54 views

Uploaded by mortaza7094

Finite Element Algorithms for Dynamic
Analysis of Geotechnical Problems

Finite Element Algorithms for Dynamic
Analysis of Geotechnical Problems

© All Rights Reserved

- Catalogo de difusores serie 5000
- Ch.08 Friction
- Advanced Mathematics &Mechanics Applications Using MATLAB
- Finite element method 2007 Dec
- Hughes Multiscale 1995
- [Doi 10.1016%2Fj.finel.2004.07.001] Bao-Lin Wang; Zhen-Hui Tian -- Application of Finite Element–Finite Difference Method to the Determination of Tra
- The influence of surface roughness and the contact pressure distribution on friction in rolling/sliding contacts
- Dormit Warangal Designs
- CHAPTER 4.pdf
- 2015_JCAM_LinLiuSadre
- Wheels
- 1-s2.0-S0043164899003154-main
- Undrained Stability of a Square Tunnel in a Soil Whose Strength Increases Linearly With Depth
- Contoh tugas fatigue-fracture mech 2017.xls
- Vizag Steel Plant Written Test Will Comprise General Awareness
- 1. Are Totally Recycled Hot Mix Asphalts a Sustainable Alternative for Road Paving
- Pavement
- 45GTC
- UBC_1990_A1 K66
- 28944_01

You are on page 1of 256

by

Hassan Sabetamal

B.Sc., Civil Engineering

M.Sc., Geotechnical Engineering

Doctor of Philosophy

at the University of Newcastle

Oct 2014

I hereby certify that the work embodied in this thesis is the result of original research and

has not been submitted for a higher degree to any other University or Institute.

I give consent to the final version of my thesis being made available worldwide when

deposited in the Universitys Digital Repository, subject to the provisions of the Copyright

Act 1968.

signed

Acknowledgements

First and foremost, I would like to express my sincere gratitude to my supervisors: Dr. Majid

Nazem, Laureate Prof. Scott Sloan and Prof. John Carter. I have been fortunate to have the

opportunity to work with these highly distinguished people.

I am indebted to Dr. Nazem for his interest and guidance. His commitment, encouragement

and support have been unfailing and limitless throughout the period of this work, and it has

been greatly appreciated.

I would like to express my sincere appreciation to Laureate Prof. Scott Sloan for his support,

suggestions, help and provision of financial assistance during this research.

I have great respect for Prof. John Carter, whose invaluable contributions, suggestions and

encouragement have been greatly appreciated.

I would like to thank all staff members at the ARC Center of Excellence for Geotechnical

Science and Engineering (CGSE), particularly the Centres coordinator, Ms Kirstin

Dunncliff, for their precious help and support.

I would like to thank my mother for her endless love, thoughtfulness, encouragement and

prayers throughout this journey. I thank her for always being patient and positive.

encouraging.

ii

Abstract

The objective of this study is to document the development of a computational procedure for

the analysis of coupled geotechnical problems involving finite deformation, inertia effects

and changing boundary conditions. The procedure involves new finite element (FE)

algorithms that were formulated and implemented into SNACa FE code developed by the

geomechanics group at the University of Newcastle, Australia. The numerical scheme was

then utilised to analyse some important offshore geotechnical problems.

The first development concerns the implementation of the governing equations of two-phase

saturated porous media in a mixed form, allowing predictions of solid displacement, pore

fluid pressure and Darcy velocity. The generalised- method was chosen to integrate the

governing equations in the time domain. The formulation was extended to consider

geometrical nonlinearity within the framework of the Arbitrary LagrangianEulerian

approach. Suitable absorbing boundary conditions were also incorporated to model the

radiation of bulk waves towards infinity at the truncated FE mesh boundaries. Some closedform solutions were also developed, which are suitable to verify the implementation of

dynamic consolidation algorithms.

contact algorithm for solidfluid mixtures accounting for large deformations and inertia

effects. The contact algorithm is based on a mortar segment-to-segment approach formulated

for cases of frictionless and frictional interfaces. The node-to-segment approach was also

employed to compare and highlight the merits of the mortar method when dealing with

dynamic coupled problems.

The computational procedure was evaluated by modelling some numerical exercises and

comparing the predicted results with alternative numerical and analytical solutions where

possible.

In the last part of the thesis, the computational framework was employed to successfully

model the problems of dynamically penetrating anchors and offshore pipeline-seabed

iii

interactions. The analysis of dynamically penetrating anchors comprises the simulation of the

penetration process and consolidation of the soil surrounding the penetrometer. The analysis

of the pipeline-seabed interaction involves the simulation of the laying process and the largeamplitude lateral motion of the pipe.

iv

Contents

Acknowledgements ..................................................................................................................ii

Abstract ................................................................................................................................... iii

Contents .................................................................................................................................... v

Preface ................................................................................................................................... viii

List of Tables and Boxes ........................................................................................................xii

List of Figures ....................................................................................................................... xiii

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

1.1 General ............................................................................................................................. 1

1.2 Scope of Research ............................................................................................................ 3

1.3 Organisation of the Thesis ............................................................................................... 5

Chapter 2: Soil as a Porous Medium - Governing Equations.............................................. 6

2.1 Introduction ...................................................................................................................... 6

2.2 Governing Differential Equations: Balance Laws ........................................................... 7

2.2.1 Balance of mixture mass ........................................................................................... 8

2.2.2 Balance of momentum ............................................................................................ 12

2.2.3 Boundary conditions ............................................................................................... 15

2.3 Variational Statement of the Balance Laws ................................................................... 15

2.4 Finite Element Discretisation ......................................................................................... 19

2.5 Arbitrary Lagrangian-Eulerian Method ......................................................................... 21

2.6 Analytical Solution ........................................................................................................ 24

2.7 Time Integration............................................................................................................. 24

2.7.1 Generalised- method............................................................................................. 26

2.7.2 Discretisation in the time domain ........................................................................... 28

2.8 Absorbing Boundary ...................................................................................................... 31

2.8.1 Adopted energy-absorbing boundary...................................................................... 34

2.8.2 Cone energy-absorbing boundary ........................................................................... 35

2.8.2.1 Implementation ................................................................................................ 40

2.9 Summary ........................................................................................................................ 43

Chapter 3: Interface Modelling: Contact Mechanics of Two-phase Saturated Porous

Media....................................................................................................................................... 45

3.1 Introduction .................................................................................................................... 45

3.2 Formulation of Frictionless Contact .............................................................................. 46

3.2.1 Kinematics at the interface ..................................................................................... 48

3.2.2 Contact interface constraints ................................................................................... 50

3.2.2.1 Displacement contribution ............................................................................... 52

3.2.2.3 Darcy velocity contribution ............................................................................. 58

3.2.3 Augmented Lagrangian regularisation .................................................................... 61

3.3 Formulation of Frictional Contact ................................................................................. 64

3.3.1 Contact kinematic states and moving friction cone ................................................ 65

3.3.2 Linearisation of contact virtual works .................................................................... 71

3.3.2.1 Displacement contribution ............................................................................... 71

3.3.2.2 Darcy velocity contribution ............................................................................. 75

3.3.2.3 Pore-pressure contribution ............................................................................... 77

3.4 Contact Formulation for the U-P Scheme ...................................................................... 78

3.4.1 Displacement contribution ...................................................................................... 79

3.4.2 Pore pressure contribution ...................................................................................... 82

3.5 Summary ........................................................................................................................ 83

Chapter 4: Numerical Evaluations ....................................................................................... 84

4.1 Introduction .................................................................................................................... 84

4.2 Response of One-dimensional Deformable Porous Medium with Incompressible

Constituents .................................................................................................................... 84

4.3 Response of One-dimensional Deformable Porous Medium with Compressible

Pore Fluid ....................................................................................................................... 88

4.4 Consolidation of Flexible Strip Footing ........................................................................ 91

4.5 Undrained Analysis of a Strip Footing .......................................................................... 95

4.6 Contact Patch Test and Verification in Unconfined Compression ................................ 98

4.7 Rapid Installation of a Pile ........................................................................................... 100

4.7.1 Installation into MC soil ....................................................................................... 102

4.7.2 Installation into MCC soil .................................................................................... 110

4.7.3 Comparative study of the MC and MCC material models ................................... 114

4.7.4 Effects of frictional interface ................................................................................ 119

4.8 Summary ...................................................................................................................... 121

Chapter 5: Numerical Analysis of Dynamically Penetrating Anchors ........................... 124

5.1 Introduction .................................................................................................................. 124

5.2 Analysis Steps of a DPA and Literature Review ......................................................... 126

5.3 Simulation of a Free-falling Torpedo Anchor ............................................................. 130

5.3.1 Soil resistance profile during penetration ............................................................. 132

5.3.2 Deceleration of the anchor .................................................................................... 134

5.3.3 Pore-pressure generation throughout the penetration ........................................... 135

5.3.4 Set-up analysis ...................................................................................................... 138

5.4 Free-falling of a Torpedo Anchor into a Normally Consolidated Clay Layer ............. 140

5.4.1 Soil resistance profile during penetration ............................................................. 141

5.4.2 Deceleration of the anchor .................................................................................... 144

5.4.3 Pore-pressure generation throughout the penetration ........................................... 145

5.4.4 Setup analysis ....................................................................................................... 150

5.5 Summary ...................................................................................................................... 153

vi

6.1 Introduction .................................................................................................................. 156

6.2 Dynamic Coupled Analysis of an Offshore PipelineSeabed System......................... 158

6.3 Dynamic Laying Process of an Elastic Pipeline and Consolidation Settlements ........ 163

6.4 Pipeline under Large Amplitude Lateral Movement ................................................... 168

6.4.1 Numerical simulation ............................................................................................ 170

6.4.1.1 Vertical penetration ........................................................................................ 171

6.4.1.2 Lateral movement .......................................................................................... 175

6.5 Summary ...................................................................................................................... 184

Chapter 7: Conclusions and Recommendations ............................................................... 187

7.1 Introduction .................................................................................................................. 187

7.2 Governing Equations of Two-phase Saturated Porous Media ..................................... 187

7.3 Contact Mechanics of Two-phase Saturated Porous Media ........................................ 189

7.4 Numerical Evaluation of the Computational Scheme .................................................. 190

7.5 Numerical Analysis of Dynamically Penetrating Anchors .......................................... 191

7.6 Numerical Analysis of PipelineSeabed Interaction Problems ................................... 192

7.7 Recommendations for Future Research ....................................................................... 193

References ............................................................................................................................. 196

Appendix A.I ........................................................................................................................ 212

Appendix A.II ....................................................................................................................... 214

Appendix A.III ..................................................................................................................... 217

ONE-DIMENSIONAL TEST PROBLEMS FOR DYNAMIC CONSOLIDATION .... 217

Appendix A.IV...................................................................................................................... 231

Appendix B ........................................................................................................................... 232

Appendix C.I ........................................................................................................................ 233

Appendix C.II ....................................................................................................................... 235

Appendix C.III ..................................................................................................................... 237

vii

Preface

The research work presented in this thesis was conducted in the Department of Civil,

Surveying and Environmental Engineering at the University of Newcastle from July 2010 to

August 2014. This work was performed under the supervision of Dr. Majid Nazem, Laureate

Prof. Scott Sloan and Prof. John Carter.

The author claims originality for the entire work described in this thesis, except the

information or ideas derived from the many references and sources which have been

acknowledged in the text. In particular, originality of the following works is claimed:

Chapter 2

i.

The field equations for two-phase porous media were derived in light of the mixture

theory extended by the concept of a volume fraction. Although these equations may

have been applied in earlier studies, the equivalent arrangement introduced in the

derivation of the equation system facilitates the description of frictional contact in

terms of the effective normal stress component on the contact interface.

ii.

saturated soils was obtained by the finite element method. A U-P-V formulation was

selected to describe both incompressible and compressible fluids, in which the

resulting mixed formulation predicted all field variables, including solid displacement

U, pore-fluid pressure P and the Darcy velocity of the pore fluid V. This dynamic

consolidation scheme was implemented by the author into the existing in-house finite

element program, SNAC. The implemented scheme provided a rigorous solution to

the governing differential equations considering the convective terms of the fluid

acceleration.

iii.

A simplified solution was also outlined in the form of the U-P approximation, which

ignores the acceleration of the fluid component. This scheme was also implemented

into SNAC by the author.

viii

iv.

The ALE operator split technique and the mesh refinement strategy presented by

Nazem et al. (2009) was incorporated in this thesis to consider the effects of finite

deformations and to avoid possible mesh distortions. Application of the ALE scheme

within the dynamic consolidation framework is specifically claimed to be original.

v.

A literature review was presented for some of the available boundary conditions for

solving wave-propagation problems in an unbounded domain.

vi.

The cone boundary of Kellezi (2000) was adopted and implemented in the U-P-V

consolidation algorithm.

vii.

for some one-dimensional problems. These solutions were useful for validating FE

codes for the dynamic consolidation of soil.

Chapter 3

A new contact algorithm based on the mortar method was formulated and implemented for

solid-fluid mixtures in the spatial frame that can accommodate inertia effects together with

finite deformation and contact sliding. Both frictionless and frictional contact formulations

were addressed for two different forms of the dynamic consolidation formulations, including

U-P-V and U-P schemes.

Chapter 4

A number of validation exercises were presented to evaluate the performance of the

developed numerical scheme. These results are claimed as original.

Chapter 5

i.

A brief literature review of the available computational methods and available model

tests on Dynamically Penetrating Anchors (DPAs) was presented.

ii.

The numerical scheme developed in this thesis was then employed to conduct coupled

analysis of DPAs. These results are claimed as original.

Chapter 6

The computational scheme was utilised to analyse a few pipeline-seabed interaction

problems. These results are claimed as original.

ix

Chapter 7

The conclusions and recommendations for future work.

The candidate used the existing node-to-segment (NTS) contact algorithm in SNAC to

analyse some problems and compare the results with the mortar contact algorithm.

However, the modification of the NTS scheme and application of the method for

dynamic coupled consolidation analyses is claimed to be original.

During the term of the candidature, a number of papers and reports were published and some

awards were granted. These are listed below:

Sabetamal, H., M. Nazem, J. P. Carter and S. W. Sloan. 2014. Large deformation dynamic

analysis of saturated porous media with applications to penetration problems. Comput

Geotech 55:117131

Carter, J. P., H. Sabetamal, M. Nazem and S. W. Sloan. 2015. One-dimensional test problems

for dynamic consolidation. Acta Geotechnica 10(1):173-178

Formulation for Dynamic Analysis of Nonlinear Saturated Porous Media Based on the

Mortar Method. Int J Num Anal Meth Geomech (under review)

Sabetamal, H., M. Nazem, S. W. Sloan and J. P. Carter. 2014. Numerical analysis of offshore

pipeline-seabed interaction. In the proceedings of the 14th Int. Conference of International

Association for Computer Methods and Recent Advances in Geomechanics, IACMAG, KyotoJapan, pp. 655 - 660.

Poland.

Sabetamal, H., M. Nazem. 2013. Finite element simulation of frictional contact problems of

saturated porous medium under finite deformation. 12th U.S. National Congress on

Computational Mechanics, USNCCM12 , Raleigh, North Carolina.

Sabetamal, H., M. Nazem, S. W. Sloan and J. P. Carter. 2012. Finite element simulation of

dynamic pile penetration into a saturated porous medium. In the proceedings of the 6th

European Congress on Computational Methods in Applied Sciences and Engineering,

ECCOMAS, Vienna-Austria, pp. 5774-5785.

dynamic pore fluid-solid interaction in fully saturated porous media. In the proceedings of

the

XI

International

Conference

on

Computational

Plasticity-Fundamentals

and

Problems. Research Report, Australian Geomecanics Society (AGS).

Awards:

Jun 2014

Sep 2014

Methods and Advances in Geomechanics 2014 Conference held in Kyoto,

Japan for the paper Sabetmal. H., M. Nazem, J.P. Carter and S.W. Sloan

2014. Large Deformation Dynamic Analysis of Saturated Porous Media with

Applications to Penetration Problems. Comput Geotech 55:117131

Sep 2013

Postgraduate Research Prize.

xi

Table 2.1: Damping and stiffness matrices for cone boundary ............................................... 42

Table 3.1: Nested augmented Lagrangian scheme for frictionless contact problems of

two-phase saturated porous media .......................................................................... 63

Table 4.1: Material parameters ................................................................................................ 85

Table 4.2: Material parameters for the wave propagation analysis ......................................... 89

Table 4.3: MohrCoulomb material parameters ...................................................................... 92

Table 4.4: Material parameters .............................................................................................. 115

Table 5.1: MCC material parameters ..................................................................................... 141

Table 6.1: MCC material parameters ..................................................................................... 159

Box 3-1: Newton scheme for the update of within time increment for frictionless

contact ..................................................................................................................... 50

Box 3-2: Newton scheme for the update of n ....................................................................... 71

Box 3-3: Newton scheme for the update of n for the U-P scheme ....................................... 80

Table A.1: Soil properties ...................................................................................................... 224

xii

List of Figures

Figure 2.1: Evaluation of the various terms of the equation of motion in the generalised-

scheme..................................................................................................................... 27

Figure 2.2: (a) Semi-infinite 1D conical rod model; (b) application of cone model for 2D

problems .................................................................................................................. 39

Figure 2.3: Six-noded isoparametric element with cone boundary applied on its lateral

edge ......................................................................................................................... 43

Figure 3.1: Geometrical description for the contact formulation ............................................ 48

Figure 3.2: Definition of gap functions: (a) Darcy velocity and pore fluid pressure; (b)

displacement ........................................................................................................... 51

Figure 3.3: Minimal distance concept during frictionless sliding............................................ 54

Figure 3.4: Geometric interpretation of Coulomb friction law for 2D problems .................... 67

Figure 3.5: Initial and current configuration of two contacting bodies in a stick case ............ 68

Figure 3.6: Frictional sliding and movement of with the moving cone: (a) initial

configuration; (b) current configuration ................................................................. 69

Figure 3.7: Sliding and movement of the friction cone ........................................................... 69

Figure 4.1: One-dimensional dynamic wave propagation problem ......................................... 85

Figure 4.2: Solid displacement response versus depth ............................................................ 86

Figure 4.3: Pore-water pressure response with time ................................................................ 86

Figure 4.4: Normal Darcy velocity versus depth ..................................................................... 87

Figure 4.5: Normalised vertical settlements versus load level ................................................ 88

Figure 4.6: Evolution of pore-water pressure at a depth of 0.2 m versus time ........................ 90

Figure 4.7: Pore-water pressure evolution at a depth of 0.2 m versus time ............................. 91

Figure 4.8: Flexible strip footing on elasto-plastic layer ......................................................... 91

Figure 4.9: Settlement versus time factor for the elasto-plastic strip footing .......................... 93

Figure 4.10: Evolution and dissipation of normalised pore pressure....................................... 94

Figure 4.11: Excess pore pressure contours and Darcy velocity vector maps ......................... 94

Figure 4.12: Rigid rough footing on a cohesive soil layer....................................................... 95

Figure 4.13: Load-displacement curves ................................................................................... 97

Figure 4.14: Deformed mesh at the end of the ALE analysis .................................................. 98

Figure 4.15: Unconfined compression models: (a) model (i), two elastic layers with a

contacting interface; (b) model (ii), equivalent case using a single elastic layer

with no contact interface ......................................................................................... 99

Figure 4.16: Pore pressure at the interface of two layers normalised by applied pressure .... 100

Figure 4.17: FE meshes and boundary conditions: (a) dense mesh; (b) fine mesh ............... 103

Figure 4.18: Deformed meshes at different times ( = 10 ): (a) t = 0.05 s; (b) t = 0.5 s;

(c) t = 1.0 s ............................................................................................................ 104

Figure 4.19: Normalised total dynamic soil resistance versus normalised penetration

depth obtained for: (a) NTS method with smooth and non-smooth cone (

= 10 ); (b) non-smooth NTS and mortar methods ( = 2 ) .............................. 105

Figure 4.20: Excess pore-pressure response at depth d = 4D and radial distance of r = 2D

( = 10 ) ............................................................................................................... 106

Figure 4.21: Normalised total dynamic soil resistance versus normalised penetration

depth ...................................................................................................................... 107

xiii

Figure 4.22: Excess pore-pressure response at depth 4D and r = 0.15D ( = 10 ): (a) time

step t = 510-5s for both analyses; (b) time step size increased to t = 110-4s

for the analysis with fine mesh only ..................................................................... 108

Figure 4.23: Excess pore-pressure counters for = 2 ......................................................... 109

Figure 4.24: Evolution of normalised total dynamic soil resistance for various dilation

angles .................................................................................................................... 109

Figure 4.25: (a) Evolution of total dynamic soil resistance for various values of p0; (b)

deformed dense mesh at the end of installation .................................................... 111

Figure 4.26: Excess pore-pressure variation throughout penetration at depth 2.5D and

different radial distances ....................................................................................... 112

Figure 4.27: Excess pore-water pressure contour at the end of installation .......................... 113

Figure 4.28: (a) Evolution of total dynamic soil resistance; (b) excess pore-water pressure

at depth 6.25D ....................................................................................................... 113

Figure 4.29: Undrained shear strength profile ....................................................................... 117

Figure 4.30: Evolution of normalised total dynamic soil resistance predicted by three soil

models ................................................................................................................... 118

Figure 4.31: Evolution of total dynamic soil resistance for smooth and rough interfaces

( = 0.25), soil permeability k = 10-8 m/s ............................................................. 120

Figure 4.32: Evolution of total dynamic soil resistance for smooth and rough interfaces,

soil permeability k = 10-3 m/s................................................................................ 120

Figure 4.33: Stresses on contact area, soil permeability k = 10-3 m/s, ( = 0.25).................. 121

Figure 5.1: (a) Deep penetrating anchor (taken from Deep Sea Anchors); (b) torpedo

anchor with fins and without fins (after Medeiros 2002) ..................................... 125

Figure 5.2: (a) FE model of torpedo anchor analysis; (b) Torpedo shape adopted for the

analysis with the mortar contact ........................................................................... 131

Figure 5.3: Total dynamic soil resistance profile................................................................... 132

Figure 5.4: Total dynamic soil resistance profile obtained by the mortar and NTS

algorithms (impact velocity = 15 m/s) .................................................................. 133

Figure 5.5: Velocity versus penetration ................................................................................. 134

Figure 5.6: Velocity versus time ............................................................................................ 135

Figure 5.7: Excess pore-water pressure evolution at depth 5D throughout the installation

phase ..................................................................................................................... 136

Figure 5.8: Excess pore-water pressure contours at two different penetration depths: (a)

5.0D; (b) end of installation (impact velocity = 15 m/s)NTS results ............... 137

Figure 5.9: Deformed meshes during the free-falling process (analysis with mortar

contact) .................................................................................................................. 138

Figure 5.10: Excess pore-water pressure dissipation versus time for elements at depth 5D . 140

Figure 5.11: Total dynamic soil resistance profile................................................................. 142

Figure 5.12: Deformed meshes during the free-falling process and gradual closure of the

pathway ................................................................................................................. 143

Figure 5.13: (a) Pore-pressure contours (corresponding to Figure 6.13(b)); (b)

displacement vector plot ....................................................................................... 144

Figure 5.14: Velocity versus penetration ............................................................................... 145

Figure 5.15: Velocity versus time .......................................................................................... 146

Figure 5.16: Excess pore-water pressure evolution at a depth of 5D throughout the

installation phase................................................................................................... 146

Figure 5.17: Excess pore-water pressure evolution at a depth of 13.4D throughout the

installation phase................................................................................................... 147

Figure 5.18: Excess pore-water pressure evolution at a depth of 14D throughout the

installation phase................................................................................................... 147

xiv

Figure 5.19: Excess pore-water pressure contours throughout the penetration process ........ 149

Figure 5.20: Excess pore-water pressure dissipation versus time for elements at depth

13.4D..................................................................................................................... 151

Figure 5.21: Excess pore-water pressure dissipation at different times after installation ..... 152

Figure 6.1: Pipe-laying from a vessel, S-lay configuration (Source:

www.theengineer.co.uk) ....................................................................................... 157

Figure 6.2: Finite element model for the pipesoil interaction problem ............................... 158

Figure 6.3: Deformed mesh at embedment depths of: (a) 0.5D; (b) 1.0D ............................. 160

Figure 6.4: Normalised total penetration resistant versus normalised embedment ............... 161

Figure 6.5: Normalised excess pore pressure versus normalised embedment at the pipe

invert ..................................................................................................................... 162

Figure 6.6: Excess pore-pressure contours at the end of the dynamic pipe embedment ....... 162

Figure 6.7: Normalised excess pore pressure at the pipe invert ............................................ 164

Figure 6.8: Normalised pore pressures contours at Tv = 0.610-6.......................................... 165

Figure 6.9: Dissipation of excess pore pressure at the pipe invert......................................... 165

Figure 6.10: Excess pore-water pressure contours at different times of consolidation for

dynamic analysis ................................................................................................... 166

Figure 6.11: Darcy velocity vector maps ............................................................................... 167

Figure 6.12: Normalised embedment versus time factor ....................................................... 168

Figure 6.13: FE model for pipesoil interaction under lateral movement ............................. 171

Figure 6.14: Normalised embedment versus time factor ....................................................... 172

Figure 6.15: Excess pore-pressure contour plots during the loading and consolidation

stages ..................................................................................................................... 174

Figure 6.16: Dynamic lateral resistance: 1st, 2nd, 3rd and 4th sweeps ................................. 176

Figure 6.17: Pipe invert trajectory during lateral movement ................................................. 177

Figure 6.18: Excess pore-pressure contours during lateral movement: (a) at breakout; (b)

at 1.25D rightwards movement (end of sweep1); (c) during backwards

movement (sweep2) .............................................................................................. 178

Figure 6.19: Excess pore-pressure contours during lateral movement: (a) sweep3; (b) at

1.25D leftwards movement (end of sweep3); (c) during forwards movement

(sweep4) ................................................................................................................ 180

Figure 6.20: Dynamic lateral resistance: 1st, 3rd, 4th sweeps and last cycle ........................ 181

Figure 6.21: Pipe invert trajectory during lateral movement ................................................. 181

Figure 6.22: Excess pore-pressure contours during lateral movement in sweep5 ................. 182

Figure 6.23: Excess pore-pressure contours during consolidation ........................................ 183

Figure 6.24: Deformed mesh at the end of the analysis ......................................................... 184

Figure A.1 : Pore pressure at x = 0.2 m in infinitely deep layer with k = 0.001 m/s ............. 225

Figure A.2 : Pore pressure at x = 0.2 m in infinitely deep layer with k = 0.0005 m/s ........... 226

Figure A.3 : Pore pressure at x = 0.2 m in finite (1 m thick) layer with k = 0.0005 m/s... 228

Figure A.4 : Pore pressure at x = 1 m in finite (1 m thick) layer with k = 0.0005 m/s .......... 229

xv

Chapter 1: Introduction

1.1 General

Numerical modelling of the dynamic loading of saturated soil bodies undergoing large

displacements and possible surface penetration is one of the most sophisticated and

challenging problems in computational geomechanics, mainly due to the extreme material

and geometrical nonlinearity, large distortions, changing boundary conditions, material rate

effects and inertia forces induced in the soil. Moreover, the presence of pore fluid in saturated

soil and the generation of excess pore-fluid pressures due to dynamic loading, together with

the subsequent consolidation arising from the dissipation of those excess pressures, combine

to increase the complexity of such problems. A fully coupled analysis is required in order to

capture all aspects of the dynamics of the saturated soil behaviour. The analyses of

dynamically penetrating anchors, such as torpedo anchors, and free-falling penetrometers are

two examples of these geotechnical problems where dynamic effects cannot be neglected.

Torpedo anchors have proven to be promising systems for anchoring taut mooring lines of

floating offshore oil and gas exploration and production units due to their relatively easy

installation process. Free-falling penetrometers have been employed as an alternative to the

static cone penetration test (CPT) to provide information on the mechanical properties of

soils in inaccessible sites such as seabeds, lakebeds, wetlands and rivers. Numerical solution

schemes for these problems require robust algorithms for time stepping, domain remeshing,

interface modelling and stress-strain integration of the soil constitutive model.

The majority of current numerical models for simulating objects penetrating into soils are

generally based upon a displacement formulation involving a single-phase soil, where the

excess pore-water pressures are not explicitly calculated. These methods, assuming saturated

conditions, can only predict the total stresses developed in soil; in general, they cannot be

used to separately find the excess pore-water pressures and effective stresses in soil. Further,

most research works devoted to the analysis of dynamically penetrating anchors ignore the

effects of installation on the pull-out or lateral capacity of the anchor. That is, deep

foundation systems are wished in place, with no effort to model the effects of the installation

phase; hence, a perfect interface is assumed between the anchor system and the surrounding

soil. The initial stress state of the soil is usually estimated based on the submerged unit

weight and the lateral earth pressure coefficient at rest, and assumes zero excess pore-water

pressure. A reliable geotechnical analysis should simulate the process of installation and

incorporate pore-fluid pressure development along with deformations, velocities and

accelerations to facilitate a thorough understanding of soil response. Subsequently, with

knowledge of the effective stresses and the excess pore-water pressures, the setup analysis

can be conducted through reconsolidation of the soil in the vicinity of the anchor. Such

problems require a fully coupled analysis that takes into account the interaction between the

soil and pore fluid by incorporating the effect of the transient flow of the pore fluid through

the inter-connected voids of the solid skeleton. Moreover, a robust algorithm is required to

model soil-structure interactions during the entire process of the simulation.

Analysis of contact problems within the framework of the finite element method (FEM) is

generally accomplished using contact mechanics, in which kinematic relations are employed

in the treatment of interactions between deformable bodies. When a two-phase saturated soil

is studied, in addition to the requirement for continuity of the contact traction, continuity

should also be maintained for the Darcy velocity and the pore-fluid pressure across the

contact interface by enforcing appropriate constraints. Effective stresses at the interface

should also be used when evaluating frictional forces. Analysis of the frictional contact

mechanics of a porous medium and its FE formulation and implementation is a highly

challenging task that has rarely been addressed in the literature.

Large deformations and rigid body rotations affect a soils stiffness and permeability, and

such important effects may not be simply disregarded in the analyses. The theory of large

deformation has been widely used with Lagrangian FE approaches to analyse finite

deformation problems of geomechanics. However, Lagrangian methods are prone to failure

and numerical errors whenever the FE mesh undergoes excessive distortion, and they are

more likely to result in a negative Jacobian for an individual element. This is because the

motion of the body and the mesh are the same in the Lagrangian approaches, viz., a given

node remains coincident with the same material point throughout the analysis. However,

numerical difficulties associated with excessive element distortion can be circumvented by

combining the merits of the Updated Lagrangian (UL) scheme and the Eulerian approach. In

the literature, this strategy has resulted in two similar methods: Arbitrary Lagrangian

Eulerian (ALE) and Coupled EulerianLagrangian (CEL).

Another challenge in dynamic FE analyses is avoiding spurious wave reflection from the

artificial mesh boundaries of the problem domain. Due to the fairly high wave velocity in

soils and rocks, modelling a large portion of the domain does not seem to be an effective way

to proceed, as the waves reflected from the boundaries will probably have enough time to

return to the area of interest during the time period of interest. Therefore, absorbing boundary

conditions should be facilitated to absorb the outgoing waves.

The objectives of the present research study are to: (i) document the development of a

computational scheme based on the FEM in order to conduct dynamic coupled analyses for

geotechnical problems involving soil-structure interactions and finite deformations; (ii)

calibrate and validate the numerical procedure, and (iii) illustrate its application and

modelling features through the simulations of two important offshore geotechnical

problemstorpedo anchors and pipeline-seabed interactions.

The first task is to formulate and implement the governing equations of two-phase saturated

porous media. The mechanical behaviour of a saturated porous medium is predicted using

mixture theory, which models the dynamic advection of fluids through a fully saturated

porous solid matrix. The resulting mixed formulation predicts all field variables, including

the solid displacement, pore-fluid pressure and Darcy velocity of the pore fluid. Chang and

Hulberts (1993) generalised- algorithm is employed to integrate the governing equations of

the two-phase saturated porous media in the time domain. The UL approach with Jaumanns

objective stress rate is utilised to consider the effects of finite deformation. This methodology

is consistent with the ALE scheme presented by Nazem et al. (2006). Accordingly, the ALE

approach is incorporated to account for geometrical nonlinearities and avoid possible mesh

distortions. The cone energy-absorbing boundary of Kellezi (2000), which consists of

dashpots and springs, is adopted to absorb the outgoing bulk waves and cope with spurious

wave reflections. Closed-form solutions developed in collaboration with others (Carter et al.

2015) are presented for the problem of step-loading applied to a layer of saturated soil with a

linear elastic skeleton and a compressible pore fluid. These solutions provide a check on the

concurrent wave transmission and consolidation processes modelled by the dynamic

consolidation algorithm.

The subject of the second development is a detailed account of the formulation of the contact

kinematics and constraints for two-phase saturated porous media. The formulation is derived

and implemented for the frictionless and frictional interfaces based on the so-called mortar

segment-to-segment approach, which allows the interpolation functions of the contact

elements to be of order n. The contact constraints arising from the requirement for continuity

of the contact traction and the fluid flow across the contact interface are enforced using a

penalty approach, which is regularised with an augmented Lagrangian method. In the coupled

consolidation-contact algorithm developed here, free-draining conditions are automatically

adopted for the nodes that lose connection with their possible contacting pair, whereas

impermeable or semi-impermeable conditions are adjusted for soil nodes that come in contact

with another surface.

The next task considers the evaluation of the numerical scheme. The implications and merits

of the mortar contact algorithm for the dynamic coupled analysis of some geomechanics

problems are illustrated by comparing its predictions with those obtained by a node-tosegment (NTS) contact algorithm. The nonlinear behaviour of the solid phase of soil in the

numerical examples is represented by either the Modified Cam Clay material model or the

MohrCoulomb model.

In the last task, the numerical procedure developed in the course of this study is employed to

model the installation and consolidation process of dynamically penetrating anchors, as well

as problems of pipeline-seabed interaction. The analysis of the pipeline-seabed system

involves simulating the laying process, the subsequent consolidation stage and the large

amplitude cyclic motions of the pipe, which requires an automatic assessment of the drainage

conditions around the pipe as it moves. The nonlinear behaviour of the solid phase of soil in

the analyses of these applications is represented by the Modified Cam Clay material model.

The work presented in this thesis is organised as follows. Chapter 2 presents the adopted

equations governing the dynamics of a saturated porous medium using mixture theory,

followed by the discretisation of the governing equation in the space and time domains, the

extension of the equations to a finite deformation regime within the framework of the ALE

method and the implementation of the absorbing boundary conditions. Further, closed-form

solutions are developed for some one-dimensional problems involving the dynamic response

of saturated porous media.

Chapter 3 provides the formulation of the contact kinematics and constraints derived for both

frictionless and frictional interfaces. The contributions of the contact constraints to the system

of equations are described and implemented.

Chapter 4 considers the evaluation of the different aspects of the computational algorithms

using either alternative numerical analyses or closed-form solutions where possible.

Chapter 5 outlines the dynamically penetrating anchors, including a literature review of the

computational methods and available model tests on these problems. The application of the

numerical scheme developed in this thesis is then illustrated by modelling the problems of

torpedo anchors. A comparison is also made between the results predicted by the mortar

method and the NTS scheme, which takes advantage of a contact surface-smoothing method

using Bzier polynomials.

problems of the offshore pipeline-seabed interaction system.

