modelling of desiccation crack depths in clay soils€¦ · speedy pathways for water ingress,...
TRANSCRIPT
Modelling of Desiccation Crack Depths in Clay Soils
A thesis submitted in fulfilment of the requirement for the
Degree of Doctor of Philosophy
By
R M SASIKA D WIJESOORIYA
BSc. (Honours)
Department of Civil Engineering
Monash University
Australia
December 2012
To
Kularatne and Leela, my beloved parents
Dileepa, my loving husband
i
ii
COPYRIGHT NOTICES
1. Under the Copyright Act 1968, this thesis must be used only under the normal
conditions of scholarly fair dealing. In particular no results or conclusions should be extracted from it, nor should it be copied or closely paraphrased in whole or in part without the written consent of the author. Proper written acknowledgement should be made for any assistance obtained from this thesis.
2. I certify that I have made all reasonable efforts to secure copyright permissions
for third-party content included in this thesis and have not knowingly added copyright content to my work without the owner's permission.
iii
DECLARATION
I hereby declare that this thesis contains no material which has been accepted for the
award of any other degree or diploma at any university or equivalent institution and
that, to the best of my knowledge and belief, this thesis contains no material previously
published or written by another person, except where due reference is made in the text
of the thesis. Where sections of this thesis include the results of joint research or
scholarly publication clear acknowledgement of the relative contributions of the
respective authors is made.
Sasika Dilrukshi Wijesooriya
Department of Civil Engineering
Monash University
Clayton, Australia
iv
v
ABSTRACT
Desiccation cracking is a major problem in many fields. In addition to introducing
speedy pathways for water ingress, cracks can also compromise the structural integrity
of the geo structures. In this regard, prediction of the depth of cracking is an important
aspect in evaluating system performance. Modelling of desiccation cracks is a major
concern for the past few decades. Despite the numerous attempts to model the crack
depths, no comprehensive modelling method is available. In this study, an attempt has
been made to model the desiccation crack depths using analytical and numerical
approaches.
The thesis presents a review of the literature identifying the gaps of the knowledge in
this field, numerical modelling of desiccation crack depths under various conditions for
different soils and laboratory experimentation to support the numerical model. The
existing theoretical methods used to analyse desiccation crack depths and new methods
have been developed to describe the predictions from the numerical program. Further
the cyclic change of climate conditions are considered for developing the numerical
model. Finally a new approach is used to predict the crack depths in which moisture
content change was used instead of suction. Finally a more rigorous approach of
predicting crack depths incorporating cohesive properties at the crack is presented using
the computer program UDEC.
The results from the numerical approach are presented and discussed in the thesis. The
crack depths are compared with either theoretical results or a bench-mark model,
highlighting pros and cons of current approaches. More accurate crack depths agreeing
with the published field observation data can be predicted when using the moisture
content change instead of suction as a model parameter. Furthermore, cohesive
properties of the crack should be considered for fracture modelling provided that soils
are not subjected to extreme dry conditions.
However, it is recommended to conduct comprehensive field experimentation to
measure the desiccation crack depths and compare the results with the numerical
vi
modelling predictions of the same soil in same field conditions to draw fully validated
conclusions.
vii
ACKNOWLEDGEMENT
The author would like to express her sincere appreciation to Professor Jayantha
Kodikara, under whose supervision this work was done. His patience, guidance,
suggestions and freely available time were invaluable. Financial support for this study,
which was provided in form of a scholarship by the Monash University, is gratefully
acknowledged.
Special gratitude is extended to Dr. Nathan Rajeev for his help as associate supervisor,
especially in experimental work. Thanks are also extended to Dr. Aruna Amarasiri for
the help given in numerical modelling work. I greatly appreciate the assistance from
civil engineering academic staff including Prof. Abdelmalek Bouazza, A/Prof. Ranjith
Gamage, Dr. Asadul Haque, Dr. Ha Bui, Dr. Dilan Robert and Dr. Susanga Costa at
Monash University.
I would like to thank all the administrative and technical staff members at the
Department of Civil Engineering, Monash University especially to Chris Powell, Jenny
Manson, Long Goh, Mike Leach, and Alan Taylor for their precious support in my
experiments and other administrative aspects.
My deepest gratitude is given to my husband, Dileepa not only for his support,
understanding and encouragement. I offer my sincere gratitude and deep respect to my
loving parents and sisters for their caring advises encouragement and love.
I would greatly appreciate the valuable discussions, assistance and above all the
friendship from my room-mates Ben Shannon and Senthil Kumar. I would also like to
extend my sincere thanks to my fellow postgraduates of the Department of Civil
Engineering for their friendship and the wonderful time I shared with them at Monash
University.
Sasika Wijesooriya
June 2012
viii
ix
TABLE OF CONTENT
DECLARATION ............................................................................................................. iv
ABSTRACT ..................................................................................................................... vi
ACKNOWLEDGEMENT ............................................................................................. viii
TABLE OF CONTENT .................................................................................................... x
LIST OF PUBLICATIONS ........................................................................................... xvi
NOMENCLATURE ..................................................................................................... xviii
ABBREVIATION ........................................................................................................... xx
LIST OF TABLES ........................................................................................................ xxii
LIST OF FIGURES....................................................................................................... xxv
Chapter 1 ........................................................................................................................... 1
INTRODUCTION............................................................................................................. 1
1.1 General ............................................................................................................... 1
1.2 The Scope of this Study ...................................................................................... 3
1.3 The Structure of the Thesis................................................................................. 4
Chapter 2 ........................................................................................................................... 8
LITERATURE REVIEW.................................................................................................. 8
2.1 Introduction ........................................................................................................ 8
2.2 Clay Liners of Landfills...................................................................................... 9
2.2.1 Evaluation of clay liner designs ................................................................ 11
2.2.2 Types of clay liners ................................................................................... 13
2.2.3 Major issues related to clay liners ............................................................. 19
2.2.4 Field Performances of clay liners .............................................................. 22
x
2.3 Desiccation Cracking........................................................................................ 43
2.3.1 Initiation and evaluation ............................................................................ 48
2.3.2 Factors affecting desiccation cracking ...................................................... 51
2.3.3 Effects of cracking .................................................................................... 52
2.3.4 Theoretical developments for desiccation cracking .................................. 53
2.4 Numerical Models ............................................................................................ 71
2.4.1 Water balance modelling software ............................................................ 72
2.4.2 Fracture modelling software ..................................................................... 77
2.5 Concluding Remarks ........................................................................................ 80
Chapter 3 ......................................................................................................................... 83
COMPARISON OF CRACK MODELLING APPROACHES ...................................... 83
3.1 Introduction ...................................................................................................... 83
3.2 Existing Analytical Approaches for Predicting Crack Depth........................... 84
3.2.1 Constant suction profile ............................................................................ 84
3.2.2 Linearly decreasing suction profile ........................................................... 88
3.2.3 Parabolic suction variation ........................................................................ 89
3.3 Numerical Modelling Approach for Crack Depth Prediction .......................... 92
3.3.1 Overview of UDEC program .................................................................... 93
3.3.2 Numerical model implementation ............................................................. 94
3.4 Comparison of Numerical and Theoretical Results........................................ 102
3.4.1 Constant suction profile (suction profile 1) ............................................ 104
3.4.2 Linearly decreasing suction profile (suction profile 2) ........................... 109
3.4.3 Parabolic suction variation (suction profile 3) ........................................ 114
3.5 Comparison of Results from Different Suction Profiles ................................ 118
3.6 Comparison of Different Theoretical Approaches ......................................... 121
3.7 Conclusion ...................................................................................................... 128
xi
Chapter 4 ....................................................................................................................... 130
EXPERIMENTAL INVESTIGATION OF SHRINKAGE AND SWELLING BEHAVIOUR ............................................................................................................... 130
4.1 Introduction .................................................................................................... 130
4.2 Hydric constant (𝛼) in Stress Analysis ........................................................... 131
4.3 Experimental Procedure ................................................................................. 136
4.3.1 Materials .................................................................................................. 136
4.3.2 Sample preparation and set-up ................................................................ 144
4.4 Results ............................................................................................................ 156
4.4.1 Swell-Shrink cycles ................................................................................. 156
4.4.2 Swelling or shrinking paths ..................................................................... 157
4.4.3 Variation of hydric coefficient (𝛼 ∗) ....................................................... 160
4.5 Other Research on Wet-Dry Cycles ............................................................... 162
4.5.1 Experimental data analysis of Sharma (1998) ........................................ 162
4.5.2 Experimental data analysis of Romero (1999) ........................................ 170
4.5.3 Experimental data analysis of Tripathy (2000) ....................................... 176
4.5.4 Experimental data analysis of Montanez (2002) ..................................... 182
4.5.5 Experimental data analysis of Monroy (2006) ........................................ 186
4.6 Summary and Discussion ............................................................................... 194
4.7 Conclusion ...................................................................................................... 199
Chapter 5 ....................................................................................................................... 201
MODELLING OF STABLE DESICCATION CRACK DEPTHS DURING CYCLIC WETTING AND DRYING .......................................................................................... 201
5.1 Introduction .................................................................................................... 201
5.2 Suction Profiles .............................................................................................. 202
5.3 The Suction in Different Climate Conditions ................................................. 206
xii
5.4 Development of Numerical Model for Compacted Clay Layers under Cyclic Atmospheric Conditions ............................................................................................ 208
5.4.1 Stress change by the direct addition of suction change........................... 209
5.4.2 Fraction of suction on change of the stress ............................................. 210
5.4.3 Stress change on the basis of moisture content change........................... 212
5.5 Soils Used for the Analysis ............................................................................ 213
5.5.1 Regina clay .............................................................................................. 213
5.5.2 Horsham clay .......................................................................................... 216
5.5.3 Altona clay .............................................................................................. 218
5.6 Results ............................................................................................................ 221
5.6.1 Crack depth prediction under different climatic condition ..................... 221
5.6.2 Crack opening and closing with time ...................................................... 223
5.6.3 Effect of placement conditions of the clay liner on initial desiccation ... 226
5.6.4 Effect of the Poisson’s ratio .................................................................... 229
5.6.5 Effect of equilibrium suction .................................................................. 231
5.7 Conclusions .................................................................................................... 233
Chapter 6 ....................................................................................................................... 235
INHERENT PROPERTIES OF UDEC ........................................................................ 235
6.1 Introduction .................................................................................................... 235
6.2 The Numerical Program UDEC ..................................................................... 236
6.2.1 UDEC operation ...................................................................................... 236
6.2.2 Theoretical background of UDEC .......................................................... 237
6.3 Problem Analysing using UDEC.................................................................... 241
6.4 Behaviour of the UDEC Model ...................................................................... 244
6.4.1 Effect of mesh size (l) ............................................................................. 244
6.4.2 Effect of number of increments............................................................... 248
xiii
6.4.3 Effect of crack length .............................................................................. 252
6.4.4 Effect of damping value .......................................................................... 254
6.4.5 Effect of block size.................................................................................. 256
6.4.6 Scaling up the model geometry ............................................................... 257
6.5 Summary and Conclusion............................................................................... 264
Chapter 7 ....................................................................................................................... 265
MODELLING OF DESICCATION CRACK DEPTHS INCORPORATING SOIL FRACTURE ENERGY ................................................................................................. 265
7.1 Introduction .................................................................................................... 265
7.2 Basics of Linear Elastic Fracture Mechanics (LEFM) ................................... 266
7.2.1 Griffith’s criterion ................................................................................... 267
7.2.2 Irwin's modification ................................................................................ 269
7.3 Past Approaches for Numerical Modelling of Fracture ................................. 273
7.3.1 Numerical modelling attempts using LEFM ........................................... 273
7.3.2 Modelling attempts with cohesive crack ................................................. 274
7.4 Modelling Crack Depths with Cohesive Properties ....................................... 276
7.4.1 Cohesive crack implementation .............................................................. 276
7.4.2 Compacted clay soils............................................................................... 279
7.4.3 Modelling Crack Depths in Soft Soils .................................................... 280
7.5 Predicting Crack Depths in Compacted Clay Soils ........................................ 285
7.5.1 Results obtained using the linearly decreasing suction profile ............... 285
7.5.2 Effects of method of implementing cohesive law ................................... 287
7.5.3 Effect of friction angle with respect to suction ....................................... 289
7.5.4 Depth of desiccation cracks with the cohesive properties ...................... 291
7.6 Predicting Crack Depths in Soft soils ............................................................. 297
7.6.1 Results for crack depth prediction using single crack ............................. 298
xiv
7.6.2 Results for crack depth prediction using multiple cracks ....................... 299
7.7 Conclusion ...................................................................................................... 302
Chapter 8 ....................................................................................................................... 304
CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER RESEARCH...... 304
8.1 Conclusions .................................................................................................... 304
8.2 Future Research Recommendations ............................................................... 307
REFERENCES .............................................................................................................. 309
xv
LIST OF PUBLICATIONS
• Wijesooriya, S., Amarasiri, A. and Kodikara, J. A Numerical Modelling Approach
for Predicting Crack Depths in Drying Clay Soils, International Journal of
Geomechanics, (in preparation).
• Wijesooriya, S. D. and J. K. Kodikara (2012). Experimental study of shrinkage and
swelling behaviour of a compacted expansive clay soil. 11th Australia New Zealand
Conference on Geomechanics (ANZ 2012). Melbourne, Australia.
• Wijesooriya, S. and Kodikara, J. (2011) Prediction of desiccation crack depths
allowing for shear failure, Asia-Pacific conference on unsaturated soils, Thailand.
• Wijesooriya, S., Amarasiri, A. and Kodikara, J. (2011) Modelling of desiccation
crack depths using UDEC in drying clay soils, International Conference of the
International Association for Computer Methods and Advances in Geomechanics,
Melbourne.
• El Maarry, Kodikara, J., Wijesooriya, S. D. and Markiewicz, W. (2011). Numerical
modelling of a desiccation mechanism for formation of Crater Floor Polygons on
Mars and Giant Polygons on Earth: Results from a Pre-Fracture Model, Earth &
Planetary Science Letters.
xvi
xvii
NOMENCLATURE
𝐴 Ratio of effective and total stress differences (𝛥𝜎3′/𝛥𝜎3)
𝐴𝑐 Cross sectional area 𝐴𝑠 Surface area 𝑎 Crack length 𝑎𝑣 Matric suction compression
index 𝐶1, 𝐶2, 𝐶3, 𝐶4, 𝐶5 Constants Cc Compression index Cs Swelling/recompression index 𝑐 Apparent cohesion, bc φtan= 𝑐′ Effective cohesion intercept 𝑐𝑢 Undrained cohesion 𝐷𝑖𝑗𝑘𝑙 Tangent stiffness tensor 𝑑 Depth of the layer 𝐸 Young’s modulus with respect
to total stress relative to pore air pressure
𝐸𝑠 Young’s modulus at saturated level
𝐸𝑇 Evapotranspiration 𝑒 Void ratio 𝑒0 Initial void ratio 𝑒𝑎 Vapour pressure of the air in
the atmosphere above the water surface
𝑒𝑠 Saturation vapour pressure of water at the temperature of the surface
𝑒𝑤 Moisture ratio 𝐹𝑖 Resultant external force 𝑓(𝑢) Turbulent exchange function 𝐺 Shear modulus 𝐺𝑓 Fracture energy
𝐺𝐼𝐶 Rate of release of strain energy with critical crack extension
𝐺𝑠 Specific gravity 𝑔𝑖 Gravitational acceleration 𝐻 Modulus with respect to matric
suction 𝐻𝑑 Depth to permanent water table
from surface Hi Initial height Hs Depth of seasonal changes in
water content from surface 𝐾 Bulk modulus 𝐾𝑎𝑝𝑝 Apparent fracture toughness 𝐾𝐼 Stress intensity factor 𝐾𝐼𝐶 Critical stress intensity factor /
Fracture toughness k Interface shear stiffness 𝑘𝑜 Earth pressure coefficient at
rest 𝑘𝑛 Normal stiffness 𝑘𝑠 Shear stiffness 𝐿 Length of the layer l Size of the mesh M Slope of failure envelope in p, q
space 𝑛 Frequency of seasonal variations in cycles/year 𝑛𝑗 Unit normal to 𝐴𝑠 𝑃 Horizontal stresses at the tip of
the crack
𝑝 Mean stress, (𝜎1 + 2𝜎3)3�
𝑝𝑛𝑒𝑡 Mean stress corrected for air pressure, (𝑝 − 𝑢𝑎)
𝑄 Horizontal stresses at the top surface of the crack
xviii
q Deviator stress, (σ1 - σ3) 𝑅 Rainfall 𝑅𝐸 Net recharge from the
surrounding soil 𝑅𝐸𝑇 Rate of evaporation 𝑅𝑂 Runoff 𝑟 Distance from the crack tip 𝑆 Suction, (𝑢𝑎 − 𝑢𝑤) 𝑆0 Suction at the surface 𝑆𝑒 Equilibrium suction 𝑆𝑠 Gradient of the SWCC 𝑠 Spacing between cracks 𝑡 Time 𝑈 Stored energy 𝑢 Displacement 𝑢𝑎 Pore air pressure 𝑢𝑤 Pore water pressure 𝑢𝚤̈ Acceleration �̇� Velocity 𝑊 Depth to water table / Depth of
the soil layer 𝑤 Moisture content 𝑧 Depth measured from surface 𝑧𝑐 Depth of cracking 𝑧𝑡 Depth of tension crack 𝑧𝑖 Initial depth of cracking 𝛼 Hydric constant 𝛼∗ Hydric coefficient 𝛼𝑇 Coefficient relating cohesion to
tensile strength, 𝛼𝑇 = 0.5 tan ϕ ′ 𝛼𝑑𝑐 Diffusion coefficient 𝛽 Given by �𝑘 𝐸𝐴𝑐⁄ 𝛾 Unit weight
dγ Dry unit weight ∆ Change of the parameter 𝛿𝑘𝑙 Kronecker delta 𝜀 Strain 𝜀𝑘𝑙 Observed strain 𝜀𝑠ℎ Shrinkage strain
shvε Volumetric shrinkage strain 𝜁 Specific surface energy
𝜃 Angle to the failure surface κ Compressibility parameter for
unload / reload condition λ compressibility parameter 𝜈 Poisson’s ratio 𝜇 Lamé constant 𝜉 Lamé constant 𝜙 Angle of shearing resistance 𝜙𝑏 Angle of shearing resistance
with respect to total stress relative to pore air pressure
𝜙′ Effective friction angle 𝜎 Normal stress 𝜎′ Effective stress 𝜎1, 𝜎3 Major principal stress, Minor
principal stress 𝜎𝑐 Stress required to opening the
crack 𝜎𝑓 Failure stress 𝜎𝑖𝑗 Stress tensor 𝜎𝑡 Tensile strength 𝜏 Shear stress 𝜏𝑓 Shear strength 𝜒 Fraction to modify the suction
contribution towards the shear stress
𝜓 Parameter given by 𝑎 W� Subscripts: x, y, z horizontal, vertical, horizontal
coordinate directions 1, 3 major, minor a, w air, water 0 surface max maximum min minimum
xix
ABBREVIATION
AE Actual rate of evaporation
CCL Compacted clay layer
EPFM Elasto plastic fracture mechanics
FEM Finite element modelling
FPZ Fracture process zone
GCL Geo-synthetic clay liner
GM Geo-membrane
LCRS Leachate collection and removal system
LE Linear Elastic
LEFM Linear Elastic Fracture Mechanics
LL Liquid Limit
PE Potential rate of evaporation
PI Plasticity Index
PL Plastic Limit
SF Cracking with shear failure method
SWCC Soil-Water characteristic curve
TMI Thornthwaite Moisture Index
UDEC Universal Distinct Element Code
USCS Unified soil classification system
xx
xxi
LIST OF TABLES
Table 2-1 Summary of Desiccation crack depths recorded ............................................ 46
Table 2-2 Variation of tensile strength with test method ................................................ 54
Table 2-3 Summary of research used numerical models ................................................ 73
Table 2-4 Comparison of several programs .................................................................... 76
Table 4-1 Mineralogy content of Altona clay ............................................................... 143
Table 4-2 Summary of the soil classification test results .............................................. 144
Table 4-3 Properties of the soil ..................................................................................... 162
Table 4-4 Summary of tests by Sharma ........................................................................ 164
Table 4-5 Summary of soil parameters ......................................................................... 170
Table 4-6 Summary of tests by Romero ....................................................................... 173
Table 4-7 Soil Properties ............................................................................................... 176
Table 4-8 Summary hydric coefficients of tests by Tripathy ........................................ 178
Table 4-9 Soil Parameters for sand bentonite mixes ..................................................... 182
Table 4-10 Summary of the tests by Montanez ............................................................ 184
Table 4-11 Properties of London Clay .......................................................................... 187
Table 4-12 Summary of tests by Monroy ..................................................................... 189
Table 4-13 Summary of hydric coefficient values ........................................................ 197
Table 5-1 Soil parameters for Regina clay .................................................................... 214
Table 5-2 Soil properties of Horsham clay ................................................................... 217
Table 5-3 Soil properties of Altona clay ....................................................................... 219
Table 6-1 Typical input parameters of the model ......................................................... 243
xxii
xxiii
xxiv
LIST OF FIGURES
Figure 2-1 Cross section of a landfill (http://www.groundwateruk.org/Image-Gallery.aspx) .............................................................................................. 10
Figure 2-2 Layers of base (bottom) lining systems ......................................................... 14
Figure 2-3 Acceptable zone based on design objectives for hydraulic conductivity, volumetric shrinkage, and unconfined compressive strength (Daniel and Wu 1993) ............................................................................................. 15
Figure 2-4 Different types of cover layers (Dwyer, 1998; Albright et al., 2004) ........... 16
Figure 2-5 Variation of moisture content with depth, Curve 1, equilibrium water content under a surface covering; curve 2, dry season; curve 3, wet season (after Kraynski, (Chen 1988)) ........................................................ 24
Figure 2-6 Effect of drying and rewetting on the ultimate moisture content of Leda clay (after warkentin (1961)) ..................................................................... 25
Figure 2-7 Typical seasonal soil suction variations with depth (Morris et al., 1992) ..... 26
Figure 2-8 Soil water characteristic curves (Fredlund and Anqing 1994) ...................... 27
Figure 2-9 Schematic Profile View of a Typical Hazardous Waste Landfill ................. 29
Figure 2-10 The relationship between the rate of actual evaporation and potential evaporation (i.e., AE/PE) and water availability (Wilson, Fredlund et al. 1994) ..................................................................................................... 34
Figure 2-11 Leachate generation rates at a modern domestic landfill in Pennsylvania (USA), Average annual precipitation at the landfill site is 1.0 m/year (after Bonaparte 1995, (Bouazza and Van Impe 1998)) ...... 36
Figure 2-12 Daily precipitation and resulting drainage rate for periods (a) before and after (b) the fall 2000 drought at Albany, GA. Daily precipitation is shown as vertical bars, drainage rate as a continuous line (Albright, Benson et al. 2006). ................................................................................... 39
Figure 2-13 Time-movement plots of several points inside and outside the National Art Gallery building, (Richards, Peter et al. 1983) .................................... 40
Figure 2-14 Variation of vertical deformation with several wet-dry cycles, (Tripathy, Subba Rao et al. 2002) .............................................................. 41
xxv
Figure 2-15 Shrinkage phases of clay upon drying (Kodikara et al., 1999) ................... 42
Figure 2-16 Theoretical patterns of desiccation cracks (a) parallel (b) square (c) hexagonal (after bezant and Cedolin, 1991, (Kodikara, Barbour et al. 2000)) ......................................................................................................... 43
Figure 2-17 Parallel cracks observed in long thin moulds.(Costa, Kodikara et al. 2008) .......................................................................................................... 43
Figure 2-18 Irregular shaped crack pattern observed on a playa surface, southern Nevada, Hammer and handle measured 330mm (Longwell 1928). Playa surface means a flood plain clay surface or dried-out lake surface. ....................................................................................................... 44
Figure 2-19 Orthogonal Cracking pattern observed in coal mine tailings, Queensland, Australia (Morris, Graham et al. 1992) ................................. 44
Figure 2-20 Hexagonal mud crack pattern observed on playa surface, Las Vegas quadrangle, Nevada. (Longwell 1928) ...................................................... 45
Figure 2-21 Vertical crack pattern of a compacted clay layer with 3 lifts (Yesiller et al. 2000) ..................................................................................................... 45
Figure 2-22 Schematic illustration of cracking (Konrad and Ayad 1997a) .................... 49
Figure 2-23 Influence of dry densities on σt of the clay with different water contents (Wang, Zhu et al. 2007) ............................................................... 55
Figure 2-24 Influence of water contents on σt of the clay with different dry densities (Wang, Zhu et al. 2007) .............................................................. 56
Figure 2-25 Details of tensile strength with constant natural density, (c) tensile strength with moisture content with polynomial fit showing the trend (d) tensile strength with degree of saturation (Lakshmikantha, 2009) ...... 56
Figure 2-26 Mohr-Coulomb failure criterion .................................................................. 60
Figure 2-27 Strength envelopes indicating tensile strength after Lee and Ingles in 1968 (Morris, Graham et al. 1992) ............................................................ 60
Figure 2-28 Variation of stress intensity factor with crack depth (Morris, Graham et al. 1994) ..................................................................................................... 65
Figure 2-29 Stress analysis for desiccation cracking (Harison and Hardin, 1994) ......... 67
Figure 2-30 Schematic representation of the proposed model by Kodikara and Choi (2006 (a)) ................................................................................................... 69
xxvi
Figure 2-31 Schematic diagrams for the (a) stress relief approach (b) energy balance approach (Costa 2009) .................................................................. 70
Figure 2-32 Flow chart of the proposed model by Konrad and Ayad (1997(a)) ............ 78
Figure 3-1 Suction profile (a) and tensile strength profile (b) with depth ...................... 85
Figure 3-2 Depths of cracking with different Poisson’s ratios tensile strength values when the constant suction profile is assumed throughout the depth .......... 87
Figure 3-3 Suction and tensile strength profiles when linearly decreasing with depth .. 88
Figure 3-4 Predicted crack depth values for linearly decreasing suction profile through linear elastic (LE) approach and allowing for shear failure (SF) approach with different surface suction values.................................. 89
Figure 3-5 Parabolic suction and tensile strength profiles .............................................. 90
Figure 3-6 Predicted Crack depths when the water table is 4m below the surface (LE-linear elastic approach, SF-elastic allowing for shear failure and LEFM-linear elastic fracture mechanics approach) ................................... 91
Figure 3-7 Predicted Crack depths using parabolic suction profile when the surface suction is 50kPa ......................................................................................... 92
Figure 3-8 Cross section of the clay layer used for the model ........................................ 95
Figure 3-9 Variation of depths of cracking with layer width when E =5MPa, ν =0.3, σt=0kPa and S0=50kPa ............................................................................. 96
Figure 3-10 Zones and blocks of the model .................................................................... 97
Figure 3-11 Variation of crack depths with mesh size when E =5MPa, ν =0.3, σt=0kPa and S0=50kPa ............................................................................. 98
Figure 3-12 Stress conditions in a soil particle at each increment ................................ 100
Figure 3-13 Effect of the Poisson’s ratio on crack depth when E =5MPa, σt =0.5 S tanϕb, W =4m and S0=50kPa ........................................................ 103
Figure 3-14 Crack depths changing with the depth to water table when Es=5MPa, E = Es + 10S , ν =0.3, σt = −αTS tanϕbcotϕ′ , W =4m, S0 =50kPa and ϕ =300 ............................................................................................... 104
Figure 3-15 Appling suction with time ......................................................................... 105
Figure 3-16 Crack of a model after opening ................................................................. 106
xxvii
Figure 3-17 Variation of crack depth with surface suction when E =5MPa, ν =0.3, W =10m, σt=constant and S0=50kPa ...................................................... 107
Figure 3-18 Variation of crack depth with surface suction when E =5MPa, ν =0.3, W =4m σt=constant and S0=50kPa ......................................................... 108
Figure 3-19 Changing the suction in UDEC ................................................................. 109
Figure 3-20 Variation of crack depth with depth to water table when E =5MPa, ν =0.3, σt = 0.5 S tanϕb ............................................................................ 110
Figure 3-21 Variation of crack depth with the surface suction when when E =5MPa, ν =0.3, σt = 0.5 S tanϕb and W =4m ..................................................... 111
Figure 3-22 (a) Strength envelope indicating tensile failure showing the effect of tensile strength, (b) Schematic of failure envelope in tension ................. 112
Figure 3-23 Variation of depth of cracking with depth to water table when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′, S0=50kPa and ϕ =300 ........... 113
Figure 3-24 Variation of crack depth with surface suction when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′, W =4m and ϕ =300 ......................................... 114
Figure 3-25 Applying suction in increments for parabolic variation ............................ 114
Figure 3-26 Effect of surface suction on depth of cracking predictions for parabolic suction variation; when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′ , W =4m and ϕ =300 .................................................................................. 115
Figure 3-27 Effect of depth to water table on depth of cracking predictions for parabolic suction variation; when E =5MPa, ν =0.3, σt =−αTS tanϕbcotϕ′, S0=50kPa and ϕ =300 ............................................. 116
Figure 3-28 Depth of cracking variation with surface suction for parabolic suction profile when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′, W=4m and ϕ =300 ...................................................................................................... 116
Figure 3-29 Depth of cracking variation with water table depth for parabolic suction profile when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′ , S0=50kPa and ϕ =300 ............................................................................................... 117
Figure 3-30 Behaviour of crack depth predictions with surface suction based on the suction profile; when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′ , W=4m and ϕ =300 .................................................................................. 118
xxviii
Figure 3-31 Behaviour of crack depth predictions with depth to water table based on the suction profile; when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′, S0=50kPa and ϕ =300 .................................... 119
Figure 3-32 Behaviour of crack depth predictions with surface suction based on suction profile allowing for shear failure; when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′, S0=50kPa and ϕ=300 ...................................... 120
Figure 3-33 Behaviour of crack depth predictions with surface suction based on the suction profile; when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′ , W=4m and ϕ =300 .................................................................................. 120
Figure 3-34 Comparison of results obtained through different theoretical assumptions with depth to water table variation when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′, S0=50kPa and ϕ =300 (SS – allowing for shear failure, LE – Linear elastic) ...................................................... 121
Figure 3-35 Comparison of results obtained through different theoretical assumptions with surface suction variation, when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′, S0=50kPa and ϕ =300 (SS – allowing for shear failure, LE – Linear elastic) ............................................................ 122
Figure 3-36 Comparison of results obtained through theoretical assumptions with depth to water table variation when E =5MPa, ν =0.3, σt =−αTS tanϕbcotϕ′, S0=50kPa and ϕ =300 (SS – allowing for shear failure, LE – Linear elastic) ..................................................................... 123
Figure 3-37 Comparison of results obtained through different theoretical assumptions with surface suction variation when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′ , S0=50kPa and ϕ =300 (SS – allowing for shear failure, LE – Linear elastic) ............................................................ 124
Figure 3-38 Comparison of values of crack depth with surface suction variation when E =5MPa, ν =0.3, σt = −αTS tanϕbcotϕ′, W=4m and ϕ =300 . 126
Figure 3-39 Design curve; zcW vs. S0γW curve when ϕ =300, ν =0.3 for suction profile given in Figure 3-19 ..................................................................... 128
Figure 4-1 Particle size distribution of Altona clay ...................................................... 138
Figure 4-2 Automatic soil compactor designed for proctor and CBR Compaction and removing sample after modified compaction test ............................. 140
Figure 4-3 Compaction curves for Altona clay ............................................................. 141
Figure 4-4 Mechanical soil mixer ................................................................................. 142
xxix
Figure 4-5 LoadTrac III consolidation machine ........................................................... 142
Figure 4-6 Compression curve for Altona clay ............................................................. 143
Figure 4-7 Sample preparation using method one ........................................................ 147
Figure 4-8 Sample preparation using method two ........................................................ 148
Figure 4-9 Values for unit weight and moisture content obtained as initial conditions. * - UW stands for Unit Weight, ** - MC stands for Moisture Content ..................................................................................... 149
Figure 4-10 Monash designed Oedometer .................................................................... 150
Figure 4-11 Schematic diagram of test set-up and placement of soil sample ............... 151
Figure 4-12 (a) Saturating the sample for swelling, (b) Wetting the sample to get the swell before saturation ....................................................................... 152
Figure 4-13 Vertical displacement of three samples compacted to similar initial conditions when subjected to wetting. ..................................................... 153
Figure 4-14 Vertical displacement of three samples compacted to similar initial conditions when subjected to drying after full swelling .......................... 154
Figure 4-15 Photos of the sample at different stages .................................................... 155
Figure 4-16 Marginal vertical displacements during wet-dry cycles for several samples ..................................................................................................... 156
Figure 4-17 Typical average vertical displacements for Altona clay ........................... 157
Figure 4-18 Swelling and shrinking paths for the first cycle ........................................ 158
Figure 4-19 Swelling and shrinking paths for the second cycle ................................... 159
Figure 4-20 Swelling and shrinking paths for the third cycle ....................................... 159
Figure 4-21 Swelling and shrinking paths for the fourth cycle ..................................... 160
Figure 4-22 Variation of α ∗ with number of cycles ..................................................... 161
Figure 4-23 Variation of α ∗ with moisture ratio .......................................................... 161
Figure 4-24 Compaction curves and the initial positions of samples ........................... 163
Figure 4-25 Swelling and shrinking curves for samples under 10kPa vertical stress, Test 4 – initially compacted under 800kPa. Test19 & 20 – initially
xxx
compacted under 3200kPa. Other tests – initially compacted under 400kPa ..................................................................................................... 167
Figure 4-26 Swelling and shrinking curves for samples under 20kPa (Test 5) and 50kPa (Test 2) vertical stress, Test 3 – initially compacted under 800kPa. Test 2 & 5 – initially compacted under 400kPa ......................... 167
Figure 4-27 Variation of α ∗ with wet-dry cycles ......................................................... 168
Figure 4-28 Variation of α ∗ with degree of saturation in each wetting or drying process ..................................................................................................... 169
Figure 4-29 Compaction curves for Boom clay ............................................................ 171
Figure 4-30 Swelling shrinking curves tested under different pressures ...................... 172
Figure 4-31 Variation of α ∗ with number of cycles ..................................................... 175
Figure 4-32 Standard Proctor curves for Soil A and Soil B .......................................... 177
Figure 4-33 Swelling and shrinking curves for several wet-dry cycles ........................ 179
Figure 4-34 Hydric coefficient change with wetting and drying for the Soil A with 6.25kPa .................................................................................................... 180
Figure 4-35 Hydric coefficient variation during several wet-dry cycles ...................... 180
Figure 4-36 Summary of variation of hydric coefficient in each wetting or drying process ..................................................................................................... 181
Figure 4-37 Compaction curves for samples with different compositions; WG – Well Graded U – Uniform and B – Bentonite ......................................... 183
Figure 4-38 Behaviour of sand samples with different initial conditions under wetting and drying ................................................................................... 185
Figure 4-39 Hydric coefficient values for wet-dry processes ....................................... 186
Figure 4-40 Compaction characteristics of London clay from standard compaction test ............................................................................................................ 188
Figure 4-41 Variation of void ratio with moisture ratio during wetting and drying ..... 193
Figure 4-42 Variation of α ∗ values with different wetting and drying paths ............... 194
Figure 4-43 Moisture content variation in a suction controlled test ............................. 195
Figure 4-44 Typical paths of expansive soils subjecting to partial wet dry cycles ....... 198
xxxi
Figure 4-45 Typical variations of α ∗ with full wet dry cycles ..................................... 200
Figure 5-1 Effect of various environmental conditions on the matrix suction profile (Peter 1979) .............................................................................................. 203
Figure 5-2 Values of suction from road site installations and postulated design curves after Richards, B.G. (1985) .......................................................... 204
Figure 5-3 Typical Suction Profile for arid and semi-arid conditions used in the present study ............................................................................................ 208
Figure 5-4 Predicted crack depths using different stress changing approaches for Regina clay soil ........................................................................................ 210
Figure 5-5 SWCC for Regina clay after (Vu, Hu et al. 2007) ...................................... 215
Figure 5-6 Void ratio vs. water content graph for Regina clay ..................................... 216
Figure 5-7 SWCC for Horsham clay (after Richards, 1985) ........................................ 218
Figure 5-8 Void ratio vs. water content curve for Horsham clay (after Richards, 1985) ........................................................................................................ 218
Figure 5-9 SWCC for Altona clay (after Chan, 2012) .................................................. 220
Figure 5-10 Void ratio vs. water content curve for Altona clay ................................... 220
Figure 5-11 Suction profiles under different climatic conditions ................................. 221
Figure 5-12 Predicted depth of cracking change with climatic conditions ................... 222
Figure 5-13 Suction profile variation due to seasonal climate change in (a) Arid area (b) Semi-arid area ............................................................................. 223
Figure 5-14 Predicted depth of cracking with seasonal change in an arid area ............ 224
Figure 5-15 Predicted depth of cracking with seasonal change in an semi-arid area ... 225
Figure 5-16 Predicted crack depth variation with initial density of the layer in an arid climate .............................................................................................. 227
Figure 5-17 Predicted crack depth variation with initial density of the layer in an semi-arid climate ...................................................................................... 228
Figure 5-18 Predicted depth of cracking variation with the Poisson's ratio in an arid area ........................................................................................................... 229
Figure 5-19 Predicted depth of cracking variation with the Poisson's ratio in an semi-arid area ........................................................................................... 230
xxxii
Figure 5-20 Suction profiles below the ground surface with different equilibrium suction values in (a) an arid climate (b) a semi-arid climate ................... 231
Figure 5-21 Predicted depth of cracking with the change of equilibrium suctions in an arid climate .......................................................................................... 232
Figure 5-22 Predicted depth of cracking with the change of equilibrium suctions in an semi-arid climate ................................................................................. 233
Figure 6-1 Sign convention for positive shear stress components ................................ 237
Figure 6-2 Actions performed during one computation cycle ...................................... 238
Figure 6-3 Area associated to grid point P .................................................................... 239
Figure 6-4 (a) Problem geometry and modelled problem (b) analysed problem due to symmetry ............................................................................................. 242
Figure 6-5 Change of mesh size in a constant size block ............................................. 245
Figure 6-6 Selection of Failure stress............................................................................ 246
Figure 6-7 Effect of mesh size; K = 5e9 Pa, G = 2e9 Pa, kn = 5e9 Pam, Damp = 0.2, N = 10000, W = 0.6m a = 0.2 m and σt = 1000Pa. .......................... 247
Figure 6-8 Stress vs. displacement curve using to get Overall E .................................. 250
Figure 6-9 Effect of number of stress increments; K = 5e9Pa, G = 2e9Pa, kn=5e9Pa, Damp=0.2, W =0.6m, σt=1000Pa, a =0.02m and l =0.2m ...................... 251
Figure 6-10 Change of load vs. displacement plot with the change of number of cycles ....................................................................................................... 252
Figure 6-11 Model geometry change with changing crack length ................................ 253
Figure 6-12 Effect of crack length, K = 5e9Pa G = 2e9Pa kn=5e9Pam, Damp=0.2, N=10000, W =1.2m, σt=1000Pa, l =0.04m and KIC=530Pam0.5 ............ 254
Figure 6-13 Effect of damping value, K = 5e9Pa, G = 2e9Pa, kn =5e9Pam, N=10000, W =0.6m, σt=1000Pa and al =10 (a =0.02m, l =0.2m) .......... 255
Figure 6-14 Stress vs. displacement curves when changing the damping value .......... 255
Figure 6-15 Change of Size of block when the crack length kept constant .................. 256
Figure 6-16 Normal Stresses from UDEC and LEFM ahead of the crack.................... 257
Figure 6-18 Scale effect on failure stress, K = 5e9Pa, G = 2e9Pa, Damp=0.2, N=10000 and σt=1000Pa ......................................................................... 258
xxxiii
Figure 6-17 Plot of models changing the size of geometry .......................................... 258
Figure 6-19 Inherent fracture energy present in current UDEC formulation................ 259
Figure 6-20 Scale effect on Fracture energy, K = 5e9Pa, G = 2e9Pa, Damp=0.2, N=10000 and σt=1000Pa ......................................................................... 260
Figure 6-21 Effect of normal stiffness on failure stress, K = 5e9Pa, G = 2e9Pa, Damp=0.2, N=10000 and σt=1000Pa ...................................................... 261
Figure 6-22 Effect of normal stiffness on fracture toughness, K = 5e9 Pa, G = 2e9 Pa, Damp=0.2, N=10000 and σt=1000Pa ................................................ 261
Figure 6-23 Effect of modulus on failure stress kn=5e9Pam, Damp=0.2, N=10000 and σt=1000Pa ......................................................................................... 263
Figure 6-24 Effect of modulus on failure stress, kn=5e9 Pam, Damp=0.2, N=10000 and σt=1000Pa ......................................................................................... 263
Figure 7-1 Semi infinite plate with the central crack of the length 2a and the directions for near tip stress field ............................................................. 269
Figure 7-2 Actual incremental fracture process in Load-Displacement space ............. 270
Figure 7-3 Distribution of the stress normal to the crack plane (Wang, 1996) ............. 272
Figure 7-4 Bridging stresses at the crack tip while crack opening considered in the cohesive crack models ............................................................................. 275
Figure 7-5 Softening curves .......................................................................................... 277
Figure 7-6 Linear softening law used in the numerical model ...................................... 278
Figure 7-7 Moisture content profiles of Saint-Alban clay test (Konrad and Ayad, 1997a) ...................................................................................................... 281
Figure 7-8 SWCC for Saint-Alban clay and the empirical equations for different moisture content ranges ........................................................................... 282
Figure 7-9 Depth of cracking variation for cohesive model with surface suction when E =5MPa, ν =0.3, σt = −αTStanϕbcotϕ′, W =4m and ϕ =300 ... 286
Figure 7-10 Depth of cracking variation for cohesive model with depth to water table when E =5MPa, ν =0.3, σt = −αTStanϕbcotϕ′, S0=50kPa and ϕ =300 ...................................................................................................... 286
Figure 7-11 Different methods of implementing cohesive law (a) Softening Law change-1 (b) Softening Law change-2 (c) Softening Law change-3 ....... 287
xxxiv
Figure 7-12 Depth of cracking for different implementations of softening law with the change of surface suction ................................................................... 288
Figure 7-13 Effect of friction angle with respect to suction on desiccation cracks ...... 290
Figure 7-14 Depth of cracking for Regina clay using different softening laws with changing surface suction .......................................................................... 292
Figure 7-15 Depth of cracking for Horsham clay using different softening laws with changing surface suction .......................................................................... 293
Figure 7-16 Depth of cracking for Altona clay using different softening laws with changing surface suction .......................................................................... 294
Figure 7-17 Fracture energies with different softening laws ........................................ 295
Figure 7-18 Effect of fracture energy on desiccation cracking in terms of residual limit .......................................................................................................... 296
Figure 7-19 Effect of fracture energy on desiccation cracking ..................................... 297
Figure 7-20 Crack depth with time ............................................................................... 298
Figure 7-21 Block with displacement and joints........................................................... 299
Figure 7-22 Block with opened cracks and displacement vectors ................................ 300
Figure 7-23 Progression of multiple cracks with time .................................................. 301
xxxv
Chapter 1
INTRODUCTION
1.1 General
Desiccation takes place from a soil surface when it is exposed to the atmosphere. The
degree of desiccation depends on soil properties, the climate condition and the depth to
the water table mainly. When desiccation progresses, especially during the dry seasons,
water escapes from the surface and soil water suctions develop making soils shrink.
While the soil is free to shrink vertically, it is normally laterally constrained against
shrinkage. This leads to formation of desiccation cracks at the upper portion of the soil
surface.
Desiccation cracking can induce severe problems in agricultural, geotechnical, and
environmental applications. In agricultural engineering, cracks can influence the
undesirable water and solute flow through soil in irrigated lands. In geotechnical
engineering, the presence of tension cracks due to desiccation may also badly influence
the stability of natural slopes (Baker and Leshchinsky, 2003; Zhan et al., 2007) and
1
2
vertical cuttings, as well as the bearing capacity of foundations (Silvestri et al., 1990). In
addition to that, they may lead to pipe settlements and breakages and dam failures
(Vasil'ev et al., 1988). In the field of environmental engineering, desiccation cracking
has the potential to render a low conductivity barrier constructed of a clay soil layer
ineffective within a short period of time (Melchior, 1997; Khire et al., 1999; Albright et
al., 2006). Deep desiccation cracks have been observed in arid areas especially in
expansive clay soils and in coal mine tailings deposits in Queensland, Australia (Morris
et al., 1992) with typical depths of about 1 to 1.5 m and, rarely, with a maximum depth
of about 4 m.
Due to these adverse effects, desiccation cracking has become a concern in geotechnical
and geo-environmental engineering, especially in design and construction of landfill
covers in arid regions. Significant amount of research (e.g., Konrad and Ayad, 1997b;
Philip et al., 2002; Kodikara and Choi, 2006; Tang et al., 2008) has been conducted for
a long period extending over decades on various aspects of this problem. However,
there are still further advancements to be made, in particular in the area of prediction of
depths of desiccation cracks in a given climate.
Soils which normally contain significant amount of clay with minerals of high
shrink/swell potential, can undergo large volume changes due to moisture content
variations. These types of soils are often called ‘expansive or reactive soils’. Expansive
soils cause distress to light-weight (e.g., foundations of light buildings) or surficial
structures (e.g., road pavements, pipelines) making considerable damage annually. As a
damage mitigative measure, vertical moisture barriers are used commonly, particularly
in the case of road pavements (Picornell and Lytton, 1987; Jayatilaka and Lytton, 1997;
Introduction 3
Aubeny and Long, 2007). However, knowledge of the maximum crack depth is valuable
to design damage mitigative measures such as moisture barriers. Furthermore, the depth
of desiccation cracks that are likely at the crest and down the slope are important
considerations in assessing slope stability (e.g., Dyer et al., 2009).
In compacted soil barriers, the infiltration rate depends on the configuration and the
depth of the fracture. On the basis of a numerical study on the infiltration of water
through a fractured clay soil, Moore and Ali (1982) remarked that the depth of cracking
played a significant role, whereas the cracking frequency was less important for overall
infiltration. However, the depth and the frequency of cracking should be considered
together to determine the overall infiltration. The worst condition would occur when
the depth of cracking and frequency are both very high.
Many analytical and numerical models have been used to model the desiccation crack
depths (Morris et al., 1992; Konrad and Ayad, 1997b; Amarasiri and Kodikara, 2011c).
In crack development in field conditions, mostly linear elastic stress analysis or linear
elastic fracture mechanics is used. Only recently, Kodikara and co-workers have
introduced methods to incorporate significant non-linearity present in fracture process.
Therefore, there is significant scope for undertaking research on crack depth prediction
under field conditions.
1.2 The Scope of this Study
The primary purpose of this study is to develop a reliable numerical modelling
technique to simulate the development of desiccation cracking in clay soils with
4
particular emphasis on compacted clay liners. Hence, this research focuses on the
phenomenon of desiccation cracking. Numerical model predictions are compared with
empirical solutions and experimental observations and the critical parameters involved
in the prediction of depth of desiccation cracking. The basic research aims can be listed
as:
i. To predict the level of desiccation cracking in a clay layer subject to certain
climate conditions;
ii. To find the range of moisture and suction change causing the cracks and their
effect on depth of cracking in compacted clay layers;
iii. To identify and obtain the value ranges of critical parameters that affect the depth
of cracking using laboratory experiments, available literature and numerical
analysis;
iv. To propose an analytical model to predict the depth of cracking; and
v. To evaluate the effects of different soil parameters on the prediction of crack
depths.
To achieve these aims, the study presented in this thesis was conducted comprising a
literature review, numerical analysis, analytical development and laboratory
experiments to support the numerical approach proposed.
1.3 The Structure of the Thesis
This thesis is divided into eight chapters, a bibliography and a list of publications
produced from the work described in this thesis.
Introduction 5
Chapter 1: Introduction
This chapter introduces the problem; “Desiccation cracking in clay liners” generally,
and explains the basic points of the thesis and points to the directions of the research. It
will also give a guide for the thesis structure.
Chapter 2: Literature Review
This chapter presents a wide review of the past research which is relevant to the
desiccation cracking in clay liners to identify the seminal work done and the gaps in the
literature.
The existing knowledge on desiccation problem of landfill liners and modelling
techniques developed so far is summarised and discussed. Fundamentals of the soil
behaviour are reviewed with emphasis on mechanics of unsaturated soils. Finally, the
gaps in the state of the art are presented.
Chapter 3: Comparison of crack modelling approaches
The crack depths are calculated analytically and numerically through different
theoretical assumptions such as:
• Linear elastic approach;
• Shear failure approach;
• Linear elasticity fracture mechanics approach; and
• Incorporation of fracture energy using cohesive crack modelling.
For numerical modelling, the UDEC computer program is selected. The results are
obtained by changing the depth to water table from the surface and the suction within
the block. These results are compared and discussed with analytical results.
