Upload
lamthuy
View
228
Download
2
Embed Size (px)
Citation preview
Department of Civil Engineering
Monash University
Compaction Behaviourof Soils
by
Nurses Kurucuk
Thesis submitted to Monash University
in partial fulfilment of of the requirements for the degree of
Doctor of Philosophy
January, 2011
c©Nurses Kurucuk
Copyright Notices Notice 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. Notice 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.
Declaration
This thesis contains no material which has been accepted for the award of any other degree or
diploma in any University or other institution. To the best of my knowledge, the thesis contains
no material previously published or written by another person, except where due reference has
been made in the text of the thesis.
Nurses Kurucuk
January 2011
Abstract
Soil compaction is widely applied in geotechnical engineering practice. It is used to maximise
the dry density of soils to reduce subsequent settlement under working loads or to reduce the
permeability of soils. The durability and stability of structures are highly related to the appro-
priate compaction achievement. The structural failure of roads and airfields, and the damage
caused by foundation settlement can often be traced back to the failure in achieving adequate
compaction. For that reason, soil compaction is important for engineering activities involving
earthworks.
Compacted soils are unsaturated by nature, which includes both air and water within their
voids. Thus, unsaturated soil mechanics principles are crucial in understanding the compaction
behaviour of soils. There are several qualitative studies, which attempt to explain the compaction
behaviour of soils and there is a vast body of literature covering the behaviour of compacted
soils. Still, fundamental research on the compaction process is limited. In addition, the current
constitutive models available for unsaturated soils assume that the soil state after compaction is
the initial state of the soil. However, compacted soils undergo a stress history which influences
the post compaction behaviour. Considering these facts, it still remains that the compaction
of soil is a complex phenomenon, which is not well explained, particularly from a quantitative
sense. Further understanding of the compaction behaviour during the compaction process will
provide important insights on the behaviour of compacted soils.
The main aim of this research project is to extend the current understanding of the com-
paction process of soils. The research focuses on three different areas: investigating the experi-
mental behaviour of soils during the static compaction process and obtaining data for compaction
modelling; developing a compaction model using the existing constitutive models for unsaturated
soils; and evaluating the performance of this model in predicting the compaction behaviour of
soils.
In the experimental part, static compaction tests were conducted on two different granular
soils, sand with 2% and 5% bentonite content by weight. The tests were undertaken on samples
with different water contents in order to observe the effect of matric suction on the compaction
behaviour. The initial matric suction of the specimens was measured using the null type axis
translation technique and the matric suction variations were monitored during the compaction
process. It was found that the unsaturated samples were always more compressible than the
saturated sample. This finding is contrary to the assumption made in most constitutive models,
and thus modelling the compaction behaviour using these models may result in some deficiencies.
In addition, in granular soils with low water content the axis translation technique was found to
vii
viii
be very time consuming for the suction measurements. This was attributed to the discontinuous
water phase within the samples.
To develop a compaction model, a volume change constitutive relationship for unsaturated
soils, defined in terms of two independent stress variables, was incorporated with pore pressure
predictions. The model was developed for undrained, semi-drained and drained loading condi-
tions. Initially, compressibility coefficients in the volume change relationship were considered
as constant parameters, i.e., the compressibility of a soil element does not change with increas-
ing vertical stress. Using constant compressibility coefficients, the compaction curve can be
predicted only for the wet side of the curve, not the dry side. Thus, variable compressibility
coefficients were derived from constitutive models proposed in the literature, and using these
coefficients, the well-known shape of the compaction curve was predicted on both dry and wet
side of the compaction curve. It was found that the shape of the compaction curve can be the-
oretically predicted using unsaturated soil mechanics principles. The main insight gained from
the model development was that the influence of matric suction on the material compressibility
with respect to net stress is the governing factor determining the inverted parabolic shape of the
compaction curve.
The performance of the compaction models were examined on their ability to predict the
compaction behaviour of soils. Data for four different soils, two sand-bentonite mixtures tested
in this study, and Boom clay and Speswhite kaolin data from literature, were used for mod-
els evaluation. Two different constitutive modelling approaches were analysed, which are the
separate stress state variables approach and combined stress state approach. It was concluded
that the samples prepared from initially slurry soils and from initially dry soils could not be
treated the same and would require the use of different sets of soil parameters. In addition, the
compaction behaviour of soils, prepared from initially dry samples, could only be modelled over
a narrow range of water contents using a single set of soil parameters. Minimum two sets of soil
parameters are required to model the compaction behaviour over a wide range of water contents
with the current constitutive models.
Acknowledgements
I would like to thank my supervisor Dr. Jayantha Kodikara for his ongoing patience and support
throughout the research project. I am grateful to Dr. Del Fredlund for his encouragement and
the valuable feedback he provided me. I should also mention Dr. Chaminda Galage’s help in
building my experimental setup.
I gratefully acknowledge the financial support received from both the Faculty of Engineering
and the Civil Engineering Department of Monash University. In addition, I acknowledge the
Monash Graduate School for providing me a scholarship in the first three and half years of my
study.
Further thanks are extended to the staff of the Civil Engineering Department. In particular
to Jenny, who was always there whenever I needed advice and support. Long and Mike were
always very helpful in the laboratory. I also appreciate Jane Moody’s meticulous work in editing
parts of my thesis.
I would like to thank Daghan for encouraging me to come to Melbourne, for the great
conversations we had together and helping me a lot in the beginning of my PhD journey. Thanks
Seb for always being very patient with me, for the great holidays we had together and for teaching
me to be a calmer person. I should also mention the contributions of my friends Chris, Scott,
Burcu, and David to my thesis in the review and editing processes. In fact, I am thankful to all
my friends for reminding me that there is life outside university.
I run out of words to express my gratitude to Cintia. She has been not only a dearest friend
but also a mother, sister and editor for me during the whole process. Finally, I would like to
thank my parents for simply being who they are and I am grateful to them for everything they
have done for me.
ix
Contents
Abstract vii
1 Introduction 1
1.1 Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research aim and scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Overview of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Literature review 5
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Qualitative studies for compaction theory . . . . . . . . . . . . . . . . . . . . . . 5
2.2.1 Proctor’s theory (1933) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.2 Hogentogler’s theory (1936) . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.3 Hilf’s theory (1956) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.4 Lambe’s theory (1959) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.5 Olson’s theory (1963) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.6 Barden and Sides’s theory (1970) . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.7 Summary of the compaction theories . . . . . . . . . . . . . . . . . . . . . 9
2.3 Suction measurement techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.1 Direct measurement techniques . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3.2 Indirect measurement techniques . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.3 Summary of the suction measurement techniques . . . . . . . . . . . . . . 16
2.4 Volume change behaviour of unsaturated soils . . . . . . . . . . . . . . . . . . . . 16
2.4.1 Stress variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.2 Volume change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.3 Pore pressure development . . . . . . . . . . . . . . . . . . . . . . . . . . 18
xi
xii CONTENTS
2.4.4 Constitutive models for unsaturated soils . . . . . . . . . . . . . . . . . . 20
2.4.5 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Selected materials for the research 25
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Physical properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.1 Particle size distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.2 Atterberg limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.3 Linear shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.4 Specific gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Compaction characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Soil water retention curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.1 Axis translation technique . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.2 Tensiometer measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5 Consolidation characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.5.1 Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.5.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4 Testing equipment and experimental procedure 45
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.2 Testing Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.2.1 Compaction device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.2.2 Accessories for the compaction device . . . . . . . . . . . . . . . . . . . . 50
4.3 Experimental procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.3.1 Material preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3.2 Static compaction test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3.3 Test difficulties and problems . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5 Experimental results 65
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
CONTENTS xiii
5.2 Initial suction measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.3 Evolution of suction during the test . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.4 Volume change characteristics of soils . . . . . . . . . . . . . . . . . . . . . . . . 77
5.5 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6 Modelling the compaction process 87
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.2 Formulation of the compaction process . . . . . . . . . . . . . . . . . . . . . . . . 87
6.2.1 Volume change calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.2.2 Pore pressure development . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.2.3 Compressibility coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3 Generation of the compaction curve . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3.1 Modelling of the compaction curve using constant compressibility coefficients 97
6.3.2 Modelling of the compaction curve using variable compressibility coefficients104
6.4 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
7 Performance of the model on predicting the static compaction curves 117
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
7.2 Experimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.2.1 Sand bentonite mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.2.2 Boom clay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.2.3 Speswhite kaolin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.3 Separate stress state variables approach . . . . . . . . . . . . . . . . . . . . . . . 121
7.3.1 Sand bentonite mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7.3.2 Boom clay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.3.3 Speswhite kaolin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.3.4 Discussion on the BBM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.3.5 Other models developed with separate stress variables approach . . . . . . 145
7.4 Combined stress variables approach . . . . . . . . . . . . . . . . . . . . . . . . . . 146
7.4.1 Sand bentonite mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.4.2 Boom clay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.4.3 Speswhite kaolin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.4.4 Discussion on the SFG model . . . . . . . . . . . . . . . . . . . . . . . . . 163
xiv CONTENTS
7.4.5 Other models developed with combined stress variables approach . . . . . 164
7.5 Overall discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
7.6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
8 Conclusions and suggestions for future research 171
8.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
8.1.1 Soil behaviour during the compaction process . . . . . . . . . . . . . . . . 171
8.1.2 Modelling the compaction process . . . . . . . . . . . . . . . . . . . . . . 172
8.1.3 Evaluation of the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.2 Suggestions for future research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
References 176
Appendices
A Matlab Codes 187
A.1 Undrained modelling using the BBM . . . . . . . . . . . . . . . . . . . . . . . . . 187
A.2 Semi-drained modelling using the BBM . . . . . . . . . . . . . . . . . . . . . . . 190
A.3 Undrained modelling using the SFG model . . . . . . . . . . . . . . . . . . . . . 193
A.4 Semi-drained modelling using the SFG model . . . . . . . . . . . . . . . . . . . . 196
B Publications 199
B.1 Prediction of compaction curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
B.2 Theoretical modelling of the compaction curve . . . . . . . . . . . . . . . . . . . . 206
B.3 Evolution of the compaction process: Experimental study - preliminary results . . 211
List of Tables
3.1 Atterberg limits of ActiveGel 150 bentonite adopted from Shannon (2008) . . . . 28
3.2 Specific gravity of the materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3 Summary of the compaction characteristics for S2B and S5B . . . . . . . . . . . 32
4.1 Loading steps for S2B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2 Loading steps for S5B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.1 Water content of S2B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2 Water content of S5B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.3 Matric suction change with loading for S2B . . . . . . . . . . . . . . . . . . . . . 75
5.4 Matric suction change with loading for S5B . . . . . . . . . . . . . . . . . . . . . 76
6.1 Comparison of Hilf and Hasan & Fredlund’s method for pore pressure generations 88
6.2 Parameter values for the pore pressure generations for Hilf’s and Hasan & Fred-
lund’s methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.3 Parameters used for the compaction curve modelling using constant compressibil-
ity coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.4 Parameters used for the compaction curve modelling using variable compressibility
coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.5 Matric suction values used for the corresponding water contents . . . . . . . . . . 105
7.1 Model parameters for S2B (BBM) . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.2 Modified model parameters for S2B . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.3 Model parameters for S5B (BBM) . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.4 Modified model parameters for S5B . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.5 Model parameters for Boom clay (BBM) . . . . . . . . . . . . . . . . . . . . . . . 135
xv
xvi LIST OF TABLES
7.6 Modified model parameters for Boom clay . . . . . . . . . . . . . . . . . . . . . . 135
7.7 Model parameters for Speswhite kaolin (BBM) . . . . . . . . . . . . . . . . . . . 138
7.8 Modified model parameters for Speswhite kaolin . . . . . . . . . . . . . . . . . . 138
7.9 Model parameters for S2B (SFG model) . . . . . . . . . . . . . . . . . . . . . . . 147
7.10 Model parameters for S5B (SFG model) . . . . . . . . . . . . . . . . . . . . . . . 149
7.11 Model parameters for Boom clay (SFG model) . . . . . . . . . . . . . . . . . . . 155
7.12 Model parameters for Speswhite kaolin (SFG model) . . . . . . . . . . . . . . . . 158
List of Figures
3.1 Particle size distribution of sand . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Particle size distributions of S2B and S5B . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Standard and modified compaction curves of S2B . . . . . . . . . . . . . . . . . . 30
3.4 Standard and modified compaction curves of S5B . . . . . . . . . . . . . . . . . . 31
3.5 Schematic of the experimental setup for axis translation technique . . . . . . . . 33
3.6 Experimental setup of axis translation technique . . . . . . . . . . . . . . . . . . 34
3.7 Water volume discharge with time under 20 kPa applied air pressure . . . . . . . 35
3.8 Evaporation of water from the outflow container . . . . . . . . . . . . . . . . . . 36
3.9 Tensiometer used for matric suction measurements . . . . . . . . . . . . . . . . . 37
3.10 Matric suction measurements of S2B with tensiometer . . . . . . . . . . . . . . . 39
3.11 Matric suction measurements of S5B with tensiometer . . . . . . . . . . . . . . . 39
3.12 SWRC of S2B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.13 SWRC of S5B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.14 Load Track III used for oedometer tests . . . . . . . . . . . . . . . . . . . . . . . 41
3.15 Slope of the normal compression line for saturated S2B . . . . . . . . . . . . . . 42
3.16 Slope of the normal compression line for saturated S5B . . . . . . . . . . . . . . 43
4.1 Schematic of the compaction device setup . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Compaction device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3 Base of the pressure cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.4 Pressure transducer mounted on a steel block . . . . . . . . . . . . . . . . . . . . 49
4.5 Water volume discharge with time under 20 kPa, 50 kPa and 100 kPa applied air
pressures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.6 Average permeability of 1 bar ceramic disk . . . . . . . . . . . . . . . . . . . . . 51
xvii
xviii LIST OF FIGURES
4.7 Calibration curve of the pressure transducer . . . . . . . . . . . . . . . . . . . . . 53
4.8 Calibration curve of the LVDT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.9 GCTS pressure booster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.10 Experimental procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.11 Sample preparation by tamping . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.12 Test stress path (matric suction change is exaggerated) . . . . . . . . . . . . . . . 59
5.1 Initial matric suction measurements for S2B . . . . . . . . . . . . . . . . . . . . . 67
5.2 Initial matric suction measurements for S5B . . . . . . . . . . . . . . . . . . . . . 69
5.3 Comparison of initial matric suction measurements using different techniques for
S2B, [a]Variation of degree of saturation with matric suction, [b]Variation of water
content with matric suction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4 Comparison of initial matric suction measurements using different techniques for
S5B, [a]Variation of degree of saturation with matric suction, [b]Variation of water
content with matric suction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.5 Evolution of matric suction with time for S2B . . . . . . . . . . . . . . . . . . . . 73
5.6 Evolution of matric suction with time for S5B . . . . . . . . . . . . . . . . . . . . 74
5.7 Long equilibration time for S5B with 2.9% water content . . . . . . . . . . . . . 76
5.8 Variation of specific volume during the tests for S2B . . . . . . . . . . . . . . . . 78
5.9 Variation of specific volume during the tests for S5B . . . . . . . . . . . . . . . . 79
5.10 Variation of dry density during the tests for S2B . . . . . . . . . . . . . . . . . . 80
5.11 Variation of dry density during the tests for S5B . . . . . . . . . . . . . . . . . . 81
5.12 Variation of degree of saturation during the tests for S2B . . . . . . . . . . . . . 82
5.13 Variation of degree of saturation during the tests for S5B . . . . . . . . . . . . . 82
5.14 Variation of degree of the coefficient of compressibility due to net stress (ms1) for
S2B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.15 Variation of degree of the coefficient of compressibility due to net stress (ms1) for
S5B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.16 Variation of the coefficient of compressibility due to net stress (ms1) with degree
of saturation for S2B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.17 Variation of the coefficient of compressibility due to net stress (ms1) with degree
of saturation for S5B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.18 Compaction curves of S2B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
LIST OF FIGURES xix
5.19 Compaction curves of S5B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.1 Pore air and pore water pressure generation using Hilf’s approach . . . . . . . . . 90
6.2 Pore air and pore water pressure generation using Hasan & Fredlund’s approach 91
6.3 Comparison of Hasan & Fredlund’s and Hilf’s approach for pore air pressure
generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.4 Approach followed for the compaction curve generation . . . . . . . . . . . . . . 98
6.5 Modelling of the compaction curves using constant compressibility coefficients for
undrained loading [a] Hilf’s method [b] Hasan & Fredlund’s method [c] Compari-
son of Hilf’s and Hasan & Fredlund’s methods at σy = 700 kPa . . . . . . . . . . 99
6.6 Modelling of the compaction curves using constant compressibility coefficients for
semi-drained loading[a] Hilf’s method [b] Hasan & Fredlund’s method [c] Com-
parison of Hilf’s and Hasan & Fredlund’s methods at σy = 700 kPa . . . . . . . . 101
6.7 Modelling of the compaction curves using constant compressibility coefficients for
drained loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.8 Comparison of the modelled compaction curves using constant compressibility
coefficients for undrained, semi-drained and drained loading conditions (σy = 700
kPa lower curves and σy = 2000 kPa upper curves) . . . . . . . . . . . . . . . . . 103
6.9 Modelling of the compaction curves using variable compressibility coefficient for
undrained loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.10 Variation of the coefficient of compressibility (ms1) due to net stress during undrained
loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.11 Variation of the coefficient of compressibility (ms1) and the net stress with water
content during undrained loading . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.12 Modelling of the compaction curves using variable compressibility coefficient for
semi-drained loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.13 Variation of the coefficient of compressibility (ms1) due to net stress during semi-
drained loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.14 Variation of the coefficient of compressibility (ms1) and the net stress with water
content during semi-drained loading . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.15 Modelling of the compaction curves using variable compressibility coefficient for
drained loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
xx LIST OF FIGURES
6.16 Variation of the coefficient of compressibility (ms1) due to net stress during drained
loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.17 Variation of the coefficient of compressibility (ms1) and the net stress with water
content during drained loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.18 Comparison of the modelled compaction curves using variable compressibility co-
efficients for undrained, semi-drained and drained loading (σy = 700 kPa lower
curves and σy = 2000 kPa upper curves) . . . . . . . . . . . . . . . . . . . . . . . 113
7.1 Normal compression line of Boom clay . . . . . . . . . . . . . . . . . . . . . . . . 120
7.2 Normal compression lines of Boom clay under constant suctions . . . . . . . . . . 121
7.3 Normal compression line of Speswhite kaolin . . . . . . . . . . . . . . . . . . . . . 122
7.4 Normal compression lines of Speswhite kaolin for constant suctions . . . . . . . . 122
7.5 λ(s) of S2B for varying suctions (BBM) . . . . . . . . . . . . . . . . . . . . . . . 124
7.6 N(s) for S2B for varying matric suctions (BBM) . . . . . . . . . . . . . . . . . . 126
7.7 BBM predictions of the variation of dry density with applied vertical stress for
S2B [a] Comparison of the undrained (U) and semi-drained (S) predictions [b]
Comparison of the experimental data (E) and model (M) predictions (semi-drained)127
7.8 Compaction curves for S2B predicted by BBM [a] Comparison of undrained (U)
and semi-drained (S) predictions [b] Comparison of the experimental data (E) and
model (M) predictions (semi-drained) . . . . . . . . . . . . . . . . . . . . . . . . 128
7.9 λ(s) of S5B for varying matric suctions (BBM) . . . . . . . . . . . . . . . . . . . 129
7.10 N(s) for S5B for varying matric suctions (BBM) . . . . . . . . . . . . . . . . . . 130
7.11 BBM predictions of the variation of dry density with applied vertical stress for
S5B [a] Comparison of the undrained (U) and semi-drained (S) predictions [b]
Comparison of the experimental data (E) and model (M) predictions (semi-drained)132
7.12 Compaction curves for S5B predicted by BBM [a] Comparison of undrained (U)
and semi-drained (S) predictions [b] Comparison of the experimental data (E) and
model (M) predictions (semi-drained) . . . . . . . . . . . . . . . . . . . . . . . . 133
7.13 λ(s) of Boom clay for varying suctions (BBM) . . . . . . . . . . . . . . . . . . . . 134
7.14 N(s) for Boom clay for varying suctions (BBM) . . . . . . . . . . . . . . . . . . . 136
LIST OF FIGURES xxi
7.15 BBM predictions of the variation of dry density with applied vertical stress for
Boom clay [a] Comparison of the undrained (U) and semi-drained (S) predictions
[b] Comparison of the experimental data (E) and model (M) predictions (semi-
drained) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.16 Compaction curves for Boom clay predicted by BBM [a] Comparison of undrained
(U) and semi-drained (S) predictions [b] Comparison of the experimental data (E)
and model (M) predictions (semi-drained) . . . . . . . . . . . . . . . . . . . . . . 139
7.17 λ(s) of Speswhite kaolin for varying suctions (BBM) . . . . . . . . . . . . . . . . 140
7.18 N(s) for Speswhite kaolin for varying suctions (BBM) . . . . . . . . . . . . . . . 141
7.19 BBM predictions of the variation of void ratio with applied vertical stress for
Speswhite kaolin [a] Comparison of the undrained (U) and semi-drained (S) pre-
dictions [b] Comparison of the experimental data (E) and model (M) predictions
(semi-drained) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.20 Compaction curves for Speswhite kaolin predicted by BBM [a] Comparison of
undrained (U) and semi-drained (S) predictions [b] Comparison of the experi-
mental data (E) and model (M) predictions (semi-drained) . . . . . . . . . . . . 143
7.21 N(s) for S2B for varying matric suctions (SFG model) . . . . . . . . . . . . . . . 148
7.22 SFG model predictions of the variation of dry density with applied vertical stress
for S2B [a] Comparison of the undrained (U) and semi-drained (S) predictions [b]
Comparison of the experimental data (E) and model (M) predictions (semi-drained)150
7.23 Compaction curves for S2B predicted by SFG model [a] Comparison of undrained
(U) and semi-drained (S) predictions [b] Comparison of the experimental data (E)
and model (M) predictions (semi-drained) . . . . . . . . . . . . . . . . . . . . . . 151
7.24 N(s) for S5B for varying matric suctions (SFG model) . . . . . . . . . . . . . . . 152
7.25 SFG model predictions of the variation of dry density with applied vertical stress
for S5B [a] Comparison of the undrained (U) and semi-drained (S) predictions [b]
Comparison of the experimental data (E) and model (M) predictions (semi-drained)153
7.26 Compaction curves for S5B predicted by SFG model [a] Comparison of undrained
(U) and semi-drained (S) predictions [b] Comparison of the experimental data (E)
and model (M) predictions (semi-drained) . . . . . . . . . . . . . . . . . . . . . . 154
7.27 N(s) for Boom clay for varying suctions (SFG model) . . . . . . . . . . . . . . . 155
xxii LIST OF FIGURES
7.28 SFG model predictions of the variation of dry density with applied vertical stress
for Boom clay [a] Comparison of the undrained (U) and semi-drained (S) pre-
dictions [b] Comparison of the experimental data (E) and model (M) predictions
(semi-drained) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.29 Compaction curves for Boom clay predicted by SFG model [a] Comparison of
undrained (U) and semi-drained (S) predictions [b] Comparison of the experi-
mental data (E) and model (M) predictions (semi-drained) . . . . . . . . . . . . 159
7.30 N(s) for Speswhite kaolin for varying suctions (SFG model) . . . . . . . . . . . . 160
7.31 SFG model predictions of the variation of dry density with applied vertical stress
for Speswhite kaolin [a] Comparison of the undrained (U) and semi-drained (S)
predictions [b] Comparison of the experimental data (E) and model (M) predic-
tions (semi-drained) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
7.32 Compaction curves for Speswhite predicted by SFG model [a] Comparison of
undrained (U) and semi-drained (S) predictions [b] Comparison of the experi-
mental data (E) and model (M) predictions (semi-drained) . . . . . . . . . . . . 162
7.33 Stress paths for attaining samples with similar suctions . . . . . . . . . . . . . . 167
7.34 Volume change behaviour of samples with similar suction but developed under
different stress paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.35 Variation of the soil compressibility with suction . . . . . . . . . . . . . . . . . . 168
Chapter 1
Introduction
1.1 Problem statement
Soil compaction is widely used in geotechnical engineering and is important for the construction
of roads, dams, landfills, airfields, foundations and for engineered barriers which include landfill
liners and barriers for underground nuclear waste disposal. The main purpose of compaction is to
maximise the dry density of soils and therefore, to achieve favorable “strength and deformation”
behaviour for geotechnical applications.
Since Proctors’ pioneering work in 1933, researchers have tried to explain the governing
mechanisms in the densification stages of soils mainly in terms of capillarity and lubrication,
viscous water and physico-chemical interactions in soil. It was not until 1956 that the compaction
process was explained by Hilf using unsaturated soil mechanics principles. However, all of these
studies were predominantly qualitative due to the limited knowledge in the unsaturated soils
mechanics at the time. In addition, there is a vast body of literature covering the behaviour
of compacted soils. However, it is the the compaction process that brings the soil to its final
compacted stage. In this respect, examination of the compaction process is crucial to extend
the current understanding of the behaviour of compacted soils and to provide a link between
compaction process and the behaviour of compacted soils.
Since the late 1950s, the mechanical behaviour of unsaturated soils has been an extensive
research topic and significant advances have been made both in experimentation techniques
and constitutive modelling of unsaturated soils. The experimental techniques have reached a
stage where the compaction process can be carefully monitored. Furthermore, in the last few
1
2 CHAPTER 1. INTRODUCTION
decades, there has been an attempt to develop constitutive models for the mechanical behaviour
of unsaturated soils. Alonso et al. (1990) proposed the first elasto-plastic constitutive model to
couple strength and volumetric behaviour of unsaturated soils. Currently, there are numerous
constitutive models available in literature (Wheeler and Sivakumar, 1995; Bolzon et al., 1996;
Gallipoli et al., 2003a; Sheng et al., 2008) which can be used to model the compaction process.
Even though the constitutive models advanced significantly, there still remains a controversy
regarding the approaches used for models’ development. Thus, modelling the compaction process
will not only provide better understanding on the behaviour of the compacted soils but will also
give insights on the applicability of the current constitutive models.
1.2 Research aim and scope
The aim of this research project is to extend the current understanding of the compaction be-
haviour of soil during the compaction process in the light of unsaturated soil mechanics principles
and constitutive models. The scope of the thesis is limited to model and investigate the com-
paction process using the constitutive models that were developed with separate stress state
variables and also with combined stress variables. Constitutive models based on a single effec-
tive stress approach are excluded from the study due to the fact that the former ones are more
widely accepted in the literature.
The overall aim can be broken into the following main objectives:
1. Examining the behaviour of soil during the static compaction process and obtaining ex-
perimental data in order to use in the evaluation of the compaction process modelling;
2. Modelling the compaction process using the existing constitutive models to extend the
current understanding of the behaviour of the compacted soils; and,
3. Evaluating the performance of the compaction process models developed with different
approaches to gain insights on the applicability of the current constitutive models.
1.3 Overview of the thesis
Chapter 2 summarises the main features of the compacted soils. Qualitative studies on the
shape of the compaction curve are reviewed. The chapter provides a detailed description of the
constitutive models employed in the modelling part of this study. Since compaction is regarded
3
as a fabrication process of soils, the experimental studies on the fabric of compacted soils are
also detailed. Key challenges in modelling the compaction process are identified.
Soils that are investigated in the research project and their selection criteria are presented
in Chapter 3. The physical and geotechnical properties of the soils and their determination
techniques are described. These material properties will be used in Chapters 5 and 7 for the
analysis of the experimental results and evaluating the performance of the compaction models
developed with different approaches.
The testing equipments used and the experimental procedure followed to achieve the ob-
jectives of the targeted experiments are presented in Chapter 4. These are important for the
interpretation of the experimental results. Test difficulties and problems encountered during the
experiments are also presented in the chapter.
Chapter 5 presents the experimental results of the constant water content static compaction
tests. The experimental behaviour of soils during the compaction process is examined. The
influence of the initial matric suction on the volume change behaviour and the variation of
the matric suction during the compaction are monitored. The challenges of conducting such
experiments are discussed. The results obtained in this chapter are used for the evaluation of
the compaction model in Chapter 7.
The formulation of modelling the compaction process is presented in Chapter 6. One dimen-
sional static compaction process is modelled for undrained, semi drained and drained loading
conditions on the basis of unsaturated soil mechanics fundamentals and constitutive models. The
main parameters governing the shape of the compaction curve are identified. The formulations
are used in Chapter 7 to obtain the modelling predictions for the evaluation of the compaction
model.
In Chapter 7 the performance of the models developed with two different approaches are
examined on their ability to predict the compaction behaviour of soils. The models are evaluated
on four different soil types which include granular and fine soils. Insights are attained on the
applicability of the current constitutive models.
Chapter 8 presents the main conclusions derived from the work presented in this thesis.
Finally, the recommendations for future works are discussed.
Chapter 2
Literature review
2.1 Introduction
This chapter provides a brief review on experimental methods and theories required for the
examination of the compaction process in soils. Particular emphasis is given to the suction
measurement techniques and volume change behaviour of unsaturated soils because they are
required for monitoring and modelling the compaction process.
In Section 2.2, the theories proposed for the explanation of the inverted parabolic shape of
the compaction curves are summarised. The experimental methods for suction measurements
during the are summarised in Section 2.3. Finally, the volume change behaviour of unsaturated
soils is presented in Section 2.4.
2.2 Qualitative studies for compaction theory
Since Proctor’s pioneering work in 1933, many researchers have attempted to explain the leading
mechanisms in the densification stages, mainly on the dry side of the optimum water content
(OWC). The compaction curve was explained in terms of capillarity and lubrication (Proctor,
1933), viscous water (Hogentogler, 1936), pore pressure theory in unsaturated soils (Hilf, 1956),
physico-chemical interactions (Lambe, 1958), and the effective stress theory (Olson, 1963). In
addition, Barden and Sides (1970) undertook an experimental study on the relation between
the engineering performance of compacted unsaturated clay and microscopic observations of the
clay structure. These studies provide predominantly qualitative explanations of the shape of
5
6 CHAPTER 2. LITERATURE REVIEW
the compaction curve. The current section focuses on these theories and summarises them in a
chronological order.
2.2.1 Proctor’s theory (1933)
Proctor assumes that water has a dual effect on the soil: capillarity and lubrication. It is
contended that a thin film of water covers the soil particles due to surface tension. When the
films come together, they join the soil particles close to each other and cause high frictional
resistance between soil particles. On the other hand, he stated that sufficient amount of water
lubricates soil particles and reduces inter-particle forces. Therefore, in relatively dry soils the
dry density is low due to capillarity effect, whereas adding more water reduces shearing strength
and increases dry density due to lubrication.
Most of Proctor’s research on the compaction curve was based on lubrication and he placed
more emphasis on the concept of lubrication than on the concept of capillarity. However, obser-
vations by Horn (1960) showed that water acts as a lubricant for some but not for all soils. It
is found that water reduces the friction by approximately two times for sheet silicates, whereas
the friction coefficient for the submerged three-dimensional network silicates (i.e., quartz and
feldspar) is five times greater than the oven dry value of three-dimensional network silicates
(Horn, 1960). Therefore, if the lubrication theory is valid, sand should have its maximum dry
density in the dry stage. However, sand has a similar bell shaped compaction curve as other
soils.
2.2.2 Hogentogler’s theory (1936)
Hogentogler (1936) advanced the theory of viscous water to the soil compaction and proposed
that the compaction of soil goes through four stages of wetting. These stages are hydration,
lubrication, swelling and saturation.
The hydration stage involves adsorption of water within the soil particles and formation of
thin films on the particles surfaces. The first layer of the absorbed water is highly cohesive
and the high viscosity of water gives the soil a high shearing strength and results in a low dry
density. Adding more water to the soil results in absorbed water grading into free water, which
is the lubrication stage. When the water layers become thicker, viscosity of the water decreases.
Therefore, the shearing strength of the soil decreases too. As a result of decreasing shearing
strength in the lubrication stage, the dry density of the soil increases. The maximum amount of
7
lubrication occurs in the OWC. Hogentogler asserted that adding more water beyond the OWC
would result in displacement of soil particles and thus, the decrease of the dry density. Through
this stage, the soil swells without any change in air volume. During the saturation stage, he
stated that all the air is displaced and the saturation line joins the zero air voids line.
At present, it is known that the compaction of soils does not result in complete saturation,
and therefore the compaction curve cannot intersect with the zero air voids line. Moreover,
research carried out by Foster et al. (1955), Rowland et al. (1956), Walker (1956) and Martin
(1959) revealed that the thickness of the absorbed water on the surface of most clays is not more
than few molecular layers.
2.2.3 Hilf’s theory (1956)
Hilf (1956) introduced the first modern type of compaction theory, which is based on the theory
of pore-water and pore-air pressure in unsaturated soils. Instead of using the common water
content-dry density curve, he used a plot based on the void ratio-water void ratio. The minimum
void ratio corresponds to the maximum dry density in the Hilf’s curve, and the shape of the
compaction curve is similar to the usual Proctor’s curve.
Hilf explained the shape of the compaction curve by the concepts of pore air pressure and
capillarity. The capillary bridges in dry soils resist the compaction pressure and therefore, the
dry density is low. However, when the water content increases the menisci flattens and cannot
resist the compaction pressure, hence the dry density increases. When the soil reaches its OWC,
the air is trapped and cannot escape. The trapped air builds up high air pressure, and therefore
reduces the effectiveness of the compaction.
The research conducted by Gilbert (1959) and Langfelder et al. (1968) support the Hilf’s pore
pressure explanation on the shape of the compaction curve that the air permeability is zero close
to the OWC. Hilf believed that the negative pressures in the moisture films are interconnected
for the water content used in the field compaction. In addition, he assumed that these pressures
result in an all-round effective compressive stress on the soil skeleton which is equal in magnitude
to the negative pressure. However, subsequent investigations conducted by Bishop (1959) and
Bishop et al. (1960) designate that the capillary pressure act as an effective stress by a factor of
χ. The value of χ is difficult to determine and direct measurements cannot be done. Therefore,
this subject is still divisive (Hilf, 1975).
8 CHAPTER 2. LITERATURE REVIEW
2.2.4 Lambe’s theory (1959)
Lambe (1958) attempted to explain the shape of the compaction curve with the help of surface
chemical theories. The electrolyte concentration in the pore water is high at low water content.
Therefore, the inter-particle repulsion between the soil particles is reduced. The reduced inter-
particle forces allow flocculation to occur at low water contents and result in low dry-densities.
On the other hand, adding more water to the soil reduces pore water electrolyte concentration
and allows double layers to develop; hence, the soil results in a more disperse structure. Higher
dry densities can be obtained as a result of more dispersed structure.
Olson (1963) stated that Lambe’s explanation of the shape of the compaction curve is based
on the double layer theory and this theory could be used to predict the shape of the swelling
curve for sodium illite. However, he stated that it does not apply to calcium illite which has a
compaction curve of the same shape of other soils. Therefore, Olson (1963) concluded that this
theory does not have general applicability.
2.2.5 Olson’s theory (1963)
Olson (1963) explained the shape of the compaction curve using the effective stress concept. For
compaction on the dry side of the OWC, the increase in the water content increases the degree of
saturation and results in higher pore pressure. Therefore, the effective stress reduces for higher
water contents. As a result of reduced effective stress, the shearing stress of the soil decreases
and allows the soil to densify more than for low water content soil.
He also stated that the lubrication theory of Proctor (1933) and the viscous water theory of
Hogentogler (1936) appear to apply to the low water content-moisture density curve. At very
low water content, the electrical forces which attract the first layer of the water to the mineral
surfaces are stronger than the forces between water molecules. Therefore, the menisci are not
formed at this stage and water act as a lubricant between soil particles. Menisci start to form
by adding more water and this is the peak of the low water content-moisture density curve. The
compacted dry density starts to reduce again by adding more water due to the development of
menisci. Olson (1963) concluded that the low water content-moisture density curve can be seen
only in soils containing high percentage of plate shaped particles.
In his analysis, Olson (1963) assumed that the shearing strength of the inter-particle contact
surfaces is a purely frictional phenomenon and he did not include the cohesion component.
However, for clays of very high plasticity the strength of the inter-particle contacts may be
9
purely cohesional. Therefore, the theory may not be fully applicable for high plasticity soils.
2.2.6 Barden and Sides’s theory (1970)
Barden and Sides (1970) undertook an experimental investigation of the engineering behaviour
of compacted clay and intended to relate this to the microscopic observations of unsaturated clay
structure. Their results indicate that at low water contents, the low dry density is caused by the
presence of large air filled macro pores which have high strength and can resist to compaction
pressure without much distortion. When the water content of the soil is increased, the air filled
macro pores become wetter and weaker. Hence, during compaction it is easier to distort them
and the dry density of the soil increases. If more water above the OWC is added to the soil, the
water layers between the soil particles become thicker and the dry density of the soil decreases.
2.2.7 Summary of the compaction theories
In summary, Proctor’s capillarity and lubrication theory, Hogentogler’s viscous water theory,
and Lambe’s physico-chemical theory were reasonably consistent with the existing knowledge
at the time the theories were developed. However, the advances of knowledge have applied
limitations to these theories (Olson, 1963). Moreover, none of these theories have been subjected
to diagnostic laboratory experiments. As a result, these theories are rather speculative. The
effective stress explanation of the shape of the compaction curve which was carried out by Hilf
(1956), Olson (1963), and Barden and Sides (1970) appears more reasonable than the explanation
of the lubrication, the viscous water and the physico-chemical theories (Hilf, 1975).
In conclusion, these works provide predominantly qualitative explanations of the shape of
the compaction curve. Further research based on new measurement techniques are needed to
explain the shape of the compaction curve and the behaviour of compacted soils.
2.3 Suction measurement techniques
Suction is an important state variable to be monitored in order to better understand the funda-
mentals of the compaction process. There are numerous methods for suction measurements and
they can be divided in to two major groups: direct and indirect suction measurement techniques.
These methods are briefly summarised and evaluated in this section to find the most appropriate
technique to be used in the current study.
10 CHAPTER 2. LITERATURE REVIEW
2.3.1 Direct measurement techniques
In direct methods, the pore pressure or the water potential of soil is directly measured. This
measurement techniques include the null type axis translation technique and tensiometers.
Axis translation technique
The axis translation technique is the most widely used method for controlling the matric suctions
(Delage et al., 2008), defined as the difference between the pore air pressure (ua) and the pore
water pressure (uw), in soil. It can also be used for matric suction measurements, where it is
referred as the null type axis translation technique.
The working principle of this method is based on the capillary phenomenon. When a capillary
tube is placed in water under atmospheric conditions, the water will rise within the tube due to
the water pressure being smaller in reference to the atmospheric pressure. The angle between
the water and the air interface will determine the magnitude of the matric suction. In a closed
system, if this capillary tube is exposed to elevated air pressures, the increase in the air pressure
will be equally transferred to the water, assuming that the water is incompressible. Thus, the
angle between the air and the water interface would not change, but the water pressure within
the tube will now be positive.
The equipment for measuring the matric suction of soils using this method was first intro-
duced by Hilf (1956). A high air entry ceramic disk is used as an interface to separate the
water and air phases. A ceramic disk is a porous disk with fine pores. The capacity of the disk
determines the maximum air pressure that can be applied to the disk without allowing the free
air to pass through the fine pores. The ceramic disk is placed on a water compartment, which
is connected to a pressure transducer. As soon as an unsaturated soil sample is placed onto the
disk, it will start absorbing water from the ceramic disk and the tension in the water phase will
be directly transferred to the pressure transducer. However, if air pressures, which are higher
than the matric suction of the sample, are applied to the system this will be transferred to the
water phase and thus to the transducer. Now the transducer will be reading positive water
pressures. The matric suction of the sample will be equal to the difference between the applied
air pressure (ua) and the water pressure (uw) readings by the transducer. This hypothesis has
been validated by Hilf (1956) and Bishop and Donald (1961).
This method can be used to measure matric suctions up to 1500 kPa. This is due to the
maximum capacity of the ceramic disks that are currently being manufactured.
11
There are some limitations and difficulties associated with this technique. This limitations
are summarised below:
• The matric suctions measured by the axis translation technique are found to be slightly
lower when compared to the measurement with tensiometers (Chahal and Yong, 1965).
This is attributed to the formation of air nucleation within the water, which will result in
a smaller contact angle between the interface of the air and water phases, when elevated
air pressures are applied.
• At high saturations, when the air phase becomes occluded within the water phase, the
application of air pressure to the system, which is transferred to the water phase, will
compress the occluded air. Thus, the angle between the water and air interface will be
altered and the correct measurements of matric suction will not be possible. This method
was experimentally evaluated by Fredlund and Morgenstern (1977) for degrees of saturation
between 76% and 95%, and by Tarantino et al. (2000) for degrees of saturation between
56% and 77%. In addition, Bocking and Fredlund (1980) conducted a theoretical study
and concluded that the matric suctions are overestimated if the soil has a compressible
structure. It was suggested by Romero (2001) that the air pressure should be changed at
a very slow rate to maintain the continuous air channels.
• Some air can diffuse under the ceramic disk when tests take long time. This can alter the
measurements of the matric suction. In addition, the diffused air will accumulate under
the ceramic disk and may result in the loss of the continuity between the pore water of
the sample and the water in the compartment under the ceramic disk. The diffused air
needs to be regularly flushed to avoid this problem. Vanapalli et al. (2008) suggested a
flushing system to periodically remove the dissolved air accumulated under the ceramic
disk. However, it should be taken into account that, flushing the system too often will
extend the testing durations as it will be described in Chapter 5.
• Condensation of vapour in the pressure chamber walls due to the temperature variations
was reported by Oliveira and Marinho (2006), which will lead to the relative humidity
changes within the chamber. The changes in the relative humidity around the soil specimen
will result in the erroneous measurements at low suctions (Marinho et al., 2009). Moreover,
the compressed air is generally applied to the pressure chamber and a moisture loss through
compressed air lines was reported by Romero (2001). In addition, Romero (2001) have
12 CHAPTER 2. LITERATURE REVIEW
conducted a study on numerical calculations for the moisture loss through compressed air
lines. Numerical analysis was also carried out by Romero (1999) to simulate the evaporative
fluxes and the corresponding matric suction changes. It was concluded that no important
suction changes are expected with evaporative fluxes lower than 9.4x10−7 (mm3/s)/mm2.
• The test durations can be long due to the matric suction equilibration time once the air
pressure is applied. The time depends on the sample size and type due to the permeability
of the soil sample (Marinho et al., 2009). Kunze and Kirkham (1962) proposed an analytical
equation to estimate the equalisation time, which considers the impedance of the ceramic
disk and the permeability of the soil sample. The equilibration time for the initial matric
suction measurements using null method was reported to be less than a day for fine grained
and residual soils (Vanapalli et al., 1999; Fredlund and Vanapalli, 2002; Rahardjo and
Leong, 2006) and around three days for gneissic soils Oliveira and Marinho (2006).
• Recently, Baker and Frydman (2009) criticised the use of axis translation technique in
suction measurements. They have stated that, in field, high suctions cannot develop with-
out the cavitation of water and the axis translation technique alters the soil behaviour
by preventing the cavitation. Thus, it should not be used to examine the behaviour of
unsaturated soils.
In summary, there are some challenges in using the axis translation technique for suction
measurements. However, when the problems are carefully considered, the axis translation tech-
nique is a reliable and efficient method for soil matric suction measurements. It also provides
some advantages over other techniques which will be pointed out in the reminder of this section.
Tensiometers
Tensiometers are devices that directly measure the pore water pressure of soils. Tensiometers
consist of a water reservoir, a ceramic disk which acts as an interface between the air and water
phases, and a pressure gauge (e.g., strain gauge diaphragm, pressure transducer).
The capacity of the standard tensiometers are recorded to be around -80 kPa pressure. This
value was attributed to the inability of water to sustain negative pressures. This supposition was
incorrect as it was experimentally shown by numerous researchers (Henderson and Speedy, 1980;
Zheng et al., 1991). It was noticed that its not the water that cannot sustain tension beyond that
values but the occurrence of heterogenous cavitation within the water that determines this limit.
13
The cavitation occurs in the boundary between the water reservoir wall and the water, and/or
between the contact of the small particles within the water reservoir and the water. The reasons
of heterogenous cavitation is well explained in Marinho and Chandler (1995) and Marinho et al.
(2009). Ridley and Burland (1993) introduced the first tensiometer which can go beyond that
pressures and it is referred as a high capacity tensiometer. The main difference from standard
tensiometers is the use of a much smaller water reservoir in order to delay cavitation.
The high capacity tensiometers are designed either by using strain gauge diaphragm or fitting
a ceramic disk to a commercial pressure transducer. The ceramic disk is fitted to the transducer
by means of O-ring (Ridley and Burland, 1993; Guan and Fredlund, 1997; He et al., 2006),
araldite (Meilani et al., 2002; Take and Bolton, 2003; Lourenco et al., 2006) or copper gasket
(Toker et al., 2004), which are also a source of air presence within the tensiometer and therefore
a source of cavitation. It is shown that a significant improvement can be done when O-ring is re-
placed by araldite or copper gasket (Toker et al., 2004). In addition, tensiometers designed using
a strain gauge diaphragm (Ridley and Burland, 1995; Marinho and Chandler, 1995; Tarantino
and Mongiovı, 2002; Cui et al., 2008; Rojas et al., 2008b) are found to have higher capacity and
longer measurement duration, compared to the ones that use pressure transducers.
Another source of cavitation is the presence of air babbles within the pores of the ceramic
disk. Full saturation should be attained in order to delay cavitation. This is generally achieved
by initially saturating the transducer under vacuum and subsequently pre-pressurising it. Pre-
pressurisation is required in order to dissolve the pore air trapped in the ceramic disk or on the
walls of the water reservoir. The effects of saturating the tensiometer under vacuum is presented
elaborately by Take and Bolton (2003). The effect of pre-pressurisation on the cavitation and
the procedures for pre-pressurisation has been extensively studied in literature by Ridley and
Burland (1995); Guan and Fredlund (1997); Tarantino and Mongiovı (2002); Meilani et al. (2002);
Take and Bolton (2003); Lourenco et al. (2006); He et al. (2006); Mahler and Diene (2007); Cui
et al. (2008). In addition, cycles of cavitation, which is induced by placing the tensiometer
in contact with a dry sample, and pre-pressurisation was also found to improve the maximum
measurement durations of the test (Tarantino and Mongiovi, 2001; Tarantino and Mongiovı,
2002; Toker, 2002; Take and Bolton, 2003).
Time to reach equilibration during tensiometer measurements was reported to depend on
the good contact between the tensiometer and the soil sample (Boso et al., 2004; Oliveira and
Marinho, 2008). It was suggested by Oliveira and Marinho (2008) to use a soil paste, which has
14 CHAPTER 2. LITERATURE REVIEW
a water content between its plastic and liquid limits, to maintain a better contact between the
porous disk and the soil. Boso et al. (2004) also showed that the use of soil sample at its liquid
limit would extend the equilibration time. In addition, as it would be expected the equilibration
time will depend on the permeability of the soil. Moreover, for tests at constant water contents
the equilibration time was reported to be about couple of hours (Tarantino and Tombolato, 2005;
Delage et al., 2007). This is a great advantage compared to the equilibration time required in
the axis translation technique.
One drawback of tensiometers compared to the other methods is the maximum measurement
duration. Although, a great care is taken for the preparation of the tensiometers, the cavitation
will eventually occur. Maximum eight days of measurement is reported by Cunningham et al.
(2003) at 800 kPa matric suction. The tensiometer has to be replaced for tests that take long
durations. An important advantage of tensiometers over axis translation technique is that they
can be used for matric suction measurements in the field.
2.3.2 Indirect measurement techniques
Indirect measurements do not measure the soil suction directly, but parameters such as water
content or relative humidity are measured and, later related to suction through a calibration
with a known values of suctions. These methods include the use of psychrometers and filter
paper.
Psychrometers
There are three types of psychrometers based on different working principles. These types in-
clude thermocouple psychrometers, transistor psychrometers and chilled-mirror dew-point psy-
chrometers. These methods can be used to measure the total suction of the soil. Each of these
psychrometers will be explained in the current section.
Thermocouple psychrometers work based on the temperature difference between two refer-
ence points, which are the reference junction and the measurement junction. This method was
used by Spanner (1951); Richards and Ogata (1958); Meeuwig (1972); Campbell (1979); Brown
and Robert (1980); Brown and James (1976) for ranges between 0.3 MPa and 8 MPa. The
equilibration time depends on the volume and initial relative humidity of the chamber, and also
the soil suction of the sample. The time required for equilibrium varies from several hours to
several days (Brown and Robert, 1980; Campbell, 1979). In order to measure the suctions with
15
an accuracy of 10 kPa, the temperature variation should be kept within a range of 0.001 ◦C
(Edil and Motan, 1984).
Transistor psychrometers are developed as an alternative to thermocouple psychrometers
to extend the suction range. The working principle is similar to thermocouple psychrometers.
There are two main difference from the previous method; the application of water, which is
manually applied by dropping water, and the measurement of the current, which is conducted
by means of a transistor. The method was used by Dimos (1991); Woodburn et al. (1993) and
Truong and Holden (1995) with an upper limit to 15 MPa. The equilibration time in general is
about 1 hour. This method was compared with tensiometer measurements (osmotic effect was
subtracted) by Boso et al. (2003) and a good agreement was observed in the range of 1 MPa
and 3 MPa.