Chapter 7 provides a summary and conclusions, along with recommendations for related

future research.

Chapter 2

Equation Chapter 2 Section 1

2.1 Introduction

A saturated porous medium comprises solid and fluid constituents that interact with each

other and affect the overall mechanical behaviour. Coupling the responses of each individual

constituent complicates the mechanics of a porous medium in comparison with a single-phase

material. For saturated soils, the coupling effects might be negligible when the permeability

is relatively high and the load is applied slowly so that the overall response of the medium is

close to fully drained. In contrast, the coupling will be very strong in the case of low

permeability and fast transient loading, in which case an accurate prediction of the soil

behaviour requires the coupling effects to be taken into account.

Terzaghi (1923) was the first to derive the partial differential equation for the onedimensional consolidation process and develop the idea of effective stress, intuitively.

Terzaghi formulated and presented the concept of effective stress in 1936. Based on a

physical approach, Biot (1941) generalised the consolidation theory for the three-dimensional

deformation of elastic porous media containing an incompressible pore fluid. Later, Biot

(1956) developed basic equations for the description of elastic wave propagation in a porous

isotropic solid saturated with a compressible viscous fluid supposing constant mass densities

for the constituents. Biot (1962) further investigated acoustic wave propagation in a porous

solid by emphasising the influence of anisotropy, viscoelasticity and dissipation in solids.

Truesdell and Touppin (1960) formed modern continuum mechanics and proposed the

general theoretical framework of mixture theory. Morland (1972) was the first to use the

concept of volume fraction in connection with the mixture theory to propose a constitutive

theory for a fluid-saturated porous medium. The effects of material and geometrical

nonlinearities in the theory of consolidation were then incorporated by Small et al. (1976)

and Carter et al. (1979), respectively. Bowen (1980, 1982) described incompressible and

compressible porous media models using the mixture theory. Prvost (1980) extended Biots

theory into the nonlinear inelastic range based on the mixture theory and constitutive

equations given by Bowen (1980). Ghaboussi and Wilson (1972) proposed the first numerical

treatment of boundary value problems based on Biots model utilising the FEM. Zienkiewicz

and Shiomi (1984) presented a few FE numerical solution schemes for Biots dynamics

theory. Coussy (1995) and de Boer (2000) comprehensively reviewed the historical

development and current state of the porous media and mixture theories.

This chapter first presents the balance laws governing interaction between the soil

deformation, pore-fluid flow and Darcy velocity of pore fluid. The balance laws are

formulated by adopting the mixture theory and using the concept of volume fraction. The

weak forms of the balance laws and their three-field FE discretisation are presented

considering the solid displacement, pore-fluid pressure and Darcy velocity as field quantities.

A continuum approach based on the theory of mixtures is employed to derive the governing

equations of saturated porous media by making use of the concept of volume fraction.

According to this approach, each phase is smeared over the entire domain of the porous

medium with a reduced density in order to create a homogenised continuum supposing an

immiscible mixture. The porous solid fills a control space, and only the fluid confined in the

pores can leave the control space. The principles of continuum mechanics are then invoked to

describe the behaviour of the equivalent medium at a macroscopic level. Accordingly, the

local balance relations should be satisfied for each individual constituent in the mixture,

together with the global balance relations, irrespective of the constitutive model. Due to the

assumption of an immiscible mixture and non-penetrating particles, the constitutive models

are required only for partial stresses and diffusive resistance. The relative motion between the

constituents is the main kinematic variable associated with the interaction volume forces.

Consider a two-phase mixture composed of a solid matrix whose voids are continuous and

completely filled with fluid. The volume fraction occupied by the phase ( = s for the solid

and = f for the fluid) is given by:

= v v

2.1

where v is the volume of the mixture in the current configuration. Thus, the saturation

condition can be expressed as:

s + f =

1

2.2

The partial mass density of each phase is related to its true mass density by:

2.3

Thus:

s + f =

2.4

where is the total mass density of the mixture. Note that f can be interpreted as the

commonly used concept of porosity n in soil mechanics. In the rest of this work, porosity n is

used to represent the volume fraction of the fluid phase, while (1-n) is used to represent the

volume fraction of the solid phase.

The following section develops principles describing the conservation of mixture mass and

the conservation of the linear momentum of mixture and the fluid phase.

The mass balance equations are formulated for each individual constituent, and the

superposition of the mass conservation equations provides the mass balance of the mixture

body.

Solid phase

The mass of the solid phase ms attributed to a deformed material body t at time t is given

by the volume integral:

ms =

(x, t )dv

2.5

where s is in general a scalar field; that is, a continuous function of location x and time t

defined in the current configuration. The law of conservation of mass implies that the

material time derivative of the solid phase mass must cease to exist. Therefore, the material

time derivative of Eq. 2.5 with use of the transport theorem is expressed in the spatial form

as:

s

d (m s )

=

+ div( s v s ) dv =

0

dt

t

2.6

where div is the divergence operator evaluated with respect to the current configuration and

vs is the intrinsic velocity of the solid phase. As t is an arbitrary part of the continuum, the

integrand in Eq. 2.6 must vanish, resulting in:

s

+ div( s v s ) =

0

t

2.7

Expressing this equation in terms of the true mass density of solid phase yields:

[ s (1 n) ]

t

0

+ div s (1 n) v s =

2.8

Porosity n is also a scalar field, viz., a continuous function of location x and time t. Assuming

the soil particles to be incompressible and homogeneous provides:

n

div [ (1 n) v s ] =

t

2.9

Fluid phase

The mass of the fluid phase mf in the deformed domain t at time t is given by the volume

integral:

mf =

(x, t )dv

2.10

The law of conservation of mass requires that the material time derivative of the fluid phase

mass must vanish. Therefore, the material time derivative of Eq. 2.10 with use of the

transport theorem is expressed in the spatial form as:

f

d (m f )

0

=

+ div( f v f ) dv =

dt

t

2.11

where vf is the intrinsic velocity of the fluid phase. The integrand in Eq. 2.11 must vanish

everywhere in the considered arbitrary domain. Hence:

f

0

+ div( f v f ) =

t

2.12

Stating this equation in terms of the true mass density of fluid phase yields:

(n f )

t

+ div(n f v f ) =

0

2.13

f

t

+f

T

n

+ f div(nv f ) + n grad( f ) v f =

0

t

2.14

Mixture

The balance equation for the mixture mass is obtained by inserting Eq. 2.9 into Eq. 2.14 as:

f

t

0

T

2.15

The intrinsic fluid mass density variation can be related to the intrinsic fluid pressure changes

through the definition of the fluid bulk modulus f as:

p

= f

f

2.16

where p is the absolute pore-fluid pressure, which is also a function of location x and time t.

It is assumed that the fluid bulk modulus is constant, and that thermal effects and any mass

exchanges are excluded from the model. By using the chain-rule of differentiation of the

calculus, it can be stated that:

f f p f dp

T

= =

[ grad( p ) ] v s

t

p t w dt

10

2.17

where

dp

is the (Eulerian) time derivative of p. Similarly:

dt

f f p (x, t ) f p (x, t ) f

grad( p )

=

= =

xi

p xi

f xi

f

2.18

Hence, by using Eqs 2.17 and 2.18 in Eq. 2.15, the conservation of mixture mass can be

written as:

n dp

n

T

+ div [ (1 n) v s ] + div(nv f ) + [ grad( p ) ] ( v f v s ) =

0

f dt

f

2.19

In the governing equations, the relative or Darcy velocity vr is adopted and defined in terms

of the fluid-phase velocity as:

=

v r n( v f v s )

2.20

Using the definition of Darcy velocity in Eq. 2.19 leads to the modified Eulerian form of the

balance equation for the mixture mass as:

div( v s ) + div(v r ) +

n dp 1

T

0

+ [ grad( p ) ] v r =

f dt f

2.21

It is notable that the sum of the third and fourth terms in Eq. 2.21 can be referred to as the

storage of fluid mass due to the compressibility of the fluid phase. For incompressible pore

fluid (i.e., f ), Eq. 2.21 reduces to:

div( v s ) + div(v r ) =

0

2.22

which is analogous to the one that governs any incompressible single-phase media. In the

present formulation, the fluid phase is considered compressible. Therefore, Eq. 2.21 will be

used for further discussion.

11

The principle of linear momentum states that the time rate of change of the linear momentum

is equal to the resultant force acting on the body. The principle of linear momentum is valid

for each constituent and for the entire mixture.

From mixture theory, the total Cauchy stress tensor is stated as the sum of the Cauchy

partial stress tensor of each constituent as:

=

f + s

2.23

According to the principle of effective stress, the total Cauchy stress tensor is the sum of the

effective stress tensor and the isotropic pore fluid pressure tensor pI as:

= + pI

2.24

where I represents the second-order unit tensor. The Cauchy partial stress tensors are related

to the effective stress tensor and pore-fluid pressure as (Prvost 1980):

f = npI

2.25

s = + (1 n) pI

2.26

Solid phase

The balance of linear momentum for the solid phase in the arbitrary domain t at time t is:

n ds + (1 n) s b dv + h s dv =

t

(1 n) a dv

s

2.27

where b represents the body force per unit volume, as denotes the acceleration of the solid

phase, which is defined as the time derivative of the solid-phase velocity (

v s

); n is the unit

t

configuration; and hs is the momentum supplied to the solid constituents from the rest of the

mixture due to other internal interaction effects between the two phases. The first term on the

12

left-hand side of Eq. 2.27 can be converted to a volume integral by applying the divergence

theorem, which results in:

div( ) + (1 n) b + h

s

(1 n) s a s dv =

0

2.28

div( s ) + (1 n) s (b a s ) + h s =

0

2.29

which presents the localised form of the balance of solid-phase momentum in Eulerian form.

Fluid phase

The balance of linear momentum for the fluid phase in the arbitrary domain t at time t is

stated as:

n ds +

b dv + h f dv =

n f a f dv

t

2.30

where hf represents the momentum supplied to the fluid phase as the fluid flows through the

voids; and af denotes the acceleration of the fluid phase as:

v f

+ grad( v f )( v f v s )

af =

t

2.31

where the first term on the right-hand side of Eq. 2.31 gives the local rate of fluid velocity

change, and the second term is the convective rate of change.

Applying the divergence theorem to the first term of Eq. 2.30 gives rise to the localised form

of the balance of fluid-phase momentum as:

div( f ) + n f (b a f ) + h f =

0

2.32

The definition of Darcys velocity in Eq. 2.20 is also adopted here to define the relative

acceleration of fluid phase ar as:

13

=

a r n(a f a s )

2.33

By replacing f and a f from Eqs 2.25 and 2.33 in Eq. 2.32, the linear momentum balance

for the fluid phase can be expressed as:

grad (np ) + n f (b a s ) f a r + h f =

0

2.34

ngrad (p ) + pgrad (n) + n f (b a s ) f a r + h f =

0

2.35

Prevost (1980) showed that interaction volume forces between the constituents ( h ) for a

two-phase saturated porous medium can be calculated using constitutive equations given by

Bowen (1980) as:

hs =

h f =

pgrad(n) + f k 1 v r

2.36

where k is the (diagonal) permeability tensor. Using Eq. 2.36 in Eq. 2.35 yields to the final

form of linear momentum balance for the fluid phase as:

1

1

0

grad( p ) + f (b a s ) f a r f k 1 v r =

n

n

2.37

An explicit form of a r in terms of solid-phase velocity and Darcy velocity can be derived by

inserting a f from Eq. 2.31 in Eq. 2.33 as:

v

v

ar =

n f + grad( v f ) v r n s

t

t

2.38

ar n

=

v

1

1

( v r + v s ) + grad( v r + v s ) v r n s

t n

t

n

14

2.39

(1 n)div( v s )

v r [ grad(n) ] v s

1

ar =

vr

v r + grad( v r ) v r

+

t

n

n

n

T

[grad(n)]

+ grad( v s ) v r

vr

2.40

vr

Mixture

For the entire mixture, the expression of the balance of linear momentum is obtained by

adding the two momentum equations given in Eqs 2.34 and 2.29, resulting in:

div( ) + s (1 n) + f n (b a s ) f a r =

0

2.41

Solving the boundary value problems using the governing differential equations presented in

Eqs 2.21, 2.37 and 2.41 requires a set of additional restraints called boundary conditions.

Appropriate Dirichlet (essential) and Neumann (natural) boundary conditions for the

governing differential equations are specified as follows:

on

and=

n t

p P

=

on

and

=

v r n qN

=

v r Vr

on

p n t p

and

=

=

us U

on

on

on

2.42

where adopted field quantities are solid displacements u s , pore-fluid pressure p and Darcy

velocities v r , t is a vector of surface traction due to the total stresses applied on the surface

boundary , t p is a vector of surface traction due to pore-fluid pressure applied on surface

boundary f , qN is the normal component of volumetric flow through flux boundary q and

n is the outward unit normal vector to surface s bounding the problem domain .

This section presents the variational statement of the balance laws governing the behaviour of

the mixture derived in the preceding section. The weak forms can be stated with respect to

15

either the deformed domain or the current configuration ( t ) at time t. The fundamental

difficulty is that the current configuration of the body is unknown a priori, as the volume of

the body changes with time as the analysis proceeds. Therefore, it is preferred to use a

previously known or reference configuration at time < t to evaluate the integration of

the weak form. The so-called UL formulation takes the last configuration as a reference body

to evaluate all sate variables.

dV . In terms of its original size, the current volume dv is JdV , where J = det F denotes the

x1 t

x1

x2 t

F=

x1

x3 t

x1

x1 t

x2

x2 t

x2

x3 t

x2

x1 t

x3

x2 t

x3

x3 t

x3

2.43

The mixture porosity at current configuration nt = n can also be evaluated from its original

value n = n0 at the reference configuration because the solid constituent is incompressible;

hence, the volume of the solid phase (1 n0 )dV is constant during the deformation process,

whereas the volume of fluid constituent changes to JdV (1 n0 )dV . Accordingly, the

porosity can be updated from reference to the current configuration as:

n=

JdV (1 n0 )dV

(1 n0 )

= 1

JdV

J

2.44

Following the standard variational principle, the differential equation governing the balance

of mixture mass (Eq. 2.21) is multiplied by an arbitrary function d p in order to transform the

differential equation onto a scalar function, where d p is chosen to be the variation of porefluid pressure. The resulting scalar function is then integrated over the domain t , leading to

a functional d p :

16

n dp

dv +

f dt

(d p) div( v ) dv + (d p) div( v ) dv + (d p)

=

d p

(d p)

2.45

[grad( p )] v r dv = 0

T

After applying the divergence theorem to the second term, the weak form becomes:

T

(d p)

=

d p

(d p)

qN ds + (d p ) div( v s ) dv + (d p )

t

n dp

dv +

f dt

2.46

[grad( p )] v r dv [grad(d p )] v r dv =

0

T

T

(d p)

=

d p

qN ds +

(d p)

n dp

JdV +

f dt

2.47

[grad( p )] v r JdV [grad(d p )] v r JdV =

0

(d p)

div( v s ) JdV +

(d p)

Similarly, the balance of linear momentum for the fluid constituent can be expressed in a

weak form. Multiplying Eq. 2.37 by the variation of the Darcy velocity d v r and integrating

the result over the domain t leads to a functional d r :

d r

=

(d v ) grad( p)dv + (d v )

r

(d v r )

f

n

f (b a s )dv (d v r )

t

f

n

a r dv

2.48

0

k v r dv =

After applying the divergence theorem to the first term, the weak form becomes:

d r

=

(d v ) t

r

(d v r )

t

ds + (d v r ) f (b a s )dv (d v r )

t

f

n

k v r dv

f

n

a r dv

0

pdiv(d v ) dv =

r

17

2.49

The weak form can be rewritten in the reference configuration by converting the domain of

integration as:

T

(d v ) t

d r

=

ds +

(d v )

(d v )

r

f (b a s ) JdV (d v r )

f

n

a r JdV

2.50

0

p div(d v ) JdV =

k v r JdV

The variation of the solid displacement d u s is taken as an arbitrary function. Multiplying Eq.

2.41 by d u s and integrating the result over the domain t leads to a functional d s :

d s

=

(d u )

s

div()dv +

(d u ) grad( p)dv + (d u )

s

(b a s )dv

2.51

0

(d u s ) f a r dv =

t

The divergence theorem is applied to the first and second terms of Eq. 2.51 and, after some

simplification, it may be written as:

T

(d u ) tds + (d u )

=

d s

[grad(d u s )]T dv

t

(b a s )dv (d u s ) f a r dv

t

0

pdiv(d u s ) dv =

2.52

The weak form can be expressed in the reference configuration by converting the domain of

integration as:

T

(d u ) tds + (d u )

=

d s

(d u )

s

s (1 n0 ) + f nJ (b a s )dV

2.53

0

pdiv(d u )JdV =

s

The numerical solution of the weak forms can be followed by either the discontinuous spacetime Galerkin method, in which time is considered an extra dimension and treated in the same

way as the spatial coordinates, or the semi-discretisation approach, which discretises space

18

and time variables using two different approaches. A straightforward application of the

Galerkin principle on the time variable may couple all time levels and destroy the crucial

property of propagation forward in time (Strang and Fix 1973). The semi-discretisation

approach utilises FEM or another alternative approach, such as the mesh-free method, to

discretise the spatial part. This process leads to a system of second-order ordinary differential

equations (ODEs) in time. In the mathematical context, the resulting system of ODEs may be

solved by means of one of many existing methods, such as RungeKutta, Adams or Crank

Nicolson. However, three distinct methodologies are usually facilitated in the framework of

the FEM: modal analysis, frequency domain analysis and direct time integration. The

classical modal and frequency domain analyses are based on the principle of superposition;

thus, they are not applicable to nonlinear systems. In contrast, the direct time integration

approach is a promising method for nonlinear problems, regardless of their time-consuming

nature. This study uses the FEM and the direct time integration procedure.

This section deals with the discretisation of the presented variational statements in the space

domain. By using the standard Galerkin approximation, the discrete versions of the weak

forms are stated. Shape functions are constructed and used to approximate solid matrix

displacements us, Darcy velocities vr and pore-fluid pressure p, together with their variations,

through the corresponding nodal values U , Vr and P , as:

u=

NsU

s

v r = N r Vr

v=

u=

NsU

s

s

v = N V

r

=

a=

u

NsU

s

s

d u=

N sd U

s

d v r = N rd Vr

=

p N=

p N p P

pP

2.54

d p = N pd P

where Ns, Nr, and Np are the interpolation functions for the solid displacements, Darcy

velocities and pore-fluid pressure, respectively. The interpolation function for the pore-fluid

pressure Np is generally chosen to be one order lower than the functions for the displacement

and fluid velocity in order to avoid numerical locking and instability (oscillations) in the

predicted pore-fluid pressure response. These issues occur because of the nature of saddlepoint behaviour associated with consolidation problems and the violation of the Babuska

Brezzi stability condition. However, pore-pressure oscillations can be circumvented for the

19

2002).

The discrete version of the weak variational statements (Eqs 2.47, 2.50 and 2.53) can be

expressed based on the approximations from Eq. 2.54, as follows:

=

d p

(d P)

NTp qN ds +

(d P)

JdV +

NTp div ( N s ) U

(d P)

NTp

N P P JdV +

2.55

0

(d P) N f grad ( N p ) v r PJdV (d P) grad ( N p ) N r Vr JdV =

T

p

(d V )

=

d r

NTr t p ds +

(d V )

r

NTr f bJdV

T

JdV

d

(

V

)

NTr f N s U

r

(d V )

(d U)

NTs tds +

(d U)

NTr

NTs f ar JdV

(d U)

div ( N

NTs bJdV

(d U)

(d V )

=

d s

T

(d Vr ) N r

T

s

)N

f

n

2.56

k 1N r Vr JdV =

0

(d U) N

T

s

JdV

NsU

(d U)

ar JdV

grad ( N s ) JdV

T

2.57

PJdV =

0

The system of equations represented by Eqs 2.55, 2.56 and 2.57 must hold for any arbitrary

selection of d P , d Vr and d U s . Therefore, the resulting equations governing the behaviour of

the mixture may be written in matrix form as:

M ss

M

rs

0

M sr

M rr

0

C

0 U

ss

0 Vr + 0

C

0 P

ps

Csr

Crr

C pr

K

0 U

0 Vr + 0

C pp P 0

0 K sp U f s

r

0 K rp 0 =

2.58

f

p

0 K pp P f

Explicit expressions for matrices M, C and K, as well as vectors f, are presented in Appendix

A.I. This provides an exact solution to the problem and is called the UPV scheme.

20

obtained by ignoring the acceleration of the fluid component in Eq. 2.41. This results in a

coupled set of equations in which U and P are the only unknowns. This type of formulation is

termed the UP formulation (Zienkiewicz and Shiomi 1984). The effect of inertia on the

pore-fluid momentum equation (Eq.2.37) is also considered insignificant within the

frequency range for which the UP approximation is valid (Chan 1988). The UP

formulation was also adopted in this thesis and applied for a few large deformation problems.

The global FE equations of the UP scheme have the following matrix form:

M s

0

Cs

0 U

+

LT

0 P

K

0 U

+

S P 0

L U f u

=

H P f p

2.59

,U

and P denote the vectors of displacement, velocity, acceleration and poreWhere U , U

fluid pressure, respectively; Ms, Cs and K are the solid mass, damping and stiffness

matrices, respectively; L, H and S represent the coupling matrix, fluid flow matrix and

compressibility matrix, respectively; and fu and fp are the vectors of external nodal forces. For

further details regarding the treatment of the UP formulation, the implementation of tangent

contributions and the solution strategy, see Sabetamal et al. (2014).

Two main sources of nonlinearitymaterial nonlinearity and geometric nonlinearitycan

arise in the analysis of porous continua. The tangent stiffness matrix K contains

contributions from geometric and material nonlinearities. Geometric nonlinearity is important

in many cases, such as the analysis of liquefaction, deep penetration of objects into soil layers

and any situation where the strain level is relatively high. However, classical FE algorithms

suppose that a small range of strains is occurring in the soil, ignoring the effects of finite

deformation, rigid body rotation and loading variations. The fundamental difficulty in large

deformation analysis is that the configuration of the body at time t is unknown, as the volume

of the body changes with time as the analysis proceeds. Moreover, the Cauchy stress tensor is

not an objective measure of stress, and it changes when rigid body rotation occurs. The Total

Lagrangian (TL) formulation was developed to incorporate these effects in a largedeformation algorithm by utilising the GreenLagrange strain tensor E and second-Piola

Kirchhoff stress tensor S instead of the Cauchy strain and stress tensors. However, the

21

definition of constitutive equations in terms of the GreenLagrange strain tensor and the

second-PiolaKirchhoff stress tensor is a complicated task. In contrast, the so-called UL

formulation takes the last configuration as a reference body to evaluate stress and strain,

resulting in a simplified stiffness matrix. See Appendix A.II for integral equations of

K accounting for material and geometric nonlinearities which is represented by K NL .

A large-strain analysis using Lagrangian methods may eventually involve severe mesh

distortion, implying that these methods usually fail to provide a solution because of the

development of a negative Jacobian in some elements. One strategy to tackle this problem is

to separate the mesh and the material displacements. This strategy is the basic idea of the

ALE method. In a Lagrangian description, the mesh follows the motion of the body. A given

node remains coincident with the same material particle throughout the motion; thus, the

mesh can become excessively distorted whenever the displacements are relatively large. In

contrast, in an Eulerian description, the mesh is fixed in space and the nodes are no longer

coincident with the material points during the analysis; consequently, mesh distortion does

not occur. However, material boundaries are difficult to model in an Eulerian description.

The ALE method attempts to combine the advantages of the Lagrangian and Eulerian

meshes. In this method, the computational grid is neither coincident with the material, nor

fixed in space. Rather, it can move arbitrarily in order to avoid possible mesh distortions.

Expressing the global equations at time t with respect to an arbitrary moving grid leads to a

convective term in the ALE equations, which accounts for the transport of material through

the grid points (e.g., Gadala 2004). Therefore, the global equations include the unknown

material deformations as well as the grid displacements. In one solution strategy, all unknown

equations are solved simultaneously for material and mesh displacements. This strategy is

known as the coupled ALE method. In contrast, in the uncoupled ALE, or the operator-split

technique, the mesh and material displacements are decoupled. Benson (1989) proposed this

method for problems of solid mechanics and showed that the cost of computations could be

reduced by a factor of two without a significant loss in accuracy. In the operator-split

technique, the analysis is performed in two steps: a regular UL step followed by an Eulerian

step. In the first step, the governing equations are solved to fulfil equilibrium and obtain the

material displacements. In the Eulerian step, the mesh is refined for the deformed domain to

find the mesh displacements. All kinematic and static variables are then remapped from the

old mesh to the new mesh. In a coupled displacementpore-water pressureDarcy velocity

22

ALE analysis, the state parameters to be transformed at integration points include the

effective stresses, hardening parameters, void ratios and coefficients of permeability, while

the pore-water pressures and Darcy velocities, as well as the solid displacements, velocities

and accelerations, are transformed from the old nodes to the new nodes. Nazem et al. (2008)

applied the ALE method for static consolidation problems considering displacement and

pore-fluid pressure as nodal variables. The remapping of state variables is usually performed

using a first-order expansion of Taylors series as:

f

f r = f + (vi vir )

xi

2.60

where f r and f denote the time derivatives of an arbitrary function f with respect to the

r

mesh and material coordinates respectively; vi is the material velocity; and vi represents the

mesh velocity. The procedure for remapping state variables has been explained in detail

elsewhere (see Nazem et al. 2009; Nazem et al. 2008), and thus will not be repeated here.

Note that the new state variables do not necessarily satisfy the equilibrium conditions or the

consistency principle of plasticity (if the material response is nonlinear), mainly because of

the inevitable numerical diffusion that occurs while remapping. To satisfy these two

conditions, additional NewtonRaphson iterations are conducted at the end of each time step.

The ALE operator-split technique and mesh refinement strategy is utilised in this thesis based

on the method presented by Nazem et al. (2009). It is notable that Carter et al. (2010) and

Nazem et al. (2012) validated the method by comparing its predictions with results obtained

from penetration experiments conducted in the laboratory.

Another important aspect of consolidation analysis is the variation of permeability that may

result from large deformations and rigid body rotations. To consider the effect of rigid body

rotations on the coefficients of permeability, Carter et al. (1979) proposed the following

transformation:

k t = RT k R

2.61

where k represents the permeability tensor and R is the rotation matrix at a Gauss point. To

consider the effects of deformations, the permeability of the soil may be expressed in terms of

its void ratio by empirical relations, in which the void ratios at each time step are updated

according to Eq. 2.44.

23

In this study, closed-form solutions were developed for some one-dimensional problems

involving the dynamic response of saturated porous media. These solutions are useful for

validating FE codes for the dynamic consolidation of soil. While they consider only elasticity

and small strains, they allow a check on the concurrent wave transmission and consolidation

process. The developed closed-form solutions and their numerical evaluations are presented

in Appendix A.III, and reference is also made to Carter, Sabetamal, Nazem and Sloan (2015).

The FE discretisation of the global equations leads to a system of second-order differential

equations in which time is a continuous variable. In a direct time integration scheme, Eq. 2.58

is integrated by a numerical step-by-step procedure. Therefore, equilibrium is only satisfied at

discrete time intervals and, depending on which time steps are used to ensure the equilibrium,

explicit and implicit integration methods can be distinguished. Although the computational

cost of implicit techniques can increase for complex problems compared to explicit methods,

conditional stability of explicit methods might be problematic for a coupled consolidation

analysis. The central difference method is a widely used explicit time integration method.

The Houbolt, Wilson-, Park and Newmark methods, along with the family of -methods, are

representative of implicit integration schemes. The approximation function used to relate the

displacement, velocity and acceleration of two consecutive time steps distinguishes the

various time integration methods.

Generally, the major concern is the contribution of low-frequency modes in the overall

behaviour of most dynamic systems. In addition, approximation of the higher eigenmodes

through the spatial FE discretisation is not accurate because many of the predicted highfrequency modes correspond to spurious artifacts of the discretisation process rather than the

actual physical behaviour of the system (Strang and Fix 1973). Therefore, it is sometimes

advantageous to damp (filter) the higher modes during the numerical integration process.

Consequently, the selected time integration scheme should possess some form of numerical

dissipation that can attenuate the inaccurate high-frequency modes. Meanwhile, it should also

allow the accurate capture of the low-frequency behaviour of the system so that it appears in

24

the solution without attenuation. Hughes and Hilber (1978) described the key characteristics

that make a time-marching scheme competitive and efficient. These include unconditional

stability for linear problems, only one set of implicit equations to be solved at each time step,

second-order accuracy, controllable numerical dissipation at higher modes, self-starting and

no tendency to pathologically overshoot the true solution.

The Newmark scheme (Newmark 1959) is one of the most popular methods in the family of

direct time integration techniques. It is based on the following approximations for

displacements and velocities:

+ t (1 2 ) U

+ 2 U

U n +1 = U n + tU

n

n

n +1

2

2.62

=U

+ t (1 ) U

+ U

U

n +1

n

n

n +1

2.63

where the increment in time (t-) is represented by t; the variables with subscripts n are the

given starting values associated with time = tn; subscripts n+1 refer to the variables at

current time t; and and are time integration parameters. The characteristics of the

Newmark method depend on its integration parameters, where it is second-order accurate if

= 0.5 and first-order accurate otherwise. The Newmark scheme attains numerical damping

characteristics by selecting values larger than 0.5 for and choosing the smallest value for

that is compatible with the stability requirements (Hughes 1983). Nonetheless, as the

numerical damping introduced to the solution affects its behaviour at low frequencies, the

accuracy of the method decreases to first-order. Low-frequency properties can be optimised

to maintain second-order accuracy while preserving high-frequency damping. This

enhancement is usually introduced into the Newmark method by expressing different terms of

the equation of motion in an average form with different degrees of forward weighting.

Following this procedure, three well-known methods have been developed, including the

WBZ- method (Wood et al. 1981), where forward weighting of the inertial term is used; the

HHT- method (Hilber et al. 1977), where forward weighing is applied on the stiffness and

load terms; and the generalised- method (Chung and Hulbert 1993), which takes advantage

of two different forward weightings on the stiffness and inertial terms. Kontoe et al. (2008)

applied these methods to a few geotechnical problems and conducted a comparative study of

integration schemes in terms of their accuracy and numerical behaviour in order to choose the

25

most appropriate integration method. Taking into account the dissipative characteristics and

accuracy of the considered methods, Kontoe et al. (2008) concluded that the generalised-

method is more accurate and has better numerical dissipation characteristics than the other

dissipative schemes. In addition, a significant advantage of this method is that it allows the

analyst to control the magnitude of numerical dissipation at a high-frequency limit without

drastically affecting the lower modes. The generalised- approach is adopted in this thesis to

solve the equation system in Eq.2.58, as presented in the next section.

2.7.1 Generalised- method

The generalised- method is a two-parameter scheme obtained from a linear combination of

the HHT- and WBZ- methods. In this approach, Newmarks recursive relations (Eqs 2.62

and 2.63) are used to approximate displacements and velocities at time tn+1. Similar

expressions are also used here to approximate the pore-fluid pressures and Darcy velocities as

follows:

t

+ 2 P

(1 2 ) P

Pn +1 = Pn + tP n +

n

n +1

2

2.64

+ P

P n +1 = P n + t (1 ) P

n

n +1

2.65

+V

Vn +1 = Vn + t (1 ) V

n

n +1

2.66

As outlined earlier, the generalised- method takes advantage of two different forward

weightings on the stiffness and inertial terms to enhance the accuracy and damping

characteristics of the Newmark scheme. Accordingly, the inertia term is evaluated at time

tn +1 m of the considered interval t, whereas all other terms are evaluated at an earlier time

tn +1 f , in which f and m are two integration parameters, with f m . Intermediate state

values are then approximated through a convex combination of the end-point values as

follows:

tn +1 m = (1 m )tn +1 + m tn

2.67

tn +1 f = (1 f )tn +1 + f tn

2.68

26

+ U

(1 m )U

U

n +1 m =

n +1

m n

2.69

+ U

U

(1 f )U

n +1 f =

n +1

f

n

2.70

U n +1 f =

(1 f )U n +1 + f U n

2.71

+ V

V

=

(1 m )V

rn +1

rn +1

m rn

2.72

Vrn +1 =

(1 f )Vrn +1 + f Vrn

2.73

P n +1 f =

(1 f )P n +1 + f P n

2.74

Pn +1 f =

(1 f )Pn +1 + f Pn

2.75

Fn +1 f =

(1 f )Fn +1 + f Fn

2.76

Figure 2.1 depicts the evaluation of different terms within time increment t based on the

generalised- scheme.

t

ft

Stiffness and

damping terms

mt

tn

inertia term

tn+1

Figure 2.1: Evaluation of the various terms of the equation of motion in the generalised-

scheme

The generalised- method is a general scheme that includes the HHT- method (m = 0), the

WBZ- method (f = 0) and the Newmark method (m = f = 0). The unconditional stability

of the scheme is guaranteed when:

27

m f 0.5 ,

1 + 2( f m )

4

2.77

= 0.5 m + f

2.78

The optimal algorithmic parameters for this scheme are presented based on the value of the

spectral radius at infinity as:

=

m

2 1

1

=

, f

,=

(1 m + f ) 2

+ 1

+ 1

4

2.79

where the dissipation is equal to zero when = 1 , but as decreases, the numerical

dissipation increases. A detailed characteristic of the generalised- method can be found in

Chung and Hulbert (1993).

The system of equations in Eq. 2.58 can be expressed according to the generalised- scheme

as follows:

M ss U

f ns+1 f

n +1 m + M sr Vrn +1 + C ss U n +1 f + C sr Vn +1 f + K U n +1 f + K sp Pn +1 f =

2.80

M rs U

f nr+1 f

n +1 m + M rr Vrn +1 + C rr Vrn +1 + K rp Pn +1 f =

2.81

C ps U

f np+1 f

n +1 f C pr Vrn +1 + C pp Pn +1 f + K pp Pn +1 f =

2.82

Incorporating Eqs 2.622.76 into the above equations, the following set of incremental

equations is obtained:

1 m

(1 f )

M ss +

C ss + (1 f )K

2

t

t

1 m

M rs

t 2

(1 f )

C ps

t

1 m

M sr + (1 f )C sr

t

1 m

M rr + (1 f )Crr

t

(1 f )C pr

28

U F s

n +1

F r

(1 f )K rp

Vr =

n +1

P F p

n +1

(1 f )(

C pp + K pp )

t

(1 f )K sp

2.83

Explicit expressions for vectors F are presented in Appendix A.IV. In order to account for

material and geometric nonlinearities, K is replaced with K NL as defined in Appendix A.II.

