Guangfeng Qu - PHD Thesis

SELECTED ISSUES ON THE PERFORMANCE OF EMBANKMENTS ON CLAY FOUNDATIONS (Spine title: Selected issues on the performance of embankments) (Thesis
format: Integrated-article)
By
Guangfeng Qn
Graduate Program in Engineering Science Department of Civil and Enviromental Engineering
A thesis submitted in partial fulfillment of the requirement for the degree of Doctor of Philosophy
Faculty of Graduate Studies The University of Western Ontario London, Ontario,Canada Guangfeng Qu 2008
1*1
Library and Archives Canada Published Heritage Branch

395 Wellington Street Ottawa ON K1A0N4 Canada
Bibliotheque et Archives Canada Direction du Patrimoine de I'edition

395, rue Wellington Ottawa ON K1A0N4 Canada
Your file Votre reference ISBN: 978-0-494-39317-8 Our file Notre reference ISBN: 978-0-494-39317-8
NOTICE: The author has granted a nonexclusive license allowing Library and Archives Canada to reproduce, publish, archive, preserve, conserve, communicate to the public by telecommunication or on the Internet, loan, distribute and sell theses worldwide, for commercial or noncommercial purposes, in microform, paper, electronic and/or any other formats. The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
AVIS: L'auteur a accorde une licence non exclusive permettant a la Bibliotheque et Archives Canada de reproduire, publier, archiver, sauvegarder, conserver, transmettre au public par telecommunication ou par Plntemet, prefer, distribuer et vendre des theses partout dans le monde, a des fins commerciales ou autres, sur support microforme, papier, electronique et/ou autres formats.
L'auteur conserve la propriete du droit d'auteur et des droits moraux qui protege cette these. Ni la these ni des extraits substantiels de celle-ci ne doivent etre imprimes ou autrement reproduits sans son autorisation.
In compliance with the Canadian Privacy Act some supporting forms may have been removed from this thesis. While these forms may be included in the document page count, their removal does not represent any loss of content from the thesis.
Conformement a la loi canadienne sur la protection de la vie privee, quelques formulaires secondaires ont ete enleves de cette these. Bien que ces formulaires aient inclus dans la pagination, il n'y aura aucun contenu manquant.
Canada
THE UNIVERSITY OF WESTERN ONTARIO FACULTY OF GRADUATE STUDIES CERTIFICATE OF EXAMINATION
Supervisor
Examining Board
Dr. Sean Hinchberger
Dr. Tim Newson
Co- Supervisor Dr. Ernest Yanful
Dr. K.Y. Lo
Dr. John Dryden
Dr. James Blatz The thesis by Guangfeng Qu Entitled Selected issues on the performance of embankments on clay foundations
is accepted in partial fulfillment of the Requirement for the degree of Doctor of Philosophy
Date March, 17,2008
Dr. Jianddong Ren Chair of Examining Board:
11
ABSTRACT This thesis examines selected issues related to the performance of earthfill embankments constructed on soft clay foundations. The primary objectives of the thesis are: to extend an existing elastic-viscoplastic (EVP) constitutive model to describe the influence of micro-structure and strength anisotropy on the engineering response of soft clay, to investigate the impact of clay structure on the performance of a full-scale test embankment on soft clay, and to evaluate the significance of three-dimensional effects on the behaviour of three test embankments constructed on soft clay foundations. Firstly, in this thesis, generalized EVP theory is used to evaluate the viscous response of 19 clays reported in the literature. It is shown that the viscous response of clay, including rate-dependent and time-dependent behaviour in different types of experiments, can be quantitively characterized using a unique set of viscous parameters. A practical methodology to determine the EVP constitutive parameters is provided. Next, an existing EVP constitutive model is extended to account for the influence of micro-structure and anisotropy on the engineering response of rate-sensitive natural clay. Microstructure and the process of destructuration are mathematically simulated using a state-dependent fluidity parameter. The EVP model also incorporates a structure tensor that can be used to describe strength anisotropy of natural clay. The extended structured and anisotropic models are shown to describe the responses of undisturbed structured clays, such as Saint-Jean-Vianney clay, Gloucester clay, and St. Vallier clay. Lastly, four case studies are used to investigate the impact of microstructure and destructuration on the performance of embankments founded on soft clay and the effects of 3-dimensional geometry on test embankment behaviour. The Gloucester test
iii
embankment is studied using the structured EVP model. This case is used to examine the impact of destructuration on strength gain in the Gloucester foundation during staged construction. In addition, three embankment cases in Vernon British Columbia, St.
Alban Quebec, and Malaysia are studied using 3-dimensional finite element analysis to examine the impact of 3-dimensional geometry on the performance of test embankments.
Key words: elastic-viscoplastic, viscosity, rate-sensitivity, natural clay, microstructure, anisotropy, case study, three-dimension.
iv
CO-AUTHORSHIP This thesis is prepared in accordance with the regulations for Manuscript format thesis stipulated by the Faculty of Graduate Studies at The University of Western Ontario. Chapters 2 and 4 of this thesis are the current versions of manuscripts in preparation for submission as papers, which will be co-authored by Guangfeng Qu and S.D. Hinchberger. Chapter 6 is a modified version of a submitted paper co-authored by G. Qu, S.D. Hinchberger and K.Y. Lo. Chapters 3 and 5 are the manuscripts currently in review coauthored by S.D. Hinchberger and G. Qu, and S.D. Hinchberger, G. Qu, and K.Y. Lo, respectively. Guangfeng Qu conducted numerical analysis and wrote the draft of the chapters. Dr. Sean Hinchberger assisted in interpretation of the results and the writing of the chapters. Dr. K.Y. Lo assisted in the interpretation of the results and the writing in Chapters 2, 5, and 6.
ACKNOWLEDGEMENT The author wishes to express his deepest gratitude and appreciation to his advisor, Dr. Sean D. Hinchberger for his insightful guidance, friendly encouragement, and continuous support throughout the research and graduate studies. The constructive and critical advice given by Dr. K.Y. Lo is greatly appreciated. The author also thanks Dr. Tim Newson, Dr. Julie Shang, Dr. M. Hesham El Naggar, and Dr. Ernest Yanful for sharing their knowledge during the general course work. The author wishes to acknowledge the Geotechnical Research Center, the Department of Civil and Environmental Engineering at University of Western Ontario for technical and clerical support. Many thanks are given to the friends and colleagues for their supports and interesting discussions during the past four years. Finally, the author wishes to thank his wife, Yanming, for her love, support, and patience.
VI
TABLE OF CONTENTS page CERTIFICATE OF EXAMINATION ABSTRACT CO-AUTHORSHIP ACKNOWLEDGEMENT TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES NOMENCLATURE CHAPTER 1 INTRODUCTION 1.1 Introduction 1.2 Definitions 1.3 Thesis Objectives and Outline 1.4 Original Contributions References CHAPTER 2 EVALUATION OF THE VISCOUS BEHAVIOUR OF NATURAL CLAY USING GENERALIZED VISCOPLASTIC THEORY 2.1 Introduction 2.2 Theoretical Background 2.2.1 Brief introduction of elastic-viscoplastic theory 2.2.2 Strain-rate controlled testing 2.2.3 Link with the isotache concept 2.2.4 Alternative flow function - the exponential law 2.2.5 Stress-controlled testing 2.3 Evaluation 16 16 17 17 20 23 24 25 27 ii iii v vi vii xi xii xix 1 1 3 5 7 10
vn
2.3.1 Rate dependency of preconsolidation pressure 2.3.2 Undrained shear strength versus strain-rate 2.3.3 Secondary compression 2.3.4 Summary 2.4 Selection of Parameters 2.4.1 The measurement of a 2.4.2 The measurement of a and yvp 2.5 Summary and Conclusion References CHAPTER 3 A VISCOPLASTIC CONSTITUTIVE APPROACH FOR RATESENSITIVE STRUCTURED CLAYS 3.1 Introduction 3.2 Theoretical Formulation 3.2.1 Overstress viscoplasticity 3.2.2 Numerical overstress 3.2.3 Modification for soil structure 3.3 Methodology 3.3.1 Laboratory tests 3.3.2 Numerical approach 3.3.3 Selection of constitutive parameters 3.4 Evaluation (Saint-Jean Vianney Clay) 3.4.1 Theoretical behaviour of the model for CIU triaxial compression 3.4.2 Calculated and measured behaviour for constant rate-of-strain triaxial compression 3.4.3 CIU triaxial creep tests 3.4.4 Theoretical response for constant rate-of-strain consolidation 3.4.5 Constant rate-of-strain consolidation 3.5 Summary and Conclusions References
27 29 32 33 34 35 36 38 41
71 71 75 75 77 78 81 81 82 83 87 87
88 90 93 94 96 100
viii
CHAPTER 4 THE STUDY OF STRUCTURE AND ITS DEGRADATION ON THE BEHAVIOUR OF THE GLOUCESTER TEST 131 131 132 136 136 139 142 142 144 144 147 151 155 187 187 188 190 193 196 202 208 210
EMBANKMENT 4.1 Introduction 4.2 Background 4.3 Methodology 4.3.1 Model 1 -Hinchberger and Rowe Model 4.3.2 Model 2 - Structured Elastic-viscoplastic (EVP) Model 4.3.3 Finite Element Mesh 4.3.4 Constitutive Parameters 4.4 Results 4.4.1 Analysis using the Unstructured EVP Model (Model 1) 4.4.2 Analysis using the Structured EVP Model (Model 2) 4.5 Summary and Conclusions References CHAPTER 5 AN ANISOTROPIC EVP MODEL FOR STRUCTURED CLAYS 5.1 Introduction 5.2 General Approaches to Anisotropic Plasticity 5.3 Microstructure Tensor 5.4 Application to Tresca's Failure Criterion 5.5 Application to an Elastic-Viscoplastic Model 5.6 Evaluation 5.7 Summary and Conclusions References CHAPTER 6 CASES STUDY OF THREE DIMENSIONAL EFFECTS ON THE BEHAVIOUR OF TEST EMBANKMENTS 6.1 Introduction 6.2 Methodology 6.3 St. Alban Test Embankment Case
234 234 235 236
ix
6.3.1 Introduction 6.3.2 Soil Conditions 6.3.3 Geometry 6.3.4 Results 6.4 Malaysia Trial Embankment Case 6.4.1 Introduction 6.4.2 Soil Conditions 6.4.3 Geometry 6.4.4 Results 6.5 The Vernon Case 6.5.1 Introduction 6.5.2 Analysis 6.5.2 Results of Vernon Approach Embankment 6.5.3 Results of Waterline Test Fill 6.6 Discussion 6.7 Summary and Conclusion References CHAPTER 7 SUMMARY AND FURTHER WORK 7.1 Summary 7.2 Suggestions for Future Research References APPENDIXES APPENDIX A APPENDIX B APPENDIX C APPENDIX D APPENDIX E APPENDIX F APPENDIX G CURRICULUM VITAE
236 237 238 238 239 239 240 241 241 243 243 244 246 247 249 250 252 279 279 280 282 283 283 290 296 306 312 324 331 337
LIST OF TABLES page Table 2.1 Table 2.2 Table 3.1 Table 3.2 Table 4.1 Table 4.2 Geotechnical properties of 19 clays Summarized a for 19 clays Properties of Saint-Jean-Vianney clay, (after Vaid et al. 1979) Constitutive parameters for Saint-Jean-Vianney clay Material parameters used in both Model 1 and Model 2 for the numerical analysis of the Gloucester test embankment Viscosity-related parameters for Gloucester clay used by Model 1 and Model 2 Table 5.1 Table 5.2 Table 5.3 Table 6.1 Comparison of elastic-viscoplastic models Constitutive parameters for Gloucester Clay Constitutive parameters for St.Vallier Clay Parameters used in the numerical analysis of the three cases 160 215 216 217 255 48 50 106 107
159
XI
LIST OF FIGURES page Figure 1.1 Figure 1.2 Cross-section of embankment and typical undrained strength profile for the underlying foundation clay Schematic of an oedometer apparatus and a typical compression curve. Figure 1.3 Figure 2.1 Figure 2.2 Definition of the orientation angle, / Illustration of models for elastic viscoplastic materials Illustration of relations between strain-rate and yield stress (or undrained shear strength) in strain-rate controlled tests Figure 2.3 Figure 2.4 The link between the EVP model and the isotache concept The influence of the power law and exponent law flow functions on the relationship between yield stress and strain-rate Figure 2.5 Figure 2.6 Figure 2.7 Figure 2.8 Figure 2.9 Figure 2.10 Figure 2.11 Typical compression curve for secondary compression. Ranges of strain-rates in laboratory tests and in situ (modified from Leroueil and Marques, 1996) Relationship between preconsolidation pressure, a'p, and strainrate, smial, in log-log scale Relationship between undrained strength, Su, and axial strainrate > zaxiai > in log-log scale Relation between undrained strength and axial strain-rate for Drammen clay and Haney clay Comparison of a estimated from rate-controlled oedometer tests and undrained triaxial tests ( See Table 2.2). Evaluation on the ability of exponential and power law flow functions to represent the relationship between preconsolidation pressure and strain-rate Comparison of a estimated from secondary consolidation tests, rate-controlled oedometer tests, and undrained triaxial tests ( See Table 2.2). 56 57 53 54 14 15 52 13
58 59 60 61 62
63
Figure 2.12
65
Xll
Figure 2.13 Figure 2.14 Figure 2.15
Comparisons of a_ac, a_oed, and acreep with aavg Typical triaxial compression curves with step-changed strainrates. Illustration of the preferred range of load increment for the measurement of Ca Normalized <r'p - s relationship at 10% vertical strain
66
67 68
Figure 2.16
(sv =10%) for Berthierville clay at a depth of 3.9-4.8m (data from Leroueil et al. 1988) Figure 2.17 Normalized cr'p - s relationship at 10% vertical strain (ev =10%) for St. Alban clay from both laboratory tests and in situ observance (data from Leroueil et al. 1988) The influence of structure on the response of Bothkennar clay during oedometer compression (from Burland 1990). The influence of structure on the response of London clay during undrained triaxial compression (from Sorensen et al. 2007 and Hinchberger and Qu 2007). The state boundary surface, critical state line, and mathematical overstress of the structured soil model. Estimation of the aspect ratio, R, for the elliptical cap. Estimation of the yield surface parameter, M oc , in the overconsolidated stress range. Estimation of the intrinsic compressibility, A,, and structure parameter, co0, from oedometer compression for SJV clay. Intrinsic compressibility of different clays (adapted from Burland 1990). Estimation of n and a'j^ from undrained triaxial compression and oedometer compression for SJV clay Influence of continued post-peak straining on the power law exponent, n. Theoretical behaviour of the structured soil model during CIU triaxial compression.
69
70 108
Figure 3.1 Figure 3.2
109 110 Ill
112
Figure 3.6
113
Figure 3.7
114
Figure 3.8
115 118 119
xm
Measured and calculated behaviour of SJV clay during CIU triaxial compression. Measured and calculated undrained shear strength versus strainrate for SJV clay. Calculated and measured behaviour during CIU triaxial creep tests on SJV clay
120 122
123 125
Figure 3.14 Figure 3.15 Figure 3.16 Figure 3.17
Calculated and measured creep-rupture life for SJV clay Calculated and measured axial strain-rate versus time during CIU triaxial creep on SJV clay. Comparison of strain-rate at failure for peak strength and creep rupture - SJV clay. Theoretical behaviour of the structured soil model during constant-rate-of-strain K'0 -consolidation. Calculated and compression. measured behaviour during oedometer
126 127 128
Figure 3.18 Figure 3.19 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4 Figure 4.5 Figure 4.6 Figure 4.7
129 130 161 163 165 167 168 169 170
Measured and calculated compression curves of SJV clay during constant rate-of-strain consolidation. (a) Geometry of the Gloucester test embankment and (b) properties of Gloucester clay Influence of clay structure on the behaviour of Gloucester clay in undrained triaxial and oedometer compression tests Rate-sensitivity of the undrained shear preconsolidation pressure of Gloucester clay strength and
Long-term oedometer creep tests on Gloucester clay (data from Lo et al. 1976) The state boundary surface and critical state line for Model 1 and Model 2. Illustration of the theoretical response of Model 1 (Hinchberger and Rowe Model) Illustration of the theoretical response of Model 2
xiv
Figure 4.8
Comparison of the measured behaviour in CRS oedometer test on Gloucester clay and the corresponding theoretical response of Model 2 Comparison of the measured settlement at Gauge SI with the calculated settlement using Model 1
171
Figure 4.9
172 173
Illustration of the linear and bilinear virgin compression curves Zones of strength gain due to consolidation, 15 years after the construction of Stage 1- Contours of (Su /Su0 )cons Zones of strength gain due to consolidation, 4 years after the construction of Stage 1. Contours of (Su /Su0 )cons Comparison of measured settlement (Gauge SI) with calculated settlement using Model 2 Comparison of the measured and calculated settlement and excess pore water pressure using Model 1 and Model 2 Zones of strength loss due to destructuration, 15 years after construction of Stage 1. Contour of [Su /Su0) Zones of net strength gain (i.e. consolidation overshadows destructuration), 15 years after construction of Stage 1. Contour ofSJSu0>l Zones of net strength loss (i.e. destructuration overshadows consolidation), 15 years after construction of Stage 1. Contour ofSJSu0<l Development of zones of net strength gain from the 4th year to the 15th year in Stage 1 Development of zones of net strength loss from the 4th year to the 15th year in Stage 1 Zones of net strength increase, 7 years after construction of Stage 2 Zones of net strength loss 7 years after construction of Stage 2 Comparison of the compression curve in laboratory test with the measured long-term field compression of Gloucester clay under the Accommodation building (from McRostie and Crawford, 2001)
11A 175
176 177 179
Figure 4.16
180
Figure 4.17
181
Figure 4.18 Figure 4.19 Figure 4.20 Figure 4.21 Figure 4.22
182 183 184 185
186
xv
Figure 5.1
Illustration of the microstructure tensor, a]-, and the generalized stress tensor, a'i}2 for transverse isotropy. 218 219
Sample orientation, i. The effect of Aon the anisotropy of cu from Tresca's failure criterion. The effect of stress ratio, a[/a'c, Tresca's failure criterion. on the anisotropy of c u from
220
Figure 5.4
221 222
Conceptual behaviour of the 'structured' soil model. The effect of sample orientation, i, on the measured and calculated peak and post-peak undrained strength of Gloucester clay. The effect of sample orientation, i, on the measured (Law 1974) and calculated (a) axial stress versus strain and (b) excess pore pressure versus strain for Gloucester clay. The comparison for sample orientations, /, of 0 and 90 on the measured (Law 1974) and calculated (a) axial stress versus strain and excess pore pressure versus strain (b) stress paths for Gloucester clay. The effect of strain-rate on the peak strength of Gloucester clay (Data from Law 1974). The effect of sample orientation, i, on the peak strength of St. Vallier clay during CIU triaxial compression tests. The effect of sample orientation, /, on the measured (Lo and Morin 1972) and calculated (a) axial stress versus strain and (b) excess pore pressure versus strain for St. Vallier clay. The effect of sample orientation, i, on the measured (Lo and Morin 1972) and calculated stress paths for St. Vallier clay. Measured and calculated peak undrained shear strength versus strain-rate for St. Vallier clay 0=0).
223
Figure 5.7
224
Figure 5.8
225 226 227
228 229
230 231
Influence of A and co on apparent yield surface Influence of destructuration on the apparent yield surface of St. Alban clay
232
xvi
Compression curves from oedometer compression tests on intact and destructured specimens of St. Alban clay Strength profile assumed and measured using field vane and undrained (UU and CIU) tests (experimental data from La Rochelle et al. 1974) Plan view and cross-section of St. Alban test embankment Generated Meshes for 3D and 2D FEM model Measured and calculated vertical displacement of point 'O' for St. Alban Embankment Spatial displacement contour of 3D model for St. Alban embankment (at failure) Spatial displacement contour V.S. fissures at failure on the top surface on St. Alban Embankment The statistic table for the prediction on the failure thickness of Malaysia test embankment (data from MHA 1989b) Strength profiles for the Malaysia case (experimental data from MHA 1989a)
233
256 257 258
Figure 6.2 Figure 6.3 Figure 6.4 Figure 6.5 Figure 6.6 Figure 6.7 Figure 6.8
259 260 261 262
263 264
Figure 6.9 Figure 6.10 Figure 6.11 Figure 6.12 Figure 6.13 Figure 6.14
Plan view of Malaysia test embankment Measured and calculated settlement of Malaysia Trial Embankment Velocity field in central cross-section of 2D model for the Malaysia trial embankment (at failure) Velocity field in central cross-section of 3D model for the Malaysia trial embankment (at failure) Plan view of Vernon embankment (modified after Crawford et al. 1995) Longitudinal section through the embankment (after Crawford et al. 1995)
265 266 267 268
269 270
Distribution of vane strength with depth Vertical displacement of Vernon Approach Embankment in 2D analysis Plan view and 3D model of Vernon approach embankment xvii
271 272
Figure 6.18
Vertical displacement of Vernon Approach Embankment in 3D analysis 273 274 275
Figure 6.19 Figure 6.20 Figure 6.21 Figure 6.22 Figure 6.23
Spatial displacement contour of Vernon approach embankment Plan view and cross section A-A of Waterline test fill Measured and calculated displacement by 2D analysis for the Waterline Test Fill Measured and calculated displacement by 3D analysis for the Waterline Test Fill Illustration of 3D effect on the bearing capacity and the cases studied.
276 277 278
xvin
NOMENCLATURE
st] J2 (j'm p' q Sy tj seij svpy sî Su <j' <y'^ <j'ny ) a'}P G v K A Cs
deviatoric stress tensor secondary invariant of deviatoric stress tensor mean effective stress mean effective stress, /?'= (<J\ +<j'2+cr\ )/3 deviatoric stress, q = (cr\ -ar\ ) Kronecker's delta total strain-rate tensor elastic strain-rate tensor viscoplastic strain-rate tensor axial strain-rate undrained compression strength apparent preconsolidation pressure static yield surface intercept dynamic yield surface intercept overstress stress dependent shear modulus Poisson's Ratio slope of the e - ln(o^) curve in the overconsolidated stress range slope of the e - \n{a'v) curve in the normally consolidated stress range slope of the e - log(cr^) curve in the overconsolidated stress range slope of the e-log(cr^) curve in the normally consolidated stress range, Cc = ln(10)A
C.
xix
Ca e n a yvp yf, yjp 4>(F)
secondary compression index void ratio power law exponent rate-sensitivity parameter {=11 n) fluidity parameter, denoting the threshold strain-rate of viscosity fluidity of the structured state and the destructed state, respectively flow function slope of the Drucker-Prager envelope in ^2j2 - a'm stress space normally consolidated stress range slope of limit state line in yJ2j2 - a'm stress space - over consolidated stress range effective friction angle angle of dilatancy effective cohesion intercept in ^/2J 2 - a' m stress space - normally consolidated stress range effective cohesion intercept of the limit state in Û2 - o'm stress space over consolidated stress range dilation parameter to define plastic potential in O/C zone aspect ratio of the elliptical cap damage strain weighting parameter destructuration-rate parameter coefficient indicating the current degree of structure parameter indicating the initial degree of structure clay orientation respect to vertical direction
Moc
</>' W
C
cs
Mv R sd A b co co0 i
xx
A 77 W B
parameter of inherent soil anisotropy coefficient of structure anisotropy at crest width of embankment base width of embankment
ABBREVIATION EVP CRS 3D 2D elastic viscoplastic constant rate of strain-rate 3-dimensional 2-dimensional
xxi
CHAPTER 1 INTRODUCTION
1.1 Introduction
Recently, some researchers have begun to recognize the important effects of clay viscosity. The most common effects of viscosity on clay behaviour include: variation of undrained strength with strain-rate, variation of preconsolidation pressure with strainrate, and creep deformation under conditions of constant effective stress (i.e. secondary compression). These phenomena cannot be accounted for in soil mechanics using
conventional elasto-plasticity theories. For example, in triaxial compression tests, a reduction of the axial strain-rate by one order-of-magnitude usually results in a decrease in the undrained strength of about 10% for several clays (e.g. Leroueil et al. 1985; Graham et al. 1983). This introduces uncertainties for geotechnical designs where
stability is assessed using the undrained strength measured from standard laboratory tests as a result of the significant difference between strain rates during experiments and those operating during field performance. In addition, results from long-term consolidation tests and field observations indicated that secondary compression accounts for a significant portion of the long-term settlement of embankments founded on some clay (e.g. Lo et al. 1976; Crawford and Bozozuk 1990; and Hinchberger and Rowe 1998). Thus, there is a need to consider these viscous effects and their impact on the performance of structures founded on or in natural clay.
2 Most natural clays are structured to some degree, in addition to being ratesensitive. Leroueil and Vaughan (1990) and Burland (1990) have perhaps undertaken the most comprehensive studies of the general influence of structure on the behaviour of natural clay. The current and subsequent studies show that structure is as important as other basic engineering properties, such as void ratio and stress history, governing the engineering response of natural clay. It has been recognized that the degradation of structure during loading (destructuration) may lead to a significant reduction in undrained strength. Burland (1990) has shown that the measured yield stress from oedometer tests on structured clays is often much higher than that for the corresponding remolded samples. In most cases, structure permits natural clay to exist at a higher void ratio than the corresponding destructured or remolded clay leading to high compressibility for stresses exceeding the in situ yield stress. Lastly, strength anisotropy is another characteristic of many natural clays, which has been recognized by many researchers (e.g. Lo 1965; Lo and Morin 1972; Pietruszczak and Mroz 1983; and Zdravkovic et al. 2002). Lo and Morin (1972)
measured the significant impact of anisotropy on the response of two natural clays from eastern Canada during undrained triaxial tests on samples trimmed at different angles, /, to the vertical axis. As i was increased from 0 to 90, the undrained strength of both St. Vallier and Gloucester clay decreased by about 30% ~ 50%. Similar behaviour has been reported by other researchers (e.g. Jardine et al 1997; Symes et al. 1984). Hence, extending constitutive models for clay to account for the effects of viscosity, structure, and anisotropy would be desirable to fully capture the key engineering characteristics of natural clays. In addition, studying the performance of
embankment on natural clays would help to improve modern geotechnical design. This thesis focuses on these two issues. The following section defines important terms used throughout the remainder of this thesis.
1.2 Definitions
To facilitate reading of the thesis, this section provides definitions and discussion of some of the important terms and concepts utilized consistently throughout the thesis. Embankment Figure 1.1 shows a typical embankment cross-section and a typical undrained strength profile for the underlying foundation clay. In Figure 1.1, the symbols, W and B, represent the crest width and the base width of the embankment, respectively. The side slope is denoted using H: V. The undrained strength profile of a soft clay deposit typically comprises three layers: a crust, a transition layer, and a soft clay layer with depth. The undrained strength is typically constant within the crust layer. It decreases to a minimum value in the underlying transition layer, and then increases with depth in the soft clay layer. Preconsolidation pressure The preconsolidation pressure (cr'p) of clay represents the maximum effective vertical stress that the clay has experienced in the past. The preconsolidation pressure can be estimated using an oedometer compression test. Figure 1.2 shows the schematic of an oedometer apparatus, together with a typical compression curve of clay plotted as void ratio versus logarithm of vertical effective pressure. Typically, the compression curve has two distinct portions. The first portion is relatively flat, representing the elastic state with low compressibility. The second portion has a greater slope and denotes the
plastic state corresponding to high compressibility and irrecoverable strains.
The
preconsolidation pressure can be determined from the transition stage between the two portions using a widely accepted Casagrande procedure (see Holtz and Kovacs 1981 and Craig 1997). Structure The term 'structure' used in this thesis specifically refers to the microstructure of clay, which arises from fabric effects and inter-particle bonding or cementation. The effect of structure on the mechanical response of natural clay is significant. Structure typically imparts additional meta-stable strength to natural clay, which leads to strength loss with large-strain. In addition, structured clay usually exists at higher void ratio than the equivalent reconstituted clay. Such a state in clay is meta-stable, and leads to high compressibility under further loading after yielding. illustrated for example by the dashed line in Figure 1.2. Anisotropy In this thesis, the term 'anisotropy' specifically refers to the variation of undrained strength with rotation of principal stresses relative to the axis of natural deposition of a clay. The undrained strength of clay is usually measured using a triaxial compression apparatus where the drainage from the specimen is not permitted during loading. In addition to the vertical specimen, clay can be trimmed at different angles, /, to the vertical axis (See Figure 1.3). The angle i denotes the sample orientation. A clay with anisotropy typically yields different undrained strengths depending on the sample orientation. In this case, to accurately access the stability of embankments and slopes, the strength anisotropy has to be taken into account in accordance with the different The influence of structure is
5 orientation of the major principal stress along the potential failure surface (see Lo and Milligan 1967). Yield surface Clays have a yield surface in generalized stress state. The yield surface is defined as a surface in stress space, which denotes stress states at which yielding begins. Inside of the yield surface, stress states are elastic. The classic yield surfaces include: Cam-clay yield surface, Modified Cam-clay yield surface, and elliptical yield surface (see Roscoe and Schofield 1963; Roscoe and Burland 1968; Chen and Mizuno 1990; Atkinson 1993). For inviscid soil, the yield surface is mainly governed by stress history. For viscous clay, the location of yield surface in stress space is also dependent on the loading strain-rate.
1.3 Thesis Objectives and Outline

This thesis has two aims: (i) to develop a general and efficient constitutive framework, which can take into account the viscosity, structure, and anisotropy of natural clays, and (ii) to study selected issues affecting the performance of earth-fill embankments built on deposits of natural clay. In Chapter 2, a general elastic-viscoplastic (EVP) theory is described and used to derive the relationships between undrained strength and strain-rate, preconsolidation pressure and strain-rate and the coefficient of secondary compression in terms of two EVP viscosity parameters. Nineteen clays from the literature are used to show that a unique set of viscous parameters can be used to describe the rate-sensitivity and timedependency of many natural clays. Chapter 3 extends the EVP model to account for clay structure by introducing a state-dependent fluidity parameter, and a damage law to describe the destructuration
6 process. Calculated and measured behavior of Saint-Jean-Vianney clay is compared for constant-rate-of-strain /^-consolidation, and both isotropically consolidated undrained triaxial compression (CIU) tests and constant stress creep tests. The ability of the
extended constitutive framework is evaluated by comparing measured and calculated creep rupture response, and the measured and calculated influence of strain-rate on the peak undrained shear strength, post-peak undrained shear strength, and apparent preconsolidation pressure of Saint-Jean-Vianney clay. Chapter 4 further investigates the influence of structure degradation on the field behaviour of a test embankment constructed at Canadian Forces Base (CFB) in Gloucester, Ontario. The calculated long-term settlement obtained using both structured and non-structured EVP models are compared with the measured response. This
comparison suggests that the extended EVP model gives improved predictions of embankment behaviour. Then, the spatial distribution of 'destructuration' in the The
Gloucester foundation is examined numerically with time after construction.
locations of possible weakened zones (destructured) in soil foundation are identified and the mechanism governing the formation of these zones is investigated. The results may have implications for the design and analysis of stage constructed embankments Chapter 5 introduces a tensor approach, which enables the EVP model to account for the strength anisotropy of natural clays. The advantages and limitations of this approach are discussed with reference to other constitutive alternatives. Then the new model is evaluated by comparing the calculated behaviour in triaxial compression tests with the measured behaviour of two anisotropic natural clays. The comparison shows that the extended EVP model is able to simulate the anisotropic undrained strength and
7 pore water pressure response of Gloucester clay and St. Vallier clay for various sample orientations. Chapter 6 examines three cases involving full-scale test embankments built on soft clay deposits. The cases are examined using both two-dimensional (2-D) plane strain finite element and three-dimensional (3D ) finite element analysis taking account of the true 3D geometry of each case. By comparing the calculated collapse fill thickness from 2D and 3D analyses, it is shown that 3D effects are quite significant for all of the test embankments examined. Finally, by comparing Finite Element (F.E.) results with a well known bearing capacity factor, it is shown that the use of bearing capacity factors commonly used for shallow foundations can be used to approximately assess 3D test embankments with an aspect ratio of base length to base width less than 2. The analysis and results presented provide practical insight into some of the key factors that should be taken into account for the design and construction of embankments and test fills on soft clay deposits. Chapter 7 presents a summary of this study and suggestions for further work.
1.4 Original Contributions

The original contributions of this thesis are summarized as following: Chapter 2 shows that the viscosity of clays can be mathematically quantified using a unique set of constitutive parameters. In addition, practical guidance is given to select and measure the viscous parameters directly from experiments without trial and error. The research in this chapter has been presented in the following manuscripts:
8 Qu, G. and Hinchberger S.D. (2007) Evaluation of the viscous behaviour of natural clay using a generalized viscoplastic theory. Geotechnique, Submitted October 2007. Prepared from the research presented in Chapter 2. Hinchberger, S.D. and Qu, G. (2007) Discussion: the Influence of structure on the timedependent behaviour of a stiff sedimentary clay. Geotechnique. Accepted.
The structure and strength anisotropy effects of natural clays are accounted for within a generalized elastic viscoplastic model described in Chapters 3 and 5. The research in this chapter has been presented in the following manuscripts:
Hinchberger, S.D. and Qu, G.(2006) A viscoplastic constitutive approach for structured rate-sensitive natural clays. Canadian Geotechnical Journal, Re-Submitted November 2007. Prepared from the research presented in Chapter 3. Hinchberger, S.D., Qu, G. and Lo, K.Y.(2007) A simplified constitutive approach for anisotropic rate-sensitive natural clay. International Journal of Numerical and Analytical Methods in Geotechnical Engineering. Submitted January 2007, resubmitted October 2007. Prepared from the research presented in Chapter 5
The case studies in Chapter 4 and 6 highlight the significant influence of clay structure and 3D geometry on the performance of test embankments founded on soft clay. The research in these chapters has been presented in the following manuscripts:
9 Qu, G. and Hinchberger, S.D. (2007) Clay microstructure and its effect on the performance of the Gloucester test embankment. Geotechnical Research Centre Report No. GEOT2007-15, the University of Western Ontario, London, Ontario, CAN. Prepared from Chapter 4. Qu, G. Hinchberger, S.D., and Lo, K.Y. (2007) Case studies of three dimensional effects on the behaviour of test embankments. August 2007. Prepared from Chapter 5. Canadian Geotechnical Journal. Submitted
10
References
Atkinson, J.H. 1993. An introduction to the mechanics of soils and foundations : through critical state soil mechanics. McGraw-Hill Book Co., New York. Burland, J.B. 1990. On the compressibility and shear strength of natural clays. Geotechnique, 40(3): 329-378. Chen, W.F., and Mizuno, E. 1990. Nonlinear analysis in soil mechanics : theory and implementation. Elsevier Science Publishing Company Inc., New York, NY, U.S.A. Craig, R.F. 1997. Soil Mechanics. E & FN Spon, New York. Crawford, C.B., and Bozozuk, M. 1990. Thirty years of secondary consolidation in sensitive marine clay. Canadian Geotechnical Journal, 27(3): 315-319. Graham, J., Noona, M.L., and Lew, K.V. 1983. yield states and stress-strain relationships in a natural plastic clay. Canadian Geotechnical Journal, 20(3): 502-516. Hinchberger, S.D., and Rowe, R.K. 1998. Modelling the rate-sensitive characteristics of the Gloucester foundation soil. Canadian Geotechnical Journal, 35(5): 769-789. Holtz, R.D., and Kovacs, W.D. 1981. An introduction to geotechnical engineering. Prentice Hall, Inc., Toronto. Jardine, R.J., Zdravkovic, L., and Porovic, E. 1997. Anisotropic consolidation, including principal stress axis rotation: Experiments, results and practical implications. In Proc. 14th Int. Conf. Soil Mech. Found. Engng. Hamburg, Vol.4, pp. 2165-2168.
11 Leroueil, S., Kabbaj, M., Tavenas, R, and Bouchard, R. 1985. Stress-strain-strain rate relation for the compressibility of sensitive natural clays. Geotechnique, 35(2): 159-180. Leroueil, S., Bouclin, G., Tavenas, F., Bergeron, L., and La Rochelle, P. 1990. Permeability anisotropy of natural clays as a function of strain. Canadian Geotechnical Journal, 27(5): 568-579. Lo, K.Y. 1965. Stability of slopes in anisotropic soils. Journal of the Soil Mechanics and Foundations Division American Society of Civil Engineers, 91(SM4): 85-106. Lo, K.Y., and Milligan, V. 1967. Shear strength properties of two stratified clays. American Society of Civil Engineers Proceedings, Journal of the Soil Mechanics and Foundations Division American Society of Civil Engineers, 93(SM1): 1-15. Lo, K.Y., and Morin, J.P. 1972. Strength anisotropy and time effects of two sensitive clays. Canadian Geotechnical Journal, 9(3): 261-277. Lo, K.Y., Bozozuk, M., and Law, K.T. 1976. Settlement analysis of the gloucester test fill. Canadian Geotechnical Journal, 13(4): 339-354. Pietruszczak, S., and Mroz, Z. 1983. On hardening anisotropy of ko-consolidated clays. International Journal for Numerical and Analytical Methods in Geomechanics,, 7(1): 19-38. Roscoe, K.H., and Burland, J.B. 1968. On the generalised stress-strain behaviour of wet clay. Cambridge Univesrity Press, Cambridge. Roscoe, K.H., Schofield, A.N., and Thurairajah, A. 1963. Yielding of clays in states wetter than critical. Geotechnique, 13(3): 211-240.
12 Symes, M.J.P.R., Gens, A., and Hight, D.W. 1984. Undrained anisotropy and principal stress rotation in saturated sand. Geotechnique, 34(1): 11-27. Zdravkovic, L., Potts, D.M., and Hight, D.W. 2002. The effect of strength anisotropy on the behaviour of embankments on soft ground. Geotechnique, 52(6): 447-457.
13
Figure 1.1 Cross-section of embankment and typical undrained strength profile for the underlying foundation clay
Typical Strength Profile Su (kPa)
Transition Layer
Clay foundation
Soft Clay Layer
Depth
Depth
14 Figure 1.2 Schematic of an oedometer apparatus and a typical compression curve.
Elastic stage
Elasto-plastic stage
*K)
\
i
Meta-stable
Ae
Schematic of an oedometer apparatus (after Holtz and Kovacs 1981)
15 Figure 1.3 Definition of the orientation angle, i
t Vertical Direction (Direction of deposition/gravity) A j = 0

l 45
Ground level
*
"S. \
^_._.=^=^.-_._._
See details.
| p,-^
' l.y
k
Z ^
.i
Clay layer
1/ *J
/
"J Horizontal Direction "77^ 77^ s7^T
l - Sample !----
16
CHAPTER 2
EVALUATION OF THE VISCOUS BEHAVIOUR OF NATURAL CLAY USING GENERALIZED VISCOPLASTIC THEORY
2.1 Introduction
In 1957, Suklje (1957) proposed the isotache concept to describe the timedependent behaviour of clay in one-dimensional compression. The isotaches were
defined as a series of t-a'v compression curves constructed from tests performed at various constant strain-rates. Since then, the concept of isotaches has been extended gradually over time to general stress space. For example, Tavenas et al. (1978) estimated isotaches in p'-q stress space for St. Alban clay using a series of drained and undrained triaxial compression tests and creep tests. Graham et al. (1983) studied the influence of strain-rate on Belfast clay using undrained triaxial compression, undrained triaxial extension and one-dimensional oedometer compression tests. The results of Graham et al. (1983) were expressed in terms of isotaches also plotted in p'-q stress space. In addition to rate-sensitivity, many natural clays exhibit significant creep or secondary compression at constant effective stress during incremental oedometer tests. Such behaviour is indicative of the viscous response of clay. Although it is generally recognized that there are similarities between the time-dependent response of clay during undrained and drained compression, so far there has not been a comprehensive study to generalize the viscous characteristics of clay for these different stress paths (e.g. triaxial compression and oedometer compression). This chapter uses a generalized viscoplastic theory to examine the viscous
A version of this chapter has been submitted to Geotechnique 2007
17 response of 19 clays reported in the literature. The main objectives of this study are: (i) to investigate if a unique set of viscous parameters can be used to describe the ratesensitivity of clay during drained and undrained triaxial compression tests and the secondary compression (or creep) response exhibited in incremental oedometer tests, (ii) to link the isotache concept (Suklje 1957; Tavenas et al. 1977; Graham et al. 1983; and Leroueil et al. 1985) with generalized elastic viscoplastic constitutive theory, and (iii) to provide guidance for the selection of viscosity parameters for viscous clays. To achieve these objectives, theoretical relationships are derived from viscoplastic theory for undrained strength and preconsolidation pressure versus strain-rate expressed in terms of two viscosity parameters called the fluidity parameter and rate-sensitivity parameter. In addition, a theoretical relationship is derived relating the fluidity and rate-sensitivity parameters to the secondary compression index. The measured behaviour of 19 clays is evaluated using the derived relationships to show that a unique set of viscoplastic parameters exist for the viscous response of clays during loading along stress paths involving drained compression or undrained shear. Such a study should be of interest to engineers and researchers in the field of soil mechanics.
2.2 Theoretical Background 2.2.1 Brief introduction of elastic-viscoplastic theory Figure 2.1 illustrates the main characteristics of elastic-viscoplastic theory. Figure 2.1a shows a general 1-D rheologic model for elastic-viscoplastic theory, which comprises a linear elastic spring in series with a plastic slider and viscous dashpot in parallel. For this type of model, the strain-rate, s , can be expressed in terms of elastic
18 (se) and viscoplastic (svp) components as follows: s =ee+svp [2.1]
Perzyna (1963) proposed an overstress viscoplastic theory to describe the ratesensitivity of materials at yield during uniaxial tension. For a steel bar in tension (see Figure 2.1b), the viscoplastic strain-rate proposed by Perzyna (1963) is:
(.
V
axial
r
axial
vp
(1) 0
for for
O'axial- Vy axial ~<Ty
>0 ^
0
[2.2]
where yîs the fluidity parameter with unit of inverse time, n is a power law exponent, <raxial is the rate-dependent axial yield stress, ay is the yield stress mobilized at very low strain-rate. The plastic potential from von Mises failure envelop is unity (1). Perzyna's theory has been extended to geologic materials (e.g. Desai and Zhang 1987, Katona and Mullert 1984, Adachi and Oka 1982), which typically possess a state boundary surface denoted by A-C-E-0 in Figure 2.1(c). Stress states inside the state boundary surface are elastic while stress states lying outside the surface are considered to be viscoplastic. The yield surfaces A-C-E and B-D-F are typically referred to as static and dynamic yield surfaces, respectively. The static yield surface (A-C-E) defines the initial onset of time-dependent viscoplastic behaviour. The dynamic yield surface (B-DF) passes through the current stress state and is used to define the plastic potential for associated flow and the degree of overstress. The term 'dynamic' yield surface is used since viscoplasticity has typically been viewed as a dynamic process (see Sheahan 1996 and Adachi and Oka 1982). In accordance with Adachi and Oka (1982) and Hinchberger and Rowe (1998), the EVP constitutive equation for normally-consolidated soil is:
19
*, = * ; + # = < V T ; + ( * F ) >
OF
da'
[2.3]
where Cjjkl is the elastic compliance tensor and a- is the effective stress tensor. The scalar function, ^ ( F ) , is the flow function governing the magnitude of the viscoplastic strain-rate and F can be any valid yield surface function from plasticity theory. The associated plastic potential, 8F do' is a unit vector normal to the dynamic yield surface
in <y'm - yJ2J2 space (see Appendix A for details). The theoretical relationships derived in the following sections also apply to the state boundary surface (ACEO) depicted in Figure 1(d), which is commonly found for soils. Two different flow functions, 0(F) , are evaluated. The first is based on the power law (Norton 1929) extended to general stress states as follows:
<W = H< } /< S ) )"