6
Chapter 4: Experimental Investigation of shrinkage and swelling behaviour
The behaviour of the soil layer with the moisture content is necessary to be observed to
predict the accurate suction profile. Hence, the laboratory tests are mainly focused on
the shrinking and swelling behaviour to represent the field soil behaviour when
subjected to climatic wet dry cycles.
Through these tests the input parameters required for the progress of the research are
obtained. The main difficulty of these tests is measuring the soil volumes accurately
with the presence of cracks.
Chapter 5: Modelling of stable desiccation crack depths during cyclic wetting and drying
In this chapter, the stable crack depths are modelled for compacted clay liners by
considering the cyclic effect of wet and dry cycles. Several clay soils, namely Regina,
Horsham and Altona, from different parts of Australia and from some other countries,
are used in the analysis. The evaluation of crack depth in different climate conditions
and initial placement conditions of the clay layer is considered.
Chapter 6: Inherent properties of UDEC for fracture modelling
This chapter is used to understand the inherent properties of UDEC for fracture
modelling. First, the fracture modelling approach of UDEC analysis is explained. The
effect of different input parameters is then analysed and discussed on the basis of
generated results.
Introduction 7
Chapter7: Modelling of desiccation crack depths incorporating soil fracture energy
In this chapter, the stress reduction while opening a crack is included in the code using
FISH, the inbuilt programming language of the numerical program UDEC. The values
of the softening curve are calculated on the basis of the fracture energy of the problem
geometry. Then the crack depths are predicted for the same soils used in Chapter 5. The
effect of cohesive properties of the fracture during its initiation and development is
studied in this chapter.
Chapter 9: Conclusions and Future recommendations
Conclusions derived from this study and recommendations for future research are
presented in this chapter.
Chapter 2
LITERATURE REVIEW
2.1 Introduction
Issues related to desiccation have been studied first qualitatively and then quantitatively
since the 1950s or so. The use of landfill liners received increasing attention with the
advent of geo-environmental engineering in the 1970s, where the importance of
addressing desiccation was highlighted. In general, the improved knowledge developed
on landfill liner design and performance has led to more efficient designs, precautions
and solutions for a wider range of existing problems of the landfill liners.
This chapter presents the existing knowledge on landfill liners, and their desiccation
problem and modelling techniques developed. Fundamentals of the soil behaviour will
be reviewed with emphasis on mechanics of unsaturated soils, as applicable to
compacted soils.
8
Literature Review 9
2.2 Clay Liners in Landfills
A landfill or a contaminated site should always be isolated from the natural environment
to minimize or eliminate the potential pollution of groundwater resources. To achieve
this, low permeable layers are placed around the waste mass. Clay layers are often used
as these low permeable layers due to the availability of clay soils, cost effectiveness of
construction and general good performance of the clay layers. The other materials used
for liners include concrete, asphalt and geo-composites such as geosynthetic clay liners.
The low conductivity layers, placed at the sides, bottom and top of the waste body are
referred to as side, base and cover liner respectively. The term ‘liners’ refers to both
side and base liners and a liner can basically can be identified as a compacted clay layer
that:
• Slows and retards the leachate migration into the ground;
• Prevents bio-gas escape to the environment; and
• Provides mechanical support to the waste mass.
However the role of a properly designed and constructed landfill cover liner is more
complex than that of bottom and side liners. The role of cover liner can be identified as
(Holzlohner et al., 1995) as to:
• Limit the percolation of surface waters through to the waste below and down to
the bottom liner;
• Prevent the direct uptake of contaminants by organisms;
• Control gas fluxes which create a hazard in the vicinity;
10
• Reduce dissolving contaminants and formation of leachate which can pollute
both soil and groundwater;
• Enhance aesthetic appearance; and
• Provide support for the aftercare options.
Gas extraction pipe
Leachate drain pipesLeachate collection and
drainage system
Impermeable clay cap
Synthetic membrane liner
Waste
Compacted low-permeability clayLeachate
collection sumpNeutral rock foundation
Figure 2-1 Cross section of a landfill (http://www.groundwateruk.org/Image-Gallery.aspx)
The knowledge of designing landfill cover liners have developed considerably over the
last two to three decades. Most of countries have their own regulatory requirements for
the construction of landfill liners. For example in Victoria, Australia, information
bulletins are published by environment agency (EPA, 2008). Similarly, other countries
have also developed various guidelines for landfill design. Holzlohner et al. (1995) and
Manassero et al. (2000) summarized major typical requirements adapted by different
countries for the design of landfill liners for hazardous and municipal waste disposal.
Accordingly, the design of covers mainly depends on several factors such as:
• climatic conditions of the site area;
• geo-mechanical properties;
Literature Review 11
• environmental risks of the contaminated area;
• waste type (i.e. hazardous, municipal or organic waste);
• intended time period; and
• cost.
In addition to government publications, design techniques and guidelines have also been
published by many researchers in technical journals and conference proceedings
(Bouazza and Van Impe, 1998; Kodikara, 2001).
2.2.1 Evaluation of clay liner designs
Prior to 1975, the emphasis and hence the regulatory restrictions placed on environment
pollution or protection was significantly less. Therefore, clay liners to isolate waste
were not designed on a proper scientific basis. The recognition of the importance of
construction procedures on the field scale performances of compacted clay liners
(CCLs) and the consequent development of guidelines for improved installation were
undertaken in late 1980s. Notable contributions towards proper engineering of
composite barriers consisting of mineral liners (CCLs) were made by Daniel (1989),
Jessberger (1994), and Benson (1994).
Since 1990, the main steps in the progress of modern solid waste containment systems
can be presented approximately in chronological order as follows (Manassero et al.,
2000):
12
• Recognition of the importance of construction procedures on the field scale
performances of compacted clay liners (CCLs) and the consequent set up of
guidelines;
• Current and correct employment of composite barriers consisting of mineral
liners (CCLs) or geo-synthetic clay liners (GCLs), placed in close contact with a
geo-membrane (GM);
• Introduction and in some cases standardization of laboratory and field tests for
the evaluation of barrier components and monitoring systems for assessing full
scale liner performances;
• Recognition of potential stability problems of different types of waste deposits,
looking in particular at sliding surfaces involving interfaces of composite barrier
systems;
• Recognition of the importance of compatibility, diffusive transport and sorption
phenomena on the overall performance of barrier systems;
• Recognition of the importance of biogas migration from landfills and the
consequent research work on gas-barrier interaction;
• Recognition of the importance of natural and manmade attenuation layers below
waste deposits in order to reduce impact of pollutants on groundwater;
• Recognition of the role played by the deformations and settlements of subgrade
layers on the performance of mineral barriers;
Literature Review 13
• Introduction of geo-synthetic clay liners (GCLs) as pollutant containment
barriers;
• Use (in some cases after appropriate modifications) of traditional geotechnical
construction techniques to install barriers around polluted subsoils (e.g. simple
and composite slurry walls, grouted and jet grouted bottom barriers, reactive
diaphragm walls, etc.);
• Introduction of new barrier types for cover systems (e.g. capillary barrier and
natural buffer barriers); and
• Introduction of performance design and related risk analyses overcoming the
prescriptive design procedures. Performance design has become feasible and
reasonably reliable owing to the goals of the aforementioned theoretical and
experimental work and practical experiences. These have led to substantial
improvements in terms of both modelling techniques and knowledge about the
different input parameters that are necessary to simulate the actual behaviour of
waste containment systems.
2.2.2 Types of clay liners
As described earlier, many different types of liners are used in landfills. Bottom layers
can be basically divided into two types, namely, single bottom liner systems and double
or multiple bottom liner systems (Bagchi, 2004). However, cover layers can be
considered in many different types such as compacted clay covers (prescriptive covers),
evapotranspiration covers (Zornberg et al., 2003), capillary barrier covers (Dwyer,
14
1998) (alternative covers) and so on. These alternative covers are designed to overcome
the problems in prescriptive cover designs. Unfortunately, however, alternative covers
do not completely solve those problems because of the complexity of the nature.
Waste
Synthetic impermeable layer
Waste
Compacted base
Compacted clay layerLeachate
collection pipes
Drainage layer
Drainage layer
Compacted clay layer
Leachate collection
pipes
Compacted subbase
Figure 2-2 Layers of base (bottom) lining systems
The main components of the bottom lining systems are the drainage layers or leachate
collection and removal system (LCRS), the compacted clay layer (CCL) and the
attenuation layer or geological barrier. Apart from the geological barrier, which of
course can only be of the natural type, the other layers can be natural or constructed of
amended soils. Generally, regulations require a (saturated) hydraulic conductivity value
of the CCL less than 10-9m/s to be achieved. These low values of hydraulic conductivity
could only be obtained if clods and inter-clod pores are eliminated, which can most
likely be achieved if the soil is compacted wet of optimum, with high compaction effort
and methods producing large shear strains, such as the use of sheep foot rollers. The
permeability of a clay soil compacted at few percentage points wet of optimum may be
10 to 1000 times less than the permeability of the same soil compacted to the same dry
density at a water content dry of optimum. The thickness of a CCL is usually between
600 and 900mm, but occasionally the thickness may reach 1.2 to 3.0m. The most
common type of compacted clay liner is one that is constructed from naturally occurring
Literature Review 15
soils that contain a significant quantity of clay (i.e., soils that are classified as CL, CH,
or SC in the Unified Soil Classification System (USCS) outlined in ASTM D2487).
Some typically acceptable zones (as shown in Figure 2-3) for selecting initial dry
density and moisture content were defined by Daniel and Wu (1993) based on low
hydraulic conductivity, low desiccation induced shrinkage and high unconfined
compressive strength.
20
Shrinkage
Strength
Hydraulic conductivity
Overall
Moisture content (%)
Dry
uni
t wei
ght (
kN/m
3 )
221816141210815
16
17
18
19
20
Figure 2-3 Acceptable zone based on design objectives for hydraulic conductivity, volumetric shrinkage, and unconfined compressive strength (Daniel and Wu 1993)
The cover layer of a landfill can be easily affected by many different other conditions
than the bottom layer. Therefore extra research attention has been placed on the landfill
cover systems, e.g., Benson (2001), Dwyer (1998). This has led to development of
various cover systems including simple soil cover, compacted clay covers, geosynthetic
16
clay covers, monolithic covers and capillary barriers (Dwyer, 1998; Albright et al.,
2004).
Compacted Clay Cover
Geosynthetic Clay Cover
Monolithic Cover
Capillary Barrier
Finetextured
soil
Finetextured
soil
Finetextured
soilFine
textured soil
Finetextured
soil
WasteWaste
Waste
Waste
WasteCompacted
Clay
GCL
CoarseSoil
Simple soil Cover
Figure 2-4 Different types of cover layers (Dwyer, 1998; Albright et al., 2004)
The prescriptive or traditional cover designs are usually based on the existence of layers
of natural fine grained soils depending on the most type of waste (i.e. municipal or
hazardous). The percolation control in these barriers is typically achieved by
constructing a compacted clay liner with low saturated hydraulic conductivity, thereby
minimizing the infiltration through the layer and also by maximizing the overland flow.
As the knowledge of cover layers grew, a strong emphasis has been placed on the
possibilities and advantages of using alternative sealing materials and cover designs.
The main objective here was to minimise the costs for cover layer systems while
Literature Review 17
increasing the effectiveness. Thereby, the requirements on the soils used and the design
of the cover layer become more and more complex and the layer thickness can increase
as much as 3m. In addition, synthetic materials such as geo-membranes and geo-
synthetic clay liners have also been incorporated into the design process.
Alternative designs mostly rely on water storage principles (i.e., controlling percolation
by water storage during wet periods and evapotranspiration during dry periods) and are
often referred to as "evapotranspirative covers". A capillary barrier layer (i.e., a layer of
fine-grained soil over a layer of coarse soil) is sometimes added to increase the water
storage capacity of the cover. However when considering the different types of designs,
some major and most popular alternative covers can be identified as designs composed
of capillary barriers, evapotranspiration layers and composite layers (Melchior, 1997;
Dwyer, 1998; Albright et al., 2004). A capillary barrier is a cover employing a finer
grained layer overlying a coarser-grained layer. This contrast in particle size limits
downward migration of water by exploiting the contrasting unsaturated hydraulic
properties of soils with different gradation. An evapotranspiration layer acts not as a
barrier, but as a sponge that stores moisture during precipitation events, and then
releases it back to the atmosphere as evapotranspiration. Composite barriers consist of
mineral layers (CCLs) or geo-synthetic clay liners (GCLs), placed in close contact with
a geo-membrane (GM).
Capillary barriers are practical for semiarid and arid regions because there is no need for
moisture conditioning, which reduces construction costs. In addition, because capillary
barriers do not have a moist, compacted clay layer, they are less susceptible to
degradation caused by desiccation cracking. Capillary barriers are constructed in
18
various forms, ranging from a simple design consisting of two layers to more complex
designs that include multiple layers of finer-grained and coarser-grained soils (Morris
and Stormont, 1997).
The hydraulic conductivity of soils typically used in evapotranspiration (ET) covers is
higher under saturated conditions than the hydraulic conductivity of typical clay barrier
materials. However, under unsaturated conditions, the hydraulic conductivity of these
soils is typically less than that of clays. ET covers have also been referred in the
technical literature as monocovers, monolithic, store-and-release and soil-plant covers.
They are usually vegetated with native plants that survive on the natural precipitation. In
addition, ET covers are less vulnerable than clay barriers to desiccation and cracking
during and after installation, are relatively simple to construct, require low maintenance,
and can be constructed with a reasonably broad range of soils (Zornberg et al., 2003).
Cote et al. in 1996 suggested a ‘self-sealing, self-healing liner’, capable of self-repairing
in situ when damaged, surrounding the waste with two layers of porous material (Kwon
and Cho, 2011). The two layers, each containing a sufficient amount of interactive
components, act as a waste-encompassing interface. Upon placement, the components
in the two layers form precipitates that fill the pores at the interface of the two layers,
and finally form a waste-encompassing layer with reduced permeability. The special
benefit of this method is that the interactive components form a new seal if seal rupture
occurs, thus preventing loss of noxious substances through the rupture. However, the
precipitates from the self-recovery reaction of the layer might be dissolved in the acidic
rains with long term conditions.
Literature Review 19
To meet the economic constraints and to fulfil the technical requirements, it is important
to know the pros and cons of the technical performance of the components and capping
systems and the uncertainties in its assessment, especially in view of the importance of
their long-term behaviour.
2.2.3 Major issues related to clay liners
The basic purpose of a well designed and constructed clay liner is to prevent or
minimize the infiltration. It is doubtful whether this expectation is satisfied in long term
conditions, where the clay liner is in operation. The deficiencies of a clay liner can
occur due to inappropriate designs or construction faults. In addition:
• Cracking due to desiccation;
• Distortion due to differential settlement in the waste body; and
• Difficult placement over compressible waste
can also be identified as factors that can compromise their long term performance.
When a compacted clay soil is exposed to the atmosphere water evaporates, potentially
inducing shrinkage in all directions. However, only the vertical direction is free to
shrink and the lateral direction is normally restrained. As a result, the soil layer can
crack, which could lead to substantial increase in seepage rates. In this context, the
relative effect of the drying/shrinkage increases with the as compacted water content.
Consequently, shrinkage strains and therefore desiccation cracks increase with
compaction water content (Istok, 1989).
20
Clay liners are normally compacted with high water content (on the wet side of the
optimum moisture content) to achieve low hydraulic conductivity. However, a small
decrease in the water content can cause a large increase in soil water tension or soil
suction. When laterally restrained, water tension induces tensile stresses on the soil
matric, which eventually leads to cracking of the clay and lateral shrinkage opening the
cracks further. However, it is found that two processes counteract desiccation in a
bottom layer (Holzlohner et al., 1997). Firstly, capillary rise of liquid water from the
subgrade compensates for the water loss by water vapour transport and the water
content can equilibrate at a reasonably high value under appropriate site conditions (i.e.,
depending on the capillarity of the soils and conductivity of the subgrade and depth to
the water table). Secondly, the compressive stress caused by the high load of the waste
body can suppress the tensile stress generation and prevent the formation of cracks.
In a cover system the desiccation processes are driven by water evaporation into the
atmosphere and by the transpiration of the plants, when present. Under this condition
the CCL becomes attractive as a water reservoir for the plants thereby accelerating the
desiccation process. In cover systems, the compressive vertical stress caused by the
overburden on the clay liner is limited. It is, therefore, questionable whether stable
conditions can be achieved in a cover system in equilibrium with its environment
without significant changes in water content from the as compacted state occuring. At
least very thick restoration layers with high storage capacity for plant-available water
are necessary to achieve such conditions (Simon and Müller, 2004).
Cracking reduces the effective thickness of the low permeability earthworks layer,
thereby compromising the overall integrity of the barrier, and conditions that would
Literature Review 21
enable self-sealing of cracks cannot be entirely relied upon (Savadis & Mallwitz 1997).
During its lifetime a landfill clay liner may be subjected to different periods of wetting
and drying. The moisture loss can be more complicated due to heat generation within
landfills. As the landfill acts as a bio-reactor, heat is generated and moisture loss can be
due to a combination of moisture migration away from the heat source and, if linked air
voids are available, evaporative losses at the interface between the geo-membrane and
the mineral barrier can take place. Doll (1997) has modelled moisture migration away
from such a heat-source in landfill liners and found that significant drying can occur
unless downward vapour diffusion due to temperature gradients can be balanced by
capillary moisture rise from underlying strata.
In addition, clay barriers may not maintain their initial as-compacted moisture content
but instead the moisture will redistribute over time until constant soil water potential
throughout the layered soil profile is attained and maintained. Compacted clay liners
placed on coarse grained soils (common practice due to installation of under drains and
leachate detection layers) will equilibrate matric potential with the soil beneath and
decrease in moisture content. Through monitoring a test pad, some researchers found
that incorporation of a drainage barrier or capillary layer beneath a compacted clay liner
causes significantly more moisture loss and associated increases in suctions in the
compacted clay liner, when compared with a clay liner underlain by a fine soil (Philip et
al., 2002).
When a landfill matures, the waste decays reducing its volume. This can cause
differential settlements of the landfill cover layer leading to differential settlements,
under which a liner can crack and fail. Similarly, when placing a layer on top of
22
compressible wastes, it may not be possible to compact to the required level and
consequently the required permeability may not be achieved. This may also lead to
undesired performance.
To overcome the above discussed issues and limitations of compacted clay layers and
understand how those problems could occur, there should be a good understanding of
field performances of compacted clay layers. The effects of these problems could be
understood by observing the field performance of clay liners. The next section describes
the literature which explains the behaviour of clay layers in field conditions.
2.2.4 Field performances of clay liners
The long-term performance of clay barrier covers is considered on two main factors:
(i) Proper construction of the barrier layer (the hydraulic conductivity requirement is
met at field scale); and
(ii) Long-term maintenance of the barrier layer (to maintain the low hydraulic
conductivity)
Factors contributing to proper construction of CCLs have been studied in detail and are
well understood (Benson et al., 1999; Kodikara, 2001). Far less attention has been given
to evaluating whether the integrity of clay barriers can be maintained under field
conditions. However, when the field performance of compacted clay liner is considered,
the moisture content variation within the layer and hydraulic conductivity and water
balance of the layer are recognized as the most important factors. Hence the
performance of the landfill liner after installation such as, evaporation, moisture content
Literature Review 23
variation, infiltration, drainage, suction and effect of wet/dry cycles has been widely
monitored (Singh and J., 2003). These field observations provide some evidence to
indicate that clay barrier covers may not be performing as intended (Corser et al., 1992).
2.2.4.1 Moisture content and suction variation within the soil layer
The variation in water content due to the evaporation mainly affects the soils close to
the ground surface. This depth changes with the soil type, location of water table and
the climate conditions. In a clay soil, plants cause significant drying only within the root
zone (Smethurst et al., 2006).
Figure 2-5 (after Kraynski, in (Chen, 1988)) shows the moisture content variation with
the depth of the soil layer. Hd and Hs represent the depth to permanent water table and
depth of seasonal changes in water content respectively. The curve number 1 is for
equilibrium water content profile expected below an ideal surface covering in which the
moisture content decreases smoothly from top on the basis of an equilibrium suction
profile. There is no gain or loss of moisture to the atmosphere. Curves number 2 and 3
show the moisture content variation with uncovered soil natural conditions. The typical
variation of water content with depth corresponding to dry-season and wet-season
conditions are shown as curve number 2 and 3 respectively. In the wet season the water
content increases rapidly near the surface of the soil. Similarly in the dry season, water
content near the surface decreases. Seasonal fluctuations are restricted to a depth Hd,
which has a potential to expand if water deficiency increases. However, this depth is
typically 1 to 2m beneath normal ground cover in Eastern Australia, about 0.5 to 1.0m
below the non-vegetated surface of Australian coal tailings, and about 3m in Manitoba,
24
Canada (Morris et al., 1992). Further, AS2870-2011 presented the depth of design soil
suction change for different locations around Australia which changes from 1.5m to
4.0m. However, the depths provided in the standard could be subject to seasonal
variations. The change in suction at the surface is considered to be the same value
throughout Australia with a value of 1.2pF. This value is used later in this study to
predict the suction profiles.
Hs
Hd
Water content
Dep
th
Curve 3Curve 2
Curve 1
Figure 2-5 Variation of moisture content with depth, Curve 1, equilibrium water content under a surface covering; curve 2, dry season; curve 3, wet season (after Kraynski, (Chen 1988))
Moisture content of the soil layer responds reasonably fast to the climatic conditions.
However, after several climatic condition cycles, the soil reaches an equilibrium
condition in laboratory conditions (Warkentin and Bozozuk, 1961). Warkentin and
Literature Review 25
Bozozuk have dried cylindrical samples of undisturbed Leda clay to three different
water contents and allowed it to swell and shrink in sub subsequent cycles. They stated
that if the drying and wetting cycles are repeated, the increment of water content
decreases in further cycles. The net loss after each cycle is small and could be due to
further orientation of particles with each drying cycle. After many cycles the volume
change becomes reversible. A typical plot showing the effect of drying and rewetting of
Leda clay by Warkentin and Bozozuk (1961) is presented in Figure 2-6. The figure
illustrates the gradual decrease of the saturated moisture content with the number of wet
dry cycles.
Initial point (water content 76.6%)
Dry
ing
curv
e
Wet
ting
curv
e
Number of wetting and drying cycles
Moi
stur
e co
nten
t (%
)
Figure 2-6 Effect of drying and rewetting on the ultimate moisture content of Leda clay (after warkentin (1961))
26
Theoretically, if climatic conditions remain unchanged long enough for the complete
equilibrium conditions and a constant water table to be reached in the soil profile, the
matric suction would decrease linearly with depth from maximum suction at the ground
surface to zero suction (atmospheric pressure) at the water table. However, this
condition is hardly ever reached in practice and generally a transient nonlinear decay of
matric suction from a value at the surface depending on the transient event (drying or
wetting) to zero at the water table is observed. The transient osmotic suctions that
depend on the salt concentration in the pore water are also considered in total suctions.
Within the depth of cracking, changes in matric suction occur much more rapidly than
changes in soil chemistry and solute suction. Hence, it is considered that matric suction
has a greater effect on cracking than solute suction within the usual timeframe of
engineering interest.
3
2 2 2
3 3 3
5 55 33
111
Suction, pF
Dep
th, m
W
W
WD
DD
Melbourne AdelaideSydney
W = Wet season Total suction D = Dry season Matrix suction
Figure 2-7 Typical seasonal soil suction variations with depth (Morris et al., 1992)
Literature Review 27
Figure 2-7 shows wet and dry season suction profiles with depth measured in southern
Australia (Morris et al., 1992). The unit of suction was given in pF which is defined as
log, (suction head in centimetres of water). It appears that the depth of shrinkage cracks
typically coincides approximately with the depth of seasonal changes in water content
and suction. The ground water table is at or is more likely to be below this depth. The
profiles of total suction (the sum of matric suction and osmotic suction) in Figure 2-7
have an approximately constant suction of about 4pF (about 1000kPa) at higher depths
mainly due to solute suction. The solute suction reaches a peak below the water table
due to the greater salt concentration produced there by downward leaching of salt from
above. It can reach 4.2pF or 4.6pF near trees (Morris et al., 1992). Suggested matric
suction profiles are shown as broken lines in Figure 2-7. They tend to be negative at the
water table, corresponding to zero suction on the logarithmic pF scale, and join the total
suction profiles towards the surface, where matric suction dominates over solute
suction.
Figure 2-8 Soil water characteristic curves (Fredlund and Anqing 1994)
(a) (b)
28
Under unsaturated conditions, the interactions between suction, water content and
hydraulic conductivity (Singh and Gupta, 2000) control the rate of water loss in the soil.
Typical plots of the relation between the soil suction and water content, known as soil-
water characteristic curves, for different soil types are shown in Figure 2-8 (a) and (b)
(Fredlund and Anqing, 1994). Figure 2-8 (a) shows the soil-water characteristic curve
for a silty soil, along with some of its key characteristics. The air-entry value of the soil
is the matric suction where air starts to enter the largest pores in the soil. The residual
water content is the water content where a large suction change is required to remove
additional water from the soil. The adsorption curve differs from the desorption curve as
a result of hysteresis of water flow. The saturated water content and the air-entry value,
generally increase with the plasticity of the soil. Other factors such as stress history also
affect the shape of the soil-water characteristic curves. The total suction corresponding
to zero water content appears to be essentially the same for all types of soils (Figure 2-8
(b)).
2.2.4.2 Water balance of the soil layer with the surrounding environment
There are three distinct moisture migration mechanisms operating over the lifetime of a
clayey barrier (Philip et al., 2002):
(1) Evaporative losses from the surface of clay layers;
(2) Equilibration of soil moisture potential between clay barrier and surrounding
soil; and
(3) Moisture migration down the thermal gradient produced by high temperatures
generated during decay of waste in the landfill.
Literature Review 29
For basal liners mechanism (1) can generally also be categorized as moisture losses that
occur during construction and pre-waste disposal (temporary works) whilst mechanism
(3) generally occurs post waste disposal and during bio-reaction (permanent works). For
basal liners mechanism (2) will take place over the entire lifetime of a landfill from
initial construction until after bio-reaction. For compacted clay final covers,
mechanisms (1) and (2) will be the dominant processes. Figure 2-9 shows a schematic
profile view of a landfill liner indicating water balance.
Top soil – percolation layer
Sand – lateral drainage layer
Clay – Barrier layer
Waste
Sand – Drainage layer
Clay – Barrier layerGeomembrane layer
Geomembrane layer
Precipitation Evapotranspiration
Runoff
Infiltration
Lateral drainage
Leachate collection
Cov
er li
ner s
yste
mB
otto
m li
ner
syst
em
Figure 2-9 Schematic Profile View of a Typical Hazardous Waste Landfill
30
According to the Blight (2003), for the zone of major seasonal wetting and drying at the
slope surface, the full water balance may be written as
�(𝑅 − 𝑅𝑂) − � 𝐸𝑇 + ∆𝑤 − 𝑅𝐸 [2-1]
where
𝑅 = rainfall,
𝑅𝑂 = runoff,
𝐸𝑇 = actual evapotranspiration,
∆𝑤 = change in stored water within the soil, and
𝑅𝐸 = net recharge from the surrounding soil.
The rainfall is relatively simple to measure, although it is very site specific, and long
records are available for many countries. Runoff is also site specific but measurable.
The upward flow of water resulting from the suction gradient into the drying (root) zone
at the slope surface is negligible in comparison with rainfall and evapotranspiration, and
may be ignored in water balance calculations (Smethurst et al., 2006).
2.2.4.3 Evaporative losses from the surface of clay layers
In the absence of artificial recharge or irrigation, water may enter the soil through the
ground surface from rainfall and leave again through the surface due to evaporation and
as a result of evapotranspiration by plants. Predicting this exchange of moisture between
the soil surface and the atmosphere is a critical issue in the design of soil covers for
acid-generating mine tailings and waste rock, and other land-based disposal systems,
Literature Review 31
since the rate of evaporation depends on both soil layer properties and environmental
conditions. Three key soil layer properties influence this exchange:
1. The supply of water and evaporative demand;
2. The ability of soil to store and transmit water; and
3. The influence of vegetation absorption through roots.
The main environmental conditions govern the evaporation (Yanful et al., 2003)
namely:
1. Air temperature;
2. Relative humidity;
3. Net radiation;
4. Wind speed; and
5. Surface cover, such as turf.
Engineers have traditionally used a term defined as potential evaporation, PE, to
estimate evaporation or evapotranspiration rates. Potential evaporation may be defined
as the upper limit or maximum rate of evaporation from a pure water surface under
given climatic conditions. The potential rate of evaporation may be computed using the
Dalton type equation (Gray, 1970);
𝑅𝐸𝑇 = 𝑓(𝑢)(𝑒𝑠 − 𝑒𝑎) [2-2]
where:
𝑅𝐸𝑇 = rate of evaporation (mm/day),
𝑒𝑠 = saturation vapour pressure of water at the temperature of the surface (mm Hg or
kPa),
32
𝑒𝑎 = vapour pressure of the air in the atmosphere above the water surface (mm Hg or
kPa),
𝑓(𝑢) = turbulent exchange function that depends on the mixing characteristics of the air
above the evaporating surface.
The use of the apparently simple expression given in equation [2-2] is considered a
direct approach (Granger, 1989). However, the application of equation [2-2] to field
problems can be difficult. Accurate evaluation of the turbulent exchange function
requires either an empirical approach or the application of rigorous aerodynamic profile
methods. Furthermore, equation [2-2] is often indeterminate when applied to field
studies because of difficulties associated with the evaluation of surface temperatures
and vapour pressure (Granger, 1989). The Dalton equation is generally not applied in
the elementary form stated in equation [2-2]. However, equation [2-2] forms the basis
for the widely used Penman method. Penman (1948) resolved the difficulty associated
with surface temperature in equation [2-2] by combining it with a second simultaneous
equation for the sensible heat flux at the surface. Penman (1948) also provides a
relatively simple method for determining the turbulent exchange function on the basis of
mean wind speed. The Penman method assumes the surface to be saturated at all times
and therefore provides an estimate of the potential rate of evaporation.
Numerous other methods are also available for calculating the rate of potential
evaporation. These include the temperature-based method proposed by Thornthwaite
(1948) and the energy-based method developed by Priestley and Taylor (1972).
Although all of the various methods for potential evaporation may predict different rates
of evaporation when applied to a specific site (Granger, 1989) the fundamental
Literature Review 33
assumption used by the methods outlined above is that water is freely available at the
surface for evaporation. In other words, the surface is an open water surface or a
saturated soil surface. However, when we consider the actual condition it is not freely
available on the surface.
The actual rate of evaporation, AE, begins to decline as the surface becomes unsaturated
and the supply of water to the surface becomes limited (Gray, 1970). Figure 2-10 shows
a typical relationship for the ratio of AE and PE, AE/PE, with water availability for a
sand surface. Gray (1970) presents similar curves for sand and clay surfaces. The rate of
actual evaporation is approximately equal to the potential rate (i.e., AE/PE equal to
100%) when the sand is saturated or nearly saturated (i.e., water content at or above the
field capacity). The rate of AE/PE decreases as the sand surface becomes drier and
eventually falls to a low, relatively constant value as the sand surface desiccates to the
permanent wilting point for plants. Similar variation can be assumed for clay soils as
well.
The shape of the drying curve shown in Figure 2-10 is well known and has been
described by others including Hillel (1980). In general, the curve is described as having
three stages of drying. Stage I drying is the maximum or potential rate of drying that
occurs when the soil surface is at or near saturation and is determined by climatic
conditions. Stage II drying begins when the conductive properties of the soil no longer
permit a sufficient flow of water to the surface to maintain the maximum potential rate
of evaporation. The rate of evaporation continues to decline during Stage II drying as
the surface continues to desiccate and reaches a slow residual value defined as Stage III
drying. Hillel (1980) states that the slow rate of evaporation during Stage III drying
34
occurs after the soil surface become sufficiently desiccated to cause the liquid-water
phase to become discontinuous. The flow of liquid water to the surface ceases and water
molecules may only migrate to the surface through the process of vapour diffusion. In
summary, it can be seen that the rate of actual evaporation from a soil surface is
controlled by both climatic conditions, which defines the potential rate of evaporation,
and soil properties such as hydraulic conductivity and vapour diffusivity.
AE
/PE
(%)
Permanent wilting point
Field capacity
Water or moisture availability
Stage IIIStage II
Stage I
Typical evaporation curve for sand
0
50
100
Figure 2-10 The relationship between the rate of actual evaporation and potential evaporation (i.e., AE/PE) and water availability (Wilson, Fredlund et al. 1994)
Since the actual rate of evaporation is controlled by both climatic conditions and soil
properties, accurate prediction of the actual rate of evaporation from soil surfaces
requires a method of analysis that includes both factors. The methods previously
outlined (i.e., the use of equation [2-2] or the Penman method) are based on climatic
conditions such as temperature, relative humidity, wind speed, and net radiation. These
methods are reliable only for special conditions where the rate of evaporation is
controlled solely by climatic conditions (i.e., Stage I drying) in Figure 2-10. The
Literature Review 35
climate-based methods of analysis for potential evaporation often over-estimate actual
evaporation rates, since the actual rate of evaporation is soil limited. This is frequently
the case for unsaturated soil surfaces in arid and semi-arid environments (Wilson,
1990).
Geotechnical engineers are often required to predict evaporative fluxes under both
environmental and soil conditions. Wilson (1990) and Wilson et al. (1994; 1997) have
conducted research on coupling soil and atmosphere for soil evaporation. They
proposed a model that includes both atmospheric conditions and soil properties and
proved its applicability for sandy, silty and clayey soils. This model can be used to
modify the potential rate of evaporation obtained from Penman method to obtain the
actual evaporation for unsaturated conditions, known as modified Penman method.
2.2.4.4 Moisture migration downward the soil layer
Moisture migration downward or percolation through the soil layer worsens the
problems created by waste in a landfill. Therefore, as noted earlier, all the precautions
are to be taken for minimizing downward leakage while designing and construction of
landfill liners. However, preventing percolation is impossible using compacted soil
layers and geo-synthetics in the field. Hence percolation rates of landfill liners have
been monitored and measured in the field widely (Benson et al., 2001; Albright et al.,
2006; Adu-Wusu et al., 2007; Henken-Mellies and Schweizer, 2011).
36
Figure 2-11 Leachate generation rates at a modern domestic landfill in Pennsylvania (USA), Average annual precipitation at the landfill site is 1.0 m/year (after Bonaparte 1995, (Bouazza
and Van Impe 1998))
The leachate amount collected at the bottom of the waste container with the installation
of cover layers was recorded by Bonaparte (1995), who also plotted the LCRS flow rate
with time (Figure 2-11). In the absence of supplemental moisture addition, the leachate
generation rate will be highest early in the facility active life, with the rate decreasing as
the landfill is filled and progressively closed. These covers can virtually eliminate
infiltration and thus the long-term leachate generation provided that the cover layer
performs as expected in long term conditions similar to at the newly built conditions.
Several laboratory studies have shown that environmental conditions, especially those
that result in desiccation and freeze–thaw cycling, cause cracking of the soil and
increases in saturated hydraulic conductivity of two or more orders of magnitude
(Drumm et al., 1997; Albrecht and Benson, 2001; Albright et al., 2006; Henken-Mellies
and Schweizer, 2011). However in general, the amount of drainage discharge per year is
Literature Review 37
independent of the geometry of the landfill such as inclination of the cover (Melchior,
1997).
Henken-Mellies (2011) presented a field testing programme under site-specific
conditions for mineral landfill surface cover systems, of which the performance has
been monitored in a long-term study in Aurach, Bavaria (Germany). In that study they
have studied three cover systems; (1) Simple soil barrier cover, (2) Geo-synthetic clay
liner, and (3) Compacted clay liner. They have monitored and plotted the precipitation,
surface runoff drainage flow, the percolation through the layer and the soil water
content in large scale lysimeter test fields from 2002 to 2008 for a CCL. After 7 years of
observation there has been no deterioration of the mineral barrier recorded. They
suggested that the proper design of a clay liner can mitigate the high amounts of
percolation.
Albright et al. (2006) monitored the field scale hydrology of clay barrier covers at three
sites located in warm–humid, cool–humid, and warm–arid climates. In that research
they monitored precipitation, drainage, volumetric soil water content, and the surface
flow over two to four years. In their records, it is noticeable that for some time after the
layer installation the percolation rate was relatively constant (precipitation also
negligible) and showed little temporal response to precipitation events. In contrast,
when precipitation resumed after the low precipitation period, percolation was
transmitted within hours of precipitation events (Figure 2-12) and there was no good
relationship between the soil moisture content and the drainage rate. So, it was
concluded that cracks had developed and those cracks were providing preferential flow
paths causing corresponding increase of hydraulic conductivity.
38
It is unknown whether preferential flow paths in the soil barriers will close over the long
term. However, other studies suggest that these flow paths are likely to persist. Melchior
(1997) reports that preferential flow regularly occurred through desiccation cracks in the
clay barrier of a conventional final cover for 4 years, when the cracks formed during a
dry period. Hydraulic conductivity tests on desiccated clays conducted by Albrecht and
Benson (2001) over a period of approximately 1 year have also shown that preferential
flow paths in clays persist even with continuous access to water. Albrecht and Benson
(2001) describe observations of soil barriers in landfill covers in Minnesota and
Wisconsin, USA, where extensive cracking of the barrier occurred. Roots were present
in many of the cracks and moisture was present on the crack surfaces, suggesting that
preferential flow through the cracks had occurred for an extended period. No evidence
was found of infilling or other mechanisms that would impede or eliminate preferential
flow through the barrier over time.
Nevertheless, the drainage rate of a clay layer is not only dependent on the cracking and
hence preferential flow paths. According to the Hillel (1980) redistribution or internal
drainage in soils can occur due to gravitational and suction gradients within the deep
part of the soil profile and leads to downward water movement. Hence, proper design
and close monitoring on the performance after the cover installation is vital.
Literature Review 39
Figure 2-12 Daily precipitation and resulting drainage rate for periods (a) before and after (b) the fall 2000 drought at Albany, GA. Daily precipitation is shown as vertical bars, drainage
rate as a continuous line (Albright, Benson et al. 2006).
2.2.4.5 Swell-shrink behaviour
The seasonal changes make the soils wet and dry cyclically. These wet-dry cycles do
not only cause desiccation cracking but also significant damage to the engineered
40
structures due to shrinking and swelling. Furthermore, it may affect not only the
geometry of the layer but also the soil-water characteristic curves (SWCC) (Jayanth et
al., 2012) and hydraulic conductivity of the clay (Thakur et al., 2005). Several field and
laboratory research works have reported this behaviour of soil (Azam, 2007; Ito and
Azam, 2010; Yu and Wei, 2011).
Figure 2-13 Time-movement plots of several points inside and outside the National Art Gallery building, (Richards, Peter et al. 1983)
The effect of swell-shrink behaviour was monitored at Adelaide, Australia for several
years (Richards et al., 1983). The ground levels both near and away from the trees had
Literature Review 41
been monitored at regular intervals on a building and the adjacent roadway, as shown in
the time-movement plots in Figure 2-13. At the end of 1971, the trees were removed
and the effect of this removal is clearly shown in Figure 2-13.
Tripathy et al. (2002) examined the swelling and shrinking effects of compacted soil
through laboratory experiments. They found that the swelling and shrinkage path
becomes reversible in terms of water content and void ratio once the specimen reaches a
stable condition after a number of wet/dry cycles. Under this stable condition, the
vertical deformations during swelling and shrinkage are the same. This stabilized
condition generally occurred after about four swell–shrink cycles. The stable swell–
shrink path changes with the changes in the surcharge pressure and swell–shrink
pattern.
Figure 2-14 Variation of vertical deformation with several wet-dry cycles, (Tripathy, Subba Rao et al. 2002)
42
Hanafy (1991) proposed a characteristic S-shaped curve to describe the potential
volume change of an expansive clayey soil for the change in void ratio relative to the
changes in water content resulting from desiccation and water absorption (Figure 2-15)
during one complete swelling–shrinkage cycle. The main phases of soil shrinkage that
accompanied water withdrawal included normal, residual and structural shrinkage
phases (Haines, 1923; Kodikara et al., 1999). During structural shrinkage a few large,
stable pores are emptied and the decrease in volume of the soil is less than the volume
of water lost. Normal shrinkage occurs when the change in soil volume equals the water
lost. Residual shrinkage occurs when air enters the soil and the reduction in soil volume
is less than volume of water lost. At the no shrinkage phase, soil does not shrink upon
further drying. Similarly the swelling curve is also divided into three phases known as
primary swell, secondary swell and no swell (Day, 1999).
Figure 2-15 Shrinkage phases of clay upon drying (Kodikara et al., 1999)
Literature Review 43
2.3 Desiccation Cracking
Desiccation cracking due to matric suction has been widely observed and reported in the
literature. Kodikara et al.(2000) has presented theoretically plausible aerial patterns of
desiccation cracks in idealised media (Figure 2-16). These aerial patterns observed in
the laboratory and field are shown in the following figures. The parallel cracks can be
observed (Figure 2-17) when thin long layers desiccate (Nahlawi and Kodikara, 2006;
Costa et al., 2008). The cracks observed in real situations are showing mostly random
(Figure 2-18), orthogonal (Figure 2-19) and hexagonal (Figure 2-20) variation. The
cross sectional view of cracks was observed (Figure 2-21) by Yesiller et al. (2000) in a
laboratory experiment for a compacted clay layer which compacted in three lifts.
Figure 2-16 Theoretical patterns of desiccation cracks (a) parallel (b) square (c) hexagonal (after bezant and Cedolin, 1991, (Kodikara, Barbour et al. 2000))
Figure 2-17 Parallel cracks observed in long thin moulds.(Costa, Kodikara et al. 2008)
44
Figure 2-18 Irregular shaped crack pattern observed on a playa surface, southern Nevada, Hammer and handle measured 330mm (Longwell 1928). Playa surface means a flood plain clay
surface or dried-out lake surface.
Figure 2-19 Orthogonal Cracking pattern observed in coal mine tailings, Queensland, Australia (Morris, Graham et al. 1992)
Literature Review 45
Figure 2-20 Hexagonal mud crack pattern observed on playa surface, Las Vegas quadrangle, Nevada. (Longwell 1928)
Figure 2-21 Vertical crack pattern of a compacted clay layer with 3 lifts (Yesiller et al. 2000)
Measuring crack depth is an approximate method due to the irregular shape and
complex geometry (Yesiller et al., 2000). The depth, width and spacing of cracks are not
46
uniform throughout the area. Generally the highest depths and widths are recorded.
However the crack depths and spacing have been reported by many researchers. Table
2-1 summarizes some of the results available in literature.
Table 2-1 Summary of Desiccation crack depths recorded
Reference Soil Type Description of area
Depth of Cracking
Width of Crack
Spacing of Cracking
(Longwell, 1928)
Playa sediments
Desert of Southern Nevada
75-150mm
After (Simpson, 1936; Lau, 1987)
Clay In a dry season at Western Texas
6.1m
After (Jahn, 1950; Lau, 1987)
Silt Layer deposited under water
1.5-3.0m
(Horberg, 1951) Clay At Lake Agassiz in North Dakota
>3.0m 150mm
(Knechtel, 1952)
Playa sediments
Playa de los pinos, New Mexico
24.5-27.5m
(Willden and Mabey, 1961)
Playa sediments
On Black Rock and Smoke Creek deserts of Nevada
>1.22m 30.5-76.3m
After (Blight and Williams, 1971; Lau, 1987)
In South Africa 0.65-1.45m
(Zein el Abedine and Robinson, 1971)
Vertisol with more than 30% clay content
In Sudan 0.65-1.35m 0.28-0.51m
After (Dasog, 1986; Lau, 1987)
Highly plastic clay
In Saskatchewan 0.28-0.6m 9-22mm 0.7-3.45m
After (Mitchell, 1986; Lau,
Intermixed and inter
Dried out surface of a hydraulic fill
>1.5m 0.6-1.0m
Literature Review 47
1987) layered silty fine sand and moderately to highly plastic clayey silt
site in Los Angeles
(Miller and Mishra, 1989)
clay Landfill liner 0.3m (up to entire depth)
>10mm n/a
(Morris et al., 1992)
Adelaide (From Peter P.) 1.8-2.0m - -
New South Wales, West Moreton, Bowen basin coalfields in Queensland
Typically 0.5m
- -
Melbourne 2.0-3.0m - -
clay Western Australia, Mine tailings
4m 1m
Lake Agassiz clay
Winnipeg 6.0-8.0m - -
After (Montgomery and Parsons, 1989; Yesiller et al., 2000)
n/a Cover layer of a landfill in Wisconsin
1.0m 13mm n/a
After (Basnett and Brungard, 1992; Yesiller et al., 2000)
clay At slopes of a liner during construction
0.3m 13-25mm n/a
After (Corser et al., 1992; Yesiller et al., 2000)
n/a Compacted cover sections in an arid part of California
>0.1m n/a n/a
(Dyer et al., 2009)
clay flood embankments in the UK
0.6-1.0m 5-25mm 0.1m
48
Lachenbruch (1961) presented a theoretical analysis for tension crack formation
considering the boundary effects through a Linear Elastic Fracture Mechanics (LEFM)
approach. Corte & Higashi (1964) conducted notable field desiccation experiments in
glass and wooden boxes. They presented some important characteristics and
relationships between shrinkage cracking related parameters. Following that research,
many workers have analysed and investigated crack formation, development and mostly
by modelling.
The state of the art work on shrinkage cracking was reviewed by Kodikara et al. (2002)
and has identified most of the studies as qualitative and behavioural. However, a
significant amount of research effort has been dedicated to predict and relate the crack
depth and spacing and to provide a theoretical base for the mechanics of crack
formation.
2.3.1 Initiation and evaluation
As compacted clays lose water through evaporation, volume change/shrinkage occurs as
the gravimetric water content decreases from the initial value (Figure 2-22(a)). When
certain restraints are applied to the material during shrinkage, full potential free
shrinkage may not develop. Then the difference between the free shrinkage and the
actual shrinkage will contribute to the development of tensile stresses within the
material (Figure 2-22(b)).
Literature Review 49
Figure 2-22 Schematic illustration of cracking (Konrad and Ayad 1997a)
When the tensile stresses exceed the tensile strength of the material, shrinkage cracks
may initiate as shown in Figure 2-22(c) in a direction perpendicular to that of the
maximum tensile stress (Mitchell and Soga, 2005, Kodikara, 2006). In the early stages
of drying from a horizontal surface of initially saturated soil, decreases in water content
are largely accommodated by reorganization of the soil particles into successively closer
arrangements. This involves one-dimensional volumetric straining under k0 conditions,
since the loss of water is largely one dimensional towards the drying surface, where it is
dominantly removed by evaporation.
50
As the developed horizontal (or lateral) stress reaches the tensile strength of the material
cracks can initiate as shown in Figure 2-22c. As drying proceeds shrinkage continues
and dominant cracks grow deeper (Figure 2-22(d)). Cracks were found continuing to
propagate rapidly once they have initiated (Morris et al., 1992). The cracks spread to
join together at the surface as roughly a polygonal crack network, which then gets
stabilized with further desiccation (Philip et al., 2002). Once the crack network had
formed, cracks continued to propagate downwards and widen, although the rate of
widening reduced and this is thought to be related in part to residual shrinkage (i.e.,
water loss greater than volume change) (Philip et al., 2002). Although the orientation of
cracks is random, they all intersect at approach angles closely approximating 90°.(Philip
et al., 2002). Depth is ultimately constrained by increasing stresses due to self- weight
of soil or overburden and the crack length is limited by intersection with other cracks.
Soil will eventually stop changing volume with moisture content decreases when it
reaches a constant void ratio with reduction in moisture content (i.e., shrinkage limit)
during the zero shrinkage phase of the shrinkage characteristic curve.
After initial crack development, soil blocks can further subdivide by repeated
occurrences of the same process within already divided blocks. Initially, the tensile
stresses are highest at the middle of the longest side of an existing block of soil formed
by earlier cracking (Morris et al., 1992).
The tensile stress at the tip of any given crack is reduced by the presence of adjacent
cracks. The reduction in tensile stress at the tips of shorter cracks is also greater than
that at the tips of longer cracks. Consequently, the growth of smaller cracks is retarded
and eventually suppressed by the growth of adjacent larger cracks. Cracking thus
Literature Review 51
becomes concentrated at a series of relatively large cracks. In a roughly uniform
horizontal stress field beneath level ground, the large cracks are roughly uniformly
spaced in any given direction (Morris et al., 1992).
This subdivision stops when the decreasing size of the block in plan (and hence the
increasing relative depth of cracking) overcomes the restraining stresses at the base of
the block and matric suctions attain equilibrium with the atmosphere; that is, when the
maximum suction consistent with the ambient conditions is reached (Morris et al.,
1992).
2.3.2 Factors affecting desiccation cracking
The onset of cracking depends on the mineralogy of the soil, climatic conditions such as
temperature and rainfall and surface vegetation cover. At a selected strength level,
plastic clays contain more water than lean clays. They therefore experience larger
volumetric contractions on drying (Philip et al., 2002). They may also have relatively
large effective cohesions and tensile strengths (Baker, 1981). These properties can lead
to development of wider, deeper cracks in plastic clays than in lean clays.
