Evaluation of the PM4SAND Constitutive Model for the Prediction of

Earthquake-Induced & Static Liquefaction in Hydraulic Fills

Master of Science Thesis

By
Aristotelis Tziolas
Student Number: 4631722
Delft, July 9, 2019

Delft University of Technology

Faculty of Civil Engineering and Geosciences

Section Geo-Engineering

|
 Graduation committee:

Prof. dr. M. A. Hicks Dr. Ir. C. Kasbergen

Delft University of Technology Delft University of Technology

Dr. Ir. R.B.J. Brinkgreve Dr. Ir. R. de Jager

Delft University of Technology Boskalis B.V.

|
Contents

Acknowledgements ............................................................................................................................. 1
Abstract ............................................................................................................................................... 1
Chapter 1: Introduction ....................................................................................................................... 1
1.1 Problem Definition .................................................................................................................... 1
1.2 Research Questions ................................................................................................................... 3
Chapter 2: Literature study ................................................................................................................. 4
2.1 Liquefaction Hazard ................................................................................................................... 4
2.1.1 Introduction ........................................................................................................................ 4
2.1.2 Types of Liquefaction ......................................................................................................... 4
2.2 Principles of Liquefaction .......................................................................................................... 5
2.2.1 Critical Void Ratio ............................................................................................................... 5
2.2.2 State Parameter.................................................................................................................. 6
2.2.3 Monotonic Loading............................................................................................................. 6
2.2.4 Cyclic Loading .................................................................................................................... 7
2.3 Simplified Procedure for the Simulation of Earthquake Loading .............................................. 9
2.3.1 Introduction ........................................................................................................................ 9
2.3.2 Cyclic Stress Ratio (CSR).................................................................................................... 10
2.3.3 Cyclic Resistance Ratio (CRR) ............................................................................................ 10
2.3.4 Correction Factors ............................................................................................................ 11
2.3.5 Simplified Liquefaction Triggering Assessment ................................................................ 12
2.4 Correlation between CRR from Lab Tests to Field Conditions ................................................ 13
2.5 Hydraulic Fill Structure ............................................................................................................ 13
2.5.1 Introduction ...................................................................................................................... 13
2.5.2 Hydraulic Filling Method .................................................................................................. 14
2.5.3 In-situ Conditions of the Reclamation Structure / In-situ Volumetric State (Dr) ............. 14
2.5.4 Placement Methods ......................................................................................................... 15
2.6 A Sand Plasticity Model Accounting for Fabric Change Effects ............................................... 16
2.6.1 Model Description ............................................................................................................ 16
2.6.2 Conceptual Desription ...................................................................................................... 17
2.7 PM4Sand Constitutive Model.................................................................................................. 21
2.7.1 Model Description ............................................................................................................ 21
2.7.2 Critical State Soil Mechanics Framework ......................................................................... 21

|
 2.7.3 Fabric Effects .................................................................................................................... 23
2.7.4 Model Parameters ............................................................................................................ 23
Chapter 3: Validation of the PM4Sand Model for Cyclic Loading ..................................................... 25
3.1 Introduction ............................................................................................................................. 25
3.2 Data Interpretation ................................................................................................................. 25
3.3 Input Model Parameters ......................................................................................................... 27
3.4 Influence of the Primary Model Parameters ........................................................................... 32
3.4.1 Influence of DR .................................................................................................................. 32
3.4.2 Influence of Go .................................................................................................................. 34
3.5 Influence of the Secondary Model Parameters....................................................................... 37
3.5.1 Influence of nb .................................................................................................................. 37
3.5.2 Influence of R.................................................................................................................... 40
3.6 Evaluation of the Final Set of Parameters ............................................................................... 43
3.7 Conclusions .............................................................................................................................. 45
Chapter 4: Verification of the PM4Sand Model for Static Loading ................................................... 46
4.1 Introduction ............................................................................................................................. 46
4.2 Data Interpretation ................................................................................................................. 46
4.3 Input Model Parameters ......................................................................................................... 48
4.5 Influence of the Primary Model Parameters ........................................................................... 50
4.5.1 Influence of DR .................................................................................................................. 50
4.5.2 Influence of Go .................................................................................................................. 52
4.6 Influence of the Secondary Model Parameters....................................................................... 53
4.6.1 Influence of nb .................................................................................................................. 53
4.6.2 Influence of R.................................................................................................................... 54
4.6.2 Influence of Q ................................................................................................................... 56
4.6.3 Influence of φ’cv ................................................................................................................ 58
4.7 Conclusions .............................................................................................................................. 60
Chapter 5: Simulation of Earthquake Induced Liquefaction in Hydraulic Fills .................................. 61
5.1 Introduction ............................................................................................................................. 61
5.2 Input Ground Motion .............................................................................................................. 61
5.3 Soil Column Interpretation ...................................................................................................... 62
5.4 Plaxis 2D Modelling ................................................................................................................. 65
5.4.1 Soil Layers with Hardening Soil Small (HSS) Model .......................................................... 65
5.4.2 Soil Layers with PM4Sand Model ..................................................................................... 66

|
 5.4.3 Mesh Generation and Time Step ...................................................................................... 67
5.4.4 Rayleigh Damping ............................................................................................................. 68
5.5 Dynamic Analysis ..................................................................................................................... 70
5.6 Results ..................................................................................................................................... 71
5.6.1 Sensitivity analysis ............................................................................................................ 72
5.6.2 Dynamic Analysis without Consolidation ......................................................................... 73
5.6.3 Dynamic Analysis with Consolidation ............................................................................... 81
5.7 Conclusions .............................................................................................................................. 92
Bibliography....................................................................................................................................... 93
APPENDIX A ....................................................................................................................................... 96
A.1 Earthquake motions ................................................................................................................ 96
A.2 Liquefaction Potential according to NCEER Method............................................................. 104
Α.3 Results for the dynamic analysis with consolidation regarding the generation of excess pore
pressures ..................................................................................................................................... 109
APPENDIX B ..................................................................................................................................... 114
Dafalias & Manzari, 2004 - Model Formulation .......................................................................... 114
B.1 Elastic/Plastic part ............................................................................................................. 114
B.2 Critical state behaviour ..................................................................................................... 114
B.3 Bounding, Dilatancy & Critical surfaces............................................................................. 115
B.4 Yield surface ...................................................................................................................... 116
B.5 Plastic modulus & Hardening/Softening rule .................................................................... 116
B.6 Effective stress reduction due to increased dilatancy ...................................................... 117
APPENDIX C ..................................................................................................................................... 118
PM4Sand Model Formulation ..................................................................................................... 118
C.1 Critical State Soil Mechanics Framework .......................................................................... 118
C.2 Bounding, Dilatancy and Critical Surfaces ......................................................................... 120
C.3 Yield Surface ...................................................................................................................... 120
C.3 Fabric effects ..................................................................................................................... 121
C.4 Stress reversal and Initial Back-Stress ratio Tensors ......................................................... 122
C.5 Elastic Components of the Model ..................................................................................... 122
C.6 Plastic Components of the Model ..................................................................................... 123
Plastic Modulus & Hardening/Softening rule .......................................................................... 124
Plastic volumetric contraction ................................................................................................. 125
Plastic volumetric dilation ....................................................................................................... 126

|
C.7 Primary Model parameters ............................................................................................... 128
C.8 Secondary Model Parameters ........................................................................................... 129

|
Acknowledgements

The current report includes the thesis project I conducted at Delft University of Technology to
graduate as a Master of Science in Geotechnical Engineering. I would like to take this opportunity
to thank several persons who have played a significant role in both the completion of the project
and to my professional and academic guidance.

First of all, I would like to thank Boskalis for offering me the opportunity to perform the current
master thesis project. Moreover, I would like to thank the committee members starting from my
daily supervisor, Richard de Jager, for his valuable advice and for guiding me towards the right
direction especially when things were not going as planned. His continuous support played an
important role in the proper completion of this project. I would also like to thank Ronald Brinkgreve
for providing me with proper instructions and clarifying my questions and doubts regarding the
Plaxis software along the entire thesis journey. Furthermore, I would like to thank both Michael
Hicks and Cor Kasbergen. Their expertise and professional instructions helped me to raise this
project on a higher level.

Finally, I want to thank my parents, my friends from TU Delft and especially my flatmates, Luka and
Uros, for their invaluable support and for being always there for me, throughout all good and bad
moments of the entire TU Delft journey.

Aristotelis Tziolas,

Delft, July 2019

|
Abstract

The earthquake-induced liquefaction is a high-risk phenomenon for dredging industries, which need
to set strict requirements in order to avoid potential disastrous effects for the project. Different
types of liquefaction exist which can be triggered over a wide range of soil types and for different
loading conditions. The liquefaction triggering due to an earthquake event is dependent on the soil
behaviour under undrained cyclic loading.

The assessment of the liquefaction hazard during an earthquake is mainly based so far on empirical
procedures. The most common used in practise is the NCEER method (Youd & Idriss, 2001) which
is established according to empirical evaluation of field observations and in-situ testing. However,
the NCEER method can be inaccurate for the design primarily due to its empirical nature as it is
capturing different soil types and loading conditions. For that purpose, advanced constitutive
models can provide more precise assessments as they can be calibrated for specific site conditions.
Such a model is the PM4Sand, which is very attractive for practical applications because there are
only a few model parameters to be determined in the calibration process.

The first part of the current thesis project includes the validation of the PM4Sand model for both
earthquake-induced and static liquefaction according to undrained Cyclic Direct Simple Shear
(𝐶𝐷𝑆𝑆) tests and undrained Direct Simple Shear (𝐷𝑆𝑆) respectively, performed on Ottawa F-65
Sand. The influence of the model parameters is examined throughout a parametric assessment
analysis. It was observed, that the model approximates well the general features of both cyclic and
static loading. Regarding cyclic loading, it produced similar responses in terms of excess pore
pressures generation and stress paths even though it slightly overpredicts the cyclic resistance for
small number of loading cycles and underpredicts the cyclic resistance for large number of loading
cycles. Regarding static liquefaction, even if the model had initially overestimated the response, it
was able to simulate successfully the static liquefaction behaviour after a recalibration process was
established.

The next part of the project includes the performance of the PM4Sand model for the prediction of
earthquake-induced liquefaction in hydraulic fills, which are analysed for several different seismic
motions. The fill is placed over a different range of relative densities and it is modelled in Plaxis
software as a 1-D soil column. The fill layers that are prone to liquefy, are modelled with the
PM4Sand model while the layers that are not susceptible to liquefaction are modelled with
Hardening Soil Small (HSS) model. The PM4Sand layer is calibrated according to factors that are
accounting for the in-situ state of the fill and the magnitude of the earthquake motions. The
dynamic analyses are performed with and without consolidation and the lateral boundaries used
are tied degrees of freedom. The results in terms of excess pore pressures generation are examined
throughout the whole earthquake motion. Moreover, the onset of liquefaction in the hydraulic fill
is captured when the excess pore pressure ratio has reached a value of around 1.0 (𝑟𝑢 ≈ 1).

It is shown, that the PM4Sand model is indeed applicable for the prediction of earthquake-induced
and static liquefaction in hydraulic fills. The effect of the in-situ state of the fill, in particular the
relative density, has a critical role on the liquefaction susceptibility, which is a lot representative to
what has been observed in reality. According to PM4Sand model, the loosely-packed fills (𝐷𝑅 =

|
30% and 𝐷𝑅 = 40%) are indeed more susceptible to liquefaction compared to the densely-packed
fills (𝐷𝑅 = 50% and 𝐷𝑅 = 60%) which showed less or even no liquefaction potential due to the
earthquake events. On the other hand, the largest drawback of the NCEER method it its empirical
nature, as for the current project it is proved to be conservative for the design. More specifically, it
predicted liquefaction for almost all the hydraulic fills (𝐷𝑅 = 30% to 𝐷𝑅 = 60%) analyzed for all
different earthquake motions.

Regarding the dynamic analyses with consolidation, the results related to the earthquake-induced
liquefaction of the fills are more representative to realistic conditions as there is a better distribution
of excess pore pressures along the soil column with respect to the dynamic analyses without
consolidation. For the latter type of analysis, in the loosely-packed fills (𝐷𝑅 = 30% and 𝐷𝑅 = 40%)
there is a better diffusion of excess pore pressures more for the signals of low dominant frequencies
regardless the peak ground acceleration values of the input signal. In the densely-packed fills (𝐷𝑅 =
50% and 𝐷𝑅 = 60%) the same phenomenon takes pace more for the signals of high dominant
frequencies. However, a localization of liquefied zones is observed in distinct parts along the fill
layer for the rest of the signals.

|
Chapter 1: Introduction

1.1 Problem Definition

Liquefaction has been so far one of the major concerns in the area of geotechnical
engineering. Loose granular soils tend to contract under cyclic loading induced by earthquake
shaking. If the soil is saturated and highly unable to drain, the resulting rearrangement of soil
particles can transfer normal stresses from the sand skeleton to pore water. The inability of
pore water to flow out causes significant development of excess pore pressures. The result is
a reduction in the effective confining stress within the soil and an associated loss of strength
and stiffness that contributes to deformations of the soil deposit (Idriss & Boulanger, 2008).

The earthquake-induced liquefaction in hydraulic fills is a well-recognized high-risk

phenomenon. Dredging industries need to set strict requirements in order to cope with these
risks as liquefaction triggering duo to a seismic event can cause devastating effects to the
project. Such a requirement is the high degree of compaction which can result in an over-
conservative design and raise significantly the cost of the project. Consequently, in the
industry there are several gaps in knowledge on how to achieve optimum design in hydraulic
fills. The key question that arises from this master thesis project is how to optimise the design
of hydraulic fills subjected to earthquake loading and in order to improve the reliability and
overall safety of the structure using finite element software. Better design approaches can
minimize the costs and reduce the time required for the completion of the project.

One of the most important available case histories that gave significant insight in the
liquefaction triggering of hydraulic fills under seismic loading is the upstream flow failure of
the Lower San Fernando Dam. In 1971, after the San Fernando earthquake stroke, a major
slide occurred due to liquefaction (loss in strength) of a zone of hydraulic sand fill located
within the lower sections of the upstream shell. The fill was mainly consisted of young
deposits (looser packing, lower stiffness). A significant remark from this case history is that:
very young, very loose, non-plastic or low-plastic soils tend to be more susceptible to
significant and rapid strength loss than older, denser, and/or more plastic soils (Robertson P.
, 2009).

Another well-documented case history that attracted considerable interest amongst

companies regarding the rational design and safe performance of hydraulic fills is the failure
of the Nerlerk berm failure. The Nerlerk berm is an underwater hydraulic fill structure
designed under inclination to support an offshore drilling platform in the Beautfort Sea. The
construction started in 1982 but suddenly interrupted before the berm had reached its
required elevation by a rapid failure including the formation of five slides. Hicks & Boughrarou
(1998) after performing finite element analysis, concluded that failure occurred under static
liquefaction of the berm which is triggered by a combination of rapid sand deposition and
limited movements in the weak underlying clay layer. However, what caused much confusion
about the berm failure is how liquefaction occurred since its state was not especially loose.

1|Page
Later, Hicks & Onisiphorou (2005) performed stochastic analysis and found out that a
liquefaction failure mechanism triggered due to the arrangement of looser zones within the
fill in what was, on average, predominantly dilative. Hence, an important lesson learned is
that a dilative fill may also liquefy due to the presence of semi-continuous loose zones arising
from deposition-induced anisotropy (Hicks & Onisiphorou, 2005).

The assessment of the liquefaction hazard during an earthquake is mainly based on empirical
methods. The most widely used framework for that purpose was provided in the 1996 NCEER
and 1998 NCEER/NSF workshops (Youd & Idriss, 2001). This procedure is based on empirical
evaluation of field observations and in-situ testing which provides a useful tool for the
evaluation of liquefaction susceptibility. However, the largest drawback of the NCEER
method is its empirical nature as it is applied for conditions that cover the aforementioned
empirical data. Thus, due to the fact that these data account for varying loading condition
and different soil types, the NCEER method may be inaccurate and unconservative for the
design.

In order to overcome the aforementioned limitations, Finite Element Method programmes

appear to be a powerful tool. The FEM software allows for modelling of complex geometries
and gives a clear insight in the failure mechanisms and the resulting deformations at both,
prior and post failure conditions. The last years several FEM software are being developed for
the evaluation of liquefaction potential under seismic loading (e.g PLAXIS UBC3D-PLM) but
certain limitations in the accuracy of the solution is encountered.

The PM4Sand is a constitutive model, which as implemented in Plaxis software, is a very useful
and powerful tool for that purpose. The PM4Sand model is a sand plasticity model (Boulanger
& Ziotopoulou, 2015) which can simulate with high accuracy the response of sands under
earthquake loading, including the pore pressure generation and liquefaction phenomena.
One of the major advantages of advanced FEM software designed for dynamic loading, such
as the PLAXIS PM4Sand, is that they are able to conduct a site-specific analysis.

2|Page
1.2 Research Questions

The main question of the current master thesis project is:

What are the possibilities of designing hydraulic fills more efficiently against earthquake
induced liquefaction by using the PM4Sand constitutive model?

What are the capabilities and the limitations of the PM4Sand model?

The key questions of this project are:

 Is the PM4Sand model applicable for the simulation of earthquake-induced

liquefaction in hydraulic fills?
 Is the PM4Sand model applicable for the simulation of static liquefaction in
hydraulic fills?
 What are the effects of the in-situ conditions on the liquefaction triggering in
hydraulic fills?
 What are the capabilities and limitations of PM4Sand model?

3|Page
Chapter 2: Literature study

2.1 Liquefaction Hazard

2.1.1 Introduction

The liquefaction hazard is a high-risk phenomenon for dredging industries which need to set
strict requirements in order to avoid potential disastrous effects for the project. Different
types of liquefaction exist which can be triggered over a wide range of soil types and for
different loading conditions. In this chapter, the concept of critical state is introduced and the
state parameter is presented which gives a clear insight of whether a soil can be susceptible
or not to liquefaction. The liquefaction triggering due to an earthquake event is dependent
on the soil behaviour under undrained cyclic loading. Accordingly, the criteria used to define
liquefaction in laboratory testing are classified either with respect to reaching a specific
percentage of shear strain at a given number of loading cycles or based on excess pore
pressure evolution after a certain value is reached. The assessment of the liquefaction hazard
during an earthquake is so far mainly based on empirical procedures. The most common used
in practise is the NCEER method (Youd & Idriss, 2001) which is based on field observations
from case histories that suffered from liquefaction. However, the NCEER procedure can be
inaccurate due to its empirical nature as it is established with respect to different soil types
and loading conditions. For that purpose, advanced constitutive models can provide more
precise assessments as they can be calibrated for specific site conditions. Such a model is the
PM4Sand which is based on previously developed sand plasticity model by (Dafalias &
Manzari, 2004) by incorporating the core of NCEER method. In the PM4Sand version the
multiaxial formulation implemented by (Dafalias & Manzari, 2004) is simplified for 2D
applications and the model is developed for plane-strain conditions. As a result, there are
some limitations which will be investigated along with the capabilities of PM4Sand later on
throughout this thesis project.

2.1.2 Types of Liquefaction

In literature, the liquefaction phenomena are mainly divided into two groups: flow
liquefaction and cyclic mobility. These definitions provide important information concerning
the conditions of liquefaction triggering.

Flow liquefaction: This type of liquefaction is related to enormous instabilities, known as flow
failures and can lead to the most detrimental effects concerning liquefaction phenomena.
Such instabilities can be present in sloping grounds. Flow liquefaction is triggered when the
𝐵
static shear stress of a soil body (𝜏𝑠𝑡𝑎𝑡𝑖𝑐 ) required to sustain equilibrium exceeds the shear
strength of the soil in its liquefied state (𝜏 𝐷 ) (Figure 2.1b). In that case, the large deformations
produced by flow liquefaction are actually driven by static shear stresses (Kramer, 1996). This
type of liquefaction is characterized by its sudden development where the failure can be
extended over large distances.

Cyclic mobility: In contrast to flow liquefaction, cyclic mobility can happen when the static
shear stress of a soil mass is lower than the shear strength of the soil in its liquefied state. This

4|Page
type of liquefaction can occur in granular soils for different in-situ relative densities (loose to
dense) and results into large deformations during a seismic event. The produced
deformations at failure develop incrementally during earthquake shaking as a result of both
static and cyclic loading. The excess pore pressures are evolving until p’=0 is reached (Figure
2.1a). A common type of cyclic mobility is lateral spreading which can develop on gently
sloping ground or on virtually flat ground surfaces. If structures are present extreme damage
can be caused (Kramer, 1996).

Another type of cyclic mobility is level-ground liquefaction. In that case, static horizontal shear
stresses that could cause lateral permanent deformations are absent. The excess pore
pressures induced by shaking dissipate by the upward movement of water causing excessive
settlements and sand boils. This type of liquefaction is dependent on the time required to
reach hydraulic equilibrium and thus, can occur also after the seismic event stops (Kramer,
1996).

Figure 2.1 a,b: Two different types of liquefaction: a) Cyclic mobility is described by the
accumulation of excess pore pressures until p’=0. b) Flow liquefaction is described as sudden
loss of shear strength driven by static shear stresses (τstaticΒ> τD) (Idriss & Boulanger, 2008)

2.2 Principles of Liquefaction

2.2.1 Critical Void Ratio

The determination of critical void ratio (𝑒𝑐𝑠 ) is of high importance in order to be able to explain
the soil behaviour under monotonic (Chapter 2.2.3) and cyclic loading (Chapter 2.2.4). In
particular, the state parameter (𝛹) (Chapter 2.2.2) is defined as a function of 𝑒𝑐𝑠 which will
determine whether the soil is susceptible to liquefaction.

The behaviour of sands under monotonic shearing is based on the concept of critical-state
theory (Casagrande, 1936). This theory refers to the condition that exists in sand when it is
being sheared continuously and no further changes in volume or stress are occurring (Idriss &
Boulanger, 2008). The void ratio at this condition is the critical void ratio (𝑒𝑐𝑠 ) and is uniquely
related to the effective confining stress (𝑝’). The relationship between 𝑒𝑐𝑠 and 𝑝’ is described
by the locus of points which represent all possible combinations between both parameters,
named as Critical State Line (𝐶𝑆𝐿) (Figure 2.2). It is important to note that the 𝐶𝑆𝐿 is not
unique but varying for different types of sand.

5|Page
Figure 2.2: Monotonic drained (constant mean effective stress) and undrained (constant void
ratio) stress paths towards the critical state line for loose of critical and dense of critical sands
(Idriss & Boulanger, 2008)

2.2.2 State Parameter

A common measure for assessing the liquefaction susceptibility of the soil is the state
parameter (𝛹) introduced (Been & Jefferies, 1985). The state parameter is defined as
𝛹 = 𝑒 − 𝑒𝑐𝑠 where 𝑒 is the current void ratio and 𝑒𝑐𝑠 is the void ratio of the 𝐶𝑆𝐿 at the
same mean effective normal stress (𝑝’) (Figure 2.3). In this way, the soil state can be
expressed with respect to its 𝐶𝑆𝐿. Loose-of-critical soils (contractive) have positive 𝛹 and
may be susceptible to liquefaction (flow liquefaction) while dense-of-critical soils (dilative)
have negative 𝛹 may not be susceptible to liquefaction (nevertheless, cyclic mobility can
occur).

Figure 2.3: State parameter (Kramer, 1996)

2.2.3 Monotonic Loading

In monotonic loading, liquefaction is defined according to the monotonic undrained stress-

strain responses (Figure 2.4). The susceptibility of the soil to liquefaction is depending on the
sate parameter 𝛹. Soil states with 𝛹 > 0 (Figure 2.4 - A) are susceptible to flow liquefaction
only if the static shear stress required to sustain equilibrium (𝜏 (𝑖) ) exceeds the shear strength

6|Page
of the soil in its liquefied state (𝜏 (𝑖𝑖) ) (Chapter 2.1.2). On the other hand, cyclic mobility can
occur in soils with 𝛹 < 0 (Figure 2.4 - B, C).

The representative stress-strain responses for liquefaction under monotonic undrained

loading are depicted in Figure 2.4, in which three different failure modes are illustrated:
liquefaction (A), dilation (B) and limited liquefaction (C) (Kramer, 1996). The failure mode (A)
is also named ‘flow liquefaction’ in which the soil (very loose) is reaching the peak undrained
strength at very low shear strain levels. After this point of instability (i), the response is
followed by a sudden collapse and subsequent flow to large strains at a constant liquefied
strength under low effective confining stress (critical state). On the other hand, dilation
(dense) (B) is characterized by an initial contractive response followed by dilation in which the
effective confining stresses are increased for larger shear strains. Apparently, in this failure
mode there is no liquefaction occurrence. The failure mode for intermediate densities (C), is
named ‘limited liquefaction’ in which the soil, after reaching the peak undrained strength at
low strain levels (point of instability i) is followed by a short period of strain-softening
response after which dilation occurs at intermediate strains.

Figure 2.4: Three different types of monotonic undrained failure: Liquefaction (A), Dilation (B)
and Limited Liquefaction (C), (Kramer, 1996)

2.2.4 Cyclic Loading

The liquefaction triggering of a soil under cyclic undrained loading arising from a sudden
earthquake event is dependent on the soil behaviour throughout this event. The behaviour is
dominated by the soil density, as loose granular soils are highly prone to liquefaction. For
loose soils (𝛹 > 0), the liquefaction triggering implies continuous contractive behaviour
which is restrained by the phase transformation line (PT) (Chapter 2.7.1). When the stress
state is reaching that line, the behaviour switches from contractive to dilative. After this point,
there is a significant loss of stiffness and a gradual shear strain accumulation.

In laboratory testing, the liquefaction triggering under cyclic loading is defined according to
two criteria. The first one indicates, that liquefaction is triggered after a specific percentage
of strain is reached. An axial strain of 5% double-amplitude (𝜀𝛼𝐷𝐴 ) is commonly adopted for
Cyclic Triaxial tests (ICU) whereas in Cyclic Direct Simple Shear (CDSS) tests, a shear strain of
3% single-amplitude is used. The second criterion is based on the 𝑟𝑢 parameter which is

7|Page
defined as the generation of excess pore water pressure (𝛥𝑢) normalized by the minor
effective consolidation stress (𝜎 ′ 3𝑐 ) in a triaxial test.

𝛥𝑢
𝑟𝑢 = (2.1)
𝜎 ′ 3𝑐

For Cyclic Direct Simple Shear (CDSS) tests, the definition of 𝑟𝑢 is similar in which the
consolidation stress from eq.(2.1) is replaced by the vertical effective stress.

The 𝑟𝑢 parameter can reach a maximum value of 1.0 when the excess pore pressure is
approaching 100% of the initial confinement (𝛥𝑢 = 𝜎 ′ 3𝑐 ). At that point, the shear strength
of the sand specimen is diminished and ‘’initial liquefaction’’ occurs (Seed & Lee, 1996).
However, the 𝑟𝑢 = 1.0 condition cannot be reached for all type of soils. In sandy silts or silty
sands the 𝑟𝑢 can reach a value of 0.90 to 0.95 resulting in a significant softening response (>
5% axial strains of double-amplitude) which corresponds to ‘initial liquefaction’ (Ishihara,
1996). Moreover, in dense sands, the 𝑟𝑢 = 1.0 condition may never be reached as it is less
prone to liquefy due to higher strength.

It is important to note that the aforementioned values are not used as strict measures but as
an index to define liquefaction and therefore can vary in the literature.

For the evaluation of the liquefaction potential, the cyclic induced shear stress is commonly
expressed in terms of Cyclic Stress Ratio (𝐶𝑆𝑅). In Cyclic Triaxial tests, 𝐶𝑆𝑅 is defined as the
maximum cyclic shear stress (𝑞𝑐𝑦𝑐 /2) divided by the isotropic consolidation stress (𝜎 ′ 3𝑐 ).

The cyclic shear strength of sands can be described by the Cyclic Resistance Ratio (𝐶𝑅𝑅) which
is termed as the 𝐶𝑆𝑅 required to reach liquefaction for a specified number of loading cycles.
In general, 𝐶𝑅𝑅 depends on the amplitude and the frequency of loading. A larger 𝐶𝑅𝑅 implies
that fewer number of loading cycles are required to trigger liquefaction (𝑟𝑢 = 100% or
𝛾𝑐𝑦𝑐 = 3%) (Figure 2.5)

Furthermore, in laboratory testing, the fabric of the soil sample plays an important role in the
liquefaction triggering under cyclic loading (Nemat & Tobita, 1982). The fabric depends on
the preparation method (inherent anisotropy) and on whether or not the sample has been
pre-strained. If the strain levels under drained cyclic loading are such that volumetric
expansion occurs, the sample shows decreased liquefaction resistance in cyclic undrained
loading. In addition, the sample preparation method (fabric) combined with a particular
loading path may result in different liquefaction resistances (e.g higher 𝐶𝑅𝑅 in triaxial loading
and significantly lower 𝐶𝑅𝑅 in simple shear loading).