and
o-'W-o-'W > o
a>W
t2-4]
m)-{?
2.1c, a'^
my my
my my
,(s)
<Q
where yvp is the fluidity parameter and n is the power law exponent. As shown in Figure is the static yield surface intercept (Point A in Figure 2.1c) and a'm(^ is the
intercept of dynamic yield surface with the mean stress axis (Point B). The stress state denoted by Point D in Figure 2.1c is a state of overstress (o'^-o'^ >0). This type of
flow function has been adopted by Adachi and Oka (1982) and Hinchberger and Rowe (1998).
20 The second form of <f>(F) considered is an exponential flow function: ^F) = f-cxp[n(-^-l)]
my
[2.5]
where again yvph the fluidity parameter and n governs the rate-sensitivity. Finally, although there are many different hardening laws, in the following discussion, kinematic strain hardening is assumed. Expansion or contraction of the static yield surface ( c r ^ ) is governed by the viscoplastic volumetric strain (s^) viz.: d<
s )
=|^<
s )
d^
[2.6]
where e is void ratio, and A and K are the compression index and recompression index, respectively.
2.2.2 Strain-rate controlled testing In this section, the influence of strain-rate in rate controlled laboratory tests is evaluated by deriving relationships between strain-rate and yield stress and strain-rate and undrained shear strength for these tests. The relationship between yield stress and strain-rate First, considering CRS (constant rate of strain) isotropic compression, a relationship between l o g a ' ^ and l o g ( ^ a , ) can be derived explicitly from elasticviscoplastic theory using Equations [2.3] and [2.4] as follows:
VP axial
fi^s/^s)"da'dF
axial .
where all of the above parameters have been defined above, and 1/3 is the plastic potential for axial strain in isotropic compression (see Appendix A). This plastic
21 potential would apply to yield surfaces such as the modified Cam-clay model or the elliptical cap model (See Figure 2.1c). Taking the logarithm of [2.7] gives:
my
l&Zai) = nto&
V
+ log(r v; ') + log(l/3)
[2.8a]
">y J
and re-arranging yields: l o g ( < } ) = a l o g ( ^ ) + [ l o g ( < s ) ) - a logfrvp) - a log(l/3) ] where a( =11 n) is the rate-sensitivity parameter, and o'^ [2.8b] is the strain-rate dependent
isotropic yield stress corresponding to the axial strain-rate, svial. At yield and failure, the elastic component of strain, e., can be neglected without significant influence on the rate sensitivity relationship (see Appendix B). Hence, the viscoplastic strain-rate in Equation [2.8b] can be expressed in terms of the total strainrate, viz. l o g ^ i - 0 )= a \og(eaxial) + log(aJ'>)- a l o g ( ^ ) - a log(i) [2.9]
Equation [2.9] shows that the power law flow function in Equation [2.4] implies a linear relationship between log^cr^j and log(T a/ ), which is plotted as a straight line A-B in Figure 2.2a. Using a similar approach to that described above, relationships between log(S*d)) and l o g f o ^ ) and log\(T'p(d)) and l o g ^ ^ ) can be derived, where S(d) is the ratedependent undrained shear strength and <j'pd) is the rate-dependent preconsolidation pressure. For most commonly used yield surfaces (e.g. Cam-clay, Modified Cam-clay, and the elliptical cap), there is a fixed relationship between the top of the yield surface
22 (see Point F in Figure 2.1c) and the yield surface intercept with the mean stress axis (see Point B in Figure 2.1c) e.g.:
g(d) g(s)
r'(d)
[2.10]
my
my
Substituting Equation [2.10] into [2.9] and modifying the plastic potential for axial strain during undrained tests gives: log{s^)=alog(saxial) + log(5 H ( s >)-alog( r -)-log(J|) [2.11]
and by similar argument, since S(ud) I cr'Â) is also constant: logfo ) = a \og(saxial) + log(cx;(s>)- a logfr *) - a l o g ^ )
[2.12]
Equations [2.11] and [2.12] are also straight lines in log-log scale as shown in Figures 2.2(c) and 2.2(b), respectively. Equation [2.12] is essentially consistent with the following relationship used by Leroueil and Marques (1996) for the variation of preconsolidation pressure versus strain-rate in oedometer tests: log{a'p)=alog(eaxial) +A [2.13]
where a and A are constants. Referring to Figure 2.2a, there are 3 characteristics of Equation [2.9] that should be discussed. First, the power law flow function (Equation[2.4]) implies a linear log-log relation between isotropic yield stress and axial strain-rate, represented by a straight A-B line in Figure 2.2a. Second, the slope of line A-B, a = IIn , represents the rate-sensitivity of the isotropic yield stress: as a increases, the yield stress becomes more rate-sensitive. Third, the linear A-B line terminates at Point A, whose coordinates are <j'^} (the static
23
yield surface) and yvp73 . Point A denotes the static yield surface intercept in Figure 2.1c (i.e. (y'^y) = cr'^)). The value of yvp 13 can be considered as a threshold strain-rate above which strain-rate effects are mobilized. For axial strain-rates less than yvp 13, the
isotropic yield stress is rate-insensitive (e.g. the EVP model retrogresses an elastic plastic model). Similar principles apply to the preconsolidation pressure and undrained strength, as shown in Figure 2.2(b) and 2.2(c) where again the slope, a, increases with the rate sensitivity. In summary, if a power law flow function (see Equation. [2.4]) is used in conjunction with elastic-viscoplastic theory, then there will be linear log-log relationships between Sd) - eaxial and cr'A) - axial, with the same magnitude of slope, a , as shown in Equations [2.11] and [2.12]. In the following sections, these derived relationships will be tested by evaluating the behaviour of 19 clays in rate-controlled undrained triaxial and oedometer compression tests reported in the literature.
2.2.3 Link with the isotache concept Figure 2.3a demonstrates the main characteristics of a power law EVP model, in a generalized stress space with an additional axis of strain-rate. In contrast with
conventional critical state theory, the EVP model implies a family of dynamic yield surfaces, which, as shown by the dashed lines (e.g. 1-2-3, 4-5-6, and 7-8-9) in Figure 2.3a, expand with increasing strain-rate. The static yield surface (A-C-E) in Figure 2.3a defines the onset of time-dependent behaviour. Viscous behaviour is mobilized only when the stress state exceeds the static yield surface. In addition, lines A-B, C-D, and EF shown in Figure 2.3a correspond to those shown in Figures 2.2(a), (b), and (c), and the
24
lines A-B, C-D, and E-F can be linked to each other by the yield surface function. As shown in Figure 2.3b, a series of isotaches can be constructed by projecting the dynamic yield surfaces in Figure 2.3a onto the yJ2J2 -a'm plane. The projected dynamic yield surfaces define the rate-dependent yielding in generalized stress space and are essentially consistent with the isotache concept proposed by Suklje (1957), and extended by Tavenas et al. (1978) and Leroueil et al. (1985). This correlation can be attributed to the common assumption of the existence of a unique cr'-e -e*9 relationship shared by both the EVP model and the isotache concept. It can be further deduced from the EVP model that the spacing among isotaches is governed by the parameter, a . For example, the spacing between isotaches 4-5-6 and 7-8-9 in Figure 2.3b is controlled by the vertical distance between the points 4 and 7 in Figure 2.2a, which is governed by the magnitude of a . Thus, a higher a leads to a series of isotaches with larger spacing, as shown in Figure 2.3c. There is one key difference between the isotache concept and EVP theory. In the EVP model, the distribution of dynamic yield surfaces has a lower limit, the static yield surface (ACE), below which the behaviour of clay is elastic and rate-insensitive. In contrast, the isotache concept assumes that isotaches contract infinitely in stress space with the reduction of strain-rate. 2.2.4 Alternative flow function - the exponential law Several viscoplastic flow functions have been proposed for use in overstress models (e.g. Adachi and Okano, 1974; Perzyna, 1963; Desai and Zhang, 1987; Fodil et al., 1997). Recently, an exponent flow function was used by Rocchi et al. (2003) who reported good agreement with the measured viscous behaviour of Batiscan clay.
25 The basic exponential function is presented in Equation [2.5]. From Equations [2.3] and [2.5], the following relationship between strain-rate and yield stress during isotropic compression can be derived:
,'(*)
acr'^ log(saxial) + l - a l o g O v / , ) - a l o g ( - ) a my
[2.14a]
where a-lln.
As shown in Equation [2.14a], an exponential-law flow function implies
a linear relationship between \og(saxial) and <J'^ . Similarly, the following equations can be obtained for undrained strength 5 ' and preconsolidation pressure, aAd)
:(d)
vp Sriog(*w) + l-a\og(y )-a\of>Q-)
'(*)
[2.14b]
,'(d) a;*' =acr'p(s)log(eaxial) + l-alog(rvp)-a\ogQ-)
a ,(.)
[2.14c]
Figure 2.4 illustrates the difference between a power law flow function and an exponential flow function (see Equation [2.9] and [2.14a], respectively). Using a log-log scale, the solid line representing a power law flow function is linear with a constant slope; whereas the dash line representing an exponential flow function is convex and its tangent slope gradually decreases with increased strain-rate. 2.2.5 Stress-controlled testing Bjerrum (1967) defined secondary compression as delayed compression
(reduction of void ratio) at constant effective stress. Conventional plasticity theories can not account for secondary deformation. Raymond and Wahls (1968) utilized Ca to
characterize creep deformation under constant effective stress as follows: Ae = Ca-Alog(t) [2.15]
26 In this section, a relationship between Ca and a = 1/n is explicitly derived from elastic viscoplastic theory. From Figure 2.5, considering secondary compression over the time interval At = t2-t1, the volumetric strain is: Aevol =Ca /(l + e0)-log(t2 ltx) = Ca /(l + c j - l o g t e I s2) [2.16]
where e0 is the initial void ratio andx and e2 are the volumetric strain-rate at tx and 12 (Note: s1/s2 =t2lt\ frm Equation [2.15]). From EVP theory, the strain-rate at tx and t2 are: t
b
=Yvp{a'w
\ I
\vmy
/(j'(s) )"
u
my-tl)
da'
[2.17a]
and
s
b
=YVP((T'W/a'(s)
/ \umy '
u
)"
my-tl)
JL
dcrl
[2.17b]
where <r'^y)_n and cr'jlt2 are the static yield surface intercepts at tt and t2, respectively. Substituting [2.17a] and [2.17b] into [2.16] gives: A*vo, = Ca 1(1 + e0) n l o g > ( 1 / <r'2) From the hardening law, Equation [2.6]:
X-K
A l n u
[2.18a]
vol =
l + en
my-tl I u
my-tl)
[2.18b]
Combining equations [2.18a] and [2.18b] gives: - = Cj[\n(10)-(A-K)] [2.18c]
Simplifying Equation [2.18c] yields:
27
a=n
= CJ(Cc-Cr)
[2.19]
where Cc and Cr are the compression index and recompression index, respectively (C e = ln(10)A and Cr = ln(10)/c). Equation [2.19] defines the theoretical connection between the rate-sensitivity parameter, a( = l/n), and the secondary compression index, Ca. The parameter, a, plays an important role in this interrelationship. The above discussions in this section describe the context in which the reported viscous behaviour of 19 clays will be evaluated to assess if there is evidence supporting a unique set of viscosity parameters for clays. It should be noted that a relationship similar to Equation [2.19] has been used by other researchers (e.g. Mesri and Choi 1979; Kim and Leroueil 2001) viz.: a=CJCc Considering Cr is typically 10% of Cc for clays (Holtz and Kovacs 1981), the above relationship is practically consistent with Equation [2.19] derived in this chapter.
2.3 Evaluation
This section first examines the influence of strain-rate on the preconsolidation pressure of some natural and remolded clays, followed by undrained strength versus strain-rate, and then the secondary compression behaviour. Table 2.1 summarizes the conventional geotechnical properties of the clay, which originated from Hong Kong, Norway, Northern Ireland, Britain, Sweden, the United States, and Canada.
2.3.1 Rate dependency of preconsolidation pressure The preconsolidation pressure (cr'p) is very important in settlement calculations.
28 For many clays, the preconsolidation pressure and the e-log(<r' v ) curve are ratedependent ( e.g. Leonardo and Ramiah 1959; Crawford 1965; Bjerrum 1967; Vaid et al. 1979; Graham et al.1983; Leroueil et al. 1983; and Leroueil et al. 1985). Figure 2.6 summarizes the usual range of strain-rates used in laboratory tests and the range mobilized in situ. It appears that in situ strain-rates are much lower than those used in laboratory. In addition, long-term field observations by Crawford and Bozozuk (1990) and Kabbaj et al. (1988) have shown that the use of laboratory measured a'p and compression curve without accounting for rate effects may lead to significant underestimation of in situ long-term settlement for viscous clay deposits. The preconsolidation pressure, er'p , versus strain-rate can be obtained from oedometer tests using any of the following test procedures: (a) Constant rate of strain (CRS) oedometer test, where cr'p and s^^ can be measured directly, (b) Conventional incremental oedometer tests (e.g. Drammen and Winnipeg clay) undertaken using different load increment duration. For this case, the strain-rate corresponding to a'p is estimated from the average strain-rates in the loading increment straddling <j'p (Graham et al. 1983). (c) Incremental creep tests (e.g. Batiscan clay). Leroueil et al.(1985) introduced a procedure using a series of incremental oedometer creep tests to construct the CRS compression curves by connecting stress points associated with the same strainrate in Ae - ln(cr'p ) space. Figure 2.7 summarizes the measured <j' plotted against saxial for 12 clays. From Figure 2.7, it appears that the relationship between log(cr'p) and log(f(a..a/) is essentially
29 linear regardless of the dramatically different magnitudes of cr'p ( from 40 to lOOOkPa) and eaxial ( from 10E-8/min to 10E-2/min). As noted above, the effect of strain-rate on (j'p can be represented by a, defined in Equation [2.12]. The parameter, a, can be measured from the best fit line through the test data in log-log scale. It can be clearly seen in Figure 2.7 that Ottawa Leda clay with a =0.104 exhibits greater rate-sensitivity (<y'p) than Winnipeg clay with a =0.03. Table 2.2 summaries the measured values of a from rate-controlled oedometer tests for 12 of the 19 clays studied. The magnitudes of a fall in a range from 0.02 to 0.1. It should be noted that a has been measured over 2 to 3 orders of magnitude strain-rate for most of the clays reported in Figure 2.7. Only 2 clays have been studied over 4 orders of magnitude strain-rate (See Berthierville and Batiscan clays).
2.3.2 Undrained shear strength versus strain-rate As a key parameter in stability analysis, the undrained shear strength (S u ) of many clays is also a function of strain-rate during loading. This observation was
confirmed by numerous studies both in the laboratory and using field vane tests (e.g. Taylor, 1943; Bjerrum 1973; Torstensson 1977; Graham et al. 1983; Kulhawy and Mayne 1990; and Hinchberger 1996). Kulhawy and Mayne (1990) concluded that the average change of Su is about 10% per log cycle change in strain-rate. In this section, the ratedependency of Su for each clay is studied and compared with the rate-dependence of the preconsolidation pressure for the corresponding clay. This is done to investigate the possible correlation between a from log(Su)-log(s ) relationship and a from
30 log(cr'p ) - log(s ) relationship. For 13 clays in Table 2.2, the Su and associated saxial were obtained either from undrained triaxial compression tests at different constant strain-rates, undrained tests with step-changed strain-rates (e.g. Winnipeg clay, London clay, and Belfast clay), or field vane tests with different rotation-rates (e.g. Backebol clay). Figure 2.8 plots the normalized undrained shear strength, SuN, (normalized by the strength, Su, at eaxial = 1.0 min -1 ) versus strain-rate in a log-log scale for 11 clays. For most clays in Table 2.2, the relationship between l o g ^ ^ ) and l o g ^ ^ , ) is apparently linear (See Figure 2.8). Two exceptions are Haney clay and Drammen clay, as shown in Figure 2.9. For Haney clay, when the strain-rate reduces to a threshold value of
2xlO~ 5 mnT 1 , the undrained strength becomes constant and unaffected by further reduction in the strain-rate. Thus, the undrained strength of Haney clay becomes rateindependent at strain-rates less than 2 xlO -5 min"1. Similar phenomenon can be seen for Drammen clay, which has a threshold strain-rate of 5xl0~ 6 min"1 (Berre and Bjerrum, 1973). Even so, it is noted that when s > 2x10 5 min"1 for Haney clay and is essentially
E >5xl0~ 6 min -1 for Drammen clay, the relation of l o g ^ ^ ) -\og(s) linear as seen for the other clays.
Therefore, a linear l o g ^ ^ ) - log(^) relationship can be obtained for each of 11 clays shown in Figure 2.8 for the range of strain-rates investigated (up to 4 orders of magnitude). For Haney and Drammen clays, the linear relationship has a termination point, corresponding to a threshold strain-rate below which Su is rate-insensitive. This phenomenon may also occur in other natural clays at very low strain-rate.
31 Interrelationship between a.uc and a_Qti As summarized in Table 2.2 and Figure 2.10, the rate-sensitivity parameter, a, measured from rate-controlled undrained tests ( a . u c ) and oedometer tests (a. o e d ) appears to be unique even though the stress paths in these two tests are different. The best
agreement has been found for Winnipeg clay where the difference between a_uc and a_oed is only 0.001; whereas St. Jean Vianney clay has the worst agreement between a_uc and . oed , where the discrepancy is 0.007: This may be attributed to the natural variability of clay samples and experimental error due to the triaxial equipment used ( See Vaid et al. 1979 and Roberson 1975). As also shown in Figure 2.10, the hollow circles representing the interrelationship between a.uc and a.oed fall within the range of the bounded lines that deviate +0.005 from the 1:1 line for the clays except Belfast clay (-0.006) and SJV clay (-0.007). In general, the rate-sensitivity parameter, a , measured from rate-
controlled undrained and rate-controlled oedometer tests appear to be consistent. However, more data over a wider range of the a values would be required for a more definite conclusion to be drawn. Exponential Flow Function Figure 2.11(a) shows a comparison of the relationship between <j'p and saxial derived from the power law (Equation [2.13]) and exponential law (Equation [2.14]) flow functions, together with data from select clays (e.g. Batiscan, Winnipeg, Gloucester, and Drammen clay). For Batiscan clay, Winnipeg clay, Drammen, and Gloucester clay, both power law and exponential laws seem to fit well with the laboratory data over the range of strain-rate measured. The difference between these two flow functions is negligible
32 compared with the data. However, Figure 2.11(b) shows the results from laboratory tests and field observations by Leroueil et al. (1983) for the Gloucester case. The data in Figure 2.11(b) does not agree well with the exponential flow rule and the data suggest a linear to slightly concave up relationship. From this comparison, it appears that a power law is more representative of data over a large range of strain-rates than the exponential law.
2.3.3 Secondary compression As previously shown, the rate-sensitivity in oedometer compression tests and in undrained triaxial compression tests can be quantified by the magnitude of a . Furthermore, it is not clear if the parameter a can be used to characterize other viscous behaviour, such as the time-dependent creep deformation that occurs in secondary compression. Table 2.2 summarizes a obtained using Equation [2.19], and Cr, Cc, and Ca for 11 clays reported in the literature. In some cases, Cr was not reported (e.g. for Winnipeg clay, Belfast clay and Ska Edeby clay). For these cases Cr was deduced from the typical relationship of C r = C c /10 from Holtz and Kovacs (1981). For all other clays, however, Cr, Cc, and Ca were obtained directly from published experimental data, using higher incremental loading stresses to minimize the possible influence of clay structure and destructuration during secondary compression (Leroueil and Vaughan 1990; Burland 1990). Appendix C summarized the determination of these parameters. Referring to Table 2.2 and Figure 2.12, it can be seen that acreep obtained from Cr, Cc, Ca, and Equation [2.19] is in a good agreement with a_otAand a_uc . The best
33
match was found for Sackville clay, where the difference is only 0.0001; whereas the worst match was for San Francisco clay with a discrepancy of 0.012. It should be noted that the samples used to measure a and a.uc were obtained respectively by
independent researchers (Arulanandan et al. 1971; Lacerda 1976) from different locations and different depths. Again, considering the natural variability of clay and the potential influence of clay structure during compression, the general match among acreep, a_otd and or.uc is very encouraging.
2.3.4 Summary In the preceding sections, the viscous behaviour in constant-stress tests and ratecontrolled tests was evaluated, using experimental data for 19 clays from Europe to North America (see Table 2.1 for the references). The key observations are summarized below: All of the clays summarized in this study exhibit an essentially linear relationship between sa and Su , or between sa and ayp for strain-rates over 2 to 4 orders of magnitude. This observation is consistent with Equations [2.11] and [2.12], which were derived from EVP theory. For Haney clay and Drammen clay, the linear relationship has a termination point at a threshold strain-rate, below which the undrained strength is rateindependent. This is consistent with the concept of a static yield surface. The strain-rate parameter, a, is unique for each of the 14 clays in Figure 2.13 regardless the type of experiment used to measure it. As shown in Table 2.2, the value of
a
creep obtained from secondary compression using Equation [2.19] agrees well with a.oed measured from rate-controlled tests (oedometer and undrained tests, and a.oed with the corresponding avg
and aac
respectively). Figure 2.13 compares auc, a
34
for 14 clays. From this figure, it can be seen that most symbols fall in a narrow range of 0.005 from the 1:1 line. This is encouraging from an engineering point of view and it suggests the viscous behaviour in rate-controlled tests (rate-dependency of yield stress) and stress-controlled test (time-dependency of secondary deformation) have an inherent correlation through the parameter, a . Appendixes D and E discusse other factors that may have an impact on a, such as temperature, plasticity index, liquidity index, and destructuration.
2.4 Selection of Parameters

An EVP model can be developed by coupling viscosity with conventional elastoplastic theory. Thus, the parameters required in such a model can be divided into two groups: elastoplastic and viscosity-related. If critical state concepts are adopted, the elastoplastic parameters include e , v , A (= Cc /ln(10)), K (= Cr /ln(10)), M , and a suitable yield surface function Ffo'^ ,eôl), where e is void ratio, v is Poisson's ratio, A and K can be estimated from oedometer compression tests, M can be obtained from the constant volume effective friction angle, and svvpol is the plastic volumetric strain. The parameters, A and K , can be determined from oedometer compression test or isotropic compression test, and M is the slope of the critical state line. Methods to determine these elastoplastic parameters can be found in the literature (e.g. Roscoe and Burland 1968; Chen and Mizuno 1990; and Atkinson 1993). In addition, the yield surface can be estimated from the stress path in undrained triaxial compression tests or determined more precisely from a series of stress-path probing tests (e.g. Tavenas et al. 1979, Leroueil et al. 1979; DeNatale 1983). To avoid repetition, only measurement of the three viscosity-
35 related parameters, a, o*^, and yvp, are discussed below.
2.4.1 The measurement of a The parameter, a , can be determined from either multiple CRS drained isotropic compression tests, multiple CRS drained oedometer compression tests, or multiple CRS undrained triaxial compression tests undertaken using various strain-rates. To minimize the influence of natural variation, compression tests can be performed on one specimen with step-changed strain-rates, from which the stress-strain curves at various constant strain-rates can be interpolated (e.g. Richard and Whiteman 1963; Graham et al. 1983). Examples of the step-changed strain-rate approach are shown in Figure 2.14. The
measured yield stresses can be plotted against the corresponding strain-rates in a log-log scale, as illustrated previously in Figure 2.2abc. The slope of the regression line passing through experimental data gives the magnitude of a. If CRS drained oedometer or isotropic compression tests are adopted, the strainrate should be low enough to avoid generating significant excess pore water pressure, since the pore water pressures vary within a specimen causing difficulties in determination of the applied effective stress. The study on Batiscan clay (Leroueil et al. 1985) indicated that excess pore pressure was undetectable for strain-rates lower than 3xl0~ 6 /min in CRS oedometer tests on specimens 19mm high and 50.8mm in diameter. In addition, a can be obtained using Equation [2.19], where Ca , Cc, and Cr can be measured from an oedometer consolidation test or K'0 triaxial test. This approach provides an additional way to check a measured from CRS compression tests and it permits evaluation of the consistency of different experiments. It is noted that this
36 approach should be carefully used for structured natural clays, considering the influence of destructuration on the measured values of Ca and Cc during the compression loading. Ideally, Ca should be measured for load increments in the intrinsic state as shown in Figure 2.15. 2.4.2 The measurement of <r'g a n d x p cr'j^ and yvp can be determined from the linear log-log relationship obtained during the measurement of a (see Figure 2.2). The first step is to examine the log-log plot, and if the clay has a threshold strain-rate below which rate effects stop, Option A can be used. If such a termination point does not exist, Option B can be adopted. Option A and B are discussed below. Option A Referring to Figure 2.9, the undrained shear strength of Haney clay is ratesensitive for strain-rates in excess of 2xl0~ 5 /min . This signifies that Point E on the static yield surface has been reached at eaxial = 2xl0" 5 /min (see Figures 2.2c and 2.3a). Correspondingly, as shown in Figure 2.9, the normalized static strength S^ is 0.65,
which gives S(us) = 268kPa (see Vaid and Campanella,1979). As a result, the static yield surface intercept, <r'j^, is: ^ A S W [2.20]
where A is a constant which can be derived from the yield surface function (see Point E and A in Figure 2.3b). For Haney clay, cr'^ =515kPa since A s 0.5 (see Vaid and Campanella, 1977), and the fluidity parameter is yvp = V3?2x 2xl0" 5 /min
37 2xlO~ 5 min _1 (see Figure 2.2c). This approach can be applied to obtain the
parameters, cr'j^ and yvp, for Drammen clay (see Figure 2.9) and any other clay which exhibit a threshold strain-rate in their log(iS'K ) versus log(aj.ia/) relationships. Alternatively, cr'^ a n d ^ can be obtained from the log(<r'p) versus log(f r a a / )
relationship. For example, Figure 2.16 shows the variation of <r' with strain-rates for Berthierville clay at a depth of 3.9-4.8m (Leroueil et al. 1988). The termination point corresponding to static state can be reached at eCDdal = 9xl0~ 7 /min. As a result, the fluidity parameter is yvp = V5/"3x 9xl(T 7 /min lxl0~ 6 /min (see Figure 2.2b). Correspondingly, in Figure 2.16, the normalized static preconsolidation pressure is 0.92, which gives cr'^)=79.8kPa (see Leroueil et al. 1988). The resultant static yield surface intercept, cr'JJ, is: <
) = A
p X <
[2-21]
where Ap is a constant which can be derived from the yield function and K'0 (see Point A and C in Figure 2.3b). OptionB For cases where laboratory testing has not been done at strain-rates slow enough to identify a termination point (e.g. Points C or E in Figures 2.2b and 2.2c), the fluidity parameter, yvp, can be estimated from engineering cases. Figure 2.17 shows the log(cr' p ) versus l o g ( ^ a , ) relationship obtained from laboratory and in situ observations for St. Alban clay (3.1-4.9m). An additional example is presented in Figure 2.11b for Gloucester clay. For these two cases (Leroueil et al.
38 1983 and 1988), the termination point has not been reached for strain-rates as low as 1(T8 min -1 . Consequently, it appears that the strain-rate, 10~8 min -1 , can be taken as the upper bound of the fluidity parameter, yvp. Figure 2.7 shows that for strain-rates lower than 10~10 muT 1 , the completion of 1% strain requires more than 190 years, which is out of practical interest for engineers. Thus, for St. Alban clay and Gloucester clay, the fluidity parameter, yvp , can be taken from the range between the upper bound (10~8 min" 1 ) and lower bound (1(T10 min -1 ). Next, according to the assumed yvp, the parameter a'^ estimated from the log( a'^) or S^s) can be easily
versus log( saxial) relationship or the log( S^s)) versus
lg(axiai) relationship (see Figures 2.2b and 2.2c). Consequently, the resultant <r'^ can be obtained through Equations [2.21] or [2.20]. However, the value of cr'^ deduced from these upper and lower bounds yvp should not result in yield stresses below the initial in situ stress state of the clay. Using Option B, the EVP constitutive model would be conservative and capable of accounting for rate-effects for the service life of most engineering projects (e.g. for 50 years if the strains do not exceed 5% ). In addition, if the actual yvp is higher than the assumed value, the prediction on long-term settlement and stability by the EVP model will be on the conservative side.
2.5 Summary and Conclusion

This chapter has evaluated the viscous response of the clays reported in the
39 literature with the intent to examine if the viscous response of clay during drained and undrained CRS laboratory tests and during drained constant stress tests can be described within a general elastic viscoplastic theory. Both power law and exponential flow laws have been used in conjunction with EVP theory to explicitly derive equations relating Su and a'p versus strain-rate. The foliowings summarize the main findings: 1) A linear log -log relationship between strain-rate and preconsolidation pressure (cr'p) or undrained shear strength (Su) can be obtained from the rate-sensitivity response of all 19 clays for the ranges of strain-rate studied. This is consistent with the theoretical equations derived from the power law EVP model. 2) This study shows strong evidence suggesting that the rate-sensitivity parameter, a , measured from three different types of experiments (CRS undrained triaxial compression, CRS oedometer compression, and oedometer creep tests) is consistent for the clays studied. This consistency indicates that the viscous responses of clay are inherently related and consequently it is possible to account for these apparently different viscous behaviours using a single phenomenological constitutive theory. 3) In the EVP framework, a power law flow function appears more appropriate than an exponential law flow function, especially for the cases involving a large range of strain-rates and it is consistent with Ca from secondary compression theory. 4) The viscous response of the clays presented in this chapter can be fully interpreted using the EVP framework with a power law flow function, provided a suitable yield function, ^(cr^,^.) with an appropriate aspect ratio, A , is chosen. The rate-
sensitivity of & and 5 and the time-dependency during secondary compression can be
40
described using a single set of viscosity parameters, a, cr'^, and^vp. Particularly, the minimum undrained strengths observed from the behaviour of Haney, Drammen, and Berthierville (3.8-4.8m) clays appears to confirm the concept of the static yield surface in the EVP model. 5) A link between the isotache concept and the EVP framework has been demonstrated. Both of these two theories have been developed based on a unique <j' -svp -e relation. 6) Novel and straightforward guidance has been provided to select and measure all three viscosity-related parameters.
41
References
Adachi, T., and Okano, M. 1974. A constitutive equation for normally consolidated clay. In Soils and Foundations, pp. 55-73. Adachi, T., and Oka, F. 1982. Constitutive equations for normally consolidated clay based on elasto-viscoplasticity. Soils and Foundations, 22(4): 57-70. Arulanandan, K., Shen, C.K., and Young, R.B. 1971. Undrained creep behaviour of a coastal organic silty clay. Geotechnique, 21(4): 359-375. Atkinson, J.H. 1993. An introduction to the mechanics of soils and foundations : through critical state soil mechanics. McGraw-Hill Book Co., New York. Bjerrum, L. 1967. Engineering geology of Norwegian normally-consolidated marine clays as related to settlements of buildings. Geotechnique, 17(2): 81-118. Bjerrum, L. 1973. Problems of soil mechanics and construction on soft clays. State of the art report Session IV.. In Pore. 8th Int. Conf. Soil Mech. Mescow, Vol.3, pp. 111159. Burland, J.B. 1990. On the compressibility and shear strength of natural clays. Geotechnique, 40(3): 329-378. Chen, W.-F., and Mizuno, E. 1990. Nonlinear analysis in soil mechanics : theory and implementation. Elsevier Science Publishing Company Inc., New York, NY, U.S.A. Crawford, C.B. 1965. Resistance of soil structure to consolidation. Canadian Geotechnical Journal, 2(2): 90-115. Crawford, C.B., and Bozozuk, M. 1990. Thirty years of secondary consolidation in sensitive marine clay. Canadian Geotechnical Journal, 27(3): 315-319.
42
DeNatale, J.S. 1983. On the calibration of constitutive models by multivariate optimization.. Ph.D Thesis, University of California. Desai, C.S., and Zhang, D. 1987. Viscoplastic model for geologic materials with generalized flow rule. In International Journal for Numerical and Analytical Methods in Geomechanics, pp. 603-620. Fodil, A., Aloulou, W., and Hicher, P.Y. 1997. Viscoplastic behaviour of soft clay. Geotechnique, 47(3): 581-591. Gasparre, A., Nishimura, S., Coop, M.R., and Jardine, R.J. 2007. The influence of structure on the behaviour of London Clay. Geotechnique, 57(1): 19-31. Graham, J., Crooks, J.H.A., and Bell, A.L. 1983. Time effects on the stress-strain behaviour of natural soft clays. Geotechnique, 33(3): 327-340. Hinchberger, S.D. 1996. The behaviour of reinforced and unreinforced embankments on rate senstive clayey foundations. Ph.D Thesis, University of Western Ontario, London. Hinchberger, S.D., and Rowe, R.K. 1998. Modelling the rate-sensitive characteristics of the Gloucester foundation soil. Canadian Geotechnical Journal, 35(5): 769-789. Holtz, R.D., and Kovacs, W.D. 1981. An introduction to geotechnical engineeering. Prentice Hall, Inc., Toronto. Kabbaj, M., Tavenas, F., and Leroueil, S. 1988. In situ and laboratory stress-strain relationships. Geotechnique, 38(1): 83-100. Kaliakin, V.N., and Dafalias, Y.F. 1990. Verification of the elastoplastic-viscoplastic bounding surface model for cohesive soils. Soils and Foundations, 30(3): 25-36.
43
Katona, M.G. 1984. Evaluation of Viscoplastic Cap Model. Journal of Geotechnical Engineering, 110(8): 1106-1125. Kavazanjian, E., and Mitchell, J. 1980. Time-dependent deformation behaviour of clays. Journal of Geotechnical Engineering, 106(6): 611-630. Kim, Y.T., and Leroueil 2001. Modeling the viscoplastic behaviour of clays during consolidation: Application to Berthierville clay in both laboratory and field conditions. Canadian Geotechnical Journal, 38(3): 484-497. Kulhawy, F.H., and Mayne, P.W. 1990. Manual on estimating soil properties for foundation design. Electric Power Research Institute, Palo Alto, Calif. Lacerda, W.A. 1976. Stress-relaxation and creep effects on soil deformation. Ph.D., University of California, Berkeley, United States California. Law, K.T. 1974. Analysis of Embankments on Sensitive Clays, University of Western Ontario, London, Ontario. Lehane, B.M., Jardine, R.J., Bond, A.J., and Frank, R. 1993. Mechanisms of shaft friction in sand from instrumented pile tests. Journal of Geotechnical Engineering, 119(1): 19-35. Leonardo and Ramiah 1959. Time effects in consolidation of clays ASTM special technical publication No.254 Leroueil, S., and Vaughan, P.R. 1990. The general and congruent effects of structure in natural soils and weak rocks. Geotechnique, 40(3): 467-488. Leroueil, S., Samson, L., and Bozozuk, M. 1983. Laboratory and field determination of preconsolidation pressures at Gloucester. Canadian Geotechnical Journal, 20(3): 477-490.
Leroueil, S., Kabbaj, M., and Tavenas, F. 1988. Study of the validity of a a - sv - v rate model in in istu conditions. Soils and Foundations, 28(3): 13-25. Leroueil, S., Kabbaj, M., Tavenas, F., and Bouchard, R. 1985. Stress-strain-strain-rate relation for the compressibility of sensitive natural clays. Geotechnique, 35(2): 159-180. Lo, K.Y. 1961. Secondary compression of clays. Journal of Geotechnical and Geoenvironmental Engineering, 87(4): 61-87. Lo, K.Y., Bozozuk, M., and Law, K.T. 1976. Settlement analysis of the gloucester test fill. Canadian Geotechnical Journal, 13(4): 339-354. Mesri, G., and Choi, Y.K. 1979. Discussion on 'Strain-rate behaviour of St. Jean Vianney clay'. Canadian Geotechnical Journal, 4: 831-834. Mesri, G., Feng, T.W., and Shahien, M. 1995. Compressibility parameters during primary consolidation. In Proc. Int. Symp. on Compression and Consolidation of Cayey Soils. Hiroshima, Jpn, Vol.2, pp. 1021-1037. Mesri, G., and Castro, A. 1987. C a / Cc concept and k'o during secondary compression. Journal of geotechnical engineering, 113(3): 230-247. Oldecop, L.A., and Alonso, E.E. 2001. A model for rockfill compressibility. Geotechnique, 51(2): 127-139. Oldecop, L.A., and Alonso, E.E. 2007. Theoretical investigation of the time-dependent behaviour of rockfill. Geotechnique, 57(3): 289-301. Perzyna, P. 1963. Constitutive equations for rate sensitive plastic materials. Quarterly of Applied Mathematics, 20(4): 321-332.
Philibert, A. (1976). "Etude de la resistance au cisaillement d'une argile Champlain. , M.Sc. Thesis, Universite de Sherbrooke, Quebec. Raymond, G.P., and Wahls, H.E. 1976. Special Report: Estimating 1-dimensional consolidation, including secondary compression, of clay loaded from overconsolidated to normally consolidated state, National Research Council, Transportation Research Board. Richardson, A.M., and Whitman, R.V. 1963. Effect of strain-rate upon undrained shear resistance of saturated remoulded fat clay. Geotechnique, 13(4): 310-324. Robertson, P.K. 1975. Strain-rate behaviour of Saint-Jean-Vianney clay. Ph.D Thesis, University of British Columbia, British Columbia,Canada. Rocchi, G., Fontana, M., and Da Prat, M. 2003. Modelling of natural soft clay destruction processes using viscoplasticity theory. Geotechnique, 53(8): 729-745. Roscoe, K.H., and Burland, J.B. 1968. On the generalized stress-strain behaviour of 'wet' clay. In Engineering Plasticity, pp. 535-609. Rowe, R.K., and Hinchberger, S.D. 1998. The significance of rate effects in modelling the Sackville test embankment. Canadian Geotechnical Journal, 35(3): 500-516. Sallfors, G. 1975. Preconsolidation pressure of soft high plastic clays, Chalmers University of Technology, Gothenburg, Sweden. Sheahan, T.C. 1995. Interpretation of undrained creep tests in terms of effective stresses. Canadian Geotechnical Journal, 32(2): 373-379. Sheahan, T.C, Ladd, C.C., and Germaine, J.T. 1996. Rate-dependent undrained shear behavior of saturated clay. Journal of Geotechnical Engineering, 122(2): 99-108.
Sorensen, K.K., Baudet, B.A., and Simpson, B. 2007. Influence of structure on the timedependent behaviour of a stiff sedimentary clay. Geotechnique, 57(1): 113-124. Soga, K., and Mitchell, J. K. (1996). "Rate dependent deformation of structured natural clays." Measuring and modeling time dependent soil behaviour, Geotechnical special publication No. 61, ASCE, Washington D.C., 243-257. Suklje, L. 1957. The analysis of the consolidation process by the isotache method. In Proc. 4th Int. Conf. on Soil Mech. and Foun. Engen. London, Vol.1. Tavenas, F., and Leroueil, S. 1978. Effects of stresses and time on yielding of clays. In Proc of the Int Conf on Soil Mech and Found Eng, 9th, Jul 11-15 1977. Edited by P.C.O.X. ICSMFE. Tokyo, Jpn. Jpn Soc of Soil Mech and Found Eng, Tokyo, pp. 319-326. Tavenas, F., Leroueil, S., La Rochelle, P., and Roy, M. 1978. Creep behaviour of an undisturbed lightly overconsolidated clay. Canadian Geotechnical Journal, 15(3): 402-423. Taylor, D.W. 1948. Fundamentals of soil mechanics. John Wiley, New York. Torstensson, B.A. 1977. Time-dependent effects in the field vane test. In Int. Symp. on Soft Clays. Bangkok, pp. 387-397. Vaid, Y.P., and Campanella, R.G. 1977. Time-dependent behavior of undisturbed clay. Journal of the Geotechnical Engineering Division, 103(7): 693-709. Vaid, Y.P., Robertson, P.K., and Campanella, R.G. 1979. Strain-rate behaviour of SaintJean-Vianney clay. Canadian Geotechnical Journal, 16(1): 35-42. Wiesel, C.E. 1973. Some factors influencing in situ vane test results. In Proc. 8th ICSMFE. Moscow, Vol. 1.2, pp. 475-479.
47
Yin, J.-H., Zhu, J.-G., and Graham, J. 2002. A new elastic viscoplastic model for timedependent behaviour of normally and overconsolidated clays: Theory and verification. Canadian Geotechnical Journal, 39(1): 157-173. Zhu, G., and Yin, J.-H. 2000. Elastic visco-plastic consolidation modelling of clay foundation at Berthierville test embankment. International Journal for Numerical and Analytical Methods in Geomechanics, 24(5): 491-508.
Table 2.1
References
Water Content (%) Liquid Limit (%) LI (%) PI (%)
Geotechnical properties of 19 clays

St
Clay Name
c c c
Cc =0.386, C r =0.184, Ca =0.036
1 23 39.9 60 60 79.6 90 43 50 44 57.4 51 42 60 62 36 0.92 0.65 1.38 52 1.27 2.74 2.74 77 0.62 45 30 21 23 18 31.5 31 16 45.4 0.77 23.7 N/A 3 N/A 125 14 8 N/A N/A 100 60 40 0.08 Taylor 1948 Graham et al.1983 Law, 1974; Lo et al,1976; Leroueil et. al,1983 Leroueil et al. 1985;
Reconsituted London clay, Lo 1961;Sorensen et al.2007; England Gasparre et al.2007 N/A
Remolded Boston blue clay, United States
Winnipeg, Central Canada
CJCC =0.018, Cr = CJ10 Cc =1.495, Cr =0.058, Ca =0.061 Cc=0.24, C r = C c /10 CJCC =0.03
Gloucester clay, Canada
Batiscan clay, Canada
St. Alban clay, Canada
7 Yin et al. 2002 Bjerrum 1967; Berre and Bjerrum, 1973; Vaid et al,1979
Haney clay, Canada
Tavenas and Leroueil,1978; Graham et al,1983; Vaid and Campanella,1977
Hong Kong marine clay, Hong Kong
Cc =0.0793, Cr =0.018, Ca =0.0025 Cc=0.45, Cr =0.055, Ca =0.016
Drammen clay, Norway
10
St. Jean Vianney clay, Canada
-1^
00
Table 2.1
References
(%) (%) (%) (%)
Geotechnical properties of 19 clays (Cont.)

