Nautica Sudeste 8
Nautica Sudeste 8
Nautica Sudeste 8
ISSN : 2326-6872
eISSN: 2326-778X
Volume 5
Yossef H. Hatzor
Ben-Gurion University of the Negev, Beer-Sheva, Israel
Guowei Ma
The University of Western Australia, Crawley,WA, Australia
Gen-hua Shi
DDA Company, Belmont, CA, USA
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business
© 2018 Taylor & Francis Group, London, UK
Typeset by MPS Limited, Chennai, India
Printed and Bound by CPI Group (UK) Ltd, Croydon, CR0 4YY
All rights reserved. No part of this publication or the information contained
herein may be reproduced, stored in a retrieval system, or transmitted in any
form or by any means, electronic, mechanical, by photocopying, recording or
otherwise, without written prior permission from the publishers.
Although all care is taken to ensure integrity and the quality of this publication
and the information herein, no responsibility is assumed by the publishers nor
the author for any damage to the property or persons as a result of operation
or use of this publication and/or the information contained herein.
Library of Congress Cataloging-in-Publication Data
Applied for
Published by: CRC Press/Balkema
Schipholweg 107C, 2316 XC Leiden,The Netherlands
e-mail: [email protected]
www.crcpress.com – www.taylorandfrancis.com
ISBN: 978-1-138-02768-8 (Hbk)
ISBN: 978-1-315-68703-2 (eBook)
Neither the authors nor the publisher shall, in any event, be liable for any damages or
expenses, including consequential damages and expenses, resulting from the use of the
information, methods, or products described in this book or related downloadable
materials. Judgments made regarding the suitability of the techniques, procedures, meth-
ods, equations, etc. for any particular application are the responsibility of the user, and the
user alone.
Dedications
Yossef dedicates this book to his beloved wife Orna who left behind a promising career
in the arts in Israel and followed him to Berkeley for his PhD studies. Her support
during those difficult times and ever since has made this work possible and has made
him a much better person.
Guowei wishes to express his deep love and appreciation to his wife and two kids for
their patience, love, and support in every aspect in life. They are the greatest motivation
for his career pursuit including the works in this book.
Table of contents
1 Introduction 1
1.1 Who should read this book? 1
1.2 How to use this book? 1
1.3 Continuous vs. discontinuous deformation 2
1.4 DDA history 8
1.5 Three decades of DDA research and development 9
1.6 DDA vs. FEM and DEM 9
1.7 Main features of DDA 10
1.7.1 Block system kinematics 10
1.7.2 Complete first order displacement approximation 11
1.7.3 Equilibrium, dynamics and energy consumption 11
1.7.4 Large Deformation 11
1.7.5 Efficiency 12
1.8 Some limitations of the original DDA 12
1.9 Block discretization 15
1.10 Higher order displacement function 17
1.11 Coupling DDA with other numerical methods 19
1.11.1 Coupling DDA with FEM 19
1.11.2 Coupling DDA with BEM 21
1.11.3 Coupling DDA with NMM 22
1.12 Improved contact algorithms 23
1.13 Incorporation of viscous damping 26
1.14 Improved friction law for discontinuities 27
1.15 Gravity turn on and sequential excavation 29
1.16 Dynamic wave propagation and blasting 30
1.17 Masonry structures 33
1.18 Improved rockbolt element 33
1.19 Granular materials 34
1.19.1 Historical overview 34
1.19.2 Modelling particulate media with DDA 35
1.20 Pore pressure and fluid flow 37
1.20.1 Historical overview 37
1.20.2 Coupling DDA and fluid flow 39
1.21 Current development of 3-D DDA 40
viii Table of contents
Yossef H. Hatzor
Lemkin Professor of Rock Mechanics, Department of Geological and Environmental
Sciences, Ben-Gurion University of the Negev, P.O.B. 653, Beer-Sheva, 84105, Israel
Tel: +972-8-6472621
Fax: +972-8-6428717
E-mail: [email protected]
Guowei Ma
Professor, School of Civil, Environmental and Mining Engineering
Faculty of Engineering, Computing and Mathematics
The University of Western Australia
35 Stirling Highway, Crawley WA6009, Australia
Tel: 61-8-6488-3102
Fax: 61-8-6488-1018
Email: [email protected]
Gen-hua Shi
Consulting Engineer in Rock Engineering and Structural Engineering
President, DDA Company. 1746 Terrace Drive, Belmont CA 94002 USA
Professor, University of Chinese Academy of Sciences, Beijing, China
Tel: 650-631-1804; Cell: 650-867-6248, Email: [email protected]
Many people should be thanked for the creation of this book. First and foremost,
we would like to thank Professor Xia-Ting Feng, the Director of the National Key
Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and
Soil Mechanics, Chinese Academy of Sciences, Wuhan, China. Prof. Feng suggested
writing this book within the context of the ISRM Book Series he has instituted while
he was serving as President of the International Society for Rock Mechanics, as the
main product of the ISRM DDA Commission, which the three authors jointly chair.
His encouragement and support along the way has assisted us in achieving this goal.
Yossef Hatzor would like to express his sincere gratitude to the Chinese Academy
of Sciences for the Visiting Professorship grant for Senior International Scientists
(No.2011T2G29) awarded to him in late 2011. The frequent visits to the Institute
in Wuhan, twice a year for the duration of the fellowship period, have provided a
perfect setting for the quiet time and concentration required for the completion of this
book. Prof. Feng is deeply thanked for being such a gracious host during that time.
The assistance of colleagues from the lab in Wuhan is greatly appreciated, particularly
Dr. Yan Guo and Mr. Hang Ruan who provided all the technical and administra-
tive assistance willingly and flawlessly. Y. Hatzor also wishes to thank his PhD and
Post Doc students who have done a lot of the actual research, the results of which
are summarized in chapters 4–7 of this book: Drs. Michael Tsesarsky, Ronnie Kamai,
Dagan Bakun-Mazor, Gony Yagoda-Biran, Huirong Bao and Benguo He. Dr. He is
also thanked for assisting in proof reading and for compiling the subject index. Finally,
Alistair Bright and the editorial team at Taylor and Francis are thanked for their
excellent work during the production process of this book.
Guowei Ma wishes to thank Dr. Gen-hua Shi for leading him into the DDA field
and mentoring his team in DDA and NMM research. Guowei has been enjoying
the experience in initiating and co-chairing the ISRM DDA Commission with Yossef.
Special thanks are dedicated to Prof. Jian Zhao at Monash University, Australia and
Dr. Yingxin Zhou at the Defense Science and Technology Agency, Singapore. With
their support, Guowei had the honor to chair the Underground Technology and Rock
Engineering (UTRE) research program and the Jurong Rock Cavern (JRC) research
project at Nanyang Technological University (NTU), which greatly substantiated and
enhanced the DDA research at NTU. His appreciation and gratitude also goes to his
team members in DDA research at both Nanyang Technological University, Singapore
and the University of Western Australia, Australia including Drs. Xinmei An, Lei He,
Youjun Ning, Huihua Zhang, Lifeng Fan, Guoyang Fu, Xiaolei Qu, and Feng Ren.
xviii Acknowledgments
Gen-hua Shi thanks Professor Zuyu Chen from the China Institute of Water
Resources and Hydropower and Professor Ying Wang from the University of Chinese
Academy of Sciences for their guidance and support during the last several years of
research on his new contact theory. Dr. Shi also thanks Drs. Xu Li, Yuxing Ben, Wei
Wu, Xinchao Lin, Yanqiang Wu, and PhD candidates Mengsu Hu, Xiaolong Cheng,
Yunfan Xiao and Fei Zheng for their proofreading and figure presentations. Dr. Shi
is also thankful for many helpful discussions and comments given by the participants
attending the workshop on 3D contact theory, which was held in Dalian China from
March 28 to April 4, 2016 and was supported by China Basic Research Program Grant
No. 2014CB047100.
Finally, we all thank our mentor, Professor Richard Goodman, who has given us the
motivation, curiosity and means to advance, albeit slightly, the state of the art in the
mechanics of discontinuous rock masses.
Foreword
Introduction
We write this book with several sorts of readership in mind, all of whom are assumed
to have at least undergraduate level training in Geomechanics. First and foremost we
think of the committed Geological Engineering practitioners that need to solve a rock
mechanics problem in a real rock mass. The emphasis here is on the word “real’’.
Although, as you will see when you become familiar with this book, discontinuous
deformation analysis (DDA) is highly technical and rests on a solid theoretical founda-
tion, the rationale behind it comes from deep recognition of the importance structural
geology has in determining the layout and means for solution of complex, yet real,
rock engineering problems. Next, we have in mind the diligent code developer, who
may not go to the field so often but is blessed with a strong mathematical background
and computational skills which enable him or her to improve the state of the art in
numerical modelling as applied to rock mechanics. The rock mechanics community
gains from such people if they set their mind to work on rock engineering problems
when they could just as well use their skills to solve other, more general, engineering
problems. Finally, we have in mind the research community, graduate students and
faculty members alike, in both civil engineering and the earth sciences. On one hand,
DDA is a vibrant field of study with many challenging problems still unresolved. On
the other DDA has matured to the level where it can be used as a powerful tool for
studying earth science and civil engineering problems that require a discrete element
approach. Therefore the research community can benefit from this book by becom-
ing more familiar with this novel numerical method, identifying limitations, and by
employing the method solve complex research problems in a rigorous yet efficient
manner.
and 3 to better understand the main developments of the method during the past two
decades as reviewed in Chapter 1.
Readers who are familiar with DDA but are trained more in numerical modelling
with less fieldwork experience, are encouraged to study Chapter 4 that provides guide-
lines for how input parameters should be obtained from the field in order to conduct
a meaningful DDA simulation that will provide realistic results. Benchmark tests for
DDA are reviewed in Chapter 5. The set of benchmark tests will be useful for code
developers so that they can test the accuracy and performance of new enhancements
proposed. In addition, users of other numerical discrete element methods may find this
chapter useful for comparison between approaches. Chapters 6 and 7 review useful
case studies where DDA has been applied to solve problems in underground and slope
engineering, respectively. It will be useful particularly to practitioners as it will delin-
eate to some extent the scope of the method and the sort of applications that may be
attempted with it. Finally in Chapter 8 we provide the complete new contact theory
developed by DDA author Dr. Gen-hua Shi, for 2D and 3D DDA, including proofs of
all mathematical postulations and illustrated examples. This last chapter defines the
state of the art currently, and it remains for future generations to implement Shi’s new
contact theory in improved numerical codes.
Figure 1.1 Photomicrograph of a thin section (27 × 46 mm) of a pyroxenite rock under crossed
polarizers (XPL) from Valle Sesia of the Italian Alps. The pyroxenite is composed mostly
of clinopyroxene and orthopyroxene with minor amounts of olivine, plagioclase and
phlogopite. It is composed of equal sized crystals with straight grain boundaries inter-
secting at 120◦ , forming a mosaic equigranular texture (photo by Bar Elisha, caption by
Prof. Yaron Katzir, Ben-Gurion University of the Negev).
Figure 1.2 Photomicrograph of a thin section of a porphyritic rhyolite dyke from the Timna igneous
complex, southern Israel.The photograph was taken under a petrographic microscope using
plane polarized light. The porphyritic rock includes phenocrysts of quartz (clear white)
and alkali feldspar (cloudy, grey-brown) set in a fine grained matrix. The rock is fractured;
fractures highlighted by copper mineralization including green malachite overgrowing earlier
copper sulfide (opaque). Photo by Bar Elisha, caption by Prof. Yaron Katzir, Ben-Gurion
University of the Negev.
4 Discontinuous deformation analysis in rock mechanics practice
Figure 1.3 Hand specimen of a fine-grained olivine basalt from northern Israel with voids. Core
diameter 54 mm.
Figure 1.4 Intersection of three joint sets giving rise to a discontinuous rock mass structure as
exposed in a rock slope of the Yalong river, Sichuan province, China. Horizontal dimension
of figure approximately 10 m.
Introduction 5
Figure 1.5 The sub-vertical El Capitan cliff at Yosemite national Park, California.
system will be controlled by the mean spacing of the discontinuities in each set, and
the shear strength of the block system will be controlled by the shear strength char-
acteristics of the surfaces of the individual sets of discontinuities. Since on average
the shear strength of discontinuities is much smaller than the shear strength of intact
rock material, most of the deformation will take place by sliding along pre-existing
planes of discontinuities rather than by shearing through the intact rock material itself.
Therefore, the total height a natural rock slope can withstand is controlled much more
by the orientation and shear strength of the discontinuities than the strength of the
intact rock material. This is typical for rocks; in soils slopes typically fail by shear
through the continuous material which generally cannot support discontinuities in the
first place due to its negligible tensile strength, and this is where rock mechanics and
soil mechanics approaches to solve slope stability problems drastically differ.
A case in point is the Snake path cliff at Masada national park in Israel (Fig. 1.6(A))
where the slope height is “only’’ 250 meters above ground surface but the material
6 Discontinuous deformation analysis in rock mechanics practice
Figure 1.6 A) The Snake path cliff at Masada national park, Israel, B) A characteristic bedding
plane in Masada dolomites, C) A rock block that rests on a bedding plane in the Snake
path cliff at Masada.
strength of the dolomite is greater than 300 MPa. Clearly, with such a high uniaxial
compressive strength the rock slope would have been safe against failure in compression
at the base even if it extended several kilometres high. But more importantly, failure
by shear through the intact rock material is also very unlikely: a series of triaxial tests
performed on solid cylinders from this dolomite (Hatzor and Palchik, 1997) provide
a cohesion of 30 MPa and an internal friction angle of 45 degrees.
When testing a bedding plane interface in this material (Fig. 1.6(B)) in triaxial shear
however, the obtained cohesion is zero and the residual friction angle only 23 degrees
(Hatzor, 2003). This means that the block shown in the right corner of Fig. 1.6(C)
which rests on a bedding plane that is dipping 19 degrees, is in fact in danger of sliding
with a very low static factor of safety, even though the intact rock material is completely
safe from failure, either by compression or shear.
The same rationale is also applicable for the design of underground excavations
in rocks. When designing caverns for storage of hazardous materials for example the
level of discontinuity of the rock mass will actually determine the safe dimensions of
the chambers beyond which active reinforcement will have to be applied. This can be
appreciated from a comparison between two rock masses in which historic caverns
were excavated in the past, one in a weak but relatively continuous rock mass and
the other in a stronger but highly discontinuous rock mass. Consider the bell shaped
Introduction 7
Figure 1.7 Control of discontinuities on stability of underground openings in rocks. A) The 1000-year-
old bell shaped caverns at Bet Guvrin, Israel, B) The 3000-year-old underground water
storage system at Tel Beer Sehva, Israel.
cavern shown in Fig. 1.7A that has been excavated at Bet Guvrin national park in a
soft but relatively continuous chalk. Although the compressive strength of the intact
rock material is very low, between 3–9 MPa, (Hatzor et al., 2002) the cavern has been
standing unsupported with a free span of 24 meters since its excavation approximately
8 Discontinuous deformation analysis in rock mechanics practice
1000 years ago. This is in stark contrast to the underground water storage system that
was discovered at Tel Beer Sheva national park that was excavated approximately
3000 years ago in a chalk with unconfined compressive strength of 30 MPa. Even
though the strength of the material is much higher, the underground chambers span
cannot exceed 7 meters, and when the ancient engineers attempted to increase the span
they have experienced immediate collapse (Hatzor and Benary, 1998). This is because
the rock mass is transected by a very dense network of discontinuities with average
spacing of 10–20 cm only (Fig. 1.7B) giving rise to a rock mass structure that consists
of many small blocks.
The numerical discontinuous deformation analysis (DDA) method has been devel-
oped for enabling robust numerical modelling of deformation in discontinuous media
with geometric and mechanical characteristics of natural rock masses specifically
in mind.
Since the development of the joint element by Goodman, Taylor, and Brekke
(Goodman et al., 1968) to enable discontinuous deformation within the framework of
the finite element method (FEM), numerical computations with geological discontinu-
ities developed rapidly and have been applied to rock engineering extensively. Shortly
after, Cundall (1971) introduced the distinct element method (DEM), a force method
employing fictitious forces to regulate sliding, to prevent block overlapping, and to
reach equilibrium. Indeed, Cundall’s DEM is now widely used in rock mechanics prac-
tice. A decade later Shi and Goodman presented a novel method for computing the
strains and displacements of a block system (Shi and Goodman, 1984, 1985) that best
explains a set of displacement and strain observations made at a sufficient number of
points. They called this approach discontinuous deformation analysis (DDA). Know-
ing only the measured displacements, at individual points or directions, the method
computes the best least square fit over all displacements, deformations, and strains of
each block as well as the opening, closing and sliding of all block interfaces. The error
minimization is constrained by a robust kinematic analysis and for the first time a
correction procedure that prevents block penetrations is introduced. Since in the orig-
inal approach the displacements were observed and the block strains and movements
were inferred, constitutive equations were not needed and thus were not introduced
in the original DDA. This backward modelling approach has later been expanded by
Shi (1988) to accommodate constitutive relations and loading for the modelled rock
mass, thereby permitting general problems in the mechanics of block systems to be
solved in either a forward or a backward mode. Given the geometry, loading, and the
material constants of each block as well as the friction angle, cohesion and damping
mechanism at the contacts between blocks, the forward DDA computes the blocks
stresses, strains, sliding forces, contact forces, and movements.
DDA uses the displacements as unknowns and solves the equilibrium equations
in the same manner as the matrix analysis of structures in the finite element method.
Although it is intended primarily for discontinuous block systems, DDA is based on
strict adherence to the rules of classical mechanics. The development of DDA required
Introduction 9
the introduction of a complete kinematic theory that enables one to obtain large defor-
mation solutions for numerous blocks, without penetration. DDA’s kinematic theory
recognizes the connections of joints around the perimeters of blocks so that an optimal
correction-logic can be applied.
Although it is theoretically possible to incorporate large numbers of blocks in a
finite element mesh by introducing large numbers of joint elements or slip lines, in real-
ity such an approach will inevitably be ill-conditioned. This is because the corrections
required to satisfy the kinematical constraints of non-closing on one joint may be in
conflict with corrections appropriate to another joint, and so on. A robust kinematic
theory required to establish the optimum correction steps was not included, in fact, in
any other previously proposed approach for modelling discontinuous deformation.
Since DDA was published in the 1980s its applicability to rock mechanics practice has
been studied extensively by the international civil engineering research community and
the results have been published in peer reviewed journal papers as well as in proceed-
ings of a series of international symposia known as ICADD (International Conference
on Analysis of Discontinuous Deformation). The first ICADD meeting was held in
Taiwan in 1995 (ICADD1, 1995) followed by a DDA “Forum’’ that was held only
once at Berkeley (Salami and Banks, 1996). The second ICADD meeting was held in
Tokyo (ICADD2, 1997), then in Vail (ICADD3, 1999), Glasgow (ICADD4, 2001),
Wuhan (ICADD5, 2002), Trondheim (ICADD6, 2003), Honolulu (ICADD7, 2005),
Beijing (ICADD8, 2007), Singapore (ICADD9, 2009), Honolulu (ICADD10, 2011),
Fukuoka (ICADD11, 2013), and Wuhan (ICADD12, 2015). In these series of proceed-
ings improvements, verifications, and applications of the method have been published
but not all of these studies were published as journal papers and therefore their results
are less accessible. Below we summarize some important DDA R&D results that were
published over the past three decades in the series of ICADD proceedings.
Although it may seem DDA’s forward analysis of discontinuous deformations for block
systems resembles the distinct element method, in fact DDA is more closely related to
the finite element method (FEM), due to the following shared attributes between the
two methods:
For these reasons, programming forward and backward DDA models will be
natural for programmers familiar with the FEM.
There are some notable advantages of DDA over FEM when considering block
systems. Firstly, DDA does not assume continuity at block boundaries, namely it is
fundamentally a discontinuous approach. Secondly, the blocks forming the “elements’’
of the “mesh’’ can be convex or non-convex with any number of edges, and can even
contain holes. Moreover, block meshes do not require the vertex of one block to be in
contact with the vertex of another block. Thirdly, modelling only few block elements
with FEM would require a much larger mesh than is required in DDA.
DDA is different by nature from the widely used distinct element method
(Cundall, 1971). DDA is a displacement method, where the unknowns in the equi-
librium equations are displacements, whereas DEM is a force method which attempts
to adjust the contact forces to be constants using damping. DDA is an implicit method,
characterized by a rigorous energy consumption approach whereas DEM is an explicit
method.
The reliability of the method is attributed to the fact that mathematical and numerical
analyses in DDA attempt to address the mechanical phenomena associated with block
movements as close as possible to the physical reality. Some of the main features in
DDA are briefly discussed below following Shi (1988).
u = a1 x + a2 y + a3 ; v = b1 x + b2 y + b3
the whole block structure and affect the modes of failure. This process, which is char-
acteristic of discontinuous deformation, affects the accuracy of the solution which is
very sensitive to it, much more than when solving a continuum mechanics problem.
1.7.5 Efficiency
Discontinuous deformation analysis uses a selective equation system to decide on the
opening, sliding and closing of the contacts between blocks. On average, in each time
step five time selections are required, therefore the computation time is about five times
greater than that of finite element computation with the same number of unknowns.
In the special case where the number of blocks equals the number of elements, the
discontinuous deformation method has about four times the degrees of freedom of
that of the finite element method for a continuum. These efficiency considerations
with respect to computing time and memory requirements apply to both statics and
dynamics of block systems.
In a recent review paper Lin (2013) discusses some key DDA issues that call for further
attention and some of his comments are briefly reviewed here. First, since the original
DDA was developed to solve an inverse problem, this determined the choice of the
governing parameters. In each discrete block, which is considered a constant strain
element, there are six unknowns: the x and y displacements, the rotation, the x and y
strains, and the shear strain. Although this form is convenient when some of the strains
are measurable in an inverse problem setting and where the underlying kinematics
of the block system must be determined, it does have some restrictions for forward
modelling mainly because the unknowns are not independent. Lin (2013) proposes a
better choice for the original displacement field (Equation 1.1) that will also retain
the number of unknowns in the system of equations at six as originally suggested by
Lin and Lee (1996):
This formulation would also have the advantage that finite rotations and finite
strains could easily be introduced.
Regarding rotation, as we shall see in Chapter 2, in the original DDA the first
order displacement (u, v) at any point (x, y) within a block is:
⎛ ⎞
u0
⎜ ⎟
⎜ v0 ⎟
⎜ ⎟
u 1 0 −(y − y0 ) (x − x0 ) 0 (y − y0 )/2 ⎜ ⎟
= ⎜ r0 ⎟ (1.2)
v 0 1 (x − x0 ) 0 (y − y0 ) (x − x0 )/2 ⎜ εx ⎟
⎜
⎟
⎜ ⎟
⎝ εy ⎠
rxy
Introduction 13
Y Y
a a d
b d
(X0,Y0) X (X0,Y0) X
c b c
(a) Before rotation (b) After rotation
Figure 1.8 Rigid block rotation by 60◦ with DDA first order approximation would result in expansion
of block area (Ohnishi et al., 1995).
where (u0 , v0 ) are the rigid displacements at the center point (x0 , y0 ) within the block;
r0 is the angle of rigid body rotation about the center of the block (x0 , y0 ); εx , εy , and
εxy are the average elastic strains of the block. Since Equation 1.2 is only the first order
approximation of displacement, the rigid rotation r0 is defined as:
1 ∂v ∂u
r0 = − (1.3)
2 ∂x ∂y
with x = x0 and y = y0 . But Equation 1.3 is accurate only for very small deformation
and therefore it cannot be expected to be valid for large rotations. As clearly demon-
strated by Ohnishi and co-workers (Ohnishi et al., 1995), suppose the displacement
of the block only involves rigid rotation, then Equation 1.3 becomes:
u = −(y − y0 )r0
(1.4)
v = (x − x0 )r0
Now consider the block in Fig. 1.8a. If it is rotated by 60 degrees counter clockwise,
according to Equation 1.4 the vertices will move to the new positions shown in Fig. 1.8b
so that the block area will actually increase. In fact, the correct displacement for this
case should be calculated as:
u = (x − x0 )(cos r0 − 1) − (y − y0 )sin r0
(1.5)
v = (x − x0 ) sin r0 + (y − y0 )(cos r0 − 1)
r tanq
r sinq
x,y
x0 ,y0 r cosq
Figure 1.9 The exact solution for the rotation follows the arc. In the original DDA formulation the
displacement is along the tangent with magnitude rθ which falls between r sin θ and r tan θ
(MacLaughlin and Sitar, 1996).
Ke (1996) has proposed an “ad hoc’’ solution to this problem, a post adjustment
procedure which allows keeping the system of linear equations as in the original DDA.
MacLaughlin and Sitar (1996) demonstrated the discrepancy between the exact and
original DDA for rigid rotation (Fig. 1.9). They proposed a more accurate displace-
ment function using the Taylor’s series polynomial approximations for sin θ and cos θ
instead of the trigonometric functions, and developed a solution which is internally
consistent with the potential energy terms. One trivial solution to the rotation problem
would simply be to shorten the time step to ensure the rotations in each time step are
sufficiently small. Alternatively, the order of the polynomial can be increased (e.g. Koo
and Chern, 1996; Ma et al., 1996).
Another difficulty with the original DDA formulation is the implementation of
stiff springs once contacts have been detected automatically. The high stiffness evoked
by the penalty parameter means that the time step size must be reduced to avoid
uncontrolled block bouncing. The advantage of the penalty method, however, is that
it is straightforward and easy to be implemented. The augmented Lagrange multiplier
method represents a good compromise (e.g. Amadei et al., 1996) as it does not require
an increase in the size of the system of equations in the original DDA and yet the
simplicity of the penalty method is retained. This would allow also introduction of soft
springs at the contacts which is particularly advantageous in dynamic simulations.
Introduction 15
Indeed, the way damping is modelled in DDA is an issue that also requires some
attention. In the original DDA damping may be modelled as “kinetic’’. DDA resets the
velocity to zero at the beginning of a time step in a “static’’ analysis and completely
transfers it to the next time step in a “dynamic’’ analysis. Any ratio between zero
and 1.0 of this velocity transfer will represent a measure of “kinetic damping’’ that is
imposed by the user in a dynamic simulation. The physical meaning of this procedure
is ambiguous however. If accurate dynamic displacement in the time domain is sought
it would inevitably lead to erroneous results when less than 100% of the velocity is
transferred to the next time step.
In the discrete element method three damping schemes have typically been
employed: 1) element level damping, 2) global damping, and 3) contact damping.
For instance, in FLAC (ITASCA, 1993) a local adaptive damping has been designed
to converge its solution scheme, but for the implicit DDA the forces should all be
in equilibrium at the end of each time step, and therefore such a scheme would not
apply. A viscous element damping can readily be introduced; Ohnishi with co-workers
(Ohnishi et al., 2012) have shown how a viscous contact damping can be implemented
at contacts, and at the boundary, so as to obtain non-reflecting boundaries. Note that
the “kinetic damping’’ discussed here should not be confused with the algorithmic
damping associated with DDA’s time integration scheme which is discussed in some
detail by Wang et al. (1996) and Doolin and Sitar (2004).
In the original DDA blocks are assumed to be stiff, of infinite strength, and simply
deformable. While this assumption is valid for the range of stresses considered in most
civil engineering applications where intact rock strength is typically much higher than
the in situ stress level, these assumptions may cease to be valid when modelling rock
deformation in high in-situ stress environments or extreme load conditions.
To enhance stress distribution within DDA’s simply deformable blocks, Ke (1995)
proposed an artificial joint concept which he has further developed to enable DDA
modeling of fracture propagation in a column subjected to uniaxial tension (Ke, 1997)
and showed how both the strength of the solid elements as well as the texture of
the modeled mass affect the fracturing behavior. Ke’s artificial joint concept consists of
cutting blocks into sub-blocks and binding the sub-blocks within a block with “cement’’
to provide an incipient joint with tensile strength.
Amadei with coworkers (Amadei et al., 1996; Lin et al., 1995) proposed sub-
blocking for DDA which was further developed to enable block fracturing. Koo and
Chern (1997) presented a new approach for modeling progressive fracture in jointed
rock masses considering both tensile (mode 1) and shear (mode 2) failure modes. They
used the Coulomb–Mohr failure criterion with tension cut off to model strength of
intact rock elements. These initial developments lead to coupling between DDA and
FEM as discussed in a further sub section.
Ma et al. (2007) proposed a Moving Least Squares (MLS) approximation where
interpolative nodes are scattered within the blocks or along their borders, in a manner
similar to coupling DDA with Element Free Galerkin (EFG) method thus inheriting
all the advantages of Mesh Free Methods directly in DDA such as an ability to model
16 Discontinuous deformation analysis in rock mechanics practice
Figure 1.10 Illustration of a nodal-based DDA model (Bao and Zhao, 2010b).
crack propagation within blocks, blocks fracture, etc. Zhu et al. (2007) proposed
a sub-block meso-damage evolution model where each block is divided into smaller
sub-blocks and two kinds of contacts are implemented between the sub-blocks using
the Augmented Lagrangian Method, one continuous and the other discontinuous.
The continuity condition can change along the sub-block contacts and the continuous
contacts can convert into discontinuous ones using the Mohr-Coulomb criterion.
Grayeli and Mortazavi (2007) proposed an internal discretization scheme based
on the Delaunay mesh generation scheme and the DDA formulation was obtained
based on a FEM discretization. An elasto-plastic yield criterion was implemented in
the modified code and its performance was compared with analytical results and com-
mercial codes (FLAC, 1998). A block fragmentation model for DDA utilizing the FEM
adaptive mesh generation technique was proposed by Jiao and Zhang (2012). Ran-
domly distributed mechanical parameters statistically satisfying Weibull’s distribution
are assigned to the blocks to simulate the heterogeneity of rock mass and the artificial
joints provide the potential paths along which the cracks generate and propagate.
These initial improvements paved the way to a new branch of current DDA
research trying to incorporate fracture mechanics into the original DDA method so
that internal block failure processes can be modeled. Bao and Zhao (2010) adopted a
Nodal-based DDA method to conduct fracture analysis (Fig. 1.10), and this direction
is explored by other researchers as well (e.g. Ben et al., 2013; Jiao et al., 2014; Zhao
et al., 2013b).
The Mohr-Coulomb criterion is commonly used for block fracturing. There are
three parameters involved including the friction angle ϕ, cohesion c and tensile
strength T0 . The three-parameter Mohr-Coulomb criterion is expressed in terms of
major and minor principal stresses σ3 and σ1 (with σ3 ≥ σ1 ) as follows:
π ϕ
σ1 = −C0 + σ3 tan2 + (1.6)
2 2
σ3 = T0 (1.7)
Introduction 17
when σ1 ≥ σ1c . C0 is the unconfined compressive strength of the block material, which
is related to the cohesion c and friction angle ϕ as follows:
π ϕ
C0 = 2c tan + (1.8)
2 2
and σ1c is a critical transitional stress between the shear and tensile fracture modes:
π ϕ
σ1c = −C0 + T0 tan2 + (1.9)
2 2
To conclude this section we would like to point out that we believe the numerical
manifold method (Shi, 1995), the presentation of which is beyond the scope of this
book, provides the best platform to proceed in this research direction. Indeed, much
of the more recent developments in this field utilize the numerical manifold method.
The original DDA uses first order approximation of the displacement field in its gov-
erning equations and we have already shown above errors that may be encountered
particularly with respect to rigid block rotation. A second order displacement function
for DDA has been proposed very early on by Koo et al. (1995) who have provided a
very comprehensive solution and implemented it in a modified DDA code which was
verified using classical benchmark tests. They have extended the first order function
by adding the complete second order term:
u = a1 + a2 x + a3 y + a4 x2 + a5 xy + a6 y2
(1.10)
v = b1 + b2 x + b3 y + b4 x2 + b5 xy + b6 y2
where u0 and v0 are rigid body translation of the block, rc0 , εcx , εcy and rcxy are the
constant terms of the rotation, normal and shear strains, εx,x , εx,y , εy,x , εy,y , rxy,x , and
rxy,y are the gradients of the normal and shear strains. The displacement functions can
be written in matrix form as:
u
= T2×12 d12×1 (1.11)
v 2×1
T = (T1 T2 )
18 Discontinuous deformation analysis in rock mechanics practice
⎛ ⎞
1
1 0 −(y − y0 ) (x − x0 ) 0 (y − y0 )
⎜ 2 ⎟
T1 = ⎜
⎝
⎟
⎠
1
0 1 (x − x0 ) 0 (y − y0 ) (x − x0 )
2
⎛ ⎞
1 2 1 2 1 2
(x − x0 ) (xy − x0 y0 ) − (y − y0 )
2 2
0 0 (y − y0 )
2
⎜2 2 2 ⎟
T2 = ⎜
⎝
⎟
⎠
1 2 1 1
0 − (x − x20 ) (xy − x0 y0 ) − (y2 − y02 ) − (x2 − x20 ) 0
2 2 2
and d = (u0 v0 rc0 εcx εcy rcxy εx,x εx,y εy,x εy,y rxy,x rxy,y )T .
A second-order displacement function incorporating a six-node triangular mesh
was derived by Grayeli and Mortazavi (2005) to enhance DDA’s capabilities for use
in practical applications. The stress distribution around a tunnel obtained with the
modified DDA was compared with FEM and the analytical Kirsch solution (Kirsch,
1898) showing good agreement.
A third order displacement function for DDA was developed by Koo and Chern
(1996). The coefficients of the displacement functions were used as the displacement
variables. The complete third order displacement functions have the following form:
where T is
1 0 x 0 y 0 x2 0 xy 0 y2 0 x3 0 x2 y 0 xy2 0 y3 0
T=
0 1 0 x 0 y 0 x2 0 xy 0 y2 0 x3 0 x2 y 0 xy2 0 y3
and dT = dT1 dT2
dT1 = d1 d2 d3 d4 d5 d6 d7 d8 d9 d10
dT2 = d11 d12 d13 d14 d15 d16 d17 d18 d19 d20
A third order displacement function has also been introduced into the original
DDA (Huang et al., 2010) exhibiting very accurate bending deformation of a can-
tilever beam. A method for approximating the continuous displacement function with
complete high order polynomials has been proposed for DDA (Wu et al., 2012) and it
is also proved to be highly accurate when solving a single block cantilever problem.
Introduction 19
For high order displacement functions, the sub-matrices and equilibrium equations
can be obtained accordingly based on the minimization of the total potential energy
similar to the original DDA.
One issue that is important to realize with regard to the implementation of higher
order displacement functions in a discontinuous framework is that under large defor-
mations the updated shapes of the blocks will differ drastically from the original shapes
because of the large internal deformation the blocks will be permitted to undergo, and
therefore new contacts will form and new interactive forces will be introduced which
will change the whole structure and may affect the modes of failure more drastically
than in a continuum mechanics problem. The difficulties associated with implementa-
tion of higher order in DDA and how they may be addressed are discussed by Wang
et al. (2007).
and FEM had been achieved by Cao et al. (2007) who obtained the displacements
and strains by proper internal discretization of deformable blocks using finite element
meshes while the contacts between the blocks were modeled by DDA. Based on the
variational principle of minimum potential energy the simultaneous equilibrium equa-
tions of the coupled method were then established. They demonstrated the validity
and advantages of their enhancement in computing the deformation of a multilayer
asphalt concrete pavement subjected to moving vehicle loads.
In order to refine the stress distribution inside the blocks and improve the flexibility
of block boundaries, a nodal based DDA was developed by Shyu (1993) by introduc-
ing the finite element mesh into the DDA blocks. Both triangular and quadrilateral
elements were studied. For triangular elements, each element has three nodes (l, m, n)
and six displacement variables in two orthogonal directions (ul , vl , um , vm , un , vn ) as
shown in Fig. 1.11. The displacement field of a triangular element with constant strain
is given by:
⎛ ⎞⎛ ⎞
1 0 xl 0 yl 0 ul
⎜0 1 0 xl 0 y l ⎟ ⎜ vl ⎟
⎟ ⎜
⎜
⎜ ⎟⎜ ⎟
⎟
u 1 0 x 0 y 0 ⎜1 0 xm 0 ym 0 ⎟ ⎜ um ⎟
= ⎜ ⎟⎜ ⎟ (1.16)
v 0 1 0 x 0 y ⎜
⎜0 1 0 xm 0 ym ⎟ ⎜ ⎟
⎟ ⎜ vm ⎟
⎜ ⎟⎜ ⎟
⎝1 0 xn 0 yn 0 ⎠ ⎝ un ⎠
0 1 0 xn 0 yn vn
For quadrilateral elements, each element has four nodes (k, l, m, n) and eight dis-
placement variables in two orthogonal directions (uk , vk , ul , vl , um , vm , un , vn ) as shown
in Fig. 1.12.
Introduction 21
The displacement field of a quadrilateral element with constant strain is given by:
⎛ ⎞
n11 0 n12 0 n13 0 n14 0
⎜ 0 n11 0 n12 0 n13 0 n14 ⎟
⎜ ⎟
⎜ ⎟
⎜ n21 0 n22 0 n23 0 n24 0 ⎟
u ⎜
1 0 x 0 y 0 xy 0 ⎜ 0 n21 0 n22 0 n23 0 n24 ⎟ ⎟
=
v 0 1 0 x 0 y 0 xy ⎜ ⎜ n31 0 n32 0 n33 0 n34 0 ⎟
⎟
⎜ 0 n31 0 n32 0 n33 0 n34 ⎟
⎜ ⎟
⎝ n41 0 n42 0 n43 0 n44 0 ⎠
0 n41 0 n42 0 n43 0 n44
(1.17)
where
⎛ ⎞ ⎛ ⎞−1
n11 n12 n13 n14 1 xk yk xk yk
⎜n n22 n23 n24 ⎟ ⎜ xl yl ⎟
⎜ 21 ⎟ ⎜1 xl yl ⎟
N=⎜ ⎟=⎜ ⎟ (1.18)
⎝ n31 n32 n33 n34 ⎠ ⎝ 1 xm ym xm ym ⎠
n41 n42 n43 n44 1 xn yn xn yn
HU = GP (1.19)
22 Discontinuous deformation analysis in rock mechanics practice
where H and G are the coefficient matrices, U is the displacement vector and P is
the surface traction vector. In order to couple BEM with the DDA, Equation 1.19 is
pre-multiplied by LG−1 as follows:
LG−1 HU = LP (1.20)
where KBEM is the stiffness matrix from the BEM, KDDA is from the DDA, KBD is the
interaction between the DDA and the BEM, uBEM is the BEM interface displacement
vector, aDDA is the coefficient vector of DDA. Both FIBEM and FIDDA are modified by the
interaction terms and thus have the superscript I.
The open-close iteration based contact algorithm in the DDA has also been imple-
mented into the dual reciprocity boundary element method (DRBEM) by Fu (2014),
Fu, Ma and Qu (2015, 2016). The governing equation is written as follows:
c =0
HU − GP + M Ü + Mg − Ed σ 0 − GF (1.22)
pq F
Fp = G
P
(1.23)
The total potential energy sys of the blocky system, which includes DDA blocks
and NMM elements, can be expressed as:
sys = NMM
sys + DDA
sys + i,j (1.24)
B,i E,j
The last term in Equation 1.24 represents the potential energy for the contact
between DDA block i and NMM element j. The corresponding matrices and vectors
in the kinematic equations can also be derived by minimizing the potential energy.
DDA contact algorithm is based on a penalty method. In the penalty method, stiff
springs are set in normal and/or shear directions between blocks to transfer the
inequality problem of contact constraint into equality problem of computing con-
tact displacements and contact forces. The penalty method proved to be effective in
many numerical areas and has been widely applied. It’s relatively simple implementa-
tion and it does not increase the dimension of the system of equilibrium equations. The
main shortcoming of the penalty method is that it can only fulfil the contact constraint
approximately, and the contact treatment precision is affected by selected penalty num-
ber. With a reasonably large penalty number the penalty method can treat the contact
of blocks well, however if the penalty number is too large the system of equilibrium
equations may become ill-conditioned resulting in un-acceptable errors. There is no
simple way to guess a priory the best value of the penalty method in the original DDA
and this does present an obstacle when high accuracy is sought. To amend this problem
many researchers have proposed different remedies since the original DDA has been
proposed.
Amadei et al. (1996) modified the original DDA contact using the Augmented
Lagrangian Method (ALM) so as to retain the simplicity of the penalty method and
yet to minimize the disadvantages of the penalty method and the classical Lagrange
Multiplier Method, the implementation of which would require an increase in the sys-
tem of the governing equations. The essential concept behind the ALM is to use both a
penalty number, p representing the stiffness of the contact spring, and a Lagrange mul-
tiplier λ∗ representing the contact force λ, for each block to block contact to iteratively
calculate the contact force. An iterative method is used to calculate the Lagrange mul-
tiplier until the distance, d, of penetration of one block into the other and the residual
force between block contacts are both below minimum specified tolerances. The final
exact contact forces can always be obtained by the iterative method even with a small
initial value of the penalty parameter, although if the penalty parameter is too small
many iterations may be required and consequently some efficiency may be lost.
To better understand the ALM modification to the original DDA contact algorithm
consider Fig. 1.13. In the original two dimensional DDA all contacts between blocks
can be transformed to contacts between an angle and an edge. P1 in Fig. 1.13 is a vertex
of block i, P2 P3 is an edge of block j. After a step displacement P1 moves to P0 . Assume
(xk , yk ) and (uk , vk ) are the coordinates and displacements of Pk (k = 0∼3) respectively.
24 Discontinuous deformation analysis in rock mechanics practice
S0 S0
dn = + EDi + GDj , ds = + E Di + G D j (1.25)
l l
1
er = [(y2 − y3 )t1r (x0 , y0 ) + (x3 − x2 )t2r (x0 , y0 )]
l
1
gr = [(y3 − y0 )t1r (x2 , y2 ) + (x0 − x3 )t2r (x2 , y2 )]
l
1
+ [(y0 − y2 )t1r (x3 , y3 ) + (x2 − x0 )t2r (x3 , y3 )]
l
1
er = [(y2 − y3 )t1r (x1 , y1 ) + (x3 − x2 )t2r (x1 , y1 )]
l
1
gr = [(x2 − x3 )t1r (x0 , y0 ) + (y2 − y3 )t2r (x0 , y0 )]
l
in which t1r (x, y) and t2r (x, y) are displacement functions of block i and j respectively.
Introduction 25
In the penalty method, two springs with stiffness of pn and ps are used in the
normal and shear directions to constrain the contact displacements to zero. Then the
strain energy of normal and shear springs can be respectively expressed as:
1 1
n = pn dn2 , s = ps ds2 (1.26)
2 2
where λ∗m and λ∗m+1 are the Lagrangian multipliers of time step m and m + 1, respec-
tively, and d is the contact displacement. Then at the mth time step the contact strain
energy can be expressed as:
1 2
s = λ + p = λ∗m d + pd (1.28)
2
where p = 12 pd 2 is the same as the spring strain energy in the penalty method, so only
λ = λ∗m d needs to be taken into account additionally. λ in the normal and shear
directions can be deduced as:
∗ ∗ S0
λ = λm dn = λm + EDi + GDj
l
(1.29)
S0
λ = λ∗
m sd = λ∗
m + E
D i + G
D j
l
In the open-close iteration of the DDA method, each contact pattern converges at
time step n when the contact pattern at iteration i is the same as the one at previous
iteration i − 1. When the number of iteration equals to 6 and the contact pattern does
not converge, the time step size will be reduced and a new iteration cycle will start.
At iteration 1 of the new cycle, the previous pattern is defined as that at iteration 5 of
the previous cycle, which may be different from the contact pattern at the end of step
n − 1. This will cause significant and unexpected change in time step size. Wu and Lin
(2013) improved the open-close iteration by reloading the contact information at the
end of previous step n − 1 as the previous iteration when the iteration is 1 in the new
cycle. The computational results indicate that the new DDA program results in correct
stress evaluation.
A new contact theory was proposed by Shi (2013), in which entrance blocks are
used for contacts between two dimensional and three dimensional blocks. The bound-
ary of an entrance block is the entrance surface, which is a contact cover system.
Each contact cover could be contact vectors, contact edges, contact angles or contact
polygons. Each contact cover defines a contact point and all closed contact points
together define the movements, rotations and deformations of all blocks. Given a ref-
erence point, the concept of entrance blocks simplifies the contact computations. The
essentials of Shi’s new contact theory are provided in Chapter 8.
In the original DDA code written by Shi viscous damping was ignored, although the
corresponding viscous damping matrix does appear in the simultaneous equations of
equilibrium. The reason for not implementing viscous damping in the original DDA
code was related to uncertainties regarding the viscous damping coefficient that should
be used. Therefore, in the original DDA code no damping other than “kinetic damp-
ing’’ exists. The lack of viscous damping however precluded accurate modeling of
dynamic impact problems such as rock falls, but DDA is well suited for modeling
such problems and therefore several research groups have attempted very early on to
incorporate viscous damping in the DDA code. The critical point here is to understand
the energy dissipation mechanism during rock fall and then to simulate it correctly by
introducing viscous damping which has a different physical meaning than the friction
coefficient or coefficient of restitution. In particular the energy of the impact is clearly
expected to be influenced strongly by the viscosity coefficient. It is important to real-
ize, however, that there is no independent way to establish the viscosity coefficient and
therefore it must be tweaked based on field observations of the particular case at hand.
Shinji et al. (1997) inserted a viscosity coefficient into the original DDA equations as
an “equivalent friction coefficient’’ and, based on field observations, suggested some
empirical relationship between the viscosity coefficient and the slope conditions, pri-
marily talus and vegetation cover. This representation of viscosity however is clearly
a gross approximation. A formal implementation of viscous damping into DDA con-
tacts has been proposed by Sasaki et al. (2005) and was used to model rock fall
problems in dangerous slopes in Japan that are frequently subjected to earthquake exci-
tation. The case of the 1994 Niigata prefecture earthquake was used as a case study
where Voigt-type viscous damping elements were implemented between blocks and
Introduction 27
Figure 1.14 Boundary condition of earthquake response analysis (Left) andVoigt-type viscous damping
of friction (Right) implemented in DDA to model earthquake induced rock falls in Niigata
prefecture, Japan (Sasaki et al., 2005).
In the original DDA the “no tension – no penetration’’ criterion is used to impose
kinematical constraints on the deformation of the constant springs. If the normal
spring undergoes tension it is removed so that no tension can exist between initially
mating surfaces. Therefore, the normal spring is set only for compression between
blocks. Because zero penetration is not feasible when calculating the spring force, a
very small amount of penetration δm is allowed in practical computations.
The Coulomb–Mohr law is applied for modeling interface frictional resistance in
DDA. If the shear force calculated by the program is higher than the available shear
strength of the interface τm the tangential spring is removed and Coulomb frictional
resistance replaces the role of the tangential spring. The shear resistance τm in DDA is
given by:
τm = C + fn tan φ (1.30)
where C and φ are the cohesion and friction angle of the discontinuity, both of which
are assumed to be constants throughout the analysis, and fn is the compressive force
of the normal spring which is updated in every time step according to the dynamic
deformation of the modelled block system.
When studying dynamic deformation in geological materials which is typically
concentrated along pre-existing discontinuities, the frictional resistance can hardly
be assumed to be constant as assumed in the original DDA. This has already been
recognized by Goodman and Seed (1966) in their classic paper which provides the
analytical solution for dynamic block displacement on frictional interfaces due to
cyclic loading. In contrast to the so called “Newmark’s method’’ for slope stability
28 Discontinuous deformation analysis in rock mechanics practice
Figure 1.15 Friction degradation in concrete surfaces as a function of sliding velocity (Bakun-Mazor
et al., 2012). Open triangles are results from pseudo-static direct shear experiments, open
diamonds are back analyzed results from shaking table experiments.
(Newmark, 1965), Goodman and Sid, allowed in their solution incremental fric-
tional degradation between displacement cycles. Clearly if the friction coefficient is
maintained constant throughout the dynamic simulation the forward modeling results
may be un-conservative because cyclic motions damage the asperities and deterio-
rate the frictional resistance the initially rough rock discontinuity surface may offer.
Velocity weakening of concrete wedge surfaces during cyclic displacement has been
measured experimentally using shaking table experiments of dynamic sliding of tetra-
hedral wedges (Bakun-Mazor et al., 2012) and the results were compared to velocity
degradation obtained during slow direct shear tests (Fig. 1.15). Similar results have
been reported for rock discontinuities as well.
Velocity weakening is but part of a more complex constitutive model for rock
joint friction that takes into account both the rate as well as the state of sliding, largely
known today as “rate and state friction’’ (Dieterich, 1972). The original rate and
state friction law has been modified and improved over the years (e.g. Dieterich, 1978,
1979; Dieterich, 1981; Dieterich and Kilgore, 1994, 1996; Kilgore et al., 1993; Ruina,
1983) and has been used to model earthquake mechanisms (e.g. Scholz, 2002). One of
the more standard representations of the rate and state constitutive law is (Dieterich,
1979; Ruina, 1983):
∗
∗ V θV
µ = µ + A ln +1 +B +1 (1.31)
V∗ Dc
Introduction 29
dθ Vθ
=− (1.32)
dt Dc
When dynamic frictional sliding takes place at steady state conditions dθ/dt =
0 Vθ
Dc
= 1 and Equation 1.31 can be written as:
∗ Vss
µss = µ + (A − B) ln +1 (1.34)
V∗
DDA is a fully dynamic method that uses inertia forces to determine the motion of
block systems. The inertia forces are updated after each time step and are dependent
on the block displacement in the previous time step. Consequently, sudden loading at
the beginning of the analysis, whether by gravity or any other applied load, produces
unwanted artificial accelerations and displacements that propagate throughout the
time steps of the analysis. MacLaughlin and Sitar (1999) identified this problem and
were the first to propose a static “gravity turn-on’’ phase which precedes the dynamic
analysis, and added two extra iterations in each time step in order to guarantee accurate
30 Discontinuous deformation analysis in rock mechanics practice
determination of the contact forces between blocks. They demonstrated the enhanced
accuracy of this approach using the classic block on an inclined plane problem.
The gravity turn on concept has already been discussed in the context of the FEM
(Desai and Abel, 1972) and its significance in geotechnical engineering has been rec-
ognized by early development of analytical (e.g. Goodman and Brown, 1963) and
numerical (e.g. Dunlop et al., 1970) solutions for rock slope stability. MacLaughlin
and Sitar (1996) explain this issue using the case of an open pit mine: in reality the
removal of the overburden will result in upward displacement of the floor, however if
the post excavation geometry is modeled without regard to the initial stress conditions
the floor of the pit will displace downward as the material deforms elastically when
gravity load is introduced.
In deep underground excavations this is particularly important. When modeling a
high in situ stress environment it may take many time steps for the model to actually
equilibrate under the imposed in situ stresses and gravitational load, a process that
has been attributed to “seating’’ of the contacts (MacLaughlin and Sitar, 1999) which
actually is a correct way for describing this problem. Therefore, if the underground
opening is removed before the stresses stabilize everywhere in the block system the
results may be overly conservative because the real available frictional resistance of
the joints hasn’t been realized yet in the model as normal stresses have not reached the
correct value (see Fig. 1.16). To address this problem MacLaughlin and Sitar (1999)
proposed a preliminary pseudo-static phase during which the unwanted displacements
are dissipated by setting the artificial velocities at the beginning of every time step in
this stage to zero.
An alternative approach would be to actually model the excavation sequence where
first the imposed in situ stresses are allowed to fully develop everywhere in the mesh
(the number of time steps this stage will require increases with the number of blocks in
the mesh because indeed all contact springs need to be equilibrated under the loads),
and only then the excavation is removed and true displacements are realized.
Sequence excavation modeling in high in situ stress environments has been demon-
strated with the NMM (Tal et al., 2014) and DDA (Hatzor et al., 2015). In the sequence
excavation code developed by Tal et al. (2014) instead of using the additional iterations
suggested by MacLaughlin and Sitar (1999) a special provision is made in the code to
run the simulation in static conditions (dynamic parameter = 0) before the excavation
is removed and when the excavation material is removed to switch the analysis to
dynamic (dynamic parameter = 1). One added advantage of the preliminary stage in
sequence excavation modeling is that it allows tweaking the best value for the contact
springs; the penalty value at which the steady state stresses in that stage are as close as
possible to the imposed stresses (before the opening is introduced) is in all likelihood
the optimal contact spring value for the specific problem at hand.
DDA is a dynamic method and as such it is suitable for modeling dynamic problems
such as seismic site response or shock wave propagation through discontinuous media.
Several modifications have been proposed to the original DDA in order to enhance
modeling capabilities of dynamic wave propagation.
Introduction 31
Figure 1.16 The significance of modeling the excavation sequence. On the right panels (Original NMM)
the opening exists from the first time step whereas on the left panels (Modified NMM)
the opening is removed only after all stresses have stabilized (gravity turn on). Note that
with gravity turn on a friction angle of 15 degrees on all joints is sufficient for stability (Tal
et al., 2014).
Tsai and Wang (1995) modified DDA to include a sliding boundary that trans-
mits the frictional forces from the ground to the superstructure. Wang et al. (1995)
conducted a pioneering study on shock wave propagation through an elastic bar dis-
cussing the interrelationship between the contact spring stiffness and the time step and
proposing new criteria to select the best choice of time step and contact spring stiffness
for dynamic contact problems. The influence of various time integration scheme on
the accuracy of the dynamic solution is discussed by Wang et al. (1996). To improve
rock blasting simulation capability a plastic constitutive relation has been added to
the original 2D-DDA code and energy dissipation was taken into account by Yang
and Ning (2005). They argued that frictional sliding along pre-existing discontinuities
cannot be assumed to be the main energy consumption mechanism in blasting simula-
tions, as is assumed in the original DDA. They implemented a Drucker-Prager plastic
yield criterion for the solid block elements so that linear strain-hardening was added
to the original 2D-DDA code. While a block reaches plastic deformation according to
32 Discontinuous deformation analysis in rock mechanics practice
the Drucker-Prager yield criterion, the plastic constitutive relation is used. If a block
expands beyond a certain threshold that is determined by the plastic constitutive law
it is considered to be fractured and a crack is recorded. For this fractured block the
tensile strength is zeroed and the increase in strain is ignored. They used this approach
to predict the fragment size that will be generated and demonstrated their approach
for bench blasting (Yang and Ning, 2005) and demolition blasting of a brick wall
(Ning et al., 2007). Blasting in a borehole has been simulated by Zhao et al. (2007)
with provisions made in the original DDA code for fracture initiation and propagation
assuming a Weibull type tensile strength distribution in the joints. This approach has
been extended for modeling full face blasting during tunneling (Zhao et al., 2010).
