Armitage [9] q s 0.45 r 0.92 0.61 For regular (smooth) clean sockets.
Rowe and Armitage [9] q s 0.6 r 1.23 0.82 For clean rough sockets.
0.41 0.27 Only for rock (or concrete) strengths (UCS)
Reese and O'neil [10] q s 0.2 r
higher than 1.9 MPa.
Zhang and Einstein [11] q s 0.4 r 0.82 0.55 Smooth sockets.
Zhang and Einstein [11] q s 0.85 r 1.74 1.16 Rough sockets.
Ng et al. [12] q s 0.19 r 0.39 0.26 Socket length-to-diameter ratios of <3.5.
qs values obtained from the correlations presented in Table 6 present a high degree
of scatter. Also, qs values are almost all higher than values of maximum shear stress
levels obtained in this work for both tuff-smooth and tuff-rough surfaces, showed in
Figure 5. Ng et al. [11] have found similar tendencies when comparing side resistance
of bored socketed piles. These authors allude that differences are because the
correlations in the literature were established using results in mostly sedimentary rocks;
witch achieved generally side resistance than granitic and volcanic rocks. Despite this,
correlations of Reese and Oneill [10] and Ng. et al. [12] are the closer to the order of
magnitude of rock-concrete shear resistance values presented in this work. Finally, it
should be considered that in direct shear tests performed in this work did not take into
account the effects of adherence between concrete an rock, and in the case of piles
tested by the authors cited in Table 6 adhesion effect is present.
4. Conclusions
An experimental study has been developed with the aim of contribute experimental
information about significant geotechnical properties of volcanic tuff and pyroclastic
sand.
Materials studied in the present work were formed in the same geological
environment. Pyroclastic sand constitutes the tuff matrix. Both materials are
characterized by their low dry unit weight and relative high liquid limit, which does not
seem to correspond to the strength parameters would be expected for materials with
these properties.
Deformational behavior of remolded specimens of pyroclastic sand are linked
directly with the deformational characteristic of it solid matrix (finer fraction). This
aspect was demonstrated in both 1-D compression tests as triaxial compression tests.
However this dependence, effective peak friction angles show high values and similar
magnitude for both dense and loose specimens. In this case, without more information
for this research, this behavior can be attributed to soil microstrucrue and particle shape.
High effective peak friction angles were observed in direct shear tests too.
Shear strength parameters identified for sand are similar to those obtained of tuff
when the mechanical behavior on interface planes used in this study is revised.
Acknowledgements
The authors thank the company J.J. Chediack and the Direccin Nacional de Vialidad, to
facilitate the provision of study samples, and for your interest in the proper interpretation of the
behavior of materials.
References
[1] Geological sheet #4172-IV, San Carlos de Bariloche. Servicio Geolgico Minero Argentino
(SEGEMAR), Bulletin No. 279, Buenos Aires, 2002. (In spanish).
[2] ASTM D4543-01, Standard Practices for Preparing Rock Core Specimens and Determining Dimensional
and Shape Tolerances, Annual Book of ASTM Standards, Vol. 04.02, 2001.
[3] ASTM D2938-95 (Reapproved 2002), Standard Test Method for Unconfined Compressive Strength of
Intact Rock Core Specimens, Annual Book of ASTM Standards, Vol. 04.08, 2002.
[4] ASTM D5607-02, Standard Test Method for Performing Laboratory Direct Shear Strength Tests of
Rocks Specimens under Normal Constant Force, Annual Book of ASTM Standards, Vol. 04.08, 2002.
[5] J. M. Duncan and C. Y. Chang, Non-linear analysis of stress and strains in soils, Journal of the soil
Mechanics and Foundationd Division, ASCE Vol. 96, Sm. 5 (1970), 1629-1653.
[6] C. W. W. Ng, T. L. Y. Yau, J. H. M. Li and W. H. Tang, New failure criterion for large diameter bored
piles in weathered geomaterials, Journal of Geotechnical and Geoenvironmental Engineering, Vol.
126, No. 6 (2001), 488-498.
[7] P. Rosemberg and N. L. Journeux, Friction and end bearing tests on bedrock for high capacity socket
design, Can. Geotech. J., 13 (1976), 324-333.
[8] R. G. Horvath and T. C. Kenney, Shaft resistance of rock socketed drilled piers, Proc. Symp. On Deep
Found., ASCE, 1979, New York, 182-214.
[9] R. K. Rowe and H. H. Armitage, A design method for drilled piers in soft rock, Can. Geotech. J., Vol. 24
(1987).
[10] L. C. Reese and M. W. O'neil, Drilled shafts: construction procedures and design methods. Publ. No.
FHWA-HI-8-042, Federal Highway Administration, USA (1988).
[11] L. Zhang and H. H. Einstein, End bearing capacity of drilled shafts in rock, Journal of Geotechnical
and Geoenvironmental Engineering, Vol. 124, No. 7 (1998), 574-584.
[12] C. W. W. Ng, T. L. Y. Yau, J. H. M. Li and W. H. Tang, Side resistance of large diameter bored piles
socketed into decomposed rocks, Journal of Geotechnical and Geoenvironmental Engineering, Vol.
126, No. 8 (2001), 642-656.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1129
Numerical assessment of shear-induced

particle interactions under different loading
conditions by means of axial and torsional
interface shear tests
Alejandro MARTINEZa,1 and J. David FROST a
a
Georgia Institute of Technology, Atlanta, United States.
Abstract. During the past few decades, numerical particulate simulations have
shown that the global and interface shear behavior of geotechnical systems is
significantly impacted by the loading conditions applied and is reflective of a
range of different localized shear behaviors. For instance, experimental and
numerical studies over a longer period of time have shown that the effective
friction angles obtained from triaxial compression, triaxial extension, plane strain
compression, plane strain extension and direct shear are different from each other
(e.g. [1]). This paper presents the results of a study that uses Discrete Element
Method (DEM) simulations to investigate this non-uniqueness of shear behavior
from a particle-to-particle interaction perspective. In particular, investigations of
interface shear simulations in axial and torsional directions in terms of relative
induced particle displacements and rotations, volume changes, principal stress
rotations, coordination numbers and slipping contact fractions allow for the study
of the soil response under different loading conditions. This numerical study is
utilized to compare to the conditions present in bearing capacity failures of
shallow foundations in granular soils, which can reflect compression, extension or
direct shear depending on the exact location within the failure zone or surface.
Keywords. Combined loading conditions, DEM, Interface shear, Soil behavior
1. Introduction
In 1979, Cundall and Strack [2] published their paper on a Discrete Element Modeling
(DEM) method for simulating the behavior of granular materials. Since then, numerous
researchers have used this tool to study the behavior of soils and other particulate
materials. The applications analyzed include soil and interface shear behavior of
spheres and more complex shapes, particle damage, soil dissolution, and the behavior
of structures such as excavations and foundations. Since 1996, there has been a steady
increase in published papers using DEM simulations [3]. The reasons include its
discontinuum approach which can simulate soil behavior more realistically than
continuum methods, as well as its ability to analyze quantities at the particle scale. As
such, several researchers have performed DEM studies to further the understanding of
the micro-scale behavior induced in different laboratory tests, such as plane-strain and
triaxial compression as well as direct and interface shear ([4], [5], [6], [7]).
1
1130 A. Martinez and J.D. Frost / Numerical Assessment of Shear-Induced Particle Interactions
It has long been recognized that the behavior and strength of granular material is
dependent on the boundary condition and loading stress path that they are subjected to.
For instance, the effective friction angles obtained from triaxial compression, triaxial
extension, plane strain compression, plane strain extension and direct shear are
different from each other (e.g. [1]). Thus, recognizing that different loading conditions
can be mobilized simultaneously in the field (e.g. at different locations along a bearing
capacity failure surface), the study of soil response under different conditions is desired.
The latest generation of a series of multi-sensor tools for site characterization
developed at the Georgia Institute of Technology, called the Multi-Piezo-Friction-
Torsion Attachment (MPFTA), offers the ability to induce different soil loading
conditions by means of axial and torsional shear [8]. These tools are capable of
obtaining the relationship between effective interface strength and surface roughness by
means of four friction sleeve sensors equipped with sleeves of varying roughnesses and
piezo-transducers [9],[10]. This paper investigates the ability of the MPFTA to induce
loading conditions similar to those found at different locations within a granular mass
underlying a shallow foundation. The end goal of this research is to investigate the
advantages of utilizing an in-situ tool capable of inducing different loading conditions
rather than analyzing the failure of shallow foundations specifically.
2. Discrete Element Modeling Simulations
2D DEM simulations were performed using the Particle Flow Code (PFC) from Itasca,
Inc. The modeling parameters used (Table 1) were obtained from calibration against
axial and torsional laboratory interface tests, described in detail in [11]. The linear
elastic contact model was used, and all particles were composed of two-particle clumps
(aspect ratio=1.5) which successfully simulate the behavior of Ottawa 20-30 sand [12].
2.1. Axial and Torsional Interface Shear Simulations
Figures 1a and 1b show the configuration of the axial and torsional interface shear
simulations, respectively. All specimens were composed of 8,500 particles and were
generated using the radius expansion technique. The sampling windows labeled 1 and 2
were used to obtain particle-level information. The axial configuration consisted of a
double-sided shear box that is 350 mm high and 55 mm wide on each side. The outer
walls applied constant stress and the inner walls were displaced downward during
shearing. The inner walls consisted of surfaces with a periodic profile with a maximum
roughness, Rmax, of 1.00 mm and an average roughness, R a, of 0.185 mm. The torsional
configuration consisted of a circular chamber of 140 mm in diameter that applied
constant stress conditions. A wall at the center of the chamber (Rmax = 1.00 mm, Ra =
0.185 mm) was rotated clockwise during shearing. All samples were consolidated until
reaching equilibrium using an interparticle friction coefficient of zero in order to create
dense specimens. Figure 1c shows typical results of stress paths as well as peak and
residual failure envelopes for axial and torsional tests obtained from simulations under
different confining stresses. The results show the non-uniqueness of interface strength
and a larger degree of strain softening during torsional simulations. The sample volume
changes during shear yielded maximum dilation angles of 8.2 and 6.4 for torsional
and axial shear, respectively. The larger dilation during torsional shear is possibly the
cause for the larger strain softening.
A. Martinez and J.D. Frost / Numerical Assessment of Shear-Induced Particle Interactions 1131
Table 1. Modeling parameters for DEM simulations

M ean Particle Particle Interparticle Shearing Other Walls Particle Particle Wall Wall
Particle- Clump Density Friction Wall Friction Friction Normal Shear Normal Shear
Clump Size, Aspect 3
Coefficient, Coefficient, Coefficient, Stiffness, Stiffness, Stiffness, Stiffness,
D50 (mm) Ratio, AR (kg/m ) p-p s w kn (N/m) ks (N/m) kn-w (N/m) ks-w (N/m)
7 7 8 8
0.90 1.50 2650 0.45 0.25 0.20 1x10 1x10 1x10 1x10
Figure 1. Configuration of DEM simulations for (a) Axial and (b) Torsional shear. Sampling windows 1 and
2 are used for different particle-level observations. (c) Peak and residual failure envelopes with stress paths
for simulations under a confining pressure of 500 kPa.
2.2. Footing Load Test Simulations
The footing load test simulations were performed using an axisymmetric geometry
(Figure 2a) with a footing-to-particle length ratio of about 40. The domain consisted of
a 300 mm x 300 mm box containing 50,000 particles. The sample was generated by
raining the particles under gravity until reaching equilibrium with an interparticle
friction coefficient of zero in order to create dense specimens. The test was performed
by displacing the footing downwards at a constant velocity of 5 mm/min. Sampling
windows 1, 2 and 3 were used to collect information at different stages of the
simulations. Figure 2b shows the contact forces at the beginning and end of the test.
3. Shear-Induced Particle Deformations
Particle displacement, particle rotation and void ratio fields were generated using
information from all particles and from 1000 measurement circles with a circle-to-
particle radius ratio of 6 evenly distributed throughout the samples. Figures 3a, 3b and
3c show the results for axial and torsional shear at 30 mm of shear displacement. The
results of both simulations show narrow shear zones of about 6 mm in thickness. Axial
shear shows displacement and rotation values as large as 10 mm and 80, respectively,
while torsional shear shows larger values of up to 25 mm and 300, respectively. Both
simulations show a significant amount of dilation within the shear zone, reaching void
ratio values of up to 0.35, as compared to the average initial void ratio of 0.17. These
results are in general agreement with the experimental results presented by [11].
Figures 4a, 4b and 4c show particle displacements, rotations and void ratio fields
for the footing load test simulations at four stages of footing displacement. At 3, 10 and
30 mm, the particle displacements concentrate in a quasi-circular region beneath the
Figure 2. (a) Simulations setup and (b) Force chains at different stages throughout the footing simulation
(different scales are used). Sampling windows 1, 2 and 3 are used for different particle-level observations.
Figure 3. (a) Particle displacement, (b) Particle rotation and (c) Void ratio fields at 30 mm of shear
displacement for axial and torsional shear simulations performed under a confining pressure of 50 kPa.
footing which agrees with the classical definition of a bearing capacity failure. The
particles located directly below the footing undergo larger displacements, of up to 25
mm, and the magnitude of displacement decreases with distance from the footing. The
particle rotation field also shows a concentration of large magnitudes beneath the
footing, with values of up to 150; however, the extent of this area is smaller than for
particle displacements. The void ratio field shows a significant degree of dilation,
reaching values as large as 0.38 directly below the footing. At 10 mm of displacement,
a shear band propagates from the corner of the footing, and at 30 mm of displacement a
shear band that defines the quasi-circular bearing failure is formed. These results show
the similarities between the deformations induced by the interface and footing cases.
4. Fabric and Principal Stress Evolution
Utilizing the sampling windows shown in Figures 1a, 1b (window 1) and 2a (1, 2 and
3), information regarding the magnitude and orientation of contacts, contact normal
forces and contact shear forces was collected. Figures 5a and 5b show normalized polar
histograms for axial and torsional shear at initial conditions and after 30 mm of shear
displacement, respectively. At the initial conditions, the major principal stress for both
simulations is oriented horizontally in the same direction as the applied constant stress,
Figure 4. (a) Particle displacement, (b) Particle rotation and (c) Void ratio fields at different stages of the
footing load test simulation
while no shear forces are present because the interparticle friction coefficient was set to
zero during sample preparation and the fabric shows a near-uniform distribution. At 30
mm of shear displacement, the major principal stress has rotated 35 for both shear
tests. Shear contact forces develop in directions that agree with the direction of the
principal stresses and the fabric evolves with an increasing number of contacts in a
direction of about 80. It should be noted that perhaps the main difference in loading
conditions is that the stress rotation takes place in a vertical plane for axial shear and in
a horizontal plane for torsional shear.
Polar histograms for the footing simulation are presented in Figure 6a and 6b at
initial conditions and at 30 mm of footing displacement for sampling windows 1, 2 and
3. Initially, the principal stresses are oriented mainly in the vertical direction as a result
of gravity, no shear forces are present and the fabric shows near-uniform distributions.
At 30 mm of footing displacement, the major principal stress direction develops in the
vertical direction for sampling window 1, in a near-horizontal direction for sampling
window 2, and inclined about 45 for sampling window 3, which all agree with the
main directions of induced particle displacement vectors shown in Figure 6c. All shear
contact forces developed in directions that agree with the direction of the principal
stresses and there was an increase in the number of contacts in directions aligned with
the major principal stresses for the three windows.
These results show the different soil loading conditions induced by a loaded
footing. During the performance of an in-situ torsional test with the MPFTA probe, the
principal stresses will rotate 90 (if originally oriented vertically as it is common for
normally consolidated deposits), in a similar way as observed in footing window 2.
Similarly, an in-situ axial shear test would result in a principal stress rotation of 35,
which is similar to the magnitude of stress rotation observed in sampling window 3.
These comparisons show the advantage for geotechnical analysis offered by utilizing
in-situ testing tools that capture the soil response under different loading conditions.
Figure 5. Normalized polar histograms from axial and torsional shear simulation of contact normal and
normal and shear contact forces at (a) Beginning of test and (b) 30 mm of shear displacement. (c) Principal
stress rotations during simulations performed under a confining pressure of 50 kPa.
Figure 6. Normalized polar histograms from footing simulation of contact normals and normal and shear
forces at different windows at (a) Beginning of test and (b) End of test. (c) Particle displacement vectors and
sampling windows at 30 mm of footing displacement. Note: the results are presented using different scales.
5. Particle Interactions
Figure 7a shows the average coordination number with shear/footing displacement for
the axial, torsional and footing simulations obtained from the sampling windows. The
results show larger coordination numbers in axial and torsional shear, possibly imposed
by the constant stress conditions. All the results show a decrease in coordination
numbers at displacements of about 0.7 mm as a result of intense dilation during the pre-
peak shear stages. At larger displacements, the results from all footing windows
reached stable values of about 2.75, indicating a critical state fabric. Similar results
were obtained from torsional shear, while axial shear did not create a critical state
fabric since the coordination numbers kept decreasing as a result of continuing dilation.
Figure 7b shows the evolution of average sliding contacts fraction. The results
show similar magnitudes and trends for axial shear and all footing windows. At small
displacements, the sliding contacts fraction increases, and at larger displacements all
footing sampling windows converged to a value of about 0.55, while the axial and
Figure 7. (a) Coordination number and (b) Sliding contacts fraction evolution during the footing and axial
and torsional shear simulations. Note: the axial and torsional shear simulations were performed under
confining pressure of 50 kPa.
Figure 8. Cumulative density functions (CDF) of induced particle rotations. (a) Footing window 1 and
torsional shear and (b) Footing windows 2 and 3 and axial shear. Note: the axial and torsional shear results
are at 20 mm of shear displacement and under a confining pressure of 50 kPa.
torsional simulations converge to values of about 0.48 and 0.19, respectively. In this
regard, axial shear appears to induce a relative magnitude of sliding contact failures
that better simulates bearing capacity failure conditions. Figures 8a shows a
comparison of Cumulative Density Functions (CDF) of induced particle rotations for
torsional shear and footing window 1, which show similar trends. Torsional shear
yielded mean and standard deviation values for particle rotations of 28.8 and 82.8,
respectively, as compared to those for footing window 1 of 32.8 and 56.2. Figure 8b
shows a similar comparison for axial shear and footing windows 2 and 3, which shows
CDFs of similar shape, with average and standard deviation values for axial shear of
10.7 and 53.8. The corresponding values for footing windows 2 and 3 are 2.6 and
37.8 and 1.7 and 24.2, respectively. These comparisons show that torsional shear
induces particle rotation distributions that resemble those induced at locations beneath
the footing, possibly because of the intense dilation rates. Similarly, axial shear induces
particle rotations that resemble those at the locations of sampling windows 2 and 3.
6. Conclusions
This paper has presented a DEM study of the micro-scale behavior of granular
assemblies sheared under different loading conditions. The similarities and differences
between the behavior observed in axial and torsional interface shear tests and various
zones within bearing capacity failures include the following:
x Axial and torsional interface shear tests were shown to induce narrow zones of
concentrated particle displacements, rotations and dilation which resemble the
shear deformation observed in internal soil shear banding.
x Particle displacements, rotations and dilation of different magnitudes and
directions were developed during the footing load test. This confirms the
concept of different testing methods (e.g. lab. tests) recreating the loading
conditions present at different locations within sheared granular masses.
x The loading conditions and soil fabric also varied with location. For instance,
locations beneath the footing did not show principal stress rotations while
other locations showed rotations of up to 90. Torsional tests induced stress
rotations of 90 while axial shear induced rotations of 35, showing that these
tests can induce loading conditions similar to those in sheared soil masses.
x The coordination number at the three footing sampling windows showed a
critical state fabric at large shear displacements. The same trend was shown in
torsional tests, while axial shear did not induce true critical state conditions.
x Axial shear induced similar magnitudes and evolution of sliding contact
fraction as the three footing sampling windows. CDFs of induced particle
rotations were shown to have similar magnitudes and shapes for torsional
shear and locations beneath the footing, while axial shear showed similar
CDFs as other locations within the bearing capacity failure zone.
x The results showed both differences and similarities in the micro-scale
behavior for axial and torsional shear and a loaded footing, underscoring the
importance of making such in-situ testing device (i.e. MPFTA) available to
the geotechnical community.
References
[1] F.H. Kulhawy and P.W. Mayne. Manual on estimating soil properties for foundation design, EPRI Report
EL-6800 (1990).
[2] P.A. Cundall and O.D.L. Strack, A discrete numerical model for granular assemblies, Gotechnique, 29
(1979), 47-65.
[3] C. OSullivan, Particle-based discrete element modeling: geomechanics perspective, Int. J. Geomech., 11
(2011), 449-464.
[4] T.M. Evans and J.D. Frost. Multiscale investigation of shear bands in sand: physical and numerical
experiments. Int. J. Num. Anal. Methods Geomech. , 34 (2010), 1634-1650.
[5] J. Kozicki, J. Tejchman and H.B. Mlhaus. Discrete simulations of a triaxial compression test for sand by
DEM. Int. J. Num. Anal. Methods Geomech. , 38 (2014), 1923-1952.
[6] J. Wang and M. Jiang. Unified soil behavior of interface shear test and direct shear test under the
influence of lower moving boundaries. Granular Matter, 13 (2011), 631-641.
[7] R.P. Jensen, P.J. Bosscher, M.E. Plesha and T.B. Edil. DEM simulations of granular media-structure
interface: effects of surface roughness and particle shape. Int. J. Num. Anal. Methods Geomech, 23
(1999), 531-547.
[8] J.D. Frost and A. Martinez. Multi-sleeve axial-torsional-piezo friction penetration system for subsurface
characterization. 18th ISSMGE Int. Conf. Soil Mech. Geotech. Eng., Paris, FR. (2013), 527-530.
[9] J.D. Frost and J.T. DeJong. In situ assessment of role of surface roughness on interface response. J.
Geotech. Geoenv. Eng., 131 (2005), 498-511.
[10] G.L. Hebeler and J.D. Frost. A multi piezo friction attachment for penetration testing. ASCE
GeoCongress, Atlanta, GA (2006), 1-6.
[11] A. Martinez, J.D. Frost and G.L. Hebeler. Experimental study of shear zones formed in sand-steel
interfaces during axial and torsional shear tests. ASTM Geotech. Test. J. 38 (2015), 1-18.
[12] Y. Lu. Reconstruction, characterization, modeling and visualization of inherent and induced digital sand
microstructures. Ph.D. Dissertation, Georgia Institute of Technology, Atlanta. 215 pp.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1137
Stress-Deformation Modeling of
Geosynthetic-Reinforced Pavement
Structures at Different Water Contents
Rafael BALTODANO GOULDINGa,1 and Paulina LEIVA PADILLA b
a
Civil Engineering Department, University of Costa Rica, and Construction
Engineering Department, Costa Rica Institute of Technology
b
Lanamme-UCR
Abstract. The use of geosynthetic materials to improve the bearing capacity and
deformability behavior of pavement structures is a technique often used by
engineers. The effectiveness of the geosynthetic as a reinforcing material relies on
certain physical mechanisms and some simplifications made about the behavior of
the soil being improved. The transmission of stresses from the soil, bases or
subgrades, to the geosynthetic material depends on the stiffness of the soil. The
compaction process of such soils becomes very important as the stiffness of
compacted soils is such that stresses can be transmitted to the geosynthetic at
relatively low deformations. However, the deformation behavior of soils varies
with the degree of saturation and a soil that is expected to have a stiff material
behavior might exhibit a more ductile deformation behavior as the water content in
the soil increases, or as a double-peak shear strength behavior as a function of
water content and consistency limits is developed. Therefore, less stresses could be
transmitted to the geosynthetic but deformations could be greater, and this can
certainly affect some analysis assumptions. A series of shear strength tests on
different fine grain soils were used to obtain deformability parameters at different
water content but at similar void ratios. These soils were used to represent
subgrade materials in a computer model of a pavement structure. The results of
this model might provide a better understanding of the influence that varying
degrees of saturation, produced by climatic conditions or poor drainage designs,
could have on the deformation behavior of pavement structures reinforced with
geosynthetics.
Keywords. Geosynthetics, Pavement Structures, Unsaturated Soil
1. Introduction
The use of geosynthetic materials to improve the bearing capacity and deformability
behavior of pavement structures is a technique often used by engineers. The
effectiveness of the geosynthetic as a reinforcing material relies on certain physical
mechanisms and some simplifications made about the behavior of the soil being
improved and the transmission of stresses from the subgrade, base or subgrade, to the
geosynthetic material depends on the stiffness of the course or subgrade. The
compaction process of such soils becomes very important as the stiffness of compacted
soils is such that stresses can be transmitted to the geosynthetic at relatively low
deformations, typical of a brittle behavior associated to stiff or dense materials.
1
Rafael Baltodano Goulding
1138 R. Baltodano Goulding and P. Leiva Padilla / Stress-Deformation Modeling
However, the deformation behavior of soils varies with the degree of saturation and a
soil that is expected to have a stiff material behavior might exhibit a more ductile
deformation behavior as the water content in the soil increases, or as a double-peak
shear strength behavior as a function of water content and consistency limits is
developed [1]. Therefore, less stresses could be transmitted to the geosynthetic but
deformations could be greater, and this can certainly affect some analysis assumptions.
Three types of soils are modeled using a finite element software to understand the
influence that variations in the degree of saturation can have in the soil and in the
geosynthetic material produced by changes in climatic conditions and poor drainage
systems. In this paper the effect of changes in degree of saturation in the behavior of
the subgrade-geosynthetic interface are analyzed.
It is important to mention that these changes in water content in the subgrade can
occur due to the water balance which obviously depends on the climate of each region
of the planet. Thus, the analysis of the water content changes is not analyzed as part of
the present work. It is recommended that a water balance analysis be made for
particular conditions being analyzed for real projects. In addition, that the best way to
account for the uncertainty in soil shear strength as a function of degree of saturation is
to perform probabilistic designs.
2.1. Materials
Over 14 types of soils were collected from different parts of the Central Valley in Costa
Rica, Central America. However, only 3 types of soil have been used as subgrade
materials to model the deformation behavior of the interface subgrade-geosynthetics in
pavement structures. The subgrade material properties are described in Table 1.
2.2. Methods
This project was divided in two phases, the first phase consisted on laboratory testing,
and the second phase is related to the computer modeling. As part of the soil properties
characterization needed to obtain design parameters, laboratory testing was conducted
at the Lanamme-University of Costa Rica laboratories and at the Construction
Engineering soil laboratory in the Costa Rica Institute of Technology.
2.2.1. Laboratory Testing previously done

