Numerical Simulation of Dynamic

Compaction within the Framework of

Unsaturated Porous Media

PhD Thesis
By

Javad Ghorbani

Supervisors:
A/Prof Majidreza Nazem
Prof. John Carter

ARC Centre of Excellence for Geotechnical Science and Engineering

Centre for Geotechnical & Materials Modelling

August 2016
 DECLARATION

“I hereby certify that the work embodied in this Thesis is the result of original research

and has not been submitted for a higher degree to any other University or Institution.”

(signed)

II
 CONTENTS
DECLARATION ...................................................................................................................... II

LIST OF FIGURES .................................................................................................................VI

LIST OF TABLES ................................................................................................................. VII

ACKNOWLEDGEMENTS .....................................................................................................IX

NOTATION .............................................................................................................................. X

ABSTRACT ........................................................................................................................... XII

PREFACE .............................................................................................................................. XV

Chapter 1. Introduction .......................................................................................................... 1

1.1 Background and motivation ........................................................................................ 2

1.2. Objectives and scope ................................................................................................... 5

1.3. Layout.......................................................................................................................... 6

Chapter 2. Multiphase flow in deforming unsaturated media; governing equations and

finite element discretization ....................................................................................................... 7

2.1 Introduction ................................................................................................................. 8

2.2 Soil-water characteristic curve .................................................................................... 9

2.3 Governing equations ................................................................................................. 11

2.3.1 Volume fraction and effective density ............................................................... 11

2.3.2 Average effective stress concept in unsaturated soils ........................................ 12

2.3.3 Conservation of mass and momentum balance .................................................. 14

2.4 Numerical simulations............................................................................................... 18

2.4.1 Finite element discretisation of the governing equations .................................. 19

2.4.2 Time discretisation of the governing equations ................................................. 23

2.4.3 The generalised- . 23

2.5 Implementation and validation of the model in dynamic elasticity .......................... 25

Chapter 3. Numerical integration of constitutive model in large deformations analysis of

partially saturated soils ............................................................................................................ 33

III
3.1 Introduction ............................................................................................................... 34

3.2 Effective stress in partially saturated soil .................................................................. 34

3.3 Generalization of available constitutive models ....................................................... 38

3.4 Constitutive models of partially saturated soils ........................................................ 38

3.4.1 Constitutive models based on net stress and suction ......................................... 38

3.4.2 Constitutive models based on a single effective stress ...................................... 39

3.5 Description of the selected model ............................................................................. 40

3.6 Integration of the model considering the occurrence of large deformation .............. 43

3.7 Stress rate objectivity ................................................................................................ 46

3.7.1 Algorithm 1: The use of Jaumann stress rate in the integration scheme ........... 46

3.8 Elastic behaviour ....................................................................................................... 47

3.8.1 Algorithm2: Calculating the secant elastic stiffness matrix .............................. 47

3.9 Transition to irreversible response ............................................................................ 48

3.9.1 Algorithm 3.a Pegasus method for case 1 .......................................................... 50

3.9.2 Algorithm 3.b Pegasus method for case 2 ......................................................... 51

3.9.3 The drift correction ............................................................................................ 52

3.9.4 Algorithm 5: Yield surface correction scheme for general elasto-plastic models
54

3.10 The integration scheme ............................................................................................. 55

3.10.1 Algorithm 6: A Modified-Euler integration scheme with automatic substepping

and drift correction for large deformation analysis of partially saturated soils. ............... 55

3.11 Numerical examples .................................................................................................. 59

3.11.1 Finite strain analysis of a one-dimensional elastic soil layer ............................. 59

3.11.2 Cylindrical cavity expansion.............................................................................. 61

3.11.3 Rigid footing on unsaturated soil ....................................................................... 64

3.11.4 Comparison of the response of partially saturated soil under a rigid footing in
small and large deformation analyses ............................................................................... 67

3.12 Summary ................................................................................................................... 70

IV
Chapter 4. Literature review of dynamic compaction ......................................................... 73

4.1 Introduction ............................................................................................................... 74

4.2 Soil compaction ......................................................................................................... 74

4.3 History ....................................................................................................................... 76

4.4 Application of dynamic compaction to different soil categories .............................. 76

4.5 Prediction of ground response to dynamic compaction ............................................ 77

4.5.1 Empirical equations ........................................................................................... 78

4.5.2 Analytical solutions ........................................................................................... 80

4.5.3 Finite element simulations ................................................................................. 83

4.6 Dynamic compaction of partially saturated soils ...................................................... 87

4.7 Summary ................................................................................................................... 88

Chapter 5. Numerical simulation of dynamic compaction in partially saturated soils ........ 89

5.1 Introduction ............................................................................................................... 90

5.2 Modelling dynamic compaction in partially saturated soil ....................................... 90

5.2.1 Discussion of viscous boundaries ...................................................................... 90

5.3 Results and discussion............................................................................................... 94

5.3.1 Mechanical response of soil to dynamic compaction ........................................ 94

5.3.2 Changes in void ratio during dynamic compaction ......................................... 100

5.3.3 Other aspects of the mechanical response of soil ............................................ 105

5.3.4 Air pressure generated during impact .............................................................. 110

5.4 Reaching higher energy levels ................................................................................ 114

5.5 Summary ...................................................................................................................... 120

Chapter 6. Conclusion ....................................................................................................... 122

6.1. Concluding remarks .................................................................................................... 123

6.2. Recommendation for future work ............................................................................... 124

Appendix A- Modifications of stiffness matrix in Updated Lagrangian method .................. 126

Appendix B- ALE convection ............................................................................................... 128

V
Appendix C- Description of material law .............................................................................. 129

REFERENCES ...................................................................................................................... 134

LIST OF FIGURES
Figure 2-1. The 6-noded element used in this study. ............................................................... 19
Figure 2-2. Soil column and boundary conditions. .................................................................. 28
Figure 2-3. Prescribed pressure at the top of the column. ....................................................... 29
Figure 2-4. Comparison of finite element predictions of displacement at the top of the column
with those obtained by the analytical solution of Li and Schanz (2011) for an elastic soil with
a degree of saturation of 0.5. .................................................................................................... 29
Figure 2-5. Comparison of finite element predictions of displacement at the top of the column
with those obtained by the analytical solution of Li and Schanz (2011) for an elastic soil with
a degree of saturation of 0.9. .................................................................................................... 30
Figure 2-6. Excess water pressure at the bottom of the column in an elastic soil with a degree
of saturation of 0.9. .................................................................................................................. 30
Figure 2-7. Excess air pressure at the bottom of the column in an elastic soil with a degree of
saturation 0.9. ........................................................................................................................... 31
Figure 2-8. Change in degree of saturation at the bottom of the column in initial saturation
degrees of 0.9. .......................................................................................................................... 31
Figure 3-1. One-dimensional column and boundary conditions. ............................................. 60
Figure 3-2. Finite element solutions for small and large deformations and analytical solution.
.................................................................................................................................................. 61
Figure 3-3. Numerical results versus analytical solution for cavity expansion. ...................... 63
Figure 3-4. Normalised effective stress and pore pressure distributions. ................................ 64
Figure 3-5. Finite element model for the problem of a rigid footing. ...................................... 65
Figure 3-6.Normalised load-displacement curves for rigid footings on the same soil in fully
saturated and semi-saturated conditions. ................................................................................. 67
Figure 3-7. Normalised load-displacement curves for rigid footings on the same soil in small
deformation and large deformation analysis. ........................................................................... 71
Figure 4-1. Macrophotographs of (a) an uncompacted area and (b) a compacted area (Marsili
et al. 1998). .............................................................................................................................. 75
Figure 4-2. Two-dimensional finite element model used by Poran and Rodriguez (1992) ..... 84
Figure 4-3. Force-time curve used by Pan and Selby (2002). ................................................. 85
Figure 5-1. Finite element model for the problem of dynamic compaction. ........................... 91
Figure 5-2. Settlement of the ground versus applied energy. .................................................. 96
Figure 5-3. Variation of displacement of the ground with changes in pre-consolidation
pressure for an energy level of 245 kN.m. ............................................................................... 97
Figure 5-4. Variation of displacement of the ground with changes in M for an energy level of
245 kN.m.................................................................................................................................. 98

VI
Figure 5-5. Variation of displacement of the ground with changes in 𝜆0for an energy level
245 kN.m.................................................................................................................................. 99
Figure 5-6. Variation of displacement of the ground with changes in Poisson’s ratio for an
energy level of 245 kN.m......................................................................................................... 99
Figure 5-7. Comparison of changes in void ratio at 15 cm beneath the centre of impact in
different levels of energy. ...................................................................................................... 100
Figure 5-8. Changes in void ratio with depth along the axis-symmetric line with different
values of M. ........................................................................................................................... 101
Figure 5-9. Changes in void ratio with depth along the axis-symmetric line with different
values of the slope of loading line. ........................................................................................ 102
Figure 5-10. Changes in void ratio with depth along the axis-symmetric line with different
values of pre-consolidation pressure. ..................................................................................... 103
Figure 5-11. Changes in void ratio with depth along the axis-symmetric line with different
values of ν. ............................................................................................................................. 104
Figure 5-12. Changes in void ratio with depth along the axis-symmetric line with different
values of 𝜅. ............................................................................................................................. 105
Figure 5-13. Predicted ground settlement under the centre of impact for different degrees of
saturation. ............................................................................................................................... 106
Figure 5-14. Maximum vertical stress generated beneath the centre of impact at the end of the
analysis versus applied energy. .............................................................................................. 107
Figure 5-15. The generated excess pore pressure 1 m beneath the centre of impact in different
saturation degrees................................................................................................................... 108
Figure 5-16. The changes in suction in 1.8 m below the centre of impact in energy levels of
180, 245 and 320 kN.m. ......................................................................................................... 109
Figure 5-17. The changes in the permeability of water in 1.8 m below the centre of impact in
energy levels of 180, 245 and 320 kN.m. .............................................................................. 110
Figure 5-18. Normalised air pressure versus applied energy. ................................................ 111
Figure 5-19.Normalised air pressure versus bubbling pressure in the energy level of 720
kN.m....................................................................................................................................... 112
Figure 5-20. Normalised air pressure versus 𝜈𝑏 in the energy level of 720 kN.m. ................ 113
Figure 5-21. Comparison of displacement in time during the impact for two levels of energies
of 845 and 980 kN.m. ............................................................................................................ 118
Figure 5-22. Displacement versus energy predicted by the UL and ALE approaches. ......... 119

LIST OF TABLES

Table 2-1. Commonly used fitting equations for Soil Water Characteristic Curve. ................ 10
Table 2-2. Typical values of intrinsic permeability for various materials Pinder and Gray
(2008). ...................................................................................................................................... 11
Table 2-3. General material parameters according to Li and Schanz (2011). ......................... 27
Table 3-1 Some effective stress equations proposed for unsaturated soils .............................. 37
Table 3-2. Material parameters according to Meroi et al. (1995). ........................................... 60

VII
Table 3-3. Material parameters used in the problem of cylindrical cavity expansion. ............ 62
Table 3-4. Constitutive model and material parameters for the problem of a rigid footing. ... 66
Table 3-5. Water retention curve parameters for the problem of a rigid footing. .................... 67
Table 3-6. Constitutive model and material parameters for the problem of a rigid footing. ... 69
Table 3-7. Water retention curve parameters for the problem of a rigid footing. .................... 70
Table 3-8. Stress errors and Normalized CPU time with different values of STol in small
deformation analysis of the rigid footing problem. ................................................................. 72
Table 3-9. Stress errors and Normalized CPU time with different values of STol in large
deformation analysis of the rigid footing problem considering stress rate objectivity. ........... 72
Table 4-1. Values of 𝑛 proposed for different types of material under dynamic compaction . 79
Table 5-1. Constitutive model and material parameters for the problem of dynamic compaction.
.................................................................................................................................................. 95
Table 5-2. Water retention curve parameters for the problem of dynamic compaction. ......... 96

VIII
 ACKNOWLEDGEMENTS

During this research work, I was the recipient of the financial support of the ARC Centre of
Excellence for Geotechnical Science and Engineering (CGSE). This financial support is greatly
appreciated.

I would like to express my sincere gratitude to my advisors A/Prof Majidreza Nazem and Prof.
John P Carter for the continuous support of my PhD study and related research, for their
patience, motivation, and immense knowledge. Their guidance helped me in all the time of
research and writing of this thesis.

My special thanks are extended to the colleagues at the Geotechnical group of the University
of Newcastle for providing a warm and friendly environment.

I would like to thank my family: my parents and my sister for supporting me spiritually
throughout writing this thesis and my life in general.

Last but not the least; I owe the greatest debt of gratitude to my wife Sougol for her unending
support, encouragement, love and patience during my candidature.

IX
 NOTATION
A list of the most frequently used variables in the thesis is given below. Matrices and vectors
are shown in bold font with the superscript 𝑇 denoting the transpose of the matrix in this thesis.
Also, scalar variables are shown in italic font throughout this thesis. Variables with two indices
denote a second-order tensor.
𝜎𝑖𝑗′ Effective Cauchy stress tensor

𝜎𝑖𝑗 Total Cauchy stress tensor

𝜘 Biot’s parameter

𝛿𝑖𝑗 Kronecker delta

χ Bishop’s parameter

𝑝′ Effective mean stress

𝑞 Deviatoric stress

𝑔 Potential surface

𝑓 Yield surface

𝑛 Porosity

𝛼∗ Yield surface size for unsaturated conditions

The slope of the compression-recompression

𝜅
line

The slopes of the normal compression lines

𝜆
for unsaturated conditions

𝑒 Void ratio

𝜈 Poisson’s ratio

𝐃𝑒𝑝 Elasto-plastic constitutive matrix

𝜀̇𝑣𝑝 The rate of plastic volumetric strain

Δ𝜆 The plastic multiplier

𝐾𝑡 Bulk modulus of the porous medium

𝐾𝑠 Bulk modulus of the solid grains

X
𝐭̅ Traction forces matrix

𝐮 Displacement vector

𝐮̇ Velocity vector

𝐮̈ Acceleration vector

The vector of actual velocity of non-solid

𝐔̇𝛽 ( 𝛽 = 𝑤, 𝑔)
phases

𝐰̇𝛽 ( 𝛽 = 𝑤, 𝑔) Darcy velocity

𝑝𝑔 Pore air pressure

𝑝𝑤 Pore water pressure

𝑝𝑐 Suction

𝐍𝐮 Shape function for displacement

𝐍𝐩𝐜 Shape function for suction

𝐍𝐩𝐰 Shape function for water pressure

𝑆𝛽 ( 𝛽 = 𝑤, 𝑔) Saturation degree

𝑆𝑒 Effective degree of saturation

Relative permeability of the water phase and

𝑘𝑟𝛽 ( 𝛽 = 𝑤, 𝑔)
the air phase

Intrinsic or absolute permeability matrix of

𝐤𝑖𝑛𝑡
the soil

𝜂𝛽 ( 𝛽 = 𝑤, 𝑔) Viscosity of each fluid

XI
 ABSTRACT
Dynamic Compaction (DC) is conducted by dropping heavy weights on the ground surface. It
has the advantage of providing rapid improvement of the geotechnical properties of soil at
relatively large depths; hence, it can reduce the overall time of the soil improvement process
dramatically. Furthermore, this method does not usually need any off-site disposal of excess
materials.

Despite the wide use of DC as an economically attractive ground improvement technique and
the abundance of experimental data for dynamic compaction, few detailed predictive methods
for DC are available. There is a demand for a reliable theory to predict and explain the
complicated nature of DC. As previous numerical simulations were mostly unable to accurately
study the problem of dynamic compaction on unsaturated soils, the primary aim of the provided
thesis is presenting and implementing a general finite element framework to investigate the
response of partially saturated soils subjected to dynamic compaction. This would enhance the
understanding of this mechanism together with providing a useful tool for studying the
determinative parameters of this problem.

Partially saturated soils are characterised by the simultaneous deformation of a porous soil, and
its pore water and pore air and usually exhibit more complex behaviour than some other porous
materials, as they often experience non-linear plastic (irrecoverable) deformation. This makes
their computational modelling less straightforward than for some other coupled problems.

In recent decades numerical modelling the response of partially saturated soils to external
loadings within the so-called mixture theory has received increasing attention. In such analyses
the mixture-theory (with or without some simplifying assumptions, e.g., considering an iso-
thermal environment and immiscibility of fluid phases) has formed the basis for analysing
thermo-hydro-mechanical and even chemical coupling in partially saturated soils.

A challenge in modelling unsaturated soil as a three-phase material arises in solving the global
equations of motion, where the presence of different phases, together with the existence of
inertia forces in each phase, makes the solution of the coupled dynamic system computationally
demanding.

XII
The selection of a consistent constitutive model within the theory of mixtures, that can
incorporate suction forces into the description of stress, provides a further complication. The
necessity of such incorporation has frequently been reported in experimental studies of
unsaturated soils. Moreover, in cases of large deformations analyses, the infinitesimal strain
theory should be replaced by an algorithm which can explain the occurrence of large
deformations in the numerical framework. These require the adoption of a unique strategy for
integration of the constitutive model for an unsaturated soil.

This thesis will address the aforementioned challenges. The generalised-𝛼 method is
introduced and implemented for solving the global equations of motions and an explicit
integration strategy for large deformations analyses of partially saturated media is also
presented for performing the numerical integration of the mechanical constitutive model. A
series of numerical examples are given throughout the thesis for the purpose of validation and
showing the ability of the developed model to capture various aspects of the response of
partially saturated soils. Among others, the changes in pore water pressure, saturation degree,
pore air pressure, and the bearing capacity of soil at different water contents are considered.
Moreover, a viscous boundary was suggested and was implemented for damping reflected
waves from the surrounding boundaries during dynamic compaction and a generalisation of
ALE method is given for this type of analysis.

The numerical study on dynamic compaction provided at the end of this thesis indicates that
the ground settlement during the impact has a direct relationship with the slope of loading line
and the applied energy. An inverse relationship also is seen between the slope of the critical
state line, Poisson’s ratio and pre-consolidation pressure with the ground settlement. The role
of determinative parameters involved in this problem were further studied particularly on the
variation of void ratio with respect to the changes in the material parameters.

Moreover, it is illustrated that drying the soils decreases the ground settlement and the
generated excess pore pressure beneath the centre of the impact area.

The relationship between impact energy and the dictated changes in suction is also shown to
be inverse; that is applying higher energies further decreases the suction. In addition, the
procedure of change in the permeability of water phase in response to the external dynamic
loads is also illustrated. It is shown that upon the impact and the consequent reduction of
suction, the permeability of water phase will increase.

XIII
The massive energy release accompanied by the application of dynamic compaction makes it
almost impossible to install some measurement instruments below the impact areas. Moreover,
the traditional numerical approaches of modelling dynamic compaction mostly neglect the state
of partially saturated soils during the analysis. Therefore, the presented numerical model can
be advantageous over these methods in capturing some aspects of the response of partially
saturated soils subjected to dynamic compaction which cannot be measured in the field studies
or by the traditional numerical approaches.

Although in this thesis the simplifying passive air pressure assumption in the analysis of
partially saturated is avoided, a numerical study on the generated excess air pressure is
performed which indicates that the excess air pressure generated during the impact can be
considered to be fairly well below the atmospheric pressure. This suggests the possibility of
performing the simplified analysis with the passive air pressure assumption for this particular
problem with an acceptable accuracy. The finding can be beneficial for future analyses of this
problem as using the simplified analysis can be highly advantageous due to reduction of the
number of nodal degrees of freedoms, computational storage, the time of the analysis and the
complexity involved in the problem.

XIV
 PREFACE

The author performed the research work presented in this Thesis in the Discipline of Civil,
Surveying and Environmental Engineering, School of Engineering, at the University of
Newcastle, Australia, under the supervision of A/Prof Majidreza Nazem and Prof John P Carter
from October 2012 to April 2016. During the term of the candidature, a number of papers have
been prepared and published.

These are listed below:

Ghorbani, J., Nazem, M. and Carter, J. (2014). Application of the generalised-α method in
dynamic analysis of partially saturated media. Computer Methods and Recent Advances in
Geomechanics: 129-133.

Ghorbani, J., Nazem, M. and Carter, J. P. (2015). Numerical study of dynamic soil compaction
at different degrees of saturation. Proceedings of the International Offshore and Polar
Engineering Conference: 815-820.

Ghorbani, J., Nazem, M. and Carter, J. P. (2016). Numerical modelling of multiphase flow in
unsaturated deforming porous media. Computers and Geotechnics 71: 195-206.

Ghorbani, J., Nazem, M. and Carter, J. P. (To be published). Dynamic analysis of unsaturated
soils subjected to large deformations. ACCM 2015.

Ghorbani, J., Nazem, M. and Carter, J. Finite element algorithm for large deformations analysis
of partially saturated porous media (To be submitted).

XV
Chapter 1. Introduction

1
1.1 Background and motivation

Dynamic Compaction (DC) is known as one of the most cost-effective soil improvement
techniques. In this method the soil at the ground surface and at depth is compacted by
repeatedly dropping heavy weights on to the ground surface. Compared to other soil
improvement methods, an important advantage of DC is its ability to provide rapid
improvement of the geotechnical properties of soil at relatively large depths; hence, the overall
time of the soil improvement process reduces dramatically. Furthermore, this method does not
usually need any off-site disposal of excess materials.

At the beginning of the 20th century, the limited capability of early cranes was a major obstacle
for further development of DC, since the amount of energy that could be imparted to the soil,
by converting the initial potential energy of the falling mass to kinetic energy at impact with
the ground surface, was limited. However, later developments in this technique made it
possible to apply much higher energy to the ground allowing deep ground improvement. Since
then, dynamic compaction has been successfully used in a wide range of civil engineering
projects worldwide, including building structures, container terminals, highways, airports,
dockyards and harbours.

The initial application of dynamic compaction was in the improvement of granular soils.
However, it was soon realised that its application could be extended to cohesive soils,
particularly to those with good drainage conditions. Consequently, a large number of reports
on field studies and experimental results regarding soil behaviour under dynamic compaction
have been published in the literature.

Despite the abundance of these experimental data and field observations, few numerical
approaches can be found in the literature to effectively predict soil behaviour under dynamic
compaction. This is mainly due to the complexities involved in the problem, such as the
occurrence of very large irreversible deformations of the soil. Current methods for predicting
soil response when subjected to dynamic compaction are generally listed in three categories:

 Empirical equations

 Analytical solutions

 Numerical simulations

2
Since 1975, when the scientific application of DC was first introduced by Ménard Techniques
Limited (Ménard and Broise (1975)), the empirical equation proposed in their work has
remained most popular in practice, mainly because of its simplicity for estimating the depth of
improvement as a function of the applied energy. However, it is notable that the term “depth
of improvement” was not clearly defined by Ménard and Broise (1975). Moreover, after this
ambiguous definition was clarified in later work, e.g., Leonards et al. (1980), it was realised
that the depth of improvement also depends on the type of soil, in contrast to Menard’s simple
equation. By observing these limitations in the late 80s and early 90s, several researchers
attempted to provide an analytical solution for this problem, such as Gunaratne et al. (1996).
Different methods were developed with different simplifying assumptions. However, these
models could not progress beyond simple representation of the soil around the impact area by
linear or non-linear spring and dashpot systems. In addition, these solutions were only validated
for values of the impact energies adopted in the experimental tests. As the complexities in
finding an analytical solution proved intractable, alternative approaches were sought by
researchers. This was concurrent with rapid growth in the computational capability of modern
computers, which eventually provided the opportunity to employ numerical methods, such as
the finite element method, for simulating dynamic compaction. Since the early 90s several
finite element models have been developed for predicting the soil response when subjected to
dynamic compaction. Each of these models covers some aspects of the problem, namely
incorporation of advanced elasto-plastic constitutive models, large deformations, and mesh
refinement due to excessive deformation at higher energies. More recently, some researchers
studied the effectiveness of DC on saturated soils with high and moderate permeability, e.g.,
Ghassemi et al. (2008). However, in such simulations the effect of the degree of saturation on
the response of soils to compaction has been ignored. This effect has long been acknowledged
in the literature and an early experimental report regarding it can be traced back to the work of
Proctor (1933). However, one important reason for the lack of consideration of this effect in
numerical simulations may be the difficulties of modelling partially saturated soils, due in turn
to the extreme non-linearity involved in the problem. Addressing these difficulties is one of the
primary goals of this thesis.

In modelling dynamic compaction in partially saturated soils, the soil should be represented as
a three-phase porous medium. Particularly in the region near the source of energy release each
of the three phases will respond differently to the dynamic loading. Indeed, partially-saturated
porous media exhibit complex behaviour making the computational analysis less

3
straightforward. The presence of a non-wetting and a wetting phase together with the existence
of inertia forces in each phase makes the solution of the coupled dynamic system
computationally demanding. Moreover, large deformations often take place during dynamic
compaction; hence the popular infinitesimal strain theory cannot be employed for higher
impact loads. Also, finding an appropriate elasto-plastic constitutive model to reflect the
nonlinear behaviour of unsaturated soils under dynamic loads is a challenge.

In the past few decades, modelling multiphase flow through porous media has gained attention
because of its importance in the areas of underground natural resource recovery, waste storage,
petroleum reservoirs, soil physics, environmental remediation, liquefaction prevention, earth
dams, and soil improvement. The equations regarding the interaction of the solid and fluid
media were first developed for quasi-static conditions by Biot (1941), and were then extended
to dynamic problems also by Biot (1956). Later, the introduction of the so-called “mixture
theory” by Truesdell (1957) and subsequent extensions of this theory provided the opportunity
to establish a new basis for Thermo-Hydro-Mechanical-Chemical analysis of porous media
considering the existence of multiphase fluids.

Among the numerical approaches for analysing partially saturated porous media, the finite
element method has received much attention in recent decades. Following the works of Lewis
and Schrefler (1982) and Li and Zienkiewicz (1992), a number of finite element simulations
have appeared in the literature. Some of these methods depend on simplifying assumptions. A
common simplifying assumption is the so-called “passive gas phase”, i.e., the gas pressure is
assumed to remain constant at atmospheric pressure during the analysis, e.g., Sheng et al.
(2003), Sheng et al. (2008). However, Morel‐Seytoux and Billica (1985), Ehlers et al. (2004)
and Khoei and Mohammadnejad (2011) pointed out that a three-phase simulation generally
predicts more realistic results than a two-phase simulation. In this thesis this particular
simplifying assumption, i.e., the assumption of a passive gas phase, is specifically avoided.

Among the numerous finite element models of problems involving partially saturated soils,
those adopting the constitutive model proposed by Bolzon et al. (1996), the original Basic
Barcelona Model (BBM) suggested by Alonso et al. (1990) and its modified versions, such as
that proposed by Sheng et al. (2003), have gained popularity. Examples of using these models
in practice can be found in the works of Li et al. (1999), Khoei and Mohammadnejad (2011)
and Sheng et al. (2003).

4
The original BBM is not formulated on the basis of a single effective stress (Bishop 1960) and
this can cause inconsistency with the proposed numerical solution method. One solution for
this inconsistency is a modification of the original BBM and one such modification was
suggested by Sheng et al. (2003). In this thesis the model of Sheng et al. (2003) is employed
in the numerical analysis of unsaturated soils due to its consistency with the numerical method
adopted to represent the elasto-plastic response of unsaturated soils under monotonic loading.

1.2. Objectives and scope

The main goal of this research was to develop a fully-coupled model for the numerical
simulation of dynamic compaction in unsaturated soils. A numerical study was performed
within the framework of the Finite Element method. To achieve this goal, the macroscopic
representation of the governing equations of partially saturated soils was adopted to account
for the simulation of a three-phase material consisting of the soil skeleton (solid), water (liquid)
and air (gas).
The governing equations are based on the balance equations for the overall momentum balance
of the mixture, the mass balance of the liquid phase and the mass balance of the gas phase. The
hydraulic behaviour of the soil is described by the soil water characteristic curves. In addition,
phase changes and chemical reactions are not permitted. Such flow through porous media is
called “immiscible flow”, which is also a subject of this study.

The finite element code SNAC developed at the University of Newcastle was adopted as the
main numerical tool used in this study, and further developments have been implemented into
this code, including:

 Introduction of a fully coupled element with pore water pressure, suction and
displacement as the primary variables.
 Time discretization by the Generalized- method for the fully coupled dynamic
analysis of partially saturated soils.
 Introduction and development of a stress integration scheme for large deformations
analysis using the model proposed by Sheng et al. (2003).

The developed computer code has been tested against various analytical solutions throughout
this thesis. An extensive numerical study on dynamic compaction was performed using the
developed code.

5
1.3. Layout

This thesis contains 6 chapters. Chapter 1 provides an introduction to the research dissertation.

Chapter 2 starts with a review of the literature concerning the theory of mixtures and the
generalized Biot theory and its application to solving the problem of multiphase immiscible
flows in deforming porous media. Then the governing equation and the basic assumptions
behind deriving these equations are presented. This chapter continues by presenting the weak
form of the global equations and the finite element discretization of them. Implementation of
the Generalized- method is then introduced. The chapter ends with a validation of the
numerical implementation in elasticity using a recently proposed analytical solution.

Chapter 3 mainly discusses the stress integration of the chosen constitutive model in the
framework of the Updated Lagrangian method. The chapter begins by a review of the definition
of effective stress in partially saturated soil. A constitutive model is then introduced and its
integration algorithm is discussed in some detail. The chapter also includes some numerical
examples for the purpose of validation of the numerical approach and showing the performance
of the algorithm and the constitutive model in different scenarios.

Chapter 4 presents a literature review of previous numerical simulations, analytical methods

and experimental observations on dynamic compaction. The chapter also discusses the
shortcoming and the drawbacks of the previous methods of simulating dynamic compaction.

In Chapter 5 the main focus is on the numerical modelling of dynamic compaction. Also,
sensitivity analyses on the parameters involved in this problem are given in this chapter.

In Chapter 6 a summary of the presented work is provided together with some

recommendations for the future research.

6
Chapter 2. Multiphase flow in deforming unsaturated
media; governing equations and finite element
discretization

7
 2.1 Introduction

The governing equations of solid and fluid interaction in a porous medium were first developed
for quasi-static situations by Biot (1941), who later extended his analysis to dynamic problems
(Biot (1956)). The subsequent introduction of ‘mixture’ theory by Truesdell (1957) and
improvements to this theory by researchers such as Green and Naghdi (1967), Barden et al.
(1969) and Bowen (1967) paved the way for the establishment of a new basis for the Thermo-
Hydro-Mechanical-Chemical analysis of porous media, allowing for the existence of
multiphase pore fluids. This made it possible to incorporate some advanced features of multi-
phase material response, including phase changes, chemical reactions, and behaviour under a
non-isothermal environment, into the analysis of porous media.

The extension of Biot’s theory to partially saturated soils was presented by Li et al. (1989) and
Li and Zienkiewicz (1992). In their research it was assumed that no phase transfer and no
chemical reactions were possible during the fluid flows. This kind of flow through porous
media is called “immiscible flow”, and is the subject of this research.

In recent decades numerical modelling of multiphase flow in deforming porous media has
received increasing attention in various areas of engineering, such as biomechanics and
geotechnics. The mixture theory has mostly been employed as the main framework for
analysing problems involving thermo-hydro-mechanical and even chemical coupling in porous
media. Partially saturated soils can be viewed as a subclass of this type of problem, since they
are characterised by the simultaneous deformation of a porous soil, and its pore water and pore
air.

However, partially saturated soils usually exhibit more complex behaviour than some other
porous materials, as they often experience non-linear plastic (irrecoverable) deformation. This
makes their computational modelling more complicated than for some other coupled problems.
The selection of a consistent constitutive model within the theory of mixtures, that can
incorporate suction forces into the description of stress, provides a further complication. The
necessity of such incorporation has frequently been reported in experimental studies of
unsaturated soils. This requires the adoption of a unique strategy for integrating the constitutive
model for these soils.

8
Another challenge in modelling unsaturated soil as a three-phase material arises in solving the
global equations of motion, where the presence of different phases (solid, gas and water),
together with the existence of inertia forces in each phase, makes the solution of the coupled
dynamic system computationally demanding. One of the aims of this research is to address the
aforementioned challenges. In this chapter, a review of some fundamental issues in partially
saturated soils is presented and then the application of the generalised-α algorithm for time
integration of the global equations of motion for unsaturated soils is demonstrated. Solutions
to these equations obtained by the finite element method are validated by recently presented
analytical solutions.

Generally, constructing a framework for numerical modelling of partially saturated soils

requires a number of main components, as follows:

 The description of the suction-saturation relationship, namely as the soil-water

characteristic curve.
 Forming the governing equations of motion under a framework which can incorporate
the individual responses of the different phases and their interactions.
 Generalisation of the concept of effective stress for partially saturated soils.
 Employing an elasto-plastic constitutive model to describe the elastic response as well
as the irreversible behaviour of unsaturated soil under complex loading paths.

This chapter mainly discusses the first two issues.