Chilled-mirror dew-point psychrometers are also developed as an alternate to the thermocou-
ple psychrometers. This method measures the temperature in which the condensation happens.
The soil sample is placed in a housing chamber which contains a mirror and a photoelectric de-
tector of condensation on the mirror. The mirror temperature is controlled by a thermoelectric
cooler. The relative humidity is estimated from the difference between the first condensation
temperature and the temperature of the soil sample. The method was successfully used by
Gee et al. (1992); Brye (2003); Leong et al. (2003); Tang and Cui (2005); Thakur and Singh
(2005) and Agus and Schanz (2005) up to maximum suctions of 60 MPa. The time required for
equilibration using this method can be as low as 5 minutes. In addition, Cardoso et al. (2007)
compared the measurements between this method and the transistor psychrometers. A good
agreement was observed in the range from 1 MPa to 7 MPa.
In summary, psyhcrometers are used to measure the total suctions at ranges from 0.3 MPa to
70 MPa. The method is not reliable in low suctions because a small change in relative humidity
results in a high suction change and thus, accuracy level is low.
Filter paper
The filter method can be used to measure both the total suction and matric suction of soils. It
is a relatively cheap method compared to the other methods described above.
It works with the principles of either vapour equilibration or water content equilibration.
When the filter paper comes to equilibrium through vapour, total suction is measured. On the
contrary, when it is placed in contact with the soil sample, it comes to equilibrium through
16 CHAPTER 2. LITERATURE REVIEW
water flow and matric suction is measured. In order to maintain the contact, filter paper can be
either placed on top (Bulut et al., 2001) or buried Houston et al. (1994) within the soil sample.
After equilibrium is reached, the water content of the filter paper is measured and it is related
to suction through calibration curves, which are developed through measurement techniques.
Whatman #42, Schleicher and Schuell #589 White Ribbon, and Fisher 9-790A are commonly
used filter papers commercially available.
The method requires approximately a week for equilibration. It is applicable to ranges from
1 MPa to 500 MPa. At lower suction ranges the relationship between total suctions and relative
humidity is very steep, thus the accuracy becomes very low (Sibley et al., 1990). In addition,
since slight temperature variations can cause significant changes in the relative humidity, the
method is also very sensitive to temperature variations.
2.3.3 Summary of the suction measurement techniques
In this section, direct and indirect suction measurement methods were briefly described. It was
shown that the indirect measurement techniques are better for the high suction ranges and not
reliable at suctions lower that 0.3 MPa (for the thermocouple psychrometer). In addition, the
filter paper method cannot be used for continuous suction measurements, which is the main aim
of the experimental part of this project. Considering these facts, the axis translation technique
and tensiometers are adopted to be used in the current study.
2.4 Volume change behaviour of unsaturated soils
The volume change behaviour of soil needs to be predicted in order to model the compaction
process. This section presents the basic concepts for the volume change calculations. Since the
compaction process is modelled for one dimensional compression, only the volume change calcu-
lations are presented and shear behaviour of soils is beyond the scope of this study. Initially, the
stress variables for unsaturated soils are summarised. As it is shown later in this section, the
pore pressure development during the compaction process is crucial for the modelling the com-
paction behaviour. A brief summary is provided on the pore pressure development approaches
available in literature. In addition, constitutive models are required for the calculations of the
compressibility coefficients. Thus, constitutive models developed for unsaturated soils are also
summarised in this section. The volume change formulations of the models, that are examined
17
in detail in the current study, are also described.
2.4.1 Stress variables
The principal of effective stress (Terzaghi, 1936) provided the key for analysing the mechanical
behaviour of unsaturated soils. For that reason, early researchers attempted to extend the
concept of the effective stress and relate it to the behaviour of unsaturated soils. The most
widely used relationship was proposed by Bishop (1959) in the following form:
σ′ = σ − ua + χ(ua − uw) (2.1)
where σ′ is the effective stress, σ is the total stress, χ is a factor varying with degree of saturation,
ua is the pore air pressure and uw is the pore water pressure.
The important assumption of defining a single effective stress is to relate all the measurable
changes in stress, compression and shearing to the changes in the effective stress. However,
early experimental studies failed to find a single χ parameter which could be used to predict
the volumetric and shearing strengths of soils (Coleman, 1962; Bishop and Blight, 1963; Blight,
1967). Jennings and Burland (1962) were the first to challenge the validity of Equation 2.1 and
whether it was correct to combine the total stress (σ), pore air pressure (ua) and pore water
pressure (uw) in a single equation. However, it became clear that the volumetric strain is not
controlled by a single parameter.
Due to the difficulties in determining the χ parameter, experimental results pushed towards
the adoption of two independent stress state variables. Matyas and Radhakrishna (1968) devel-
oped state surfaces for the void ratio (e) and degree of saturation (Sr), based on two separate
stress state approach. Fredlund and Morgenstern (1977) argued that the use of any two of the
three stress variables σ− ua, σ− uw and ua − uw would be sufficient to fully describe the stress
state of unsaturated soils. This approach is adopted widely in literature and has formed the
basis of the most of the constitutive models for unsaturated soils.
There are also models developed using the effective stress principal for unsaturated soils
in literature (Loret and Khalili, 2002; Laloui and Nuth, 2009). However, this approach is not
considered in this study due to the time restrains and the former approach being more widely
accepted. The topic is very controversial and requires meticulous separate examination; and
thus, is out of the scope of this study.
18 CHAPTER 2. LITERATURE REVIEW
2.4.2 Volume change
The description of volume change in terms of two independent stress variables was proposed by
Fredlund & Morgenstern (1976). For one dimensional loading, the equation is in the following
form:
εv =∆VvV0
= ms1∆(σy − ua) +ms
2∆(ua − uw) (2.2)
where εv is the volumetric strain, ∆Vv is the overall volume change, V0 is the initial volume,
∆(σy − ua) is the change in the vertical net stress, ∆(ua − uw) is the change in matric suction,
ms1 is the compressibility of a soil element with respect to vertical net stress (σy − ua) and ms
2
is the compressibility of a soil element due to matric suction (ua − uw).
2.4.3 Pore pressure development
One of the primary aspects associated to model the compaction process is the calculations
of the pore pressure development as it can be seen from Equation 2.2. There are two different
approaches available in the literature that can be adopted for pore pressure developments, which
are Hilf’s (1948) and Hasan & Fredlund’s (1980) approaches. These two methods are summarised
in this section.
Hilf’s approach (1948)
Hilf (1948) developed a relationship between pore pressure and applied stress, which is based on
one-dimensional soil compression, Boyle’s law and Henry’s law, and is expressed as follows:
∆ua =
1
1 +(1 − Sr0 + hSr0)n0
(ua0 + ∆ua)mv
∆σy (2.3)
where ∆ua is the change in the pore air pressure, Sr0 is the initial degree of saturation, h is
the solubility coefficient of air in water, n0 is the initial porosity, ua0 is the initial absolute pore
air pressure, mv is the coefficient of volume change in saturated soil, and ∆σy is the change in
applied vertical stress.
Hilf (1948) developed this equation assuming that both air and water phases are undrained,
and volume reduction is due to air dissolving in water and compression of free air. Both liquid
and solid phases were considered to be incompressible. In this approach, it is also assumed that
19
the changes in pore air pressure are equal to the changes in pore water pressure, which makes
the change in matric suction (∆(ua − uw)) insignificant.
Hasan & Fredlund’s approach (1980)
Hasan & Fredlund & (1980), on the other hand, developed an equation for pore pressure (ua
and uw) predictions during undrained loading using pore pressure parameters. They used pore
pressure parameters in the following form:
dua = Badσy (2.4)
duw = Bwdσy (2.5)
where Ba is the pore air pressure parameter and Bw is the pore water pressure parameter. The
pore pressure parameters are defined as:
Ba =R2R3 −R4
1 −R1R3(2.6)
Bw =R2 −R1R4
1 −R1R3. (2.7)
In Equations (2.6) and (2.7) parameters R1, R2, R3 and R4 are expressed as following:
R1 =(ms
2/ms1 − 1 − (1 − Sr + hSr)n/(uam
s1)
ms2/m
s1 + SrnCw/ms
1
(2.8)
R2 =1
ms2/m
s1 + SrnCw/ms
1
(2.9)
R3 =ma
2/ma1
ma2/m
a1 − 1 − (1 − Sr + hSr)n/(uams
1)(2.10)
R4 =1
ma2/m
a1 − 1 − (1 − Sr + hSr)n/(uams
1)(2.11)
where Sr is degree of saturation, n is the porosity, Cw is the compressibility of water, ma1 is
the compressibility of the air phase with respect to vertical net stress (σy − ua) and ma2 is the
compressibility of the air phase with respect to matric suction (ua − uw).
20 CHAPTER 2. LITERATURE REVIEW
In this method, it is assumed that the soil particles are incompressible, but the fluid phase is
compressible (air and water). Therefore, the volume reduction of a soil element is considered to
be due to the compression of free and dissolved air as well as water. In contrast to the approach
proposed by Hilf (1948), the matric suction changes were taken into account. It was accepted
that the air and water pressures become equal at saturation. Therefore, matric suction will
decrease from the initial value to zero at saturation.
2.4.4 Constitutive models for unsaturated soils
Constitutive models can be used to estimate the variations in the compressibility coefficients,
which are defined in Equation 2.2. In this study, the constitutive models developed with separate
stress and combined stress state variables approach are shortly summarised. The mathematical
formulations for the models that were analysed in detail are presented in this section. These
equations will be later used to model the compaction process.
Separate stress state variables approaches
Alonso et al. (1987) introduced an elasto-plastic critical state framework, which form the basis
for the first constitutive model for unsaturated soils. The mathematical formulation of this
model was later presented in Alonso et al. (1990) and is commonly know as the Barcelona Basic
Model (BBM). The model considers the net stress (σ− ua) and matric suction (s = ua− uw) as
independent stress state variables.
The volumetric strain (εv) for one dimensional loading can be written in the following form
when derived from the BBM:
εv =dν
ν=λ(s)
ν
d(σy − ua)
(σy − ua)+κsν
d(ua − uw)
(ua − uw) + uatm(2.12)
where ν is the specific volume of the soil element, λ(s) is the slope of the NCL at matric suctions
defined in semi logarithmic plot, d(σy − ua) is the change in vertical net stress, d(ua − uw) is
the change in matric suction and κs is the slope of the volume change vs. matric suction curve
defined in semi logarithmic plot.
In Equation 2.12, λ(s) is defined as:
λ(s) = λ(0)[(1 − r)e−βs + r] (2.13)
21
where λ(0) is the slope of the NCL for saturated soil, r is a constant parameter related to the
maximum stiffness of the soil, and β is a constant parameter which controls the rate of increase
of soil stiffness with matric suction.
Integrating Equation 2.12 for vertical net stress changes (σy − ua) under a constant matric
suction leads to:
ν = N(s) − λ(s)ln(σy − ua)
(σy − ua)c(2.14)
where N(s) is the specific volume at an initial stress state, (σy − ua)c) and (σy − ua)c is a
reference stress in which changes of matric suction causes only elastic deformation.
In the BBM, N(s) is defined as:
N(s) = N(0) − (λ(0) − λ(s))ln(σy − ua)c − κsln
(s+ uatmuatm
)(2.15)
where N(0) is the specific volume, at a reference stress state,(σy − ua)c, for saturated soil.
The BBM was later modified by Wheeler and Sivakumar (1995). They abandoned the use
of a reference stress (pc) at which the change of matric suction only produces elastic volume
deformations (a yield curve is a straight vertical line in matric suction and stress plot). Atmo-
spheric pressure was taken as a reference pressure in their volume change equation. In addition,
they used the empirical values for the slope of the NCLs (λ(s)) and the initial specific volumes
(N(s)) of unsaturated soil samples rather than introducing an equation for the variation of the
slope of the NCLs with the change in matric suction. Wheeler and Sivakumar (1995) noticed in
their experimental results that the change in the slope of the NCLs (λ(s)) and the initial specific
volume (N(s)) vary differently (the slope of the NCLs increases with increasing matric suction)
than it was proposed in the BBM; however, they did not developed a complete mathematical
model for this behaviour.
Vaunat et al. (2000) and Thu et al. (2007) slightly modified the BBM and incorporated
SWRC into the model. The same assumption for the slope of the NCLs (λ(s)) as proposed in
the BBM was used.
In addition, Chiu and Ng (2003) incorporated SWRC into the framework proposed by
Wheeler and Sivakumar (1995). For the variation of the slope of the NCLs (λ(s)), they used
a similar approach proposed in the BBM (see Equation 2.13). Different from the BBM, they
considered that the “r” parameter is bigger than one to fit their experimental data better. This
22 CHAPTER 2. LITERATURE REVIEW
modification was also proposed by Wheeler et al. (2002) in order to capture the increase in
compressibility with the increase in matric suction. This is a reasonable modification and well
captured the behaviour of tested soil in their studies because the soil compressibility increases
with the increase in matric suction.
In summary, the BBM is the first and most widely accepted elasto-plastic model for unsatu-
rated soils. The other models based on two stress state variables approach are mainly derived in
a similar manner or further improved by adding other important behaviour of unsaturated soils
which was not covered in the BBM (e.g., incorporation of SWRC). Thus, the BBM is chosen for
thorough examination in this study.
Combined stress state variables approaches
The combined stress variables approach models differ from the separate stress variables approach
models by combining the net stress and matric suction in their model formulations. These models
are briefly discussed in this section.
The combined stress state approach proposed by Bolzon et al. (1996), modified the BBM
by using average skeleton stress (σ − ua + Sr(ua − uw)) and matric suction (ua − uw) as stress
variables instead of net stress (σ − ua) and matric suction (ua − uw). This modification gives
advantages over the BBM when the transition zone between unsaturated and saturated soils is
considered. This model uses a similar function to the one proposed in the BBM for the variation
of soil compressibility with matric suction (λ(s)).
Santagiuliana and Schrefler (2006) included hydraulic hysteresis to the model proposed by
Bolzon et al. (1996). However, during the compaction process only wetting path is followed and
there is no need to consider the hysteretic nature of the soil. Thus, this modification would not
improve the compaction process predictions and will give similar results to the BBM.
Gallipoli et al. (2003a) proposed an elasto-plastic constitutive model incorporating the effect
of soil fabric into the model. They used average skeleton stress (σ − ua + Sr(ua − uw)) as
a stress variable and introduced a new stress variable as a second stress variable which is a
function of matric suction and degree of saturation (ξ = f(s)(1 − Sr)). In their model, they
proposed two soil parameters which can be found by calibrating the model with the experimental
results. Therefore, the initial specific volume (N(s)) and the slope of the NCLs (λ(s)) do not
necessarily reduce with increasing matric suction but depend on the experimental results and
are incorporated into the model with the proposed soil parameters.
23
Tarantino and De Col (2008) developed a compaction model where they used different work
conjugates than the ones used in the BBM as suggested by Houlsby (1997). The work conjugates
they adopted are the average skeleton stress (σ−ua+Sr(ua−uw)) and modified suction (s∗ = ns).
In addition, Tarantino and De Col (2008) considered the change in matric suction with loading
by incorporating the model proposed by Gallipoli et al. (2003b). Their model is similar to the
model developed by Gallipoli et al. (2003a) and only uses different stress state variables.
The model proposed by Sheng et al. (2008), referred as the SFG model in this study, is de-
veloped in a different way than the models presented above. They developed their mathematical
formulations, extending the volume change equation for saturated soil to unsaturated soil by
adding a suction parameter into it. The volumetric strain (εv) of the SFG model is suggested in
the following form:
εv =dν
ν= λvp
d(σy − ua)
(σy − ua) + s+ λvs
d(ua − uw)
(σy − ua) + s(2.16)
where ν is the specific volume of the soil element, λvp is the slope of the normal compression line
(NCL) for saturated soil when both axes are plot in logarithmic scale, d(σy − ua) is the change
in vertical net stress, d(ua − uw) is the change in matric suction, s is matric suction and λvs is
the slope of the volume change vs. matric suction curve when both axes are plot in logarithmic
scale.
Integrating Equation 2.16 for vertical net stress changes (σy − ua) under a constant matric
suction or for matric suction (s) changes under a constant vertical net stress(σy − ua) lead to:
lnν = lnN(s) − λvpln(σy − ua) + s
(σy − ua)0 + s(2.17)
where N(s) is the specific volume at an initial stress state ((σy − ua)0,s).
N(s), which is the specific volume at an initial stress state ((σy − ua)0,s) for matric suctions
smaller than the saturation suction (ssa), s < ssa is defined as
lnN(s) = lnN0 − λvpln(σy − ua)0 + s
(σy − ua)0(2.18)
and; when N(s) is under net vertical stress (σy −ua)0 equal to unity and matric suctions bigger
than the saturation suction (ssa), s≥ssa, it is defined as
24 CHAPTER 2. LITERATURE REVIEW
lnN(s) = lnN0 − λvpln(σy − ua)0 + ssa
(σy − ua)0− λvp
(1 − ssa + 1
s+ 1
)(2.19)
where N0 is the specific volume at initial stress state for saturated soil when both axes are plot
in logarithmic scale.
To conclude, the models developed with combined stress variables approach use the BBM
or models with similar approach (Wheeler and Sivakumar, 1995) in their mechanical modelling
formulations but use different stress state variables instead of net stress and matric suction
except for the SFG model. For that reason, the SFG model is also selected in this study for
detailed evaluation.
2.4.5 Concluding remarks
In the first section of this chapter, it is shown that the compaction process of soils is not well
examined particularly from a quantitative sense. Compaction process can be analysed by both
examining the variation of the stress variable during the compaction process and by modelling
the compaction process.
Experimental methods that are required to monitor suction during the compression process
were summarised in this section. The axis translation technique and tensiometers were found
to be applicable for the experimental part of this study. The psychrometers were disregarded
because they are more reliable at high suction ranges. In the current study, the compression
behaviour of sand-bentonite mixtures, which have relatively low suctions, was monitored. In
addition, filter paper cannot be used for continuous suction measurements, which is the main
aim of the experimental part in this study.
In addition, in order to model the compaction process, the volume change theories for un-
saturated soils are also presented. Initially, the stress variables for unsaturated soils are briefly
summarised. Subsequently, the approaches available in literature for pore pressure development
are explained. Finally, the unsaturated constitutive models that can be used for the calculations
of the compressibility coefficients are presented. Two models, the BBM developed by two sepa-
rate stress variables approach and the SFG model based on combined stress variables approach
were selected to be incorporated in the compaction model.
Chapter 3
Selected materials for the
research
3.1 Introduction
The aim of this research project is to fundamentally understand the governing parameters of the
shape of the compaction curve and the volume change characteristics of soils. For that reason,
materials that have complex chemical properties (e.g., clays) and are hard to work with during
experimental procedure were avoided. Initially sand was chosen as a primary material to be
investigated. However, it was difficult to monitor the low suction of the sand with the available
equipment. The suction within the material was increased by introducing a small percentage of
fines to the sand. This resulted in an improved investigation into the effect of matric suction
on the overall shape of the compaction curve. Bentonite was selected as the fine material to
be mixed with sand, considering that the results might be useful for the researchers who study
engineered barriers and clay liners.
Sand bentonite mixtures with 2% and 5% bentonite content by weight were considered for
investigation. The sand with 2% bentonite content had the lowest measurable suction with the
available equipment, therefore it was selected as a material to be studied. In addition, bentonite
in the amount of 5% by weight was mixed with poorly graded sand to produce a material with
higher suctions and was determined as the second material to be tested. Moreover, considering
the research conducted by Mollins et al. (1996) and Montanez (2002) where they found that
sand particles are in contact even for high water contents for similar sand bentonite mixtures;
25
26 CHAPTER 3. SELECTED MATERIALS
the amount of bentonite mixed with sand (2% and 5%) is unlikely to produce any swelling in
the selected materials. Therefore, conditions for experiments are favourable.
This chapter describes the characteristics of the materials investigated in this study cover-
ing physical and geotechnical properties and techniques of their determination. These material
properties will be used in other chapters for analysis and modelling. Firstly, the physical prop-
erties of the selected soils are summarised and this is followed by the compaction characteristics
of the materials. Finally, the soil water characteristic curves and the consolidation curves of the
materials and their determination methods are described.
3.2 Physical properties
This research project has been carried out on two types of soils. Sand and bentonite were mixed
in the amount of 2% and 5% bentonite content by weight. These soils hereafter will be referred
to as S2B for the 2% sand-bentonite mixture and S5B for the 5% sand-bentonite mixture.
The sand selected for the study was the finest sand commercially available during this research
work and was provided by Unimin Australia. It contains 99% of quartz minerals and 1% of
respirable free crystalline silica in the form of quartz.
Bentonite, named as ActiveGel 150, was obtained from the same supplier. It is specified
by the supplier (Unimin Australia, 2002) to contain more than 87% of smectite, less than 2%
of albite and less than 11% of crystalline silica. As characterised by the supplier, its swelling
volume is 35 ml/2g, cation exchange capacity is 95 meq/100g and bulk density 1 Mg/m3. The
water content as received from the supplier was 10% (Shannon, 2008). All the physical properties
of bentonite are adopted from Shannon (2008) who worked on the water flow in unsaturated
reactive soils with the same soil.
The index properties of the individual soils (sand, bentonite) and the mixtures (S2B, S5B)
are summarised in the following part of this section. These include the particle size distribution,
Atterberg limits, linear shrinkage and the specific gravity of the materials where applicable.
3.2.1 Particle size distribution
Two methods were employed to determine the particle size distribution of the materials. Siev-
ing method was used for the determination of the particle size distribution of sand and laser
diffractometer was used for bentonite.
27
10−2
10−1
100
0
10
20
30
40
50
60
70
80
90
100
Diameter (mm)
Per
cent
age
finer
by
wei
ght (
%)
Cu = 1.71Cc = 1.07
Figure 3.1: Particle size distribution of sand
The particle size distribution of the sand was determined using standard method of analysis
by sieving according to the Australian Standards (AS1289.3.3.1-1995) . The particle size dis-
tribution of sand is shown in Figure 3.1. The sand has 100% particles finer than 0.6 mm and
20% particles of finer than 0.2 mm. The uniformity coefficient (Cu = D60/D10) of the sand is
determined to be 1.71 and the coefficient of curvature (Cc = (D30)2/(D10D60)) is 1.07. The
sand used in the investigation is classified as poorly graded sand (SP) according to the Unified
Soil Classification System (USCS).
The particle size distributions of sand bentonite mixture with 2% and 5% content of bentonite
by weight are shown in Figure 3.2. Although the particle size distributions of the two soils are
quite similar, the small difference in the bentonite content changes the suction of the material
enough to serve the experimental purpose (see Chapter 5 and 7).
3.2.2 Atterberg limits
The Atterberg limits of the bentonite ActiveGel 150 were determined according to the Australian
Standards (AS 1289.3.1.1 1995 and AS 1289.3.2.1 1995) . The Casagrande method was performed
for the liquid limit determination. Plastic limit of the soil was found by the standard thread-
rolling method. Table 3.1 shows the Atterberg limits of this soil and the results are in agreement
with the values specified by the soil provider (Unimin, 2002). ActiveGel 150 bentonite has 80%
28 CHAPTER 3. SELECTED MATERIALS
10−4
10−3
10−2
10−1
100
0
10
20
30
40
50
60
70
80
90
100
Diameter (mm)
Per
cent
age
finer
by
wei
ght (
%)
S2BS5B
Figure 3.2: Particle size distributions of S2B and S5B
fraction finer than 2 µm and a Plasticity index of 514%, resulting in an activity (A = PI/(%
finer than 2 µm)) of 6.425 which is a highly active soil. Bentonite is classified as high plasticity
clay (CH) according to the USCS.
Table 3.1: Atterberg limits of ActiveGel 150 bentonite adopted from Shannon (2008)
Liquid limit (LL) 550%
Plastic limit (PL) 36%
Plasticity index (PI) 514%
3.2.3 Linear shrinkage
A linear shrinkage test was performed according to the AS 1289.3.4.1 . The mould was lubricated
using hydraulic oil and the bentonite was placed in the mould with the water content at the
liquid limit of 550%. The mould was left in a constant temperature environment (22 ◦C) until
shrinkage had stopped. The linear shrinkage was found to be 50.4%.
3.2.4 Specific gravity
The specific gravities (Gs) of the sand, bentonite and the mixtures (2% and 5% bentonite) were
found using a pycnometer (Micro-meritics AccuPyc 1330). Before the measurements, soils were
29
dried in the oven of 105 ◦C overnight and were subsequently placed in a desiccator for a day
to avoid any surface moisture while cooling down. In addition, the instrument purges soils in
dry helium and vacuums ten times in the test chamber to remove any residual surface moisture.
The average values of the specific gravity after eight consecutive measurements are presented in
Table 3.2.
Table 3.2: Specific gravity of the materials
Soil type Gs
Sand 2.65
Bentonite 2.69
Sand-bentonite (2%) 2.65
Sand-bentonite (5%) 2.65
3.3 Compaction characteristics
The compaction characteristics of sand-bentonite mixtures, S2B and S5B, were determined and
are described in this section. Dynamic compaction tests for both standard and modified com-
pactive efforts were conducted according to the procedures described in the Australian Standards
(AS 1289.5.1.1-2003 and AS 1289.5.2.1-2003 ). The applied energy for standard compaction
test is 596 kJ/m3 and for modified compaction test is 2703 kJ/m3 in Australian Standards.
Figure 3.3 shows the compaction curves of S2B for standard and modified compactive efforts.
At a standard compactive effort, the maximum dry density of S2B was found to be about 1.72
Mg/m3 at an optimum water content of 11.5% and at a degree of saturation of about 50%.
In addition, at a modified compactive effort, the maximum dry density was found to be 1.78
Mg/m3 at an optimum water content of 10% and at a degree of saturation of about 50%.
For a standard compactive effort and 2% bentonite content, Ameta and Wayal (2008) attained
a maximum dry density of 1.71 Mg/m3 at 15% optimum water content. The maximum dry
density is similar but the optimum water content is higher compared to the results obtained
in this study. The compaction characteristics of bentonite used in this project was studied by
Shannon (2008), who obtained a maximum dry density of 1.20 Mg/m3 at an optimum water
content of 20% for the ActiveGel 150 bentonite. Bentonite used in both studies have similar
maximum dry densities of about 1.20 Mg/m3, whereas different optimum water contents of
30 CHAPTER 3. SELECTED MATERIALS
0 5 10 15 201.6
1.65
1.7
1.75
1.8
1.85
1.9
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
StandardModified
Sr = 70%Sr = 50%
Figure 3.3: Standard and modified compaction curves of S2B
20% and 47%, the lower being the optimum water content of the bentonite used in the current
study. Therefore, this results in a lower overall optimum water content in the mixture of the
sand bentonite used in the current study.
Chalermyanont and Arrykul (2005), however, attained a maximum dry density of 1.97
Mg/m3 at an optimum water content of 10%, at a standard compactive effort. They used
a finer sand with a coefficient of uniformity (Cu) of 11.7 and coefficient of curvature (Cc) of 1.0,
which was classified as well graded silty sand (SW-SM) according to the USCS. Therefore, they
obtained a higher maximum dry density than the one obtained in this study.
In addition, Cavalcante Rocha and Didier (1993) have found a maximum dry density of 1.69
Mg/m3 at an optimum water content of about 10.5%. They have used a uniform sand and their
results are pretty close to the ones attained in this study. Therefore, it can be concluded that
the compaction characteristics of sand containing 2% bentonite by weight (S2B) are within a
range that is comparable with the results found in the literature.
Figure 3.4 shows the compaction curves of S5B for standard and modified compactive efforts.
At a standard compactive effort, the maximum dry density of S5B was found to be about 1.75
Mg/m3 at an optimum water content of 13% and at a degree of saturation of about 63%. In
addition, at a modified compactive effort, the maximum dry density was found to be 1.81 Mg/m3
at an optimum water content of 13% and at a degree of saturation of about 65%. The usual
31
0 5 10 15 201.6
1.65
1.7
1.75
1.8
1.85
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
StandardModified
Sr = 60% Sr = 80% Sr = 100%
Figure 3.4: Standard and modified compaction curves of S5B
decrease of the optimum water content with the increasing compactive effort has not been seen
for this soil type.
For a standard compactive effort and 5% bentonite content, Cavalcante Rocha and Didier
(1993) attained similar results with a maximum dry density of 1.73 Mg/m3 at an optimum water
content of 11%. Chalermyanont and Arrykul (2005), however, attained a maximum dry density
of 1.94 Mg/m3 at an optimum water content of 10.5%. They used a finer sand, classified as well
graded silty sand (SW-SM) according to the USCS. Therefore, they obtained a higher maximum
dry density than the one obtained in this study. Komine and Ogata (1999), on the other hand,
found a maximum dry density of 1.61 Mg/m3 at an optimum water content of 19.4%. The
maximum dry density attained by Komine and Ogata (1999) is quite low compared to the other
results found in the literature. The comparison of the results of current study and the ones
found in literature for 5% bentonite content shows that the results attained in this study are
within expected range.
The compaction test results for both soil types used in the current study are also presented
in Table 3.3. Comparison of the compaction curves for two soils shows that when the bentonite
content increases the maximum dry density and the optimum water content of the soil also
increases for both compactive efforts. Since the sand used was poorly graded, an increase in
bentonite content resulted in voids being filled with more fines instead of an increase in the
32 CHAPTER 3. SELECTED MATERIALS
Table 3.3: Summary of the compaction characteristics for S2B and S5B
SoilStandard Modified
γdmax OWC γdmax OWC(Mg/m3) (%) (Mg/m3) (%)
S2B 1.72 11.5 1.78 10S5B 1.75 13 1.81 10
volume of soil. The results of Komine and Ogata (1999) and Cavalcante Rocha and Didier
(1993), who used uniform sand in their research, also show an increase in the maximum dry
density with an increase in bentonite content. Chalermyanont and Arrykul (2005), on the other
hand, observed a decrease in the maximum dry density with an increasing bentonite content.
However, they used a well graded fine sand, which lead to a lower void ratio within the soil.
Bentonite has a higher optimum water content than the optimum water content of sand.
Therefore, this results in increase of optimum water content with increasing bentonite content.
3.4 Soil water retention curve
The soil water retention curve (SWRC), which is also referred to as the soil water characteristic
curve or moisture retention curve is a relationship between suction and water content (gravi-
metric or volumetric) or degree of saturation. Several ways of suction measurements in order to
determine the SWRC of the soil were described in Section 2.3.
The SWRCs for the soils used in this study, S2B and S5B, are presented in this section
together with the type of equipments and methods used for their determination. Two differ-
ent methods for matric suction measurements were employed, axis translation technique and
tensiometer measurements. Tensiometer measurements were conducted for comparison of the
results obtained by axis translation technique.
Both of the techniques used for matric suction measurements are described in detail below.
Firstly, the equipments and methods used to determine the SWRCs will be summarised and
that will be followed by the results of both techniques.
3.4.1 Axis translation technique
The drying curves for the samples S2B and S5B were determined using axis translation technique.
Due to the high sand content in both of the specimens (S2B and S5B), the drying curves were
measured within a range of 0-60 kPa matric suction.
33
Pressure gauge panel
Sample
Valve Scale
Outflowcontainer
Air supply
PC
Tempe cell
Figure 3.5: Schematic of the experimental setup for axis translation technique
Equipment
The schematic diagram and the photo of the experimental setup for axis translation technique
are given in Figures 3.5 and 3.6. The setup consist of three main components: a pressure cell
assembly (tempe cell), a pressure panel and data acquisition system. The tempe cell was designed
and developed by previous researchers at Monash University workshop. A high air entry ceramic
disk was fitted inside the tempe cell for matric suction measurements.
Method
The permeability of the high air entry ceramic disk was checked before the test was commenced
in order to find out if there were any cracks on the disk or any leakage around the disk. In
addition, the capacity of the ceramic disk was unknown and thus, it was necessary to check
whether it was enough to serve within the experimental range which was about 60 kPa. To
calculate the permeability of the ceramic disk the tempe cell was filled with deaired water and
a pressure of 20 kPa was applied in the tempe cell to create a hydraulic head. The thickness
and the diameter of the high-air entry ceramic disk was 10 mm and 70 mm respectively. The
permeability of the disk was calculated using Darcy’s law (Darcy, 1856) given with the following
equation:
k =VwAt
h
H(3.1)
34 CHAPTER 3. SELECTED MATERIALS
Figure 3.6: Experimental setup of axis translation technique
where; k is the permeability of the ceramic disk in m/s, Vw is the water volume discharge in
m3, A is the cross sectional area of the ceramic disk in m2, t is the time in s, h is the thickness
of the ceramic disk in m and H is the hydraulic head in m.
The water volume discharge with time under 20 kPa applied air pressure is shown in Figure
3.7. The permeability of the ceramic disk is calculated by multiplying the slope of the curve by
h/(AH) and is found to be 1.34x10−10 m/s. The permeability of the disk was compared with
the ones produced by Soil Moisture Equipment Corporation and it was found that it can serve
well above the experimental range which was about 60 kPa. In addition, the permeability of the
disk was compared with the ones used by Padilla et al. (2006); and based on those values, the
ceramic disk has a permeability value equivalent to the permeability of between 5 and 15 bar
high-air entry ceramic disks.
The soil mixed with water at optimum water content was placed in the metal ring with a
desired dry density (all the specimens were matured for 28 days after mixed with water for
hydration and equal distribution of moisture within the soil, see Section 4.3.1). The diameter of
the ring used was 64 mm and the height was 18 mm. The dry density chosen for this study was
1.58 Mg/m3. The choice for the initial void ratio value will be described in Section 4.3.1.
Once the soil samples were prepared with a desired dry density, they were placed on a porous
stone covered with a filter paper and were subsequently placed in a container. The container was
35
0 1 2 3 4 5 6 7
x 104
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5x 10
−6
Time, t (s)
Wat
er v
olum
e di
scha
rge,
Vw (
m3 )
y = 5.1e−011*x + 9.5e−007
Figure 3.7: Water volume discharge with time under 20 kPa applied air pressure
filled with deaired water until the water level was slightly below the top of the specimens. This
allowed the saturation of the specimens from bottom up and therefore, release of the entrapped
air from the soil. The soil samples were left overnight for saturation.
To saturate the ceramic disk, it was placed in a vacuum flask and submerged in deaired
water. The vacuum flask was connected to a vacuum pump for 15 minutes and the high air
entry ceramic disk was left for 24 hours under vacuum to ensure the complete removal of air
bubbles trapped in it. The ceramic disk was then placed in the base of the tempe cell ensuring
that no air was trapped under the disk. The tubing connection between the base of the tempe
cell and outflow container was carefully checked for air bubbles assuring that no air bubbles were
left in the tube.
In order to accurately measure water volume changes, it was necessary to account for water
mass or volume losses due to evaporation from the outflow container. As can be seen from Figure
3.6 the lid of the outflow container was placed on the container and only a small hole was drilled
to open it to atmosphere. Figure 3.8 shows the amount of water evaporated from the container
with time. The evaporation rate was found to be 0.0002 ml/min. This amount was taken into
account for the water content or degree of saturation calculations.
After saturating the soil samples, they were removed from water and placed on the ceramic
disk without the porous stone and filter papers. A thin layer of kaolin paste was placed between
36 CHAPTER 3. SELECTED MATERIALS
0 500 1000 1500 2000 2500 30000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Time (min)
Eva
pora
tion
(ml)
y = 0.000183*x + 0.104
Figure 3.8: Evaporation of water from the outflow container
the sample and the ceramic disk in order to ensure a good contact between them. The tempe
cell was after mounted and was checked for air leakage. The outflow container was placed onto
the scale and the scale was connected to a data acquisition system which took readings every
4 minutes. The level of the scale was arranged so that the water level in the outflow container
would be slightly above the top level of the soil sample. The soil sample was also saturated
within the tempe cell until the system came to equilibrium.
The targeted air pressures were applied after the saturation and the system was left to equi-
librate. The drainage of the water from the specimen continued until it reached the equilibrium
condition with applied air pressure or matric suction. The air pressure was increased by 1 kPa
from 0 to 10 kPa, 2 kPa from 10 kPa to 16 kPa, 5 kPa from 16 kPa to 30 kPa and 10 kPa from
30 kPa to 60 kPa for S2B. For S5B the pressure was increased by 1 kPa from 0 to 6 kPa, 2 kPa
from 6 kPa to 20 kPa, 5 kPa from 20 kPa to 30 kPa and 10 kPa from 30 kPa to 60 kPa. The
same procedure was repeated for every air pressure increment.
After the last pressure increment the pressure within the tempe cell was released, the tempe
cell was dismantled and the soil sample was taken out as quickly as possible to avoid the water
intake into the soil sample from the ceramic disk. The moist soil samples were weighted and
then placed in an oven of 105 ◦C for 24 hours to find the final water content. The water content
for each of the applied air pressure was then back calculated based on the final water content
37
CERAMICDISK
PLASTICBODY TUBE
TRANSDUCER
Figure 3.9: Tensiometer used for matric suction measurements
and the drying SWRCs were plotted. The results are shown in Section 3.4.3.
3.4.2 Tensiometer measurements
Matric suction measurements were also conducted with tensiometers in order to compare the
results obtained by the axis translation technique. Measurements with tensiometers took less
time than the measurements conducted with the axis translation technique. 3 tensiometers were
available in the laboratory and therefore 3 samples with different water contents could be tested
at the same time. This corresponds to 3 points on the SWRC.
Equipment
Tensiometers with a capacity of 100 kPa were used for matric suction measurements. Tensiome-
ters were provided by ICT Australia. The tensiometer consists of a ceramic disk with a capacity
of 100 kPa air entry value, a plastic body tube assembly that connects to the ceramic disk, a
transducer and a data acquisition system. A photo of the tensiometer is shown in Figure 3.9.
Method
The soil samples prepared with the predetermined water content were placed in small containers
with a dry density equivalent to 1.58 Mg/m3 (same as the dry density in the axis translation
technique). Tensiometers were inserted subsequently into the soil samples carefully from the
top to avoid soil disturbance, and therefore, the density of the soil samples. Samples were then
38 CHAPTER 3. SELECTED MATERIALS
covered with multiple layers of cling wrap and aluminium foil to avoid any evaporation of water
from the specimens. Tensiometers were left to equilibrate to the matric suction level of the
specimens. Experiments were repeated until the measurements were accomplished, in case there
was a breakage in tensiometers before the equilibration. The water content of the specimens
were measured both before and after the test to ensure that no evaporation occurred during
equilibration since the experiment took considerable time to complete.
3.4.3 Results
Matric suction measurements conducted with tensiometers are shown in Figures 3.10 and 3.11.
It took longer time for dryer samples to equilibrate compared to relatively saturated samples for
both soils (S2B and S5B). Since samples prepared with different water content would result in
different microstructure and therefore in different unsaturated permeability, this was an expected
outcome evident from the work of Oliveira and Marinho (2008). The samples prepared at
dry of optimum water content will have an aggregate microstructure, compared to the matrix
microstructure when prepared at the wet of optimum water content (Delage et al., 1996; Romero
et al., 1999). The equilibration time depends on the transfer rate of water between the sample
and the ceramic disk or the tensiometer. As stated by Delage et al. (2008), when the soil sample
has an aggregate microstructure the water transfer is through the inter-aggregates contacts and
inside the inter-aggregates smaller pores; thus, it would take longer time for drier samples to
equilibrate. It was also noted that, S5B has relatively shorter equilibration time compared to
S2B. This is attributed to its finer nature compared to S2B. The presence of higher amount
of fines in soil will result in a better contact between the sample and the ceramic disk, and
eventually this would lead to a shorter equilibration time.
The SWRCs for S2B and S5B are shown in Figures 3.12 and 3.13. Matric suction measure-
ments obtained by both axis translation technique and tensiometer are shown on the figures.
The matric suction measurements obtained by axis translation technique are higher than the
tensiometer measurements. This is due to the different methods used in sample preparation.
The measurements conducted using axis translation technique present the drying path, whereas
the tensiometer results lie on the wetting path of the SWRCs. It should also be noted that
the drying path was obtained from one sample, which had the same pore size distribution for
every measured matric suction or degree of saturation. On the other hand, different samples
with varying water contents were prepared for tensiometer measurements. Sample preparation
39
0 10 20 30 40 5010
−1
100
101
102
Time, t (hours)
Mat
ric s
uctio
n, s
(kP
a)
w = 2.6%w = 3.8%w = 5.2%w = 7.3%w = 9.6%w = 12%
Figure 3.10: Matric suction measurements of S2B with tensiometer
0 10 20 30 4010
0
101
102
Time, t (hours)
Mat
ric s
uctio
n, s
(kP
a)
wc = 3%wc = 3.6%wc = 6.3%wc = 7.7%wc = 9.4%wc = 14%
Figure 3.11: Matric suction measurements of S5B with tensiometer
40 CHAPTER 3. SELECTED MATERIALS
with different water contents and same density results in different pore size distribution of the
material. Therefore, the drying curve results were obtained from a unique pore size distributed
sample. Whereas, the wetting curves were obtained from a varying pore size distributed samples.
100
101
102
0
10
20
30
40
50
60
70
80
90
100
Matric suction, s (kPa)
Deg
ree
of s
atur
atio
n, S
r (%
)
Axis translationTensiometer
Figure 3.12: SWRC of S2B
100
101
102
10
20
30
40
50
60
70
80
90
100
Matric suction, s (kPa)
Deg
ree
of s
atur
atio
n, S
r (%
)
Axis translationTensiometer
Figure 3.13: SWRC of S5B
41
The air entry values obtained from the drying curves are about 3 kPa and 4 kPa for S2B and
S5B respectively (Figures 3.12 and 3.13). The matric suctions for S5B are slightly higher than
the matric suctions obtained for S2B due to the finer nature of S5B.
3.5 Consolidation characteristics
Oedometer tests were performed in order to find the saturated normal compression lines for both
soils (S2B and S5B). The slope of the normal compression line (λvp) for saturated samples is
required for the modelling part of this study and will be used in Chapter 7.
3.5.1 Equipment
Load Track III was used for oedometer tests and was supplied by Geocomp Corporation. Load
Track III consists of a loading frame, consolidation cell and a computer for data acquisition.
The equipment is shown in Figure 3.14. The load frame contains an embedded control system
that generates a desired force on a soil sample and measures the displacement. In addition, it
communicates with the transducers to provide a real-time control of the loading frame.
Figure 3.14: Load Track III used for oedometer tests
42 CHAPTER 3. SELECTED MATERIALS
100
101
102
103
104
1.6
1.62
1.64
1.66
1.68
1.7
Vertical effective stress, σy − u
w (kPa)
Spe
cific
vol
ume,
ν
Experimental data
λ = 0.013
κ = 0.006
Figure 3.15: Slope of the normal compression line for saturated S2B
3.5.2 Method
Soil samples were compacted at an optimum water content to a desired density (γdry = 1.58Mg/m3)
in a consolidation ring. Subsequently, the ring was placed in a consolidation cell and the cell was
filled with deaired water to achieve complete saturation of the samples. The samples were satu-
rated for 12 hours before the consolidation commenced. After the saturation the consolidation
cell was placed on the loading frame and loading specifications were installed to the computer
software that controls the frame.
3.5.3 Results
The consolidation results for both soils are presented in Figures 3.15 and 3.16. The slopes of
the normal compression lines and the slopes of the unloading curves are calculated from the
experimental results and are shown on the graphs.
The slope of the normal compression line (λ) for soil sample S2B is 0.013 and the slope
of the unloading curve (κ) is calculated as 0.006. (Figure 3.15). For sample S5B, the slope
of the normal compression line (λ) is 0.021 and the slope of the unloading curve (κ) is 0.008
(Figure 3.16). Sample S5B has a higher loading and unloading slope than S2B due to the higher
bentonite content in it.
Evans and Ryan (2005) attained a slope of 0.032 for the normal compression line where the
43
100
101
102
103
104
1.45
1.5
1.55
1.6
1.65
Vertical effective stress, σy − u
w (kPa)
Spe
cific
vol
ume,
ν
Experimental data
λ = 0.021
κ = 0.008
Figure 3.16: Slope of the normal compression line for saturated S5B
soil had 5% bentonite content by weight. They used a sand that was classified as sandy clay
(SC) according to the USCS. This result is slightly higher than the result found in this study
due to the different nature of the sand used in both projects. The sand used by Evans and
Ryan (2005) was a finer sand (SC) than the one used in the current work (SP). Therefore, it is
expected to deform more under the applied vertical stress and have a higher slope of the normal
compression line. In addition, the slope of the unloading curve was found to be 0.002 by Evans
and Ryan (2005), which is lower than the one found in the current study. This might be due to
the finer nature of the sand used by Evans and Ryan (2005). Since the plastic deformation in
clayey soil is higher compared to sandy soils, it is acceptable for the clayey soil to have a lower
unloading slope, which relate to elastic recovery of deformation.
3.6 Concluding remarks
A bentonite content of 2% and 5% has been selected to be mixed with poorly graded sand to
be investigated in this study. Both of the selected materials have a suction range that could be
measured with the equipment available in the laboratory.
The physical properties and compaction characteristics of the selected materials were deter-
mined according to Australian Standards. The sand used is classified as SP and the bentonite
44 CHAPTER 3. SELECTED MATERIALS
is classified as CH according to the USCS. The maximum dry density of S2B was found to be
1.72 Mg/m3 at an optimum water content of about 11.5% at a standard compactive effort. In
addition, the maximum dry density of the same soil for modified compactive effort was found
to be 1.78 Mg/m3 at an optimum water content of 10%. The maximum dry density of S5B
was found to be 1.75 Mg/m3 at an optimum water content of 13%, at standard compactive
effort, and 1.81 Mg/m3 at an optimum water content of 13% for modified compactive effort.
An increase in maximum dry density and water content was observed with increase in bentonite
content.
The water retention behaviour of the materials was determined using two different methods,
axis translation technique and tensiometer measurements. The air entry value of S2B was
calculated as 3 kPa and the air entry value of S5B was calculated to be 4 kPa. In the overall
water retention curve S5B had slightly higher matric suctions than S2B owing it to its finer
fraction. In addition, the matric suctions measured with axis translation technique were on the
drying path and the matric suctions measured with tensiometer were on the wetting path of
the SWRCs. For that reason, tensiometer measurements for the same water content or degree
of saturation had lower matric suction values than the measurements using axis translation
technique.
Consolidation tests were performed to determine the slope of the normal compression line.
The slopes of the normal compression lines for S2B and S5B were measured to be 0.013 and
0.021. The S5B has a higher slope of the normal compression line than the slope of the S2B due
to the higher clay content. In addition, the slopes of the unloading curves for S2B and S5B are
0.06 and 0.08, respectively.
Chapter 4
Testing equipment and
experimental procedure
4.1 Introduction
Compaction characteristics of soils can be better understood through carefully monitored ex-
periments. Therefore, the testing instruments and their working techniques are crucial in the
interpretation of the results. Another key parameter that leads to a successful outcome of a
research project is the choice of the experimental methods. This chapter is devoted to the both
instrumentation and experimental procedure followed throughout this research project.
The primary objective of the proposed experimental procedure was to conduct constant
water content static compaction tests to represent the compaction process in man made fills or
hydraulic barriers. In addition, the experimental methods were chosen to best represent the
compaction models so that the models could be verified.
The targets of the experiments can be summarised as follows:
1. To measure the initial matric suction of the soil samples;
2. To measure the matric suction changes of the soil specimens during the constant water
static compaction test; and,
3. To measure the volume change of the soil specimens during the constant water static
compaction test.
45
46 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
This chapter describes the equipments used and the experimental procedures followed to
achieve the objective and the targets of the experiments. The first part details the equipments
and accessories used during the testing. The second part describes the paths and methods
of the experimental procedure. Finally, the problems and difficulties encountered during the
experiments are presented.
4.2 Testing Equipment
The soil samples were statically compacted with a device that can be used to measure the initial
matric suction and the matric suction changes during the compaction. The device measures
matric suction with the axis translation technique. The compaction devices and its working
principles are briefly described in this section. In addition, the accessories used in the setup of
the compaction device are summarised.
4.2.1 Compaction device
Static compaction experiments were conducted with SWC-150 Fredlund Device manufactured
by Geotechnical Consulting and Testing Systems (GCTS). The device was modified by adding
a pressure transducer to the bottom of the plate (below the ceramic disk) in order to monitor
the pore water pressure during the static compaction test. This modification also helped to
measure the initial matric suction values of the specimens prepared with predetermined water
content based on the null-type axis translation technique proposed by Hilf (1956). In addition,
a linear variable differential transformer (LVDT) was used in order to track volume changes of
the samples during compaction. The pressure transducer and LVDT were connected to the data
taker. The schematic of the modified equipment setup is shown in Figure 4.1. The setup consists
of a pressure cell assembly, a pressure panel with dual gauges, a loading frame, a pressure booster
and a data acquisition system. The entire setup, excluding the pressure booster and the data
acquisition system, is shown in Figure 4.2.
The SWC-150 Device was originally designed to replace pressure plate devices. Different from
the pressure plate device, changes in the water volume can be continuously monitored without
dismantling the device (Perera et al., 2005). In addition, the device has a loading frame as shown
in Figure 4.2, which helps applying vertical loads to simulate the field overburden pressure. In
this study the loading frame was used to apply a load to the soil sample in order to implement
47
Air SupplyPressurebooster
Pneumaticloader
Loadingframe
LVDT
Releasevalve
Data taker
Pressuretransducer
Steelblock
Computer
B A
Opening L Opening R
Airpressureopening
Water volumechange tube
C
Pressuregauge
Valve
Valves
High
pressureLow
pressure
Low
High
Air
Water
Figure 4.1: Schematic of the compaction device setup
Loading frame
Pneumatic loader
Pressure transducer
Top plate
Bottom plate
Water volume change tube
LVDT
Figure 4.2: Compaction device
48 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
To pressure transducer
To water volume tube
Three way valve
To water volume tube
Figure 4.3: Base of the pressure cell
the static compaction test. The system can be flushed with an air pump to remove the diffused
air accumulated beneath the high air entry ceramic disk.