Soils used for this research were obtained from sites located within the Central Valley
in Costa Rica, Central America, where most of the population resides [1].
Laboratory testing, already publish in previous papers, included Water Content,
Atterberg Limits, Grain Size Distribution, Standard Proctor, UU Triaxial Tests, and
Unconfined Compression Tests. Preparation of the soil specimens included sieving
through the # 4 sieve in order to create a particular type of soil that can be classified as
a fine grain size either soil plastic or non-plastic [1] [2].
Compacted specimens were prepared using the undercompaction method [4]. This
method allows achieving homogeneous specimens and reproducible densities.
Individual soil layers (lifts) were compacted to a lower density than the final desired
R. Baltodano Goulding and P. Leiva Padilla / Stress-Deformation Modeling 1139
value to normalize the cumulative compaction effect, as each subsequent layer is

compacted.
Table 1. Types of subgrade soils. [2][3]
Sr E
Soil
% MPa -
56 7.00 0.49
60 2.00 0.50
1, Sandy Silt (ML), e=0.70*
74 0.90 0.48
80 0.50 0.36
57 54.20 0.50
70 37.40 0.50
2, Silt (ML), e=0.89*
76 22.20 0.50
85 24.30 0.50
48 4.40 0.43
58 9.40 0.49
3, Silty Clay (CL), e=0.50*
68 8.20 0.45
81 6.40 0.45
*e: Void ratio
E: Youngs Modulus
: Poissons Ratio
Number of layers and undercompaction percentages were chosen according to the

requirements of each particular type of soil and void ratio. The soils were mixed with
different amounts of water to produce the target degree of saturation. Since the void
ratio and soil type were kept constant the only independent variable is the degree of
saturation. There are two dependent variables: suction stress and soil structure, but
since the differences in soil structure for the same soil type and void ratio are affected
by the interparticle forces captured in the suction stress term, we attribute all changes in
soil structure and shear strength observed to the suction stress variation with degree of
saturation. It is important to mention that the term suction stress is not equal to matric
suction in fine-grained soils. The term suction stress according to [5] is a macroscopic
concept capturing all microscopic effects that affect the effective stress in unsaturated
soil, i.e., Van der Waals forces, diffuse double layer forces, matric suction, changes in
physical and chemical forces as the degree of saturation varies, and those forces are
part of the definition of structure [6].
2.2.2. Computer Modeling

A computer model using the ABAQUS software was utilized to investigate the effects
that changes in water content in the subgrade have in the interaction behavior between
the subgrade and the geosynthetics. The ABAQUS model was defined by the following
number of steps:
Two axisymmetric parts definition were created for the model; the first part
consisted in the three upper layers: asphalt concrete layer, base course, and subbase
course, and the second in the subgrade layer. The geometry and mechanical properties
defined for the model can be seen in table 2.
The geosynthetic materials were defined as a skin and were attached to the bottom
of the base course, and membrane elements were assigned as a property for the
reinforcement. The elastic modulus used for the model definition was 4230 MPa, and
the Poisson ratio was 0.30. For each region the mechanical properties and sections were
defined and assigned according to Table 2.
Table 2. Structure model geometry and mechanical properties.
> z W d

DW
Asphalt concrete layer 400 0.40 150
Base course 50 0.35 270
Subbase course 15 0.30 300
Subgrade Variable Variable -
The structure parts assembly was inserted. The analysis steps sequence was
created: initial condition, and load application. Interaction properties between the
geosynthetic material and the subgrade was defined and assigned according to
conventional procedures for these types of materials. The pavement structure was
loaded with 0.83 MPa using a radius of 124 mm. A mesh for the model was created; it
was finest in the load area for the data collection. Finally, the analysis file was created,
and the model was analyzed.
For the analysis, first a model without geosynthetic was developed, and then a
model with a geosynthetic material over the subgrade layer was also created, in the
ideal condition of contact, when the friction is full bounded. Then a sensitivity analysis
was performed for the frictional properties in the geosynthetic and the subgrade.
3. Results
The stress and strain distribution in the subgrade upper fiber for the reinforced
pavement structure for soil 1 as a subgrade material is shown in Figures 1 and 2. The
stress and strain distributions for the sensibility analysis about the friction coefficient
influence are shown in Figures 3 and 4. For the soils 2 and 3, the same pavement
structure was analyzed, the stress and strain distribution for the three soils are shown in
Figures 5 thru 8.
4. Analysis
When the horizontal stress transmitted to the subgrade is analyzed as a function of

degree of saturation, it was observed that for the sandy silt variations of up to 34%
smaller occur from the 80% to the 56% degree of saturation. And in the case of the
strains associated to those values the decrease from 80% to 56% degree of saturation is
approximately of 1%. In the other hand, the vertical stress transmitted to the subgrade
under similar conditions exhibit around 80% increase from the 80% to the 56% degree
of saturation, and approximately 2% less strain for the dryer specimens. It was also
seen that at approximately 2.5 meters from the point of application of the load the
effect has been dissipated.

&,^

&s^^


& ,
'

& s
'
In terms of the friction developed between the geosynthetic materials and the
subgrade, it was observed that there is a significant effect, as expected, from the
behavior exhibit by a pavement structure unreinforced and a reinforced one. However,
it was also evident from the computer modeling that there is a smaller effect in the
behavior of reinforced pavement structures as a function of degree of saturation
compared to the unreinforced structures. There is about a 30% increase in the
horizontal stresses if the pavement is unreinforced, and the strains associated to this
condition vary from 4% for the unreinforced structure to 1.5% for the reinforced. And
there is also a little effect in terms of strain for all the reinforced structures regardless of
the degree of saturation, but between reinforced and unreinforced a 5% variation was
observed. It was also seen that at approximately 2.5 meters from the point of
application of the load the effect has been dissipated for the horizontal effects and at
1.5 meters for the vertical effects.

&, & s

&, & s

Analyzing the behavior of all three types of soils is possible to see the effect that
structure and plasticity can have in the behavior of the reinforced pavement structure. It
was determined that soils with greater plasticity and smaller grain size tend to have
larger horizontal and smaller vertical stresses. In terms of strain it was observed that the
materials with larger grain size and less plasticity have greater variations in horizontal
strains (up to 1.5% variation) than materials with smaller grain size and more plasticity
( up to 0.8% variation) as it gets away from the point of application of the load.
The vertical strain behavior varies from 12% to 1% as a function of not only
plasticity but also degree of saturation. This is consistent with the expected behavior
that coarser material have when subjected to traffic loads. However, it is important to
mention that in fine-grained subgrades the double-peak behavior in shear strength and
deformability characteristics can have a significant effect in soils with lower plasticity
indexes [1]. In addition, in some cases coarser materials with a particular degree of
saturation can have a similar behavior than finer materials at a different degree of
saturation, so to understand the water balance of each project and the expected water
content variations it is important to better predict the pavement structure behavior
during its lifetime.
5. Conclusions
It was observed that there is, as expected, significant effect in stresses and strains
transmitted to the subgrade when a pavement structure is reinforced using a
geosynthetic material.
It was determined from the computer model that there is significant effect in
stresses and strains transmitted to the subgrade as a function of degree of saturation. In
this model special attention was dedicated to degree of saturation from 80% and dryer.
The 80% degree of saturation is close to the optimum water content typically used for
pavement construction. Wetter conditions were also analyzed but are not concluded in
this paper. However, degrees of saturation higher than the optimum water content are
possibly starting to behave as saturated soils.
It was also observed that there is no significant effect in the frictional forced
between the subgrade and the geosynthetic. It is important to mention that these values
were estimated considering only undrained conditions, i.e. undrained shear strength
parameter. Some research is currently running laboratory testing in some universities
where the effect of matric suction in the frictional forces is being actually measured.
It is recommended that uncertainties associated to water content changes produced
by changes in the water balance proper for each project be analyzed using a
probabilistic approach. This is an acceptable method to quantify changes in shear
strength and deformation behavior of the reinforced pavement structure.
References
[1] Baltodano-Goulding, Rafael, Deformation Behavior in Unsaturated Soils, UNSAT 2010 Fifth
International Conference in Unsaturated Soils (2010)
[2] Baltodano-Goulding, Rafael, Suction Stress Influence on Earth Retaining Structures, GeoFlorida 2010
Conferences (2010)
[3] Korte-Leiva, Diana, Suction Stress Influence on Earth Retaining Structures Graduation Project, Costa
Rica Institute of Technology, Cartago, Costa Rica
[4] Ladd, R.S., Preparing Test Specimens using Undercompaction, Geotechnical Testing Journal GTJDJ,
(1978), 16-23
[5] Lu, N; Likos, W.J., Suction Stress Characteristic Curve for Unsaturated Soil, Journal of Geotechnical
and Geoenvironmental Engineering, (2006), 131-141
[6] Mitchell, J.K. Fundamentals of Soil Behavior, John Wiley and Sons, New York, 1993.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1145
Constitutive Model Input Parameters for

Numerical Analyses of Geotechnical Problems:
An In-Situ Testing Case Study
Crystal COXa,1 and Paul MAYNE b
a
GeoEnvironmental Resources, Inc.
b
Georgia Institute of Technology
Abstract. Numerical modeling and finite element analyses are increasingly

becoming cost effective methods for assessing geotechnical problems. The use of
in-situ testing methods, notably the SDMT, along with stiffness reduction
approaches utilizing G- curves are used to obtain stiffness related input
parameters for numerical modeling. This paper presents the steps used to obtain
stiffness parameters based on shear wave velocity measurements taken from
SCPTu and DMT data and presents a case study of a monitored embankment.
Keywords. constitutive model, modeling, numerical analysis, in-situ, CPT, DMT,

stiffness, stiffness degradation
1. Introduction
Numerical-method based analyses encountered in common geotechnical software

utilize various constitutive soil models and stiffness related inputs to characterize site
conditions and predict soil response. Customarily, rigorous laboratory testing is needed
to determine site specific stiffness related inputs for these models. Often, reference
tables are used to provide stiffness estimates based on soil type. In-situ testing can
provide a site specific and cost effective method to determine stiffness related
characteristics of geomaterials that may include sands, silts and clays.
Stress related deformation calculations involve linear elastic, nonlinear elastic and
elasto-plastic models (with and without strain hardening) and their respective input
parameters. The purpose of this paper is to address the determination of site-specific
soil characterization and stiffness parameters for use in numeric and/or finite element
analyses utilizing in-situ testing methods, in particular seismic flat dilatometer testing
(SDMT). A case study of an instrumented embankment utilizing linear elastic
constitutive relations in Newport News, Virginia is also presented.
2. Stiffness and Stiffness Degradation
Stiffness related inputs for numerical analyses primarily include variants of Youngs
(elastic) modulus, E along with poissons ratio, . Elastic modulus variants of secant
modulus in drained triaxial test at 50% strength (E50), tangent modulus for primary
oedometer loading (Eoed), as well as unloading/reloading modulus (Eur) are commonly
needed values.
1
1146 C. Cox and P. Mayne / Constitutive Model Input Parameters
Poissons ratio values determined through laboratory triaxial testing and/or

geophysical testing should generally be used as shown in Table 1.
Table 1. General values for Poissons Ratio based on soil type.

Soil Type Poissons Ratio ()
Granular Material 0.3
Drained Cohesive Material 0.3 0.4
Undrained Cohesive Material 0.5
Small strain stiffness and nonlinear soil behavior rely on shear modulus value
determination. The fundamental shear modulus, G0, is first determined for a soil profile
using seismic shear wave velocity testing via SDMT, seismic piezocone testing
(SCPTu), spectral analysis of surface waves (SASW) or other methods. Dilatometer
testing of the subsurface then provides the basis to produce a G- modulus reduction
curve for each representative soil type. Next, the G- modulus reduction curve is
translated to an E- modulus reduction curve using elastic theory. The site specific E-
modulus reduction curve along with data obtained from SDMT testing can then be used
to determine modulus inputs for use in numerical simulations.
2.1. Stiffness Degradation from Dilatometer Testing
Hyperbolic G- modulus reduction curves follow the typical behavior indicated by

Fig. 1 (Hardin & Drnevich, 1972). The Hardin-Drnevich degradation curve is defined
as:
=
#########################################################################################################################(=*
= /
/
where ref = reference strain
The general Hardin-Drnevich relation has been further modified to include scaling
factors in order to achieve a best fit hyperbolic model of modulus reduction for various
soil types based on laboratory testing. The scaling factors are seen in the inclusion of a
power exponent ( ) as shown in Eq. 2 (Vardenega & Bolton 2011), or alternatively, a
multiplicative factor (a) as shown in Eq. 3 (Santos & Correia, 2000).
The modified Hardin-Drnevich G- expression curves are the basis for modulus
reduction curves used in several numerical modeling suites. The use of a = 0.385 in Eq.
3 after Santos and Correia (2000) is common to many hardening soil models (Benz,
2007).
The use of Eq. 3 after Santos & Correia is most common to numerical analysis
suites in determining parameters for the hardening soil model. The degradation curves
as shown in Fig. 2 based on Eq. 2 and 3 reduce in distinctly different curves.
C. Cox and P. Mayne / Constitutive Model Input Parameters 1147
HardinDrnevichShearModulusDegradationCurve
1.2
0.8
0.6
G/Go
0.4
0.2
0
0.01 0.1 1 10 100

/
ref
Figure 1. Shear modulus reduction curve

(after Hardin and Drnevich 1972)
=
####################################################################################################################(*
/
=
/
=
###################################################################################################################(?*
= /
/
To construct the site specific G- modulus degradation curve, the working shear
strain DMT corresponding with GDMT must be determined (Cox & Mayne, 2015).
Once the G- modulus degradation curve is determined using in-situ testing, a
corresponding E- modulus degradation curve can be constructed using Hookes law
and elastic theory as shown in Figure 3.
Then, the secant modulus in triaxial testing at 50 percent strength E50 can also be
determined using values obtained from SDMT testing. Where according to Vermeer
(2001),
z^ ################################################################################################################################(*##
The unloading/reloading modulus in the drained/undrained triaxial test, Eur, cannot
readily be determined using data obtained from DMT testing and must be calculated
using accepted relationships if not using laboratory testing such as that given by
Vermeer (2001),
Figure 2. Reduction curves from fitted experimental data studies
zx z^ #################################################################################################################################(*##
One will note that when viewing the stiffness degradation curve, E50 is the smallest of
the modulus values discussed. Most numerical programs maintain an elastic stiffness
cutoff at Eur (corresponding to Gur), where hardening plasticity accounts for further
stiffness reductions.
Advanced hardening models include the values of Go and 0.7 as inputs to define
the nonlinearity and small strain stiffness relationships for various geomaterials. Once
Go is determined from seismic shear wave velocity testing, the stiffness degradation
curve as shown in Figure 2 can be used to define 0.7.
Figure 3. Elastic Modulus reduction curve using SDMT

