Aashto Geocell Önemli

Use of Geocell in Pavement Design: Final Report
Technical Report 0-6833-1

Cooperative Research Program
CENTER FOR TRANSPORTATION INFRASTRUCTURE SYSTEMS

THE UNIVERSITY OF TEXAS AT EL PASO
EL PASO, TX 79968
IUUQTXXXVUFQFEVFOHJOFFSJOHDUJT
in cooperation with the

Federal Highway Administration and the
Texas Department of Transportation
Technical Report Documentation Page
1. REPORT NO. 2. GOVERNMENT ACCESSION NO. 3. RECIPIENTS CATALOG NO.
FHWA/TX-21/0-6833-1
4. TITLE AND SUBTITLE 5. REPORT DATE
Use of Geocell in Pavement Design: Final Report November 11, 2018
Published: November 2021
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S) 8. PERFORMING ORGANIZATION REPORT No.

Sundeep Inti, Ph.D. https://orcid.org/0000-0002-8631-446X 0-6833-1
Megha Sharma, Ph.D.
Cesar Tirado, Ph.D.
Vivek Tandon, Ph.D. https://orcid.org/0000-0003-2469-2956
9. PERFORMING ORGANIZATION NAME AND ADDRESS 10. WORK UNIT NO.
Center for Transportation Infrastructure Systems
The University of Texas at El Paso 11. CONTRACT OR GRANT NO.
500 W. University Ave. 0-6833

El Paso, Texas 79968
12. SPONSORING AGENCY NAME AND ADDRESS 13. TYPE OF REPORT AND PERIOD COVERED
Texas Department of Transportation Final Report
Research and Technology Implementation Division September 2014 – August 2017
125 E. 11th Street 14. SPONSORING AGENCY CODE
Austin, TX 78701
15. SUPPLEMENTARY NOTES
Project performed in cooperation with the Texas Department of Transportation and the Federal Highway
Administration.
16. ABSTRACT
This study focused on identifying mechanisms responsible for improved bearing capacity and benefits derived
from geocell. The study performed Finite Element Analyses (FEA) and verified the results by performing
laboratory tests. In addition, the study developed a design system and performed Life Cycle Cost Analysis to
identify the benefits of geocell reinforcement and provided construction guidelines. Based on the FEA and
laboratory evaluation, the study identified that the geocell is beneficial when construction needs to be
performed with a lower/marginal base and subgrade material is available.
17. KEYWORDS 18. DISTRIBUTION STATEMENT

Geocell, Base, Subgrade, Finite Element, Strain, No restrictions. This document is available to the
Stress public through the National Technical Information
Service, Springfield, Virginia, 22161; www.ntis.gov
19. SECURITY CLASSIF. (OF 20. SECURITY CLASSIF. (OF 21. NO. OF PAGES 22. PRICE
THIS REPORT) THIS PAGE)
238
Unclassified Unclassified
Use of Geocell in Pavement Design
Final Report
Conducted for
Texas Department of Transportation
By
Sundeep Inti, Ph.D.
Megha Sharma, Ph.D.
Cesar Tirado, Ph.D.
Vivek Tandon, Ph.D
Research Report 0-6833-1
November 2018
Published November 2021
Department of Civil Engineering

The University of Texas at El Paso
El Paso, TX 79968
(915) 747-6924
Authors’ Disclaimer:
The contents of this report reflect the view of the authors, who are responsible for the
facts and the accuracy of the data presented herein. The contents do not necessarily
reflect the official views or policies of the Texas Department of Transportation (TxDOT)
or the Federal Highway Administration (FHWA). This report does not constitute a
standard, specification, or regulation.
NOT INTENDED FOR CONSTRUCTION, BIDDING, OR PERMIT

PURPOSES
Sundeep Inti., Ph.D.

Megha Sharma, Ph.D.
Cesar Tirado, Ph.D.
Vivek Tandon, Ph.D.
ACKNOWLEDGMENTS
The successful progress of this project could not have happened without the valuable assistance from
several TxDOT personnel. The authors acknowledge Mr. Mark McDaniel, Mr. Wade Blackmon, and
Mr. Boon Thian for their valuable guidance, input, and support. The authors would also like to
acknowledge Mr. Brett Haggerty, Mr. Aldo Madrid, and Mr. Richard Williammee for facilitating
collaboration with TxDOT Districts. In addition, the authors would like to acknowledge Mr. Jose Luis
Arias, Mr. Armando Esquivel, Mr. Angel Rodarte, Mr. Rafael Silva, and Ms. Sofia Martin for their
assistance in performing tests. In the end, the authors would like to thank Ms. Sonya Badgley of RTI
for her support.
EXECUTIVE SUMMARY
The geotechnical construction has evolved over the years with the application of geosynthetics,
starting with the nonwoven to the more complex geo-composites. Most of these systems are two‐
dimensional. Cellular confinement systems add the third dimension to geosynthetics, which opens
more applications, ranging from providing strength to geosystems to protect against erosion.
Cellular Confinement Systems, popularly known as “Geocells,” are durable, lightweight, three-
dimensional fabricated systems that are expandable on‐site to form a honeycomb‐like structure.
Geocells are filled with infill material and compacted. The composite forms a rigid to the semi‐
rigid structure. The depth of the geocells and the size of each cellular unit can vary as per design
requirements. In addition, the surface of the geocell can be textured to increase soil‐geocell wall
friction. Geocells have been intermittently used in roads; they provide lateral reinforcement, which
increases the bearing capacity of the subbase and subgrade pavement layers.
This study focused on identifying mechanisms responsible for improved bearing capacity and
benefits derived from geocells. The study performed Finite Element Analyses (FEA) and verified
the results by performing laboratory tests. In addition, the study developed a design system and
performed Life Cycle Cost Analysis to identify the benefits of geocell reinforcement and provided
construction guidelines.
Based on the FEA and laboratory evaluation, the study identified the following:
• Geocells are beneficial when construction needs to be performed with a lower/marginal
base and subgrade material.
• The geocells can be used in an urban environment where the height of the pavement
structure is restricted due to the curb and gutter.
• The geocell reinforced layer reduced the stresses up to 20% six inches away from the center
of the loading plate and up to 50% nine inches away from the center of the loading plate
compared with no geocell condition (unreinforced layer).
• The geocells allow placement on top of the weak quality subgrade and can also reduce the
thickness of the base layer.
• It is uneconomical to use geocells of less than 4 in. height and when the base modulus is
higher than 20 ksi.
• The life cycle cost analysis indicated that it is economical to construct pavements with
geocells when the weak/marginal material is available.
vii
viii
TABLE OF CONTENTS
Page
EXECUTIVE SUMMARY ........................................................................................................ vii
LIST OF TABLES ..................................................................................................................... xiii
LIST OF FIGURES .................................................................................................................... xv
1. PROJECT OUTLINE .......................................................................................................... 1
1.1 NATURE OF THE PROBLEM ........................................................................................ 1
1.2 TECHNICAL OBJECTIVES ............................................................................................ 1
1.3 RESEARCH PLAN AND REPORT ORGANIZATIONS.............................................. 1
2. LITERATURE REVIEW .................................................................................................... 3
2.1 GENESIS ............................................................................................................................. 3
2.1.1 Construction with Geocell .......................................................................................... 4
2.1.2 Load Support Mechanism .......................................................................................... 4
3. FINITE ELEMENT MODELING (FEM) AND ANALYSIS OF RESULTS ............... 25
3.1 PARAMETRIC STUDY .................................................................................................. 25
3.2 INFLUENCE OF BASE MODULUS (BOTH INFILL AND COVER LAYERS
WITH SIMILAR MODULUS VALUES) ....................................................................... 29
3.3 INFLUENCE OF COVER THICKNESS....................................................................... 30
3.4 INFLUENCE OF COVER BASE MATERIAL............................................................. 39
3.5 INFLUENCE OF SUBGRADE MODULUS .................................................................. 39
3.6 INFLUENCE OF GEOCELL LAYER THICKNESS OR GEOCELL HEIGHT ..... 40
4. SELECTION OF MATERIAL, EXPERIMENTAL DESIGN, AND
LABORATORY EVALUATION ...................................................................................... 51
4.1 EXPERIMENTAL DESIGN ........................................................................................... 51
4.2 LABORATORY EVALUATION .................................................................................... 51
4.3 MATERIAL SELECTION .............................................................................................. 54
4.3.1 Geocell ........................................................................................................................ 54
4.3.2 Base and Subgrade Selection ................................................................................... 54
4.4 LABORATORY EVALUATION PROCESS ................................................................ 56
4.4.1 Laboratory Set-Up .................................................................................................... 56
4.5 DATA REDUCTION AND CLEANING........................................................................ 61
4.5.1 Noise Removal from Zero Reading (Datum) .......................................................... 61
4.5.2 Noise Removal from Actual Data and Data Reduction ......................................... 62
4.5.3 Presentation of Data ................................................................................................. 64
4.5.4 Summary of Data ...................................................................................................... 67
ix
5. CRITICAL EVALUATION OF EXISTING DESIGN METHODS ............................. 71
5.1 TEXAS PAVEMENT DESIGN PROCEDURE............................................................. 71
5.1.1 Geosynthetics in Pavement Structure ..................................................................... 72
5.1.2 Geosynthetics for Geotechnical Reinforcement ..................................................... 72
5.2 LOW VOLUME ROAD DESIGN AASHTO ................................................................. 72
5.2.1 Aggregate-Surface Roads ......................................................................................... 73
5.3 DESIGN METHODS FOR LOW VOLUME UNPAVED ROADS
WITH GEOCELL REINFORCEMENT ....................................................................... 86
5.3.1 PRESTO GEOSYSTEM Design Method ............................................................... 86
5.3.2 Design Method Presto Geosystems for Low Volume Roads –
Unpaved Roads (Pokharel 2010) ............................................................................. 91
5.3.3 Example ..................................................................................................................... 93
5.3.4 Conclusions ................................................................................................................ 94
6. MODEL DEVELOPMENT, STATISTICAL ANALYSIS, AND VALIDATION ....... 97
6.1 PATH TO DEVELOPMENT OF MODEL.................................................................... 97
6.2 DATA FOR DEVELOPING MODEL ............................................................................ 97
6.3 DEVELOPMENT AND VALIDATION OF MODEL FOR
ESTIMATING BENEFIT OF GEOCELL REINFORCED LAYER .......................... 98
6.4 MULTILINEAR REGRESSION MODEL AND VERIFICATION
OF ASSUMPTIONS ......................................................................................................... 99
6.4.1 Initial Model .............................................................................................................. 99
6.4.2 Linear Relationship ................................................................................................ 101
6.4.3 Variation in Residuals Same for Large and Small ŷ Values ............................... 102
6.4.4 Distribution of Residuals ........................................................................................ 103
6.4.5 Multicollinearity...................................................................................................... 103
6.4.6 Independent Observations ..................................................................................... 105
6.5 MODEL DEVELOPMENT USING CROSS-VALIDATION TECHNIQUES ........ 108
6.5.1 Full cross-validation or leave-one-out validation method (LOOV) ................... 110
6.5.2 KK-FOLD Cross-Validation .................................................................................. 111
6.5.3 Bootstrapping Cross-Validation ............................................................................ 112
6.6 SUMMARY OF MODELS ............................................................................................ 113
7. LIFE CYCLE COST ANALYSIS ................................................................................... 115
7.1. DETERMINISTIC APPROACH ................................................................................. 115
7.1.1 Establish alternative pavement design strategies for the analysis period ......... 115
x
7.1.2 Compute Net Present Worth (NPW) or Equivalent Uniform Annual
Costs (EUAC) .......................................................................................................... 117
7.1.3 Analyze Results ....................................................................................................... 118
7.1.4 Reevaluate Design Strategies ................................................................................. 118
7.2. ESTIMATION OF GEOCELL LAYER MODULUS ................................................ 119
7.2.1 Comparison of Geocell vs. No Geocell .................................................................. 119
7.3. ESTIMATION OF COST OF GEOCELL REINFORCED LAYER ....................... 125
7.3.1 Design 1: FM 55 Ellis County of Dallas District .................................................. 126
7.3.2 Design 1: US 83 Uvalde County, San Antonio District ........................................ 131
7.3.3 Design 1: US 67 at FM 1001 Titus County, Atlanta District............................... 134
8. CLOSURE ......................................................................................................................... 137
8.1 SUMMARY ..................................................................................................................... 137
8.2 CONCLUSION ............................................................................................................... 137
8.3 RECOMMENDATION .................................................................................................. 139
9. REFERENCES.................................................................................................................. 141
A. APPENDIX A: SITE INSTRUMENTATION ............................................................... 145
A.1 INTRODUCTION.......................................................................................................... 145
A.2 SITE LOCATION AND MATERIAL PROPERTIES .............................................. 145
A.2.1 Site Location ........................................................................................................... 145
A.2.2 Soil Classification ................................................................................................... 146
A.3 INSTRUMENTATION ................................................................................................. 152
A.4 CONSTRUCTION OF REINFORCED AND UNREINFORCED SECTIONS ...... 156
A.5 DATA COLLECTION AND ANALYSIS ................................................................... 161
B. APPENDIX B- FINITE ELEMENT MODEL DEVELOPMENT............................... 167
B.1 SOIL MODELS .............................................................................................................. 168
B.1.1 Linear Elastic .......................................................................................................... 168
B.1.2 Mohr-Coulomb ....................................................................................................... 168
B.1.3 FHWA Soil Constitutive Model ............................................................................ 170
B.2 SHELL ELEMENT TYPE FOR GEOCELL MODELING ...................................... 172
B.2.1 Shell Element .......................................................................................................... 172
B.2.2 Thick Shell Element (TSHELL)............................................................................ 173
B.2.3 Geocell Dimensions and Properties ...................................................................... 173
B.3 CONTACT MODEL ...................................................................................................... 174
B.3.1 Automatic Single Surface Contact ........................................................................ 174
xi
B.3.2 Discrete Beam Element Interface ......................................................................... 175
B.4 NODE COMPATIBILITY FROM GEOCELL AND GEOMATERIAL
AT LAYER INTERFACE ............................................................................................ 176
B.5 BOUNDARY CONDITIONS ........................................................................................ 177
B.6 LEVEL OF SOPHISTICATION OF FINITE ELEMENT MODELS ..................... 178
B.6.1 Simplified Single Layer FE Model........................................................................ 178
B.6.2 Pavement with Single Cell ..................................................................................... 179
B.6.3 Geocell-Reinforced Pavement with Geocell Panel Modeled Using
Rhomboidal Pattern............................................................................................... 180
B.6.4 Geocell-Reinforced Pavement with Geocell Panel Modeled Using
Pseudo-Sinusoidal Pattern .................................................................................... 183
B.7 SELECTION OF LOADING CONDITIONS: PLATE SIZE AND LOCATION ... 184
B.8 RESULTS FROM PRELIMINARY INVESTIGATION: EVALUATION OF
SOIL MODELS AND SHELL ELEMENT TYPE FOR MODELING
THE GEOCELL ............................................................................................................ 185
B.8.1 Geocell-Reinforced Pavement with Geocell Panel Modeled
Using Pseudo-Sinusoidal Pattern ......................................................................... 185
B.8.2 Comparison to University of Kansas Model ........................................................ 190
B.8.3 Preliminary Evaluation of Contact....................................................................... 190
B.8.4 Evaluation of Geocell Shape: Geocell Panel Simulated with
Rhomboidal Pattern............................................................................................... 191
B.8.5 Evaluation of Element Type for Modeling the Geocell ....................................... 192
B.8.6 Use of Shell Elements (SHELL) for Modeling Geocell and Automatic
Surface Contact Type ............................................................................................ 192
B.8.7 Use of Thick Shell Elements for Modeling Geocell and Automatic
Surface Contact Type ............................................................................................ 193
B.9 EVALUATION OF CONTACT TYPE FOR SOIL-GEOCELL INTERFACE ...... 196
B.9.1 Assessment of Contact Type Using Simplified Single-Layer Model ................. 196
B.9.2 Evaluation of Contact Type Using Single-Cell 3-D FE Model ........................... 198
C. APPENDIX C- GEOCELL INFORMATION ............................................................... 201
D. APPENDIX D .................................................................................................................... 209
D.1 PAIRED T-TEST ........................................................................................................... 217
D.1.1 Procedure ................................................................................................................ 217
D.2 PEARSON CORRELATION ....................................................................................... 218
xii
LIST OF TABLES
Page
Table 1-1 List of Tasks and Associated Chapters........................................................................... 2
Table 2-1 Various Studies Performed on Use of Geocells. ............................................................ 8
Table 2-2 Test Setups Developed for Evaluation of Geocell. ...................................................... 14
Table 2-3 Numerical Modeling of Geocell-Reinforced Layer. .................................................... 17
Table 2-4 Finite Element Modeling Geocell Reinforced Layer and Contact Models. ................. 19
Table 2-5 Existing Pavement Design Methods of Geocell-Reinforced Layers ............................ 22
Table 2-6 Summary of Research Performed at Kansas State University ..................................... 24
Table 4-1 Experimental Test Plan................................................................................................. 53
Table 4-2 Geocell Selection and Properties .................................................................................. 54
Table 4-3 Test Procedures for Evaluation of Base and Subgrade Material .................................. 54
Table 4-4 Measured Engineering Properties of Base and Subgrade Materials ............................ 55
Table 4-5 Factors Evaluated in Laboratory .................................................................................. 56
Table 4-6 Strain Gauges................................................................................................................ 59
Table 4-7 Vertical Deformation (below loading plate)................................................................. 68
Table 4-8 Stresses Measured on Top of Subgrade (below loading plate) .................................... 68
Table 4-9 Vertical Deformation (below loading plate)................................................................. 69
Table 4-10 Hoop Strain (cell below loading plate) ....................................................................... 69
Table 4-11 Summary of Measured Stress and Strain with 12” Loading Plate ............................. 70
Table 5-1 Geosynthetic Application (Pavements in Texas). ........................................................ 72
Table 5-2 Suggested Seasons Length (Months) for the Six U.S. Climatic Regions. .................... 75
Table 5-3 Suggested Seasonal Roadbed Soil Resilient Moduli, MR (psi), as a Function of the
Relative Quality of the Roadbed Material. ................................................................................... 75
Table 5-4 Effective Roadbed Soil Resilient Modulus Values, MR (psi) that may be used in the
Design of Flexible Pavements for Low-Volume Roads. Suggested Values Depend on the U.S.
Climatic Region and the Relative Quality of the Roadbed Soil. .................................................. 75
Table 5-5 Chart for Computing Total Pavement Damage (for both Serviceability and Rutting
Criteria) Based on a Trial Aggregate Base Thickness. ................................................................. 78
Table 5-6 Computation of Total Pavement Damage (for both Serviceability and Rutting Criteria)
Based on Trial Aggregate Base Layer. ......................................................................................... 80
Table 5-7 Computation of Total Pavement Damage (for both Serviceability and Rutting Criteria)
Based on 10” Aggregate Base Layer. ........................................................................................... 82
Table 5-8 Chart for Computing Total Pavement Damage (for both Serviceability and Rutting
Criteria) Based on a Trial Aggregate Base Thickness. ................................................................. 86
Table 5-9 Methodology for Designing Gravel Roads (Unreinforced).......................................... 87
Table 5-10 Correlation of Subgrade Soil Strength Parameters for Cohesive (Fine-Grained) Soils
....................................................................................................................................................... 87
Table 5-11 Methodology for Designing Gravel Roads with Geocell Reinforced Base ............... 88
Table 5-12 Recommended Peak Friction Angle Ratio (Presto Geosystems) ............................... 89
Table 5-13 Recommended Peak Friction Angle Ratio (Presto Geosystems) for Design ............. 92
Table 6-1 Range of Input Variables Used for Developing the Model .......................................... 98
Table 6-2 Correlation Matrix ...................................................................................................... 105
Table 6-3 Variance Inflation Factors (VIF) ................................................................................ 105
Table 6-4 Verification of Assumptions in Multiple Regression ................................................ 106
xiii
Table 6-5 Multiple Linear Regression Model Parameters for Reduced Model (LOOV) ... 110
Table 6-6 Multiple Linear Regression Model Parameters for Reduced Model (KK-FOLD) 111
Table 6-7 Multiple Linear Regression Model Parameters for Reduced Model (Bootstrapping
CV) ............................................................................................................................................. 112
Table 6-8 Summary of Models Developed Using Cross-Validation Techniques....................... 114
Table 7-1 Cost of Base Materials Considered in this Study ....................................................... 126
Table 7-2 Inputs for Pavement Design (Ellis County, Dallas District) ...................................... 127
Table 7-3 Cost Estimate for Geocell Reinforced Layer (Ellis County, Dallas District)............. 129
Table 7-4 Input Data for Pavement Design (Uvalde County, San Antonio District) ................ 131
Table 7-5 Cost Estimate of Geocell-Reinforced Layer (Uvalde County, San Antonio District) 133
Table 7-6 Input Data for Pavement Design (Titus County, Atlanta District) ............................. 134
Table 7-7 Cost Estimate of Geocell-Reinforced Layer (Titus County, Atlanta District) ........... 136
Table A-1 Stress at Different Confinements (Subgrade soil) ..................................................... 149
Table A-2 Average Stresses at Different Confinements ............................................................. 151
Table A-3 Strain Gauges............................................................................................................. 152
Table A-4 Results of MODULUS Software ............................................................................... 163
Table A-5 Results of Field Testing and Bisar Modeling ............................................................ 163
Table B-1 Material Properties Used for Linear-Elastic Model. .................................................. 169
Table B-2 Material Properties Used for Mohr-Coulomb Material Model. ................................. 170
Table B-3 Soil Properties Used for Modeling Base and Subgrade Used for Parametric Study and
for Evaluation of Geocell Element Types and Contact .............................................................. 171
Table B-4 FHWA Soil Material Properties Used for Evaluation of the University of Kansas Study
(after Yang. 2010) ....................................................................................................................... 172
Table B-5 Geocell Dimensions and Properties. .......................................................................... 173
Table B-6 Dimensions of Single Cell FE Model. ....................................................................... 180
Table B-7 Dimensions and Properties of Geocell-Reinforced Pavement FE Model with Geocell
Panel Simulated Using Rhomboidal Shaped Cells. .................................................................... 181
Table B-8 Dimensions and Properties of Geocell-Reinforced Pavement FE Model with Geocell
Panel Simulated Using Pseudo-Sinusoidal Shaped Cells. .......................................................... 184
Table B-9 Summary of Numerical Models Used for Evaluating Contact Types. ...................... 197
Table B-10 Summary of Longitudinal and Vertical Stress at the Mid-Height of Model Observed
at the Edge of Cell#1, at the Geocell, and at the Edge of Cell#2 in contact with the Geocell. .... 198
Table D-1 Raw Data ................................................................................................................... 209
Table D-2 Training Data Set LOOV-CV .................................................................................... 211
Table D-3 Testing Data Set LOOV-CV...................................................................................... 212
Table D-4 Training Data Set KK FOLD CV .............................................................................. 213
Table D-5 Testing Data Set KK FOLD CV ................................................................................ 214
Table D-6 Training Data Set Bootstrapping CV......................................................................... 215
Table D-7 Testing Data Set Bootstrapping CV .......................................................................... 216
xiv
LIST OF FIGURES
Page
Figure 2-1 Geocell Supplied. .......................................................................................................... 3
Figure 2-2 Application of Geocell (US Army Corps). ................................................................... 4
Figure 2-3 Geocell Construction Sequence. ................................................................................... 5
Figure 2-4 Mechanics of Geocell Reinforcement. .......................................................................... 5
Figure 3-1 Quarter Model used in the Parametric Study. ............................................................. 28
Figure 3-2 Locations of Output Evaluated from FEA. ................................................................. 28
Figure 3-3 Comparison of Compressive Stress at 6” from Loading Center
of Loading Plate (Geocell vs. No Geocell). ................................................................ 28
Figure 3-4 Stress Distribution (Geocell vs. No Geocell) along the Subgrade. ............................. 29
Figure 3-5 Strain Distribution (Geocell vs. No Geocell) along the Subgrade. ............................. 30
Figure 3-6 Stress Distribution (Geocell 4” and 6” vs. No Geocell)along the Subgrade
(Influence of Infill Modulus). ..................................................................................... 31
Figure 3-7 Strain Distribution (Geocell 4” and 6” vs. No Geocell) along the Subgrade
Figure 3-8 Vertical Deformation (Geocell 4” and 6” vs. No Geocell)
Figure 3-9 Hoop Strains on Geocell 4” and 6” (Influence of Infill Modulus). ............................. 33
Figure 3-10 Strain Distribution (Geocell 4” and 6” vs. No Geocell) along the Subgrade
(Influence of Cover Thickness). ................................................................................ 34
Figure 3-11 Vertical Deformation (Geocell 4” and 6” vs. No Geocell)
(Influence of Cover Thickness). ............................................................................... 35
Figure 3-12 Hoop Strain on Geocell 4” and 6” vs. No Geocell
(Influence of Cover Thickness). ............................................................................... 35
Figure 3-13 Stress Distribution (Geocell 3”, 4” and 6” vs. No Geocell) along the Subgrade
(Influence of Good Cover and Poor Infill). .............................................................. 36
Figure 3-14 Strain Distribution (Geocell 3”, 4” and 6” vs. No Geocell) along the Subgrade
Figure 3-15 Vertical Deformation (Geocell 3”, 4” and 6” vs. No Geocell)
(Influence of Good Cover and Poor Infill). ............................................................. 38
Figure 3-16 Hoop Strains on Geocell 3”, 4” and 6” vs. No Geocell
Figure 3-17 Stress Distribution (Geocell 3” vs. No Geocell) along the Subgrade
(Influence of Subgrade Modulus). ............................................................................ 41
Figure 3-18 Strain Distribution (Geocell 3” vs. No Geocell) along the Subgrade
Figure 3-19 Vertical Deformation (Geocell 3” vs. No Geocell)
Figure 3-20 Hoop Strain on Geocell 3” (Influence of Subgrade Modulus). ................................. 43
xv
Figure 3-24 Hoop Strain on Geocell 4” (Influence of Subgrade Modulus). ................................. 46
Figure 3-28 Hoop Strains on Geocell 6” (Influence of Subgrade Modulus). ............................... 49
Figure 4-1 Photo of Fabricated Box.............................................................................................. 56
Figure 4-2 Sample Preparation Using Vibratory Compactor ....................................................... 57
Figure 4-3 Applied Load Cycle .................................................................................................... 58
Figure 4-4 Locations of Stress and Strains Measurement Transducers ........................................ 58
Figure 4-5 Half Bridge Vs. Quarter Bridge Strain Gauge Circuits............................................... 59
Figure 4-6 Photo of Earth Pressure Cell ....................................................................................... 60
Figure 4-7 a) MTS Data Acquisition System, b) LMS Data Acquisition System ........................ 61
Figure 4-8 Noise removal from pressure cell readings ................................................................. 62
Figure 4-9 a) Pressure Cell 1 (Original Data), b) Pressure Cell 2 (Original Data),
c) Pressure Cell 3 (Original Data), d) Pressure Cell 3 (Kernel Regression). .............. 63
Figure 4-10 A) Strain gauge (Original Data), B) Strain gauge (Kernel Regression) ................... 64
Figure 4-11 Vertical Deformation A) Geocell, B) No Geocell. ................................................... 65
Figure 4-12 Vertical Stresses on Subgrade a) Geocell, b) No Geocell. ........................................ 65
Figure 4-13 a) Location and Direction of Strain Gauges b) Vertical Strain Observed
in the Center Geocell. ............................................................................................... 66
Figure 4-14 Hoop Strains on Geocell. .......................................................................................... 66
Figure 5-1 The Six Climatic Regions in the United States (AASHTO, 1993). ............................ 74
Figure 5-2 Design Chart for Aggregate-Surfaced Roads Considering Allowable
Serviceability Loss (AASHTO, 1993). ....................................................................... 76
Figure 5-3 Design Chart for Aggregate-Surfaced Roads Considering Allowable Rutting
(AASHTO, 1993)........................................................................................................ 77
Figure 5-4 Example Growth of Total Damage Versus Base Layer Thickness for Both
Serviceability and Rutting Criteria. ............................................................................ 81
Figure 5-5 Chart to Convert a Portion of the Aggregate Base Layer Thickness
to an Equivalent Thickness of Subbase (AASHTO, 1993)......................................... 81
Figure 5-6 Design Chart for Aggregate-Surfaced Roads Considering Allowable
Serviceability Loss (Geocell Reinforced Layer Calculations). (AASHTO, 1993) ..... 83
Figure 5-7 Design Chart for Aggregate-Surfaced Roads Considering Allowable Rutting
(Geocell Reinforced Layer Calculations) (AASHTO, 1993). .................................... 84
Figure 5-8 Chart to Convert a Portion of the Aggregate Base Layer Thickness to an
Equivalent Thickness of Subbase (Geocell Reinforced Layer Calculations)
(AASHTO, 1993)........................................................................................................ 85
Figure 5-9 Design Calculation Geocell Reinforced and Unreinforced Unpaved Roads
(Presto Geosystems).................................................................................................... 90
Figure 6-1 Summary of Initial Regression Model (Full Model)................................................. 100
Figure 6-2 Linear Relationship Between Independent Variable and Dependent Variables. ...... 102
xvi
Figure 6-3 Residuals vs. Estimated (fitted) Values .................................................................... 103
Figure 6-4 Verification of Distribution of Residuals .................................................................. 104
Figure 6-5 Summary of Reduced Model. ................................................................................... 107
Figure 6-6 Example of Leave One Out Validation. .................................................................... 109
Figure 6-7 Example of KK-fold Segmented Cross-Validation................................................... 109
Figure 6-8 Example of Bootstrapping......................................................................................... 110
Figure 6-9 Predicted Data vs. Measured Data on Training and Testing Data Sets .................... 111
Figure 6-10 Predicted Data vs. Measured Data on Training and Testing Data Sets .................. 112
Figure 6-11 Predicted Data vs. Measured Data on Training and Testing Data Sets .................. 113
Figure 7-1 Cash flow diagram for a pavement ........................................................................... 117
Figure 7-2 Cash flow diagram for NPW and EUAC .................................................................. 118
Figure 7-3 Pavement Section Evaluated in the Laboratory ........................................................ 120
Figure 7-4 Stress Distribution a) Geocell and b) No Geocell ..................................................... 120
Figure 7-5 Benefit of Geocell in Reduction of Stress (Laboratory Results)............................... 121
Figure 7-6 Comparison of unreinforced section with BISAR equivalent Section
(Laboratory Results) ................................................................................................. 121
Figure 7-7 Comparison of geocell reinforced section with BISAR equivalent Section
(Laboratory Results) ................................................................................................. 122
Figure 7-8 Stress Distribution a) Geocell and b) No Geocell ..................................................... 123
Figure 7-9 Benefit of Geocell in Reduction of Stress (FEM) ..................................................... 124
Figure 7-10 Equivalent Modulus Calculation (FEM) ................................................................. 124
Figure 7-11 Estimated Geocell Layer Modulus with Various Infills (4inch Geocell) ............... 125
Figure 7-12 Alternative Designs Developed Using FPS 21 and Geocell Reinforced
Layer (Ellis County, Dallas) ................................................................................... 128
Figure 7-13 Estimated Geocell Reinforced Layer Costs using Probabilistic
LCCA (Ellis County, Dallas). ................................................................................. 130
Layer for Uvalde County, San Antonio District. .................................................... 132
Figure 7-15 Estimated Geocell Reinforced Layer Costs using Probabilistic LCCA
(Uvalde County, San Antonio). .............................................................................. 132
Layer for Titus County, Atlanta District. ................................................................ 135
Figure 7-17 Estimated Geocell Reinforced Layer Costs using Probabilistic LCCA
(Titus County, Atlanta). .......................................................................................... 135
Figure A-1 Lamar County, Paris TX Site Location. (Source: Wikipedia). ................................ 145
Figure A-2 Test Site Location (Source: ArcMAP). .................................................................... 146
Figure A-3 Mech. Sieve Soil Classification. .............................................................................. 147
Figure A-4 Liquid Limit. ............................................................................................................ 147
Figure A-5 Subgrade Mohr's Circles. ......................................................................................... 148
Figure A-6 MD Curve Subgrade Soil. ........................................................................................ 148
Figure A-7 MD Curve of Flexbase Material............................................................................... 150
Figure A-8 Flex-Base Mohr Circles. .......................................................................................... 150
Figure A-9 Flex base classification............................................................................................. 151
Figure A-10 Photo of Earth Pressure Cell. ................................................................................. 152
Figure A-11 Pavement Section of Geocell-Reinforced at Testing Site. ..................................... 153
Figure A-12 Pavement Section of FM 906 at Testing Site. ........................................................ 153
xvii
Figure A-13 Cross Section and Instrumentation of FM 906 at Testing Site
for Unreinforced Section........................................................................................ 154
Figure A-14 Wiring Arrangement at FM 906 at Testing Site for Unreinforced Section. ........... 154
Figure A-15 Cross Section and Instrumentation of FM 906 at Testing Site for Geocell
Reinforced Section. ................................................................................................ 155
Figure A-16 Wiring Arrangement at FM 906 at Testing Site for Geocell Reinforced Section. . 155
Figure A-17 LMS Data Acquisition System............................................................................... 156
Figure A-18 Marked Sensor Location for Unreinforced Section. .............................................. 157
Figure A-19 Installation of Pressure Cells. ..................................................................................158
Figure A-20 Installation of Geosynthetics. ..................................................................................159
Figure A-21 Installation of Instrumented Geocells. ....................................................................160
Figure A-22 Geocell Construction Sequence. .............................................................................161
Figure A-23 Example of pressure cell result from one FWD drop............................................. 162
Figure A-24 Comparison of Pressure Cell Response between Non-Geocell and
Geocell Sites. ......................................................................................................... 164
Figure A-25 FWD Geophone Response of Non-Geocell and Geocell Spot 3. ........................... 165
Figure A-26 FWD Geophone Response of Non-Geocell and Geocell Spot 5. ........................... 165
Figure B-1 Geocell and Infill Material. ...................................................................................... 167
Figure B-2 Stress-Strain Relationship for Linear Elastic Material Model. ................................ 169
Figure B-3 Stress-Strain Relationship with Plastic Behavior. .................................................... 169
Figure B-4 Comparison of Mohr-Coulomb Yield Surfaces in Shear Stress-Pressure Space. .... 171
Figure B-5 Representation of (a) Four-Node (quad) BLT Shell Element with 5 Local DOFs,
1 Integration Point in the Plane and 5 Through-Thickness Integration Points,
and (b) Eight Node Thick Shell Element with 5 Local DOFs, Single (green)
or Reduced (red) Integration Points in the Plane and 5 Through-Thickness
Integration Points. .................................................................................................... 173
Figure B-6 Automatic Single Surface Contact to Model Geocell Soil Interface ........................ 174
Figure B-7 Contact Using Discrete Beams (Spring with Normal and
Tangential Components). .......................................................................................... 175
Figure B-8 (a) Unreinforced Pavement Indicating Location Where Geocell must be
Inserted and (b) Mesh Transition at the Layer interface from
Thick Shell (TSHELL) Elements to Solid Elements. .............................................. 176
Figure B-9 Side View of FE Model Showing Boundary Conditions, Shown as Triangles,
with Base Layer with no Restrained Conditions to Allow for Expansion. .............. 177
Figure B-10 Simplified Single Layer FEM with Soil Cells of Base Material Separated
with (a) BLT Shell (SHELL) Element Types, and (b) Thick Shell (TSHELL)
Element Types. ....................................................................................................... 178
Figure B-11 Loading Cycles Used for Simplified FE Model. .................................................... 179
Figure B-12 (a) Laboratory Setup, (b) FE Model of Laboratory Setup, and (c) Setup of
Strain Gauges on Geocell for Evaluating Soil-Geocell Interaction. ...................... 179
Figure B-13 Cyclic Loading with Constant Peak Pressure. ........................................................ 180
Figure B-14 Finite Element Model of Pavement Structure with Geocell Panel Simulated
using Rhomboidal Shape Cells: (a) Top View Highlighting Quarter Model
and (b) Embedding of Geocell Reinforcement in Base Material. .......................... 181
Figure B-15 Cyclic Loading Increase from 0 to 80 psi. ............................................................. 182
Figure B-16 Monotonically Increasing Static Load.................................................................... 182
xviii
using Pseudo-Sinusoidal Shaped Cells: (a) Top View Highlighting Quarter
Model and (b) Embedding of Geocell Reinforcement in Base Material. .............. 183
Figure B-18 Position of Load in the FE Model at (a) Center of Geocell and
(b) Joint of Geocell. ............................................................................................... 184
Figure B-19 Vertical Deflection along the Depth of the Pavement Structure. ........................... 185
Figure B-20 Vertical Stress (z-direction) in (a) Geocell-Reinforced
(b) Unreinforced Sections. ..................................................................................... 186
Figure B-21 Vertical Deflection with Respect to Depth at Peak Load of First Cycle
for a Two-Layer Pavement System Consisting of Subgrade-1
(E=1.74 ksi, c=1.45 psi, φ=15°) and Different Base Properties. ............................ 187
Figure B-22 Percentage Reduction in Surface Deformation with respect to Base Modulus
for a Pavement with Subgrade-1 (E=1.74 ksi, c=1.45 psi, φ =15°) Material. ....... 188
Figure B-23 Permanent Deformation for Geocell-Reinforced and Unreinforced Pavement
Sections using FHWA Soil Model Properties: Kansas River Sand base
(E=0.48 ksi, c=0, φ=41°) and Clay Subgrade (E=1.5 ksi, c=15.2 psi, φ =0°)
and Repeated 80 psi Loading. ................................................................................ 188
Sections using FHWA Soil Model Properties: Kansas River Sand base
(E=0.48 ksi, c=0, φ=41°) and Clay Subgrade (E=1.5 ksi, c=15.2 psi, φ =0°)
and Increasing Repeated Loading Reaching 80 psi Followed by 80 psi
Repeated Loads. ...................................................................................................... 189
Figure B-25 Impact of Ramp Loading on Permanent Deformation for Geocell-Reinforced
and Unreinforced Pavement Sections using Kansas River Sand Base and
Clay Subgrade. ....................................................................................................... 189
Figure B-26 Pressure to Surface Displacement Curves for Unreinforced and
Geocell-Reinforced Sections using (a) FHWA Constitutive Soil Model
and (b) Duncan Chang Constitutive Model used in the
University of Kansas Study (after Yang, 2010). ..................................................... 190
Figure B-27 Node Penetration and Overlapping of Element at the Joints. ................................. 191
Figure B-28 (a) Top View of FE Model Showing Nodal Unbonding to Accommodate
Geocell and (b) Zoomed in View at Joint Showing Infill Material and
Geocell Modeled using Shell Elements. ................................................................. 191
Figure B-29 Stresses Developed at the Edges of the Model. ...................................................... 192
Figure B-30 Longitudinal Displacement (dx) at 10th Loading Cycle after FE model with
Geocell using Shell Elements and Automatic Single Surface Contact Element. ... 193
Figure B-31 Vertical Stress (Δx) at 10th Loading Cycle after FE Model with Geocell using
Shell Elements and Automatic Single Surface Contact Element........................... 193
Figure B-32 Longitudinal Displacement (dx) at 1st Loading Cycle after FE Model with
Geocell using Thick Shell Elements and Automatic Single Surface
Contact Element. ..................................................................................................... 194
Figure B-33 Vertical Stress (σx) at 1st Loading Cycle after FE Model with Geocell
using Thick Shell Elements and Automatic Single Surface Contact Element....... 194
Figure B-34 Time Step Size as Reported by the d3hsp Output File Required for the
Analysis of Simplified FE Model using (a) Shell Elements and
(b) TSHELL Elements. .......................................................................................... 195
xix
Figure B-35 Time Required for 1 Second of Analysis for 480 Element FE model
with Different Number of 1.65 GHz Processors. .................................................... 195
Figure B-36 Simplified FE Model with Highlighted Elements used for Evaluating
theStress Transfer Through Contact and Geocell. .................................................. 197
Figure B-37 Hoop Strains at Edge of Geocell for (a) Laboratory and (b) FEM. ........................ 199
Figure B-38 Hoop Strains at Mid-Geocell for (a) Laboratory and (b) FEM. ............................. 199
Figure B-39 Strains at Center of Geocell for (a) Laboratory and (b) FEM. ............................... 200
Figure C-1 Tenax Geocell Properties. ........................................................................................ 201
Figure C-2 Presto Geocell Properties. ........................................................................................ 203
Figure C-3 Strain Gauge Specifications. .................................................................................... 205
Figure C-4 Pressure Cell Specifications. .................................................................................... 207
Figure C-5 Data Smoothing: Kernel Regression vs. Moving Average. ..................................... 208
xx
1. PROJECT OUTLINE
1.1 NATURE OF THE PROBLEM
To enhance the bearing capacity of the subgrade, the geocell has been promoted as one of the
products that reduce the magnitude of traffic load on the subgrade layer by laterally distributing
traffic load through lateral confinement. In the recent decade, the product is also promoted for
reducing base layer thickness when lower quality or recycled base material is used. Since the use
of geocell for a reduction in base layer thickness in conjunction with lower quality subgrade layer
has not been studied, the focus of this study is to evaluate geocell ability in reducing the base layer
thickness when poor quality subgrade and base materials are readily available. To successfully
implement and use geocells regularly, a design system needs to be developed and will be the
additional focus of this project.
1.2 TECHNICAL OBJECTIVES
The following are the objectives of this project:

i. To evaluate the existing pavement design methods with geocell reinforcement and
state-of-practice in terms of geocell application in pavement design and construction.
ii. Characterize mechanism for improved support.
iii. Develop an experiment design to characterize the mechanisms for the improved support
that geocell provides to the layers' bearing capacity, resulting in thickness reduction of
base or subbase layers.
iv. Develop a finite element model to replicate the laboratory experiment set up and
perform a parametric analysis.
v. Perform laboratory evaluation and conduct statistical analyses of the collected
laboratory data to identify parameters that significantly influence the performance of
the pavement system constructed with geocell.
vi. Develop a design system for future pavement construction consisting of geocell.
vii. Perform lifecycle cost analysis with pavements using the geocell reinforced layer.
viii. Develop specifications and construction steps for future highway construction
consisting of geocell.
ix. Document the validity and practicality of the selected design and construction
approach.
1.3 RESEARCH PLAN AND REPORT ORGANIZATIONS
The objectives mentioned above are classified into various tasks listed in Table 1-1, and a complete
description of each task is explained in separate chapters mentioned in Table 1-1. The whole
project is classified into nine tasks and is explained in this report.
1
Table 1-1 List of Tasks and Associated Chapters.
Task Description Objectives Chapter
Study the published research on geocell. Understand the working mechanism of geocell and identify the research
Task 1 Information Search Chapter 2
gaps. Perform preliminary testing. (This task covers the first two objectives of the study).
Task 2 and Task 3 were started simultaneously. This task is performed based on the results of task 1 and
Finite element modeling preliminary results from Task 3. In this task, finite element models with and without geocell are developed and
Task 2 (FEA) and analysis of calibrated with the preliminary results of Task 3. In addition, this task explains the origin of the FEA model Chapter 3
results developed for the study, the selection of numerous parameters, and their values. Chapter 3 provides the complete
information on the FEA of this study.
In this task, the experiments were designed. Two subgrades and three base materials from various locations in
Selection of Material, Texas were collected. The testing mold was fabricated. Electronic devices such as strain gauges and pressure cells
Task 3 Experiment Design, and were used to measure the responses of the samples during testing. Chapter 4 provides information regarding Chapter 4
Laboratory Evaluation material properties, experiment design, and evaluation of laboratory results. The test results were compared with
FEA results and presented in Chapter 3.
The main idea of this task is to demonstrate the development of a mathematical model for capturing the geocell
reinforced layer benefit over the unreinforced layer using multi-linear regression. Later the assumptions (linear
Task 4 Statistical Analysis relationship, collinearity, etc.) in the multiple regression are verified through statistical tools, and the model is Chapter 5
enhanced further. Cross-validation techniques are used for model development and are discussed. Some statistical
tests were performed to fine-tune the laboratory test results and are discussed in Chapter 4.
This task provides information on existing pavement design methods for low-volume roads with the incorporation
Critical Evaluation of
Task 5 of geocell layers. Chapter 6 covers three areas: Texas Pavement Design Method, Low Volume Pavement Design Chapter 6
Existing Design Methods
Method (AASHTO), Proposed methods on Geocell incorporation in Pavement Design.
Validation of the This task provides input parameters for designing low-volume roads in Texas with geocell reinforced layers.
Task 6 Chapter 7
Proposed Design Method Cross-validation techniques are used for validating the proposed method.
The economic feasibility of geocell in pavement construction is evaluated in this task. In addition, lifecycle cost
Task 7 Cost Analysis analysis is performed to estimate the maximum allowable cost of the geocell reinforced layer for various district Chapter 8
roads in Texas.
The main idea of this task is to guide the practitioners in material selection, design, construction, and safety of Reported
Task 8 Practitioners Guidelines
pavements with geocell layers. Separately
Results of Site This task involves instrumenting the pavements constructed with a geocell reinforced base and no geocell for
Task 9 Appendix A
Instrumentation measuring the responses under traffic load.
2
2. LITERATURE REVIEW
The pavement construction has evolved over the years with the inclusion of geosynthetic material,
starting with the simpler non‐woven to the more complex geocomposites. Technological
advancement modified geosynthetic from two‐dimensional to cellular confinement systems that
added a third dimension to geosynthetic. The Cellular Confinement Systems are popularly known
as “Geocell.” Geocell is a durable, lightweight, three-dimensional fabricated system that is
expandable on‐site to form a honeycomb‐like structure (Figure 2-1). Geocell is filled with the soil
and compacted to enhance the bearing capacity of the subgrade layer. The depth of the geocell, as
well as the size of each cellular unit, typically varies depending on the supplier as well as design
requirements. The infill material can be non-cohesive recycled material as the geocell provides
confinement and friction (geocell wall texture).
Figure 2-1 Geocell Supplied.