The base of the pressure cell is shown in Figure 4.3. The ceramic disk is embedded into the
groove of the base cell. There is an O-ring between the ceramic disk and the base to create a
sealing interface. However, Zhang et al. (2009) have experienced air leakage problem from this
interface. Therefore, the system was checked for air leakage while measuring the permeability of
the ceramic disk and the results are given in Section 4.2.2.
The base of the pressure cell was connected to the calibrated water volume tubes which were
placed on the pressure panel (see Figure 4.2) via plastic tubes. A three way valve (see Figure
4.3) was added in one of the plastic tubes to connect the base with the pressure transducer which
monitors the water pressure below the ceramic disk (valve C in Figure 4.1). This modification
was done in order to measure the initial matric suction using the null-type axis translation
technique and monitor the pore water pressure, or matric suction, in a constant water content
test. The pressure transducer was mounted on a steel block, which is connected to the base of
the pressure cell, as shown in Figure 4.4. A release valve was also mounted on the steel block to
remove the air bubbles from the plastic tube that connects the base of the pressure cell and the
pressure transducer.
49
Pressure transducer
Release valve
To the base of the pressure cell
Figure 4.4: Pressure transducer mounted on a steel block
The pressure cell is connected to the pressure panel via a plastic tube from the upper part
of the pressure cell. The device can apply air pressures between 0 and 2000 kPa.
The pressure cell was placed under a loading frame which can apply 10 kN load. Considering
the diameter of the specimen used in the experiment (75 mm for S2B and 64 mm for S5B) this
corresponds to about 2250 kPa and/or 3200 kPa applicable pressure to S2B and S5B samples,
respectively. A LVDT was placed between the specimen and the cell in order to monitor the
vertical displacement, or the volume change, of the specimen. The sample was placed in a steel
ring of 1.5 mm thickness.
The original device had been previously used for SWRC measurements by Mavroulidou et al.
(2009) and Chao (2007) for clay, by Li and Zhang (2007) for five types of decomposed granite
- all classified as coarse grained soils, by Chakraborty (2009) for soil bentonite mixtures with
soil being mainly sand, by Perera (2003) for both disturbed granular and undisturbed fine
grained samples and by Zapata et al. (2009) for granular material. In the current study, the
device was modified and used for the null-type initial matric suction measurements and for the
measurements of the matric suction changes during compaction for the first time according to
literature.
50 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
4.2.2 Accessories for the compaction device
The compaction device was used with auxiliary devices in order to conduct the desired exper-
iments (constant water content static compaction tests). A pressure transducer was used to
measure the pore water pressure below the ceramic disk, a LVDT was used to monitor the de-
formation of the samples, an air booster was used to increase the air pressure available by the
university central source and a data taker was used to measure and record the parameters of the
pressure transducer and the LVDT. These devices are briefly summarised in this section.
High air entry value ceramic disk
The ceramic disk was purchased glued to a metal ring by an epoxy glue from GCTS. The
diameter and the height of the ceramic disk are 75 mm and 10 mm respectively. A ceramic disk
with an air entry value of 1 bar was selected to use throughout the experimental program.
The ceramic disk was placed in a vacuum flask and submerged in deaired water for saturation.
The vacuum was applied to the vacuum flask to remove any air bubbles left within the ceramic
disk. The ceramic disk was left under vacuum overnight for saturation.
The saturated ceramic disk was placed in its groove (see Figure 4.3) to check for any possible
cracks and leakage within the system. An air pressure was applied gradually within the cell and
the bubbling pressure of the ceramic disk was found to be about 200 kPa. This value is within
a given range for the same disk by the manufacturer. Therefore, the ceramic disk was found to
have no major cracks and the sealing interface between the ceramic disk and the base of the cell
seems to have no leakage problems.
To check the permeability of the disk, the pressure cell was filled with deaired water and an
air pressures of 20 kPa, 50 kPa and 100 kPa were applied inside the pressure cell to create a
hydraulic head. The permeability of the ceramic disk was calculated based on the Darcy’s law
(Darcy, 1856) from the Equation 3.1. The water volume discharge (Vw) from the ceramic disk
with time (t) under 20 kPa, 50 kPa and 100 kPa applied air pressures are shown in Figure 4.5.
The water volume discharge through the ceramic disk within a ceratin time is also presented,
under 20 kPa, 50 kPa and 100 kPa, in Figure 4.6. The slopes of these curves give the average
permeability of the ceramic disk, which is calculated as 4x10−8m/s. The permeability of the
ceramic disk is within an expected range as indicated by Padilla et al. (2006). The leakage
problem that was experienced by Zhang et al. (2009) at the interface of the base cell and the
ceramic disk might be due to damage of the O-ring,
51
0 500 1000 15000
1
2
3
4
5
6
7x 10
−6
Time (s)
Wat
er v
olum
e di
scha
rge
(m3 )
20 kPa50 kPa100 kPa
y = 2.8e−008
y = 1.1e−008
y = 4.13e−009
Figure 4.5: Water volume discharge with time under 20 kPa, 50 kPa and 100 kPa applied airpressures
0 20 40 60 80 1000
0.5
1
1.5
2
2.5
3
3.5
4
4.5x 10
−6
Hydraulic head (kPa)
Wat
er v
olum
e di
scha
rge
(m3 )
y = 4e−008R2 = 0.9713
Figure 4.6: Average permeability of 1 bar ceramic disk
52 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
Garga and Zhang (1997) studied the rate of air dissolution in deaired water and have found
that the deaired water exposed to atmosphere for 10 hours had the same oxygen content of
that of tap water. Therefore, a test was conducted to study the amount of air, passing through
ceramic disk in the form of dissolved air or free air (from the interface of base cell and ceramic
disk). To do that, the saturated ceramic disk was placed on the based of the pressure cell and an
air pressure of 10 kPa was applied for a period of 5 days. At the end of the experiment (after 5
days) the water volume tubes were flushed by an air pump and the changes of the water volume
at the each tube were recorded. It was found that 0.6 ml of air was diffused from the ceramic
disk in 5 days. Therefore, during the static compaction test the water under the ceramic disk
was flushed regularly (once a day) to avoid the accumulation of diffused air under the disc which
might lead to loss of the continuous water connection between the ceramic disk and the water
under the ceramic disk (Vanapalli et al., 2008).
Pressure transducer
Druck PDCR 4011 pressure transducer was used to measure the pore water pressure of the
system under the ceramic disk. The pressure transducer was selected according to two main
reasons. First, the pressure transducer required to be in a range of being able to measure
the matric suction change during the static compaction tests. Second, it needed to be sensitive
enough to detect the small matric suction changes during the compaction tests which was hard to
detect with high capacity transducers due to their lower accuracy level. The Druck PDCR 4011
transducer has an operating pressure range of 0 - 70 kPa and the accuracy of the transducer is
0.04% full scale best fit straight line (FSB). It can successfully operate between -20◦ and +80◦C.
The transducer was calibrated with known pressures of the SWC-150 Device. The calibration
curve is shown on Figure 4.7.
Linear variable differential transformer
Volume changes in the soil samples were tracked using a KYOWA DTH-A-10 LVDT. The LVDT
has a capacity of 10 mm and an accuracy of 0.1% rated output (RO). The instrument can
operate from 0◦ to 60◦C successfully.
The LVDT was calibrated using an LVDT calibrator. The calibration curve is shown in
Figure 4.8.
53
−50 0 50 100 1500
10
20
30
40
50
60
70
80
Pressure transducer reading
App
lied
air
pres
sure
, ua (
kPa)
y = 0.4618*x + 10.37
Figure 4.7: Calibration curve of the pressure transducer
−3000 −2000 −1000 0 1000 2000 30000
1
2
3
4
5
6
7
8
9
10
LVDT reading
Dis
plac
emen
t (m
m)
y = 0.002*x + 5.5
Figure 4.8: Calibration curve of the LVDT
54 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
Figure 4.9: GCTS pressure booster
Pressure booster
Air pressure source at laboratory was limited to 700 kPa and PCP-PBOOST pressure booster
manufactured by GCTS was used to increase the available pressure. The pressure booster can
increase the output pressure to 2000 kPa. A photo of the pressure booster is shown in Figure
4.9.
Data taker
The data of the pressure transducer and the LVDT were measured by DT500 data taker. The
data taker was programmed and the data was archived by DeTransfer software. The data taker
was connected to a computer running Windows XP.
4.3 Experimental procedure
The flow chart of the experimental procedure is given in Figure 4.10. Initially, the sand bentonite
samples were prepared and left for hydration. Hydrated samples were then placed into the
metal ring by the procedure described in Section 4.3.1 and the initial matric suctions of the
samples were measured. The device measures the matric suction using null type axis translation
technique. Axis translation technique cannot be used when the air phase in the specimens
becomes discontinuous (Bocking and Fredlund, 1980). In addition, Olson and Langfelder (1965)
55
Experimental Procedure
Suction
measurements
Static compaction
test
Sample preparation
and curing
Compaction device
(modified SWC-150)
Null type axis translation technique
Continuous suction measurement
during compaction test
Figure 4.10: Experimental procedure
and Langfelder et al. (1968) showed that the air phase becomes occluded at about optimum
water content. However, the axis translation technique was experimentally evaluated for soils
having a degree of saturation between 76% and 95% by Fredlund and Morgenstern (1977) and
for soils having a degree of saturation between 56% and 77% by Tarantino et al. (2000). Dynamic
compaction test results (Figures 3.3 and 3.4) for both soils used in the current study shows that
at the end of the compaction test the final degree of saturation of the soil samples will be less
than 95%. Therefore, the matric suction of the samples were measured and monitored in the
compaction device using null type axis translation technique. The experimental procedure is
described in detail in the following sections.
4.3.1 Material preparation
The selection of appropriate material preparation techniques in granular soils is crucial, since the
behaviour of the soils can be affected by the method of preparation (Mulilis et al., 1977). There
56 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
are several material preparation techniques for granular soils found in literature which could
be categorised according to their moisture condition, soil placement method and the medium
through soil is placed. The most common ones that are used in the preparation of sandy soils are
air pluviation, water pluviation and moist tamping methods. In addition, the slurry deposition
method developed by Kuerbis and Vaid (1988) for the preparation of well graded and silty sands
is also considered in this study.
The samples prepared by air pluviation represent the natural deposition of wind blown aeolian
deposits (Kuerbis and Vaid, 1988) and water pluviated samples simulate the natural deposition
of sand through water in many natural environments (Vaid and Negussey, 1984). The driving
forces for both of these methods is gravimetry and therefore leads to a segregation of the fine
particles during preparation procedure. Moist tamping method, on the other hand, best models
the soil fabric of construction fills (Lambe, 1951). In addition, the slurry deposition method best
simulates the soil fabric found within a natural fluvial deposit (Kuerbis and Vaid, 1988). For
uniformly graded soils, air and water pluviated samples are more homogenous than the moist
tamped samples (Vaid et al., 1999; Frost and Park, 2003); whereas, the slurry deposition method
produces the most homogenous samples for sand with fines.
Most of these methods were used in the preparation of soil samples for saturated soil me-
chanics problems. There is not much information available in literature on the applicability of
these methods for the preparation of unsaturated soil samples. There are two important criteria
for the selection of material reconstitution method in this study:
1. the sample should be well mixed and have a uniform void ratio throughout; and,
2. the sample preparation technique should simulate the soil fabric found in soil fills.
Air pluviation, water pluviation, and slurry deposition method cannot be used in this study,
since the initial water content of the samples can not be controlled. In these methods, the initial
water content can only be controlled by saturating or drying the initially prepared samples which
would lead to a different soil fabric properties than the desired initial ones. By the moist tamping
method, on the other hand this issue could be easily overcome. In addition, the moist tamping
method is originally found by Lambe (1951) for the construction of fills. Although the moist
tamping method results in less uniform soil fabric compared to the other noted methods, the
study conducted by Frost and Park (2003) shows that the change in void ratio with height of
the sample is not that significant. Therefore, moist tamping by layers is chosen as a material
57
Figure 4.11: Sample preparation by tamping
preparation technique in this study. This method was also selected by Rojas et al. (2008a) and
Lauer (2005) for the preparation of unsaturated soil samples.
Initially, sand and bentonite were mixed in an air dry state by hand for 15 minutes. The
desired amount of water was then added and mixed for another 20 minutes to evenly distribute
the moisture. The samples were first attempted to be mixed in a mixer; however, due to the
uneven mix of the fine and coarse particles the use of mixer was abandoned. The major difficulty
of using a mixer during the soil mixing procedure was the separation of fine and coarse particles
while the blade of the mixer was rotating. The fine particles were accumulating on the inner
wall while the coarse particles were staying in the centre of the mixer container.
The mixture was then wrapped with several layers of plastic wrap and aluminum foil. All
the samples were placed in air tight plastic zip bags and stored in plastic containers in the high
humidity room for a minimum of 28 days. Montanez (2002) conducted a study to investigate
the influence of the hydration time on the soil specimens. He stated that it is necessary to allow
a minimum period of 28 days for hydration in order to establish an equal distribution of suction
within the material.
Hydrated soil samples were placed in a metal ring in 5 thin equal layers by softly tamping
the specimen with a metal disk (or hammer) as shown in Figure 4.11. The diameter of the metal
disc was slightly less than the diameter of the steel ring so that it can freely move within the
ring.
The dry density was desired to be the lowest dry density attainable since one of the purposes
of this study was to observe not only the final soil properties of the compacted soils but also the
evolution of the compression properties during the compaction process. There were two concerns
for deciding the initial dry density. The first aim was to attain the lowest density possible.
Therefore, while soil was saturating under zero applied total or net vertical stress it should not
58 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
collapse. The second aim was to stop the swelling of the sample while saturating. Prepared soil
samples contain only 2% and 5% bentonite content by weight and therefore, swelling was not
considered to be an issue (Montanez, 2002). Taking into account these concerns, the initial dry
density of the samples was chosen to be 1.43 Mg/m3 initially. However, after testing the samples
with 2% bentonite content, it was noticed that as water content was increasing it was harder to
maintain such low density. Therefore, the dry density of the samples with 5% bentonite content
was changed to 1.58 Mg/m3. The tests for the samples with lower initial density (γd = 1.43
Mg/m3) were not repeated and are presented in Chapter 5 due to long duration of the tests. The
diameter and height of S2B samples were 75 mm and 25 mm respectively. The sample diameter
was reduced to 64 mm when testing S5B samples in order to shorten the long equilibration time
observed during the initial tests (S2B samples).
4.3.2 Static compaction test
Hydrated soil samples were statically compacted in order to produce the static compaction curves
of the soils and to compare them with the proposed compaction model. The stress path followed
during the test and the loading rate, the preparation of the compaction device for the test,
measurement during the test and the dismantling of the device are explained in this section.
Test stress path and loading rate
Figure 4.12 shows the stress path followed during the compaction experiment. Due to the
reduction in volume during the constant water content tests, there was a decrease in matric
suction with increasing net vertical stress or applied vertical stress (since air phase is considered
to be drained applied vertical stress is equal to the net vertical stress).
There are two main types of loading rates found in literature; time dependent loading rate
and equilibration or stabilisation dependent loading rate. It was shown in numerous studies
that the time dependent rate of loading has an influence on the volume change characteristics of
unsaturated soils (Huat et al., 2006; Rojas et al., 2008a). Huat et al. (2006) conducted an exper-
imental study on the influence of loading rate to the volume change behaviour of unsaturated
granitic residual soil. They found that the higher the rate of loading is, the bigger the volume
change of the soil is. This was explained with the higher pore water pressure development or the
sudden matric suction reduction during the higher loading rates (Fredlund, 1979). Contrary to
the findings of Huat et al. (2006), Rojas et al. (2008a) found that the higher the rate of loading
59
σy - ua
u a-
u w
Figure 4.12: Test stress path (matric suction change is exaggerated)
is, the smaller the volume change of the pyroclastic silty sand is. This was explained by the creep
phenomena in soil, as time passed the creep also increased, resulting in bigger volume changes of
the soil specimens. Therefore, the effect of a time dependent loading rate on the volume change
characteristics of soils seems to be not only dependent on loading rate but also soil type and the
experimental techniques used.
In this study, suction equilibration loading rate was chosen so that the matric suction change
could be monitored in each loading step. The duration of the loading time was changed according
to the time required at each step for stabilisation of the matric suction and volume change. There
are similar loading rate approaches presented in the literature (Tang et al., 2008). The loading
steps applied to the soil specimens during the tests are shown in Tables 4.1 and 4.2. The loading
increment was initially chosen to be 80 kPa for S2B and 110 kPa for S5B, differing due to the
change in sample diameters (the lower rates were harder to maintain with the equipment). Later
in the test it was observed that the matric suction and volume change of the sample were low
and therefore the loading increment was increased.
Table 4.1: Loading steps for S2B
Vertical stress Number of Increment
(kPa) increments (kPa)
0 - 640 8 80
640 - 1280 4 160
1280 - 2250 3 320
60 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
Table 4.2: Loading steps for S5B
Vertical stress Number of Increment
(kPa) increments (kPa)
0 - 880 8 110
880 - 1760 4 220
1760 - 3080 3 440
Equipment preparation and test setup
Before placing the saturated ceramic disk into the groove on the base cell, valve C was turned
to the direction of the water volume change tubes and valves A and B were left open (Figure
4.1) . The groove was filled with deaired water and the ceramic disk was then placed and gently
pushed into the groove. This made some amount of water under the ceramic disk (replacing
the volume of the ceramic disk) to move into the water volume tubes. Water was then flushed
with the help of an air pump from the opening L and R until the trapped air bubbles under the
ceramic disk were completely removed. Subsequently, valve C was turned to the direction of the
pressure transducer and the release valve on the block was opened for water to flow. The release
valve was then closed ensuring that no air bubbles were left in the plastic tube that connects
the base of the pressure cell and the transducer. A back pressure of about 20 kPa was applied
to the transducer to avoid air cavitation due to the long time of the test and valve C was turned
again to the direction of the water volume tube. Ensuring that air bubbles were removed from
the entire setup a thin layer of kaolin paste was spreaded on the ceramic disk and the steel
ring with the soil sample was placed onto the kaolin paste. Kaolin paste was placed to ensure
a good connection between the ceramic disk and the granular soil sample. Kaolin paste was
used for similar purpose (to ensure good conductivity between the soil and the ceramic disk)
in tensiometers by Tarantino et al. (2000) and Boso et al. (2004), and in thermal conductivity
suction sensors by Zhan et al. (2007). The pressure cell was quickly mounted to start the test
in order to avoid any water intake of the specimen from the ceramic disk. The valves A and
B were then closed, the valve C was turned to the direction of the pressure transducer and the
experiment was ready for initial matric suction measurements using null-type axis translation
technique.
61
Measurements during the test
An air pressure higher than the back pressure was applied inside the cell from the top of the
specimen and the water pressure below the ceramic disk was monitored by the help of the pressure
transducer. The applied air pressure was gradually changed after equilibrium was reached (i.e.,
there was no changes in water pressure below the ceramic disk). Final water pressure values are
plotted against the applied air pressures. The function of this curve needed to be a 45◦line when
the applied air pressure was greater than the initial matric suction of the specimen(taking into
account the Hilf’s approach for undrained systems). The initial matric suction of the specimen
was determined from the intersection of the curve with the abscissa of the graph (Hilf, 1956).
After initial matric suction was measured, the air pressure inside the cell was left as it was
and a vertical load was applied by the loading cell. The soil samples were statically compacted
using the loading steps shown on Tables 4.1 and 4.2. The duration of the loading time was
changed according to the time required at each step for stabilisation of the matric suction and
volume change. At the end of the last loading increment, the air pressure within the pressure
cell was reduced to zero, the valve C was turned to the direction of water volume tube and the
valves A and B were opened to allow saturation of the specimen under constant applied vertical
stress. This was conducted in order to see the volume change of the samples when the matric
suction was reduced (collapse).
Dismantling the compaction device
At the end of the saturation procedure the valves A and B were closed. The sample was quickly
taken out from the pressure cell. The samples were left in the oven overnight to find the final
water content of the samples.
4.3.3 Test difficulties and problems
Some difficulties and problems were experienced during the experimental procedure. These
problems are summarised below:
1. Matric suction readings for samples with different water contents, in the initial experiments,
were not going beyond 1 to 2 kPa, which probably was the matric suction of the ceramic
disk. This problem was attributed to the poor contact between the sample and the ceramic
disk due to the granular nature of the soil used in the experiments. Therefore, a thin layer
62 CHAPTER 4. TESTING EQUIPMENT AND EXPERIMENTAL PROCEDURE
of kaolin paste was spreaded on the ceramic disk and the soil sample was placed on it to
ensure a better connection. This resolution is also suggested by other researchers (e.g.,
Tarantino et al. (2000); Boso et al. (2004); Zhan et al. (2007)).
2. Some tests, particularly the ones with lower water contents or higher suctions, failed to be
successfully completed. This problem was thought to rise due to the cavitation of the water
in the tube which connects the pressure transducer and the bottom plate (tube between
pressure transducer and valve C). Thus, back pressure was applied to the water within this
tube. Still occasionally using this method was not successful and some tests needed to be
ceased; however, it was found more efficient.
3. During the initial matric suction measurement, after the application of air pressure to the
sample, the water pressure or suction equilibration took longer time than it was expected
(see Section 5.3). This lead to very long duration of the tests (in lower water contents
reaching up to 3 months). However, in literature, the equilibration time for the initial
matric suction measurements using null method was reported to be less than a day for fine
grained and residual soils (Vanapalli et al., 1999; Fredlund and Vanapalli, 2002; Rahardjo
and Leong, 2006).
4. Due to the long duration of the tests the device required to be flushed regularly in order
to stop the accumulation of the defused air beneath the ceramic disk. However, after this
procedure an additional time was required for the pressure transducer to equilibrate, which
extended the testing time further. This problem was raised more in S5B samples and in
the samples with a relatively lower water content (see Chapter 5 ).
5. It was very hard to maintain initially identical sample size for granular soils used in the
tests. Since the deformation of the granular soils is already very small, it was important to
know precisely the initial sizes of the samples. This lead to some difficulties in the inter-
pretation of the trial test. Therefore, extra care needed to be taken during the preparation
of the samples.
4.4 Concluding remarks
In this chapter the equipments used in the experimental part of the current study and the
experimental procedure were presented. SWC-150 Fredlund Device that is originally designed to
63
control matric suction was modified to measure matric suction. A pressure transducer was added
to the bottom plate of the pressure cell to measure the pore water pressure under the ceramic disk.
This modification was done in order to measure initial matric suctions and the matric suction
changes during the constant water static compaction tests. The equipment measures matric
suctions based on null-type axis translation technique. In addition, a LVDT was mounted to the
device to monitor volume changes of the sample during static compaction. The equipment was
checked for any possible leakage and it was found out that it can operate successfully. However,
it was decided to flush the equipment regularly due to the air diffusion within the water.
An experimental procedure was developed to conduct constant water content (undrained
water) static compaction tests. Soil specimens were prepared using moist tamping method. The
stress path followed was also described. Each loading increment was applied after the matric
suction from previous loading increment reached equilibrium. Loading increments were also
increased gradually, the sample was initially loaded slow and the rate was increased later on due
to the reduction in the total volume change.
The difficulties experienced during the experimental measurements were also summarised in
the last section. Since granular soil was used in experimentation, a thin layer of kaolin paste was
spreaded on the ceramic disk to maintain good contact between the disk and the soil sample.
Furthermore, after experimentation, the application of back pressure to the pressure transducer
was found to be beneficial to prevent water cavitation due to the long duration of the tests.
Chapter 5
Experimental results
5.1 Introduction
Static undrained (water) compaction tests (constant water content) were conducted on two
different granular soils, sand with 2% (S2B) and 5% (S5B) bentonite, with varying water contents
in order to cover a wide range in the compaction curve and thus to observe the effect of suction on
the compaction behaviour. Soil samples were compacted with vertical stresses ranging between
80 kPa and 2250 kPa for S2B and 110 kPa and 3100 kPa for S5B. Initial matric suctions of
the specimens were measured using the null type axis translation technique. Matric suction
variation was also monitored during the compaction process and its influence on the overall
volume change behaviour was observed. The physical properties of tested materials and the
experimental procedure were described in detail in Chapters 3 and 4.
This chapter presents the experimental results of the static compaction tests conducted in
the current study. Firstly, the initial matric suction measurements are summarised and this is
followed by the evolution of the matric suction during the tests. Finally, the volume change
behaviour of tested soils were discussed in detail. The results from this chapter will be used in
Chapter 7 where experimental results were compared with the modelling predictions.
5.2 Initial suction measurement
The initial matric suction of soil specimens was measured using the null type axis translation
technique (Hilf, 1956) before the compaction commenced. Matric suction measurements and
65
66 CHAPTER 5. EXPERIMENTAL RESULTS
Table 5.1: Water content of S2B
Test Water content (%) Dry densitybefore after Mg/m3
2b6.9wc 5.6 6.9 1.432b8.4wc 7.6 8.4 1.462b8.6wc 7.7 8.6 1.462b10.2wc 6.0 10.2 1.442b10.7wc 9.8 10.7 1.46
compaction tests were conducted within the same device (Figure 4.2), which allowed continuous
matric suction measurements during the tests.
The water contents of tested samples and their initial dry densities (γd) are given in Tables
5.1 and 5.2. Tests are named according to their bentonite and water contents. For both soils,
five compaction tests with different water contents were successfully performed and their results
are presented here. More tests were conducted; however, as it will be explained later in this
chapter, there were difficulties in finalising the tests and thus, these tests were disregarded. The
water content of the samples were measured before and at the end of the each test. It can be
seen from Tables 5.1 and 5.2 that there was a slight increase in the water content of the samples.
The increase in water content is unlikely to happen during the compaction process and/or after
the application of air pressure on the samples, because the applied air pressure to the samples
was higher than the matric suction of the samples. Eliminating the time spend during loading,
it seems that the water content increased either during the preparation stage of the test (before
the application of pore air pressure (ua)) or during the dismantling of the equipment. The time
spent during equipment preparation was much longer than the time spent when dismantling
the equipment. It took a considerable time, ranging from five to twenty minutes, to remove air
bubbles from the system after the placement of the sample onto the ceramic disk. During this
time, some water may have been imbibed from the ceramic disk. Thus, the final water content
of the samples was considered to be the water content during the compaction tests. In addition
to that, these values are in agreement with matric suction values measured using tensiometer as
shown in Figures 5.3 and 5.4. Initial dry density (γd) of the samples were chosen to be about
1.43Mg/m3 at first. However, it was hard to maintain that low dry density with increasing water
content because samples tend to collapse to a higher dry density (γd). Therefore, the initial dry
density for samples with 5% bentonite content was changed to 1.58Mg/m3 (see Section 4.3.1 for
more details).
Initial matric suction measurements for both soil types (S2B and S5B) and different water
67
Table 5.2: Water content of S5B
Test Water content (%) Dry densitybefore after Mg/m3
5b6.5wc 6.2 6.5 1.595b7.5wc 7.0 7.5 1.585b10.3wc 9.7 10.3 1.585b10.9wc 10.3 10.9 1.585b12.7wc 11.8 12.6 1.58
0 20 40
0
20
40
0 20 40
0
20
40
0 20 40
0
20
40
0 20 40
0
20
40
0 20 40
0
20
40
Pore air pressure, ua (kPa)
Por
e w
ater
pre
ssur
e, u
w (
kPa)
y = 1.0474x − 7.7773 y = 1.0207x − 5.198
y = 1.0007x − 4.8872
y = 1.0452x − 4.131y = 1.025x − 5.6921
wc = 6.9% wc = 8.4%
wc = 10.7%
wc = 8.6% wc = 10.2%
Figure 5.1: Initial matric suction measurements for S2B
68 CHAPTER 5. EXPERIMENTAL RESULTS
content are shown in Figures 5.1 and 5.2. The applied air pressures (ua) and the final pore water
pressures (uw), values after equilibration, are plotted against each other. Taking into account
Hilf’s approach (Hilf, 1956) for undrained loading condition, after applied air pressure was greater
than the initial matric suction of the specimens, the increase in pore air pressure (∆ua) would be
equal to the increase in pore water pressure (∆uw). This leads to a linear relationship between
pore air (ua) and pore water pressure (uw), with a slope equal to unity. Thus, the initial pore
water pressure (uw) of the specimen would be the intersection of the line with the ordinate of the
graph. Equations of the relationship between pore air (ua) and pore water pressure (uw), which
have the information of the initial matric suction, are also shown on the graphs for every water
content. As it was expected, the matric suction reduced with increasing water content, except
for the slight inconsistency between the samples 2b8.4wc and 2b8.6wc, and between 2b10.2wc
and 2b10.7wc. The difference, however, is very small. In addition, there is a discrepancy in the
water content of the sample 2b10.2wc, which showed higher variation than rest of the tested
samples. Therefore, this anomaly was ignored and accepted to result due to experimental errors.
The water retention curves obtained by different methods for both soil types are presented
in Figures 5.3 and 5.4. The matric suction measurements using the null type axis translation
technique are also plotted on the graphs. The null type measurements are close to the matric
suction measured with tensiometer. The matric suction measurement conducted using the axis
translation technique differ from the other two methods. While using axis translation technique,
the drying curve was measured. On the contrary, the measurements recorded by tensiometer
and null method lie on the wetting path of the water retention curve. In addition to that,
the soil specimens were prepared at an optimum water content for the measurements of drying
water retention curves. On the other hand, the soil specimens were prepared with different
water contents for the suction measurements using the null-type axis translation technique and
tensiometers. Specimens prepared with different water contents have different soil structure
which influence the water retention behaviour of the soils (Vanapalli et al., 1999; Catana, 2006).
The specimens compacted with an initial water content on the dry of optimum conditions have
relatively large pore spaces compared to the specimens compacted on the optimum or wet of
optimum conditions. Therefore, the specimens compacted with water contents on the optimum
or wet of optimum have higher storage capacity (Cui and Delage, 1996; Vanapalli et al., 2004).
The effect of structure on water retention behaviour, however, is not as big for coarse grained
soils compared to fine soils due to its relatively homogenous structure. Thus, the discrepancy in
69
60 80 100 1200
40
80 y = 0.9753x − 49.63
40 60 80 1000
50
100 y = 1.156x − 34.57
40 60 80 10020
60
100 y = 0.9587x − 15.34
20 40 60 8020
40
60
80 y = 0.999x − 10.88
40 60 80 10020
60
100
Pore air pressure, ua (kPa)
Por
e w
ater
pre
ssur
e, u
w (
kPa)
y = 0.9539x − 7.51
wc = 6.5% wc = 7.5%
wc = 10.3% wc = 10.9%
wc = 12.6%
Figure 5.2: Initial matric suction measurements for S5B
70 CHAPTER 5. EXPERIMENTAL RESULTS
suction measurements between axis translation technique and the other two methods resulted
mainly due to the measurement of wetting and drying paths in the water retention curve; and
the differences within the soil fabric of the samples prepared with different initial water contents.
It should also be noted that, the samples measured with tensiometer had higher dry densities
(γd = 1.58Mg/m3) than the samples measured by the null method (γd = 1.43 to 1.46 Mg/m3)
for S2B. However, it is not expected to have noticeable suction differences due to the variation
in density for granular soils which already have low suctions. Considering these arguments, the
results obtained by tensiometer and null method are in good agreement. Moreover, S2B has a
bigger hysteresis compared to S5B (Figures 5.3 and 5.4). S2B seems to be more sensitive to the
structure compared to S5B.
5.3 Evolution of suction during the test
Figures 5.5 and 5.6 present the evolution of matric suction during the tests (initial matric suction
measurement and loading). The applied air pressures (ua) and measured pore water pressures
(uw) are also shown on the figures. Matric suction values are calculated by subtracting the pore
water pressure (uw) from the applied air pressure (ua). In the initial part of the experiments,
the pore air pressures (ua) were increased to greater values than the initial matric suctions of
the samples (the initial matric suction of the specimens was approximated from the tensiometer
measurements conducted before). After exceeding the initial matric suction values, a minimum
of three air pressure increments (∆ua) were applied in order to obtain three points on pore
air pressure (ua) and pore water pressure (uw) graphs (see Figures 5.1 and 5.2). Following the
calculation of the initial matric suction values, the air pressure was kept constant and the loading
of the specimens was commenced.
During the initial matric suction measurement part of the tests, it can be seen that the
increase of air pressure (ua) resulted in a sudden increase of pore water pressure (uw). Initially,
the increase in the pore water pressure (∆uw) was slightly lower than the increase in the air
pressure (∆ua), thus resulting in a sudden matric suction increase. This is due to the time
taken of the pore air pressure and therefore, pore water pressure (uw) to evenly distribute
within the samples. Thereafter the pore water pressure gradually (uw) increased until it reached
equilibration.
The equilibration time for the initial part of the tests (matric suction measurements) was
71
[a]10
−110
010
110
20
10
20
30
40
50
60
70
80
90
100
Matric suction, s (kPa)
Deg
ree
of s
atur
atio
n, S
r (%
)
Axis translationTensiometerNull method
[b]10
−110
010
110
20
5
10
15
20
25
Gra
vim
etric
wat
er c
onte
nt, w
c (%
)
Matric suction, s (kPa)
Axis translationTensiometerNull method
Figure 5.3: Comparison of initial matric suction measurements using different techniques forS2B, [a]Variation of degree of saturation with matric suction, [b]Variation of water content withmatric suction
72 CHAPTER 5. EXPERIMENTAL RESULTS
[a]10
010
110
210
20
30
40
50
60
70
80
90
100
Matric suction, s (kPa)
Deg
ree
of s
atur
atio
n, S
r (%
)
Axis translationTensiometerNull method
[b]10
010
110
20
5
10
15
20
25
30
Matric suction, s (kPa)
Gra
vim
etric
wat
er c
onte
nt, w
c (%
)
Axis translationTensiometerNull method
Figure 5.4: Comparison of initial matric suction measurements using different techniques forS5B, [a]Variation of degree of saturation with matric suction, [b]Variation of water content withmatric suction
73
0 100 200−50
0
50
100
0 50 100 150−50
0
50
100
0 20 40 60−50
0
50
100
0 100 200−50
0
50
100
0 20 40 60−50
0
50
100
Time, t (hours)
Por
e pr
essu
re, u
a, uw a
nd m
atric
suc
tion,
s (
kPa)
ua
uw
matric suction
wc = 6.9%
wc = 10.7%
wc = 10.2%wc = 8.6%
wc = 8.4%
Figure 5.5: Evolution of matric suction with time for S2B
very long for the samples with low water content or low degree of saturation. An example for this
is shown for S5B with 2.9% water content in Figure 5.7 (also see Figure 5.5 for wc = 6.5%). In
some cases experiments were ceased due to the accumulation of dissolved air under the ceramic
disk. The device was regularly flushed, however, it took a long time after flushing for the system
to re-equilibrate in low water content tests, which further extended the testing period. This
can be seen in Figure 5.7 where after 350 hours the device was flushed. There was a sudden
reduction in pore water pressure with flushing and it took about five days for the system to
reach the previous pore water pressure (uw). Therefore, many tests with low water contents
were stopped and due to the time taken tests were not repeated.
In the literature, the equilibration time for the initial matric suction measurements using null
method was reported to be less than a day for fine grained and residual soils (Vanapalli et al.,
1999; Fredlund and Vanapalli, 2002; Rahardjo and Leong, 2006). However, during this project,
the equilibration time for the samples with low water content was found to be much longer. In
74 CHAPTER 5. EXPERIMENTAL RESULTS
0 500 10000
50
100
150
0 100 200−100
0
100
200
0 200 400−50
0
50
100
0 50 100−50
0
50
100
0 200 400−50
0
50
100
Time, t (hours)
Por
e pr
essu
re, u
a, uw a
nd m
atric
suc
tion,
s (
kPa)
ua
uw
matric suction
wc = 6.5%
wc = 7.5%
wc = 10.3%
wc = 12.6%
wc = 10.9%
Figure 5.6: Evolution of matric suction with time for S5B
the example shown here the equilibration time was about two and half months (Figure 5.7). This
might be due to the problems in applicability of the axis translation technique to the samples
with low degrees of saturation (Sr = 3− 10%). Most of the soils tested in literature were clayey
soils which had high suction values and it is well known that with this method matric suctions
up to 1500 kPa can be measured. Therefore, clayey soils with low water contents, which would
have matric suction values beyond 1500 kPa, were not measured before. This is the first time
in literature where suctions of granular samples with low water contents were tested using the
null method. The long equilibration time can be attributed to the low water content which in
some cases would result in discontinuous water phase. Most likely vapour equilibration took
place within the sample which lasted for a very long time. Another question that might rise
out of this study is whether the axis translation technique can be used to measure correctly
the matric suction of soils that have discontinuous water phase. Even though, the difficulties in
the applicability of the axis translation technique to the soils with occluded air phase (i.e., with
75
Table 5.3: Matric suction change with loading for S2B
Test ∆(ua − uw)(kPa)
2b6.9wc 1.12b8.4wc 0.82b8.6wc 0.62b10.2wc 02b10.7wc 0.2
high degrees of saturation) was extensively studied (Olson and Langfelder, 1965; Bocking and
Fredlund, 1980; Romero, 2001), there is not any previous work conducted on the applicability
of this technique to soils with a discontinuous water phase. However, in this study, it is found
that the matric suction values measured using null method are close to the values measured
with tensiometer (Figures 5.3 and 5.4). This is a promising result on the applicability of the
axis translation technique to dry granular soils. However, the method seems to be very time
consuming and therefore, not practical. Another reason for long equilibration time could be the
low permeability of the granular soils with low water content. Thus, further studies should be
undertaken before coming to a general conclusion.
After the initial matric suction measurements, samples were loaded incrementally as it was
described in Section 4.3.2. There was an instantaneous suction decrease when samples were
loaded in each loading step (Figures 5.5 and 5.6). Similar behaviour of sudden matric suction
decrease with loading was observed with a tensiometer by Delage et al. (2007). This resulted
due to a sudden increase in pore air pressure (ua) and therefore, pore water pressure (uw) with
loading. The increase in pore water pressure (uw) due to loading reduced at higher stresses due
to relative reduction in deformation. The excess pore air pressure (ua) dissipated in a short time
after loading, because air phase was drained. However, suction equilibration took a longer time
due to rearrangement of water. Water rearrangement occurred because of soil deformation which
resulted in a different pore size distribution within the sample. Equilibration time for suction or
pore water pressure (uw) was longer for samples with 5% bentonite content than samples having
2% bentonite content. This was due to the lower permeability of the samples with 5% bentonite
content. In general, the equilibration after loading took a longer time for samples that had lower
water contents in both types of soils. As it was discussed previously in this section this might
be due to the discontinuity of the water phase within the samples.
Tables 5.3 and 5.4 show the matric suction change with loading for samples with different
76 CHAPTER 5. EXPERIMENTAL RESULTS
0 500 1000 1500 20000
50
100
150
200
250
300
Time, t (hours)
Por
e pr
essu
re, u
a, uw a
nd m
atric
suc
tion,
s (
kPa)
ua
uw
matric suction
wc = 2.9%
Figure 5.7: Long equilibration time for S5B with 2.9% water content
Table 5.4: Matric suction change with loading for S5B
Test ∆(ua − uw)(kPa)
5b6.5wc 105b7.5wc 125b10.3wc 65b10.9wc 55b12.6wc 3.9
77
water contents. There was a matric suction reduction for both soil types at all tested water
contents. Most of the matric suction reduction took place at the first loading step. There was
no or very little matric suction reduction during the following loading steps. Matric suction
changes with loading gradually reduced with increasing water content for both soils. Similar
behaviour was observed by Tarantino and De Col (2008) for kaolin and by Olson and Langfelder
(1965) for Champaign till and Fayalite clay. At first sight the results seems to be unexpected,
because there is a common belief that suction reduces with loading on the wet side of the
optimum water content and stays almost constant in the dry side of the optimum water content.
Examining the previous studies carried on by Li (1995); Romero (1999); Montanez (2002), it is
expected an increasing suction change with increasing water content which is opposite to the
results obtained from this study. It should be noted, however, that post compaction suctions
were measured in those studies. When samples were unloaded, it resulted in an elastic volume
increase of the samples and samples moved from a loading wetting curve to an unloading drying
curve. This phenomena is well explained in Tarantino and Tombolato (2005) and Tarantino and
De Col (2008). Therefore, the past studies, where post compaction suctions were measured, may
not be directly comparable to suctions measured during the compaction process.
5.4 Volume change characteristics of soils
The change in specific volume (ν) with applied vertical stress (σy) is presented in Figure 5.8
for S2B and in Figure 5.9 for S5B. The results are also plotted in semi logarithmic graphs and
the normal compression lines (NCLs) for saturated soil samples are included in the figures for
comparison. It should be noted that the compression curves are for constant water content tests
and are not for constant suction tests except for the results of saturated soils, which are specified
on the figures. However, since the suction reduction with loading is low they can be regarded as
constant suction tests (see Section 5.3). Even though, the deformation of the samples were quite
low due to high sand proportion within the samples, a clear trend of deformation increase with
increase in water content can be noticed. In Figure 5.8, the sample with 10.2% water content
experienced an unexpected deformation. Moreover, the water content change during this test
was recorded to be much higher than the other tests (Table 5.1). Therefore, this experiment is
disregarded in the interpretation of the overall results and it will not be presented hereafter.
The slope of the normal compression lines (λ(s)) for unsaturated soil samples are higher
78 CHAPTER 5. EXPERIMENTAL RESULTS
[a]
0 500 1000 1500 2000 25001.3
1.4
1.5
1.6
1.7
1.8
1.9
Net vertical stress, σy − u
a (kPa)
Spe
cific
vol
ume,
ν
wc = 6.9%wc = 8.4%wc = 8.6%wc = 10.2%wc = 10.7%s = 0
[b]
10−2
100
102
104
1.6
1.7
1.8
1.9
Net vertical stress, σy − u
a (kPa)
Spe
cific
vol
ume,
ν
wc = 6.9%wc = 8.4%wc = 8.6%wc = 10.7%s = 0
Figure 5.8: Variation of specific volume during the tests for S2B
79
[a]
[b]
Figure 5.9: Variation of specific volume during the tests for S5B
80 CHAPTER 5. EXPERIMENTAL RESULTS
0 500 1000 1500 2000 25001.4
1.45
1.5
1.55
1.6
1.65
Net vertical stress, σy − u
a (kPa)
Dry
den
sity
, γd (
t/m3 )
wc = 6.9wc = 8.4wc = 8.6wc = 10.7
Figure 5.10: Variation of dry density during the tests for S2B
than the slope of the saturated soils in both type of soils. An increase in the slope of the
NCL with increase in matric suction was also observed by Estabragh et al. (2004),Cuisinier and
Masrouri (2005) and Monroy et al. (2008). Moreover, this can be also observed in Tarantino
and De Col (2008), where kaolin samples were initially prepared with different water contents
and later saturated before loading. The samples initially prepared with less water contents had
higher slopes of the NCLs (λ(s)) than the slope of the NCL for initially slurry soils (λ(0)). These
results contradict with the assumption taken in most of the constitutive models (Alonso et al.,
1990; Bolzon et al., 1996; Sheng et al., 2008), which state a decreasing slope with increasing
suction. Due to this discrepancy, these models may fail to successfully simulate the compaction
process. This will be examined in Chapter 7.
Figures 5.10 and 5.11 show the changes in dry density with increasing applied net vertical
stress (σy − ua). The dry density increased exponentially with loading. In addition, as it would
be expected from the compaction curve, the samples with higher water content have higher dry
density.
Changes in degree of saturation with loading are given in Figures 5.12 and 5.13. The variation
of degree of saturation with loading have a similar trend with dry density. It can be seen that
at the end of the loading none of the samples reached saturation. Saturation in these samples
could be only reached at high stresses if particles were compressible. It is important to notice
that the granular soils have very low degrees of saturation even at high applied stresses. In
81
Figure 5.11: Variation of dry density during the tests for S5B
addition, tested soils (S2B and S5B) reached saturation at low water contents in the wetting
path of their water retention curves (Figures 5.3 and 5.4). S2B had a degree of saturation about
50% at 1 kPa matric suction and S5B about 55% at 6 kPa matric suction. Probably, the pores
were already filled with water where the radius between the air and the water interface was at
its minimum angle. Therefore, addition of more water would only resulted in an increase of the
degrees of saturation but not in a reduction of the suctions. There is a common observation
that the water contents at saturation are lower in the wetting path than the drying path due
to the air being trapped within the sample during wetting. However, at such low degrees of
saturation, it is very unlikely for the air to be trapped. Moreover, these values are very close to
the degrees of saturation at their optimum water content. In the wetting path at zero suctions, as
the degrees of saturation increases when soils get finer, the optimum water content of these soils
also increases. However, further research needs to be conducted to find the direct correlation
between the optimum water content and the degree of saturation at zero suction for the wetting
path to assure this finding.
The variation of the coefficient of compressibility (ms1) with applied vertical net stress (σy −
ua) is shown in Figures 5.14 and 5.15. The compressibility coefficient (ms1) decreases with
increasing applied net vertical stress (ms1) up to 500 kPa for S2B and up to 1000 kPa for S5B,
and later remains almost constant for higher applied stresses. In general, the compressibility
82 CHAPTER 5. EXPERIMENTAL RESULTS
0 500 1000 1500 2000 250015
20
25
30
35
40
45
50
Net vertical stress, σy − u
a (kPa)
Deg
ree
of s
atur
atio
n, S
r (%
)
wc = 6.9wc = 8.4wc = 8.6wc = 10.7
Figure 5.12: Variation of degree of saturation during the tests for S2B
Figure 5.13: Variation of degree of saturation during the tests for S5B
coefficient (ms1) also decreases with decreasing water content at a particular stress. However,
since the deformation differences between samples are very small; this is not very obvious in the
current results.
The variation of the coefficient of compressibility (ms1) with degree of saturation is shown
in Figures 5.16 and 5.17. The compressibility coefficient (ms1) reduces with increasing degree of
83
0 500 1000 1500 2000 25000
1
2
3
4
5
6x 10
−4
Net vertical stress, σy − u
a (kPa)
Coe
ffici
ent o
f com
pres
sibi
lity,
m1s (
1/kP
a) wc = 6.9wc = 8.4wc = 8.6wc = 10.7
Figure 5.14: Variation of degree of the coefficient of compressibility due to net stress (ms1) for
S2B
0 1000 2000 3000 40000
1
2
3x 10
−4
Net vertical stress, σy − u
a (kPa)
Coe
ffici
ent o
f com
pres
sibi
lity,
m1s (
1/kP
a) wc = 6.5%wc = 7.5%wc = 10.3%wc = 10.9%wc = 12.6%
Figure 5.15: Variation of degree of the coefficient of compressibility due to net stress (ms1) for
S5B
saturation during the compaction process.
Finally, Figures 5.18 and 5.19 show the static and dynamic compaction curves for S2B and
S5B. Even though, the deformations of the samples were low, there is still a clear trend of
a density increase with an increase in water content for both soils. In addition, the static
84 CHAPTER 5. EXPERIMENTAL RESULTS
20 25 30 35 40 45 500
1
2
3
4
5
6x 10
−4
Degree of saturation, Sr (%)
Coe
ffici
ent o
f com
pres
sibi
lity,
m1s , (
1/kP
a) wc = 6.9wc = 8.4wc = 8.6wc = 10.7
Figure 5.16: Variation of the coefficient of compressibility due to net stress (ms1) with degree of
saturation for S2B
20 30 40 50 600
1
2
3x 10
−4
Degree of saturation, Sr (%)
Coe
ffici
ent o
f com
pres
sibi
lity,
m1s (
1/kP
a) wc = 6.5%wc = 7.5%wc = 10.3%wc = 10.9%wc = 12.6%
Figure 5.17: Variation of the coefficient of compressibility due to net stress (ms1) with degree of
saturation for S5B
compaction curves have a similar trend to dynamic compaction curves. The energy applied
with dynamic compaction for the Standard tests was 596kJ/m3 and for Modified tests was
2703kJ/m3. On the other hand, the stresses applied for static compaction curves were 560kPa
and 2250kPa for S2B and 560kPa and 2650kPa for S5B.
85
0 5 10 15 201.5
1.55
1.6
1.65
1.7
1.75
1.8
1.85
1.9
1.95
2
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
Standard (596 kJ)Modified (2703 kJ)560 kPa2250 kPa
Sr = 70%Sr = 50%
Figure 5.18: Compaction curves of S2B
0 5 10 15 201.6
1.65
1.7
1.75
1.8
1.85
1.9
1.95
2
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
Standard (596 kJ)Modified (2703 kJ)560 kPa2650 kPa Sr = 100%
Sr = 80%
Figure 5.19: Compaction curves of S5B
5.5 Concluding remarks
This chapter presented the experimental results of the constant water content static compaction
tests conducted during this study. Before compaction, the initial matric suction of the specimens
was measured using null type axis translation technique. Matric suction measurements were
conducted in the same device with compaction tests. This allowed continuous matric suction
86 CHAPTER 5. EXPERIMENTAL RESULTS
measurements during loading.
Initial matric suction measurements took a long time for the low water content samples.