Water Liquid content Limit LI St 7.5 N/A PI 70 61 47 1.74 19 90 0.67 60
Clay Name Graham et al,1983 Hinchberger 1996;Row and Hinchberger 1998. Wiesel,1973;Mesri et al. 1995 102 62 84.8 76.5 100 55 46 70 70 93 31 99 1.05 1.67 1.34 1.15 1.15 4 65 24 43 43 48 8 22 28 ~7 >50 25
c c c
CJCC =0.05, Cr=CJ10 Cr =0.07;, Cc =0.646, Ca =0.0312 CJCC =0.05, Cr=CJW
11
Belfast clay, N.Ireland
12
Sackville clay, Canada
13
Ska Edeby clay, Sweden
14
Backebol clay (7m), Sweden Sallfors 1975; Torstensson 1977;Leroueil et al 1985. Leroueil et al. 1988; Zhu and Yin 2000; Kim and Leroueil,2001; Leroueil et al. 1985 Leroueil et al. 1985 Arulanandan et al. 1971; Lacerda 1976; Kavazanjian and Mitchell 1980 Crawford, 1965
15
Berthierville clay, Canada
Cr =0.027, Cc =0.497, Ca =0.027
16
St Cesaire clay, Canada
17
Louiseville clay, Canada
18
San Francisco Bay Mud. United States
Cr=0.1, Cc =0.75, Ca =0.05
19
Leda clay, Canada
-1^
Table 2.2
"-
-oed
CI creep
Summarized a for 19 clays
Clay Name
References
from triaxial compression (13 clays) From Eq.[2.19] (11 clays) 0.018 0.023 0.024 0.031 0.037 0.030 0.035 0.047 0.041 0.041 0.044 0.051 0.045 0.046 0.048 0.052 0.052 0.056 0.041 0.041 0.020 0.043 0.033
from oedometer compression (12 clays)
(the average value) 0.021 0.024 0.027 0.038 0.040 0.041 0.041 0.043 0.047 0.049 0.051
1 Taylor 1948;
Lo 1961;Sorensen et al.2007; Gasparre et al.2007
Reconsituted London clay, England Remolded Boston blue clay , United States Winnipeg, Central Canada Gloucester clay, Canada Batiscan clay, Canada St. Alban clay, Canada Haney clay, Canada Graham et al.1983 Law,1974; Lo et al,1976; Leroueil et. al,1983 Leroueil et al. 1985; Mesri et al. 1995 Tavenas and Leroueil, 1978; Graham et al,1983 Vaid and Campanella,1977
Hong Kong marine clay
9 Vaid et al.1979 Graham et al,1983
Drammen clay, Norway
Yin et al. 2002 Bjerrum 1967; Berre and Bjerrum, 1973
10
St. Jean Vianney clay, Canada
11
Belfast clay, N.Ireland
Table 2.2
-uc -oed
creep
Summarized a for 19 clays (Cont.)
Clay Name References
from triaxial compression (13 clays) From Eq.[2.19] (11 clays) 0.053 0.056 0.058 0.056 0.067 0.069 0.065 0.104 0.077 0.057 0.053 0.054 0.053
from oedometer compression (12 clays )
(the average value) 0.053 0.055 0.056 0.057 0.067 0.069 0.071 0.104
12 Wiesel,1973;Mesri et al. 1995
Sackville clay, Canada Hinchberger 1996; Row and Hinchberger 1998.
13
Ska Edeby clay, Sweden
14
Backebol clay (7m), Sweden
15 Leroueil et al. 1985 Leroueil et al. 1985 Arulanandan et al. 1971; Lacerda 1976; Kavazanjian and Mitchell 1980 Crawford, 1965
Berthierville clay, Canada
Sallfors 1975; Torstensson 1977;Leroueil et al 1985. Leroueil et al. 1988; Zhu and Yin 2000; Kim and Leroueil,2001;
16
St Cesaire clay, Canada
17
18
Louiseville clay, Canada San Francisco Bay Mud. United States
19
Leda clay, Canada
52 Figure 2.1 Illustration of [G24]models for elastic viscoplastic materials
b)
c'P âxial
-VP I
a
axial
, a
v
>
c)
d)
Critical state line
Figure 2.2
Illustration of relations between strain-rate and yield stress (or shear strength) in strain-rate controlled tests
(a)
- Derived from the EVP mbdel with a power law flow function
. '.. 0 , ( L - ' - A
_ . rpy _ -r
j .
t j
i
\
: ~Ta i
y <
i
1
-, '! . -
...^
' 1 1
1-
Axial Strain Rate, min"1, in log scale
(b)
Derived from the EVP model with a power taw flow function
> -
-, ,
Sf 'eyS
or
1 / *
/-X*"
'7\JfvZ-'Qyt
_~"~~*' ~-y 1 ;'
\
t
i
~ \ i 1
*1
1-
Axial Strain Rate, min '\ in log scale
(c)
Il'l.
u mvi
diffpnjlheif =V^ m utfel with'ra p Swef: IS iv tlovMtinSJj 5h: .iillJRLI! i
:ii
i.i
....
'
i.i
-:-f
:i
J f-i4-
"*p^
r-t-f
: ..1
,,L|J ill
- I ' l l
i iii i
.,i..L
-r I r !
S~7 5"TIT <
.......... 1 )
* >
i ;H
ii
-: 11
-
._
-2-1 < H-- j

<
ur4-4 H J_j
i | fill
xittfi
J_L,
;v
(7 ill
Lrv*f 3 !:
I -
i..f...
r~i|:|--
nt
i :lj
11
Axial Strain Rate, min 1 , in log scale
Figure 2.3
The link between the EVP model and the isotache concept
(a) Yield surface family for the EVP model
(b) The correlation between dynamic yield surfaces and isotaches on stress space.
Critical state line
Isotach contour
a (s) 4 7 B
my
a
m
Figure 2.3
The link between the EVP model and the isotache concept (Cont.)
(c) The relationship between the magnitude of a and the spacing for isotaches
y[2J2
Isotaches
high
-104 v W i
Figure 2.4
The influence of the power law and exponent law flow functions on the relationship between yield stress and strain-rate
CO
Q.
j
CO CO
b
CD
i_
3 CO CO Q.
^
CO
>^
O Q.
CO
Strain Rate, mirf1, in log scale
57 Figure 2.5 Typical compression curve for secondary compression.
Ae
EOP: End of excess pore pressure dissipation
log(time)
58
Figure 2.6
Ranges of strain-rates in laboratory tests and in situ (modified from Leroueil and Marques, 1996)
' \ f
strain rate range strain rate range in Varje shear test ,1 In abpratory test -A, -j- i ' ' "S . I. ... . . . . j ^ . -1 f i 1- -t Berttatefvite-Ol 'jpRS)oeBometr ^taiirrate range" SniFifeldtohaVieur
l1
|M&?4 -J|..M t pictocpster

" H "^ " *!r - 1
j i |T
5b3
Jriftrsined shear test
Time required for"!*?* strpfn
<
Years years
1
MsJwtris i,daysi .;
i... , i
$e$ofids i 1 Mins > .
foMfl^strair. -(o-12 10 11
r $
y ^ 10"9 10"8
year 10"7
l.'tto^h 10"6 10"s
10-13
10-'
1CH
10"3 10"2 10"1 10 Strain rate, / m i n
10-9
10"8
10"7
10"6
10 s
10"4
10"3
10-2
10-1
10
101
102
103
Equivalent strain rate with unit of %/hr
Figure 2.7
Relationship between preconsolidation pressure, a'p, and strain-rate, m i a i , i n log-log scale
CD Q.
CO CO
c o
TO
"o (/)
C
8
CD
c
CO
co
<
Q. Q.
10 2
10"1
Strain rate, /min

A O n T O + V
X
Batiscan clay Backebol clay Louiseville clay St Cesaire clay Gloucester Clay Belfast clay Winnipeg clay Drammen clay St.Jean Vianney clay Ottawa Leda clay Berthierville clay St. Alban clay
60 Figure 2.8 Relationship between undrained strength, Su, and axial strain-rate, saxial, in log-log scale
a=0. 0 3 7 . ^ - * a=0.045
.9
a=0.023 I
.8 - a=0.0244\^ = = 5 j f = ==3S== .7
o|=0.031' Jf
a =0.0441 -
I
n
w
6
.5
a=0.046 a a=0.065\ 0
a=0.053/
^-^^3^ <*^^"^
D V A Backebol clay Belfast clay Winnipeg clay Ottawa Leda clay Remolded Boston blue clay Gloucester Clay St.Jean Vianney clay Sackfill clay Hong Kong Marine clay San Francisco Bay Mud London clay
1
CO CO II z
co3
|
.3
e
A
'
10-' 10"
!
10"'
10-
10":
10-'
10
Strain rate, /min
61 Figure 2.9 Relation between undrained strength and axial strain-rate for Drammen clay and Haney clay
0 o 0.9
Drammen clay Haney clay
s
1
1
!E
T
o.8
i I oy rr
E
II
ate
1
0.7
a=0.046| .
^ C O 3
CO
II 2
1 I
^T
f9dX JZf
i<5/^
1
i
i !
i
T i X
y |
0.6
co-
\ !
'
1 I
\
!
I
10
!
1
1
5
2El-5/min
0.5
1 0 -7
1Q
1 4E-d/min 1 i -6 ' ' -5
10-4
1 0 -3
1Q -2
10-1
10
Strain rate, /min
62 Figure 2.10 Comparison of a estimated from rate-controlled oedometer tests and undrained triaxial tests ( See Table 2.2).
oed
VS
-acreep
CO CO CD CD
c
CO
k-
T3 C T3
CD
o o cb *-*
CO
k_
E
p
0.00
0.00
a from rate-controlled oedometer tests, a
63 Figure 2.11 Evaluation on the ability of exponential and power law flow functions to represent the relationship between preconsolidation pressure and strainrate
(a) Experiment results from Batiscan clay, Backebol clay, Gloucester clay, and Drammen clay
400
a.
300
Batiscan clay
Backebol clay Gloucester Clay

Drammen clay Regression line using exponential flow function Regression line using power law flow function
E c o
'*-
200
;u "o
c o o 100 c
CO
2 cc
<
Q. Q.
-10-3
10-2
-10-1
Strain rate, /min
Figure 2.11
Evaluation on the ability of exponential and power flow functions to represent the relationship between preconsolidation pressure and strainrate (Cont.)
(b) Results measured from lab and in situ for Gloucester clay (data from Leroueil et al. 1983)
200
CO
a.
CO 3
(A V)
,o
TO
CO
o c o o c
CO
CD
<
C O O. CL
10-2
Strain rate, /min
65
Figure 2.12
Comparison of a estimated from secondary consolidation tests, ratecontrolled oedometer tests, and undrained triaxial tests (See Table 2.2).
O aoedv.s.au
of v s a
a> o
*-
(A
o o o
o
a.
CD
(0
E o T3 a; o
o
4-*
CO
c o o
a> o
0 C
o o
</i
tn
CO
a) E p 0.00
a from rate-controlled undrained tests, a
uc
66
2.13 Comparisons of a.uc , a_ot&, and a with aavg
.10
.08
lower bound 005
06 -Hong Kong njailrrerclay^J

Raticran H a w 1
.04
.02 4
0.00 0.00 .02 .04 .06 .08

avg
.10
67
Figure 2.14
Typical triaxial compression curves with step-changed strain-rates.
0.6 0.5 o
1c Confining pressure.kPa
Axial strain rate = 5%/h 0.5%/h
f 0.4 0.05%/h
Belfast clay (Graham, et al. 1983)
0.3 16%/h 0.2 0.1 0.0 -
r
i
0.255/h
~""~-'
Winnipeg clay (Graham, et al. 1983)
10
15
20
25
30
Axial Strain, %
2.15
Illustration of the preferred range of load increment for the measurement of C
Compression curve on intact specimen
Structure effect
Compression curve on remolded specimen Estimate Ca for load increments exceeding Point A
5*
Vertical Effective Stress, a' v , in log scale (kPa)
Figure 2.16
Normalized a' -e relationship at 10% vertical strain (sv -10%) for Berthierville clay at a depth of 3.9-4.8m (data from Leroueil et al. 1988)
Ate v =10%
lab:
Berthierville at a depth of 3.9-4.8;m
in situ:
CO
0.9
0.8
10-9
-I o-4
-to-3
Strain rate, /min
70 re 2.17 Normalized a'p-s relationship at 10% vertical strain (e v = 10%) for St.
Alban clay from both laboratory tests and in situ observance (data from Leroueil et al. 1988)
Ate =10%
lab:
âint Alban clay jat a depth of 3.1 -4.9m
in situ:
10-4
10-3
Strain rate, /min
71
CHAPTER 3 A VISCOPLASTIC CONSTITUTIVE APPROACH FOR RATESENSITIVE STRUCTURED CLAYS

3.1 Introduction
It is generally recognized that most geologic materials are structured to some degree (e.g. Leroueil and Vaughan 1990; and Burland 1990; Malandraki and Toll 2000). For natural clay, there are two general forms of structure: (i) macrostructure which refers to visible features such as fissures, joints, stratification and other discontinuities in an otherwise intact soil mass (Lo and Milligan 1967; Lo 1970; Bishop and Little 1967; and Lo and Hinchberger 2006) and (ii) microstructure which arises from fabric effects and inter-particle bonding or cementation (Mitchell 1970). Although both types of structure can strongly influence the engineering response of natural clay, macrostructure such as fissures and joints can be seen with the naked eye and treated in engineering mechanics either by introducing joints and/or contacts between discrete elements of intact material (Cho and Lee 1993; Chen et al. 2000; Li et al. 2007) or by adopting a mass strength for the clay (Lo 1970 and Lo and Hinchberger 2006). In contrast, the influence of
microstructure is comparatively more difficult to assess in part due to its microscopic nature. Consequently, the majority of studies reported in the literature over the past 20 to 30 years have focused on either characterizing the influence of microstructure on the strength and stiffness of natural clay (Leroueil and Vaughan 1990; Burland 1990; Gasparre et al. 2007; Sorensen et al. 2007; etc.) or on constitutive proposals that include
A version of this chapter has been submitted to Canadian Geotechnical Journal 2007
72 the effects of microstructure (Baudet and Stallebrass 2004; Callisto and Rampello 2004; Karstunen et al. 2005). Typically, clay microstructure, hereafter referred to as structure, is mechanically characterized by comparing the response of natural intact clay to that of the corresponding reconstituted material. Examples of the influence of structure on the mechanical response of natural clay are given in Figures 3.1 and 3.2. Figure 3.1
compares the response of undisturbed and reconstituted Bothkennar clay during oedometer compression (Burland 1990) and Figure 3.2 compares similar behaviour for London clay during triaxial compression (see Sorensen et al. 2007 and Hinchberger and Qu 2007). Additional examples of the behaviour of structured clay during oedometer and triaxial compression tests can be found in Mesri et al. (1975), Philibert (1976), and Locat and Lefebvre (1985). As shown in Figure 3.1, structure permits natural clays to exist at higher void ratios than the equivalent reconstituted materials. Such a state in clay is typically
metastable leading to high compressibility when loaded past its preconsolidation pressure (Vaid et al. 1979; Leroueil et al. 1985). In addition, structure imparts additional strength to the soil skeleton above that which can be typically accounted for by state-parameters such as void ratio and stress history (see Figure 3.2). Again, this additional strength is typically metastable leading to significant post-peak strength loss with large-strain (Lo and Morin 1972) and creep-rupture at deviator stresses exceeding the large-strain postpeak strength (Philibert 1976). Behaviour such as that depicted in Figures 3.1 and 3.2 has lead various researchers to conclude: (i) that the effect of structure on the mechanical response of natural clay is as significant as state parameters such as void ratio and stress
73 history, which are commonly used in traditional soil mechanics models (Leroueil and Vaughan 1990) and (ii) it is critical to include structure and loss of structure during straining in constitutive models for natural clays (Baudet and Stallebrass 2004). Recently, both rate-independent (Liu and Carter 1999; Baudet and Stallebrass 2004; Callisto and Rampello 2004; and Karstunen et al. 2005) and rate-dependent (e.g. Kim and Leroueil 2001 and Rocchi et al. 2003) constitutive models have been proposed to model the mechanical response of structured clay. However, since most structured clays exhibit significant strain-rate sensitivity, creep and stress relaxation (Vaid et al. 1979; Leroueil et al. 1983; Silvestri et al. 1984; Leroueil et al. 1985), constitutive models that account for the viscous behaviour of clay are desirable. In terms of time-dependent constitutive proposals, the 1-dimensional elastic-viscoplastic model described by Kim and Leroueil (2001) has been shown to provide an encouraging description of the Berthieville test embankment (Kim and Leroueil 2001). Although 1-dimensional models can be sufficient for practical problems involving 1-dimensional settlement, they are not suited for the study of problems involving 2- or 3-dimensional behaviour. Rocchi et al. (2003) proposed an elastic-viscoplastic constitutive model for 2-dimensional analysis of structured clay. This model (Rocchi et al. 2003) was a useful step forward, however, it has been shown to only roughly describe the engineering behaviour of structured clay during K'0 compression. Currently, a time-dependent constitutive model capable of
describing the mechanical behaviour of structured clay for generalized 2-dimensional loading and stress-paths other than K'a -compression does not exist. The primary objective of this chapter is to describe the extension of an existing elastic-viscoplastic constitutive model (Hinchberger 1996; Rowe and Hinchberger 1998;
74 Hinchberger and Rowe 1998) to describe the influence of structure on the engineering behaviour of rate-sensitive structured natural clay. In the extended model, soil structure is accounted for mathematically using a state-dependent viscosity parameter, and a damage law that describes 'destructuration' of the clay. The model is tested by
comparing calculated and measured behavior of Saint-Jean-Vianney (SJV) clay for constant-rate-of-strain K'a -consolidation, and both isotropically consolidated undrained triaxial compression (CIU) tests and constant load CIU triaxial creep tests. Though these comparisons, it is shown that a single elastic-viscoplastic constitutive model can describe behaviour such as accelerated creep rupture, the influence of strain-rate on the peak undrained shear strength, large-strain post-peak undrained shear strength, and the apparent preconsolidation pressure of a structured natural clay. In addition, the
constitutive model does not rely on multiple or nested yield surfaces, which simplifies the formulation. The research presented in this chapter suggests a potential mathematical link between the time-dependent response of natural clay during tests involving either constant volume shear or volumetric compression. The model and its extensions should be of interest to researchers and practitioners in the field of soil mechanics or geomechanics.
75
3.2 Theoretical Formulation

3.2.1 Overstress viscoplasticity In the following sections, the elastic-viscoplastic model proposed by Hinchberger and Rowe (1998) is extended to account for the effect of 'structure' on the engineering behaviour of natural rate-sensitive clay. The Hinchberger and Rowe (1998) model is a three-parameter elastic-viscoplastic formulation based on the elliptical cap yield surface (Chen and Muzino 1990), Drucker-Prager failure envelope, Perzyna's theory of overstress viscoplasticity (Perzyna 1963) and concepts from the critical state framework (Roscoe et al. 1963). Full details of the constitutive model are presented by Rowe and Hinchberger (1998) and Hinchberger (1996). The following is a brief summary of the model. In the normally-consolidated stress range, the constitutive relationship is:
"
2G 3(1 + e)(j'm
ll
2G 3(1 + e)a'm
lJ
^ V " da'
W))- (<V<Ni M < ^ <s])

0 M<><<
[31]
where stj is the deviatoric stress tensor, a'm is the mean effective stress, Stj is Kronecker's delta, G is the stress dependent shear modulus, K is the slope of the e - ln(ciy) curve in the over-consolidated stress range, e is the void ratio, yvp is the fluidity parameter, &^ I o'^ is mathematically the overstress ratio (described below) anddF/dcrJj is the plastic potential, which is derived as a unit norm vector. The flow function, <|>(F), in Equation [3.1] is a power law (Norton 1929) similar to functions used
76 by Adachi and Oka (1982) and Kantona (1984). For normally consolidated clay, associated flow has been adopted (see Figure 3.3). Accordingly, the plastic potential, dFjda'y , is derived using the elliptical cap equation: F=(o' m - l ) 2 +2J 2 R 2 -(og> -if =0 [3.2]
where cyjjîs the dynamic yield surface intercept, 1 and R are parameters defining the aspect ratio of the elliptical cap, and J 2 is the second invariant of the deviatoric stress tensor, stj. In the constitutive formulation, the Drucker-Prager failure envelope is used to define the critical state viz.:
F=M cs a^+c; s -V2l7=0
[3.3]
where M cs is the slope of the Drucker-Prager envelope and c'cs is the effective cohesion intercept in *J2J2 - o'm stress space. The cap parameters 1 and R are determined so that the top of the cap (point B in Figure 3.3) is coincident with the critical state line or Drucker-Prager envelope. Lastly, strain hardening of the static yield surface, da'$, proportional to incremental plastic volumetric strain, <9e^, viz.: is
(X-K)
Thus, for normally consolidated clay, there are eight constitutive parameters that must be determined for this model: the compression and recompression indices, X
and K , the critical state parameters, M cs and c'cs, the static yield surface intercept, cr'^-1, the aspect ratio of the elliptical cap, R , the power law exponent, n , and the fluidity
77 parameter, yvp . addressed below. In the overconsolidated stress range, the constitutive equation (Equation [3.1]) incorporates the Drucker-Prager envelope: F=M0Ca'm+c;c-V2J^ [3.5] The distinction between static and dynamic yield surface will be
where M o c and c'oc define the slope and cohesion intercept of the yield envelop in yJ2i2 - <*'m stress space. As a result, the state boundary surface or yield surface of the soil is denoted by A-B-C in Figure 3.3 and defined by Equations [3.2] and [3.5]. As noted above, the Drucker-Prager equation is also used to define the critical state line. In this study, a non-associated flow rule was required to describe the volumetric response of Saint-Jean-Vianney Clay in the overconsolidated state. Accordingly, the parameter M has been utilized with Equation [3.5] to define the plastic potential, dg/da'^ (see point
D in Figure 3.3) and the resultant dilatant behaviour of Saint-Jean-Vianney clay for plastic states of stress approaching the critical state from the dry side. In total, eleven constitutive parameters must be measured to fully define the elastic-viscoplastic material behaviour. Although the number of constitutive parameters is significant, the parameters can be estimated from standard incremental oedometer consolidation, and undrained triaxial compression tests undertaken at different strainrates. 3.2.2 Numerical overstress Figure 3.3 illustrates the static (or reference) yield surface, the dynamic yield surface and the definition of overstress adopted in the elastic-viscoplastic formulation. In
78 viscoplastic theory, the static yield surface defines the yield loci mobilized at very low strain-rates. Stress states that lie inside the static yield surface are elastic. The intercept of the static yield surface with the mean stress axis is a ' ^ . The dynamic yield surface is used to define the level of overstress and the plastic potential, dF/dcr'y, for time-
dependent plastic flow. The intercept of the dynamic yield surface with the mean stress axis is c'^y*. In accordance with overstress viscoplasticity (Perzyna 1963), stress states are permitted to exceed the yield surface of the material (in this case the static yield surface). Points D and E in Figure 3.3 illustrate two states of overstress. Referring to Figure 3.3, a dynamic yield surface is defined passing through states of overstress (e.g. Points D and E) and, a'jff/a', 10% overstress). defines the overstress ratio (&lff/a'j= 1.1 implies
The resultant rate-of-plastic flow, sjf , is governed by the flow
function, <|>(F), in Equation [3.1]. As a result, a series of isotaches exists in yJ2J2 -a'm stress space (see Figure 3.3), which defines states of equivalent overstress or flow potential, (|)(F). Suklje (1957) proposed similar isotaches for equal volumetric strainrates in the e-a'v plane. In this chapter, the concept is applied to the magnitude of the viscoplastic strain-rate tensor in ^2J2 -<y'm stress space.
3.2.3 Modification for soil structure Most structured soils exhibit characteristics such as creep rupture during both drained and undrained triaxial creep tests (Vaid et al. 1979; Lefebvre et al. 1982). Previously, this type of behaviour has been modeled using overstress viscoplasticity theory (Perzyna 1963). For example, Adachi et al. (1987) introduced a state-dependent
79 fluidity parameter in the Adachi and Oka model (1982) to account for accelerated creep rupture of Umeda clay (Sekiguchi, 1984). Aubrey et al. (1985) describe a similar modification of the Adachi and Oka (1982) model utilizing a damage law. Recently, Kimoto et al. (2004) incorporated state-dependent viscosity parameters in the Adachi and Oka (1982, 1995) model to describe strain softening of structured clay. However, in spite of the potential of this approach, relatively little attention has been given to the use of state-dependent viscosity parameters to describe the behaviour of time-dependent structured clay for generalized stress states. Here, Equation [3.1] is extended using a state-dependent fluidity parameter to describe the engineering behaviour of structured rate-sensitive clay. In the new
formulation, the parameter, oa0, is introduced to mathematically define the structure viz.:
coo={rTlfsPYn
and n=lla
[3.6]
[3.7]
where yj p is the fluidity of the structured or undisturbed clay fabric, y^p is the fluidity of the destructed or intrinsic fabric, and n is the power law exponent from the power law in Equation [3.1], and a is the rate-sensitivity parameter (see Chapter 2). The structure parameter, oo0, is related to common engineering parameters as shown below. Next, the concept of damage strain, e d , is used to define the transition from an initially viscous state (the structured state) to a more fluid destructed state viz.: ded = A /(l-A)(d e :P 1 ) 2 +A(de s vp ) 2 [3.8]
In Equation [3.8], originally proposed by Rouainia and Wood (2000), ded is the
80
incremental damage strain, d e ^ and de^p are incremental plastic volumetric and plastic octahedral shear strains, respectively. A is a weighting parameter, which is assumed to be 0.5 similar to Baudet and Stellebrass (2004). Lastly, an exponential damage law is introduced to describe the process of structure degradation: aX8d)=[l + (co 0 n -l.o)e- b e d ] 1 / n [3.9]
where b is a material parameter governing the rate of destructuration, ed is the damage strain, 6)0 defines the initial structure and co(ed) defines the state-dependent structure level. On inspection of Equation [3.9], it can be seen that co(ed)=a)0 for undisturbed clay (e.g. ed = 0 ) and that co(ed) decreases to 1 as the plastic strain and consequent damage strain becomes large. damage strain: fp(ed)=yjp lcon{ed) [3.10] Accordingly, the fluidity parameter is a function of
and the viscoplastic strain-rate tensor is: ^ p =Y v p (e d )(<t)(F))^ [3.H]
In the extended elastic-viscoplastic constitutive model, the fluidity of the structured clay fabric, yJp, is assumed to be significantly lower than the fluidity of the destructured fabric, y^p. For the undisturbed structured state, plastic deformation of the clay fabric is initially restrained by the low structured fluidity, yj p , permitting overstress to build up relative to the static state boundary surface. However, with continued plastic straining, damage causes the soil viscosity to break down and the clay fabric to become more fluid. This process is commonly referred to as destructuration (Baudet and
81 Stallebrass 2004). Thus, it is hypothesized that structure is caused by viscous bonding between particles and that destructuration is a stress relaxation phenomenon whereby the viscosity of the structured soil is gradually reduced due to plastic strain until eventually the destructured or intrinsic state is reached (governed by the Hinchberger and Rowe (1998) model). This hypothesis is tested using Saint-Jean-Vianney clay. The following sections describe the methodology of this study and selection of material parameters for the structured soil model.
3.3 Methodology 3.3.1 Laboratory tests Vaid et al. (1979) reported the results of constant rate-of-strain K -consolidation, and CIU undrained triaxial compression and triaxial creep tests performed to study the time-dependent behaviour of SJV clay. The experimental program and methodology is described in detail by Vaid et al. (1979) and Campanella and Vaid (1972). Only those details required for the present study are repeated here. The specimens utilized in the laboratory program (Vaid et al., 1979) were trimmed from block samples retrieved from the site of the SJV slide. Consequently, the samples are considered to be of a high quality and the measured laboratory behaviour is considered representative of the undisturbed natural clay; notwithstanding that limitations of the apparatus used by Vaid et al. (1979) may have affected the measured response. In addition, Vaid et al. (1979) observed that the block sample used in their testing program had distinct upper and lower layers and that the behaviour of these layers was significantly different during the laboratory testing. Thus, laboratory results are separated into upper and lower layers.
82
Properties of Saint-Jean-Vianney clay are summarized in Table 3.1 and the experimental results are reproduced and compared with the constitutive model in Figures 3.6, 3.8 and 3.11 to 3.19, inclusive.
3.3.2 Numerical approach Calculated behaviour has been obtained by modeling laboratory test conditions using the finite element program AFENA (Carter and Balaam 1995), which has been modified by the authors to include a rate-sensitive 'structured' clay model. In all cases, 8-noded rectangular isoparametric elements were used assuming axisymmetric geometry. For each test, the number of elements and time-steps were varied to ensure convergence of the calculations. An undrained finite element formulation was used to obtain the calculated behaviour of SJV clay for isotropically consolidated undrained (CIU) triaxial compression tests. A fully drained formulation (e.g. uncoupled) was adopted for constant rate-of-strain K'0 -consolidation tests. Chen and Muzino (1990) describe similar
formulations for elastoplastic analysis. A fully coupled formulation was used to model undrained creep and creep-rupture (Hinchberger 1996). In all cases, an incremental solution approach was adopted. For all compression tests on SJV clay, compression was simulated by applying boundary displacements at a rate that matched the displacement-rate in the corresponding laboratory test. Smooth rigid end conditions were assumed. For triaxial compression of overconsolidated clay, there is a limitation on the stress state that can be applied to the specimen. It is normally assumed that the excess pore pressure in a triaxial specimen cannot exceed the cell pressure. Thus, for calculated behaviour during CIU triaxial compression and creep tests, stress states exceeding the triaxial limit were corrected back
83
to the triaxial limit and the nodal forces required to make this correction were applied in the force vector for the subsequent increment to maintain equilibrium. This type of approach is commonly used in incremental elastoplastic analysis to correct stress states that exceed the failure criterion. As noted above, CIU triaxial creep tests were modeled using a fully coupled formulation. To simulate the behaviour of SJV clay during CIU creep, uniform deviator stresses were specified at the top mesh boundary. This approach ignores stress
concentrations within the sample due to the relatively stiff end-caps. Axial loads were applied incrementally over a period of 30 seconds and maintained at a constant level for the duration of the test. This was done to avoid numerical instability. A smooth rigid boundary was adopted at the bottom of the finite element mesh and pore pressures were constrained by the triaxial limit as described above (in this case ue < G'celi = 40 kPa).
3.3.3 Selection of constitutive parameters The material parameters utilized in this chapter can be divided into three groups as summarized in Table 3.2. The three groups include: (i) conventional elastoplastic constitutive parameters, which define the variation of soil stiffness and strength versus the state variables void ratio and stress history, (ii) the intrinsic viscosity parameters (y^p and n), which govern the fluidity and rate sensitivity of the soil skeleton, and (iii) structure and destructuration parameters ( co0andb ), which govern the structure component of the constitutive behaviour. The following is a brief description of how the parameters were derived from the experimental tests. Additional details can be found in Qu and Hinchberger (2007). A single set of elastic-plastic parameters were used for the analyses reported
84
below. First, the critical state parameter, M cs = 1.34, was determined from the measured structured friction angle (()>' = 40) of SJV clay reported by Vaid et al. (1979). Figures 3.4 and 3.5 illustrate selection of the state boundary surface parameters from laboratory results in Saihi et al. (2002). In the N/C stress range, the aspect ratio of the elliptical cap yield surface, R = 0.7 , was estimated from the stress path response of normally consolidated SJV clay during CIU triaxial compression. Figure 3.4 illustrates this
parameter selection. Similarly, in the O/C stress range, the yield surface parameter, M oc = 0.48, was estimated from CIU triaxial tests on overconsolidated specimens of SJV clay as shown in Figure 3.5. The intercept of the Drucker-Prager envelop in the O/C stress range, c^., is a dependent parameter determined by the yield surface intercept (either a ^ or a ^ ) . The initial void ratio, e 0 =1.15, was estimated from the natural
moisture content and specific gravity reported in Table 3.1 and Poisson's Ratio, v=0.33, was assumed. Only one of the plastic constitutive parameters, M =0.01, was obtained by some trial and error. Initially, calculations were performed assuming associated flow in the O/C stress range; however, such calculations tended to overestimate the post-peak pore pressures at large-strain by about 20%. Consequently, reduced dilatancy was
assumed. Further details are provided in Chapter 3. In the constitutive model, compressibility of the intrinsic soil skeleton is defined by the recompression and compression indexes, K and X, respectively. To estimate A,, normalized one-dimensional compression curves for SJV clay where plotted as shown in Figure 3.6 (e.g. data from each test was normalized by the mobilized preconsolidation pressure c'p). Figure 3.6 also shows the assumed intrinsic compression line (ICL). For
85 SJV clay, the intrinsic compression index, A,, was taken to be 0.26 giving an ICL parallel too but below that of clays from Drammen, Tilbury, St. Andrews, Tilbury, Alvangen and several ocean cores over the stress range lOOkPa to l,000kPa (see Figure 3.7). From Figure 3.7, it can be seen that the assumed ICL for SJV clay is parallel to that measured for other natural clays up to a vertical stress of about 2000kPa. For stresses exceeding 2000kPa, it is anticipated that the intrinsic compression index, X, of SJV clay would reduce since ICL's are typically concave upward (Burland 1990). The recompression index, K , was taken to be 0.02, which can be easily deduced from the measured
compression behaviour shown in Figure 3.6. The power law exponent, n, can be estimated from data presented in Figure 3.8. For elastic-viscoplastic models based on a power law flow function, (|)(F), n can be obtained by plotting either undrained shear strength, S u , or apparent preconsolidation pressure, a' , versus strain-rate on a log-log scale. The power law exponent (n = 22) is inversely proportional to the slope of this plot (see Figures 3.8a, b and c). In addition, it should be noted that for many natural clays, including SJV clay, the power law exponent, n , remains constant during 'destructuration' or straining as shown in Figure 3.9 for Winnipeg clay, Belfast clay and London clay (see Figure 3.2 for the stress-strain response of London clay). Exceptions to this have been noted by Sorensten et al. (2007). The fluidity parameters (yj p and y,vp) and the static yield surface intercept ( a ^ ) are inter-related parameters and the most difficult parameters to assess for overstress viscoplastic formulations. Detailed guidance on the selection of these parameters can be found in Chapter 3. In the absence of specific testing to determine the fluidity of SJV clay, a structured fluidity of lxlO"10 min"1 was assumed from long-term observations at
86 the Berthierville, Gloucester and St. Alban test sites (see Leroueil 2006). These are also Champlain clay deposits. The structured fluidity (lxlO 1 0 min"1) defines the transition from inviscous behaviour for strain-rates less than or equal to yj p to viscous behaviour for strain-rates greater than y j p . Referring to Figure 3.8(a) and (b), the static yield surface intercept for both the upper (ay} =405kPa) and lower (a^y* =518kPa) layers of SJV clay can be estimated from Point 'A' using Equations [3.2] and [3.5], which define the state boundary surface. Alternatively, the static yield surface intercept can be deduced from the apparent preconsolidation pressure versus strain-rate assuming ysvp =lxl0" 1 0 min _ 1 , K'o=0.5 and using Equation [3.2]. Referring to Figure 3.8(c), the static yield surface intercept estimated from the consolidation response varies from 370kPa to 410kPa, which is comparable to that deduced in Figure 3.8(a) for the upper clay layer. Thus, it has been inferred that the constant rate of strain consolidation tests were performed on clay from the upper layer although this is not explicitly stated by Vaid et al. (1979). The structure parameter, co0, can be obtained from either: (i) undrained shear strength versus strain-rate (see Figures 3.8a and 3.8b) or (ii) from the intrinsic, o'p_{, and structured, a' s , preconsolidation pressures in oedometer compression (see Figure 3.6). In this chapter, the latter approach is used. From Figure 3.6 and Equation [3.2], it can be shown that the structure parameter is a>0 = a'p_s IG'^ =1.68. Lastly, from Equation [3.9], the constitutive parameter b governs the rate-of-destructuration and the magnitude of strain at which the intrinsic state is reached. Referring again to Figure 3.6, it can be seen that SJV clay reaches the intrinsic state during 1-D compression at an axial strain of
87 approximately 13% (e.g. 0.28/(1 + 1.15)=0.13), which can be achieved using Equation [3.9] for b = 120. Similar analysis of the behaviour during CIU triaxial compression can be used to deduce b = 4000. The rate of damage is significantly higher during CIU triaxial compression, which can be attributed to strain localization.
3.4 Evaluation (Saint-Jean Vianney Clay) 3.4.1 Theoretical behaviour of the model for CIU triaxial compression
Figure 3.10 illustrates the basic features of the extended constitutive model during CIU triaxial compression test on heavily over-consolidated structured clay. Initially, in accordance with the test conditions, the clay is isotropically consolidated to a stress state significantly lower than the static yield surface (see Point 1 in Figure 3.10). During undrained triaxial compression, the effective stress-path moves from point 1 to 2 where the triaxial limit is reached and then from point 2 to 3 on the Drucker-Prager envelope where first yield occurs. Continued strain-rate controlled compression causes the stressstate to exceed the static yield surface moving from point 2 to point 4 along the triaxial limit. During this stage of the test, significant overstress builds in the model due to the low fluidity of the structured soil skeleton, yj p . The initial low fluidity, yj p restrains plastic deformation of the soil skeleton and the resultant load-displacement response is essentially elastic from point 1 to 4. From point 1 to 4, components of the plastic strainrate tensor are finite but very small: non-associated flow is assumed (see Point 3). At point 4, the overstress becomes large enough to cause significant plastic straining and consequent destructuration, e d , which begins to dominate the constitutive behaviour. During further compression, the soil fluidity increases from yj p to y ^ in
88
accordance with Equations [3.9] and [3.10] and the overstress built up during compression from point 1 to point 4 dissipates from point 4 to 6. The strain-softening is treated as a stress-relaxation problem and the rate of softening is governed by the parameter b in Equation [3.9].
3.4.2 Calculated and measured behaviour for constant rate-of-strain triaxial compression As discussed above, Vaid et al. (1979) carried out constant rate-of-strain CIU triaxial compression tests on specimens of SJV clay consolidated to an isotropic effective stress of 40kPa. A low consolidation pressure was chosen to study the engineering behaviour of SJV clay in a highly overconsolidated state. Figures 3.11(a) and (b) show calculated and measured behaviour for the upper and lower layers during CIU triaxial compression. As noted above, Vaid et al. (1979) reported the existence of both upper and lower clay layers in the block sample used for their investigation. Figure[G36] 3.12 compares the measured and calculated peak and post-peak (large-strain) deviator stress, a d , versus strain-rate and Table 3.2 summarizes the constitutive parameters used in the computations. Focusing on Figure 3.11(a) and (b), it can be seen that there is reasonable agreement between the calculated and measured stress-strain response of SJV clay. Differences between measured and calculated behaviour prior to reaching the peak strength can be attributed to the elastic properties used in the analyses. The finite element calculations were undertaken using an average elastic modulus deduced from all of the undrained compression tests whereas the lower clay appears to be stiffer and the upper clay is less stiff than the average value selected. Other than some variations, which can
89 be attributed to natural variation of SJV clay, the agreement between measured and calculated behaviour is reasonable. Referring to the excess pore pressure response plotted in Figures 3.11(a) and (b), there is generally good agreement between the measured and calculated excess pore pressures at the peak strength and at the large-strain post-peak state (e.g. for axial strains exceeding 2.5%). After reaching the peak strength, however, there is considerable
divergence of the calculated and measured behaviour particularly for the upper layer as shown in Figure 3.11(a). Variations between the measured and calculated excess pore pressure are less for the lower layer shown in Figure 3.11(b). Overall, the calculated and measured response is for the most part similar and variances can be attributed to the use of average constitutive parameters (E and b [G37]) for the computations and the natural variation of the SJV clay. Figure 3.12 compares calculated and measured peak and post-peak shear strength for both the upper and lower clay layers. From Figure 3.12, it can be seen that there is very good agreement between measured and calculated peak undrained shear strength versus strain-rate for both upper and lower clay layers. The constitutive model
overestimates the large-strain post-peak strength ( e ^ ^ > 3%) versus strain-rate for the lower clay layer but gives good predictions of the post-peak strength for the upper layer. In general, the data presented in Figure 3.12 suggests that the lower clay is more structured than the upper clay and over prediction of the post-peak strength of the lower layer is a consequence of using the same structure parameter for both layers (see Table 3.2). A higher structure parameter is[G38] also indicated by Figure 3.8(b) for the lower layer. Considering that the structure parameter used in the computations was estimated
90 from the e - l o g a ^ response of SJV clay, the general behaviour of the model is quite satisfactory.
3.4.3 CIU triaxial creep tests In addition to triaxial compression, CIU triaxial creep tests were performed at constant deviator stresses ranging from 430 kPa to 630 kPa. Figures 3.13(a) and (b) compare measured and calculated axial strain versus log-time for the upper and lower layers, respectively. Figure 3.14 shows measured and calculated creep rupture time. The creep tests were simulated using the same constitutive parameters used in the previous section (see Table 3.2). Referring to Figures 3.13(a) and (b), all of the laboratory test specimens were able to initially support the applied deviator stress, od, for some time prior to failure. Failure was manifest by accelerated creep rates (creep rupture) accompanied by a rapid reduction in pore pressure. The constitutive formulation described in this chapter is able to
simulate such behaviour. The theoretical response shown in Figures 3.13(a) and (b) is very similar to the measured response. In the theoretical calculations, the applied
deviator stresses exceed the long-term static state boundary surface of the material. The material begins to creep at a rate governed by the structured fluidity, yj p . However, with time, damage or destructuration occurs (governed by Equations [3.8] and [3.9]) causing the fluidity of the soil skeleton to increase from yj p to eventually y,vp. This process of destructuration leads to accelerating axial creep rates and creep rupture. For the structure parameter assumed in the analysis, co0 =1.68, the clay fluidity increases by almost five orders of magnitude during the calculations due to damage strain.
91 Detailed inspection of Figure 3.13(a) for the upper clay shows that the calculated initial strain during each test exceeds the measured strain. This can be attributed to the use of an average modulus of elasticity for the computations. For the upper layer, the measured and calculated times to failure are close. For the lower clay layer shown in Figure 3.13(b), the calculated and measured initial strains are in good agreement except for the test undertaken at a deviator stress, a d , of 630kPa. The high initial strain
measured during this test is not consistent with the other experimental observations suggesting possible sample disturbance. Again from Figure 3.13(b), there is good
agreement between calculated and measured time to creep rupture (rupture life) at a deviator stress of 575 kPa. For o d =630kPa and 440 kPa, however, the calculated creep rupture times become approximate as discussed below. For both upper and lower layers (see the lower parts of Figs. 13(a) and (b)), the rapid generation of excess pore pressure that occurs at failure (rupture) is predicted by the model. Figure 3.14 compares calculated and measured creep rupture life during CIU triaxial creep tests. As observed above for the upper clay, there is good agreement between the theoretical and measured rupture life. For the lower clay, the agreement between measured and calculated behaviour is less accurate and can be characterized as approximate. The difference between calculated and measured rupture life at
Gd = 630kPa can be attributed at least in part to the time required to apply loads to the specimen during the computations (30 seconds). Considering, however, that natural variation of Champlain clays can be quite significant (Robertson 1975), the constitutive model appears to give useful predictions of creep rupture life even for the lower clay layer. From an engineering point of view, predicting instability or meta-stability is an
92 important characteristic whereas estimating the exact time of the instability is of lesser importance. Figure 3.15 shows the measured and calculated creep rates prior to creep rupture illustrating the main limitation of the constitutive model. The constitutive model predicts constant creep rates prior to creep rupture whereas the measured creep rates reduce with time. Similar measured behaviour has been reported for other clays (Bishop and
Lovenby 1973 and Tavenas et al. 1978). However, on reinspection of Figures 3.13(a) and (b), the axial strain that accumulates prior to creep rupture is generally less than 0.1%, which would be difficult to detect or measure in situ. Thus, from a practical point of view, the inability to predict diminishing creep rates with time is not a significant limitation. The phenomenon, however, can be predicted by allowing some strain-
hardening of the Drucker-Prager envelop similar to that done by Lade and Duncan[G39] (1973) or by assuming some rotational hardening of the state boundary surface introducing additional constitutive parameters (see Appendix F). Lastly, Figure 3.16 compares calculated and measured strength (or creep stress) versus strain-rate in both creep tests and triaxial compression tests. Figure 3.16(a) shows that the relationship between undrained strength and strain-rate in the constant rate-ofstrain tests is consistent with the relationship between creep stress and the minimum strain-rate in the constant stress tests (Vaid et al. 1979). This measured behaviour agrees well with the theoretical relations. As shown in Figure 3.16(b), the measured creep rate versus Gd exhibits a similar pattern, but with significant natural variation for the lower clay layer whereas the data for the upper layer is more uniform. In general, from Figure 3.16, it can be seen that the constitutive model provides a good estimate. In addition,
93 from a theoretical point of view, this figure shows that for a given clay the power law exponent, n, in the constitutive model can be deduced from the minimum strain-rate measured immediately prior to creep rupture.
3.4.4 Theoretical response for constant rate-of-strain consolidation Figures 3.17(a) though (e) illustrate the theoretical behaviour of the constitutive formulation for constant rate-of-strain K^ -consolidation. Figures 3.17(a) and (b) show a typical stress path during strain-rate controlled K^,-compression. Figure 3.17(c) shows the theoretical void ratio versus mean effective stress during K^ -consolidation. The structure parameter is shown versus mean effective stress, a'm, and damage strain, e d , in Figures 3.17(d) and (e), respectively. Referring to Figure 3.17(a), the structured soil is assumed to initially behave as an elastic material during K^ -compression as the stress state moves from point 1 to 2. At point 2, the stress state reaches the static yield surface where there is a transition from elastic to plastic behaviour. During continued compression, the material response from point 2 to 4 remains predominantly elastic even though the stresses exceed the static yield surface. This is evident in Figures 3.17(a), (b) and (c), inclusive. From point 2 to 4, the structured fluidity and level of overstress are too low to cause significant plastic flow. Thus, although the magnitude of the viscoplastic strain-rate tensor is finite, it is very small and the material response is predominantly elastic (Figure 3.17c). As a result, the ratio of vertical to horizontal stress from 1 to 4 is governed by Poisson's ratio. At point 4, the combined overstress and structured fluidity are such that the viscoplastic strain-rate becomes sufficiently large to dominate the constitutive behaviour
94 causing destructuration. There is a change in the stress path from 4 to 5 as the specimen adjusts to the plastic strain-rate tensor (or plastic potential). During the transition from 4 to 5, the static yield surface begins to expand due to strain hardening. Continued
compression beyond point 5 causes further destructuration and consequently increased fluidity. Eventually, at point 6, the initial structure is destroyed and the compression curves of the structured and intrinsic soil skeleton merge (see Figure 3.17c). Although it is not entirely evident in Figures 3.17(a) through (e), the theoretical behaviour during K^ -compression is analogous to that described for CIU triaxial compression. Initially, significant overstress builds in the model from point 2 to 4 due to the low fluidity of the structured soil fabric. The overstress reaches a maximum at point 4 and begins to dissipate as the magnitude of the plastic strain-rate tensor becomes large enough to dominate the material behaviour. From point 4 to 6 in Figure 3.17, there is stress-relaxation similar to that described in Figure 3.2 for CIU triaxial compression. As shown in Figure 3.17(c), the overstress ratio, G'^ lo'^, is greatest at the mobilized
apparent preconsolidation pressure (point 4) and lowest at point 6 where destructuration is complete. In accordance with the theory of overstress viscoplasticity, reducing the rate-of-compression relative to that illustrated in Figure 3.17(c) will cause points 3, 4, 5, 6 and 7 to shift to the left toward 1, 2 and 7: the long-term compression curve. Conversely, increasing the strain-rate will cause points 3, 4, 5, 6 and 7 to shift to the right. The strain-rate parameter, n, controls the degree of rate-sensitivity (or shift). 3.4.5 Constant rate-of-strain consolidation Vaid et al. (1979) conducted a series of constant rate-of-strain consolidation tests on 6.1 cm diameter and 2.5cm high specimens using a K^-triaxial cell (Campanella and
95 Vaid, 1972). Drainage was permitted from the top of each specimen and excess-pore pressures measured at the bottom. The compressibility of SJV clay was studied using strain-rates slow enough to allow almost complete dissipation of excess pore pressures. However, Vaid et al. (1979) reported that the maximum excess pore pressure in the fastest test was 7% of the applied vertical stress for stresses above 600kPa. Leakage was also observed during the slowest tests as discussed below (Robertson 1975). Figure 3.18(a) shows calculated and measured normalized compression curves during K^ -compression. Figure 3.18(b) shows variation of the tangent virgin
compression index (Cc) versus vertical effective stress. Although Vaid et al. (1979) did not specify the origin of the test specimens, it has been inferred that all tests were performed on the upper layer (see previous discussion of Figure 3.8c). In addition, the constitutive parameters used to obtain the calculated behaviour are the same as those listed in Table 3.2 and used in prior computations. From Figure 3.18(a), there is good agreement between the measured and calculated normalized compression curves. The measured response is affected by natural variation. Referring to Figure 3.18(b), the calculated and measured variation of Cc versus vertical effective stress agrees well. In the proposed constitutive model, the initial high compressibility mobilized after reaching the apparent yield stress is due to the high structured fluidity, yj p , which permits the material to exist in a metastable state at a higher void ratio than would be expected for the equivalent destructured or reconstituted material. The post yield compressibility is governed by Equations [3.8] and [3.9]. To conclude, Figure 3.19 shows measured and calculated variation of the e-loga' v response versus strain-rate. The calculated behaviour was obtained using a
96 static yield surface intercept of o^Sy = 405 kPa for tests undertaken at rates of 1.68xl0~2 and 6.78xl0"2%/min and o'^^TOkPa for the tests undertaken at 1.17xl0"3 and 3.48x10"
3
%/min. Both static yield surface intercepts were deduced from the data plotted in Figure
3.8(c). Overall, there is good agreement between the calculated and measured response of SJV clay during strain-rate controlled K^ -compression. The data presented in this figure further illustrates the challenge of working with natural clays, which often exhibit significant natural variation. However, in spite of the need for some interpretation of the results, the proposed constitutive model is able to describe the general trends in behaviour of SJV clay during drained K^ -compression using parameters that are readily deduced from standard rate controlled laboratory tests.
3.5 Summary and Conclusions