2.1 GENESIS
The development of geocell can be credited to the US Corps of Engineers for evaluating the
feasibility of constructing bridge approach roads over the soft ground in 1975. Geocell was
extensively used during the Vietnam War and the Gulf operations in the late 1980s (Figure 2-2).
In the civilian sector, geocell was first used for load support systems in the early 1980s in the US,
followed by slope erosion control and channel lining in 1984 in the US, and earth retention in
Canada in 1986. Today applications are many and broadly include:
• Load support systems:
o The increased bearing capacity of the foundation layer.
o Reinforcement and support systems for embankments on the weak ground;
o Reduction in pavement sections for all types of roads, laydown areas, and parking
lots.
• Gravity walls for earth retention and surcharge load support.
• Erosion control:
3
o Embankment slopes and natural slopes;
o Water channel and water bondage linings.
Figure 2-2 Application of Geocell (US Army Corps).

2.1.1 Construction with Geocell
The sequence of construction is illustrated in Figure 2-3. The subgrade layer is leveled and
compacted before the placement of geocell. Initially, the geocell was spread on top of the subgrade.
However, a geomembrane layer is placed on top of the subgrade to minimize contamination. The
spreading of geocell and interconnection of geocell is performed manually using ties and metal
anchors or wooden stakes. The spread open geocell is in‐filled using a loader or similar equipment
and compacted using a roller compactor. An additional layer of infill material is placed on top to
provide a smoother ride.
2.1.2 Load Support Mechanism
The proposed mechanics of geocell, as a load-carrying system, is illustrated in Figure 2-4. The
moving traffic imparts vertical as well as lateral stresses in the base and subgrade layers. The
geocell walls counteract the induced lateral stresses as its movement is restricted by adjacent cells.
If q₀ is the vertical pressure, the lateral stresses generated along the walls of the individual cells
would be K₀q₀ where K₀ is the coefficient of earth pressure “at rest.” The K₀ depends on the
angle of internal friction (φ) of the infill soil. This increases the shear strength of the confined
infill, which distributes the applied load over a wider area. This horizontal stress acting normal to
the cell wall increases the vertical frictional resistance between the infill and the geocell wall,
reducing the stresses induced on the layer below the geocell. This phenomenon allows transferring
relatively heavy vertical loads onto relatively weak soils.
A detailed literature review was performed to identify the state-of-the-art and research gaps to be
bridged to achieve this study's objectives. The literature review aimed to understand numerous
factors in the published studies like the study's objective, laboratory setup, numerical modeling,
and design. The comprehensive examination of published studies supported developing the
laboratory setup, experimental design and proposed a design procedure for low volume roads with
geocell. Although reviewed literature identified various applications of geocell, the literature
relevant to pavements is summarized and tabulated in the following pages.
4
Table 2-1 summarizes numerous studies performed on the application of geocell in pavement
applications. The summary included in Table 2-1 can be categorized into three groups: 1)
application of geocell reinforcement to enhance the performance of the soft base material through
geocell, 2) the behavior of geocell reinforced layer due to different loading patterns, and 3) the
benefits of geocell in enhancing the bearing capacity of soft subgrade material.
Figure 2-3 Geocell Construction Sequence.
Figure 2-4 Mechanics of Geocell Reinforcement.

Usage of poor infill in pavements using geocell was studied by (Jie Han et al. 2011), Pokharel et
al. (2009), Thakur et al. (2011), Pokharel et al. (2010), Tanyu et al. (2013), Emersleben and Meyer
(2008)). These researchers used recycled asphalt pavement (RAP), sand, or quarry waste as infill
material. The behavior of the geocell reinforced layer under various loading patterns was also
reviewed. Some researchers used the static load (Pokharel et al. (2009), Pokharel et al. (2010),
Thakur et al. (2011)), repeated loads like cyclic loading (Pokharel et al. (2009), Tanyu et al. (2013),
5
accelerated pavement testing using wheel loads in the lab (Pokharel et al. 2009), field evaluation
either using falling weight deflectometer or truck load (Emersleben and Meyer (2008), Imad L.Al
Qadi & John J Hughes (2000), Ofer Jieft et al. (2011))
Jie Han et al. (2011) observed that the stress distribution angle increased due to the geocell
reinforcement and less stress on top of the subgrade. Emersleben and Meyer (2008) evaluated
geocell of different heights (4”, 6” and 8”) in the field as well as in the laboratory. They concluded
that the geocell reinforcement decreases deflection due to an increase in reinforced layer modulus.
Additionally, the study identified that increase in geocell height from 4” to 8” enhances the
performance of the pavement, which was also observed by Pokharel et al. (2009). Al Qadi & John
J Hughes (2000) observed that the resilient modulus of the infill material doubled due to the geocell
reinforcement. An increase of base resilient modulus by 40-50% was witnessed by Tanyu et al.
(2013). Ofer Jieft et al. (2011) perceived that the modulus of infill increased 2.75 times through
the field, laboratory, and finite element evaluation.
Table 2-2 summarizes the laboratory setups of various research projects. Five parameters were
identified to be summarized in this table: 1) the dimensions of the pavement tested, 2) the geocell
material and height, 3) the thickness of the layer(s), 4) the load type, magnitude, and size of the
plate for performing tests, and 5) test parameters and transducers for measuring response due to
applied loads.
Dimensions: To minimize the end effect, the researchers fabricated boxes to evaluate geocell in
the laboratory; however, the box size differed from one study to the other. For example,
Emersleben and Meyer (2008) fabricated a box of 6.6’ x 6.6’ x 6.6’ whereas Tanyu et al. 2013
fabricated a box of 9.8’ x 9.8’ x 11.5’ size. Pokharel et al. (2009) constructed a pavement section
of 10’ x 8’ x 6’ for performing accelerated testing at Kansas State University. Even though the box
dimension varied between studies, the built structure can hold more than one layer of pavement
section, and dimensions were large enough to minimize the end effect.
Geocell Type and Size: The literature review identified two types of material used in geocell
production: High-density-polyethylene (HDPE) and Novel-polymeric-alloy (NPA). Three
different geocell heights were studies: 4”, 6”, and 8.” Although different geocell have different
openings (major and minor diameter), Emersleben and Meyer (2008) were the only ones
evaluating the influence of geocell opening on performance.
Layer Thicknesses: Since the initial focus of the geocell was to enhance the bearing capacity of
subgrade, most of the laboratory studies used a poor performing subgrade over which either geocell
reinforced or unreinforced layer was placed followed by a cover layer. However, the thickness of
the reinforced layer, unreinforced layer, and cover layer varied across studies. Tanyu et al. (2013)
used expanded polystyrene (EPS), and Emersleben and Meyer (2008) used the Glyben to simulate
the weak subgrade to minimize preparation time.
Load: Type, Magnitude, and Shape of Plate: The load applied for evaluation varied between
studies ranging from static to repeated loads to moving loads. In repeated load tests, the majority
of studies targeted 80 psi which resembles the truck tire load. For example, Jie Han et al. (2011),
Pokharel et al. (2009), and Emersleben and Meyer (2008) targeted 80 psi that simulated a truck
tire load. Emersleben and Meyer (2008), in the field evaluation, used a heavy truck weighing 41
tons traveling at a speed of 25 mph. The load magnitude used by Tanyu et al. (2013) is different
from other studies. They classified load into two classes: 1) Construction load (load on the geocell
6
layer during the construction phase, a higher load but fewer number cycles), 2) Traffic load (load
that comes with the base layer during regular traffic, a lower load but higher number of cycles).
Tanyu et al. (2013) applied 100 psi to simulate construction load and 20 psi to simulate traffic
load. The size of the loading plate varied between studies as well. Emersleben and Meyer (2008)
used a 12” diameter plate, Tanyu et al. (2013) used a 10” diameter plate, and Pokharel (2009) and
Thakur et al. (2011) used a 6” diameter plate. Even though the static and preliminary testing in
Kansas studies used a 6” and 9” loading plates, the later studies (repeated load cycle) used a 12”
diameter plate. The reasoning behind the larger diameter plate was not mentioned in the literature.
Test Parameters and Transducers: Jie Han et al. (2011) used pressure cells to monitor the
vertical stress on the subgrade. The purpose of the pressure cell was to record the change in stress
distribution angle due to the geocell reinforcement. Pokharel (2009) recorded the rut depths for
specific wheel loads, vertical stress on the subgrade, and strains on geocells. Tanyu et al. (2013)
measured strains in geocell and vertical deformations. It is noted that the researchers used three
electronic transducer types: pressure cell to record vertical stress on the subgrade, strain gauges to
record the strains on the geocell, linear variable differential transformer (LVDT), or dial gauges to
measure the vertical deformations.
Tables 2-3 and 2-4 summarize studies exclusively on numerical and finite element modeling of
geocell reinforced layers. Table 2-3 focuses on the published research on numerical modeling of
the geocell reinforced layer. This table includes literature from the geocell application in roads and
building foundations. Additionally, the table summarizes information on how the geocell
reinforced layer was modeled and how the material was characterized to identify the influence of
geocell. Most of the researchers, mentioned in this table, modeled geocell and infill material
together as a composite material and characterized its behavior either using Drucker-Prager Model
(Mhaiskar and Mandal, 1996), Duncan-Chang (Evan et al., 1994; Bathurst and Knight, 1998;
Madhavi Latha and Rajagopal, 2008; Madhavi Latha and Rajagopal, 2009; and Madhavi Latha
and Somwanshi, 2009) or Mohr-Coulomb Model ((Madhavi Latha and Rajagopal, 2007; and Han
et al., 2008).
Modeling the geocell and infill as a composite material leads to lesser computational effort than
modeling them separately. However, if the geocell and infill material needs to be modeled
separately, the contact between the surfaces plays a vital role in the model. Hence, studies
conducted to evaluate performance models are summarized in Table 2-4. The outcomes from Table
2-3 and Table 2-4 guided selecting the element type, meshing, contact models, material models.
The complete details of the model developed for this study are discussed in the next chapter,
“Finite Element Modeling.”
The existing pavement design methods using geocell reinforced layers are included in Table 2-5.
Currently, two design methods are available 1) proposed by Pokharel (2009) that is a modification
of Giroud & Han (2004) for Geogrid reinforced pavements, and 2) the second method is proposed
by Presto Geosystems (2008). The design methods are for unpaved roads (no asphalt layer) and
are empirical (based on the laboratory and field test results). A complete analysis of these two
methods is presented in the chapter “Critical Evaluation of Existing Design Methods.”
There are multiple studies performed on geocell by various researchers of Kansas University that
can be considered a torchbearer for studying the usage of geocell in pavements. Hence, a summary
of all the studies performed at Kansas University is summarized in Table 2-6.
7
Table 2-1 Various Studies Performed on Use of Geocells.
Author & Title Objective Application Experiment/Field Findings Compaction Other
Year Testing
Jie Han et al. Performance of To evaluate the To use RAP as A specific number of wheel An increase in stress distribution 4 Ton Infill RAP and
(2011) geocell effect of geocell infill in the passes were applied to angles of 7° and 10° higher than the Compactor. Fine RAP
reinforced RAP reinforcement geocell. evaluate the advantage of unreinforced section was observed. Plate (FRAP)
bases over weak using RAP base geocell reinforcement on This demonstrates that geocell vibrator was
subgrade under courses over weak rut depth. In addition, the reinforcement reduced vertical stress used.
full scale moving subgrade. angle of stress distribution by distributing the load to a wider
wheel loads from the surface to the area.
subgrade interface was
evaluated along with rut
depth.
Ansgar The use of To evaluate the Large-scale static Field Application: Geocell
• The load-carrying capacity Vibration Infill material
Emersleben geocell in road influence of the load tests were were placed within the increased with increasing cell Plate sand with a
& Norbert construction over geocell layer on the carried out to gravel base layers of two height and decreasing cell Compactor maximum size
Meyer (2008) soft soil: vertical load-deformation evaluate the different asphalt paved diameter. is 0.075 in.
stress and FWD behavior of the soil influence of a road constructions. Vertical
• The load-carrying capacity Pressure cells:
measurements geocell layer on stresses on the subgrade improved up to 1.5 times due to the Placed at 13.8
the load- were measured using earth reinforcement of dry sand. in. away from
deformation pressure cells. FWD • A stress reduction between 30% load plate. The
behavior of the measurements were also and 36% was observed. distance
soil. conducted. • During in-situ testing, the average between
stress reduction on the subgrade pressure cells
layer was approximately 30%. is 6 in.
• FWD tests were also conducted,
and results suggested a decrease in
deflections up to 15% and an
increase in layer modulus of about
10%.
Pokharel et Behavior of To understand the Static and Evaluated Influence of • The stiffness of the reinforced sand Not River sand
al. (2009) geocell behavior of geocell repeated loads geocell confinement using was approximately1.5 times higher Mentioned
reinforced reinforced bases were applied on single geocell than that of the unreinforced sand.
granular bases under static and reinforced and • Single geocell reinforcement
under static and repeated loads by unreinforced base increased the maximum failure load
repeated loads using a single layers. The by two times from that of the
geocell stiffness of layers unreinforced sand.
for the static load
8
Year Testing
was calculated. • Single geocell reinforcement

Elastic reduced plastic deformation and
deformation and increased the percent of elastic
plastic deformation under repeated loading.
deformations were It took 10 cycles to reach 80% or
calculated per more of elastic deformation. The
each cycle of elastic deformation reached 95% of
repeated loading. the total deformation at the end of
150 loading cycles.
Pokharel et Experimental To understand the Static and Evaluated Influence of • Under static loading, the Not River sand &
al. (2009) study on bearing behavior of geocell repeated loads geocell confinement using improvement factor (ratio of the Mentioned quarry waste
capacity of reinforced bases were applied on single geocell slope of an initial portion of the
geocell under static and reinforced and load-displacement curve for
reinforced bases repeated loads by unreinforced base reinforced and unreinforced base)
using a single layers. for geocell-reinforced sand was 1.75
geocell in terms of ultimate bearing capacity
The layer stiffness and 1.5 in terms of stiffness.
due to static load
• Permanent deformation reduced to
was calculated. 1.5 times of unreinforced base
Elastic and plastic (quarry waste).
deformations were
• Reinforced quarry waste had a
calculated per
higher percentage of elastic
each cycle of
deformation. At the same time,
repeated loading.
reinforcing river sand had a lower
percentage of elastic deformation
compared to unreinforced sand. The
inferior quality of sand was cited as
a reason for higher deformation.
Giroud & Jie Design Method The proposed This design The theoretical study only Proposed a method for calculating NA NA
Han (2004) for geogrid method for method can be (Assumed proportionality base course thickness for a reinforced
reinforced calculating the used for geogrid between rut depth and unpaved road is calculated using a
unpaved Roads. thickness of the reinforced bearing capacity different equation.
I. Development base layer of unpaved roads. mobilization factor.
of design Method unpaved roads. Tension membrane effect
is not considered.)
9
Year Testing
Giroud & Jie Design Method Calibrate the This design The theoretical study only • The calibrated design proposed in
Han (2004) for geogrid design method method can be (Data for calibrating the the above paper using field and
reinforced using field wheel used for geogrid models was taken from laboratory data.
unpaved roads. load tests and reinforced other researchers. Details • Analyzed test data for three case
II. Calibration laboratory cyclic unpaved roads. of data were not discussed studies.
and Applications plate loading tests in this paper)
on an unreinforced
and reinforced base
layer.
Thakur et al. Creep Permanent Using RAP as a Evaluated Influence of • Confinement of RAP significantly The material Infill RAP
(2011) deformation of deformation or base layer with geocell confinement using increased its strength. in geocell is
unreinforced and rutting due to creep geocell single geocell • NPA geocell significantly reduced compacted
geocell deformation is one reinforcement the initial deformation and the rate in three
reinforced of the concerns of of creep of the RAP. layers 2 in.,
recycled asphalt RAP usage in base 2 in., 0.8
pavements courses. in—95% of
Confinement due to MDD.
Geocell can reduce
the creep.
Imad L.Al Field evaluation Analysis of a Various pavement FWD tests were performed • In sections where 4” thick geocell Material Infill subbase
Qadi & John of geocell use in pavement sections were periodically for up to 3 were used, the resilient modulus backfilled material.
J Hughes flexible performance constructed using years. FWD measurements of the aggregate layer increased using
(2000) pavements constructed with geocell, geogrids, were used to calculate the almost twofold due to the material backhoe and
geocell and and geosynthetics surface modulus and back- confinement. movement
geosynthetics on on a high-traffic calculate the resilient • The aggregate confinement of
the weak subgrade. road. modulus of the subgrade provided by the geocell and the equipment
based on known subgrade-subbase separation over geocell
thicknesses and reasonably provided by the geotextile was
assumed resilient moduli of improved the performance of discussed
pavement layers based on pavement constructed on a weak
material testing and field subgrade for a heavily trafficked
experience. pavement.
• In this study combination of
geocell, geogrid, and geosynthetic
were used. So, specific benefit
due to geocell was difficult to
evaluate.
10
Year Testing
Thakur et al. Creep Rutting due to Using of RAP as a Three laboratory tests were • Confinement of RAP significantly The material NPA geocell
(2011) deformation of creep deformation base layer with conducted in a test box increased its strength. in geocell Used
unreinforced and is one of the geocell (100 in. by 100 in. by 6 in. • NPA geocell significantly reduced compacted
geocell concerns of RAP reinforcement high) to investigate the role the initial deformation and the in three
reinforced usage in base of lateral confinement in rate of creep of the RAP. layers 2in., 2
recycled asphalt courses. reducing creep deformation in., 0.8 in—
pavements Confinement due to of RAP. Creep tests were 95% of
Geocell can reduce conducted at a room MDD.
the creep temperature of 77°C. Each Compaction
deformation of test lasted for 7 to 10 days. method not
RAP. mentioned.
Pokharel et Accelerated The objective was Assess the Laboratory tests on 20 ft. • Reinforced geocell with Quarry Vibratory Infill
al. (2009) pavement testing to evaluate the influence of infill by 16 ft. by 5.9 ft. waste performed poorly. compactor Materials;
of geocell effectiveness of materials on pavement trial area divided •Geocell with aggregate as infill Crushed
reinforced geocells as geocell into four sections. performed better. Limestone,
unpaved roads reinforcement for performance Analyzed four different •RAP as infill performance is better Quarry Waste,
over weak granular base types of pavement sections. than aggregate and RAP
subgrade courses over weak •Welds of geocell were broken under
subgrade. wheel loads.
• Geocell layer reduced the vertical
stress by dispersing the load to a
wide area (12” regular base ≈ 6.5”
Geocell layer)
Pokharel et Investigation of This study To understand the Repeatability of test • Elastic modulus of geocell is an River sand Infill
al. (2010) factors experimentally behavior of method was verified in the essential factor compacted Materials:
influencing the investigated the geocell reinforced study and later confirmed • Unconfined geocell had a lower to 70% of 1) River Sand
behavior of factors influencing bases under static the test set up yields stiffness but a higher ultimate load relative 2) Quarry
single geocell the behavior and repeated loads similar results (repeated capacity than the confined geocell density and Waste
reinforced bases (stiffness and by using a single test) • Performance of geocell depends on quarry waste
under static bearing capacity) geocell infill material. Cohesionless infill to 95%
loading of single geocell- material behaved well than cohesive MDD.
reinforced bases, • Single geocell reinforce base had a
including shape, lower stiffness and bearing capacity
type, embedment, than the multiple geocell reinforced
the height of bases.
11
Year Testing
geocells, and • A thinner unreinforced or geocell-

quality of infill reinforced base on a firm subgrade
materials. had a higher bearing capacity than
the thicker unreinforced or geocell-
reinforced base, respectively.
Emersleben Bearing capacity To evaluate the Model tests were Prepared and tested geocell • The decrease in cell diameter Sand infill
& Meyer improvement of influence of a performed in the using a box size of 6.6 ft3. reduced rutting and vertical stress.
(2008) gravel base geocell layer on the laboratory with Different geocell heights • Increase in cell height reduced
layers in road load-deformation various geocell were tested (4”, 6”, and rutting and vertical stress.
constructions behavior of soil. parameters and 8”). Different Geocell • Geocell layer increased the bearing
using Geocells Geocell made with applied to a trial diameters were tested capacity by reducing 30% or more
different materials, section in the (6.3”, 9”, and 12”). Earth vertical stress on to subgrade.
with various cell field. Pressure cells with a • The deflections on asphalt surface
heights and diameter of 2” were used to were reduced by 15%, and back-
different cell measure the vertical earth calculated layer modulus increased
diameters. pressure on the subgrade. by 10%
Pressure cells were placed
within the sand.
Tanyu et al. Laboratory Strain gauges (bonded The presence of geocells reduced the 90% of Subbase
(2013) evaluation of metallic foil) were used to elastic deflection of the working relative material.
geocell measure strains of geocell. platforms by 30-50%, improved the compaction
reinforced gravel Vertical deflections in the resilient modulus of the subbase by is based on a
subbase over test pit were measures 40-50%, and the modulus of subgrade standard
poor subgrades using position transducers. reaction by more than two times. proctor. The
CR9000 data logger was material in
used to acquire data from the geocell
position transducers and compacted
strain gauges. with the
vibratory
compactor.
12
Year Testing
Ofer Jieft et Modulus Geocell improves The performance The modulus improvement The MIF value obtained from the field Granular Sub-
al. (2011) improvement the layer modulus of geocell was factor (MIF) verified in test, laboratory test, and finite element base
factor for geocell and reduces layer evaluated for an multiple research projects, studies is 2.75.
reinforced bases thickness unpaved trial and field demos provide a
pavement reliable method for
subjected to traffic quantifying the NPA
for an industrial geocell contribution to the
access road in pavement structure for use
India. The in the design of unpaved
performance of and paved roads and
pavement after railways.
nine months of
traffic (one
monsoon) is
evaluated.
13
Table 2-2 Test Setups Developed for Evaluation of Geocell.
Author Title Specimen Geocell Size Layer Thickness Load Type Load Magnitude and Loading Test Parameters
and Year Size shape Plate and Means
Jie Han et Performance Test Geocell used Seven different sections Tests were carried 42’-long reaction NA Rut Depth, Vertical
al. (2011) of geocell section were in three 1) 12” RAP over Subgrade out for 15000-wheel frame and an 18 KN Stress on subgrade,
reinforced size was heights 3”, 4”, 2) 16” geocell reinforced RAP passes. Tests stopped single axle load with stress distribution
RAP bases 20’ by 16’ and 6” high. 3) 4” geocell reinforced RAP for a rut depth of 3.” dual tires Wheel angle, maximum
over weak by 5.9’. Single Geocell 4) 2 layers of 4” geocell Load—Tire Pressure recorded tensile
subgrade Divided dimensions reinforced RAP separated by 80 psi. The frequency strain
under full into four 8.2” long and 1.2” RAP cover. The top cover of the moving wheel
scale moving sections 10” wide is 2.8” thick was 0.167 Hz (i.e., 6
wheel loads 5) 10” unreinforced FRAP s/Pass), and the
6) 1 layer 4” geocell wheels were run at a
reinforced FRAP over 4” speed of 11.3 km/h
unreinforced FRAP. 2” FRAP within the test pit.
cover.
7) 3” geocell reinforced
FRAP. 4” unreinforced FRAP.
3” FRAP Cover
Ansgar The use of 6.6’ x 6.6’ 1) HDPE 8.3” Laboratory Set Up: Glyben Static Load. Vertical load up to 12” Lab Set-Up: Load
Emersleben geocell in x 6.6’ long and 9.8” used as a subgrade (4” thick). Field Test: Heavy 150KN. Circular Steel (dia) Carrying Factor:
& Norbert road wide. Heights It is compacted as 10cms Truck plate used for Ratio of footing
Meyer construction 4”, 6” and 8”. layers. The non-woven applying load. pressure between
(2008) over soft soil: Cell walls material is used to separate the Field Setup: Heavy reinforced and
vertical stress perforated with subgrade and infill material. Truck with five axles unreinforced soil for
and FWD 0.4” diameter Field Set-Up: 6” Gravel layer and a weight of the same settlement;
measurements holes. over subgrade. 8” geocell thick approximately 41 tons Vertical Stresses.
2) Thermally layers. With little cover and a crossed the road at Field Set-Up:
solidified non- 4” Asphalt layer different speeds. Stresses on subgrade
woven were measured
geosynthetic. during vehicle
With diameters passes. FWD
6.3”, 8.7” and measurements were
11.8”. Height taken after the road
8”. was subjected to
traffic.
14
Pokharel et Experimental 2’ x 2’ in Single-cell 4” Filled sand in geocell in three Static and repeated Static Load: Applied 6” (dia) The stiffness of the
al. (2009) study on plan height, 8” layers. Two 2” and one 0.75” load until geocell unreinforced and
bearing diameter with layer. 70% relative density is reinforced sand failed reinforced sands
capacity of two the compaction. In case quarry (72.5 psi). using pressure
geocell perforations of waste as fill material Repeated Load: The displacement
reinforced 0.155 in.2 each compacted to 95% of MDD repeated load test was curves.
bases on both sides. only conducted on the Repeated Load: %
NPA geocell reinforced sand at an Elastic
applied pressure of 50 Deformation
psi (70% of static (Elastic
failed pressure). 1 displacement to the
cycle/minute. 150 total displacement
Cycles. On quarry induced by each
waste, the cyclic load load cycle)
is 80 psi
Thakur et Creep 2’ x 2’x 4” high geocell 4.75” and 4” high geocell with Static Creep Behavior 6” (dia) Deformations in two
al. (2011) deformation 0.5’ 1” cover. No subgrade was Analysis: 40 psi on perpendicular
of used in this study. reinforced & transverse directions
unreinforced unreinforced for ten were measured with
and geocell days. For Comparison three digital dial
reinforced purposes, wholly gauges mounted on
recycled confined RAP (RAP the loading plate,
asphalt compacted in a and averages of
pavements compaction mold) was three were used for
also tested for the calculation. (5
same pressure seven Minute Interval)
days and measured
axial strains with time.
Pokharel et Accelerated 20’ X 16’ 6.7” thick Four pavement sections were Single axle dual tire- Tire pressure applied 21.6” Rut depths for
al. (2009) pavement X 6’ geocell. tested. wheel loading 80 psi. Wheels are width of specific wheel
testing of (Divided 6” geocell + 1) Unreinforced base 12” running at 7 mph with dual tire passes and the angle
geocell into four 0.75” cover aggregate layer over weak a frequency of 6s per of stress distribution
reinforced sections subgrade. pass. Tests terminated from the surface to
unpaved roads for 4 2) 3 sections had 6.7” NPA after 305 passes as the the base course-
over weak analyzing geocell reinforced sections rut depth in one subgrade interface
subgrade 4 over the weak subgrade section is more than using pressure cells.
sections) 5”.
15
Emersleben Bearing Lab Set-Up: 3.2’ thick
6.6’ x 6.6’ 4”,6”, and 8” Lab test: Static Load. The load was applied 12” Vertical pressure on
& Meyer capacity x 6.6’ geocell height.
subgrade (Glyben), 4”,6”, and Field Test: A heavy until contact pressure the subgrade, FWD
(2008) improvement 6.4”, 9.2”, and
8” high geocell were used. truck with five axles. of 72.5 psi for a lab layer modulus, and
of gravel base 12” diameterField Sections: Section 1: 7” setup. thickness
layers in road geocell. asphalt layer, 8” geocell, 6” For field testing, the
constructions gravel; 2) 7” asphalt truck weight was
using geocell pavement, 16 “gravel base, approximately 41
and subgrade; tons. Speed 25 mph.
3) 7” asphalt pavement, 28” FWD: Performed after
gravel. pavement being
subject to traffic (age
of pavement not
mentioned)
Tanyu et Laboratory 9.8’ x 9.8’ HDPE geocell Geocell reinforced section 9” Two types of load 1) Construction load 10” Strains in Geocell.
al. (2013) evaluation of x 11.5.’ diameter 8” & & 18” thick, unreinforced 9” were applied to 7.9 kips for 1,000 Deflection of the
geocell 12”. Height 6” and 18” thick. Subgrade 18” simulate cycles (based on layers.
reinforced and 8”. thick (EPS was used to 1) Construction typical truckload
gravel simulate poor subgrade). equipment traffic during road
subbase over Subgrade was laid over 8’ load expected on constructions in
poor thick soil. working platforms Wisconsin).
subgrades during the 2) The magnitude of
construction phase. the second load was
2)traffic load on selected at 1.6 kips
subbase after the and was applied for
pavement had been 10,000 cycles. (load
constructed and the applied for 0.1s
pavement system was followed by 0.9s rest
opened to service period). (Reduced
(traffic phase) load is due to reduced
traffic load over the
subbase).
3) The maximum
applied stress for the
first loading condition
was 50 psi and varied
up to11.6 psi
16
Table 2-3 Numerical Modeling of Geocell-Reinforced Layer.
Author and Title Model Software Geocell Model & Infill Soil Model Mesh
Year Program
Evan et al. Geocell mattress effects on the Geocell reinforced sand on SSTIPG/2-D Duncan-Chang model, equivalent
(1994) embankment settlements, vertical and top of soft subgrade linearly elastic planar reinforcement
horizontal deformations of supporting an embankment
foundations and embankments: load
settlement
Mhaiskar and Investigations on soft clay subgrade Geocell reinforced sand on ANSYS/3-D Geocell reinforced soil was modeled as a
Mandal (1996) strengthening using geocells top of clay subgrade, composite material using Drucker-Prager
supporting a rectangular Model
footing
Bathurst and Analysis of geocell reinforced soil Geocell reinforced sand GEOFEM/2-D Geocell reinforced soil was modeled as a
Knight (1998) covers over large span conduits over a steel conduit composite material using Duncan-Chang
Model
Madhavi Latha Parametric finite element analyses of Geocell reinforced sand on GEOFEM/2-D Geocell reinforced soil was modeled as a 3 Node triangles
and Rajagopal Geocell Supported Embankment top of clay subgrade composite material using the Mohr- within each
(2007) supporting an embankment Coulomb model rectangle
load
Han et al. The behavior of geocell reinforced Single-cell reinforced sand FLAC/3-D Mohr-Coulomb Model/Linearly elastic
(2008) sand under a vertical load supporting the rectangular membrane
footing
Madhavi Latha Equivalent continuum simulations of Geocell reinforced sand GEOFEM/2-D Geocell reinforced soil was modeled as a
and Rajagopal geocell reinforced sand beds supporting a strip footing composite material using Duncan-Chang
(2008) supporting strip footings Model
Madhavi Latha Numerical simulation of the behavior Geocell reinforced sand GEOFEM/2-D Geocell reinforced soil was modeled as a
and Rajagopal of geocell reinforced sand in supporting a strip footing composite material using Duncan-Chang
(2009) foundations Model
Madhavi Latha Effect of reinforcement form on the Geocell reinforced sand FLAC/3-D Geocell reinforced soil was modeled as a
and Somwanshi bearing capacity of square footings supporting a square footing (Fast composite material using Duncan-Chang
(2009) on sand Lagrangian Model
Analysis of
Continua)
17
Author and Title Model Software Geocell Model & Infill Soil Model Mesh
Year Program
Xiaoming Yang Numerical analyses of geocell Characterize the FLAC-3D Infill material is modeled using Duncan The geocell
(2010) reinforced granular soils under static performance of the geocell (finite Chang Model; geocell was modeled pockets were
and repeated loads reinforced soil under static difference using a linear elastic plate model. A modeled in a
and repeated loads program) Mechanistic empirical model was diamond shape
developed for Geocell reinforced soil
under repeated loads based on stress-
dependent response model in MEPDG
Mehdipour et Numerical study on the stability of Behavior of geocell FLAC-2D The Young's modulus of geocell encased
al. (2013) geocell reinforced slopes by reinforced slopes soil was obtained from the elastic
considering the bending effect modulus of the unreinforced soil and the
tensile modulus of the geocell
reinforcement using an empirical
equation proposed by
Madhavilatha(2007). The interface shear
stress-strain relationship between the
geocell and the foundation soil was
modeled based on the Mohr-Coulomb
sliding criterion
18
Table 2-4 Finite Element Modeling Geocell Reinforced Layer and Contact Models.
Title Objectives Findings Contact Model Geocell Model & Infill Soil Model
Numerical Modeling of Perform a parametric study • Geocell confinement was very The interaction between the Infill ballast was modeled as non-
behavior of railway to investigate the effects of effective in reducing vertical surrounding ballast/sub-ballast and associative elastic-plastic material,
ballasted structure with geocell confinement on deformations, primarily when the geocell was modeled with obeying 3D Drucker Prager Yield
geocell confinement / ballasted embankments the low-quality material was contact elements having "hard” Criterion. The foundation was
Ben Leshchinsky, Hoe I when encountering a soft used. normal contact (no penetration), modeled as an elastic material to
Ling (2013) subgrade, weaker ballast, or • Geocell assists in redistributing and tangential contact was modeled demonstrate the effects of a
varying reinforcement the stresses more evenly, as 2/3 of the tangent of the friction compressible, soft soil without
stiffnesses. possibly preventing the angle (45), which was applied using considering any time-dependent
development of high shear penalty friction algorithm. behavior.
strains and failure, especially The geocell was modeled as an
on softer subgrades. elastic material. The shape of the
• Lateral spreading along the geocell was modeled with a
slope of the railroad rhomboidal shape instead of the
substructure was greatly actual pseudo sinusoidal shape used
reduced. in the tests. This prevented meshing
• More uniform subgrade stress issues that could occur due to the
distribution. Also, the complex nature of the mesh under
magnitudes of stresses were 3D configurations.
reduced significantly, in turn
mobilizing more of the
subgrade's shear strength and
preventing shear failure.
3-Dimensional This paper presents a more It was found that the geocells The geocell and the soil interfaces The elastic-perfectly plastic Mohr-
numerical modeling of realistic modeling approach distribute the load laterally and to were linearly modeled with Mohr- Coulomb model was used to
geocell reinforced sand to model geocell in the 3D a relatively shallow depth as Coulomb Yield Criterion (FLAC simulate the behavior of the
beds/ framework. 3D simulations compared to the unreinforced 3D). foundation and the infill soil. The
A Hegde and were performed by the case and the geogrid reinforced geocell was modeled using the
T.G.Sitharam (2015) actual 3D honeycomb case. Therefore, the performance geogrid structural element. The
shape of geocell using the of the foundation bed was directly Linear elastic model was used to
FLAC3D. influenced by the modulus and simulate the behavior of the
the height of the geocell. geocell.
19
Numerical study on the Behavior of geocell The interface shear stress-strain
stability of geocell reinforced slopes relationship between the geocell
reinforced slopes by and the foundation soils was
considering the bending modeled based on the Mohr-
effect/ Coulomb sliding criterion (FLAC
Iman Mehdipour, 2D)
Mahmoud Ghazavi,
Reza Ziaie Moayed
(2013)
Joint Strength and Wall The current study discusses The results of the experimental The interface behavior of the Elastic perfectly plastic Mohr-
Deformation the joint strength and the study revealed that the geogrid element used in FLAC3D Coulomb criterion was used to
Characteristics of a wall deformation deformation of the geocell wall can be numerically represented at model the behavior of the soil.
Single-cell Geocell characteristics of a single decrease with the increase in the each geogrid node by a rigid Geocell material was modeled
Subjected to Uniaxial cell when subjected to friction angle of the infill attachment in the normal direction using the simplistic linear elastic
Compress/ A Hegde and uniaxial compression. material. and a spring slider in the tangent model.
T.G.Sitharam (2014) The experiment and the numerical plane to the geogrid surface. The The shape of the Geocell is
results were found to be in good orientation of the spring slider modeled using the quarter
agreement with each other. changes concerning relative shear symmetry, assuming the shape of
A simple analytical model based displacement between the geogrid the geocell is cylindrical.
on the theory of thin cylinders is and the soil.
also proposed to calculate the The shear behavior of the geogrid
accumulated strain of the geocell soil interface is cohesive and
wall. frictional and controlled by the
parameters, namely, interface shear
modulus, interface cohesive
strength, and interface friction
angle.
A partially rigid interface with
interface coefficient R(inter)=0.7
was assumed between the soil and
the geocell material. The interface
coefficient relates the strength of
the soil to the strength of the
interfaces.
(FLAC 3D)
20
Accelerated Pavement To test a geocell design Trial sections with a 50 mm Geocells were meshed doubly The base material was modeled
Testing of Low Volume with different infill HMA layer reached the failure curved thin or thick shell. An with Mohr-Coulomb plasticity.
Paved Roads with materials and a thin HMA criteria of 12.5 mm rut depth after embedded region was used to place HMA layers were considered linear
Geocell Reinforcement. layer under simulated full- 10,000 passes due to excessive the geocells in the base layer. elastic. Geocell as linear elastic.
Brandon Bortz, scale traffic on a marginal stress in the subgrade. Embedded regions are a group of
Mustaque Hossain subgrade, using accelerated The redesigned sections with 100 elements that are within a "host"
(2015) pavement testing (APT). mm HMA layer carried 1.2 region.
To develop a finite element million passes without reaching The response of the host elements
model for the geocell 12.5 mm failure rut depth. constrained embedded elements;
reinforced paved roads The geocells with marginal therefore, no contact friction could
considering the quality of materials as infills appear to be be attributed to the geocell wall.
the infill material to study viable in low volume paved road (Abaqus)
the design of such applications.
pavements.
Numerical analyses of Characterize the Mohr-Coulomb yield criterion Infill material is modeled using
geocell reinforced performance of the geocell (FLAC 3D) Duncan Chang Model; geocell was
granular soils under reinforced soil under static modeled using a linear elastic plate
static and repeated loads and repeated loads model. A Mechanistic empirical
Xiaoming Yang (2010) model was developed for Geocell
reinforced soil under repeated loads
based on stress-dependent response
model in MEPDG
21
Table 2-5 Existing Pavement Design Methods of Geocell-Reinforced Layers
Author Title Properties Model Used Modification Factors Design Thickness Approach
and Year Measured
Jie Han et Performance of Rutting and NPA geocell reinforcement increased
al. (2011) geocell reinforced stress the life of unpaved roads compared with
RAP bases over distribution an unreinforced section by 1.8 times
weak subgrade under angle with one layer of 10 cm high geocell.
full scale moving Geocell higher than 10 cm may be
wheel loads beneficial, but it makes compaction
more difficult.
Pokharel Behavior of geocell
Stiffness of The high percent of elastic deformation
et al. reinforced granular
reinforced layer is beneficial to the service life of the
(2009) bases under static
and % Elastic road.
and repeated loads
deformation in
total
deformation
Giroud & Design Method for Distribution of Design Method 1)Base course is characterized Developed an 1) Assumed the bearing capacity factor
Jie Han Geogrid Reinforced stress, the was developed by its CBR. Resilient modulus equation to 3.14 for unreinforced bases and 5.71 for
(2004) Unpaved Roads. strength of base based on is estimated by using CBR. calculate the reinforced bases with geogrid, whereas
1. Development of course material, 1) stresses at the 2) Subgrade characterized by thickness based on 5.14 for geotextile reinforced unpaved
Design Method the interlock base CBR & Undrained cohesion. normal stress pi at roads.
between course/subgrade Resilient modulus is estimated the interface 3) Proposed an equation for calculating
geosynthetic and soil interface. by using CBR. between base bearing capacity mobilization
base course 2) determining 3) Geogrid is characterized by course and coefficient. The equation has three
material, the rut depth as Aperture Stability modulus. subgrade soil unknown parameters that need to
geosynthetic in- a function of the combined with calibrate with experimental data.
plane stiffness. stresses at the subgrade soil 4) Explained that the geogrid properties
base needs with three will influence the stress distribution
course/subgrade unknowns 1) angle. Higher aperture stability modulus
soil interface bearing capacity increases the distribution angle. As well
and the bearing mobilization as the thickness of the base layer has a
capacity of the coefficient, 2) positive influence on the distribution
subgrade soil. bearing capacity angle.
factor, 3)
distribution angle.
Pokharel Accelerated Rut depth and Calculated the stress distribution angle
et al. pavement testing of stress under wheel loads for various sections.
(2009) geocell reinforced distribution The calculated stress angle can be used
angle to calibrate the design.
22
Author Title Properties Model Used Modification Factors Design Thickness Approach
and Year Measured
unpaved roads over
weak subgrade
Presto Design method for Maximum Presto Presto Geosystems The design is based on the theory that
Geosystem low volume unpaved allowable stress Geosystems recommended peak friction the geocell reinforced layer absorbs a
roads with presto geo on the subgrade recommended angle ratio based on the various portion of vertical load, thus reducing
reinforcement. peak friction geocell texture types and infill the ultimate load on the subgrade. Due
angle ratio based to the load transfer to geocell, the
on the various pavement section can either carry higher
geocell texture loads or have an extended life than the
types and infill un-reinforced pavement section. The
amount of stress absorbed is given in the
equation below.
𝐻𝐻
𝜎𝜎𝑟𝑟 = 2 � � 𝜎𝜎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡
𝐷𝐷
Where, H = Geocell Height; D =
Effective Geocell Diameter
δ = Angle of shearing resistance
between granular infill material and
Geocell walls; δ = rΦ;
23
Table 2-6 Summary of Research Performed at Kansas State University
Parameters Static Plate Loading Tests Cycle Plate Loading Tests Full-Scale Moving Wheel Tests
Sample size 24 in. by 24 in. 7.2 ft. by 6.6 ft. by 6.6 ft. 20 ft. X 16 ft. (Often divided into four different
sections)
800 mm X 800mm 23.8 in. by 23.8 in.
Infill Materials River Sand, Quarry Waste, RAP, Base aggregates River Sand, Quarry Waste, RAP, Base aggregates River Sand, Quarry Waste, RAP, Base aggregates
Subgrade No Subgrade for single geocell testing. Using Weak Subgrade Using Weak Subgrade
The remaining tests performed on poor subgrade soils
Load Maximum Capacity 130 psi Maximum Capacity 25,000 kips 18 kips single axle with dual tires with tire
pressure 80 psi
Plate size 6 in. 6 in. by 12 in. 21.7 in. width of dual tire
Maximum load 130 psi on Quarry waste Peak force of 9 kips and a trough force 0.1 kips wave Up to 15,000 wheel passes
applied frequency 0.77Hz
Other loads 36, 70, 115 psi The repeated load test was only conducted on the reinforced Few studies terminated at a lower number of
applied sand at an applied pressure of 50 psi (70% of static failed cycles (305) for unpaved trial sections
pressure). 1 cycle/minute. 150 Cycles
Geocell Heights 4 in., 4.7 in. 6 in., 9 in., and 12 in. 6.6 in., 9 in., 10 in., and 12 in.
Parameters Vertical stress, deformations at each load at every five- Vertical stress strains on geocell, deformations
Tested minute interval until failure of the test section
Conclusions Stiffness improvement Factor (Ratio of the slope of an Elastic Deformation (Elastic displacement to the total Rut depths for a specific number of passes of
initial portion of the vertical stress-displacement curve for displacement induced by each load cycle) wheel load and the angle of stress distribution
the geocell confined base to that of the unreinforced base) from the surface to the base course-subgrade
interface. Pressure cells were placed at the
interface of the subgrade and base course.
The stiffness factor increased from 1.2 to 2.0 (Quarry Reduced permanent deformation by reducing the vertical stress Geocell reduced the rut depth and vertical stresses
waste as infill material is weak due to high percentage of at the interface of base and subgrade and increasing the transferred to the subgrade by distributing the load
fines) elasticity of RAP bases. over a wider area.
Bearing capacity increased to 1.9 to 3.2 times. The permanent deformation of 3 in., the ratio of loading cycles Sufficient cover up to 2 in. To 3 in. thick was
for the reinforced section to that for the unreinforced section necessary to minimize damage to geocell under
was approximately 10. trafficking.
Creep Tests were performed on infill material RAP. The
amount and rate of creep deformation decreased with
confinement by geocell. Tests conducted at 40 psi or 80 psi
(7 days & 10 days)
24
3. FINITE ELEMENT MODELING (FEM) AND ANALYSIS
OF RESULTS
This chapter provides information relevant to the 3-D FEM developed to analyze geocell-
reinforced pavement structures for studying the behavior of the geocell-reinforced pavement
structure subjected to repeated loading simulating laboratory testing and traffic loading conditions.
A summary of selected model components is included herein, while the details relevant to the
model and selection of each component are included in Appendix B.
Although FE analyses (FEA) can identify the level of reinforcement provided by the geocell, the
generation of a mesh for FEA is complicated due to several factors like the interaction between
geocell and adjacent soil, transfer of load, and confinement provided by the geocell, among others.
Additionally, the modeling of geocell required a significant number of elements and nodes to
model the honeycomb shape of geocell, which requires significant computational time. Therefore,
to develop a 3-D FE model (FEM) that better addresses the needs imposed by the characteristics
of the geocell-reinforced pavement, distinctive FEMs with different levels of sophistication were
developed before the development of the final 3-D model. These models were developed to
evaluate the following aspects:
• Soil material model
• Boundary conditions of reinforced-layer
• Shape of geocell
• Shell element type
• Geocell-soil interaction
To perform FEA, a general-purpose finite element program LS-DYNA was selected because this
program allows dynamic FEA and includes a comprehensive list of material and contact
models/algorithms. Moreover, the program can also be installed on the High-Performance Cluster
(HPC), a computer system that groups class Linux clusters and symmetric shared-memory
multiprocessor systems that significantly improve simulation program speed performance. The
HPC allows executing parallel programs or multiple instances of the same program, each driven
by a different parameter set. Based on the analysis detailed in Appendix B, the developed model
used for the parametric analysis is summarized in Table 3-1.
3.1 PARAMETRIC STUDY
The following five parameters were selected for evaluating the influence of geocell reinforced
layer on pavement performance:
1) cover thickness (layer on top of the geocell reinforced layer),
2) the thickness of the geocell reinforced layer,
3) modulus of infill material,
4) subgrade modulus, and
5) modulus of the cover material (high and low modulus material on top of the geocell
reinforced layer).
25
Table 3-2 displays the plan for performing the parametric study. It also shows the properties of the
materials, layer, and cover thicknesses. For comparing the effectiveness of geocell, all the cases
shown in Table 3-2 are also performed with no geocell. Base 1 and Base 2 were selected as per the
laboratory evaluation plan discussed in Section 4. Base 3 and Base 4 used in FEA were considered
part of the parametric study.
Table 3-1 Dimension and Properties of Geocell-Reinforced Pavement FE Model with
Geocell Panel Simulated Using Pseudo-Sinusoidal Shaped Cells.
Pavement Structure Thickness
Layers Geocell Reinforced Unreinforced
Pavement Structure Pavement
Structure
Top Base Layer (in.) Varying 1, 2, 3 and 4 Varying 1, 2, 3, and 4
Geocell Reinforced Base (in.) Varying 3, 4 and 6 Varying 3, 4 and 6
Subgrade (in.) 24 24
Finite Element Model Properties (Quarter Model)
Number of Solid Elements 45,344 45,344
Number of Thick Shell Elements (Geocell) 1120 -
Number of Discrete Beams 2568 -
Total Number of Elements 49,032 45,344
Total Number of Nodes 52,547 52,547
Finite Element Model Size (Quarter Model)
Longitudinal Dimension, x-axis (in.) 24
Transversal Dimension, y-axis (in.) 22
The following four pavement performance criteria were used for evaluating the influence of
geocell reinforced layer:
1. vertical stress distribution on subgrade top
2. vertical strain distribution on subgrade top
3. vertical deformation in the top layer (base)
4. hoop strains on geocell
Figure 3-1 displays the typical quarter model used for the study and the various layers in the model.
Each model consists of a cover layer above the geocell reinforced layer and a 24-in. subgrade layer
below the reinforced layer. The legend depicting properties evaluated consists of cover thickness
and modulus followed by geocell height and infill material modulus, and the last legend indicates
subgrade modulus {e.g., COVER4” (12 ksi)_GEOCELL4”(12 ksi)_SUBG(4.5 ksi)}.
Figure 3-2 shows the location of various outputs evaluated in the study. Instead of measuring the
output on a single element, an average of three elements (nearby) was considered. All the model
combinations in the parametric research are run for twenty load cycles using a ten-inch load plate.
Figure 3-3 illustrates the stresses measured on subgrade at six inches from the center of loading
plate for twenty load cycles for both geocell and no geocell model for COVER4” (12 ksi)
_GEOCELL4” (12 ksi)_SUBG(4.5 ksi). It is observed that there is no significant change in stresses
with load cycles.
26
Table 3-2 Parametric Study Plan.
Geocell 3" Geocell 4" Geocell 6"
Base Modulus Base Modulus Base Modulus
Cover Cover High Marginal Low High Marginal Marginal Low High Marginal Marginal
Subgrade Low
Thickness Quality (15 ksi) (12 ksi) (2 ksi) (15 ksi) (12 ksi) (7 ksi) (2 ksi) (15 ksi) (12 ksi) (7 ksi)
(2 ksi)
Same as
A A
Infill
1"
Stiff (30
B, D A, D A, D, B
ksi)
Same as
A A
Infill Good
2"
Stiff (30 (4.5 ksi)
B, D A, D A, D, B
ksi)
Same as
A A
Infill
3"
Stiff (30
B, D A, D A, D, B
ksi)
Same as A, B, C, A, B, A, B, C,
B, E B, E B, E B, C, E C B, C, E C B, C, E
Infill Good D, E C, E D, E
4"
Stiff (30 (4.5 ksi)
D D A
ksi)
Same as
E E E E E, A E E, A, D E
Infill Good
Stiff (30 (4.5 ksi)
6" B, D B, D B, D
ksi)
Same as Poor
A, D A, D
Infill (2 ksi)
A Evaluate the effect of the cover thickness (only for 4" and 6" height geocells, BASE 2 (Marginal) properties)
B Evaluate geocell height (base thickness)
C Evaluate base infill material (modulus)
D Evaluate the influence of low and high modulus cover material
E Evaluate subgrade modulus
27
Figure 3-1 Quarter Model used in the Parametric Study.
Figure 3-2 Locations of Output Evaluated from FEA.
Figure 3-3 Comparison of Compressive Stress at 6” from Loading Center of Loading Plate
(Geocell vs. No Geocell).
28
All the outputs from FEA except hoop strains on geocell are compared with the stresses and strains
obtained from BISAR software. Even though the output from the model is evaluated from the
twenty load cycles, only one load cycle data is employed for comparison between BISAR and
FEA. The influence of the geocell reinforced layer on performance is included in the following
sections.
3.2 INFLUENCE OF BASE MODULUS (BOTH INFILL AND COVER LAYERS WITH
SIMILAR MODULUS VALUES)
The cases mentioned with alphabet ‘C’ of Table 3-2 are compared in this analysis. Eight cases
(four with geocell 4” thick and four with geocell 6”) have been used for comparison purposes. All
the cases have a cover thickness of four in. and the same modulus as the infill material. The
subgrade is twenty-four inches thick with a modulus value of 4.5 ksi.
The four infill materials of varying modulus values (15 ksi, 12 ksi, 7 ksi, and 2 ksi) were evaluated.
For 12 ksi infill modulus material, the stress and strain distribution on the top of the subgrade are
summarized in Figures 3-4 and 3-5, respectively. It is apparent that there is no significant
difference in stresses and strains for geocell and no geocell cases nine inches away from the loading
plate. Similar trends were observed for the infill modulus of 15, 7, and 2ksi. Therefore, the analysis
later focused on up to nine inches from the loading plate.
Figure 3-4 Stress Distribution (Geocell vs. No Geocell) along the Subgrade.
29
Figure 3-5 Strain Distribution (Geocell vs. No Geocell) along the Subgrade.
The FEA results are summarized in Figures 3-6 through 3-9 regarding stresses, strains, vertical
deformation, and hoop strains to identify the influence of base properties. The following
conclusions can be drawn based on the summarized results:
• The stresses, strains, and vertical deformation estimated from FEA (unreinforced or no
geocell layer) are higher than those estimated using BISAR.
• The geocell reinforced layer reduced the stresses up to 20% six inches away from the center
of the loading plate and up to 50% nine inches away from the center of the loading plate
compared with the unreinforced layer (no geocell condition).
• A similar trend was observed in the vertical strains on the subgrade.
• Geocell reinforced layer often produced higher deformation than the no geocell layers. This
trend is like the one observed in the laboratory evaluations as well.
• The hoop strain on the first geocell increased with a lower base modulus. However, a
significant strain reduction is observed in the second geocell and minimal strain on the
third. It indicates that the geocell was more effective with lower modulus base materials.
3.3 INFLUENCE OF COVER THICKNESS
All the cases with alphabet “A” of Table 3-2 are considered in this section. The Geocell with 4”
and 6” height is analyzed along with various cover thicknesses of 1, 2, 3, 4, and 6 inches. The
geocell infill and cover materials have a similar modulus value of 12 ksi. The subgrade layer is
twenty-four inches thick with a modulus value of 4.5 ksi. The FEA analysis results are summarized
in Figures 3-10 through 3-16, and the following conclusions could be drawn from the analysis:
• The stresses, strains, and vertical deformation estimated from finite element analysis (no
geocell) are higher than the BISAR.
• There is a significant reduction in stresses (below loading plate) when the cover thickness
is 3 inches or higher.
30
Figure 3-6 Stress Distribution (Geocell 4” and 6” vs. No Geocell) along the Subgrade (Influence of Infill Modulus).
31
Figure 3-7 Strain Distribution (Geocell 4” and 6” vs. No Geocell) along the Subgrade (Influence of Infill Modulus).
32
Figure 3-8 Vertical Deformation (Geocell 4” and 6” vs. No Geocell) (Influence of Infill Modulus).
Figure 3-9 Hoop Strains on Geocell 4” and 6” (Influence of Infill Modulus).
33
Figure 3-10 Strain Distribution (Geocell 4” and 6” vs. No Geocell) along the Subgrade (Influence of Cover Thickness).
34
Figure 3-11 Vertical Deformation (Geocell 4” and 6” vs. No Geocell) (Influence of Cover Thickness).
Figure 3-12 Hoop Strain on Geocell 4” and 6” vs. No Geocell (Influence of Cover Thickness).
35
Figure 3-13 Stress Distribution (Geocell 3”, 4” and 6” vs. No Geocell) along the Subgrade (Influence of Good Cover and Poor Infill).
36
Figure 3-14 Strain Distribution (Geocell 3”, 4” and 6” vs. No Geocell) along the Subgrade (Influence of Good Cover and Poor
Infill).
37
Figure 3-15 Vertical Deformation (Geocell 3”, 4” and 6” vs. No Geocell) (Influence of Good Cover and Poor Infill).
Figure 3-16 Hoop Strains on Geocell 3”, 4” and 6” vs. No Geocell (Influence of Good Cover and Poor Infill).
38
• The geocell reinforced layer reduced the stresses up to 20% at six inches away from the
center of the loading plate and up to 50% at nine inches away from the center of the loading
plate compared with no geocell layer. Similar trends were observed in the vertical strains
on the subgrade.
• Geocell reinforced layer often produced higher deformation than the no geocell layers.
• The hoop strain on the first geocell decreased with an increase in cover thickness. The
strain reduced in the second and third geocell significantly.
3.4 INFLUENCE OF COVER BASE MATERIAL
All the cases with the alphabet “D” are considered for this study. For this analysis, three geocell
heights (Geocell 3”, 4”, and 6”) and five cover thicknesses (1, 2, 3, 4, and 6 inches) were modeled
along with geocell infill material of 2 ksi and cover layer material of 30 ksi. The subgrade modulus
was maintained at 4.5 ksi with a thickness of twenty-four inches. The results are summarized in
Figure 3-13 through Figure 3-16, and the following conclusions could be drawn from the analysis:
• The stresses, strains, and vertical deformation estimated from finite element analysis (no
geocell) are higher than the BISAR.
• There is a sharp reduction in stresses (below loading plate) when the cover thickness is 3
inches higher.
• The three-inch geocell reinforced layer reduced the stresses up to 10-30% at six inches
away from the center of the loading plate and up to 15-60% at nine inches away from the
center of the loading plate compared with no geocell layer. Similar trends were observed
regarding the vertical strains on the subgrade.
• The four-inch geocell reinforced layer reduced the stresses up to 10-25% at six inches away
from the center of the loading plate and up to 0-30% at nine inches away from the center
of the loading plate compared with no geocell layer. Similar trends were observed
regarding the vertical strains on the subgrade. The performance of geocell diminished with
the use of stiffer base material cover above 4 inches.
• The six-inch geocell reinforced layer reduced the stresses up to 0-25% at six inches away
from the center of the loading plate and up to 0-30% at nine inches away from the center
of the loading plate. Similar trends were observed regarding the vertical strains on the
subgrade. The performance of geocell diminished with the stiffer base material cover above
4 inches.
• Geocell reinforced layer often produced higher vertical deformation than the no geocell
layers.
• The hoop strain on the first geocell decreased with an increase in cover thickness. The
higher the thickness of geocell more hoop strains were observed—however, the strain
reduced in the second and third geocell significantly.
3.5 INFLUENCE OF SUBGRADE MODULUS