This was attributed to the discontinuous water phase within the samples. There was a good
agreement between the matric suction measurements with null method and tensiometers. Post
loading equilibration time was also found to be very long for samples with low water content
compared to the time reported in the literature. Matric suction reduction was observed during
loading for all tested samples with varying water contents. The matric suction change decreased
with increasing water content and decreasing clay content.
The slopes of the normal compression lines for the constant water content tests were found
to be higher than the slopes of the saturated samples for both samples. This finding is different
from the assumption taken in most of the constitutive models which states a decrease in the
slopes of the normal compression line with increasing suctions. This might result in deficiencies
to model the compaction behaviour with existing soil constitutive models and will be further
examined in Chapter 7.
Full saturation was never reached at the end of the compaction tests. On the contrary, the
degrees of saturation attained at the end of the compaction process were very low and were
close to the values of their water retention curves at zero suction in the wetting paths. This
strengthens the idea of direct correlation between the degree of saturation at optimum water
content and the maximum degree of saturation in the wetting path of the water retention curve.
Further investigation needs to be undertaken to validate the correlation observed in this study.
The coefficient of compressibility (ms1) of the soils decreased with both increasing vertical
net stress (σy−ua) and degree of saturation during the test. In addition, the compaction curves
obtained with static loading had similar shapes to the dynamic compaction curves.
In conclusion, the matric suction of granular soils were measured using null type axis transla-
tion technique for the first time in the literature. Even though, the experiments were very time
consuming, they gave insights to the possibilities in matric suction measurement techniques for
granular soils with low water content.
Chapter 6
Modelling the compaction process
6.1 Introduction
Compacted soils undergo a compaction process which influences the post compaction behaviour.
However, information on the evolution of the compaction process is limited particularly from a
quantitative sense. Thus, examination of the compaction process is a crucial topic in order to
extend the current knowledge on the behaviour of compacted soils.
One way of accomplishing this aim is to model the compaction process. Modelling the
compaction process will help to understand the evolution of the stress and phase variables during
compaction and the fundamental reasons for the inverted parabolic shape of the compaction
curve.
6.2 Formulation of the compaction process
The main parameters required to be predicted to model the compaction process are the volume
change, the pore pressure development, and the compressibility coefficients. Theoretical concepts
used to predict these parameters were presented in Section 2.4 and are briefly summarised here
for clarity.
6.2.1 Volume change calculation
The constitutive relationship proposed by Fredlund & Morgenstern (1976) was used for the
volume change and thus, dry density calculations (see Equation 2.2). The volume change consti-
87
88 CHAPTER 6. MODELLING THE COMPACTION PROCESS
Table 6.1: Comparison of Hilf and Hasan & Fredlund’s method for pore pressure generations
Hilf (1948) Hasan & Fredlund (1980)compressibility free air 4 4
dissolved air 4 4water x 4
soil x xmatric suction ∆ua = ∆uw 4 x
∆(ua − uw) = 0 4 x
tutive relationship is defined in terms of two independent stress variables (σy−ua and ua−uw).
As applicable to one dimensional loading, the constitutive relationship can be expressed as fol-
lows:
εv =∆VvV0
= ms1∆(σy − ua) +ms
2∆(ua − uw) (2.2)
where εv is the volumetric strain, ∆Vv is the overall volume change, V0 is the initial volume,
∆(σy − ua) is the change in the vertical net stress, ∆(ua − uw) is the change in matric suction,
ms1 is the compressibility of a soil element with respect to vertical net stress (σy − ua) and ms
2
is the compressibility of a soil element due to matric suction (ua − uw).
The volume changes (∆Vv) were incrementally calculated by multiplying the right hand side
of the Equation 2.2 with the initial volume (V0). The corresponding dry density (γd) was then
calculated once the final total volume of the soil element is known.
Examination of the Equation 2.2 clearly indicates that in order to calculate the volume change
(∆Vv) with the applied vertical stress (σy), the pore pressures (ua, uw) and the compressibility
coefficients (ms1, ms
2) need to be predicted at each stress state.
6.2.2 Pore pressure development
One of the primary aspects associated to model the compaction process is the pore pressure
development. Two different approaches were identified in literature, namely the Hilf’s (1948)
and the Hasan & Fredlund’s (1980) approaches. The major differences between two methods
are summarised in Table 6.1 and their mathematical formulations were presented in Chapter 2.
In this section, both approaches are analysed for undrained loading conditions in detail. The
assumptions taken for the undrained, semi-drained and drained loading calculations are also
explained.
For the analysis of Hilf’s approach, the compressibility coefficient for saturated soil sample
89
Table 6.2: Parameter values for the pore pressure generations for Hilf’s and Hasan & Fredlund’smethods
mv = ms1 ms
2 ma1 ma
2 Cw ua0 h Sr0 n0(kPa−1) (kPa−1) (kPa−1) (kPa−1) (kPa−1) (kPa)
Hilf 1.45E-04 - - - - 101.3 0.02 0.85 0.50Hasan & Fredlund 1.45E-04 1.02E-04 1.15E-04 0.00 4.00E-07 101.30 0.02 0.85 0.50
(mv) was replaced with the unsaturated compressibility coefficient (ms1) (see Equation 2.3). This
modification was adopted considering that the soil being compressed is unsaturated and thus,
will better represent the real situation. If compressibility coefficient for saturated soil (mv)
is used higher deformations would be expected and this will lead to higher pore air pressure
generation within the soil sample.
∆ua =
1
1 +(1 − Sr0 + hSr0)n0
(ua0 + ∆ua)ms1
∆σy (2.3)
where ∆ua is the change in the pore air pressure, Sr0 is the initial degree of saturation, h is the
solubility coefficient of air in water, n0 is the initial porosity, ua0 is the initial absolute pore air
pressure, ms1 is the coefficient of volume change in unsaturated soil, and ∆σy is the change in
applied vertical stress.
Undrained loading
During undrained loading, both the air and water phases were considered to be undrained.
An example of pore pressure development during compression, using Hilf’s approach is shown
in Figure 6.1. The soil parameters used for the predictions are given in Table 6.2 and were
obtained from Fredlund and Rahardjo (1993). It was assumed by Hilf (1948) that the changes
in the pore water pressures (∆uw) will be equal to the changes in the pore air pressures (∆ua).
Thus, matric suction changes (∆(ua − uw)) will remain constant. For that reason, when Hilf’s
approach is adopted for pore pressure calculations, the second part or the matric suction part
of the volume change equation (see Equation 2.2) can be neglected.
During the pore water pressure (uw) calculations, the initial matric suction (s = ua − uw)
values were obtained from the calculations of Hasan & Fredlund’s method. They assume that
the pore water pressure (uw) will be equal to the atmospheric pressure when the soil sample
reaches saturation (Sr = 100%). Therefore, even the initial matric suction (s0 = ua0 −uw0) was
unknown it could be back calculated.
90 CHAPTER 6. MODELLING THE COMPACTION PROCESS
0 500 1000 1500 2000−200
0
200
400
600
800
1000
1200
Vertical stress, σy (kPa)
Por
e ai
r an
d w
ater
pre
ssur
es u
a, uw (
kPa)
ua
uw
Figure 6.1: Pore air and pore water pressure generation using Hilf’s approach
An example of pore pressure development using Hasan & Fredlund’s approach is shown in
Figure 6.2 and the parameter values used to generate the curves are given in Table 6.2. Different
than the Hilf’s approach, Hasan and Fredlund (1980) assumed that the suction becomes zero at
full saturation. Thus, the suction is progressively reducing with loading.
The pore air (ua) predictions for both approaches are presented in Figure 6.3. Constant
soil coefficients are used for both methods. Although, the results of the two methods are quite
close, it can be seen from the figure that Hilf’s method generates higher pore air pressures (ua)
than Hasan & Fredlund’s method for the same applied vertical stress. This difference could be
explained by two arguments:
1. Hasan & Fredlund (1980) take into consideration matric suction changes (∆(ua−uw)) with
applied vertical stress (σy), whereas Hilf’s method assumes a constant matric suction with
the increasing vertical stress. Taking matric suction changes (∆(ua − uw)) into account
(second part of the Equation 2.2) reduces the compressibility and therefore results in lower
density of the soil element. It should be noted that at the end of the compression the
two samples have different densities; Hilf’s approach has higher and Hasan & Fredlund’s
approach has lower density, thus they develop different pore air pressures at the same
applied vertical stress.
2. The compressibility of water is considered in Hasan & Fredlund’s approach and this results
91
0 500 1000 1500 2000−200
0
200
400
600
800
1000
1200
1400
1600
Vertical stress, σy (kPa)
Por
e ai
r an
d w
ater
pre
ssur
es u
a, uw (
kPa)
ua
uw
Sr = 1
Figure 6.2: Pore air and pore water pressure generation using Hasan & Fredlund’s approach
in a slightly different pore air pressure development. However, since the compressibility of
water is very low (Cw = 4.00E − 07kPa−1) the effect of this is only minor.
Semi-drained loading
In semi-drained loading, it was accepted that pore pressures (ua and uw) develop in every loading
step, and after the pore air pressure (ua) and therefore, the pore water pressure (uw) dissipate.
Therefore, each loading step started with the atmospheric air pressure (ua0 = 101.3kPa). This
results in different porosity (n) and degree of saturation (Sr) of the soil element than undrained
and drained loading types.
This type of loading was implemented in order to better capture the experimental method.
As it was explained in Chapters 4 and 5, air phase is drained during the experiments but it
requires some time for the buildup air pressure to dissipate after each loading increment. Thus,
it is expected that air pressure builds up in every load increment and later dissipates.
Drained loading
During drained loading, it was considered that the air is continuous throughout the soil element
and can drain freely, thus, keeping the pore air pressure (ua) constant and equal to atmospheric
air pressure. However, water was considered to be undrained. It is well known that in un-
92 CHAPTER 6. MODELLING THE COMPACTION PROCESS
0 500 1000 1500 20000
200
400
600
800
1000
1200
Vertical stress, σy (kPa)
Por
e ai
r pr
essu
re, u
a (kP
a)
Hilf (1948)Hasan & Fredlund (1980)
Figure 6.3: Comparison of Hasan & Fredlund’s and Hilf’s approach for pore air pressure gener-ation
saturated soils, due to the presence of suction, water cannot drain until it reaches the OWC
where air becomes discontinues and is trapped within water (Langfelder et al., 1968; Hilf, 1956).
Moreover, it could be argued that by a sudden increase in vertical stress (σy), there would be
a temporary increase in the pore water pressure (uw) due to the increase in pore air pressure
(ua) and this might result in water drainage (which might be the case in dynamic compaction).
However, this study concentrates on static compaction and the loading rate is considered to be
rather small.
In practice, this loading type could be the case in coarse grained soils where the pores are
large and the pore air pressure could quickly dissipate. In addition, the compaction behaviour
for drained loading is modelled in order to compare with undrained loading results and illustrate
the effect of pore pressure on the overall shape of the compaction curve.
6.2.3 Compressibility coefficients
Compressibility coefficients (ms1,m
s2) are vital for compaction behaviour modelling. In this
section the basic assumptions and approaches used for their predictions are explained.
93
Constant compressibility coefficients
Initially, the compressibility coefficients (ms1 and ms
2) were considered as constant parameters,
i.e., compressibility of a soil element does not change with increasing vertical stresses (σy) or
vertical net stresses (σy − ua) and changes in matric suction (∆(ua − uw)). Compaction curves
generated with this assumption are shown and discussed in Section 6.3.1.
Variable compressibility coefficients
An insight for the variation of the compressibility coefficients was gained from the study con-
ducted by Cho & Santamarina (2001) on the variation of the shear wave velocity of granite
powder for different degrees of saturation. They found that the shear wave velocity (Vs), hence
the shear modulus (G = ρVs) increases with decreasing degrees of saturation. This outcome is
confirming that material stiffness is increasing for decreasing saturations and is not a constant
parameter. Therefore, it can be stated that the compressibility of the material is increasing with
increasing degrees of saturation. In addition, Lloret et al. (2003) showed that the coefficient
of compressibility (ms1) decreases with both increasing suctions (s = ua − uw) and vertical net
stresses (σy − ua).
A functional form for the compressibility coefficients (ms1 and ms
2) could be approximated
from the constitutive models available in literature. Two of these models are considered in
this study. Barcelona Basic Model (BBM) proposed by Alonso et al. (1990) and the model,
which will be referred hereafter as SFG model, proposed by Sheng et al. (2008). The BBM
is the first introduced constitutive model for unsaturated soils. It is a separate stress state
variables approach model, which considers the vertical net stress (σy − ua) and matric suction
(s = ua−uw) as independent variables. Other separate stress state variables models in literature
(Wheeler and Sivakumar, 1995; Vaunat et al., 2000; Thu et al., 2007; Chiu and Ng, 2003) are
derived in a similar manner to the BBM and only minor changes were added to adapt the specific
modelling purposes. For that reason, the BBM was selected for detailed examination in the
current study. In contrast to the BBM, the SFG model is a combined stress variables approach
model, which combines the vertical net stress (σy − ua) and matric suction (s = ua − uw) in its
model formulations. There are other combined stress-suction approach models in literature (such
as Bolzon et al. (1996); Santagiuliana and Schrefler (2006); Gallipoli et al. (2003a); Tarantino and
De Col (2008)); however, those models use the BBM or models with similar approach (Wheeler
and Sivakumar, 1995) in their mechanical modelling formulations. Therefore, the SFG model is
94 CHAPTER 6. MODELLING THE COMPACTION PROCESS
the second selected approach for detailed examination in this study.
As applicable to one dimensional loading, volumetric strain (εv) can be written the following
form for the BBM:
εv =dν
ν=λ(s)
ν
d(σy − ua)
(σy − ua)+κsν
d(ua − uw)
(ua − uw) + uatm(2.12)
where ν is the specific volume of the soil element, λ(s) is the slope of the NCL at suctions defined
in semi logarithmic plot, d(σy −ua) is the change in vertical net stress, d(ua−uw) is the change
in suction and κs is the slope of the volume change vs. suction curve defined in semi logarithmic
plot.
In Equation 2.12, λ(s) is defined as:
λ(s) = λ(0)[(1 − r)e−βs + r] (2.13)
where λ(0) is the slope of the NCL for saturated soil defined in semi logarithmic plot, r is a
constant parameter related to the maximum stiffness of the soil, and β is a constant parameter
which controls the rate of increase of soil stiffness with matric suction.
From the BBM, the compressibility coefficients (ms1 and ms
2) can be derived as:
ms1 =
λ(s)
ν(σy − ua)(6.1)
ms2 =
κsν(σy − ua)
(6.2)
.
In addition, volumetric strain (εv) for one dimensional loading is suggested in the following
form in the SFG model:
εv =dν
ν= λvp
d(σy − ua)
(σy − ua) + s+ λvs
d(ua − uw)
(σy − ua) + s(2.16)
where ν is the specific volume of the soil element, λvp is the slope of the normal compression line
(NCL) for saturated soil when both axes are plot in logarithmic scale, d(σy − ua) is the change
in vertical net stress, d(ua − uw) is the change in matric suction, s is matric suction and λvs is
the slope of the volume change vs. matric suction curve when both axes are plot in logarithmic
scale.
95
This gives the compressibility coefficients (ms1 and ms
2) as:
ms1 =
λvp(σy − ua) + s
(6.3)
ms2 =
λvs(σy − ua) + s
(6.4)
.
Equations 2.12 and 2.16 are integrable for vertical net stress changes (σy−ua) under constant
matric suctions (s = ua−uw). Integrating Equations 2.12 and 2.16 for vertical net stress changes
(σy − ua) under a constant matric suction or for matric suction (s) changes under a constant
vertical net stress(σy − ua) lead to:
ν = N(s) − λ(s)ln(σy − ua)
(σy − ua)c(2.14)
lnν = lnN(s) − λvpln(σy − ua) + s
(σy − ua)0 + s(2.17)
where N(s) is the specific volume at an initial stress state ((σy − ua)0,s) and (σy − ua)c is a
reference stress in which changes of matric suction causes only elastic deformation.
In the BBM, N(s) is defined as (in Equation 2.14):
N(s) = N(0) − (λ(0) − λ(s))ln(σy − ua)c − κsln
(s+ uatmuatm
)(2.15)
where N(0) is the specific volume, at a reference stress state ((σy − ua)c), for saturated soil
defined in semi logarithmic plot.
In the SFG model, N(s), which is the specific volume at an initial stress state ((σy − ua)0,s)
for matric suctions smaller than the saturation suction (ssa), s < ssa is defined as:
lnN(s) = lnN0 − λvpln(σy − ua)0 + s
(σy − ua)0(2.18)
and; when N(s) is under net vertical stress (σy −ua)0 equal to unity and matric suctions bigger
than the saturation suction (ssa), s≥ssa, it is defined as
lnN(s) = lnN0 − λvpln(σy − ua)0 + ssa
(σy − ua)0− λvp
(1 − ssa + 1
s+ 1
)(2.19)
96 CHAPTER 6. MODELLING THE COMPACTION PROCESS
where N0 is the specific volume at initial stress state for saturated soil when both axes are plot
in logarithmic scale.
These equations will be used for the compaction curve generation. The variation of the
compressibility coefficients with degree of saturation and their effect on the overall compaction
curve are further discussed in Section 6.3.2 under the light of selected constitutive models.
6.3 Generation of the compaction curve
The compaction curves were generated by using the theories summarised in Chapter 2 and
Section 6.2. First, the pore pressures (ua, uw) were predicted and were used for the calculations
of the volumetric strain (εv = ∆V/V ). Subsequently, the volume change (∆V ) results were
used to find the corresponding dry densities. The computations were performed for a range of
moisture contents (w) which also define the values of the initial degrees of saturation (Sr0).
A summary of the approach that was followed and all the combinations used for the gener-
ation of the compaction curves are illustrated in Figure 6.4. The following combinations were
performed and each combination is described in detail in the following sections:
1. undrained modelling using Hilf’s method for pore pressure development and constant com-
pressibility coefficient (ms1)
2. undrained modelling using Hasan & Fredlund’s approach for pore pressure development
and constant compressibility coefficients (ms1 and ms
2)
3. undrained modelling using Hilf’s method for pore pressure development and variable com-
pressibility coefficient (ms1)
4. semi-drained modelling using Hilf’s method for pore pressure development and constant
compressibility coefficient (ms1)
5. semi-drained modelling using Hasan & Fredlund’s approach for pore pressure development
and constant compressibility coefficients (ms1 and ms
2)
6. semi-drained modelling using Hilf’s method for pore pressure development and variable
compressibility coefficient (ms1)
7. drained modelling using constant compressibility coefficient (ms1)
8. drained modelling using variable compressibility coefficient (ms1)
97
The compaction curves predictions are explained under two different classifications. First,
all the loading types which are developed using constant compressibility coefficients (ms1,m
s2)
are summarised and this is followed by the compaction curves generated using variable com-
pressibility coefficient (ms1). In this chapter, the shape of the compaction curve is qualitatively
examined. According to the model formulations and the assumptions in the both constitutive
models (the SFG model and the BBM) presented in Chapter 2 and Section 6.2.3, the shape of
the compaction curve will be similar and only the quantitative results will differ. For simplicity,
in this chapter, the compaction curves are only generated using the SFG model. In Chapter 7,
however, the performances of the both models are comprehensively examined.
6.3.1 Modelling of the compaction curve using constant compressibil-
ity coefficients
In this section, compaction curves were generated using constant compressibility coefficients
(ms1,m
s2) for undrained, semi-drained and drained loading conditions. During undrained and
semi-drained loading, the pore pressures (ua, uw) were predicted by the two aforementioned
methods (Hilf, 1948 and Hasan & Fredlund, 1980). For modelling purposes, the compressibility of
the soil element is accepted to be equal to the compressibility of water at the time it reaches 100%
saturation (Sr) (soil does not compress further after it reaches full saturation). A comparison of
the results of the three loading types as well as the use of constant compressibility coefficients
in the modelling of the compaction curve is discussed.
Undrained loading
The compaction curves generated assuming constant compressibility coefficients (ms1 and ms
2)
during undrained loading for different energy levels are presented in Figure 6.5. It should be
noted that the first method (Hilf, 1948) does not take into account the changes in suction and
assumes a constant suction during the compression of the soil specimen. Therefore, in the volume
change equation (Equation 2.2) the second component of the equation was neglected, whereas
in the Hasan & Fredlund’s method the second part of the volume change equation is taken into
account. Parameters used for the modelling of the compaction curves are presented in Table 6.3.
Figure 6.5 shows that both methods produced a reasonable shape on the wet side of the
compaction curve, but not on the dry side. In addition, the effect of applied vertical stress
(σy) or applied energy can be seen in the simulations by the both approaches. Moreover, matric
98 CHAPTER 6. MODELLING THE COMPACTION PROCESS
Compaction Curve
Undrained
Compaction
Drained
Compaction
Semi-drained
Compaction
BBM
Constant
Compressibility Coefficients
Variable
Compressibility Coefficients
Fredlund & Rahardjo’s
approach (1993)
Hilf’s approach
(1948)
Pore pressure development
SFG model
Volume change calculations
Figure 6.4: Approach followed for the compaction curve generation
99
[a]0 5 10 15 20 25
1.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
σy = 700 kPa
σy = 2000 kPa
Saturation curve
[b]0 5 10 15 20 25
1.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
σy = 700 kPa
σy = 2000 kPa
Saturation curve
[c]0 5 10 15 20 25
1.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
Hilf (1948)Hasan & Fredlund (1980)Saturation curve
Figure 6.5: Modelling of the compaction curves using constant compressibility coefficients forundrained loading [a] Hilf’s method [b] Hasan & Fredlund’s method [c] Comparison of Hilf’s andHasan & Fredlund’s methods at σy = 700 kPa
100 CHAPTER 6. MODELLING THE COMPACTION PROCESS
Table 6.3: Parameters used for the compaction curve modelling using constant compressibilitycoefficients
mv = ms1 ms
2 ma1 ma
2 Cw ua0 h Gs e0(kPa−1) (kPa−1) (kPa−1) (kPa−1) (kPa−1) (kPa)1.45E-04 1.02E-04 1.15E-04 0.00 4.00E-07 101.30 0.02 2.65 0.68
suction changes (∆(ua−uw)) that were included in Hasan & Fredlund (1980)’s method resulted in
some reduction in the dry density. This was due to the matric suction (s = ua − uw) reduction
in the specimen with compression (see Section 6.2.2). When matric suction (s = ua − uw)
reduced, the change in matric suction (∆(ua − uw)), which is the second part of the volume
change equation (Equation 2.2), became negative; and therefore resulted in lower total volume
changes (∆V ) which led to lower dry densities.
Semi-drained loading
The compaction curves generated during semi-drained loading type and using constant compress-
ibility coefficients (ms1 and ms
2) are illustrated in Figure 6.6. Figure 6.6(a) shows the results of
the compaction curves using the approach of Hilf (1948) for pore pressure (ua, uw) predictions.
The results obtained by using Hasan and Fredlund (1980)’s approach for pore pressure genera-
tion (ua, uw) were shown in Figure 6.6(b) for different energy levels (σy = 700 kPa and σy = 2000
kPa). The comparison of the compaction curves modelled by employing these two methods at
the applied vertical stress of 700 kPa are presented in Figure 6.6(c).
Similar results to undrained loading were obtained while compaction curves were generated
for semi-drained loading conditions. Again, a reasonable shape of the compaction curve was
attained on the wet side of the compaction curve, but not on the dry side.
Drained loading
In this section, the compaction curves modelled for drained loading are analysed. During drained
loading it was assumed that pore air pressure (ua) stayed constant and equal to atmospheric
pressure. Therefore, the pore air pressure (ua) predictions were not required and only the volume
change equation (Equation 2.2) was necessary to use. The use of the second part of the volume
change equation (Equation 2.2) is attainable only if matric suction change (∆(ua − uw)) with
compression can be predicted. Here, the volume change (∆V ) resulted due to the change in
matric suction (∆(ua − uw)) was neglected and the calculated volume change (∆V ) was only
101
[a]0 5 10 15 20 25
1.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
σy = 700 kPa
σy = 2000 kPa
Saturation curve
[b]0 5 10 15 20 25
1.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
σy = 700 kPa
σy = 2000 kPa
Saturation curve
[c]0 5 10 15 20 25
1.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
Hilf (1948)Hasan & Fredlund (1980)Saturation curve
Figure 6.6: Modelling of the compaction curves using constant compressibility coefficients forsemi-drained loading[a] Hilf’s method [b] Hasan & Fredlund’s method [c] Comparison of Hilf’sand Hasan & Fredlund’s methods at σy = 700 kPa
102 CHAPTER 6. MODELLING THE COMPACTION PROCESS
0 5 10 15 20 251.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
σy = 700 kPa
σy = 2000 kPa
Saturation curve
Figure 6.7: Modelling of the compaction curves using constant compressibility coefficients fordrained loading
due to the change in the vertical net stress (σy − ua).
The compaction curves generated with the assumption of using constant compressibility
coefficient (ms1) for drained loading are presented in Figure 6.7. In this figure, the compaction
curves were generated for applied vertical stresses of 700 kPa and 2000 kPa. The shape of the
modelled curves on the dry side were similar to the ones modelled for undrained and semi-
drained loading. However,it can be seen that due to the elimination of the pore air pressure (ua)
generation the compaction curves joined the saturation line on the wet side of the curve.
Discussion on using constant compressibility coefficients for the compaction mod-
elling
The comparison of the compaction curves generated using constant coefficients of compressibility
(ms1,m
s2) for undrained, semi-drained and drained loading types are demonstrated in Figure 6.8.
During the modelling of the drained loading type only the first part of the volume change
equation (Equation 2.2) was used. Therefore, for the impartial comparison of the three loading
types the results shown in the figure for undrained and semi-drained loadings are the curves
generated with Hilf’s method (Hilf’s method also uses only the first part of the volume change
equation). In this figure, compaction curves are shown for applied vertical stress (σy) at 700
kPa (the lower curve) and 2000 kPa (the upper curve).
103
0 5 10 15 20 251.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
UndrainedSemi−drainedDrainedSaturation curve
Figure 6.8: Comparison of the modelled compaction curves using constant compressibility coef-ficients for undrained, semi-drained and drained loading conditions (σy = 700 kPa lower curvesand σy = 2000 kPa upper curves)
Figure 6.8 shows that the modelling results of drained loading produced the highest densities
and the modelling results of undrained loading generated the lowest densities under the same
applied vertical stresses (σy). It was an expected result as higher pore air pressure (ua) would
make the soil element harder to compress. In addition, another difference could be noticed close
to saturation. The model results attained during drained loading coincided with the saturation
curve. This happened due to the assumption that the air phase is continuous and can drain freely
for all degrees of saturation (Sr). However, it is well known that after the soil specimen reaches
the OWC, air is discontinuous and is trapped into water and thus, cannot drain (Langfelder et al.
(1968)). However, during the modelling of the semi-drained and the undrained loading types
there was high pore air pressure (ua) generation which could be expected to act like trapped air.
Therefore, the results obtained for the wet side of the compaction curves during the semi-drained
and the undrained loading types better represents the real situation. The comparison between
the modelling results of the semi-drained and the undrained loading conditions will be examined
further in Chapter 7 when the models’ performance is evaluated with the experimental results.
It is clear from Figure 6.8 that using constant compressibility coefficients (mv = ms1 and ms
2),
a reasonable shape for the compaction curve can be attained on the wet side of the compaction
curve, but not on the dry side of the curve. Moreover, the matric suction changes (∆(ua − uw))
104 CHAPTER 6. MODELLING THE COMPACTION PROCESS
that were included in Hasan & Fredlund (1980)’s method resulted in some reduction in the dry
density (see Figure 6.5(c) and Figure 6.6(c)), but did not change the shape of the compaction
curve at lower water contents.
In order to examine the likely suction change during compaction, consideration was given to
the experimental results produced by Montanez (2002). Montanez (2002) provided the suction
contours for compacted sand-bentonite mixtures associated with domains defined by compaction
curves at various apparent energy levels. He found that the matric suction (ua − uw) only
decreases marginally with the density increase and in many cases can be considered constant.
Similar results for the suction change (∆(ua−uw)) can be found in literature (Li (1995); Delage
and Graham (1995). In addition, even if there is a small reduction in matric suction (ua − uw)
with compression, the ms2 applicable to the unloading process can be considered smaller than
that applicable to loading process. Therefore, Hilf’s analysis of assuming constant matric suction
during compaction appears to be close to the real situation and requires less soil parameters (no
need for parameters such as ma1 ,m
a2 etc.). Therefore, in the reminder of this thesis pore pressure
(ua and uw) predictions will be calculated using Hilf’s method.
6.3.2 Modelling of the compaction curve using variable compressibility
coefficients
In Section 6.3.1 the reasons for using only the first part of the volume change equation (Equation
2.2) and the discussion on neglecting the compressibility coefficient due to matric suction (ms2)
were presented. In this section, the attention is given to the compressibility coefficient due to net
stress (ms1) and the compaction curves are generated using variable coefficient of compressibility
(ms1) for undrained, semi-drained and drained loading conditions. Equation 6.3 was used for the
compressibility coefficient (ms1) calculations. It was accepted that the suction (s = ua − uw)
stays constant and equal to the initial suction (s0) during compression.
Equations 2.2, 2.3 and 6.3 were used in incremental forms to compute the incremental and
the total volume changes and the corresponding dry densities during undrained and semi-drained
compression. In addition, as it was taken into account in the previous section (see Section 6.3.1)
the compressibility of the soil element was accepted to be equal to the compressibility of water
(Cw) at the time it reaches saturation (Sr = 100%), which is only the case for drained loading.
Subsequently, a comparison of the results of the three loading types modelled using variable
compressibility coefficient (ms1) are discussed.
105
0 5 10 15 20 251.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
σy = 700 kPa
σy = 2000 kPa
Saturation curve
Figure 6.9: Modelling of the compaction curves using variable compressibility coefficient forundrained loading
Table 6.4: Parameters used for the compaction curve modelling using variable compressibilitycoefficients
Cw ua0 λvp h Gs e0(kPa−1) (kPa)4.00E-07 101.30 1.45E-01 0.02 2.65 0.68
Undrained loading
In this section, the compaction curves are generated assuming variable compressibility coefficient
(ms1) during undrained loading for different energy levels. The results are given in Figure 6.9.
Parameters used for the modelling of the compaction curves are presented in Table 6.4. The
initial matric suctions (s0 = ua0 − uw0) for varying water contents were calculated with the
assumption explained in Section 6.2.2 and are given in Table 6.5.
Figure 6.9 shows that the inverted parabolic shape of the compaction curve can be predicted
using variable compressibility coefficient (ms1) during undrained loading not only on the wet side
of the compaction curve but also on the dry side of the compaction curve. The effect of applied
stress was also well captured for this type of loading (the OWC shifts to the left).
Table 6.5: Matric suction values used for the corresponding water contents
water content (%) 2 5 10 15 20 25suction (kPa) 1183 1003 721 768 238 27
106 CHAPTER 6. MODELLING THE COMPACTION PROCESS
100
101
102
103
104
0
1
2
3
4
5
6
7
8x 10
−4
Vertical net stress, σy − u
a (kPa)
m1s (
kPa−
1 )
w = 2%w = 5%w = 10%w = 15%w = 20%
Figure 6.10: Variation of the coefficient of compressibility (ms1) due to net stress during undrained
loading
Figure 6.10 shows the variation of the coefficient of compressibility (ms1) with vertical net
stress (σy − ua) and water content (w) during undrained compression. It can be seen that the
coefficient of compressibility due to net stress (ms1) decreases with decreasing water content
(w) or initial degree of saturation (Sr0), as it was assumed in previous published work of this
study (see Appendix B). In addition, coefficient of compressibility (ms1) also decreases with
increasing vertical net stress (σy − ua). However, during the compression process, the degree of
saturation increases from the initial value, but the coefficient of compressibility (ms1) decreases.
This decrease of compressibility happens owing to the increase of net stress as the compression is
progressed. Therefore, it can be argued that the coefficient of compressibility (ms1) increases with
increasing initial degree of saturation (Sr0) but decreases with increasing degree of saturation
(Sr) during compression precess. It is also apparent that much of the compaction takes place in
the early part of the process where the soil compressibility decreases rapidly.
In Figure 6.11, the predicted values of the coefficient of compressibility (ms1) for different
values of water content (w) during undrained loading are shown together with vertical net stress
(σy−ua) development. The predictions of the coefficient of compressibility (ms1) and vertical net
stress (σy − ua) for both σy = 700 kPa and σy = 2000 kPa applied vertical stress are presented
on the figure. The curves shown by dotted lines represent the vertical net stress (σy − ua)
development with varying water content (w) and energy level. The curves shown by continues
107
5 10 15 200
1
2
3
4x 10
−4
Water content, w (%)
m1s (
kPa−
1 )
5 10 15 200
500
1000
1500
2000
Ver
tical
net
str
ess,
σy −
ua (
kPa)
σy − u
a (σ
y = 700 kPa)
σy − u
a (σ
y = 2000 kPa)
m1s (σ
y = 700 kPa)
m1s (σ
y = 2000 kPa)
Figure 6.11: Variation of the coefficient of compressibility (ms1) and the net stress with water
content during undrained loading
lines represent the predicted values of the coefficient of compressibility (ms1) for varying water
content (w) and energy levels.
As can be seen from Equation 2.2, when the volume change due to matric suction change
(∆s = ∆(ua − uw)) is disregarded, the volumetric strain (∆V/V ), and thus the dry density,
becomes a function of the compressibility coefficient (ms1) and the change in vertical net stress
(∆(σy − ua)). It is important to note that in Figure 6.11 while the coefficient of compressibility
(ms1) increases, the vertical net stress (σy − ua) decreases with increasing water content (w).
Therefore, the maximum dry density could be attained at the optimum combination of these
two parameters (ms1 and ∆(σy − ua)).
It can be seen in Figure 6.11 that the vertical net stress (σy−ua) development decreases with
increasing water contents (w). This is due to the increase in pore air pressure (ua) generation
with increasing water content (w). On the other hand, the decrease in vertical net stress (σy−ua)
increases the coefficient of compressibility (ms1) as can be seen from Equation 6.3 and Figure
6.10. In addition, the initial matric suction (s0) decreases with increasing water content (w),
which also gives rise to higher coefficient of compressibility (ms1) (see Equation 6.3). As a result,
coefficient of compressibility (ms1) increases with increasing water content (w).
The optimum combination of the coefficient of compressibility (ms1) and vertical net stress
108 CHAPTER 6. MODELLING THE COMPACTION PROCESS
0 5 10 15 20 251.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
σy = 700 kPa
σy = 2000 kPa
Saturation curve
Figure 6.12: Modelling of the compaction curves using variable compressibility coefficient forsemi-drained loading
(σy − ua) happens in lower water contents for the soil samples compacted with higher energy
than the soil samples compacted with lower energy. Figure 6.11 shows that the change in vertical
net stress (σy − ua) is higher than the change in the coefficient of compressibility (ms1) for two
different energy levels. Therefore, the vertical net stress (σy − ua) has higher impact on the
volumetric strain (∆V/V ) than the impact of the coefficient of compressibility (ms1) for the soil
samples compacted with higher energy levels. This results in the shift of the OWC to the left,
where higher net stresses could be attained for increased energy levels.
Semi-drained loading
The compaction curves generated during semi-drained loading and using variable compressibility
coefficient (ms1) are illustrated in Figure 6.12. The parameters used for the modelling are same
as the ones used for undrained loading type (Tables 6.4 and 6.5). Similar results to undrained
loading are obtained while modelling is done for semi-drained loading. A reasonable shape of
the compaction curve is attained on both wet and dry side of the compaction curve.
Figure 6.13 presents the variations of the coefficient of compressibility (ms1) with vertical
net stress (σy − ua) and water content (w) during semi-drained compression. The coefficient
of compressibility due to net stress (ms1) decreases with decreasing water content (w) or initial
degree of saturation (Sr0) and the coefficient of compressibility (ms1) also decreases with increas-
109
100
101
102
103
104
0
1
2
3
4
5
6
7
8x 10
−4
Vertical net stress, σy − u
a (kPa)
m1s (
kPa−
1 )
w = 2%w = 5%w = 10%w = 15%w = 20%
Figure 6.13: Variation of the coefficient of compressibility (ms1) due to net stress during semi-
drained loading
ing vertical net stress (σy − ua), which is similar to the undrained loading condition. The main
difference with undrained loading is that during semi drained loading higher vertical net stress
(σy − ua) values could be attained for high water contents due to the low pore air pressure (ua)
generations.
The predicted values of the coefficient of compressibility (ms1) and the vertical net stress
(σy − ua) for different values of water content (w) during semi-drained loading are shown in
Figure 6.14. The predictions of the coefficient of compressibility (ms1) and the vertical net stress
(σy−ua) for both σy = 700 kPa and σy = 2000 kPa applied vertical stresses are presented on this
figure. The curves with dotted lines represent the vertical net stress (σy − ua), while the curves
plot with continues lines represent the predicted values of the coefficient of compressibility (ms1)
for varying water content (w) and energy levels.
Similar to undrained loading, vertical net stress (σy − ua) decreases with increasing water
contents (w) and the coefficient of compressibility (ms1) increases with increasing water contents
(w). The decrease in vertical net stress (σy − ua), however, is less with increasing water content
(w) compared to undrained loading because every loading increment was considered to start
with atmospheric pore air pressure which resulted in lower pore air pressure (ua) generation.
110 CHAPTER 6. MODELLING THE COMPACTION PROCESS
5 10 15 200
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8x 10
−4
Water content, w (%)
m1s (
kPa−
1 )
5 10 15 20400
600
800
1000
1200
1400
1600
1800
2000
2200
Ver
tical
net
str
ess,
σy −
ua (
kPa)
σy − u
a (σ
y = 700 kPa)
σy − u
a (σ
y = 2000 kPa)
m1s (σ
y = 700 kPa)
m1s (σ
y = 2000 kPa)
Figure 6.14: Variation of the coefficient of compressibility (ms1) and the net stress with water
content during semi-drained loading
Drained loading
The compaction curves generated for drained loading are presented in this section. During
drained loading it was assumed that pore air (ua) pressure stays constant and equal to atmo-
spheric pressure (101.3kPa). Therefore, the pore air pressure (ua) predictions were not nec-
essary. After the predictions of the coefficients of compressibility (ms1) for the different water
contents, which is only a function of initial suction (s0 = ua0−uw0), the volume change equation
(Equation 2.2) was used for the volume change and subsequently dry density calculations. The
compaction curves modelled for this type of loading are shown in Figure 6.15. The shape of the
modelled compaction curves are similar to the ones modelled for undrained and semi-drained
loading. However, it can be seen that, as was the case for the constant compressibility coefficients
(ms1,m
s2) modelling, due to the elimination of pore air pressure (ua) generation the compaction
curves join the saturation line on the wet side of the curve.
Figure 6.16 shows the variations of the coefficient of compressibility (ms1) with vertical net
stress (σy−ua) and water content (w) during drained loading. The coefficient of compressibility
due to net stress (ms1) decreases with decreasing water content (w) and the coefficient of com-
pressibility (ms1) also decreases with increasing vertical net stress (σy − ua). For water contents
of 15% and 20%, the degree of saturation (Sr) reached 100% and therefore, the compressibility
111
0 5 10 15 20 251.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
σy = 700 kPa
σy = 2000 kPa
Saturation curve
Figure 6.15: Modelling of the compaction curves using variable compressibility coefficient fordrained loading
coefficient (ms1) dropped to the value of the water compressibility(Cw).
Figure 6.17 shows the predicted values of the coefficient of compressibility (ms1) for different
values of water content (w) during drained loading together with vertical net stress (σy − ua)
development. The predictions of the coefficient of compressibility (ms1) and vertical net stress
(σy − ua) for both σy = 700 kPa and σy = 2000 kPa applied vertical stress are presented on the
figure. The curves shown by dotted lines represent the vertical net stress (σy −ua) development
with varying water content (w) and energy level. The curves shown by continues lines represent
the predicted values of the coefficient of compressibility (ms1) for varying water content (w) and
energy levels.
During drained loading, the vertical net stress (σy−ua) stays constant with increasing water
contents (w). This is due to the disregard of the pore air pressure (ua) generation for drained
loading. The coefficient of compressibility (ms1), on the other hand, first increases and after
starts decreasing with increasing water contents (w), which is different than undrained and semi-
drained loading types. The decrease of the coefficient of compressibility (ms1) starts when the
soil element reaches 100% degree of saturation (Sr). The increasing part of the compressibility
coefficient (ms1) resulted due to the decrease of initial suction (s0) with increasing water content.
In contrast to the undrained and to a certain extent to semi-drained loading, the vertical net
stress (σy−ua) has no effect on the compressibility coefficient (ms1) with changing water content,
112 CHAPTER 6. MODELLING THE COMPACTION PROCESS
100
101
102
103
104
0
1
2
3
4
5
6
7
8x 10
−4
Vertical net stress, σy − u
a (kPa)
m1s (
kPa−
1 )
w = 2%w = 5%w = 10%w = 15%w = 20%
Figure 6.16: Variation of the coefficient of compressibility (ms1) due to net stress during drained
loading
2 4 6 8 10 12 14 16 18 20−2
0
2
4
6
8
10
12
14
16x 10
−5
m1s (
kPa−
1 )
2 4 6 8 10 12 14 16 18 20400
600
800
1000
1200
1400
1600
1800
2000
2200
Ver
tical
net
str
ess,
σy −
ua (
kPa)
Water content, w (%)
σy − ua (σy = 700 kPa)
σy − ua (σy = 2000 kPa)
m1s (σy = 700 kPa)
m1s (σy = 2000 kPa)
Figure 6.17: Variation of the coefficient of compressibility (ms1) and the net stress with water
content during drained loading
113
0 5 10 15 20 251.6
1.8
2
2.2
2.4
Water content, w (%)
Dry
den
sity
, γd (
Mg/
m3 )
UndrainedSemi−drainedDrainedSaturation curve
Figure 6.18: Comparison of the modelled compaction curves using variable compressibility co-efficients for undrained, semi-drained and drained loading (σy = 700 kPa lower curves andσy = 2000 kPa upper curves)
since it stays constant with varying water content (w). Therefore, the compaction curve owns
its shape to the shape of the coefficient of compressibility (ms1) with water content (w).
Discussion on using variable compressibility coefficients for the compaction mod-
elling
The compaction curves generated with the assumption of using variable coefficient of compress-
ibility (ms1) for undrained, semi-drained and drained loading are compared in Figure 6.18. In
this figure, compaction curves were modelled at the applied vertical stresses (σy) of 700 kPa (the
lower curve) and 2000 kPa (the upper curve).
Figure 6.18 shows that the well known inverted parabolic shape of the compaction curve was
predicted using variable compressibility coefficient (ms1) for all three types of loading conditions.
The highest dry density was produced for drained loading condition and the lowest dry density
was attained for undrained loading condition. The same results were obtained for the modelling
with constant coefficients (ms1, ms
2) of compressibility. Moreover, the dependency of the curves
to different stress levels (or energy input) was reasonably well captured, where the OWC shifts
to the left with increasing stress level.
The analysis showed that it is the compressibility coefficient due to net stress (ms1) that
114 CHAPTER 6. MODELLING THE COMPACTION PROCESS
controls the volume changes during the compression. As it can be seen from Equation 6.3 the
main controlling parameters for the coefficient of compressibility due to net stress (ms1) are the
suction (s = ua − uw) and the slope of the NCL for saturated soils (λvp). However, during the
generation of the compaction curve for a specific type of soil, the differences in the coefficient
of compressibility (ms1) between two different water contents (w) were only resulted due to the
difference in suction between these two water contents. Therefore, it can be argued that during
the generation of the compaction curve for a certain type of soil the suction (s = ua−uw) is the
governing parameter.
The coefficient of compressibility (ms1) increases with increasing initial degree of saturation
(Sr0) but decreases with increasing degree of saturation (Sr) during compression process in all
types of loading conditions. In addition, most of the compaction takes place in the early part of
the process where the soil compressibility decreases rapidly.
6.4 Concluding remarks
This chapter presented theoretical concepts for modelling the compaction behaviour for soils,
during one dimensional undrained, semi-drained and drained loading, using unsaturated soil
mechanics principles and constitutive models. The main parameters required to be predicted for
the modelling of the compaction curve are found to be the the pore pressure development and
the compressibility coefficient due to net stress (ms1). The volume changes were estimated by
the volume change constitutive relationship proposed by Fredlund and Morgenstern (1976) for
unsaturated soils. Two approaches, Hilf (1948) and Hasan & Fredlund (1980) for pore pressure
development were examined. The second approach, however, has been subsequently disregarded
in favour of the first one which presents some advantages. The most important advantage is
that it requires less parameters for the predictions of pore pressure with applied vertical stress
(σy). Moreover, it was found that the matric suction component of the volume change equation
(Equation 2.2), which was taken into account in the second approach, does not have a significant
influence on the shape of the compaction curve but the influence of suction on the compressibility
coefficient due to net stress (ms1) is a governing parameter.
It was found that by using constant compressibility coefficients (ms1,m
s2) the shape of the
compaction curve can be only produced on the wet side of the curve but not on the dry side.
On the other hand, the well-known shape of the compaction curve was modelled both on the
115
dry and wet side of the compaction curve using the variable compressibility coefficient (ms1). In
addition, the dependency of the compaction curve to the applied vertical stress (σy) or energy
level is well captured, where the OWC shifts left with the higher applied vertical stress (σy).
It was shown that the dry density is a function of the compressibility coefficient (ms1) and the
change in vertical net stress (∆(σy − ua)). During undrained and semi drained loading, while
the coefficient of compressibility (ms1) increases, the vertical net stress (σy − ua) decreases with
increasing water content (w). This is due to the increase in pore air pressure (ua) generation with
increasing water content (w). During drained loading, while the coefficient of compressibility
(ms1) increases, the vertical net stress (σy − ua) stays constant with increasing water content
(w). Therefore, the maximum dry density could be attained at the optimum combination of the
compressibility coefficient (ms1) and the change in vertical net stress ∆(σy − ua)).
The optimum combination of the coefficient of compressibility (ms1) and vertical net stress
(σy − ua) happens in lower water contents for the soil samples compacted with higher energy
than the soil samples compacted with lower energy. This results in the shift of the OWC to the
left.
It was also shown that the variation in the drainage conditions during compression may
influence the results. The highest dry densities were achieved for drained loading where there
was no pore air pressure (ua) generation and the lowest dry densities were attained for undrained
loading. The influence of the loading types on the modelling of the compaction curve will be
further examined through targeted experiments, where model’s performance is analysed (see
Chapter 7).
In conclusion, this chapter highlights the fact that the inverted parabolic shape of the com-
paction curve can be theoretically predicted using unsaturated soil mechanics principles. The
main insight gained was that the changes in the matric suction (s = ua−uw) are not significantly
important for the evolution of the compaction states, but that the influence of the matric suction
(s = ua − uw) on the material compressibility with respect to net stress (ms1) is the governing
factor determining the compaction density. Therefore, it is argued that the inverted parabolic
shape of the compaction curve is a direct function of the variation of the material compressibility
(ms1) with vertical net stress (σy − ua).
Chapter 7
Performance of the model on
predicting the static compaction
curves
7.1 Introduction
In Chapter 6, it was found that the variable compressibility coefficients must be adopted for
producing the well known shape of the compaction curve and that the variation of the com-
pressibility coefficient due to net stress (ms1) is the governing parameter on the shape of the
compaction curve. The calculation of the compressibility coefficient due to net stress (ms1)
requires the use of soil constitutive models. In literature, there are different approaches for con-
stitutive model development and the topic is still controversial. As it was explained in Chapter
6, two of those approaches were used in the compaction process modelling. This includes the
separate stress variables and the combined stress variables approach. The basis of choosing these
two approaches were discussed in the previous chapter.
In this chapter, the models developed with the two approaches are evaluated on their ability
to predict the static compaction curve. This will not only reveal the quantitative predictions by
the models but more importantly will enlighten the controversial topic in the constitutive model
development from a different perspective.
The compaction curve data of four different soils are used for the evaluation. These include
117
118 CHAPTER 7. PERFORMANCE OF THE MODEL
the data of sand-bentonite mixtures (S2B and S5B) obtained in the experimental part of this
study (see Chapter 5); and, Boom clay and Speswhite kaolin data from the literature.
Firstly, the experimental data used for the analysis is presented. Subsequently, the model
performance developed by adopting separate stress variables approach is examined. This is
followed by the evaluation of the model developed using the combined stress variables approach.
Finally, the effectiveness of the models are compared and their limitations are discussed.
7.2 Experimental data
The experimental data obtained from the current study (S2B and S5B) and data from literature
(for Boom clay and Speswhite kaolin) were used to evaluate the performance of the compaction
models using two different constitutive modelling approaches. The data from literature were
used in order to evaluate the models for a wider range of soil types, thus drawing more gener-
alised conclusions. The experimental data were used first to find the model parameters that are
necessary for implementations and later to compare the model predictions with experimental
results. The model parameters are given on the basis of both semi logarithmic relationship and
when both axes are plot in logarithmic scale as will be noted later in this section because of the
differences taken on the derivation of the constitutive models’ equations with different constitu-
tive models. The physical properties of the materials that were used in the implementations are
also summarised in this section.
7.2.1 Sand bentonite mixtures
Experimental results for sand bentonite mixtures were presented in Chapter 5. Experiments
were conducted on two different types of granular soils (S2B and S5B). Samples with different
water contents were statically compacted and their matric suctions were measured during the
compaction process using axis translation technique. Normal compression lines (NCLs) for the
saturated and unsaturated samples are given in Figures 5.8 for S2B and in Figure 5.9 for S5B.