In this chapter, an existing elastic-viscoplastic constitutive model has been extended using a state-dependent viscosity parameter to describe the response of ratesensitive structured clay. The formulation has been tested by comparing calculated and measured behaviour of Saint-Jean-Vianney clay during CJU triaxial compression, CIU triaxial creep and constant rate-of-strain K -consolidation tests. Considering the
challenges introduced by natural variation, the agreement between calculated and measured response was found to be reasonable notwithstanding some differences. The proposed formulation has been shown to describe the predominant effects of stain-rate on the engineering behaviour of Saint-Jean-Vianney clay. With the exception of the damage parameter, b , in Equation [3.9], the behaviour of Saint-Jean-Vianney clay could be
97 adequately described during both drained and undrained tests using a single set of constitutive parameters. Many natural clays develop strain-localization or shear banding at failure. Accordingly, the higher damage parameter, b , required to describe shear failure (b =4000 versus b =120) may be due to strain localization. In addition, it has been recognized that the stiffness of triaxial apparatus can influence the rate of strain softening (Lo 1972). Consequently, it must be recognized that the apparatus used by Vaid et al. (1979) may have had an impact on the post-peak response of Saint-Jean-Vianney clay. Most likely, both strain-localization and apparatus effects are incorporated in the damage parameter, b , deduced above. In terms of the rate-sensitivity of SJV clay, the research reported in this chapter suggests that rate sensitivity and structure during drained and undrained laboratory compression can be linked mathematically using state-dependent viscosity terms and overstress viscoplasticity theory (Perzyna 1963). As shown above, the structure
exhibited in undrained compression and creep tests performed in a triaxial cell could be described using a structure parameter deduced from oedometer compression tests. This is considered to be new and of interest to researchers in the field of geomechanics. Lastly, many natural structured clays exhibit significant anisotropy (Lo and Morin 1972). Both anisotropy and rate-sensitivity seem to be inherently linked for structured clays such as St. Louis clay and St. Vallier clay (Lo and Morin 1972). In this study anisotropy has been neglected. This is considered to be satisfactory for the present investigation given the stress paths investigated (triaxial and K^ -consolidation) and the absence of principal stress rotations. Regardless, there are normally significant principal
98 stress rotations in situ and the impact of anisotropy on the engineering response of structured clays could be quite significant. Overall, based on the analyses presented, the following conclusions are drawn: The proposed formulation is able to generally describe much of the engineering behaviour of SJV clay using constitutive parameters that are determined from standard rate-controlled laboratory tests. The formulation can describe metastable phenomena such as accelerated creep rupture and high compressibility after reaching the apparent preconsolidation pressure of structured clay in addition to the influence of strain-rate on the peak shear strength, post-peak shear strength and the apparent preconsolidation pressure of SJV clay. It appears that the rate-dependent behaviour during undrained and drained laboratory tests can be simulated with a power law and unique single power law exponent for SJV clay and for the range of strain-rate studied. The structure parameter deduced from drained oedometer compression tests can also simulate the influence of structure during undrained triaxial creep and compression tests. At times, the proposed constitutive model becomes approximate particularly where natural variations may have affected the measured response and average soil parameters have been used in the computations. Specifically, there are differences
between calculated and measured post-peak excess pore pressures during CIU compression (see Figure 3.10a - upper layer) and creep rupture life (see Figure 3.14 lower layer). The main limitation of the proposed constitutive model is that it cannot describe
99 the reduction of creep rates with time that are normally measured prior to creep rupture. This is a minor limitation, however, since the strains that accumulate prior to creep rupture are small and probably immeasurable in situ. With the exception of the damage parameter b , a single set of constitutive parameters can describe both drained and undrained behaviour of Saint-Jean-Vianney clay over the range of strain-rates considered. Accordingly, it appears that it may be possible to mathematically link the time-dependent behaviour of structured clay for stress paths causing either shearing or compression. Traditionally, there has been an artificial distinction between such behaviours (Tavenas et al. 1978). The damage parameter, b , appears to be affected by strain localization. This should be investigated further. The constitutive model is proposed as an alternative to structured models that use multiple or nested yield surfaces (Rocchi et al. 2003). Although the model can describe most of the rate-sensitive and structured behaviour of SJV clay, the model should be extended to include the effects of anisotropy, which can be pronounced in natural clays.
References
Adachi, T., and Oka, F. 1982. Constitutive equations for normally consolidated clay based on elasto-viscoplasticity. Soils and Foundations, 22(4): 57-70. Adachi, T., and Oka, F. 1995. An Elastoplastic Constitutive Model for Soft Rock with Strain-Softening. International Journal for Numerical and Analytical Methods in Geomechanics, 19(4): 233-247. Adachi, T., Oka, F., and Mimura, M. 1987. Mathematical structure of an overstress elasto-viscoplastic model for clay. Soils and Foundations, 27(3): 31-42. Aubry, D., Kodaissi, E., and Meimon, Y. 1985. A viscoplastic constitutive equation for clays including a damage law. In Fifth International Conference on Numerical Methods in Geomechanics. Nagoya, pp. 421-428. Baudet, B., and Stallebrass, S. 2004. A constitutive model for structured clays. Geotechnique, 54(4): 269-278. Bishop, A.W., and Little, A.L. 1967. Influence of size and orientation of sample on apparent strength of London clay at Maldon, Essex. Oslo, Norway, Vol.1, pp. 8996. Bishop, A.W., and Lovenbury, H.T. 1969. Creep characteristics of two undisturbed clays. In Proc. 7th ICSMFE. Mexico, Vol.1, pp. 29-37. Burland, J.B. 1990. On the compressibility and shear strength of natural clays. Geotechnique, 40(3): 329-378. Callisto, L., and Rampello, S. 2004. An interpretation of structural degradation for three natural clays. Canadian Geotechnical Journal, 41(3): 392-407.
Campanella, R.G., and Vaid, Y.P. 1972. A simple Ko triaxial cell. Canadian Geotechnical Journal, 9(3): 249-260. Carter, J.P., and Balaam, N.P. 1990. AFENA-A general finite element algorithm: users manual, School of Civeil Engineering and Mining Engineering,University of Sydney, Australia. Chen, S.G., Cai, J.G., Zhao, J., and Zhou, Y.X. 2000. Discrete element modelling of an underground explosion in a jointed rock mass. Geotechnical and Geological Engineering, 18(2): 59-78. Chen, W.-F., and Mizuno, E. 1990. Nonlinear analysis in soil mechanics : theory and implementation. Elsevier Science Publishing Company Inc., New York, NY, U.S.A. Cho, T.F., and Lee, C. 1993. New discrete rockbolt element for finite element analysis. International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, 30(7): 1307-1310. Gasparre, A., Nishimura, S., Coop, M.R., and Jardine, R.J. 2007. The influence of structure on the behaviour of London Clay. Geotechnique, 57(1): 19-31. Graham, J., Crooks, J.H.A., and Bell, A.L. 1983. Time effects on the stress-strain behaviour of natural soft clays. Geotechnique, 33(3): 327-340. Hinchberger, S.D. 1996. The behaviour of reinforced and unreinforced embankments on rate senstive clayey foundations. Ph.D Thesis, University of Western Ontario, London. Hinchberger, S.D., and Rowe, R.K. 1998. Modelling the rate-sensitive characteristics of the Gloucester foundation soil. Canadian Geotechnical Journal, 35(5): 769-789.
Hinchberger, S.D., and Qu, G. 2007. Discussion: the Influence of structure on the timedependent behaviour of a stiff sedimentary clay. Geotechnique, Accepted. Karstunen, M., Krenn, H., Wheeler, S.J., Koskinen, M , and Zentar, R. 2005. Effect of anisotropy and destructuration on the behavior of Murro test embankment. International Journal of Geomechanics, 5(2): 87-97. Katona, M.G. 1984. Evaluation of Viscoplastic Cap Model. Journal of Geotechnical Engineering, 110(8): 1106-1125. Kim, Y.T., and Leroueil 2001. Modeling the viscoplastic behaviour of clays during consolidation: Application to Berthierville clay in both laboratory and field conditions. Canadian Geotechnical Journal, 38(3): 484-497. Kimoto, S., Oka, F., and Higo, Y. 2004. Strain localization analysis of elasto-viscoplastic soil considering structural degradation. Computer Methods in Applied Mechanics and Engineering, 193(27-29): 2845-2866. Lade, P.V., and Duncan, J.M. 1973. Ccubical triaxial tests on cohesionless soil. Journal of Soil Mechanics and Foundations Division ASCE(99(SM10)): 793-812. Lefebvre, G., Langlois, P., Lupien, C , and Lavallee, J.-G. 1982. Laboratory testing and in situ behaviour of peat as embankment foundation. In Canadian Geotechnical 35th Conference: Water Retaining Structures. Montreal, Quebec, Canada. Canadian Geotechnical Soc, Montreal, Que, Can, pp. 113-142. Leroueil, S., and Vaughan, P.R. 1990. The general and congruent effects of structure in natural soils and weak rocks. Geotechnique, 40(3): 467-488.
Leroueil, S., Samson, L., and Bozozuk, M. 1983. Laboratory and field determination of preconsolidation pressures at Gloucester. Canadian Geotechnical Journal, 20(3): 477-490. Leroueil, S., Kabbaj, M., Tavenas, F., and Bouchard, R. 1985. Stress-strain-strain rate relation for the compressibility of sensitive natural clays. Geotechnique, 35(2): 159-180. Li, S.H., Wang, J.G., Liu, B.S., and Dong, D.P. 2007. Analysis of critical excavation depth for a jointed rock slope using a face-to-face discrete element method. Rock Mechanics and Rock Engineering, 40(4): 331-348. Liu, M.D., and Carter, J.P. 1999. Virgin compression of structured soils. Geotechnique, 49(1): 43-57. Lo, K.Y. 1970. The operational strength of fissured clays. Geotechnique, 20(1): 57-74. Lo, K.Y. 1972. An approach to the problem of progressive failure. Canadian Geotechnical Journal, 9: 407-429. Lo, K.Y., and Milligan, V. 1967. Shear strength properties of two stratified clays. American Society of Civil Engineers Proceedings, Journal of the Soil Mechanics and Foundations Division American Society of Civil Engineers, 93(SM1): 1-15. Lo, K.Y., and Morin, J.P. 1972. Strength anisotropy and time effects of two sensitive clays. Canadian Geotechnical Journal, 9(3): 261-277. Lo, K.Y., and Hinchberger, S.D. 2006. Stability analysis accounting for macroscopic and microscopic structures in clays. In Proc. 4th International Conference on Soft Soil Engineering. Vancouver, Canada, pp. pp. 3-34.
Locat, J., and Lefebvre, G. 1985. Laboratory investigations on the lime stabilisation of sensitive clays. In Proc. 40th Can. Geotech. Conf. Regina, pp. 121-130. Malandraki, V., and Toll, D. 2000. Drained probing triaxial tests on a weakly bonded artificial soil. Geotechnique, 50(2): 141-151. Mesri, G., Rokhsar, A., and Bohor, B.F. 1975. Composition and compressibility of typical samples of Mexico city clay. Geotechnique, 25(3): 527-554. Mitchell, J.K. 1976. Foundamental of soil behavioiur. Wiley, New York. Norton, F.H. 1929. The Creep of Steel at High Temperature. McGraw-Hill Book Co., New York. Perzyna, P. 1963. Constitutive equations for rate sensitive plastic materials. Quarterly of Applied Mathematics, 20(4): 321-332. Philibert, A. 1976. Etude de la resistance au cisaillement d'une argile Champlain. M.Sc. Thesis, Universite de Sherbrooke, Quebec. Qu, G., and Hinchberger, S.D. 2007. Evaluation of the viscous behaviour of natural clay using a generalized viscoplastic theory. Geotechnique, submitted. Robertson, P.K. 1975. Strain rate behaviour of Saint-Jean-Vianney clay. Ph.D Thesis, University of British Columbia, British Columbia,Canada. Rocchi, G., Fontana, M., and Da Prat, M. 2003. Modelling of natural soft clay destruction processes using viscoplasticity theory. Geotechnique, 53(8): 729-745. Roscoe, K.H., Schofield, A.N., and Thurairajah, A. 1963. Yielding of clays in states wetter than critical. Geotechnique, 13(3): 211-240. Rouainia, M., and Wood, D.M. 2000. Kinematic hardening constitutive model for natural clays with loss of structure. Geotechnique, 50(2): 153-164.
Rowe, R.K., and Hinchberger, S.D. 1998. The significance of rate effects in modelling the Sackville test embankment. Canadian Geotechnical Journal, 35(3): 500-516. Saihi, F., Leroueil, S., La Rochelle, P., and French, I. 2002. Behaviour of the stiff and sensitive Saint-Jean-Vianney clay in intact, destructed, and remoulded conditions. Canadian Geotechnical Journal, 39(5): 1075-1087. Sekiguchi, H. 1984. Theory of undrained creep rupture of normally consolidated clay based on elasto-viscoplasticity. Soils and Foundations, 24(1): 129-147. Silvestri, V. 1984. The preconsolidation pressure of Champlain clay, Part II. Canadian Geotechnical Journal, 21(3): 600-602. Sorensen, K.K., Baudet, B.A., and Simpson, B. 2007. Influence of structure on the timedependent behaviour of a stiff sedimentary clay. Geotechnique, 57(1): 113-124. Suklje, L. 1957. The analysis of the consolidation process by the isotache method. In Proc. 4th Int. Conf. on Soil Mech. and Foun. Engen. London, Vol.1. Tavenas, F., Leroueil, S., La Rochelle, P., and Roy, M. 1978. Creep behaviour of an undisturbed lightly overconsolidated clay. Canadian Geotechnical Journal, 15(3): 402-423. Vaid, Y.P., Robertson, P.K., and Campanella, R.G. 1979. Strain rate behaviour of SaintJean-Vianney clay. Canadian Geotechnical Journal, 16(1): 35-42.
106 Table 3.1 Properties of Saint-Jean-Vianney clay, (after Vaid et al. 1979)
Liquid Limit Plastic Limit Plasticity index Natural water content Degree of saturation Specific gravity of solid Percent finer than 2um Unconfined compressive strength (test duration approx. 3min) Sensitivity Activity P.I.(%<2um)
36% 20% 16% 42% 100% 2.75 50% 640kPa Approx. 100 32%
Table 3.2
Constitutive parameters for Saint-Jean-Vianney clay
Parameters Initial Structure, Q)Q Destructuration Parameter, b Weighting Parameter, A Power Law Exponent, n Structured Fluidity, yvsp, (min" 1 ) Intrinsic Fluidity, yjp, (min -1 ) Aspect Ratio of Elliptical Cap, R Moc Mv
C
Clay 1.68 4000 (120) 0.5 22 l.OxlO"10 9-lxlO" 6 0.7 1.34 0.48 0.011 0 0.33 1.15 0.02 0.26
cs
Poisson's Ratio, v Initial Void Ratio, e0 Recompression Index, K Compression Index, X
* The damage exponent, b , was 120 for K'0 compression and 4000 for triaxial compression
108 Figure 3.1 The influence of structure on the response of Bothkennar clay during oedometer compression (from Burland 1990[G40]).
2.2
2.0
1.8 1.6 - Attributed to structure O 'a \f^)/A \
en
1-t
>
o O
1-2 1.0 " BothKennar clay from 6.5m depth
\N.
In situ state ICL (Intrinsic compression line)
\ ,
10
100
1000
10000
Vertical Effective Stress, o ' v , kPa
109 ure 3.2 The influence of structure on the response of London clay during undrained triaxial compression (from Sorensen et al. 2007 and Hinchberger and Qu 2007[G41]).
a,a = 0.05% lh = 8.3 - 6/ min

i 1 1 1 1 1 1 1 1 1 1 1 1 1 i
0.0
.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
Axial Strain, %
110 Figure 3.3 The state boundary surface, critical state line, and mathematical overstress of the structured soil model [G42].
Triaxial limit
Critical state line
Resultant isotaches of constant (/>{F)
\ \ \
Associated flow rule \ \ \
\ V -*-4 Dynamic yield surface /(d) \ l (typical) \

'(d) my
_'(J)
my
Ill Figure 3.4 Estimation of the aspect ratio, R , for the elliptical cap[G43].
nrj-
1UUU
(kPa) 900 800 700 600 500 400 300 200 100 0 -r
i i T
T / ~-^S3'^
Crjtical state line M=1.34 :
y/
Elliptical cap with an aspect ratio of 0.7
^^Q
-i
- \r
"i
100
200
300
400
500
600
700
800
900
1000
<ym , kPa
112 3.5 Estimation of the yield surface parameter, M o c , in the overconsolidated stress range[G44].
uuu -
Triaxial lirnit 900 800 700

J
yCr\t\ca\ state line / ' ' \ Mcs=1.34
/ The slope of the Lin)K/state line
" - - - - ^ - - D y n a m i c yield surface 600 500 400 300 200 100 0/ IW
) )
/ uî
I/'X
i
1 200 300
: 3 C
Undrained triaxial tests i - O - on intact SJV clay \ (Saihi et al. 2002) \
>X 100
-i( 400 500
V-
\-wi
600
700
800
900 1000 cxm , kPa
Figure 3.6
Estimation of the intrinsic compressibility, X, and structure parameter, co 0 , from oedometer compression for SJV clay[G45].
0.0
=*?=
0.1 4
Estimated
to
CO
0.2 4
IntFinsic Gdmpression line k=0.26
a:
o >
a> o> c CO x: O
g
0.3 4
Deviation due to natural variation
0.4 4
0.5 4
1.68x10"^%/min
Data from Vaidetal. 1979 0.6

! ,
2
3.48x10"%/min 1.17x10" 3 %/min
e"
e^1
e1
Normalized Vertical Effective Stress, c' v /a' p (kPa), in natural log scale
114 Figure 3.7 Intrinsic compressibility of different clays (adapted from Burland
LL: Liquit Limit LL=109
St. Jean Vianney clay

St. Andrew clay Oslofjord Ocean cores Tilbury Alvangen G. of mexico Gosport Pisa clay Avonmouth Drammen Grangemou Drammen Detroit Milazzo S. Joaqui Milazzo Po Valley
4H
LL=80
Assumed A,=0.26 for St. Jean Viartney cla' with LL=35
3H
LL=63
g
-Q-
2 2H
LL=46
-*-
\ A L L = 6 4
=--11=62
LL=4oS5*fe
- 1
-
10"
10
10
102
103
104
105
Vertical Effective Stress, a' , kPa
115 Figure 3.8 Estimation of n anda'my from undrained triaxial compression and oedometer compression for SJV clay (a) Peak Undrained Shear Strength - Upper layer[G47].
...... Measured Peak Strength {Upper Layer) 1000 900 Data from Vaid et al. 1979 800 | 700 600 500
05 Q.
400 h 300
'2 J , = 224kPa Note: From Figure 3.3, at Point A:
200 -
r ;''
= lxlO"" ) min"'
I I
100
At
1Q-12
1Q-11 1Q-10
10-9
1Q-8
10-7
1Q-6
1Q-5
1Q-4
-| Q 3
10"2
10" 1
10
Strain Rate (/min)
116 Figure 3.8 Estimation of n ando'^ from undrained triaxial compression and oedometer compression for SJV clay (cont.) (b) Peak Undrained Shear Strength - Lower layer.
1000 900
Measured Peak Strength (Lower Layer)
800 700 600

500
(kPa)
CM
400
CM
300
200
Note: From Figure 3.3, at Point A:
\rlp =1x10-' din"

100 -10-12 10" 8 10"7 10"6 10"5 10"4 10" 3 10"2 10"1
-| Q-11
1 0 -io
<| Q-9
10
Strain Rate (/min)
117 Figure 3.8 Estimation of n anda'^ y from undrained triaxial compression and oedometer compression for SJV clay(Cont) (c) Preconsolidation pressure versus strain-rate.
a=0.045 n=1/a=22
o-'1,:' = 52UPa <T"" = 45lkPa
a"''. = 0.796CT'''' = 417 kPa

Upper Layer]
o-f,! = 0.796er"" = 358/tPu
Measured apparent preconsolidation pressures (Vaidetal. 1979)
X;"=lxlO""min
J I I L.
-10-12
1 0 -n
-\Q-W 10'9
10"8
10"7
10"6
10"5
10"4
10-3
10"2
10"1
10
Strain Rate (/min)
118 Figure 3.9 Influence of continued post-peak straining on the power law exponent, n.
(a) Stress-strain behaviour for Belfast clay and Winnipeg clay during CU tests
0.7 CT1C: Confining pressure.kPa Axial strain rate = 5%/h 0.5%/h
Winnipeg clay (Graham, et al. 1983) o.i H 0.0 10 15 20 25 Axial Strain, % 30
(b) Relationshi[G48]p between axial strain and n
100
CD test on CU test on CU test on CU test on CU test on undisturbed London sample i undisturbed London sample reconsituted London sample J undisturbed Belfast clay undisturbed Winnipeg clay Data from Sorensen et al. (2007) and Hinchberger and Qu (2007) Data from Figure 3.2 Data from Figure 3.9(a)
80
c C D c
o a. x
60
Best fit n=43 for London clay
c
i
40
Best fit n=29 for Winnipeg clay
20
Axial Strain,%
119 Figure 3.10 Theoretical behaviour of the structured soil model during CIU triaxial compression[G49].
1
(b)
1
. Triaxial Limit 14* Critical State Line
,**Jtry 6 /
/
^^
>
. X-<Dynamic Yield
2/
^vj( Initial Static Yield Surface Slight Contraction Due to uft *Iegati 1
IY
Vertical Strata v
1
11
J
1
u
my6 ^ my
my6
(c)
lg(<0)
Slope is governed by the parameter b
Destructured State
Vertical Straia e
120 Figure 3.11 (a) Measured and calculated behaviour of SJV clay during CIU triaxial compression, Upper layer
700 I
Axial Strain (%)
Axial Strain (%)
Figure 3.11 Measured and calculated behaviour of SJV clay during CIU triaxial compression. (Cont.) (b) Lower layer[G50]
800 700
(0
0.
600 500 400
<D
55
Q
300 200
100 1,
2
>
<D Q
Measured - 2.8x10"1%/mirt Measured - 2.0x10"2%/min Measured - 7.2x10"4%/min Calculated -2.8x10"1%/mir Calculated - 2.0x10_2%/mii Calculated - 7.2x10"4%/mii 1.0 1.5 2.0 Axial Strain (%) 2.5 3.0
0.5
1.0
1.5 Axial Strain (%)
2.0
2.5
3.0
Figure 3.12
Measured and calculated undrained shear strength versus strain-rate for SJVclay[G51].
2000
C O
Q _
-*
' 1000 900 800 700 600 500 400 300 200
CO
Calculated Measured Calculated Measured
Peak and Peak and Peak and Peak and
Residual Strength Residual Strength Residual Strength Residual Strength
(Lower: Layer) (Lower: Layer) (Upper; Layer) (Upper; Layer)
~b "c
-t
* C O
i
C D
iJ-Largeistrain Strength (3%)
C O C D
C O
T3 C D
'c0
^T3 C Z) 100 0.0001
0.001
0.01
0.1
10
Strain Rate (%/min)
Figure 3.13
Calculated and measured behaviour during CIU triaxial creep tests on SJV clay
(a)
Upper layer[G52]
1.5-
ji i
ii [i
I
i
I
ii
;
Stress Level=430KPa n Rtrpss Level=470KPa J *. _ Stress I evel=47nKPa Stress Level=430KPa -i
"= = ~,, ,ifj
M e a s u r e d ,, a :>ul u
stress Levei=470KPa
Calculated
35
xial Strain,
1.0-
ii
ii
<
.5-
[ ._
>i
j
l
fr
1
ii
1
1
!
1
on-
100 Time (min)
1000
10000
_*
. ** - <
20-
I !
t I|
N
!|
0-
n
j
! 1
20-
1
1
1
1
!
1
!
!
i
1
40-
1 i i
10000
1 100 Time (min)
Figure 3.13 Calculated and measured behaviour during CIU triaxial creep tests on SJV clay (b) Lower layer[G53]
Stress Level=630KPa Stress Level=575KPa n Stress Level=550KPa Stress Level=630KPa Stress Level=575KPa ^ ^ Stress Level=550KPa
Measured Calculated
10000
125 Figure 3.14 Calculated and measured creep-rupture life for SJV clay[G54]
650
600 H
550 H
500
f A -V -A
Upper Layer Lower Layer Upper Layer Lower Layer
Measured'
Df Calculated
450 -\
400 10 100 1000 10000
Rupture Life, tf, (min)
126 Figure 3.15 Calculated and measured axial strain-rate versus time during CIU triaxial creep on SJV clay.
(a) Measured
(b) Calculated
I i :070.1 550-
10000
Figure 3.16
Comparison of strain-rate at failure for peak strength and creep rupture SJV clay,
(a)
Upper layer
900 800 700 jk. # Measured Critical Creep Rate In Creep Tests Measured Strain Rate at Failure for Triaxial Shear Tests Theoretical
1e-7
1e-6
1e-5
1e-4
1e-3
1e-2
Strain Rate /min
(b)
Lower layer
900
A # 800 Measured Critical Creep Rate In Creep Tests Measured Strain Rate at Failure for Triaxial Shear Tests Theoretical
700
600
500
For Lower Layer Clay

1e-6
1e-5 1e-4
1e-7
1e-3
1e-2
Strain Rate /min
Figure 3.17
Theoretical behaviour of the structured soil model during constant-rateof-strain K^ -consolidation.
Elasti
BasticZone
Vlscoplastic Zone
am Points: 1. Starting Stress State
3^5
\ \
v\
Overstress Created by Structure, <a
2. Static Yield Point 3. Apparent Yield (No Structure)
v
\ \
\ 6
4. Apparent Yield (Structured Soil) 5. Stress State Adjusted To The Plastic Potential 6. End of Destructuratton
(c)
Long Term Compression Curve
Ae.
CD ^ Elastic Zone (Or
\7
Viscoplastic Zone 4,5

L1,2,3,4,5
(d)
1.0L
Structured Zone I Non-structured Zone
129 Figure 3.18 (a) Calculated and measured behaviour during oedometer compression[G55]. Normalized compression curves.
Normalized Vertical Effective Stress, <f /o' kPa, in log scale 0.1 1
1 1 1
5
I 1 1
0.0
Structure effect
0.1
Estimated; * 0.2 ^
v%
X yr * J& Intrinsic compression line
Theoretical (Structured mode \
0.3
0.4 -
.
0.5
6.78x10" 2 %/min 1.68x10" 2 %/min 3.48x10" 3 %/min 1.17x10" 3 %/min
_ .
Measrued data from Vaid et al. 1979!
,
o
0.6
(b)
Variation of compressibility (C c ) during loading
2.0
Calculated Cc variation Measured Cc variation under the loading rate of 1.68 x 10"2%/min
1.5
o
C O
o O
1.0
0.5 A
0.0
500
1000
1500
2000
2500
Vertical Effective Stress, c'v, kPa
gure 3.19 Measured and calculated compression curves of SJV clay during constant rate-of-strain consolidation[G56].
100
1000
Vertical Effective Stress, a'v, kPa
131
CHAPTER 4
THE STUDY OF STRUCTURE AND ITS DEGRADATION ON THE BEHAVIOUR OF THE GLOUCESTER TEST EMBANKMENT
4.1 Introduction
Staged construction is often used to build embankments founded on soft clay deposits. This construction method can be used to improve the stability of embankments during construction by permitting time for excess pore pressure dissipation and consolidation to occur, leading to increased strength in most cases. However, for some embankments built on clay, staged construction has failed to produce significant strength gain even after large vertical settlements and several years of consolidation (Holtz and Broms 1972 and Stermac et al. 1967). Such behaviour has been attributed to strength loss due to 'disturbance' or 'yielding' of the clay, which exceeds that normally expected due to consolidation and reduction of void ratio. This phenomenon is thought to be due to the breakdown of microstructure, which is a combination of fabric effects and cementation bonding between clay particles (Mitchell 1976). Many natural clays have microstructure to some extent (here after referred to as structure). Structure can have a pronounced impact on the stress-strain-time response of clay and the performance of embankments built on those deposits of clay. As discussed in Chapter 3, structure has two important effects: (i) it imparts additional strength above and beyond that normally anticipated from conventional soil mechanics models and (ii) it permits clay to exist at higher void ratios than expected for the corresponding
132 'unstructured' or reconstituted material. With respect to structured clay, only Karstunen et al. (2005) has studied the influence of clay structure on the performance of embankments (the Muro test embankment). However, most structured clays are very rate sensitive and Karstunen et al. (2005) undertook their study using a rate-independent elastic plastic constitutive model. As such, an analysis that takes into account both the time dependent response of structured clay and the influence of structure on the performance of a test embankment should be of interest to geotechnical engineers. In this Chapter, the Gloucester test embankment (Bozozuk and Leonards 1972) is examined using the finite element method with a constitutive model that accounts for the viscous behaviour of Gloucester clay and its structure (see Chapter 3). The primary objective of this chapter is to evaluate the impact of structure on the long-term settlement of the Gloucester test embankment. The second objective is to numerically study the distribution of 'destructuration' in the Gloucester foundation clay with time and to compare the strength loss due to 'destructuration' to the strength increase that would normally be expected due to compression and decreasing void ratio. The results of this study should provide insight into the response of natural clay deposits to embankment construction and/or staged loading.
4.2 Background
Gloucester test embankment The Gloucester test embankment is located on a site near the southern outskirts of Ottawa, Canada. The site is commonly referred to as Canadian Geotechnical Research Site No. 1 at Gloucester. McRostie and Crawford (2001) recently summarized the
history of this site and the historic research activities undertaken there, including the
133 Gloucester test embankment and performance monitoring of an Accommodation building. Stage 1 of the Gloucester test embankment was constructed in 1967. After 15 years, a second stage was constructed beginning in June, 1982. Prior to construction of the Stage 1 embankment, the upper 1.2m of the foundation deposit was excavated to minimize the influence of the upper fissured clay. Then both stages 1 and 2 were built within the excavation. Figure 4.1(a) shows the geometry of Gloucester test embankment. The Gloucester test embankment was heavily instrumented and well documented (Bozozuk and Leonards 1972; Lo et al. 1976; Fisher et al. 1982), as shown in Figure 4.1(a). However, only the centre settlement gauges (SI and S3) shown in Figure 4.1(a) are considered in the following discussion. The Gloucester foundation comprises an upper soft clay layer and a lower medium to stiff clay layer (see Figure 4.1b). The upper clay layer extends from a depth of 0 to 4m and comprises grey-brown, oxidized clay that is occasionally stratified with silt. The average preconsolidation pressure in this layer is about 60kPa. Below that, the lower layer extends from 4m to 19m and possesses relatively high preconsolidation pressures, ranging from 80 kPa to 170 kPa. Finally, this layer is underlain with varved clay and glacial till (Bozozuk and Leonards 1972; Lo et al. 1976; Leroueil et al. 1983). The performance of the Gloucester test embankment has been studied by several researchers. Fisher[i57] et al. (1982) predicted the field behaviour prior the construction of the Stage 2 fill using an elastoplastic finite element model and the Gibson and Lo (1967) model, and incorporating some engineering judgment. Lo et al. (1976) simulated the performance of the Stage 1 fill, also using the Gibson and Lo (1967) model. Finally, Hinchberger and Rowe (1998) utilized an elasto-viscoplastic constitutive model to
134 simulate both Stages 1 and 2 of the Gloucester case. In spite of past efforts to model the Gloucester test embankment, so far the case has not been studied using a constitutive model that accounts for the impact of structure and destructuration on the response of the Gloucester clay.
Clay structure Figures 4.2a and 4.2b illustrate the influence of clay structure on the behaviour of the Gloucester clay during undrained triaxial compression and oedometer compression tests (Law 1974 and Leroueil et al. 1983). Figure 4.2a shows the stress-strain curves for two specimens consolidated isotropically using different confining stresses (<T'C = 83kPa and a'c = AOkPa). The apparent preconsolidation pressure of these two specimens[i58] obtained from a depth of 2.4m is about 60 kPa (Lo et al. 1976). In Figure 4.2a, the specimen consolidated to stresses less than & (<J\ = 40kPa <60kPa) exhibits a peak strength and subsequent reduction of strength with continued compression. When the consolidation pressure exceeds the preconsolidation pressure of the clay
(cr'c = 83kPa >60kPa), the clay structure is degraded during the consolidation process and the specimen exhibits a strain-hardening response. The shaded area in Figure 4.2a can be attributed to bonding and fabric or more generally structure, assuming the intrinsic or fully destructurated state is reached during CRS triaxial compression (Such may not always be the case). Figure 4.2b shows a typical compression curve from an oedometer compression test[i59] on Gloucester clay (Leroueil et al. 1983). After yielding, high compressibility is observed and with further straining, the compression index Cc reduces
135 and becomes approximately constant for stresses exceeding Point A in Figure 4.2b. The compression curve at large-strain in Figure 4.2b is essentially the intrinsic compression line (Burland 1990). These characteristics of Gloucester clay are consistent with the studies performed on other natural clays by Kabbaj et al. (1988), Burland (1990), and Leroueil and Vaughan[i60] (1990). Rate-sensitivity The undrained shear strength and preconsolidation pressure of Gloucester clay are both rate-sensitive. Laboratory data from both Leroueil et al. (1983) and Law (1974) effectively characterize the rate-sensitivity of Gloucester clay. Figures 4.3a and 4.3b show that the undrained shear strength and preconsolidation pressure versus strain rate is essentially linear in a log-log plot. For an order change in strain rate, both the strength and preconsolidation pressure vary by about 8%. A straight line can be fit through the data with a slope a = 0.033. It should also be noted that the slope of the log-log plots is consistent for both cr'p and Su.
Secondary compression Lo et al.(1976) presented the results of long-term oedometer creep tests on undisturbed Gloucester clay using a Rowe Cell. The data is reproduced in Figure 4.4, which shows that more than half of the total compression during these tests occurred after complete dissipation (see Aw=0) of the excess pore pressures. Such behaviour is
commonly referred to as secondary compression or delayed compression (Bjerrum 1967) and is clearly an important characteristic of Gloucester clay. Summary
136 As discussed above, Gloucester clay is a structured, rate-sensitive, and timedependent natural clay. From the data presented above, the structure and rate-sensitivity are significant characteristics of Gloucester clay and ideally both characteristics should be accounted for in the analysis of this case. Below, a structured EVP model is used to investigate the influence of structure and rate-sensitivity on the performance of the Gloucester test embankment.
4.3 Methodology
This chapter uses an unstructured elastic-viscoplastic constitutive model (Hinchberger and Rowe 1998) and a structured elastic-viscoplastic constitutive model (see Chapter 3) to examine the performance of the Gloucester test embankment. Both models are described in detail elsewhere. The following is a brief summary of both models and their constitutive parameters. 4.3.1 Model 1 -Hinchberger and Rowe Model Hinchberger and Rowe (1998) developed an unstructured EVP model, which hereafter is referred to as Model 1 in this chapter. Model 1 is an overstress elasticviscoplastic (EVP) model based on Perzyna's theory of overstress viscoplasticity (Perzyna 1963 and 1971). The constitutive equation is:
e = ,J +? =^-+^^j-8u e* ,J ,J y
2G 3(l + e)(r'm
+ejf
"
2G
3(1+ e)tr'
=+-J^~Sii+yvpX (<l)(F)) J
'
dF
[4.1]
137
my
-1
a' ( J > >(7' ( l )