Nonlinearities arise because all matrices and vectors are configuration-dependent, including

some of the terms in the external force vectors. In addition, the material stress-strain model

itself may be nonlinear. The nonlinear system of equations in Eq. 2.83 can be rewritten as:

R ( X) = 0

2.84

Rs

R = Rr

R p

2.85

U

X = Vr

P

2.86

where:

(1 f )

(1 m )

(1 m )

Rs

M ss +

Css + (1 f )K U +

M sr + (1 f )Csr Vr

=

2

t

t

s

+ (1 f )K sp Pr Fn +1 =

0

=

Rr

(1 m )

(1 m )

M rs U +

M rr + (1 f )Crr Vr + (1 f )K rp =

P Fnr+1 0

2

=

Rp

(1 f )

C ps U (1 f )C pr Vr + (1 f )(

C pp + K pp )=

P Fnp+1 0

t

2.87

2.88

2.89

At each time step, Eq. 2.84 needs to be solved by an iterative process until a prescribed

tolerance is achieved. The solution to the system in Eq. 2.84 may be found using the

NewtonRaphson algorithm. Letting superscript i denote the iteration number, this scheme

takes the form:

=

Xi Xi 1 + d Xi

29

2.90

R

d X = R ( Xi 1 )

X

i

2.91

and the Jacobian matrix R/X is evaluated at Xi-1. Differentiating Eq. 2.84 and neglecting

second derivatives with respect to X gives the required Jacobian matrix as:

R

X

1 m

(1 f ) i 1

1 m

M ss +

Css + (1 f )K i 1

M sr + (1 f )Cisr1

(1 f )K isp1

2

t

t

t

1 m

1 m

i 1

i 1

M rs

M rr + (1 f )Crr

(1 f )K rp

t 2

t

(1 f ) i 1

i 1

C ps

C pp + K ipp1 )

(1 f )Cipr1

(1 f )(

t

t

2.92

It is notable that if the stiffness matrix K is formed using the so-called elasto-plastic rate

constitutive continuum formulations, the Jacobian matrix will be only an approximation to

R/X, and the rate of convergence of the iteration will not be quadratic. However, with the

use of small time increments in the vicinity of highly nonlinear behaviour, the number of

iterations required for each time step is usually low. The selection of an appropriate time step

size may be accomplished automatically by using an error-control mechanism within an

automatic time stepping algorithm. Such an algorithm was introduced and used successfully

by Sloan and Abbo (1999) for quasi-static consolidation problems. It is notable that, in

uncoupled analyses, the performance of numerical models depends on the choice of stress

integration and load-stepping scheme (Oritz and Simo 1986; Borja 1991), while for coupled

problems, it is also necessary to select an appropriate time-stepping scheme (Booker and

Small 1975; Vermeer and Verruijt 1981).

The concept of consistent linearisation proposed by Simo and Taylor (1985) for a singlephase system can be used to derive a tangent operator consistent with the integration

algorithm used in the incremental problem, by which quadratic convergence may be obtained.

Borja (1989) applied the concept of consistent linearisation to some quasi-static elasto-plastic

consolidation problems. To avoid further complexity, a continuum elasto-plastic operator is

utilised in this thesis.

Two convergence criterions are considered for terminating the iteration procedure: (a) the

displacement criterion, which measures a relative error in the displacement component of X

as:

30

d Ui

Ui

2.93

the displacement tolerance; and (b) the force convergence criterion, which sets a limit for the

norm of the out-of-balance forces compared to the norms of the incremental and accumulated

external forces Fext as:

R in +1

R

Fext

2.94

where R is the force threshold. If either value of the relative displacement or the out-ofbalance forces is seen to increase, the solution has begun to diverge from the true solution

and the analysis should be aborted. If the unbalanced forces at the start of the time step are

not small, the solution may tend to drift from equilibrium as it is marched forward. It is

notable that there is no need to check convergence of the nodal velocities and accelerations;

this is inherently included with the convergence of the nodal displacements because the nodal

velocities and accelerations are expressed in terms of the incremental nodal displacement by

the time discretisation scheme. Typical values for the iteration tolerance are in the range of

10-3 to 10-6, with the lower limit ensuring that the drift from equilibrium is very small.

Another challenge in dynamic FE analyses is to cope with spurious wave reflections from the

artificial boundaries of the problem domain. Due to the fairly high wave velocity in soils and

rocks, modelling a large portion of the domain is not an effective way of proceeding, as the

waves reflected from the boundaries will probably have enough time to return to the area of

interest. Therefore, special boundary techniques must be facilitated to incorporate the

radiation condition of the truncated infinite domain into the finite numerical model. Two

fundamental approaches are usually used in the literature to suppress undesired reflections.

The first method uses absorbing boundary conditions (ABCs) on the restraining boundary to

simulate the radiation of energy towards infinity. These boundary conditions can be

interpreted as constitutive equations for interaction forces between the near and far fields.

The second approach utilises infinite elements on the artificial limits of the system. An

31

infinite element is in fact a semi-infinite radial strip with some nodes at infinity, and its

adopted shape functions represent the asymptotic behaviour of the solution at infinity.

Several artificial boundaries have been developed in the past two decades to eliminate the

effects of reflected waves from the response of wave propagation analysis. These ABCs are

classified into two categories: global (also consistent transmitting) and local boundary

conditions. The nonlocal boundary condition is described through integro-differential

operators, which are computationally expensive and difficult to implement. This ABC is

generally utilised in the framework of frequency-domain analysis because the method is

based on the principle of superposition, viz., linear behaviour is assumed for both the near and

far fields. In contrast, local ABCs are used in the framework of time domain analysis; hence,

they can be applied in nonlinear FE analysis (near field nonlinear, far field linear). Consistent

boundaries are rigorous and thus exactly satisfy the radiation boundary, while the radiation

condition in local type is approximately satisfied. The thin layer and coupled boundaryFE

methods are two distinct types of consistent boundaries. The viscous boundary of Lysmer and

Kuhlemeyer (1969) is the most widely used local absorbing boundary in numerical

engineering mechanics. It has been implemented in many general-purpose FE programs,

including Abaqus, ADINA and Ansys. The popularity of this method is due to its simple

physical interpretation in the form of lumped dampers (dashpot), whose absorption

characteristics are independent of frequency and can be easily implemented in FE codes.

However, it is known that the performance of the standard viscous boundary deteriorates as

the position approaches the source of excitation (Castellani 1974; Wolf 1988). This relies on

the fact that some of the wave energy does not radiate on the closed region, including surface

waves and body waves with angles of occurrence of less than 30. More importantly, it fails

for static loads because dashpots have no static stiffness. A detailed survey on local and

nonlocal boundary conditions, as well as their areas of application, can be found in Givoli

(1991) and Lehmann (2007).

A number of ABCs has been proposed for saturated porous media as extensions of the

methods for solid dynamics. These methods are essentially derived by utilising paraxial

approximations proposed by Clayton and Engquist (1977). Modaressi and Benzenati (1994)

developed an ABC for the U-P approximation of saturated porous media based on the

paraxial element, neglecting the second longitudinal wave (P2 wave). Degrande and De

Roeck (1993) derived a global ABC in the frequency domain by utilising an analytical

32

solution for the wave propagation problem. Akiyoshi et al. (1994) used zeroth-order paraxial

boundaries to devise local boundary conditions for U-P, U-w and U-u formulations of a

linear and isotropic saturated porous medium, where w and u in their formulations denote the

relative fluid velocity and fluid displacement, respectively. These boundary conditions are

almost equivalent to the viscous boundary; consequently, they cannot model the stiffness of

the unbounded domain. Moreover, when developing the explicit solution for the boundary

conditions, the second compression waves have been ignored. Modaressi (1995) presented a

note on the work of Akiyoshi et al. (1994) and incorporated the effects of P2 waves in their

work. Later, Akiyoshi et al. (1998) introduced Lames constants and applied the paraxial

method to Biots two-phase theory and generalised the previously proposed boundary

condition for isotropic, transverse isotropic and anisotropic two-phase saturated porous media

in a U-w formulation. Zerfa and Loret (2004) proposed an ABC for transient analyses of

saturated porous media, which consists of applying viscous tractions along the artificial

boundary. They derived the viscous tractions by assuming drained conditions, infinite

permeability, in which case no coupling occurs between the solid and fluid phases.

Another approach to modelling the unbounded nature of the domain is to use the infinite

element method (IEM). This approach is analogous to the FEM because it uses the idea of

interpolation functions to predict the different variables based on their nodal values. What

distinguishes these two methods is the use of decay terms in the shape functions of the

infinite elements (IEs) to ensure the decay of the variables at large distances. The problem

domain in this method is divided into two regions: the near-field, where the modelling is

purely based on the conventional FEM; and the far field, where the IEs are used to ensure the

role of infinity. Decay terms, which generally have a reciprocal or exponential form, are

obtained based on the analytical functions describing the far field of the problem. The idea of

the IEM was first introduced by Ungless (1973) and Bettess and Zienkiewicz (1977) in the

context of static single-phase elastic media. Zienkiewicz et al. (1983) proposed and

elaborated a novel boundary IEM, which is available in Abaqus software to model infinity for

static analyses. Simoni and Schrefler (1987), Selvadurai and Karpurapu (1989), and Xia and

Zhang (2006) also presented an application of IEs in fluid flow and soil consolidation. The

IEM was extended to the wave propagation problem in solid media by other researchers,

including Chow and Smith (1981), Medina and Taylor (1983), Zhao and Valliappan (1993),

and Zhao et al. (1992). Khalili et al. (1997) extended the concept of IEM to one-dimensional

wave propagation in two-phase saturated media, and Khalili et al. (1999) then generalised it

33

for a two-dimensional case of two-phase mixtures saturated with a slightly compressive fluid.

Zhao (2009) published a book about dynamic and transient IEs.

A survey in the literature reveals that most of the ABCs developed in the context of twophase saturated mixtures are in the form of viscous boundary conditions. These methods are

applicable for some variant solutions of Biots theory, including U-P, U-u and U-w

formulations. However, in this thesis, the proposed solution scheme for the governing

equations is in a mixed U-P-V form. Considering the complexity of deriving such boundary

conditions for this solution scheme, an alternative approach that is relevant to the aim of the

study may be preferable. In the next section, two alternative approaches are outlined, and one

of them is described and implemented into SNAC.

a saturated porous medium (Biot 1956). Dilatational waves can be decoupled into two waves:

P1 and P2. P1 waves propagate fasterindependent of frequencyand attenuate slower

compared to P2 waves, known as Biots slow wave. P2 waves are proportional to the square

root of the permeability and frequency (e.g., Akiyoshi et al. 1994). S-waves are transmitted

only in the solid constituent and are mainly governed by their shear stiffness, while the

propagation of acoustic waves essentially depends on the frequency of excitation, hydraulic

permeability and mechanical properties of the constituent materials (Corapcioglu and Tuncay

1996; Straughan 2008).

For the dynamic problems of concern in this thesis, two types of the bulk wavesP2 waves

and S-wavesare predominant. This is mainly because of the conditions of materially

incompressible solidfluid aggregates, low-frequency excitations and the relatively low

permeabilities present in the problems of interest. Under such circumstances, very low

relative motions between the solid matrix and the viscous pore fluid are likely, and body

waves are mostly transmitted via the structure of the solid skeleton. Hence, a local

transmitting boundary, such as the standard viscous boundary of Lysmer and Kuhlemeyer

(1969), may be utilised to ensure the absorption of the arriving elastic energy. However, as

outlined earlier, the standard viscous boundary embodies no static stiffness. Therefore, it

cannot model a static problem, and rigid body movement can occur in low frequencies.

34

Alternatively, two different approaches may be followed to consider the effects of the quasistatic response of the far field. The first approach is to use the IEM in combination with the

standard viscous boundary. In this methodology, the near field is discretised with the FEM,

whereas the far field is discretised using the mapped IEM in the quasi-static form (Marques

and Owen 1984; Selvadurai and Karpurapu 1989). In addition, the standard viscous boundary

can be utilised to absorb the dynamic waves at the FEIE interface. The second alternative is

to use the cone boundary proposed by Kellezi (2000), which consists of both dashpots and

springs. In this thesis, the second procedure (the cone boundary) is adopted in order to deal

with both body waves and surface waves. However, as the first approach is also perceived to

be an effective method, it is proposed for future research works.

This section presents the concept of the cone energy-absorbing boundary and derives the

coefficients of springs and dampers for the 2D plane strain, as well as the axisymmetric

conditions. The cone boundary is then implemented into SNAC to deal with the radiation of

the body and Rayleigh waves.

The constitutive relation for travelling S-waves in the context of 1D wave theory at any point

y for an (x, y) coordinate system is expressed as:

( y, t ) = cs2

u

y

2.95

denotes the shear wave velocity as:

cs =

E

2 (1 + )

2.96

where E denotes the modulus of elasticity and is the Poissons ratio. In plane strain

conditions, body waves propagate radially outwards along a cylindrical wavefront in which a

cylindrical P- or S-wave travelling in the positive y direction can be approximated by:

35

=

u ( y, t )

1

f ( y cs t )

y

2.97

=

w( y, t )

1

f ( y c pt )

y

2.98

where w denotes the displacement in y direction and cp is the P-wave velocity defined by:

cp =

E (1 )

(1 + )(1 2 )

2.99

Using Eq. 2.97 in Eq. 2.95, the following expression for boundary stress for S-wave

propagation is obtained:

c2

s u ( y, t ) + cs u ( y, t )

( y, t ) =

t

2y

2.100

From Eq. 2.100 and Eq. 2.95, the differential equation for outgoing S-waves can be expressed

as:

cs

0

t + 2 y + cs y u =

2.101

Likewise, the boundary stress condition for dilatational wave propagation and the differential

equation for outgoing P-waves can be written, respectively, as:

c 2p

w( y, t ) + c p w( y, t )

( y, t ) =

2 y

cp

+ cp w =

0

+

y

t 2 y

2.102

2.103

where is the normal stress. So far, the shape of the 1D propagating medium has not been

addressed. The equation of motion for the considered model can be determined in the same

way as for the rod model, by the product of two complementary boundary operators, which in

the case of S-waves is:

36

1 1

1 1

+

+

+ u =

0

cs t 2 y y cs t 2 y y

2.104

Ignoring the term 1/4y2 in Eq. 2.104, the differential equation of the model can be written as:

1 2u 1 u 2u

+

=

0

cs t 2 y y y 2

2.105

which resembles the differential equation of a horn with linear area variation (Graff et al.

1975). Therefore, the semi-infinite truncated conical rod represents the physical interpretation

of the considered model. In the 1D strength-of-materials approach, it is notable that a conical

bar model is commonly employed to represent the soil medium. The choice of a conical bar is

based on the fact that stresses act on an area that increases with depth due to geometric

spreading. Wolf and Deeks (2004) presented detailed descriptions of the cone models.

According to Eqs 2.100 and 2.102, the missing part of the linear cones from a truncated

boundary can be modelled by a mechanical system containing two series of springs and

dashpots that are oriented normal and tangential to the boundary. The stiffness of the springs

varies linearly in inverse proportion to the double-apex axis of the linear cones. Further, the

characteristics of the springs and dampers are frequency-independent. These physical

characteristics make it possible to solve the equations of motion in the time domain and

consider nonlinearity in the near field. Kellezi (2000) suggested that these models can be

used as transmitting boundaries for body waves in plane strain analyses. Multiplying both

terms of Eqs 2.100 and 2.102 by surface area A results in equations of equilibrium at the

artificial boundary as:

Q + K hpl u (r , t ) + Ch

u (r , t ) =

0

t

2.106

N + K vpl w(r , t ) + Cv

w(r , t ) =

0

t

2.107

where y = r is the apex of the truncated cone; Q and N are the shear and axial forces,

respectively; K hpl and K vpl are the coefficients of springs; and Ch and Cv denote coefficients

of dampers defined as:

37

K hpl = A(r )

K = A(r )

pl

v

cs2

2r

c 2p

2r

2.108

2.109

Ch = A(r ) cs

2.110

Cv = A(r ) c p

2.111

where superscript pl denotes plane strain conditions. It will be shown that the dampers

coefficients for axisymmetric conditions are also obtained from Eqs 2.110 and 2.111.

Rayleigh surface waves (R-waves) in plane strain analyses propagate along an infinitely long

rectangular surface with height equal to one R-wave length R. Accordingly, Kellezi (2000)

suggested using an appropriate boundary condition at lateral boundaries from the free surface

to a depth equal to R in order to replicate the stress conditions imposed by the R-wave. For

plane strain analyses, only dampers are employed to absorb the outgoing R-waves. The

required changes in the above dashpot coefficients (Eqs 2.110 and 2.111) are to replace cs by

R-wave velocity (cR,) and change cp to scR as:

ChR = A(r ) cR

2.112

2.113

s=

2(1 )

(1 2 )

2.114

cR = cs

(0.862 + 1.14 )

(1 + )

38

2.115

Figure 2.2 shows the application of the cone transmitting boundaries for the cases of 1D and

2D models. According to Figure 2.2(b), the wave direction vector is denoted by r, and the

vector normal to cone boundary is represented by n.

Source of excitation

x

R

r

n

r

Ch

Kh

(a)

(b)

Figure 2.2: (a) Semi-infinite 1D conical rod model; (b) application of cone model for 2D

problems

In the derivation of the cone boundary, it was assumed that the vectors r and n coincided, so

the methodology needed to be generalised for the cases in which these vectors took different

directions, as shown in Figure 2.2(b). Kellezi (1998) introduced the interaction factor in the

stiffness terms, which scales r (the distance from the boundary node to the source location) to

account for non-coincidence between r and n. In fact, is a function of the scalar product of

the vectors r and n, and it is equal to one for a circular boundary surface. Kellezi (1998) took

the interaction factor equal to = 1.31.5. However, it is suggested that is determined

experimentally by conducting numerical tests and comparing them with closed-form

solutions.

In axisymmetric conditions, the propagation of body waves has a spherical wavefront. In the

context of a 1D model, outgoing spherical P-waves can be approximated by:

( y, t )

w=

1

f ( y c pt )

y

2.116

Comparing Eq. 2.116 with the one used for waves with a cylindrical wavefront (Eq. 2.98)

reveals that spherical waves attenuate more quickly than cylindrical waves, as the wave

amplitude decreases at a rate of 1/y. If a similar procedure to the plane strain condition is

39

followed, the same coefficients for dampers are obtained for axisymmetric conditions (Eqs

2.110 and 2.111); however, coefficients for the springs are doubled as:

K axi

h = A( r )

K vRaxi = A(r )

cs2

r

2.117

s 2 cR2

2 r

2.118

body waves in plane strain conditions also propagate in a cylindrical wavefront. Therefore,

the linear cone models can be employed to construct a lateral transmitting boundary for Rwaves in axisymmetric conditions. Accordingly, the following coefficients for the springs are

obtained:

K hRaxi = A(r )

cR2

2 r

2.119

K vRaxi = A(r )

s 2 cR2

2 r

2.120

and the dampers coefficients are achieved from Eqs 2.112 and 2.113.

2.8.2.1 Implementation

This section presents the implementation of the cone boundary into SNAC for 2D plane strain

and axisymmetric analyses. To implement the cone boundary, the damping matrix Css and the

stiffness matrix K (Eq. 2.58) of elements located at the boundaries of the problem domain

should be modified. The stiffness matrix at the boundary is formulated as:

K=

K + K B

2.121

where KB is the contribution from the cone boundary to the element stiffness matrix. The

consistent damping matrix is also modified as:

40

C=

Css + C B

ss

2.122

where CB is the contribution from the cone boundary to the element damping matrix.

The shear and normal stresses arising from wave propagation spread continuously along the

boundary surface. Therefore, both the dashpots and springs need to be applied continuously

along the boundary of the mesh and should not be considered discrete dashpots or springs

positioned at boundary nodes. However, the contributions of each element should be

expressed based on equivalent nodal springs and dashpots before they can be assembled into

the global matrices. The contribution of each element to the global matrices takes the form:

=

KB

K cNs d

2.123

N C N d

2.124

T

s

=

CB

T

s

where denotes the element side over which the damper and the spring are applied; Ns is the

interpolation functions consistent with the solid displacement (Eq. 2.54); and Kc and Cc are

constitutive stiffness and damping matrices, respectively (see Table 2.1). The above

equations only contribute to the global stiffness and damping terms of the nodes along the

element side. Thus, the integrations in Eqs 2.123 and 2.124 must be evaluated for each

element side.

The surface integrals are transformed into corresponding 1D form in a natural coordinate

system as:

1

K B = BT K c B J ldT

1

C B = BT Cc B J ldT

1

2.125

2.126

where l is unity for plane strain problems and equals 2r for axisymmetric problems; B

contains the interpolation functions on the element side; J is the Jacobian determinant

obtained from mapping the element side from the global element to the parent element; and T

denotes the natural ordinate that varies from -1 to +1 over the element length (see Figure 2.3).

41

Plane strain analysis

Body waves

c

Cc = p

0

0

cs

c

Cc = p

0

c 2p 0

2 r 0 cs2

Kc =

Rayleigh waves

Axisymmetric analysis

sc

Cc = R

0

Kc =

0

cR

c 2p 0

r 0 cs2

sc

Cc = R

0

Kc =

Kc = 0

0

cs

0

cR

s 2 cR2 0

2 r 0 cR2

For the six-noded isoparametric element in Figure 2.3, assuming that the dashpots and

springs are applied along the right-hand side edge of the element, B takes the form:

B

B= 1

0

0 0 0 0 B2

B1 0 0 0 0

0

B2

0 0 B3

0 0 0

0

B3

0 0 0

0 0 0

2.127

1

T (T 1)

2

B2= (1 T 2 )

=

B1

=

B3

2.128

1

T (T + 1)

2

The Jacobian determinant for each point on the element side is given by:

1

dx 2 dy 2 2

J

=

+

dT dT

2.129

To conduct a dynamic coupled analysis, it is usually necessary to establish the initial stresses

in the problem domain due to soil self-weight through a quasi-static coupled analysis. This

becomes inevitable if a soil constitutive model such as Modified Cam Clay is utilised, as the

42

initial effective stresses are required to define the location of yield surfaces at each Gauss

point.

-1

Figure 2.3: Six-noded isoparametric element with cone boundary applied on its lateral

edge

To avoid numerical difficulties and rigid body movement due to the absence of fixity at the

FE mesh boundaries, the analyses may need to be conducted in two stages. In the first step

concerning the application of body force, the bottom boundary is fixed both in the horizontal

and vertical directions, whereas the lateral boundaries are only fixed in the horizontal

direction. In the second stage, all fixities are removed from the boundaries and, subsequently,

equivalent reaction forces are applied on the corresponding boundary nodes. Finally, the

absorbing boundaries are applied and the analysis proceeds to the dynamic stage. An

alternative approach is to apply absorbing boundaries with large coefficients for the springs

in order to model fixed boundaries. After the body forces have been established, the springs

coefficients are corrected based on the real material properties and the analysis proceeds to

the dynamic stage.

It is also notable that in nonlinear analyses, the constitutive damping and stiffness matrices of

spring-dashpot systems are updated during NewtonRaphson iterations.

2.9 Summary

The first part of this chapter presented the balance laws governing the response of two-phase

saturated porous media. The variational statements of the resulting differential equations were

derived, and the FE discretisation of these equations was also presented. The treatment of

43

large deformations and the mesh refinement strategy within an ALE scheme was briefly

outlined. Closed-form solutions were also developed for some 1D problems involving the

dynamic response of saturated porous media (see Appendix A.III).

The second part of this chapter detailed the discretisation of the governing equations in the

time domain using the generalised- method. The arising incremental form of these equations

was then presented for use in Newton iterations.

The third part of this chapter provided a detailed literature review on energy-absorbing

boundaries. The concept of the cone absorbing boundary was elaborated, and the coefficients

of springs and dampers proposed by Kellezi (2000) were derived for 2D plane strain and

axisymmetric conditions. Finally, the cone boundary was implemented into SNAC in order to

deal with radiation of body and Rayleigh waves.

Chapter 3

44

Saturated Porous Media

3.1 Introduction

The analysis of soil-structure interaction in the framework of the finite element method

(FEM) is generally accomplished using contact mechanics, in which kinematic relations are

employed in the treatment of interactions between deformable bodies. Accordingly, the

contributions of each of the bodies in contact are included in the governing equations of the

entire domain using an algorithm suitable for solving constrained optimisation problems. The

main contact contribution arises from the non-penetration condition of the contacting bodies

(i.e., as a purely geometrical constraint). This condition is fulfilled by enforcing constraints

on the displacement vector of the active nodes located at the contact interface. When twophase saturated soil is studied, in addition to the requirement for continuity of the contact

traction, continuity must also be maintained for the Darcy velocity and pore-fluid pressure

across the interface via the enforcement of appropriate constraints. Analysis of the contact

mechanics of a porous medium and its FE formulation and implementation are challenging

tasks. Only a limited number of studies have proposed solution schemes for this type of

problem, and they usually ignore inertia effects and assume a frictionless interface. The

frictionless contact problems of biphasic cartilage layers that are applicable in biomechanics

have been addressed by Donzelli and Spilker (1998) and Chen et al. (2005) for small-strain

and finite-deformation regimes, respectively, using the Lagrange multiplier method. Ateshian

et al. (2012) also described studies for the quasi-static frictionless contact problems of

biomechanics.

The aim of this chapter is to formulate FE contact implementation for solid-fluid mixtures in

the spatial frame that can accommodate inertia effects together with finite deformation and

contact sliding. Both frictionless and frictional contact formulations are addressed. Therefore,

for a node that is in contact with a corresponding contacting pair, contact contributions

arising from constraining solid displacement, Darcys velocity and pore pressure are added to

the tangent stiffness matrix and the residual vector during Newton iterations.

45

In a frictional contact element, two conditionsstick and slipare distinguished on the basis

of the level of interface frictional force compared with the Coulomb frictional force. The

formulation of frictional contact is developed in terms of effective forces, which are integrals

of the effective stresses along the interface. To differentiate between the stick and slip cases,

the concept of a moving friction cone (MFC) is used (Wriggers and Haraldsson 2003), which

is a relatively efficient methodology for deriving the contact kinematics.

enforce the necessary contact constraints. The formulation of the contact kinematics and

constraints adopted in this thesis is based on the so-called mortar segment-to-segment

approach. The mortar method is adopted because high-order approximation functions can

then be used to interpolate different field variables. This is necessaryparticularly for twophase saturated consolidation problemsbecause the interpolation function for pore-fluid

pressure is generally chosen to be one order lower than the functions for the displacement and

fluid velocity in order to avoid numerical locking and instability (oscillations) in the

predicted pore-fluid pressure response. In the developed coupled consolidationcontact

algorithm, free-draining conditions are automatically adopted for the nodes that lose

connection with their possible contacting pair, whereas impermeable or semi-impermeable

conditions are adopted for nodes that come into contact with another surface.

This chapter discusses the development of the contact algorithm and describes various

contributions arising from the contact algorithm to the tangent stiffness matrix and residual

vector.

The so-called one-pass node-to-segment (NTS) discretisation method is widely used to

analyse large sliding and large deformation problems in contact mechanics (Hallquist et al.

1985; Wriggers and Simo 1985). In this method, an arbitrary sliding of a node over the entire

contact area is permissible through a relatively simple treatment of the contact kinematics.

However, it has been highlighted that this approach cannot pass the contact patch test

(Papadopoulos and Taylor 1990; Taylor and Papadopoulos 1991), wherein a flat contact

surface should be able to transfer a spatially uniform pressure from one body to another. In

46

contrast, the two-pass NTS scheme can satisfy the patch test for low-order elements such as

bilinear quadrilateral elements, but coupling with higher-order elements such as quadratics is

not feasible without losing accuracy in the displacements and stresses in the contact area.

This can be a particular disadvantage in soil mechanics, in which higher-order shape

functions are often used to improve accuracy and avoid possible locking effects.

Additionally, if the NTS approach is adopted for contact problems in two-phase saturated

porous media, the occurrence of errors in the flow and pore pressure of the interstitial fluid in

the contact surface is often another consequence. It is notable that, for algorithms that do not

pass the patch test, the associated errors do not necessarily decrease with mesh refinement.

Some details regarding the NTS method and the patch test are found in Zavaris and Lorenzis

(2009).

Another deficiency of the method is oscillation in the contact forces predicted by the NTS

technique because of the non-smooth surface profile of low-order elements. In large sliding

transient applications, non-smooth transition from one element to the next can induce nonphysical inertial discontinuities that contribute to oscillations in the solution. Such oscillatory

responses were observed for dynamic coupled problems by Sabetamal et al. (2014), for

dynamic penetration problems by Nazem et al. (2012) and for quasi-static problems by Sheng

et al. (2006) and Simo and Meschke (1993). The two-pass method also fails the Babuska

Brezzi condition (Brezzi and Fortin 1991), and it has the well-known deficiency of locking

due to over-constraint, in which ill-conditioning and poor convergence behaviour are typical

manifestations (Puso and Laursen 2004).

To overcome these issues, the so-called mortar-type method is adopted based on the domain

decomposition technique for nonmatching grids (Bernardi et al. 1992). Indeed,

nonconforming grids are the rule rather than the exception in contact and impact problems.

Even if the meshes match at initial contact, slipping may produce a nonmatching mesh in the

deformed configuration. Nonetheless, the mortar discretisation technique is a segment-tosegment approach that projects segments on one side of the contact surface onto the

corresponding segments of the opposing side. One of the surfaces in contact is called nonmortar and the other is called mortar, as shown in Figure 3.1. The contact constraints are

treated continuously in a weak integral form rather than locally at a number of finite

collection points, as it is considered in the NTS scheme. The integral variational terms are

evaluated by quadrature formulas on the discretised subareas by locating integration points on

47

the non-mortar boundary. The current position of the mortar segment defines the boundary of

an inadmissible region for the current positions of the integration points on the non-mortar

segment. The terms in the contact virtual work related to the deformation of the contact

bodies should be properly linearised because the mortar integrals and associated segment-tosegment projections are strongly dependent on the deformation of the contact bodies. The

mortar discretisation technique was first introduced to large deformation contact formulations

of single-phase materials by Puso and Laursen (2004) and Fischer and Wriggers (2005).

Comprehensive details and derivations can also be found in Wriggers (2006) and Laursen

(2002). This method is extended in the following sections to two-phase saturated materials.

t (B nm)

nm

nm

xnm

xm

t (B m)

nm

am

Non-mortar

guN

Mortar

(a)

(b)

Consider two deformable bodiesB nm and B mthat undergo finite deformations and come

into contact over respective non-mortar and mortar surfaces, denoted by nm = t (nm) and

m = t (m), where t indicates an operator relating the mechanical deformation between the

initial and current configuration at time t, as shown in Figure 3.1. The location of point x on

( denotes the respective body) can be related to the initial configuration X on via the

total displacement field U as:

x=

X + U

48

3.1

Kinematic relations must be formulated to constrain the separate bodies with respect to their

interactions. The interactive forces at a frictionless contact interface for a two-phase saturated

porous media arise from normal total stresses, fluid flow and porefluid pressure.

Geometrically, contact between two deformable bodies is described using the so-called

normal gap guN, which gives the minimum distance between two points on the contacting

surfaces as:

guN = (x nm x m ) n m

3.2

where x nm = x nm ( , t ) denotes the coordinates of a fixed point on the non-mortar surface and

x m = x m ( , t ) is the closest point of projection on the mortar surface. and [ 0,1] are the

discrete element-wise convective coordinates on the non-mortar and mortar surfaces,

respectively, with denoting the convective coordinates at the projection point. Here and in

the following, a bar over a quantity denotes its evaluation at the minimum distance point. The

projection point is obtainable from the orthogonality constraint expressed as:

(x

nm

xm ) am =

0

3.3

The normal vector n m and the tangent vector a m in Eqs 3.2 and 3.3 are associated with the

mortar surface. The tangential basis in the spatial frame is augmented by the outwards normal

to m at the projection point x m as:

e3 a m

n =

am

m

3.4

Generally, Eq. 3.3 is nonlinear in because of the curvature of the contacting surfaces and

large deformations. Therefore, the solution of the closest point on the mortar surface must

be found in an iterative manner for the discretised surfaces. The incremental update of

within a time increment can be obtained using Newtons method, as presented in Box 3.1.

49

Box 3-1: Newton scheme for the update of within time increment for frictionless

contact

Initialise:

=

i 0=

, i n (last converged solution)

LOOP over NEWTON iterations : i = 1,, convergence

Compute:

(i ) = i

=

(x

nm

xim ) aim = 0

Compute

i , =

aim

i +=

i

1

+ ( x nm xim ) xim,

i

, set i = i +1

i ,

END LOOP

The geometric and kinetic constraint conditions for non-adhesive contact are given by the

KuhnTucker complementary conditions, which can be expressed as:

guN 0,

tuN 0,

guN t N = 0

g uN t N = 0

3.5

where the normal stress tuN is equivalent to the negative contact pressure. If the bodies are

separated on a surface area ( guN > 0 ), contact stresses do not exist ( tuN = 0 ). When

interpenetration occurs between the two bodies ( guN < 0 ), contact constraints must be

established to keep guN = 0 throughout the solution, while also allowing consistent contact

pressure to be transmitted. The last requirement of Eq. 3.5, g uN tuN = 0 , is interpreted as a

persistency condition, implying that the rate of separation between contact surfaces must be

zero when the contact pressure is non-zero. Moreover, the continuity requirement of the

Darcy velocities and the balance condition of the porefluid pressures must be fulfilled across

the interface by imposing two other constraints.

According to Figure 3.2, the normal relative Darcy velocity at the point of contact is defined

as:

50

g vN = ( v rnm v rm ) n m

3.6

=

g p (p nm p m )

3.7

where v nm

and p nm denote Darcy velocities and pore-fluid pressure at a fixed point on a nonr

mortar surface, while v mr and p m represent their counterparts at the projection point on the

mortar surface, respectively. The constraints on the Darcy velocities, as well as the pore-fluid

pressures, are applied to set g vN and g p to zero across the interface. This guarantees,

respectively, the linear momentum balance of the fluid phase and the conservation of the

mixture mass for the entire domain B (B nm U B m).

non-mortar surface

non-mortar surface

pnm

nm

am

mortar surface

mortar surface

(xnm- xm)

xnm

xm

x2

x1

(a)

(b)

Figure 3.2: Definition of gap functions: (a) Darcy velocity and pore fluid pressure;

(b) displacement

The overall boundary value problem subject to the constraints given in Eq. 3.5, as well as

those related to the flow continuity, including g vN = 0 and g p = 0 , requires a solution to be

found in a constrained space. As a result, some limitations are induced by these physical

constraints on the admissible variational statements. A number of different methods are used

in the literature to remove these restrictions and enforce the contact constraints within the

FEM. The penalty approach is usually preferred to the Lagrange multiplier method, largely

because it provides a more stable solution strategy in the presence of nonlinearities such as

51

nonlinear elasticity, curved contacting surfaces and a variable number of contact constraints.

Instability usually ensues when the size of the global equation system changes as a result of

the activation and deactivation of contact constraints. The matter of a variable number of

contact restraints is more significant for problems involving saturated two-phase materials, as

a contacting boundary condition might be changed from a permeable boundary to an

impermeable one during the process of soil-structure interaction. In the penalty approach, the

contact forces are functions of displacements, pore pressures or Darcy velocities, so that no

new unknowns are introduced, and thus the equation system is not expanded as a result of the

interactions between bodies. However, the accuracy of the penalty method depends on the

choice of the penalty parameter. The penalty factors must be large enough to control the

interpenetrations and satisfy the continuity requirements. However, as is well known, penalty

parameters that are too large may produce severe numerical problems in the solution or

simply make a solution impossible. Such numerical issues might include a dramatic decrease

in the size of the critical time step in explicit procedures or an ill-conditioned coefficient

matrix in implicit solution schemes. The critical size of the penalty parameter that initiates illconditioning usually depends on both the original coefficient matrix and the precision of the

digital computer used to perform the calculations. When the global equation solver uses

Gaussian elimination without pivoting, ill-conditioning will be manifested as a loss of

numerical accuracy during the elimination stage, leading to either slow convergence or even

divergence. The penalty approach is used here, but it is regularised with an augmented

Lagrangian method in an attempt to improve the accuracy of the solution while avoiding illconditioning problems.