For these soil types, the NCLs of the unsaturated samples are plotted for constant water content
tests and not for constant suction tests. However, they are considered to be approximately
equivalent to the constant suction NCLs due to the low suction change during the compaction
of these samples (see Figures 5.5 and 5.6, and Tables 5.3 and 5.4).
The specific gravity (Gs) of S2B is 2.65. The slope of the NCL for saturated S2B is 0.0131 and
119
the specific volume (ν) under 1 kPa applied stress was found to be 1.7164 for semi logarithmic
relationship; on the other hand, the slope of the NCL and the specific volume (ν) under 1 kPa
applied stress when both axes are plot in logarithmic scale were found to be 0.0079 and 1.7176,
respectively.
The specific gravity (Gs) of S5B is 2.65. The slope of the NCL of S5B is 0.0207 and its cor-
responding specific volume (ν) under 1 kPa applied vertical stress is 1.6651 for semi logarithmic
relationship. Moreover, the slope of the NCL of S5B is 0.0132 and the specific volume (ν) under
1 kPa applied vertical stress is 1.6678 when both axes are plot in logarithmic scale.
7.2.2 Boom clay
Static compaction curve data for Boom clay were obtained from Romero (1999), where he studied
the thermo-hydro-mechanical behaviour of unsaturated Boom clay. In this study, static com-
paction tests were conducted for five different water contents ranging from 3% to 22.3%. The
samples were compacted to the desired density with their initially determined water content
and later they were removed from the compaction mould for suction measurements. The dry
samples that had high suction values (higher than 1500 kPa) were placed in a psychrometer and
their total suctions were measured using vapour equilibration technique. On the other hand,
the matric suction of the wetter samples were measured using axis translation technique. Post
compaction suction was measured by both methods.
Romero (1999) conducted consolidation test for saturated soil up to the maximum vertical
stress of 1 MPa. However, the maximum vertical stress applied during the static compaction
tests for unsaturated samples was 10 MPa. Thus, other studies were also examined to find out
the behaviour of saturated Boom clay under a wider range of applied vertical stresses as shown
in Figure 7.1. Bouazza et al. (1996) also investigated the consolidation behaviour of saturated
Boom clay up to 1 MPa vertical stress. Deng et al. (2010), however, conducted oedometer test
under wider range of stresses, where the maximum applied vertical stress was 32 MPa. Thus,
the slope of this curve is adopted for implementations in this study.
The specific gravity (Gs) of Boom clay is 2.70 (Romero, 1999). The slope of the NCL is
0.1446 and the specific volume (ν) under 1 kPa applied stress was found to be 2.7703 for semi
logarithmic relationship; on the other hand, the slope of the NCL and the specific volume (ν)
under 1 kPa applied stress when both axes are plot in logarithmic scale were found to be 0.0993
and 3.5744, respectively (Deng et al., 2010). The elastic stiffness parameter for the changes in
120 CHAPTER 7. PERFORMANCE OF THE MODEL
101
102
103
104
105
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
Vertical effective stress, σv − u
w (kPa)
Spe
cific
vol
ume,
ν
Romero (1999)Bouazza et al. (1996)Deng et al. (2010)
ν = −0.1446ln(σv − ua) + 2.7703
Figure 7.1: Normal compression line of Boom clay
suction (κs) is 0.00732 (Vaunat and Gens, 2005).
Slopes of the NCLs for unsaturated samples were determined from the study of Romero
(1999), where the constant suction contours were plotted on the static compaction curves data.
The results are shown in Figure 7.2. It is interesting to note that the slope of the NCL for
saturated sample is lower than the slopes of the NCL for unsaturated clay samples. In addition,
unsaturated samples have higher specific volumes (ν) than the saturated samples under low
stresses. This is most probably due to the aggregation of soil into cloddy structure within the
drier samples.
7.2.3 Speswhite kaolin
Speswhite kaolin data were obtained from Tarantino and De Col (2008). They conducted static
compaction tests for twelve different water contents ranging between 8.5% and 31%. The matric
suction of the samples higher than 18% water content were measured by tensiometer. Suction
measurements of the drier samples were conducted by psychrometer; thus, the total suction of
these samples were measured. Continuous matric suction measurements during compression were
conducted for the relatively wet samples (wc = 22− 31% ) in which a high capacity tensiometer
was used for the matric suction measurements. On the other hand, post compaction suctions
were measured for the samples that had lower water contents (or higher suctions).
121
102
103
104
105
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
2.1
Vertical net stress, σv − u
a (kPa)
Spe
cific
vol
ume,
νs = 0s = 450s = 3000s = 6000s = 8000s = 10000s = 32000s = 102000
Figure 7.2: Normal compression lines of Boom clay under constant suctions
The specific gravity (Gs) of Speswhite kaolin is 2.61 (Chung, 2006). The slope of the NCL
is 0.1537 and the specific volume (ν) under 1 kPa applied stress was found to be 2.851 for semi
logarithmic relationship (see Figure 7.3); on the other hand, the slope of the NCL and the specific
volume (ν) under 1 kPa applied stress when both axes are plot in logarithmic scale were found
to be 0.0748 and 3.0156, respectively. The elastic stiffness parameter for changes in suction (κs)
is 0.00013 (Tarantino and De Col, 2008).
The NCLs for unsaturated samples under constant suction are given in Figure 7.4. These data
were derived from the constant suction contours presented on the compaction curve in Tarantino
and De Col (2008). Similar to the Boom clay data, unsaturated samples have higher slopes than
the saturated sample. In addition, the slope tends to reduce with increasing stresses. The initial
specific volume of the unsaturated samples (N(s)) are higher than the saturated sample (N(0)).
7.3 Separate stress state variables approach
In this section, the models developed with two stress state variables approach are examined on
their ability to predict the static compaction curve. The performance of the Barcelona Basic
Model (BBM) developed by Alonso et al. (1990) is evaluated in detail. The BBM is the first and
most widely used elasto-plastic model for unsaturated soils and other models based on two stress
state variables approach (Wheeler and Sivakumar, 1995; Vaunat et al., 2000; Chiu and Ng, 2003;
122 CHAPTER 7. PERFORMANCE OF THE MODEL
101
102
103
1.8
1.9
2
2.1
2.2
2.3
2.4
2.5
Vertical effective stress, σv − u
w (kPa)
Spe
cific
vol
ume,
νν = −0.1537ln(σv − uw) + 2.851
Figure 7.3: Normal compression line of Speswhite kaolin
101
102
103
104
1.8
1.9
2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
Vertical net stress, σv − u
a (kPa)
Spe
cific
vol
ume,
ν
s = 0s = 400s = 600s = 800s = 1000s = 1200s = 2000s = 3000s = 5000
Figure 7.4: Normal compression lines of Speswhite kaolin for constant suctions
123
Thu et al., 2007) are mainly derived in a similar manner or further improved by adding other
important behaviour of unsaturated soils which was not covered in the BBM (e.g., incorporation
of SWRC). Thus, the BBM is chosen for thorough examination in this study. The application of
the BBM on the development of the compaction process is presented in Chapter 6. Here, only
the results of the implementations are shown and discussed.
Firstly, the BBM is used for predicting the compaction behaviour of four different soils;
orderly, S2B, S5B, Boom clay and Speswhite kaolin. The performance of using the BBM in
compaction model development is assessed for each soil type. Model parameters that are neces-
sary for the implementations were determined using the experimental data and/or when avail-
able the model parameters found in literature. Subsequently, the static compaction curves were
predicted using the model parameters and the modelling results were compared with the experi-
mental results. Although the experiments were conducted using semi-drained loading conditions,
the implementations of undrained loading conditions are also presented for comparison. During
undrained loading both air and water phases were considered to be undrained. Whereas, in
semi-drained loading, the air phase was assumed to be undrained during each loading step but
the buildup air pressure was released before the next loading increment; thus, the air pressure
was set to atmospheric pressure in the beginning of every loading increment. Like in undrained
loading, the water phase was also assumed to be undrained in semi-drained loading simulations
(see Section 6.2.2 for details in the loading conditions). Suction was considered to stay constant
during loading for constant water content tests in both type of simulations. Implications of
assuming constant suction throughout loading are discussed later in this section. To incorporate
other constitutive models, that are developed with separate stress variables approach, in com-
paction model development are also evaluated briefly. Finally, the advantages and shortcomings
of using the current constitutive models in compaction model development are discussed.
7.3.1 Sand bentonite mixtures
The performance of the compaction model which is developed using the BBM is evaluated on
two granular soils; S2B and S5B. Firstly, the results for S2B is presented and this is followed
with the implementations for S5B.
The experimental results on the variation of the slopes of the NCLs (λ(s)) with matric suction
are shown in Figure 7.5 for the S2B. These results were calculated using the experimental data
presented in Figure 5.8. Model estimations for the slope of the NCLs calculated using the
124 CHAPTER 7. PERFORMANCE OF THE MODEL
0 2 4 6 80.01
0.015
0.02
0.025
0.03
0.035
0.04
Matric suction, s (kPa)
λ(s)
experimentmodel (Eq. 2.13)modified model
Figure 7.5: λ(s) of S2B for varying suctions (BBM)
equation proposed by the BBM (see Equation 2.13) are also presented on the graph and the
parameters used for the predictions are given in Table 7.1. Experimental observations revealed
that the slope of the NCL (λ(s)) first increases and later starts to decrease slightly as matric
suction increased. However, the matric suction range covered in the experiments were very low
due to the difficulties in the measurements of high suctions in granular materials (see Section
5.3). For this reason, the evolution of the slope of the NCLs at higher suctions is not very
certain. Nevertheless, the model estimates are far beyond the experimental data mainly due to
the assumption adopted in the BBM, which states that the slopes of the NCLs decreases with
increasing matric suctions. It can be seen from the experimental results presented in Figure 7.5
that the slopes of the NCLs have their lowest value at saturated state. Thus, the model will not
be able to predict the correct deformations for the S2B samples. The model parameters were
modified in order to fit most of the experimental data and modified parameters are presented in
Table 7.2. The estimates with the modified parameters are also shown in Figure 7.5 and they fit
most of the experimental data better than the original model predictions except at zero suction.
It should be noted that, the slope of the NCL (λ(0)) for the saturated sample was modified but
the soil parameters κs, pc, r and β were kept constant. The consequence of such change is only
a shift of the curve, as shown in Figure 7.5, which is the relationships between the slope of the
NCLs (λ(s)) and the matric suction.
The experimental data for the initial specific volumes (N(s)) under the reference stress (pc)
125
Table 7.1: Model parameters for S2B (BBM)
N(0) λ(0) κs pc r β(kPa) (kPa−1)
1.7164 0.0131 0.00006 10 0.769 0.18985
Table 7.2: Modified model parameters for S2B
N(0) λ(0) κs pc r β(kPa) (kPa−1)
1.835 0.036 0.00006 10 0.769 0.18985
are presented in Figure 7.6 for S2B. The initial specific volumes (N(s)) were calculated using the
Equation 2.15. The modelling estimates for the variation of the initial specific volume (N(s))
with matric suction are also shown on the figure. The BBM model estimated a decrease in the
initial specific volume (N(s)) with increasing matric suctions. This is due to the assumption
taken by the model, which states that ”at reference stress (pc) the suction increase leads to an
elastic shrinkage in the soils”. The experimental results in Figure 7.6, however, show that the
initial specific volumes (N(s)) increase with increasing matric suctions. Similar to the slope
of the NCLs (λ(s)), the initial specific volumes (N(s)) have their lowest value at saturation.
The model predictions of N(s) poorly represent the experimental behaviour and therefore were
modified. The modification was done only by increasing the initial specific volume (N(0)) of the
saturated state and keeping the other model parameters constant (κs, pc, r and β). There is a
better agreement between the model predictions with modified parameters (see Table 7.2) and
the experimental results except for the data at saturation.
Using the modified parameters, the model predictions for the variation of dry density with
applied vertical stress are shown in Figure 7.7. Figure 7.7(a) presents the comparison between
the undrained and semi-drained simulations; whereas, Figure 7.7(b) compares the semi-drained
model predictions with the experimental results (experiments were also semi-drained). The
difference between the undrained and semi-drained simulations is not very notable because the
degrees of saturation reached during the compaction are low (at about between 28% and 43%).
Since the pore air volume is high at low degrees of saturation, the development of pore air
pressure is low. This subject was explained in detail in Section 6.2.2.
Figure 7.7(b) demonstrates that the model predictions for the dry densities at the lower
water contents (w = 6.9% and w = 8.4%) fit better the experimental results than the model
predictions at relatively wetter samples (w = 8.6% and w = 10.7%). At 6.9% water content, the
model estimates for both the slope of the NCL (λ(s)) and the initial specific volume (N(s)) are
126 CHAPTER 7. PERFORMANCE OF THE MODEL
0 2 4 6 81.7
1.72
1.74
1.76
1.78
1.8
1.82
1.84
Matric suction, s (kPa)
N(s
)
experimentmodel (Eq. 2.15)modified model
Figure 7.6: N(s) for S2B for varying matric suctions (BBM)
in good agreement with the experimental data, which in return resulted in a reasonable match
between the experimental and the predicted values for the dry density. In addition, at 8.4%
water content (5.2kPa matric suction), the model estimated a higher slope of the NCL (λ(s))
and a higher initial specific volume (N(s)) (initially less dense sample) than the experimental
data. These two values (λ(s), N(s)) compensated each other and resulted in a good agreement
between the predicted and experimental dry density. At 8.6% (5.7kPa matric suction) and 10.7%
water contents, the model estimated less compressible (lower λ(s)) and initially looser (higher
N(s)) sample compared to the experimental data, which led to a lower predicted dry density.
Figure 7.8 presents the model predictions of the static compaction curves for S2B. The
comparison between the undrained and semi-drained implementations are shown in Figure 7.8(a).
Figure 7.8(b) illustrates the comparison between the modelled and experimental compaction
curves. There is no much difference between the undrained and semi-drained predictions due to
the low degrees of saturation attained during the compaction process (see Figure 5.12).
At 6.9% and 8.4% water contents, the model predictions are in good agreement with the
experimental results. However, the model underestimates the experimental results at higher
water content as shown in Figure 7.8(b). As discussed earlier, this is due to crude estimates of
the the slope of the NCLs (λ(s)) and the initial specific volumes (N(s)) by the model.
The experimental results for the variation of the slopes of the NCLs (λ(s)) with matric suc-
tion are shown in Figure 7.9 for S5B. The results were obtained from the experimental data
127
[a]
0 500 1000 1500 2000 25001.44
1.46
1.48
1.5
1.52
1.54
1.56
1.58
1.6
1.62
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
wc = 6.9% Uwc = 8.4% Uwc = 8.6% Uwc = 10.7% Uwc = 6.9% Swc = 8.4% Swc = 8.6% Swc = 10.7% S
[b]
0 500 1000 1500 2000 25001.44
1.46
1.48
1.5
1.52
1.54
1.56
1.58
1.6
1.62
1.64
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
wc = 6.9% Mwc = 8.4% Mwc = 8.6% Mwc = 10.7% Mwc = 6.9% Ewc = 8.4% Ewc = 8.6% Ewc = 10.7% E
Figure 7.7: BBM predictions of the variation of dry density with applied vertical stress for S2B[a] Comparison of the undrained (U) and semi-drained (S) predictions [b] Comparison of theexperimental data (E) and model (M) predictions (semi-drained)
128 CHAPTER 7. PERFORMANCE OF THE MODEL
[a]
6 7 8 9 10 111.55
1.56
1.57
1.58
1.59
1.6
1.61
Water content (%)
Dry
den
sity
(M
g/m
3 )
560 kPa U2250 kPa U560 kPa S2250 kPa S
[b]
6.5 7 7.5 8 8.5 9 9.5 10 10.5 111.55
1.56
1.57
1.58
1.59
1.6
1.61
1.62
1.63
1.64
1.65
Water content (%)
Dry
den
sity
(M
g/m
3 )
560 kPa M2250 kPa M560 kPa E2250 kPa E
Figure 7.8: Compaction curves for S2B predicted by BBM [a] Comparison of undrained (U)and semi-drained (S) predictions [b] Comparison of the experimental data (E) and model (M)predictions (semi-drained)
129
0 10 20 30 40 500.01
0.015
0.02
0.025
0.030.03
Matric suction, s (kPa)
λ(s)
experimentmodel (Eq. 2.13)modified model
Figure 7.9: λ(s) of S5B for varying matric suctions (BBM)
Table 7.3: Model parameters for S5B (BBM)
N(0) λ(0) κs pc r β(kPa) (kPa−1)
1.6174 0.021 0.00008 10 0.765 0.451
presented in Figure 5.9. The slope of the NCLs (λ(s)) initially increases with increasing matric
suctions and after reaching a peak at about 35kPa matric suction, it starts to decrease. Model
estimations for the NCLs are also presented in Figure 7.9. These values were calculated using
the equation proposed in the BBM (see Equation 2.13) and the model parameters presented in
Table 7.3. Different than the experimental observation, the BBM predicts decreasing slope of
the NCLs (λ(s)) with increasing matric suctions. The model predicts less compressible soil than
the experimental observations and thus, this will lead to poor predictions of the soil deformation.
Therefore, the model parameters were modified and the predictions with the modified param-
eters are also presented in Figure 7.9. The modified model parameters are given in Table 7.4.
Modification was done by increasing the slope of the NCL (λ(0)) for saturated soil and keeping
the other parameters constant (κs, pc, r and β). This modification resulted only in a shift of
the relationship between the slope of the NCLs (λ(s)) and the matric suction on the ordinate.
The experimental data for the initial specific volumes (N(s)) under the reference stress (pc)
are presented in Figure 7.10 for the S5B. The initial specific volumes (N(s)) increased with
increasing matric suctions. The model estimates for the initial specific volumes (N(s)) are
130 CHAPTER 7. PERFORMANCE OF THE MODEL
Table 7.4: Modified model parameters for S5B
N(0) λ(0) κs pc r β(kPa) (kPa−1)
1.7 0.028 0.00008 10 0.765 0.451
0 10 20 30 40 501.6
1.65
1.7
1.75
Matric suction, s (kPa)
N(s
)
experimentmodel (Eq. 2.15)modified model
Figure 7.10: N(s) for S5B for varying matric suctions (BBM)
also presented in Figure 7.10 and are far from the experimental results. Model estimates were
obtained from the Equation 2.15. The BBM predicts a decrease in the initial specific volumes
(N(s)) with increasing matric suctions due to the assumption taken by the model. Similar to the
slope of the NCLs (λ(s)), the initial specific volumes (N(s)) have their lowest value at saturated
state. The model parameters were modified in a similar manner to S2B by increasing the initial
specific volume for saturated sample (N(0)) and keeping the other parameters constant (κs, pc,
r and β). The predictions were made in a way that they would fit most of the experimental
data. However, due to the nature of the equation proposed in the model (Equation 2.15), it
was very difficult to fit the model predictions to the entire range of suctions. The modification
was done to fit the values between 10kPa and 20kPa. This selection was made considering the
results of the slope of the NCLs (λ(s)) (Figure 7.9). The model predictions for the slope of the
NCLs (λ(s)) were best capturing the experimental data about that suction range.
Using the modified model parameters, the model predictions for the variation of dry density
with applied vertical stress are shown in Figure 7.11. Figure 7.11(a) presents the comparison
between the undrained and semi-drained simulations; whereas, Figure 7.11(b) compares the semi-
131
drained model predictions with the experimental results which were also conducted under semi-
drained loading conditions. The difference between the undrained and semi-drained simulations
is not very notable due to the low degrees of saturation attained during the compaction process
(between 28% and 59%). It was explained in Section 6.2.2 that the pore air pressure buildup is
crucial at high degrees of saturation.
Figure 7.11(b) shows that the model predictions for the dry densities, at 10.3% and 10.9%
water contents fit better the experimental results than the model predictions at 6.5%, 7.5% and
12.6% water contents. This is purely related to the agreement between the model estimates
and the experimental values on the slope of the NCLs (λ(s)) and the initial specific volumes
(N(s)). At 10.3% and 10.9% water contents, the model estimates are in good agreement with
the experimental data for both the slope of the NCLs (λ(s)) and the initial specific volumes
(N(s)), which led to a reasonable match between the experimental and the predicted values
for dry densities. At 6.5% and 7.5% water content, the model estimated lower values for both
the slope of the NCLs (λ(s)) and initial specific volumes (N(s)) (initially denser soil) than
the experimental data. This resulted in higher dry density predictions. It can be noticed in
Figure 7.11(b) that the difference between the predicted and experimental values increases with
increasing applied vertical stresses. This is the consequence of the model being more sensitive to
the slope of the NCLs (λ(s)) at increased stresses (see Equation 2.14). At 12.6% water content,
the model estimated less compressible (lower λ(s)) and initially looser (higher N(s)) sample
compared to the experimental data, which led to a lower predicted dry density.
Figure 7.12 presents the model predictions of the static compaction curves for S5B. The com-
parison between the undrained and semi-drained implementations are shown in Figure 7.12(a).
In addition, Figure 7.12(b) illustrates the comparison between the modelled and experimen-
tal compaction curves. Due to the low degrees of saturation attained during the compaction
process (see Figure 5.13), there is no much difference between the undrained and semi-drained
predictions.
At 10.3% and 10.9% water contents, the model predictions are in good agreement with the
experimental results. However, at 6.5% and 7.5% water contents, the model over estimates the
experimental results, and at 12.6% water content, the model under estimates the experimental
dry density. As discussed earlier, this is due to crude estimates of the the slope of the NCLs
(λ(s)) and the initial specific volumes (N(s)) by the model.
132 CHAPTER 7. PERFORMANCE OF THE MODEL
[a]
0 500 1000 1500 2000 2500 3000 35001.56
1.58
1.6
1.62
1.64
1.66
1.68
1.7
1.72
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
wc = 6.5% Uwc = 7.5% Uwc = 10.3% Uwc = 10.9% Uwc = 12.6% Uwc = 6.5% Swc = 7.5% Swc = 10.3% Swc = 10.9% Swc = 12.6% S
[b]
0 500 1000 1500 2000 2500 3000 35001.56
1.58
1.6
1.62
1.64
1.66
1.68
1.7
1.72
1.74
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
wc = 6.5% Mwc = 7.5% Mwc = 10.3% Mwc = 10.9% Mwc = 12.6% Mwc = 6.5% Ewc = 7.5% Ewc = 10.3% Ewc = 10.9% Ewc = 12.6% E
Figure 7.11: BBM predictions of the variation of dry density with applied vertical stress for S5B[a] Comparison of the undrained (U) and semi-drained (S) predictions [b] Comparison of theexperimental data (E) and model (M) predictions (semi-drained)
133
[a]
6 7 8 9 10 11 12 131.655
1.66
1.665
1.67
1.675
1.68
1.685
1.69
1.695
1.7
Water content (%)
Dry
den
sity
(M
g/m
3 )
560 kPa U2650 kPa U560 kPa M2650 kPa M
[b]
6 7 8 9 10 11 12 131.62
1.64
1.66
1.68
1.7
1.72
1.74
Water content (%)
Dry
den
sity
(M
g/m
3 )
560 kPa M2650 kPa M560 kPa E2650 kPa E
Figure 7.12: Compaction curves for S5B predicted by BBM [a] Comparison of undrained (U)and semi-drained (S) predictions [b] Comparison of the experimental data (E) and model (M)predictions (semi-drained)
134 CHAPTER 7. PERFORMANCE OF THE MODEL
0 20 40 60 80 100 1200.08
0.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
0.26
Suction, s (MPa)
λ(s)
experimentmodel (Eq. 2.13)modified model
Figure 7.13: λ(s) of Boom clay for varying suctions (BBM)
7.3.2 Boom clay
The experimental results for the slopes of the NCLs (λ(s)) of Boom clay are shown in Figure 7.13.
These results were calculated using the experimental data presented in Figure 7.1. The model
estimations for the NCLs calculated using the equation proposed by BBM (see Equation 2.13)
are also presented in this figure and the parameters used for the predictions are given in Table
7.5. Experimental observations reveal that the slopes of the NCLs (λ(s)) first increases and later,
after a peak value at about 3 MPa suction, start decreasing with increasing suctions. It should be
noted that the slopes of the NCLs for unsaturated Boom clay have greater values than the slope
of the NCL for its saturated state. This observation is contrary to the assumption taken by BBM
(Alonso et al., 1990). Therefore, the modelling predictions for the slope of the NCLs (λ(s)) are
far beyond the experimental results. In order to better implement the experimental behaviour,
the model parameters were modified to fit most of the experimental data and the predictions
with the modified parameters are also presented in the same figure. The modified parameters
are given in Table 7.6 and the values are similar to the ones found in other studies (Vaunat
and Gens, 2005) for the same soil. This time the predictions fit most of the experimental data
reasonably well except for the saturated sample. Similar to the modification done for granular
soils (see Section 7.3.1) only the slope of the NCL (λ(0)) and the initial specific volume for
saturated sample (N(0)) were modified.
135
Table 7.5: Model parameters for Boom clay (BBM)
N(0) λ(0) κs pc r β(kPa) (kPa−1)
2.4373 0.1446 0.0732 10 0.564 0.0000544
Table 7.6: Modified model parameters for Boom clay
N(0) λ(0) κs pc r β(kPa) (kPa−1)
3.1 0.26 0.0732 10 0.564 0.0000544
The experimental data for the initial specific volumes (N(s)) under reference stress (pc) are
presented in Figure 7.14 for varying suctions. Model predictions using Equation 2.15 and the
initial model parameters (see Table 7.5) are also shown in the figure. The BBM model predicts
a decrease in the initial specific volume (N(s)) with increasing suction. On the other hand,
the experimental results in Figure 7.14 show that the initial specific volume (N(s)) is at its
lowest value for saturated Boom clay. Similar to the model predictions for the slopes of the
NCLs (λ(s)), the model predictions for the initial specific volumes (N(s)) poorly represent the
experimental behaviour, and therefore the model parameters were modified. There is a good
agreement between the model predictions with modified parameters (see Table 7.6) and the
experimental results except for the data at saturation.
Using the modified parameters, the model predictions on the variation of dry density with ap-
plied vertical stress is shown in Figure 7.15 for both undrained and semi-drained simulations. In
semi-drained loading simulations, the loading was ceased when the samples reached saturation.
At low water contents, both type of simulations are predicting similar results (7.15(a)). However,
at high water contents, undrained model simulations produce lower dry densities than the pre-
dictions of the semi-drained simulations. This is due to the high pore air pressure development
at high water contents during undrained loading. Although, the experiments were conducted
under semi-drained loading conditions, undrained loading simulations are also presented here for
comparison, because the well known inverted parabolic shape of the dynamic compaction curve
can be better produced under undrained loading conditions and it better represents practice.
Examination of Figure 7.15(b) shows that the model underestimates the experimental results
at low water contents (3%) and overestimates the results at high water contents (9.8%, 14.7%,
18.7%) except for the sample with 22.3% water content. This is similar to the estimations of
the slopes of the NCLs (λ(s)) (see Figure 7.13) by the model. Therefore, it can be stated that
the model predictions are mainly influenced by the slopes of the NCLs (λ(s)) for Boom clay.
136 CHAPTER 7. PERFORMANCE OF THE MODEL
0 20 40 60 80 100 1202.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
3
3.1
3.2
Suction, s (MPa)
N(s
)
experimentmodel (Eq. 2.15)modified model
Figure 7.14: N(s) for Boom clay for varying suctions (BBM)
At 22.3% water content, the model prediction for the slopes of the NCL (λ(s)) is higher than
the experimental value and therefore, the predicted dry densities are expected to be higher than
the experimental values. However, the modelled and experimental results are in good agreement
(Figure 7.15(b)). This is considered to arise due to the suction reduction at high water contents
during loading in the experiments, which was not considered in the current model development
for simplicity.
The modelling predictions for the static compaction curve for Boom clay are presented in
Figure 7.16 for undrained and semi-drained simulations. As before, there is not much difference
between undrained and semi-drained simulations at low water contents. However, considerable
difference is observed after 9.8% water content (Figure 7.16(a)). In addition, at high water
contents, the predicted dry densities start dropping during undrained simulations resulting in a
well known inverted parabolic shape of the compaction curve. Semi-drained simulations do not
produce similar inverted parabolic shape because it is accepted that there is no trapped air at
high water contents, which would normally result in high pore air pressure buildup within the
sample, and the compaction is ceased after reaching the saturation.
Figure 7.16(b) shows the comparison between the model predictions and experimental data
for the compaction curves. The model predictions poorly represent the experimental results at
3% water content. Moreover, at higher water contents, the model overestimates the experimental
results, as it was stated previously, except at 22.3% water content. This is due to the overesti-
137
[a]
0 2000 4000 6000 8000 100000.8
1
1.2
1.4
1.6
1.8
2
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
w = 3% Uw = 9.8% Uw = 14.7% Uw = 18.7 Uw = 22.3 Uw = 3% Sw = 9.8% Sw = 14.7% Sw = 18.7 Sw = 22.3 S
[b]
0 2000 4000 6000 8000 100000.8
1
1.2
1.4
1.6
1.8
2
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
w = 3% Mw = 9.8% Mw = 14.7% Mw = 18.7% Mw = 22.3% Mw = 3% Ew = 9.8% Ew = 14.7% Ew = 18.7% Ew = 22.3% E
Figure 7.15: BBM predictions of the variation of dry density with applied vertical stress for Boomclay [a] Comparison of the undrained (U) and semi-drained (S) predictions [b] Comparison ofthe experimental data (E) and model (M) predictions (semi-drained)
138 CHAPTER 7. PERFORMANCE OF THE MODEL
Table 7.7: Model parameters for Speswhite kaolin (BBM)
N(0) λ(0) κs pc r β(kPa) (kPa−1)
2.4971 0.1537 0.00013 10 0.74 0.0018
Table 7.8: Modified model parameters for Speswhite kaolin
N(0) λ(0) κs pc r β(kPa) (kPa−1)
4.12 0.49 0.00013 10 0.74 0.0018
mation of the slopes of the NCLs (λ(s)) by the model. The influence of the slope of the NCL
(λ(s)) on the specific volume or the dry density increases as the applied stress increases. In other
words, as stress increases the model becomes more sensitive to the slope of the NCL (λ(s)) than
to the initial specific volume (N(s)). This can be clearly seen from Equation 2.14. Therefore,
the differences between experimental and predicted values increase as the applied vertical stress
increases (Figure 7.15(b) and 7.16(b)).
7.3.3 Speswhite kaolin
Figure 7.17 presents the experimental and modelling results for the variation of the slopes of the
NCLs (λ(s)) with suction for Speswhite kaoline. Experimental points were calculated from the
data presented in Figure 7.4. Modelling predictions for the slope of the NCLs were calculated
using the equation proposed by the BBM (see Equation 2.13). Similar to sand-bentonite mixtures
and Boom clay, again the experimental observations are far beyond the model estimations;
therefore, the model parameters were modified and the modified model predictions are also
presented in Figure 7.17. Model parameters and modified model parameters are given in Tables
7.7 and 7.8, respectively. Again, the modification is done only to the slope of the NCL (λ(0))
and the initial specific volume for saturated sample (N(0)). Both parameters were increased
which makes the soil sample more compressible than the model predictions.
The experimental data for the initial specific volume (N(s)) under reference stress (pc) are
presented in Figure 7.18 for varying suctions. The model estimates using Equation 2.15 are
shown in the figure together with the modified modelling estimations. Similar to the estimations
observed in sand-bentonite mixtures and Boom clay, BBM predicts a decrease in the initial
specific volume (N(s)) with increasing suction. On the other hand, experimental data in Figure
7.18 illustrate that the initial specific volume (N(s)) is at its lowest value for saturated sample.
139
[a]
0 5 10 15 20 25
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
Water content (%)
Dry
den
sity
(M
g/m
3 )
1 MPa U2 MPa U3 MPa U4 MPa U5 MPa U6 MPa U8 MPa U10 MPa U1 MPa S2 MPa S3 MPa S4 MPa S5 MPa S6 MPa S8 MPa S10 MPa S
[b]
0 5 10 15 20 25
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
Water content (%)
Dry
den
sity
(M
g/m
3 )
1 MPa M2 MPa M3 MPa M4 MPa M5 MPa M6 MPa M8 MPa M10 MPa M1 MPa E2 MPa E3 MPa E4 MPa E5 MPa E6 MPa E8 MPa E10 MPa E
Figure 7.16: Compaction curves for Boom clay predicted by BBM [a] Comparison of undrained(U) and semi-drained (S) predictions [b] Comparison of the experimental data (E) and model(M) predictions (semi-drained)
140 CHAPTER 7. PERFORMANCE OF THE MODEL
0 1000 2000 3000 4000 50000.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Suction, s (kPa)
λ(s)
experimentmodel (Eq. 2.13)modified model
Figure 7.17: λ(s) of Speswhite kaolin for varying suctions (BBM)
Similar to the model predictions of the slope of the NCL (λ(s)), the model predictions of the
initial specific volume (N(s)) poorly represent the experimental behaviour and therefore were
modified. There is a better agreement between the model predictions with modified parameters
(see Table 7.8) and the experimental results except for the data at zero suction.
Figure 7.19 shows the variation in void ratio with applied vertical stress for Speswhite kaolin.
Both undrained and semi-drained simulations are shown in Figure 7.19(a). Undrained simula-
tions have lower deformations than the semi-drained ones due to the higher pore air pressure
buildup within the undrained samples. Figure 7.19(b) illustrates the comparison between the
model predictions and experimental results for the samples with 24%, 25% and 28% water con-
tent. In overall, there is a good agreement between the experimental and predicted values, with
only a slight underestimation of the predicted values at 28% water content. This discrepancy is
most likely to happen due to suction reduction with loading which was not taken into account
in the model simulations.
Figure 7.20 shows the predicted static compaction curve of Speswhite kaolin for both undrained
and drained loading types. Firstly, in Figure 7.20(a) the comparison between undrained and
semi-drained simulations are given and later, in Figure 7.20(b) experimental data are compared
with the model predictions. Dry density stays almost constant until 18% water content for both
type of simulations and later slightly decreases with increasing water content in undrained sim-
ulations; whereas, it increases in semi-drained simulations. The slope of the NCL (λ(s)) and
141
0 1000 2000 3000 4000 5000
2.6
2.8
3
3.2
3.4
3.6
3.8
4
4.2
Suction, s (kPa)
N(s
) experimentmodel (Eq. 2.15)modified model
Figure 7.18: N(s) for Speswhite kaolin for varying suctions (BBM)
the initial specific volume (N(s)), which are the parameters that influence the deformation, stay
constant until 18% water content (that corresponds to 2000 kPa suction), as shown in Figures
7.17 and 7.18. Therefore, the dry densities also stay almost constant for both type of simulations
(undrained and semi-drained). After 18% water content (less suction than 2000 kPa), both the
slope of the NCL (λ(s)) and the initial specific volume (N(s)) start to increase, which means
the soil is more compressible and initially in a looser state. Therefore, as it is stated that the
initial specific volume (N(s)) is more dominant at low applied stresses and the slope of the NCL
(λ(s)) at high applied stresses. It can be seen that at higher applied stresses the dry density
reduction is much less or even dry density stays almost constant due to the higher influence of
the the slope of the NCL (λ(s)) on the model. In addition, the development of pore air pressure
within the samples resulted in slight reduction of the dry density in undrained simulations.
The comparison of the experimental data and model predictions shown in Figure 7.16(b)
reveals that the model can predict the compaction curve at mid water contents reasonably
well but model underestimates the dry densities at higher water contents. Nevertheless, if
the reduction of suction with loading was included in the model, reasonable predictions could
probably be attained at higher water contents as it was shown by Tarantino and De Col (2008).
On the other hand, at low water contents (up to 18%) model overestimates the experimental
data and inclusion of suction reduction with loading would not help to attain better predictions.
Firstly, because inclusion of suction reduction would result in higher predicted dry densities
142 CHAPTER 7. PERFORMANCE OF THE MODEL
[a]
0 200 400 600 800 1000 12000.5
1
1.5
2
2.5
3
Applied vertical stress (kPa)
Voi
d ra
tio (
e) w = 24% Uw = 25% Uw = 28% Uw = 24% Sw = 25% Sw = 28% S
[b]
0 200 400 600 800 1000 12000.5
1
1.5
2
2.5
3
Applied vertical stress (kPa)
Voi
d ra
tio (
e) w = 24% Mw = 25% Mw = 28% Mw = 24% Ew = 25% Ew = 28% E
Figure 7.19: BBM predictions of the variation of void ratio with applied vertical stress forSpeswhite kaolin [a] Comparison of the undrained (U) and semi-drained (S) predictions [b]Comparison of the experimental data (E) and model (M) predictions (semi-drained)
143
[a]
5 10 15 20 25 30 350.9
0.95
1
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
Water content (%)
Dry
den
sity
(M
g/m
3 )
300 kPa U600 kPa U900 kPa U1200 kPa U300 kPa S600 kPa S900 kPa S1200 MPa S
[b]
5 10 15 20 25 30 350.9
0.95
1
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
Water content (%)
Dry
den
sity
(M
g/m
3 )
300 kPa M600 kPa M900 kPa M1200 kPa M300 kPa E600 kPa E900 kPa E1200 MPa E
Figure 7.20: Compaction curves for Speswhite kaolin predicted by BBM [a] Comparison ofundrained (U) and semi-drained (S) predictions [b] Comparison of the experimental data (E)and model (M) predictions (semi-drained)
144 CHAPTER 7. PERFORMANCE OF THE MODEL
than the current ones, which would increase the error. Secondly, at low water contents suction
reduction with loading is not an expected phenomena and generally evident from experimental
results (Li, 1995; Tarantino and De Col, 2008).
7.3.4 Discussion on the BBM
The BBM was evaluated on its ability to predict the static compaction curve of four different
type of soils. The following conclusions about the model are gained:
• The model assumes that the slope of the NCLs (λ(s)) decreases with increasing suction.
However, experimental observations revealed that the slope of the NCLs (λ(s)) initially
increases and after reaching a peak starts to decrease. In addition, the slope of the NCLs
(λ(s)) for the unsaturated specimens is higher than the saturated sample, which is contra-
dictory to the assumption made in the BBM.
• The model predicted decreasing initial specific volumes (N(s)) with increasing suctions.
The initial specific volumes of the unsaturated samples were predicted to be lower than
the saturated samples. However, similar to the slope of the NCLs (λ(s)), the experimental
measurements revealed that the initial specific volumes (N(s)) of the unsaturated samples
are higher than the saturated one.
• To overcome the model deficiencies stated above, the modelling parameters for the slope
of the NCL (λ(0)) and the initial specific volume (N(0)) for the saturated soil sample were
increased. This led to a more compressible soil, which is consistent with the experimen-
tal results. After introducing these modifications, the model predictions are in a better
agreement with the experimental results.
• In the compaction process modelling, the suction change inclusion with loading was not
considered for simplicity. The inclusion of the suction change with loading into the model
would result in better model predictions after the mid water contents. However, it would
not change the results in the dry part of the compaction curve because of the low suction
change in relatively dry soil samples with loading. The inclusion of suction into the BBM
could be done by incorporating a SWRC relationship as suggested by Gallipoli et al.
(2003b). In addition, for a wide range of water contents, it is hard to determine the model
parameters for the SWRC with the change in void ratio with the same accuracy level.
Gallipoli et al. (2003b) evaluated the model for Speswhite kaolin where the maximum
145
applied matric suction was only 300 kPa. This model was also used by Tarantino and
De Col (2008) for the same soil, but wider range of water contents. They determined the
model parameters for matric suctions up to 1200 kPa. However, it can be clearly noticed
that if the suction range is further increased the same model parameters cannot be used.
• After the modification, the BBM can predict the compaction curves for granular and
fine soils in a narrow range of water contents with single set of soil parameters. The
correct predictions for the whole compaction curves would require an additional set of soil
parameters to be used.
• As can be seen from the modified predictions of the slope of the NCLs (λ(s)) and the
initial specific volume (N(s)), the transition zone between the saturated and unsaturated
soils cannot be well captured by the BBM.
7.3.5 Other models developed with separate stress variables approach
Other models developed with separate stress variables approach are mainly the modified versions
of the constitutive model proposed by Alonso et al. (1990). These models and their overall effect
on the prediction of the compaction curve is briefly discussed in this section.
Wheeler and Sivakumar (1995) developed an elasto-plastic critical state framework for un-
saturated soils by modifying some of the assumptions in the BBM. They abandoned the use
of a reference stress (pc) at which the change of suction only produces elastic volume deforma-
tions (a yield curve is a straight vertical line in suction and stress plot). Atmospheric pressure
was taken as a reference pressure in their volume change equation. This might result in some
problems when calculating the volume changes at stresses lower than the atmospheric pressure.
In addition, they used the empirical values for the slope of the NCLs (λ(s)) and the initial
specific volumes (N(s)) of unsaturated soil samples rather than introducing an equation for the
variation of the slope of the NCLs with the change in suction (see Equation 2.13). However,
the use of empirical values for these parameters is not practical and very time consuming when
trying to model a soil behaviour for wide range of water contents. In fact, model’s purpose is
to reduce the amount of experiments in order to predict the behaviour of soils. Wheeler and
Sivakumar (1995) noticed in their experimental results that the change in the slope of the NCLs
(λ(s)) and the initial specific volume (N(s)) vary differently (the slope of the NCLs increases
with increasing suction) than it was proposed in the BBM; however, they did not developed a
146 CHAPTER 7. PERFORMANCE OF THE MODEL
complete mathematical model for this behaviour.
Vaunat et al. (2000) and Thu et al. (2007) slightly modified the BBM and incorporated SWRC
into the model. However, the same assumption for the slope of the NCLs (λ(s)) as proposed
in the BBM was used. Therefore, this will lead to a similar predictions for the compaction
behaviour as it is in the BBM.
Chiu and Ng (2003) incorporated SWRC into the framework proposed by Wheeler and
Sivakumar (1995). For the variation of the slope of the NCLs (λ(s)), they used a similar approach
proposed in the BBM (see Equation 2.13). Different from the BBM, they considered that the ”r”
parameter is bigger than one in order to fit their experimental data better. This modification
was also proposed by Wheeler et al. (2002) in order to capture the increase in compressibility
with an increase in suction. This is a reasonable modification and well captured the behaviour
of tested soil (Speswhite kaolin) in their studies because the soil compressibility increases with
the increase in suction. However, it should be noted that the examples they presented were in a
narrow range of suctions and do not exceed 300 kPa of suction. When wider range of suctions or
water contents are considered, it was shown that the compressibility initially increases and after
reaching a peak value starts to decrease (see Figures 7.13 and 7.17). Therefore, this is consid-
ered to be a worthwhile modification up to the suctions where the peak in the compressibility is
reached.
7.4 Combined stress variables approach
In this section, the models developed with combined stress variables approach are examined on
their ability to predict the static compaction curve. The performance of the model proposed by
Sheng et al. (2008) and named as SFG model by the developers is evaluated in detail. There
are other combined stress variables approach models proposed in literature (Bolzon et al., 1996;
Gallipoli et al., 2003a; Tarantino and De Col, 2008). However, those models use the BBM or
models with similar approach (Wheeler and Sivakumar, 1995) in their mechanical modelling
formulations and only use different work conjugates as suggested by Houlsby (1997). Therefore,
the SFG model is selected in this study for detailed evaluation. The application of the SFG
model on the development of the compaction process is presented in Chapter 6. In this section,
only the results of the model implementations are presented.
The SFG model is used for predicting the compaction behaviour of four different soils, which
147
Table 7.9: Model parameters for S2B (SFG model)
N0 λvp (σy − ua)0 ssa(kPa) (kPa)
1.6866 0.0079 10 2
are S2B, S5B, Boom clay and Speswhite kaolin. The performance of the SFG model on pre-
dicting the static compaction curves is evaluated for each soil type. Model parameters that are
necessary for implementations were determined using the experimental data. Subsequently, the
static compaction curves were predicted using the model parameters; and the modelling results
were compared with the experimental results. The experiments were conducted using semi-
drained loading condition but the implementations of undrained loading are also presented for
comparison. Suction was considered to stay constant during loading for constant water content
tests in the model implementations. Implications of assuming constant suction throughout load-
ing are discussed later in this section. Other models developed using combined stress variables
approach are also evaluated briefly in this section. Finally, the advantages and shortcomings of
the models are discussed.
7.4.1 Sand bentonite mixtures
The performance of the compaction model developed using the constitutive model proposed by
Sheng et al. (2008) is evaluated in this section for sand-bentonite mixtures. Initially, the results
for S2B are presented and this is followed with the results for S5B.
The model parameters used for the implementations of S2B are given in Table 7.9. These
parameters were calculated from the experimental results given in Figure 5.8. The experimental
results were derived when both axes are plot in logarithmic scale of the specific volume (ν) and
the applied net vertical stress (σy − ua) instead of a semi logarithmic relationship as it was in
the BBM. This is due to the the derivation of the modelling equations for the SFG model (see
Section 6.2.3).
The experimental values and the model estimates for the initial specific volumes (N(s)) are
given in Figure 7.21. The model predicts lower initial specific volumes (N(s)) with increasing
matric suctions. On the contrary, the experimental results revealed that the initial specific
volume (N(s)) for S2B increases with increasing matric suctions. Thus, the model assumes
initially denser samples compared to the experimental values.
Figure 7.22 shows the model predictions for the variation of the dry density with the applied
148 CHAPTER 7. PERFORMANCE OF THE MODEL
0 2 4 6 81.68
1.7
1.72
1.74
1.76
1.78
1.8
1.82
1.84
Matric suction, s (kPa)
N(s
)
experimentmodel (Eqs. 2.18 and 2.19)
Figure 7.21: N(s) for S2B for varying matric suctions (SFG model)
vertical stress for S2B. Although the experiments were conducted under semi-drained conditions,
the model predictions are presented for both the undrained and semi-drained loading conditions
in Figure 7.22(a). The model predicted slightly lower dry densities under the same applied
vertical stresses for undrained loading condition. This is due to the higher buildup air and
water pressures, which made the compaction harder. The difference is very small due to the low
degrees of saturation attained during the compaction (see Figure 5.12). Figure 7.22(b) shows
the comparison between the experimental values and the model predictions. For all the samples,
the model predicted higher dry densities than the experimental results due to the lower initial
specific volume (N(s)) estimates by the model. In addition, the gradient of the curve predicted
by the model is lower than the gradient of the experimental curve. This means that even if the
model correctly predicted the initial specific volumes (N(s)), it would still fail to predict the
correct dry densities. This model deficiency may result due to two reasons: the underestimation
of the slope of the NCL (λvp) or the over emphasis of the suction in the volume change equation
(Equation 2.17). For S2B, matric suctions are very low due to the granular nature of the soil.
Thus, the lower gradients of the predicted curves are due to the underestimation of the slope of
the NCL for saturated soil sample (λvp). Moreover, the model predicts very similar dry densities
for samples with different water contents, which is different from the experimental observations.
Considering that the matric suctions in S2B samples with different water contents are close to
each other (see Figure 5.1), the model can only depict this experimental observation by adopting
149
Table 7.10: Model parameters for S5B (SFG model)
N0 λvp (σy − ua)0 ssa(kPa) (kPa)
1.6179 0.0132 10 4
different slopes for the NCLs.
Figure 7.23 presents the model predictions of the compaction curves for S2B. The comparison
of the undrained and semi-drained predictions for different water contents are shown in Figure
7.23(a). The model predictions for undrained loading are slightly lower than the predictions for
semi-drained loading conditions, due to the buildup air and water pressure in undrained loading
condition. As stated earlier, the difference is very small due to the low degrees of saturation in the
tested samples. Figure 7.23(b) shows the comparison of the model predictions and experimental
results for different stress levels. The model overestimated the dry densities. As discussed earlier,
this is due to the underestimation of the initial specific volume (N(s)) by the model.
A conference paper was published before (see Appendix B) for the the comparison of the
experimental results and predictions with the SFG model for S2B. In that paper, the model
predictions better depict the experimental results for the compaction curve. However, the slope
of the NCL for saturated soil sample (λvp) was not measured at the time and was chosen to be
a value which would result in a best model prediction. After the measurement of the slope of
the NCL (λvp) which lead to a poor prediction as shown in Figure 7.23(b), it was noticed that
the experimental value is much lower than the previously adopted one. This fact also reveals
that the model underestimates the compressibility or slope of the NCL for the unsaturated soil
samples.
The model parameters used for the implementations of S5B are given in Table 7.10. These
parameters were calculated from the experimental results given in Figure 5.9. The experimental
results were derived from a double logarithmic relationship (when both axes are plot in logarith-
mic scale) instead of a semi logarithmic relationship due to the approach adopted in the SFG
model.
The experimental values and the model estimates for the initial specific volumes (N(s)) are
given in Figure 7.24. Similar to the BBM, the SFG model predicts lower initial specific volumes
(N(s)) with increasing suctions. On the contrary, the experimental results revealed that the
initial specific volume (N(s)) for S5B increases with increasing matric suctions.