To enable accurate modeling and fast computation of blasting phenomena with
DDA the boundaries of the modeled domain should be relatively close to the blasting
point. If however the boundaries of the modeled domain are positioned too close to the
basting point, inevitable reflections from the boundaries may obscure the computation.
To avoid this problem non-reflective boundaries must be introduced. Two types of non-
reflective boundaries for DDA are typically considered (Ning and Zhao, 2012): viscous
boundary condition (VBC) and superposition boundary condition (SBC).
Jiao et al. (2007) were the first to improve the existing fixed boundary condition
in the original DDA by introducing a viscous boundary condition into the original
DDA code. They adopted the VBC proposed by Lysmer and Kuhlemeyer (1969) which
is based on the use of independent dashpots in the normal and shear directions of
specific boundaries (Fig. 1.17). Their proposed VBC was verified using analytical
solution for P wave propagation through an elastic bar and attenuation of blasting
wave. Bao et al. (2012) introduced a new viscous boundary submatrix with high
absorbing efficiency which was developed specifically for DDA, based on the viscous
boundary condition originally introduced by Lysmer and Kuhlemeyer (1969). In their
derivation, the analytical velocity of a dashpot is employed instead of using the finite
Introduction 33
Figure 1.18 Observed finite displacement in an old masonry structure used to constrain paleo peak
ground acceleration.
difference method to obtain the velocity. Their VBC is verified using analytical solutions
for P wave propagation through an elastic bar and S wave propagation through a
stack of horizontal layers. A recent comprehensive review and original developments
of boundary settings for DDA is provided by Fu et al. (2015).
The robust simplex integration scheme in DDA and complete dynamic formulation
makes it a very attractive candidate for studying deformation of masonry structures.
This promising research direction has been pioneered by Professor Bicanic from the uni-
versity of Glasgow (Bicanic and Stirling, 2001; Bicanic et al., 2003) who has originally
used the semicircular masonry arch under self-weight as a benchmark test for DDA to
show how a departure from the classical interface conditions (no sliding, no crushing,
no tension) influences masonry arch stability. Observed finite block displacements in
historic masonry structures (Fig. 1.18) have been used to constrain paleo peak ground
acceleration (Kamai and Hatzor, 2008; Yagoda-Biran and Hatzor, 2010) by inserting
sinusoidal input accelerations to the modeled structure and finding and frequency and
acceleration at which the damage obtained numerically best fits the damage mapped
in the field.
DDA can also be utilized to consider alternative reinforcement schemes for ancient
masonry structures. This has been demonstrated in the case of Angkor, one of the
most significant monuments of the Khmer culture in Cambodia (Hayashi et al., 2012;
Koyama et al., 2013; Sasaki et al., 2011; Tian et al., 2012).
The original DDA includes rockbolt submatrices where the bolt is modeled as a fixed
line on two ends with a given stiffness and length. The block system is then allowed
34 Discontinuous deformation analysis in rock mechanics practice
to deform and the bolts to extend, but the original bolt elements are linear, of infi-
nite stiffness, and a perfect coupling between the bolt element and the rock mass
is assumed. The location of the end point of the bolt and its stiffness are defined
by the user. During each time step the resistance provided by the bolt stiffness is
incorporated into the global stiffness matrix by minimizing the potential energy. The
bolt stiffness is then mobilized with the occurrence of relative displacement of the
two end points. The applicability of this simple rockbolt element has been demon-
strated for both underground (Bin Shi et al., 2010; Hatzor et al., 2015; Yeung, 1991,
1993) and rock slope (Hatzor et al., 2004; Shi, 2012; Tsesarsky and Hatzor, 2009)
engineering.
The original DDA bolt is capable of simulating mechanically anchored bolt sys-
tems, the usage of which has become less popular over time. Clapp and MacLaughlin
(2003) further pointed out several deficiencies of the bolt element in the original DDA:
1 The original bolt element keeps deforming infinitely throughout the analysis as it
is assumed to have linear elastic stiffness and infinite strength.
2 No option for measurement points along the bolt trace are provided in the original
code to check the shear stress and displacement distribution along the bolt length.
3 The original bolt element does not interact realistically with the rock mass.
4 The single segment spring bolt allows forces to be distributed uniformly along the
length of the bolt instead of dispersing the load gradually away from the load
point.
A more representative bolting system which takes into consideration the bond
and its interaction with the bolt element on one side and with the rock mass on the
other is therefore required. Moosavi and Grayeli (2006) implemented bond strength
characteristics into the original DDA bolt element for simulation of the performance
of fully grouted cable bolts during deformation of a discontinuous rock mass where
most of the deformation is assumed to concentrate along the interface between the
rock and the cable bolt.
A unified rock bolt element for 2D-DDA has been developed by Professor Zhao
Zhiye and his research group in Singapore (Zhao et al., 2013a). Different types of
rockbolts are modeled in a unified framework and different failure modes are possible
in the proposed analysis procedure. The details of the coupling between the rock bolts
and rock blocks and its implementation in 2D-DDA are discussed by Nie et al. (2014).
on efforts to develop circular or elliptical disc elements for DDA. The first efforts in this
direction were presented by Ohnishi and his group (Ohnishi et al., 1995; Ohnishi and
Miki, 1996) who have also provided the mathematical derivation and implemented
this into the original DDA with the elliptical elements enabling internal deformation
of the initially circular discs. This research direction was followed by many workers in
the soil mechanics and geotechnical engineering fields (e.g. O’Sullivan and Bray, 2001;
Thomas et al., 1996).
The deterioration of railroad track material (ballast) was studied by applying cyclic
loading to a stack of coarse granular fragments and measuring the obtained plastic
deformation with DDA (Ishikawa et al., 1997). To better represent the high fric-
tion angle of such materials polygon elements generated by Voronoi tessellation were
employed and DDA results with polygon and circular disc elements were compared.
They concluded that the contact mechanism in the original DDA must be improved
in order to properly model coarse granular deformation. They have later extended
their analysis to study time dependent deformation of granular materials with DDA
(Ishikawa and Ohnish, 2001) and further incorporated also elliptic disc elements in
the analysis (Ohnishi et al., 2005).
DDA has been used to model ball milling in mining industry context with more
numerical efficient contact sorting and contact detection algorithms by simulating
rigid, non-fracturing, 2D ore particles and implementing viscos damping both in the
normal and shear directions (Balden et al., 2001).
Early in 2001, an algorithmic breakthrough was achieved allowing 3D tran-
sient dynamic modelling of particulate systems of real-shaped particles (Munjiza and
Latham, 2002). Whereas with spheres it is relatively simple to establish whether par-
ticles are in contact from the position of their centers and their radii, and to establish
forces and trajectory paths associated with collisions, considerable algorithmic sophis-
tication is required for collisions of rock fragment-shaped particles. Guo and Lin (2007)
used 2D DDA with polygonal particle shapes to study, numerically, the mechanical
behavior of coarse granular media and obtained surprisingly good agreement with
laboratory experiments.
Some researchers have tried extending DDA to problems which are typically con-
sidered as soil mechanics problems. For example, the effect of soil expansion in
response to changes in water on slope stability has been studied with DDA by intro-
ducing time step wise linear material property degradation in response to increasing
water content (Lin and Qiu, 2010).
After solving (u0 , v0 , γ0 ), Equation 1.37 may not compute (u, v) of a specified point
P(x, y) within disk i if γ0 is very small, as the radii is lengthened. To avoid such an
error, Equation 1.36 is used to calculate the components of (u, v) due to γ0 .
For elliptical disks (Ohnishi and Miki, 1996), assuming that (x0 , y0 ) is the centre
of gravity of an ellipse, an ellipse is represented as:
ab
x= cos s + x0
b2 cos2 s + a2 sin2 s
0 ≤ s ≤ 2π (1.38)
ab
y= sin s + y0
b2 cos2 s + a2 sin2 s
where a and b are the major and minor axes, respectively and s is an angular parameter.
When only translations (u0 , v0 ) are involved, the displacements (u, v) at any point in
the block are given by:
u 1 0 u0
= (1.39)
v 0 1 v0
Assuming that only rotation (r0 ) of the ellipse around (x0 , y0 ) is involved, the shape
of the ellipse can exactly be represented by:
ab
x= cos(s + r0 ) + x0
b2 cos2 s + a2 sin2 s
0 ≤ s ≤ 2π (1.40)
ab
y= sin(s + r0 ) + y0
b2 cos2 s + a2 sin2 s
When the rotation angle r0 is small, the displacements are simplified to the
following equation:
u −(y − y0 )
= γ0 (1.41)
v x − x0
Introduction 37
When only shear strain rxy is involved, the shape of an ellipse is represented by:
ab
rxy
x= cos s − + x0
b2 cos2 s + a2 sin2 s 2
(1.43)
ab rxy
y= sin s + + y0
b2 cos2 s + a2 sin2 s 2
Assuming small shear strain rxy , the displacements are simplified as follows:
⎛ ⎞
y − y0
u ⎜ ⎟
= ⎝ x −2 x ⎠ γxy (1.44)
v 0
2
Summing Equations 1.39, 1.41, 1.42, and 1.44, we obtain the displacements
for any point within the elliptical disk, which is identical to that of a polygon in the
original DDA:
⎛ ⎞
u0
⎜v ⎟
⎜
⎜ ⎟
0⎟
u 1 0 −(y − y0 ) (x − x0 ) 0 (y − y0 )/2 ⎜ r0 ⎟
= ⎜ ⎟ = Ti di (1.45)
⎜ ⎟
v 0 1 (x − x0 ) 0 (y − y0 ) (x − x0 )/2 ⎜ εx ⎟
⎜ ⎟
⎝ εy ⎠
rxy
The remaining formulation of particle-based DDA is the same as the original DDA.
method for locating the free surfaces in unconfined flow problems, which was then
coupled to the mechanical equations of motion for the coupled hydro-mechanical
analysis. A laboratory experiment of fluid flow in a 2-D rectangular network was
simulated to verify the residual flow algorithm in their modified DDA code, and an
acceptable agreement between the measured and calculated results was presented.
A coupled hydro-mechanical DDA model (HYDRO-DDA) has also been pro-
posed by Rouainia et al. (2001) where the discontinuous medium is represented with
DDA and is interfaced with a continuum formulation for flow through porous media.
The two frameworks communicate via mapping of an equivalent porosity field from the
solid to the fluid phase and an inverse mapping of the pressure field back to DDA. The
fluid system, by means of Dracy’s law using a FEM mesh, responds to pressure or flux
boundary conditions and to porosity changes caused by changes in the discontinuity
patterns. DDA is used to model the blocky discontinuous solid phase, responding to
force or displacement boundary conditions and to the state of effective stress. The
procedure operates by establishing an initial fluid pressure distribution in a static rock
framework and passing this information to DDA which produces a deformed solid
framework shape. This change in geometry is passed back to the hydro part of the
code for refinement of fluid pressure distribution and returned back to DDA until the
results converge according to some tolerance. The motivation behind their study came
from the petroleum engineering industry where the fracturing of a mudrock sealing cap
of an over pressured sandstone reservoir was of interest, but of course this approach
can be further utilized to model other petroleum engineering related issues such as, for
example, borehole breakouts and hydraulic fracturing. Indeed, a different fluid-solid
coupling scheme for DDA was proposed by Ben et al. (Ben et al., 2012) for single
phase compressible fluid, and it was later utilized for modeling hydraulic fracturing
with DDA (Ben et al., 2013).
Challenging geotechnical engineering problems, which would typically fall into
the realm of classical soil mechanics, have been attempted to be analyzed with DDA.
Oh et al. (2002) modeled wave induced sea bed response to the dynamic impact of
waves incorporating into DDA effective stresses and Biot’s equations, perhaps for the
first time, arguing that an effective, rather than total, stress approach is needed in such
cases. Moreover, the progressive soil deformation will become discontinuous over time
under these unique loading conditions and thus the justification for attempting a dis-
continuous approach, on contrast to methods that have been applied until then to
address these problems numerically. Koyama et al. (2012) modeled triggering large
landslides due to torrential rainfall by introducing pore pressures into the unsaturated
zone using FEM and modeling the dynamic deformation with DDA. The validity of
their numerical approach was tested using large scale physical experiments. Chen et al.
(2013) introduced seepage forces into DDA to analyze the stability of coastal breakwa-
ter structures against tsunami damage. Hydro-mechanical coupling with DDA has also
been applied to investigate the seepage rate into oil – storage caverns under different
in situ stress conditions after the excavation is formed (Zhao et al., 2013b).
A three-dimensional fluid-structure coupling between Smoothed Particles Hydro-
dynamics (SPH) and 3D-DDA for modelling rock-fluid interactions has recently been
presented by Mikola and Sitar (2013). The Navier-Stokes equation is simulated using
the SPH method and the motions of the blocks are tracked by a Lagrangian algorithm
based on a newly developed, explicit, 3D-DDA formulation. The coupled model is
Introduction 39
employed to investigate the water entry of a sliding block and the resulting wave(s)
providing a new computational tool for coastal and offshore engineering.
k
q=− (∇h + ρr ∇z) (1.46)
µr
where k is the hydraulic conductivity tensor and h is the hydraulic head. µr and ρr are
the relative viscosity and relative density of the fluid respectively and z is the elevation.
The finite element approximation (Smith and Griffiths, 1998) is used to solve the
hydraulic head, with suitable boundary conditions and fluid material properties.
The fluid pressure acting on the vertex of every DDA block is determined from the
fluid flow analysis. Where a solid block vertex lies within a finite element, the pressure
at the centroid of the element is calculated and applied as force components to the
vertex of the block. The force components on the vertices of a block are written as
follows:
L 1
F1 = p1m + (p2m − p1m )
2 3
(1.47)
L 2
F2 = p1m + (p2m − p1m )
2 3
where L = (x2 − x1 )2 + (y2 − y1 )2 and p1m and p2m are the pressure at the centroid
of the corresponding finite elements in the mesh where the two block vertices lie,
respectively. The horizontal and vertical components of the force components F1 and
F2 are determined and added to the external load and given as boundary conditions
in the DDA. This produces a changed element geometry according to block system
kinematics and consequently leads to a change in the fluid pressure distribution.
40 Discontinuous deformation analysis in rock mechanics practice
3-D DDA is being extensively studied currently, mainly its basic theory and contact
algorithm. Shi (2001a, b, c) provided basic formulations for matrices such as mass
matrix, stiffness matrix, point load matrix, body load matrix, initial stress matrix and
fixed point matrix for different potential terms. Liu et al. (2004) and Yeung et al.
(2003; 2004) highlighted the application of 3-D DDA for dealing with 3-D mechan-
ical interactions of blocks, using the ‘common-plane’ technique. Jiang and Yeung
(2004) developed a point-to-face model for 3-D DDA, but this model did not con-
sider the vertex-to-vertex, vertex-to-edge and edge-to-edge contact modes. Moosavi
et al. (2005) investigated dynamic 3-D DDA using analytical solutions, but they did
not investigate the effect of the numerical parameters on the results. Beyabanaki and
Jafari (2005) presented modified point-to-face frictionless contact constraints for 3-D
DDA. Wu et al. (2005) developed a new contact-searching algorithm for frictionless
vertex-to-face contact problems, and presented the 3-D DDA formulation for normal
contact force. Yeung et al. (2007) and Wu (2008) presented algorithms for edge-to-edge
contacts. Beyabanaki et al. (2008) presented a new algorithm to search and calculate
geometrical contacts in 3-D DDA. Beyabanaki et al. (2009a, b) implemented trilinear
and serendipity hexahedron FEM meshes into 3-D DDA. Beyabanaki et al. (2009b, c;
2010) further formulated 3-D DDA with high-order displacement functions.
In terms of geometric algorithms, a polyhedron is discretized into geometric ele-
ments including vertexes, edges and facets. The 3-D contact detection of polyhedral
blocks includes vertex-to-vertex, vertex-to-edge, vertex-to-face, edge-to-edge, edge-to-
face and face-to-face contacts (Ahn et al., 2011). Intuitively, all other contact patterns,
such as vertex-to-vertex, vertex-to-edge, face-to-face, edge-to-face and edge-to-edge,
respectively, can be converted into vertex-to-face contacts (Wu et al., 2005). Jiang
et al. (2004) proposed a vertex-to-face contact algorithm based on geometric analysis
for 3-D DDA. Keneti and Jafari (2008) developed a new contact detection algorithm
considering main planes and dominant contacts to identify contact points and types.
Beyabanaki and Mikola (2008) offered an approach to identify the contact pattern
between two blocks using a closest point searching algorithm as well as other 3-D
contact algorithms for improved efficiency. Wu et al. (2014) presented an efficient and
robust spatial contact detection algorithm using a novel multi-shell cover system and
decomposition of geometrical sub-units, which greatly reduced the contact detection
volume and iterations. Zhang et al. (2016) identified contact types between polyhe-
dral blocks using an extended hierarchy territory algorithm and a new loop search
procedure.
Jiao et al. (2015) presented a new 3D spherical DDA model for rock failure
(SDDARF3D) for simulating the whole process of rock failure. An integrated sys-
tem coupling 3D binocular photogrammetry and DDA for stability analysis of tunnels
in blocky rock mass was proposed by Zhu et al. (2015) recently.
These studies demonstrate the extensive research currently being pursued on con-
tact definition, identification and implementation in DDA. In Chapter 8, we present the
complete new contact theory developed by Dr. Gen-hua Shi which is general and there-
fore applicable to both two and three dimensional DDA. It remains for the research
community to find ways to implement these new algorithms in the future, in efficient
computational platforms.
Chapter 2
As stated in Chapter 1, the DDA method possesses unique features in terms of system
variables, single block matrix formulation, simplex integration, contact algorithm and
time integration, etc.
The traditional DDA method adopts three displacement variables (u0 , v0 , r0 ) and
three strain variables (εx , εy , rxy ) as unknowns for an individual block. The equilibrium
equations of the DDA method are derived from the minimization of the total potential
energy. The Coulomb’s law of friction controls the contact modes between blocks
and an iterative process called open-close iterations is enforced at each time step to
satisfy the no-tension and no-penetration criteria and ensure the contact accuracy. The
contacts between blocks are identified and continuously updated during the entire
analysis process.
δ(U + W) = δ( ) = 0 (2.1)
where U is the elastic potential energy of the whole system, W is the work done by the
external loads including these among the blocks, and is the total potential energy of
the whole system.
Based on the minimized potential energy, the system of equations can be
represented as follows:
Kd + Cḋ + M d̈ = F (2.2)
where K is the stiffness matrix, M is the mass matrix, C is the viscosity matrix, F is the
external force vector, d, ḋ and d̈ are the displacement, velocity and acceleration vector
respectively.
42 Discontinuous deformation analysis in rock mechanics practice
The sub-matrices from each block and the contact sub-matrices among the blocks
are assembled into the global matrices according to the degrees-of-freedom of each
block. The global equilibrium equations can be written in the following form,
⎡ ⎤⎧ ⎫ ⎧ ⎫
K11 K12 K13 · · · K1n ⎪ ⎪ d1 ⎪
⎪ ⎪
⎪ F1 ⎪⎪
⎢K21 ⎥ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎢ K22 K23 ⎪ d2 ⎪
· · · K2n ⎥ ⎪ ⎪ ⎪
⎪F2 ⎪ ⎪
⎪
⎢ ⎥ ⎨ ⎬ ⎨ ⎬
⎢K31 K32 K33 · · · K3n ⎥ d3 = F3 (2.4)
⎢ ⎥
⎢ .. .. .. .. ⎥ ⎪ ⎪ ⎪
⎪ .. ⎪ ⎪ .. ⎪
⎪
. ⎦⎪ .⎪ ⎪ .⎪
..
⎣ . . . . ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
⎩ ⎭ ⎩ ⎪
⎪ ⎪ ⎪ ⎭
Kn1 Kn2 Kn3 · · · Knn dn Fn
where sub-matrices Kii depend on the material properties of block i and j and sub-
matrices Kij are defined by the contacts between blocks i and j. di and Fi are 6 × 1
sub-matrices, and Fi is the loading distributed to the six unknown variables of block i.
The matrix Kij (i, j = 1, 2, . . ., n) is derived by the differentiation:
∂2
(r, s = 1, 2, . . ., 6) (2.5)
∂dri ∂dsj
The details of the derivation for the matrix Kij and force vector Fi can be referred
to in Sections 2.2 and 2.3.
u = a1 + a2 x + a3 y
(2.7)
v = b1 + b2 x + b3 y
where (u, v) are the displacements at point (x, y). At any point in the block (x0 , y0 ), the
displacements (u0 , v0 ) are determined as follows:
u0 = a1 + a2 x0 + a3 y0
(2.8)
v0 = b1 + b2 x0 + b3 y0
Theory of the discontinuous deformation analysis (DDA) 43
u = a2 (x − x0 ) + a3 (y − y0 ) + u0
(2.9)
v = b2 (x − x0 ) + b3 (y − y0 ) + v0
Substituting the parameter derived from Equation 2.10 into Equation 2.9 and
we get:
1
u = εx (x − x0 ) + γxy − r0 (y − y0 ) + u0
2
(2.11)
1
v= γxy + r0 (x − x0 ) + εy (y − y0 ) + v0
2
Written in a matrix form, the complete first order approximation of the displace-
ments of any point within the block can be given as follows:
⎛ ⎞
u0
⎜ v0 ⎟
% & ' (⎜ ⎟
u 1 0 −(y − y0 ) (x − x0 ) 0 (y − y0 )/2 ⎜ ⎟
⎜ r0 ⎟ = Ti di (2.12)
=
v 0 1 (x − x0 ) 0 (y − y0 ) (x − x0 )/2 ⎜ ε
⎜ ⎟ x
⎟
⎝ εy ⎠
rxy
where (x, y) are the coordinates of any point within the block, (x0 , y0 ) are the coordi-
nates of a point within the block (normally the centroid of the block is selected). u0
and v0 are the rigid body translations at the point (x0 , y0 ) along x and y directions
respectively, r0 is the rotation angle in radian around the point (x0 , y0 ), εx , εy and rxy
are the normal and shear strains of the block. di is the vector form of the unknowns
(u0 , v0 , r0 , εx , εy , rxy ) as shown in Fig. 2.1 for block i and Ti is the matrix form of the
first order displacement function.
Due to the adoption of the linear displacement function term, the angular rotation
will produce a strain term, which may be very large if the angular rotation within a
time step is large, leading to unexpected block expansion and to distortion of blocks.
A number of techniques have been proposed by researchers to solve this problem as
mentioned in Section 1.8.
44 Discontinuous deformation analysis in rock mechanics practice
Based on Equations 1.4 and 1.5 suggested by Ohnishi and co-workers (Ohnishi
et al., 1995), MacLaughlin and Sitar (1996) proposed to use the second order term to
account for the effects of angular rotation:
r02
cos r0 = 1 − (2.13)
2
r02
u = u0 − (x − x0 ) − (y − y0 )r0
2 (2.14)
r2
v = v0 − (x − x0 )r0 − (y − y0 ) 0
2
M
u = (x − x0 )(cos r0 − 1) − (y − y0 ) sin r0 + t1j dj (j = 3)
j=1
(2.15)
M
v = (x − x0 ) sin r0 + (y − y0 )(cos r0 − 1) + t2j dj (j = 3)
j=1
where M is the number of block unknowns, d i and tij is the elements of the displacement
function t i .
Theory of the discontinuous deformation analysis (DDA) 45
An iterative method was developed by Cheng and Zhang (2000) which can reduce
the block distortion to a minimum.
The exact solution for the x and y components (u, v) of the displacement of an
arbitrary point (x, y) within a deformable block are given by:
Consider that,
where:
sin r0
f1 = −(x − x0 ) − (y − y0 )
1 + cos r0
sin r0
f2 = −(y − y0 ) − (x − x0 )
1 + cos r0
In the treatment of the terms f1 and f2 in iteration i within each time step, the
value of r0 can be taken as the value of r0 in step i = 1. This process goes along with
the global iteration analysis within each time step.
The equilibrium equations of the DDA method are derived by minimizing the total
potential energy of the block system. For single blocks, the total potential energy may
include elastic strains, initial stress, point loading, line loading, body force, bolting
connection, inertia force, displacement constraints at a point or in a direction, etc.
where the integration is over the entire area of block i. For each displacement step, the
blocks are assumed to be linearly elastic.
The relationship between stress and strain is expressed as follows:
σ i = Ei εi (2.20)
46 Discontinuous deformation analysis in rock mechanics practice
where S is the area of block i as the integration is over the entire area of the block.
The derivatives are calculated to minimize the strain energy e :
∂ 2 e S ∂2
krs = = d T Ei di = SEi (r, s = 1, . . ., 6) → Kii (2.24)
∂dri ∂dsi 2 ∂dri ∂dsi i
x = (x2 − x1 )t + x1
0≤t≤1 (2.29)
y = (y2 − y1 )t + y1
The loading is a variant along the loading line and represented as follows:
Fx = Fx (t)
0≤t≤1 (2.31)
Fy = Fy (t)
The potential energy of the line loading (Fx (t), Fy (t)) is:
) 1 ) 1
Fx (t) T Fx (t)
l = − (u v) ldt = −di
T
Ti ldt (2.32)
0 Fy (t) 0 Fy (t)
48 Discontinuous deformation analysis in rock mechanics practice
) 1 ) 1
∂ l ∂ T Fx (t) T Fx (t)
fr = − = d TiT ldt = Ti ldt → Fi (2.33)
∂dri ∂dri i 0 Fy (t) 0 Fy (t)
) 1 ) 1
Fx (t) Fx (t)
TiT ldt = TiT ldt (2.34)
0 Fy (t) 0 Fy (t)
) 1
l
((x2 − x1 )t + (x1 − x0 ))ldt = (x2 + x1 − 2x0 )
0 2
) (2.35)
1
l
((y2 − y1 )t + (y1 − y0 ))ldt = (y2 + y1 − 2y0 )
0 2
Therefore,
⎛ ⎞
Fx
⎜ ⎟
⎜ ⎟ Fy
⎜ ⎟
⎜ 1 ⎟
⎜− (y + y − 2y )F + 1 (x + x − 2x )F ⎟
⎜ 2 2 1 0 x 2 1 0 y⎟
) 1 ⎜ 2 ⎟
Fx ⎜ ⎟
TiT ldt =l⎜
⎜
1
(x + x − 2x )F
⎟ → Fi
⎟ (2.36)
0 Fy ⎜ 2
2 1 0 x
⎟
⎜ ⎟
⎜ 1 ⎟
⎜ (y2 + y1 − 2y0 )Fy ⎟
⎜ 2 ⎟
⎜ ⎟
⎝ 1 1 ⎠
(y2 + y1 − 2y0 )Fx + (x2 + x1 − 2x0 )Fy
4 4
⎡ ⎤ ⎡ ⎤
S 0 S 0
⎢ 0 S ⎥ ⎢0 S⎥
)) ⎢ ⎥ ⎢ ⎥
⎢ −Sy + y0 S Sx − x0 S ⎥ ⎢ 0⎥
⎢
Ti dxdy = ⎢
T ⎥ = ⎢0 ⎥ (2.38)
S − x S 0 ⎥ ⎢0 0⎥
⎢ x 0 ⎥ ⎢ ⎥
⎣ 0 Sy − y0 S ⎦ ⎣0 0⎦
(Sy − y0 S)/2 (Sx − x0 S)/2 0 0
1
dl = [(x1 − x2 )(dx1 − dx2 ) + (y1 − y2 )(dy1 − dy2 )]
l
1
= [(x1 − x2 )(u1 − u2 ) + (y1 − y2 )(v1 − v2 )]
l
lx lx T T lx T T lx
= u 1 v1 − u 2 v2 = di Ti − d j Tj (2.41)
ly ly ly ly
1
lx = (x1 − x2 )
l
(2.42)
1
ly = (y1 − y2 )
l
dl
f = −s (2.43)
l
50 Discontinuous deformation analysis in rock mechanics practice
2
1 s s l l
b = − fdl = dl 2 = diT TiT x − djT TjT x
2 2l 2l ly ly
s l s l
= diT TiT x (lx ly )Ti di − diT TiT x (lx ly )Tj dj
2l l y l ly
s l
+ djT TjT x (lx ly )Tj dj
2l ly
s T s s
= di Ei EiT di − diT Ei GjT dj + djT Gj GjT dj (2.44)
2l l 2l
where
lx
Ei = TiT
ly
lx
Gj = TjT
ly
∂ 2 b s ∂2 s
krs = = d T Ei EiT di = Ei EiT → Kii (2.45)
∂dri ∂dsi 2l ∂dri ∂dsi i l
∂ 2 b s ∂2 s
krs = =− diT Ei GjT dj = − Ei GjT → Kij (2.46)
∂dri ∂dsj l ∂dri ∂dsj l
∂ 2 b s ∂2 s
krs = =− diT Ei GjT dj = − Gj EiT → Kji (2.47)
∂drj ∂dsi l ∂drj ∂dsi l
∂ 2 b s ∂2 s
krs = = djT Gj GjT dj = Gj GjT → Kjj (2.48)
∂drj ∂dsj 2l ∂drj ∂dsj l
⎛ 2 ⎞
∂ u(t)
⎜ ∂t 2 ⎟ 2
fx
= −M ⎜ ⎟ = −MTi ∂ di (t) (2.49)
fy ⎝ 2
∂ v(t) ⎠ ∂t 2
∂t 2
)) ))
fx ∂2 d(t)
i = − (u v) dxdy = M diT TiT Ti dxdy (2.50)
fy ∂t 2
52 Discontinuous deformation analysis in rock mechanics practice
Assume di (0) = 0 is the block displacement at the beginning of the time step, is
the time step, and di () = di is the block displacement at the end of the time step.
Using the time integration:
∂ 2 m
krs = = pTiT Ti → Kii (2.60)
∂dri ∂dsi
krs (r, s = 1, . . ., 6) forms a 6 × 6 matrix and it is added to the sub-matrix Kii in the
global Equation 2.4.
And the following vector can be added to the global force vector:
∂ m T um
fr = − = pTi → Fi
∂dri vm
Denote (lx , ly )(lx2 + ly2 = 1) as the direction along which the displacement δ exists. The
spring displacement is d = δ − (lx u + ly v) and the spring force is represented by:
where p is the stiffness of the spring, which is a very large number, normally from 10E
to 1000E, to guarantee the displacement of the spring is 10−3 to 10−4 times the total
displacement. If p is large enough, the computation results will be independent of the
selections of p.
The strain energy of the spring is:
p 2 p lx u l p
c = d = (u v) (lx ly ) − pδ(u v) x + δ2 (2.62)
2 2 ly v ly 2
% &
l
Let L = x and c = T T L
ly
Then
l
(u v) x = diT TiT L = diT ci
ly
u
(lx ly ) = LiT TiT di = ciT di
v
Therefore, we have
p p
c = diT ci ciT di − pδdiT ci + δ2 (2.63)
2 2
A 6 × 6 matrix is obtained by taking the derivatives of the strain energy of the stiff
spring:
∂ 2 c p ∂2
krs = = d T ci ciT di = pci ciT → Kii (2.64)
∂dri ∂dsi 2 ∂dri ∂dsi i
krs (r, s = 1, . . ., 6) forms a 6 × 6 matrix and it is added to the sub-matrix Kii in the
global Equation 2.4.
We minimize the spring strain energy c by taking the derivatives of c at d = 0:
∂ c (0) ∂ T
fr = − = pδ d ci = pδci → Fi (2.65)
∂dri ∂dri i
The DDA blocks are interacting with neighboring blocks through contacting and
separating, and their movement must obey the imposed contact criterion, i.e., ‘no
Theory of the discontinuous deformation analysis (DDA) 55
penetration, no tension’ (Shi, 1988). The effect of the contact can be represented by
applying two stiff contact springs in the normal and shear directions or frictional forces
along the sliding edge. The normal and shear contact springs are added if the blocks
are in contact and are not sliding relative to one another, and deleted if the blocks
separate or the normal contact force is tensile. If the blocks are in contact and sliding
relative to one another, a normal spring is added together with the frictional forces.
The solution of the analysis requires the exact number of contacts and their relevant
information. However, the total number of contacts is unknown prior to the solution of
the problem. This phenomenon is particularly serious in multi-body contact problems.
For such a problem, a possible solution is to use a trial-and-error iteration procedure,
which in DDA is called “open-close’’ iterations (Shi, 1988). Open-close iterations are
applied to identify the contacts and to arrange the correct locations of contact springs
for each time step. After determining the contact points and the associated contact
forces, the normal contact and shear contact sub-matrices or friction force sub-matrices
are formulated, calculated, and added to the global simultaneous equations.
Figure 2.5 Vertex-to-edge contact and vertex-to-vertex contact between two blocks.
Figure 2.6 Edge-to-edge contacts between two blocks.
α ≤ 180◦ β ≤ 180◦ OP 3 OP 2
α ≤ 180◦ β > 180◦ OP 3 OP 4
α > 180◦ β ≤ 180◦ OP 1 OP 2
α > 180◦ β > 180◦ OP 1 OP 4
Contacts are determined by calculating the penetration distance dn and the stiff
springs are applied to ensure no-penetration. For the contact between two convex
angles, a normal spring is used between the vertex and its corresponding entrance
line, which has a smaller entrance distance. For the contact between a convex angle
and a concave angle, a stiff spring is applied if only one of the two entrance lines is
passed, and two stiff springs are applied if both entrance lines are passed. For the
vertex-to-edge contact, only one stiff spring is used.
For each vertex-entrance line contact pair, there are three possible contact modes:
open, sliding and locked. At the beginning of each time step, the contact modes for
all contact pairs are assumed to be locked except those contact pairs inherited from
the previous time step. For a locked contact pair, a normal spring is applied to push
the vertex away from the entrance line in the normal direction and a shear spring is
applied to avoid the tangential displacement between the vertex and the entrance line.
Penalty method
The penalty method, most widely used due to its simplicity, was originally adopted in
the DDA method to enforce the contact constraints at block interfaces. This method
does not increase the number of governing equations because of contacts, and the
solution is easily obtained by simply adding contact sub-matrices to the stiffness matrix.
However, the penalty method is just an approximation of the actual contact con-
straints and it reaches the exact result when the penalty parameters kN and kT are
approaching infinity. In practical applications, however, it is impossible to use very
large penalty parameters, as this would lead to an ill-conditioned coefficient matrix.
The formulation for the contact contribution for the lock state is:
1 1
c = kN (dN )2 + kT (dT )2 (2.66)
2 2
where dN and dT are the normal and tangential gap functions.
58 Discontinuous deformation analysis in rock mechanics practice
c = λN dN + λT dT (2.67)
Table 2.2 Rules for Open-Close Iteration following Doolin and Sitar (2002).
Previous
Type state Contact condition New state Operation
N is the normal contact force, S is the shear contact force, ψ is the direction of shear displacement, and f is the
frictional force.
1. Initialize data
2. For each ti , i = 1, . . . , n time steps do
3. Identify/update contacts
4. Assemble stiffness matrix
5. Integrate
6. Repeat (Open-close iteration)
7. Add/subtract contact matrices
8. Solve the unknowns d
9. Resolve contact conditions
10. Until no-tension and no penetration
11. Update block geometry
12. Update block stresses, velocities
13. End for
dn = (2.69)
l
Normal contact
−−
If P1 passes through edge P2 P3 (Figure 2.8), dn should be negative. is the area of
−−
triangle P1 P2 P3 , and l is the length of edge P2 P3 at the end of the current time step:
*
l= (x2 + u2 − x3 − u3 )2 + (y2 + v2 − y3 − v3 )2
*
≈ (x2 − x2 )2 + (y2 − y3 )2 (2.70)
1 x1 + u1 y1 + v1
u1
= 1 x2 + u2 y2 + v2 ≈ Sn0 + {y2 − y3 x3 − x 2 }
v1
1 x3 + u3 y3 + v 3
(2.71)
u2 u3
+ {y3 − y1 y1 − y 3 } + {y1 − y2 y2 − y 1 }
v2 v3
1 x1 y1
S0 = 1 x2 y2 (2.72)
1 x3 y3
% & % &
1 u1 u2
dn = S0 + {y2 − y3 x3 − x 2 }+ {y3 − y1 x1 − x 3 }
l v1 v2
% &
u
+ {y1 − y2 x2 − x 1 } 3 (2.73)
v3
−−
Since P1 belongs to block i while edge P2 P3 belongs to block j, the displacements
of P1 , P2 and P3 can be represented as follows:
u1
= Ti (x1 , y1 )di
v1
u2
= Tj (x2 , y2 )dj (2.74)
v2
u3
= Tj (x3 , y3 )dj
v3
Then
⎛ ⎞ ⎛ ⎞
d1i d1j
⎜d2i ⎟ ⎜d2j ⎟
⎜ ⎟ ⎜ ⎟
⎜d ⎟ ⎜d ⎟
S0 ⎜ 3i ⎟ ⎜ 3j ⎟
dn = + (e1 e2 e3 e4 e5 e6 ) ⎜ ⎟ + (g1 g2 g3 g4 g5 g6 ) ⎜ ⎟
l ⎜d4i ⎟ ⎜d4j ⎟
⎜ ⎟ ⎜ ⎟
⎝d5i ⎠ ⎝d5j ⎠
d6i d6j
(2.75)
where
1
er = ((y2 − y3 )t1r (x1 , y1 ) + (x3 − x2 )t2r (x1 , y1 ))
l
1
gr = ((y3 − y1 )t1r (x2 , y2 ) + (x1 − x3 )t2r (x2 , y2 ))
l
1
+ ((y1 − y2 )t1r (x3 , y3 ) + (x2 − x1 )t2r (x3 , y3 )) (2.76)
l
62 Discontinuous deformation analysis in rock mechanics practice
6 2
p 6
p 2 S0
nc = dn = er dri + gr drj +
2 2 l
r=1 r=1
⎛ 2 2
p⎝
6 6
= er dri + gr drj
2
r=1 r=1
6 6
6
2S0
+2 er dri gr drj + er dri
l
r=1 r=1 r=1
6
2S0 S02
+ gr drj + (2.77)
l l2
r=1
Four 6 × 6 sub-matrices and two 6 × 1 sub-matrices are obtained and then added
to Kii , Kij , Kji , Kjj , Fi and Fj respectively, by minimizing the strain energy nc .
The derivative of nc
6 2
∂2 nc p ∂2
krs = = er dri
∂dri ∂dsi 2 ∂dri ∂dsi
r=1
⎛ ⎞
e1
⎜ e2 ⎟
⎜ ⎟
⎜ e3 ⎟
= p⎜ ⎟
⎜e4 ⎟ (e1 e2 e3 e4 e5 e6 ) → Kii (2.78)
⎜ ⎟
⎝ e5 ⎠
e6
6 6
∂2 nc p ∂2
krs = = er dri gr drj
∂dri ∂dsj 2 ∂dri ∂dsj
r=1 r=1
⎛ ⎞
e1
⎜e2 ⎟
⎜ ⎟
⎜e3 ⎟
= p⎜ ⎟
⎜e4 ⎟ (g1 g2 g3 g4 g5 g6 ) → Kij (2.79)
⎜ ⎟
⎝e5 ⎠
e6
The derivative of nc
∂2 nc p ∂2
6
6
krs = = er dri gr drj
∂drj ∂dsi 2 ∂dri ∂dsi
r=1 r=1
⎛ ⎞
g1
⎜g2 ⎟
⎜ ⎟
⎜g3 ⎟
= p⎜ ⎟
⎜g4 ⎟ (e1 e2 e3 e4 e5 e6 ) → Kij (2.80)
⎜ ⎟
⎝g5 ⎠
g6
Shear contact
−−
Assume P0 with coordinates (x0 , y0 ) is the projection of vertex P1 on the edge P2 P3 ,
and the coordinates of P0 can be represented as follows:
x0 = (1 − t)x2 + x3 t
(2.84)
y0 = (1 − t)y2 + y3 t
Ignoring the second-order infinitesimal small terms and rearranging Equation 2.85,
gives,
% &
S0 1 u − u0
ds = + (x3 − x2 y3 − y2 ) 1
l l v1 − v 0
% & % &
S0 1 u 1 u
= + (x3 − x2 y3 − y2 ) 1 + (x2 − x3 y2 − y3 ) 0 (2.86)
l l v1 l v0
% &
x1 − x0
S0 = (x3 − x2 y3 − y2 ) (2.87)
y1 − y 0
Then
⎛ ⎞ ⎛ ⎞
d1i d1j
⎜d2i ⎟ ⎜d2j ⎟
⎜ ⎟ ⎜ ⎟
S0 ⎜d3i ⎟ ⎜d3j ⎟
ds = + (e1 e2 e3 e4 e5 e6 ) ⎜ ⎟
⎜d4i ⎟ + (g1 g2 g3 g4 g5 g6 ) ⎜
⎜d4j ⎟
⎟
l ⎜ ⎟ ⎜ ⎟
⎝d5i ⎠ ⎝d5j ⎠
d6i d6j
(2.89)
Theory of the discontinuous deformation analysis (DDA) 65
where
1
er = ((y3 − y2 )t1r (x1 , y1 ) + (x3 − x2 )t2r (x1 , y1 ))
l
1
gr = ((−x1 + 2(1 − t)x2 − (1 − 2t)x3 )t1r (x2 , y2 )
l
+ (−y1 + 2(1 − t)y2 − (1 − 2t)y3 )t2r (x2 , y2 )) (2.90)
1
+ ((x1 − (1 − 2t)x2 − 2tx3 )t1r (x3 , y3 )
l
+(y1 − (1 − 2t)y2 − 2ty3 )t2r (x3 , y3 ))
6 2
p 6
p 2 S0
sc = ds = er dri + gr drj +
2 2 l
r=1 r=1
⎛ 2 6 2
p⎝
6
= er dri + gr drj
2
r=1 r=1
6 6
6
2S0
+2 er dri gr drj + er dri
l
r=1 r=1 r=1
6
2S0 S02
+ gr drj + (2.91)
l l2
r=1
Four 6 × 6 sub-matrices and two 6 × 1 sub-matrices are obtained and then added
to Kii , Kij , Kji , Kjj , Fi and Fj respectively, by minimizing the strain energy sc .
The derivative of sc
6 2
∂2 sc p ∂2
krs = = er dri
∂dri ∂dsi 2 ∂dri ∂dsi
r=1
⎛ ⎞
e1
⎜e2 ⎟
⎜ ⎟
⎜e3 ⎟
= p⎜ ⎟
⎜e4 ⎟ (e1 e2 e3 e4 e5 e6 ) → Kii (2.92)
⎜ ⎟
⎝e5 ⎠
e6
The derivative of sc
6 6
∂2 sc p ∂2
krs = = er dri gr drj
∂dri ∂dsj 2 ∂dri ∂dsj
r=1 r=1
⎛ ⎞
e1
⎜e2 ⎟
⎜ ⎟
⎜e3 ⎟
= p⎜ ⎟
⎜e4 ⎟ (g1 g2 g3 g4 g5 g6 ) → Kij (2.93)
⎜ ⎟
⎝e5 ⎠
e6
The derivative of sc
⎛ ⎞
g1
6 ⎜g ⎟
⎜ 2⎟
∂ sc (0) pS0 ∂ pS0 ⎜g ⎟
⎜ 3 ⎟ → Fj
fr = − =− dr drj = − (2.97)
∂drj l ∂drj l ⎜ 4⎟
⎜g ⎟
r=1
⎝g5 ⎠
g6
Frictional force
For sliding mode, besides the normal spring, a pair of frictional forces instead of a
shear spring should be added. The frictional force is calculated based on the normal
contact compressive force from the former iteration:
where pn is the normal spring stiffness, ϕ is the friction angle, dn is the normal
penetration distance taken from the previous iteration, and c is cohesion.
The friction force Ff is along the direction
1
(x3 − x2 y3 − y2 )
l
and
*
l = (x3 − x2 )2 + (y3 − y2 )2
−−
is the length of line P2 P3 .
The potential energy of the frictional force Ff at P1 on block i is:
F x3 − x 2 F T T x3 − x 2
f = (u1 v1 ) = di Ti (x1 , y1 ) = FdiT H (2.99)
l y3 − y 2 l y3 − y 2
⎛ ⎞
e1
⎜e2 ⎟
⎜ ⎟
1 T x3 − x2 ⎜ ⎟
T ⎜ e3 ⎟
H = Ti (x1 , y1 ) = Fdi ⎜ ⎟ (2.100)
l y3 − y 2 ⎜ e4 ⎟
⎝ e5 ⎠
e6
68 Discontinuous deformation analysis in rock mechanics practice
⎛ ⎞
g1
⎜g2 ⎟
⎜ ⎟
1 x − x2 ⎜g3 ⎟
G = TjT (x0 , y0 ) 3 = FdjT ⎜
⎜g4 ⎟
⎟ (2.103)
l y3 − y 2 ⎜ ⎟
⎝g5 ⎠
g6
t 2
dn+1 = dn + t ḋn + [(1 − 2β)d̈n + 2βd̈n+1 ] (2.105)
2
t is the time step size for an incremental dynamic formulation, dn and dn+1 denote the
approximation to the values d(t) and d(t + t) for a time step t, β is an acceleration
weighting parameter and γ is a velocity weighting parameter.
It can be shown that the Newmark–β method is unconditionally stable when the
following condition is met (Hughes, 1983):
1
2β ≥ γ ≥ (2.107)
2
Newmark–β method with two parameters β = 0.5 and γ = 1.0 is used in DDA
(Doolin and Sitar, 2004); therefore Equations 2.105 and 2.106 can be simplified as
follows:
2
ḋn+1 = ḋn + t d̈n+1 = dn+1 − ḋn (2.108)
t
2 2
d̈n+1 = 2
dn+1 − ḋn (2.109)
t t
It can be seen that this scheme satisfies the criterion in Equation 2.107 and therefore
it is unconditionally stable. According to Hughes (1983), the condition to prevent
bifurcation is as follows:
1 2
β≥ γ + /4 (2.110)
2
However, the Newmark–β scheme with two parameters β = 0.5 and γ = 1.0 vio-
lates this condition, inducing bifurcation. For an undamped analysis, the critical
sampling frequency at which bifurcation occurs is:
− 12
1 1 2
bif = γ+ −β (2.111)
4 2
For β = 0.5 and γ = 1.0, bif = 4. On the other hand, = tω, meeting the
constraint of Equation 2.111, requires that:
4
t < (2.112)
ωmax
Distinguished from other numerical methods where numerical integrations are typi-
cally used, the DDA method uses Simplex for all integrations. Simplex integrations are
accurate solutions on n-dimensional, generally shaped, domains.
A simplex is an oriented simplest domain. Simplex integration is used to calculate
ordinary integrations and it is not limited for integrations on a simplex only. A complex
shape can always be sub-divided into simplex, so simplex integration can be calculated
in each simplex and the sum of the simplex integrations is the ordinary integration over
the complex shape.
A two-dimensional block is assumed to be a polygon with vertices arranged
anticlockwise such as,
P 1 P 2 P3 · · · Pn Pn+1 = P1
and
Pi = (xi , yi )
) n )
n
1 xk yk
xdxdy = xD(x, y) = xk+1 (x + xk+1 ) (2.114)
Area P0 Pk Pk+1 6 yk+1 k
k=1 k=1
) n )
n
1 xk yk
ydxdy = yD(x, y) = xk+1 (y + yk+1 ) (2.115)
Area P0 Pk Pk+1 6 yk+1 k
k=1 k=1
) n )
x2 dxdy = x2 D(x, y)
Area P0 Pk Pk+1
k=1
n
1 xk yk 2
= xk+1 (x + x2k+1 + xk xk+1 ) (2.116)
12 yk+1 k
k=1
) n )
y dxdy =
2
y2 D(x, y)
Area P0 Pk Pk+1
k=1
n
1 xk yk 2
= xk+1 (y + yk+1
2
+ yk yk+1 ) (2.117)
12 yk+1 k
k=1
Theory of the discontinuous deformation analysis (DDA) 71
) n )
xydxdy = xyD(x, y)
Area P0 Pk Pk+1
k=1
n
1 xk yk
= xk+1 (2xk yk + 2xk+1 yk+1 + xk+1 yk + xk yk+1 )
24 yk+1
k=1
(2.118)
For more details on simplex integration for DDA see Shi (1996).
2.6 SUMMARY
We reviewed the main theory of 2D DDA in this chapter, including formulation of the
governing equations based on the minimum potential energy theory, contact algorithm,
time-step integration and simplex integration schemes used in DDA. Since the original
DDA method was introduced by Shi (1988) many developments have been presented
by researchers in the field, some of which have been reviewed in Chapter 1. Many
unresolved issues still remain and it is expected future research will yield advances
in topics such as improvement of the computational efficiency during the open-close
iteration process, determination of optimal time interval in time domain integration,
damping effects, and coupling with other physical processes such as water and heat
flow. The basic theory of 3D DDA, which is an extension of this chapter, will be
introduced in the next chapter.
Chapter 3
The theoretical analysis of DDA in two dimensions presented in Chapter 2 reveals the
following general characteristics of DDA: (1) The principle of minimum total potential
energy is used to find a solution, similar to the FEM. (2) Dynamic and static problems
can be solved by applying the same formulations. (3) Different constitutive laws for
the materials can be incorporated. (4) Different types of contact criterion, boundary
condition, loading condition, and volumetric force, can be modelled. With regard to
modelling three-dimensional problems, however, the two-dimensional approach can
only provide a gross approximation of the real behavior. The shortcomings of two-
dimensional representation become evident when considering for example discrete
block interactions in masonry structures, expected failure modes in multiple free sur-
face conditions such as in tunnel portals, or in blocky rock masses consisting of more
than two joint sets where the interaction between the finite blocks and the free sur-
faces control the mechanical deformation. In order to solve such problems accurately, a
robust three-dimensional approach is required. In this chapter, the basic formulations
of 3D DDA is presented.
• Each loading step (i.e. time step) satisfies the conditions of very small displacement
and deformation.
• Block stresses and strains are constant.
are normal and shear strain vectors of the block centroid (see Fig. 3.1(c) and (d)),
respectively.
Assuming constant strains and constant stresses within each block, the first order
approximation of displacement can be written as follows:
⎧
⎨ u = a0 + a1 x + a2 y + a3 z
v = b0 + b1 x + b2 y + b3 z (3.1)
⎩
w = c0 + c1 x + c2 y + c3 z
Substituting the coordinates of the initial position of the centroid into Equation 3.1,
the displacement of the centroid can be written as:
⎧
⎨ u0 = a0 + a1 x0 + a2 y0 + a3 z0
v0 = b0 + b1 x0 + b2 y0 + b3 z0 (3.2)
⎩
w0 = c0 + c1 x0 + c2 y0 + c3 z0
Combining Equations 3.1 and 3.2 we get the displacement components of the
block:
⎧
⎨ u = u0 + a1 (x − x0 ) + a2 (y − y0 ) + a3 (z − z0 )
v = v0 + b1 (x − x0 ) + b2 (y − y0 ) + b3 (z − z0 )
⎩
w = w0 + c1 (x − x0 ) + c2 (y − y0 ) + c3 (z − z0 )
where:
In the discussion that follows, the potential energy of each force or stress and their
differentiations are computed separately.
The total potential energy can be written as:
⎡ ⎤⎛ ⎞
K11 K12 K13 · · · K1n D1
⎢K21 K22 K23 · · · K2n ⎥⎜D2 ⎟
⎢ ⎥ ⎜ ⎟
1 ⎢ ⎥⎜ ⎟
= (DT1 DT2 DT3 . . . DTn )⎢K31 K32 K33 · · · K3n ⎥⎜D3 ⎟
2 ⎢ .. .. .. .. .. ⎥⎜ .. ⎟
⎣ . . . . . ⎦⎝ . ⎠
Kn1 Kn2 Kn3 · · · Knn Dn
⎧ ⎫
⎪ F1 ⎪
⎪ ⎪
⎪
⎪ ⎪
⎪
⎨ F2 ⎪
⎪ ⎬
+ (DT1 DT2 DT3 . . . DTn ) F3 + C (3.5)
⎪
⎪ .⎪
⎪ .. ⎪
⎪ ⎪
⎪
⎩ ⎪
⎪ ⎭
Fn
where sub-matrices Kii depend on the material properties of block i and sub-matrices
Kij are defined by the contacts between blocks i and j. Di and Fi are 12 × 1 sub-matrices,
and Fi is the loading distributed to the twelve unknown variables of block i.
Based on the minimized potential energy approach we can write,
∂
= 0, r = 1, 2, . . . , 12 (3.6)
∂dri
where sub-matrices Kij present the ith row and the jth column stiffness
which are 12 × 12 sub-matrices; Di is the ith block deformation variables
(d1i , d2i , d3i , . . . d12i )T ; and Fi is the loading on the ith block. Sub-matrices
Kij (i = j) are determined by the block material properties, whereas submatrices
Kij (i = j) are related to the contacts between the ith and the jth blocks.
The matrix Kij (i, j = 1, 2, . . . , n) is derived by the differentiation:
∂2
(r, s = 1, 2, . . . , 12) (3.8)
∂dri ∂dsj
The equilibrium equations of the DDA method are derived by minimizing the total
potential energy of the block system. For a single block, the total potential energy may
include components from elastic strains, initial stress, point loading, line loading, body
force, bolting connection, inertia force, etc., similar to those introduced in Chapter 2.
To solve the derivative of deformation variables from each separated potential energy,
the sub-matrices are calculated first. Then, the global equilibrium equation of the entire
block system is constructed by assembling the sub-matrices.
where the integration is over the entire volume of block i. For each displacement step,
the blocks are assumed to be linearly elastic. The relationship between stress and strain
is generally expressed as follows:
σ i = Ei εi (3.11)
78 Discontinuous deformation analysis in rock mechanics practice
or in a matrix form of
⎛ ⎞ ⎡ ⎤⎛ ⎞
σx 1−v v v εx
⎜ σy ⎟ ⎢ v 1−v v 0 ⎥ ⎜ εy ⎟
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎜ σz ⎟ E ⎢ v vi 1−v ⎥ ⎜ εz ⎟
⎜ ⎟= ⎢ ⎥⎜ ⎟
⎜ τ yz ⎟ (1 − v2 )(1 − 2v) ⎢ 0.5 − v 0 0 ⎥ ⎜ ⎟
⎜ ⎟ ⎢ ⎥ ⎜ γ yz ⎟
⎝τ zx ⎠ ⎣ 0 0 0.5 − v 0 ⎦ ⎝γ zx ⎠
τ xy 0 0 0.5 − v γ xy
(3.12)
The block strain energy can then be rewritten as:
⎛ ⎞
σx
⎜ σy ⎟
))) ⎜ ⎟
+ 1 ⎜ σz ⎟
= εx εy εz γyz γzx γxy ⎜ ⎟
⎜ τyz ⎟ dxdydz
elastic 2 ⎜ ⎟
⎝τzx ⎠
τxy
)))
1
= [Di ]T [Ei ][Di ]dxdydz
2
Vi T
= d E i di (3.13)
2 i
where Vi is the volume of the ith block.