3. Case Study Of A Monitored Embankment: Newport News, VA
A linear elastic constitutive model used in commercially available settlement software

was used to assess the applicability of utilizing stiffness parameters determined from
in-situ testing methods. The embankment studied was part of a protection system for
an electron beam accelerator constructed in Newport News, Virginia in 1986 with
initial analysis presented by Mayne & Frost. The embankment was constructed on site
soils consisting of interlayered sands, silts and clays of the Norfolk Formation to an
approximate depth of 7 m underlain by preconsolidated low plasticity clays and silts
along with silty sands of the Yorktown Formation.
Figure 4. Cross Section of Test Embankment in Newport News, Virginia,

(after Mayne & Frost)
Instrumentation used to monitor deformations of a test embankment included

horizontal inclinometer pipe, settlement plates, borros points, pneumatic piezometers
and open standpipes. A cross section of the embankment is shown in Fig. 4. Settlement
estimates were calculated using various methods that included estimates based on
oedometer testing (OED 86), estimates based on DMT testing (DMT 86) and a FEM
model (FEM 86) based on work by Duncan (Mayne & Frost, 1988).
In the current case study, inputs to the settlement software program were
based on DMT and SCPTu data available at the site. A multi-layer theory stress
anlaysis was conducted based on Hankel transforms utilizing numerical methods
presented by Yue (Rocscience, 2009). The requested input for elastic settlement
calculations for each soil layer was Es, defined as the one-dimensional Youngs
modulus. The input for consolidation settlement for each soil layer using a linear
elastic soil model was mv, the coefficient of volume compressibility. The operative
constrained modulus from DMT testing MDMT was used as the in-situ input for Es
following Eq. 4. The reciprocal of the operative constrained modulus MDMT (1/ MDMT)
was used as the in-situ input for mv. Table 2 includes the layer definitions and inputs
used in the settlement analysis.
The results of the in-situ generated input parameter analysis are shown in Fig.
5 and a comparison of the estimated settlements from multiple calculation methods
shown in Fig. 6.
Figure 5. Settlement Results from Numerical Analysis using Input

Parameters from In-Situ Testing
DistanceAlongEmbankmentSection(m)
0 5 10 15 20 25 30
0
1 9/30/1986
10/10/1986
10/29/1986
2 DMT86
Settlement(cm)
DMT15
FEM86
3
OED86
Figure 6. Comparison of Estimated Settlements for a Monitored

Embankment, Newport News, VA
Table 2. Layer Properties Used in Numerical Analysis from In-Situ Testing
Layer Depth (m) Soil Type Go MDMT Es

(Mpa) (Mpa) (Mpa)
1 0 - 1.1 Sandy Silt 86 100 100
2 1.1 - 1.7 Silty Sand 89 125 125
3 1.7 - 2.0 Silty Clay 71 7 7
4 2.0 - 2.3 Silty Sand 80 28 28
5 2.3 - 3.0 Silty Clay 71 3 3
6 3.0 - 3.3 Sandy Silt 75 12 12
7 3.3 - 4.5 Silty Sand 89 125 125
8 4.5 - 4.8 Silt 80 35 35
9 4.8 - 5.5 Clay 80 12 12
10 5.5 - 7.5 Silty Sand 222 105 105
11 7.5 - 10 Silty Sand 216 135 135
12 10 - 12 Silty Sand 264 90 90
13 12 - 15 Silty Sand 331 80 80
14 15 - 21 Silty Sand 282 70 70
15 21 - 25 Silty Sand 230 50 50
Results utilizing numerical methods with stiffness inputs derived from in-situ
testing (DMT 15) were a very close match to calculations performed prior using DMT
data (DMT 86). The estimated deformations were in agreement with measured
deformations recorded from instrumentation. A FEM model of the embankment
utilizing the in-situ derived parameters for non-linear elasto-plastic soil conditions is
under development.
4. Conclusion
In-situ testing can provide a site specific and cost effective method to determine
stiffness related characteristics of geomaterials that may include sands, silts and clays.
Seismic testing to determine fundamental shear modulus G0 values along with DMT
testing can be used to produce G- modulus reduction curves that are then translated to
an E- modulus reduction curve using elastic theory.
A case study of an instrumented embankment utilizing linear elastic
constitutive relations in Newport News, Virginia was examined. Linear elastic input
parameters of Es and mv were assigned to the settlement model. The model predicts the
observed settlement accurately and provides a simplistic and cost effective method to
determine site specific stiffness parameters for numerical analyses in all soil types.
References
[1] Amoroso, S., Monaco P. & Marchetti, D. (2012). Use of the Seismic Dilatometer (SDMT) to estimate
in-situ G- decay curves in various soil types. Geotechnical and Geophysical Site Characterization,
Vol. 1 (Proc. ISC-4, Pernambuco), Taylor & Francis Group, London: 447-452.
[2] Benz, T. (2007).Small-strain stiffness of soils and its numerical consequences.PhD Thesis, Universitat
Stuttgart.
[3] Bolton, M. (1986).The strength and dilatancy of sands.Gotechnique36(1): 65-78.
[4] Campanella, R.G. & Robertson, P.K. (1991) Use and interpretation of a research dilatometer,
Canadian Geotechnical Journal, 28 (1): pp.113-126.
[5] Cox, C.L. & Mayne, P.W. (2015) Soil stiffness constitutive model parameters for geotechnical
problems: A dilatometer testing approach Proceedings DMT 15, Rome, Italy, June 2015.
[6] Lee, J., Salgado, R. & Carraro, J.A. (2004). Stiffness degradation and shear strength of silty
sandsCanadian Geotechnical Journal, 41(5): pp. 831-843.
[7] Marchetti, S. (1980). In situ test by flat dilatometer Journal of Geotechnical Engineering Division
ASCE, 106 (GT3): pp. 299-321.
[8] Marchetti, S. (1997). The flat dilatometer: design applications Third Geotechnical Engineering
Conference, Cairo University: 421-448.
[9] Mayne, P.W. & Frost, D.D. (1988) Dilatometer experience in Washington D.C., Transporation
Research Record 1169, National Academy Press, Washington, D.C., 16-23.
[10] Mayne, P.W., Schnider, J.A., & Martin, G.K. (1999) Small-and large strain soil properties from
seismic flat plate dilatometer tests, Prefailure Deformation Characteristics of Geomaterials, Vol. 1
(Proc. Torino),Balkema, Rotterdam: 419-426.
[11] Orbzud, R. & Truty, A. (2010). The hardening soil model A practical Guidebook Technical Report
Z_Soil.PC 100701, Lausanne, August 2010.
[12] Monaco, P., Totani, G. & Calabrese, M. (2006), DMT predicted vs. observed settlements: a review of
the available experience, Proceedings from the Second International Flat Dilatomer Conference.
Publisher. 244-252.
[13] Plaxis.Plaxis-GiD Material Models Manual Version 1 Plaxis., pp. 102-1
[14] Rocscience. (2009) Settle3d Theory Manual RocScience.
[15] Santos, J.A., & Gomes Correia, A., (2000). Shear modulus of soils under cyclic loading at small to
medium strain levelProc.12th World Conference on Earthquake Engineering, paper ID 0530,
Auckland, New Zealand.
[16] Santos, J.A., Gomes Correia, A., Modaressi, A., Lopez-Caballero, F., & Gomes, R. (2003). Validation
of an elasto-plastic model to predict secant shear modulus of natural soils by experimental results
Deformation Characteristics of Geomaterials, Di Benedetto et al., Eds., Swets&Zeitlinger, Lisse.
[17] Totani, G., Marchetti, S., Monaco, P. & Calabrese, M. (2001).Use of the Flat Dilatometer Test (DMT)
in geotechnical design Proceedings, In Situ 2001,International Conference on In Situ Measurement of
Soil Properties.Bali, Indonesia: 487-494.
[18] Vardenega, P.J. and Bolton, M. (2011). Practical methods to estimate the non-linear shear stiffness of
fine grained soils International Symposium on Deformation Characteristics of Geomaterials, Vol. 1,
Hanrimwon Company, Seoul, Korea: 372-379.
[19] Vermeer, P.A. (2001). On single anchored retaining walls PlaxisBulletin 10.
[20] Vucetic, M. and Dobry, R. (1991). Effect of soil plasticity on cyclic response Journal of
Geotechnical Engineering (ASCE), Vol. 117 (1): 89-107.
[21] Yue, Z.Q. (1995) On generalized Kelvin solution in multilayered elastic medium Journal of
Elasticity, 40(1), 143
[22] Yue, Z.Q. (1996) On elastostatics of multilayered solids subjected to general surface traction
Quarterly Journal of Mechanics and Applied Mathematics 49(Part 3), 471-499.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1153
Geotechnical characterization of calcareous

sands in the northeastern coast of
Venezuela
Nelly VIEIRA Faria a 1 and Diana DE PONTE Fernandes a
a
Geohidra Consultores, Caracas, Venezuela
Abstract. The term calcareous may be applied to sediments, sedimentary rocks or

soils which are formed or contain high proportion of calcium carbonate (CaCO3).
The calcareous sediments are usually deposited in shallow water and close to the
coast because the precipitated calcium carbonate comes mainly from organisms
that required terrigenous nutrients. These sediments exist in different forms, in
each case having distinctive properties between them, depending on the
sedimentary facies that controls their deposition process. In general there is limited
literature related to carbonate materials in the Caribbean. Most of recent
information comes from Australia where they have been extensively studied. This
article shows the characterization of carbonate soils in the northeastern region of
Venezuela. For this study, data from two campaigns conducted offshore Venezuela
are studied. Water depth in the area is approximately 100 m. Samples were taken
with 3" Shelby (pushed) and 2 Shelby (driven). Additionally, CPT tests were
performed. Basic classification tests including carbonate content and advanced
geomechanical tests such as direct simple shear (ASTM D6528, [1]) and
controlled-strain loading consolidation tests (ASTM D4186, [1]) are performed.
All data obtained in field and laboratory is processed and analyzed in order to
estimate their characteristics, behavior and in-situ state. This paper presents the
characterization of the calcareous sands incorporating geological and geotechnical
data with the aim at a future comparison with analogous materials in other parts of
the world. Main findings can be summarized as: a) the methodology for
reconstituting this sort of soil for maximum and minimum density should be
improved, and b) samples did not suffer fragmentation and did not (or barely)
experienced compressibility under the applied stresses.
Keywords. Calcareous, sand, geotechnical characterization
1. Introduction
Venezuelan offshore geotechnical studies in areas with presence of carbonate materials

are very incipient, since the expansion of the oil and gas industry on these areas is
recent. For this reason, information about the characteristics and geomechanical
properties of these materials is very limited. Due to the magnitude of the structures and
investments required for the execution of oil projects, it is essential to obtain a
complete characterization of the subsoil where these structures will be emplaced. As is
well known, these materials have erratic and difficult to predict behavior, and from this
1
1154 N. Vieira Faria and D. De Ponte Fernandes / Geotechnical Characterization
it becomes evident the necessity of providing a detailed characterization of this

materials.
The purpose of this paper is to review and analyze the data from laboratory tests
(e.g. DSS and CRS), conducted on reconstituted samples and to relate it to in situ CPT
data in an attempt to relate both and to establish the main limitations of the former.
2. Regional Geology
The study area is located at the North of the Paria Peninsula, Sucre State, Venezuela,
near the border with Trinidad.
The area is mostly characterized by the effects of the rupture zone of the South
American plate in northern Trinidad, where the Atlantic Plate subducts under the
Caribbean Plate.
In the current projects undergoing in the North of Paria, the presence of corals are
observed beneath a clay layer under the seafloor. This fact is confirmed by the work of
Delgado [3], where a 10-25 meters thick layer of carbonated material is described
below a clay seabed.
The coral itself is the shell of a living organism, it needs abundant sunlight and
nutrients; corals lived in the area during the Pliocene/Pleistocene in an environment of
warm shallow water.
During the Pleistocene, several transgressive-regressive events happened, as a
product of thaws-glaciations, but the most significant decrease was generated 20,000-
25,000 years ago (Middle-Upper Pleistocene), when the sea level reached 120 meters
below the current, generating favorable conditions for the development of carbonate
platforms in several areas of the Caribbean, among which is the area located at the
North of the Paria Peninsula, with the presence of reef complexes.
Periodical transgressions have been reported for the Late Pleistocene/Holocene and
some studies have indicated that the sea level changes during the Quaternary played a
significant role in the generation of marine terrace deposits in the northeastern
Venezuela. The main consequence of the increase in the sea level was local extinction
or population displacement, such that coastal species would be found around the new
coastline.
The breakdown of calcareous substrates among coral reefs, or bioerosion of the
mass occurred and generated the calcareous sand reported in the study.
Above the carbonate platform sequences recent clays, from the Orinoco Delta have
been deposited.
3. Properties of calcareous soils
The term calcareous may be applied to sediments, sedimentary rocks or soils which
are formed or contain high portion of calcium carbonate (CaCO3). According to
reference [5] this concept is applicable when the identification of carbonated mineral
refers to calcite, dolomite, aragonite, siderite and others.
The calcareous sediments are tipically deposited in shallow waters and nearshore,
since the precipitated calcium carbonate comes mostly from organisms that required
terrigenous nutrients. The fundamental condition controlling accumulation of
N. Vieira Faria and D. De Ponte Fernandes / Geotechnical Characterization 1155
calcareous sediments is that the sites of deposition are free from entrance of
considerable quantities of muddy and sandy sediments.
On the Venezuelan continental shelf, calcareous sediments are formed by the
accumulation of the skeletons of no more than four fossil groups, mostly corals and
calcareous algae, molluscs and / or bryozoans.
3.1. Calcareous sands
The main components of the calcareous sands are calcite, dolomite and aragonite;
therefore, these sands have high calcium carbonate content. The calcareous sands are
unique in terms of features, including their mineralogy, roundness, sphericity and
internal porosity, among others, and can be presented in cemented or uncemented state.
Because of their mineralogy, these sands are more susceptible to cracking and breaking
of their grains under the application of stress. This can be seen if sieve analysis is
performed before and after a compression test, if an increase of their finest grains can
be observed, this is an indication of the breaking of original particles.
Additionally because of the angularity and shape of the grains and the
susceptibility of the precipitation of salts at the grain contacts these sands tend to form
metastable structure of high void ratio that are highly compressible in a brittle behavior.
These differences make clear that its geotechnical behavior, including its strength,
differs considerably from the siliciclastic sands.
4. Calcareous sands on the northeastern continental shelf of Venezuela
The studied calcareous sands were recovered at the North of the Paria Peninsula, Sucre
State, Venezuela. The thickness of calcareous material in the area varies from 7 to 30 m.
The recovered sediments were classified as sands with high content of calcium
carbonate (over 60%), originated from the fragmentation of corals and bivalves shells
or gastropods.
4.1. Characterization of tested sands
The calcareous bioclastic granular material varies between sand and gravel. The
greatest thicknesses of this material are found in the south (16 and 19 meters); while in
the north the thickness is between 6 and 10.5 meters. The visual description was made
following two procedures, depending on the size of the grains: one directly in grains
with size greater than 0.2 cm and another for smaller grains using a polarizing
microscope CX31-P with camera included.
Broadly, the sphericity of the studied samples is medium-high, ranging between
0.7 and 0.9 [4]; whereas the roundness of the grains varies from subangular to
subrounded and angular to subangular. In grains observed under a microscope, growth
of crystals of aragonite around the soil grains were identified, and in few cases, this
growth kept the grains together, showing contacts between particles, indicating a very
weak level of cementing by grains overgrowth (Figure 1).
< 0.2 cm
Figure 1. Morphology of the grains.
5. Laboratory tests
For the characterization of the granular materials found on the northeast coast of
Venezuela, samples from two different offshore geotechnical campaigns for the same
area of study were used. The water depth is about 100 m. These samples were
recovered by pushed 3" Shelby samplers and by percussion using Shelbys 2 " thick
walled. It is clear that these sampling methods, particularly the thick walled Shelbys do
not allow the recovery of undisturbed samples. Recovery was also poor (<40%). The
samples were preserved and transported according to ASTM D4220, [1]. The
carbonated soils classification was performed according to reference [5]. Additionally,
Cone Penetration Test (ASTM D5778, [1]) tests were performed.
Based on the above, sand samples with carbonate content greater than 60%, were
analyzed. Next, the results of 22 samples and 5 CPTs on granular materials between 7
and 30 m depth below seabed are analyzed.
5.1. Basic Tests
The following basic classification tests were performed to the recovered samples:
Visual classification (ASTM D2488, [1]), water content (ASTM D2216, [1]), Particle-
size analysis (ASTM D422, [1]), Plastic and Liquid Limits (ASTM D2488, [1]) and
Carbonate Content (ASTM D4373, [1]). The results of these tests are presented in
Figure 2.
According to reference [5] and the results of basic classification tests, samples can
be described as olive-gray siliceous carbonated clayey sands (low plasticity clay) with a
high proportion of fragments of corals, or as olive-gray carbonated clayey sands (low
plasticity clay).
Figure 2. Basic tests results
5.2. Advanced tests
Advanced tests for the geomechanic characterization include procedures and

techniques for assessing the response of a material under a certain stress. Due to the
type of materials and the sampling methods, keeping the original structure of the soil
(density/cementation and, in general, the fabric) during sampling is not feasible. Only
reconstituted samples were used, and the properties obtained define the best bounds on
the in-situ conditions. A comparison of the CPT tests with those properties could lead
to a better insight of the in situ expected behavior. Hence the use of these results to
predict field behavior involves using engineering criteria and is accompanied by
unavoidable uncertainties.
For this article the general framework defined by [6] to try to establish bounds on
the maximum and minimum peak friction angles on the uncemented material was
followed.
Direct Simple Shear test (DSS, ASTM D6528, [1]) under constant were performed
in samples reconstituted in two states: one state with the highest void ratio (loosest
material) and other state with the lowest void ratio (densest material) that was possible
to obtain in the laboratory. To the loosest sample a stress reasonably higher than the
maximum expected in field conditions was applied, which was defined as 50% higher
than the estimated stress applied by the temporal foundation mat of a jacket platform.
This should provide a lower bound to the peak friction angle, and in most cases this test
should progress in a stable path to the critical state. To the densest sample a stress
lower than the minimum expected was applied (1/4 of the maximum stress), this time
to provide an upper bound on the dilation contribution (if any) to the peak friction
angle. To obtain the loosest condition of the samples, the method consisted in dropping
freely the dried and thoroughly mixed material within the rings of the equipment, this
in order to prevent segregation. Meanwhile, the densest condition was obtained placing
the samples into the rings in several layers, providing small bumps and vibration and
trying to increase the density without breaking the grains. Figure 3 shows the peak
friction angles obtained in each case. It also shows that in both cases, angles were
greater than 40.
Figure 3. Results of DSS Tests