Figure 7.25 presents the model predictions for the variation of the dry density with the applied
150 CHAPTER 7. PERFORMANCE OF THE MODEL
[a]
0 500 1000 1500 2000 25001.57
1.58
1.59
1.6
1.61
1.62
1.63
1.64
1.65
1.66
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
wc = 6.9% Uwc = 8.4% Uwc = 8.6% Uwc = 10.7% Uwc = 6.9% Swc = 8.4% Swc = 8.6% Swc = 10.7% S
[b]
0 500 1000 1500 2000 25001.5
1.55
1.6
1.65
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
wc = 6.9% Mwc = 8.4% Mwc = 8.6% Mwc = 10.7% Mwc = 6.9% Ewc = 8.4% Ewc = 8.6% Ewc = 10.7% E
Figure 7.22: SFG model predictions of the variation of dry density with applied vertical stressfor S2B [a] Comparison of the undrained (U) and semi-drained (S) predictions [b] Comparisonof the experimental data (E) and model (M) predictions (semi-drained)
151
[a]
6 7 8 9 10 111.615
1.62
1.625
1.63
1.635
1.64
Water content (%)
Dry
den
sity
(M
g/m
3 )
560 kPa U2250 kPa U560 kPa S2250 kPa S
[b]
6 7 8 9 10 111.55
1.56
1.57
1.58
1.59
1.6
1.61
1.62
1.63
1.64
1.65
Water content (%)
Dry
den
sity
(M
g/m
3 )
560 kPa M2250 kPa M560 kPa E2250 kPa E
Figure 7.23: Compaction curves for S2B predicted by SFG model [a] Comparison of undrained(U) and semi-drained (S) predictions [b] Comparison of the experimental data (E) and model(M) predictions (semi-drained)
152 CHAPTER 7. PERFORMANCE OF THE MODEL
0 10 20 30 40 501.6
1.65
1.7
1.75
Matric suction, s (kPa)
N(s
)
experimentmodel (Eqs. 2.18 and 2.19)
Figure 7.24: N(s) for S5B for varying matric suctions (SFG model)
vertical stress for S5B. Figure 7.25(a) shows the model predictions for both the undrained and
semi-drained loading conditions. The model predicted slightly lower dry densities under the same
applied vertical stresses due to the higher buildup air and water pressures, for undrained loading
condition. However, the difference is very small because of the low degrees of saturation attained
at the end of the compaction (see Figure 5.13). Figure 7.25(b) shows the comparison between
the experimental values and the model predictions. For all the samples, the model predicted
higher dry densities than the experimental results due to the lower initial specific volume (N(s))
estimates by the model. Different than S2B predictions, the model predicted more disperse dry
densities for S5B samples with different contents. This resulted due to the higher variation in
suction between the samples with different water contents.
Figure 7.26 shows the model predictions of the compaction curves for S5B. The comparison
of the undrained and semi-drained predictions are shown in Figure 7.26(a). The model pre-
dicted slightly lower dry densities for undrained loading due to the higher buildup air and water
pressures. However, the difference is very small due to the low degrees of saturation within the
samples (see Figure 5.13). Figure 7.26(b) shows the comparison of the model predictions and
experimental results for different stress levels. The model overestimated the dry densities due
to the underestimation of the initial specific volume (N(s)).
153
[a]
0 500 1000 1500 2000 2500 3000 35001.64
1.66
1.68
1.7
1.72
1.74
1.76
1.78
1.8
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
wc = 6.5% Uwc = 7.5% Uwc = 10.3% Uwc = 10.9% Uwc = 12.6% Uwc = 6.5% Swc = 7.5% Swc = 10.3%Swc = 10.9% Swc = 12.6% S
[b]
0 500 1000 1500 2000 2500 3000 35001.58
1.6
1.62
1.64
1.66
1.68
1.7
1.72
1.74
1.76
1.78
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
wc = 6.5% Mwc = 7.5% Mwc = 10.3% Mwc = 10.9% Mwc = 12.6% Mwc = 6.5% Ewc = 7.5% Ewc = 10.3% Ewc = 10.9% Ewc = 12.6% E
Figure 7.25: SFG model predictions of the variation of dry density with applied vertical stressfor S5B [a] Comparison of the undrained (U) and semi-drained (S) predictions [b] Comparisonof the experimental data (E) and model (M) predictions (semi-drained)
154 CHAPTER 7. PERFORMANCE OF THE MODEL
[a]
6 7 8 9 10 11 12 131.7
1.71
1.72
1.73
1.74
1.75
1.76
1.77
1.78
Water content (%)
Dry
den
sity
(M
g/m
3 )
560 kPa U2650 kPa U560 kPa S2650 kPa S
[b]
6 7 8 9 10 11 12 131.62
1.64
1.66
1.68
1.7
1.72
1.74
1.76
1.78
Water content (%)
Dry
den
sity
(M
g/m
3 )
560 kPa M2650 kPa M560 kPa E2650 kPa E
Figure 7.26: Compaction curves for S5B predicted by SFG model [a] Comparison of undrained(U) and semi-drained (S) predictions [b] Comparison of the experimental data (E) and model(M) predictions (semi-drained)
155
Table 7.11: Model parameters for Boom clay (SFG model)
N0 λvp (σy − ua)0 ssa(kPa) (kPa)
2.8438 0.0993 10 35
10−6
10−4
10−2
100
102
2.2
2.4
2.6
2.8
3
3.2
3.4
3.6
3.8
Suction, s (MPa)
N(s
) experimentmodel (Eqs. 2.18 and 2.19)
Figure 7.27: N(s) for Boom clay for varying suctions (SFG model)
7.4.2 Boom clay
The model parameters used for Boom clay are given in Table 7.11. The parameters are derived
from the experimental results of Deng et al. (2010) which are presented in Figure 7.1. The
tests conducted only on saturated soil samples are adequate for the determination of the model
parameters necessary for the SFG model.
Figure 7.27 shows the variation of the initial specific volume (N(s)) with suction. Both
experimental measurements and model predictions calculated with Equation 2.19 are presented
in the figure. Similar to the sand bentonite mixtures, the SFG model predicts decreasing initial
specific volumes (N(s)) with increasing suction for Boom clay. The experimental measurements,
however, show that the initial specific volume initially increases and after a peak value (at about
3000 kPa suction) starts decreasing with increasing suctions. The experimentally measured
initial specific volumes (N(s)) for unsaturated soil samples have greater values than the saturated
sample.
The model estimates for the variation of dry density with applied vertical stress for Boom
156 CHAPTER 7. PERFORMANCE OF THE MODEL
clay are shown in Figure 7.28. Figure 7.28(a) shows the comparison of the undrained and semi
drained modelling predictions. The difference between the two loading conditions is negligible at
low water contents and only starts to be noticed after 18.7% water content. It was also discussed
in Chapter 6 that air pressure and water pressure start to be influential after the OWC under
undrained loading conditions, first due to the high water content and therefore lower space left
for the air within the soil skeleton and second due to the air being trapped within the water.
Figure 7.28(b) shows the comparison of the model predictions and the experimental results
for the variation of dry density with applied vertical stress. Contrary to the predictions for
sand bentonite mixtures, even though the model estimates that the initial specific volumes
(N(s)) are lower than the experimental ones, the model predicts lower dry densities than the
experimentally measured values for Boom clay. This results due to high influence of suction in
the model derivation. This can be clearly observed comparing the model predictions for low
water content and higher water contents (or samples having high suction with samples having
low suctions). The error between the model predictions and experimental results reduces with
increasing water contents or reducing suctions. The evaluation of the model with sand bentonite
mixtures, which had low suctions compared to Boom clay, revealed the deficiency of the model’s
sensitivity to the slope of the NCL. The evaluation of the model with Boom clay, on the other
hand, clearly reveals the deficiency of the model, which highly emphasises the suctions. Even
the sample with 22.3% water content, which had the lowest suction between the tested samples,
failed to reach the experimentally measured dry densities. Therefore, it seems that the influence
of the suction in the model equation needs to be reduced.
Figure 7.29 shows the model estimates of the compaction curve for different stress levels. The
comparison of undrained and semi-drained model predictions are given in Figure 7.29(a). The
effect of buildup air pressure, which causes in lower predicted dry densities, is only noticeable
after 18.7% water content. In addition, the effect of buildup pore air pressure within the samples
becomes higher with increasing applied vertical stresses.
Figure 7.29(b) presents the comparison of the experimental and the modelling results for the
compaction curves with under different applied vertical stresses. The model highly underesti-
mates the experimental results due to the high dependency of the model equation (Equation
2.17) to suction. This can be clearly noticed at higher water contents (due to suction reduction),
when the error rate between the experimentally measured and predicted values reduces. The
effect of suction on volume change behaviour is so strong that in low water contents (at about
157
[a]
0 2000 4000 6000 8000 100001.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
w = 3% Uw = 9.8% Uw = 14.7% Uw = 18.7% Uw = 22.3% Uw = 3% Sw = 9.8% Sw = 14.7% Sw = 18.7% Sw = 22.3% S
[b]
0 2000 4000 6000 8000 100001.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
Applied vertical stress (kPa)
Dry
den
sity
(M
g/m
3 )
w = 3% Mw = 9.8% Mw = 14.7% Mw = 18.7% Mw = 22.3% Mw = 3% Ew = 9.8% Ew = 14.7% Ew = 18.7% Ew = 22.3% E
Figure 7.28: SFG model predictions of the variation of dry density with applied vertical stress forBoom clay [a] Comparison of the undrained (U) and semi-drained (S) predictions [b] Comparisonof the experimental data (E) and model (M) predictions (semi-drained)
158 CHAPTER 7. PERFORMANCE OF THE MODEL
Table 7.12: Model parameters for Speswhite kaolin (SFG model)
N0 λvp (σy − ua)0 ssa(kPa) (kPa)
2.5385 0.0748 10 30
3% water content) the effect of applied vertical stress is not noticeable. This is due to the higher
suctions of the sample (102 MPa) than the maximum applied vertical stress (10 MPa). The
effect of applied vertical stress on the volume change of the sample starts to be noticed after
9.8% water content, because at this water content the suction value is about 10 MPa which is
close to the maximum applied vertical stress. Therefore, in the SFG model the ratio between
suction and applied vertical stress is important.
7.4.3 Speswhite kaolin
The model parameters for Speswhite kaolin are given in Table 7.12. This parameters are calcu-
lated from the experimental measurements presented in Figure 7.3. The experimental results in
Figure 7.3 are plotted in semi logarithmic graphs. The model parameters for the SFG model,
however, are derived from double logarithmic graphs (when both axes are plot in logarithmic
scale), thus the values differ from the values shown in Figure 7.3.
Figure 7.30 shows the variation of initial specific volumes (N(s)) with suction. Both experi-
mentally measured values and model estimates are presented in the figure. The model predicts
reducing initial specific volumes (N(s)) with increasing suction. Experimentally measured val-
ues, on the other hand, initially increase and after a peak value start to decrease. The model
predicts initially lower specific volumes (N(s)), which means the model assumes that the unsat-
urated samples are initially denser compared to the experimental conditions.
The modelling results for the variation of void ratio with applied vertical stress are shown
in Figure 7.31. Figure 7.31(a) shows the comparison of the undrained and semi-drained model
predictions. Similar to the other soil types examined, the difference between the undrained
and semi-drained loading conditions increases with increasing water content. Figure 7.31(b)
presents the comparison between the experimental results and the model predictions. Since the
model predicted much lower initial void ratios (or initial specific volumes (N(s))), the model
estimates for the void ratio at low applied vertical stresses are much lower than the experimentally
measured values. The difference between the predicted and experimental measurements reduces
with increasing applied vertical stresses. This is explained by the fact that the high influence
159
[a]
0 5 10 15 20 25
1.25
1.3
1.35
1.4
1.45
1.5
1.55
1.6
1.65
1.7
Water content (%)
Dry
den
sity
(M
g/m
3 )
1 MPa U2 MPa U3 MPa U4 MPa U5 MPa U6 MPa U8 MPa U10 MPa U1 MPa S2 MPa S3 MPa S4 MPa S5 MPa S6 MPa S8 MPa S10 MPa S
[b]
0 5 10 15 20 25
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
Water content (%)
Dry
den
sity
(M
g/m
3 )
1 MPa M2 MPa M3 MPa M4 MPa M5 MPa M6 MPa M8 MPa M10 MPa M1 MPa E2 MPa E3 MPa E4 MPa E5 MPa E6 MPa E8 MPa E10 MPa E
Figure 7.29: Compaction curves for Boom clay predicted by SFG model [a] Comparison ofundrained (U) and semi-drained (S) predictions [b] Comparison of the experimental data (E)and model (M) predictions (semi-drained)
160 CHAPTER 7. PERFORMANCE OF THE MODEL
0 1000 2000 3000 4000 50002
2.5
3
3.5
4
4.5
5
Suction, s (kPa)
N(s
)
experimentmodel (Eqs. 2.18 and 2.19)
Figure 7.30: N(s) for Speswhite kaolin for varying suctions (SFG model)
of suction in the model derivation starts to be less influential with increasing applied vertical
stresses. The model’s sensitivity to suction reduces after the applied vertical stress exceeds the
suctions of the samples and the model becomes more sensitive to the applied vertical stresses.
The gradient of the curves in Figure 7.31(b) are much lower for the model estimates compared
to the experimentally measured values. As it was stated before, this again reveals that model
places unreasonable emphasis on suction in its derivations and in return underestimates the
compressibility (slope of the NCLs) of the soil.
Figure 7.32 shows the model predictions of the compaction curve for Speswhite kaolin. The
comparison between undrained and semi-drained modelling predictions are presented in Figure
7.32(a). Similar to the other examined soils (sand bentonite mixtures and Boom clay), the
influence of buildup pore pressure is noticeable at high water contents. The comparison between
the modelling predictions and the experimental results of the compaction curve is shown in
Figure 7.32(b) for different stress levels. The model overestimates the dry densities up to 15%
water content at all stress levels due to initially overestimated dry densities (or underestimated
initial specific volumes (N(s))). After 15% water content, the model still overestimates the
experimental results at low applied stresses, however, predicts better at high applied vertical
stresses. As stated before, this is due to the ratio between the suction and applied vertical stress.
After reaching the applied stresses higher than the initial suction value, the model becomes less
sensitive to suction and more sensitive to the slope of the NCL and the applied stresses.
161
[a]
0 200 400 600 800 1000 12000.98
1
1.02
1.04
1.06
1.08
1.1
1.12
1.14
1.16
Applied vertical stress (kPa)
Voi
d ra
tio (
e) wc = 24% Uwc = 25% Uwc = 28% Uwc = 24% Swc = 25% Swc = 28% S
[b]
0 200 400 600 800 1000 12000.9
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
Applied vertical stress (kPa)
Voi
d ra
tio (
e) wc = 24% Mwc = 25% Mwc = 28% Mwc = 24% Ewc = 25% Ewc = 28% E
Figure 7.31: SFG model predictions of the variation of dry density with applied vertical stressfor Speswhite kaolin [a] Comparison of the undrained (U) and semi-drained (S) predictions [b]Comparison of the experimental data (E) and model (M) predictions (semi-drained)
162 CHAPTER 7. PERFORMANCE OF THE MODEL
[a]
5 10 15 20 25 30 351.22
1.24
1.26
1.28
1.3
1.32
1.34
Water content (%)
Dry
den
sity
(M
g/m
3 )
300 kPa U600 kPa U900 kPa U1200 MPa U300 kPa S600 kPa S900 kPa S1200 kPa S
[b]
5 10 15 20 25 30 350.9
0.95
1
1.05
1.1
1.15
1.2
1.25
1.3
1.35
1.4
Water content (%)
Dry
den
sity
(M
g/m
3 )
300 kPa M600 kPa M900 kPa M1200 MPa M300 kPa E600 kPa E900 kPa E1200 kPa E
Figure 7.32: Compaction curves for Speswhite predicted by SFG model [a] Comparison ofundrained (U) and semi-drained (S) predictions [b] Comparison of the experimental data (E)and model (M) predictions (semi-drained)
163
7.4.4 Discussion on the SFG model
The SFG model was evaluated on its ability to predict the static compaction curve of four
different soils. The following conclusions about the model are gained:
• The model predicts decreasing initial specific volumes (N(s)) with increasing suctions. In
addition, the initial specific volumes for unsaturated samples are predicted to be lower than
the saturated samples. However, the experimental measurements reveal that the initial
specific volumes (N(s)) of unsaturated samples are higher than the saturated samples.
This model deficiency results in overestimation of the dry densities under low applied
vertical stresses.
• Although a direct relationship between the slope of the NCL and suction, similar to the
equation proposed in BBM (see Equation 2.13), is not presented in the model formulation,
the effect of suction in the compressibility of the soil samples is considered in the model.
However, similar to the BBM, in the SFG model, the higher the suction is, the less com-
pressible the soil is (see Equation 2.17). The experimental results, however, reveals the
contrary (see Figures 5.8, 5.9, 7.2 and 7.4).
• The examination of the gradient of the dry density and applied vertical stress relationship
reveals that the model over emphasises the suction effect on the compressibility. The effect
of suction on the compressibility is so high that the samples with high suctions are very
hard to compact, thus the suction effect needs to be reduced. This can be done by adding
a value between zero and one to the power of suction in the model formulation. This is
very similar to multiplying the suction parameter with χ parameter to reduce the effect of
suction as it is adopted in single stress approach models (effective stress models) (Khalili
et al., 2004; Laloui and Nuth, 2009). However, when this is done explicit integration of
the model equation (see Equation 2.16) could not be preformed.
• The examination of the gradient of the dry density and applied vertical stress relationship
also reveals that the slope of the NCL is underestimated. Although, it looks very convenient
to only use the slope of the NCL for saturated soil samples in the model formulations, the
model fails to depict the compressibility of the samples. This was proven in the evaluation
of the sand bentonite mixtures, where the suctions were very low and therefore the low
gradient did not result due to the over emphasis of the suction in the model but only due
164 CHAPTER 7. PERFORMANCE OF THE MODEL
to the low values of the compressibility coefficients. In addition, during the evaluation of
the granular soils (S2B and S5B), it was shown that the adoption of a single slope for the
NCL (λvp), which is for saturated samples in the SFG model, is not adequate and results in
very close predicted dry densities for samples with low suction but different water contents.
Therefore, there might be a need for the adoption of different values for the slope of the
NCLs for samples with different water contents, in place of the slope of the NCL (λvp) for
saturated soil sample.
• The ratio between suction and applied vertical stress is important in the model formula-
tions. When suctions are high the effect of applied stresses are almost considered to be
negligible on the volume change of the samples.
Sheng et al. (2008) evaluated the model for compacted bentonite data presented by Lloret
et al. (2003) and found a reasonable fit to the data. At low suctions, the model was shown to
correctly predict the void ratio of the soil with applied net stresses. At high suctions, the model
underestimates the void ratio (the model predicts less dense samples than the experimental
results). At low suctions, the net stress and suction ratio is high; whereas, at high suctions,
the net stress and suction ratio is lower. This leads to reasonable model predictions at low
suctions. In addition, the model parameter, which is the slope of the NCL (λvp), is assumed to
be much higher (λvp−model = 0.085) than the experimentally measured value (λvp−exp = 0.069)
for the same soil (Lloret et al., 2003), when calculated from a double logarithmic relationship
as defined in the model derivations (when both axes are plot in logarithmic scale). The model
places too much emphasis on suction, which would normally lead to a less compressible soil.
Thus, the use of a higher value for the slope of the NCL (λvp) than the experimentally measured
value, substituted the model’s overemphasis on the suction parameter. Therefore, the model
predictions seems to fit the experimental data better than it would normally fit.
7.4.5 Other models developed with combined stress variables approach
Other models developed with combined stress variables approach use the BBM or models with
similar approach (Wheeler and Sivakumar, 1995) in their mechanical modelling formulations
but use different stress state variables instead of net stress and suction. These models and their
overall effect on the prediction of the compaction curve are briefly discussed in this section.
165
Bolzon et al. (1996) modified the BBM by using average skeleton stress (σ−ua+Sr(ua−uw))
and suction (ua−uw) as stress variables instead of net stress (σ−ua) and suction (ua−uw). This
modification gives advantages over the BBM when the transition zone between unsaturated and
saturated soils is considered. However, this model uses a similar function to the one proposed in
the BBM for the variation of the soil compressibility with suction (λ(s)). Therefore, using this
model will give similar results to the BBM.
Santagiuliana and Schrefler (2006) included hydraulic hysteresis to the model proposed by
Bolzon et al. (1996). However, during the compaction process only wetting path is followed and
there is no need to consider the hysteretic nature of the soil. Thus, this modification would not
improve the compaction process predictions and will give similar results to the BBM.
Gallipoli et al. (2003a) proposed an elasto-plastic constitutive model incorporating the effect
of soil fabric into the model. They used average skeleton stress (σ − ua + Sr(ua − uw)) as
a stress variable and introduced a new stress variable as a second stress variable which is a
function of suction and degree of saturation (ξ = f(s)(1 − Sr)). In their model, they proposed
two soil parameters which can be found by calibrating the model with the experimental results.
Therefore, the initial specific volume (N(s)) and the slope of the NCLs (λ(s)) do not necessarily
reduce with increasing suction but depend on the experimental results and are incorporated
into the model with the proposed soil parameters. In fact, this model is very similar to the
modified BBM suggested in Section 7.3 when ”r” parameter is considered to be bigger than one
(see Equation 2.13). However, the model proposed by Gallipoli et al. (2003a) needs less soil
parameters and will produce better results in the transition zone due to the use of the average
skeleton stress as a stress variable. Gallipoli et al. (2003a) validated their model for Speswhite
kaolin for a very narrow range of suctions. Looking at the experimental results of the soils
examined in this study, it is clear that the compressibility first increases and after reaching a
peak starts decreasing. Thus, still the soil behaviour cannot be modelled using a single set of
soil parameters for a wider range of water contents but will require more sets of parameters.
Tarantino and De Col (2008) developed a compaction model where they used different work
conjugates than the ones used in the BBM as suggested by Houlsby (1997). The work conjugates
they adopted are the average skeleton stress (σ−ua+Sr(ua−uw)) and modified suction (s∗ = ns).
In addition, Tarantino and De Col (2008) considered the change in suction with loading by
incorporating the model proposed by Gallipoli et al. (2003b). Their model is similar to the
model developed by Gallipoli et al. (2003a) and expected to predict similar results.
166 CHAPTER 7. PERFORMANCE OF THE MODEL
7.5 Overall discussion
The evaluation of the compaction models developed using two different constitutive modelling
approaches revealed that both modelling approaches have significant shortcomings in capturing
the volume change behaviour of the compacted soils, except the models prosed by Gallipoli et al.
(2003a) and Tarantino and De Col (2008). The main shortcomings for both approaches are as
follows:
1. The assumption of a decreasing slope of the NCLs (λ(s)) or decreasing compressibility
with increasing suctions; and,
2. The assumption of a decreasing initial specific volume (N(s)) with increasing suctions.
These assumptions are correct when the suction is applied by drying the initially slurry
soil. Slurry soil have a homogeneous fabric compared to compacted soils and drying the soil
would not change the fabric significantly. However, the same suctions could be attained in soils
by two main different ways which would result in completely different fabric within the soil
structure. The stress paths for both type of conditions are presented in Figure 7.33. First,
initially slurry soil can be dried to certain suctions (path AB) and subsequently compacted to
certain stresses (path BC). Second, initially dry soil can be mixed with ceratin amount of water
(path DB’) and after compacted to desired stresses (path B’C’). Although, this two samples
will have similar suctions, they would have completely different volume change behaviour which
is presented in Figure 7.34. Sivakumar and Wheeler (2000) conducted experiments where they
initially compacted the samples with different initial suctions and later saturated them. They
also showed that those two samples have different compaction behaviour.
It was shown in this study that unsaturated loose samples are more compressible than the
saturated samples at the same stress level. The compressibility or the slope of the NCLs (λ(s))
of the unsaturated samples initially increases up to a certain suction and later starts to decrease
with increasing suctions as shown in Figures 7.5, 7.9, 7.13 and 7.17. Most likely, at high applied
vertical stresses, the NCLs for unsaturated samples will join the NCL for saturated sample. This
was experimentally observed by Tarantino and De Col (2008). They initially prepared samples
with different water contents and compacted them under relatively low stresses. Subsequently,
they saturated the samples and loaded them to higher stresses in order to obtain the NCLs. The
samples which were prepared with lower water contents have higher slopes of the NCLs than
167
Net stress, σ −ua
Suc
tion,
s
A
1
C’B’
B C
2
s1
D
Figure 7.33: Stress paths for attaining samples with similar suctions
ln Net stress, ln(σ−ua)
Spe
cific
vol
ume,
ν
C’
D
B’
s = s1
Initially dry soil
A
B C
s = s0
s = s1 Initially slurry soil
Figure 7.34: Volume change behaviour of samples with similar suction but developed underdifferent stress paths
168 CHAPTER 7. PERFORMANCE OF THE MODEL
Suction, s
λ(s)
, N(s
)
2a 2b
set ofparameters set of
parameters
set ofparameters 1
Figure 7.35: Variation of the soil compressibility with suction
the samples prepared with higher water contents. Eventually, all samples joined the NCL for
saturated sample.
The discrepancy observed between the experimental results and the current constitutive
models arises due to the cloddy fabric within the unsaturated samples generalised with initially
dry soils. The current models do not capture the fabric effect of the unsaturated samples
except the models developed by Gallipoli et al. (2003a) and Tarantino and De Col (2008).
However, these models still need to use different sets of soil parameters for the compaction
model development as shown in Figure 7.35. Initially slurry soils require a set of soil parameters
and samples prepared as shown in the stress path 2 (see Figure 7.33) require two different sets of
soil parameters, one for the part where compressibility increases with suction (2a) and another
one for the part where compressibility decreases with increasing suction (2b). Alternatively, the
relationships between the slope of the NCLs (λ(s)) and suction, and between the initial specific
volume (N(s)) and suction need to be captured by a single equation.
In addition, the fabric effect can be easily incorporated to the models formulated using similar
approach to the BBM by adopting the ”r” parameter bigger than one up to the suction level
where the peak value for the slope of the NCL is attained (2a) and after that suction value by
adopting ”r” parameter smaller than one (2b). For initially slurry soils another set of parameters
should be determined by using an ”r” parameter smaller than one (1). This basically highlights
that compacted soils from initially dry soil and initially slurry soil cannot be treated as same.
169
7.6 Concluding remarks
This chapter presented the performance of the compaction models on predicting the static com-
paction behaviour during the compaction process. The models were evaluated with the data of
four different soils. These soils are two sand-bentonite mixtures with different bentonite contents
(S2B and S5B), Boom clay and Speswhite kaolin. The experiments used for the model evalu-
ation were conducted for semi-drained loading conditions. However, the model predictions for
undrained loading conditions were also presented in the chapter to show the comparison between
the two loading conditions. It is found that during undrained loading, lower dry densities were
attained under the same applied vertical stresses. This is due to the higher pore pressure devel-
opment during the undrained loading conditions, which made the compaction process harder (see
Chapter 6). The difference between the predicted dry densities of these two loading conditions
became notable for Boom clay and Speswhite kaolin but not sand-bentonite mixtures. This is
due to the low degrees of saturation attained for S2B and S5B samples during the compaction
process.
Firstly, the performance of using constitutive models developed with separate stress state
variables approach in the compaction process modelling were evaluated. The BBM was chosen for
detailed evaluation and other models developed with a similar approach were briefly examined.
The BBM and the models developed by Vaunat et al. (2000) and Thu et al. (2007) assume
decreasing values with increasing suctions, for the slope of the NCLs (λ(s)) and the initial
specific volumes (N(s)). However, experimental observations revealed that the slope of the
NCLs (λ(s)) initially increases and after reaching a peak value starts to decrease with increasing
suctions, for all the soil types. In addition, the initial specific volumes (N(s)) have a similar
trend to the slope of the NCLs for Boom clay and Speswhite kaolin, and increasing values with
suction for sand-bentonite mixtures. Different than those models, Chiu and Ng (2003) adopted
an increasing slope of the NCLs with increasing suctions by modifying the ”r” parameter to
bigger than one in the BBM. However, this still contradicts with the experimental observations.
Chiu and Ng (2003) would result in reasonable predictions up to the suctions where peak values
for the slope of the NCLs (λ(s)) were attained but would fail to predict the values beyond that
suctions. In the current study, the BBM was modified by increasing the slope of the NCL (λ(0))
and the initial specific volume (N(0)) for saturated soil. This modification led to better model
predictions; however, still the compaction behaviour was not captured using a single set of soil
parameters for a wide range of water contents or suctions.
170 CHAPTER 7. PERFORMANCE OF THE MODEL
Secondly, the performance of using constitutive models developed with combined stress state
variables approach in the compaction process modelling was evaluated. The SFG model was
chosen for detailed evaluation and other models developed with a similar approach were briefly
examined. It is found that the SFG model estimates decreasing values for the initial specific
volumes (N(s)) with increasing suctions. This is due to the over emphasis placed on suction
in the model development. In addition, even a direct relationship for the slope of the NCLs
(λ(s)) is not proposed in this model, it indirectly predicts a decreasing compressibility with
increasing suctions. Both of these assumptions contradict with the experimental observations.
Therefore, the predictions using the SFG model fails to depict the experimental behaviour during
the compaction process. The models developed by Bolzon et al. (1996) and Santagiuliana and
Schrefler (2006) estimates decreasing slope with increasing suctions. Thus, those models would
not correctly predict the compaction behaviour either. The models proposed by Gallipoli et al.
(2003a) and Tarantino and De Col (2008) use similar formulations to the BBM with different
work conjugates for their volume change predictions. However, these models do not have restrains
like decreasing or increasing compressibility with changing suctions. Instead, they calibrate their
models according to the experimental results. Therefore, they can capture both decreasing or
increasing compressibility according to the experimental results. In addition, they use average
skeleton stress (σ− ua +Sr(ua− uw)) as a stress variable in their modelling formulations which
would result in a better transition between the unsaturated and saturated soil states. Thus,
the models developed by Gallipoli et al. (2003a) and Tarantino and De Col (2008) are the
most promising ones in predicting the compaction process. However, using a single set of soil
parameters for the prediction of the compaction behaviour over a wide range of water contents
would not be possible.
In conclusion, it is shown that the samples prepared from initially slurry soils and initially
dry soils could not be treated the same. They would require the adoption of different sets
of soil parameters in order to model their compaction behaviour. In addition, compaction
behaviour of the samples prepared form initially dry soils can be predicted using a single set of
soil parameters only for a narrow range of water contents. Minimum of two sets of soil parameters
are required to model the compaction behaviour over a wide range of water contents with the
existing constitutive models.
Chapter 8
Conclusions and suggestions for
future research
8.1 Conclusions
The objective of this research project was to advance the current understanding of the com-
paction process of soils. The research concentrated on the following areas: investigating the
behaviour of soils during the static compaction process; modelling the compaction process using
the unsaturated soil mechanics principles; and evaluating the performance of the model. The
conclusions gained from the study are presented in this chapter.
8.1.1 Soil behaviour during the compaction process
On the experimental side, the main contribution of this work is the observation of the stress
and phase variables during the static compaction process. Constant water content tests were
conducted on two different granular soils, sand with 2% and 5% bentonite content, with varying
water contents to observe the effect of matric suction on the compaction behaviour.
It is found that matric suction decreases with loading for all tested samples in the present
research program. The matric suction change increases with reducing water content and increas-
ing clay content. For the matric suction range tested, the slopes of the NCLs for unsaturated
samples are found to be higher than the saturated samples. The compressibility coefficient (ms1)
decreases with both increasing vertical net stress (σy − ua) and degree of saturation (Sr) during
171
172 CHAPTER 8. CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH
the tests.
Complete saturation was never attained during the compaction tests for both samples. On
the contrary, the degrees of saturation reached at the end of the compaction process were low
and similar to their values at their saturation matric suctions in the wetting path of the water
retention curves. This is attributed to the correlation between the degree of saturation at
optimum water content and the degree of saturation achieved at zero matric suction in the
wetting path of the water retention curve.
Matric suction was measured using null type axis translation technique for granular soils
for the first time in literature. It is found that this method is not practical for the matric
suction measurements at low degrees of saturation or water contents. This is attributed to the
discontinues water phase within the soil samples.
8.1.2 Modelling the compaction process
On the modelling part, one dimensional static compaction process is modelled for undrained,
semi drained and drained loading conditions. Model is developed by incorporating pore pres-
sure predictions into a volume change constitutive relationship for unsaturated soils. The pore
pressure generation was calculated using two approaches, Hilf (1948) and Hasan & Fredlund
(1980). The first approach is adopted for the pore pressure calculations because it requires less
parameters for the predictions of the pore pressure with applied vertical stress (σy).
It is found that by using constant compressibility coefficients (ms1,m
s2) the shape of the
compaction curve cannot be produced on the the dry side but only on the wet side of the
compaction curve. Thus, variable compressibility coefficients were derived from constitutive
models proposed in the literature, and using these coefficients, the well-known shape of the
compaction curve can be modelled both on the dry and wet side of the compaction curve using
variable compressibility coefficient (ms1). Moreover, it is found that the matric suction component
of the volume change equation (Equation 2.2) does not have a significant influence on the shape
of the compaction curve but the influence of matric suction on the compressibility coefficient
due to net stress (ms1) is a governing parameter.
It is shown that the dry density is a function of the compressibility coefficient (ms1) and the
change in vertical net stress (∆(σy − ua)). During undrained and semi drained loading, while
coefficient of compressibility (ms1) increases, vertical net stress (σy−ua) decreases with increasing
water contents (w). This happens due to the increase in pore-air pressure (ua) development
173
with increasing water contents (w). Therefore, the maximum dry density is attained at the
optimum combination of the compressibility coefficient (ms1) and the change in vertical net
stress ∆(σy − ua)). The optimum combination of the coefficient of compressibility (ms1) and
vertical net stress (σy − ua) takes place in lower water contents for the soil samples compacted
with higher energy than the soil samples compacted with lower energy. This results in the shift of
the OWC to the left. In addition, during drained loading, while the coefficient of compressibility
(ms1) increases, vertical net stress (σy − ua) stays constant with increasing water content (w).
For that reason, during drained loading the OWC is on the full saturation line.
8.1.3 Evaluation of the model
On the model’s performance part, the compaction models developed with two different con-
stitutive modelling approaches, namely separate stress variables and combined stress variables
approaches, are evaluated on their ability to capture the compaction behaviour of soils.
For the models developed with separate stress variables approach the following conclusions
are derived:
• The BBM, and the models developed by Vaunat et al. (2000) and Thu et al. (2007) assumed
decreasing values for the slope of the NCLs (λ(s)) and the initial specific volumes (N(s))
with increasing suctions. However, experimental observations show that the slope of the
NCLs (λ(s)) initially increases and after reaching a peak value starts to decrease with
increasing suctions, for the tested soil types. In addition, the initial specific volumes (N(s))
have a similar trend to the slope of the NCLs for Boom clay and Speswhite kaolin, but
increasing values with increasing matric suction for sand-bentonite mixtures. Therefore,
using these two constitutive models for compaction process modelling would fail to predict
the volume change behaviour of the soils.
• Chiu and Ng (2003) adopted an increasing slope for the NCLs with increasing suctions.
However, this still contradicts with the experimental observations. Chiu and Ng (2003)
would result in reasonable predictions up to the suction values where the slope of the NCLs
(λ(s)) starts decreasing but would fail to predict the values beyond those suctions.
• Those models can be improved by:
1. modifying the ”r” parameter to values bigger than one up to the suction values where
peak compressibility is observed; and,
174 CHAPTER 8. CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH
2. adopting an ”r” parameter smaller than one, and adopting higher values for the slope
of the NCL (λ(0)) and the initial specific volume (N(0)) for saturated soils.
For the models developed with combined stress variables approach the following conclusions
are obtained:
• The SFG model, and the models developed by Bolzon et al. (1996) and Santagiuliana and
Schrefler (2006) estimate decreasing values for the slope of the NCLs (λ(s)) and initial
specific volumes (N(s)) with increasing matric suctions. Therefore, those models would
not predict the compaction behaviour of soils correctly.
• The models developed by Gallipoli et al. (2003a) and Tarantino and De Col (2008) use
similar formulations for their volume change calculations but different stress variables with
the BBM. However, both of these models do not have restrains like decreasing or increasing
compressibility with changing matric suctions but they calibrate their models according to
the experimental data. Therefore, they can have either decreasing or increasing compress-
ibility according to the experimental results. Thus, they are the most promising ones in
predicting the compaction process. However, they would still require more than one set of
soil parameters to model the compaction behaviour for the wide range of water contents.
• The SFG, and the models developed by Bolzon et al. (1996) and Santagiuliana and Schrefler
(2006) can be improved with the modifications suggested below:
1. The SFG model over emphasises the matric suction in the model formulations. The
effect of matric suction can be reduced by adding a value between zero and one to the
power of matric suction in the model formulation. This is very similar to multiplying
the matric suction parameter with χ parameter to reduce the effect of matric suction
as it is adopted in single stress approach models (effective stress models) (Khalili
et al., 2004; Laloui and Nuth, 2009). However, after this modification the explicit
integration of the model equation (see Equation 2.16) could not be preformed.
2. The models developed by Bolzon et al. (1996) and Santagiuliana and Schrefler (2006)
can be modified similar suggestions proposed above for the models developed with a
separate stress variables approach.
In conclusion, it is shown that the samples prepared to different water contents from initially
slurry soils and initially dry soils could not be treated the same. They would require the adoption
175
of different sets of soil parameters in order to model their compaction behaviour for a wide range
of water contents. In addition, compaction behaviour of the samples prepared from initially
dry soils can be predicted using minimum two sets of soil parameters for a wide range of water
contents. In order to model the compaction behaviour using single set of soil parameters for
samples prepared from initially dry soils, the relationships between the slope of the NCLs (λ(s))
and suction, and between the initial specific volume (N(s)) and suction need to be captured by
a single equation.
8.2 Suggestions for future research
The research conducted in this study significantly contributed to the current knowledge on the
compaction process and therefore the compaction behaviour of soils. Future research, however, is
needed to further extend the understanding of the compaction behaviour. This can be achieved
by improving the experimental techniques and the constitutive models suggested for unsaturated
soils.
On the experimental side, the following work is suggested:
1. It is found that for granular soils, there are some difficulties in suction measurements for
the samples with low water contents using the null type axis translation technique. The
validity of this technique for soils with low water content or degrees of saturation should
be further investigated by using different soil types and conducting more tests. The mea-
surements should be compared by the measurements obtained using different techniques.
The fundamental reasons for the long duration of the tests need to be investigated and
clearly identified.
2. It is found that there is a correlation between the saturations at the OWC and the sat-
urations at zero suctions in the wetting path of the soil water retention curve for tested
soils. This evidence should be further investigated for different types of soils in order to
generalise this finding.
3. Modelling the compaction process will provide better understanding of the post compaction
behaviour of soils. Thus, further experiments using different types of soils should be
conducted to provide the link between the compaction process and the post compaction
behaviour of soils.
176 CHAPTER 8. CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH
On the modelling side the following work is suggested:
1. It is shown in this study that the current constitutive models are unable to predict the
compaction behaviour of soils using single set of soil parameters for a wide range of water
contents and a minimum of three sets of parameters are required; one for the samples
prepared at different water contents from initially slurry soil and two for the samples
prepared from initially dry soils. This can be improved by proposing a single equation for
the relationship between the slope of the NCLs (λ(s)) and suction, and incorporating this
equation into the current existing constitutive models.
2. It has been inferred that the compaction process produce the basis for the post compaction
behaviour. Therefore, it is believed that the yield behaviour of soils (e.g., yield curves)
could be correlated to the compaction curves at differen stress levels.
References
Agus, S. S. and Schanz, T. (2005). Comparison of four methods for measuring total suction.
Vadose Zone Journal, 4(4):1087.
Alonso, E. E., Gens, A., and Hight, D. W. (1987). Special problem soils. General report. PROC
IX ECSMFE, Dublin, Ireland. Balkema, Rotterdam, pages 1087–1146.
Alonso, E. E., Gens, A., and Josa, A. (1990). A constitutive model for partially saturated soils.
Geotechnique, 40(3):405–430.
Ameta, N. K. and Wayal, A. S. (2008). Effect of bentonite on permeability of dune sand.
Electronic Journal of Geotechnical Engineering, 13:1–7.
Australian Standards 1289.3.1.1 (1995). Determination of the liquid limit of a soil - Four point
Casagrande method. Australian Standards.
Australian Standards 1289.3.2.1 (1995). Determination of the plastic limit of a soil - Standard
method. Australian Standards.
Australian Standards 1289.3.4.1 (1995). Determination of the linear shrinkage of a soil - Standard
method. Australian Standards.
Australian Standards 1289.3.6.1 (1995). Methods of testing soils for engineering purposes - Soil
classification tests - Determination of the particle size distribution of a soil - Standard method
of analysis by sieving. Australian Standards.
Australian Standards 1289.5.1.1 (2003). Soil compaction and density tests - Determination of the
dry density/moisture content relation of a soil using standard compactive effort. Australian
Standards.
Australian Standards 1289.5.2.1 (2003). Soil compaction and density tests - Determination of
the dry density/moisture content relation of a soil using modified compactive effort. Australian
Standards.
Baker, R. and Frydman, S. (2009). Unsaturated soil mechanics: critical review of physical
foundations. Engineering Geology, 106(1-2):26–39.
Barden, L. and Sides, G. R. (1970). Engineering behavior and structure of compacted clay.
Journal of the Soil Mechanics and Foundations Division, 96(4):1171–1200.
Bishop, A. W. (1959). The principle of effective stress. Teknisk Ukeblad, 106(39):859–863.
Bishop, A. W., Alpan, I., Blight, G. E., and Donald, I. B. (1960). Factors controlling the strength
of partly saturated cohesive soils. In Proceedings of the ASCE Conference on Shear Strength
of Cohesive Soils, Boulder, Colorado, USA, 1960.
177
178 REFERENCES
Bishop, A. W. and Blight, G. E. (1963). Some aspects of effective stress in saturated and partially
saturated soils. Geotechnique, 13(3):177–197.
Bishop, A. W. and Donald, I. B. (1961). The experimental study of partly saturated soil in the
triaxial apparatus. In Proceedings of the 5th International Conference on Soil Mechanics and
Foundation Engineering, volume 1, pages 13–21, Paris, France, 1961.
Blight, G. E. (1967). Effective stress evaluation for unsaturated soils. Journal of the Soil
Mechanics and Foundations Division, Proc. ASCE 93 SM2 (1967):125–146.
Bocking, K. and Fredlund, D. G. (1980). Limitations of the axis translation technique. In
Proceedings of the 4th International Conference on Expansive Soils, pages 117–135, Denver,
Colorado, USA, June, 1980.
Bolzon, G., Schrefler, B. A., and Zeinkiewicz, O. C. (1996). Elastoplastic soil constitutive laws
generalized to parially saturated states. Geotechnique, 46(2):279–289.
Boso, M., Romero, E., and Tarantino, A. (2003). The use of different measurement techniques
to determine water retention curves. In Proceedings of International Conference on Mechanics
of Unsaturated Soils, volume 1, pages 169–181, Weimar, Germany, 2003.
Boso, M., Tarantino, A., and Mongiovı, L. (2004). Shear strength behaviour of a reconstituted
clayey silt. In Unsaturated Soils-Advances in Testing, Modelling and Engineering Applications:
Proceedings of the Second International Workshop on Unsaturated Soils, page 1, Anacapri,
Italy, June, 2004.
Bouazza, A., Impe, W. F., and Haegeman, W. (1996). Some mechanical properties of reconsti-
tuted Boom clay. Geotechnical and Geological Engineering, 14(4):341–352.
Brown, R. W. C. and James, M. (1976). A screen-caged thermocouple psychrometer and cal-
ibration chamber for measurements of plant and soil water potential. Agronomy Journal,
72(5):851.
Brown, R. W. J. and Robert, S. (1980). Extended field use of screen-covered thermocouple
psychrometers1. Agronomy Journal, 68(6):995.
Brye, K. R. (2003). Long-term effects of cultivation on particle size and water-retention charac-
teristics determined using wetting curves. Soil science, 168(7):459.
Bulut, R., Lytton, R. L., and Wray, W. K. (2001). Soil suction measurements by filter paper.
In Vipulanandan, C., Addison, M. B., and Hasen, M., editors, ASCE Geotechnical Special
Publication No. 115, pages 243–261, Houston, Texas, USA, 2001. 3.
Campbell, G. S. (1979). Improved thermocouple psychrometers for measurement of soil water
potential in a temperature gradient. Journal of Physics E: Scientific Instruments, 12:739.
Cardoso, R., Romero, E., Lima, A., and Ferrari, A. (2007). A comparative study of soil suction
measurement using two different high-range psychrometers. Experimental Unsaturated Soil
Mechanics, pages 79–93.
Catana, C. M. (2006). Compaction and water retention characteristics of Champlain Sea Clay.
PhD thesis, University of Ottawa.
Cavalcante Rocha, J. and Didier, G. (1993). Laboratory permeability tests: evaluation of hy-
draulic conductivity of sand liners. In Fourth International Landfill Symposium, pages 299–304,
Sardinia, 1993.
179
Chahal, R. S. and Yong, R. N. (1965). Validity of the soil water characteristics determined with
the pressurized apparatus. Soil Science, 99(2):98.
Chakraborty, S. (2009). Numerical modeling for long term performance of soil-bentonite cut-off
walls in unsaturated soil zone. Master’s thesis, Louisiana State University.
Chalermyanont, T. and Arrykul, S. (2005). Compacted sand-bentonite mixtures for hydraulic
containment liners. Songklanakarin Journal of Science and Technology, 27(2):313–323.
Chao, K. C. (2007). Design principles for foundations on expansive soils. PhD thesis, Colorado
State University.
Chiu, C. F. and Ng, C. W. W. (2003). A state-dependent elasto-plastic model for saturated and
unsaturated soils. Geotechnique, 53(9):809–830.
Cho, G. C. and Santamarina, J. C. (2001). Unsaturated particulate materials-particle-level
studies. Journal of Geotechnical and Geoenvironmental Engineering, 127(1):84.
Chung, H. I. (2006). Dewatering and decontamination of artificially contaminated sediments
during electrokinetic sedimentation and remediation processes. KSCE Journal of Civil Engi-
neering, 10(3):181–187.
Coleman, J. D. (1962). Stress/strain relations for partly saturated soils. Geotechnique, 12(4):348–
350.
Cui, Y. and Delage, P. (1996). Yielding and plastic behaviour of an unsaturated compacted silt.
Geotechnique, 46(2):291–311.
Cui, Y., Tang, A., Mantho, A. T., and De Laure, E. (2008). Monitoring field soil suction using
a miniature tensiometer. Geotechnical Testing Journal, 31(1):95.
Cuisinier, O. and Masrouri, F. (2005). Hydromechanical behaviour of a compacted swelling soil
over a wide suction range. Engineering Geology, 81:204–212.
Cunningham, M. R., Ridley, A. M., Dineen, K., and Burland, J. B. (2003). The mechanical
behaviour of a reconstituted unsaturated silty clay. Geotechnique, 53(2):183–194.
Darcy, H. (1856). Les fontaines publiques de la ville de Dijon. Dalmont, Paris, -:647.
Delage, P., Audiguier, M., Cui, Y. J., and Howat, M. D. (1996). Microstructure of a compacted
silt. Canadian Geotechnical Journal, 33(1):150–158.
Delage, P. and Graham, J. (1995). Mechanical behaviour of unsaturated soils: understanding the
behaviour of unsaturated soils requires reliable conceptual models. In Proceedings of the 1st
International Conference on Unsaturated Soils, pages 1223–1256, Paris, France, September,
1995.
Delage, P., Le, T. T., Tang, A. M., Cui, Y. J., and Li, X. L. (2007). Suction effects in deep
Boom clay block samples. Geotechnique, 57(2):239–244.
Delage, P., Romero, E. E., and Tarantino, A. (2008). Recent developments in the techniques
of controlling and measuring suction in unsaturated soils. In Unsaturated Soils. Advances
in Geo-Engineering: Proceedings of the 1st European Conference, E-UNSAT, pages 33–52,
Durham, United Kingdom, July, 2008.
180 REFERENCES
Deng, Y. F., Tang, A. M., Cui, Y. J., Nguyen, X. P., Li, X. L., and Wouters, L. (2010).
Laboratory hydro-mechanical characterisation of Boom clay at Essen and Mol. In Clays in
natural & engineered barriers for radioactive waste confinement: 4th International Meeting,
Nantes, France, 2010.
Dimos, A. (1991). Measurement of soil suction using transistor psychrometer. Technical report,
Internal Report IR/91–3, Special Research Section, Materials Technology Department, Vic
Roads.
Edil, T. B. and Motan, S. E. (1984). Laboratory evaluation of soil suction components. Geotech-
nical Testing Journal, ASTM, 7(4):173–181.
Estabragh, A. R., Javadi, A. A., and Boot, J. C. (2004). Effect of compaction pressure on
consolidation behaviour of unsaturated soil. Canadian Geotechnical Journal, 41:540–550.
Evans, J. and Ryan, C. (2005). Time-dependent strength behavior of soil-bentonite slurry wall
backfill. In Waste Containment and Remediation: Proceedings of the Geo-Frontiers 2005
Congress, Geotechnical Special Publication, number 142, January, 2005.
Foster, W. R., Savins, J. G., and Waite, J. M. (1955). Lattice expansion and rheological behavior
relationships in water-montmorillonite systems. Clays and Clay Minerals, 395:296–316.
Fredlund, D. G. (1979). Second Canadian Geotechnical Colloquium: Appropriate concepts and
technology for unsaturated soils. Canadian Geotechnical Journal, 16(1):121–139.
Fredlund, D. G. and Morgenstern, N. R. (1976). Constitutive relations for volume change in
unsaturated soils. Canadian Geotechnical Journal, 13(3):261–276.
Fredlund, D. G. and Morgenstern, N. R. (1977). Stress state variables for unsaturated soils.