8|Page
Figure 2.5: Cyclic strength curve of sand (left) after the conduction of three cyclic triaxial
tests of the same volumetric and stress state (DR, σ’3c). The tests are performed for different
cyclic stress amplitudes (right).

2.3 Simplified Procedure for the Simulation of Earthquake Loading

2.3.1 Introduction

In the 1996 NCEER and 1998 NCEER/NSF workshops (Youd & Idriss, 2001), a widely used
method for the assessment of the liquefaction resistance of soils under level to gently sloping
ground was constituted. The procedure is based on empirical evaluation of field observations,
laboratory and in-situ testing. The NCEER method requires the determination of two
parameters in order to evaluate the cyclic resistance of soils: the Cyclic Stress Ratio (𝐶𝑆𝑅)
(Chapter 2.3.2) and the Cyclic Resistance Ratio (𝐶𝑅𝑅) (Chapter 2.3.3).

The shear stresses induced by earthquake loading in a soil body below a depth 𝑧 can be
estimated using the simplified procedure proposed by (Seed & Idriss, 1971). The subsurface
is simplified as a soil column of unit width and length which behaves as a rigid body (Figure
2.7). The earthquake induced shear stresses are calculated by the following expression:

𝑊 𝛾𝑧 𝛼𝑚𝑎𝑥
𝜏𝑚𝑎𝑥 = 𝐹 = 𝑚𝑎 = 𝑎𝑚𝑎𝑥 = 𝛼𝑚𝑎𝑥 = 𝜎𝜈𝜊 (2.5)
𝑔 𝑔 𝑔

where 𝐹: horizontal seismic force, 𝑚: total mass of soil column, 𝑎𝑚𝑎𝑥 : maximum horizontal
acceleration at the ground surface, 𝑊: weight of the soil column, 𝛾: average total unit weight
of the soil above depth 𝑧 and 𝜎𝜈𝜊 : total vertical stress at the same depth.

It is clear, that the simplified procedure assumes a constant distribution of both peak
horizontal acceleration (𝑎𝑚𝑎𝑥 ) and stress ratio (𝜏𝑚𝑎𝑥 /𝜎′𝜈𝜊 ) along the soil column (Figure 2.6).
However, this assumption is not the case in reality because it does not take into account the
subsurface ability to deform.

9|Page
Figure 2.6: Simplified representation of earthquake loading along a soil profile

2.3.2 Cyclic Stress Ratio (CSR)

The 𝐶𝑆𝑅 represents the dynamic stress conditions induced by the earthquake and is
calculated as the normalized cyclic horizontal shear stress (𝜏𝑐𝑦𝑐 ) by the vertical effective
overburden stress (𝜎 ′ 𝜈𝜊 ):
𝜏𝑐𝑦𝑐
𝐶𝑆𝑅 = (2.6)
𝜎 ′ 𝜈𝜊

The 𝐶𝑆𝑅 can be calculated by the following expression (Idriss I. , 1999)proposed by (Seed &
Idriss, 1971):
𝜏𝑐𝑦𝑐 𝜏𝑚𝑎𝑥 𝜎𝜈𝜊 𝛼𝑚𝑎𝑥
𝐶𝑆𝑅 = = 0.65 = 0.65 ′ 𝑟 (2.7)
𝜎′𝜈𝜊 𝜎′𝜈𝜊 𝜎 𝜈𝜊 𝑔 𝑑

where 𝜏𝑚𝑎𝑥 is the peak cyclic shear stress induced by earthquake, 𝜎𝜈𝜊 and 𝜎 ′ 𝜈𝜊 are the vertical
total and effective overburden stress respectively, 𝛼𝑚𝑎𝑥 is the maximum horizontal
acceleration at the ground surface, 𝑔 is the gravity acceleration and 𝑟𝑑 is a stress reduction
coefficient. The variation of 𝑟𝑑 with depth for earthquakes with different magnitude can be
calculated based on expressions proposed by (Idriss I.M, 1999)(Appendix A.2)

2.3.3 Cyclic Resistance Ratio (CRR)

The 𝐶𝑅𝑅 is described as the capacity of soil to resist liquefaction or as the cyclic stress ratio
required to trigger liquefaction. In particular, 𝐶𝑅𝑅 is the required 𝐶𝑆𝑅 to trigger liquefaction
at a certain number of loading cycles.

The Cyclic Resistance Ratio (𝐶𝑅𝑅) against liquefaction triggering is most commonly estimated
based on in-situ tests with case-history-based liquefaction correlations (Idriss & Boulanger,
2008). The 𝐶𝑅𝑅 curve is described as the boundary line that separates the figure into two
areas that are indicating either liquefaction or non-liquefaction occurrence (Figure 2.7).

10 | P a g e
The 𝐶𝑅𝑅 curve is illustrated in terms of either 𝑆𝑃𝑇 or 𝐶𝑃𝑇 penetration resistance. In the first
case, it is represented as a function of the parameter (𝑁1 )60 (Seed & De Alba, 1986) which
stands for an 𝑆𝑃𝑇 blow count normalized to an overburden pressure of approximately
100𝑘𝑃𝑎 and a hammer efficiency of 60%. Regarding Cone Penetration Tests, the 𝐶𝑅𝑅 curve
is determined from the normalized, dimensionless and corrected 𝐶𝑃𝑇 resistance 𝑞𝑐1𝑁
(Robertson & Wride, 1998) for clean sands. It is important to note, that these are standard
curves and correction factors need to be applied (see following sections).

Figure 2.7: Liquefaction curve representing the boundary regarding occurrence and non-
occurrence of liquefaction for different combinations of Cyclic Resistance Ratio (𝐶𝑅𝑅) based
on 𝑆𝑃𝑇 data for different combinations of 𝐶𝑅𝑅 and (𝑁1 )60 . (Idriss & Boulanger, 2008) for an
earthquake of 7.5 magnitude.

2.3.4 Correction Factors

MSF correction factor

In order to convert the 𝐶𝑅𝑅 curves to a magnitude different than 7.5, (Idriss, 1999) proposed
the following relation that connect the so-called Magnitude Scale Factor (MSF) for sands to
the earthquake magnitude 𝑀:

𝑀
𝑀𝑆𝐹 = 6.9exp (− ) − 0.058 ≤ 1.8 (2.8)
4

𝐾𝜎 correction factor

The 𝐶𝑅𝑅 of sand depends on the effective confining stress (𝜎’𝜈𝑐 ) because the latter accounts
for the tendency of sand to contract or dilate. Based in previous research (Vaid &
Sivathayalan, 1996), it is proven that that the 𝐶𝑅𝑅 is decreased with an increasing confining

11 | P a g e
stress for a given value of relative density (or for a given value of relative state parameter in
Chapter 2.2.2).

This effect of overburden stress can be captured by the 𝐾𝜎 correction factor introduced by
(Seed, 1983) as:

𝐶𝑅𝑅𝜎’𝜈𝑐≠1 𝑎𝑡𝑚
𝐾𝜎 = (2.9)
𝐶𝑅𝑅𝜎’𝜈𝑐=1 𝑎𝑡𝑚

where 𝜎′𝜈𝑐 is the effective confining stress.

𝐾𝜎 correction factor

The 𝐶𝑅𝑅 of sand is highly affected by the presence of initial static shear and can be captured
by the following correction factor proposed by (Seed, 1983):

𝐶𝑅𝑅𝛼≠1
𝐾𝛼 = (2.10)
𝐶𝑅𝑅𝛼=0

where 𝛼 is the static shear stress ratio defined as:

𝜏𝑠
𝑎= (2.11)
𝜎’𝜈𝑐

where 𝜏𝑠 is the horizontal shear stress

Later, Boulanger (2003a) showed that 𝐾𝛼 can be related to the sand’s state as it depends on
both relative density (𝐷𝑅 ) and confining stress (𝜎’𝜈𝑐 ) which are both accounting for the dilative
tendency of sand (Figure 2.8). The effect of relative density is attributed to the fact that sand
is more strongly dilatant in shear as 𝐷𝑅 is increased (at the same 𝜎’𝜈𝑐 ). The effect of confining
stress (𝜎’𝜈𝑐 ) is attributed to the fact that sand is more strongly dilatant in shear at the lower
confining stresses (for the same 𝐷𝑅 ).

Hence, the correction factor 𝐾𝛼 bridges cyclic and monotonic behaviour, or cyclic mobility and
flow liquefaction. This factor, which is applied when the presence of initial static shear stress
is presented, is used e.g in slope geometries. In case of very loose sands, for an increasing
slope angle the failure is dominated by flow liquefaction as the soil is becoming weaker. On
the other hand, in case of very dense sands, the failure is dominated by cyclic mobility as the
soil is becoming stronger.

2.3.5 Simplified Liquefaction Triggering Assessment

The susceptibility of a soil to liquefaction can be assessed according to (Seed & Idriss, 1971) in
terms of the factor of safety (𝐹𝑆). This factor is defined as the ratio between the liquefaction
resistance (𝐶𝑅𝑅) to the seismic induced loading (𝐶𝑆𝑅) as:

𝐶𝑅𝑅
𝐹𝑆 = (2.12)
𝐶𝑆𝑅

12 | P a g e
In case 𝐹𝑆 < 1, the soil’s potential to liquefaction is high. In the aforementioned equation,
the 𝐶𝑅𝑅 is dependent on factors concerning the earthquake magnitude (𝑀𝑆𝐹), the
confinement level (𝐾𝜎 ) and sloping ground effects (𝐾𝛼 ) as:

𝐶𝑅𝑅 = 𝐶𝑅𝑅7.5 ∙ 𝑀𝑆𝐹 ∙ 𝐾𝜎 ∙ 𝐾𝛼 (2.13)

2.4 Correlation between CRR from Lab Tests to Field Conditions

The different states of consolidation stress measured in Cyclic Simple Shear and Cyclic Triaxial
(𝐼𝐶𝑈) tests are the cause of differences in the measured 𝐶𝑅𝑅. The 𝐶𝑅𝑅 of anisotropically
consolidated specimens (𝐾𝑜 ≠ 1) can be related to the CRR of isotropically consolidated
specimens (𝐾𝑜 = 1) by the relationship proposed by (Isihara, 1985) (Appendix A.1 – eq.(A.9)).

The relation between the CRR from a simple shear test to the 𝐶𝑅𝑅 from the Triaxial test (𝐼𝐶𝑈)
can be calculated according to (Idriss & Boulanger, 2008) (Appendix A.2 – eq.(A.10)).

When a second direction of cyclic loading is added, the 𝐶𝑅𝑅 decreases by 10 − 15%, as
summarized by (Pyke , Chan, & Seed, 1974) and (Isihara, 1996). For level ground conditions,
the earthquake loading is best approximated as a two-directional simple shear loading, so the
𝐶𝑅𝑅 from a unidirectional simple shear test would be reduced by 10% to represent in-situ
conditions. The relationship between in-situ 𝐶𝑅𝑅𝑓𝑖𝑒𝑙𝑑 and either 𝐶𝑅𝑅𝑇𝑋 or 𝐶𝑅𝑅𝑆𝑆 can be
computed according to (Appendix A.2 – eq.(A.11-A.12)).

2.5 Hydraulic Fill Structure

2.5.1 Introduction

Land reclamation is generally defined as the process of raising the elevation of a seabed or
riverbed to a level in which water is being replaced by dry soil in order to create a new land
(Van't Hoff & Nooy van der Kolff, 2012). This method has been undertaken in different parts
of the world aiming to substantially increase the amount of land that is available for various
activities due to significant increase in world’s population, rise of sea levels (global warming)
and upraising energy demand. Land reclamation can be implemented in various projects
ranging from onshore to offshore applications. Related examples are the expansion of
commercial and industrial activities such as ports (Port of Rotterdam), airports (Kansai
International Airport in Japan) and the construction of artificial islands on the ocean (Palm
Island in Dubai).

There are many different methods that can be used for land reclamation purposes. These
methods can be divided mainly into two types, the filling and non-filling. The filling type can
be categorized into dry earth movement and hydraulic filling. The current research is mainly
focused on the hydraulic filling method.

13 | P a g e
The hydraulic filling method can be achieved either underwater or above water. Underwater
placement methods provide a lower relative density (𝐷𝑅 < 50%) than the placement
methods used above the water level (𝐷𝑅 > 50%). Hence, for underwater placement the fill
can be more susceptible to liquefaction (looser-packing, lower stiffness).

2.5.2 Hydraulic Filling Method

The hydraulic filling method aims to create new land by raising the ground level from below
water to a surface with dry land. This procedure of new land creation is achieved by applying
the following steps (Van't Hoff & Nooy van der Kolff, 2012):

 Acquisition of the fill material (mainly granular soil) from either borrow or dredging
area which is proximate to the site. The latter case is undertaken by floating
equipment (dredgers)
 Transportation of the fill mass from the borrow area to the project site by dredger,
barge or pipeline
 Placement of the construction material based on soil flow. The flow is accommodated
by the use of (processed) water by which a soil mixture is created. This mixture can
be pumped and lied on the seabed according to the current conditions

2.5.3 In-situ Conditions of the Reclamation Structure / In-situ Volumetric State (Dr)

It is of high importance to determine the in-situ conditions and the intrinsic properties of the
hydraulic fill structure. The anticipated geometry as well as the in-situ relative density need
to be assessed. This will provide the appropriate information to be used as an input in the
PM4Sand model.

The most common framework used for the assessment of the relative density is the indirect
methods based on in-situ testing (𝐶𝑃𝑇 and 𝑆𝑃𝑇). Over the years, several correlations have
been established to relate the cone resistance (𝑞𝑐 ) as well as the blow counts (𝑁𝑆𝑃𝑇 ) with the
relative density. For Standard Penetration Tests (𝑆𝑃𝑇), the number of blow counts (𝑁𝑆𝑃𝑇 ) is
used to estimate the relative density of sands or gravels according to British Standard
classification (Table 2.3).

Table 2.3: Relative density in relation to 𝑁𝑆𝑃𝑇 (Van't Hoff & Nooy van der Kolff, 2012)

14 | P a g e
The relative density of a hydraulic fill before compaction is highly influenced by its placement
method. The disposal energy plays an important role in the resulting density of the fill. In
particular, higher densities can be achieved by raising the disposal energy which leads to
better compaction. In Table 2.4 several placement methods with the corresponding ranges
of relative densities for silica sands are indicated.

Table 2.4: Relative densities of a hydraulic fill achieved over different placement methods
(Van't Hoff & Nooy van der Kolff, 2012)

2.5.4 Placement Methods

The procedure of hydraulic fill placement can be divided into four types. Each type is briefly
described in this section. More detailed information about technical specifications regarding
fill placement can be seen in the book: ‘’Hydraulic fill manual’’ (Van't Hoff & Nooy van der
Kolff, 2012)

Underwater placements in bulk of fill material (dumping) (Dr=30-50%)

The filling process is guided by 𝑇𝑆𝐻𝐷𝑠 vessels or barges. The soil is loaded into the vessels
and subsequently transported to the reclamation area. Then, the soil mixture is dumped
through the bottom doors of the 𝑇𝑆𝐻𝐷 and consequently falls down towards the seabed.
After this step, the surrounding water will dilute the mixture during its route to the seafloor,
hence lowering its concentration and increasing the seabed area that will be covered by
dumped material. It is important to note, that due to the low relative densities provided by
this placement method, the fill material may likely be susceptible to liquefaction

Rainbowing (Dr=40-60% under water, Dr =60-80% over water)

This technique is often preferred in shallow locations where water depths are limited for
access by 𝑇𝐻𝑆𝐷𝑠. In that case, a suction hopper dredger can rather approach the designated
area. The fill mixture is pumped and sprayed from the dredger resulting in a rainbow arch

15 | P a g e
which can reach distances up to 150𝑚. Beach replenishment/recreation, port expansion and
artificial islands are the major reclamation applications related to this method.

Placement of fill material using a discharge pipeline (Dr=60-70% over water)

The discharging pipeline method is the most common technique used for reclamation.
Pumping through steel pipelines is applied for reclamation above water in areas which are out
of reach of rainbowing. The pipeline may be assisted by pontoons or floating jackets with the
latter connecting the transportation vessels with the reclamation area (Van't Hoff & Nooy van
der Kolff, 2012). The fill material is gently distributed towards the seabed.

Spraying (Dr=20-40% under water)

This method is applied in case of soft subsoil conditions where lower shear strengths are
present. The soil mixture is sprayed smoothly, preventing mud wave formation, instabilities
or inclusions of soft soil deposit and within the fill structure. Spraying is also used when the
sand has to be placed accurately within predefined boundaries according to slope design
requirements. In the present research, this method is not considered as fill is underlain by
bedrock.

Figure 2.21: The rainbowing method of placement (Van't Hoff & Nooy van der Kolff, 2012)

2.6 A Sand Plasticity Model Accounting for Fabric Change Effects

2.6.1 Model Description

PM4Sand is a stress-ratio controlled, critical state compatible, sand plasticity model applied
for earthquake geotechnical engineering as described by (Boulanger & Ziotopoulou, 2015).
The formulation of the model is based on previous works from (Dafalias & Manzari, 2004)
which incorporates the dependence of fabric and fabric evolution on previous loading history.
One of the main assumptions is that only changes in stress-ratio can produce plastic strains.

Moreover, PM4Sand is an effective stress model which accounts for the generation of excess
pore pressure under undrained cyclic loading and the progressive increase in shear strain
accumulation after a specific value of excess pore pressure ratio is reached. In the following
sections a short introduction regarding the formulation of the model from (Dafalias &

16 | P a g e
Manzari, 2004) is elaborated and in Chapter 2.7 the PM4Sand model is described as
implemented in Plaxis 2D software. The detailed description of the (Dafalias & Manzari, 2004)
model can be found in Appendix B. It is important to note, that PM4Sand is a simplification
of the aforementioned model and has some limitation as it is implemented for 2D applications
only.

2.6.2 Conceptual Desription

The elastic response of the (Dafalias & Manzari, 2004) model is defined with respect to the
yield surface. The yield surface is represented as a linear wedge in the 𝑞 − 𝑝 space (Figure
B.1). The elastic response is depending on the density and the stress state below the yield
surface through the elastic modulus. (eq. B.1-B.4)

The model also embodies bounding (𝑀𝑏 ), dilatancy (𝑀𝑑 ) and critical surfaces (𝑀) (Figure
B.1). These surfaces are defined in such a way that are dependent on the soil state through
the state parameter 𝛹 (Chapter 2.2.2). During the loading process the quantities 𝑀𝑏 and 𝑀𝑑
are varying according to the soil state in order to have 𝑀𝑏 = 𝑀𝑑 = 𝑀 when the critical state
condition is reached (Chapter 2.2.1). Moreover, for a denser-than-critical state (𝛹 < 0) one
has 𝑀𝑑 < 𝑀 < 𝑀𝑏 while for a looser-than-critical state (𝛹 > 0) one has 𝑀𝑑 > 𝑀 > 𝑀𝑏 (eq.
B.7-B.8)

The bounding surface (𝑀𝑏 ) represents the onset of softening behaviour and the progressive
soil shearing towards critical state. In dense sands, the stress state is reaching the bounding
surface to represent the explicit peak in hardening which is followed by sudden softening. On
the other hand, in loose sands, the bounding surface lies on or close to the critical surface (𝑀)
so as to show hardening behaviour towards failure without any dilation.

The dilatancy surface (𝑀𝑑 ) represents the boundary between contractive and dilative
volumetric behaviour, also called as phase transformation line (PT) (Chapter 2.2.4). In dense
sands, this surface lies below the critical surface (𝑀) in which the dilation leads to a peak
behaviour. On the other hand, in loose sands, the dilatancy surface lies close or on to the
critical surface because there is no dilation.

The critical state surface (𝑀) represents the stress ratio at which 𝛹 = 0 (or 𝑒 = 𝑒𝑐 )(eq. B.6).
After this state is reached, the soil deforms continuously under constant volume which
represents an ultimate strength state for large strains (Chapter 2.2.1).

The model accounts for hardening and softening by kinematic rotation of the yield surface in
the stress space. This is accompanied by the plastic modulus 𝐻 (Chapter B.5). For stress states
above the bounding surface (𝐻 < 0) softening occurs whereas for stress states below the
bounding surface (𝐻 > 0) hardening occurs. The amount of hardening or softening is
depending on the distance of the current stress state to the bounding surface (eq. B.10). More
specifically, the larger the distance, the higher the amount of hardening or softening.

17 | P a g e
Monotonic Loading

In order to get a better understanding on how the model works, both drained (𝑇𝑆𝑃) and
undrained (𝐸𝑆𝑃) stress paths for a dense-of critical (𝛹 < 0) and a loose-of-critical (𝛹 > 0) soil
element confined under mean effective stress 𝑝𝑜′ are illustrated in Figure 2.22.

In monotonic drained loading (𝑇𝑆𝑃), as long as the stress state 𝜂 lies inside the yield surface
only elastic strains are produced for a change in stress-ratio (𝑑𝜂 ≠ 0). When the stress state
crosses the yield surface with 𝑑𝜂 > 0, then plasticity occurs. Meanwhile, the stress state is
lying below the bounding surface (𝐻 > 0), so hardening behaviour is observed.

Initially, the quantity 𝐻 is large since the distance of the stress state to the bounding surface
is also large and therefore, the plastic strains are limited. The dense-of critical element has a
higher 𝐻 since from eq.(B.11) the void ratio 𝑒 is lower. As the loading increment 𝑑𝜂 further
increases, 𝐻 is decreasing (increasing void ratio) which represents less hardening behaviour
towards critical state. Furthermore, the yield surface is moving upwards (increase in 𝑎) so as
𝜂 to remain on 𝑓 = 0 (kinematic and isotropic hardening) and the absolute value of Ψ is
decreasing for both soil elements (eq. B.6).

Figure 2.22: Monotonic stress paths for dense-of-critical and loose-of-critical soil element.

Similarly, the response is only elastic for stress states inside the wedge until the yield surface
is crossed again where plasticity occurs. As the yield surface is moving towards critical state
the amount of induced plastic strains is increasing since the distance between the bounding
surface (𝑀𝑏 ) and the stress state is becoming smaller. During the aforementioned stages,
the response is contractive since the stress state is below the dilatancy surface (𝑀𝑑 ). When
the stress state crosses the dilatancy surface 𝑀𝑑 the response transitions from contraction to
dilation. Furthermore, at the point where the stress state overpasses the bounding surface
𝑀𝑏 , 𝐻 < 0 and thus, softening behaviour is observed. If the loading is to be continued, critical
state is reached as 𝛹 → 0 and 𝑀𝑏 , 𝑀𝑑 → 𝛭.

18 | P a g e
In monotonic undrained loading (𝐸𝑆𝑃) the same logic applies for elasticity and plasticity in the
stress paths. However, the rate with which the plastic strains are being generated is different.
The reduction of mean effective stress 𝑑𝑝′ is larger in the loose-of-critical element as a result
of higher excess pore pressure generation. The plastic modulus 𝐻 is reduced more rapidly
because the stress state is approaching faster the critical state line than in the dense-of-critical
element. Therefore, the loose-of-critical element produces plastic strains faster.

Fabric Effects

One of the main limitations of the previous version of the (Dafalias & Manzari, 2004) model is
that the decrease of effective stresses towards 𝑝′ = 0 during cyclic loading was not well
captured with the stress-path loops being disturbed before reaching that state. That is
because the model didn’t account for the significant changes in fabric during the dilatant
phase of plastic deformation, which has a significant effect on the contractive response upon
reversal of loading (Dafalias & Manzari, 2004).

The latest addition of the model enhances to overcome this limiation by introducing the fabric
dilatancy tensor 𝒛 which evolves as follows:
𝑝𝑙
𝑑𝑧 = −𝑐𝑧 〈−𝑑𝜀𝑣 〉(𝑠𝑧𝑚𝑎𝑥 + 𝑧) (2.14)

Also the parameter 𝐴𝑑 that accounts for dilatancy 𝑑 (eq. B.12) is modified as:

𝐴𝑑 = 𝐴𝑜 (1 + 〈𝑠𝑧〉) (2.15)

where 𝑧𝑚𝑎𝑥 is the maximum value of 𝑧, 𝑐𝑧 controls the evolution of 𝑧, 𝐴𝑜 is a constant, 𝑠 =

±1 according to 𝜂 = 𝑎 ± 𝑚, 𝑧 = 0 (initially) and <> are MacCauley brackets setting negative
values to zero.

In undrained loading (eq. B.13), for a given 𝐾 and 𝐻, an increase in the reduction of 𝑝 requires
an increase in 𝑑. This increase in 𝑑 is required for the proper simulation of effective stress
reduction and modulus degradation under undrained cyclic loading (Dafalias & Manzari,
2004).

Cyclic Loading
𝑝𝑙
During loading from 𝜂 = 0 with 𝑑𝜂 > 0 (Figure 2.23), the response is contractive (𝑑𝜀𝑣 > 0)
𝑝𝑙
as long as 𝜂 < 𝑀𝑑 . This yields 𝑑𝑧 = 0 because 〈−𝑑𝜀𝑣 〉 = 0. When 𝜂 > 𝑀𝑑 , considering
𝑝𝑙
eq.(B.12) and eq.(B.2)2, the response is switching to dilative (𝑑𝜀𝑣 < 0) and eq.(2.14) is
𝑝𝑙
activated giving 𝑑𝑧 = −𝑐𝑧 |𝑑𝜀𝑣 |(𝑧𝑚𝑎𝑥 + 𝑧) < 0. As long as the stress state is at 𝜂 = 𝛼 + 𝑚,
eq.(B.4) yields 𝐴𝑑 = 𝐴𝑜 because 〈𝑠𝑧〉 = 〈+𝑧〉 = 0. When a loading reversal occurs (i.e 𝑑𝜂 <
0 at 𝜂 = 𝛼 − 𝑚), then 𝑠 = −1 and 〈𝑠𝑧〉 = 〈−𝑧〉 = |𝑧| which gives 𝐴𝑑 = 𝐴𝑜 (1 + |𝑧|). This
increase in 𝐴𝑑 will enhance the dilatancy 𝑑 in eq.(B.12) and result in the reduction of 𝑝 in
eq.(B.13). At the same time, 𝑑𝜂 < 0 will produce plastic strains according to eq.(B.2).

19 | P a g e
 𝑝𝑙
Upon new reversal (i.e 𝑑𝜂 > 0 at 𝜂 = 𝛼 + 𝑚) the response is contractive (𝑑𝜀𝑣 > 0) and
similarly 𝑧 < 0 and eq.(2.14) is once again deactivated yielding 𝑑𝑧 = 0 until the next dilative
response in triaxial extension. Hence, in each increment of 𝑑𝜂 upon reversal which exhibits
dilative behaviour, 𝐴𝑑 is increased from 𝐴𝑜 to 𝐴𝑜 (1 + |𝑧|), then decreased back to 𝐴𝑜 and
eventually increased again due to the evolution of the fabric-dilatancy tensor 𝑧. Note that, 𝐴𝑑
𝑝𝑙
will remain constant for any loading in the contractive region (𝑑𝜀𝑣 > 0).

Figure 2.23: Cyclic undrained stress path of a soil element.

Multiaxial Formulation

The bounding surface model implemented by (Dafalias & Manzari, 2004) can capture loading
conditions in triaxial compression and extension in which the major principal stress direction
is vertical and horizontal respectively. However, it is known that the principal stress direction
within a soil mass is different from element to element. After the construction of an
embankment (Figure 2.24), the major principal stress direction (𝜎1 ) is varying from vertical

20 | P a g e
(triaxial compression) to horizontal (triaxial extension) along the slip surface. The
aforementioned principal stress rotation is accompanied during an earthquake event as well.

Figure 2.24: Limiting equilibrium stress conditions in an embankment.

Furthermore, most sands have anisotropic strength characteristics (Saada, 1988) also known
as stress-induced anisotropy. Under three-dimensional stress conditions, this type of
anisotropy is dependent on many factors such as: the major (𝜎1 ), intermediate (𝜎2 ) and
minor (𝜎3 ) principal stresses, the inclination of the major principal stress direction (𝜎1 ) to the
𝜎1 +𝜎2 +𝜎3
vertical, the deviator stress (𝜎1 − 𝜎3 ) and the mean normal stress ( ) (Uthayakumar
3
& Vaid, 1998).

In order to capture the three-dimensional stress conditions, the model is generalized in

multiaxial stress space (Dafalias & Manzari, 2004). In this formulation, the stress-ratio is
generalized by the deviatoric stress ratio.

2.7 PM4Sand Constitutive Model

2.7.1 Model Description

The PM4Sand model is a simplification of the multiaxial formulation of the sand plasticity
model developed by (Dafalias & Manzari, 2004) and is implemented in plain-strain conditions
(2D model). The surfaces (Figure 2.25) are simplified by removing the Lode angle dependency
in such a way that the same friction angles are used for both compression and extension.
Consequently, the model has some limitations and is only applicable to plane-strain problems
as the various relationships are implemented in terms of in-plane stress only.