Friction angles found in this study are high compared to the results that are usually
reported on these materials; however according to [8] friction angles of 60 have also
been reported in carbonated sands in Australia.
CRS consolidation tests (ASTM D4186) were also performed in order to verify the
degree of compressibility of these materials. Samples were reconstituted with the
maximum void ratio obtained in laboratory. In these tests, the maximum stress applied
was again 320 kPa. The results of the compressibility coefficient obtained in the four
samples were less than 0.2. This value indicates that under the applied stress,
reconstituted samples are little or non-compressible. Particle-Size analysis was
performed before and after tests execution in order to verify the fragmentation that
shell and coral fragments may suffer, after stress loading. Comparison of those showed
that under these stress conditions, the material does not suffer fragmentation.
6. CPT Results
From the geotechnical campaign of 2011, data from 5 CPTs performed in boreholes
located less than 10 m away from the sampling holes are available. The tip resistance
values, qc of one borehole is shown in Figure 4 which shows the typical erratic behavior
of carbonated sands. According to [8], the classification charts of CPT for non-
calcareous soils can be used for calcareous soils as well. However, for a rational
interpretation, laboratory tests are required. Based on [9], the degree of cementation of
the carbonated sands can be deduced from the tip values qc, of the CPT test. In this
paper, this classification for carbonated soils is also used for siliceous carbonated soils,
as shown in Figure 4.
Due to the high variability of CPT values, with very high peaks of values of qc
followed for very low values of it, these results cannot be directly used for engineering
calculations. Instead, it is used smoothed records from medians of qc values calculated
in spaces from 10 to 30 cm, depending on the variability of the log. Figure 4 shows an
example of a CPT record from de study, where it can be observed the erratic results of
qc and the smoothed record with the median every 12 cm, that is used for analysis and
calculations.
Figure 4. Values of tip resistance qc from CPT.
7. Conclusions and Recommendations
The recovered sediments from 10 to 28 m depth were classified as sands rich in

calcium carbonate (over 60%), which originated from fragmentation of paleocorals and
bivalves shells or gastropods. From the records of the CPT, there is evidence that some
of these carbonated sand strata are well cemented, as its value of qc>4 MPa. Even
though the CPT shows degree of cementation variable and localized in this material,
remain the uncertainty whether this cementation is reliable under stress loads applied
by the structures
The results of CRSs indicated that the samples are little or non-compressible under
the applied stress, and that the material do not suffer fragmentation. In this sense,
additional tests applying greater loads must be executed, in order to reach a complete
characterization of this property.
The results of advanced tests show that the reconstitution of material with
minimum void ratio, with the methodology used at the lab, could not reach the
maximum density that these materials could achieve. Thereby, the method of
reconstitution of samples for maximum density in these materials should be improved
and also, it is recommended to test methods of artificial cementing of carbonated
samples in the lab according to [10].
We recommend further testing in other areas of the region to expand the

information presented in this paper. Alike, it is recommended to perform cyclic tests to
characterize the behavior of materials under dynamic loads. Table 1 shows the relevant
characteristics of the studied materials.
Table 1. Relevant characteristics of the studied materials
Feature Characteristics
Mineralogy CaCO3 main contribution of aragonite
Grain shape Skeletal grains, subangular and subrounded in fewer quantity
Gs 2,75
emx 1,05 - 0,78
emin 0,76 - 0,64
43 - 57
Cc < 0,2
Cementation Varying in depth
Degree of cementation Very Weakley cemented to well cemented
Due to the erratic behavior and natural state of the carbonate soils, the result of
the evaluation of these materials on disturbed samples is referential. For this reason, it
is important to develop engineering methods based directly on CPT records, to predict
the behavior of offshore structures. These records provide closer values of the natural
state of carbonate soils and their behavior.
8. Acknowledgements
The authors express their sincere appreciation to Eng. J. Sgambatti for reviewing this
paper.
References
[1] ASTM Standard test method for testing materials, Annual book of ASTM Standard. Vol. 04.08 and
04.09. (2012).
[2] FUNVISIS, (2007). Estudio de Riesgos Geomorfolgicos y Sismolgicos del Proyecto Mariscal Sucre.
Technical report.
[3] Delgado, M., (1982). Estudio Ssmico de alta resolucin de un rea de inters petrolero al norte de la
Pennsula de Paria, Estado Sucre. Graduate thesis, Universidad Central de Venezuela. Unpublished.
[4] W. C. Krumbein and L. L. Sloss, Stratigraphy and sedimentation, San Francisco, W. H. Freeman, 1963.
[5] Clark, A.R. & Walker, B.F., A Proposed Scheme for the Classification and Nomenclature for Use in the
Engineering. Description of Middle Eastern Sedimentary Rocks; Geotechnique Vol. 27, No. 1. (1977),
93-99.
[6] Bolton, M.D. The Strength and dilatancy of sands, Geotechnique, Vol. 36, (1986), 65-78.
[7] N. Yamamoto, M.F. Randolph & I. Einav, Numerical study of shallow foundations on calcareous sand,
Proceedings of Frontiers in Offshore Geotechnics: ISFOG, Pages 443-449. (2005).
[8] Lunne, T., Robertson, P. and Powell, J. Cone Penetration Testing in Geotechnical Practice, Spon Press,
London, 2002.
[9] King, R. W., van Hooydonk, W. R., Kolk, H. F. and Windle, D. Geotechnical Investigations of
Calcareous Soils on The North West Shelf, Australia, Proceedings of the Offshore Technology
Conference, Paper N 3772. (1980).
[10] Coop and Atkinson, The mechanics of cemented carbonate sands, Geotechnique Volume 43, Issue 1,
01 March 1993, (1993), 53 67.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1161
Laboratory Determination of Parameters

for Transversely Anisotropic Model of Stiff
Clay
Monika ERNKOVa,1, Martin KRUPIKAa, Tom MOHYLAa, Jan BOHa,
Josef ROTTa and David MANa
a
Charles University in Prague, Faculty of Science, Albertov 6, 12843Prague, Czech
Republic
Abstract. The overconsolidated Brno clay (Tegel) was subjected to the extensive
laboratory testing. The obtained data were used in the anisotropic hypoplastic
model and a backanalysis of coefficient of earth pressure at rest K0 was
performed. The resulting value of K0 = 0.75 was not estimated by more
conventional methods.
Keywords. Anisotropy, backanalysis, in-situ stress, stiff clay, tunnelling
1. Introduction
The recently opened urban inner ring road of the town of Brno (Czech Republic)
includes the Krlovo Pole Tunnels, consisting of two mostly parallel tunnel tubes with
a separation distance of about 70 m and the length of approximately 1250 m. The
height and width of the cross section are about 11.5 m and 14 m, respectively, and the
overburden thickness varies from 6 m to 21 m. The tunnels have been built in the
developed urban environment and mostly, except for the portal portions, in the strata of
high plasticity Neogene (Miocene) clay, called Tegel (Figure 1).
The tunnels have been driven by the New Austrian Tunelling Method. The face of
one of the tunnels included the geometry of two exploratory galleries of a triangular
cross-section with arched sides, constructed as a part of the geotechnical site
investigation. The width and height of the galleries reached 4.75 m and 4.5 m,
respectively. The primary lining of the galleries consisted of 0.10 m of sprayed
concrete in combination with steel profiles.
To estimate the horizontal stresses in the clay massif four unsupported circular side
drifts of the diameter of 1.9 m were excavated from the exploratory galleries. The
convergences of these cylindrical cavities were measured during the site investigation
[1] and a coefficient of earth pressure at rest K0 in overconsolidated Tegel was back-
calculated.
In backanalysing the final tunnel, galleries, and the side drifts the recently
developed hypoplastic model for clay was used, which included stiffness anisotropy
[2].The basic set of the parameters for the hypoplastic model for clays, as well as the
additional parameters for the enhanced model, were determined using the laboratory
experiments.
1162 M. Cernkov et al. / Laboratory Determination of Parameters
Figure 1. Longitudinal geological cross-section along the tunnels [1].
2. Brno Tegel
The Miocene Brno Tegel was deposited in the Carpathian Fore-deep, north of the
Vienna Basin. The area was strongly influenced by tectonic phenomena, which
controlled the depths of the Neogene sea, and therefore the height of the clayey
sediments varied.
In Pliocene and Quartenary periods erosion of the marine sediments took place.
However the depth of the erosion has not been determined to date. From the
geotechnical viewpoint the clay exhibits moderate apparent overconsolidation, with
yield stress ratio (apparent overconsolidation ratio OCR) reaching cca 10. However it
has not been explained to date, whether the apparent OCR was caused by mechanical
unloading (erosion) or by ageing.
The Tegel is a calcareous clay, of grey-blue to grey-green colour, the upper parts,
down to 15 to 20 metres, brownish due to weathering. Often there are crystals of calcite
and/or limonite. Crystals of gypsum are frequent especially in the upper brown layers,
their occurrence diminishing in the grey layers. Locally the clay can be sandy.
The Tegel sedimented in the marine environment and during its subsequent
exposure it was subject to intense chemical changes, especially due to freshwater
leaching. This might explain the intense fissuring, reported typically in site
investigations: The percolation of fresh water is concentrated into pre-existing tectonic
cracks. The resulting chemical changes, formation of gypsum, calcite (10-34%) and
limonite, were most intense in the zones. Assuming that the chemical changes were
accompanied by volume changes, the fissuring may have originated in a similar way
like freezing and shrinkage [3].
Table 1 summarizes the mineralogical composition of the three clay samples. The
analyses were carried out by X-ray diffraction. The typical grading curve (sieving and
hydrometer analysis) is shown in Figure 2, and index properties summarized in
Table 2.
M. Cernkov et al. / Laboratory Determination of Parameters 1163
Table 1. X-ray diffraction analysis from Brno-Slatina.

Depth Quartz Calcite Smectite Kaolinite Muscovite
(m) (%) (%) (%) (%) (%)
24 27 29 15 3 8
27 33 34 0 10 7
36 23 31 7 7 26
Figure 2. Grading curves of Brno Tegel.
Table 2. Index properties of Brno Tegel.

Liquid Limit Plastic Limit Plasticity Index Colloidal
Depth (m)
(%) (%) (%) Activity (%)
24 56 31 25 0,515
27 40 30 10 0,289
36 61 35 26 0,577
3. Laboratory Determination of Parameters of Brno Tegel
Undisturbed samples were taken by an open thin-walled steel sampler of the inner
diameter of 114 mm, manufactured to the DIN specification [4]. Undisturbed and
reconstituted specimens were tested in the oedometer, triaxial, and shear box
apparatuses.
3.1. Strength
The critical state strength was determined using

x Conventional triaxial reconstituted specimens;
x Frictionless triaxial reconstituted specimens;
x Translational and ring shear box.
The stress paths and the failure line produced by the undrained triaxial tests with
pore pressure measurements (CIUP) of reconstituted specimens with frictionless
platens are shown in Figure 3. CIUP triaxial tests on undisturbed specimens of 38 mm
diameter were used in calibrating the 'large-strain' shear stiffness. The stress paths and
the calibration are in Figure 4.
Figure 3. Stress paths of CIUP of reconstituted specimens with frictionless platens [5].
Figure 4. Triaxial CIUP Stress paths and calibration of undisturbed specimens using the 'basic' hypoplastic
model (p is the effective mean stress; [6]).
Further to triaxial testing the critical state friction angle c was also estimated by
testing reconstituted specimens in direct shear. Both the translational and the ring shear
devices were used (see Table 3). There are many rather conflicting theoretical and
experimental findings about the possibility of reaching the critical state of soils in the
ring shear, or in simple shear (e.g. [7, 8, 9], and many others). It seems that ring shear
tests underestimate the critical state friction angle. Nevertheless it was used in the
current project as a quick estimate of critical state friction angle c.
Table 3. The critical state friction angle c [] of Brno Tegel.

Test Series Triaxial Triaxial Ring Shear
Conventional Platens Frictionless Platens (Bromhead Type)
Feda et al., 1995 [3] 27.5 - -
Svoboda et al., 2010 [6] 18.9 - 19.9
Fencl, 2012 [5] 21.0 23.6 19.8
3.2. 1-D Compressibility
The compressibility of Tegel was studied using oedometer testing of both undisturbed
and reconstituted specimens up to the vertical effective stresses of 13 MPa. The
intrinsic and sedimentation compression lines, and the compression curves of
undisturbed specimens normalized using the void index [10] can be seen in Figure 5.
Figure 5. Normalized compression curves of Brno Tegel.
The quasi-preconsolidation pressures were determined from the compressibility

curves by Casagrande's method. The obtained values at individual depths of sampling
are shown in Figure 6.
Figure 6. The profile of quasi-preconsolidation pressures vmax.

3.3. Small and Very Small Strains Stiffness Anisotropy of Tegel
The model for cross-anisotropic stiffness requires five further model parameters G tp0,
G, xGv, xGE, pp0 (the index t denotes the transverse direction to the plane of isotropy
the vertical direction, while p represents the in-plane direction the horizontal
direction; [11]). In the model the very small strain shear modulus Gtp0 requires two
parameters Ag and ng. Their values have been determined using bender element results
[6].
The main anisotropy parameter G = Gpp0/Gtp0 was estimated from the
measurements of vertically transmitted shear wave velocities through two specimens.
The first one trimmed vertically, the second one horizontally. The experiments were
carried out under all-round effective cell pressure corresponding to the estimated in-situ
mean effective stress. As shown in [11], stiff to hard clays exhibit just mild stress-
induced anisotropy, and therefore the isotropic stress state during the bender elements
testing was not expected to influence the results significantly. Average of times of
arrival at frequencies 3,5,7,9 kHz was used to evaluate the shear wave velocity. These
frequencies showed no effect of the overshooting phenomenon. The measured Gpp0 was
consistently higher than Gtp0 (Figure 7a). To quantify G the data were fitted by a
straight line (Figure 7a). The resulting coefficient G 1.45 is shown in Figure 7b.
Figure 7. Shear moduli obtained by bender element measurements (a) and the resulting coefficient G (b).
The remaining anisotropy parameters were quantified using small isotropic probes,
and constant radial stress shearing probes, applied to the triaxial undisturbed
specimens. Again the specimens were first isotropically consolidated to estimated in-
situ stresses. Submersible LVDTs were mounted on the specimens in two pairs: two
measuring the axial, and two measuring the radial strains during the probes ( a and r,
respectively). The strains obtained from the radial LVDTs were supplemented by the r
computed from the conventional measurements of volume changes using usual 'flow
pumps' (volume-pressure controllers). The radial strains during the r = const shearing
probes were negligible, as indicated by both local mesurements and by radial strains
computed from the volume changes (Figure 8a). In Figure 8b there are results of the
isotropic probes, which were approximated by a linear fit r = 0.6 a.
Figure 8. Axial (a) and radial (r) strains obtained by shearing (a) and isotropic (b) stress probing of
undisturbed triaxial specimens.
From the compliance matrix for transversely isotropic soil, for constant radial
stress probes it follows that tp0 is directly proportional to the radial strain rate [12].
Negligible radial strains of Figure 8(a) therefore imply tp0 0. For the isotropic probes
it then follows for Eh0/Ev0 E = a/r, ( means strain rate). Therefore for the
experimentally found r = 0.6 a (Figure 8b) E 1.67. G 1.45 from the bender
elements measurements (Figure 7b) then yields xGE 0.73, which is close to xGE 0.8
suggested by Man and Rott [11] on the basis of the experimental database from the
literature.
4. The Full Set of Parameters and the Use in Numerical Model
The parameters for transversely anisotropic hypoplastic model, which were determined
and calibrated on laboratory tests, are listed in Table 4.
Table 4. Parameters for transversely anisotropic hypoplastic model.
Parameter Parameter
18.8 kNm-3 0.8
c ~22 0.33
* 0.128 Ag 5300 kPa
* 0.015 ng 0.50
N 1.5 tp0 ~0.00
R 0.0001 G 1.45
mrat 0.5 xGE 0.73
r 0.2 xG (1.0)
Figure 9. Value of K0 (a) and horizontal displacement (b) in the vicinity of exploratory adit [12].
To obtain the relation between the shear moduli ratio G = Gpp0/Gtp0 and K0, three pairs
of G-K0 were backcalculated in such a manner that horizontal and vertical convergence
ratio uh / uv = 1.25, measured in the monitoring, was obtained. The values of G = 1,00;
1,35 and 1,70 were based on literature review. K0 = 0.75 can be deduced with
laboratory determined value of G = 1.45 (Figure 9a). The simulations of horizontal
displacements agree well with the monitoring data (Figure 9b).
5. Conclusions
The soil parameters for transversely anisotropic model have been calibrated using the
oedometer and triaxial tests on undisturbed and reconstituted specimens. The numerical
simulations based on the presented laboratory testing successfully back-calculated the
in-situ monitoring data, and also enabled us to estimate with reasonable confidence the
at-rest coefficient K0, which for the pseudo-overconsolidated clay in question proved
impossible to be determined by other direct or indirect methods.
References
[1] J. Pavlk, L. Klmek and D. Rupp, Geotechnical exploration for the Dobrovskho tunnel, the most
significant structure on the large city ring road in Brno. Tunel 13 (2004), No 2, 2-12.
[2] D. Man, Clay hypoplasticity model including stiffness anisotropy. Gotechnique 64 (2014), No. 3, 232-
238.
[3] J. Feda, J. Boh, and I. Herle, Shear resistance of fissured Neogene clays. Engineering Geology 39,
(1995), 171-184.
[4] DIN 4021 Aufschluss durch Schrfe und Bohrungen sowie Entnahme von Proben (Exploration by
excavation and borings, sampling in German), DIN, Oktober 1990.
[5] M. Fencl, Critical state strength of Brno Clay (in Czech), MSc Thesis, Charles University in Prague,
(2012), 64 pp.
[6] T. Svoboda, D. Man and J. Boh, Class A predictions of a NATM tunnel in stiff clay. Computers and
Geotechnics 37 (2010), No. 6, 817-825.
[7] A. Sadrekarimi and S.M. Olson, Critical state friction angle of sands. Gotechnique 61 (2011), No 9,
771-783.
[8] J.H. Atkinson, W.H.W. Lau and J.J.M. Powell, Measurements of soil strength in simple shear test. Can.
Geotech. J. 28 (1991), 255-262.
[9] D.W. Airy and D.M. Wood, An evaluation of direct simple shear tests on clay. Gotechnique 37 (1987),
No 1, 25-35.
[10] J.B. Burland, ON the compressibility and shear strength of natural clays. Gotechnique 40 (1990), No
3, 329-378.
[11] D. Man and J. Rott, Small strain stiffness anisotropy of natural sedimentary clays: review and a
model. Acta Geotechnica 9 (2014), 299-312.
[12] J. Rott, D. Man, J. Boh, M. Krupika and T. Mohyla, Evaluation of K 0 of stiff clay by back-analysis
of convergence measurements from unsuported cylindrical cavity. Submitted to Acta Geotechnica.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1169
Interpretations of measurements for both

rigid and flexible retaining system in
cohesive cemented sandy soils in Lebanon
Youssam KAZAN
Order of engineers of Beirut, Lebanese university, Faculty of engineering, Lebanon
Abstract. In this present paper we present deflection measurements of rigid and

flexible, anchored and not anchored retaining walls of a deep excavation in
cemented sands located in the north east suburb of Beirut. Qualitative
interpretation of these data shows specific phenomena for this kind of middle size
projects. It leads us to proceed in a specific modeling more developed than simple
back analysis in order to simulate the corresponding curves. As reminder, we make
here also a general back view on the approaches adopted by engineers in order to
rectify and define the good approach of geotechnical practices in such materials.
Keyw ords. Cemented sand, deflection, back analysis, inclinometer
1. Introduction
Cemented sandy soils are abundant in Lebanon. They are spread in almost fifty percent
of coastal areas [1]. They are sometimes cemented by siliceous clay or by high
concentration of calcareous contents. In this case, they are real *-c soils, and the
cohesion has a big impact on the behavior of structures. The cohesion value is
depending on water presence and conditions. It is very sensitive to thermal variation
and altered easily whenever exposed to open air and meteoric change.
In some other cases, cemented sandy soils are non-cemented, they are considered as
classic frictional soils and they are not treated in this paper.
These soils are usually compact and very dense, and sometimes gravelly. They are
classified very good in most classification system. These soils may be loose; they
will be treated as normal pure frictional sand. The characterization will be done in the
standards of soil mechanics. In other cases these geological formations are highly
cemented and rocky. Their characterization does not lead to confusion in applying the
rock mechanics approaches. Sometimes they may have cavities [2], the treatment of
this problem is beyond the scope of this paper.
However, the intermediate case, when cemented sand is within the limit of soft rock
and hard soils, leads to some confusion [3]. In the geological context of Lebanon, the
soil is formed by granular quartz in a fine matrix. The sampling becomes difficult and
the results of in-situ tests will be widely dispersed. It is very difficult to obtain intact
sample to use in laboratory tests devices and machines. The visual end laboratory
examination of remained samples considered them as non cohesive soil, and this is not
realistic.
1170 Y. Kazan / Interpretations of Measurements for Both Rigid and Flexible Retaining System
Various types of retaining systems, flexible or rigid, are used in deep excavations in
such soils. However, the monitoring of these screen after excavations are rare, and it is
almost impossible to find published values before the last decade [4]. The only
criterion of validation of the method and parameters used in the calculation of these
geotechnical structures is the stability of the excavation during execution work. Thus,
retaining system, if they are recommended and executed, are often over designed.
In the case history presented in this paper, we found in a single small project (and
the same soil mass) both flexible and rigid retaining walls (anchored and not anchored).
Compared with the measurements in huge project, these data may be considered not
important; however there will be very interesting consideration of the comparative
phenomena:
Figure 1. General view of the excavation and piles

The right side of figure 1 we can see a flexible retaining system executed by
shotcrete and stabilized by nails and one row of tiebacks to retain the west side. The
neighbors forbid the use of anchors; the southern side was retained by 60 cm o. d. piles
executed after 4 m of free excavation, as viewed in front of photo. The anchored
retaining system in the east side is described paragraph 3 later (see figure 5).
2. Historical Geotechnical and Structural Studies context
The difference between in situ mechanical characteristics of frictional cohesive soil and
laboratory proposed geotechnical parameters lead engineers in Lebanon to make back
analysis of several deep excavation in the last decades. Studies were especially focused
on marl [5], [6] and on cemented sands [7]. A common conclusion of this research
analysis was that the in situ modulus of elasticity has the most influence on the
structures, as shown in the following table corresponding to typical retaining structure
in sandy soil.
Y. Kazan / Interpretations of Measurements for Both Rigid and Flexible Retaining System 1171
Table 1 Comparison between initially proposed values and values deduced from back analysis for typical
cemented sand.
Soil parameters upper layer main layer
Initially c (KPa) 10 25
proposed * degree 32 36
values E (MPa 5 20
Values c (KPa) 32 35
Deducted from * degree 10 36
Back analysis E (MPa 20 10+5 z
The typical example related in this paragraph concern an anchored rigid retaining
wall (*80 in situ cast concrete spaced 1.5 m axe to axe with only two rows of anchors).
The figure 2 above shows the corresponding calculated displacement with a maximum
displacement of 80 mm. The monitoring of deflection did not exceed 2 cm of
deflection and was accepted by the engineer. The refine back analysis simulated
exactly the deflection curve and new set of parameters was proposed in table 1 above.
Figure 2 Measured and calculated deflection in cemented sand formation.