2.2 Soil-water characteristic curve

The soil-water characteristic curve (SWCC) is defined as the relationship between the
volumetric water content, or the effective degree of saturation, 𝑆𝑒 , and the suction in the soil,
𝑝𝑐 , where 𝑝𝑐 is the suction head and Se is the effective saturation defined according to

𝑆𝑤 − 𝑆𝑟𝑤
𝑆𝑒 = (2.1)
𝑆𝑟𝑎 − 𝑆𝑟𝑤

where 𝑆𝑤 represents the degree of saturation, 𝑆𝑟𝑤 is the residual degree of saturation at
extremely dry conditions, and 𝑆𝑟𝑎 is the residual degree of saturation at the fully saturated
condition.

9
The soil-water characteristic curves are generally identified by a point of “air-entry” or the “air
entry value”, which is the point where the largest voids in the soil structure begin to lose
available water content as suction increases. At the other extreme, “residual conditions”
commence at the point where it becomes very difficult for soils to experience further
desaturation.

A number of the most frequently used mathematical expressions for the soil-water
characteristic curve are presented in Table 2-1.

Table 2-1. Commonly used fitting equations for Soil Water Characteristic Curve.

Reference Equation Fitting Parameters

1
Gardner (1958) 𝑆𝑒 = ′ 𝛼 ′ , 𝑛′
1 + 𝛼 ′ 𝑝𝑐 𝑛
𝑏
Brooks and Corey (1964) 𝑆𝑒 = (𝑃𝑑 /𝑝𝑐 )𝜈 𝜈𝑏
′ ′
Van Genuchten (1980) 𝑆𝑒 = 1/[1 + (𝛼 ′ . 𝑝𝑐 )𝑛 ]1−1/𝑛 𝛼 ′ , 𝑛′

1
𝑆𝑒 = 𝑛′
Fredlund and Xing (1994) 𝑝 ′ 𝛼 ′ , 𝑛′ , 𝑚 ′
{ln [exp(1) + (𝛼𝑐′ ) ]}𝑚

In the above table, 𝑃𝑑 represents the ‘‘air-entry value’’ in the Brooks-Corey function (Brooks
and Corey 1964).

The SWCC equation can also be used to derive the relation between the permeability and the
degree of saturation of soil. The coefficient of permeability of an unsaturated soil is a function
of properties such as the degree of saturation, soil structure, and the viscosity of the fluid. An
example of a permeability function based on the work of Brooks and Corey (1964) is presented
below:

(2+3𝜈 𝑏 )/𝜈 𝑏
𝑘𝑟𝑤 = 𝑆𝑒 (2.2)

(2+𝜈 𝑏 )/𝜈 𝑏
𝑘𝑟𝑔 = (1 − 𝑆𝑒 )2 [1 − 𝑆𝑒 ] (2.3)

10
where 𝑘𝑟𝑤 and 𝑘𝑟𝑔 are the relative permeability of the water phase and the air phase,
respectively, which are defined as the ratio between the actual permeability and the
permeability at fully saturated conditions. According to this definition the relative permeability
is bounded between 1 and 0. The actual permeability for each phase, 𝐤𝛽 (𝛽 = 𝑤, 𝑔), can be
obtained from the following equation:

𝑘𝑟𝛽
𝐤𝛽 = 𝐤 𝑖𝑛𝑡 . (2.4)
𝜂𝛽

in which 𝐤 𝑖𝑛𝑡 is the intrinsic or absolute permeability matrix of the soil, 𝜂𝛽 denotes the viscosity
of each fluid. Typical values of intrinsic permeability are listed in Table 2-2 .for different types
of material.

Table 2-2. Typical values of intrinsic permeability for various materials Pinder and Gray (2008).

Materials 𝑘𝑖𝑛𝑡 (cm2 )

Well-sorted gravel 10−3 − 10−4

Fractured rock 10−3 − 10−6

Well-sorted sand 10−5 − 10−7

Peat 10−7 − 10−8

Fine sand, loam 10−8 − 10−11

Layered clay 10−9 − 10−11

Limestone 10−12 − 10−13

Roof tile 10−12 − 10−13

Granite 10−14 − 10−15

Concrete 10−14 − 10−16

2.3 Governing equations

2.3.1 Volume fraction and effective density

11
It is assumed that the medium under consideration is composed of three phases, including solid
(s), liquid (w) and gas (g) phases. These phases are continuously distributed throughout space.
The degree of saturation, 𝑆𝛽 (𝛽 = 𝑤, 𝑔), and the density of each phase 𝜌𝛼 (𝛼 = 𝑠, 𝑤, 𝑔) are
obtained from:

𝛺𝛽
𝑆𝛽 = (2.5)
𝛺

𝑀𝛼
𝜌𝛼 = (2.6)
𝛺𝛼

where 𝛺 represents the volume and M is the mass. This gives the average density of the mixture,
𝜌, as:

𝜌 = (1 − 𝑛)𝜌𝑠 + 𝑛𝑆𝑤 𝜌𝑤 + 𝑛𝑆𝑔 𝜌𝑔 (2.7)

where n is the porosity.

2.3.2 Average effective stress concept in unsaturated soils

During the last seven decades, the concept of ‘effective stress’ has remained a topic of
considerable debate in the analysis of fluid-saturated porous materials. It has also been
important in the study of porous materials due to its unique capability to extend the available
constitutive models for saturated soils to partially fluid-filled geomaterials, e.g., Khalili et al.
(2004), Sheng (2011) and Vlahinić et al. (2011). Several researchers have independently
published similar definitions of the effective stress for partially saturated materials such as
Lambe (1960), Jennings (1961), Hassanizadeh and Gray (1980), Khalili and Khabbaz (1998)
and Alonso et al. (2010).

The single effective stress method emerged following the proposition by Bishop (1960) and is
similar to that adopted for fully saturated conditions. Bishop proposed the following equation
for the effective stress in partially saturated soils:

𝜎𝑖𝑗′ = (𝜎𝑖𝑗 − 𝑝𝑔 𝛿𝑖𝑗 ) + 𝜒(𝑝𝑔 − 𝑝𝑤 )𝛿𝑖𝑗 (2.8)

12
where 𝜎𝑖𝑗′ is the effective stress, 𝜎𝑖𝑗 represents the total stress, 𝑝𝑔 denotes the pore air pressure,
𝑝𝑤 is the pore water pressure, 𝜒 is called the effective stress parameter or Bishop’s parameter,
ranging from 0 to 1 for dry and saturated conditions, respectively, and 𝛿𝑖𝑗 is the Kronecker
delta. The term (𝜎𝑖𝑗 − 𝑝𝑔 δ𝑖𝑗 ) in Equation (2.8) is commonly called the net stress, and 𝑝𝑔 − 𝑝𝑤
represents the matric suction, also known as the capillary pressure.

A popular form of Equation (2.8) that was introduced by Lewis and Schrefler (1982) is
achieved by assuming that the Bishop parameter, 𝜒 is identical to the degree of saturation, 𝑆𝑤 ,
so that:

𝜎 ′ 𝑖𝑗 = (𝜎𝑖𝑗 − 𝑝𝑔 𝛿𝑖𝑗 ) + 𝑆𝑤 (𝑝𝑔 − 𝑝𝑤 )𝛿𝑖𝑗 (2.9)

A review of the concept of effective stress and proposed definitions for unsaturated soils are
also presented in Khalili et al. (2004). Also, the application of an alternative definition of
effective stress and its role in the global system of equations for unsaturated soils is shown in
the work of Khalili et al. (2008). The concept will be discussed further in Chapter 3.

Houlsby (1997) described 𝜎 ′ 𝑖𝑗 , defined in Equation (2.9), as the “average soil skeleton stress”
tensor. This representation of effective stress can also be found in Bolzon et al. (1996), Lewis
and Schrefler (1998), Hutter et al. (1999), Jommi (2000), Wheeler et al. (2003), Ehlers et al.
(2004), Tamagnini (2004), Oka et al. (2006), Nuth and Laloui (2008) and Khoei and
Mohammadnejad (2011), with different names adopted, such as “skeleton stress”. This
definition is consistent with multiphase mixture theory.

In this thesis Equation (2.9) is selected to define the average effective 𝜎𝑖𝑗′ in the mechanics of
partially saturated porous media. By also incorporating the Biot parameter, 𝜘, Equation (2.9)
becomes:

𝜎𝑖𝑗′ = 𝜎𝑖𝑗 − 𝜘. 𝑝𝑎𝑣𝑒 . 𝛿𝑖𝑗 (2.10)

where 𝑝𝑎𝑣𝑒 denotes the mean pore pressure exerted by the fluid phase(s) on the solid grains,
and is obtained by the averaging technique proposed by, according to Gray and Hassanizadeh
(1991) and Houlsby (1997):

𝑝𝑎𝑣𝑒 = 𝑆𝑤 . 𝑝𝑤 + 𝑆𝑔 . 𝑝𝑔 (2.11)

13
Substituting the capillary pressure, 𝑝𝑐 , into Equation (2.11) gives:

𝑝𝑎𝑣𝑒 = 𝑝𝑤 + (1 − 𝑆𝑤 ). 𝑝𝑐 (2.12)

The Biot coefficient, 𝜘, is described by the relationship:

𝐾
𝜘 = 1 − 𝐾𝑡 ≤ 1 (2.13)
𝑠

where 𝐾𝑡 is the bulk modulus of the porous medium and 𝐾𝑠 is the bulk modulus of the solid
grains.

2.3.3 Conservation of mass and momentum balance

According to the principle of conservation of mass, inside an arbitrary volume 𝛺, mass cannot
be eliminated or increased unless there is an outward or inward flow of materials. A flow of
each phase 𝛼 (𝛼 = 𝑠, 𝑤, 𝑔) through a surface d𝛤 can be described as 𝜌𝛼 𝑛𝛼 . 𝑽𝜶 . 𝐧∗ . d𝛤, where
𝐧∗ is the unit vector normal to the surface, 𝜌𝛼 𝑛𝛼 is the partial mass density of each material,
𝛺𝛼
and 𝑽𝜶 is the velocity of the material. 𝑛𝛼 = 𝛺 represents the volume fraction of each phase in

the mixture with 𝛺 𝛼 and 𝛺 as the volume occupied by each phase and the total volume of the
mixture, respectively.

In its integral form, the principle of mass conservation is written as:

𝑑
∫ 𝜌 𝑛 𝑑𝛺 = − ∮ 𝜌𝛼 𝑛𝛼 . 𝐕 𝜶 . 𝐧∗ . 𝑑𝛤 (2.14)
𝑑𝑡 𝛺 𝛼 𝛼 𝛤

By applying the Gauss theorem, the conservation of mass can be written in the form of a
differential equation, as follows:

𝜕(𝜌𝛼 𝑛𝛼 ) 𝜕(𝜌𝛼 𝑛𝛼 𝑉𝑖 𝛼 )
+ =0 (2.15)
𝜕𝑡 𝜕𝑥𝑖

Considering 𝐕 𝒔 = 𝐮̇ for the solid phase and the porosity as 𝑛 = 𝑛𝑤 + 𝑛𝑔 , the following
expression for the conservation of mass for the solid phase can be obtained as:

14
 𝜕((1 − 𝑛)𝜌𝑠 ) 𝜕((1 − 𝑛)𝜌𝑠 𝑢̇ 𝑖 )
+ =0 (2.16)
𝜕𝑡 𝜕𝑥𝑖
𝜕(ℜ𝒊 ) 𝜕𝜌𝑠
Assuming = ℜ𝑖,𝑖 and = 0, the following equation is obtained from expanding the
𝜕𝑥𝑖 𝜕𝑥𝑖

previous equation:

(1 − 𝑛)𝜕𝜌𝑠 𝜌𝑠 𝜕𝑛 𝜕((1 − 𝑛)𝜌𝑠 )

− + (1 − 𝑛)𝜌𝑠 𝑢̇ 𝑖,𝑖 + 𝑢̇ 𝑖 =0 (2.17)
𝜕𝑡 𝜕𝑡 𝜕𝑥𝑖

The actual velocity of the other phases, 𝐔̇𝛃 , can be written as:

𝐰̇ 𝛃
𝐔̇𝛃 = 𝐮̇ + (2.18)
𝑛𝑆𝛽

where 𝐰̇ 𝛃 represents the Darcy velocity (where 𝛽 = 𝑤, 𝑔).

Also, the conservation of mass for the fluid phases is represented by:

𝛽
𝜕(𝑛𝑆𝛽 𝜌𝛽 ) 𝜕(𝑛𝑆𝛽 𝜌𝛽 𝑈̇𝑖 )
+ =0 (2.19)
𝜕𝑡 𝜕𝑥𝑖

Substituting Equation (2.18) into (2.19) gives:

𝜕𝑛𝑆𝑤 𝜌𝑤 𝜕𝜌𝑤
+ 𝑛𝑆𝑤 𝜌𝑤 𝑢̇ 𝑖,𝑖 + 𝑤̇𝑖𝑤 𝑤
+ 𝜌𝑤 𝑤̇𝑖,𝑖 =0 (2.20)
𝜕𝑡 𝜕𝑥𝑖

and

𝜕𝑛𝑆𝑔 𝜌𝑔 𝑔 𝜕𝜌𝑔 𝑔
+ 𝑛𝑆𝑔 𝜌𝑔 𝑢̇ 𝑖,𝑖 + 𝑤̇𝑖 + 𝜌𝑔 𝑤̇𝑖,𝑖 = 0 (2.21)
𝜕𝑡 𝜕𝑥𝑖
𝐷(∗) 𝜕(∗) 𝜕(∗)
Dividing Equation (2.17) by 𝜌𝑠 and considering = + 𝑢̇ 𝑖 results in the following
𝐷𝑡 𝜕𝑡 𝜕𝑥𝑖

equation:

(1 − 𝑛)𝐷𝜌𝑠 𝐷𝑛
− + (1 − 𝑛) 𝑢̇ 𝑖,𝑖 = 0 (2.22)
𝜌𝑠 𝐷𝑡 𝐷𝑡

Similarly, dividing Equation (2.19) by 𝑆𝛽 𝜌𝛽 gives the following equations for the water and
gas phases, respectively:

15
 𝑛𝐷(𝑆𝑤 𝜌𝑤 ) 𝐷𝑛 1 𝑤 𝑤̇𝑖𝑤 𝜕𝜌𝑤
+ + 𝑛𝑢̇ 𝑖,𝑖 + 𝑤̇ + =0 (2.23)
𝑆𝑤 𝜌𝑤 𝐷𝑡 𝐷𝑡 𝑆𝑤 𝑖,𝑖 𝑆𝑤 𝜌𝑤 𝜕𝑥𝑖

and

𝑔
𝑛𝐷(𝑆𝑔 𝜌𝑔 ) 𝐷𝑛 1 𝑔 𝑤̇𝑖 𝜕𝜌𝑔
+ + 𝑛𝑢̇ 𝑖,𝑖 + 𝑤̇𝑖,𝑖 + =0 (2.24)
𝑆𝑔 𝜌𝑔 𝐷𝑡 𝐷𝑡 𝑆𝑔 𝑆𝑔 𝜌𝑔 𝜕𝑥𝑖

Adding Equation (2.22) to (2.23) and (2.24) leads to:

(1 − 𝑛)𝐷𝜌𝑠 𝑛𝐷(𝜌𝑤 ) 𝑛𝐷(𝑆𝑤 ) 1 𝑤 𝑤̇𝑖𝑤 𝜕𝜌𝑤

+ + + 𝑢̇ 𝑖,𝑖 + 𝑤̇ + =0 (2.25)
𝜌𝑠 𝐷𝑡 𝜌𝑤 𝐷𝑡 𝑆𝑤 𝐷𝑡 𝑆𝑤 𝑖,𝑖 𝑆𝑤 𝜌𝑤 𝜕𝑥𝑖

and

𝑔
(1 − 𝑛)𝐷𝜌𝑠 𝑛𝐷(𝜌𝑔 ) 𝑛𝐷(𝑆𝑔 ) 1 𝑔 𝑤̇ 𝜕𝜌𝑔
+ + + 𝑢̇ 𝑖,𝑖 + 𝑤̇𝑖,𝑖 + 𝑖 =0 (2.26)
𝜌𝑠 𝐷𝑡 𝜌𝑔 𝐷𝑡 𝑆𝑔 𝐷𝑡 𝑆𝑔 𝑆𝑔 𝜌𝑔 𝜕𝑥𝑖

The rate of change in the solid phase density can also be defined by:

(𝜘 − 𝑛) 𝐷𝑝
1 𝐷𝜌𝑠 [ (1
𝐾𝑠 𝐷𝑡 − − 𝜘)𝑢̇ 𝑖,𝑖 ]
=
𝜌𝑠 𝐷𝑡 (1 − 𝑛) (2.27)

and the rate of change in the fluid density is:

1 𝐷𝜌𝛽 1 𝐷𝑝𝛽
= , (𝛽 = 𝑤, 𝑔) (2.28)
𝜌𝛽 𝐷𝑡 𝐾𝛽 𝐷𝑡

Substituting Equations (2.27) and (2.28) into (2.25) and (2.26) leads to the following:

(𝜘 − 𝑛) 𝐷𝑝𝑎𝑣𝑒 𝑛 𝐷𝑝𝑤 𝑛𝐷𝑆w 1 𝑤 𝑤̇𝑖𝑤 𝜕𝜌𝑤

[ − (1 − 𝜘)𝑢̇ 𝑖,𝑖 ] + + + 𝑢̇ 𝑖,𝑖 + 𝑤̇𝑖,𝑖 + =0 (2.29)
𝐾𝑠 𝐷𝑡 𝐾𝑤 𝐷𝑡 𝑆𝑤 𝐷𝑡 𝑆𝑤 𝑆𝑤 𝜌𝑤 𝜕𝑥𝑖

and

𝑔
(𝜘 − 𝑛) 𝐷𝑝𝑎𝑣𝑒 𝑛 𝐷𝑝𝑔 𝑛𝐷𝑆𝑔 1 𝑔 𝑤̇ 𝜕𝜌𝑔
[ − (1 − 𝜘)𝑢̇ 𝑖,𝑖 ] + + + 𝑢̇ 𝑖,𝑖 + 𝑤̇𝑖,𝑖 + 𝑖 =0 (2.30)
𝐾𝑠 𝐷𝑡 𝐾𝑔 𝐷𝑡 𝑆𝑔 𝐷𝑡 𝑆𝑔 𝑆𝑔 𝜌𝑔 𝜕𝑥𝑖

where the time derivative of 𝑝𝑎𝑣𝑒 from Equation (2.11) is expressed by:

16
 𝐷𝑝𝑎𝑣𝑒 𝐷𝑝𝑤 𝐷𝑝𝑐 𝐷𝑆𝑤
= + (1 − 𝑆𝑤 ) − 𝑝 (2.31)
𝐷𝑡 𝐷𝑡 𝐷𝑡 𝐷𝑡 𝑐

By ignoring the hysteresis effect in the SWCC and neglecting the dependency of saturation
degree to the deformation of the soil, the rate of saturation change can also be described by:

𝐷𝑆𝑤 𝜕𝑆𝑤 𝐷𝑝𝑐

= (2.32)
𝐷𝑡 𝜕𝑝𝑐 𝐷𝑡
𝜕𝑆𝑤
where can be obtained from the water retention curve. Also, by assuming:
𝜕𝑝𝑐

(𝜘 − 𝑛)𝑆𝑤 𝑛𝑆𝑤
𝐶1 = + (2.33)
𝐾𝑠 𝐾𝑤

(𝜘 − 𝑛) (𝜘 − 𝑛) 𝜕𝑆𝑤 𝜕𝑆𝑤
𝐶2 = 𝑆𝑤 (1 − 𝑆𝑤 ) − 𝑆𝑤 𝑝𝑐 +𝑛 (2.34)
𝐾𝑠 𝐾𝑠 𝜕𝑝𝑐 𝜕𝑝𝑐

(𝜘 − 𝑛)𝑆𝑔 𝑛𝑆𝑔
𝐶3 = + (2.35)
𝐾𝑠 𝐾𝑔

(𝜘 − 𝑛) 𝑛 (𝜘 − 𝑛) 𝜕𝑆𝑤 𝜕𝑆𝑤
𝐶4 = 𝑆𝑔 (1 − 𝑆𝑤 ) + 𝑆𝑔 − 𝑆𝑔 𝑝𝑐 −𝑛 (2.36)
𝐾𝑠 𝐾𝑔 𝐾𝑠 𝜕𝑝𝑐 𝜕𝑝𝑐

and after some manipulation of Equations (2.29) and (2.30), the following governing equations
will be obtained:

𝑤 𝑤̇𝑖𝑤 𝜕𝜌𝑤
𝐶1 𝑝̇𝑤 + 𝐶2 𝑝̇𝑐 + 𝜘𝑆𝑤 𝑢̇ 𝑖,𝑖 + 𝑤̇𝑖,𝑖 + =0 (2.37)
𝜌𝑤 𝜕𝑥𝑖

where the super-imposed dot represents the time derivative of a variable, and

𝑔
𝑔 𝑤̇𝑖 𝜕𝜌𝑔
𝐶3 𝑝̇𝑤 + 𝐶4 𝑝̇𝑐 + 𝜘𝑆𝑔 𝑢̇ 𝑖,𝑖 + 𝑤̇𝑖,𝑖 + =0 (2.38)
𝜌𝑔 𝜕𝑥𝑖

17
The linear momentum balance equation for the total mixture can be written as:

𝑔
𝐷 𝑤̇𝑖 𝐷 𝑤̇𝑖𝑤
𝜎𝑖𝑗,𝑗 + 𝜌𝑏𝑖 − 𝜌𝑢̈ 𝑖 − 𝑛 [(1 − 𝑆𝑤 )𝜌𝑔 ( ) + 𝑆𝑤 𝜌𝑤 ( )] = 0 (2.39)
𝐷𝑡 𝑛(1 − 𝑆𝑤 ) 𝐷𝑡 𝑛𝑆𝑤

where 𝐮̈ is the acceleration of the solid phase and 𝐛 is the body force vector.

The linear momentum balance equation for each fluid phase results in the generalised Darcy
equation for multiphase flow, as:

𝛽
𝛽 𝐷 𝑤̇𝑗
𝑤̇𝑖 = (𝑘𝛽 )𝑖𝑗 [−𝑝𝛽,𝑗 + 𝜌𝛽 (𝑏𝑗 − 𝑢̈𝑗 − 𝐷𝑡 (𝑛𝑆 ))] , (𝛽 = 𝑤, 𝑔) (2.40)
𝛽

According to Lewis and Schrefler (1999), Zienkiewicz et al. (1999) and Zienkiewicz et al.
(1999), the relative acceleration of the fluids can be assumed to be much smaller than the
acceleration of the solid phase, so that:

𝐷
(𝐰̇𝛽 ) = 0 , (𝛽 = 𝑤, 𝑔) (2.41)
𝐷𝑡

Note that 𝐤 𝑤 and 𝐤𝑔 are the permeability matrices of the medium governing the flow of the
pore fluids, defined in Equation (2.4).

2.4 Numerical simulations

There are a few numerical approaches for solving the equations governing the behaviour of
partially saturated porous media. Some of these methods depend on common simplifying
assumptions in the analysis of unsaturated soils, which are explained in the following.

Morel-Seytoux and Billica (1985) and Wu and Forsyth (2001) studied the problem of multi-
phase flow inside a rigid porous body with no deformation in the solid phase. However, by a
series of tests on an unsaturated sand column, Narasimhan and Witherspoon (1978) showed
that agreement with experimental results cannot be achieved unless deformation of the soil
skeleton is taken into account.

Another common simplification is the assumption of a so-called passive gas phase, i.e., the gas
pressure is assumed to remain constant at atmospheric pressure during the analysis, whereby

18
the continuity equation of the gas flow is ignored. Perhaps this method was developed to be
consistent with the constitutive models formulated on the basis of net stress and suction (Sheng
et al. 2003, Sheng et al. 2008). Rigorous comparisons between two-phase and three-phase
modelling are rarely found in the literature, due to a lack of experimental data and the different
nature of the constitutive models adopted for partially saturated and fully saturated conditions.
By assuming a rigid solid skeleton, Morel-Seytoux and Billica (1985) illustrated that the results
obtained from a three-phase analysis provide a better fit to the experimental results. Ehlers et
al. (2004) also studied the problem of draining of an elastic soil column, which had been
examined experimentally by Liakopoulos (1964). After comparing the results obtained by two-
phase and three-phase models, they concluded that in contrast with the experimental evidence,
the two-phase formulation predicted the final stage of saturation sooner, while the three-phase
model fitted the experimental data very well in the final stage. More recently, Khoei and
Mohammadnejad (2011) also examined the performance of two-phase and three-phase models
in a back analysis of the liquefaction potential of San Fernando dam, and pointed out that the
latter model provided more realistic results for the studied problem.

In this thesis a fully dynamic, coupled solution method, based on previous works by Li et al.
(1990), Li and Zienkiewicz (1992), Schrefler and Xiaoyong (1993), Schrefler et al. (1995), Li
et al. (1999), Rahman and Lewis (1999), Schrefler and Scotta (2001), Ehlers et al. (2004), Nuth
and Laloui (2008) and Khoei and Mohammadnejad (2011), is employed to simulate the
behaviour of a partially saturated porous medium.

2.4.1 Finite element discretisation of the governing equations

Displacement, pore water pressure, and suction are represented by the nodal quantities 𝐮, 𝐏𝐰
and 𝐏𝐜 , respectively. In this study, plane problems are investigated and 6-noded elements have
been adopted to discretise the soil body. It is assumed that the element is isoparametric with
respect to displacement, while only the three corner nodes are used to represent pore pressure
and suction (see Figure 2-1).
Displacement- pore
pressure- suction

Displacement

Figure 2-1. The 6-noded element used in this study.

19
The weak form of the governing equation for the fluid phases in the 𝐮 − 𝑝𝑤 − 𝑝𝑐 framework,
can be written as:

𝜕𝑝𝑤 𝜕𝑝𝑤
∫𝛺 𝛿𝑝𝑤 (𝐶1 𝑝̇𝑤 + 𝐶2 𝑝̇𝑐 + 𝜘𝑆𝑤 𝛁𝐮̇ + 𝛁 (𝐤 𝐰 [− + 𝜌𝑤 (𝐛 − 𝐮̈ )]) + 𝐤 𝐰 [− + 𝜌𝑤 (𝐛 −
𝜕𝑥𝑖 𝜕𝑥𝑖

1 𝜕𝜌𝑤 𝜕𝑝𝑤 𝜕𝑝
𝐮̈ )] 𝜌 ) 𝑑𝛺 + ∫𝛺 (𝛿𝑝𝑐 + 𝛿𝑝𝑤 ) (𝐶3 𝑝̇𝑤 + 𝐶4 𝑝̇𝑐 + 𝜘𝑆𝑔 𝛁𝐮̇ + 𝛁 (𝐤 𝐠 [− − 𝜕𝑥𝑐 +
𝑤 𝜕𝑥𝑖 𝜕𝑥𝑖 𝑖

𝜕𝑝𝑤 𝜕𝑝 1 𝜕𝜌𝑔 𝜕𝑝𝑤

𝜌𝑔 (𝐛 − 𝐮̈ )]) + 𝐤 𝐠 [− − 𝜕𝑥𝑐 + 𝜌𝑔 (𝐛 − 𝐮̈ )] 𝜌 ) 𝑑𝛺 − ∫𝛤 𝛿𝑝𝑤 [𝐤 𝐰 [− + 𝜌𝑤 (𝐛 −
𝜕𝑥𝑖 𝑖 𝑔 𝜕𝑥𝑖 𝑞𝑤 𝜕𝑥𝑖

𝜕𝑝𝑤 𝜕𝑝
̅̇ 𝑤 ] 𝑑Г − ∫ (𝛿𝑝𝑐 + 𝛿𝑝𝑤 ) [𝐤 𝐠 [−
𝐮̈ )] − 𝑤 ̅̇ 𝑔 ] 𝑑Г = 0
− 𝜕𝑥𝑐 + 𝜌𝑔 (𝐛 − 𝐮̈ )] − 𝑤 (2.42)
𝛤 𝑞𝑔 𝜕𝑥𝑖 𝑖

and for the solid phase:

∫𝛺 𝛿𝑢(𝛁𝛔 + 𝜌𝐛 − 𝜌𝐮̈ )𝑑𝛺 + ∫𝛤 𝛿𝑢(𝐭̅ − 𝛔𝐧∗ )𝑑Г = 0 (2.43)

𝑡

It is also assumed that:

(𝜘 − 𝑛) 𝑛𝑆𝑤 𝑛𝑆𝑔
𝐶1∗ = 𝐶1 + 𝐶3 = + + (2.44)
𝐾𝑠 𝐾𝑤 𝐾𝑔

(𝜘 − 𝑛) 𝜕𝑆𝑤 𝑛 𝜕𝑆𝑤
𝐶2∗ = 𝐶2 = 𝑆𝑔 (1 − 𝑆𝑤 − 𝑝𝑐 ) + 𝑆𝑔 −𝑛 (2.45)
𝐾𝑠 𝜕𝑝𝑐 𝐾𝑔 𝜕𝑝𝑐

(𝜘 − 𝑛) 𝜕𝑆𝑤 𝑛
𝐶3∗ = 𝐶4 + 𝐶2 = (1 − 𝑆𝑤 − 𝑝𝑐 ) + 𝑆𝑔 (2.46)
𝐾𝑠 𝜕𝑝𝑐 𝐾𝑔

(𝜘 − 𝑛)𝑆𝑔 𝑛𝑆𝑔
𝐶4∗ = 𝐶3 = + (2.47)
𝐾𝑠 𝐾𝑔

where 𝐾𝑤 and 𝐾𝑔 are the bulk moduli of water and air, respectively.

The finite element discretisation is achieved by considering the following shape functions for
the variables 𝐮, 𝑝𝑤 and 𝑝𝑐 , as described below:

20
𝐔 = 𝐍𝐮 𝐮 (2.48)

𝐏𝐰 = 𝐍𝐩𝐰 𝑝𝑤 (2.49)

𝐏𝐜 = 𝐍𝐩𝐜 𝑝𝑐 (2.50)

where 𝐍𝐮 , 𝐍𝐩𝐰 and 𝐍𝐩𝐜 are shape functions for displacement, pore water pressure and suction,
respectively. Based on the above interpolation functions, the following system of fully coupled
algebraic equations is derived in matrix form from the equations above:

𝐌𝐮 𝐔̈ + 𝐂𝐔̇ + ∫𝛺 𝐁 𝑇 𝛔′ d𝛺 − 𝐐𝐰 𝐏𝐰 − 𝐐𝐜 𝐏𝐜 = 𝐅𝐮 (2.51)

𝐌𝐰 𝐔̈ + 𝐐𝑇𝐰 𝐔̇ + 𝐂𝐰𝐰 𝐏̇𝐰 + 𝐂𝐰𝐜 𝐏̇𝐜 + 𝐇𝐰𝐰 𝐏𝐰 + 𝐇𝐰𝐜 𝐏𝐜 = 𝐅𝐰 (2.52)

𝐌𝐜 𝐔̈ + 𝐐𝐓𝐜 𝐔̇ + 𝐂𝐜𝐰 𝐏̇𝐰 + 𝐂𝐜𝐜 𝐏̇𝐜 + 𝐇𝐰𝐜

𝑇
𝐏𝐰 + 𝐇𝐜𝐜 𝐏𝐜 = 𝐅𝐜 (2.53)

where 𝐁 is the strain-displacement matrix and 𝐂 is the damping matrix. The definitions of the
various matrices and vectors are given below:

𝐌𝐮 = ∫𝛺 𝐍𝐮𝑇 𝜌𝐍𝐮 𝑑𝛺 (2.54)

𝑇
𝐌𝐰 = ∫𝛺 (𝛁𝐍𝐩𝐰 ) ( 𝐤𝑔 𝜌𝑔 + 𝐤 𝑤 𝜌𝑤 )𝐍𝐮 𝑑𝛺 (2.55)

𝐐𝐜 = ∫𝛺 𝐁 𝑇 𝜘(1 − 𝑆𝑤 )𝐦𝐍𝐩𝐜 𝑑𝛺 (2.56)

𝐐𝐰 = ∫𝛺 𝐁 𝑇 𝜘𝐦𝐍𝐩𝐰 𝑑𝛺 (2.57)

𝐌𝐂 = ∫𝛺 (𝛁𝐍𝐩𝐜 )𝑇 𝐤𝑔 𝜌𝑔 𝐍𝐮 𝑑𝛺 (2.58)

𝑇
𝐇𝐰𝐰 = ∫𝛺 (𝛁𝐍𝐩𝐰 ) ( 𝐤𝑔 + 𝐤 𝑤 )(𝛁𝐍𝐩𝐰 )𝑑𝛺 (2.59)

𝑇
𝐇𝐜𝐜 = ∫𝛺 (𝛁𝐍𝐩𝐜 ) ( 𝐤𝑔 )(𝛁𝐍𝐩𝐜 )𝑑𝛺 (2.60)

21
 𝑇
𝐇𝐰𝐜 = ∫𝛺 (𝛁𝐍𝐩𝐰 ) ( 𝐤𝑔 )(𝛁𝐍𝐩𝐜 )𝑑𝛺 (2.61)

𝑇
𝐂𝐰𝐰 = ∫𝛺 𝐍𝐩𝐰 𝐶1∗ 𝐍𝐩𝐰 𝑑𝛺 (2.62)

𝑇 ∗
𝐂𝐜𝐜 = ∫𝛺 𝐍𝐩𝐜 𝐶2 𝐍𝐩𝐜 𝑑𝛺 (2.63)

𝑇
𝐂𝐰𝐜 = ∫𝛺 𝐍𝐩𝐰 𝐶3∗ 𝐍𝐩𝐜 𝑑𝛺 (2.64)

𝑇 ∗
𝐂𝐜𝐰 = ∫𝛺 𝐍𝐩𝐜 𝐶4 𝐍𝐩𝐰 𝑑𝛺 (2.65)

where 𝐦 is {1,1,1,0,0,0} for a three-dimensional case.