^ my ^ my
*(/) =
^ my
[4.2]
o <o'
my
,(d)
{s)
my
where stj is the deviatoric stress tensor, a'm is the mean effective stress, dtj is Kronecker's delta, G is the stress dependent shear modulus, K is the slope of the e - l n ( o ^ ) curve in the over-consolidated stress range, yvp is the fluidity parameter, a( = l/n) is rate-sensitivity parameter, n = \la is the power law exponent, t r ^ i s the
intercept of static yield surface with mean stress axis, o'^ is the intercept of dynamic dF
yield surface, and a ^ / a ^
is the overstress ratio.
is the plastic potential,
which is derived as a unit norm vector. The flow function, <j>{F), in Equation [4.2] is a power law function. Figure 4.5 shows the state boundary surface for the Hinchberger and Rowe (1998) model. In the normally consolidated stress range, the state boundary surface is defined using an Elliptical cap equation: / =K-/)2+2J2/?2-(<>-/)2=0 [4.3]
where / and R are parameters defining the aspect ratio of the elliptical cap, and J2 is the second invariant of the deviatoric stress tensor, s;i. In the overconsolidated stress range, the Drucker-Prager equation is used as: f=MocG'm+c'-j2T2=0
[4.4]
where Moc is the slope of the Drucker-Prager envelope and c is the effective cohesion
138 intercept in -yj2J2 - o'm stress space. The resultant state boundary surface, which
delineates elastic and viscoplastic stress states, is illustrated by A1-A2-A3 in Figure 4.5. The kinematic strain-hardening of the state boundary surface occurs according to: ^=^-<êZ
AK
[4.5]
where do'^
is the incremental expansion of the state-boundary surface, e is void ratio,
A and K are the compression and recompression indices, and devvpol is the incremental plastic volumetric strain. In accordance with overstress viscoplastic theory, stress states are permitted to exceed the state boundary surface of the soil. Point G in Figure 4.5 shows a typical state of overstress.
a
In accordance with Equation [4.1], the overstress ratio at Point G is
mv} I awmv} a n ^ m e resultant plastic strain-rate tensor is: my my

mv
.
a>U) w
ejf = fp
1
,
my
dF do':V
[4-6]
Lastly, the Drucker-Prager equation is also used to define the critical state for the Hinchberger and Rowe model (1998) viz.:
f=Mcsa'm -J2T2=0
[4.7]
where Mcs is the slope of the classic critical state line (see Figure 4.5). The elliptical cap is defined such that the top of the cap coincides with the critical state line. In summary, there are nine constitutive parameters required for this model. Each of the constitutive parameters, e0, A , K , Moc, Mcs, a'^ , R, yvp, and a , can be derived
from standard laboratory tests as described in Chapter 2.
139 Figure 4.6 illustrates the constitutive response of Model 1 during CRS (constant rateof-strain) oedometer and CRS undrained triaxial compression tests. In accordance with EVP theory, Model 1 gives classic idealized compression behaviour during oedometer compression. The compressibility is governed by the recompression K and compression X indices in the overconsolidated (elastic) and normally consolidated (viscoplastic) stress ranges, respectively. In addition, the preconsolidation pressure, o'p , is rate-sensitive. Since a power law flow function is used in the Hinchberger and Rowe (1998) model (see (j)(F)in Equation [4.2] ), there is a linear relationship between log(a')-log(e) with a slope of a. The static yield surface intercept, a'^J, and fluidity parameter, y vp , define
the viscous range of the material. For strain-rates less than yvp (e.g. eMial < y v p ), the model becomes a rate-independent elastic-plastic model and the preconsolidation pressure is rate-insensitive. The model becomes viscous and rate-sensitive for strainrates greater than yvp (see also Chapter 2). Similar behaviour occurs during CRS triaxial compression. For strain-rates less
thany vp , the material response is rate-independent and there is a classic elastic-plastic strain-hardening stress-strain response. For strain-rates exceeding y v p , the undrained shear strength is rate-sensitive and there is a linear relationship between log(Su) - log(e) with a slope of a. 4.3.2 Model 2 - Structured Elastic-viscoplastic (EVP) Model The Gloucester test embankment was also analyzed in this study using the structured EVP model described in Chapter 3 (Model 2). The following is a brief
140 summary of the model and the additional parameters required. Model 2 is an extended Hinchberger and Rowe (1998) model. In this model, structure is defined by:
where oo0 is the initial structure, cc = l / n ( n is the power law exponent in Equation [4.2]), yj p is the fluidity of the undisturbed structured clay, Y^p is the fluidity of the corresponding remoulded or destructured clay, and c ' and G'
{
are the structured and
intrinsic preconsolidation pressures as defined in Figure 4.2. The structured fluidity, yj p , is much lower than the destructured fluidity, yfp, and consequently co0 is greater than one. To simulate destructuration, the structure parameter is assumed to be a function of plastic strain viz.: aKd)=[l + ((O0n-1.0)e-b^ and ed=\'o ^\-A){deZl)2+A(de:P)2dt [4.10] [4.9]
where ed is the plastic damage strain, e ^ and ej p are plastic volumetric and octahedral shear strains, A is a weighting parameter (assumed to be 0.5), and b is a parameter that controls the rate-of-destructuration and the magnitude of plastic damage strain required to reach the intrinsic state. This damage law was first proposed by Rouainia and Wood (2000) for use with a rate-insensitive elastoplastic model. The resultant viscoplastic strain-rate tensor of Model 2 is:
141 vvp ?F
where (f)(F) = \<j'^) lo'^J
- 1 is a power law flow function and the term a = \ln
defines the slope of the S u and o'p versus strain-rate in a log-log scale. On inspection of Equations [4.8] through [4.11], the clay fluidity term in Equation [4.11], yl" l[co{ed)], increases from the initial value, yvsp, and it approaches yjp with large plastic damage strain. As noted above, the magnitude of plastic damage strain required to reach the destructured state is governed by b in Equation [4.9]. Figure 4.7 shows the response of Model 2 during CRS oedometer and CRS triaxial compression tests. Similar to Model 1, Model 2 implies linear log(S u )-log(e) behaviour and linear log(a' ) - log(e) behaviour. These relationships apply to both the undisturbed structured state and the intrinsic or destructured state. The initially low structured fluidity, yj p , imparts additional strength to the clay above the post-peak strength; whereas the post-peak strength is mobilized at large-strain or in the intrinsic state. In Model 2, the additional strength is attributed to viscous effect[i61]s and is denoted by the shaded area in Figure 4.7(c), which represents additional overs tress caused by the initially low structured fluidity, yj p . Similarly, during the CRS oedometer compression, the initially low structured fluidity, yjp permits the clay to exist at a higher void ratio after yielding than would otherwise be expected for the corresponding destructured or remoulded clay. The shaded area in Figure 4.7(a) is additional overstress due to yj p (structure). As a consequence of using EVP theory, the structured state in both
142 CRS oedometer and CRS triaxial compression tests is metastable, since the fluidity will increase with plastic damage strain and eventually reach the intrinsic fluidity as a consequence of the soil being in a state of overstress. The above discussions have described the constitutive models used to examine the performance of the Gloucester test fill. Model 1 (Hinchberger and Rowe 1998) has 11 constitutive parameters; whereas, Model 2 (the Structured EVP model) has 13 constitutive parameters. The following sections describe the finite element mesh and constitutive parameters used to re-examine this case. 4.3.3 Finite Element Mesh The Gloucester test embankment is examined using finite element analysis. The corresponding finite element mesh, shown in Figure 4.1a, consisted of 800 six-noded plane-strain triangles and 1702 nodes. A rigid boundary condition was set at a depth of 20.2m and smooth boundaries are assumed at the embankment centerline and 75m beyond the centerline. The construction process for Stages 1 and 2 was simulated by adding the corresponding elements representing the embankment fill, layer by layer, into the finite element mesh and incrementally increasing the body force due to gravity within the added elements. 4.3.4 Constitutive Parameters The constitutive parameters for Gloucester clay have been summarized in Table 4.1 and 4.2. The elastoplastic parameters, e0, v , A , K, Mcs, Moc , and R , were estimated by Hinchberger (1996), from standard laboratory tests provided by the National Research Council of Canada (NRC). The hydraulic conductivity, k, of the Gloucester foundation soil was assumed to be dependent on the void ratio as follows:
143
k = k0cxp(^-fi-)
[4.12]
where k0 is the hydraulic conductivity at the reference void ratio, and Ck is the slope of the e - log(fc) plot. The elastoplastic constitutive parameters are summarized in Table 4.1 and compared with laboratory results in Figure 4.1b. The same elastoplastic parameters were used with Models 1 and 2. Table 4.2 summarizes the viscosity-related parameters. The parameters, yvp and n = \la , required in Model 1 for Gloucester clay were directly adopted from the study by Hinchberger and Rowe (1998). The structure-dependent parameters (y]p, yjp, co0 and b) required in Model 2 were estimated using a similar approach described in Chapter 3. As shown in Figure 4.2, the structure parameter is coo =<7'p-s / V , =1.18, which was estimated from the intrinsic, a* {, and structured, <y'p_s, preconsolidation pressures in oedometer compression on intact Gloucester clay specimens (Leroueil et al. 1983). For analysis purpose, the structured fluidity, yvsp , was taken to be lxlO^min" 1 (from Hinchberger and Rowe 1998). The intrinsic fluidity yvp is an inter-dependent parameter, which can be calculated using Equation [4.8]. The destructuration-rate parameter, b [i62], was estimated using Equation [4.9] and to the magnitude of plastic strain required to reach the intrinsic state during oedometer compression (see Point A in Figure 4.2). Figure 4.8 compares the theoretical response of Model 2 during CRS oedometer compression test with the actual response of Gloucester clay obtained at depth between 3.45m and 3.90m (Leroueil et al. 1983). The embankment fill was modeled as a Mohr-Coulomb material with an effective
144 friction angle <> = 35, a dilation angle y/ = 0, c'=0kPa, and a unit weight of 18.4 /' kN/m3.
4.4 Results
This section presents the results of finite element analysis of the Gloucester test embankment. Calculated long-term settlements using Model 1 and Model 2 are
compared with the measured data to assess the relative importance of structure in the Gloucester case. In addition, the calculated change of strength in the Gloucester
foundation is studied. In most cases, strength gain due to consolidation is relied on during the design of staged embankment. However, for natural clay, destructuration due to plastic strain in the clay foundation may result in strength loss. In situ tests such as field vane and cone penetration tests are typically used to investigate changes in the undrained strength of foundation deposits to verify the design assumption (Stermac et al. 1967, Holtz and Broms 1972, and Koskinen et al. 2002). In the following sections, the distribution of strength gain due to consolidation and strength loss due to destructuration are studied with the intent to gain insight into the effect of structure and destructuration on the staged construction.
4.4.1 Analysis using the Unstructured EVP Model (Model 1) Settlement versus time response Figure 4.9 shows the measured long-term settlement at Settlement Gauge SI beneath the center of the Gloucester test embankment (see Figure 4.1 for the location). In addition, the estimated time required for the primary consolidation (1.5[G63][G64] years) is also shown in Figure 4.9 as reported by Lo (1976). From Figure 4.9, it can be seen that
145 a significant amount of the total settlement of Stage 1 (34cm) can be attributed to secondary compression (20cm +/-). With further load of the Stage 2 fill, the settlement increased from 34cm up to about 70cm. The calculated settlements versus time using Model 1 are presented in Figure 4.9. Two types of theoretical curves have been generated using: (i) a linear virgin compression curve, and (ii) bilinear virgin compression curves, as shown in Figures 4.10(a) and 4.10(b). For the bilinear calculations, the clay was assumed to be more compressible immediately after reaching the preconsolidation pressure and less compressible at stresses of either 20 or 40 kPa higher than the preconsolidation pressure. Such an approach can approximately account for destructuration in compression. From Figure 4.9, it can be seen that Model 1 gives calculated behaviour that is in reasonable agreement with measured behaviour for Stage 1. For Stage 2, however, the use of a linear virgin compression curve (see Figure 4.10a) leads to calculated settlements significantly higher than those measured in Stage 2. Good agreement between calculated and measured settlements can be obtained using a bilinear virgin compression curve with a transition stress of 20kPa (see Figure 4.10b). Although a bilinear virgin compression curve is common for structured clay (see Figure 4.2), the transition of compression index from X to X12 occurs at stresses about 60kPa higher than the preconsolidation pressure (see Point A in Figure 4.2). Consequently, the bi-linear curve required to fit the field response is not entirely consistent with the actual compression response of Gloucester clay. Strength Increase Due To Consolidation One of the important objectives of staged construction is to improve the strength
146 of the clay deposit by allowing time for consolidation and consequent decrease in void ratio. Here, the term 'consolidation' includes both primary and secondary consolidation. Studies on remoulded clay have shown that the decrease in void ratio during consolidation leads to expanding of yield surface and a corresponding increase in undrained shear strength. In addition, Bjerrum and Lo (1963) have shown that the
undrained strength increased with the period of aging as well. Thus, it was considered to be insightful to examine the relative strength increase in the Gloucester foundation caused by consolidation. Model 1 uses a classical critical state concept in which consolidation and the resultant expansion of the yield surface (see Equation [4.9]) causes strength increase. Thus, the undrained shear strength, S u , is a function of void ratio and stress history. From the elliptical cap equation, it can be shown that the initial undrained strength prior to Stage 1 constructing, Su0, is related to (j'^l viz.: SB0=AOi65] [4.13] where A is a constant and depends on the aspect ratio of the yield surface, and a ^ is the initial yield surface intercept prior to the construction in Stage 1 (see Figure 4.1b). In addition, the relative increase in undrained strength due to hardening during consolidation is:
r
S
u
>i
JCon
mv
o'^+do'}*
o-' (s)
my
[4.14]
where d o ^ is the incremental expansion of the yield surface due to strain-hardening (from Equation [4.9]).
147 Figure 4.11 shows the contours of (Su I Su0 )cms 15 years after construction of Stage 1. The contours represent zones of relative strength increase due to consolidation. From Figure 4.11, it can be seen that the magnitude of (Su ISM)cgm is 1.4 in Zone A
below the embankment centerline. This indicates a 40% increase in undrained strength. The zones influenced by consolidation extend 13m from the centre line and over 10m deep into the foundation. Thus, from Model 1, significant increase in the undrained strength of the Gloucester clay would be anticipated after 15 years of consolidation. Figure 4.12 shows the contours of (Su I Su0)cgm 4 years after construction of Stage 1. The expansion of the (Su I'Su0)cons =1.2 contour from the 4 th year to the 15th year of Stage 1 suggests that the magnitude of strength gain due to secondary consolidation increases considerably with time; whereas the extent of strength gain does not change significantly. From Model 1, up to 20% strength gain would be anticipated 4 years after construction of Stage 1.
4.4.2 Analysis using the Structured EVP Model (Model 2) Long-term settlement Figure 4.13 shows the calculated settlement versus time at Gauge SI, using the structured EVP model (Model 2). From Figure 4.13, it can be seen that Model 2 gives settlement predictions that are very close to those measured for both Stages 1 and 2. This is an improvement over the unstructured EVP model, since referring back to Figure 4.8, Model 2 gives a stress-strain response during CRS oedometer compression that is consistent with that measured for Gloucester clay (see Figure 4.8) and using structure parameters that are easily deduced from laboratory tests. Figure 4.14(a) presents the
148 measured and calculated settlement during Stage 1 at Gauge SI and S3, using Model 1 and Model 2 respectively. This figure shows that the results from Model 2 agree better with the measured data, compared with that from Model [i66] 1. Furthermore, both models were found to give settlement with depth comparable to that measured in the case. Figure 4.14(b) shows the measured and calculated excess pore water pressure 1 year after the construction of Stage 2. Both Model 1 and Model 2 slightly overestimate the excess pore water pressure at all depths. Model 1 gives slightly higher excess pore pressures within the upper soft clay layer and lower excess pore pressure within the underlying medium to stiff clay layer, compared with Model 2. Strength Loss Due To Destructuration For Model 2, there are two competing effects on the strength change during consolidation. The first is strength increase due to consolidation and consequent
hardening of the yield surface. This strength increase with time during the Gloucester case can be examined by contouring (Su ISM)com using Equation [4.14]. However, for
the structured EVP model, there is also loss of strength caused by destructuration. From Equations [4.9] and [4.10], the relative strength loss due to destructuration can be represented viz:
V""0 )Str
where co0 is the intact structure, co{ed) represents the state-dependent structure. Referring to Equation [4.15], destructuration causes a reduction of peak undrained strength. For the Gloucester clay with &)0=1.18, the maximum strength loss is 15%, representing complete transformation to the intrinsic state.
149 The distribution of destructuration 15 years after the construction of Stage 1 is shown in Figure 4.15. [Su/Su0 ) . The degree of destructuration is denoted by contours of
From Figure 4.15, it can be seen that the strength loss due to
destructuration is concentrated in two zones (Zone A and Zone B). Zone A is located in the upper 5m-thick soft clay layer close to the centerline, where most of the vertical compression in the foundation occurs ( Leroueil et al. 1983 and Figure 4.14(b). The large plastic strain in this layer causes considerable destructuration. [Su/Su0 ) The magnitude of
in the center of Zone A ranges from 0.9 to 0.85 suggesting the undrained
strength would decrease by 10% to 15%. In Zone B under the embankment toe, the magnitude of \SU I Su0 J varies from 0.95 to slightly higher than 0.9, indicating a
relative decrease in undrained strength by 5% tol0%. Combined influence of Consolidation and destructuration In reality, the variation of undrained strength is affected by both destructuration and consolidation. The combined effects can be deduced from Model 2 by plotting the ratio,5/5 M0 :
(
c
( ^
( <i \
vsu"V
[4.16]
V " J Com V " JStr
where \SU I Su0 J
is the ratio representing the relative strength loss due to denotes the relative strength gain due to
destructuration, and the ratio, [Su/Su0)Cgnt, consolidation.
Figures 4.15 and 4.16 present the calculated strength variation in the Gloucester foundation under the combined influence of destructuration and consolidation at the end
150 of Stage 1. Figure 4.16 shows zones of net strength gain (i.e. Su /Su0 >1) and Figure 4.17 shows zones of net strength loss (i.e. Su /Su0 <1). Referring Figure 4.16, it appears that the strengthened zones are located mainly within the upper soft clay layer in Zone A. However, the actual strength increase is only 30% in Zone A, compared to 40%, which was deduced from conventional unstructured soil mechanics (Model 1). Figure 4.17 presents the contour of Su/Su0 <1, denoting the net strength loss in the foundation.
From Figure 4.17, it can be seen that there is an extensive weakened zone (Zone B) under the toe of embankment, which reaches a depth of 3m. A second weakened zone (Zone C) is located between depths of 8m and 13m and it extends laterally extending about 7m from the centre line of the embankment. In summary, 15 years after the construction of Stage 1, the strengthening effect due to consolidation is significant; however, destructuration reduces the magnitude and extent of the strengthened zone that would be expected from conventional soil mechanics. In addition, there is some weakening in Zone B, which is likely on the potential failure surface and may be of practical interest for stability analysis and design of staged construction on structured natural clay. Time effect on the strength variation Figures 4.18 and 4.19 show the development of strengthened and weakened zones in the Gloucester foundation from the 4th year to the 15th year after the construction of Stage 1. As shown in Figure 4.18, the extent of strengthened zones does not change significantly during this period. However, the magnitude of strengthening increases
considerably, as suggested by the expanded Su / Su0 =1.2 contour. The weakened zone in Zone B under the embankment toe maintains the same size while the intensity of
151 weakening increases, as suggested by the expanded S M /S u0 =0.95 contour in Figure 4.19,. The other weakened zone located in Zone C expands in size slightly (about lm downward). Simultaneously the degree of strength loss in this zone increases, but it does not exceed 5%.
Influence of additional load (Stage 2) Figure 4.20 shows the geometry after construction of the Stage 2 fill. Correspondingly, the construction of Stage 2 increases the stresses imposed by the embankment on the clay foundation. As a result, additional consolidation and
destructuration is induced by Stage 2, as discussed below. Figures 4.20 and 4.21 show the distribution of net strength gain (Su /Su0 >1) and strength loss( Su /SuQ <1) .respectively, at the end of Stage 1 and 7 years after the construction of Stage 2. The net strength gain has been estimated using Equation [4.16]. Figure 4.20 shows that the zone of strength increase expands laterally by about 2m due to the increased width of the fill base. In addition, the magnitude of strengthening also increases, as shown by the expanded contour line for Su /Su0=1.2. In Figure 4.21, the
weakened zone in Zone B shifts laterally by about 2m outward, as a result of the new location of the embankment toe. The weakened zone in Zone C also expands downward and further outward from the centerline[G67].

This chapter has examined the influence of clay structure and its degradation on the field performance of the Gloucester test embankment. Two different constitutive
152 models, an unstructured EVP model and structured EVP model, have been used to evaluate the long-term settlement of the Gloucester test embankment and to assess the change in undrained strength with time. The following is a summary of the conclusions drawn from this study: 1. More than 50% of Stage 1 settlements occurred at constant effective stress, due to secondary compression. A similar observation on the long-term (33 years) settlement of an Accommodation building founded on the same site has been reported by McRostie and Crawford (2001), as shown in Figure 4.22. As a result, it can be concluded that a constitutive model that can account for the viscous behaviour of Gloucester clay is required to predict the long-term performance of infrastructure founded on Gloucester clay. 2. The structured EVP model (Model 2) was capable of describing the long-term performance of the Gloucester case and gave improved results compared to those obtained using the unstructured EVP model, and using soil parameters consistent with those measured in the Gloucester case. 3. During Stage 1 of the Gloucester case (the first 15 years), the undrained strength of the foundation deposit is subject to the combined influence of destructuration and consolidation. In Zone A (see Figure 4.16) directly below the embankment, the strengthening effect due to consolidation overshadows the strength loss due to destructuration. In Zone B near the toe and Zone C at depth below centerline, however, the opposite was observed. In these zones, destructuration overshadows consolidation causing a net decrease in undrained strength of between 5% and 10% in Zone B and 5% in Zone C.
153 4. The distribution of net strengthening and weakening in the Gloucester foundation can be interpreted as follows. In Zone A below the embankment centerline, the clay is confined and the predominantly volumetric compression results in hardening of the yield surface. This hardening effect leads to strength gain that overshadows In Zones B and C, however, there is more
strength loss caused by destructuration.
deviatoric or shear strain and comparatively less volumetric strain. The net effect is a reduction of strength due to the limited volumetric strains. 5. There is additional evidence reported in the literature of other cases that show the existence of weakened and strengthened zones in clay foundations below embankments. For example, eight years after construction of the 2-meter-high
Murro[G68] Embankment, an in situ investigation (Koskinen et al. 2002) was carried out to access the increase of the undrained shear strength. As pointed out by Karstunen et al. (2005) within the top 7 meters in the foundation, the undrained shear strength increased; Below that, however, the undrained shear strength decreased. Another example is the 1.5-meter-high the Ska-Edeby embankment (Holtz and Lindskog 1972). For this case, vane shear tests in 1961 and 1970 suggested that the vane shear strength increased about 5kPa within the upper 3 meter in the clay deposit, whereas there was a reduction in shear strength for the clay at the depth between 3m and 5m. These trends of strength variation under the Murro Embankment and the Ska-Edeby embankment agree well with that calculated for the Gloucester embankment, where the undrained strength is strengthened in the top 10m, but weakened between the depths of 10m and 15m. 6. Analysis of the Gloucester case has shown the possible existence of a weakened or disturbed zone of clay near the toe of the Gloucester test embankment. It is
154 conceivable that such a zone may influence the stability of embankments founded on highly structured clay deposits and reduce the effectiveness of staged construction. As a result, it is concluded that the importance of stress concentrations near the toe of embankments should be explored further. 7. The structure of natural clay plays an important role in the distribution of strengthened and weakened (disturbed) zones in embankment foundations. For a clay deposit with more structure than the Gloucester case, the structure degradation would lead to a greater reduction in undrained strength. As a result, the weakened zone would expand; whereas the strengthen zone would contract relative to those deduced for the Gloucester case. For Gloucester clay with &) 0 =1.18, the maximum reduction of
undrained strength is 15% ; While for St. Jean Vianney clay with coQ =1.1, the maximum reduction ratio would be 42%. It can be inferred that the strength loss due to
destructuration would be more serious for the same embankment founded on St. Jean Vianney clay than that on Gloucester clay.
155
References
Adachi, T., and Oka, F. 1982. Constitutive equations for normally consolidated clay based on elasto-viscoplasticity. Soils and Foundations, 22(4): 57-70. Baudet, B., and Stallebrass, S. 2004. A constitutive model for structured clays. Geotechnique, 54(4): 269-278. Bjerrum, L. 1967. Engineering geology of Norwegian normally-consolidated marine clays as related to settlements of buildings. Geotechnique, 17(2): 81-118. Bjerrum, L., and Lo, K.Y. 1963. Effect of aging on shear-strength properties of normally consolidated clay. Geotechnique, 13(2): 147-157. Bozozuk, M., and Leonards, G.A. 1972. The Gloucester test fill. In Proceedings of the ASCE Specialty Conference on Performance of Earth and Earth-Supported Structures, pp. 299-317. Burland, J.B. 1990. On the compressibility and shear strength of natural clays. Geotechnique, 40(3): 329-378. Callisto, L., and Rampello, S. 2004. An interpretation of structural degradation for three natural clays. Canadian Geotechnical Journal, 41(3): 392-407. Carter, J.P., and Balaam, N.P. 1990. AFENA-A general finite element algorithm: users manual, School of Civeil Engineering and Mining Engineering,University of Sydney, Australia. Chen, W.-F., and Mizuno, E. 1990. Nonlinear analysis in soil mechanics : theory and implementation. Elsevier Science Publishing Company Inc., New York, NY, U.S.A.
156 Fisher, D.G., Rowe, R.K., and Lo, K.Y. 1982. Prediction of the second stage behaviour of Gloucester Test Fill. In Geotechnical Research Report. Hinchberger, S.D., and Rowe, R.K. 1998. Modelling the rate-sensitive characteristics of the Gloucester foundation soil. Canadian Geotechnical Journal, 35(5): 769-789. Hinchberger, S.D., and Rowe, R.K. 2005. Evaluation of the predictive ability of two elastic-viscoplastic constitutive models. Canadian Geotechnical Journal, 42(6): 1675-1694. Holtz, R.D., and Lindskog, G. 1972. Soil movement below a test embankment. In Proceedings of the ASCE Specialty Conference on the Performance of EarthSupported Structures. Lafayette, Indiana. Purdue University, pp. 273-284. Holtz, R.D., and Broms, B.B. 1972. Long-term loading tests at Ska - Edeby. In Proceedings of the ASCE Specialty Conference on Performance of Earth and Earth-Supported Structures. Sweden. Purdue University, Vol.1, pp. 435- 464. Kabbaj, M., Tavenas, F., and Leroueil, S. 1988. In situ and laboratory stress-strain relationships. Geotechnique, 38(1): 83-100. Karstunen, M., Krenn, H., Wheeler, S.J., Koskinen, M., and Zentar, R. 2005. Effect of anisotropy and destructuration on the behavior of Murro test embankment. International Journal of Geomechanics, 5(2): 87-97. Katona, M.G. 1984. Evaluation of Viscoplastic Cap Model. Journal of Geotechnical Engineering, 110(8): 1106-1125. Kim, Y.T., and Leroueil 2001. Modeling the viscoplastic behaviour of clays during consolidation: Application to Berthierville clay in both laboratory and field conditions. Canadian Geotechnical Journal, 38(3): 484-497.
Law, K.T. 1974. Analysis of Embankments on Sensitive Clays. Ph.D Thesis, University of Western Ontario, London, Ontario. Leroueil, S., and Vaughan, P.R. 1990. The general and congruent effects of structure in natural soils and weak rocks. Geotechnique, 40(3): 467-488. Leroueil, S., Samson, L., and Bozozuk, M. 1983. Laboratory and field determination of preconsolidation pressures at Gloucester. Canadian Geotechnical Journal, 20(3): 477-490. Liu, M.D., and Carter, J.P. 2002. A structured Cam Clay model. Canadian Geotechnical Journal, 39(6): 1313-1332. Lo, K.Y., Bozozuk, M., and Law, K.T. 1976. Settlement analysis of the gloucester test fill. Canadian Geotechnical Journal, 13(4): 339-354. McRostie, G.C., and Crawford, C.B. 2001. Canadian Geotechnical Research Site No. 1 at Gloucester. Canadian Geotechnical Journal, 38(5): 1134-1141. Perzyna, P. 1963. Constitutive equations for rate sensitive plastic materials. Quarterly of Applied Mathematics, 20(4): 321-332. Perzyna, P. 1971. Thermodynamics of rheological materials with internal changes, 10(3): 391-408. Rouainia, M., and Wood, D.M. 2000. Kinematic hardening constitutive model for natural clays with loss of structure. Geotechnique, 50(2): 153-164. Rowe, R.K., and Hinchberger, S.D. 1998. The significance of rate effects in modelling the Sackville test embankment. Canadian Geotechnical Journal, 35(3): 500-516. Silvestri, V. 1984. The preconsolidation pressure of Champlain clay, Part II. Canadian Geotechnical Journal, 21(3): 600-602.
Stermac, A.G., Lo, K.Y., and Barsvary, A.K. 1967. The performance of an embankment on a deep deposit of varved clay. Canadian Geotechnical Journal, 2(3): 234-253. Yin, J.-H., and Graham, J. 1996. Elastic visco-plastic modelling of one-dimensional consolidation. Geotechnique, 46(3): 515-527.
159
Table 4.1
Material parameters used in both Model 1 and Model 2 for the numerical analysis of the Gloucester test embankment
Depth (m)
0.0-2.0 2.0-5.2 5.2-7.2 7.2-13.4 13.4-20.2
e0
Mcs
0.9 0.9 0.9 0.9 0.9
R
1.65 1.65 1.65 1.65 1.65
(xlO m/min)
10.0 7.2 6.0 6.0 7.2
Ck
M oc
0.025 0.65 1.8 0.3 0.025 0.65 1.8 0.3 0.025 0.32 1.8 0.3 0.025 1.35 2.4 0.3
0.25 0.9 0.25 0.9 0.5 0.9
0.25 0.9 0.25 0.9
0.025 0.75 1.8 0.3
160 Table 4.2 Viscosity[i69]-related parameters for Gloucester clay used by Model 1 and Model 2
Model 1 (Hinchberger and Rowe Model) a = l/n 0.033

YVP
Model 2 Structured EVP model a = lln 0.033 77 lxl0" 8 /min <o 1-18 b 50
ixlO' 8 /min
161 Figure 4.1 (a) Geometry[i70] of the Gloucester test embankment and (b) properties of Gloucester clay (a) Geometry[i71] of the Gloucester test embankment and the according finite element mesh (modified from Hinchberger 1996)
Distance from the centre (m)

5 10 15 20
Preconsoiidation pressure (kPa)
Fissures
50
100
150
1 * Soft clay # layer
clay layer
Boundaries: A-B: Smooth, rigid, no drainage; B-C: Smooth, rigid, drained; C D : Smooth, rigid, no drainage
Oedometer(NRC) IAssumed a' (sii
10
15
20
25
ure 4.1 (b) Comparison[i72] of the adopted parameters with laboratory results of Gloucester clay (from Hinchberger 1996)
Compression Index, X 1 T~ 2 I
Hydraulic Conductivity (m/min)
Preconsolidation Pressure (kPa)
10
15
50
i
100
150
:
I I I
-|llll|llli|r-
NUMERICAL
NUMERICAL
GROUNDWATER O LEVEL!
m
FIELD
o o o
-t-O
4h 6b 8h
Q. Q
8 4QIJ&.
o *p o o
Jo-
To D" "b" o
\<8>. <\c
O \Q O
10 12
14
...o.. oo
O"
Oi LAB YIELD SURFACE INTERCEPT ELLIPTICAL CAP MODEL
.'M
o\.
& o
\0
10 12
<D 3
<JE>
Q o
O
o
LAB (TO C)
14
16
my0
o
O
16 18 20
LABORATORY oedometer (NRC) preconsolidation; pressure
h\ o
\ 18 ......... \
i L _i i
_J
iL_i_
20
Figure 4.2
Influence of clay structure on the behaviour of Gloucester clay in undrained triaxial and oedometer compression tests
a) Typical undrained[i73] triaxial compression[i74] [G75]response
2.0
1.5 H
Structured Specimen (a'c=40kPa)
(0
1.0 H Destructured Specimen (a'c=83kPa)

(0
0.5 4
0.0
Data from Law 1974 8 10
Axial Strain,%
Figure 4.2
Influence of clay structure on the behaviour of Gloucester clay in undrained triaxial and oedometer compression tests (Cont.)
b) Typical oedometer[i76][G77] compression curve
p-iap-s
0.0
10
0'p, =80.5kPa
-0.1
f p-s ==95KPa
-0.2 O (0
<*>o i
=G,p-s/t'p-i =1.18
5 'o > H o w
"0.3
to
-0.4 4
JO
O
-0.5 -
ICU: Intrinsic compression line

-0.6 4
Measured Data from Leroueil et al. (1983) on Gloucester clay obtainted from 4.05-4.18m
-0.7
10
100
Vertical effective stress (a'u), kPa, in log scale
165 Figure 4.3 Rate-sensitivity of the undrained shear strength and pre[G78]consolidation pressure of Gloucester clay (a) Undrained strength
Data from Law 1974 CAU peak strength o CAU large strain strength
10 -3
10-2
10-1
Strain Rate (/min)
166 Figure 4.3 Rate-sensitivity of the undrained shear strength and pre[G79]consolidation pressure of Gloucester clay (Cont.) (b) Preconsolidation pressure
200
C O
Measured data by Leroueil et al. (1983) at a depth of 4.08-4.15m Measured data by Leroueil et al.(1983) at a depth of 3.45-3.9m Usipg Constant Rate of Strain (CRS) oedometer tests
tn tn
a) 100 90 80 70 60 -
c o
CD
Estimated in-situ preconsolidation pressure from depths between 2.4 and 4.9m (Leroueil et al 1983)
"5 CO c o o
CD
c
CD Q.
50 40
-10-8
Estimated ranges of preconsolidation pressure from conventional oedometer tests (Leroueil et al 1983)
10-e
<
10-'
-i 0 -5
10"
10":
10":
Strain rate, /min
167 Figure 4.4 Long-term oedometer creep[G80] tests on Gloucester clay (data from Lo et al. 1976)
24
c o "co
CO
4 4 Long-term consolidation tests
(Loetal. JI976)
O
AU=
6 4
o
84
Sarjnple at -2.4 rpetre Sarinple at - 4.2 pnetre
10 10 101 102 103

10 4 105 10 6
Time, mins
Figure 4.5
The state boundary[i81] surface and critical state line for Model 1 and Model 2.
pj2
State boundary surfac;e:Al-A2-A3 Dynamic yield surfa :e: B1-B2-B3
AM
, /classical Critical State Line
/ B 2 / ^ ^^"-"A2/ B3 A3
"^
1Y1
Elliptical Caps /
dF
P
^ \ G \~~~ State of overstress at 'G' \ / Dynamic yield \* surface \ Bl Elastic^/ \ domain State boundary/ \ surface A l\
oc
/ /
f
^-,
a,
/(s)
'(d)
169 Figure 4.6 Illustration of the theoretical response[i82] of Model 1 (Hinchberger and Rowe Model)
(a)
(c)
< eB <
log(eO
(b)
(d)
I
a = \ln
sr
log(e)
log(e)
170 Figure 4.7 Illustration of the theoretical[i83] response of Model 2
(a)
CJ
(c)
C
C
1 Metastable Structure
Sc
ec
*j
1 _~B'
B A
SB
m
U
A
$uA
&A < ^B <
log(crp)
(b)
(d)
I
~ " ~ ~ Y " " T B'
Post
P ^ strength
= 77
= 7)
log()
= r7
srf
log(e)
gure 4.8
Comparison of the measured behaviour in CRS oedometer test on Gloucester clay and the corresponding theoretical response of Model 2
200
-0.7
Measured Data from Leroueil et al. (1983) on Gloucester clay obtained from 3.45-3.90m Theoretical response (Model 2)
50 100 150
-0.8 200 Vertical effective stress (o'v), kPa
Figure 4.9
Comparison of the measured settlement at Gauge SI with the calculated settlement using Model 1
1.5 years 5 years 15 years 20 years
1000
2000
3000
4000
5000
6000
7000
8000
Time (days)
173 Figure 4.10 Illustration of the linear and bilinear virgin compression curves
(a) linear approach

ln(<7)
Long Term Compression Curve
>ev0
Ae
(b) Bilinear approach
Transition Residual phase
Figure 4.11
Zones of strength gain due to consolidation, 15 years after the construction of Stage 1- Contours[i84] of {Su/Su0)cons

25 5 10 T 15 20 T 25
Preconsolidation pressure (kPa) Stage 1 50 100 150
-!r~
Zone A
j> Soft clay layer
. Medium to stiff clay layer
Oedometer(NRC) - Assumed a' J*\ Contours of ( 4 / 5 0 X , ) m
10
15
20
25
Figure 4.12
Zones of strength gain due to consolidation, 4 years after the construction of Stage 1. Contours[i85] of (Su /Su0)cgns
Distance from the centre (m) 10 15
Soft clay
la er
, Medium to stiff clay layer
4 years after the construction of Stage 1 15 years after the constructio of Stage 1
Oedometer(NRC) -Assumed a'my(s^
10
15
20
25
Figure 4.13
Comparison of measured settlement (Gauge SI) with calculated settlement using Model 2
1.5 years
0 i
5 years
10 years
15 years
20 years \
A
Stage 1
-10 -20 -
Stage 2
?
4-1
-30 -40 -
E <u -50 C/)
-60 Measured Data in Stage 1 Measured Data in Stage 2 ""^^^~ Caculated by Model 2
o
i i
" i
-70 -80 -
- I
i-
^ ^
1000
2000
3000
4000
5000
6000
7000
8000
Time (days)
Figure 4.14
Comparison of the measured and calculated settlement and excess pore water pressure using Model 1 and Model 2
(a) Comparison of the measured settlement (Gauges SI and S3) with the calculated settlement using Model 1 and Model 2
1.5 years
5 years
10 years
13 years
i.
-40 4 -50 -60 -70 -80 Measured Settlment (S1) in Stage 1 Calculated by Model 2 Calculated by Model 1 Measured Settlement (S3) in Stage 1 Calcualted by Model 2 Calculated by Model 1
<D C/D
1000
2000
3000
4000
5000
6000
7000
8000
Time (days)
Figure 4.14
Comparison of the measured and calculated settlement and excess pore water pressure using Model 1 and Model 2 (Cont.)
(b) Comparison of the measured extra pore water pressure[i86] with the calculated settlement using Model 1 and Model 2
c g
'4-
CO
>
LU
o
1 Year after the constructure of Stage 2
Geonor Hydraulic Standpipe IRAD Vibrating Wire Calculated using Model 1 Calculated using Model 2 3 4 5
Excess Pressure Head (m)
Figure 4.15
Zones of strength loss due to destructuration, 15 years[i87] after construction of Stage 1. Contour of [Su I Su0) 'Str
25
5
T
10
T
15
20
T
25
Preconsolidation pressure (kPa) 50 100 150
"ir
if f. Soft clay layer , .Medium to stiff ' clay layer
Contours of
(sjSua\ 20
>Oedometer(NF\C) I Assumed &
10
15
25
Figure 4.16
Zones of net strength gain (i.e. consolidation overshadows destructuration), 15 years[i88] after construction of Stage 1. Contour of
SJSU0>1

25 5 10 T 15 T 20 25
Preconsolidation pressure ' (kPa)
20 I
Stage 1
50
100
150
ir
^ ' Soft clay layer
Zone A
a #
Medium to stiff clay layer
g 5 10 i
LU
Zone C
Oedometer(NPtC) Assumed o'my<s)> Contours of Su I Su(j > 1
10
15
20
25
Figure 4.17
Zones of net strength loss (i.e. destructuration overshadows consolidation), 15 years[i89] after construction of Stage 1. Contour of
su/su0<i

25
10
15
20
25
Preconsolidation pressure (kPa)
20 4
Stage 1
.NylZone B
50
lio
150
T f Soft clay layer
Zone A E, 15
C O "^ ro > LU
^Medium to stiff clay layer
Contours of ; S / Sll0 < 1
Oedometer(NRC) I Assumed a' (s!l
10
15
20
25
Figure 4.18
Development of zones of net strength gain from the 4th year to the 15th year in Stage 1

5 10 15 20
Preconsolidation pressure (kPa)
5D 100 150
i f *9
Soft clay layer
Medium to stiff clay layer
15 years after the construction of Stage 1 4 years after the construction of Stage 1
Oedometer(NRC) Assumed a' Js\
10
15
20
25
Figure 4.19
Development of zones of net strength loss from the 4th year to the 15th year in Stage 1

25 5 T 10 T 15 T 20 25
20
i r
' Soft clay layer
Zone A
15
a
c q
> m 10
, Medium to stiff clay layer
5|
Contours of S / S,, < 1
15 years after the construction of Stage 1 4 years after the construction of Stage 1 Oedometer (NFJC) I Assumed o',
10
15
20
25
Figure 4.20
Zones of net strength increase, 7 years [i90] after construction of Stage 2

5 10 15 20
25
Preconsolidation pressure (kPa) I 50 100 150
IS *
Soft clay layer
. Medium to stiff clay layer
54
Contours of Sh/Sll0 >1 Oedometer (NF\C) I Assumed 0'
At the end of Stage1(i.e. just before the construction of Stage 2) 7 years after the construction of Stage 2
10
15
20
25
Figure 4.21
Zones of net strength loss 7 years after construction of Stage 2

10
25 Stage 2
15
20
25
20
Stage 1
ir
Soft clay ' layer
Zone A
15
a s
y; 1.0
Lateral shift due to the additional loads in Stage 2
Medium : to stiff * clay layer
Contours of
S^ I Sll0 < 1 At the end of Stage 1 (i.e. just before the construction^ Stage 2) 7 years after the construction of Stage 2
Oedometer(NRC) Assumed a'my(s))
10
15
20
25
186
Figure 4.22 Comparison of the compression curve in laboratory test with the measured long-term field compression of Gloucester clay under the Accommodation building (from McRostie and Crawford, 2001)
80
100
In situ observation 0.5 year 1 year Average of laboratory tests 2 years
4H
E o o
en a> a.
10
20
40
60
80
100
Vertical effective stress (cr'v), kPa
187
CHAPTER 5 AN ANISOTROPIC EVP MODEL FOR STRUCTURED CLAYS
5.1 Introduction
Structured[G91] clay deposits are widely distributed throughout the world. As a result, many countries build significant infrastructure on or in these difficult soils. During loading, these clays can exhibit engineering characteristics such as ratesensitivity, drained and undrained creep, accelerated creep rupture and significant anisotropy (Lo et al. 1965; Lo et al. 1972; Tavenas et al. 1978; and Vaid et al. 1979). Some of these characteristics, in particular anisotropy, have been attributed to the microscopic structure of clay. For many structured clays, both anisotropy and viscosity appear to be significant. Lo and Morin (1972) found that anisotropy, strain-rate and time effects were pronounced for St. Louis and St. Vallier clay from Eastern Canadian,. Tavenas et al. (1978) observed similar behaviour for other clays from eastern Canada. The engineering significance of both anisotropy and strain-rate effects has been well established. Recently, Hinchberger and Rowe (1998) and Kim and Leroueil (2001) demonstrated the importance of viscous effects for embankments founded on soft clay deposits. Similarly, Zdravkovic et al. (2002) demonstrated the effect of anisotropy on embankment behaviour. Thus, a
constitutive model that can describe both anisotropy and viscous effects in 'structured' clays would be useful in geomechanics. This chapter describes a constitutive approach to model the time-dependent
188 plastic behaviour of rate-sensitive anisotropic structured clay. The main objective of the chapter is to demonstrate a novel approach to the anisotropic behaviour of viscous 'structured' clay at yield and failure. As a consequence of this study, some observations are also made regarding the anisotropic elastic behaviour of 'structured' clay. The
constitutive approach described in the following sections utilizes non-linear elasticity theory, overstress viscoplasticity (Perzyna 1963), a Drucker-Prager failure envelope, and an elliptical cap yield surface (Chen and Mizuno 1990). Structure is accounted for by adopting a viscosity parameter that is initially high (the structured viscosity) and that decreases to the residual or intrinsic viscosity due to plastic strain or damage strain (see chapter 3). The structured viscosity is made anisotropic using a tensor approach similar to that described by Boehler (1987), Pietruszczak and Mroz (2001) and Cudny and Vermeer (2004). The intrinsic viscosity is assumed to be isotropic. Theoretical
behaviour is compared with the measured response of Gloucester clay and St. Vallier clay (Lo and Morin 1972) during undrained triaxial compression tests on samples trimmed in different orientations, /, to the vertical axis. The comparisons show that the constitutive model is capable of accounting for both anisotropy and strain-rate effects on the engineering behaviour of these clays.
5.2 General Approaches to Anisotropic Plasticity