The derivatives are calculated to minimize the strain energy elastic :
∂ 2 e Vi ∂2
krs = = ([Di ]T [Ei ][Di ]) = Vi Ei , r, s = 1, . . . , 12 → Kii (3.14)
∂dri ∂dsi 2 ∂dri ∂dsi
which is added to the sub-matrix Kii in the global Equation 3.7.
∂ initial
fr = − = Vi σ 0 (r = 1, . . . , 12) → Fi (3.16)
∂dri
The potential energy of the line loading (Flx , Fly , Flz ) is:
⎛ ⎞ ⎛ ⎞
) Flx ) Flx
lineload = − (u v w) ⎝Fly ⎠ dl = −[Di ]T [Ti (x, y, z)]T ⎝Fly ⎠ dl (3.20)
l Flz l Flz
Ax + By + Cz + D = 0 (3.22)
Since the coordinates of the centre of gravity of block i are assumed to be (x0 y0 z0 ),
Sx Sy Sz
x0 = , y0 = , z0 = (3.27)
Vi Vi Vi
... ... ... ...
where Sx = xdxdydz, Sy = ydxdydz, Sz = zdxdydz, Vi = dxdydz.
Then,
⎛ ⎞ ⎛ ⎞
Vi 0 0 Vi 0 0
⎜ 0 V 0 ⎟ ⎜0 Vi 0⎟
⎜ i ⎟ ⎜ ⎟
⎜ 0 0 Vi ⎟ ⎜0 0 Vi ⎟
⎜ ⎟ ⎜ ⎟
⎜ 0 −(Sz − z0 Vi ) (Sy − y0 Vi ) ⎟ ⎜ 0⎟
⎜ ⎟ ⎜0 0 ⎟
⎜ (Sz − z0 Vi ) 0 −(Sx − x0 Vi ) ⎟ ⎜
⎟ 0⎟
))) ⎜ ⎜0 0 ⎟
⎜ −(S − y V ) (S − x V ) 0 ⎟ ⎜0 0 0⎟
[Ti ] dxdydz = ⎜
T y 0
⎜ (Sx − x0 Vi )
i x 0 i ⎟=⎜
⎟ ⎜0
⎟
⎜ 0 0 ⎟ ⎜ 0 0⎟ ⎟
⎜ 0 (Sy − y0 Vi ) 0 ⎟ ⎜0 0 0⎟
⎜ ⎟ ⎜ ⎟
⎜ 0 0 (Sz − z0 Vi ) ⎟ ⎜ 0⎟
⎜ ⎟ ⎜0 0 ⎟
⎜ 0 (Sz − z0 Vi ) (Sy − y0 Vi )/2 ⎟ ⎜⎟ 0⎟
⎜ ⎜0 0 ⎟
⎝ (Sz − z0 Vi )/2 0 (Sx − x0 Vi )/2⎠ ⎝ 0 0 0⎠
(Sy − y0 Vi )/2 (Sx − x0 Vi ) 0 0 0 0
(3.28)
82 Discontinuous deformation analysis in rock mechanics practice
∂ g T
fr = − = fx Vi fy V i fz V i 0 0 0 0 0 0 0 0 0 → Fi (3.29)
∂dri
*
l = (x1 − x2 )2 + (y1 − y2 )2 + (z1 − z2 )2
and
1
dl = [(x1 − x2 )(dx1 − dx2 ) + (y1 − y2 )(dy1 − dy2 ) + (z1 − z2 )(dz1 − dz2 )]
l
1
= [(x1 − x2 )(u1 − u2 ) + (y1 − y2 )(v1 − v2 ) + (z1 − z2 )(w1 − w2 )]
l
⎛ ⎞ ⎛ ⎞
lx lx
= [Di ]T [Ti (x, y, z)]T ⎝ly ⎠ − [Dj ]T [Tj (x, y, z)]T ⎝ly ⎠ (3.30)
lz lz
1
lx = (x1 − x2 )
l
1 (3.31)
ly = (y1 − y2 )
l
1
lz = (z1 − z2 )
l
dl
f = −s (3.32)
l
where
⎛ ⎞
lx
Ei = TiT ⎝ly ⎠
lz
⎛ ⎞
lx
Gj = TjT ⎝ly ⎠
lz
∂ 2 b s ∂2 s
krs = = diT Ei EiT di = Ei EiT → Kii (3.34)
∂dri ∂dsi 2l ∂dri ∂dsi l
∂ 2 b s ∂2 s
krs = =− diT Ei GjT dj = − Ei GjT → Kij (3.35)
∂dri ∂dsj l ∂dri ∂dsj l
∂ 2 b s ∂2 s
krs = =− d T Ei GjT dj = − Gj EiT → Kji (3.36)
∂drj ∂dsi l ∂drj ∂dsi i l
84 Discontinuous deformation analysis in rock mechanics practice
∂ 2 b s ∂2 s
krs = = d T Gj GjT dj = Gj GjT → Kjj (3.37)
∂drj ∂dsj 2l ∂drj ∂dsj j l
Assume di (0) = 0 is the block displacement at the beginning of the time step,
is the time step, and di () = di is the block displacement at the end of the time step.
Using the time integration we get:
The potential energy of the inertia force at the end of each time step is:
)))
∂2 d(t)
inertia = M diT TiT Tidxdydz
∂t 2
)))
2 2 ∂di (0)
= Mdi
T T
Ti Ti dxdydz di − (3.42)
2 ∂t
Theory of the discontinuous deformation analysis in three dimensions 85
where p is the stiffness of the spring, which is a very large value, normally from 10E
to 1000E, to guarantee the displacement of the spring is 10−1 to 10−3 times the total
displacement. If p is large enough, the computation results will be independent of the
selections of p.
The strain energy of the spring is:
p 2
constraint = d
2
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
lx u lx
p p
= (u v w) ⎝ly ⎠ (lx ly lz ) ⎝ v ⎠ − pδ(u v w) ⎝ly ⎠ + δ2 (3.47)
2 lz w lz 2
⎛ ⎞
lx
Let L = ⎝ly ⎠ and c = T T L
lz
86 Discontinuous deformation analysis in rock mechanics practice
Then:
⎛ ⎞
lx
(u v w) ⎝ly ⎠ = diT TiT L = diT ci
lz
⎛ ⎞
u
(lx ly lz ) ⎝ v ⎠ = LiT Ti di = ciT di
w
Therefore:
p p
constraint = diT ci ciT di − pδdiT ci + δ2 (3.48)
2 2
A 12 × 12 matrix is obtained by taking the derivatives of the strain energy of the
stiff spring:
∂2 constraint p ∂2
krs = = d T ci ciT di = pci ciT → Kii (3.49)
∂dri ∂dsi 2 ∂dri ∂dsi i
krs (r, s = 1, . . . , 12) forms a 12 × 12 matrix and it is added to the sub-matrix Kii in the
global Equation 3.7.
The spring strain energy constraint is minimized by taking the derivatives of c at
d = 0:
∂ constraint (0) ∂ T
fr = − = pδ d ci = pδci → Fi (3.50)
∂dri ∂dri i
Therefore, the strain energy fixpoint caused by the constrained springs can be
expressed as:
kfix
fixpoint = [(uf − ufs )2 + (vf − vfs )2 + (wf − wfs )2 ]
2
Theory of the discontinuous deformation analysis in three dimensions 87
⎡ ⎤
kfix (uf − ufs )
= [(uf − ufs )(vf − vfs )(wf − wfs )]⎣ (vf − vfs ) ⎦
2 (wf − wfs )
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
kfix uf ufs kfix ufs
= (uf vf wf ) ⎝ vf ⎠ − kfix (uf vf wf ) ⎝ vfs ⎠ + (ufs vfs wfs ) ⎝ vfs ⎠
2 wf wfs 2 wfs
(3.52)
Since:
⎛ ⎞
uf
⎝ vf ⎠ = Ti di
wf
By minimizing the potential energy fixpoint , the following matrix can then be
added to the submatrix Kii in the global stiffness matrix:
∂2 fixpoint kfix T
krs = = T Ti → Kii (3.54)
∂dri ∂dsi 2 i
krs (r, s = 1, . . . , 12) forms a 12 × 12 matrix and it is added to the sub-matrix Kii in the
global Equation 3.7.
88 Discontinuous deformation analysis in rock mechanics practice
In Chapter 8 the complete new contact theory developed by Shi is presented in great
detail. Here we summarize the contact algorithms as used in the original 3D-DDA.
Contact transfer
Since the calculation in discontinuous deformation analysis utilizes a time step march-
ing scheme, the contact states at the previous time step are inherited as the initial states
before the next time step begins. Therefore, all the geometrical and physical parame-
ters in the block system must be transferred from the previous step to the next step.
The transferred parameters include:
The transferred information of close contacts includes contact position and contact
state, as follows:
Open-close iterations
In two-dimensional DDA the entrance line theory is applied to examine whether a
vertex penetrates into an edge, and then the open-close iteration law is introduced.
In three-dimensional DDA the normal contact can be presented by the normal vector
to a face, so the entrance line theory can be employed to determine the entrance
face. Choosing a square block face as an entrance face, the contact distance value
can be determined by the contact faces, and the open-close states can then further be
determined. For contacts not involving faces, such as edge-edge contacts, the general
approach is to find three points of the edges to construct contact face, then determine
the contact normal vector. The details of contact mode changes procedure are listed
in Table 3.1. N, T is normal and shear forces respectively, and TOL is a prescribed
limit value.
In each time step, choosing lock position requires to solve the global equilibrium
equation iteratively. During the iterative solution process stiff springs are applied and
removed based on the open-close condition, hence it is so-called an ‘open-close iter-
ation’ process. If the normal spring experiences tension at a contact, the spring is
removed; if there are penetrations between contact vertices, the spring is applied at
that point. Each contact pair has three modes: open, sliding and lock, and the mode
changes are summarized in Table 3.1. The open-close iteration procedure continues
until the following two constraints are satisfied in the entire system:
These two conditions must be satisfied in all contacts. If these two conditions are
not satisfied after 6 iterations, the time step size is reduced and the open-close iterations
resume from the end of the previous time step.
kn 2
nc = d (3.59)
2 n
where,
1 x1 + u1 y1 + v1 z1 + w1
1 x2 + u 2 y2 + v2 z2 + w2
=
1 x4 + u 4 y4 + v4 z4 + w4
1 x3 + u 3 y3 + v 3 z3 + w 3
1 x1 y1 z1 1 u1 y1 z1 1 x1 v1 v1
1 x2 y2 z2 1 u2 y2 z2 1 x2 v2 v2
= + +
1 x4 y4 z4 1 u4 y4 z4 1 x4 v4 v4
1 x3 y3 z3 1 u 3 y3 z3 1 x 3 v3 v3
1 x1 y1 w1 1 u1 v1 w1
1 x2 y2 w2 1 u2 v2 w2
+ +
1 x4 y4 w4 1 u4 v4 w4
1 x3 y3 w3 1 u3 v3 w3
Let,
1 x1 y1 z1
1 x2 y2 z2
Vol =
1 x4 y4 z4
1 x3 y3 z3
If the displacements of each block in a time step are sufficiently small, the last
term of involving high-order terms of displacement can be ignored, and the above
Theory of the discontinuous deformation analysis in three dimensions 93
Vol
dn = + EiT di + GjT dj (3.61)
A
where:
⎡ ⎤
1 y2 z2 1 x2 z2 1 x2 y2
1
EiT = ⎣−1 y4 z4 , 1 x4 z4 , 1 x4 y4 ⎦ · TiT (x1 , y1 , z1 )
A 1 y3 z3 1 x3 z3 1 x3 y3
⎡ ⎤
1 y1 z1 1 x1 z1 1 x1 y1
1
GiT = ⎣−1 y4 z4 , −1 x4 z4 , 1 x4 y4 ⎦ · TjT (x2 , y2 , z2 )
A 1 y 3 z3 1 x3 z3 1 x3 y3
⎡ ⎤
1 y1 z1 1 x1 z1 1 x1 y1
1
+ ⎣−1 y2 z2 , 1 x2 z2 , −1 x2 y2 ⎦ · TjT (x4 , y4 , z4 )
A 1 y3 z3 1 x3 z3 1 x3 y3
⎡ ⎤
1 y1 z1 1 x1 z1 1 x1 y1
1 ⎣
+ 1 y2 z2 , 1 x2 z2 , 1 x2 y2 ⎦ · TjT (x3 , y3 , z3 )
A 1 y z 1 x z 1 x y
4 4 4 4 4 4
−−→ −−→ x2 y2 z2
A = p2 p3 × p2 p4 ≈ x3 y3 z3
x4 y4 z4
Therefore, the strain energy of the normal contact spring forces is:
2
kn kn Vol
nc = dn2 = + EiT di + GjT dj (3.62)
2 2 A
94 Discontinuous deformation analysis in rock mechanics practice
∂2 nc
krs = = kn Ei EiT → Kii (3.63)
∂dri ∂dsi
∂2 nc
krs = = kn Ei GjT → Kij (3.64)
∂dri ∂dsj
∂2 nc
krs = = kn Gj EiT → Kji (3.65)
∂drj ∂dsi
∂2 nc
krs = = kn Gj GjT → Kjj (3.66)
∂drj ∂dsj
∂ nc (0) Vol
fr = − = −kn Ei → Fi (3.67)
∂dri A
∂ nc (0) Vol
fr = − = −kn Gj → Fj (3.68)
∂drj A
Shear contact
When the contact vertex and entrance face are in a “no sliding’’ mode the shear spring
is adopted in addition to the normal spring, to obtain a “lock’’ mode. As shown in
Fig. 3.7, assuming the vertex P0 moves to a new point P0 at the end of a time step, L is
Theory of the discontinuous deformation analysis in three dimensions 95
−−→
the projection of the vector P0 P1 on the contact face P2 P3 P4 P5 P6 . The magnitude of
L is given by:
1
−−→2
| L | = ds = P0 P1 − dn2 (3.69)
ks 2 ks −− →2
sc = ds = P0 P1 − dn2
2 2
⎛ ⎞T ⎛ ⎞
x + u1 − x0 − u0 x1 + u 1 − x 0 − u 0
ks ⎝ 1 ks
= y1 + v1 − y0 − v0 ⎠ ⎝ y1 + v1 − y0 − v0 ⎠ − dn2 (3.70)
2 z +w −z −w z +w −z −w 2
1 1 0 0 1 1 0 0
∂2 sc
krs = = ks TiT (x1 , y1 , z1 )Ti (x1 , y1 , z1 ) − ks EiT (x1 , y1 , z1 )Ei (x1 , y1 , z1 ) → Kii
∂dri ∂dsi
(3.72)
∂2 sc
krs = = −ks TiT (x1 , y1 , z1 )Tj (x0 , y0 , z0 ) − ks EiT (x1 , y1 , z1 )Gj (x0 , y0 , z0 ) → Kij
∂dri ∂dsj
(3.73)
∂2 sc
krs = = −ks TjT (x0 , y0 , z0 )Ti (x1 , y1 , z1 ) − ks GjT (x0 , y0 , z0 )Ei (x1 , y1 , z1 ) → Kji
∂drj ∂dsi
(3.74)
∂2 sc
krs = = ks TjT (x0 , y0 , z0 )Tj (x0 , y0 , z0 ) − ks GjT (x0 , y0 , z0 )Gj (x0 , y0 , z0 ) → Kjj
∂drj ∂dsj
(3.75)
For the sliding mode, in addition to the normal spring, a pair of frictional forces instead
of a shear spring needs be added. Based on the Coulomb’s friction law, the frictional
force is:
δτ is the direction of the frictional force. Substituting Equation 3.79 into Equation 3.80,
we get:
⎛ ⎡ ⎤⎞ ⎛ ⎞
n 1 n1 n1 n2 n1 n3 a
δτ = ⎝1 − ⎣n2 n1 n2 n2 n2 n3 ⎦⎠ ⎝b⎠ = r1 i + r2 j + r3 k (3.81)
n3 n1 n3 n2 n3 n3 c
*
Let R = r21 + r22 + r23 , the potential energy of the frictional force Ff at P1 at
block i is:
⎛ ⎞ ⎛ ⎞
Ff r1 Ff T T r1
f = (u1 v1 w1 )⎝r2 ⎠ = di Ti (x1 , y1 , z1 )⎝r2 ⎠ = Ff diT H (3.82)
R r R r
3 3
where:
⎛ ⎞
r1
1 T
H = Ti (x1 , y1 , z1 )⎝r2 ⎠
R r 3
∂ f (0)
fr = − = − Ff H → Fi (3.83)
∂dri
where
⎛ ⎞
r1
1
G = TjT (x0 , y0 , z0 ) ⎝r2 ⎠
R r3
∂ f (0)
fr = − = Ff G → Fj (3.85)
∂drj
One of the advantage of DDA is to use the closed-form simplex integration method
over the blocks. This section will focus on the derivation of analytical formula of the
2D polygon and 3D polyhedron. It is a straightforward extension from the 2D version
of DDA.
Assume the vertex list of the i-th plane loop is arranged anticlockwise as defined
in Shi (1996), as shown in Fig. 3.9:
[i] [i] [i] [i] [i] [i]
P1 P2 P3 . . . Pn(i) Pn(i)+1 = P1
and
[i] [i] [i] [i]
Pj = (xj , yj , zj ) j = 1, 2, . . . , n(i)
Then, total s polygon loops are assumed to compose the current block and set
P0 = (0, 0, 0). Computed by simplex integrations, integrals for 1, x, y, z, x2 , y2 , z2 , xy,
yz, xz are represented by the coordinates of the boundary vertices only.
))) n(i) )
s
S =
1
1dxdydz = 1D(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
s
1 [i] [i]
= xk [i]
yk zk (3.86)
6
i=1 k=1 [i] [i]
x [i]
xk+1 zk+1
k+1
))) n(i) )
s
Sx = xdxdydz = xD(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
1 [i]
s
[i]
= xk yk
[i]
zk (x1 + xk + xk+1 ) (3.87)
24
i=1 k=1 [i] [i]
x [i]
xk+1 z
k+1 k+1
))) n(i) )
s
S =
y
ydxdydz = yD(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
1 [i]
s
[i]
= xk yk
[i]
zk (y1 + yk + yk+1 ) (3.88)
24
i=1 k=1 [i] [i]
x [i]
xk+1 z
k+1 k+1
))) n(i) )
s
Sz = zdxdydz = zD(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
s
1 [i] [i]
= xk yk
[i]
zk (z1 + zk + zk+1 ) (3.89)
24
i=1 k=1 [i] [i]
x [i]
xk+1 zk+1
k+1
100 Discontinuous deformation analysis in rock mechanics practice
))) n(i) )
s
2
Sx = x2 dxdydz = x2 D(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
s
1 [i] [i]
= xk [i]
yk zk
60
i=1 k=1 [i] [i]
x [i]
xk+1 zk+1
k+1
))) n(i) )
s
2
Sy = y2 dxdydz = y2 D(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
1 [i]
s
[i]
= xk [i]
yk zk
60
i=1 k=1 [i] [i]
x [i]
xk+1 z
k+1 k+1
))) n(i) )
s
z2
S = z dxdydz =
2
z2 D(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
s
1 [i] [i]
= xk yk
[i]
zk
60
i=1 k=1 [i] [i]
x [i]
xk+1 zk+1
k+1
))) n(i) )
s
Sxy = xydxdydz = xyD(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
1 [i]
s
[i]
= xk yk
[i]
zk
120
i=1 k=1 [i] [i]
x [i]
xk+1 z
k+1 k+1
))) n(i) )
s
Syz = yzdxdydz = yzD(x, y, z)
[i] [i]
V i=1 k=1 P0 Pk Pk+1
[i] [i] [i]
x
n(i) 1
y1 z1
s
1 [i] [i]
= xk yk
[i]
zk
120
i=1 k=1 [i] [i]
x [i]
xk+1 zk+1
k+1
3.5 SUMMARY
Three-dimensional DDA is still under development mainly due to the complex nature
of the contact detection algorithm, contact status conversion, and convergence effi-
ciency of the open-close iteration process in three dimensions. The basic theory of
two-dimensional DDA was introduced in Chapter 2, and is extended here to three
dimensions. It should therefore be easier to follow the concepts presented in this chapter
after studying Chapter 2. In addition to the complex contact algorithm in 3D DDA,
challenges also exist in conducting failure and multi-physics analysis with 3D DDA;
some recent advances are reviewed in Chapter 1. With the new contact theory recently
proposed by Shi (see Chapter 8), we expect that further development of 3D DDA will
be accelerated and numerical simulations of engineering problems involving highly
fractured rock masses may become possible in the future.
Chapter 4
4.1 INTRODUCTION
We have seen in the previous chapters that DDA is based on rigorous mathematical
principles, but it is important to realize that its realistic application requires good
understanding of the geology of the rock mass that is being modeled numerically.
Engineers that are working on the mathematical and/or computational development
of DDA are not always familiar with the basic geological principles that control rock
mass structure. However, such an understanding is essential for realistic representa-
tion of the rock mass structure in the field by a DDA block system that will produce
meaningful results after forward numerical modeling. The most important thing to
realize at the block construction stage is what should be included, and what could be
safely omitted, when attempting to represent a complex geological structure with sets
of lines that intersect one another to form a discrete element system. This would be
the main purpose of this chapter. We will begin by addressing the issue of scale and the
geometrical parameters that form the DDA block system (number of joint sets, joint
orientation, spacing, and length) and will continue to discuss realistic ranges for input
mechanical parameters of both intact elements (Young’ modulus and Poisson’s ratio)
as well as discontinuities (cohesion, tensile strength, and friction angle). Optimizing
the numerical control parameters, with particular emphasis on time step size and con-
tact spring stiffness, will be discussed in Chapter 5 in the context of benchmarking of
the DDA method.
As briefly discussed in Chapter 1, rock masses are never continuous, at any scale
of reference. Choices must be made, therefore, as to which discontinuities to
represent in the modeled domain and which to omit, because with a finite compu-
tational capability, regardless of the computational platform, a complete numerical
simulation incorporating all discontinuities at all scales is simply not feasible. It
is also not necessary because the solution to most rock engineering problems in
discontinuous rock masses can be obtained by correctly and selectively represent-
ing in the modeled system the relevant discontinuities that control the rock mass
deformation.
104 Discontinuous deformation analysis in rock mechanics practice
Figure 4.1 Scales of interest in underground (left) and slope (right) engineering (Hoek, 2007).
The issue of relative scale is clearly of the essence here. Consider the famous chart
drawn by Professor Evert Hoek (Hoek, 2007) that delineates the scales of reference in
underground or rock slope engineering problems (Figure 4.1).
If the factor of safety against failure in compression in the sidewalls of a deep
underground excavation is of interest, then the microstructure of the rock is important,
because as has been shown by many (e.g. Fredrich et al., 1990; Hatzor and Palchik,
1997; Sammis and Ashby, 1986), microstructure, and particularly the size of initial
flaws such as grain boundaries and pores, controls brittle strength of crystalline rocks
(see Figure 4.2). Moreover, if a fracture evolution study is intended, for example in
order to obtain the extent of the excavation damage zone (EDZ) as controlled by
micromechanical processes (e.g. Cai and Kaiser, 2005), then the micro scale cannot be
ignored.
Indeed, when the level of in situ stress approaches the strength of intact rock
around the excavation it would be prudent to employ continuum-based methods so
that the influence of microstructure on fracture evolution can be modeled accurately.
In most instances, however, the influence of microstructure on intact rock strength
will be studied experimentally in the lab to obtain the ultimate strength of the rock
Geological input parameters for realistic DDA modeling 105
Figure 4.2 Influence of maximum grain size on unconfined compressive strength of dolomites.
and the results will be used in the numerical model as known magnitudes, to enable
computation of factor of safety around the excavation.
Fracture mechanics approaches are appropriate when the rock mass is continuous
at the relevant field scale, namely no discontinuities transect the rock mass to a distance
of several opening diameters in the case of an underground opening or to a distance
into the excavation that roughly equals the slope height, as delineated by the outermost
circles in Figure 4.1. If this indeed is the case, the stability of the project may be
studied using continuum approaches allowing complicated fracture propagation and
coalescence processes in which the microscale is important. If however the relevant
field scale is transected by discontinuities, it would be safe to assume that most of the
deformation, either shear sliding, opening, or rotation, will take place along preexisting
joints because of the much lower shear, cohesive, or tensile strength they possess in
comparison to the intact rock material (see Figure 4.3). In such cases, it would be more
realistic to adopt a discontinuous approach, and the microscale flaws may be safely
ignored.
It is well established that the zone of influence of underground excavations extends
to several opening diameters, as also portrayed in the left panel of Figure 4.1. This
is the zone where stress relaxation takes place followed by loosening and where the
excavation damage zone develops (e.g. Li et al., 2011). With regard to rock loads that
might develop on passive support elements such as steel arches or concrete segments,
Terzaghi (1946) has already predicted, based on many tunneling case studies in the
Alps, that the height of loosening would be roughly one half the excavation span in
106 Discontinuous deformation analysis in rock mechanics practice
Figure 4.3 Comparison between shear strength of intact rock and a bedding plane in the dolomitic
rock mass of Masada.
Figure 4.4 Left – Rockmass consisting of columnar jointed basalt (photo courtesy of Dr. Li Shao-Jun,
Institute of Soil and Rock Mechanics, Chinese Academy of Science); Right – Detail of a DDA
model representing a tunnel in columnar jointed basalt rock mass. The opening span in
the model is 19.7 m, the mean spacing between the lava basal planes is 5 m, and between
the columnar joints is 0.3 m.
blocky rock masses. This prediction by the famous geotechnical engineer proves to
be valid and is confirmed by results of numerical analyses (see Hatzor and Benary,
1998). To demonstrate this, consider a blocky rock mass model created with DDA for
a columnar jointed basalt rock mass based on the geometrical properties of the rock
mass encountered during construction of the Baihetan hydropower station in south
west China (Jiang et al., 2014) as shown in Figure 4.4. The extent of the loosening
Geological input parameters for realistic DDA modeling 107
0.005
0
0 2 4 6 8 10 12 14
–0.005
Displacement (m)
–0.01
–0.015
V1
–0.02
V2
–0.025 V3
V4
–0.03
V5
–0.035
Time (sec)
Figure 4.5 Vertical displacements in the roof of the tunnel shown in Figure 4.4 as computed with DDA.
Point V1 located in immediate roof,V2 @ 2.5 m,V3 @ 9 m,V4 @ 25 m,V5 @ 50 m above
immediate roof.
zone can be inferred from the vertical displacement output obtained from five mea-
surement points that are positioned in the roof (Figure 4.5). All measurement points
up to a vertical distance of 10 meters from the immediate roof output ongoing vertical
displacement, namely the loosening zone in this rock mass, according to DDA predic-
tions, extends to at least half the excavation span which here is 20 meters, as indeed
predicted by Terzaghi for blocky rock masses, on the basis of many field observations.
To be able to generate such a realistic block system that will yield meaningful
numerical results attention must be paid to the correct characterization of the rock mass
structure, with particular emphasis on the number of joint sets, the mean orientation,
spacing, and length of each joint set in the rock mass.
Figure 4.6 Schematic illustration of a plane with the dip and strike vectors. Here the strike trends
North–South and the dip points to the West.
lie on the plane and are perpendicular to one another. The normal to the plane (some-
times referred to as “pole’’) is perpendicular to both the dip and the strike vectors
and can point either upwards or downwards from the plane. By convention, the dip
vector always points downwards (geologists always tend to look down to the ground).
When discussing poles however it is necessary to specify if upward or downward poles
are considered. To provide complete information on the dip that fully characterizes
the plane both the amount of inclination of the dip vector with respect to a horizontal
plane and the direction of the dip vector (the azimuth as measured on a horizontal
plane) must be specified.
The dip and dip direction of each detected joint should be recorded and the poles to
all joints should be plotted using stereographic projection. This way the poles of joints
belonging to the same set will plot in close proximity to one another, thus making it easy
for the geological engineer or the engineering geologist to quickly identify the principal
joint sets that comprise the rock mass structure, the degree of dispersion in each set,
and the mean orientation of each set. Either ‘lower’ or ‘upper’ hemisphere projections
can be used for stereographic projection of poles, although the upper hemisphere (UH)
projection has the advantage that the poles orientation as plotted on the reference circle
has the same attitude as the dip direction. The principles of stereographic projection
as used in rock engineering are reviewed by several authors (Goodman, 1976, 1989;
Priest, 1993) and readers that are not familiar with this useful technique are strongly
advised to consult these excellent sources.
Examples of well-organized rock mass structures as inferred from joint surveys
performed in the field on rock exposures are shown in Figure 4.7 and Figure 4.8.
In Figure 4.7 only the joints (tensile fractures) are plotted, excluding bedding
planes. Quick inspection of Figure 4.7 clearly reveals that the rock mass is transected
by two principal joint sets labeled in the Figure as J2 and J3 . The two joint sets are sub-
vertical with orthogonal strike directions: J2 strikes ESE–WNW and J3 strikes NNE–
SSW. In Figure 4.8 all discontinuities (including joints and bedding planes) measured
in the exposed rock slopes of King Herod’s palace at the northern face of Masada
mountain are plotted using upper hemisphere projection of poles. The principal joint
Geological input parameters for realistic DDA modeling 109
Figure 4.7 Left – Joint survey in Ramleh open pit mine, Right – Principal joint sets in the rock mass
as inferred from lower hemisphere projection of poles (joint survey performed by Dagan
Bakun-Mazor).
Figure 4.8 The rock mass structure at the north face of Masada mountain (left), as represented by upper
hemisphere projection of poles (right). Joint survey performed by Michael Tsesarsky.
sets are again sub vertical striking NNE–SSW (J2 ) and ESE–WNW (J3 ). The bedding
planes dip shallowly to the north.
Figure 4.9 Horizontal bedding planes at the old city walls foundations, Jerusalem.
the co-existence of several sets of discontinuities in any given rock mass is that over its
geological history the rock has been subjected to different in situ stress regimes that have
left their imprint on the rock mass in the form of well-defined sets of discontinuities.
Rock discontinuities typically cluster into sets of common orientation that can be
grouped into several common genetic categories including: bedding planes, joints (or
tensile fractures), shears (or faults), folds, and foliation planes. Below we will briefly
describe each of these sets.
Bedding planes
Sedimentary rocks that form by slow deposition from water bodies will typically
exhibit bedding planes of “infinite’’ extent with respect to the dimensions of the engi-
neering problem at hand (see Figure 4.9). Bedding planes may be filled with infilling
or alteration material (see Figure 4.10) and often exhibit irregular surface geometry.
The roughness of bedding planes, and planes of discontinuities in general, adds to
their shear resistance whereas the infilling material typically decreases their avail-
able shear strength. Because of their large persistence (lateral extent) bedding planes
almost always participate in the formation of rock blocks in a sedimentary rock mass.
Bedding plane orientation is derived from the regional structure and typically dictates
the optimum orientation of the designed engineered structure. Initially bedding planes
will form parallel to the direction of sedimentation and will therefore be horizontal.
Crustal deformation processes however, such as folding and faulting, may lead over
time to inclined bedding planes. An extreme example is the rock salt diapir on the
western margins of the Dead Sea near Sedom (see Figure 4.11), where the initially
horizontal bedding planes are now vertical to sub-vertical as a result of salt intrusion
into the surrounding country rock.
Geological input parameters for realistic DDA modeling 111
Figure 4.11 Vertically dipping bedding planes in the Sedom rock salt wall, Dead Sea.
Joints
The term joints, when not relating to discontinuities in general, typically refers to
tensile fractures formed in the rock mass parallel to the direction of the maximum
112 Discontinuous deformation analysis in rock mechanics practice
Figure 4.12 A) two parallel sub-vertical joints in Yellow Mountain, China; B) A horizontal excavation
induced tensile fracture with characteristic plumose structures. The tensile fracture was
induced by block sliding along a steeply inclined preexisting joint at the bell-shaped caverns
of Bet Guvrin.
With increasing gouge thickness the shear resistance of shears and faults may be
reduced to a point where the available shear strength equals the shear strength of
the infilling material itself, when the infilling thickness approaches or supersedes the
asperity height (Goodman, 1976). The slip striations often observed on shear planes
are assumed to delineate the direction of motion which is typically parallel to the
direction of the maximum shear stress active under the governing paleo tectonic stress
field. Indeed, slip striations along well defined shear planes, have been used to recover
the paleo in-situ tectonic stress tensor, by inversion (Reches, 1987). The relationship
between the statistical characteristics of the roughness profile and slip history is dis-
cussed by Sagy and Brodsky (2009). Examples of slip markings on the planes of a
strike slip and normal faults are shown in Figure 4.13.
Folds
Folds are regional structural features that typically can be traced on geological maps
(see Figure 4.14). Folds often represent ductile deformation of originally deep buried
rocks under high temperature and pressure conditions. At the project scale all discon-
tinuities belonging to a particular fold may have the same orientation, however in the
regional scale the folds may exhibit changing orientation according to the particular
structural configuration (see Figure 4.15).
Figure 4.13 Faults and shears exhibiting distinct slip striations:A) Rough striated strike slip fault surface
in carbonate rocks, SouthernAlps, Italy (Photo by Nicholas van der Elst); B) Smooth normal
fault surface in Andesite rocks, Flowers Pit Fault, Oregon; C) Striated normal fault surface
in Silicate rocks, Dixie valley, Nevada. (photos and captions courtesy of Dr. Amir Sagy,The
Geological Survey of Israel).
Geological input parameters for realistic DDA modeling 115
Figure 4.14 The External Zone of the French Alps. View from Huez to the southwest across the
Romanche glacially-carved valley and Le Bourg d’Osians. Folded and faulted Jurassic and
Triassic sediments overlying Hercynian basement. Deformation in the area is dominated
by polyphase shortening during the Alpine orogeny (photo and caption courtesy of Dr. Itai
Haviv, Ben-Gurion University of the Negev).
Figure 4.15 The Front Ranges of the southern Canadian Rocky Mountains near Banff.View from Sulphur
Mountain towards the southeast. Mount Rundle stretches from left to right at the horizon.
Resistant Paleozoic carbonates are exposed along the ridge crests. Sulphur Mountain
Thrust stretches roughly parallel to the valley. Deformation in the area is characterized by
southwest-dipping thrust sheets and is part of the Laramide orogeny (photo and caption
courtesy of Dr. Itai Haviv, Ben-Gurion University of the Negev).
116 Discontinuous deformation analysis in rock mechanics practice
Figure 4.16 Precambrian mica schist composed of the minerals biotite, quartz, plagioclase and garnet as
exposed in Eilat area, southern Israel. Top: shallow dipping foliations parallel to the layering
of original bedding planes. Bottom: sub-vertical foliations parallel to layering of original
bedding planes (photos and captions courtesy of Prof. Dov Avigad, Hebrew University of
Jerusalem).
Geological input parameters for realistic DDA modeling 117
Figure 4.17 Example of exfoliation planes in Gneiss,Tenaia Lake,Yosemite national park, California.
Figure 4.18 Example of exfoliation planes in Gneiss, Sigiriya (Lion Rock) monument, Sri Lanka.
118 Discontinuous deformation analysis in rock mechanics practice
If the upward joint normal rises at angle δ above the horizontal in direction β
measured counterclockwise from the north, then the direction cosines for the joint
normal are:
l = cos δ cos β
m = cos δ sin β
n = sin δ (4.1)
For a cluster of normals in a given joint set the preferred, or mean, joint set
orientation will trend parallel to the resultant vector of the cluster R, with direction
cosines (lR , mR , nR ):
li
lR =
R
mi
mR =
R
ni
nR = (4.2)
R
1
2
2
2
R = li + mi + ni (4.3)
Geological input parameters for realistic DDA modeling 119
Figure 4.20 A joint set intersected by a sampling line of general orientation (after Priest, 1985).
The angle of rise δ and the direction of the rise β of the resultant normal are
obtained as in Equation 4.1 but with some rules for the correct sign:
With these simple equations, the engineer or engineering geologist can readily
determine the mean joint set orientation in each set based on data from a field survey.
Let us denote the true frequency of the joint set λ and the apparent frequency
measured by the arbitrary scan line direction λs , which is less than or equal to λ.
Let the acute angle between the joint set normal and the scan line be δ. A line of
length l parallel to the joint set normal is expected to intersect a total number of
discontinuities N given by:
N = λl (4.5)
for a sufficiently long value of l. The length of the general scanline with angle δ
to the joint set normal would be l/cos δ in order to intersect the same number of
discontinuities N. Hence the observed joint set frequency is given by:
N
λs = (4.6)
l/cos δ
and since the true frequency is given by λ = N/l we get the famous “Terzaghi’s
correction’’ (Terzaghi, 1965):
λs = λ cos δ (4.7)
Ruth Terzaghi proposed this bias correction originally in 1965 in the context of
drill hole sampling. The meaning of this correction is that the number of disconti-
nuities from a given set, intersected by a sampling line that makes an acute angle δ
to the set normal, reduces with increasing values of δ and approaches zero when δ
approaches 90◦ . We now understand why orientation data from linear sampling lines
may be severely biased.
Following the discussion provided by Priest (1985) we realize that in the general
case there are m joint sets, each of which is assumed to contain parallel joint planes,
with joint set frequency λi and acute angle between ni and the scanline equals δi where
i = 1, 2, 3, . . . , m. Thus using Equation 4.7 above, the frequency of the ith set measured
along the sampling line is:
and the total frequency along the sampling line λs is given by the sum of the frequency
components as follows:
m
λs = λs,i (4.9)
i=1
The total sample size Ns , obtained from a sampling length ls , is given by:
Ns = λs ls (4.10)
The number Ns,i , of discontinuities from the ith set in the sample is given by:
If we examine Equation 4.11 we see that the number of discontinuities in the ith set
in the sample depends in part upon the value of λi and the angle δi (from Equation 4.8).
Geological input parameters for realistic DDA modeling 121
While it is reasonable that the sample size would depend upon the true frequency of
the set λi , it is unreasonable that it should depend upon the arbitrary angle δi . Terzaghi
(1965) suggested that this dependence could lead to errors in interpreting the results of
discontinuity surveys. In the theoretical case where all planes in a joint set are perfectly
parallel this dependence can be removed by dividing Nsi by cos δi to give a weighted
sample size Ni :
Ns,i
Ni = = λi ls (4.12)
cos δi
Since each joint set consists of individual members which are not perfectly parallel
we can apply this weighting factor to each individual joint by using individual weighting
factor 1/cos δi for each mapped joint, where δi is the acute angle between the individual
joint normal and the arbitrary scan line direction. We can thus define a “weighting
factor’’ ω for each member in a set as:
1
ω= (4.13)
cos δ
As we know, the angle θ between two lines is given by:
where α and β are the dip and dip direction of each line, so we can write the weighting
factor ω as:
1
ω= (4.15)
|cos(αn − αs ) cos βn cos βs + sin βn sin βs |
where αs and βs are the trend and plunge of the scanline and αn and βn are the trend
and plunge of the normal to the discontinuity plane. Alternatively, the angle δ can be
measured directly along a great circle common to the joint normal and the scanline
vector using the stereographic projection. A simple procedure to contour the joint
densities on the stereographic projection is proposed by Priest (1985).
N
KF = (4.16)
N − R
the density distribution of which has a bell shaped geometry, the probability that a
normal will make an angle ψ or less with the mean orientation is given by:
1
cos ϕ = 1 + ln(1 − P) (4.17)
KF
thus one can express the spread of values about the mean, the dispersion, corresponding
to any degree of certainty. Fisher has also shown that the standard deviation of the
hemispherical normal distribution is given by:
1
ϕ= √ (4.18)
KF
The length of the resultant R for the un-weighted case is given by Equation 4.3,
and the un-weighted number of discontinuities in a set is N. The Fisher constant can
also be found for a weighted population by application of the weighting factor ω. As
shown by Priest (1985), The weighted number of the population is given by:
N
Nw = ωj (4.19)
J=1
where each of the weighting factors is greater than or equal to 1.0. Consequently Nw
will usually be greater than N, with typical values of Nw /N between 2 and 5. This
artificial increase of sample size is of little concern for contouring as the concentrations
are expressed in percent. However, the value of N is of critical importance when
estimating the precision of the data. Priest (1985) suggested that this obstacle can
be overcome by normalizing (again …) each of the weighting factors ωj so the total
normalized weighted sample size is equal to N:
N
ωj = ωj (4.20)
Nw
so that
N
ωj = N (4.21)
j=1
The ratio N/Nω is a constant and does not change therefore the relative weighting
values. The weighted resultant is given by:
*
Rw = R2x + R2y + R2z (4.22)
where
N
Rx = nj,x
j=1
N
Ry = nj,y (4.23)
j=1
N
Rz = nj,z
j=1
Geological input parameters for realistic DDA modeling 123
and
for upper hemisphere projection. Priest (1985) shows the same method but for lower
hemisphere projection with a different sign convention for the direction cosines.
The original solution by Fisher did not consider the use of weighting factors to
correct for sampling bias and hence implicitly assumed that all weighting values ωj were
equal to unity. In this case, therefore, each normal vector is of unit magnitude and the
total weighted sample size Nw is equal to N and normalization is not required for the
population size. However, normalization for the resultant size must be performed so
that it is never greater than the population size. The weighted Fisher constant is thus
given by (Priest, 1985):
N
KF,w = (4.25)
N − Rw
It should be noted that the error introduced by normalizing the weighting factors
(Equation 4.20) is much less than the error which may be introduced by ignoring the
bias correction all together. Still it would be a good engineering practice to first plot
un-weighted poles, examine the results, and proceed with the bias correction when
necessary.
Figure 4.21 An example of joint set spacing and length distribution from Masada (joint survey
performed by Michael Tsesarsky).
Figure 4.22 Modeling mechanical layering with DDA by controlling input joint length, spacing, and
bridge. A) a trace map with definition of terms, B) the generated DDA block system after
block cutting and dead end trimming (after Bakun-Mazor et al., 2009).
Figure 4.23 DDA block system for the upper terrace of Herod’s palace, Masada, assuming joint spacing
and length distribution as measured in the field (Figure 4.21).
plotted in Figure 4.8B. Inspection of Figure 4.8A clearly reveals that the joints cannot
be assumed infinitely persistent, therefore simulating this rock mass with infinite joint
trace length with respect to the modeled domain assuming zero bridge length would be
totally unrealistic and overly conservative. The results of the joint survey performed
at the site (Figure 4.21) can be utilized to generate a jointing pattern that will reflect
the finite joint persistence, even at the outcrop scale, as demonstrated in Figure 4.23.
However, when attempting to run forward analysis the jointed domain disintegrates
when subjected to the slightest input earthquake vibration, but we know that this slope
has survived several strong earthquake events in its 2000 year history. Although the
slopes geometry has been modified over the years by local block sliding and toppling
126 Discontinuous deformation analysis in rock mechanics practice
Figure 4.24 A) Joint map of the upper terrace of Herod’s palace in Masada as obtained from good
resolution air photos of the outcrop, B) DDA mesh cut from the map shown in (A) after
digitization.
Geological input parameters for realistic DDA modeling 127
Figure 4.25 A) DFN model incorporating mechanical layering effects used for preprocessing, B) result-
ing DDA mesh (after Bakun-Mazor et al., 2009).
from the left and right sides (E-W) respectively, the bulk of the slope remained intact
over the years, suggesting that such a representation of the rock mass, while more
realistic than when using infinite joint traces, is still overly conservative.
When representative surface exposures are available, joint trace mapping using
advanced photogrammetric technics (e.g. Ferrero et al., 2009) may be used to obtain a
digital representation of the rock mass structure. Then using adequate data processing
128 Discontinuous deformation analysis in rock mechanics practice
procedures this information can be used to obtain digital trace maps of the joints on
the exposed outcrops that can be imported into a discrete element package to generate
the modeled discontinuous domain, semi-automatically. Alternatively, or if access is
not possible for photogrammetric measurements, high resolution air photos of the
exposed outcrop may be digitized, either manually or automatically, to obtain the
most accurate joint trace map of the field exposure. A map of the outcrop with its
joints obtained by manually tracing the joints detected from air photos is shown in
Figure 4.24A. This map was digitized manually and the line data were used for input
in DDA line generation code (DL). Then the robust DDA block cutting code (DC) was
used to generate the block system shown in Figure 4.24B.
The advantage of the block system shown in Figure 4.24B is that it includes irreg-
ular block shapes that are interlocked to form a more cohesive rock mass structure.
No automatic procedure can produce such a result. The engineering geologist, or the
rock engineer, is encouraged therefore to go out to the field, measure the joints, and
characterize the rock mass structure as accurately as possible, so that the basic block
system used for forward analysis will bear close resemblance to the actual situation in
the field.
When the statistical distributions of the jointing parameters are well known, and
are too complicated to simulate with the basic options available in the DL code, existing
tools such as DFN (Dershowitz et al., 2000; Dershowitz and Einstein, 1988) may be
used for preprocessing and then the resulting lines can be input directly into DDA’s
block cutting code DC. Such an approach has been used to simulate mechanical layering
effects in the rock mass shown in Figure 4.7 (See Figure 4.25). This effort enabled
accurate assessment of roof stability of natural karstic caverns discovered below an
active open pit mine that needed to be preserved for future scientific investigations (see
Bakun-Mazor et al., 2009).
Finally, statistical joint generation considerations pertinent to three dimensional
DDA are discussed by Shi (2005) and Ma and Fu (2011).
The input parameters for intact rock elements should be easy to obtain from pre-
liminary site investigations and standard laboratory tests performed on representative
cores from the site. Some points to remembered when assigning intact rock parameters:
1 The introduced values should be representative of the rock in the field, therefore the
engineer or engineering geologist should select samples for testing at the laboratory
wisely so as not to introduce bias in the sampling and testing results.
2 Running robust simulations with DDA of a real engineering problem without first
obtaining good knowledge of the actual material properties in the field is not
recommended as material properties of rocks are very much site specific.
3 When running sensitivity analyses for gaining some deeper theoretical understand-
ing of the studied problem, material properties can be obtained from publications
or data bases.
4 The units for the solution are determined by the input values of the physical and
mechanical parameters, namely density and Young’s modulus, therefore they must
be consistent. For example, if a solution in SI units is desired, then the unit weight
of the elements should be in N/m3 and the unit mass will be the assigned unit weight
divided by gravitational acceleration in units of m/s2 . For consistency therefore,
the Young’s modulus must be in this case in units of N/m2 . Naturally, if unit weight
is input in units of kN/m3 then Young’s modulus should be input in units of kPa.
Any other internally consistent unit system can be used. The output values of the
analysis will be in the same units.
5 As will be explored in Chapter 5, for every problem there is optimal relationship
between Young’s modulus and the contact spring stiffness, or penalty parameter.
The issue of what is the optimal penalty value to insert is a subject of ongoing
research as the accuracy of the solution very much depends on it, particularly
in dynamic problems, as will be explored in Chapter 5. In the absence of any
knowledge or previous experience Shi’s rule of thumb (Shi, 1996) may be used:
G0 = (E0 )(L0 ) where G0 is the stiffness of the contact spring, E0 is Young’s modulus
of the block, and L0 is the “average’’ block diameter. Note that by using this rule
of thumb G0 becomes proportional to the size of the block or to the contact force,
as it should. While the value of G0 may be optimized to obtain best numerical
accuracy, Young’s modulus should not be tweaked at all and must reflect the best
representation of the physical reality in the field.
6 When there are reasons to believe that the rock mass stiffness is significantly
different than that of intact rock elements it is possible to introduce rock mass
modulus to the intact rock elements instead of values obtained from laboratory
tests performed on intact element. Many empirical relationships between rock
mass classifications and rock mass stiffness exist in the literature (e.g. Hoek and
Diederichs, 2006), and these may be used for preliminary purposes. This approach
could be effective also when attempting to model deformation in rock masses with
filled joints, as joint infilling deformation is not incorporated in DDA other than
the linear elastic deformation of the contact springs.
DDA verification
5.1 INTRODUCTION
Any numerical method, regardless of the underlying theory and robustness of the
mathematical formulation, requires verification and validation once it is implemented
into computer code. By verification we mean ensuring that there is an acceptable
agreement between results obtained by running the code with the computer and by
solving a closed form analytical solution of exactly the same problem. Of course if we
could solve everything analytically there would not be a need for us to use numerical
approaches. But we cannot, consider tunneling for example: once the immediate roof
of a tunnel consists of more than three blocks, there is no analytical solution we can
employ to solve the forces between the blocks and to determine the most likely mode
of failure, not to mention the time dependent displacement of each and every block in
the system. Similarly in a rock slope problem: when the number of interacting blocks
exceeds more than two, application of analytical approaches becomes very difficult.
Once the number of blocks in a slope or around a tunnel increases to realistic numbers
of 101 to 102 , depending on the intensity of the jointing pattern in the rock mass, we
become completely dependent on numerical approaches. Indeed, we will never know
if our numerical solution is correct when the problem we are solving does not have an
analytical solution. What we can do is attempt to determine if our numerical solution
is valid. We do this by comparing results obtained numerically to results obtained from
field or laboratory experiments. Consider the previous tunnel roof example: although
there is no analytical solution for say the vertical displacement of the central block in
a roof consisting of three blocks, we can measure that displacement over time, if we
can construct a similar model in the lab or if we can find such a structure in the field
and monitor it properly. Then we can go back to the numerical prediction and test its
validity. Such an approach is called validation.
Verifications and validations are extremely important for establishing the applica-
bility of a numerical approach to real life problems. They allow us a deeper insight not
only to the accuracy of the method used, but also to the sensitivity of the numerical
solution to the numerical control parameters the code needs in order to run properly.
Verifications are also useful as benchmark tests. They are a proving method for any
numerical code, so when we develop a new code or modify an existing code, before we
attempt to solve real life problems with it, we must prove that it can pass those bench-
mark tests satisfactorily. A comprehensive review of DDA validations summarizing
132 Discontinuous deformation analysis in rock mechanics practice
Figure 5.1 The plane failure of Mt. Toc in the southern Italian Alps. A mass of about 270–300 million
m3 of rocks slid in 1963 on a preexisting plane of weakness into the Vajont reservoir at a
velocity sufficiently high to cause a wave that overtopped the Vajont dam, resulting in more
than 2000 causalities and devastating the downstream town of Longarone, Italy.
work that has been done a decade after its publication was published by MacLaughlin
and Doolin (2006). A set of DDA benchmark tests was published a decade later by
Yagoda Biran and Hatzor (2016), and in between many more excellent work has been
done in an attempt to verify and validate DDA (e.g. MacLaughlin and Hayes, 2005;
Scheele and Bates, 2005; Wu et al., 2007). In this chapter verifications and validation
studies performed by the DDA research group at Ben-Gurion University of the Negev
are discussed.
The most common problem in rock slope engineering is that of a block resting on an
inclined plane. The famous landslide of Mount Toc in the Italian Alps that overtopped
the Vajont dam (see Barla and Paronuzzi, 2013) is agreed by all to have originated in
a single plane sliding mechanism, the sliding surface of which is shown in Figure 5.1.
Single plane sliding is typically encountered in rock masses of sedimentary origin,
where bedding planes of infinite extent transect the rock mass structure. When the
bedding planes are horizontal or dipping gently we may assume there is sufficient
frictional resistance to restrain any possible block motion along those weakness planes,
in static conditions. If the bedding planes are smooth, clean and tight the friction angle
may be assumed to be at least as high as 20 degrees, the residual friction angle of most
carbonate and silicate minerals being typically higher than this value. If the surfaces
DDA verification 133
Figure 5.2 2D-DDA verification for dynamic sliding of a block on an inclined plane with constant friction
under gravitational loading (modified after Kamai (2006) following MacLaughlin (1997)). The
modeled plane inclination is 28◦ .
are rough, the friction angle can increase quite significantly, up to twice or three
times this value, but if the bedding planes are filled with clay infilling thicker than the
roughness amplitude, the friction angle can be as low as half that value, depending on
the mineralogical composition of the clay and its water content. So clearly, in addition
to measuring the slope inclination properly in the field, we must have a good estimate
of the frictional resistance of the sliding planes, before we begin the analysis. The
possible range of friction angles in rock discontinuities has been discussed by many
authors (e.g. Barton, 1976; Brady and Brown, 2004; Goodman, 1989; Hoek and Bray,
1981; Jaeger et al., 2007).
5.2.1 2D-DDA
The first to use the block on an incline problem to verify dynamic sliding under con-
stant gravitational acceleration with 2D-DDA was MacLaughlin (1997) in her Ph.D.
dissertation. This has since become a popular bench mark test for DDA because the
formulation of the limit equilibrium equation as originally proposed by MacLaughlin
(1997) is straight forward. The displacement s at time t of a block resting on a
plane inclined at angle α with friction angle φ and subjected to constant gravitational
acceleration g is given by Equation 5.1 (MacLaughlin, 1997):
1 1
st = at 2 = g(sin α − cos α tan φ)t 2 (5.1)
2 2
A comparison between the analytical solution (Equation 5.1) and 2D-DDA as com-
puted by Kamai (2006) in her M.Sc. thesis following the original work of MacLaughlin
134 Discontinuous deformation analysis in rock mechanics practice
(1997) is shown in Figure 5.2 for a range of friction angles, between 5 and 25 degrees,
and slope inclination of 28 degrees. Inspection of Figure 5.2 immediately reveals strik-
ing agreement between the analytical and numerical solutions. This benchmark test,
originally proposed by MacLaughlin (1997), proves therefore that dynamic block slid-
ing along a single plane can be computed with great accuracy with DDA under constant
friction and constant acceleration.