As the contact kinematics were derived based on total stresses, due to the assumption of a

frictionless contact, the treatment of the contact constraints arising from the displacement

contribution is independent of the other field variables. As a result, it can be treated in a

manner similar to a single-phase medium, as presented by Fischer and Wriggers (2005).

However, for the sake of completeness, a brief outline of the method of constraint

enforcement is presented here. The contribution of each contact segment to the total energy

of the two bodies can be incorporated using the penalty approach as:

52

1

uc = u gu2N d

2 c

3.8

tuN =

u guN . For

continuous spring stiffness giving the contact pressure as pN =

minimisation of the total potential energy, the variation of Eq. 3.8 is evaluated as:

ccu =

g uN d g u N d =

uN

d guN d

3.9

where:

d guN = (d U nm d U m ) n m

3.10

ccu is added to the functional d s in Eq. 2.52 as the contact contribution following from the

non-penetration condition. Linearisation of Eq. 3.9 is also necessary for the iterative solution

of the nonlinear global system of equations using Newtons method. It is given as:

ccu =

g

u

uN

3.11

found in Wriggers (2006). For 2D cases, it is evaluated by:

d g uN = (d U ,m + U ,m d + x,m d ) n m +

guN

a

m 2

3.12

The variation d can be calculated for large frictionless sliding problems by observing that

the distance defined in Eq. 3.2 must be minimised at any time. That is, the distance vector

must be orthogonal to the tangential vector at any time, as depicted in Figure 3.3. As a result,

the time derivative of the orthogonality constraint follows as:

d

(x nm x m ) a m =

0

dt

53

3.13

gu

T = t0

T=t

which yields:

( x

nm

x m x,m a m + ( x nm x m ) x ,m + x,m =

0

3.14

Using the definition of the displacement gap function (Eq. 3.2) in Eq. 3.14 provides:

m 2

x, guN n

( x nm x m ) a m + guN n m x ,m

3.15

From this equation and the relation between a time derivative and a variation, the following

expression is obtained:

=

d

a

m 2

x, guN n

(d U nm d U m ) x, + guN n m d U ,m

3.16

The weak forms given by Eq. 3.9 and Eq. 3.11 are discretised in this thesis by considering a

quadratic geometric approximation of both the non-mortar and mortar surfaces. As a result,

every non-mortar surface segment contains three non-mortar nodes. This results in the

following discretisation:

3

U nm ( ) = U inm N inm ( q )

3.17

i =1

3

U m ( ) = U im N im ( q )

i =1

54

3.18

where N denotes the quadratic shape functions. A Gauss-point rule is also employed to

evaluate the integrations over every non-mortar segment in which q and q , used in the

above equations, denote the convective coordinates of a Gauss point on the non-mortar

segment and its projection on the mortar segment, respectively. Thus, the approximated

variation of the potential energy and the linearisation of the constraint can be written in

matrix form, respectively, as:

ccu = d UT R uN

3.19

=

ccu d UT K u U

3.20

where the included residual vector R uN and the contact stiffness matrix K u are evaluated by:

nq

R uN = u guN Bu n L q wq

3.21

T T

T T

+

(

Bu nnT BTu 2 uN

B

na

B

B

an

B

,

,

u

u

u

u

nq

a guN cnT

L w

Ku = u

T

q q

cn

q =1

T T

T T

+ guN Bu , nn Bu , + 2 Bu aa Bu )

a

3.22

q =1

in which L q is the tangent on the non-mortar side; wq denotes the weighting related to each

local quadrature point q; nq is the number of integration points per element surface; Bu

contains displacement shape functions; and Bu , and Bu , contain, respectively, the first- and

second-order derivatives of the shape functions with respect to as:

N1nm ( q )1I

nm

N 2 ( q )1I

N nm ( )1I

q

3

Bu =

Bu ,

=

N1m ( q )1I

m

N 2 ( q )1I

N m ( )1I

3 q

0

=

Bu ,

m

N1, ( q )1I

N m ( )1I

2, q

N 3,m ( q )1I

and:

55

0

m

N1, ( q )1I

N m ( )1I

2, q

N 3,m ( q )1I

3.23

xT = (x1nm

x nm

2

=

a BTu , x =

n

x3nm

x1m

x m2

x3m )

e3 a

=

c a=

BTu , x

,

e3 a

3.24

The discrete form of the contact virtual work is thus equal to the sum of the individual

quadrature point contributions.

The pore pressures at each Gauss point on the non-mortar segment p nm and the point of

contact on the mortar segment p m are interpolated from the corresponding nodal values pi

using linear shape functions as:

(1 q ) p1nm + q p2nm

p nm ( q ) =

p m ( p=

) q p1m + (1 q ) p2m

3.25

The difference in pore pressure across the interface (gap function) defined in Eq. 3.7 can be

expressed in matrix form as:

g p = PT B p

3.26

3.27

where:

1 q

Bp =

q

(1 q )

3.28

The contribution of the pore-fluid pressure constraint to the total energy of the two bodies in

contact can be formulated for each contact segment as:

56

1

cp = p g 2p d

2 c

3.29

where p denotes the penalty parameter associated with pore-fluid pressure. The variation of

Eq. 3.29 provides:

=

ccp

g d g d q

=

d g pd

3.30

where:

=

d g p (d p nm d p m )

3.31

ccp is associated with the virtual work of the internal flux, which is added as a contact

contribution to the functional d p in Eq. 2.46; and qN is equal to the negative of the flux

that would be required for the difference in pressure across the interface to vanish (i.e.,

qN = v r n = p g p ). It is notable that p has the same units as membrane permeability (i.e.,

length cubed per force, per time). By choosing a sufficiently large value for p , the

continuity requirement p nm = p m is enforced to within an acceptable tolerance. In a special

case when one of the contacting bodies is a non-porous solid, qN = 0 should be used inside

the contact surface. An approximation of Eq. 3.30 results in the residual vector, which for

every mortar segment can be expressed as:

nq

R p = p B p g p L q wq

3.32

q =1

The incremental form of the constraint condition in Eq. 3.30 may be written as:

=

ccp p ( p nm p m ) (d p nm d p m ) d p (d p nm d p m ) p,m d

c

p d p g p d

3.33

m

,

=

ccp d PT K p P + K pu U

57

3.34

where:

nq

K p = p B p BTp L q wq

q =1

=

K pu p

q =1

3.35

nq

B p , g p + B p B

a guN cnT

2

T

p ,

P ( ) Bu a + guN Bu , n L q wq

T

with:

0

0

=

1

1

B p ,

3.36

Accordingly, the virtual work performed by the pore-fluid pressure can be included in the

global equation system during the Newton iteration as:

{d U

dV

T

r

dP } 0

K pu

0 U

0 0 Vr =

{d UT

0 K p

P

0

dV

T

r

d P } 0

R p

T

3.37

Darcys velocity at each Gauss point on the non-mortar segments v nm

r

contact on the mortar segments v mr are interpolated from the corresponding nodal values v nm

ri

and v mri using quadratic shape functions Nr as:

3

nm

nm

v nm

r ( ) = v ri N ri ( q )

i =1

v ( ) = v N ( q )

m

r

i =1

m

ri

3.38

m

ri

The total relative Darcy velocity g v at the point of contact at the interface is calculated as:

=

g v ( v rnm v mr )

58

3.39

G v = BTv Vr

3.40

nm

nm

m

m

m

VrT = ( v nm

r1 , v r2 , v r3 , v r1 , v r2 , v r3 )

3.41

where:

N rnm

( q )1I

1

nm

N r2 ( q )1I

nm

N r3 ( q )1I

Bv = m

N r1 ( q )1I

N m ( )1I

r2 q

N rm ( q )1I

3

3.42

Only the normal component of the relative Darcy velocity g vN is set to zero by enforcing the

relevant constraint, while flow tangential to the interface is allowed.

Therefore, the constraint condition for the normal relative Darcy velocity can be formulated

in a similar fashion as:

=

ccvr

g d g d t

=

vN

vN

vN

d g vN d

3.43

ccvr is associated with the virtual work of the internal nodal forces. This term is added to the

functional d r in Eq. 2.49 as a contact contribution. Here, v is the penalty parameter

related to the Darcy velocity. It is worthwhile noting that tvN = v g vN represents the negative

of the force required to enforce no relative flux across the interface. This force acts at a Gauss

point on a non-mortar segment, and a force equal in magnitude but opposite in direction acts

on the mortar segment at the point of projection. The constraint equation in this method is

fulfilled in the limit v g vN 0 . The virtual gap vector of the Darcy velocity d g vN is

stated as:

m

m

d g vN = (d v nm

r d vr ) n

59

3.44

Note that the variation of the normal vector should not be included in the above equation, as

Eq. 3.43 represents the virtual work due to internal nodal forces arising from the virtual

Darcy velocity. Thus, the residual vector for one mortar segment is obtained from Eq. 3.43

as:

nq

R v = v B v nG Tv n L q wq

3.45

q =1

ccvr=

(g vNd g vN + g vN d g vN )d

3.46

where:

g vN =

( v

nm

r

m

m

v mr v mr , ) n m + ( v nm

r v r ) n

m

m

d g vN = d v mr , n m + (d v nm

r d v r ) n

3.47

n m =

am

a

m 2

( U

+ X, ) n m

3.48

=

ccvr d VrT [ K v Vr + K vu U ]

3.49

where:

nq

K v = v B v nnT BTv L q wq

3.50

q =1

K vu

2

2

nq

a

a

v

L q wq 3.51

q =1

1

T T

T T

T T

2 B v na B v Vr n B v , + g vN B v an B v ,

with:

60

K u

=

1

a guN cn

2

aBTu + guN Bu , nT

3.52

Finally, the virtual work performed by the Darcy velocity may be rewritten in matrix form as:

{d U

dV

T

r

0 0 U

0

d P } K vu =

K v 0 Vr

0

0 0 P

T

{d U

dV

T

r

d P } R v

0

T

3.53

To enforce contact constraints based on a user-defined tolerance and avoid possible illconditioning, the penalty method may be regularised with an augmented Lagrangian scheme

(Simo and Laursen 1992). The augmentation of the global contact problem is given by the

following equations:

tuN= u + u guN

q=

p + p g p

N

3.54

tvN= v + v g vN

where u , p and v are the Lagrange multipliers. The most common technique used in

mechanics for the solution of augmented Lagrangian problems is Uzawas algorithm. The

basic idea of the method is to solve the discrete versions of Eq. 3.54 and their corresponding

constraint formulations given in Eqs 3.9, 3.30 and 3.43 with the values of u , p and v

fixed to some estimate. These estimates are updated within an iterative scheme, and the

iteration process is continued until the exact multipliers (satisfying a user-defined tolerance)

are obtained. Table 3.1 presents the Uzawa-type algorithm adopted in this thesis. It is notable

that the problem of step 2 in Table 3.1 is highly nonlinear itself, so that before proceeding to

step 3, a converged solution is required. Although the iterations in steps 2 and 3 might be

conducted simultaneously (Wriggers et al. 1985), such methodology is not likely to be

superior to the adopted nested iteration scheme. This is largely because the multipliers are

completely fixed in the solution phase 2, and thus the theory of regularisation is linear

(Laursen 2002). This is consistent with the formulation based on the penalty approach

derived in the previous sections. Consequently, the only necessary changes in the earlier

61

expressions of the contact stiffness matrices and residual vectors are to replace t N = g N

with t =

+ g N ( symbolises either u or v) and substitute qN = p g p with

N

q=

P + p g p . These amendments are straightforward, as presented in Appendix B. An

N

additional important benefit of using the nested scheme is that it preserves the quadratic

convergence of the inner loop when a consistent NewtonRaphson solution scheme is

conducted.

The penalty parameters u , p and v are estimated for each contact pair as:

N

Ei Ai

Vi el

3.55

ki Ai

Vi el

3.56

li Ai

kiVi el

3.57

u = u

i =1

p =p

i =1

v = v

i =1

where N is the number of element faces representing the contact surface in the contact pair; Ai

denotes the area of the element face; Vi el represents the volume of each element; Ei is the

average Youngs modulus; ki is the mean hydraulic permeability of the element; and li

denotes the average length of the drainage path for each contact element. u , p and v are

user-defined non-dimensional scale factors. When a porous non-mortar surface is in contact

with an impermeable mortar surface, the Darcy velocity associated with mortar points is

equal to zero ( v mr = 0 ). The contact constraint on the Darcy velocity will then set the normal

Darcys velocity component at the corresponding non-mortar points equal to zero, and only

flow that is tangential to the interface is allowed.

62

Table 3.1: Nested augmented Lagrangian scheme for frictionless contact problems of

two-phase saturated porous media

1- Initialisation (main loop)

Initialise the augmented Lagrangian counter i and assume the first estimates for the multipliers

as:

i=0

u(i ) = u

n +1

(i )

pn+1

= pn

(i )

vn+1

= vn

Solve nonlinear global equation system for the current increments ( U (n+k )1 , Pn+( k1) , and Vn+( k1) )

where the contact tractions for the active contact pairs ( t N < 0 ) are incorporated in the solution

as:

(i )

t=

u(ni+)1 + u gu(Nk )n+1

N n+1

(i )

p(in)+1 + p g (pkn+)1

q=

N n+1

(i )

tvN

=

v(ni+)1 + v gv(Nk )n+1

n +1

Check the convergence of the Newton scheme and continue the iterations (k) until a user-defined

tolerance for the solution of the global system presented in Eq. 2.83 is attained.

3- Update

Update the Lagrange multipliers and the iteration counter as:

u(i +1) = t N(i )

n +1

n +1

n +1

( i +1)

vn+1

n +1

=t

(i )

vN n+1

i i +1

4- Return

Return to step 2 and repeat the procedure until all contact constraints are satisfied within a

tolerance.

63

In frictional contact mechanics, it is necessary to differentiate between two distinct kinematic

states of the contacting bodies: sticking and sliding. These conditions are distinguished on the

basis of the level of interface frictional force compared with the Coulomb force represented

as:

=

fs

tT t N

3.58

where tT and t N are, respectively, the tangential and normal effective stress components of

the total traction t at the contact interface, and denotes the interface friction coefficient.

The key difference between the formulations of frictional contact in solid mechanics and

saturated two-phase media arises from the evaluation of effective stress t N . The effective

stress at the contact interface can be obtained from the adopted governing equations of the

two-phase saturated porous media in Section 2.5. The first integral in Eq. 2.57 is associated

with the virtual work of the boundary traction due to total stresses, which in fact defines the

total force consistent with displacement interpolation Ns. Similarly, the first integral in Eq.

2.56 corresponds to the virtual work of the boundary pore-fluid pressure and defines the porefluid force consistent with Darcys velocity interpolation Nr. Therefore, the surface traction

resulting from effective stresses can be obtained as the difference between these two integrals

provided that Ns = Nr.

The concept of the moving friction cone (MFC) is utilised (Wriggers and Haraldsson 2003) to

differentiate between the stick and slip cases instead of using the conventional return

mapping procedure (e.g., Giannokopoulos 1989). In the MFC scheme, the contact virtual

work is fulfilled based on the total virtual gap vector g u defined as:

=

g u x nm x m

3.59

in which the gap vector is not resolved to its tangential and normal components. This

simplifies the formulation of contact kinematics because the variation of the normal vector is

not involved in the formulation, and the variation of the projection point is also zero for the

case of stick state, wherein the point of solution is fixed during deformation. The constraint

condition for displacement based on the total virtual gap is then formulated as:

64

ccu =

g d g d

u

3.60

=

d g u (d u nm d u m )

3.61

Note that, in Eq. 3.60, no distinction is made regarding the normal and tangential components

of the interface force.

The concept of the MFC and the states of sticking and slipping, together with their

corresponding constraints, are further detailed in the following section.

Sticking between two bodies implies that relative tangential movement between them is not

allowed. The tangential gap vector g uT is introduced to describe the relative movement

between the bodies by:

g ustT = ( x nm x m )

1

a

(a a)

3.62

st

st

The frictional force is evaluated with the imposition of the sticking constraint g=

g=

0.

uT

uT

In contrast, when sliding occurs, points on the contact surface cnm can move relative to the

other surface cm . Therefore, the relative velocity g uslT is not zero where g uslT denotes the

corresponding tangential slip. The relative velocity can be evaluated by the convective or LIE

derivative of g uslT , which physically provides the frame indifference time rate of change of

the tangential slip. The time derivation is carried out in the reference configuration so that

g uslT is first pulled back to the reference configuration where the tangential direction is fixed

=

g uslT =

g uT

g uslT

am

am

65

=

g uslT a m

3.63

Afterwards, g uslT is pushed back to the current configuration, and that the direction of

tangential movement is fixed during an incremental path change (i.e., dg uslT = a m d with

d = dt ). Note that the length and orientation of the sliding path are not known a priori and

depend on the loading and the adopted constitutive friction law. Consequently, the calculation

of the sliding path over the time interval [t0, t] is accomplished by integrating the relative

velocity as:

t

guslT = a m dt

3.64

t0

The frictional force tT = tTsl is related to normal effective stress by the Coulomb friction law,

which gives rise to the slip constraint as:

tTsl = t N

g uslT

g uslT

3.65

The negative sign in Eq. 3.65 is consistent with the physical understanding that the relative

sliding velocity is in the opposite direction to the friction force. As the calculation of the

frictional force depends on whether the contacting node is in a stick state or in a slip state, a

trial state must be assumed before any solution can be achieved. First, the stick condition is

considered the trial state, with 0 denoting the position of the first contact of a non-mortar

point with the mortar surface or its last position after sliding at time tn: 0 = n . The slip

function fs is then evaluated to check whether the sticking condition is satisfied and whether

the trial state is valid. If the trial state is not the case, it is changed to a sliding state and the

solution procedure continuous. The evaluation of the slip condition fs requires computing tT

and t N from a trial state as:

=

t tr =

u (x nm x m )

u gu

3.66

tTtr = (1 n n)t tr

3.67

t Ntr = t tr n tvN

3.68

66

where tvN = v g vN defined in Eq. 3.43 represents pore-water pressure across the interface (the

negative of the force required to enforce no relative flux across the interface). Using the

values of tTtr and t Ntr in Eq. 3.58 provides:

f str

=

tTtr t Ntr

3.69

A slip surface such as that shown in Figure 3.4 can be defined according to the slip function

(Eq. 3.69) in analogy to a yield surface in plasticity. If the friction force tT represents a point

within the cone ( f str < 0 ), the corresponding state is identified as stick, and no sliding will

occur. Conversely, a point on the surface of the cone ( f str = 0 ) indicates a sliding contact. The

point of origin of the friction cone changes as sliding occurs (i.e., MFC). Figure 3.5 illustrates

the concept of the MFC when two bodies stick together.

tT

tN

According to Figure 3.5, when contact is established at time t0, a stick state is assumed and

the position of the vector x0m (0 ) is fixed at 0 . This position is determined by fulfilling the

minimal distance between xnm and the mortar surface m. While the point xnm is in the stick

state, its position inside the friction cone may change relative to m at any time tn+1 due to

deformation, but the projection point is fixed to 0 or the last position after sliding at time

tn: 0 = n . As soon as xnm touches the boundary of the MFC, sliding occurs and the state of

contact is changed from stick to slide, as shown in Figure 3.6. The total stress t is then

transferred between the contact partners x nm and x m , which is in the direction of the total gap

vector gu (see Figure 3.6).

67

B nm

t (B nm)

t (B m)

Bm

T = t0

T = tn+1

Figure 3.5: Initial and current configuration of two contacting bodies in a stick case

It is worth noting that, for the frictionless contact presented in Section 3.2, the contact

formulation was expressed using the normal component of gap vector guN, which gave the

distance vector in the normal direction. The update of the projection point was also

accomplished by keeping the minimal distance between the contacting points (see

Figure 3.3). However, in the MFC scheme, as the entire load is transferred in the direction of

gu, a relative movement can take place in the orthogonal direction g u . Therefore, the

moving direction is obtained following the orthogonality constraint as:

t gu =0

3.70

The total contact stress vector t can be expressed based on its normal t N and tangential t T

components as:

t = t N + tT = t N n m + tT

a

am

3.71

By inserting the slip stress from Eq. 3.65 and including the sliding velocity (Eq. 3.63), the

total contact stress may be expressed as:

t N n m t N tvN

t sl =

t=

g uslT

g uslT

t

am

t N n m + sign(t N ) sign(tT )(1 vN ) m

=

tN a

68

3.72

a: T = t0

b: T = t

Figure 3.6: Frictional sliding and movement of with the moving cone: (a) initial

configuration; (b) current configuration

Comparing Eqs 3.72 and 3.70, the moving direction can be stated as:

g u = a m sign(t N ) sign(tT )n m a m

3.73

provides the direction of the normal and frictional forces. Figure 3.6(b) depicts the moving

direction g u , assuming sign

=

(t N ) sign

=

(tT ) 1 .

Updated cone

at time step tn is changed to a new position x nm

n+1 at

n

the next time step tn+1 in the direction of g u . Therefore, a new projection point on mortar

segment n +1 must be evaluated, as depicted in Figure 3.7. The new position should lie on the

friction cone boundary to satisfy:

69

0

3.74

=

f sn+1 u guTn+1 sign( guTn+1 ) ( u guNn+1 v g vNn+1 )sign( =

guNn+1 ) 0

3.75

where guTn +1 and guNn +1 are, respectively, the tangential and normal components of g un +1 as:

=

guTn+1 g un+1 ( n )

a mn+1 ( n )

a mn+1 ( n )

3.76

guN=

g un+1 n nm+1 ( n )

n+1

3.77

and g vNn+1 is the normal component of the relative Darcys velocity defined in Eq. 3.6. Note

that, in Eq. 3.75, the penalty parameter for the tangential direction T and normal N

direction are the same (i.e., =

=

u ). Eq. 3.75 can then be expressed as:

T

N

f sn+1 guTn+1 sign( guTn+1 ) ( guNn+1

=

v

g

)sign( =

g uN ) 0

u vN

n+1

n+1

3.78

Applying the Newton iteration to Eq. 3.78, a new position n +1 may be obtained. For the

Newton iteration, the derivative of Eq. 3.78 with respect to is required (see Appendix C.I),

which can be expressed as:

2

=

i , sign( guTi ) ( x nm x m (i ) ) aim, aim +

aim aim,

m

m

aim

sign( guNi )

m m m

v , ni ai

3.79

where = v u . The adopted iteration scheme is presented in Box 3.2. Note that the

directions sign( guT ) and sign( guN ) in Eq. 3.78 are fixed within a time step by taking the

n+1

n +1

70

Initialisation: i = 0, i = ni

LOOP over NEWTON iterations: i = 1,, convergence

Compute

(i ) = i

= guTi sign( guT0 ) ( guNi

v

g vNi )sign(=

guN ) 0

u

0

i

Compute

=

i , sign( guT0 ) (x nm x m (i )) aim, aim + sign( guN0 )

aim aim,

m

m

( guTi g vTi )ni ai , + ( g vNi guNi )

aim

m

m

m

v , ni ai

i +=

i

1

i

, set i = i +1

i ,

END LOOP

This section presents the linearisation of contact virtual works for frictional contact.

Stick case

As the total gap g u and total virtual gap d g u vectors are evaluated at a fixed point n in a

stick case, the linearisation of Eq. 3.60 can be expressed as:

d g ustick

d

n+1

n+1

c

71

3.80

The weak forms in Eqs 3.60 and 3.80 are discretised by considering the quadratic geometric

approximation of both non-mortar and mortar surfaces, as given in Eqs 3.17 and 3.18. Using

matrix notation, Eq. 3.80 can be written as:

u stick

c

=

c

nq

d U

q =1

( u Bu BTu n L q wq )U

3.81

which provides the tangent matrix for the stick case as:

nq

stick

u

= u Bu BTu L q wq

3.82

q =1

The corresponding residual vector of a mortar segment is evaluated using the constraint

formulation (Eq. 3.60) as:

nq

R ustick = u Bu G u L q wq

3.83

G u = BTu x

3.84

q =1

where:

Accordingly, virtual work performed by virtual displacement can be included in the global

equation system as:

{d U

d VrT

K ustick

d PT } 0

0

0 0 U

0 0 Vr =

{d UT

0 0 P

d VrT

R ustick

d PT } 0

3.85

Slip case

If the evaluation of the trial state indicates that sliding occurs, a new position n +1 is obtained

on the boundary of the moving cone (Box 3.2), this position is also fixed, and the variation of

the total gap vector is given by:

m

d=

g uslip d g uslip (=

n +1 ) d U nm

n +1 d U n +1 ( n +1 )

n+1

n+1

72

3.86

The contact virtual work (Eq. 3.60) can now be linearised for the actual deformation state as:

d g uslip

d + u g uslip

d g uslip

d

n+1

n+1

n+1

n+1

c

3.87

where:

g uslip

=

n+1

( U

nm

n +1

U mn +1 ) U mn +1,

d g uslip

=

d U nm+1,

n+1

3.88

slip

ccu=

u ( U nnm+1 U nm+1 ) (d U nnm+1 d U nm+1 ) d c

c

3.89

in which the increment of is also required. can be derived from the linearisation of the

product of Eq. 3.78 and the absolute value of the tangent vector as:

a

3.90

g un+1 n +1 sign( guTn+1 ) ( guNn+1 g vNn+1 )sign( guNn+1 ) a n +1 =

0

a n +1

sign( g ) ( x nm x m ) x m a 2

uT

,

a

v m n a sign( g )

r ,

uN

3.91

nm

m

( U U ) sign( guT )a sign( guN )n a

= + ( x nm x m ) sign( g uT )U ,m + sign( g uN ) ( x nm x m ) ( v nm

r v r )

n

m

m

nm

m

U , a + sign( guN ) ( v r v r ) n a

U , n

The above equations can be cast in matrix form using the following notations:

73

G=

BTu x =

a BTu , x =

n

u

e3 a

e3 a

c a=

BTu , x

=

,

G=

BTv Vr A=

BTv , Vr

v

v

3.92

( sign( guT )aT sign( guN )nT a ) BTu +

1

sign( guN ) T

= G Tu sign( guT ) + (G Tu G Tv )

a n nT a ) BTu ,

(

R

a

T

T

+ sign( guN )n a B v V

3.93

where:

sign( guN ) T

2

T

(G u G Tv ) ( anT naT ) c

sign( guT ) G u c a +

a

R=

T

A v n a sign( guN )

3.94

Note that, according to the update of in Box 3.2, the sign functions are constant within a

time step.

=

ccu slip u d UT Bu BTu Ud u d UT Bu , G u + Bu a d

c

3.95

A Gauss-point role is used to evaluate the integrations over every non-mortar segment. By

replacing from Eq. 3.93 in Eq. 3.95, the linearisation of the contact virtual work may be

written in a compact form as:

=

ccu d UT [ K u U + K uv V ]

where:

74

3.96

n +1

n +1

nq

T

1

G sign( guTn+1 ) +

=

K u u Bu BTu + Bu , G u + Bu a u

L q wq

T

R

q =1

B

T

g

sign(

)

u ,

uN n+1

T

3.97

anT naT )

(

(G u G v )

nq

1

Bu , G u + Bu a sign( guNn+1 )nT BTv a L q wq

q =1 R

=

K uv v

Finally, the virtual work resulting from the displacement can be rewritten in matrix form as:

{d U

d VrT

K u

T

dP } 0

0

0 K uv U

0 0 Vr =

{d UT

0 0 P

d VrT

R uslip

d PT } 0

3.98

nq

slip

u

= u Bu G u L q wq

3.99

q =1

For frictionless contact, Darcy velocity contributions (Section 3.2.2.3) were formulated based

on a conditional equation (Eq. 3.13) for the projection point , which implies that, at any

times, the distance vector (Eq. 3.2) must be orthogonal to the tangent vector a m . In the MFC

scheme, the total moving direction and variation of the projection point are determined by the

constitutive equation for friction, which is restricted to Coulombs friction law. This resulted

in Eq. 3.93 for , which is used here for the linearisation of the contact virtual work arising

from the Darcy velocity (Eq. 3.46).

Slip case

Using Eq. 3.93, the linearisation of the contact virtual work arising from the Darcy velocity

(Eq. 3.46) may be expressed as (see Appendix C.III):

slip

=

U

ccvr slip d VrT K vslip Vr + K vu

75

3.100

where:

K vslip

B v n nT BTv , Vr + 2 aT BTv Vr nT c +

nq

a

v

T T

=

v B v nn B v +

K L q wq

q =1

1

T

g vN B v , n + 2 B v an c

1

B v n nT BTv , Vr + 2 aT BTv Vr nT c +

nq

1

T

v g vN B v , n +

L q wq

B v an c

q =1

T T

T T

T T

a 2 B v na B v Vr n B v , + g vN B v an B v ,

slip

K vu

3.101

3.102

n+1

n+1

sign(

)

uN n+1

K u = (G Tu G Tv )

anT naT ) + T

(

R

a

Bu ,

G T sign( g )

uTn+1

K v =

R

3.103

3.104

n+1

Finally, the virtual work performed by the Darcy velocity may be rewritten in matrix form as:

{d U

dV

T

r

0

d P } K vuslip

0

T

0

0 U

slip

0 Vr

K

=

v

0

0 P

{d U

dV

T

r

76

d P } R v

0

T

3.105

Stick case

In a stick case, the position of the contact at each Gauss point is fixed with respect to the

surface coordinate; thus, the coordinate does not change. As a result, the second and third

terms in Eq. 3.33 can be neglected, leading to:

ccp stick= d PT K stick

p P

3.106

where:

nq

K stick

= p B p BTp L q wq

p

3.107

q =1

Accordingly, the virtual work performed by the pore-fluid pressure can be rewritten in matrix

form as:

{d U

d VrT

0 0 0 U

d PT } 0 0 0 Vr =

{d UT

0 0 K p P

d VrT

d PT } 0

R p

3.108

Slip case

When a slip case is studied, should not be neglected. As the concept of a moving cone

boundary is used, must be evaluated by Eq. 3.93. Otherwise, can be found because

the minimal distance must be kept at any time (Eq. 3.13), as it was used for the case of the

frictionless interface in Section 3.2.2.2. Nonetheless, replacing in Eq. 3.33 may provide:

slip

slip

slip

ccp=

d PT K slip

p P + K pu U + K pv Vr

where:

77

3.109

nq

T

K slip

p = p B p B p L q wq

3.110

q =1

n +1

n +1

nq

sign( guNn+1 )

1

B p , g p + B p BTp , P( ) (G Tu G Tv )

=

K slip

anT naT ) + T L q wq 3.111

(

pu

p

a

q =1 R

Bu ,

T

G u sign( guTn+1 )

nq

=

p

K slip

pv

q =1

R

n+1

3.112

Accordingly, the virtual work performed by the pore-fluid pressure can be included in the

global equation system during the Newton iteration as:

{d U

dV

T

r

dP } 0

K pu

0

0

K pv

0 U

0 Vr =

{d UT

K p P

dV

T

r

d P } 0

R

p

T

3.113

U-P formulation provides an approximate solution to the dynamic behaviour of fluidsaturated porous media (see Section 2.4). However, this scheme is a prevalent approach in

geomechanics because of its relatively simple formulation and implementation compared to

the U-P-V method. Darcys velocity does not appear explicitly in the U-P formulation (Eq.

2.59), so that pore pressures at the points of contact can be used to derive the corresponding

normal effective stresses. Meanwhile, a constraint is applied to fulfil the balance condition of

the pore pressure at the contact interface where necessary. It is notable that the presented

contact formulation in Section 3.2.2.2 also holds for the frictionless contact of the U-P

scheme. For the stick case of the frictional contact, the formulations presented in

Sections 3.3.2.1 and 3.3.2.3 can also be used for the U-P scheme. However, for the slip case,

contact contributions arising from displacement and pore-fluid pressure must be

reformulated. The treatment of frictional contact mechanics for solidfluid mixture based on

the U-P scheme is presented in this section.

78

Slip case

When sliding occurs, new projection points on mortar segment n +1 should be calculated. The

new position lies on the boundary of the friction cone to satisfy Eq. 3.74, which can be

expressed based on the gap functions as:

=

f sn+1 guTn+1 sign( guTn+1 ) ( guNn+1

pnm+1 )sign( =

guNn+1 ) 0

3.114

where pnm+1 denotes the corresponding pore pressure at the point of contact. Applying the

Newton iteration to Eq. 3.114, a new position n +1 may be obtained. For the Newton iteration,

the derivative of Eq. 3.114 with respect to is required, which can be expressed as:

2

=

i , sign( guTi ) ( x nm x m (i ) ) aim, aim

1 m

aim aim,

m

m

+

guTi ni ai , + p (i ) guNi

m

u

ai

+ sign( guNi )

1

p,m aim

3.115

The increment of can be calculated from the linearisation of the product of Eq. 3.114 and

the absolute value of the tangent vector as:

a

1

3.116

g un+1 n +1 sign( guTn+1 ) ( guNn+1 pnm+1 )sign( guNn+1 ) a n +1 =

0

a n +1

u

79

Box 3-3: Newton scheme for the update of n for the U-P scheme

Initialisation: i = 0, i = ni

LOOP over NEWTON iterations: i = 1,, convergence

Compute

(i ) = i

= guTi sign( guT0 ) ( guNi

1 m

0

pi )sign( g=

uN )

u

0

i

Compute

2

=

i , sign( guT0 ) ( x nm x m (i ) ) aim, aim

1 m

aim aim,

m

m

(

)

n

a

g

p

g

+

uTi i i ,

i

uNi

m

u

ai

+ sign( guN0 )

1

p,m aim

i +=

i

1

i

, set i = i +1

i ,

END LOOP

sign( g ) ( x nm x m ) x m a 2

,

uTn+1

m

+ sign( g

uN n+1 ) p, a

n +1

n +1

a

n

U ,m a

= + ( x nm x m ) sign( guTn+1 )U ,m + sign( guNn+1 ) U ,m n

n +1

n +1

u

a

u

80

3.117

1

G Tu sign( guT )1[22] +

a

T U

1 +

Bu ,

=

R + 1 sign( g ) a PT B m

N

u

p

a

u

+ 1 sign( g ) ( B m )T P

uN

p

u

3.118

sign( guN ) T

1 T m

2

T

T

T

sign( guT ) G u c a +

G u ( an na ) + P B p a c

a

u

R=

+ 1 sign( g )PT B a

uN

p ,

3.119

0

0

m

Bp =

q

(1 q )

3.120

where:

and:

The linearisation of the virtual work given in Eq. 3.89 may now be expressed as:

ccu slip d UT K u U + K up P

=

3.121

where:

T

Bu , G u + Bu a

+

B

B

u

u

nq

L q wq 3.122

K u = u T

sign( guNn+1 )

T

T

q =1 G u sign( g uT )1[22] +

( an na )

n+1

a

BT

u ,

1

1 T T m

+ sign( guNn+1 ) a P B p

a

u

81

nq

R sign( g

=

K up

q =1

) Bu , G u + Bu a ( B mp ) a L q wq

T

uN n+1

3.123

Accordingly, the virtual work performed by the displacement can be included in the global

equation system during the Newton iteration as:

{d U

K u

0

d PT }

K up U

T

= {d U

0 P

R uslip

d PT }

3.124

The incremental form of the constraint condition in Eq. 3.33 may be written as:

=

ccp slip p d PT B p BTp Pd p d PT ( B p , g p + B p BTp , P ( ) ) d

c

3.125

Replacing (Eq. 3.118) in the above equation and expressing the result in a compact form

yields:

ccp slip d PT K p P + K pu U

=

3.126

where:

nq

T

Kp =

p B p BTp +

sign( guNn+1 ) B p , g p + B p BTp , P( ) ( B mp ) L q wq

R u

q =1

T

T

T sign( guN ) ( an na )

nq

1

G u a

K pu p B p , g p + B p B p , P ( )

L w

=

BT q q

q =1 R

+sign( guT )1[22]

u ,

1

a

+ sign( guN ) PT B mp

82

3.127

3.128

Thus, the virtual work performed by the pore-fluid pressure can be included in the global

equation system during the Newton iteration as:

{d U

0

K pu

d PT }

0 U

T

= {d U

K p P

0

slip

R p

d PT }

3.129

is obtained by:

p

nq

R slip

p = p B p ( n +1 ) g p ( n +1 ) L q wq

3.130

q =1

3.5 Summary

This chapter detailed the treatment of the contact mechanics of two-phase saturated porous

media. The development of the contact algorithm was described, and various contributions

arising from the contact algorithm to the tangent stiffness matrix and residual vector were

presented. The formulation of the contact kinematics and constraints adopted in this chapter

was based on the so-called mortar segment-to-segment approach. The formulation was

derived for two different forms of dynamic coupled equations: the U-P-V and U-P schemes.