Journal of the Geotechnical Engineering Division, 103(5):447–466.
Fredlund, D. G. and Rahardjo, H. (1993). Soil mechanics for unsaturated soils. John Wiley &
Sons.
Fredlund, D. G. and Vanapalli, S. K. (2002). Methods of soil analysis, Part 4 - Physial methods,
chapter Shear strength of unsaturated soils, pages 324–360. Soil Science Society of America,
Book Series 5.
Frost, J. D. and Park, J. Y. (2003). A critical assessment of the moist tamping technique. ASTM
Geotechnical Testing Journal, 26(1):57–70.
Gallipoli, D., Gens, A., Sharma, R., and Vaunat, J. (2003a). An elasto-plastic model for un-
saturated soil incorporating the effects of suction and degree of saturation on mechanical
behaviour. Geotechnique., 53(1):123–136.
Gallipoli, D., Wheeler, S. J., and Karstunen, M. (2003b). Modelling the variation of degree of
saturation in a deformable unsaturated soil. Geotechnique, 53(1):105–112.
Garga, V. K. and Zhang, H. (1997). Volume changes in undrained triaxial tests on sands.
Canadian Geotechnical Journal, 34(5):762–772.
Gee, G. W., Campbell, M. D., Campbell, G. S., and Campbell, J. H. (1992). Rapid measurement
of low soil water potentials using a water activity meter. Soil Science Society of America
Journal, 56(4).
181
Gilbert, O. H. (1959). The influence of negative pore water pressures on the strength of com-
pacted clays. Master’s thesis, Massachusetts Institute of Technology.
Guan, Y. and Fredlund, D. G. (1997). Use of the tensile strength of water for the direct
measurement of high soil suction. Canadian Geotechnical Journal, 34(4):604–614.
Hasan, J. U. and Fredlund, D. G. (1980). Pore pressure parameters for unsaturated soils.
Canadian Geotechnical Journal, 17(3):395–404.
He, L., Leong, E. C., and Elgamal, A. (2006). A miniature tensiometer for measurement of high
matric suction. Geotechnical Special Publication, 2(147):1897–1907.
Henderson, S. J. and Speedy, R. J. (1980). A Berthelot-Bourdon tube method for studying water
under tension. Journal of Physics E: Scientific Instruments, 13:778.
Hilf, J. W. (1948). Estimating construction pore pressures in rolled earth dams. In Proceedings of
the 2nd International Conference on Soil Mechanics and Foundation Engineering, volume 3,
pages 234–240, Rotterdam, The Netherlands, 1948.
Hilf, J. W. (1956). An investigation of pore water pressure in compacted soils. Bureau of
Reclamation, Tech. Mem. 654.
Hilf, J. W. (1975). Foundation Engineering Handbook, chapter Compacted fill, pages 249–316.
Chapman & Hall Ltd., London, United Kingdom.
Hogentogler, C. A. (1936). Essentials of soil compaction. In Proceedings of the Highway Research
Board, pages 309–316.
Horn, H. M. (1960). An investigation of the frictional characteristics of minerals. PhD thesis,
University of Illinois, Urbana, Illinois.
Houlsby, G. T. (1997). The work input to an unsaturated granular material. Geotechnique,
47(1):193–196.
Houston, S. L., Houston, W. N., and Wagner, A. M. (1994). Laboratory filter paper suction
measurements. ASTM Geotechnical Testing Journal, 17(2):185–194.
Huat, B. B. K., Ali, F. H. J., and Choong, F. H. (2006). Effect of loading rate on the vol-
ume change behavior of unsaturated residual soil. Geotechnical and Geological Engineering,
24:1527–1544.
Jennings, J. E. B. and Burland, J. B. (1962). Limitations to the use of effective stresses in partly
saturated soils. Geotechnique, 12(2):125–144.
Khalili, N., Geiser, F., and Blight, G. E. (2004). Effective stress in unsaturated soils: review
with new evidence. Internatioanal Journal of Geomechanics, 4(2):115–126.
Komine, H. and Ogata, N. (1999). Experimental study on swelling characteristics of sand-
bentonite mixture for nuclear waste disposal. Journal of the Japanese Geotechnical Society:
Soils and Foundation, 39(2):83–97.
Kuerbis, R. and Vaid, Y. P. (1988). Sand sample preparation: the slurry deposition method.
Soils and Foundations, 28(4):107–118.
Kunze, R. J. K. and Kirkham, D. (1962). Simplified accounting for membrane impedance in
capillary conductivity determinations. Soil Science Society of America Journal, 26(5):421.
182 REFERENCES
Laloui, L. and Nuth, M. (2009). On the use of the generalised effective stress in the constitutive
modelling of unsaturated soils. Computers and Geotechnics, 36(1-2):20–23.
Lambe, T. W. (1951). Soil testing for engineers. Soil Science, 72(5):406.
Lambe, T. W. (1958). The structure of compacted clay. Journal of the Soil Mechanics and
Foundations Division, ASCE, 84:1–34.
Langfelder, L. J., Chen, C. F., and Justice, J. A. (1968). Air permeability of compacted cohesive
soils. Journal of the Soil Mechanics and Foundations Division, 94(4):981–1002.
Lauer, C. (2005). Investigations on the unsaturated stress-strain behaviour and on the SWCC of
Hostun Sand in a double-walled triaxial cell. In Advanced Experimental Unsaturated Soil Me-
chanics: Proceedings of the International Symposium on Advanced Experimental Unsaturated
Soil Mechanics, page 185, Trento, Italy, June 2005.
Leong, E. C., Tripathy, S., and Rahardjo, H. (2003). Total suction measurement of unsaturated
soils with a device using the chilled-mirror dew-point technique. Geotechnique, 53(2):173–182.
Li, X. and Zhang, L. (2007). Prediction of SWCC for coarse soils considering pore size changes.
Spinger Proceedings in Physics, 112:401–412.
Li, Z. M. (1995). Compressibility and collapsibility of compacted unsaturated loessial soils. In
Proceedings of the 1st International Conference on Unsaturated Soils, pages 139–144, Paris,
France, September, 1995.
Lloret, A., Villar, M. V., Sanchez, M., Gens, A., Pintado, X., and Alonso, E. E. (2003). Me-
chanical behaviour of heavily compacted bentonite under high suction changes. Geotechnique,
53(1):27–40.
Loret, B. and Khalili, N. (2002). An effective stress elastic-plastic model for unsaturated porous
media. Mechanics of Materials, 34(2):97–116.
Lourenco, S. D. N., Gallipoli, D., Toll, D. G., and Evans, F. (2006). Development of a commercial
tensiometer for triaxial testing of unsaturated soils. Geotechnical Special Publication - ASCE,
2(147):1875–1886.
Mahler, C. F. and Diene, A. A. (2007). Tensiometer development for high suction analysis in
laboratory lysimeters. Experimental Unsaturated Soil Mechanics, 112:103–115.
Marinho, F. A. M. and Chandler, R. J. (1995). Cavitation and the direct measurement of soil
suction. In Proceedings of 1st International Conference on Unsaturated Soils, volume 1, pages
623–630, Paris, France, September, 1995.
Marinho, F. A. M., Take, W. A., and Tarantino, A. (2009). Measurement of matric suction using
tensiometric and axis translation techniques. Laboratory and Field Testing of Unsaturated
Soils, pages 3–19.
Martin, R. T. (1959). Water-vapor sorption on kaolinite: Hysteresis. Clays and Clay Minerals,
6:259–277.
Matyas, E. L. and Radhakrishna, H. S. (1968). Volume change characteristics of partially satu-
rated soils. Geotechnique, 18(4).
183
Mavroulidou, M., Zhang, X., Cabarkapa, Z., and Gunn, M. J. (2009). A study of the laboratory
measurement of the soil water retention curve. In Proceedings of the 11th International Con-
ference on Environmental Science and Technology, CEST, volume A, pages 907–915, Chania,
Crete, Greece, September, 2009.
Meeuwig, R. O. (1972). A low-cost thermocouple psychrometer recording system. In Psychrom-
etry in water relations research: Proceedings of the Symposium on Thermocouple Psychrome-
ters, page 131, Utah State University, USA, March, 1971.
Meilani, I., Rahardjo, H., Leong, E. C., and Fredlund, D. G. (2002). Mini suction probe for
matric suction measurements. Canadian Geotechnical Journal, 39(6):1427–1432.
Mollins, L. H., Stewart, D. I., and Cousens, T. W. (1996). Predicting the properties of bentonite-
sand mixtures. Clay Minerals, 31(2):243–252.
Monroy, R., Zdravkovic, L., and Ridley, A. (2008). Volumetric behaviour of compacted London
clay during wetting and loading. In Unsaturated Soils. Advances in Geo-Engineering: Proceed-
ings of the 1st European Conference, E-UNSAT, pages 315–320, Durham, United Kingdom,
July 2008.
Montanez, J. E. C. (2002). Suction and volume changes of compacted sand-bentonite mixtures.
PhD thesis, Imperial College of Science, University of London.
Mulilis, J., Seed, H., Chan, C., Mitchell, J., and Arulanandan, K. (1977). Effects of sample
preparation on sand liquefaction. Journal of the Geotechnical Engineering Division, 103(2):91–
108.
Oliveira, O. M. and Marinho, F. A. M. (2006). Study of equilibration time in the pressure plate.
Geotechnical Special Publication, 2(147):1864–1874.
Oliveira, O. M. and Marinho, F. A. M. (2008). Suction equilibration time for a high capacity
tensiometer. Geotechnical Testing Journal, 31(1):101–105.
Olson, R. E. (1963). Effective stress theory of soil compaction. Journal of the Soil Mechanics
and Foundations Devision, ASCE, 89(SM2):27–45.
Olson, R. E. and Langfelder, L. J. (1965). Pore water pressures in unsaturated soils. Journal of
Soil Mechanics and Foundations Division, ASCE, 91:127–150.
Padilla, J. M., Perera, Y. Y., Houston, W. N., Perez, N., and Fredlund, D. G. (2006). Quan-
tification of air diffusion through high air-entry ceramic disks. In Proccedings of the Fourth
International Conference on Unsaturated Soils, volume 2, pages 1852–1863, Arizona, USA,
April, 2006.
Perera, Y. Y. (2003). Moisture equilibria beneath paved areas. PhD thesis, Arizona State Uni-
versity, Arizona, USA.
Perera, Y. Y., Houston, W. N., and Fredlund, D. G. (2005). A new soil-water characteristic curve
device. In Advanced Experimental Unsaturated Soil Mechanics: Proceedings of the Interna-
tional Symposium on Advanced Experimental Unsaturated Soil Mechanics, page 15, Trento,
Italy, June, 2005.
Proctor, R. R. (1933). Fundamental principles of soil compaction. Engineering News Record,
111(9):245–248.
184 REFERENCES
Rahardjo, H. and Leong, E. C. (2006). Suction measurements. In Proceedings of 4 th Inter-
national Conference on Unsaturated Soils. Geo Institute, pages 81–104, Arizona, USA, April,
2006.
Richards, L. A. and Ogata, G. (1958). Thermocouple for vapor pressure measurement in biolog-
ical and soil systems at high humidity. Science, 128(3331):1089.
Ridley, A. M. and Burland, J. B. (1993). A new instrument for the measurement of soil moisture
suction. Geotechnique, 43(2):321–4.
Ridley, A. M. and Burland, J. B. (1995). Measurement of suction in materials which swell.
Applied Mechanics Reviews, 48:727.
Rojas, J. C., Mancuso, C., and Vinale, F. (2008a). A modified triaxial apparatus to reduce
testing time: equipment and preliminary results. In Unsaturated Soils. Advances in Geo-
Engineering: Proceedings of the 1st European Conference, E-UNSAT, pages 103–109, Durham,
United Kingdom, July, 2008.
Rojas, J. C., Pagano, L., Zingariello, M. C., and Mancuso, C. (2008b). A new high capacity
tensiometer: First results. In Unsaturated Soils. Advances in Geo-Engineering: Proceedings
of the 1st European Conference, E-UNSAT, pages 205–211, Durham, United Kingdom, July,
2008.
Romero, E. (1999). Characterisation and thermo-hydro-mechanical behavior of unsaturated Boom
clay: An experimental study. PhD thesis, Technical University of Catalonia, Spain.
Romero, E. (2001). Controlled suction techniques. In Proceedings of the 4th Simposio Brasileiro
de Solos Nao Saturados, Porto Alegre, Brasil, 2001.
Romero, E., Gens, A., and Lloret, A. (1999). Water permeability, water retention and mi-
crostructure of unsaturated compacted Boom clay. Engineering Geology, 54(1-2):117–127.
Rowland, R. A., Weiss, E. J., and Bradley, W. F. (1956). Dehydration of monoionic montmoril-
lonites. Clays and Clay Minerals, -:85–95.
Santagiuliana, R. and Schrefler, B. A. (2006). Enhancing the Bolzon-Schrefler-Zienkiewicz con-
stitutive model for partially saturated soil. Transport in Porous Media, 65:1–30.
Shannon, B. (2008). Laboratory modeling of unsaturated water flow and heave in reactive soils.
Final Year Project Report, Monash University, Australia.
Sheng, D., Fredlund, D. G., and Gens, A. (2008). A new modelling approach for unsaturated
soils using independent stress variables. Canadian Geotechnical Journal, 45(4):511–534.
Sibley, J. W., Smyth, G., and Williams, D. J. (1990). Suction-moisture content calibration of
filter papers from different boxes. Geotechnical Testing Journal, 13(3):257–262.
Sivakumar, V. and Wheeler, S. J. (2000). Influence of compaction procedure on the mechanical
behaviour of an unsaturated compacted clay. Part 1: Wetting and isotropic compression.
Geotechnique, 50(4):359–368.
Spanner, D. C. (1951). The Peltier effect and its use in the measurement of suction pressure.
Journal of Experimental Botany, 2(5).
Take, W. A. and Bolton, M. D. (2003). Tensiometer saturation and the reliable measurement of
soil suction. Geotechnique, 53(2):159–172.
185
Tang, A. M. and Cui, Y. J. (2005). Controlling suction by the vapour equilibrium technique
at different temperatures and its application in determining the water retention properties of
MX80 clay. Canadian Geotechnical Journal, 42(1):287–296.
Tang, A. M., Cui, Y. J., and Barnel, N. (2008). Compression-induced suction change in a
compacted expansive clay. In Unsaturated Soils. Advances in Geo-Engineering: Proceedings
of the 1st European Conference, E-UNSAT, pages 369–374, Durham, United Kingdom, July,
2008.
Tarantino, A. and De Col, E. (2008). Compaction behaviour of clay. Geotechnique, 58(3):199–
214.
Tarantino, A. and Mongiovi, L. (2001). Experimental procedures and cavitation mechanisms in
tensiometer measurements. Geotechnical and Geological Engineering, 19(3):189–210.
Tarantino, A. and Mongiovı, L. (2002). Design and construction of a tensiometer for direct
measurement of matric suction. In Proceedings of 3nd International Conference on Unsaturated
Soils, volume 3, pages 319–324, Recife, Brasil, 2002.
Tarantino, A., Mongiovı, L., and Bosco, G. (2000). An experimental investigation on the inde-
pendent isotropic stress variables for unsaturated soils. Geotechnique, 50(3):275–282.
Tarantino, A. and Tombolato, S. (2005). Coupling of hydraulic and mechanical behaviour in
unsaturated compacted clay. Geotechnique, 55(4):307–317.
Terzaghi, K. (1936). The shearing resistance of saturated soils and the angle between the planes
of shear. In Proc. First International Conference on Soil Mechanics, pages 54–56.
Thakur, V. K. S. and Singh, D. N. (2005). Swelling and suction in clay minerals. In Advanced
Experimental Unsaturated Soil Mechanics: Proceedings of the International Symposium on
Advanced Experimental Unsaturated Soil Mechanics, page 27, Trento, Italy, June, 2005.
Thu, T. M., Rahardjo, H., and Leong, E. C. (2007). Elastoplastic model for unsaturated soil
with incorporation of the soil-water characteristic curve. Canadian Geotechnical Journal,
44(1):67–77.
Toker, N. K. (2002). Improvements and reliability of MIT tensiometers and studies on soil
moisture characteristic curves. Master’s thesis, Massachusetts Institute of Technology.
Toker, N. K., Germaine, J. T., Sjoblom, K. J., and Culligan, P. J. (2004). A new technique for
rapid measurement of continuous soil moisture characteristic curves. Geotechnique, 54(3):179–
186.
Truong, H. V. P. and Holden, J. C. (1995). Soil suction measurement with transistor psychrome-
ter. In Proceedings of the 1st International Conference on Unsaturated Soils,, volume 2, pages
659–665, Paris, France, September, 1995.
Unimin (2002). Product Profile ActiveGel 150. Unimin Australia.
Vaid, Y. P. and Negussey, D. (1984). Relative density of pluviated sand samples. Soils and
Foundations, 24(2):101–105.
Vaid, Y. P., Sivathayalan, S., and Stedman, D. (1999). Influence of specimen-reconstituting
method on the undrained response of sand. ASTM geotechnical testing journal, 22(3):187–
195.
186 REFERENCES
Vanapalli, S. K., Fredlund, D. G., and Pufahl, D. E. (1999). The influence of soil structure and
stress history on the soil-water characteristics of a compacted till. Geotechnique, 49(2):143–
160.
Vanapalli, S. K., Nicotera, M. V., and Sharma, R. S. (2008). Axis translation and negative water
column techniques for suction control. Geotechnical and Geological Engineering, 26(6):645–
660.
Vanapalli, S. K., Salinas, L. M., Avila, D., and Karube, D. (2004). Suction and storage char-
acteristics of unsaturated soils. In Unsaturated soils: proceedings of the Third International
Conference on Unsaturated Soils, UNSAT 2002, 10-13 March 2002, Recife, Brazil, volume 3,
page 1045. Taylor & Francis.
Vaunat, J. and Gens, A. (2005). Analysis of the hydration of a bentonite seal in a deep radioactive
waste repository. Engineering Geology, 81(3):317–328.
Vaunat, J., Romero, E., and Jommi, C. (2000). An elastoplastic hydro-mechanical model for
unsaturated soils. In Experimental evidence and theoretical approaches in unsaturated soils,
International Workshop on Unsaturated Soils, pages 121–138, Trento, Italy, April, 2000.
Walker, G. F. (1956). The mechanism of dehydration of Mg-vermiculite. Clays and Clay Min-
erals, 4:101–115.
Wheeler, S. J., Gallipoli, D., and Karstunen, M. (2002). Comments on use of the Barcelona Basic
Model for unsaturated soils. International Journal for Numerical and Analytical Methods in
Geomechanics, 26:1561–1571.
Wheeler, S. J. and Sivakumar, V. (1995). An elasto-plastic critical state framefork for unsatu-
rated soil. Geotechnique, 45(1):35–53.
Woodburn, J. A., Holden, J. C., and Peter, P. (1993). Transistor psychrometer: a new instrument
for measuring soil suction. Geotechnical Special Publication, (39):91–102.
Zapata, C. E., Perera, Y. Y., and Houston, W. N. (2009). Matric suction prediction model in
new AASHTO mechanistic-empirical pavement design guide. Transportation Research Record:
Journal of the Transportation Research Board, 2101:53–62.
Zhan, T. L. T., Ng, C. W. W., and Fredlund, D. G. (2007). Instrumentation of an unsaturated
expansive soil slope. Geotechnical Testing Journal, 30(2):113.
Zhang, X., Mavroulidou, M., Gunn, M. J., and Cabarkapa, Z. (2009). Experience gained using
various axis translation technique apparatus to determine soil water retention curve. In Ex-
perimental studies in unsaturated soils and expansive soils: Proceedings of the 4th Asia Pacific
Conference on Unsaturated Soils, pages 269–274, Newcastle, Australia, November, 2009.
Zheng, Q., Durben, D. J., Wolf, G. H., and Angell, C. A. (1991). Liquids at large negative
pressures: water at the homogeneous nucleation limit. Science, 254(5033):829–832.
Appendix A
Matlab Codes
A.1 Undrained modelling using the BBM
clc;
clear all;
%CONSTANT SUCTION WITH LOADING
%UNDRAINED WITH BBM
%air solvability in water
h = [value];
%slope of the NCL in semi log axis
lvp = [value];
%specific gravity of soil
Gs = [value];
%model parameter
r = [value];
%model parameter
B = [value];
%model parameter
pc = [value];
%model parameter
ks = [value];
%initial specific volume for saturated soil
N_0 = [value];
%change in vertical stress
d_sigma = [value];
%final applied vertical stress
sigma_final = [value];
%atmospheric pressure
p_atm = [value];
%wc and suction_0 needs to be the same size vector
%wc of the samples (add as many as you want)
187
188 APPENDIX A. MATLAB CODES
wc = [values];
%initial suction of the samples
suction_0 = [values];
%calculating lvps and Ns using BBM assumptions and equations
for t = [1:length(suction_0)]
lvp_s(t) = lvp*((1 - r)*exp(-B*suction_0(t)) + r);
N_s(t) = N_0 - (lvp - lvp_s(t))*log(pc) - ks*log((suction_0(t) + p_atm)/p_atm);
end
%programming trick to keep the final values all together
sigma_ff = [];
u_a_ff = [];
v_ff = [];
e_ff = [];
S_rff = [];
n_ff = [];
m1s_ff = [];
gamma_d_ff = [];
%loop that is rotating for different wc
for i = 1:length(wc)
%resetting the initial air pressure
u_a0(i,1) = p_atm;
%keep the final values within the next loop together
sigma_f = [];
u_a_f = [];
v_f = [];
e_f = [];
S_rf = [];
n_f = [];
m1s_f = [];
gamma_d_f = [];
%the loop that is rotating for different vertical stress
for sigma = [d_sigma:d_sigma:sigma_final]
%defining the index for the loop
j = sigma/d_sigma;
%net stress
net_stress(i,j) = sigma - u_a0(i,j) + p_atm;
%specific volume
v(i,j) = N_s(i) - lvp_s(i)*log(net_stress(i,j)/pc);;
%void ratio
e(i,j) = v(i,j) - 1;
%degree of saturation
S_r(i,j) = wc(i)*Gs/e(i,j);
%porosity
n(i,j) = e(i,j)/(e(i,j) + 1);
%coefficient of compressibility due to net stress
189
m1s(i,j) = lvp_s(i)/v(i,j)/(net_stress(i,j));
%dry density
gamma_d(i,j) = Gs/(1 + e(i,j));
%changes in air pressure with loading
du_a(i,j+1) = 1/2/m1s(i,j)*(-m1s(i,j)*u_a0(i,j)...
- n(i,j)+ n(i,j)*S_r(i,j) - n(i,j)*h*S_r(i,j)...
+ m1s(i,j)*d_sigma + (-2*n(i,j)*m1s(i,j)*d_sigma...
+ 2*m1s(i,j)*u_a0(i,j)*n(i,j)...
+ n(i,j)^2*h^2*S_r(i,j)^2 ...
+ 2*m1s(i,j)^2*d_sigma*u_a0(i,j)...
+ 2*n(i,j)^2*h*S_r(i,j)...
- 2* n(i,j)^2*S_r(i,j)^2*h...
- 2*m1s(i,j)*u_a0(i,j)*n(i,j)*S_r(i,j)...
+ m1s(i,j)^2*u_a0(i,j)^2...
+ n(i,j)^2-2*n(i,j)^2*S_r(i,j)...
+ n(i,j)^2*S_r(i,j)^2 + m1s(i,j)^2*d_sigma^2 ...
+ 2*m1s(i,j)*u_a0(i,j)*n(i,j)*h*S_r(i,j) ...
+ 2*n(i,j)*S_r(i,j)* m1s(i,j)*d_sigma...
- 2*n(i,j)*h*S_r(i,j)*m1s(i,j)*d_sigma)^(1/2));
%absolute air pressure
u_a0(i,j+1) = u_a0(i,j) + du_a(i,j+1);
%keep the final values within the loop together
sigma_f = [sigma_f,sigma];
u_a_f = [u_a_f, u_a0(i,j)];
v_f = [v_f, v(i,j)];
e_f = [e_f, e(i,j)];
S_rf = [S_rf, S_r(i,j)];
n_f = [n_f, n(i,j)];
m1s_f = [m1s_f, m1s(i,j)];
gamma_d_f = [gamma_d_f, gamma_d(i,j)];
end
%keep the final values within the loop together
sigma_ff = [sigma_ff; sigma_f];
u_a_ff = [u_a_ff; u_a_f];
v_ff = [v_ff; v_f];
e_ff = [e_ff; e_f];
S_rff = [S_rff; S_rf];
n_ff = [n_ff; n_f];
m1s_ff = [m1s_ff; m1s_f];
gamma_d_ff = [gamma_d_ff; gamma_d_f];
end
190 APPENDIX A. MATLAB CODES
A.2 Semi-drained modelling using the BBM
clc;
clear all;
%CONSTANT SUCTION WITH LOADING
%SEMI DRAINED WITH BBM
%air solvability in water
h = [value];
%slope of the NCL in semi log axis
lvp = [value];
%specific gravity of soil
Gs = [value];
%model parameter
r = [value];
%model parameter
B = [value];
%model parameter
pc = [value];
%model parameter
ks = [value];
%initial specific volume for saturated soil
N_0 = [value];
%change in vertical stress
d_sigma = [value];
%final applied vertical stress
sigma_final = [value];
%atmospheric pressure
p_atm = [value];
%wc and suction_0 needs to be the same size vector
%wc of the samples (add as many as you want)
wc = [values];
%initial suction of the samples
suction_0 = [values];
%calculating lvps and Ns using BBM assumptions and equations
for t = [1:length(suction_0)]
lvp_s(t) = lvp*((1 - r)*exp(-B*suction_0(t)) + r);
N_s(t) = N_0 - (lvp - lvp_s(t))*log(pc) - ks*log((suction_0(t) + p_atm)/p_atm);
end
%programming trick to keep the final values all together
sigma_ff = [];
u_a_ff = [];
v_ff = [];
e_ff = [];
S_rff = [];
n_ff = [];
m1s_ff = [];
gamma_d_ff = [];
191
%loop that is rotating for different wc
for i = 1:length(wc)
%programming trick to keep the final values all together
sigma_f = [];
u_a_f = [];
v_f = [];
e_f = [];
S_rf = [];
n_f = [];
m1s_f = [];
gamma_d_f = [];
%the loop that is rotating for different vertical stress
for sigma = [d_sigma:d_sigma:sigma_final]
%defining the index for the loop
j = sigma/d_sigma;
%resetting the initial air pressure
u_a0(i,j) = 101.3;
%net stress
net_stress(i,j) = sigma - u_a0(i,j) + 101.3;
%specific volume
v(i,j) = N_s(i) - lvp_s(i)*log(net_stress(i,j)/pc);;
%void ratio
e(i,j) = v(i,j) - 1;
%degree of saturation
S_r(i,j) = wc(i)*Gs/e(i,j);
%porosity
n(i,j) = e(i,j)/(e(i,j) + 1);
%coefficient of compressibility due to net stress
m1s(i,j) = lvp_s(i)/v(i,j)/(net_stress(i,j));
%dry density
gamma_d(i,j) = Gs/(1 + e(i,j));
if S_r(i,j) >= 0.98;
net_stress(i,j) = net_stress(i,j-1);
v(i,j) = v(i,j-1);
e(i,j) = e(i,j-1);
S_r(i,j) = S_r(i,j-1);
n(i,j) = n(i,j-1);
m1s(i,j) = m1s(i,j-1);
gamma_d(i,j) = gamma_d(i,j-1);
else
%changes in air pressure with loading
du_a(i,j+1) = 1/2/m1s(i,j)*(-m1s(i,j)*u_a0(i,j)...
- n(i,j)+ n(i,j)*S_r(i,j) - n(i,j)*h*S_r(i,j)...
+ m1s(i,j)*d_sigma + (-2*n(i,j)*m1s(i,j)*d_sigma...
192 APPENDIX A. MATLAB CODES
+ 2*m1s(i,j)*u_a0(i,j)*n(i,j)...
+ n(i,j)^2*h^2*S_r(i,j)^2 ...
+ 2*m1s(i,j)^2*d_sigma*u_a0(i,j)...
+ 2*n(i,j)^2*h*S_r(i,j)...
- 2* n(i,j)^2*S_r(i,j)^2*h...
- 2*m1s(i,j)*u_a0(i,j)*n(i,j)*S_r(i,j)...
+ m1s(i,j)^2*u_a0(i,j)^2...
+ n(i,j)^2-2*n(i,j)^2*S_r(i,j)...
+ n(i,j)^2*S_r(i,j)^2 + m1s(i,j)^2*d_sigma^2 ...
+ 2*m1s(i,j)*u_a0(i,j)*n(i,j)*h*S_r(i,j) ...
+ 2*n(i,j)*S_r(i,j)* m1s(i,j)*d_sigma...
- 2*n(i,j)*h*S_r(i,j)*m1s(i,j)*d_sigma)^(1/2));
%absolute air pressure
u_a0(i,j+1) = u_a0(i,j) + du_a(i,j+1);
%net stress
net_stress(i,j) = sigma - u_a0(i,j) + 101.3;
%specific volume
v(i,j) = N_s(i) - lvp_s(i)*log(net_stress(i,j)/pc);
%void ration
e(i,j) = v(i,j) - 1;
%degree of saturation
S_r(i,j) = wc(i)*Gs/e(i,j);
%porosity
n(i,j) = e(i,j)/(e(i,j) + 1);
%coefficient of compressibility due to net stress
m1s(i,j) = lvp_s(i)/v(i,j)/(net_stress(i,j));
%dry density
gamma_d(i,j) = Gs/(1 + e(i,j));
end;
%programming trick to keep the final values all together
sigma_f = [sigma_f,sigma];
u_a_f = [u_a_f, u_a0(i,j)];
v_f = [v_f, v(i,j)];
e_f = [e_f, e(i,j)];
S_rf = [S_rf, S_r(i,j)];
n_f = [n_f, n(i,j)];
m1s_f = [m1s_f, m1s(i,j)];
gamma_d_f = [gamma_d_f, gamma_d(i,j)];
end
%programming trick to keep the final values all together
sigma_ff = [sigma_ff; sigma_f];
u_a_ff = [u_a_ff; u_a_f];
v_ff = [v_ff; v_f];
e_ff = [e_ff; e_f];
S_rff = [S_rff; S_rf];
n_ff = [n_ff; n_f];
m1s_ff = [m1s_ff; m1s_f];
gamma_d_ff = [gamma_d_ff; gamma_d_f];
end
193
A.3 Undrained modelling using the SFG model
clc;
clear all;
%CONSTANT SUCTION WITH LOADING
%UNDRAINED WITH SFG
%air solvability in water
h = [value];
%slope of the NCL in semi log axis
lvp = [value];
%specific gravity of soil
Gs = [value];
%initial net stress where N is calculated
p_0 = [value];
%initial specific volume for saturated soil
N_0 = [value];
%saturation suction
s_sa = [value];
%change in vertical stress
d_sigma = [value];
%final applied vertical stress
sigma_final = [value];
%atmospheric pressure
p_atm = [value];
%wc and suction_0 needs to be the same size vector
%wc of the samples (add as many as you want)
wc = [values];
%initial suction of the samples
suction_0 = [values];
%calculating Ns using SFG assumptions and equations (choose one of them)
for t = [1:length(suction_0)]
%for s > s_sa
N(t) = exp(log(N_0) - lvp*log((p_0 + s_sa)/p_0)...
- lvp*(s_sa +1)/(p_0 - 1)*log((suction_0(t)...
+ 1)*(p_0 + s_sa)/(p_0 + suction_0(t))/(s_sa + 1)));
%for s < s_sa
N(t) = exp(log(N_0) - lvp*log((p_0 + s_sa)/p_0) ...
- lvp*(1 - (s_sa + 1)/(suction_0(t) + 1)));
end
%programming trick to keep the final values all together
sigma_ff = [];
u_a_ff = [];
v_ff = [];
e_ff = [];
S_rff = [];
n_ff = [];
m1s_ff = [];
194 APPENDIX A. MATLAB CODES
gamma_d_ff = [];
%loop that is rotating for different wc
for i = 1:length(wc)
%resetting the initial air pressure
u_a0(i,1) = p_atm;
%keep the final values within the next loop together
sigma_f = [];
u_a_f = [];
v_f = [];
e_f = [];
S_rf = [];
n_f = [];
m1s_f = [];
gamma_d_f = [];
%the loop that is rotating for different vertical stress
for sigma = [d_sigma:d_sigma:sigma_final]
%defining the index for the loop
j = sigma/d_sigma;
%net stress
net_stress(i,j) = sigma - u_a0(i,j) + p_atm;
%specific volume
v(i,j) = N_s(i) - lvp_s(i)*log(net_stress(i,j)/pc);;
%void ratio
e(i,j) = v(i,j) - 1;
%degree of saturation
S_r(i,j) = wc(i)*Gs/e(i,j);
%porosity
n(i,j) = e(i,j)/(e(i,j) + 1);
%coefficient of compressibility due to net stress
m1s(i,j) = lvp_s(i)/v(i,j)/(net_stress(i,j));
%dry density
gamma_d(i,j) = Gs/(1 + e(i,j));
%changes in air pressure with loading
du_a(i,j+1) = 1/2/m1s(i,j)*(-m1s(i,j)*u_a0(i,j)...
- n(i,j)+ n(i,j)*S_r(i,j) - n(i,j)*h*S_r(i,j)...
+ m1s(i,j)*d_sigma + (-2*n(i,j)*m1s(i,j)*d_sigma...
+ 2*m1s(i,j)*u_a0(i,j)*n(i,j)...
+ n(i,j)^2*h^2*S_r(i,j)^2 ...
+ 2*m1s(i,j)^2*d_sigma*u_a0(i,j)...
+ 2*n(i,j)^2*h*S_r(i,j)...
- 2* n(i,j)^2*S_r(i,j)^2*h...
- 2*m1s(i,j)*u_a0(i,j)*n(i,j)*S_r(i,j)...
+ m1s(i,j)^2*u_a0(i,j)^2...
+ n(i,j)^2-2*n(i,j)^2*S_r(i,j)...
+ n(i,j)^2*S_r(i,j)^2 + m1s(i,j)^2*d_sigma^2 ...
+ 2*m1s(i,j)*u_a0(i,j)*n(i,j)*h*S_r(i,j) ...
+ 2*n(i,j)*S_r(i,j)* m1s(i,j)*d_sigma...
195
- 2*n(i,j)*h*S_r(i,j)*m1s(i,j)*d_sigma)^(1/2));
%absolute air pressure
u_a0(i,j+1) = u_a0(i,j) + du_a(i,j+1);
%keep the final values within the loop together
sigma_f = [sigma_f,sigma];
u_a_f = [u_a_f, u_a0(i,j)];
v_f = [v_f, v(i,j)];
e_f = [e_f, e(i,j)];
S_rf = [S_rf, S_r(i,j)];
n_f = [n_f, n(i,j)];
m1s_f = [m1s_f, m1s(i,j)];
gamma_d_f = [gamma_d_f, gamma_d(i,j)];
end
%keep the final values within the loop together
sigma_ff = [sigma_ff; sigma_f];
u_a_ff = [u_a_ff; u_a_f];
v_ff = [v_ff; v_f];
e_ff = [e_ff; e_f];
S_rff = [S_rff; S_rf];
n_ff = [n_ff; n_f];
m1s_ff = [m1s_ff; m1s_f];
gamma_d_ff = [gamma_d_ff; gamma_d_f];
end
196 APPENDIX A. MATLAB CODES
A.4 Semi-drained modelling using the SFG model
clc;
clear all;
%CONSTANT SUCTION WITH LOADING
%SEMI DRAINED WITH SFG
%air solvability in water
h = [value];
%slope of the NCL in semi log axis
lvp = [value];
%specific gravity of soil
Gs = [value];
%initial net stress where N is calculated
p_0 = [value];
%initial specific volume for saturated soil
N_0 = [value];
%saturation suction
s_sa = [value];
%change in vertical stress
d_sigma = [value];
%final applied vertical stress
sigma_final = [value];
%atmospheric pressure
p_atm = [value];
%wc and suction_0 needs to be the same size vector
%wc of the samples (add as many as you want)
wc = [values];
%initial suction of the samples
suction_0 = [values];
%calculating Ns using SFG assumptions and equations (choose one of them)
for t = [1:length(suction_0)]
%for s > s_sa
N(t) = exp(log(N_0) - lvp*log((p_0 + s_sa)/p_0)...
- lvp*(s_sa +1)/(p_0 - 1)*log((suction_0(t)...
+ 1)*(p_0 + s_sa)/(p_0 + suction_0(t))/(s_sa + 1)));
%for s < s_sa
N(t) = exp(log(N_0) - lvp*log((p_0 + s_sa)/p_0) ...
- lvp*(1 - (s_sa + 1)/(suction_0(t) + 1)));
end
%programming trick to keep the final values all together
sigma_ff = [];
u_a_ff = [];
v_ff = [];
e_ff = [];
S_rff = [];
n_ff = [];
m1s_ff = [];
197
gamma_d_ff = [];
%loop that is rotating for different wc
for i = 1:length(wc)
%programming trick to keep the final values all together
sigma_f = [];
u_a_f = [];
v_f = [];
e_f = [];
S_rf = [];
n_f = [];
m1s_f = [];
gamma_d_f = [];
%the loop that is rotating for different vertical stress
for sigma = [d_sigma:d_sigma:sigma_final]
%defining the index for the loop
j = sigma/d_sigma;
%resetting the initial air pressure
u_a0(i,j) = 101.3;
%net stress
net_stress(i,j) = sigma - u_a0(i,j) + 101.3;
%specific volume
v(i,j) = N_s(i) - lvp_s(i)*log(net_stress(i,j)/pc);;
%void ratio
e(i,j) = v(i,j) - 1;
%degree of saturation
S_r(i,j) = wc(i)*Gs/e(i,j);
%porosity
n(i,j) = e(i,j)/(e(i,j) + 1);
%coefficient of compressibility due to net stress
m1s(i,j) = lvp_s(i)/v(i,j)/(net_stress(i,j));
%dry density
gamma_d(i,j) = Gs/(1 + e(i,j));
if S_r(i,j) >= 0.98;
net_stress(i,j) = net_stress(i,j-1);
v(i,j) = v(i,j-1);
e(i,j) = e(i,j-1);
S_r(i,j) = S_r(i,j-1);
n(i,j) = n(i,j-1);
m1s(i,j) = m1s(i,j-1);
gamma_d(i,j) = gamma_d(i,j-1);
else
%changes in air pressure with loading
du_a(i,j+1) = 1/2/m1s(i,j)*(-m1s(i,j)*u_a0(i,j)...
- n(i,j)+ n(i,j)*S_r(i,j) - n(i,j)*h*S_r(i,j)...
198 APPENDIX A. MATLAB CODES
+ m1s(i,j)*d_sigma + (-2*n(i,j)*m1s(i,j)*d_sigma...
+ 2*m1s(i,j)*u_a0(i,j)*n(i,j)...
+ n(i,j)^2*h^2*S_r(i,j)^2 ...
+ 2*m1s(i,j)^2*d_sigma*u_a0(i,j)...
+ 2*n(i,j)^2*h*S_r(i,j)...
- 2* n(i,j)^2*S_r(i,j)^2*h...
- 2*m1s(i,j)*u_a0(i,j)*n(i,j)*S_r(i,j)...
+ m1s(i,j)^2*u_a0(i,j)^2...
+ n(i,j)^2-2*n(i,j)^2*S_r(i,j)...
+ n(i,j)^2*S_r(i,j)^2 + m1s(i,j)^2*d_sigma^2 ...
+ 2*m1s(i,j)*u_a0(i,j)*n(i,j)*h*S_r(i,j) ...
+ 2*n(i,j)*S_r(i,j)* m1s(i,j)*d_sigma...
- 2*n(i,j)*h*S_r(i,j)*m1s(i,j)*d_sigma)^(1/2));
%absolute air pressure
u_a0(i,j+1) = u_a0(i,j) + du_a(i,j+1);
%net stress
net_stress(i,j) = sigma - u_a0(i,j) + 101.3;
%specific volume
v(i,j) = N_s(i) - lvp_s(i)*log(net_stress(i,j)/pc);
%void ration
e(i,j) = v(i,j) - 1;
%degree of saturation
S_r(i,j) = wc(i)*Gs/e(i,j);
%porosity
n(i,j) = e(i,j)/(e(i,j) + 1);
%coefficient of compressibility due to net stress
m1s(i,j) = lvp_s(i)/v(i,j)/(net_stress(i,j));
%dry density
gamma_d(i,j) = Gs/(1 + e(i,j));
end;
%programming trick to keep the final values all together
sigma_f = [sigma_f,sigma];
u_a_f = [u_a_f, u_a0(i,j)];
v_f = [v_f, v(i,j)];
e_f = [e_f, e(i,j)];
S_rf = [S_rf, S_r(i,j)];
n_f = [n_f, n(i,j)];
m1s_f = [m1s_f, m1s(i,j)];
gamma_d_f = [gamma_d_f, gamma_d(i,j)];
end
%programming trick to keep the final values all together
sigma_ff = [sigma_ff; sigma_f];
u_a_ff = [u_a_ff; u_a_f];
v_ff = [v_ff; v_f];
e_ff = [e_ff; e_f];
S_rff = [S_rff; S_rf];
n_ff = [n_ff; n_f];
m1s_ff = [m1s_ff; m1s_f];
gamma_d_ff = [gamma_d_ff; gamma_d_f];
end
Appendix B
Publications
Kurucuk, N., Kodikara, J., and Fredlund, D. G., 2009. Evolution of the compaction process:
Experimental study - preliminary results. Unsaturated Soils - Theoretical & Numerical Advances
in Unsaturated Soil Mechanics: Proceedings of the 4th Asia-Pacific Conference on Unsaturated
Soils, pages 887-893, Newcastle, Australia, November 2009.
Kurucuk, N., Kodikara, J., and Fredlund, D. G., 2008. Theoretical modelling of the compaction
curve. Advances in Geo-Engineering: Proceedings of the 1st European Conference, E-UNSAT,
pages 375-379, Durham, United Kingdom, July 2008.
Kurucuk, N., Kodikara, J., and Fredlund, D. G., 2007. Prediction of compaction curves.
Proceedings of the 10th ANZ Conference on Geomechanics, Common Ground, pages 115-119,
Brisbane, Australia, October 2007.
199
200 APPENDIX B. PUBLICATIONS
B.1
114 grouNd improvemeNt
cO
mm
On
gr
Ou
nd
07
Common Ground ProCeedinGs 10th AustrAliA new ZeAlAnd ConferenCe on GeomeChAniCs BrisBAne
cO
mm
On
gr
Ou
nd
07 Prediction of compaction curves
Nurses Kurucuk Faculty of Engineering, Monash University, Clayton, VIC, Australia
Jayantha Kodikara Faculty of Engineering, Monash University, Clayton, VIC, Australia
Delwyn Fredlund Golder Associates, Saskatoon, Saskatchewan, Canada.
Keywords: soil compaction, soil suction, stiffness, unsaturated soil, pore pressure development
ABSTRACT
Compaction of soil is one of the major activities in geotechnical engineering involving earthworks. Compaction curve generally features an inverted parabolic shape and is used to find the optimum water content that maximises dry density. Since its introduction by Proctor in 1933, several researchers have provided qualitative explanations for the general shape of the compaction curve. Furthermore, there is a vast body of literature covering the behaviour of compacted soil. However, fundamental research on the compaction process and the evolution of compaction characteristics are limited. Therefore, in order to understand the driving mechanisms of compaction, this paper investigates the effect of soil suction and stiffness in the shape of the compaction curve, from unsaturated soil mechanics standpoint. On the basis of unsaturated soil mechanics principles, a theoretical prediction of the parabolic shape of the compaction curve is provided.
1 INTRODUCTION
Soil compaction is widely used in geo-engineering and is important for the construction of roads, dams, landfills, airfields, foundations, hydraulic barriers, and ground improvements. Compaction is applied to the soil, with the purpose of finding optimum water content to maximise its dry density, and therefore, to decrease soil’s compressibility, increase its shearing strength, and in some cases, to reduce its permeability. Proper compaction of materials ensures the durability and stability of earthen constructions.
A typical compaction curve presents different densification stages when the soil is compacted with the same apparent energy input but different water contents. The water content at the peak of the curve is called optimum water content (OWC) and represents the water content at which dry density is at its maximum for a given compaction energy.
Since Proctor`s pioneering work in 1933, many researchers have attempted to explain the leading mechanisms in the densification stages, mainly on the dry side of optimum water content. The compaction curve was explained in terms of capillarity and lubrication (Proctor, 1933), viscous water (Hogentogler, 1936), pore pressure theory in unsaturated soils (Hilf, 1956), physico-chemical interactions (Lambe, 1960), and concepts of effective stress theory (Olson, 1963). More recently, Barden & Sides (1970) undertook an experimental research on the relation between the engineering performance of compacted unsaturated clay and microscopic observations of clay structure. In addition, Lee & Suedkamp (1972) conducted research on the shape of the compaction curve for different soils. These works provide predominantly qualitative explanations of the shape of the curve.
Despite this research work, and the importance and high demand for the compaction process in engineering practice, it still remains that the compaction of soil is quite complex and not well explained, particularly from a quantitative sense. Therefore, there is need for research to be undertaken at a fundamental level to understand the compaction characteristics of soil and the inverted parabolic shape of the compaction curve.
This paper presents a theoretical prediction of the compaction curve for sand using unsaturated soil mechanics principles. In order to achieve this aim, approaches developed by Hilf (1948) and Fredlund & Rahardjo (1993) for undrained loading process are simulated. Volume changes of a compacted soil are calculated using the volume change theory from unsaturated soil mechanics, and are used to predict theoretical compaction curves.
201
115grouNd improvemeNt
cO
mm
On
gr
Ou
nd
07
Common Ground ProCeedinGs 10th AustrAliA new ZeAlAnd ConferenCe on GeomeChAniCs BrisBAne
cO
mm
On
gr
Ou
nd
072 MODELLING OF COMPACTION CURVES
2.1 Theoretical background
Theoretical concepts utilized for the development of soil compaction curves are presented in this section. Initially, Hilf’s (1948) and Fredlund & Rahardjo’s (1993) approaches for pore pressure development are presented. This is continued with Fredlund & Morgenstern’s (1976) volume change theory for a compacted soil and the derivation of the dry density of a soil.
2.1.1 Pore pressure development during static compaction
One of the main simulations for the generation of the compaction curve is that of pore pressure development. Hilf (1948) developed a relationship between pore pressure and applied stress, which is based on one-dimensional soil compression, Boyle’s law, and Henry’s law, and is expressed as follows:
0 0 0
0
11
1a y
a a v
uS hS n
u u m
(1)
where; au = change in absolute pore air pressure, 0S = initial degree of saturation, h = coefficient
of solubility, 0n = initial porosity, 0au = initial absolute air pressure, vm = coefficient of volume
change in saturated soil, and y= change in applied vertical stress.
Hilf (1948) developed this equation assuming that air and water phases are undrained, and volume reduction is due to air dissolving in the water and compression of free air. Both liquid and solid parts were considered to be incompressible. He also assumed that the change in pore air pressure is equal to the change in pore water pressure, and therefore, matric suction change was insignificant. These assumptions will be reviewed further in Section 2.2. On the other hand, Fredlund & Rahardjo (1993) developed an equation for undrained loading for pore pressure changes by using pore pressure parameters:
a a yu B (2) w w yu B (3)
where; aB = pore air pressure parameter, wB = pore water pressure parameter, and y= change in
applied vertical stress.
Pore air ( )aB and pore water ( )wB pressure parameters are defined as follows:
2 3 4
1 31aR R RB
R R (4); 2 1 4
1 31wR R RB
R R. (5)
In Equations 4 and 5, 1R , 2R , 3R and 4R are defined as:
2 1 11
2 1 1
/ 1 1 /
/ /
s s sa
s s sw
m m S hS n u mR
m m SnC m (6);
22 1 1
1/ /s s s
w
Rm m SnC m
(7)
2 13
2 1 1
// 1 1 /
a a
a a sa
m mRm m S hS n u m
(8); 4
2 1 1
1/ 1 1 /a a s
a
Rm m S hS n u m
(9)
where; 1sm = compressibility of soil particles with respect to net stress ( )y au , 2
sm =
compressibility of soil particles referenced to suction ( )a wu u , 1am = compressibility of air phase
202 APPENDIX B. PUBLICATIONS
116 grouNd improvemeNt
cO
mm
On
gr
Ou
nd
07
Common Ground ProCeedinGs 10th AustrAliA new ZeAlAnd ConferenCe on GeomeChAniCs BrisBAne
cO
mm
On
gr
Ou
nd
07 with respect to net stress ( )y au , 2
am = compressibility of air phase with respect to suction
( )a wu u , au = absolute pore air pressure, S = degree of saturation, h = coefficient of solubility,
n = porosity, and wC = compressibility of water.
This approach assumes that soil particles are incompressible, but the fluid phases are compressible. In contrast to the approach by Hilf (1948), matric suction changes were taken into account. It was accepted that air and water pressures become equal at saturation. Therefore, matric suction will decrease from the initial value to zero at saturation
2.1.2 Computation of volume change and dry density
The volume change constitutive relationship proposed by Fredlund & Morgenstern (1976) for unsaturated soils is used for the calculations of compaction curves, and is expressed as follows:
1 20
( ) ( )s svy a a w
V m u m u uV
(10)
where; vV = overall volume change of a soil element, 0V = initial total volume of the soil element,
1sm = compressibility of soil particles with respect to net stress ( )y au , 2
sm = compressibility of
soil particles referenced to suction ( )a wu u , ( )y au = change in net stress, and ( )a wu u =
change in soil suction.