What happens when the acceleration is not constant with time, for example in
the case of a strong earthquake? This problem has been studied independently by both
Newmark (1965) and Goodman and Seed (1966), although the approach is universally
referred to as “Newmark’s method’’ today. Both Newmark (1965) and Goodman and
Seed (1966) showed that the down slope, horizontal, yield acceleration ay for a block
resting on a plane with inclination α and friction angle φ (cohesion is ignored) is
given by:
where g is the gravitational acceleration. The “yield acceleration’’ is, as the name
implies, the acceleration at which sliding will ensue. Consider now a harmonic
acceleration function at of the form:
at = A sin(ωt + θ) (5.3)
in which A and ω are the amplitude and angular frequency of the input harmonic
acceleration function. θ is the phase angle required to satisfy the initial condition
a = ay at the instant sliding begins (t = 0) and is found by solving Equations 5.2 and
5.3 simultaneously:
2 2
sin−1 tan(φ − α) g A sin−1 ay A
θ= = (5.4)
ω ω
The displacement of the block can be determined by integrating Equation 5.3
twice whenever at > ay using the definition of θ; the first integration provides the time
dependent velocity vt and the second the time dependent displacement dt . A graphical
illustration of this concept is shown in Figure 5.3, following the original paper by
Goodman and Seed (1966).
The first who have attempted to verify DDA using Goodman and Seed’s approach
were Hatzor and Feintuch (2001). Note however that in the original paper by Hatzor
and Feintuch the frictional resistance of the sliding interface was used to evaluate the
yield acceleration ay (Equation 5.2) and therefore θ (Equation 5.4), but frictional resis-
tance was neglected once dynamic sliding ensued, obviously resulting in less than per-
fect agreement between the analytical and DDA solutions. Kamai and Hatzor (2008)
corrected this oversight and included the frictional resistance of the sliding interface
during dynamic sliding in their DDA verification for the same problem (see Figure 5.4).
The down slope acceleration of the sliding block can be determined by subtracting
the resisting forces from the driving forces:
, - , -
at = kg sin(ωt) cos α + g sin α − g cos α − kg sin(wt) sin α tan φ (5.5)
DDA verification 135
Acceleration
ky
1 ky2
ky3
t
t1 t2
velocity
t
t1 t3
Displacement
t
t1 t3
Figure 5.3 Method originally proposed by Goodman and Seed (1966) for integration of accelerograms
to determine downslope displacement of a block resting on an inclined plane. Dynamic
friction degradation is assumed, as implied by the decreasing value of yield acceleration
between cycles. Reproduced with permission from ASCE.
Figure 5.4 Free body diagram for the block on an incline problem subjected to harmonic horizontal
acceleration loading function (Kamai, 2006).
136 Discontinuous deformation analysis in rock mechanics practice
Figure 5.5 2D-DDA verification for single block on an inclined plane subjected to harmonic accelera-
tion input function (after Kamai and Hatzor (2008)). The modeled plane inclination here is
20 degrees.
where k calibrates the proportion between a and g. As explained above, the displace-
ment of the block at any time t is determined by double integration on the acceleration,
with θ as reference datum, in a conditional manner:
if a > ay or v > 0:
)t )) /
22 0
dt = v= a = g (sin α − cos α tan φ) t 2 − θt
θ θ
A
+ 2 [(cos α + sin α tan φ) (ω cos(ωθ) (t − θ) − sin(ωt) + sin(ωθ))] (5.6)
ω
otherwise dt = dt−1
where the initial velocity and displacement of the sliding block are assumed to be zero.
The downslope displacements, d(t), are calculated while ay is exceeded for the first
time at θ1 , or when the block’s velocity is positive. If neither condition is fulfilled the
block is assumed to be at rest, and sliding will commence only once ay is exceeded again
at θ2 , and so on. The excellent agreement between DDA and the analytical solution
(Equation 5.6) for three different friction angles is shown in Figure 5.5. The relative
DDA verification 137
Figure 5.6 Three dimensional DDA mesh for the block on an incline problem, the direction of loading
is indicated in the inset (Yagoda-Biran, 2013).
error (EN ) between the analytical and numerical solutions (Equation 5.7) proves to be
less than 1% in this verification.
dA − d N
EN = · 100% (5.7)
dA
5.2.2 3D-DDA
We have seen in Chapter 3 that in 3D-DDA the number and type of contacts are
significantly greater than in 2D-DDA and naturally this might affect the accuracy of
the dynamic solution. To verify the accuracy of 3D-DDA when solving dynamic sliding
of a single block on a single plane we use the same block on an incline problem as
discussed in the previous section but this time in three dimensions, following work
performed by Yagoda-Biran (2013) in her Ph.D. dissertation. The three dimensional
mesh as computed by Yagoda-Biran is shown in Figure 5.6. The plane inclination is
45◦ , its height 10 m, and its depth 5 m. The base block is fixed in space by imposing
seven fixed points, and the sliding block is a 1 m × 1 m × 0.5 m box.
The general expression for time dependent displacement under harmonic acceler-
ation input when the initial velocity and displacement are not zero is (Yagoda-Biran,
2013):
1
d(t) = g(sin α − cos α tan φ)t 2
2
A
− sin(ωt) (cos α + sin α tan φ) + ḋ0 t + d0 (5.8)
ω2
138 Discontinuous deformation analysis in rock mechanics practice
Figure 5.7 Verification of 3D-DDA for dynamic sliding of a block subjected to gravitational loading
(Yagoda-Biran, 2013).
Parameter value
dd – dynamic parameter 1
g0 – normal contact spring stiffness 4 * 108 N/m
g1 – time step interval 0.001 s
g2 – maximum displacement ratio 0.001
Block material density 2700 kg/m3
Young’s modulus 40 GPa
Poisson’s ratio 0.18
When the block is subjected to gravity loading only (A = 0), and begins at rest (ḋ0 = 0,
d0 = 0), Equation 5.8 is reduced to Equation 5.1. The analytical vs. 3D-DDA downs-
lope displacement history as computed by Yagoda-Biran (2013) for three values of
friction angle is shown in Figure 5.7. Note that all three friction angles are smaller
than the inclination of the slope and therefore sliding begins immediately the instant
gravity is applied. After 0.2 s of sliding the relative error as defined by Equation 5.7
drops to below 1%, indicating the excellent agreement between the analytical and 3D-
DDA solutions. The input numerical control parameters used by Yagoda-Biran (2013)
are listed in Table 5.1.
While it is encouraging to learn that dynamic sliding of a single block can be studied
very accurately with 3D-DDA when the sliding block is initially at rest, it would be
interesting to check if the same agreement could be obtained when the sliding block
has some initial velocity. This would be particularly useful for landslide studies where
the sliding mass is assumed to be creeping downslope at some very low initial velocity,
DDA verification 139
Figure 5.8 a) Downslope displacement histories of a block on an inclined plane subjected to gravity
and different initial velocities, b) The relative numerical error (Yagoda-Biran, 2013).
before some physical or mechanical barrier is removed, thus allowing the sliding mass
to run out. For the analytical solution we can use again Equation 5.8 but this time
with consideration of the initial velocity, namely ḋ0 = 0. The agreement between the
analytical and 3D-DDA solutions for initial velocities of 0.01 m/s, 0.1 m/s, and 1 m/s
as computed by Yagoda-Biran (2013) is shown in Figure 5.8. The numerical control
parameters remain as in Table 5.1, except for the time step interval that had to be
reduced by one order of magnitude to 0.0001 s in order to obtain an agreement within
1% with the analytical solution for all initial velocities. This level of accuracy is attained
after 0.5 s of sliding.
After verifying that dynamic sliding under gravitational loading can be performed
extremely accurately with 3D-DDA when only a single block is considered, it would be
instructive to examine the accuracy of the 3D-DDA solution for dynamic sliding under
harmonic acceleration input. Yagoda-Biran (2013) subjected the block model shown
in Figure 5.6 to a sinusoidal input acceleration function with amplitude of A = 2 m/s2
and frequency of 1 Hz. A friction angle of 50◦ was assigned to the interface which
was inclined at 45◦ , so that the input acceleration must have exceeded the theoretical
yield acceleration before sliding began. The numerical control parameters were the
same as listed in Table 5.1 with two exceptions: 1) the time step interval was further
reduced by one order of magnitude to 0.0001 s as in the previous example, and 2) the
normal contact spring stiffness was increased by more than one order of magnitude to
7 ∗ 109 N/m. The results of this verification are shown in Figure 5.9 where throughout
the analysis the relative error never exceeds 3%.
So far in our discussion we have limited the input motion direction to be parallel
to one of the axes of the coordinate system. But in a three dimensional approach we
need not do that, indeed there is no reason for the dynamic input to be restricted in
any way to a single direction as ground motions in all directions are recorded during
140 Discontinuous deformation analysis in rock mechanics practice
Figure 5.10 Three dimensional force representation for the block on an incline problem (Bakun-Mazor,
2011).
nx = sin(α) sin(β)
ny = sin(α) cos(β)
nz = cos(α) (5.9)
A block of a unit mass is assumed, therefore the force equilibrium equations pre-
sented below can be discussed in terms of accelerations. Let the resultant force vector
DDA verification 141
that acts on the system at each time-step be: r = [rx , ry , rz ]. The driving force vector
that acts on the block (m), namely the projection of the resultant force vector on the
sliding plane at each time step, is:
m = (n̂ × r) × n̂ (5.10)
The normal force vector that acts on the block at each time step is:
At the beginning of a time step, if the velocity of the block is zero then the resisting
force vector due to the interface friction angle φ is:
− tan(φ)pm̂ , tan(φ)p < |m|
f= (5.12)
−m , else
where m̂ is a unit vector in direction m. If at the beginning of a time step the velocity
of the block is not zero, then:
f = − tan(φ)pv̂ (5.13)
s=m + f (5.14)
1
Si = (si−1 + si ) (5.15)
2
V i = V i−1 + Si t (5.16)
1
Di = Di−1 + V i−1 t + Si t 2 (5.17)
2
142 Discontinuous deformation analysis in rock mechanics practice
Figure 5.11 a) Comparison between Newmark solution,Vector Analysis, and 3D-DDA for horizontal
motion parallel to X axis for the model shown in Figure 5.10. The relative errors are
computed against Newmark’s solution, b) Comparison between VA and 3D-DDA for
horizontal input motion parallel to both X and Y axes simultaneously. Here VA is used as
a reference for relative error computation (after Bakun-Mazor, 2011).
Table 5.2 Input parameters for 3D-DDA verification performed by Bakun-Mazor (2011).
Mechanical Properties:
Elastic Modulus, MPa 20 200000
Poisson’s Ratio 0.25 0.25
Density, kg/m3 1000 1700
Friction angle, Degrees 30 30–36
Numerical Parameters:
Dynamic control parameter 1 1
Number of time steps 628 8000
Time interval, Sec 0.01 0.005
Assumed max. disp. Ratio, m 0.01 0.01
Penalty stiffness, MN/m 10 10000–20000
Once verified, Bakun-Mazor’s VA approach can now be used to verify 3D-DDA for
single plane sliding but under input acceleration in both the x and y directions as well
as vertical gravitational acceleration (Figure 5.11b). The resultant input acceleration
vector here is r = [rx ry rz ] = [0.5 sin(10t), 0.5 sin(5t), −1]g, and the friction angle is
again φ = 30◦ . The two components of the input horizontal acceleration are plotted
as shaded lines and the acceleration values are shown on the right-hand side axis as
before. The relative error in the final position of the sliding block does not exceed 8%.
The input values used by Bakun-Mazor (2011) for the verification are listed in
Table 5.2.
This mode of failure is commonly referred to as “Wedge Failure’’ and has been discussed
by many authors in the past. Three dimensional limit equilibrium analysis for this
failure mode using the stereographic projection is provided by Goodman (1976, 1989)
for dry joints and by Londe et al. (1969, 1970) for water filled joints. An analytical
procedure to solve the factor of safety is discussed by Hoek and Bray (1981).
Wedge failures are encountered in rock masses consisting of more than one joint
set, as the intersection of two sets are necessary to produce three-dimensional wedges
in the rock mass. If there are more than two joint sets in the rock mass different
wedges may form, but the number of failure modes will not increase with increasing
number of joint sets in the rock mass. In fact, if rotations are ignored the only possible
failure modes are opening, single, and double plane sliding. Moreover, in order to
create a finite block a minimum of three joint sets is required if only one free surface
is considered (Goodman and Shi, 1985). As can be seen in Figure 5.12A, the “block
mold’’ (Hatzor, 1993) that remained in the rock slope after the failure of the wedge
has three joints in its boundaries, two steeply dipping and one gently dipping. The
wedge slides along the line of intersection of the two steeply inclined joints and opens
from the moderately dipping joint at the top of the block. The same can be seen also
in Figure 5.12B, but this time the boundary planes along which sliding took place are
dipping more moderately. But again, sliding only takes place on two joints in direction
144 Discontinuous deformation analysis in rock mechanics practice
Figure 5.12 Examples of wedge failures in the field. A) a wedge in a Gypsum quarry at the
Ramon crater (southern Israel), B) a wedge in the slopes of Jinping mountain, Sichuan
province, China.
Figure 5.13 Three dimensional representation of the wedge problem (Bakun-Mazor, 2011).
parallel to the line of intersection of the two boundary planes, and the wedge opens
from the third joint at the top.
The case of wedge failure has been used by Bakun-Mazor (2011) to verify
3D-DDA. A three dimensional model of a wedge that is sliding simultaneously on
two joints, plane 1 and plane 2, in direction parallel to the line of intersection (I1,2 ) is
shown in Figure 5.13.
The normal to plane 1 is n̂1 = [nx1 , ny1 , nz1 ] and the normal to plane 2 is n̂2 =
[nx2 , ny2 , nz2 ]. The line of intersection Î12 , along which sliding takes place in this failure
mode is given by:
As discussed in the case of plane failure in three dimensions, the resultant force in
each time step is r = [rx , ry , rz ], and the driving force in each time step is now:
The normal force acting on plane 1 in each time step is p = [px , py , pz ], and the
normal force acting on plane 2 in each time step is q = [qx , qy , qz ], where:
As in the case of single face sliding, the direction of the resisting force (f ) depends
upon the direction of the velocity of the block. Therefore, as before, in each time step:
⎧
⎪
⎪ − tan(φ1 )p + tan(φ2 )q m̂ , V = 0 and tan(φ1 )p + tan(φ2 )q < |m|
⎨
f = −m , V = 0 and tan(φ1 )p + tan(φ2 )q ≥ |m|
⎪
⎪
⎩
− tan(φ1 )p + tan(φ2 )q v̂ , V = 0
(5.22)
where V = [Vx , Vy , Vz ] is the velocity vector. The sliding force, namely the block
acceleration during each time step, is s = [sx sy sz ] and is calculated as the force balance
between the driving and the frictional resisting forces:
Si = m + f (5.23)
1
Si = (si−1 + si ) (5.24)
2
The velocity for time step i is therefore:
V i = V i−1 + Si t (5.25)
1
Di = Di−1 + V i−1 t + Si t 2 (5.26)
2
Sensitivity analyses were performed by Bakun-Mazor (2011) to discover the max-
imum value of the time interval for the trapezoidal integration method without
compromising accuracy, where the relative error was defined as:
DVectorAnalysis − D3D DDA
Erel = · 100% (5.27)
DVectorAnalysis
146 Discontinuous deformation analysis in rock mechanics practice
Figure 5.14 3D-DDA vs. analytical solution for dynamic wedge sliding. a) Wedge response to one
component of horizontal sinusoidal input motion and self-weight, lower panel presents
the relative error calculated according to Equation 5.27. b) Wedge response to 3D loading
using data from the ImperialValley earthquake where the three components, multiplied by
a factor of 5, are shown in the lower panel (after Bakun-Mazor et al., 2009).
It was found that the results are sensitive to the time interval size when the friction
angle is greater than the slope inclination, and that the upper limit for time interval
for acceptable accuracy was 0.001 s. The obtained accuracy with 3D-DDA for this
problem can be appreciated from Figure 5.14. The input values used for the analysis
are listed in Table 5.2.
In all verifications until now, we have introduced the force directly to the center of
mass of the sliding block. During strong earthquakes however, the ground is shaking
and the motion is transmitted to the overlying blocks through the frictional interface
between the block and the ground or between one block to another (see Figure 5.15).
Therefore, a study of block response to shaking foundation along a frictional interface
is very relevant for geotechnical earthquake engineering. Verification of this failure
mode was originally done with 2D-DDA by Kamai (2006) who was the first to propose
a semi-analytical solution to this problem that enabled verification, which was later
expanded to three dimension by Yagoda-Biran (2013), for verifying 3D-DDA. Other
workers have also studied this problem following the original work of Kamai (2006)
and explored various aspects of it with regards to DDA (e.g. Akao et al., 2007; Sasaki
et al., 2007).
DDA verification 147
Figure 5.15 Column drum response to shaking foundations at the Acropolis, Athens.
Figure 5.16 2D-DDA model used by Kamai (2006) to study block response to cyclic motion of frictional
interface.
5.4.1 2D-DDA
The studied block system consists of three blocks: a fixed foundation block (Block 0),
the intermediate block subjected to horizontal cyclic displacement (Block 1), and the
overlying responding block (Block 2). In order to avoid rotational movements that
will complicate the analytical solution the responding block is made sufficiently flat,
as shown in Figure 5.16.
Block 1 is subjected to a horizontal displacement input function in the form of a
cosine, starting from 0:
d(t) = D 1 − cos 2πft (5.28)
where D and f are the amplitude and frequency of motion, respectively. The only
force acting on Block 2 other than gravity is the frictional force, which immediately
148 Discontinuous deformation analysis in rock mechanics practice
m2 a2 = Ffriction
m2 a2 = µm2 g (5.29)
a2 = µg
where µ is the friction coefficient. The direction of the driving force is determined by
the direction of the relative velocity between Blocks 1 and 2 (v1∗ ). When Block 1 moves
to the right relative to Block 2 (positive x direction here), the frictional force pulls
Block 2 in the same direction, and determines the sign of a2 . When Block 2 is at rest
in relation to the Block 1, the frictional force is determined by the acceleration of the
bottom block (a1 ). The threshold acceleration, under which the two blocks move in har-
mony, is equal to the friction coefficient multiplied by the gravitation acceleration (µg).
When the acceleration of Block 1 passes this threshold value, the frictional forces act
in the same direction as a1 . The relative velocity of Block 1 is given by:
v1∗ = v1 − v2 (5.30)
In the solution of Equation 5.31 Kamai (2006) employed Matlab (MATLAB, ver-
sion 7) software package because the analytical solution must be computed iteratively
as the relative velocity and the direction of the force are dependent upon one another.
The accumulated displacement of Block 2 in response to the cyclic shaking of Block 1
as computed by Kamai (2006) using the analytical approach and 2D-DDA is shown in
Figure 5.17. Three different displacement amplitudes (D) are modeled with a constant
input frequency of 1 Hz and friction coefficient of 0.6 at the sliding interface. The
obtained magnitude of accumulated displacement is directly proportional to the input
amplitude, as expected. Note that the three displacement curves follow the periodic
behavior of the input displacement function (T = 1 sec.), and that divergence between
curves starts after 0.25 s where the input displacement function has an inflection point.
The relative error is mostly between 1% and 2%.
5.4.2 3D-DDA
Yagoda-Biran (2013) expanded Kamai’s approach into three dimension in order to
verify 3D-DDA using the model shown in Figure 5.18. The analytical approach is
similar to the one-dimensional approach originally proposed by Kamai (2006) except
that here displacements, velocities and accelerations are vectors.
DDA verification 149
Figure 5.17 2D-DDA verification of block response to sliding interface problem (after Kamai, 2006).
Figure 5.18 3D-DDA model for sliding interface used by Yagoda-Biran (2013).
Each of the two moving blocks, blocks 1 and 2, has time dependent displacements
d(t), velocities ḋ(t) and accelerations d̈(t). The displacement induced to block 1, d1 , is
in the form of a cosine function similar to Equation 5.28:
d1 (t) = A 1 − cos 2πf t (5.32)
here A and f are the amplitude and frequency of motion, respectively. The acceleration
of block 2 isd̈2 = µ∗ g, as before. The direction of the frictional force, and therefore
of d̈2 , is determined by the direction of the relative velocity between thetwo blocks,
∗ ˆ ∗
ḋ ≡ ḋ1 − ḋ2 , defined by the unit vector of the relative velocity, ḋ∗ . When ḋ = 0, the
acceleration of block 2 (d̈2 ) is determined by the acceleration of block 1 (d̈1 ). When
the acceleration of block 1 exceeds the yield acceleration µ∗ g, over which block 2 no
longer moves in harmony with block 1, the frictional force direction is determined
150 Discontinuous deformation analysis in rock mechanics practice
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
time (sec)
Figure 5.19 Comparison between the semi-analytical (lines) and 3D-DDA (symbols) solutions for
changing motion amplitude. Input motion frequency 1 Hz, friction coefficient of sliding
interface 0.6 (Yagoda-Biran, 2013).
ˆ
by the direction of ḋ∗ , but the magnitude of d̈2 is equal to µ∗ g. Yagoda-Biran (2013)
formulated this rationale as follows:
∗
if ḋ = 0 and d̈2 ≤ µ∗ g then d̈2 = d̈1
ˆ
and d̈2 > µ∗ g then d̈2 = µ∗ g · d̈1 (5.33)
∗ ˆ
if ḋ = 0 then d̈2 = µ∗ g · ḋ∗
The results of the validation performed by Yagoda-Biran (2013) for changing input
amplitude and input friction are shown in Figure 5.19 and Figure 5.20, respectively
and the numerical input parameters are listed in Table 5.3. The obtained agreement
between the semi-analytical and numerical solutions is excellent, with the relative error
remaining most of the time less than 1%.
The above verification is in fact the same as performed by Kamai (2006) as in
both cases the input motion is constrained to one direction, namely a one dimensional
solution. Yagoda-Biran (2013) added another motion component into block 1 paral-
lel to the y-axis (see Figure 5.18) with different amplitude and frequency to enable a
two-dimensional verification. The result for three different sets of x and y motions for
DDA verification 151
Figure 5.20 Comparison between the semi-analytical (lines) and 3D-DDA (symbols) solutions for
changing interface friction. Input motion frequency of 1 Hz and amplitude of 0.5 remained
constant (Yagoda-Biran, 2013).
Parameter Value
interface friction of 0.6 is shown in Figure 5.21, indicating good agreement, with the
relative error remaining less than 1% throughout most of the analysis. The displace-
ments evolution in both the x and y directions with time for a single set of motions
(Ax = 0.3 m, fx = 2 Hz; Ay = 0.2 m, fy = 4 Hz) is shown in Figure 5.22.
Finally, to perform a true three dimensional verification Yagoda-Biran (2013)
added a third motion component parallel to the z direction. Adding time dependent
displacement in the z direction affects the response of block 2 as it changes the normal
force between the two blocks, and therefore the frictional force between them. This
in turn changes the acceleration of block 2 (d̈2 ) and consequently yields a different
152 Discontinuous deformation analysis in rock mechanics practice
Figure 5.21 Comparison between analytical (curves) and 3D-DDA (symbols) solutions. Each set of
curves and symbols corresponds to a different set of amplitude and frequency for the
input displacements, noted beside the data (Yagoda-Biran, 2013).
Figure 5.22 Displacement evolution with time for x amplitude and frequency of 0.3 m and 2 Hz and y
amplitude and frequency of 0.2 m and 4 Hz, respectively (Yagoda-Biran, 2013).
DDA verification 153
Figure 5.23 Comparison between semi-analytical (heavy line) and 3D-DDA (light lines) solutions
for three-dimensional input motion for five different values of contact spring stiffness
(Yagoda-Biran, 2013).
where î, ĵ, k̂ are unit vectors parallel to the x, y, and z axes, respectively. The results of
the verification study with three components of induced displacements are shown in
Figure 5.23 for five different values of normal contact spring stiffness (k).
The range of stiffness values that best fits the analytical solution is between 1 ∗ 107
and 1 ∗ 109 N/m, with stiffness of k = 1 ∗ 107 N/m, or 0.0003 E ∗ L, being the optimal
selection, where E is the Young’s modulus of the block and L is the length of the line
across which the contact springs are attached. When considering 3D-DDA, it might
be more relevant to compare k to E ∗ A, where A is the area across which the contact
springs are attached. In this case, k = 1 ∗ 107 is ∼0.0001 E ∗ A, not much different from
E ∗ L. For the results obtained with k = 1 ∗ 107 N/m, the relative error stays below 3%
154 Discontinuous deformation analysis in rock mechanics practice
for the entire analysis, and the error is well below 10% for k = 1 ∗ 108 and 1 ∗ 109
N/m as well. It is important to note here that the optimal value for normal contact
spring stiffness is found in this case to be between 2 and 4 orders of magnitude less
than E ∗ L, the value recommended by Shi (1996).
We have considered so far sliding along frictional interfaces, a failure mechanism that
is relevant mainly to rock slope stability. When the blocks are slender however, cyclic
earthquake forces may induce rocking, rather than sliding, that may ultimately lead
to toppling. The limits between sliding, sliding and toppling, and only toppling were
originally discussed by Goodman and Bray (1976) and their analytical solution was
used by Manchu Ronald Yeung in his PhD thesis (Yeung, 1991) to verify 2D-DDA.
Yagoda-Biran (2013) in her PhD thesis has modified those boundaries for the case of
pseudo-static loading and used her solution to verify both 2D-DDA as well as 3D-DDA
methods (see also Yagoda Biran and Hatzor, 2013). All these studies are concerned
with limiting equilibrium and do not address dynamic rocking motions that do not
necessarily culminate in block toppling. The interesting problem of dynamic rocking
of slender blocks subjected to sinusoidal input acceleration has been studied by has
been studied by Makris & Roussos (2000). In Makris and Roussos solution the time
dependent acceleration is input directly to the center of mass of the block, a free body
of which is shown in Figure 5.24.
The centers of rotation of the freestanding column can be either 0 or 0 . Assuming
there is no vertical base acceleration, the equations of motion are (Makris and Roussos,
2000):
Figure 5.24 Free body diagram and sign convention for dynamic rocking of slender blocks (after Makris
and Roussos, 2000). Reproduced with permission from ICE Publishing.
DDA verification 155
where I0 is the mass moment of inertia, m the block mass, and üg the ground accel-
eration. All the geometrical parameters are defined in Figure 5.24. Inserting the
definition
of Io into Equation 5.35 and Equation 5.36, introducing the parameter
p (p = 3g/4R), linearizing the equations due to the slender geometry of the col-
umn (small α), using a sinusoidal input ground acceleration in the form of üg (t) =
ap sin(ωp t + ψ) (where ap , ωp and ψ are the amplitude, frequency and phase when rock-
ing initiates, respectively) and integrating the equations yields (for complete solution
see Makris and Roussos, 2000):
ap
1
θ(t) = A1 sinh pt + A2 cosh pt − α +
3 sin ωp t + ψ , θ≤0
1 + ωp2 p2 g
(5.37)
ap
1
θ(t) = A3 sinh pt + A4 cosh pt + α + 3 sin ωp t + ψ ,
θ≥0
1 + ωp2 p2 g
(5.38)
The dynamic equations for the angular velocity are (Makris and Roussos, 2000):
ωpap
θ̇(t) = pA1 cosh pt + pA2 sinh pt +
3 cos ωp t + ψ , θ≤0
1 + ωp2 p2 g
(5.39)
ωpap
θ̇(t) = pA3 cosh pt + pA4 sinh pt +
3 cos ωp t + ψ , θ≥0
1 + ωp2 p2 g
(5.40)
Yagoda-Biran and Hatzor (2010) used the analytical solution proposed by Makris
and Roussos (2000) to verify 2D-DDA for this difficult case of bouncing angle to edge
contacts. The results of this verification for input peak acceleration 1% lower and
1% higher than the peak acceleration required for toppling according to Makris and
Roussos (2000) solution are shown in Figure 5.25 A and B, respectively, for column
width and height of b = 0.2 m and h = 0.6 m.
In the DDA model the column rests on a fixed base and is subjected to dynamic
input at its centroid. The friction angle along the interface is set to 89 degrees to avoid
sliding, as the analytical solution ignores sliding. The input numerical control param-
eters are: energy dissipation coefficient = 1, maximum displacement ratio = 0.0075,
time step size = 0.0025 sec, normal contact spring stiffness = 83 ∗ 106 N/m, E = 3 GPa,
and ν = 0.25. A remarkably good agreement between the analytical and numerical
solutions is indicated, as can be seen from the plotted numerical error that after initial
perturbations rapidly decreases below 1%.
156 Discontinuous deformation analysis in rock mechanics practice
ü (g)
0.2 0.2
0
0 1 2 3 4 5 6
t = 1.28 sec 1
θ/α
0
–1
0 1 2 3 4 5 6
angular velocity
t = 2.36 sec 1
(rad/sec) 0
–1
0 1 2 3 4 5 6
103 103
t = 4.88 sec 102 102
error(%)
101 101
100 100
10–1 10–1
10–2 10–2
0 1 2 3 4 5 6
A time (sec)
0.2
0
0 1 2 3 4 5 6
t = 1.76 sec 1
θ/α
–1
0 1 2 3 4 5 6
angular velocity
t = 3.32 sec 1
(rad/sec)
–1
0 1 2 3 4 5 6
103
t = 3.92 sec 102
error(%)
101
100
10–1
10–2
10–3
0 1 2 3 4 5 6
B time (sec)
Figure 5.25 2D-DDA verification of dynamic rocking of slender blocks (modified afterYagoda-Biran and
Hatzor, 2010). A) Peak acceleration 1% lower (A) and 1% higher (B) than peak acceleration
required for toppling according to Makris and Roussos (2000) solution.
DDA verification 157
We have shown so far that rock mechanics problems concerning stability and failure
modes of discrete blocks can be addressed effectively and accurately with DDA as long
as the input control parameters are correctly applied. We now turn our attention to
wave propagation phenomena through a block system. This problem is important in
site response analysis as well as in blasting simulations, both issues that are central to
geotechnical earthquake engineering and to mining. Since the formulation of DDA is
dynamic in nature, the main issue to clarify here is how restricting is DDA’s simply
deformable blocks assumption in such simulations? As we have seen in the introductory
chapters on the theory of DDA, the blocks in the original DDA are assumed to be stiff
and simply deformable, namely the computed stresses and strains are assumed to be
distributed homogenously throughout the block; i.e. constant stress and strain within
the block is considered in the solution. Is this assumption seriously compromising our
ability to compute wave propagation accurately with DDA? We will explore this issue
in this section.
Figure 5.26 Model of elastic bar used for P-wave propagation verification.
Since the assumed cross sectional area of the bar is 1 m × 1 m, the peak stress
amplitude entering the bar from the force pulse (Equation 5.41) which is applied at
the left end of the bar (Figure 5.26) is 1 MPa.
The analytical solution for wave velocity in a rod subjected to one dimensional
P wave propagation is given by (Kolsky, 1964):
4
E
Vp = (5.42)
ρo
where ρo is the density and E is the Young’s modulus of the material. The relative error
between analytical and computed stress or velocity is defined here as follows:
|A1 − A0 |
e= × 100% (5.43)
A0
where A1 is the computed wave amplitude or velocity at a reference measurement point
in the model, and A0 is the analytical wave amplitude or velocity at a given point.
The verification is performed using four different time intervals of 0.01 ms,
0.05 ms, 0.1 ms, and 0.5 ms and four different block lengths of 0.5 m, 1 m, 2 m,
and 5 m. The theoretical P-wave velocity in the bar is expected to be 4344 m/s
DDA verification 159
Figure 5.27 Relative stress (A) and velocity (B) errors at center of the elastic bar as function of time
interval and element size.
(Equation 5.42), and as mentioned earlier, the theoretical amplitude of the P wave
is expected to be 1 MPa (Equation 5.41). The numerical P-wave velocity is calculated
from the travel time of the wave between the incident point at the left end of the bar
and the measurement point at the center of the bar. The numerical amplitude of the
wave is computed directly by DDA as the magnitude of the horizontal normal stress
component at the measurement point.
The relative stress and velocity errors are plotted in Figure 5.27 as a function of
time step size and block length. Inspection of Figure 5.27A clearly reveals that the
time interval has a very significant effect on the stress error, but apparently less on the
velocity (Figure 5.27B). Regarding the influence of element size the reverse is true: while
the stress error seems to be unaffected by block size, the velocity error clearly is. As
would be intuitively expected, the velocity error indeed decreases when the element size
decreases from 5 meters to 2 meters. However, when the block size further decreases
to 1 meter the velocity error increases, and continues to increase when the block size
is further reduced to 0.5 m (see Figure 5.27B). This result appears to be in agreement
with the rule of thumb regarding the optimal ratio between the element size and wave
length proposed by Lysmer and Kuhlemeyer (1969): η = 1/12 = 0.08. Considering that
the period of the incident P wave used in our verification study is 0.01 s, and that the
velocity of the wave is 4344 m/s, the wavelength here is 43.43 m. The optimal element
size according to Lysmer and Kuhlemeyer (1969) rule of thumb should be 3.62 m and
here we are getting and optimal block length of 2 m, not very different. The exact
numerical errors with respect to η are shown in Figure 5.28 where indeed it can be
appreciated that near η = 0.08 both stress and velocity errors are low; the minimum
error however appears to be at η close to 0.046, at about half the recommended value
by Lysmer and Kuhlemeyer (1969). The influence of time interval on the obtained
waveform at the center of the bar for a block length of 1 m is shown in Figure 5.29.
The complete waveform reflects both the wave stress and velocity. Clearly, the accuracy
160 Discontinuous deformation analysis in rock mechanics practice
14%
18%
∆t = 0.01ms 12%
∆t = 0.1ms
12% 8%
10%
6%
8%
6% 4%
4% 2%
2%
0% Block length (m)
0% Block length (m) 0 2 4 6 8 10
0 2 4 6 8 10 η
η
A 0.000 0.046 0.092 0.138 0.184 0.230 B 0.000 0.046 0.092 0.138 0.184 0.230
Figure 5.28 Influence of block length/wave length ratio (η) on stress (A) and velocity (B) errors
Figure 5.29 Obtained waveform at center of bar as a function of time step size for element size of 1 m.
of the waveform increases with decreasing time interval most likely due to the strong
influence of the time interval on the stress accuracy.
It has been observed by many workers that in dynamic simulations the penalty
parameter, or the contact spring stiffness, has a very strong influence on the solution
accuracy. Doolin and Sitar (2002) for example pointed out that increasing the stiff-
ness of the contact springs may increase the overall accuracy of the solution in slope
simulations with DDA.
Seven different contact stiffness values, between 10E to 640E, are tested where E
is the Young’s modulus of block material (here E = 50 MPa). The sensitivity analysis
results (Figure 5.30) suggest that while the stress error is not very sensitive to the penalty
parameter, the velocity error is, although not in a straightforward way. The velocity
error decreases with increasing penalty parameter up to an optimal value, here 160E,
beyond which it increases again. The optimal contact spring stiffness found here is
approximately three time higher than the rule of thumb suggested by Shi (1996) of
k = Eo Lo where k is the penalty number (or contact spring stiffness), Eo is the Young’s
modulus of the block, and Lo is the block diameter.
DDA verification 161
Figure 5.30 Influence of contact spring stiffness on the relative stress (a) and velocity (b) error
in elastic bar with element size of 1 m. Note that while the stress error is less sensitive
to contact spring stiffness, the velocity error decreases with increasing penalty value
(in units of Yong’s modulus E) up to an optimal value (here 160E) beyond which the error
increases again.
Figure 5.31 Influence of contact spring stiffness on waveform at center of bar, element size 1 m.
The influence of the penalty parameter on the waveform is shown in Figure 5.31.
Because of the great sensitivity of the velocity error to the k value, simulations with low
k values yield artificial attenuation of the waveform resulting in lower wave velocity
and amplitude. This effect is remedied well with increasing the value of the penalty
parameter, but up to a point, as can be inferred from Figure 5.30.
Figure 5.32 Configuration of the 1-D S-wave propagation model, note implementation of non-reflective
boundary at the bottom of the modeled domain (indicated dimensions are in meters).
P waves, but here we study vertical shear wave propagation through a stack of horizon-
tal layers, where the layers are made very wide with respect to their height (Figure 5.32).
Non reflective boundaries (Bao et al., 2012) are applied at the foundation to restrain
artificial reflections that might distort the measurements, taken at a point positioned
at a distance of 50 m from the upper surface, where the input motion is applied.
Vertical propagation of S waves through a horizontally layered system will only
induce horizontal displacements, it can therefore be considered as a one-dimensional
S-wave propagation, the velocity of which is given by:
4
E
Vs = (5.44)
2ρ0 (1 + ν)
where ν is the Poisson’s ratio. The relative error between the analytical and numerical
solutions is found using Equation 5.43, as before.
The layers are 100 m wide and their height varies in the sensitivity analyses between
0.5 m, 1 m, 2 m, and 5 m. The input horizontal motion at the upper surface in the
verification study is:
The theoretical S-wave velocity for the stacked layers model shown in Figure 5.32
is 2747 m/s (Equation 5.44) given the material properties listed in Table 5.4. The
relative errors with respect to S wave velocity are shown in Figure 5.33 for varying time
intervals and contact spring stiffness. The results suggest that S wave velocity errors
are not sensitive to the time interval but that the deviation from the exact solution
increases with decreasing block size (Figure 5.33a). The influence of contact stiffness
DDA verification 163
Figure 5.33 Verification of vertical S wave propagation through horizontally layered stack. a) Influence
of time step size, b) influence of contact stiffness.
Figure 5.34 Site response analysis with DDA. a) DDA model, b) Equivalent SHAKE model.
The two-dimensional DDA model (Figure 5.34a) is created with layer length to
layer height ratio sufficiently high (15) so as to simulate one dimensional vertical
propagation of shear waves from the excited foundation block through the stack of
the horizontal layers, topped by the surface layer. A real earthquake time history is
applied to the four fixed points at the foundation block, in the horizontal direction only.
The shear waves are then allowed to propagate vertically upward through the stack of
15 horizontal layers, each 1 meter high. The response is measured at two measurement
points M1 and M2 at the foundation block and surface layer, respectively. The same
geometrical configuration is modeled with SHAKE, 15 horizontal layers of infinite
lateral extent each of 1 meter height (Figure 5.34b). The only difference in the loading
scheme is that while in the DDA the foundation block is excited by time dependent
displacements, in SHAKE the excitation at the bedrock layer is in acceleration. In
both methods the excitation is restricted to the foundation block and the response is
measured at the top layer (M2) with respect to the foundation layer (M1). The input
motions for DDA and SHAKE are shown in Figure 5.35.
The damping ratio which is necessary for meaningful comparisons between
SHAKE and DDA is obtained here by controlling the time step size in DDA utiliz-
ing the inherent algorithmic damping (see Doolin and Sitar, 2004) as no other sort of
damping is applied in the DDA version we used for verifications. A damping ratio of
2.3% thus obtained with DDA is input to the SHAKE model to enable quantitative
comparison of the results.
The spectral amplifications obtained with the two different methods are plotted
in Figure 5.36. With DDA the maximum amplification is 29.07 and the resonance
frequency is 13.95 Hz; with SHAKE the maximum amplification is 28.93 at frequency
14.21 Hz. The agreement between the two methods is striking, suggesting that accu-
rate site response analysis is possible with DDA, even when higher order terms are
neglected due to first order approximation and the simply deformable blocks assump-
tion. Moreover – it is clearly demonstrated here that loading the foundation block with
displacement or acceleration time histories is equivalent, an issue that has focused some
debate recently (e.g. Wu, 2010). Results of site response analysis for non-homogenous
layers also show excellent agreement between the two methods (see Bao et al., 2014).
DDA verification 165
Figure 5.35 Input motion used for verification study (CHI-CHI 09/20/99).
Figure 5.36 Spectral amplification obtained with DDA and SHAKE for 15 horizontal layers of
homogenous material properties.
Jiang et al., 2010; Zhang et al., 2013), dynamic slope stability (e.g. Miki et al., 2010;
Sasaki et al., 2005; Sasaki et al., 2007; Wu, 2010), and stability of masonry structures
against earthquake vibrations (e.g. De Luca et al., 2004; Kamai and Hatzor, 2008;
Koyama et al., 2013; Sasaki et al., 2011; Stefanou et al., 2006).
Sensitivity analyses reviewed here show that the DDA method can provide good
accuracy for solution of wave propagation problems, provided that the numerical
control parameters are properly conditioned. Some important points to remember:
Underground excavations
6.1 INTRODUCTION
Since its publication in 1993 (Shi, 1993) DDA has been applied extensively to rock
engineering projects from around the world, while it was still being validated, verified
and modified by the rock mechanics research community, an effort that is still on-going
at the present time. Among the initial test grounds for DDA were some high profile
cases, such as the Three Gorges Project (e.g. Dong et al., 1996), the Masada world
heritage site (Hatzor et al., 2004), the Vajont slide (Sitar et al., 2005), the Jinping
hydroelectric station (e.g. Chen and Deng, 2008), the Tanjiashan landslide triggered
by the Wenchuan Earthquake (Wu et al., 2009), the Bayon temple at Angkor Thom
and Pharaoh Khufu’s pyramid (Ohnishi et al., 2012), to name but a few. DDA is
now being applied in rock engineering projects worldwide on a routine basis, as can
be appreciated by surveying the professional literature in recent years. In this chapter
DDA application in the engineering of underground excavations is demonstrated using
two end members representing two extreme cases of stability concern: 1) shallow
underground excavations where gravitational loading controls the deformation, and
2) deep underground excavations where deformation is controlled by the level of in-
situ stress. In both cases the rock mass is assumed discontinuous so that the mode of
deformation is controlled by the interactions between pre-existing blocks. The chapter
closes with demonstration of rockbolting design with DDA where the anisotropy of
the rock mass controls the length of bolts required for safety.
Shallow underground excavations are at risk due to the low level of in situ stresses. In
lack of high stresses, the available shear strength of discontinuities is hardly mobilized
once the opening is created, and therefore loosening, sliding, and falling of blocks into
the excavation space are common. Of particular concern is the height of the arching
mechanism, which typically develops in the roof, below which loosening and falling
of blocks is to be expected. The assumed height of the arching mechanism therefore
determines the minimum rock bolt length required to ensure stability.
There is no analytical solution for the height and thickness of the stressed zone
in the arching mechanism which develops in the roof of an opening excavated in an
initially discontinuous rock mass. Terzaghi (1946) introduced his famous rock load
168 Discontinuous deformation analysis in rock mechanics practice
on tunnel supports classification, which provides the height of the loosened zone in
the roof as function of the structure of the rock mass, based on empirical observations
from tunnels excavated primarily in the Austrian Alps. As much as we have found
during years of research, Terzaghi’s rock load classification is valid and can be used as
a rule of thumb in lack of any other design tools. Although the result would typically
be conservative, in some cases we found it to be accurate. Of course, if the structure of
the rock mass deviates from the classes of rock structures discussed by Terzaghi, then
more sophisticated design tools must be employed, namely numerical discrete element
methods.
Figure 6.1 DDA model of the rock mass structure at Zedekiah cave with location of measurement
points for further analysis depicted.
Figure 6.2 An idealization of the immediate roof after fracturing at mid-section by a tension crack.
must have been fractured in tension at mid section by a vertical crack propagating
normal to the bedding direction. Note that the compressive strength of this rock in
direction parallel to bedding is 16.4 MPa excluding the possibility of failure in crushing
at mid-section.
If indeed that scenario had materialized and the immediate roof split at mid-section
into two blocks, the resulting configuration could be analyzed by means of the Voussoir
beam approach (see Figure 6.2). However, with the given geometrical and mechanical
parameters the beam would have failed in a “snap through’’ mechanism (see Brady
and Brown, 2004) when analyzed this way due to the large span (30 m) with respect
to the thickness of the layer (0.85 m) and the relatively soft material stiffness.
The only way to explain the exiting stability of the model shown in Figure 6.1,
and indeed of the main chamber of Zedekiah cave that has stood unsupported for over
two millennia, is to allow in the analysis for the interaction between discrete blocks in
the roof. To do this effectively, we must employ a numerical discrete element analysis.
170 Discontinuous deformation analysis in rock mechanics practice
Figure 6.3 Deformation of immediate roof in Zedekiah cave as modeled with DDA. Note geometry
of arching in the discontinuous roof, as delineated by the orientation of the principal stress
trajectories.
The final position of the immediate roof after gravity is turned on in the model
as obtained with DDA is shown in Figure 6.3. Note that after some downward shear
displacement along the abutments takes place the immediate roof stabilizes, leaving
a permanent gap between the immediate roof layer and the overlying layers, which
remain more or less in place. The reason for the stabilization of the immediate roof after
some initial shear along the abutments takes place, is the effective transfer of normal
stresses through the interacting blocks, thus better mobilizing the shear strength of the
joints and arresting any further downward displacement.
The dynamic deformation process can be better understood by inspection of
Figure 6.4 where the vertical downward displacement and horizontal stress evolu-
tion in the four measurement points shown in Figure 6.1 are plotted in panels (a) and
(b) respectively as function of real time. Consider measurement point 1 at the imme-
diate roof marked by open diamonds in the figure. Once the opening is created at
time 0 s, it begins to exhibit downward vertical displacement at a constant velocity,
achieved by shearing deformation of the entire immediate roof layer along the abut-
ments. After three seconds the vertical displacement is suddenly arrested and at the
same time the horizontal stresses at that point abruptly increase from zero to 700 kPa.
With the given friction coefficient for the joints in the simulation (µ = 0.87) this level
of horizontal stress is sufficiently high to arrest any further downward displacement
under the gravitational pull on the system.
We see through this example the importance of modeling the deformation of the
entire block system rather than separating the problem into few representative ele-
ments that can be handled more easily using existing analytical or semi-analytical
Underground excavations 171
Figure 6.4 DDA results of deflection and horizontal stress evolution in the four measurement points
shown in Figure 6.1 (after Bakun-Mazor et al., 2009).
Figure 6.5 Layout of the underground water system discovered under Tel Beer-Sheva revealing the
alignment of the excavated chamber walls with the strikes of the principal joint sets shown
as embedded rose diagrams (after Hatzor and Benary, 1998).
excavation span, influence the stability of the underground structure? Studying historic
excavations in horizontally layered and vertically jointed rock masses can shed some
light on this problem and provide constraints on predictions otherwise obtained by
running countless numerical simulations on the computer.
A case in point is the underground water storage system discovered under Tel Beer-
Sheva, an archeological site near the modern city of Beer-Sheva (where Ben-Gurion
University of the Negev is located). The water storage system dates back to the Israelite
period, ca. 3000 years before present, and is believed to have been used to store
water during siege on the ancient city walls, by diverting water from a nearby stream
in aqueducts that were excavated below the city walls and connected between the
stream and the storage system. The system, comprised of several square chambers, was
excavated in horizontally layered and vertically jointed chalk, keeping the chamber
walls aligned with the strike directions of the principal joint sets that are roughly
orthogonal to one another (see Figure 6.5).
The roof of the system has collapsed during excavation as attested by the applica-
tion of plaster on the exposed sidewalls extending above the level of the original roof,
Underground excavations 173
Figure 6.6 Plan (top) and cross sections (bottom) of the collapsed roof of the underground water
system at Tel Beer Sheva (after Hatzor and Benary, 1998).
and by the massive support pillar constructed by the ancient engineers. Inspection of
the system reveals that the construction of the support pillar was sufficient to arrest
any further deformation of the roof, which indeed remained unsupported until discov-
ered by modern archeologists several decades ago. A plan of the roof and two cross
sections (Figure 6.6) show that the roof collapsed into a more stable dome structure,
the contours of which delineate, most likely, the geometry of the arching mechanism
that developed in the roof once the original excavation was attempted.
The well documented failure of the immediate roof at Tel Beer Sheva provides us
with a good opportunity to check the predictive capability of DDA when studying the
stability of stratified and jointed rock mass structures.
The rock mass consists of horizontal bedding planes with average thickness
of 0.5 m, and two principal joint sets (J1 and J2 in Figure 6.5) both with mean
174 Discontinuous deformation analysis in rock mechanics practice
Figure 6.7 Mechanical behavior of the chalk at the site. Left – Anisotropic strength and elasticity (β is
the angle between the compression direction and the normal to the bedding), Right –
normal stress dependent friction angle of the joints (Tsesarsky, 2005).
spacing of 0.25 m. The rock is an upper Cretaceous chalk, locally known as Taqiya
Formation, exhibiting anisotropic mechanical behavior with ultimate strength and
elasticity much higher when tested parallel to bedding, the relevant direction here.
Direct shear tests of the joints show that the friction angle of the joints decreases with
increasing normal stress (see Figure 6.7).
The sensitivity of the structure to joint friction and spacing was studied by Hatzor
and Benary (1998) with DDA using a mesh similar to the one shown in Figure 6.8,
where the spacing between the joints and the friction on the joints were varied between
the simulations. Since here we are only interested in the failure of the immediate
roof only the upper mesh, used for detailed modeling of the deformation of a sin-
gle jointed layer, is considered. The results of DDA simulations are plotted graphically
in Figure 6.9 where the required friction angle for layer stability is reported for seven
joint spacing configurations. In the most slender block geometry with joint spacing of
Sj = 25 cm, the layer consists of 28 blocks, with block dimension Sj /t = 0.5 and scaled
block length of Sj /S = 0.04. In the extreme case where the layer is completely contin-
uous, the joint spacing is Sj = 700 cm, the layer consists of one block only with block
dimension of Sj /t = 14 and a scaled block length Sj /S = 1.
The relevant case for comparison with the Voussoir beam analogue discussed in
the previous section would be for joint spacing of Sj = 350 cm, where the layer is split
into two blocks at mid-section. As can be inferred from inspection of Figure 6.9, a
friction angle of 45◦ would be required to keep this configuration stable. Indeed this
result is in agreement with the Voussoir beam analogue which predicts that the two
block beam will be safe against shear along the abutments with the given geometrical
and mechanical parameters provided that the available friction angle for the joints is
greater than 40◦ (see Hatzor and Benary, 1998). Due to the relatively high compressive
strength and elasticity of this chalk in direction parallel to bedding (the direction of
the thrust force in the beam), failure by crushing or by a “snap through’’ mechanism
(see Brady and Brown, 2004) are ruled out by the Voussoir beam methodology.
Underground excavations 175
Figure 6.8 DDA mesh similar to the one used for sensitivity analyses (after Tsesarsky and Hatzor,
2006).
Figure 6.9 Results of sensitivity analysis of immediate roof layer in a typical chamber at Tel Beer-Sheva
(after Hatzor and Benary, 1998).
176 Discontinuous deformation analysis in rock mechanics practice
Figure 6.10 Comparison between first order DDA and UDEC with deformable blocks for the case
of Tel Beer Sheva. The optimal scaled block length of Sj /S = 0.25 is confirmed by both
methods (after Barla et al., 2010).
Once we introduce more joints into the layer, the Voussoir beam analogue ceases
to be valid of course, and we must rely on predictions provided by numerical analysis.
Results of DDA simulations as plotted in Figure 6.9 provide a useful insight into this
problem. While the demand for joint friction decreases with increasing joint spacing
from Sj = 25 cm to 175 cm, it increases when the block length increases from what
appears to be an optimal scaled length of Sj /S = 0.25, where the required friction for
stability is only 25◦ , to what appears to be the worst case of Sj /S = 1.
The discovery of an optimal scaled block length of Sj /S = 0.25 is very worthwhile
in itself, and can be applied to excavations through coal seams for example. In terms
of optimal block dimension it is found to be Sj /t = 3.5.
We can now understand the failure of the immediate roof at Tel Beer-Sheva the first
time the excavation was attempted by the skillful ancient engineers. With the average
joint spacing in the field being only 25 cm, the required friction angle for stability
is 80◦ , much higher than the available friction angle for the joints at the site (see
Figure 6.9).
Note that the accuracy of first order DDA with simply deformable blocks assump-
tion is compromised when the block size is large with respect to the modeled domain,
because the blocks cannot bend when using first order approximation. With a sin-
gle block forming the roof layer, as is the case in our example with joint spacing of
700 cm, the layer will most likely bend before it shears along the abutment, and that
complex failure mechanism cannot be modeled accurately with first order DDA. To
demonstrate this consider Figure 6.10 where first order DDA is compared with UDEC
with deformable blocks for the exact same problem. When UDEC is applied with a
coarse grid the results are strikingly similar. When the UDEC mesh is refined, however,
we see that block bending in the case of 700 cm spacing allows roof stabilization even
with a relatively low friction angle. Note also that when the block size decreases with
Underground excavations 177
respect to the size of the modeled domain the agreement between first order DDA and
UDEC with deformable blocks greatly improves.
Interestingly, with both methods the existence of an optimal scaled block length
of Sj /S = 0.25 is confirmed, suggesting that this could indeed be a rule of thumb for
assessing the risk of roof collapse when excavating through horizontally layered and
vertical jointed rocks. This would require the geological engineers to properly asses
the layer thickness and the joint spacing at the site.
Figure 6.11 DDA mesh of a rock mass consisting of horizontal layers and vertical joints with mechanical
layering.
Figure 6.12 Position of measurement points for recording computational output data.
concentrations due to the opening near the boundaries. The relationships between
tunnel span B, tunnel height H, and crown height C remain constant in all simulations.
19 DDA models are analyzed with varying span vs. depth ratios as listed in
Table 6.1. Because the mean joint spacing in the modeled rock mass remains the same,
the total number of blocks in the mesh increases with increasing tunnel depth and
tunnel diameter. The greatest number of blocks occurs at the model with the deepest
tunnel of the widest span, namely Model 10 in Table 6.1 (11,104 blocks). Model 1
with the smallest tunnel diameter and shallowest cover results in the smallest number
of blocks in the mesh (871). Depending on the kind of processor, it is evident that a
significant demand of CPU time will would be needed to complete these simulations.
The performance of the modeled opening can be assessed by the deflection and
horizontal stress evolution at mid-section of the immediate roof. As was seen in the
case study of Zedekiah cave, the mere fact that the immediate roof exhibits vertical
Underground excavations 179
Table 6.1 DDA models used to study the span vs. overburden problem in horizontally layered and
vertically jointed rock masses. Legend: B is tunnel span, H is tunnel height, c is crown height,
b is distance to modeled domain boundary, and h is overburden thickness.
1 10 5 2.5 20 6 871
2 14 7 3.5 28 6 1325
3 14 7 3.5 28 11 1780
4 17 8.5 4.25 34 6 2123
5 20 10 5 40 6 2221
6 20 10 5 40 10 2715
7 20 10 5 40 14 3197
8 40 20 10 80 15 8638
9 40 20 10 80 20 9948
10 40 20 10 80 25 11104
11 30 15 7.5 60 15 5720
12 30 15 7.5 60 22 6947
13 30 15 7.5 60 30 8157
14 25 12.5 6.25 50 12.5 4287
15 25 12.5 6.25 50 25 5925
16 25 12.5 6.25 50 19 5086
17 25 12.5 6.25 50 30 6591
18 22 11 5.5 44 11 3240
19 22 11 5.5 44 22 4683
deformation from the first time step still does not necessarily mean that the roof is
doomed. Even in the most stable configuration some vertical shear along the abutments
will always precede the arching mechanism. Once the arching mechanism is developed
in the roof two observations will be made: 1) the vertical deflection will cease, and
2) horizontal stresses will reach a constant value the magnitude of which would be
sufficient to keep the blocks in place given sufficient joint friction.