2.7.2 Critical State Soil Mechanics Framework

The PM4Sand model is implemented according to critical state soil mechanics framework
(Bolton, 1986). In this formulation, the relative state parameter 𝜉𝑅 (Boulanger, 2003a) is used
instead of the parameter Ψ (Been & Jefferies, 1985). The 𝜉𝑅 is the state parameter 𝛹
(eq.(B.6)) normalized by the difference between the maximum void ratio (𝑒𝑚𝑎𝑥 ) and the

21 | P a g e
minimum void ratio (𝑒𝑚𝑖𝑛 ) that are used to define relative density (𝐷𝑅 ). This provides
improved correlation regarding the shearing behaviour of sands (Konrad, 1988).

Therefore, the critical state line is ‘empirically’ determined according to the following
equation:

𝑅
𝐷𝑅,𝐶𝑆 = (2.16)
𝑝’
𝑄 − 𝑙𝑛 (100 )
𝑝’𝐴

The relative state parameter 𝜉𝑅 is simply defined as the difference between the relative
density at critical state (𝐷𝑅,𝐶𝑆 ) and the current relative density (𝐷𝑅 ) as:

𝜉𝑅 = 𝐷𝑅,𝐶𝑆 − 𝐷𝑅 (2.17)

Regarding eq.(2.16) and eq.(2.17): 𝑝’ is the current mean effective stress and 𝑝𝐴 is the
atmospheric pressure. The parameters 𝑄 and 𝑅 were shown by (Bolton, 1986) to be about
10 and 1.0 respectively, for quartzitic sands. The parameter 𝑄 defines the mean stress level
at which the CSL shifts sharply downwards due to considerable particle crushing (Boulanger,
2003a). By increasing the value of 𝑅 to 1.5 a better approximation is provided for Direct
Simple Shear (DSS) tests (Boulanger & Ziotopoulou, 2015). An example of the critical state
line in the DR-p plane with the parameters 𝑄 = 10 and 𝑅 = 1.5 and the effect of changes in
𝑄 and 𝑅 on the critical state line are illustrated in Figure 2.25 and Figure 2.26 correspondingly.

Figure 2.25: Illustration of the 𝐶𝑆𝐿 in the 𝐷𝑅 − 𝑝′ space for 𝑄 = 10 and 𝑅 = 1.5.

22 | P a g e
Figure 2.26: influence of 𝑄 and 𝑅 on the position of the 𝐶𝑆𝐿.

2.7.3 Fabric Effects

The fabric-dilatancy tensor proposed by (Dafalias & Manzari, 2004) in eq.(2.14) is modified for
the present model (eq.C.15-C.16). This modification was included so as the fabric tensor to
evolve with plastic shear strain rather than plastic volumetric strain that occurs during dilation
only. Moreover, the influence of fabric history and the cumulative fabric term are included
in the present formulation.

The rate of evolution of fabric tensor in eq.(C.15) is decreasing with increasing values of the
cumulative fabric term in eq.(C.16) which disables the undrained cyclic stress-strain response
to lock-up into a repeating stress-strain loop and enables the progressive accumulation of
shear strains (Boulanger & Ziotopoulou, 2015).

2.7.4 Model Parameters

The model parameters are grouped into two categories:

 a primary set of parameters (𝐷𝑅𝑜 , 𝐺𝑜 , ℎ𝑝𝑜 ) that are most important for model
calibration
 a secondary set of parameters that may be modified from the recommended default
values in special circumstances

23 | P a g e
Primary Model Parameters

Apparent relative density(𝐷𝑅𝑜 ): This parameter accounts for the dilatancy and stress-strain
responses of the model. The input value of 𝐷𝑅𝑜 determines the tendency of the soil for
contraction or dilation.

The apparent relative density defines initial bounding (𝑀𝑏 ) and dilatancy (𝑀𝑑 ) surfaces in
relation to critical (𝑀) surface through the relative state parameter (𝜉𝑅 ) according to eq.(2.16)
and eq.(2.17). The distance between 𝑀𝑏 and 𝑀𝑑 with respect to 𝑀 is increasing for higher
values of 𝐷𝑅 (larger 𝑀𝑏 and lower 𝑀𝑑 ) which corresponds to a more dilative response.

It is important to note, that the input value of 𝐷𝑅𝑜 is referred as an ‘’apparent relative density’’
rather than a strict measure and can be adjusted as part of the calibration process. The term
‘’apparent’’ is used because there may be situations where the user need to modify 𝐷𝑅 in
order to improve the calibration according to relationships or test data.

Shear modulus coefficient 𝐺𝑜 : This parameter controls the elastic shear modulus 𝐺 for small
strains according to:

𝑝
𝐺 = 𝐺𝑜,𝑃𝑀4 = 𝐺𝑜 𝑝𝐴 √ (2.18)
𝑝𝐴

Contraction rate parameter ℎ𝑝𝑜 : This parameter controls the evolution of plastic volumetric
strains during contraction. During calibration its value can be adjusted to match specific Cyclic
Resistance Ratio (𝐶𝑅𝑅) values from in-situ tests (𝑆𝑃𝑇, 𝐶𝑃𝑇, 𝑣𝑠 ) based on liquefaction
triggering correlations (Idriss & Boulanger, 2008) or laboratory tests. For example, for in-situ
tests, this parameter can be calibrated to match the 𝐶𝑅𝑅 corresponding to 𝛾 = 3% in direct
simple shear loading at 𝑁𝑐 = 15 loading cycles over a range of relative densities (eq.(C.45-
C.46)) for an earthquake of 7.5 magnitude.

Secondary Model Parameters

The secondary set of parameters can be found in Appendix C.8.

24 | P a g e
Chapter 3: Validation of the PM4Sand Model for Cyclic Loading

3.1 Introduction

The PM4Sand is a constitutive model which is able to simulate successfully the undrained
behaviour of sands during earthquake loading, including the generation of pore pressures and
liquefaction phenomena. What makes this model very attractive for practical applications is
that there are mainly three primary model parameters to be determined in the calibration
process: the apparent relative density 𝐷𝑅𝑜 , the shear modulus coefficient 𝐺𝑜 and the
contraction rate parameter ℎ𝑝𝑜 while the secondary parameters are suggested to be used
with default values. However, the user can modify the secondary parameters in the
calibration process for particular loading conditions.

The response of the model varies for different cyclic stress levels and in-situ conditions thus,
it has to be examined and compared to what has been observed in real practice. Therefore,
the PM4Sand model, which is implemented in Plaxis 2D software, is investigated at element
level to evaluate whether it can reproduce similar responses with respect to what has been
observed in laboratory tests.

For the evaluation process, Cyclic Direct Simple Shear (𝐶𝐷𝑆𝑆) tests performed on Ottawa F-
65 Sand are analysed. Then, the PM4Sand model is calibrated according to the cyclic strength
curves obtained from these tests and the responses for different loading conditions and
confinement levels are assessed. Moreover, the influence of the parameters in the response
of the model is examined by performing a parametric assessment analysis.

3.2 Data Interpretation

The validation of the PM4Sand model is investigated by analysing undrained Cyclic Direct
Simple Shear (𝐶𝐷𝑆𝑆) tests performed on Ottawa F-65 Sand
(https://datacenterhub.org/resources/ottawa_f_65). The evaluation of the model’s
performance on the aforementioned type of sand has already been investigated by (Bastidas,
2016) using single-element numerical simulations in FLAC 2D. In this study, the validation is
accomplished using element tests in Plaxis software.

The Ottawa F-65 sand is classified as a white grained silica sand with rounded grains and a
quartz content of 99.7%. Its index properties are given in Table 3.1. The samples are prepared
by the dry funnel deposition method and the tests were performed in a GEOTAC Direct Simple
Shear apparatus (Bastidas, 2016).

Index properties
Specific gravity (Gs) 2,65
Medium grain size (D50, mm) 0,20
Maximum void ratio (e max) 0,83
Minimum void ratio (e min) 0,51
Critical state friction angle (φ'cv ) 33,00
Table 3.1: Index properties of Ottawa F-65 Sand

25 | P a g e
The tests were performed for loose samples normally consolidated under vertical effective
stresses of 50𝑘𝑃𝑎, 100𝑘𝑃𝑎 and 400kPa. The samples are loaded under various cyclic stress
levels and prepared under different relative densities (Table 3.2). For the sake of convenience,
the average values of relative densities over the three ranges is adopted (40%, 40% and
45%). Then, the cyclic strength curves for each confining level can be constructed from the
different combinations of Cyclic Stress Ratio (𝐶𝑆𝑅) and number of cycles (𝑁) to reach a 3%
peak shear strain in direct simple shear loading (Figure 3.2). The exponential relationships
(𝐶𝑆𝑅 = 𝑎𝑁 −𝑏 ) between the 𝐶𝑆𝑅 and 𝑁 are also depicted.

DR Average DR σ' νc CSR

(%) (%) (kPa)
37-46 40 50 0.080-0.107
34-47 40 100 0.086-0.114
37-48 45 400 0.087-0.115

Table 3.2: Relative density, confining stress, Cyclic Stress Ratio for the CDSS tests.

Figure 3.1: Cyclic strength curves obtained from CDSS tests performed on Ottawa Sand
(Bastidas, 2016) for confining stresses of 50kPa, 100kPa and 400kPa.

At this point an important observation is that normally one would expect a reduction in the
𝐶𝑅𝑅 by increasing the confining stress level under the same (or similar) relative density level
(Chapter 2.3.4). However, according to the test results for Ottawa F-65 Sand, the reverse
effect is observed. The cyclic resistance curves obtained from the laboratory are increasing
with an increasing consolidation stress. In particular, the 𝐶𝑅𝑅 is increasing by 11% from
𝜎’𝜈𝑐 = 50𝑘𝑃𝑎 to 𝜎’𝜈𝑐 = 100𝑘𝑃𝑎 and at 𝜎’𝜈𝑐 = 400𝑘𝑃𝑎 the CRR is the same as at 𝜎’𝜈𝑐 =
100𝑘𝑃𝑎 (Figure 3.1).

26 | P a g e
3.3 Input Model Parameters

The primary model parameters 𝐷𝑅𝑜 , 𝐺𝑜 and ℎ𝑝𝑜 can be calibrated according to laboratory or
in-situ test data while the secondary model parameters can be used with its default values.
The relative density 𝐷𝑅𝑜 accounts for the dilatancy and the stress-strain responses of the
model as it defines the tendency of the soil for contraction or dilation through the relative
state parameter (Chapter 2.8.4). The shear modulus coefficient 𝐺𝑜 controls the elastic shear
modulus 𝐺 at small strains (eq.(2.18)) while the contraction rate parameter ℎ𝑝𝑜 governs the
evolution of plastic volumetric strains and can be calibrated to match specific Cyclic Resistance
Ratio (𝐶𝑅𝑅) values.

In this thesis project, the apparent relative density 𝐷𝑅𝑜 is determined from laboratory data
while the shear modulus coefficient 𝐺𝑜 is estimated from correlation since there is no
information about the dynamic soil properties of Ottawa F-65 Sand (e.g the shear wave
velocity profile). The contraction rate parameter ℎ𝑝𝑜 can be calibrated according to two
criteria. The first one indicates that the onset of liquefaction is captured after reaching a
specific percentage of shear strain (𝑒. 𝑔 𝛾 = 3%) in direct simple shear loading for a certain
number of loading cycles (𝑁𝑐 ). The second one indicates that liquefaction is triggered when
the pore pressure ratio (𝑟𝑢 = 𝛥𝑢/𝜎’𝜈𝑐 ) is reaching a value of about 1.0 for a specified number
of loading cycles (𝑁𝑐 ).

In this study, the ℎ𝑝𝑜 parameter is calibrated with respect to both criteria. Initially, according
to 𝛾 = 3% as implemented by (Bastidas, 2016) in order to evaluate any differences between
the responses from FLAC and Plaxis and with the aim to get a first insight about the overall
response of the model. Secondly, it is calibrated according to 𝑟𝑢 = 1.0 as this criterion is
regarded as the most representative for the prediction of earthquake induced liquefaction in
hydraulic fills (Chapter 5.6).

Apparent Relative Density 𝐃𝐑

This parameter is determined as the value of relative density measured in the laboratory.
Thus, for the specimens confined under vertical effective stresses of 𝜎’𝜈𝑐 = 50𝑘𝑃𝑎, 𝜎’𝜈𝑐 =
100𝑘𝑃𝑎 and 𝜎’𝜈𝑐 = 400𝑘𝑃𝑎 the relative density is defined as 𝐷𝑅 = 0.40, 𝐷𝑅 = 0.40 and
𝐷𝑅 = 0.45 respectively. The ‘apparent’ value of 𝐷𝑅 means that the aforementioned
quantities are an initial estimation for the calibration and the user can adjust them later on in
order to better approximate the responses of interest.

Shear Modulus Coefficient 𝐆𝐨

Due to the lack of available dynamic soil test data (e.g the shear wave velocity profile) from
which to calibrate 𝐺𝑜 , this parameter can be related over a range of relative densities
according to the following equation which is a combination between eq.(C.45) and eq.(C.49)
as:

𝐺𝑜 = 167√46𝐷𝑅2 + 2.5 (3.1)

27 | P a g e
Contraction rate parameter 𝐡𝐩𝐨

This parameter can be calibrated either according to reaching a specific percentage of shear
strain (𝑒. 𝑔 𝛾 = 3%) in direct simple shear loading for specified number of loading cycles (𝑁𝑐 )
or in order to reach a pore pressure ratio (𝑟𝑢 = 𝛥𝑢/𝜎’𝜈𝑐 ) of about 1.0. The 𝑟𝑢 = 1.0 condition
represents the onset of liquefaction in which shear strains exceed the elastic shear strain
threshold.

The ℎ𝑝𝑜 parameter is initially calibrated according to the 𝛾 = 3% criterion at 𝑁𝑐 = 15 loading

cycles for the CRR curve at an overburden stress of 100kPa. In this way, having an initial
estimate about the set of model parameters, the performance of the PM4Sand model is
evaluated for all confining stress levels (𝜎’𝜈𝑐 = 50𝑘𝑃𝑎, 𝜎’𝜈𝑐 = 100𝑘𝑃𝑎 𝑎𝑛𝑑 𝜎’𝜈𝑐 = 400𝑘𝑃𝑎)
(Figure 3.2). In addition, the ℎ𝑝𝑜 parameter is recalibrated with respect to the pore pressure
ratio (𝑟𝑢 ≈ 1) which is the most relevant criterion for the prediction of earthquake induced
liquefaction in hydraulic fills in real practice.

Plaxis Soil Element Tests

The performance of the PM4Sandmodel for the liquefaction triggering is verified by

reproducing undrained Cyclic Direct Simple Shear (𝐶𝐷𝑆𝑆) tests in the Plaxis SoilTest facility at
element level. The coefficient of earth pressure at rest is defined as: 𝐾𝑜 = 1 − sin(𝜑′ 𝑐𝑣 ) =
1 − sin 33𝑜 = 0.46.

The apparent relative density 𝐷𝑅 for the confining stress levels of 𝜎’𝜈𝑐 = 50𝑘𝑃𝑎, 𝜎’𝜈𝑐 =
100𝑘𝑃𝑎 and 𝜎’𝜈𝑐 = 400𝑘𝑃𝑎 is taken as 𝐷𝑅 = 0.40, 𝐷𝑅 = 0.40 and 𝐷𝑅 = 0.45 respectively
(Table 3.2). The shear modulus coefficient 𝐺𝑜 is calculated according to eq.(3.1) as 𝐺𝑜 = 524,
𝐺𝑜 = 524 and 𝐺𝑜 = 624 correspondingly. The contraction rate parameter ℎ𝑝𝑜 is adjusted
(ℎ𝑝𝑜 = 0.37) to match the 𝐶𝑅𝑅 corresponding to 𝛾 = 3% (𝐶𝑅𝑅 = 0.093) at 𝑁𝑐 = 15 loading
cycles from Plaxis element tests at the confining stress of 𝜎’𝜈𝑐 = 100𝑘𝑃𝑎 (Figure 3.2). In the
study from (Bastidas, 2016) the only difference is that the contraction rate parameter ℎ𝑝𝑜 had
a slightly lower value of 0.33. Then the ability of the model to reproduce the other cyclic
strength curves at different levels of CSR and at different overburden stresses (𝜎’𝜈𝑐 =
50𝑘𝑃𝑎, 𝜎’𝜈𝑐 = 400𝑘𝑃𝑎 ) is evaluated.

28 | P a g e
Figure 3.2: Comparison of cyclic strength curves obtained from CDSS tests performed on
Ottawa Sand to the simulated response from Plaxis PM4Sand model

29 | P a g e
It is observed, that the cyclic strength curves produced by the PM4Sand are steeper than the
experimental responses, especially for the lower confinement of 50𝑘𝑃𝑎 (Figure 3.2).
Consequently, even though the model was calibrated to match a specific liquefaction criterion
(𝛾 = 3% 𝑎𝑡 𝑁𝑐 = 15 ), it is not able to reproduce the same conditions at different 𝐶𝑆𝑅 levels.
In general, the model overestimates the cyclic resistance for small number of loading cycles
and underestimates the cyclic resistance for large number of loading cycles. This effect is
decreased with an increasing overburden stress. The aforementioned conclusion does not
mean that the model is not capable of reproducing the general features of cyclic loading as in
real practice, there is a certain 𝐶𝑅𝑅 − 𝑁 range that is of interest rather than the whole part
of the 𝐶𝑅𝑅 curve. Related to the study from (Bastidas, 2016) the cyclic strength curves
obtained by Plaxis approximated well the corresponding responses from FLAC.

In earthquake design it is important to define which part of the 𝐶𝑅𝑅 − 𝑁 curve (Figure 3.2) is
of high significance. In general, the majority of seismic events are having a magnitude (𝑀𝑤 )
between 6 and 8. Thus, in terms of uniform loading cycles, this range is corresponding to a
range between 6 and 22 cycles (Figure A.21) (Appendix A.2). Thus, from now on, this range
will be examined as one can assume that higher or lower values are not relevant for common
design purposes (Figure 3.2).

The evolution of excess pore water pressures (Figure 3.4) along with the stress-strain
responses (Figure 3.3) produced by PM4Sand are compared to the experimental results for
𝐷𝑅 = 0.40 and 𝜎’𝜈𝑐 = 100𝑘𝑃𝑎 performed at a 𝐶𝑆𝑅 = 0.086.

The PM4Sand simulations accumulate a lower 𝑟𝑢 than the experimental responses during the
first two loading cycles (𝑁𝑐 = 2) (Figure 3.4). After 𝑁𝑐 = 2, the rate of 𝑟𝑢 accumulation is
similar to the experimental responses until 𝑁𝑐 = 14. After 𝑁𝑐 = 14, the simulations
accumulate 𝑟𝑢 more rapidly and between 𝑁𝑐 = 17 and 𝑁𝑐 = 18, the shear strains exceed the
elastic shear strain threshold with the 𝑟𝑢 being increased from 0.6 to about 1.0. However,
the increase of 𝑟𝑢 is occurring more gradually in the experiments reaching a value of about 1.0
at 𝑁𝑐 = 25.

Figure 3.3 a, b: Simulated and experimental stress-strain responses and stress path

30 | P a g e
Figure 3.4: Simulated and experimental excess pore pressure generation.

Related to the excess pore pressure generation (𝑟𝑢 ), the decrease of effective stresses in the
simulated stress path during the first two loading cycles is more restraint than in the
experiment (Figure 3.3 b) which is translated to a lower contraction per loading cycle.
Between 𝑁𝑐 = 2 and 𝑁𝑐 = 14 soil densification is observed in both simulated and
experimental responses. At the point of elastic shear strain threshold exceedance (𝑁𝑐 = 18),
rapid soil degradation is observed in the simulations which leads to significant stiffness and
strength reduction, as the accumulations of shear strains increase considerably due to cyclic
mobility (Figure 3.6 a). This temporary condition occurs under an isotropic stress state. During
this condition, the samples are having dilative tendencies when loaded leading to an increase
in vertical effective stresses and contractive tendencies when unloaded leading to an decrease
in vertical effective stresses (Figure 3.3-b). On the contrary, at the point where the
experimental responses are reaching an excess pore pressure ratio (𝑟𝑢 ) of about 1.0 (𝑁𝑐 = 25)
the aforementioned soil degradation effect is absent. It takes about 6 more cycles (Figure
3.3a - 𝑁𝑐 = 31) for the specimens to present significant stiffness reduction.

Regarding the aforementioned results, the PM4Sand model well approximated the general
features of cyclic loading in terms of stress-strain responses, stress path and excess pore
pressure generation while better approximated soil behaviour can be obtained after gaining
deeper insight into the model performance. For that purpose, the possibility of reducing the
steepness of the 𝐶𝑅𝑅 − 𝑁 curves is examined in the following section by using different set
of parameters. Moreover, a parametric assessment is performed in order to evaluate the
influence of the model parameters to the soil response and the possibility of better
approximating the 𝑟𝑢 − 𝑁 response (Figure 3.4) is examined by recalibrating the contraction
rate parameter ℎ𝑝𝑜 for different values of primary and secondary model parameters.

31 | P a g e
3.4 Influence of the Primary Model Parameters

3.4.1 Influence of DR

The influence of the apparent relative density 𝐷𝑅 with respect to the 𝐶𝑅𝑅 − 𝑁 response and
to the evolution of excess pore pressures (𝑟𝑢 ) is evaluated by testing seven new set of
parameters (Table 3.2). The first pair includes the variation of 𝐷𝑅 by keeping all the other
parameters constant. In the following two sets, the recalibration of the contraction parameter
ℎ𝑝𝑜 is performed according to the first liquefaction criterion (𝛾 = 3% 𝑎𝑡 𝑁𝑐 = 15 𝑐𝑦𝑐𝑙𝑒𝑠)
while the last three are about the recalibration of the parameter ℎ𝑝𝑜 parameter according to
the second liquefaction criterion (𝑟𝑢 = 1.0 𝑎𝑡 𝑁𝑐 = 25 𝑐𝑦𝑐𝑙𝑒𝑠) at a CSR of 0.086.

Table 3.2: Parametric assessment of 𝑫𝑹 in the model response by using different set of
parameters.

Figure 3.5: Influence of the apparent relative density 𝑫𝑹 on th𝑒 𝐶𝑅𝑅 − 𝑁 curves.

32 | P a g e
It is observed, that the apparent relative density 𝐷𝑅 has a direct influence in the 𝐶𝑅𝑅 − 𝑁
response (Figure 3.5). Without recalibration of ℎ𝑝𝑜 , by using lower values of 𝐷𝑅 (0.35) the
cyclic strength curve is moving downwards in a parallel manner compared to the curve at
𝐷𝑅 = 0.40 by decreasing the liquefaction resistance by 13% between 6 and 22 uniform
loading cycles. On the other hand, using higher values of 𝐷𝑅 (0.45) the cyclic strength curve
is moving upwards in a parallel manner compared to the curve at 𝐷𝑅 = 0.40 by increasing the
liquefaction resistance by 13% in the range between 6 and 22 uniform loading cycles.
Moreover, by recalibrating the contraction parameter ℎ𝑝𝑜 and using a lower value of 𝐷𝑅 , the
steepness of the 𝐶𝑅𝑅 − 𝑁 curve is slightly reduced by 2% at 𝑁 = 6 whereas a higher value
of 𝐷𝑅 leads to an even steeper curve (3% increase at 𝑁 = 6). The aforementioned
observation is really important as the apparent relative density 𝐷𝑅 , as an input parameter,
has a crucial effect in the liquefaction resistance for the design of hydraulic fills under
earthquake loading over the whole range of interest (𝑁 = 6 − 22 uniform loading cycles).

The influence of the apparent relative density on the model response can be explained by
eq.(C.2-C.4) (Appendix C). It is clear that, the 𝐷𝑅 defines initial bounding and dilatancy
surfaces through the relative state parameter 𝜉𝑅 . Therefore, for a given cyclic loading level,
the looser the soil is the more rapidly the bounding surface is approaching the critical surface
where plastic deformations are evolving under constant mean effective stresses.

Figure 3.6: Influence of the apparent relative density 𝑫𝑹 on the generation of excess pore
pressures.

Related to the excess pore pressure generation (𝑟𝑢 ), the simulations accumulate similar 𝑟𝑢
during the first five loading cycles (𝑁𝑐 = 5) for all values of 𝐷𝑅 (Figure 3.6). After this stage,
the impact of 𝐷𝑅 in the soil behaviour until a 𝑟𝑢 of around 1.0 is reached is clear. Without
recalibration of ℎ𝑝𝑜 , the simulations show stiffer response for an increasing value of 𝐷𝑅 (0.45)
as more cycles are required to reach that state (𝑁𝑐 = 31) (critical state is reached slower).
Conversely, the simulations show softer response for a decreasing value of 𝐷𝑅 (0.35), as less
number of cycles are required (𝑁𝑐 = 11) (critical state is reached faster). It should be noted,
that the effect of the recalibration of ℎ𝑝𝑜 parameter to the generation of excess pore
pressures is almost negligible as the response remains the same.

33 | P a g e
The influence of the apparent relative density 𝐷𝑅 on the evolution of excess pore pressures
can be explained as follows: using lower values of 𝐷𝑅 will define lower bounding surface (𝑀𝑏 ).
This will result in lower elastic shear modulus 𝐺 (eq.(C.20)) through the stress-ratio effects
(eq.(C.21)). Consequently, the bulk modulus 𝐾 is going to be lower which eventually restricts
the elastic volumetric strain increment (eq.(C.18)) and thus the evolution of excess pore
pressures. As a result, the looser the sand is the faster the generation of excess pore pressures
is accomplished

Moreover, the contraction rate parameter ℎ𝑝𝑜 is recalibrated according to the second
liquefaction criterion (𝑟𝑢 = 1.0 𝑎𝑡 𝑁𝑐 = 25 𝑐𝑦𝑐𝑙𝑒𝑠) for three different values of 𝐷𝑅 at a CSR
of 0.086 (Figure 3.7). This was done in order to investigate the possibilities of approaching
more accurately the 𝑟𝑢 − 𝑁 curves obtained from the laboratory.

Figure 3.7: Influence of the recalibration of the contraction rate parameter ℎ𝑝𝑜 regarding the
𝑟𝑢 ≈ 1 criterion at 𝑁𝑐 = 25 cycles at a CSR of 0.086.

It is observed that the recalibration of the ℎ𝑝𝑜 parameter led to a precise approximation
concerning the 𝑟𝑢 ≈ 1 state at a CSR of 0.086 (Figure 3.7) by varying apparent relative density
𝐷𝑅 . One important observation is, that the generation of excess pore pressures until 𝑟𝑢 ≈ 1
from PM4Sand model is almost the same for all three different set of parameters.

3.4.2 Influence of Go

The influence of the shear modulus coefficient 𝐺𝑜 with respect to the 𝐶𝑅𝑅 − 𝑁 response and
to the evolution of excess pore pressures (𝑟𝑢 ) is evaluated by testing four new set of
parameters (Table 3.3). The first pair includes the variation of 𝐺𝑜 by keeping all the other
parameters constant. In the following two sets, the recalibration of the contraction parameter
ℎ𝑝𝑜 is performed according to the first liquefaction criterion (𝛾 = 3% 𝑎𝑡 𝑁𝑐 = 15 𝑐𝑦𝑐𝑙𝑒𝑠).

34 | P a g e
Table 3.3: Parametric assessment of 𝑮𝒐 in the model response by using different set of
parameters.

Figure 3.8: Influence of the shear modulus coefficient 𝑮𝒐 on the 𝐶𝑅𝑅 − 𝑁 curves.

It is observed, that the influence of the shear modulus coefficient 𝐺𝑜 in the 𝐶𝑅𝑅 − 𝑁 response
(Figure 3.8) is of less significance, compared to the corresponding effect of apparent relative
density 𝐷𝑅 (Figure 3.8). Without recalibration of ℎ𝑝𝑜 , by using lower values of 𝐺𝑜 (100) the
cyclic strength curve is moving downwards with respect to the curve at 𝐺𝑜 = 524 by
decreasing the liquefaction resistance by 6% at 𝑁 = 6 and by 2% at 𝑁 = 22 uniform loading
cycles. Consequently, the effect of 𝐺𝑜 in cyclic strength for large number of loading cycles is
becoming trivial. On the other hand, using higher values of 𝐺𝑜 (1000) the cyclic strength
curve is moving upwards compared to the curve at 𝐺𝑜 = 524 by increasing the liquefaction
resistance by 11% at 𝑁 = 6 and by 7% at 𝑁 = 22 uniform loading cycles. Moreover, by
recalibrating the contraction parameter ℎ𝑝𝑜 and using a lower value of 𝐺𝑜 , the steepness of

35 | P a g e
the CRR-N curve is slightly reduced by 4% at 𝑁 = 6 whereas a higher value of 𝐺𝑜 leads to an
even steeper curve (7% increase at 𝑁 = 6).

Figure 3.9: Influence of the shear modulus coefficient 𝑮𝒐 on the generation of excess pore
pressures.