Also, high temperatures alone do not produce wide, deep cracks if high temperatures
occur during wet seasons. Wide, deep cracking is associated with plastic soils and high
temperatures during dry seasons when the water table drops to considerable depth in the
soil profile. Furthermore, small changes in moisture content during post-compaction can
lead to large changes in volume and increased crack potential (Philip et al., 2002).
52
Landfill covers are relatively thin systems, placed over a relatively unstable and
consolidating mass. Their placement (at an angle) and the likely differential movement
at their base (Harison and Hardin, 1994), will also subject them to stresses, which
cannot be supported without cracking as the soil loses its plasticity once dried to below
the plastic limit.
2.3.3 Effects of cracking
A cracked soil has much higher hydraulic conductivity than the same soil at the same
water content in a no crack state. Thus any consolidation can be expected to proceed
much more quickly. However, this effect can be masked (or indeed reversed) by low
hydraulic conductivity in regions of desiccated soil between cracks (Blake et al., 1973)
and in the zone of shrinkage ahead of the crack tips. Hydraulic conductivity tests
performed by Albrecht and Benson (2001) indicated that shrinkage cracking of clays
increases their hydraulic conductivities by one to three orders of magnitude. In addition,
permanent changes in SWCC occur after an initial drying.(Sadek et al., 2007).
Other than that, cracks affect a soil's compressibility, its time rate of consolidation, its
strength, and the rate at which water can re-enter. Thus, much geotechnical construction
is affected directly or indirectly by the presence of cracks in a soil mass. A soil with
cracks is more compressible than an intact version of the same soil at the same water
content.
Literature Review 53
2.3.4 Theoretical developments for desiccation cracking
With the increased concern about desiccation cracking researchers tried to understand
and explain the phenomena using theoretical approaches. The crack depth, spacing,
initiation and evaluation were analysed in detail through several different theoretical
approaches. Basically it was believed that cracks occur when the tensile stress within
the block exceeds the tensile strength of the soil. Hence the behaviour of tensile strength
of a soil is also important to understand when dealing with desiccation cracking.
2.3.4.1 Tensile strength of soil
Tensile strength of a soil is a critical parameter when observing the stress related issues
in soil. It varies with several different parameters. However, it is not commonly
examined by researchers, perhaps due to the experimental difficulty in measuring it. The
problem with tensile strength is that values measured for a given material vary with the
type of test used and with the specimen size (Harison et al., 1994). Furthermore,
determination of tensile strength should be changed according to the condition of the
sample such as saturated and unsaturated or consolidated, unconsolidated and over
consolidated samples (Lakshmikantha, 2009).
The most common experimental methods of measuring tensile strength and the values
they obtained have been presented in Table 2-2 (Harison et al., 1994). V95 refers to the
volume of that part of the specimen subjected to at least 95% of maximum tensile stress.
The ratios of tensile strength from the ring test to the tensile strength from each of the
54
other tests are listed in the last column of Table 2-2. This analysis shows that the tensile
strength should not be treated as a material property.
Table 2-2 Variation of tensile strength with test method
Test method Specimen dimensions (mm) V95(mm3) Strength ratio
Ring test Ro/Ri = 8; R = 50 and t = 25 8.29 1
Three-point bending 200 × 50 × 25 ; 100 × 50 × 25 156 ; 78.1 1.78 ; 1.56
Four-point bending 200 × 50 × 25 2187 3.00
Brazilian test R = 50 and t = 25 10596 4.10
Direct-pull dog bone d = 25 and L = 35 17181 4.50
Note: Ro = Outside radius; Ri = Inside radius; t = Thickness; d = Diameter; and L = Length
Wang et al. (2007) has investigated the tensile strength of the clay by a uniaxial tension
loading assembly on cylindrical compacted specimens. The results have been presented
in graphs showing the influence of the dry density (Figure 2-23) and the influence of the
water content (Figure 2-24) on the tensile strength (𝜎𝑡) of clay. The results showed that
the values of 𝜎𝑡 increase with dry density which can be regarded as the result from the
change of compaction effort because a greater compaction effort is required to compact
a denser specimen. The tensile strength (𝜎𝑡) of the tested clay was decreasing with
increasing water contents. This behaviour can be regarded as the macroscopic exhibition
of the change of interaction force among soil particles and suction potential in the soil
columns induced by the change of water contents.
Figure 2-25 shows the experimental data for tensile strength variation with moisture
content and degree of saturation presented by Lakshmikantha (2009). Tensile strength
was determined using equipment designed by Rodriquez in 2002. The reduction in
Literature Review 55
tensile strength at lower degrees of saturation was explained as it may be attributed to
weak or broken capillary bonds. Finally he concluded that, the effect of density is more
pronounced for specimens with lower moisture content than those with higher moisture
content and the tensile strength does not increase continually with increasing suction,
but reaches a peak and then reduces with further increase in suction.
Figure 2-23 Influence of dry densities on 𝜎𝑡 of the clay with different water contents (Wang, Zhu et al. 2007)
56
Figure 2-24 Influence of water contents on 𝜎𝑡 of the clay with different dry densities (Wang, Zhu et al. 2007)
Figure 2-25 Details of tensile strength with constant natural density, (c) tensile strength with moisture content with polynomial fit showing the trend (d) tensile strength with degree of
saturation (Lakshmikantha, 2009)
Literature Review 57
2.3.4.2 Theoretical approaches describing crack depth and spacing
In soil mechanics, cracking can be described as soil failure. For that reason, different
theoretical assumptions can be used through different approaches. Elastic theory,
elastic-plastic behaviour and LEFM theory have widely been used in past research.
Apart from that, the energy balance approach and stress relief approach also have been
used. In early days, the crack depth was predicted using the Rankine theory of earth
pressure which was proposed by Taylor in 1948 (Bagge, 1985; Terzaghi et al., 1996).
For the drained condition,
𝑧𝑐 =2𝑐′𝛾
𝑡𝑎𝑛 (45 +𝜙′2
) [2-3]
and for the undrained condition,
𝑧𝑐 =2𝑐𝑢
𝛾 [2-4]
where,
𝑧𝑐 = depth of tension crack.
𝑐′ =effective cohesion intercept.
𝑐𝑢= undrained shear strength
𝜙′ = effective friction angle.
𝛾 = the unit weight of soils.
58
2.3.4.3 Linear elastic approach
Morris et al. (1992) have reviewed the behaviour of unsaturated soils and the mechanics
of cracking based on three different approaches. The linear elastic approach will be
discussed here. They suggested that the immediate surface of a soil deposit is dominated
by the matric suction (𝑢𝑎 − 𝑢𝑤) during desiccation and assumed that it drives the
shrinkage and cracking process during drying. The strains were assumed as one
dimensional just before the cracking and then the suction stresses decrease linearly from
the maximum at the surface to zero at the ground water table. The relation for the
suction profile was given by,
𝑆 = 𝑆0(1 −𝑧𝑐
𝑊) [2-5]
The tensile strength of the soil was selected as function of suction. Hence the relation to
obtain the tensile strength was considered as,
𝜎𝑡 = 𝛼𝑇𝑆𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′ [2-6]
Then they have presented a relationship to predict crack depths which is given by,
𝑧𝑐 =𝑆0
𝑆0𝑊 + 𝜈𝛾
1 − 2𝜈 − (1 − 𝜈)𝛼𝑇 𝑡𝑎𝑛 𝜙𝑏 𝑐𝑜𝑡 𝜙
[2-7]
Although, the relationship presented was simple and easy to understand, the stress
conditions are not one dimensional after a crack starts to open, especially close to the
tips of the opened crack. Hence the authors recommend further refinements of the
presented solution.
Literature Review 59
2.3.4.4 Elastic-Plastic approach
When considering the elastic-plastic behaviour, the most widely used failure criterion is
the Mohr-Coulomb theory as illustrated in Figure 2-26. The shear strength of a soil at a
point on a particular plane in terms of effective stresses can be expressed as a linear
function given by equation [2-8].
𝜏𝑓 = 𝑐′ + 𝜎′𝑡𝑎𝑛𝜙′ [2-8]
where,
𝜏𝑓 = shear strength.
𝑐′ = effective cohesion intercept.
𝜎′ = effective normal stress.
𝜙′ = effective friction angle.
The relationship for principle stresses and shear strength parameters can be obtained
from Figure 2-26 as,
(𝜎1′ − 𝜎3′) = 2𝑐′ 𝑐𝑜𝑠𝜙′ + (𝜎1′ + 𝜎3′)𝑠𝑖𝑛𝜙′ [2-9]
where,
𝜎1′ , 𝜎3′ = effective principal stresses
There are other failure criteria, such as Mohr-Paul, Griffith, Griffith-Brace and
Modified Mohr-Coulomb theories (Figure 2-27) which have been proposed to describe
the soil failure under tensile forces.
60
τ
τtFailure envelope
τf
σt
σ1'
σ3'
σ1'
σ1' σ'c'
ϕ′
σf'2θ
σ3' σ3'
σf'
θ
Figure 2-26 Mohr-Coulomb failure criterion
τ
σn
Mohr-Coulomb
Mohr-Paul
Griffith-Brace
Griffith
Figure 2-27 Strength envelopes indicating tensile strength after Lee and Ingles in 1968 (Morris, Graham et al. 1992)
Literature Review 61
A method to determine the maximum depth of tension cracks for both drained and
undrained conditions has been presented by Bagge (1985) . The analysis was based on
effective stresses and includes the effect of negative pore pressures (Bagge, 1985). The
relationship for the crack depth under undrained conditions is given by,
𝑧𝑐 =2𝐴𝑐𝑢 − 𝛾𝑤𝑊�𝑘𝑜 + 𝐴(1 − 𝑘𝑜)� − 𝛼𝑇𝑐′
�𝑘𝑜 + 𝐴(1 − 𝑘𝑜)�(𝛾 − 𝛾𝑤)
[2-10]
And for the drained conditions,
𝑧𝑐 =2𝑐′(𝑐𝑜𝑠𝜙′ − (1/2)𝛼𝑇(1 + 𝑠𝑖𝑛𝜙′)
1 − 𝑠𝑖𝑛𝜙′ ) − 𝛾𝑤𝑊
𝛾 − 𝛾𝑤
[2-11]
𝑧𝑐 = depth of tension crack.
𝑐′ =effective cohesion intercept.
𝑐𝑢= undrained shear strength
𝜙′ = effective friction angle.
𝛾, 𝛾𝑤 = the unit weight of soils and water
𝐴 = ratio of effective and total horizontal stress differences, (i.e.; Δσ3′/Δσ3)
𝑊 = depth to ground water table
𝑘𝑜 = earth pressure coefficient at rest
𝛼𝑇 = coefficient relating cohesion with shear strength at tensile failure
In 1992, Morris et al. has presented a similar solution for crack depth prediction
obtained directly from the Mohr-Coulomb diagram (Figure 2-26). Here failure was
assumed to take place in shear when the minor principal stress attempts to exceed the
62
tensile strength (Morris et al., 1992). The suction profile was assumed to be linearly
decreasing as given in equation [2-5]. The tensile strength was also considered to be
varying according to equation [2-6]. The derived relationship for crack depth was given
by,
𝑧𝑐 =𝑆0
𝑆0𝑊 + (1 − 𝑠𝑖𝑛𝜙′)𝛾
𝑡𝑎𝑛𝜙𝑏[2𝑐𝑜𝑠𝜙′ − 𝛼𝑇(1 + 𝑠𝑖𝑛𝜙′]𝑐𝑜𝑡𝜙′
[2-12]
Furthermore, they have derived another relation to obtain the crack depth by relating
cracking with a transition between tensile and shear failure. This is an extension to the
work (Bagge, 1985) into a more general stress system and for unsaturated soils more
generally. The relation was then presented as,
'cottan)23('sin3
tan'cos18)3(0
0
φφαφφφ
γ
bT
b
c
M
MWS
Sz
+−−
−+
=
[2-13]
where,
φφ
sin3cos6−
=M
Suction variation with depth and the tensile strength is given by equations [2-5] and
[2-6] respectively. Both equations [2-12] and [2-13] predict similar results for crack
depth as claimed by the authors.
Literature Review 63
2.3.4.5 Linear Elasticity Fracture Mechanics (LEFM) approach
Although fracture mechanics developed mainly in the 1960's, the basic concept of crack
propagation was established by Griffith in the 1920s. This theoretical approach has been
followed by many researchers to predict the crack depth.
In 1961, Lachenbruch analysed the stresses near the crack using the LEFM approach to
predict the tension crack depths. He presented several relationships to predict the stable
and unstable crack depths which are induced by different stress conditions such as
diurnal thermal stresses, surficial tensile stresses due to cooling, and deep tensile
stresses due to tectonic forces. Equation [2-14] represents the stable cracking depth due
to diurnal thermal activities of permafrost (Lachenbruch, 1961).
𝑧𝑐 ≈1
1.21𝜎2𝐸𝐺𝑐
𝜋(1 − 𝜈2) [2-14]
Where,
𝑧𝑐 = critical crack depth
𝐸 = Young’s modulus
𝐺𝑐 = the rate of release of strain energy with critical crack extension
𝜎 = constant tensile stress
𝜈 = Poisson’s ratio
After one crack has formed, the stresses around the opened crack area would relieve.
Hence no parallel crack can be initiated close to it in which the stress relief exceeded
10%, for the stress would be below the strength there (Lachenbruch, 1961). Also, the
64
second crack should initiate within the area where about 5% stress relief occurred.
Hence the author suggests calculating the spacing of the crack based on the above
predicted crack depths.
A LEFM solution for predicting crack depths has been presented in Morris et al. (1992).
That research presented the stress intensity factors corresponding to three modes of
stress distribution. The summation of three stress intensity factors equal to the critical
stress intensity factor, which can be obtained in a separate calculation,
𝐾𝐼𝐶2 =
−2𝜁𝐸1 − 𝜈2 [2-15]
This work have been extended and presented in more detail in a later publication by
Morris et al. (1994). They developed a graph showing the relation between the mode 1
stress intensity factor and the crack depth as shown in Figure 2-28. The curve intersects
the crack depth axis at 𝑧𝑐 = 0 and at,
𝑧𝑚𝑎𝑥 =1.6420𝑆0
𝑆0𝑊 + 𝜈𝛾
(1 − 2𝜈)
[2-16]
where,
𝑧𝑚𝑎𝑥 is the maximum possible crack depth for the steady state suction profile. The
critical crack depth (𝑧𝑐) values can be obtained by the intersections of the critical stress
intensity factor line and the mode 1 stress intensity factor curve. The smaller depth is
the pre-existing crack depth and the higher depth is the propagated final crack depth
(Morris et al., 1994). The authors concluded that LEFM over-predicts crack depths as
clay cracking requires significant dissipation of plastic energy.
Literature Review 65
Harison and Hardin (1994) have presented methods for analysing earth structures
subjected to tensile stresses that may produce cracking based on LEFM. Methods of
solution are proposed for desiccation cracking of a clay layer above the water table;
cracking of layered earth structures subjected to external loads; cracking of clay caps for
landfills due to differential settlement; and radial cracking in pressurized boreholes.
Figure 2-28 Variation of stress intensity factor with crack depth (Morris, Graham et al. 1994)
The crack depths were derived using a stress intensity factor which was determined
using stress analysis for a cracked half-space (Figure 2-29) by applying tractions to the
crack surface, where the magnitude of applied traction is determined by the stresses in
the uncracked body (Harison and Hardin, 1994). They presented several equations to
66
obtain the stress intensity factors. From those equations, the variation of stress intensity
factor with crack depth have been calculated and were illustrated in a figure similar to
the graph of Morris et al. shown in Figure 2-28. Here the variation of different surface
suction values was also shown. Although, the trend of the variation of results is similar
to Morris et al. (1994), the values obtained for crack depth are much less.
Konrad and Shen (1997) presented a simplified theoretical approach for the prediction
of the spacing between thermal cracks in asphalt pavements based on a finite element
formulation of linear elastic fracture mechanics applied to an idealized layered system
in which the properties of the frozen layers are temperature dependent. They have
presented equations by which one can determine the extent of stress relief caused by the
fictitious normal stress distributed on the crack walls over a depth equal to the thickness
of the asphalt layer. The size of the stress relief zone depends upon the crack depth and
the neighbouring crack will form at the edge of this stress relief zone. Spacing between
cracks then readily follows (Konrad and Shen, 1997). The calculated crack spacing was
found to be between 25 and 90m for asphalt base/subbases while the observed crack
spacings are 8 to 9m.
In another study, cracks were analysed by postulating that they occur successively (Sun
et al., 2009). Formulae for the secondary crack spacing were then derived after stress
analysis. The effect of the Poisson ratio on the crack spacing was analysed and a linear
relation between the Poisson’s ratio and crack spacing was obtained. The authors
concluded that the Poisson’s ratio of a soil is important in soil cracking studies.
Literature Review 67
Figure 2-29 Stress analysis for desiccation cracking (Harison and Hardin, 1994)
2.3.4.6 Other approaches
Two other approaches; namely, stress-relief and energy balance have been used to
predict the crack spacing and depth in past research. This section will summarize the
key findings from that research briefly.
68
(i) Stress relief approach
A simplified analytical model to explain desiccation cracking of clay layers in
laboratory cracking tests was presented by Kodikara and Choi (2006) based on a stress
relief approach. This model predicts the consecutively propagating cracks on the basis
of the maximum tensile stress that develops at the mid sections of the clay layer due to
the restraints provided at the basal interface. The highly non-linear process was
simulated by recursively applying the analytical model incorporating the variation of
material properties through correlations with the moisture content and following the
moisture reduction during desiccation experiments. The schematic diagram of the
analytical model presented here is shown in Figure 2-30.
The free shrinkage strain was observed correlating linearly to moisture content
reduction which can be identified as a new approach in this field. The relation to obtain
the maximum tensile stress at the middle of the layer for a fully elastic interface
condition was then given by,
𝜎𝑥,𝑚𝑎𝑥 = 𝐸𝛼𝛥𝑤 �1 −1
𝑐𝑜𝑠ℎ(𝛽𝐿 2⁄ )� [2-17]
where, 𝛼 is the Hydric constant and 𝛥𝑤 is the moisture reduction, giving 𝜀𝑠ℎ = 𝛼𝛥𝑤,
𝜎𝑥,𝑚𝑎𝑥 is the maximum tensile stress, 𝐸 is the Young’s modulus of the clay, 𝛽 =
�𝑘 𝐸𝐴𝑐⁄ , 𝐴𝑐 is the cross sectional area, 𝑘 (in units of pressure) is the interface shear
stiffness and 𝐿 is the length of the clay layer. They used empirical correlations to
represent non-linear material properties of clay and the basal interface.
Literature Review 69
Figure 2-30 Schematic representation of the proposed model by Kodikara and Choi (2006)
The approximate spacing to depth ratio of parallel cracks that form in long desiccating
soil layers subjected to uniform tensile stress was examined by Kodikara et al.(2011).
The formation of sequential crack patterns was examined theoretically on the basis of an
energy balanced and stress relief approach.
Kodikara and Choi (2006) have derived a relation for non-dimensional crack spacing as
shown in equation [2-18] to capture the stress release by a crack penetrating to the full
depth of an elastic clay layer. It is assumed that the crack occurs reasonably fast and
there is no change in tensile stress and tensile strength due to further drying, and that a
second crack would initiate when the horizontal tensile stress reaches 𝛼 (≤1) times the
tensile strength.
�𝑠𝑑
� = −�1
(1 − 𝜈)+
𝐸𝑘𝑑(1 − 𝜈2)
𝑙𝑛(1 − 𝛼) [2-18]
where, s is the spacing between cracks, 𝐸 is the elastic modulus, 𝜈 is the Poisson’s ratio,
𝑘 is the interface shear stiffness and d is the depth of the layer.
c
L/2 L/2
evaporation
tensile stressdistribution
basex
Clay layer
Interface
τf
uup
τ
stiffness, k
characteristics
70
Figure 2-31 Schematic diagrams for the (a) stress relief approach (b) energy balance approach (Costa 2009)
(ii) Energy balanced approach
The formation of crack simultaneously and sequentially was analysed under this energy
balanced approach and relations derived to predict the non dimensional strain energy
loss (Kodikara et al., 2011). In this method, it was postulated that the strain energy
released during crack formation was balanced by the energy consumed during the
formation of cracks. By using the non-dimensional energy balance at crack formation
the authors have derived a relation given by,
�𝑠𝑑
� = −1.82𝑙𝑛 (1 − (𝐾𝐼𝐶
𝜎𝑡)2 (1 − 𝜈2)
𝑑
[2-19]
where, 𝜎𝑡 is the tensile strength of the material, 𝐾𝐼𝐶 is the fracture toughness and the
other parameters are same as in equation [2-18].
Kodikara et al. (2011) have then compared both the stress-relief and energy balanced
approaches and presented solutions to predict non-dimensional energy loss and crack
Literature Review 71
spacing for the general case. The relation to predict non-dimensional crack spacing was
given by,
�𝑠𝑑
�𝑚𝑖𝑛
≈ −1.5�1
(1 − 𝜈)+
𝐸𝑘𝑑(1 − 𝜈2)
𝑙𝑛 (1 − (𝐾𝐼𝐶
𝜎𝑡)2 (1 − 𝜈2)
𝑑
[2-20]
The authors recommend further research in this area due to the large number of
simplifying assumptions made in the analysis.
2.4 Numerical Models
Demonstrating and predicting the effectiveness of a cover over the long term presents a
challenge, as there is usually not enough data available from long-term field monitoring
of soil covers. The laboratory-measured values are often not quite representative of field
conditions due the challenges associated with obtaining representative samples for
laboratory tests and the differences that exist in field and laboratory conditions. Hence
as a reasonable substitute, numerical models have been used widely in predicting long
term conditions. However, their ability to accurately simulate conditions existing in the
field must be verified before the predictions can be trusted.
Generally, the modelling approaches can be considered as numerical modelling,
analytical modelling and empirical modelling. Empirically based models may not be
applicable to general conditions outside where they were developed whereas theoretical
methods can be too restrictive to cover a wide range of conditions that exist in field.
Research undertaken using numerical modelling is presented in this section.
72
To improve the accuracy of computer predictions, it is important that the input data, for
example, soil properties and the data that describes the physical processes in the field,
be representative of field conditions. Soil properties may be derived by comparing
computer model simulations to experimental field results, and modifying uncertain soil
properties and physical conditions until the model reasonably simulates the field results.
This process, referred to as model calibration, allows for the determination of soil
properties that are more representative of field conditions (Adu-Wusu et al., 2007).
Model calibration also helps to understand the processes and identify important factors
that influence the moisture movement in soils. The calibration process helps to gain
insight into the internal workings and sensitivities of a particular model. The calibrated
model can then be used as a tool to aid in the design process, for example, to predict
long-term behaviour of a soil cover or to compare different design scenarios. The
numerical models discussed here are based on two main categories; (i) Water balance
modelling, (ii) Fracture modelling.
2.4.1 Water balance modelling software
Existing models used for water balance modelling include: Groundwater Loading
Effects from Agricultural Management Systems: GLEAMS (Leonard et al., 1987),
Hydrologic Evaluation of Landfill Performance: HELP,(Schroeder et al., 1994),
UNSAT-H (Fayer, 2000), HYDRUS-1D (Simunek et al., 1998), Soil Water Infiltration
and Movement: SWIM (Krysanova et al., 1997), Soil Cover (Wilson G. W., 1994) and
Vadose/W (Krahn, 2004). Generally, these models simulate subsurface flow based on
either a simple water balance approach (GLEAMS and HELP) or more complex
Literature Review 73
numerical solutions of Richard’s equation (UNSAT-H, HYDRUS-1D, VS2DTI, SWIM,
SoilCover and Vadose/W) (Adu-Wusu et al., 2007).
The usefulness of most of these as design tools has been assessed by researchers. The
following is a summary of some of their research findings.
Table 2-3 Summary of research used numerical models
Author and Paper Title
Aim of the research
Liner/ Cover Details
Observations Model
1 Milind V. Khire, Craig H. Benson, and Peter J. Bosscher, 1999
Field data from a capillary barrier and Model predictions with UNSAT-H
To compare the field measured water balance data and estimated water balance data using UNSAT-H
Capillary barrier;
• 75 cm of clean, uniformly graded medium sand (SP) • 15 cm uncompacted, sparsely vegetated sandy silt(SM-ML) • Slope 37%
• Precipitation
• Climatic data • Run off • Soil water
content • Percolation
UNSAT-H and HELP
• Volumetric water content
• Soil water storage
• Evapotranspiration
2 Smethurst, J. A., Clarke, D. & Powrie, W. (2006)
Seasonal changes in pore water pressure in a grass-covered cut slope in London Clay
To quantify the hydrological environment and pore water pressures in a cut slope in London Clay
Cut Slope
• 0.4m Top soil
• 2.5m weathered London clay
• 20m London clay
• Lambeth Group Deposit
• Slope 160
• Water content Variation with Soil Suction along the depth when drying & wetting (1 Yr)
• Evapotranspiration
• Daily Rainfall
• Soil moisture deficit
• Volumetric Soil water content
• Volumetric water
CROPWAT
• Soil moisture deficit
74
content along the depth (lab & Field)
• Suction
• Pore water Pressure
3 Sadek, S.,Ghanimeh, S.,El-Fadel, M., 2007
Predicted performance of clay-barrier landfill covers in arid and semi-arid environments
To predict the impact of cracks in the clay barrier, and estimate their potential effect on the percolation rates through the cover system
• 15cm vegetative Soil
• 45cm Barrier Soil (Ks =10-
5 cm/s) • Prepared Sub
Grade
• Daily Precipitation (1 Yr)
• Relative Humidity
• Maximum and minimum Temperature
• Potential Evapotranspiration
• Volumetric Water Content
Hydrus-2D numerical model
• Volumetric Water content along the depth
• Suction along the depth
• Cumulative Precipitation
• Cumulative Evapotranspiration (3 yrs)
• Cumulative Flux
• Daily Flux
• Monthly Flux
4 Yanful, E. K., Mousavi, S. M.
Yang, M., 2003
Modelling and measurement of evaporation in moisture-retaining soil covers
To model water flow through a single clayey till cover and a layered soil cover using a coupled liquid flow, vapour diffusion and heat transfer finite-element model (SoilCover)
Clayey till cover (25cm)
• Halton clayey till
• Evaporation
• Drainage • Total water flux • Volumetric water
content along the depth
SoilCover
• Volumetric water content along the depth
Layered soil covers
• 47mm Coarse Sand • 125mm Clayey Till • 78mm Fine Sand
• Relative humidity
• Temperature • Evaporation • Drainage • Total water flux • Volumetric water
content along the
SoilCover
• Evaporation
• Drainage
• Total water flux
Literature Review 75
depth • Volumetric water content along the depth
5 Adu-Wusu, C.,Yanful, E. K., Lanteigne, L., O'Kane, M., 2007
Prediction of the water balance of two soil cover systems
To Confirm the in situ soil properties of the various cover materials used for the test covers through model calibration using Vadose/W and to Determine the suitability of the model in describing existing field conditions and hence its effectiveness as a design tool
TP # 1
• 90cm Non compacted Layer
• GCL Barrier • Waste rock
platform • Slope 20%
• Daily minimum and maximum temperature (1 Yrs)
• Daily minimum and maximum relative humidity
• Daily wind speed
• Daily precipitation
• Soil porosity
• Volumetric specific heat capacity
• Thermal conductivity
• Coefficient of compressibility
• Saturated hydraulic conductivity
• Unsaturated hydraulic conductivity function
• Soil water characteristic curves
Vadose/W 2D model
• Potential evaporation
• Temperature • Suction • Soil water
storage • Percolation • Runoff
TP # 3
• 90cm Non compacted Layer
• 60cm Sandy silt Barrier
• Waste rock platform
• Slope 20%
76
Dwyer (2003) compared water balance simulations of six test covers applying the
computer models HELP and UNSAT-H to field data to verify the accuracy of the
models in describing field conditions. He reported that neither model predicted runoff
and percolation with reasonable accuracy. Furthermore, HELP does not contain
algorithms to simulate unsaturated flow rigorously, which can result in gross errors
when simulating capillary breaks (Khire et al., 1999). Roesler and Benson in 2002
(Adu-Wusu et al., 2007) reported that UNSAT-H could not simulate field data with
reasonable accuracy even after input parameters were modified similarly with HELP.
Table 2-4 Comparison of several programs
Software Water flow theory Evaporation modeling Lateral drainage
Vadose W Caters for unsaturated water flow by SWCC and hydraulic conductivity function
Flux boundary by coupling ground heat mass and vapor flow
Run off allowed
Unsat H liquid water flow using the Richards equation, water vapor diffusion using Fick’s law, and sensible heat flow using the Fourier equation using SWCC and hydraulic conductivity function
two methods are used, isothermal mode and thermal mode. In isothermal mode the potential evapotranspiration concept uses by solving Penman equation. In thermal mode evaporation calculates as a function of the vapour density difference between the soil and the reference height
No lateral drainage
Help assumes Darcian flow by gravity influences through homogeneous layers
Penman method, incorporating wind and humidity effects as well as long wave radiation losses
lateral drainage allowed
SoilCover is a one-dimensional finite-element package that models transient liquid and
water vapour flow, based on a theoretical model for predicting the rate of evaporation
Literature Review 77
from soil surfaces presented by Wilson et al. (1994) on the basis of a system of
equations for coupled heat and mass transfer in soil (Yanful et al., 2003). Simulations of
water balance using SoilCover were compared favourably with measured data from soil
column experiments in the laboratory (Yanful et al., 2003). The model’s ability to
predict field response data was also validated by Shuniark in 2003 (Adu-Wusu et al.,
2007), who reported that the model reasonably simulated measured data from four test
covers in the field. The program SoilCover has been subsequently replaced by
Vadose/w program.
2.4.2 Fracture modelling software
A rational highly idealized framework for the prediction of the spacing between primary
shrinkage cracks in cohesive soils undergoing desiccation was presented by Konrad and
Ayad (Konrad and Ayad, 1997b). The proposed framework was based on the theory of
LEFM, which is used to describe, in a simple manner, the phenomenon of crack
propagation. The principle of effective stresses was used to describe stress partitioning
in soils, and a fictitious stress superposition concept is used to predict the average
spacing between primary cracks. Crack propagation was analysed with a trapezoidal
distribution of total horizontal tensile stress as derived from the material constitutive
equations. A model named CRACK (Figure 2-32) was proposed that can be applied to
soft soils, consolidated natural soils, and compacted clays to predict crack spacing.
The model CRACK appears to possess most of the essential features of the analysis of
shrinkage crack development in clayey soils. It requires the knowledge of several soil
properties, which are relatively difficult to determine in the laboratory. The computed
78
results, however, are sensitive to input parameters. Nevertheless, the model CRACK
was shown to produce reasonable values at crack initiation and time of crack initiation
when compared with the field observations (Ayad et al., 1997).
Figure 2-32 Flow chart of the proposed model by Konrad and Ayad (1997(a))
Literature Review 79
Kodikara et al. (2004) modelled the curling deformation of desiccating clay using the
FLAC computer program. They observed that, the computer program simulated the
observed curling behaviour reasonably well despite the simplified assumptions made
during the analysis.
Inci (2008) used Finite Element Modelling (FEM) to simulate the strain response of soil
due to the loss of moisture through thermal shrinkage. He has observed that FEM
simulates the strains and the displacements correctly. However, the author recommends
including crack potential determination for alternative water content profiles and
various crack depths in future work. Experimental validation was also expected to be
done.
Amarasiri et al. (2010b) analysed long thin layers of slurry clay contained in moulds of
different dimensions using the computer program UDEC duplicating laboratory
experiments carried out on similar specimens. The computer program was successful in
modelling fracture which can be identified in the form of number of cracks, residual soil
height, moisture content at crack initiation and time progression of cracks etc. The
authors recommend the programme for desiccation crack modelling stating that
extension of the methodology could potentially be of value in modelling desiccation
cracking of field geo-engineering applications and other instances where time varying
material properties are prevalent.
Further studies using UDEC model have been undertaken as described in a number of
publications (Amarasiri et al., 2010a; Amarasiri and Kodikara, 2011b; Amarasiri et al.,
2011) The modelling approach proposed by them appeared to capture the essential
80
behaviour of the desiccation cracking with varying behaviour of soil properties and
deformability of soil.
The studies mentioned above emphasise the importance of validating computer codes
and suggest the need for improvement to existing codes or development of new ones.
Inability to reproduce field data may not be due only to suspect or inaccurate computer
programs but also to the accuracy of the parameters used as input. In some of the studies
mentioned above, better results were obtained when the input parameters were
modified. However, modelling of unsaturated soils involves many parameters and
factors that could easily influence the outcome. Given the influence of soil properties
such as the SWCC and unsaturated hydraulic conductivity-suction functions on model
output, it is necessary that variability in soil properties measured in the laboratory and
field be accounted for during modelling.
2.5 Conclusions
From the review of the literature related to desiccation cracking, several conclusions
were drawn as given below.
Waste containment systems have used different types of surrounding layers to minimize
pollution. The selection of layer design mainly depends on the regulations of the area,
waste type, geology, climatic conditions and the budget allocated. Since these
conditions are highly variable from one place to another there is no any particular type
of liner design recommended for waste containment systems. Furthermore, there are
Literature Review 81
still further developments to be made on more effective designs such as liners with self-
healing ability.
Almost all liner designs consist of a compacted clay layer. The main reason for using a
compacted clay layer is to minimize infiltration pollutants and water. However, given
the long term conditions it is impossible to minimise the infiltration due to preferential
flow paths which are developed with desiccation. Long term monitoring of the
performance of liners is an expensive and difficult task. However, careful monitoring
for long term stability after a reasonable period following liner installation is yet to be
observed in order to understand the liner behaviour with the maturity of landfill.
Desiccation is one of the severe problems a clay liner can face. Cracks develop in clay
liners when they desiccate. Much research needs to carried out in the form laboratory
and field experiments, theoretical models and numerical models in order to understand
the mechanism of cracking. Factors that affect desiccation cracking have been
understood qualitatively. However, the precise mechanism of cracking is imperfectly
understood and the depth and spacing of the cracks are not predicted accurately by any
of the proposed models.
When observing the cracking behaviour in the field, it has been observed that the crack
initiation and propagation highly depend on the cyclic change of the climate.
Desiccation crack modelling related to the moisture content variation during wet-dry
cycles seems to be important, but has not been undertaken in the past. Therefore, this is
considered as a subject of the current study.
Of the various computer programs used to predict the crack depths in the literature,
most are not sufficiently capable to capture the actual behaviour taking into account the
82
likely variation in moisture contents, material properties and boundary conditions.
However, the modelling approach used in UDEC appears to be able to reproduce the
desiccation fracturing process. This program will be further considered in the research
undertaken in the current study.
Ground moisture modelling has been also undertaken by several approaches using a
number of computer programs. Among them, Vadose/W appears to be able to
accurately simulate trends in the field response and more representative values of soil
properties.
Chapter 3
COMPARISON OF CRACK MODELLING APPROACHES
3.1 Introduction
The previous chapter presented a review of desiccation cracking and associated
mechanisms. This chapter presents results from some analytical models and numerical
models capable of producing crack depths in clay soils. The numerical modelling
approach for predicting desiccation crack depths of clay layers is presented using the
computer program, UDEC. Using this model, non-dimensional results are produced to
estimate crack depths. Numerical results are compared with the results obtained from
the analytical model of Morris et al. (1992), which has been referred to extensively by
researchers in the geo-environmental field.
83
84
3.2 Existing Analytical Approaches for Predicting Crack Depth
In the previous chapter, it was highlighted that the suction profile that develops in the
soil is very important in crack development. In this regard, three suction profiles as
given below are considered for the analysis as suggested by Morris et al. (1992).
i. Constant suction
ii. Linearly decreasing suction
iii. Parabolic suction variation
Three methods are used to estimate the crack depths, namely linear elastic, elastic with
shear failure and LEFM approaches.
3.2.1 Constant suction profile
Constant suction profile is an idealised condition which can happen in clay layers when
the soil dries uniformly down the depth. Usually, slow drying conditions can lead to this
development. The schematic view of the profile is shown in Figure 3-1 (a). Using
Equation [2-6], the tensile strength of the soil can be expected to be uniform in the
ground as shown in Figure 3-1(b).
Under elastic conditions, the relation between the components of stress and the
components of the strain can be described according to the Hooke’s law. In the
following derivations, stresses were analysed to obtain crack depths from the basic
theoretical assumptions. The x-direction is considered as horizontal and perpendicular
to the crack, the y-direction is horizontal and parallel to the crack, and the z-direction is
Comparison of Crack Modelling Approaches 85
vertical (down towards the earth). Tensile stresses and strains are considered to be
negative.
Dep
th
Suction
W
S0
z
Dep
th
Tensile strength
W
σt
z z
Figure 3-1 Suction profile (a) and tensile strength profile (b) with depth
The plane strain condition is also assumed meaning the crack is infinitely long and the
problem could be analysed in 2 dimensions. Hence assuming plane strain conditions,
strains in the x and y directions can be considered as zero (z direction = vertically
down). When the strain in the y direction is 𝜀𝑦 and the stresses in the x, y and z
directions are 𝜎𝑥, 𝜎𝑦 and 𝜎𝑧 respectively, the relationships for strain using applied stress
parameters in x, y and z directions can be written as,
𝜀𝑦 = 0 =1𝐸
�𝜎𝑦 − 𝜈(𝜎𝑥 + 𝜎𝑧)�
∴ 𝜎𝑦 = 𝜈(𝜎𝑥 + 𝜎𝑧) [3-1]
and,
𝜀𝑥 = 0 =1𝐸
�𝜎𝑥 − 𝜈�𝜎𝑦 + 𝜎𝑧��
86
∴ 𝜎𝑥 = 𝜈�𝜎𝑦 + 𝜎𝑧� [3-2]
where 𝐸 is a compression modulus for changes in total stress and 𝜈 is Poisson's ratio.
By substituting the value of 𝜎𝑦 from equation [3-1] in eq. [3-2], it is possible to obtain a
relationship between stresses in x and y directions.
𝜎𝑥 = 𝜈2(𝜎𝑥 + 𝜎𝑧) + 𝜈𝜎𝑧
Hence,
𝜎𝑥 =𝜈
1 − 𝜈𝜎𝑧 [3-3]
In natural ground conditions, however, matric suction also takes a part of the stress and
strain values. When considering the effect of matric suctions as the potential strain, the
effective strain in y direction can be written as
𝜀𝑦 =1𝐸
�𝜎𝑦 − 𝜈(𝜎𝑥 + 𝜎𝑧)� − 𝛼𝛥𝑤
where, 𝛼Δ𝑤 = 𝑆 𝐻⁄ . 𝑆 is the matric suction and 𝐻 is the modulus with respect to matric
suction. It should be noted that Morris et al. (1992) used directly 𝑆 𝐻⁄ as the strain
component without using the change in moisture content. As noted previously, for plane
strain condition, strains in the x and y directions are equal to zero,
𝜀𝑦 = 0 =1𝐸
�𝜎𝑦 − 𝜈(𝜎𝑥 + 𝜎𝑧)� − 𝛼𝛥𝑤
𝜎𝑦 = 𝜈(𝜎𝑥 + 𝜎𝑧) + 𝐸(𝛼𝛥𝑤) [3-4]
and
𝜀𝑥 = 0 =1𝐸
�𝜎𝑥 − 𝜈�𝜎𝑦 + 𝜎𝑧�� − 𝛼Δ𝑤
Comparison of Crack Modelling Approaches 87
𝜎𝑥 = 𝜈�𝜎𝑦 + 𝜎𝑧� + 𝐸(𝛼Δ𝑤) [3-5]
Using equation [3-4] and [3-5]
𝜎𝑥 = �𝜈
1 − 𝜈� 𝜎𝑧 + 𝐸
(1 + 𝜈)(1 − 𝜈2) 𝛼Δ𝑤
or in terms of suction, S,
𝜎𝑥 = �𝜈
1 − 𝜈� 𝜎𝑧 +
𝐸𝐻
1(1 − 𝜈) 𝑆 [3-6]
where 𝜎𝑧 = 𝛾𝑧, 𝛾 is the unit weight of the soil, 𝛼 is the hydric constant and 𝛥𝑤 is the
moisture content change. A similar relationship was derived by Morris et al. (1992)
using elastic theory.
𝑧𝑐 =(1 − 2𝜈)
𝜈𝛾𝑆0 +
(1 − 𝜈)𝜈𝛾
𝜎𝑡 [3-7]
The results obtained from equation [3-7] are presented in Figure 3-2.
Figure 3-2 Depths of cracking with different Poisson’s ratios tensile strength values when the constant suction profile is assumed throughout the depth
0
1
2
3
4
5
6
7
8
0 20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Suction (kpa)
v=0.3
v=0.35
v=0.4
t = -10kPa
t = -20kPa σt
σt
88
3.2.2 Linearly decreasing suction profile
In this section, the boundary conditions of the clay layer are assumed to be a constant
atmosphere flux boundary at the top and steady state conditions of moisture equilibrium
with the bottom boundary as the water table. The resulting suction profile would be a
linearly decreasing profile with depth as shown in Figure 3-3 (Morris et al., 1992).
Dep
th
Suction
W
S0S
zD
epth
Tensile strength
W
σt(0)σt
Figure 3-3 Suction and tensile strength profiles when linearly decreasing with depth
The values of suction in various depths were calculated using equation [2-5] and the
tensile strength was calculated similarly to the previous section using equation [2-6].
The crack depth corresponding to this type of suction profile through linear elastic
analysis is given by equation [2-7]. When allowed for shear failure in elastic approach,
the relationship becomes the form in equation [2-13]. The results from those equations
are presented in Figure 3-4. It can be seen that the shear failure criterion for fracture
Comparison of Crack Modelling Approaches 89
gives higher crack depths than those obtained using the linear elastic method with
tensile strength failure.
Figure 3-4 Predicted crack depth values for linearly decreasing suction profile through linear elastic (LE) approach and allowing for shear failure (SF) approach with different surface
suction values
3.2.3 Parabolic suction variation
The real field suction profiles are neither linearly decreasing nor parabolic exactly. The
actual variations were given in Figure 2-5 for moisture content and in Figure 2-7 for
suctions. However, in many cases, a parabolic variation can be considered to be a better
approximation for the field suction profile. The variation of suction with depth for this
2 4 6 8 10 0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
Depth to water table (m)
Dep
th o
f cra
ckin
g (m
) LE-Suction=100kPa LE-suction=50kPa LE-Suction=25kPa SF-Suction=100kPa SF-Suction=50kPa SF-Suction=25kPa
90
variation is given by equation [3-8]. The tensile strength has the same relation with
suction as used in previous sections.
𝑆 = 𝑆0 �1 −𝑧2
𝑊2� [3-8]
D
epth
Suction
W
S0S
z
Dep
th
Tensile strength
W
σt(0)σt
z
Figure 3-5 Parabolic suction and tensile strength profiles
Following the same analysis approach as used for constant and linear suction profiles,
the crack depth is calculated for a parabolic suction profile. The corresponding crack
depth for linear elastic approach is given in equation [3-9] while the shear failure
criterion gives equation [3-10].
𝑧𝑐 =𝑊2
2�−
𝐶1𝛾𝑆0𝐵
± ��𝐶1𝛾𝑆0𝐶2
�2
+4
𝑊2� [3-9]
where,
𝐶1 = 𝜈1−𝜈
and 𝐶2 = 1−2𝜈1−𝜈
− 0.5𝑡𝑎𝑛𝜙𝑏
Comparison of Crack Modelling Approaches 91
𝑧𝑐 =𝑊2
2�−
𝛾𝑆0𝐶3
± ��𝛾
𝑆0𝐶3�
2+
4𝑊2�
[3-10]
where,
𝐶3 =1 − 𝑠𝑖𝑛𝜙′
𝑡𝑎𝑛𝜙𝑏(2𝑐𝑜𝑠𝜙′ − 0.5𝑠𝑖𝑛𝜙′ − 0.5)
Figure 3-6 Predicted Crack depths when the water table is 4m below the surface (LE-linear elastic approach, SF-elastic allowing for shear failure and LEFM-linear elastic fracture
mechanics approach)
The crack depths obtained from the above equations are presented in Figure 3-6 and
Figure 3-7 below with different surface suction values and water table depths. Those
figures show the crack depths for parabolic suction variation within the soil mass in the
LE and SF approaches. However, the LEFM approach uses its own stress distribution
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 20 40 60 80 100
Crac
k D
epth
(m)
Surface suction (kPa)
LE SF LEFM
92
profile. The values of the LEFM approach have been calculated using equation [2.16]
which is discussed in section 2.3.4.5.
Figure 3-7 Predicted Crack depths using parabolic suction profile when the surface suction is 50kPa
These theoretical results will be compared and discussed with the numerically obtained
values later in this chapter.
3.3 Numerical Modelling Approach for Crack Depth Prediction
Numerical modelling is extensively used in understanding and predicting complex
natural phenomena due to its effectiveness. Hence, it is being used herein for modelling
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to water table (m)
LE SF LEFM
Comparison of Crack Modelling Approaches 93
desiccation cracking. As presented in section 2.4. The Universal Distinct Element Code
(UDEC) computer program is chosen for the analysis of desiccation cracks in a
hypothetical compacted clay liner based on the linear elasticity method and allowing for
shear failure in bulk media.
3.3.1 Overview of UDEC program
UDEC is a widely used numerical program in rock modelling. It simulates two
dimensional discontinuous media based on the distinct element method (Itasca, 2004).
The discontinuous media are represented as assemblages of discrete blocks. The
discontinuities or joints are treated as boundary conditions between blocks. Individual
blocks behave as either rigid or deformable material. Deformable blocks are subdivided
into a mesh of finite-difference elements, and each element responds according to a
prescribed linear or nonlinear stress-strain law. UDEC has several built-in material
behaviour models, for both the intact blocks and the discontinuities, which permit the
simulation of the response of discontinuous geologic or other similar materials.
The method of analysis in UDEC is based on the Newton’s equation. This equation of
motion is damped to reach a force equilibrium state under the applied initial and
boundary conditions. Two forms of velocity proportional damping are provided in
UDEC; first, adaptive global damping (referred as auto), second, local damping. Auto
damping is used to adjust the damping constant automatically using the viscous
damping forces. In local damping the damping force on a node is proportional to the
magnitude of the unbalanced force. UDEC also contains a built-in programming
language FISH. User defined functions can be written with FISH, to extend UDEC’s
usefulness. FISH offers a unique capability to UDEC users who wish to tailor analyses
94
to suit specific needs. Therefore this software has been identified as an attractive
program to model cracking.
The desiccation cracking is a three dimensional problem in the field, where cracks
would propagate down the depth and generally across the depth. However when one
crack is considered in a finite soil block, the 2D conditions are acceptable since any soil
block is reasonably uniform along the crack length which is the normal plane of the
plane of analysis. The size of the model and the modeling time can be greatly reduced
using the plane strain approximation. It should be remembered, however, that the 2D
approximation has limitations, especially in relation to prediction of cracking in the x-y
plane.
3.3.2 Numerical model implementation
This section is to reproduce the theoretical results shown in the section 3.2 numerically.
Hence, a compacted clay layer was analysed as shown in Figure 3-8. Since the intention
is to study the propagation of an isolated crack in plane strain, an interface (or joint) is
placed where the crack is intended to propagate. The boundary conditions are such that
the top surface is free of stresses, the edge boundaries are restrained against lateral
deformation and the bottom boundary is restrained against vertical deformations. The
groundwater level is placed at the bottom of the layer.
Comparison of Crack Modelling Approaches 95
Depth of clay liner
Water table
Possible Crack
Length of the clay liner
ZcZ
3.3.2.1 Problem geometry
The clay block was created in the UDEC program using co-ordinates. For crack
simulation in UDEC the blocks should be split into several subsidiary blocks, using pre-
decided joints which open up with the change of stresses representing crack formation.
The water table was considered to be at the bottom boundary of the soil block. This
condition was accomplished by using the suction stresses within the block. Hence, the
depth to water table was changed by changing the depth of the block. Since a finite
block is considered, the distance to the boundaries of the block from the crack
influences in the results considerably. The size of the zones within the deformable
blocks is also a considerable factor in obtaining precise results. Both sides and bottom
of the soil layer were supported by roller supports.
Figure 3-8 Cross section of the clay layer used for the model
96
Several test models were conducted with the same input parameters but changing the
size of the block to find the distance to the boundary from the crack to reduce boundary
effects. Some typical properties of compacted clay were used in the model setup. The
width of the clay layer was increased from 5m to 50m until the model gave consistent
results. This was done with two different block heights of 4 m and 10 m. Figure 3-9
gives the results obtained.