Using these stability guidelines we can group the simulation results of the 19 mod-
els into three main categories: 1) Unstable, 2) Marginally Stable, 3) Stable. In the
“Unstable’’ category vertical deflection will never be arrested and horizontal stresses
may never be developed, thus, ultimately, the entire roof may collapse and in very shal-
low opening the ground surface may break (Figure 6.13). In the “Marginally Stable’’
configuration we may loosen the immediate roof, however stable arching may develop
from the middle roof and above (Figure 6.14). Finally, in the “Stable’’ category effective
arching will arrest any further vertical deflection of the entire roof (Figure 6.15).
When plotting the results of these 19 simulations in span (B) vs. cover height (h)
space an interesting and surprising conclusion is obtained. Instead of confirming the
famous rule of thumb proposed by Terzaghi (1946) predicting that in blocky rock
masses the height of the rock requiring support will be 0.5B, here we get a nonlinear
function. In relatively small span tunnels the demand for cover remains constant on a
low value, much lower than the two tunnel diameters rule of thumb. Once the tunnel
diameter exceeds a certain size, here found to be 18 m, the demand for cover increases
very steeply. Although it never reaches the rule of thumb h = 2B, it does reach h = 1B
Figure 6.13 Graphical illustration of “Unstable’’ configuration.
Table 6.2 Geometrical characteristics of the rock mass used in the 19 models.
at the most extreme case, here for tunnel diameter of 30 m. When the opening diameter
continues to increase beyond that point the rate of change of demand for cover with
respect to tunnel diameter becomes very small, namely, once we are deep enough, we
are quite safe.
Of course these results are assumed to be valid only for the rock mass simulated
here, the geometrical and mechanical properties of which are listed in Table 6.2 and
Table 6.3. However, when plotting results from several other case studies in hori-
zontally layered and vertically jointed rock masses but with different mechanical and
geometrical properties (Hatzor and Benary, 1998; Bakun-Mazor et al., 2009; Hatzor
et al., 2010), the results fall nicely along the nonlinear trend presented in Figure 6.16,
182 Discontinuous deformation analysis in rock mechanics practice
Figure 6.16 Overburden vs. span relationships in blocky rock masses as predicted by DDA. The solid
line separates between unsafe and safe configurations. Three case studies in blocky rock
masses in three different lithologies plot near the numerically predicted function possibly
suggesting it’s universality for blocky rock masses.
suggesting that this predicted trend indeed may be true for a large range of blocky rock
masses.
Figure 6.17 The canyon in Jinsha river where Baihetan hydropower project is being constructed. The
left bank consists of columnar basalts with basal planes dipping gently to the right (east)
whereas in the right bank a thick sequence of southerly dipping sandstones overlies the
columnar basalt sequence.
mesh from the beginning of the forward modeling stage, excessive displacements will
be computed by DDA because it takes a relatively large number of time steps until the
imposed initial stresses are fully developed and the available shear strength of all joints
is fully mobilized (Tal et al., 2014). The number of time steps required to achieve this
equilibrium increases with the number of blocks comprising the DDA mesh, and there-
fore some experience is required in determination of the time step at which the opening
should be removed in the model. In lack of such experience, the user can find by trial
and error the time step at which the recorded stresses in some strategically located
measurement points reach the imposed values, and then re-run the simulation with
the opening being removed after that time step. Clearly, a modification to the original
DDA code is necessary, where essentially the modeled domain first equilibrates under
the imposed initial stresses in a “static’’ deformation stage, followed by “dynamic’’
forward modeling stage at the beginning of which the opening is removed.
Another important issue to realize when modeling deep underground excavations
with DDA is the distance between the opening and the boundaries of the modeled
domain. Clearly, when the average block diameter is in the order of several meters and
the depth of the excavation is in the order of several thousand meters it is not feasible
to model the entire problem with a DDA mesh comprised of so many discrete blocks.
Since we are interested primarily in the deformation near the tunnel, the modeled
184 Discontinuous deformation analysis in rock mechanics practice
domain boundaries may be set close to the opening, to a distance of say 3 to 5 tunnel
diameters, to allow for complete dissipation of opening induced stress concentrations.
However, if we do this, the boundaries of the modeled domain should be viscous, so as
not to reflect waves that were supposed to propagate to infinity through the rock mass
back into the analyzed domain, an artifact that will distort the computation. Applying
viscous boundaries close to the opening therefore enables reducing the total number
of blocks in the mesh, the interactions of which must be computed and results stored,
in every time steps. This will save enormous computation time without compromising
solution accuracy. Several methods for applying viscous, or non-reflective, boundaries
specifically for DDA have recently been proposed (Jiao et al., 2007; Bao et al., 2012).
In the discussion that follows extensive use is made of both sequential excavation and
non-reflective boundaries enhancements for DDA.
Figure 6.18 Cross section through a 20 m span diversion tunnel in the left abutment of Baihetan Dam
showing the eastwardly dipping basal planes (10/90) and the steeply dipping columnar joints
(80/270). The mean spacing of the basal planes is 5 m, and of the columnar joints 0.3 m.
Figure 6.19 Initial configuration of the mesh used to find the extent of the EDZ in the diversion tunnel
(left) with the imposed initial in situ stresses (right). The assumed initial in-situ stress field
is shown in inset.
The vertical displacement of the measurement points in the roof is shown in Fig-
ure 6.21. Note that the three lowermost measurement points never stop their vertical
motion, meaning the vertical displacement is not arrested by an effective arching mech-
anism, which must develop higher in the roof. Judging from the output displayed
in Figure 6.21 maximum arching stresses must develop somewhere between 10 and
186 Discontinuous deformation analysis in rock mechanics practice
Figure 6.20 Sequence excavation modeling. Left – top heading removal after 17,000 time steps
(0.109 s), Center – mid section removal after 120,000 time steps (0.798 s), Right – bench
removal after 230,000 time steps (1.49 s).
Figure 6.21 Vertical displacement of five measurement points in the roof. The height of the arching
stresses must be between 10 and 25 meters above the immediate roof.
25 meters about the immediate roof. This means that the height of the loosening zone
in this rock mass extends more than Terzaghi’s rule of thumb of 0.5B for blocky rock
masses, which is typically considered conservative.
The reason of course for this extended height of the loosening zone is the peculiar
structure of the rock mass. The steep inclination of the columnar joints restricts the
ability of the arching stresses to be transferred properly from one block to another and
thus an extremely large section of the immediate roof is loosened once the opening
is created. In the sidewalls, however, the extent of the loosening zone is much more
restricted, with sliding indicated in the right sidewall (Figure 6.22) and block toppling,
as expressed by the measurement forward rotation in point 9, in the left sidewall
Figure 6.23. In both side walls the loosening zone does not extend beyond 3 meters,
in stark contrast to the situation in the roof where the loosening zone extends beyond
10 meters.
Underground excavations 187
Figure 6.22 Horizontal displacement vs. time in right sidewall. Measurement point 6 outputs suggest
sliding of the wedges in the immediate sidewall into the opening, but measurement points
7 and 8 output suggests that this deformation is arrested less than 3 meters from the
sidewall.
Figure 6.23 Deformation in the left sidewall is dominated by forward rotation of the immediate blocks
on the face as indicated by the plotted rotations in the figure. The deformation does not
extend to more than 3 meters from the face.
6.3.3 Rockbolting
Rockbolting design in discontinuous rock masses is based largely on experience. Most
empirical rock mass classifications provide some guidelines for dimensioning rockbolts
188 Discontinuous deformation analysis in rock mechanics practice
in jointed rocks. Although they are not intended to be design tools, many engineers do
rely on guidelines offered by empirical classification methods in the design of support
of underground excavations. One of the most useful design guidelines for dimen-
sioning rockbolts for underground excavation in discontinuous rocks was offered
by T. A. Lang, based first on his experience gained during the construction of the
Snowy Mountain project in Australia, and on many tests he has later conducted in
the US (Lang, 1961, 1972; Lang and Bischoff, 1982). A good and accessible review of
T. A. Lang guidelines for dimensioning rock bolts in discontinuous rock is provided
by Brady and Brown (2004).
For blocky rock masses, Lang proposed that the length of the rock bolts (L) should
be the longest of the following three options:
With regard to the spacing between the bolts, Lang suggested that it be the smaller
of the following two options:
• s = 0.5L
• s = 1.5b
Let us now consider the rock mass structure shown in Figure 6.18. With an opening
span of 20 m by Lang guidelines the length of the bolts should be L = 0.25B = 5 m.
If we consider the size of the blocks formed by columnar joints as the “keyblock’’
in the rock mass, then by Lang guidelines the spacing between the bolts should be
s = 1.5b = 0.45 m. From a practical standpoint, however, this bolt spacing is too small,
and therefore the other suggested option of s = 0.5L = 2.5 m should be used.
An example of a “conservative’’ rock-bolting pattern based on Lang’s guidelines is
shown in Figure 6.24 and the response of the bolted roof as computed with DDA
is shown in Figure 6.25. As would have been anticipated based on the discussion in
the previous section, this rockbolting pattern does not have any reinforcement effect
on the rock mass in the roof, because the loosening zone extends far beyond the
“conservative’’ bolt length of 6 m.
But this should not discourage us from using the well tested empirical guidelines
proposed by Lang. Rather, the application of these guidelines must be done with
particular attention paid to the particular geometry of the rock mass at the site. The
key issue here is to realize that the “average keyblock width’’, and important design
parameter in Lang’s guidelines, is not the same in this case for the roof and the sidewalls.
In the sidewalls rockbolts are installed in direction roughly normal to the columnar
joints, therefore the relevant block size for the sidewalls is given by the spacing between
the columnar joints, namely 0.3 m. Since using the L = 3b rule will obviously lead
to too short length, we can use the L = 0.25B rule (18 m < B < 30 m) to obtain a
design rockbolt length of L = 5 m for the sidewalls. This length will supply ample
reinforcement to the loosened zone in the sidewalls and predicted by DDA. In the roof
rockbolts are installed in direction roughly normal to the basal planes, therefore the
Underground excavations 189
Figure 6.24 A “conservative’’ rockbolting design pattern based on T. A. Lang’s guidelines for the diver-
sion tunnel used as an example in the previous section. Bolt length in roof is 6 m and bolt
spacing is 1.2 m.
Figure 6.25 Roof deformation with (B) and without the rockbolts as inferred from vertical displacement
of five measurement points in the roof. Note that application of rockbolts according to the
layout shown in Figure 6.24 hardly makes any difference and the same height of loosening
zone is obtained above the immediate roof, with or without, the bolts.
190 Discontinuous deformation analysis in rock mechanics practice
relevant block width is determined by the mean spacing of the basal planes. Using the
rule of L = 3b for the roof here would require a bolt length of 15 m, which according
to our modeling results should be sufficient to arrest any downward displacement in
the roof.
We have seen in these two sections how the geometry of the rock mass structure
controls both the extent of the loosening zone behind the free faces of the excavation,
as well as the correct way to apply safe rockbolting support. In the next section we will
examine how pre-existing blocks in the rock mass may be ejected as hazardous rock-
bursts when the excavation is created in a rock mass that is originally discontinuous
and subjected to a high level of in situ stresses.
6.3.4 Rockbursts
It is widely accepted that rockbursts are generated in deep underground openings by
fracture of initially intact rock. Therefore, analytical approaches based on fracture
mechanics have been employed to discuss the governing mechanism (e.g. Cook, 1966;
Fairhurst and Cook, 1966; Cook, 1976), to assess the extent of the excavation damage
zone (e.g. Perras and Diederichs, 2016), and to propose effective support measures (e.g.
Kaiser and Cai, 2012). Here we propose a different outlook on rockbursts. We argue
that when a deep underground opening is excavated in an originally discontinuous rock
mass that is subjected to very high initial stresses, before fracture of intact rock takes
place, removable blocks (as defined by Goodman and Shi, 1985) may be ejected as
projectiles from the surrounding rock in response to strain relaxation. Of course, if the
rock is completely intact, or if removable keyblocks are not formed by the intersection
of pre-existing discontinuities around the opening, this assumed mechanism would not
take place, and the rockburst potential can indeed be analyzed using continuum based
approaches.
To illustrate our concept consider Figure 6.26 where a circular tunnel is excavated
out of a discontinuous rock mass that is subjected to an initial hydrostatic stress field of
50 MPa. The inclination of the joints is 45◦ to right and left, and the input friction angle
for the joints is 65◦ . Under gravitational loading, the removable key block in the left
sidewall should not move once the opening is created, because the joints offer sufficient
frictional resistance under static conditions. But because so much elastic strain energy
is stored in the rock before the excavation is formed due to the high in-situ stresses,
once the opening is formed some of this stored energy is transferred into kinetic energy
that is sufficient to overcome the frictional resistance of the joints. When this happens,
if the block is removable as in this example, it will be shot out of the rock mass into
the opening space as a projectile that for all practical purposes may be referred to as a
“rock burst’’.
To appreciate the magnitude of the peak acceleration and velocity such key block
ejections can reach, consider the time histories plotted in Figure 6.27 as obtained with
DDA for the mesh shown in the right panel. The tunnel is removed at time 0.3 s, the
imposed initial stress is hydrostatic and equals 30 MPa, the dip of the joints is 45◦ to
both right and left, the input friction angle for the joints is 65◦ , the Young’s modulus
of the rock is 20 GPa, and Poison’s ratio is 0.2. Note that an extremely high peak
Underground excavations 191
acceleration is obtained the first instant the key block is ejected from the host rock, but
it quickly drops to zero while the velocity remains constant (note that this simulation
is done with no gravity). The peak velocity obtained in this DDA simulation is similar
to values of 3 m/s reported by Kaiser and Cai (2012), 10 m/s reported by Ortlepp and
Stacey (1994), and to values between 0.6 to 2.5 m/s measured by Milev et al. (2001)
using high speed video camera in situ.
The energy balance associated with opening the excavation in an initially contin-
uous, homogeneous, linear-elastic (CHILE) rock has been calculated analytically by
He et al. (2016) who have also showed that the zone of influence in terms of energy
increase extends to a distance of three diameters from the tunnel center. Once the
opening is formed, the energy increase in that annulus around the excavation must be
balanced by the following three energy components: 1) the elastic strain energy that
goes into intact rock elements, 2) the kinetic energy that moves removable blocks in
the affected zone, and 3) the energy that dissipates by shear displacement of blocks
along joints. Energy components 1 and 2 are readily available from DDA output, pro-
vided that a measurement point is positioned at the center of each block in the affected
annulus. Since the energy increase in the affected annulus is found analytically (He
et al., 2016), the shear component of the total energy budget can be found by simple
subtraction.
The evolution of the kinetic energy of the entire block system in the affected annu-
lus as computed with DDA is shown in Figure 6.28 for several joint friction coefficients.
The restraining effect of joint friction is readily apparent, where an increase of friction
192 Discontinuous deformation analysis in rock mechanics practice
Figure 6.27 Acceleration, velocity, and displacement of the ejected keyblock shown in inset as
computed with DDA.
tan ϕ=0.75
(kJ)
tan ϕ=1
1000
Kinetic energy of
ejected block
0
0.29 0.3 0.31 0.32 0.33 0.34
Time (s)
Figure 6.28 Evolution of kinetic energy of the block system in the affected zone of the opening as a
function of joint friction.
coefficient from 0.25 to 0.5 is sufficient to reduce the total kinetic energy of the block
system in the affected domain roughly by a factor of 2.
The influence of initial stress and joint friction on the kinetic energy of the ejected
key block, or the rockburst, is shown in Figure 6.29. With increasing initial in situ stress
the kinetic energy of the rockburst naturally increases when the frictional resistance
Underground excavations 193
500 m = 0.5; k = 2
Opening
Kinetic energy of rock burst
400 Stress
equilibrium
300
(kJ)
0
0.2995 0.3000 0.3005 0.3010 0.3015 0.3020
Time (sec)
300
Ejected
Kinetic energy of rock burst
keyblock
63 m
200 8m
(kJ)
45°30°
100 66 m
s1 = 60 MPa
s2 = 30 MPa
0
0.25 0.45 0.65 0.85 1.05
Figure 6.29 Kinetic energy of ejected keyblock as a function of initial stress (top) and joint friction
(bottom). DDA mesh used in simulation shown in inset.
offered by the joints is kept constant (top panel). Similarly, for a given initial in situ
stress the kinetic energy of the rockburst decreases with increasing frictional resistance
of the joints (lower panel), as would be expected.
In order to probe further into the influence of rock mass characteristics on rock-
burst intensity, we will decrease somewhat the size of the studied annulus to reduce
the total number of monitored blocks in the DDA mesh, from a distance of three to
a distance of 1.5 tunnel diameters measured from tunnel center. It can be shown that
the energy density concentration is highest in this restricted domain (He et al., 2016).
We can investigate the influence of various rock mass parameters on the distribu-
tion of the energy components by changing the input parameters in DDA and checking
194 Discontinuous deformation analysis in rock mechanics practice
Figure 6.30 Influence of Young’s modulus on the distribution of the energy components.
the dynamic response of the system. The rock mass parameters that are considered are
the Young’s modulus of intact rock, the mean spacing of the joints, friction coefficient
of the joints, and the inclination of unfavorably oriented joints. Young’s modulus of
intact rock elements scales the rock mass stiffness, the mean spacing between joints
scales the mean block size in the rock mass, the friction coefficient of the joints scales
the shear resistance of the rock mass, and the inclination of the unfavorable oriented
joints scales the potential for obtaining block displacements into the opening from a
kinematic standpoint. While all empirical rock mass classification systems address, in
one way or another, the first three parameters, only the RMR method provides a way
to address unfavorably inclined joints.
The influence of Young’s modulus on the distribution of the energy components
is shown in Figure 6.30. A drastic decrease in elastic strain energy is indicated when
the modulus increases from 20 GPa which is characteristic of soft rocks, to 60 GPa
which is characteristic of stiff rocks. This decrease in elastic strain energy also implies
that much less elastic strain energy was stored in the rock mass before the opening was
created and therefore that there is much less energy to be balanced after the excavation
is created by the aforementioned three mechanisms. This means that with increasing
rock mass stiffness the rockburst potential in fact decreases.
The influence of the mean block size in the rock mass on the energy distributions
after the excavation is made is shown in Figure 6.31, where two important trends can
be observed: 1) the dissipated energy by shear sliding increases with increasing block
size, and 2) the kinetic energy of the block system decreases with increasing block size.
Both results are intuitive, suggesting that with increasing block size in the rock mass
the energy of rockburst is expected to decrease.
The influence of joint friction on the distribution of energy components after the
excavation is made is shown in Figure 6.32. As would be expected, the dissipated
energy by shear sliding increases with increasing joint friction, and the kinetic energy
of the block system is reduced with increasing joint friction. This result implies that
Underground excavations 195
Figure 6.31 Influence of mean joint spacing on the distribution of the energy components.
Figure 6.32 Influence of joint friction on the distribution of the energy components.
with increasing shear resistance of the rock mass the energy of rockbursts is expected
to decrease.
Finally, the influence of unfavorably oriented joints is shown in Figure 6.33. Two
joint sets are considered, one horizontal and the other inclined, with the dip angle
α measured as shown in the upper left panel. The obtained result is not necessarily
intuitive. We see that the highest kinetic energy for the block system is obtained when
the dip of the inclined joint set is vertical, and it decreases with decreasing dip of the
inclined joint set. This can be explained by the degree of interlocking in the block system
which appears to decrease with increasing α as indicated by the red and yellow lines
delineating the displacement vectors of the removable blocks in the modeled annulus.
Note that in the left upper panel showing a block system with gently inclined joints
196 Discontinuous deformation analysis in rock mechanics practice
Figure 6.33 Influence of joint inclination on the distribution of the energy components.
Table 6.4 Input rock mass parameters to DDA and corresponding rockmass quality ranking by the
major systems.
the freedom of point B to move towards the opening is constrained by point A that
must move first. Such constraints on the motion of blocks are absent in the case of
the vertical joints (upper right panel) where indeed all the blocks delineated in red and
yellow arrows are free to move simultaneously into the opening once the excavation
is created.
These sensitivity analyses naturally lead to the discussion of the relationship
between rock mass quality and the energy of rockbursts. As mentioned earlier, only
the RMR method allows for consideration of unfavorably oriented joints and there-
fore here we shall use the RMR method for ranking rock mass quality. Of course,
Underground excavations 197
Figure 6.34 Four block systems used for DDA modeling of fours rock mass qualities listed in Table 6.4,
decreasing in quality from upper left to lower right.
correlations between all major rock mass classifications are readily available (e.g. Hoek
et al., 1995).
In Table 6.4 input parameters for DDA simulations that represent different rock
mass quality rankings according to the different empirical rock mass classification
systems are listed. Note that the inclination of the unfavorable joint set is not included
here, and in all simulations a symmetric joint inclination of 60 degrees is assumed. The
number of monitored blocks in the analyzed annulus increases with decreasing rock
mass quality because of the smaller joint spacing that is assumed for poor quality rocks.
The DDA block systems that correspond to each of the four rock mass categories listed
in Table 6.4 are shown in Figure 6.34.
198 Discontinuous deformation analysis in rock mechanics practice
Figure 6.35 Stored energy in tunnel before it is removed and kinetic energy of the block system in the
analyzed annulus as a function of rockmass quality.
The four block systems shown in Figure 6.34 where subjected to an initial hydro-
static stress of 55 MPa with all other input parameters as listed in Table 6.4. The
deformation of all monitored blocks in the analyzed annulus was studied and the elas-
tic strain energy, the kinetic energy, and the dissipated energy by shear sliding along
joints were calculated for each rockmass quality. The energy that was stored in the
tunnel area before it was removed and the kinetic energy of the block system in the
analyzed annulus around the excavation once it is formed are plotted in Figure 6.35
as a function of rockmass quality. It can be appreciated that the greatest rockburst
energy is obtained with rockmass qualities that are between fair and good, in RMR
terms between RMR = 65 and 75.
When the rock mass quality is lower than 65 it is feasible that tunneling induced
stress concentrations will indeed fracture intact rock elements because of their relatively
low compressive strength with respect to the magnitude of the stress concentrations
around the tunnel. When the rock becomes sufficiently strong so that it can sustain the
stress concentrations, the risk of rockbursts in the form of block ejections increases.
Here we show that when rockbursts in the form of block ejections are feasible, their
kinetic energy will be highest for fair rock (RMR = 65) and it will decrease drastically
when the RMR value is greater than 75. Clearly, it would have been impossible to
arrive at these conclusions without performing dynamic discontinuous deformation
analysis.
Chapter 7
Rock slopes
7.1 INTRODUCTION
The discipline of slope stability in soil like materials is very developed with advanced
analytical and numerical solutions for both static and dynamic problems. These
approaches universally assume continuity of the medium where separation between the
elements comprising the sliding mass is not allowed and all deformation is assumed to
be concentrated along a well defend sliding surface. When the analyzed mass remains
continuous throughout the deformation process, it becomes much easier to consider
complex constitutive laws for the material, and to address pore pressure evolution
with ongoing deformation. Moreover, addressing time dependency and creep defor-
mation are easier when assuming continuity. Thus, the deformation and mechanism
of failure of landslides have typically been analyzed assuming continuity, where the
main questions were the triggering mechanism, the critical state at which the material
failed, the amount of displacement, and possibly also the velocity evolution of the
sliding mass. Famous landslides have been analyzed this way, for example the Vajont
landslide (sometimes referred to as the Vaiont, or Mount Toc, landslide) in the Italian
Alps (e.g. Mencl, 1966; Skempton, 1966; Jager, 1979; Trollope, 1980; Veveakis et al.,
2007) which was in fact a rock slide (see Barla and Paronuzzi, 2013). Furthermore,
continuity has been assumed to predict relationships between earthquake magnitude
and expected displacement of the sliding mass (e.g. Jibson, 2007) based on analytical
solutions which consider a single two-dimensional slice, such as those proposed by
Newmark (1965).
As we have seen in this book, rock masses are rarely continuous. Even if a rock
slope starts as a continuous body, once sliding begins it may quickly disintegrate into
discrete blocks, either along pre-existing discontinuities or along deformation induced
fractures. During sliding these blocks will interact with one another, while the entire
discontinuous mass will move together along some well-defined shear surface, either
pre-existing, or evolving. Clearly, analytical closed – form solutions do not exist for
such complex dynamic processes. With the introduction of numerical discrete element
methods, however, it became possible to compute the deformation of landslides with-
out ignoring the interaction between the blocks comprising the rockmass, and even to
allow for sliding induced disintegration of the mass to smaller blocks during sliding.
Sitar, MacLaughlin and coworkers (Sitar et al., 2005) applied DDA to the case of
the Vajont slide by discretizing the sliding mass in advance, before sliding initiated,
into increasing number of blocks, and studying the influence of block kinematics
200 Discontinuous deformation analysis in rock mechanics practice
on the velocity of the sliding mass, an issue of intensive debate among Vajont slide
researchers. They found that peak velocity increased by up to 50% as the number of
blocks increased, indicating that internal disintegration of the landslide mass results in
increasing acceleration and higher peak velocity. The have also simulated pore pres-
sure rise due to shear heating using their DDA velocity results, and showed that this
mechanism could also explain peak velocity rise as the number of blocks increases.
They concluded that the increase in peak velocity due to disintegration suggests that
as much attention should be paid to the geometry of discontinuities, as is paid to shear
strength and pore pressure in conventional slope stability studies.
In addition to the strong influence of the initial geometry of the sliding mass, the
dynamic loading input is also very important in determining the final configuration of
the sliding mass. Wu (2010) tested various methods of seismic record input into DDA
using the Chiufenerhshan landslide that was triggered by the 1999 M = 7.3 Chi-Chi
earthquake in Taiwan as a case study. By discretizing the enormous rock avalanche
into many blocks, he was able to obtain quite accurately the post landslide topography
as mapped in the field.
Clearly, both the initial geometry of the discontinuous sliding mass as well as
the dynamic loading input should be considered simultaneously, to best simulate the
deformation mode and the final configuration of a sliding mass. Zhang et al. (2015)
applied DDA to the case of the Donghekou landslide which was triggered by the 2008
M = 8 Wenchuan earthquake that affected an area of 100,000 km2 with nearly 90,000
fatalities in Sichuan province, China. They studied the effect of seismic loading, the
number of blocks, and the block size on the kinematics of the deformation and on the
runout, namely on the final position of the slide after the event. Similarly, the formation
mechanism of the Tangjiashan landslide, also triggered by the same earthquake, was
studied with DDA by Wu et al. (2009). They were able to constrain the duration of
sliding (35 s), the peak velocity (30 m/s), to discuss the stress distribution along the
body of the slide during the event, and to assess the degradation of the shear resistance
during sliding.
The strong kinematic theory of DDA enables dynamic analysis of landslides that
also involve, in addition to sliding, rigid block rotations such as in rock falls and
toppling. Rock falls have been studied extensively by Ohnishi and coworkers with DDA
(Yang et al., 2004; Sasaki et al., 2005; Wu et al., 2005) and the results applied to real
rock slopes in Japan and elsewhere. The classical analytical solution for block toppling
(Goodman and Bray, 1976) has been checked and validated in 2D (Yeung, 1991; Yeung
and Goodman, 1992) and extended to 3D (Yagoda Biran and Hatzor, 2013) with two
dimensional and three dimensional DDA, respectively. The intriguing back-slumping
mode where the blocks rotate backwards while sliding forward was initially discussed
by Prof. Wittke (Wittke, 1965), and has been further explored with DDA (Kieffer,
1998; Goodman and Kieffer, 2000).
Existing rock slopes are stable under static conditions, naturally, and the major sta-
bility concern is what kind of external effects will destabilize the slope and will prompt
motion. External effects are typically driven by environmental changes. Changes in the
hydrology of the slope can induce pore pressure rise in the joints that may decrease the
frictional resistance of the siding planes (Hoek and Bray, 1981) and may even change
the failure mode of rock wedges from single plane to double plane sliding, or vice
versa (Hatzor and Goodman, 1997). Seasonal climatic changes may cause thermal
Rock slopes 201
expansion of wedges in the tension crack that may drive the block forward by a
wedging – ratcheting mechanism (Bakun-Mazor et al., 2013). Slow degradation of
shear strength along the sliding plane due to changing humidity conditions and water
chemistry may induce creep deformation along a potential sliding plane that can
culminate in rupture of rock bridges and runout (Kemeny, 2005).
These effects are time dependent with a period of days to years. Earthquakes, in
contrast, subject the slope to cyclic vibrations at much shorter periods of a second,
or less. The only way to consider such rapid loading changes are through a dynamic
analysis in the time domain, and DDA proves to be a capable method for doing just
that, as shown in the verification examples presented in Chapter 5 using sinusoidal
input motions. Seismically induced ground motions are never sinusoidal, but rather
contain a rich frequency content, but this does not constrain the analysis in any way
as it is performed in DDA in the time domain.
Outstanding issues pertinent to numerical rock slope analysis such as application
of viscous damping at the domain boundaries and at contact points, dynamic shear
strength degradation, dynamic block fracturing and pore pressure evolution during
sliding, are actively being studied presently by the DDA research community. Here we
shall focus on three issues that help illustrate the applicability of DDA to rock slope
engineering: rotational failure modes, dynamic input motion, and reinforcement.
Rock slopes exhibit several, frequently occurring, failure modes including single plane
sliding, double plane sliding, forward rotation or “toppling’’, and backward rotation
or “back-slumping’’. These failure modes involve the displacement or rotation of a
single block and typically limit equilibrium analysis can be applied to find the factor
of safety against failure. When the failing rock slope consists of multiple, interacting
blocks numerical, discrete element, approaches must be adopted and a factor of safety
cannot be clearly defined. Rather, the stability of the slope is discussed in terms of some
displacement tolerance. When the failure mechanism involves further decomposition
of blocks into smaller blocks during sliding as in large landslides or rock avalanches,
fracture mechanics failure criteria must be employed in conjunction with the robust
kinematical analysis performed with DDA.
Block sliding modes, either single plane or double plane, have been studied exten-
sively in the general literature, and DDA verifications for sliding modes were discussed
in Chapter 5. These modes, therefore, are not discussed here again, although in the
discussion of dynamic input motion later in this chapter single plane sliding is used to
illustrate some concepts. Large rockslides during which the sliding mass breakups into
multiple blocks, have also been studied extensively in the DDA literature as reviewed
in Chapter 1 and in the introduction to this chapter, and therefore are not discussed
here as well. We therefore limit the discussion in this section to failure modes involving
in-plane rotation: toppling, back-slumping, and overhanging slopes.
7.2.1 Toppling
The classical rock slope problem of block toppling as discussed by Goodman and Bray
(1976) involves a set of slender parallel blocks that are resting on a stepped base, with
202 Discontinuous deformation analysis in rock mechanics practice
Figure 7.1 Examples of toppling failures (after Goodman, 2013). A) Block Toppling, B) Flexural Toppling,
C) Block – Flexural Toppling. Illustrations courtesy of Prof. R. E. Goodman.
Figure 7.2 Definition of terms for limit equilibrium analysis of the block on an incline problem.
the dips of the boundary joints of the blocks pointing at high angles into the slope.
Recently, Goodman (2013) published a general classification of toppling failures as
found in the field, some examples of which are reproduced in Figure 7.1.
Ashby (1971) and later Hoek and Bray (1981) discussed the static limiting condi-
tions for a single block that rests on an inclined plane to undergo sliding, toppling, or
both sliding and toppling, using the parameters as shown in Figure 7.2. They presented
their results in a kinematic chart showing the boundaries between the failure modes as
a function of slope inclination α, friction angle φ, and the aspect angle of the block δ
(Figure 7.3A). Bray and Goodman (1981) who expanded the study to multiple blocks,
treated boundary 3 as a “dynamic’’ boundary, and modified it as shown in Figure 7.3B.
With the publication of DDA the problem of block toppling was revisited by Yeung
(1991) who found agreement between DDA and boundaries 1 and 2 and the modi-
fied boundary 3 as proposed by Bray and Goodman (1981), but found discrepancies
with regard to boundary 4. He found that in some cases, within the “sliding and top-
pling’’ region as delineated in Figure 7.3B, only toppling is obtained with DDA. It has
already been pointed out (Sagaseta, 1986) that the state of equilibrium at boundary 4
is dynamic, rather than static. Yeung (1991) treated boundary 4 as such and derived
a complete analytical solution for it, following Sagaseta (1986). This evolution of the
research demonstrates that when the numerical approach is accurate it can assist us in
finding the correct solution even to problems that actually have an analytical solution;
it becomes a guide of sorts, as the failure mode in DDA is not an assumption, but
rather a result, of the analysis.
Rock slopes 203
Figure 7.3 A) Boundaries between failure modes for block toppling as proposed by Hoek and Bray
(1981), B) Modified boundary 3 as proposed by Bray and Goodman (1981).
h b 1
N tan φ −N = m(h2 + b2 )θ̈ (7.3)
2 2 12
1
ü = θ̈ h2 + b2 (7.4)
2
The friction angle that will satisfy boundary 4 for every combination of α and δ is
found by solving the four equations (for complete derivation see Yagoda-Biran (2013)
or Yagoda Biran and Hatzor (2013)):
Figure 7.4 Modified boundary 4 (after Yeung, 1991). A) Block dynamics at boundary 4, B) Kinematic
conditions for sliding and toppling with modified boundary 4, for φ = 30◦ . Illustrations from
Yagoda-Biran (2013).
or:
The kinematic chart with the modified boundary 4 is shown in Figure 7.4B for
the case of φ = 30◦ . Indeed, Yeung (1991) obtained good agreement between 2D-DDA
and the modified boundary 4 as shown in Figure 7.4B.
Note that in order to present the kinematic chart in two dimensions one of the
three parameters α, δ, φ must be set at a fixed value, which limits the usefulness of
the presentation. Yagoda-Biran (2013) showed an original way to present the chart in
a three dimensional (α, δ, φ) space so that no single value of these three independent
parameters needs to be fixed, and her chart is reproduced in Figure 7.5A. The Stable
mode is above the red (α = δ) surface and to the left of the blue (α = φ) surface. The
Sliding mode is above the green (φ = δ) surface and to the right of the blue (α = φ)
surface. The Sliding and Toppling mode is below the green (φ = δ) surface and in front
of the curved surface representing Equation 7.6 (note that in this view the curved
surface is actually behind the green surface). The Toppling mode is below the red
(α = δ) surface and behind the curved surface representing Equation 7.6
In Figure 7.5B the boundaries are mapped as viewed from vector (−1, −1, −1)
which points from the object to the eye. When using this viewing direction, vector
(1, 1, 1) is reduced to a point, and the surfaces separating the different modes are
reduced to lines. In this mapping the 3D space appears as a 2D space where it is easier
to perceive the boundaries between the four modes.
Yagoda-Biran (2013) also considered pseudo-static earthquake inertia forces and
how they might affect the kinematic boundaries. Typically, the peak ground accelera-
tion (PGA) of the earthquake record is converted into a pseudo-static horizontal force
Rock slopes 205
90
80
70 90
60 80
70 sliding
50 stable
60
d
40 50
d
30 40
30
20 20
10 10 toppling sliding
0 and
0 0 toppling 0
80 20 20
60 40 40
40 60 f
f 80 a 60
20 60
20 40 80 80
0 0 a
A B
Figure 7.5 The kinematic chart proposed byYagoda-Biran (2013) for the four regions. A) Point of view
similar to Figure 7.4B, B) Isometric point of view, viewing vector (−1, −1, −1). See text for
explanation.
Figure 7.6 Force diagram for a block on an incline with pseudo- static force F. The hinge of rotation is
marked by a star (Yagoda-Biran, 2013).
F
tan β = =k (7.7)
W
Yagoda-Biran (2013) modified the kinematic chart using the angle β and defined
the kinematic boundaries for the case of pseudo-static loading F applied at the centroid
206 Discontinuous deformation analysis in rock mechanics practice
Table 7.1 Kinematic constraints assuming pseudo-static loading as defined by Yagoda-Biran (2013),
where ψ = α + β. The complete analytical derivation of the kinematic constraints for each
boundary can also be found in Yagoda-Biran and Hatzor (2013).
Figure 7.7 Verification of the modified static (A) and modified pseudo-static (B) kinematic charts
with 3D-DDA (Yagoda-Biran, 2013) using the isometric view where surfaces are reduced
to lines.
of the block. Her results, some of which required elaborate analytical derivations, are
summarized in Table 7.1.
The results of Yagoda-Biran (2013) imply that when a horizontal force F = kW
acts on the centroid of the block, the kinematic boundaries become a function of three
angles: φ, δ and ψ, instead of only α as in the case of gravitational loading alone. The
new angle ψ can also be expressed in terms of k instead of β:
k + tan α
ψ = tan−1 (7.8)
1 − k tan α
Figure 7.8 Changing failure modes with increasing input PGA as computed with 2D-DDA for two block
geometries: δ = 50◦ (square symbols) and δ = 15◦ (x symbols). For both geometries α = 10◦
and φ = 30◦ . The mode of the block changes with changing value of k, as indicated near the
symbols (Yagoda Biran and Hatzor, 2013).
change the failure mode of tetrahedral wedges from single to double plane sliding, or
vice versa.
Figure 7.9 Three categories of rock slumping modes as defined by Kieffer (1998). A) Flexural slump-
ing, B) Block slumping, C) Block flexure slumping (after Goodman and Kieffer, 2000).
Here the “bedding planes’’ are steeply dipping and the “cross-joints’’ are gently dipping.
Reproduced with permission from ASCE.
In any case, rotation around a center located outside of the block may take place in
this geometrical setting.
The forces acting on the block are shown in Figure 7.10. Assuming the friction
angles on the two sliding planes are equal (φ1 = φ2 ) three equilibrium equations are
necessary for solution of the contact forces N1 and N2 and the mobilized friction
angle φ, where α1 and α2 are the inclinations of the basal plane and the steep joint,
respectively (Kieffer, 1998):
Fv = 0 : W = N1 cos α1 + N1 tan φ1 sin α1 + N2 cos α2 + N2 tan φ2 sin α2 (7.9)
M0 = 0 : Wdw + N2 tan φ2 d2 = N2 d2 (7.10)
MC = 0 : Wx = N2 tan φ2 AC + N1 tan φ1 OC (7.11)
Consider now the block geometry shown in Figure 7.11A which is susceptible to
the block slumping mode as the weight vector of the block plots on the steeply inclined
plane. Simultaneous solution of the three equilibrium equations for the geometry of the
block in Figure 7.11A yields a mobilized friction angle value of φ = 22◦ . A 2D-DDA
model for the same block is shown in Figure 7.11B. Forward modeling of the block is
performed with 2D-DDA with input friction angles for the joints decreasing from 43◦
Rock slopes 209
Figure 7.10 Free body diagram used for static limit equilibrium analysis of the block slumping mode.
Figure 7.11 A) Block geometry used for limit equilibrium analysis, B) DDA model used for verification.
to 22◦ . In all these simulations the block remains stable under gravitational load.
However, once the input friction angle for the joints is reduced to 21◦ , block motion is
initiated, exactly as predicted by the limit equilibrium analysis for the modeled block
geometry. The motion of the block as computed with DDA clearly exhibits the block
210 Discontinuous deformation analysis in rock mechanics practice
Figure 7.12 DDA forward modeling with input friction angle for the joints of 20◦ . A) t = 0.8 s,
B) t = 1.6 s, C) t = 2.5 s (after Hatzor, 2003).
slumping mode where friction is mobilized simultaneously on both the steeply dipping
plane at the back as well as on the gently dipping plane at the base (Figure 7.12).
Note that once backward slumping is initiated joint water pressures will rapidly
dissipate, as a joint with a wide base and sharp edge at the top will form behind
the block at onset of motion. Consideration of water pressures and how they might
affect static stability in backward slumping is therefore less important in the analysis
of this failure mode. The formation of an A joint (terminology from R. E. Goodman,
personal communication) which is wide at the base and narrows towards the top, is
clearly evident in the final snap shot of the DDA simulation (Figure 7.12C).
Figure 7.13 A simple example of an overhanging slope in a horizontally layered and vertically jointed
rock mass where the tension crack is formed by coalescence of cross joints at a distance L
and B from the tip and the base of the block, respectively (afterTsesarsky and Hatzor,2009).
Reproduced with permission from ASCE.
The extruding slope and the vertical tension crack at the back form a polygon for
the geometry of which a structural kernel lies between B/3 and 2B/3 (Figure 7.14).
When the weight vector plots within the structural kernel the stress distribution at the
base of the block is uniformly compressive (Figure 7.14A). When the resultant plots
outside the kernel (Figure 7.14B) part of the base experiences tensile stresses. When
the resultant vector plots outside the polygon the block will overturn, provided that
the base is disconnected (Figure 7.14C).
The X coordinate of the center of mass (XCM ) of the polygon is:
3B · tan α + 23 h
XCM = (7.12)
h(B + h)
The relationship between the location of the center of mass (XCM /B) and the
eccentricity ration (B/L) and how this affects the expected failure mode of the slope
can be found analytically and the results are plotted in Figure 7.15. It can be seen from
212 Discontinuous deformation analysis in rock mechanics practice
Figure 7.14 Stress distribution and the base of an overhanging slope where the resultant plots
inside the kernel (A), outside the kernel (B) and outside the block (C), after Tsesarsky
and Hatzor (2009). Reproduced with permission from ASCE.
Figure 7.15 Analytical relationship between the location of the center of mass with respect to the base
of the slope (X cm /B) and the eccentricity ratio (er = B/L),afterTsesarsky and Hatzor (2009).
Reproduced with permission from ASCE.
inspection of Figure 7.15 that when the eccentricity ratio er is smaller than 0.38 the
slope is prone to forward rotation, and when er > 0.62 the slope is safe. The critical
geometry that requires further analysis is when the eccentricity ratio is between these
two bounds.
Rock slopes 213
Figure 7.16 Results of DDA for an overhanging slope with eccentricities ratio between 0.3 and 0.65
showing the displacement of the tip of the slope vs. number of time steps. Friction angle
of discontinuities 41◦ . After Tsesarsky and Hatzor (2009). Reproduced with permission
from ASCE.
DDA can be employed to perform such an analysis where the eccentricity ratio is
being changed between models. The results of such analyses are shown in Figure 7.16.
Indeed the expected relationship between the eccentricity ratio and the failure mode of
the overhanging slope is confirmed with DDA. The role of reinforcement in stabilizing
such a slope is discussed in the final section of this chapter.
In the previous section we used static and pseudo-static approaches to determine block
stability by means of limit equilibrium analysis. But life is really dynamic…and since
DDA is a dynamic method there is no reason to restrict ourselves to static or pseudo-
static approaches that only provide us with the state of equilibrium. With a truly
dynamic approach we can find in addition to the state of limiting equilibrium also
the rate and amount of displacement as well as the dynamic deformation pattern in
cases of multi-block problems. In order to perform a true dynamic analysis how-
ever, a time dependent loading function must be incorporated correctly in the loading
matrix instead of the constant gravitational acceleration. Methods to input the dynamic
motion into the analyzed domain, and what kind of input motions should be selected,
are explored in this section.
Figure 7.17 Record of the Loma Prieta earthquake as measured on a rock site in the San Francisco
Bay area.
track of the San Francisco Bay bridge in California. Shi used a modified record of the
Loma Prieta earthquake, which struck that region in 1989, the three components of
which are shown in Figure 7.17.
Seismometers that are attached to the ground typically provide motions in three
directions, one vertical and two horizontal, typically aligned in E–W and N–S direc-
tions. For a two dimensional analysis only one horizontal component is sufficient, and
therefore it would be required to find the resultant of the two horizontal components
in the direction of the analyzed cross section which need not necessarily be aligned in
one of those two directions. When running the dynamic simulation with direct input
of horizontal and vertical accelerations it becomes possible to assess the significance
of the vertical motions which are typically ignored.
and to assume the entire rock mass in which the underground opening is embedded
responds to the earthquake in the same manner.
In the case of surface structures that are resting on a shaking foundation such as
masonry structures, it would be more accurate to introduce the loading function, either
time dependent displacements, velocities, or accelerations, into a moving foundation
block that represents the shaking ground. This way the waves can propagate through
the block structure allowing for possible amplifications and ground-structure interac-
tions. This method which was used by Kamai and Hatzor (2008) in a verification study
of frictional block response to shaking foundation (see discussion in Chapter 5), has
been referred to as “Method 2’’ (Wu, 2010).
Kamai and Hatzor (2008) in their verification study used a flat responding block
so as to avoid possible rotations that are not considered in the analytical solution they
developed for this problem (see Figure 5.16). Therefore, wave propagations were not
studied in their research. Indeed, when they attempted to apply this loading method
to existing historic masonry structures that have experienced shaking in the past (the
Nabatean Mamshit town in the Negev and the Crusaders Nimrod castle in the Golan)
they encountered difficulties in obtaining with DDA the same structural deformation
pattern as mapped in the field. The way in which the connection between the block
representing the shaking ground and the overlying blocky structure is modeled with
DDA will certainly control the wave propagation pattern from the base to the top of the
masonry structure and consequently the resulting structural deformation. Therefore,
this issue of how to best represent a masonry structure with DDA for true dynamic
analysis requires further research.
In the case of landslides that are comprised of many blocks, it is not clear which of
the two methods is preferable, and indeed this issue has been a subject of some debate
in the literature. Clearly if one seeks to perform an analysis similar to the Newmark
method where the final displacement of a block resting on a pre-existing plane is sought,
then Method 1 as used in DDA would be the exact equivalent to Newmark’s method.
Then by extension in case of multiple blocks where an analytical solution does not
exist, Method 1 could be assumed to provide sufficiently accurate results.
This however is challenged by Wu (2010) who argues otherwise. Consider for
example the two-block assembly shown in Figure 7.18, similar in essence to the block
geometry used in the verification study of block response to cyclic motion of a frictional
interface in Chapter 5 (Figure 5.16). The initial configuration is shown in Figure 7.18A
where Point C in Block B and Point C in Block A are in contact. Assume a constant
acceleration is applied horizontally to the right. In Method 1, where the input accel-
eration goes into the centroid of Block B (Figure 7.18B), the final position of Point C
will be to the right of Point C . In Method 2, however, where loading is applied to the
foundation block A and Block B is allowed to slide in response to the induced motion
in Block A, the final position of Point C will be to the left of Point C (Figure 7.18C).
Wu (2010) argued that this is in conflict with d’Alembert’s principle of mechanics, and
therefore Method 1 should not be used in such cases.
Ning and Zhao (2012) studied extensively dynamic displacement of blocks on
frictional planes with DDA and provided analytical solutions to a variety of cases
including consideration of the role of vertical motions. With regard to the conflict
with d’Alembert’s principle of mechanics when applying Methods 1 as argued by Wu
(2010), Ning and Zhao (2012) note that the relative displacement between the two
216 Discontinuous deformation analysis in rock mechanics practice
Figure 7.18 Comparison between loading methods 1 and 2 under constant acceleration acting to the
right. (A) Initial block configuration, (B) Final position of point C in Method 1, (C) Final
position of point C in Method 2 (after Wu (2010)).
blocks, after all, is the same in the two methods. Therefore, if the input motion is
applied to the base block in the opposite direction in Method 2, then the resulting
position of Block 1 would be the same as if it was loaded in Method 1. They conclude
therefore that the two methods are equivalent for single block sliding, but care has
to be taken with regard to the direction of application of the dynamic load. If the
existing record was measured on bedrock, equivalent in our analogy to Block A, then
when the record is applied to the sliding mass directly (Method 1) as a volume force,
it should be applied in the opposite direction. If however the record is applied to the
base (Method 2), then it should be applied as constraint displacement time histories
and it can retain the same directions as measured in the ground during the earthquake.
Figure 7.19 Experimental topographic effect measurement campaign at Masada. A) Topographic map
and location of geophones, B) Cross section A-A (Zaslavsky et al., 1998).
using ambient seismic noise, b) reference station technique, and c) receiver function
estimates based on earthquake data.
Individual and average horizontal-to-vertical spectral ratios for Sites 1 and 2 as
obtained from micro-tremors are shown in Figure 7.20. The dominant feature of all
spectral ratios is the high spectral ratio level at a frequency of about 1.4 Hz. At this
frequency we also observe differences between the EW and NS components. Such
differences are characteristic of topographic effects. At the summit of Mt. Masada, the
average spectral ratios reach maxima of about 2.5 in the EW direction and about 2
in the NS direction. It should be pointed out, however, that the Nakamura method
provides in general a relatively reliable estimate of the predominant frequency of the
site (resonance frequency) but it is less reliable for estimating the amplification level,
especially at other frequencies.
The magnitude of spectral amplification is best estimated when comparing between
sites at the top of the mountain and a reference site at the base (the reference station
technique). In our case this would be between stations 1 and 2 at the top and station
no. 4 at the base (see Figure 7.19B). The comparison between the top and base of
the mountain was performed by Zaslavsky and Shapira (1998) utilizing a recorded
earthquake of magnitude ML = 2.9 and at a distance of 545 km from the site that struck
Southeast Cyprus during the time of the survey at Masada. Only small variations in the
site response of the two sites are indicted (Figure 7.21). The spectral amplification ratios
show a prominent peak at about 1.3 Hz. The horizontal ground motion oriented EW is
amplified by a factor of about 3.5, while it is about 2.0 in the NS direction. These results
clearly indicate that Mt. Masada exhibits a preferential direction of resonance motion.
Finally, the horizontal-to-vertical spectral ratios for Sites 1 and 2 for the S wave
window during the earthquake are plotted in Figure 7.22. The receiver function clearly
exhibits resonance mode in frequency range 1.2–1.4 Hz, with amplification values of
about 3.5.
218 Discontinuous deformation analysis in rock mechanics practice
Figure 7.20 Individual and average (heavy lines) horizontal-to-vertical spectral ratio for Mt. Masada
obtained from microtremors recorded at Sites 1 and 2 shown in Figure 7.19 (Zaslavsky
et al., 1998).
The final empirical response function for Masada as developed by Zaslavsky et al.
(1998) is shown in Figure 7.23. Three characteristic modes are found, at frequencies
of 1.06, 3.8, and 6.5 Hz.
We have shown the pronounced topographic site effect measured experimentally
at Masada to demonstrate that such local effects may be rather significant and should
therefore be incorporated in dynamic analysis if realistic results are sought.
Rock slopes 219
Figure 7.21 Spectral ratios for Sites 1 and 2 with respect to Site 4 computed from an earthquake
the struck Cypress during the survey and was recorded at the site (Zaslavsky et al., 1998).
Figure 7.22 Horizontal-to-vertical spectral ratios obtained from earthquake data for Sites 1 and 2 at
Mt. Masada (Zaslavsky et al., 1998).
Figure 7.23 The empirical topographic response spectra from Masada as computed by Zaslavsky et al.
(1998).
Alternatively, one could use the original record without any modifications, and apply
it to the base of the model in a Method 2 type of loading, and then let the waves prop-
agate upwards in the mesh during the numerical simulation. Taking such an approach,
however, will require generating a mesh that will represent the entire mountain with
its particular geological structure. For the case of Mt. Masada for example, with the
250 m high topography and a reasonable cross section width of at least 500 meters
(see Figure 7.19), this will result in a block system comprised of more than 100,000
blocks. This estimated number of blocks in the modeled domain is obtained if only
one horizontal and one vertical joint sets are assumed, each with a mean spacing of
1 m. The reality in the field is more complicated of course. Computing such a mesh in
loading method 2 would require an enormous computational effort that is not avail-
able for standard computational platforms. Therefore if the response function of the
site is known, it would be much more efficient to apply the loading function in Mode
1 directly at the top of the mountain, without compromising accuracy.
To demonstrate this approach consider the record of the M = 7.1 Nuweiba earth-
quake which struck the Gulf of Aqaba in 1995. The motion was recorded in the city of
Eilat, some 100 km north of the epicenter in a measurement station that was located
on 50 meters of alluvial sediments (Figure 7.24). Zaslavsky et al. (2000) performed
“deconvolution’’ of the measured record to obtain an analytical rock response function
using information they collected from nearby boreholes with regard to the dynamic
properties of the strata which are required for this analytical procedure. The rock
record thus obtained is shown in Figure 7.25.
Rock slopes 221
Figure 7.24 Record of the Nuweiba M = 7.1 Earthquake of Nov. 1995 as recorded in the city of Eilat,
some 100 km north of the epicenter.
In a dynamic analysis of Masada rock slopes, Hatzor et al. (2004) used the rock
record developed by Zaslavsky et al. (2000) for the Nuweiba earthquake as measured
in Eilat, and performed convolution with the experimentally measured topographic
response spectrum at Masada (Figure 7.23). They have thus obtained the expected
ground motion at the top of the mountain, should a similar earthquake hit the region
in the nearby, seismically active, Dead Sea rift valley assuming such an earthquake will
have similar characteristics as the Nuweiba earthquake (Figure 7.26).
With the available dynamic motion for the top of Masada it is now possible to
perform dynamic DDA analysis for block at the top of the mountain in Method 1, with-
out having to simulate the dynamic deformation of the whole mountain in Method 2,
without compromising accuracy.
Figure 7.26 Convolution of the Nuweiba “rock response’’ record with the experimentally measured
Masada topographic response spectrum for obtaining the input motion to be used at the
top of the mountain.
Rock slopes 223
Figure 7.27 Example of a block that is separated from the rock mass by a tension crack filled with
debris and resting on an inclined base at the top of the Snake-Path cliff, Mount Masada
(Bakun-Mazor, 2011). The block is held in place by virtue of friction alone.
any rock slope expected to experience amplifications due to topography during strong
earthquakes. Consider the prismatic block that was mapped at the top of the East face
of Mt. Masada in a location locally known as the “Snake Path’’ Cliff (Figure 7.27). We
shall refer to this block as Block 1 in the discussion that follows. The Block rests on a
bedding plane dipping 19 degrees to ESE and has separated over time from the rock
mass by opening of a vertical joint with accumulated siding to a distance of 200 mm
over time. In a different study we explore the role of thermal fluctuations on the
opening of the joint and sliding of the bock in a thermally induced wedging – ratcheting
mechanism (Bakun-Mazor et al., 2013). Here we shall only consider motion of this
block due to seismic vibrations. Our goal would be to try to constrain the earthquake
magnitude in the seismically active Dead-Sea rift valley below Masada that could have
caused the mapped block displacement of 200 mm.