Figure 3.10: Influence of the shear modulus coefficient 𝑮𝒐 on the generation of excess pore
pressures.

Related to the excess pore pressure generation (𝑟𝑢 ), it is shown, that without recalibration of
the contraction rate parameter ℎ𝑝𝑜 , the variation of 𝐺𝑜 is not affecting the generation of
excess pore pressures at all until an 𝑟𝑢 of around 1.0 is reached (Figure 3.9). For that reason,
the recalibration of ℎ𝑝𝑜 according to the second liquefaction criterion (𝑟𝑢 = 1.0) is not
investigated for the shear modulus coefficient 𝐺𝑜 . However, the fact that there is no influence
of 𝐺𝑜 throughout the whole response before 𝑟𝑢 = 1.0 seems odd in the following sense: as
explained in section 3.5.1 the evolution of excess pore pressures is governed by the restriction
of the elastic volumetric strain tensor to evolve which depends on the bulk modulus 𝐾
(through the elastic shear modulus 𝐺). Due to the fact that the shear modulus coefficient 𝐺𝑜
is directly proportional to the elastic shear modulus 𝐺 one would expect considerable effects
on the evolution of 𝑟𝑢 with the variation in 𝐺𝑜 .

36 | P a g e
On the other hand, the effect of the variation of shear modulus coefficient 𝐺𝑜 is significant in
the rate of shear strain accumulation after 𝑟𝑢 = 1.0 (𝑁𝑐 = 18) (Figure 3.9). It is clear, that a
higher value of 𝐺𝑜 leads to significant increase in the accumulation of shear strains per loading
cycle after an 𝑟𝑢 of around 1.0 is reached compared to a lower 𝐺𝑜 in which the shear strain
accumulation is more gradual. After an 𝑟𝑢 = 1.0 (𝑁𝑐 = 18) a shear strain of approximately
3% is reached after 1 cycle for 𝐺𝑜 = 100 (𝑁𝑐 = 19) and after 5 cycles for 𝐺𝑜 = 1000 (𝑁𝑐 =
23). This effect can be explained as follows: the factor that regulates whether the
accumulation of shear strains after an 𝑟𝑢 of around 1.0 is reached is accomplished in a gradual
or progressive manner is the fabric tensor (eq.C.15)). This tensor is dependent on the plastic
shear strain increment (𝑑𝒆𝑝𝑙 ) (eq.(C.24) through the ratio between plastic volumetric strain
𝑝𝑙
increment (𝑑𝜀𝑣 ) and dilatancy (𝐷) (eq.(C.15)). From eq.(C.24 to C.26) the(𝑑𝒆𝑝𝑙 ) is directly
proportional to the loading index 𝑳 which is inversely proportional to the plastic modulus 𝐾𝑝
(eq.(C.28)). The latter is directly proportional to 𝐺𝑜 through the elastic shear modulus 𝐺
(eq.(C.31)). As a results higher values of 𝐺𝑜 lead to a more progressive shear strain
accumulation after an 𝑟𝑢 of around 1.0 is reached.

Moreover, it is shown, that the recalibration of ℎ𝑝𝑜 by using higher values of 𝐺𝑜 (1000) led
to slower generation of excess pore pressures (𝑁𝑐 = 15) while the effect of recalibration of
ℎ𝑝𝑜 by using lower values of 𝐺𝑜 (100) on the evolution of 𝑟𝑢 is almost negligible (𝑁𝑐 = 19)
(Figure 3.10).

3.5 Influence of the Secondary Model Parameters

3.5.1 Influence of nb

The bounding surface parameter 𝑛𝑏 is used to define the initial bounding surface (eq.(C.3))
and consequently affects the distance between the bounding and critical state surface.
Therefore, it defines the rate with which the bounding surface is approaching the critical
surface. Moreover, the 𝑛𝑏 affects the dilatancy 𝐷 (eq.C.34 to eq.C.44) which governs the
evolution of plastic volumetric strains (eq.C.23)).

The influence of the bounding surface parameter 𝑛𝑏 with respect to the 𝐶𝑅𝑅 − 𝑁 response
and to the evolution of excess pore pressures (𝑟𝑢 ) is evaluated by testing six new set of
parameters (Table 3.4). The first pair includes the variation of 𝑛𝑏 by keeping all the other
parameters constant. In the following set, the recalibration of the contraction parameter ℎ𝑝𝑜
is performed according to the first liquefaction criterion (𝛾 = 3% 𝑎𝑡 𝑁𝑐 = 15 𝑐𝑦𝑐𝑙𝑒𝑠) while
the last three are about the recalibration of the parameter ℎ𝑝𝑜 parameter according to the
second liquefaction criterion (𝑟𝑢 = 1.0 𝑎𝑡 𝑁𝑐 = 25 𝑐𝑦𝑐𝑙𝑒𝑠) at a CSR of 0.086

37 | P a g e
Table 3.4: Parametric assessment of 𝒏𝒃 in the model response by using different set of
parameters.

Figure 3.11: Influence of the bounding surface parameter 𝒏𝒃 on the 𝐶𝑅𝑅 − 𝑁 curves.

It is observed, that bounding surface parameter 𝑛𝑏 do not affect the 𝐶𝑅𝑅 − 𝑁 response
besides when using 𝑛𝑏 = 0.10 without the recalibration of ℎ𝑝𝑜 (Figure 3.11). In that case, the
cyclic strength curve is moving upwards in a parallel manner compared to the curve at 𝑛𝑏 =
0.50 by increasing the liquefaction resistance by around 7.5% between 6 and 22 uniform
loading cycles. However, in all other cases, the effect of 𝑛𝑏 in cyclic strength throughout the
whole range of interest is becoming trivial.

38 | P a g e
Figure 3.12: Influence of the bounding surface parameter 𝒏𝒃 on the generation of excess pore
pressures.

Related to the excess pore pressure generation (𝑟𝑢 ), the simulations accumulate similar 𝑟𝑢
during the first twelve loading cycles (𝑁𝑐 = 12) besides when using 𝑛𝑏 = 0.10 without the
recalibration of ℎ𝑝𝑜 (Figure 3.12). After this stage, the simulations show stiffer response for
a decreasing value of 𝑛𝑏 especially for the lowest (0.10) as more cycles are required to reach
an 𝑟𝑢 of around 1.0 (𝑁𝑐 = 29) (critical state is reached slower). It should be noted, that the
recalibration of ℎ𝑝𝑜 parameter do not have a big impact to the generation of excess pore
pressures as a slight increase in loading cycles is observed ((𝑁𝑐 = 20).

Moreover, the contraction rate parameter ℎ𝑝𝑜 is recalibrated according to the second
liquefaction criterion (𝑟𝑢 = 1.0 𝑎𝑡 𝑁𝑐 = 25 𝑐𝑦𝑐𝑙𝑒𝑠) for two different values of 𝑛𝑏 at a CSR of
0.086 (Figure 3.13). This was done in order to investigate the possibilities of approaching more
accurately the 𝑟𝑢 − 𝑁 curves obtained from the laboratory.

Figure 3.13: Influence of the recalibration of the contraction rate parameter ℎ𝑝𝑜 regarding
the 𝑟𝑢 ≈ 1 criterion at 𝑁𝑐 = 25 cycles at a CSR of 0.086.

39 | P a g e
It is observed, that the recalibration of the ℎ𝑝𝑜 parameter led to a precise approximation
concerning the 𝑟𝑢 ≈ 1 state at a CSR of 0.086 (Figure 3.10) by varying the bounding surface
parameter 𝑛𝑏 . One important observation is, that the generation of excess pore pressures
until 𝑟𝑢 ≈ 1 from PM4Sand model is almost the same for all three different set of parameters.

3.5.2 Influence of R

In PM4Sand model, the critical state line (CSL) is defined with respect to two parameters: the
critical state parameter 𝑅 and 𝑄 as proposed by (Bolton, 1986). These parameters are
affecting the curvature and position of the CSL (Figure C.1). In this study the influence of the
𝑅 parameter is going to be examined as after parametric assessment the cyclic response of
the model is not considerably affected by the parameter Q.

The influence of the critical state parameter 𝑅 with respect to the 𝐶𝑅𝑅 − 𝑁 response and to
the evolution of excess pore pressures (𝑟𝑢 ) is evaluated by testing four new set of parameters
(Table 3.5). The first pair includes the variation of 𝑅 by keeping all the other parameters
constant. In the following two sets, the recalibration of the contraction parameter ℎ𝑝𝑜 is
performed according to the first liquefaction criterion (𝛾 = 3% 𝑎𝑡 𝑁𝑐 = 15 𝑐𝑦𝑐𝑙𝑒𝑠) while the
last three are about the recalibration of the parameter ℎ𝑝𝑜 parameter according to the second
liquefaction criterion (𝑟𝑢 = 1.0 𝑎𝑡 𝑁𝑐 = 25 𝑐𝑦𝑐𝑙𝑒𝑠) at a CSR of 0.086.

Table 3.5: Parametric assessment of 𝑹 in the model response by using different set of
parameters

40 | P a g e
Figure 3.14: Influence of the bounding surface parameter 𝑹 on the 𝐶𝑅𝑅 − 𝑁 curves.

It is observed that the critical state parameter 𝑅 has a direct influence in the 𝐶𝑅𝑅 − 𝑁
response (Figure 3.14). Without recalibration of ℎ𝑝𝑜 , by using lower values of 𝑅 (1.0) the
cyclic strength curve is moving upwards in a parallel manner compared to the curve at 𝑅 =
1.5 by increasing the liquefaction by 20% between 6 and 22 uniform loading cycles. On the
other hand, using higher values of 𝑅 (2.0) the cyclic strength curve is moving downwards in
a parallel manner compared to the curve at 𝑅 = 1.5 by decreasing the liquefaction resistance
by 16% in the range between 6 and 22 uniform loading cycles. Moreover, by recalibrating
the contraction parameter ℎ𝑝𝑜 for different values of 𝑅 the steepness of the 𝐶𝑅𝑅 − 𝑁 curves
are not modified at all.

The influence of the critical state parameter 𝑅 on the model response can be explained by
eq.(C.2-C.4) and Figure C.1 (Appendix C). By increasing the parameter 𝑅 the critical state line
is moving downwards. This will lead to a decrease in 𝐷𝑅,𝐶𝑆 which gives a lower relative state
parameter 𝜉𝑅 enabling the model to approach the 𝐶𝑆𝐿 faster. This effect is comparable to
the apparent relative density as both parameters are related to the relative state parameter.

41 | P a g e
Figure 3.15: Influence of the critical state parameter 𝑹 on the generation of excess pore
pressures.

Related to the excess pore pressure generation (𝑟𝑢 ) and without recalibration of ℎ𝑝𝑜 , the
simulations show stiffer response for a decreasing value of 𝑅 (1.0) (Figure 3.15) as more cycles
are required to reach that state (𝑁𝑐 = 41) (critical state is reached slower). Conversely, the
simulations show softer response for an increasing value of 𝑅 (2.0), as less number of cycles
are required (𝑁𝑐 = 9) (critical state is reached faster). It should be noted, that the effect of
the recalibration of ℎ𝑝𝑜 parameter to the generation of excess pore pressures is almost
negligible as the response remains the same.

Figure 3.16: Influence of the recalibration of the contraction rate parameter ℎ𝑝𝑜 regarding the
𝑟𝑢 ≈ 1 criterion at 𝑁𝑐 = 25 cycles at a 𝐶𝑆𝑅 of 0.086.

It is observed, that the recalibration of the ℎ𝑝𝑜 parameter led to a precise approximation
concerning the 𝑟𝑢 ≈ 1 state at a CSR of 0.086 (Figure 3.16) by varying the critical state
parameter 𝑅. One important observation is, that the generation of excess pore pressures
until 𝑟𝑢 ≈ 1 from PM4Sand model is almost the same for all three different set of parameters.

42 | P a g e
3.6 Evaluation of the Final Set of Parameters

Before evaluating the final set of parameters it is important to define which 𝐶𝑅𝑅 − 𝑁 curves
obtained from Ottawa F-65 Sand are of importance for the design of a typical hydraulic fill. In
real practise the fills are placed over a 10 − 15𝑚 height. This means that the effective
confining stress levels can reach 150𝑘𝑃𝑎 at maximum. Consequently, the first two graphs in
Figure 3.2 are of primary focus for this study.

After concluding that the PM4Sand model produced far steeper cyclic strength curves than
the experimental responses for 𝜎’𝜈𝑐 = 50𝑘𝑃𝑎, a new calibration process is employed. A
lower dilatancy is needed to better approximate the 𝐶𝑅𝑅 − 𝑁 curves for the lower confining
stress of 50𝑘𝑃𝑎. This process included reduced values for the bounding surface parameter
𝑛𝑏 (0.25 & 0.10 compared to 0.50 at default) and the recalibration of the contraction
parameter ℎ𝑝𝑜 according to the first liquefaction criterion (𝛾 = 3% 𝑎𝑡 𝑁𝑐 = 15 𝑐𝑦𝑐𝑙𝑒𝑠).
From Figure 3.17, it is clear, that lower values of 𝑛𝑏 reduced significantly the steepness of the
curves and gave a much better approximation.

Figure 3.17: Recalibration of the ℎ𝑝𝑜 parameter by using lower values for the bounding surface
parameter 𝑛𝑏 to better approximate the curves for 𝜎’𝜈𝑐 = 50𝑘𝑃𝑎.

43 | P a g e
Figure 3.18: Influence of the apparent relative density 𝑫𝑹 on the generation of excess pore
pressures.

Related to the excess pore pressure generation (𝑟𝑢 ), the recalibration of ℎ𝑝𝑜 improved the
𝑟𝑢 − 𝑁 response (Figure 3.18) at a CSR of 0.080. Using lower values for the bounding surface
parameter 𝑛𝑏 led to a softer soil behavior by reducing the number of cycles required to reach
an 𝑟𝑢 of around 1.0 from 𝑁 = 31 to 𝑁 = 17 for 𝑛𝑏 = 0.25 and to 𝑁 = 18 for 𝑛𝑏 = 0.10.

Regarding the parameter selection there are two different guidelines from which one can
select the final set for the design under earthquake induced liquefaction: the 𝐶𝑅𝑅 − 𝑁
response and the 𝑟𝑢 − 𝑁 response. The first one gives a clear insight about the overall
response of the model and is defined according to criteria that refer to reaching specific
percentage of shear strains at a given number of loading cycles. The second one is based on
the excess pore pressure evolution in which after a certain value is reached, liquefaction is
defined. In this study the most representative guideline is reaching a pore pressure ratio of
approximately unity (𝑟𝑢 ≈ 1) as it the most common criterion used in real practise for the
determination of liquefaction triggering in hydraulic fill structures. In this study, the
parameter selection is going to be defined for both cases as after the parametric assessment
a clear insight about the model performance is acquired.

Regarding the samples consolidated at 𝜎’𝜈𝑐 = 50𝑘𝑃𝑎, the same set of parameters are
selected for both the 𝐶𝑅𝑅 − 𝑁 response and the evolution of excess pore pressure as there
was a decent match for both cases (Table 3.6). About the specimens consolidated at 𝜎’𝜈𝑐 =
100𝑘𝑃𝑎, the 𝐶𝑅𝑅 − 𝑁 response was well approximated by using the initial set of parameters.
However, a lower value for 𝐺𝑜 is selected to decrease the overestimation at 𝑁 = 6 uniform
cycles. With respect to the 𝑟𝑢 ≈ 1 criterion it was concluded, that the model can approximate
the 𝑟𝑢 − 𝑁 responses with the recalibration of ℎ𝑝𝑜 parameter and varying any of the primary
and secondary model parameters. For practical reasons, the concept of reducing the
bounding surface parameter 𝑛𝑏 by recalibrating ℎ𝑝𝑜 for 𝑟𝑢 ≈ 1 is selected as it has minor
influence in the 𝐶𝑅𝑅 − 𝑁 response.

44 | P a g e
Table 3.6: Parametric assessment of 𝑹 in the model response by using different set of
parameters

3.7 Conclusions

The comparison of the experimental from the Ottawa F-65 Sand to the simulated responses
from PM4Sand model gave significant insight about the capabilities and limitation of the
PM4Sand model:

 The PM4Sand model well approximates the general features of cyclic loading
behaviour
 The PM4Sand model produces 𝐶𝑅𝑅 slopes which are consistent with typical cyclic
strength curves from literature
 The PM4Sand model predicts steeper cyclic strength curves than in the experimental
responses
 This steepness is decreased with an increasing confinement level which is means that
there is a dependence of the 𝐶𝑅𝑅 on the overburden stress level. This is consistent
with (Boulanger & Ziotopoulou, 2017) according to the 𝐾𝜎 behaviour of the model
 In general, the model overpredicts the CRR for small number of loading cycles and
underpredicts the 𝐶𝑅𝑅 for large number of loading cycles
 The apparent relative density 𝐷𝑅 , as a primary input parameter, has the most crucial
effect in the cyclic strength and also significant influence in the generation of excess
pore pressures
 The shear modulus coefficient 𝐺𝑜 is having minor influence in the cyclic strength and
is not affecting the generation of excess pore pressures at all until an 𝑟𝑢 of around 1.0
is reached
 The overestimation of 𝐶𝑅𝑅 at the lower confinement level (𝜎’𝜈𝑐 = 50𝑘𝑃𝑎) can be
reduced by using a lower value for the bounding surface parameter 𝑛𝑏 and then
recalibrating the contraction rate parameter ℎ𝑝𝑜 against the cyclic strength curve
 The generation of excess pore pressures in the experiments the experiments showed
a relatively large buildup during first two cycles while the model carry difficulties on
coping with that effect

45 | P a g e
Chapter 4: Verification of the PM4Sand Model for Static Loading

4.1 Introduction

As previously described in Chapter 3, the PM4Sand model is designed in order to capture

liquefaction phenomena which occur during dynamic loading arising from an earthquake
event. Moreover, in literature it is indicated, that the soil structure can also fail due to
monotonic undrained loading. This type of failure is triggered when the static shear stress of
a soil body required to sustain equilibrium exceeds the shear strength of the soil in its liquefied
state (Chapter 2.1.2). This is most commonly known as static liquefaction or flow liquefaction
and can lead to disastrous effects for the structure. The aforementioned background
triggered the question on whether the PM4Sand model can replicate static liquefaction
phenomena, despite the fact, that it is preliminary designed to model liquefaction arising from
earthquake loading.

For the evaluation process, undrained Direct Simple Shear (𝐷𝑆𝑆) tests performed on Ottawa
F-65 Sand are analysed. Then, the PM4Sand model is calibrated according to the monotonic
shear strength curves obtained from these tests and the responses for different conditions
confinement levels are assessed. Moreover, the influence of the parameters in the response
of the model is examined by performing a parametric assessment analysis.

4.2 Data Interpretation

The validation of the PM4Sand model for the simulation of static liquefaction is investigated
by analysing undrained Direct Simple Shear (𝐷𝑆𝑆) tests performed on Ottawa F-65 Sand
(https://datacenterhub.org/resources/ottawa_f_65). The model is examined at element level
to evaluate whether it can reproduce similar responses with respect to what has been
observed in laboratory tests for monotonic undrained loading. The evaluation of the model’s
performance for the aforementioned type of sand has already been investigated by (Bastidas,
2016) using single-element numerical simulations in FLAC 2D. In her study, the input model
parameters were used as obtained from the cyclic calibration while in this study a separate
calibration for static loading is implemented. The validation of the PM4Sand model is
accomplished using element tests in Plaxis software. The model is calibrated according to the
monotonic shear strength curves obtained from lab tests and the responses for different
confinement levels are assessed. Moreover, the influence of the parameters in the response
of the model is examined by performing a parametric assessment analysis.

The tests were performed for loose samples normally consolidated under vertical effective
stresses of 100𝑘𝑃𝑎 and 400𝑘𝑃𝑎 with an average relative density of about 31%. The samples
were sheared under a constant height condition which is equivalent to an undrained shearing
condition in the DSS (Dyvik, 1987).

46 | P a g e
The stress-strain responses, stress paths and evolution of excess pore water pressures are
depicted in Figures 4.1-4.3. It can be observed, that the samples are reaching a Quasi-Steady
State (QSS) condition at a shear strain of around 3% and 7% for an overburden stress of
100𝑘𝑃𝑎 and 400𝑘𝑃𝑎 respectively (Figure 4.1). For higher shear strains, the samples show
dilative behaviour as an increase in effective stresses and a decrease in excess pore water
pressures is observed for both confining stresses of 100𝑘𝑃𝑎 and 400𝑘𝑃𝑎 (Figure 4.2-4.3).
The higher confined samples (400𝑘𝑃𝑎) are showing less dilative response compared to the
lower confined (100𝑘𝑃𝑎) which is according to what has been observed in real practise.

Figure 4.1 a,b: Stress-strain responses obtained from monotonic DSS tests performed on
Ottawa Sand for confining stresses of 100kPa and 400kPa (Bastidas, 2016)

Figure 4.2 a,b: Stress paths obtained from monotonic DSS tests performed on Ottawa Sand
for confining stresses of 100kPa and 400kPa (Bastidas, 2016)

Figure 4.3 a,b: Evolution of excess pore water pressures obtained from monotonic DSS tests
performed on Ottawa Sand for confining stresses of 100kPa and 400kPa (Bastidas, 2016)

47 | P a g e
4.3 Input Model Parameters

The primary model parameters 𝐷𝑅 , 𝐺𝑜 and ℎ𝑝𝑜 are calibrated with respect to the
aforementioned laboratory test results while the secondary model parameters are used with
its default values. The relative density 𝐷𝑅 accounts for the dilatancy and the stress-strain
responses of the model as it defines the tendency of the soil for contraction or dilation
through the relative state parameter (Chapter 2.8.4). The shear modulus coefficient 𝐺𝑜
controls the elastic shear modulus 𝐺 at small strains (eq.(2.18)) while the contraction rate
parameter ℎ𝑝𝑜 governs the evolution of plastic volumetric strains.

In the current thesis project, the apparent relative density 𝐷𝑅 is determined as the relative
density measured in the lab while the shear modulus coefficient 𝐺𝑜 is estimated from
correlation. The contraction rate parameter ℎ𝑝𝑜 is calibrated to match the peak monotonic
shear strength at a confining stress of 100𝑘𝑃𝑎. Then the ability of the model to reproduce
the other monotonic responses at different confining stress levels (50𝑘𝑃𝑎, 400𝑘𝑃𝑎) is
evaluated.

Apparent Relative Density 𝐃𝐑

This parameter is determined as the value of relative density measured in the laboratory.
Thus, for the specimens confined under vertical effective stresses of 𝜎’𝜈𝑐 = 100𝑘𝑃𝑎 and
𝜎’𝜈𝑐 = 400𝑘𝑃𝑎 the relative density is defined as 𝐷𝑅 = 0.31. The ‘apparent’ value of 𝐷𝑅
means that the aforementioned quantities are an initial estimation for the calibration and the
user can adjust them later on in order to better approximate the responses of interest.

Shear Modulus Coefficient 𝐆𝐨

This parameter can be related over a range of relative densities according to the following
equation which is a combination between eq.(2.64) and eq.(2.67) as:

𝐺𝑜 = 167√46𝐷𝑅2 + 2.5 (4.1)

Contraction rate parameter 𝐡𝐩𝐨

This parameter is adjusted to approximate the peak monotonic shear strength at a confining
stress of 100𝑘𝑃𝑎.

Plaxis Soil Element Tests

The performance of the model for the liquefaction triggering under static loading is verified
by reproducing undrained Direct Simple Shear (𝐷𝑆𝑆) tests in the Plaxis SoilTest facility at
element level. The coefficient of earth pressure at rest is defined as: 𝐾𝑜 = 1 − sin(𝜑′ 𝑐𝑣 ) =
1 − sin 33𝑜 = 0.46.

The apparent relative density 𝐷𝑅 for the confining stress levels of 𝜎’𝜈𝑐 = 100𝑘𝑃𝑎 and 𝜎’𝜈𝑐 =
400𝑘𝑃𝑎 is defined as 𝐷𝑅 = 0.31. The shear modulus coefficient 𝐺𝑜 is calculated according
to eq.(4.1) as 𝐺𝑜 = 439. The contraction rate parameter ℎ𝑝𝑜 is adjusted (ℎ𝑝𝑜 = 0.15) to

48 | P a g e
approximate the peak monotonic shear strength from Plaxis element tests at the confining
stress of 𝜎’𝜈𝑐 = 100𝑘𝑃𝑎 (Figure 3.5). Then the ability of the model to reproduce the other
monotonic responses at different overburden stresses (𝜎’𝜈𝑐 = 50𝑘𝑃𝑎, 𝜎’𝜈𝑐 = 400𝑘𝑃𝑎 ) is
evaluated.

Figure 4.4 a,b: Simulated and experimental stress-strain responses at overburden stresses of
100kPa and 400kPa for DR=0.31.

Figure 4. 5a,b: Simulated and experimental stress paths at overburden stresses of 100kPa
and 400kPa for DR=0.31.

Figure 4.6 a,b: Simulated and experimental evolution of excess pore water pressures at
overburden stresses of 100kPa and 400kPa for DR=0.31.

It is observed, that the stress-strain responses produced by the PM4Sand are showing higher
dilative response after the peak shear strength is reached while the laboratory responses are
much smoother (Figure 4.4-4.6). This effect is decreased with an increasing overburden

49 | P a g e
stress. In the following section, a parametric assessment is performed to evaluate the
influence of the parameters in the model response and whether the PM4Sand model can
approximate better the experimental responses.

4.5 Influence of the Primary Model Parameters

4.5.1 Influence of DR

The influence of the apparent relative density 𝐷𝑅 in the model response is examined by testing
six new set of parameters (Table 4.1). The first pair includes the variation of 𝐷𝑅 by keeping all
the other parameters constant at 𝜎′𝜈𝑐 = 100𝑘𝑃𝑎. In the following four sets, the recalibration
of the contraction parameter ℎ𝑝𝑜 is performed by using lower values for relative density.

Table 4.1: Parametric assessment of 𝑫𝑹 in the model response by using different set of
parameters.

It is observed, that the initial softening behaviour is slightly decreased and the following
dilative tendency is increased with an increasing relative density 𝐷𝑅 at 𝜎′𝜈𝑐 = 100𝑘𝑃𝑎 (Figure
4.7). Consequently, the effect of relative density in the peak monotonic strength is minor
while its influence is clear primarily in the post-peak response.

Regarding the recalibration of the ℎ𝑝𝑜 parameter, it is observed, that by decreasing 𝐷𝑅 , the
post-peak response is becoming considerably softer by approximating better the
experimental repsonses (Figure 4.8). The simulations are showing ‘limited liquefaction’
behaviour for 𝐷𝑅 = 0.25 and 𝐷𝑅 = 0.28 and ‘liquefaction’ behaviour for 𝐷𝑅 = 0.20 and 𝐷𝑅 =
0.23 (Chapter 2.2.3).

50 | P a g e
Figure 4.7: Influence of the apparent relative density 𝑫𝑹 in the model response at 𝜎′𝜈𝑐 =
100𝑘𝑃𝑎.

Figure 4.8: Influence of the apparent relative density 𝑫𝑹 by recalibrating 𝒉𝒑𝒐 in the model
response at 𝜎′𝜈𝑐 = 100𝑘𝑃𝑎.

51 | P a g e
4.5.2 Influence of Go

The influence of the shear modulus coefficient 𝐺𝑜 in the model response is examined by
testing two new set of parameters (Table 4.2). This pair includes the variation of 𝐺𝑜 by keeping
all the other parameters constant.

Table 4.2: Parametric assessment of 𝑮𝒐 in the model response by using different set of
parameters

It is observed, that the 𝐺𝑜 parameter is having a direct influence on the response at small
strains. In particular, the residual strength is reached more gradually (at higher strain levels)
for a decreased value of shear modulus coefficient (𝐺𝑜 = 100)(Figure 4.9).

Figure 4.9: Influence of the shear modulus coefficient 𝑮𝒐 in the model response at 𝜎′𝜈𝑐 =
100𝑘𝑃𝑎.

52 | P a g e
4.6 Influence of the Secondary Model Parameters

4.6.1 Influence of nb

The influence of the bounding surface parameter 𝑛𝑏 in the model response is examined by
testing two new set of parameters (Table 4.3). This pair includes the variation of 𝑛𝑏 by keeping
all the other parameters constant.

Table 4.3: Parametric assessment of 𝒏𝒃 in the model response by using different set of
parameters

It is observed, that the initial softening behaviour is becoming trivial for all values of bounding
surface parameter (Figure 4.10). The dilative tendency that is followed after the peak shear
strength is reached, is decreased with a decreasing value of 𝑛𝑏 . This can be attributed to the
fact, that for lower values of 𝑛𝑏 (and by considering that the default value of dilatancy surface
parameter 𝑛𝑑 is 0.10) the bounding surface (𝑀𝑏 ) lies closer to the dilatancy surface (𝑀𝑑 )
(Figure 2.22). Consequently, once the stress state crosses 𝑀𝑑 , the 𝑀𝑏 is then being
approached faster. Therefore, using lower values of 𝑛𝑏 leads to a softer post-peak response
which is closer to the experimental results.