The frictionless contact was first presented and then extended to the frictional contact using

the concept of the moving friction cone.

83

4.1 Introduction

The numerical scheme developed in the course of this thesis has been implemented into

SNACa FE code developed by the geomechanics group at the University of Newcastle,

Australia. Some numerical examples are presented in this chapter to evaluate the performance

of the computational scheme and to verify its implementation. The first set of examples

involves the validation of the mixed dynamic consolidation formulation presented in Chapter

2, as well as the proposed contact algorithm, described in Chapter 3. A dynamic coupled

analysis of pile installation into a saturated soil layer is then studied. In all analyses, a relative

error tolerance of 10-4 for the unbalanced forces and solid displacement was utilised during

the NewtonRaphson iterations. The value of the spectral radius at infinity was assumed

to be 0.818 for the analyses presented in Sections 4.2 and 4.3, and = 0.50 for all other

simulations in this thesis.

Incompressible Constituents

de Boer et al. (1993) presented an exact solution for the 1D problem of contemporaneous

wave propagation under a time-dependent load. The solution holds for a saturated mixture

with two incompressible constituents in the small strain regime, neglecting the variations in

volume fractions during the deformation process. As a result of the incompressibility

constraints, only one independent dilatational wave in the two-phase mixture propagates.

To evaluate the dynamic coupled algorithm presented in Chapter 2, the wave propagation

responses of a porous layer of initial thickness h0 = 10 m subjected to a dynamic load were

studied, and the results were compared with the corresponding analytical solution (de Boer et

al. 1993). Although the problem was 1D, plane strain conditions were assumed here, but the

horizontal solid displacements and normal fluid motion were restrained on both sides of the

plane strain FE mesh (see Figure 4.1). The upper boundary was drained and subjected to a

time-dependent load w = f(t), where f(t) was chosen to be a sine function, as depicted in

84

Figure 4.1. A rigid impermeable boundary was adopted at the bottom of the mesh, preventing

solid displacements and vertical fluid motion.

w = f(t)

Time (s)

Permeable

0.08

0.16

0.24

0.32

0.4

w (kPa)

6

4

2

0

h0

Rigid/impermeable

The assumed soil properties are listed in Table 4.1. The predicted displacement response of

the solid as a function of depth, measured from the free surface, is depicted in Figure 4.2 at

two different times. The results indicated agreement between the numerical and analytical

results.

Parameter

Value

Porosity

n = 0.33

s = 1.34 Mg/m3

f = 0.33 Mg/m3

Permeability of soil

k = 10-2 m/s

= 0.2

= 30 MPa

Figure 4.3 plots the pore-water pressure response versus time, indicating negative values

(suction) in the vicinity of the loading surface. As identified by de Boer et al. (1993), this

result was due to the recovery of the elastic skeleton matrix close to the surface during the

85

cyclic sinusoidal loading, where the pore water did not squeeze out but was absorbed into the

pores and accompanied by fluid suction.

0.04

0.03

0.03

t=0.135s

0.02

t=0.155s

0.02

0.01

0.01

0.00

0.0

1.0

2.0

3.0

4.0

5.0

6.0

Depth (m)

Figure 4.4 also shows the generation of fluid suction in the vicinity of the free surface,

depicting the normal component of the Darcy velocity at times 0.08s and 0.4s.

Z=0.4m

Z=1.0m

Z=6.0m

5

4

3

2

1

0

-1

-2

-3

0.00

0.04

0.08

0.12

0.16

0.20

0.24

0.28

0.32

0.36

Time (s)

86

0.40

According to Figure 4.4, the direction of the Darcy velocity, predicted by the numerical

solution, was downwards for a zone within 1 m of the free surface, whereas it changed to an

upwards direction at deeper depths.

-0.60

0

-0.40

-0.20

0.00

0.20

0.40

Depth (m)

t=0.08s

t=0.40s

To assess the large deformation capability of the code, the soil column was subjected to a

uniformly distributed step load q at the free surface according to the load type illustrated in

Figure 4.5. Seven load levels of 0.1E, 0.2E, 0.3E, 0.4E, 0.5E, 0.6E and 0.8E were applied on

the column, and the predicted results were compared with those reported by Meroi et al.

(1995). The assumed material parameters were E = 1GPa, = 0.0, n = 0.3 and k = 0.01 m/s.

Figure 4.5 plots the applied pressure normalised by E versus the total consolidation

settlements (at large time) normalised by the column depth (h0). As shown, the results

obtained by SNAC agreed with those reported by Meroi et al. (1995).

87

1

Meroi(1995)- Large strain

Analitical solution

0.8

SNAC- Small strain

SNAC- Large strain

q/E

0.6

0.4

0.2

t

0.1 Sec

0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

s/h0

Compressible Pore Fluid

In the first example, the assumption of fluid incompressibility limited the analysis to

modelling only one type of independent dilatational wave. Generally, dilatational waves can

be decoupled into two types: fast and slow compression waves. The second wave is also

known as Biots slow wave. The common feature of the fast compression wave is that the

porous skeleton and the interstitial fluid move essentially in phase, so that the relative motion

of the fluid within the pore channels is relatively unimportant. In contrast, the porous skeleton

and the interstitial fluid move essentially out of phase in the slow compression wave.

Consequently, the relative motion of the fluid within the pore channels becomes very

important. However, the slow compression wave is usually very weak due to high attenuation

caused by the viscous drag between the fluid and the solid skeleton of the porous medium.

The propagation of these acoustic waves essentially depends on the frequency of excitation,

the hydraulic permeability of the porous medium and the mechanical properties of the

constituent materials.

88

In this section, the propagation of plane waves through a porous medium and the associated

coupled consolidation was studied for the case of a deep layer of saturated soil subjected to a

step loading applied at the surface. In particular, a 10 m deep soil column with a linear elastic

skeleton saturated with a compressible pore fluid was analysed for the case of a sudden

increase in pore-fluid pressure of magnitude p0 applied at the surface. The mesh and

boundary conditions for the analysis were similar to the previous example, except that the

upper boundary was impermeable (see Figure 4.1). The material parameters are listed in

Table 4.2.

Parameter

Value

Porosity

n = 0.4

s = 1.59 Mg/m3

f = 0.40 Mg/m3

Permeability of soil

k = 10-3 m/sec

Compressibility of soil

mv = 2x10-10 m2/N

Verruijt (2010) and Carter et al. (2015) proposed closed-form solutions for this problem, with

the latter using Laplace transforms of the equations governing the dynamic behaviour of a

porous medium. Carter et al.s (2015) analytical solution (see appendix A.III) was employed

here to evaluate the performance of the dynamic consolidation algorithm when modelling the

wave propagation characteristics of porous saturated soil. In the numerical analysis, the

increase in pore-water pressure at the surface p0 was applied at a uniform rate over a period

of 10 s and thereafter held constant with time. Apart from the difference in the means of

applying the loadingthat is, an instantaneous increase in pore-water pressure in the case

of the analytical solution and a linear increase of the surface pore pressure over a very small

period (10 s) for the numerical solutionall other problem parameters were identical in the

case of the analytical and numerical treatments.

Figure 4.6 depicts the pore-water pressure response as a function of time at a depth of 0.2 m.

There was reasonably close agreement between the numerical solution (labelled the U-P-V

89

analysis) and the analytical solution, although the numerical results exhibited some

oscillations around the exact solution. Better agreement between the numerical and analytical

solutions may be possible if different numerical integration and numerical damping schemes

and perhaps a different numerical approximation of the instantaneous boundary loading are

adopted, but this issue was not explored in any detail. But, importantly, it can be clearly

observed that both the numerical and analytical solutions predict that two distinct waves of

dynamic pore pressure are developed and pass through the given location. The fast

dilatational wave arrived at a time of approximately 90 s, whereas the slower wave arrived

after a delay of around 85 s, having about half of the velocity of the first wave. After the

initial shock due to the arrival of the dynamic waves at a depth of 0.2 m, the pore pressure

p / p0

gradually increased.

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0E+00

U-P-V analysis

1.0E-04

2.0E-04

3.0E-04

Time (s)

4.0E-04

5.0E-04

Figure 4.7 depicts the pore-water pressure response at much larger times, during which the

pore pressure at this depth increased and eventually approached the value p0 applied at the

soil surface. The mechanism causing the last stages of this increase was consolidation, as the

pore fluid flowed through the solid skeleton of the soil.

This example demonstrates the effectiveness of the numerical algorithm when analysing a

problem that involves a significant transient (short-term) dynamic response, dominated by

inertia effects, followed by the increasing importance of the consolidation phenomenon in the

porous medium at intermediate and large times.

90

1.0

p / p0

0.8

0.6

0.4

0.2

0.0

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10

Time (s)

This example considers the consolidation behaviour of a smooth, flexible strip footing resting

on an elasto-plastic layer. The layer was modelled by a rounded Mohr Coulomb (MC) yield

criterion (Abbo and Sloan 1995) with an associated flow rule. Although the physical

implications and limitations of such a flow rule have been highlighted for consolidation

simulations (Small 1977), the analyses presented here were conducted to enable a direct

comparison of the results with those of Manoharan and Dasgupta (1995). Figure 4.8 depicts

the FE model of the footing, containing 280 quadratic triangular elements and 595 nodal

points.

a

permeable

smooth / impermeable

smooth / impermeable

rough / impermeable

16a

91

8a

The assumed material parameters are listed in Table 4.3. The load q was applied as a uniform

prescribed pressure on the footing area and imposed over an initial period of Tv =

0.01 ,

where Tv denotes the dimensionless time factor defined by:

Tv =

cv t

a2

4.1

where t represents actual time and cv denotes the 2D coefficient of consolidation given by:

cv =

kE

2 w (1 + )(1 2 )

4.2

Parameter

Value

E = 2 MPa

= 0.3

c = 10 kN/m2

= 20

= 20

sat = 20 kN/m3

Permeability of soil

k = 10-5 m/day

time step to avoid inertia effects during the loading process so that the total loads were

applied over the actual time increment of t =10 days, resulting in a time factor increment of

Tv =

0.01 by considering cv = 0.001a . The subsequent consolidation in each case was then

2

modelled up to a total time factor of Tv = 1000 . Figure 4.9 depicts the predicted consolidation

settlements at the centre of the footing. As shown, the results of the analysis agreed with

those presented previously (Manoharan and Dasgupta 1995).

92

0

0.05

0.1

Settlement / a

0.15

0.2

0.25

0.3

0.35

0.4

q/c'=5

q/c'=10

q/c'=15

0.45

0.5

0.0001 0.001

0.01

0.1

10

100

1000

Tv

Figure 4.9: Settlement versus time factor for the elasto-plastic strip footing

Figure 4.10 depicts the development of pore-fluid pressure at the centre of the footing, as

well as its dissipation with time. The pore pressure reached its maximum value at the end of

the loading process and had entirely dissipated when Tv = 1.0 . Figure 4.10 shows that the

results obtained by SNAC (labelled as the U-P-V analysis) and those reported by Manoharan

and Dasgupta (1995) were in agreement. Therefore, the capability of the dynamic

consolidation algorithm was verified for consolidation processes in which strong coupling

occurred between the three field variables of solid displacement, pore-fluid pressure and fluid

velocity (U-P-V). Some vector plots of Darcy velocity are shown in Figure 4.11, with

corresponding excess pore-pressure contours at different values of the time factors.

According to Figure 4.11, higher pore pressure at the edges tends to cause inward flow

initially, which can delay the dissipation of excess pore pressure at the centre-line.

93

0.9

0.8

U-P-V analysis

0.7

p / q0

0.6

0.5

0.4

0.3

0.2

0.1

0

-0.1

0.0001

0.001

0.01

0.1

10

100

Tv

(a) Tv = 0.05

(b) Tv = 0.09

(c) Tv = 0.22

Figure 4.11: Excess pore pressure contours and Darcy velocity vector maps

94

This section is also concerned with the performance of the numerical algorithms when

dealing with geometrical nonlinearity. Drained and undrained loading conditions represent

the extremes of consolidation behaviour and can be used to validate FE models. The

undrained behaviour of a strip footing under dynamic loading and large deformations was

previously studied by Sabetamal et al. (2014) to validate a dynamic consolidation algorithm

formulated in a displacementpressure (U-P) form. The analysis is repeated here utilising the

displacementpressurevelocity (U-P-V) algorithm to illustrate its performance in large

deformation regimes. Figure 4.12 shows the boundary conditions and the mesh for the righthand half of the problem. The soil domain consisted of 872 plane strain elements and 1,817

nodal points. The footing was modelled by six elements, and its behaviour was assumed to be

elastic. The results of the undrained analysis conducted in terms of total stressconsidering

nodal displacements as the only degrees-of-freedom and assuming a Tresca material model of

the soilwere compared with the results of a dynamic coupled consolidation analysis (U-PV) using a MC material model.

0.5B

Impermeable

Displacement analysis:

4B

Coupled Analysis:

6B

The results of these two analyses should coincide with each other provided the soil medium is

initially unstressed and the soil properties satisfy (Small 1977):

Eu =

3E

2(1 + )

95

4.3

cu 2 N

=

c 1 + N

4.4

where subscript u and superscript ' denote undrained and drained quantities, respectively; c

represents the drained cohesion of the soil skeleton; ' is its friction angle; and N is obtained

according to:

N =

1 + sin

1 sin

4.5

Figure 4.12 shows the values of the material parameters used in this example. Note that '

and in Figure 4.12 are the dilation angle of the soil and its mass density, respectively. A

non-zero value of the mass density was considered here to account for inertia effects only,

while the initial geostatic stresses were assumed to be zero. To avoid further complexities,

any increase in the shear strength of the soil due to rate effects was specifically ignored.

In the first analysis, it was assumed that the soil behaved as a Tresca material model

deforming under undrained conditions, and only the displacement degrees-of-freedom were

considered in the analysis. To simulate a rough footing, the nodal points on the footing were

fixed in the horizontal direction. These nodes were also tied to each other in the vertical

direction to represent a rigid footing. A large deformation analysis was conducted with a

uniform pressure applied to the footing at a uniform rate of 30cu per second for a period of

1 sec. Figure 4.13 plots the settlement of the footing, normalised by its width, versus the

applied pressure, normalised by cu,. A clear collapse load, such as Prandtls undrained

collapse pressure of q = 5.14 su , applicable in a small strain analysis, was not identifiable in

this analysis. The higher soil stiffness and footing resistance predicted in this analysis

resulted from inertia effects and large deformations, noting again that material rate effects

were not considered here.

In the second analysis, a non-associated MC material model was used to predict the soil

response by conducting a coupled consolidation analysis, while the mortar contact algorithm

described in this study was employed to model the interface between the elastic footing and

the elasto-plastic soil.

The relative stiffness of the footing and the soil can be expressed through factor K given by

(Brown 1969):

96

K=

E f (1 2f )h3

4.6

Eb3

where E f , f , h and b are Youngs modulus of the footing, Poissons ratio, thickness and

half width, respectively, and E is Youngs modulus of soil. To model a rigid footing, K was

assumed to be 1,000. To simulate a rough interface, the friction coefficient between the soil

and footing was 1.0. This relatively high value led to a purely stick state at the contact

surface, and it prevented the soil from moving horizontally immediately underneath the

footing. This loading condition was the same as that assumed in the previous analysis. The

soil response was predicted using three different values of dilation angles: 0, 2 and 5.

Figure 4.13 plots the normalised settlement of the footing versus the normalised pressure for

each dilation angle. As shown, the soil response predicted by the undrained analysis was

essentially identical to the results obtained by the coupled analysis, provided the dilation

angle was zero. For non-zero values of the dilation angle, an increase in the undrained

strength of the soil was observed because, based on the MC material model, the soil tended to

dilate when shearing occurs, but the volume change was restricted in the model by the

combination of a relatively low value of permeability and the relatively rapid loading. Small

(1977) also observed that a consolidation analysis with a non-zero dilation angle in a MC

stress-strain model of the soil skeleton may significantly overestimate the undrained shear

strength of the soil. The soil responses predicted by the U-P-V algorithm were also compared

with U-P predictions, showing excellent matches between the results.

Applied pressure / cu

32

24

Undrained total stress analysis

Dynamic U-P-V analysis: '=0

Dynamic U-P-V analysis: '=2

16

Dynamic U-P analyses

Static collapse load

0

0

0.4

0.8

1.2

1.6

Settlement / B

97

2.4

It is worth noting that the large deformation results presented in Figure 4.13 were obtained

using the ALE method. The UL method could not simulate the dynamic response under rapid

loading due to a severe mesh distortion, leading to termination of the analysis at a settlement

of ~0.3B. Figure 4.14 shows one of the deformed meshes at the end of the analysis. It was

evident that no sliding between the soil and footing occurred during settlement; that is, a pure

stick condition at the contact surface was satisfied. It was also observed that for this

particular problem, three Gauss points on every non-mortar segment (soil surface) were

enough to transfer contact tractions properly.

The aim of this section is to further assess the mortar contact algorithm developed in this

study. A relatively simple way to verify that the contact algorithm functioned as expected and

passed the patch test was to model the unconfined compression of a saturated porous layer.

Two rectangular elastic layers of porous media were compressed between two smooth and

rigid plates, and the interface between the two layers was modelled using the mortar method.

Due to symmetry about the y-z plane, only half of the geometry was modelled (see model (i)

in Figure 4.15(a)). To evaluate the patch test, a different number of elements was used in the

top and bottom layers of model (i), providing non-conforming meshes and guaranteeing that

nodes on opposing contact surfaces did not occupy the same location.

98

y

x

q

Contact interface

Free draining

a

(a)

Free draining

a

(b)

Figure 4.15: Unconfined compression models: (a) model (i), two elastic layers with a

contacting interface; (b) model (ii), equivalent case using a single elastic layer with no

contact interface

The average pressure q was then applied on the top surface in a ramp-and-hold profile in

which the total load was ramped over a time interval equivalent to the dimensionless time

factor Tv = 0.001. Tv is defined here by:

Tv =

cv t

3a 2

4.7

where a is the half-width of the layer and cv represents the 1D coefficient of consolidation

given by:

cv =

k E (1 )

w (1 + )(1 2 )

4.8

in which E is the drained Youngs modulus of the soil skeleton of each layer, and denotes

their drained Poissons ratio. After the total pressure was applied, consolidation was

permitted and monitored at different dimensionless time intervals. In an alternative analysis,

an equivalent model using a single layer with the same overall dimensions, and subjected to

the same external boundary conditions, was considered (see model (ii) in Figure 4.15(b)).

99

Figure 4.16 depicts a plot of the normalised nodal pore-water pressures across the entire

contact interface for model (i) for three different values of the dimensionless time factor. It

also depicts the pore-fluid pressures predicted by the alternative analysis, model (ii) for nodes

coinciding with the contact interface of the model (i). Figure 4.16 indicates agreement

between the two analyses, as well as the accuracy of the contact algorithm used to solve the

unconfined compression problem.

0.6

0.5

P/q

0.4

0.3

Equivalent analysis - no contact

Top contact nodes - Tv=0.001

Top contact nodes - Tv=0.01

Top contact nodes - Tv=0.1

Bottom contact nodes

0.2

0.1

0

0

0.2

0.4

0.6

0.8

x/a

Figure 4.16: Pore pressure at the interface of two layers normalised by applied pressure

Several methods have been proposed for analysing the quasi-static penetration of objects into

soil layers. A simple method involves using the classical theory of bearing capacity, which

considers penetration the continuous failure of a rigid plastic material and calculates soil

resistance using limit-equilibrium or slip-line analysis. Cavity expansion theory provides

another approximate solution for penetration problems and generally provides more realistic

results than bearing capacity theory because it takes into account the elasto-plastic behaviour

of the soil throughout the penetration. However, this approach ignores the history of soil

straining and the particular geometry of the penetrometer. For problems of deep steady

penetration, Baligh (1985) suggested the strain path method, which takes into account the

strain path and deformation history during penetration. However, applying this method to

frictional soil is a complicated task; therefore, it has mostly been used for undrained soils.

100

Teh and Houlsby (1991) obtained more realistic results by coupling the strain path approach

with a large deformation FEM to eliminate the violation of equilibrium occurring in the strain

path method. A detailed review of the various approaches adopted in cone penetration

analysis can be found in Yu and Mitchell (1998).

The FEM has been utilised by researchers to overcome the limitations of the methods

previously mentioned. The first attempts to use the FEM to predict collapse loads were based

on small strain theory (e.g., Sloan and Randolph 1982; de Borst 1982; Griffiths 1982).

Subsequently, large deformation theory was incorporated to account for geometry changes

during penetration and the effect of soil stiffness on its resistance (e.g., Budhu and Wu 1992;

Kiousis et al. 1988; Van den Berg 1991; Voyiadjis et al. 1997; Yu et al. 2000;

Liyanapathirana 2009). However, most of these analyses assume an existing pre-bored

narrow cone hole prior to the advancement of the penetrometer in order to avoid numerical

difficulties. Moreover, the interaction between the soil and the penetrometer is modelled

using interface elements, which are only suitable for predefined interfaces with small

interfacial deformations. Sheng et al. (2007) developed a contact algorithm to model the

interface between soil and structure and solved a few penetration problems.

Unlike static cone penetration tests, which have been studied using numerical and analytical

methods, dynamic penetration problems are still a challenging field of research in

geomechanics. Carter et al. (2010) used the FEM, based on the ALE strategy, for the

dynamic analysis of miniature free-falling penetrometers. Abelev et al. (2009) used the

commercial software Abaqus to model the dynamic penetration of a free-falling penetrometer

utilising the ALE scheme and the Von Mises soil model. However, based on a displacement

formulation, such analyses can only predict the total stresses developed in the soil; as a result,

they fail to estimate the component excess pore-water pressures and effective stresses. To

investigate the process of pore-pressure generation, its subsequent dissipation and the

effective stress in the soil, a fully dynamic coupled analysis can be facilitated by combining

deformations and pore-fluid pressures as basic unknowns.

The dynamic coupled analysis of a pile rapidly penetrating into a saturated soil layer is

studied in this section. In addition to providing the response of soil and pore-pressure

generation resulting from the installation of piles, some numerical insights are given into the

implications of using a NTS contact algorithm when analysing dynamic coupled problems.

101

Insights are also given into the effects of the soil model, mesh size and interface friction on

the soil response.

A rigid pile was installed into a layer of soil represented by the non-associated MC and

Modified Cam Clay (MCC) material models. To simulate relatively rapid installation, the pile

was pushed into the soil by applying a prescribed displacement of 8.75D in 1 s, where D

represents the pile diameter. Figure 4.17 shows the geometry, boundary conditions and FE

meshes. Two FE meshes, including a dense mesh with 9,517 nodes and 4,675 elements and a

fine mesh with 2,815 nodes and 1,353 elements, were used in this example. Figure 4.17(a)

and Figure 4.17(b) show the dense and fine meshes, respectively. The radii of the soil

elements underneath the pile were 0.25D and 0.125D in the fine and dense meshes,

respectively. Advantage was taken of the axial symmetry of the problem.

4.7.1

In the first series of analyses, it was assumed that the soil to behave as an ideal nonassociated MC material. The ratio between the Youngs modulus of the soil and its drained

cohesion E/c was 100, drained cohesion c' = 10 kPa, unit weight = 20 kN/m3, Poissons

ratio = 0.3, friction angle = 30, dilation angle =0, 2, 5,10 and coefficient of

permeability k, is 10-6 m/s. The contact interface between the pile and soil can be modelled as

smooth or rough. In the frictional contact formulation presented in Chapter 3, the relative

tangential displacement at the interface was divided into a slip and a stick, and the classical

Coulomb friction criterion was used to estimate the frictional forces due to normal contact

forces acting on the interface. This type of nonlinearity usually increases the number of

iterations required to achieve equilibrium in each increment, hence increasing the

computational time. In some cases, the frictional forces cause severe distortion of the contact

elements. Nonetheless, the friction between the pile and soil was ignored in these analyses

(see Section 4.7.4 for problems with frictional interfaces).

The initial stresses in the soil were generated assuming that the water table was at the ground

surface, the coefficient of earth pressure at rest was K0 = 0.43 and no surcharge was applied

at the soil surface.

102

Axis of symmetry

15.0D

Axis of symmetry

15.0D

0.5D

0.5D

9.0D

9.0D

(a)

(b)

Figure 4.17: FE meshes and boundary conditions: (a) dense mesh; (b) fine mesh

Figure 4.18 shows the deformed shapes of the dense mesh at times 0.05, 0.5 and 1.0 s,

representing the successful analysis of the penetration of the pile from the ground surface to

the desired depth of 8.75D. Figure 4.19 plots the soil resistance versus the penetration of the

pile, normalised by its diameter. Note that the two right-hand segments of the pile, which

came into contact with the soil, were considered the master surface in the NTS contact

formulation. Two numerical contact modelsa non-smooth discretisation and a smooth

discretisationwere used to represent the master surface. In the non-smooth discretisation,

the two linear segments of the pile were treated as straight lines, and their intersection point

indicated a sharp corner. This type of discretisation usually produces oscillation in the

predicted soil response, which is evident in the curves labelled non-smooth cone and nonsmooth NTS in Figures 4.19(a) and 4.19(b), respectively.

103

(a)

(b)

(c)

Figure 4.18: Deformed meshes at different times ( = 10 ): (a) t = 0.05 s; (b) t = 0.5 s;

(c) t = 1.0 s

One of the major reasons for this behaviour is that the direction of the normal vector to the

master surface at a sharp tip is not unique, as the derivatives of the linear functions describing

the linear segments at the tip are not identical. To reduce the intensity of oscillation and avoid

sharp transitions, either a smooth discretisation of the master segment can be adopted for the

NTS scheme (e.g., Wriggers 2006), or a mortar contact algorithm with a quadratic

discretisation scheme can be utilised. Sheng et al. (2006) used Bzier polynomials to

discretise the sharp corner between the cone and the shaft, showing its efficiency in

decreasing oscillations in the problem of static pile penetration. A Bzier-type smooth

discretisation technique was also employed here to analyse the penetration problem. The soil

response predicted using this method is denoted by smooth cone in Figure 4.19(a). Note that

the oscillation in the soil response was significantly decreased using the Bzier-type smooth

discretisation technique. Figure 4.19(b) depicts the results of the two analyses obtained by the

non-smooth NTS, as well as the mortar contact algorithms. For both discretisations, the same

104

penalty parameters were used. It was observed that the mortar algorithm significantly

decreased the oscillations in the response because of an increase of the polynomial order of

the shape functions. The remaining jumps in the response predicted by the mortar method

resulted from the fact that, although the discretisation was quadratic within the integration

area of a mortar-type element, it was not C1- continuous.

q/c'

q/c'

0

10

20

30

40

50

60

10

15

20

25

Non-Smooth NTS

Mortar method

Non-Smooth cone

Smooth Cone

Penetration / D

Penetration / D

(a)

(b)

Figure 4.19: Normalised total dynamic soil resistance versus normalised penetration

depth obtained for: (a) NTS method with smooth and non-smooth cone ( = 10 );

(b) non-smooth NTS and mortar methods ( = 2 )

Figure 4.20 represents the excess pore-water pressure variation during the pile installation at

a depth of 4D and initial radial distance of 2D.

105

60

Non-smooth cone

Smooth cone

40

20

-20

-40

0.0

0.2

0.4

0.6

0.8

1.0

Time (s)

r = 2D ( = 10 )

According to Figure 4.20, the smoothed contact discretisation decreased the oscillation in the

numerically predicted pore-water pressure, but it did not entirely eliminate the oscillatory

response of the soil medium. Another possible reason for the oscillation of the solution is that

the remapping of the state variables, as well as the contact path-dependent variables, may

disturb the equilibrium or the consistency principle of plasticity, thus producing some

unbalanced forces.

Note that all numerical solutions presented in the following and later in the thesis utilise the

mortar scheme unless otherwise stated.

Figure 4.21 plots the soil resistance (normalised by its drained cohesion) versus the

penetration (normalised by the pile diameter) predicted by the fine and dense meshes, which

gave rather close results in terms of the soil resistance. Figure 4.22(a) depicts the same

comparisons in terms of the excess pore-pressure response. Although the results obtained by

the two meshes were close, the excess pore-water pressure predicted by the fine mesh

experienced more fluctuation compared to the prediction obtained using the dense mesh.

106

q/c'

0

10

20

30

0

1

Penetration / D

2

3

4

Fine mesh

5

Dense mesh

6

7

8

9

Figure 4.21: Normalised total dynamic soil resistance versus normalised penetration

depth

To decrease this fluctuation and obtain a more stable result, the time step (110-4 sec) was

increased by a factor of two. These analysis results are depicted in Figure 4.22(b). According

to Figure 4.22(b), increasing the size of the time step reduced the oscillation observed in the

excess pore-water pressure response predicted by the fine mesh. It has already been

recognised that a small initial time step may produce spatial oscillations in pore pressures

near free-draining boundaries (Sandhu et al. 1977; Kanok-Nukulchal and Suaris 1982; Reed

1984). As a result of this phenomenon, Vermeer and Verruijt (1981) recommended that the

step size should not be reduced below a threshold value. However, hiding the oscillations in

the pore pressures by using large time steps is flawed logic, as the true transient response can

never be obtained in that way. Large pore-pressure oscillations that are adjacent to freedraining boundaries can alternatively be viewed as a signal that the mesh needs to be refined

in these zones (see Sloan and Abbo 1999).

107

Time (s)

0.2

0.4

0.6

0.8

0.0

1.0

0.2

0.4

0.6

0.8

1.0

50

50

0

Fine mesh

-50

Dense mesh

-100

-150

-200

-250

0.0

Time (s)

0

Fine mesh

-50

Dense mesh

-100

-150

-200

-250

-300

-300

-350

-350

(a)

(b)

(a) time step t = 510-5s for both analyses; (b) time step size increased to t = 110-4s

for the analysis with fine mesh only

Figure 4.22 and Figure 4.20 represent the excess pore-water pressure response for two points

initially located at a depth of 4D and radial distances of 0.15D and 2D, respectively, and

show that the tensile excess pore-water pressure (a suction increment) was generated close to

the pile shaft because the soil elements in this region underwent high plastic volumetric

expansion predicted by the MC model (the assumed dilation angle was 10). The excess porewater pressure then tended to become compressive at further radial distances because the

magnitude of the mean total stresses was comparatively larger than the deviatoric stresses at

larger radii. Figure 4.23 plots the excess pore-pressure counters for a dilation angle of 2,

which shows high-suction pore pressures adjacent to the pile shaft.

Figure 4.24 depicts the effect of the dilation angle on soil resistance by plotting the

normalised soil resistance versus the normalised penetration for dilation angles of 0, 2, 5

and 10. According to Figure 4.24, the variation of the dilation angle changed the soil

resistance considerably. By increasing the dilation angle above zero, the tendency of the soil

to harden will increase because it is forced to deform plastically at a constant volume over a

relatively short loading period compared to the typical drainage time.

108

q/c'

0

10

15

20

25

0

1

2

Penetration / D

3

4

5

6

Dilation = 0

Dilation = 2

Dilation = 5

Dilation = 10

8

9

Figure 4.24: Evolution of normalised total dynamic soil resistance for various dilation

angles

109

4.7.2

The pile installation problem was analysed again using the MCC material model. The

material parameters used in this analysis were as follows: slope of the normal compression

line = 0.25; slope of the unloadingreloading line = 0.05; initial void ratio e0 = 1.8; overconsolidation ratio OCR = 2 (at all depths); lateral stress ratio K0 = 1.0; total unit weight

k = 10-8 m/s. The same geometry, FE meshes and boundary conditions, as depicted in

Figure 4.17, were adopted in this analysis. To conduct a coupled consolidation analysis with

the MCC model, it was essential to generate the geostatic stress field due to soil self-weight,

and to guarantee that the initial stress state of the entire soil profile was within the state

boundary surface. The initial stress state was established by applying the body force and an

effective overburden pressure of p0 over a long period to allow all excess pore water

pressures to dissipate. As a result, the initial isotropic effective stress state (prior to pile

penetration) at depth z had a magnitude of 10z + p0 kPa. Four values were selected for p0: 30,

50, 70 and 150 kPa. After the initial stresses were established, all nodal displacements,

velocities and accelerations were set to zero, and the pile was then pushed into the soil by

applying a prescribed displacement at a rate of 8.75D/sec. Analyses were conducted utilising

both the fine and dense meshes.

Figure 4.25(a) shows the soil resistance versus the penetration of the pile, normalised by its

diameter. According to Figure 4.25(a), the soil resistance increased with an increase in the

overburden pressure. In addition, for a constant value of overburden pressure, the soil

resistance predicted by the MCC model increased as the pile penetrated into the soil layer, but

it did not converge to a steady state. This is mainly because the shear strength of the soil

predicted by the MCC model depends on the initial effective stresses, which increases

linearly with depth. In this instance, it can be assumed that the soil behaved in an undrained

manner, as the pile was pushed into the ground at a relatively rapid rate. As an illustration of

the constant volume deformation, Figure 4.25(b) depicts the deformed shape of the dense

mesh at the end of the penetration for the case where p0 = 50 kPa. It was observed that the

soil near the ground surface in the vicinity of the pile tended to heave, while the soil elements

within an approximate horizontal distance of one shaft radius underwent significant radial

compression in order to accommodate the pile in the undrained soil medium.

110

0

200

400

600

800

1000

0

1

Penetration / D

2

3

pp0=30kPa

0 = 30 kPa

pp0=50kPa

0 = 50 kPa

pp0=70kPa

0 = 70 kPa

pp0=150kPa

0=150 kPa

4

5

6

7

8

9

(a)

(b)

Figure 4.25: (a) Evolution of total dynamic soil resistance for various values of p0;

(b) deformed dense mesh at the end of installation

and, in particular, the material nonlinearity in this example, it was found that an appropriate

size for time steps was important to attain a converged solution. In this example, for the

analysis of the penetration phase, two time step sizes were used (510-5 s and 8.510-6 s),

whereas a constant time step size of 510-5 s was used to simulate the entire process of

penetration in the previous example (see Section 4.7.1).