The imposed Dirichlet boundary conditions of the primary variables on the boundaries are:

𝐮=𝐮
̅ on 𝛤𝑢 (2.66)

𝑝𝑤 = 𝑝̅𝑤 on 𝛤𝑝𝑤 (2.67)

𝑝𝑤 = 𝑝̅𝑐 on 𝛤𝑝𝑐 (2.68)

whereas the Neumann boundary conditions as the prescribed traction and fluxes are:

𝐭̅ on 𝛤𝑡 (2.69)

̅̇ 𝑤 = 𝐤 𝑤 [−𝛁𝑝𝑤 + 𝜌𝑤 (𝐛 − 𝐮̈ )]. 𝐧∗
𝑤 on 𝛤𝑞𝑤 (2.70)

̅̇ 𝑔 = 𝐤𝑔 [−𝛁𝑝𝑔 + 𝜌𝑔 (𝐛 − 𝐮̈ )]. 𝐧∗
𝑤 on 𝛤𝑞𝑔 (2.71)

̅̇ 𝛽 (𝛽 = 𝑤, 𝑔)are the prescribed values of the outflow rate of non-solid phase on the
where 𝑤
permeable boundaries 𝛤𝑞𝛽 (𝛽 = 𝑤, 𝑔). Therefore, the load and flow vectors are defined by:

̅ 𝛤
𝐅𝐮 = ∫𝛺 𝐍𝐮𝑇 𝜌𝐛𝑑𝛺 + ∫𝛤 𝐍𝐮𝑇 𝐭𝑑 (2.72)
𝑡

𝑇 𝑇 ̅ 𝑇 ̅
𝐅𝐰 = ∫𝛺 (𝛁𝐍𝐩𝐰 ) ( 𝐤𝑔 𝜌𝑔 + 𝐤 𝑤 𝜌𝑤 )𝐛 𝑑𝛺 − ∫𝛤 𝐍𝐩𝐰 𝑤̇𝑤 𝑑𝛤 − ∫𝛤 𝐍𝐩𝐰 𝑤̇𝑔 𝑑𝛤 −
𝑞𝑤 𝑞𝑤 ∩ 𝛤𝑞𝑔

𝑇 ̅ 𝑇 𝐰̇ 𝑇 𝐰̇𝑔𝑇
∫𝛤 𝐍𝐩𝐰 𝑤̇𝑔 𝑑𝛤 − ∫𝛺 𝐍𝐩𝐰 ( 𝜌 𝑤 𝛁𝜌𝑤 + 𝛁𝜌𝑔 ) 𝑑𝛺 (2.73)
𝑃𝑐 ∩ 𝛤𝑞𝑤 𝑤 𝜌𝑔

22
 𝑇 𝑇
𝑇 ̅ 𝑇 𝐰̇𝑔
𝐅𝐜 = ∫𝛺 (𝛁𝐍𝐩𝐜 ) 𝐤𝑔 𝜌𝑔 𝐛 𝑑𝛺 − ∫𝛤 𝐍𝐩𝐜 𝑤̇𝑔 𝑑𝛤 − ∫𝛺 𝐍𝐩𝐜 𝛁𝜌𝑔 𝑑𝛺 (2.74)
𝑞𝑔 𝜌 𝑔

2.4.2 Time discretisation of the governing equations

Chung and Hulbert (1993) introduced the generalised-𝛼 method for the analysis of dynamic
problems, in which the features of the HHT method developed by Hilber et al. (1977) and the
WBZ method proposed by Wood et al. (1980) are combined. The fundamental idea of this
method is that the different terms in the equation of motion are calculated at different times
within each time step. Similar to the HHT and WBZ methods, the generalised-𝛼 method uses
Newmark’s equations (Newmark 1959) for the displacement and velocity variations, but
employs two additional algorithmic parameters 𝛼𝑚 and 𝛼𝑓 . A discussion of the properties of
the generalised-𝛼 method in a nonlinear regime is given by Erlicher et al. (2002). Kontoe et al.
(2008) also compared the performance of the WBZ, HHT, and the generalised-𝛼 method and
pointed out that the generalised-𝛼 algorithm shows superior performance in the coupled
analysis of geotechnical problems compared to the other algorithms. Nazem et al. (2009) also
employed the generalised-𝛼 method to solve several highly nonlinear dynamics problems in
geomechanics.

2.4.3 The generalised-α method for dynamic analysis of partially saturated media

In the generalised-α method it is assumed that all variables are known at time t and the aim is
to find their values at time 𝑡 + Δ𝑡, where Δ𝑡 denotes the size of the current time step. In this
method the inertia forces in the momentum equation are measured at time 𝑡 + (1 − 𝛼𝑚 )Δ𝑡,
whereas the other terms are measured at time 𝑡 + (1 − 𝛼𝑓 )Δ𝑡, where 𝛼𝑚 and 𝛼𝑓 are two
integration parameters. The integration parameters in this method satisfy the following
relations:

2𝜌∞ − 1
𝛼𝑚 = (2.75)
𝜌∞ + 1

𝜌∞
𝛼𝑓 = (2.76)
𝜌∞ + 1

1
𝛾 = − 𝛼𝑚 + 𝛼𝑓 (2.77)
2

1
𝛽 = 4 (1 − 𝛼𝑚 + 𝛼𝑓 )2 (2.78)

23
where 𝜌∞ is the desired value of the spectral radius at infinity, and 𝛾 and 𝛽 are Newmark’s
parameters. The unconditional stability of the method is guaranteed provided that

𝜃 ≥ 0.5 (2.79)

𝛾 ≥ 0.5 (2.80)

𝛼𝑚 ≤ 𝛼𝑓 ≤ 0.5 (2.81)

1 1
𝛽 ≥ 4 + 2 (𝛼𝑓 − 𝛼𝑚 ) (2.82)

In addition, the acceleration and velocities are updated by Newmark’s equation by Newmark
(1959) according to

1 1 1
𝐔̈𝑡+Δ𝑡 = 𝛽Δ𝑡 2 (𝐔𝑡+Δ𝑡 − 𝐔 𝑡 ) − 𝛽Δ𝑡 𝐔̇𝑡 − (2𝛽 − 1)𝐔̈ 𝑡 (2.83)

𝛾 𝛾 𝛾
𝐔̇𝑡+Δ𝑡 = 𝛽Δ𝑡 (𝐔𝑡+Δ𝑡 − 𝐔 𝑡 ) − (𝛽 − 1)𝐔̇𝑡 − Δ𝑡(2𝛽 − 1)𝐔̈𝑡 (2.84)

and using the 𝜃 method

𝑡+Δ𝑡 1 1 𝑡
𝐏̇𝐰 = 𝜃Δ𝑡 (𝐏𝐰 𝑡+Δ𝑡 − 𝐏𝐰 𝑡 ) − (𝜃 − 1)𝐏̇𝐰 (2.85)

𝑡+Δ𝑡 1 1 𝑡
𝐏̇𝐜 = 𝜃Δ𝑡 (𝐏𝐜 𝑡+Δ𝑡 − 𝐏𝐜 𝑡 ) − (𝜃 − 1)𝐏̇𝐜 (2.86)

By knowing that

𝐔̇𝑡+(1−𝛼𝑓 )Δ𝑡 = (1 − 𝛼𝑓 )𝐔̇ 𝑡+Δ𝑡 + 𝛼𝑓 𝐔̇𝑡 (2.87)

𝐔𝑡+(1−𝛼𝑓 )Δ𝑡 = (1 − 𝛼𝑓 )𝐔 𝑡+Δ𝑡 + 𝛼𝑓 𝐔𝑡 (2.88)

𝑡+(1−𝛼𝑓 )Δ𝑡 𝑡+Δ𝑡 𝑡

𝐏̇𝐰 = (1 − 𝛼𝑓 )𝐏̇𝐰 + 𝛼𝑓 𝐏̇𝐰 (2.89)

𝑡+(1−𝛼𝑓 )Δ𝑡 𝑡+Δ𝑡 𝑡

𝐏̇𝐜 = (1 − 𝛼𝑓 )𝐏̇𝐜 + 𝛼𝑓 𝐏̇𝐜 (2.90)

𝐏𝐰 𝑡+(1−𝛼𝑓 )Δ𝑡 = (1 − 𝛼𝑓 )𝐏𝐰 𝑡+Δ𝑡 + 𝛼𝑓 𝐏𝐰 𝑡+Δ𝑡 (2.91)

24
𝐏𝐜 𝑡+(1−𝛼𝑓 )Δ𝑡 = (1 − 𝛼𝑓 )𝐏𝐜 𝑡+Δ𝑡 + 𝛼𝑓 𝐏𝐜 𝑡+Δ𝑡 (2.92)

The residual vectors for an iterative procedure like the Newton-Raphson method can be
obtained. Coupling the generalized–α method with a Newton-Raphson iteration gives the
following equations for the vector of residuals at two consecutive iterations. These two residual
vectors are related to each other by:

𝐫𝐮𝑖,𝑡+Δ𝑡 𝐫𝐮𝑖−1,𝑡+Δ𝑡 𝑑𝐔𝑖,𝑡+Δ𝑡

𝑖,𝑡+Δ𝑡
{𝐫𝐰𝑖,𝑡+Δ𝑡 } = {𝐫𝐰𝑖−1,𝑡+Δ𝑡 } + 𝐉 {𝑑𝐏𝐰 }=𝟎 (2.93)
𝑖,𝑡+Δ𝑡 𝑖−1,𝑡+Δ𝑡 𝑖,𝑡+Δ𝑡
𝐫𝐜 𝐫𝐜 𝑑𝐏𝐜

where 𝐉 is the Jacobian matrix defined by:

∂𝐫𝐮 ∂𝐫𝐮 ∂𝐫𝐮

∂𝐔 ∂𝐏𝐰 ∂𝐏𝐜
∂𝐫𝐰 ∂𝐫𝐰 ∂𝐫𝐰
𝐉= (2.94)
∂𝐔 ∂𝐏𝐰 ∂𝐏𝐜
∂𝐫𝐜 ∂𝐫𝐜 ∂𝐫𝐜
[ ∂𝐔 ∂𝐏𝐰 ∂𝐏𝐜 ]

which will be of the following form after simplification Ghorbani et al. (2015):

𝐉=
1 𝛾
(1 − 𝛼𝑚 )𝐌𝐮 + (1 − 𝛼𝑓 )𝐂 + (1 − 𝛼𝑓 )𝐊 −𝐐𝐰 (1 − 𝛼𝑓 ) −𝐐𝐜 (1 − 𝛼𝑓 )
𝛽Δ𝑡 2 𝛽Δ𝑡
1 𝛾 1 1
(1 − 𝛼𝑚 )𝐌𝐰 + (1 − 𝛼𝑓 )𝐐𝑇𝐖 (1 − 𝛼𝑓 )𝐂𝐰𝐰 + (1 − 𝛼𝑓 )𝐇𝐰𝐰 (1 − 𝛼𝑓 )𝐂𝐰𝐜 + (1 − 𝛼𝑓 )𝐇𝐰𝐜
𝛽Δ𝑡 2 𝛽Δ𝑡 𝜃Δ𝑡 𝜃Δ𝑡
1 𝛾 1 1
[ (1 − 𝛼𝑚 )𝐌𝐜 + (1 − 𝛼𝑓 )𝐐𝑇𝐜 (1 − 𝛼𝑓 )𝐂𝐜𝐰 + (1 𝑇
− 𝛼𝑓 )𝐇𝐰𝐜 (1 − 𝛼𝑓 )𝐂𝐜𝐜 + (1 − 𝛼𝑓 )𝐇𝐜𝐜 ]
𝛽Δ𝑡 2 𝛽Δ𝑡 𝜃Δ𝑡 𝜃Δ𝑡

(2.95)

where

𝐊 = 𝑑(∫𝛺 𝐁 T 𝛔′ d𝛺 )/𝑑𝐔 (2.96)

2.5 Implementation and validation of the model in dynamic elasticity

The Jacobian matrix given in Equation (2.95) is not symmetric. However, Khoei and
Mohammadnejad (2011) suggested that in the case of small deformations and having a
constitutive model with associated flow rule or elasticity a symmetric matrix can be achieved
if matrices 𝐌𝐰 and 𝐌𝐜 from Equations (2.55) and (2.58) are assumed to be zero; and also if

25
𝐂𝐰𝐜 form Equation (2.64) is replaced by the transpose of 𝐂𝐜𝐰 from Equation (2.65). However,
these assumptions are not made in the current thesis. Therefore, a non-symmetric solver is
developed for solving the global system of motion.

The generalised-α method described in the previous section together with a non-symmetric
solver has been implemented in SNAC, the finite element code developed by the Geotechnical
group at the University of Newcastle. In order to validate the code, an unsaturated elastic sand
column subjected to a dynamic load is solved numerically and the results are compared with
an analytical solution for this problem presented by Li and Schanz (2011).

The one-dimensional soil column and the boundary conditions are shown in Figure 2-2. The
length of the column, H, is 10 𝑚. It is assumed that the side walls of the column are rigid,
frictionless and impermeable. The bottom boundary is also impermeable and fixed. Due to
these restrictions the vertical displacement u, the pore water pressure 𝑝𝑤 , and the suction 𝑝𝑐 ,
are the only degrees of freedom. At the top of the column, a vertical pressure 𝜎 = 1.0 Pa is
prescribed (almost) instantaneously (Figure 2-3). The material properties are also presented in
Table 2-3.

The equations proposed by Brooks and Corey (1964) were chosen to account for the soil-water
characteristic curve and the relative permeability in this problem.

The derivative of the fluid saturation with respect to the capillary pressure is expressed by:

𝜕𝑆𝑤
𝑝𝑐 = −𝜈 𝑏 (𝑆𝑤 − 𝑆𝑟𝑤 ) (2.97)
𝜕𝑝𝑐

The pore size distribution index 𝜈 𝑏 is set to 1.5, the residual water saturation 𝑆𝑟𝑤 is set to 0,
and the air entry saturation 𝑆𝑟𝑎 is set to 1.

26
Table 2-3. General material parameters according to Li and Schanz (2011).

Parameter type Symbol Value Unit

Porosity 𝑛 0.23 -

Density of the solid skeleton 𝜌𝑠 2650 kg m−3

Density of the water 𝜌𝑤 997 kg m−3

Density of the air 𝜌𝑎 1.01 kg m−3

Drained bulk modulus of the mixture 𝐾 1.02 × 109 Pa

Shear modulus of the mixture 𝐺 1.44× 109 Pa

Bulk modulus of the solid skeleton 𝐾𝑠 3.5 × 1010 Pa

Bulk modulus of the water 𝐾𝑤 2.25 × 109 Pa

Bulk modulus of the air 𝐾𝑎 1.10 × 105 Pa

Intrinsic permeability 𝑘 2.5 × 10−12 m2

Viscosity of the water 𝜂𝑤 1.0 × 10−3 Ns m−2

Viscosity of the air 𝜂𝑎 1.8 × 10−5 Ns m−2

The vertical displacement at the top of the column versus time predicted by the finite element
method as well as by the analytical approach given by Li and Schanz (2011) are plotted in
Figure 2-4 and Figure 2-5 for initial degrees of saturation of 0.5 and 0.9, respectively. There is
good agreement between the results obtained by these two approaches, and the amplitude and
wave length are more or less identical in the analytical and finite element simulations.

As Li and Schanz (2011) pointed out, the magnitude of the displacements are independent of
𝑆𝑤 because the effective bulk modulus does not change for the range of degrees of water
saturation considered here. The propagation of pore pressure and air pressure waves through
the column is also presented in Figure 2-6 and Figure 2-7, respectively.

27
 σ

smooth / impermeable permeable

smooth / impermeable
H

rough / impermeable

displacement nodes
displacement, pore pressure and suction nodes

Figure 2-2. Soil column and boundary conditions.

28
 1.2

0.8
Applied Pressure (Pa)

0.6

0.4

0.2

0
0 0.002 0.004 0.006 0.008 0.01
Time (s)

Figure 2-3. Prescribed pressure at the top of the column.

0.00E+00
FEM
-1.00E-09
Analytical

-2.00E-09

-3.00E-09
Displacement(m)

-4.00E-09

-5.00E-09

-6.00E-09

-7.00E-09
0 0.02 0.04 0.06 0.08 0.1
Time(s)

Figure 2-4. Comparison of finite element predictions of displacement at the top of the column with those
obtained by the analytical solution of Li and Schanz (2011) for an elastic soil with a degree of saturation of 0.5.

29
 0.00E+00
FEM
-1.00E-09
Analytical

-2.00E-09
Displacement(m)

-3.00E-09

-4.00E-09

-5.00E-09

-6.00E-09

-7.00E-09
0 0.02 0.04 0.06 0.08 0.1
Time(s)

Figure 2-5. Comparison of finite element predictions of displacement at the top of the column with those
obtained by the analytical solution of Li and Schanz (2011) for an elastic soil with a degree of saturation of 0.9.

6.00E-06

5.00E-06

4.00E-06
Water pressure (kPa)

3.00E-06

2.00E-06

1.00E-06

0.00E+00

-1.00E-06

-2.00E-06
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
t (s)

Figure 2-6. Excess water pressure at the bottom of the column in an elastic soil with a degree of saturation of
0.9.

30
 5.00E-06

4.00E-06

3.00E-06
Air pressure (kPa)

2.00E-06

1.00E-06

0.00E+00

-1.00E-06

-2.00E-06
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
t (s)

Figure 2-7. Excess air pressure at the bottom of the column in an elastic soil with a degree of saturation 0.9.

5.00E-08

4.00E-08
Change in degree of saturation

3.00E-08

2.00E-08

1.00E-08

0.00E+00

-1.00E-08
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
t (s)

Figure 2-8. Change in degree of saturation at the bottom of the column in initial saturation degrees of 0.9.

31
Very good agreement can also be seen with the solutions reported by Li and Schanz (2011) at
the same degree of saturation. It is worth investigating how the degree of saturation changes
during the analysis. Changes in saturation are dictated by the changes in suction in a partially
saturated medium. This can be seen in the graphs presented in Figure 2-8 where the change in
degree of saturation is plotted versus time. According to this graph, when a compression phase
occurs in the column, the pore pressure increases and consequently the suction decreases. On
the other hand, in case of expansion of the column, suction generates along the column and
consequently the degree of saturation decreases.

The presented results show how well the proposed method is performing in this type of
analysis. However, inclusion of an elasto-plastic model to the developed framework will be
discussed in the following chapter.

2.6 Summary

In the beginning of this chapter the origin of the theory of mixtures and later improvements on
this theory were explained. Then, a simplified version of this theory, which can be applied for
modeling partially saturated geomatrials, was presented. The global equations are based on the
assumption of immiscibility of fluids and the environment is also assumed to be iso-thermal.
The FEM discretization of the global equation as well as the generalized-α algorithm for time
discretisation were presented. At the end of this chapter, the numerical solution based on the
proposed algorithm was compared with an analytical solution to verify the correctness of the
results. The presented results indicate very good agreement between these two methods. In the
next chapter the implementation of an elasto-plastic model chosen to account for the non-linear
response of partially saturated soils will be discussed.

32
Chapter 3. Numerical integration of
constitutive model in large deformations
analysis of partially saturated soils

33
3.1 Introduction
This chapter contains discussion about the implementation of an elasto-plastic constitutive
model for solving large deformation problems in partially saturated soils. In the beginning the
concept of effective stress for partially saturated soils is reviewed. Then the formulation of the
chosen constitutive model is presented. In a subsequent part, a strategy for performing explicit
integration of the constitutive model for large deformation problems is suggested. Finally,
numerical examples are presented and these have been designed to serve the following purpose:

 validation of the implementation of the global equations in a finite strain analysis by

comparing the numerical solution with an analytical solution (Example 1);
 validation of the implementation of the chosen constitutive model with a closed-form
solution (Example 2);
 demonstration of the capability of the model in representing the mechanical behaviour
of unsaturated soils in the presence of suction forces, compared to the condition of zero
suction for fully saturated states (Example 3); and
 comparison of the non-linear behaviour of soil assuming both small and large
deformation and assessment of the role of the integration algorithm parameters
(Examples 4).

3.2 Effective stress in partially saturated soil

In three-phase geomaterials, where air and water are present together with the solid soil
skeleton, the basic constitutive assumptions are not so straightforward. Moreover, the
definition of the hardening variables controlling the mechanical behaviour of the solid phase is
not direct, and providing experimental evidence is also challenging. This is mainly due to the
complex interaction between the different phases and the effects of this interaction on the
overall behaviour of the mixture. It is also well-known that the presence of two different fluids
(a wetting and a non-wetting phase) denotes the existence of capillary forces which strongly
affect the mechanical response of the material as Tamagnini and Gennaro (2008) noted.

Terzaghi (1936) emphasised the importance of effective stress in saturated soils by stating that
“….All the measurable effects of a change in stress, such as compression, distortion, and a
change in shearing resistance, are exclusively due to changes in effective stress”. Despite the
general agreement on a unified definition of effective stress in saturated soils and its role in

34
describing the plastic behaviour of saturated soils, the topic is still an important topic of
research in unsaturated soil mechanics.

During the last seven decades, the concept of ‘effective stress’ has remained an extremely
debatable topic in the analysis of partially fluid-saturated porous materials. It is also considered
as perhaps the first contributor to the study of such materials, due to its unique capability to
extend the available constitutive models for saturated soils to partially fluid-filled geomaterials,
by researchers such as, Khalili et al. (2004), Sheng (2011) and Vlahinić et al. (2011).

For partially saturated materials, several researchers have independently published similar
definitions of the effective stress, which can be generally categorised into two methods. The
first method emerged following the proposition by Bishop (1960), which was explained
previously in Section 2.3.2. Perhaps Jennings and Burland (1962) were among the first to raise
doubts about the validity of this definition, pointing out that it is incapable of explaining the
process of compression (known as collapse) on wetting. They conducted a series of
consolidation tests on several unsaturated soils, and demonstrated that upon reducing suction
by wetting the soil, all samples collapsed (experienced compression), rather than expanding.
This observation was particularly true for soils below a critical degree of saturation which was
estimated to be approximately 20% for silts and sands, and as high as 85-90% for clays
(Fredlund and Rahardjo 1993). However, it was mentioned that this behaviour cannot be
predicted by Bishop’s proposition where reducing the suction by wetting and the resulting
decrease in effective stress, should, by definition, increase the void ratio.

Another shortcoming of Bishop’s definition of effective stress, often mentioned in the relevant
literature, is related to finding an appropriate value for Bishop’s parameter, 𝜒. Coleman (1962)
pointed out that 𝜒 is related to the material structure and concluded that it should not be
surprising if a correlation with the degree of saturation is not found. Aitchison (1967) also
mentioned the complexity involved in finding 𝜒 by pointing out that a particular value of 𝜒
may only be valid for a particular stress path.

Moreover, the difficulties of following conventional laboratory stress paths in an effective

manner using Bishop’s effective stress definition was another challenge (Gens et al. 2006). The
observations of such shortcomings persuaded many researchers to look for an alternative
approach. However, the idea of a possible extension of the effective stress concept for the

35
unsaturated state has caused many stress measures to be proposed. Table 3-1 shows a list of
some suggestions made in search for a single effective stress for unsaturated conditions.

As a consequence of the lack of success with finding a single acceptable definition of effective
stress during the 1960s, Burland (1964) and Burland (1965) suggested that two independent
variables should be employed to describe the behaviour of unsaturated soils. Fredlund and
Morgenstern (1977) subsequently presented a theoretical stress analysis of an unsaturated soil
based on continuum mechanics.

The unsaturated soil was considered to be incompressible and the effect of chemical reactions
in the soil was neglected. They stated that any pair of the three stress state variables,
(𝜎𝑖𝑗 − 𝑝𝑔 𝛿𝑖𝑗 ), (𝜎𝑖𝑗 − 𝑝𝑤 𝛿𝑖𝑗 ) and (𝑝𝑔 − 𝑝𝑤 ), can be used to describe the behaviour of an
unsaturated soil, where 𝛔 is the total stress, 𝑝𝑔 is the pore air pressure and 𝑝𝑤 is the pore water
pressure. To avoid unnecessary complexity, a preferred choice was the pair (𝜎𝑖𝑗 − 𝑝𝑔 𝛿𝑖𝑗 ) and
(𝑝𝑔 − 𝑝𝑤 ) because, firstly, the suction 𝑝𝑐 = (𝑝𝑔 − 𝑝𝑤 ) is a scalar value, and the second motive
follows if it is assumed that the entrapped air remains at atmospheric pressure, so that the net
stress can represent the total stress and suction is also simplified as a negative pore pressure.

One important advantage of using net stress and suction as the two independent variables is the
easier control of these variables in laboratory tests (Sheng et al. 2013). Therefore, using these
parameters experimental results can be directly used in constructing constitutive models.

However, adoption of the net stress and suction could fail to provide a direct transition between
saturated and unsaturated conditions, i.e., in this case Terzaghi's effective stress cannot be
obtained in a fully saturated state. Also, with the assumption of passive air pressure, the use of
net stress for saturated states implies that the constitutive laws should be expressed in terms of
the total stress instead of the effective stress. Thus, it becomes very difficult to extend many
available constitutive models, developed for saturated soils, to unsaturated soils.

The drawbacks of using two independent stress variables together with knowledge of the
hydro-mechanical coupling that necessarily occurs in unsaturated soil mechanics are probably
responsible for the reappearance, re-examination and reacceptance of the Bishop-type effective
stress.

36
Table 3-1 Some effective stress equations proposed for unsaturated soils

Equation Description of variables Reference

𝜎𝑖𝑗′ = (𝜎𝑖𝑗 − 𝑝𝑔 𝛿𝑖𝑗 ) + 𝜒(𝑝𝑔 − 𝑝𝑤 )𝛿𝑖𝑗 Bishop (1960)

𝛽 = determined statistically based

on experiments for an individual
case.
𝜎𝑖𝑗′ = 𝜎𝑖𝑗 − 𝛽𝑝′′ 𝛿𝑖𝑗 Jennings (1961)

𝑝′′ = pore-water pressure

deficiency

Hassanizadeh
𝜎𝑖𝑗′ = (𝜎𝑖𝑗 − 𝑝𝑔 ) + 𝑆𝑤 (𝑝𝑔 − 𝑝𝑤 )𝛿𝑖𝑗
and Gray (1980)
𝑝𝑐 −0.55
𝜎𝑖𝑗′ = (𝜎𝑖𝑗 − 𝑝𝑔 𝛿𝑖𝑗 ) + ( ) 𝑝𝑐 𝛿𝑖𝑗 𝑖𝑓 𝑝𝑐 > 𝑃𝑐𝑒 𝑝𝑐𝑒 = suction value determining
𝑝𝑐𝑒 Khalili and
the states of being saturated and
Khabbaz (1998)
𝜎𝑖𝑗′ = (𝜎𝑖𝑗 − 𝑝𝑤 𝛿𝑖𝑗 ) 𝑖𝑓 𝑝𝑐 ≤ 𝑝𝑐𝑒 unsaturated.

𝑆𝑤𝑚 = a microscopic degree of

𝑆𝑤 − 𝑆𝑤𝑚
𝜎𝑖𝑗′ = (𝜎𝑖𝑗 − 𝑝𝑔 𝛿𝑖𝑗 ) + 〈 〉 (𝑝𝑔 − 𝑝𝑤 )𝛿𝑖𝑗 saturation which concerns with
1 − 𝑆𝑤𝑚 Alonso et al.
the water within micropores.
(2010)

Some previous limitations of the single effective stress principle, which had often been cited
in the literature as key shortcomings of this method, were criticized in more recent research.
For example, Khalili et al. (2004) showed that it is possible to find a relationship between
Bishop’s parameter 𝜒 and the degree of saturation, 𝑆𝑤 for most soils, while the existence of
such a relationship had previously been questioned by Coleman (1962). In addition, later
interpretation of experimental data invalidated the critique made by Jennings and Burland
(1962) on the inability of a single effective stress to explain the collapse phenomenon upon
flooding. It was subsequently shown by Khalili et al. (2004) that the collapse behaviour is
actually linked to a plasticity mechanism.

Moreover, models based on the Bishop-type effective stress have other advantages, such as the
smooth transition between saturated and unsaturated states and a more straightforward finite
element implementation.

Throughout this thesis, the popular Equation (2.9) is therefore used as the definition of effective
stress for partially saturated soils.

37
 3.3 Generalization of available constitutive models

To construct a framework for the generalisation of the available constitutive models for
saturated soils to unsaturated conditions, much research has been carried out during the last
two decades to study the effects of suction on soil stiffness. The majority of the experimental
work has focused on investigating the soil response under different stress paths with net stress
and suction as the independent variables. Common stress paths investigated are:

 A suction reduction under a constant net mean stress following an initial isotropic
compression under a constant suction.
 One-dimensional compression under a constant suction followed by a suction decrease.
 Shear under a constant net mean stress followed by a suction reduction.

Sheng (2011) mentioned some fundamental issues that should be taken into account in the
generalisation of constitutive models formulated for saturated soils to a partially saturated state,
including:

 Volume change due to suction or saturation variations, and

 The dependence of shear strength on suction.

3.4 Constitutive models of partially saturated soils

Early constitutive models for unsaturated soils were based on the theory of elasticity, relating
the strain of the soil skeleton to the two independent stress state variables, net stress and matric
suction. Nonlinear elastic models were also employed for simulation of partially saturated soils,
e.g., Gens et al. (1998) and Thomas and He (1995). In order to overcome the shortcomings of
elastic constitutive models, numerous efforts have been made in recent years in developing
elasto-plastic models for partially saturated soils.

Generally, at least two main classes of elasoplastic models exist for unsaturated soils:

(i) Models based on two independent variables, net stress and suction, and
(ii) Models based on a single effective stress (Bishop-type).

3.4.1 Constitutive models based on net stress and suction

38
The first critical state model and perhaps the most popular model for partially saturated soils
was proposed by Alonso et al. (1990), which is called the Basic Barcelona Model (BBM).
There are some features defining the fundamental characteristics of this model including:

1) It is an extension of the well-known Modified Cam-Clay model (MCC) by Roscoe

and Burland (1968), developed for describing the behaviour of saturated soils, i.e.,
similar to MCC model a logarithmic relation between the mean stress and the void
ratio exists in the BBM. However, the normal consolidation line depends on suction.
2) A logarithmic relation between suction and void ratio is also assumed.
3) As mentioned previously, two independent stress state variables, net stress and
matric suction, are adopted for the formulation of the model.
4) The evolution of the pre-consolidation pressure with increasing matric suction is
explicitly defined and denoted as loading collapse (LC) yield curve.
5) The model uses two yield surfaces. The so-called Suction Increase (SI) yield surface
is related to suction increase.
6) The occurrence of yielding on the LC curve does not affect the current position of
SI.
7) The BBM has seven additional parameters in comparison to the Modified Cam-
Clay model. The extra parameters are associated with relating suction to two yield
surfaces involved in the problem.
8) Unlike the MCC-model, the BBM has a non-associated flow rule.
Since this leading work of Alonso et al. (1990), a number of elastoplastic constitutive models
have been developed in order to predict the behaviour of unsaturated soils by adopting the net
stress and suction as the main stress variables, such as those proposed by Josa Garcia-Tornel
et al. (1992), Cui et al. (1995), Wheeler and Sivakumar (1995), Sheng et al. (2008) and Pedroso
and Farias (2011). Some shortcomings of BBM have been improved by adding more flexibility
for the definition of the LC yield curve, the evolution of the critical state line with matric
suction, and using one yield surface.

3.4.2 Constitutive models based on a single effective stress

A few models for partially saturated soils formulated in terms of Bishop stress and matric
suction have been developed by researchers, including those suggested by Bolzon et al. (1996),
Eberhardsteiner et al. (2003), Gallipoli et al. (2003), Sheng et al. (2003), Borja (2004),

39
Tamagnini (2004), Tamagnini and Pastor (2004), Loret and Khalili (2002), Russell and Khalili
(2004), Khalili et al. (2008) and Santagiuliana and Schrefler (2006).

Among the finite element models of problems involving partially saturated soils, the model
proposed by Bolzon et al. (1996), the original Basic Barcelona Model (BBM) suggested by
Alonso et al. (1990) and its modified versions, such as that proposed by Sheng et al. (2003),
have gained popularity. Examples of using these models in practice can be found in the works
of Li et al. (1999), Khoei and Mohammadnejad (2011) and Sheng et al. (2003).