All the cases with alphabet “E” were considered for evaluating the influence of subgrade modulus
on geocell height (Geocell 3”, 4”, and 6”) with cover thicknesses of 4 inches. Three geocell infill
and cover materials (2ksi, 12 ksi, 15 ksi) and two subgrade moduli, 4.5 ksi, and 2 ksi were utilized
for this assessment. The results are summarized in Figure 3-17 through Figure 3-28, and the
following conclusions can be drawn:
39
• It is observed that the 3" geocell height has no significant influence on the reduction of
stresses at the top of the subgrade. In contrast, the 4" geocell height reduced stresses up to
30-45% at 9" away from the loading point. The 6" geocell performed better by reducing
the stresses about 10% at 6" away from loading point and around 50% at 9" away from the
loading point.
• The hoop strains on the geocell were minimally influenced by a change in subgrade
modulus (4.5 ksi and 2.0 ksi).
• The vertical strains and deformation for the 2.0 ksi subgrade (for all geocell) are higher
than the 4.5 ksi subgrade.
3.6 INFLUENCE OF GEOCELL LAYER THICKNESS OR GEOCELL HEIGHT
All the cases with the alphabet “B” were considered to evaluate the influence of geocell height.
Three geocell heights (Geocell 3”, 4”, and 6”), three geocell infills, and cover materials (2 ksi, 12
ksi, 15 ksi) were used in this assessment with a cover thickness of 4 in. The results are summarized
in Figure 3-17 through Figure 3-28, and the following conclusions can be drawn from the analysis:
• In terms of geocell height, the geocell 3" performance is inferior to other geocell heights as it
only reduced stresses up to 10% at 6" from the loading plate and 20% at 9" from the loading
plate, whereas geocell 4" and 6" reduced the stresses around 20% at 6" and 40-50% at 9" from
loading plate.
• Geocell 3" has minimal influence in the presence of weak subgrade material.
Based on the overall evaluation, the observed trends have been summarized in Table 3-3, and the
benefits of geocell height are summarized in Table 3-4. The summarized results indicate that 4”
and 6” geocell heights provide the benefit of using geocell. In addition, the benefit of geocell
reduces with an increase in base modulus (same material for infill and cover) as observed by
various researchers and laboratory evaluation results. The test results also indicate that stress on
top of the subgrade should be used to design pavements constructed with geocell.
40
Figure 3-17 Stress Distribution (Geocell 3” vs. No Geocell) along the Subgrade (Influence of Subgrade Modulus).
41
Figure 3-18 Strain Distribution (Geocell 3” vs. No Geocell) along the Subgrade (Influence of Subgrade Modulus).
42
Figure 3-19 Vertical Deformation (Geocell 3” vs. No Geocell) (Influence of Subgrade Modulus).
Figure 3-20 Hoop Strain on Geocell 3” (Influence of Subgrade Modulus).
43
44
45
Figure 3-24 Hoop Strain on Geocell 4” (Influence of Subgrade Modulus).
46
47
48
Figure 3-28 Hoop Strains on Geocell 6” (Influence of Subgrade Modulus).
49
Table 3-3 Observed Performance Trends for Various Parameters
Increase of Good Cover Decrease in Increase in
Cover
Infill and Poor Subgrade Geocell
Thickness
Modulus Infill Modulus Depth
Stress on Subgrade Decrease
Strain on Subgrade Increase

No Clear
Vertical Deformation
Trend
Hoop Strain on geocell
Table 3-4 Performance of Three Geocell for Various Parameters
Subgrade Poor infill material (2 ksi) and good cover

Parameter Infill Modulus (ksi) Cover Thickness (in.) (same as infill)
Modulus (ksi) material (30 ksi)
Above
GEOCELL 2 7 12 15 1" 2" 3" 4" 6" 2 4.5 1" 2" 3" 4" 6"
15
Geocell 3" 7 7 7 7 0 7 7 10 10 10 0 7 10 10 10 7 7
Geocell 4" 10 10 10 7 0 7 7 10 10 10 7 10 10 10 10 7 0
Geocell 6" 10 10 10 10 0 7 7 10 10 10 10 10 10 10 7 7 0
Good Performance 10
Average Performance 7
No Influence 0
50
4. SELECTION OF MATERIAL, EXPERIMENTAL DESIGN,
AND LABORATORY EVALUATION
This chapter provides information regarding the selection of material, experimental design, and
laboratory evaluation process. One of the key objectives of this project is to develop pavement
design inputs for geocell reinforced layers. To accomplish the above objective, the performance
of the geocell reinforced layer is assessed by modeling the geocell reinforced pavement system in
finite element software and comparing the test results obtained from laboratory testing.
Furthermore, cyclic load with a rest period (typically used in resilient modulus evaluation) was
applied in laboratory testing and finite element modeling to simulate traffic loading.
4.1 EXPERIMENTAL DESIGN
Based on published literature, various parameters that contribute to the performance of pavement
reinforced with geocell were identified for evaluation laboratory and FEA analysis (discussed in
chapter 3).
The literature and preliminary investigation identified the following four parameters that influence
performance:
1) Cover thickness (layer over geocell) and quality of cover material,
2) Geocell layer thickness,
3) Infill material (in geocell) modulus, and
4) Subgrade modulus.
Since the focus of this study was on low-volume roads, it was decided not to include an asphalt
concrete surface layer in the analysis. Therefore, the modeling analysis included (bottom to top) a
subgrade layer placed at the bottom followed by a geocell reinforced layer (geocell layer with infill
base material). At the end (at the top), a layer consisting of either poor quality infill or better-
quality base material was placed on top of the geocell infill material and referred to as a cover
layer.
To document the benefit of geocell reinforcement, the analysis was also performed for pavement
systems without geocell reinforcement by placing a layer of similar thickness without geocell.
Although the geocell placement protocol suggests placing fabric between the geocell and subgrade
to avoid contamination, it was not followed because of computational constraints in this study.
Additionally, initial finite element analysis consisted of a loading plate of 10 inches; however, the
design analysis suggested that a 12-in. diameter loading plate would be a better option. Therefore,
additional analysis was performed using a 12 in. plate. Finally, the influence of geocell was
evaluated by measuring the following performance parameters:
1) Vertical stress and strain distribution below the base layer or at the top of subgrade,
2) Vertical deformation below the base layer or at the top of subgrade, and
3) Hoop strains on the side of geocell.
4.2 LABORATORY EVALUATION
Like the finite element analysis, the laboratory evaluation was performed for evaluating the
performance of the geocell reinforced layer with one exception (quality of cover material). The
51
reason for using the same quality material (for geocell infill and cover layer) was to avoid
contamination because we were reusing the material for testing.
The range of parameters selected and the overall laboratory assessment plan are summarized in
Table 4-1. To document the benefit of geocell reinforcement, the analysis was also performed for
pavement systems without geocell reinforcement by placing a layer of similar thickness without
geocell. The initial laboratory test plan consisted of a loading plate of 10 inches; however, the
design analysis suggested that a 12-in. diameter loading plate would be a better option. There are
three reasons that the 12” plate is used in the testing:
• The input modulus in Texas pavement design software FPS 19 or 21 was based on falling
weight deflector (FWD) back-calculated modulus. According to the test procedure
presented in the report “The Falling Weight Deflectometer for Nondestructive Evaluation
of Rigid Pavements (1985) by CTR UT Austin, the diameter of the loading plate is (11.8”).
• The equivalent diameter of each cell in the geocell mattress is around 11” <Geocell
Diameter<12.5”. The preliminary testing is performed with a 10” which fitted into the cell.
Vertical stresses are concentrated in a single geocell rather than distributed across the
mattress. The vertical stress distribution on the subgrade top due to 10” loading plate was
appalling, like higher vertical stresses (compared with unreinforced base) below the center
of loading plate and significant drop moving away from the center of loading. So, the actual
benefit of geocell reinforcement is hard to comprehend. But using a 12” loading plate
(larger than the single geocell diameter opening), the vertical stress on subgrade right
below the loading plate was dropped compared with the unreinforced layer, and a smooth
reduction in vertical stresses away from the center of the loading plate.
• Based on the literature, “Experimental Study on Geocell Reinforced Bases under Static and
Dynamic Loading” (Sanat Kumar Pokharel, 2010) from Kansas in large-scale testing (like
the tank testing and type of geocells used in this study) used a 12” loading plate (for cyclic
loading). In the same report, the preliminary testing (static loading) was performed using a
smaller diameter plate (6” plate), the reason for shifting to higher diameter plate was not
mentioned.
Therefore, additional analysis was performed using 12 in. plate. The influence of geocell was
evaluated by measuring the following performance parameters:
• Vertical stress and strain distribution below the base layer or at the top of subgrade,
• Vertical deformation below the base layer or at the top of the subgrade
• Hoop strains on the side of geocell.
52
Table 4-1 Experimental Test Plan
Tests with Geocell and No Geocell (58)
Tests Not Required
53
4.3 MATERIAL SELECTION
4.3.1 Geocell
Although several geocell manufacturers were contacted, only two (Presto and Tenax) provided
geocell for this study. Presto provided geocell of 4.25” and 6” height geocell, while Tenax provided
3” and 4” height geocell, as summarized in Table 4-2. Although the main difference from the
pavement system point of view is only the height, the geocell has different joints, thickness,
construction, etc. Therefore, the details provided by the manufacturers are included in Appendix
B – Figure B-1 and Figure B-2. Other than height, the manufacturer-provided specifications were
considered in the finite element analysis. In addition, the dimensions of the geocell (width and
length of each geocell opening) and thickness of geocell are measured in the laboratory for
verification purposes.
Table 4-2 Geocell Selection and Properties
Geocell Height Manufacturer Properties
3” and 4” Tenax Appendix C
4.25” and 6” Presto Appendix C
4.3.2 Base and Subgrade Selection

As per the tentative plan of the study, three base materials and two subgrades with fair to inferior
quality are needed. The needed materials were selected from various locations in Texas with the
help of the project management committee.
The base materials were obtained from Dallas (Collin County), San Antonio (La Hoya Quarry),
and El Paso Districts. Similarly, the subgrades were obtained from Paris and El Paso District. To
perform testing in the laboratory, approximately five tons of each material was obtained from
either plant or field. The obtained material was stored in labeled barrels to minimize cross-
contamination and accidental use of the wrong material. The following tests were performed to
measure properties needed for finite element analysis and design of pavement system. The test
procedures followed are included in Table 4-3, while the measured engineering properties are
included in Table 4-4.
Table 4-3 Test Procedures for Evaluation of Base and Subgrade Material
Material Property Test Procedure
Maximum Dry Density (MDD) and Optimum Moisture Content (OMC) Tex-114-E
Particle Size Analysis (Gradation) Tex-110-E
Plastic Limit of Soils Tex-105-E
Liquid Limit of Soils Tex-104-E
Triaxial Compression for Disturbed Soils and Base Materials Tex-117-E
In-Place Density of Soils and Base Materials (Sand Cone Method) Tex-115-E
Resilient Modulus AASHTO T-307
54
Table 4-4 Measured Engineering Properties of Base and Subgrade Materials
Gradation
using the Portable

Resilient Modulus
Seismic Pavement
Elastic Modulus
(KSI) estimated
MDD (PCF)
OMC (%)
Analyzer
Location
Material
UCS, psi
microns (Silt
4.75mm<S<
50mm< G <
(KSI)
Passing 75
0.0075mm
PI
or Clay)
4.75mm
T
FM 545-
Collin
Base 1 County- 125.70 6.10 45.80 53.70 0.50 NP 39.10 11.23 3.00 25 28 15
Dallas
District
La Hoya
Quarry, San
Base 2 130.00 6.00 54.10 28.50 2.50 NP 36.90 2.00 4.30 15 23 12
Antonio
Texas
El Paso (Jobe
Base 3 106.00 16.70 17.00 78.00 5.00 3.00 0.00 8.70 6.90 6 6 5
Plant)
Subgrade El Paso (Jobe
130.00 6.70 32.40 65.20 2.50 NP 33.80 7.90 3.90 24 20 4.5
1 Plant)
Between
Subgrade Commerce
88.00 16.00 0.00 88.70 11.30 8.00 0.00 7.20 6.20 13 Too Soft 2.5
2 and Paris of
Texas
MDD: Maximum Dry Density, OMC: Optimum Moisture Content, PI - Plasticity Index, Φ- Angle of Internal Friction, C – Cohesion
(psi), T = Triaxial Classification, UCS = Unconfined Compressive Strength (psi), G-Gravel, S-Sand.
55
4.4 LABORATORY EVALUATION PROCESS
4.4.1 Laboratory Set-Up

Although the initial plan was to use a cylindrical container for testing, two rectangular tanks of
3′ × 4′ × 5′ were fabricated (Figure 4-1) to simulate the field pavement section in the laboratory.
This tank can accommodate three pavement layers (subgrade, geocell reinforced base, and cover
over geocell reinforced layer) of less than 36 in. Although the quality of the subgrade layer varied,
the thickness of the subgrade layer was maintained at 24 in. Likewise, the thickness of the base
layer was varied depending on the height of the geocell selected. Various laboratory factors are
evaluated in Table 4-5, based on the combination proposed in Table 4-1.
Figure 4-1 Photo of Fabricated Box

Table 4-5 Factors Evaluated in Laboratory
S.no Factor Mechanism Means
1 Cellular Strains developed in Geocell will Strain gauges will be fastened
confinement be observed at various locations of to Geocell, and strains will be
Geocell under loading. recorded.
2 Stress distribution Stresses at various locations on thePressure cells will be placed on
subgrade top will be examined. the subgrade, and stresses will
be recorded.
3 Vertical The settlement of reinforced base The settlements will be
Deformation under loadings will be estimated. documented through LVDT
placed on top of the base.
4 Height of The influence of Geocell Three different geocell heights
reinforced base. reinforced base height on various (3 in., 4 in., and 6 in.) will be
properties will be studied. investigated.
56
4.4.1.1 Sample Preparation
The specimens were prepared by placing and compacting subgrade in layers at the optimum
moisture content and compacted with the help of a vibratory compactor (Figure 4-2).
The required quantity for 24 in. subgrade is sundried and stored in 1-gallon buckets. The soil is
mixed thoroughly with a measured amount of water in a concrete mixer for consistent moisture
distribution. The subgrade is compacted in two 12 in. layers. To further verify moisture
distribution, random samples were collected and tested from the compacted layer. The compaction
is verified by using the sand cone method as per the specification Tex-115E. The subgrade is
compacted to achieve a minimum of 95% of maximum dry density.
Figure 4-2 Sample Preparation Using Vibratory Compactor

After compacting the subgrade layer, trenches for placing pressure cells were dug. After placing
the pressure cells, they are covered with 2 in. of subgrade (100% passing 3/16 in. or 4.75mm) and
manually compacted with a rammer.
After placing the subgrade layer, the base layer is placed following a similar process. The base
layer is placed in two layers; a geocell reinforced layer and a cover. The base required to fill the
geocell is estimated based on the geocell height and the proctor density details. The estimated
material (base material and water) is placed in the geocell layer and compacted carefully. The
density achieved in the geocell is verified by the sand cone method (performed at the box's corners
to avoid damaging the sample near the testing area). The targeted density is 95% of the maximum
dry density. The cover above the geocell reinforced layer is placed similarly. The elastic modulus
of unreinforced layers is estimated using the Portable Seismic Pavement Analyzer (PSPA).
A cyclic load is applied to the prepared sample using either a 10 in. or 12 in diameter plate.
Maximum pressure of 80 psi (550 kPa) is applied in each cycle, followed by a rest period. For each
laboratory test, 10,000 cycles of load and the rest were applied using an MTS loading machine.
The shape of the loading cycle applied is shown in Figure 4-3. For each load cycle, the data (stress
on the subgrade, hoop strains on geocell, load, and vertical deformation) is collected for the factors
shown in Table 4-5. First, the stress distribution beneath the geocell is evaluated by placing three
pressure cells on the subgrade. Next, strain gauges are attached to geocell to estimate the hoop
57
strains developed in the geocell wall. Finally, the vertical deformation is estimated on the top of
the base layer (beneath the load plate) using the LVDT. The location of strain gauges and pressure
cells in the testing are also shown in Figure 4-4. The specifications of the electronics used in the
experiments are discussed in the later sections.
Figure 4-3 Applied Load Cycle
Figure 4-4 Locations of Stress and Strains Measurement Transducers

4.4.1.2 Strain Gauge
In this study, the deformations are measured using the strain gauges due to the loading cycle and
converted to estimate induced strains. The strain gauges were selected based on their sensitivity
and suitability with the data acquisition system. A half bridge strain gauge circuit was selected
because of better sensitivity. The sensitivity of the bridge can be doubled in a half bridge
58
configuration compared to a quarter bridge. Figure 4-5 shows the difference between the quarter
bridge and half bridge gauge circuits.
Each strain gauge is glued to geocell and adequately protected. Table 4-6 shows the model of strain
gauge used and protection means. Despite employing the precautions in gluing the strain gauges
and protecting them, they are occasionally damaged during compaction. Therefore, the working
condition of strain gauges before and after compaction of the sample is verified for each test. In
some cases, if one of the strain gauges on the half bridge is broken, then the circuit is converted
(wiring) into a quarter bridge, and the strains are recorded for the quarter bridge. More details on
the strain gauges used in the study are presented in Figure B-3 in Appendix B.
Figure 4-5 Half Bridge vs. Quarter Bridge Strain Gauge Circuits
Table 4-6 Strain Gauges

Item Description
Strain Gauge KFH-6-120-C1-11L1M3R (6mm strain gauge, 120Ω,
three pre-wired)
Glue (to glue strain gauge to Ethyl based cyanoacrylate
geocell)
Protection of strain gauge Performix Plasti Dip (flexible protection)
Protection of strain gauge wires PVC Tubing
4.4.1.3 Earth Pressure Cells
For estimating stress on the subgrade layer, the pressure cells were placed on top of the subgrade.
Geokon Model 3500 series (2.5 MPa and 600 kPa) were selected for evaluation, and the
specifications of the pressure cells are included in Appendix B Figure B-4. Each pressure cell is a
semiconductor strain gauge earth pressure cell (circular 9”), with the thermistor in SS housing, 0-
5 VDC output as shown in Figure 4-6.
59
Figure 4-6 Photo of Earth Pressure Cell
The pressure cells are placed 2” to 3” below the subgrade instead of precisely on the top to protect
the pressure cells from the aggregates present in the base material that can influence the stress
readings (concentrated loads). The soil cover on top of the geocell is placed with no material
greater than 3/16” (or 4.5 mm). Three pressure cells were placed in the subgrade, one beneath the
loading plate and two feet away from each other.
4.4.1.4 Data Acquisition System
Two data acquisition systems were used in the testing, as shown in Figure 4-7. The MTS loading
system provides one data system. This data acquisition system records the load and vertical
deformation data from the transducers provided by the MTS. This data acquisition system can
collect data at a frequency of 100 data points per second. The LMS data acquisition system was
employed to record the stresses and strains of the pavement system because it can accommodate
16 channels, i.e., it can record stresses and strains from 16 locations at a time. In addition, it can
record the data at different frequencies. In this study, a frequency of 128 data points per second is
chosen, close to the frequency of other data acquisition systems.
60
Figure 4-7 a) MTS Data Acquisition System, b) LMS Data Acquisition System
4.5 DATA REDUCTION AND CLEANING
The procedures explained below are coded in MATLAB. The codes are developed to take raw data
from the data acquisition without modifications (.dat, .txt forms) and generate the output into an
excel file (.xls).
Before applying the cyclic load on the prepared sample, an initial reading of all electronics is taken
to make sure the measuring devices are working and take the initial reading. The initial reading is
considered a datum, and the data collected during the testing are corrected by subtracting the initial
data.
4.5.1 Noise Removal from Zero Reading (Datum)
Since load and deformation transducers are electronics-based, the transducers tend to measure
electronic noise. This noise can be cyclic or non-cyclic. To minimize the influence of noise, the
data collected when the system is stationary can be evaluated. Since the system is essentially
stationary, the feedback signal obtained from the transducers is noisy. The measured data noise
can be normally distributed with some mean and standard deviation. The noise signal can be
removed by performing Fast Fourier Transform or normalized through the root mean square error
before any analysis. Both techniques can be utilized to identify suitable technique which minimizes
the influence of noise. During the initial data collection, the chances of recording noises are higher,
and the data is refined using the Fast Fourier Transform (FFT). As mentioned in the report, the
pressure cells were placed below the base layer, and the pressure cells carried the dead load of the
base material.
No additional load was applied to check the noise because pressure cells carry a dead load of the
base material. Since no other additional load is applied, the received signal should be constant due
to dead weight. However, Figure 4-8a shows the received signal is not constant and absorbing
some noise from the lab environment. Hence, the influence of noise was minimized by performing
Fast Fourier Transform on the received signal.
61
For example, the pressure cell's zero reading (no additional load applied apart from dead load) is
shown in Figure 4-8. The raw data collected was ranged from 7350 to 7700 Pa (1.06 to 1.11 psi).
The data was refined by removing the noise using the Fast Fourier Transform (FFT), and the actual
reading from the pressure cells was 7650 Pa (1.11 psi). Therefore, theoretically, the pressure cell
was placed below 12 inches of base (Base 1), and around 3 inches into subgrade (Base 2), the
stress on the pressure cell was approximately 1.10 psi. The noise minimized signal shows the input
as 1.11 psi close to the expected value.
The test data (during load) was corrected using a correction factor of -7650 Pa (for this case, only
on pressure cell 1).
Figure 4-8 Noise removal from pressure cell readings

4.5.2 Noise Removal from Actual Data and Data Reduction
Each test is performed for 10,000 load cycles, and each load cycle is around 1.3 seconds. Based
on the data recorded frequency, around 1.7 million data points from each channel (strain gauges,
pressure cells, load cell, and LVDT) are collected from each test. The data needs to reduce to
perform further assessments. In addition to the reduction of data, the quality of data also needs to
be evaluated.
The data recorded through pressure cells, LVDT, strain gauges, and load cells follow the waveform
of the applied load (cyclic pulse). Earlier, some researchers used moving average techniques that
may not suit this analysis (might manipulate the waveform). In this study, we used nonparametric
regression to minimize the data and remove the noise from the readings. Nonparametric regression
provides an effective means by which complex displacement/pressure patterns occurring over a
wide range of values can be captured without the constraint of an assumed functional form.
Therefore, kernel regression (nonparametric regression) or kernel smoothing is used in this study.
Kernel smoothing generally utilizes locally weighted averages of the data defined by a kernel. The
use of 95% confidence interval and root mean square with Kernel smoothing should minimize the
62
influence of noise and identify minimum levels of magnitudes that can be measured with pressure
cells and strain gauges such that erroneous conclusions will not be drawn from the collected data.
The 1.7 million data points can be refined by removing the noise. The data can be reduced to the
required number of data points (150,000 considered in this study) without impacting the observed
waveform. It is observed that the pressure cells (closer to the loading plate), load cells, LVDT are
not impacted by noise during testing. The noise is observed from the test data on the strain gauges
and pressure cell 3 (2 feet away from the loading plate). Figure 4-9 shows the data observed during
the testing. Pressure cell 3 (Figure 4-9c) has some noise in the readings refined by Kernel
regression (Figure 4-9d). Similarly, Figure 4-10 shows the effectiveness of Kernel regression from
the noise from the strain gauge data and reducing the data points without impacting the waveform.
Kernel regression smoothens the data by reducing the influence of noise. It is different from the
averaging data or moving average, as Kernel regression follows the signal waveform. Thus,
averaging may shift the data. To clarify, comparisons of Kernel regression and moving average
are shown in Appendix C Figure C-5.
Figure 4-9 a) Pressure Cell 1 (Original Data), b) Pressure Cell 2 (Original Data), c)
Pressure Cell 3 (Original Data), d) Pressure Cell 3 (Kernel Regression).
63
Figure 4-10 A) Strain gauge (Original Data), B) Strain gauge (Kernel Regression)
4.5.3 Presentation of Data
Instead of presenting the data in waveforms, it is presented as the readings during loading (when
the load is maximum in the cycle as shown in load cycle Figure 4-3) and rest (when the load is
minimum in the load cycle). The following figures present the data for the laboratory test
performed on geocell of 4 in. height, cover thickness of 6”, subgrade modulus of 2.5 ksi, and base
modulus of 5 ksi (for both geocell reinforced and unreinforced sections): a) Figure 4-11 shows the
vertical deformation, b) Figure 4-12 shows the stresses on the subgrade, c) Figure 4-13 shows the
vertical strains on the geocell, and d) Figure 4-14 shows the hoop strains observed on the geocell.
All the data is refined and reduced using Kernel regression.
4.5.3.1 Vertical Deformation
Figure 4-11 shows the vertical deformation on the surface top measured using an LVDT attached
to the loading frame. Again, both reinforced and unreinforced bases produced a similar
deformation. The results observed in other test samples also followed a similar trend.
4.5.3.2 Vertical Stress Distribution on Subgrade
Figure 4-12 shows the vertical stress captured on three pressure cells (PC1, PC2, and PC3) along
the subgrade. It is observed that the stresses on the subgrade are reduced on PC1 and PC2 locations
significantly by placing a geocell reinforced layer. The stresses observed on the PC3 are almost
negligible in both reinforced and unreinforced sections.
64
Figure 4-11 Vertical Deformation A) Geocell, B) No Geocell.
Figure 4-12 Vertical Stresses on Subgrade a) Geocell, b) No Geocell.

4.5.3.3 Vertical Strain on Geocell
Figure 4-13a shows the location and direction of strain gauges attached to monitor the strains
distributed across the geocell mattress. Figure 4-13b shows that the vertical strain on the single-
cell below the loading plate. The initial strain is around 90 microstrains in the initial cycles and
increases up to 115 microstrains by the end of 10,000 load cycles. This indicates that the geocell
is barely compressed in the vertical direction.
65
4.5.3.4 Hoop Strains on Geocell
Figure 4-14 shows the hoop strain distribution across the cells from the center of the loading plate.
The first and second cells show higher strains compared with the third cell. This indicates that the
cell below the loading is expanding and transferring the load in the lateral direction (to the second
cell). After the second cell, the load distribution is almost negligible.
Figure 4-13 a) Location and Direction of Strain Gauges b) Vertical Strain Observed in the
Center Geocell.
Figure 4-14 Hoop Strains on Geocell.
66
4.5.4 Summary of Data
The test results obtained from the laboratory testing for 10 in. and 12 in. loading plates are
summarized in Tables 4.7 through 4.11. In addition, the stress and strain measured with the 10 in.
loading plate are included in Tables 4.7 through 4.10, while Table 4.11 has results summarized for
12 in. loading plate. Initially, the laboratory and FEM applied load using a 10 in. loading plate.
However, the pavement design required a 12 in. loading plate because of FPS-21; therefore, some
analysis was performed in the laboratory using a 12 in. loading plate to verify finite element results
obtained using a 12 in. loading plate.
The measured vertical deformation below the loading plate is summarized in Table 4.7, obtained
from the LVDT of the MTS system. The results indicate an increase in vertical deformation with
a reduction in the base modulus. However, the measured vertical deformation was lower in the
absence of geocell than those measured with the geocell reinforcement. This can be attributed to
lower compaction levels obtained within the individual cells of geocell. Researchers from the
University of Kansas also observed this pattern.
67
Table 4-7 Vertical Deformation (below loading plate)
Vertical Deformation, mils
Geocell 1 (Presto 6") Geocell 2 (Presto 4") Geocell 3 (Tenax 4") Geocell 4 (Tenax 3") No Geocell
Base Base Base Base Base
Good Marginal Poor Good Marginal Poor Good Marginal Poor Good Marginal Poor Good Marginal Poor
(15 ksi) (12 ksi) (5 ksi) (15 ksi) (12 ksi) (5 ksi) (15 ksi) (12 ksi) (5 ksi) (15 ksi) (12 ksi) (5 ksi) (15 ksi) (12 ksi) (5 ksi)
Subgrade
182 190 1154 187 182 389 186 132 1358 331 244 1467 166 137 323
Cover (4.5 ksi)
4” Subgrade
609 593 1142 732 1035 1240 587 535 947 469 189 327 454 754 693
(2.5 ksi)
Subgrade
1153 886 437 1261
Cover (4.5 ksi)
6” Subgrade
205 387 1078 205 413 1238 153 413 482 121 1053 1026 453 753 692
(2.5 ksi)
Table 4-8 Stresses Measured on Top of Subgrade (below loading plate)

Pressure Cell Reading (Below Load Plate), psi
Geocell 1 (Presto 6") Geocell 2 (Presto 4") Geocell 3 (Tenax 4") Geocell 4 (Tenax 3") No Geocell
Subgrade
25.0 17.6 22.7 30.3 34.5 27.8 31.2 34.6 34.3 33.8 31.9 30.6 16.3 18.0 27.2
Cover (4.5 ksi)
4” Subgrade
13.9 19.3 17.7 22.1 29.1 22.7 21.7 28.7 29.5 29.7 36.2 35.0 14.9 25.6 18.4
(2.5 ksi)
Subgrade
26.4 19.0 19.3 37.2 18.4
Cover (4.5 ksi)
6” Subgrade
12.2 16.1 15.1 15.5 19.2 15.5 15.7 20.5 22.2 20.7 20.5 13.0 14.9 25.6 18.4
(2.5 ksi)
68
Table 4-9 Vertical Deformation (below loading plate)
Vertical Compressive Strain (Cell below loading plate), microstrain
Geocell 1 Geocell 2 Geocell 3 Geocell 4 No Geocell
Subgrade
27.15 22.62 13.34 13.5766 45.02 2.86 11.79 15.72 278 89.3 Broken 15.6 NA NA NA
Cover (4.5 ksi)
4” Subgrade
50.49 30.96 54.66 3.93 30.25 6.43 16.43 4.17 112.9 21.2 Broken 32.04 NA NA NA
(2.5 ksi)
Subgrade
13.34 77.89 66.4 27.39
Cover (4.5 ksi)
6” Subgrade
7.62 36.32 14.35 7.62 8.69 9.05 10.84 8.69 20.25 10.6 10.72 11.91 NA NA NA
(2.5 ksi)
Table 4-10 Hoop Strain (cell below loading plate)

Hoop Strain (Cell below loading plate), microstrain
Geocell 1 Geocell 2 Geocell 3 Geocell 4 No Geocell
Subgrade
70.2651 186.9 1197 57.284 306.95 133 Broken 478 498 98.8 2.73 74.43 NA NA NA
Cover (4.5 ksi)
4” Subgrade
Broken 719.8 171.85 41.5 68.48 681.93 89.44 165.9 Broken 84.68 152.8 102.78 NA NA NA
(2.5 ksi)
Subgrade
611.304 1197 Broken 363 123.38 881.65
Cover (4.5 ksi)
6” Subgrade
15.24 509.24 44.54 15.24 144.8 735 103.97 144.82 40.7 56.09 182.81 27.75 NA NA NA
(2.5 ksi)
69
Table 4-11 Summary of Measured Stress and Strain with 12” Loading Plate
Infill Modulus 5 ksi , Subgrade 4.5 ksi Infill Modulus 5 ksi , Subgrade 2.5 ksi Infill Modulus 12 ksi , Subgrade 4.5 ksi
Geocell Geocell Geocell No No Geocell Geocell Geocell No No No
Geocell 6" Geocell 4" Geocell3"+ No Geocell No Geocell
6" + 4" + 3"+Cover Geocell Geocell 6" + 4" + 3"+Cover Geocell Geocell Geocell
+ Cover 4" + Cover 6" Cover 6" Base 10" Base 9"*
Cover 4" Cover 6" 6" Base 10" Base 9"* Cover 4" Cover 4" 4" Base 10" Base 8"* Base 7"*
Vertical Deformation (mm) 27.2 25.6 27.1 26.7 24 25.6 26.8 25.1 13.7 15.0 15.8 14.5
PC1 (psi) 22 26.4 24.5 34.2 33.5 26.4 27 28.5 29.1 28.4 25.1 28.8 31.6 27 30.0 35.1
Vertical Compressive Strain