In general, four main approaches have been developed to describe the anisotropic behaviour of clayey soils at yield and failure excluding those based on nested yield surfaces. The approaches are: (i) Rotational Kinematic Hardening Laws: The yield surface is assumed to rotate under the influence of an anisotropic stress field (Davies and Newson
189 1992; Whittle and Kavvadas 1994; Wheeler et al. 2003). Rotational
hardening models have been used to describe the response of embankments built on natural clay soils (Zdravkovic et al. 2002; Oztoprak and Cinicioglu 2005). (ii) Transformed[G92] Stress Tensor: A fabric tensor is used to modify the stress tensor, <r',,, obtaining the transformed stress tensor, T'
y
Yield and failure
*^
criterion are subsequently developed using T~ instead of &tj (Miura et al. 1986; Tobita 1988; Tobita and Yanagisawa 1992; Sun et al. 2004;). (iii) The Fabric Tensor[G93] Approach: A fabric tensor is used to modify the plastic energy dissipation formulation to develop new state boundary surfaces (Muhunthan et al. 1996). (iv) The Structure Tensor Approach: Boehler (1987), Pietruszczak and Mroz
(2001), and Cudny and Vermeer (2004) used the stress tensor, c' iy , and a microstructure tensor, atj, to obtain an anisotropic scalar coefficient, n , that can be used to give anisotropic characteristics to scalar parameters such as the cohesion intercept, c', and effective friction angle, <f>'. Pietruszczak and Mroz (2001) demonstrated the use of this approach to obtain an anisotropic Mohr-Coulomb failure criterion. In summary, all of these approaches are useful, however, the common limitations of the first three are generally: (i) complex formulations, (ii) numerous material parameters required and (iii) parameters that generally cannot be determined using conventional laboratory tests. However, the fourth approach described by Pietruszczak and Mroz (2001) is relatively straightforward and it can be implemented into viscoplastic
190 formulations (Pietruszczak et al. 2004) with the introduction of only one additional constitutive parameter for the case of transverse isotropy.
5.3 Microstructure Tensor

Transverse Isotropy In accordance with Pietruszczak and Mroz (2001), material anisotropy can be described using a microstructure tensor, atj, which describes the spatial distribution of microstructure. In its general form, the microstructure tensor is:
a a.
xy xz
a J =
a
xx
a
yx vc
a
a
a
yy zy
yz z
[5.1] For clays deposited under the influence of gravity with horizontal bedding or laminations, the principal directions of anisotropy are vertical and horizontal. In this case, the
microstructure tensor, ay , is coaxial with the axes of orthotropy of the material and it can be simplified to:
a
xx
0
T
a- =
y
0 0
0" 0
T
[5.2] where the superscript, T, denotes transverse isotropy. For a transverse isotropic material, the microstructure tensor can also be written in terms of the mean and deviatoric components viz.:
191 r a,-, = 0 0 [5.3] fit[ + flj + a3 a3 am 2a3 1ax a , A 2ax +a3 a 0 0
o"
0 =
T
am
a3 _
-arA/2 0 0 0 0 a -aa, A/2 m m 0 0 a +a A

m m
where a is the mean structure and A describes deviations from the mean. When aT is normalized by am, the microstructure tensor becomes: a a A/2
m m
11
am
am l-A/2 0 0
0 0 0 l-A/2 0
a
n
0 -a m A/2 0
am m
0 0 +am A m
0 0 1+A
[5.4] where the normalized microstructure tensor, a j , quantifies the spatial distribution of structure with respect to the mean structure and the parameter, A, defines the degree of inherent material anisotropy. The absolute magnitude of A is zero in the case of isotropy and it increases as the degree of anisotropy increases. Researchers such as Oda and Nakayama (1989) have shown that it may be possible to relate A to measurements of soil fabric. For most naturally deposited clays, the major principal anisotropic direction is vertical (e.g. azz>axx=a ). Correspondingly A is positive (see Figure 5.1a).
However, for heavily overconsolidated clays such as London clay (Ward et al. 1959) high
192 horizontal stresses may lead to higher undrained strength in specimens of horizontal orientation compared to those of vertical orientation. Accordingly, A could be negative if the major principal direction of anisotropy is horizontal (see Figure 5.1b). For clays with sub-horizontal bedding or laminations, a transform tensor, Q, can be applied. For example, in the case of plane strain:
fl
= gx
cos" i sin" i sin2/ 2 2 = sin / cos / -sin 2/ - 0.5 sin 2i 0.5 sin 2/ cos 2i
[5.5] where i is the angle of the bedding or laminations relative to the horizontal axis. Thus, the transform tensor can be applied to cases where the major principal directions of the microstructure tensor are not oriented along the vertical direction. From Pietruszczak and Mroz (2001), a scalar parameter, rj, can be derived to define the anisotropy of a material using the generalized effective stress state, otj , and
_r
2
microstructure tensor, atj or atj .
The diagonal components of atj
represent the
resultant stresses on each of the principal planes of orthotropy (see Figure 5.1c):
(
9
(
.2 \
'2
'2
'2
+(J
'2
\ ' )yy
.l\
(J 2
U=(yx
=C7
'y
2=a
=xx
2
xy +yy +yz
=<T
xy
.2
+
+xz
yi
.2
.2
)=<** [5.6]
2 .2
zy
zz
The anisotropic scalar parameter, r\, can be obtained by taking the normalized projection of the microstructure tensor on the generalized stress state viz.:
193
2 -
_ _
.- 2.
fr>.ir 0" U
tr\fT9 j
[5.7] which in the case of a vertically orientated (e.g. i=0) specimen subject to a triaxial stress state simplifies to: ( 1 - A / 2 W 2 +(1-A/2)cr' 2 +(1 + A)(T'2 7 = 7 2 .2 .2
^
+<7
+(J
[5.8] Equation [5.7] conforms to the Representation Theorem of Isotropic Functions (Wang et al. 1970) and as such, r\ is independent on the choice of orthogonal coordinate system viz.
[5.9] where Q is the transform tensor. The scalar parameter, r\, accounts for the influence of stress orientation and material orientation as illustrated in the following section.
5.4 Application to Tresca's Failure Criterion

To illustrate the use of the microstructure tensor, consider Tresca's failure criteria, which is often used in soil mechanics: f(a\,r])=a\ [5.10] -a'3 -r]cu0 (<J\ >cr' 3 )
194 where cu0 is the isotropic undrained shear strength and 7 represents the influence of 7 anisotropy on the undrained strength of clay. As shown above, the scalar coefficient, 77, is derived from the microstructure tensor, aj,, and the stress tensor, o'. The magnitude of 7 depends on the relative orientation and magnitude[G94] of both a I and <7y. 7 Consider a series of undrained triaxial compression tests on clay specimens trimmed at different orientations, i, to the vertical. In accordance with Equation [5.9], the effect of sample rotation can be taken into account by transforming the structure tensor using Equation [5.5] taking i equal to the angle formed by the specimen axis and the vertical (see Figure 5.2). Now, given the following arbitrary triaxial stress state:
a\ 0 0
0 a\ 0
0 0 a\
"l 0 0" = 0 1 0 0 0 4
[5.11] where a\ is the cell pressure in a triaxial test, and &'a is the axial stress, the influence of sample orientation, i, on 7 is shown in Figure 5.3 for A equal to 0, 0.1 and 0.2, 7 respectively. Referring to Figure 5.3, when the anisotropic parameter A equals zero, 7 is 7 constant and equal to one. For this case, the resultant undrained shear strength is isotropic. As A is increased from 0.0 to 0.1 and 0.2, respectively, the undrained shear strength becomes increasingly more anisotropic and the strength of vertical samples ( i = 0) exceeds that of horizontal samples ( i = 90). Conversely, the strength of
horizontal samples ( i = 90) exceeds that of vertical samples ( i = 0) when A is
195 negative. Thus, the parameter r\ can be used to modify Tresca's failure criteria obtaining anisotropic undrained shear strength similar to that observed by Lo and Milligan (1967). It can be shown that, for soils that reach a unique effective stress ratio at failure: Tli = T W + ( T W
_ 1
W )C0S2i
t 5 - 12 ]
which is identical to the relationship used by Lo (1965) to describe the anisotropic undrained shear strength of Welland clay in Canada. Figure 5.4 illustrates the influence of the stress ratio, o\jo\ the following triaxial stress state: , on r]cu0. Consider
\
ij =
0
<y'c
0 0
0 0
<*\.
= 0'c
1 0 0 1
0 0
[5.13]
_0 0
a\l&
which permits investigation of the influence of o\ \&'c , on the anisotropic parameter T|. Referring to Figure 5.4, for stress ratios less than one, the undrained shear strength is higher for horizontal specimens than for vertical specimens since the major principal stress is acting in the radial direction. For stress ratios that exceed one, r\ increases to a maximum of almost 1.2 for vertical specimens and stress ratios in the order of 6. Similar trends can be observed for specimens trimmed at i = 45 and i = 90, respectively. Thus, the anisotropy is not only dependent on the orientation of the stress field relative to the microstructure of the clay, i, but also on the stress ratio, o\ /CJ'C . It should be noted that the parameter 7 approaches its upper limit as &a > 5<r'c. Although this complicates the 7 determination of A somewhat, it has benefits that will be explored later in this chapter.
196
5.5 Application to an Elastic-Viscoplastic Model

Overstress Viscoplasticity The formulation presented in the following sections is based on the Hinchberger and Rowe Model (Hinchberger 1996; Hinchberger and Rowe 1998). This model has a state boundary surface defined by an elliptical cap yield function (Chen and Mizuno 1990) and Drucker-Prager envelope (see Figure 5.5a); it has provision for either isotropic and anisotropic non-linear behaviour in the elastic stress range; and the plastic response is defined within the framework of Perzyna's theory of overstress viscoplasticity (Perzyna 1963) utilizing concepts from critical state soil mechanics (Roscoe et al. 1963). A summary of the Hinchberger and Rowe (1998) model can be found in Table 5.1; however, in principal the following constitutive formulation could be adapted to any overstress viscoplastic model. The basic constitutive equation (from Hinchberger and Rowe, 1998) is:
=e+evp
2G-+-Xl + e)a'm
dF
y
>'
dot
[5.14a] where the flow function, <|)(F) , is a power law viz.
<KF)=
my my
os
-1
[5.14b] In Equations [5.13] and [5.14], e~ is the strain-rate tensor, stj is the deviatoric stress tensor, a ' is the mean effective stress, 8U is Kronecker's delta, G is the stress
197 dependent shear modulus, K is the slope of the e - l n ( a ^ ) curve in the overconsolidated stress range, and e is the void ratio. The scalar function <j)(F) is called the flow function, o,{^ is the overstress (see Hinchberger and Rowe 2005), c'j^ is the static yield surface intercept and dF/dcr'j is the normalized plastic potential for associated plastic flow. An associated flow rule has been adopted in this chapter. It should be noted that although the associated flow rule and isotropic plastic potential simplify the formulation, such an assumption introduces a limitation in the model since the plastic potential of most clays is anisotropic and in some cases non-associated (Graham et al. 1983 and Newson 1998). The time-dependent plastic behaviour of clay is thus governed by the viscosity parameter, u,, and the strain-rate exponent n . Viscosity, u., is the inverse of fluidity
(y'1' = l/(i) and as u, increases the soil becomes less fluid and viscous effects increase. The rate-sensitivity is governed by n . As n increases the rate-sensitivity decreases.
Consequently, through varying n and \x, viscous rate-sensitive, viscous rate-insensitive and inviscous plasticity can be modeled. The latter can be obtained by using an iterative solution scheme to keep the stress-state on either the static yield surface or the DruckerPrager envelope (Zienkiewicz and Cormeau 1974).
Modification for Structure Burland (1990) suggested that the engineering behavior of natural clays can be described with reference to the remolded or intrinsic state. In accordance with this concept, it has been hypothesized (See Chapter 3) that the viscous component of clay structure can be defined in terms of the intrinsic and structured viscosities viz.:
198
1
0)0
=
Vr-mt
[5.15]
J
where a}, is the initial structure parameter, fistr is the initial viscosity of the undisturbed 'structured' clay and jj,^ is the remolded or intrinsic viscosity (Hichberger and Qu 2007). As a result, 'structured' clay is considered to have a high initial viscosity relative to the residual or intrinsic viscosity[G95]. During loading, it is assumed that the initial viscosity, (0,str, is gradually damaged by plastic strain until eventually the clay is completely destructured and the viscosity has degraded to the intrinsic viscosity, jx^ . This process is commonly referred to as 'destructuration' (Rouainia and Wood 2000). Degradation of the clay viscosity is
assumed to occur as a function of damage strain viz.: tfed) = /"int + <A,r - # * )"*'" [5-16]
where b is a parameter that controls the rate-of-destructuration of clay and the damage strain, s d , is:
de d =V(l-A)(de7 0l ) 2 +A(d e ; p ) 2
[5.17]
In Equation [5.17], (see Rouainia and Wood 2000), A is a weighting parameter and e vp vol and e^ are plastic volumetric and octahedral shear strains (V3y oct ), respectively. In this chapter, the weighting parameter, A , has been assumed to be 0.5. It is also recognized that the current model does not account for shear banding or strain localization and that the parameter & in Equation [5.16] includes these effects for shearing modes of failure. In summary, the Hinchberger and Rowe model (Hinchberger and Rowe 1998) has
199 been modified by adopting a state-dependent viscosity parameter and the resultant plastic strain-rate tensor is: BF dF
da'.
HM
(<KF))
[5.18]
The conceptual behaviour of the 'structured' clay model is described below. Conceptual Behaviour of the 'Structured' Model The conceptual behaviour of the structured model has been described extensively by Hinchberger and Qu (2007) for over consolidated materials such as St. Vallier (Lo and Morin 1972) and Saint-Jean Vianney clays (Vaid and Campanella 1977; Vaid et al. 1979). Figure 5.5 illustrates the model behaviour for lightly over consolidated materials during CIU triaxial compression tests. Referring to Figures 5.5b and 5.5c, after initial isotopic consolidation to point 1 in Figure 5.5b, triaxial compression of the soil specimen at a constant rate of strain will cause the effective stress path to move on the elastic wall from point 1 to 2 where yielding occurs. During continued compression, the 'structured' soil skeleton will
undergo plastic straining as the stress path moves from 2 to 3; however, the plastic strainrate during this phase of compression is very low due to the high viscosity of the 'structured' soil skeleton. Thus, the material behaviour is still predominantly elastic from 2 to 3 as shown in Figure 5.5b as overstress builds up relative to the long-term or static yield surface (Hinchberger and Rowe 2005). At point 3, the overstress and resultant plastic strain-rate becomes high enough to begin destructuration of the clay and consequent increased fluidity of the clay skeleton. From point 3 to 4, there is stabilization of the overstress during which the peak strength is
200 reached. From point 4 to 5, however, the damage rate is high and there is a significant reduction of overstress (stress relaxation) caused by the shear thinning or degrading soil viscosity. Thus, strain softening is modeled as a stress-relaxation phenomenon. As
compression continues, it is assumed that eventually the plastic strain causes the viscosity of the soil skeleton to decrease to the intrinsic viscosity; although this state may not be reached during triaxial compression. The conceptual behaviour described in Figure 5.5 applies to lightly over consolidated materials such as Gloucester clay. In addition, during undrained triaxial creep tests, application of a constant deviator stress exceeding that denoted by point A in Figure 5.5a and 5.5b will cause time-dependent plastic creep followed by eventual creep rupture of the material. However, applied deviator stresses below that denoted by point A will cause time-dependent plastic creep that will eventually stabilize when the stress state reaches the static yield surface. Sheahan (1995) summarizes such behaviour for natural clays. In general, the model adopted in this chapter is identical that described in Chapter 3 except that, in this study, an associated flow rule has been assumed in conjunction with separate 'structured' and 'destructured' bounding surfaces. Appendix G describes an alternative approach utilizing non-associated plasticity. Modification for Anisotropy To account for both time-dependency and anisotropy of natural clay at yield and failure, it is hypothesized that the 'structured' viscosity of clay is anisotropic whereas the intrinsic viscosity is isotropic. Studies by Law (1974) and Lo and Morin (1972) contain experimental observations supporting this assumption for some clays from Eastern Canada.
201 Using the microstructure tensor, the structured viscosity can be modified as follows: H{ed,r\) = (jimt + {wstr ~/"int yhSd) [5.19]
where T| is the anisotropic scalar parameter defined by Equations [5.7] and [5.8]. Equation [5.19] can also be expressed in terms of the initial structure, CO;, viz.: co(sd ,7])= (l + (ryo/ - l)e-* ) ' " [5.20]
where oo(ed,T|) defines the remaining structure at any point after some destructuration, e d , has occurred. The resultant anisotropic viscoplastic strain-rate tensor is: dF a<rf, dF
^(ed,r?)
(<KF))
[5.21]
Thus, a structure parameter and microstructure tensor have been used to extend the Hinchberger and Rowe (1998) model to obtain an anisotropic rate-sensitive constitutive model for structured clays. Clay structure is treated as a viscous bonding phenomenon and the source of anisotropy is assumed to be the anisotropic distribution of viscous bonds. The main characteristics of the constitutive model are summarized in Table 5.1. It is noted that for a tensor approach, adopting an associated plastic potential law would lead to underestimation of the deviatoric plastic strain for natural clay subject to at isotropic stress path. A non-associated flow rule can be utilized to overcome this limitation and improve the prediction of the tensor approach on plastic strain under loading. However, additional parameters have to be introduced for the non-associated plastic potential law and these parameters must be deduced from non-standard laboratory
202 tests. Thus, for the sake of simplicity, the tensor approach in this chapter adopts an associated plastic potential law and consequently only one parameter, A, is required to describe the anisotropy characteristic of natural clay.
5.6 Evaluation
Methodology This section compares calculated and measured behaviour of both Gloucester clay and St. Vallier clay during undrained triaxial compression tests on specimens trimmed at various orientations, i, to the vertical. Only tests performed at consolidation pressures less than the in situ overburden stress were considered in the analysis. In addition, the test results used below were obtained using high quality triaxial specimens trimmed from block samples. For Gloucester clay and St. Vallier clay, a series of isotropically consolidated undrained (CIU) triaxial compression tests were evaluated. The measured behaviour has been reported by Law (1974) for Gloucester clay and Lo and Morin (1972) for St. Vallier clay. The calculated behaviour presented in Figure 5.6 through 5.13 was obtained using the finite element (FE) program AFENA (Carter and Balaam 1990), which has been modified by the authors to account for time-dependent plasticity and structure. A FE analysis was undertaken for each test starting from the initial stress state reported during the test. The sample was loaded by prescribing displacements to the top of the mesh at a rate corresponding to the compression rate reported for each test. The FE calculations were performed using 6-noded linear strain triangles in conjunction with axi-symmetric conditions. The top and bottom mesh boundaries were assumed to be smooth (e.g.
203
friction was neglected) and rigid. The FE calculations are summarized in Figure 5.6 through 5.9 for Gloucester clay and Figure 5.10 through 5.13 for St. Vallier clay. The constitutive parameters used in the analysis are listed in Tables 5.2 and 5.3 for Gloucester and St. Vallier clay, respectively. Gloucester clay Law (1974) conducted a series of CIU triaxial compression tests on specimens of Gloucester clay trimmed at 0, 30, 45, 60 and 90 to the vertical[i96] (see Figure 5.2). The test results are summarized in Figure 5.6, which shows the measured and calculated peak undrained shear strength of Gloucester clay versus sample orientation, i. Figure 5.6 also shows the measured and calculated post-peak strength at 8% axial strain and the calculated intrinsic or residual strength of Gloucester clay at large-strain. The intrinsic or residual state was assumed in the FE interpretation even though it is difficult to reach the residual state in a triaxial apparatus. From Figure 5.6, it is evident that the measured and calculated peak undrained shear strength of Gloucester clay are strongly anisotropic. The peak strength of vertical specimens (e.g. /=0) is typically 40% higher than for horizontal specimens (e.g. i = 90). In general, there is overall good agreement between the calculated and measured peak strength for all sample orientations, i. At an axial strain of 8%, there is also good agreement between the calculated and measured post-peak shear strength. The measured strength of Gloucester clay at 8% axial strain is only slightly lower than the calculated strength for the values of / considered. At the intrinsic state, which is reached at 12% axial strain (assumed), the theoretical strength of Gloucester clay is isotropic (see Figure 5.6). Overall, it is concluded that the trends of
204
calculated undrained shear strength are comparable to the measured trends of undrained strength versus sample orientation, i. Calculated and measured deviator stress and excess pore pressure versus axial strain are compared in Figure 5.7 up to 12% axial strain. From Figure 5.7, it can be seen that there is adequate agreement between the measured and calculated behaviour notwithstanding the notable differences in the elastic range as discussed below. In the post-peak stress range, the measured rate of strength reduction is somewhat higher than the calculated rate for specimens trimmed at / of 0 and 30. However, the theoretical response is considered to be a reasonable idealization considering the probable impact of such factors as natural variability on the laboratory measurements. In addition, the
calculated excess pore pressures are generally within 15% of the measured excess pore pressures for axial strains up to 10%. The difference between calculated and measured pore pressures can be attributed to the isotropic elastic theory used to obtain the calculated behaviour as discussed in the following paragraph. Figure 5.8 shows the calculated and measured stress path during triaxial compression tests on specimens at i=0 and 90, respectively. In accordance with
Graham and Houlsby (1983), an anisotropic elastic parameter, (3 = E V / E h , can be derived from the deviation of the measured stress path from the theoretical isotropic stress path for / = 0 (see Figure 5.8). For Gloucester clay, the anisotropic parameter, (3, is approximately 1.6 assuming a Poisson's ratio of 0.3, where Ev and Eh are the vertical and horizontal elastic modulus, respectively. Reanalysis of the CIU triaxial test using cross-anisotropic elastic theory in the elastic-viscoplastic constitutive model produced Curve '2' in Figure 5.8, which is in close agreement with the measured stress path.
205 To conclude, Figure 5.9 summarizes the effect of strain-rate on the undrained strength of Gloucester clay. Again, the constitutive model is capable of describing the overall variation of undrained shear strength versus strain-rate, and as such, the constitutive results are considered to be encouraging. For Gloucester clay, the
constitutive model is capable of describing both the variation of peak and post-peak strength versus sample orientation and the effects of strain-rate on the mobilized strength. The peak strength of Gloucester clay varies by about 10% per order of magnitude change in the strain-rate. This is quite significant and in many cases it should not be ignored by engineers (Marques et al. 2004). St. Vallier clay To complete the evaluation, the anisotropic behaviour of St. Vallier clay during CIU triaxial compression tests was also considered. St. Vallier clay is considered
because it exhibits different anisotropy behaviour from Gloucester clay, which may be of interest. The behaviour of St Vallier clay during CIU triaxial compression was reported by Lo and Morin (1972). Figure 5.10 through 5.13, inclusive, compare the calculated and measured behaviour and the constitutive parameters for this case are summarized in Table 5.3. Overall, there is also good agreement between the calculated and measured peak undrained shear strength versus sample orientation of St. Vallier clay (see Figure 5.10). From Figure 5.10, it can be seen that the undrained shear strength of St. Vallier clay is highly anisotropic. The peak undrained shear strength of vertical specimens, /=0, is 1.8 times that of the horizontal specimens (i=90), which is a significant difference. Figure 5.11 compares the calculated and measured deviator stress and excess pore pressure
206 versus axial strain for vertical and horizontal specimens. The overall trends in the measured and calculated data are considered to be consistent. Similar to Gloucester clay (see Figure 5.9), slight differences between the measured and calculated data may also be attributed to the anisotropic elastic response of St. Vallier clay. Figure 5.12 shows the calculated and measured stress paths for triaxial compression tests on specimens at i = 0 and 90. For St. Vallier clay, the elastic anisotropic parameter, (3, is 1.14 . Similarly, Curve 2 in Figure 5.12 shows the stress path calculated using cross-anisotropic elasticity in conjunction with the structured elastic-viscoplastic model for i = 0. Again, the calculated and measured behaviour agree. Thus, it appears that the constitutive framework is able to also account for the variation of peak undrained shear strength of St. Vallier clay versus sample orientation. The effect of strain-rate on the measured and calculated undrained peak shear strength of St. Vallier clay is summarized in Figure 5.13. Referring to Figure 5.13, an order of magnitude increase in the applied strain-rate causes a 15% increase in the peak undrained shear strength of St. Vallier clay. In comparison, the peak strength of
Gloucester clay increased by only 10% for an order of magnitude increase in the applied strain-rate. The increased rate-sensitivity is accounted for by decreasing the exponent, n, in the constitutive model for St. Vallier clay (see Table 5.3). The results in Figure 5.13 further highlight the significant influence of strain-rate on the engineering behaviour of clays from Eastern Canada. The influence of destructuration on anisotropy Figure 5.14 illustrates apparent yield states derived from the anisotropic structurered constitutive model assuming a>0 = 1.52 and A = 0.45. Since the yield stress
207 in an EVP model is governed by strain-rate, the term ' apparent' is used to denote isotaches in general stress space and the apparent yield surfaces depicted in Figures 5.14a and 5.14b correspond to normalized isotaches. From Figure 5.14a, it can be seen that as A increases, the apparent yield surface predicted by the proposed constitutive model becomes increasingly anisotropic. Similarly, in Figure 4.15b, as the structure For
parameter, a> , decreases, the apparent yield surface becomes more isotropic.
comparison purposes, Figure 5.15 shows the effect of destructuration on St. Alban clay. This figure summarizes the influence of anisotropic consolidation (K'0 ranges from 0.5 to 0.6) on the apparent yield surface of St. Alban clay measured using drained triaxial probing tests (Leroueil et al. 1979). From Figure 5.15, it can be seen that as the
volumetric strain increases from 8% to 20% during K'0 consolidation, the apparent yield surface of St. Alban clay becomes more isotropic. The behaviour depicted in Figure 5.15 is consistent with that shown in Figure 5.14b for CD0 =1.52 and A = 0.45. The parameter co0 =1.52 can be estimated from the structured and intrinsic compression curve for St. Alban clay shown in Figure 5.16[197]. Figure 5.16 also demonstrates the destructuration of St. Alban clay with increase of volumetric strain. Based on Figure 5.16, the structured and intrinsic compression curves for St. Alban clay are almost equal for volumetric strains exceeding 20% (e.g. the intact clay is destructured). Figuress 5.15 and 5.16 suggest that destructuration of a natural anisotropic clay can lead to a reduction of its anisotropy even for K\ -consolidation where K'0 =0.5-0.6. Similar behaviour has been reported for Winnipeg clay and Onsoy Clay, by Graham et al. (1983) and Lune et al. (2006), respectively. Furthermore, as shown in Figure 5.6, the undrained peak strength of Gloucester clay exhibits significant anisotropy (e.g. 40%
208
difference for undrained strengths for / = 0 and i = 90); whereas the post-peak strength of Gloucester clay is nearly isotropic (e.g. 7% different for i = 0 and / = 90). This suggests that destructuration in the undrained triaxial compression tests[G98] reduces the strength anisotropy of Gloucester clay.

This chapter presented a constitutive approach for modeling the rate-dependent, anisotropic behaviour of structured clay. The foundation of the constitutive framework is an existing overstress elastic viscoplastic model (Hinchberger 1996; Hinchberger and Rowe 1998), which has been extended using a state-dependent viscosity parameter to account for the effects of clay 'structure' (Hinchberger and Rowe 2005). A tensor approach similar to that described by Boehler (1987), Pietruszczak and Mroz (2001) and Cudny and Vermeer (2004) has been used to incorporated anisotropic viscoplasticity into the model, which has been shown to describe some of the key engineering characteristics of two clays from Eastern Canada. For St. Vallier and Gloucester clay, the effects of strain-rate (or time) and sample orientation are clearly significant and a constitutive framework that can account for these two effects is considered to be desirable. Other factors that may affect the response of natural clay during triaxial compression tests include but are not limited to end effects and strain localization both of which have been ignored. In accordance with Equations [5.16] and [5.17], the damage parameter, b , governs the rate of structural degradation of the clay skeleton. Presently, it is not clear how this parameter may be affected by strain localization and future development should focus on this important issue. However, the model captures most of the anisotropic and time-dependent characteristics of these clays,
209 which are clearly very significant. Based on the analyses and discussions presented above, the following observations and conclusions may be made: (i). The extended Hinchberger and Rowe (1998) model can describe the effect of strainrate and sample orientation on the peak undrained shear strength of Gloucester clay and St. Vallier clay. (ii). The constitutive approach described above can also approximate the nearly isotropic post-peak strength of Gloucester clay during CIU triaxial compression. (iii). The plastic response of St. Vallier clay is more anisotropic ( A = 0.3 and c ui=0 /c ui=90 =1.8 ) than that of Gloucester clay (A = 0.15 and c ui=0 /c ui=90 =1.4 ). However, the opposite can be observed for the elastic anisotropy where (3 = 1.6 for Gloucester clay compared to (5 = 1.15 for St. Vallier clay. As a result, it is
concluded that the degree of elastic and viscoplastic anisotropy may not necessarily be interrelated for structured clay. (iv). For Gloucester clay, it is concluded that both cross-anisotropic elasticity (e.g. Graham and Houlsby 1983; Love 1927) and anisotropic viscoplasticity should ideally be accounted for in a constitutive model for this material. The need to account for the anisotropic elasticity of St. Vallier clay is less evident.
210
References
Boehler, J.P. 1987. Applications of tensor functions in solid mechanics. Springer, Wien. Burland, J.B. 1990. On the compressibility and shear strength of natural clays. Geotechnique, 40(3): 329-378. Carter, J.P., and Balaam, N.P. 1990. AFENA-A general finite element algorithm: users manual, School of Civeil Engineering and Mining Engineering,University of Sydney, Australia. Chen, W.F., and Mizuno, E. 1990. Nonlinear analysis in soil mechanics : theory and implementation. Elsevier Science Publishing Company Inc., New York, NY, U.S.A. Cudny, M., and Vermeer, P.A. 2004. On the modelling of anisotropy and destructuration of soft clays within the multi-laminate framework. Computers and Geotechnics, 31(1): 1-22. Davies, M.C.R., and Newson, T.A. 1992. Critical state constitutive model for anisotropic soil. In Proceedings of the Wroth Memorial Symposium, 07/2707/29/92. Oxford, UK. Publ by Thomas Telford Services Ltd, London, Engl, p. 219. Graham, J., and Houlsby, G.T. 1983. Anisotropic elasticity of a natural clay. Geotechnique, 33(2): 165-180.
211 Graham, J., Noona, M.L., and Lew, K.V. 1983. Yield states and stress-strain relationships in a natural plastic clay. Canadian Geotechnical Journal, 20(3): 502-516. Hinchberger, S.D. 1996. The behaviour of reinforced and unreinforced embankments on rate senstive clayey foundations. Ph.D Thesis, University of Western Ontario, London. Hinchberger, S.D., and Rowe, R.K. 1998. Modelling the rate-sensitive characteristics of the Gloucester foundation soil. Canadian Geotechnical Journal, 35(5): 769789. Hinchberger, S.D., and Rowe, R.K. 2005. Evaluation of the predictive ability of two elastic-viscoplastic constitutive models. Canadian Geotechnical Journal, 42(6): 1675-1694. Hinchberger, S.D., and Qu, G. 2007. A viscoplastic constitutive approach for structured rate-sensitive natural clay. Canadian Geotechnical Journal, ReSubmitted November 2007. Kim, Y.T., and Leroueil 2001. Modeling the viscoplastic behaviour of clays during consolidation: Application to Berthierville clay in both laboratory and field conditions. Canadian Geotechnical Journal, 38(3): 484-497. Law, K.T. 1974. Analysis of Embankments on Sensitive Clays. Ph.D Thesis, University of Western Ontario, London, Ontario. Lo, K.Y. 1972. An approach to the problem of progressive failure. Canadian Geotechnical Journal, 9: 407-429.
Lo, K.Y., and Milligan, V. 1967. Shear strength properties of two stratified clays. American Society of Civil Engineers Proceedings, Journal of the Soil Mechanics and Foundations Division American Society of Civil Engineers, 93(SM1): 1-15. Lo, K.Y., and Morin, J.P. 1972. Strength anisotropy and time effects of two sensitive clays. Canadian Geotechnical Journal, 9(3): 261-277. Love, A.E.H. 1927. A treatise on the mathematical theory of elasticity. Cambridge University Press, Cambridge,England. Marques, M.E.S., Leroueil, S., and de Almeida, M.d.S.S. 2004. Viscous behaviour of St-Roch-de-1'Achigan clay, Quebec. Canadian Geotechnical Journal, 41(1): 25-38. Miura, K., Miura, S., and Toki, S. 1986. Deformation behavior of anisotropic dense sand under principal stress axes rotation. Soils and Foundations, 26(1): 36-52. Muhunthan, B., Cudny, M., and Masad, E. 1996. Fabric effects on the yield behavior of soils. Soils and Foundations, 36(3): 85-97. Newson, T.A. 1998. Validation of a non-associated critical state model. Computers and Geotechnics, 23(4): 277-287. Oda, M., and Nakayama, H. 1989. Yield function for soil with anisotropic fabric. Journal of Engineering Mechanics, 115(1): 89-104. Oztoprak, S., and Cinicioglu, S.F. 2005. Soil behaviour through field instrumentation. Canadian Geotechnical Journal, 42(2): 475-490. Perzyna, P. 1963. Constitutive equations for rate sensitive plastic materials. Quarterly of Applied Mathematics, 20(4): 321-332.
Pietruszczak, S., and Mroz, Z. 2001. On failure criteria for anisotropic cohesivefrictional materials. International Journal for Numerical and Analytical Methods in Geomechanics, 25(5): 509-524. Pietruszczak, S., Lydzba, D., and Shao, J.F. 2004. Description of creep in inherently anisotropic frictional materials. Journal of Engineering Mechanics, 130(6): 681-690. Roscoe, K.H., Schofield, A.N., and Thurairajah, A. 1963. Yielding of clays in states wetter than critical. Geotechnique, 13(3): 211-240. Rouainia, M., and Wood, D.M. 2000. Kinematic hardening constitutive model for natural clays with loss of structure. Geotechnique, 50(2): 153-164. Sun, D.A., Matsuoka, H., Yao, Y.P., and Ishii, H. 2004. An anisotropic hardening elastoplastic model for clays and sands and its application to FE analysis. Computers and Geotechnics, 31(1): 37-46. Tavenas, F., and Leroueil, S. 1978. Effects of stresses and time on yielding of clays. In Proc of the hit Conf on Soil Mech and Found Eng, 9th, Jul 11-15 1977. Edited by P.C.O.X. ICSMFE. Tokyo, Jpn. Jpn Soc of Soil Mech and Found Eng, Tokyo, pp. 319-326. Tavenas, F., Leroueil, S., La Rochelle, P., and Roy, M. 1978. Creep behaviour of an undisturbed lightly overconsolidated clay. Canadian Geotechnical Journal, 15(3): 402-423. Tobita, Y. 1988. Yield condition of anisotropic granular materials. Soils and Foundations, 28(2): 113-126.
214 Tobita, Y., and Yanagisawa, E. 1992. Modified stress tensors for anisotropic behavior of granular materials. Soils and Foundations, 32(1): 85-99. Vaid, Y.P., and Campanella, R.G. 1977. Time-dependent behavior of undisturbed clay. Journal of the Geotechnical Engineering Division, 103(7): 693-709. Vaid, Y.P., Robertson, P.K., and Campanella, R.G. 1979. Strain rate behaviour of Saint-Jean-Vianney clay. Canadian Geotechnical Journal, 16(1): 35-42. Wang, C.C. 1970. A new representation theorem for isotropic functions. Archive For Rational Mechanics And Analysis, 36: 166-223. Ward, W.H., Samuels, S.G., and Butler, M.E. 1959. Further studies of the properties of London Clay. Geotechnique, 9(2): 33-59. Wheeler, S.J., Naatanen, A., Karstunen, M., and Lojander, M. 2003. An anisotropic elastoplastic model for soft clays. Canadian Geotechnical Journal, 40(2): 403418. Whittle, A.J., and Kavvadas, M.J. 1994. Formulation of MIT-E3 constitutive model for overconsolidated clays. Journal of Geotechnical Engineering, 120(1): 173198. Zdravkovic, L., Potts, D.M., and Hight, D.W. 2002. The effect of strength anisotropy on the behaviour of embankments on soft ground. Geotechnique, 52(6): 447457. Zienkiewicz, O.C., and Cormeau, I.C. 1974. Viscoplasticity - plasticity and creep in elastic solids - a unified numerical solution approach. International Journal for Numerical Methods in Engineering, 8(4): 821-845.
Table 5.1
Comparison of elastic-viscoplastic models
Hinchberger and Rowe (1998)