Figure 3-9 Variation of depths of cracking with layer width when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡=0kPa and 𝑆0=50kPa
Initially, with the increase of the length of the clay layer the resulting crack depth also
increased. Finally the crack depths became stabilized after a certain layer length. For
both layers with 4m and 10m heights, the consistent crack depth is almost 3 m and the
2
2.2
2.4
2.6
2.8
3
3.2
0 10 20 30 40 50
Dep
th o
f Cra
ckin
g (m
)
Length of the layer (m)
W = 4m
W = 10m
Comparison of Crack Modelling Approaches 97
layer length needs to be at least 30 m. According to this observation, a relationship was
deduced for the crack depth and layer length so that the layer length should be more
than 10 times the crack depth (i.e. Length ≥ 10 × 𝑧𝑐, where 𝑧𝑐 is the crack depth). In
other words, the boundary of the layer should be more than 5 times the crack depth
distant from the crack. Initially the crack depth can be estimated from a theoretical
method and according to that the length of the clay layer can be decided. The advantage
of selecting the correct size of the block without selecting larger sizes is minimizing the
run time of models.
Figure 3-10 Zones and blocks of the model
The effect of mesh size on the results was examined as the next step. A block of 4 m
height× 5 m (𝑊 =4 m) width and 10 m× 10 m (𝑊 =10 m) layers were selected to get
the optimum mesh size. The mesh consists of triangular zones with the maximum edge
length of 0.5m as shown in Figure 3-10. The models were run until the results did not
change as the mesh size was reduced further. Figure 3-11 shows the variation of crack
Join
Block Zone
Zone edge length
98
depth with mesh size. In all models in this series, the soil was assigned zero tensile
strength, 5MPa compression modulus (𝐸), 1000 suction increments and 50kPa surface
suction. The tensile strength and the suction were kept constant throughout the depth.
Here also two model series were analysed with 10m and 4m deep water tables.
When decreasing the mesh size in both models with different block heights, (𝑊) the
crack depths reduce and finally achieve a constant value. In both cases, the mesh
dependency ceases when the mesh size is less than 0.1m. Hence, a mesh size of 0.1m
was selected as the optimum mesh size for all the models used in this analysis.
Figure 3-11 Variation of crack depths with mesh size when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡=0kPa and 𝑆0=50kPa
Other than the mesh size and size of the block, the damping method and method of
changing stresses with time also generally affects the results in UDEC models. Hence,
2
2.2
2.4
2.6
2.8
3
3.2
3.4
0 0.1 0.2 0.3 0.4 0.5
Dep
th o
f Cra
ckin
g (m
)
Mesh Size (m)
Comparison of Crack Modelling Approaches 99
those factors also tested for the above model. As discussed earlier, damping auto and
local was used for the testing. Although, no significant change was observed, damping
local was selected for the modelling due to the fact that local damping is more suitable
to minimize oscillations that may arise when abrupt failure occurs in the model (Itasca,
2004). Local damping is also recommended for static analysis in most instances.
UDEC needs many time steps to reach the equilibrium state since it was explicit method
of analysis. Hence, the effect of number of time steps between increments on results
was also tested. However, the number of time steps used was not too low starting from
500 steps. The results obtained showed only a negligible difference and, therefore, 1000
steps were selected as a reasonable number considering the unbalanced force and the
model run time. However, this approach was reviewed depending on the reduction of
unbalanced forces, which was the ultimate criterion for achieving equilibrium.
The suction while desiccating the soil layer was changed by increasing the tensile
stresses in cumulative increments. The following constitutive relation was used for
modelling suction stress (Rajeev and Kodikara, 2011).
𝛥𝜎𝑖𝑗 = 𝐷𝑖𝑗𝑘𝑙(𝛥𝜀𝑘𝑙 − 𝛥𝜀𝑠ℎ𝛿𝑘𝑙) [3-11]
where 𝛥𝜎𝑖𝑗 represents the incremental stresses induced by expansion of the soil
(compressive stress/strain are negative), 𝐷𝑖𝑗𝑘𝑙 is the tangent stiffness tensor and 𝛥𝜀𝑘𝑙 is
the observed (or mechanical) strain and 𝛿𝑘𝑙 is the Kronecker delta. The strain increment
𝛥𝜀𝑠ℎ is the isotropic free shrinkage increment experienced by the soil due to a suction
increment or a moisture decrease. This approach is acceptable for continuous drying
events as simulated in this chapter.
100
Z
ΔS = ∆σx
ΔS = ∆σy
ΔS = ∆σz
x
y
Three suction profiles were utilized similar to Section 3.2: a constant suction (Figure
3-1), a linearly decreasing suction along the depth (Figure 3-3) and a parabolic suction
variation (Figure 3-5). The linearly decreasing suction profile can be given by the
equation,
𝑆 = 𝑆0(1 − 𝑧 𝑊⁄ ) [3-12]
and the parabolic variation follows equation,
𝑆 = 𝑆0(1 − 𝑧2 𝑊2⁄ ) [3-13]
According to those, the suction is decreasing from the top surface (maximum suction) to
the ground water table (zero suction). The tensile strength of the soil and the joint was
also changed with the suction according to the relation,
𝜎𝑡 = 𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′ [3-14]
in each case. The value of 𝜙𝑏 is selected as (𝜙 − 50) following Morris et al. (1992)
according to their interpretation of data presented by Fredlund et al. (1978).
When plane strain conditions are assumed for the numerical analysis, the strains in x
and y directions and stress in z direction are, 𝜀𝑥 = 0, 𝜀𝑦 = 0 and 𝜎𝑧 = 𝛾𝑧.
Figure 3-12 Stress conditions in a soil particle at each increment
Comparison of Crack Modelling Approaches 101
The following description shows how the suction increments are applied in the UDEC
model. If we consider an element within the desiccating soil, the stress increments can
be shown as in Figure 3-12. The suction increment (Δ𝑆) is applied in all three directions
which will increase the stress in each direction by Δ𝜎. This consideration is approximate
only for soil closest to saturation. Since clay liners are generally compacted wet of
optimum, it is considered this is reasonable.
At the beginning before applying suction, the strains in each direction can be written as,
𝜀𝑥 =1𝐸
�𝜎𝑥 − 𝜈�𝜎𝑦 + 𝜎𝑧�� [3-15]
𝜀𝑦 =1𝐸
�𝜎𝑦 − 𝜈(𝜎𝑥 + 𝜎𝑧)� [3-16]
𝜀𝑧 =1𝐸
�𝜎𝑧 − 𝜈�𝜎𝑥 + 𝜎𝑦�� [3-17]
After the first cycle 𝜎𝑥 , 𝜎𝑦 and 𝜎𝑧 become 𝜎𝑥 + Δ𝑆, 𝜎𝑦 + Δ𝑆 and 𝜎𝑧 + Δ𝑆 respectively
where Δ𝑆 equals, Δ𝜎𝑥, Δ𝜎𝑦 and Δ𝜎𝑧 respectively. Then the new strains become,
𝜀𝑥 =1𝐸
�(𝜎𝑥 + Δ𝑆) − 𝜈�𝜎𝑦 + 𝜎𝑧 + 2Δ𝑆��
Since 𝜀𝑥 = 0,
𝜎𝑥 = 𝜈�𝜎𝑦 + 𝜎𝑧� − (1 − 2𝜈)Δ𝑆 [3-18]
Similarly, in the y direction,
𝜀𝑦 =1𝐸
��𝜎𝑦 + Δ𝑆� − 𝜈(𝜎𝑥 + 𝜎𝑧 + 2Δ𝑆)�
Since 𝜀𝑦 = 0,
𝜎𝑦 = 𝜈(𝜎𝑥 + 𝜎𝑧) − (1 − 2𝜈)Δ𝑆 [3-19]
102
But for the z direction where 𝜀𝑧 is not equal to zero,
𝜀𝑧 =1𝐸
�𝜎𝑧 − 𝜈�𝜎𝑥 + 𝜎𝑦�� +(1 − 2𝜈)
𝐸Δ𝑆
When substituting the value of yσ (equation [3-19]) in equation [3-18],
𝜎𝑥 = 𝜈[𝜎𝑧 + {𝜈(𝜎𝑥 + 𝜎𝑧) − (1 − 2𝜈)Δ𝑆}] − (1 − 2𝜈)Δ𝑆
𝜎𝑥 =
𝜈(1 − 𝜈) 𝜎𝑧 −
(1 − 2𝜈)(1 − 𝜈) Δ𝑆
[3-20]
The stress relationship for stress in the x direction (equation [3-20]) is similar to the
relationship derived by Morris et al. (1992).
3.4 Comparison of Numerical and Theoretical Results
Numerical models were run to predict the analytical results discussed in section 3.2. The
same three suction profiles were followed with two different theoretical assumptions,
namely the linear elastic method and the cracking with shear failure method. For these
analyses, the same input parameters as used by Morris et al. (1992) were selected; in
order to compare both results easily.
Numerical models were run with two layer thicknesses, 4m and 10m at which depths
the water table was placed in each case. For all models, the Poisson’s ratio (ν) was
chosen as 0.3. The results significantly change with the Poisson’s ratio as shown in
Figure 3-13. Hence the Poisson’s ratio should be carefully selected when modelling
crack depth numerically. The effective cohesion was assumed to be zero for all the
models in the linear elasticity method.
Comparison of Crack Modelling Approaches 103
Figure 3-13 Effect of the Poisson’s ratio on crack depth when 𝐸 =5MPa, 𝜎𝑡 = 0.5 𝑆 𝑡𝑎𝑛𝜙𝑏, 𝑊 =4m and 𝑆0=50kPa
The modulus of elasticity was selected as a constant value of 5MPa although it is a
function of suction. However, the effects of varying modulus of elasticity were also
examined. Young’s modulus was changed with suction based on 𝐸 = 𝐸𝑠 + 10 × 𝑆 where
𝐸 is the Young’s modulus, 𝐸𝑠 is the modulus at saturation, and 𝑆 is suction at any depth
based on relations between modulus and suction provided by Kodikara et al. (2004) and
Sawangsuriya et al. (2009). 𝐸𝑠 was taken to be 5MPa in a study on the effects of varying
modulus.
Figure 3-14 shows the results obtained numerically with and without changing Young’s
modulus as a function of suction. There is no significant difference between the two
results especially when the lower water table depths are used. Hence, it was decided that
changing Young’s modulus has no discernible effect on the results obtained
numerically.
0
0.5
1
1.5
2
2.5
3
3.5
0.25 0.3 0.35 0.4
Dep
th o
f Cra
ckin
g (m
)
Poisson's Ratio
104
Figure 3-14 Crack depths changing with the depth to water table when 𝐸𝑠=5MPa, 𝐸 = 𝐸𝑠 +10𝑆, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑊 =4m, 𝑆0=50kPa and 𝜙 =300
3.4.1 Constant suction profile (suction profile 1)
In this section, the desiccation cracks were produced using the numerical model UDEC
similar to the theoretical assumptions in Section 3.2.1. The constant suction profile was
applied incrementally as shown in Figure 3-15. Suction increment Δ𝑆 was calculated by
dividing the final suction by the number of suction increments. For the linear elastic
analysis the in-built elastic isotropic constitutive model was selected for the block
material and for the joint materials area contact-coulomb slip with residual strength was
selected.
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to Water Table (m)
Analytical, Morris et al. (1992)
Numerical with constant E
Numerical with changing E
Comparison of Crack Modelling Approaches 105
Dep
th
Suction
Wf
S0,fS0,4S0,3S0,2S0,1
ΔS
S0,f-1
Figure 3-15 Appling suction with time
When the suction increases in the zones within blocks with time, the joint opens
creating a gap inbetween the blocks. Figure 3-16 shows the joint of the model and
velocity vectors of zones after opening the crack. The length of that opening from top to
the last displaced point was recorded as the crack depth. The crack depths obtained are
shown in Figure 3-17 and Figure 3-18 in comparison to the analytical results.
The crack depths obtained from the numerical model were larger than the calculated
results when the tensile strength of the soil was zero in Figure 3-17 and Figure 3-18.
This is possible since Morris et al. (1992) do not allow for stress redistribution after
crack opening and in contrast the numerical model does open the crack, which
redistributes the developed tensile stresses. In order to verify this explanation, a
numerical model was also designed with a very high tensile strength at the joint to
106
preclude cracking and obtain the stress distribution expected numerically. When the
tensile strength was set to a high value, there was no crack propagation in the soil block,
but the stress distribution profile assumed in the numerical analyses by Morris et al.
(1992) could be replicated. The depth from the surface to the point where stresses
change from tension to compression was considered as the crack depth.
Figure 3-16 Crack of a model after opening
With high tensile strength (in this case 100kPa was selected) the predicted results were
similar to the analytical results using Morris et al. (1992) (Figure 3-17) with a 10 m
deep clay layer. The same kind of variation was observed when the depth to water table
was decreased to 4 m (i.e., decrease the depth of suction profile) while keeping other
parameters constant as shown in Figure 3-18. This variation would be due to the
differences in implementation of the results in two models, where the UDEC model
Depth of the crack
Comparison of Crack Modelling Approaches 107
used the incremental implementation in contrast to the Morris et al analytical model.
This checking was not undertaken for 10m depth water table since results would be
expected to be similar and it would require substantial computer time to complete.
Figure 3-17 Variation of crack depth with surface suction when 𝐸 =5MPa, 𝜈 =0.3, 𝑊 =10m, 𝜎𝑡=constant and 𝑆0=50kPa
The comparison shows that the numerical model produces deeper desiccation cracks
than the Morris et al. (1992) analytical model would produce. This analytical model
gives the value for the crack depth at a depth where the horizontal tensile force equals or
exceeds the given tensile strength. However, in reality the crack develops progressively.
When the crack opens by a small amount, the stresses applied at the face of the crack
then becomes zero and stresses concentrate at the tip of the crack. This causes the crack
0
1
2
3
4
5
6
10 20 30 40 50
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (kPa)
Linear Elasticity Solutions (W = 10m)
Analytical (Morris et al; 1992)
Numerical, σt=100kPa
Numerical , σt=0kPa
σt
σt
108
to extend further. This phenomenon happens until the stresses at the tip of the crack are
getting less than the tensile strength of the soil. UDEC model allows for progressive
crack opening and stress redistribution along the crack which is unaccounted for in the
analytical model. This reasoning was verified by the results obtained for the numerical
models with high tensile strength where no physical crack propagation took place
(Figure 3-17 and Figure 3-18). In those models, the stresses were not redistributed since
no cracks opened up. Hence those crack depths obtained are similar to the
corresponding analytical results.
Figure 3-18 Variation of crack depth with surface suction when 𝐸 =5MPa, 𝜈 =0.3, 𝑊 =4m 𝜎𝑡=constant and 𝑆0=50kPa
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
10 20 30 40 50
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (kPa)
Linear Elasticity Solutions (W = 4m)
Analytical Numerical, σt=100kPa Numerical, σt=0kPa σt
σt
(Morris et al; 1992)
Comparison of Crack Modelling Approaches 109
3.4.2 Linearly decreasing suction profile (suction profile 2)
When the suction profile decrease linearly with the depth as discussed in Section 3.2.2,
the crack depths were computed numerically in order to compare with the analytical
results. Both linear elastic and elastic allowing for shear failure was used for
comparison. In the linear elastic approach, shear failure in the bulk medium is not
accommodated for, similarly to Morris et al.’s analytical formulation. Therefore, the
models were rerun with shear strength of the bulk medium specified as Mohr-Coulomb
model in UDEC.
Depth
Suction
Wf
S0,fS0,4S0,3S0,2S0,1
W1
W2
W3
W4
ΔS
ΔW
Figure 3-19 Changing the suction in UDEC
As shown in Figure 3-19, suction at the surface increases by an amount of Δ𝑆 and the
depth of suction profile goes down by an amount of Δ𝑊 in each step. This
representation is similar to progression of drying due to diffusion of soil moisture by
evaporation. Δ𝑆 was calculated by dividing the final surface suction by a number of
suction increments and Δ𝑊 was calculated by dividing the final depth to water table by
110
the same number of suction increments. The suction values inbetween the top surface
and the bottom water table were calculated using the corresponding suction profile
given in equation [3-12].
With this suction profile, the variation of crack depth was obtained while changing the
depth of suction profile and surface suction. Results obtained through the approaches of
linear elastic and allowing for shear failure are presented in separate figures. The
variation of crack depth was obtained with the depth to water table and surface suctions
in both approaches. In all figures the numerical results are compared with the analytical
results.
Figure 3-20 Variation of crack depth with depth to water table when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = 0.5 𝑆 𝑡𝑎𝑛𝜙𝑏
0.0
1.0
2.0
3.0
4.0
5.0
1 2 3 4 5 6 7 8 9 10
Dep
th o
f Cra
ckin
g (m
)
Depth to water table (m)
Analytical Results, S0=25kPa Numerical Results, S0=25kPa Analytical Results, S0=50kPa Numerical results, S0=50kPa
Linear Elastic Approach
Comparison of Crack Modelling Approaches 111
Figure 3-21 Variation of crack depth with the surface suction when when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = 0.5 𝑆 𝑡𝑎𝑛𝜙𝑏 and 𝑊 =4m
Figure 3-20 shows the variation of the depth of cracking with the depth to water table
when subjected to 50kPa and 25kPa surface suction. Figure 3-21 shows the variation of
depth of cracking with different surface suctions comparing with the corresponding
analytical results. Similarly to the results previously presented, the numerical model
predicts higher crack depths than the analytical model. In fact, the numerically predicted
crack depths are almost double the analytical predictions. When surface suction
increases beyond 80kPa in Figure 3-21, the clay layer cracks in its entire depth.
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (kPa)
Linear Elastic Approach, W = 4m
112
Capp = C' + S tanϕb
τ
ϕ
σt σn
q
S
M
pnet
q
pnet
σt = 0.5 S tan ϕb
σt = σt (0)
(b)(a)
In the modelling allowing for shear failure, the shear failure envelope and the tensile
strength limit were imposed as shown in Figure 3-22. The shear failure envelope
modified for tensile failure is shown in Figure 3-22(a) (after Bagee (1985)), which is
known as Mohr-Paul failure diagram. Figure 3-22(b) shows the effect of suction on
failure criteria developed to include tension (after Morris (1992)). In this method, the
soil blocks were discretized as continuum elements with Mohr-Coulomb plasticity
behaviour and the joint in the middle with coulomb slip with residual strength behaviour
in shear.
The results are shown in Figure 3-23 and Figure 3-24. First one compares the change of
crack depth with the change of depth to water table with analytical results. Then the
variation of crack depth with surface suction is given in the next figure. The Young’s
modulus and the Poisson’s ratio were selected as 5MPa and 0.3 respectively. Cohesion
and tensile strength properties were set to change as functions of changing suction
within the soil block as the crack progressively opened. Following the functional
Figure 3-22 (a) Strength envelope indicating tensile failure showing the effect of tensile strength, (b) Schematic of failure envelope in tension
Comparison of Crack Modelling Approaches 113
relationships used previously, the cohesion was given by 𝑐 = 𝑆0(1 − 𝑧 𝑊⁄ )𝑡𝑎𝑛𝜙𝑏 and the
tensile strength was given by 𝜎𝑡 = 0.5𝑐.
Figure 3-23 Variation of depth of cracking with depth to water table when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa and 𝜙 =300
From the results shown in Figure 3-23 and Figure 3-24 it is clear again that the
numerical model predicts higher results. However, the difference predicted by the
curves is getting less compared to the results obtained from the linear elastic method.
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to Water Table (m)
Analytical, (Morris et al. (1992)
Numerical, UDEC
shear failure approach
114
Figure 3-24 Variation of crack depth with surface suction when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 =−𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑊 =4m and 𝜙 =300
3.4.3 Parabolic suction variation (suction profile 3)
Dep
th
Suction
Wf
S0,fS0,4S0,3S0,2S0,1
W4
W3
W2
W1
Figure 3-25 Applying suction in increments for parabolic variation
0
0.5
1
1.5
2
2.5
3
3.5
0 20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (kPa)
Shear Failure method solutions, W=4m
Comparison of Crack Modelling Approaches 115
As the final suction profile, a parabolic variation was followed as discussed in section
3.2.3 following equation [3-8] which is a better representation of the real field
condition. In this series of models similar input parameters were also used as the soil
properties and the suction applied in cumulative increments as shown in Figure 3-25.
Both linear elastic approach and shear failure approach were used in this analysis. The
results were then compared with the analytical results. Figure 3-26 and Figure 3-27
illustrate the results obtained through the linear elastic approach with surface suction
variation and depth to water table respectively. Also Figure 3-28 and Figure 3-29 show
the comparison of crack depth predictions through shear failure approach as before,
with surface suction and depth to water table.
Figure 3-26 Effect of surface suction on depth of cracking predictions for parabolic suction variation; when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑊 =4m and 𝜙 =300
0
0.5
1
1.5
2
2.5
3
3.5
4
0 20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface suction (kPa)
Analytical Numerical
Linear elastic approach
116
Figure 3-27 Effect of depth to water table on depth of cracking predictions for parabolic suction variation; when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa and 𝜙 =300
Figure 3-28 Depth of cracking variation with surface suction for parabolic suction profile when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑊=4m and 𝜙 =300
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to water table (m)
Analytical
Numerical
Linear elastic approach
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
0 20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (kPa)
Analytical Numerical
Shear failure approach
Comparison of Crack Modelling Approaches 117
Figure 3-29 Depth of cracking variation with water table depth for parabolic suction profile when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa and 𝜙 =300
The parabolic suction profile in clay layers is more realistic and predicts more
reasonable values for crack depths in comparison to the observed crack depths in the
field given in Table 2-1. When the linear elastic theoretical assumptions are used, the
numerical model predicts larger crack depths in comparison to those of the analytical
model ((Morris et al., 1992). In the shear failure approach, the results from both
methods appear to produce closer crack depth values.
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0 2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to Water Table (m)
Analytical
Numerical
Shear failure approach
118
3.5 Comparison of Results from Different Suction Profiles
The suction profile is the most important input information in desiccation crack
modelling. This is evident from the fact that the crack depths change quite significantly
as the suction profile changes. It can be expected that a more realistic suction profile
leads to more realistic crack depth values. However, due to the high complexity of the
suction profile below the ground surface it is reasonable to assume a representative
simple profile for analysis.
Figure 3-30 Behaviour of crack depth predictions with surface suction based on the suction profile; when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑊=4m and 𝜙 =300
The suction profiles 1, 2 and 3 were compared in Figure 3-30 and Figure 3-31 to
examine the effects of suction profiles. The results shown in both figures are for the
linear elastic approach with the same input parameters. Both figures show that the least
crack depths are predicted by the suction profile 3. In addition, the minimum difference
0
1
2
3
4
5
6
7
10 30 50 70 90
Dep
th o
f cra
ckin
g (m
)
Surface suction (kPa)
suc prof 1-Analytical Suc prof 1-Numerical suc prof 2-Analytical suc prof 2-Numerical suc prof 3-Analytical suc prof 3-Numerical
Comparison of Crack Modelling Approaches 119
between the numerical and analytical results is also observed when the suction profile 3
is used for the stress variation.
Figure 3-31 Behaviour of crack depth predictions with depth to water table based on the suction profile; when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa and 𝜙 =300
Similar results can be observed when shear failure is allowed for in the model (Figure
3-32 and Figure 3-33). The results given in Figure 3-32 and Figure 3-33 illustrate that
both analytical and numerical results produce similar results when the shear failure was
allowed. As shown in Figure 3-33, the crack depths predicted by the numerical method
are less than those predicted by the analytical method for the suction profile 3. A
possible reason for this occurrence is that the numerical method redistributes stresses
during crack opening and therefore, shear failure may occur to a larger extent thereby
reducing the crack depths. However, it can be expected that this result will depend on
the parameters used to define the shear failure of the soil.
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to water table (m)
Suc prof 2-Analytical Suc prof 3-Analytical Suc prof 2-Numerical Suc prof 3-Numerical
Linear Elastic Approach
120
Figure 3-32 Behaviour of crack depth predictions with surface suction based on suction profile allowing for shear failure; when 𝐸=5MPa, 𝜈=0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa and 𝜙=300
Figure 3-33 Behaviour of crack depth predictions with surface suction based on the suction profile; when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑊=4m and 𝜙 =300
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
2 4 6 8 10
Dep
th o
f cra
ckin
g (m
)
Depth to Water Table (m)
suc prof 2-Analytical suc prof 2-Numerical suc prof 3-Analytical suc prof 3-Numerical
Shear failure Approach
0
0.5
1
1.5
2
2.5
3
3.5
0 20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (kPa)
suc prof 2-Analytical suc prof 2-Numerical suc prof 3-Analytical suc prof 3-Numerical
Shear failure Approach
Comparison of Crack Modelling Approaches 121
3.6 Comparison of Different Theoretical Approaches
The values obtained through the different theoretical approaches are compared in this
section for the suction profiles 2 and 3. The suction profile 1 is highly idealistic in the
case of deep clay layers and predicts larger crack depths than expected. The suction
profile 1 is more applicable to a case where the ground is covered by an impermeable
barrier and the suction has equilibrated with the water table.
Figure 3-34 Comparison of results obtained through different theoretical assumptions with depth to water table variation when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa
and 𝜙 =300 (SS – allowing for shear failure, LE – Linear elastic)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to Water Table (m)
SS-Analytical SS-Numerical LE-Numerical LE-Analytical
122
Figure 3-35 Comparison of results obtained through different theoretical assumptions with surface suction variation, when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa and 𝜙 =300 (SS –
allowing for shear failure, LE – Linear elastic)
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (m)
SS- Analytical
SS- Numerical
LE- Numerical
LE- Analytical
Comparison of Crack Modelling Approaches 123
Figure 3-36 Comparison of results obtained through theoretical assumptions with depth to water table variation when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa and 𝜙 =300
(SS – allowing for shear failure, LE – Linear elastic)
Figure 3-34 to Figure 3-37 show the comparisons of results from linear elastic and shear
failure approaches with depth to water table and surface suction. Generally, the crack
depths from the numerical results are higher than those from the analytical results, for
the reasons explained earlier. Similarly, when shear failure is allowed for, the
discrepancy between the crack depths is reduced more than in the linear elasticity
approach.
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to Water Table (m)
SS-Analytical
SS-Numerical
LE-Numerical
LE-Analytical
124
Figure 3-37 Comparison of results obtained through different theoretical assumptions with surface suction variation when 𝐸=5MPa, 𝜈=0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑆0=50kPa and 𝜙=300 (SS –
allowing for shear failure, LE – Linear elastic)
In order to improve the analytical model with linear elasticity to incorporate stress
redistribution, it was attempted to redistribute the stresses due to opening of the crack in
the following section. With redistribution of stresses, the crack tip attracts higher
stresses, which may be able to be captured using LEFM theoretical approach. As
appears in LEFM theory, the stress intensity factor was selected as proposed by
Bentham et al. (1973) based on asymptotic approximation, given in equation [3-21].
𝐾𝐼 = (0.6825𝑃 + 0.439𝑄)�𝜋𝑧𝑐 [3-21]
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0 20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (m)
SS- Analytical
SS- Numerical
LE- Numerical
LE- Analytical
Comparison of Crack Modelling Approaches 125
where 𝐾𝐼 is the stress intensity factor, 𝑃 and 𝑄 are the horizontal normal stresses at the
tip of the crack and top surface of the crack respectively. The stresses based on stress
intensity factor were calculated using equation given by,
𝜎𝑥 =𝐾𝐼
√2𝜋𝑟 [3-22]
where 𝑟 is the distance from the crack tip. The stress from equation [3-22] is added to
that in equation [3.20] to get the updated tensile stress, which is then used to calculate
the depth of cracking compared with the tensile stress value as undertaken by Morris et
al. (1992). The new equation to calculate the depth of cracking comes out as a cubic
equation as given below.
𝐶42𝑧𝑐
2 − (2𝐶4𝐶5 − 𝐶4𝑧𝑖)𝑧𝑐2 + �𝐶5
2 + 2𝐶4𝐶5𝑧𝑖�𝑧𝑐 − �𝐶52𝑧𝑖 +
𝐾𝐼2
2𝜋� = 0
[3-23]
𝐶4 =𝜈
(1 − 𝜈) 𝛾 +(1 − 2𝜈)(1 − 𝜈)
𝑆0
𝑊−
0.5𝑆0𝑡𝑎𝑛𝜙𝑏
𝑊
𝐶5 =(1 − 2𝜈)(1 − 𝜈)
𝑆0
𝑊− 0.5𝑆0𝑡𝑎𝑛𝜙𝑏
where 𝑧𝑖 is the depth of cracking from equation [3-20]. Equation [3-23] was solved and
the results are shown in Figure 3.38 referring to it as ‘modified analytical’.
The depth of cracking also can be produced using LEFM approach and considering the
fracture energy as presented by Morris et al. (1994). In this instance the crack depth is
determined by the condition that the stress intensity factor falls below the critical stress
intensity factor. The method proposed by this approach is discussed in the Section
2.3.4.2 under the LEFM approach. By using the equation proposed for crack depths
(equation [2-16]) the crack depths were calculated and are shown in Figure [3-38]
126
referred to as ‘LEFM-Morris et al. (1994)’. The results from this method depend on the
used critical stress intensity factor used.
Figure 3-38 Comparison of values of crack depth with surface suction variation when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛𝜙𝑏𝑐𝑜𝑡𝜙′, 𝑊=4m and 𝜙 =300
The results obtained through the LEFM approach are similar to the results produced by
UDEC model through the linear elastic approach. Similarly, the linear elastic analytical
approach with stress redistribution (‘modified analytical’) produces quite similar results
to both above methods. Therefore, it is possible to say that an improved analytical
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (m)
LEFM- Morris et al. (1994) Modified analytical LE- Analytical SS- Numerical SS- Analytical LE- Numerical
Comparison of Crack Modelling Approaches 127
solution has been developed to predict the crack depths when shear failure is not taking
place. As explained earlier, the shear strength approach produces smaller crack depths.
It is useful to present the results obtained so far in non-dimensionless form to generalise
the results for other geometries. The properties that govern the desiccation cracking
process were identified as surface suction, tensile strength, suction profile, depth to
water table, soil unit weight, friction angle and the Poisson’s ratio. The crack depth (𝑧𝑐)
can then be written as,
𝑧𝑐 = 𝑓(𝜈, 𝑊, 𝜎𝑡, 𝑆0, 𝛾, 𝑆𝑢𝑐𝑡𝑖𝑜𝑛 𝑝𝑟𝑜𝑓𝑖𝑙𝑒 𝑎𝑛𝑑 𝜙) [3-24]
By considering Poisson’s ratio, suction profile and friction angle constant (and therefore
tensile strength variation known), only 𝜈, 𝑊, 𝜎𝑡 , 𝑆0 and 𝛾 need to be considered.
Therefore, from the Buckingham 𝜋 theorem (Buckingham, 1914; Buckingham, 1915),
two non-dimensional quantities, namely 𝑧𝑐 𝑊⁄ and 𝑆0 𝛾𝑊⁄ are considered. A parametric
study was undertaken to produce sufficient results to develop a relationship between
these two non-dimensional quantities. The results are given in Figure 3-39, for 𝜈 = 0.3,
𝜙 = 300 for the linearly decreasing suction profile (suction profile 2). The curves have
been produced for both linear elasticity and allowing shear failure. These curves can be
used to estimate crack depths approximately for the given set of other parametric values.
It is possible to develop similar crack depth curves to cover other ranges of parameters
that were held constant in this instance.
128
Figure 3-39 Design curve; 𝑧𝑐 𝑊⁄ vs. 𝑆0 𝛾𝑊⁄ curve when 𝜙 =300, 𝜈 =0.3 for suction profile given in Figure 3-19
3.7 Conclusions
This chapter undertook a study of crack depth evaluation using the UDEC program
utilising its existing capabilities. Three suction profiles were considered, namely
constant, linearly varying and parabolically varying with depth. Perhaps the last suction
profile is the most realistic for a deep clay layer. The results obtained from the
numerical analyses were compared with the analytical model developed by Morris et al.
(1992) to predict crack depths. Two methods of analysis were considered, namely
treating the soil as linear elastic allowing tensile failure and allowing shear failure to
occur within the soil. The results indicated that in the case of the linear elastic approach
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4
z c/W
S0/γW
LE, UDEC Modified analytical SS, UDEC
Comparison of Crack Modelling Approaches 129
with tensile failure, the UDEC results predicted much deeper cracks than the analytic
model did. This result was attributed to the stress redistribution that takes place when a
crack opens up, which was not considered in the Morris et al’s analytical model. For
shear failure however, the numerical and analytical models produce closer results.
Morris et al’s analytical model was then extended to allow for stress distribution, which
eventually gave results closer to those predicted by UDEC. As a typical example, the
results were presented in non-dimensional form for general use.
A third approach where the LEFM theory was also proposed by Morris et al. The results
from this analysis was also compared with the UDEC results and it was found that
Morris et al results produced slightly higher crack depths than the UDEC results.
However, these results are highly dependent on the critical crack intensity factor or
fracture toughness chosen for the soils. In the current form of UDEC, fracture energy
(or facture toughness) associated with a joint or crack opening is not considered
directly. In Chapters 6 and 7, incorporation of fracture energy into UDEC model will be
considered in more detail.
The main issue of using UDEC is the run time in incremental nonlinear analysis. The
model run time changes significantly with the size of the block, number of grid points
and the method of analysis used. When running 10m deep 40m wide clay layer with
50cm mesh for the shear failure method, one model took about a month to complete
with a run in a core 2 duo, 4GB ram computer.
Chapter 4
EXPERIMENTAL INVESTIGATION OF SHRINKAGE AND SWELLING BEHAVIOUR
4.1 Introduction
Through various field and laboratory studies, it is well established that expansive clay
soils experience seasonal movement in the ground surface in wet and dry periods. Some
of these studies focused on the swelling and the shrinkage behaviour of expansive soils
(Richards et al., 1983; Sharma, 1998; Tripathy et al., 2002). However, very few studies
have been conducted to observe the cyclic swell-shrink behaviour of expansive soils for
several numbers of cycles.
The degree of swelling and shrinking depends on the amount of clay minerals within the
soil and the type of the clay minerals. However, the behaviour can be assumed similar
for any expansive soil and varies only in degree (Sharma, 1998; Kodikara, 2012). When
a soil layer is subjected to a sufficient number of wet-dry cycles, it can act
predominantly in an elastic or reversible manner, which is generally referred to as
130
Experimental Investigation of Shrinkage and Swelling Behaviour 131
environmental stabilization (Gould et al., 2011) or aging. Aging of newly compacted
clay soil and its effect on soils behaviour has been studied in the past by several workers
including Tripathy et al. (2002). How the aging of expansive soils will affect their
swell-shrink behaviour is not fully explained and more experimental work needs to be
undertaken in this area.
In Chapter 2, it was highlighted that stress induced by desiccation as needed for fracture
modelling can be represented either by suction increment or moisture increment. The
stress induced by the moisture content change of soil can be analysed using a hydric
constant (𝛼) with the change of moisture content (Kodikara and Choi, 2006; Peron et
al., 2007). However the behaviour of this α value due to the cyclic swell-shrink
behaviour is unknown. Hence this chapter is dedicated to present the laboratory
experiments carried out to examine the variation of 𝛼 value by analysing the data
published in the literature and undertaking new experiments. During experiments, the
behaviour of the soil subjected to wet-dry cycles will be analysed for a number of
different expansive soils with different initial and test conditions. A more general
variation of 𝛼 will be presented on the basis of the results obtained.
4.2 Hydric constant (𝜶) in Stress Analysis
In order to analyse the behaviour of unsaturated soils accurately, it is important to
compute the stress-strain behaviour of the soil subject to boundary and initial
conditions. Unfortunately, there is still not a universally accepted model to capture the
complex behaviour observed by compacted soils. For instance, D’onza et al. (2011)
132
presented a blind test on different constitutive models independently calibrated by
different teams of researchers on the same set of suction-controlled triaxial and
oedometer tests performed on compacted silty soil samples. These calibrated
constitutive models were then used to predict the behaviour of some other soil tests and
to compare the results blindly. None of the models agreed fully with the experimental
data showing the need for improvements of constitutive models that are currently
available to predict the unsaturated soil behaviour. None of these test results included
desiccation cracking of soils, and hence, such behaviour was not modelled.
Generally, the stresses induced due to moisture content change were analysed by most
researchers using the suction variation. Hence the swell-shrink behaviour was also
described as a function of suction; e.g., Barcelona Basic Model, BBM, (Alonso et al.,
1987; Kodikara, 2012). However, it is evident that the measurements of suction in
expansive soils are not straightforward especially under field conditions. Currently, a
number of techniques such as axis translation (increasing air pressure to generate a
curved meniscus while keeping water pressure atmospheric) tensiometer, thermal
conductivity sensor and osmotic methods need to be employed to get the full suction
range covered by laboratory methods. Some of these techniques are not able to be
employed in the field or may not work well. Therefore, although suction is a well-
defined potential of the soil for water flow, its measurement and its use as a stress state
variable is still posing problems to advancement of unsaturated soil mechanics.
In 2009, Baker and Frydman questioned the use of suction in geotechnical constitutive
theory (Baker and Frydman, 2009). In their review, they pointed out that all devices for
measuring soil suction actually measure soil-water potential (internal or matric) rather
Experimental Investigation of Shrinkage and Swelling Behaviour 133
than the state of pressure (tension) in the soil water, except for the tensiometer. The
actual water tension in unsaturated soil was recognized as difficult to be measured or
controlled. Furthermore they explained that the matric potential consists of two major
components namely, adsorption and capillary, and only the capillary component may be
explained in terms of mechanical pressure (tension). They argued that the geotechnical
literature generally ignores the adsorption potential when interpreting “suction”
measurements although under a significant range of conditions adsorption may be the
main component of the matric potential. This omission results in prediction of
unrealistically large water tensions which cannot be realized (due to cavitation) under
field conditions, when the air pressure is normally atmospheric.
In addition, suction variation does not give the full range of moisture content variation
in repeating wet-dry cycles (to be discussed later in this chapter) and displays
significant hysteresis during drying and wetting, increasing the complexity of the
problem. Due to these drawbacks, in this research moisture content variation is used
over suction as advocated by some other researchers (Costa, 2009; Gould et al., 2011;
Kodikara, 2012).
When the stresses are analysed in terms of moisture content variation, the potential
strain can be obtained using a constant of proportionality known as the hydric constant
(𝛼) and the difference of moisture content similar to heat flow analysis (i.e. ∆𝜀𝑠ℎ =
𝑘∆𝑇, where 𝑘 is the thermal conductivity, ∆𝑇 is the thermal gradient and ∆𝜀𝑠ℎ is the
shrinkage strain change).
∆εvsh = α∆w [4-1]
134
However, unlike for thermal expansion coefficient, α is not a constant and its behaviour
has not been studied in detail with the cyclic wetting and drying for compacted soils. A
theoretical explanation of the hydric coefficient is presented as follows. According to
the MPK framework presented by Kodikara (2012), the net stress in unsaturated soil can
be presented as a function of void ratio, e and the moisture ratio, 𝑒𝑤, where:
𝑒𝑤 = 𝑤𝐺𝑠 [4-2]
Therefore, this functional relationship can be presented as:
𝜎 = 𝑓(𝑒, 𝑒𝑤) [4-3]
By partially differentiating [4-3] gives:
𝑑𝜎 = �𝜕𝜎𝜕𝑒
�𝑒𝑤
𝑑𝑒 + �𝜕𝜎
𝜕𝑒𝑤�
𝑒𝑑𝑒𝑤 [4-4]
From the cyclic formula,
�𝜕𝜎
𝜕𝑒𝑤�
𝑒= −
� 𝜕𝑒𝜕𝑒𝑤
�𝜎
�𝜕𝑒𝜕𝜎�
𝑒𝑤
[4-5]
Substituting equation [4-3] in equation [4-4] gives:
𝑑𝜎 = �𝜕𝜎𝜕𝑒
�𝑒𝑤
𝑑𝑒 −� 𝜕𝑒
𝜕𝑒𝑤�
𝜎
�𝜕𝑒𝜕𝜎�
𝑒𝑤
𝑑𝑒𝑤
𝑑𝜎 = �𝜕𝜎𝜕𝑒
�𝑒𝑤
𝑑𝑒 − �𝜕𝑒
𝜕𝑒𝑤�
𝜎�
𝜕𝜎𝜕𝑒
�𝑒𝑤
𝑑𝑒𝑤
𝑑𝜎 = �𝜕𝜎𝜕𝑒
�𝑒𝑤
(𝑑𝑒 − �𝜕𝑒
𝜕𝑒𝑤�
𝜎𝑑𝑒𝑤) [4-6]
Experimental Investigation of Shrinkage and Swelling Behaviour 135
when,
�𝜕𝑒
𝜕𝑒𝑤�
𝜎= 𝛼∗ [4-7]
and,
�𝜕𝜎𝜕𝑒
�𝑒𝑤
= 𝐾
where 𝐾 is the bulk modulus at the particle moisture ratio and 𝛼∗ is the hydric
coefficient given by the partial differential of void and moisture ratios at a particular net
stress (equation [4-7]). Then equation [4-6] becomes,
𝑑𝜎 = 𝐾(𝑑𝑒 − 𝛼∗𝑑𝑒𝑤) [4-8]
If this relation is for only one directional stress then the 𝐾 should be replaced by 3𝐾.
𝛼∗𝑑𝑒𝑤 represents the free shrinkage. The relation to obtain the potential free shrinkage
can be written as,
𝛼∗ 𝑑𝑒𝑤
(1 + 𝑒0)= ∆𝜀𝑣𝑠ℎ [4-9]
Hence, from equations [4-1] and [4-9],
𝛼∗ 𝑑𝑒𝑤
(1 + 𝑒0)= 𝛼∆𝑤 [4-10]
Using the relation presented in equation [4-2] and substituting it in equation [4-10] leads
to the relation,
𝛼 = 𝛼∗ 𝐺𝑠
(1 + 𝑒0) [4-11]
136
The relation given by equation [4-11] can be used to calculate the hydric constant. It is
clear that the hydric coefficient depends on the stress level. This coefficient (𝛼∗) can be
measured experimentally by following a relatively easy test. The following sections
explain the swelling and shrinking experiments conducted to develop the relationships
of void and moisture ratios and thereby the variation of the hydric coefficient.
4.3 Experimental Procedure
Some swell-shrink tests were conducted to calculate the hydric coefficient. The soils
used in the investigations, their basic characteristics, the method of sample preparation
and experimentation adopted, observations made and methods of calculations are
presented in this section.
4.3.1 Materials
The expansive soil material was collected from the clay deposits in North Altona in
Melbourne, Australia at the depths of 0.4m to 2.0m. It is light brown in colour when
dry, becoming dark when wet. This clay is referred to as Altona clay in this thesis.
Several basic geotechnical tests were carried out according to the Australian Standards
in order to characterise the properties of the clay used here and these are listed below:
• Particle size distribution analysis;
• Atterberg limits;
• Specific gravity;
Experimental Investigation of Shrinkage and Swelling Behaviour 137
• Compaction tests (Standard and Modified);
• One dimensional consolidation test.
4.3.1.1 Particle size distribution analysis
Soil particles down to 75μm sieve (clay soil), were prepared by the wet sieving method
in accordance with the Australian Standard (AS1289.3.6.1, 2009), and were analysed
using an optical particle size analyser instead of the hydrometer as noted in the
Standard, due to its higher level of accuracy. The soil samples were soaked in water for
one hour before being agitated and then washed on a 425μm sieve until the water was
clear and all of the material had been washed through the sieve. The washed material
with water was collected in a tray placed under the sieve and was oven dried at 105°C to
collect the washed soil particles in powder form.
A mass of 5g of the soil powder was suspended in distilled water and poured into the
sample dispersion unit of the particle size analyser for the analysis. In this device, the
soil particles within the solution were passed through a focused laser beam. The laser
light is scattered at an angle that is inversely proportional to the soil particle size, and
the angular intensity of the scattered light was then measured by a series of
photosensitive detectors. The map of scattering intensity versus angle was used to
calculate the particle size using a model in the analyser software. The optical particle
size analysis was repeated three times and the average results are plotted in Figure 4-1.
The results show that Altona clay is relatively well graded with 63 % of clay, 30 % of
silt and less than 7% of sand.
138
Figure 4-1 Particle size distribution of Altona clay
4.3.1.2 Atterberg limits and specific gravity
The Australian Standard was followed in determining the liquid limit (AS1289.3.1.1,
2009) and the plastic limit (AS1289.3.2.1, 2009). The moisture content was also
measured in accordance with the instructions in Australian Standards (AS1289.2.1.1,
2005). The value of the liquid limit for Altona clay was found to be 70.2% and the
plastic limit was 21.8%. The plasticity index was obtained as 48.4%. Under the unified
soil classification system (USCS), this soil is classified as clay of high plasticity (CH)
on the basis of Atterberg limits.
The specific gravity of Altona clay was determined as 2.614. The test was repeated
twice to get the value more accurately. The linear shrinkage value was also measured
according to the standard (AS1289.3.4.1, 2008 ) giving a value of 16%. This means that
0
10
20
30
40
50
60
70
80
90
100
0.0001 0.001 0.01 0.1 1 10
Perc
enta
ge P
assi
ng (%
)
Practicle Size (mm)
Clay Silt Sand
Gravel
Experimental Investigation of Shrinkage and Swelling Behaviour 139
the soil has the potential to shrink 16% when dried from the liquid limit to oven dry
condition.
4.3.1.3 Compaction tests
Standard and modified compaction tests were conducted according to the Australian
Standards, (AS1289.5.1.1, 2003 ) and (AS1289.5.2.1, 2003 ) respectively. Soil was
prepared and the test was conducted following the standard. An automatic soil
compactor (shown in Figure 4-2) was used to eliminate the inconsistencies in
compaction measured by an operator.
The automatic blow pattern of the compactor ensures the optimum compaction for each
layer of soil. The rammer automatically travels across the mould and the table rotates
the mould in equal steps on a base that is extremely stable. The number of blows,
weight of the rammer and height of the rammer drop can be set at the beginning.
140
Figure 4-2 Automatic soil compactor designed for proctor and CBR Compaction and removing sample after modified compaction test
The results of the compaction tests are shown in Figure 4-3. The optimum moisture
content for standard proctor compaction was approximately 23.5% whereas for
Automatic compaction machine
Controls
Compaction hammer
Compaction mould
Equipments for Compaction test
Removing jack
Experimental Investigation of Shrinkage and Swelling Behaviour 141
modified compaction the optimum moisture content was around 16%. The dry densities
for both standard and modified compaction tests at the optimum points are 1.5g/cm3
and 1.8g/cm3 respectively.
Figure 4-3 Compaction curves for Altona clay
4.3.1.4 One dimensional consolidation test
One-dimensional consolidation tests were also carried out in accordance with Australian
standards (AS1289.6.6.1, 1998) using an oedometer. Soil was prepared to a slurry of
152% moisture content which is approximately twice the liquid limit to ensure that air
bubbles were minimised and workability was increased using a mechanical soil mixer
(Figure 4-4). The slurry was tested after leaving for at least 24 hours for equalization.
The slurry was placed in a 63 mm diameter (18.2 mm height) cylinder and LoadTrac III
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
4 8 12 16 20 24 28 32 36
Dry D
ensi
ty (g
/cm
3 )
Gravimetric Water Content (%)
Modified ProctorStandard Proctor100% Saturation80% Saturation60% Saturation40% Saturation
142
(Geomech Corp.) which is an automated consolidation oedometer rig. The experimental
set up is shown in Figure 4-5.
Using the software provided with the LoadTrac, the user inputs the stress step values (in
kPa) and minimum and maximum times (in hours). Various pressures were preset to
allow measurements of the soil strain under predetermined pressures. The machine
places the predetermined stress onto the sample based on the area given and the load is
applied for the inputted amount of time. The pressure is built up slowly to allow the
pore water pressure to dissipate. The sample was allowed to swell at each unloading
step, which was also performed at predetermined steps. The linear variable differential
transformer (LDVT) and load cells were calibrated prior to use.
Figure 4-4 Mechanical soil mixer Figure 4-5 LoadTrac III consolidation machine
Experimental Investigation of Shrinkage and Swelling Behaviour 143
Figure 4-6 represents the variation of void ratio with the pressure on a log scale. From
this figure, the soil compressibility parameters for loading (λ0) and unloading (κ0) under
saturated conditions were obtained as λ0 of 0.391 and κ0 of 0.038 respectively.
Figure 4-6 Compression curve for Altona clay
4.3.1.5 Mineralogy
Table 4-1 Mineralogy content of Altona clay
Quartz Albite Orthoclase Kaolin Smectite Calcite Halite Anatase
59% 2% 3% 2% 31% 3% <1% <1%
The study of mineralogy of the soil was not within the scope of this study. Hence,
mineralogy analysis was not conducted in this research. However, the similar soil
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
1 10 100 1000 10000
Void
Rat
io
Pressure (kPa)
144
obtained from the same area has been analysed in the past (Gallage et al., 2008) from
which it is possible get the results for the composition of this soil , as given in Table 4-
1.
4.3.1.6 Summary of results
The properties of soil obtained from the tests described above are summarised in Table
4-1.