The original design recommends four rows of anchors in order to stabilize 14 m
excavation depth with adjacent building. However the engineer decided practically,
during the construction, to use only two rows with permanent monitoring.
3. Case study
3.1. Geometrical site description
It is a fifteen stories building with two basements and a ground floor. The surface of the
building excavation is approximately 800 m2. The plot is limited by road on the East
side and another road on the partially west side. An adjacent building with one
basement is located in the north east side.
Figure 3. Schematic plan of the excavation

The field has a trapezoid shape. The plot follows a very slight slope along the
eastern direction. The foundation level of future construction is planned to eight meters
deep.
3.2. Geotechnical profile and parameters
It is almost a homogenous soil. We do not consider in our analysis the thin top layer, a
quaternary loose silty formation containing essentially limestone gravels (less than one
meter thickness).
The soil is composed of two cemented sandy layers. The upper layer consists of red
brown, medium plastic, fine silty sand with gravel with a depth of about 3 m. The
lower layer consists of a red brown, plastic, dense to very dense, fine clayey sand with
gravel.
Figure 4. Geological and geotechnical stratigraphy of the site

During the site investigation works, the water table level was measured at 6 meters
depth. However the excavation works were undertaken during the dry winter 2012-
2013. The water level was below 11 m depth. It has no influence on the soil
deformations.
Tests identifications and consistency, sieve analysis and Atterberg limits, were
made on a selected samples to characterize the soil mass (or clayey) mentioned above.
Results were accorded with the mechanical tests performed on ten samples.
Interpretation of all these laboratory tests leads to propose the parameters in the
following table.
Table 2, Mechanical parameters of both layers proposed by the laboratory tests.

parameters upper layer Lower layer
c (KPa) 10 15
* degree 32 30
E (MPa) 10 25
sec (KN/m3) 19 21
sat (KN/m3) 21 23
These proposed values for the mechanical parameters were considered smaller the
real in-situ parameters as usually done in geotechnical practice in Lebanon (as
described in paragraph 2). The Engineer decides to use the observational method in
order to monitor the deformation of soil during excavation to optimize the cost of
retaining systems.
3.3. Retaining systems
The absence of approximate constructions in some sides allowed the engineer to

execute a flexible anchored retaining shotcrete wall in the west side. The inclinometer
I2 was installed just beside the limit before excavation (Figure 5). However, it was
recommended to execute rigid retaining wall in the east side to avoid any instability
problem on the main road.
It was forbidden by the neighbor to execute anchors in the south free side, the
engineer decide to execute vertical non anchored piles in the lower part and leave the
upper four meters without retaining in order to minimize the lateral pressure on the in
situ cast concrete piles as shown in figure 5.
Figure 5. Cross section (and side view in the middle) shows three different retaining systems
4. Measurements
Three inclinometers were installed in three sides of the site. Inclinometers I1 and I2
were installed in the soil of adjacent plots in the west and south side respectively. I3
was installed in one pile in the east side during the piles execution. Inclinometer
measurements in each were made in a sequence of half a meter. Figures 6 and 7 show
the curves of deflection corresponding to same excavation phases.
4.1. Flexible wall
The flexible retaining system in west site is a 15 cm thickness of shotcrete with one
row of anchors of 15 tons spacing of 3 m. the first curve was measured before
execution of anchor. Displacement is similar to a free edge beam, with a total of 2 mm.
The anchor was installed and a new stage of construction was done until 7m depth and
the second curve was measured. The influence of anchor is clear. In this stage some
passive nails were executed 3 to 4 m spaces. The third curve corresponds to the final
stage of excavation.
One more curve was measured, three months after the final stage of excavation. We
can clearly see additional deflections of about 2 mm. These deformations may result
from three different reasons:
+ plasticity and creep behavior of soil
+ relaxation of anchors and nails
+ Variation of the groundwater conditions
Modeling and interpretation of these phenomena must be undertaken carefully
Figure 6. Deflection of flexible retaining screen

4.2. Non anchored rigid wall
For construction rules, the retaining system in the south side must be without anchors.
He cannot propose flexible system as the west site, the engineer propose * 60 cm
outside diameter spaced 1.5 m. Excavation depth is about 8 m depth, the calculation
design non admissible deflection. So the final solution was to execute the rigid piles
after 3 or 4 m of free excavation as shown in figure 1 and 5. This solution is acceptable
because of cohesion in cemented sand. The total depth of piles was only 8 m, four of
them embedded under the excavation level.
Interpretation of curves in figure 7 shows that deflection in the upper free
excavation (above 4 m depth) is clearly higher than deflection in the lower part of
excavation.
Comparison between the two last curves shows the influence of non elasticity in
soil. The simulation studies should be focused on whether this phenomenon is due to
the variation of water condition or viscosity in soil.
Only more significant deflection in the upper part is detected (about 10 mm) after
3 months. The origin of this phenomenon may by the variation of the interface
condition between soil and in inclinometer casing.
4.3. Anchored rigid wall
This corresponds to the east part. It is a combined structure: L section reinforced

concrete wall (3 m height) executed on * 60 cm diameter spaced 1.5 m. The foundation
of the upper part (L section) was the cap beam (2m width) of the piles.
Figure 7. Deflection of rigid retaining walls

A total of 6 anchors (stressed 20 tons each) of 12 m length were installed at 4.5 m

depth in this retaining system (1m below the bottom of cap beam).
The deflection of piles in the end of excavation not exceeds 2 mm in comparison
with the 8 mm of the flexible west side. This shown the utility of rigid retaining system
and the pre stressed tiebacks.
However a big values was measured after three months, it can be estimated by two
times the elastic deflection. This phenomenon is to be clarified by special interpretation
and modeling.
5. Conclusions and perspectives
We presented in this paper results of monitoring of three different retaining systems

(flexible and rigid, anchored and not anchored) in same small deep excavation. The
homogeneous soil conditions for the three systems lead us to focus interpretation on
both geotechnical and structural parameters separately.
The future back analysis elements, using one set of mechanical parameters, will be
focus on the interpretation of the following issues:
+ Emplacement and depth of row of anchors
+ Influence of pre-stressing of these anchors.
+ Thickness of the flexible shotcrete wall.
+ The elastic geo-material proprieties
+ The origin of additional relaxation deformations
References
[1] Dubertret L. 1931, Carte gologique du Liban : Revue de gographie de Lyon 31.3: 265-266
[2] Kazan Y. 2002 Gologie du Liban vue par un gotechnicien. Revue Franaise de gotechnique n 100.
[3] Khawlie M. et al. 1987 ; Terrain analysis for development studies. Geological Society of Hong Kong
Bulletin N 3
[4] Fawaz A. et al. 2005 Analyse inverse et modlisation dessais pressiomtriques ralises sur un sol de
Beyrouth. Revue Franaise de gotechnique n 112 : p43-48.
[5] Saydi Ch. and al. 2013 Monitoring, Back analysis and Redesign of a Deep excavation Retaining system
in marl formation.. MATEC Web of Conferences journal, proceeding of CMMS 2013, Rabat
[6] Kazan Y. 2014 Modelling and Redesign of Retaining system of a deep excavation in marl formation
using back analysis results. Proceeding of the ISRM European Rock Mechanics Symposium, May 2014.
[7] Chour A. Monitoring and back analysis of retaining system deformation in sandy soils and sandstone.
Interior report of Lebanese University, July 2014.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1177
Primary Consolidation Settlement of

Embankments Constructed on Young Bay
Mud of San Francisco Bay
Abhijit BATHE a , W. Martin McCABE a, L. Sebastian BRYSON b, 1, Frank
LOBEDAN c, and Samuel WON c
a
URS Corporation, Seattle, WA, USA
b
University of Kentucky, Lexington, KY, USA
c
Port of Oakland, Oakland, CA, USA
Abstract. Active Airports are complex environments for geo-construction.

Construction activities must not interfere with takeoff and landing operations and
the final earthwork must provide adequate bearing support for the aircraft and
support equipment. This case study discusses challenges faced in the design and
construction of two 5 to 6 ft (1.52 to 1.83 m) high embankments for new taxiways
at Oakland International Airport in Oakland, CA. The taxiway embankments were
constructed partly on a manmade wetland located on top of Quaternary Young Bay
Mud (YBM). YBM is a thick deposit of soft, saturated, unconsolidated silty clay in
the San Francisco Bay area which exhibits low shear strength, high compressibility
and low permeability, making it a very problematic foundation material. A ground
improvement program was undertaken to expedite approximately 90 percent of
estimated primary consolidation settlement to meet the construction schedule. The
ground improvement program consisted of using a maximum permitted surcharge
near the operating runway and Prefabricated Vertical Drains (PVD). The
settlement of both surcharged embankments was monitored using settlement plates.
In addition, soil parameters were updated using ongoing measurements to revise
the estimated final settlement and time required to achieve geotechnical
stabilization. This case study discusses the parameters used for acceptable
geotechnical stabilization, compares the estimated and actual magnitude and
duration of settlement with and without ground improvement, and comments on
conventionally used soil parameters for the Bay Mud.
Keywords. Consolidation Settlement, Prefabricated Vertical Drains, Preloading,

San Francisco Bay Mud, Airport Taxiway
1. Introduction
The Oakland International Airport is located on the east side of San Francisco Bay in
Alameda County, California. The site of Oakland International Airport was reclaimed
in the late 1950s from tidal marshes and shallow water areas of San Francisco Bay by
constructing a perimeter dike and then hydraulically placing fill from a borrow pit
southwest of the site. This perimeter dike forms the western shoreline of the Oakland
Airport property. The airport was then constructed on the hydraulically placed fill. The
1
Corresponding author, Department of Civil Engineering, 161 Raymond Bldg., University of Kentucky,
Lexington, KY 40506 USA, email: [email protected].
1178 A. Bathe et al. / Primary Consolidation Settlement of Embankments Constructed
airport is undertaking several projects to upgrade the Runway Safety Areas (RSA)
pursuant to the nationwide congressionally-mandated aviation safety program to
improve RSAs at all commercial airports with the United States. The RSA is a defined
area surrounding the runway, the surface of which must be prepared or suitable for
supporting aircraft in the event of a runway excursion (i.e. aircraft veering off either
side of the runway), overruns and underruns.
One of these projects consisted of extending the runway by 520 ft (158.5 m), and
adding two taxiways; Taxiway (TWY) W2 and W7, as shown in Figure 1. The taxiway
embankments were constructed partly on a manmade wetland created by the perimeter
dike and located on top of Quaternary Young Bay Mud (YBM), which is a thick
deposit of soft, saturated, unconsolidated silty clay in the San Francisco Bay area which
exhibits low shear strength, high compressibility and low permeability, making it a
very problematic foundation material.
Figure 1. Construction areas.

A ground improvement program was undertaken to expedite approximately 90
percent of estimated primary consolidation settlement to meet the construction
schedule. The ground improvement program consisted of using maximum permitted
surcharge near the operating runway and Prefabricated Vertical Drains (PVD) with 4-ft
(1.22-m) triangular spacing. The PVD program was implemented only for Taxiway W7
due to operational restrictions. However, the settlement of both surcharged
embankments was monitored using settlement plates. In addition, soil parameters were
updated using ongoing measurements to revise the estimated final settlement and time
required to achieve geotechnical stabilization. An acceptable level of geotechnical
stabilization was achieved despite various challenges encountered during construction
of the embankments for these two taxiways. These challenges included:
1. Restriction on height of surcharge, which required construction in stages to
avoid interference with aircraft takeoff and landing operations.
2. Limited time to achieve an acceptable level of geotechnical stabilization such
that the pavement subgrade was stable enough to provide adequate bearing for
the aircraft and support equipment.
3. Mitigation of differential settlement resulting from two embankments with
varying thicknesses constructed at different times and keeping post-
construction total settlement at acceptable magnitudes. The Runway (RWY)
12 extension portion received about 3 ft (0.91 m) of new fill in an area that
was previously subjected to hydraulically placed sand fill, whereas Taxiway
W7 connecting to Runway 12 received about 6 ft (1.83 m) of new fill.
A. Bathe et al. / Primary Consolidation Settlement of Embankments Constructed 1179
4. The designed geotechnical stabilization procedure needed to be adjusted in

real-time based on observed settlement, changes in time available to achieve
settlement, and restrictions on height of construction equipment.
2. Site Soil and Groundwater Conditions
San Francisco Bay is located in the central portion of the Coast Ranges geomorphic
province of California. San Francisco Bay is a structural trough partially filled with
non-marine and marine sedimentary deposits and locally with artificial fill around the
bay margins. Bedrock geology in the Berkeley and Oakland hills along the east side of
San Francisco Bay is structurally and lithologically complex. Along the bay plain,
these rock units are overlain by alluvial sediments, old Bay Clay, and along the edge of
the bay, an uppermost layer of fine grained marine deposits known as Young Bay Mud
(YBM). YBM varies considerably in thickness, generally less than 10 ft (3.05 m) thick
near the shore to more than 100 ft (30.48 m) thick far offshore. These generally soft
muds interfinger with silt, sand and gravel near the mouths of streams that carry these
coarser materials to the Bay [1]. In addition to these terrigenous clastic interlayers, peat
and shell beds also exist within the YBM [2].
A geotechnical investigation was performed to provide information on the
thickness and compressibility of YBM in the low-lying wetland pond area at the
intersection of Taxiway W and W7. This area was where the greatest thickness of new
fill was placed. The investigation also focused on Taxiway W2, which was an area that
had seen little or no previous fill. Geotechnical exploration was performed in stages
and consisted of borings with standard penetration test (SPT) sampling and cone
penetration test (CPT) probes. The locations of the borings are shown in Figure 2. The
locations of Taxiways W7 and W2 are as shown in Figure 2a and 2b, respectively.
(a) (b)
Figure 2. Boring and settlement plate locations.

The subsurface profile typically consisted of sandy fill overlying YBM.
Groundwater was found at depths ranging from one to 10 ft (3.05 m) below the ground
surface, with the shallowest groundwater at locations where the ground surface
elevations were also somewhat lower such as at Taxiway W2 and W7.
The condition of the sandy fill was observed to be variable, with density varying
typically between loose to medium dense. Borings at ground surface elevations of
approximately 0 ft in the low-lying wetland area of Taxiway W7 revealed an upper
layer of 8 to 9 ft (2.44 to 2.74 m) of medium dense sand. At Taxiway W7, the sand was
observed to be overlying approximately 18 ft (5.49 m) of YBM and 11 ft (3.35 m) of
Old Bay Mud, separated by 2 ft (0.61 m) of loose, wet silty sand. The groundwater was
observed at a depth of one foot (0.3 m). Borings at ground surface elevations of 2.5 to 3
ft (0.76 to 0.91 m) at Taxiway W2 revealed 4 to 7 ft (1.22 to 2.13 m) of loose sand
overlying YBM. While the borings drilled at Taxiway W2 did not fully penetrate the
YBM layer, historic borings in the vicinity of Taxiway W2 indicated that the YBM was
approximately 10 ft (3.05 m) thick in this area. The YBM was observed to consist of
very soft, highly plastic, wet, soft clay with occasional seashells. The Old Bay Mud
consisted of medium stiff to stiff, medium plastic, wet, dark gray lean clay and was
observed to be homogenous. The groundwater was observed to fluctuate with tides.
Similar native soil layers were observed at the Runway 12 extension areas where a fill
thickness was observed to be 17 ft (5.18 m) with ground surface elevation of about 6.0
ft (1.83 m).
3. Geometry of the Embankments
The taxiway embankments took into consideration the grading criteria for runway and
taxiway safety areas in their final configuration. Thus, for Taxiway W7 a trapezoidal
cross-sectional geometry consisting of a 6-ft (1.83-m) high embankment with
approximately a 80-ft (24.38-m) top width and 0.8 (horizontal):1 (vertical) side slopes;
resulting in 210-ft by 705-ft (64-m by 214.88-m) footprint of the embankment.
Similarly, a footprint of 270-ft by 480-ft (82.3-m by 146.3-m) was considered for
Taxiway W2 with embankment average height of 5 ft (1.52 m) above the existing
ground surface.
4. Consolidation Settlement Calculation
A summary of the soil parameters for the YBM used for the consolidation calculations
are shown in Table 1.
Table 1. Soil parameters for the Young Bay Mud.
Dry Coefficient of
Moisture
Unit Compression Consolidation
Project Location Boring Content
Weight Index (ft2/yr)
(%)
(pcf)
Taxiway TW-7 and 61 to
33 to 71 0.27
W2 TW-8 86
Taxiway TW-1 and 95 to 108 88 0.35
Oakland 3.7 to 22.7
W7 TW-2 97 92 0.28
International
Airport UB-1 88 50 0.38
Dike
UB-4 58 64 0.26
Study
UB-12 100 45 0.42
Not 39 to
ATCT 33 to 119 0.16 to 0.20
specified 66
Other NW SF
NA NA NA 0.40
projects Bay
around the Hamilton 70 to
NA 40 to 90 0.25 to 0.45
bay AFB 90
3 2 2
1 kN/m = 6.3658 pcf; 1m /yr = 10.7639 ft /yr
Consolidation settlement due to the new Taxiways W2 and W7 and due to Runway
12 extension was calculated using soil parameters obtained from following sources:
1. Index and consolidation laboratory tests performed on borings drilled for this
project within the footprint of the taxiways,
2. Available soil data along the runway and other taxiways for this and previous
projects,
3. Studies performed along the dike for another project for perimeter dike
evaluation, and
4. Settlement monitoring observations performed previously for the Air Traffic
Control Tower at the Oakland International Airport.
4.1. Time Rate of Consolidation
For Taxiways W2 and W7, the primary objective was to achieve geotechnical
stabilization by achieving at least 90 percent of primary consolidation settlement within
the main taxiway pavement footprint. Adequate geotechnical stabilization of the
proposed embankments was defined as achieving 90 percent of primary consolidation
settlement with the assumption that remaining primary and secondary consolidation
settlement would not result in more than 1 inch (25.4 mm) over 100 ft (30.48 m)
distance differential settlement. The intent was to minimize the potential for long-term
flexible pavement maintenance for the airport.
The time constraints for the project were such that approximately 10 to 12-weeks
were available to achieve geotechnical stabilization for Taxiway W7 and about 30-
weeks were available for Taxiway W2. Based on available lab data and literature
review, the average initial vertical coefficient of consolidation ( cv ) for the airport was
estimated to be 10 ft2/yr (0.93 m2/yr). Time rate of consolidation calculation for PVD
conservatively used a horizontal:vertical coefficient of consolidation ratio of one (i.e.
ch cv 1 ).
4.2. Initial Analyses
For the project, the final embankment was termed as preload and the additional load
termed as surcharge. The settlement calculations for the preload alone estimated
approximately 24 in (609.6 mm) of primary consolidation settlement along the
centerline of Taxiway W7. Additional consolidation analyses estimated that about 29 in
(736.6 mm) of primary consolidation settlement would be expected with roughly 2.5 ft
(0.76 m) of surcharge. The time rate of consolidation calculations associated with the
2.5-ft (0.76-m) of surcharge indicated that approximately 21.5 in would be achieved
within the available construction timeframe. This was the same settlement magnitude
as 90 percent of primary consolidation settlement without surcharge. Similarly,
consolidation settlement of approximately 12 in (304.8 mm) within the available
timeframe was estimated for Taxiway W2 with 3 ft (0.91 m) of surcharge.
It is noted that the consolidation analyses were performed for the YBM assuming
the layer to be normally consolidated. Although some consolidation tests from previous
studies suggested pre-consolidation of the YBM, this study found no obvious evidence
of desiccation of upper portions of YBM. Thus, for conservatism the YBM was
assumed to be normally consolidated.
5. Design for Geotechnical Stabilization
Various methods such as use of light weight fill, deep compaction, use of stone
columns and preloading were originally explored to mitigate the anticipated
consolidation settlement in the given timeframe. However, they were discarded either
due to cost or schedule constraints. Limited surcharging with use of Prefabricated
Vertical Drains (PVD) was considered as a viable option for both taxiways and
performed in the foundation treatment area extending 10 ft (3.05 m) beyond the edge of
flexible pavement.
These drains are flexible, prefabricated, rectangular shaped, composite products.
The drain conduit consists of a polyolefin inner core completely wrapped by a filter
fabric. The PVD was shall be installed using a mandrel that advanced through the
compressible soils to the required depth using a constant rate of advancement. For the
project, the Mebra Drain MD-7407 PVD was used.
For Taxiway W7, 3-ft (0.91-m) triangular PVD spacing was used inside the RSA
and Taxiway Object Free Areas (TOFA) due to shorter duration for consolidated
settlement (9-weeks), and 4-ft (1.22-m) PVD spacing was used for the non-RSA areas
with PVD installed to a minimum depth of 30 ft (9.14 m). Only 2.5 ft (0.76 m) of
surcharge was allowed for Taxiway W7 due to FAA height restrictions. For Taxiway
W2, 4-ft (1.22-m) triangular PVD spacing was used for all foundation treatment areas
with no surcharge. A PVD depth of 17 ft (5.18 m) was recommended for Taxiway W2.
It was anticipated that with this preload and surcharge; required geotechnical
stabilization could be achieved within the available duration.
6. Field Instrumentation
Settlement plates were used to monitor the magnitude and rate of settlement during
construction. Settlement plates consisted of a steel plate with coupling for attaching the
central rod and protective PVC casing. Four settlement plates were installed along the
centerline of Taxiway W7. These settlement plates were designated 7A, 7B, 7C, and
7D and are shown in Figure 2a. Four settlement plates were installed along the
centerline of Taxiway W2. These settlement plates were designated 2A, 2B, 2C, and
2D and are shown in Figure 2b. A typical cross-section of the embankment showing the
PVD and settlement plates is shown in Figure 3.
Figure 3. Typical embankment cross-section with settlement plate.