(Cell below loading plate), 17.74 190 22.75 NA NA 147 135 22.75 NA NA 53 126 31 NA NA NA
microstrain
Hoop Strain (Cell below
1220 1197 Broken NA NA 1438 589 50.5 NA NA Broken 218 413 NA NA NA
loading plate), microstrain
* Stresses calculated using BISAR not laboratory testing. For comparison purpose only.
70
5. CRITICAL EVALUATION OF EXISTING DESIGN
METHODS
Currently, there are no pavement design methods for pavements consisting of the geocell
reinforced layer. To design pavements reinforced with geocell, this study focused on the following:
1) Texas Pavement Design Method, 2) Low Volume Pavement Design Method (AASHTO), and
3) Proposed methods on Geocell incorporation in Pavement Design. Since laboratory experiments
and finite element analysis (FEA) were performed with no asphalt layer on top, more emphasis is
placed on evaluating low-volume unpaved road design methods in practice.
5.1 TEXAS PAVEMENT DESIGN PROCEDURE
In Texas, low-volume flexible pavements are designed using the Flexible Pavement System
software (FPS-21). The program uses a “District Temperature Constant” that assigns a cold region
multiplier to those areas of the state more susceptible to thermal cracking as the only environmental
input. However, the current recommendation is to nullify this parameter by assigning a value of
“31” corresponding to the climate in Central Texas.
The program uses a “confidence level” approach to account for in-place subgrade stiffness
variability, construction variability, and traffic growth. A multiplier is assigned to the cumulative
traffic loading as the desired level of confidence or reliability increases. As a result, the system
can generate designs that may fail under occasional heavy wheel loads. This circumstance is
particularly acute for designs that have low cumulative loading in regions with a poor subgrade.
For this reason, designs obtained from the program must be checked with the “Modified Texas
Triaxial Design Method,” which is included in FPS-21 in the post-design check module. In
addition, a mechanistic design check is provided to evaluate the expected fatigue life of the HMAC
layers and full-depth rut life of the structure with options to use several strain-based performance
models.
FPS-21 uses back-calculated modulus to characterize the pavement layer strength (stiffness) based
on Falling Weight Deflectometer (FWD) deflection measurements, which is different from the
resilient modulus used in the AASHTO design procedure. However, the elastic modulus of
pavement layers is the essential input for finding the pavement structure's layer thickness.
Therefore, it is incumbent upon the designer to have a current set of deflection data for the project
under consideration from which moduli can be generated and institutional knowledge of material
moduli when virgin or recycled materials are incorporated in the design.
Design Confidence Level: This parameter is meant to address variability in material quality,
construction processes, and forecasted traffic as a means of assuring the structure performs as
desired. It does not account for premature failures that may occur due to inadequate construction
quality controls. An overall multiplier to the cumulative traffic loading is applied and can be
modified if more confidence is desired. FPS-21 uses an alphabetic code tied to a reliability or
confidence level: A-80%, B-90%, C-95%, D-99%, E-99.9%. A confidence level as low as level
‘A’ can be considered for designs below 1 million equivalent single axles loads (ESALs); level
‘B’ is recommended for 1 to 5 million ESALs; and level ‘C’ for above 5 million ESALs.
Each material layer used in the structure requires a modulus value characterizing the average
stiffness of that material. The construction process, inherent material variability (initially and over
service life), and environment and traffic loading effects will typically introduce considerable
71
variance. Overestimating this material property can reduce the service life while underestimating
can result in the uneconomical pavement.
Usage of the geocell reinforced layer in pavements is not incorporated in the Texas pavement
design guide. However, the following information includes planar geosynthetic (geogrid and
geotextile) in the pavement.
5.1.1 Geosynthetics in Pavement Structure
According to the pavement design guide of Texas, a geosynthetic is defined as “A manmade
material that consists of one or more products used to provide added benefit to the infrastructure,”
Applications of geosynthetics have been in asphalt concrete overlays of existing asphalt concrete
and hydraulic (Portland) cement concrete surfaces, unbound (flexible) base, soft subgrade,
drainage, and encapsulation. The two geosynthetics primarily used are geotextiles and geogrids.
The applications of geosynthetics in Texas roads are shown in Table 5-1.
Table 5-1 Geosynthetic Application (Pavements in Texas).
Application Name Description
1 Pavement Surface The first application should be separated and not confused
Layer with any other application in a pavement structure. This
Reinforcement application is specific to hot-mix asphalt concrete.
2 Geotechnical Refers to pavement layers only inclusive of subgrades and
Reinforcement subbases
3 Drainage and Drainage and moisture control are features of the pavement
Moisture Control structure, not reinforcements. Their functions are to enhance
and lengthen pavement performance by reducing the
influence of moisture.
5.1.2 Geosynthetics for Geotechnical Reinforcement
Many geosynthetics have multiple uses and can serve more than one function. For instance,
geogrids are often used in a way to restrain base material during compaction or loading. Still, they
also serve as a separation layer to prevent excessive migration and intermingling of pavement
layers at interfaces.
Current usage in Texas has been for both restraints of pavement materials and separation of
materials. The department acknowledges the benefit of geosynthetics in pavement layers; however,
there has been insufficient conclusive research to develop guidance about the reinforcement of
unbound materials in pavement structures now. As a result, usage of geosynthetics is limited to
separation and restraint and is not accounted for pavement design using FPS-21 design.
5.2 LOW VOLUME ROAD DESIGN AASHTO

According to AASHTO, pavement structural design for low-volume roads is divided into three
categories
1. Flexible Pavements,
2. Rigid Pavements, and
3. Aggregate-surfaced roads.
72
For designing low-volume roads, the recommended reliability level is 50 percent. However, the
user may design for higher reliability levels of 60 to 80 percent, depending on the actual projected
level of traffic and the feasibility of rehabilitation, the importance of corridor, etc. The following
section describes the design procedure for low-volume unpaved or aggregate surfaced roads.
5.2.1 Aggregate-Surface Roads
Unlike flexible or rigid design procedures, the design procedure for aggregate-surfaced roads
requires a graphical solution. It is important to note that the effective modulus of the roadbed soil
developed for flexible pavement design should not be used instead of the procedure described here.
Because the primary basis for all rational pavement performance prediction methods is cumulative
heavy axle load applications, using the 18-kip equivalent single axle load (ESAL) design approach
for low-volume roads is necessary. For the aggregate-surfaced (gravel) roads, the maximum traffic
level considered is 100,000 18-kip ESAL applications, while the practical minimum level (during
a single performance period) is 10,000.
The primary design requirements for aggregate surface roads include:
1) The predicted future traffic, W18 for the period
2) The lengths of the seasons
3) Seasonal resilient moduli of the roadbed soil
4) Elastic modulus, EBS (psi) of an aggregate base layer
5) Elastic modulus ESB (psi) of an aggregate subbase layer
6) Design serviceability loss, ΔPSI
7) Allowable rutting, RD (inches), in the surface layer, and
8) Aggregate loss, GL (inches), of the surface layer
Figure 5-1 provides a map showing six different climatic regions of the United States and the
environmental characteristics associated with each. Based on these regional characteristics, Table
5-2 can be used to define the season lengths needed for determining the adequate roadbed soil
resilient modulus for flexible pavement design or the effective modulus of subgrade reaction for
rigid pavement design.
Table 5-3 provides roadbed soil resilient modulus values that may be used for low-volume road
design if the user can classify the general quality of the roadbed material as a foundation for the
pavement structure. Based on the suggested season lengths identified in the previous section,
effective roadbed soil resilient modulus values (for flexible pavement design only) can be
generated for each of the six U.S. climatic regions. These MR values are presented in Table 5-4.
The design procedure is explained through a numerical example (fabricated case study). An
aggregate surface road is needed to withstand the traffic of 35,000 ESALs in zone V of Figure 5-
1 (zone V covers most regions in Texas). Two types of pavement materials base material with
30,000 psi and subbase with 15,000 psi moduli are available for construction. The existing
subgrade is weak, with a modulus of 5,000 psi. The design procedure is explained stepwise in
Table 5-5.
An alternative pavement needs to be designed using the geocell reinforced layer. A six in. geocell
layer is considered for demonstration purposes. The equivalent modulus of the geocell reinforced
layer is taken as 40,000 psi (the actual modulus might vary) using a 15,000-psi modulus material
as an infill.
73
Figure 5-1 The Six Climatic Regions in the United States (AASHTO, 1993).
74
Table 5-2 Suggested Seasons Length (Months) for the Six U.S. Climatic Regions.
U.S. Climatic Season (Roadbed Soil Moisture Condition)
Region Winter Spring-Thaw Spring/Fall Summer
(Roadbed (Roadbed (Roadbed Wet) (Roadbed Dry)
Frozen) Saturated)
I 0.0 0.0 7.5 4.5
II 1.0 0.5 7.0 3.5
III 2.5 1.5 4.0 4.0
IV 0.0 0.0 4.0 8.0
V 1.0 0.5 3.0 7.5
VI 3.0 1.5 3.0 4.5
Table 5-3 Suggested Seasonal Roadbed Soil Resilient Moduli, MR (psi), as a Function of
the Relative Quality of the Roadbed Material.
U.S. Climatic Season (Roadbed Soil Moisture Condition)
Region Winter Spring-Thaw Spring/Fall Summer
(Roadbed (Roadbed (Roadbed Wet) (Roadbed Dry)
Frozen) Saturated)
Very Good 20,000 2,500 8,000 20,000
Good 20,000 2,000 6,000 10,000
Fair 20,000 2,000 4,500 6,500
Poor 20,000 1,500 3,300 4,900
Very Poor 20,000 1,500 2,500 4,000
Table 5-4 Effective Roadbed Soil Resilient Modulus Values, MR (psi) that may be used in
the Design of Flexible Pavements for Low-Volume Roads. Suggested Values Depend on the
U.S. Climatic Region and the Relative Quality of the Roadbed Soil.
U.S. Climatic Relative Quality of Roadbed Soil
Region Very Poor Poor Fair Good Very Good
I 2,800 3,700 5,000 6,800 9,500
II 2,700 3,400 4,500 5,500 7,300
III 2,700 3,000 4,000 4,400 5,700
IV 3,200 4,100 5,600 7,900 11,700
V 3,100 3,700 5,000 6,000 8,200
VI 2,800 3,100 4,100 4,500 5,700
75
Figure 5-2 Design Chart for Aggregate-Surfaced Roads Considering Allowable Serviceability Loss (AASHTO, 1993).
76
Figure 5-3 Design Chart for Aggregate-Surfaced Roads Considering Allowable Rutting (AASHTO, 1993).
77
Table 5-5 Chart for Computing Total Pavement Damage (for both Serviceability and Rutting Criteria) Based on a Trial
Aggregate Base Thickness.
STEPS Description Referred
Tables,
Figures
Step 1 Select different base thicknesses, DBS, which should bound the probable solution. Three base layer thicknesses
8”, 10”, and 12” are selected for demonstration purposes. The data for three bases is shown in Table 5-6. The Table 5-6
details of the data are explained in the following steps.
Step 2 Select the serviceability and rutting criteria for the design. A serviceability index of 3 and rutting depth of 3
inches is considered for demonstration.
Step 3 Enter the approximate seasonal resilient moduli of the roadbed (MR) and the aggregate base material, EBS (psi), Figure 5-1,
in columns 2 and 3, respectively, of Table 5-6. For demonstration purposes, the weak quality of subgrade (5000 Table 5-2,
psi) as shown in Table 3.2 is considered for various seasons. In addition, the season lengths are considered for Table 5-3,
Zone V (as Figure 5-1 for Texas) from Figure 5-2. Table 5-4
Step 4 Enter the seasonal 18-kip ESAL traffic in Column 4 of Table 5-6. If truck traffic is distributed evenly throughout
the year, the lengths of the seasons should be used to proportion the total projected 18-kip ESAL traffic for each
Table 5-6
season. However, if the road is load-zoned (restricted) during specific critical periods, the total traffic may be
distributed only among those seasons when truck traffic is allowed.
Step 5 Within each of the tables (for each base thickness), estimate the allowable 18-kip ESAL traffic for each of the
four seasons using the serviceability-based nomograph (Figure 5-2), and enter in column 5. For example, if the
Figure5-2,
resilient modulus of the roadbed soil (during the cold season) is such that the allowable traffic exceeds the upper
Table 5-6
limit of the nomograph, assume a practical value of 500,000 18-kip ESAL. In Figure 5-2, an example for an 8”
thick base layer is displayed with arrows.
Step 6 Each table estimates the allowable 18-kip ESAL traffic for each of the four seasons using the rutting-based
nomograph in Figure 5-3 (Figure 5-3, an example for 8” thick base layer is shown with arrows) and enter in Figure 5-3
column 7. Again, if the resilient modulus of the roadbed soil is such that the allowable traffic exceeds the upper Table 5-6
limit of the nomograph, assume a practical value of 500,000 18-kip ESAL
78
Chart for Computing Total Pavement Damage (for both Serviceability and Rutting Criteria) Based on a Trial Aggregate Base Thickness
(cont.)
Step 7 Compute the annual damage values in each base layer thickness for the serviceability criteria by dividing the Table 5-6
projected seasonal traffic (column 4) by the allowable traffic in that season (column 5). Next, enter these seasonal
damage values in column 6 of Table 5-6, corresponding to serviceability criteria. Next, follow these exact
instructions for rutting criteria, i.e., divide column 4 by column 7 and enter column 8.
Step 8 Compute the total damage for both the serviceability and rutting criteria by adding the seasonal damages. The Table 5-6,
total damage (serviceability) for the 8”,10”, and 12” is 2.610,1.156, and 0.975, respectively. Similarly, the total Figure 5-4.
damage (rutting) is 1.478,0.560 and 0.353, respectively. Any total damage greater than one indicates the
corresponding base layer thickness fails to satisfy the design criteria. For example, the 8” thick base layer has a
serviceability damage value of 2.610 and a rutting damage value of 1.478, which means that the pavement fails to
fulfill both serviceability and rutting criteria. When damage indices for all the assumed cover thicknesses are
estimated, a graph of total damage versus base layer thickness should be prepared. The average base layer
thickness, DBS, required is determined by interpolating in this graph for total damage equal to 1.0. Figure 3.4
shows that the 9” thick base layer satisfies the rutting criteria but fails the serviceability. A 10.5” thick base
satisfies both criteria. Hence an 11” thick base is ideal for construction.
Step 9 The base layer thickness determined in step 8 should be used for design if the effects of aggregate loss are
negligible. If, however, the aggregate loss is significant, then the design thickness is determined using the
following equation: DBS = DBS+(0.5 X GL). Where GL = total estimated aggregate (gravel) loss (inches) over the
performance period. If, for example, the total estimated gravel loss was 2 inches and the average base thickness
required was 11 inches, the design thickness of the aggregate base layer should be 𝐷𝐷𝐵𝐵𝐵𝐵𝑖𝑖 = 11+(0.5 X 2) = 12
inches. The initial estimated base thickness is 12 inches.
Step The final step of the design chart procedure for aggregate-surfaced roads is to convert a portion of the aggregate Figure 5-5
10 base layer thickness to an equivalent thickness of subbase material. As shown in Figure 5-5, the nomograph can
be used to calculate the subbase layer thickness. For example, out of the 12” inches of the estimated base, only
8” is used, and a third layer (15,000 psi as subbase ESB modulus) is considered subbase material. Using the actual
base thickness 𝐷𝐷𝐵𝐵𝐵𝐵𝑓𝑓 as 8”, subbase modulus 𝐸𝐸𝐵𝐵𝑠𝑠 as 15,000 psi, decrease in base thickness 𝐷𝐷𝐵𝐵𝐵𝐵𝑖𝑖 − 𝐷𝐷𝐵𝐵𝐵𝐵𝑓𝑓 = 12" −
8" = 4", and base modulus (30,000 psi) values on Figure 5.5 (shown in blue line), the required thickness of
subbase layer is approximately equal to 6”. Hence, the final pavement design is an 8” thick base layer (30,000 psi
modulus) over a 6” thick subbase layer (15,000 psi modulus)
79
Table 5-6 Computation of Total Pavement Damage (for both Serviceability and Rutting Criteria) Based on Trial Aggregate
Base Layer.
Trial (1) Season (Roadbed (2) (3) Base (4) (5) (6) (7) (8)
Base Moisture Condition) Roadbed Elastic Projected Allowable Seasonal Allowable Seasonal
(in.) Resilient Modulus, 18-kip 18-kip Damage, 18- kip Damage
Modulus, EBS (psi) ESAL ESAL 𝑾𝑾𝟏𝟏𝟏𝟏 ESAL 𝒘𝒘𝟏𝟏𝟏𝟏
MR (psi) Traffic, Traffic, (𝑾𝑾𝟏𝟏𝟏𝟏 )𝑷𝑷𝑷𝑷𝑷𝑷 Traffic, (𝑾𝑾𝟏𝟏𝟏𝟏 )𝑹𝑹𝑹𝑹𝑹𝑹
W18 (W18) PSI (W18) RUT
8” Winter (Frozen) 20,000 30,000 8,750 400,000 0.022 130000 0.067
Spring/Thaw
1,500 30,000 4,375 4,900 0.893 8400 0.521
(Saturated)
Spring/Fall (Wet) 3,300 30,000 8,750 10000 0.875 20000 0.438
Summer (Dry) 4,900 30,000 13,125 16000 0.820 29000 0.453
Total Total Total
35,000 2.610 1.478
Traffic Damage Damage
10” Winter (Frozen) 20,000 30,000 8,750 400,000 0.022 250000 0.035
Spring/Thaw
1,500 30,000 4,375 9,000 0.486 20000 0.219
(Saturated)
Spring/Fall (Wet) 3,300 30,000 8,750 32000 0.273 50000 0.175
Summer (Dry) 4,900 30,000 13,125 35000 0.375 100000 0.131
Total Total Total
35,000 1.156 0.560
12” Winter (Frozen) 20,000 30,000 8,750 500,000 0.018 400,000 0.022
Spring/Thaw(Saturated) 1,500 30,000 4,375 15,000 0.292 28000 0.156
Spring/Fall (Wet) 3,300 30,000 8,750 22000 0.398 100000 0.088
Summer (Dry) 4,900 30,000 13,125 35000 0.375 100000 0.131
Total Total Total
35,000 0.975 0.353
80
Figure 5-4 Example Growth of Total Damage Versus Base Layer Thickness for Both
Serviceability and Rutting Criteria.
Figure 5-5 Chart to Convert a Portion of the Aggregate Base Layer Thickness to an
Equivalent Thickness of Subbase (AASHTO, 1993).
81
5.2.1.1 Aggregate-Surface Roads using Geocell Reinforced Layer
Initially, a design with a 10-inch-thick layer is developed, and later it is adjusted to a 6-inch
reinforced layer and subbase layer based on pavement damage calculations shown in Table 5-7.
The values shown in column 5 of Table 5-7 are estimated as shown in Figure 5-6, and the values
in column 7 are estimated as shown in Figure 5-7. The total damage values for serviceability and
rutting are less than 1. Hence, the design is acceptable. Since only 6 out of 10 inches of the base is
reinforced, the final step is to convert a portion of the 4” aggregate base layer (unreinforced)
thickness to an equivalent thickness of subbase material. The nomograph as shown in Figure 5-8,
can be used to calculate the subbase layer thickness. Out of 10” inches of the estimated base only
6” is considered as a base and a third layer (15,000 psi as subbase ESB modulus) is considered as
a subbase material. Using the actual base thickness 𝐷𝐷𝐵𝐵𝐵𝐵𝑓𝑓 as 6”, subbase modulus 𝐸𝐸𝐵𝐵𝑠𝑠 as 15,000
psi, decrease in base thickness 𝐷𝐷𝐵𝐵𝐵𝐵𝑖𝑖 − 𝐷𝐷𝐵𝐵𝐵𝐵𝑓𝑓 = 10" − 6" = 4", and base modulus (40,000 psi)
values on Figure 5-6 (shown in blue line), the required thickness of the subbase layer is
approximately equal to 7”. Hence, the final pavement design is 6” thick geocell reinforced base
layer (40,000 psi modulus) over 7” thick subbase layer (15,000 psi modulus). A comparison of the
unreinforced and geocell reinforced layer is shown in Table 5-8.
Table 5-7 Computation of Total Pavement Damage (for both Serviceability and Rutting
Criteria) Based on 10” Aggregate Base Layer.
Trial (1) Season (2) (3) Base (4) (5) (6) (7) (8) Seasonal
Base (Roadbed Roadbed Elastic Projected Allowable Seasonal Allowable Damage
(in.) Moisture Resilient Modulus, 18-kip 18-kip Damage, 18- kip 𝑤𝑤18
Condition) Modulus, EBS (psi) ESAL ESAL 𝑊𝑊18 ESAL (𝑊𝑊18 )𝑅𝑅𝑅𝑅𝑅𝑅
MR (psi) Traffic, Traffic, (𝑊𝑊18 )𝑃𝑃𝐵𝐵𝑃𝑃 Traffic,
W18 (W18)PSI (W18)RUT
10” Winter
20,000 40,000 8,750 500,000 0.02 500,000 0.018
(Frozen)
Spring/Tha
w 1,500 40,000 4,375 14,000 0.31 50,000 0.088
(Saturated)
Spring/Fall
3,300 40,000 8,750 26000 0.34 150,000 0.058
(Wet)
Summer
4,900 40,000 13,125 39000 0.34 210,000 0.063
(Dry)
Total Total Total
35,000 1.00 0.226
82
Figure 5-6 Design Chart for Aggregate-Surfaced Roads Considering Allowable Serviceability Loss (Geocell Reinforced Layer
Calculations) (AASHTO, 1993).
83
Figure 5-7 Design Chart for Aggregate-Surfaced Roads Considering Allowable Rutting (Geocell Reinforced Layer
Calculations) (AASHTO, 1993).
84
Figure 5-8 Chart to Convert a Portion of the Aggregate Base Layer Thickness to an Equivalent Thickness of Subbase (Geocell
Reinforced Layer Calculations) (AASHTO, 1993)
85
Table 5-8 Chart for Computing Total Pavement Damage (for both Serviceability and
Rutting Criteria) Based on a Trial Aggregate Base Thickness.
Thickness Modulus (ksi)
Pavement Type Subbase
Base Subbase Total Base Modulus
Modulus
Unreinforced Section 8" 6" 14" 30 15
Geocell Reinforced 6"* 7" 13" 40** 15
* Geocell Height
** Equivalent Modulus of Geocell layer is considered as 40 ksi with an infill material of 15 ksi
5.3 DESIGN METHODS FOR LOW VOLUME UNPAVED ROADS WITH GEOCELL
REINFORCEMENT
5.3.1 PRESTO GEOSYSTEM Design Method
Conventional flexible pavement design methods are based on historical performance data collected
from either full-scale road tests or ongoing testing and monitoring of pavement performance within
various geographical areas. Federal and local agencies have determined structural values of
conventional road construction materials based on years of in-service performance history. While
many new materials (Stabilizers, geosynthetics, etc.) have been introduced in recent years to
enhance the structural value of conventional construction materials, it isn't easy. It can take several
years to obtain structural values for these components to use with existing design methods. For
this reason, there are no agency-accepted structural values or equivalencies that can be used with
current pavement design methods for the geocell system.
The design of geocell confined granular pavements over soft soils is relatively straightforward and
has been well documented for general design purposes. The design method outlined by Presto is
for the low-volume roads where minor deformations are tolerable or for the design of pavement
subbase layers over soft soils. They are not intended for the design of flexible pavement structures
with paved surfaces. The calculations are only valid for granular pavement design over cohesive
subgrade soils with CBR values less than 5.
Empirically derived bearing capacity coefficients are first used to determine the maximum
allowable stress on a subgrade with either known or estimated shear strength. The maximum
allowable stress is the stress that would cause the subgrade's local punching/shear failure under
sustained loading conditions. The maximum allowable stress is the limiting stress for design
purposes. Next, the Boussinesq theory is used to determine the required depth of granular cover
beneath the design wheel load to ensure that the maximum allowable stress is not exceeded.
5.3.1.1 Design for Unreinforced Base layers (Unpaved Surface Layer)
Table 5-9 shows the estimation of base layer thickness (height) for unpaved roads using
Boussinesq theory. The thickness of the base layer is a factor of the design load, loaded area, and
bearing capacity of the subgrade. Table 5-10 displays the close correlation between California
Bearing Ratio (CBR), shear strength, and Standard Penetration Testing (SPT). The values of shear
strength can be considered from this table if no test data is available.
.
86
Table 5-9 Methodology for Designing Gravel Roads (Unreinforced)
A: Design of Unreinforced Pavement (No paved surface layer)
Step Description Equation
A1 Determine the maximum 𝑞𝑞𝑎𝑎 = 𝑁𝑁𝑐𝑐 𝑐𝑐𝑢𝑢
allowable stress on the Where: Nc Bearing capacity coefficient- based on design traffic
subgrade Cu Subgrade shear strength;
The typical values of shear strength for various subgrades are
shown in Table 2.2
A2 Determine the required 𝑅𝑅
𝑍𝑍𝑢𝑢 =
thickness of the granular 1
pavement, Zu �(1 − 𝑞𝑞𝑎𝑎 )2/3 − 1
𝑝𝑝
𝑃𝑃
𝑅𝑅 = �
𝑝𝑝𝑝𝑝
Where
R= Radius of loaded area (i.e., the effective radius of single or
dual tires); p=Contact Pressure; P= Design wheel load
Table 5-10 Correlation of Subgrade Soil Strength Parameters for Cohesive (Fine-Grained)
Soils
Shear Strength
Bearing Ratio
Identification
Cu kPa(psi)
Undrained
California
(blows/ft)
CBR (%)
Field
SPT
<0.4 <11.7 (1.7) <2 Very Soft (extruded between fingers when squeezed)
0.4-0.8 11.7-24.1 2-4 Soft (molded by light finger pressure)
(1.7)-(3.5)
0.8-1.6 24.1-47.6 4-8 Medium (molded by strong finger pressure)
(3.5)-(6.9)
1.6-3.2 47.6-95.8 8-15 Stiff (readily indented by thumb but penetrated with great
(6.9)-(13.9) effort)
3.2-6.4 95.8-191 15-30 Very stiff (readily indented by thumbnail)
(13.9)-(27.7)
>6.4 > 191 (27.7) > 30 Hard (indented with difficulty by thumbnail)
5.3.1.2 Design for Geocell Reinforced Base layers (Unpaved Surface Layer)
The design procedure for the geocell reinforced base layer is shown stepwise in Table 5-11. The
design is based on the theory that the geocell reinforced layer absorbs a portion of vertical load,
thus reducing the ultimate load on the subgrade. Due to the load transfer to geocell, the pavement
section can either carry higher loads or have an extended life than the unreinforced pavement
section.
87
Table 5-11 Methodology for Designing Gravel Roads with Geocell Reinforced Base
B1 Determine the maximum allowable stress 𝑞𝑞𝑎𝑎 = 𝑁𝑁𝑐𝑐 𝑐𝑐𝑢𝑢
on the subgrade Where: Nc Bearing capacity coefficient- based on design traffic
Cu Subgrade shear strength
B2 Calculate the vertical stress 𝛔𝛔vt at the top 3
⎡ 2
⎤
of the geocell layer 1
⎢
𝜎𝜎𝑎𝑎𝑣𝑣 = 𝑝𝑝 1 − � 𝑅𝑅 2 � ⎥ ; Where zt Depth from surface to top of Geocell walls
⎢ 1+�𝑧𝑧 � ⎥
𝑡𝑡
⎣ ⎦
B3 Calculate the vertical stress 𝛔𝛔bt at the 3
⎡ 2
⎤
bottom of the geocell layer 1
𝜎𝜎𝑠𝑠𝑣𝑣 = 𝑝𝑝 ⎢1 − � 𝑅𝑅 2 � ⎥; Where zb Depth from surface to bottom of Geocell wall
⎢ 1+�𝑧𝑧 � ⎥
𝑏𝑏
⎣ ⎦
B4 Calculate the horizontal stress at the top 𝜎𝜎ℎ𝑣𝑣 = 𝐾𝐾𝑎𝑎 𝜎𝜎𝑎𝑎𝑣𝑣 ; 𝜎𝜎ℎ𝑠𝑠 = 𝐾𝐾𝑎𝑎 𝜎𝜎𝑠𝑠𝑣𝑣
and bottom of the Geocell section ∅
Where Ka is the coefficient of active earth pressure 𝐾𝐾𝑎𝑎 = tan 2 �45 − �
2
B6 Calculate the average horizontal stress at (𝜎𝜎ℎ𝑣𝑣 + 𝜎𝜎ℎ𝑠𝑠 )

the center of the Geocell layer 𝜎𝜎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 =
2
B7 Calculate the reduction in stress, 𝛔𝛔r , 𝐻𝐻
𝜎𝜎𝑟𝑟 = 2 � � 𝜎𝜎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡
directly beneath the center of the loaded 𝐷𝐷
area due to stress transfer to the geocell Where, H = Geocell Height ; D = Effective Geocell Diameter
walls using the following equation δ = Angle of shearing resistance between granular infill material and Geocell walls;
δ = rΦ; The values of r and δ for a different type of geocells are presented in Table
2.4
B8 Determine the allowable design stress, qG 𝑞𝑞𝐺𝐺 = 𝑞𝑞𝑎𝑎 + 𝜎𝜎𝑟𝑟
, on the subgrade with the Geocell
B9 Determine the total required thickness of 𝑅𝑅
𝑍𝑍𝐺𝐺 =
the granular pavement, ZG, with the 1
Geocell reinforced layer �(1 − 𝑞𝑞𝐺𝐺 )2/3 − 1
𝑝𝑝
88
The amount of stress absorbed is:
𝑯𝑯
𝝈𝝈𝒓𝒓 = 𝟐𝟐 � � 𝝈𝝈𝒂𝒂𝒂𝒂𝒂𝒂𝒂𝒂 𝒕𝒕𝒂𝒂𝒕𝒕𝒕𝒕 5-1
𝑫𝑫
where, H = geocell Height; D = effective geocell diameter, and δ = angle of shearing resistance
between granular infill material and geocell walls; δ = rΦ;
Table 5-12 shows typical values of angle of shearing resistance between geocell infill and wall for
various types of geocell texture and infill material. The reduction of vertical stress is a function of
the angle of shearing resistance between infill material and geocell walls, the aspect ratio (H/D),
and the average horizontal stress on the center of geocell
Table 5-12 Recommended Peak Friction Angle Ratio (Presto Geosystems)
Granular Infill Material Cell Wall Type r=δ/Φ
Smooth 0.71
Coarse Sand/Gravel Textured 0.88
Textured-Perforated 0.90
Smooth 0.78
#40 Silica Sand Textured 0.90
Smooth 0.72
Crushed Stone Textured 0.72
5.3.1.3 Angle of Shearing Resistance between Infill Material and Geocell Walls
The friction between the infill material and the geocell wall is one of the key inputs in the design
method proposed by Presto. The friction between the geocell and infill material is typically less
than the internal friction of infill material. Table 5-12 shows the recommended friction angle
between geocell surface type and infill material as a ratio of reduction of internal friction angle of
infill material. As per the table, textured-perforated geocell generates the highest shearing
resistance, and silica sand is the most effective infill material.
5.3.1.4 Aspect Ratio
The aspect ratio is the ratio between the height of the geocell to the equivalent diameter of the
geocell opening. Therefore, the higher aspect ratio indicates more geocell material in the reinforced
base.
5.3.1.5 Average Horizontal Stress on the Center of Geocell
As per the equation, the reduction in vertical stress is proportional to the horizontal stress coming
onto the center of the geocell. It suggests that the higher the horizontal stress, the more significant
is the benefit of geocell. Higher horizontal stress can be achieved if the geocell is placed near the
loading point. The efficiency decreases with the increase in the geocell placement depth from the
applied load. Figure 5-9 shows an example of the design of a low-volume road with and without
geocell. A 6-inch geocell with a 4-inch cover is used for the analysis. Infill base and subgrade
properties are shown in Figure 5-9. In unreinforced conditions, a 12-inch-thick base is required,
whereas a 10-inch-thick (6-inch geocell layer + 4-inch cover) base is required for the use of geocell
reinforcement. The geocell reinforcement reduces the base thickness by 2 inches.
89
Input Parameters
Geocell Equivalent Diameter (inch) 11
Cover over geocell(inch) 4
Geocell depth (inch) 6
Radius of Loading Plate (inch) 6
Contact Pressure (psi) 100

Angle of Internal Friction (infill
30
material)
Subgrade Shear Strength Cu (psi) 5
Bearing Capacity Coefficient Nc 2.8
Stress Reduction Calculation

Stress Reduction and Bearing Capacity
Stress Profile
Calculations
Vertical Horizontal
Depth in inches Description Value
Stress (psi) Stress (psi)
1 99.6 71.2 Active Earth Pressure Coefficient 0.72
2 96.8 69.3 Horizontal Stress on Top of Geocell (psi) 59.3
3 91.1 65.1 Horizontal Stress on Bottom of Geocell (psi) 26.4
4 82.9 59.3 Average Horizontal Stress (psi) 42.88

Aspect Ratio (Height of Geocell/ Equivalent
5 73.8 52.8 Diameter of Geocell) 0.55
6 64.6 46.2 Reduction in stress 7.07
7 56.2 40.2
8 48.8 34.9 Allowable Bearing Capacity No Reinforcement 14.00
9 42.4 30.3 Increased Bearing Capacity 21.07
10 36.9 26.4
Summary Reinforced Base Details

Required Depth of Geocell Reinforced
10.0 Cover over Geocell (inch) 4.0
Layer (inches)
Required Depth for unreinforced base
12.0 Geocell Reinforced Layer (inch) 6.0
layer
% Reduction in Base Layer Thickness
16.7 Subbase below geocell reinforced layer (inch) 0.0
(inches)
Figure 5-9 Design Calculation Geocell Reinforced and Unreinforced Unpaved Roads
(Presto Geosystems)
90
5.3.2 Design Method Presto Geosystems for Low Volume Roads – Unpaved Roads
(Pokharel 2010)
In this study, the design method developed by Giroud and Han (2004 a and b) for planar
geosynthetic-reinforced (geotextile and geogrid) unpaved roads is modified to accommodate
geocell reinforcement. The complete methodology for the design of the unpaved road is shown in
Table 5-13. In addition, Gabr (2001) and Pokharel et al. (2010) observed that geocell reinforcement
reduces the stress distribution angle. This phenomenon is attributed to the geocell confinement of
the base course to increase and maintain the modulus of the base course.
A modulus improvement factor was proposed by Han et al. (2007) to account for this benefit:
𝐸𝐸𝑠𝑠𝑐𝑐(𝑟𝑟𝑎𝑎𝑟𝑟𝑟𝑟𝑓𝑓𝑟𝑟𝑟𝑟𝑐𝑐𝑎𝑎𝑟𝑟)
𝐼𝐼𝑓𝑓 = � �
𝐸𝐸𝑠𝑠𝑐𝑐(𝑢𝑢𝑟𝑟𝑟𝑟𝑎𝑎𝑟𝑟𝑟𝑟𝑓𝑓𝑟𝑟𝑟𝑟𝑐𝑐𝑎𝑎𝑟𝑟)
5-2
where Ebc(reinforced) is the modulus of the reinforced base, and Ebc(unreinforced) is the modulus of the
unreinforced base. According to Pokharel et al. (2010), the modulus improvement factor is
between 1.7 to 2.0 based on the infill material. Regular base material and sand have an
improvement factor of 2, which means geocell reinforcement increases the infill modulus by two
times. If recycled asphalt pavement (RAP) is used as an infill, a value of 1.7 is proposed.
The design equation proposed by Pokharel et al. (2010) is shown in the equation below (Each
parameter in the equation is defined in Table 3.1)
𝑟𝑟 1.5
(0.868 + 0.52 � � 𝑙𝑙𝑙𝑙𝑙𝑙𝑁𝑁) 𝑃𝑃
ℎ= ℎ × �� 2 − 1� 𝑟𝑟
𝜂𝜂{1 + 0.204[𝑅𝑅𝐸𝐸 − 1]} 𝑝𝑝𝑟𝑟 𝑚𝑚𝑁𝑁𝑐𝑐 𝑐𝑐𝑢𝑢
5-3
In the Giroud and Han (2004 a and b) design, the modulus ratio (RE) for unreinforced and planar
geosynthetic-reinforced unpaved roads is limited to 5. They recommended this limit, considering
that base courses cannot be well compacted over soft subgrade. The three-dimensional
confinement by geocell can overcome this problem and help the base course reach and maintain
its higher modulus.
Han et al. (2007) reported that the geocell-reinforced bases had modulus ratios ranging from 4.8
to 10. However, their study's calculated modulus ratios from cyclic plate loading tests and
accelerated moving wheel tests ranged from 3.4 to 7.6. Therefore, Pokharel (2010) kept the
maximum limit of the modulus ratio to 7.6 for Novel Polymeric Alloy (NPA) geocell. The modulus
ratio can be calculated based on the equation shown in step 3 of Table 5-13.
In general, it is proposed to use a nonwoven geotextile as a separation between the geocell
reinforced layer and the underlying soft subgrade. Since a nonwoven geotextile sheet is commonly
used below geosynthetic reinforced bases, the bearing capacity factor (Nc) for geocell-reinforced
unpaved roads can be reasonably assumed to be equal to 5.14. Therefore, in the unreinforced base
scenario, the recommended value is 3.14.
The proposed methodology is demonstrated through an example in the following section.
91
Table 5-13 Recommended Peak Friction Angle Ratio (Presto Geosystems) for Design
1 Bearing capacity 𝑠𝑠 𝑟𝑟 2
mobilization coefficient 𝑚𝑚 = � � �1 − 0.9𝑒𝑒𝑒𝑒𝑝𝑝 �− � � ��
75𝑚𝑚𝑚𝑚 ℎ
2 Modulus improvement The experimental If values estimated were for sand infill material, it was 2.0, and for RAP base
factor If courses, it is 1.7. As the improvement factors were carried out with a 2 cm cover, the modulus of the
unreinforced material was multiplied by ‘If’ for the thickness equal to the height of geocell plus 2cm
cover. The remaining thickness of the base course was considered unreinforced, and no modulus
improvement factor was applied.
3 Modulus Ratio 𝐸𝐸𝑠𝑠𝑐𝑐 3.48𝐶𝐶𝐶𝐶𝑅𝑅𝑠𝑠𝑐𝑐 0.3
𝑅𝑅𝐸𝐸 = 𝐼𝐼𝑓𝑓 = max �7.6, 𝐼𝐼𝑓𝑓 � ��
𝐸𝐸𝐵𝐵𝑎𝑎 𝐶𝐶𝐶𝐶𝑅𝑅𝐵𝐵𝑎𝑎
Where Ebc is the resilient modulus of the base course (MPa), and Esg is the resilient modulus of the
unreinforced (MPa), CBRbc= California Bearing Ratio (CBR) of base courses; and CBRsg California
Bearing Ratio of Subgrade.
Han et al. (2007) reported the geocell-reinforced bases had modulus ratios ranging from 4.8 to 10.
However, this study's calculated modulus ratios from cyclic plate loading tests and accelerated
moving wheel tests ranged from 3.4 to 7.6. Therefore, it is reasonable to set the maximum limit of
the modulus ratio to 7.6 for NPA geocell-reinforced unpaved roads until additional data is available.
4 The height of the base 𝑟𝑟 1.5
(0.868 + 0.52 � � 𝑙𝑙𝑙𝑙𝑙𝑙𝑁𝑁) 𝑃𝑃
required for the ℎ= ℎ × �� 2 − 1� 𝑟𝑟
designed traffic 𝜂𝜂{1 + 0.204[𝑅𝑅𝐸𝐸 − 1]} 𝑝𝑝𝑟𝑟 𝑚𝑚5.14𝑐𝑐𝑢𝑢
where; r is the radius of tire contact (m); N is the number of passes
P is the wheel load (kN); RE is the modulus ratio of base to subgrade (limited to 5.0 for unreinforced
and planar geosynthetic-reinforced roads)
cu is the undrained cohesion of the subgrade soil (kPa); η is the conversion factor between field and
laboratory performances is the allowable rut depth (mm); fs is the factor equal to 75mm
m the bearing capacity mobilization coefficient
𝑟𝑟 1.5 𝑟𝑟 1.5
# for unreinforced case, the value 0.52 � � is replaced with 0.661 � �
ℎ ℎ
# Nc is the bearing capacity factor 3.14 for unreinforced roads, 5.71 for geogrid-reinforced roads, and
5.14 for geotextile-reinforced roads.
92
5.3.3 Example
Find the number of wheel passes for a 7.87 in. (20 cm) thick unpaved RAP base course section
reinforces with 6” (15 cm) high NPA geocell above a weak subgrade. The CBR values of the
subgrade and the base course are 2% and 20%, respectively. The allowable rut is 3” (75mm). The
design wheel load is 8970 lbs (40kN), and the tire pressure is 80 psi (552 kPa). Consider Cu of
subgrade as 5.8 psi (40kPa).
Solution: The solution is presented as the steps described in Table 5-13
5.3.3.1 Design of Geocell-Reinforced Base Layer
5.3.3.1.1 Step 1: Calculation of Bearing Capacity Mobilization Coefficient
𝑠𝑠 𝑟𝑟 2
𝑚𝑚 = � � �1 − 0.9𝑒𝑒𝑒𝑒𝑝𝑝 �− � � ��
75𝑚𝑚𝑚𝑚 ℎ
The radius of the equivalent tire contact area is
0.5
𝑃𝑃 0.5 40
𝑟𝑟 = � � = � � = 0.15 𝑚𝑚, ℎ
𝑝𝑝𝑝𝑝 3.14 × 552
= 0.20 𝑚𝑚(𝑡𝑡ℎ𝑖𝑖𝑐𝑐𝑖𝑖𝑡𝑡𝑒𝑒𝑠𝑠𝑠𝑠 𝑙𝑙𝑜𝑜 𝑏𝑏𝑡𝑡𝑠𝑠𝑒𝑒 𝑖𝑖𝑡𝑡𝑐𝑐𝑙𝑙𝑢𝑢𝑢𝑢𝑖𝑖𝑡𝑡𝑙𝑙 𝑙𝑙𝑒𝑒𝑙𝑙𝑐𝑐𝑒𝑒𝑙𝑙𝑙𝑙)
75 0.15 2
𝑚𝑚 = � � �1 − 0.9𝑒𝑒𝑒𝑒𝑝𝑝 �− � � ��= 0.49
75𝑚𝑚𝑚𝑚 0.20
5.3.3.1.2 Step 2: Modulus improvement factor If

The modulus improvement factor (If) for RAP as infill proposed by Pokharel (2010) is 1.7 within
the geocell and 0.787 in. (2 cm) cover, and 1.0 for the remaining unreinforced portion. The
weighted average modulus improvement factor
(15 + 2) × 1.7 + (20 − 15 − 2) × 1.0
𝐼𝐼𝑓𝑓 = � � = 1.59
20
5.3.3.1.3 Step 3: Estimation of Modulus Ratio
𝐸𝐸𝑏𝑏𝑏𝑏 3.48×200.3
𝑅𝑅𝐸𝐸 = 𝐼𝐼𝑓𝑓 = max �7.6, 𝐼𝐼𝑓𝑓 � ��=6.81
𝐸𝐸𝑠𝑠𝑠𝑠 2.0
The factor k’
′
0.15 1.5
𝑖𝑖 = 0.5 � � = 0.34
0.20
5.3.3.1.4 Step 4: Estimation of Number of Load Cycles the Pavement Can Withstand
Since the thickness of the base layer and geocell height are defined in the example, the solution
will be how many load cycles this pavement section can withstand. The equation in step 4 of Table
5-13 is used. The pavement section can sustain 5500 load cycles.
𝑟𝑟 1.5
(0.868 + 0.52 � � 𝑙𝑙𝑙𝑙𝑙𝑙𝑁𝑁) 𝑃𝑃
ℎ= ℎ × �� 2 − 1� 𝑟𝑟
𝜂𝜂{1 + 0.204[𝑅𝑅𝐸𝐸 − 1]} 𝑝𝑝𝑟𝑟 𝑚𝑚5.14𝑐𝑐𝑢𝑢
93
(0.868 + 0.34 × 𝑙𝑙𝑙𝑙𝑙𝑙𝑁𝑁) 40000
0.20 = × �� − 1� × 0.15
1 × {1 + 0.204[6.81 − 1]} 𝑝𝑝0.152 × 0.49 × 5.14 × 40000
2.14 = 0.868 + 0.34 × log 𝑁𝑁