53 | P a g e
Figure 4.10: Influence of the bounding surface parameter 𝒏𝒃 in the model response at 𝜎′𝜈𝑐 =
100𝑘𝑃𝑎.

4.6.2 Influence of R

The influence of the critical state parameter 𝑅 in the model response is examined by testing
six new set of parameters (Table 4.4). The first pair includes the variation of 𝑅 by keeping all
the other parameters constant for all different confining stress levels. In the following four
sets, the recalibration of the contraction parameter ℎ𝑝𝑜 is performed by using lower values
for 𝑅 at 𝜎′𝜈𝑐 = 100𝑘𝑃𝑎.

Table 4.4: Parametric assessment of 𝑹 in the model response by using different set of
parameters

It is observed, that the initial softening behaviour along with the following dilative tendency
is reduced with an increasing 𝑅 (Figure 4.11). This can be attributed to the fact, for higher

54 | P a g e
values of 𝑅 the critical state line is moving downwards (Figure C.1). This will lead to a decrease
in 𝐷𝑅,𝐶𝑆 which gives a lower relative state parameter 𝜉𝑅 (eq. C.2) enabling the model to
approach the 𝐶𝑆𝐿 faster. From Figure 4.11, it is clear, that the effect of the critical state
parameter 𝑅 is significant both in the peak monotonic strength and in the post-peak response.

Regarding the recalibration of the ℎ𝑝𝑜 parameter, it is observed, that by increasing 𝑅, the
post-peak response is becoming considerably softer by approximating better the
experimental responses (Figure 4.12). The simulations are showing ‘limited liquefaction’
behaviour for 𝑅 = 1.75 and ‘liquefaction’ behaviour for 𝑅 = 2.0 and 𝑅 = 2.25 (Chapter
2.2.3).

Figure 4.11: Influence of the critical state parameter 𝑹 in the model response at 𝜎′𝜈𝑐 =
100𝑘𝑃𝑎.

55 | P a g e
Figure 4.12: Influence of the critical state parameter 𝑹 by recalibrating 𝒉𝒑𝒐 in the model
response at 𝜎′𝜈𝑐 = 100𝑘𝑃𝑎.

4.6.2 Influence of Q

The influence of the critical state parameter 𝑄 in the model response is examined by testing
six new set of parameters (Table 4.4). The first pair includes the variation of 𝑄 by keeping all
the other parameters constant for all different confining stress levels. In the following four
sets, the recalibration of the contraction parameter ℎ𝑝𝑜 is performed by using lower values
for 𝑄 at 𝜎′𝜈𝑐 = 100𝑘𝑃𝑎.

Table 4.5: Parametric assessment of 𝑸 in the model response by using different set of
parameters

It is observed, that the initial softening behaviour along with the following dilative tendency
is reduced with an decreasing 𝑄 (Figure 4.13). This can be attributed to the fact, for lower
values of 𝑄 the critical state line is moving downwards (Figure C.1). This will lead to a decrease
in 𝐷𝑅,𝐶𝑆 which gives a lower relative state parameter 𝜉𝑅 (eq. C.2) enabling the model to

56 | P a g e
approach the 𝐶𝑆𝐿 faster. From Figure 4.13, it is clear, that the effect of the critical state
parameter 𝑄 is significant both in the peak monotonic strength and in the post-peak response.
Moreover, it is important to note, that the effect of 𝑄 in the model response is realized with
a lower rate compared to the corresponding effect of the parameter R (Figure 4.11).

Regarding the recalibration of the ℎ𝑝𝑜 parameter, it is observed, that by decreasing 𝑄, the
post-peak response is becoming considerably softer by approximating better the
experimental responses (Figure 4.14). The simulations are showing ‘limited liquefaction’
behaviour for 𝑄 = 9.5 and 𝑄 = 9.0 ‘liquefaction’ behaviour for 𝑄 = 8.5 (Chapter 2.2.3).

Figure 4.13: Influence of the critical state parameter 𝑸 in the model response at 𝜎′𝜈𝑐 =
100𝑘𝑃𝑎.

57 | P a g e
Figure 4.14: Influence of the critical state parameter 𝑸 by recalibrating 𝒉𝒑𝒐 in the model
response at 𝜎′𝜈𝑐 = 100𝑘𝑃𝑎.

4.6.3 Influence of φ’cv

The influence of the critical state friction angle 𝜑′𝑐𝑣 in the model response is examined by
testing two new set of parameters (Table 4.6). This pair includes the variation of 𝜑′𝑐𝑣 by
keeping all the other parameters constant.

Table 4.6: Parametric assessment of 𝝋′𝒄𝒗 in the model response by using different set of
parameters

58 | P a g e
It is observed, that the initial softening behaviour along with the following dilative tendency
is slightly reduced with a decreasing value of 𝜑′𝑐𝑣 . (Figure 4.15). Consequently, the effect of
the critical state friction angle is of less significancet for both the peak monotonic strength
and the post-peak response.

Figure 4.15: Influence of the critical state parameter 𝝋′𝒄𝒗 in the model response at 𝜎′𝜈𝑐 =
100𝑘𝑃𝑎.

59 | P a g e
4.7 Conclusions

 Plaxis PM4Sand model approximates well the general features of static loading
 Regarding the initial parametric selection, the model shows limited liquefaction
behaviour in which the dilative (or post-peak) response after the quasi-steady state is
reached is higher than in the experimental responses
 This high dilative tendency is decreased with an increasing confinement level which
means that higher overburden stresses suppress the dilative response
 The following dilative (or post peak) response after the peak shear strength is reached
can be smoothened and thus approximate better the experimental responses by:

 using lower values for the apparent relative density 𝐷𝑅 and recalibrating the
contraction rate parameter ℎ𝑝𝑜
 using lower values for the bounding surface parameter 𝑛𝑏
 using higher values for the critical state parameter 𝑅 and recalibrating the
contraction rate parameter ℎ𝑝𝑜
 using lower values for the critical state parameter 𝑄 and recalibrating the
contraction rate parameter ℎ𝑝𝑜

60 | P a g e
Chapter 5: Simulation of Earthquake Induced Liquefaction in Hydraulic
Fills

5.1 Introduction

In this chapter the performance of the Plaxis PM4Sand model for the prediction of earthquake
induced liquefaction in hydraulic fills is evaluated. Several different seismic motions
characterized by various fundamental frequencies and peak ground accelerations are
analysed. Then, the influence of these earthquake motions in the liquefaction triggering of
hydraulic fills placed over different relative densities are evaluated. The acceleration-time
signals are obtained from a large database regarding strong-motion seismographs
(www.kyoshin.bosai.go.jp) and need to be initially evaluated before applied to the model.

The hydraulic fill structure is modelled in Plaxis as a 1-D soil column. The stratigraphy of the
soil column is interpreted based on correlations from the literature. This will provide the
appropriate information for the estimation of the required input parameters and also for the
calibration of the PM4Sand model parameters that are going to be used for the liquefiable
layers. The susceptibility of the hydraulic fill to liquefaction which is placed over a different
range of relative densities is initially assessed with the NCEER method. The fill layers that are
prone to liquefy are modelled with PM4Sand model whereas the layers that are not
susceptible to liquefaction are modelled with Hardening Soil Small (HSS) model.

The acceleration-time signals are applied as a dynamic displacement on bedrock level which
generates seismic waves that propagate towards the surface through the soil layers. After
applying the Fourier Transform in each signal, the frequency domain is obtained from which
dominant frequencies are identified. These frequencies are required to determine the size of
the mesh-elements and the dynamic properties of the soil column.

The dynamic analysis is performed with and without consolidation and the lateral boundaries
used are tied degrees of freedom. This type of boundary is ideal for 1-D wave propagation
which allows to reduce the geometry of the problem. The results in terms of the development
of excess pore pressures are examined throughout the whole earthquake motion and the
differences between those two modes are investigated. The onset of liquefaction in the
hydraulic fill is captured when the excess pore pressure ratio has reached a value of around
1.0 (𝑟𝑢 ≈ 1).

5.2 Input Ground Motion

The acceleration-time motions are obtained from the database (www.kyoshin.bosai.go.jp). In

particular, sevel different earthquake signals are analysed with a magnitude of 𝑀𝑤 = 8. The
maximum horizontal accelerations were measured at bedrock level and the values are ranging
between 𝑎𝑚𝑎𝑥 = 0.14𝑔 and 𝑎𝑚𝑎𝑥 = 0.22𝑔. Moreover, the dominant frequencies acquired
after applying the Fourier Transform in each of the acceleration-time signals are varying
between 𝑓𝑓𝑢𝑛𝑑 = 0.37𝐻𝑧 and 𝑓𝑓𝑢𝑛𝑑 = 5.47𝐻𝑧. The earthquake motions can be

61 | P a g e
distinguished into two classes with respect to this fundamental frequency content. The first
class is comprised by motions characterized by low fundamental frequencies (𝑓𝑓𝑢𝑛𝑑 =
0.37𝐻𝑧, 𝑓𝑓𝑢𝑛𝑑 = 0.40𝐻𝑧, 𝑓𝑓𝑢𝑛𝑑 = 0.59𝐻𝑧 and 𝑓𝑓𝑢𝑛𝑑 = 0.70𝐻𝑧) while the second containts
signals of higher dominant frequencies (𝑓𝑓𝑢𝑛𝑑 = 3.31𝐻𝑧, 𝑓𝑓𝑢𝑛𝑑 = 3.60𝐻𝑧 and 𝑓𝑓𝑢𝑛𝑑 =
5.47𝐻𝑧). Before applying the signal to Plaxis an evaluation process is required which will
determine whether the signal is going to be modified or applied as initially measured. This
process consists of the following steps:

 The acceleration motions can be cut to capture only the strong ground motion. In
this analysis, the duration of the acceleration motions is ranging between 140𝑠𝑒𝑐
and 200𝑠𝑒𝑐. The signals are going to be cut at 100𝑠𝑒𝑐 as after this point the
ground shaking is negligible (Appendix A.1 – Figure A.1-A.7)
 If the acceleration in the end of the ground motion is zero the velocity should be
also zero. A non-zero value of the velocity could be due to noise or disturbance of
the wave propagation. In this case, filtering and baseline correction should be
applied to the acceleration-time signal. As this was the case, a bandpass
Butterworth filter is applied in which the frequency range was set between 0.1𝐻𝑧
and 11𝐻𝑧. The Butterworth filter is a type of signal processing filter designed to
have a frequency response as flat as possible in the passband.
 The acceleration-time history to be applied in the model represents the signal at
bedrock level. In case the ground motion is obtained at surface, a deconvolution
process should be applied to capture the signal variation from bedrock to surface
due to the different dynamic soil properties of the layers. In this case, the
deconvolution is not applied because the acceleration ground motion corresponds
to the signal at bedrock.
 It is known that the filtering or the correction process may modify the initial input
signal, especially in terms of Peak Ground Acceleration (PGA). Thus, the new signal
needs to be scaled to the original PGA.
 After finishing the filtering and correction process, the input signal is applied in
Plaxis at bedrock level as a dynamic displacement. This generates seismic waves
propagating towards surface through the model. Moreover, Fourier Transform is
applied to the acceleration signal in order to obtain the frequency amplitude
spectrum from which the fundamental frequencies can be defined (Appendix A.1 –
Figure A.8-A.14). These frequencies are required to define the Rayleigh damping
coefficients in Chapter 5.4.4

5.3 Soil Column Interpretation

The hydraulic fill structure is simulated as a 1-D soil column (Figure 5.1). The fill material
consists of sand which is placed 1𝑚 over water using the discharge pipeline method with 𝐷𝑅 =
60% and 10𝑚 underwater using the rainbowing method with different relative densities
varying between 𝐷𝑅 = 30% and 𝐷𝑅 = 60%. For deeper layers the relative density is

62 | P a g e
apparently increased (𝐷𝑅 = 70% from 11m to 26m and 𝐷𝑅 = 80% from 26m to 41m) and
the bedrock level is assumed to be at 41m depth.

The strength parameters of the soil column are calculated as a function of the relative density
from eq.(5.1-5.3). The latter is set equal to the relative density of the considered layer (Figure
5.3) since there are no available laboratory or in-situ data.

Figure 5.1: Simulated hydraulic fill of 11m as a 1-D soil column.

63 | P a g e
The friction angle and the unit weight are calculated according to (Brinkgreve, R.B.J.,2010) as:

𝐷𝑅
𝜑’ = 28 + 12.5 [⁰] (5.1)
100
𝐷𝑅
𝛾𝑢𝑛𝑠𝑎𝑡 = 15 + 4 [𝑘𝑁/𝑚3 ] (5.2)
100
𝐷𝑅
𝛾𝑠𝑎𝑡 = 19 + 1.6 [𝑘𝑁/𝑚3 ] (5.3)
100

Table 5.1: Interpretation of the soil column.

Liquefaction Susceptibility

The liquefaction susceptibility of the hydraulic fill and of the whole soil column is initially
evaluated with the NCEER method. The Cyclic Stress Ratio (𝐶𝑆𝑅) of each layer is calculated
according to the semi-empirical method (Seed & Idriss, 1971) from eq.(2.7). The Cyclic
Resistance Ratio (𝐶𝑅𝑅) is determined from eq.(2.24) where factors concerning the
earthquake magnitude, the confinement level and sloping ground effects are considered.

As level ground conditions was the case, the 𝐾𝑎 factor is taken as unity. The 𝐶𝑅𝑅7.5 is
calculated from eq.(2.11) where the (𝑁1 )60 value is assessed from eq.(2.14) for each value of
relative density 𝐷𝑅 . Then the factory of safety 𝐹𝑆 is determined from eq.(2.23) which provides
a clear insight for the layers that are prone to liquefy (𝐹𝑆 < 1).

The results of the liquefaction susceptibility analysis for the hydraulic fill earthquake are given
in Appendix A.2 (Table A.16-A.19) regarding the seven earthquake motions. The results of the
whole soil column for a typical fill of 𝐷𝑅 = 30% relative density are indicated in Appendix A.2
(Table A.15). Note that, in the aforementioned Tables, the results for four acceleration
motions are indicated as some of the signals have the same peak ground acceleration. It is
clear that, according to NCEER method, the liquefaction potential of the hydraulic fill placed
with relative densities between 𝐷𝑅 = 30% and 𝐷𝑅 = 60% for all different seismic motions is
high (𝐹𝑆 < 1).

64 | P a g e
5.4 Plaxis 2D Modelling

5.4.1 Soil Layers with Hardening Soil Small (HSS) Model

The Hardening Soil Small (HSS) model is used to model all layers in static analysis as well as for
soil layers in dynamic analysis that are not susceptible to liquefaction (𝐹𝑆 > 1). The layers
with high liquefaction potential (𝐹𝑆 < 1) are modelled with the PM4Sand model. The
following correlations are used to define the input parameters of the HSS model. These
parameters are mainly dependent on the relative density of the soil and given by (Brinkgreve,
R.B.J.,2010) as:
𝑟𝑒𝑓
𝐸50 ≈ 60𝐷𝑅 (5.4)

𝑟𝑒𝑓 𝑟𝑒𝑓
𝐸𝑜𝑒𝑑 = 𝐸50 ≈ 60𝐷𝑅 (5.5)

𝑟𝑒𝑓 𝑟𝑒𝑓
𝐸𝑢𝑟 ≈ 2𝐸𝑜𝑒𝑑 (5.6)

𝑟𝑒𝑓 𝐷𝑅
𝐺𝑜 = 60 + 68 (5.7)
100
𝐷𝑅
𝛾0.7 = (2 − ) 10−4 (5.8)
100
𝐷𝑅
𝑚 = 0.7 − (5.9)
320

𝐾𝑜,𝑛𝑐 = 1 − sin 𝜑′ (5.10)

𝐷𝑅
𝑅𝑓 = 1 − (5.11)
800
𝑟𝑒𝑓 𝑟𝑒𝑓
where 𝐸50 is the secant stiffness from triaxial test at reference pressure, 𝐸𝑜𝑒𝑑 is the tangent
𝑟𝑒𝑓
stiffness from oedometer test at reference pressure, 𝐸𝑢𝑟 is the unloading/reloading stiffness,
𝑟𝑒𝑓
𝐺𝑜 is the reference shear stiffness at small strains, 𝛾0.7 is the shear strain in which the shear
modulus 𝐺 has reduced to 72.2%, 𝑚 is the rate of stress-level dependency in stiffness
behaviour, 𝐾𝑜,𝑛𝑐 is the horizontal to vertical ratio in primary 1D compression and 𝑅𝑓 is the
failure ratio.

The list of the Hardening Soil Small (HSS) parameters for all soil layers that are going to be
used for the analysis are indicated in Table 5.2.

65 | P a g e
Table 5.2: Hardening Soil Small (HSS) parameters for each soil layer.

5.4.2 Soil Layers with PM4Sand Model

The PM4Sand model is used in dynamic analysis to model the soil layers that are susceptible
to liquefaction (𝐹𝑆 < 1). The relevant layers consist of the hydraulic fill (𝐷𝑅 = 30 − 60%)
which is extended from 1𝑚 to 11𝑚 depth (Figure 5.1).

Apparent Relative Density DR

The apparent relative density (𝐷𝑅 ) of the hydraulic fill is taken as the relative of the whole
layer: 𝐷𝑅 = 0.30, 𝐷𝑅 = 0.40, 𝐷𝑅 = 0.50 and 𝐷𝑅 = 0.60 for each case.

Shear Modulus Coefficient Go

The shear modulus coefficient (𝐺𝑜 ) is calculated from eq.(3.1) as a function of relative
density (𝐷𝑅 ).

Contraction Rate Parameter hpo

The contraction rate parameter (ℎ𝑝𝑜) is calibrated to match specific Cyclic Resistance Ratio
(𝐶𝑅𝑅) in Plaxis element test corresponding to a 𝛾 = 3% failure criterion at 𝑁𝑐 = 15 loading
cycles in direct simple shear loading. The calibration is done for the middle of the fill layer
which is at 6𝑚 depth. It is important to note, that the target 𝐶𝑅𝑅 (Table 5.3) considered for
the calibration of ℎ𝑝𝑜 takes into account the earthquake magnitude (𝑀𝑆𝐹) and overburden
stress effects (𝐾𝜎 ) as shown in eq.(2.44).

Table 5.3: Target CRR from which to calibrate the contraction rate parameter ℎ𝑝𝑜 .

66 | P a g e
In the following table the primary set of the model parameters used in PM4Sand are indicated
with the secondary parameters kept as default (Boulanger & Ziotopoulou, 2015).

Table 5.4: Primary set of model parameters used in PM4Sand.

5.4.3 Mesh Generation and Time Step

In dynamic finite element models, the determination of the size of the mesh-elements along
with the time step of the calculation is of high importance to ensure proper seismic wave
propagation through the model. This means that the element size would preferably not be
too large. However, a too small element size would lead to large computational time so the
user has to select an optimum average element size for each layer.

The maximum element size per layer is defined according to (Kuhlmeyer & Lysmer, 1973) as:

𝜆 𝑣𝑠,𝑙𝑎𝑦𝑒𝑟
𝛥𝑙𝑎𝑣𝑒𝑟𝑎𝑔𝑒 ≤ = (5.12)
8 8𝑓𝑚𝑎𝑥

where 𝜆 is the wavelength, 𝑣𝑠,𝑙𝑎𝑦𝑒𝑟 is the shear wave velocity of the considered layer and
𝑓𝑚𝑎𝑥 is the maximum frequency component of the input signal. The maximum frequency
component is obtained after applying the Fourier Transform on the acceleration signal (Figure
5.2) as 𝑓𝑚𝑎𝑥 = 11𝐻𝑧.

The dynamic time step in Plaxis is set equal to the time step of the acceleration input signal.
In this case, the time step is 𝛥𝑡 = 0.005𝑠. Moreover, the waves have to be prevented from
traveling within more than one element per time step. This can be evaluated by the following
expression where the critical time step is defined as:

𝛥𝑙𝑎𝑣𝑒𝑟𝑎𝑔𝑒
𝛥𝑡 ≤ (5.13)
𝑣𝑠,𝑙𝑎𝑦𝑒𝑟

The shear wave velocity for each layer is calculated as:

𝐺
𝑣𝑠,𝑙𝑎𝑦𝑒𝑟 = √ (5.14)
𝜌

where 𝐺 is the elastic shear modulus and 𝜌 is the density of the soil layer.

The elastic shear modulus 𝐺 for the layers modelled with HSS small model is calculated as:

67 | P a g e
 𝑚
𝜎′3 sin(𝜑′)
𝐺 = 𝐺𝑜,𝐻𝑆 = 𝐺𝑜,𝑟𝑒𝑓 ( ) (5.14)
𝑝𝑟𝑒𝑓 sin(𝜑′)

where 𝑝𝑟𝑒𝑓 is the reference confining pressure taken as 100kPa and 𝜎′3 is the minor principal
stress in a triaxial test.

The elastic shear modulus 𝐺 for the fill layer modelled with PM4Sand model is calculated in
two different ways: 𝑖) according to eq.(2.18) where 𝐺𝑜 is taken from eq.(3.1) (PM4Sand model
formulation) and 𝑖𝑖) according to eq.(5.14) (HSS Small model formulation). Then, the
maximum element size and maximum time step for each layer is computed from eq.(5.12)
and eq.(5.13) and the results are indicated in Table 5.5. It is important to note that the value
of 𝑣𝑠,𝑙𝑎𝑦𝑒𝑟 in the aforementioned equations is set equal to the average value of shear wave
velocity for the considered soil layer. The analytical calculations for the whole soil column are
indicated in Appendix A.2.

Table 5.5 i), ii): Average shear wave velocity, maximum frequency, maximum element size,
maximum time step for each layer by taking into account the elastic shear modulus
𝐺 calculated according to i) PM4Sand model and ii) HSS Small model.

5.4.4 Rayleigh Damping

In Plaxis, two major damping components are introduced, representing the dissipation of
energy from an earthquake due to the wave propagation through the model. The first one is
the hysteric damping which accounts for damping at large strains. However, the soil
behaviour may be irreversible even at low strain levels. For that purpose, the Rayleigh
damping is introduced which can capture damping at small strains.

The damping matrix 𝐶 is composed of the mass matrix 𝑀 and the stiffness matrix 𝐾 and given
according to the Rayleigh damping formulation as:

𝐶⃗ = 𝑎𝑅 𝑀
⃗⃗⃗ + 𝛽𝑅 𝐾
⃗⃗ (5.15)

where 𝑎𝑅 and 𝛽𝑅 are the Rayleigh damping coefficients.

68 | P a g e
The Rayleigh damping coefficients can be calculated after the two target frequencies have
been defined. According to (Hudson, 1994) the first target frequency is set equal to the
fundamental frequency of the considered soil profile as:
𝑣𝑠,𝑎𝑣𝑒𝑟𝑎𝑔𝑒
𝑓1 = (5.16)
4𝐻

where 𝐻 is the thickness and 𝑣𝑠,𝑎𝑣𝑒𝑟𝑎𝑔𝑒 is the average shear wave velocity of the considered
soil layer.

The second target frequency is set as the closest odd number regarding the ratio of the
fundamental frequency of the acceleration signal at bedrock to the first target frequency as:

𝑓𝑓𝑢𝑛𝑑
𝑓2 ≈ (5.17)
𝑓1

where 𝑓𝑓𝑢𝑛𝑑 is the frequency with the maximum Fourier amplitude (Figure 5.2).

The Rayleigh damping coefficients 𝑎𝑅 and 𝛽𝑅 are then calculated as:

2𝜉𝜔1 𝜔2
𝛼𝑅 = (5.18)
𝜔1 +𝜔2

2𝜉
𝛽𝑅 = (5.19)
𝜔1 +𝜔2

where:

𝜔𝑖 = 2𝜋𝑓𝑖 (5.20)

where 𝜉 is the target damping ratio and 𝜔𝑖 is the mode angular circular frequency.

In literature, the suggested values for damping ratio are varying between 0.5% and 2%. In this
case, a value of 1% is used.

Table 5.6: Rayleigh damping coefficients by considering the elastic shear modulus 𝐺
according to the PM4Sand model formulation.

69 | P a g e
Table 5.7: Rayleigh damping coefficients by considering the elastic shear modulus
𝐺 according to the HSS Small model formulation.

The target frequencies 𝑓1 and 𝑓2 along with the corresponding Rayleigh damping coefficients
𝑎𝑅 and 𝛽𝑅 define frequencies that are going to be absorbed by the soil layers. This means
that frequencies higher than 𝑓1 are going to be over-damped. In that way, the frequency
range between 𝑓1 and 𝑓2 is set automatically with lower damping values by Plaxis. This allows
frequencies belonging to that range to be taken into account in the analysis. Using e.g higher
values of 𝑓1 results in a lower range of over-damped frequencies. This allows for more
frequencies to be taken into account in the analysis and thus having a bigger influence in the
results.

5.5 Dynamic Analysis

The dynamic analysis in Plaxis can be excecuted in two different ways: with consolidation or
without consolidation. The latter is implemented in the latest version of Plaxis 2D 2018 and
provides less mesh dependence and better spread of pore-water pressures according to the
manual. In this study, both types of dynamic analysis are tested out and the differences in the
results in terms of excess pore pressures evolution are investigated. Regarding the lateral
boundaries, if free-field boundaries are to be used then the material at the boundaries needs
to be replaced by drained layers. This is required to prevent complete loss of strength at the
boundaries which may result in large deformations. In this case, the selected lateral boundary
conditions are tied degrees of freedom as the hydraulic fill structure is simulated as a 1-D soil
column. This allows to simulate a reduced geometry of the problem. The nodes at the left
and right model boundaries are connected such that the nodes will undergo the same
displacement At the bottom boundary a compliant base is applied by adding a surface to
absorb downwards propagating waves.

The geometry of the simulated hydraulic fill and of the whole soil column is depicted in
Figure 5.2.

70 | P a g e
Figure 5.2: Model geometry. Hydraulic fill simulated as a 1-D soil column.

5.6 Results

This section includes the results obtained from Plaxis regarding the generation of excess pore
pressures in the hydraulic fill layers with relative densities of 𝐷𝑅 = 30%, 𝐷𝑅 = 40% , 𝐷𝑅 =
50% and 𝐷𝑅 = 60% for all seven different earthquake motions (Appendix A.1). In real
practise, it is suggested to scale the series of acceleration signals into a specific value of peak

71 | P a g e
ground acceleration. In this thesis project, this is not carried out as the resulting excess pore
pressures evolution in the end of the earthquake motion (and thus the earthquake-induced
liquefaction) is not influenced by the peak ground acceleration. Firstly, dynamic analyses
without consolidation are performed for all the aforementioned fill cases followed by dynamic
analyses with consolidation. The differences in the results in terms of excess pore pressures
evolution are investigated for both types of analysis. Soil layers that reach an excess pore
pressure ratio (𝑟𝑢 ) of around 1.0 are identified as liquefied.

5.6.1 Sensitivity analysis

Mesh element size

Initially, the dynamic analyses in Plaxis are performed with a very coarse mesh. A domain
comprised of coarse mesh elements is suggested to be employeed in order to save
computational time. However, this triggered the question on whether finer mesh elements
would result in different results. Consequently, the influence of the mesh element size in the
generation of excess pore pressures is investigated by refining the domain. Two finer element
sizes are tested for the analyses: an element size of 0.90𝑚 and 0.46𝑚 compared to the initial
analyses where the element size was 1.45𝑚. It is observed, that when using finer mesh
elements the results remain the same in terms of excess pore pressures generation. This can
be attributed to the fact, that the default element size of 1.45𝑚 is within the range of
allowable limits regarding the maximum value (Table 5.5) that leads to a proper wave
propagation along the soil column.

Time step

The dynamic time step in Plaxis is set equal to the time step of the acceleration input signal.
This is set in the dynamic phase where the time step is set as the dynamic time interval (100𝑠)
divided by the maximum number of steps stored in the software. In order to save
computational time, the user can reduce the maximum number of substeps while increasing
the number of substeps, as long as the aforementioned division is even to the dynamic time
step (0.01𝑠) (eq. 5.18).

𝐷𝑦𝑛𝑎𝑚𝑖𝑐 𝑡𝑖𝑚𝑒 𝑖𝑛𝑡𝑒𝑟𝑣𝑎𝑙

𝑇𝑖𝑚𝑒 𝑠𝑡𝑒𝑝 = (5.18)
𝑀𝑎𝑥 𝑛𝑜. 𝑠𝑡𝑒𝑝𝑠 ∗ 𝑆𝑢𝑏𝑠𝑡𝑒𝑝𝑠

Target frequency

As described in Chapter 5.4.4, the target frequencies are defined through the elastic shear
modulus 𝐺 in two different ways: i) according to the formula for HSS Small model and ii)
according to the formula for PM4Sand model. These are giving different values for the target
9frequency 𝑓1. By applying (i) for the calculation of target frequencies leads to slightly higher
values of 𝑓1 and consequently to a lower range of over-damped frequencies. This allows for
more frequencies to be taken into account in the analysis and one should expect bigger
influence in the results. For that reason, both ways are tested and the differences in terms of
excess pore pressure evolution are examined. It is observed, that the range between the two

72 | P a g e
target frequencies has a minor effect in the generation of excess pore pressures for all
different hydraulic fills. On the other hand, the influence of the range between the target
frequencies is more explicit in the resulting displacements. In particular, a bigger range leads
to slightly higher horizontal displacements along the fill column.