Because the block has separated from the rock mass, it is held in place by virtue
of friction alone. In order to establish the factor of safety against sliding in static
conditions the available friction angle of the bedding plane has to be determined.
224 Discontinuous deformation analysis in rock mechanics practice
Figure 7.28 Triaxial and direct shear test results of a bedding planes at Masada. The triaxial tests
are performed on inclined saw-cut planes and therefore provide the residual frictional
resistance (φres = 23◦ ). The direct shear tests are performed or rough bedding planes and
dilation is permissible during the tests. Therefore the results provide the peak friction
angle for the rough joints (φpeak = 41◦ ).
Triaxial tests of an inclined saw-cut filled with typical infilling material found in the
field inside the bedding planes yielded a residual friction angle of 23 degrees, whereas
direct shear tests performed on rough surfaces under an imposed constant normal
stress yielded a peak friction angle of 41◦ (Figure 7.28).
With the geometry of the block and the frictional resistance established, we can
proceed with forward dynamic analysis with DDA. In order to do this accurately we
need to first select a good value for the contact spring stiffness to be used in the dynamic
DDA simulations. This was done by Bakun-Mazor (2011) by subjecting the block to
sinusoidal input motions in loading method 1, namely the time-dependent acceler-
ations were applied at the centroid of the block. Two input frequencies were used,
corresponding to the two resonance modes determined experimentally for Masada,
1.3 Hz and 3.8 Hz, because the optimal penalty value is sensitive also to the frequency
of the input vibrations. The analysis was repeated with changing contact spring stiff-
ness values until the minimum error was obtained for the two frequencies. A peak
friction angle of 41 degrees was assumed for the sliding interface. The results are
shown in Figure 7.29. A contact spring stiffness of k = 500 GN/m was used as input
for forward modeling with DDA, with a time step of 0.005 s. The method for choosing
the optimal stiffness value using the relative errors obtained with the two frequencies
is discussed by Bakun-Mazor et al. (2013).
The horizontal yield acceleration (ayield ) for an inclined bedding plane dipping
19◦ with peak friction angle of 41◦ as in the case of Block 1 is readily obtained
using pseudo-static analysis (see Goodman and Seed, 1966) at ayield = 0.404 g. The
ayield value thus obtained constrains the epicenter location of the maximum expected
earthquake (Mw = 7.5) at the Dead Sea rift capable of triggering sliding of Block 1, to
a distance of up to 20 km from Masada (see Figure 7.30).
Rock slopes 225
Figure 7.29 Finding the optimal contact spring stiffness for forward analysis by comparing DDA results
with the analytical (Newmark’s) solution for the two resonance modes measured for the
top of Masada (Figure 7.23). A friction angle of 41◦ is assumed for the base plane. The yield
acceleration is 0.404 g (Bakun-Mazor, 2011).
DDA results for seismic analysis of Block 1 subjected to amplified Nuweiba records
corresponding to Mw = 6.0, 6.5, 7.0, 7.5 Dead Sea rift earthquakes at an epicenter dis-
tance of 1 km from Masada are shown in Figure 7.31 (Bakun-Mazor et al., 2013). For
moderate earthquakes (Mw ≤ 6.5) the block displacement per single event is expected to
be lower than 42 mm, whereas for strong earthquakes (Mw ≥ 7.0) the block is expected
to slide more than 447 mm along the inclined bedding plane in a single event. The
mapped opening of the tension crack in the field is only 200 mm (see Figure 7.27), and
this value constrains feasible earthquake scenarios that could have struck the region
since this block was situated in its current geomorphological setting.
Figure 7.30 Attenuation relationships for the Dead Sea fault adjacent to Mount Masada based on
Boore and Joyner (1997) empirical relationship between earthquake magnitude, distance,
and peak ground acceleration in rock sites. The dashed lines are without considering
topographic amplification. Solid lines are with topographic effect and are relevant to the
top of Masada. Shaded region indicates static stability of the modeled block at the top of
Masada (Block 1). After Bakun-Mazor et al. (2013).
Figure 7.31 Plastic displacement of Block 1 as computed with DDA for different earthquake magni-
tudes assumed to take place at a distance of 1 km from Masada. This chart indicates that
slope could not have experienced an earthquake magnitude greater than 6.5 in its current
geometrical configuration. After Bakun-Mazor et al. (2013).
Rock slopes 227
Figure 7.32 Dynamic deformation of the Upper terrace of the North face of Masada. Top – response
to the record shown in Figure 7.26 scaled to PGA = 0.2 g with no rockbolt reinforcement,
Bottom – response to the record shown in Figure 7.26 scaled up to PGA = 0.6 g with spot
bolting applied only in the West slope.
provided that the bolts are loaded within their yield strength and good anchoring is
ensured when installed in the field. The dimensioning of the bolt in the field, in terms
of bolt length and diameter, can be represented in DDA by the corresponding bolt stiff-
ness that goes into the code as an input parameter, where typically a Young’s modulus
of steel is assumed (kbolt = AE/L).
The bolt action to restrain block motions due to dynamic loading is well illustrated
by the case of upper terrace in the North face of Masada discussed, the geological
structure of this wad discussed in Chapter 4. The cross section shown in Figure 4.24B
was subjected to the earthquake record shown in Figure 7.26 in Mode 1, and the
response of the cross section to the input record when scaled to PGA = 0.2 g is shown
in upper panel of Figure 7.32. It can be clearly seen that given such an earthquake the
228 Discontinuous deformation analysis in rock mechanics practice
35m
0m
0m 13m
Figure 7.33 The “Gibborim’’ overhang in Haifa near the Carmel tunnels. Left – cross section,
Right – photo.
West slope will exhibit toppling failure, endangering the visitors on the path that is
located along the West slope just below the upper terrace. However, installing a simple
pattern of rockbolts as shown in the lower panel of Figure 7.32 will arrest any such
toppling, even if the input record is up scaled to PGA = 0.6 g, due to the expected
topographic effect in Masada (Figure 7.23).
Consider now the “Gibborim’’ overhang in Haifa near the Carmel tunnels project
with cross section and photo shown in Figure 7.33, that was studied by Tsesarsky
and Hatzor (2009). Forward rotation of the slope is anticipated if a tension crack
is assumed at a distance of 5 m from the foot of the slope (B = 5). Indeed, forward
rotation of the slope is indicated with DDA, as shown in Figure 7.34. A friction angle
of 37◦ is assumed both for the horizontal and vertical joints, and therefore initially the
overhanging polygon rotates as a single rigid body under gravitational loading, as no
connection is provided at the base of the slope. This is a realistic assumption as rock
joints in general possess very little, if any, tensile strength.
The forward rotation of the overhang can be arrested if a rockbolting pattern
that extends beyond the assumed location of the tensile crack is installed. This is
demonstrated in Figure 7.35 where the deformation of the slope with the installed
rockbolts after 20 seconds of gravitational loading is shown.
A very useful byproduct of DDA with rockbolts is the tension force evolution in
every bolt as the analysis progresses in the time domain. The developed tension in the
bolt will depend on the assumed stiffness of the bolt, which relates to the installed bolt
diameter. The tension evolution in the bolts when a bolt diameter of 2 inch is assumed
is plotted in Figure 7.36 where the bolts are numbered from No. 1 at the foot of the
slope to No. 9 at the slope tip. The extension of the bolts can be appreciated from this
Figure 7.34 Deformation of the Gibborim overhang as computed with 2D-DDA if no rockbolt
reinforcement is applied (assumed joint friction angle is 37◦ ).
Figure 7.35 Designed rockbolting reinforcement pattern for the Gibborim overhang with an assumed
tension crack at a distance from the base of B = 5 m.
230 Discontinuous deformation analysis in rock mechanics practice
Figure 7.36 Bolt tension evolution computed with DDA for the Gibborim overhang. Bolt stiffness is
calculated for steel Young’s modulus of 200 GPa, bolt diameter of 2 inch, and respective
bolt length in each position. No interaction between rockbolt and rock mass is consid-
ered, namely the bolt is fully bonded throughout its entire length (after Tsesarsky and
Hatzor, 2009). Reproduced with permission from ASCE.
plot and their stabilizing effect that can be inferred once the tension forces in all bolts
approach constant values, depending on the location of the bolt in the slope.
The role of bolt stiffness is clearly portrayed in Figure 7.37 where the displacement
evolution of the slope tip is plotted for the two bounding eccentricity ratios with no
reinforcement applied, and then with three values of installed bolt diameters from 1 to
3 inch representing increasing bolt stiffness. Clearly, with increasing bolt stiffness the
displacement of the slope is restrained. With all three modeled diameters the forward
rotation is arrested, and therefore the decision of which bolt diameter to use in the field
becomes an engineering decision involving both economic and practical considerations.
The increasing anchoring force with increasing elevation of the bolt in a block
undergoing forward rotation has also been observed by Yagoda Biran and Hatzor
(2013) who computed the anchoring force with 2D-DDA for a block that either slides
or topples, depending on the input friction angle (Figure 7.38). The observed difference
in developed bolt tension between sliding and toppling modes reveals once again the
insight that could be gained from the results of proper numerical analysis where the
obtained failure mode is a final result, and not an initial assumption, as in DDA.
Because the bolt connection is modeled in DDA as a stiff spring, there is a linear
relationship between the amount of extension in the bolt and the resulting tension
force that will be developed through the assumed input stiffness of the bolt in the
Rock slopes 231
Figure 7.37 The role of increasing bolt stiffness (via increasing bolt diameter in reality) on slope defor-
mation for the Gibborim overhang, as expressed by the displacement evolution of the
slope tip with time steps. The dashed lines are the computed displacement of the tip
without rockbolt reinforcement for the two boundary eccentricity ratios (after Tsesarsky
and Hatzor, 2009). Reproduced with permission from ASCE.
Figure 7.38 Comparison between the anchoring force required to restrain motion in sliding and top-
pling modes using 2D-DDA in static loading. The same block geometry is used in both
simulations, the change in mode is induced by changing the input friction angle from φ = 15◦
(sliding) to φ = 30◦ (toppling) (after Yagoda Biran and Hatzor, 2013).
232 Discontinuous deformation analysis in rock mechanics practice
analysis. In the sliding mode the block undergoes displacement, as a single rigid body,
of equal amount everywhere in the block. This is why the developed anchoring force is
not related to the position of the anchor in the block; It would be the same everywhere
in the block as long as the sliding mode is preserved throughout the analysis. In the
toppling mode, however, the block undergoes forward rotation where the horizontal
component u increases with increasing vertical position in the block. Consequently,
the bolt tension will increase from the base to the top of the block, and as can be
inferred from Figure 7.38, in a linear trend. This result implies that in overhanging
slopes where forward rotation is suspected, attention must be paid to the expected
loading of the bolts with increasing elevation in the slope, to ensure that the design
capacity of the bolt is not exceeded in the higher most areas of the slope. Such an
assessment can readily be performed with DDA as we have shown here.
Chapter 8
EDITORIAL NOTE
This chapter presents the new contact theory developed by Dr. Gen-hua Shi and has
been written exclusively by Dr. Shi especially for this book.
FOREWORD
Contacts between two general blocks are the fundamental problem for discontinuous
analysis. There are different contact points in different block positions, and there
may be infinite contact point pairs in the same block position. A new concept of
“entrance block’’ is introduced in this chapter for solving the contacts between two
general blocks. The boundary of an entrance block is a contact cover system. Contact
covers may consist of contact vectors, edges, angles or polygons. Each contact cover
defines a contact point and all closed-contact points define the movements, rotations
and deformations of all blocks as in real cases. Given a reference point, the concept of
entrance block simplifies the contact computation in the following ways:
1 The shortest distance between two blocks can be computed by the shortest distance
between the reference point and the surface of the entrance block.
2 As the reference point outside the entrance block moves onto the surface of
entrance block, the first entrance takes place. This first entrance point on the
entrance block surface defines the contact points and related contact locations.
3 If the reference point is already inside the entrance block, it will exit the entrance
block along the shortest path. The corresponding shortest exit point on the
entrance block surface defines the contact points and related contact locations.
All blocks and angles here are defined by inequality equations. Algebraic operations
on blocks and angles are described here. Since the blocks and angles are point sets with
infinite points, the geometric computations are difficult, and therefore the geometric
computations are performed by related algebraic operations.
8.1 INTRODUCTION
of discontinuous computation is to detect where and when contacts happen, which are
the major objectives of this chapter. This is one of the most difficult issues in numer-
ical computation involving discontinuities and the simulation of sliding, holding, and
impacting. As a result, contact computation has been drawing much attention from
many scholars in the past decades.
A “distinct object’’ as used here is referred to as a “block’’ in the aspect of geome-
try. A block system (Poetsch, 2011) contains more than one block, but only when two
blocks are close enough can they possibly contact and these two blocks are then defined
as neighboring blocks. Before computing contacts, algorithms for finding neighboring
blocks (e.g. Belytschko and Neal, 1991; Wu et al., 2014) should be used in contact
analysis.
For the contact problem of two neighboring blocks, several methods, including
the common plane (Cundall 1988; Nezami et al., 2006), the single surface algorithm
(Benson and Hallquist 1990), hierarchy territory algorithm (Zhong and Nilsson 1996;
He 2010), the “master-slave’’ approach (Jelenić and Crisfield 1996), and the block-
particle algorithm (Li et al., 2004) were proposed.
These methods have solved many contact problems (Konyukhov and Schweizerh
2013), such as the contact between rounded convex angles, balls, 2D convex angles,
and 3D planes. However, it is more advantageous that a global contact theory, one
that is topologically proven and applied to the objects with arbitrary shapes, should
be developed. With the use of such global contact theory, difficult contact problems,
such as the contact between symmetric angles, the rotation of parallel edges, the sep-
aration of two contacted polygons, the contact between two concave blocks, the
contact of blocks with small edge length and under large penetration distance, and
the destabilization process of a densely compacted block system, can be easily solved.
The difficulty of contact judgment is associated with the arbitrary shape of the
boundary and the infinite interior points. In order to tackle contacts, appropriate
mathematical tools are needed for transforming geometric and topological computa-
tions to algebraic computations. Blocks and the components of blocks, such as solid
angles, polygons, and edges, are defined as point sets. The operation of simple point
sets has been provided by Minkowski and Geometrie der Zahlen (1910). However, the
mathematical tool for dealing with complex point sets is still lacking.
In this study, new mathematical tools are first proposed and used to operate point
sets (Sections 8.2 and 8.3). A new concept “the entrance block’’ is proposed and
associated with an operation of point sets (Section 8.3). With the use of the entrance
block, the relationship between the two blocks reduces to the relationship between a
reference point and the entrance block, thus considerably simplifying the complexity
of contact computation (Section 8.4). The entrance block is defined and represented by
its boundaries and conforms with many theorems (Section 8.4). With these theorems,
the entrance blocks in different conditions are formed and discussed in Sections 8.5
to 8.12, including contacts between 2D angles, 2D blocks, 3D angles, and 3D blocks.
These solids can be of any shape, i.e. convex or concave, with rounded or sharp
angles. In Sections 8.5 to 8.12, the solutions of the entrance block both in algebraic
and geometrical ways are proven. In fact, the solving of the entrance block is to build
and solve inequality equations that employ point sets.
Using the entrance block and its contact covers, the possible contact positions and
degrees of freedom can be identified. Thus the problem as to where to apply possible
and suitable contact springs or penalties is solved.
Shi’s new contact theory 235
To date, the idea of contact theory (Shi 2013a, 2013b) has been fully devel-
oped and employed. In this theory, the geometrical and topological computation is
transferred into algebraic computation. Subsequently, the proposed contact theory
becomes computational. Most of the computation methods, including discontinuous
deformation analysis (Shi 1988), discrete element method (Cundall 1988; Li et al.,
2004; Nezami et al., 2006), numerical manifold method (Shi 1991), and finite element
method (Goodman et al., 1968; Taylor 2004) can use this theory to compute contacts.
A, B are blocks.
∂A, ∂B are the boundaries of blocks A, B respectively.
int(A) is the set of inner points: x ∈ A, and x ∈
/ ∂A.
x, y, z are values or real numbers.
236 Discontinuous deformation analysis in rock mechanics practice
normal vector of a block or polygon, which points the inside of the block and
polygon.
i
n nj means n i = t
nj .
ni ↑↑ nj means ni = t nj , t > 0.
The computation of a and b follows the vector operation rule (as shown in
Figures 8.1 & 8.2),
a ± b = (xi ± xj , yi ± yj , zi ± zj )
A − B = A + (−B).
Therefore:
5
A − B = A + (−B) = (a − b).
a∈A,b∈B
A one-dimensional line A along e 1 passing (0, 0, 0) is represented as:
↔
5
e1 = t e 1.
∀t
A 2D solid angle e 1 e 2 means the swept area of a ray when rotating the ray from
vector e 1 to e 2 along the direction from ox to oy, i.e. anti-clockwise direction. Vectors
e 1 and e 2 are the edge vectors of solid angle e 1 e 2 (Figures 8.3 & 8.4).
t1 ≥0,t2 ≥0
e 1 × e 2 (0, 0, +1).
e 1 × e 2 (0, 0, −1).
238 Discontinuous deformation analysis in rock mechanics practice
− e 1 e 2 = (− e 1 )(− e 2 ).
e 1 e 2 · · · e u−1 e u .
Shi’s new contact theory 239
(Figures 8.5 & 8.6). The boundary of 3D solid angle e 1 e 2 · · · e u−1 e u is 2D solid
angles e r e r+1 . Solid angle e r e r+1 rotates from e r to e r+1 in right-hand rule. The
n12 = e r+1 × e r .
240 Discontinuous deformation analysis in rock mechanics practice
Figure 8.7 A 3D concave solid angle (left) and a 3D convex solid angle (right).
If e r e r+1 is a concave angle (Figure 8.6), its inner normal vector (Figure 8.7):
n12 = e r × e r+1 .
− e 1 e 2 · · · e u−1 e u
⊥ n1r = {x · n1r = 0}
where n1r is the normal vector of the plan A (Figure 8.8).
Shi’s new contact theory 241
Half space A passing (0, 0, 0) can be represented as the set of x satisfying the
equation (Figure 8.8),
r=1
6
u
= ↑ n1r .
r=1
242 Discontinuous deformation analysis in rock mechanics practice
When u = 2,
6
A =↑ n11 ↑ n12
a1 a2 · · · ap−1 ap
where ap+1 = a1 and a1 a2 · · · ap−1 ap ap+1 rotates in the right-hand rule, i.e. the rotation
axis points outward to block A (Figures 8.12–14).
This polygon is also represented by its boundary
Denote a1k is any vertex on polygon face k, the inner normal vector m1k of face k
points the inside of block A (Figure 8.12) and
(a1k ai ) = a i .
The inner vector m1k can be defined as,
p
m1k = a i+1 × a i .
i=1
244 Discontinuous deformation analysis in rock mechanics practice
e1 e2 · · · eu−1 eu , eu+1 = e1 .
Define:
(eer ) = e r .
e = e 1 e 2 · · · e u−1 e u .
Similar to the condition of 3D solid angle (Figure 8.10), the inner normal vectors of
boundary solid angle e r e r+1 are n1r and point the inside of block A.
A(0), B(0).
Denote the following as the set of all boundary vectors or edges of block A and B
respectively
A(1), B(1).
Shi’s new contact theory 245
Denote the following as the set of all boundary 2D solid angles or polygons of block
A and B respectively
A(2), B(2).
Ai ⊂ ∂Ai .
where m1k is the inner normal vector of face k and points the inside of A (Figures 8.15 &
8.16).
246 Discontinuous deformation analysis in rock mechanics practice
The complicated contact conditions between two general blocks A and B can be simpli-
fied as the relations that a referring point is located inner, outer, or lies on the boundary
of an entrance block. The entrance block is also represented by inequality equations.
The computation of an entrance block is geometric computation. The way of
computing entrance block here is to transfer geometric computation to algebraic
computation.
There are three different kinds of relation, i.e. contact conditions, between two blocks
A and B.
Case 1: A and B are separated (Figure 8.17): A ∩ B = ∅
In this condition, the distance between A and B is: |A, B| = ε > 0,
Case 2: A and B are overlapped, i.e. penetrated (Figures 8.18 & 8.19) A ∩ B = ∅
Case 3: A and B contacted (Figures 8.20 & 8.21)
A ∩ B = ∅, A ∩ B ⊂ (∂A ∩ ∂B) ⇔
A ∩ B = ∅, int(A ∩ B) = ∅.
248 Discontinuous deformation analysis in rock mechanics practice
A ∩ B ⊂ (∂A ∩ ∂B) = ∅.
To know these different cases, the “entrance block’’ is defined in the next section.
The reference point a0 moves together with block A. Moving block A without rotation,
every point of E(A, B) is a position of a0 while block A and block B have common
points.
Proof:
E(A, B) = E(a0 , B) = B − a0 + a0 = B.
A = {|x − a0 | ≤ r}.
250 Discontinuous deformation analysis in rock mechanics practice
Figure 8.22 The entrance block E(A,B) remains the same after parallel movements of block A.
Proof:
|a0 , B| ≤ r ⇔
∃ b ∈ B, |a0 − b| ≤ r ⇔
∃ b ∈ B, b ∈ A ⇔
∃ b ∈ A ∩ B ⇔ a0 ∈ E(A, B).
Shi’s new contact theory 251
A = {|x − a0 | ≤ r1 },
B = {|x − b0 | ≤ r2 }.
Denote ball
Proof:
∀ a ∈ A, b ∈ B, E(a, b) = b − a + a0 ,
|E(a, b) − b0 | = |(b − b0 ) − (a − a0 )|
≤ |b − b0 | + |a − a0 | ≤ r2 + r1
⇒ E(A, B) ⊂ C.
If
|c − b0 | ≤ (r1 + r2 ),
∃ a ∈ A, b ∈ B, b = a − a0 + c ⇒
E(A, B) ⊃ E(a, b) = b − a + a0 = c
⇒ E(A, B) ⊃ C.
252 Discontinuous deformation analysis in rock mechanics practice
Figure 8.25 The Entrance ball of the contact between two balls.
Therefore
E(A, B) = C.
This is a basic formula of the ball to ball contact (Figures 8.25 & 8.26).
Shi’s new contact theory 253
Proof:
A ∩ B = ∅ ⇔ ∀a ∈ A, ∀b ∈ B, a = b
⇔ ∀a ∈ A, ∀b ∈ B, b − a + a0 = a0 ⇔
a0 ∈
/ E(A, B).
Proof:
|A, B| = ε ⇔ ∃ a ∈ A, b ∈ B, |a, b| = ε,
∀a ∈ A, ∀b ∈ B, |a, b| ≥ ε ⇔
∃ a ∈ A, b ∈ B, |a0 , b − a + a0 | = ε,
∀a ∈ A, b ∈ B, |a0 , b − a + a0 | ≥ ε
⇔ |a0 , E(A, B)| = ε.
(A ∩ B) ⊂ (∂A ∩ ∂B) = ∅,
∃ ε > 0, ∀|δ| < ε, (A + δ) ∩ B = ∅ ⇔
(A ∩ B) ⊂ (∂A ∩ ∂B) = ∅, a0 ∈ int(E(A, B)).
Shi’s new contact theory 255
Figure 8.29 E(A,B) and E(B,A), where block A is a triangle and block B is a concave polygon.
Proof:
(A ∩ B) ⊂ (∂A ∩ ∂B) = ∅,
∃ ε > 0, ∀|δ| < ε, (A + δ) ∩ B = ∅ ⇔
(A ∩ B) ⊂ (∂A ∩ ∂B) = ∅,
∃ ε > 0, ∀|δ| < ε, ∃a ∈ A, b ∈ B, a + δ = b,
b − a + a0 = a0 + δ, a0 + δ ∈ E(A, B) ⇔
(A ∩ B) ⊂ (∂A ∩ ∂B) = ∅,
a0 ∈ int(E(A, B)).
The definition of entrance block transfers the entrances and contacts of two blocks
to the entrances and contacts of one point and one block.
256 Discontinuous deformation analysis in rock mechanics practice
Figure 8.30 Computed E(A,B) where blocks A and B are the same concave polygons.
Figure 8.31 E(A,B) and E(B,A) where blocks A is a triangle and B is a rectangle.
Shi’s new contact theory 257
Figure 8.32 Computed E(A,B) where block A is a convex polygon and B is a concave polygon.
Figure 8.33 Computed E(A,B) where blocks A and B are the same concave polygons.
As shown in Figures 8.29 & 8.31, E(A, B) and E(B, A), they are centrally symmet-
rical. Similarly, in Figures 8.30 & 8.33, E(A, B) are also centrally symmetrical in the
conditions that A, B are of the same shape.
In the discontinuous computation, the entrance blocks are required. The following
theorems will make the computation of entrance blocks more efficient.
a = b ⇔ E(a, b) = a0 .
Proof:
a = b ⇔ E(a, b) = b − a + a0 = a0 .
7
n
E(A, B) = E(A(m), B(n − m)),
m=0
E(A, B) = E(∂A, B) ∪ E(A, ∂B).
Shi’s new contact theory 259
Proof:
∀a ∈ A, ∀b ∈ B,
∃ δ, a + δ ∈ A(m) ⊂ A, b + δ ∈ B(n − m) ⊂ B ⇒
E(a, b) = E(a + δ, b + δ) ⊂ E(A(m), B(n − m)) ⇒
7n
E(A, B) = E(A(m), B(n − m)) ⇒
m=0
E(A, B) = E(∂A, B) ∪ E(A(n), B(0)) ⇒
E(A, B) = E(∂A, B) ∪ E(A, ∂B).
This theorem indicates that E(A(m), B(n − m)) forms a finite cover system on E(A, B).
If A1 ⊂ A, B1 ⊂ B then
E(A1 , B) ⊂ E(A, B),
E(A, B1 ) ⊂ E(A, B).
Proof:
A1 ⊂ A → E(A1 , B) = B − A1 + a0
⊂ B − A + a0 = E(A, B)
⇒ E(A1 , B) ⊂ E(A, B).
B1 ⊂ B ⇒ E(A, B1 ) = B1 − A + a0
⊂ B − A + a0 = E(A, B)
⇒ E(A, B1 ) ⊂ E(A, B).
260 Discontinuous deformation analysis in rock mechanics practice
Proof:
E(A, B) = B − A + a0
= −(A − B + b0 ) + (a0 + b0 )
= −E(B, A) + (a0 + b0 ).
The entrance block E(A, B) and entrance block E(B, A) are centrally symmetrical
(Figures 8.38 & 8.39).
The condition:
Proof:
int(A ∩ B) = ∅ ⇒ a = b,
∃ ε > 0, ∀|δ| < ε, b + δ ∈ B
⇒ b + δ − a + a0 = a0 + δ ∈ E(A, B)
⇒ a0 ∈ int(E(A, B)).
Proof:
Assuming
The boundaries of entrance solid angle determine the entrance solid angles completely.
This section will be the theoretical basis of the computation of the entrance solid angle
boundary. The boundaries of 2D entrance solid angles are the subsets of the contact
vectors.
In the previous codes of 2D discontinuous deformation analysis (DDA, Shi 1988)
and 2D numerical manifold method (NMM, Shi 1991), the algorithms for solving the
contacts are also based on computing these boundaries and consistent with the results
264 Discontinuous deformation analysis in rock mechanics practice
of this general entrance solid angle boundary. However, the new theory can simplify
these contact algorithms of 2D DDA and NMM. In complicated cases the new theory
can find the contact covers precisely.
A = e 1 e 2 + e,
B = h1 h2 + h.
v 0 ∈ int( e 1 e 2 ∩ h1 h2 ).
∂E(A, B) = ∅.
Shi’s new contact theory 265
Proof:
∃ε > 0,
C = {v 0 · x/ |x| ≤ ε},
C ⊂ ( e 1 e 2 ∩ h1 h2 ).
C − v 0
∂(C − v 0 ) = ∅.
E(A, B) = B − A + a0
= h 1 h 2 − e 1 e 2 + h − e + a0
⊃ C − v 0 + h − e + a0 .
∂(C − v 0 + h − e + a0 ) = ∅
⇒ ∂E (A, B) = ∅.
Proof:
If a ∈ int(A), b ∈ B,
E(a, b) = b − a + a0 ∈ int(b − A + a0 )
= int(E(A, b)) ⊂ int(E(A, B)).
If a ∈ A, b ∈ int(B),
E(a, b) = b − a + a0 ∈ int(B − a + a0 )
= int(E(a, B)) ⊂ int(E(A, B)).
If a ∈ int(A) or b ∈ int(B),
E(a, b) = b − a + a0
is an inner point of E(A, B).
Therefore
∂E(A, B) ⊂ E(∂A, ∂B).
A = e 1 e 2 + e,
B = h1 h2 + h.
Proof:
From Section 5.3,
∂E(A, B) ⊂ E(∂A, ∂B) = E(A(1), B(1)).
If e i and hj are not parallel and
a ∈ e i + e, b ∈ hj + h, i, j = 1, 2,
a = e, b = h
Shi’s new contact theory 267
then
E(a, b) = b − a + a0
∈ int(hj + h − e i − e + a0 )
= int(E( e i + e, hj + h)) ⊂ int(E(A, B)).
E( e i + e, hj + h) ∩ ∂E(A, B) ⊂
(E(e, hj + h) ∪E ( e i + e, h)).
If e i hj , from the theorem of Section 8.4.2
E( e i + e, hj + h) = (E(e, hj + h) ∪ E( e i + e, h)).
Therefore
n11 = (0, 0, 1) × e 1 ,
n21 = (0, 0, 1) × h1 ,
n22 = (0, 0, −1) × h2 .
Proof:
If
int( e 1 e 2 ∩ ↑ n2j ) = ∅,
∃ e 0 ∈ e 1 e 2 , e 0 · n2j > 0,
E(e, b) ∈
int(hj − e 0 + h − e + a0 ∪ h1 h2 + h − e + a0 )
= int(E( e 0 + e, hj + h) ∪ E(e, h1 h2 + h))
⇒ int h1 h2 ∩ ↑ n1i = ∅.
Proof:
If
int(h1 h2 ∩ ↑ n1i ) = ∅,
∃ h0 ∈ h1 h2 , h0 · n1i > 0,
E(a, h) ∈
int(h0 − ei + h − e + a0 ∪ − e 1 e 2 + h − e + a0 )
= int(E( e i + e, h0 + h) ∪ E( e 1 e 2 + e, h))
e 1 e 2 ∩ ↑ n2j ⊂ ∂ e 1 e 2 ∩ ∂ ↑ n2j ⇔
E e 1 e 2 , ↑ n2j =↑ n2j + a0 ⇔
The equations of the first line are defined as contact condition and E(e, hj + h)
is defined as one contact vector (Figure 8.44).
int(h1 h2 ∩ ↑ n1i ) = ∅ ⇔
h1 h2 ∩ ↑ n1i ⊂ (∂h1 h2 ∩ ∂ ↑ n1i ) ⇔
E(h1 h2 , ↑ n1i ) = − ↑ n1i + a0 ⇔
h1 × h2 (0, 0, +1) , h1 · n1i ≤ 0, h2 · n1i ≤ 0.
The equations of the first line are defined as contact condition, and E( e i + e, h)
e i hj ,
∃ a ∈ int(e + e i ), b ∈ int(h + hj ),
n1i n2j ⇒
∃ e 0 ∈ e 1 e 2, e 0 · n2j > 0,
E(a, b) ∈ int(E( e i + e, hj + h)) ⊂
270 Discontinuous deformation analysis in rock mechanics practice
int(E( e 0 + e i + e, hj + h) ∪ E( e i + e, h1 h2 + h)) ⊂
Therefore
n1i −n2j .
E( e i + e, hj + h)
= E(e, hj + h) ∪ E(hj + e, h).
Shi’s new contact theory 271
If
− e i hj ,
E( e i + e, hj + h)
= E(e, hj + h) = E( e i + e, h).
Proof:
E(e, hj + h) = hj + h − e + a0 ,
E( e i + e, h) = − e i + h − e + a0 .
If
e i hj ,
E( e i + e, hj + h)
= h j + h − e i − e + a0
= h j + h − h j − e + a0
↔
= h j + h − e + a0
= E(e, hj + h) ∪ E(hj + e, h).
Here the contact vectors E(e, hj + h) and E(hj + e, h) are two connected covers
of E( e i + e, hj + h).
If
hj − e i ,
E( e i + e, hj + h)
= h j + h − e i − e + a0
= h j + h − e + a0
= E(e, hj + h) = E( e i + e, h).
a0 ∈ E(e, hj + h) ⇔ e ∈ hj + h,
a0 ∈ E( e i + e, h) ⇔ h ∈ e i + e.
272 Discontinuous deformation analysis in rock mechanics practice
∂E(A, B) ⊂ E(A(0),B(1))
∪ E(A(1), B(0))
7 7
= E(e, hj + h) ∪ E( e i + e, h)
j=1,2 i=1,2
E(e, hj + h),
5
C(0, 1) = E(e, hj + h)
contact vector
e 1 × e 2 (0, 0, +1) ,
e 1 · n2j ≤ 0, e 2 · n2j ≤ 0.
Shi’s new contact theory 273
E( e i + e, h),
7
C(1, 0) = E( e i + e, h)
contact vector
h1 × h2 (0, 0, +1) ,
h1 · n1i ≤ 0, h2 · n1i ≤ 0.
Based on Section 8.5.1, if the step movement ρ is small enough, the entrances will be
simplified as entrances of solid angles. The contact vectors or contact vector covers on
the contact surface between 2D solid angles are computed in the following sections.
A = e + e 1 e 2,
B = h + h1 h2 ,
e 1 × e 2 (0, 0, +1),
h1 × h2 (0, 0, −1).
The inner normal vectors of e 1 , e 2 , h1 , h2 are
n11 = e 1 × e 2 × e 1 ,
n12 = e 2 × e 1 × e 2 ,
n21 = h2 × h1 × h1 ,
n22 = h1 × h2 × h2 .
274 Discontinuous deformation analysis in rock mechanics practice
Figure 8.47 Contact vectors of a 2D convex solid angle and a 2D concave solid angle.
e 1 × e 2 (0, 0, +1),
h1 × h2 (0, 0, −1),
then
int( e 1 e 2 ∩ h1 h2 ) = ∅ ⇔
E(A, B) = B − e + a0 ,
∂E(A, B) = ∂B − e + a0 .
Proof:
int( e 1 e 2 ∩ h1 h2 ) = ∅ ⇔
int( e 1 e 2 ∩ ↑ n21 ) = ∅,
int( e 1 e 2 ∩ ↑ n22 ) = ∅ ⇔
Shi’s new contact theory 275
E( e 1 e 2 + e, ↑ n22 + h)
= (E( e 1 e 2 , ↑ n21 ) + h − e) ∪
(E( e 1 e 2 , ↑ n22 ) + h − e)
=↑ n21 ∪ ↑ n22 + h − e + a0
= B − e + a0 ,
∂E(A, B) = ∂B − e + a0 .
E(e, h1 + h) = h − e + a0 + h1 ,
E(e, h2 + h) = h − e + a0 + h2 .
A = e + e 1 e 2,
B = h + h1 h2 ,
276 Discontinuous deformation analysis in rock mechanics practice
e 1 × e 2 (0, 0, 1),
h1 × h2 (0, 0, 1).
n11 = e 1 × e 2 × e 1 ,
n12 = e 2 × e 1 × e 2 ,
n21 = h1 × h2 × h1 ,
n22 = h2 × h1 × h2 .
A = {(x − e) · n11 ≥ 0}
B = {(x − h) · n21 ≥ 0}
Figure 8.50 Angle method for finding contact vectors of two 2D solid angles.
Also
A = (↑ n11 + e)
∩(↑ n12 + e).
B = (↑ n21 + h)
∩(↑ n22 + h).
Based on Section 8.5.2, if e1 e2 and h1 h2 have a common inner vector v 0 , E(A, B)
If
e 1 × h1 (0, 0, 1),
E(e, h1 + h)
is a boundary vector of the entrance angle (Figures 8.50 & 8.51).
If
e 2 × h2 (0, 0, 1) ,
E( e 2 + e, h)
Figure 8.51 Rotation method for finding contact vectors of two 2D solid angles.
Proof:
int( e 1 e 2 ∩ h1 h2 ) = ∅ ⇔
e 1 , e 2 , h1 , h2 rotates from ox to oy.
Shi’s new contact theory 279
If
e 1 × h1 (0, 0, 1) ⇒
e 1 · n21 ≤ 0, e 2 · n21 ≤ 0 ⇒
↔
∂E( e 1 e 2 + e, (↑ n21 ) + h) = h − e + a0 + h1 ⇒
E(e, h1 + h) ⊂ E(A, B) ⊂
E( e 1 e 2 + e, (↑ n21 ) + h),
E(e, h1 + h) ⊂ ∂E( e 1 e 2 + e, (↑ n21 ) + h) ⇒
E(e, h1 + h) ⊂ ∂E(A, B)
If
e 1 × h1 (0, 0, −1) ⇒
h1 · n11 ≤ 0, h2 · n11 ≤ 0 ⇒
E((↑ n11 ), h1 h2 ) =↑ n11 + a0 ⇒
E( e 1 + e, h) ⊂ E(A, B) ⊂
E((↑ n11 ) + e, h1 h2 + h),
E( e 1 + e, h) ⊂ ∂E((↑ n11 ) + e, h1 h2 + h) ⇒
Figures 8.50 & 8.51 illustrate two methods for finding the contact vectors, which
obey the upper theorem. The previous 2D DDA and NMM adopted the angle method.
If
e 2 × h2 (0, 0, 1) ⇒
h1 · n12 ≤ 0, h2 · n12 ≤ 0 ⇒
E((↑ n12 ), h1 h2 ) =↑ n12 + a0 ⇒
E( e 2 + e, h) ⊂ E(A, B) ⊂
E((↑ n12 ) + e, h1 h2 + h),
E( e 2 + e, h) ⊂ ∂E((↑ n12 ) + e, h1 h2 + h) ⇒
If
e 2 × h2 (0, 0, −1) ⇒
e 1 · n22 ≤ 0, e 2 · n22 ≤ 0 ⇒
E(e, h1 + h) ⊂ E(A, B) ⊂
E( e 1 e 2 + e, (↑ n22 ) + h),
E(e, h1 + h) ⊂ ∂E( e 1 e 2 + e, (↑ n21 ) + h) ⇒
E(e, h1 + h) ⊂ ∂E(A, B)
These two methods can be applied to the case of the contact vectors between two
vertically opposite convex angles (Figures 8.52 & 8.53).
A1 = {|x − a0 | ≤ r},
A = E(A1 , A2 ),
A2 = (↑ n11 + e0 )
∩ (↑ n12 + e0 ) ,
e0 = e + r(n11 + n12 )/(1 + n11 · n12 ).
Proof:
Define an inner angle A2 by:
(e0 − e) · n11 − r = 0,
(e0 − e) · n12 − r = 0 ⇒
A2 = (↑ n11 + e0 ) ∩ (↑ n12 + e0 ).
A = E(A1 , A2 ).
282 Discontinuous deformation analysis in rock mechanics practice
Theorem of entrance of round corner solid angle and concave solid angle
Round corner solid angle A is the same. Concave solid angle B is the union of two half
planes,
If
Proof:
A = E(A1 , A2 ).
Figure 8.55 Round corner angle is the entrance angle of a 2D solid angle and a disk.
∪ (↑ n22 + h − rn22 ).
The angle top h0 of E(A, B) (Figure 8.56) satisfies the following equations
Figure 8.56 Contact vectors of a round corner convex solid angle and a concave solid angle.
(h0 − h) · n21 + r = 0,
(h0 − h) · n22 + r = 0
⇒ ∃ t, h0 − h = t(n21 + n22 ),
E(A, B) = (↑ n21 + h0 )
∪ (↑ n22 + h0 ).
A0 = {(x − e) · n11 ≥ 0}
∩ {(x − e) · n12 ≥ 0}
B0 = {(x − h) · n21 ≥ 0}
∩ {(x − h) · n22 ≥ 0}
A1 = {|x − a0 | ≤ r1 },
B1 = {|x − b0 | ≤ r2 }.
Let
Define two inner angles A2 and B2 by the following equations: (Figure 8.57)
A = E(A1 , A2 ) = A2 − A1 + a0 ,
B = E(B1 , B2 ) = B2 − B1 + b0 ,
E(A1 , B1 ) = {|x − b0 | ≤ r1 + r2 }.
Figure 8.58 The entrance angle of two 2D round corner convex solid angles is also a round corner
convex solid angle.
Figure 8.59 Another example of round corner convex entrance solid angle.
Proof:
Let e0 be the reference point of A.
E(A, B) is the entrance block of the disk E(A1 , B1 ) and the solid angle E(A2 , B2 ), i.e.
also a round corner convex solid angle (Figures 8.58 & 8.59).
An entrance block is determined by its boundary. The following theorems are for
understanding and finding the boundaries of 2D entrance blocks. The boundaries of
2D entrance blocks are the subsets of the contact edges.
Proof:
Assuming
a ∈ A, b ∈ B.
If a ∈ int(A)
E(a, b) = b − a + a0 ∈ int(b − A + a0 )
= int(E(A, b)) ⊂ int(E(A, B)).
E(a, b) = b − a + a0 ∈ int(B − a + a0 )
= int(E(a, B)) ⊂ int(E(A, B)).
E(a, b) = b − a + a0 ∈ int(E(A, B)).
Therefore
A = a1 a2 · · · ap−1 ap , ap+1 = a1 ,
B = b1 b2 · · · bq−1 bq , bq+1 = b1 .
288 Discontinuous deformation analysis in rock mechanics practice
Figure 8.60 The entrance point of a vertex and an inner point is still an inner point of entrance block.
Proof:
From Section 8.6.1,
a ∈ int(ai ai+1 ), i = 1, . . . , p,
b ∈ int(bj bj+1 ), j = 1, . . . , q,
E(a, b) = b − a + a0
∈ int(bj bj+1 − ai ai+1 + a0 )
= int(E(ai ai+1 , bj bj+1 )) ⊂ int(E(A, B)).
E(ai ai+1 , bj bj+1 ) ∩ ∂E(A, B) ⊂
(E(∂ai ai+1 , bj bj+1 ) ∪ E(ai ai+1 , ∂bj bj+1 )).
Figure 8.61 The entrance point of inner points of edges is an inner point of the entrance block.
Therefore
e1 = ai+1 , e2 = ai−1 .
e 1 = (ee1 ), e 2 = (ee2 ),
ai = e 1 e 2 .
Proof:
If
int(ai ∩ ↑ m2j ) = ∅.
∃e0 , ee0 ⊂ A, e 0 = (ee0 ), e 0 · m2j > 0,
E(e, b) = b − e + a0 ∈
int((B − e + a0 ) ∪ (bj bj+1 − ee0 + a0 ))
= int(E(e, B) ∪ E(ee0 , bj bj+1 ))
⊂ int(E(A, B) ∪ E(A, B)) = int(E(A, B))
⇒ E(e, b) ∈
/ ∂E(A, B).
h1 = bj+1 , h2 = bj−1 .
Denote
h1 = (hh1 ), h2 = (hh2 ),
bj = h1 h2 .
Proof:
Referring to Figure 8.63, if
int(bj ∩ ↑ m1i ) = ∅,
∃ h0 , hh0 ⊂ B, h0 = (hh0 ), h0 · m1i > 0,
E(a, h) = h − a + a0 ∈
int((h − A + a0 ) ∪ (hh0 − ai ai+1 + a0 ))
= int(E(A, h) ∪ E(ai ai+1 , hh0 ))
⊂ int(E(A, B) ∪ E(A, B)) = int(E(A, B)),
⇒ E(a, h) ∈
/ ∂E(A, B).
Shi’s new contact theory 291
The equations are defined as contact condition, and E(ai , bj bj+1 ) is defined as
contact edge.
int(bj ∩ ↑ m1i ) =
int(h1 h2 ∩ ↑ m1i ) = ∅ ⇔
E(h1 h2 , ↑ m1i ) = ↑ m1i + a0 ⇔
h1 × h2 (0, 0, +1), h1 · m1i ≤ 0, h2 · m1i ≤ 0.
The equations are defined as contact condition, and E(ai ai+1 , bj ) is defined as
contact edge.
point pairs. These vertex-to-edge contact point pairs are called “contact points’’ corre-
sponding to overlapped contact edges. These contact pairs may have two states: open
or closed, which will be determined by an open-close iteration. After small rotation or
deformation, the contacts of blocks A and B happen only along these contact points.
All closed contact points together define the movements, rotations and deforma-
tions of all blocks just as in real cases.
Assuming A and B are the same 2D blocks as in Section 8.7.2, with the only
difference that
E(ai ai+1 , bj ),
E(ai ai+1 , bj+1 ),
E(ai , bj bj+1 ),
E(ai+1 , bj bj+1 )
are overlapped covers of this line segment (Figure 8.64). Each cover refers to a possible
vertex-to-edge contact. Therefore the entrance of parallel edges can be transferred to
the entrances of a vertex and an edge.
Proof:
m1i m2j ⇒
∃δ,
C = {|x| ≤ δ},
(C ∩ ↑ m1i ) + a ⊂ A,
(C ∩ ↑ m2j ) + b ⊂ B ⇒
E(a, b) ∈
int((C ∩ ↑ m2j ) − (C ∩ ↑ m1i ) + b − a + a0 ) =
int(E((C ∩ ↑ m1i ) + a, (C ∩ ↑ m2j ) + b)) ⊂
int(E(A, B)).
⇒ E(a, b) ∈ int(E(A, B)), E(a, b) ∈
/ ∂E(A, B).
, bj bj+1 ),
E(ai7
C(0, 1) = E(ai , bj bj+1 )
contact edge
e 1 × e 2 (0, 0, +1),
e 1 · m2j ≤ 0, e 2 · m2j ≤ 0.
E(ai ai+1 , bj ),
7
C(1, 0) = E(ai ai+1 , bj )
contact edge
The contact surface of 2D blocks is a cover system formed by contact edges. Each
contact edge corresponds to a contact point, resulting in a possible vertex-to-edge
contact.
In the process of computing, time steps are used. In each time step, the displacement
should be smaller than the set maximum displacement ρ. At the beginning of each time
step, the contact edges are computed. From the relative position of reference point a0
and the contact edges, the closed contact edges are found. The closed contact points
control the movements, rotations and deformations throughout this time step.
A = a1 a2 · · · ap−1 ap , ap+1 = a1 ,
B = b1 b2 · · · bq−1 bq , bq+1 = b1 ,
Shi’s new contact theory 295
which rotates in the right hand rule. As convex, A and B can be represented by
simultaneous inequality equations. For block A, denote:
m1i = (0, 0, 1) × e 1 ⇒
Normal vectors m1j and m2j point the inside of block.
Proof:
Based on Section 8.7.3,
int(ai ∩ ↑ m2j ) = ∅ ⇔
E(ai , ↑ m2j ) = ↑ m2j + a0 .
ai + ai ⊃ A, ↑ m2j + bj ⊃ B, ⇒
E(ai + ai , ↑ m2j + bj ) ⊃ E(A, B).
E(ai , bj bj+1 ) is not only a contact but also a boundary edge of E(A, B).
Proof:
Based on Section 8.7.3,
int(bj ∩ ↑ m1i ) = ∅ ⇔
E(bj , ↑ m1i ) =↑ m1i + a0 .
↑ m1i + ai ⊃ A, bj + bj ⊃ B ⇒
E(↑ m1i + ai , bj + bj ) ⊃ E(A, B).
E(ai ai+1 , bj ) is both a contact edge and a boundary edge of E(A, B) (Figures 8.65
to 8.68).
Figure 8.66 Entrance block of two equal lateral blocks which are in different directions.
Proof:
From Section 8.7.5,
Therefore,
A = a1 a2 · · · ap−1 ap , ap+1 = a1 ,
B = b1 b2 · · · bq−1 bq , bq+1 = b1 ,
m1i = (0, 0, 1) × e 1 , i = 1, . . . , p.
h1 = (bj bj+1 ), h2 = (bj bj−1 ),
m2j = (0, 0, 1) × h1 , j = 1, . . . , q.
Normal vectors m1i and m2j point the inside of block. Based on the theorems of
Sections 8.7.3 to 8.7.5, the following theorems for general 2D blocks exist.
All contact edges are connected or intersected with each other to form the boundaries
of the entrance block ∂E(A, B) (Figure 8.69).
Figure 8.70 Deformation and rockbolt reinforcement modeling with 2D-DDA for a 30 m span
underground powerhouse.
DDA works on deformable block systems. Each block has linear displacements, a
constant stress and a constant strain. In DDA, multi-time steps are used. Both static and
dynamic processes are performed in the dynamic computation. Static computation is
the stabilized dynamic computation. Therefore, discontinuous and large displacement
computation can be realized for both static and dynamic processes in DDA.
For each time step, there are usually several open-close iterations in DDA. Before
going to the next time step, open or close modes are adjusted within iterations until
Shi’s new contact theory 301
Figure 8.71 2D DDA block mesh of parallel tunnels and collapse under shock wave.
every contact position has the same contact mode before and after the equations
are solved. Here, for each open-close iteration in each time step, DDA solves the
global equilibrium equations. As the principal law of stability analysis, a friction
law is ensured in DDA computation. The friction law is mathematically translated
to inequality equations.
Every single block of 2D DDA can be a generally shaped convex or concave poly-
gon. Based on simplex integration (Shi 1996), the stiffness matrices, the inertia matrices
and all other matrices of DDA are analytical solutions.
DDA is a discontinuous version of the FEM and a visible version of the limit
equilibrium method. DDA also serves as an implicit version of the DEM method.
302 Discontinuous deformation analysis in rock mechanics practice
Figure 8.72 2D DDA computed results of a thin arch dam under five times normal water pressure.
Figure 8.73 Stability analysis of a RCC gravity dam as computed with 2D DDA where the friction angle
on the joints is 17◦ .
DDA has all advantages of dynamic relaxation yet the convergence is strict and the
result is accurate.
More importantly, DDA is a very well-tested method by analytical solutions,
physical model tests and large engineering projects.
Shi’s new contact theory 303
Figure 8.74 2D DDA computation of the critical section of a dam abutment slope with the assumed
sliding line and using a friction angle of 18◦ for the joints.
Figure 8.76 Rock and shotcrete fall in an underground powerhouse without horizontal initial stress
as computed with 2D NMM.
Figure 8.77 2D NMM simulation of the deformation of an underground powerhouse where the side-
walls are parallel to the bedding planes direction with one tenth of the real elastic modulus
used as input.
(Shi, 1991, 1996). In this method, the movements and deformations of continuous
and discontinuous structures or materials are computed in a unified form. The meshes
of the numerical manifold method consist of finite covers. The finite covers overlap
each other and cover the entire material volume. On each cover, the manifold method
Shi’s new contact theory 305
defines an independent displacement function, called cover function. The global dis-
placement functions are the weighted averages of local independent cover functions
on the common part of several covers. For more details, see Shi (1991).
Using the finite cover systems, continuous, jointed or blocky materials can be
computed in a mathematically consistent manner. For manifold computation, the
mathematical covers and physical mesh are independent. Therefore, the mathemat-
ical covers are free to define and are easily changed. The mathematical covers can be
easily moved, split, removed, or added. Moving the covers, the large deformations
and moving boundaries can be computed. Because the joints can divide a cover into
two or more independent covers, which have independent displacement functions, the
general discontinuity can be modeled.
Both the finite element method (FEM) for continua and the discontinuous
deformation analysis (DDA) for block systems are special cases of NMM.
In the current development stage of NMM, by using the finite cover approach,
more flexible deformations and movements of joint and block systems can be
computed.
As for the applications of the new theory of contact, several engineering cases
of 2D NMM are shown in Figures 8.76 to 8.80. The current version of 2D NMM
adopts the same contact algorithm used in 2D DDA, which is consistent with the new
theory of contact while the following two conditions are fulfilled: the time step is
306 Discontinuous deformation analysis in rock mechanics practice
Figure 8.80 2D NMM results of a multi-lane highway tunnel in jointed rocks with and without
concrete lining.
small enough and the stiffness of the contact spring is large enough. Figure 8.76 shows
NMM computation results of rock and shotcrete fall of an underground powerhouse.
Figure 8.77 shows the computed deformation of an underground powerhouse by 2D
NMM. Figure 8.78 shows circular sliding computed by 2D NMM. Figure 8.79 shows
the deformation of a coalmine computed by 2D NMM. Figure 8.80 is the 2D NMM
computation results of a highway tunnel with and without concrete lining.
Contacts between two 3D solid angles are much more complex and difficult than
these between two 2D solid angles or blocks. However, under the entrance theory,
Shi’s new contact theory 307
the 3D entrance solid angles can be computed in a similar way as that in the 2D
cases. Similarly, a 3D entrance solid angle can be defined and represented by its
boundaries. The following theorems are used to find the boundaries of a 3D entrance
solid angle.
Assume the maximum step movement is ρ. The contact distance is ε = 2ρ. Denote
D as a ball
D = {|x − a0 | ≤ ε}.
Therefore, for a given time step, only a small part of entrance block E(A, B)
E(A, B) ∩ D = ∅
is needed to be considered.
A = e 1 e 2 · · · e u−1 e u + e,
B = h1 h2 · · · hv−1 hv + h.
v 0 ∈ int( e 1 e 2 · · · e u−1 e u ∩ h1 h2 · · · hv−1 hv ),
∂E(A, B) = ∅.
308 Discontinuous deformation analysis in rock mechanics practice
Proof:
∃ ε > 0,
C = {v 0 · x/|x| ≤ ε},
C ⊂ ( e 1 e 2 · · · e u−1 e u ∩ h1 h2 · · · hv−1 hv ).
C − v 0
∂(C − v 0 ) = ∅
E(A, B) = B − A + a0
= h1 h2 · · · hv−1 hv − e 1 e 2 · · · e u−1 e u + h − e + a0
⊃ C − v 0 + h − e + a0 .
∂(C − v 0 + h − e + a0 ) = ∅
⇒ ∂E(A, B) = ∅.
∂E(A, B) = ∅
e 1 e 2 · · · e u−1 e u ∩ h1 h2 · · · hv−1 hv ⊂
(∂ e 1 e 2 · · · e u−1 e u ∩ ∂h1 h2 · · · hv−1 hv ).