N=5500 Cycles
Developing a design with an unreinforced section to withstand 5500 load cycles and later
comparing it with a geocell reinforced layer will help in perceiving the benefit of geocell.
5.3.3.2 Design of Unreinforced Base Layer
The following are the parameters required for designing the unreinforced base layer.
Since there is no reinforcement, the modulus improvement factor will be equal to 1.𝐼𝐼𝑓𝑓 = 1.
The modulus ratio will be
3.48 × 200.3
𝑅𝑅𝐸𝐸 = �� = 4.27
2.0
The equation for calculating the base layer thickness for the unreinforced section proposed by
Pokharel (2010) is as follows
𝑟𝑟 1.5
(0.868 + 0.661 � � 𝑙𝑙𝑙𝑙𝑙𝑙𝑁𝑁) 𝑃𝑃
ℎ= ℎ × �� 2 − 1� 𝑟𝑟
𝜂𝜂{1 + 0.204[𝑅𝑅𝐸𝐸 − 1]} 𝑝𝑝𝑟𝑟 𝑚𝑚3.14𝑐𝑐𝑢𝑢
By inserting the numerical values for all parameters in the above equation, the geocell height can
be estimated by trial and error.
0.15 1.5
(0.868 + 0.661 � � 𝑙𝑙𝑙𝑙𝑙𝑙5500) 𝑃𝑃
ℎ= ℎ × �� − 1� × 0.15
𝜂𝜂{1 + 0.204[𝑅𝑅𝐸𝐸 − 1]} 2
𝑝𝑝0.15 𝑚𝑚3.14 × 40000
By trial and error, the value of h is equal to 18.11 in. (0.46 m).
5.3.3.3 Benefit of Geocell Reinforcement
In this example, the geocell reinforced layer height required for the same traffic loading is 7.87 in.
(20 cm where 15 cm is geocell + 2 cm is cover + 3 cm is subbase), whereas the unreinforced layer
thickness is 18 inches (46 cm). Hence a saving of 10 inches (26 cm) is achieved by geocell
reinforcement.
5.3.4 Conclusions
1. Low volume road design can be classified into paved (asphalt on the top) and unpaved (no
asphalt on top).
2. The factor that helps to classify the low-volume roads is the traffic. According to the Texas
pavement design guide, low volume roads have traffic less than one million 1ESALs in its
life. However, according to AASHTO, the maximum number of ESAL applications
considered for low volume flexible or rigid pavement design is 700,000 to 1 million.
94
Therefore, the practical minimum traffic level that can be considered for any flexible or
rigid pavement during a given performance period is about 50,000 ESAL applications.
3. The design procedure suggested for low-volume paved roads by AASHTO and TxDOT is
like the design procedure for regular pavements. However, the main difference is the
reliability or confidence level. A low confidence level is suggested for designing low-
volume roads. TxDOT recommended a confidence level as low as ‘An (80%)’ for designs
below 1 million ESALs. The level of reliability recommended by AASHTO for low-
volume road design is 50 percent. The user may design for higher levels of 60 to 80 percent
depending on the importance of the project.
4. AASHTO aggregate surface low volume road method is a graph-based design. A detailed
explanation of the existing method and incorporation of the geocell reinforced base is
presented through an example.
5. Two proposed design methods for unpaved roads with the geocell reinforced base are
examined. One method is proposed by PRESTO GEOSYSTEM (2008), and the other is
designed by Pokharel (2010). However, these methods are not in practice.
6. The theory behind the PRESTO GEOSYSTEM design is that the geocell reinforced layer
transfers load laterally, and a lower load is transmitted to the subgrade. Three main criteria
that benefit the geocell reinforced layer are:
I. the higher aspect ratio (Height of Geocell/ Diameter of Geocell)
II. the texture of the geocell (perforated geocell wall with the textured wall is more
efficient than smooth geocell)
III. placement of geocell (closer to the loading point highly efficient)
7. Pokharel's (2010) design method is a modification of unpaved roads design with planar
geosynthetics proposed by Giroud and Han (2004 a and b). The design philosophy is based
on the experimental results, which showed that the geocell reinforced layer increases the
modulus of the infill material, and increased modulus is considered in the design. The
modulus improvement through geocell is in the range of 1.7 to 2.0 times.
8. PRESTO GEOSYSTEM (2008) and Pokharel (2010) design methods revealed a significant
benefit of using geocell reinforcement for pavements with a poor subgrade. However, both
the methods have limitations like:
a. Pokharel's (2010) design method is based on a limited test result (geocell from one
manufacturer and four base materials), and its applicability on a different type of
materials (other than tested) is questionable. For example, the design procedure
developed by Pokharel (2010) is valid only for NPA geocell; no recommendations are
proposed for HDPE geocell.
b. The PRESTO GEOSYSTEM (2008) design method does not consider the quality of
the infill material. The primary design criteria friction angle between the infill and
geocell wall considers the classification of infill (sand, silt, stone) but not the quality
of the infill base. It is observed in both our laboratory testing and FEA that with higher
quality (high modulus) infill reduces the benefit of geocell, but PRESTO
GEOSYSTEM design ignores this parameter.
95
96
6. MODEL DEVELOPMENT, STATISTICAL ANALYSIS,
AND VALIDATION
The main idea of this chapter is to show the developing mathematical model for capturing the
geocell reinforced layer benefit over the unreinforced layer using multilinear regression. Later the
assumptions (linear relationship, collinearity, etc.) in the multiple regression are verified through
statistical tools, and the model is enhanced further. Finally, cross-validation techniques for model
development are discussed. The developed model is later employed in the design of pavements
with geocell reinforced layer and will be reported in the subsequent chapter “ Design Method for
Pavements Consisting of Geocells.” Some statistical tests (removal of noise for laboratory tests)
are discussed in Chapter Four and are omitted.
6.1 PATH TO DEVELOPMENT OF MODEL

The laboratory test results and FEA indicated that the geocell performance is a function of geocell
height, infill modulus, cover thickness, and subgrade modulus (Table 3-3). Since the combination
of geocell layered infill materials and existing subgrade conditions vary within geographical parts
of Texas, a mathematical model will be helpful in the pavement design process followed within
Texas. In addition, the developed model will help provide the information needed by the pavement
designers.
The primary step in the model development is identifying the performance factor to be modeled.
It is observed from the FEA and laboratory testing that the geocell reinforced layer reduced the
vertical stresses on the subgrade, which is due to an increase in stiffness (modulus) of the base
layer (geocell reinforced base). Since stress on top of the subgrade is the criteria for design, it was
decided to develop a model that predicts a percent reduction (%Red. stress). The percentage of
vertical stress reduction on subgrade depended on the combination of geocell height, infill
material, cover thickness, and subgrade modulus. Therefore, the percentage vertical stress
reduction due to the geocell reinforced layer on the subgrade can be a possible parameter that can
be modeled to represent the performance of the geocell layer. The probable model is shown in the
following equation:
%𝑹𝑹𝒂𝒂𝑹𝑹.𝑷𝑷𝒕𝒕𝒓𝒓𝒂𝒂𝑺𝑺𝑺𝑺 = 𝒇𝒇(𝑮𝑮𝒂𝒂𝑮𝑮𝑮𝑮𝒂𝒂𝑮𝑮𝑮𝑮 𝑯𝑯𝒂𝒂𝑯𝑯𝒂𝒂𝑯𝑯𝒕𝒕, 𝑪𝑪𝑮𝑮𝒂𝒂𝒂𝒂𝒓𝒓 𝑹𝑹𝑯𝑯𝑯𝑯𝑮𝑮𝑻𝑻𝒕𝒕𝒂𝒂𝑺𝑺𝑺𝑺, 𝑷𝑷𝒕𝒕𝒇𝒇𝑯𝑯𝑮𝑮𝑮𝑮 𝒂𝒂𝒕𝒕𝑹𝑹 𝑷𝑷𝑺𝑺𝑺𝑺𝒂𝒂𝒓𝒓𝒂𝒂𝑹𝑹𝒂𝒂 𝑴𝑴𝑮𝑮𝑹𝑹𝑺𝑺𝑮𝑮𝑺𝑺𝑺𝑺)

6-1
Based on the percentage stress reduction, the equivalent modulus due to geocell reinforcement can
be back-calculated using BISAR. Sample calculation of vertical stress reduction on subgrade due
to geocell reinforced layer and back-calculation of equivalent modulus is shown in the LCCA
calculations section (Section 7).
6.2 DATA FOR DEVELOPING MODEL
The percentage of stress reduction due to the geocell reinforced layer is calculated for the various
input variables ranges shown in Table 6-1. The complete details of combinations of tests are
provided in Table 3- and Table 4-1. The parametric study in FEA used subgrades of 2 ksi, 4.5 ksi,
6 ksi, 9 ksi, and 15 ksi. The base properties evaluated include 2 ksi, 3 ksi, 5 ksi, 7 ksi, 9 ksi, 12 ksi,
15 ksi, 20 ksi, and 30 ksi modulus material. Even though the proposed test plan consisted of only
testing two subgrades and four base materials, additional tests were performed to develop an
appropriate model suitable for Texas (five subgrades and nine bases). It should be noted that the
97
additional tests were performed not only with geocell reinforced layers but also with unreinforced
layers for estimating the benefit of the geocell layer.
Table 6-1 Range of Input Variables Used for Developing the Model
Parameter Range Remarks
Base Modulus 2 ksi to 30 ksi Base as infill as well as cover material.
Geocell Height 3”,4”, and 6” Three geocell are used in the FEA.
Subgrade Modulus 2 ksi to 15 ksi Tested from very poor subgrade to good subgrade
Cover Thickness 4” and 6” Only two cover thicknesses were considered.
6.3 DEVELOPMENT AND VALIDATION OF MODEL FOR ESTIMATING BENEFIT

OF GEOCELL REINFORCED LAYER
In developing the model, this study proposes using multiple regression analysis, a statistical tool
for investigating relationships between variables (Sykes 1993). The main advantage of regression
analysis is that the analyst can assess the statistical significance of the estimated relationships. The
statistical significance provides the degree of confidence that the actual relationship is close to the
expected relationship. One form of regression analysis that is widely used is multi-linear regression
analysis, which finds the linear relationships between variables. The multilinear regression
analysis is often of the following form:
𝒚𝒚 = 𝜷𝜷𝟎𝟎 + 𝜷𝜷𝟏𝟏 𝒙𝒙𝟏𝟏 + … … . . + 𝜷𝜷𝒕𝒕 𝒙𝒙𝒕𝒕 + 𝝐𝝐
6-2
where:
𝑦𝑦 = 𝑢𝑢𝑒𝑒𝑝𝑝𝑒𝑒𝑡𝑡𝑢𝑢𝑒𝑒𝑡𝑡𝑡𝑡 𝑣𝑣𝑡𝑡𝑟𝑟𝑖𝑖𝑡𝑡𝑏𝑏𝑙𝑙𝑒𝑒 (𝑝𝑝𝑟𝑟𝑒𝑒𝑢𝑢𝑖𝑖𝑐𝑐𝑡𝑡𝑖𝑖𝑡𝑡𝑙𝑙)
𝑒𝑒𝑟𝑟 = 𝑖𝑖𝑡𝑡𝑢𝑢𝑒𝑒𝑝𝑝𝑒𝑒𝑡𝑡𝑢𝑢𝑒𝑒𝑡𝑡𝑡𝑡 𝑣𝑣𝑡𝑡𝑟𝑟𝑖𝑖𝑡𝑡𝑏𝑏𝑙𝑙𝑒𝑒
𝐶𝐶𝑟𝑟 = 𝑟𝑟𝑒𝑒𝑙𝑙𝑟𝑟𝑒𝑒𝑠𝑠𝑠𝑠𝑖𝑖𝑙𝑙𝑡𝑡 𝑐𝑐𝑙𝑙𝑒𝑒𝑜𝑜𝑜𝑜𝑖𝑖𝑐𝑐𝑖𝑖𝑒𝑒𝑡𝑡𝑡𝑡
𝜖𝜖 = 𝑒𝑒𝑟𝑟𝑟𝑟𝑙𝑙𝑟𝑟
The regression coefficient explains the correlation between the independent variable and the
dependent variable. Although multi-linear regression analysis is commonly used, the main
limitation of this method is that there is no sure mechanism for the connected variables. Therefore,
this study considers a multiple linear regression analysis as a starting point for the model
development and modifies the model further to fit the testing data.
Thus, the model will be developed in three stages (explained in later sections) as follows:
a. Identify the parameters contributing towards the estimation of the benefit of geocell
reinforced layer,
b. Develop a suitable approach for developing a model, and
c. Perform statistical tests on the established model to verify its performance.
98
6.4 MULTILINEAR REGRESSION MODEL AND VERIFICATION OF ASSUMPTIONS
The raw data (nearly sixty data points) for developing the model is shown in Table C-1; as
discussed earlier, the model is developed using multiple regression. There are certain underlying
assumptions in the multiple regression that needs to be verified.
It is essential to know that the validity of the statistical global and individual tests rely on several
assumptions. So, if the assumptions are inaccurate, the results might be biased or misleading.
However, even if the values in the multiple regression equations are “off” slightly, our estimates
using a multiple regression equation will be closer than any that could be made otherwise. The
following are the assumptions for multiple regressions:
1. There is a linear relationship.
2. The variation in the residuals is the same for both large and small values of ŷ.
3. The residuals follow the normal probability distribution.
4. The independent variables should not be correlated.
5. The residuals are independent.
Assumptions and statistical tools used for verification are demonstrated in the following sections.
6.4.1 Initial Model
Using the raw data shown in Appendix C, a multilinear regression model is developed. The
dependent variable is the % vertical stress reduction on the subgrade top, whereas geocell height,
infill base modulus, subgrade modulus, and cover thickness are independent variables. The
summary of the model is shown in Figure 6-1. The figure shows the regression statistics and
analysis of variance. The r-squared value of the model is 0.622, and the adjusted r-square is 0.593,
which indicates a functional association between independent and dependent variables.
6.4.1.1 Analysis of Variance (ANOVA)
The question that arises is, can the dependent variable be estimated without relying on the
independent variables? The test used to check this condition is the “global test,” which tests
whether all the independent variables may have zero regression coefficients. The null hypothesis
and the alternate hypothesis statements need to be defined as below in testing the hypothesis.
Null Hypothesis 𝐻𝐻0 : 𝛽𝛽1 = 𝛽𝛽2 = 𝛽𝛽3 = 𝛽𝛽4 = 0
The Alternative Hypothesis 𝐻𝐻1 : 𝑁𝑁𝑙𝑙𝑡𝑡 𝑡𝑡𝑙𝑙𝑙𝑙 𝑡𝑡ℎ𝑒𝑒 𝛽𝛽𝑗𝑗′ 𝑠𝑠 𝑡𝑡𝑟𝑟𝑒𝑒 0
If the hypothesis test fails to reject the null hypothesis, it implies that the regression coefficients
are all zero and, logically, are of no value in estimating the dependent variable. Therefore, f
distribution is employed at a significance level of 0.05 (95%) to test the null hypothesis that the
multiple regression coefficients are zero.
Testing the null hypothesis is typically based on a p-value. The p-value is defined as the probability
of observing an F-value as significant or more significant than the F test statistic, assuming the
null hypothesis is true. If the p-value is less than the selected significance level, then the null
hypothesis can be rejected.
The ANOVA in Figure 6-1 shows that the F-statistics p-value for all dependent variables except
subgrade modulus is less than the selected significance level; hence the null hypothesis that all the
coefficients are all zero is rejected.
99
Summary of Model
Regression Statistics
Multiple R-Square 0.622
Adjusted R-Square 0.593
P-value 2.71E-10
Residual Standard Error 5.133
Observations 59
Analysis of Variance (ANOVA)

Sum of Mean
df Squares Square F Significance
(SS) (MS)
Geocell Height 1 1143.54 1143.54 43.398 0.000***
Infill Modulus 1 808.1 808.1 30.669 0.000***
Subgrade Modulus 1 1.15 1.15 0.044 0.83
Cover Thickness 1 262.91 262.91 9.978 0.002***
Significance Codes : 0'***', 0.001 '**', 0.01'*', 0.05 '.'
Regression Coefficients
Standard
Coefficients Error t stat P-Value
Intercept -9.09 4.624 -1.967 0.054
Geocell Height 4.46 0.574 7.776 0.000***
Infill Modulus -0.511 0.0983 -5.198 0.000***
Subgrade Modulus -0.023 0.22 -0.109 0.91
Cover Thickness 2.203 0.697 3.159 0.003**
Figure 6-1 Summary of Initial Regression Model (Full Model)
100
6.4.1.2 Evaluating Individual Regression Coefficients
The next step is to test the independent variables individually to determine which regression
coefficients may be zero or non-zero. The strategy is to use four sets of hypotheses: one for geocell
height, infill modulus, subgrade modulus, and cover thickness. The null and alternative hypotheses
are shown below.
Geocell Height Infill Modulus Subgrade Modulus Cover Thickness
𝑯𝑯𝟎𝟎 : 𝜷𝜷𝟏𝟏 = 𝟎𝟎 𝐻𝐻0 : 𝛽𝛽1 = 0 𝐻𝐻0 : 𝛽𝛽1 = 0 𝐻𝐻0 : 𝛽𝛽1 = 0
𝑯𝑯𝟏𝟏 : 𝜷𝜷𝟏𝟏 ≠ 𝟎𝟎 𝐻𝐻1 : 𝛽𝛽1 ≠ 0 𝐻𝐻1 : 𝛽𝛽1 ≠ 0 𝐻𝐻1 : 𝛽𝛽1 ≠ 0
These hypotheses are again tested at 0.05 significance level, using Student’s t distribution. If the
p-value is less than the selected significance level, then the null hypothesis can be rejected. The
ANOVA in Figure 6 -1 shows that the t-statistic’s p-value for all dependent variables except
subgrade modulus is less than the selected significance level; hence the null hypothesis that all the
coefficients are all zero is rejected for geocell height, infill modulus, and cover thickness. Since
the value of t-statistic’ p-value for subgrade modulus is 0.91, which is higher than the considered
significance level 0.05, the null hypothesis is correct that the coefficient of subgrade modulus is 0.
That means the subgrade modulus does not correlate with the dependent variable.
The following sections show the verification of multiple regression assumptions.
6.4.2 Linear Relationship
The scatter diagrams or residual plots are used to verify that the relationship between the
independent variables and the dependent variable is linear.
6.4.2.1 Scatter Diagrams
Evaluating multiple regression equations requires a scatter diagram that plots the dependent
variable against each independent variable. These graphs help us visualize the relationships and
provide initial information about the direction (positive or negative), linearity, and relationship
strength. Figure 6-2 shows the scatter plots for all the independent variables with the dependent
variables. Geocell height has a positive correlation, indicating that the vertical stress reduction on
the subgrade is higher by increasing the geocell depth. Infill modulus has a negative relation that
indicates that by increasing the infill base modulus, the effectiveness of the geocell reinforced
layer is reducing. The relation between the cover thickness and vertical stress reduction is barely
a linear relationship. Even though the statistical tests performed on the coefficient of cover
thickness showed an independent variable, the scatter plots suggest no relationship between them.
The remedial action in such a scenario will be transforming the independent variable into the
logarithmic or polynomial function. Since only two cover thicknesses (4 inches and 6 inches) are
considered in the parametric study, the transformation will also yield the same relation. Increasing
the range of cover thicknesses in the raw data might improve the linear relationship assumption.
As identified in the ANOVA and student t-test, the subgrade modulus has no significant influence
on the dependent variable. This is further strengthened by the absence of a linear relationship
shown in Figure 6-2.
101
Figure 6-2 Linear Relationship Between Independent Variable and Dependent Variables.
6.4.2.2 Residual Plots

Plots of the residuals can help us evaluate the linearity of the multiple regression equations. To
investigate, the residuals (the difference between actual data and estimated data using a model) are
plotted on the vertical axis against the predicted variable, ŷ. The linearity assumption is valid if
the points are scattered and there is no apparent pattern. However, if the graph shows a non-linear
pattern of the results, the relationship is probably not linear. In that case, different transformations
of the variables in the equation are required. Figure 6-3 shows the residuals vs. fitted plot that
indicates no pattern (red dotted line), and hence the linearity assumption is fulfilled.
6.4.3 Variation in Residuals Same for Large and Small ŷ Values
This requirement indicates that the variation in the residuals is constant, regardless of whether the
predicted values are large or small. To check for homoscedasticity, the residuals are plotted against
ŷ. This is the same graph that we used to evaluate the assumption of linearity. Based on the scatter
diagram, it is reasonable to conclude that this assumption has not been violated. Figure 6-3 shows
the residuals vs. fitted plot, indicating that the estimated values are scattered irrespective of
whether the fitted values are small or large.
102
Residuals vs. Fitted
20
15
10
Residuals
-5
-10
-15
0 5 10 15 20 25
Fitted Values
Figure 6-3 Residuals vs. Estimated (fitted) Values

6.4.4 Distribution of Residuals
The distribution of residuals is evaluated to ensure that the inferences made in the global and
individual hypothesis tests are valid. Ideally, the residuals should follow a normal probability
distribution and is shown as a normal probability plot. If the plotted points are close to a straight
line drawn from the lower left to the upper right of the graph, the normal probability plot supports
the assumption of normally distributed residuals. For example, both the plots in Figure 6-4 satisfy
the requirement of residual distribution as a normal distribution. Therefore, this plot supports the
assumption of normally distributed residuals.
6.4.5 Multicollinearity
Multicollinearity exists when independent variables are correlated. Correlated independent
variables make it difficult to make inferences about the individual regression coefficients and their
individual effects on the dependent variable. In practice, it is nearly impossible to select entirely
unrelated variables. However, a general understanding of the issue of multicollinearity is essential.
Firstly, multicollinearity does not affect a multiple regression equation’s abilities to predict the
dependent variable. However, multicollinearity may show unexpected results when the
relationship between independent and dependent variables is evaluated.
103
Figure 6-4 Verification of Distribution of Residuals
A second reason to avoid correlated independent variables is that they may lead to erroneous
results in the hypothesis tests. Several clues that indicate problems with multicollinearity:
1. An independent variable known to be a vital predictor has a regression coefficient that is
not significant.
2. A regression coefficient that should have a positive sign turns out to be negative, or vice
versa.
3. When an independent variable is added or removed, there is a drastic change in the values
of the remaining regression coefficients.
Table 6-2 shows the correlation matrix where the relationship between all the variables (both
dependent and independent) is summarized. The correlation between dependent variables is
highlighted in italic font. The values indicate that there is some relationship between the dependent
variable(s). For example, the infill modulus and subgrade modulus has a correlation of 0.34. The
question arises is, does this correlation is significant in influencing the model predictions?
A general rule is that if the correlation between two independent variables is between -0.70 and
0.70, there is likely no problem using both independent variables. However, a more precise test
is to use the variance inflation factor (VIF). The value of VIF can be found as follows:
1
𝑉𝑉𝐼𝐼𝑉𝑉 =
1 − 𝑅𝑅𝑗𝑗2
6-3
The term 𝑅𝑅𝑗𝑗2 refers to the coefficient of determination, where the selected independent variable is
used as a dependent variable, and the remaining independent variables are used as independent
variables. A VIF greater than 8 is considered unsatisfactory, indicating that the independent
variable should be removed from the analysis. Before performing VIF, eigenvalues of the
correlation matrix and a ratio of maximum eigenvalue to minimum eigenvalue are used to identify
collinearity existence.
104
Table 6-2 Correlation Matrix
Geocell Infill Cover Subgrade Reduction in Vertical
Height Modulus Thickness Modulus Stress on Subgrade
Geocell Height 1.00 0.14 -0.19 0.09 0.57
Infill Modulus 0.14 1.00 -0.02 0.34 -0.39
Cover Thickness -0.19 -0.02 1.00 0.08 0.16
Subgrade Modulus 0.09 0.34 0.08 1.00 -0.09
Reduction in Vertical Stress
0.57 -0.39 0.16 -0.09 1.00
on Subgrade
For initial detection, eigenvalues of correlation matrix for Table 6-2 are 1.66, 1.40,1.10,0.66, and
0.19. A variance in the eigenvalues indicates the existence of collinearity; however, the calculated
eigenvalues have very low variation (maximum value of 1.66 and a minimum value of 0.19).
Kappa value (max eigenvalue / min eigenvalue) > 100 indicates higher collinearity. The value of
Kappa is 8.93 (<100), which indicates that there is no significant correlation between independent
variables. Despite early indications that collinearity is not of major concern, variance inflation
factor (VIF) are estimated for verification, and the VIF’s of all the independent variables is shown
in Table 6.3. Since all the independent variables have VIFs less than 8, the collinearity between
independent variables is minimal.
Table 6-3 Variance Inflation Factors (VIF)
Geocell Height Infill Modulus Subgrade Modulus Cover Thickness
1.06 1.15 1.14 1.05
6.4.6 Independent Observations

The fifth assumption about regression and correlation analysis is that successive residuals are not
autocorrelated. This identifies no pattern to the residuals, the residuals are not highly correlated,
and there are no long runs of positive or negative residuals.
Autocorrelation frequently occurs when the data is collected over a period, and the Durbin-Watson
test can be performed to identify the presence of autocorrelation. If the residuals are correlated,
problems arise while conducting hypothesis tests about the regression coefficients. Also, a
confidence interval or a prediction interval, where the multiple standard errors of the estimate are
used, may not yield the correct results. The autocorrelation, reported as r, is the strength of the
association among the residuals. The r has the same characteristics as the coefficient of correlation.
For example, values close to -1.00 or 1.00 indicate a strong association, and values near 0 indicate
no association. Instead of directly conducting a hypothesis test on r, the following Durbin-Watson
statistic tests were performed:
∑𝑟𝑟𝑣𝑣=2(𝑒𝑒𝑣𝑣 − 𝑒𝑒𝑣𝑣−1 )2
𝑢𝑢 =
∑𝑟𝑟𝑣𝑣=1(𝑒𝑒𝑣𝑣 )2
6-4
105
The value of the Durbin-Watson statistic can range from 0 to 4. The value of d is 2.00 when there
is no autocorrelation among the residuals. When the value of d is close to 0, this indicates a positive
correlation. Values beyond 2 indicate negative autocorrelation. Negative autocorrelation seldom
exists in practice.
To conduct a test for autocorrelation, the null and alternate hypotheses are:
Ho: No residual correlation (p=0)
H1: Positive residual correlation (p>0)
But the residuals vs. fitted values graph shown in Figure 6-3 shows no autocorrelation (no trend).
Hence, the assumption that successive residuals are independent is satisfied.
Table 6-4 shows the summary of verification of assumptions in multiple regression. As per the
table, all conditions are satisfied. However, in the model, the subgrade modulus can be neglected
(failure to have a linear relationship, t-stat on coefficient statistical insignificant, F-test ANOVA
indicates that the influence of subgrade modulus is negligible for estimating the vertical stress
reduction subgrade). The modified model is shown in Figure 6-5.
Table 6-4 Verification of Assumptions in Multiple Regression
Assumption Statistical Tool Result
A linear Scatter plot Figure 6-2, Figure 6-2 shows independent variable subgrade
relationship Residual vs. Fitted plot modulus has no linear relationship with %
between each (Figure 6-3) percentage of vertical stress reduction. Cover
independent thickness has a weak linear relationship with the
variable with the dependent variable. Increasing the range of cover
dependent thickness in the data might change the linear
variable relationship. The remaining independent variables
satisfy the linear relationship. Figure 6-3 shows no
trend between residuals vs. fitted plots, which
satisfies the linear relationship.
Variation in Residual vs. Fitted plot There is no variation in the residuals whether the
Residuals Same (Figure 6-3) fitted values are large or small. Hence satisfies the
for Large and assumption.
Small ŷ Values
Distribution of Figure 6-4 Verification of Figure 6-4 shows that the residuals are normally
Residuals shall be Distribution of Residuals distributed. Hence this assumption is satisfied.
normally
distributed
There is no Correlation Matrix Table Table 6-2 shows that there is a correlation between
multicollinearity 6-2, Variance Inflation the independent variables. However, the statistical
between the Factors (VIFs) Table 6-3. test VIFs show that the values are less than 8, which
independent indicates the impact of collinearity is negligible.
variables. Thus, the assumption is satisfied.
Independent Residual vs. Fitted plot Figure 6-3 shows that there is no autocorrelation (no
Observations. (Figure 6-3) trend). Hence the assumption that successive
successive residuals are independent is satisfied
residuals should
be independent
106
Summary of Model
Regression Statistics
Multiple R-Square 0.622
Adjusted R-Square 0.601
P-value 4.69E-11
Residual Standard Error 5.084
Observations 59
Analysis of Variance (ANOVA)

Sum of Mean
df Squares Square F Significance
(SS) (MS)
Geocell Height 1 1143.54 1143.54 44.24 0.000***
Infill Modulus 1 808.1 808.1 31.263 0.000***
Cover Thickness 1 262.91 262.91 10.204 0.002***
Regression Coefficients
Standard
Coefficients Error t stat P-Value
Intercept -9.149 4.55 -2.011 0.049
Geocell Height 4.46 0.567 7.859 0.000***
Infill Modulus -0.514 0.0919 -5.601 0.000***
Cover Thickness 2.196 0.687 3.194 0.002**
Figure 6-5 Summary of Reduced Model.
107
6.5 MODEL DEVELOPMENT USING CROSS-VALIDATION TECHNIQUES
Statistically, the best course of action would be to use all the data for model building and statistical
methods to get reasonable error estimates. However, when a model is developed using all data,
one of the fundamental problems is “Overfitting.” Overfitting occurs when a model inappropriately
picks up trends in the total data that do not generalize to a new data set. When this occurs, the
model based on the initial data set doesn’t estimate the dependent variable. Even from a non-
statistical perspective, users of the developed models emphasize the need for an untouched set of
samples to evaluate performance. Although it is highly desirable to have a test set of data to verify
the developed model, it is often difficult to get real-life data in the model development stage. One
of the ways to alleviate the problem stated above (overfitting and lack of test data) is to use the
cross-validation technique. Cross-validation maximizes the value of the available limited data and
allows the identification of the model's accuracy with new data.
There are two types of cross-validation techniques 1) full cross-validation or leave-one-out
validation method (LOOV) and 2) segment (KK-fold and Bootstrapping) cross-validation. A
detailed discussion of each method is included in Appendix C.
1) Full cross-validation or leave-one-out validation method (LOOV) is done when
sufficient data is unavailable to satisfy statistical requirements. With this technique, the
same number of models as data size is developed. One data point (one set of independent
variables) is removed in each iteration, and the model is developed with the remaining data
set. The validity of the developed model is evaluated using the removed data set and
estimating the dependent variable. This process is repeated for the remaining data set by
eliminating one data set at a time. After each test, the model's accuracy is verified through
a metric (such as coefficient of determination, root means square error, etc.). The model
which performed better is considered as the final model. Figure 6-6 shows an example of
how this validation technique is applied. The full cross-validation method is known to be
one of the best validation tests. Although some researchers suggest that full cross-
validation is not an ideal test because the tests are made each time but using the “same
data.”
2) Segmented cross-validation method is a good validation test that is used when there is
enough data. For this test, the data is segmented into samples of different sizes or
percentages. The model is developed on one sample data and tested on the other. Multiple
iterations can be performed by randomly picking the samples. The size of the segments
could vary depending on the data size and user choice.
There are various approaches for performing segmented cross-validation. K-fold or KK-fold
cross-validation (CV) is one such technique. First, the samples are randomly partitioned into
kk sets (called folds) of roughly equal size in their basic KK-fold CV. Next, a model is fitted
using all the samples of the first subset. Then, the prediction error of the fitted model is
calculated using the held-out samples. Finally, the sample operation is repeated for each fold,
and the model’s performance is calculated by averaging the error across all folds. This process
can be repeated numerous times (iterations). Thus, many performance estimates can be done
using KK-fold CV, unlike LOOV, where the number of tests will be fixed. The schematic of
k-fold cross-validation is presented in Figure 6-7.
Another type of CV that is widely used is “BOOTSTRAPPING.” Bootstrapping takes a
random sample with replacement. Some data is selected for testing the model, and the gap
108
created by moving testing data is replaced with the existing data. So, there are chances that
each sample is used more than once in each test. This replacement helps to keep the size of
model data constant. The process is repeated multiple times (50-75 times). This procedure
yields low variance compared to LOOV and KK-fold CV. Figure 6-8 shows a simple example
of how Bootstrapping works.
Data 1 2 3 4 5 6 7 8 9 10
Model
Model Training Data
Test Data
Test 1 2 3 4 5 6 7 8 9 10 1
Test 2 1 3 4 5 6 7 8 9 10 2
Test 3 1 2 4 5 6 7 8 9 10 3
Test 4 1 2 3 5 6 7 8 9 10 4
Test 5 1 2 3 4 6 7 8 9 10 5
Test 6 1 2 3 4 5 7 8 9 10 6
Test 7 1 2 3 4 5 6 8 9 10 7
Test 8 1 2 3 4 5 6 7 9 10 8
Test 9 1 2 3 4 5 6 7 8 10 9
Test 10 1 2 3 4 5 6 7 8 9 10
Figure 6-6 Example of Leave One Out Validation.
Figure 6-7 Example of KK-fold Segmented Cross-Validation.
109
Figure 6-8 Example of Bootstrapping.
6.5.1 Full cross-validation or leave-one-out validation method (LOOV)
The training and testing data sets are shown in Appendix C, Table C-2, and Table C-3. The model
developed using the LOOV is shown in Table 6-5. The model generated from the LOOV method
(R-square 0.620) is different from the reduced model developed using the full data. Figure 6-9
shows the performance of the developed model on the training and testing datasets.
Table 6-5 Multiple Linear Regression Model Parameters for Reduced Model (LOOV)
Coefficients Standard t value p-value Significance
Error
Intercept -7.6948 5.177 -1.486 0.145
Geocell 4.6497 0.676 6.875 0.000 ***
Height
Infill -0.496 0.110 -4.489 0.000 ***
Modulus
Cover 1.7401 0.777 2.241 0.0231 **
Thickness
110
Figure 6-9 Predicted Data vs. Measured Data on Training and Testing Data Sets
(Model Developed by LOOV Methodology).
6.5.2 KK-FOLD Cross-Validation

In the KK-fold CV method, the training data is segmented into four groups. The model is
developed on three segments and tested on the 4th one. This process is repeated 25 times. The
training data set and testing data sets are shown in Appendix C, Table C-4, and Table C-5. The
model developed using the KK-fold is shown in Table 6-6. The model generated from the KK-
fold method (R-square 0.60) is different from the reduced model developed using the full data.
Figure 6-10 shows the prediction capability of the developed model on the training and testing
datasets.
Table 6-6 Multiple Linear Regression Model Parameters for Reduced Model (KK-FOLD)
Error
Intercept -8.985 5.272 -1.704 0.096
Geocell 4.566 0.671 6.802 0.000 ***
Height
Infill -0.502 0.107 -4.705 0.000 ***
Modulus
Cover 2.088 0.823 2.536 0.015 **
Thickness
111
(Model Developed by KK FOLD CV Methodology)
6.5.3 Bootstrapping Cross-Validation

The training data set and testing data sets are shown in Appendix C, Table C-6, and Table C-7.
The model developed using bootstrapping is shown in Table 6-7. The model generated from the
bootstrapping method (R-square 0.570) is different from the reduced model developed using the
full data. Figure 6-11 shows the prediction capability of the developed model on the training and
testing datasets.
Table 6-7 Multiple Linear Regression Model Parameters for Reduced Model
(Bootstrapping CV)
Error
Intercept -8.8573 5.5228 -1.604 0.1166
Geocell 4.6424 0.6988 6.644 0.000 ***
Height
Infill Modulus -0.5287 0.1101 -4.802 0.000 ***
Cover 1.9325 0.8223 2.350 0.0238 **
Thickness
112
(Model Developed by Bootstrapping Methodology)
6.6 SUMMARY OF MODELS

Each cross-validation technique explained in Appendix C has advantages over another method.
For this study, the three cross-validation techniques were employed, and the technique which
generated the best model will be considered for validation. In all the CV techniques explained
below, only 75% of data is used (training data set), and 25% (testing dataset) is kept for testing.
Furthermore, only significant variables found in the previous analysis are considered in the model
development, i.e., the influence of subgrade modulus is ignored in the model.
The performance of the developed model on testing data is statistically measured using the paired
t-tests and Pearson correlation. Pearson’s correlation coefficient measures the strength of the linear
relationship between two sets of variables. The paired t-test is a statistical analysis used to test a
hypothesis about the mean population difference between paired (predicted vs. measured)
observations. Pearson’s correlation coefficient measures the strength of the linear relationship
between two sets of variables.
The details of the paired t-test and Pearson correlation are given in Appendix C (C.1).
The three cross-validation techniques generated different models. The most appropriate model
among the three was selected based on their performance (analyzed using paired t-test and Pearson
correlation) on the testing data. The summary of the models and statistical tests are presented in
Table 6-8. For paired t-test, the following hypothesis is considered:
Null Hypothesis: 𝜇𝜇𝑟𝑟 = 0
Alternative Hypothesis: 𝜇𝜇𝑟𝑟 ≠ 0
where 𝜇𝜇𝑟𝑟 𝑖𝑖𝑠𝑠 the true difference in population values, and D is hypothesized value.
113
The results in Table 6.8 show that the models developed using CV’s supports (p-value > 0.05) the
alternative hypothesis that the true difference in predicted vs. measured values of the testing data
set are not equal to zero. It is to be noted that the D value in the hypothesis is considered zero (no
difference between predicted and measured). Therefore, the results might change if an allowance
is considered (D>0) in the hypothesis. Based on the 95% confidence levels, the model developed
using LOOV-CV generated better results (less spread of predictions) and had a Pearson correlation
of 0.82. Hence, the most appropriate model of this data is achieved through LOOV-CV, and the
model is below:
%𝑅𝑅𝑒𝑒𝑢𝑢.𝐵𝐵𝑣𝑣𝑟𝑟𝑎𝑎𝐵𝐵𝐵𝐵 = −7.69 + 4.65 × 𝐺𝐺𝑒𝑒𝑙𝑙𝑐𝑐𝑒𝑒𝑙𝑙𝑙𝑙 𝐻𝐻𝑒𝑒𝑖𝑖𝑙𝑙ℎ𝑡𝑡 − 0.50 × 𝐼𝐼𝑡𝑡𝑜𝑜𝑖𝑖𝑙𝑙𝑙𝑙 𝑀𝑀𝑙𝑙𝑢𝑢𝑢𝑢𝑙𝑙𝑢𝑢𝑠𝑠
+ 1.74 × 𝐶𝐶𝑙𝑙𝑣𝑣𝑒𝑒𝑟𝑟 𝑡𝑡ℎ𝑖𝑖𝑐𝑐𝑖𝑖𝑡𝑡𝑒𝑒𝑠𝑠𝑠𝑠
6-5
Hence the model shows that the geocell depth (height) is the major influencing factor, followed by
the cover thickness and infill modulus. Remember that the model developed is based on the input
ranges shown in Table 6-1. Since the cover thickness showed a vague linear relationship with the
dependent variable additional testing is required to expand the range of cover thickness.
Table 6-8 Summary of Models Developed Using Cross-Validation Techniques
Coefficients paired t-test *

95%
correlation
R-Square
Modeling
Geocell Height
Infill Modulus
Pearson
Type of
Confidence
Significance
Hypothesis
Thickness
Intercept
(p-value)
Levels
Cover
Lower Upper
Without Cross
-9.15 4.46 -0.514 2.19 0.62 Na Na Na Na Na
Validation (CV)
LOOV - CV -7.69 4.65 -0.50 1.74 0.62 0.12 Alternative -0.66 4.81 0.82
KK-FOLD CV -8.98 4.56 -0.50 2.09 0.60 0.19 Alternative -0.90 3.89 0.85
Bootstrapping CV -8.85 4.64 -0.53 1.93 0.56 0.16 Alternative -3.87 0.75 0.88
*paired t-test performed by implementing the model developed using training data on the testing
data.
114
7. LIFE CYCLE COST ANALYSIS
This chapter provides information on the procedure of life cycle cost analysis of pavements. Life
cycle cost analysis (LCCA) is a data-driven tool that provides a detailed account of the total costs
of a project over its expected life. The purpose of performing LCCA is to provide necessary
information to the decision-makers in selecting the best design. Therefore, LCCA needs to estimate
the various costs in a pavement life cycle like initial construction, maintenance, user, etc., and
discount the anticipated costs to present worth.
There are presently two methodologies being in use for LCCA.
1) Deterministic Approach to LCCA applies procedures and techniques without regard for the
variability of the inputs. The primary disadvantage of this traditional approach is that it does not
account for the variability associated with the LCCA input parameters.
2) Risk Analysis Approach (or Probabilistic Approach) characterizes uncertainty. The Interim
Technical Bulletin on LCCA by the FHWA (J. W. Smith 1998) advocates this approach because
it combines probability descriptions of analysis inputs with computer simulations to generate the
entire range of outcomes and the likelihood of occurrence.
Currently, the deterministic approach is widely used in many states of the US. Still, the risk
analysis approach is recommended by Rangaraju et al., 2008 based on research conducted to
evaluate LCCA practices among state highway agencies for pavement selection. According to the
study, 81 percent of agencies (17 out of 21) are still using a deterministic approach, 14 percent (3
out of 21) use a combination of deterministic and probabilistic approaches, and only one state is
using a probabilistic approach. However, the risk analysis approach is not in practice due to
complexity in considering the inputs required for analysis. Therefore, if the deterministic approach
is used, a sensitivity analysis of LCCA results is recommended to accompany LCCA results.
7.1. DETERMINISTIC APPROACH
The following are the procedural steps involved in conducting a life-cycle cost analysis (LCCA).
1. Establish alternative pavement design strategies for the analysis period
2. Determine performance periods and activity timing
3. Estimate agency costs
4. Estimate user costs
5. Develop expenditure stream diagrams
6. Compute Net Present Worth (NPW) or Equivalent Uniform Annual Costs
(EUAC)
7. Analyze results
8. Reevaluate design strategies
7.1.1 Establish alternative pavement design strategies for the analysis period
7.1.1.1 Analysis period
It is the period used to evaluate the total investment required to build and maintain the pavement
at an agreed-upon quality level. Although FHWA’s LCCA Policy Statement recommends an
analysis period of at least 35 years for all pavement projects (including new or total reconstruction
projects as well as rehabilitation, restoration, and resurfacing projects), an analysis period range of
30 to 40 years is also acceptable (J. W. Smith 1998). In Norway, an analysis period is reduced to
115
10 to 20 years because 40 years’ analysis is considered an extended period. According to Scheving
(2011), an analysis of more than 20 years will not be suitable because of changes in the economic
situation, traffic, and technology.
7.1.1.2 Alternate Pavement Designs
Develop alternate designs or different maintenance preservation strategies that perform well during
the desired analysis period. Alternate design implies that each design will perform equally, provide
a similar level of service over the same performance period, and have similar life-cycle costs
(Stephanos November 2008).
7.1.1.3 Determine performance periods and activity timing
After selecting the analysis period and alternate pavement design, the next step is to determine the
timing and performance of each maintenance activity. This step also involves determining the
maintenance duration and how the ongoing traffic will be managed during maintenance.
7.1.1.4 Estimate Agency Costs
Agency costs include all costs incurred directly by the agency over the life of a pavement.
Although numerous activities are conducted during the construction, reconstruction, or major
pavement rehabilitation, only those specific to a pavement alternative should be included in the
agency costs (VDOT 2002). As the present study is focused on comparing various pavement
designs, costs typical to all alternatives cancel out; these cost factors are generally noted and
excluded from LCCA calculations (J. W. Smith 1998). Agency costs include initial preliminary
engineering, contract administration, construction supervision, and construction costs, as well as
future routine and preventive maintenance, resurfacing and rehabilitation cost, and the associated
administrative cost. Preliminary engineering and other administrative costs were not contemplated
in this study.
The first step in estimating the agency costs is to calculate the quantities and unit prices. The
quantities were calculated based on the structural design and maintenance. The quantities were
estimated typically for one mile of a highway. In this study, the unit prices are taken from the
average low bid prices published by the Texas Department of Transportation (TxDOT) (“Average
Low Bid Unit Prices” 2016).
At the end phase of the pavement life cycle, salvage value will be assigned to the pavement,
resulting in a negative cost (savings). Salvage value is considered rather than demolition of
pavement based on the practicability issues. Therefore, salvage value represents the value of an
investment alternative at the end of the analysis period.
7.1.1.5 Estimate User Costs
In the simplest sense, user costs are expenses incurred by the highway user over the project's life.
User costs are an aggregation of three separate cost components: vehicle operating costs (VOC),
user delay costs, and crash costs.
7.1.1.6 Develop Expenditure Stream Diagrams or Cash Flow Diagrams
A cash flow diagram is a pictorial representation of a financial problem that portrays the timing of
the cash flows and their character (either cost or saving). Thus, a cash flow diagram helps to
comprehend the entire investment quickly. For example, figure 7-1 shows a typical cash flow
diagram for a highway project.
116
Figure 7-1 Cash flow diagram for a pavement
7.1.2 Compute Net Present Worth (NPW) or Equivalent Uniform Annual Costs (EUAC)
NPW is calculated by discounting all project costs to the present year. All project costs
throughout the analysis period are expressed in the form of a single cost in terms of the present
year monetary value (shown in Figure 7-2). The relative cost of alternatives can then be directly
compared from this single representative value; the NPW is estimated using the following
equation
𝑟𝑟
1
𝑁𝑁𝑃𝑃𝑉𝑉 = 𝐼𝐼𝐶𝐶 + � 𝑃𝑃𝑀𝑀𝐶𝐶𝑘𝑘
(1 + 𝑖𝑖)𝑘𝑘
𝑘𝑘=1
7-1
where IC= Initial construction costs
i= discount rate
k= year of expenditure
PMCk=maintenance treatment cost at year k
n=analysis period
Equivalent Uniform Annual Costs (EUAC) is another way to look at the results of an LCCA.
In this case, all alternative project costs are converted to a uniform annual cost over the analysis
period. Whereas NPW discounts all costs to a single base year which can be compared, EUAC
discounts (shown in Figure 7-2) all alternative activities to a yearly cost which is then compared.
EUAC is particularly useful when budgets are established on an annual basis. The EUAC
calculations are performed as per the equation below
𝑖𝑖(1 + 𝑖𝑖)𝑟𝑟
𝐸𝐸𝐸𝐸𝐸𝐸𝐶𝐶 = 𝑁𝑁𝑃𝑃𝑉𝑉 � �
(1 + 𝑖𝑖)𝑟𝑟 − 1
7-2
117
Figure 7-2 Cash flow diagram for NPW and EUAC
7.1.3 Analyze Results
The next step is to analyze the estimated costs. Since LCCA requires they guide more inputs and
the accuracy of the analysis, more scrutiny on the inputs is needed. Sensitivity analysis is one such
tool that assists decision-makers in understanding the likelihood of results happening. A sensitivity
analysis is recommended as a minimum to study the impact of the individual inputs on LCCA
results.
7.1.4 Reevaluate Design Strategies
If required, the whole design strategies can be reevaluated, and new design strategies can be
developed. Once the new designs are generated, the LCCA can be repeated, as explained in the
previously mentioned steps.
In this study, instead of performing LCCA on geocell reinforced layered pavements, we used
LCCA to estimate the unit cost for constructing the geocell reinforced layer.
The calculated unit cost will help to decide the viability of using geocell in pavement construction.
This approach is chosen for two reasons:
1) The market price of geocell varies among manufacturers. However, none of the
manufacturers disclosed the actual price of the delivered product, which restricted LCCA
(due to lack of geocell layer price) calculations.
2) The requirement of geocell is primarily dependent on the quality of pavement layer
materials (subgrade, base) and traffic. Through the proposed approach, this study
calculated the possible unit price of the geocell layer for pavements at different
geographical locations such as Dallas, Paris, Fort Worth, Atlanta, and San Antonio.
The following section describes how this study estimated the modulus of the geocell reinforced
layer by using the outputs from both laboratory testing and finite element modeling. The
subsequent steps describe the use of the estimated geocell layer modulus in pavement design and
LCCA analysis.
118
7.2.ESTIMATION OF GEOCELL LAYER MODULUS
To perform LCCA of the pavement, it is necessary to develop alternative pavement designs for
comparing costs. Additionally, TxDOT uses FPS-21 for pavement design analysis. Thus,
alternative designs need to be developed using FPS-21. Since FPS-21 requires layer modulus as
an input for performing analysis, the first step is to estimate the modulus of the layer reinforced
with geocell. However, the literature review identified that a process for estimating modulus of
layer reinforced with geocell is not in practice. Therefore, this study developed a process based on
FPS-21 design principles and evaluation results and is described in the following paragraphs.
Based on the laboratory evaluation and FEA, the geocell reinforcement allows:
1. reduction in the thickness of the base layer, or
2. the increased service life of the pavement, or
3. use of low-quality construction materials.
In both laboratory testing and finite element analysis, the geocell reduces the stress on the
subgrade. Therefore, the modulus of the geocell reinforced layer was estimated based on trial and
error such that similar subgrade stresses were estimated. This estimated geocell modulus was later
used in FPS-21.
7.2.1 Comparison of Geocell vs. No Geocell
The process of estimation of geocell layer equivalent modulus is demonstrated through a
laboratory evaluated pavement section shown in Figure 7-3. Since the evaluation included geocell
reinforcement and no-reinforcement testing, the change in observed subgrades stresses can be
attributed to the geocell reinforcement. The stresses estimated on top of the subgrade for both
geocell reinforced section (Figure 7-4a) and unreinforced section (Figure 7-4b) for 10,000 load
cycles are shown in Figure 7-4. It is observed that the geocell reinforced section reduced the stress
from 34 psi (No-Geocell) to 26.5 psi (Geocell) below the center of the load plate (PC1). The stress
reduction faded away with an increase in distance from the load plate (PC2 and PC3).
Figure 7-3 Pavement Section Evaluated in the Laboratory
119
Figure 7-4 Stress Distribution a) Geocell and b) No Geocell
The overall stress reduction in the presence of geocell reinforcement is summarized in Figure 7-5.
The stresses are maximum just below the load plate and gradually decrease to zero psi at 18 in.
away from the load plate; however, the path followed differs in the presence of geocell
reinforcement. Due to geocell reinforcement, the stresses reduce to less than 5 psi at a distance of
7 in. from the center of the load plate, while unreinforced sections experience similar stress levels
at a distance of 15 in. from the center of the load plate. At a distance of 18 in. from the center of
the load plate, the stresses are negligible with and without geocell reinforcement. The results
suggest that fewer stresses are transferred to the subgrade in the presence of reinforcement and are
similar to the condition where the high-quality base modulus is used rather than reinforcement. To
include the benefit of reinforcement in FPS-21, the study proposes to use an equivalent modulus
of the reinforced layer. The equivalent modulus is estimated such that the subgrade stresses at the
center of the load plate and 18 in. away from the load plate are similar to the one observed in the
laboratory. Although approximate, the use of equivalent modulus obtained due to reinforcement
can be included for estimating the design service life of the pavement. Using BISAR, a three-
layered pavement section is developed, and stresses on the subgrade are calculated. Figure 7-6
shows the stresses estimated from the BISAR vs. unreinforced pavement section. Then the
modulus of the middle layer is increased (by trial) such that it matches with the stress generated
from the geocell reinforced layer. Figure 7-7 shows that the equivalent modulus estimated for the
geocell reinforced layer is approximately 28 ksi, whereas the modulus of the infill material in the
geocell is only 4.5 ksi. The geocell increased the modulus of infill by approximately six times. It
is to be noted that using BISAR, the definite trend of stress distribution on subgrade by geocell
layer cannot be fitted. This is due to the behavior of the geocell reinforced layer, which acts as a
semi-rigid pavement (stress on the second geocell is significantly dropped than the first geocell.
..
120
Figure 7-5 Benefit of Geocell in Reduction of Stress (Laboratory Results)
Figure 7-6 Comparison of unreinforced section with BISAR equivalent Section