The model proposed by Bolzon et al. (1996) is a generalised version of the model of Pastor et
al. (1990) and was developed to predict the elasto-plastic behaviour of unsaturated soils using
a single measure of effective stress. The model is developed initially by the simple introduction
of Bishop's stress in the model of Pastor et al. (1990) to take into account the volume changes
induced by suction. This is achieved by substitution of Bishop’s definition of mean stress for
partially saturated soil in the yield surface and hardening parameters. Some more modifications
were also made on the original model of Pastor et al. (1990) to incorporate suction on the
evolution of yield surface.

However, the model is formulated in the generalized plasticity framework, and thus the absence
of a yield surface in this framework particularly limits the application of explicit stress
integration schemes and the accompanying yield correction methods. On the other hand, the
original BBM is not formulated on the basis of a single effective stress. This can cause
inconsistency with the proposed numerical solution method. One solution for this inconsistency
is a modification of the original BBM and one such modification was suggested by Sheng et
al. (2003), as described in the following section of this chapter. In this thesis the model of
Sheng et al. (2003) is employed in the numerical analysis of unsaturated soils due to its
consistency with the numerical method adopted to represent the elasto-plastic response of
unsaturated soils under monotonic loading.

3.5 Description of the selected model

The adopted model is an extension of the Modified Cam Clay (MCC) model first proposed by
Roscoe and Burland (1968) for saturated soils. An associated plastic flow rule is assumed
where the yield surface, 𝑓, and the plastic potential surface, 𝑔, are described by

40
 𝑞2 2
𝑝′ 𝑝′
𝑓=𝑔= + 𝑀 (1 + ) (3.1)
𝛼∗2 𝛼∗ 𝛼∗

Similar to MCC, 𝑀 is the slope of the critical state line in p'-q stress space. Also, 𝛼 ∗ determines
the yield surface size for unsaturated conditions, 𝑝′ represents the effective mean stress and 𝑞
is the deviatoric stress.

To incorporate suction forces into the model, the evolution of the yield surface is expressed by
an expression similar to that adopted in the BBM, but the net stress is replaced by a Bishop-
type constitutive stress. This is illustrated in the following equation where the hardening
parameter for unsaturated soil, 𝛼 ∗ , depends on suction:

𝛼∗ 𝛼 (𝜆0 −𝜅)/(𝜆−𝜅)
=( ) (3.2)
𝑝𝑟 𝑝𝑟

where 𝛼 is the yield surface size at the fully saturated condition, 𝑝𝑟 is a reference mean stress,
usually equal to 1 kPa Sheng et al. (2003), 𝜅 represents the slope of the compression-
recompression line, 𝜆 and 𝜆0 are the slopes of the normal compression lines for the saturated
and unsaturated conditions, respectively. The variation of 𝜆 with respect to suction 𝑝𝑐 is also
described by:

𝑝𝑐2
𝜆0 ((1 − 𝑟) exp(−𝛽𝑝𝑐 ) + 𝑟) 𝑝𝑐 > 𝛽
𝑝𝑐1 𝑝𝑐2
𝜆 = 𝜆0 ((0.875 + √0.015625 − (𝛽𝑝𝑐 − 𝑝𝑐1 )2 )(1 − 𝑟) + 𝑟) 𝛽
< 𝑝𝑐 ≤ 𝛽
(3.3)
𝑝𝑐1
{ 𝜆0 𝑝𝑐 ≤ 𝛽

where 𝑝𝑐1 and 𝑝𝑐2 are dimensionless quantities given approximately by (Sheng et al. (2003))

𝑝𝑐1 = −0.05111961064754801 (3.4)

𝑝𝑐2 = +0.03567832315393903 (3.5)

Moreover, the hardening law is defined by

∗
(1 + 𝑒)𝛼 ∗ 𝑝
𝛼̇ = 𝜀̇ (3.6)
𝜆0 − 𝜅 𝑣

where 𝑒 is the void ratio and 𝜀̇𝑣𝑝 is the rate of plastic volumetric strain.

41
The stress increment now depends on suction as well as the strain increment, and can be
expressed as:

Δ𝛔′ = 𝐃𝑒𝑝 Δ𝛆 + 𝚲Δ𝑝𝑐 (3.7)

where the elasto-plastic constitutive matrix, 𝐃𝑒𝑝 , is obtained from

𝜕𝑔 𝜕𝑓 𝑇
𝐃𝑒 ( ′ ) ( ′ ) 𝐃𝑒
𝜕𝛔 𝜕𝛔
𝐃𝑒𝑝 = 𝐃𝑒 − (3.8)
𝜕𝑓 𝑇 𝜕𝑔
𝐾𝑝 + ( ′ ) 𝐃𝑒 ( ′ )
𝜕𝛔 𝜕𝛔

in which 𝐃𝑒 is the elastic constitutive matrix, and 𝚲 is defined by

𝜕𝑔 𝜕𝑓
𝐃𝑒 ( ′ ) ( )
𝜕𝛔 𝜕𝑝𝑐
𝚲=− (3.9)
𝜕𝑓 𝑇 𝜕𝑔
𝐾𝑝 + ( ′ ) 𝐃𝑒 ( ′ )
𝜕𝛔 𝜕𝛔

The plastic strain increment Δ𝛆p is obtained from

𝜕𝑔
Δ𝛆p = Δ𝜆 ( ) (3.10)
𝜕𝛔′

where the plastic multiplier Δ𝜆 depends on the suction rate as well as the strain increment, and
is given by

𝜕𝑓 𝑇 𝑒 𝜕𝑓
( ′ ) 𝐃 Δ𝛆 + ( ) 𝛥𝑝𝑐
𝜕𝛔 𝜕𝑝𝑐
Δ𝜆 = (3.11)
𝜕𝑓 𝑇 𝜕𝑔
𝐾𝑝 + ( ′ ) 𝐃𝑒 ( ′ )
𝜕𝛔 𝜕𝛔
𝜕𝑓
in which (𝜕𝑝 ) is the derivative of yield function with respect to suction, and the plastic
𝑐

modulus 𝐾𝑝 is expressed by:

𝜕𝑓 𝜕𝛼 ∗ 𝜕𝑔
𝐾𝑝 = − ( ∗ ) ( 𝑝 ) ( ′ ) (3.12)
𝜕𝛼 𝜕𝜀𝑣 𝜕𝑝
More discussion on these matrices, stress invariants and the derivatives are provided in
Appendix C.

42
It is worth mentioning that due to the incorporation of suction forces generating effective stress
in the soil skeleton according to Equation (3.7), the coupling matrix 𝐐𝐜 in Equation (2.56) can
be replaced by a modified coupling matrix for suction forces, 𝐐∗𝐜 , as follows:

𝐐∗𝐜 = ∫𝛺 𝐁 𝐓 [𝛼(1 − 𝑆𝑤 )𝐦 − 𝚲]𝐍𝐩𝐜 𝑑𝛺 (3.13)

3.6 Integration of the model considering the occurrence of large deformation

Numerical simulation of soil behaviour has remained a challenge in geotechnical research and
practice despite the progress made in both the theory of plasticity and numerical methods. The
key reason is that unlike non-porous solids the behaviour of geomaterials is highly dependent
on their constituent particles, which can have a broad spectrum of size and type.

Additionally, the presence of liquid and gas and their presence in various proportions in the
voids of the porous solid skeleton of the soil can also change its behaviour. This makes finding
a unique model to predict the behaviour of all types of soil extremely challenging. Even when
advances in constitutive modelling are made the resulting model may still show some weakness
in simulating particular soil behaviour, for example under cyclic or monotonic loadings. As a
result, the literature concerning the development of constitutive modelling of geomaterials is
very rich and thus a comprehensive review of it is beyond the scope of this research.

But, it is worth mentioning that some of these models can be mathematically complicated and
they often have plenty of input parameters. Generally speaking, many of these models at least
have one thing in common, and that is the rate of stress is usually expressed as a function of
the rate of strain. Hence, obtaining the updated stress values in each step requires the integration
of the constitutive equations over the specified strain intervals.

It is impossible to integrate most of these constitutive models analytically due to their extreme
non-linearity. Therefore, a numerical integration method should be used to approximate the
exact solution.

However, these methods, if incautiously chosen, may also impose additional complexity on the
implementation procedure and more importantly can negatively affect the performance of the
constitutive equations in actual computations as mentioned by Hughes (1984). This makes the
whole procedure of implementation of a model into a computational framework even more
cumbersome. Conversely, a robust and efficient numerical procedure can facilitate the

43
implementation of constitutive models into a finite element program greatly facilitating the
analysis of large-scale, complex geotechnical problems.

A large number of numerical integration algorithms have been developed over the last four
decades. These methods can be listed under two main categories: explicit and implicit
algorithms.

Generally, the explicit methods are frequently endorsed in the literature because of their
simplicity of implementation and robustness. However, it has often been mentioned that
explicit algorithms suffer from a lack of convergence in high strain increments. The frequent
appearance of such a statement in the literature has perhaps three main reasons:

 Due to the weak performance of the Forward Euler Method.

 Not being aware of the fact that the performance of explicit methods in handling large
strain increments can be greatly enhanced if the increments are subdivided into smaller
substeps to avoid numerical instability.
 Not considering that an explicit procedure can be rendered even more promising by
introducing an automated sub-stepping and error control to the algorithm.

On the other hand, implicit integration schemes have often been cited to have significant
advantages over explicit approaches in dealing with large strain increments; thereby increasing
the speed of analysis. However, most implicit methods, such as the Closest Point Projection
Method (CPPM) (Simo and Taylor 1985), usually face difficulties in convergence in the
following situations (Anandarajah 2011):

 When the initial guess of the solution (required at the start of the algorithm) is too far
from the correct solution. This can occur particularly in elasto-plastic models with
associated softening and hardening behaviour.
 Particularly in models based on Cam-Clay, that may experience a change from a
positive value of the hardening parameter (determining the yield surface size) to a
physically inappropriate negative value during the iteration. Such an occurrence, if not
detected and appropriately managed, usually causes a termination of the analysis.
 In some more recent advanced constitutive models for geomaterials that describe the
elastic and plastic moduli as functions of the mean normal pressure. In such models
numerical instability becomes highly likely when the magnitude of the mean normal

44
 pressure is zero or less than some arbitrarily small value. The plastic modulus may also
vary from a positive to a negative value. Consequently, convergence difficulties are
encountered around the limiting pressure.

A numbers of research papers have proposed solutions to enhance the implicit algorithms, such
as those by de Souza Neto et al. (1994) and Pérez-Foguet et al. (2001). Nevertheless, including
such solutions into the implicit methods has the added cost of increasing the complexity of
implementation. Also, if Newton iteration is used in the implicit integration algorithm, finding
the second derivatives of the yield function and the plastic potential becomes necessary. This
can also increase the complexity of implementation.

To avoid these problems, a combination of a sub-stepping technique with an explicit integration

method has been applied to the chosen constitutive model in this research. Typical examples
of explicit integration algorithms with sub-stepping and their applications can be seen in
Hashash and Whittle (1992), Wissmann and Hauck (1983), Sloan (1987), Sloan et al. (2001),
Nazem et al. (2009), Sołowski and Gallipoli (2010), Ghorbani et al. (2014) and Zhao et al.
(2005).

The scheme proposed by Sloan (1987) was originally designed for traditional elasto-plastic
models where the elastic moduli are independent of mean normal effective stress and the
hardening law is described by a single parameter. Sloan et al. (2001) then extended the method
to encompass constitutive models with non-linear elastic behaviour. Later, Sheng et al. (2003)
generalized this scheme for unsaturated soils and studied its performance on a modified BBM
(Alonso et al. 1990) model. The generalisation of the integration scheme was applied by
dealing with suction as an additional strain parameter, because, similar to strain, suction also
has a direct effect on hardening as well as some of the other parameters in the model. The
integration algorithm uses a modified Euler scheme to measure the local error in each sub-
increment and to determine a criterion for automatic sub-incrementation in the subsequent
steps. At the end of each sub-increment, the stresses may diverge from the prescribed tolerance
𝐹𝑇𝑜𝑙 around the yield function 𝑓, as follows:

̃ ̃ ̃
𝑓(𝛔′𝑡+𝛥𝑡 , 𝛼 ∗𝑡 ) > 𝐹𝑇𝑜𝑙 (3.14)

45
where 𝑡̃ is the pseudo-time used in sub- incrementation, which should not be mistaken for the
actual time of the analysis 𝑡. Violation of the yield criterion has been commonly referred to as
“yield surface drift” in the literature.

In the following sections details of the integration algorithm at each stage, as well as a
description of each step, are explained.

3.7 Stress rate objectivity

In large-deformation analysis, constitutive equations must be frame-independent; that is

quantitative descriptions of the stress rate variables do not change when they are observed
under different conditions or using different frames of reference. Thus it is essential for the
constitutive equations to be written in terms of ‘objective rates’ to demonstrate correct
properties when the material is subjected to rigid body rotations. This is generally guaranteed
by using an objective stress rate, such as the Jaumann stress rate, 𝛔̇ ′𝐽 , in the stress–strain
relations, as follows:

𝛔̇ ′𝐽 = 𝛔̇ ′ + 𝛔′ 𝐰 − 𝐰𝛔′ (3.15)

where the spin tensor 𝐰 represents the angular velocity of the material which can be described
in indicial form as below

1
𝑤𝑖𝑗 = 2 (𝑢̇ 𝑖,𝑗 − 𝑢̇𝑗,𝑖 ) (3.16)

with 𝑢̇ 𝑖,𝑗 as the velocity gradient referred to spatial coordinates. More discussion on the use of
an objective stress rate in conjunction with elasto-plastic models can also be found in Bathe et
al. (1975), Pinsky et al. (1983), Rodríguez-Ferran and Huerta (1998), Gadala and Wang (2000)
and Nazem et al. (2009).

3.7.1 Algorithm 1: The use of Jaumann stress rate in the integration scheme

By using the Jaumann stress rate in the small-strain constitutive equations, the updated stress
in the next time step is found by integration of the equation over the prescribed strain, Δ𝛆,
applied in the time interval Δ𝑡:

Δ𝛆
𝛔′𝑡+Δ𝑡 = 𝛔′𝑡 + ∫0 𝐃𝐞𝐩 𝑑𝛆 + 𝛔
̌′𝑡 (3.17)

46
where

̌′𝑡 = 𝐐𝛔′𝑡 𝐐T
𝛔 (3.18)

with

𝐐 = (𝐈 − 𝜓𝐰)−1 (𝐈 + (1 − 𝜓)𝐰) (3.19)

and

𝐐T = (𝐈 + 𝜓𝐰)−1 (𝐈 − (1 − 𝜓)𝐰) (3.20)

In the above equations 𝐈 represents the identity matrix. 𝜓 is an integration parameter which can
vary between 0 and 1. It is shown by Hughes and Winget (1980) that the orthogonality of 𝐐
exist for 𝜓 = 0.5. Thus the following algorithm can be used to obtain the last term on the right
side of Equation (3.17).

(1) Enter with effective stresses 𝛔′𝑡 and the spin tensor 𝐰.
(2) Calculate 𝐐 and 𝐐T by equations (3.18) and (3.19).
̌′𝑡 = 𝐐𝛔′𝑡 𝐐T .
(3) Find 𝛔
̌′𝑡 .
(4) Exit with 𝛔

3.8 Elastic behaviour

In the model, the elastic part of the constitutive relation is generally dependent on the stresses;
thereby due to similarity of the elastic part of the model to the MCC model it is possible to find
an analytical solution by integrating the incremental relation between the mean stress and the
elastic volumetric strain. This permits direct calculation of the corresponding secant elastic
stiffness matrix as follows:

3.8.1 Algorithm2: Calculating the secant elastic stiffness matrix

1) Enter with strain increment, Δ𝛆, stress values and material parameters.
2) Calculate 𝑝𝑡′̃ +Δ𝑡̃ according to:
′
(1 + 𝑒)𝑝 𝑒
𝑝′̇ = 𝐾𝜀̇𝑣𝑒 = 𝜀̇𝑣 (3.21)
𝜅

47
 (1 + 𝑒) 𝑒
𝑝𝑡′̃+Δ𝑡̃ = 𝑝𝑡′̃ exp( Δ𝜀𝑣 ) (3.22)
𝜅
̅
3) Calculate the corresponding secant elastic bulk modulus 𝐾
𝑝𝑡′̃ (1 + 𝑒) 𝑒
̅=
𝐾 𝑒 [𝑒𝑥𝑝 ( Δ𝜀𝑣 ) − 1] (3.23)
Δ𝜀𝑣 𝜅

4) If the Poison’s ratio 𝑣 is constant (as chosen for this research), the shear modulus can
be calculated as:
(3 − 6𝑣)
𝐺̅ = ̅
𝐾 (3.24)
2(1 + 𝑣)

̅ and 𝐺̅ , the elastic constitutive matrix 𝐃𝑒 can be formed and the elastic stress
5) Having 𝐾
increment 𝛥𝛔′𝑒 is obtained by:

Δ𝛔′𝑒 = 𝐃𝑒 Δ𝛆 (3.25)

6) Exit with Δ𝛔′𝑒 .

3.9 Transition to irreversible response

In the primary step of the stress integration the whole strain increment is initially assumed to
generate an elastic stress increment of Δ𝛔′𝑒 . Thus, the elastic stress state at the new time 𝛔′𝑡+Δ𝑡
is obtained by

𝛔′𝑡+Δ𝑡 = 𝛔′𝑡 + Δ𝛔′𝑒 + 𝛔

̌′𝑡 (3.26)

̌′𝑡 , is also found by Algorithm 1.

where the rigid-body rotation effect, 𝛔

When plastic yielding occurs, the following condition happens:

𝑓(𝛔′𝑡+Δ𝑡 , 𝛼 ∗𝑡+Δ𝑡 ) > 𝐹𝑇𝑜𝑙

The equation above succeeds its former counterpart at time 𝑡 + Δ𝑡 in two possible scenarios
where:

(1) The stress state at the former time 𝑡 illustrated purely elastic behaviour:
𝑓(𝛔′𝑡 , 𝛼 ∗𝑡 ) < −𝐹𝑇𝑜𝑙
(2) Or, it showed a probability of experiencing plastic behaviour at the time 𝑡,:

48
 𝑓(𝛔′𝑡 , 𝛼 ∗𝑡 ) ≤ 𝐹𝑇𝑜𝑙

In the second case the probability of unloading behaviour should be investigated by the
algorithm where:

𝜕𝑓 𝑇
( ′𝑡 ) . Δ𝛔′𝑒 < 0 (3.27)
𝜕𝛔

or, alternatively, it can be said that the angle 𝜃 between the gradient of the yield surface and
the calculated elastic stress is larger than 90°. This condition can be numerically introduced to
the code with the help of a prescribed tolerance, as follows:

𝜕𝑓 𝑇
( ′𝑡 ) . Δ𝛔′𝑒
𝜕𝛔
cos 𝜃 = < −𝐿𝑇𝑜𝑙 (3.28)
𝜕𝑓
‖ ′𝑡 ‖ . ‖Δ𝛔 ‖
′𝑒
𝜕𝛔

However, if cos 𝜃 ≥ −𝐿𝑇𝑜𝑙 then purely plastic behaviour is seen.

The two cases dictate using different approaches for calculating the intersection point of the
yield surface. However, in both cases roots are calculated by solving the following equation
numerically:

̌′𝑡 ), 𝛼 ∗𝑡 (𝑝𝑐𝑡 + 𝛸Δ𝑝𝑐 )) = 0

𝑓(𝛔′𝑡 + 𝛸(Δ𝛔′𝑒 + 𝛔 (3.29)

In the above equation 𝛸 = 1 indicates purely elastic behaviour while 𝛸 = 0 assumes purely
irreversible deformation, and therefore 0 < 𝛸 < 1. Also,

𝛸Δ𝛔′𝑒 = 𝐃𝑒 . (𝛸Δ𝛆) (3.30)

where 𝛸Δ𝛆 provides the portion of the strain increment which generates purely elastic
behaviour. Since suction is handled the same way as strain the same definition also applies to
𝛸Δ𝑝𝑐 .

Equation (3.29) can be solved numerically by the use of algorithms such as the Bi-section
method or the Pegasus method (Dowell and Jarratt 1972). However, as previously noted, the

49
nature of finding the solution becomes different in the two cases listed above, as in the second
case the possibility of the existence of multiple roots should also be considered.

In the latter case, Sloan et al. (2001) suggest breaking up the strain increment into sub
increments. This makes it possible to investigate smaller intervals for the existence of a root.

(1) Enter with effective stresses 𝛔′𝑡 , the rigid-body rotation effect, 𝛔
̌′𝑡 and 𝐿𝑇𝑜𝑙.
(2) If 𝑓(𝛔′𝑡 , 𝛼 ∗𝑡 ) < −𝐹𝑇𝑜𝑙 use Algorithm 3.a (defined below) for finding 𝛸.
(3) If 𝑓(𝛔′𝑡 , 𝛼 ∗𝑡 ) ≤ 𝐹𝑇𝑜𝑙 check:
If cos 𝜃 < −𝐿𝑇𝑜𝑙 use Algorithm 3.b (defined below) for finding 𝛸,
Else If cos 𝜃 ≥ −𝐿𝑇𝑜𝑙, Set 𝛸 = 0.
(4) Exit with 𝛸.

3.9.1 Algorithm 3.a Pegasus method for case 1

1) Enter with the stresses and hardening parameter, the strain increment, the suction
increment, initial guess values for the two parameters of 𝛸 0 and 𝛸1 determining the
domain for finding the intersection with the yield surface, and the maximum number of
iterations MAXITS.
2) Calculate:
̅𝑒 (𝛔′𝑡 , Δ𝜀𝑣 ). Δ𝜀𝑣 + 𝛔
Δ𝛔′𝑒0 = 𝛸 0 . (𝐷 ̌′𝑡 )
and

̅𝑒 (𝛔′𝑡 , Δ𝜀𝑣 ). Δ𝜀𝑣 + 𝛔

Δ𝛔′e1 = 𝛸1 . (𝐷 ̌′𝑡 ) (3.31)

3) Calculate:
𝐹 0 = 𝑓(𝛔′𝑡 + Δ𝛔′𝑒0 , 𝛼 ∗𝑡 )
and
𝐹1 = 𝑓(𝛔′𝑡 + Δ𝛔′𝑒1 , 𝛼 ∗𝑡 )
4) Perform steps 5-9 MAXITS times.
5) Calculate:
(𝛸1 − 𝛸 0 )
𝛸 = 𝛸1 − 𝐹1 (3.32)
(𝐹1 − 𝐹 0 )

6) Calculate:

50
 ̅𝑒 (𝛔′𝑡 , Δ𝜀𝑣 ). Δ𝜀𝑣 + 𝛔
Δ𝛔′𝑒 = 𝛸. (𝐷 ̌′𝑡 )

𝐹 𝑓 = 𝑓(𝛔′𝑡 + Δ𝛔′𝑒 , 𝛼 ∗𝑡 )
7) If |𝐹 𝑓 | ≤ 𝐹𝑇𝑜𝑙 go to step 10.
8) If 𝐹 𝑓 . 𝐹 0 < 0 then:
Set 𝛸1 = 𝛸 and 𝐹1 = 𝐹 𝑓 ,
else:
𝐹1 𝐹0
Set 𝐹1 =
𝐹 0 +𝐹 𝑓
9) Set 𝛸 0 = 𝛸 and 𝐹 0 = 𝐹 𝑓 .
10) Convergence not achieved → Terminate the analysis.
11) Exit with 𝛸.

3.9.2 Algorithm 3.b Pegasus method for case 2

1) Enter with the stresses and hardening parameter, the strain increment, the suction
increment and the maximum number of iterations MAXITS.
2) Set 𝛸 0 = 0, 𝛸1 = 1, 𝐹 0 = 𝑓(𝛔′𝑡 , 𝛼 ∗𝑡 ) and 𝐹 𝑓 = 𝐹 0 .
3) Loop over 4 and 5 MAXITS times.
4) Calculate:

Δ𝛸 = (𝛸1 − 𝛸 0 )/𝑁𝑆𝑈𝐵 (3.33)

5) Loop over steps 6 and 7 NSUB times.

6) Calculate:

̅𝑒 (𝛔′𝑡 , Δ𝜀𝑣 ). Δ𝛆 + 𝛔
Δ𝛔′𝑒 = 𝛸. (𝐷 ̌′𝑡 ) (3.34)

𝛔′𝑡+Δ𝑡 = 𝛔′𝑡 + Δ𝛔′𝑒

where

𝛸 = 𝛸 0 + Δ𝛸 (3.35)

7) If 𝑓(𝛔′𝑡+Δ𝑡 , 𝛼 ∗𝑡 ) > 𝐹𝑇𝑜𝑙, then

Set 𝛸1 = 𝛸
If 𝑓(𝛔′𝑡+Δ𝑡 , 𝛼 ∗𝑡 ) < −𝐹𝑇𝑜𝑙

51
 Set 𝐹1 = 𝑓(𝛔′𝑡+Δ𝑡 , 𝛼 ∗𝑡 ) and go to step 9
Else
Set 𝛸 0 = 0 and 𝐹 0 = 𝐹 𝑓 and exit loop over steps 6 and 7.
Else
Set 𝛸 0 = 0 and 𝐹 0 = 𝑓(𝛔′𝑡+Δ𝑡 , 𝛼 ∗𝑡 ).
8) Intersection not found after MAXITS iterations →Terminante the analysis.
9) Exit with 𝛸 0 and 𝛸1 .
10) Call the Algorithm 3.a with 𝛸 0 and 𝛸1 to identify the yield surface intersection.

3.9.3 The drift correction

Sloan (1987) stated that if the integration is performed precisely, the drift will be negligible
and any correction algorithm to return back the stress within the prescribed tolerance is
optional.
However, particularly Potts and Gens (1985) and Crisfield and Tassoulas (1993) opposed that
idea and proposed the use of a scheme to remedy the yield surface drift at the end of a pseudo-
time.
The drift correction here is based on the scheme presented by Potts and Zdravković (1999). In
the beginning of the sub-increment the value of the yield function 𝑓 0 is located within the
prescribed tolerance as follows:
̃ ̃
𝑓 0 (𝛔′𝑡 , 𝛼 ∗𝑡 ) ≤ 𝐹𝑇𝑜𝑙
At the end of the integration, the stress state and hardening parameters are updated to a new
state 1 where
𝛔′1 = 𝛔′0 + 𝛿 𝛔′
and
̃ ̃ ̃ ̃
𝑓 1 (𝛔′𝑡+Δ𝑡 , 𝛼 ∗𝑡+Δ𝑡 ) > 𝐹𝑇𝑜𝑙
Here, a correction becomes essential to return back 𝑓 1 within the prescribed tolerance. Giving
the corrected value 𝑓 2 , it is aimed to obtain:
𝑓 2 ≤ 𝐹𝑇𝑜𝑙
The correction is made by preserving the value of total strain during each actual time increment
in the analysis. This can be achieved by keeping the applied strain in the actual time intact,
d𝛆 = 0
which gives

52
d𝛆𝑝 = −𝑑𝛆𝑒 (3.36)

Also, knowing that:

d𝛆𝑒 = (𝐃𝑒 )−1 (𝛔′2 − 𝛔′1 ) (3.37)

leads to:

𝑑𝛆𝑝 = −(𝐃𝑒 )−1 (𝛔′2 − 𝛔′1 ). (3.38)

The plastic strain is proportional to the gradient of the plastic potential 𝑔,

𝜕𝑔
𝑑𝛆𝑝 = Δ𝜆 (3.39)
𝜕𝛔′

thus the stress at point 2 is

𝜕𝑔
𝛔′2 = 𝛔′1 − Δ𝜆𝐃𝑒 (3.40)
𝜕𝛔′

A Taylor expansion of the yield function at point 2 gives:

𝑓 (𝛔′1 + 𝛿𝛔′ , 𝛼 ∗1 + 𝛿𝛼 ∗ )

′1 ∗1 )
𝜕𝑓 𝑇 𝑒 𝜕𝑔 𝜕𝑓 𝜕𝛼 ∗ 𝜕𝑔
= 𝑓(𝛔 , 𝛼 − Δ𝜆 ( ′ ) 𝐃 ( ′ ) + ( ∗ ) ( 𝑝 ) Δ𝜆 ( ′ ) (3.41)
𝜕𝛔 𝜕𝛔 𝜕𝛼 𝜕𝜀𝑣 𝜕𝑝

Having the equations gives:

𝑓(𝛔′ , 𝛼 ∗ )
Δ𝜆 =
𝜕𝑓 𝑇 𝑒 𝜕𝑔 𝜕𝑓 𝜕𝛼 ∗ 𝜕𝑔 (3.42)
( ′ ) 𝐃 ( ′ ) − ( ∗ ) ( 𝑝 )( ′ )
𝜕𝛔 𝜕𝛔 𝜕𝛼 𝜕𝜀𝑣 𝜕𝑝

To calculate Δ𝜆 from the above equation it would be best to use the values of stresses and the
hardening parameter at point 2. Because these values are not known, Potts and Zdravković
(1999) suggested using the values of the stresses and hardening parameter at point 0, as an
alternative. This is due to the fact that the calculated values at point 1 already exceeded 𝐹𝑇𝑜𝑙,
and thereby using them may cause substantial errors.

53
Following the calculation of Δ𝜆 an enhanced stress state, which is closer to the yield surface,
is obtained from:

𝛔′ = 𝛔′𝑡̃ + 𝛿𝛔′ , (3.43)

𝛼 ∗ = 𝛼 ∗𝑡̃ + 𝛿𝛼 ∗ . (3.44)

In most cases it is anticipated that by updating the stress and hardening parameters at the end
of the drift correction, the stress state is located within the prescribed tolerance.

However, it is still possible that after the drift correction they will diverge from the correct
solution. Thus, the use of a drift correction at the end of a pseudo-time step should be as
minimal as possible. This can be achieved by coupling proper values for the yield surface
tolerance and the integration tolerance.

However, in some cases using a correction which is normal to the yield surface instead of the
above method has been reported as successful, e.g., Crisfield and Tassoulas (1993), Nayak and
Zienkiewicz (1972), Sloan and Randolph (1982) and Sloan et al. (2001).

In this method, it is assumed that the hardening parameter remains intact during the correction;
thereby providing:

𝑓(𝛔′ , 𝛼 ∗ )
Δ𝜆 =
𝜕𝑓 𝑇 𝜕𝑓 (3.45)
( ) ( ′)
𝜕𝛔′ 𝜕𝛔

3.9.4 Algorithm 5: Yield surface correction scheme for general elasto-plastic models

Algorithm 5: Yield surface correction scheme for general elasto-plastic models

(1) Enter with stress, 𝛔′𝑡̃ and hardening 𝛼 ∗𝑡̃ at point zero explained above.
(2) Loop over step 3-6 MAXITC times.
(3) Calculate:

𝑓(𝛔′ , 𝛼∗ )
Δ𝜆 =
𝜕𝑓 𝑇 𝑒 𝜕𝑔 𝜕𝑓 𝜕𝛼∗ 𝜕𝑔
( ) 𝐃 ( ) − ( )( )( )
𝜕𝛔′ 𝜕𝛔′ 𝜕𝛼∗ 𝜕𝜀𝑝𝑣 𝜕𝑝′

(4) Correct stresses and hardening parameter:

54
 𝜕𝑔
𝛔′ = 𝛔′𝑡̃ − Δ𝜆𝐃𝒆 ( ),
𝜕𝛔′ (3.46)

𝛼 ∗ = 𝛼 ∗𝑡̃ + Δ𝜆𝐵. (3.47)

(5) If |𝑓(𝛔′ , 𝛼 ∗ )| > |𝑓(𝛔′𝑡̃ , 𝛼 ∗𝑡̃ )| then the previous corrections should be abandoned.
Instead the following should be performed:

𝑓(𝛔′ , 𝛼 ∗ )
Δ𝜆 = ,
𝜕𝑓 𝑇 𝜕𝑓
( ′) ( ′)
𝜕𝛔 𝜕𝛔

𝜕𝑓
𝛔′ = 𝛔′𝑡̃ − Δ𝜆 ( ), (3.48)
𝜕𝛔′

𝛼 ∗ = 𝛼 ∗𝑡̃ .

(6) If |𝑓(𝛔′ , 𝛼 ∗ )| ≤ 𝐹𝑇𝑜𝑙, then go to step 9.

(7) Set 𝛔′𝑡̃ = 𝛔′ and 𝛼 ∗𝑡̃ = 𝛼 ∗ .
(8) It did not converge after MAXITC steps → Terminate the analysis.
(9) Exit with 𝛔′ and 𝛼 ∗ .

3.10 The integration scheme

A generalization of the method of Sloan et al. (2001) in the framework of the updated
Lagrangian (UL) method is employed here. A detailed description of this method applied to
small deformation problems in a simplified formulation (biphasic assumption) of partilly
saturated soils is presented in Sheng et al. (2003).

The use of this explixit integration to obtain variables in any arbitrary pseuodo-time 𝑡̃ indicates
an implicit declaration that the algorithm has already been successful at the previous time 𝑡̃ −
Δ𝑡̃. However, this condition can also be checked in the second step of the following algorithm.

3.10.1 Algorithm 6: A Modified-Euler integration scheme with automatic substepping and

drift correction for large deformation analysis of partially saturated soils.