Shi’s new contact theory 309
Proof:
If a ∈ int(A), b ∈ B,
E(a, b) = b − a + a0 ∈ int(b − A + a0 )
= int(E(A, b)) ⊂ int(E(A, B)).
If a ∈ A, b ∈ int(B),
E(a, b) = b − a + a0 ∈ int(B − a + a0 )
= int(E(a, B)) ⊂ int(E(A, B)).
If a ∈ int(A) or b ∈ int(B),
E(a, b) = b − a + a0
Proof:
From Section 8.9.3,
hs hs+1 + h, s = 1, . . . , v.
If solid angle e r e r+1 and hs hs+1 are not parallel, for
a ∈ int( e r e r+1 + e), b ∈ int(hs hs+1 + h),
E(a, b) = b − a + a0
∈ int(hs hs+1 + h − v 0 − e + a0 )
= int(E(v 0 + e, hs hs+1 + h))
⊂ int(E( e r e r+1 + e, hs hs+1 + h))
⊂ int(E(A, B)).
E(a, b) ∈ int(E(A, B)).
E( e r e r+1 + e, hs hs+1 + h) ∩ ∂E(A, B) ⊂
(E(∂ e r e r+1 + e, hs hs+1 + h) ∪ E( e r e r+1 + e, ∂hs hs+1 + h)).
If solid angle e r e r+1 and hs hs+1 are parallel (Figure 8.82), from theorem in
Section 8.4.2
E( e r e r+1 + e, hs hs+1 + h) =
E(∂ e r e r+1 + e, hs hs+1 + h)
∪E( e r e r+1 + e, ∂hs hs+1 + h).
Therefore
Proof:
e r + e, r = 1, . . . , u,
hs hs+1 + h, s = 1, . . . , v,
Shi’s new contact theory 311
Figure 8.83 Entrance point of a face angle inner point and an edge vector inner point.
are boundary edge vector and angle of 3D angles A and B, respectively. For any points
a ∈ int( e r + e), b ∈ int(hs hs+1 + h),
If er and hs hs+1 are not parallel (Figure 8.83),
E(a, b) = b − a + a0
∈ int(hs hs+1 + h − e r − e + a0 )
= int(E( e r + e, hs hs+1 + h)
⊂ int(E(A, B)).
E(a, b) ∈ int(E(A, B)).
E( e r + e, hs hs+1 + h) ∩ ∂E(A, B) ⊂
(E(∂ e r + e, hs hs+1 + h) ∪ E( e r + e, ∂hs hs+1 + h)).
312 Discontinuous deformation analysis in rock mechanics practice
If solid angle e r and hs hs+1 are parallel, from theorem of Section 8.4.2
E( e r + e, hs hs+1 + h) =
E(∂ e r + e, hs hs+1 + h)
∪ E( e r + e, ∂hs hs+1 + h).
Therefore
Proof:
e r e r+1 + e, r = 1, . . . , u,
hs + h, s = 1, . . . , v,
are boundary angle and vector of 3D angles A and B respectively. For any points
a ∈ int( e r e r+1 + e), b ∈ int(hs + h),
If er e r+1 and hs are not parallel (Figure 8.83),
E(a, b) = b − a + a0
∈ int(hs + h − e r e r+1 − e + a0 )
= int(E( e r e r+1 + e, hs + h)
⊂ int(E(A, B)).
E(a, b) ∈ int(E(A, B)).
E( e r e r+1 + e, hs + h) ∩ ∂E(A, B) ⊂
(E(∂ e r e r+1 + e, hs + h) ∪ E( e r e r+1 + e, ∂hs + h)).
If solid angles e r e r+1 and hs are parallel, from theorem of section 8.3.2
E( e r e r+1 + e, hs + h) =
E(∂ e r e r+1 + e, hs + h)
∪ E( e r e r+1 + e, ∂hs + h).
Shi’s new contact theory 313
Therefore
∂E(A, B) ⊂ E(A(0), B(2)) ∪ E(A(2), B(0)) ∪
E(A(1), B(1)).
A = e 1 e 2 · · · e u−1 e u + e,
B = h1 h2 · · · hv−1 hv + h.
The inner normal vectors of e r e r+1 and hs hs+1 are n1r and n2s respectively. The
following
e 1 e 2 · · · e u−1 e u + e,
hs hs+1 + h, s = 1, . . . , v.
are 3D angle A and a boundary angle of B respectively.
Proof
If
int( e 1 e 2 · · · e u−1 e u ∩ ↑ n2s ) = ∅,
E(e, b) = b − e + a0 ∈ int(E(e, hs hs+1 + h)) ⊂
int((hs hs+1 − v 0 + h − e + a0 ) ∪ (B − e + a0 ))
= int(E(v 0 + e, hs hs+1 + h) ∪ E(e, B))
⊂ int(E(A, B)).
⇒ E(e, b) ∈ int(E(A, B)), E(e, b) ∈
/ ∂E(A, B).
Similarly, the following
h1 h2 · · · hv−1 hv + h,
e r e r+1 + e, r = 1, . . . , u.
are 3D angle B and a boundary angle of A respectively. n1r is the inner normal vector
of e r e r+1 .
314 Discontinuous deformation analysis in rock mechanics practice
⇒ int(h1 h2 · · · hv−1 hv ∩ ↑ n1r ) = ∅.
Proof:
If
int(h1 h2 · · · hv−1 hv ∩ ↑ n1r ) = ∅,
∃v 0 ∈ h1 h2 · · · hv−1 hv , v 0 · n1r > 0.
int((h − A + a0 ) ∪ ( v 0 + h − e − e r e r+1 + a0 ))
⊂ int(E(A, B)).
⇒ E(a, h) ∈ int(E(A, B)), E(a, h) ∈ / ∂E(A, B).
e 1 e 2 · · · e u−1 e u ∩ ↑ n2s
e r n2s ≤ 0, r = 1, . . . , u,
The equations are defined as contact condition, and E(e, hj + h) is defined as
contact vector.
int(h1 h2 · · · hv−1 hv ∩ ↑ n1r ) = ∅ ⇔
h1 h2 · · · hv−1 hv ∩ ↑ n1r
⊂ (∂h1 h2 · · · hv−1 hv ∩ ∂ ↑ n1r ) ⇔
E(h1 h2 · · · hv−1 hv , ↑ n1r ) =↑ n1r + a0 ⇔
hs · n1r ≤ 0, s = 1, . . . , v,
n1r ∩ h1 h2 · · · hv−1 hv = (0, 0, 0).
Figure 8.84 Contact solid angle of a 3D solid angle and a face solid angle.
Figure 8.85 Contact condition of a 3D solid angle and a face solid angle.
A = e 1 e 2 · · · e u−1 e u + e,
B = h1 h2 · · · hv−1 hv + h.
316 Discontinuous deformation analysis in rock mechanics practice
The inner normal vectors of e r−1 e r and e r e r+1 are n11 and n12 respectively. The
inner normal vectors of hs−1 hs and hs hs+1 are n21 and n22 respectively.
Assuming edges of e r and hs are convex,
e r n12 × n11 ,
hs n22 × n21 .
E(a, b) ∈ ∂E(A, B) ⇒
int((↑ n11 ∩ ↑ n12 ) ∩ (↑ n21 ∩ ↑ n22 )) = ∅.
Proof:
Denote
hs1 = n21 × n22 × n21 ,
hs2 = n22 × n21 × n22 .
If
∃i, j, 1 ≤ i ≤ 2, 1 ≤ j ≤ 2,
e ri · n0 > 0,
hsj · n0 > 0.
If i = 1, P1 = e r−1 e r , if i = 2, P1 = e r e r+1 .
If j = 1, P2 = hs−1 hs , if j = 2, P2 = hs hs+1 .
Assume e r and hs are not parallel,
E( e r , hs ) ⊂ E(P1 , hs ),
E( e r , hs ) ⊂ E( e r , P2 ).
E(a, b) ∈ int(E( e r , hs ))
Shi’s new contact theory 317
⊂ int(E(P1 , hs ) ∪ E( e r , P2 ))
⊂ int(E(P1 , P2 ) ∪ E(P1 , P2 ))
= int(E(P1 , P2 ))
⊂ int(E(A, B)) ⇒
E(a, b) ∈
/ ∂E(A, B).
Denote
e r1 = n11 × n12 × n11 ,
hs1 = n21 × n22 × n21 ,
hs2 = n22 × n21 × n22 ,
then
↔
↑ n11 ∩ ↑ n12 = e r1 e r2 + e r ,
↔
↑ n21 ∩ ↑ n22 = hs1 hs2 + hs ,
↔ ↔
∂(↑ n11 ∩ ↑ n12 ) = ( e r1 + e r ) ∪ ( e r2 + e r ),
↔ ↔
∂(↑ n21 ∩ ↑ n22 ) = (hs1 + hs ) ∪ (hs2 + hs ).
↔
↔
( e r1 e r2 + e r ) ∩ (hs1 hs2 + hs ) ⊂
↔
↔
∂( e r1 e r2 + e r ) ∩ ∂(hs1 hs2 + hs ) ⇔
∃n0 , n0 e r × hs ,
↔
↔
E( e r1 e r2 + e r , hs1 hs2 + hs ) =↑ n0 + a0 ⇔
∃n0 , n0 e r × hs ,
e r1 · n0 ≤ 0, e r2 · n0 ≤ 0,
hs1 · n0 ≥ 0, hs2 · n0 ≥ 0.
The equations are defined as contact condition, and E( e r + e, hs + h) is defined as
e r + e, r = 1, . . . , u,
hs hs+1 + h, s = 1, . . . , v.
are parallel edge vector and boundary angle of 3D angles A and B respectively.
318 Discontinuous deformation analysis in rock mechanics practice
Figure 8.88 Contact of an edge vector with a parallel face solid angle int( e r ∩ h s h s+1 ) = ∅.
A0 = e r + e, B0 = hs hs+1 + h,
e r hs hs+1 .
E(e, hs hs+1 + h) ∪
E( e r + e, hs + h) ∪
E( e r + e, hs+1 + h).
Figures 8.88–8.91 illustrate the contacts of several cases where an edge vector is
parallel to the face of a solid angle.
320 Discontinuous deformation analysis in rock mechanics practice
Figure 8.89 Contact solid angles of an edge vector with a parallel face solid angle
int( e r ∩ h s h s+1 ) = ∅.
Assuming
A = e 1 e 2 · · · e u−1 e u + e,
B = h1 h2 · · · hv−1 hv + h,
e r hs hs+1 ,
The inner normal vectors of e r−1 e r and e r e r+1 are n11 and n12 , respectively
(Figure 8.92).
Shi’s new contact theory 321
Figure 8.90 Contact of an edge vector with a parallel face solid angle int( e r ∩ h s h s+1 ) = ∅.
E(a, b) ∈ ∂E(A, B) ⇒
int((↑ n11 ∩ ↑ n12 ) ∩ ↑ n21 ) = ∅.
Proof:
If
E(a, b) ∈ int(E( e r + e, hs hs+1 + h)) ⊂
int(E( e r + v 0 + e, hs hs+1 + h) ∪ E( e r + e, B)) ⊂
Figure 8.91 Contact solid angles of an edge vector with a parallel face solid angle
int( e r ∩ h s h s+1 ) = ∅.
Shi’s new contact theory 323
e r e r+1 + e, r = 1, . . . , u,
hs hs+1 + h, s = 1, . . . , v.
are
n11 , n21 .
Assuming
n11 n21 .
e r e r+1 hs hs+1 .
E( e r + e, hs + h)∪
E( e r + e, hs+1 + h)∪
E( e r+1 + e, hs + h)∪
E( e r+1 + e, hs+1 + h).
Proof:
If
n11 n21 ⇒
∃v 0 , e r e r+1 + v 0 ⊂ e 1 e 2 · · · e u−1 e u ,
v 0 · n2s > 0.
E(a, b) ∈ int(E( e r e r+1 + e, hs hs+1 + h)) ⊂
int(E( e r e r+1 + v 0 + e, hs hs+1 + h) ∪ E( e r e r+1 + e, B)) ⊂
Figures 9.93 to 9.96 illustrate the contacts of several cases where the faces of two
solid angles are parallel.
A = e 1 e 2 · · · e u−1 e u + e,
B = h1 h2 · · · hv−1 hv + h.
Shi’s new contact theory 325
Figure 8.93 Contact of parallel boundary solid angles int( e r e r+1 ∩ h s h s+1 ) = ∅.
Figure 8.94 Contact solid angles of parallel boundary solid angles int( e r e r+1 ∩ h s h s+1 ) = ∅.
Figure 8.95 Contact solid angles of parallel boundary solid angles int( e r e r+1 ∩ h s h s+1 ) = ∅.
E( e r e r+1 + e,
h),
C(2, 0) = ∪contact E( e r e r+1 + e,
angle h)
int(h1 h2 · · · hv−1 hv ∩ ↑ n1r ) = ∅
E( e r + e, hs + h),
C(1, 1) = ∪contact E( e r + e, hs + h)
angle
Based on Section 8.9.1, if the step movement ρ is small enough, the entrances will be
simplified as a series of solid angles. There are different entrance cases between two
3D solid angles. Only the first entrance is essential for discontinuous computations.
This case is
a0 ∈ ∂E(A, B).
The contact surfaces between two 3D solid angles are covered by contacts of 2D
solid angles and are computed in the following sections.
A is a 3D angle defined by
A = e 1 e 2 · · · e u−1 e u + e.
The inner normal vectors of e r e r+1 is n1r . From the theorem of entrance of solid
angle and half space of Section 8.8.6, the following theorem takes place.
e 1 e 2 · · · e u−1 e u ∩ ↑ n21
e r · n21 ≤ 0, r = 1, . . . , u,
(x − e) · n12 ≥ 0.
Shi’s new contact theory 329
B:
(x − h) · n21 ≥ 0,
(x − h) · n22 ≥ 0.
A = (↑ n11 ∩ ↑ n12 ) + e,
B = (↑ n21 ∩ ↑ n22 ) + h.
e 1 and h1 are edge vectors
e 1 = n12 × n11 ,
h1 = n22 × n21 .
Let r = 1, s = 1, the theorem of intersection of half space from Section 8.9.7 can
be simplified as
h11 = n21 × n22 × n21 ,
h12 = n22 × n21 × n22 ,
then
↔
↑ n11 ∩ ↑ n12 = e 11 e 12 + e 1 ,
↔
↑ n21 ∩ ↑ n22 = h11 h12 + h1 ,
↔ ↔
∂(↑ n11 ∩ ↑ n12 ) = ( e 11 + e 1 ) ∪ ( e 12 + e 1 ),
↔ ↔
∂(↑ n21 ∩ ↑ n22 ) = (h11 + h1 ) ∪ (h12 + h1 ).
Let r = 1, s = 1, the theorem of entrance of solid edge to solid edge from Section
8.9.7 can be simplified as follows.
↔
↔
E( e 11 e 12 + e 1 , h11 h12 + h1 ) =↑ n0 + a0 ⇔
∃n0 , n0 e 1 × h1 ,
e 11 · n0 ≤ 0, e 12 · n0 ≤ 0,
h11 · n0 ≥ 0, h12 · n0 ≥ 0.
It was assumed here, e 1 and h1 are not parallel.
330 Discontinuous deformation analysis in rock mechanics practice
Proof:
Using the same n0 , the sufficient condition can be proved. If the necessary condition
is not true,
∃n0 , i, j, 1 ≤ i ≤ 2, 1 ≤ j ≤ 2,
n0 e 1 × h1 , h1j n0 > 0 ⇒
e 1i n0 > 0,
↔
e 1 + e 1i ⊂↑ n11 ∩ ↑ n12 ,
↔
h1 + h1j ⊂↑ n21 ∩ ↑ n22 ,
↔
↔
e 1 + e 1i + h1 =↑ n0 ,
↔
↔
h1 + h1j + e 1 =↑ n0 ⇒
E(↑ n ∩ ↑ n12 , ↑ n21 ∩ ↑ n22 ) ⊃
↔
↔
E( e 1 + e 1i , h1 + h1j )
↔ ↔
↔ ↔
= E( e 1 + e 1i − h1 , h1 + h1j − e 1 )
↔ ↔
↔ ↔
= E( e 1 + e 1i + h1 , h1 + h1j + e 1 )
= E(⊥ n0 + e 1i , ⊥ n0 + h1j ) = E(↑ n0 , ↑ n0 )
=↑ n0 − ↑ n0 + a0 .
Here
↑ n0 − ↑ n0
∂E(A, B) = ∅.
Therefore
E(A, B) =↑ n0 + h − e + a0 ,
∂E(A, B) = ⊥ n0 + h − e + a0 .
B = B1 ∪ B2 ,
B1 =↑ n21 + h,
B2 =↑ n22 + h.
A = e 1 e 2 · · · e u−1 e u + e.
Shi’s new contact theory 331
Proof:
E(A, B) = E(A, B1 ∪ B2 ) = E(A, B1 ) ∪ E(A, B2 ).
int(A ∩ B) = ∅ ⇔
int(A ∩ B1 ) = ∅, int(A ∩ B2 ) = ∅.
int(A ∩ B1 ) = ∅, ⇔ E(A, B1 ) = B1 + a0 − e ⇔
E(A, B1 ) =↑ n21 + h + a0 − e.
int(A ∩ B2 ) = ∅ ⇔ E(A, B2 ) = B2 + a0 − e ⇔
E(A, B2 ) =↑ n22 + h + a0 − e.
Therefore
int(A ∩ B) = ∅ ⇔
E(A, B) =↑ n21 ∪ ↑ n22 + h + a0 − e = B + a0 − e.
∂E(A, B) = ∂B − e + a0 .
Denote
h1 = n22 × n21 ,
h21 = n22 × n21 × n21 ,
h22 = n21 × n22 × n22 .
↔
∂E(A, B) = (h1 + h21 + h − e + a0 ) ∪
↔
(h1 + h22 + h − e + a0 ).
332 Discontinuous deformation analysis in rock mechanics practice
h1 = n22 × n21 .
The inner normal vector of er er+1 is n1r (Figures 8.98 & 8.99).
Proof:
Denote
h11 = n21 × n22 × n21 ,
h12 = n22 × n21 × n22 .
Shi’s new contact theory 333
Figure 8.99 Projection along edge vector of the contact of a 3D solid angle and a convex vector edge.
↔
B = h1 + h11 h12 + h,
E(A, B) =
↔
E( e 1 e 2 · · · e u−1 e u + e, h1 + h11 h12 + h) =
↔
E( e 1 e 2 · · · e u−1 e u , h1 + h11 h12 ) + h − e =
↔ ↔
E( e 1 e 2 · · · e u−1 e u − h1 , h1 + h11 h12 ) + h − e =
↔ ↔
E( e 1 e 2 · · · e u−1 e u + h1 , h1 + h11 h12 ) + h − e
∃ei, ∃ej,
↔ ↔
e 1 e 2 · · · e u−1 e u + h1 = e i e j + h1 .
e 11 = h1 × e i × h1 ,
e 12 = h1 × e j × h1 .
It can be assumed
e 11 × e 12 = h1 .
↔ ↔
e 1 e 2 · · · e u−1 e u + h1 = e 11 e 12 + h1 .
E(A, B) =
↔ ↔
E( e 1 e 2 · · · e u−1 e u + h1 , h1 + h11 h12 ) + h − e
↔ ↔
= E( e 11 e 12 + h1 , h11 h12 + h1 ) + h − e
↔
= E( e 11 e 12 , h11 h12 ) + h1 + h − e.
334 Discontinuous deformation analysis in rock mechanics practice
int(A ∩ B) = ∅ ⇔ int( e 11 e 12 ∩ h11 h12 ) = ∅ ⇔
∃d 1 , d 2 , E( e 11 e 12 , h11 h12 )
= d 1 d 2 + a0 ⇔
↔
∃d 1 , d 2 , E(A, B) = d 1 d 2 + h1 + h − e + a0 .
↔
h 1 + d 1 + h − e + a 0 ,
↔
h 1 + d 2 + h − e + a 0 .
A = e 1 e 2 · · · e u−1 e u + e,
B = h1 h2 · · · hv−1 hv + h.
n1r = + e r+1 × e r , r = 1, . . . , u.
n2s = +hs+1 × hs , s = 1, . . . , v.
Based on Section 8.9.10, all contact angles are searched and examined as follows
(Figure 8.100).
E(e, hs hs+1 + h) ⊂ ∂E(A, B)
Shi’s new contact theory 335
Proof:
Based on Section 8.9.6,
E(e, hs hs+1 + h) = hs hs+1 + h − e + a0 ,
∂E(A, ↑ n2s + h) = ⊥ n2s + h − e + a0 ⇒
E(e, hs hs+1 + h) ⊂ ∂E(A, ↑ n2s + h),
E(e, hs hs+1 + h) ⊂ E(A, B) ⇒
E(e, hs hs+1 + h) ⊂ ∂E(A, B).
Proof:
Based on Section 8.9.6,
int(h1 h2 · · · hv−1 hv ∩ ↑ n1r ) = ∅ ⇒
E(↑ n1r , h1 h2 · · · hv−1 hv ) = − ↑ n1r + a0
E(↑ n1r + e, B) = − ↑ n1r + h − e + a0 .
E( e r + e, hs + h) ⊂ ∂E(A, B)
Proof:
From Section 8.9.7,
↔
↔
int( e r1 e r2 + e r ∩ hs1 hs2 + hs ) = ∅ ⇒
∃n0 , n0 e r × hs ,
↔
↔
E( e r1 e r2 + e r , hs1 hs2 + hs =↑ n0 + a0 .
Shi’s new contact theory 337
↔
↔
E( e r1 e r2 + e r + e, hs1 hs2 + hs + h ⊃ E(A, B).
⇒↑ n0 + h − e + a0 ⊃ E(A, B).
= (− e r )hs + h − e + a0 ⊂ ⊥ n0 + h − e + a0 ⇒
E( e r + e, hs + h)
↔
↔
⊂ ∂E( e r1 e r2 + e r + e, hs1 hs2 + hs + h).
E( e r + e, hs + h) ⊂ E(A, B) ⇒
E( e r + e, hs + h) ⊂ ∂E(A, B).
Proof:
From Section 8.9.10,
Therefore,
A = e 1 e 2 · · · e u−1 e u + e,
B = h1 h2 · · · hv−1 hv + h.
e r e r+1 , r = 1, . . . , u,
n1r = + e r+1 × e r ,
Shi’s new contact theory 339
Figure 8.103 Entrance 3D concave solid angle of two 2D concave solid angles.
n1r = − e r+1 × e r .
The inner normal vector of hs hs+1 is n2s . For all face angles of B
hs hs+1 , s = 1, . . . , v,
If face hs hs+1 is convex,
n2s = +hs+1 × hs .
If face hs hs+1 is concave,
n2s = −hs+1 × hs .
Based on Section 8.9.10, all contact angles are searched and examined as follows.
The following theorem is the conclusion of Sections 8.9.6 and 8.9.7.
The contact angles connected and intersected with each other to form a 3D angle
which is the 3D entrance angle E(A, B). If no 3D angle can be formed, E(A, B) is the
whole space.
Three dimensional contacts are much more complicated and difficult than the two
dimensional counterpart. However, the 3D entrance blocks are computed in a similar
way with 2D cases. The way to compute 3D entrance block is also to compute the
boundary of entrance block. The following theorems are for understanding and finding
the boundaries of 3D entrance blocks. The boundaries of 3D entrance blocks are the
subsets of the contact polygons.
Proof:
If a ∈ int(A), b ∈ B,
E(a, b) = b − a + a0 ∈ int(b − A + a0 )
= int(E(A, b)) ⊂ int(E(A, B)).
If a ∈ A, b ∈ int(B),
E(a, b) = b − a + a0 ∈ int(B − a + a0 )
= int(E(a, B)) ⊂ int(E(A, B)).
If a ∈ int(A) or b ∈ int(B),
E(a, b) = b − a + a0
Proof:
From Section 8.11.1,
Assume
a1 a2 · · · ap−1 ap , ap+1 = a1
is a polygon of A. The inner normal vector is m1k ,
b1 b2 · · · bq−1 bq , bq+1 = b1
is a polygon of B. The inner normal vector is m2l , (Figures 8.104).
If m1k and m2l are not parallel, for
a ∈ int(a1 a2 · · · ap−1 ap ),
b ∈ int(b1 b2 · · · bq−1 bq ),
E(a, b) = b − a + a0 ,
∃v1 , v2 , (v1 v2 ) · m2l = 0,
⊂ int(E(A, B)).
E(a, b) ∈ int(E(A, B)).
E(a1 a2 · · · ap−1 ap , b1 b2 · · · bq−1 bq ) ∩ ∂E(A, B) ⊂
(E(∂a1 a2 · · · ap−1 ap , b1 b2 · · · bq−1 bq ) ∪ E(a1 a2 · · · ap−1 ap , ∂b1 b2 · · · bq−1 bq )).
Therefore
Proof:
ai ai+1 , i = 1, . . . , p,
b1 b2 · · · bq−1 bq .
are edge and polygon of 3D blocks A and B respectively, (Figure 8.105). For any points
a ∈ int(ai ai+1 ),
b ∈ int(b1 b2 · · · bq−1 bq ),
If
Therefore, if
Figure 8.105 Entrance point of a face polygon inner point and an edge inner point.
Therefore
b1 b2 · · · bq−1 bq , bq+1 = b1
is a polygon of B. The normal vector of polygon is m2l . Given a vertex e of A, all
vertices connected with e by edges are (Figures 8.106 & 8.107)
e1 e2 · · · eu−1 eu , eu+1 = e1 .
Proof:
If
int( e 1 e 2 · · · e u−1 e u ∩ ↑ m2l ) = ∅,
E(e, b) = b − e + a0 ∈
Shi’s new contact theory 345
Similarly,
a1 a2 · · · ap−1 ap , ap+1 = a1
is a polygon of A. The normal vector of polygon is m1k .
Given a vertex h of B, all vertices connected with h by edges are
h1 h2 · · · hv−1 hv , hv+1 = h1 .
Denote hs = (hhs ), n2s as the inner normal of hs hs+1 .
Proof:
If
int(h1 h2 · · · hv−1 hv ∩ ↑ m1k ) = ∅,
∃ h0 , hh0 ⊂ B, (hh0 ) · m1k > 0.
E(a, h) = h − a + a0 ∈
int(h − a1 a2 · · · ap−1 ap + a0 ) ⊂
int((h − A + a0 )∪(hh0 − a1 a2 · · · ap−1 ap + a0 ))
= int(E(A, h)∪E(a1 a2 · · · ap−1 ap , hh0 ))
⊂ int(E(A, B)) ⇒
E(a, h) ⊂ int(E(A, B)) ⇒ E(a, h) ∈ / ∂E(A, B).
inner normal vectors of hs−1 hs and hs hs+1 are n21 and n22 respectively.
346 Discontinuous deformation analysis in rock mechanics practice
e r n12 × n11 ,
hs n22 × n21 .
Denote
e r1 = n11 × n12 × n11 ,
hs1 = n21 × n22 × n21 ,
hs2 = n22 × n21 × n22 ,
Shi’s new contact theory 347
Proof:
If
int((↑ n11 ∩ ↑ n12 ) ∩(↑ n21 ∩ ↑ n22 )) = ∅,
∃n0 , n0 e r × hs ,
∃i, j, 1 ≤ i ≤ 2, 1 ≤ j ≤ 2,
e ri · n0 > 0,
hsj · n0 > 0.
Therefore there is a vector n1 that satisfies the related equations.
The following theorems are based on the assumption of blocks A and B contacting
along the parallel edges and polygons. Assume A and B are the same blocks as in
Sections 8.11.2 and 8.11.3.
ai ai+1 , i = 1, . . . , p,
b1 b2 · · · bq−1 bq , j = 1, . . . , q.
Proof:
Similar as in Section 8.4.2,
Here
∂ai ai+1 = ai ∪ ai+1 ,
q
∂b1 b2 · · · bq−1 bq = ∪j=1 bj bj+1 .
P(ai ai+1 ) is a polygon having ai ai+1 as its boundary, and P(ai+1 ai ) is another
polygon having ai+1 ai as its boundary.
The inner normal vectors of
are n11 and n12 respectively.
ai ai+1 · m2l = 0,
Proof:
If
int((↑ n11 ∩ ↑ n12 ) ∩ ↑ m2l ) = ∅,
∃k, 1 ≤ k ≤ 2,
e rk m2l > 0.
E(a, b) ∈ int(E(ai ai+1 , b1 b2 · · · bq−1 bq )) ⊂
int(E(Pk , b1 b2 · · · bq−1 bq ) ∪ E(ai ai+1 , B)) ⊂
⊂ int(E(A, B) ∪ E(A, B)) = int(E(A, B)).
⇒ E(a, b) ∈ int(E(A, B)), E(a, b) ∈
/ ∂E(A, B).
a1 a2 · · · ap−1 ap , ap+1 = a1
is a polygon of A. The inner normal vector is m1k .
b1 b2 · · · bq−1 bq , bq+1 = b1
is a polygon of B (Figures 8.110), where the inner normal vector is m2l .
A0 = a1 a2 · · · ap−1 ap ,
B0 = b1 b2 · · · bq−1 bq ,
m1k m2l .
350 Discontinuous deformation analysis in rock mechanics practice
Here
p
∂a1 a2 · · · ap−1 ap = ∪i=1 ai ai+1 .
q
∂b1 b2 · · · bq−1 bq = ∪j=1 bj bj+1 .
m1k m2l
Proof:
If
m1k m2l ⇒
∃δ,
C = {|x| ≤ δ},
Shi’s new contact theory 351
(C ∩ ↑ m1k ) + a ⊂ A,
(C ∩ ↑ m2l ) + b ⊂ B ⇒
E(a, b) ∈
int((C ∩ ↑ m2l ) − (C ∩ ↑ m1k ) + b − a + a0 ) =
int(E((C ∩ ↑ m1k ) + a, (C ∩ ↑ m2l ) + b)) ⊂
int(E(A, B)).
⇒ E(a, b) ∈ int(E(A, B)), E(a, b) ∈
/ ∂E(A, B) ⇒
⇒ m1k −m2l .
E(e, b1 b2 · · · bq−1 bq )
C(0, 2) = ∪contact polygon E(e, b1 b2 · · · bq−1 bq )
E(eer , hhs ),
C(1, 1) = ∪contact polygon E(eer , hhs )
The simplest 3D contact case is the ball-to-ball contact with an obvious entrance
ball. Geometrically, balls are convex. Most of the blocks are convex. According to
the convex block theorem (Section 8.4.9), the entrance block of two convex blocks
is convex. Therefore, the computation of contact polygons of 3D convex entrance
blocks is about few times of the computation of ball cases. Complex blocks are union
of convex blocks.
The contact surface of a 3D entrance block is formed by contact polygons.
Each contact polygon corresponds to a vertex-to-polygon contact or an edge-to-edge
contact.
In the process of computing, time steps are used. At the beginning of each time step,
the contact polygons are computed. From the relative position of reference point a0
and the contact polygons, the closed contact point pairs are found. The closed contact
point pairs are defined by contact polygons and control the movements, rotations and
deformations throughout this time step.
a1 a2 · · · ap−1 ap , ap+1 = a1 .
354 Discontinuous deformation analysis in rock mechanics practice
The normal vector m1k of face k points inward. The faces of block B are polygons
which rotate the outward block. The vertices of face l are
b1 b2 · · · bq−1 bq , bq+1 = b1 .
The normal vector m1k of face l points the inside of block B.
a1k is any point on the face plane k of A, (a1k ai ) = a i .
b1l is any point on the face plane l of B, (b1l bj ) = bj .
p
m1k = a i+1 × a i .
i=1
q
m2l = bj+1 × bj .
j=1
The inner normal vector of angle er er+1 is n1r . Since A is convex, er er+1 is a
convex angle,
n1r = + e r+1 × e r .
h1 h2 · · · hv−1 hv , hv+1 = h1 , (hhs ) = hs .
The inner normal vector of angle hs hs+1 is n2s , which points the inside of block B.
n2s = +hs+1 × hs .
Proof:
Based on Section 8.9.6,
e 1 e 2 · · · e u−1 e u + e ⊃ A ⇒
Proof:
Based on Section 8.9.6,
int(h1 h2 · · · hv−1 hv ∩ ↑ m1k ) = ∅ ⇒
E(↑ m1k , h1 h2 · · · hv−1 hv ) = − ↑ m1k + a0 .
h1 h2 · · · hv−1 hv + h ⊃ B ⇒
E(↑ m1k + a1 , B) ⊂ − ↑ m1k + h − a1 + a0 .
↑ m1k + a1 ⊃ A, ⇒ E(↑ m1k + a1 , B) ⊃ E(A, B).
356 Discontinuous deformation analysis in rock mechanics practice
As defined in Section 8.11.5, eer is an edge of A; e and er are vertices of block A. hhs
is an edge of B; h and hs are vertices of block B. The inner normal vectors of e r−1 e r
and e r e r+1 are n11 and n12 respectively. The inner normal vectors of hs−1 hs and
hs hs+1 are n21 and n22 respectively. As blocks A, B are convex, the edges of e r and hs
are convex
hs = (hhs ) n22 × n21 .
Denote
hs1 = n21 × n22 × n21 ,
hs2 = n22 × n21 × n22 ,
Proof:
From Section 8.9.7,
↔
↔
int( e r1 e r2 + e r ∩ hs1 hs2 + hs ) = ∅ ⇒
∃n0 , n0 e r × hs ,
↔
↔
E( e r1 e r2 + e r , hs1 hs2 + hs ) =↑ n0 + a0 .
Shi’s new contact theory 357
↔
↔
e r1 e r2 + e r + e ⊃ A, hs1 hs2 + hs + h ⊃ B ⇒
↔
↔
E( e r1 e r2 + e r + e, hs1 hs2 + hs + h) ⊃ E(A, B).
⇒↑ n0 + h − e + a0 ⊃ E(A, B).
Proof:
From Section 8.11.8,
Therefore,
a1 a2 · · · ap−1 ap , ap+1 = a1 .
358 Discontinuous deformation analysis in rock mechanics practice
Normal vector m1k of polygon k points the inside of block A. The faces of block
B are polygons which rotate outward. The vertices of face l are
b1 b2 · · · bq−1 bq , bq+1 = b1 .
The normal vector m2l of polygon l points the inside of block B.
a1k is any point on the polygon plane k of A,
(a1k ai ) = a i
(b1l bj ) = bj
p
m1k = a i+1 × a i .
i=1
q
m2l = bj+1 × bj .
j=1
Based on Section 8.11.8, all contact polygons are searched and examined. The
following theorem is the conclusion of Sections 8.11.4 and 8.11.5.
The contact polygons of 3D entrance block E(A, B) are connected and intersected
with each other to form E(A, B). This process, which is called cutting, is a stan-
dard computing geometric process to form 3D entrance blocks from contact polygons.
In this process, all contact polygons are input polygons.
Shi’s new contact theory 359
8.13 CONCLUSIONS
In this study, a general contact theory for 2D and 3D discontinuous computation was
proposed along with a new definition for operating the point sets named as the entrance
block. The entrance block is defined and represented by its boundary.
Further, the solutions of the contact boundary are analytically derived. The contact
boundary is proved to be a cover system, with each cover corresponding to a possible
360 Discontinuous deformation analysis in rock mechanics practice
contact. The contact boundaries are of similar formulas in all conditions (2D or 3D,
angles or blocks, convex or concave), e.g., all the local contacts which are either
in vertex-to-polygon form or in edge-to-edge form, have similar or almost identical
formulas for the contacts of 3D solid angles, the contacts of 3D convex blocks, and
the contacts of general 3D blocks. These consistencies will make the computation of
an entrance block much easier.
Given a reference point, the entrance block can be solved and thus simplify the
contact computation in the following ways:
First, the entrance block concept simplifies the processes of contact searching prior
to the occurrence of contacts. The shortest distance between two blocks is the shortest
distance between the reference point and the contact surface. The point pairs with the
shortest distance can be found from the contact cover system.
Second, the entrance block concept simplifies the definition and searching routines
of the first entrance, as soon as the reference point outside the entrance block moves
onto one certain contact surface.
Last but not least, the entrance block concept simplifies the searching routines of
the shortest path of exit. If the reference point is already inside the entrance block,
it will exit the entrance block in the shortest path. The shortest path connected with
the closest point on the contact surface can be computed and the contact points are
determined.
When two real blocks approach each other, there are most likely two possibilities:
vertex to polygon contacts and edge to edge contacts in loose block system. Due to the
small number of contacts, loose blocks are much easier to compute.
In destabilization or damage computation, solids are changed from the continuous
state to discontinuous state. The analyses were based on starting with newly divided
and unperturbed block meshes, where there are no voids in between. The quantities
of contacts in this case are higher than those in loose blocks.
In such a case, contacts often happen along parallel edges and polygons or along
a pair of parallel polygons where there are infinite contact point pairs. Among these
infinite contact point pairs, only a few vertex-to-polygon and edge-to-edge contact
point pairs control the movements of blocks, especially rotations.
With this new theory, it is now possible to compute contacts with infinite contact
point pairs, such as contacts in densely compacted blocks. In this theory, these vertex-
to-polygon contact point pairs and edge-to-edge contact point pairs are called contact
points, which correspond with overlapping contact polygons. As a result, after small
rotation or deformation, blocks contact only along these few contact points.
Based on this theory, all closed contact points control the movements, rotations
and deformations of blocks, as in real cases.
Given a block system, the contact points should be the same as in any computation
method, as well as the means of finding contact points.
As the discontinuous computation follows time steps, if one contact point in one
time step was not correct, the whole computation would not be useful and large pene-
tration would happen. However the method of “shortest exit’’, i.e., using the shortest
path for the reference point to exit the entrance block, will recover the block sys-
tem from a penetrated state “to a’’ contact state. This will make the discontinuous
computation more reliable.
The proposed contact theory is applicable to all computational methods, in areas
of civil, structural, geological, mechanical, and robotic engineering.
References
Ahn, T.Y. & Song, J.J. (2011) New contact-definition algorithm using inscribed spheres for 3D
discontinuous deformation analysis. International Journal of Computational Methods, 8 (2),
171–191.
Akao, S., Ohnish, Y., Nishiyama, S. & Nishimura, T. (2007) Comprehending DDA for
block behaviour under dynamic condition. In: Ju, Y., Fang, X. & Bian, H. (eds.) The
Eighth International Symposium on Analysis of Discontinuous Deformation, Beijing, China.
pp. 135–140.
Amadei, B., Lin, C. & Dewyer, J. (1996) Recent extensions to DDA. In: Salami, M.R. & Banks,
D. (eds.) First International Forum on Discontinuous Deformation Analysis and Simulations
of Discontinuous Media. Berkeley, CA, TSI Press. pp. 1–30.
Ashby, J.P. (1971) Sliding and Toppling Modes of Failure in Models and Jointed Rock Slopes.
M Sc Thesis. London, Inst. of Geol. Sciences., University of London, Imperial College.
Ashford, S.A., Sitar, N., Lysmer, J. & Deng, N. (1997) Topographic effects on the seismic
response of steep slopes. Bulletin of the Seismological Society of America, 87, 701–709.
Bahat, D. (1991) Tectonofractography. Berlin, Springer-Verlag. 453 pp.
Bai, T. & Gross, M.R. (1999) Theoretical analysis of cross-joint geometries and their
classification. Journal of Geophysical Research-Solid Earth, 104, 1163–1177.
Bakun-Mazor, D. (2011) Modeling Dynamic Rock Mass Deformation with the Numerical DDA
Method. PhD Thesis. Beer Sheva, Dept. of Geological and Environmental Sciences, Ben-
Gurion University of the Negev. 102 pp.
Bakun-Mazor, D., Hatzor, Y.H. & Dershowitz, W.S. (2009a) Modeling mechanical layering
effects on stability of underground openings in jointed sedimentary rocks. International
Journal of Rock Mechanics and Mining Sciences, 46, 262–271.
Bakun-Mazor, D., Hatzor, Y.H. & Glaser, S.D. (2009b) 3D DDA vs. analytical solutions for
dynamic sliding of a tetrahedral wedge. In: ICADD9 Nanyang Technological University,
Singapore, 25–27 November 2009.
Bakun-Mazor, D., Hatzor, Y.H. & Glaser, S.D. (2012) Dynamic sliding of tetrahedral wedge:
The role of interface friction. International Journal for Numerical and Analytical Methods
in Geomechanics, 36, 327–343.
Bakun-Mazor, D., Hatzor, Y.H., Glaser, S.D. & Santamarina, J.C. (2013) Thermally vs. seis-
mically induced block displacements in Masada rock slopes. International Journal of Rock
Mechanics and Mining Sciences, 61, 196–211.
Balden, V., Scheele, F. & Nurick, G. (2001) Discontinuous deformation analysis in ball milling.
In: Bicanic, N. (ed.) The Fourth International Conference on Analysis of Discontinuous
Deformation, Glasgow, Scotland, UK. pp. 337–348.
Bao, H.R. & Zhao, Z. (2010a) An alternative scheme for the corner-corner contact in the two-
dimensional discontinuous deformation analysis. Advances in Engineering Software, 41,
206–212.
362 References
Bao, H.R. & Zhao, Z.Y. (2010b) Modelling crack propagation with nodal-based discontinuous
deformation analysis. In: Ma, G. & Zhou, Y. (eds.) Analysis of Discontinuous Deforma-
tion: New Developments and Applications. Singapore, Nanyang Technological University.
pp. 161–167.
Bao, H.R. & Zhao, Z.Y. (2013) Modelling brittle fracture with the nodal-based discontinuous
deformation analysis. International Journal of Computational Methods, 10 (6), 1350040
(26 pp.).
Bao, H.R., Hatzor, Y.H. & Huang, X. (2012) A new viscous boundary condition in the two-
dimensional discontinuous deformation analysis method for wave propagation problems.
Rock Mechanics and Rock Engineering, 45, 919–928.
Bao, H.R., Yagoda-Biran, G. & Hatzor, Y.H. (2014a) Site response analysis with two-
dimensional numerical discontinuous deformation analysis. Earthquake Engineering and
Structural Dynamics, 43, 225–246.
Bao, H.R., Zhao, Z. & Tian, Q. (2014b) On the implementation of augmented Lagrangian
method in the tow-dimensional discontinuous deformation analysis. International Journal
for Numerical and Analytical Methods in Geomechanics, 38, 551–571.
Bard, P.Y. & Tucker, B.E. (1985) Underground and ridge and site effects: Compari-
son of observation and theory. Bulletin of the Seismological Society of America, 75,
905–922.
Barla, G. & Paronuzzi, P. (2013) The 1963 Vajont landslide: 50th anniversary. Rock Mechanics
and Rock Engineering, 46, 1267–1270.
Barla, G., Monacis, G., Perino, A. & Hatzor, Y.H. (2010) Distinct element modelling in static and
dynamic conditions with application to an underground archaeological site. Rock Mechanics
and Rock Engineering, 43, 877–890.
Barton, N. (1976) The shear strength of rock and rock joints. International Journal of Rock
Mechanics and Mining Sciences, 13, 255–279.
Barton, N., Lien, R. & Lunde, J. (1974) Engineering classification of jointed rock masses for
the design of tunnel support. Rock Mechanics, 6, 189–236.
Beer, M. & Meek, J.L. (1982) Design curves for roofs and hanging-walls in bedded rock based on
Voussoir beam and plate solutions. Transactions of the Institution of Mining and Metallurgy,
91, 18–22.
Belytschko, T. & Neal, M.O. (1991) Contact-impact by the pinball algorithm with penalty
and Lagrangian methods. International Journal for Numerical Methods in Engineering, 31,
547–572.
Ben, Y.X., Xue, J., Miao, Q.H. & Wang, Y. (2012) Coupling fluid flow with discontinuous
deformation analysis. In: Zhao, J., Ohnishi, Y., Zhao, G.F. & Sasaki, T. (eds.) Advances in
Discontinuous Numerical Methods and Applications in Geomechanics and Geoengineering.
pp. 107–112.
Ben, Y.X., Wang, Y. & Shi, G.H. (2013) Development of a model for simulating hydraulic
fracturing with DDA. In: Proceedings of the 11th International Conference on Analysis
of Discontinuous Deformation (ICADD’13), 27–29 August. Fukuoka, Taylor & Francis
Group. pp. 169–175.
Benson, D.J. & Hallquist, J.O. (1990) A single surface contact algorithm for the post-buckling
analysis of shell structures. Computers Methods in Applied Mechanics and Engineering, 78,
141–163.
Beyabanaki, S.A.R. & Jafari, A. (2005) Modified point-to-face frictionless contact constraints
for three-dimensional discontinuous deformation analysis (3-D DDA). In: Proceedings of the
Seventh International Conference on Analysis of Discontinuous Deformation (ICADD-7),
Hawaii. pp. 24–36.
Beyabanaki, S.A.R., Mikola, R.G. & Hatami, K. (2008) Three-dimensional discontinuous
deformation analysis (3-D DDA) using a new contact resolution algorithm. Computers and
Geotechnics, 35 (3), 346–356.
References 363
Beyabanaki, S.A.R., Jafari, A. & Yeung, M.R. (2009a) Second-order displacement functions for
three-dimensional discontinuous deformation analysis (3-D DDA). International Journal of
Science and Technology, 16 (3), 216–225.
Beyabanaki, S.A.R., Jafari, A., Biabanaki, S.O.R. & Yeung, M.R. (2009b) A coupling model
of 3-D discontinuous deformation analysis (3-D DDA) and finite element method. Arabian
Journal for Science and Engineering, 34 (1B), 108.
Beyabanaki, S.A.R., Jafari, A., Biabanaki, S.O.R. & Yeung, M.R. (2009c) Nodal-based three-
dimensional discontinuous deformation analysis (3-D DDA). Computers and Geotechnics,
36 (3), 359–372.
Beyabanaki, S.A.R., Jafari, A. & Yeung, M.R. (2010) High-order three-dimensional discon-
tinuous deformation analysis (3-D DDA). International Journal for Numerical Methods in
Biomedical Engineering, 26 (12), 1522–1547.
Bicanic, N. & Stirling, C. (2001) DDA analysis of the Couplet/Heyman minimum thickness
arch problem. In: Bicanic, N. (ed.) The Fourth International Conference on Analysis of
Discontinuous Deformation, Glasgow, Scotland, UK. pp. 165–170.
Bicanic, N., Stirling, C. & Pearce, C.J. (2003) Discontinuous modelling of masonry bridges.
Computational Mechanics, 31, 60–68.
Bieniawski, Z.T. (1970) Time-dependent behaviour of fractured rock. Rock Mechanics, 2,
123–137.
Bieniawski, Z.T. (1974) Geomechanics classification of the rock masses and its application in
tunneling. In: Proceedings of the Third Congress ISRM, Denver. pp. 7–32.
Bieniawski, Z.T. (1976) Rock mass classification in rock engineering. In: Bieniawski, Z.T.
(ed.) Exploration for Rock Engineering, Proc. of the Symp. Vol. 1. Cape Town, Balkema.
pp. 97–106.
Bin Shi, G., Bai, J., Wang, M., Sun, B., Wang, Y. & Shi, G. (2010) Study on failure characteristics
and support measure of layer structure-cataclasm rock mass. In: Ma, G. & Zhou, Y. (eds.)
Analysis of Discontinuous Deformation: New Developments and Applications. Singapore,
Nanyang Technological University. pp. 135–143.
Bonnet, J. & Peraire, J. (1991) An alternating digital tree (ADT) algorithm for 3D geomet-
ric searching and intersection problem. International Journal for Numerical Methods in
Engineering, 31, 1–17.
Boore, D.M. & Joyner, W.B. (1997) Site amplifications for generic rock sites. Bulletin of the
Seismological Society of America, 87, 327–341.
Bouchon, M. & Barker, J.S. (1996) Seismic response of a hill: The example of Tarzana,
California. Bulletin of the Seismological Society of America, 86, 66–72.
Brady, B.H.G. & Brown, E.T. (2004) Rock Mechanics for Underground Mining. 3rd edition.
Dordrecht, Kluwer Academic Publisher.
Bray, J.W. & Goodman, R.E. (1981) The theory of base friction models. International Journal
of Rock Mechanics and Mining Sciences, 18, 453–468.
Brekke, T.L. Personal Communication. U. C. Berkeley graduate level course entitled: Geological
engineering of underground openings.
Cai, M. & Kaiser, P.K. (2005) Assessment of excavation damaged zone using a micromechanics
model. Tunnelling and Underground Space Technology, 20, 301–310.
Cao, C.Y., Zhong, Y. & Shi, G.-H. (2007) Application of DDA-FEM coupled method in pave-
ment analysis. In: Ju, Y., Fang, X. & Bian, H. (eds.) The Eighth International Symposium
on Analysis of Discontinuous Deformation, Beijing, China. pp. 111–116.
Celebi, M. (1987) Topographical and geological amplifications determined from strong-motion
and aftershock records of the 3 March 1985 Chile earthquake. Bulletin of the Seismological
Society of America, 77, 1147–1167.
Chavez-Garcia, F.J., Sanchez, L.R. & Hatzfeld, D. (1996)Topographic site effects and HVSR –
A comparison between observations and theory. Bulletin of the Seismological Society of
America, 86, 1559–1573.
364 References
Chen, G. & Ohnish, Y. (1999) A non-linear model for discontinuities in DDA. In: Amadei, B.
(ed.) The Third International Conference on Analysis of Discontinuous Deformation, Vail,
Colorado. pp. 57–64.
Chen, G., Zheng, L. & Zhang, Y. (2013) Practical applications of DDA to disaster preven-
tion. In: Chen, G., Ohnishi, Y., Zheng, L. & Sasaki, T. (eds.) The 11th International
Symposium on Analysis of Discontinuous Deformation. Fukuoka, Taylor & Francis Group.
pp. 15–28.
Chen, W. & Deng, J. (2008) Application of key block theory and DDA to the stability analysis
of underground powerhouse of Jinping hydropower station I. In: Ju, Y., Fang, X. & Bian,
H. (eds.) The Eighth International Symposium on Analysis of Discontinuous Deformation,
Beijing, China. pp. 243–247.
Cheng, Y.M. & Zhang, Y.H. (2000) Rigid body rotation and block internal discretiza-
tion in DDA analysis. International Journal for Numerical and Analytical Methods in
Geomechanics, 24, 567–578.
Chern, J.C., Koo, C.Y. & Chen, S. (1995) Development of second order displacement function
for DDA and manifold method. In: Working Forum on the Manifold Method of Material
Analysis, Vicksburg. pp. 183–202.
Choo, L.Q., Zhao, Z., Chen, H. & Tian, Q. (2016) Hydraulic fracturing modeling using the
discontinuous deformation analysis (DDA) method. Computers and Geotechnics, 76, 12–22.
Clapp, K.K. & MacLaughlin, M. (2003) Preliminary development of a fully grouted rock bolt
element for discontinuous deformation analysis. In: Lu, M. (ed.) The Sixth International
Conference on Analysis of Discontinuous Deformation. Trondheim, Balkema Publishers.
pp. 111–118.
Clatworthy, D. & Scheele, F. (1999) A method for sub-meshing in discontinuous deformation
analysis (DDA). In: Amadei, B. (ed.) The Third International Conference on Analysis of
Discontinuous Deformation, Vail, Colorado. pp. 85–94.
Cook, N.G.W. (1966) Rock Mechanics Applied to the Study of Rockbursts. Johannesburg,
South African Institute of Mining and Metallurgy.
Cook, N.G.W. (1976) Seismicity associated with mining. Engineering Geology, 10, 99–122.
Crawford, A.M. & Curran, J.H. (1981) The influence of shear velocity on the frictional resis-
tance of rock discontinuities. International Journal of Rock Mechanics and Mining Sciences,
18, 505–515.
Crawford, A.M. & Curran, J.H. (1982) The influence of rate-dependent and displacement-
dependent shear resistance on the response of rock slopes to seismic loads. International
Journal of Rock Mechanics and Mining Sciences, 19, 1–8.
Cundall, P.A. (1971) A computer model for simulating progressive, large scale movements in
blocky rock system. In: Symposium of International Society of Rock Mechanics, Nancy,
France. pp. 11–18.
Cundall, P.A. (1988) Formulation of a three-dimensional distinct element model-Part I: A
scheme to detect and represent contacts in a system composed of many polyhedral blocks.
International Journal of Rock Mechanics and Mining Sciences & Geomechanics, 25,
107–116.
Davis, L.L. & West, L.R. (1973) Observed effects of topography on ground motion. Bulletin of
the Seismological Society of America, 63, 283–298.
De Luca, A., Giordano, A. & Mele, E. (2004) A simplified procedure for assessing the seismic
capacity of masonry arches. Engineering Structures, 26, 1915–1929.
Dershowitz, B., LaPointe, P., Eiben, T. & Wei, L.L. (2000) Integration of discrete feature network
methods with conventional simulator approaches. SPE Reservoir Evaluation & Engineering,
3, 165–170.
Dershowitz, W.S. & Einstein, H.H. (1988) Characterizing rock joint geometry with joint system
models. Rock Mechanics and Rock Engineering, 21, 21–51.
References 365
Desai, C.S. & Abel, J.F. (1972) Introduction to the Finite Element Method. New York, NY, Van
Nostrand-Reinhold.
Diederichs, M.S. & Kaiser, P.K. (1999a) Authors’ reply to discussion by A.L. Sofianos regard-
ing Diederichs M.S. and Raiser P.K. Stability of large excavations in laminated hard rock
masses: The voussoir analogue revisited. International Journal of Rock Mechanics and
Mining Sciences, 36, 97–117, 995–997.
Diederichs, M.S. & Kaiser, P.K. (1999b) Stability of large excavations in laminated hard rock
masses: The voussoir analogue revisited. International Journal of Rock Mechanics and
Mining Sciences, 36, 97–117.
Diederichs, M.S. & Kaiser, P.K. (1999c) Tensile strength and abutment relaxation as failure
control mechanisms in underground excavations. International Journal of Rock Mechanics
and Mining Sciences, 36, 69–96.
Dieterich, J.H. (1972) Time-dependent friction in rocks. Journal of Geophysical Research, 77,
3690–3697.
Dieterich, J.H. (1978) Time-dependent friction and the mechanism of stick-slip. Pure and
Applied Geophysics, 116, 790–806.
Dieterich, J.H. (1979) Modeling of rock friction, 1: Experimental results and constitutive
equations. Journal of Geophysical Research, 84, 2161–2168.
Dieterich, J.H. (1981) Constitutive properties of faults with simulated gouge. In: Carter, N.L.,
Friedman, M., Logan, J.M. & Stearns, D.W. (eds.) Mechanical Behavior of Crustal Rocks.