Final set of parameters

Mesh element size Time step Dynamic time interval Max no. steps Substeps Target frequency f1 Target frequency f2(Hz)
(m) (s) (s) (-) (-) (Hz) (Hz)
1.45 0.01 100 2000 5 3.2 1
Table 5.8: Final set of parameters that are used in the dynamic analysis.

5.6.2 Dynamic Analysis without Consolidation

Excess Pore Pressures Generation

Dr=30%

i) After 24𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.3).
Between 24𝑠 and 32𝑠 an increase in excess pore pressure is identified mainly in the middle
hydraulic fill layer that propagates upwards and downwards through the soil column. After a
rapid increase in excess pore pressures at 35𝑠 the liquefaction condition is reached (𝑟𝑢 ≈ 1)
at 38𝑠. This condition is extended nearly throughout the entire fill layer. The liquefaction
condition that has been reached up to that point remains constant and restraint in the parts
of the hydraulic fill that have liquefied and it does not propagate further through the soil
column until the end of the earthquake signal (100𝑠).

Figure 5.3: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 30% for
dynamic analysis without consolidation: (𝑖, 𝑣)

v) After 30𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.3).
Between 30𝑠 and 33𝑠 a slight increase in excess pore pressure is identified in the hydraulic fill
layer and at 35𝑠 the liquefaction condition is reached (𝑟𝑢 ≈ 1) in scattered zones along the

73 | P a g e
fill layer. Between 35𝑠 and 46𝑠 the liquefaction state is spreaded in between these zones by
creating a uniform liquefied layer. The liquefaction condition that has been reached up to
that point remains constant and restraint in the parts of the hydraulic fill that have liquefied
and it does not propagate further through the soil column until the end of the earthquake
signal (100𝑠).

The results regarding the remaining acceleration signals (𝑖𝑖, 𝑖𝑖𝑖, 𝑖𝑣, 𝑣𝑖, 𝑣𝑖𝑖) and the
corresponding evolution of excess pore pressures along the hydraulic fill are indicated in
Figure 5.4. It is observed, that the liquefied parts are localized in discrete zones within the
hydraulic fill and are not extending along the entire fill column.

Figure 5.4: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 30% for
dynamic analysis without consolidation (𝑖𝑖, 𝑖𝑖𝑖, 𝑖𝑣, 𝑣𝑖, 𝑣𝑖𝑖)

Dr=40%

i) After 24𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.5).
Between 24 and 35𝑠 an increase in excess pore pressure is identified mainly in the middle
hydraulic fill layer that propagates upwards and downwards through the soil column. After a
rapid increase in excess pore pressure at 38𝑠 the liquefaction condition is reached (𝑟𝑢 ≈ 1)
which extends from bottom to mid-upper part of the fill and at 43𝑠 the liquefaction state is
propagated gently upwards. The liquefaction condition that has been reached up to that point
remains constant and restraint in the parts of the hydraulic fill that have liquefied and it does
not propagate further through the soil column until the end of the earthquake signal (100𝑠).

ii) After 32𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.5).
Between 32𝑠 and 43𝑠 an increase in excess pore pressure is identified mainly in the middle

74 | P a g e
hydraulic fill layer that propagates upwards and downwards through the soil column. After a
rapid increase in excess pore pressure at 49𝑠 the liquefaction condition is reached (𝑟𝑢 ≈ 1)
in the mid-upper part of the fill. Between 49𝑠 and 57𝑠 the liquefaction state is extended
from top to nearly the bottom fill by creating a uniform liquefied layer.

Figure 5.5: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 40% for
dynamic analysis without consolidation.

iv) After 35𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.5).
Between 35𝑠 and 41𝑠 an increase in excess pore pressure is identified mainly in the mid-upper
hydraulic fill layer that propagates upwards and downwards through the soil column. After a
rapid increase in excess pore pressure at 43𝑠 the liquefaction condition is reached (𝑟𝑢 ≈ 1)
in the mid-upper layer. Between 43𝑠 and 51𝑠 a significant rise in the excess pore pressures
is observed likewise in the mid-lower fill layer and at 54𝑠 this part has also liquefied (𝑟𝑢 ≈ 1).
Between 54𝑠 and 57𝑠 the liquefaction state is propagated between the two aforementioned
liquefied layers and eventually almost the entire hydraulic fill has liquefied.

v) After 30𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.5).
Between 30𝑠 and 32𝑠 a slight increase in excess pore pressures is identified that propagates
upwards and downwards through the soil column. After a rapid increase in excess pore

75 | P a g e
pressures at 35𝑠 the liquefaction condition is reached (𝑟𝑢 ≈ 1) which extends from bottom
to mid-upper part of the fill and at 38𝑠 the liquefaction state is propagated gently upwards.

The results regarding the remaining acceleration signals (𝑖𝑖𝑖, 𝑣𝑖, 𝑣𝑖𝑖) and the corresponding
evolution of excess pore pressures along the hydraulic fill are indicated in Figure 5.6. It is
observed, that, the liquefied parts are restricted in distinct zones within the hydraulic fill and
are not extending along the entire fill column.

Figure 5.6: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 40% for
dynamic analysis without consolidation

At this point, it important to highlight, that the excess pore pressures evolution along the
hydraulic fill and eventually the liquefaction triggering is dependent on the dominant
frequency of the considered signal regardless the peak ground acceleration values. In
particular, there is a better diffusion of excess pore pressures in the loosely-packed fills (𝐷𝑅 =
30% and 𝐷𝑅 = 40%) for the signals of low dominant frequencies (Figure 5.3, 5.5). However,
for the rest of the signals, there is a localization of liquefied zones in distinct parts along the
fill (Figure 5.4, 5.6).

Dr=50%

iii) After 22𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.7).
Between 22𝑠 and 30𝑠 an increase in excess pore pressures is identified mainly in the middle
hydraulic fill layer that propagates upwards and downwards through the soil column.
Between 32𝑠 and 38𝑠 a rapid increase in excess pore pressure is observed in three distinct
zones of the fill layer which have eventually liquefied (𝑟𝑢 ≈ 1). Between 38𝑠 and 41𝑠 the
liquefaction state is spreaded in between those zones forming almost an entire liquefied layer
extending from bottom to mid-upper fill part.

vi) After 22𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.7).
Between 22𝑠 and 30𝑠 an increase in excess pore pressure is identified in the mid-upper
hydraulic fill layer that propagates upwards and downwards through the soil column.
Between 32𝑠 and 35𝑠 a rapid increase in excess pore pressures is observed in three distinct
zones of the fill layer which eventually have liquefied (𝑟𝑢 ≈ 1). Between 35𝑠 and 38𝑠 the
liquefaction state is spreaded in between those zones forming a uniform liquefied layer

76 | P a g e
extending from bottom to mid-upper fill part and at 41𝑠 the liquefaction state is slightly
propagated upwards.

vii) After 24𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.7).
Between 24𝑠 and 30𝑠 a slight increase in excess pore pressure is identified in the mid-upper
hydraulic fill layer that propagates upwards and downwards through the soil column. After a
rapid increase in excess pore pressures at 32𝑠 the liquefaction condition is reached (𝑟𝑢 ≈ 1)
which extends from bottom to mid-upper part of the fill and at 38𝑠 the liquefaction state is
propagatedgently upwards.

Figure 5.7: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 50% for
dynamic analysis without consolidation

The results regarding the remaining acceleration signals (𝑖, 𝑖𝑖, 𝑖𝑣, 𝑣) and the corresponding
evolution of excess pore pressures along the hydraulic fill are indicated in Figure 5.8. The
liquefied parts are localized in a singular restricted zone within the fill layer which is not the
case in the fills of lower densities (𝐷𝑅 = 30%, 𝐷𝑅 = 40% ) in which the liquefied parts are
developing in more discrete zones and in a less restricted manner. Therefore, the effect of
the relative density as an input parameter has a significant role in the liquefaction triggering.

77 | P a g e
Figure 5.8: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 50% for
dynamic analysis without consolidation

Dr=60%

iii) After 22𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.9).
Between 22𝑠 and 32𝑠 an increase in excess pore pressure is identified that propagates
upwards and downwards through the soil column. Between 32𝑠 and 38𝑠 a rapid increase in
excess pore pressure is observed in two distinct parts of the fill which have eventually
liquefied (𝑟𝑢 ≈ 1). Between 38𝑠 and 43𝑠 the liquefaction state is spreaded in between those
parts forming an entire liquefied layer extending from bottom to mid-upper fill.

78 | P a g e
Figure 5.9: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 60% for
dynamic analysis without consolidation.

vi) After 27𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.9).
Between 27𝑠 and 32𝑠 an increase in excess pore pressures is identified mainly in the middle
hydraulic fill layer that propagates upwards and downwards through the soil column.
Between 35𝑠 and 38𝑠 a rapid increase in excess pore pressure is observed in three distinct
zones of the fill layer which have eventually liquefied (𝑟𝑢 ≈ 1). Between 38𝑠 and 41𝑠 the
liquefaction state is spreaded in between those zones forming an entire liquefied layer
extending from bottom to mid-upper fill part.

vii) After 24𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.9).
Between 24𝑠 and 30𝑠 an increase in excess pore pressures is identified mainly in the middle
hydraulic fill layer that propagates upwards and downwards through the soil column. After a
rapid increase in excess pore pressures at 32𝑠 the liquefaction condition is reached (𝑟𝑢 ≈ 1)
which extends from bottom to middle part of the fill.

Regarding the aforementioned results, the liquefaction condition in the end of the earthquake
is more restricted compared to the fills of lower relative densities (𝐷𝑅 = 30%, 𝐷𝑅 = 40%) in
which the liquefied parts are extending almost along the entire fill column (Figure 5.3, 5.5).
Consequently, the relative density as an input parameter has a significant role in the
liquefaction triggering.

The results regarding the remaining acceleration signals (𝑖, 𝑖𝑖, 𝑖𝑣, 𝑣) and the corresponding
evolution of excess pore pressures along the hydraulic fill are indicated in Figure 5.10. It is
observed, that the liquefied parts are localized in a singular restricted zone within the fill layer
regarding two signals (𝑖, 𝑣) while for the rest (𝑖𝑖, 𝑖𝑣), the liquefaction is not triggered at all.
This can be attributed to the fact, that the relative density is significant in the evolution of
excess pore pressures and eventually in the liquefaction triggering.

79 | P a g e
Figure 5.10: Excess pore pressure ratio development in the hydraulic fill of DR = 60% for
dynamic analysis without consolidation

At this point, it is important to note, that there is a better diffusion of excess pore pressures
in the densely-packed fills (𝐷𝑅 = 50% and 𝐷𝑅 = 60%) for the signals of high dominant
frequencies (Figure 5.7, 5.9). This phenomenon is contrary to the one observed in the loosely-
packed fills (𝐷𝑅 = 30% and 𝐷𝑅 = 40%), in which there is a better excess pore pressure
diffusion for the low dominant frequency signals. However, there is a localization of a singular
liquefied zone within the fill for the rest of the signals (Figure 5.8, 5.10).

80 | P a g e
5.6.3 Dynamic Analysis with Consolidation

The new version of Plaxis 2D 2018 is incorporating the mode of dynamic analysis with
consolidation. This type of analysis provides less mesh dependence and better spread of pore-
water pressures according to the Plaxis manual. Moreover, in the current mode, the
coefficients of permeability need to be introuced to the software. Depending on the
magnitude of these coefficients, the rate of excess pore pressures dissipation is defined. In
particular, for higher values of the coefficient of permeability, the excess pore pressures are
dissipating faster as the flow rate for drainage is higher.

Due to the lack of experimental data (e.g grain size distribution curves) from which to
determine those coefficients, empirical values are adopted (Table 5.8) which are taking into
account a wide range of soils with different permeabilites. The coefficients of permeability
for the hydraulic fills placed over relative densities of 𝐷𝑅 = 30%, 𝐷𝑅 = 40% , 𝐷𝑅 = 50% and
𝐷𝑅 = 60% can be considered to vary between the values from low permeable to medium
permeable soils. Therefore, the coefficients of permeability are considered to range between
the values of 10−6, 10−5, 10−4 , 10−3 𝑚/𝑠.

In the following section, a sensitivity analysis regarding the influence of the aforementioned
coefficients in the generation of excess pore pressures is carried out for all different hydraulic
fills and the differences in the results with respect to the dynamic analysis without
consolidation are investigated. In the dynamic analyses with consolidation, one
representative acceleration motion is depicted for each relative density which is followed by
the corresponding dynamic analysis without consolidation for the same signal. This is done,
in order to explain the main differences between the two different types of analysis. The
results for the other representative acceleration signals are indicated in Appendix A.3.

Table 5.8: Typical coefficients of permeability for different soils

81 | P a g e
Excess Pore Pressures Generation

Dr=30%

 (ii) 𝑓𝑓𝑢𝑛𝑑 = 0.37𝐻𝑧, 𝛼𝑚𝑎𝑥 = 0.16𝑔

k=10-6m/s

After 35𝑠 excess pore pressures start to develop in the hydraulic fill layer (Figure 5.11).
Between 35𝑠 and 43𝑠 an increase in excess pore pressure is identified mainly in the mid-upper
hydraulic fill layer that propagates upwards and downwards through the soil column. After a
rapid increase in excess pore pressures at 46𝑠 the liquefaction condition (𝑟𝑢 ≈ 1) is reached
in the mid-upper fill part. Between 46𝑠 and 54𝑠 the liquefaction state is reached in two more
discrete zones along the fill column. Moreover, between 54𝑠 until the end of the earthquake
(100𝑠) a slight increase in excess pore pressures is observed in between the liquefied parts.
This is can be attributed to the fact, that with 𝑘 = 10−6 𝑚/𝑠, the developed excess pore
pressures tend to dissipate upwards and downwards along the soil column in between the
liquefied layers. Hence, the liquefaction condition (𝑟𝑢 ≈ 1) is better spread along the fill layer.
However, the magnitude of the current permeability coefficient is very low. Therefore, there
is not enough time for enough drainage and for a more extensive diffusion of the liquefaction
state throughout the fill layer. As a result, some parts of the fill layer did not liquefy at the
end of the earthquake (100𝑠). It is important to note, the current dynamic analysis with
consolidation, for this extremely low value of permeability coefficient (10−6 𝑚/𝑠), nearly
resembles to the undrained response for the corresponding dynamic analysis without
consolidation (Figure 5.12).

k=10-5m/s

Between 35𝑠 and 43𝑠 the generation of excess pore pressures is almost identical compared
to the analysis with 𝑘 = 10−6 𝑚/𝑠 (Figure 5.11). However, the impact of the permeability
coefficient has a crucial role after this point. In particular, the liquefaction state is better
spread along the fill column between 46𝑠 until the end of the earthquake (100𝑠).
Cosequently, by applying a permeability coefficient of higher magnitude, the dissipation of
excess pore pressures that have developed in the liquefied layers is accomplished in a faster
rate. Hence, the liquefaction condition (𝑟𝑢 ≈ 1) is more uniformly diffused throughout the
fill layer. As a result, at the end of the ground motion the entire fill has liquefied after 𝑟𝑢 ≈ 1
is reached.

82 | P a g e
Figure 5.11: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 30% for
dynamic analysis with consolidation.

k=10-4m/s & k=10-3m/s

Between 35𝑠 and 43𝑠 the rate of excess pore pressures generation is increased compared to
the analyses with 𝑘 = 10−6 𝑚/𝑠 and 𝑘 = 10−5 𝑚/𝑠 (Figure 5.11). This increase originates
from the top fill part and propagates downwards through the soil column. At 46𝑠 the
liquefaction condition (𝑟𝑢 ≈ 1) is reached in the top fill part. Moreover, between 46𝑠 and
57𝑠, the liquefaction state is uniformly propagated from top to bottom fill part without
forming discrete or scattered zones throughout. Similarly, by applying a permeability
coefficient of an even higher magnitude (𝑘 = 10−4 𝑚/𝑠 & 𝑘 = 10−3 𝑚/𝑠), the excess pore
pressures developed in the liquefied layers are dissipating more rapidly with respect to the
ones of lower magnitudes (𝑘 = 10−6 𝑚/𝑠 & 𝑘 = 10−5 𝑚/𝑠). Hence, the liquefaction
condition (𝑟𝑢 ≈ 1) is even more uniformly diffused through the fill layer. As a result, at 57𝑠
the entire fill has liquefied. It is important to note, that the liquefaction phenomenon
capturing the entire fill is occuring much earlier compared to the 𝑘 = 10−5 𝑚/𝑠 case in which
the fill column has liquefied in the end of the earthquake (100𝑠).

83 | P a g e
Figure 5.12: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 30% for
dynamic analysis without consolidation

Dr=40%

 (vi) 𝑓𝑓𝑢𝑛𝑑 = 5.47𝐻𝑧, 𝛼𝑚𝑎𝑥 = 0.22𝑔

k=10-6m/s

Between 11𝑠 and 16𝑠 an increase in excess pore pressures is identified mainly in the upper
hydraulic fill layer that propagates downwards through the soil column (Figure 5.13). After a
rapid increase in excess pore pressures at 19𝑠 the liquefaction condition (𝑟𝑢 ≈ 1) is reached
in both the mid-upper and bottom fill part. Between 19𝑠 until the end of the earthquake
(100𝑠) a slight increase in excess pore pressures mainly in the top fill part. Similarly as
described before, the current dynamic analysis with consolidation, for an extremely low value
of permeability coefficient (10−6 𝑚/𝑠), nearly resembles the undrained response for the
corresponding dynamic analysis without consolidation (Figure 5.14).

k=10-5m/s

Between 11𝑠 and 16𝑠 the generation of excess pore pressures is very similar compared to the
analysis with 𝑘 = 10−6 𝑚/𝑠 (Figure 5.13). However, the impact of the permeability coefficient
has a crucial role after this point. In particular, the liquefaction state is better diffused along
the fill column between 19𝑠 and 57𝑠 in which the entire fill has liquefied. The liquefaction
phenomenon is capturing the entire fill layer much earlier (57𝑠) compared to the
𝑘 = 10−6 𝑚/𝑠 case in which the entire fill has not liquefied even at the end of the earthquake
(100𝑠).

84 | P a g e
Figure 5.13: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 40% for
dynamic analysis with consolidation.

k=10-4m/s

Between 11𝑠 and 16𝑠 the rate of excess pore pressures generation is increased compared to
the analyses with 𝑘 = 10−6 𝑚/𝑠 and 𝑘 = 10−5 𝑚/𝑠 (Figure 5.13). This increase originates
from the top fill part and propagates downwards through the soil column. As described
before, by applying a permeability coefficient of an even higher magnitude (𝑘 = 10−4 𝑚/𝑠),
the drainage flow rate is higher with respect to the ones of lower magnitudes
(𝑘 = 10−6 𝑚/𝑠 & 𝑘 = 10−5 𝑚/𝑠). Hence, the liquefaction condition (𝑟𝑢 ≈ 1) is even more
uniformly diffused through the fill layer. As a result, at 24𝑠 the entire fill has liquefied. It is
important to note, that the liquefaction phenomenon is occuring much earlier compared to
the 𝑘 = 10−5 𝑚/𝑠 case in which the entire fill had liquefied at 57𝑠.

k=10-3m/s

Between 11𝑠 and 14𝑠 the rate of excess pore pressures generation is increased compared to
the analyses with 𝑘 = 10−6 𝑚/𝑠, 𝑘 = 10−5 𝑚/𝑠 and 𝑘 = 10−4 𝑚/𝑠 (Figure 5.9). This increase
originates from the top fill part and propagates downwards through the soil column. At 16𝑠
the liquefaction condition (𝑟𝑢 ≈ 1) is reached in the top fill part. Moreover, between 16𝑠
and 24𝑠, the liquefaction state is even more uniformly propagated along the entire fill part

85 | P a g e
without forming discrete or scattered zones throughout and at 24𝑠 the entire hydraulic fill
has liquefied.

Figure 5.14: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 40% for
dynamic analysis without consolidation

Dr=50%

 (ii) 𝑓𝑓𝑢𝑛𝑑 = 0.37𝐻𝑧, 𝛼𝑚𝑎𝑥 = 0.16𝑔

k=10-6m/s

Between 35𝑠 and 54𝑠 an increase in excess pore pressures is identified mainly in the mid-
upper hydraulic fill layer that propagates upwards and downwards through the soil
column (Figure 5.15). It is important to note, that the excess pore pressure generation is
carried out in a considerably slower rate than in the loosely-packed fills (𝐷𝑅 = 30%, 𝐷𝑅 =
40%). Thus, the apparent relative density 𝐷𝑅 , as an input parameter, has a crucial role in
the generation of excess pore pressure and hence in the liquefaction triggering of the fill.
After a rapid increase in excess pore pressures at 57𝑠 the liquefaction condition (𝑟𝑢 ≈ 1)
is reached in the middle fill part. Between 57𝑠 until the end of the earthquake (100𝑠) a
slight increase in excess pore pressures is observed around the aforementioned liquefied
layers. As described before, the excess pore pressures developed in the liquefied layers
tend to dissipate upwards and downwards along the soil column and hence, spread the
liquefaction condition (𝑟𝑢 ≈ 1) throughout the fill column. However, the magnitude of
the permeability coefficient is very low (𝑘 = 10−6 𝑚/𝑠). Therefore, there is not enough
time for drainage and for a more extensive diffusion of the liquefaction state throughout
the fill layer. As a result, at the end of the earthquake (100𝑠) the liquefaction is restricted
in a singular continuous layer. Similarly, the current dynamic analysis with consolidation,
for an extremely low value of permeability coefficient (10−6 𝑚/𝑠), nearly resembles the
undrained response for the corresponding dynamic analysis without consolidation (Figure
5.16).

86 | P a g e
k=10-5m/s

Between 35𝑠 and 51𝑠 the generation of excess pore pressures is very similar compared to the
analysis with 𝑘 = 10−6 𝑚/𝑠 (Figure 5.15). However, the impact of the permeability coefficient
has a crucial role after this point. In particular, once the liquefaction state is reached in the
middle fill part (57𝑠) this condition is better diffused towards the top of the fill column and
gently downwards between 57𝑠 and until 89𝑠. Consequently a permeability coefficient of
higher magnitude leads to a higher drainage flow rate and thus a faster dissipation of excess
pore pressures that are developed in the liquefied layers. As a result, the liquefaction
condition (𝑟𝑢 ≈ 1) is more uniformly diffused throughout the fill layer. At 89𝑠 a continuous
liquefied layer has been formed extending from top to mid-lower fill column. An important
observation at this point is, that at the end of the ground motion (100𝑠) the entire fill has not
liquefied which was actually the case in the corresponding loosely-packed fills (𝐷𝑅 =
30%, 𝐷𝑅 = 40%). This is attributed to the fact, that the higher relative density of the fill
(𝐷𝑅 = 50%) restricts the generation of excess pore pressures and therefore, the spread of
the liquefaction condition to a smaller extent compared to the loosly-packed fills (𝐷𝑅 =
30%, 𝐷𝑅 = 40%).

Figure 5.15: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 50% for
dynamic analysis with consolidation.

87 | P a g e
k=10-4m/s

Between 35𝑠 and 46𝑠 the rate of excess pore pressures generation is increased compared to
the analyses with 𝑘 = 10−6 𝑚/𝑠 and 𝑘 = 10−5 𝑚/𝑠 (Figure 5.15). This increase originates
from the top fill part and propagates downwards through the soil column. At 51𝑠 the
liquefaction condition (𝑟𝑢 ≈ 1) is reached in the top fill part. An important observation here
is, that the liquefaction region has moved upwards at the top part of the fill. Moreover,
between 51𝑠 and 68𝑠, the liquefaction state is propagated uniformly from top until the
middle fill part. Similarly, by applying a permeability coefficient of an even higher magnitude
(𝑘 = 10−4 𝑚/𝑠), the excess pore pressures developed in the liquefied layers are dissipating
more rapidly along the soil column. Hence, a continuous liquefied layer has been formed at
68𝑠 extending from top to middle fill column. Furthermore, the final liquefaction state (68𝑠)
is occuring much earlier compared to the 𝑘 = 10−5 𝑚/𝑠 case in which the same conditions
are observed at 89𝑠.

k=10-3m/s

Between 35𝑠 and 43𝑠 the rate of excess pore pressures generation is increased compared to
the analyses with 𝑘 = 10−6 𝑚/𝑠, 𝑘 = 10−5 𝑚/𝑠 and 𝑘 = 10−4 𝑚/𝑠 (Figure 5.15). This
increase originates from the top fill part and propagates downwards through the soil column.
At 49𝑠 the liquefaction condition (𝑟𝑢 ≈ 1) is reached in a small region at the top fill part.
However, between 49𝑠 until the end of the earthquake (100𝑠), the liquefaction state is not
further diffused along the fill column but remains localized in the top fill part that has
liquefied.

Figure 5.16: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 50% for
dynamic analysis without consolidation

88 | P a g e
Dr=60%

 (i) 𝑓𝑓𝑢𝑛𝑑 = 0.59𝐻𝑧, 𝛼𝑚𝑎𝑥 = 0.16𝑔

k=10-6m/s

Between 35𝑠 and 41𝑠 an increase in excess pore pressures is identified mainly in the lower
hydraulic fill layer that propagates upwards through the soil column (Figure 5.17). In a similar
manner with the case of 𝐷𝑅 = 50%, the excess pore pressure generation is carried out in a
considerable slower rate than in the loosely-packed fills (𝐷𝑅 = 30%, 𝐷𝑅 = 40%). After a
rapid increase in excess pore pressures at 43𝑠 the liquefaction condition (𝑟𝑢 ≈ 1) is reached
in the bottom fill part. Between 43𝑠 until the end of the earthquake (100𝑠) the liquefaction
condition remains localized in the bottom fill part and does not propagate further along the
soil column. Moreover, the magnitude of the permeability coefficient is very low
(𝑘 = 10−6 𝑚/𝑠). Therefore, there is not enough time for drainage and for a more extensive
diffusion of the liquefaction state throughout the fill layer. As a result, at 43𝑠 the liquefaction
condition is restricted to a singular continuous layer at the bottom of the fill and remains
unchanged until the end of the earthquake (100𝑠). Similarly, the current dynamic analysis
with consolidation, for an extremely low value of permeability coefficient (10−6 𝑚/𝑠), nearly
resembles the undrained response for the corresponding dynamic analysis without
consolidation (Figure 5.18).

k=10-5m/s

Between 35𝑠 and 38𝑠 the generation of excess pore pressures is slightly increased compared
to the analysis with 𝑘 = 10−6 𝑚/𝑠 (Figure 5.17). However, the impact of the permeability
coefficient has an important role after this point. In particular, once the liquefaction state is
reached in the bottom fill part (43𝑠) this condition is better diffused upwards until the middle
of the fill column between 43𝑠 and until 46𝑠. Hence, a continuous liquefied layer has been
formed at 46𝑠 extending from bottom to middle fill column. It is important to note, that
between 46𝑠 until the end of the earthquake (100𝑠) the liquefaction condition remains
localized in the aforementioned liquefied fill part and does not propagate further along the
soil column. Consequently, a higher fill density (𝐷𝑅 = 60%) restricts the spread of the
liquefaction condition to an even lesser extent.

89 | P a g e
Figure 5.17: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 60% for
dynamic analysis with consolidation.

k=10-4m/s

Between 35𝑠 and 41𝑠 the rate of excess pore pressures generation is increased compared to
the analyses with 𝑘 = 10−6 𝑚/𝑠 and 𝑘 = 10−5 𝑚/𝑠 (Figure 5.17). This increase originates
from the top fill part and propagates downwards through the soil column. At 43𝑠 the
liquefaction condition (𝑟𝑢 ≈ 1) is reached in the top fill part. Similarly with the 𝐷𝑅 = 50%
case, the liquefaction region has moved upwards at the top fill part. Moreover, between 43𝑠
and 46𝑠, the liquefaction state is propagated uniformly from top until the mid-upper fill part.
It is important to note, between 46𝑠 until the end of the earthquake (100𝑠) the liquefaction
condition remains localized in the aforementioned liquefied fill part which is attributed the
the high relative density of the fill.

90 | P a g e
k=10-3m/s

Between 35𝑠 and 38𝑠 the rate of excess pore pressures generation is increased compared to
the analyses with 𝑘 = 10−6 𝑚/𝑠, 𝑘 = 10−5 𝑚/𝑠 and 𝑘 = 10−4 𝑚/𝑠 (Figure 5.17). This
increase originates from the top fill part and propagates downwards through the soil column.
At 38𝑠 the liquefaction condition (𝑟𝑢 ≈ 1) is reached in a small region at the top fill part.
However, between 38𝑠 until the end of the earthquake (100𝑠), the liquefaction state is not
further diffused along the fill column but remains localized in the top fill part that has
liquefied.