55
(1) Initial input: stresses 𝛔′𝑡 , Δ𝛆, suction 𝑝𝑐𝑡 , suction change Δ𝑝𝑐 , hardening parameter 𝛼 ∗ ,
the spin tensor 𝐰, Error tolerance values for yield surface 𝐹𝑇𝑜𝑙 and the error tolerance
of stresses 𝑆𝑇𝑜𝑙.
(2) If 𝑓(𝛔′𝑡 , 𝛼 ∗𝑡 ) > 𝐹𝑇𝑜𝑙 → Terminate the analysis.
Else:
̌′𝑡 based on Algorithm 2.
(3) Calculate 𝛔
(4) Calculate secant elastic stiffness matrix and elastic stress increment according to
Algorithm 1 and obtain 𝛔′𝑡+Δ𝑡 = 𝛔′𝑡 + 𝛥𝛔′𝑒 + 𝛔
̌′𝑡 .
(5) If 𝑓(𝛔′𝑡+Δ𝑡 , 𝛼 ∗𝑡 ) ≤ 𝐹𝑇𝑜𝑙 the behaviour is purely elastic; accept 𝛔′𝑡+Δ𝑡 . In this case no
change should happen in hardening parameter. Thus, 𝛼 ∗𝑡+Δ𝑡 = 𝛼 ∗𝑡 and exit.
(6) Else: call Algorithm 3 for computing 𝛸.
(7) Update the stresses at the onset of plastic yielding as 𝛔′𝑡 ← 𝛔′𝑡 +
̅𝑒 (𝛔′𝑡 , Δ𝜀𝑣 ). Δ𝛆 + 𝛔
𝛸. (𝐷 ̌′𝑡 ) , Then compute the portion of Δ𝛆 and change of suction that
correspond to plastic deformation according to Δ𝛆 ← (1 − 𝛸)Δ𝛆 and Δ𝑝𝑐 ← (1 −
𝛸)Δ𝑝𝑐 .
(8) Set 𝑡̃ = 0 and Δ𝑡̃ = 1
(9) While 𝑡̃ < 1:
(10) Calculate Δ𝛔′𝑖 and Δ𝛼 ∗𝑖 for 𝑖 = 1,2 using:

Δ𝛔′𝑖𝑒 = Δ𝑡̃. 𝐃𝑒 . Δ𝛆 (3.49)

Δ𝛔′𝑖 = −Δ𝜆𝑖 . 𝐃𝑒 . 𝐵𝑖 (3.50)

and

Δ𝛼 ∗𝑖 = Δ𝜆𝑖 . 𝐵𝑖 (3.51)

where

𝜕𝛼 ∗ 𝜕𝑔
𝐵 = ( 𝑝)( ′ ) (3.52)
𝜕𝜀𝑣 𝜕𝑝

also

56
 𝑇
𝜕𝑓 𝜕𝑓
( ′𝑖 ) 𝐃𝑒 Δ𝛆 + ( 𝑡 ) Δ𝑝𝑐
𝜕𝛔 𝜕𝑝𝑐
Δ𝜆𝑖 = 𝑚𝑎𝑥 𝑇 ,0 (3.53)
𝜕𝑓 𝜕𝛼 ∗𝑖 𝜕𝑔 𝜕𝑓 𝜕𝑔
− ( ∗𝑖 ) ( 𝑝 ) ( ′ ) + ( ′𝑖 ) . 𝐃𝑒 . ( ′𝑖 )
{ 𝜕𝛼 𝜕𝜀𝑣 𝜕𝑝 𝜕𝛔 𝜕𝛔 }

̃
𝛔′1 = 𝛔′𝑡

̃
𝛔′2 = 𝛔′𝑡 + Δ𝛔′1
̃
𝛼 ∗𝑖 = 𝛼 ∗𝑡
̃
𝛼 ∗2 = 𝛼 ∗𝑡 + Δ𝛼 ∗1

̃ ̃ ̃ 1
𝛔′𝑡+Δ𝑡 = 𝛔′𝑡 + 2 (Δ𝛔′1 + Δ𝛔′2 ) (3.54)

̃ ̃ 1
𝛼 ∗𝑡+Δ𝑡 = 𝛼 ∗𝑡 + (Δ𝛼 ∗1 + Δ𝛼 ∗2 ) (3.55)
2

(11) Determine the relative error for the current substep:

‖Δ𝛔′2 − Δ𝛔′1 ‖ |Δ𝛼 ∗2 − Δ𝛼 ∗1 |

𝐸 = 𝑚𝑎𝑥 { , , 𝐸𝑃𝑆} (3.56)
‖𝛔′𝑡̃+Δ𝑡̃ ‖ |𝛼 ∗𝑡̃+Δ𝑡̃ |

where EPS is a machine constant indicating the smallest relative error that can be
calculated.
(12) If 𝐸 > 𝑆𝑇𝑂𝐿 the substep has failed. First calculate:

𝑆𝑇𝑂𝐿
𝒜 = 𝑚𝑎𝑥 {0.9√ , 0.1} (3.57)
𝐸

then set:
∆𝑡̃ ← 𝑚𝑎𝑥{𝒜𝑡̃, 𝛥𝑡̃𝑚𝑖𝑛 },

and then go to step 7.

̃ ̃ ̃ ̃
(13) The substep is accepted, update stress and hardening variables to 𝛔′𝑡+Δ𝑡 and 𝛼 ∗𝑡+Δ𝑡 .

57
 ̃ ̃ ̃ ̃
(14) If 𝑓(𝛔′𝑡+Δ𝑡 , 𝛼 ∗𝑡+Δ𝑡 ) > 𝐹𝑇𝑜𝑙, then a drift correction becomes necessary. Call Algoithm
5 for drift correction.
(15) Extrapolate to get the size of next subincremenet as follows:

𝑆𝑇𝑜𝑙
𝒜 = 𝑚𝑖𝑛 {0.9√ , 1.1}
𝐸

If drift correction was called, limit the next step size further by:

𝑆𝑇𝑜𝑙
𝒜 = 𝑚𝑖𝑛 {0.9√ , 1},
𝐸

then set

∆𝑡̃ ← 𝒜∆𝑡̃,
𝑡̃ ← 𝑡̃ + ∆𝑡̃.
(16) Limit the next step size to the minimum step size and control that the pseudo-time does
not proceed beyond 𝑡̃= 1 by setting:
∆𝑡̃ ← 𝑚𝑎𝑥{∆𝑡̃, Δ𝑡̃𝑚𝑖𝑛 }

and then

∆𝑡̃ ← 𝑚𝑎𝑥{∆𝑡̃, 1 − 𝑡̃}.

(17) Exit with stress and hardening parameter when 𝑡̃ = 1.

A recommendation for the minimum substep size Δ𝑡̃𝑚𝑖𝑛 are of the order of 10−4 according to
Sloan et al. (2001). Throughout this thesis 𝐿𝑇𝑜𝑙 is also considered to be the order of 10−6 and
𝐹𝑇𝑜𝑙 is kept of the order of 10−9 in the presented examples. However, a suggestion for
choosing 𝐹𝑇𝑜𝑙 can be any arbitrary values of the orders between 10−5 and 10−12. MAXITS
and 𝑁𝑆𝑈𝐵 are also set to ten in the presented analyses whereas they can also be considered
any values between five and ten. MAXITC in Algorithm 5 or the maximum number of iteration
for yield surface corrections is also set to 100 in the presented examples.

58
 3.11 Numerical examples

In this section some numerical simulations using the algorithms proposed above are presented.
In the first example the validity of the algorithm for finite strain analysis is verified by
comparing the numerical predictions against an analytical solution proposed by Malvern
(1969).

Then a cylindrical cavity expansion is solved to verify the correctness of the implementation
of the constitutive model with an existing analytical solution.

In the third example, the bearing capacity of a rigid footing resting on the same soil in two
different scenarios of initial saturation is studied to demonstrate how the numerical model
captures the effect of suction forces as well as the changes in the hydro-mechanical properties
of the soil during an elasto-plastic analysis.

Finally, the last example compares two analyses of a particular soil in a partially saturated state
in both small and large deformation analyses. All finite strain analyses have been conducted
by activating algorithm 1, providing the stress integration process during a large deformation
analysis.

3.11.1 Finite strain analysis of a one-dimensional elastic soil layer

In the first example, the validity of the algorithm and the further modifications on stiffness
matrix and the effect of updating the geometry (presented in Appendix A) in finite strain
analysis is demonstrated using the analytical solution for one-dimensional compression of an
elastic soil proposed by Malvern (1969).

This solution has been frequently used in the literature for the purpose of validation of coupled
finite element codes for finite strain, e.g., Meroi et al. (1995) and more recently by Shahbodagh
et al. (2015).

A one-dimensional fully saturated elastic column with a height of 10 m and a width of 1 m is

considered in this analysis. The boundary conditions together with a schematic representation
of the finite element mesh are shown in Figure 3-1. The material properties are identical to
those adopted by Meroi et al. (1995) and are listed in Table 3-2.

59
 σ

permeable

smooth / impermeable

smooth / impermeable
H

rough / impermeable

displacement nodes
Coupled nodes

Figure 3-1. One-dimensional column and boundary conditions.

The finite strain solution is compared with the results obtained from the numerical analysis
assuming small and large deformations in

Figure 3-2. In this figure the stress applied to the top boundary 𝜎 is normalized by the elastic
modulus of the soil skeleton, 𝐸, and displacements are normalized by the height of the column,
𝐻. Good agreement is seen between the numerical and analytical finite strain results.

Moreover, it is seen that the difference between the results of small deformations and large
deformations becomes more significant as the deformation increases.

Table 3-2. Material parameters according to Meroi et al. (1995).

60
 Parameter type Symbol Value Unit

Porosity 𝑛 0.3 -

Density of the water 𝜌𝑤 1000 kg m−3

Elastic modulus of the mixture 𝐸 1.0 × 109 Pa

Poisson’s ratio 𝜈 0 -

Permeability 𝑘 0.01 ms −1

0.7

0.6

0.5

0.4
σ/E
0.3
FEM (finite strain)

0.2 FEM (small deformation)

0.1 Analytical solution

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Displacement/H

Figure 3-2. Finite element solutions for small and large deformations and analytical solution.

3.11.2 Cylindrical cavity expansion

It is noted that in the absence of suction forces, i.e., for fully saturated conditions, the
constitutive model adopted here simplifies to the Modified Cam-Clay model, and the global
equations also simplify to those of Biot’s consolidation theory for saturated soils under static
loading conditions. These special conditions can be illustrated by the following problem.

61
In order to verify the implementation of the numerical method presented here, its results are
compared with the exact analytical solution for the small strain expansion of a cylindrical
cavity under static loading conditions. The analytical solution for this problem using the
Modified Cam Clay model assuming undrained conditions was presented by Collins and Yu
(1996).

Again, there is very good agreement between the numerical predictions with those presented
by Collins and Yu (1996). In Figure 3-4, 𝜎𝑟′ and 𝜎𝜃′ represent the normalised radial effective
stress and the normalised tangential effective stress, respectively.

Table 3-3. Material parameters used in the problem of cylindrical cavity expansion.

Parameter type Symbol Value Unit

Slope of critical state line 𝑀 0.888 -

Drained bulk modulus of the

𝜆 0.161 -
mixture

Poisson’s ratio of the mixture 𝜇 0.3 -

Bulk modulus of the solid skeleton 𝜅 0.062 -

Specific volume at 𝑝′ = 1 𝛤 2.579 -

Over Consolidation Ratio (OCR) - 1.0 -

Initial yield surface size 𝛼∗ 170.8 kPa

Intrinsic permeability 𝑘 2.5 × 10−12 m2

Viscosity of the water 𝜂𝑤 1.0 × 10−3 Ns m−2

Note: when 𝑆𝑤 = 1.0, 𝛼 ∗ = 𝛼, 𝜆 = 𝜆0

In order to accurately model a cavity in an infinite medium, Sheng et al. (2000) showed that
the minimum ratio between the external radius and the internal radius should be at least 20.
This assumption is adopted here by setting the internal and the external radii of the finite
element mesh to 1 m and 20 m, respectively. The material parameters assumed in the analysis
are presented in Table 3-3. A uniformly distributed pressure of 100 kPa was applied on the
external cavity boundary, while an increasing internal pressure was applied to the cavity and
monitored until a radial displacement of 0.5 m was achieved.

62
The normalised total radial stress on the cavity boundary versus the radial displacement of that
boundary, normalised by the initial radius, as predicted by the FEM, as well as the analytical
solution, is plotted in

Figure 3-3. This figure shows that there is good agreement between the numerical results and
the analytical solution presented by Collins and Yu (1996). The distribution of effective stress
together with the excess pore pressure distribution is illustrated in Figure 3-4.

In Figure 3-4 the total stress, effective stresses and pore pressure have all been normalised by
dividing each quantity by the undrained shear strength of the soil 𝑠𝑢 , which according to the
adopted soil model is calculated as:

𝑠𝑢 = 0.5 𝑀 𝑒𝑥𝑝(𝛤 − 𝑒0 − 1) = 49.52 kPa (3.58)

7
normalised cavity pressure, φ

6 Exact

FEM-Small deformation
5

3
1 1.1 1.2 1.3 1.4 1.5
radial displacement normalised by the initial radius

Figure 3-3. Numerical results versus analytical solution for cavity expansion.

63
 5

4.5

4
Normalized stress and pore water pressure

3.5

2.5 𝑝𝑤 - Analytical
Normalized Excess Pore Pressure
solution
𝜎𝑟′ - Analytical
Normalized solution
Radial Stress
2 𝜎𝜃′ - Analytical solutionStress
Normalized Tangential

𝜎𝑧′ - Analytical
Normalized Sigzsolution
1.5
𝑝𝑤 - pressure-FEm
Pore FEM

1 𝜎𝑟′ - FEM
Radial-FEM
𝜎𝜃′ - FEM
Tangential-FEM
0.5
𝜎𝑧′ - FEM
Aigz-FEM

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
r/a

Figure 3-4. Normalised effective stress and pore pressure distributions.

3.11.3 Rigid footing on unsaturated soil

In this example, the bearing capacity of a foundation on a particular soil for two cases of fully
saturated and semi-saturated conditions is studied. The geometry of the plane strain model is
presented in Figure 3-5. The top boundary is assumed permeable; so no generation of pore
water or pore air pressure across the boundary is permitted and the side boundaries are
restrained so that no lateral displacement is possible. The overall width of the footing is 2 m.
Two values of 0.5 and 1.0 were considered for the degree of saturation in this example. The

64
soil behaviour under fully saturated conditions was predicted by the Modified Cam Clay model.
Statically applied loading only is considered here.

1.0 m
permeable
smooth / impermeable

smooth / impermeable

10 m
rough / impermeable

10 m
displacement nodes
displacement, pore pressure and suction nodes

Figure 3-5. Finite element model for the problem of a rigid footing.

The material properties assumed in these calculations are presented in Table 3-4. The equation
proposed by Brooks and Corey (1964) was used here to model the water retention inside the
soil skeleton, with the appropriate parameters given in Table 3-5. The analysis was performed
in two steps. In the first step, the initial stresses due to the self-weight of the soil were generated.
The footing load was then applied to the soil over a period of 10 s. This loading is rapid enough
to allow effectively undrained conditions to prevail in a saturated soil, yet slow enough that
any inertial effects may be ignored in the analysis.

The predicted soil response for fully saturated as well as semi-saturated conditions is presented
in Figure 3-6. In this figure, the applied load has been normalised by the initial

65
overconsolidated pressure, 𝛼, which is assumed to be constant in the soil layer, and the
displacement is normalised by the half width of the foundation.

Table 3-4. Constitutive model and material parameters for the problem of a rigid footing.

Parameter type Symbol Value Unit

Slope of critical state line 𝑀 0.888 -

The slope of loading line in fully

𝜆0 0.161 -
saturated state

Poisson’s ratio of the mixture 𝜇 0.3 -

The slope of unloading/reloading

𝜅 0.062 -
line in fully saturated state

Specific volume at 𝑝′ = 1 𝛤∗ 2.579 -

Over Consolidation Ratio (OCR) - 1.0 -

Initial yield surface size 𝛼 60 kPa

Unsaturated material parameter 𝛽 0.012 1/(kPa)

Unsaturated material parameter 𝑟 0.9 -

Density of the solid skeleton 𝜌𝑠 2700 kg m−3

Density of the water 𝜌𝑤 997 kg m−3

Density of the air 𝜌𝑎 1.1 kg m−3

Elastic modulus of the mixture 𝐸 1.2 × 107 Pa

Bulk modulus of the solid skeleton 𝐾𝑠 3.5 × 1010 Pa

Bulk modulus of the water 𝐾𝑤 2.25 × 109 Pa

Bulk modulus of the air 𝐾𝑎 1.01 × 105 Pa

Intrinsic permeability 𝑘 2.5 × 10−12 m2

Viscosity of the water 𝜇𝑤 1.0 × 10−3 Ns m−2

Viscosity of the air 𝜇𝑎 1.8 × 10−5 Ns m−2

66
Table 3-5. Water retention curve parameters for the problem of a rigid footing.

Parameter type Symbol Value Unit

Bubbling pressure 𝑃𝑑 50 kPa

Residual wetting fluid saturation 𝑆𝑟𝑤 0.0 -

Air entry saturation 𝑆𝑟𝑎 1.0 -

Pore size distribution index 𝑣𝑏 3.5 -

The plots in Figure 3-6 indicate that the response of the unsaturated soil is generally stiffer than
the saturated soil, and the ultimate bearing capacity of the unsaturated soil is larger than the
capacity of the fully saturated soil. This is mainly due to the existence of suction forces in the
partially saturated states compared to the state of zero suction in the fully saturated state which
increase the yield surface size resulting in higher resistance to the external loading.

1.2

0.8

0.6
F/α
Semi-saturated
0.4

Fully Saturated
0.2

0
0 0.05 0.1 0.15 0.2 0.25 0.3
Displacement / (B/2)

Figure 3-6. Normalised load-displacement curves for rigid footings on the same soil in fully saturated and semi-
saturated conditions.

3.11.4 Comparison of the response of partially saturated soil under a rigid footing in small
and large deformation analyses

67
The fourth example studies the sensitivity of the integration algorithm to the prescribed stress
tolerance, 𝑆𝑇𝑜𝑙.

The geometry of the problem is the same as example 3 while a nearly dry soil with a water
content of 10.54% is considered in this analysis.

The constitutive model and material parameters are shown in Table 3-6 and the water retention
parameters are also listed in Table 3-7. From Figure 3-7, it is clearly seen that in the range of
small deformations (footing displacements up to almost 0.12 m in this example) there is not
much difference between the predictions of small deformations and large deformations
analyses.

However, imposing additional displacement causes increase the disagreement in the predicted
responses.

In this example, the whole domain is discretised by of 625 nodes and 288 elements giving 1456
total degrees of freedom to the problem. As suggested by Sloan et al. (2001), the assessment
of stress errors can be obtained by :

𝑇 0.5
{∑𝑙𝑖=1 [(𝛔′𝑟𝑒𝑓 − 𝛔′ ) (𝛔′𝑟𝑒𝑓 − 𝛔′ )] }
𝑖
𝐸𝜎 = 0.5 (3.59)
𝑇
{∑𝑙𝑖=1 [(𝛔′𝑟𝑒𝑓 ) (𝛔′𝑟𝑒𝑓 )] }
𝑖

The element beneath the centre of footing is chosen for the assessment in this analysis. The
results of error assessment and computational time of the analysis for different values of STol
are shown in Table 3-8. The stress values corresponding to the prescribed stress tolerance of
0.0001 were considered as the reference stress,𝛔′𝑟𝑒𝑓 , in Equation (3.59). Similarly, CPU time
are normalized by CPU time corresponding to the analysis with STol=0.0001.

68
Table 3-6. Constitutive model and material parameters for the problem of a rigid footing.

Parameter type Symbol Value Unit

Slope of critical state line 𝑀 1.07 -

The slope of loading line in fully

𝜆0 0.28 -
saturated state

Poisson’s ratio of the mixture 𝜇 0.3 -

The slope of unloading/reloading

𝜅 0.05 -
line in fully saturated state

Specific volume at 𝑝′ = 1 𝛤∗ 2.6 -

Over Consolidation Ratio (OCR) - 1.0 -

Initial yield surface size 𝛼∗ 60 kPa

Unsaturated material parameter 𝛽 0.012 1/(kPa)

Unsaturated material parameter 𝑟 0.9 -

Density of the solid skeleton 𝜌𝑠 2700 kg m−3

Density of the water 𝜌𝑤 997 kg m−3

Density of the air 𝜌𝑎 1.1 kg m−3

Bulk modulus of the solid skeleton 𝐾𝑠 3.5 × 1010 Pa

Bulk modulus of the water 𝐾𝑤 2.25 × 109 Pa

Bulk modulus of the air 𝐾𝑎 1.01 × 105 Pa

Intrinsic permeability 𝑘 2.5 × 10−12 m2

Viscosity of the water 𝜇𝑤 1.0 × 10−3 Ns m−2

Viscosity of the air 𝜇𝑎 1.8 × 10−5 Ns m−2

69
Table 3-7. Water retention curve parameters for the problem of a rigid footing.

Parameter type Symbol Value Unit

Bubbling pressure 𝑃𝑑 30 kPa

Residual wetting fluid saturation 𝑆𝑟𝑤 0.0 -

Air entry saturation 𝑆𝑟𝑎 0.5 -

Pore size distribution index 𝑣𝑏 4.0 -

Table 3-9 also follows the same procedure of normalization for the large deformation analysis
with Algorithm 1 being active in the whole integration procedure. As the values of 𝐸 𝜎 in
Table 3-8 and

Table 3-9 suggest, considering stress objectivity in the analysis makes the algorithm more
sensitive to the prescribed stress tolerance.

3.1 Summary

This chapter began with a review of the concept of effective stress for partially saturated soils.
Following this discussion, the chosen elasto-plastic model for partially saturated soil and its
formulation were described. Then, a novel algorithm for stress integration of the elasto-plastic
model for partially saturated soils subjected to large deformation was presented.

The numerical model presented here was then validated by comparing its predictions with some
existing closed-form solutions. The capability of the finite element code developed for this
work in capturing the mechanical changes imposed by desaturation of soils is also investigated.
The error assessment of the presented algorithm in small and large deformation analysis
demonstrated in the last example showed the algorithm shows more sensitivity to the prescribed
integration parameter in large deformation analysis.

In the following chapter, various aspects of modelling the problem of dynamic compaction of
unsaturated soils will be discussed.

70
 1.4

1.2

0.8

F/α*

0.6

Updated Lagrangian

0.4

Small deformation
0.2

0
0 0.1 0.2 0.3 0.4 0.5 0.6
Displacement / (B/2)

Figure 3-7. Normalised load-displacement curves for rigid footings on the same soil in small deformation and
large deformation analysis.

71
Table 3-8. Stress errors and Normalized CPU time with different values of STol in small deformation analysis of
the rigid footing problem.

STol 𝐸𝜎 Normalized CPU Time

0.913043478
0.1 4.47E-12

0.956521739
0.01 4.47E-12

0.001 4.25E-12 0.956521739

0.0001 0.0 1

Table 3-9. Stress errors and Normalized CPU time with different values of STol in large deformation analysis of
the rigid footing problem considering stress rate objectivity.

STol 𝐸𝜎 Normalized CPU Time

0.1 1.84341E-06 0.954545

0.01 5.06187E-07 0.954545

0.001 1.73665E-08 1

0.0001 0.0 1

72
Chapter 4. Literature review of dynamic compaction

73
4.1 Introduction

This chapter begins with a discussion on the process of dynamic compaction and its application.
Previous numerical simulations, analytical solutions and experimental studies on dynamic
compaction are described.

4.2 Soil compaction

In the past engineers had to adapt their designs to the restrictions dictated by natural ground
conditions. However, recent advances in geotechnical engineering have provided the
opportunity to change the characteristics of sites by employing ground improvement
techniques. Consequently, soil improvement has become an important part of engineering
projects in which the initial in situ soil strength parameters do not meet the minimum
requirements.

Based on their application procedure, soil improvement techniques can be categorised into
different groups including replacement, reinforcement, adhesion and densification. Although
these methods differ in practice, they mainly aim to reduce the overall cost of a project, and
reduce the risk of foundation settlement. The application of these pre-emptive methods not
only reduces the project time significantly, but may also reduce considerably the total
expenditure on a civil engineering project. The great body of field data provided by researchers
in the area of ground improvements indicate that these methods can be confidently used in
different types of soil.

Soil treatment is called compaction if densification takes place by applying an external load.
In this case, loading the soil causes its particles to come closer to each other, as illustrated in
Figure 4-1. Consequently, some soil properties, such as its shear strength, will normally
increase.

In compaction methods, the external load can be applied in different ways. For instance,
movement of a truck or heavy roller can result in compaction. In one of the compaction
methods, the external load is produced by dropping a heavy weight (usually up to 40 tonnes
from a height up to 30 metres) on the ground, repeatedly.

74
 (a) (b)

Figure 4-1. Macrophotographs of (a) an uncompacted area and (b) a compacted area (Marsili et al. 1998).

This technique is called dynamic compaction (DC) because the dynamic load produced by
dropping the heavy weight forces the ground beneath to be compacted. This process involves
using a crane to lift the weight and releasing it to strike the ground surface. Subsequently, the
energy imparted to the soil by the falling weight, causing it to compact to a denser state, is
transmitted to greater depths in the soil in the form of stress waves. In each drop, the resulting
compacting enhances the soil strength and reduces its compressibility. Based on the original
soil conditions and its desired characteristics after DC, a grid pattern is usually selected from
which the locations of the drops are determined. The degree of compaction depends on many
factors such as the mass of the falling weight, the drop height, and the spacing of the drop
locations, and can be assessed in the field by conducting in situ tests such as CPT or SPT at the
end of each pass of the falling weight.

75
4.3 History

The process of ground improvement by dropping heavy weights has an ancient history. Menard
and Broise (1975) reported that historical use of dynamic compaction (DC) can be found by
looking at early Chinese paintings indicating the fact that this method could be centuries old.
Kérisel (1985) also reported that this technique had been used by the Romans in ancient times.
In the 21st century, the invention of large cranes provided the opportunity to apply higher
energies to the ground. In the 1970s, dynamic compaction was scientifically introduced to the
construction industry by Menard Techniques Limited (Ménard and Broise 1975). Initially, the
lower capacity of early cranes had limited the amount of applied energy, but further
improvements in these techniques eventually made it possible to apply very large energies to
the ground with each drop. Since its introduction, dynamic compaction has exhibited its
versatility and simplicity of use in different types of civil engineering projects around the
world, including building structures, container terminals, highways, airports, dockyards, and
harbours, as reported by Moseley and Slocombe (1978), Leonards et al. (1980), Mayne et al.
(1984), Lukas (1986), Dise et al. (1994), Hansbo (1996), Van Impe and Bouazza (1996),
Merrifield and Davies (2000), Satyapriya and Gallagher (2000), Nashed et al. (2004), Miao et
al. (2006), Perucho and Olalla (2006), Feng et al. (2011), Zekkos et al. (2013) and Jia et al.
(2015).

4.4 Application of dynamic compaction to different soil categories

Dynamic compaction (DC) was initially used for the improvement of granular soils, and was
found to be economically attractive in comparison with other techniques such as removal and
replacement. Later, many researchers including Mayne et al. (1984), Lo et al. (1990), Hansbo
(1996) and Perucho and Olalla (2006) demonstrated that dynamic compaction can also be
effectively applied to cohesive soils and the term “dynamic consolidation” was adopted for
referring to the interaction taking place between the solid and fluid phases, as they mostly show
an undrained response when the energy is first applied. In both cohesive and granular soils,
there seems to be a common agreement among different studies that the induced displacement
of the soil particles due to dynamic compaction usually increases the density, bearing capacity
and stiffness of the soil. However, Wang et al. (2009) observed that the existence of fine
particles (diameter <0.075 mm) in the soil skeleton can significantly alter the response of the
soil to dynamic loading. That is mainly because fine content can affect the permeability of soil

76
which determines the dissipation speed of the excess pore water pressures generated by the
impact.

Gunaratne et al. (1996) demonstrated that within the short duration of the impact load, free
draining soils with relatively high values of permeability show no significant pore pressure
development while impervious soils behave in an undrained fashion. Therefore, in predicting
the response of soil subjected to dynamic compaction it is important to categorise the soil based
on the size of the particles it contains.

Based on values of permeability, Lukas (1986) identified the ranges of soils which could be
effectively improved by dynamic compaction, identifying three different zones of permeability
(or hydraulic conductivity). The first zone mostly consists of granular soils like sands and
gravels as well as municipal solid waste on which dynamic compaction was applied at its
primary stage of development. Cohesive soils, silts and sandy silts are mainly in the second
and third zones. According to Lukas (1986), deposits with a very low permeability (<
10−8 m/sec) and poor drainage would not be suitable for dynamic compaction treatment due
to the negative effects of excess pore pressure. On the other hand, the most favourable category
of soils for application of DC is granular soils. Thus, the efficiency of dynamic compaction
strongly depends on the drainage conditions and the speed of pore pressure dissipation.

This categorisation of soil becomes particularly important in the simulation of ground response
under dynamic loads because, unlike coarse granular material, less permeable soils may
experience a consolidation phase after receiving the impact loads. This effect was reported by
Wang et al. (2000). They illustrated that the behaviour of a clay fill is improved by DC together
with the installation of drains to facilitate drainage. The results indicate that depending on the
drainage condition, post-improvement of cohesive soils as a result of consolidation can be even
more important than the sudden settlement during the DC.

4.5 Prediction of ground response to dynamic compaction

Many researchers have studied dynamic compaction experimentally. Generally, laboratory

experimental studies of DC can be divided into two categories including box models and
centrifuge models. Nevertheless, despite the wide use of DC as an economically attractive
ground improvement technique and the abundance of experimental data for dynamic
compaction, few detailed predictive methods for DC are available. There is a demand for a

77
reliable theory to predict and explain the complicated nature of DC. This would enhance the
understanding of this mechanism together with providing a useful engineering tool for studying
the determinative parameters of this problem. This may also result in finding a more credible
and rational method for estimating the overall cost and time of a DC project.

4.5.1 Empirical equations

After the pioneering publications in the 1970s describing field observations of dynamic
compaction treatment, researchers have tried to find a relation between the depth of
improvement, as a measurement of the degree of ground improvement, and the impact energy.
Such relations can provide a quick solution for design purposes in a given project.

A simplified empirical equation for an approximate evaluation of the depth of improvement

after dynamic compaction was proposed by Menard and Broise (1975), which is still in popular
use in practice today, and is given by

𝑑𝑚𝑎𝑥 = √𝑊𝐻 (4.1)

where 𝑑𝑚𝑎𝑥 is the depth of improvement in metres, 𝑊 is the mass of the tamper in tonnes and
𝐻 is the drop height in metres. Several researchers have since investigated and calibrated the
original equation to adapt it to different DC sites under various conditions as follows:

𝑑𝑚𝑎𝑥 = 𝑛√𝑊𝐻 (4.2)

where 𝑛 = an empirical constant that accounts for influencing factors such as soil type and DC
procedure. A range of values has been proposed for 𝑛 in different types of material, some of
which are listed in Table 4-1.

Slocombe (1993) also presented a nonlinear relationship between tamping energy and the depth
of improvement in the equation below.

𝑑𝑚𝑎𝑥 = 0.586√𝑊𝐻 − 0.009𝑊𝐻 (4.3)

However, it is notable that no clear and unambiguous definition of the “depth of improvement”
was provided by Menard and Broise (1975). Later, different studies were carried out to
eliminate this ambiguity. For example, Leonards et al. (1980) suggested that the depth where

78
the value of 𝑁 measured in a standard penetration test increased at least three times in
comparison with its original value can be considered as the depth of improvement. Another
approach was proposed by Oshima and Takada (1998) in which the depth of improvement is
located at the depth where the relative density of the soil is increased by at least 5%.

Table 4-1. Values of 𝒏 proposed for different types of material under dynamic compaction

Reference Proposed value of 𝑛 Deposit

Lukas (1980) 0.65-0.8 Miscellaneous fill

Leonards et al. (1980) 0.5 Dry sands

Mayne et al. (1984) 0.3-0.8 Sands

Van Impe and Bouazza (1996) 0.5-0.65 Municipal soil waste

Rollins et al. (1998) 0.4 Collapsible soils

Kumar and Puri (2001) 0.53 Sand, gravels and silt

Shui et al. (2006) 0.35 Backfill and silt

Feng et al. (2010) 0.35 Backfill, sand and Clay

Zekkos et al. (2013) 0.4 ± .05 Municipal soil waste

However, such empirical equations might not be confidently used to design many dynamic
compaction applications, as there are several influencing parameters that are not incorporated
into these simple equations. Moreover, this simple method of design cannot provide an answer
for prediction of the depth of improvement inside the improved zone or after several rounds of
compaction. By conducting a series of numerical simulations, Ghassemi et al. (2008) showed
that Menard’s equation overestimates the depth of improvement for higher impact loads.

Because of the mentioned drawbacks in such empirical approaches, much research has been
conducted to find a more reliable estimation of ground behaviour under DC. These models can
be used as a supplementary means for solving other dynamic geotechnical engineering
problems and are useful in preliminary design phases and feasibility studies. These methods
can be generally categorized into analytical solutions and numerical simulation. The latter has
become popular in recent decades due to the availability of modern computers and their
capability to conduct large-scale computations efficiently.