Figure 4.26 depicts the excess pore-water pressure variation at a depth of 2.5D during the

installation process. According to Figure 4.26, the magnitude of excess pore-water pressure at

a point in the soil increased as the pile advanced towards it, but once the pile passed that

location, the magnitude of excess pore-water pressure decreased sharply and approached a

steady state value. The sudden drop in the excess pore-water pressure was significant for the

soil elements located within a radial distance of 1.0D from the pile shaft. It was also observed

that the pile could only influence a region ~3.0D in the radial direction and ~1.0D in the

111

vertical direction, as measured from the pile tip. Figure 4.27 depicts a contour plot of the

excess pore-water pressures developed at the end of the pile installation.

Time (s)

0.0

0.2

0.4

0.6

0.8

1.0

250

200

r=0.25D

150

r=0.50D

100

r=0.0

r=0.75D

r=1.0D

50

r=1.50D

r=2.0D

r=1.25D

r=1.75D

r=3.0D

-50

Figure 4.26: Excess pore water pressure variation throughout penetration at depth 2.5D

and different radial distances

Figure 4.28(a) plots the total dynamic soil resistance versus the normalised penetration that

was predicted using the fine and dense meshes. The results obtained for the two meshes were

almost identical for penetrations less than ~1.0D, but they tended to differ slightly for deeper

penetrations. At the end of the pile-installation phase of the analysis, the soil resistance

predicted by the dense mesh was approximately 10 per cent greater than the resistance

estimated by the fine mesh. Figure 4.28(b) plots the development of excess pore water

pressure at a point located at a radial distance of 0.5D and a depth of 6.25D, as predicted by

the fine and dense meshes. The pore pressure did not change before the pile approached the

point. However, (compressive) excess pore-water pressure developed as the pile further

penetrated into the soil. The maximum magnitudes of the pore-water pressure predicted by

the fine and dense meshes were ~225 kPa and ~185 kPa, respectively.

112

0

150

300

450

0

1

250

3

4

5

Fine mesh

Dense mesh

Penetration / D

200

Fine mesh (r=0.5D)

150

100

50

0

-50

0.0

0.2

0.4

0.6

0.8

1.0

Time (s)

(a)

(b)

Figure 4.28: (a) Evolution of total dynamic soil resistance; (b) excess pore-water

pressure at depth 6.25D

113

4.7.3

The problems considered in Sections 4.7.1 and 4.7.2 were presented in an attempt to illustrate

some important aspects of the numerical modelling, as well as the soil behaviour under fast

dynamic penetration. The importance of using a smooth, continuous geometry at the point of

transition between the conical tip and the cylindrical shaft was illustrated using the MC soil

model. In addition, the effect of the dilation angle on the mobilised soil resistance and the

excess pore-water pressure were highlighted, in which a non-zero dilation angle resulted in an

unrealistically high prediction of suction pore pressure (see Figure 4.22). Conversely, in the

analysis using the MCC soil model, greater emphasis was placed on the generation of excess

pore-water pressure during the penetration, and the predictions revealed a pattern in which

the excess pore pressure at a given depth first increased, then decreased and ultimately

approached a steady state as the cone tip approached and then passed beneath that given

location. Moreover, the results of this analysis with a constant rate of penetration can be

compared and contrasted with the analysis of a torpedo anchor analysis (to be presented in

Chapter 5), where varying rates of penetration are applied.

A direct comparison between the results obtained by the two soil models cannot be made in a

straightforward way. The MC model is an elastic-perfectly plastic material model that entails

no volumetric hardening because the yield surface is fixed in stress space. The adopted yield

function only depends upon the stresses, and the size of the yield surface is determined by

constant strength parameters. In contrast, the yield surface in the MCC model is allowed to

expand or contract with a constant shape and, correspondingly, involves isotropic hardening

or softening due to changes in plastic volumetric strain. Therefore, a reasonable comparative

study might be feasible only by setting the model parameters such that the same, or similar,

elastic and plastic behaviours are predicted by the two models, including comparable profiles

of undrained shear strength.

To make such a comparison, soil parameters were selected so as to provide nearly the same

undrained shear strength su profile and the same or similar rigidity index Ir = G/su. The shear

modulus of soil G is constant for the MC material, but it is stress-dependent in the MCC

model, noting that the effective Poissons ratio is held constant for the adopted MCC model,

and G is determined consistently from the bulk modulus. Table 4.4 provides the assumed soil

parameters for the two models.

114

MCC

MC

= 20 kN/m3

= 20 kN/m3

e0 = 1.8

E = 1400 kPa

= 0.3

= 0.3

' = 25

' = 18

= 0.36

= 0

= 0.07

c = 0.20 kPa

OCR = 2

k = 10-8 m/s

-8

k = 10 m/s

A theoretical formula for predicting the undrained shear strength of K0-consolidated soil can

be derived based on the deviatoric stress at failure as:

su = J f cos

4.9

where Jf and are stress invariants denoting deviatoric stress and Lodes angle at failure,

respectively. Potts and Zdravkovic (1999) gave an explicit form of su based on the MCC

model parameters as:

2 (1 + 2 K 0OC )

OCR

NC

2

=

(1 + 2 K 0 ) (1 + B )

su 0 g ( ) cos

NC

2

6

(1 + 2 K 0 ) OCR (1 + B )

4.10

where o denotes initial vertical effective stress and B and g ( ) are defined by:

g ( ) =

B=

sin

sin sin

cos +

3

3 (1 K 0NC )

g (30 ) (1 + 2 K 0NC )

115

4.11

4.12

where g ( ) specifies the shape of the yield surface in the deviatoric plane. If the Von Mises

circle that circumscribes the MC hexagon is chosen for the yield surface, g ( ) can be

expressed for triaxial compression ( = 30 ) as:

g ( ) =

2 3 sin

3 sin

4.13

If the soil element is normally consolidated, K 0NC = K 0OC and OCR = 1, Eq. 4.10 reduces to:

(1 + 2 K 0NC ) 1 + B 2

su = 0 g ( ) cos

3

2

4.14

K 0NC = 1 sin

4.15

K 0OC = K 0NC OCR sin

4.16

sin 1 + B 2

su = 0

2B 2

4.17

Worth (1984) also derived this for the undrained shear strength of a 1D normally

consolidated clay in plane strain conditions.

Using Eqs 4.10, 4.11 and 4.12, the undrained shear strength of 1D over-consolidated clay in

plane strain conditions can be expressed as:

su = 0

sin 1 + B

2B 2

2

OCR

NC

(1 + 2 K 0 ) OCR

4.18

considering the geometry of Mohrs circle of stress at failure, resulting in:

1 + K0

su c cos +

=

2

116

0 sin

4.19

Figure 4.29 depicts the undrained shear strength profile for the two soil models assuming

K 0NC

= K=

0.43 and an effective surcharge pressure p0 = 30 kPa. The value of the rigidity

0

index is constant for the MCC model (i.e., Ir = 45, independent of depth). However, it varies

with depth in the MC model because G is constant but su varies with depth. The assumed

parameters listed in Table 4.4 provide an average value of Ir = 45 for the MC model over the

depth range of interest in this problem.

Another aspect of the soil model that can be important for comparison purposes is the shape

of its yield surface in the deviatoric plane. All previously described analyses that were

conducted using the MCC model utilised a Von Mises circular shape for the yield surface in

the deviatoric plane, which may over-predict the strength compared to the MC model.

Therefore, it was decided to conduct an additional analysis with the MCC model in which its

smooth, non-circular yield surface coincided with the MC hexagon at all vertices in the

deviatoric plane (see Sheng et al. 2000). It is notable that the analysis procedure, adopted fine

mesh, penetration rate and boundary conditions were all in accordance with those adopted in

Section 4.7.1. Figure 4.30 depicts the predicted soil resistance q normalised by su versus

normalised penetration depth.

su (kPa)

0

10

20

30

0

MCC

MC

Depth (m)

117

According to Figure 4.30, all analyses predicted almost the same response. This should not be

surprising because, in addition to adopting approximately the same undrained shear strength

profile and (average) rigidity index, the zero dilation angle prevented the MC soil from

hardening under undrained conditions (Small 1977), thus over-predicting the soil resistance.

Conversely, rapid penetration combined with low soil permeability would have imposed

effectively undrained behaviour, during which lightly over-consolidated MCC soil would

have engaged an initial yield surface and reached the critical state line, resulting in undrained

failure through an effective stress path, which entails no expansion of the initial yield surface

if OCR = 2. As shown in Figure 4.30, the normalised soil resistance predicted by MCC1 was

less than that predicted by MCC2, which was consistent with the adopted shapes of the yield

surface in the deviatoric plane.

q/su

0

10

15

Penetration / D

7

MCC1 - Smooth non-circular shape

MC

Figure 4.30: Evolution of normalised total dynamic soil resistance predicted by three

soil models

118

4.7.4

The analyses explained in the two preceding sections assumed a smooth interface between

the soil and pile. This section provides results when a rough interface is adopted. The

analyses are only presented for the MCC soil model, and frictional contact was modelled

using the mortar scheme. The adopted mesh and soil parameters were identical to those

presented in Section 4.7.2.

The lateral stress ratio and overburden pressure were assumed K0 = 0.67 and p0 = 50 kPa,

respectively. The soil was defined as non-mortar, with 20 Gauss points within a segment for

numerical integration. The introduction of high penalty parameters on the contact surface

caused convergence difficulties in the solution of the equations. After some trial and error,

the appropriate choice of penalty parameter was u = 105 .

Figure 4.31 plots the predicted total reaction forces for both the frictionless and frictional

cases, in which = 0.25 was assumed for the friction coefficient. It was observed that the

difference between the results was almost negligible, implying that the amount of effective

normal stresses at the interface were relatively small compared to the corresponding excess

pore pressures, and that the soil was in an undrained condition. Consequently, the assumption

of a smooth interface for problems involving fast penetration may be relevant as long as the

soil permeability is relatively low.

To evaluate the effect of soil permeability on the developed frictional forces, the value of the

soil permeability was increased to k = 10-3 m/s. Although this amount of permeability is

unrealistic for clay-type materials, the analysis result may be indicative of the roughness

effects on the interface. Figure 4.32 depicts the predicted results for two different values of

= 0.20 and = 0.25. It was observed that the reaction forces diverged after penetration

depths of 1D, where the pile shaft came into contact with the soil surface. This is even better

captured in Figure 4.33, which depicts the reaction forces in normal and tangential directions

for the value of = 0.25. The tangential or sleeve resistance continuously increased as the

pipe further penetrated into the soil, ultimately reaching ~56 kPa at the end of the installation

process, where the total tip resistance had reached ~ 320 kPa.

119

350

300

250

200

Frictionless

150

=0.25

100

50

0

0

Penetration /D

Figure 4.31: Evolution of total dynamic soil resistance for smooth and rough interfaces

( = 0.25), soil permeability k = 10-8 m/s

400

350

300

250

200

Frictionless

=0.2

150

=0.25

100

50

0

0

Penetration /D

Figure 4.32: Evolution of total dynamic soil resistance for smooth and rough interfaces,

soil permeability k = 10-3 m/s

120

350

Total contact pressure tN = tN+p

Dynamic soil resistance (kPa)

300

250

200

150

100

Tangential contact stress tT

50

0

0

Penetration /D

Figure 4.33: Stresses on contact area, soil permeability k = 10-3 m/s, ( = 0.25)

4.8 Summary

The first part of this chapter detailed a series of validation exercises to evaluate the

performance of the developed numerical scheme. The analytical solutions of de Boer et al.

(1993) for the wave propagation responses of a porous layer subjected to a dynamic load

were used to verify the formulation of the U-P-V algorithm for dynamic consolidation

problems. The dynamic coupled consolidation formulation was also validated for both small

and large deformation analysis by test problems presented by Meroi et al. (1995). The

propagation of plane waves through a porous medium with a compressible pore fluid was

examined, and the results were compared with closed-form solutions presented by Carter et

al. (2014). Two waves of dynamic pore pressurethe fast dilatational wave and Biots slow

wavewere identified, matching with the analytical solution of Carter et al. (2015). This

example also demonstrated the effectiveness of the numerical algorithm when analysing a

problem that involves a significant transient (short-term) dynamic response that is dominated

by inertia effects, followed by the increasing importance of the consolidation phenomenon in

the porous medium at intermediate and large times.

121

The consolidation of a flexible strip footing under a uniform pressure was modelled, and the

predicted results were compared with those presented by Manoharan and Dasgupta (1995).

The capability of the dynamic consolidation algorithm was verified for elasto-plastic

consolidation processes, during which strong coupling occurred between the three field

variables of solid displacement, pore-fluid pressure and fluid velocity.

The undrained behaviour of a strip footing under dynamic loading and large deformations

was modelled in order to illustrate the performance of the code in large deformation regimes,

particularly when combined with the ALE scheme and the contact algorithm. Analyses were

conducted utilising both the U-P and U-P-V consolidation schemes, and the predicted results

were compared with an alternative numerical solution that only considers the displacement

degrees-of-freedom in the analysis. The alternative solution assumed that the soil behaves as

a Tresca material model deforming under undrained conditions. The frictional contact

algorithm in this example proved to be capable of transferring vertical pressure and

horizontal stresses withstanding against the tangential forces, in which a rough footing was

modelled through a relatively high-friction coefficient at the interface. It was found that three

Gauss points were enough for this particular problem, mainly because of the small

deformations at the interface (due to the rough interface), so that the meshes were nearly

conforming. The mortar contact algorithm was also employed to check the contact patch test

by modelling the unconfined compression of a saturated porous layer.

The second part of this chapter presented the analysis of the penetration of a rigid pile into a

saturated soil layer. A thorough investigation was conducted on the response of soil and porewater pressure generation resulting from the fast penetration of a pile. The results revealed

how soil resistance is mobilised and pore-water pressures are generated during the

penetration process. Further, the effects of different soil models, soil parameters, mesh sizes

and contact discretisation schemes on the predicted results were studied.

It was shown that the ALE method can tackle large deformation problems in geomechanics in

which the displacements, velocities and accelerations are coupled with pore-water pressures

and Darcys velocity and involve rapid loading as well as changing boundary conditions.

The numerical results also showed that large-suction pore-pressure changes are developed for

soil elements close to the pile when the soil is represented by the MC material model with a

122

non-zero dilation angle. For the lightly over-consolidated soil modelled by the MCC material

model, compressive excess pore-water pressures are generated in the soil around the pile tip

and shaft. Their magnitudes first increase when the pile tip is above or at the level of the

point of interest in the soil, and then decrease once the pile tip has moved below the

evaluation point, finally approaching a steady value at the end of the installation phase.

A smooth discretisation scheme with the NTS method was used to avoid sharp transitions in

contact element segments. The results showed that smoothing the transition point between the

cone tip and its shaft can noticeably decrease the oscillatory responses observed in the

predicted total dynamic soil resistance, as well as the excess pore-water pressure.

123

Anchors

5.1 Introduction

Dynamically penetrating anchors (DPAs) have proven to be promising systems for anchoring

taut mooring lines of floating offshore oil and gas exploration and production units because

of their relatively easy installation process. The kinetic energy of a DPA attained by gravity

throughout free-fall through the water column provides the required dynamic penetration

force, making it more practical and cost-effective than other offshore structures such as

suction piles, driven piles, drilled and grouted piles, and drag embedment anchors. The costs

of installation of these methods can dramatically increase with water depth.

The deep penetrating anchor (Lieng et al. 1999, 2000) and a less sophisticated torpedo anchor

are two types of DPAs that are conceptually similar (see Figure 5.1), and both are referred to

here as DPAs. The deep penetrating anchor is a dart-shaped, thick-walled steel cylinder with

flat plates or flukes attached to its upper section, as shown in Figure 5.1(a). A torpedo usually

consists of a pipe pile (1218 m in length and 0.761.07 m in diameter) filled with scrap

metal and concrete, close-ended and fitted with a conical tip and sometimes including fins at

the top end, which provide stability during free-fall. To retain the vertical trajectory of the

pile inside the soil and water, the centre of gravity is maintained below the centre of

buoyancy by increasing the weight of the shaft, which for various designs varies between 241

and 961 kN. Torpedo anchors were first commercially employed in the Campos Basin,

offshore Brazil (Medeiros 2002). The impact velocity of the torpedo piles reported by

Medeiros (2002) varied between 10 and 22 m/s for hanging heights, from which free-fall

commenced between 30 and 150 m as measured from the seabed, and the penetration depth

usually varied between 8 m and 22 m.

124

(a)

(b)

Figure 5.1: (a) Deep penetrating anchor (taken from Deep Sea Anchors); (b) torpedo

anchor with fins and without fins (after Medeiros 2002)

In place, DPA behaviour is expected to be similar to that of conventional piles, where uplift

forces are resisted by the friction developed at the anchor-soil interface.

A torpedo anchor is typically installed into soft to medium clay soils, where it performs best.

However, the existence of oil and gas deposits in calcareous sediments of ocean floors with

water depths exceeding 900 m (Watson and Randolph 1998) motivated the investigation of

the anchor performance in calcareous sand. The field tests of Medeiros (2001, 2002) in

uncemented calcareous sand conducted in the Campos Basin showed an average tip

penetration of 15 m for a drop height of 30 m. Laboratory experiments conducted by

Richardson et al. (2005) suggested that DPAs could be used in calcareous sediments. They

carried out some centrifuge tests on DPAs in medium-density calcareous sand with a uniform

particle distribution. The experiment results indicated that, at a similar impact velocity, the

embedment depth in calcareous sands is, on overage, half of those measured in normally

consolidated clay, which agreed with the field tests reported by Medeiros (2002). The holding

capacity of the anchor was evaluated at 12 times the anchors dead weight in air, and it

corresponded to 35 times the anchors dead weight for normally consolidated clay.

embedment depth, pull-out capacity and the prediction of stresses in their structure remain a

challenge. Current design procedures include an estimation of the penetration depth through a

125

theoretical model and predicting the pull-out capacity. Consequently, a simulation of the

installation process is neglected, whereas the installation of a DPA leads to considerable

disturbance and remoulding of the soil in the vicinity of the anchor.

The first step in the analysis of a DPA involves the simulation of the installation phase in

order to predict the penetration depth, soil resistance and development of excess pore-water

pressure. However, in the majority of research works devoted to the analysis of DPAs, the

effect of installation on pull-out or the lateral capacity of the anchor is ignored. That is, in

most analyses conducted to date, deep foundation systems are wished in place with no effort

to model the installation phase; hence, a perfect interface between the anchor system and the

surrounding soil is assumed.

The initial stress state of the soil is usually estimated based on the submerged unit weight, the

lateral earth pressure coefficient at rest, and assumes zero excess pore-water pressure. The

estimation of the penetration depth of the anchor usually relies on the theoretical framework

developed by True (1976), which is based on the equation of motion and the soil resistance

components during penetration, including end-bearing force, side-adhesion force and inertial

drag force. This method has been utilised by different researchers to predict the embedment

depth (e.g., designers of DPAs and Medeiros 2002; OLoughlin et al. 2004; Audibert et al.

2006; Sousa et al. 2010). However, this method has some limitations because of the

simplifying assumptions made in the model development. Raie (2009) summarised the

limitations of Trues method as: (1) strain rate effect is only velocity-dependent, while the

geometry of the anchor can influence the strain rate; (2) average coefficients are used for the

penetrator effective mass, inertial drag coefficient and side-adhesion factor, while they can be

dependent on the geometry of the penetrator and the velocity of the penetrator throughout

penetration; and (3) the results of Trues method are limited to the velocity profile and final

embedment depth.

Sturm and Andresen (2010) presented a FE model for torpedo anchors using the commercial

software package Abaqus, with a user-defined contact algorithm and a UL formulation.

However, they simulated the installation process quasi-statically with a constant penetration

126

rate and neglected any inertia effects. They used the Tresca material model to simulate soil

behaviour and evaluated the excess pore pressure using knowledge of the mean stress

distribution and shear strain in the soil. Raie (2009) developed an alternative procedure based

on Computational Fluid Dynamics (CFD) to predict the embedment depth and installation

effects, including shear distributions on the soil-anchor interface and the soil state parameters.

This method is based on the principles of fluid dynamics, where stress at any point of the

media is equal to the pressure at that point independent of the direction; that is, the vertical

and horizontal stresses on the soil elements are assumed to be identical, with this being an

unrealistic assumption for soil.

The second stage in the simulation of DPAs is the set-up analysis. With knowledge of the

effective stresses and excess pore-water pressures generated during installation, the set-up

analysis can be performed by reconsolidating the soil in the vicinity of the anchor. In DPA

systems, excess pore pressure is generated because of two main factors: shearing of soil

during installation, and increase in total stress because of the vertical and mostly radial soil

volume changes. The excess pore-water pressures result in lower frictional resistance, which

leads to lower pull-out capacity of the anchor. As the soil consolidates, the pull-out capacity

of the anchor increases because of the dissipation of excess pore pressures and the

corresponding increases with time of the effective stresses.

Carter et al. (1979) and Randolph and Wroth (1979) presented analyses of the stress changes

resulting from the expansion of a cylindrical cavity and the subsequent consolidation of the

soil for soil idealised as an elastic or elastic-perfectly plastic material. Randolph et al. (1979)

applied the cavity expansion theory to model pile installation and estimated the radial

consolidation of soil around a driven pile. The results of their study compared well with some

field measurements and showed that the major pore-water pressure gradients around driven

piles are radial. Hence, for driven piles, the dominant consolidation factor is considered

horizontal rather than vertical. Richardson et al. (2009) performed a series of centrifuge tests

on model torpedo anchors installed in kaolin clay, with a scale of 1:200, and used the cavity

expansion method (CEM) to predict the consolidation behaviour of the soil around the

torpedo anchor. They concluded that the CEM can be a relatively accurate procedure to

predict the consolidation of soil surrounding torpedo anchors without fins, as well as those

with four flukes. Moreover, the experimental results reported by Richardson et al. (2009)

showed that, for typical coefficient values of radial consolidation ch = 330 m2/yr, the time

127

required for 50 per cent consolidation of a prototype dynamic anchor (with a shaft diameter

of 1.2m) ranges from approximately 35 to 350 days, whereas 90 per cent consolidation takes

place within 2.424 years. Mirza (1999) also stated that full capacity of an offshore pile is

attained after a consolidation period of 12 years. Jeanjean (2006) reported that 90 days or

fewer are required to accomplish 90 per cent of the operational capacity of suction anchors.

FE analysis results reported by Raie et al. (2009) indicated that, for kaolin clay with

permeability of 8 109 m / s and 8 1010 m / s , the required times for a 99.9 per cent degree

of consolidation are 42 and 420 days, respectively. Medeiros (2002) reported an increase of

275 per cent on pull-out capacity within 10 days for a finless torpedo anchor (0.76 m in

diameter, 12 m in length and 240 kN in weight). Lieng et al. (1999) used 1D consolidation

analysis for the case of clay deposits in Norway by taking into account radial consolidation

with a drainage passage equal to 4R (R is the radius of the anchor). They concluded that after

two weeks, 70 per cent of anchor capacity is achievable.

The third and final step of the torpedo analysis includes an estimation of the pull-out capacity

of the anchor. The FEM and the American Petroleum Institute (API 2002) method are the two

most common techniques for estimating the holding capacity of torpedo anchors. The API

method takes advantage of the conventional theory of pile-bearing capacity based on the total

stresses and predicts the undrained holding capacity of deep penetrating anchors in cohesive

soils. API (2000) recommended two formulas to predict skin friction and end-bearing

capacity in cohesive clay to apply in a pile-bearing capacity equation. In fact, this method

provides an upper limit for the holding capacity, as it is assumed that maximum skin friction

and end-bearing capacity are mobilised at the same time. However, to mobilise the maximum

skin friction and end-bearing resistance, a certain amount of displacement should take place,

and this amount of displacement can sometimes go beyond that of anchor serviceability. The

API provided some curves that can be used to estimate the displacements corresponding to

maximum skin friction and end-bearing capacity mobilisations. OLoughlin et al. (2004)

adjusted the API method to account for the problems of dynamic finless torpedo anchors in

clay soils, and Richardson et al. (2005) adopted the method for the case of calcareous sand.

Gilbert et al. (2008) also utilised the API method to estimate the fast pull-out capacity of

model torpedo anchors. Modelling the soil as a DruckerPrager material and assuming the

anchor to be whished in place, Sousa et al. (2010) employed the FEM to evaluate the longterm load capacity of a typical torpedo anchor subjected to vertical and inclined loads. The

128

analysis was conducted with different designs of flukes and different soil undrained shear

strengths. They evaluated embedment depth using Trues (1974) method and observed that to

simulate infinite media for the purpose of static uplift capacity analysis, one needs to

extended the domain up to 20 times the anchor diameter. Moreover, they stated that axial

load capacities evaluated using the FEM agree with the results estimated using the API

methodology.

A brief survey of the literature reveals that the analysis of DPAs needs further research in

order to realistically model and evaluate their behaviour. Most available studies are based on

experimental or approximate analytical solutions, and the current FE simulations are

generally based on a displacement formulation (neglecting the pore-water pressures) and

consider simplifying assumptions in the modelling. The major limitation is a lack of

knowledge of the effective stress state and pore pressure around the pile. No laboratory tests

have yet been reported with measurements of excess pore pressures or effective stresses in

the soil during or following the dynamic penetration of objects. This is largely because of the

fast and transient nature of the problem, which requires a set of sophisticated piezometers and

instrumentation techniques in low permeability materials such as clay. Therefore, the

problem still remains as to how pore pressures and stresses are affected by the installation of

DPAs. A more realistic model must then incorporate pore-fluid pressure development along

with deformations, velocities and accelerations, to facilitate a thorough understanding of soil

response. Moreover, subsequent dissipation can be investigated by providing the initial

undrained or partially drained distributions of pore pressure. Such problems require a fully

coupled analysis that takes into account the interaction between soil and pore fluid by

incorporating the effect of the transient flow of the pore fluid through the inter-connected

voids of the solid skeleton.

This chapter illustrates an important application of the developed computational scheme and

attempts to shed light on the response of soil throughout the installation and set-up of DPAs.

129

A rigid, finless torpedo anchor falling freely into a saturated soil layer represented by the

MCC material model was studied in this example. The aim of the analysis was to study the

total penetration depth of the anchor, mobilised soil resistance, deceleration characteristics of

the anchor during the installation phase, and excess pore-water pressures generated in the

surrounding soil together with their dissipation with time. Figure 5.2(a) presents the geometry

of the torpedo anchor, the finite element mesh containing 3,416 triangular elements and 7,028

nodes, the boundary conditions and the soil properties. The radial thickness of the soil

elements underneath the anchor is equal to one-third of the anchor shaft radius. It is notable

that the analysis results presented in Section 4.7 regarding mesh sensitivity suggested that the

predicted soil resistance and pore pressure were relatively the same for the mesh sizes of

0.125D and 0.25D. Therefore, the mesh size of 0.167D (next to the anchor shaft in the radial

direction) was chosen for this analysis, and test simulations proved that it can provide reliable

predictions while also being computationally efficient. The submerged weight of the anchor

(W), its diameter (D) and its length (L) were assumed to be 40 kN, 0.6 m and 6.0 m,

respectively. To avoid further material nonlinearity and complexity, the shear strength

increase due to strain rate effects was ignored in this example. The coefficient of friction at

the interface was 0.20, and a penalty parameter equal to u = 1 106 was used to enforce the

contact constraints. The analysis was conducted utilising both NTS (with a smoothed contact

geometry utilising Bzier polynomials) and mortar contact schemes. Figure 5.2(b) shows the

geometry of a torpedo with a curved surface at the cone and top of the pile. This geometry

was adopted for the analysis with the mortar algorithm only. The anchor diameter and its

submerged weight were the same as that shown in Figure 5.2(a), except that its length was

6.8 m.

The analysis of the torpedo anchor installation included four stages. In the first stage, a body

force loading because of the self-weight of the soil was used to establish the initial stress field

in the soil body. The second stage of the analysis included applying an overburden pressure

of p0 = 50kPa to the surface of the soil layer over a long period to allow the dissipation of

excess pore pressures. This overburden alleviates numerical problems associated with the

very low shear strength of the soil at the mudline when using the Modified Cam Clay soil

model and normally consolidated conditions. Besides providing numerical stability, this

130

such as the coefficient of consolidation and the initial undrained shear strength are essentially

invariant with depth.

Axis of symmetry

0.5D

D=0.6m

L=6.0m

Permeable

K0 = 0.67

= 0.25

e0 = 1.8

OCR = 2

= 25

= 0.3

k = 10-8 m/s

= 20 kN/m3

20D

= 0.05

7.5D

(a)

(b)

Figure 5.2: (a) FE model of torpedo anchor analysis; (b) Torpedo shape adopted for the

analysis with the mortar contact

After generating a non-zero stress field, the location of the yield surface at each integration

point in the FE mesh was adjusted according to the initial effective stresses and the

designated value of the over-consolidation ratio (OCR). The third stage of the analysis

involved applying the absorbing boundary conditions. In this step, reaction forces were first

extracted from the boundary nodes. Then all boundary restraints were released, and nodal

131

reaction forces were applied to bring the system to equilibrium. The reaction forces were

typically applied in 10 increments. The absorbing boundary conditions were then activated,

and analysis proceeded to the next stage. Finally, the torpedo was deployed and allowed to

impact the soil vertically at a specified initial velocity. In this example, three different values

of the impact velocity10, 15 and 20 m/swere adopted.

Figure 5.3 plots the predicted total dynamic soil resistance versus the penetration (normalised

by D) for the three impact velocities. These results were obtained using the NTS algorithm.

The intensity of the numerical oscillations was severe (for the selected penalty parameter

u = 1 106 ), so rather moderated responses are shown in Figure 5.3 for better visual clarity.

As expected, the total depth of penetration depended on the impact velocity. For impact

velocities of 10, 15 and 20 m/s, the predicted depths of installation were 8.78D, 13.26D and

17.86D, respectively. The soil resistance increased as the pile penetrated into the soil layer,

but it did not converge to a steady state. This is mainly because the shear strength of the soil

predicted by the MCC model depended on the initial effective stresses, which increased

linearly with depth.

900

750

600

450

300

Impact velocity = -15 m/sec

150

Impact velocity = -20 m/sec

0

0

12

15

Penetration / D

132

18

Carter et al. (2010) showed that the effect of the strain rate on the shear strength of the soil

should not necessarily be ignored in problems involving the fast penetration of objects into

soil layers. However, strain rate effects were not considered in this study because the

undrained shear strength was not a direct input parameter of the MCC model. On the

contrary, the rate-independent value of the undrained strength is predicted by the model and

could be related directly to the parameters of the MCC model and the initial state of the

stresses in the soil (e.g., Potts and Zdravkovic 1999).

700

600

500

400

300

Mortar contact

200

Bezier polynomial

100

0

0

10

12

14

Penetration / D

Figure 5.4: Total dynamic soil resistance profile obtained by the mortar and NTS

algorithms (impact velocity = 15 m/s)

For the impact velocity of 15 m/s, the analysis results obtained by the mortar algorithm were

compared with the NTS scheme in Figure 5.4, showing significant improvement in the

numerical oscillation. The use of the mortar-type method in the FE contact model facilitated a

curved surface between the torpedo and the soil because of the quadratic shape functions (see

Figure 5.2(b)). Therefore, the FE discretisation of the pile did not include any sharp corners,

thus reducing the numerical oscillations in the soil response. Further, in large sliding transient

applications, non-smooth transition from one element to the next can induce non-physical

inertial discontinuities, which contribute to oscillations in the solution. It is worth noting that

the non-mortar body (soil surface) should always have a finer mesh than the corresponding

mortar body (rigid pile) to ensure that there is a limited number of non-continuous

differentiable locations in the integration area.

133

The deceleration characteristics of the torpedo can be investigated by studying its velocity

changes during penetration. Figure 5.5 and Figure 5.6 show the variations of torpedo velocity

versus the normalised penetration and time, respectively. It was observed that the torpedo

accelerated slightly in the early stages of the penetration for two main reasons: (a) the sum of

soil resistance and the frictional forces at the interface between the soil and torpedo was

initially less than the submerged weight of the torpedo; and (b) the impact velocity of the

torpedo was less than its terminal velocity in water, allowing the torpedo to continue to

accelerate even after the initial impact. After the resisting forces exceeded the submerged

weight of the torpedo, it started to decelerate at an approximately linear rate. This rate was

consistent with the soil resistance profiles plotted in Figure 5.3. According to Figure 5.5, the

deceleration occurred after the torpedo had penetrated approximately 1D into the soil layer. It

should be noted that the depth at which the anchor began to decelerate could decrease with

increasing impact velocity if the increase in shear and inertial drag resistance with velocity

had been considered (OLoughlin et al. 2013).

Velocity (m/s)

0

10

15

20

0

2

4

Penetration / D

6

8

10

12

14

16

18

20

134

25

Time (s)

0.00

0

0.20

0.40

0.60

0.80

1.00

Velocity (m/s)

10

15

Impact velocity = 10 m/s

20

Impact velocity = 20 m/s

25

To investigate the generation of excess pore pressures, the development of excess pore

pressures for an impact velocity of 15 m/s were monitored at seven different points located at

a depth of 5.0D and radial distances of 0.0, 0.17D, 0.33D, 0.5D, 0.67D, 1.5D and 4.0D

throughout the installation process. Figure 5.7 plots the excess pore-water pressures

developed at these seven points versus time, noting that the total installation time was 0.85 s.

According to this figure, the excess pore-water pressures experienced a relatively steady state

following a peak value in each case. Typically, the magnitude of excess pore-water pressure

at a point in the soil increased as the pile advanced towards it, but once the pile passed that

location, the magnitude decreased sharply and approached the steady state value. The sudden

drop in the excess pore water pressure was significant for the soil elements located within a

radial distance of 1.0D from the pile shaft. The behaviour detected in Figure 5.7 is due to the

fact that both the mean and deviatoric stresses increase as long as the tip of the penetrometer

is above or at the level of the evaluation point in the soil, and then both decline once the tip

has moved below the evaluation point. Similar behaviour was also detected in Section 4.7.2.

It was also observed that the installation of the pile could significantly influence a region

~3.0D in the radial direction and ~1.0D in the vertical direction, as measured from the pile

tip.

135

To visualise the size and shape of the region affected by the anchor installation, Figure 5.8

presents two contour plots of excess pore-water pressure. These contour plots were captured

at embedment depths of 5.0D and at the end of the installation. Figure 5.8 shows that a

compressive excess pore-water pressure bulb grew around the anchor tip and shaft during

installation. This bulb extended a distance of approximately 3.0D in the radial direction and

about 1.0D in the vertical direction, as measured from the anchor tip, whereas the maximum

compressive values were developed at the anchor tip and extended to its shoulder.

300

250

Excess pore water pressure (kPa)

r=0.17D

200

r=0.33D

150

r=0.50D

r=0.67D

100

r=0.0

r=1.50D

50

r=4.0D

-50

0.0

0.2

0.4

0.6

0.8

Time (s)

Figure 5.7: Excess pore-water pressure evolution at depth 5.0D throughout the

installation phase

Moreover, a tensile (suction pore-pressure change) region was located at a distance of about

1.0D vertically underneath the anchor tip, and it advanced downwards along with, and ahead

of, the anchor. Another suction zone was also developed near the ground surface. As outlined

earlier, these suction zones develop because of the tendency towards plastic dilation induced

by the shear yielding.