7. Stages of Embankment Construction
The construction schedule for the embankment construction (preload and surcharge) of
Taxiway W7 is given in Table 2. Construction started in June 2013 and had a duration
of 21-weeks.
Table 2. Embankment construction schedule for Taxiway W7.
Week Construction Activity

Week 0 to 5 Runway was shortened on Runway 12 end, after constructing Taxiway Z to
access the shortened Runway. Due to the presence of wetland and to provide a
platform for construction equipment, the contractor placed 2 ft of embankment
fill prior to installation of settlement plates. It was estimated that about 2-in
settlement due to this activity was not recorded. The contractor installed
settlement plates 7B, 7C and 7D and began installation of PVD. Due to delays
in the Federal Aviation Administration (FAA) approval process regarding
height restrictions for equipment working adjacent to active runways, the PVD
installation process was delayed and PVDs inside RSA and TOFA were
cancelled.
Week 5 to 7 Contractor placed preload and surcharge without adequate compaction to meet
the schedule. Contractor was provided additional time to complete paving.
Week 14 7B was damaged due to construction activity, but restored. Based on the
observed settlement, the coefficient of consolidation was updated to 15 ft2/year
resulting in a revised maximum settlement estimate of 18 in.
Week 18 Contractor was asked to place an additional 1-ft of surcharge in order to
expedite the settlement.
Week 21 It was concluded that required geotechnical stabilization was achieved in about
14-weeks from the placement of preload and surcharge. The surcharge was
removed and pavement base and flexible pavement was installed.
The embankment construction (preload and surcharge) schedule for Taxiway W2

is given in Table 3. Construction began in October 2013 and was ongoing for 24-weeks.
Similar to W7, the FAA restricted equipment heights such that PVDs could not
practically be installed for Taxiway W2 while the runway is operational. Consequently,
it was decided that additional embankment fill surcharge would be considered based on
the consolidation settlement monitoring results in lieu of the PVD.
Table 3. Embankment construction schedule for Taxiway W2.
Week Construction Activity

Week 1 Runway was shortened on Runway 30 end, after constructing Taxiway U to
access the shortened Runway. Due to the presence of low strength, saturated
soils, the Contractor had construct a platform for construction equipment,
consisting of geotextile and triaxial geogrid on the existing ground surface prior
to placing fill materials. The contractor installed the settlement plates 2A, 2B,
and 2C and began placement of preload.
Week 20-22 Approximately 3 ft of additional embankment fill surcharge was placed to
expedite the settlement and provide the contractor with additional time to
complete paving. The fill was heaped in the center portion of future Taxiway
where the YBM layer was the thickest.
Week 24 It was concluded that required geotechnical stabilization was achieved in about
24-weeks from the placement of preload and surcharge. The surcharge was
removed and pavement base and flexible pavement was installed.
8. Comparison of Estimated and Observed Settlements
The estimated (i.e. calculated) and observed settlement data for Taxiway W7 are given
in Figure 4. Based on the lab results obtained at Taxiway W7 and data from similar
projects, the compression index for the YBM was conservatively estimated to be 0.34
resulting in a total settlement of 24 in (609.6 mm) without surcharge. As shown in
Figure 4, a theoretical curve was plotted assuming a coefficient of consolidation of 10
ft2/yr (0.93 m2/yr) and maximum drainage path of 2 ft (0.61 m), owing to 4-ft (1.22-m)
triangular PVD spacing, for the embankment with 2.5-ft (0.76-m) surcharge. This
resulted in a total settlement of about 29 in (736.6 mm). The coefficient of
consolidation was subsequently revised to a value of 15 ft2/yr (1.39 m2/yr) to match the
observed settlement measurements. This produced a maximum settlement value of
approximately 18 in (457.2 mm).
Time (months)
0 1 2 3 4 5
0
5 Phase 1 Phase 2 Phase 3

Revised Calculation Parameters:
Embankment Max. Settlement = 18-inch
Settlement (inch)
Coefficient of Consolidation = 15 ft2/yr

10
15 Original Calculation Parameters:

Embankment Max. Settlement = 24-inch
Original Settlement Estimate

20 Observed Settlement @ B
Observed Settlement @ C
Observed Settlement @ D
Revised Time Rate of Settlement
25
Figure 4. Settlement monitoring data for Taxiway W7.

It was speculated that the initial variance between the observed and estimated
settlement may have been due to the stress history of the YBM. Observations from this
project and others projects performed by the authors in the San Francisco Bay area tend
to suggest that if stress history or stiffening effects of YBM are not considered, ground
settlement with a relatively significant fill cover should be expected to be about 2/3rd of
the total calculated consolidation settlements. For this study, based on the observed
settlement data, the maximum settlement estimate for the embankment with surcharge
at Taxiway W7 was reduced by 25 percent to be conservative and the targeted
settlement was determined to be 18 in (457.2 mm). The settlement data for Taxiway
W2 are shown in Figure 5. In the case of Taxiway W2, a similar reduction of about 25
percent was observed for the total settlement. The interesting observation for Taxiway
W2 was that the observed coefficient of consolidation was more than twice that of
Taxiway W7 without installation of PVD. This variation is most probably due to lack
of pre-consolidation owing to shallow overburden on the YBM and not due to reduced
thickness compared to Taxiway W7 YBM.
Time (months)
0 1 2 3 4 5 6 7 8 9 10
0
Phase 3 Phase 4
1
Revised Time Rate of Settlement
2 Observed Settlement @ A
Observed Settlement @ B
3
Observed Settlement @ C
4 Original Settlement Estimate
Settlement (inch)
5
Revised Calculation Parameters:
6 Embankment Max. Settlement = 9-inch
7
10
Original Calculation Parameters:
11 Embankment Max. Settlement = 12-inch
12
Figure 5. Settlement monitoring data for Taxiway W2.
9. Conclusions
This case study discussed the methods used to produce acceptable geotechnical
stabilization within a very narrow timeframe at an active airport. Comparison between
estimated and actual magnitude and duration of settlement show the settlement with
relatively significant fill cover should be expected to be about 2/3rd of the total
calculated consolidation settlements if stress history or stiffening effects of YBM are
not considered. In addition, observations from this project suggest that for similar
situations with unconsolidated YBM with up to 5 ft (1.52 m) of drainage path,
application of preload can be a viable option rather than installation of PVD.
References
[1] Schloker, J. (1968). Regional geology, Supplemental Report IS-3, Association of Bay Area
Governments, Bay Area Regional Planning Program, pp. 47.
[2] Nicholas, D.R. and Wright, N.A., (1971). Preliminary map of historic margins of marshland, San
Francisco Bay, California, U.S. Geological Survey Open-File Report 71-216, scale 1:125,000, U.S.
Geological Survey.
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1186
Seismic Wave Velocities of Soft Clays

Measured with Piezoelectric Crystals in an
Oedometer
Miguel DIAZ PARDAVEa,1 and Efran OVANDO-SHELLEY b
a
MSc Student, Instituto de Ingeniera, Universidad Nacional Autnoma de Mxico
b
Instituto de Ingeniera, Universidad Nacional Autnoma de Mxico
Abstract. We present some of the results of an experimental investigation aimed at

characterizing soil samples in terms of seismic wave velocities estimated from measurements
with piezoelectric crystals. Specimens of Mexico City Clay were mounted on an instrumented
consolidometer in which we obtained compression and shear wave velocities. The paper
examines the close field effect on shear wave velocity measurements by looking at the signals
recorded in the piezoelectric receivers in response to sinusoidal signals induced at different
frequencies that ranged between 1 and 7 kHz. We use the specimens transfer functions to
obtain their natural frequencies; with them and with a dimensionless parameter (the quotient of
the distance between the emitter and the receiver and wave length), we put forth a procedure to
determine the minimum frequency required to avoid near field effects and, hence, to obtain
reliable values of shear wave velocities.
Keywords. Wave velocity, near field effect, excitation frequency, transfer

function, systems fundamental frequency
1. Introduction
Piezoelectric crystals installed in oedometers to measure seismic wave propagation

velocities have been used in the past, [1, 2, 3, 4, 5]. The experimental gear developed
thence was used to look at the effects of various stress conditions on shear wave
velocities subjected to loads, restraining lateral deformations for a variety of soil types.
In this investigation we equipped an oedometer with piezoelectric crystals to evaluate
these effects on extremely soft samples of the former Texcoco Lake, north of central
Mexico City. This paper presents and discusses some of these results. We look at the
near field effect in our oedometer and put forth a means to evaluate it in terms of
analyses of waveforms in the frequency domain. Figure 1 provides a sketch of our
experimental set up.
1
Corresponding Author: Miguel Daz Pardav, direccin, E-mail:
M. Diaz Pardave and E. Ovando-Shelley / Seismic Wave Velocities of Soft Clays 1187
2. Materials and experimental techniques
We report the results of shear wave velocities measurements in samples of soft

Texcoco Lake Clay. The main soil stratum at the site of sample retrieval is constituted
by soft clay strata with an overall thickness of 34 m, overlain by a non plastic
desiccation crust and underlain by coarse sandy silts. The clay strata are highly
compressible lakebed deposits of volcanic origin.
We used titanium zirconate bender elements 15 mm long, 6.4 mm wide and 0.38
mm thick and compression disks 6.4 mm diameter and 0.38 mm think as well as a
function generator manufactured by the Institute of Engineering of UNAM, a digital
oscilloscope of model HP54540A, manufactured by Hewlett Packard USA, and a
dedicated Lab-view data acquisition program. Data were stored in a computer provided
with a code to analyse and process time signals. The sketches in Figure 1 illustrate our
testing equipment.
Soil specimens were retrieved with thin walled aluminium tube samplers that
were cut to the required length, 72 mm to 80 mm, 69 mm inner diameter. The samples
were left inside the tube and mounted in the oedometer, as indicated in the figure.
The samples were subjected to incremental loads that were left in place until
primary consolidation was over, 12 to 24 hours. Compression and shear wave
velocities were measured before applying a new increment. In this paper we only
discuss and comment results of bender element testing.
Figure 1. General set up of the array implemented to carry out the experimental program.
Arrival times of waves travelling from the source to the receiver can be determined
identifying the start of the signal at the source and the first deflection in the receiver
signal [6, 7, 8, 9, 10]. Although conceptually simple, the procedure may be quite
problematic, as reported by several investigators [7, 11, 12].
In estimating arrival times, instrumentation delay time should be taken into
account; otherwise arrival times will be in error. Other errors will arise because of
deviations and lack of uniformity in waveforms recorded in the receiver produced by
1188 M. Diaz Pardave and E. Ovando-Shelley / Seismic Wave Velocities of Soft Clays
the so-called near field effect which introduces ambiguity to the determination of
arrival times [8, 12, 13, 14].
The graphs in Figure 2 illustrate the near field effect. Receiver signals were
obtained from source signals whose frequency varied from 1 to 7 kHz. As seen, the
first inflection point in these signals cannot be precisely determined for the lower
frequency signals. The figure also illustrates, as a reference, a source signal with a
frequency of 1 kHz.
Figure 2. Wave forms obtained with sinusoidal pulses of different frequencies. Depth of sample: 19.4
m; applied stress: 60 kPa. Conditions before testing: Sr=93.4%, w= 93.5%; PI=247%.
Furthermore, looking at Figure 2, it is evident that the points chosen to indicate the
first arrival of shear waves vary for each signal recorded at the receiver from which
different shear wave velocities may be deduced. Note, however, that arrival times tend
to be equal as the frequency of the sinusoidal pulses increases, i.e. as near field effects
become less important. The graphs in Figure 3 show shear wave velocities obtained by
using sinusoidal pulses having different frequencies. The influence of the frequency of
input signals on the measured values of Vs is discussed later.
It has been suggested that near field effects in the receiver signals may be
mitigated or even eliminated altogether choosing an adequate frequency for the source
signals. The following parameter is a useful index to assess whether near field effects
will or will not be present in receiver signals:

(1)

where Ls is distance travelled by the waves, from the source to the receiver bender
elements; is the wavelength; f is the frequency of the input signal and Vs is the shear
wave velocity.
There are several propositions as to what should the values of be, to reveal the
possibility of having near field effects. It was initially suggested that they will not be
present when the parameter [13]. Later its was recommended that
occlusion of shear wave arrivals was especially severe when values varied between
0.25 and 4 [12] but others [11] have suggested that the near field effect can be
altogether avoided when Rd> 4. Finally, it has also been estimated that errors in Vs
determinations can be kept at less than 5% if Rd is not less than 0.625 [14].
Figure 3. Shear wave velocities as a function of effective stress for different frequencies. Conditions of
sample: retrieval depth: 15 m; Sr = 97 %; w= 79.9 %; PI= 260.6 %.
In what follows we look at an alternative means of avoiding the near field effect
and put forth a procedure for estimating shear wave velocity values unambiguously. In
order to achieve that, we first look at the transfer functions, H(), obtained from bender
element tests in which we applied sinusoidal pulses of various frequencies. The transfer
function represents a dynamic system that is constituted in this case by the source
bender element, the soil, and the receiver bender element. It is also assumed that this
system is independent of frequency. Transfer functions were obtained with:

(2)

where , and are the Fourier transforms of the source and the receiver
signals; stands for the representation of the transfer function in the frequency
domain.
In Figure 4 we present the transfer functions obtained from a sample of
Mexico City clay excited with sinusoidal pulses generated with a bender element. The
pulses had frequencies equal to 1, 2, and 7 kHz. As seen there, the functions
we obtained are all similar, for all practical purposes, i.e. with these results we confirm
that the transfer function we obtained is unique albeit this statement is true in a limited
sense since it pertains to the specific testing conditions and to the effective stress-strain
states present while doing the tests. These functions contain several peaks and in the
figure we identified the ones having the larger amplitudes, f1, f2, f3.
Figure 4. Transfer functions obtained applying sinusoidal pulses with different frequencies. Testing
conditons: applied effective stress: 39.2 kPa. Sample index properties: Sr = 97 %; w= 79.9 %; PI= 260.6 %.
Depth of extraction: 15 m.
From the curves in Figure 4 we identified the systems natural frequency as the
frequency at the peak with the large spectral amplitude, f2, where f2 =fo. Referring to
Eq. (1), we calculated Rd substituting the value of the natural frequency in it, fo, and
using the values of Vs determined from the application of sinusoidal pulses having
different frequencies. In the graph shown in Figure 5, we show the values of Rd thus
obtained and in Figure 5 these values are plotted against the frequency of the input
signal. As seen there, Rd becomes independent of frequency when the input frequency,
fin, becomes larger than 4 kHz. Under these circumstances Rd is nearly constant, and
equal to 1.69. We call this value Rd* and the frequency at which this happens, the
critical frequency, fcrit.
To avoid near field effects, we suggest, first, to determine Rd*; Vs values may then
be determined confidently, using input frequencies equal or larger than Rd*. In other
words,

(3)
Note, however, that Rd* and fcrit are stress dependent parameters. Data presented in
Figure 6, for example, were obtained applying and effective vertical stress of 58.8 kPa.
It is interesting to point out that if instead of using we use fo, the error in the
values of Vs obtained with Eq. (3) becomes unimportant.
Figure 5. Variation of Rd as frequency of the input signal is changed. Sample conditions same as in Figure 2
Figure 6. Rd as a function of the frequency of the input signal. Sample and testing conditions same as in
Figure 4.
4. Conclusions
Near field effects render the definition of arrival times in bender element testing in the
laboratory difficult and may induce large, unacceptable errors in the determination of
shear wave velocity. In this research we showed that, for our testing gear and the soils
we tested in our research, shear wave velocities depend strongly on the frequency of
sinusoidal pulses used to excite the source element.
Our results also revealed that there exists a critical frequency above which shear
wave velocity, Vs, becomes independent of frequency. We showed that this frequency,
called here fcrit, is found by exciting the sample with pulses having different
frequencies. Once fcrit is found, the parameter Rd becomes constant and equal to Rd*,
provided the applied frequency is equal or larger to fcrit. Vs can then be determined
unambiguously.
Our data suggest that good approximations to the actual value of Vs are obtained if
one replaces the value of fcrit with the value of f0, the natural frequency of the dynamic
system that is constituted by the source bender element, the soil, and the receiver
bender element. We also showed that f0 can be readily obtained from this systems
transfer function.
The values of f0, fcrit and Rd* depend on stress level and stress history and more
research is necessary to assess duly this dependency. It is also a matter of future
research to determine if and how these values are affected by testing conditions and
soil type.
References
[1] Thomann, J. G. and Hryciw, R. D., Laboratory Measurement of Small Strain Shear Modulus under ko
Conditions, Geotechnical Testing Journal, 13, 2, (1990) 97105.
[2] Lo Presti D.C.F. et. al., Maximun shear modulus measurement using bender element in oedometer tests,
Rivista Italiana di Geotecnia (1993).
[3] Fam, M.A., and Santamarina, J.C., Study of geoprocesses with complementary mechanical and
electromagnetic wave measurements in an oedometer, Geotechnical Testing Journal, 18, 3, (1995)
307 314.
[4] Shibuya, S., Hwang, S.C. and Mitachi, T., Elastic shear modulus of soft clays from shear wave velocity
measurement, Gotechnique, 47, 3, (1997) 593-601.
[5] Comina, C., S. Foti, G. Musso, and E. Romero, EIT oedometer: an advanced cell to monitor spatial and
time variability in soil, ASTM Geotechnical Testing Journal, 31, 5, (2008) 404412.
[6] Shirley, D. J., 1978, An improved shear wave transducer, Journal of Acoustic Society of America, Vol.
63, No. 5, pp. 16431645
[7] Dyvik, R. and Madshus, C., 1985, Lab Measurements of Gmax Using Bender Elements, Norwegian
Geotechnical Institute Publication, 161, Oslo.
[8] Jovicic, V., Coop M. R., and Simic, M., 1996, Objective Criteria for Determining Gmax from Bender
Element Tests, Geotechnique, Vol. 46, No. 2, pp. 357362.
[9] Lee, J. S. and Santamarina, J. C., 2004-2005, Bender Elements: Performance and Signal Interpretation,
J. Geotech. Geoenviron. Eng., Vol. 131, No. 9, pp. 10631070.
[10] da Fonseca, V., C. Ferreira, and M. Fahey. 2009. A Framework Interpreting Bender Element Tests,
Combining Time-Domain and Frequency-Domain Methods. ASTM Geotechnical Testing Journal 32,
no. 2: 91107.
[11] Brignoli, E., Gotti, M, and Stokoe, K., H. II, (1996), Measurement of Shear Waves in Laboratory
Specimens by Means of Piezoelectric Transducers Geotechnical Testing Journal, GTJODJ, Vol. 19,
No4, December, pp 384-397.
[12] Viggiani, G., Atkinson, JH, (1995), Interpretatin of bender elements tests, Gotechnique 45, No 1,
pp 149-154.
[13] Sanchez-Salinero, I., Roesset, J.M., and Stokoe, K.H., II. (1986). Analytical studies of body wave
propagation and attenuation. Civil Engineering Department, The University of Texas at Austin,
Austin, Tex. Geotechnical Engineering Report No.GR86-15.
[14] Arroyo, M., Muir Wood, D, & Greening P.P. (2003), Source near-field effects and pulse test in soil
samples, Gotechnique 53, No 3, pp 337-345
IOS Press, 2015
doi:10.3233/978-1-61499-601-9-1194
Determination of the Maximum Density of