(Laboratory Results)
121
Figure 7-7 Comparison of geocell reinforced section with BISAR equivalent Section
(Laboratory Results)
The following subsections discuss the estimation of geocell modulus based on the results of the
finite element modeling. The significant difference between the FEM and laboratory testing is the
number of load cycles. In the laboratory testing, 10,000 cycles of the load are applied, whereas, in
the finite element analysis, only 20 cycles of load are applied due to two reasons:
1) It was observed that there is no significant change in stresses and/or deformation with more
load cycles (preliminary testing was performed up to 400 cycles of load).
2) Reduce the computational time of the FEM model (more cycles lead to a substantial
increase in computational time).
The stress distribution obtained from FEA for the pavement section shown in Figure 7-3 is shown
in Figure 7-8. The stress distribution trends obtained from the laboratory and the FEA are similar
(Figure 7-4 and 7-8). However, due to geocell reinforcement, the stress on top of the subgrade (at
the center of the load plate) is estimated to be 28 psi (FEA) in comparison to 26.5 psi estimated
from the laboratory testing. Similarly, the stresses on top of the subgrade without any
reinforcement resulted in 37 psi from FEA, while only 34 psi was estimated from the laboratory
testing.
The equivalent geocell layer modulus for FEM results was estimated similar to the steps performed
for the laboratory test results and are shown in Figures 7-9 and 7-10. The estimated equivalent
geocell layer modulus is 31 ksi (based on FEA), while 28 ksi was estimated based on the laboratory
test results. Thus, the estimated equivalent geocell reinforced layer modulus from both FEM and
laboratory are within 10% of each other, indicating FEA can be used for LCCA.
To estimate the benefit of geocell reinforcement, the equivalent modulus of various infill base
materials was obtained for the base layer reinforced with a 4-inch geocell. The FEA was performed
for various infill base materials (3 ksi to 30 ksi modulus) using the above procedure on top of the
subgrade layer (4.5 ksi modulus). The estimated equivalent modulus is summarized in Figure 7-
11, and it is apparent that the geocell reinforced layer modulus is dependent on the infill modulus.
The obtained equivalent geocell modulus increased with an increase in infill modulus from 3 to 12
122
ksi; however, an increase in infill modulus reduced the effectiveness of geocell reinforcement. For
instance, the infill modulus of 12 ksi provided an equivalent geocell modulus of 48 ksi, indicating
a four times increase. In comparison, only 37.5 ksi equivalent geocell modulus was obtained for
infill material with 30 ksi modulus. The analysis is consistent with published literature indicating
that the effectiveness of geocell reinforcement reduced with an increase in the quality of infill
material.
Figure 7-8 Stress Distribution a) Geocell and b) No Geocell
123
Figure 7-9 Benefit of Geocell in Reduction of Stress (FEM)
Figure 7-10 Equivalent Modulus Calculation (FEM)
124
Figure 7-11 Estimated Geocell Layer Modulus with Various Infills (4inch Geocell)
7.3. ESTIMATION OF COST OF GEOCELL REINFORCED LAYER
Although the use of geocell has been documented, the cost of geocell and geocell installation has
not been documented. Since one of the project's tasks was to calculate and perform LCCA, the
cost calculations were performed to identify the installation cost of the geocell reinforced layer
that will provide service life similar to that of the conventional base material. Also, this study made
two assumptions for estimating the cost of the geocell reinforced layer:
1) The cost of the stabilized base is higher than the flexible base, and
2) The cost of a higher modulus base is higher than the lower modulus base.
Thus, the viable cost of the geocell reinforced layer is back-calculated using LCCA. The cost
estimation and viability of geocell reinforcement are documented through three pavement designs:
a) Ellis county of Dallas district, b) Titus county of Atlanta district, and c) Ulvade county of San
Antonio district. For the pavement design, the geocell reinforced layer modulus is considered to
be 48 ksi, which assumes an infill modulus of 12 ksi (Figure 3.20). The respective districts
provided the pavement designs, and geocell reinforcement was incorporated using FPS 21
pavement design software. The unit prices for various districts are taken from Average Low Bid
Unit Prices from the Texas Department of Transportation. Since the prices for similar materials
varied within the district, the cost of the geocell reinforced layer is estimated by including cost
variability. The base material cost considered in this study is included in Table 7-1. Each base
material's cost is considered a triangular probability distribution with minimum, mean, and
maximum values. The mean cost of the stabilized base is assumed to be $60 per cubic yard for all
four districts. Geocell reinforced layer cost is estimated using the mean values (deterministic
LCCA) and the costs' variability (probabilistic LCCA).
125
Table 7-1 Cost of Base Materials Considered in this Study
Costs Probability Good Base (Base with modulus Average Base
District Stabilized Base
Distribution greater than 35 ksi) (< 35 ksi, >20 ksi)
Atlanta $45, $50, $55 $35, $40, $45 $55, $60, $65
Triangular
Paris $50, $57, $62 $45, $52, $56 $65, $75, $85
(Minimum, Mean,
San Antonio $20, $30, $40 $12, $25, $35 $55, $60, $65
Maximum)
Dallas $38, $45, $56 $30, $40, $46 $55, $60, $65
For the probabilistic approach, the Monte Carlo Simulation (Latin Hypercube Sampling) was
performed by varying the base prices of Table 7-1 and performing 5,000 simulations. The
following sections show the estimated costs of a geocell reinforced layer providing similar service
life for various districts. The following assumptions were made for LCCA analysis:
1. A discount rate of 4% is assumed for estimating net present worth.
2. The salvage value of pavement is not considered because of a lack of data on the
reusability of geocell reinforced layers.
3. The unit prices for the pavement layers need to be verified according to each county.
7.3.1 Design 1: FM 55 Ellis County of Dallas District

The data for performing LCCA for FM 55 is shown in Table 7-2, including the design criteria,
traffic, and construction data (from 0.14 Miles South of Nash Howard Road to US 77). Based on
the data provided, three alternative pavement designs were developed (Figure 7-12) and evaluated
for 40 years of service life. Alternative 1 is the district's typical pavement design, including a
stabilized base (35 ksi) and a good quality base (40 ksi). The Alternative 2 pavement design is
based on the lowest base quality (25 ksi) that can provide service life similar to Alternative 1.
Finally, alternative 3 is the pavement designed with geocell reinforcement.
The designs Alternatives 2 and 3 were developed based on discussion with district personnel and
FPS-21 constraints. Alternatives 1 and 2 have a subgrade of 200 in. while Alternative 3 has a
subgrade of 196 in. Since the subgrade is of lower quality, the top of 4 in. of the subgrade is
typically scarified and mixed with lime to stabilize the layer. As a result, construction equipment
can move on top of the surface with minimal damage to the layer. Traditionally, the design doesn’t
include stabilized subgrade layer because lime eventually washes away with rainwater. Thus,
Alternative 3 evaluates a design where the top 4 in. of the subgrade is replaced with a geocell
reinforced layer. Alternative 2 assumes that only locally available base material of mediocre
quality is available, and the layer is not stabilized. In this situation, the thickness of the base layer
and asphalt layer needs to be modified to meet the service life. In comparison, the total pavement
thickness is minimal for Alternative 3, indicating that the base thickness can be reduced even with
the inferior base material is used to infill geocell.
The cost estimates for three alternatives are summarized in Table 7-3. In terms of Alternatives 1
and 2, the cost of Alternative 2 is higher because of the increase in the HMA layer from 2 to 3 in.
Overall, the cost of Alternative 1 turns out to be $601,593, while the cost is $697,319 for
Alternative 2. Since the construction cost of the geocell reinforced layer is unavailable, the cost
without the geocell layer is calculated, which is $423,769. Based on the difference between the
126
cost of Alternative 1 and 3, it can be stated that if the geocell reinforced layer construction is less
than or equal to $177,824 (or $14 per square yard), then there is no increase in overall cost.
Table 7-2 Inputs for Pavement Design (Ellis County, Dallas District)
Basic Design Criteria
-Length of the Analysis Period (Years) 20.0
-Minimum Time to First Overlay (Years) 8.0
-Minimum Time Between Overlays (Years) 8.0
-Design Confidence Level (80.0%) A
-Serviceability Index of the Initial Structure 4.5
-Final Serviceability Index (P2) 2.5
-Serviceability Index P1 after an Overlay 4.5
-District Temperature Constant 31.0
-Subgrade Elastic Modulus by County (ksi) 6.0
-Interest Rate of Time Value of Money (%) 7.0
Traffic Data
-ADT at the beginning of Analysis Period (Vehicles/Day) 1700.0
-ADT at the end of Twenty Years (Vehicles/Day) 2400.0
-One Direction 20 Year 18 kip ESAL (millions) 0.182
-Average Approach Speed to the Overlay Zone (MPH) 60.0
-Average Speed Through Overlay Zone (Overlay Direction) MPH 45.0
-Average Speed Through Overlay Zone (Non-Overlay Direction) MPH 60.0
-Proportion of ADT arriving each hour of construction (Percent) 6.0
-Percent Trucks in ADT 5.2
Construction and Maintenance Data
-Minimum Overlay Thickness (Inches) 2.0
-Overlay Construction Time (Hours/Day) 12.0
-Asphaltic Concrete Compacted Density (Tons/C.Y.) 1.9
-Width of Each Lane (Feet) 11.0
127
HMA 500 KSI 3"
HMA 500 KSI 2"
HMA 500 KSI 2"

FLEXIBLE BASE 40 KSI 8"

CEMENT TREATED BASE
10"
35 KSI
GEOCELL BASE 48 KSI 4"
SUBGRADE 6 KSI 200" SUBGRADE 6 KSI 200" SUBGRADE 6 KSI 196"
Alternative 1 Alternative 2 Alternative 3
Figure 7-12 Alternative Designs Developed Using FPS 21 and Geocell Reinforced Layer
(Ellis County, Dallas)
128
Table 7-3 Cost Estimate for Geocell Reinforced Layer (Ellis County, Dallas District)
Alternative 1 Actual Design
Unit No of Thickness Width Length Quantity Total
Description Units
Price Lanes (in) (feet) (miles) (Cu.Yd.) Cost
Hot Mix Asphalt (500 ksi) CY 115 2 2 11 1 717 $82,459
Flexible Base (40 ksi) CY 56 2 8 11 1 2868 $160,616
Cement Treated Base (35 ksi) CY 60 2 10 11 1 3585 $215,111
Subgrade (6 ksi) CY 2 2 200 11 1 71704 $143,407
Alternative 1 Cost $601,593
Alternative 2 Modified Design with Poor Base Material

Description Units
Subgrade (6 ksi) CY 2 2 200 11 1 71704 $143,408
Alternative 3 With Geocell

Description Units
Geocell Base CY ? 2 4 11 1 1434 ?
Subgrade (4.5 ksi) CY 2 2 196 11 1 70270 $140,539
Difference $177,824
Geocell Cost Per
$14
Square Yard
129
The cost of the geocell reinforced layer using the Monte Carlo Simulation is shown in Figure 7-
13. The mean estimated cost based on 5,000 simulations is $14.33 with a standard deviation of
$1.42. Thus, the minimum cost is $10.11, and the maximum cost is $18.86, and 90% of the time,
the cost of the geocell layer is estimated to be between $11.92 and $16.69. Thus, the construction
of the geocell reinforced layer is faster than the other base reinforcement techniques (cement or
lime stabilized). However, these LCCA calculations are based on the assumption that there is no
time difference between the alternatives.
Figure 7-13 Estimated Geocell Reinforced Layer Costs using Probabilistic LCCA (Ellis
County, Dallas).
Similarly, the cost estimates were performed for Uvalde County of San Antonio District (Table
7-4 design data, Figure 7-14 alternative designs, and Table 7-5 cost estimates). In this design, the
maintenance cost was also included in LCCA. This design requires maintenance (overlay) after 11
years of service. Hence, the Net Present Worth is calculated as discussed in Section 2 of this report.
The optimal construction cost for a geocell reinforced layer is estimated to be $ 15 per square yard.
The estimated cost can be interpreted as if the cost of the geocell layer is less than or equal to $15;
then there is no cost increase due to the use of geocell. The cost of the geocell reinforced layer
using the Monte Carlo Simulation is shown in Figure 7-15. The mean estimated cost based on
5,000 simulations is $15.01, with a standard deviation of $1.55. The minimum cost is $9.59, and
the maximum cost is $20.08, and 90% of the time, the cost of the geocell layer is estimated to be
between $12.45 and $17.57.
The cost estimate for Titus county of Atlanta District are included in Table 7-6 (design data),
Figure 7-16 (alternative designs), and Table 7-7 (cost estimate). The unit cost for geocell is
estimated at $ 12 per square yard. This design requires maintenance (overlay) after 14.7 years of
service. Hence, the Net Present Worth is calculated as discussed in Section 2 of this report. A
discount rate of 4% is considered in the net present worth estimation. The optimal construction
cost for a geocell reinforced layer is estimated to be $ 12 per square yard. The estimated cost can
be interpreted as if the cost of the geocell layer is less than or equal to $12. Then there is no cost
increase due to the use of geocell. The cost of the geocell reinforced layer using the Monte Carlo
Simulation is shown in Figure 7-17. The mean estimated cost based on 5,000 simulations is $12.46
with a standard deviation of $0.62. The minimum cost is $10.71, and the maximum cost is $14.17,
and 90% of the time, the cost of the geocell layer is estimated to be between $11.43 and $13.49.
130
7.3.2 Design 1: US 83 Uvalde County, San Antonio District
Table 7-4 Input Data for Pavement Design (Uvalde County, San Antonio District)
Length of the Analysis Period (Years) 20.0
Minimum Time to First Overlay (Years) 8.0
Minimum Time Between Overlays (Years) 8.0
Design Confidence Level (80.0%) C
Serviceability Index of the Initial Structure 4.5
Final Serviceability Index (P2) 2.5
Serviceability Index P1 after an Overlay 4.2
District Temperature Constant 31.0
Subgrade Elastic Modulus by County (ksi) 14.9
Interest Rate of Time Value of Money (%) 7.0
Traffic Data
ADT at the beginning of Analysis Period (Vehicles/Day) 2800.0
ADT at the end of Twenty Years (Vehicles/Day) 3800.0
One Direction 20 Year 18 kip ESAL (millions) 2.672
Average Approach Speed to the Overlay Zone (MPH) 70.0
Average Speed Through Overlay Zone (Overlay Direction) MPH 65.0
Average Speed Through Overlay Zone (Non-Overlay Direction) MPH 70.0
Proportion of ADT arriving each hour of construction (Percent) 5.0
Percent Trucks in ADT 22.3
Minimum Overlay Thickness (Inches) 2.0
Overlay Construction Time (Hours/Day) 10.0
Asphaltic Concrete Compacted Density (Tons/C.Y.) 1.9
Width of Each Lane (Feet) 12.0
131
Initial Life After Overlay Initial Life After Overlay Initial Life After Overlay
11.0 Years 26.1 Years 11.0 Years 26.1 Years 11.0 Years 26.1 Years
Overlay (500 KSI) 2"

HOT MIX HOT MIX
ASPHALT (500 ASPHALT (500
Overlay (500 KSI) 2" KSI) KSI) 5"
HOT MIX HOT MIX Overlay (500 KSI) 2"
3"
ASPHALT (500 ASPHALT (500 HOT MIX HOT MIX
3"
ASPHALT (500 ASPHALT (500
FlEXIBLE BASE (65 FlEXIBLE BASE
7"
KSI) (65KSI)
FLEXIBLE FLEXIBLE FLEXIBLE BASE 65 FLEXIBLE BASE 65
8"
CEMENT OR CEMENT OR BASE BASE 24" ksi ksi
LIME LIME 30 KSI 30 KSI
STABILIZED STABILIZED 12" FLEXIBLE BASE FLEXIBLE BASE
6"
BASE BASE 30 KSI 30 KSI
35 KSI 35 KSI GEOCELL BASE GEOCELL BASE
4"
48 KSI 48 KSI
SUBGRADE SUBGRADE SUBGRADE SUBGRADE

200" 200" SUBGRADE SUBGRADE
14.9 KSI 14.9 KSI 14.9 KSI 14.9 KSI 196"
14.9KSI 14.9KSI
Alternative 1 Alternative 2 Alternative 3
Figure 7-14 Alternative Designs Developed Using FPS 21 and Geocell Reinforced Layer for
Uvalde County, San Antonio District.
Figure 7-15 Estimated Geocell Reinforced Layer Costs using Probabilistic LCCA (Uvalde
County, San Antonio).
132
Table 7-5 Cost Estimate of Geocell-Reinforced Layer (Uvalde County, San Antonio District)
Alternative 1 Actual Design Initial Construction
Description Units
Dense Graded HMA (500 ksi) CY 115 2 3 12 1 1173 $134,933
Lime (Cement) Stabilized (35 ksi) CY 60 2 12 12 1 4693 $281,600
Subgrade (14.9 ksi) CY 2 2 200 12 1 78222 $156,444
Maintenance (11 Years)
Overlay CY 115 2 2.5 12 1 978 $112,444
Net Present Worth $728,152
Description Units
Subgrade (14.9 ksi) CY 2 2 200 12 1 78222 $156,444
Overlay CY 115 2 0.5 12 1 196 $22,489
Description Units
Geocell CY ? 2 4 12 1 1564 ?
Subgrade (14.9 ksi) CY 2 2 200 12 1 78222 $156,444
Overlay CY 115 2 2.5 12 1 978 $112,444
Difference Compared with Alternative 1 $211,200
Geocell Cost Per
$15
Square Yard
133
7.3.3 Design 1: US 67 at FM 1001 Titus County, Atlanta District
Table 7-6 Input Data for Pavement Design (Titus County, Atlanta District)
Length of the Analysis Period (Years) 20.0
Minimum Time to First Overlay (Years) 10.0
Minimum Time Between Overlays (Years) 8.0
Design Confidence Level (80.0%) C
Serviceability Index of the Initial Structure 4.5
Final Serviceability Index (P2) 3.0
Serviceability Index P1 after an Overlay 4.2
District Temperature Constant 25.0
Subgrade Elastic Modulus by County (ksi) 10.0
Interest Rate of Time Value of Money (%) 6.0
Traffic Data
ADT at the beginning of Analysis Period (Vehicles/Day) 6800.0
ADT at the end of Twenty Years (Vehicles/Day) 9400.0
One Direction 20 Year 18 kip ESAL (millions) 3.738
Average Approach Speed to the Overlay Zone (MPH) 65.0
Average Speed Through Overlay Zone (Overlay Direction) MPH 30
Average Speed Through Overlay Zone (Non-Overlay Direction) MPH 40
Proportion of ADT arriving each hour of construction (Percent) 6.0
Percent Trucks in ADT 14.3
Minimum Overlay Thickness (Inches) 1.5
Overlay Construction Time (Hours/Day) 12.0
Asphaltic Concrete Compacted Density (Tons/C.Y.) 1.98
Width of Each Lane (Feet) 12.0
Based on the LCCA, using a geocell reinforced layer does not necessarily increase the cost of construction
if the cost of a geocell reinforced layer is between $12 and $15 (varies between districts). However, if the
construction cost if more than $15, then there is an increase in the initial cost of construction.
134
Figure 7-16 Alternative Designs Developed Using FPS 21 and Geocell Reinforced Layer for Titus
County, Atlanta District.
Figure 7-17 Estimated Geocell Reinforced Layer Costs using Probabilistic LCCA (Titus
County, Atlanta).
135
Table 7-7 Cost Estimate of Geocell-Reinforced Layer (Titus County, Atlanta District)
Alternative 1 Actual Design
Unit No of Thickness Width Length
Description Units Quantity (Cu.Yd.) Total Cost
Price Lanes (in.) (feet) (miles)
Cement Stabilized Subgrade (30 ksi) CY 60 4 10 12 1 7822 $469,333
Subgrade (10 ksi) CY 2 4 200 12 1 156444 $312,889
Maintenance (14.7 Years)
Overlay CY 115 4 2 12 1 1564 $179,911
Alternative 1 Cost $1,602,946
Subgrade (10 ksi) CY 2 4 200 12 1 156444 $312,889
Overlay (500 ksi) CY 115 4 2 12 1 1564 $179,911
Alternative 3 Modified Design with Geocell Layer
Geocell CY ? 4 4 12 1 3129 ?
Subgrade (14.9 ksi) CY 2 4 196 12 1 153315 $306,631
Overlay CY 115 4 2 12 1 1564 $179,911
Difference Compared with Alternative 1 $350,435
Geocell Cost Per
$12
Square Yard
136
8. CLOSURE
8.1 SUMMARY
Cellular Confinement Systems, popularly known as “Geocell,” are durable, lightweight, three-
dimensional fabricated systems that are expandable on‐site to form a honeycomb‐like structure.
Geocell filled with infill material and compacted to form a rigid to the semi‐rigid layer. The height
of the geocell and the size of each cellular unit vary between manufacturers. Some geocell surface
has been textured to increase soil‐geocell wall friction. In this study, the mechanism for improved
bearing capacity and benefits derived from geocell were evaluated. Also, the influence of in-fill
material quality on the performance of geocell has been evaluated, which is a critical issue when
only lower/marginal quality material is available. Since geocell provides lateral reinforcement, the
possibility of reducing the layer thickness to achieve similar performance levels to the one
expected from the thicker base and sub-base systems was also evaluated. The economic feasibility
of geocell use in pavements is also evaluated using LCCA. This study also provides a guideline
for practitioners in selecting materials, design, construction, and safety of pavements with geocell
cell layers. Finally, the conclusion is drawn, and the recommendation of this study is explained in
the following sections.
8.2 CONCLUSION
I. The following conclusions can be drawn from FEA:
1. Influence of Base Modulus
a. The geocell reinforced layer reduced the stresses up to 20% six inches away from
the center of the loading plate and up to 50% nine inches away from the center of
the loading plate compared with no geocell condition (unreinforced layer).
b. Similar trends were observed in the vertical strains on the subgrade.
c. The hoop strain on the loaded geocell increased with a lower base modulus.
However, a significant strain reduction is observed in the next geocell and minimal
strain on the third. Overall, the geocell was more efficient with lower modulus base
materials.
2. Influence of Cover Layer Thickness:
a. The stress (below loading plate) magnitude reduced significantly with a cover
thickness of 3 in. or higher.
b. The hoop strain on the loaded geocell decreased with an increase in cover thickness.
Similar to the base modulus, the strain reduced significantly reduced in adjacent
geocell.
c. The benefit of geocell diminished with a stiffer base cover of more than 4 inches.
3. Influence of Geocell-Reinforced Layer Thickness:
Since the thickness of the geocell reinforce layer depended on the height of the geocell,
the following conclusions are based on geocell height:
a. The three-inch geocell reinforced layer reduced the stresses up to 10-30% at six
inches away from the center of the loading plate and up to 15-60% at nine inches
away from the center of the loading plate compared with no geocell layer.
b. The four-inch geocell reinforced layer reduced the stresses up to 10-25% at six
inches away from the center of the loading plate and up to 0-30% at nine inches
away from the center of the loading plate compared with no geocell layer.
137
c. The six-inch geocell reinforced layer reduced the stresses up to 0-25% at six inches
away from the center of the loading plate and up to 0-30% at nine inches away from
the center of the loading plate.
d. Comparing the three geocell, the geocell 3" performance is inferior to other geocell
heights as it only reduced stresses up to 10% at 6" from the loading plate and 20%
at 9" from the loading plate, whereas geocell 4" and 6" reduced the stresses around
20% at 6" and 40-50% at 9" from loading plate.
e. Geocell 3" has minimal influence in the presence of weak subgrade material.
4. Influence of Subgrade Modulus:
a. It is observed that geocell of 3 in. height has no significant influence on reducing
stresses at the top of subgrade in the presence of lower quality subgrade (4.5 and 2
ksi).
b. The hoop strains on the geocell were minimally influenced by a change in subgrade
modulus (4.5 ksi and 2.0 ksi).
c. The vertical strains and deformation for the 2.0 ksi subgrade (for all geocell) are
higher than the 4.5 ksi subgrade
The FEA results indicate that 4” and 6” geocell heights will be more effective than 3”
geocell. Although the geocell reinforcement increases the modulus of a reinforced layer,
the increase in infill modulus beyond 12 ksi is not as pronounced compared to the infill
modulus of 12 ksi or lower.
II. The following conclusion can be drawn from the Laboratory Evaluation:
1. Vertical Deformation: The magnitude of deformation was similar regardless of reinforced
and unreinforced base layers.
2. Vertical stress distribution on subgrade: The stress magnitude and pattern estimated from
the laboratory tests is similar to that obtained from FEA.
3. The vertical strain on geocell: the strain on geocell increased from 90 to 115 microstrains
by the end of 10,000 cycles indicating that the geocell is barely compressed in the vertical
direction.
4. Geocell Hoop Strains: The geocell below the loading is expanding and transferring the load
in the lateral direction (to the adjacent), and beyond that, the load distribution is almost
negligible.
The FEM and laboratory test results showed similar trends. A maximum difference of 10%
(vertical stress on top of the subgrade) was observed between them, indicating that FEM analysis
can be performed for any new material selected for analysis.
III. The laboratory results indicated similar performance for geocell obtained from two
sources.
IV. Summary of models used for Cross-validation Techniques is shown in the figure below:
V. The performed LCCA resulted in the following:

a. Suppose the geocell reinforced layer cost is between $12 and $15 (depending on
District) per square yard. In that case, the initial construction cost will be lower
138
while providing a similar service life to the conventional design due to a reduced
base layer thickness.
8.3 RECOMMENDATION
The recommendation of this study are as follows:
1. This study recommends using geocell, especially in the weak quality base material
(modulus of 12 ksi or less).
2. Cover thickness (layer over geocell) and quality of cover material; geocell layer thickness;
infill material (in geocell) modulus; and subgrade modulus are four factors that influence
the performance of geocell reinforced pavements. Therefore, this study recommends:
a. A minimum cover thickness of 4 in. is recommended.
b. A minimum of 4” geocell height should be used to optimize the benefits provided
by geocell reinforcement.
c. A cover thickness of 6” is recommended to maximize performance.
3. A site has been instrumented in Paris District, and future performance data will be collected
to evaluate long-term performance.
4. Two to four sites should be constructed and monitored before routine use of geocell can be
recommended.
5. Since the top four inches of the subgrade is scarified and stabilized with lime in districts
with the weak subgrade, the geocell can be placed on top of the scarified surface and
infilled with the recovered material will reduce the thickness of the base layer or required
quality of base layer and will reduce the construction cost.
139
140
9. REFERENCES
1. AASHTO (1993), “AASHTO Guide for Design of Pavement Structures,” AASHTO,
Washington, D.C.
2. Allison, P. D. D. (1999). Multiple regression: A primer. Thousand Oaks, CA: SAGE
Publications.
3. Al-Qadi, I. L., and Hughes, J. J. (2000). "Field evaluation of geocell use in flexible
pavements." Transportation Research Record: Journal of Transportation Research Board,
1709, 26-35.
4. “Average Low Bid Unit Prices.” 2016. Accessed December 15th, 2016.
http://www.txdot.gov/business/letting-bids/average-low-bid-unit-prices.html.
5. Bathurst, R. J., and Knight, M. A. (1998). “Analysis of geocell reinforced-soil covers over
large span conduits.” Computers and Geotechnics, 22(3), 205–219.
6. Bortz, B., and Hossain, M. (2015). Accelerated pavement testing of low-volume paved
Roads with geocell reinforcement. Final Report FHWA-KS-14-14, Manhattan, KS.
7. Emerslebn, A. and Meyer, N. (2008). “The use of geocells in road constructions over soft
soil: vertical stress and falling weight deflectometer measurements.” EuroGeo4 paper
number 132, 8 pgs.
8. EN ISO 13426-1 (2003). “Geotextiles and geotextile-related products- strength of internal
structural junctions, Part 1: Geocells.” 13 pgs.
9. Esbensen, K., Guyot, D., Westad, F., & Houmoller, L. P. (1994). Multivariate data
analysis: in practice: an introduction to multivariate data analysis and experimental design.
Oslo, Norway: CAMO.
10. Evans, M. D. (1994). Geocell mattress effects on embankment settlements. Vertical and
Horizontal Deformations of Foundations and Embankments, ASCE.
11. Feicheng Sanjie Engineering Materials Co., Ltd., China. “Geocell for road construction
specification.” Retrieved April 18, 2011, from http://www.tasanjie.com/english/cp-15.htm.
12. Han, J., Pokharel, S.K., Yang, X., Manandhar, C., Leshchinsky, D., Halahmi, I., and
Parsons, R.L. (2011). “Performance of geocell-reinforced RAP bases over weak subgrade
under full-scale moving wheel loads.” ASCE Journal of Materials in Civil Engineering
accepted.
13. Han, J., Yang, X., Leshchinsky, D., and Parsons, R. (2008). “Behavior of geocell-
reinforced sand under a vertical load.” Transportation Research Record: Journal of the
Transportation Research Board, 2045, 95–101.
14. Haufe, A., Schweizerhof, K., and DuBois, P. (2013). “Properties and Limits: Review of
Shell Element Formulations.” LS-DYNA Developer Forum 2013, DYNAmore GmbH,
Filderstadt, Germany.
15. Hegde, A., and Sitharam, T. G. (2014). “Joint strength and wall deformation characteristics
of a single-cell geocell subjected to uniaxial compression.” International Journal of
Geomechanics, 4014080.
16. Henry, K. S., Olson, J. P., Farrington, S. P., and Lens, J. (2005). "Improved Performance
of Unpaved Roads During Spring Thaw." USACE ERDC/CRREL TR-05-1, Engineer
Research and Development Center, Cold Region Research and Engineering Laboratory,
Hanover, NH.
141
17. James Walls, and Michael R Smith. 1998. “Life Cycle Cost Analysis in Pavement Design-
Interim Technical Bulletin.” FHWA-SA-98-079. Federal Highway Administration.
18. Klisiński, M. (1985). “Degradation and Plastic Deformation of Concrete.” Polish Academy
of Sciences, Institute of Fundamental Technology Research (IFTR) Report 38.
19. Latha, G. M., Dash, S. K., and Rajagopal, K. (2009). “Numerical Simulation of the
Behavior of Geocell-Reinforced Sand in Foundations.” American Society of Civil
Engineers.
20. Latha, G. M., and Somwanshi, A. (2009). “Effect of reinforcement form on the bearing
capacity of square footings on sand.” Geotextiles and Geomembranes, 27(6), 409–422.
21. Lewis, B. A. (2004). Manual for LS-DYNA Soil Material Model 147. FHWA-HRT-04-095,
McLean, VA.
22. Long, S. J., & Freese, J. (2006). Regression models for categorical dependent variables
using Stata (2nd ed.)
23. LSTC. (2016). LS-DYNA Theory Manual. Livermore Software Technology Corporation,
Livermore, CA.
24. Mehdipour, I., Ghazavi, M., and Moayed, R. Z. (2013). “Numerical study on stability
analysis of geocell reinforced slopes by considering the bending effect.” Geotextiles and
Geomembranes, 37, 23–34.
25. Mengelt, M. J., Edil, T. B., and Benson, C. H. (2000). "Reinforcement of Flexible
Pavements Using Geocells." Geo Engineering Report No. 00-04, University of Wisconsin-
Madison, Madison, WN.
26. Mengelt, M.J., Edil, T.B., and Benson, C.H. (2006). “Resilient modulus and plastic
deformation of soil confined in a geocell.” Geosynthetic International, 13 (5), 195-205.
27. Mhaiskar, S.Y. and Mandal, J.N. (1992a). “Comparison of geocell and horizontal inclusion
for paved road structure.” Earth Reinforcement Practice, Ochiai, Hayashi, and Otani.
Balkema, Rotterdam.
28. Mhaiskar, S.Y. and Mandal, J.N. (1992b). “Subgrade stabilization using geocells.” ASCE
Geotechnical Special Publication, 2 (30), 1092-1103.
29. Mhaiskar, S. Y., and Mandal, J. N. (1994). "Three-dimensional geocell structure:
performance under repetitive loads." 5th International Conference on Geotextiles,
Geomembranes, and Related Products, Singapore, 155-158.
30. Mhaiskar, S. Y., and Mandal, J. N. (1996). “Investigations on soft clay subgrade
strengthening using geocells.” Construction and Building Materials, 10(4), 281–286.
31. Neher, H. P., and Lachler, A., “Numerical modeling of a diaphragm wall production
process in Rotterdam compared to monitoring data,” 6th European Conf. on Numerical
Methods in Geotechnical Eng., Graz, Austria, pp. 417-22, 2006.
32. Pokharel, S.K., Han, J., Leshchinsky, D., Parsons, R.L., and Halahmi, I. (2009a).
“Experimental evaluation of influence factors for single geocell-reinforced sand.”
Transportation Research Board 88th annual meeting, January 11-15, 2009, Washington,
DC.
33. Pokharel, S.K., Han, J., Leshchinsky, D., Parsons, R.L., and Halahmi, I. (2009b).
“Behavior of geocell-reinforced granular bases under static and repeated loads.”
International Foundation Congress & Equipment Expo 2009, March 15-19, 2009, Orlando,
Florida.
142
34. Pokharel, S.K., Han, J., Parsons, R.L., Qian, Y., Leshchinsky, D., and Halahmi, I. (2009c).
“Experimental study on bearing capacity of geocell-reinforced bases.” 8th International
Conference on Bearing Capacity of Roads, Railways and Airfields, June 29 - July 2, 2009,
Champaign, Illinois.
35. Pokharel, S.K., Han, J., Leshchinsky, D., Parsons, R.L., and Halahmi, I. (2010).
“Investigation of factors influencing behavior of single geocell-reinforced bases under
static loading.” Journal of Geotextile and Geomembrane, 28 (6), 570-578.
36. Pokharel, S.K. (2010). Experimental Study on Geocell-Reinforced Bases under Static and
Dynamic Loadings. Ph.D. dissertation, CEAE Department, the University of Kansas
37. Pokharel, S.K., Han, J., Manandhar, C., Yang, X.M., Leshchinsky, D., Halahmi, I., and
Parsons, R.L. (2011). “Accelerated pavement testing of geocell-reinforced unpaved roads
over weak subgrade.” Journal of the Transportation Research Board, No. 2204, Low-
Volume Roads, Vol. 2, Proceedings of the 10th International Conference on Low-Volume
Roads, July 24–27, Lake Buena Vista, Florida, USA, 67-75.
38. PRS Mediterranean Ltd., Israel. Stabilizing an unstable world, Neoloy based nanoweb,
Paving the way for tomorrow’s roads. Retrieved on September 20, 2011, from
http://www.prs-med.com/.
39. Presto Geosystems, USA. Retrieved on August 11, 2011, from http://www.prestogeo.com/.
40. Presto Geosystems (2007). “Geoweb load support system, Technical overview,” 19 pgs.
41. Rangaraju, Prasada Rao, Serji N. Amirkhanian, Zeynep Guven, and South Carolina. 2008.
“Life Cycle Cost Analysis for Pavement-Type Selection.” South Carolina Department of
Transportation. http://www.clemson.edu/t3s/scdot/pdf/projects/spr656final.pdf.
42. Russel Lenz.W. 2011. “Pavement Design Guide.” Pavement Design Guide. Texas
Department of Transportation.
43. Reid, J. D., Coon, B. A., Lewis, B. A., Sutherland, S. H., and Murray, Y. D. (2004).
Evaluation of LS-DYNA Soil Material Model 147. FHWA-HRT-04-094.
44. Saleh, M., and Edwards, L. (2011). “Application of a Soil Model in the Numerical Analysis
of Landmine Interaction with Protective Structures.” Ballistics 2011: 26th International
Symposium, E. Baker and D. Templeton, eds., DEStech Publications, Inc., Miami, FL, 256–
275.
45. Schumacher, M., Holländer, N., & Sauerbrei, W. (1997). Resampling and cross-validation
techniques: a tool to reduce bias caused by model building? Statistics in Medicine, 16(24),
2813-2827.
46. Sharma, M., Inti, S., Tirado, C., & Tandon, V. (2016). Evaluating the Benefits of Geocell
Reinforcement of the Base Course in Flexible Pavement Structures Using 3-D Finite
Element Modeling. In International Conference on Transportation and Development 2016
(pp. 728-739).
47. Thakur, J.K., Han, J., Leshchinky, D., Halahmi, I., and Parasons, R.L. (2011). “Creep
deformation of unreinforced and geocell-reinforced recycled asphalt pavements.”
Accepted for publication at the GeoFrontier International Conference 2010, March 13-16,
Dallas, Texas, USA.
48. Thakur, J.K. (2010). Experimental Study on Geocell-Reinforced Recycled Asphalt
Pavement (RAP) Bases under Static and Cyclic Loadings. M.S. Thesis, CEAE Department,
the University of Kansas.
143
49. Uyanık, G. K., & Güler, N. (2013). A study on multiple linear regression analysis. Procedia
- Social and Behavioral Sciences, 106, 234–240. doi:10.1016/j.sbspro.2013.12.027
50. Webster, S. L. (1979a). "Investigation of Beach Sand Trafficability Enhancement Using
Sand-Grid Confinement and Membrane Reinforcement Concepts; Report 1, Sand Test
Sections 1 and 2." Technical Report GL-79-20, Geotechnical Laboratory, US Army Corps
of Engineers Waterways Experimentation Station, Vicksburg, MS.
51. Webster, S. L. (1979b). "Investigation of Beach Sand Trafficability Enhancement Using
Sand-Grid Confinement and Membrane Reinforcement Concepts; Report 2, Sand Test
Sections 3 and 4." Technical Report GL-79-20, Geotechnical Laboratory, US Army Corps
of Engineers Waterways Experimentation Station, Vicksburg, MS.
52. Wimsatt, A.J., Chang-Albitres, C.M., Krugler, P.E., Scullion, T., Freeman, T.J. and
Valdovinos, M.B.. 2009. “Considerations for Rigid Vs. Flexible Pavement Designs When
Allowed as Alternate Bids: Technical Report.” Texas Transportation Institute, Texas A&M
University System.
53. Yang, X. (2010). “Numerical Analyses of Geocell-Reinforced Granular Soils under Static
and Repeated Loads.” The University of Kansas.
54. Yang, X., Han, J., Pokharel, S.K., Manandhar, C., Parsons, R.L., Leshchinsky, D., and
Halahmi, I. (2011). “Accelerated pavement testing of unpaved roads with geocell-
reinforced sand bases.” Geotextiles and Geomembranes, accepted.
55. Kim, Y-R, Hong, Minki, Allen, D.H. & Park, S-W. (2012) “Statistical and dimensional
analysis of hot-mix asphalt mixture characteristics on asphalt pavement analyzer rutting
behavior,” International Journal of Pavement Engineering.
56. Yuu, J., Han, J., Rosen, A., Parsons, R.L., and Leshchinsky, D. (2008). “Technical review
of geocell-reinforced base courses over weak subgrade,” Proceedings of the First Pan
American Geosynthetics Conference & Exhibition, Cancún, Mexico, 2-5 March 2008,
1022-1030.
144
APPENDIX A: SITE INSTRUMENTATION
A.1 INTRODUCTION
An additional task was added to the project to instrument a portion of FM 906 in Lamar County (
Paris, Texas) constructed using geocells. The purpose of instrumentation was to document the
benefits of geocell in pavements. The pavement instrumented included two sections a) no geocell
(NGC) and b) constructed with geocell (GC). In addition, falling Weight Deflectometer (FWD)
testing was performed at the end of the construction to ensure that the instruments were working
correctly. The location of the test section and pertinent information related to construction is
included in this appendix.
A.2 SITE LOCATION AND MATERIAL PROPERTIES
A.2.1 Site Location
The project site for geocell installation is a local road located in Lamar County, Paris, TX, a low-
volume road (FM 906). The location of the road within Texas is shown in Figure A-1, while Figure
A-2 shows the exact location of the test section.
Figure A-1 Lamar County, Paris TX Site Location (Source: Wikipedia).
145
Figure A-2 Test Site Location (Source: ArcMAP).
A.2.2 Soil Classification
At the construction time, subgrade and base material were collected and tested in the laboratory
for soil classification, and results are included in the following sections.
Subgrade Soil Classification:
Subgrade soil was classified using the procedure proposed in Tex-142-E (Laboratory
Classification of Soils for Engineering Purposes). A 500 g sample was used for the soil
classification. The gradation chart is shown in Figure A-3 as represented by a particle–size
distribution curve. Based on the gradation, the soil was classified as a poorly graded soil SP. The
sample classified consisted of 4% gravel, 69% sand, and 27% fines. In addition, Tex-104-E, Tex-
105-E, and Tex-106-E determined Atterberg Limits. The soil's liquid limit (LL) and plastic limit
(PL) were 21 and 0, respectively, having a plastic limit (PL) of 21 (Figure A-4). Therefore, the soil
is classified as poorly graded sand with clay (SP-SC).
146
Figure A-3 Mech. Sieve Soil Classification.
Figure A-4 Liquid Limit.
Texas Tri-axial Test:

Subgrade Material:
Subgrade material was prepared as per Tex-114-E (Part III) and tested by the Accelerated Method
for Triaxial Compression of Soils in Tex-117-E (Part II). A total of 6 specimens at an optimum
moisture content of 10.3% and maximum dry density of 112.7 pcf (Figure A-5) were tested. All 6
specimens were 4 in. in diameter and 6 in. in height. Specimen were prepared for capillary wetting
for 24 hours, after which the 6 specimens were tested at a single lateral pressure of 0 psi, 3 psi, 5
psi, 10 psi, 15 psi, and 20 psi. In addition, Mohr’s diagram and failure envelope were produced for
shear strength evaluation (Figure A-6).
147
Figure A-5 Subgrade Mohr's Circles.
Figure A-6 MD Curve Subgrade Soil.
The measured cohesion and friction angle of subgrade material was 4.7 psi and 44.1°, respectively.
The resultant cohesion of sand was not equivalent as initially expected. Therefore, the specimens
were not subjected to the 21 days mentioned in section 5.12., standard Tex-117-E based on the PI
value. Instead, the accelerated test was performed for these specimens. The angle of internal
friction does apply to this test since the Mohr’s Circles would have been reduced by the capillary
wetting of the 21 days, and consequently, the slope would remain the same. Capillary wetting
148
reduced the shear stresses for Mohr’s Circles and reduced the y-intercept of the Coulomb failure
envelope line (Cohesion).
The resulting data are summarized in Table A-1, as shown below. For example, at 0 psi lateral
pressure, the resulting normal stress comes out to 3.83 psi, and the shear stress at 8.98 psi. At 5 psi
lateral pressure, the failure normal and shear stresses were calculated to be 11.08 psi and 14.24
psi, respectively. Finally, at a maximum lateral pressure of 20 psi, the resulting failure normal and
shear stresses were estimated to be 36.29 psi and 38.15 psi, respectively. Thus, with an increase in
lateral pressure, both normal stress and shear stress increased linearly.
Table A-1 Stress at Different Confinements (Subgrade soil)
Failure Stresses
Lateral Stress Normal Stress Shear Stress
(psi) (psi) (psi)
0 3.83 8.98
0 3.83 8.98
3 8.91 13.84
3 8.10 11.95
5 11.08 14.24
10 20.40 24.36
15 29.36 33.63
20 36.29 38.15
Flexbase material:
The Texas Triaxial Test determined the shear strength of the flexible base material following the
accelerated method (Tex-117-E, Part II). A total of 9, 6”x8” cylinders were prepared for three
different lateral confining pressures. Three prepared specimens were tested under zero confining
pressure, three at 3.0 psi confining pressure, and the remaining three were tested under a 15 psi
confining pressure. These samples were prepared in accordance with Test Method Tex-113-E,
using a Soil Compactor Analyzer (SCA). Specimens were intended to be mold at their Optimum
Moisture Content (OMC) of 6.1% and a maximum dry density of 132.5 pcf (Figure A-7). Once
the material was molded, extruded, and allowed to sit for 24 hours while monitoring moisture
travel from the specimen to the porous stones, if any. Based on Mohr circles results (Figure A-8),
the flex base classification, internal friction angle (ϕ), and cohesion were found. Therefore, flex
base classification was classified under Class 1 (Figure A-9).
149
Figure A-7 MD Curve of Flexbase Material.
Figure A-8 Flex-Base Mohr Circles.
150
Classification Chart
60
50
40
Shear Stress (psi)
Class 1
Class 2
30
Class 3
20
Class 4
10
Class 5
Class 6
0
0 5 10 15 20 25 30
Normal Stress (psi)
Figure A-9 Flex base classification
Average failure stresses occurred at different lateral stresses, as shown in (Table A-2). The
minimum average shear stress at failure (27.14 psi) transferred at an average normal stress of 11.46
psi with 0 psi lateral stress. The highest failure shear stresses occurred at 15 psi confinement with
the average normal stress of 37.42 psi and average shear stress of 53.11.
Table A-2 Average Stresses at Different Confinements
Failure Stresses
Lateral Stress Normal Stress Shear Stress
(psi) (psi) (psi)
0 11.46 27.14
3 16.61 32.23
15 37.42 53.11
151
A.3 INSTRUMENTATION
Geocell: The geocell manufacturer used in this study is Presto 4.25” geocell. The properties of the
selected geocell are shown in Appendix B (Figure B-2).
Strain Gauge: The strain gauges are used to measure the deformation occurring due to loading
and are converted to estimate the induced strains. Each strain gauge is glued to geocell and
adequately protected. Table A-3 shows the strain gauge model and protection means. The
specification of the strain gauge is mentioned in Appendix B (Figure B-3).
Table A-3 Strain Gauges
Item Description
Strain Gauge KFH-6-120-C1-11L1M3R (6mm strain gauge, 120Ω,
3 pre-wired)
Glue (to glue strain gauge to Ethyl based cyanoacrylate
geocell)
Protection of strain gauge Performix Plasti Dip (flexible protection)
Protection of strain gauge wires PVC Tubing
Earth Pressure Cells: The pressure cells were placed on the top of the subgrade to estimate the
stresses at the subgrade level. Geokon Model 3500 series (2.5 MPa and 600 kPa) pressure cell was
selected for evaluation, and the specifications of these pressure cells are illustrated in Appendix B
(Figure B-4). Each pressure cell is a semiconductor strain gauge earth pressure cell (circular 10”),
with the thermistor in SS housing, 0-5 VDC output as shown in Figure A-10.
Figure A-10 Photo of Earth Pressure Cell.
A trench of 1’ x 1’ x 3” is required to fit the front portion of the pressure cell and 3” to 5” along
the road for keeping the wiring safe.
Pavement Section: The FM 906 constructed with geocell and pavement profile is shown in
Figures A-11 and A12, respectively. The pavement profile consisted of subgrade, 2.0 in. of the
salvaged base, and 8 in. of the cement-treated base. In addition, a portion of FM 906 eastbound
152
was left untreated. Approximately 125 ft. of the untreated section was constructed without geocell,
while another 250 ft. was constructed with geocell reinforcement. To maintain a similar profile,
the unreinforced section consisted of 10 in. of the flexible base. In contrast, the reinforced section
consisted of 4 geocells filled and covered (6 in.) with the flexible base material.
Figure A-11 Pavement Section of Geocell-Reinforced at Testing Site.
Figure A-12 Pavement Section of FM 906 at Testing Site.
The location of sensors for the unreinforced section is shown in Figure A-13, while Figure A-14
shows the wiring arrangement in the test section. The placement of the pressure cell at the
unreinforced section is shown in Figure A-13c.
153
Figure A-13 Cross Section and Instrumentation of FM 906 at Testing Site for
Unreinforced Section.
Figure A-14 Wiring Arrangement at FM 906 at Testing Site for Unreinforced Section.
Figure A-15 represents the cross-section of the geocell reinforced section of the pavements at the
test site. The pressure cell and strain gauge placement at the geocell reinforced section is shown in
Figure A-15c. Figure A-16 shows the wiring arrangement in the test section.
154
Figure A-15 Cross Section and Instrumentation of FM 906 at Testing Site for Geocell
Reinforced Section.
Figure A-16 Wiring Arrangement at FM 906 at Testing Site for Geocell Reinforced
Section.
Data Acquisition System: For recording the stresses and strains of the pavement system, the LMS
data acquisition system was employed because the system can accommodate 16 channels, i.e., it
can record stresses and strains from 16 locations at a time. In addition, it can record the data at
different frequencies (Figure A-17). In this study, a frequency of 128 data points per second is
chosen, close to the frequency of other data acquisition systems.
155
Figure A-17 LMS Data Acquisition System.
A.4 CONSTRUCTION OF REINFORCED AND UNREINFORCED SECTIONS
Before the instrumentation of sites and construction, the site was visited, and a plan for placement
was developed based on the discussion in the previous section. The UTEP research team arrived
and left the construction site on July 10th and July 13th, respectively. On the first day of arrival, the
site was surveyed to identify and mark sensor locations. In addition to location marking, the
construction process was also discussed with the contractors.
Instrumentation of Unreinforced Section: On July 11th, the contractor provided the traffic
control and started removing the un-stabilized flexible base material to a depth of 10 in. from the
surface and placed the removed material on the westbound lane for the movement of traffic. After
material removal, the sensor locations were marked (Figure A-18), and a trencher was employed
to dig a trench to place pressure cells and wiring. The pressure cells were placed on the well-
compacted leveled surface and covered with playground sand (Figure A-19) before placing
flexible base material. To ensure proper contact, the surface below the pressure cell was compacted
using a sledgehammer and leveled to minimize settlement of pressure cells due to compaction,
thus, minimizing damage to the pressure cell.
156
Figure A-18 Marked Sensor Location for Unreinforced Section.
157
Figure A-19 Installation of Pressure Cells.
After installing pressure cells, the wires were placed inside PVC pipes and were brought to a box
at the side of the road for data collection. In the end, the contractor was allowed to place and
compact the flexible base material on top.
158
Instrumentation of Geocell-Reinforced Section: On July 12th, the placement of the geocell
reinforced section was performed. The construction sequence similar to the unreinforced was
followed for the placement of pressure cells. After the placement of pressure cells, a geosynthetic
fabric (Figure A-20) was placed on top of the pressure cells to minimize the migration of material.
After placement of geosynthetic, the instrumented geocells were placed on top, as shown in Figure
A-21. The geocells came in panel form (Figure A-20) that can be stretched and connected (Figure
A-21). Since geocells were placed in half road width, the standard panels were cut and connected
to the standard panel. The connection of geocells is simple using the Atrakey provided by the
supplier and requires minimal connection time. To make sure that the geocells were stretched, the
geocells were stretched, and rebar was placed on maintaining the stretched condition. After placing
instrumented and non-instrumented geocells, the strain gauges wires were placed inside the pipes
and moved towards the control box together with pressure cell wires.
Figure A-20 Installation of Geosynthetics.
159
Figure A-21 Installation of Instrumented Geocells.
160
After the placement of geocells, the rebars were removed, and pegs were placed to maintain the
width of geocells. Next, the contractor started moving material and filling geocells closer to the
median and continued till it reached the shoulder (Figure A-22).
Figure A-22 Geocell Construction Sequence.
A.5 DATA COLLECTION AND ANALYSIS
After construction and compaction of the geocell reinforced layer, the UTEP research team went
to the site on July 13th to verify the working of pressure cells and strain gauges. For verification,
Falling Weight Deflectometer (FWD) was requested by the Paris District office. The FWD testing
was performed on top of the geocell reinforced, unreinforced, and cement-treated sections. The
purpose of FWD tests was to measure the response of pressure cells and strain gauges and obtain
a section profile (of three pavement sections). To measure profile, the FWD tests were performed
at every 20 or 25 ft. by dropping two seating loads and two drops for measurement. To measure
the response of pressure cells and strain gauges, the tests were performed by placing the FWD load
plate on top of the pressure cell. To ensure the FWD load plate is placed on top of the pressure
cells, the pressure cells locations were marked on the surfaces, and the FWD operator was guided
to place the load plate on top of the pressure cells as close as possible. Although efforts were made,
the FWD load plate was not precisely aligned on top of the pressure cells at many locations. Each
FWD test included two seating drops and two measurements with approximately 9,000 lbs. of the
load. Additionally, the FWD tests on the top of the sensors were repeated by moving and
161
repositioning the FWD load plates. A typical response by pressure cell with the drop is shown in
Figure A-23.
Figure A-23 Example of pressure cell result from one FWD drop.
To evaluate and validate the data obtained from pressure cells, the profile data obtained from the
FWD testing was entered in the MODULUS software to estimate the modulus of the base,
subgrade, and influence of reinforcement on the base modulus. The modulus values obtained from
the MODULUS software were then entered in the BISAR program to identify induced pressures
at the pressure cell due to applied FWD load. The modulus values obtained from MODULUS
software for unreinforced pavement section, reinforced pavement section, and Cement Treated
(CT) section are included in Table A-4. It can be seen that the CT pavement section is exhibiting
significantly higher modulus values in comparison to the reinforced section. However, the
reinforced layer exhibits a higher modulus in comparison to the unreinforced pavement section.
The data in Table A-4 suggests that the subgrade modulus is roughly 12 ksi for the CT base section,
11 ksi for the unreinforced section, and 8.0 ksi for the geocell reinforced section. The base layer
modulus is roughly 115 ksi for the CT section, 33 ksi for the unreinforced section, and 22 ksi for
cover (on top of geocell) for the geocell reinforced section. The reduction in modulus values for
the geocell reinforced section can be attributed to the lower level of compaction allowed during
the construction process. Additionally, the CT layer had been compacted more than expected due
to the construction process followed. The 4.0 in. the geocell reinforced layer had an average
modulus of 76 ksi, indicating a benefit of geocell reinforcement. The future testing of the
instrument section will govern the performance of the geocell reinforced layer.
162
Table A-4 Results of MODULUS Software
Modulus, psi
Cement Treated Unreinforced Geocell Reinforced
6 in. 4 in.
10 in. Base Subgrade 10 in. Base Subgrade Subgrade
Cover Geocell
Average 115,068 12,328 32,657 10,809 22,124 75,696 7,877
Minimum 105,980 11,697 8,073 9,807 17,144 55,061 5,946
Maximum 122,973 12,614 48,407 11,624 27,954 91,886 10,854
To estimate pressure experienced by the pressure cell at the subgrade level due to FWD load, the
BISAR software was used. The base modulus of 33 ksi and subgrade modulus of 11 ksi was
assumed to calculate the pressure values. The pavement structure in Bisar software was set as 10-
inch of the top layer and 200-inch of subgrade, and the stress on pressure cell was assumed to be
buried 3.0 in. below the base layer. Thus, the pavement layer on top of the pressure cell was
assumed to be 13 in. The top layer and subgrade layer Poisson’s ratio was assumed to be 0.33 and
0.35, respectively. The FWD load plate was assumed to be 10.82 in., and load values were entered
as per the collected data. The results of the BISAR output and pressure measured from pressure
cells are summarized in Table A-5 indicate that the pressure cells are working as expected while
some of the strain gauges are not functioning because of damage during the construction process.
In some instances, pressure cells estimated higher, lower, or similar results to the estimated
pressures obtained from BISAR. The lower values can be attributed to misalignment of the FWD
plate; the higher values can be attributed to construction equipment passing by while FWD testing
was performed. Additionally, the compaction of layers was also occurring due to the movement of
construction equipment which further contributed to the discrepancy. Based on the results, it can
be concluded that the pressure cells and strain gauges are functioning with few exceptions.
Table A-5 Results of Field Testing and Bisar Modeling
ID NGPC (psi) GPC (psi) GSG (µε) Bisar (psi)
6-1 11.34 4.76 - 19.63
6-2 9.04 1.42 - 18.23
5-1 48.82 3.68 112.60 18.23
5-2 44.80 11.53 351.27 16.74
4-1 22.96 13.22 63.78 18.42
4-2 12.87 13.67 235.75 17.12
3-1 31.57 2.10 179.89 18.89
3-2 23.51 13.47 188.05 17.67
2-1 45.39 76.75 85.21 17.86
2-2 28.82 83.58 85.99 17.12
1-1 4.34 28.67 133.50 18.79
1-2 3.28 31.01 169.00 17.11
Note NGPC-Non-Geocell Pressure Cell; GPC-Geocell Pressure Cell; GSG-Geocell Strain Gauge; “-” means the data
was unavailable.
163
A comparison between no geocell and geocell reinforced n the comparison of pressure cell
response; as shown in Figure A-24, most of the pressure cells of Geocell sites had lower stress
response than the Non-Geocell sites except pressure cells 1 and 2. Although reported, some
numbers are significantly higher and impossible because the applied pressure was around 80 psi
on top and is expected to be lower 13 in. below the surface. The possible reason for discrepancy
can be attributed to interference caused by construction equipment. The deflections measured near
pressure cells 3 and 5 for the geocell, reinforced section, and the unreinforced section are
summarized in Figures A-25 and A-26, respectively. The overall results suggest that the magnitude
of deflections is lower in the presence of geocell reinforcement. However, future performance will
provide a better indicator of the benefit of geocells.
90
Vertical Stress
80
Pressure Cell Response (Psi)
Non-Geocell Site
70
60 Geocell Site
50
40
30
20
10
0
PC 6-1
PC 6-2
PC 5-1
PC 5-2
PC 4-1
PC 4-2
PC 3-1
PC 3-2
PC 2-1
PC 2-2
PC 1-1
PC 1-2
Figure A-24 Comparison of Pressure Cell Response between Non-Geocell and Geocell Sites.
164
Deflection of Spot 3
D1 D2 D3 D4 D5 D6 D7
0
-10
-20
-30
Deflection (mils)
-40
-50
-60
Non-Geocell
-70
Geocell
-80
-90
Figure A-25 FWD Geophone Response of Non-Geocell and Geocell Spot 3.
Deflection of Spot 5
D1 D2 D3 D4 D5 D6 D7
0
-10
-20
Deflection (mils)
-30
-40
-50 Non-Geocell
-60 Geocell
-70
Figure A-26 FWD Geophone Response of Non-Geocell and Geocell Spot 5.
165
166
APPENDIX B: FINITE ELEMENT MODEL
DEVELOPMENT
The Finite Element (FE) numerical analysis method was used to study the behavior of geocell-
reinforced pavement structures under traffic loading. The finite element analysis of geocell
reinforced soils requires truly three-dimensional models because of the all-around confinement of
soil by geocell pockets. For this purpose, a 3-D FE model was developed to address better the
geometry of geocell panels that expand into a honeycomb formation when placed on-site, as shown
in Figure B-1.
Figure B-1 Geocell and Infill Material.

Although FE analyses (FEA) can identify the level of reinforcement provided by the geocell, the
generation of a mesh for FEA is complicated due to several factors like the interaction between
geocell and adjacent soil, transfer of load, and confinement provided by the geocell, among others.
Additionally, the modeling of geocell required a significant number of elements and nodes to
model the honeycomb shape of geocell, which requires significant computational time. Therefore,
to develop a 3-D FE model (FEM) that better addresses the needs imposed by the characteristics
of the geocell-reinforced pavement, distinctive FEMs with different levels of sophistication were
developed before the development of the final 3-D model. These models were developed to
evaluate the following aspects:
• Soil material model
• Boundary conditions of reinforced-layer
• Shape of geocell
• Shell element type
• Geocell-soil interaction
To perform FEA, a general-purpose finite element program LS-DYNA was selected because this
program allows dynamic FEA and includes a comprehensive list of material and contact
models/algorithms. Moreover, the program can also be installed on the High-Performance Cluster
(HPC), a computer system that groups class Linux clusters and symmetric shared-memory
multiprocessor systems that significantly improve simulation program speed performance. The
167
HPC allows executing parallel programs or multiple instances of the same program, each driven
by a different parameter set.
B.1 SOIL MODELS
Many researchers modeled the geocell and infill material as a composite material using FEM or
finite difference methods (Bathurst and Knight 1998; Latha et al. 2009; Latha and Somwanshi
2009; Mhaiskar and Mandal 1996), but very few have modeled them as a separate material (Bortz
and Hossain 2015; Evans 1994; Han et al. 2008; Yang 2010). In the geocell reinforced base layer,
the infill material (linear elastic or plastic) and geocell (elastic) respond simultaneously to loading,
but the working mechanism of each material is different. Therefore, to model a geocell reinforced
base layer more precisely, the behavior of each material (infill and geocell) needs to be evaluated
separately and was utilized in this study.
The mechanical behavior of geomaterials may be modeled at various degrees of accuracy. Hooke's
law of linear elasticity, the most straightforward stress-strain relationship, may not provide
adequate responses to represent soil behavior under traffic loading adequately. Researchers have
proposed many constitutive models to describe various aspects of soil behavior. However, the
more sophisticated a soil model is, the more input parameters are necessary, requiring additional
laboratory tests. A limited number of models use parameters obtained from traditional laboratory
tests. Among these, material models like Mohr-Coulomb and Duncan-Chang models have been
used by researchers to simulate permanent deformation under repeated loading and require
standard tri-axial soil tests (Han et al., 2008; Yang 2010). To simulate the behavior of geomaterials,
three soil models were selected based on their compatibility with LS-DYNA software.
B.1.1 Linear Elastic
The linear elastic model is used to describe materials that behave as follows:
1. The strain in the material is minor (linear).
2. The stress is proportional to the strain, σ ∝ ε (linear).
3. When the loads are removed, the material returns to its original shape, and the unloading
path is the same as the loading path (elastic).
4. There is no dependence on the rate of loading or straining (elastic).
The stress-strain (loading and unloading) curve for the linear elastic is shown in Figure B-2. Finite
element programs only require the modulus of elasticity and Poisson’s ratio as input for
mechanical properties.
The properties considered for the parametric study for both the base and subgrade materials are
shown in Table B-1. In all cases, the base material properties were kept the same for both the top
unreinforced base material and the infill material of the geocell-reinforced layer.
B.1.2 Mohr-Coulomb
With the increase in loads, the soil behavior changes from linear to non-linear, as shown in Figure
B-3. The Mohr-Coulomb model is a well-known and straightforward linear elastic-perfectly plastic
model, which can be used as a first approximation of soil behavior. The linear elastic part of the
Mohr-Coulomb model is based on Hooke’s law of isotropic elasticity. The perfectly plastic part is
168
based on the Mohr-Coulomb failure criterion. This model has been used by Han et al. (2008) and
Yang (2010) for modeling infill material.
Figure B-2 Stress-Strain Relationship for Linear Elastic Material Model.
Table B-1 Material Properties Used for Linear-Elastic Model.

Layer Property Layer
Base Subgrade
Modulus (ksi) 5 30 50 5
Poisson’s Ratio, ν 0.35
Unit Weight (lb/ft3) 115 147 155 115
Peak Perfectly
Plastic
a b
Residual
Stress
Work
Softening
Strain
Figure B-3 Stress-Strain Relationship with Plastic Behavior.
The linear elastic perfectly-plastic Mohr-Coulomb model requires a total of five parameters, listed
as follows:
• Young’s Modulus (E)
• Poisson’s Ratio (ν)
• Cohesion (c)
• Friction angle (φ)
• Dilatancy angle (ψ)
169
To perform the parametric study, the Mohr-Coulomb constitutive material model required material
properties obtained from the literature using typical values for granular and clayey materials and
shown in Table B-2. Three different base materials and two subgrade materials were considered.
It must be mentioned that base modulus values were deliberately selected lower because it has
been documented that the geocell reinforcement seemed to be more efficient in materials with low
moduli and was also one of the objectives of this study.
Table B-2 Material Properties Used for Mohr-Coulomb Material Model.
Layer Modulus Poisson's Cohesion Friction Unit Weight
(ksi) Ratio (psi) Angle, ϕ (pcf)
Subgrade 1 5 0.35 0.725 40° 146
Subgrade 2 5 0.35 15.167 20° 155
Base 1 5 0.35 0.725 40° 186
Base 2 10 0.35 0.725 40° 186
Base 3 15 0.35 0.725 40° 186
B.1.3 FHWA Soil Constitutive Model
This model is a modified Mohr-Coulomb model available in LS-DYNA that was extended to
include excess pore-water effects, strain softening, strain hardening, strain-rate effects, and
elements deletion (Lewis, 2004; Reid et al. 2004). These enhancements to the standard soil
material models were made to increase accuracy, robustness, and ease of use.
The modified yield surface is a hyperbola fitted to the Mohr-Coulomb surface. At the crossing of
the pressure axis (zero shear strength), the modified surface is smooth and perpendicular to the
pressure axis. The yield surface is given as
− P sin φ + J 2 K (θ ) + ahyp 2 sin 2 φ − c cos φ

2
F=
B-1
where P is pressure, φ is the internal friction angle, c is cohesion, J2 is the second
invariant of the stress deviator, ahyp is a Drucker-Prager hyperbolic coefficient parameter for
determining how close to the standard Mohr-Coulomb yield surface the modified surface is fitted
and is defined as
c
ahyp = cot (φ )
20
B-2
K(θ) is the Klisiński modified Mohr-Coulomb function of the angle θ in a deviatoric plane
(Klisiński 1985), defined as
K (θ ) =
( )
4 1 − e 2 cos 2 θ + ( 2e − 1)
2
( ) ( )
2 1 − e 2 cos θ + ( 2e − 1) 4 1 − e 2 cos 2 θ + 5e 2 − 4e
B-3
170
where e is an eccentricity parameter describing the ratio of triaxial extension strength to triaxial
compression strength responsible for third invariant (J3) effects, ranging 0.5 < e ≤ 1.0, and initially
modeled as e = 0.7, and angle θ obtained from:
3 3J 3
cos 3θ =
2 J 23/2
B-4
where J3 is the third invariant of the stress deviator (Lewis, 2004), if ahyp is input as zero, the
standard Mohr-Coulomb surface is recovered.
As shown in Figure B-4, the modified yield surface is a hyperbola fitted to the Mohr-Coulomb
surface. At the crossing of the pressure axis (zero shear strength), the modified surface is smooth
and perpendicular to the pressure axis.
Shear Stress
Standard Mohr-Coulomb
Klisiński's Modified
Pressure
Figure B-4 Comparison of Mohr-Coulomb Yield Surfaces in Shear Stress-Pressure Space.
The properties used for evaluating the FHWA material model are shown in Table B-3. Four
different types of base materials, labeled from 1 to 4, i.e., from stiffer to less stiff materials, and
two subgrade materials were used for the parametric study. The properties for each material were
determined using Tex-117-E Test Procedure for Triaxial Compression for Soil and Base Materials
following Reid et al. (2004) and Saleh and Edwards (2011) for samples collected at different sites.
Table B-3 Soil Properties Used for Modeling Base and Subgrade Used for Parametric
Study and for Evaluation of Geocell Element Types and Contact
Soil Layer ID Modulus Poisson's Cohesion Friction Unit Weight
Type (ksi) Ratio (psi) Angle, ϕ (pcf)
Clay Subgrade 1 1.74 0.30 1.450 15° 100
Clay Subgrade 2 2.32 0.30 1.450 15° 95
Sand Base 1 5.80 0.30 0.015 34° 127
Sand Base 2 8.70 0.30 0.145 29° 124
Sand Base 3 11.6 0.30 0.000 29° 127
Sand Base 4 17.4 0.30 0.000 29° 127
Since previously documented studies used FLAC-3D rather than LS-Dyna, a FEM model was
developed with similar pavement structure and mechanical properties as the University of Kansas
for Kansas DOT. Although the Kansas study used the Duncan-Chang constitutive material model
for modeling the infill material, the properties utilized in their study were used for calculating the
parameters for the FHWA soil model. The pavement structure and material properties used in the
FHWA soil model are summarized in Table B-4 to compare the results with the University of
Kansas study (Yang 2010).
171
B.2 SHELL ELEMENT TYPE FOR GEOCELL MODELING
One of the problems encountered while modeling the geocell and geomaterials is the element
aspect ratio or the computational constraints. Although geomaterials can be modeled as solid
elements of any shape and size, the geocell can be modeled in specific shapes and sizes because
of their thickness. It is entirely possible to create FEM of both geocell as well as geomaterials of
the same thickness. However, this will significantly increase the number of elements which will
require a significant computational time. On the other hand, if geomaterial element and geocell
element thicknesses are different, the compatibility of two different element types and transference
of deformation between the two materials must be evaluated and understood. To model thin
geocell, various thin element types available in LS-DYNA were explored. The available elements
can be categorized as per their behavior to be: as a membrane and as a plate (or shell) elements.
Shell elements can sustain loading using bending stresses, unlike membrane elements. LS-DYNA
has several formulations for shell elements that meet this requirement. Since some formulations of
thin and thick shell elements meet the model's needs, two were selected for evaluation.
Table B-4 FHWA Soil Material Properties Used for Evaluation of the University of Kansas
Study (after Yang. 2010)
Layer Soil Thickness Modulus Poisson's Cohesion Friction Unit
Type (in.) (ksi) Ratio, ν (psi) Angle, Weight
ϕ (pcf)
Top Base AB-3 3.5 5.8 0.35 0.682 47.2° 142
(Unreinforced)
Bottom Base Kansas 6.0 0.48 0.35 0 41.1° 114
(Reinforced) River
Sand
Subgrade Clay 40 1.5 0.35 15.2 0° 124
B.2.1 Shell Element
Among the list of thin shell element types available in LS-DYNA, the Belytschko-Lin-Tsay Shell
(BLT) formulation is recommended as it is more computationally efficient (faster) than the
Hughes-Liu shell element formulation (LSTC 2016). The BLT formulation is based on the
Reissner-Mindlin kinematic assumption, which combines co-rotational and velocity-strain
formulation. The BLT element formulation is suitable for four nodes (quad) shell elements and
offers a single-point integration with hourglass control. Each element has five local degrees of
freedom (DOF) per node (dx, dy, dz, rx, ry) and two (through-thickness) integration points by
default (Figure B-5a). In addition, they have a bi-linear nodal interpolation.
172
Figure B-5 Representation of (a) Four-Node (quad) BLT Shell Element with 5 Local DOFs,
1 Integration Point in the Plane and 5 Through-Thickness Integration Points, and (b) Eight
Node Thick Shell Element with 5 Local DOFs, Single (green) or Reduced (red) Integration
Points in the Plane and 5 Through-Thickness Integration Points.
B.2.2 Thick Shell Element (TSHELL)
The other shell formulation evaluated was the thick shell elements (TSHELL). Unlike thin shell
elements, thick shell elements, also called isoparametric solid-shell elements, consist of eight-
noded elements like solid brick elements (see Figure B-5b) that have enhancements based on the
Hughes-Liu and the BLT shell types formulations. Like BLT elements, they also have 5 local DOF
per node (dx, dy, dz, rx, ry, rz) and can have a one-point quadrature (used as default) or selected
reduced integration in the shell plane, and two through shell integration points used as default.
Three-dimensional constitutive material models can be applied directly to these types of elements.
B.2.3 Geocell Dimensions and Properties
Three types of geocell were simulated in the parametric study with different heights. Table B-5
shows the dimensions and material properties corresponding to the geocell types evaluated.
Geocell were modeled as linear elastic materials.
Table B-5 Geocell Dimensions and Properties.
Geocell Type Presto Tenax 3/200 Tenax 6/200
GW20V and 4/200
Geocell Dimensions
Longitudinal 9.2 8.0 8.0
Length (in.)
Transversal Length 7.9 8.0 8.0
(in.)
Height (in.) 4.0 and 6.0 3.0 and 4.0 6.0
Thickness (in.) 0.040 0.040 0.040
Material Properties
Density (lb/ft3) 59.3 59.3 59.3
Modulus (ksi) 60.0 60 60
Poisson’s Ratio 0.45 0.45 0.45
173
B.3 CONTACT MODEL
One of the most critical aspects of understanding the behavior of geocell-reinforced pavements
comes from the interaction between the geocell and the surrounding geomaterials. In the
composite case, the modeling of geocell reinforcement becomes significantly simplified when a
fully bonded model is considered. In a fully bonded model, shell nodes belonging to the geocell
reinforcement are shared with solid elements representing the host infill base material. Thus, the
solid elements (i.e., the base material) constrain the embedded geocell's translational degrees of
freedom. This approach has been followed by Bortz and Hossain (2015) using geocell modeled
as a shell in an embedded region. This treatment of the contact issue considerably simplifies the
modeling. Still, it fails to adequately address the interaction between the soil and the geocell, i.e.,
no separation or sliding between the geomaterial and the geocell is allowed. Other authors have
preferred to include a shear stress-strain interface relationship based on the Mohr-Coulomb sliding
criterion (Yang 2010; Mehdipour et al. 2013; Hegde and Sitharam 2014). The advantages offered
by this type of interface consist of faster execution times and somewhat simplified meshing.
To better address the inclusion of the geocell within soil material, a contact model is needed. LS-
DYNA allows the insertion of different types of contact models. For modeling the geocell-soil
interface, both the Automatic Surface-to-Surface and the Automatic Single Surface contact types
and the use of springs were considered. These types of models allow sliding and friction between
the soil and geocell.
B.3.1 Automatic Single Surface Contact
Automatic contact models require the definition of a slave material and master material. The latter
usually is the stiffer material. Each slave node is checked for penetration through the master
surface, as shown in Figure B-6. The Automatic Single Surface contact type is among the most
widely used contact options in LS-DYNA. The slave surface is typically defined as a list of parts
and, unlike Automatic Surface-to-Surface contact type, no master surface is defined. Instead,
contact is considered between all the parts in the slave list, including the self-contact of each part.
Sliding permitted
Node penetration check
Figure B-6 Automatic Single Surface Contact to Model Geocell Soil Interface
If the slave node penetrates, an interface force fs is applied between the slave node and its contact
point. The magnitude of this force is proportional to the amount of penetration that can be thought
of as the addition of an interface spring. The interface force is a function of a stiffness factor ki and
penetration l of slave node ns through master segment si given as
fs = −lk i ni if l < 0
B-5
174
where ni is normal to the master segment at the contact point. Stiffness ki for master segment si is
defined as
For brick elements
f si K i Ai2
ki =
Vi
B-6
For shell elements
f si K i Ai
ki =
max(ldiag )
B-7
where Ki is the bulk modulus, Vi is volume, Ai is the face area of the element in si, ldiag is the shell
diagonal length, and fsi is a scale factor for the interface stiffness (normally fsi = 0.10) (LSTC 2016).
Friction is based on a Coulomb formulation. The model implements a friction algorithm that makes
use of an elastic-plastic spring. The algorithm is based on an iterative process that starts by
calculating the yield force based on the friction and the normal force, followed by the calculation
of the incremental movement of the slave node to update the interface force and check yield
condition. Finally, an exponential interpolation function smooths the transition between the static
and dynamic coefficients of friction based on the relative velocity between the slave node and the
master, as described in the LS-DYNA theory manual (LSTC 2016).
B.3.2 Discrete Beam Element Interface
The use of discrete beam elements for establishing a geocell-soil interface was also considered.
For the implementation of springs, spring elements that consider normal and tangential
components are needed. i.e., the normal component dn, and two other orthogonal components ds
and dt, as shown in Figure B-7. For this purpose, discrete elements can be used. These elements
are point-to-point physical connections between deformable bodies. As springs (discrete elements)
have only one DOF, i.e., a normal component, discrete beams were used instead as they have up
to 6 DOFs. Resultant forces and moments of a discrete beam are output in the local (r,s,t)
coordinate system. Moreover, the length of a discrete beam may be zero or nonzero.
Figure B-7 Contact Using Discrete Beams (Spring with Normal and Tangential
Components).
175
B.4 NODE COMPATIBILITY FROM GEOCELL AND GEOMATERIAL AT LAYER
INTERFACE
Using thick shell elements to simulate the geocell embedded within soil required special treatment
in modeling the layer interface. As thick shell elements are analogous to solid elements, the node
connectivity between thick and thin geocell wall thickness becomes an issue. Figure B-8a
illustrates the region at the interface of the top cover material and the base material for an
unreinforced pavement, indicating where the thick shell elements will be inserted to simulate the
geocell reinforcement. To accommodate the thick shell elements, a gap was created slightly wider
than the geocell thickness to insert the geocell element. As a result, elements with the same gap
thickness propagated towards the surface were created, as shown in Figure B-8b. A similar
approach was followed for the base to subgrade interface. The thin solid elements were bonded
with the solid elements (soil), while spring elements connected the thick shell elements to the solid
elements. With this approach, it was possible to allow vertical movement of the geocell and see its
effect on the material placed on top and bottom of the geocell.
Figure B-8 (a) Unreinforced Pavement Indicating Location Where Geocell must be
Inserted and (b) Mesh Transition at the Layer interface from Thick Shell (TSHELL)
Elements to Solid Elements.
176
B.5 BOUNDARY CONDITIONS
A preliminary study was performed to evaluate the effect of boundary conditions on geocell
reinforcement behavior and whether the change in boundary conditions would destabilize FEA
under repeated loading conditions. For this purpose, a mesh optimization was conducted to ensure
the model's limits, i.e., boundary conditions were placed far enough where stress and deflection
values were less than 10% of the peak values obtained under the load. The identified location was
18 in. away from the point of application of the load. Though models with wider dimensions could
be developed, the model must be constrained in size due to a large number of nodes and elements
generated since the size of the elements on the pavement layers is determined by the size of the
elements along the geocell. About 15 elements were used along the geocell along the half-sine-
shaped outline, i.e., from joint to joint, causing many elements to be generated within the infill
material, all adapted to the size of the elements on the geocell. The number of nodes and elements
used in the three-dimensional models is described in the sections referring to the developed
models.
In addition, it was desired to design the FE model with a reinforced layer that allows the expansion
of the geocell panels. For this purpose, boundary conditions were removed just for the geocell-
reinforced base layer, while the other layers were restrained from lateral movement. A
representation of the boundaries conditions is shown in Figure B-9. While evaluating the material
constitutive models for simulating the geomaterials, it was found that no instability occurred due
to the removal of lateral constraints on the nodes at the edge of the base layer.
Figure B-9 Side View of FE Model Showing Boundary Conditions, Shown as Triangles,
with Base Layer with no Restrained Conditions to Allow for Expansion.
177
B.6 LEVEL OF SOPHISTICATION OF FINITE ELEMENT MODELS
Three different levels of sophistication of the FEM were developed. First, a simplified 3-D FEM
was developed consisting of a single layer to evaluate the most appropriate shell element and
contact model for modeling the geocell embedded in geomaterial. A second 3-D FEM was
developed consisting of a pavement section with a single reinforced cell to evaluate the contact
type compared to the responses observed in the laboratory after a shell element type was selected
from the first simplified model. Finally, an extension of this latter model was developed to simulate
pavements reinforced with the honeycomb-type geocell reinforcement. This model was used to
carry a parametric study to evaluate the behavior of geocell reinforcement in different pavements.
B.6.1 Simplified Single Layer FE Model
The 3-D model was reduced to a 2-D (one-thick element wide) single-layer model to identify the
most appropriate contact model. Additionally, this model was also used for identifying the type of
shell element that best addresses the behavior and interaction of the geocell with geomaterial. This
model was developed to check the efficiency of the shell elements for transferring the stress from
one cell to the adjacent one.
The model consisted of a single layer with properties of base course material, divided into three
infill cell pockets separated by the geocell reinforcement, as shown in Figure B-10. The model
consisted of 480 elements, divided into three parts representing the cells, and loaded with a 6-in.
diameter, 40 psi pressure load directly applied to Cell #1. Symmetry conditions were applied on
the left boundary to account for a half-model. The applied pressure of 80 psi was selected to
reproduce the pressure conditions experienced by the reinforced base layer when placed
underneath a 4-in. cover of the base material. Ten loading cycles were simulated, each of them
consisting of a 0.10 sec loading period followed by a 0.90 sec rest period, as shown in Figure B-
11. All cells were modeled using material properties of Base 4 material, described in Table B-1.
Figure B-10 Simplified Single Layer FEM with Soil Cells of Base Material Separated with
(a) BLT Shell (SHELL) Element Types, and (b) Thick Shell (TSHELL) Element Types.
178
50
40
Load (psi)
30
20
10
0
0 1 2 3 4 5 6 7 8 9 10
Number of cycles
Figure B-11 Loading Cycles Used for Simplified FE Model.
Two shell element types were evaluated: thin Belytschko-Lin-Tsay Shell (BLT) and thick shell
elements (TSHELL). Both element types were assigned linear elastic material properties
corresponding to Presto GW20V, as shown in Table B-5.
B.6.2 Pavement with Single Cell
This model was developed for evaluating the soil-geocell interface behavior. The 3-D FE model
was developed for simulating a laboratory test consisting of a 3 ft diameter cylindrical tank filled
with a 24-in. thick compacted subgrade, as shown in Figure B-12. A single geocell was placed on
top of the compacted subgrade and filled with infill material. In the laboratory, a 6-in. diameter
and 0.80 in. thick circular steel plate was placed at the center of the geocell pocket and loaded with
an MTS load frame was analogous to the laboratory testing, consisting of Subgrade 1 material and
Base 1 for cell infill, whose properties are described in Table B-3.
Figure B-12 (a) Laboratory Setup, (b) FE Model of Laboratory Setup, and (c) Setup of
Strain Gauges on Geocell for Evaluating Soil-Geocell Interaction.
179
The FE model was developed with these dimensions and loaded with a 6-in. diameter, 80 psi
pressure load applied directly at the center of the cell to simulate these conditions. In addition,
symmetry conditions were applied on boundaries to account for a quarter-model to reduce the
number of elements. Similarly, loading cycles were simulated, each of them consisting of a 0.10
s loading period followed by a 0.90 s rest period, as shown in Figure B-13.
For comparison to unreinforced pavements in the parametric study, a model without geocell was
developed by removing the shell elements and discrete beams from the FE models of geocell-
reinforced structures and then merging nodes that were in contact. Properties for this model are
shown in Table B-4 as well. Table B-6 summarizes the properties of the FE model; Shell elements,
both thin BLT and TSHELL, used for modeling the geocell and geocell-soil interaction was
simulated using discrete beams (springs).
100
80
Load (psi)
60
40
20
0
0 2 4 6 8 10
Number of cycles
Figure B-13 Cyclic Loading with Constant Peak Pressure.
Table B-6 Dimensions of Single Cell FE Model.
Layer Thickness
Geocell Reinforced Base (in.) 4
Subgrade (in.) 24
Number of Solid Elements 2376
Number of Shell Elements (Geocell) 72
Number of Discrete Beams 91
Total Number of Elements 2539
Total Number of Nodes 3011
B.6.3 Geocell-Reinforced Pavement with Geocell Panel Modeled Using Rhomboidal Pattern
A three-layered pavement structure was developed simulating a geocell-reinforced pavement

structure with a geocell expanded panel modeled using a rhomboidal (diamond-shaped) mesh
pattern (Figure B-14). Modeling the geocell using a rhomboidal pattern is less complicated than
the pseudo-sinusoidal honeycomb pattern used in different studies (Yang 2010; Leshchinsky and
Ling 2013). The structure consisted of geocell-reinforced base material on top of the subgrade. In
addition, a base material (cover) was placed on top of the geocell reinforced base for protection of
the geocell. Finally, a quarter model was simulated to reduce the number of elements by taking
advantage of the symmetry. For this model, dimensions and characteristics, such as the number of
nodes and elements, are shown in Table B-7.
180
using Rhomboidal Shape Cells: (a) Top View Highlighting Quarter Model and (b)
Embedding of Geocell Reinforcement in Base Material.
Table B-7 Dimensions and Properties of Geocell-Reinforced Pavement FE Model with
Geocell Panel Simulated Using Rhomboidal Shaped Cells.
Structure
Top Base Layer (in.) 4 4
Geocell Reinforced Base (in.) 4 4
Number of Shell Elements (Geocell) 2400 -
This model was developed as a preliminary model before developing a more complex model that
involved modeling the geocell mat using a pseudo-sinusoidal pattern. The purpose was to verify
whether this model could capture the behavior of the geocell reinforced pavement to traffic loading
without incurring instabilities of the model to avoid using the more complex modeling of the
pseudo-sinusoidal mesh pattern.
The loading consisted of multiple haver-sine cyclic loads of 80 psi (550 kPa), each with a loading
period of 0.1 sec and a rest period of 0.9 sec, as shown in Figure B-13. The load was applied at the
center of the geocell using a 10 in. diameter plate.
181
Other loading scenarios were also considered stability problems aroused from the previous loading
at the element level, mainly when the finite element model used the Mohr-Coulomb material
constitutive model. An approach to overcome this problem was proposed by implementing a
cyclic increasing load that once it reached an 80-psi pressure, the cycles were kept constant. This
method was recommended in Yang's (2010) University of Kansas study to maintain the model's
accuracy. According to the research, the author argued the deformation of the model be more
sensitive to the load step at the beginning of the loading and suggested increasing the load steps
1.2 times the previous load step, starting from 0.072 psi until a magnitude of 80 psi, a typical tire
contact pressure, is reached. This approach is shown in Figure B-15.
Figure B-15 Cyclic Loading Increase from 0 to 80 psi.