136

kPa

(a)

(b)

Figure 5.8: Excess pore-water pressure contours at two different penetration depths:

(a) 5.0D; (b) end of installation (impact velocity = 15 m/s)NTS results

Figure 5.8 also demonstrates that the pathway of the anchor was left continuously open from

the top of the anchor up to the crater at the soil surface with a diameter of less than the anchor

diameter. Poorooshasb and James (1989) studied the hole closure behind dynamically

installed projectiles in kaolin clay in laboratory centrifuge tests. They observed open

pathways from the top of the projectiles to a certain extent. However, the entering crater at

the soil surface was always closed. In some tests, the pathways were continuously open up to

the vicinity of the entry crater, with a typical diameter of one-third to half of the projectile

diameter. Aubeny and Shi (2006) assumed that a cylindrical void forms because of the

installation of impact penetrometers with an aspect ratio of L/D = 4.25, where L denoted the

length of the penetrometer and D its diameter. OLouglin et al. (2013) also assumed hole

closure behind the shaft of DPAs to occur only if L/D 2.

Figure 5.9 depicts two deformed meshes during the free-falling process for the analysis

conducted by mortar contact with the anchor geometry, as depicted in Figure 5.2(b). It was

observed that the pathway of the anchor was left continuously open from the top of the

anchor up to the crater at the soil surface. This could result in a lower pull-out capacity

137

because it would not comprise the reverse end-bearing resistance at the upper end of the

anchor shaft. Section 5.4 presents another example, which shows the occurrence of the hole

closure behind the anchor.

Figure 5.9: Deformed meshes during the free-falling process (analysis with mortar

contact)

The set-up analysis aims to address the installation effects on the holding capacity. The

results of this stage of the analysis indicate the rate of dissipation of excess pore pressures

and the corresponding soil strength recovery. These are important factors in evaluating the

pull-out capacity of DPAs, which may be predicted at different times after installation.

138

When the full embedment depth was achieved and the torpedo came to rest, the

computational process automatically proceeded to the set-up analysis, and consolidation of

the soil was permitted. This process was continued until the generated excess pore-water

pressures dissipated entirely. Figure 5.10 depicts the dissipation curves for the soil elements

located at a depth of 5.0D and different radial distances measured from the pile tip. Note that

the pore pressure close to the pile initially fell off very rapidly. A similar phenomenon was

reported by Seed and Rees (1955) and Eide et al. (1961), who showed very rapid increases in

the bearing capacity of a driven pile within a short time after driving. This was also identified

by Randolph and Worth (1979), who studied the radial consolidation of soil following the

creation of a cylindrical cavity using an analytical solution. In their solution, the initial

distribution of pore pressure is assumed to be a function of the rigidity index Ir =G/su of the

soil, where G denotes the shear modulus of the soil. The dissipation of the excess pore

pressures is then governed by the extent of the pore-pressure zone around the cavity, which is

quantified by the rigidity index. Richardson et al. (2009) reported that Randolph and Worths

(1979) theoretical solution for the value of Ir = 500 provides a relatively accurate

representation of the measured increase in capacity with time for DPAs. Nevertheless, it was

observed from Figure 5.10 that a degree of consolidation of 90 per cent (the degree of

consolidation is equal to 1 minus the ratio between the current excess pore-water pressure and

the initial excess pore-water pressure) for elements within a 1.0D radial distance of the

torpedo shaft was achieved 13~33 days after installation. The same degree of consolidation

took place within 33~72 days for elements between 1.0D and 1.83D. Figure 5.10 also shows

that a degree of consolidation of ~96 per cent was attained for the entire affected zone at

depth 5.0D after ~260 days.

It was concluded that most of the pull-out capacity of the torpedo anchor (soil resistance) was

available much earlier than the completion of consolidation, as most of the excess pore

pressure dissipated within a matter of days or weeks. Further discussions are presented in the

next example to further enhance the understanding of the consolidation process.

139

120

100

r=0.0

80

r=0.17D

60

r=0.33D

r=0.50D

40

r=0.67D

r=1.0D

r=1.50D

r=1.83D

20

r=3.0D

0

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

1.E+00

1.E+01

1.E+02

1.E+03

Time (day)

Figure 5.10: Excess pore-water pressure dissipation versus time for elements at depth

5.0D

Layer

This section attempts to illustrate some other aspects of DPAs, such as the hole closure

behind the anchor, the effect of overburden on the predicted soil response and a more detailed

study on the generation and dissipation of pore-water pressure.

The installation of a torpedo anchor into a normally consolidated clay soil was simulated. The

soil behaviour was captured by the MCC constitutive model with the parameters listed in

Table 5.1. These are typical properties of the kaolin clay used for experimental research at the

University of Western Australia (Stewart 1992).

The submerged weight of the anchor, its diameter and its length were assumed to be

W = 156kN, D = 1.0 m and L = 11m, respectively. The FE mesh contained 3,433 triangular

elements and 7,079 nodes with a mesh size of 0.167D next to the anchor shaft.

140

Parameter

Value

Friction angle

= 23

= 0.205

= 0.044

e0 = 2.14

OCR = 1

Poissons ratio

= 0.3

sat = 17 kN/m3

w = 10 kN/m3

Permeability of soil

k = 510-9 m/s

The boundary conditions and geometry of the torpedo anchor were similar to those adopted in

the previous example (see Figure 5.2), except that the torpedo length and diameter were 11 m

and 1.0m, respectively. The coefficient of friction at the interface was 0.20, and a penalty

parameter equal to 1105 was used to enforce the contact constraints utilising the mortar

contact algorithm. The analysis process was similar to the previous example and included

establishing the body force, updating the location of the yield surface based on the initial

effective stresses and OCR = 1, applying the absorbing boundary conditions and deploying

the torpedo with impact velocity of 4.7 m/s. Note that no surcharge was applied on the soil

surface. Finally, the setup analysis was carried out. The results of this analysis are presented

in the following sections.

Figure 5.11 plots the predicted total dynamic soil resistance versus penetration, normalised by

D. It was observed that the soil resistance increased linearly with embedment depth and did

not experience a rapid increase during the early stages of the penetration, as noticed in the

previous example (see Figure 5.5). This was to be expected, as no surcharge was applied in

141

this problem, whereas a uniform pressure of 50 kPa was applied in the previous example,

thus providing a non-zero undrained shear strength at the mudline.

As shown in Figure 5.11, a drop in the soil resistance occurred at a penetration depth of

greater than 12D, where the full length of the anchor shaft had embedded into the soil layer

and closure of the created pathway had begun (see Figure 5.13(e)). As a result, the sum of the

resisting forces acing on the torpedo decreased because of the contribution of the buoyant soil

weight falling behind the anchor. As the anchor penetrated further, the resistance rose again.

400

350

300

250

200

150

100

50

0

0

10

12

14

Penetration / D

The displaced volume of soil multiplied by the submerged unit weight of soil is termed the

soil buoyancy, which acts as a resisting force against the anchor penetration. The soil

buoyancy increases with depth where no hole closure is assumed. Where hole closure occurs,

soil buoyancy increases until the anchor has fully penetrated the soil and then remains

constant with depth.

Figure 5.12 depicts the process of anchor penetration and the gradual closure of its pathway.

As the anchor impacted the soil, some heave was created around the anchor and, because of

the very low shear strength of the soil at the mudline, the heaved soil was also displaced

laterally (see Figure 5.12(b),(c) and Figure 5.13(b)).

142

(a)

(b)

(c)

(d)

(e)

(f)

Figure 5.12: Deformed meshes during the free-falling process and gradual closure of the

pathway

143

(a)

(b)

displacement vector plot

This form of deformation generated suction pore pressure at shallower depths within the soil

adjacent to the anchor shaft, as shown in Figure 5.13(a). At greater depthsmore than about

4Dthe soil was displaced predominantly outwards in the radial direction to accommodate

pile in the soil.

Figure 5.12(e) shows the occurrence of the pathway closure at penetrations greater than

12.0D. It was also observed that the soil heave gradually disappeared as the anchor penetrated

deeper, and a greater volume of soil was required to fill the hole behind the anchor. It is

worth noting that, for the adopted impact velocity, soil separation did not appear to occur

over the full length of the anchor.

The acceleration and deceleration characteristics of the torpedo were investigated by studying

its velocity profile during embedment. Figure 5.14 presents the variations of torpedo velocity

versus the normalised penetration. From this figure, it was observed that the torpedo

accelerated considerably at penetrations up to ~7D, where it ultimately reached a maximum

velocity of 9.4m/s; that is, two times the anchors impact velocity. Afterwards, the torpedo

144

started to decelerate at a linear rate, as depicted in Figure 5.15. The torpedo finally came to

rest at a penetration depth of ~13.8D within a period of 2.107 s.

Velocity (m/s)

0

10

0

2

Penetration / D

4

6

8

10

12

14

16

The laboratory tests reported by OLoughlin et al. (2013) indicated that the net increase in

velocity of model DPAs during penetration is reduced (for a given anchor) as the impact

velocity increases. For instance, their test with a zero-impact velocity showed that velocity

increases from 0 m/s at a clay surface to a maximum velocity of ~6.5 m/s at ~0.5 anchor

lengths. This observation implies that the closer the impact velocity approaches to the

anchors terminal velocity, the less acceleration will occur within the soil. As soil strength

increases with depth, at some depth, the shear resistance outweighs the submerged weight of

the anchor, and the anchor decelerates.

To study the generation of excess pore-water pressure, the evolution of excess pore pressures

was monitored at some evaluation points located at depths of 6.67D, 13.4D and 14.0D.

Figure 5.16, Figure 5.17 and Figure 5.18 plot the excess pore pressures developed at these

points versus time, noting again that the total installation time was 2.107 s.

145

Time (s)

0.00

0

0.50

1.00

1.50

2.00

2.50

Velocity (m/s)

2

4

6

8

10

Figure 5.16 represents the typical form of pore-water pressure generation at points along the

shaft of the anchor (points with levels above the transition point between the conical tip and

the cylindrical shaft). As the tip of the penetrometer was above or at the level of the

evaluation point in the soil, both the mean and deviatoric stresses increased, which resulted in

rapid increases in the pore pressure. Once the tip had moved below the evaluation point, both

stresses declined and a sharp drop occurred in the magnitude of the pore-water pressure. A

steady state was then attained for the rest of the penetration process. This typical form of

pore-water pressure evolution was also observed in Section 5.3.3.

120

100

r=0.17D

80

r=0.67D

r=1.00D

60

r=1.33D

40

r=1.67D

20

r=2.00D

r=4.00D

0

-20

0.00

0.25

0.50

0.75

1.00

1.25

Time (s)

1.50

1.75

2.00

2.25

2.50

installation phase

146

250

r=0.00

200

150

r=0.67D

100

r=1.00D

r=1.33D

r=1.67D

r=2.00D

50

r=4.00D

0

-50

0.00

0.25

0.50

0.75

1.00

1.25

Time (s)

1.50

1.75

2.00

2.25

2.50

Figure 5.17: Excess pore-water pressure evolution at a depth of 13.4D throughout the

installation phase

140

r=0.00

r=0.67D

r=1.00D

r=1.33D

120

100

80

r=1.67D

60

r=2.00D

40

20

r=4.00D

0

-20

0.00

0.25

0.50

0.75

1.00

1.25

Time (s)

1.50

1.75

2.00

2.25

2.50

Figure 5.18: Excess pore-water pressure evolution at a depth of 14D throughout the

installation phase

In this example, the anchor tip reached an embedment depth of 13.8D at the end of the

installation, so all evaluation points located at levels 13.4D (see Figure 5.17) and 14.0D (see

Figure 5.18) should have only experienced the sharp increase in the pore pressure as the pile

tip reached the level of points or their close vicinity. However, the anchor shaft did not reach

147

the level of the points; thus, the pore pressures remained relatively constant until

consolidation in the soil occurred. Another observation from Figure 5.16, Figure 5.17 and

Figure 5.18 was that all evaluation points experienced a suction pore-pressure change before

the rise of compressive pore pressures. This was due to the development of a tensile region

underneath the anchor tip, which advanced along with, and ahead of, the pile tip. This is

better visualised by some contour plots, which are presented below.

To visualise the size and shape of the region affected by anchor installation, some contour

plots of excess pore-water pressure are presented in Figure 5.19. These contour plots (see

Figures 5.19(a)(d)) correspond to the deformed meshes shown in Figure 5.12(b), (c), (e) and

(f), respectively. Figure 5.19 shows the gradual generation of a compressive excess porewater pressure bulb around the anchor tip during the penetration. A tensile (suction porepressure change) region was created at the soil surface and shallower depths adjacent to the

anchor shaft (see also Figure 5.13(a)). As the anchor advanced in the soil layer, the

compressive excess pore-pressure bulb grew around the torpedo pile and extended in the

radial and vertical directions. The maximum compressive values were developed at the

anchor tip and extended to its shoulder. Moreover, a tensile region was located at a distance

of about 2D vertically underneath the anchor tip, and it advanced downwards along with, and

ahead of, the anchor. When closure occurred behind the shaft (see Figure 5.19(c)), a sudden

and local increase in the pore pressure was observed in the zone of contact between the

falling soil mass and the anchors rigid tail. However, this sudden and intense increase in the

pore-water pressure quickly disappeared as the pore fluid flowed to the surrounding soil with

comparatively much lower pore pressures. At the end of the installation (see Figure 5.19(d)),

the region affected by the torpedo installation had an extent of about 4.0D in the radial

direction (almost over the full length of the shaft and with pore pressures larger than 10 kPa)

and around 2D in the vertical direction, as measured from the anchor tip. Although the impact

velocity in this example was low, the affected region was larger in extent (in both directions)

compared to the previous example with an impact velocity of 15 m/s (see Figure 5.8). The

increase in the zone of influence may result from lower permeability, which is half of that

assumed in the previous example, and larger pipe diameter. Further, the rigidity index for this

example was Ir = 73, whereas for the first problem it was Ir = 37. However, it should be noted

that associating G with the current in situ stress state resulted in a low, unrealistic value of G,

hence low values of the ratio G/su at higher values of OCR (Zytinsky et al. 1987).

148

(a)

(b)

(c)

Figure 5.19: Excess pore-water pressure contours throughout the penetration process

149

The rate of dissipation of excess pore pressures following torpedo anchor installation was

studied in Section 5.3.4. In this section, further discussions are provided making use of some

more dissipation curves and contour plots.

After the full embedment depth was achieved and the torpedo came to rest, the computational

process automatically proceeded to the setup analysis, and consolidation of the soil was

analysed. This process was continued until the generated excess pore-water pressures

dissipated entirely. It was necessary to use very small time steps at the beginning of the setup

analysis, where high pore-water pressure gradients occurred. The time step size was then

gradually increased as the solution became smoother. Typically, the size of the time steps in

the consolidation phase can be up to 108 times larger than the size of the increments in the

installation phase. For instance, 84,307 equal time increments with a size of 0.2510-4 s were

used for the installation phase of this problem, whereas the time step size at the last stages of

the consolidation process was 10,000 s, being 108 times larger.

Figure 5.20 depicts the dissipation curves for the soil elements located at a depth of 13.4D

and different radial distances measured from the pile tip. These dissipation curves correspond

to the pore-pressure evolution curves shown in Figure 5.17. The pore pressure close to the

pile cone initially fell off rapidly. It was observed that a degree of consolidation of 90 per

cent for elements within 1.0D radial distance from the torpedo shaft was achieved 51~82

days after installation, whereas the same degree of consolidation took place within 82~121

days for elements between 1.0D and 2.0D. Figure 5.17 also shows that a degree of

consolidation of ~99 per cent was attained for the entire affected zone at a depth of 13.4D

after ~515 days. The dissipation of excess pore pressures with time is further illustrated by

some contour plots in the following.

150

250

r=0.00

200

r=0.17D

r=0.67D

150

r=1.00D

r=1.33D

100

r=1.67D

r=2.00D

50

r=4.00D

0

1.E-04

1.E-03

1.E-02

1.E-01

1.E+00

1.E+01

1.E+02

1.E+03

Time (day)

Figure 5.20: Excess pore-water pressure dissipation versus time for elements at depth

13.4D

Figure 5.21 depicts some excess pore-pressure contour plots throughout the consolidation

process. Pore-pressure dissipations ensued immediately around the pile tip and cone, in which

a degree of 30 per cent consolidation took place after nine hours (see Figure 5.20). The flow

of pore water from locations with higher values of pore pressures towards locations with

lower magnitudes diminished the concentration of excess pore-pressure contours. This was

followed by an increase in the size of the contours in both directions, particularly in the radial

direction. It was observed that the extent of the pore-pressure contour with a magnitude of

10 kPa increased from a radial distance of ~4.5D to ~6.5D within about 7.3 days. It also grew

~1D in the vertical direction towards to the tensile region located underneath the pipe tip and

ultimately diminished the entire suction zone. Afterwards, dissipations occurred

predominantly in the radial direction up to ~ t = 28.4 days. The 10 kPa contour then

contracted both in width and height (decreasing from the top) because of the combined action

of radial and vertical pore-pressure dissipations in the soil.

151

Elapsed time: t = 0

t = 13.6 days

t = 9 hours

t = 1.7 days

t = 28.4 days

t = 71 days

t = 7.3 days

t = 148 days

Figure 5.21: Excess pore-water pressure dissipation at different times after installation

152

5.5 Summary

This chapter presented one of the important applications of the computational procedure

developed in this thesis. Numerical analysis of DPAs was detailed by modelling torpedo

anchors free-falling into saturated soil layers.

The first part of this chapter provided a brief literature review of the computational methods

and available model tests on DPAs. A survey of the literature revealed that despite the

increasing relevance of DPAs in offshore applications, the estimation of embedment depth,

pull-out capacity and the prediction of stresses in their structure remains a significant

challenge. The problem still remains as to how pore pressures and stresses are affected by the

installation of DPAs. To address these limitations, two numerical analyses were conducted in

the second and third part of this chapter.

The second part of this chapter employed the numerical scheme to simulate the free-falling

penetration of a rigid, finless torpedo anchor into a lightly over-consolidated clay soil. The

aim of the analysis was to study the total penetration depth of the anchor, mobilised soil

resistance, deceleration characteristics of the anchor during its installation phase, generation

of excess pore-water pressures in the surrounding soil and its subsequent dissipation. A

comparison was also made between the results predicted by the node-to-segment (NTS) and

the mortar algorithms. The analysis was conducted for three different values of the impact

velocity: 10, 15 and 20 m/s. A significant improvement was observed in the numerical

oscillation when the mortar algorithm was used instead of the NTS scheme (see Figure 5.4).

The analysis result suggested that the embedment depth of the anchor depends mainly on its

initial kinetic energy. Further, it was observed that the anchor may actually continue to

accelerate under the action of gravity during the early stages of penetration, but it must

eventually start to decelerate and ultimately come to rest, largely because of the finite

shearing resistance of the seabed soil. The deceleration occurred at an approximately linear

rate, in which the rate of deceleration increased for higher-impact velocities. The numerical

results showed that for the lightly over-consolidated soil simulated by the Modified Cam

Clay (MCC) material model, excess pore-water pressures were generated in the soil

surrounding the pile tip and its shaft. The magnitude of excess pore-water pressure first

increased when the pile tip was above or at the level of the point of interest in the soil, and

153

then decreased once the pile tip had moved below the evaluation point, finally approaching a

steady value at the end of the installation phase. It was also found that the installation of pile

can significantly influence a region ~3.0D in the radial direction and ~1.0D in the vertical

direction, measured from the pile tip. In this example, the pathway of the anchor was left

continuously open from the top of the anchor up to the crater at the soil surface with a

diameter of less than the anchor diameter. The setup analysis indicated that the pore pressure

close to the pile initially fell off very rapidly. It was shown that most of the pull-out capacity

of the torpedo anchor (soil resistance) can be available much earlier than the completion of

consolidation, as most of the excess pore pressure dissipates within a matter of days or

weeks.

The third part of this chapter presented a detailed analysis of a torpedo anchor free-falling

into a normally consolidated clay. The torpedo had a larger diameter and mass, but a lower

impact velocity than that assumed in the first example. Due to the lower shear strength of the

soil at the mudline (no surcharge was applied on the soil surface), some heave was created

and the soil displaced laterally upon the impact of the anchor. This form of deformation

caused suction pore pressure in shallower depths within the soil adjacent to the anchor shaft

(see Figure 5.13(a)). The closure of the anchors pathway was also observed for this example

and modelled successfully. Due to the low-impact velocity (far less than the anchors

terminal velocity) and relatively high weight of the anchor, the torpedo accelerated

considerably and then started to decelerate at penetrations larger than ~7.0D. When closure

occurred behind the shaft (see Figure 5.19(c)), a sudden and local increase in the pore

pressure was observed in the zone of contact between the falling soil mass and the anchors

rigid tail, but it was rapidly dissipated as the pore water quickly moved outwards from the

vicinity of the anchor. The results of the setup analysis indicated that pore-pressure

dissipation proceeded immediately after installation for elements located around the pile tip

and cone, in which a degree of 30 per cent consolidation took place within nine hours (see

Figure 5.20). Three stages were identified during the consolidation process: (1) initial and

fast dissipations occurred around the pile tip and cone, followed by an increase in the extent

of compressive excess pore pressure, both in the radial and vertical directions within ~7.3

days of installation; (2) afterwards, dissipations occurred predominantly in the radial

direction up to ~27.4 days; and (3) the rest of the consolidation process occurred with a

decreased rate in the radial and vertical directions.

154

It should be added that the analyses conducted in this chapter did not consider the effect of

the strain rate on the undrained shear strength, although it is well known that it can

significantly affect the soil resistance. This was not included in the analysis because the

undrained shear strength was not a direct input parameter of the MCC model. However, this

may be easily considered by adjusting the size of the yield surfaces based on the updated

undrained shear strength parameters at the end of each increment. Such analyses were not

followed in the study in any detail, but they may be the subject of future numerical studies.

155

6.1 Introduction

In the deep-water industry, offshore pipelines are used to transport products to shore or fieldprocessing facilities. They are usually installed by laying them from a vessel (Figure 6.1),

typically using an S-lay or J-lay configuration (e.g., Jensen 2010). The motion of the lay

vessel and any hydrodynamic action on the hanging span will cause the pipe to move

dynamically. The pipe embedment is essentially difficult to predict, largely because of the

dynamic effects involved in the lay process. In addition, the laying procedure imposes an

undrained condition on the seabed soil, as the soil is impermeable relative to the rate at which

the lay procedure occurs. Consequently, excess pore pressures are generated, and their

dissipations with time involve consolidation settlements. Pipelines are generally operated at

high temperature and pressure to ease the flow of hydrocarbons and prevent solidification of

the materials. Accordingly, axial compressive stresses are generated due to thermal expansion

in the pipe, which may lead to lateral buckling. To relieve the axial stresses, buckles are

allowed to form; they are engineered such that no excessive bending takes place. The induced

lateral pipe movement within an engineered buckle is several diameters. The changing

operating conditions and shutdowns result in cycles of expansion and contraction that

produce cyclic lateral movements of the partially embedded pipelines. This form of

movement generates soil berms, which grow ahead of the laterally sweeping pipe.

Design practice for pipeseabed interaction in soft soil typically assumes undrained

behaviour throughout the pipe-laying process and subsequent pipe operation. In reality, the

generation of excess pore pressures around a partially embedded pipe, together with their

subsequent dissipation, may have a significant effect on the vertical penetration and

horizontal breakout resistance of the pipe.

Therefore, a reliable geotechnical analysis of a pipeline must simulate its embedment under

dynamic motion during the laying process, while a suitable soil model is also required to

capture the shear-induced pore pressure.

156

Hanging

span

www.theengineer.co.uk)

Although a few experimental studies (e.g., Cheuk et al. 2007; Dingle et al. 2008) have

provided insights into the mechanisms involved in pipesoil interaction under largeamplitude cyclic movements, a better understanding of the soilpipe response requires robust

and realistic numerical modelling. Such numerical simulation, for instance, should consider

the drainage conditions around the pipe as it moves because it can greatly affect the pipesoil

forces. Further, the analysis involves large deformations and significant soilstructure

interaction. In most analyses conducted to date, the pipesoil interface has been assumed to

be fully smooth or fully rough in order to avoid numerical difficulties arising from frictional

forces developed at the soilpipe interface.

The aim of this chapter is to utilise the developed computational scheme in order to study a

few pipelineseabed interaction problems and demonstrate an important application of the

proposed numerical method in geomechanics. The first analysis considers the dynamic

installation of a pipe on a seabed soil and studies the soil resistance and excess pore-water

pressure generation. The NTS and mortar algorithms are employed, and the predicted results

are compared. The second example studies another pipeline problem and gives some

highlights on the effects of inertia forces. Finally, the soilpipe interaction during largeamplitude cyclic movements is investigated.

157

The process of laying a rigid pipe on deformable soft seabed soil was simulated in order to

evaluate the excess pore-water pressures induced by the installation process and the total

penetration resistance at the pipesoil interface. A 2D FE mesh containing 2,246 plane strain

triangular elements and 1,068 nodal points was employed for the simulation, as depicted in

Figure 6.2. The rigid pipe was modelled using 24 quadratic triangular elements, producing a

7D

6D

The MCC constitutive model was adopted to represent the soil in the study using the model

parameters listed in Table6.1. The pipe-soil interface was assumed to be frictionless.

The analysis was started by establishing the initial stresses in the soil domain due to the soil

self-weight and an overburden pressure of 50 kPa applied on the top boundary. The initial

stresses were generated assuming the soil to be one-dimensionally (K0) consolidated with

K0 = 0.67, and the water table was assumed to be located at the soil surface. During this stage

of the analysis, the horizontal components of the solid movement and fluid flow were

prevented on the side boundaries. The bottom boundary was also fixed against vertical and

158

horizontal solid displacements, whereas fluid flow tangential to the boundary only was

allowed.

Parameter

Value

Friction angle

= 25

= 0.205

= 0.044

e0 = 2.14

OCR = 2

Poissons ratio

= 0.3

sat = 15 kN/m3

w = 10 kN/m3

Permeability of soil

k = 10-8 m/s

After generating a non-zero stress field, the location of the yield surface at each integration

point in the FE mesh was adjusted according to the initial effective stresses and the

designated value of the OCR. Subsequently, energy-absorbing boundaries that comprised

springs and dashpots were applied to the boundary nodes to absorb the energy of the

impacting waves and eliminate possible wave reflections. Finally, the pipe was installed with

a velocity of 0.5D/s to an embedment of 1.0D by conducting a large deformation ALE

analysis. Drainage was allowed at the top soil surface through the nodal points not in contact

with the impermeable pipe surface. However, as there is no contact between the pipe and the

soil initially, zero pore-fluid pressure (free-draining conditions) was prescribed for all nodal

points on the top boundary at the beginning of the simulation. As soon as contact was

established between the soil and the pipe, the condition of the corresponding nodal points in

contact (on the soil side only) were automatically changed from free-draining to the

impermeable condition by removing the zero boundary condition applied to the pore-fluid

pressures. In contrast, if a nodal point lost contact with the pipe during the analysis, the freedraining condition was recovered. In addition, the constraint on the normal component of the

relative Darcy velocity was enforced consistently across the interface during the entire

simulation process. Accordingly, appropriate drainage conditions around the pipe were taken

159

into account throughout the entire pipe-laying process. This particular feature of the

modelling is necessary if the pipesoil interaction analysis involves lateral movement of the

pipe. Figure 6.3 depicts the deformed mesh at an embedment of 0.5D and at the end of the

analysis (penetration of 1D), and it illustrates the quality of the optimised meshes and the

performance of the coupled ALE analysis.

(a)

(b)

Figure 6.3: Deformed mesh at embedment depths of: (a) 0.5D; (b) 1.0D

The variation of the normalised total penetration resistance with depth is depicted in

Figure 6.4. The total resistance force R was normalised using the undrained shear strength

su0 = 15 kPa at the pipe invert obtained from the assumed MCC parameters for K0consolidated soil. Ten Gauss points were used to integrate the contact contributions over

every non-mortar segment.

Figure 6.4 also shows the results of analysis conducted using the NTS procedure. As shown,

the mortar approach provided a more continuous and stable response compared to the NTS

method. There were some jumps in the penetration resistance predicted by the NTS scheme

for embedment depths greater than about 0.55D, together with pronounced oscillatory

responses at penetration depths larger than about 0.80D.

160

R / Dsu0

0

0

0.1

0.2

Embedment / D

0.3

0.4

0.5

Mortar method

NTS method

0.6

0.7

0.8

0.9

1

Figure 6.5 depicts the evolution of excess pore-water pressure at the pipe invert throughout

the penetration process, normalised by q = R/D. A sudden increase in the pore-water pressure

occurred when initial contact was established between the pipe and the soil. As the pipe was

embedded deeper and the contact area increased, the rate of the excess pore pressure

evolution decreased. Figure 6.5 also shows the difference between the results obtained using

the NTS and mortar methods. Similar to the predicted soil resistance, as depicted in

Figure 6.4, the excess pore-water pressure response calculated by the NTS method

experienced some jumps and instability during the course of the embedment.

To visualise the extent of the affected zone, Figure 6.6 shows a contour plot of the excess

pore-water pressure at the end of the analysis.

161

1.6

1.2

Mortar method

NTS method

0.8

0.4

0

0.0

0.2

0.4

0.6

0.8

1.0

Embedment / D

Figure 6.5: Normalised excess pore pressure versus normalised embedment at the pipe

invert

According to Figure 6.6, the maximum excess pore-water pressure value was observed at a

point in contact with the pipe invert. Pore-water pressures larger than 2su0 could be seen up to

2.5D under the pipe invert. Moreover, a limited suction zone was generated near the ground

surface.

Figure 6.6: Excess pore-pressure contours at the end of the dynamic pipe embedment

162

Settlements

The process of rapidly laying an elastic pipe on a deformable seabed, and the subsequent

consolidation phase of the soil, were simulated. The adopted FE mesh, boundary conditions

and material parameters were identical to the previous example, except that OCR = 1 was

assumed in this analysis. Further, the coefficient of friction at the pipesoil interface was

taken as 0.2. The analysis started by establishing the initial stresses in the soil due to its selfweight and an effective overburden pressure of 30 kPa. The initial stresses were generated

assuming the soil to be 1D consolidated, with K0 = 0.58 and the water table located at the soil

surface. Subsequently, energy-absorbing boundaries were applied to the boundary nodes of

the FE mesh. Finally, the pipe was laid on the soil by applying a uniform pressure q = 6.6su0

to it over the dimensionless time increment of Tv = 0.510-6, where su0 = 9kPa is the

undrained shear strength at the pipe invert, obtained from the assumed MCC parameters for

K0-consolidated soil. The dimensionless time Tv is defined as:

Tv =

cv t

D2

6.1

where t represents the actual time and cv denotes the coefficient of consolidation given by:

cv =

k

mv w

6.2

where k is the soil permeability, mv is the volume compressibility and w denotes the unit

weight of water. The virgin compressibility in the Cam Clay model can be expressed as:

mv =

(1 + e0 ) p0

6.3

To evaluate the effects of inertia forces on the predicted pipesoil response, two different

analyses were conducted. In the first analysis (dynamic solution), the total pressure q was

applied at a uniform rate of 6.6su0 per second for a period of t = 1s, whereas the loading rate

for the other analysis (static solution) was 0.066su0 per second for a period of t = 100s. The

soil permeability was decreased to k = 10-10m/s for the slow-rate analysis to keep Tv the

163

same for both simulations. Accordingly, the analysis with the slow rate effectively provided

the static undrained solution, as the inertia effects were negligible during the loading process.

Throughout the embedment process and the subsequent consolidation stage, drainage was

allowed at the soil surface through the nodal points not in contact with the impermeable pipe

surface. The adjustment of the drainage condition around the pipe was carried out

automatically during the analysis, as described in the previous example.

Figure 6.7 depicts the evolution of excess pore-water pressure p at the pipe invert, normalised

by q, throughout the penetration process.

1.25

1

Dynamic solution

p/q

0.75

Static solution

0.5

0.25

0

0

0.1

0.2

0.3

0.4

Tv

0.5

x 10-6

Figure 6.7 shows that the excess pore pressure at the pipe invert was almost the same for both

analyses (i.e., it was not significantly affected by the inertia forces). To visualise the entire

affected zone, Figure 6.8 shows a contour plot of excess pore-water pressure, corresponding

to a time factor Tv = 0.610-6 for the dynamic analysis. Pore-water pressures larger than 0.2q

were observed in a vertical interval up to ~3D under the pipe invert.

The axial and lateral resistance of the pipeline are significantly affected by the degree of

consolidation following installation. The dissipation of compressive excess pore-water

pressures results in an increase in the shear strength of the soil near the pipe and alters the

strength distribution around it, leading to an increase in the breakout resistance.

164

When the loading process was completed, the analysis proceeded to the consolidation stage.

Figure 6.9 depicts the time history of the normalised excess pore-water pressure at the pipe

invert, indicating that the developed excess pore-water pressure had entirely dissipated when

Tv = 10.

Figure 6.10 depicts excess pore-pressure contours at different times of the dissipation

process. It was observed that the dissipations took place from the pipe invert extending to the

entire pipesoil interface and then through the upper boundary.

1.25

1

p/q

0.75

Dyamic solution

0.5

Static solution

0.25

0

1E-6

1E-5

1E-4

1E-3

1E-2

1E-1

Tv

165

(a) Tv = 0.610-6

(b) Tv = 0.05

(c) Tv = 0.1

(d) Tv = 0.5

Figure 6.10: Excess pore-water pressure contours at different times of consolidation for

dynamic analysis

The curved shape of the pipe led to a different initial distribution of excess pore pressure

because the distribution of contact stress was not uniform around the pipe. As a result, the

initial pore-pressure distribution created flow away from the pipe invert, which was clearly

detected by observing the pore-fluid flow, as depicted in Figure 6.11. This flow pattern could

166

foundations, such as the one shown in Figure 4.11, higher pore pressure at the edges tends to

cause inward flow initially, which delays the dissipation of excess pore pressure at the centreline.

m/s

Tv = 0.065

Tv = 0.505

Figure 6.12 shows the variation of the normalised settlement of the pipe during the entire

analysis. There was a marked difference between the static and the dynamic solutions. For the

static analysis, the pipe embedment at the end of the installation (Tv = 0.510-6) was 0.41D,

and it did not increase until the consolidation settlement began (Tv ~ 0.1), whereas in the

dynamic analysis, the pipe continued to penetrate after the loading stage up to Tv = 0.610-6.

Although the consolidation settlement for both analyses started and ended almost at the same

dimensionless time, the pipe embedment increase due to the dissipation of excess pore

pressures was larger for the dynamic case (see Figure 6.12).

Therefore, it appeared that a dynamic approach was necessary for coupled problems of pipe

seabed interaction involving very fast loading for which static coupled solutions might not be

appropriate. This conclusion is in agreement with practical observations of the as-laid

pipeline embedment, which is typically much greater than would be expected from the static

pipeline weight. Although the increased penetration could be attributed to different

mechanisms, three major mechanisms here include a stress concentration at the pipe invert

(touchdown point), induced inertia forces and subsequent potentiality for partial liquefaction

of the soil under the pipeline.