Since soil particles are incompressible, it is accepted that deformation is primarily due to compression of the pore fluid (i.e., the air and water mixture). The independent stress state variable concept is utilized in the derivation; namely, net stress ( )y au (causes a reduction in
volume with compression), and matric suction stress ( )a wu u (generally results in volume increase with compression). Consequently, overall volume change of the soil sample is calculated by multiplying the right hand side of Equation 10 by the initial volume of the soil sample. Furthermore, the final total volume can be obtained using the volume change of the soil, and this gives the corresponding dry density.
2.2 Estimation of soil compaction curves
Figure 1: Pore air pressure development with applied stress In order to produce compaction curve for a given set of parameters, the pore air pressure development for Hilf`s (1948) and Fredlund & Rahardjo’s (1993) approaches are plotted in Figure 1. The parametric values used in this calculation are given in this figure, and were obtained by
203
117grouNd improvemeNt
cO
mm
On
gr
Ou
nd
07
Common Ground ProCeedinGs 10th AustrAliA new ZeAlAnd ConferenCe on GeomeChAniCs BrisBAne
cO
mm
On
gr
Ou
nd
07consulting graphs given by Fredlund & Rahardjo (1993). It should be noted that Fredlund &
Rahardjo`s (1993) approach takes into consideration the changes in suction with compression, whereas Hilf’s (1948) analysis assumes that the suction change is negligible during compression and suction is equal to the initial matric suction. Therefore, Hilf`s (1948) approach produces larger pore air pressure values than Fredlund & Rahardjo’s (1993) approach for a given applied stress.
Using the volume change equation (Equation 10) and air pressure computed with applied stress (Figure 1), compaction curves are produced for both approaches as shown in Figure 2. In Figure 2 compressibility coefficients are assumed to be constant and their values are given on the figure. It should be noted that compaction curve which is plotted using Hilf’s (1948) approach does not include the suction component of volume change theory (or assumed m2
s is assumed equal to zero in Equation 10).
Figure 2: Compaction curves for constant coefficient of compressibility
It is clear that using constant compressibility coefficients (mv, m1s, m2
s, in fact it can be shown that mv=m1
s) a reasonable shape for the compaction curve can be produced on the wet side of OWC, but not on the dry side of OWC. Moreover, suction change that is included in Fredlund & Rahardjo’s (1993) approach results in some reduction in dry density, but does not change the shape of the compaction curve at lower water contents. In order to examine the likely suction change during compaction, consideration is given to the experimental results produced by Montanez (2002), as shown in Figure 3. Montanez (2002) produced suction contours for compacted sand-bentonite mixtures associated with domains defined by compaction curves at various apparent energy levels. He found that matric suction only decreases marginally with a density increase and in many cases can be considered constant. Therefore, Hilf’s analysis of assuming constant suction during compaction appears to be close to the real situation, but direct data measured during the compaction process may be useful to validate this assumption.
1.90
2 6
Contours of equal suction
Water content (%)
Dry
den
sity
( M
g/m
)
4 8 1210 1614 2018 2422 2826
1.80
1.70
1.60
1.50
1.40
3
Compaction energyHeavyStandard
5153
kPa
1498
kPa
780
kPa
374
kPa
182
kPa
84 k
Pa
Sr = 100%
Figure 3: Data for uniform sand-bentonite (10%) mixture (adopted from Montanez, 2002)
204 APPENDIX B. PUBLICATIONS
118 grouNd improvemeNt
cO
mm
On
gr
Ou
nd
07
Common Ground ProCeedinGs 10th AustrAliA new ZeAlAnd ConferenCe on GeomeChAniCs BrisBAne
cO
mm
On
gr
Ou
nd
07 Furthermore, even if there is a small reduction in suction with compaction, m2
s applicable to the unloading process can be considered smaller than that applicable to loading process. Therefore, the effect of the suction component (as in the second component of Equation 10) can be considered not very significant. With this insight, the focus on the compaction process falls on the coefficient of soil compressibility due to net stress, m1
s. In the following part of the paper, the influence of coefficient of compressibility due to net stress (m1
s) on the shape of the compaction curve is presented. Direct data related to the variation of m1
s with saturation are not available. However, some insight can be gained from the shear-wave velocity measured by Cho & Santamarina (2001) in granite powder at various saturations, shown in Figure 4, to explain the relationship between m1
s
and degree of saturation.
Figure 4: Shear-wave velocity data for granite powder (adopted from Cho & Santamarina, 2001)
It is clear from Figure 4 that the shear-wave velocity sV , hence shear modulus 2sG = V of the
soil increases with decreasing saturation (or conversely soil compressibility decreases). Therefore, it can be argued that m1
s will decrease with decreasing saturation in some form. On this basis, a relationship between m1
s and degree of saturation is plotted as shown in Figure 5.
205
119grouNd improvemeNt
cO
mm
On
gr
Ou
nd
07
Common Ground ProCeedinGs 10th AustrAliA new ZeAlAnd ConferenCe on GeomeChAniCs BrisBAne
cO
mm
On
gr
Ou
nd
07Figure 5: Change in m1
s values with initial degree of saturation When the variation of m1
s with degree of saturation is taken into account, the resulting compaction curves are shown in Figure 6. Clearly, now the proper shape of the compaction curve can be produced. Moreover, it is clear that the dependency of compaction curves at different stress levels (or energy input) is reasonably well captured, where optimum water content shifts to the right with higher stress levels.
Figure 6: Compaction curves for different energy levels
3 CONCLUSIONS
This paper presents the theoretical concepts related to the prediction of the compaction curve. For the first time, it highlights the fact that shape of the compaction curve can be predicted using unsaturated soil mechanics principles. The main insight gained was that the changes in matric suction is not important for the evolution of the compaction states, but that the influence of matric suction on the material compressibility with respect to net stress is the governing factor determining the compaction density. Therefore, it can be reasoned that the inverted parabolic shape of the compaction curves is a direct function of the variation of the material compressibility with degree of saturation. More research, however, is necessary to reinforce these concepts.
REFERENCES
Barden, L., and Sides, G. R. (1970), Engineering behaviour and structure of compacted clay, Journal Soil Mechanics and Foundations Division, ASCE, 96, No. SM4, 1171. Cho, G. C., and Santamarina, J. C. (2001), Unsaturated particulated materials – Particle level studies, Journal of Geotechnical and Geoenvironmental Engineering, 127(1), 84-96. Fredlund, D. G., and Morgenstern, N. R. (1976), Constitutive relations for volume change in unsaturated soils, Canadian Geotechnical Journal, 14, 3, 261-276. Fredlund, D. G., and Rahardjo, H. (1993), Soil mechanics for unsaturated soils, p. 191-194, John Wiley & Sons. Hilf, J. W. (1956), An Investigation of Pore Water Pressures in Compacted Cohesive Soils, Technical Memorandum 654, U. S. Department of the Interior, Bureau of Reclamation, Denver, Colorado. Hogentogler, C. A. (1936), Essentials of soil compaction, Proceedings Highway Research Board, National Research Council, Washington, D.C., 309-316. Lambe, T. W. (1960), Structure of compacted clay, Transactions, ASCE, 125, 682-705. Lee, D. Y., and Suedkamp, R. J. (1972), Characteristics of irregularly shaped compaction curves of soil, Highway Research Board, 381, 1-9. Montanez, J. E. C. (2002), Suction and volume changes of compacted sand-bentonite mixtures, PhD thesis, University of London, Imperial College of Science, London, England. Olson, R. E. (1963), Effective stress theory of soil compaction, Journal Soil Mechanics and Foundations Division, ASCE, 89, No. SM2, 27-45. Proctor, R. R. (1933), Fundamental Principles of Soil Compaction, Engineering News-Record, 111,286.
206 APPENDIX B. PUBLICATIONS
B.2
Unsaturated Soils: Advances in Geo-Engineering – Toll et al. (eds)© 2008 Taylor & Francis Group, London, ISBN 978-0-415-47692-8
Theoretical modelling of the compaction curve
N. Kurucuk & J. KodikaraDepartment of Civil Engineering, Faculty of Engineering, Monash University, Clayton, VIC, Australia
D.G. FredlundGolder Associates, Saskatoon, Saskatchewan, Canada
ABSTRACT: Soil compaction is one of the major activities in geotechnical engineering involving earthworks.The compaction curve is used to find the optimumwater content thatmaximizes dry density. Since its introductionby Proctor in 1933, several researchers have provided qualitative explanations for the inverted parabolic shapeof the compaction curve. However, fundamental research on the compaction process and the evolution ofcompaction characteristics are limited, particularly from a quantitative sense. In order to understand the drivingmechanisms of soil compaction, this paper investigates the effect of soil suction, stiffness and pore air pressureon the shape of the compaction curve, from an unsaturated soil mechanics standpoint. This paper presents anapproach to predict the soil compaction curve during undrained loading. Particular attention is focused on thederivation of the compressibility coefficient due to net stress. Model predictions of the compaction curve arecompared with some experimental results from the literature.
1 INTRODUCTION
Soil compaction is widely used in geo-engineeringand is important for the construction of roads, dams,landfills, airfields, foundations, hydraulic barriers,and ground improvements. Compaction is applied tothe soil, with the purpose of finding optimum watercontent in order to maximize its dry density, and there-fore, to decrease compressibility, increase shearingstrength, and in some cases, to reduce permeability.Proper compaction of materials ensures the durabilityand stability of earthen constructions.A typical compaction curve presents different den-
sification stages when the soil is compacted with thesame apparent energy input but different water con-tents. The water content at the peak of the curve iscalled the optimum water content (OWC) and repre-sents the water content at which dry density is at itsmaximum for a given compaction energy.Since Proctor’s pioneering work in 1933, many
researchers have attempted to explain qualitativelythe leading mechanisms in the densification stages,mainly on the dry side of optimum water content. Thecompaction curvewas explained in terms of capillarityand lubrication (Proctor, 1933), viscouswater (Hogen-togler, 1936), pore pressure theory in unsaturated soils(Hilf, 1956), physico-chemical interactions (Lambe,1960), and concepts of effective stress theory (Olson,1963). More recently, Barden & Sides (1970) under-took experimental research on the relation between the
engineering performance of compacted unsaturatedclay andmicroscopic observations of clay structure. Inaddition, Lee & Suedkamp (1972) conducted researchon the shape of the compaction curve for differentsoils.Despite this research work, and the importance and
high demand for the compaction process in engineer-ing practice, the compaction of soil is quite complexand notwell explained, particularly from a quantitativesense. Theoretical modelling of the soil compactioncurve will provide a better understanding of the mainparameters that affect the shape of the compactioncurve, and understanding the behaviour of compactedmaterials. Therefore, there is need for research tobe undertaken at a fundamental level to understandthe compaction characteristics of soil and the invertedparabolic shape of the compaction curve.This paper presents a theoretical explanation of the
compaction curve using unsaturated soil mechanicsprinciples. Particular attention is focused on the pre-diction of the compressibility coefficient due to netstress. Likely predictions of the model are comparedwith the experimental results from literature.
2 THEORETICAL BACKGROUNDFOR MODELLING
Theoretical concepts utilized for the development ofsoil compaction curves are presented in this section.
375
207
Initially, Hilf’s (1948) approach for pore pressuredevelopment is presented. This is continued withFredlund &Morgenstern’s (1976) volume change the-ory for a compacted soil and the derivation of the drydensity of soil.
2.1 Pore pressure development during staticcompaction
One of the main simulations for the generation of thecompaction curve is that of pore pressure develop-ment. Hilf (1948) developed a relationship betweenpore pressure and applied stress, which is based onone-dimensional K0 soil compression, Boyle’s law,and Henry’s law, and is expressed as follows:
ua =1
1 + (1− S0+ hS0)n0(ua0+ ua)mv
σ y (1)
where; ua = change in absolute pore air pressure,S0 = initial degree of saturation, h = coefficient ofsolubility, n0 = initial porosity, ua0 = initial absoluteair pressure, mv = coefficient of volume change insaturated soil, and σy = change in applied verticalstress.Hilf (1948) developed this equation assuming that
air and water phases are undrained, and volumereduction is due to air dissolving in the water andcompression of free air. Both liquid and solid partswere considered to be volumetrically incompressible.Hilf also assumed that the change in pore air pres-sure is equal to the change in pore water pressure,and therefore, matric suction change was insignifi-cant. Experimental results on suction change duringcompaction can be found in literature (e.g. Li 1995,Montanez 2002). It is shown that matric suction onlydecreases marginally with a density increase and maybe approximated to be constant. Therefore, Hilf’sanalysis assuming constant suction during compactionappears to be close to the real situation. Further justifi-cation for assuming constant matric suction during thecompaction test is presented in Kurucuk et al. (2007).
2.2 Computation of volume change and dry density
The volume change constitutive relationship as appli-cable to K0 loading, which is defined in termsof two independent stress variables as proposed byFredlund &Morgenstern (1976) for unsaturated soils,is used for the calculation of compaction curves:
εv =VvV
= ms1 σy − ua + ms2 (ua − uw) (2)
where; εv = volumetric strain, Vv = overall volumechange of soil element, V = initial total volume of soil
element, ms1 = compressibility of soil particles with
respect to net stress (σy − ua), ms2 = compressibility
of soil particles referenced to matric suction (ua− uw),( σy − ua) = change in net stress, and ( ua − uw) =change in soil suction.Since soil particles are incompressible, it is
accepted that deformation is primarily due to compres-sion of the pore fluid (i.e., the air and air/water mix-ture). The independent stress state variable concept isutilized in the derivation; namely, net stress (σy − ua)(causes a reduction in volume with compression), andmatric suction stress (ua − uw) (generally results involume increase with compression). Once the overallvolume change is computed, the corresponding drydensity can be easily computed.
3 MODELLING ASSUMPTIONS
Kurucuk et al. (2007) showed that the assumptionof constant coefficients of compressibility duringcompaction does not produce a proper shape of thecompaction curve especially on the dry side of theoptimum water content. Their analysis showed that itis ms
1 that controls the volume changes during com-paction because the associated change in suction maybe neglected. The parameter ms
1 was represented as afunction of saturation and decreases with decreasingsaturation. However, the experimental results pre-sented by Loret et al. (2003) showed that ms
1 decreasedwith both suction and net stress. Therefore, followingthe functional form suggested by Sheng et al. (2007),the volumetric strain, ignoring suction change, may bepresented as:
εv =dV
V= λvp
d (σnet − ua)
(σnet − ua) + s0(3)
where; εv = volumetric strain, (σnet− ua) = mean netstress, ua = pore air pressure, s0 = suction, λvp =slope of the normal compression line (NCL) of thesaturated soil, and V = initial total volume of the soilelement.This gives ms
1 as:
ms1 =λvp
σy − ua + s0(4)
This assumption will be used and discussed furtherin the modelling of the compaction curve. It is reason-able to replace mv in Equation 1 by ms
1. A numericalexample of the variation ofms
1 during compaction pro-cess is given in the following section. Equations (1),(3) and (4) were used in incremental forms to com-pute the incremental and total volume change and thecorresponding dry density values during compaction.
376
208 APPENDIX B. PUBLICATIONS
4 NUMERICAL EXAMPLES
The performance of the proposed model is demon-strated by comparing the experimental results pre-sented by Montanez (2002) and Kenai et al. (2006).Figures 1 and 2 show the compaction curves for sand-bentonite mixture with bentonite content of 5% and15% by weight. Montanez’s experimental data presentvalues for theStandardProctorTest (BS, external grossenergy input = 637 kJ/m3 or kPa). In Figures 1 and2, two model predictions are also shown. The curvesshown by dashed lines represent the static compactioncurve predicted by the model for undrained (air/water)loading up to external quasi-static pressure, σy, of637 kPa. The curves shown by solid lines are for equalenergy input, calculated by integrating the appliedstress σy with respect to volumetric strain. The actualenergy input into the soil was computed on the basisof the values applicable at the optimum water con-tent, which were found to be 16 kJ/m3 and 18 kJ/m3
respectively.
Figure 1. Comparison of predicted and experimental com-paction curves for well graded sand with 5% bentonite (afterMontanez, 2002).
Figure 2. Comparison of predicted and experimental com-paction curves for well graded sandwith 15%bentonite (afterMontanez, 2002).
Figure 3 shows an example of compaction curvefor clay sandy soil (liquid limit = 39%, plasticityindex = 15%) adopted from Kenai et al. (2006).Experimental results shown in figure are for static(σy = 2100 kPa) and dynamic (external gross energyinput 3000 kJ/m3) compaction tests. Both predictedcompaction curves are produced from quasi-staticcompaction up to external pressures, σy, of 2100 kPaand 4000 kPa respectively.
Model parameters used for prediction of the abovecompaction curves are shown in Table 1, 2 and 3.Initial pore air pressure (ua0) is taken to be equalto atmospheric pressure (101.3 kPa). For a certainsoil, a lower initial porosity was assumed and thecomputations were performed for a range of mois-ture contents which also define the values of initialdegree of saturation (S0). The water solubility value isadopted from Fredlund & Rahardjo (1993). The val-ues of λvp (slope of the NCL) are selected to best fitthe experimental results and compared with the mea-sured values from literature. These values are foundto be generally in the range of experimentally mea-sured values. Table 2 shows the initial equilibriumsuctions measured for compacted specimens at differ-entmoisture contents given byMontanez (2002). Theyare presented as constant suction contours which are
Figure 3. Comparison of predicted and experimental com-paction curves for clay sandy soil (after Kenai et al., 2006).
Table 1. Parameter values for the proposed model.
Well graded sand with Well graded sand5% bentonite with 15% bentonite
Parameter Value Value
h∗ 0.02 0.02λvp 0.045 0.13n0 34 % 36 %Gs 2.656 2.660
∗Water solubility
377
209
Table 2. Initial matric suction (s0) values.
Well graded sand with Well graded sand with5% bentonite content 15% bentonite content
w (%) s0 (kPa) w (%) s0 (kPa)
3.9 4630 4.9 178005.8 1130 6.4 115007.8 350 8.5 25509.7 130 10.6 126011.7 54 12.7 85013.6 32 14.9 53015.6 26 17.0 29017.5 22 19.1 270
Table 3. Parameter values for the proposed model.
Clay sandy soil
Parameter Value
h∗ 0.02λvp 1.3n0 46 %Gs 2.66
generally perpendicular to the water content axis giv-ing approximately constant suction values for a givencompaction water content. This ignores the curving ofthese contours close to saturation towards the left even-tually becoming almost parallel to the full saturationline.
Figures 1 and 2 show comparisons between exper-imental and predicted values of compaction curvesfor sand-bentonite mixtures. The experiments wereperformed under dynamic conditions (Proctor com-paction), whereas model prediction assumed staticundrained conditions for both air and water. Despitethese differences, it is clear that reasonable predictionsof the shape of the compaction curve can be obtainedwith the proposed approach.
Figure 3 shows comparison between experimen-tal and predicted values of the compaction curve forsandy clay soil. For this example, experiments wereperformed under both dynamic and static conditions.It should be noted that in this example, experimentalresults did not include the initial suction values. There-fore, initial suction values are assumed to be same aswell graded sand with 15% bentonite (Table 2).
Differences in the predicted and experimentalbehaviour can be traced to a number of sources. Onepossibility is the drainage of air, particularly on thedry side of the optimum, which can lead to higherdry densities. This analysis, however, shows that the
Figure 4. Variations of ms1 with initial degree of saturation
(S0) and during compaction for well graded sand with 5%bentonite.
development of air pressure on the dry side is not verysignificant, but will depend on ms
1. It can be seen thatthe predicted and experimental density difference onFigure 1 is higher than that of Figure 2. In additionto the likely influence of the other assumptions madein the analysis, this difference seems to indicate thatthe drainage of air may lead to higher densities in thedry side. It is likely that the pore sizes in well gradedsandwith 5%bentonite content is higher than the samesand with 15% bentonite. These issues will be furtherexamined through future targeted experiments.
Figure 4 shows the variation of ms1 with initial
degree of saturation (S0) and during compaction forthe well graded sand with 5% bentonite content. It canbe seen that coefficient of compressibility due to netstress (ms
1) decreases with decreasing initial degreeof saturation (S0), as assumed previously by Kuru-cuk et al. (2007). However, during the compactionprocess, the degree of saturation increases from the ini-tial value, but the coefficient of compressibility (ms
1)decreases. This decrease of compressibility happensowing to the increase of net stress as the compactionprogresses. It is also apparent that much of the com-paction takes place in the early part of the processwhere the soil compressibility decreases rapidly.
5 CONCLUSION
This paper presents theoretical concepts to predictthe compaction curve for soil during undrained K0or isotropic loading using unsaturated soil mechan-ics principles. It highlights the fact that the well-known inverted parabolic shape of the compactioncurve may be theoretically predicted using unsatu-rated soil mechanics principles, arguably for the firsttime in literature. This was demonstrated using pub-lished experimental results, but it was necessary tomake some assumptions. The controlling parameter
378
210 APPENDIX B. PUBLICATIONS
governing the compaction process was identified asthe coefficient of compressibility with respect to netstress or ms1. It was also identified that the variationin drainage conditions during compaction may influ-ence the results. Future experiments will be targetedto develop a comprehensive set of data to examinethe modelling assumptions and improve modellingcapability.
ACKNOWLEDGEMENTS
Thanks are rendered to Monash University for pro-viding a Monash Graduate Scholarship and financialassistance to the first author for her PhD candidature.
REFERENCES
Barden, L. & Sides, G.R. 1970. Engineering behaviour andstructure of compacted clay. Journal Soil Mechanics andFoundations Division, ASCE, 96, No. SM4: 1171.
Fredlund, D.G. & Rahardjo, H. 1993. Soil mechanics forunsaturated soils. John Wiley & Sons, Inc.
Fredlund, D.G. & Morgenstern, N.R. 1976. Constitutiverelations for volume change in unsaturated soils. CanadianGeotechnical Journal, 14, 3: 261–276.
Hilf, J.W. 1948. Estimating construction pore pressuresin rolled earth dams. Proceedings of 2nd InternationalConference in Soil Mechanics and Foundation Engineer-ing, 3: 234–240. Rotterdam, The Netherlands.
Hilf, J.W. 1956. An investigation of pore water pressures incompacted cohesive soils. Technical Memorandum 654,U.S. Department of the Interior, Bureau of Reclamation,Denver, Colorado.
Hogentogler, C.A. 1936. Essentials of soil compaction.ProceedingsHighwayResearchBoard, National ResearchCouncil, Washington, D.C., 309–316.
Kenai, S., Bahar, R. & Benazzoug, M. 2006. Experimentalanalysis of the effect of some compaction methods onmechanical properties and durability of cement stabilizedsoil. Journal of Material Science, 41: 6956–6964.
Kurucuk, N., Kodikara, J. & Fredlund, D.G. 2007. Pre-diction of compaction curves. 10th ANZ Conference onGeomechanics, 2: 115–119.
Lambe, T.W. 1960. Structure of compacted clay. Transac-tions, ASCE, 125: 682–705.
Lee, D.Y. & Suedkamp, R.J. 1972. Characteristics of irregu-larly shaped compaction curves of soil. HighwayResearchBoard, 381: 1–9.
Li, Z.M. 1995. Compressibility and collapsibility of com-pacted unsaturated loessial soils. Unsaturated Soils. Proc.1st Int. Conf. on Unsaturated Soils (UNSAT 95), Paris,France (ed. Alonzo, E.E. and Delage, P.), Rotterdam:Balkema, Vol. 1: 139–144.
Lloret, A., Villar, M.V., Sanchez, M., Gens, A., Pintado, X.& Alonso, E.E. 2003. Mechanical behaviour of heav-ily compacted bentonite under high suction changes.Géotechnique, 53(1): 27–40.
Montanez, J.E.C. 2002. Suction and volume changes of com-pacted sand-bentonite mixtures. PhD thesis, University ofLondon, Imperial College of Science, London, England.
Olson, R.E. 1963. Effective stress theory of soil compaction.Journal SoilMechanics and FoundationsDivision, ASCE,89, No. SM2: 27–45.
Proctor, R.R. 1933. Fundamental Principles of Soil Com-paction, Engineering News-Record, 111: 286.
Sheng, D., Fredlund, D.G. & Gens, A. 2007. A new mod-elling approach for unsaturated soils using independentstress state variables. Research Report No. 261.11.06,University of Newcastle, NSW 2308, Australia.
379
211
B.3
Unsaturated Soils – Buzzi, Fityus & Sheng (eds)© 2010 Taylor & Francis Group, London, ISBN 978-0-415-80480-6
Evolution of the compaction process: Experimental study – preliminaryresults
N. Kurucuk & J. KodikaraDepartment of Civil Engineering, Faculty of Engineering, Monash University, Clayton, Australia
D.G. FredlundGolder Associates, Saskatoon, Canada
ABSTRACT: Fundamental research on the compaction characteristics and the evolution of the compactionprocess is limited in literature, particularly froma quantitative standpoint. Improved understanding of unsaturatedsoil mechanics over the last few decades has allowed more detailed study of the driving mechanism of thecompaction process. This paper presents an approach to predict the soil compaction curve during static loadingusing unsaturated soil mechanics principles. Particular attention is given to the effect of suction on the overallcompaction process. The results obtained by the model are compared with the experimental results obtainedusing a special oedometer. Sand-bentonite mixtures were prepared with a range of initial water contents. Thesoils were statically compacted in the special oedometer. Soil samples were compacted with vertical stressesranging between 80 kPa and 1600 kPa. The initial suction was measured and controlled using the axis translationtechnique. A pressure transducer was also installed on the lower part of oedometer and was connected to thewater under the ceramic disk. It was possible to measure pore-water pressure changes (and therefore suctionchanges) during the loading process. Some results are presented and discussed in the paper.
1 INTRODUCTION
Soil compaction is widely used in geo-engineeringand is important for the construction of roads, dams,landfills, airfields, foundations, hydraulic barriers,and ground improvements. Compaction is appliedto the soil with the purpose of finding the optimumwater content corresponding to themaximum dry den-sity. Compaction decreases the soil’s compressibility,increases its shearing strength, and in some cases,reduces its permeability. Proper compaction of mate-rials ensures the durability and stability of earthenstructures.A typical compaction curve presents different den-
sification stages when the soil is compacted withthe same apparent energy input but different ini-tial water contents. The water content at the peakof the curve is called the optimum water content(OWC) and represents the water content at whichdry density is a maximum for a given compactionenergy.Since Proctor’s pioneering work in 1933, many
researchers have attempted to explain qualitativelythe leading mechanisms in the densification stages,mainly on the dry side of optimum water content.The compaction curve has been explained in termsof capillarity and lubrication (Proctor, 1933), vis-cous water (Hogentogler, 1936), pore pressure theory
in unsaturated soils (Hilf, 1956), physico-chemicalinteractions (Lambe, 1960), and the concept of effec-tive stress theory (Olson, 1963). More recently,Barden & Sides (1970) undertook an experimentalstudy on the relationship between engineering perfor-mance of compacted unsaturated clay andmicroscopicobservations of clay structure. Lee&Suedkamp (1972)also conducted research on the shape of the com-paction curve for different soils.Despite this research work, and the importance and
high usage of the compaction process in engineeringpractice, it still remains that the compaction of soilis quite complex and not well explained, particularlyfrom a quantitative standpoint. Theoretical modelingof soil compaction can provide a better understandingof themain parameters that affect the shape of the com-paction curve. It will also provide a better understand-ing of the behavior of compactedmaterials. Therefore,there is need for research to be performed at a funda-mental level in order to understand the compactioncharacteristics of soil and the inverted parabolic shapeof the compaction curve.An approach to predict the soil compaction curve
during static loading was developed and presented inprevious works by the authors (Kurucuk et al., 2007and Kurucuk et al., 2008). The approach was sup-ported by data from the literature. This paper presentsan experimental study conducted in order to compare
887
212 APPENDIX B. PUBLICATIONS
with the proposedmodel. Predictions by the model arecompared with the experimental data presented in thispaper.
2 FORMULATION OF THE MODEL
Theoretical concepts utilized for the development ofsoil compaction curves are presented in this section.Initially, Hilf’s (1948) approach for pore pressuredevelopment is presented and this is followed withthe Fredlund & Morgenstern’s (1976) volume changetheory for a compacted soil.
2.1 Pore pressure development during staticcompaction
One of the primary aspects associated with the gen-eration of the compaction curve is the simulation ofpore pressure development. Hilf (1948) developed arelationship between pore pressure and applied stress,which is based on one-dimensional K0 soil compres-sion, Boyle’s law, and Henry’s law, and is expressed asfollows:
ua =1
1 +(1 − S0 + hS0) n0(ua0 + ua)mv
σ y (1)
where; ua = change in absolute pore air pressure,S0 = initial degree of saturation, h = coefficient ofsolubility, n0 = initial porosity, ua0 = initial absoluteair pressure, mv = coefficient of volume change insaturated soil, and σ y = change in applied verticalstress.Hilf (1948) developed this equation assuming that
air and water phases were undrained, and volumereduction was due to air dissolving in the water andthe compression of free air. Both liquid and solidphases were considered to be volumetrically incom-pressible. It was also assumed that the change in poreair pressure was equal to the change in pore waterpressure, and therefore, matric suction change wasinsignificant. Experimental results on suction changeduring compaction can be found in literature (e.g., Li,1995; Montanez, 2002). It is shown that matric suc-tion only decreases marginally with a density increaseand may be approximated as a constant value. There-fore, Hilf’s assumption that suction remains constantduring compaction appears to be close to the reality.Further justification for assuming that suction remainsconstant during compaction test has been presented inKurucuk et al. (2007).
2.2 Computation of volume change and dry density
The volume change constitutive relationship corre-sponding to K0 loading can be defined in terms of two
independent stress variables as proposed byFredlund &Morgenstern (1976) for unsaturated soils.The relationship will be used for the calculationsassociated with the compaction curves:
εv =VvV
= ms1( σ y − ua) + ms2 (ua − uw) (2)
where; εv = volumetric strain, Vv = overall volumechange of soil element, V = initial total volume of soilelement, ms1 = compressibility of soil particles withrespect to net stress (σy − ua), ms2 = compressibilityof soil particles referenced to matric suction (ua− uw),( σ y − ua) = change in net stress, and ( ua − uw) =change in soil suction.Since soil particles are incompressible, it is accepted
that deformation is primarily due to compression ofthe pore fluid (i.e., the air and air/water mixture). Theindependent stress state variable concept is utilized inthe derivation; namely, net stress (σy − ua) (causes areduction in volume with compression), and matricsuction stress (ua − uw) (generally results in volumeincrease with compression). Once the overall volumechange is computed, the corresponding dry densitycan be computed.
2.3 Modelling assumptions
Kurucuk et al. (2007) showed that the assumptionof a constant coefficient of compressibility duringcompaction does not produce a proper shape of thecompaction curve, particularly on the dry side of opti-mum water content. The analysis showed that it iswas ms1 that controlled volume changes during com-paction because the associated change in suction maybe neglected. The parameter ms1 was represented as afunction of saturation and decreased with decreasingsaturation. However, the experimental results pre-sented byLloret et al. (2003) showed thatms1 decreasedwith both suction and net stress. Therefore, followingthe functional form suggested by Sheng et al. (2008),the volumetric strain, ignoring suction change, can bepresented as:
εv =dV
V= λvp
d (σnet − ua)
(σnet − ua) + s0(3)
where; εv = volumetric strain, (σnet − ua) = meannet stress, ua = pore air pressure, s0 = suction, λvp =slope of the normal compression line (NCL) of thesaturated soil, and V = initial total volume of the soilelement.This gives the following equation for ms1 as:
ms1 =λvp
(σy − ua) + s0(4)
888
213
This assumption will be used and discussed furtherin the modeling of the compaction curve. It is reason-able to replacemv in Equation 1 withms1. A numericalexample of the variation of ms1 during the compactionprocess is given in the following section. Equations(1), (3) and (4) were used in incremental forms to com-pute the incremental and total volume change and thecorresponding dry density values during compaction.
3 EXPERIMENTAL PROGRAM
Experimental tests were performed in a specialoedometer for comparison with the model predictions.Material characteristics, the experimental device,details of the performed tests and the steps followedare explained in the following sections.
3.1 Material characteristics
The material selected for this research is a sand-bentonite mixture. Bentonite in the amount of 2% byweight was mixed with poorly graded sand to producea material with some fines. The fines allow the genera-tion of higher initial suctions. The amount of bentonitemixed with sand is unlikely to produce some swellingin the soil sample (sand particles are in contact evenfor high water contents) (Montanez, 2002). Particlesize distributions (PSD) for the sand and bentonite areshown in Figures 1 and 2.
The bentonite used for the experiments is commer-cially available as ActiveGel 150 bentonite. Atterberglimits of this clay are given in Table 1.
The specific gravity of the sand, bentonite and themixture was measured using a pycnometer (Micro-meritics AccuPyc 1330). Measured values are pre-sented in Table 2.
Dynamic compaction tests were conducted accord-ing to the procedures described in Australian Standards(AS 1289.5.1.1-2003). Figure 3 shows the compactioncurve of sand-bentonite mixture.
0
10
20
30
40
50
60
70
80
90
100
0.010.11
Diameter (mm)
% P
ass
ing Cu = 1.71
Cc = 1.07
Figure 1. PSD for sand.
0
10
20
30
40
50
60
70
80
90
100
1101001000
Particle Size ( m)
%P
assi
ng
Figure 2. PSD for bentonite (adopted from Shannon, 2008).
Table 1. Atterberg limits.
Liquid limit (%) 550Plastic limit (%) 36Plasticity index (%) 514
Table 2. Specific gravity of the soil.
Soil type Gs
Sand (SP) 2.65Bentonite 2.69Mixture 2.65
1.66
1.68
1.70
1.72
1.74
1.76
2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0
Water content (%)
Dry
den
sity
(t/
m3)
Sr = 80%
Figure 3. Standard compaction curve of sand-bentonitemixture.
3.2 Material preparation
Sand and bentonite were mixed in an air dry statewith a mixer for 15 minutes initially. The desiredamount of water was then added and mixed for another20 minutes to evenly distribute the moisture. The mix-ture was then wrapped with several layers of plasticwrap and aluminum foil. All the samples were placedin air tight plastic zip bags and stored in contain-ers in the moisture room for a minimum of 28 days.Montanez (2002) conducted a study to investigate theinfluence of the hydration time on the soil specimens.
889
214 APPENDIX B. PUBLICATIONS
He stated that it is necessary to allow a minimumperiod of 28 days for hydration in order to establish anequal distribution of suction within the material.
3.3 Experimental apparatus
Static compaction experiments were conducted withFredlund’s SWCCDevice (Figure 4). Devicewasmod-ified by adding pressure transducer to the bottom ofthe plate (below the ceramic disk) in order to conductthe desired experiments (i.e., constant water contenttest). A Druck—PDCR 4011 pressure transducer wasused. This modification helped to measure the initialsuction values of the prepared specimens and alsomea-sure changes in suction with loading. Volume changesin the specimen were tracked using a linear variabledifferential transformer (LVDT-KYOWADTH-A-10).
3.4 Methods
Hydrated soil samples were placed in a metal ring byhand tapping (loose) to ensure similar initial void ratiosfor all samples prepared with different water contents.The metal ring with soil was then placed after on theceramic disk inside the experimental device.Before starting the loading process, the initial suc-
tion of the soil specimen was measured. Air pressurewas applied to the specimen and the water pressurebelow the ceramic disk was monitored with the helpof the pressure transducer. The applied air pressurewas gradually changed after equilibrium was reached(i.e., there was no changes in water pressure below theceramic disk). Final water pressure values are plottedagainst the applied air pressure. The function of thiscurve is a 45◦ line (taking into account theHilf’s (1948)approach for undrained systems, ua = uw). Theinitial suction was determened from the intersection ofthe curve with the abscissa of the graph (Fredlund &Rahardjo, 1993).
Figure 4. Experimental SWCC pressure plate device.
Table 3. Loading steps.
Load (kPa) No of increments Increment (kPa)
0− 640 8 80640− 1280 4 1601280− 2250 3 320
uu a-u w
Figure 5. Stress path followed (constant water content test)(small suction change is exaggerated).
Table 4. Suction change in first load-ing step for different water contents.
Water content (%) ( ua − uw) (kPa)
6 1.18 0.810 0.3
Soil samples were statically compacted using ver-tical stresses ranging between 80 kPa and 2250 kPa.Loading steps followed are given in the Table 3. Theduration of the loading time was changed according tothe time required at each step for stabilization of thesuction and volume change. There are similar loadingrate approaches presented in the literature (e.g., Tanget al., 2008).Figure 5 shows the stress path followed during the
compression process. Since the tests conducted areconstant water content tests, there should be a decreasein suction with increasing net stress. However, con-sidering the soil type used and taking into account theexperimental results, it is noted that the change in suc-tion with loading is very small (Table 4). Moreover,suction changes can only be seen in the first loadingstep and suction stays constant during the followingloading steps. Therefore, as stated in Section 2.3, itappears to be acceptable to assume the suction remainsconstant with loading. In addition during loading, the
890
215
soil passes through plastic yielding with the increase ofthe yield stress concurrently with the applied loading.
4 RESULTS AND DISCUSSION
Results from the conducted experiments are presentedin this section. These results are compared with thepredictions produced by the proposed model.
4.1 Conducted experiments
Static undrained (water) compaction tests were con-ducted on 5 samples with different water contents andsimilar initial densities. The properties of the spec-imens prepared for static compaction test are givenin Table 5. The water content of each specimen wasmeasured before and after the test and also presentedin Table 5. It is clear that the water contents remainedalmost constant during the tests (Note: it is a constantwater content test but during the preparation or dis-mantling stage the specimen might take on water fromthe ceramic disk).
Specimens prepared with 6% and 8% water contentwere tested two times in order to ensure the repeatabil-ity of the test. An important point to be noted here isthat every sample was assumed to be identical but therewere slight differences in initial properties (Table 5),which would present some dissimilarity between tests.Taking into account this fact, the results show that thetests were repeatable.
4.2 Test results
Figure 6 shows the changes in specific volume(1 + e) with net stress (σy − ua). It can be seen thatthere is a general trend that specific volume is lowerfor soils with a higher water contents.
The variation of coefficient of compressibility (ms1)
with net stress is shown in the following figure(Figure 7). The compressibility coefficient (ms
1)increases with increasing water content at a particularnet stress. For example, the highest coefficient of com-pressibility at 1000 kPa net stress is for the specimen
Table 5. Specimen properties.
Water contentInitial drydensity Before test After test Suction
Test (t/m3) (%) (%) (kPa)
Test1 (6%) 1.44 5.56 6.93 7.5Test1 (8%) 1.46 7.62 8.37 5.34Test1 (10%) 1.46 9.82 10.67 4.95
1.6
1.65
1.7
1.75
1.8
1.85
1.9
1 10 100 1000 10000
Net Stress + Suction (kPa)
Sp
ecif
ic v
olu
me
test1 (6%)test1 (8%)test1 (10%)
Figure 6. Changes in specific volume with loading andwater content.
0.0E+00
2.0E-05
4.0E-05
6.0E-05
8.0E-05
1.0E-04
1.2E-04
1.4E-04
1.6E-04
1.8E-04
2.0E-04
0 500 1000 1500 2000 2500
Net Stress (kPa)
m1s
-co
eff.
of
com
p. (
1/kP
a) test1 (6%)test1 (8%)test1 (10%)
Figure 7. Changes in ms1 with loading and water content.
prepared with 10% water content and the lowest coef-ficient of compressibility is for the specimen preparedwith 6% water content.
Figure 8 presents the changes in slope of the com-pression curve (λ) with net stress (σy − ua). The slopeof the compression (λ) curve is again higher for higherwater contents. Moreover, the slope has a tendency todecrease with increasing net stress (σy − ua) exceptat higher stresses at which slope tends to increaseagain. The slope of the compression line (λ) is alsodecreasing with an increasing suction (ua −uw) for thesame net stress (σy −ua). This observation is in agree-ment with the model proposed by Alonso et al. (1990),generally referred as Barcelona Basic Model (BBM),and Sheng et al. (2008). Some researchers have indi-cated an increase in slope with increasing suction (e.g.Monroy et al., 2008) and some point out initially anincrease and after decrease with suction (e.g. Wheelerand Sivakumar, 1995, Estabragh et al., 2004, Cuisinerand Masrouri, 2005).
The variation in degree of saturation with compres-sion is shown in Figure 9. As it is expected, there isan increase in degree of saturation with increasingnet stress. However, it should be noted that all threecases are well below saturation and saturation was notreached at the end of compression.
Finally, the change in dry density with compressionand different water contents is shown in Figure 10. Thespecimens compacted with higher water contents are
891
216 APPENDIX B. PUBLICATIONS
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0 500 1000 1500 2000 2500 3000
Net Stress (kPa)
Slo
pe
of
the
com
pre
ssio
n c
urv
e test1 (6%)test1 (8%)test1 (10%)
Figure 8. Changes in the slope of the compression line withloading and water content.
12
17
22
27
32
37
42
47
0 500 1000 1500 2000 2500
Net Stress (kPa)
Sr
(%)
test1 (6%)test1 (8%)test1 (10%)
Figure 9. Changes in degree of saturation with compressionand water content.
1.4
1.45
1.5
1.55
1.6
1.65
1.7
0 500 1000 1500 2000 2500
Net Stress (kPa)
Dry
den
sity
(t/
m3)
test1 (6%)test1 (8%)test1 (10%)
Figure 10. Changes in dry density with compression andwater content.
more compressible and therefore, result in higher finaldensities.
4.3 Discussion
The static compaction curve obtained during the exper-imental program and the predicted compaction curveis presented in Figure 11. All initial parameters areobtained from the conducted experiments except theslope of the normal compression line is for saturated
1.5
1.54
1.58
1.62
1.66
4 5 6 7 8 9 10 11
Water content ( %)
Dry
den
sity
(t/
m3)
Experimental results (560 kPa)Model prediction (560 kPa)
Figure 11. Static compaction curve.
soil (λvp). Further oedometer test needs to be com-pleted to determine this parameter more accurately.In this paper, an approximate value (assuming thatslope of the normal compression line for saturatedsoil is higher than unsaturated soil) from the testsconducted for unsaturated soils is estimated (λvp =0.028).
It can be seen from Figure 11 that the predictedand experimental compaction curves are in reason-able agreement. There are slight differences betweenpredicted and experimental values for the specimensprepared at 6% and 8% water contents. This can beexplained using two different hypotheses.
First, as it can be seen from Equation 3 and 4 that themain controlling parameters for deformation is suc-tion (ua − uw) and the slope of the compaction curvefor saturated soils (λvp). In contrast to Alonso et al.(1990), Sheng et al. (2008) introduced the use of aconstant parameter for the slope of the normal com-pression line which is measured from the saturated soilspecimens (λvp) and the resulting deformation due tonet stress is a function of suction and λvp as shown inEquation 3. In addition to that, during the generationof the compaction curve for a specific type of soil,the differences in deformation between two differentwater contents were only assumed to be a function ofsuction (Equation 4). Taking into account the granularnature of the soil used, it is clear that the suction differ-ences between the two different water contents will bevery small and this leads to negligible density differ-ences. Hence, it may be stated that better results canbe obtained for fine soils than coarse soils. However,the proposed model needs to be modified with differ-ent constitutive soil model assumptions from literature(such as BBM) and compared for clarification.
The second possible explanation could be that suc-tion is assumed to stay constant during loading andit is actually slightly changing as shown in Table 4.The effect of suction (as a second component in Equa-tion 2) on the compaction curve is shown in Kurucuket al. (2007). It was stated there that taking into accountthe changes in suction will result in lower dry densities.Moreover, it can be seen in Table 4 that the changes insuctions are higher for lower water content. Therefore,
892
217
a bigger difference is anticipated between the pre-dicted and experimental results in lower water contentsoils. This difference appears to reducewith increasingwater content.
5 CONCLUSION
This paper presents experimental study conducted inorder to evaluate a proposed theoretical model forthe compaction curve. The proposed model predictsthe compaction curve for soil during undrained K0 orisotropic loading using unsaturated mechanics princi-ples. It suggests that thewell-known inverted parabolicshape of the compaction curve may be theoreticallypredicted using unsaturated soil mechanics princi-ples. Obtained experimental results are comparedwithmodel predictions but it was necessary to make someassumptions. It is shown that the proposed model canpredict the compaction curve reasonably well. Futureexperiments will be targeted to develop a comprehen-sive set of data using different types of soil. Moreover,the model will be modified in the light of several con-stitutive models (e.g. Alonso et al., 1990, Wheeleret al. 2003) and the capability of the models will beexamined.
ACKNOWLEDGEMENTS
Thanks are rendered to Monash University for pro-viding a Monash Graduate Scholarship and financialassistance to the first author for her PhD candidatureand to Dr. Chaminda Gallage for his advice regardingexperimental setup.
REFERENCES
Alonso, E.E., Gens, A. and Josa, A. (1990). A constitutivemodel for partially saturated soils. Geotechnique, 40(3),405–430.
AS 1289.5.1.1 (2003), Soil compaction and density tests—Determination of the dry density/moisture content rela-tion of a soil using standard compactive effort, StandardsAustralia, Sai Global, Sydney.
Barden, L. & Sides, G.R. (1970). Engineering behaviour andstructure of compacted clay. Journal Soil Mechanics andFoundations Division, ASCE, 96, No. SM4: 1171.
Cuisinier, O. and Masrouri, F. (2005). Hydromechanicalbehaviour of a compacted swelling soil over awide suctionrange. Engineering Geology, 81, 204–212.
Estabragh, A.R. Javadi, A.A., and Boot, J.C. (2004). ‘‘Effectof compaction pressure on consolidation behaviour ofunsaturated soil.’’ Canadian Geotechnical Journal, 41,540–550.
Fredlund, D.G. & Rahardjo, H. (1993). Soil mechanics forunsaturated soils. John Wiley & Sons, Inc.
Fredlund, D.G. & Morgenstern, N.R. (1976). Constitutiverelations for volume change in unsaturated soils. Cana-dian Geotechnical Journal, 14, 3: 261–276.
Hilf, J.W. (1948). Estimating construction pore pressures inrolled earth dams. Proceedings of 2nd International Con-ference in Soil Mechanics and Foundation Engineering,3: 234–240. Rotterdam, The Nederlands.
Hilf, J.W. (1956). An investigation of pore water pressures incompacted cohesive soils. Technical Memorandum 654,U.S. Department of the Interior, Bureau of Reclamation,Denver, Colorado.
Hogentogler, C.A. (1936). Essentials of soil compaction.ProceedingsHighwayResearchBoard, NationalResearchCouncil, Washington, D.C., 309–316.
Kurucuk, N., Kodikara, J. & Fredlund, D.G. (2007). Pre-diction of compaction curves. 10th ANZ Conference onGeomechanics, 2: 115–119.
Kurucuk, N., Kodikara, J. & Fredlund, D.G. (2008). The-oretical modelling of the compaction curve1st EuropeanConference onUnsaturated Soils, Durham, UK, 375–379.
Lambe, T.W. (1960). Structure of compacted clay. Transac-tions, ASCE, 125: 682–705.
Lee, D.Y.&Suedkamp, R.J. (1972). Characteristics of irregu-larly shaped compaction curves of soil.HighwayResearchBoard, 381: 1–9.
Li, Z.M. (1995). Compressibility and collapsibility of com-pacted unsaturated loessial soils.Unsaturated Soils. Proc.1st Int. Conf. on Unsaturated Soils (UNSAT 95), Paris,France (ed. Alonzo, E.E. and Delage, P.), Rotterdam:Balkema, Vol. 1: 139–144.
Lloret, A., Villar, M.V., Sanchez, M., Gens, A., Pintado, X.&Alonso, E.E. (2003). Mechanical behaviour of heav-ily compacted bentonite under high suction changes.Geotechnique, 53(1): 27–40.
Monroy, R., Zdravkovic, L. and Ridley, A. (2008) Volumetricbehaviour of compacted London Clay during wetting andloading. 1st European Conference on Unsaturated Soils,Durham, UK, 315–320.
Montanez, J.E.C. (2002). Suction and volume changes ofcompacted sand-bentonite mixtures. PhD thesis, Univer-sity of London, Imperial College of Science, London,England.
Olson, R.E. (1963). Effective stress theory of soil com-paction. Journal Soil Mechanics and Foundations Divi-sion, ASCE, 89, No. SM2: 27–45.
Proctor, R.R. (1933). Fundamental Principles of Soil Com-paction, Engineering News-Record, 111: 286.
Sheng, D., Fredlund, D.G. & Gens, A. (2008). A new mod-elling approach for unsaturated soils using independentstress variables. Canadian Geotechnical Journal, 45,511–534.
Shannon, B. (2008). Laboratory modeling of unsaturatedwater flow and heave in reactive soils. Final Year ProjectReport, Monash University, Australia.
Tang, A.M., Cui, Y.J. and Barnel, N. (2008). Compression-induced suction change in a compacted expansive clay.1st European Conference on Unsaturated Soils, Durham,UK, 369–374.
Wheeler, S.J. and Sivakumar, V. (1995). ‘‘An elasto-plasticcritical state framefork for unsaturated soil.’’ Geotech-nique, 45(1), 35–53.
Wheeler, S.J., Sharma, R.S. and Buisson, S.R. (2003). ‘‘Cou-pling of hydraulic hysteresis and stress—strain behaviourin unsaturated soils.’’ Geotechnique, 53(1), 41–54.
893