Carbonate Silty Sands
Naemeh NAGHAVI ALHOSSEINI a 1, M. Hesham EL NAGGAR b, Abouzar SADREKARIMI c
a
PhD Candidate, Dept. of Civil and Environmental Engineering,
Western Univ., Canada
b
Professor, Dept. of Civil and Environmental Engineering,
c
Assistant Professor, Dept. of Civil and Environmental Engineering,
Abstract. The range and values of the limit densities are important in interpreting
laboratory results. In this study, the available procedures reported in the literature
and ASTM standard are evaluated to identify their limitations in maximum density
testing on well-graded carbonate silty sand samples with different fines content.
Four levels of fines content in the samples are studied and it is revealed that
particle degradation and segregation can affect the results of density tests. As a
result, appropriate test methods that preserve the original soil gradation and create
uniform soil distribution should be employed in determining soil densities.
Keywords. Carbonate silty sands, crushing, gradation, maximum density
1. Introduction
A critical review of the available methods for determining limit densities (ASTM
D4253 [1], ASTM D698 [2] and method by Yamamuro and Lade [3]) indicates that no
universal method exists for determining these values, and different procedures are used
for different types of soils and project limitations. For example ASTM D4253 [1]
cannot be used for soils that have more than 15% of fines and ASTM D698 [2] is used
for the projects that have large amounts of material available for this test. Also, particle
degradation and breakage could affect the results of density tests obtained from
ASTM D4253 [1] in carbonate sands. Therefore, it is important to determine soil
densities through procedures that do not alter the soil gradation. For example, the
ASTM D4253 [1] method is followed to determine the maximum density
(corresponding to the minimum void ratio, emin) of carbonate silty sands. Sieve analyses
are conducted on the test specimens before and after the density determination and the
results demonstrate changes in soil gradation, which could be attributed to rounding or
crushing of particles. As a result, the maximum density determined by using the ASTM
D4253 method [1] may not be applicable because of this degradation, and the index
properties calculated using these density values could not be valid.
Natural sands contain varying amounts of fine and there are difficulties associated
with particle segregation, electrostatic charges, and bulking in finding minimum void
ratio (emin) for sands with fines content more than 15% [4]. Different methods for the
1
Corresponding author, PhD Candidate, Dept. of Civil and Environmental Engineering, Western Univ.,
London, ON, Canada N6A 5B9; E-mail: [email protected].
N. Naghavi Alhosseini et al. / Determination of the Maximum Density 1195
determination of emin are not expected to give results of general validity [5], which
could be due to varying physical properties of sand. This is especially important for the
fine sand as the higher effect of inter-particle forces can lead to different minimum void
ratio depending on the method used for finding the value of minimum void ratios. For
the fine sizes of natural non-spherical and irregular shape particles, van der Walls
attraction and double layer repulsion compete with gravitational forces and play more
important role in arrangement of soil particles in aggregates through attractive and
repulsive forces between particles. These inter-particle forces increase tendency of fine
particles to adhere to each other [6].
Kuerbis et al. [7] and Lade and Yamamuro [8] showed that the minimum void ratio
of silty sand decreases as the silt content increases from 0 to 20%.
Guerrero found that well-graded sands have higher dry densities compared to
poorly graded sands with the same fine content. The author concluded that adding fines
to sand leads to a gradual increase of the maximum dry density up to a maximum
value. After reaching that value, increasing the fine content reduces the density [9].
The fine content, which results in maximum dry density, is called the threshold fines
content by Thevanayagam et al. [10].
Available methods for determining the sand minimum void ratio need large
amounts of material, are time consuming and have limitations. As the method of
deposition affects the values of minimum void ratios, different methods of maximum
density determination are examined to check the particle degradation and values of dry
densities.
In most studies reported in the literature, uniform clean sand was mixed with fines
and the resulting soil was gap-graded. Also, only quartz or silica sands with round
shapes were studied. There are a few studies in the literature on sand with angular
particles or crushable sand [11], [6] and [12].
In this research, samples of well-graded natural carbonate silty sands are prepared
by mixing different sizes of sieved and washed natural particles. In order to study the
effect of the fine contents and crushing on the determination of minimum void ratios,
different mixtures are made with changing the amount of fines (from 10 to 40% with
10% increments) while preserving the original shape of the soil gradation curve.
2. Material and preparation
The soil investigated in this study was collected from the lower region of the Boler
Mountain in London, Ontario. Figure 1 presents the average gradation of the original
soil, which classifies it as silty sand (SM) according to the Unified Soil Classification
System [13]. Based on SEM images, the grains were subangular to angular.
In order to study the effect of fine contents (FC) on the determination of minimum
void ratio (emin), four different mixtures with 10, 20, 30 and 40% of FC were prepared.
This was accomplished by sieving, washing and separating the original soil into
different ranges of particle sizes, and subsequently remixing them in order to produce
new gradations with different fines contents. As illustrated in Figure 1, compared to the
original grain size distribution, the mixtures gradation curves were only shifted
vertically and the original shape of the particle size distribution was preserved. The
effect of particle crushing caused by different methods was studied by comparing the
gradation curves following each test with that of the initial soil.
1196 N. Naghavi Alhosseini et al. / Determination of the Maximum Density
100
90
80
70
Percent Finer
60
50 FC=40%
40 FC=30%
30 FC=20%
20 Natural Gradation of
Boler Mountain Soil
10 FC=10%
0
0.01 0.1 1
Grain Size D (mm)
Figure 1. Soil gradations of Boler Mountain silty sand with different fine contents
Specific gravity of the particles (GS) for different mixtures was determined as 2.72
following the ASTM D854 [14] standard guidelines. A plasticity index of 2.5 was also
obtained from Atterberg limit test [15], which is considered a low plasticity index.
Based on the acid dissolution technique [16], the soil mixtures were determined to have
a carbonate content of 26 to 40%.
3. Methodology
Four different methods were used to determine emin, including 1) the ASTM D4253
method [1], 2) a simulated ASTM D698 [2] method in a small mold, 3) Yamamuro and
Lades method [3], and 4) Yamamuro and Lades method [3] in a small mold.
The standard procedure for ASTM D4253 [1] requires that the material be
compacted in a mold under a specified dead weight of 14-kPa while being subjected to
vertical vibrations. ASTM D4253 [1] test method produces a moderate amount of
particle breakage for many types of cohesionless soils and results in an increase of the
measured maximum index density. For some soils with 5 to 15% fines, compaction test
(ASTM D698 [2]) may be useful in determination of the appropriate maximum dry
density. [1] Application of the standard proctor compaction procedure as an alternative
method for obtaining maximum dry density needs large amounts of material.
Based on ASTM D698 [2], the maximum unit weight of a soil can be obtained by
using a standard compaction procedure, which minimizes particle segregation and
breakage. As there was not enough material to perform this procedure, principles of the
method were applied using a smaller size mold with a diameter and height of 50 and
100 mm, respectively. For this test, the effort energy per unit volume of soil was the
same as standard proctor test and a metal rod with a diameter of 30 mm diameter was
used to compact soil in layers.
The procedure developed by Yamamuro and Lade [3] was also used to determine
maximum dry densities of the mixtures. According to their method, a total soil mass of
800g is poured in 16 layers (each of them approximately 50g) in a graduated cylinder
and the cylinder is tapped twice on its four opposing sides with a rubber mallet after
deposition of each layer. Their procedure was simulated by pouring 320g of soil in 16
layers (each of them approximately 20g) in a smaller size mold with a diameter and
height of 50 and 100 mm, respectively.
4. Observations and results
A series of experiments were performed on sand with varying FC to determine the

maximum dry density. The results of these experiments are compared in Figure 2. The
differences in maximum dry densities established from each method clearly illustrate
the wide ranges of maximum dry density, which potentially results in producing
different relative densities.
The maximum dry density measured from the different methods are almost the
same for FC = 10%, except for the one calculated from the standard method of ASTM
D4253 [1]. In the ASTM approach, gradation curve of soil changes after the test and
this shows that some of the particles are crushed or rounded, which alters soil gradation
leading to a highly compacted material with a higher dry density. Compared to the
results obtained from moist compaction, the dry density is lower for the mixtures with
higher FC despite of some particle degradation in these mixtures. Lower dry density for
the mixtures with higher fine FC could be due to the segregation of fines and using dry
soil in the tests. Three other methods used in this study resulted in higher values of dry
density with the largest dry density obtained from the standard compaction in the small
mold. For this technique, dry density increases with increasing FC from 10 to 30% and
then it decreases up to FC = 40%.
Almost the same values of dry density for mixtures with 10 and 20% of FC were
obtained from the method of Yamamuro and Lade [3] in the small mold and graduated
cylinder. For the mixtures with more FC, values calculated from depositing material in
graduated cylinder were larger than the values from the small mold.
Comparison of the gradation curves after performing the ASTM D4253 [1] test on
the soil revealed changes in particle sizes that could be due to particle abrasion or
crushing. Whereas, gradation curves after standard compaction in small mold of the
soil showed almost no change in gradation. Gradation curves before and after
performing ASTM D4253[1] test on mixture with fine content of 20% is presented in
figure 3. A parameter is defined in the literature to quantify the relative breakage of soil
particles by dividing the area between gradation curves by the breakage potential which
is the area above the original gradation curve [17]. The relative breakage was equal to
0.101, 0.092, 0.114 and 0 for the mixtures of 10%, 20%, 30% and 40% of fines,
respectively.
2.1
Maximum Dry Density (g/cm3)
2.0
1.9
1.8
1.7 ASTM D4253

Yamamuro and Lade's Method
1.6 Yamamuro and Lade's Method in Small Mold
Standard Compaction in Small Mold

1.5
10 20 30 40
Fine Content (%)
Figure 2. Variation of the maximum dry densities with fine contents
100
90
80
70
Percent Finer
60
50
40
30
20 After Test
10 Before Test
0
0.01 0.1 1
Grain Size D (mm)
Figure 3. Particle degradation of mixture with 20% of fines after performing the ASTM D4253[1]
Particle segregation was visually observed during the ASTM D4253 [1] procedure;
however no particle segregation was observed in the second (standard compaction in
small mold) and the third (depositing soil in 16 layers in the graduated cylinder)
methods, i.e., a uniform distribution of soil particle sizes was observed. The application
of Yamamuro and Lade method [3] in small mold resulted in a decrease of maximum
dry density and the values were smaller for the mixtures with FC = 30 and 40%. It may
be concluded that the deposition of dry sand with fine content in layers in a smaller
mold would cause more segregation.
The literature shows a decrease in the minimum void ratios by increasing FC up to
20 to 30% and then an increasing trend for higher amounts of FC [7] - [10]. The trend
of maximum dry density obtained from the standard compaction test in the smaller
mold is consistent with these observations.
Depositing material in 16 layers in a graduated cylinder or in a small size mold and
the standard compaction method for soils with FC = 10% provided consistent results
without particle crushing. However, the maximum dry density attained from ASTM
D4253 [1] was high, but because of the alteration of particles this maximum density is
not reliable. For higher levels of FC (20, 30 and 40%) the method of ASTM D4253 [1]
and depositing material in a graduated cylinder and in a smaller mold resulted in lower
values of dry density due to particle segregation whereas the standard compaction
carried out in a smaller mold resulted in large values of density without altering soil
particle distributions.
5. Conclusions
Four methods for the determination of maximum dry density were evaluated for
crushable silty sands in this study.
It was shown that particle crushing during maximum density test can increase the
value of maximum density by 9% for silty sands with 10% fine content, and because of
this degradation, the original gradation of soil particles were not contributed to form
the packing with maximum density. Also particle segregation in mixtures with higher
values of fines content can decrease the value of maximum density by 14% in the
presence of 30% fines in the mixture. The case that is mostly presented in the literature
consists of poorly graded clean sand mixed with fines. Trend of maximum dry densities
changed when silty sands of natural gradation were tested.
The standard compaction test carried out in a smaller mold gave larger values of
dry density with very little particle crushing. For soils with 10% of fines content, all
methods except the ASTM D4253 [1] resulted in approximately the same values of dry
density. Therefore, for the mixture of 10%FC there was no need to use larger volumes
of soil to perform moist compaction test as dry material in small mold resulted in
trustable values for dry density.
Given the consistency of the soil gradation, the density values obtained from the
standard proctor compaction test in small mold are considered to be more
representative for reliable evaluation of minimum void ratio and thus the relative
density of the carbonate silty sands. It is suggested that the standard proctor
compaction test be used if enough soil is available and if not, performing standard
compaction in small mold can give reasonable results. For dry soil, the method of
Yamamuro and Lade [3] provides consistent results for 30 and 40% of fines and for
lower amount of fines, following the same method in a small mold results in reasonable
values.
References
[1] ASTM Standard D 4253-00, 2000 (2006). Standard Test Methods for Maximum Index Density and
Unit Weight of Soils Using a Vibratory Table, ASTM International, West Conshohocken, PA 2006,
DOI: 10.1520/D4253-00R06, www.astm.org.
[2] ASTM Standard D 698-12, 2012, Test Methods for Laboratory Compaction Characteristics of Soil
Using Standard Effort (12 400 ft-lbf/ft3 (600 kN-m/m3)), ASTM International, West Conshohocken,
PA 2012, DOI: 10.1520/D0698-12, www.astm.org.
[3] J.A. Yamamuro and P.V. Lade, Static liquefaction of very loose sands. Canadian Geotechnical
Journal, 34 (1997) 905917.
[4] J.C. Santamarina and G. C. Cho, Determination of critical state parameters in sandy soils-imple
procedure, Geotech. Test. J., 24 (2001), 185192.
[5] K. Terzaghi, R. Peck, and G. Mesri, Soil Mechanics in Engineering Practice, 3rd ed., John Wiley &
Sons Inc., NewYork, 1996.
[6] P.V. Lade, C. D. Jr. Liggio and J.A. Yamamuro, Effects of non-Plastic fines on minimum and
maximum void ratios of sand, Geotechnical Testing Journal, GTJODJ, 21 (1998), 336347.
[7] R. Kuerbis, D. Nagussey and Y.P. Vaid, Effect of gradation and fines content on the undrained response
of sand. Proc. Hyd. Fill Struc. Geotech. Spec. publ. 21, ASCE, New York, 1988, 330-345.
[8] P.V. Lade and J.A. Yamamuro, Effects of non-plastic fines on static liquefaction of sands. Canadian
Geotechnical Journal, 34 (1997), 918-928.
[9] A. M. A. Guerrero, Effects of the Soil Properties on the Maximum Dry Density Obtained from the
Standard Proctor Test, University of Central Florida, MSc Thesis, 2004.
[10] S. Thevanayagam, T. Shenthan, S. Mohan and J. Liang, Undrained fragility of clean Sands, silty sands,
and sandy silts, J. Geotech. Geoenviron. Eng. 128 (2002), 849-859.
[11] H. J. Fraser, Experimental study of the porosity and permeability of clastic sediments, Journal of
Geology, 43 (1935), 910-1010.
[12] M. R. Muszynski, Determination of maximum and minimum densities of poorly graded sands using a
simplified method, Geotechnical Testing Journal, 29 (2006).
[13] ASTM Standard D 2487-11, 2011, Standard Practice for Classification of Soils for Engineering
Purposes (Unified Soil Classification System). ASTM International, West Conshohocken, PA 2011,
DOI: 10.1520/D2487-11, www.astm.org.
[14] ASTM Standard D 854-10, 2010, Standard Test Methods for Specific Gravity of Soil Solids by Water
Pycnometer. ASTM International, West Conshohocken, PA 2010, DOI: 10.1520/D0854-10,
www.astm.org.
[15] ASTM Standard D 4318-10, 2010, Standard Test Methods for Liquid Limit, Plastic Limit, and
Plasticity Index of Soils, ASTM International, West Conshohocken, PA 2010, DOI: 10.1520/D4318-10,
www.astm.org.
[16] J. K. Mitchell, Fundamentals of Soil Behaviour, John Wiley & Sons Inc., NewYork, 1993.
[17] B. O. Hardin, Crushing of soil particles, Journal of Geotechnical Engineering, 111 (1985), 1177-1192.
IOS Press, 2015
Subject Index
3D imaging 431, 551 clay(s) 156, 358, 583, 761, 858,
FE 816 915, 1002, 1057
accelerometers 374 clayfills 323
accumulation of displacement 890 CNS tests 1105
acoustic emission 342 cohesionless soils 1089
acrylamide grout 955 collapse 462
adaptive substepping 784 combined loading conditions 1129
ageing 1010 compaction energy 40
air dissolution 687 compaction method 850
airport taxiway 1177 compressibility 1121
anisotropy 156, 315, 1161 compression behavior 583, 1057
application 192 compression wave 291
argillaceous rocks 342 consolidated-undrained triaxial
artificial soils 663, 1057 test 407
asymptotic states 695 consolidation 745, 882, 906, 947
back analysis 1161, 1169 consolidation settlement 1177
Bartlett test 559 constitutive model 679, 784, 1145
bearing capacity 1081 constitutive modeling 1105
bender element(s) 283, 374, 441, 470 contact model 671
bender element test 315, 398 contractancy 768
bitumen 986 controlled low strength material
bituminous mixtures 3 (CLSM) 559
boiled sand 800 coordination number 535
Box-Cox transformation 559 coupled behavior 1024
calcareous 1153 coupled hydromechanical
calibration chamber 866 behavior 687
Cam-Clay model 939 CPT 703, 1145
capillary rise 1113 crack propagation 79
carbonate sand 551 creep 486, 494, 503, 898, 915,
carbonate silty sands 1194 931, 947
cavity expansion 727 creep deformation 575
CBR 40 critical state 768, 1010
cell pressure control system 299 critical state soil mechanics 115, 939
cement 1002 cross-anisotropy 824
cement breakage 1065 crushing 1194
cementation 874, 1010 cubical cell apparatus 390
cementation degradation 978 cyclic behavior 898
cemented paste backfill 441 cyclic consolidation 267
cemented sand(s) 350, 978, 1169 cyclic loading 267, 866, 890
cemented soil(s) 711, 963 cyclic resistance 1010
centrifuge 331 cyclic softening 939
characterize 745 cyclic triaxial test 679
chemical grout 955 damping ratio 939
1202
de-bonding 1065 frequency domain analysis 398