K=(l + e)ar' m
IK
Hinchberger and Qu (2007)

a n d G,V
Present Chapter
Elastic Model
Yield Function
f = CJ^M+C^C - V ^ T = " N / c Critical State
f=(<-lf-2J 2 R 2 + (<;-lf=0
Limit Surface f=CT^M+< c -7^7 = 0 -O/C Yield 1

my os my os my
Flow Function, 4>(F) -1

my
my
os
-1
MM
rK.)
M^7)
<j;my
Hardening Law
_ ( ! + ) -C (X-K)
to
Table 5.2
Constitutive parameters for Gloucester Clay Value 1.8 48.5 30 1.20 0.3 0.02 0.63 2.5 0.5 0.15 1E8 50 1.18
Parameter Initial void ratio, e tic Yield Surface Intercept o'^ , (kPa) Visoplastic Strain-rate Exponent, n Mstruc
V
Recompression Index, K Compression Index, X Aspect Ratio of Elliptical Cap, R A (weighting Parameter) Anisotropy Parameter A Structured viscosity, |j, str , (min) Damage Exponent, b Structure parameter, co0
Table 5.3
Constitutive parameters for St.Vallier Clay Value 1.6 70 25 1.85 0.3 0.01 0.65 1.8 0.5 0.3 2.5E9 200 1.37
Parameter Initial void ratio, e tic Yield Surface Intercept o'^, (kPa) Visoplastic Strain-rate Exponent, n
M s truc
Recompression Index, K Compression Index, A Aspect Ratio of Elliptical Cap, R A (weighting Parameter) Anisotropy Parameter, A Structured viscosity, /ustr, (min) Damage Exponent, b Structure parameter, co0
218 Figure 5.1 Illustration of the microstructure tensor, al, and the generalized stress tensor, ov for transverse isotropy. (a) Structure Tensor with positive A (b) Structure Tensor with negative A (c) Generalized stress state
(a)
(b)
(C)
219
Figure 5.2
Sample orientation, /.
i = 45
"7
/
a.
/ = 60
1 / \
1/ Y
4
&
H* d LW
/ = 90
...- .1
.3
The effect of A on the anisotropy of c u from Tresca's failure criterion.
2 h
/: Orientation angle 7]: Anisotropy scalar
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
T) i=90
221 Figure 5.4 The effect of stress ratio, (5'al<5'c, on the anisotropy of c u from Tresca's failure criterion.
1.4
o
3
1.2
o
II
_cc
CO
1.0 - -
< ^ -~/F^'~-.
>> 0.8
Q. O
<n 'F c
0.6-
<
0.4
n
0.2
i=0 , A - 0.2 i=45, A = 0.2 i=90,A = 0.2

i i i
0.0
10
Vertical Stress Ratio, a'fa'
222 Figure 5.5 Conceptual behaviour of the 'structured' soil model.
fij2
Structured Overconsolidated Yield envelope / /
Af =1.2
Destructured Critical State Line (Classical)
M = 0.9
Associated Plastic Flow
Dynamic yield surface for point '3' (Expands to states of overstress e.g. '3'defining dFfdcj'y )
my
my
(a)
/277A
(b)
(c)
ure 5.6
The effect of sample orientation, i, on the measured and calculated peak and post-peak undrained strength of Gloucester clay.
30
25
_o-~-
, Measured (Lavy 1974)
20
Calculated
15 Intrinsic strength at ea=20%(assumed)
10
r-
-20
-10
10
20
30
40
50
60
70
80
90
Orientation angle, i
224
Figure 5.7
The effect of sample orientation, i, on the measured (Law 1974) and calculated (a) axial stress versus strain and (b) excess pore pressure versus strain for Gloucester clay. (a)
cfl Q.
60 - i=0 Measured - i=30 Measured - i=45 Measured 50 o i=90 Measured
60
- i=0 Calculated - i=30 Calculated - i=45 Calculated 50 T i=ou uaicuiaiea i=90 Calculated
40
40
30
30
20
20
10
10
10
12
10
12
Vertical Strain (%)
Vertical Strain (%)
(b)
30
20
10
10
12
10
12
Vertical Strain (%)
Vertical Strain (%)
225 Figure 5.8 The comparison for sample orientations, /, of 0 and 90 on the measured (Law 1974) and calculated[i99] (a) axial stress versus strain and excess pore pressure versus strain (b) stress paths for Gloucester clay.
(a)
60
O
C O Q.
o 50
-*_
<r> D
irf\
\-0 Measured i=90 Measured i=0 Calculated i=90 Calculated
Q.
40
^vj
30
"~"-~^ft== ^ *
</> CO 2 a.
"îr
So
I?
"* ~
o a.
<n o>
CD
20
a
10
HI
10
12
10
12
Vertical Strain (%)
Vertical Strain (%)
(b)
226 Figure 5.9 The effect of strain-rate on the peak strength[ilOO] of Gloucester clay (Data from Law 1974).
1 -i
'
^p^
rengt ratio
.9
Calculated -^
Lab CAU * s ^ LabUU
-c
^\a=Un=0.033 o 1
To .8
Und rained shear
o
.7
.6 10- 5
io-<
10 3
10"2
10-1
10
Strain rate, %/min
227
Figure 5.10
The effect of sample orientation, i, on the peak strength of St. Vallier clay during CIU triaxial compression tests.
140 S. P
CD
120 100
80 60 40 4 20 4
Calculated
co
Measured J r o m i o and MorirL(1912X
(/)
CD .C
co
CD Q_ "O CD C 'CO
45
90
228 Figure 5.11 The effect of sample orientation, /, on the measured (Lo and Morin 1972) and calculated (a) axial stress versus strain and (b) excess[G101] pore pressure versus strain for St. Vallier clay.
180
Calculated,i=0
Measured, r=Q
CO
a.
en
D
i
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Vertical Strain (%)

o Calculated i=90
MoaGuk)d,i=0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Vertical Strain (%)
229
Figure 5.12
The effect of sample orientation, i, on the measured (Lo and Morin 1972) and calculated stress paths for St. Vallier[G102] clay.
iiu O
120 -
<?l
9=2.55 in (ar a3) - a m ' space p = 1.14 = E v / E h ( See Graham and Houlsby, 1983)
100 -
I
Calculated i=0 J W
1
/
Curve '2' (Cross-anisotropic elasticity)
80 -
60 -
K./L* ; < ~
_jfjpf ii
1/
Calculated i=90 40 -
t.
[
I
i
0 Measured Stress Path i=0 --*Measured Stress Path i=90
20 -
0 -
50
100
150
200
a ', kPa
230 Figure 5.13 Measured and calculated peak undrained shear strength versus strain-rate for St. Vallier clay (/=0).
85
!
CO
ou
D.
4-
lear treng
75 -
j
70 -
C O
C O
T3 CD C CO
65 -
GaJeiilatedMeasured
60
by Lo aniMoriri (1972)
55
10":
10'
10-3
10-2
Strain rate, %/min
Figure 5.14
Influence of A and co on apparent yield surface
(a[G103]) Variation of A
V2-V<"
A For a given structure parameter 6)0=1.52
1.32Su Su
(b[G104]) Variation of m
Julio*.
T
my
For a given anisotropic parameter
A = 0.45
1.32Su Su
M=0.91
C00 = 1 . 5 2 (Intact state) CO = 1.31 Partly G) = 1.1 -T destructured
CO = 1 (Intrinsic state)
<7' lo'w
m
my
1.0
Figure 5.15
Influence of destructuration on the apparent yield surface of St. Alban clay
^2/a'(d)
my
1.4
a 1c
Surfaced Surfaced Surface ;3 53kPa 58kPa 80kPa
K' 0.5 0.6 0,5
Sv/V 8% 14% 20%
1.2 4
Surface! 1(53KPa)
^yv
o\c : Vertical consolidation stress K'0 : Consolidation stress ratio, cr'1c/a'3c. 5vA/: Volumetric compression strain in consolidation
Compression curves from oedometer compression tests on intact and destructured specimens of St. Alban clay
(J i p-i
<j _ _ p-s
10
100
a' p . s =35kPa
a j=23kPa
=^C3TT\ ^
\
>o
-10
, ::_ _a
' , _ ! G P-S' P-\ -1-52
\"
\ Intact sample
Remolded sample
-15
-20 4
-25
10
100
Mean effective stress (o'm), kPa, in log scale
234
CHAPTER 6
CASES STUDY OF THREE DIMENSIONAL EFFECTS ON THE BEHAVIOUR OF TEST EMBANKMENTS
6.1 Introduction
Trial embankments or test fills are usually constructed to assist in the design of important embankments on difficult foundations (Dascal et al. 1972; Tavenas et al. 1974; La Rochelle et al. 1974). In some cases, these structures are built to investigate
geotechnical technology such as geosynthetic reinforcement (Rowe and Soderman 1984; Alfaro and Hayashi 1997; Rowe et al. 1995) or prefabricated wick drains (Crawford et al.1992; Burgado et al. 2002). Since trial embankments are often heavily instrumented and the foundation soils extensively investigated, these structures make interesting cases for researchers and engineers to study. For most studies involving trial embankments, the geometry is usually simplified to the two-dimensional (2D) plane strain case neglecting the three-dimensional (3D) geometry and its effect (Tavenas et al.1974; Indraratna et al.1992; Crawford et al. 1995; Zdravkovic et al. 2002). Thus, it is important to evaluate to what extend 3D effects may influence the behaviour of test fills so as to improve the interpretation of their behaviour and performance. This chapter uses the finite element software ABAQUS to investigate 3D effects on the behavior of tiiree full-scale test fills: the St. Alban test embankment (La Rochelle et al. 1974), the Malaysia trial embankment(MHA 1989b; Indraratna et al. 1992), and the Vernon test fill (Crawford et al. 1995). Back analysis of these cases highlights some important considerations in the design and interpretation of test embankments. From this
A version of this chapter has been submitted to Canadian Geotechnical Jouranl 2007
235 study, a shape factor commonly used in bearing capacity calculations is evaluated for use with 2D embankment collapse calculations. The analysis and evaluation described in the following sections should be of interest to geotechnical engineers and researchers involved in the study of embankments on soft soils.
6.2 Methodology
In this chapter, the 2D and 3D behaviour of embankments built on soft cohesive soil was studied using the elasto-plastic finite element software ABAQUS. In each case, the embankment fill and foundation soils were discretized using linear eight-node brick elements and four-node plane strain elements for 3D and 2D analysis, respectively. The typical finite element mesh comprised a rough rigid boundary which was extended at the bottom of the soft foundation soil and smooth rigid boundaries on its lateral sides. Some typical meshes are illustrated throughout the paper. The foundation soil was modeled as an elastic-perfectly plastic material. The constitutive parameters include: undrained strength (c u ), undrained elastic modulus( Eu), Poisson ratio( v), bulk unit weight( 7), and coefficient of earth pressure( K0) defined in terms of total stresses. A typical undrained strength profile comprised a crust underlain by soft soil layers. In all cases, the foundation clay was assumed to have a constant Eu for each layer and failure was assumed to be governed by the Mohr-Coulomb failure criterion with (j)u = 0, cu varying with depth, and a dilation angle, y/, of 0. Typical foundation layers were idealized as having an undrained shear strength, cu0, at the top of the layer and a gradient of cu with depth, pc . The fill was also modeled as an elastoplastic material with a constant Young's modulus('), Poisson ratio(v = 0.3), bulk
236 unit weight( y), effective friction angle, ((/)') , and cohesion intercept, (c'), and dilation angle(i^). Soil properties for each case are summarized in Table 6.1 and discussed below. The construction process was numerically simulated by activating both the weight (body force) and stiffness of fill elements layer by layer using an incremental, iterative, and load-adjusting solution scheme to ensure convergence of elastoplastic solutions.
6.3 St. Alban Test Embankment Case 6.3.1 Introduction

In 1972, the geotechnical research group of Laval University built four test embankments at Saint Alban to investigate the behavior of embankments on sensitive Champlain clay. Deposits of Champlain clay are widespread in eastern Canada. In this case, three of the four test embankments were built with different side slopes to study the influence of slope inclination on the deformation behavior (La Rochelle et al. 1974). One test embankment was constructed to failure to study the collapse of embankments on sensitive soft clays. This case was used by Zdravkovic et al.(2002) to demonstrate the effect of strength anisotropy on the embankment behaviour using the 2D MIT-E3 constitutive model (Whittle and Kavvadas 1994) which requires 15 material parameters. In contrast, this chapter investigates 3D effects using a simple undrained strength profile and elastoplastic analysis. The St. Alban case has also been investigated by Trak et al.(1980), who used 2D limit equilibrium analysis to show that the factor of safety of test embankment 'A' at failure(La Rochelle et al. 1974) was 1.20 and 0.93 using the vane strength profile and the
237
relationship cu - 0.22cr' , respectively.
From this assessment, Trak et al. (1980)
concluded that using cu = 0.22c'' for the strength profile was slightly conservative in this case. It is noted that 3D geometric effect was neglected in the limit equilibrium analysis; however, in this chapter, 2D and 3D FEM analyses are performed to investigate the St. Alban test embankment. 6.3.2 Soil Conditions The foundation at St. Alban comprised a 1.5m-thick weathered crust and a 8m thick layer of soft silty marine clay, which was underlain by a 4m thick soft clayey silt. Below the clayey layers was a deposit of fine to medium sand extending to 24.4m depth. For subsequent analysis of this case, undrained conditions were assumed due to the relatively short period of construction (10 days from Oct. 4 to Oct. 13, 1972) and the low permeability of Champlain clays ranging from 10~10 m/s to 10~9 m/s (Tavenas et al. 1983). As shown by Tavenas et al. (1983), such an assumption is not strictly correct near the drainage boundaries of the clay but should be adequate to assess the relative behaviour of 2D and 3D embankments. The Champlain clay deposit has been studied utilizing in situ field vane tests (17 in situ vane tests), cone tests ( 8 static cone tests), and laboratory tests on tube and block samples (La Rochelle et al. 1974; Tavenas et al. 1974; Leroueil et al. 1979). As pointed out by La Rochelle et al. (1974), the vane and cone tests across the site at St. Alban indicated that there was no significant variation of undrained shear strength horizontally across the site but a relatively typical variation of undrained shear strength with depth. The clay sensitivity varied between 14 and 22. The measured undrained strength from field vane and triaxial tests are shown in Figure 6.1, together with the undrained strength
238
profile used in subsequent FE analysis in this case (the solid line in Figure 6.1). The undrained strength profile for the 1.6m-thick weathered crust layer was corrected to 15kPa due to the probable existence of fissures. The undrained shear strength was then assumed to increase at a rate of 2.1kPa per meter from 7kPa at a depth of 1.5m. The assumed undrained elastic modulus and bulk unit weight are summarized in Table 6.1. The fill materials consisted of uniform medium to coarse sand with an effective friction angle, <p', of 44 based on drained triaxial tests (La Rochelle et al. 1974). For analysis, the dilation angle, y/, was assumed to be half of the effective friction angle.
6.3.3 Geometry A plan view and cross section of the test embankment are shown in Figure 6.2. The ratio of crest length to width was 4:1 at the designed height of 4.6m and the corresponding ratio of length to width at the base was about 2:1. The right (front) side slope in Figure 6.2a was 1.5H:1V and the other slopes were 2H:1V. A 1.5m high berm was placed on the left side and at the ends to ensure the failure occurred on the right side. The test embankment failed at a fill thickness of 4.0m before reaching its design height. For analysis, the 2D analysis considered the central cross section A-A as shown in Figure 6.2b;whereas, the 3D analysis took into account symmetry and thus only half of the trial embankment was modeled. The 2D and 3D finite element meshes are shown in Figure 6.3a and 6.3b. 6.3.4 Results As shown in Figure 6.4, the calculated failure fill thickness for 2D and 3D FE analysis were 3.6m and 4.0m respectively, based on the vertical displacement curves of the embankment at a central point "O" (See Figure 6.2). Despite the 10% difference
239 between predicted failure thicknesses, the magnitude of settlement from both 2D and 3D analysis at the same fill thickness are very similar. The contour of spatial displacement at failure is shown on the 3D model in Figure 6.5, which indicates that the development of failure in the longitudinal direction is restricted by the length of the fill (due to the stabilization effect of both end- slopes). In Figure 6.6, the observed and computed extents (plan view) of failure are compared, and the agreement is reasonable. It can also be seen that the mobilized failure mass for the St. Alban embankment has strong 3D characteristics. As such, failure of the St. Alban test fill is 3D in nature and use of a 2D model results in underestimation of the failure thickness by about 10% compared with that obtained by 3D analysis. It is interesting to note that 3D analysis yields a higher factor of safety than 2D analysis for the St. Alban case. As a result, cu = 0.22<r'p investigated by Trak et al. (1980) is actually a reasonable estimate of strength profiles, considering that 3D analysis would give a factor of safety 10% higher than that obtained by Trak et al. (1980) using 2D analysis (0.93).
6.4 Malaysia Trial Embankment Case 6.4.1 Introduction

The Malaysia trial embankment was built to failure at Muar flat in the valley of the Muar River in Malaysia. Muar clay is a very soft clay which caused frequent
instability problems during construction of the Malaysian North-South Expressway. The Malaysia trial embankment was built between 27th Oct. 1988 and 4th Feb. 1989. The fill was placed at a rate of about 0.4m/week until it failed at a fill thickness of 5.4m. The Malaysia case was fully instrumented and well documented (MHA 1989a).
240 A series of comprehensive field and laboratory tests were carried out before the embankment construction, which provided parameters for researchers and engineers to predict the embankment behavior. An International Symposium entitled: "Trial
Embankment on Malaysian Marine Clays" was held in November 1989 and 31 class A' predictions of the embankment performance were received from experienced researchers and engineers (MHA 1989b); each employing different methods of analysis ranging from stability charts and limit equilibrium analysis, to undrained and drained finite element analysis (e.g. Brand and Premchitt 1989). In one case, a centrifuge model test was used (Nakase and Takemura 1989). Subsequent to the symposium, Indraratna et al. (1992) reported a calculated failure thickness of 5.0m using a modified Cam-clay model and 2D FEM analysis. All predictions of the failure thickness are summarized in Figure 6.7, where it can be seen that there was a wide variation of predicted failure thickness ranging from 2.8m to 9.5m. The majority of predictors underestimated the failure thickness: the average predicted failure thickness was 4.7m, whereas the actual failure thickness was 5.4m. This discrepancy reflects the difficulty of geotechnical prediction and also
suggests there may be some characteristics that have not been fully explored by the predictors.
6.4.2 Soil Conditions Figure 6.8 summarizes the Malaysian soil profile. As reported by MHA( 1989a), the subsoil consists of a 2m-thick weathered crust underlain by a 6m-thick deposit of very soft silty clay and a 10m thick layer of silty clay. The upper clay deposits overly a 0.5m peat layer, 3.5m sandy clay, and then dense sand. According to field and laboratory tests, the undrained strength increases linearly with depth below the weathered crust.
241 For the analysis of this case, the undrained strength of the crust was corrected to one third of the field vane strength to account for the likely presence of fissures (Lo and Hinchberger 2006). The engineering parameters of the fill and foundation subsoil used in the analysis are summarized in Table 6.1.
6.4.3 Geometry A plan view and cross-section of the Malaysia trial embankment are presented in Figure 6.9a and 6.9b. With respect to the designed thickness of 6m, the ratio of crest length to width was 2, while the aspect ratio at the base of the main fill was 1.4. Similar to the St. Alban case, a 2.5m high berm was placed around three sides of the fill to force the failure toward one side.
6.4.4 Results Figure 6.10a shows the calculated fill thickness versus vertical displacement at point "O" (See Figure 6.9) beneath the center of the embankment. From this Figure, it can be seen that the displacement increases linearly as the fill thickness increases during the initial stage of construction. When approaching to the critical height, however, a small increase in fill thickness results in large displacement. As shown in Figure 6.10b, initially the net fill height increases with fill thickness until reaching a maximum value at a critical point, where upon it decreases with the addition of fill indicating that the incremental vertical displacement exceeds the corresponding increment in fill thickness. When further loads are applied, the net fill height decreases and full collapse occurs. Thus, the state with the maximum net height is considered as a critical state and the corresponding thickness is taken as the calculated failure thickness. According to Figure 6.10, the calculated failure thickness for 2D and 3D analysis are 4.2m and 5.2m
242
respectively. Therefore, in the case of Malaysia test fill, consideration of 3D geometric effects results in a 20% increase in the calculated failure thickness relative to the 2D analysis. In Figure 6.10a, the deformation curve of the 2D model follows closely with that of the 3D model until failure occurs. It appears that in this case the 3D geometry does not significantly influence the deformation prior to imminent collapse. This is consistent with the fact that the calculated settlement by predictors using 2D analysis agree well with the measured data of the Malaysia trial embankment in spite of the relatively large discrepancy of predicted failure thickness (Brand and Premchitt 1989). Figure 6.11 and 6.12 show the velocity fields at failure for 2D and 3D models, respectively. The trend of movement of the failure mass is shown by the direction of the velocity vectors, and the length of velocity vectors represents the relative magnitude of movement. Both 2D and 3D failure surfaces were estimated based on the velocity fields at the respective failure thickness ( 4.2m for 2D analysis and 5.2m for 3D analysis). As shown in Figures 6.11 and 6.12, the estimated failure surfaces for 2D and 3D analysis are generally comparable though 2D and 3D models predicted different failure thickness. As shown above, the failure thicknesses predicted using 2D and 3D analysis of the Malaysia trial fill differ by about 20%. Thus neglecting 3D geometric effects will lead to underestimation of the failure thickness. From an engineering point of view, it seems to be conservative to adopt the plane strain assumption. However, to evaluate the strength of subsoil based on the behaviour of a trial embankment, the 2D model will, in return, lead to an overestimation of the available foundation strength profile, which could lead to inadequate designs for long embankment on such soft soils. This will be
243 discussed and highlighted further during evaluation of the Vernon case below. 6.5 The Vernon Case 6.5.1 Introduction The final case considered is the Vernon embankment presented by Crawford et al. (1992, 1995). In this case, two consecutive failures occurred during construction of an approach embankment on soft clay in British Columbia. The embankment failures
occurred in spite of the fact that two test fills were built successfully on either side of the failures and that the test fills were higher than the approach embankment that failed. Figure 6.13 shows a site plan of the Vernon approach embankment and the location of the two test fills. The Vernon case comprised the West Abutment Test fill which was constructed to a maximum fill thickness of 11.5m, with wick drains in the foundation; and the east test fill or Waterline Test Fill which was built to a maximum fill thickness of 12m, without wick drains. Both test fills were constructed in 1986 and remained stable for approximately 3 years before being incorporated into the Vernon approach embankment. Figure 6.14 shows a cross-section of both the Waterline and West Abutment Test fills. Since the wick drains may influence the test fill behaviour, only the Waterline test fill is selected for comparative analysis with the Vernon approach embankment. Construction of the Vernon approach embankment commenced in early December 1988 and progressed slowly to a fill thickness of between 7m and 9.5m by June 30th, 1989. At this time, the embankment failed (first failure) on the north side encompassing a portion of the West Abutment Test Fill. The extent of the first failure is shown in Figures 6.13 and 6.14. At the time of the first failure, the West Abutment Test
244
Fill had been in place for approximately 3 years, and according to the results of monitoring, the excess pore pressures generated during construction of this test fill had dissipated (see Crawford et al. 1992). The failed approach embankment was redesigned with 5m thick and 30m wide berms on both sides of the original embankment and reconstruction commenced in August 1989 at a very slow rate. In March 1990, a second failure occurred that was much larger in extent and included most of the first failure. The second failure occurred at a fill thickness of about 11.2m and it involved both sides of the approach fill. The extent of the second failure is also shown in Figures 6.13 and 6.14. The approach embankment was eventually completed using berms and lightweight fill; however, the case raises an obvious but perplexing issue: In what way were the results of the two test fills misleading?
6.5.2 Analysis The subsurface conditions in the Vernon Case are summarized in Figure 6.15. In the Vernon case, the foundation soils comprised about 4m of interlayered sand, silt and clay underlain by a 5m thick crust comprising stiff to very stiff clay then a deep deposit of soft to firm silty clay. Figure 6.15 summarizes the results of field vane tests done in 1960 and 1985 in addition to the undrained strength profiles investigated in this study. For the purpose of analyzing the Vernon case, the undrained strength of the crust was reduced to 40kPa in accordance with Lo (1970) and Lo and Hinchberger (2006) to account for the probable effect of fissures on the mass strength of the crust. The
undrained strength was assumed to be constant at 40kPa from the ground surface to 6m deep then it was assumed to decrease linearly from 6m to a depth of 9m below which the
245
strength increased linearly with depth.
Three different strength profiles were
investigated: Profiles L(0.84M), M and H(1.08M) which denote lower, middle and upper strength profiles (see Figure 6.15). The L and H profiles are 84% and 108% of the M profile as a whole. Lo and Hinchberger (2006) studied the 3D effect in this case using 2D axisymmetric FE analysis, where the 3D geometry of test embankment was simplified as axisymmetric. In addition, the three profiles used by Lo and Hinchberger (2006) have the same crust strength but slightly different strength for the soft clay underlying the crust. This chapter utilizes a real 3D model to account for the geometry characteristic of the test embankments. Table 6.1 summarizes the material parameters used in the analysis. foundation clay was modeled as an undrained material with a unit weight of 16 The kN/m3,
friction angle (pu = 0, and undrained shear strength, cu, that varied with depth (see Figure 6.15). The fill was considered to be a drained material with a unit weight of 20.4 kN /m 3 , and effective friction angle, <f>', of 30, in accordance with that reported in the case (Crawford et al. 1992, 1995) The Vernon case was studied using both 2D and 3D finite element analysis as described in the following: (i) The first failure was evaluated using 2D and 3D finite element analyses to assess the mobilized undrained strength of the clayey foundation and to investigate the role of 3D effects on the approach embankment performance, (ii) Next, the Waterline Test Fill was analyzed using 2D and 3D finite element analyses. The backcalcualted strength profiles by 2D (H) and 3D (L) analysis of the Waterline Fill are compared with the mobilized strength obtained from the first failure of the Vernon
246 approach embankment. The purpose of these analyses was to investigate the degree to which 3D effects may have affected the performance and consequent lessons learned from the failure of the Vernon approach embankment.
6.5.2 Results of Vernon Approach Embankment As shown in Figure 6.13 and 6.14, the Vernon approach embankment was constructed between the Waterline test fill and West Abutment test fill. In order to evaluate the first failure, Station 27+80 was considered for 2D analysis since it is situated at the midpoint of the first failure. Figure 6.16 shows the calculated failure thickness (8.2m, 9.8m, and 10.8m) for the L-profile, M- profile, and H-profile respectively. Compared with the actual failure thickness of 9.9m, the M-profile provides the best fit and it is thus considered to be the approximate mobilized strength profile for the first failure from 2D analysis, notwithstanding that there could be other interpretations. However, the assumed plane strain condition of the 1st failure may not strictly satisfy the actual condition for the Vernon approach embankment. As shown in Figure 6.13 and 6.17, the height and width of the approach embankment increase toward the bridge site. The longitudinal slope of the embankment crest was also about 3.2% (See Figure 6.14). Thus, each cross section in the approach embankment varied geometrically and consequently the degree of divergence from a plane-strain condition is unknown. In light of this, a 3D analysis was done to compare with the 2D analysis discussed above and to explore to what extent the first failure may have been affected by 3D effects. Accordingly, the true 3D geometry of the approach embankment was modelled as shown in Figure 6.17. The crest width was constant at 22m with 1.5H:1V side slopes.
247
The crest aspect ratio, length/width, was approximately 9.8 and the average base aspect ratio was 4.2. The plan view and cross sections of both ends are shown in Figure 6.17. Figure 6.18 compares the results of 3D analysis and 2D analysis using the Mprofile. The predicted failure thickness from 3D analysis was 10.3m, which was 0.4m higher than the 2D prediction. Since the calculated failure thickness of the Vernon
approach embankment is only 4% higher for the 3D case compared to that calculated for 2D analysis, it is concluded that the choice of Station 27+80 for 2D analysis of the first failure was acceptable. Figure 6.19 shows the displacement contours of the 3D model at the failure thickness, together with the vectors indicating the direction and relative magnitude of movement of the ground surface. The extent of the calculated failure mass is between station 27+35 and station 28+20 and spreading about 50m outward from central line. Referring to Figure 6.13 and 6.14, the observed limit of the first failure agrees well with that calculated by 3D analysis.
6.5.3 Results of Waterline Test Fill Since the Waterline fill was used to conclude the final approach embankment would be stable, the performance of this test fill was analyzed in detail. For 2D analysis, the central cross section of the Waterline test fill was considered because of its symmetrical geometry. Figure 6.20 shows the geometry considered. The measured and calculated displacement of centre point 'O' below the Waterline fill are presented in Figure 6.21. From Crawford et al (1995), the measured vertical
displacement curve was essentially linear, which indicates that the behaviour of the Waterline fill was predominantly elastic. The predicted failure thickness from 2D
248
analysis were 8.3m for the L-profile, 10.8m for the M-profile, and 11.8m for the Hprofile, respectively. Considering that the Waterline test fill was stable at a thickness of 11.8m, the 2D analysis suggests that the H-profile is the lower bound strength available in situ. The M-profile, however, was back calculated from the first failure of the Vernon approach embankment and this discrepancy warrants further investigation. Accordingly, a 3D model was undertaken to account for the geometry of the Waterline test fill. The results of the 3D analysis are presented in Figure 6.22. From Figure 6.22a, it can be seen that the predicted failure thickness is 11.8m using the Lprofile and that the embankment is stable at 11.8m for both the M- and H-strength profiles. Thus, it can be deduced from the 3D analysis that the L-profile is a lower bound for the available in situ foundation strength. Based on the analysis and discussion above, there is a consistent interpretation of the Vernon case. If the M-strength profile shown in Figure 6.15 is adotped, then a 3D analysis indicates that the Waterline test fill is stable at a fill thickness of 11.8m whereas the approach embankment fails at a fill thickness of 9.8m. This is what was observed in this case. A 2D-analysis on the Waterline test fill, however, yielded the H-profile as the lower bound strength profile, which may lead to an inadequate design for the approach embankment. The FE analyses suggest that 3D effects may have contributed to the failure of the approach embankment before reaching the height of the adjacent Waterline Fill, notwithstanding that natural soil variability may have also played a role. From the FE analysis, the ratio of 3D collapse thickness of the Waterline test fill to the 2D collapse thickness is 1.4, which is significant.
249
6.6 Discussion
From the detailed analysis of the above cases, the base aspect ratio (L/B) of length over width can be utilized to represent the 3D geometry of test embankments. In
addition, the ratio of the calculated failure thickness by 3D and 2D FE analysis (Hf3D I Hf2D) can be used to quantify 3D effects. base aspect ratio (L/B) in Figure 6.23. represented by Hf3D/Hf2D Hf3D I Hf
2D
is plotted against the
As shown in Figure 6.23, the 3D effect
are inversely proportional to the base aspect ratio. It is
qualitatively consistent with the shape factor equation utilized by Skempton (1951) to account for geometric effects on the bearing capacity of spread foundations, e.g.: ^ ^
Quit,2D
=1+
L I B
[6.1]
where L and B are the length and width of the foundation; qultiD and qult 2D are the ultimate bearing capacity of a rectangular foundation and the bearing capacity of a infinitely long foundation, respectively. Equation [6.1] is plotted in Figure 6.23 for comparison with the St. Alban, Malaysia, and Vernon cases. For the Vernon approach embankment and the St. Alban test embankment with aspect ratios (L/B) equal to or larger than 2.0, the corresponding case points in Figure 6.23 plot very close to Equation [6.1]. This suggests that 3D geometry effects on test embankments with a aspect ratio greater than 2 are reasonably close to those deduced from Equation [6.1]. For the cases of Malaysia trial fill and the Waterline test fill, the difference between the predicted failure thicknesses by 2D and 3D analysis increases from 20% up to 40% as the aspect ratio decreasing from 1.4 to 1.2. The corresponding points representing these two cases in Figure 6.23 (from 2D and 3D
250 FE analyses) lie above Equation [6.1], indicating that 3D effects on test embankments with an aspect ratio less than 2 is greater than that expected for foundation bearing capacity. It is noted that other factors such as side slopes and berms of the test fills may also influence the 3D effect to some degree, and that these factors may account for some of the difference noted in Figure 6.23.
6.7 Summary and Conclusion

Three full-scale test fills have been evaluated utilizing finite element analysis (ABAQUS) accounting for 2D and 3D geometries, respectively. resulting from these case studies are summarized as follows: Considerable difference (10% to 40%) of die predicted failure thicknesses obtained by 2D and 3D analysis was found for all three cases. Considering 3D geometry results in an increase in the predicted failure thickness of test fills. This finding is qualitatively consistent with the shape factor equation in bearing capacity theory (Skempton, 1951) for base aspect ratios (L/B) greater than 2. However, when the base aspect ratio is less than 2, the 3D effect on test embankments becomes considerably greater than that suggested by bearing capacity theory. 3D analysis agrees better with the field behaviour. Beside the influence of 3D geometry, the calculated extent of failure by 3D analysis agrees fairly well with the field observations in the St. Alban and Vernon cases. Assuming plane strain conditions in the analysis of test fills may potentially lead to the overestimation of the available soil strength and consequently inadequate design of long embankments on the same site. As shown in Vernon case, neglecting the 3D The key findings
251 geometry and its impact on the behaviour of the Waterline test fill could have been misleading for the design of the long approach embankment. It is recommended that 3D geometry should be considered in the design and interpretation of test fills whose base ratios of length to width are less than 2. In this situation, the influence of 3D geometry should be taken into account to reasonably evaluate the behaviour of the trial fills, including failure thickness, strength profiles, and stress in the reinforcement in the test embankment.
252
References
Alfaro, M.C., and Hayashi, S. 1997. Deformation of reinforced soil wall-embankment system on soft clay foundation. Soils and Foundations, 37(4): 33-46. Bergado, D.T., Fannin, R.J., Holtz, R.D., and Balasubramaniam, A.S. 2002. Prefabricated vertical drains (PVDs) in soft Bangkok clay: A case study of the new Bangkok International Airport project. Canadian Geotechnical Journal, 39(2): 304-315. Brand, E.W., and Premchitt, J. 1989. Comparison of the predicted and observed performance of the Muar test embankment. In Proceeding of the international symposium on trial embankments Malaysia marine. Edited by R.R. Hudson, C.T. Toh, and S.F. Chan. Kuala Lumpur. The Malaysian Highway Authority, Vol.2, pp. 10-18. Crawford, C.B., Jitno, H., and Byrne, P.M. 1994. Influence of lateral spreading on settlements beneath a fill. Canadian Geotechnical Journal, 31(2): 145-150. Crawford, C.B., Fannin, R.J., and Kern, C.B. 1995. Embankment failures at Vernon, British Columbia. Canadian Geotechnical Journal, 32(2): 271-284. Dascal, O., Tournier, J.P., Tavenas, F., and La Rochelle, P. 1972. Failure of test embankment on sensitive clay. In Proceeding of ASCE Specialty Conference on Performance of Earth and Earth-Supported Structures. Purdue University, Lafayette, Vol.1, pp. 129-158. Indraratna, B., Balasubramaniam, A.S., and Balachandran, S. 1992. Performance of test embankment constructed to failure on soft marine clay. Journal of Geotechnical Engineering, 118(1): 12-33.
La Rochelle, P., Trak, B., Tavenas, F., and Roy, M. 1974. Failure of a test embankment on a sensitive Champlain clay deposit. Canadian Geotechnical Journal, 11(1): 142-164. Leroueil, S., Tavenas, F., Brucy, F., La Rochelle, P., and Roy, M. 1979. Behavior of destructured natural clays. Journal of the Geotechnical Engineering Division, 105(6): 759-778. Lo, K.Y. 1970. The operational strength of fissured clays. Geotechnique, 20(1): 57-74. Lo, K.Y., and Hinchberger, S.D. 2006. Stability analysis accounting for macroscopic and microscopic structures in clays. In Proc. 4th International Conference on Soft Soil Engineering. Vancouver, Canada, pp. pp. 3-34. MHA 1989a. Factual report on performance of the 13 trial embankments. In Proceeding of the international symposium on trial embankments Malaysia marine. Edited by R.R. Hudson, C.T. Toh, and S.F. Chan. Kuala Lumpur. The Malaysian Highway Authority, Vol.1. MHA 1989b. The Embankment built to failure. In Proceeding of the international symposium on trial embankments Malaysia marine. Edited by R.R. Hudson, C.T. Toh, and S.F. Chan. Kuala Lumpur. The Malaysian Highway Authority, Vol.2. Nakase, A., and Takemura, J. 1989. Prediction of behaviour of trial embankment built to failure. In International Symposium On Trial Embankments On Malaysia Marine Clays. Kuala Lumpur. November 6-8, Vol.2, pp. 3-1,3-13. Rowe, R.K., and Soderman, K.L. 1985. Approximate method for estimating the stability of geotextile-reinforced embankments. Canadian Geotechnical Journal, 22(3): 392-398.
254 Rowe, R.K., Gnanendran, C.T., Landva, A.O., and Valsangkar, A.J. 1995. Construction and performance of a full-scale geotextile reinforced test embankment, Sackville, New Brunswick. Canadian Geotechnical Journal, 32: 512-534. Skempton, A.W. 1951. The bearing capacity of clay. In Building Research Congress. London. Tavenas, F., Leblond, P., Jean, P., and Leroueil, S. 1983. Permeability of natural soft clays, part I: methods of laboratory measurement. Canadian Geotechnical Journal, 20(4): 629-644. Tavenas, F.A., Chapeau, C , La Rochelle, P., and Roy, M. 1974. Immediate settlements of three test embankments on champlain clay. Canadian Geotechnical Journal, 11(1): 109-141. Trak, B., La Rochelle, P., Tavenas, F., Leroueil, S., and Roy, M. 1980. New approach to the stability analysis of embankments on sensitive clays. Canadian Geotechnical Journal, 17(4): 526-544. Whittle, A.J., and Kavvadas, M.J. 1994. Formulation of MIT-E3 constitutive model for overconsolidated clays. Journal of Geotechnical Engineering, 120(1): 173-198. Zdravkovic, L., Potts, D.M., and Hight, D.W. 2002. The effect of strength anisotropy on the behaviour of embankments on soft ground. Geotechnique, 52(6): 447-457.
Table 6.1
Parameters used in the numerical analysis of the three cases
St. Alban test embankment Soft clay 1.6-40

c0 = IkPa c0 = 25kPa A.=0 c0 = 8kPa pc_ =lA8kPa/m p c> =2.\kPalm cu0 = AQkPa cu0 = 40kPa pCj =-2.67kPa/m
Clay Deposit
Crust
Soft clay 9-80

cu0 = 32kPa pc_ =l.03kPa/m
0-1.6
Malaysia trial embankment Underlying Crust soft clay 0-2 2-40
Vernon case (M-Profile) Transition Crust layer 0-6 6-9
cu0 = 15kPa
A, = 0
Depth, m Undrained Shear Strength, kPa ysat(KN/m3) 17 16.0 0.9 2.5E7 19.8 20 17
19
17
K0 EK(KPa)
1.0 1.5E7
Fill Materials
0'= 44 yr = 22 E'= 2E8kPa c'= 5kPa v = 0.3 ysal=19kN/m3
0.9 1.0 2.5E7 8E6 ^'=31 ^ = 1 5 E'=5.1E6kPa c'=5kPa v = 0.3 y . =20.5kN/m3
1.04 0.85 0.85 1.8E7 1.5E7 2.5E7 ^'=33 ^ = 1 6 E'=\5ElhPa c'=5kPa v = 0.3 ysal=20.5kN/m3
256
Figure 6.1
Strength profile assumed and measured using field vane and undrained (UU and CIU) tests (experimental data from La Rochelle et al. 1974)
Undrained Strength, (kPa) 0 10

20 2m i i i i i r i 30 40 50 60 i i I i ; i i i i i i i I i i i i I I 70
I I I I I
Assumed Strength Profile
Om
-2m
7kPa
o.
<D
-4m
Q In-situ Average Vane Strength -6m 4 In-situ Minimum Vane Strength; In-situ Maximum Vane Strength
-8m 4 ncreasing rate = 2.1 -10m
257 Figure 6.2 Plan view and cross-section of St. Alban test embankment
(a) Plan view
- 4 . 6 i
6.1
7.6
6.9-
\^^^2^ At^*" X ' 6 Point 'O' X ?
(b)
A-A Section
258 Figure 6.3 Generated Meshes for 3D and 2D FEM model
(a) Plane strain analysis
(b) 3D analysis
259 Figure 6.4 Measured and calculated vertical displacement of point 'O' for St. Alban Embankment
o.oo -.02 Hf2D=3.6m B(3D=4.0m

- . 0 4 -
-.06 -.08 -.10 -.12 4-.14 H 0.0 . I 2D analysis 3D analysis R23 Centre Point 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Embankment thickness,m
260 Figure 6.5 Spatial displacement contour of 3D model for St. Alban embankment (at failure)
U, Magnitude + 9.842e-01 +9.022e-01 + 8.202e-01 +7.382e-01 + 6.SSle-01 + S.741e-01 + 4.921e-01 +4.101e-01 +3.281e-01 +2.4Sle-01 +1.640e-01 + 8.202e-02 + 0.000e+00
261 Figure 6.6 Spatial displacement contour V.S. fissures at failure on the top surface on St. Alban Embankment
Projected contours from Figure 6.5 Observed extent of fissures Calculated extent of failure
Measured (La Rochelle et al. 1974)
3DFEM Analysis
...
sm
7*~t X.
262
Figure 6.7
The statistic table for the[il05] prediction on the failure thickness of Malaysia test embankment (data from MHA 1989b)
C/2
pre
a o o a
o ^ 1) ran
11 10 9 8 7 6 5 4 3 2 1 0 4
A.ctual failure thickness = 5.4 m
Predicted thickness, m
263 Figure 6.8 Strength profiles[i 106] for the Malaysia case (experimental data from MHA 1989a)
Dense Sand
i i i i | i i i i [ i i i i I i f i i I i i i i I i i i i | i i
10
20
30
40
50
60
70
Undrained Strength, (kPa)
6.9
Plan view of Malaysia test embankment
h
! _^^
H
_ ^ ^
Main Fill ^ ^ s .
Berm
A-A Section
s
Berm
7 1
Main Fill
Point ' 0 '
Berm
IZ
Plan View
Figure 6.10
Measured and calculated settlement[il07] of Malaysia Trial Embankment
b)
Failure Thickness : Hf 3D = 5.2m
2D analysis 3D analysis 8 Fill Thickness,m
266 Figure 6.11 Velocity field in central cross-section of 2D model for the Malaysia trial embankment (at failure)
5m Scale: [
267 Figure 6.12 Velocity field in central cross-section of 3D model for the Malaysia trial embankment (at failure) Scale: p ^
Velocity Field on A-A section
268 Figure 6.13 Plan view of Vernon embankment (modified after Crawford et al. 1995)
LIMIT OF SECOND FAILURE
LIMIT FIRST
STRUCTURE
KEY : 0 &
1960 BORINGS 1985 BORINGS 1990 BORINGS
269 Figure 6.14 Longitudinal section through the embankment (after Crawford et al. 1995)
i
FAILURE 10 MAR. '90
FAILURE 30 JUNE '89
SURFACE, 10 MAR. M 9 0
FINALffAVEMENTSFC.
WATERLINE FILL :X *
WEST ABUTMENT FILL
j.
1...
-A*"
SETTLEMENT PLATE
MAXIMUM SETTLEMENT |WICKS 1986 llllllllllllllllllllllllllllllll
8
+
8
+
WEST
SURVEY STATION
EAST
ure 6.15
Distribution of vane strength [i 108]with depth
Crust Transition Layer
-10
-10
-20 -4Clay Layer H Profile . Strength Profile(M) L Profile
4 -20
-30 +
Measured Vane Strength 1985 (Crawford 1992) Measured Vane Strength 1960 (Crawford 1992)
4 -30
-40 0 20 40 60 80 100
_L
120
J_
140
-40
Undrained Strength, (kPa)
Figure 6.16
Vertical displacement of Vernon Approach Embankment in 2D analysis [G109]
10
E -8
2D ^ n a f y g j g ^ ^ g ^ ^ ^ i Q f i i g j r 2D analysis(M Strength Profile) 2D analysis(L Strength Profile)
*f
4f
Fill Thickness,m
0.5
E
c 0.0
C D
"^^r~Z:Z
***JM*
E -0.5
CD
'A^*^ 2D analysis(H Strength Profile) 2D analysis(M Strengthi Woffle!)'' 2D analysis(L StrengthProfile)
co
Q. C O
-1.0 -1.5 -2.0
b
"O C
_o
V-4 L.
C D
>
* -2.5
measured data (Crawford et al.,1995)

-i
10
12
Fill Thickness,m
272 Figure 6.17 Plan view and 3D model of Vernon approach embankment[G 110]
2^
Station 26+20
Station 27+8C
Figure 6.18
Vertical displacement of Vernon Approach Embankment in 3D analysis [ G i l l ]
0 1
10
11
12
Fill Thickness,m
o.o
-.5 -1.0 -1.5 -2.0 -2.5
^"W^'^^^gi^gmci^
3D analysis(M Strength Profile) J2D analysis(M sirength Profile)
î
0 1
10
11
12
Fill Thickness,m
274
Figure 6.19
Spatial displacement contour of Vernon approach embankment[Gl 12]
Station 26+20
Station 27+35 Station 27+80 Station 28+20
275 Figure 6.20 Plan view and cross section A-A of Waterline test [ G l 13]fill
54.2
16.;
21
, 7,6
32
I A
4
l-
11,40
Poorer
17,1
10
>
17,1
x
Point O
A-A Section
Figure 6.21
Measured and calculated displacement by 2D analysis for the Waterline Test Fill
12 11 J2D analysis(H Strength Profile) 10 |2D analysis(M Strength Profile) 9 8 7 6 5 4 3 2 1 0 6 7 8 9 10 10 11 12 13 * measured data (Crawford et al. 1a92) The fill thickness of the Waterline test embankment H=11.8ni 2D ahalysjs(L Strength Profile)
./
=.-.. = , , = . ^ = 4 * ^ ;
A
^
Fill Thickness,m
10
11
12
13
Fill Thickness,m
Figure 6.22
Measured and calculated displacement by 3D analysis for the Waterlme Test Fill
(a) 12 11 10 9 3D knalysis(M Strength'Profile) 3D jtna|ysis(L Strength Profile)
-^
e
Net Fill Height,
8 7 6 5 4 3 2 1 0 8 9 ; The fill thickness of the Waterlinq test embankment: j

4 __ 4 ..._, 4 ; ~H=Tx.8m.|.i.
(b)
10 11 12 13 Fill Thickness,m
10
11
12 13
Fill Thickness,m
278
Figure 6.23
Illustration of 3D effect on the bearing capacity and the cases studied.
1.5
Shape factor from Equation [6.1]
Waterline fill
1.4 4
.
Q
CM
1.3
1.2 4
Malaysia St. Alban fill Vernon approach embankment
1.1
1.0 0 1
10
11
Base aspect ratio, L/B
279
CHAPTER 7 SUMMARY AND FURTHER WORK

7.1 Summary
In this thesis, a general constitutive framework has been developed to account for viscosity, structure, and strength anisotropy of natural clay. In addition, selected issues affecting the performance of embankments on clay foundations were investigated. This thesis first introduces a simple elastic viscoplastic model and describes an approach to determine the viscosity-related parameters required in this model. Some of the fundamental principles in EVP theory are validated based on the viscous responses of 19 clays reported in the literature by different researchers. By introducing a state-dependent fluidity parameter, an existing EVP model is extended to account for structure and its degradation on the engineering behaviour of natural clay. The extended constitutive model was successfully used to simulate the rupture response and rate sensitivity of Saint-Jean-Vianney clay. Then a tensor approach was coupled with the EVP model to simulate the strength anisotropy of natural clays. This model was shown to be able to simulate the measured orientation-dependent strengths and pore water pressure responses in undrained triaxial tests on two natural clays. The Gloucester test embankment was examined to investigate the influence of structure and destructuration on its field performance. This study shows that the use of a structured EVP model improves the numerical simulation over long-term settlement of the Gloucester test embankment, compared with the use of an unstructured EVP model.
280
In addition, contours of strength change in the Gloucester foundation highlight the influence of destructuration, which reduces the strength gain due to consolidation and even leads to net strength loss in some local zones in the clay deposit. Thus, considering destructuration is important to evaluate the in situ settlement and the stability of infrastructures on or in structured natural clays. Lastly, three full-scale test embankments built on soft clay deposits were studied to investigate the influence of three-dimensional (3-D) geometry on their in situ performance. Both two-dimensional (2-D) plane strain finite element analysis and threedimensional (3-D) finite element analysis are performed for each case. By comparing the calculated collapse fill thickness from 2D and 3D analyses, it is shown that 3D effects are quite significant for all test embankments, which have dramatically different fill thicknesses and underlying clay deposits. Finally, a suggestion is provided to estimate the 3D effect based on the aspect ratio of the fill base length to the base width.
7.2 Suggestions for Future Research