Table 4-2 Summary of the soil classification test results
Colour Light brown / beige
Linear shrinkage 16%
Liquid limit 70.2%
Plastic limit 21.8%
Plasticity index 48.4%
Soil class Inorganic clays of high plasticity (CH)
Standard optimum moisture content 23.5%
Compressibility parameters for loading 0.391
Compressibility parameters for unloading 0.038
Specific gravity 2.614
4.3.2 Sample preparation and set-up
A targeted experimental procedure needed to prepare a large number of samples with
the same density, moisture content and soil homogeneity was essential. Hence, selection
Experimental Investigation of Shrinkage and Swelling Behaviour 145
of an appropriate method for sample preparation and initial conditions required careful
consideration and balance.
4.3.2.1 Sample size
The samples were prepared in a basic size oedometer ring of 76mm in diameter. When
deciding the height of the sample, three conditions were taken into consideration:
1. Loss of energy input due to boundary friction;
2. Maintaining homogeneity throughout the sample; and
3. Producing sufficient swelling and shrinking effect while testing.
The height of the sample needed to be small enough to avoid a significant degree of
energy loss and to avoid heterogeneity of the sample (Romero, 1999). However, the
height should be sufficient to produce considerable volume change with the moisture
change. Considering these facts, the sample height was selected as 12mm.
4.3.2.2 Sample preparation
As stated before, samples were prepared carefully and identically in order to maintain
homogeneity and repeatability. The initial conditions of the samples were selected as
17±1% moisture content and 14.6±0.2kN/m3 dry unit weight. The starting point is
shown in the compaction curve in Figure 4-3, where it lies in the dry side of the
optimum. The behaviour of the soil compacted at wet side of optimum is reasonably
well known and some results can be found already in the literature (Romero, 1999;
Tripathy et al., 2002; Monroy, 2006). The dry side of the optimum has been neglected
146
although in actual field conditions there is a possibility of ending up on the dry side of
compaction under different compaction energies. Hence, the dry side of optimum was
considered in this research.
The soil was first oven dried and sieved to remove the sand size particles. The clay
powder was then mixed with the required amount of water manually for around 15
minutes by spraying until moisture was distributed evenly. The wet soil was placed in a
resealable polythene bag and sealed. This soil bag was sealed again in another bag to
make sure, the soil sample was completely air tight and left for at least 24 hours prior to
use.
Stainless steel oedometer rings were used in the test. The diameters of the rings were
76.5±0.1mm and heights were 19.1±0.1mm. The rings were cleaned, greased and
weighed at the beginning of the test.
Static compaction was selected for compacting samples. Static compaction shows much
higher potential in repeatability and uniformity over dynamic compaction (Sivakumar,
1993). Two methods were tested for ultimate sample preparation. Method one involves
compacting soil in one large ring and then cutting samples out of the compacted soil
using the cutting ring. The bottom of the sample was trimmed off and was then inserted
into the oedometer ring. Figure 4-7 shows the method followed in obtaining samples
using this method (method one). The second method involves compacting the soil
directly into the oedometer ring. The required quantity of soil was calculated on the
basis of the dry density to be achieved. Figure 4-8 illustrates method two.
Experimental Investigation of Shrinkage and Swelling Behaviour 147
Figure 4-7 Sample preparation using method one
Load
Cutting ring
Soil block
Loading plate
Cutting a sample
Cutting a sample
Compacting large volume
Before Compaction
Pushing the sample out
of the cutting ring
Oedometer cell
Prepared sample
148
Figure 4-8 Sample preparation using method two
Several samples were compacted using both preparation methods. The initial unit
weights and moisture contents of samples after compacting were compared. Figure 4-9
shows the results obtained from the sample preparation methods one and two. The
results for unit weight from method one show some scatter although the moisture
content is almost the same in all samples. Unit weight and moisture content results
obtained from method two show a small variation. The variation of unit weight in
different samples for method one may be due to the uneven distribution of soil in the
large container. The sample disturbance during cutting and moving the sample from one
ring to another may also have influenced conditions to some extent. Other than the un-
Prepared sample
Before Compaction
Oedometer cell
Compacting sample
Experimental Investigation of Shrinkage and Swelling Behaviour 149
acceptable initial condition variation, the first method is also difficult to execute
precisely in comparison to the second method. Hence, it was decided to compact the
samples directly into the oedometer ring using static compaction for all tests performed
in this research.
Figure 4-9 Values for unit weight and moisture content obtained as initial conditions. * - UW stands for Unit Weight, ** - MC stands for Moisture Content
4.3.2.3 Set-up
The intension of the test programme was to observe the change of void ratio over the
time when the sample was subjected to controlled wet and dry conditions. Several types
of apparatus can be used for measuring swell-shrink properties, and the conventional
oedometer is one of them. Monash University Civil Engineering laboratory has
oedometers designed and manufactured at Monash as shown in Figure 4-10. This
oedometer is relatively simple, easy to use and takes less space than the conventional
oedometer. Furthermore, wetting and drying tests take long periods of time and it was
10
12
14
16
18
20
22
24
0
2
4
6
8
10
12
14
16
18
20
0 1 2 3 4 5 6 7 8 9 10
Moi
stur
e Co
nten
t (%
)
unit
wei
ght (
kN/m
3)
Sample Number
Method 1 - UW* Method 2 - UW
Method 1 - MC** Method 2 - MC
150
not possible to utilise a conventional oedometer for a long period of time due to the
requirements of other students. Hence it was decided to use the Monash designed
oedometer for this study.
As shown in Figure 4-11, the Monash oedometer is loaded from above and the load is
transferred through a loading rod. An electronic dial gauge can be attached to it for the
measurement of displacements. The equipment can accommodate different sizes of
rings, but a conventional sized cell and ring were used. However, the main limitation of
the test set-up is the pressure that can be applied by the loads that can be placed above.
Typically about 100kPa pressure is easily achievable.
Figure 4-10 Monash designed Oedometer
Loading plate Load
Dial-gauge
Oedometer Cell
Experimental Investigation of Shrinkage and Swelling Behaviour 151
Figure 4-11 Schematic diagram of test set-up and placement of soil sample
The tests conducted had two main parts, wetting and drying, and a total of 50 samples
was tested. The samples were subjected to several wetting and drying cycles until the
volume change followed the same path for both wetting and drying. This condition was
considered as the environmentally stabilized condition for the environment imposed.
The tests were conducted with 5kPa constant pressure applied on the sample in a
temperature controlled room. The temperature was maintained at 25°C during wetting
and increased up to 40±5°C while drying using halogen working lamps. The pressure
used was typically a nominal pressure and is representative of an unloaded clay layer.
When obtaining the swelling path, the moisture content was increased by adding the
required amount of water through the holes on the top cap (Figure 4-12(a)) and by
inundating the sample in water when the full potential of swell was assumed to have
been obtained (Figure 4-12(b)). To trace one wetting or drying path 6 to 10 samples
were tested, and each sample provided only one point in the 𝑒 − 𝑒𝑤 space. The quantity
Oedometer Cell
Dial-gauge
Load Loading plate
Porous disks
Lid with holes
152
of water to be added was decided allowing 3 to 5mm water evaporation from the sample
during the swelling period. During other periods the cell was covered with plastic to
prevent evaporation.
Figure 4-12 (a) Saturating the sample for swelling, (b) Wetting the sample to get the swell before saturation
Before selecting the test duration for swelling and shrinking, several tests were
conducted letting the sample swell until it did not show a significant increase in height.
For drying similar tests were performed. Figure 4-13 and Figure 4-14 show the vertical
displacement behaviour with time for the first swelling and shrinking after first swelling
respectively. Figure 4-13 illustrates that, almost 90% of the swelling occurred during the
very first day after wetting. Hence it was decided to allow the samples to swell for three
days while carrying out the wet dry cycles. However, the drying process usually took a
longer period than swelling took and it was necessary to allow for about 8 days for
drying, as shown in Figure 4-14. In wet dry cycles, the samples were allowed to dry for
10 days ensuring more accurate results.
(a) (b)
Syringe with water
Water surrounding the sample
Experimental Investigation of Shrinkage and Swelling Behaviour 153
Figure 4-13 Vertical displacement of three samples compacted to similar initial conditions when subjected to wetting.
While running a test, the vertical displacement of the sample was recorded every 24hrs.
Then at the end of the test, the moisture content and unit weight were measured
immediately after dismantling the test set-up. Even before the full shrinkage, cracks and
considerable lateral shrinkage were observed. Figure 4-15 shows the photos of the
sample at different stages of the test. The moisture content of the sample was measured
by oven drying the soil obtained from different places of the sample.
To measure the unit weight the following procedure was used in a water displacement
technique. Immediately after dismantling the sample, a small portion of the sample was
removed without applying too much force to avoid any change in properties. The
specimen was then weighed and coated with paraffin wax (density = 0.96g/cm3) and
0
0.5
1
1.5
2
2.5
0 2 4 6 8 10 12
Vert
ical
dis
plac
emen
t (m
m)
No. of days
Sample 1 Sample 2 Sample 3
154
again weighed with the paraffin wax coat by hanging it underneath an electronic
balance. After that, the specimen was submerged in a beaker of distilled water and the
submerged mass was measured. Using equation [4-12] the unit weight was calculated.
𝑈𝑛𝑖𝑡 𝑤𝑒𝑖𝑔ℎ𝑡 𝑜𝑓 𝑡ℎ𝑒 𝑠𝑎𝑚𝑝𝑙𝑒 =𝑤(𝑖𝑛 𝑎𝑖𝑟)
𝑤 (𝑖𝑛 𝑎𝑖𝑟) − 𝑤 (𝑖𝑛 𝑤𝑎𝑡𝑒𝑟) × 𝑢𝑛𝑖𝑡 𝑤𝑒𝑖𝑔ℎ𝑡 𝑜𝑓 𝑤𝑎𝑡𝑒𝑟 [4-12]
where, w is the mass of the sample. The calculated soil unit weight was corrected for the
wax volume. The measured vertical displacement of the soil was not used in the void
ratio calculation due to the cracks and lateral strain.
Figure 4-14 Vertical displacement of three samples compacted to similar initial conditions when subjected to drying after full swelling
-2.5
-2
-1.5
-1
-0.5
0 0 2 4 6 8 10 12
Vert
ival
dis
plac
emen
t (m
m)
No. of days
Sample 1 Sample 2 Sample 3
Experimental Investigation of Shrinkage and Swelling Behaviour 155
Figure 4-15 Photos of the sample at different stages
After 1st drying
After 2nd wetting After 2nd drying
After 1st wetting
After 3rd wetting
After 4th wetting
After 3rd drying
After 4th drying
156
4.4 Results
Trends and variations of different parameters captured during the tests were presented in
this section. The vertical displacements and void ratios were taken into consideration.
4.4.1 Swell-Shrink cycles
The maximum and minimum vertical displacements observed during wetting and drying
cycles were plotted in Figure 4-16 and Figure 4-17. The samples were compacted at the
same initial conditions under the same vertical pressure as discussed previously. The
tests were conducted continually for several wet dry cycles.
Figure 4-16 Marginal vertical displacements during wet-dry cycles for several samples
The results indicate that the initial wetting and initial drying (first cycle) have different
vertical displacements variations. The next cycles show a repetitive variation. From
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
1st Wetting
1st Drying
2nd wetting
2nd Drying
3rd wetting
3rd Drying
4th wetting
4th Drying
Swel
ling
heig
ht (m
m)
Sample 45
Sample 46
Sample 47
Sample 48
Sample 49
Sample 50
Experimental Investigation of Shrinkage and Swelling Behaviour 157
Figure 4-17, it can be seen that Altona clay shows more than 30% net expansion from
the initial height. The change in vertical displacement during wet dry cycles is expected
to be around 20% after aging. This results show a good agreement with the results of
Tripathy et al. (2002).
Figure 4-17 Typical average vertical displacements for Altona clay
4.4.2 Swelling or shrinking paths
The variation of void ratio in each cycle is examined and the void ratio is plotted against
the moisture ratio. Figure 4-18, Figure 4-19, Figure 4-20 and Figure 4-21 show the first,
second, third and fourth cycles respectively. In each figure wetting is followed by
drying to obtain the full cycle.
In Figure 4-18, the initial point indicates the point where the sample was compacted.
The lower part of the swelling curve from the initial point was obtained by drying the
0
5
10
15
20
25
30
35
40
45
1st Wetting
1st Drying 2nd wetting
2nd Drying
3rd wetting
3rd Drying
4th wetting
4th Drying
Swel
ling
Perc
enta
ge (s
wel
ling
heig
ht/I
nitia
l hei
ght,
%)
158
sample. It can be observed that the curves behave wildly at the initial wetting and drying
cycles before converging to a stabilized curve achieved in the fourth cycle.
The stabilized curve is shown in Figure 4-21 where drying and wetting both are lying on
top of each other. In the beginning, the void ratio does not change with the moisture
content very much. However, after approximately 0.2 moisture ratio the void ratio
shows a significant change until close to the saturation followed by a flatter part again
to reach the saturation line. This behaviour agrees with results of Tripathy et al. (2002)
and the explanation given by Gould et al.(2011).
Figure 4-18 Swelling and shrinking paths for the first cycle
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Void
Rat
io, e
Moisture Ratio, wGs
1st cycle
Initial point
Wetting curve
Drying curve
Starting point of 1st
cycle
End point of 1st
cycle
Experimental Investigation of Shrinkage and Swelling Behaviour 159
Figure 4-19 Swelling and shrinking paths for the second cycle
Figure 4-20 Swelling and shrinking paths for the third cycle
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Void
Rat
io, e
Moisture Ratio, wGs
2nd cycle
Drying curve
Wetting curve
Starting point of 2nd cycle
End point of 2nd
cycle
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Void
Ratio
, e
Moisture Ratio, wGs
3rd cycle
Drying curve
Wetting curve
Starting point of 3rd cycle
End point of 3rd
cycle
160
Figure 4-21 Swelling and shrinking paths for the fourth cycle
4.4.3 Variation of hydric coefficient (𝜶∗)
The variation of hydric coefficient was examined in relation to number of cycles and
moisture content in Figure 4-22 and Figure 4-23.
It can be seen that 𝛼∗ shows similar behaviour in all swell shrink cycles, although the
first cycle shows little variation in comparison to the other cycles, as shown in Figure
4-22. However, for both shrinking and swelling curves similar variation of 𝛼∗ values
can be observed. A careful examination of the figure shows slightly higher values for 𝛼∗
while shrinking than while swelling.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Void
Rat
io, e
Moisture Ratio, wGs
4th cycle
Starting point of 4th cycle
End point of 4th
cycle
Experimental Investigation of Shrinkage and Swelling Behaviour 161
Figure 4-22 Variation of 𝛼∗ with number of cycles
At low moisture contents, 𝛼∗ is low and increases with the moisture content
approximately up to 1 and remains relatively constant for a long range of moisture
content. Finally, the value of 𝛼∗ drops drastically as shown in Figure 4-23.
Figure 4-23 Variation of 𝛼∗ with moisture ratio
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
Hyd
ric C
oeff
icie
nt (α
*)
Wetting
Drying
1st cycle 2nd cycle 3rd cycle 4th cycle
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 0.2 0.4 0.6 0.8 1 1.2
Hyd
ric C
oeff
icie
nt (α
*)
Moisture Ratio, wGs
1st wetting 1st Drying 2nd wetting 2nd Drying 3rd Wetting 3rd Drying 4th Wetting 4th Drying
162
4.5 Other Research on Wet-Dry Cycles
It was decided to analyse different types of soils in order to draw a general conclusion.
Hence, existing data obtained by digitising the results in several published research
works were analysed.
4.5.1 Experimental data analysis of Sharma (1998)
In this study, 10% Wyoming sodium bentonite with speswhite kaolin was used as
expansive soil represented by ‘BK’. Speswhite kaolin alone was denoted by “K’. The
basic properties of the soil are given in Table 4-3. The compaction curve and the initial
position of the soil are shown in Figure 4-24. Tri-axial tests were used to measure the
swell-shrink behaviour and the axis translation technique was the basic approach in
measuring suction. The characteristics of the tests used for this analysis are shown in
Table 4-4.
Table 4-3 Properties of the soil
Parameter Value
BK K
Specific Gravity (𝐺𝑠) 2.65
Liquid limit 93% 65%
Plasticity Index 60% 33%
Optimum Moisture content 29%
Optimum dry density 1.44Mg/m3
Experimental Investigation of Shrinkage and Swelling Behaviour 163
Figure 4-24 Compaction curves and the initial positions of samples
1
1.1
1.2
1.3
1.4
1.5
15 20 25 30 35 40
Dry
Den
sity
(Mg/
m3)
Moisture Content (%)
164
Table 4-4 Summary of tests by Sharma
Test No.
Material Static compaction Pressure (kPa)
Drainage
No. of days
Test method Mean net stress (kPa)
Suction changing rate (kPa/hr)/Value of k for non-linear variation of suction
No of cycles
Suction Hydric coefficient (𝜶∗)
1 2 3
wet dry wet dry wet
1 BK 400 Single 68 wetting/drying 10 2.5 2 400-50-380-0-370
0.298 0.721 0.357 0.869
2 BK 400 Single 28 wetting/drying 50 1.6 1 400-100-400 -0.172
0.596
3 BK 800 Single 67 wetting/drying 50 0.8 1.5 400-100-400-10 0.295 0.563 0.338
4 BK 800 Double 37 wetting/drying 10 2.0 1.5 400-100-400-10 0.374 0.726 0.569
5 BK 400 Double 56 wetting/drying 20 1.6 1.5 400-100-400-10 0.321 0.630 0.523
8 BK 400 Double 42 1. wetting/drying 10 ±0.00016 1 300-20-300 0.187 0.884
2. loading 10-175 300
10 BK 400 Double 66 1.loading/unloading
10-100-10
200
2. wetting/drying 10 ±0.00016 1 200-20-200 0.504 0.851
3.loading/unloadi 10-250- 200
Experimental Investigation of Shrinkage and Swelling Behaviour 165
ng 10
12 K 400 Double 13 wetting/drying 10 Wetting- ±0.00048
Drying- ±0.00024
1 400-100-400 0.468 0.497
14 K 400 Double 76 wetting/drying 10 ±0.00016 2.5 400-50-400-20-400-5
0.467 0.661 0.453 0.636 0.480
16 BK 400 Double 71 1.wetting/drying 10 ±0.00016 1 400-20-400 0.195 0.905
2. loading 10-40 400
3.wetting/drying, P’ varied to v kept constant
40-70-1 ±0.00016 1 400-20-105
18 K 400 Double 39 1. wetting/drying 10 ±0.00016 1 200-20-200 0.495 0.684
2. loading 10-300 200
19 BK 3200 Double 36 wetting/drying 10 ±0.00016 1 400-50-175 0.785 1.044
20 BK 3200 Double 48 wetting/drying 10 ±0.00016 1 100-1-100 0.905 1.007
Experimental Investigation of Shrinkage and Swelling Behaviour 166
The wetting and drying paths obtained from the tests are shown in Table 4-4, and are
plotted in the void ratio-moisture ratio space as shown in Figure 4-25 and Figure 4-26.
The results of the samples tested under 10kPa net vertical stress are shown in Figure
4-25 while the results at 20kPa and 50kPa pressures were plotted in Figure 4-26. The
curves generally show a flatter line while swelling and a steeper line while shrinking. It
can be observed that some samples (Tests 2, 5, 8 and 16) collapsed while wetting when
they reached the pressure line corresponding to the net vertical stress in comparison
with the MPK framework. Also shown in these figures are the likely compaction curves
interpolated from the two results given.
Figure 4-27 shows the variation of the hydric coefficient with wet dry cycles. The value
of 𝛼∗ shows a rough variation due to the small intervals selected in determining the
gradients. However, a general trend can be observed. While the initial swelling 𝛼∗ is
about 0.5, it drops at the end of the swell and again increases up to approximately 0.8
during the shrinking period. A similar trend follows for the next cycle as well with 𝛼∗
values of 0.65 and 0.9 for wetting and drying respectively.
To observe the variation of 𝛼∗ closely, that variation was plotted against the degree of
saturation for each cycle as shown in Figure 4-28. Figure 4-28(a) shows the variation of
𝛼∗ during the first wetting period. 𝛼∗ for different samples between 0.55 and 0.1 while
slightly decreasing with the increased degree of saturation. Figure 4-25 shows the first
wetting, where 𝛼∗ ranges between 0.4 and 1. During the first drying 𝛼∗ increases rapidly
and then tends to decrease at higher degrees of saturation. The second wetting shows the
typical increasing and deceasing behaviour of 𝛼∗ for full swell or full shrink. The last
figure (Figure 4-28(d)) shows that the hydric coefficient increased up to the value 1.
Experimental Investigation of Shrinkage and Swelling Behaviour 167
Figure 4-25 Swelling and shrinking curves for samples under 10kPa vertical stress, Test 4 – initially compacted under 800kPa. Test19 & 20 – initially compacted under 3200kPa. Other
tests – initially compacted under 400kPa
Figure 4-26 Swelling and shrinking curves for samples under 20kPa (Test 5) and 50kPa (Test 2) vertical stress, Test 3 – initially compacted under 800kPa. Test 2 & 5 – initially compacted
under 400kPa
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4
Void
Rat
io
Moisture Ratio
Test 1 Test 4 Test 8 Test 14 Test 16 Test 19 Test 20
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4
Void
Rat
io
Moisture Ratio
Test 2 Test 3 Test 5
168
Figure 4-27 Variation of 𝛼∗ with wet-dry cycles
Sharma’s results were not carried out for full swelling and full shrinking as it was
undertaken in this thesis and by Tripathy et al. (2002). Hence, the typical behaviour of
the hydric coefficient is hard to observe. However, as for previous tests described, the
general trend of increasing close 1.0 during wetting and dropping to a small value under
drying passing through 1.0 or more is clear.
Another important observation one can make from Sharma’s data is that wetting always
causes a lesser change in void ratio leading to low hydric coefficient values. On the
other hand, drying causes a higher amount of void ratio change for a small moisture
reduction leading to higher hydric coefficients generally close 1.0. From this
observation, it may be concluded that a hydric coefficient close to 1.0 may be relevant
for shrinking from the compacted state.
0
0.2
0.4
0.6
0.8
1
1.2
Hyd
ric co
effic
ient
test 1
test 3
test 4
test 5
test 10
test 12
test 14
test 18
Wetting Wetting Wetting Drying Drying
Experimental Investigation of Shrinkage and Swelling Behaviour 169
Figure 4-28 Variation of 𝛼∗ with degree of saturation in each wetting or drying process
170
4.5.2 Experimental data analysis of Romero (1999)
The material tested is known as Boom Clay. Boom clay powder at 𝛾𝑑 = 10.8kN/m3 and
starting at an initial water content of 27.4% was controlled air dried to a final target of
15.4%. Then the samples were compacted statically to achieve the desired dry unit
weight (i.e. 16.7 or 13.7kN/m3). Other material properties of the soil are listed in Table
4-5. The static compaction curve for Boom clay with similar compaction energy to
Proctor compaction curve and the initial state of the soil specimens are shown in Figure
4-29.
Table 4-5 Summary of soil parameters
Parameter Value
Specific Gravity (𝐺𝑠) 2.7
Liquid limit 55.7±0.9%
Plasticity Index 26.9±1.0%
Soil Class (USCS, ASTM D2487) CH
De-structured dry unit weight 10.8 kN/m3
De-structured void ratio 1.46
Compressibility parameter for saturated conditions [λ(0)]
0.150 (𝛾𝑑 = 13.7 kN/m3) 0.136 (𝛾𝑑 = 16.7 kN/m3)
Compressibility parameter for unload/reload conditions [κ]
0.01
This research was mainly focused on the volumetric behaviour of unsaturated clays
(swelling, collapse, and shrinkage) under suction, stress and temperature changes. The
tests were performed in two types; oedometer and isotropic tests. In both methods,
isothermal wetting and drying cycles were carried out at a constant net vertical stress at
two different temperatures.
Experimental Investigation of Shrinkage and Swelling Behaviour 171
Figure 4-29 Compaction curves for Boom clay
For oedometer tests, specimens with a diameter of 50mm and a height of 10mm were
used for temperature controlled oedometer testing and these were compacted in a single
layer. The maximum pre-consolidation net vertical stress is 4.5±0.16MPa at 22°C and
4.07±0.12MPa at 80°C for the heavily over-consolidated sample. The maximum net
horizontal stress measured at ambient temperature is around 1.74±0.07MPa. For
isotropic tests, specimens 38mm in diameter and 76mm in height were statically
compacted in a stainless steel rigid mould in three layers using static one dimensional
compression.
In oedometer tests, the samples at two different temperatures (22°C and 80°C) were
loaded at approximately constant water content (around 15%) and at constant air
pressure (𝑢𝑎=0.5MPa) until the desired net vertical stresses were reached. The matric
13
14
15
16
17
18
19
4 9 14 19 24
Dry
Uni
t Wei
ght,
γ d (k
N/m
3 )
Water content (%) Standard Proctor energy: 0.59 compaction MJ/m3 - 22 C Standard Proctor energy: 0.59 compaction MJ/m3 - 80 C Wan (1996) Modified dynamic compaction test:2.69 MJ/m3 Sr = 100%
Initial Conditions
172
suction paths were applied systematically by maintaining a constant air pressure of
𝑢𝑎 =0.5MPa and controlling the water pressure. The following suction paths were
followed in the first wetting path: 0.45MPa, 0.2MPa, 0.06MPa and 0.01MPa.
Subsequently, the process was reversed following the same suction steps for drying. The
next cycles also followed the same suction path until the desired matric suction was
achieved.
Figure 4-30 Swelling shrinking curves tested under different pressures
In isotropic tests, the wet-dry cycles, under constant isotropic stress and temperature,
were applied by varying the matric suction. The samples were isotropically loaded at
constant water content until the desired net mean stresses were reached. The suction
paths were then imposed in steps similarly to the oedometer tests. All the tests analysed
in this section are summarised in Table 4-6.
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Void
ratio
moisture ratio
0.026 MPa 0.085 MPa 0.3 MPa 0.55 MPa 0.6 MPa 1.2 MPa I 0.085 MPa I 0.6 MPa
Experimental Investigation of Shrinkage and Swelling Behaviour 173
Table 4-6 Summary of tests by Romero
Test No.
Initial
Unit
Weigh
t (𝜸𝒅),
kN/m3
Tempe
rature
(°C)
Height/
Radius
Ratio
No.
of
days Test
Mean
net
stress
(MPa)
No. Of
Suction
changin
g steps
No of
cycles Suction (MPa)
Gradient of 𝒆 − 𝒆𝒘 curve
(𝜶∗)
1 2
wet dry wet dry
C17-
0.026C 16.7 22 0.4 74 Oedometer 0.026 11 1.5 2-0.01-0.45-0.01-0.3
0.547 0.473
H17-
0.026C 16.7 80 0.4 103 Oedometer 0.026 11 1.5 2-0.01-0.45-0.01-0.3
0.596 0.465
C17-
0.085A 16.7 22 0.4 160 Oedometer 0.085 16 2
2-0.01-0.45-0.01-0.45-
0.01 0.412 0.452 0.514
C17-
0.085B 16.7 22 0.57 105 Oedometer 0.085 12 2 2-0.01-0.45-0.01-0.2
0.408 0.373 0.345
H17-0.085 16.7 80 0.4 124 Oedometer 0.085 10 1.5 1.6-0.01-0.45-0.01 0.51 0.449 0.578
C17-0.300 16.7 22 0.4 89 Oedometer 0.3 10 1.5 2-0.01-0.45-0.01 0.254 0.314 0.358
H17-0.300 16.7 80 0.4 92 Oedometer 0.3 10 1.5 1.6-0.01-0.45-0.02 0.315 0.222 0.149
C17-0.550 16.7 22 0.4 36.5 Oedometer 0.55 12 2 2-0.01-0.45-0.01-0.45 0.094 0.033 0.089 0.12
174
1
H17-0.550 16.7 80 0.4 42 Oedometer 0.55 12 2 1.6-0.01-0.45-0.01-0.45 0.441 0.225 0.223 0.68
C14-
0.600B 13.7 22 0.4 66 Oedometer 0.6 11 2 2-0.01-0.45-0.01-0.2
0.548 0.108
C14-
0.600C 13.7 22 0.57 70 Oedometer 0.6 11 2 2-0.01-0.45-0.01-0.2
0.456 0.117
H14-
0.600D 13.7 80 0.4 73 Oedometer 0.6 11 2 1.6-0.01-0.45-0.01-0.2
0.628 0.104
C14-1.200 13.7 22 0.4 92.5 Oedometer 1.2 10 2 1.8-0.01-0.45-0.01 0.461 0.197
H14-
1.200B 13.7 80 0.4 68 Oedometer 1.2 10 1.5 1.6-0.01-0.45-0.01
0.308 0.051
I17-0.085A 16.7 99 Isotropic 0.085 12 2 2-0.01-0.45-0.01 0.606 0.653 0.599
I14-0.600 13.7 111 Isotropic 2-0.01-0.45-0.01 0.883 0.78
Experimental Investigation of Shrinkage and Swelling Behaviour 175
Figure 4-30 shows the void ratio variation with moisture content in different tests under
different vertical pressures. The constant pressure lines shown in the figure are
estimated based on two points calculated using a compressibility parameter. I 0.085MPa
and I 0.6MPa denote isotropic tests while the rest are oedometer tests. It can be
observed that three samples have undergone collapse at the pressure line corresponding
to the vertical stress (C14-0.600B, C14-0.600 and I14-0.600) in comparison with the
MPK framework. Others show a similar trend to Sharma’s data. The swelling curve
shows a flatter variation following a steeper shrinking curve. The average gradients of
each curve are shown in Figure 4-31.
Figure 4-31 Variation of 𝛼∗ with number of cycles
From Figure 4-31, it can be seen that, generally, the hydric coefficient is greater in
shrinking periods than in swelling periods, although there are variations from this trend
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1st Swelling 1st Shrinking 2nd Swelling 2nd Shrinking
Hyd
ric C
oeff
icie
nt
C17-0.085A
C17-0.085B
C17-0.300
H17-0.550
C14-0.600B
C14-0.600C
H14-0.600D
C14-1.200
H14-1.200B
I17-0.085A
I14-0.600
176
in some instances. In this data set the hydric coefficient is changing from 0.05 to 0.7 in
swelling periods and from 0.25 to 0.9 in shrinking periods.
4.5.3 Experimental data analysis of Tripathy (2000)
Two types of black cotton soils were used in these experiments, referred to as Soil A
and Soil B hereafter. The basic properties of the soil are shown in Table 4-7. The
samples were tested in an oedometer cell under controlled environment. The initial
positions of the samples used in this study are shown in Figure 4-32 with proctor
compaction curves for each soil.
Table 4-7 Soil Properties
Soil Property Soil A Soil B
Liquid limit (%) 100 74
Plasticity index (%) 58 42
Shrinkage limit (%) 10.6 13.5
Specific gravity 2.68 2.73
% passing sieve No. 200 98 80
Clay content (<0.002 mm; %) 62 52
Silt content (%) 36 28
Fine sand content (%) 2 20
Free swell (%) 340 225
During the tests, full swelling - full shrinking and full swelling – partial shrinking were
tested for several wetting drying cycles. The paths during swelling and shrinking were
plotted in void ratio – moisture content space. The tests were conducted under different
Experimental Investigation of Shrinkage and Swelling Behaviour 177
vertical net stresses, 6.25kPa and 50kPa. The swelling and shrinking paths for different
tests are shown in Figure 4-33.
It can be that when the vertical pressure is less the soil shows substantial or even wild
expansive/shrinking behaviour during first 2 cycles and then gradually comes to an
equilibrate state. All 4 tests, given in Figure 4-33 show similar behaviour after several
wet dry cycles.
Figure 4-32 Standard Proctor curves for Soil A and Soil B
The variation of hydric coefficient with moisture content in each wet or dry process is
shown in Figure 4-34 for soil A tested under 6.25kPa vertical net stress. This shows that
the variation of hydric coefficient is almost similar in every process except in 1st wetting
process and 2nd wetting process.
11
12
13
14
15
16
17
18
10 20 30 40 50
Dry
Den
sity
(kN
/m3)
Moisture Content (%)
Soil A
11
12
13
14
15
16
17
18
10 20 30 40 50
Dry
Den
sity
(kN
/m3)
Moisture Content (%)
Soil B
178
Table 4-8 Summary hydric coefficients of tests by Tripathy
Test Soil A with 6.25kPa
Soil A with 6.25kPa – Partial Shrinkage
Soil A with 50kPa
Soil A with 100kPa
Soil B with 6.25kPa
Soil B with 50kPa
Cycle 1 wetting 0.946 0.943 0.913 0.858
drying 1.002 1.011 0.956 0.879
Cycle 2 wetting 2.429 0.954 1.091 1.696
drying 0.997 0.994 0.877 0.840
Cycle 3 wetting 1.037 0.949 0.951 0.927
drying 0.908 0.973 0.927 0.876
Cycle 4 wetting 0.957 0.952 0.953 0.989
drying 0.957 0.952 0.953 0.989
Cycle 5 wetting 0.979 0.938 0.958 0.936 0.859
drying 0.979 0.938 0.958 0.936 0.859
The variation of hydric coefficient in tests conducted with soil A under 6.25kPa, soil A
under 50kPa and Soil B under 6.25kPa for several cycles are shown in Figure 4-35. The
first wetting and second wetting shows somewhat abnormal behaviour compared to the
rest of the tests. Furthermore, it can be observed that for low vertical net stress,
abnormality of the expansive behaviour increases in the initial wetting cycles. Another
observation is that when the vertical stress is low, the number of cycles required for soil
to stabilize increases. However after the 2nd cycle soil becomes mostly stable.
Experimental Investigation of Shrinkage and Swelling Behaviour 179
Figure 4-33 Swelling and shrinking curves for several wet-dry cycles
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0.0 0.5 1.0 1.5 2.0
Void
Rat
io
Moisture Ratio
(a) Soil A under 6.25kPa
Cycle 1 Cycle 2 Cycle 3 Cycle 4
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Void
Rat
io
Moisture Ratio
(b)Soil A under 50kPa
Cycle 1 Cycle 2 Cycle 3 Cycle 4
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
0.8 1 1.2 1.4 1.6
Void
Rat
io
Moisture Ratio
(c) Soil A under 6.25kPa - Partial Shrinkage
Cycle 1 Cycle 2 Cycle 3 Cycle 4
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Void
Rat
io
Moisture Ratio
(d) Soil B under 6.25kPa
Cycle 1 Cycle 2 Cycle 3 Cycle 4
180
Figure 4-34 Hydric coefficient change with wetting and drying for the Soil A with 6.25kPa
Figure 4-35 Hydric coefficient variation during several wet-dry cycles
The hydric coefficient was closely examined in each process for all the tests as shown in
Figure 4-36. All tests show a similar variation for the hydric coefficient α∗, except for
the partial drying test. This varies from zero (at lower and higher degrees of saturation)
to one at the intermediate saturation levels, which is significant, after stabilization.
However the degree of saturation corresponding to the same value of α∗ is considerably
different for different soil types and different vertical stresses (Figure 4-36).
0
0.5
1
1.5
2
2.5
3
3.5
0 10 20 30 40 50 60 70 80
Hyd
ric C
oeff
icie
nt
Moisture Content
Cycle 1-wetting Cycle 2-wetting Cycle 1-drying Cycle 2-drying Cycle 3-wetting Cycle 3-drying Cycle 4-wetting Cycle 4-drying
0
0.5
1
1.5
2
2.5
3
3.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Hyd
ric C
oeff
icie
nt
No of Cycles
A 6.25 A 50 B 6.25
Experimental Investigation of Shrinkage and Swelling Behaviour 181
Figure 4-36 Summary of variation of hydric coefficient in each wetting or drying process
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.5 1 1.5
Hyd
ric C
oeff
icie
nt
Degree of saturation
First Wetting A 6.25 A 6.25 half A 50 B 6.25
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.5 1 1.5
Hyd
ric C
oeff
icie
nt
Degree of saturation
First Drying A 6.25 A 6.25 half A 50 B 6.25
0
0.5
1
1.5
2
2.5
3
3.5
0 0.5 1 1.5
Hyd
ric C
oeff
icie
nt
Degree of saturation
Second Wetting A 6.25 A 6.25 half A 50 B 6.25
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.5 1 1.5
Hyd
ric C
oeff
icie
nt
Degree of saturation
Second Drying A 6.25 A 6.25 half A 50 B 6.25
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.5 1 1.5
Hyd
ric C
oeff
icie
nt
Degree of saturation
Third Wetting A 6.25 A 6.25 half A 50 B 6.25
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.5 1 1.5
Hyd
ric C
oeff
icie
nt
Degree of saturation
Third Wetting A 6.25 A 6.25 half A 50 B 6.25
0 0.2 0.4 0.6 0.8
1 1.2 1.4
0 0.2 0.4 0.6 0.8 1 1.2
Hyd
ric C
oeff
icie
nt
Degree of saturation
Forth Wetting & Drying A 6.25 A 6.25 half A 50 B 6.25
182
4.5.4 Experimental data analysis of Montanez (2002)
In this study, sand bentonite mixes were used to analyse the expansive behaviour of soil.
Well and uniformly graded sand was mixed with different proportions of bentonite. The
soil properties are presented in Table 4-9.
Table 4-9 Soil Parameters for sand bentonite mixes
Well-graded sand Uniform sand Bentonite
Bentonite (%) 0 5 10 15 0 10 100
Specific gravity 2.655 2.657 2.658 2.660 2.657 2.660 2.688
Liquid Limit (%) 69 110 150 71 337
Plasticity Index (%)
46 88 122 45 286
Classification (BS) SPg SW SW SPC CE
Samples were prepared by the static compaction technique. Soil was compacted in a
larger mould and the samples were then cut out in required the size using a cutting ring
which was finally transferred into the oedometer ring. The initial conditions of the
samples are shown in Figure 4-37.
A sample was then subjected to several wet dry cycles in order to evaluate the
behaviour of several properties including void ratio and water content with suction. The
tests were conducted with different initial moisture contents, dry densities, bentonite
percentages and sand types. The summary of the tests is presented in Table 4-10.
Experimental Investigation of Shrinkage and Swelling Behaviour 183
Figure 4-37 Compaction curves for samples with different compositions; WG – Well Graded U – Uniform and B – Bentonite
The results obtained from these experiments are shown in Figure 4-38. Figure 4-38(a)
shows the behaviour with different initial moisture contents. The trends are not very
clear although, when the moisture content is less, it is apparent that soil reaches its
stable position more quickly than the soils compacted to higher moisture contents.
Considering the tests analysed previously, it may be that the soil shows erratic
1
2
3
4
5
6
1.5
1.6
1.7
1.8
1.9
2
2.1
2.2
2 7 12 17 22
Dry
Den
sity
(Mg/
m3 )
Moisture Content (%)
WG sand with 10% B
Standard
7
1.6
1.7
1.8
1.9
2
2.1
2.2
2 7 12 17
Dry
Den
sity
(Mg/
m3 )
Moisture Content (%)
WG sand with 5% B
Standard Heavy
8
1.6
1.7
1.8
1.9
2
2.1
2.2
2 7 12 17
Dry
Den
sity
(Mg/
m3 )
Moisture Content (%)
WG sand with 15% B
Standard Heavy
9
1.4
1.5
1.6
1.7
1.8
1.9
2 12 22
Dry
Den
sity
(Mg/
m3 )
Moisture Content (%)
U sand with 10% B
Standard Heavy
184
behaviour during the first cycle when the initial moisture content is low. However, it is
hard to explain the behaviour using one cycle only.
Table 4-10 Summary of the tests by Montanez
Sample No.
initial 𝒘
Initial dry
density (Mg/m3)
Sand type
% of Bentonite
Sample type
𝜶∗ value
1st drying
1st wetting
2nd drying
1 5.94 1.82 WG 10 c 0 0.441
2 11.95 1.85 WG 10 c 0.032 0.469 0.413
3 12.11 1.76 WG 10 c 0.032 0.416 0.446
4 12.14 1.66 WG 10 c 0.028 0.328 0.312
5 13.97 1.88 WG 10 c 0.022 0.37
6 22 1.66 WG 10 r 0.074 0.187 0.146
7 11.74 1.78 WG 5 c 0.083 0.149 0.133
8 11.41 1.79 WG 15 c 0.013 0.506 0.441
9 12.15 1.49 U 10 c 0.008 0.331 0.328
c = compacted r = reconstituted WG = well-graded sand U = Uniform sand
Figure 4-38(b) shows the behaviour of samples with different initial dry densities. It can
be clearly seen that the soil is moving towards a stable condition during cycling. The
curves for each wetting and drying do not show much difference when the dry densities
are lower. This may be due to the fact that soil structure does not need to rearrange
significantly if the soil has not been compressed heavily at the beginning. Figures 4-38
(c) and (d) show the effect of composition, indicating the commonly known fact that the
more reactive fines are in the soil, the higher the swelling and shrinking effect are.
Experimental Investigation of Shrinkage and Swelling Behaviour 185
Some soils show that the shrinking curve is above the swelling curve, which is different
to other results examined so far. This behaviour can be expected when the initial
condition of the soil is below its equilibrium position.
Figure 4-38 Behaviour of sand samples with different initial conditions under wetting and drying
According to the results presented in Figure 4-38, the gradients obtained during the first
drying (the tests were started by drying the sample) are much lower than seen in the
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
Void
Rat
io
Moisture Ratio
(a) Different initial moisture contents
6%, Sample 1 11.50%, Sample 2 14%, Sample 5 22 %, Sample 6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 0.2 0.4 0.6 0.8
Void
Rat
io
Moisture ratio
(b) Different initial dry densities
1.65 g/cm3, Sample 4 1.75 g/cm3, Sample 3 1.85 g/cm3, Sample 5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 0.2 0.4 0.6 0.8
Void
ratio
Moisture Ratio
(c) Different Bentonite contents
B 5%, Sample 7 B 10%, Sample 2 B 15%, Sample 8
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
Void
Rat
io
Moisture Ratio
(d) Different sand Types with 10% B
WG Sand, Sample 4
186
previous results. Also, the wetting stage shows higher 𝛼∗ values than the drying stage.
The values of the hydric coefficient were computed and then plotted in Figure 4-39. It
can be seen that the hydric coefficient is generally lower than in previous tests,
presumably due to sand providing a different form of resistance to swelling and
shrinking.
Figure 4-39 Hydric coefficient values for wet-dry processes
4.5.5 Experimental data analysis of Monroy (2006)
The material chosen for this study consisted of brown, weathered London Clay,
obtained from Harlesden, in north London. The basic properties of the soil are listed in
Table 4-11. The compaction characteristics of the soil are shown in Figure 4-40.
0
0.1
0.2
0.3
0.4
0.5
0.6
1st drying 1st wetting 2nd drying
Hyd
ric C
oeff
icie
nt
1
2
3
4
5
6
7
8
9
Experimental Investigation of Shrinkage and Swelling Behaviour 187
Table 4-11 Properties of London Clay
Parameter Value
Specific Gravity (𝐺𝑠) 2.7
Liquid limit 85%
Plasticity Index 60%
Soil Class (BS5930: 1981) CV (very high plasticity clay)
Clay fraction 58%
Silt fraction 40% Sand fraction 2%
Compressibility parameter for saturated conditions [Cc(0)]
0.349
Compressibility parameter for unload/reload conditions [Cs(0)]
0.081
The soil was compacted statically into the oedometer ring under increasing load. The
maximum load applied to compact the sample was approximately 720kPa. The initial
condition of the sample is marked in Figure 4-40. The suction was changed in order to
examine the behaviour during wetting and drying. Loading and unloading stages were
also carried out in the oedometer, and the vertical pressure was kept at a constant value
during these tests. However, in some tests the vertical stress has been changed. A
complete summary of the tests used in this study is given in Table 4-12.
188
Figure 4-40 Compaction characteristics of London clay from standard compaction test
1.36
1.4
1.44
1.48
1.52
1.56
1.6
1.64
14 16 18 20 22 24 26 28 30 32 34
Dry
Den
sity
(g/c
m3)
Moisture Content (%)
Initial Condition of the sample
Experimental Investigation of Shrinkage and Swelling Behaviour 189
Table 4-12 Summary of tests by Monroy
Test Initial 𝒆
Initial dry
density (Mg/m3)
Initial 𝑺𝒓 (%)
No of
days
Test path Suction (kPa) Vertical net stress
(kPa)
Hydric Coefficient 𝜶∗
Cycle 1 Cycle 2 Cycle 3
wet dry wet dry wet dry
o13 0.942 1.39 67.6 93 wetting 840 - 405 7 - 100 -0.096
Loading
Unloading
405
405 - 430
100 – 630
630 - 100
Wetting (V Constant) 430 - 115 100 - 270 0.076
Loading
Unloading
115
115 - 130
270 – 600
600 - 415
wetting 130 - 0 415 0.106
Loading 0 415 - 700
o14 0.947 1.387 67.4 51 wetting 940 - 430 7 -
Loading 430 7 - 600
Unloading 430 - 445 600 - 220
190
wetting 445 - 0 220 0.058
Drying 0 - 340 220 0.349
Wetting 340 - 0 220 0.184
o15 0.941 1.391 66.7 146 Loading 720 - 610 105 - 615
Unloading 610 - 630 615 - 105
Wetting 630 - 0 105 0.376
Drying 0 - 400 105 0.332
Loading 400 - 340 105 - 585
o16 0.945 1.388 67.7 43 Wetting 870 - 415 7 -
Loading 415 7 - 220
Wetting 415 - 0 220 -0.371
Drying 0 - 300 220 0.472
Wetting 300 - 0 220 0.211
o21 0.947 1.387 68.2 73 Wetting 1010 - 620 7 - 65 0.095
Wetting 620 - 0 65 0.164
Experimental Investigation of Shrinkage and Swelling Behaviour 191
Drying 0 - 120 65 0.216
Wetting 120 - 0 65 0.115
Drying 0 - 120 65 0.105
Loading 120 - 85 65 - 590
Unloading 85 - 120 590 - 180
o25 0.962 1.376 69.8 80 Wetting 820 - 0 7 0.415
Loading 0 7-30
Drying 0 - 100 30 0.385
Wetting 100 - 0 30 0.068
Drying 0 - 100 30 0.164
Wetting 100 - 0 30 0.107
Drying 0 - 100 30 0.127
o26 0.953 1.382 68.7 102 wetting 810 - 0 7 -
Loading 0 7 - 30
Drying 0 - 50 30 -
192
Loading 50 30 - 65
Drying 50 - 90 65 -
Loading 90 65 - 205
wetting 90 - 0 205 0.217
Drying 0 - 100 205 0.143
wetting 100 - 0 205 0.597
Drying 0 - 100 205 0.639
o27 0.956 1.38 68.4 89 wetting 870 - 0 7 0.428
Loading 0 7 - 30
Drying 0 - 80 30 0.249
Loading 80 30 - 425
Experimental Investigation of Shrinkage and Swelling Behaviour 193
The behaviour of soil during the tests is plotted in the void ratio and moisture ratio
space as shown in Figure 4-41. The figure shows a complete path in each test including
the loading and unloading cycles. Constant pressure lines shown in the figure were
estimated using the standard proctor curve and compressibility parameter for the
saturated condition. The gradient of the wetting portion gives lower values than the
drying portion of the curves. This observation is in agreement with the observations
from most of the other results presented earlier. Some samples (o16, o26) showed
collapse when they reached the corresponding pressure line during wetting, which is in
line with the MPK framework suggested by Kodikara (2012).
Figure 4-41 Variation of void ratio with moisture ratio during wetting and drying
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Void
Rat
io
Moisture Ratio
o13 o14 o15 o16 o21 o25 o26 o27
194
The gradients of wetting and drying curves (𝛼∗) in Figure 4-41 are plotted in Figure
4-42. The values corresponding to wetting are lower than the values corresponding to
the drying that followed.
Figure 4-42 Variation of 𝛼∗ values with different wetting and drying paths
4.6 Summary and Discussion
The present study and Tripathy’s study considered the full swell shrink cycles for
several cycles. The other studies involved conducting one cycle of wetting and drying
and in most cases, the cycles are not full drying or full wetting. Most of the experiments
were carried out in isotropic conditions and suction was controlled to change the
moisture content. The current research, Tripathy’s research and Montanez’s research
have been selected initial conditions of the sample below its equilibrium curve (at a
certain moisture ratio) while other research selected a position above the likely
equilibrium position.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1st wetting 1st drying 2nd wetting 2nd drying 3rd wetting 3rd drying
Hyd
ric C
oeff
icie
nt
o13
o14
o15
o16
o21
o25
o26
Experimental Investigation of Shrinkage and Swelling Behaviour 195
The suction (𝑆 = 𝑢𝑎 − 𝑢𝑤) of the sample is controlled by changing the pore water
pressure using the axis translation technique. The basic principle of the axis translation
technique is to elevate the air pressure, while keeping the water pressure zero or
positive. By changing the pore water pressure from a negative to a positive value, the
possibility of cavitation is prevented, not only in the measuring system, which was the
intention of changing pore water pressure, but also within the soil pores. This implies
that any influence of cavitation of the pore water which may be an important
phenomenon under in-situ stress conditions, is not accounted for in the laboratory tests
using the axis translation technique (Sharma, 1998).