79
4.5.2 Analytical solutions

The complicated behaviour of soil subjected to DC is not thoroughly understood. However,

there are a few idealised models that have been developed particularly for analytical simulation
of the soil response during DC. Scott and Pearce (1975) proposed some recommendations in
the analytical simulation of dynamic compaction. Later, Mayne and Jones (1983) suggested a
one-dimensional solution based on wave transmission theory in elastic media for predicting the
response of soil under DC. Mayne and Jones (1983) assumed that the force-time graph during
the impact is a triangular impulse, and the stress distribution under the weight is trapezoidal.
With these assumptions the maximum stress under the centre of impact, 𝜎𝑧𝑚𝑎𝑥 , is expressed by

𝑐𝑠 √𝑊𝐻𝐵
𝜎𝑧𝑚𝑎𝑥 = (4.4)
(𝐵 + 𝑧)2

where 𝑧 is the depth beneath the loaded surface, W denotes the mass of the falling weight, H is
the height of fall, 𝑐𝑠 represents the shear wave velocity of the soil, and B is the width of the
falling weight which is obtained from 𝑟 𝑜 √𝜋, 𝑟 𝑜 being the plan radius of the falling object. In
this equation the units of mass, length and time are ft, lb and sec, respectively. However,
incorporation of even some of the most prevalent observations in DC, such as the occurrence
of irreversible strain, is beyond the capability of this simple model.

Based on the same theory of one-dimensional wave propagation, Chow et al. (1990) presented
a semi-analytical solution for DC by assuming an elastic soil column beneath the impact load
with a length equal to the expected depth of improvement. They considered an axisymmetric
column which is laterally confined by the surrounding soil. The surrounding soils were
modelled by a number of springs and dashpots which represent numerically the dynamic soil
stiffness and damping, respectively. However, because soil, particularly in the vicinity of the
impact area, will undergo a very high strain level, a linear spring and a linear dashpot cannot
accurately model this phenomenon.

Later, Chow et al. (1992) considered the non-linear response of the soil by employing a strain-
hardening constitutive model for loose cohesionless soils in a similar spring-dashpot
representation of the soil. However, both of these models have had restricted use and their
applicability has not been verified for different types of soil and different initial conditions.

80
Moreover, no clear definition or approximate procedure for determining the “expected depth
of improvement”, as a key assumption of their analysis, was addressed in this work.

Thilakasiri et al. (1996) suggested that there are three different regions in the soil which behave
differently under applied dynamic load. There is a region of intense plastic shearing
immediately under the falling weight surrounded by a second plastic region and a third elastic
outer region. Equations of motion are written separately for the three zones, distinguished by
their mode and degree of deformation, while also satisfying compatibility. Moreover, the
different degrees of deformation of the distinct zones are also considered in the model by using
nonlinear stiffness properties where necessary. Based on these assumptions, an analytical
model was developed to estimate the surface stress and the surface deformation. A numerical
procedure was adopted to solve these equations of motion.

However, none of the methods mentioned here was able to predict the pore pressure distribution
generated during DC and the effect of subsequent consolidation. Knowledge of the pore
pressure generation during DC is particularly important in predicting the post-impact behaviour
of the soil during the consolidation stage. Despite the abundance of field observations of
induced pore pressures in the literature, e.g., Gunaratne et al. (1996), Ménard and Broise
(1975), Mayne et al. (1984), Hansbo (1996), Wang et al. (2000), Perucho and Olalla (2006)
and Jia et al. (2015), analytical or numerical approaches for pore pressure prediction are scarce,
due largely to the complex dynamic response of soil in the presence of pore water. Moreover,
the depth of improvement can significantly decrease due to the negative influence of pore water
pressure. Also there is evidence that local liquefaction is probable when impact loads are
repeatedly applied before the excess pore pressures have sufficient time to dissipate (Ménard
and Broise 1975). In the absence of a drainage system, further drops of the tamper are usually
no longer effective and impose additional unnecessary costs on the project. Therefore, it is
essential to predict the level of the induced pore water pressures during heavy tamping.

An analytical model for predicting the generation of pore water pressure during DC was first
pointed out by Gunaratne et al. (1996) based on laboratory dynamic consolidation data.
Elasticity theory was first employed to determine the vertical stress Δ𝜎𝑣 at an arbitrary depth 𝑧
vertically under the centre of the falling weight, by following equation

81
 1
Δ𝜎𝑣 = Δ𝜎0 [1 − ] (4.5)
𝑟 2
[1 + ( 𝑧0 ) ]3/2

where Δ𝜎0 is the dynamic stress beneath the falling weight of radius 𝑟0 .

One limitation of this approach is that it does not account for the time delay between the impact
and the arrival of the resulting stress wave at the depth z. Moreover, the theory of elasticity
may provide a relatively good approximation for stress distribution in the range of limited
energies applied in laboratory studies, but it fails to provide solutions for higher tamping
energies that cause large strains inside the soil mass.

Based on experimental data, Nashed et al. (2004) suggested the following equations for
measurement of the total energy loss, 𝑤𝑐 , per unit volume of soil mass

𝑤𝑐 = 𝑤𝑅 + 𝑤𝐵 (4.6)

This is composed of energy loss due to Rayleigh waves 𝑤𝑅 and body waves 𝑤𝐵 which are
expressed as

𝛼𝑒 −2𝛼𝑟
𝑤𝑅 = 𝐹(0.67𝑊𝐻) (4.7)
𝜋𝑟

𝛼𝑒 −2𝛼𝑅
𝑤𝐵 = 𝐹(0.33𝑊𝐻) (4.8)
𝜋𝑅 2

where, 𝐹 is the integral function proposed by Nashed et al. (2004), 𝛼 is a material damping
parameter, and 𝑅 = √𝑟 2 + 𝑧 2 .

Later, Nashed et al. (2006) used these equations to develop a semi-analytical solution for
impact-induced pore pressure generation and the consequent densification process in saturated
sand deposits. Consolidation theory was employed to explain the post-impact behaviour during
the consolidation process. The implementation of these equations was performed using the
finite difference method. The wick drain effect on the densification process was also
investigated. However, this model also neglects the non-linear response of soil during DC, as
the energy loss equations are obtained for an elastic half-space.

82
4.5.3 Finite element simulations

As the number of substantial parameters incorporated in the analysis of this problem grows, it
becomes more difficult to use analytical methods. Hence, due to the impressive growth in
computational capability of modern computers, numerical solutions have been gradually
preferred to analytical solutions. However, numerical simulation of DC has remained a
challenging problem for many researchers. Some of these important challenges are:

 The occurrence of large deformations, finding an appropriate constitutive model for

describing the non-linear response of the soil during DC,
 Mesh distortion and singularity issues arising during the finite element analysis, and
 Describing the behaviour of soil at the consolidation stage.

Poran and Rodriguez (1992) were among the first who simulated dynamic compaction within
the framework of finite element method (see Figure 4-2). Large deformation of soil beneath
the impact area was considered along with using two Drucker-Prager-type elasto-plastic
models to study soil response under DC. Furthermore, a remeshing technique was employed to
enhance the shape of the distorted mesh and improve convergence of the analysis. Their model
showed good agreement with experimental results particularly in relatively loose sand, but after
the first few drops the predicted results deviated from the experimental data.

Pan and Selby (2002) used Abaqus to model the response of loose granular soil under an impact
load of a rigid body. A Mohr-Coulomb type constitutive model with a non-associated flow rule
was employed to simulate the elasto-plastic behaviour of the soil mass. They divided their
model vertically into three separate regions. In each zone, soil parameters were updated after
each drop to simulate the variation of the mechanical properties of the soil after each impact.
However, there is no explanation in their work about how they updated the soil parameters.

83
 Figure 4-2. Two-dimensional finite element model used by Poran and Rodriguez (1992)

The dynamic load was applied using two methods:

 A force time curve based on field deceleration data of falling weights (see Figure 4-3).
 Assignment of an initial velocity based on the height of the drop, 𝑉 = √2𝑔ℎ.

It was illustrated that both of these procedures can reasonably model P-wave transmission in
the soil, although their observations indicated that the depth of improvement was overestimated
compared to the predictions of empirical equations.

84
 10

Force (MN)

0
0 0.01 0.02 Time(s) 0.03 0.04 0.05

Figure 4-3. Force-time curve used by Pan and Selby (2002).

An Updated-Lagrangian formulation was employed by Lee and Gu (2004) to simulate large

straining involved in dynamic compaction processes. In addition, an elasto-plastic cap model
was used together with an assumption that the elastic soil modulus does not need to be updated
during multiple impacts. This was based on observations that plastic behaviour of the soil plays
a more important role than its elastic response during dynamic compaction. The numerical
results were sufficiently accurate compared to centrifuge modelling data of DC. The influence
of drop energy, momentum of the falling weight, and tamper radius on the depth of
improvement was also exhibited. They recommended that dropping masses from lower heights
together with selecting heavier weights can increase the depth of influence.

Ghassemi et al. (2008) numerically investigated the capability of Menard’s empirical equation
for predicting the depth of improvement. The effect of initial relative density, tamper radius,
number of drops and the impact energy were studied using a finite element model. It was found
that suitable results can be obtained by Menard’s empirical relation particularly when the
weight and the height of drop are both within the normal practical ranges. However, for higher
impact energies, Menard’s relation overestimates the depth of improvement. Their predicted
results also indicated that the value of the empirical coefficient in the Menard formula depends
strongly on the applied energy and tamper radius. Based on this finding, a new relation for
predicting the improvement depth was proposed and applied to a number of DC projects.

85
Mostafa (2010) focused on numerical simulation of DC of cohesive soils by employing the
Modified Cam-Clay model. The finite element model was implemented into Abaqus in both
two and three dimensional models. However, no comparison of two and three dimensional
results was presented in that study.

López‐Querol et al. (2008) were among the first to employ a coupled finite element code
formulated in terms of displacements for both solid and fluid phases to simulate dynamic
compaction. The behaviour of granular soils was modelled within the framework of Biot’s
equations for saturated media by selecting the constitutive law developed by Pastor et al. (1990)
to simulate the nonlinear response of the soil. They illustrated that the results of a 𝐮 − 𝑝𝑤 − 𝑤
formulation (with 𝐮 as displacement, 𝑤 the relative velocity of fluid and 𝑝𝑤 as pore pressure)
are similar to those of a displacement-pore pressure formulation. However, they suggested that
one of the main drawbacks of the 𝐮 − 𝑝𝑤 − 𝑤 formulation is that it involves a higher number
of degrees of freedom compared to the 𝐮 − 𝑝𝑤 approach. Therefore, it reduces the speed of
the analysis.

More recently, Ghassemi et al. (2010) investigated dynamic compaction efficiency in saturated
soils, recognising that soils under fully saturated conditions generally exhibit less improvement
by DC due to absorption of a part of the applied energy by the liquid phase. This can reduce
the speed of consolidation of the soil. They carried out a sensitivity analysis to see how the
depth of improvement in DC is affected by the hydraulic conductivity of saturated soils. Their
results indicate that in granular material permeability does not play an important role during
the impact phase while it can substantially alter the soil response during the consolidation
phase. When consolidation occurs, soils with higher permeability will have a larger improved
zone.

Moreover, by comparing the results with those of obtained from an analysis of the same soil in
a fully-dried condition, they came to this conclusion that the extent of the improved zone under
fully saturated conditions is less than that of a completely dry soil. Finally, they concluded that
most of the improvement due to DC takes place during the undrained phase and only a small
portion of it occurs during the consolidation phase.

86
4.6 Dynamic compaction of partially saturated soils

In practice, a large number of engineering problems involve the presence of unsaturated soil
zones where the voids between the soil particles are filled with a mixture of air and water.
These zones are usually ignored in numerical simulations and the soil is assumed to be either
entirely saturated or completely dry. This is despite the experimental results and field
observations which indicate significant differences between the mechanical behaviour of
saturated, partially saturated and dry soils, e.g., Mitchell et al. (1965), Howard et al. (1981),
Daniel and Benson (1990), Adams and Wulfsohn (1998), Karube and Kawai (2001), Cokca et
al. (2004), Rahardjo et al. (2004) and Ruiz et al. (2005). Indeed, there are numerous studies
indicating the fact that moisture content can significantly change the dynamic response of a
soil.

A typical observation reported by many researchers is that soils in different degrees of

saturation show different responses to dynamic loads, especially in compaction studies. This is
generally expressed in the form of a compaction curve where the water content at which the
dry density is at its maximum can be found for a given compaction energy, e.g., Proctor (1933),
Hilf (1956), Lee and Suedkamp (1972). Despite previous efforts to incorporate pore pressure
effects in the dynamic analysis of soil media, there are few which can explain the effect of
water content on the mechanical behaviour of a soil.

It has been shown that numerical and analytical methods for simulating DC have generally not
progressed beyond providing solutions for fully saturated conditions. Also, dynamic
compaction is usually avoided for fully saturated soils unless the drainage conditions allow fast
dissipation of pore pressure Mayne et al. (1984). But up to now, the majority of improvements
have been achieved in clayey fill deposits that are only partially saturated. This includes fills
elevated well above the water table and with good surface drainage. In this case, improvement
occurs as the particles are compacted before the deposits become fully saturated. After
saturation occurs, negligible further improvement will be realized regardless of the amount of
energy applied.

Lukas (1986) also suggested that the water content of the clayey soils prior to dynamic
compaction should be less than the plastic limit of the deposit. Field experience with dynamic
compaction and laboratory Proctor testing indicates that compaction efficiency is related to the
soil moisture content and that an optimum compaction moisture content probably exists. In the

87
case of dynamic compaction, Rollins et al. (1998) evaluated compaction efficiency on six field
test cells with different water contents. They illustrated that there is an optimum water content,
similar to other compaction methods, at which the maximum efficiency can be achieved for a
certain level of compaction energy.

Thus, higher compaction efficiency could probably be achieved by pre-wetting the soil to the
optimum water content prior to dynamic compaction. In addition, field test data indicate that
pre-wetting of soil can also reduce vibrations associated with dynamic compaction.

However, Rollins and Rogers (1994) pointed out that if the moisture content becomes too great,
large pore water pressures will be generated upon impact preventing the occurrence of further
densification. This suggests that in order to find an accurate and comprehensive solution soil
should be modelled as a three-phase porous medium to accommodate various degrees of water
saturation. This is especially true for the soil mass surrounding the source of energy release, as
each of the three phases is likely to respond differently to dynamic loading.

4.7 Summary
In this chapter a review of former numerical simulations, analytical solutions and field
observations of dynamic compaction was presented. The drawbacks and shortcomings of the
previous methods in capturing hydro-mechanical changes in dynamic compaction of partially
saturated soils were discussed in the last part. In the next chapter, numerical modelling of
dynamic compaction in partially saturated media is presented.

88
Chapter 5. Numerical simulation of dynamic
compaction in partially saturated soils

89
 5.1 Introduction
Following the literature review on dynamic compaction presented in the previous chapter, in
this chapter a numerical study of dynamic compaction in partially saturated soils is presented.
The chapter also covers some important aspects of modelling the dynamic behaviour of
unsaturated soils such as the use of energy absorbing boundary conditions.

5.2 Modelling dynamic compaction in partially saturated soil

In this example we use the constitutive model described in Chapter 3 to model the non-linear
response of soil subjected to dynamic compaction. A cylindrical falling weight of diameter 2 m
and a mass of 10000 kg is considered in the analysis. The weight was modelled as rigid body
due to its relatively high stiffness compared to that of the soil skeleton.

A schematic representation of the FEM model is presented in Figure 5-1. All boundaries are
considered impermeable except the top boundaries. Due to the axis-symmetric nature of the
problem two viscous boundaries with artificial damping are considered along the right hand
side and the bottom of the mesh to damp out reflected waves. 6-noded iso-parametric elements
are used over the whole domain. These elements are 6-noded displacement elements which are
coupled with two 3-noded elements corresponding to the degrees of freedom for suction and
pore pressure. The cylindrical dropped weight was also located above the soil on the top left
hand side of the mesh.

5.2.1 Discussion of viscous boundaries

The right and bottom boundaries are assumed to be energy absorbent to prevent spurious
reflections from the fixed boundaries. This is an alternative approach to the most commonly
used element boundaries of either zero stress (Neumann condition) or zero displacement
(Dirichlet condition) which act as perfect reflectors. This type of boundary is based on the
standard viscous boundary presented for solids by Lysmer and Kuhlemeyer (1969), but
extended to unsaturated soils in this study.

The basic idea of this method involves applying traction forces at the artificial boundaries. The
traction forces then dictate any reflected stresses to have zero magnitude, as follows:

𝑏 𝑙
𝜕𝑢𝑏
𝜎 + 𝛼 𝜌𝑐𝑝 =0 (5.1)
𝜕𝑡

90
 1.0 m
permeable

smooth / impermeable/viscous
Axisymmetric line

10 m
rough /impermeable/viscous
10 m

displacement nodes
displacement, pore pressure and suction nodes

Figure 5-1. Finite element model for the problem of dynamic compaction.

𝑏 𝑙
𝜕𝑢𝜏𝑏
𝜏 + 𝛽 𝜌𝑐𝑠 =0 (5.2)
𝜕𝑡

where 𝑢𝑏 and 𝑢𝜏𝑏 are the normal and tangential displacement at the boundary while 𝜎 𝑏 and 𝜏 𝑏
represent the normal and tangential stress at the boundary, respectively. 𝜌 represents the
material density whereas 𝑐𝑝 and 𝑐𝑠 are the pressure and shear wave velocity. For a solid soil
skeleton these values are given by:

91
 (𝜆̃ + 2𝐺)
𝑐𝑝 = √ (5.3)
𝜌

𝐺
𝑐𝑠 = √ (5.4)
𝜌

where 𝐺 and 𝜆̃ are the Lamé parameters of the theory of elasticity.

Also, 𝛼 𝑙 and 𝛽 𝑙 are dimensionless parameters, for which the optimal value of 𝛼 𝑙 and 𝛽 𝑙 was
proposed as 1 by Lysmer and Kuhlemeyer (1969). Later, the study by Cohen (1980) showed
that the method does not show great sensitivity to the values adopted for 𝛼 𝑙 and 𝛽 𝑙 .

The standard viscous boundary is perhaps the most popular scheme, as it offers an acceptable
capability of damping waves at the low cost of implementation. However, it may show
inadequate accuracy in damping Rayleigh waves.

Degrande and De Roeck (1993) indicated that in the low frequency range, discussed in Chapter
2, the fully saturated porous medium behaves like a frozen mixture. In this condition it is
acceptable to assume the relative motion of the pore fluid with respect to the solid skeleton is
negligible. This has also been shown by Heider et al. (2011), where the following values for
𝑐𝑝 and 𝑐𝑠 were proposed for dynamic analysis of fully saturated soils in the low frequency
range:

𝐺
𝑐𝑠 = √ (5.5)
𝜌𝑠 + 𝜌𝑠

𝜆̃ + 2𝐺
𝑐𝑝 = √ (5.6)
𝜌𝑠 + 𝜌𝑤

where 𝜌𝑠 and 𝜌𝑤 are the density of solid and water phases, respectively.

92
Albers (2009) also proposed a general solution for the wave propagation problem inside an
elastic partially saturated porous medium. Based on this result, the following equations for 𝑐𝑝
and 𝑐𝑠 in a three-phase material consisting of solid, liquid and gas phases in the low frequency
range can be obtained:

𝐺
𝑐𝑠 = √ (5.7)
𝜌𝑠 + 𝜌𝑤 + 𝜌𝑎

𝜆̃ + 2𝐺
𝑐𝑝 = √ (5.8)
𝜌𝑠 + 𝜌𝑤 + 𝜌𝑎

with 𝜌𝑎 is the density of gas phase (air in the case of partially saturated soils).

The damping matrix of elements sharing a side with the viscous boundary 𝐂 𝐄 includes the
contribution of the material damping matrix 𝐂 as well as the contribution of damping due to
the boundary, 𝐂 𝐀𝐁𝐂 , and is calculated by:

𝐂 𝐄 = 𝐂 + 𝐂 𝐀𝐁𝐂 (5.9)

and

𝐂 𝐀𝐁𝐂 = ∫ 𝐍𝐮𝐓 𝐜̃𝐍𝒖 𝑑𝐿 (5.10)

𝐿𝐴𝐵𝐶

where 𝐿𝐴𝐵𝐶 is the surface of the element over which the absorbing boundaries act and 𝐜̃ is
defined by:

𝑐𝑠 0
𝐜̃ = [ 0 𝑐𝑝 ] (5.11)

The constitutive and material parameters selected for this example are shown in Table 5-1 and
the parameters of the soil water characteristic curve defined by Brooks and Corey (1964) and
also used in this example are shown in Table 5-2. In the first step of the analysis the body force
was applied to generate the geostatic stresses inside the mesh and to prepare the initial

93
conditions for the next step of the analysis. In this next step the rigid body was released to
impact the surface of the soil with different velocities. The analysis was terminated when the
rigid body came to rest. By changing the initial velocity, the problem was studied for a range
of applied energies.

5.3 Results and discussion

Finding a set of experimental data based on case studies of dynamic compaction in unsaturated
soils can be challenging. One of the key challenges in obtaining such data can be difficulties in
the measurement of suction in the field before and after dynamic compaction of the site.
Furthermore, most case studies and reports available in the literature mainly discuss the
application of dynamic compaction on granular soils, which are not the main subject of this
thesis. Therefore, to the best of the author’s knowledge, rigorous validation of the numerical
results with a case study is not possible. However, a set of field observations and previous
numerical simulations of the application of dynamic compaction on cohesive materials are used
here to verify the obtained numerical results.

5.3.1 Mechanical response of soil to dynamic compaction

Figure 5-2 shows the settlement of the ground predicted by the numerical method versus the
applied energy in one impact of the falling mass. As the graph suggests, increasing the applied
energy will increase the settlement of the ground.

One important parameter in models based on Cam-Clay plasticity is the pre-consolidation

pressure as it also defines the initial size of the yield surface. This also represents the level of
stress that the soil had previously experienced. Therefore, it is anticipated that when other
parameters of soils are kept constant during the analysis the soil with a higher pre-consolidation
pressure shows a higher resistance to dynamic loadings.

To study this effect, eleven different analyses were conducted with pre-consolidation pressures
of 50, 60, 70, 80, 90, 100, 110, 120, 130 and 150 kPa, while keeping the other parameters of
the problem constant. The predicted results are shown in Figure 5-3 for an applied energy level
of 245 kN.m.

94
Table 5-1. Constitutive model and material parameters for the problem of dynamic compaction.

Parameter type Symbol Value Unit

The slope of critical state line 𝑀 1.2 -

The slope of loading line in fully

𝜆 0.25 -
saturated state

Poisson’s ratio of the mixture 𝜇 0.3 -

The slope of unloading/reloading line

𝜅 0.05 -
in fully saturated state

Specific volume at 𝑝′ = 1 kPa 𝛤∗ 2.8 -

The initial size of the yield surface 𝛼∗ 100 kPa

Unsaturated material parameter 𝛽 0.012 1/(kPa)

Unsaturated material parameter 𝑟 0.9 -

Density of the solid skeleton 𝜌𝑠 2650 kg m−3

Density of the water 𝜌𝑤 997 kg m−3

Density of the air 𝜌𝑎 1.1 kg m−3

Elastic modulus of the mixture 𝐸 1.2 × 104 kPa

Bulk modulus of the solid skeleton 𝐾𝑠 3.5 × 107 kPa

Bulk modulus of the water 𝐾𝑤 2.25 × 106 kPa

Bulk modulus of the air 𝐾𝑎 1.01 × 102 kPa

Intrinsic permeability 𝑘 2.5 × 10−12 m2

Viscosity of the water 𝜂𝑤 1.0 × 10−3 Ns m−2

Viscosity of the air 𝜂𝑎 1.8 × 10−5 Ns m−2

In this graph the pre-consolidation pressure, 𝛼 ∗ is normalised by the unit weight of the falling
object, 𝑀𝑜 per impact area, 𝐴𝑜 as follows:

𝛼∗
𝛼̅ ∗ =
𝑀𝑜 (5.12)
( 𝑜)
𝐴

where 𝛼̅ ∗ is the normalized pre-consolidation pressure, and

95
𝐴𝑜 = 𝜋. (𝑟 𝑜 )2 (5.13)

where 𝑟 𝑜 is the radius of falling weight.

Table 5-2. Water retention curve parameters for the problem of dynamic compaction.

Parameter type Symbol Value Unit

Bubbling pressure 𝑃𝑑 40 kPa

Residual wetting fluid saturation 𝑆𝑟𝑤 0.4 -

Air entry saturation 𝑆𝑟𝑎 0.8 -

Pore size distribution index 𝑣𝑏 3.0 -

As the graph suggests the ground settlement has an inverse relationship with the pre-
consolidation pressure.

0 120 240 360 480 600 720

0

-0.25

-0.5
Displacement (m)

-0.75

-1
Energy (kN.m)

Figure 5-2. Settlement of the ground versus applied energy.

96
The variation of ground settlement with the slope of the Critical State Line, 𝑀, is also shown
in Figure 5-4 for an energy level of 245 kN.m. The values for the slope of Critical State Line
were chosen to be 0.85, 0.9, 1.0, 1.1, and 1.2 in five different analyses in which the other
parameters were maintained fixed. As the graph suggests an inverse relationship is seen
between 𝑀 and the ground settlement, i.e., potentially stronger soil undergoes less permanent
displacement, as expected.

Normalised pre-consolidation pressure

1.5 2 2.5 3 3.5 4 4.5 5

-0.45

-0.5

-0.55
Displacement (m)

-0.6

-0.65

-0.7

-0.75

Figure 5-3. Variation of displacement of the ground with changes in pre-consolidation pressure for an energy
level of 245 kN.m.

97
The changes in ground settlement versus 𝜆 are also shown in Figure 5-5 for an applied energy
level of 245 kN.m. Five different analyses were performed in which 𝜆was chosen to be 0.15,
0.2, 0.25, 0.3, 0.35 while the rest of the parameters were unchanged. As this Figure shows,
increasing the slope of the loading line, 𝜆, can significantly increase the predicted ground
settlement. This is also as anticipated, since increasing values of 𝜆 imply more compressible
soil.

The variation of permanent ground settlement with changes in Poisson’s ratio is also shown in
Figure 5-6. Keeping all other parameters constant and assuming an applied energy level of 245
kN.m, five different analyses were performed with Poisson’s ratios of 0.15, 0.2, 0.25, 0.3, 0.35
and 0.4. The presented results indicate an inverse relationship between the ground settlement
and Poisson’s ratio. This is also as expected, since higher values of Poisson’s ratio imply
smaller volume compressibility, with a value of 0.5 corresponding to incompressible elastic
behaviour of the soil.

Mostafa (2010), who presented a numerical simulation of dynamic compaction of dried

cohesive soils using the Abaqus software, also reported a direct relationship between the slope
of loading line and the ground settlement while an inverse relationship was reported between
the slope of critical state line, Poisson’s ratio and pre-consolidation pressure with the ground
settlement.

M
0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
-0.5

-0.55

-0.6
Displacement
(m) -0.65

-0.7

-0.75

-0.8

Figure 5-4. Variation of displacement of the ground with changes in M for an energy level of 245 kN.m.

98
 𝜆
0.15 0.2 0.25 0.3 0.35
-0.48

-0.5

-0.52

-0.54

-0.56
Displacement(m)
-0.58

-0.6

-0.62

-0.64

-0.66

Figure 5-5. Variation of displacement of the ground with changes in 𝝀 for an energy level 245 kN.m.

ν
0.15 0.2 0.25 0.3 0.35 0.4
-0.54

-0.55

-0.56

Displacement(m) -0.57

-0.58

-0.59

-0.6

Figure 5-6. Variation of displacement of the ground with changes in Poisson’s ratio for an energy level of 245
kN.m.

99
5.3.2 Changes in void ratio during dynamic compaction

By conducting a digital image analysis of semi-saturated clay fill before and after dynamic
compaction, Hu et al. (2005) showed that larger compaction energy levels can cause higher
reduction in the porosity of clay fills. Although an exact comparison with the presented work
of Hu et al. (2005) is not possible, a comparative study on the variation of void ratio 15 cm
beneath the centre of impact under different levels of energy is shown in Figure 5-7, which
indicates a direct relationship between the applied energy and changes in void ratio, as also
reported by Hu et al. (2005).

The effects of some of the parameters of the constitutive model on the predicted profile of void
ratio with depth are also shown in the series of studies presented below. In all these studies the
energy level was assumed to be 245 kN.m and except for the particular parameter being studied
all other soil parameters were maintained constant across all analyses.

-0.27

-0.29

-0.31

-0.33

Δe -0.35

-0.37

-0.39

-0.41

-0.43
125 225 325 425 525 625

Energy (kN.m)

Figure 5-7. Comparison of changes in void ratio at 15 cm beneath the centre of impact in different levels of
energy.

100
Figure 5-8, Figure 5-9, Figure 5-10, Figure 5-11 and Figure 5-12 present the predicted
variations of void ratio with depth (along the axis-symmetric line) corresponding to changes in
the slope of critical state line, the slope of the virgin compression line, the pre-consolidation
pressure, Poisson’s ratio, and the slope of unloading/reloading line, respectively.

Δe
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1
0

-1

-2

-3

M=0.85 -4
Depth (m)
M=0.9
-5
M=1.0

M=1.1
-6
M=1.2

-7

-8

-9

-10

Figure 5-8. Changes in void ratio with depth along the axis-symmetric line with different values of M.

101
 Δe

-0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05
0

-1

-2

λ=01.5
0.15
-3

λ=0.2 Depth (m)

-4
λ=0.25

λ=0.3 -5

λ=0.35
-6

-7

-8

-9

-10

Figure 5-9. Changes in void ratio with depth along the axis-symmetric line with different values of the slope of
loading line.

102
 Δe

-0.5 -0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05
0

α*=50 kPa -1

α*=60 kPa
-2

α*=70 kPa

-3
α*=80 kPa

Depth (m)
α*=90 kPa -4

α*=100 kPa
-5
α*=110 kPa

-6
α*=120 kPa

α*=130 kPa
-7

α*=140 kPa
-8
α*=150 kPa

-9

-10

Figure 5-10. Changes in void ratio with depth along the axis-symmetric line with different values of pre-
consolidation pressure.

103
 Δe

-0.46 -0.36 -0.26 -0.16 -0.06 0.04

0

-1

-2

-3

Poisson's ratio=0.15 Depth (m)

Poisson's ratio=0.2 -4
Poisson's ratio=0.25

Poisson's ratio=0.3
-5
Poisson's ratio=0.35

Poisson's ratio=0.4
-6

-7

-8

-9

-10

Figure 5-11. Changes in void ratio with depth along the axis-symmetric line with different values of ν.

104
 Δe
-0.5 -0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05
0

-1

-2

-3

κ=0.03
κ=0.04 -4 Depth (m)
κ=0.05
κ=0.06
-5
κ=0.07
κ=0.08
-6

-7

-8

-9

-10

Figure 5-12. Changes in void ratio with depth along the axis-symmetric line with different values of 𝜿.

5.3.3 Other aspects of the mechanical response of soil

Figure 5-13 also shows a comparison of the predicted settlement of the soil surface for different
degrees of saturation. As Rollins et al. (1998) suggests, it is anticipated that the ground
settlement should increase with wetting of the soil. This behaviour is clearly seen in the
predictions shown in Figure 5-13 as decreasing the degree of saturation implies higher suctions,

105
which results in a larger hardening parameter in partially saturated soils; and therefore, less
settlement is anticipated.

Saturation Degree

0.3 0.4 0.5 0.6 0.7

-0.45

-0.47

-0.49

-0.51

-0.53

-0.55
Displacement
(m)
-0.57

-0.59

-0.61

-0.63

-0.65

Figure 5-13. Predicted ground settlement under the centre of impact for different degrees of saturation.

Figure 5-14 also shows how the maximum vertical stress generated beneath the centre of
impact, as predicted by the finite element analysis, is increased by increasing the applied
energy.

106
 -90

-110

-130

-150

-170
Vertical Stress
(kPa)

-190

-210

-230

-250
100 200 300 400 500 600 700 800

Energy (kN.m)

Figure 5-14. Maximum vertical stress generated beneath the centre of impact at the end of the analysis versus
applied energy.

A comparison of the generated excess pore water pressure 1 m beneath the centre of impact for
different degrees of saturation is presented in Figure 5-15. From the graph it is seen that

107
increasing the degree of initial saturation will increase the generated excess pore pressure
generated inside the porous medium.

42

38 Sw=0.3

Sw=0.4
34
Sw=0.5

30
Sw=0.6

Sw=0.7
26

Sw=0.8

Excess pore 22
water pressure
(kPa)
18

14

10

-2
0.00 0.03 0.06 0.09 0.12 0.15 0.18
Time (s)

Figure 5-15. The generated excess pore pressure 1 m beneath the centre of impact for different saturation
degrees.

108
As the compaction occurs, suction beneath the impact area tends to decrease due to the positive
generated air pressure. The changes in the values of suction for three different energy levels of
180 kN.m, 245 kN.m and 320 kN.m, at a point 1.8 m beneath the centre of impact are shown
in Figure 5-16. The results indicate that the suction decreases as higher values of energies are
applied.