Figure 5.18: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 60% for
dynamic analysis without consolidation

91 | P a g e
5.7 Conclusions

 PM4Sand model is indeed applicable for the prediction of earthquake induced

liquefaction in hydraulic fills
 The empirical nature of the NCEER method is its largest drawback, as it accounts for
varying loading condition and different soil types. Consequently, this method can be
conservative for the design, as for the current thesis project, it predicted liquefaction
for almost all the hydraulic fills (𝐷𝑅 = 30% to 𝐷𝑅 = 60%) analyzed for all different
earthquake motions
 On the other hand, PM4Sand model provides the opportunity to conduct a site-
specific analysis for each case by giving results that are much more realistic to what
has been observed in real practise
 The effect of the in-situ state of the fill has a critical role on the liquefaction
susceptibility in PM4Sand model which is more representative to what has been
observed in reality. In particular, the relative density 𝐷𝑅 , as an input parameter, is
the most predominant factor concerning the liquefaction triggering of hydraulic fills
 According to PM4Sand model, the loosely-packed fills (𝐷𝑅 = 30% and 𝐷𝑅 = 40%)
are indeed more susceptible to liquefaction compared to the densely-packed fills
(𝐷𝑅 = 50% and 𝐷𝑅 = 60%) which showed less or even no liquefaction potential due
to the earthquake events
 Regarding the dynamic analyses with consolidation, the results related to the
earthquake-induced liquefaction of the fills is more representative to realistic
conditions as there is a better distribution of excess pore pressures along the soil
column
 Regarding the dynamic analyses without consolidation, there is a better diffusion of
excess pore pressures in the loosely-packed fills (𝐷𝑅 = 30% and 𝐷𝑅 = 40%) for the
signals of low dominant frequencies regardless the peak ground acceleration values
of the input signal. In the densely-packed fills (𝐷𝑅 = 50%and 𝐷𝑅 = 60%) the same
phenomenon takes pace more for the signals of high dominant frequencies. In the
rest of the signals there is localization of liquefied zones in distinct parts within the fill
layer

92 | P a g e
Bibliography

Ambraseys, N. (1988). Engineering Seismology. Int. J. Earth-Geology-University of

Thessaloniki and Institute of quake Eng. Struct.Dyn., 101-105.

Bastidas, A. M. (2016). Ottawa F-65 Sand Characterization. University of California at Davis:

Department of Civil and Environmental Engineering.

Been, K., & Jefferies, M. (1985). A state parameter for sands. Geotechnique 35, (2), 99-112.

Bolton, M. (1986). The strength and dilatancy of sands. Geotechnique 36, (1), 65-78.

Boulanger, R. W. (2003a). Relating Kα to relative state parameter index. J. Geotechnical and

Environmental Eng., ASCE 129(8), 770-73.

Boulanger, R. W., & Ziotopoulou, K. (2015). PM4Sand (Version3): A Sand Plasticity Model for
Earthquake Engineering Applications. Center for Geotechnical Modeling Department
of Civil and Environmental Engineering, University of California Davis, California.

Brinkgreve, R., Engin, E., & Engin, H. (2010). Validation of empirical formulas to derive model
parameters for sands. Technical Report, TU Delft, Plaxis.

Casagrande, A. (1936). Characteristics of Cohesionless Soils Affecting the Stability of Slopes

and Earth Fills. Journal of the Boston Society of Civil Engineers, 257-276.

Chang, F., & Krinitzsky, E. (1977). Duration, spectral content and predominant period of
strong motion earthquake records from western United States. US Army Corps of
Engineers Waterways Experiment Station.

Dafalias, Y., & Manzari, M. (2004). Simple plasticity sand model accounting for fabric change
effects. Journal of Engineering Mechanics, ASCE, 130(6), 622-634.

Dyvik, R. T. (1987). Comparison of truly undrained and constant volume direct simple shear
test. Geotechnique 37(1), 3-10.

Gonzalo , C. (4234). likvefaction. 118(3), 2342-2323.

Hanks, T., & McGuire, R. (1981). The character of high frequency strong ground motion. Bull.
Seism. Soc. Am. 120, (6)(25), 2071-2095.

Hicks, M., & Boughrarou, R. (1998). Finite element analysis of the Nerlerk underwater berm
failures. Geotechnique 35, (2), 169-185.

Hicks, M., & Onisiphorou, C. (2005). Stochastic evaluation of static liquefaction in a

predominantly dilative sand fill. Geotechnique 55, (2), 123-133.

Idriss, I. (1999). An Update to the Seed-Idriss Simplified Procedure for Evaluating

Liquefaction Potential. Proceedings of TRB Workshop on New Approaches to
Liquefaction, Federal Highway Administration. Washington DC, 10 January 1999.

93 | P a g e
Idriss, I., & Boulanger, R. (2008). Soil Liquefaction During Earthquakes. USA: Lynx
Communication Group Inc.

Isihara, K. (1984). Post-earthquake failure of tailings, journal=Soil and Foundation.

Isihara, K. (1985). Stability of natural deposits during earthquakes. 11th Int. Conf. on Soil
Mechanics and Foundation Engineering.

Isihara, K. (1996). Soil behaviour in Eartqhquake Geotechnics. The Oxford Engineering

Science Series, (46).

Konrad, J.-M. (1988). Interpretation of flat plate dilatometer tests in sands in terms of the
state parameter. Geotechnique 38, (2), 263-77.

Kramer, S. (1996). Geotechnical Earthquake Engineering. Prentice Hall-Inc, New Jersey, USA.

Kuhlmeyer, R., & Lysmer, J. (1973). Finite element method accuracy for wave propagation
problems. Technical Report, Journal of the soil mechanics and foundation division,
99, 421-427.

Li, X., & Wang, Y. (1998). Linear representation of steady-state line for sand. Journal of
Geotechnical and Geoenvironmental Engineering, 124(12), 1215-7.

Nemat, & Tobita. (1982). Influence of fabric on liquefaction and densification potential of
cohesionless sand. 43-62.

Pyke , R., Chan, C., & Seed, H. (1974). Settlement and Liquefaction of Sands under Multi-
directional Shaking. Earthquake Engineering Research Center, No. EERC 74-2.
University of California at Berkley.

Robertson, P. (2009). Evaluation of Flow Liquefaction and Liquefied Strength. Journal of

Geotechnical and Geoenvironmental Engineering, 842-853.

Robertson, P., & Wride, C. (1998). Evaluating cyclic liquefaction potential using the cone
penetration test. Canadian Geotechnical Journal 35, (3), 442-459.

Saada, A. (1988). Hollow cylinder torsional devices: their advantages and limitations. In
Symposium on advanced triaxial testing of soil and rock. Edited by R.T. Donaghe, R.C.
Chaney, and Marshall L. Silver. American Society for Testing and Materials,Special
Technical Publication 977, (pp. 766-795).

Schofield, A., & Wroth, C. (1968). Critical State Soil Mechanics. New York, McGraw-Hill.

Seed, H. (1976). Evaluation of Soil Liquefaction Effects on Level Ground During Earthquakes.
American Society of Civil Engineers, presentet at ASCE National Convention,
Philadelphia, 1-104.

Seed, H. (1983). Earthquake resistant design of earthquake dams, in. Proceedings,

Symposium on Seismic Design of Embankments and Caverns, (pp. 41-64).
Pennsylvania, ASCE, NY.

94 | P a g e
Seed, H., & De Alba, P. (1986). Use of CPT and SPT Test for evaluating the liquefaction
resistance of soils. Use of in situ Tests in Geotechnical Engineering, 281-302.

Seed, H., & Idriss. (1982). Ground motions and soil liquefaction during earthquakes: Berkley.
Earthquake Engineering Research Institute Monograph.

Seed, H., & Idriss, I. (1968). AN ANALYSIS OF GROUND MOTION DURING THE 1957 SAN
FRANCISCO EARTHQUAKE. Bulletin of the Seismological Society of America 58, (6).

Seed, H., & Idriss, I. (1971). Simplified procedure for evaluating the liquefaction potential.
Journal of Soil Mechanics and Foundation Division 97, (9), 1249-1273.

Seed, H., & Lee, K. (1996). Liquefaction of saturated sand during cyclic loading. Journal of Soil
Mechanics and Foundation Engineering, 105-134.

Uthayakumar, M., & Vaid, Y. (1998). Static liquefaction of sands under multiaxial loading.
Canadian Geotechnical Journal 35, 273-283.

Vaid, Y., & Finn, W. (1979). Static shear and liquefaction potential. J. Geotechnical Div. ASCE
105, (GT10), 1233-246.

Vaid, Y., & Sivathayalan, S. (1996). Static and cyclic liquefaction potential of Fraser Delta
Sand in simple shear and triaxial tests. Canadian Geotechnical J., (33), 281-89.

Van't Hoff, J., & Nooy van der Kolff, A. (2012). Hydraulic Fill Manual: For Dredging and
Reclamation Works. CRC Press, Taylor and Francis Group, The Netherlands.

Youd, T., & Idriss, I. (2001). "Liquefaction resistance of soils: summary report from the 1996
NCEER and 1998 NCEER/NSF workshops on evaluation of liquefaction resistance of
soils. Journal of Geotechnical and Geoenvironmental Engineering, 127(4), 297-313.

95 | P a g e
APPENDIX A

A.1 Earthquake motions

Figure A.1: Filtered and corrected acceleration-time record with Seismosignal software,
measured in the north-south direction and scaled at 0.16g, (i)

Figure A.2: Filtered and corrected acceleration-time record with Seismosignal software,
measured in the east-west direction and scaled at 0.16g, (ii)

96 | P a g e
Figure A.3: Filtered and corrected acceleration-time record with Seismosignal software,
measured in the north-south direction and scaled at 0.16g, (iii)

Figure A.4: Filtered and corrected acceleration-time record with Seismosignal software,
measured in the north-south direction and scaled at 0.16g, (iv)

97 | P a g e
Figure A.5: Filtered and corrected acceleration-time record with Seismosignal software.,
measured in the north-south direction and scaled at 0.16g, (v)

Figure A.6: Filtered and corrected acceleration-time record with Seismosignal software,
measured in the east-west direction and scaled at 0.16g, (vi)

98 | P a g e
Figure A.7: Filtered and corrected acceleration-time record with Seismosignal software,
measured in the east-west direction and scaled at 0.16g, (vii)

Figure A.8: Frequency amplitude spectrum of the input acceleration ground motion (𝑃𝐺𝐴 =
0.16𝑔) with Seismosignal software, (i)

99 | P a g e
Figure A.9: Frequency amplitude spectrum of the input acceleration ground motion (𝑃𝐺𝐴 =
0.16𝑔) with Seismosignal software, (ii)

Figure A.10: Frequency amplitude spectrum of the input acceleration ground motion
(𝑃𝐺𝐴 = 0.20𝑔) with Seismosignal software, (iii)

100 | P a g e
Figure A.11: Frequency amplitude spectrum of the input acceleration ground motion
(𝑃𝐺𝐴 = 0.16𝑔) with Seismosignal software, (iv)

101 | P a g e
Figure A.12: Frequency amplitude spectrum of the input acceleration ground motion
(𝑃𝐺𝐴 = 0.22𝑔) with Seismosignal software, (v)

Figure A.13: Frequency amplitude spectrum of the input acceleration ground motion
(𝑃𝐺𝐴 = 0.22𝑔) with Seismosignal software, (vi)

102 | P a g e
Figure A.14: Frequency amplitude spectrum of the input acceleration ground motion
(𝑃𝐺𝐴 = 0.14𝑔) with Seismosignal software, (vii)

103 | P a g e
A.2 Liquefaction Potential according to NCEER Method

Table A.15: Liquefaction susceptibility analysis for the hydraulic fill of 𝐷𝑅 = 30% and the
whole soil column using the NCEER method

104 | P a g e
Table A.16: Liquefaction susceptibility analysis for the hydraulic fill of 𝐷𝑅 = 30% using the
NCEER method for seven different earthquake motions.

Table A.17: Liquefaction susceptibility analysis for the hydraulic fill of 𝐷𝑅 = 50% using the
NCEER method

Table A.18: Liquefaction susceptibility analysis for the hydraulic fill of 𝐷𝑅 = 60% using the
NCEER method

105 | P a g e
 Depth(m) σ’ν p' σ’h γunsat ρ m G o,ref Go vs vs,average fmax Δlmax Δtmax
Soil column
(m) (kPa) (kPa) (kPa) (kN/m3) (kg/m3) m (kPa) (kpA) (m/s) (m/s) (Hz) (m) (s)
0 0 0 0 0 102
DR=60% 0,5 9 7 4 17,4 1,77 0,5125 100800 18472 102 109 11 1,23 0,011
1 17 13 7 26350 122
1,5 14 11 6 14144 93
2 19 14 9 16333 99
2,5 24 18 11 18260 105
3 28 21 13 20003 110
3,5 33 25 15 21606 114
4 38 28 17 23098 118
4,5 43 32 19 24499 122
5 47 36 22 25824 125
5,5 52 39 24 27085 128
6 57 43 26 28289 131
DR=30% 16,2 1,65 _____ _____ 129 11 1,47 0,011
6,5 62 46 28 29444 134
7 66 50 30 30555 136
7,5 71 53 32 31628 138
8 76 57 35 32665 141
8,5 81 60 37 33670 143
9 85 64 39 34647 145
9,5 90 68 41 35596 147
10 95 71 43 36521 149
10,5 100 75 45 37423 151
11 104 78 47 38303 152
11,5 116 87 49 76182 205
12 121 91 51 77759 207
12,5 127 95 53 79302 209
13 132 99 55 80813 211
13,5 137 102 57 82294 213
14 142 106 59 83747 215
14,5 147 110 62 85173 217
15 152 114 64 86574 218
15,5 157 118 66 87951 220
16 162 121 68 89305 222
16,5 167 125 70 90638 224
17 172 129 72 91949 225
17,5 177 133 74 93241 227
18 182 137 76 94514 228
DR=70% 18,5 187 140 79 17,8 1,81 0,4813 107600 95768 230 229 11 2,61 0,011
19 192 144 81 97005 231
19,5 197 148 83 98225 233
20 202 152 85 99429 234
20,5 207 156 87 100618 235
21 213 159 89 101792 237
21,5 218 163 91 102951 238
22 223 167 93 104096 240
22,5 228 171 95 105228 241
23 233 175 98 106347 242
23,5 238 178 100 107454 243
24 243 182 102 108548 245
24,5 248 186 104 109630 246
25 253 190 106 110701 247
25,5 258 194 108 111761 248
26 263 197 110 112811 249
26,5 272 204 105 116790 251
27 278 208 107 117776 252
27,5 283 212 109 118753 253
28 288 216 111 119720 254
28,5 293 220 113 120677 255
29 298 224 115 121625 256
29,5 303 227 117 122564 257
30 308 231 119 123495 258
30,5 314 235 121 124417 259
31 319 239 122 125331 260
31,5 324 243 124 126236 261
32 329 247 126 127134 262
32,5 334 251 128 128024 263
33 339 254 130 128907 264
33,5 344 258 132 129782 264
DR=80% 18,2 1,86 0,45 114400 265 11 3,01 0,011
34 350 262 134 130650 265
34,5 355 266 136 131511 266
35 360 270 138 132366 267
35,5 365 274 140 133213 268
36 370 278 142 134054 269
36,5 375 281 144 134889 270
37 380 285 146 135717 270
37,5 386 289 148 136540 271
38 391 293 150 137356 272
38,5 396 297 152 138166 273
39 401 301 154 138971 274
39,5 406 305 156 139770 274
40 411 308 158 140563 275
40,5 416 312 160 141351 276
41 421 316 162 142134 277

Table A.19: Calculation of the average shear wave velocity, maximum element size and
maximum time step for the hydraulic fill with 𝐷𝑅 = 30% and for the whole soil column.

106 | P a g e
Calculation of CRR7.5

The CRR in Figure 2.14a can analytically be expressed by the following equation proposed by
Rauch (1998) (Youd & Idriss, 2001):

1 (𝑁1 )60 50 1
𝐶𝑅𝑅7.5 = + + 2
− (𝐴. 1)
(𝑁 )
34 − 1 60 135 [(10 (𝑁 ) )
∙ 1 60 + 45] 200

Calculation of rd

The stress reduction coefficient 𝑟𝑑 is unity at surface and reduces with depth to account for
the subsurface ability to deform. The variations of 𝑟𝑑 with depth for earthquakes of different
magnitude are depicted in Figure 2.8 based on expressions proposed by (Idriss, 1999):

𝑟𝑑 = 𝑒𝑥𝑝[𝑎(𝑧) + 𝛽(𝑧) ∙ 𝑀] (𝐴. 2)

𝑧
𝑎(𝑧) = −1.012 − 1.126 sin ( + 5.133) (𝐴. 3)
11.73
𝑧
𝛽(𝑧) = 0.106 + 0.118 sin ( + 5.142) (𝐴. 4)
11.28

Kα correction factor

Figure A.20: Κα factor for sands of different relative densities and overburden stresses for an
increasing static shear stress ratio (Boulanger, 2003a)

Moreover, Boulanger (2003a) introduced the following relationship to capture these effects:

−𝜉𝑅
𝐾𝑎 = 𝑎 + 𝑏𝑒 𝑐 (𝐴. 5)

𝑎 = 1267 + 636𝑎2 − 634𝑒 𝑎 − 632𝑒 −𝑎 (𝐴. 6)

2 +1.31𝑙𝑛(𝑎+0.0001)
𝑏 = 𝑒 −1.11+12.3𝑎 (𝐴. 7)

𝑐 = 0.138 + 0.126𝑎 + 2.52𝑎3 (𝐴. 8)

107 | P a g e
where 𝑎 is the static shear stress ratio, 𝜉𝑅 is the relative state index which is further elaborated
in Chapter 2.8.2.

Correlation between CRR from lab tests to field conditions

1 + 2𝛫𝜊
𝐶𝑅𝑅(𝐾𝑜 ≠1) = 𝐶𝑅𝑅(𝐾𝑜 =1) (𝐴. 9)
3
1 + 2(𝛫𝜊 )𝑆𝑆
𝐶𝑅𝑅𝑆𝑆 = ( ) 𝐶𝑅𝑅𝑇𝑋 (𝐴. 10)
3

1 + 2(𝛫𝜊 )𝑓𝑖𝑒𝑙𝑑
𝐶𝑅𝑅𝑓𝑖𝑒𝑙𝑑 = 0.9 ( ) 𝐶𝑅𝑅𝑇𝑋 (𝐴. 11)
3

1 + 2(𝛫𝜊 )𝑓𝑖𝑒𝑙𝑑
𝐶𝑅𝑅𝑓𝑖𝑒𝑙𝑑 = 0.9 ( ) 𝐶𝑅𝑅𝑆𝑆 (𝐴. 12)
1 + 2(𝛫𝜊 )𝑆𝑆

Correlation between earthquake magnitude Mw and number of equivalent stress sycles

Figure A.21: Correlation between earthquake moment magnitude M and number of

equivalent stress cycles

108 | P a g e
Α.3 Results for the dynamic analysis with consolidation regarding the

generation of excess pore pressures

Figure A.22: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 30% for
dynamic analysis with consolidation – Comparison with the corresponding dynamic analysis
without consolidation

109 | P a g e
Figure A.23: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 30% for
dynamic analysis with consolidation – Comparison with the corresponding dynamic analysis
without consolidation

Figure A.24: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 30% for
dynamic analysis with consolidation – Comparison with the corresponding dynamic analysis
without consolidation

110 | P a g e
Figure A.25: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 40% for
dynamic analysis with consolidation – Comparison with the corresponding dynamic analysis
without consolidation

Figure A.26: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 40% for
dynamic analysis with consolidation – Comparison with the corresponding dynamic analysis
without consolidation

111 | P a g e
Figure A.27: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 50% for
dynamic analysis with consolidation – Comparison with the corresponding dynamic analysis
without consolidation

Figure A.28: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 60% for
dynamic analysis with consolidation – Comparison with the corresponding dynamic analysis
without consolidation

112 | P a g e
Figure A.29: Excess pore pressure ratio development in the hydraulic fill of 𝐷𝑅 = 60% for
dynamic analysis with consolidation – Comparison with the corresponding dynamic analysis
without consolidation

113 | P a g e
APPENDIX B

Dafalias & Manzari, 2004 - Model Formulation

B.1 Elastic/Plastic part

In the triaxial formulation (Dafalias & Manzari, 2004), the elastic and plastic strain increments
are given by the following equations:

𝑑𝑞 𝑑𝑝′
𝑑𝜀𝑞𝑒𝑙 = , 𝑑𝜀𝑣𝑒𝑙 = (𝐵. 1)
3𝐺 𝐾

𝑝𝑙 𝑑𝜂 𝑝𝑙 𝑝𝑙
𝑑𝜀𝑞 = , 𝑑𝜀𝑣 = 𝑑|𝑑𝜀𝑞 | (𝐵. 2)
𝐻

where 𝐺 is the elastic shear modulus, 𝐾 is the elastic bulk modulus, 𝐻 is the plastic modulus
and 𝑑 is the dilatancy which will be described later on.

Figure B.1: Schematic illustration of yield, bounding and dilatancy lines in the q-p space

The parameters 𝐺 and 𝐾 are dependent on both the current stress state and the density of
the soil according to:

𝑝 (2.97 − 𝑒)2
𝐺 = 𝐺𝑜 𝑝𝐴 √ ∙ (𝐵. 3)
𝑝𝐴 1+𝑒

2(1 + 𝜈)
𝐾= 𝐺 (𝐵. 4)
3(1 − 2𝜈)

where 𝑝 is the current means stress and 𝑒 is the void ratio. Moreover, 𝐺𝑜 is a constant, 𝜈 is
the Poisson’s ratio and 𝑝𝐴 is the atmospheric pressure. It is clear that e.g for dense sands, a
lower void ratio (𝑒) results in a higher value of 𝐺 in eq.(B.3) which through eq.(B.4) gives a
larger bulk modulus 𝐾.

B.2 Critical state behaviour

The soil at critical state is described by the following relation (Li & Wang, 1998):

114 | P a g e
 𝑝𝑐 𝛽
𝑒𝑐 = 𝑒𝑜 − 𝜆𝑐 ( ) (𝐵. 5)
𝑝𝐴

where 𝑒𝑐 is the void ratio at critical state and 𝜆𝑐 , 𝛽 are constants.

The state parameter 𝛹 which represents a generalization of the current state with respect to
the critical state is introduced by (Been & Jefferies, 1985) as:

𝛹 = 𝑒 − 𝑒𝑐 (𝐵. 6)

Figure B.2: Demonstration of the critical state line (Li & Wang, 1998).

In general, loose sands have 𝛹>0 and may be susceptible to liquefaction whereas dense sands
have 𝛹<0 and are less prone to liquefaction.

B.3 Bounding, Dilatancy & Critical surfaces

The model embodies bounding (𝑀𝑏 ), dilatancy (𝑀𝑑 ) and critical surfaces (𝑀) (Figure B.1).
During the loading process the quantities 𝑀𝑏 and 𝑀𝑑 are varying with the material state in
order to have 𝑀𝑏 = 𝑀𝑑 = 𝑀 = 0 when 𝑒 = 𝑒𝑐 . Moreover, for a denser-than-critical state
(𝛹 < 0) one has 𝑀𝑑 < 𝑀 < 𝑀𝑏 while for a looser-than-critical state (𝛹 > 0) one has 𝑀𝑑 >
𝑀 > 𝑀𝑏 . This can be seen from the following expressions (Li & Dafalias 2000):

𝑀𝑏 = 𝑀𝑒𝑥𝑝(−𝑛𝑏 𝛹) (𝐵. 7)

𝑀𝑑 = 𝑀𝑒𝑥𝑝(𝑛𝑑 𝛹) (𝐵. 8)

where 𝑛𝑏 , 𝑛𝑑 are constants. It is clear that, from eq.(B.7) and eq.(B.8) as 𝛹 → 0 => 𝑀𝑏 , 𝑀𝑑 →
𝑀.

The bounding surface (𝑀𝑏 ) represents the onset of softening behaviour and the progressive
soil shearing towards critical state. In dense sands, the stress state is reaching the bounding
surface to represent the explicit peak in hardening which is followed by sudden softening. On

115 | P a g e
the other hand, in loose sands, the bounding surface lies on or close to the critical surface (𝑀)
so as to show hardening behaviour towards failure without any dilation.

The dilatancy surface (𝑀𝑑 ) represents the boundary between contractive and dilative
volumetric behaviour, also called as phase transformation line (PT). In dense sands, this
surface lies below the critical surface (𝑀) in which the dilation leads to a peak behaviour. On
the other hand, in loose sands, the dilatancy surface lies close or on to the critical surface
because there is no dilation.

The critical state surface (𝑀) represents the stress ratio at which = 𝑒𝑐 (𝛹 = 0). After this
state is reached, the soil deforms continuously under constant volume.

B.4 Yield surface

The yield surface is defined by the following equation:

𝑓 = |𝜂 − 𝑎| − 𝑚 = 0 (𝐵. 9)

which represents a wedge in the 𝑞 − 𝑝 space (Figure B.1). The parameters 𝑎 and 𝑚 stand for
the center and the size of the yield surface respectively.

B.5 Plastic modulus & Hardening/Softening rule

The plastic modulus 𝐻 depends on the difference between the bounding stress ratio (𝑀𝑏 )
and the current stress ratio (𝜂) according to the following equation:

𝐻 = ℎ(𝑀𝑏 − 𝜂) (𝐵. 10)

which represents triaxial loading in compression. In triaxial extension, the plastic modulus is
rewritten as 𝐻 = ℎ(𝑀𝑏 + 𝜂).

The hardening coefficient ℎ is given by the following equation:

𝑏𝑜 𝑝 −1⁄2
ℎ= , 𝑏𝑜 = 𝐺𝑜 ℎ𝑜 (1 − 𝑐ℎ 𝑒) ( ) (𝐵. 11)
|𝜂 − 𝜂𝑖𝑛 | 𝑝𝐴

Moreover, ℎ𝑜 and 𝑐ℎ are scalar parameters, 𝜂𝑖𝑛 is the value of 𝜂 at the beginning of loading
and is updated when a loading reversal occurs. It is clear that e.g for dense sands, a lower
void ratio (𝑒) results in a higher value of 𝑏𝑜 which through eq.(B.11)1 gives a larger plastic
modulus 𝐻 in eq.(B.10).

According to the dilatancy theory (Rowe, 1962), the dilatancy (𝑑) is determined by the
‘distance’ of the dilatancy stress ratio (𝑀𝑑 ) and the current stress ratio (𝜂) as:

𝑑 = 𝐴𝑑 (𝑀𝑑 − 𝜂) (𝐵. 12)

where 𝐴𝑑 is a function of state. In triaxial extension, the dilatancy is rewritten as 𝑑 =

ℎ(𝑀𝑏 + 𝜂).

116 | P a g e
If (𝑀𝑑 − 𝜂) < 0, 𝑑 is also negative and plastic volumetric dilation occurs. If (𝑀𝑑 − 𝜂) > 0, 𝑑
is also positive and plastic volumetric contraction occurs.

The hardening, softening and failure responses are represented through the plastic modulus
𝐻 in which 𝐻 > 0, 𝐻 < 0 and 𝐻 = 0 respectively in eq.(B.10). For example, when 𝜂 > 𝑀𝑏 ,
softening response occurs with the current stress-ratio 𝜂 moving outside the bounding
surface. This results in negative (𝑀𝑏 − 𝜂) as the bounding surface contracts until critical state
is reached whereby 𝜂 → 𝑀𝑏 → 𝑀 as 𝛹 → 0.

B.6 Effective stress reduction due to increased dilatancy

After decomposing the total volumetric strain into elastic and plastic part for undrained
loading:

𝑝𝑙 (2.27)2 ,(2.28)2
𝑑𝜀𝑣 = 𝑑𝜀𝑣𝑒𝑙 + 𝑑𝜀𝑣 = 0 ⇒

𝑑𝑝′ 𝑝𝑙 (2.28)1 𝑑𝑝′ 𝑑𝜂

+ 𝑑|𝑑𝜀𝑞 | = 0 ⇒ + 𝑑 | | = 0 =>
𝐾 𝐾 𝐻
𝑑𝜂
𝑑𝑝′ = −𝑑 | |𝐾 (𝐵. 13)
𝐻

117 | P a g e
APPENDIX C

PM4Sand Model Formulation

C.1 Critical State Soil Mechanics Framework

The PM4Sand model is implemented according to critical state soil mechanics framework
(Bolton, 1986). In this formulation, the relative state parameter ξR (Boulanger, 2003a) is used
instead of the parameter Ψ (Been & Jefferies, 1985). The ξR is the state parameter 𝛹 in
eq.(B.6) normalized by the difference between the maximum void ratio (emax) and the
minimum void ratio (emin) that are used to define relative density (𝐷𝑅 ). This provides
improved correlation regarding the shearing behaviour of sands (Konrad, 1988).