The Handin Volume. Washington, DC, American Geophysical Union. pp. 103–120.
Dieterich, J.H. & Kilgore, B.D. (1994) Direct observation of frictional contacts: New insights
for state-dependent properties. Pure and Applied Geophysics, 143, 283–302.
Dieterich, J.H. & Kilgore, B.D. (1996) Imaging surface contacts: Power law contact distribu-
tions and contact stresses in quartz, calcite, glass and acrylic plastic. Tectonophysics, 256,
219–241.
Dong, P.H. & Osada, M. (2007) Effects of dynamic friction on sliding behaviour of block in
DDA. In: Ju, Y., Fang, X. & Bian, H. (eds.) The Eighth International Symposium on Analysis
of Discontinuous Deformation, Beijing, China. pp. 129–134.
Dong, X., Wu, A. & Ren, F. (1996) A preliminary application of DDA to the three gorges project
on Yangtze river, China. In: Salami, M.R. & Banks, D. (eds.) First International Forum on
Discontinuous Deformation Analysis and Simulations of Discontinuous Media. Berkeley,
CA, TSI Press. pp. 310–317.
Doolin, D.M. & Sitar, N. (2002) Displacement accuracy of discontinuous deformation anal-
ysis method applied to sliding block. Journal of Engineering mechanics (ASCE), 128 (11),
1158–1168.
Doolin, D.M. & Sitar, N. (2004) Time integration in discontinuous deformation analysis.
Journal of Engineering mechanics (ASCE), 130 (3), 249–258.
Dunlop, P., Dunne, J.M. & Seed, H.B. (1970) Finite element analysis of slopes in soils. Journal
of the Soil Mechanics and Foundations Division, ASCE, 96, 471–493.
Einstein, H.H., Veneziano, D., Baecher, G.B. & Oreilly, K.J. (1983) The effect of discontinuity
persistence on rock slope stability. International Journal of Rock Mechanics and Mining
Sciences, 20, 227–236.
Fairhurst, C. & Cook, N.G.W. (1966) The phenomenon of rock splitting parallel to the direc-
tion of maximum compression in the neighbourhood. In: 1st Congress ISRM, Lisbon.
pp. 296–689.
Ferrero, A.M., Forlani, G., Roncella, R. & Voyat, H.I. (2009) Advanced geostructural survey
methods applied to rock mass characterization. Rock Mechanics and Rock Engineering, 42,
631–665.
Fisher, F.R.S. (1953) Dispersion on a sphere. Philosophical Transactions of the Royal Society of
London, 217, 295–305.
366 References
FLAC (1998) Fast Lagrangian Analysis of Continua (FLAC). In: Group, I.C. (ed.)
Minneapolis, MN.
Fredrich, J.T., Evans, B. & Wong, T.F. (1990) Effect of grain-size on brittle and semibrittle
strength – Implications for micromechanical modeling of failure in compression. Journal of
Geophysical Research-Solid Earth and Planets, 95, 10907–10920.
Fu, G.Y. (2014) Realistic Rock Mass Modelling and Its Engineering Applications. PhD Disser-
tation Thesis. Crawley, WA, School of Civil, Environmental and Mining Engineering, The
University of Western Australia.
Fu, G.Y., Ma, G.W. & Qu, X.L. (2015a) Dual reciprocity boundary element based block
model for discontinuous deformation analysis. Science China-Technological Sciences, 58 (9),
1575–1586.
Fu, X., Sheng, Q., Zhang, Y., Zhou, Y. & Dai, F. (2015b) Boundary setting method for the
seismic dynamic response analysis of engineering rock mass structures using the discontinuous
deformation analysis method. International Journal for Numerical and Analytical Methods
in Geomechanics. Available from: 10.1002/nag.2374.
Fu, G.Y., Ma, G.W. & Qu, X.L. (2016) Boundary element based discontinuous deformation
analysis. International Journal for Numerical and Analytical Methods in Geomechanics.
Available from: 10.1002/nag.2661.
Gilbert, E.G., Johnson, D.W. & Keerthi, S.S. (1988) A fast procedure for computing the dis-
tance between complex objects in three-dimensional space. IEEE Journal on Robotics and
Automation, 4, 193–203.
Goodman, R.E. (1976) Methods of Geological Engineering in Discontinuous Rocks. San
Francisco, CA, West Publishing Company.
Goodman, R.E. (1989) Introduction to Rock Mechanics. 2nd edition. New York, NY, John
Wiley & Sons.
Goodman, R.E. (2013) Toppling – A fundamental failure mode in discontinuous materials –
Description and analysis. In: Geotechnical Special Publication No. 231 “Geo-Congress 2013:
“Stability and Performance of Slopes and Embankments III’’ 231.
Goodman, R.E. & Bray, J.W. (1976) Toppling of rock slopes. In: Specialty Conference on Rock
Engineering for Foundations and Slopes. Boulder, CO, American Society of Civil Engineering.
pp. 201–234.
Goodman, R.E. & Brown, C.B. (1963) Dead load stresses and the instability of slopes. Journal
of the Soil Mechanics and Foundations Division, ASCE, 89, 103–104.
Goodman, R.E. & Kieffer, D.S. (2000) Behavior of rock in slopes. Journal of Geotechnical and
Geoenvironmental Engineering, 126, 675–684.
Goodman, R.E. & Seed, H.B. (1966) Earthquake induced displacements in sands and
embankments. Journal of the Soil Mechanics and Foundations Division, ASCE, 92 (SM2),
125–146.
Goodman, R.E. & Shi, G.H. (1985) Block Theory and Its Application to Rock Engineering.
Englewood Cliffs, NJ, Prentice-Hall, Inc.
Goodman, R.E., Taylor, R.L. & Brekke, T.L. (1968) A model for the mechanics of jointed rock.
Journal of the Soil Mechanics and Foundations Division, ASCE, 94 (3), 637–659.
Grayeli, R. & Mortazavi, A. (2005) Implementation of a quadratic triangular mesh into the
DDA method. In: MacLaughlin, M. & Sitar, N. (eds.) The Seventh International Symposium
on Analysis of Discontinuous Deformation, Honolulu, Hawaii. pp. 91–101.
Grayeli, R. & Mortazavi, A., (2007) Elasto-plastic discontinuous deformation analysis using
Mohr-Coulomb model. In: Ju, Y., Fang, X. & Bian, H. (eds.) The Eighth International
Symposium on Analysis of Discontinuous Deformation, Beijing, China. pp. 195–200.
Gross, M.R. (1993) The origin and spacing of cross joints – Examples from the Monterey
Formation, Santa-Barbara Coastline, California. Journal of Structural Geology, 15,
737–751.
References 367
Guo, P. & Lin, S. (2007) Numerical simulation of mechanical characteristics of coarse granular
materials by discontinuous deformation analysis. In: Ju, Y., Fang, X. & Bian, H. (eds.) The
Eighth International Symposium on Analysis of Discontinuous Deformation, Beijing, China.
pp. 57–60.
Hatzor, Y.H. (1993) The block failure likelihood – A contribution to rock engineering in
blocky rock masses. International Journal of Rock Mechanics and Mining Sciences, 30,
1591–1597.
Hatzor, Y.H. (2003) Keyblock stability in seismically active rock slopes – Snake Path Cliff,
Masada. Journal of Geotechnical and Geoenvironmental Engineering, 129, 1069.
Hatzor, Y.H. & Benary, R. (1998) The stability of a laminated Voussoir beam: Back analysis
of a historic roof collapse using DDA. International Journal of Rock Mechanics and Mining
Sciences, 35, 165–181.
Hatzor, Y.H. & Feintuch, A. (2001) The validity of dynamic block displacement prediction using
DDA. International Journal of Rock Mechanics and Mining Sciences, 38, 599–606.
Hatzor, Y.H. & Goodman, R.E. (1997) Three-dimensional back-analysis of saturated rock slopes
in discontinuous rock – A case study. Geotechnique, 47, 817–839.
Hatzor, Y.H. & Palchik, V. (1997) The influence of grain size and porosity on crack initiation
stress and critical flaw length in dolomites. International Journal of Rock Mechanics and
Mining Sciences, 34, 805–816.
Hatzor, Y.H., Talesnick, M. & Tsesarsky, M. (2002) Continuous and discontinuous stabil-
ity analysis of the bell-shaped caverns at Bet Guvrin, Israel. International Journal of Rock
Mechanics and Mining Sciences, 39, 867–886.
Hatzor, Y.H., Arzi, A.A., Zaslavsky, Y. & Shapira, A. (2004) Dynamic stability analy-
sis of jointed rock slopes using the DDA method: King Herod’s Palace, Masada, Israel.
International Journal of Rock Mechanics and Mining Sciences, 41, 813–832.
Hatzor, Y.H., Wainshtein, I. & Mazor, D.B. (2010) Stability of shallow karstic caverns in
blocky rock masses. International Journal of Rock Mechanics and Mining Sciences, 47 (8),
1289–1303.
Hatzor, Y.H., Feng, X.T., Li, S., Yagoda-Biran, G., Jiang, Q. & Hu, L. (2015) Tunnel rein-
forcement in columnar jointed basalt: The role of rock mass anisotropy. Tunneling and
Underground Space Technology, 46, 1–11.
Hayashi, M., Yamada, S., Araya, M., Koyama, T., Fukuda, M. & Iwasaki, Y. (2012) Study for
reinforcement planning of masonry structure with cracks at Bayon main tower, Angkor. In:
Zhao, J., Ohnishi, Y., Zhao, G.F. & Sasaki, T. (eds.) Advances in Discontinuous Numerical
Methods and Applications in Geomechanics and Geoengineering. pp. 247–252.
He, B.G., Zelig, R., Hatzor, Y.H. & Feng, X.T. (2016) Rockburst generation in discontinuous
rock masses. Rock Mechanics and Rock Engineering, 49, 4103–4124.
He, L. (2010) Three Dimensional Numerical Manifold Method and Rock Engineering Applica-
tions. PhD Thesis. Singapore, Nanyang Technological University.
He, M.C., Nie, W., Zhao, Z.Y. & Guo, W. (2012) Experimental investigation of bedding plane
orientation on the rockburst behavior of sandstone. Rock Mechanics and Rock Engineering,
45, 311–326.
He, M.C., Sousa, L.R.E., Miranda, T. & Zhu, G.L. (2015) Rockburst laboratory tests database –
Application of data mining techniques. Engineering Geology, 185, 116–130.
Hoek, E. (2007) Practical Rock Engineering. Rocscience.
Hoek, E. & Bray, J.W. (1981) Rock Slope Engineering. 3rd edition. London, Institution of
Mining and Metallurgy.
Hoek, E. & Brown, E.T. (1997) Practical estimates of rock mass strength. International Journal
of Rock Mechanics and Mining Sciences, 34, 1165–1186.
Hoek, E. & Diederichs, M.S. (2006) Empirical estimation of rock mass modulus. International
Journal of Rock Mechanics and Mining Sciences, 43, 203–215.
368 References
Hoek, E., Kaiser, P.K. & Bawden, W.F. (1995) Support of Underground Excavations in Hard
Rock. Rotterdam, A.A. Balkema. p. 215.
Huang, T., Zhang, G.X. & Peng, X.C. (2010) The analysis of structure deformation using DDA
with third order displacement function. In: Ma, G. & Zhou, Y. (eds.) Analysis of Discontinu-
ous Deformation: New Developments and Applications. Singapore, Nanyang Technological
University. pp. 177–183.
Hudson, J.A. & Priest, S.D. (1979) Discontinuities and rock mass geometry. International
Journal of Rock Mechanics and Mining Sciences, 16, 339–362.
Hudson, J.A. & Priest, S.D. (1983) Discontinuity frequency in rock masses. International
Journal of Rock Mechanics and Mining Sciences, 20, 73–89.
Hudson, J.A., Backstrom, A., Rutqvist, J., Jing, L., Backers, T., Chijimatsu, M., Christiansson,
R., Feng, X.T., Kobayashi, A., Koyama, T., Lee, H.S., Neretnieks, I., Pan, P.Z., Rinne,
M. & Shen, B.T. (2009) Characterising and modelling the excavation damaged zone in
crystalline rock in the context of radioactive waste disposal. Environmental Geology, 57,
1275–1297.
Hughes, T.J.R. (1983) Analysis of transient algorithms with particular reference to stability
behavior. In: Belytschko, T. & Hughes, T.J.R. (eds.) Computational Methods for Transient
Analysis. Amsterdam, Elsevier Science.
ICADD1 (1995) In: Li, J.C., Wang, C.-Y. & Sheng, J. (eds.) The First International Conference
on Analysis of Discontinuous Deformation, Changli, Taiwan.
ICADD2 (1997) In: Ohnishi, Y. (ed.) The Second International Conference on Analysis of
Discontinuous Deformation, Kyoto, Japan.
ICADD3 (1999) In: Amadei, B. (ed.) The Third International Conference on Analysis of
Discontinuous Deformation, Vail, Colorado.
ICADD4 (2001) In: Bicanic, N. (ed.) The Fourth International Conference on Analysis of
Discontinuous Deformation, Glasgow, Scotland, UK.
ICADD5 (2002) In: Hatzor, Y.H. (ed.) The Fifth International Conference on Analysis of
Discontinuous Deformation. Wuhan, Balkema.
ICADD6 (2003) In: Lu, M. (ed.) The Sixth International Conference on Analysis of Discontin-
uous Deformation. Trondheim, Balkema Publishers.
ICADD7 (2005) In: MacLaughlin, M. & Sitar, N. (eds.) The Seventh International Symposium
on Analysis of Discontinuous Deformation, Honolulu, Hawaii.
ICADD8 (2007) In: Ju, Y., Fang, X. & Bian, H. (eds.) The Eighth International Symposium on
Analysis of Discontinuous Deformation, Beijing, China.
ICADD9 (2009) In: Ma, G. & Zhou, Y. (eds.) The Ninth International Symposium on Analysis
of Discontinuous Deformation. Research Publishing, Singapore.
ICADD10 (2011) In: Zhao, J., Ohnishi, Y., Zhao, G.F. & Sasaki, K. (eds.) The Tenth Inter-
national Symposium on Analysis of Discontinuous Deformation. Honolulu, HI, CRC Press,
Taylor &Francis Group.
ICADD11 (2013) In: Chen, G., Ohnishi, Y., Zheng, L. & Sasaki, T. (eds.) The 11th International
Symposium on Analysis of Discontinuous Deformation. Fukuoka, Taylor & Francis Group.
Ishikawa, T. & Ohnish, Y. (2001) A study on the time dependency of granular materials with
DDA. In: Bicanic, N. (ed.) The Fourth International Conference on Analysis of Discontinuous
Deformation, Glasgow, Scotland, UK. pp. 271–279.
Ishikawa, T., Ohnishi, Y. & Namura, A. (1997) DDA applied to deformation analysis of course
granular material (Ballast). In: Ohnishi, Y. (ed.) The Second International Conference on
Analysis of Discontinuous Deformation, Kyoto, Japan. pp. 252–262.
ITASCA (1993) FLAC – Fast Lagrangian Analysis of Continua. Minneapolis, MN, ITASCA
Consulting Group Inc.
Jaeger, J., Cook, N.G. & Zimmerman, R. (2007) Fundamentals of Rock Mechanics. 4th edition.
Wiley-Blackwell.
References 369
Jager, C. (1979) The vaiont rock slide. In: Rock Mechanics and Rock Engineering. Cambridge,
Cambridge University Press.
Jelenić, G. & Crisfield, M.A. (1996) Non-linear master-slave relationships for joints in 3D
beams with large rotations. Computer Methods in Applied Mechanics and Engineering, 135,
211–228.
Jiang, Q., Feng, X.T., Hatzor, Y.H., Hao, X.J. & Li, S.J. (2014) Mechanical anisotropy of colum-
nar jointed basalts: An example from the Baihetan hydropower station, China. Engineering
Geology, 175, 35–45.
Jiang, Q.A., Feng, X.T., Xiang, T.B. & Su, G.S. (2010) Rockburst characteristics and numerical
simulation based on a new energy index: A case study of a tunnel at 2,500 m depth. Bulletin
of Engineering Geology and the Environment, 69, 381–388.
Jiang, Q.H. & Yeung, M.R. (2004) A model of point-to-face contact for three-dimensional
discontinuous deformation analysis. Rock Mechanics and Rock Engineering, 37 (2),
95–116.
Jiao, Y., Huang, G., Zhao, Z., Zheng, F. & Wang, L. (2015) An improved three-dimensional
spherical DDA model for simulating rock failure. Science China-Technological Sciences,
58 (9), 1533–1541.
Jiao, Y.Y. & Zhang, X.L. (2012) DDARF-A simple solution for simulating rock fragmenta-
tion. In: Zhao, J., Ohnishi, Y., Zhao, G.F. & Sasaki, T. (eds.) Advances in Discontinuous
Numerical Methods and Applications in Geomechanics and Geoengineering. pp. 85–98.
Jiao, Y.Y., Zhang, X.L., Zhao, J. & Liu, Q.S. (2007) Viscous boundary of DDA for model-
ing stress wave propagation in jointed rock. International Journal of Rock Mechanics and
Mining Sciences, 44 (7), 1070–1076.
Jiao, Y.Y., Zhang, X.L. & Zhao, J. (2011) Two-dimensional DDA contact constitutive model
for simulating rock fragmentation. Journal of Engineering Mechanics, 138 (2), 199–209.
Jiao, Y.Y., Zhang, X.L., Zhang, H.Q. & Huang, G.H. (2014) A discontinuous numerical
model to simulate rock failure process. Geomechanics and Geoengineering: An International
Journal, 9, 133–141.
Jiao, Y.Y., Zhang, X.L., Zhang, H.Q., Li, H.B., Yang, S.Q. & Li, J.C. (2015a) A coupled
thermo-mechanical discontinuum model for simulating rock cracking induced by temperature
stresses. Computers and Geotechnics, 67, 142–149.
Jiao, Y.Y., Zhang, H.Q., Zhang, X.L., Li, H.B. & Jiang, Q.H. (2015b) A two-dimensional
coupled hydromechanical discontinuum model for simulating rock hydraulic fracturing.
International Journal for Numerical and Analytical Methods in Geomechanics, 39 (5),
457–481.
Jibson, R.W. (2007) Regression models for estimating coseismic landslide displacement.
Engineering Geology, 91, 209–218.
Jing, L., Ma, Y. & Fang, Z. (2001) Modeling of fluid flow and solid deformation for frac-
tured rocks with discontinuous deformation analysis (DDA) method. International Journal
of Rock Mechanics and Mining Sciences, 38, 343–355.
Jun, L., Xianjing, K. & Gao, L. (2004) Formulations of the three-dimensional discontinuous
deformation analysis method. Acta Mechanica Sinica, 20 (3), 270–282.
Kaiser, P.K. & Cai, M. (2012) Design of rock support system under rockburst condition. Journal
of Rock Mechanics and Geotechnical Engineering, 4, 215–227.
Kamai, R. (2006) Estimation of Historical Seismic Ground-Motions Using Back Analysis of
Structural Failures in Archaeological Sites. M Sc Thesis. Beer-Sheva, The Department of
Geological and Environmental Sciences, Ben-Gurion University of the Negev. 127 pp.
Kamai, R. & Hatzor, Y.H. (2005) Dynamic back analysis of structural failures in archeologi-
cal sites to obtain paleo-seismic parameters using DDA. In: MacLaughlin, M. & Sitar, N.
(eds.) The Seventh International Symposium on Analysis of Discontinuous Deformation,
Honolulu, Hawaii. pp. 121–136.
370 References
Kamai, R. & Hatzor, Y.H. (2008) Numerical analysis of block stone displacements in ancient
masonry structures: A new method to estimate historic ground motions. International Journal
for Numerical and Analytical Methods, 32, 1321–1340.
Ke, T.C. (1995) DDA combined with the artificial joint concept. In: Li, J.C., Wang, C.-Y. &
Sheng, J. (eds.) The First International Conference on Analysis of Discontinuous Deforma-
tion, Changli, Taiwan. pp. 124–139.
Ke, T.C. (1996) The issue of rigid-body rotation in DDA. In: Salami, M.R. & Banks, D. (eds.)
Proceedings of the First International Forum on Discontinuous Deformation Analysis (DDA)
and Simulations of Discontinuous Media, June 12–14. Berkeley, CA, TSI Press. pp. 318–325.
Ke, T.C. (1997) Application of DDA to simulate fracture propagation in solid. In: Ohnishi,
Y. (ed.) Proceedings of the Second International Conference on Analysis of Discontinuous
Deformation. Kyoto, Japan Institute of Systems Research. pp. 155–185.
Ke, T.C. & Bray, J. (1995) Modeling of particulate media using discontinuous deformation
analysis. Journal of Engineering Mechanics ASCE, 121, 1234–1243.
Kemeny, J. (2005) Time-dependent drift degradation due to the progressive failure of rock
bridges along discontinuities. International Journal of Rock Mechanics and Mining Sciences,
42, 35–46.
Keneti, A.R., Jafari, A. & Wu, J.H. (2008) A new algorithm to identify contact patterns between
convex blocks for three-dimensional discontinuous deformation analysis. Computers and
Geotechnics, 35 (5), 746–759.
Kieffer, S.D. (1998) Rock Slumping – A Compound Failure Mode of Jointed Hard Rock Slopes.
PhD Dissertation Thesis. Berkeley, Dept. of Civil and Environmental Engineering, U. C.
Berkeley.
Kilgore, B.D., Blanpied, M.L. & Dieterich, J.H. (1993) Velocity dependent friction of granite
over a wide range of conditions. Geophysical Research Letters, 20, 903–906.
Kirsch, G. (1898) Die theorie der elastizitat und die bedurfnisse der festigkeitslehre. Zeitschrift
des Vereines deutscher Ingenieure, 42, 797–807.
Kolsky, H. (1964) Stress waves in solids. Journal of Sound and Vibration, 1, 88–110.
Konyukhov, A. & Schweizerhof, K. (2013) Computational Contact Mechanics: Geometrically
Exact Theory for Arbitrary Shaped Bodies. New York, NY, Springer.
Koo, C.Y. & Chern, J.C. (1996) The development of DDA with third order displacement func-
tion. In: Salami, M.R. & Banks, D. (eds.) Proceedings of the First International Forum
on Discontinuous Deformation Analysis (DDA) and Simulation of Discontinuous Media.
Albuquerque, TSI Press. pp. 342–349.
Koo, C.Y. & Chern, J.C. (1997) Modelling of progressive fracture in jointed rock by
DDA method. In: Ohnishi, Y. (ed.) The Second International Conference on Analysis of
Discontinuous Deformation, Kyoto, Japan. pp. 186–200.
Koo, C.Y., Chern, J.C. & Chen, S. (1995) Development of second order displacement function
for DDA. In: Li, J.C., Wang, C.-Y. & Sheng, J. (eds.) The First International Conference on
Analysis of Discontinuous Deformation, Changli, Taiwan. pp. 91–108.
Koyama, T., Irie, K., Nagano, K., Nishiyama, S., Sakai, N. & Ohnishi, Y. (2012) DDA sim-
ulations for slope failure/collapse experiment caused by torrential rainfall. In: Zhao, J.,
Ohnishi, Y., Zhao, G.F. & Sasaki, T. (eds.) Advances in Discontinuous Numerical Methods
and Applications in Geomechanics and Geoengineering. pp. 119–125.
Koyama, T., Hashimoto, R., Kikumoto, M., Yamada, S., Araya, M., Iwasaki, Y. & Ohnishi, Y.
(2013) Application of coupled elasto-plastic NMM–DDA procedure for the stability analysis
of Prasat Suor Prat N1 Tower, Angkor, Cambodia. Geosystem Engineering, 16, 62–74.
Landers, J.A. & Taylor, R.L. (1986) An Lagrangian Formulation for the Finite Element Solution
of Contact Problems. Report No. CR 86.008 to Naval Civil Eng. Lab., Cal.
Lang, T.A. (1961) Theory and practice of rock bolting. Transactions of the American Institute
of Mining Engineering, 220, 333–348.
References 371
MacLaughlin, M.M. & Doolin, D.M. (2006) Review of validation of the discontinuous deforma-
tion analysis (DDA) method. International Journal for Numerical and Analytical Methods,
30, 271–305.
MacLaughlin, M.M. & Hayes, M.A., (2005) Validation of DDA block motions and failure
modes using laboratory models. In: MacLaughlin, M.M. & Sitar, N. (eds.) Proceedings of
ICADD-7, Honolulu, Hawaii. pp. 71–78.
MacLaughlin, M.M. & Sitar, N. (1996) Rigid body rotations in DDA. In: Proceed-
ings of the First International Forum on Discontinuous Deformation Analysis (DDA)
and Simulations of Discontinuous Media, June 12–14. Berkeley, CA, TSI Press;
pp. 620–635.
MacLaughlin, M.M. & Sitar, N. (1999) A gravity turn-on routine for DDA. In: Amadei, B.
(ed.) 3rd International Conference on Analysis of Discontinuous Deformation (ICADD-3)
ARMA, Vail, CO. pp. 65–74.
Makris, N. & Roussos, Y.S. (2000) Rocking response of rigid blocks under near-source ground
motions. Geotechnique, 50, 243–262.
Marone, C. (1998) The effect of loading rate on static friction and the rate of fault healing
during the earthquake cycle. Nature, 391, 69–72.
MATLAB, version 7. The MathWorks, Inc., Natick, MA.
Mauldon, M. (1998) Estimating mean fracture trace length and density from observations in
convex windows. Rock Mechanics and Rock Engineering, 31, 201–216.
Mauldon, M. & Mauldon, J.G. (1997) Fracture sampling on a cylinder: From scanlines to
boreholes and tunnels. Rock Mechanics and Rock Engineering, 30, 129–144.
Mauldon, M., Dunne, W.M. & Rohrbaugh, M.B. (2001) Circular scanlines and circular win-
dows: New tools for characterizing the geometry of fracture traces. Journal of Structural
Geology, 23, 247–258.
Mencl, V. (1966) Mechanics of landslides with non-circular slip surfaces with special reference
to vaiont slide. Geotechnique, 16, 329–337.
Miki, S., Sasaki, T., Koyama, T., Nishiyama, S. & Ohnishi, Y. (2010) Development of cou-
pled discontinuous deformation analysis and numerical manifold method (NMM-DDA).
International Journal of Computational Methods, 7 (1), 131–150.
Mikola, R.G. & Sitar, N. (2013) Next generation discontinuous rock mass models: 3-D and
rock-fluid interaction. In: Chen, G., Ohnishi, Y., Zheng, L. & Sasaki, T. (eds.) The 11th
International Symposium on Analysis of Discontinuous Deformation. Fukuoka, Taylor &
Francis Group. pp. 81–90.
Milev, A.M., Spottiswoode, S.M., Rorke, A.J. & Finnie, G.J. (2001) Seismic monitoring of
a simulated rockburst on a wall of an underground tunnel. Journal of the South African
Institute of Mining and Metallurgy, 101 (5), 253–260.
Minkowski, H. & Geometrie der Zahlen. Leipzig and Berlin, B.G. Teubner, 1910, JFM
41.0239.03, MR 0249269.
Moosavi, M. & Grayeli, R. (2006) A model for cable bolt-rock mass interaction: Integration
with discontinuous deformation analysis (DDA) algorithm. International Journal of Rock
Mechanics and Mining Sciences, 43, 661–670.
Moosavi, M., Jafari, A. & Beyabanaki, S.A.R. (December 2005) Dynamic three-dimensional
discontinuous deformation analysis (3-D DDA) validation using analytical solution.
In: Proceedings of the Seventh International Conference on Analysis of Discontinuous
Deformation (ICADD-7), Hawaii, USA. pp. 37–48.
Morgan, W.E. & Aral, M.M. (2015) An implicitly coupled hydro-geomechanical model for
hydraulic fracture simulation with the discontinuous deformation analysis. International
Journal of Rock Mechanics and Mining Sciences, 73, 82–94.
Munjiza, A. & Andrews, K.R.F. (1998) NBS contact detection algorithm for bodies of similar
size. International Journal for Numerical Methods in Engineering, 43 (1), 131–149.
References 373
Munjiza, A. & Latham, J.P. (2002) Grand challenge of discontinuous deformation analysis.
In: Hatzor, Y.H. (ed.) The Fifth International Conference on Analysis of Discontinuous
Deformation. Wuhan, Balkema. pp. 69–74.
Narr, W. & Suppe, J. (1991) Joint spacing in sedimentary-rocks. Journal of Structural Geology,
13, 1037–1048.
Newmark, N. (1965) Effects of earthquakes on dams and embankments. Geotechnique, 15,
139–160.
Nezami, E.G., Hashash, Y.M.A., Zhao, D. & Ghaboussi, J. (2006) Shortest link method for con-
tact detection in discrete element method. International Journal for Numerical and Analytical
Methods in Geomechanics, 30 (8), 783–801.
Nie, W., Zhao, Z.Y., Ning, Y.J. & Sun, J.P. (2014) Development of rock bolt elements in two-
dimensional discontinuous deformation analysis. Rock Mechanics and Rock Engineering,
47, 2157–2170.
Ning, Y.J. & Zhao, Z. (2012a) A detailed investigation of block dynamic sliding by the discon-
tinuous deformation analysis. International Journal for Numerical and Analytical Methods
in Geomechanics, 37 (15), 2373–2393.
Ning, Y.J. & Zhao, Z.Y. (2012b) Nonreflecting boundaries for the discontinuous deformation
analysis. In: Zhao, J., Ohnishi, Y., Zhao, G.F. & Sasaki, T. (eds.) Advances in Dis-
continuous Numerical Methods and Applications in Geomechanics and Geoengineering.
pp. 147–153.
Ning, Y.J., Yang, J. & Chen, P.-W. (2007) Application study of DDA method in blasting numer-
ical simulation. In: Ju, Y., Fang, X. & Bian, H. (eds.) The Eighth International Symposium
on Analysis of Discontinuous Deformation, Beijing, China. pp. 117–122.
Ning, Y.J., Yang, J., Ma, G.W. & Chen, P.W. (2010) Contact algorithm modification of DDA
and its verification. In: Ma, G. & Zhou, Y. (eds.) Analysis of Discontinuous Deforma-
tion: New Developments and Applications. Singapore, Nanyang Technological University.
pp. 73–81.
Obert, L. & Duvall, W.I. (1967) Rock Mechanics and the Design of Structures in Rock.
New York, NY, John Wiley & Sons.
Oh, Y.N., Jeng, D.S., Chen, S. & Chien, L.K. (2002) A parametric study using discontinu-
ous deformation analysis to model wave-induced seabed response. In: Hatzor, Y.H. (ed.) The
Fifth International Conference on Analysis of Discontinuous Deformation. Wuhan, Balkema.
pp. 113–120.
Ohnishi, Y. & Miki, S. (1996) Development of circular and elliptical disc elements for DDA. In:
Salami, M.R. & Banks, D. (eds.) First International Forum on Discontinuous Deformation
Analysis and Simulations of Discontinuous Media. Berkeley, CA, TSI Press. pp. 44–51.
Ohnishi, Y., Chen, G. & Miki, S. (1995) Recent development of DDA in rock mechanics practice.
In: Li, J.C., Wang, C.-Y. & Sheng, J. (eds.) The First International Conference on Analysis
of Discontinuous Deformation, Changli, Taiwan. pp. 26–47.
Ohnishi, Y., Nishiyama, S., Akao, S., Yang, M. & Miki, S. (2005) DDA for elastic elliptic ele-
ment. In: MacLaughlin, M. & Sitar, N. (eds.) Proceedings of ICADD-7, Honolulu, Hawaii.
pp. 103–112.
Ohnishi, Y., Koyama, T., Sasaki, T., Hagiwara, I., Miki, S. & Shimauchi, T. (2012) Application
of DDA and NMM to practical problems in recent new insight. In: Zhao, J., Ohnishi,
Y., Zhao, G.F. & Sasaki, T. (eds.) Advances in Discontinuous Numerical Methods and
Applications in Geomechanics and Geoengineering. pp. 31–42.
Ortlepp, W.D. & Stacey, T.R. (1994) Rockburst mechanisms in tunnels and shafts. Tunnelling
and Underground Space Technology, 9, 59–65.
Osada, M. & Tanityama, H. (2005) Preliminary consideration for analyzing ground deforma-
tion due to fault movement. In: MacLaughlin, M. & Sitar, N. (eds.) Proceedings of ICADD-7,
Honolulu, Hawaii. pp. 113–120.
374 References
O’Sullivan, C. & Bray, J.D. (2001) A comparative evaluation of two approaches to discrete
element modeling of particulate media. In: Bicanic, N. (ed.) The Fourth International
Conference on Analysis of Discontinuous Deformation, Glasgow, Scotland, UK. pp. 97–110.
Parker, R.N., Densmore, A.L., Rosser, N.J., de Michele, M., Li, Y., Huang, R.Q.,
Whadcoat, S. & Petley, D.N. (2011) Mass wasting triggered by the 2008 Wenchuan
earthquake is greater than orogenic growth. Nature Geoscience, 4, 449–452.
Perras, M.A. & Diederichs, M.S. (2016) Predicting excavation damage zone depths in brittle
rocks. Journal of Rock Mechanics and Geotechnical Engineering, 8, 60–74.
Pietrzak, G. & Curnier, A. (1999) Large deformation frictional contact mechanics: Continuum
formulation and augmented Lagrangian treatment. Computer Methods in Applied Mechanics
and Engineering, 177, 351–381.
Poetsch, M. (2011) The Analysis of Rotational and Sliding Modes of Failure for Slopes, Foun-
dations, and Underground Structures in Blocky Rock. Institute fur Felsmechanik und Tunnel
bau Tech, Univ. GRAZ, Heft.
Priest, S.D. (1985) Hemispherical Projectiom Methods in Rock Mechanics. George Allen &
Unwin Publishers. Note: this publisher has been purchased by Wiley and they gave us the
permision for Figure 4.20.
Priest, S.D. (1993) Discontinuity Analysis for Rock Engineering. London, Chapman & Hall.
Priest, S.D. & Hudson, J.A. (1976) Discontinuity spacings in rock. International Journal of
Rock Mechanics and Mining Sciences, 13, 135–148.
Priest, S.D. & Hudson, J.A. (1981) Estimation of discontinuity spacing and trace length
using scanline surveys. International Journal of Rock Mechanics and Mining Sciences, 18,
183–197.
Reches, Z. (1987) Determination of the tectonic stress tensor from slip along faults that obey
the coulomb yield condition. Tectonics, 6, 849–861.
Rohrbaugh, M.B., Dunne, W.M. & Mauldon, M. (2002) Estimating fracture trace inten-
sity, density, and mean length using circular scan lines and windows. AAPG Bulletin, 86,
2089–2104.
Rouainia, M., Pearce, C. & Bicanic, N. (2001) Hydro-DDA modeling of fractured mudrock
seals. In: Icanic, N. (ed.) The Fourth International Conference on Analysis of Discontinuous
Deformation, Glasgow, Scotland, UK. pp. 413–423.
Ruf, J.C., Rust, K.A. & Engelder, T. (1998) Investigating the effect of mechanical discontinuities
on joint spacing. Tectonophysics, 295, 245–257.
Ruina, A. (1983) Slip instability and state variable friction laws. Journal of Geophysical
Research, 88, 10359–10370.
Rutqvist, J. & Stephansson, O. (2003) The role of hydro-mechanical coupling in fractured rock
engineering. Hydrogeology Journal, 11, 7–40.
Sagaseta, C. (1986) On the modes of instability of a rigid block on an inclined plane. Rock
Mechanics and Rock Engineering, 19, 261–266.
Sagy, A. & Brodsky, E.E. (2009) Geometric and rheological asperities in an exposed fault zone.
Journal of Geophysical Research-Solid Earth, 114, B02301.
Salami, M.R. & Banks, D. (1996) In: Salami, M.R. & Banks, D. (eds.) First International Forum
on Discontinuous Deformation Analysis and Simulations of Discontinuous Media. Berkeley,
CA, TSI Press.
Sammis, C.G. & Ashby, M.F. (1986) The failure of brittle porous solids under compressive stress
states. Acta Metallurgica, 34, 511–526.
Sanchez-Sesma, F.J. & Campillo, M. (1991) Diffraction of P, SV and Rayleigh waves by topo-
graphic features: A boundary integral formulation. Bulletin of the Seismological Society of
America, 81, 2234–2253.
Sasaki, T., Hagiwara, I., Sasaki, K., Horikawa, S., Ohnishi, Y. & Nishiyama, S.
(2005) Earthquake response analysis of a rock-fall by discontinuous deformation analysis.
References 375
Shi, G.H. (2013b) Basic theory of two dimensional and three dimensional contacts. In: Chen,
G.Q., Ohnishi, Y., Zheng, L. & Sasaki, T. (eds.) ICADD-11: The 11th International Confer-
ence on Analysis of Discontinuous Deformation. London and Fukuoka, Taylor and Francis.
pp. 3–14.
Shi, G.H. & Goodman, R.E. (1984) Discontinuous deformation analysis. In: Proc. 25th U. S.
Symposium on Rock Mechanics. pp. 269–277.
Shi, G.H. & Goodman, R.E. (1985) Two dimensional discontinuous deformation analysis.
International Journal for Numerical and Analytical Methods in Geomechanics, 9, 541–556.
Shinji, M., Ohno, H., Otsuka, Y. & Ma, G. (1997) The viscosity coefficient of the rock-
fall simulation. In: Ohnishi, Y. (ed.) The Second International Conference on Analysis of
Discontinuous Deformation, Kyoto, Japan. pp. 201–210.
Shyu, K. (1993) Nodal-Based Discontinuous Deformation Analysis. PhD Thesis. Berkeley, CA,
University of California.
Shyu, K., Wang, X. & Chang, C.T. (1999) Dynamic behaviors in discontinuous elastic
media using DDA. In: Proceedings of the Third International Conference on Analysis of
Discontinuous Deformation from Theory to Practice. pp. 243–252.
Sitar, N., MacLaughlin, M.M. & Doolin, D.M. (2005) Influence of kinematics on landslide
mobility and failure mode. Journal of Geotechnical and Geoenvironmental Engineering,
131, 716–728.
Skempton, A.W. (1966) Bedding-plane slip residual strength and vaiont landslide. Geotechnique,
16, 82–84.
Smith, I.M. & Griffiths, D.V. (1998) Programming the Finite Element Method. Chichester, John
Wiley & Son Press.
Sofianos, A.I. (1999) Discussion of the paper by M.S. Diederichs and P.K. Kaiser “Stabil-
ity of large excavations in laminated hard rock masses: The Voussoir analogue revisited’’.
International Journal of Rock Mechanics and Mining Sciences, 36, 97–117, 991–993.
Sofianos, I. (1996) Analysis and design of an underground hard rock Voussoir beam roof.
International Journal of Rock Mechanics and Mining Sciences, 33, 153–166.
Solberg, P. & Byerlee, J.D. (1984) A note on the rate sensitivity of frictional sliding of Westerly
granite. Journal of Geophysical Research, 89, 4203–4205.
Stead, D., Eberhardt, E. & Coggan, J.S. (2006) Developments in the characterization of com-
plex rock slope deformation and failure using numerical modelling techniques. Engineering
Geology, 83, 217–235.
Stefanou, I., Sulem, J. & Vardoulakis, I. (2006) Continuum modeling of masonry structures
under static and dynamic loading. In: Kourkoulis, S.K. (ed.) Fracture and Failure of Nat-
ural Building Stones – Applications in the Restoration of Ancient Monuments. Dordrecht,
Springer. pp. 123–136.
Tal, Y., Hatzor, Y.H. & Feng, X.T. (2014) An improved numerical manifold method for simu-
lation of sequential excavation in fractured rocks. International Journal of Rock Mechanics
and Mining Sciences, 64, 116–128.
Tang, C., Zhu, J., Qi, X. & Ding, J. (2011) Landslides induced by the Wenchuan earthquake and
the subsequent strong rainfall event: A case study in the Beichuan area of China. Engineering
Geology, 122, 22–33.
Taylor, R.L. (2004) Finite element solution of contact problems: From 1974 to 2004. In:
Advances in Computational Mechanics, Celebrating the 60th Birthday of Tom Hughes, 7
April 2004.
Terzaghi, K. (1946) Load on tunnel supports. In: Proctor, R.V. & White, T.L. (eds.) Rock
Tunneling with Steel Supports. Ohio, Commercial Shearing Inc. pp. 47–86.
Terzaghi, R.D. (1965) Sources of error in joint surveys. Geotechnique, 15, 287–304.
Thomas, P.A., Bray, J.D. & Ke, T.C. (1996) Discontinuous deformation analysis for soil mechan-
ics. In: Salami, M.R. & Banks, D. (eds.) First International Forum on Discontinuous
References 377
Deformation Analysis and Simulations of Discontinuous Media. Berkeley, CA, TSI Press.
pp. 454–461.
Tian, J.Q., Nishiyama, S., Koyama, T. & Ohnishi, Y. (2012) Masonry retaining wall under
static load using discontinuous deformation analysis. In: Zhao, J., Ohnishi, Y., Zhao, G.F. &
Sasaki, T. (eds.) Advances in Discontinuous Numerical Methods and Applications in
Geomechanics and Geoengineering. pp. 169–174.
Trollope, D.H. (1980) The vaiont slope failure. Rock Mechanics, 13, 71–88.
Tsai, J.S. & Wang, W.C. (1995) Dynamic response of sliding structure subjected to seismic exci-
tation. In: Li, J.C., Wang, C.-Y. & Sheng, J. (eds.) The First International Conference on
Analysis of Discontinuous Deformation, Changli, Taiwan. pp. 420–432.
Tsesarsky, M. (2005) Stability of Underground Openings in Stratified and Jointed Rock. PhD
Thesis. Beer Sheva, Dept. of Geological and Environmental Sciences, Ben-Gurion University
of the Negev. 187 pp.
Tsesarsky, M. & Hatzor, Y.H. (2006) Tunnel roof deflection in blocky rock masses as a function
of joint spacing and friction – A parametric study using discontinuous deformation analysis
(DDA). Tunnelling and Underground Space Technology, 21, 29–45.
Tsesarsky, M. & Hatzor, Y.H. (2009) Kinematics of overhanging slopes in discontinuous
rock. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 135, 1122–
1129.
Veveakis, E., Vardoulakis, I. & Di Toro, G. (2007) Thermoporomechanics of creeping land-
slides: The 1963 Vaiont slide, northern Italy. Journal of Geophysical Research-Earth,
112, F3.
Wang, C.Y., Chuang, C.C. & Sheng, J. (1996) Time integration theories for the DDA method
with finite element meshes. In: Salami, M.R. & Banks, D. (eds.) 1st International Forum
on Discontinuous Deformation Analysis (DDA) and Simulation of Discontinuous Media.
Berkeley, CA, TSI Press. pp. 263–287.
Wang, C.Y., Sheng, J. & Chen, M.H. (1995) Dynamic contact analysis scheme applied in the
DDA method. In: Li, J.C., Wang, C.-Y. & Sheng, J. (eds.) The First International Conference
on Analysis of Discontinuous Deformation, Changli, Taiwan. pp. 433–459.
Wang, L.-Z., Jiang, H.-Y., Yang, Z.-X., Xu, Y.-C. & Zhu, X.-B. (2013) Development of
discontinuous deformation analysis with displacement-dependent interface shear strength.
Computers and Geotechnics, 47, 91–101.
Wang, W., Zhang, H., Zheng, L., Zhang, Y.B., Wu, Y.Q. & Liu, S.G. (2017) A new approach
for modeling landslide movement over 3D topography using 3D discontinuous deformation
analysis. Computers and Geotechnics, 81, 87–97.
Wang, X.B., Ding, X.L., Lu, B. & Wu, A. (2007) DDA with higher order polynomial displace-
ment functions for large elastic deformation problems. In: Ju, Y., Fang, X. & Bian, H. (eds.)
The Eighth International Symposium on Analysis of Discontinuous Deformation, Beijing,
China. pp. 89–94.
Wittke, W. (1965) Methods to analyze the stability of rock slopes with and without additional
loading (in German). Rock Mechanics and Engineering Geology, Suppl. II, 52.
Wu, A., Ding, X., Lu, B. & Zhang, Q. (2007) Validation for rock block stability kinemat-
ics and its application to rock slope stability evaluation using discontinuous deformation
analysis. In: Ju, Y., Fang, X. & Bian, H. (eds.) Proceedings of ICADD-8, Beijing, China.
pp. 27–32.
Wu, A., Yang, Q., Ma, G., Ju, B. & Li, X.S. (2009) Study on the formation mechanism of
Tanjiashan Landslide triggered by Wenchuan earthquake using DDA simulation. In: Ma,
G. & Zhou, Y. (eds.) The Ninth International Symposium on Analysis of Discontinuous
Deformation. Singapore, Research Publishing.
Wu, A.Q., Zhang, Y. & Lin, S.Z. (2012) Complete and high order polynomial displacement
approximation and its application to elastic mechanics analysis based on DDA. In: Zhao, J.,
378 References
Ohnishi, Y., Zhao, G.F. & Sasaki, T. (eds.) Advances in Discontinuous Numerical Methods
and Applications in Geomechanics and Geoengineering. pp. 43–53.
Wu, J.H. (2007) Applying discontinuous deformation analysis to assess the constrained area
of the unstable Chiu-fen-erh-shan landslide slope. International Journal for Numerical and
Analytical Methods, 31, 649–666.
Wu, J.H. (2008) New edge-to-edge contact calculating algorithm in three-dimensional discrete
numerical analysis. Advances in Engineering Software, 39 (1), 15–24.
Wu, J.H. (2010) Seismic landslide simulations in discontinuous deformation analysis. Computers
and Geotechnics, 37, 594–601.
Wu, J.H., Juang, C.H. & Lin, H.M. (2005) Vertex-to-face contact searching algorithm for three-
dimensional frictionless contact problems. International Journal for Numerical Methods in
Engineering, 63 (6), 876–897.
Wu, J.H. & Lin, H.M. (2013) Improvement of open-close iteration in DDA. In: The Eleventh
International Conference on Analysis of Discontinuous Deformation, Fukuoka, Japan.
pp. 185–191.
Wu, J.H., Ohnishi, Y. & Nishiyama, S. (2005a) A development of the discontinuous defor-
mation analysis for rock fall analysis. International Journal for Numerical and Analytical
Methods, 29, 971–988.
Wu, J.H., Ohnishi, Y., Shi, G.H. & Nishiyama, S. (2005b) Theory of three-dimensional discon-
tinuous deformation analysis and its application to a slope toppling at Amatoribashi, Japan.
International Journal of Geomechanics, 5 (3), 179–195.
Wu, W., Zhu, H., Zhuang, X., Ma, G. & Cai, Y. (2014) A multi-shell cover algorithm for
contact detection in the three dimensional discontinuous deformation analysis. Theoretical
and Applied Fracture Mechanics, 72, 136–149.
Yagoda-Biran, G. (2013) Seismic Hazard Analysis Using the Numerical DDA Method. PhD
Thesis. Beer Sheva, Dept. of Geological and Environmental Sciences, Ben-Gurion University
of the Negev. 153 pp.
Yagoda-Biran, G. & Hatzor, Y.H. (2010) Constraining paleo PGA values by numerical
analysis of overturned columns. Earthquake Engineering & Structural Dynamics, 39,
463–472.
Yagoda Biran, G. & Hatzor, Y.H. (2013) A new failure mode chart for toppling and sliding
with consideration of earthquake inertia force. International Journal of Rock Mechanics and
Mining Sciences, 64, 122–131.
Yagoda Biran, G. & Hatzor, Y.H. (2016) Benchmarking the numerical discontinuous deforma-
tion analysis method. Computers and Geotechnics, 71, 30–46.
Yang, J. & Ning, Y.J. (2005) Numerical simulation of bench blasting by 2D-DDA method.
In: MacLaughlin, M. & Sitar, N. (eds.) Proceedings of ICADD-7, Honolulu, Hawaii.
pp. 159–165.
Yang, M., Fukawa, T., Ohnishi, Y., Nishiyama, S., Miki, S., Hirakawa, Y. & Mori, S. (2004)
The application of three-dimensional DDA with a spherical rigid block to rockfall simulation.
International Journal of Rock Mechanics and Mining Sciences, 41, 476–476.
Yeung, M.C.R. & Goodman, R.E. (1992) Use of Shi’s Discontinuous Deformation Analysis on
Rock Slope Problems. In: Seed, R.B. & Boulanger, R.W. (eds.) Stability and Performance of
Slopes and Embankments-II, Vols 1 and 2.
Yeung, M.R. (1991) Application of Shi’s Discontinuous Deformation Analysis to the Study of
Rock Behaviour. PhD Thesis. Berkeley, CA, Civil Engineering, University of California.
Yeung, M.R. (1993) Analysis of a mine roof using the DDA method. International Journal of
Rock Mechanics and Mining Sciences, 30, 1411–1417.
Yeung, M.R., Jiang, Q.H. & Sun, N. (2003) Validation of block theory and three-dimensional
discontinuous deformation analysis as wedge stability analysis methods. International
Journal of Rock Mechanics and Mining Sciences, 40 (2), 265–275.
References 379
Yeung, M.R., Jiang, Q.H. & Sun, N. (2007) A model of edge-to-edge contact for three-
dimensional discontinuous deformation analysis. Computers and Geotechnics, 34 (3),
175–186.
Yeung, M.R., Sun, N., Jiang, Q.H. & Blair, S.C. (2004) Analysis of large block test data
using three-dimensional discontinuous deformation analysis. International Journal of Rock
Mechanics and Mining Sciences, 41, 521–526.
Zaslavsky, Y. & Shapira, A. (2000) Experimental study of topographic amplification using the
Israel seismic network. Journal of Earthquake Engineering, 4, 43–65.
Zaslavsky, Y., Shapira, A. & Arzi, A.A. (2000) Amplification effects from earthquakes and
ambient noise in the Dead Sea rift (Israel). Soil Dynamics and Earthquake Engineering, 20,
187–207.
Zaslavsky, Y., Shapira, A. & Leonoy, J. (1998) Topography effects and seismic hazard assessment
at Mt. Massada – Dead Sea. The Geophysical Institute of Israel. Report no. 522/62/98.
Zelig, R., Hatzor, Y.H. & Feng, X.T. (2015) Rock burst simulations with 2D-DDA. In: Pro-
ceedings of the 49th U.S. Rock Mechanics Symposium, June 28–July 1 2015. San Francisco.
Zhang, C., Feng, X.-T., Zhou, H., Qiu, S. & Wu, W. (2013) Rockmass damage develop-
ment following two extremely intense rockbursts in deep tunnels at Jinping II hydropower
station, southwestern China. Bulletin of Engineering Geology and the Environment, 72,
237–247.
Zhang, H., Liu, S.G., Chen, G.Q., Zheng, L., Zhang, Y.B., Wu, Y.Q., Jing, P.D., Wang, W., Han,
Z., Zhong, G.H. & Lou, S. (2016a) Extension of three-dimensional discontinuous deforma-
tion analysis to frictional-cohesive materials. International Journal of Rock Mechanics and
Mining Sciences, 86, 65–79.
Zhang, H., Liu, S.G., Han, Z., Zheng, L., Zhang, Y.B., Wu, Y.Q., Li, Y.G. & Wang, W. (2016b)
A new algorithm to identify contact types between arbitrarily shaped polyhedral blocks
for three-dimensional discontinuous deformation analysis. Computers and Geotechnics, 80,
1–15.
Zhang, L. & Einstein, H.H. (1998) Estimating the mean trace length of rock discontinuities.
Rock Mechanics and Rock Engineering, 31, 217–235.
Zhang, L., Einstein, H.H. & Dershowitz, W.S. (2002) Stereological relationship between trace
length and size distribution of elliptical discontinuities. Geotechnique, 52, 419–433.
Zhang, Y., Wang, J., Xu, Q., Chen, G., Zhao, J.X., Zhenge, L., Han, Z. & Yu, P. (2015)
DDA validation of the mobility of earthquake-induced landslides. Engineering Geology,
194, 38–51.
Zhao, G.F., Zhao, X.B. & Zhu, J.B. (2014) Application of the numerical manifold method
for stress wave propagation across rock masses. International Journal for Numerical and
Analytical Methods in Geomechanics, 38, 18.
Zhao, Z., Zhang, Y. & Wei, X. (2010) Discontinuous deformation analysis for parallel hole cut
blasting in rock mass. In: Ma, G. & Zhou, Y. (eds.) Analysis of Discontinuous Deforma-
tion: New Developments and Applications. Singapore, Nanyang Technological University.
pp. 169–176.
Zhao, Z.Y., Gu, J. & Bao, H. (2007) Understanding fracture patterns of rock mass due to blast
load – A DDA approach. In: Ju, Y., Fang, X. & Bian, H. (eds.) The Eighth International
Symposium on Analysis of Discontinuous Deformation, Beijing, China. pp. 150–156.
Zhao, Z.Y., Zhang, Y. & Bao, H.R. (2011) Tunnel blasting simulations by the dis-
continuous deformation analysis. International Journal of Computational Methods, 8,
277–292.
Zhao, Z.Y., An, X.M. & Zhou, Y.X. (2013) DDA/NMM developments and applications in
Nanyang Technological University, Singapore. In: Chen, G., Ohnishi, Y., Zheng, L. & Sasaki,
T. (eds.) The 11th International Symposium on Analysis of Discontinuous Deformation.
Fukuoka, Taylor & Francis Group. pp. 67–80.
380 References
Zheng, H. & Li, X.K. (2015) Mixed linear complementarity formulation of discontinuous
deformation analysis. International Journal of Rock Mechanics and Mining Sciences, 75 (1),
23–32.
Zhong, Z.H. & Nilsson, L. (1989) A contact searching algorithm for general contact problems.
Computers & Structures, 33 (1), 197–209.
Zhu, H., Wang, S. & Cai, Y. (2007) A discontinuous sub-block meso-damage evolution model
for rock mass. In: Ju, Y., Fang, X. & Bian, H. (eds.) The Eighth International Symposium
on Analysis of Discontinuous Deformation, Beijing, China. pp. 162–167.
Zhu, H., Wu, W., Chen, J., Ma, G.W, Liu, X. & Zhuang, X. (2016) Integration of three
dimensional discontinuous deformation analysis (DDA) with binocular photogrammetry
for stability analysis of tunnels in blocky rockmass. Tunnelling and Underground Space
Technology, 51, 30–40.
Subject index
ISSN: 2326-6872
3. Rock Dynamics
Author: Ömer Aydan
2017
ISBN: 978-1-138-03228-6 (Hbk)