The decrease of suction in partially saturated soils may increase the permeability of water
phase. As the presence of suction inside the porous skeleton naturally demands more water to
be pushed out of the pores; the suction can act like an obstacle slowing down the movement of
the water phase inside the porous skeleton of the soil. The variations of permeability of water
with respect to the dynamic load for three different levels of energy 180 kN.m, 245 kN.m and
320 kN.m, at a point 1.8 m beneath the impact area are presented in Figure 5-17. The results
show a decrease in permeability of the water phase during the analysis.

-2

180 kN.m
-4

245 kN.m
-6
Excess suction
(kPa) 320 kN.m
-8

-10

-12

-14

-16
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18
Time (s)

Figure 5-16. The changes in suction in 1.8 m below the centre of impact in energy levels of 180, 245 and 320
kN.m.

109
 6.00E-12

5.00E-12 180 kN.m

4.00E-12 245 kN.m

320 kN.m
3.00E-12

The changes in the

permeability of
water (m/s) 2.00E-12

1.00E-12

0.00E+00

-1.00E-12
0.00 0.03 0.06 0.09 0.12 0.15 0.18
Time (s)

Figure 5-17. The changes in the permeability of water in 1.8 m below the centre of impact in energy levels of
180, 245 and 320 kN.m.

5.3.4 Air pressure generated during impact

The variation of air pressure during impact is also of interest. This prediction is provided in
Figure 5-18 in terms of air pressure change normalised with respect to atmospheric pressure
versus the applied energy at a point located 1.0 m below the impact area. The results indicate
that despite avoiding the passive air pressure assumption in the current study, the value of the
air pressure change does not exceed more than 15% of atmospheric pressure.

The effects of the two parameters of the soil-water characteristic curve proposed by Brooks
and Corey (1964), viz., the bubbling pressure, 𝑃𝑑 and the pore size distribution index 𝑣 𝑏 were
also investigated for the case of an applied energy level of 720 kN.m. The predicted results of
changes in air-pressure with 𝑃𝑑 and 𝑣 𝑏 are presented in Figure 5-19 and Figure 5-20,

110
respectively, and these correspond to 17 different analyses. It is shown that the air-pressure
values do not exceed 16% of atmospheric pressure in the studied domain.

0.15

0.14

0.13

0.12

Normalised air
pressure

0.11

0.1

0.09

0.08

0.07
100 175 250 325 400 475 550 625
Energy (kN.m)

Figure 5-18. Normalised air pressure versus applied energy.

111
 0.16

0.15

0.14

Normalised air
pressure

0.13

0.12

0.11
0 30 60 90 120 150 180 210
Bubbling pressure (kPa)

Figure 5-19.Normalised air pressure versus bubbling pressure in the energy level of 720 kN.m.

As argued by researchers such as Sheng et al. (2003), in many geotechnical problems it is

reasonable to assume that the air pressure in the voids of the soil remains at atmospheric
pressure. The numerical results presented here also indicate the possibility of conducting an

112
analysis with the assumption of passive air pressure for this particular problem. This may
simplify future studies of the problem of dynamic compaction, as the number of nodal degrees
of freedom and the problem complexity will diminish by the adoption of this assumption.
Nevertheless, there may be some particular problems that specifically demand the application
of a fully-coupled simulation. One example is the air storage in a porous aquifer presented by
Schrefler and Scotta (2001).

0.16

0.15

0.14

Normalised air
pressure

0.13

0.12

0.11
1 1.5 2 2.5 3 3.5 4

Pore size distribution index

Figure 5-20. Normalised air pressure versus 𝜈 𝑏 for the energy level of 720 kN.m.

113
However, it is emphasised that the global equations derived in Chapter 2 were based on the
assumption of the immiscibility of the pore fluids. The numerical modelling of miscible fluids
in the theory of mixture is an ongoing area of study in theory and applications. Therefore, the
use of a simplified formulation may not be extendable to cases where significant phase transfer
may occur.

5.4 Reaching higher energy levels

The numerical approach adopted for the examples presented above was the Updated-
Lagrangian method, which has proved to be adequate for problems that do not exhibit extreme
mesh distortion within the limited range of applied energy (in this study when the ratio of the
ground settlement to the radius of the falling object is 0.97). However, the severe distortion of
elements that occurs at higher impact energies will make the determinant of the Jacobian matrix
of some finite elements negative, which would then result in immediate termination of the
numerical analysis.

There are two main fundamental methods for describing the governing equations of motion:

 the Lagrangian description, and

 the Eulerian description.

In a Lagrangian and Updated-Lagrangian scheme, such as the method adopted above, the finite
element mesh is ‘glued’ to the material over the entire domain, and as the mesh distorts
following the material movement, the volume changes but the mass remains intact. Thus,
despite the imposed change in density due to volume change, the law of conservation of mass
will automatically be satisfied within each element.

The Lagrangian description of movements has been used in many finite element codes, e.g.,
Zienkiewicz and Taylor (2000), and has often been cited as having some superiorities over the
Eulerian method, as follows:

1. The generated mesh does not need to proceed beyond the problem domain. This can
increase the efficiency of the computational method being used.
2. Particularly in solid mechanics and modelling path-dependent materials, tracking the
material movement becomes much easier as the grid is attached to the material during

114
 the whole process of computation. Thus, internal variables and constitutive variables
can be obtained with sufficient accuracy.
3. Since the tracking of material boundaries is easier, numerical simulation of domains
with complex geometries can be easier than the Eulerian method.
4. Due to automatic preservation of mass inside each element, no convective term is
needed in the global equations. Thus, the code is conceptually simpler and the
implementation becomes easier.

However, the above advantages can be offset by mesh distortion Benson (1992). A possible
solution to improve the performance of Lagrangian codes is to perform rezoning of the mesh
or to re-mesh the problem domain, a process also known as h-adaptivity. These procedures
require generation of a new mesh over the distorted mesh and continuing the analysis with the
undistorted mesh. All internal and external variables will be transformed from the old mesh to
the new mesh with the help of an approximation procedure to preserve the momentum and
energy balance during the analysis.

The h-adaptive technique has been used for modelling problems such as high velocity impact,
penetration, metal forming, etc. e.g., Zienkiewicz and Zhu (1992), Khoei and Lewis (1999),
Perić et al. (1999). However, depending on the problem, the h-adaptive analysis can become
very time-consuming and if frequent re-meshing becomes necessary the efficiency of the
analysis can be decreased. This can also be combined with the occurrence of some material
diffusion resulting in loss of material history data during the re-meshing procedure Liu and Gu
(2005).

On the contrary, in an Eulerian analysis the material movement is separated from the grid
movement by letting the material flow inside each element while the grid remains fixed during
the analysis. Hence, the volume of each element remains unchanged while mass and density
vary depending on the direction and magnitude of the flow. Thus, it becomes necessary that at
every time step, the spatial distribution of materials should also be considered in the analysis
in addition the time rate of change of the material properties. This additional term in the
Eulerian calculations is called the convective derivative, the material derivative or the
substantial derivative in the literature (McDonald (1996) and Anderson and Wendt (1995)) and
𝐷
is indicated by 𝐷𝑡, where

115
𝐷 𝜕 𝜕
= 𝜕𝑡 + 𝑣𝑖 𝜕𝑥 (5.14)
𝐷𝑡 𝑖

𝜕
In the above equation 𝜕𝑡 is known as the local time derivative, which physically describes the
𝜕
time rate of change in the demanded properties at a certain point; whereas 𝑣𝑖 𝜕𝑥 is called the
𝑖

convective term and corresponds to the time rate of change due to the movement of the material
from one location to another in the grid. Thus, the total Eulerian time derivative indicates that
the flow property of the material element is not only affected by the variation of material
property itself but also its movement.

Therefore, the Eulerian method is in fact a spatial description of the movement by nature and
more often is known by its application in the finite difference method (e.g., see Anderson and
Wendt (1995), Wesseling (2009), Pletcher et al. (2012)).

The main advantage of this method is that the separation of movements of grid and material
permits the computational method not to be affected by the extreme movement of the nodes.
However, Eulerian codes usually suffer from the following drawback. The tracking of material
movements becomes difficult as only the time history of material properties at the fixed
Eulerian grid is available during the computation. This can also cause difficulties in finding the
exact position of free surfaces, deformable boundaries, and moving material interfaces during
the analysis. Moreover, complex geometries can also be problematic in the Eulerian codes as
some special treatments (e.g., numerical mapping) are needed to convert the geometry into a
regular domain.

As stated above, Eulerian and Lagrangian methods have their own drawbacks and advantages.
Also, in numerical modelling of many problems, extreme mesh distortion may occur in a small
segment of the domain. Therefore, there has been a steady search for an innovative method
which can take the advantages of both while overcoming their drawbacks. One such approach
is the Arbitrary Lagrangian–Eulerian (ALE) method.

The term “arbitrary” comes from the fact that the grid can have arbitrary movement to keep the
mesh regular. This is the key difference of the ALE method compared with the Lagrangian
approach, where the grid is adhered to the material and the Eulerian approach in which the grid
is fixed in space. Therefore, a “reference configuration” is applied to describe the motion. This
new configuration is neither the fixed or spatial configuration nor the material configuration.

116
Similar to the Eulerian method, due to the separation of the mesh and material movements, a
convective term appears in the global equations.

The ALE technique was initially introduced in fluid mechanics problems by Belytschko and
Kennedy (1978). Hughes et al. (1981) then applied this method in finite element simulation of
viscous incompressible flows and also proposed a kinematical theory for the ALE method.
Kennedy and Belytschko (1982) also developed a coupled model for the analysis of three-
dimensional fluid-structure interaction. The method was also applied to solid mechanics by
researchers such as Haber (1984) where ALE was used in modelling large-deformation
frictional contact and fracture mechanics problems. Later, Liu et al. (1986) proposed the finite
element formulation of ALE together with a stress update technique for path-dependent
materials.

Benson (1989) introduced the operator-split technique into the ALE scheme. In the operator-
split approach, a set of partial differential equations is decoupled into smaller sets and then is
solved sequentially. Although the operator-split approach can reduce the accuracy of the
solution to some extent, it can result in a less expensive and more robust algorithm. Applying
the operator-split method to Equation (5.14) results in a general two-step algorithm. In the first
step of the algorithm the local derivative of the material is obtained by a Lagrangian algorithm,
and subsequently the corresponding substantial derivative will map the Lagrangian mesh to the
reference mesh in an Eulerian step.

The operator-split approach was adopted by Nazem et al. (2008) to develop an ALE code for
coupled analysis of fully saturated soils. The method was implemented into the in-house FEM
code, SNAC, which is the numerical tool employed in this work. In this thesis, a generalisation
of this code has been implemented in SNAC to account for numerical modelling of fully
coupled problems in partially saturated soils.

Given the description above about the operator-split technique, a computation with ALE
method generally consists of the following steps:

(1) Perform a Lagrangian analysis within the Updated Lagrangian framework in every
time step.
(2) Update the coordinates of nodal points based on incremental displacements.
(3) Perform the relocation of the boundary nodes and the interior nodes.

117
 (4) Map the variables from the old mesh to the new mesh.
(5) Check equilibrium and conduct further iterations if necessary.

Steps 1, 2 and 3 are fully covered in Chapter 3, and in Nazem (2006), Nazem et al. (2008) and
Nazem et al. (2009) and will not be repeated here. A discussion of the mapping of global and
local variables into the reference configuration is provided in Appendix B.

A comparison of displacement predicted during the analysis with applied energy level of 845
kN.m of UL and ALE method is shown in Figure 5-21. It is seen that the UL analysis failed at
a displacement of 97 cm. Comparison of these two obtained displacements with the ALE
method and those obtained with the UL analyses is also shown in Figure 5-22.

ALE- 845 kN.m

-0.2

Failed UL analysis - 845 kN.m

-0.4

Diplacement
-0.6
(m)

-0.8

-1

-1.2
0 0.03 0.06 0.09 0.12 0.15 0.18 0.21
Time (s)

Figure 5-21. Comparison of displacement in time during the impact for two levels of energies of 845 and 980
kN.m.

118
 0

-0.2

-0.4

-0.6 UL

Displacement
(m)
ALE
-0.8

-1

-1.2
100 250 400 550 700 850 1000
Energy (kN.m)

Figure 5-22. Displacement versus energy predicted by the UL and ALE approaches.

119
5.5 Summary

Following the validation of the method presented in the previous chapters, in the current
chapter a numerical study of dynamic compaction in partially saturated soils was presented. In
the beginning of the chapter, a viscous boundary for dynamic analysis of partially saturated
media was proposed and was implemented to the finite element code to damp the spurious
reflections from the surrounding boundaries.

Then, the numerical study on dynamic compaction investigated the role of a range of hydro-
mechanical properties involved in the problem. The effect of constitutive model parameters on
void ratio and the ground settlement associated with compaction were shown and the effect of
soil-water characteristic curve parameters on the amount of generated air pressure was also
studied.

The ground settlement variation was observed to have a direct relationship with the slope of
loading line and the applied energy while an inverse relationship was observed between the
slope of critical state line, Poisson’s ratio and pre-consolidation pressure with the ground
settlement.

Moreover, it was shown that desaturation of soils decreases the ground settlement while it has
an inverse relationship with the generated excess pore pressure beneath the centre of impact
area. The changes in void ratio were observed to have a direct relationship with the applied
energy. It was also shown that the excess pore water pressure increases as the saturation
degrees increases in the soil.

In addition, through a series of analyses it was shown that the impact energies and the changes
in suction have inverse relationship as increasing the impact energies reduces the suction in
soil. As a result, the permeability of water was shown to increase beneath the impact area.

As was discussed in the previous chapters, a common simplifying assumption in the analysis
of partially saturated media can be made by considering the air phase passive. This type of
analysis can be beneficial as the number of nodal freedoms, the time of the analysis and the
complexity involved in the problem in some cases can reduce dramatically. Moreover, as
described in Chapter 2, in small deformations analyses, a symmetric Jacobian matrix for
solving the global system of equations can be formed, when material models with associated

120
flow rule are assumed in the problem. This can also reduce the amount of computational storage
in that type of analysis.

To observe the possibility of conducting such an analysis, the generated pore air pressure was
the subject of study at the end of this chapter. The provided numerical study showed that the
excess air pressure generated during the impact and the consequent massive energy release can
be considered to be fairly well below the atmospheric pressure; therefore, the author suggests
considering the possibility of conducting a more simplified analysis with this assumption for
this particular problem with an acceptable accuracy.

121
Chapter 6. Conclusions

122
6.1. Concluding remarks

In this study, a mathematical framework for large deformation dynamic analysis of partially
saturated soils was introduced, implemented and applied to the problem of dynamic
compaction.

In Chapter 2 a review on the application of the theory of mixture was given. The chapter also
discussed the derivation of the global equation in a fully coupled dynamic analysis of
immiscible flows in deforming porous media within the framework of finite element
simulation. A novel time discretization technique of the global equation was presented by the
application of the generalized-𝛼 method. The chapter ended with validation of the
implementation in dynamic elasticity.

In Chapter 3, implementation of a constitutive model to predict the non-linear response of

partially saturated soils was described and discussed. As the problem of dynamic compaction
cannot be properly modelled with infinitesimal strain theory, particularly in the presence of
large deformations, a general explicit integration scheme was proposed with incorporation of
stress rate objectivity. The proposed explicit integration method was then validated by a series
of numerical examples provided in this chapter. At the end of this chapter, a sensitivity analysis
was also performed for the proposed algorithm.

After, preparing the algorithm for performing large deformations analysis of partially saturated
porous media, in the following two chapters numerical modelling of dynamic compaction was
mainly discussed. In Chapter 4, a discussion on the application of dynamic compaction on soil
improvement was made. Then, a series of analytical solutions, experimental studies and
previous finite element simulations were reviewed to identify the shortcomings and drawbacks
of the previous methods in modelling dynamic compactions.

In Chapter 5 a limited numerical study of dynamic compaction was provided. The presented
study highlighted the effects of the determinative hydro-mechanical parameters of partially
saturated soils on the behaviour of those soils under dynamic compaction.

Capturing many features of the response of partially saturated soils during dynamic compaction
in field studies can be very challenging. Part of the reason can be the extreme energy release
imposed by dynamic compaction to the ground beneath the falling weight and the possibility
of damaging the measurement instruments installed in that area. Therefore, the provided

123
numerical simulation can be advantageous. Particularly, in revealing particular aspects of the
behaviour of partially saturated soil which cannot be examined neither in the real conditions
nor by the traditional theoretical, experimental and numerical approaches. A brief summary of
the outcomes is given below.

The ground settlement during the impact has a direct relationship with the slope of the loading
line and the applied energy. An inverse relationship also was seen between the slope of critical
state line, Poisson’s ratio, pre-consolidation pressure and desaturation of soil with the ground
settlement. The role of the determinative parameters involved in dynamic compaction were
further studied particularly on the variation of void ratio after dynamic compaction.

Moreover, the generated excess pore pressure beneath the centre of impact area in different
saturation degrees were also studied. In addition, the relationship between impact energy on
reducing the suction is also shown to be inverse; that is applying higher energies further reduces
the suction beneath the impact area.

Additionally, the procedure of the change in the permeability of the water phase in response to
the external dynamic loads was also presented. It was shown that increasing the applied energy
and the subsequent reduction in suction increases the permeability of the water phase during
the compaction.

It is noted that the presented model can be used as a supplementary means for solving other
dynamic geotechnical engineering problems and can be useful in the preliminary design phases
and feasibility studies regarding dynamic compaction. In addition, the numerical results
provided here can be used as a benchmark in future simulations regarding modelling dynamic
nonlinear response of partially saturated media.

6.2. Recommendation for future work

Based on the findings and results of this dissertation, the following topics are suggested for
further investigation:
 The research described here has focused mainly on the theoretical and numerical
formulation of the problem investigated, i.e., predicting the dynamic response of
unsaturated soils. The proposed method can be applied to a wide range of other

124
 geotechnical problems including dynamic rolling compaction of partially saturated
soils and slope stability problems.
 In this study, partially saturated soils were studied assuming the immiscibility of the
pore fluids. The application of this method in cases where considerable phase transfer
may occur should be done with extra care. However, a rigorous validation of this
assumption demands further experimental and theoretical research.
 In Chapter 5 the study of soil response to dynamic compaction was restricted to cases
involving application of just a single impact. This is due to the fact that the chosen
constitutive model is not able to provide a realistic simulation of the soil response
during multiple impacts, when cycles of loading and unloading are involved in the
problem. Therefore, application of an appropriate constitutive model for this type of
behaviour can be considered in future research.
 The present study was limited to two-dimensional analyses. In this type of analysis, it
is not possible to evaluate the role of the grid spacing of the falling weight on the
application and outcomes of dynamic compaction. Hence, a three-dimensional analysis
based on the proposed algorithm can be another subject of interest for future research.
 The presented framework can be coupled with temperature-related variables to account
for modelling Thermo-Hydro-Mechanical simulations of some geotechnical problems
such as hydration of geosynthetic clay liners. However, additional effort is needed to
implement an appropriate soil water characteristic curve (SWCC) and constitutive
model that consider temperature in their formulations.
 The development of new soil-water characteristic curves considering the effect of the
void ratio, the stress state and the temperature has received broad attention in
unsaturated soil mechanics in recent decades. Hence, these models, if consistently
coupled with the theory of mixture; can be considered in the future study of unsaturated
soil.
 In Chapter 5, a discussion was provided on the consequences of conducting the same
analysis with the assumption of passive air pressure. A comparison of results obtained
by two-phase and three-phase analyses together with verifications with experimental
results can be considered in future research.

125
Appendix A- Modifications of stiffness matrix in Updated Lagrangian method

In the Updated Lagrangian framework the reference and actual configurations are the same.
Therefore, the initial volume of the element, 𝛺, in small deformations should be replace by the
known updated volume of the element, 𝛺 𝑡 in the current time of the analysis 𝑡 in all global
coupled systems of matrices introduced in Chapter 2. .

The generalization of the stiffness matrix in fully saturated soils Nazem (2006) to unsaturated
̅ as follows:
soil can be obtained if stiffness matrix 𝐊 from Equation (2.96) is replaced by 𝐊

̅ = 𝐊 + 𝐊 𝐧𝐥 + 𝐊 𝛔 − 𝐊 𝐩 − 𝐊 𝐩
𝐊 (A.1)
𝒘 𝒄

̅𝑇 𝛔
𝐊 𝐧𝐥 = ∫𝛺𝑡 𝐁 ̅ 𝑑𝛺
̿′ 𝐁 (A.2)

′
𝐊 𝛔 = ∫𝛺𝑡 𝐁 𝑇 𝛔
⏞ 𝐁 𝑑𝛺 (A.3)

̅𝑇 𝐏
𝐊 𝐩𝒘 = ∫𝛺𝑡 𝐁 ̅𝐰 𝐁
̅ 𝑑𝛺 (A.4)

̅𝑇 𝐏
𝐊 𝐩𝒄 = ∫𝛺𝑡 𝐁 ̅𝐜 𝐁
̅ 𝑑𝛺 (A.5)

where

𝜕𝑁𝑢𝑖
( ) 0
𝜕𝑥
𝜕𝑁𝑢𝑖
( ) 0
̅ 𝜕𝑦
𝐁 = (A.6)
𝜕𝑁𝑢𝑖
0 ( )
𝜕𝑥
𝜕𝑁𝑢𝑖
0 ( )
[ 𝜕𝑦 ]

In two-dimensional plane strain analysis for the ith node of an arbitrary element and x and y
are the current configuration of the node. Also,

126
 ′ ′
𝜎11 𝜎12 0 0
𝜎′ ′
𝜎22 0 0
̿′ = 21
𝛔 ′ ′
0 0 𝜎11 𝜎12
[ 0 0 ′
𝜎21 ′
𝜎22 ] (A.7)

′
2𝜎11 0 0
′ ′
⏞ =[ 0
𝛔 2𝜎22 0 ] (A.8)
′ ′ ′ ′
𝜎12 𝜎12 𝜎22 − 𝜎11

𝑝𝑤 0 0 0
̅𝐰 = [ 0
𝐏
𝑝𝑤 0 0
]
0 0 𝑝𝑤 0
0 0 0 𝑝𝑤 (A.9)

and

𝑝𝑐 0 0 0
̅𝐜 = (1 − 𝑆𝑤 ) [ 0
𝐏
𝑝𝑐 0 0
]
0 0 𝑝𝑐 0
0 0 0 𝑝𝑐 (A.10)

127
Appendix B- ALE convection

Generally, the convection term in the ALE method should be used to perform the remapping
on two groups of variables:

 Variables at the integration points (such as the effective stress and hardening
parameters)
 Nodal variables (displacement, pore water pressure and suction).

The equation for mapping both sets of variables is similar. However, the procedure of
implementation differs. According to Hughes et al. (1981), the mapping of any arbitrary
variable, 𝑓, should follow the below rule:

𝜕𝑓
𝑓̇ 𝑟 = 𝑓̇ + (𝑢̇ 𝑖𝑟 − 𝑢̇ 𝑖 ) 𝜕𝑥 (B.1)
𝑖

where 𝑓̇ 𝑟 is the time derivative of 𝑓 with respect to the reference configuration, whereas 𝑓̇ is
the rime derivative of 𝑓 with respect to the material coordinates. Also, 𝑢̇ 𝑖𝑟 is the mesh velocity
and 𝑣𝑖 is the material velocity. Applying Equation (B.1) to the suction and multiplication of
Equation (B.1) by Δ𝑡 results in the following equations:

𝜕𝑝
𝑝𝑐𝑟 = 𝑝𝑐 + (𝑢𝑟 − 𝑢) 𝜕𝑥𝑐 (B.2)
𝑖

where 𝑢𝑟 is the mesh displacement.

128
Appendix C- Description of material law

The elastic constitutive matrix 𝐃𝒆 in a three-dimensional space can be obtained by:

̅ + 4/3𝐺̅
𝐾 ̅ − 2/3𝐺̅
𝐾 ̅ − 2/3𝐺̅
𝐾 0 0 0
̅ − 2/3𝐺̅
𝐾 ̅ + 4/3𝐺̅
𝐾 ̅ − 2/3𝐺̅
𝐾 0 0 0
̅ ̅ ̅ − 2/3𝐺̅ ̅ + 4/3𝐺̅
𝐃𝒆 = 𝐾 − 2/3𝐺 𝐾 𝐾 0 0 0
0 0 0 𝐺̅ 0 0 (C.1)
0 0 0 0 𝐺̅ 0
[ 0 0 0 0 0 𝐺̅ ]

̅ and 𝐺̅ was explained in details in Chapter 3. The mean

where the procedure of finding 𝐾
effective stress is:

1 ′ ′ ′
𝑝′ = (𝜎 + 𝜎22 + 𝜎33 ) (C.2)
3 11

The derivative of mean effective stress with respect to effective stress variables is obtained by

1
1
𝜕𝑝′ 1 1
= (C.3)
𝜕𝛔′ 3 0
0
[0]

where

′
𝜎11
′
𝜎22
𝜎′
𝛔′ = 33
′
𝜎12
′
𝜎13 (C.4)
′
[𝜎23 ]

in a general three-dimensional space.

129
The deviatoric stress tensor, 𝑠𝑖𝑗 , is defined by

𝑠𝑖𝑗 = 𝜎𝑖𝑗 − 𝑝𝛿𝑖𝑗 (C.5)

where 𝛿𝑖𝑗 represents the Kronecker delta defined by

0 𝑖𝑓 𝑖 ≠ 𝑗
𝛿𝑖𝑗 = { (C.6)
1 𝑖𝑓 𝑖 = 𝑗

The second invariant of deviatoric stress tensor, 𝐽2 , is defined by

1 1 2 2
𝐽2 = 𝑠𝑖𝑗 𝑠𝑖𝑗 = [(𝜎11 − 𝜎22 )2 + (𝜎33 − 𝜎22 )2 + (𝜎11 − 𝜎33 )2 ] + 𝜎12 + 𝜎32
2 6 (C.7)
2
+ 𝜎13

and the devatoric stress 𝑞 is related to the second invariant of deviatoric stress tensor by

𝑞 = √3𝐽2 (C.8)

The derivative of the deviatoric stress with respect to stress variables can be obtained by:

𝜕𝑞 𝜕𝑞 𝜕𝐽2
= . (C.9)
𝜕𝛔 𝜕𝐽2 𝜕𝛔

The derivative of the second invariant of deviatoric stress tensor with respect to stress variables
is given by:

𝜕𝐽2
= 𝐐. 𝛔 (C.10)
𝜕𝛔

130
where, matrix 𝐐 has the following form:

2/3 −1/3 −1/3 0 0 0

−1/3 2/3 −1/3 0 0 0
𝐐 = −1/3 −1/3 2/3 0 0 0
0 0 0 2 0 0
0 0 0 0 2 0 (C.11)
[ 0 0 0 0 0 2]

Since the flow rule in the model proposed by Sheng et al. (2003) is assumed to be associative,
the derivatives of yield surface and potential surface with respect to the stress variables are
identical.

Giving the description above, the following relationship can be obtained for the derivative of
yield surface and potential surface with respect to the effective stress variables:

𝜕𝑓 𝜕𝑔 𝜕𝑓 𝜕𝑞 𝜕𝑓 𝜕𝑝′
= = . + . (C.12)
𝜕𝛔′ 𝜕𝛔′ 𝜕𝑞 𝜕𝛔′ 𝜕𝑝′ 𝜕𝛔′

where the derivative of yield surface with respect to the deviatoric stress is obtained by:

𝜕𝑓 2𝑞
= (C.13)
𝜕𝑞 𝛼 ∗

The derivative of the yield surface with respect to the mean effective stress is calculated
according to:

𝜕𝑓 𝑀2 (𝛼 ∗ + 2𝑝′ )
= (C.14)
𝜕𝑝′ 𝛼 ∗2

The derivative of the yield surface with respect to the hardening parameter, 𝛼 ∗ , is determined
by:

131
𝜕𝑓 𝑀2 𝑝′ 𝛼 ∗ + 2(𝑀𝑝′ )2 + 2𝑞 2
= − (C.15)
𝜕𝛼 ∗ 𝛼∗3

The derivative of yield surface with respect to the suction is defined by:

𝜕𝑓 𝜕𝑓 𝜕𝛼 ∗ 𝜕λ
= . . (C.16)
𝜕𝑝𝑐 𝜕𝛼 ∗ 𝜕λ 𝜕𝑝𝑐

where the derivative of hardening parameter with respect to the slope of loading line λ is:

𝜕𝛼 ∗ 𝛼 ∗ (𝜅 − 𝜆0 )log(𝛼)
= (C.17)
𝜕λ (𝜅 − 𝜆)2

The derivative of λ with respect to suction is given by

𝑝𝑐2
−𝛽𝜆0 (1 − 𝑟) exp(−𝛽𝑝𝑐 ) 𝑝𝑐 >
𝛽
𝜕λ 𝛽𝜆0 (𝑝𝑐1 − 𝛽𝑝𝑐 )(1 − 𝑟) 1
𝑝𝑐 𝑝𝑐2
= < 𝑝𝑐 ≤
𝜕𝑝𝑐 √(0.015625 − (𝛽𝑝𝑐 − 𝑝𝑐1 )2 ) 𝛽 𝛽
(C.18)
𝑝𝑐1
0 𝑝𝑐 ≤
{ 𝛽

The derivative of hardening parameter with respect to the volumetric plastic strain is obtained
by:

𝜕𝛼 ∗ 𝜕𝛼 ∗ 𝜕𝛼
= . (C.19)
𝜕𝜀𝑣𝑝 𝜕𝛼 𝜕𝜀𝑣𝑝

where the derivative of the hardening parameter in a fully saturated state with respect to the
plastic volumetric strain is:

132
𝜕𝛼 (1 + 𝑒)𝛼
𝑝 = (C.20)
𝜕𝜀𝑣 𝜆0 − 𝜅

Finally, the derivative of unsaturated hardening parameter with respect to the saturated
hardening parameter (condition of zero suction) is defined by:

𝜕𝛼 ∗ 𝛼 ∗ (𝜆0 − 𝜅)
= . (C.21)
𝜕𝛼 𝛼 (λ − 𝜅)

133
REFERENCES

Adams, B. A. and Wulfsohn, D. (1998). Critical-state behaviour of an agricultural soil. Journal

of agricultural engineering research 70(4): 345-354.

Aitchison, G. (1967). The separate roles of site investigation, quantification of soil properties,
and selection of operational environment in the determination of foundation design on
expansive soils. Israel Soc Soil Mechanics & Fdn Eng Asian Conf Soil Mech & Fdn Haifa.

Albers, B. (2009). Analysis of the propagation of sound waves in partially saturated soils by
means of a macroscopic linear poroelastic model. Transport in porous media 80(1): 173-192.

Alonso, E., Gens, A. and Josa, A. (1990). A constitutive model for partially saturated soils.
Géotechnique 40(3): 405-430.

Alonso, E. E., Gens, A. and Josa, A. (1990). A constitutive model for partially saturated soils.
Géotechnique 40(3): 405-430.

Alonso, E. E., Pereira, J.-M., Vaunat, J. and Olivella, S. (2010). A microstructurally based
effective stress for unsaturated soils. Géotechnique 60(12): 913-925.

Anandarajah, A. (2011). Computational methods in elasticity and plasticity: solids and porous
media, Springer Science & Business Media.

Anderson, J. D. and Wendt, J. (1995). Computational fluid dynamics, Springer.

Barden, L., Madedor, A. and Sides, G. (1969). Volume change characteristics of unsaturated
clay. Journal of Soil Mechanics & Foundations Div.

Bathe, K. J., Ramm, E. and Wilson, E. L. (1975). Finite element formulations for large
deformation dynamic analysis. International Journal for Numerical Methods in Engineering
9(2): 353-386.

Belytschko, T. and Kennedy, J. M. (1978). Computer models for subassembly simulation.

Nuclear engineering and design 49(1): 17-38.

Benson, D. J. (1989). An efficient, accurate, simple ALE method for nonlinear finite element
programs. Computer methods in applied mechanics and engineering 72(3): 305-350.

Benson, D. J. (1992). Computational methods in Lagrangian and Eulerian hydrocodes.

Computer methods in applied mechanics and engineering 99(2): 235-394.

134
Biot, M. A. (1941). General theory of three‐dimensional consolidation. Journal of Applied
Physics 12(2): 155-164.

Biot, M. A. (1956). Theory of propagation of elastic waves in a fluid‐saturated porous solid. I.

Low‐frequency range. The Journal of the Acoustical Society of America 28: 168.

Bishop, A. W. (1960). The principles of effective stress, Norges Geotekniske Institutt.

Bolzon, G., Schrefler, B. and Zienkiewicz, O. (1996). Elastoplastic soil constitutive laws
generalized to partially saturated states. Géotechnique 46(2): 279-289.

Borja, R. I. (2004). Cam-Clay plasticity. Part V: A mathematical framework for three-phase

deformation and strain localization analyses of partially saturated porous media. Computer
methods in applied mechanics and engineering 193(48): 5301-5338.