Controlling suction by changing (increasing and decreasing) using same value for
several cycles may not represent the moisture content also change (decrease and
increase) in constant value for every cycle as expected. Figure 4-43 shows the moisture
content variation in a suction controlled test given by Romero (1999). The moisture
content changes considerably for the same suction in various cycles due to hysteresis.
Figure 4-43 Moisture content variation in a suction controlled test
0
0.1
0.2
0.3
0.4
0.5
0.51 0.53 0.55 0.57 0.59 0.61 0.63
Suct
ion
(Mpa
)
Moisture Ratio
196
The summary of the hydric coefficient values is presented in Table 4-13. From these
values it is obvious that 𝛼∗ values very much depend on the initial conditions of the soil.
Further results of 𝛼∗ variation during partial wet-dry cycles are given in Figure 4-44.
Full wet-dry cycles show similar 𝛼∗ values close to 1.0 during both wetting and drying
when the moisture ratio range is approximately 0.3 to 1.0. At lower moisture ratio
values, 𝛼∗ value decreases gradually and becomes zero. Similar variation can be seen at
higher moisture ratio values more than 1.0.
Experimental Investigation of Shrinkage and Swelling Behaviour 197
Table 4-13 Summary of hydric coefficient values
Cycle 1 Cycle 2 Cycle 3 Cycle 4 Cycle 5
wetting drying wetting drying wetting drying wetting /drying wetting/drying
𝛼∗ range Current 0 - 1.15 0 - 1.7 0.15 - 1.04 0.45 - 0.9 0.6 - 0.84 0.55 - 0.84 0.6 - 1.05
Sharma 0 – 0.91 0.5 – 1.04 0.34 – 0.57 0.64 – 0.87 0.48
Romero 0.09 – 0.6 0.22 – 0.88 0.05 – 0.78 0.12 – 0.68
Tripathy 0.85 – 0.95 0.88 – 1.01 0.95 - 2.43 0.84 – 1.00 0.93 - 1.04 0.88 – 0.97 0.95 – 0.99 0.86 – 0.98
Montanez 0.01 -0.08 0.149 – 0.51 0.15 – 0.45
Monroy 0.06 – 0.43 0.14 – 0.47 0.07 – 0.6 0.11 – 0.64 0.11 – 0.11 0.13
Overall 0 – 1.15 0 - 1.7 0.05 – 2.43 0.11 – 1.00 0.11 – 1.04 0.13 – 0.97 0.6 – 1.05 0.86 – 0.98
Average 𝛼∗
Current 1.00 1.30 0.95 0.90 0.83 0.83 0.97 0.97
Sharma 0.39 0.75 0.45 0.75 0.48
Romero 0.42 0.43 0.30 0.40
Tripathy 0.92 0.96 1.54 0.93 0.97 0.92 0.96 0.93
Montanez 0.03 0.36 0.32
Monroy 0.21 0.31 0.21 0.30 0.11 0.13
Overall 0.59 0.63 0.63 0.60 0.60 0.63 0.96 0.95
Experimental Investigation of Shrinkage and Swelling Behaviour 198
As discussed before, the initial conditions highly relate to the values of 𝛼∗. Figure 6.45
shows a conceptual understanding of how the soils approach the stable condition when
subjected to several partial wetting drying cycles. It is also apparent 𝛼∗ depends on the
net stress level.
Figure 4-44 Typical paths of expansive soils subjecting to partial wet dry cycles
The wetting path is generally curved, which starts flatter and gets steeper in the middle
then can get flatter towards the end especially if the saturations are very high. When the
initial conditions are above the equilibrium line, 𝛼∗ values in the drying path are almost
Moisture Ratio
Void Ratio
Typical path a soil can follow, when the initial conditions of the soil lie below the stable curve and are subjected to partial wetting and drying (e.g. Figure 4-33, Figure 4-38)
Stable curve corresponding to the stress level at which drying and wetting are
Typical path a soil can follow, when the initial conditions of the soil lie above the stable curve and are subjected to partial wetting and drying (e.g. Figure 4-25, Figure 4-26, Figure 4-30, Figure 4-41)
Experimental Investigation of Shrinkage and Swelling Behaviour 199
1.0 and 𝛼∗ values at wetting paths generally less. This makes the soil approach the
stable curve as cycles progress. However, when the initial conditions are selected below
the equilibrium line, the 𝛼∗ values during wetting and drying approach the stable curve
from behind as illustrated in Figure 4-44.
4.7 Conclusions
Modelling the behaviour of partly saturated soils during cycles of wetting and drying,
especially those containing a significant amount of reactive clay minerals, seems to pose
a high level of difficulty. Behaviour of soil depends on many factors, such as initial
moisture content, initial dry density, net vertical stress, composition of the soil and
degree of wetting and drying processes.
Despite all these restrictions, any expansive soil appears to come to an equilibrium
condition where the soil behaves reversibly while wetting and drying. The path
followed by the soil to reach the equilibrium curve depends on the factors listed in the
previous paragraph. Typical paths followed by expansive soils are shown in Figure 4-44
when they are subjected to partial wetting and drying.
The first two cycles of the wetting and drying process can be highly unpredictable. Most
soils show an erratic behaviour when adjusting the soil structure generated by
compaction to initial wet/dry cycles, particularly if the soil is heavily compacted
initially and the vertical stress on the soil during cycling is small.
200
The hydric coefficient is less when the soil swells and comparatively high when the soil
shrinks, if the soil is above the equilibrium line. If it is below, the opposite can happen
where the hydric coefficient is less during drying and high during wetting.
For full wetting drying cycles, typical variations of 𝛼∗ for several cycles are shown in
Figure 4-45. After the first few cycles, α∗ starts from zero and then increases up to one
or closer to one, and remains at this value for a significant amount of wetting or drying.
At the extreme of wetting or drying, the value will drop close to zero and the cycle
continues as shown in Figure 4-45. The first few cycles can show a deviation from this
behaviour when the soil structure is significantly adjusting to the drying wetting
environment.
Figure 4-45 Typical variations of 𝛼∗ with full wet dry cycles
𝛼∗ Value
1st Wetting
2nd Wetting
3rd Wetting
1st Drying
2nd Drying
3rd Drying
1
Chapter 5
MODELLING OF STABLE DESICCATION CRACK DEPTHS DURING CYCLIC WETTING AND DRYING
5.1 Introduction
As discussed in Chapter 2, seasonal movements of soil can be observed in response to
the change of soil moisture content resulting in a change of suction stresses. It was also
highlighted in Chapter 4 that this soil shrinking and swelling becomes stable with the
aging of soil. After the soil becomes stabilised, it behaves in reversibly during wetting
and drying (Tripathy et al., 2002; Wijesooriya and Kodikara, 2012). Mitchell (2012)
provided models to predict the suction profiles below the ground surface after this stable
condition had been reached. The standard values of suctions to be used under such
conditions generally were also given in the Australian Standards (Residential slabs and
footings, (Shannon, 2012)) for design purposes.
Significant differences in suction profiles can be observed depending on the water table
depth. With the use of the suction profile proposed by Mitchell (2012), the stable crack
201
202
depths can be calculated using elastic theory. In this chapter crack depths were
calculated using the above method for several clay soils from different parts of Australia
and from some other countries. The evaluation of crack depth in different climate
conditions and initial placement conditions of the clay layer was considered.
5.2 Suction Profiles
The suction variation below the ground has been measured by several researchers and
presented as the observed suction profiles during different climate conditions (Corte and
Higashi, 1964; Richards, 1985; Morris et al., 1992; Wijesooriya and Kodikara, 2012).
The position of the water table was considered as a key factor in each of these works
and the depth of soil from the surface was also identified as critical when the suction
profile changes significantly according to seasonal effects. The water table was
considered as shallow when the water table is less than 6m deep in clay soil, 3m in
sandy clay and silts and 1m in sand (Corte and Higashi, 1964). Otherwise it was
considered as a deep water table.
When shallow water table depths were presented, several appropriate suction profiles
were presented by Morris et al. (1992) as discussed in Chapter 2 (section 2.3.4.2) and
Chapter 3 (section 3.2). However, the depth to water table could increase when the soil
layer is subjected to arid climatic conditions.
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 203
Figure 5-1 Effect of various environmental conditions on the matrix suction profile (Peter 1979)
Peter (1964) suggested that the equilibrium suction profile should be given by the
extension of the hydrostatic water table, regardless of the climate, where shallow water
tables exist. However, it is well known that the suction profile fluctuates seasonally and
during wet and dry conditions. The effect of the position of the water table and various
environmental factors on the matric suction profile is presented in Figure 5-1. The
broken line in the figure shows the equilibrium suction profile and the solid line gives
the observed actual suction profile.
When deep water tables are present, the moisture conditions within the soil are
controlled by the moisture balance between rainfall and evapotranspiration (Russam and
Coleman, 1961; Richards, 1985). Richards (1985) predicted equilibrium suction profiles
using empirical relations between soil suction and climatic indices such as
Thornthwaite’s Moisture Index (TMI) in terms of total suction. These suction profiles
204
are based on the curve proposed by Russam and Coleman (1961) as shown in Figure 5-2
for optimum drainage conditions considering the practical purposes.
Perera et al. (2004) later modified the equilibrium suction prediction model for different
soils based on the particle size and plasticity index, especially with reference to
pavement layers. Although these models indicated great predictive capability based on
the error analysis, the Russam and Coleman (1961) model was used in this analysis
since it was more applicable in Australian conditions and has direct reference to clay
layers.
Figure 5-2 Values of suction from road site installations and postulated design curves after Richards, B.G. (1985)
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 205
A more detailed theoretical model for predicting the suction profile below the ground
has been developed by Mitchell (2012). In his model, the suction variation with depth
was represented according to the soil characteristics and the time. The model given in
equation [5-1] was derived by solving the moisture diffusion equation (equation [5-2])
as the solution of a linear homogeneous equation of the fourth order (equation [5-3]).
𝑆(𝑧, 𝑡) = 𝑆𝑒 + ∆𝑆𝑒−�
𝑛𝜋𝛼𝑑𝑐
𝑧𝑐𝑜𝑠 �2𝑛𝜋𝑡 − �
𝑛𝜋𝛼𝑑𝑐
𝑧� [5-1]
where 𝑆(𝑧, 𝑡) is the suction at any depth 𝑧 in metres at the time of 𝑡 in years, 𝑆𝑒 is the
equilibrium suction below the depth of seasonal suction change (or reactive zone depth),
∆𝑆 is the amplitude of suction variation at the surface. All the suction values have the
unit pF, which is defined as pF=log10 |suction in cm of water|. 𝑛 is the frequency of
seasonal variations given by cycles per year, 𝛼𝑑𝑐 is the diffusion co-efficient of soil with
the units of m2 per year.
𝜕𝑆𝜕𝑡
= 𝛼𝑑𝑐𝜕2𝑆𝜕𝑧2
[5-2]
The empirical relationship to calculate the diffusion co-efficient was proposed by
Lytton (1994) for expansive soils. This empirical equation is given by,
𝛼𝑑𝑐 = 0.0029 − 0.000162𝑆𝑠 − 0.0122(𝑎𝑣) [5-3]
where 𝛼𝑑𝑐 is given in cm2/s, 𝑆𝑠 is the slope of the suction water content curve and 𝑎𝑣 is
the matric suction compression index (slope of the void ratio-suction curve). The value
of 𝑆𝑠 can be obtained from equation [5-4] (Lytton, 1994) using the Atterberg limits of
the soil.
206
𝑆𝑠 = −20.29 + 0.1555(𝐿𝐿%) − 0.117(𝑃𝐼%) + 0.0684(% 𝑁𝑜. 200) [5-4]
where 𝐿𝐿% is the percentage of the liquid limit, 𝑃𝐼% is the percentage of the plasticity
index and % 𝑁𝑜. 200 is the percentage of soil passing through the US No. 200 sieve
equivalent to 75μm mesh size.
Equation [5-1] defines the suction decay along the depth and its oscillation with time
which is symmetric about the equilibrium suction 𝑆𝑒. Although this model covers the
seasonal effects on the suction profile and the variation of suction with the depth it is
based on some unrealistic assumptions. Basically, it assumes that the suction varies
purely due to the climatic conditions and the moisture with in the soil is neglected,
which is not accurate if the water table is not deeper than 10m (Aubeny and Long,
2007). Furthermore, at or near the surface the model (equation [5-1]) is valid only if the
soil layer has a very shallow root depth or bare surface which is highly unlikely.
However due to the usefulness in modelling the seasonal variation of suction profile
with depth as represented by equation [5-1], this model was selected as the basis for the
numerical model.
5.3 The Suction in Different Climate Conditions
Despite all the different empirical, theoretical and numerical models developed to
predict the suction below ground, it is essential to measure the actual suction profile in
the field under natural conditions. These real observations should be used as a
benchmark to validate the other developed models. Hence the literature provides some
useful data on observed suction profiles in different areas of the world.
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 207
Peter (1964) has observed the matric and total suction profiles throughout the different
parts of Australia giving especial consideration to Adelaide. He reported that in semi-
arid climates, the matric suction varies from 50kPa to 200kPa in different locations of
Adelaide. In desert environments such as in Woomera, he reported matric suctions of
1.5MPa to 2.0MPa. However, under extreme conditions, the total suction was measured
as high as 15.0MPa.
According to the chart provided by Richards (1985) as shown in Figure 5-2, suctions
can be estimated on the basis of a climatic index. Considering that Adelaide, Horsham
and Bordertown are in semi arid climates, the matric suctions at approximately 0.5m
below the ground can be obtained from the Russam and Coleman curve in Figure 5-2.
The values are between 100kPa (3pF) and 2000kPa (4.3pF) when the climatic index is
considered to change from 20 to -20 for semi arid conditions. Furthermore, Richards
observed more than 10MPa at the surface at a road site in Horsham, Victoria although
the area was considered a semi-arid area. Assuming arid conditions occur when the TMI
is less than -20, then matric suctions below 0.5m are observed greater than 2MPa. In
deserts during the dry season the exposed surface soil displayed suctions from 9.8MPa
to 98MPa and during wet season suctions from 4MPa to 12.4MPa (Russam and
Coleman, 1961).
Mitchell (1979) used surface suction value of 30MPa during the drying season and
1MPa during the wetting season in semi-arid climates. The difference between the
equilibrium suction and the surface value was selected as 1.5pF. However, Australian
Standards for slabs and footings (AS2870, 2011) suggest a value of 1.2pF for the
change of suction with depth for every part of Australia for design purposes.
208
Considering the observed suctions in the past, it was decided to select a matric suction
variation from 10kPa to 2.5MPa (2.0pF to 4.4pF, with 1.2pF difference to equilibrium
suction) from wet season to dry season in semi-arid climate conditions. Also, in arid
climate conditions the suction was changed from 20kPa to 5MPa (2.3pF to 4.7pF, with
1.2pF difference to equilibrium suction) from the wet season to dry season. The suction
profiles are illustrated in Figure 5-3 for arid and semi-arid conditions.
Figure 5-3 Typical Suction Profile for arid and semi-arid conditions used in the present study
5.4 Development of Numerical Model for Compacted Clay Layers under Cyclic Atmospheric Conditions
A numerical model was developed similar to that in Chapter 3 except that the suction
profile is different. Furthermore, the stress change was calculated on the basis of water
0
1
2
3
4
5
6
7
8
1 2 3 4 5
Dep
th (m
)
Suction (pF)
Semi-arid - wettest profile Semi-arid - driest profile Arid - wettest profile Arid - driest profile
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 209
content change instead of the suction change. Initially three methods were considered to
achieve the applicable stress change with the change in moisture content:
• Stress change by the direct addition of suction difference (same as Chapter 3);
• Fraction of suction added to change the stress to allow for desaturation; and
• Stress change on the basis of moisture content change.
The procedure of implementing stress change in each method is discussed in the
following section.
5.4.1 Stress change by the direct addition of suction change
The suction values were calculated using equation [5-1] corresponding to the midpoint
depth in each zone in the problem model geometry then the value obtained was applied
as the tensile stress of that zone. This application of stress was based on the assumption
that the degree of saturation will remain high close to the tip of the propagating crack,
and that the soil will behave similarly whether it is subjected to an externally applied
stress change or the internally developed suctions. This was discussed in more detail in
the Chapter 3.
However, this condition is not applicable when the soil is subjected to higher suctions as
soil gets drier, although heavy clays can remain saturated to suctions of 1MPa or so.
Therefore, the applied suctions will not be correct especially under arid conditions. The
results from this section are shown in Figure 5-4 in comparison to the other methods of
suction application. When this approach is followed, 8m crack depth was observed for
2MPa surface suction value, which represented semi arid conditions. Morris et al.
210
(1992) and some other researchers (Table 2-1, Chapter 2) have indicated that typically
observed crack depths in semi arid condition do not exceed 4m. Therefore, deeper crack
depths of 8m are unlikely in semi arid climates. Accordingly, the use of this assumption
is less appropriate to predict the crack depths for arid conditions.
Figure 5-4 Predicted crack depths using different stress changing approaches for Regina clay soil
5.4.2 Fraction of suction on change of the stress
As the suction increases beyond the air entry value the soil may significantly desaturate
and the 𝐸 𝐻⁄ ratio will fall below the value (1 − 2𝜈). Lau (1987) has investigated the
𝐸 𝐻⁄ ratio experimentally for Regina clay soil and he found that at low suction 𝐸 𝐻⁄ is
not a constant but at higher applied stresses, it is almost constant. He suggested that the
0
1
2
3
4
5
6
7
8
0 2 4 6 8 10 12
Dep
th o
f cra
ckin
g (m
)
Maximum surface suction (MPa)
Theoretical value Directly added suction Fraction of suction Stress due to W change
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 211
normal range of the 𝐸 𝐻⁄ should vary between 0 and 0.2, with the latter corresponding
to near saturated condition.
The relation between the water content change and the suction change can be written as,
1(1 − 𝜐) 𝐸𝛼∆𝑊 =
𝐸𝐻
1(1 − 𝜐) Δ𝑆 [5-5]
Δ𝜎 =𝐸𝛼∆𝑊
(1 − 2𝜈) =𝐸𝐻
Δ𝑆(1 − 2ν)
Hence, the stress change in one direction in UDEC, Δ𝜎, can be written as:
Δ𝜎 =𝐸
3𝐻(1 − 2𝜈) Δ𝑆 [5-6]
The values for 𝐸 𝐻⁄ ratio were used as suggested by Lau in his experimental
investigation using the data from Fredlund (1964). The typical 𝐸 𝐻⁄ value for Regina
clay with 0.4 Poisson’s ratio was taken as 0.11. Then the stresses for the UDEC model
(Δ𝜎) were calculated using equation [5-7].
The results obtained through the above method were shown in Figure 5-4. As expected,
the compacted crack depths are smaller than those obtained by using direct suction
change as the stress change and show a better correlation with the theoretical crack
depth values shown in Figure 5-4. However, the 𝐸 𝐻⁄ ratio can vary with the soil type
and desaturation. Hence, it was decided to select another method that can predict crack
depths more rationally.
212
5.4.3 Stress change on the basis of moisture content change
Both methods described earlier used stress change through the suction change.
However, Chapter 4 described the possibility of considering the moisture content
change instead of suction change. In the third approach adopted, the moisture content
change was used to incorporate the stress change within the soil.
The suction profiles assumed for arid climates and semi-arid climates are the same as
shown in Figure 5-3. The suction corresponding to each zone was calculated using
equation [5-1]. However to relate suction to water content, the soil water characteristics
curve (SWCC) is used to obtain the water content change corresponding to the suction
change applied.
The applied stress is given by equation [5-7]. The horizontal strain due to matric suction
can be written as,
𝜀𝑥 =∆𝑆𝐻
= 𝛼∆𝑊 [5-7]
By substituting Equation [5-8] in equation [5-7]
Δ𝜎 =𝐸𝛼∆𝑊
(1 − 2𝜈) [5-8]
By substituting the value of α in Equation [4-11] in Chapter 4 for Equation [5-9] the
stress change becomes,
Δ𝜎 =𝐸𝐺𝑠
(1 − 2𝜈) ∆𝑊𝛼∗
(1 + 𝑒0) [5-9]
Equation [5-10] was then used to calculate the stress change in the continuum. The void
ratio was obtained from either void ratio vs. suction or void ratio vs. water content
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 213
curve. Then, from the void ratio vs. moisture ratio curve the hydric coefficient (𝛼∗) was
obtained.
The results obtained using this method, are presented in Figure 5-4. From these results it
can be seen that the predicted crack depths are less than those compacted by the other
two methods and the theoretical results. However, the actual observed crack depths for
typical clay soils are well matched with these results as shown in Table 2-1 in Chapter
2. Hence it was decided continue use of this method for further analysis.
5.5 Soils Used for the Analysis
In order to examine typical crack depths, it was decided to use several clay soils from
different areas. Two Australian soils and a Canadian soil were selected. Three soils
were named on the basis of their original location as Regina clay, Horsham Clay and
Altona clay.
5.5.1 Regina clay
Regina Saskatchewan, Canada, has a semi-arid continental climate with warm summers
and cold, dry winters. The layers of clay and silt in the area are generally between 3.5m
and 6.5m (Strunk et al., 2009). Measured suctions in these soils are about 3MPa for the
upper 4 m of clay (Vu et al., 2007).
Regina clay is widely used in unsaturated soils research and hence the soil properties are
widely available in the literature (Azam et al., 2012). For this study the soil parameters
214
for Regina clay were selected from the papers published by Fredlund and his team
(Fredlund and Rahardjo, 1993; Fredlund, 2002; Vu et al., 2007). The parameters used
for this study are shown in Table 5-1.
Table 5-1 Soil parameters for Regina clay
Parameter value
Density 15.4 kN/m3
Young’s modulus 10 MPa
Poisson’s Ratio 0.4
Liquid limit 70
Plasticity Index 38
% Passing No. 200 97.8
av2 (gradient of 𝑒 vs. 𝑆 curve) 0.09
Specific gravity 2.83
Friction angle 20˚
For the numerical model the relation between the water content and suction (i.e. SWCC)
is required. According to the SWCC presented in the literature (Vu et al., 2007), the
curve was represented in two equations as shown in Figure 5-5 to simplify the
calculation in the numerical model. If the calculated suction is greater than 387kPa, the
logarithmic equation was used otherwise the 2nd order polynomial equation was used.
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 215
Figure 5-5 SWCC for Regina clay after (Vu, Hu et al. 2007)
Then to calculate the void ratio change, the void ratio vs. water content graph (after
Fredlund, 2002) was used as shown in Figure 5-6 at low stress levels. The hydric
coefficient was obtained from the same figure. However, for simplicity, only three
values were used for hydric coefficient (𝛼∗) as shown in Figure 5-6. The gradient of the
curve changes with the water content. However, only three gradient values (0, 0.5 and
1.0) were used in the numerical model for specific water content ranges. These
gradients in each water content range are shown in Figure 5-6.
216
Figure 5-6 Void ratio vs. water content graph for Regina clay
5.5.2 Horsham clay
Horsham is an area with a semi-arid climate with an annual average of 102 wet days and
mean annual maximum and minimum temperatures respectively of 21.7˚C and 7.7˚C.
The TMI of the area is -23, which is on the arid side. The months May to October fall
below the annual means. Differences in elevation between the brown soils on the higher
parts and the lower-lying grey clay soils are small (Skene, 1959).
Richards (1985) used Horsham clay for his research and published soil properties as
shown in Table 5-2.
y = -7E-06x3 + 0.001x2 - 0.0105x + 0.4876 R² = 0.9988
0
0.5
1
1.5
2
2.5
0 20 40 60 80
Void
ratio
, e
Water content (%)
α*=0.5 α*= 1
α*= 0
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 217
Table 5-2 Soil properties of Horsham clay
Parameter value
Density 16.81 kN/m3
Young’s modulus 12 MPa
Poisson’s Ratio 0.4
Liquid limit 65
Plastic limit 22
Plasticity Index 43
% Passing No. 200 75
av2 (gradient of 𝒆 vs 𝑺) 0.084
Specific gravity 2.8
Friction angle 20˚
𝒔𝒔 (gradient of SWCC) -10.0835
Diffusion coefficient, 𝜶𝒅𝒄 (m2/yr) 11.06
The curves used to obtain the relationship of water content with suction and void ratio
are shown in Figure 5-7 and Figure 5-8 respectively (Richards, 1985). Equations shown
in these figures were used in the numerical model to calculate water content and void
ratio. The hydric coefficient values used for different water content ranges are shown in
Figure 5-8.
218
Figure 5-7 SWCC for Horsham clay (after Richards, 1985)
Figure 5-8 Void ratio vs. water content curve for Horsham clay (after Richards, 1985)
5.5.3 Altona clay
Altona is a temperate to semi-arid area in Western Melbourne. This area has a residual
clay layer 1.5 m below the ground surface with basalt rock found at varying depth
limiting the thickness of the layer (Gallage et al., 2008). The Monash Geomechanics
y = -3.448ln(x) + 43.582 R² = 0.9679
0
10
20
30
1 10 100 1000 10000
Wat
er c
onte
nt, %
Suction (kPa)
y = -7E-05x3 + 0.0047x2 - 0.0777x + 0.7607 R² = 0.9988
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 5 10 15 20 25 30 35
Void
ratio
, e
Water content, %
α*=0 α*=0.5 α*= 1
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 219
Research Group often uses soils from this area and the experimental investigations in
the previous chapter were also carried out using the same soil. The parameters used for
the numerical modelling in this section are listed in Table 5-3.
Table 5-3 Soil properties of Altona clay
Parameter Value
Density 15.3kN/m3
Young’s modulus 6.34MPa (Shannon, 2012)
Poisson’s Ratio 0.4
Liquid limit 70.2
Plastic limit 21.8
Plasticity Index 48.4
% Passing No. 200 93
av2 (gradient of 𝒆 vs 𝑺) 0.033 (Chan, 2012)
Specific gravity 2.61
Friction angle (at joint) 20
𝒔𝒔 (gradient of SWCC) -8.6755
Diffusion coefficient, 𝜶𝒅𝒄 (m2/yr) 12.31
The SWCC and void ratio vs. water content curve for Altona clay are shown in Figure
5-9 (after Chan, 2012) and Figure 5-10 respectively. The relationships used to calculate
220
the water content and void ratio are shown in the figures. The hydric coefficient values
are also shown in Figure 5-10.
Figure 5-9 SWCC for Altona clay (after Chan, 2012)
Figure 5-10 Void ratio vs. water content curve for Altona clay
y = 2E-08x2 - 0.0002x + 0.4501 R² = 1
y = -0.071ln(x) + 0.8159 R² = 0.9863
0
0.1
0.2
0.3
0.4
0.5
0.01 0.1 1 10 100 1000 10000 100000 1000000
Volu
met
ric m
oist
ure
cont
ent
Suction (kPa)
y = 1.68x2 + 1.4941x + 0.3128 R² = 0.9945
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.1 0.2 0.3 0.4 0.5
Void
ratio
Water Content
α*=0.5 α*= 1
α*=0
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 221
5.6 Results
The variation of predicted crack depth values with different influential parameters for
different clay soils are presented here.
5.6.1 Crack depth prediction under different climatic condition
The position of the suction profile beneath the ground surface changes significantly with
the climate condition of the area. Figure 5-11 shows the different suction profiles under
different climatic conditions which were used to obtain the depths of cracking as shown
in Figure 5-12. All three clay soils used in the study showed similar crack depths under
the same climatic condition.
Figure 5-11 Suction profiles under different climatic conditions
0
1
2
3
4
5
6
7
8
1.5 2 2.5 3 3.5 4 4.5 5 5.5
Dep
th (m
)
Suction (pF)
Surface suction=10MPa Surface suction=8MPa Surface suction=6MPa Surface suction=5MPa Surface suction=4MPa Surface suction=2.5MPa Surface suction=1.5MPa Surface suction=0.5MPa Surface suction=0.1MPa
222
The wettest profile used was 100kPa at the surface and the equilibrium suction at depth
was 6.2kPa, which showed no cracks at all in all three soils. The depth of cracking value
of 4.5m was observed for Regina and Altona clay soils when 10MPa surface suction
was applied although for Horsham clay crack depth was about 0.5m less.
Figure 5-12 Predicted depth of cracking change with climatic conditions
In the field generally, 2 to 4m depths of cracks were observed when the matric suction
at shallow depths was recorded around 3MPa values (Russam and Coleman, 1961;
Corte and Higashi, 1964; Richards, 1985; Wijesooriya and Kodikara, 2012). The
predicted crack depths appear to match those observed crack depths since the crack
0
1
2
3
4
5
6
7
8
0.0 2.0 4.0 6.0 8.0 10.0
Dep
th o
f cra
ckin
g (m
)
Maximum surface suction (MPa)
Regina clay Horsham clay Altona clay
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 223
depths for 3MPa matric suction are around 2m for all three soils. It should be noted
however, in these predictions, that the matric suction values given for the surface and
the in the field suctions are measured at shallow depths. Hence, considering the fact that
the surface suction is always higher than the suction values at shallow depths during the
drying periods, the observed crack depths should be a little higher than the predicted
values confirming the accuracy of the predicted results.
5.6.2 Crack opening and closing with time
Figure 5-13 Suction profile variation due to seasonal climate change in (a) Arid area (b) Semi-arid area
0
1
2
3
4
5
6
7
8
1.5 2.5 3.5 4.5 5.5
Dep
th (m
)
Suction (pF)
t=0.1 t=0.2 t=0.3 t=0.4 t=0.5 t=0.6 t=0.7 t=0.8 t=0.9 t=1
(a) Arid 0
1
2
3
4
5
6
7
8
1.5 2.5 3.5 4.5 5.5
Dep
th (m
)
Suction (pF)
t=0.1 t=0.2 t=0.3 t=0.4 t=0.5 t=0.6 t=0.7 t=0.8 t=0.9 t=1
(b) Semi-arid
224
During the year with the change of seasons, the suction profile also changes
significantly. Assuming that a year has only one cycle of climatic change Figure 5-13
was produced. The driest time of the year was obtained when t=0.5yr which represents
around January in Australia. The wettest profile was obtained when t=1yr representing
around July. The broken lines in Figure 5-13 show the change of suction profile from
the wettest condition to driest condition and the solid lines show the change from driest
to wettest condition in each time step.
Figure 5-14 Predicted depth of cracking with seasonal change in an arid area
Depending on the suction or water content profile in the soil, the depth of the crack may
vary. The cracks can be opened further if drying continues or the opened cracks can
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
0 0.2 0.4 0.6 0.8 1
Dep
th o
f cra
ckin
g (m
)
Time (years)
Regina clay
Horsham clay
Altona clay
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 225
close due to soil wetting, along with erosion from the sides of the crack or plastic flow.
Hence it is important to observe the behaviour of the cracks during the seasonal
movements. However, the results shown in the following figures do not show the
continuation of suction profiles through the seasonal variation. Instead they show the
crack depths observed when the suction profile changes from the initial conditions to
the relevant suction profile at a particular time. Hence the results may deviate from the
actual results, when actual progression of suction and associated changes in crack depth
and width are considered with the moisture dynamics associated with actual wetting and
drying.
Figure 5-15 Predicted depth of cracking with seasonal change in an semi-arid area
-0.5
0
0.5
1
1.5
2
2.5
0 0.2 0.4 0.6 0.8 1
Dep
th o
f cra
ckin
g (m
)
Time (years)
Regina clay
Horsham clay
Altona clay
226
Figure 5-14 and Figure 5-15 show the crack depths corresponding to the suction profiles
at times of 0.1yr steps in arid climate and semi-arid climate respectively. For Australian
conditions the cycle will start from the wettest month July (0yr or 1yr) and reach the
driest condition in January at 0.6yrs. The results show that the crack depths increase
quickly after a certain suction level and then reach a peak when the suction becomes a
maximum. During wetting the crack depths decrease sharply as suction decreases and
then remain unchanged. These results suggest that the crack depth can show significant
change during the year with its depth peaking rapidly during the dry period. These
results appear to be consistent with the field observations that cracks close during the
winter and again open up during summer. However, their actual dynamics may be
affected by debris flowing into the cracks and causing changes in the normal pattern of
behaviour.
5.6.3 Effect of placement conditions of the clay liner on initial desiccation
As shown in Chapter 4, after a clay liner is placed it undergoes several cycles before it
comes to a stable condition. However, the first drying period a clay layer is subjected to
after the initial placement is critical in developing desiccation cracks (Kodikara, 2006).
In this context, studying the effect of initial placement conditions is important.
Generally, the maximum density based on the standard proctor compaction curve was
selected as the initial placement density in the numerical model. However, in order to
capture the effect of placement conditions on cracking the density of the soil was
changed.
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 227
In the models for both arid and semi-arid climates, the results obtained by changing the
density only are shown in Figure 5-16 and Figure 5-17. The results in both figures do
not show a significant effect on the depth of cracking. All three soils have produced
similar results although Horsham clay is a little different from the other two in showing
little decrease in crack depth with density.
Figure 5-16 Predicted crack depth variation with initial density of the layer in an arid climate
2
2.2
2.4
2.6
2.8
3
3.2
1400 1500 1600 1700 1800 1900
Dep
th o
f cra
ckin
g (m
)
Initial Placement density (kN/m3)
Regina clay Horsham clay Altona clay
228
Figure 5-17 Predicted crack depth variation with initial density of the layer in an semi-arid climate
The results suggest that the initial placement density used in the UDEC does not
influence on crack depth prediction significantly.
0
0.5
1
1.5
2
2.5
3
1400 1500 1600 1700 1800 1900
Dep
th o
f cra
ckin
g (m
)
Initial Placement density (kN/m3)
Regina clay Horsham clay Altona clay
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 229
5.6.4 Effect of the Poisson’s ratio
The effect of Poisson’s ratio on the depth of cracking was considered for both arid and
semi-arid conditions. The values of Poisson’s ratio were changed from 0.36 to 0.46
where the general values of the Poisson’s ratio for clay soils can be considered to lie
within this range. The results are shown in Figure 5-18 and Figure 5-19 for the arid
climates and semi-arid climates respectively, which show that the depth of cracking
reduces with the increase of Poisson’s ratio.
Figure 5-18 Predicted depth of cracking variation with the Poisson's ratio in an arid area
All soils show the same trend and similar values of crack depth at any value of
Poisson’s ratio. In an arid climate (Figure 5-18) the depth of cracking has decreased
0
0.5
1
1.5
2
2.5
3
3.5
0.3 0.35 0.4 0.45 0.5
Dep
th o
f cra
ckin
g (m
)
Poisson's ratio
Regina clay Horsham clay Altona clay
230
from 5m to 3m for 0.1 changes in the Poisson’s ratio value. Also in semi-arid climates
(Figure 5-19) the change of depth of cracking is 4m to 1.5m given the same change in
Poisson’s ratio.
Figure 5-19 Predicted depth of cracking variation with the Poisson's ratio in an semi-arid area
These results indicate that at lower suctions the effect of the Poisson’s ratio is higher
than at higher suctions. However, soils with higher values of the Poisson’s ratios are
less problematic in any climate condition against the desiccation cracking.
0
0.5
1
1.5
2
2.5
3
0.3 0.35 0.4 0.45 0.5
Dep
th o
f cra
ckin
g (m
)
Poisson's ratio
Regina clay Horsham clay Altona clay
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 231
5.6.5 Effect of equilibrium suction
As the next parameter the effect of change of equilibrium was selected. Several values
of amplitude of suction at the surface variation (∆𝑆) have been used in the past research.
Mitchell (1979) has used 1.5pF and the Australian footing standard suggests the use of
1.2pF suction for design purposes. However, the equilibrium value can be changed
according to variation in other conditions such as the depth to water table and soil
properties.
Figure 5-20 Suction profiles below the ground surface with different equilibrium suction values in (a) an arid climate (b) a semi-arid climate
To capture the change in the effect of equilibrium suction due to different reasons, the
models were run with suction profiles as shown in Figure 5-20. The suction profiles
0
1
2
3
4
5
6
7
8
3 3.5 4 4.5 5
Dep
th (m
)
Suction (pF)
Se=0.80MPa
Se=0.63MPa
Se=0.50MPa
Se=0.40MPa
Se=0.32MPa
Se=0.25MPa
Se=0.20MPa
Se=0.16MPa
(a) Arid
0
1
2
3
4
5
6
7
8
2.5 3 3.5 4 4.5 D
epth
(m)
Suction (pF)
Se=0.39MPa
Se=0.31MPa
Se=0.25MPa
Se=0.20MPa
Se=0.16MPa
Se=0.12MPa
Se=0.10MPa
Se=0.08MPa
(b) Semi-arid
232
representing arid and semi-arid climates are shown in Figure 5-20 (a) and (b)
respectively. The amplitude of suction at the surface was changed from 0.8pF to 1.5pF
while keeping the surface suction constant in both climate conditions. The equilibrium
suctions were changed from 0.16MPa to 0.8MPa in arid climate conditions and from
0.08MPa to 0.39MPa in semi-arid climates.
Figure 5-21 Predicted depth of cracking with the change of equilibrium suctions in an arid climate
The results are shown in Figure 5-21 and Figure 5-22 for arid and semi-arid climates
respectively. As expected, the crack depths increase similarly with the increase of
equilibrium suction in all soils both in arid and semi-arid climate conditions.
0
1
2
3
4
5
6
0.00 0.20 0.40 0.60 0.80 1.00
Dep
th o
f cra
ckin
g (m
)
Equlibrium suction (MPa)
Regina clay
Horsham clay
Altona clay
Modelling of Stable Desiccation Crack Depths during Cyclic Wetting and Drying 233
Figure 5-22 Predicted depth of cracking with the change of equilibrium suctions in an semi-arid climate
5.7 Conclusions
The UDEC program was used in this chapter to predict the crack depths for several soil
types. The predicted crack depths using the numerical model represent the generally
observed values of depths of cracking in the field reasonably well.
Three soils were selected to represent the clay soils in different climatic locations and
soil types. Generally all three clay soils behave similarly with the change of influencing
parameters. Especially for Regina and Altona clay, the predicted crack depths are
almost the same while the Horsham clay shows around 0.5m less value. However, it is
0
0.5
1
1.5
2
2.5
3
3.5
0.00 0.10 0.20 0.30 0.40 0.50
Dep
th o
f cra
ckin
g (m
)
Equlibrium suction (MPa)
Regina clay Horsham clay Altona clay
234
reasonable to conclude that all three soils produce similar depths of cracking under the
same conditions.
The effects of several parameters on depth of cracking were analysed and discussed.
The climate condition of the area has a great influence on the depth of cracking and the
predicted crack depths range between zero and 8m when the aridity increases. The
higher compaction during the placement of a layer tends to decrease the depth of
cracking while a lower Poisson’s ratio increases the crack depth. In a situation where
the water table is significantly below the ground level, increasing equilibrium suction
while keeping the surface suction constant tends to increase the crack depth. This may
represent a scenario where the atmospheric conditions remain the same (therefore the
surface suction remains the same) but the water table is dropping thereby increasing the
equilibrium suction level.
While the analysis presented in this chapter encompasses the major stresses that control
desiccation cracking in a certain climatic location, the actual crack development and
dynamics may be influenced by other factors that were not considered in this analysis.
These factors include the effects of continuous drying and wetting dynamics of the soil
and associated influence on the crack dynamics, and the effects of other intervening
events such as debris flow into the cracks and associated changes in stress and moisture
development. Nevertheless, the computed crack depths appear to be reasonable on the
basis of the field observations.
Chapter 6
INHERENT PROPERTIES OF UDEC FOR FRACTURE MODELLING
6.1 Introduction
In previous chapters, the UDEC program was used for fracture modelling utilising the
capabilities available in the program. In other words, the fracture was considered to
initiate when the tensile strength of the soil was reached. However, the fracture energy
associated with opening a new surface was not controlled. It was not clear whether the
UDEC inherently has a kind of fracture energy in its current formulation. Therefore, this
chapter investigates the inherent properties of UDEC formulation with respect to
fracture formation.
235
236
6.2 The Numerical Program UDEC
6.2.1 UDEC operation
The newest UDEC version 5.0 (released in 2011) and the previous version (released in
2004) were used for the modelling in this research. UDEC in its general form:
• allows simulation of motion of blocks (including slip and opening) in a
discontinuous medium;
• treats the discontinuous medium as an assembly of discrete polygonal blocks
with round corners. The motion and interaction of blocks are computed on the
basis of an explicit solution scheme which allows tracing of the mechanical
evaluation of the system even if the process becomes mechanically unstable;
• uses the linear and non-linear force displacement laws to govern the relative
motion along the discontinuities for the normal and shear directions;
• assumes blocks to be rigid or deformable. In the deformable configuration (as
used in this research) it discretises a block into zones and assigns continuum
constitutive models to each of them; and
• provides an in built library of materials models for deformable blocks and
discontinuities to suit the problem.
In UDEC, the geometry of the model is defined in terms of a fixed orthogonal
horizontal-vertical (x-y) system where x is positive in the rightwards direction and y is
positive in the upward direction. Displacements and force refer to the fixed x-y system
and counter clockwise moments and rotations are positive. The tensile stresses and
corresponding material elongations are positive and compressive stresses and
Inherent Properties of UDEC 237
corresponding contractions are negative for the block material. Joint normal stress is
positive in compression while joint normal displacement is positive in opening. Positive
shear stresses are shown in Figure 6-1.
Figure 6-1 Sign convention for positive shear stress components
UDEC only handles problems in two-dimensions either in plane stress or plane strain.
Real problems are generally three-dimensional. However, useful predictions of the
behaviour of systems can often be obtained from two-dimensional analyses depending
on the problem.
6.2.2 Theoretical background of UDEC
UDEC computation is based on a time marching algorithm in which governing
equations are integrated explicitly in time in a series of successive cycles or steps.
Actions performed in one cycle are shown in Figure 6-2. A central finite difference
scheme is used to integrate equations of motion. Any grid point in a deformable block is
238
associated with the surrounding zones. The area influencing one grid point is shown in
Figure 6-3.
The quation of motion for velocities at grid point P is calculated as,
𝑢𝚤̈ = ∫ 𝜎𝑖𝑗 𝑛𝑗 𝑑𝐴𝑠+𝐹𝑖
𝑚+ 𝑔𝑖 [6-1]
𝜎𝑖𝑗 is the zone stress tensor. 𝐴𝑠 is the surface enclosing the area lumped at the grid point
P and 𝑛𝑗 is the unit normal to 𝐴𝑠 . 𝑢𝚤̈ is the acceleration. 𝑔𝑖 is the gravitational
acceleration. 𝐹𝑖 is the resultant external forces applied to the grid point P, can be
obtained as a sum of three terms,
𝐹𝑖 = 𝐹𝑖𝑙 + 𝐹𝑖
𝑐 + 𝐹𝑖𝑧 [6-2]
Apply Equations of motion to all
gridpoints in blocks
Update block grid point co-ordinates and
contacts between blocks (delete or create
contacts as needed)
Compute unbalanced force and exit loop if force if force is below limit and
solve command has been issued
Apply material constitutive equations to all zones in blocks and contacts between
blocks
Derive stresses and nodal forces
Derive velocities and displacements
Figure 6-2 Actions performed during one computation cycle
Inherent Properties of UDEC 239
Forces 𝐹𝑖𝑙 are the external applied loads. Forces 𝐹𝑖
𝑐 result from the contact forces and
exist only for grid points along the block boundary. The force contributions from the
internal stresses in the zones adjacent to the grid point are calculated as,
𝐹𝑖𝑧 = ∫ 𝜎𝑖𝑗 𝑛𝑗 𝑑𝐴𝑠 [6-3]
By finite difference integration of equation [6-1],
𝑢(𝑡+∆𝑡) = 𝑢(𝑡) + �̇�(𝑡+∆𝑡/2)∆𝑡 [6-4]
With velocities stored at the half time step point, it is possible to express displacement
as in equation [6-4] where the superscripts denote the time at which the corresponding
variable is evaluated. Because the force depends on displacement, the
force/displacement calculation is done at one time instant.
Zone 1
P
Zone 2
Zone 3
Zone 4Zone 5
Zone 6
Fi
Figure 6-3 Area associated to grid point P
During each time step, strains are related to nodal displacements in the usual fashion,
240
𝜖�̇�𝑗 = 12
��̇�𝑖,𝑗 + �̇�𝑗,𝑖� [6-5]
This equation (equation [6-5]) does not imply a restriction to small strains due to the
incremental treatment. The constitutive relations for deformable blocks are used in an
incremental form so that implementation on nonlinear problems can be accomplished
easily. The actual form of the equation is,
∆𝜎𝑖𝑗𝑒 = 𝜉Δ𝜖𝜐𝛿𝑖𝑗 + 2𝜇Δ𝜖𝑖𝑗 [6-6]
where, ξ and µ are the Lamé constants, ∆𝜎𝑖𝑗𝑒 is the elastic increment of the stress tensor,
Δ𝜖𝑖𝑗 is the incremental strain, Δ𝜖𝜐 is the increment of volumetric strain and 𝛿𝑖𝑗 is the
Kronecker delta function.
In equations of motion, damping is used to simulate dissipation of kinetic energy in geo-
materials as deformation takes place. Several damping methods can be used in UDEC.
Local damping and auto damping are generally used for quasi-static problems (i.e. when
loading or unloading rates are low enough that inertia effects can be disregarded). The
damping scheme is designed to converge to the static solution (if it exists) as fast as
possible.
For local damping, the direction of the damping force is such that energy is always
dissipated. For deformable blocks, the equation of motion given by equation [6-7] is
replaced by
𝑢𝚤̇ (𝑡+∆𝑡/2) = �̇�(𝑡−∆𝑡/2) + �∑ 𝐹𝑖(𝑡) − 𝛼�∑ 𝐹𝑖
(𝑡)�𝑠𝑔𝑛 ��̇�(𝑡−∆𝑡/2)�� ∆𝑡𝑚𝑛
[6-7]
to incorporate local damping. Viscous damping is used in auto damping. However, the
viscosity constant is continuously adjusted in such a way that the power absorbed by
Inherent Properties of UDEC 241
damping is a constant proportional to the rate of change of kinetic energy in the system.
The adjustment to the viscosity constant is made by a numerical servo-mechanism that
seeks to keep the following ratio, 𝑅, equal to a given ratio.
𝑅 =∑ 𝑃∑ �̇�𝑘
[6-8]
where, 𝑃 is the damping power for a node, �̇�𝑘 is the rate of change of nodal kinetic
energy.
A limiting time step for integration of the dynamic equations of motion is chosen to lead
to a stable computation of internal block deformation and stable computation of inter-
block relative displacements. Generally higher numbers of time steps are preferred with
local or auto damping conditions.
6.3 Description of a problem analysed using UDEC
In order to analyse the scale effects, the problem analysed in this chapter was a centre
crack in a finite square body subjected to far field constant stress (𝜎𝑓) normal to the axis
of the crack, as shown in Figure 6-4(a). The problem was simplified due to the
symmetry of the problem, as shown in Figure 6-4(b).
In the numerical model, two blocks were connected to each other through a joint to
represent the geometry shown in Figure 6-4. The block can be defined as the
fundamental geometric entity for the distinct element calculation in the program UDEC.
The deformable blocks were to be composed of elastic material following an isotropic
elastic constitutive model and the joint was behaving according to the Mohr-Coulomb
242
area of contact with tensile strength. The existing crack differentiates from the joint by
applying no tensile strength for the length of crack. 2𝑊 is the width of the block and 2 𝑎
is the length of the crack.
The bulk modulus (𝐾 ) and shear modulus (𝐺 ) were calculated using the Young’s
modulus (𝐸) and the Poisson’s ratio (𝜈) as:
𝐾 = 𝐸3(1−2𝜈)
[6-9]
𝐺 = 𝐸2(1+𝜈)
[6-10]
and the joint normal stiffness (𝑘𝑛) and shear stiffness (𝑘𝑠) were calculated using the
equation,
𝑘𝑛 = 𝑘𝑠 = 10 �𝑚𝑎𝑥 �𝐾+4/3𝐺∆𝑧𝑚𝑖𝑛
�� [6-11]
𝜎𝑓
𝜎𝑓
2a 2W
2W
𝜎𝑓
2W a
W
𝜎𝑓
Figure 6-4 (a) Problem geometry and modelled problem (b) analysed problem due to symmetry
Inherent Properties of UDEC 243
where ∆𝑧𝑚𝑖𝑛 is the minimum size of the zones. The input parameters given in Table 6-1
are not general but specifically related to a soil since the intention was to obtain a clear
understanding of the program behaviour with the change of different parameters.
Table 6-1 Typical input parameters of the model
Parameter Block Joint Crack
Size 2 𝑊 × 𝑊 (𝑊 − 𝑎) 𝑎
Density 1835 kg/m3 n/a n/a
Bulk modulus 5e8 Pa n/a n/a
Shear modulus 2e8 Pa n/a n/a
Joint normal stiffness n/a 5e8Pa.m 5e8Pa.m
Joint shear stiffness n/a 5e8Pa.m 5e8Pa.m
Tensile strength n/a 1000Pa 0
Cohesion n/a 0 0
Friction angle n/a 25˚ 25˚
Dilation angle n/a 0 0
Residual cohesion n/a 0 0
Residual friction n/a 25˚ 25˚
Residual tensile strength n/a 1000Pa 0
In the model far field stress was applied at the boundary of the block. To avoid unstable
conditions in the model the stress was applied consecutively from zero to the final stress
limit in small increments and then the failure stress was obtained. The procedure
followed in applying the stress is discussed in a later section.