Although the developed constitutive model is able to simulate the main characteristics of the natural clays studied, it is acknowledged that the behaviour of natural clay is complicated and the numerical simulation for some clays is very challenging. The following summarizes several interesting issues deserving further investigation. There are currently few studies on the static yield state and the threshold strainrate require to reach it except those reported by Sheahan et al (1995) and Hinchberger (1996). The possible reason is that the magnitude of the threshold strain rate according to the static yield state is far lower than the strain-rates commonly used in laboratory tests.
281 As an alternative, the static yield state may be evaluated using long-term consolidation tests, where a very low strain-rate would be reached after a long period of creep. This thesis studied the viscosity of clay macroscopically. It would be helpful to explore the microscopic mechanism of the viscosity behaviour for clays. The rateprocess theory, proposed for the atom and molecule level by Glasstone et al. (1941), assumes a balance of input energy and the energy barrier among the equilibrium position for particles. This theory has been introduced into soil mechanics to interpret the viscosity mechanism (Mitchell et al. 1968; Feda 1989). Obviously, as a mixture of water, particles, and possibly air, clay is far more complicated than metal and consequently the application of Gloasstone's rate-process in soil mechanic faces considerable challenges. However, it is worth further study to gain insight into the viscosity mechanism at a microscopic level for clay viscosity. Clay structure has been modeled by adopting a state-dependent fluidity parameter in this thesis. However, as mentioned previously, the rates of structure damage with plastic strain during undrained triaxial compression and oedometer compression tests appear different. This discrepancy may be attributed to the strain localization during undrained triaxial compression. To investigate this issue, a numerical simulation including strain localization in triaxial tests may be helpful to address this discrepancy and improve the understanding in the development of shear bands in specimens during undrained compression tests.
282
References
Feda, J. 1989. Interpretation of creep of soils by rate process theory. Geotechnique, 39(4): 667-677. Glasstone, S., Laidler, K.J., and Eyring, H. 1941. The theory of rate process. McGraw Hill, New York. Hinchberger, S.D. 1996. The behaviour of reinforced and unreinforced embankments on rate senstive clayey foundations. Ph.D Thesis, University of Western Ontario, London. Mitchell, J.K., Campanella, R.G., and Singh, A. 1968. Soil creep as rate process. American Society of Civil Engineers Proceedings, Journal of the Soil Mechanics and Foundations Division, 94(SM1): 231-253. Sheahan, T.C. 1995. Erratum: interpretation of undrained creep tests in terms of effective stresses. Canadian Geotechnical Journal, 32(3): 557. Sheahan, T.C. 1995. Interpretation of undrained creep tests in terms of effective stresses. Canadian Geotechnical Journal, 32(2): 373-379.
283
APPENDIXES
APPENDIX A
ON THE PLASTIC POTENTIAL IN EVP MODEL This appendix introduces the definition of plastic potential, and addresses the necessity of the plastic potential normalization in EVP theory. Lastly, the typical plastic potentials for undrained triaxial compression tests, oedometer compression tests, and isotropic compression tests are summarized. Definition ofplastic potential The plastic potential defines the direction of plastic strain increments. In stress space, the plastic potential governs the relative magnitudes of volumetric and deviatoric incremental plastic strains. For example, in Figure Al, the plastic potential for Point B is vertical, suggesting no volumetric incremental plastic strain at the current stress state; whereas the plastic potential for Points Tl or T2 suggests no deviatoric incremental plastic strain at an isotropic stress state. Point A depicts a stress state with both volumetric and deviatoric components of strain. Necessity of normalization on plastic potential This section illustrates the need of the normalization on plastic potential in EVP theory through a simple example. As shown in Figure Al, an elliptical yield surface is assumed in this appendix with an associated flow rule.
284
' < T : - 2 C Tmy / 3 ^ 2 1 m
+ {fir2)2-(2<7'my/3)2
[Al]
where / is the yield surface function, which can be used to derive the plastic potential if an associated flow rule is assumed. Considering two isotropic stress states (see Points Tl and T2 in Figure Al), their volumetric plastic potentials are shown as following: df _v'm-2<j'my/3 do'm 2 Consequently, a'm 6 [A2]
a/ TI = 1 0 >6 ao-:
JL do'
IU
=15
Equation [A3] shows that the magnitude of plastic potential, , increases with
the magnitude of o'm , which is inconsistent with EVP theory in which [G114]- is a a"m unit vector. As previously mentioned, the plastic potential defines the direction of plastic strain according to the stress state. In this case, the stress states for both Tl and T2 are isotropic, and consequently the plastic potentials for these two points should be same. Thus, there is a need to normalize plastic potential in the EVP theory to address this discrepancy. In this appendix, the plastic potential function, , is normalized viz.:
285
do'. d<rf.
\2
a<x'
3/
\2
[A4]
In Equation [A4], the plastic potential is normalized as a unit vector in the o'm yj2J2 stress space. As a result, the normalized plastic potentials for Points Tl and T2 can be expressed as:
71
Tl
do'
[A5]
a' =60
3<r'
3^2^
3a'
T2
a<xl
a' =90
+ dJlT 2
It can be clearly seen that the normalized plastic potentials for the stress points Tl and T2 are same and consistent with the definition of plastic potential in EVP theory. Typical plastic potentials for standard experiments For oedometer compression tests, the corresponding boundary condition can be
represented by: horizontal =0 , ewrf = e atfa , , and dev= yf2/3(eaxua-ehorizimlal )=4lT$ e ^ .
3/
2 ' <=90
=1
[A6]
Assuming that the elastic strain at yielding is negligible compared to the plastic strain, the axial plastic potential in oedometer test can be expressed viz.:
286
3f
df
Ô'axial oedometer dv'axial
da'
+ 3^2^
3/
de axial V( j v 0 /) 2 +feJ 2
y
[A7]
For undrained triaxial compression tests, the boundary condition includes

^horizontal = "
axial ^
vol
= 0
>
a n d
dev
= ^
, mial
. For isotropic compression, the

an
boundary condition includes: e axial =e horizontal .fl/
''
voi - 3 fflrifl /' vol axial
d * - 0 .
The normalized axial plastic potentials for undrained triaxial tests and isotropic compression tests are derived respectively: 3/
. axial undrained
de axial
AdeJ^deJ
d
j(0)^M ^
[A8]
axial
4faJ+faJ W^f
[A9]
Tables Al and A2 summarize the boundary conditions and the normalized axial plastic potentials for undrained triaxial compression test, oedometer compression test, and isotropic compression test, respectively. These values are utilized in Equations from [2.7] to [2.12] in Chapter 2.
Table Al Summarized boundary conditions in standard experiments
Isotropic compression test
Oedometer compression test
Undrained triaxial compression test
'axial
axial
axial
'axial
'horizontal
' axial
' axial ' ^
^vol ~ V^ horizontal "*" ^ axial >
3e axial
'axial
^dev
V r, V & axial ' ^horizontal
* axial
'axial
288 Table A2 Summary of normalized plastic potentials for standard experiments
Normalized plastic potential
Isotropic compression
Oedometer compression
Undrained triaxial compression

0
V375
V275
l/3(=0.33)
(=0.82)
(=0.77)
289 Figure Al Illustration of plastic potential in stress space
(V^-K/3) 2
Om ,kPa
290
APPENDIX B
THE RELATIVE MAGNITUDE OF e' AND ej IN EVP MODEL
y y
Typically, in an EVP model, the total strain-rate comprises two components, elastic strain-rate and viscoplastic strain-rate. The objective of this appendix is to evaluate the assumption that the elastic strain-rate at or after yielding is negligible compared with viscoplastic strain-rate. A numerical simulation of compression tests was performed to investigate the relative magnitudes of the elastic, et, and viscoplastic, ejf, strain-rates. A schematic diagram (Figure Bl) shows the stress path of a constant rate of strain (CRS) isotropic compression test. The isotropic compression test is chosen because the plastic potential along this stress path can be conveniently assumed to be unity, although this is approximate. The assumed constitutive parameters for the numerical analysis are listed as following: o =50kPa, a =0.033( n = 30 ), f
K7A=0.1,
=1.0 xl0~ 8 /min, X =0.65,
and the applied volumetric strain-rate,eml =1x10"^/min.
Figure B2a compares the relative magnitudes of the elastic and viscoplastic strainrates during the isotropic compression. At beginning of loading, the ratio of elastic strainrate to total strain-rate is close to unity (see the dashed line in Figure B2a), suggesting elastic strain-rate dominates during this period. Then, elastic strain-rate abruptly decreases as the stress approaches the apparent yield stress or isotache corresponding to the applied strain-rate (see Figure B2b for the determination of the apparent yield stress). Figure B2a shows that at or after yielding, the elastic strain-rate is about 4% of the total strain-rate. Accordingly the viscoplastic strain-rate increases up to 96% of the total strain-
291 rate at or after yielding (see the solid line in Figure B2a). A sensitivity study was done to investigate the influence of K IX. This study found that the ratio, Kl X, has some influence on the ratio of the elastic strain-rate to the total strain-rate, eeml to evol, at yielding. As shown in Figure B3, the ratio of evollsvol at yielding increases with the ratio of Kl X. For KIX values ranging from 5% to 30%, the ratio of evolleml at yielding varies from 2% to 12%. Thus, the maximum difference between eeml andevo/ at yielding is 12%. For the typical KlX ratio of 0.1 for most soils (Holtz and Kovacs 1981), the elastic strain-rate is within 5% of the magnitude of viscoplastic strain-rate. Therefore, from an engineering point of view, the strain-rate at or after yielding is approximately equal to the viscoplastic strain-rate. The following further evaluates the assumption of neglecting the elastic strain-rate in the context of the rate-sensitivity analysis. Figure B4 shows the log( a'p ) and log( e ) relations in terms of the strain-rate accounting for elastic strain-rate and the strain-rate neglecting elastic strain-rate, respectively. The measured data from Batiscan clay is also shown in Figure B4. It can be seen that the effect of neglecting ee is minor, even in the case of KlX =0.3. It is then concluded that elastic strain-rate component can be neglected in the rate-sensitivity analysis of clays. Although this conclusion is derived from the isotropic compression tests, this statement is considered applicable for oedometer compression tests and undrained triaxial compression tests.
292 Figure B1 Stress path in CRS isotropic compression test
Dynamic yield surface A with increased "p
Static yielding surface
Stress path in CRS isotropic compression
Figure B2 Comparison of the elastic strain-rate with the viscoplastic strain-rate during CRS isotropic compression^ 115]. (a) Variation of elastic and viscoplastic strain-rates with loading stress
Effective vertical stress, kPa 35 40 45 50 55 60 Apparent yield stress 65
1.4
1.2 -
o
'->
Visocplastic strain-rate / Total strain-rate Elastic strain-rate / Total strain-rate
CO
CD CD i
1.0 .8 .6
4 -|
'co
CO
-I
.2 0.0
(b) Determining the apparent yield stress
0.000 ,it)
-.005 -.010
"eg o
c
CD >
-.015 .020 40 50 60 Effective vertical stress, kPa, in log scale
294 Figure B3 Relationship between eevolleml and KIX in isotropic compression tests
0.04
0.00
KA,
Figure B4 Influence of neglecting elastic strain-rate in the rate-sensitivity analysis.

Neglecting elastic strain-rate (assuming KJ'X = 0.3) Neglecting elastic strain-rate (assuming K/X = 0.1) Total strain rate Measured rate-senstivity for Batiscan Clay
-10-3
1Q-2
-10-1
Strain rate, /min
296
APPENDIX C
DETERMINATION OF THE PARAMETERS, Cr,Cc, Ca Cr and Cc are the recompression index and compression index respectively. Figure CI shows a typical response of clay in an oedometer compression test, in terms of the void ratio versus the effective vertical stress in a semi-log scale. Cr and Cc can be determined by the following equations: Cr=Ae/A\og(a'v) Cc = Ae/Alog(cr' v ) for<7'v<<7'p for <r'v > &p [CI] [C2] is
where e is the void ratio, a\ is the effective vertical compression pressure, and o' the preconsolidation pressure.
The determination of Cr and Cc is graphically shown in Figure CI, where Cr characterizes the pseudo - elastic segment of the compression curve and Cc describes the plastic segment. Ca is the secondary compression index. Figure C2 shows a typical compression curve from a drained constant stress creep test. It can be seen that Ca is measured from the segment of compression curve after the dissipation of excess pore water pressure (see EOP in Figure C2). Raymond and Wahls (1976) and Mesri and Godlewski (1979) defined Ca viz: Ca=Ae/Al0g(O [C3]
where e is the void ratio and t is the time elapse after the beginning of creep tests. In
297 addition, Figure C2 graphically shows the measurement of Ca . Alternatively, C^ = A/Alog(0 is also used to describe the secondary compression. The relationship between Ca and C^ is: Ca=Caex{l + e0) [C4] C^ are all
where e0 is the initial void ratio. It is noted that Cr, Cc, Ca , and dimensionless parameters[Gl 16][G117]
298 Figure CI Measurement[G118] of Cr and Cc
CD
'*-
o"
Cd
TO
-a o
>
Vertical Effective Stress, a' v , in log scale (kPa)
299 Figure C2 Measurement[G119] of Ca
Ae
EOP: End of pore pressure dissipation
log(time)
300 Figure C3 Measurement of Ca from secondary compression tests on London[il20]
clay (data from Lo 1961)
&
1^w\J
sJTO w**0 o^50
** 330 3X> 350 IB 15

41
to 340380 360 1825 t345 385 365 1S30
3 4 5 Time (min)
104
301 Figure C4 Measurement of Ca from secondary compression tests on[G121] Gloucester clay (data from Lo et al. 1976)
"5
o o o
Ooo
o
^
Au=0 ^ 'Q. S ^ S.
LABORATORY DATA O Loetal. 1976 Depth 4.3m Stress Increment 43.2 - 82.7 kPa
CCCE=0.022, e0=1.8
Ca=0.061
12 h
I I I I I I I
| | I 1 | J
I I I I M 1
I t I M I I
I I I I | I
| I i I I I I
10
100
1000
Elapsed Time (min)
10000
100000
100000C
Figure C5 Measurement of Ca from secondary compression tests on Drammen (data from Bjerrum 1967; Berre and Bjerrum 1973)
0.1
TIME IN YEARS 10 100
1000 3000
Figure Co Measurement of Ca from secondary compression tests on Sackville clay (data from Hinchberger 1996)
-1 -
---o....
"'0..
o.
-2
o. o.
Q
^
c
CO
-3r* -^ ' .
O Q
55
1? -4 x ^ < _ LABORATORY DATA O Data from Hinchberger (1996) Stress Increment 50-100kPa
ô.
\ ""-'Q.. ^ ^
Depth 3.8m
Cas r^-.
Cocs=0.0115 e0=1.7 Ca=0.0311
i i i i i i 111
ill
1
1 1 1
Ô
M i l
i i i i 11
10
Elapsed Time (min)
100
1000
304
Figure C7 Measurement of Ca from secondary compression tests on Berthierville[il22] clay (data from Leroueil et al. 1988)
0.00
LABORATORY DATA
Note: The test at the highest increment stress (135kPa) is chosen to obtain Ca, because the influence of clay destructuration on the secondary compression is assumed to be less significant for the clay sample at high increment stress than the clay samples at low increment stress.
- Data from Leroueil et al. 1988 Depth Stress Increment 135kPa -.05 - 2.23-3.48m
.10 h
.15 h
-.20
Cae=0.01 e0=1.7 Ca=0.027
Cas T
135kPa
-.25
_i
i i i 1111
i i i 1111
_l_uj
_1
I I 1 1 1
10
100
1000
10000
100000
Elapsed Time (min)
305 Figure C8 Measurement of Ca from secondary compression tests on St. Alban[il23]
clay (data from Tavenas et al. 1988)
St Alban Clay LABORATORY DATA (Tavenas et al. 1978) Long-term Oedometer creep test Depth Stress Increment 3m 28.0 kPa Ccce=0.015 e0=2.43 Ccc=0.05
j i i i 11 i n
i i i i 1 1 1 1 1
_i
' i i'' i
ûL
10
100
1000
Elapsed Time (min)
10000
100000
100000C
306
APPENDIX D
FACTORS AFFECTING a This appendix investigates several factors that may have an impact on a, such as temperature, plasticity index, sensitivity, liquidity index, and destructuration. The influence of temperature has been investigated by several researchers (e.g. Boudali et al.1994; Graham et al. 2001;Marques et al. 2004). Marques et al. (2004) presented a detailed study on the temperature effect on the behaviour of St-Roch-deF Achigan clay. In Figure Dl, it can be seen that the slope for the log( <r' ) and log(e aiaal ) relationships appears to be independent on the change of temperature from 10C to 30C and 50C. Similar observations were reported by Boudali et al. (2004). Therefore, the parameter a appears not to be sensitive to temperature. The parameter, a , seems independent on the plastic index (PI). Table 2.1 summarizes the soil properties (e.g. water content and plasticity index) for the clays. The values of a are plotted against the plasticity index (PI) for 18 clays in Figure D2. There is no clear evidence for the correlation between a and PI. Thus, it seems that the ratesensitivity, represented by a, is independent on PI. This finding is consistent with the study by Graham et al. (1983). The correlations of a with St (Sensitivity), and LI (Liquidity index) are presented in Figures D3, and D4 respectively. As shown in Figures D3, the correlation of a with St can be approximately represented by a linear line, which shows the trend for most clays presented except St. Alban clay and Batiscan clay. In Figure D4, the correlation between a and LI can be represented by the following equations:
307 Best fit line: a = 0.05xLI [Dl] [D2] [D3]
Upper bound: a = 0.08xL7 Lower bound: a = 0.03xZi
As shown in Figure D4, most clays fall in the range defined by the two bound lines. However, it is noticed that the three Leda clays (St. Alban clay, Batiscan clay, and Ottawa Leda clay) are located outside of the range defined by Equations [D2] and [D3]. These three Leda clays are the Champlain Sea Clay from eastern Canada, which is characterized by the extraordinarily high water content and liquidity index. Therefore, the proposed relationship between a and LI may be not applicable for some Leda clay. Hinchberger and Qu (2007) discussed the influence of destructuration on a. The comparison of a measured at different strains for London clay, Belfast clay and Winnipeg clay respectively shows that the a measured at various strains appears to be consistent. Thus, a is considered independent on the structure damage during
loading[il24]. (more details is referred to Appendix E).
308
Figure Dl
Influence of temperature on the rate-sensitivity parameter, a forStRoch-de-F Achigan clay (modified from Marques et al. 2004)
200
Temperature 10 c
13 o ieo.
.5
140<
309 Figure D2 Variation of viscosity exponent, a, with Plasticity index[il25] ( for clays listed in Tables 2.1 and 2.2)
.10
.08
8
.06 4
>
.04
O O
0
o
0 O O
o o o
.02 4
0.00
10
r 20
30
40
50
60
70
Plasticity lndex.%
310 Figure D3 The correlation between[i 126] a and St (Sensitivity)
.12
a =0.025 + 0.0016*St
.10
:
/
.08 4
/
/
.06
a
.04
Batiscan clay
...v/.n...
/
""SrAlbah clay
.02
0.00
20
40
60
80
100
120
140
St
31 D4 The correlation between a and[il27] LI (Liquidity index)
.12 Upper bound .10 4 a=g.08*LI

/ Ottawa Leda clay
Best fit line oc=0.05*LI Leda ciay
.08
/
;/ .
Lower bound a=0.03*LI
.06 4
_. Batiscan clay " 0 ~ St. Alban clay
.04
.02
0.00
LI
312
APPENDIX[G128] E INFLUENCE OF STRUCTURE ON THE TIME-DEPENDENT BEHAVIOUR OF A STIFF SEDIMENTARY CLAY Sorensen et al. (2007) have decided to study the influence of microstructure on the time-dependent response of undisturbed and reconstituted London clay using drained and undrained triaxial compression tests (CIU and CID) with step changes in the applied strain-rate. The paper presents interesting behaviour and Sorensen et al. (2007) should be commended for demonstrating the viscous response of London clay. The primary influence of microstructure on the engineering response of London clay can be seen in Figure El a, which compares the stress-strain response in the undisturbed and reconstituted states. Figure Elb shows similar behaviour from triaxial compression tests on Rosemere clay from Eastern Canada (Philibert 1976). From Figure El, it can be seen that there are similarities in the relative stress-strain response of both materials in spite of their vastly different index properties (e.g. IL = 0 versus I I ~ 1.2). The stress-strain response of both clays during triaxial compression is characterized by: (i) reaching a peak shear strength followed by post-peak strength reduction with largestrain, (ii) predominantly strain hardening response of the reconstituted or disturbed materials, and (iii) at large-strain, the post-peak strength of the undisturbed clay approaches that of the reconstituted and 'cut' materials, respectively. The difference in behaviour (the shaded areas in Figures El a and Elb) is typically attributed to the effects of microstructure or weak bonding between the clay particles and aggregates of clay particles. Such behaviour is analogous to that typically observed in oedometer
consolation tests on undisturbed and reconstituted materials (Burland 1990).
A version of this appendix has been accepted in Geotechnique 2007
313 Regarding the time-dependency or rate-sensitivity of London clay, Sorensen et al. (2007) quantify viscous effects using the jump in deviatoric stress induced immediately after changing the axial strain-rate. Although such an approach has merit, the following presents an alternative interpretation of the rate-sensitive response of London clay using the theory of overstress viscoplasticity (Perzyna 1963). The current authors hope that this alternative interpretation will provide additional insight into the viscous response of undisturbed and reconstituted London clay. Theoretical Background Perzyna (1963) originally proposed the theory of overstress viscoplastic for the yielding of steel at high temperature. This theory has been subsequently adapted to geologic materials by researchers such as Adachi and Oka (1982), Katona and Mulert (1984), Desai and Zhang (1987) and Hinchberger and Rowe (1998) to name a few. For an elastic-viscoplastic material, the strain-rate tensor can be decomposed into elastic and viscoplastic components as follows: e = 85+63" [El]
At yield or failure, the viscoplastic strain-rate typically dominates (Chapter 2). A form of the viscoplastic strain-rate tensor is (e.g. Katona and Mulert 1984 and Desai and Zhang 1987):
^=^(f))y/îjhj{^/^y-^k^^j]
[E2]
where ^ is a viscosity constant with units of inverse time (typically s _I ), f is the yield function from classical plasticity theory, (f) is called the flow function and it is derived from f , and [3f /da^ J is the plastic potential, which is derived as a vector of unit length.
314 The Macauley brackets ( ) in Equation [E2] imply (j)(f) = 0 for f < 0 and
HfHq/qoY-liorf>0.
The flow function, (|>(f), in Equation [E2] is a power law (Norton 1929) where q0 represents the long-term strength (reached at very low strain-rates), q is the strain-rate dependent deviator stress at yield and the term q/q 0 is the overstress (e.g. q/q 0 =1.1 implies 10% overstress). An upper bound estimate of q0 , q 0 = 1 2 5 k P a , can be
obtained for London clay from the deviator stress reached after 4 days of stress relaxation (see Figure 3 in Sorensen et al. 2007). Considering axial strain-rate only, the viscoplastic strain-rate at yield is approximately:
e;W((4/<7j"-l)(V273)
[E3]
where V2/3 is an estimate of the plastic potential, 3f / 3 o u , derived assuming constant volume deformation. Although London clay exhibits dilatant behaviour during the
triaxial tests (see the pore pressure response in Figure 8, Sorensen et al. 2007), the plastic potential has a negligible impact on the following discussion and derivation. Taking the logarithm of Equation [E3] and rearranging, it can be shown that (Qu and Hinchberger 2007): log(q) = alog(e axial )+A s [E4]
for q/q 0 > 1.1. In Equation [E4], As =log(q0u.a) and a = \ln . Leroueil and Marques (1996) and Soga and Mitchell (1996) have used a similar relationship to evaluate the ratesensitivity of various clays. Thus, elastic-viscoplastic constitutive models based on a power law flow function
315 (e.g. Adachi and Oka 1982, Katona and Mulert 1984, Hinchberger 1996, Hinchberger and Rowe 1998, and Desai and Zhang 1987) imply a linear relationship between log(q) and log(e) for stress states at yield or failure. In such a theory, the rate-sensitivity
(variation of q versus ) at yield or failure is governed by a , which is the inverse of the power law exponent, n . The following is a re-evaluation of the strain-rate effects measured by Sorensten et al. (2007) for London clay using the above theoretical framework. Interpretation of Rate-Effects Figure El a shows the deviator stress, q, versus axial strain response reported by Sorensen et al. (2007). The data is re-plotted in Figure E2 using a semi-log scale. From Figure E2, it can be seen that there is relatively uniform variation of log(q) versus axial strain, notwithstanding that rate-effects appear to be less pronounced for the reconstituted material at axial strains in excess of about 5%. Extracting deviator stress versus axial strain-rate from Figures El (a) and E2, a series of essentially parallel linear lines can be obtained in log(q) - log space. Figure E3 summarizes the log(q) versus log(e) data extracted from undrained triaxial compression tests on undisturbed London clay at axial strains of 1, 1.5, 2, 2.5, 3, 4, and 4.5%. Figure E4 shows similar data for the reconstituted material at axial strains of 1, 2, 3, 4, and 5%. The slope, a , of the lines in Figures E3 and E4 represents the ratesensitivity of London clay. When compared in Figure E5, the data suggests that the mean value of a is about 0.023 (n=44) and that both the undisturbed and reconstituted materials have essentially the same rate-sensitivity. Furthermore, the rate-sensitivity
316 parameter, a , estimated from drained triaxial tests on intact material (see Figure 9 in Sorensen et al. 2007) is also plotted in Figure E5. It can be seen that the rate sensitivity parameter estimated from CID triaxial tests is the same as that deduced from the CIU tests. Thus, the rate-sensitivity is identical for both drained and undrained triaxial
compression and for the intact and remolded materials. For comparative purposes, Figure E6 shows the results of step tests on Belfast and Winnipeg clay (Graham et al. 1983). The strain-rate parameter, a , is plotted in Figure E7 for both clays. From Figure E7, it can be seen that a varies from 0.035 to 0.041 (24< n < 29) for Belfast clay, and from 0.033 to 0.036 (28 < n < 30) for Winnipeg clay. Both clays are more rate-sensitive than London clay. In addition, Belfast, and Winnipeg clay do not show reduced rate-sensitivity with continued straining (or destructuration) after reaching the peak strength; even for axial strains in excess of 15%. In contrast, the ratesensitivity of London clay diminishes with large axial strains in excess of about 5%; however, additional testing is required to confirm this behaviour. Summary From the above discussion and interpretation, it can be concluded that the ratesensitivity of undisturbed London clay is the same as that of the reconstituted material. Thus, the structure of London clay appears to have a negligible impact on its rate sensitivity, whereas, the primary influence of structure appears to be exhibited by the shaded areas in Figures la and 2. The above interpretation, has utilized a power law in conjunction with Perzyna's theory of overstress viscoplastic (Perzyna 1963) and clearly other interpretations are possible. However, Sorensen et al. (2007) hope that this
discussion provides an alternative perspective to that of Sorensen et al. (2007) for
317 consideration.
318 Figure El Stress-strain behaviour during triaxial compression tests on London clay (Reconstituted and Undisturbed) and Rosemere clay (Undisturbed and 'Precut'). (a) London Clay (Sorensen et al. 2007)
600
Axial Strain, %
(b) Rosemere Clay (Philibert 1976)
Axial Strain (%)
319
Figure E2
Stress-strain response of London clay in semi-log scale (Re-plotting Figure la in a semi-log scale)
Figure E3
The rate sensitivity parameter, a, measured from undrained triaxial compression tests on undisturbed London clay
Undrînkttriaxia! triaxial corripression test)onjurjdisthibed Lohdon clay:
(k=0.023 at Peak oc=0.021 at e=2.5% prerp#akj-4 apO.Q21 at e=2% pre-pedk! apO.026 at e=^1.5% ;pre-peaka^0.b2!4ate=1% pre-peak a=0.021 af!e=4.0% post-peaks a=0.017ate=4.5% post-peak!
10-7
10e
10-5
10"4
10-3
10- 2
Strain rate, /min, in log scale
Figure E4
The rate sensitivity parameter, a , from undrained compression tests on OC reconstituted London clay
triaxial
UndraineditrWxjial compressiqn!te?t qn OC! reconstituted Londonjclay
a=0.016ate=?5% a=0.022 at e*4%; 0=0.022 at p=3%
I
"\
\ ^
i ! i
ix=0.023 at e=2%
!|a=0.023:ate=1 < ^ i
10- 7
10-6
10-5
10"4
10 3
10 2
Strain rate, /min, in log scale
Figure E5
Summary of a obtained from undrained triaxial compression tests on reconstituted London clay, and drained and undrained tests on undisturbed London clay
.10 -Jt CD test on undisturbed sample - CU test on undisturbed sample -O CU test on reconsituted sample .08
O
o
.06
c
03
.04
.02
0.00 0.00
Axial Strain
323
Figure E6
Stress-strain relations in CAU tests on Belfast clay and Winnipeg clay
<J1C: Confining pressure.kPa Axial strain rate = 5%/h 0,.5/<^
Winnipeg clay (Graham, et al. 1983) .1 H 0.0 0.00
.05
.10
.15
.20
.25
.30
Axial Strain, %
324
Figure E7
Parameter a measured for Belfast clay and Winnipeg clay
Belfast clay (drahaml Hlj al. 1983) oc=0.035 M Reak Va=0.040 M * f W/o post-peak o oc=6.041 i H e(=N5% post-peak
2i
Winnipeg clay (JGrahanfi,! M bil. 1983) a=0.033 at Peak ] I a=0.036 at e=10% post-piak] a=0.033 at e=15% post-pieak
10" 7
10
10 E
10J
10- 3
10- 2
Strain rate, /min
325
APPENDIX F
ON THE DECREASE OF STRAIN-RATE IN THE O/C CREEP TESTS During the undrained creep test on Saint-Jean-Vianney clay at dry side in stress space, the axial strain-rate was found to decrease with time prior to creep rupture. Considering the incremental strain from the completion of loading to the creep rupture was less than 0.2% for each creep test, the decrease of strain-rate is negligible from an engineering point of view. However, theoretically, the overstress[G129] concept alone can not explain this phenomenon. It is noted that the influence of this decrease of strain rate in. creep is minor considering-4he~4elal4ftefe^ ereep-was-ftet-e*eeed 0.2%, which is out of
engineering irrtefestrAlse-the in-situ creeps are often' in drained'Conditions, wMeh-fetiew
has been successfully used to simulate "the s#-ain-rate -decrease- during -the drained creep test and the undrained creep test with stress state in the "wet: side (e.g. Kutter et al. 1.992 and-Hinehberger 4 996). To investigate the possible reasons for this phenomenon, this appendix reevaluates this creep tests on SJV clay using modified approaches with various assumptions to simulate the decrease of strain-rate. The hypotheses adopted in the modified approaches are described below, together with the comparison of the calculated and measured response of SJV clay. In the first approach, it is assumed that the Drucker-Prager envelop would be hardened due to plastic work, as suggested by Lade and Duncan (1973). The slope of the
326 Drucker-Prager envelop in the -J2J2 -<Jm stress space can be represented using the effective friction angle viz: M
CS
=2
->
S w
*'
, i
[Fl]
J
3-srn0
where Mcs is the slope of the Drucker-Prager envelop in Figure Fl, and 0' is the effective friction angle. The hardening law can be expressed using an exponential equation: Mcs=M1-(Mz-Ml)xe'CWp wp=jcr'yde? [F2] [F3]
where Mt and M2 represent the initial and final slopes, respectively, c is the hardening parameter, and wp is the plastic work. The magnitude of the final slope, M 2 =1.34, was obtained according to <p'= 40 reported by Vaid et al.(1979). The other two parameters, M[ =1 and c =20 were obtained using a trial and error approach. As shown in Figure Fl, the increase of the slope of the Drucker-Prager envelope due to hardening leads to a contraction of dynamic yield surface and consequently a decrease in the overstress. As a result, the calculated strain-rate during the creep tests would reduce with time. Figure F2 shows the comparison of the measured and calculated strain-rate versus time during the undrained creep tests. It appears that the decrease of strain-rate can be simulated by accounting for the hardening of the Drucker-Prager envelope. In the second approach, it is assumed that the stress path in the central part of the triaxial specimen was permitted to follow the elastic stress path during initial loading, not
327
the triaxial limit (see the dash line in Figure F3). In this approach[G130], more overstress develops relative to the static yield surface in the central part of the specimen: a consequence of the assumed stress state. Compared with analyses where the triaxial limit was enforced throughout the specimen (see the solid stress-path line in Figure F3), the higher level of overstress in the modified analysis causes significantly higher strain-rates and more dilatancy early on in the simulation. Thus the overstress and consequent creep rates reduce with time as the stress state moves right toward the static yield surface, producing calculated creep rates similar to measured creep rates, as shown in Figure F4. In summary, both of the two approaches used in this appendix are capable of simulating the decrease of strain-rate and subsequent creep rupture during the undrained triaxial creep tests on SJV clay. Another alternative is to assume rotational hardening of the state boundary surface, which would give similar results with those two approaches. In addition, the decrease of strain-rate can also be attributed to external factors, for example, the sample bulging under constant loads and consequent stress decrease on the specimen top. However, given the lack of experimental evidence to support these
hypotheses, a definitive conclusion can not be drawn as to the reason for the decrease of strain-rate during the undrained creep tests at the dry side in stress space for SJV clay. Further experiments on the overconsolidated natural clay are desired to testify these hypotheses or investigate the external factors.
328
Fl Illustration of the hardening[il31] of the Drucker-Prager envelope.
Inereaseof M due lo .Hardening effect
Dynamic yield surfaces corresponding MI and

M2
, N
\ \
x
\J*\
Contraction of the dviuunlc vield surfaces during creep tests
\ 'X \ \ \ \
329 Figure F2 The measured and calc[G132]ulated strain-rate variation during the creep tests accounting for the hardening of the Drucker-Prager envelope
430
jCalculated
Measured
10-H
10- 5
10
100
1000
10000
Time (min)
330
Figure F3 Comparison of stress paths in CIU undrained creep on Saint-Jean-Vianney clay
Critical State Line
y'U)
my
331 Figure F4 The measured and calculated stress-rate versus time using the second approach
10
rr
1 1 1
-o
ad=430
10-'
a d =470-
Axial Creep Strain Rate, %/MIN
7 I / I / I / 1 / 1 r
X^F
^^ \v^ I
104'
i\
; ^*
1000 10000
5
10*
, A (Measured i Calculated |
Time (min)
332
APPENDIX G
A NON-ASSOCIATED VISCOPLASTIC APPROACH The main body of the research in Chapter 5 has focused on the use of associated viscoplasticity to describe the engineering behaviour of 'structured' anisotropic timedependent clay. This appendix describes an alternative approach based on a non-
associated flow rule in the over consolidated stress range (i.e. the dry side) and an associated flow rule in the normally consolidated stress range (see Figure Gl). Based on the results presented below, it can be seen that the engineering behaviour of Gloucester clay can be described using either the approach presented in the main body or using the approach summarized in Figure Gl. Figures G2 to G4 compare the calculated and measured behaviours of Gloucester clay during undrained triaxial compression tests. Figure G2 shows the measured and calculated peak and post-peak strengths for specimens with the orientations of i=0,30 o ,45,60 o ,and90. The corresponding curves of deviator stress versus axial
strain and excess pore pressure versus axial strain are presented in Figure G3. Deviator stress versus axial strain curves for i = 0 and i = 90 are compared in Figure G4. As shown in Figure G2, a non-associated approach is also able to reproduce the measured peak and post-peak strengths of Gloucester clay. The calculated deviator
stresses increase up to the peak strength after which there is a reduction of strength with axial strain after mobilization of the peak strength (see Figures G3 and G4). The
calculated behavior agrees well with the measured behaviour. From Figures G3 and G4, the general trends of the calculated and measured pore water pressure with strain are
333
comparable, although the excess pore pressures are underestimated by the non-associated approach. Overall, the non-associated approach can reproduce the major characteristics of the stress- strain behavior and strength for Gloucester clay in undrained triaxial tests.
Figure Gl Conceptual behaviour of the non-associated soil model
Unassociated Plastic Potential Law

A/2^
Destructured Critical State Line
I I
M,j=0.0 Af.=0.03 M , = -0.03

Associated Plastic Potential Law
M=0.9
Typical Stress Path - >
Static yield surface
335 Figure G2 The effect of sample orientation, i, on the measured and calculated peak and post-peak undrained strength of Gloucester clay, (using a nonassociated approach)
30
Using a non-associated approach Peak
Measured strength (Law 1975) Calculated strength (This paper)
10
Measured post-peak strength at 8% strain Calculated post-peak strength at 8% strain Calculated post-peak strength at 20% strain
5 -4
-20
-10
10
20
30
40
50
60
70
80
90
336 Figure G3 The effect of sample orientation, i, on the measured and calculated (a) axial stress versus strain and (b) excess pore pressure versus strain for Gloucester clay, (using a non-associated approach)
(a)
CO Q.
60
- i=0 Calculated - i=30 Calculated - i=45 Calculated i=90 Calculated
e o-
50
40
30
i-^
20
10
8 10 12 Vertical Strain (%)
(b)
30
0.
CO
20
o 10
Q_ to
<n
g
111
i>
ol
337
Figure G4
The comparison for sample orientations, /, of 0 and 90 on the measured and calculated axial stress versus strain and excess pore pressure versus strain
a. .*_
CO
(0 Q.
l
l
10
12
Vertical Strain (%)
338
CURRICULUM VITAE
Name : PLACE OF BIRTH: POST-SECONDARY EDUCATION AND DEGREES: Ph.D Master of Science Bachelor University of Western Ontario University of Tianjin University of Tianjin 2003-2008 2000-.2003 1996-.2000 Guangfeng Qu Hehei, China
HONORS & SCHOLARSHIPS 2006 2005 2003-2006 2003-2007 2003 Novak Award John Booker Award IGSS (International graduate student scholarship) Graduate Special Scholarship Outstanding Graduation Thesis of Master of Science
1999 Tianjin University Academically Outstanding Student Honor with privilege of being directly admitted into the graduate school without Mandatory Admission Examinations 1998 and 2000 Tianjin University People's Scholarship (The First Class)
RELATED WORK EXPERIENCE: 2000 2003-2007 Engineer in Jun Hua Foundation Engineering Technology Group Teaching and Research Assistant, University of Western Ontario
Publications
Qu, G. and Hinchberger S.D. (2007) Evaluation of the viscous behaviour of natural clay using a generalized viscoplastic theory. Geotechnique, In review Hinchberger, S.D. and Qu, G. (2007) Discussion: the Influence of structure on the time-dependent behaviour of a stiff sedimentary clay. Geotechnique. In press Qu, G. Hinchberger, S.D., and Lo, K.Y. (2007) Case studies of three dimensional effects on the behaviour of test embankments. Canadian Geotechnical Journal. In review. Hinchberger, S.D. and Qu, G.(2006) A viscoplastic constitutive approach for structured rate-sensitive natural clays. Canadian Geotechnical Journal, ReSubmitted November 2007 Hinchberger, S.D., Qu, G. and Lo, K.Y.(2007) A simplified constitutive approach for anisotropic rate-sensitive natural clay. International Journal of Numerical and Analytical Methods in Geotechnical Engineering. In review Qu, G. and Hinchberger, S.D. (2007) Clay microstructure and its effect on the performance of the Gloucester test embankment. Geotechnical Research Centre Report No. GEOT2007-15, the University of Western Ontario, London, Ontario.

Guangfeng Qu - PHD Thesis

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Guangfeng Qu - PHD Thesis

Hochgeladen von

Copyright:

Verfügbare Formate

SELECTED ISSUES ON THE PERFORMANCE OF EMBANKMENTS ON CLAY FOUNDATIONS (Spine title: Selected issues on the performance of embankments) (Thesis

Graduate Program in Engineering Science Department of Civil and Enviromental Engineering

Library and Archives Canada Published Heritage Branch

Bibliotheque et Archives Canada Direction du Patrimoine de I'edition

THE UNIVERSITY OF WESTERN ONTARIO FACULTY OF GRADUATE STUDIES CERTIFICATE OF EXAMINATION

Dr. Sean Hinchberger

Dr. Tim Newson

Co- Supervisor Dr. Ernest Yanful

Dr. John Dryden

Date March, 17,2008

Dr. Jianddong Ren Chair of Examining Board:

234 234 235 236

Figure 2.13 Figure 2.14 Figure 2.15

Figure 3.1 Figure 3.2

109 110 Ill

Figure 3.3 Figure 3.4 Figure 3.5

115 118 119

Figure 3.9 Figure 3.10

Figure 3.11 Figure 3.12 Figure 3.13

Figure 3.14 Figure 3.15 Figure 3.16 Figure 3.17

126 127 128

129 130 161 163 165 167 168 169 170

Figure 4.10 Figure 4.11 Figure 4.12

Figure 4.13 Figure 4.14 Figure 4.15

176 177 179

182 183 184 185

Figure 5.2 Figure 5.3

Figure 5.5 Figure 5.6

225 226 227

Figure 5.9 Figure 5.10 Figure 5.11

Figure 5.12 Figure 5.13

Figure 5.14 Figure 5.15

Figure 5.16 Figure 6.1

256 257 258

259 260 261 262

265 266 267 268

Figure 6.15 Figure 6.16 Figure 6.17

Vertical displacement of Vernon Approach Embankment in 3D analysis 273 274 275

276 277 278

Ca e n a yvp yf, yjp 4>(F)

plastic state corresponding to high compressibility and irrecoverable strains.

1.3 Thesis Objectives and Outline

Gloucester foundation is examined numerically with time after construction.

1.4 Original Contributions

Typical Strength Profile Su (kPa)

Soft Clay Layer

14 Figure 1.2 Schematic of an oedometer apparatus and a typical compression curve.

Schematic of an oedometer apparatus (after Holtz and Kovacs 1981)

15 Figure 1.3 Definition of the orientation angle, i

t Vertical Direction (Direction of deposition/gravity) A j = 0

18 (se) and viscoplastic (svp) components as follows: s =ee+svp [2.1]

O'axial- Vy axial ~<Ty

<W = H< } /< S ) )"

+ log(r v; ') + log(l/3)

As shown in Equation [2.14a], an exponential-law flow function implies

vp Sriog(*w) + l-a\og(y )-a\of>Q-)

,'(d) a;*' =acr'p(s)log(eaxial) + l-alog(rvp)-a\ogQ-)

Combining equations [2.18a] and [2.18b] gives: - = Cj[\n(10)-(A-K)] [2.18c]

Simplifying Equation [2.18c] yields:

respectively). Figure 2.13 compares auc, a

2.4 Selection of Parameters

35 related parameters, a, o*^, and yvp, are discussed below.

Time required for"!? strpfn