debris flows 1081 friction 307
deflection 1169 friction angle 623
deformation 711, 824, 1018 frost heave 1024
deformation behavior 192 frozen soil 1024
deformation modulus 315 geo-changes 745
degree of compaction 40 geomaterials 3
degree of saturation 40 geosynthetic(s) 1032, 1137
density 832 geosynthetics-encased columns
digital image analysis 906 (GECs) 792
digital image correlation 600 geotechnical characterization 1153
dilatancy 591, 761, 768, 808 geotechnical modeling 511
dilatancy angle 623 gradation 1194
direct shear 494, 503 grain breakage 431
direct-shear test 858 grain contacts 719
Distinct Element Method (DEM) 79, grain crushing 1065
535, 655, 671, 695, 1129 grain-displacing hydrates 1018
disturbance effects 663 granular flow 647, 695
DMT 1145 granular material(s) 423, 494, 503,
double cell 299 591
downhole test 358 granular temperature 647
drained triaxial compression test 575 gravel 527
dry density 40 grouted sand 955
dynamic soil properties 398 Hall effect sensors 454
earthquakes 792 hardening-soil 737
effective friction angle 407 Henrys law 687
elastic limit strain 745 heterogeneity 543, 840
elastic shear modulus 470 high plasticity 1057
elastoplastic modeling 1024 high-speed microscope camera 567
element testing 382 hollow cylinder apparatus 462
element tests 323 hollow cylindrical torsional shear 519
engineered fill 527 horizontal deformation 906
environmentally friendly 511 hot asphalt mixtures 994
erodimeter 1089 hydrates 1018
excess pore-water pressure 792 hydraulic conductivity 40, 1002
excitation frequency 1186 hydro-mechanical behaviour 222
explicit integration 784 hydro-mechanical modelling 639
fabric 423 image 816
fabric/microstructure of soil 816 image analysis 423, 519, 1097
FEM 737 image processing 431
fiber optic 608 image segmentation 840
fiber optics sensing 382 in situ 1145
fiber reinforcement 350 in situ and laboratory test 231
field tests 608 in situ stress 1161
fine-grained soil(s) 407, 1049 in situ tests 727
finite element analysis 671 inclinometer 1169
finite element modelling 816 inherent anisotropy 840, 850
fissured clay behaviour 615 inhomogeneity 323
fractals 1113 injection pressure 955
1203
instability line 655 non-isothermal 784

interface shear 1129 non-linearity 3
intermediate soils 824 nonconvex surface 784
intrinsic parameters 583 numerical analysis(es) 1040, 1145
isotache 931 numerical modelling 703, 719
K0 275 OCR 358, 931
Kaolin-Silt clay 331 oedometer 535
keyword 454 oedometric test 431
keyword triaxial tests 454 one-dimensional compression 415,
kinematic behavior 478 486
kinematics 600 one-dimensional consolidation 931
laboratory 390, 1073 one-dimensional consolidation
laboratory test(s) 3, 291, 366, 415, test 1057, 1097
1049, 1113 oriented structure 315
laboratory testing 115, 447, 1040 over-consolidated clay 898
large size disk transducer 366 packings 768
linearity 3 partially drained conditions 703
liquefaction 519, 679, 792, 800, 832 particle breakage 79, 551
liquefaction resistance 299 particle image velocimetry
load tests 382 (PIV) 623, 647, 808
local deformation 519 particle shape 591
local stresses 866 particle-size distribution 591
localization 808 pavement structures 1137
loess 631, 850, 874 peat 543
long-term cyclic loading 890 peat behavior 947
low-strain shear modulus 331 penetration and centrifuge tests 1081
Marshall Mix Design Method 994 percolation suction 687
material damping ratio 527 permeability 882
maximum density 1194 phase transformation 687
measurement 192 piezoelectric transducer 331
measuring small displacements 454 pile foundation 583
mechanical characterization 382 pile installation 600
mechanical properties 994 pile-soil interface 866
membrane filter technique 299 plasticity 761
membrane penetration 275 PLAXIS 737
micro-mechanics 535, 567 Poissons ratio 291
microbially induced calcite pore pressure 939
precipitation (MICP) 283, 970, pore-water pressure 407
978 poromechanics 727
microstructure 423, 551, 970 porosity 882
model test 800 porous rock 1065
modeling 1145 prefabricated vertical drains 1177
monotonic 1010 preloading 1177
MPM 703 pressure melting 1024
multiphase formulation 703 pyroclastic sand 1121
multiple liquefactions 307 random factor analysis of
natural heterogeneity and shells 1097 variance (ANOVA) 559
near field effect 1186 rate-effects 695, 858
non-destructive methods 231 re-liquefaction 307
1204
reconstituted soil 840 small strain nonlinearity 608

reinforced soil 1032 small strain stiffness 291, 366, 737,
remolding energy 776 890
residual soil(s) 415, 447, 663 small-strain shear modulus 283
residual strength 858 smectite 1057
resonant-column 374, 527 soft clay 486, 1097
retaining wall 623, 808 soft grouting 1040
REV 615 soft soils 963
reverse extrusion test 407 soil behavior 1129
rheological model 947 soil compaction 40
rheological modelling 3 soil improvement 970
rockfill 79 soil modulus 631
rocks in suspension 1081 soil-structure interface testing 1105
rotational hardening 761 stacked-ring 307
rough pores 1113 static fatigue 719
sampling 663 stationarity 559
sampling quality 231 stiff clay 1161
San Francisco Bay mud 1177 stiffness 156, 824, 970, 1002, 1145
sand(s) 275, 315, 431, 478, 816, 840, stiffness degradation 939, 1145
866, 1010, 1073, 1153 stiffness profile 382
sand aging 719 strain 623, 631
sand particles 567 strain-dependent 527
sand with fibers 753 strain localization 832
sandy soils 832 strain measurements 382
scale effect(s) 543, 615 strain path control 275
sedimentary soft rock 575 strength 970
seismic isolation 1040 stress 631
sensitive clays 776 stress (strain) distribution 323
sensitivity 583 stress history 931
sensitivity framework 615 stress path tests 156
settlement 1032 stress relaxation 494, 503
shales 222, 882 stress state 832
shallow slope instability 1049 stress strain behavior 963
shape 551 structure 115, 447
shear 478 Structured Cam-Clay model 711
shear band 808 subloading surface elastoplasticity 784
shear displacement behaviour 1105 subsidence due to oxidation 947
shear modulus 315, 374, 753 suction effect 79
shear strength 156, 315, 350, 407, suffusion 1089
768, 840, 874, 955, 978, 1121 sulfate attack 1002
shear wave 291, 366 surcharge 931
shear wave velocity 192, 283, 331, swelling behavior 1057
358, 470, 753 synchrotron computed
simulation 503 tomography 478
site investigation 703 systems fundamental frequency 1186
size effect 79 theoretical analysis 1113
slip line theory 1081 three-component model 486, 503
small deformation stiffness 441 thyxotropy 874
small strain modulus 382 time dependency 79, 915
1205
tomography 840 uniaxial compression test 567

torsional shear 527, 832 unload/reload cycles 486
traffic loading 737 unsaturated rocks 342, 1113
transfer function 1186 unsaturated soil(s) 299, 462, 639,
transient flow 727 679, 890, 1137
transitional soils 986 vertical deformation 470
transparent soil 511 vertical drain 906
travel time 470 viscous behaviour 3
triaxial compression test 315 viscous property(ies) 486, 575
triaxial test(s) 267, 275, 283, 350, volcanic ash 639
358, 431, 543, 986, 1073 volcanic tuff 1121
true triaxial test 850 volume change 350
tunnelling 1161 water content 40
tyres 1073 water retention properties 222
ultimate bearing capacity 1032 water seepage power 1089
unbound granular materials 3 wave propagation 390
unconfined compression test 575 wave velocity 1186
undrained 267 wetting pore-collapse 639
undrained cyclic loading test 299 x-ray computed
undrained failure 1049 micro-tomography (XCT) 431
undrained instability 655 x-ray CT 423, 800
undrained shear strength 358, 832, x-ray tomography 600
1081 yield surface 761
IOS Press, 2015
Author Index
Acharya, R. 559 Camacho-Tauta, J. 398, 737
Adami, A. 874 Camenen, J.F. 390
Agaiby, S.S. 358 Capdevila, J.A. 631
Ahmad, S. 331 Crdenas, O.E. 462
Ahn, T. 1018 Carraro, J.A.H. 1097
Aiassa Martinez, G. 1032 Cascante, G. 398
Airey, D. 283, 986 Casini, F. 1065
Al-Raoush, R.I. 478 Cavaretta, I. 390
Ali, H. 398 Ceccato, F. 703
Alkayyal, H. 792 ernkov, M. 1161
Alnuaim, A.M. 331 Chaves-Agudelo, J. 939
Alonso, E.E. 79 Chekired, M. 591
Alshibli, K. 478 Chiaro, G. 832
Altunbas, A. 808 Cho, H.-I. 192
Alvarado, G. 470 Chow, J.K. 511
Amirpour Harehdasht, S. 591 Chung, C.-K. 906
Amundsen, H.A. 923 Cinicioglu, O. 623, 808
And, E. 423, 431 Clari, J.J. v, 753
Anh, N.N. 800 Combe, G. 600
Aoyagi, Y. 307 Consoli, N.C. 441
Arantes, C.Q. 994 Coop, M. 115, 156, 567
Arrua, P. 1032 Coop, M.R. 415, 447
Awwad, T. 792 Correia, A.A.S. 963
Ayala, J. 470 Cotecchia, F. 615
Baltodano Goulding, R. 1137 Coutinho, R.Q. 454
Barreto, D. 655 Covassi, P.A. 850, 1121
Bathe, A. 1177 Cox, C. 1145
Belokas, G. 761 Dafalias, Y.F. 761
Bendahmane, F. 1089 Dano, C. 898
Bernucci, L.L.B. 994 De Ponte Fernandes, D. 1153
Beuth, L. 703 Degago, S.A. 776, 915
Bezuijen, A. 1002 Di Benedetto, H. 3
Bheemasetti, T.V. 559 Di Emidio, G. 1002
Bilotta, E. 1040 Diaz Pardave, M. 1186
Boh, J. 1161 Dienstmann, G. 727
Boukpeti, N. 1097 Druckrey, A.M. 478
Boumezerane, D. 947 Duraisamy, Y. 283
Bowman, E.T. 647 Duttine, A. 494, 503
Branque, D. 687 Eberhardt, M. 1032
Brosse, A. 156 Eichenberger, J. 639
Bryson, L.S. 955, 1177 El Naggar, M.H. 331, 1194
Caballero, S.R. 559 Eliadorani, A.A. 275
Caicedo, B. 342, 1081, 1113 Emdal, A. 923
1208
Espitia, J. 342 Kim, J.-Y. 906

Ezaoui, A. 494, 503 Kim, S. 1018
Fabbri, A. 687 Kiyota, T. 832
Fauzi, U.J. 519 Klar, A. 231, 382
Favero, V. 222, 882 Koseki, J. 299, 307, 519, 575, 832
Feng, K. 978 Krupika, M. 1161
Ferrari, A. 222, 639, 882 Kuwano, J. 267, 800
Ferreira, C. 231, 374 Kuwano, R. 291, 366, 800
Festugato, L. 441 Kwa, K.A. 986
Flora, A. 1040 Lai, B.T. 687
Flrez Valencia, L. 840 Laloui, L. 222, 639, 882, 970
Fonseca, J. 551, 816 Lee, H. 1018
Fourie, A. 441 Lee, J.Y. 1018
Fratalocchi, E. 1002 Leguizamn Barreto, L.C. 840
Friedli, B. 608 Leiva Padilla, P. 1137
Frost, J.D. 1129 Lemos, L. 663
Gadegaard, S.S. 858 Lemos, L.J.L. 963
Ghoreishian Amiri, S.A. 1024 Levenberg, E. 382
Gollin, D. 647 Li, C. 639
Gomes Correia, A. 374 Li, X. 671
Gorosito, S.I. 1121 Li, Z. 671
Grimstad, G. 915, 947, 1024 Liao, T. 527
Guevara C., M. 583 Lim, G.T. 1097
Hamidi, A. 350 Lings, M. 390
Hamlin, S. 390 Lirer, S. 1040
Han, J. 898 Liu, M.D. 711
Herle, I. 323, 423 Lloret, A. 462
Hernndez-Rodrguez, F. 939 Lobedan, F. 1177
Hicher, P.-Y. 898 Lodahl, M.R. 1057
Horiuchi, Y. 800 Lombardi, D. 1040
Horpibulsuk, S. 711 Maghous, S. 727
Hoshino, R. 519 Marot, D. 1089
Hosseini Kamal, R. 156 Martinez, A. 1129
Hussien, M.N. 591 Martnez, D.A. 824
Ibraim, E. 390, 1073 Man, D. 695, 1161
Jardine, R. 156 Massoudi, N. 527
Jardine, R.J. 866 Matsumaru, T. 679
Jarrar, M. 478 Matsumoto, M. 575
Jerman, J. 695 Mayne, P. 1145
Jung, M. 527 Mayne, P.W. 358
Kadivar, M. 1024 McCabe, W.M. 1177
Kafle, B. 890 McHood, M. 527
Karatza, Z. 431 Mehli, M. 915
Karray, M. 591 Michalowski, R. 719
Kawabe, S. 486 Miyashita, Y. 519, 575
Kayabali, K. 407 Mohyla, T. 1161
Kazan, Y. 1169 Molina, F. 737
Kiguchi, M. 679 Mollamahmutoglu, M. 407
Kim, D.-S. 192 Montoya, B. 978
1209
Mortara, G. 1105 Santos Jr., O.F. 454

Muir Wood, D. 390 Sato, T. 299
Mukabi, J.N. 745 Schnaid, F. 727
Nadimi, S. 816 Scipioni, C. 1002
Naghavi Alhosseini, N. 1194 Severo, R.N.F. 454
Namikawa, K. 575 Shepley, P. 647
Nappa, V. 1040 Shi, X.S. 323
Nash, D.F.T. 931 Sibille, L. 1089
Nawir, H. 519 Silva, M. 600, 866
Nielsen, J.K. 858 Sim, W.W. 535
Nikzad, N. 350 Simonini, P. 703
Nishimura, S. 824, 1049 Soares, M. 231
Nishimura, T. 679 Soltanbeigi, B. 808
OSullivan, C. 535 Sorensen, K.K. 858, 1057
Ooi, G.L. 511 Stokoe, K.H. 527
Ooi, J.Y. 431 Stutz, H. 1105
Ortiz, R. 955 Suriol, J. 462
Otsubo, M. 535 Suwal, L.P. 291, 366
Ovando-Shelley, E. 1186 Tachibana, S. 267
Panta, A. 1049 Takada, S. 315
Papanicolopulos, S.-A. 431 Takeshi, S. 307
Park, H.-J. 192 Tan Tian, J. 299
Pedarla, A. 559 Tang, S. 1113
Pereira, C. 374 Tapias, M. 79
Pineda, J.A. 1097 Tatsuoka, F. 40, 486, 494, 503
Pokhrel, A. 366 Taylor, N. 816
Prada-Sarmiento, L.F. 655, 840 Terzis, D. 970
Puppala, A.J. 559 Thakur, V. 776, 923
Puzrin, A.M. 608 Tika, T. 1010
Ramos Can, A. 840 Tique Alviar, D.O. 840
Ramos-Can, A.M. 655 Todisco, M.C. 415
Rasool, A.M. 267 Toyota, H. 315
Reyes-Aldasoro, C.C. 551 Tsuha, C. de H.C. 866
Reyes-Ortiz, O. 737 Tudisco, E. 1065
Rezende, L.R. 994 Uchida, S. 382
Rimoy, S. 866 Useche Infante, D. 1032
Rinaldi, V.A. v, 631, 850, 874 Uzuoka, R. 679
Rocchi, I. 415, 447 Vaid, Y. 275
Rochim, A. 1089 Vallejo, L.E. 1081, 1113
Rodrigues, C. 663 Van, M.A. 543
Romero, E. 462 Venda Oliveira, P.J. 963
Rott, J. 1161 Verastegui Flores, R.D. 1002
Roubtsova, V. 591 Vettorelo, P. 753
Rouhanifar, S. 1073 Viana da Fonseca, A. 231, 398
Sadrekarimi, A. 1194 Vieira Faria, N. 1153
Salotti, A. 494, 503 Viggiani, G. 423, 431
Sancak, E. 623 Viggiani, G.M.B. 1065
Santos, Jaime 374 Villalobos, F. 470
Santos, Joo 374 Vitone, C. 615
1210
Vranna, A.D. 1010 Wong, H. 687

Wahyudi, S. 307 Wuttke, F. 890, 1105
Wang, H. 299 Yang, Z. 866
Wang, W. 567 Yanqui, C. 768
Wang, Y.-H. 511, 671 Yilmaz, N.P. 407
Wang, Z. 719 Yoo, H. 1081
Weber, R.C. 462 Yuan, Y. 931
Whittle, A.J. 931 Zeballos, M.E. v, 1121
Wiebicke, M. 423 Zhang, Y. 784
Wils, L. 551 Zhou, A. 784
Won, S. 1177 Zwanenburg, C. 543

6th International Symposium On Deformation Characteristics of Geomaterials Argetina 2015

Uploaded by

Copyright:

Available Formats

6th International Symposium On Deformation Characteristics of Geomaterials Argetina 2015

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

6th International Symposium On Deformation Characteristics of Geomaterials Argetina 2015

Uploaded by

Copyright:

Available Formats

DEFORMATION CHARACTERISTICS

ISSN 2212-781X (print)

ISBN 978-1-61499-600-2 (print)

Distributor in the USA and Canada

PRINTED IN THE NETHERLANDS

We would like to express our gratitude to the International Society of Soil

Organizing Committee TC101

International Advisory Committee

David Airey (Australia)

Herv DiBenedetto (France)

Peer-Reviewers of Buenos Aires Conferences 2015

Abdelkareem Alzo'ubi Celso Romanel

Gabriele Della Vecchia Juan Jos Clari

Maria Teresa Zandarin Paulo Albuquerque

Tatsuya Ishikawa Werner Bilfinger

Bishop Lecture and Plenary Lecture

Rockfill Mechanics. Experimental Observations and DEM Modelling 79

Behavior of Compacted Unsaturated Soil in Isotropic Compression, Cyclic

Influence of Initial Stress/Strain State on the Coefficient of Earth Pressure

Frequency Domain Method in Bender Element Testing Experimental

Environmentally Friendly Transparent Soil 511

Behavioural Features of Fissured Clays: Experimental Evidence and Modelling 615

Simulation of Traffic Loading on an Embankment by the Finite Element

Advanced Laboratory Investigation of Axial Cyclic Loading in Silica Sands 866

Capillary Rise in Corners of Rough Pores: Laboratory and Theoretical

Subject Index 1201

Advanced Testing and modelling of

-La troisime Confrence Bishop-

Keywords. unbound granular materials, geomaterials, bituminous mixtures,

2. General framework for rheological modelling: 3 component model and

d M {h, dir (d )} d (2)

represented by elastic bodies, i.e. spring of stiffness Ei in the 1-dimensional case or 3-

A specific procedure was developed to quantify the asymptotic hypoelastic tensor

3. Behaviour in the small strain domain

3.1. Unbound Granular Materials (UGM)

3.1.1. Experimental devices: T4C StaDy & Triaxial StaDy

Scheme of "T4CStady" apparatus System of strains measurement

Piston axis Suspension of ring (x3)

// "high" and "low"

Internal barrels Aluminium paper target

Targets for vertical

Ring hanging by 3 strips

Support of radial strain Pair (x3) of sensors

measurement sensor ("inner" and "outer")

measuring radial strain

Targets for shear

Moreover, both devices are equipped with piezoelectric sensors (compression

Back analysis of wave travel times provides dynamic elastic parameters.

3.1.2. Some noticeable results obtained with the 2 devices

Figure 5. Illustration of Positive and Negative (P & N) viscosity (from [43])

3.1.2.2. Hypoelastic tensor

Symmetry of the hypoelastic tensor M0

Mrreq Mreq Mrzeq Mreq 

DBGS and DBGSP formulation

Simple Pluviation method

Anisotropy of hypoelastic tensor

3.1.3. Example of site back calculation

M15 mixture (4 tests)

200 Hostun (m=0.50) (29 tests) z

d M {h, dir (d )} d (2)

Mrreq Mreq Mrzeq Mreq

200 Hostun (m=0.50) (29 tests) z

M15 mixture (m=0.74) (4 tests) z 10-5

ax (t) = Aax sin(t + E )

rad II (t) = - Arad II sin(t + II )

rad III (t) = - Arad III sin(t + III ) (11)

* Arad III j III * j III

s 1st CH surface fd(1) s sinv

(a) first loading to s : EH (b) unloading from sinv