244
6.4 Behaviour of the UDEC Model
The behaviour of UDEC with the change of each parameter was analysed and is
discussed in this section. Mesh size, sizing effect, damping, modulus etc were selected
specifically for the analysis.
6.4.1 Effect of mesh size (l)
Each material block is divided into finite difference zones as discussed before. The
maximum edge length of those triangular zones was given as the input which is referred
to mesh size.
The effect of mesh size change was observed while keeping the other inputs constant,
such as block size, properties of the bulk media and joint, size of the crack and so on.
The change of the size of the mesh in the block is shown in Figure 6-5. From Figure
6-5(a) to (f) mesh size decreases from 0.2 m to 0.005 m. The block size was selected as
1.2 × 0.6 m and the length of the crack was 0.2m.
The failure stress for the above shown different models was obtained from stress vs.
(load step) time curve, where the time at the failure was found from the unbalanced
force vs. time graph. The method followed for obtaining failure stress is shown below in
Figure 6-6.
Inherent Properties of UDEC 245
Figure 6-5 Change of mesh size in a constant size block
Since there was no clear point for the failure in the stress vs. displacements graph, an
alternative method to find the failure stress was needed. The unbalanced force (Figure
6-6(a)) at the beginning shows almost zero value before the joint contacts start to open.
However, as the failure stress was reached and the blocks started to separate from each
other, a sudden change in the unbalanced force curve can be observed. The time when
(a)
(b)
(d)
(c)
(e)
(f)
Mesh size = 0.2m
Mesh size = 0.005m
Mesh size = 0.05m
Mesh size = 0.01m
Mesh size = 0.1m
Mesh size = 0.02m
246
this happened was identified as the failure point and the corresponding time and the
failure stress were determined from the stress vs. time graph as shown in Figure 6-6 (b).
Figure 6-6 Selection of Failure stress
(a)
(b)
σ’
Inherent Properties of UDEC 247
A typical failure stress obtained using this method is shown in Figure 6-6. All results
are shown in Figure 6-7 where log (failure stress) vs. log (mesh size) is shown. These
results show the effect of mesh size indicated as l. 𝑘𝑛 is the joint normal stiffness and N
denotes the number of stress increments used to achieve the final stress. N is described
in more detail in Section 6.4.2. As can be seen from Figure 6-7, there is slight increase
in failure stress with increasing mesh size. However, it can be ignored for the cases
considered.
Figure 6-7 Effect of mesh size; 𝐾 = 5e9 Pa, 𝐺 = 2e9 Pa, 𝑘𝑛 = 5e9 Pam, Damp = 0.2, N = 10000, 𝑊 = 0.6m 𝑎 = 0.2 m and 𝜎𝑡 = 1000Pa.
𝝈 is the failure stress and 𝝈𝒕 is the tensile strength of the joint.
Potyondy and Cundall (2004), (as proposed by (Anderson, 1991)) has presented a
relationship for failure stress as,
𝜎΄ = 𝐾𝐼𝐶𝐶(𝜓) √𝜋
𝑎−12� [6-12]
0.1
1.0
0.004 0.04
log
(σ/σ
t)
log L
kn=5e9Pa kn=5e8Pa
248
where, 𝐶(𝜓) = [𝑠𝑒𝑐 �𝜋𝜓2
�]12� [1 − 0.025𝜓2 + 0.06𝜓4] and 𝜓 = 𝑎
W� . 𝐾𝐼𝐶 is the mode 1
fracture toughness, σ΄ is the failure tensile stress, 𝑎 is the crack length, and 2𝑊 is the
height of the soil block.
According to equation [6-12] there is no direct relationship between the failure stress
and mesh size, similar to the results predicted from UDEC in Figure 6-7. However, the
maximum stress which is governed by tensile strength shows proportionality with the
length of joint. So, when the tensile stresses are applied on a length of 𝑊, failure occurs
in a length of (𝑊 − 𝑎), and the ratio between the applied tensile stress vs. failure stress
can be written as,
𝜎′𝜎𝑡
= 𝑊−𝑎𝑎
[6-13]
When, 𝑊 =0.6m, applied tensile stress=1000Pa and 𝑎 =0.2m (as the values used to
obtain results in Figure 6-7), the failure stress is 666Pa. The results produced by the
numerical program are also closer to this value.
6.4.2 Effect of number of increments
The effect of change of the number of stress increments (N) was examined in this
section. In UDEC, the stresses should be applied consecutively in small steps to avoid
instability of the model. The following part of the code was used to apply stresses at the
boundaries.
def bstress
str1=1.0e3 ; Failure stress
Inherent Properties of UDEC 249
sdif=str1/10000 ; Stress increment, N = 10000
nsb1=0 ; Initial normal stress at the boundary
nsblt=1*str1 ; Normal stress limit at the boundary
loop while nsb1 < nsblt
nsb1=nsb1 + sdif
command
bound st 0 0 sdif range yr -0.001,0.001 ; Applying stresses at bottom boundary
bound st 0 0 sdif range yr 9.599,9.601 ; Applying stresses at top boundary
solve force 0.1
endcommand
endloop
end
As it can be seen in the FISH code, the final stress was divided into 10000 divisions and
each division was added cumulatively to the stress. This number of divisions was
referred to as the number of stress increments (N).
Here, the effect of the number of suction increments was captured, while keeping all
other parameters constant. The variation of failure stress and overall modulus of the
model was observed. The failure stress was obtained as described in Section 6.4.1.
Overall modulus was obtained from the stress vs. displacement curve as shown in
Figure 6-8.
250
Figure 6-8 Stress vs. displacement curve using to get Overall 𝐸
The failure stress and overall modulus values are shown in Figure 6-9. Both failure
stress values and overall modulus values do not show significant change when changing
the number of increments, indicating no major effect from the number of increments.
∆𝜎
∆𝑢
E* = ∆𝜎∆𝑢
Inherent Properties of UDEC 251
Figure 6-9 Effect of number of stress increments; 𝐾 = 5e9Pa, 𝐺 = 2e9Pa, 𝑘𝑛=5e9Pa, Damp=0.2, 𝑊 =0.6m, 𝜎𝑡=1000Pa, 𝑎 =0.02m and 𝑙 =0.2m
However, when the load vs. displacement curves are examined it can be observed that,
these were changing from a rough variation to a smooth linear variation when
increasing the number of stress increments at the initial part of the curve (Figure 6-10).
Hence it can be recommended to use small increments when changing the conditions of
a model such as stress, velocity etc.
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
0
100
200
300
400
500
600
0 2000 4000 6000 8000 10000 12000 14000
Mod
ulus
*109 ,
E (P
a)
Stre
ss, σ
(Pa)
Number of increments
Stress
Overall E
252
Figure 6-10 Change of load vs. displacement plot with the change of number of cycles
6.4.3 Effect of crack length
In this section, models were used to check the effect on failure stress. The crack lengths
of the models were changed from 0.1m to 1.0m as shown in Figure 6-11 and the results
were plotted in Figure 6-12. By manipulating equation [6-12], the failure stress over
tensile strength can be given by the relationship,
(a) (d)
(e) (b)
(c) (f)
N=2000
N=4000
N=8000
N=6000
N=10000
N=120000
Inherent Properties of UDEC 253
𝜎΄𝜎𝑡
=𝐾𝐼𝐶
𝜎𝑡 √𝜋 (
1𝐶(𝜓) √𝑎
) [6-12a]
Hence in order to obtain a straight line response log (failure stress/tensile strength) vs.
log (1/ C(ψ) √𝑎 ) was plotted. The results show good agreement to the theoretical
gradient as shown in Figure 6-12.
It can be seen that when the crack length increases the failure stress reduces. When the
crack length increases the uncracked length reduces, which reduces the energy required
to break the material and the joint can fail with lesser stress. This highlights that the
model behaves according to the traditional fracture mechanics theoretical concepts.
a2.4m
1.2m
aaa
1.2m1.2m1.2m
Figure 6-11 Model geometry change with changing crack length
254
Figure 6-12 Effect of crack length, 𝐾 = 5e9Pa 𝐺 = 2e9Pa 𝑘𝑛=5e9Pam, Damp=0.2, N=10000, 𝑊 =1.2m, 𝜎𝑡=1000Pa, 𝑙 =0.04m and 𝐾𝐼𝐶=530Pam0.5
6.4.4 Effect of damping value
In UDEC, a solution is reached when the rate of change of kinetic energy in the model
approaches a negligible value (Itasca, 2004). This is accomplished by damping the
equations of motion, to reach a force equilibrium state as quickly as possible under the
applied initial and boundary conditions. Hence, the effect of damping was also
examined as another parameter, since it may affect the crack propagation specifically.
Here the damping value was changed from 6% to 25%, and other parameters kept
constant. The failure stresses and fracture toughness values were plotted in Figure 6-13.
As the graph shows, the effect of damping within this range is negligible. However, to
see the minor effects of damping on the stability of the model, the stress vs.
displacement curves of each model were examined (Figure 6-14). The slight increase in
the failure stress with increasing damping value can be noted. Nevertheless, these
0.1
1
0.3 3
Log
(σ′/
σ t)
log [1/a0.5*C(ψ)]
Theoretical Gradient Line
Inherent Properties of UDEC 255
effects can be considered negligible and, therefore, small. A 6% damping value was
selected as a reasonable value.
Figure 6-13 Effect of damping value, 𝐾 = 5e9Pa, 𝐺 = 2e9Pa, 𝑘𝑛=5e9Pam, N=10000, 𝑊 =0.6m, 𝜎𝑡=1000Pa and 𝑎 𝑙� =10 (𝑎 =0.02m, 𝑙 =0.2m)
Figure 6-14 Stress vs. displacement curves when changing the damping value
0
100
200
300
400
500
600
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Damping value
Stress
KIC
D=6% D=25%
256
6.4.5 Effect of block size
In this section, the size of the block has been changed while keeping the crack length
and other soil and joint properties constant. The change of the geometries of the models
is shown in Figure 6-15.
To find the applicability of the equivalence between the theoretical assumptions in
Linear Elastic Fracture Mechanics (LEFM) and the theoretical formulation in the UDEC
analysis, the stress along the axis of the crack was plotted in Figure 6-16, in which the
LEFM results were derived considering the centre cracked rectangular plate under
uniform tension.
Figure 6-15 Change of Size of block when the crack length kept constant
For this configuration, the stress intensity factor, 𝐾𝐼 is given by,
𝐾𝐼 = 𝜎𝑓√𝜋𝑎 . 𝑓(𝛼, 𝛽) [6-14]
nW
nW
nW
nW
2n
2n
2n
2n a
a
a
a
n = 1
n = 2
n = 3
n = 4
Inherent Properties of UDEC 257
where, 𝛼 = 2𝑎𝑊
, 𝛽 = 2𝐻𝑑𝑊
, where 𝐻𝑑 is the height of the block and 𝑓(𝛼, 𝛽)= 1.0415 in this
particular case (Murakami, 1986) and 𝜎𝑓is the far field stress applied on the finite plate.
The normal stress near the crack tip was given by:
𝜎𝑛 = 𝐾𝐼√2𝜋𝑟
[6-15]
where r is the distance from the crack tip.
Figure 6-16 Normal Stresses from UDEC and LEFM ahead of the crack
6.4.6 Scaling up the model geometry
The failure stress is affected by scale of the problem in fracture mechanics. Hence, the
geometry of the model was scaled up by a factor ‘n’ to examine the behaviour of the
model response, as shown in Figure 6-17. The effect of problem scale on failure stresses
and the fracture toughness was observed in this series of models. The soil and joint
0
100
200
300
400
500
600
700
800
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Nor
mal
Str
ess (
Pa)
Distance from the crack tip (m)
Numerical
Theoretical
258
properties were selected as constant values. Furthermore, the mesh size was kept
constant in all the models.
6.4.6.1 Effect of joint normal stiffness while scaling up
Figure 6-18 Scale effect on failure stress, 𝐾 = 5e9Pa, 𝐺 = 2e9Pa, Damp=0.2, N=10000 and 𝜎𝑡=1000Pa
0.100
1.000
0.02 0.2 2
log
(σ'/σ t
)
log na
kn=5e8Pa kn=5e9Pa kn=5e10Pa kn=5e11Pa
n = 1
n = 2
n = 3
n = 4
Figure 6-17 Plot of models changing the size of geometry
nW
nW
n
nW
2n
2n
2n
2nW na
na
na
na
Inherent Properties of UDEC 259
The log of failure stress was plotted against log of crack length with different joint
stiffness values as shown in Figure 6-18.
When lower fracture energies were used associated with higher normal stiffness values,
the curves tend to show close to a gradient of -1/2 representing LEFM behaviour. When
the fracture energy is increased by lowering the normal stiffness values, the normalised
failure stress becomes flatter for small crack lengths prior to changing to approximately
-1/2 gradient. The flatter part of the curve indicates the influence of fracture process
zone for smaller crack lengths, showing quasi-brittle or more ductile behaviour.
However, as the crack length increases, the influence of the fracture process zone
diminishes, eventually approaching the LEFM response. From this graph we can
suggest that the relevance of cohesive law is crucial when higher fracture energies are
used with higher crack lengths and when the lower fracture energies are considered.
The fracture energy related to the curve representing 𝑘𝑛 = 5𝑒9𝑃𝑎 is around 0.3Pam,
which is more relevant to very soft clay soils and this value is lower than that of natural
compacted clay.
Displacement
Tens
ile st
ress
Fracture energy
Figure 6-19 Inherent fracture energy present in current UDEC formulation
260
Therefore, in the current implementation of UDEC, there is an inherent fracture energy
associated with the initial elastic response of the joint. This fracture energy component
is shown in Figure 6-19. However, normal stiffness 𝑘𝑛 should be not be used to control
the fracture energy after failure because it is related to the physical behaviour prior to
cracking. A more rational way is to control the post failure stiffness to represent the
actual fracture energy of crack formation. This aspect will be considered in the next
chapter.
Figure 6-20 Scale effect on Fracture energy, 𝐾 = 5e9Pa, 𝐺 = 2e9Pa, Damp=0.2, N=10000 and 𝜎𝑡=1000Pa
Figure 6-20 presents theoretical and UDEC results for fracture energy with different
normal stiffness values and block size. When the normal stiffness value is constant, the
size of the block does not have an effect on fracture energy. This can be expected since
the fracture energy is related to the tensile strength and 𝑘𝑛 as noted earlier (both of them
are constant).
1.00E-07
1.00E-06
1.00E-05
1.00E-04
1.00E-03
1.00E-02
0 0.5 1 1.5
log
Gf
a^0.5
kn=5e8 UDEC
kn=5e9 UDEC
kn=5e10 UDEC
kn=5e11 UDEC
kn=5e8 Theoretical
kn=5e9 Theoretical
kn=5e10 Theoretical
kn=5e11 Theoretical
Inherent Properties of UDEC 261
Figure 6-21 Effect of normal stiffness on failure stress, 𝐾 = 5e9Pa, 𝐺 = 2e9Pa, Damp=0.2, N=10000 and 𝜎𝑡=1000Pa
Figure 6-22 Effect of normal stiffness on fracture toughness, 𝐾 = 5e9 Pa, 𝐺 = 2e9 Pa, Damp=0.2, N=10000 and 𝜎𝑡=1000Pa
0.100
1.000
1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+12 1.E+13
log
(σ′/
σ t)
log Kn
1.0
10.0
100.0
1000.0
1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+12 1.E+13
K IC
log kn
UDEC changing kn
Theoretical
262
Figure 6-23 shows the effect of normal stiffness on the normalised strength while all
other aspects of the model were kept constant. When the normal stiffness is low, the
joint imparts a significant resistance to failure due to associated high fracture energy.
However, when the joint stiffness becomes very large, the strain energy coming from
the continuum section (its modulus was kept constant in this instance) can easily break
the joint overcoming the fracture energy resulting from the low joint stiffness.
Therefore, the curve tends to flatten out as the joint stiffness increases.
When increasing the normal stiffness it can be seen that the numerical results deviate
from the theoretical value, although for smaller normal stiffness values both values
show similar results. That happens when smaller fracture energies using unloading
modulus also act same but in numerical model which does not allow the fracture
toughness to fall in high numbers.
6.4.6.2 Effect of bulk and shear modulus
After observing the variation of joint stiffness, the effect of bulk and shear modulus was
tested in this section.
Figure 6-23 shows the effect of modulus on failure stress. When the stiffness of the
block material increases to higher values then the LEFM is no longer applicable since
the failure stress does not show considerable variation with the size of the geometry.
This shows that the inherent fracture behaviour in UDEC comes into the picture only for
deformable blocks. That indicates that when the blocks have a smaller modulus,
deformable material can store and release energy for fracture propagation.
Inherent Properties of UDEC 263
Figure 6-23 Effect of modulus on failure stress 𝑘𝑛=5e9Pam, Damp=0.2, N=10000 and 𝜎𝑡=1000Pa
Figure 6-24 Effect of modulus on failure stress, 𝑘𝑛=5e9 Pam, Damp=0.2, N=10000 and 𝜎𝑡=1000Pa
0.100
1.000
0.02 0.2 2
log
(σ′/
σ t)
log a
K=5e8 K=5e9 K=5e10 K=5e11
0
200
400
600
800
1000
1200
1400
1600
1800
0 0.2 0.4 0.6 0.8 1 1.2 1.4
K IC
a0.5
K=5e8 UDEC K=5e9 UDEC K=5e10 UDEC K=5e11 UDEC K=5e8 Theoretical K=5e9 Theoretical K=5e10 Theoretical K=5e11 Theoretical
264
6.5 Summary and Conclusions
The behaviour of the output results from the UDEC numerical model was observed with
the change of different input parameters focused on energy associated with the
numerical modelling.
The failure stress increases with the increase of mesh size. This result shows good
agreement with the results presented in Chapter 3. This shows the need of selecting the
precise size of the mesh before using the program for the actual problem.
The increments used to implement the stress change need to be selected carefully so that
no high unbalanced forces are generated in the model. Similarly, the studies should
include a selection of an appropriate damping value to be used in the model. This can
become important when cracks are close to each other where opening of one crack may
affect the response of the other crack.
The hidden fracture energy associated with current implementation of UDEC was found
to be due to the initial normal stiffness of the joint. When the normal stiffness is low,
the associated fracture energy can lead to high fracture process zones and therefore,
plastic-like behaviour in fracture development. Under these circumstances, the model
results will deviate from classical LEFM behaviour. However, the fracture energy
cannot be properly controlled with the current implementation. This aspect will be
considered in more detail in the next chapter.
Chapter 7
MODELLING OF DESICCATION CRACK DEPTHS INCORPORATING SOIL FRACTURE ENERGY
The numerical modelling work undertaken up to now did not consider soil fracture
energy directly. This chapter will detail how it can be incorporated in the numerical
modelling. The cohesive fracture approach is presented as a way of consistently and
conveniently incorporating soil fracture energy in desiccation crack formation. The
cohesive fracture method is considered to be particularly suited to soil since soil can
exist in a wide range of consistencies during desiccation.
7.1 Introduction
Fracture mechanics is a failure theory that determines the material failure by fracture
using an energy criterion possibly in conjunction with yielding. It considers that in order
to generate a surface within a material, energy is required and this energy is referred to
as the fracture energy. Basically fracture mechanics can be considered in two main
265
266
forms: Linear Elastic Fracture Mechanics (LEFM) and Elasto-Plastic Fracture
Mechanics (EPFM). LEFM is more applicable for brittle materials such as glass and
some metals while EPFM gives excellent results for ductile materials such as certain
metals, alloys and polymers.
LEFM considers that there is a stress singularity at the crack tip and any plastic zone
that develops due to this stress singularity is relatively small compared to the size of the
structure subject to loading. For soil, however, it is not very clear whether LEFM will
be applicable for all soil consistency states from the liquid limit to below the plastic
limit. Arguments for the use of LEFM (Lakshmikantha et al., 2012) and for the use of
EPFM (Hallett and Newson, 2001) have been raised by various researchers. Therefore,
a common approach that caters for both conditions is needed for soil cracking.
In this chapter, the use of a cohesive crack approach that uses the stress softening while
opening the crack is followed. This approach is applicable to both elastic and plastic
conditions and hence can cater for all soil consistency states. The values of the softening
curve are calculated on the basis of the fracture energy. The approach is implemented
in the numerical code of UDEC and is applied to predict the crack depths for the same
soils used in the Chapter 6. The effect of cohesive properties of the fracture during the
initiation and progression of fracture is studied.
7.2 Basics of Linear Elastic Fracture Mechanics (LEFM)
LEFM was originated by Griffith (1921; 1924) formulating an energy based criterion
for propagation of cracks and then was extended by Irwin (1957) developing the stress
Modelling of desiccation crack depths incorporating soil fracture energy 267
intensity factor and fracture process zone. Later Rice (1968b (a); 1968a (b)) introduced
the J-integral as another way of computing the critical energy release rate applicable to
fracture. However, LEFM is valid as long as the material behaves elastically and the
fracture process zone is small compared to the loaded area of the structure. In other
words, the main assumption is that the whole area of the material is elastic except the
vanishingly small area at the tip of the crack.
7.2.1 Griffith’s criterion
It is well known that the growth of a crack requires the creation of two new surfaces and
hence consumes more energy because the surfaces carry more energy than material
bodies. Based on this concept, Griffith found an expression for the stored elastic energy
U of the crack by solving the elasticity problem (Figure 7-1) of a finite elliptical crack
with a length 2𝑎 at the centre of a semi finite, homogeneous, isotropic plate with
Young’s modulus 𝐸 which is subjected to applied tensile stress σ given by,
𝑈 =𝜋𝜎2𝑎2
𝐸
[7-1]
assuming that the thickness of the plate is of unit length. When the surface energy per
unit area of the material is 𝜁, the total required energy to open the crack by creating two
new surfaces of a crack of length 2a is,
𝑊 = 4𝑎𝜁 [7-2]
268
When the crack extends further, the required energy is provided by the stored energy,
which can be expressed mathematically as in equation [7-3] based on the assumption
that the energy consumed in crack propagation consisted of the surface energy only.
𝑑𝑈𝑑𝑎
=𝑑𝑊𝑑𝑎
[7-3]
By substituting equations [7-1] and [7-2] in equation [7-3] and rearranging it for the
stress as in equation [7-4] this gives the stress required for crack opening according to
Griffith as:
𝜎𝑐 = �2𝐸𝜁𝜋𝑎
[7-4]
Two important conclusions can be drawn from equation [7-4]. First, the critical stress
level for a given crack length varies with the material due to the variation of the surface
energy. Second, the critical stress level decreases with increase of the crack length and
the critical stress level is inversely proportional to the square root of the prevailing crack
length. Another important observation is that the crack would not propagate if the initial
crack length is zero since the stress required becomes infinite. This means that LEFM
cannot predict the crack initiation in its pure from.
Modelling of desiccation crack depths incorporating soil fracture energy 269
2a
σ
σ
x
y r
θ
Figure 7-1 Semi infinite plate with the central crack of the length 2a and the directions for near tip stress field
7.2.2 Irwin's modification
Griffith’s findings are based on very brittle materials such as glass. Hence, for ductile
materials the sole consideration of surface energy for cracking comprises the results. As
a remedy for this drawback, other parameters were introduced such as critical energy
release rate 𝐺𝐼𝐶 and stress intensity factor 𝐾𝐼𝐶.
7.2.2.1 Critical energy release rate
The critical energy release rate is a material property and within the elastic range it is
constant and can be graphically shown in a diagram as shown in Figure 7-2. The
shaded area illustrates the potential energy change due to the crack opening for ∆𝒂
length. The potential energy change is the difference between the external work done
and the stored but recoverable elastic strain energy. The energy release rate (ERR) can
be mathematically expressed as:
270
𝐺𝐼𝐶 =𝑑𝑈𝑑𝑎
=𝜋𝜎𝑐
2𝑎𝐸
[7-5]
By rearranging the terms the relation for the critical stress can be given as:
𝜎𝑐 = �𝐸𝐺𝐼𝐶
𝜋𝑎
[7-6]
This equation describes the inter-relation between the material property, stress level and
the crack size by means of critical ERR, critical stress and crack length. Furthermore, it
gives the advantage of omitting the surface energy parameter and hence omitting the
error of not considering the plastic flow near the crack and associated non linearity.
Load
DisplacementO
a
a + Δa
A
Area representing
energy supply for fracture
B
Area representing elastic strain
energy
Load-displacement
curve
Figure 7-2 Actual incremental fracture process in Load-Displacement space
Modelling of desiccation crack depths incorporating soil fracture energy 271
7.2.2.2 Stress Intensity Factor (SIF)
When a body containing a crack is subjected to tensile stresses, a strong stress
concentration develops around the crack tip. This stress concentration varies in a
manner inversely proportional to the square root of the length from the crack tip
regardless of the shape and boundary conditions, on which, however, the intensity of the
stress concentration will depend. For the same intensity, the stresses around and close to
the crack tip are identical (Bažant and Planas, 1998). The intensity of stress is greatly
affected by the mode of the fracture also. However, in the present research, only mode I
fracture associated with pure tensile loading is considered as applicable to desiccation
cracking predominantly.
The in plane stress acting at a point ahead of the crack tip was described by Irwin using
the stress intensity factor as,
𝜎𝑖𝑗 =𝐾𝐼
√2𝜋𝑟𝑓𝑖𝑗(𝜃) [7-7]
The expression 𝑓𝑖𝑗(𝜃) is a known function of 𝜃 and 𝑟 and 𝜃 are the cylindrical polar
coordinates as shown in Figure 7-1. 𝐾𝐼 is the stress intensity factor for mode I crack
opening. This factor 𝑓𝑖𝑗(𝜃) is dimensionless and hence is independent of stress and the
size of the structure except the shape factors of the geometry.
Clearly the actual stress acting on the body is greater than the stress given by equation
[7-7]. Hence the solution is valid only near the crack tip, which is referred as the
fracture process zone as shown in Figure 7-3.
272
The expression to obtain the SIF changes with the geometry of the structure. The
expressions for most possible problems have been summarised by Sih (1973). The SIF
in an infinite body as shown in Figure 7-1, can be expressed as:
𝐾𝐼 = 𝜎√𝜋𝑎 [7-8] St
ress
Fracture Singularity dominated zone
Distance
KI/√2πr
σ (theoretical with externally applied load)
Stress limit due to plastic zone
Figure 7-3 Distribution of the stress normal to the crack plane (Wang, 1996)
7.2.2.3 The relationship between the SIF and ERR
Since the near tip stress field is unique to the material of the structure and the energy
flow rate into the crack tip should depend on this stress field, a unique relationship
between SIF and ERR can be derived.
Simply by equating equation [7-6] and [7-8], the expression for critical conditions can
be derived as,
Modelling of desiccation crack depths incorporating soil fracture energy 273
𝜎𝑐 = �𝐸𝐺𝐼𝐶
𝜋𝑎 =𝐾𝐼𝐶
√𝜋𝑎
𝐾𝐼𝐶2 = 𝐸𝐺𝐼𝐶 [7-9]
This expression applies in the plane stress condition and if the plane strain conditions
are considered equation [7-9] should be adjusted to,
𝐾𝐼𝐶2 =
𝐸𝐺𝐼𝐶(1 − 𝜈2) [7-10]
These equations (equation [7-9] and [7-10]) can also be used for the other conditions
less than the critical condition. Hence they can be identified as the general relationship
connecting the SIF and ERR.
7.3 Past Approaches for Numerical Modelling of Fracture
As discussed in the Chapter 2, several numerical models have been used by researchers
to model the desiccation process. However, LEFM and the cohesive crack method
appear to be leading in the state of the art fracture analysis.
7.3.1 Numerical modelling attempts using LEFM
Many computer programs used for fracture modelling (Haberfield, 1987; Haberfield,
1990; Lim et al., 1994) have considered the fracture toughness and the SIF at the crack
tip. On most occasions, the fracture was allowed to open when the measured SIF at the
crack tip stress field has reached the critical value. When the critical SIF (𝐾𝐼𝐶 ) is
involved in the numerical analysis, a finer mesh is often required near the crack tip and
274
re-meshing near the crack tip is required with the advancement of the fracture. This can
be very time-consuming and computationally expensive. Furthermore, it assumes that
the applicability of material linear elasticity in the presence of an infinite tensile stress at
the crack tip. Hence it ignores the presence of a finite tensile strength for the material.
LEFM is more valid for brittle soils (Prat et al., 2008). It assumes infinite tensile
stresses near the crack tip, which is not valid when the material has limited tensile
strength and displays significant plastic behaviour around the crack process zone. The
significant plastic behaviour in clay beams as experienced in tests by (Hallett and
Newson, 2005) shows the requirement of Elasto-Plastic Fracture Mechanics (EPFM) for
some soil consistencies.
7.3.2 Modelling attempts with cohesive crack
As the drawbacks and difficulties identified in fracture mechanics approach, in
geotechnical engineering a cohesive crack approach was introduced as a powerful
alternative for fracture modelling.
The cohesive crack method was introduced in the early 1960s by Dugdale (Dugdale,
1960) and Barenblatt (Barenblatt, 1962). It was introduced to represent different
nonlinear processes occurring at the front of a pre-existing crack of quasi-brittle
materials by which the peak and post-peak fracture behaviour can be determined. In the
late 1970s, this method was extended by proposing that the cohesive crack may be
assumed to develop anywhere, even if no pre-existing macro crack is actually present as
considered before (Hillerborg et al., 1976; Modeer, 1979).
Modelling of desiccation crack depths incorporating soil fracture energy 275
w
FPZ
0
Stress
Distance from the crack tip
Tensile strength
Figure 7-4 Bridging stresses at the crack tip while crack opening considered in the cohesive crack models
In the cohesive crack model, the entire fracture process zone is considered to be lumped
into the crack line and is characterised in the form of a stress-displacement law which
exhibits softening. In other words, the cohesive crack law lets the bridging stresses
across a crack that is opening up drop progressively from the tensile strength to zero as
the two crack faces move apart, as shown in Figure 7-4. Hence it allows for the
recognition of the finite tensile strength as well as the plastic energy dissipation in the
crack-processing zone.
Recently Amarasiri et al. used this cohesive modelling approach for soft rock
modelling. They used a cohesive model featuring a bilinear softening curve to model the
mode I fracture in a three point bending test in a single edge notched beam (Amarasiri
276
and Kodikara, 2011b). The change of displacement with the applied load at the crack
mouth which was obtained from the UDEC modelling was compared with the
laboratory results and found good agreement. They have concluded that the cohesive
law approach is a convenient and rational method for fracture modelling in soft rock
geo-mechanics.
Later Amarasiri has successfully reproduced another three point bending test laboratory
results for clay beam (Amarasiri and Kodikara, 2011a) confirming the applicability of
the cohesive modelling approach for fracture modelling in geo materials.
7.4 Modelling Crack Depths with Cohesive Properties
7.4.1 Cohesive crack implementation
Cohesive crack models are used in fracture modelling since they describe in full the
progressive cracking process. Different analytical softening curves can be used
depending on the different softening behaviours in various materials used for the
analysis, which can be referred as rectangular, linear and bilinear (Bažant and Planas,
1998). These typical curves are shown in Figure 7-5. The fractures in compacted clay
layers follow the linear softening curve according to the findings of Amarasiri et al.
(2010a). Therefore, a linear softening law was described in the following models.
Modelling of desiccation crack depths incorporating soil fracture energy 277
Dugdale
Crack opening (w)
Linear
Bilinear
Nor
mal
Stre
ss (σ
)
Rectangular
Figure 7-5 Softening curves
The constitutive model for the block material was selected as linear elastic with the
same soil parameters used in the previous sections. Then the joint model was selected as
the area contact-Coulomb slip with residual strength. Linearly decreasing suction profile
(suction profile 2) was used in cumulative incremental fashion as discussed in previous
sections.
The tensile strength of the joint is considered as a function of suction similarly to
previous sections. The cohesive law used in the numerical model is shown in Figure
7-6. The cohesive law is applied to the model in several steps in the FISH code.
According to the tensile stress at the point the fracture toughness has been calculated
using the power law equation
𝐾𝑎𝑝𝑝 = 0.154𝜎𝑡1.022
[7-11]
as given in Amarasiri et al.(2010a). Then the fracture energy is calculated from Irwin’s
equation,
278
𝐺𝑓 =𝐾𝑎𝑝𝑝
2
𝐸
[7-12]
Crack opening (w)
Nor
mal
Stre
ss (σ
)
b
yield limit
ca residual
limit
Figure 7-6 Linear softening law used in the numerical model
The gradient of line ab in Figure 7-6 is the joint normal stiffness from which the yield
limit can be calculated. Fracture energy (𝐺𝑓) is the area under the abc curve in Figure
7-6. Hence from the geometry of the curve the residual limit can be calculated. Since
the yield limit and residual limit is known, the tensile stress can be changed according to
the softening curve. As the normal displacement exceeds the yield limit the tensile stress
will be changed according to the softening tensile value from the bc arm of Figure 7-6.
Also when the normal displacement of the joint exceeds the residual limit, the tensile
stress can be assigned as zero.
The use of cohesive cracks to model desiccation cracking is more complex than in
cracking where material properties are held constant because the cohesive softening
Modelling of desiccation crack depths incorporating soil fracture energy 279
laws themselves may change while the crack is opening up. This changing process has
been described in detail by Kodikara and Amarasiri (1961).
7.4.2 Compacted clay soils
7.4.2.1 Using the linearly decreasing suction profile
According to the results of Chapter 3, the UDEC program generally produces higher
crack depths than the analytical solutions which are similar to the predictions from
LEFM approach. By this method it was decided to apply fracture energy externally to
the crack tip. Hence several models were run with cohesive law implemented to the
joint.
To implement the cohesive law, the linear softening law was used. And the suction
profile was applied as in Chapter 3. The material properties and boundary conditions
were the same as in the models in Chapter 3 excepting the cohesive properties at the
crack.
7.4.2.2 Using observed suction profiles
Following from Chapter 6, development of cracks under different climate conditions
was analysed incorporating soil fracture energy. The analysis method was improved by
incorporating cohesive properties for Regina, Altona and Horsham clay soils. The
procedure used for applying cohesive crack was similar to the method described in the
Section 3.7 in the Chapter 3. However, two softening laws were followed in this section,
280
namely, the Dugdale softening law and linear softening law. The overall method used in
the numerical model is described briefly in the following paragraph.
The fracture energy was calculated on the basis of the tensile strength (Equation 7-11
and 7-12). Then the yield limit as shown in Figure 7-6 was calculated by dividing the
tensile strength by the normal stiffness and then the residual limit was calculated using
the fracture energy. Using these values the residual tensile stress is calculated and
assigned to the crack as a compression force. The stress change due to the suction
change was calculated as in equation [5-10] (in Chapter 5) and applied progressively to
the block of length 40m and height 8m.
7.4.3 Modelling crack depths in soft soils
In order to demonstrate the modelling of soft natural soils, the field test undertaken by
Konrad and Ayad (1997a) was used. However, only one crack was allowed to propagate
in order to compare the effect of cohesive properties on the joint.
7.4.3.1 Test results and soil parameters
Konrad and Ayad (1997a) used the natural clay at Saint-Alban in Saint Lawrence
Valley for the desiccation test. The test was carried out for the clay soil at 2m below the
ground level since the clay layer was almost homogeneous between 2m and 7m depths
from the ground according to the vane and piezocone penetration tests. The moisture
contents at various depths were obtained using time domain reflectometry (TDR) probes
Modelling of desiccation crack depths incorporating soil fracture energy 281
up to 2.7m depth. Then the moisture content profiles were recorded with time as shown
in Figure 7-7.
Figure 7-7 Moisture content profiles of Saint-Alban clay test (Konrad and Ayad, 1997a)
The clay between 1.9m and 2.7m depth showed 25% and 50% plastic and liquid limits
respectively with gravimetric moisture content greater than 100%, hence could be
considered as soft clay. The Poisson’s ratio for Saint Alban clay was considered as 0.3
and the Young’s modulus at 2m depth was selected as 5MPa according to the recorded
range from 4MPa to 6MPa (Ayad et al., 1997). The specific gravity of the soil was 2.8
(Konrad and Seto, 1994).
0
10
20
30
40
50
60
70
80
0 20 40 60 80 100 120 D
epth
from
2m
bel
ow th
e su
rfac
e (c
m)
Moisture content (%)
t=0h
t=18h
t=24h
t=42h
t=65h
t=73h
t=97h
t=145h
t=193h
t=241h
282
7.4.3.2 Material properties in the numerical analysis
Similarly to the previous chapter (Chapter 6) the suction was obtained from the SWCC
as given by Konrad and Ayad (1997a) for Saint-Alban clay. The empirical equations
used for inputting the soil characteristics in the numerical model for different moisture
content ranges are shown in Figure 7-8. Here the gravimetric moisture content was used
instead of volumetric moisture content given in the above reference.
The Young’s modulus of the Saint-Alban clay was selected as 5MPa. However, due to
the fact that the soil properties changed with the moisture content, it was decided to
incorporate this change in the numerical model, although it had not indicated a
considerable effect in previous analyses undertaken on other soils.
Figure 7-8 SWCC for Saint-Alban clay and the empirical equations for different moisture content ranges
0
20
40
60
80
100
120
1 10 100 1000
Gra
vim
etric
moi
stur
e co
nten
t, w
(%)
Suction, S (kPa)
S = 3843.4e-0.109w R² = 0.9975
S = 520.78e-0.03w R² = 0.9993
S = 19727e-0.08w R² = 0.9746
25.5
74.5
Modelling of desiccation crack depths incorporating soil fracture energy 283
The Young’s modulus increase with the increase of the suction (Kodikara et al., 2004)
and the following relation was proposed for Werribee clay,
𝐸 = 24(1 − 2𝜈)(𝑆)0.92 [7-13]
The Werribee clay testing was undertaken from a slurry state, whereas the Saint-Alban
clay had some strength and stiffness at the beginning. Hence, equation [7-13] was
modified to incorporate the initial conditions and the new equation is given as:
𝐸 = 𝐸0 + 24(1 − 2𝜈)(𝑆 − 𝑆0) [7-14]
where, 𝐸 is the Young’s modulus, 𝐸0 is the initial Young’s modulus which was selected
to be 5MPa, 𝜈 is the Poisson’s ratio and 𝑆0 is the initial suction which is taken to be
5.6kPa calculated from the SWCC.
The variation of tensile strength with drying was also considered. It is reasonable to
assume that tensile strength increases linearly with the suction (Morris et al., 1992;
Nahlawi, 2004). According to Ayad (1997), the tensile strength of the material at the in
situ moisture content was about 9kPa, as obtained by pressure testing on cylindrical
specimens. Taking the soil suction at this point to be 5.6kPa as consistent with the
SWCC, the following relation was derived:
𝜎𝑡 = 1.74𝑆 [7-15]
The fracture energy of the Saint-Alban clay soil was calculated from the experimental
data presented in Ayad et al. (1997). This was calculated by dividing the work done on
the specimen by the surface area fractured. The work done on the specimen was
obtained from the area under the load displacement curve. Although the fracture energy
obtained through this method is a rough estimation, the calculated average value from
284
three samples was considered as reasonable. The fracture energy value of 0.74N/m was
obtained for the clay at the in situ moisture content to be used in the numerical analysis.
7.4.3.3 Model implementation and applying stresses
After deciding the initial soil parameters and their variations with others, it was required
to implement those in the numerical program. In UDEC the implementation of the stress
and other soil parameter variation can be done using the in-built language FISH. The
method of applying stress change is described below.
The successive moisture content profiles shown in Figure 7-7 as drying progressed were
implemented progressively in the numerical model. The soil properties such as Young’s
modulus and tensile strength were modified according to the moisture content change or
the corresponding suction change, as described above. The suction due to the moisture
content change was applied as a stress change to the soil layer.
The stress increment in the model was applied similarly to the method used in Chapter
5. However, equation [5-10] that was used to calculate the stress increment in Chapter 5
was modified to incorporate moisture content instead of void ratio in this series of
models. The modified equation is given as:
∆𝜎 =𝐸
3(1 − 2𝜈)∆𝑤𝐺𝑠
1 + 𝑤𝐺𝑠 [7-16]
This stress increment was implemented in the FISH code according to the moisture
content profiles. A soil block of 3m length and 1.2m height was selected for the
Modelling of desiccation crack depths incorporating soil fracture energy 285
numerical model with a very fine mesh. A linear softening cohesive law was used to
manage residual tensile stresses while opening the crack.
7.5 Predicting Crack Depths in Compacted Clay Soils
After implementing the model as described in the section 7.4.1, the results for the crack
depths were obtained. The effect of the different implementations of cohesive law and
the effect of the shear angle with respect to the suction were checked and the results are
discussed below.
7.5.1 Results obtained using the linearly decreasing suction profile
The results obtained from the cohesive model are compared with the previously
obtained similar linear elastic model results as shown in Figure 7-9 and Figure 7-10
(LE-UDEC curve and cohesive-UDEC (normal) curve). The implementation of
cohesive crack does not seem to change the results significantly especially when the
watertable is smaller. Hence, to make sure that the cohesive part of the code is working
properly, a series of models was analysed by making artificially high fracture energy
values. The results shown in Figure 7-9 for cohesive-UDEC (high) indicate the fracture
energy to be 100 times higher than it should be according to equation [7-11].
When the fracture energy is extremely high, as expected, the crack length is low
compared with the previous values. The higher fracture energy increases the residual
286
limit, that is the gradient of bc line in Figure 7-6 also increases. Hence the opening of
the crack makes much more difficult resulting in lesser crack depths.
Figure 7-9 Depth of cracking variation for cohesive model with surface suction when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛 𝜙𝑏 𝑐𝑜𝑡 𝜙′, 𝑊 =4m and 𝜙 =300
Figure 7-10 Depth of cracking variation for cohesive model with depth to water table when 𝐸 =5MPa, 𝜈 =0.3, 𝜎𝑡 = −𝛼𝑇𝑆 𝑡𝑎𝑛 𝜙𝑏 𝑐𝑜𝑡 𝜙′, 𝑆0=50kPa and 𝜙 =300
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
20 40 60 80 100
Dep
th o
f Cra
ckin
g (m
)
Surface Suction (kPa)
LE-UDEC Cohesive-UDEC (Normal) Cohesive-UDEC (high)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
2 4 6 8 10
Dep
th o
f Cra
ckin
g (m
)
Depth to Water Table (m)
LE - UDEC Cohesive-UDEC (normal)
Modelling of desiccation crack depths incorporating soil fracture energy 287
7.5.2 Effects of method of implementing cohesive law
The cohesive law was implemented in the numerical code using a large number of steps.
This final cohesive profile was achieved through three different ways as shown in
Figure 7-11.
A AA C
B
B1
B1
B B
B2
B2
B3B3
C3 C3C2 C2C1C1 CCDisplacement
Tens
ile S
tress
(a) (c)(b)
Figure 7-11 Different methods of implementing cohesive law (a) Softening Law change-1 (b) Softening Law change-2 (c) Softening Law change-3
Figure 7-11 illustrates the progressions of cohesive laws to the final stage shown as
ABC curve. B1, B2, B3, C1, C2 and C3 are intermediate positions during the stepping
process in the numerical model. In the final curve given as ABC, the point B represents
the yield limit on the x axis and the tensile strength on the y axis. The point C is the
residual limit on the displacement axis. In Figure 7-11(a), both the yield limit and
residual limit were changed in the numerical process to reach the final ABC while
maintaining a constant stiffness (i.e. the gradient of the AB line). This approach was
named Softening Law change (SL change)-1. In Figure 7-11(b) both the yield limit and
the residual limit were kept constant. Only the tensile strength and the stiffness were
changed. This approach was referred to as Softening Law change (SL change)-2.
Finally, as shown in Figure 7-11(c), the stiffness, tensile strength and the yield limit
288
were maintained unchanged during stepping and only the residual limit was increased
until the final curve was reached. This approach was called Softening Law change (SL
change)-3 for reference. The depths of cracking obtained through these approaches were
shown in Figure 7-12.
Figure 7-12 Depth of cracking for different implementations of softening law with the change of surface suction
In Figure 7-12, SL change 1, 2 and 3 refers to the above mentioned three softening law
implementing approaches and no SL refers to the crack model without any user defined
cohesive properties at the crack opening. As Figure 7-12 illustrates all models show
similar results for crack depths through each approach despite the without cohesion
approach. However, with further drying, the curves tend to separate from each other
0
1
2
3
4
5
6
7
8
0.0 5.0 10.0 15.0 20.0
Dep
th o
f cra
ckin
g (m
)
Maximum surface suction (MPa)
H-no SL H-SL change-1 H-SL change-2 H-SL change-3
Modelling of desiccation crack depths incorporating soil fracture energy 289
although that difference is not more than 100mm. From this result, it was decided that
the method of softening law implementation is not significant and softening law change-
1 approach was used in all the later modelling, since it can be considered as the most
realistic approach among the three approaches as the tensile strength increases with the
suction.
7.5.3 Effect of friction angle with respect to suction
In previous sections of this thesis, the friction angle with respect to suction, 𝜙𝑏, was
considered as a constant based on Morris et al. (1992). However, it was based on a
broad interpretation of data published in literature and may not be relevant for the entire
range of suction.
Bishop (1959) proposed an equation for shearing resistance using the effective stress
based Mohr–Coulomb failure criterion given in equation [7-17]. A 𝜒 parameter was
used to modify the suction value that contributes to the effective stress.
𝜏 = 𝑐΄ + [(𝜎 − 𝑢𝑎) + 𝜒(𝑢𝑎 − 𝑢𝑤)𝑡𝑎𝑛𝜙΄] [7-17]
Fredlund et al. (1978) used an approach that proposed another relation for shear strength
with normal stress and matric suction. This relation is given as:
𝜏 = 𝑐΄ + (𝜎 − 𝑢𝑎) tan 𝜙΄ + (𝑢𝑎 − 𝑢𝑤)𝑡𝑎𝑛𝜙𝑏 [7-18]
By rearranging equation [7-18],
𝜏 = 𝑐΄ + �(𝜎 − 𝑢𝑎) + (𝑢𝑎 − 𝑢𝑤) �𝑡𝑎𝑛𝜙𝑏
tan 𝜙΄�� tan 𝜙΄
[7-19]
290
By comparing equation [7-17] with equation [7-19],
𝜒 =𝑡𝑎𝑛𝜙𝑏
tan 𝜙΄
[7-20]
Khalili and Khabbaz (1998) proposed a relationship for χ parameter to determine the
shear strength in unsaturated soils as given below.
𝜒 = �(𝑢𝑎 − 𝑢𝑤)
(𝑢𝑎 − 𝑢𝑤)𝑏�−0.55
[7-21]
where, (𝑢𝑎 − 𝑢𝑤)𝑏 is the suction at the air entry. On the basis of experimental data, they
indicated that 𝜒 is not a constant (so that the internal angle of friction with respect to
matric suction) and decreases with increasing matric suction.
Figure 7-13 Effect of friction angle with respect to suction on desiccation cracks
0
1
2
3
4
5
6
7
8
0 2 4 6 8 10
Dep
th o
f cra
ckin
g (m
)
Maximum surface suction (MPa)
A-no SL-Constant ϕ^b A-no SL-Varing ϕ^b A-with SL-Constant ϕ^b A-with SL-Varing ϕ^b
Modelling of desiccation crack depths incorporating soil fracture energy 291
By equating equation [7-20] and [7-21],
𝑡𝑎𝑛𝜙𝑏 = tan 𝜙΄ �(𝑢𝑎 − 𝑢𝑤)
(𝑢𝑎 − 𝑢𝑤)𝑏�−0.55
[7-22]
According to equation [7-22], the internal friction angle with respect to matric suction
was changed in the FISH code to obtain the depths of cracking and the results are shown
in Figure 7-13.
Figure 7-13 has shown the development of depth of desiccation cracking with different
modelling conditions. In the legend the first capital A refers to the soil which was
Altona clay in this scenario, SL is Softening Law, showing the application of cohesive
properties to the crack or not and 𝜙𝑏 is the friction angle with respect to matric suction.
The above figure shows that in both cases, i.e. with and without cohesive crack
properties, changing 𝜙𝑏 does not show any effect at wet stages (lower suctions).
However when the desiccation progresses with higher suction stresses, changing 𝜙𝑏
shows a slight increase in crack depth. With higher suction values, the 𝜒 parameter
decreases and controls the suction effect on the shear strength and the tensile strength
that leads to increase the depth of cracking. However, since the difference between the
two curves is not significant and at low suctions no difference is detected, the effect of
𝜙𝑏 was considered as not significant.