The idea behind using the concept of relative state parameter ξR (defined in terms of relative
density) instead of the parameter Ψ (defined in terms of void ratio) is straightforward. For a
geotechnical engineer it is more convenient to refer to relative density (𝐷𝑅 ) rather than void
ratio (𝑒) in order to distinguish between loose and dense sands.

The determination of the CSL requires extended experimental data along with the emax, emin
and the in-situ void ratio. Moreover, in this formulation, the out-of-plane stress is ignored in
the calculations (e.g of the mean stress) which makes the realistic representation of the CSL
complicated.

Therefore, the critical state line is ‘empirically’ determined according to the following
equation:

𝑅
𝐷𝑅,𝐶𝑆 = (𝐶. 1)
𝑝’
𝑄 − 𝑙𝑛 (100 )
𝑝’𝐴

The relative state parameter ξR is simply defined as the difference between the relative
density at critical state (𝐷𝑅,𝐶𝑆 ) and the current relative density (𝐷𝑅 ) as:

𝜉𝑅 = 𝐷𝑅,𝐶𝑆 − 𝐷𝑅 (𝐶. 2)

Regarding eq.(C.1) and eq.(C.2): 𝑝’ is the current mean effective stress and 𝑝𝐴 is the
atmospheric pressure. The parameters 𝑄 and 𝑅 were shown by (Bolton, 1986) to be about 10
and 1.0 respectively, for quartzitic sands. The parameter 𝑄 defines the mean stress level at
which the CSL shifts sharply downwards due to considerable particle crushing (Boulanger,
2003a). By increasing the value of 𝑅 to 1.5 a better approximation is provided for Direct
Simple Shear (DSS) tests (Boulanger & Ziotopoulou, 2015). An example of the critical state
line in the DR-p plane with the parameters 𝑄 = 10 and 𝑅 = 1.5 and the effect of changes in
𝑄 and 𝑅 on the critical state line are illustrated in Figure C.1a and Figure C.1b correspondingly.

118 | P a g e
Figure C.1 a: Illustration of the CSL in the DR-p’ space for Q=10 and R=1.5

Figure C.1 b: Influence of Q and R on the position of the CSL

119 | P a g e
C.2 Bounding, Dilatancy and Critical Surfaces

The model incorporates bounding, dilatancy and critical surfaces following the form (Dafalias
& Manzari, 2004). The bounding 𝑀𝑏 , dilatancy 𝑀𝑑 and critical state 𝑀 ratios are given by the
following formulas:

𝑀𝑏 = 𝑀𝑒𝑥𝑝(−𝑛𝑏 𝜉𝑅 ) = 2 sin(𝜑𝑝𝑒𝑎𝑘 ) (𝐶. 3)

𝑀𝑑 = 𝑀𝑒𝑥𝑝(𝑛𝑑 𝜉𝑅 ) = 2 sin(𝜑𝑑 ) (𝐶. 4)

𝑀 = 2 sin(𝜑𝑐𝑣 ) (𝐶. 5)

where 𝜑𝑝𝑒𝑎𝑘 is the peak friction angle, 𝜑𝑑 is the friction angle at the onset of dilatancy, 𝜑𝑐𝑣
is the critical state effective friction angle, 𝑀𝑏 is the value of 𝒓 = 𝒔/𝑝 at peak strength, 𝑀𝑑 is
the value of 𝒓 at the onset of dilation, 𝑀 is the value of 𝒓 at critical state strength.

The parameters 𝑀𝑏 and 𝑀𝑑 are defined by the model parameters 𝑛𝑏 and 𝑛𝑑 in relation to
𝑀. During shear loading, 𝜉𝑅 → 0 (shearing towards critical state line) while 𝑀𝑏 , 𝑀𝑑 → 𝑀.

C.3 Yield Surface

The yield surface is illustrated as a small cone in the deviatoric stress space (Figure C.2)2 and
is formulated as:

𝑓 = √(𝒔 − 𝑝𝛼): (𝒔 − 𝑝𝛼) − √1⁄2 𝑝𝑚 = 0 (𝐶. 6)

The back-stress ratio tensor 𝛼 denotes the centre of the yield surface in the deviatoric stress
ratio space and the parameter 𝑚 defines the size of the yield surface. The above expression
can be rewritten in terms of stress ratio (𝑟) as:

1
𝑓 = √(𝒓 − 𝛼): (𝒓 − 𝛼) − √ 𝑚 = 0 (𝐶. 7)
2

In that way, the yield function can be visualized as the distance between the stress ratio 𝑟 and
the back-stress ratio 𝛼 (Figure C.2)2. In the current simplified model, the stresses from
multiaxial formulation (chapter 2.7.2) are generalized in terms of in-plane stress as:
𝜎𝑥𝑥 𝜎𝑥𝑦
𝛔 = (𝜎 𝜎𝑦𝑦 ) (𝐶. 8)
𝑥𝑦

𝜎𝑥𝑥 + 𝜎𝑦𝑦 1 + 𝐾𝑜
𝑝= = 𝜎𝑦𝑦 (𝐶. 9)
2 2
𝜎𝑥𝑥 − 𝑝 𝜎𝑥𝑦
𝒔 = 𝜎 − 𝑝𝜤 = ( 𝜎 𝜎𝑦𝑦 − 𝑝) (𝐶. 10)
𝑥𝑦

𝜎𝑥𝑥 − 𝑝 𝜎𝑥𝑦
𝒔 𝑝 𝑝
𝒓= =( 𝜎 𝜎𝑦𝑦 − 𝑝) (𝐶. 11)
𝑝 𝑥𝑦
𝑝 𝑝

120 | P a g e
Figure C.2: Illustration of bounding, critical, dilatancy and yield lines for a dense-of-critical
state in q-p’ space (left) (Dafalias & Manzari, 2004) and in rxy-ryy stress-ratio space (right)
(Boulanger & Ziotopoulou, 2015).

The bounding and dilatancy surfaces (Appendix B.3) are defined in terms of the image back-
stress ratios 𝛼 𝑏 and 𝛼 𝑑 as:

1
𝛼 𝑏 = √ [𝑀𝑏 − 𝑚]𝒏 (𝐶. 12)
2

1
𝛼 𝑑 = √ [𝑀𝑑 − 𝑚]𝒏 (𝐶. 13)
2

where 𝒏 denotes the deviatoric unit normal to the yield surface as:
𝒓−𝜶
𝒏= (𝐶. 14)
√1 𝑚
2

C.3 Fabric effects

The fabric-dilatancy tensor proposed by (Dafalias & Manzari, 2004) in eq.(2.29) is modified for
the present model as:

𝑐𝑧 〈−𝑑𝜀𝜈𝑝𝑙 〉
𝑑𝒛 = − 𝑧 (𝑧𝑚𝑎𝑥 𝒏 + 𝒛) (𝐶. 15)
1 + 〈2𝑧𝑐𝑢𝑚 − 1〉 𝐷
𝑚𝑎𝑥

where:

𝑑𝑧𝑐𝑢𝑚 = |𝑑𝒛| (𝐶. 16)

This modification was included so as the tensor 𝒛 to evolve with plastic shear strain
𝑝𝑙 𝑝𝑙
(〈−𝑑𝜀𝜈 〉/𝐷) rather than plastic volumetric strain (〈−𝑑𝜀𝜈 〉) that occurs during dilation only

121 | P a g e
(𝐷 < 0). Moreover, the influence of fabric history and the cumulative fabric term are
included in the present formulation.

In eq.(C.16), 𝑧𝑐𝑢𝑚 is the cumulative value of absolute changes of the fabric tensor 𝒛 which
increases with increasing plastic deviatoric strains. The rate of evolution of z in eq.(C.15) is
decreasing with increasing values of zcum, which disables the undrained cyclic stress-strain
response to lock-up into a repeating stress-strain loop and enables the progressive
accumulation of shear strains (Boulanger & Ziotopoulou, 2015).

C.4 Stress reversal and Initial Back-Stress ratio Tensors

The present model keeps track of the initial back stress ratio (𝑎𝑖𝑛 ) in order to include the
effects of loading history. This ratio is updated at the reversal in loading direction which is
identified whenever the following condition holds:

(𝑎 − 𝑎𝑖𝑛 ): 𝒏 < 0 (𝐶. 17)

At the reversal, the initial stress ratio 𝑎𝑖𝑛 is updated to the current one 𝛼. The initial stress
𝑎𝑝𝑝 𝑡𝑟𝑢𝑒
ratio is subdivided into three initial stress ratios, namely the apparent (𝛼𝑖𝑛 ) , true (𝛼𝑖𝑛 )
𝑝
and the previous initial stress (𝛼𝑖𝑛 ) (eq.(C.32)). This is done in order to avoid the over-
stiffening effect at small loading reversals (Boulanger & Ziotopoulou, 2015).

C.5 Elastic Components of the Model

The elastic strain increments are given by (Dafalias & Manzari, 2004) according to the
following equations:

𝑑𝑠 𝑑𝑝
𝑑𝒆𝑒𝑙 = , 𝑑𝜀𝑣𝑒𝑙 = (𝐶. 18)
2𝐺 𝐾

The elastic bulk modulus K is related to the shear modulus G through the Poisson’s ratio ν as:

2(1 + 𝜈)
𝐾= 𝐺 (𝐶. 19)
3(1 − 2𝜈)

The elastic shear modulus (𝐺) is computed according to:

𝑧
𝑝 1 + 𝑐𝑢𝑚
𝑧𝑚𝑎𝑥
𝐺 = 𝐺𝑜 𝑝𝐴 √ 𝐶𝑆𝑅 ( 𝑧 ) (𝐶. 20)
𝑝𝐴 1 + 𝑧𝑐𝑢𝑚 𝐶𝐺𝐷
𝑚𝑎𝑥

where 𝐺𝑜 the small strain shear modulus coefficient influenced by environmental factors, 𝑝𝐴
is the atmospheric pressure,

It is important to note that in this formulation the elastic shear modulus G is not dependent
on the void ratio (𝑒) as initially proposed by (Dafalias & Manzari, 2004). The void ratio (𝑒) was
omitted because: 𝑖) the effects of void ratio variations in 𝐺 are small relatively to the

122 | P a g e
corresponding effects of confining stress, 𝑖𝑖) the calibration of 𝐺 according to shear wave
velocity profiles is accomplished by using only the constant 𝐺𝑜 instead of using both 𝐺𝑜 and
𝑒 (Boulanger & Ziotopoulou, 2015).

Moreover, the elastic shear modulus 𝐺 is dependent on stress ratio effects and fabric history.
The stress ratio effects are captured by the 𝐶𝑆𝑅 factor (Yu & Richart, 1984) according to the
following equation:

𝑀 𝑚𝑆𝑅
𝐶𝑆𝑅 = 1 − 𝐶𝑆𝑅,𝑜 ( ) (𝐶. 21)
𝑀𝑏

𝐶𝑆𝑅,𝑜 and 𝑚𝑆𝑅 are set to 0.5 and 4 internally which keeps the effect of stress ratio (𝒓) smala
on elastic modulus at low stress ratios, but lets the effect increase to a 60% reduction when
the stress ratio is on the bounding surface (𝒓 → 𝑀𝑏 ) (Boulanger & Ziotopoulou, 2015).

The effect of fabric history on the elastic shear modulus G is captured by the last bracketed
term in eq.(C.20). As the plastic shear strains are increasing , 𝑧𝑐𝑢𝑚 is also increasing which
results in the progressive degradation of the shear modulus G. This term was added to take
into account progressive destruction of any minor cementation bonds with increasing plastic
shear strains that resulted in a decreased G after an earthquake event (Boulanger &
Ziotopoulou, 2015).

Similarly, the elastic bulk modulus 𝐾 which is directly proportional to 𝐺 in eq.(C.31) is also
degrading with an increasing 𝑧𝑐𝑢𝑚 . This reduces the rate of strain-hardening at large shear
strains after the phase transformation line is reached and enables the model to approximate
the hysteretic stress-strain response of a soil as it liquefies (Boulanger & Ziotopoulou, 2015).

Moreover, 𝐶𝐺𝐷 is a factor accounting for the shear modulus degradation at large values of
𝑧𝑐𝑢𝑚 (set internally as 2). The parameter 𝑧𝑚𝑎𝑥 is calculated at the initiation of loading by
considering the initial relative state parameter 𝜉𝑅𝑜 as:

𝑧𝑚𝑎𝑥 = 0.7𝑒𝑥𝑝(−6.1𝜉𝑅𝑜 ) ≤ 20 (𝐶. 22)

C.6 Plastic Components of the Model

The increment of plastic volumetric strain along with the plastic deviatoric strain increment is
given as:
𝑝𝑙
𝑑𝜀𝑣 = 〈𝐿〉𝐷 (𝐶. 23)

𝑑𝒆𝑝𝑙 = 〈𝐿〉𝑹′ (𝐶. 24)

where 𝐿 is the loading index, 𝐷 is the dilatancy, R is the direction of depl, R’ is the deviatoric
component of R. Under the assumption of no Lode angle dependency, the tensors R and R’
are:

1
𝑹 = 𝒏 + 𝐷𝑰 (𝐶. 25)
3

123 | P a g e
 1
𝑹′ = 𝑹 − 𝐷𝑰 (𝐶. 26)
3

where n is the unit normal to the yield surface (Figure C.2). Note that, the aforementioned
assumption implies also that R’=n.

The dilatancy 𝐷 (without fabric effects) is given by the following equation:

𝐷 = 𝐴𝑑𝑜 [(𝑎𝑑 − 𝑎): 𝑛] (𝐶. 27)

The loading index 𝐿, as derived from (Dafalias & Manzari, 2004) is:

𝟏 𝝏𝒇 1
𝑳= : 𝒅𝝈 = [𝒏: 𝑑𝑠 − 𝒏: 𝒓𝑑𝑝] =
𝑲𝒑 𝝏𝝈 𝐾𝑝

2𝐺𝒏: 𝑑𝒆 − 𝒏: 𝒓𝐾𝑑𝜀𝑣
= (𝐶. 28)
𝐾𝑝 − 2𝐺 − 𝐾𝐷𝒏: 𝒓

where 𝐾𝑝 is the plastic modulus.

By using the calculated loading index, the stress increment can be calculated as:

𝑑𝝈 = 2𝐺𝑑𝒆 + 𝐾𝑑𝜀𝑣 𝑰 − 〈𝐿〉〈2𝐺𝒏 + 𝐾𝐷𝑰〉 (𝐶. 29)

Plastic Modulus & Hardening/Softening rule

This model accounts for hardening and softening by kinematic rotation of the yield surface in
the stress space. This rotation is accomplished by updating the back-stress ratio (𝑎) which
defines the centre of the yield surface.

The evolution of the back-stress-ratio 𝑑𝜶 is given:

2
𝑑𝛼 = 〈𝐿〉 ℎ(𝑎𝑏 − 𝑎) (𝐶. 30)
3

where h is the hardening coefficient.

The plastic modulus is modified in this formulation and given by the following equation:

√(𝑎𝑏 − 𝑎): 𝑛
𝐾𝑝 = 𝐺ℎ𝑜 𝑎𝑝𝑝
𝐶𝑟𝑒𝑣
[𝑒𝑥𝑝 ((𝑎 − 𝑎𝑖𝑛 ): 𝑛) − 1] + 𝐶𝛾1
𝐶𝑘𝑎
∙ 𝑧𝑝𝑒𝑎𝑘 (𝐶. 31)
1 + 𝐶𝑘𝑝 ( 𝑧 ) 〈(𝑎𝑏 − 𝑎): 𝑛〉√1 − 𝐶𝑧𝑝𝑘2
𝑚𝑎𝑥

where:
𝑎𝑝𝑝
(𝑎 − 𝑎𝑖𝑛 ): 𝑛 𝑎𝑝𝑝
𝐶𝑟𝑒𝑣 = 𝑡𝑟𝑢𝑒
𝑓𝑜𝑟 (𝑎 − 𝑎𝑖𝑛 ): 𝑛 ≤ 0
(𝑎 − 𝑎𝑖𝑛 ): 𝑛

124 | P a g e
 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 𝐶𝑟𝑒𝑣 = 1 (𝐶. 32)

The main difference compared to (Dafalias & Manzari, 2004) is that the plastic modulus
degradation is directly dependent on fabric changes instead of effective stress reduction in
eq.(B.11) as a result of increased dilatancy through eq.(2.30). In particular, the last term in
eq.(C.34) causes a reduction in 𝐾𝑝 whenever the fabric is favorable (𝒛: 𝑛 > 0) and for
increasing plastic shear strains.

Similarly to (Dafalias & Manzari, 2004), the plastic modulus 𝐾𝑝 is proportional to 𝐺 and to the
distance between the current stress state (𝛼) and its image on bounding surface (𝑎𝑏 ) and
𝑎𝑝𝑝
inversely proportional to the difference between 𝛼 and the initial back stress ratio 𝑎𝑖𝑛
(eq.(C.31)).
𝑎𝑝𝑝
The initial back-stress ratio 𝑎𝑖𝑛 is varying between the apparent back-stress ratio 𝑎𝑖𝑛 and
𝑡𝑟𝑢𝑒
the true back-stress ratio 𝑎𝑖𝑛 through the parameter 𝐶𝑟𝑒𝑣 in order to avoid the over-
𝑡𝑟𝑢𝑒
stiffening of the stress-strain response. In particular, the response is stiffer when 𝑎𝑖𝑛 = 𝑎𝑖𝑛
𝑎𝑝𝑝
(larger 𝐾𝑝 ) while a softer response is the case when 𝑎𝑖𝑛 = 𝑎𝑖𝑛 (lower 𝐾𝑝 ). In this way, the
𝑡𝑟𝑢𝑒
model allows for the reloading stiffness to be larger (𝑎𝑖𝑛 = 𝑎𝑖𝑛 ) until the stress state
𝑎𝑝𝑝
exceeds the previous initial back-stress ratio (𝑎𝑖𝑛 = 𝑎𝑖𝑛 ) to avoid the over-stiffening effect
(Boulanger & Ziotopoulou, 2015).

The parameter ℎ𝑜 is used to adjust the ratio between plastic and elastic modulus. This
parameter is internally set as a function of apparent relative density 𝐷𝑅𝑜 as:

(0.25 + 𝐷𝑅𝑜 )
ℎ𝑜 = ≥ 0.3 (𝐶. 33)
2

The parameter 𝐶𝛾1 is a constant which is set as ℎ𝑜 /200 to avoid division by zero.

Plastic volumetric contraction

Plastic volumetric contraction occurs whenever (𝑎𝑑 − 𝑎): 𝑛 > 0. Then from eq.(C.27), 𝐷 is
also positive and given by the following equation:

(𝑎𝑑 − 𝑎): 𝑛
𝐷 = 𝐴𝑑𝑐 [(𝑎 − 𝑎𝑖𝑛 ): 𝑛 + 𝐶𝑖𝑛 ]2 (𝐶. 34)
(𝑎𝑑 − 𝑎): 𝑛 + 𝐶𝐷

(𝑎𝑑 − 𝑎): 𝑛
𝑤𝑖𝑡ℎ 𝐷 ≤ 1.5𝐴𝑑𝑜 (𝐶. 35)
(𝑎𝑑 − 𝑎): 𝑛 + 𝐶𝐷

The effect of fabric on plastic volumetric contraction is captured by the following expression:

𝐴𝑑𝑜 (1 + 〈𝒛: 𝑛〉)

𝐴𝑑𝑐 = (𝐶. 36)
ℎ𝑝 𝐶𝑑𝑧

125 | P a g e
where 𝐴𝑑𝑐 is increasing with the evolution of 𝒛 and therefore enhancing the plastic
volumetric contraction.

Moreover, 𝐶𝐷 is set internally as 0.16, 𝐶𝑖𝑛 is increasing with the evolution of z to enhance
the contraction rate at the start of an unloading cycle and 𝐶𝑑𝑧 improves modelling of the
cyclic strength of denser sands (Boulanger & Ziotopoulou, 2015).

The parameter Ado can be expressed according relationship to the dilatancy (Bolton, 1986)
as:

𝜑𝑝𝑒𝑎𝑘 − 𝜑𝑐𝑣 = −0.8𝜓 (𝐶. 37)

where 𝜑𝑝𝑒𝑎𝑘 is the peak friction angle of shearing resistance,𝜑’𝑐𝑣 the constant volume
friction angle and 𝜓 the dilatancy angle. By further analysing eq.(2.87):

𝑀𝑏 𝑀
sin−1 ( 2 ) − sin−1 ( 2 )
𝐴𝑑𝑜 = 2.5 (𝐶. 37)
𝑀𝑏 − 𝑀 𝑑

The parameter hp is a function of the contraction parameter ℎ𝑝𝑜 and the relative state
parameter 𝜉𝑅 :

ℎ𝑝 = ℎ𝑝𝑜 𝑒𝑥𝑝(−0.7 + 7(0.5 − 𝜉𝑅 )2,5 ) 𝑓𝑜𝑟 𝜉𝑅 ≤ 0.5

ℎ𝑝 = ℎ𝑝𝑜 𝑒𝑥𝑝(−0.7) 𝑓𝑜𝑟 𝜉𝑅 > 0.5 (𝐶. 38)

in which the parameter ℎ𝑝𝑜 can be calibrated to achieve specific CRR values.

From the aforementioned equations, the dilatancy 𝐷 is proportional to the difference 𝑎𝑑 −

𝑎, to the parameter 𝐴𝑑𝑜 and to favorable fabric (𝒛: 𝑛 > 0) and inversely proportional to the
parameter ℎ𝑝𝑜 .

Plastic volumetric dilation

Plastic volumetric dilation occurs whenever (𝑎𝑑 − 𝑎): 𝑛 < 0. Then from eq.(C.27), 𝐷 is also
negative. In this formulation, the rotated dilatancy surface 𝑀𝑑𝑅 is introduced to capture the
effect of fabric evolution on plastic volumetric dilation in order to accommodate earlier
dilation at low stress ratios under certain loading paths (Boulanger & Ziotopoulou, 2015). The
rotated surface (Figure C.3) is determined as:

𝑀𝑑
𝑀𝑑𝑅 = (𝐶. 39)
𝐶𝑟𝑜𝑡1

1
𝛼 𝑑𝑅 = √ [𝑀𝑑𝑅 − 𝑚]𝒏 (𝐶. 40)
2

126 | P a g e
where 𝐶𝑟𝑜𝑡1 is a function of fabric and loading reversals.

Then the dilatancy 𝐷 is calculated in two different ways: for rotated dilatancy surface and for
non-rotated dilatancy surface according to:

〈−𝒛: 𝑛〉 (𝑎𝑑𝑅 − 𝑎): 𝒏

𝐷𝑟𝑜𝑡 = 𝐴𝑑 ∙ (𝐶. 41)
√2𝑧𝑚𝑎𝑥 𝐶𝐷𝑅

𝐷𝑛𝑜𝑛−𝑟𝑜𝑡 = 𝐴𝑑 (−〈−(𝑎𝑑 − 𝑎): 𝒏〉) (𝐶. 42)

𝑖𝑓 𝐷𝑛𝑜𝑛−𝑟𝑜𝑡 < 𝐷𝑟𝑜𝑡 => 𝐷 = 𝐷𝑛𝑜𝑛−𝑟𝑜𝑡

〈𝑀𝑏 − 𝑀𝑐𝑢𝑟 〉
𝑒𝑙𝑠𝑒 𝐷 = 𝐷𝑛𝑜𝑛−𝑟𝑜𝑡 + (𝐷𝑟𝑜𝑡 − 𝐷𝑛𝑜𝑛−𝑟𝑜𝑡 ) (𝐶. 43)
〈𝑀𝑏 − 𝑀𝑐𝑢𝑟 + 0.01〉

Figure C.3: Schematic illustration of the calculation of dilatancy D based on the stress state
with regards to the 𝑀𝑑𝑅 , 𝑀𝑑 and 𝑀𝑏 surfaces during a half-cycle of loading that goes from
contraction to dilation (Boulanger & Ziotopoulou, 2015).

The parameter 𝐴𝑑 is calculated as:

𝐴𝑑𝑜 (𝐶𝑧𝑖𝑛2 )
𝐴𝑑 = 3 (𝐶. 44)
𝑧2 〈−𝒛: 𝑛〉
(𝑧𝑐𝑢𝑚 ) (1 − ) (𝐶𝜀 )2 (𝐶𝑝𝑧𝑝 )(𝐶𝑝𝑚𝑖𝑛 )(𝐶𝑧𝑖𝑛1 ) + 1
𝑚𝑎𝑥 √2𝑧𝑝𝑒𝑎𝑘

where the first term (in the denominator) accounts for the progressive accumulation of strains
under symmetric loading by decreasing the dilatancy, the second term facilitates strain-
hardening when the plastic shear strain reaches the prior peak value, the third term 𝐶𝜀 is a
calibration constant that control the rate of plastic shear strain accumulation, the fourth term
𝐶𝑝𝑧𝑝 causes the effects of fabric on dilation to be diminished whenever the current value of p
is approaching the value of 𝑝𝑧𝑝 , the fifth term 𝐶𝑝𝑚𝑖𝑛 provides a minimum amount of shear

127 | P a g e
resistance for a soil after it has temporarily reached an excess pore pressure ratio of 100%,
the sixth term 𝐶𝑧𝑖𝑛1 facilitates strain-hardening when stress reversals are not causing fabric
changes (Boulanger & Ziotopoulou, 2015).

C.7 Primary Model parameters

The apparent relative density can initially be estimated from CPT or SPT correlations by the
following correlations (Idriss & Boulanger, 2008):

(𝑁1 )60
𝐷𝑅 = √ (𝐶. 45)
𝐶𝑑

0.264
𝑞𝑐1𝑁
𝐷𝑅 = 0.465 ( ) − 1.063 (𝐶. 46)
𝐶𝑑𝑞

where DR is expressed as a ratio rather than a percentage.

The values of Cd and Cdq can be taken according to (Idriss & Boulanger, 2008) as Cd=46. and
Cdq=0.90.

In case laboratory tests are available, the relative density can be defined from the following
expression:
𝑒𝑚𝑎𝑥 − 𝑒
𝐷𝑅 = (𝐶. 47)
𝑒𝑚𝑎𝑥 − 𝑒𝑚𝑖𝑛

where emax and emin are the maximum and the minimum void ratio and e is the void ration at
the current state.

The apparent relative density influences the stress-strain responses through the elastic shear
modulus 𝐺 in terms of stress ratio effects (C.20-C.21). In particular, 𝐷𝑅 is directly proportional
to 𝐶𝑆𝑅 (the ratio 𝑀⁄𝑀𝑏 is decreasing with an increasing 𝐷𝑅 ). Therefore, higher values of 𝐷𝑅
will result in larger shear modulus 𝐺.

The above correlations are used to provide a reasonable estimate for the apparent DR so that
the corresponding model response is also reasonable.

The shear modulus coefficient G0 can be calibrated to match estimated or measured shear
wave velocities along a soil profile, according to:

𝐺 = 𝜌(𝑣𝑠 )2 (𝐶. 48)

Alternatively, Go can be related over a range of typical densities by the following expression
proposed by Boulanger & Ziotopoulou (2017).

𝐺𝑜 = 167√(𝑁1 )60 + 2.5 (𝐶. 49)

128 | P a g e
C.8 Secondary Model Parameters

Bounding surface parameter nb: This parameter affects the distance of the bounding surface
to the critical state through eq.(C.3) and thus the rate with which the bounding surface (𝑀𝑏 )
is approaching the critical surface (𝑀). For a lower nb the bounding surface lies closer to the
critical surface. Consequently, the bounding surface is approaching the critical surface faster
after 𝐫 → 𝑀𝑏 .

The bounding surface parameter controls the dilatancy and thus the peak effective friction
angle. Its default value is 0.50 but for looser-than-critical states a value of nb/4 is used.

Dilatancy surface parameter nd: This parameter affects the distance of the dilatancy surface
to the critical state in eq.(2.57) and represents the stress ratio (𝐫) in which contraction
transitions to dilation. Its default value is 0.10 but for looser-than-critical states a value of 4nd
is used.

Atmospheric pressure pA: Default value of 101.3 kPa is used

Maximum and minimum void ratios, emax and emin: Influences the computation of relative state
parameter ξR and how volumetric strains are translated into changes ξR. The default values
used are 0.8 and 0.5 respectively. These parameters are most likely not refined for practical
problems as the calibration of other parameters will have a bigger impact on monotonic or
cyclic strengths.

Critical state friction angle φ’cv: Defines the position of the critical surface (M) and its default
value is 33o.

Poisson’s ratio ν: Default value of 0.3 is used

Critical state line parameters Q and R: These parameters determine the critical state line in
Figure 2.27 through eq.(2.48). Default values of 10 and 1.5 are used respectively to better
approximate Direct Simple Shear (DSS) tests results (Boulanger & Ziotopoulou, 2015).

129 | P a g e