Bowen, R. (1967). Toward a thermodynamics and mechanics of mixtures. Archive for Rational
Mechanics and Analysis 24(5): 370-403.

Brooks, R. H. and Corey, A. T. (1964). Hydraulic Properties of Porous Media. Hydrology

Papers 3.

Burland, J. B. (1964). Effective stresses in partly saturated soils. Discussion on ‘some aspects
of effective stress in saturated and partly saturated soils’, by G.E. Blight and A.W. Bishop.
Géotechnique 14: 65-68.

Burland, J. B. (1965). Some aspects of the mechanical behaviour of partly saturated soils. . In:
Moisture Equilibria and Moisture Changes in Soils Beneath Covered Areas, a Symposium in
Print. G. D. Aitchison. Butterworths, Sydney, Australia: 270-280.

Chow, Y., Yong, D., Yong, K. and and Lee, S. (1992). Dynamic Compaction Analysis. Journal
of Geotechnical Engineering 118(8): 1141–1157.

Chow, Y. K., Yong, D. M., Yong, K. Y. and Lee, S. L. (1990). Monitoring of dynamic
compaction by deceleration measurements. Computers and Geotechnics 10(3): 189-209.

Chung, J. and Hulbert, G. (1993). A time integration algorithm for structural dynamics with
improved numerical dissipation: the generalized-α method. Journal of applied mechanics
60(2): 371-375.

135
Cohen, M. (1980). Silent boundary methods for transient wave analysis, California Institute of
Technology, Earthquake Engineering Research Laboratory.

Cokca, E., Erol, O. and Armangil, F. (2004). Effects of compaction moisture content on the
shear strength of an unsaturated clay. Geotechnical & Geological Engineering 22(2): 285-297.

Coleman, J. D. (1962). Stress/strain relations for partly saturated soils. Géotechnique 12(4):
348-350.

Collins, I. F. and Yu, H. (1996). Undrained cavity expansions in critical state soils.
International journal for numerical and analytical methods in geomechanics 20(7): 489-516.

Crisfield, M. and Tassoulas, J. L. (1993). Non-Linear Finite Element Analysis of Solids and
Structures, Volume 1. Journal of Engineering Mechanics 119(7): 1504-1505.

Cui, Y., Delage, P. and Sultan, N. (1995). An elasto-plastic model for compacted soils.
Proceedings Of The First International Conference On Unsaturated Soils/Unsat'95/, Paris,
France.

Daniel, D. E. and Benson, C. H. (1990). Water content-density criteria for compacted soil
liners. Journal of Geotechnical Engineering 116(12): 1811-1830.

de Souza Neto, E., Peric, D. and Owen, D. (1994). A model for elastoplastic damage at finite
strains: algorithmic issues and applications. Engineering Computations 11(3): 257-281.

Dise, K., Stevens, M. G. and Von Thun, J. L. (1994). Dynamic Compaction to Remediate
Liquefiable Embankment Foundation Soils. In-situ ground improvement by deep dynamic
compaction ASCE 45: 1-25

Dowell, M. and Jarratt, P. (1972). The “Pegasus” method for computing the root of an equation.
BIT Numerical Mathematics 12(4): 503-508.

Eberhardsteiner, J., Hofstetter, G., Meschke, G. and Mackenzie-Helnwein, P. (2003). Coupled

material modelling and multifield structural analyses in civil engineering. Engineering
Computations 20(5/6): 524-558.

Ehlers, W., Graf, T. and Ammann, M. (2004). Deformation and localization analysis of
partially saturated soil. Computer methods in applied mechanics and engineering 193(27):
2885-2910.

136
Erlicher, S., Bonaventura, L. and Bursi, O. S. (2002). The analysis of the Generalized-α method
for non-linear dynamic problems. Computational mechanics 28(2): 83-104.

Feng, S.-J., Shui, W.-H., Gao, L.-Y., He, L.-J. and Tan, K. (2010). Field studies of the
effectiveness of dynamic compaction in coastal reclamation areas. Bulletin of Engineering
Geology and the Environment 69(1): 129-136.

Feng, S., Shui, W., Tan, K., Gao, L. and He, L. (2011). Field Evaluation of Dynamic
Compaction on Granular Deposits. Journal of Performance of Constructed Facilities 25(3):
241-249.

Fredlund, D. G. and Morgenstern, N. R. (1977). Stress state variables for unsaturated soils.
Journal of Geotechnical and Geoenvironmental Engineering 103(ASCE 12919).

Fredlund, D. G. and Rahardjo, H. (1993). Soil mechanics for unsaturated soils, John Wiley &
Sons.

Fredlund, D. G. and Xing, A. (1994). Equations for the soil-water characteristic curve.
Canadian Geotechnical Journal 31(4): 521-532.

Gadala, M. and Wang, J. (2000). Computational implementation of stress integration in FE

analysis of elasto-plastic large deformation problems. Finite Elements in Analysis and Design
35(4): 379-396.

Gallipoli, D., Gens, A., Sharma, R. and Vaunat, J. (2003). An elasto-plastic model for
unsaturated soil incorporating the effects of suction and degree of saturation on mechanical
behaviour. Géotechnique. 53(1): 123-136.

Gardner, W. (1958). Mathematics of isothermal water conduction in unsaturated soil. Highway

Research Board Special Report(40).

Gens, A., Garcia‐Molina, A., Olivella, S., Alonso, E. and Huertas, F. (1998). Analysis of a full
scale in situ test simulating repository conditions. International journal for numerical and
analytical methods in geomechanics 22(7): 515-548.

Gens, A., Sánchez, M. and Sheng, D. (2006). On constitutive modelling of unsaturated soils.
Acta Geotechnica 1(3): 137-147.

Ghassemi, A., Pak, A. and Shahir, H. (2008). Validity of Menard relation in dynamic
compaction operations. Proceedings of the Institution of Civil Engineers
Ground Improvement(GI1): 1–9.

137
Ghassemi, A., Pak, A. and Shahir, H. (2010). Numerical study of the coupled hydro-mechanical
effects in dynamic compactionof saturated granular soils. Computers and Geotechnics 37: 10-
24.

Ghorbani, J., Nazem, M. and Carter, J. P. (2015). Application of the generalised-α method in
dynamic analysis of partially saturated media. Computer Methods and Recent Advances in
Geomechanics - Proceedings of the 14th Int. Conference of International Association for
Computer Methods and Recent Advances in Geomechanics, IACMAG 2014.

Ghorbani, J., Noorzad, A. and Shahnazari, H. (2014). A smart increment technique and its
application to a bounding surface model. Numerical Methods in Geotechnical Engineering,
CRC Press: 45-47.

Gray, W. G. and Hassanizadeh, S. M. (1991). Unsaturated flow theory including interfacial

phenomena. Water Resources Research 27(8): 1855-1863.

Green, A. E. and Naghdi, P. M. (1967). A theory of mixtures. Archive for Rational Mechanics
and Analysis 24(4): 243-263.

Gunaratne, M., Ranganath, M., Thilakasiri, S., Mullins, G., Stinnette, P. and Kuo, C. (1996).
Study of pore pressures induced in laboratory dynamic consolidation. Computers and
Geotechnics 18(2): 127-143.

Haber, R. B. (1984). A mixed Eulerian-Lagrangian displacement model for large-deformation

analysis in solid mechanics. Computer methods in applied mechanics and engineering 43(3):
277-292.
Hansbo, S. (1996). Dynamic consolidation of mixed fill: a cost-effective alternative to piling:
a case record. Géotechnique 46(2): 351-355.

Hashash, Y. M. and Whittle, A. J. (1992). Analysis of braced diaphragm walls in deep deposits
of clay, Department of Civil Engineering and Environmental Engineering, Massachusetts
Institute of Technology.

Hassanizadeh, M. and Gray, W. G. (1980). General conservation equations for multi-phase

systems: 3. Constitutive theory for porous media flow. Advances in Water Resources 3(1): 25-
40.

Heider, Y., Markert, B. and Ehlers, W. (2011). Dynamic wave propagation in infinite saturated
porous media half spaces. Computational mechanics 49(3): 319-336.

138
Hilber, H. M., Hughes, T. J. and Taylor, R. L. (1977). Improved numerical dissipation for time
integration algorithms in structural dynamics. Earthquake Engineering & Structural Dynamics
5(3): 283-292.

Hilf, J. W. (1956). An investigation of pore water pressure in compacted cohesive soils.

Houlsby, G. (1997). The work input to an unsaturated granular material. Géotechnique 47(1):
193-196.

Howard, R. F., Singer, M. J. and Frantz, G. A. (1981). Effects of soil properties, water content,
and compactive effort on the compaction of selected California forest and range soils. Soil
Science Society of America Journal 45(2): 231-236.

Hu, R., Yue, Z., Tham, L. and Wang, L. (2005). Digital image analysis of dynamic compaction
effects on clay fills. Journal of geotechnical and geoenvironmental engineering 131(11): 1411-
1422.

Hughes, T. J. (1984). Numerical implementation of constitutive models: rate-independent

deviatoric plasticity. Theoretical foundation for large-scale computations for nonlinear material
behavior, Springer: 29-63.

Hughes, T. J., Liu, W. K. and Zimmermann, T. K. (1981). Lagrangian-Eulerian finite element

formulation for incompressible viscous flows. Computer methods in applied mechanics and
engineering 29(3): 329-349.

Hughes, T. J. R. and Winget, J. (1980). Finite rotation effects in numerical integration of rate
constitutive equations arising in large-deformation analysis. International Journal for
Numerical Methods in Engineering 15(12): 1862-1867.

Hutter, K., Laloui, L. and Vulliet, L. (1999). Thermodynamically based mixture models of
saturated and unsaturated soils. Mechanics of Cohesive‐frictional Materials 4(4): 295-338.

Jennings, J. (1961). A revised effective stress law for use in the prediction of the behaviour of
unsaturated soils. Pore pressure and suction in soils: 26-30.

Jennings, J. and Burland, J. (1962). Limitations to the use of effective stresses in partly
saturated soils. Géotechnique 12(2): 125-144.

Jia, M., Zhao, Y. and Zhou, X. (2015). Field Studies of Dynamic Compaction on Marine
Deposits. Marine Georesources & Geotechnology: 1-8.

139
Jommi, C. (2000). Remarks on the constitutive modelling of unsaturated soils. Experimental
evidence and theoretical approaches in unsaturated soils: 139-153.

Josa Garcia-Tornel, A., Balmaceda, A., Gens Solé, A. and Alonso Pérez de Agreda, E. (1992).
An Elasto-plastic model for partially saturated soils exhibiting a maximum of collapse. 3rd
InternationalConference on Computational Plasticity, Barcelona.

Karube, D. and Kawai, K. (2001). The role of pore water in the mechanical behavior of
unsaturated soils. Geotechnical & Geological Engineering 19(3-4): 211-241.

Kennedy, J. and Belytschko, T. (1982). Theory and application of a finite element method for
arbitrary Lagrangian-Eulerian fluids and structures. Nuclear engineering and design 68(2):
129-146.

Kérisel, J. (1985). The history of geotechnical engineering up until 1700. 11th International
Conference on Soil Mechanics and Foundation Engineering, San Francisco.

Khalili, N., Geiser, F. and Blight, G. (2004). Effective stress in unsaturated soils: review with
new evidence. International Journal of Geomechanics 4(2): 115-126.

Khalili, N., Habte, M. and Zargarbashi, S. (2008). A fully coupled flow deformation model for
cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses. Computers
and Geotechnics 35(6): 872-889.

Khalili, N. and Khabbaz, M. (1998). A unique relationship of chi for the determination of the
shear strength of unsaturated soils. Géotechnique 48(5).

Khoei, A. and Mohammadnejad, T. (2011). Numerical modeling of multiphase fluid flow in

deforming porous media: a comparison between two-and three-phase models for seismic
analysis of earth and rockfill dams. Computers and Geotechnics 38(2): 142-166.

Khoei, A. R. and Lewis, R. W. (1999). Adaptive finite element remeshing in a large

deformation analysis of metal powder forming. International Journal for Numerical Methods
in Engineering 45(7): 801-820.

Khoei, A. R. and Mohammadnejad, T. (2011). Numerical modeling of multiphase fluid flow

in deforming porous media: A comparison between two- and three-phase models for seismic
analysis of earth and rockfill dams. Computers and Geotechnics 38(2): 142-166.

Kontoe, S., Zdravkovic, L. and Potts, D. M. (2008). An assessment of time integration schemes
for dynamic geotechnical problems. Computers and Geotechnics 35(2): 253-264.

140
Kumar, S. and Puri, V. K. (2001). Soil improvement using heavy tamping—a case history.
ISET J. Earthq. Tech 38(2-4): 123-133.

Lambe, T. W. (1960). A mechanistic picture of shear strength in clay. Research Conference on

shear strength of cohesive soils, ASCE.

Lee, F. and Gu, Q. (2004). Method for Estimating Dynamic Compaction Effect on Sand.
Journal of Geotechnical and Geoenvironmental Engineering 130(2): 139-152.

Lee, P. Y. and Suedkamp, R. (1972). Characteristics of irregularly shaped compaction curves

of soils.

Leonards, A. G., Holtz, D. R. and Cutter, A. W. (1980). Dynamic Compaction of Granular

Soils. Journal of the Geotechnical Engineering Division 106(1): 35-44.

Lewis, R. and Schrefler, B. (1982). A finite element simulation of the subsidence of gas
reservoirs undergoing a water drive. Finite element in fluids 4: 179-199.

Lewis, R. and Schrefler, B. (1999). The Finite Element Method in the Static and Dynamic
Deformation and Consolidation of Porous Media. Meccanica 34(3): 231-232.

Lewis, R. W. and Schrefler, B. A. (1998). The finite element method in the static and dynamic
deformation and consolidation of porous media, Wiley.

Li, P. and Schanz, M. (2011). Wave propagation in a 1-D partially saturated poroelastic
column. Geophysical Journal International 184(3): 1341-1353.

Li, X., Ding, D., Chan, A. and Zienkiewicz, O. (1989). A coupled finite element method for
the soil-pore fluid interaction problems with immiscible two-phase fluid flow. Proc. 5th Int.
Symp. on Numerical Methods in Eng., Lausanne.

Li, X., Thomas, H. and Fan, Y. (1999). Finite element method and constitutive modelling and
computation for unsaturated soils. Computer Methods in Applied Mechanics and Engineering
169(1): 135-159.

Li, X. and Zienkiewicz, O. (1992). Multiphase flow in deforming porous media and finite
element solutions. Computers & structures 45(2): 211-227.

141
Li, X., Zienkiewicz, O. and Xie, Y. (1990). A numerical model for immiscible two‐phase fluid
flow in a porous medium and its time domain solution. International Journal for Numerical
Methods in Engineering 30(6): 1195-1212.

Liakopoulos, A. (1964). transient flow through unsaturated porous media. D. Eng. Berkeley,
University of California at Berkeley. PhD Thesis.

Liu, G.-R. and Gu, Y. (2005). An introduction to meshfree methods and their programming.
Berlin, Springer.

Liu, W. K., Belytschko, T. and Chang, H. (1986). An arbitrary Lagrangian-Eulerian finite

element method for path-dependent materials. Computer methods in applied mechanics and
engineering 58(2): 227-245.

Lo, K., Ooi, P. and Lee, S. (1990). Unified Approach to Ground Improvement by Heavy
Tamping. Journal of Geotechnical Engineering 116(3): 514-527.

López‐Querol, S., Fernández‐Merodo, J., Mira, P. and Pastor, M. (2008). Numerical modelling
of dynamic consolidation on granular soils. International journal for numerical and analytical
methods in geomechanics 32(12): 1431-1457.

Loret, B. and Khalili, N. (2002). An effective stress elastic–plastic model for unsaturated
porous media. Mechanics of Materials 34(2): 97-116.

Lukas, R. G. (1980). Densification of loose deposits by pounding. Journal of the Geotechnical

Engineering Division 106(4): 435-446.

Lukas, R. G. (1986). Dynamic compaction for highway construction. Design and construction
guidelines. Washington, DC, U.S. Federal Highway Administration. 1: 204–219.

Lysmer, J. and Kuhlemeyer, R. L. (1969). Finite dynamic model for infinite media. Journal of
the Engineering Mechanics Division 95(4): 859-878.

Malvern, L. E. (1969). Introduction to the Mechanics of a Continuous Medium.

Marsili, A., Servadio, P., Pagliai, M. and Vignozzi, N. (1998). Changes of some physical
properties of a clay soil following passage of rubber- and metal-tracked tractors. Soil and
Tillage Research 49(3): 185-199.

Mayne, P., Jones, J., Jr., and Dumas, J. (1984). Ground Response to Dynamic Compaction. J.
Geotech. Engrg. 110(6): 757–774.

142
Mayne, P. W. and Jones , J. S. (1983). Impact stresses during dynamic compaction. Journal of
Geotechnical Engineering 109(10): 1342-1346.

McDonald, P. H. (1996). Continuum Mechanics. Boston, MA, PWS Publishing Co.

Menard, L. and Broise, Y. (1975). Theoretical and practical aspects of dynamic consolidation.
Ground Treatment by Deep Compaction, London, Institution of Civil Engineers.

Ménard, L. and Broise, Y. (1975) Theoretical and practical aspect of dynamic consolidation.
Géotechnique 25, 3-18

Meroi, E., Schrefler, B. and Zienkiewicz, O. (1995). Large strain static and dynamic
semisaturated soil behaviour. International Journal for Numerical and Analytical Methods in
Geomechanics 19(2): 81-106.

Merrifield, C. M. and Davies, M. C. R. (2000) A study of low-energy dynamic compaction:

field trials and centrifuge modelling. Géotechnique 50, 675-681

Miao, L., Chen, G. and Hong, Z. (2006). Application of Dynamic Compaction in Highway: A
Case Study. Geotechnical & Geological Engineering 24(1): 91-99.

Mitchell, J. K., Hooper, D. R. and Campenella, R. G. (1965). Permeability of compacted clay.

Journal of the Soil Mechanics and Foundations Division 91(4): 41-66.

Morel-Seytoux, H. J. and Billica, J. A. (1985). A Two-Phase Numerical Model for Prediction

of Infiltration: Applications to a Semi-Infinite Soil Column. Water Resources Research 21(4):
607-615.

Morel‐Seytoux, H. J. and Billica, J. A. (1985). A two‐phase numerical model for prediction of

infiltration: Applications to a semi‐infinite soil column. Water Resources Research 21(4): 607-
615.

Moseley, M. P. and Slocombe, B. C. (1978). In-Situ Treatment of Clay Fills. Conference on

Clay Fills. London, Institution of Civil Engineers: 165-169.

Mostafa, K. (2010). Numerical modeling of dynamic compaction in cohesive soils, The

University of Akron, PhD thesis.

143
Narasimhan, T. N. and Witherspoon, P. A. (1978). Numerical model for saturated-unsaturated
flow in deformable porous media: 3. Applications. Water Resources Research 14(6): 1017-
1034.

Nashed, R., Thevanayagam, S. and Martin, G. (2006). Simulation of Dynamic Compaction

Processes in Saturated Silty Soils. GeoCongress 2006@ sGeotechnical Engineering in the
Information Technology Age, ASCE.

Nashed, R., Thevanayagam, S., Martin, G. and Shenthan, T. (2004). Liquefaction mitigation in
silty soils using dynamic compaction and wick drains. Proc. 13th World Conf. on Earthq. Eng.

Nayak, G. and Zienkiewicz, O. (1972). Elasto‐plastic stress analysis. A generalization for

various contitutive relations including strain softening. International Journal for Numerical
Methods in Engineering 5(1): 113-135.

Nazem, M. (2006). Numerical algorithms for large deformation problems in geomechanics,

University of Newcastle.

Nazem, M., Carter, J. and Airey, D. (2009). Arbitrary Lagrangian–Eulerian method for
dynamic analysis of geotechnical problems. Computers and Geotechnics 36(4): 549-557.

Nazem, M., Carter, J. P., Sheng, D. and Sloan, S. W. (2009). Alternative stress-integration
schemes for large-deformation problems of solid mechanics. Finite Elements in Analysis and
Design 45(12): 934-943.

Nazem, M., Sheng, D., Carter, J. P. and Sloan, S. W. (2008). Arbitrary Lagrangian–Eulerian
method for large-strain consolidation problems. International journal for numerical and
analytical methods in geomechanics 32(9): 1023-1050.

Newmark, N., M. (1959). A Method of Computation for Structural Dynamics. Journal of the
Engineering Mechanics Division 85(3): 67-94.

Nuth, M. and Laloui, L. (2008). Effective stress concept in unsaturated soils: clarification and
validation of a unified framework. International journal for numerical and analytical methods
in geomechanics 32(7): 771-801.

Oka, F., Kodaka, T., Kimoto, S., Kim, Y. S. and Yamasaki, N. (2006). An Elasto‐Viscoplastic
Model and Multiphase Coupled FE Analysis for Unsaturated Soil, ASCE.

Oshima, A. and Takada, N. (1998). Evaluation of compacted area of heavy tamping by cone
point resistance. Proc., Int. Conf. Centrifuge.

144
Pan, J. L. and Selby, A. R. (2002). Simulation of dynamic compaction of loose granular soils.
Advances in Engineering Software 33(7–10): 631-640.

Pastor, M., Zienkiewicz, O. and Chan, A. (1990). Generalized plasticity and the modelling of
soil behaviour. International journal for numerical and analytical methods in geomechanics
14(3): 151-190.

Pedroso, D. M. and Farias, M. M. (2011). Extended Barcelona Basic Model for unsaturated
soils under cyclic loadings. Computers and Geotechnics 38(5): 731-740.

Pérez-Foguet, A., Rodrı́guez-Ferran, A. and Huerta, A. (2001). Consistent tangent matrices for
substepping schemes. Computer methods in applied mechanics and engineering 190(35):
4627-4647.

Perić, D., Vaz, M. and Owen, D. (1999). On adaptive strategies for large deformations of
elasto-plastic solids at finite strains: computational issues and industrial applications.
Computer methods in applied mechanics and engineering 176(1): 279-312.

Perucho, A. and Olalla, C. (2006) Dynamic consolidation of a saturated plastic clayey fill.
Proceedings of the ICE - Ground Improvement 10, 55-68

Perucho, A. and Olalla, C. (2006). Dynamic consolidation of a saturated plastic clayey fill.
Proceedings of the Institution of Civil Engineers-Ground Improvement 10(2): 55-68.

Pinder, G. F. and Gray, W. G. (2008). Essentials of multiphase flow in porous media, John
Wiley & Sons.

Pinsky, P. M., Ortiz, M. and Pister, K. S. (1983). Numerical integration of rate constitutive
equations in finite deformation analysis. Computer Methods in Applied Mechanics and
Engineering 40(2): 137-158.

Pletcher, R. H., Tannehill, J. C. and Anderson, D. (2012). Computational fluid mechanics and
heat transfer, CRC Press.

Poran, J. and Rodriguez, A. (1992). Finite element analysis of impact behavior of sand. Soils
and Foundations 32(4): 68-80.

Potts, D. M. and Gens, A. (1985). A critical assessment of methods of correcting for drift from
the yield surface in elasto-plastic finite element analysis. International journal for numerical
and analytical methods in geomechanics 9(2): 149-159.

145
Potts, D. M. and Zdravković, L. (1999). Finite Element Analysis in Geotechnical Engineering:
Theory, Thomas Telford Ltd.

Proctor, R. (1933). Fundamental principles of soil compaction. Engineering News Record

111(9): 245-248.

Rahardjo, H., Heng, O. B. and Choon, L. E. (2004). Shear strength of a compacted residual soil
from consolidated drained and constant water content triaxial tests. Canadian Geotechnical
Journal 41(3): 421-436.

Rahman, N. A. and Lewis, R. W. (1999). Finite element modelling of multiphase immiscible

flow in deforming porous media for subsurface systems. Computers and Geotechnics 24(1):
41-63.

Rodríguez-Ferran, A. and Huerta, A. (1998). Comparing two algorithms to add large strains to
small-strain FE code. Journal of Engineering Mechanics 124(9): 939-948.

Rollins, K., Jorgensen, S. and Ross, T. (1998). Optimum Moisture Content for Dynamic
Compaction of Collapsible Soils. Journal of Geotechnical and Geoenvironmental Engineering
124(8): 699-708.

Rollins, K. and Rogers, G. (1994). Mitigation Measures for Small Structures on Collapsible
Alluvial Soils. Journal of Geotechnical Engineering 120(9): 1533-1553.

Roscoe, K. H. and Burland, J. (1968). On the generalized stress-strain behaviour of wet clay.

Ruiz, T., Delalonde, M., Bataille, B., Baylac, G. and de Crescenzo, C. D. (2005). Texturing
unsaturated granular media submitted to compaction and kneading processes. Powder
Technology 154(1): 43-53.

Russell, A. R. and Khalili, N. (2004). A bounding surface plasticity model for sands exhibiting
particle crushing. Canadian Geotechnical Journal 41(6): 1179-1192.

Santagiuliana, R. and Schrefler, B. (2006). Enhancing the Bolzon–Schrefler–Zienkiewicz

constitutive model for partially saturated soil. Transport in porous media 65(1): 1-30.

Satyapriya, C. and Gallagher, P. E. (2000). Dynamic compaction of surface mine spoils to limit
settlements within Commercial developments. ASTM special technical publication 1384: 163-
172.

146
Schrefler, B. and Xiaoyong, Z. (1993). A fully coupled model for water flow and airflow in
deformable porous media. Water Resources Research 29(1): 155-167.

Schrefler, B. A. and Scotta, R. (2001). A fully coupled dynamic model for two-phase fluid flow
in deformable porous media. Computer methods in applied mechanics and engineering
190(24–25): 3223-3246.

Schrefler, B. A., Zhan, X. and Simoni, L. (1995). A coupled model for water flow, airflow and
heat flow in deformable porous media. International Journal of Numerical Methods for Heat
& Fluid Flow 5(6): 531-547.

Scott, R. and Pearce, R. (1975). Soil compaction by impact. Géotechnique 25(1): 19-30.

Shahbodagh, B., Khalili, N. and Esgandani, G. (2015). A numerical model for nonlinear large
deformation dynamic analysis of unsaturated porous media including hydraulic hysteresis.
Computers and Geotechnics 69: 411-423.

Sheng, D. (2011). Constitutive modelling of unsaturated soils: Discussion of fundamental

principles. Unsaturated soils 1: 91-112.

Sheng, D. (2011). Review of fundamental principles in modelling unsaturated soil behaviour.

Computers and Geotechnics 38(6): 757-776.

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.

Sheng, D., Gens, A., Fredlund, D. G. and Sloan, S. W. (2008). Unsaturated soils: from
constitutive modelling to numerical algorithms. Computers and Geotechnics 35(6): 810-824.

Sheng, D., Sloan, S. and Yu, H. (2000). Aspects of finite element implementation of critical
state models. Computational mechanics 26(2): 185-196.

Sheng, D., Sloan, S. W., Gens, A. and Smith, D. W. (2003). Finite element formulation and
algorithms for unsaturated soils. Part I: Theory. International Journal for Numerical and
Analytical Methods in Geomechanics 27(9): 745-765.

Sheng, D., Smith, D. W., W Sloan, S. and Gens, A. (2003). Finite element formulation and
algorithms for unsaturated soils. Part II: Verification and application. International Journal for
Numerical and Analytical Methods in Geomechanics 27(9): 767-790.

147
Sheng, D., Zhang, S. and Yu, Z. (2013). Unanswered questions in unsaturated soil mechanics.
Science China Technological Sciences: 1-16.

Shui, W. H., Wang, T. H. and Wang, Y. L. (2006). SPT for dynamic compaction with 10000
kN·m high energy on foundation backfilled with crushed stone. Chinese Journal of
Geotechnical Engineering 28(10): 1309-1312.

Simo, J. C. and Taylor, R. L. (1985). Consistent tangent operators for rate-independent

elastoplasticity. Computer methods in applied mechanics and engineering 48(1): 101-118.

Sloan, S. (1987). Substepping schemes for the numerical integration of elastoplastic stress–
strain relations. International Journal for Numerical Methods in Engineering 24(5): 893-911.

Sloan, S. and Randolph, M. F. (1982). Numerical prediction of collapse loads using finite
element methods. International journal for numerical and analytical methods in geomechanics
6(1): 47-76.

Sloan, S. W. (1987). Substepping schemes for the numerical integration of elastoplastic stress–
strain relations. International Journal for Numerical Methods in Engineering 24(5): 893-911.

Sloan, S. W., Abbo, A. J. and Sheng, D. (2001). Refined explicit integration of elastoplastic
models with automatic error control. Engineering Computations 18(1/2): 121-194.

Slocombe, B. C. (1993). Ground improvement. M. P. Moseley, CRC, Boca Raton, Fla.: 21-39.

Sołowski, W. T. and Gallipoli, D. (2010). Explicit stress integration with error control for the
Barcelona Basic Model: Part I: Algorithms formulations. Computers and Geotechnics 37(1):
59-67.

Tamagnini, R. (2004) An extended Cam-clay model for unsaturated soils with hydraulic
hysteresis. Géotechnique 54, 223-228

Tamagnini, R. and Gennaro, V. D. (2008). Implicit integration of an extended Cam-clay model

for unsaturated soils. Unsaturated Soils. Advances in Geo-Engineering, Taylor & Francis: 713-
719.

Tamagnini, R. and Pastor, M. (2004). A thermodynamically based model for unsaturated soil:
a new framework for generalized plasticity. Unsaturated Soils. Advances in Testing, Modelling
and Engineering Applications: 121-134.

Terzaghi, K. (1936). The shear resistance of saturated soils.

148
1st International Conference on Soil Mechanics and Foundation Engineering Cambridge, MA.

Thilakasiri, H., Gunaratne, M., Mullins, G., Stinnette, P. and Jory, B. (1996). Investigation of
impact stresses induced in laboratory dynamic compaction of soft soils. International journal
for numerical and analytical methods in geomechanics 20(10): 753-767.

Thomas, H. and He, Y. (1995). Analysis of coupled heat, moisture and air transfer in a
deformable unsaturated soil. Géotechnique 45(4): 677-689.

Truesdell (1957). Sulle Basi Della Thermomeccanica. C. Rend. Lincei 22: 33-38.

Van Genuchten, M. T. (1980). A closed-form equation for predicting the hydraulic

conductivity of unsaturated soils. Soil Science Society of America Journal 44(5): 892-898.

Van Impe, W. F. and Bouazza, A. (1996). Densification of domestic waste fills by dynamic
compaction. Canadian Geotechnical Journal 33(6): 879-887.

Vlahinić, I., Jennings, H. M., Andrade, J. E. and Thomas, J. J. (2011). A novel and general
form of effective stress in a partially saturated porous material: the influence of microstructure.
Mechanics of Materials 43(1): 25-35.

Wang, R., Zhu, C., Wang, J. and Yang, M. (2000). Improvement of Soft Clays by Dynamic
Compaction with PVDs. Advances in Grouting and Ground Modification: 311-324.

Wang, S., Chan, D. and Lam, C. (2009). Experimental study of the effect of fines content on
dynamic compaction grouting in completely decomposed granite of Hong Kong. Construction
and Building Materials 23: 1249–1264.

Wesseling, P. (2009). Principles of computational fluid dynamics, Springer Science &

Business Media.

Wheeler, S., Sharma, R. and Buisson, M. (2003). Coupling of hydraulic hysteresis and stress–
strain behaviour in unsaturated soils. Géotechnique 53(1): 41-54.

Wheeler, S. J. and Sivakumar, V. (1995). An elasto-plastic critical state framework for

unsaturated soil. Géotechnique 45(1): 35-53.

Wissmann, J. W. and Hauck, C. (1983). Efficient elastic-plastic finite element analysis with
higher order stress-point algorithms. Computers & structures 17(1): 89-95.

149
Wood, W. L., Bossak, M. and Zienkiewicz, O. C. (1980). An alpha modification of Newmark's
method. International Journal for Numerical Methods in Engineering 15(10): 1562-1566.

Wriggers, P. and Laursen, T. A. (2006). Computational contact mechanics, Springer.

Wu, Y.-S. and Forsyth, P. A. (2001). On the selection of primary variables in numerical
formulation for modeling multiphase flow in porous media. Journal of contaminant hydrology
48(3): 277-304.

Zekkos, D., Kabalan, M. and Flanagan, M. (2013). Lessons learned from case histories of
dynamic compaction at municipal solid waste sites. Journal of Geotechnical and
Geoenvironmental Engineering 139(5): 738-751.

Zhao, J., Sheng, D., Rouainia, M. and Sloan, S. W. (2005). Explicit stress integration of
complex soil models. International journal for numerical and analytical methods in
geomechanics 29(12): 1209-1229.

Zienkiewicz, O. C., Chan, A., Pastor, M., Schrefler, B. and Shiomi, T. (1999). Computational
geomechanics, Wiley Chichester.

Zienkiewicz, O. C. and Taylor, R. L. (2000). The finite element method: Solid mechanics,
Butterworth-heinemann.

Zienkiewicz, O. C. and Zhu, J. Z. (1992). The superconvergent patch recovery and a posteriori
error estimates. Part 2: Error estimates and adaptivity. International Journal for Numerical
Methods in Engineering 33(7): 1365-1382.

150
‘

151