7.5.4 Depth of desiccation cracks with the cohesive properties
The crack depths while desiccating a soil layer were obtained for Regina, Horsham and
Altona clay soils. For each soil the effectiveness of using cohesive properties at the
292
crack was analysed using two different softening laws and then compared with the crack
depths without any cohesive properties. Figure 7-14, Figure 7-15 and Figure 7-16 show
the results for each soil.
Figure 7-14 Depth of cracking for Regina clay using different softening laws with changing surface suction
0
1
2
3
4
5
6
7
8
0.0 2.0 4.0 6.0 8.0 10.0
Dep
th o
f cra
ckin
g (m
)
Maximum surface suction (MPa)
R-no SL R-SL-1 R-SL-2
Modelling of desiccation crack depths incorporating soil fracture energy 293
Figure 7-15 Depth of cracking for Horsham clay using different softening laws with changing surface suction
In all three figures, the highest crack depths were observed when no cohesive properties
were assigned to the crack or, in other words, when there is no actual fracture energy
specified to the crack propagation. On the other hand, with the Dugdale softening law
the lowest crack depths were recorded. These results are acceptable since the fracture
energy tends to reduce the crack depth since some of the strain energy is consumed by
opening the fracture. It is obvious that as shown in Figure 7-16 with the increase of
fracture energy, the depth of cracking reduces. With the use of cohesive properties,
there is a fictitious stress between the cracked surfaces that holds the two surfaces
0
1
2
3
4
5
6
7
8
0.0 2.0 4.0 6.0 8.0 10.0
Dep
th o
f cra
ckin
g (m
)
Maximum surface suction (MPa)
H-no SL H-SL-1 H-SL-2
294
together, which is also referred to as residual tensile strength. Therefore, this residual
strength limits the development of deep cracks.
Figure 7-16 Depth of cracking for Altona clay using different softening laws with changing surface suction
(In the legends of Figure 7-14, Figure 7-15 and Figure 7-16, first capital letter refers to the name of the soil such as A for Altona clay, H for Horsham clay and R for Regina clay. Then the SL-1 and 2 are the two softening laws where SL-1 refers to the Dugdale softening law and SL-2 refers to the linear softening law. Finally no SL is the control model where no softening law was used, in other words no cohesive properties were given by the user to the numerical program.)
However, it is noticeable for the above figures that the cohesive properties at the crack
show a considerable effect only at low suctions. When the soil is subjected to higher
0
1
2
3
4
5
6
7
8
0.0 2.0 4.0 6.0 8.0 10.0
Dep
th o
f cra
ckin
g (m
)
Maximum surface suction (MPa)
A-no SL A-SL-1 A-SL-2
Modelling of desiccation crack depths incorporating soil fracture energy 295
suctions the effect of cohesive properties appears to become negligible. It may be that
with high suctions the soil will develop high strain energy to propagate the crack and
the fracture energy may not have a big influence on the final crack depth.
Displacement
Ten
sile
str
ess
Without cohesive properties Linear softening law(SL-2)
Dugdale softening law (SL-1)
Fracture energy Fracture energy
Fracture energy
Yield limit
Yield limit
Yield limit
Residual limit
Figure 7-17 Fracture energies with different softening laws
With the fracture energies used, the predicted crack depths do not show considerable
reduction compared to the crack depths without cohesive properties especially at higher
suctions. Hence to confirm the effect of fracture energy in crack development, several
series of models were run by increasing fracture energy artificially.
Generally fracture energy in soils up to about 173N/m (0.173kPa.m) have been reported
according to the literature (Costa, 2009). However for illustration of its effect, from
small to very high fracture energies were used here. Figure 7-18 and Figure 7-19
illustrate the effect of fracture energy in terms of the residual limit and fracture energy
itself. The fracture energy was increased from zero up to 3.5kPa.m in predicting the
crack depths while keeping the other conditions the same.
296
From Figure 7-18 and Figure 7-19, it can be clearly seen that the development of crack
depth is highly dependent on the increase of fracture energy, as expected. The softening
law used in this series of models was the linear softening law as shown in Figure 7-17.
When the fracture is increased, the residual limit increases but not the tensile strength or
the yield limit. Hence, with the increase of fracture energy, the residual strength
between the cracking surfaces increases and the tensile strength does not decrease to
zero until the crack opens wider. For these reasons, the crack depth drops considerably
with the increase of fracture energy. However, although this demonstrates the influence
of fracture energy and residual limit, it should be noted that these results are artificial
since such large residual limits are not practical for clay soils.
Figure 7-18 Effect of fracture energy on desiccation cracking in terms of residual limit
0
1
2
3
4
5
6
7
8
0 50 100 150 200 250
Dep
th o
f cra
ckin
g (m
)
Residual limit (mm)
20MPa 14MPa 10MPa 8MPa
Modelling of desiccation crack depths incorporating soil fracture energy 297
Figure 7-19 Effect of fracture energy on desiccation cracking
(20MPa, 14MPa, 10MPa and 8MPa in above figures, indicate the surface suction value of the soil block.)
7.6 Predicting Crack Depths in Soft soils
Cohesive properties at the crack mostly have an influence when low suctions are
incorporated. Hence, the soft soils response to the cohesive properties would be an
interesting observation.
As discussed in the section 7.4.2, the model was developed to reproduce the test
conducted by Konrad and Ayad (1997a) using slurry clay. However, other than the
0
1
2
3
4
5
6
7
8
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Dep
th o
f cra
ckin
g (m
)
Fracture Energy (kPa.m)
20MPa 14MPa 10MPa 8MPa
298
multiple cracks as observed during the field test, single crack was also used in this
study.
7.6.1 Results for crack depth prediction using single crack
Figure 3-25 shows the depth of the crack predicted with the same moisture content
profiles with time. But a single joint was allowed to open. The progression of crack
depth was modelled according to the increment of suction as given in Figure 7-7
profiles with time. The crack depths were modelled with and without cohesive
properties. As with the previously experience, the crack depths are similar in both cases
especially when the suctions reach higher values.
Figure 7-20 Crack depth with time
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 20 40 60 80 100
Dep
th o
f cra
ckin
g (m
)
Time (hr)
With cohesion
Without cohesion
Modelling of desiccation crack depths incorporating soil fracture energy 299
7.6.2 Results for crack depth prediction using multiple cracks
The block of soil was modelled allowing for multiple cracks to occur. This was
undertaken by providing joints at equal spacing of 50 mm. The distribution of cracks
and the displacements of zones within the block after 18 hours are shown in Figure
7-21. The figure shows only the top left section of the soil block analysed for clarity. As
the cracks progress, the displacement vectors become more lateral from the initial
vertical downward direction with the increasing suction.
Figure 7-21 Block with displacement and joints
300
Figure 7-22 Block with opened cracks and displacement vectors
The condition of soil block after 65 hours is shown in Figure 7-22. It is clear that the
opening of the joints provided is selective, where several cracks at some spacing open
Modelling of desiccation crack depths incorporating soil fracture energy 301
first and these cracks extend deeper before other joints start to open in between. Several
minor cracks also initiated at the surface, but they did not extend further. Some of these
cracks closed as the dominant cracks propagate deeper. This behaviour may be partially
attributed to the current modelling approach, where cracks can open anywhere when the
tensile stress exceeds the tensile strength. However, a similar response is also seen in
field situations, where bifurcation of crack patterns occurs with minor cracks at small
spacing which stop growing and even close giving way to the progression of major
cracks to greater depths.
Figure 7-23 Progression of multiple cracks with time
0
0.2
0.4
0.6
0.8
1
1.2
0 20 40 60 80 100
Dep
th o
f cra
ckin
g (m
)
Time (hr)
maximum crack depth
most obseverd crack depth
302
The depth of the largest crack was recorded with time as the information on primary
crack growth. The depths of the minor cracks were monitored and the average depth
was recorded with time. These results are shown in Figure 7-23.
The primary cracks develop at considerable speed for almost three days and then limit
the progression. The secondary cracks show fast development within the first 24 hours
and subsequently slow down their progression.
7.7 Conclusions
Compacted clay and soft soils from different geographical regions were modelled
incorporating the fracture energy as cohesive properties. The results of these analyses
were compared with the results from numerical models that did not incorporate fracture
energy (i.e., without using cohesive models). In order to model compacted clays with
cohesive cracks, three types of softening laws were used and results were obtained for a
wide range of suctions. Soft soils were analysed with actual ground and suction
conditions observed in the field. A single crack and multiple cracks were used in the
analysis for soft soils.
The softening cohesive laws used included linearly decreasing and constant as described
by Dugdale model. However, experimental evidence indicates that linear softening law
is more appropriate for soil modelling.
When the suctions in the soil become very large, the influence of the soil fracture
energy diminishes and therefore the use of the cohesive model becomes less important.
Hence, the results obtained are similar to models that do not use fracture energy.
Modelling of desiccation crack depths incorporating soil fracture energy 303
Therefore, in modelling crack depths in extremely dry conditions such as in deserts or
during droughts, the use of cohesive properties does not have any considerable effect on
the crack depth predictions. Hence, it is fair to say that the normal fracture analysis with
a tensile criterion would be sufficient to model extremely dry conditions.
However, for low to moderate dry conditions observed generally in temperate or semi-
arid areas, cohesive properties could have a significant influence on the crack depth
predictions. Without cohesive properties, the numerical analysis over-predicts the crack
depth and when the cohesive properties are introduced, the desiccation crack depths are
reduced. It can be noted, that the crack depths predicted using the cohesive cracks are
within the range typically observed in the field, although the field measurements of
depth of cracking are not accurate on most occasions, due to the extreme difficulty in
measuring the actual depths.
Analyses also indicated that the use of single isolated crack in crack depth prediction
could lead to deeper cracks in comparison to crack depths obtained from multiple crack
analysis. In some field conditions, however, it is possible to generate isolated cracks. In
addition, the advantages of using single crack analysis are the ease of modelling and
reduced model run times. It can also be argued that the use of isolated crack depth could
be conservative in design as it provides the maximum crack depth possible. Multiple
cracking can be used as a solution to replicate actual field conditions.
Chapter 8
CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER RESEARCH
The conclusions drawn from the present study are summarised and presented in this
chapter. Following the conclusions, improvements for the research and new pathways
are suggested based on the knowledge gathered from this study.
8.1 Conclusions
1. Desiccation is one of the major problems a clay liner can face due to the fractures.
Due to its complexity, the precise mechanism of cracking has remained
undiscovered for a long time.
2. Desiccation crack modelling related to the moisture content variation during wet-
dry cycles seems to be more accurate and relatively new in this field. No research
has been conducted to predict the desiccation crack depths using numerical model
considering the moisture content variations.
304
Conclusions and Recommendations for Further Research 305
3. Most of the various computer programs used to predict the crack depths in the
literature are not capable enough to use for fracture modelling in geo-materials.
However the computer programme UDEC was found to be suitable software to
reproduce the actual behaviour of soil in the field when fracturing.
4. In UDEC program, the mesh size does not show much effect on the failure stress.
The number of steps used to change the initial conditions does not affect the final
results. A similar result was obtained when changing the damping value.
However, it is recommended to use a higher number of steps while changing the
conditions of the model and more careful consideration for damping should be
given for dynamic model problems.
5. All other parameters such as, crack length, size of the block and scale of the
model geometry have an influence in failure stress, indicating that clay type soils
with higher fracture energies behaves more plastically than following LEFM
theories unless large problem geometries are used. The soils with lower fracture
energies (i.e. higher stiffness) clearly follow the LEFM theories. Therefore,
applying cohesive properties to the fracture opening may increase the accuracy of
the model in desiccation fracture modelling.
6. The numerical program generally predicts higher crack depths than the existing
analytical predictions due to the relaxation of stresses that are concentrated at the
crack tip just after opening a crack.
7. The moisture content can successfully be used to overcome the over-prediction of
crack depths in the numerical program. The hydric co-efficient is required to
apply moisture content changes to the numerical program.
306
8. Expansive soils come to an equilibrate condition where the soil behaves elastically
while wetting and drying. The path followed by the soil to reach the equilibrium
curve depends on initial moisture content, initial dry density, net vertical stress,
composition of the soil and level of wetting and drying processes.
9. The value of hydric coefficient is dependent on the initial condition of the soil
before equilibrium and the applied stress level. After reaching the equilibrium
stage, in each full wetting or drying process, the α* remains at a value of one for
more than 80% except during the beginning and the end of the process.
10. The cracks seem to grow deeper very quickly once they open up at the surface
within a small amount of time. The closure is also fast. The climate condition of
the area has a great influence on the depth of cracking. The lower Poisson’s ratios
increase the crack depth. However, the deviation of equilibrium suction from the
surface suction towards the wet side decreases the crack depth.
11. To apply cohesive properties to the fractures in soil, linear softening law is more
appropriate according to the experimental and numerical evidences. The fractures
in soils that are subjected to extremely dry condition are not influenced by
cohesive properties. Hence, normal fracture analysis without cohesive properties
would be sufficient to model extremely dry conditions. However, for typical dry
conditions observed generally in arid or semi-arid areas, cohesive properties can
have a significant influence on the crack depth prediction.
12. Both single and multiple conditions can be used to model crack depths effectively,
however, using multiple cracks is recommended only if the spacing of the cracks
is known. Modelling using single crack can be recommended as a safer method to
Conclusions and Recommendations for Further Research 307
be used in design calculations, since this method leads to the maximum crack
depths possible.
13. The predicted crack depths may vary from the actual observed values. Becuase, in
actual conditions continuous desiccation or continuous wetting is highly unlikely
and other disturbing activities for cracking can also happen, such as animals
movements, plastic flow within the soil and erosion by rain or wind. The lack of
accurate field observation data for depth of desiccation cracks to compare with the
numerical predictions is a serious limitation.
14. Finally, all clay soils produce similar depths of cracking under the same
conditions. Hence it is reasonable to predict the crack depths for any clay soil
from the results presented here, provided that particle size and other properties of
the soil are similar to the tested soils.
8.2 Future Research Recommendations
Based on the knowledge gathered from this study, the following further steps research
are recommended.
The most important recommendation would be developing a comprehensive data base
of field desiccation tests. The observations of crack initiation, growth measuring the
depth, width and spacing of crack under different climate conditions and soil types
should be recorded.
Considering the difficulty of the above research, laboratory experiments can be planned
using transparent boxes in order to measure the correct crack depth. However, the boxes
308
must be large enough to observe the natural crack depths. Further, this laboratory test
should be conducted with different soil types replicating different climate conditions.
The effect of cyclic climate variation on depth of cracking need to be monitored either
using a laboratory test or using a field test. Development of the crack and the healing
behaviour should be modelled physically for several wet and dry cycles.
The numerical model proposed in this study should be developed further to model the
spacing of the surficial cracks and the precise sequence of the crack development.
Further, the behaviour of the cracks under several wet dry cycles should be modelled.
Three dimensional numerical modelling also should be developed for a better
representation of the actual field situation.
A 3-D numerical model should be developed by coupling the atmospheric and climatic
conditions and the fracture properties giving more realistic results.
References 309
REFERENCES
Adu-Wusu, C.; Yanful, E. K.; Lanteigne, L. and O'Kane, M., (2007). Prediction of the water balance of two soil cover systems. Geotechnical and Geological Engineering 25(2): 215-237.
Albrecht, B. A. and Benson, C. H., (2001). Effect of desiccation on compacted natural clays. Journal of Geotechnical and Geoenvironmental Engineering 127(1): 67-75.
Albright, W. H.; Benson, C. H.; Gee, G. W.; Abichou, T.; Tyler, S. W. and Rock, S. A., (2006). Field performance of three compacted clay landfill covers. Vadose Zone Journal 5(4): 1157-1171.
Albright, W. H.; Benson, C. H.; Gee, G. W.; Roesler, A. C.; Abichou, T.; Apiwantragoon, P.; Lyles, B. F. and Rock, S. A., (2004). Field water balance of landfill final covers. Journal of Environmental Quality 33(6): 2317-2332.
Alonso, E. E.; Gens, A. and Hight, D. W., (1987). Special problem soils. General report. proceedings of the 9th European Conference on Soil Mechanics and Foundation Engineering, Dublin 1087-1146.
Amarasiri, A. and Kodikara, J., (2011a). Use of Material Interfaces in DEM to Simulate Soil Fracture Propagation in Mode I Cracking. International Journal of Geomechanics 11(4): 314-322.
Amarasiri, A. L.; Costa, S. and Kodikara, J. K., (2010a). Determination of Cohesive Properties for Mode I Fracture from Compacted Clay Beams. Canadian Geotechnical Journal (on review).
Amarasiri, A. L. and Kodikara, J. K., (2011b). Determination of cohesive properties for mode I fracture from beams of soft rock. International Journal of Rock Mechanics and Mining Sciences (1997) 48(2): 336-340.
Amarasiri, A. L. and Kodikara, J. K., (2011c). Numerical Modelling of Desiccation Cracking Using the Cohesive Crack Method. International Journal of Geomechanics.
Amarasiri, A. L.; Kodikara, J. K. and Costa, S., (2010b). Numerical modelling of laboratary tests on desiccation cracking. International Journal of Rock Mechanics and Mining Sciences.
310
Amarasiri, A. L.; Kodikara, J. K. and Costa, S., (2011). Numerical modelling of desiccation cracking. International Journal for Numerical and Analytical Methods in Geomechanics 35(1): 82-96.
Anderson, T. L., (1991). Fracture mechanics: fundamentals and applications. Boca Raton, Florida, CRC Press.
AS1289.2.1.1, (2005). Methods of testing soils for engineering purposes - Soil moisture content tests - Determination of the moisture content of a soil - Oven drying method (standard method). Standards Australia, Sydney, Australia.
AS1289.3.1.1, (2009). Methods of testing soils for engineering purposes - Soil classification tests - Determination of the liquid limit of a soil - Four point Casagrande method. Standards Australia, Sydney, Australia.
AS1289.3.2.1, (2009). Methods of testing soils for engineering purposes - Soil classification tests - Determination of the plastic limit of a soil - Standard method. Standards Australia, Sydney, Australia.
AS1289.3.4.1, (2008 ). Methods of testing soils for engineering purposes - Soil classification tests - Determination of the linear shrinkage of a soil - Standard method Standards Australia, Sydney, Australia.
AS1289.3.6.1, (2009). Soil classification tests—Determination of the particle size distribution of a soil - Standard method of analysis by sieving. Standards Australia, Sydney, Australia.
AS1289.5.1.1, (2003 ). Methods of testing soils for engineering purposes - Soil compaction and density tests - Determination of the dry density/moisture content relation of a soil using standard compactive effort Standards Australia, Sydney, Australia.
AS1289.5.2.1, (2003 ). Methods of testing soils for engineering purposes - Soil compaction and density tests - Determination of the dry density or moisture content relation of a soil using modified compactive effort Standards Australia, Sydney, Australia.
AS1289.6.6.1, (1998). Methods of testing soils for engineering purposes - Soil strength and consolidation tests - Determination of the one-dimensional consolidation properties of a soil - Standard method Standards Australia, Sydney, Australia.
AS2870, S. A. L., (2011). AS2870-2011 Residential slabs and footings. Sydney, Australia.
Aubeny, C. and Long, X., (2007). Moisture diffusion in shallow clay masses. Journal of Geotechnical and Geoenvironmental Engineering 133(10): 1241-1248.
References 311
Ayad, R.; Konrad, J. M. and Soulie, M., (1997). Desiccation of a sensitive clay: application of the model CRACK. Canadian Geotechnical Journal 34(6): 943-951.
Azam, S., (2007). Study on the swelling behaviour of blended clay-sand soils. Geotechnical & Geological Engineering 25(3): 369-381.
Azam, S.; Li, Q.; Shah, I.; Hu, Y. and Chowdhury, R., (2012). Field monitoring of seasonal effects in Regina clay. International Journal of Geoengineering Case Histories. ISSMGE In Revision.
Bagchi, A., (2004). Design of Landfills and Integrated solid waste Management. Hoboken, New Jersey, John Wiley & Sons, Inc.
Bagge, G., (1985). Tension cracks in saturated clay cuttings. Proce. of 11th international conference on soil mechanics and foundation engineering, San Francisco, (Balkema) 393-395.
Baker, R., (1981). Tensile strength, tension cracks, and stability of slopes. Soils & Foundations 21(2): 1-17.
Baker, R. and Frydman, S., (2009). Unsaturated soil mechanics: Critical review of physical foundations. Engineering Geology 106(1-2): 26-39.
Baker, R. and Leshchinsky, D., (2003). Spatial distribution of safety factors: Cohesive vertical cut. International Journal for Numerical and Analytical Methods in Geomechanics 27(12): 1057-1078.
Barenblatt, G. I. (1962). The Mathematical Theory of Equilibrium Cracks in Brittle Fracture. Advances in Applied Mechanics. T. v. K. G. K. F. H. v. d. D. H.L. Dryden and L. Howarth, Elsevier. Volume 7: 55-129.
Basnett, C. and Brungard, M., (1992). The clay desiccation of a landfill composite lining system. Geotechnical Fabrics Report, IFAI 10 (1): 38–41.
Bažant, Z. P. and Planas, J., (1998). Fracture and size effect in concrete and other quasibrittle materials, CRC Press.
Benson, C.; Abichou, T.; Albright, W.; Gee, G. and Roesler, A., (2001). Field evaluation of alternative earthen final covers. International Journal of Phytoremediation 3(1): 105-127.
Benson, C. H. and Daniel, D. E., (1994). Minimum thickness of compacted soil liners: I. Stochastic models. Journal of Geotechnical Engineering - ASCE 120(1): 129-152.
Benson, C. H.; Daniel, D. E. and Boutwell, G. P., (1999). Field performance of compacted clay liners. Journal of Geotechnical and Geoenvironmental Engineering 125(5): 390-403.
312
Bishop, A. W., (1959). The principle of effective stress. Teknisk Ukeblad 106(39): 859-863.
Blake, G.; Schlichting, E. and Zimmermann, U., (1973). WATER RECHARGE IN A SOIL WITH SHRINKAGE CRACKS. Soil Sci Soc Am Proc 37(5): 669-672.
Blight, G. E., (2003). The vadose zone soil-water balance and transpiration rates of vegetation. Geotechnique 53(1): 55-64.
Blight, G. E. and Williams, A. A. B., (1971). Cracks and Fissures by Shrinkage and Swelling. Proceedings of the Regional Conference for Africa - Soil Mechanics and Foundation Engineering 5(1): (1)15.
Bouazza, A. and Van Impe, W. F., (1998). Liner design for waste disposal sites. Environmental Geology 35(1): 41-54.
Buckingham, E., (1914). On Physically Similar Systems: Illustrations of the Use of Dimensional Equations. Phys. Rev. 4: 345-376.
Buckingham, E., (1915). Model Experiments and the Form of Empirical Equations. Trans. ASME 37: 263.
Chan, D., (2012). Study of pipe-soil-climate interaction of buried water and gas pipes. PhD, Monash University.
Chen, F. H., (1988). Foundations on Expansive Soils (Developments in Geotechnical Engineering), Elsevier Science Publishers B.V.
Corser, P.; Pellicer, J. and Cranston, M., (1992). Observations on long-term performance of composite clay liners and covers. GEOTECHNICAL FABRICS REPORT VOL 10(NUMBER 8): 6.
Corte, A. and Higashi, A., (1964). Experimental research on desiccation cracks in soil. Hanover, NH 03755-1290 USA, U.S. Army Cold Regions Research and Engineering Laboratory.
Costa, S., (2009). Study of desiccation cracking and fracture properties of clay soils. Doctor of Philosophy thesis, Monash University.
Costa, S.; Kodikara, J. K. and Thusyanthan, N. I., (2008). Modelling of Desiccation Crack Development in Clay Soils. The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG).
D'Onza, F.; Gallipoli, D.; Wheeler, S.; Casini, F.; Vaunat, J.; Khalili, N.; Laloui, L.; Mancuso, C.; Main, D.; Nuth, M.; Pereira, J. M. and Vassallo, R., (2011). Benchmark of constitutive models for unsaturated soils. Geotechnique 61(Compendex): 283-302.
References 313
Daniel, D. E., (1989). In situ hydraulic conductivity tests for compacted clay. Journal of Geotechnical Engineering 111(8): 957-970.
Daniel, D. E. and Wu, Y.-K., (1993). Compacted clay liners and covers for arid sites. Journal of Geotechnical Engineering 119(2): 223-237.
Dasog, G. S., (1986). Properties, Genesis and Classification of Clay Soils in Saskatchewan. Ph.D. Thesis, University of Saskatchewan.
Day, R. W., (1999). Geotechnical and foundation engineering: design and construction, McGraw-Hill.
Drumm, E. C.; Boles, D. R. and Wilson, G. V., (1997). Desiccation cracks result in preferential flow. Geotechnical News 15(2): 22-25.
Dugdale, D. S., (1960). Yielding of steel sheets containing slits. Journal of the Mechanics and Physics of Solids 8(2): 100-104.
Dwyer, S. F., (1998). Alternative landfill covers pass the test. Civil Engineering 68(9): 50-52.
Dwyer, S. F., (2003). Water balance measurements and computer simulations of landfill covers, University of New Mexico.
Dyer, M.; Utili, S. and Zielinski, M., (2009). Field survey of desiccation fissuring of flood embankments. Proceedings of the Institution of Civil Engineers: Water Management 162(Compendex): 221-232.
EPA, (2008). Best Practice guidelines for landfills accepting category C prescribed industrial waste, Information Bulletin. Publication 1208.
Fayer, M. J., (2000). UNSAT-H Version 3.0: Unsaturated Soil Water and Heat Flow Model, Pacific Northwest National Laboratory, Richland, Washington.
Fredlund, D. G., (2002). use of soil-water characteristic curves in the implementation of unsaturated soil mechanics. UNSAT 2002, Proceedings, Third international conference on unsaturated soil. Recife, Brazil.
Fredlund, D. G. and Anqing, X., (1994). Equations for the soil-water characteristic curve. Canadian Geotechnical Journal 31(4): 521-532.
Fredlund, D. G.; Morgenstern, N. R. and Widger, R. A., (1978). The shear strength of unsaturated soils. Canadian Geotechnical Journal = Revue Canadienne de Geotechnique 15(3): 313-321.
Fredlund, D. G. and Rahardjo, H., (1993). Soil Mechanics for Unsaturated Soils. New York, John Wiley & Sons.
Gallage, C.; Chan, D.; Gould, S. and Kodikara, J. K., (2008). Field Measurement of the Behaviour of an In-Service Water Reticulation Pipe Buried in Reactive Soil
314
(Altona North, Vic). ARC Pipe Linkage Project : Prediction and controlling of pipe failures in buried water and gas pipe systems, Monash University.
Gould, S. J. F.; Kodikara, J.; Rajeev, P.; Zhao, X.-L. and Burn, S., (2011). A void ratio – water content – net stress model for environmentally stabilized expansive soils. Canadian Geotechnical Journal 48(6): 867-877.
Granger, R. J., (1989). An examination of the concept of potential evaporation. Journal of Hydrology 111(1-4): 9-19.
Gray, D. M., (1970). Handbook on the principals of hydrolog.
Griffith, A. A., (1921). The Phenomena of Rupture and Flow in Solids. Philosophical Transactions of the Royal Society of London. Series A, Containing Papers of a Mathematical or Physical Character 221(ArticleType: research-article / Full publication date: 1921 / Copyright © 1921 The Royal Society): 163-198.
Griffith, A. A., (1924). The theory of rupture Selected papers on Foundations of Linear Elastic Fracture Mechanics (A99-25742 06-39), Bethel, CT/Bellingham, WA, Society for Experimental Mechanics/Society of Photo-Optical Instrumentation Engineers (SEM Classic Papers. Vol. CP 1; SPIE Milestone Series. Vol. MS 137), 1997: 96-104.
Haberfield, C. M., (1987). The performance of the pressure meter and socketed piles in weak rock. PhD thesis, Monash University.
Haberfield, C. M. J., I W (1990). Determination of the fracture toughness of a saturated soft rock Canadian Geotechnical Journal 27(3): 276-284.
Haines, W. B., (1923). The volume-changes associated with variations of water content in soil. The Journal of Agricultural Science 13(03): 296-310.
Hallett, P. D. and Newson, T. A., (2001). A simple fracture mechanics approach for assessing ductile crack growth in soil. Soil Science Society of America Journal 65(4): 1083-1088.
Hallett, P. D. and Newson, T. A., (2005). Describing soil crack formation using elastic–plastic fracture mechanics. European Journal of Soil Science 56(1): 31-38.
Hanafy, E. D. E., (1991). Swelling/shrinkage characteristic curve of desiccated expansive clays. Geotechnical Testing Journal 14(2): 206-211.
Harison, J. A. and Hardin, B. O., (1994). Cracking in clays: solutions to problems in earth structures. International Journal for Numerical & Analytical Methods in Geomechanics 18(7): 467-484.
Harison, J. A.; Hardin, B. O. and Mahboub, K., (1994). Fracture toughness of compacted cohesive soils using ring test. Journal of Geotechnical Engineering - ASCE 120(5): 872-891.
References 315
Henken-Mellies, W. U. and Schweizer, A., (2011). Long-term performance of landfill covers - Results of lysimeter test fields in Bavaria (Germany). Waste Management and Research 29(1): 59-68.
Hillel, D., (1980). Fundamentals of soil physics, Academic Press, Inc. (London) Ltd.
Hillerborg, A.; Modéer, M. and Petersson, P. E., (1976). Analysis of crack formation and crack growth in concrete by means of fracture mechanics and finite elements. Cement and Concrete Research 6(6): 773-781.
Holzlohner, U.; August, H. and Meggyes, T., (1997). Water balance, the risk of desiccation in earthen liners.
Holzlohner, U.; August, H.; Meggyes, T. and Brune, M., (1995). Landfill Liner Systems.
Horberg, L., (1951). Intersecting Minor Ridges and Periglacial Features in the Lake Agassiz Basin, North Dakota. The Journal of Geology 59(1): 1-18.
Inci, G., (2008). Numerical Modeling of Desiccation Cracking in Compacted Soils. The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG). Goa, India.
Irwin, G., (1957). Analysis of stresses and strains near the end of a crack traversing a plate. Journal of Applied Mechanics 24: 361–364.
Itasca, (2004). User's Guide, Itasca Consulting Group, Inc., Minnesota, USA.
Ito, M. and Azam, S., (2010). Determination of swelling and shrinkage properties of undisturbed expansive soils. Geotechnical and Geological Engineering 28(4): 413-422.
Jahn, A., (1950). Osobluve Formy Poligonalne Natakachw Dolonie Wieprza. Acta Geologica POlonica Vol. 1(No.2): pp. 150-157.
Jayanth, S.; Iyer, K. and Singh, D. N., (2012). Continuous Determination of Drying-path SWRC of Fine-grained Soils. Geomechanics and Geoengineering: An International Journal In Print.
Jayatilaka, R. and Lytton, R. L., (1997). PREDICTION OF EXPANSIVE CLAY ROUGHNESS IN PAVEMENTS WITH VERTICAL MOISTURE BARRIERS, National Technical Information Service,Springfield, VA 22161 USA: 310.
Jessberger, H. L., (1994). Geotechnical aspects of landfill design and construction. Part 1: principles and requirements. Part 2: material parameters and test methods. Part 3: selected calculation methods for geotechnical landfill design. Proceedings - ICE: Geotechnical Engineering 107(2): 99-122.
316
Khalili, N. and Khabbaz, M. H., (1998). A unique relationship for χ for the determination of the shear strength of unsaturated soils. Geotechnique 48(5): 681-687.
Khire, M. V.; Benson, C. H. and Bosscher, P. J., (1999). Field data from a capillary barrier and model predictions with UNSAT-H. Journal of Geotechnical and Geoenvironmental Engineering 125(6): 518-527.
Knechtel, M. M., (1952). Pimpled plains of eastern Oklahoma. Geological Society of America Bulletin 63(7): 689-699.
Kodikara, J.; Barbour, S. L. and Fredlund, D. G., (1999). Changes in clay structure and behaviour due to wetting and drying. Proceedings of the 8th Australian-New Zealand conference on geomechanics, Hobart, Tasania 179-186.
Kodikara, J. K., (2001). Design of compacted clay liners for waste containment. Geo Environment.
Kodikara, J. K., (2006). Conceptual approaches for considering shrinkage cracking in the cement stabilized pavement mix design. 22nd ARRB Conference. Canberra, Australia.
Kodikara, J. K., (2012). New framework for volumetric constitutive behaviour of compacted 1 unsaturated soils. Canadian Geotechnical Journal (on review).
Kodikara, J. K.; Barbour, L. and Fredlund, D. G., (2000). Dessication cracking of soil layers. proceedings of the asian conference in unsaturated soils, UNSAT ASIA 2000: 693-698.
Kodikara, J. K.; Barbour, L.; Fredlund, D. G. and Choi, X., (2011). Theoretical analysis of desiccation cracking of a long soil layer. On Review.
Kodikara, J. K.; Barbour, S. L. and Fredlund, D. G., (2002). Structure Development in Surficial Heavy Clay Soils: A Synthesis of Mechanisms. Australian Geomechanics: Journal and News of the Australian Geomechanics Society Volume 37(Issue 3 ): 25-40.
Kodikara, J. K. and Choi, X., (2006). A simplified analytical model for desiccation cracking of clay layers in laboratory tests. Geotechnical Special Publication: 2558-2569.
Kodikara, J. K.; Nahlawi, H. and Bouazza, A., (2004). Modelling of curling in desiccating clay. Canadian Geotechnical Journal 41(3): 560-566.
Konrad, J.-M. and Ayad, R., (1997a). Desiccation of a sensitive clay: Field experimental observations. Canadian Geotechnical Journal 34(6): 929-942.
Konrad, J. M. and Ayad, R., (1997b). An idealized framework for the analysis of cohesive soils undergoing desiccation. Canadian Geotechnical Journal 34(4): 477-488.
References 317
Konrad, J. M. and Seto, J. T. C., (1994). Frost heave characteristics of undisturbed sensitive Champlain Sea clay. Canadian Geotechnical Journal 31(2): 285-298.
Konrad, J. M. and Shen, M., (1997). Prediction of the spacing between thermal contraction cracks in asphalt pavements. Canadian Journal of Civil Engineering 24(2): 288-302.
Krahn, J., (2004). Vadose Zone modelling with Vadose/W: An Engineering Methedology, GEO SLOPE International LTD, Alberta, Canada.
Krysanova, V.; Müller-Wohlfeil, D.-I. and Becker, A. (1997, Wed Aug 21 21:44:51 CEST 2002). "The Register of Ecological Models (REM) " General Model Information: SWIM, from http://ecobas.org/www-server/rem/mdb/swim.html.
Kwon, O. and Cho, W., (2011). Field applicability of self-recovering sustainable liner as landfill final cover. Environmental Earth Sciences 62(8): 1567-1576.
Lachenbruch, A. H., (1961). Depth and spacing of tension cracks. Journal of Geophysical Research 66(12): 4273-4292.
Lakshmikantha, M. R., (2009). Experimental and theoretical analysis of cracking in drying soils. PhD Thesis, Technical University of Catalonia.
Lakshmikantha, M. R.; Prat, P. C. and Ledesma, A., (2012). Experimental evidence of size effect in soil cracking. Canadian Geotechnical Journal 49(3): 264-284.
Lau, J. T. K., (1987). Desiccation Cracking of Soils. M. Sc. thesis, University of Saskatchewan.
Leonard, R. A.; Knisel, W. G. and Still, D. A., (1987). GLEAMS: GROUNDWATER LOADING EFFECTS OF AGRICULTURAL MANAGEMENT SYSTEMS. Transactions of the American Society of Agricultural Engineers 30(5): 1403-1418.
Lim, I. L.; Johnstone, I. W. and Choi, S. K., (1994). Assessment for mixed-mode fracture toughness testing for rocks. International Journal of Rock Mechanics and Minning Sciences 31: 265-272.
Longwell, C. R., (1928). Three common types of desert mud cracks. American Journal of Science 15(86): 136-145.
Lytton, R. L., (1994). Prediction of movement in expansive clays, College Station, TX, USA, Publ by ASCE 1827-1845.
Manassero, M.; Benson, C. H. and Bouazza, A., (2000). Solid waste containment systems. Proceedings of Geo Eng2000 1: 520-642.
Melchior, S., (1997). In situ studies of the performance of landfill caps (compacted soil liners, geomembranes, geosynthetic clay liners, capillary barriers). Land Contamination and Reclamation 5(3): 209-216.
318
Miller, C. J. and Mishra, M., (1989). Modeling of leakage through cracked clay liners-1: State of the art. Journal of the American Water Resources Association 25(3): 551-556.
Mitchell, J. K., (1986). Personal Communications on Desiccation Cracks in a Hydraulic Fill Site at Terminal Island in Los Angeles Harbour.
Mitchell, P. W., (1979). The structural analysis of footing on expansive soil. Adelaide, Australia.
Modeer, M., (1979). A fracture mechnics approach to failure analyses of concrete materials. TVBM-1001 Thesis, Lund University of Technology.
Monroy, R., (2006). The influence of load and suction changes on the volumetric behaviour of compacted London Clay. PhD Thesis, University of London.
Montanez, J. E. C., (2002). Suction and volume changes of compacted sand-bentonite mixtures. Ph. D. Thesis, Imperial College of Science, University of London.
Montgomery, R. J. and Parsons, L. J., (1989). The Omega Hills final cap test plot study: three year data summary. Annual Meeting of the National Solid Waste Management Association, National Solid Waste Management Association, Washington DC.
Moore, C. A. and Ali, E. M., (1982). Permeability of Cracked Clay Liners. Proc. 8th Annual Research Symp. on Land Disposal of Hazardous Waste, U.S. Environmental Protection Agency, Municipal Environment Res. Lab., Cincinnati: 174-178.
Morris, C. E. and Stormont, J. C., (1997). Capillary barriers and subtitle D covers: Estimating equivalency. Journal of Environmental Engineering 123(1): 3-10.
Morris, P. H.; Graham, J. and Williams, D. J., (1992). Cracking in drying soils. Canadian Geotechnical Journal 29(2): 263-277.
Morris, P. H.; Graham, J. and Williams, D. J., (1994). Crack depths in drying clays using fracture mechanics. Geotechnical Special Publication, Atlanta, GA, USA, ASCE 40-53.
Murakami, Y., (1986). Stress Intensity Factors Handbook. Pergamon, New York, Pergamon Press.
Nahlawi, H., (2004). Behaviour of a Reactive Soil During Desiccation. M. Sc., Monash University.
Nahlawi, H. and Kodikara, J. K., (2006). Laboratory experiments on desiccation cracking of thin soil layers. Geotechnical and Geological Engineering Journal 24(6): 1641-1664.
References 319
Penman, H. L., (1948). Natural Evaporation from Open Water, Bare Soil and Grass. Proceedings of Royal Society London A: Mathematical, Physical and Engineering sciences. Vol 193. No. 1032: 120-145.
Perera, Y. Y.; Zapata, C. E.; Houston, W. N. and Houston, S. L., (2004). Long-Term Moisture Conditions under Highway Pavements, Los Angeles, California, USA, ASCE 102-102.
Peron, H.; Hu, L. B.; Laloui, L. and Hueckel, T., (2007). Mechanisms of desiccation cracking of soil: Validation, Rhodes 277-282.
Philip, L. K.; Shimell, H.; Hewitt, P. J. and Ellard, H. T., (2002). A field-based test cell examining clay desiccation in landfill liners. Quarterly Journal of Engineering Geology and Hydrogeology 35(4): 345-354.
Picornell, M. and Lytton, R. L., (1987). CHARACTERIZATION OF THE METEOROLOGICAL DEMAND FOR THE DESIGN OF VERTICAL MOISTURE BARRIERS, Transportation Research Board Business Office, Washington, DC 20001, USA.
Potyondy, D. O. and Cundall, P. A., (2004). A bonded-particle model for rock. International Journal of Rock Mechanics and Mining Sciences 41(8 SPEC.ISS.): 1329-1364.
Prat, P. C.; Ledesma, A.; Lakshmikantha, M. R.; Levatti, H. and Tapia, J., (2008). Fracture Mechanics for Crack Propagation in Drying Soils. The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG). Goa, India: 1060-1067.
Priestley, C. H. B. and Taylor, R. J., (1972). On the Assessment of Surface Heat Flux and Evaporation Using Large-Scale Parameters. Monthly Weather Review 100(2): 81-92.
Rajeev, P. and Kodikara, J., (2011). Numerical analysis of an experimental pipe buried in swelling soil. Computers and Geotechnics 38(7): 897-904.
Rice, J. R., (1968a). Mathematical Analysis in the Mechanics of Fracture, Chapter 3 of Fracture: An Advanced Treatise, Academic Press, N.Y.
Rice, J. R., (1968b). A path independent integral and the approximate analysis of stress concentration by notches and cracks. Journal of Applied Mechanics 35: 379-386.
Richards, B. G., (1985). Moisture Flow and Equilibria in Unsaturated Soils for Shallow Foundations. Golden Jubilee of the International Society for Soil Mechanics and Foundation Engineering: Commemorative Volume. Barton, ACT. Institution of Engineers, Australia: 71-101.
320
Richards, B. G.; Peter, P. and Emerson, W. W., (1983). The effects of vegetation on the swelling and shrinking of soils in Australia ( Adelaide). Geotechnique 33(2): 127-139.
Romero, E., (1999). Characterization and thermo-hydro-mechanical behavoiur of unsaturated Boom clay: an experimental study. Ph.D., Technical University of Catalonia.
Russam, K. and Coleman, J. D., (1961). The Effect of Climatic Factors on Subgrade Moisture Conditions*. Geotechnique 11(1): 22-28.
Sadek, S.; Ghanimeh, S. and El-Fadel, M., (2007). Predicted performance of clay-barrier landfill covers in arid and semi-arid environments. Waste Management 27(4): 572-583.
Sawangsuriya, A.; Edil, T. and Benson, C., (2009). Effect of suction on resilient modulus of compacted fine-grained subgrade soils. Transportation Research Record (2101): 82-87.
Schroeder, P. R.; Dozier, T. S.; Zappi, P. A.; McEnroe, B. M.; Sjostrom, J. W. and Peyton, R. L., (1994). The Hydrologic Evaluation of Landfill Performance (HELP) Model: Engineering Documentation for Versoin 3. U.S. Environment Protection Agency Office of Research and Development, Washington, DC.
Shannon, B. M., (2012). Fracture propagation of cohesive soils under tensile loading and desiccation. PhD degree, Monash University.
Sharma, R. S., (1998). Mechanical behaviour of unsaturated highly expansive clays. PhD Thesis, University of Oxford.
Sih, G. C., (1973). Mechanics of Fracture: Methods of Analysis and Solutions of Crack Problems. Noordhoff, Holland, Springer.
Silvestri, V.; Sarkis, G.; Bekkouche, N. and Soulie, M., (1990). Evapotranspiration, trees and damage to foundations in sensitive clays. Canadian Geotechnical Conference: 533-538.
Simon, F. G. and Müller, W. W., (2004). Standard and alternative landfill capping design in Germany. Environmental Science and Policy 7(4): 277-290.
Simpson, W. E., (1936). Foundation Experiences with Clay in Texas, Civil Eng.
Simunek, J.; Huang, K. and Genuchten, M. T. v., (1998). The HYDRAUS Code for Simulating the Movement of Water, Heat and Multiple solutes in Variably Saturated Media U.S. Salinity Laboratory, USDA, ARS, Riverside, California. Version 6.0.
Singh, D. N. and Gupta, A. K., (2000). Modelling hydraulic conductivity in a small centrifuge. Canadian Geotechnical Journal 37(5): 1150-1155.
References 321
Singh, D. N. and J., K. S., (2003). Estimation of Unsaturated Hydraulic Conductivity Using Soil Suction Masurements Obtained By An Insertion Tensiometer. Canadian Geotechnical Journal 40(2): 476-483.
Sivakumar, V., (1993). A critical state Framework for unsaturated soils. Ph.D. Thesis, University of Sheffield.
Skene, J. K. M., (1959). Report on soils of the Horsham soldier settlement project. Soil Survey Report No. 31, Soils Section, Department of Agriculture, Australia.
Smethurst, J. A.; Clarke, D. and Powrie, W., (2006). Seasonal changes in pore water pressure in a grass-covered cut slope in London Clay. Geotechnique 56(8): 523-537.
Strunk, R. L.; Dobchuk, B. S.; Nieminen, G.; Barbour, S. L. and O’Kane, M. A., (2009). Performance monitoring of soil cover field trials at the Regina municipal landfill. Proceedings of SWANA’s WASTECON 2009 Conference. Long Beach, CA.
Sun, J.; Wang, G. and Sun, Q., (2009). Crack spacing of unsaturated soils in the critical state. Chinese Science Bulletin 54(12): 2008-2012.
Tang, C.; Shi, B.; Liu, C.; Zhao, L. and Wang, B., (2008). Influencing factors of geometrical structure of surface shrinkage cracks in clayey soils. Engineering Geology 101(3-4): 204-217.
Terzaghi, K.; Peck, R. B. and Mesri, G., (1996). Soil Mechanics in Engineering Practice (3rd Edition), John Wiley & Sons.
Thakur, V. K. S.; Sreedeep, S. and Singh, D. N., (2005). Parameters Affecting Soil Water Characteristic Curves of Fine-grained Soils. Journal of Geotechnical and Geoenvironmental Engineering, ASCE 31(4): 521-524.
Thornthwaite, C. W., (1948). An Approach toward a Rational Classification of Climate. Geographical Review 38(1): 55-94.
Tripathy, S., (2000). Behaviour of compacted expansive soils under swell–shrink cycles. Ph. D. Thesis, Indian Institute of Science.
Tripathy, S.; Subba Rao, K. S. and Fredlund, D. G., (2002). Water content - Void ratio swell-shrink paths of compacted expansive soils. Canadian Geotechnical Journal 39(4): 938-959.
Vasil'ev, I. M.; Sinyakov, L. N. and Mel'nikov, V. A., (1988). Strength of thawed and frozen soils of dam cores under conditions of triaxial independent loading by tension and compression. Hydrotechnical Construction 22(1): 27-33.
Vu, H. Q.; Hu, Y. and Fredlund, D. G., (2007). Analysis of soil suction changes in expansive Regina clay. 60th Canadian Geotechnical Conference. OttawaGeo 2007, Ottawa, ON: 1-8.
322
Wang, C. H., (1996). Introduction to fracture mechanics. Defence Science and Technology Organisation DSTO-GD-0103.
Wang, J. J.; Zhu, J. G.; Chiu, C. F. and Zhang, H., (2007). Experimental study on fracture toughness and tensile strength of a clay. Engineering Geology 94(1-2): 65-75.
Warkentin, B. P. and Bozozuk, M., (1961). shrinking and swelling properties of two canadian clays. 5th internatinal conference on S.M. and F.E., Dunod, Paris, France 851-855.
Wijesooriya, S. D. and Kodikara, J. K., (2012). Experimental study of shrinkage and swelling behaviour of a compacted expansive clay soil. 11th Australia New Zealand Conference on Geomechanics (ANZ 2012). Melbourne, Australia.
Willden, R. and Mabey, D. R., (1961). Giant desiccation fissures on the Black Rock and Smoke Creek Deserts, Nevada. Science 133(3461): 1359-1360.
Wilson G. W., F. D. G., Barbour S.L., (1994). Coupled soil-atmosphere modelling for soil evaporation. Canadian Geotechnical Journal 31: 151-161.
Wilson, G. W., (1990). Soil evaporative fluxes for geotechnical engineering problem. Ph.D. thesis, University of Saskatchewan, Saskatoon, SK, Canada.
Wilson, G. W.; Fredlund, D. G. and Barbour, S. L., (1994). Coupled soil-atmosphere modelling for soil evaporation. Canadian Geotechnical Journal 31(2): 151-161.
Wilson, G. W.; Fredlund, D. G. and Barbour, S. L., (1997). The effect of soil suction on evaporative fluxes from soil surfaces. Canadian Geotechnical Journal = Revue Canadienne de Geotechnique 34(1): 145-155.
Yanful, E. K.; Mousavi, S. M. and Yang, M., (2003). Modeling and measurement of evaporation in moisture-retaining soil covers. Advances in Environmental Research 7(4): 783-801.
Yesiller, N.; Miller, C. J.; Inci, G. and Yaldo, K., (2000). Desiccation and cracking behavior of three compacted landfill liner soils. Engineering Geology 57(1-2): 105-121.
Yu, C. and Wei, J. C., (2011). Swell-shrinkage behavior of remolded expansive soils. Haikou. 243-249: 3050-3055.
Zein el Abedine, A. and Robinson, G. H., (1971). A study on cracking in some vertisols of the Sudan. Geoderma 5(3): 229-241.
Zhan, T. L. T.; Ng, C. W. W. and Fredlund, D. G., (2007). Field study of rainfall infiltration into a grassed unsaturated expansive soil slope. Canadian Geotechnical Journal 44(4): 392-408.
References 323
Zornberg, J. G.; LaFountain, L. and Caldwell, J. A., (2003). Analysis and design of evapotranspirative cover for hazardous waste landfill. Journal of Geotechnical and Geoenvironmental Engineering 129(5): 427-438.