Comparisons between both loading cycles, repeated loading vs. increasing loading, were carried
for constitutive models with plastic behavior to study whether the impact of geocell-reinforcement
on the pavement responses is significantly different due to the rate of loading in repeated loads.
A third loading scenario consisted of a monotonically increasing static load, like the repeated
increasing load but without resting periods, shown in Figure B-16. In this situation, the pressure
increases well beyond 80 psi until it reaches 1,200 psi or soil failure, whichever occurs first. This
loading scenario was used to compare a similar approach evaluated in the University of Kansas
study.
Figure B-16 Monotonically Increasing Static Load.
182
B.6.4 Geocell-Reinforced Pavement with Geocell Panel Modeled Using Pseudo-Sinusoidal
Pattern
A three-layered pavement structure was developed simulating a geocell-reinforced pavement

structure used in the parametric study. The structure consisted of geocell-reinforced base material
on top of the subgrade. Base material was placed on top of the geocell reinforced base as
recommended for protection of the geocell. This model was modified to simulate different heights
of the geocell and top cover for the parametric analysis. Geocell infill pockets were modeled using
pseudo-sinusoidal curve shapes of the geocell honeycomb structure, as shown in Figure B-17. To
model the geocell, shell elements were used while solid elements were implemented to model the
soil of unreinforced layers and the infill soil material.
A quarter model was used to reduce the total number of elements, as symmetry conditions exist in
the finite element model, as shown in Figure B-17a. The 8-node constant stress solid elements
were selected for the modeling of the base and subgrade structures. These solid elements use one-
point integration, and hourglass control was included. Considering the lattice pattern of geocell
and its soil embedment in the base layer, proper element arrangement, and nodal connectivity are
required to adjust to the geocell lattice pattern and avoid node penetration between the two
different materials parts. The resulting mesh consisted of quad elements representing the geocell
and a mix of solid hexahedral and prisms elements for the soil. As the reinforced layer was
developed first, while the top unreinforced layer and subgrade were developed later, the element
pattern of the geocell and infill material propagated through the upper and bottom layers, as seen
in Figure B-17b. Dimensions of properties of the FE model are provided in Table B-8.
using Pseudo-Sinusoidal Shaped Cells: (a) Top View Highlighting Quarter Model and (b)
Embedding of Geocell Reinforcement in Base Material.
Both subgrade and base layers were modeled using the FHWA soil model, using the properties
shown in Table 3-1 for the parametric study. In the parametric study, a haversine cyclic load of 80
psi (550 kPa), as shown in Figure B-13, was applied at the center of the geocell using a 10 in.
diameter plate.
183
Table B-8 Dimensions and Properties of Geocell-Reinforced Pavement FE Model with
Geocell Panel Simulated Using Pseudo-Sinusoidal Shaped Cells.
Structure
Top Base Layer (in.) Varying 1, 2, 3 and 4 Varying 1, 2, 3, and 4
Geocell Reinforced Base (in.) Varying 3, 4 and 6 Varying 3, 4 and 6
Number of Thick Shell Elements (Geocell) 1120 -
B.7 SELECTION OF LOADING CONDITIONS: PLATE SIZE AND LOCATION
In all 3-D FE simulations shown in the parametric study are presented in Section 4. The cases used
for developing and improving the model were applied at the center of the geocell, as shown in
Figure B-18a. However, as per the TxDOT request, the load applied at the joint of the geocells, as
shown in Figure 3-18b, has also been taken into consideration, and the analysis of the FE
simulations with this type of loading arrangement will be shared with TxDOT after the analysis of
the results is completed and summarized. In addition, different sizes of the load plates, i.e., 6 in.,
9 in., and 12 in., were evaluated during the development of the model. Nevertheless, the FE models
used for the parametric study presented in the “Parametric Study” later in this chapter for
evaluating the benefit of geocell simulated a 9-in. diameter plate.
Figure B-18 Position of Load in the FE Model at (a) Center of Geocell and (b) Joint of
Geocell.
184
B.8 RESULTS FROM PRELIMINARY INVESTIGATION: EVALUATION OF SOIL
MODELS AND SHELL ELEMENT TYPE FOR MODELING THE GEOCELL
B.8.1 Geocell-Reinforced Pavement with Geocell Panel Modeled Using Pseudo-Sinusoidal

Pattern
B.8.1.1 Linear Elastic Model
The initial investigation consisted of modeling the base and subgrade layers as linear elastic
materials. The geocell-reinforced infill material, consisting of base material, was modeled as linear
elastic with properties shown in Table B-1. The FEM is simulating a geocell-reinforced pavement
with geocell panel modeled using a rhomboidal pattern, as shown in Figure B-14 and with
properties shown in Table B-7, was used for analyzing the behavior of the pavement layers.
Only one loading cycle was attempted in the linear elastic FE analysis, as other cycles would yield
identical results as no plastic deformation occurs. The cycles, as mentioned previously, consisted
of a loading period of 0.1 sec and a rest period of 0.9 sec, reaching a peak pressure of 80 psi, as
shown in Figure B-13. Pavement responses were compared to those of unreinforced pavements
with similar loading conditions.
The initial attempt using linear elastic material models failed to exhibit any significant difference
in the responses between the geocell-reinforced pavements and their respective unreinforced cases.
For instance, the plot is shown in Figure B-19 for a 5 ksi base layer and 5 ksi subgrade, and 6-in.
diameter load plate applying a pressure of 80 ksi at the center of the cell no significant difference
regarding deflection with respect to depth. Similar responses were obtained for all other loading
scenarios and pavement distresses.
Figure B-19 Vertical Deflection along the Depth of the Pavement Structure.
In addition to finding no significant effect of the geocell reinforcement, this model could not
capture plastic deformation. This behavior was consistent with the findings previously carried in
other numerical simulation studies using linear elastic materials where minimal or no significant
185
changes in distresses are observed when geocell reinforcement is included. Consequently, other
material models were selected to accommodate plastic responses.
B.8.1.2 Mohr-Coulomb Material Model
As previously described, the Mohr-Coulomb is the best-known soil material model that can
address the soil’s plastic behavior. Unlike the linear elastic model, the use of this model incites the
inclusion of multiple loading cycles into the analysis to allow permanent deformation responses to
occur in the soil. The properties used for evaluating this constitutive model are shown in Table B-
2. The FE model of a geocell-reinforced pavement with a geocell panel modeled using a
rhomboidal pattern was used as the linear elastic model. The Mohr-Coulomb material model
provided different soil responses when geocell reinforcement was included, as shown in Figure B-
20. This figure shows the vertical stress of both pavements, geocell-reinforced and unreinforced,
loaded with 12 in. diameter plates and a pavement structure comprised of a 5-ksi base and 5-ksi
subgrade layers, both with the cohesion of 0.72 psi and friction angle of 40°. Fringe contours are
set equal in both figures for better comparison purposes, and stress is expressed in Pascals (Pa).
Significant higher stresses develop in the reinforced base that covers a wider area than in the
unreinforced case. Lower stresses occur at the top of the subgrade in the reinforced pavement when
compared to the unreinforced one.
Figure B-20 Vertical Stress (z-direction) in (a) Geocell-Reinforced (b) Unreinforced

Sections.
However, the use of this model became unstable after a few loading cycles, resulting in an abrupt
termination of the analysis due to the excessive deformation. The number of cycles before the error
depended heavily on the properties used in the analysis, with less stiff and more cohesive materials
failing earlier. No material could resist failure beyond ten loading cycles. Figure B-20 shows
indications of the elements prone to become unstable, mainly seen in the regions in darker blue
along the surface of the soil, which by the way happen to be unloaded. In addition to this problem,
large deformations occurred that seemed excessive when compared to other studies.
The problem encountered with this material model in LS-DYNA made its implementation
unsuccessful. An attempt was made with ABAQUS to replicate the model. Despite apparent
186
success in providing reasonable deformations, the analysis proved time prohibitive. The program
lacks solvers with various iterative methods to solve the linear system of equations formed during
the implicit analysis. Moreover, the program was not available in the HPC system. Another soil
model was selected within LS-DYNA to overcome this problem, which was developed based on
the Mohr-Coulomb approach. The selected model was the FHWA Soil model, whose description
was provided in a previous section.
B.8.1.3 FHWA Soil Model
Like the Mohr-Coulomb model, the FHWA soil model can accommodate permanent deformation.
Four different base materials were evaluated with two distinct subgrades, described in Table B-3.
The pavements were subjected to multiple cycles with a constant peak pressure of 80 psi, as shown
in Figure B-13.
Figure B-21 shows the deflection with respect to depth under the center of load occurring at the
peak load of the first cycle for both reinforced and unreinforced base layers. The plots shown in
Figure B-21 correspond to the modeling of pavements consisting of Subgrade-1 (1.74 ksi) and
Bases 1-3, ranging from 5.8 to 11.6 ksi (refer to Table B-3). In these figures, geocell reinforcement
reduces the surface deformation when compared to unreinforced sections. Geocell reinforcement's
impact on surface deformation decreases as the base becomes more rigid.
Figure B-21 Vertical Deflection with Respect to Depth at Peak Load of First Cycle for a
Two-Layer Pavement System Consisting of Subgrade-1 (E=1.74 ksi, c=1.45 psi, φ=15°) and
Different Base Properties.
187
Figure B-22 shows the decrease in surface deformation in terms of percent reduction between the
unreinforced and reinforced sections with respect to the base modulus. As the modulus of the base
layer improves, the effect of geocell reinforcement diminishes. The benefit of using a geocell-
reinforced base layer was not evident when a good base quality material was used.
Figure B-22 Percentage Reduction in Surface Deformation with respect to Base Modulus
for a Pavement with Subgrade-1 (E=1.74 ksi, c=1.45 psi, φ =15°) Material.
The implementation of the FHWA soil model proved to be successful in terms of reasonable
deformations compared to the Mohr-Coulomb responses and considerably more stable too. The
model is also prone to encounter issues with respect to element stability after several loading
cycles. However, in some instances, the model behaved reasonably well up to 200 cycles, where
just one element located right under the center of the load suffered excessive deformation that
caused an error in the execution. Despite this issue, the model provided enough cyclic responses
to visualize the permanent deformation behavior of the pavement. To overcome the problem
arising from the instability of the elements, a ramp-like increasing repeated loading until a pressure
of 80 psi is reached, as shown in Figure B-15, was adopted and compared to the repeated loading
with no change in the loading pattern and magnitude.
To see the effect of the loading condition on the permanent response behavior of the pavement
structure due to these two different loading cycles, a pavement with weak mechanical properties
like those used in the University of Kansas, as shown in Table B-4, was modeled. The permanent
deformation due to the first loading scenario (Figure B-13), which consisted of the repeated 80 psi
loads, is shown in Figure B-23. In contrast, the permanent deformation of the second loading
scenario (Figure B-15) is shown in Figure B-24 for both reinforced and unreinforced sections.
Number of Cycles
0 1 2 3 4 5 6 7 8 9 10
0.00
Deformation (in.)
Permanent
-0.10 Geocell Reinforced

Unreinforced
-0.20
-0.30
-0.40

Sections using FHWA Soil Model Properties: Kansas River Sand base (E=0.48 ksi, c=0,
φ=41°) and Clay Subgrade (E=1.5 ksi, c=15.2 psi, φ =0°) and Repeated 80 psi Loading.
188
Number of Cycles
0 5 10 15 20 25 30 35 40
Permanent Deformation 0.00
Geocell Reinforced
-0.10
Unreinforced
(in.)
-0.20
-0.30
-0.40
Sections using FHWA Soil Model Properties: Kansas River Sand base (E=0.48 ksi, c=0,
φ=41°) and Clay Subgrade (E=1.5 ksi, c=15.2 psi, φ =0°) and Increasing Repeated Loading
Reaching 80 psi Followed by 80 psi Repeated Loads.
The results summarized in Figure B-23 indicate that higher permanent deformation is observed
within the initial cycles (first two cycles) when a constant repeated load of 80 psi (Figure B-13) is
applied to the pavement. However, the benefit of geocell reinforcement is observed after the
application of three or more cyclic loads. On the other hand, the increasing repeated load (Figure
B-15) indicates that the permanent deformation is independent of reinforcement (Figure B-24).
For further comparison, the permanent deformation data for both loading scenarios are
summarized in Figure 3-25. In this figure, the surface displacements during the earlier cycles are
not comparable, as the increasing cyclic loading already amounted to given permanent deformation
of about 0.30 in. However, after a few 80 psi cycles, the responses of both loading scenarios
approach each other, indicating that there is no need to initiate the loading cycles from low
magnitudes but instead start with an 80-psi magnitude. Furthermore, no evidence was found that
ramping the repeated loads made the model more stable by extending the analysis in cycles before
reaching an error termination due to excessive element deformation. Moreover, and most
importantly, time is saved as the ramping to reach a pressure of 80 psi (about 30 cycles) requires
analyses times of about 30 hours using the HPC system. Additional discussion about execution
times and problems encountered during the modeling are provided in a later section.
Number of Cycles
0 1 2 3 4 5 6 7 8 9 10
0.00
Geocell reinforced_increasing load
Deformation (in.)
-0.10 Unreinforced_increasing load

Permanent
Geocell reinforced_constant load

-0.20 Unreinforced_contant load
-0.30
-0.40
Figure B-25 Impact of Ramp Loading on Permanent Deformation for Geocell-Reinforced

and Unreinforced Pavement Sections using Kansas River Sand Base and Clay Subgrade.
189
B.8.2 Comparison to University of Kansas Model
A complete comparison of the responses using the FHWA soil model with those of the numerical
modeling developed at the University of Kansas was carried to check the behavior of our model
with respect to the Duncan-Chang model used in their study. The properties used in their model,
shown in Table B-4, were adapted to the FHWA constitutive soil model. Following similar loading
conditions, a monotonically increasing load was applied to our model.
For comparison, the pressure to surface displacement curves for both unreinforced and geocell-
reinforced sections was developed and are shown in Figure B-26. From Figure B-26a, showing
the surface displacements for the FHWA soil model, a significant difference in surface
displacements between the reinforced and unreinforced section can be seen after the pressure of
500 kPa (72 psi) is reached. This behavior occurs earlier, at about 200 kPa (29 psi), when the
Duncan-Chang model is used (Figure B-26b). However, both responses agree on the benefit of
using geocell reinforcement for very weak unbound aggregate materials. Both responses are not
similar, and the reason behind this is due to the different soil models used and possibly to distinct
rates of loading. SI units are shown in Figure B-26, as the results provided in Yang (2010) used
this unit system.
Figure B-26 Pressure to Surface Displacement Curves for Unreinforced and Geocell-
Reinforced Sections using (a) FHWA Constitutive Soil Model and (b) Duncan Chang
Constitutive Model used in the University of Kansas Study (after Yang, 2010).
B.8.3 Preliminary Evaluation of Contact
The development of the 3-D FE model of the geocell-reinforced pavement used to evaluate the
most appropriate constitutive model for simulating the permanent deformation of geomaterials
described in the previous section brought forth different problems during the simulation. Some of
the issues were related to the contact between the geocell simulated with shell elements and the
infill material simulated with solid elements.
One of the conditions required by LS-DYNA for modeling the contact between the shell and solid
elements is the need to have an offset between the shell and the solid elements to avoid overlapping
nodes between the two types of elements. This implies that two nodes cannot share the same
coordinate and that a gap must be considered to account for the shell thickness provided by the
user. This helps the contact model's search algorithm find nodes' penetrations and establish contact
between nodes. If the user does not correct this, LS-DYNA will automatically adjust the nodes by
190
changing the coordinates before the analysis, leading to overlapping the elements at the contact. This
issue is seen markedly at the joints in our preliminary finite element models, as shown in Figure B-27.
Figure B-27 Node Penetration and Overlapping of Element at the Joints.

To correct this problem, a gap was included between the soil infill and the geocell on both sides
of the geocell, as shown in Figure B-28a. In addition, a separation between the cell infill material
was modeled to accommodate a 1 mm thick geocell modeled using shell elements, as shown in
Figure B-28b.
Figure B-28 (a) Top View of FE Model Showing Nodal Unbonding to Accommodate
Geocell and (b) Zoomed in View at Joint Showing Infill Material and Geocell Modeled
using Shell Elements.
B.8.4 Evaluation of Geocell Shape: Geocell Panel Simulated with Rhomboidal Pattern
During the evaluation of the most appropriate constitutive model for modeling the base and
subgrade materials, the geocell panel was modeled using a rhomboidal pattern with an automatic
surface to surface contact model between the infill soil and the geocell. In addition, a tied contact
interface was provided between the geocell and the interfaced above and below the geocell
reinforced base, i.e., the top cover and the subgrade, respectively. The rationale to use a tied contact
interface between layers was to allow full transfer of stress from the upper to the lower soil layer.
After some load cycles, the geocell tarted to experience an unstable behavior, which varied
191
depending on the properties of the base material, but consistently occurred at the early stages of
the loading series, sometimes happening before the tenth loading cycle. As the soil accumulated
permanent deformation, large stresses developed and increased in the geocell at the model's edges,
as shown in Figure B-29. Moreover, nodal penetrations occurred at the joints.
Figure B-29 Stresses Developed at the Edges of the Model.

To further investigate and find a solution to this problem that originated from the contact model, a
simplified model was developed to study the behavior of the different contact types available in
the program. Furthermore, to address the issue of the high stresses occurring at the joints, another
pattern was considered for modeling the geocell, which consisted of moving from the rhomboidal
shape towards a pseudo-sinusoidal shaped pattern.
B.8.5 Evaluation of Element Type for Modeling the Geocell
The simplified single-layer FE model described in B.6.7 was used to evaluate the shell elements'
efficiency in transferring the stresses from one cell to the adjacent one (see Figure B-10). Different
scenarios were considered to evaluate both the contact types and shell element types. In addition,
longitudinal stresses were collected at mid-depth of the model to identify the effect of the geocell
on the infill material of the adjacent cell. The results of these scenarios are described in the
following sections.
B.8.6 Use of Shell Elements (SHELL) for Modeling Geocell and Automatic Surface Contact
Type
The first approach to model geocell consisted of using shell elements and the Automatic Single
Surface contact type. The results obtained for this model shown indicated that the geocell applied
no lateral displacement to the infill material inside Cell #2 despite the geocell being pushed by the
infill material in Cell #1, as shown in Figure B-30. This indicated that the automatic single surface
contact model could not detect penetration of the nodes belonging to the geocell into the infill
material. When the automatic single surface contact type is used, the master surface is projected
192
based on normal nodal vectors; thus, the algorithm of the contact type was unable to detect node
penetration in the direction opposite to the normal vector of the shell elements. Therefore, no
longitudinal stresses developed on Cell #2 or Cell #3, as shown in Figure B-31. The Automatic
Surface-to-Surface contact type has similar limitations and runs into the same problem. The use of
shell elements and automatic surface contact types proved inadequate for modeling geocell
reinforcement embedded in soil material when solid elements constrain the two sides of the shell
elements.
Figure B-30 Longitudinal Displacement (dx) at 10th Loading Cycle after FE model with
Geocell using Shell Elements and Automatic Single Surface Contact Element.
Figure B-31 Vertical Stress (Δx) at 10th Loading Cycle after FE Model with Geocell using
Shell Elements and Automatic Single Surface Contact Element.
B.8.7 Use of Thick Shell Elements for Modeling Geocell and Automatic Surface Contact
Type
The second approach to model geocell reinforcement consisted of using thick shell elements,
identified by LS-DYNA as TSHELL, and the Automatic Surface to Surface contact type. In this
case, lateral displacement transferred from Cell #1 to Cell #2 through the geocell and Cell #2 to Cell
#
3 to a lesser extent, as shown in Figure B-32. Likewise, stresses transferred through the geocell
from Cell #1 to Cell #2, as shown in Figure B-33. For this purpose, TSHELL proved to provide
meaningful results in conjunction with the Automatic Surface to the Surface contact type.
193
Figure B-32 Longitudinal Displacement (dx) at 1st Loading Cycle after FE Model with
Geocell using Thick Shell Elements and Automatic Single Surface Contact Element.
Figure B-33 Vertical Stress (σx) at 1st Loading Cycle after FE Model with Geocell using
Thick Shell Elements and Automatic Single Surface Contact Element.
With similar results, thick shell elements could be used with both automatic surface to surface or
automatic single surface contact types. However, a disadvantage in using thick shell elements is
that the analysis time increases considerably as thick shells are more computationally expensive
than traditional ones. The critical time steps Δte,crit for shells and thick shells are provided in
Equations 3.8 and 3.9, respectively as:
For shell elements
A
∆te,crit =
cs lmax
B-8
For thick shell elements
V
∆te,crit =ele
cb Aele,max
B-9
194
where cs is the speed of sound, cb is the P-wave velocity, A is the area of the smallest shell element
in the model, lmax is either the maximum side or diagonal length of the smallest shell element in
the model, and Vele and Aele,max are the volume and area of the smallest thick shell element used in
the model, respectively (Haufe et al. 2013). For this reason, the critical time steps in the FE
analysis are much smaller for thick shell elements than for conventional thin shell elements, as
shown in Figure B-34, leading to a larger number of time steps.
Figure B-34 Time Step Size as Reported by the d3hsp Output File Required for the
Analysis of Simplified FE Model using (a) Shell Elements and (b) TSHELL Elements.
Figure B-35 shows the time required for 1 second of simulated time when a High-Performance
Computing (HPC) cluster with IBM POWER5 1.65 GHz processors for the 480-element small FE
model when thick shell elements are used at a different number of processors.
0:15
simulated time (Hr:Min)
Time per 1 sec. of
0:10
0:05
0:00
0 5 10 15 20 25 30
No. of Processors
Figure B-35 Time Required for 1 Second of Analysis for 480 Element FE model with
Different Number of 1.65 GHz Processors.
Thick shell elements addressed the contact better when placed between solid elements than typical
thin shell elements. For this reason, thick shell elements were preferred for simulating the geocell.
To better understand the contact type and its effect on the transfer of stress from one cell to another,
different types of contact were evaluated while keeping the thick shell element type for the
modeling of geocell.
195
B.9 EVALUATION OF CONTACT TYPE FOR SOIL-GEOCELL INTERFACE
B.9.1 Assessment of Contact Type Using Simplified Single-Layer Model
To better address the behavior and interaction of the geocell with soil, several contact types were
evaluated using the same small FE model described in the previous section. This model was
developed to check the efficiency of the contact type on transferring the stresses from one cell to
the adjacent one through the geocell. Likewise, as shown in Figure B-11, ten loading cycles were
simulated with a peak pressure load of 40 psi on a 6-in. diameter load, and the soil had properties
of sand as described in Table B-1. To evaluate the contact type and geocell on the infill material
of the adjacent cell, longitudinal and vertical stresses were collected at mid-depth of the model, as
shown in Figure B-10.
Three contact types were evaluated:
a) Automatic Surface-to-Surface Contact,
b) Automatic Single Surface Contact, and
c) Springs (discrete beam elements).
The responses obtained at the specified locations for the simplified models with these types of
contacts were compared to those obtained by other simplified models with no geocell
reinforcement (no implementation of shell elements) and different contact scenarios. In addition,
the responses using the BISAR computer program based on layer elastic theory are provided for
comparison purposes. The following FE models were used for comparing the responses of the
geocell reinforced material.
A linear elastic model based on the relationships developed by Boussinesq and implemented in
BISAR. All FE models are based on three individual infill cells separated by geocell, as shown in
Figure B-36. Contact type and element used for modeling geocell vary depending on the following
conditions:
a) FHWA Soil: FEM simulates only soil material with no geocell separating the infill cells. No
interaction of geocell to soil is required. FHWA soil model is used.
b) Automatic Surface Contact Soil to Soil: an FE unreinforced model using FHWA soil model for
the base material, simulating three individual infill cells (no geocell) connected only with an
automatic surface-to-surface contact type (automatic single surface offers identical results).
c) SOLID Geocell + Automatic Single Surface: FEM simulates three individual infill cells
separated by geocell simulated with brick elements (instead of shell elements) and connected
with an automatic single surface contact type (automatic surface-to-surface offers identical
results).
d) TSHELL Geocell + Fully Bonded: a geocell-reinforced FE model with fully bonded conditions
between geocell and soil. Geocell simulated using thick shell (TSHELL) elements.
e) TSHELL Geocell + Automatic Single Surface: FEM simulates three individual infill cells
separated by geocell simulated with thick shell elements (TSHELL) and connected with an
automatic single surface contact type (automatic surface-to-surface offers identical results).
f) Spring: kn = 57 lb/in, kt = 5.7 lb/in FE model simulating three individual infill cells separated
by geocell simulated with the thick shell (TSHELL) elements and connected with springs with
normal stiffness kn = 57 lb/in and tangential stiffness kt = 5.7 lb/in.
196
g) Spring: kn = 570 lb/in, kt = 5.7 lb/in FE model simulating three individual infill cells separated
by geocell simulated with thick shell (TSHELL) elements and connected with springs with
normal stiffness kn = 570 lb/in and tangential stiffness kt = 5.7 lb/in.
These models are summarized along with their characteristics in Table B-9.
Figure B-36 Simplified FE Model with Highlighted Elements used for Evaluating the Stress
Transfer Through Contact and Geocell.
Table B-9 Summary of Numerical Models Used for Evaluating Contact Types.
Analysis Type Geocell Geocell-Soil Interaction
Type of Model Layer Elastic Finite
Element Type Contact Type
Theory Element
1. Linear Elastic (BISAR)  - -
2. FHWA Soil  No No
Automatic Surface to
3. Aut. Surf. Contact Soil to Soil*  No
Surface
Automatic Surface to
4. SOLID Geocell + Automatic Single Surf. *  Brick (Solid)
Surface
5. TSHELL Geocell + Fully Bonded  Thick Shell No
6. TSHELL Geocell + Automatic Single Automatic Surface to
 Thick Shell
Surf. * Surface
7. Spring: kn = 57 lb/in, kt = 5.7 lb/in  Thick Shell Spring
8. Spring: kn = 570 lb/in, kt = 5.7 lb/in  Thick Shell Spring
* Automatic Surface to Surface contact type yielded similar results to Automatic Single Surface.
Table B-10 summarizes the soil responses regarding the longitudinal stress, σx, and vertical stress,
σy, at mid-height of the base layer as reported by the analysis of the simplified FE models and
linear elastic theory at the locations:
• Edge of Cell#1 in contact with geocell
• Edge of Cell#2 in contact with geocell (see Figure B-36).
• This table shows that the inclusion of geocell caused a decrease in stress at the elements
adjacent to the cells compared to the values obtained from the FE model with no geocell.
Moreover, the magnitude of the stress of the elements adjacent to the geocell when the
geocell is modeled as a thick shell is somewhat like those obtained after the geocell is
modeled as a solid element. The use of geocell also caused a decrease in magnitude in
stress at the element belonging to the second cell, and vertical stress decreased in the
element belonging to Cell #1, next to the geocell.
197
Table B-10 Summary of Longitudinal and Vertical Stress at the Mid-Height of Model
Observed at the Edge of Cell#1, at the Geocell, and at the Edge of Cell#2 in contact with the
Geocell.
Longitudinal Stress (σx), psi Vertical Stress (σz), psi
Type of Model
Soil Cell 1 Soil Cell 2 Soil Cell 1 Soil Cell 2
1. Linear Elastic (BISAR) 5.60 4.93 4.34 2.75
2. FHWA Soil 4.39 4.44 8.01 3.99
3. Aut. Surf. Contact Soil to Soil 5.05 4.84 5.79 2.62
4. SOLID Geocell + Automatic Single Surface 4.71 4.10 5.74 1.40
5. TSHELL Geocell + Fully Bonded 6.06 5.95 7.84 3.75
6. TSHELL Geocell + Automatic Single Surface 4.45 4.28 5.94 1.39
7. Spring: kn = 57 lb/in, kt = 5.7 lb/in 0.90 0.84 5.09 0.39
8. Spring: kn = 570 lb/in, kt = 5.7 lb/in 3.50 3.55 6.98 1.72
The use of discrete elements (springs) provided a more stable behavior during simulation when
cyclic loading is applied and provided reliable results when used with thick shell elements.
However, little information and no standard laboratory method exists to determine the values that
should be used for the normal and tangential stiffness. Table B-10 suggests that the discrete
elements and the automatics single surface contacts seem to work well when used with thick shell
elements, i.e., stress is transferred from one cell to another through the geocell. A single cell FEM
was developed to compare which contact type provided more similar results to those obtained from
the laboratory to evaluate the contact types further.
B.9.2 Evaluation of Contact Type Using Single-Cell 3-D FE Model
A test was conducted in a 3-ft. diameter cylindrical tank with a single geocell was placed on top
of a 24-in. thick compacted subgrade and infill materials in the geocell pocket were compacted in
three layers to study the behavior of the geocell and the soil interaction. The vertical load was
applied through an MTS load frame, and the plate settlement was recorded. Strain gauges were
glued at the mid-height of the geocell, around the cell circumference, as shown in Figure B-12. A
quarter bridge circuit arrangement was used to connect the strain gauges. The strain was recorded
using the LMS Scadas Mobile data acquisition system. A repeated load was applied in the middle
of the cell, as shown in Figure B-13. The finite element model described in B.3.2 was used for
comparison purposes and was developed to simulate the laboratory conditions. Likewise, the
strains were recorded at the middle of the geocell, selecting elements located close to the strain
gauges’ positions.
As per the laboratory test results, the hoop strains (circumferential strains) were maximum at the
center and decreased towards the edge. Though the FEM results for both contact models (automatic
single surface and discrete elements) provided similar results, the hoop strains at the edge of the
geocell differed from the laboratory results when an automatic single surface contact model was
used. Hoop strains determined from the numerical model with a soil-geocell interface using
discrete beam elements, i.e., 6-DOF spring elements, were closer to laboratory strain gauge
measurements, at the center and edge of the geocell, and along the perimeter of the geocell, as
well. Figure B-37, Figure B-38, and Figure B-39 show the hoop strains observed on edge, mid-
geocell, and the center of geocell, respectively (see Figure B-12c), for 100 cycles, as obtained from
laboratory test and the FE model. It must be noted that the data plotted from the laboratory shows
198
only the responses at peak loading, while the FE model data shows the responses at the peak and
rest period as well.
The measured strains at the edge of geocell observed were in the range of 240 μɛ. The FEM model
generated strains in the range of 290 με during loading and 180 με during the rest period. The key
difference in the results was the elastic strain in the FEM model, which was higher than the
laboratory results. The difference in responses could be attributed to the impossibility of achieving
zero loads during unloading in the lab testing for each cycle and geocell characterized as a perfectly
elastic material in the FE model. Similar behavior can be observed in the responses at mid-geocell
and the center of geocell.
Based on the similarity of the results provided by the geocell in the single-cell FEM with the
laboratory results, it was decided to use to expand this model to simulate the honeycomb pattern
of the geocell. This suggests that the discrete elements were used as a contact type, thick shell
elements were used to simulate the geocell, and a pseudo-sinusoidal pattern was used for modeling
the geocell panel, as shown in Figure B-17. The resulting model is described in Table B-8 and was
used for carrying a parametric study for evaluating the effect of geocell on the pavement responses.
600 600
a) b)
400 400
Strain (µε)
Strain (µε)
200 200
0 0
0 20 40 60 80 0 20 40 60 80
Cycles Cycles
Figure B-37 Hoop Strains at Edge of Geocell for (a) Laboratory and (b) FEM.
800 800
a) b)
600 600
Strain (µε)
Strain (µε)
400 400
200 200
0 0
0 20 40 60 80 0 20 40 60 80
Cycles Cycles
Figure B-38 Hoop Strains at Mid-Geocell for (a) Laboratory and (b) FEM.
199
600 600
a) b)
400 400
Strain (µε)
Strain (µε)
200 200
0 0
0 20 40 60 80 0 20 40 60 80
Cycles Cycles
Figure B-39 Strains at Center of Geocell for (a) Laboratory and (b) FEM.
200
APPENDIX C: GEOCELL INFORMATION
Figure C-1 Tenax Geocell Properties.
201
Figure C-2 Presto Geocell Properties.
202
Figure C-2 continued
203
Figure C-3 Strain Gauge Specifications.
204
Figure C-3 continued.
205
Figure C-4 Pressure Cell Specifications.
206
Figure C-4 continued.
207
Figure C-5 Data Smoothing: Kernel Regression vs. Moving Average.
208
APPENDIX D
Table D-1 Raw Data

Geocell Infill Cover Subgrade Modulus Percentage Reduction of Vertical
Height (in.) Modulus Thickness (in.) (ksi) Stress on Subgrade Top Below the
(ksi) Center of Loading Plate
4 3 6 4.5 16
4 5 6 4.5 19
4 7 6 4.5 11
4 9 6 4.5 20
4 12 6 4.5 24
4 15 6 4.5 11
4 30 6 4.5 4
6 3 4 4.5 39
6 5 4 4.5 30
6 7 4 4.5 21
6 9 4 4.5 20
6 12 4 4.5 21
6 15 4 4.5 20
4 7 6 6 15
4 12 6 6 17
4 15 6 6 17
4 20 6 6 9
4 30 6 6 27
4 12 6 9 20
4 15 6 9 17
4 20 6 9 15
4 30 6 9 4
4 15 6 15 20
4 20 6 15 14
4 30 6 15 22
3 3 4 4.5 2
3 9 4 4.5 5
3 15 4 4.5 4
3 30 4 4.5 2
4 12 6 8 20
4 12 6 3 17
4 12 2 4.5 21
6 7 4 6 22
209
6 12 4 6 18
6 15 4 6 19
6 20 4 6 9
6 30 4 6 12
6 12 4 9 21
6 15 4 9 15
6 30 4 9 5
6 15 4 12 11
6 20 4 12 11
6 30 4 12 14
3 5 4 2.5 1
3 7 4 2.5 5
3 12 4 2.5 13
3 15 4 2.5 7
3 20 4 2.5 2
3 7 4 6 5
3 12 4 6 14
3 15 4 6 8
3 12 4 9 1
3 15 4 9 3
3 20 4 9 5
3 15 4 15 3
3 20 4 15 3
6 30 4 4.5 9
6 20 4 9 17
210
Table D-2 Training Data Set LOOV-CV
Sample no Geocell Infill Modulus Cover Thickness % Reduction in
Height (in.) (ksi) (in.) Stress
1 4 3 6 16
3 4 7 6 11
4 4 9 6 20
5 4 12 6 24
6 4 15 6 11
7 4 30 6 4
8 6 3 4 39
9 6 5 4 30
11 6 9 4 20
12 6 12 4 21
14 4 7 6 15
15 4 12 6 17
17 4 20 6 9
18 4 12 6 20
20 4 20 6 15
21 4 30 6 4
22 4 15 6 20
23 4 20 6 14
25 3 9 4 5.3
26 3 15 4 3.9
27 3 30 4 1.8
28 4 12 6 20
29 4 12 6 17
30 4 12 2 21
31 6 7 4 22
32 6 12 4 18
33 6 15 4 19
35 6 30 4 12
37 6 15 4 15
39 6 15 4 11
40 6 20 4 11
41 6 30 4 14
42 3 5 4 1
43 3 7 4 5
44 3 12 4 13
47 3 7 4 5
48 3 12 4 14
49 3 15 4 8
50 3 12 4 1
51 3 15 4 3
52 3 20 4 5
53 3 15 4 3
54 3 20 4 3
56 6 20 4 17
211
Table D-3 Testing Data Set LOOV-CV
Sample no Geocell Height Infill Modulus Cover Thickness % Reduction in Stress
(in.) (ksi) (in.)
2 4 5 6 19
10 6 7 4 21
13 6 15 4 20
16 4 15 6 17
19 4 15 6 17
24 3 3 4 2
34 6 20 4 9
36 6 12 4 21
38 6 30 4 5
45 3 15 4 7
46 3 20 4 2
55 6 30 4 9
2 4 5 6 19
10 6 7 4 21
13 6 15 4 20
212
Table D-4 Training Data Set KK FOLD CV
1 4 3 6 16
2 4 5 6 19
5 4 12 6 24
6 4 15 6 11
7 4 30 6 4
8 6 3 4 39
9 6 5 4 30
11 6 9 4 20
12 6 12 4 21
13 6 15 4 20
14 4 7 6 15
16 4 15 6 17
19 4 15 6 17
20 4 20 6 15
21 4 30 6 4
22 4 15 6 20
23 4 20 6 14
24 3 3 4 2
25 3 9 4 5.3
26 3 15 4 3.9
27 3 30 4 1.8
28 4 12 6 20
29 4 12 6 17
30 4 12 2 21
31 6 7 4 22
33 6 15 4 19
34 6 20 4 9
35 6 30 4 12
36 6 12 4 21
37 6 15 4 15
39 6 15 4 11
40 6 20 4 11
41 6 30 4 14
42 3 5 4 1
43 3 7 4 5
44 3 12 4 13
46 3 20 4 2
47 3 7 4 5
48 3 12 4 14
49 3 15 4 8
51 3 15 4 3
52 3 20 4 5
53 3 15 4 3
55 6 30 4 9
213
Table D-5 Testing Data Set KK FOLD CV
(in.) (ksi) (in.)
3 4 7 6 11
4 4 9 6 20
10 6 7 4 21
15 4 12 6 17
17 4 20 6 9
18 4 12 6 20
32 6 12 4 18
38 6 30 4 5
45 3 15 4 7
50 3 12 4 1
54 3 20 4 3
56 6 20 4 17
3 4 7 6 11
4 4 9 6 20
10 6 7 4 21
214
Table D-6 Training Data Set Bootstrapping CV
1 4 3 6 16
2 4 5 6 19
3 4 7 6 11
4 4 9 6 20
6 4 15 6 11
8 6 3 4 39
9 6 5 4 30
10 6 7 4 21
11 6 9 4 20
13 6 15 4 20
14 4 7 6 15
16 4 15 6 17
17 4 20 6 9
19 4 15 6 17
20 4 20 6 15
21 4 30 6 4
22 4 15 6 20
23 4 20 6 14
24 3 3 4 2
28 4 12 6 20
29 4 12 6 17
30 4 12 2 21
31 6 7 4 22
32 6 12 4 18
33 6 15 4 19
34 6 20 4 9
35 6 30 4 12
36 6 12 4 21
38 6 30 4 5
39 6 15 4 11
40 6 20 4 11
41 6 30 4 14
42 3 5 4 1
43 3 7 4 5
44 3 12 4 13
46 3 20 4 2
47 3 7 4 5
48 3 12 4 14
50 3 12 4 1
51 3 15 4 3
52 3 20 4 5
53 3 15 4 3
54 3 20 4 3
55 6 30 4 9
215
Table D-7 Testing Data Set Bootstrapping CV
(in.) (ksi) (in.)
5 4 12 6 24
7 4 30 6 4
12 6 12 4 21
15 4 12 6 17
18 4 12 6 20
25 3 9 4 5.3
26 3 15 4 3.9
27 3 30 4 1.8
37 6 15 4 15
45 3 15 4 7
49 3 15 4 8
56 6 20 4 17
216
D.1 PAIRED T-TEST
Use the paired t-test to determine whether the difference between sample means for paired data is
significantly different from the hypothesized difference between population means.
Null Hypothesis: 𝜇𝜇𝑟𝑟 = 𝐷𝐷
Alternative Hypothesis: 𝜇𝜇𝑟𝑟 ≠ 𝐷𝐷
Where 𝜇𝜇𝑟𝑟 𝑖𝑖𝑠𝑠 the true difference in population values, and D is hypothesized value.
D.1.1 Procedure
Using sample data, find the standard deviation, standard error, degrees of freedom, test statistic,
and the P-value associated with the test statistic.
Step 1: Compute the standard deviation (sd) of the differences computed from n matched pairs.
sd = sqrt [ (Σ(di - d)2 / (n - 1) ]
D-1
where di is the difference for pair i, d is the sample mean of the differences, and n is the
number of paired values.
Step 2: Standard error. Compute the standard error (SE) of the sampling distribution of d.
SE = sd * sqrt{ ( 1/n ) * [ (N - n) / ( N - 1 ) ] }
D-2
where sd is the standard deviation of the sample difference, N is the number of matched
pairs in the population, and n is the number of matched pairs. When the population size is
much larger (at least 20 times larger) than the sample size, the standard error can be
approximated by: SE = sd/sqrt ( n )
Degrees of freedom. The degree of freedom (DF) is DF = n - 1.
Step 3: Test statistics. The test statistic is a t-statistic (t) defined by the following equation.
t = [ (x1 - x2) - D ] / SE = (d - D) / SE
D-3
where x1 is the mean of sample 1, x2 is the mean of sample 2, d is the mean difference
between paired values in the sample, D is the hypothesized difference between population
means, and SE is the standard error.
Step 4: The P-value is the probability of observing a sample statistic as extreme as the test statistic.
Since the test statistic is a t-statistic, use the t Distribution Calculator to assess the probability
associated with the t statistic, having the degrees of freedom computed above. Interpret Results
217
Step 5: If the sample findings are unlikely, given the null hypothesis, the researcher rejects the null
hypothesis. Typically, this involves comparing the P-value to the significance level and rejecting
the null hypothesis when the P-value is less than the significance level.
D.2 PEARSON CORRELATION

Pearson’s correlation coefficient measures the strength of the linear relationship between two sets
of variables. Correlation coefficient:
∑(𝑒𝑒 − 𝑒𝑒̅ )(𝑦𝑦 − 𝑦𝑦�)
𝑟𝑟 =
(𝑡𝑡 − 1)𝑠𝑠𝑥𝑥 𝑠𝑠𝑦𝑦
D-4
Characteristics:
1. Shows the direction and strength of the linear relationship between two variables
2. Can assume any value from -1.00 to +1.00
3. When 𝑟𝑟 = 0, this indicates there is no linear relationship between the variables.
4. When 𝑟𝑟 ≈ 0, this indicates a weak linear relationship between the variables.
5. When 𝑟𝑟 ≈ 1, this indicates a direct or positive linear relationship between the variables.
6. When 𝑟𝑟 ≈ −1, this indicates an inverse or negative linear relationship between the
variables.
218

Aashto Geocell Önemli

Uploaded by

Copyright:

Available Formats

Aashto Geocell Önemli

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aashto Geocell Önemli

Uploaded by

Copyright:

Available Formats

Use of Geocell in Pavement Design: Final Report

Technical Report 0-6833-1

CENTER FOR TRANSPORTATION INFRASTRUCTURE SYSTEMS

in cooperation with the

7. AUTHOR(S) 8. PERFORMING ORGANIZATION REPORT No.

500 W. University Ave. 0-6833

17. KEYWORDS 18. DISTRIBUTION STATEMENT

Research Report 0-6833-1

Department of Civil Engineering

NOT INTENDED FOR CONSTRUCTION, BIDDING, OR PERMIT

Sundeep Inti., Ph.D.

The following are the objectives of this project:

Figure 2-1 Geocell Supplied.

Figure 2-2 Application of Geocell (US Army Corps).

Figure 2-3 Geocell Construction Sequence.

Figure 2-4 Mechanics of Geocell Reinforcement.

was calculated. • Single geocell reinforcement

geocells, and • A thinner unreinforced or geocell-

3.1 PARAMETRIC STUDY

Figure 3-2 Locations of Output Evaluated from FEA.

Figure 3-9 Hoop Strains on Geocell 4” and 6” (Influence of Infill Modulus).

3.5 INFLUENCE OF SUBGRADE MODULUS

Figure 3-20 Hoop Strain on Geocell 3” (Influence of Subgrade Modulus).

Figure 3-24 Hoop Strain on Geocell 4” (Influence of Subgrade Modulus).

Figure 3-28 Hoop Strains on Geocell 6” (Influence of Subgrade Modulus).

Strain on Subgrade Increase

Table 3-4 Performance of Three Geocell for Various Parameters

Subgrade Poor infill material (2 ksi) and good cover

Tests with Geocell and No Geocell (58)

Tests Not Required

4.3.2 Base and Subgrade Selection

using the Portable

4.4.1 Laboratory Set-Up

Figure 4-1 Photo of Fabricated Box

Figure 4-2 Sample Preparation Using Vibratory Compactor

Figure 4-3 Applied Load Cycle

Figure 4-4 Locations of Stress and Strains Measurement Transducers

Table 4-6 Strain Gauges

4.4.1.3 Earth Pressure Cells

Figure 4-8 Noise removal from pressure cell readings

Figure 4-12 Vertical Stresses on Subgrade a) Geocell, b) No Geocell.

Figure 4-14 Hoop Strains on Geocell.

Table 4-8 Stresses Measured on Top of Subgrade (below loading plate)

Table 4-10 Hoop Strain (cell below loading plate)

Vertical Compressive Strain

5.2 LOW VOLUME ROAD DESIGN AASHTO

B6 Calculate the average horizontal stress at (𝜎𝜎ℎ𝑣𝑣 + 𝜎𝜎ℎ𝑠𝑠 )

Cover over geocell(inch) 4

Geocell depth (inch) 6

Radius of Loading Plate (inch) 6

Contact Pressure (psi) 100

Bearing Capacity Coefficient Nc 2.8

Stress Reduction Calculation

1 99.6 71.2 Active Earth Pressure Coefficient 0.72

2 96.8 69.3 Horizontal Stress on Top of Geocell (psi) 59.3

3 91.1 65.1 Horizontal Stress on Bottom of Geocell (psi) 26.4

4 82.9 59.3 Average Horizontal Stress (psi) 42.88

6 64.6 46.2 Reduction in stress 7.07

Significance Codes : 0'', 0.001 '', 0.01'', 0.05 '.'

Significance Codes : 0'', 0.001 '', 0.01'', 0.05 '.'

Significance Codes : 0'', 0.001 '', 0.01'', 0.05 '.'

Significance Codes : 0'', 0.001 '', 0.01'', 0.05 '.'