167

0

0.1

Dynamic solution

Embedment/D

0.2

Static solution

0.3

0.4

0.5

0.6

0.7

0.8

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2

Tv

Other possible causes of the increased embedment in practice are an additional vertical force

near the touchdown point resulting from catenary effects, as well as remoulding or

displacement of the soil resulting from small-amplitude cyclic movements of the pipeline

throughout the laying process, which may cause significant self-burial. However, the last two

phenomena were not considered in the analysis described here. It is worth noting that cyclic

movements may remould and soften the surrounding soil, and such disturbance can lead to

significant changes in the operative shear strength and the basic constitutive properties of the

soil. The use of an appropriate soil constitutive model is then essential to account for the

combined effects of remoulding and reconsolidation during the small-amplitude cyclic

movements.

Controlled on-bottom lateral buckling of partially embedded pipelines is a novel and costeffective solution to relieve the developed axial compressive stresses resulting from cycles of

thermal expansion and the contraction of operative pipelines. Harrison et al. (2003), Nystrom

et al. (2002) and Kaye et al. (1995) reported using designed lateral buckles in some case

studies. The main uncertainty in the buckling design is the soil resistance faced by the pipe

during lateral movement. To properly engineer lateral buckles, it is necessary to make

168

accurate predictions of the initial as-laid pipe embedment and the subsequent soilpipe

response because of the combined loading from the thermal expansion and pipe self-weight.

The conventional approach for modelling the lateral response of a pipe on soft clay has been

based on a frictional behaviour. In this approach, the pipesoil contact is simulated using a

frictional springslider system, such that the lateral resistance consists of two components: a

frictional part, which is linearly related to the pipe weight via a friction factor; and a passive

component associated with the pipe embedment. Verlay and Lund (1995) suggested such an

empirical relation to calculate the limiting horizontal force on the pipe, taking into account

the weight and predicted as-laid embedment. Bruton et al. (2006) then updated the model

proposed by Verlay and Lund (1995) based on a large database of experimental results and

offered new relations to predict breakout resistance and subsequent steady-state lateral

resistance at large displacements. Cheuk et al. (2007) conducted a series of large-scale plane

strain model tests to study soilpipe interaction during large-amplitude cyclic movements.

They identified four key stages involved in the force-displacement response of the system:

initial breakout of the pipe; suction release, which causes a sudden drop in soil resistance

upon separation of the pipe from the soil behind it; resistance increase associated with the

growth of an active berm ahead of the pipe; and additional resistance achieved during the

collection of a pre-existing dormant berm. They suggested a simple upper-bound solution to

model the observed response. Dingle et al. (2008) also described mechanisms of the vertical

embedment and lateral breakout of a model pipe using centrifuge and particle image velocity

(PIV) facilities. White and Cheuk (2008) reported a geotechnical centrifuge test and

incorporated the additional berm resistance in the empirical relation that Bruton et al. (2006)

proposed for the residual lateral resistance. In their study, the growth of a berm for undrained

conditions is related to the thickness of the soil that is ploughed away by the advancing pipe

from the conservation of volume. They assumed that the ploughing depth is unaffected by the

berm size while, as the berm grows, the pipe may rise relative to the original seabed level.

Sometimes the pipe can even completely slide off the berm and come out of the trench,

particularly in the case of an over-penetrated pipe that has experienced a high vertical load

during installation compared to the operative load during bulking. Further, in the theoretical

framework established by White and Cheuk (2008), the lateral resistance after a reversal is

assumed to equal the initial response with the same value of breakout and residual

resistances, whereas the initial breakout resistance will not be mobilised, as the pipe is no

longer at the initial embedment. However, the assumption is justified by the fact that a brittle

response after reversal is shown by experiments, as suction is mobilised on the back face of

169

the pipe. Nevertheless, the changing geometry is not the only factor affecting cyclic

resistance. Pore-pressure dissipations may occur during lateral sweeping and also during the

period of start-up and shutdown events. This leads to reconsolidation of the disturbed soil

within the berm and can significantly increase berm resistance.

Therefore, pipesoil interaction during lateral buckling involves complex changes in seabed

geometry and likely episodes of undrained or partially drained behaviour interspersed with

consolidation periods. A large deformation FE analysis is required to capture the effects of

these factors in the response of a soilpipe system undergoing large-amplitude cyclic

movements. Such analysis has rarely been addressed in the literature, so a numerical

simulation is presented in this section to study this type of problem utilising the developed

computational scheme.

An elastic pipe of 0.8 m in diameter was first laid rapidly onto deformable seabed soil by

applying a uniform pressure load to it, and the soil was then allowed to consolidate. When the

consolidation process was completed and all generated excess pore pressures were dissipated,

the vertical force was reduced to a specific value. Finally, the pipe was swept back and forth

across the model (at a constant velocity) while the vertical load remained constant and the

pipe was free to move up or down.

As depicted in Figure 6.13, a 2D FE mesh containing 3,240 plane strain triangular elements

and 6,695 nodal points was employed for the simulation. The elastic pipe was modelled using

24 quadratic triangular elements, producing a smooth continuous surface that was considered

the mortar surface. The MCC constitutive model was also adopted for this problem to

represent the soil in the study using the model parameters listed in Table 6-1. Further, the

coefficient of friction at the pipesoil interface, , was taken as 0.1. The analysis started by

establishing the initial stresses in the soil due to its self-weight and an effective overburden

pressure of 75 kPa. The initial stresses were generated assuming the soil to be 1D (K0)

consolidated with K0 = 0.67, and the water table was assumed to be located at the soil

surface. During this stage of the analysis, the horizontal components of the solid movement

and fluid flow were prevented on the side boundaries. The bottom boundary was also fixed

170

against vertical and horizontal solid displacements, whereas fluid flow tangential to the

boundary only was allowed. Drainage was allowed at the top soil surface through the nodal

points not in contact with the impermeable pipe surface. Finally, the pipe was laid on the soil

by applying uniform pressure q = 4.75su0 to it over the dimensionless time increment of

Tv = 1.410-6, where su0 = 38.8 kPa.

5.5D

12.0D

Figure 6.14 depicts the variation of the normalised settlement of the pipe during the loading

process and subsequent consolidation stage. From Figure 6.14, the pipe embedment at the end

of the loading (Tv = 1.410-6) was 0.35D; however, the pipe penetration continued to increase

up to Tv = 1.5410-6, when it came to rest at an embedment depth of 0.41D. Afterwards, the

consolidation settlement occurred between Tv = 1.210-2 and Tv = 10, leading to a total

embedment depth of 0.5D at the end of the consolidation stage, which is typical of the

embedment observed for on-bottom pipelines on clay.

171

Embedment/D

0.1

0.2

0.3

0.4

0.5

0.6

1E-8 1E-7 1E-6 1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2

Tv

Figure 6.15 depicts the contours of the excess pore-water pressure field at different time

factors during the embedment process (see Figure 6.15(a)(d)) and the consolidation stage

(see Figure 6.15(e)(g)). It was evident from these contour plots that the maximum pore

pressure was observed at a point below the pipe invert. The rate of excess pore-pressure

development at the pipe invert was similar to that shown in the previous example (see

Figure 6.7), where it was higher at the initial stages because of the small contact area and

high stress concentration. As the pipe embedded deeper into the soil, the pore-pressure

contours grew in size and shape (see Figure 6.15). The extent of the compressive porepressure contours indicated the amount of soil undergoing compression and shearing because

of the pipe loading. The shape of the contours was analogous to the curved surface geometry

of the pipe. A small zone of suction could be seen advancing downwards as the pipe was

embedded deeper. This suction zone slightly altered the shape of the contours at some stages

(see Figure 6.15(a),(b)).

172

Tv = 0.4510-6

(a)

Tv = 0.5910-6

(b)

Tv = 1.410-6

(c)

Tv = 1.5410-6

(d)

173

Tv = 8.1310-3

(e)

Tv = 4.0710-2

(f)

Tv = 0.36

(g)

Figure 6.15: Excess pore-water pressure contour plots during the loading and

consolidation stages

174

The displaced soil during pipe embedment caused a gradual formation of soil heave at the

two sides of the pipe, leading to pipesoil contact over more of the pipe perimeter when

compared with a whished-in-place pipe. The enlarged contact surface increases the resistance

to both axial and lateral movement. Further, the level of thermal insulation provided by the

soil would increase because of the reduced exposure of the pipe to the free water circulating

around the pipe. The excess pore pressures at and around the heave may affect pipe stability

and, in particular, its lateral buckling, as elevated pore pressures could soften heaved soil,

thereby possibly providing additional resistance to a buckling pipe. However, the time delay

for the dissipation of pore pressures around the heave would be relatively small compared to

the pipe invert, as it is closer to the free surface.

Figure 6.15(e)(g) show the gradual pore-pressure dissipation and associated consolidation

settlements. The concentration of excess pressure contours under the pipe invert diminished

with increasing time, and equilibrium was finally attained with the pipe weight and the soil

effective stresses. As dissipation occurred, the pore-pressure contours became almost parallel

to the free surface and faded.

After the pipe was fully embedded and the consolidation process had finished, the vertical

force was reduced to q = 2.7su0, which was 57 per cent of the force required for embedment,

corresponding to an over-penetration ratio of q/q = 1.76. The unloading was applied during a

period of Tv = 1.410-6, which was the same rate as adopted in the initial loading stage. This

relatively rapid unloading process resulted in some negative pore pressure (suction) in the

soil around the pipe surface. The pipe was then moved horizontally while keeping the vertical

load constant and allowing the pipe to freely move up or down. The rate of lateral

displacement was 0.625D/s, and it was applied in a rightwards direction for a period of 2 s

(sweep1). This rate of movement could be sufficiently high to invoke dynamic effects in the

analysis and effectively impose undrained conditions within the soil around the pipe. It is

notable that analyses with different rates of movement, as well as static solutions to the

problem, could be obtained as explained, for instance, in Section 6.3, but such analyses were

not followed here.

175

When the pipe movement at sweep1 was finished (the pipe invert had been displaced 1.25D

horizontally in the rightwards direction), the generated active berm was deposited and the

direction of pipe movement was reversed to pass its initial position during sweep2. The

backwards movement was continued in the leftwards direction (sweep3) to the same amount

(-1.25D) and then reversed again to the first place (sweep4). Figure 6.16 depicts the lateral

dynamic load-displacement response of the pipe over the first four sweeps. For sweep1, the

initial breakout of the pipe occurred at a lateral displacement of ~0.1D, followed by a steady

residual resistance up to the horizontal displacement of 0.60D, during which the pipe had an

upward trajectory, according to Figure 6.17, which is typical of light pipes. From this point

on, the pipe approached a relatively steady elevation, and the lateral resistance increased

gradually because of the steady growth of the active berm ahead of the pipe.

-1

Sweep1

Sweep2

-2

Sweep3

Sweep4

-3

-4

-1.5 -1.25 -1 -0.75 -0.5 -0.25

1.25 1.5

Horizontal displacement / D

Figure 6.16: Dynamic lateral resistance: 1st, 2nd, 3rd and 4th sweeps

176

0

Sweep1

0.1

Sweep2

Sweep3

0.2

Sweep4

0.3

0.4

0.5

-1.5

-1.25

-1

-0.75

-0.5

-0.25

0.25

0.5

0.75

1.25

1.5

Horizontal displacement / D

According to Figure 6.18(a), suction pore pressures were generated behind the pipe as soon

as it was displaced laterally, whereas compressive pore pressures were developed ahead of

the pipe. As the pipe moved further, the suction pore pressures behind the pipe increased and

extended to the soil within the berm ahead of the pipe (see Figure 6.18(b)). Upon reversal of

the movement direction (sweep2), the lateral resistance was released (see Figure 6.16). It is

notable that the suction behind the pipe generates a tensile force. This force resists against

the lateral movement of the pipe which was not considered in the analysis through the contact

algorithm. Its inclusion is a relatively trivial task, but it was not followed in any detail in this

thesis. Nevertheless, throughout sweep2, the pipe slid backwards over the created trench and

slightly deepened surface profile of the trench, as depicted in Figure 6.17. When the pipe

approached its initial place, it faced the pre-existing soil berm created during the embedment

process. Accordingly, a steeper increase in soil resistance was experienced by the pipe. The

pipe continued its backwards movement (sweep3) and pushed the berm ahead, in which the

second breakout occurred at the ~-0.20D lateral movement (see Figure 6.16).

177

Sweep1

(a)

(b)

Sweep2

(c)

Figure 6.18: Excess pore-pressure contours during lateral movement: (a) at breakout;

(b) at 1.25D rightwards movement (end of sweep1); (c) during backwards movement

(sweep2)

178

A similar behaviour was then repeated as observed for sweep1, such that a steady residual

resistance was followed after a breakout while the pipe slightly moved upwards, and then the

lateral resistance increased gradually because of the steady growth of the berm ahead of the

pipe. However, in sweep3, the embedment of the pipe invert decreased more than that

observed for sweep1 (see Figure 6.17). This suggests that the left berm mobilised more

resistance compared to the right berm initially faced by the pipe in sweep1. Accordingly, the

upwards movement of the pipe was larger compared to sweep1 (see Figure 6.17), in which

the pipe pushed a smaller volume of the soil and ultimately led to a decreased breakout and

residual lateral resistances. The higher passive resistance of the left berm resulted from the

existence of suction pore pressures generated during sweeps12. Figure 6.18(c) and

Figure 6.19(a) show excess pore-pressure contours as the pipe approached the left berm. The

suction pressures increased the soil resistance and prevented slip surfaces developing within

the zone of high suctions. However, if the pipe weight was enough to sustain greater

resistance without moving upwards, these suctions would result in a greater breakout

resistance.

At the extremity of sweep3, the generated active berm was deposited and the pipe was

reversed to travel the created trench through sweeps45. Figure 6.16 depicts that the lateral

resistance after the reversal was released. The lateral displacement in sweep5 was increased

to 2.25D to collect the right dormant berm as deposited in sweep1, which was at the lateral

distance of ~1.25D. As the position of the dormant berm was passed during sweep5, this

material was collected and the lateral resistance increased sharply, as depicted in Figure 6.20.

The increased berm resistance, perhaps because of the generation of high-suction pore

pressures within the berm, caused the pipe to slide over the berm and come out of the trench.

This can be observed in Figure 6.21, which depicts the pipe invert trajectory throughout

sweep5 and the last cycle. The lateral resistance faced by the pipe decreased as the pipe

moved over the berm. A similar pattern was observed for the last cycle, during which the pipe

was displaced from the right side to the left side to a horizontal displacement of -2.25D.

Figure 6.22 shows excess pore-pressure contours for sweep5. High values of compressive

pore pressures concentrated in the pipe invert were observed, and the corresponding contours

moved with the pipe as it was displaced during the sweep. The movement of the pipe also

generated some suction pore pressures behind the pipe at points that are passed over.

179

sweep3

(a)

(b)

sweep4

(c)

Figure 6.19: Excess pore-pressure contours during lateral movement: (a) sweep3; (b) at

1.25D leftwards movement (end of sweep3); (c) during forwards movement (sweep4)

180

However, they quickly disappeared from the top surface as the drainage condition

automatically changed and free draining was adopted. Further, the concentration of excess

pressure contours under the pipe invert diminished as it approached the left dormant berm

where high values of suction were generated.

6

Sweep1

Sweep3

Sweep5

Last cycle

-2

-4

-6

-2.25 -2 -1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25

Horizontal displacement / D

Figure 6.20: Dynamic lateral resistance: 1st, 3rd, 4th sweeps and last cycle

-0.3

Pipe invert embedement / D

-0.2

Sweep5

-0.1

Last cycle

0

0.1

0.2

0.3

0.4

0.5

-2.25 -2 -1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25

Horizontal displacement / D

181

sweep5

(a)

(b)

(c)

Figure 6.22: Excess pore-pressure contours during lateral movement in sweep5

182

(a) Tv = 0

(b) Tv = 6.7310-6

(c) Tv = 1.3610-2

(d) Tv = 4.0710-2

Figure 6.23: Excess pore-pressure contours during consolidation

183

In the last part of the analysis, the pipe was moved to its first position and left for a while

until all excess pore pressures dissipated. Figure 6.23 visualises the gradual dissipation of

excess pore pressures through some contour plots. Figure 6.23(a) corresponds to the start of

the consolidation process, in which the time factor was assumed to be zero, implying no time

delay after the pipe arrived at its first position. The dissipation process was accomplished at

~Tv = 0.8.

Figure 6.24 depicts the quality of the deformed mesh at the end of the dynamic simulation

and the excellent performance of the coupled ALE analysis.

6.5 Summary

The dynamic coupled analysis of pipelineseabed interaction problems was detailed in this

chapter to illustrate another important application of the developed numerical scheme. Three

examples were presented to address different features of the modelling.

The first example considered the laying process of a rigid pipe to evaluate the excess pore

pressures induced by the installation process and the total penetration resistance at the pipe

soil interface. The pipe was installed with a velocity of 0.5D/s to an embedment of 1D. A

184

comparison was made between the two contact algorithms, and it was illustrated that the

mortar algorithm can provide more stable results than the NTS scheme. Generally, a sudden

increase in the pore pressure occurred when initial contact was established between the pipe

and the soil, and then the rate of increase decreased as the pipe was embedded deeper. The

maximum excess pore-water pressure value was observed at a point in contact with the pipe

invert. Pore-water pressures larger than 2su0 were detected up to 2.5D under the pipe invert.

Moreover, a limited suction zone was generated near the ground surface.

The second example modelled the process of laying an elastic pipe rapidly on a deformable

seabed, as well as the subsequent consolidation phase of the soil. The excess pore-pressure

distribution was found to form contours with a shape analogous to the pipes curved-surface

geometry. To evaluate the effect of inertia forces on the predicted results, an equivalent static

analysis was conducted, which showed that the excess pore pressure at the pipe invert was

almost the same for both analyses (i.e., it was not significantly affected by the inertia forces).

However, the significant difference between a static coupled analysis and its equivalent

dynamic coupled analysis was found in their settlement responses immediately after the

loading stage. Unlike the static solution, the pipe continued to penetrate in the dynamic

solution for a while after the vertical loading process, resulting in a higher embedment depth.

Therefore, it was concluded that a dynamic approach was necessary for coupled problems of

pipeseabed interaction involving very fast loading, for which static coupled solutions may

not be appropriate. The results of the consolidation stage revealed that pore water essentially

moved away from the pipe invert during the dissipation process because of the initial form of

excess pore-pressure distribution. This was clearly detected by observing the direction of

pore-fluid Darcy velocity during the consolidation process (see Figure 6.11). The time history

of the normalised excess pore-water pressure at the pipe invert indicated that the developed

excess pore-water pressure entirely dissipated within a time factor of Tv = 10.

The last part of this chapter presented a literature review of the pipelineseabed interaction

problems under large-amplitude cyclic movements. Most of the available studies were found

to be experimental, and the adopted numerical solutions were mainly based on analytical

plasticity solutions. The plasticity solutions for the vertical collapse load of a shallowly

embedded pipeline may closely predict the deformation mechanisms during pipe penetration,

but they do not allow large deformation effects, such as heave, to be incorporated. In

addition, including the dynamic effects involved in the lay process is essentially difficult.

185

Further, the steady lateral response of the soilpipe system is governed by the growth of a

soil berm, which is created as the pipe rises from the initial embedment. However, the

changing geometry was not the only factor affecting the pipesoil response. Pore-pressure

dissipations may occur during lateral sweeping and for the period of start-up and shutdown

events. This gives rise to the reconsolidation of the disturbed soil within the berm and can

significantly increase the berm resistance. Therefore, there is a need for a numerical

simulation that can address the different phenomena involved. The last example in this

chapter attempted to conduct such an analysis and provided some insights into the behaviour

of soilpipe interaction under large-amplitude lateral movements.

According to the analysis results, it was observed that for the adopted pipe weight and the

MCC soil parameters, the initial break-out of the pipe occurred at a lateral displacement of

~0.1D and was followed by a steady residual resistance up to the horizontal displacement of

0.60D. During this stage, the pipe had an upward trajectory, which is typical of light pipes. At

further lateral displacements, the pipe approached a relatively steady elevation, and the lateral

resistance increased gradually because of the steady growth of the active berm ahead of the

pipe. Suction pore pressures were generated behind the pipe as soon as it was displaced

laterally, whereas compressive pore pressures were observed ahead of the pipe. Similar

behaviour was repeated when the pipe was moved backwards and faced the left berm.

However, because of the developed suction within the left berm, it showed more resistance

against the pipe and, accordingly, caused the pipe to move upwards. High values of

compressive pore pressures concentrated in the pipe invert were observed while it was

travelling the created trench during the sweeps, in which the corresponding contours were

moving with the pipe. It was also observed that the movement of the pipe generated some

suction pore pressures behind the pipe at points that were passed over. However, they quickly

disappeared from the top surface as the drainage condition automatically changed and free

draining was adopted. At further cycles of movement, as the position of the dormant berms

was passed, the lateral resistance increased sharply. The generation of high-suction pore

pressure within the deposited soil berms resulted in a higher passive resistance and ultimately

caused the pipe to slide over the berm and come out of the trench.

186

7.1 Introduction

This thesis presented the development and implementation of a computational scheme, in the

framework of the finite element method, for the analysis of coupled geotechnical problems

involving finite deformation, inertia effects and changing boundary conditions. The

numerical scheme was employed to simulate the installation and consolidation of torpedo

anchors, as well as problems of pipeline-seabed interaction. The mechanical behaviour of a

two-phase saturated porous medium was predicted using mixture theory, which models the

dynamic advection of fluids through a fully saturated porous solid matrix. High-order contact

algorithms were formulated to account for contact problems of the saturated porous medium

based on a mortar segment-to-segment scheme. An Arbitrary LagrangianEulerian (ALE)

approach was utilised to consider geometrical nonlinearities and avoid possible mesh

distortions. Suitable absorbing boundary conditions were adopted to absorb the outgoing bulk

waves and eliminate spurious wave reflections.

In Chapter 2, the field equations for two-phase porous media were derived in light of the

mixture theory extended by the concept of volume fraction. A numerical solution of

governing differential equations for saturated soils was obtained by the finite element

method. A U-P-V formulation was selected to describe both incompressible and compressible

fluids, in which the resulting mixed formulation predicted all field variables, including solid

displacement U, pore-fluid pressure P and the Darcy velocity of the pore fluid V. This

dynamic consolidation scheme provided an exact solution to the governing differential

equations considering the convective terms of fluid acceleration. A simplified solution was

also explained in the form of the U-P approximation, which ignores the acceleration of the

fluid component.

Chung and Hulberts (1993) generalised- algorithm was selected to discretise the governing

equations in the time domain. This scheme possesses some form of numerical dissipation that

187

can attenuate the inaccurate high-frequency modes. Meanwhile, it also allows accurate

capture of the low-frequency behaviour of the system so that it appears in the solution

without attenuation. This enhancement was attained by expressing different terms of the

equation of motion in an average form with different degrees of forward weighting. In this

scheme, two different forward weightings were applied on the stiffness and inertial terms

such that the inertia term was evaluated at time tn +1 m of the considered interval t, whereas

the stiffness and all other terms were evaluated at an earlier time tn +1 f , in which f and m

were two integration parameters, with f m .

A literature review was presented of some of the available boundary conditions for solving

wave-propagation problems in an unbounded domain. Based on the conclusions of this

review, a well-established absorbing boundarythe cone boundarywas adopted and

implemented in the U-P-V consolidation algorithm. These boundary conditions can be

interpreted as constitutive equations for interaction forces between the near and far fields.

The semi-infinite truncated conical rod represented a physical interpretation of the considered

model. The missing part of the linear cones from a truncated boundary was then modelled by

a mechanical system containing two series of springs and dashpots oriented normal and

tangential to the boundary.

An ALE operator split scheme, which attempted to combine the advantages of the Lagrangian

and Eulerian approaches, was incorporated to consider the effects of finite deformations and

avoid possible mesh distortions. In this method, the computational grid was not coincident

with the material, nor was it fixed in space. Rather, it could move arbitrarily to avoid possible

mesh distortions. The ALE operator split technique and the mesh refinement strategy

incorporated in this thesis were based on the method presented by Nazem et al. (2009).

Closed-form solutions were also developed for some one-dimensional problems involving the

dynamic response of saturated porous media. These solutions were useful for validating FE

codes for the dynamic consolidation of soil. While they only considered elasticity and small

strains, they allowed a check on the concurrent wave transmission and consolidation process.

188

In Chapter 3, a new contact algorithm was formulated and implemented for solid-fluid

mixtures in the spatial frame that can accommodate inertia effects together with finite

deformation and contact sliding. Both frictionless and frictional contact formulations were

addressed. For a two-phase saturated soil, in addition to the requirement for continuity of the

contact traction, continuity also had to be maintained for the Darcy velocity and the porefluid pressure across the contact interface via the enforcement of appropriate constraints.

Therefore, for a node that was in contact with a corresponding contacting pair, contact

contributions arising from constraining solid displacement, Darcys velocity and pore

pressure were added to the tangent stiffness matrix and the residual vector during the global

Newton iterations. In a frictional contact element, two conditionsstick and slipwere

distinguished on the basis of the level of interface frictional force compared with the

Coulomb frictional force. The formulation of frictional contact was developed in terms of

effective forces. To differentiate between the stick and slip cases, the concept of a moving

friction cone (Wriggers and Haraldsson 2003) was used, which was a relatively efficient

methodology in deriving the contact kinematics. The penalty method regularised with an

augmented Lagrangian approach was employed to enforce the necessary contact constraints.

The formulation of the contact kinematics and constraints adopted in this thesis was based on

the so-called mortar segment-to-segment approach. For two-phase saturated consolidation

problems, the interpolation function for the pore-fluid pressure was generally chosen to be

one order lower than the functions for the displacement and fluid velocity. As a result, the

mortar method was adopted, as high-order approximation functions could then be used to

interpolate different field variables. Moreover, a consistent coupling of the node-to-segment

(NTS) elements with elements of a higher order was not possible, whereas the mortar method

could appropriately overcome this problem because the contact constraints were fulfilled in a

weak integral form. The NTS element assumed constant contact stresses around each slave

node, while the mortar-type element considered a quadratic approximation of the stress

vector. In the developed coupled consolidation-contact algorithm, free-draining conditions

were automatically adopted for the nodes that lost the connection with their possible

contacting pair, whereas impermeable or semi-impermeable conditions were adopted for

nodes that came into contact with another surface. This feature was important when a

structure like a pipeline slid over a soil surface and imposed a variable drainage condition.

189

The mortar contact formulation was derived for two different forms of dynamic coupled

equations: the U-P-V and U-P schemes.

In Chapter 4, a number of validation exercises were presented to evaluate the performance of

the developed numerical scheme. The dynamic coupled consolidation formulation was

validated for both small and large deformation analysis using the test problems presented by

de Boer et al. (1993) and Meroi et al. (1995). The closed-form solutions developed in

Chapter 3 were also employed to check the propagation of plane waves through a porous

medium with a compressible pore fluid. These examples demonstrated the effectiveness of

the numerical algorithm when analysing a problem that involved a significant transient

(short-term) dynamic response, dominated by inertia effects, followed by the increasing

importance of the consolidation phenomenon in the porous medium at intermediate and large

times.

The capability of the dynamic consolidation algorithm was verified for the elasto-plastic

consolidation processes by modelling the consolidation of a flexible strip footing under a

uniform pressure and comparing the predicted results with those presented previously by

Manoharan and Dasgupta (1995).

The undrained behaviour of a strip footing under dynamic loading and large deformations

was studied. This example attempted to illustrate the performance of the code in large

deformation regimes, particularly when combined with the ALE scheme and the contact

algorithm.

To further assess the contact algorithm, unconfined compression of a saturated porous layer

was modelled by which the contact patch test was examined.

The last part of Chapter 4 analysed the penetration of a rigid pile into a saturated soil layer. A

thorough investigation of the response of soil- and pore-pressure generation resulting from

the fast penetration of a pile was conducted, which revealed how soil resistance is mobilised

and pore pressures are generated and affected by the soil models, soil parameters, mesh size

190

and different contact discretisation schemes. The results indicated that smoothing the

transition point between the conical tip and cylindrical shaft can noticeably decrease the

oscillatory responses observed in the predicted total dynamic soil resistance, as well as the

excess pore-water pressure. The ALE method has tackled large deformation problems in

geomechanics in which displacements, velocities and accelerations are coupled with porewater pressures and Darcys velocity of the pore water, and involve rapid loading as well as

changing boundary conditions.

In Chapter 5, the numerical scheme was applied to the problems of dynamically penetrating

anchors (DPAs). First, a brief literature review of the available computational methods and

available model tests on DPAs was presented. It was concluded that the current design

procedures mainly include estimation of the penetration depth through a theoretical model

and predicting the pull-out capacity. Consequently, simulation of the installation process is

neglected, while installation of a DPA can lead to considerable disturbance and remoulding

of the soil in the vicinity of the anchor. The available FE simulations are generally based on a

displacement formulation (neglecting the pore-water pressures) and use simplifying

assumptions in the modelling. No laboratory tests have yet been reported which measure

excess pore pressures or effective stresses in the soil during or following the dynamic

penetration of objects. Two numerical examples concerning the installation of DPAs and the

subsequent consolidation analysis were presented in this chapter. The numerical scheme

successfully simulated these problems.

The first example studied the free-fall of a torpedo anchor into a lightly over-consolidated

clay layer. The analysis was conducted for three different values of the impact velocity10,

15 and 20 m/sand the evolution of excess pore pressure, total dynamic soil resistance and

deceleration characteristics of the anchor were studied. The analysis results predicted by the

NTS and the mortar algorithms were compared, and a significant improvement was observed

in the numerical oscillation when the mortar algorithm was utilised instead of the NTS

scheme. It was observed that the anchor may actually continue to accelerate under the action

of gravity during the early stages of penetration, but eventually it must start to decelerate and

ultimately come to rest, largely because of the finite shearing resistance of the seabed soil. It

191

was shown that most of the pull-out capacity of the torpedo anchor (soil resistance) could be

available much earlier than the completion of consolidation, as most of the excess pore

pressure dissipates within a matter of days or weeks.

The second example studied the free-fall of a torpedo anchor into a normally consolidated

clay layer. The numerical scheme simulated the entire installation process where the pathway

of the anchor was completely closed at the end of the installation. A higher weight and lower

impact velocity were adopted for the anchor in this example. The results suggested that an

anchor with a lower impact velocity could accelerate considerably during most of its

penetration, signifying that the closer the impact velocity is to the anchors terminal velocity,

the less acceleration will occur within the soil. Three stages were identified during the

consolidation process: (1) initial and quick dissipations that occur around the pile tip and

cone immediately after the installation, while the extent of compressive excess pore pressure

simultaneously increases both in the radial and vertical directions (this process lasts around

one week); (2) afterwards, dissipations occur predominantly in the radial direction up to

about a month; and (3) the rest of the consolidation process takes place with a decreased rate,

in which it takes ~515 days to reach a degree of 99.9 per cent of the consolidation. This stage

of consolidation occurs because of a combined action of radial and vertical dissipation of

pore pressures.

In Chapter 6, the numerical scheme was applied to the problems of pipeline-seabed

interaction problems.

The initial embedment of the pipe was observed to induce excess pore pressures around the

pipe wall. The maximum pore-pressure generation was experienced at the soil directly below

the pipe invert and was found to diminish towards the free surface along the pipe wall.

During the soil consolidation, the excess pore pressures around the pipe wall were found to

dissipate with time, considerably influenced by the horizontal length from the pipe invert as

pore water essentially moves away from the pipe invert towards the free surface. This trend

192

of pore-pressure dissipation was visualised by the direction of the pore-water flow during the

consolidation stage.

A comparison between the results predicted by the NTS and the mortar algorithms indicated

that the mortar algorithm would provide more stable results.

It was shown that the excess pore pressure at the pipe invert was not significantly affected by

the inertia forces. However, the embedment depth increased markedly because of the

dynamic forces.

that the initial break-out of the pipe occurred at a lateral displacement of ~0.1D and was then

followed by a steady residual resistance up to the horizontal displacement of 0.60D. During

this stage, the pipe had an upward trajectory, which was typical of light pipes. Suction pore

pressures were generated behind the pipe as soon as it was displaced laterally, whereas

compressive pore pressures were generated ahead of the pipe. The high value of compressive

pore pressure was shown to occur at the pipe invert while it was travelling along the trench

created during the sweeps. It was also observed that the movement of the pipe generated

some suction pore pressures behind the pipe at points that it passed over. However, they

quickly disappeared from the top surface as the drainage condition is automatically changed

to free-draining.

The accuracy of the predictions in a dynamic analysis largely depends upon the adopted timemarching scheme in the time domain. The dissipative characteristic of the time integration

scheme at a higher mode is a particular feature of the selected method that affects its accuracy

and efficiency. However, temporal discretisation errors are inevitable for all single-step timemarching schemes because of the discontinuous distribution of acceleration in the time

domain. The errors associated with the time discretisation might be reduced by increasing the

number of time increments. However, as the optimal time step size may change during the

computation process, particularly for nonlinear systems, a time step control algorithm is

required to automatically adjust the time increments to attain maximum accuracy while

193

preserving the feasibility, stability and efficiency of the solution. The development of an

automatic time-stepping procedure for the dynamic consolidation algorithm is recommended.

A survey of the literature revealed that most of the absorbing boundary conditions developed

in the context of two-phase saturated mixtures are in the form of viscous boundary

conditions. Therefore, they cannot model a static problem, and rigid body movement can

occur for low frequencies. An alternative approach is to use the infinite element method

(IEM) in combination with the standard viscous boundary. In this methodology, the near field

is discretised with the finite element method (FEM), whereas the far field is discretised using

the mapped IEM in the quasi-static form. In addition, the standard viscous boundary can be

utilised to absorb the dynamic waves at the FEIE interface.

It was shown that mortar contact is a robust method for geomechanics problems; it provides

quite smooth results compared to the NTS scheme, particularly when inertia forces are

involved. Based on the satisfactory results of 2D FE computations, this method may be

extended to 3D formulations.

The verification of the numerical scheme developed in this thesis, involving all its embodied

features, should be further investigated. This could be accomplished by conducting

simulations of laboratory tests on model free falling penetrometers and pipelines.

In the analysis of the dynamically penetrating anchors, the effect of strain rate on the

undrained shear strength was not considered. It is well known that this parameter can

significantly affect the soil resistance and predicted embedment depth. This was not included

in the analyses presented in this thesis, as the undrained shear strength was not a direct input

parameter of the Modified Cam Clay model. However, this may be easily considered by

adjusting the size of the yield surfaces based on the updated undrained shear strength

parameters at the end of each increment.

It was interesting to simulate the pull-out capacity of the dynamically penetrating anchors at

different times of installation under monotonic and time-varying cyclic loads.

simulation of torpedo anchors with fins and off vertical installations/pull-outs.

194

Small-amplitude cyclic movements of pipelines throughout the laying process may cause

significant self-burial. It is desirable to apply the numerical procedure developed in this thesis

to analyse such problems. However, small-amplitude cyclic movements may remould and

soften the surrounding soil, and such disturbance can lead to significant changes in the

operative shear strength and basic constitutive properties of the soil, such as the critical state.

The use of an appropriate soil constitutive model is then essential to account for the

combined effects of remoulding and reconsolidation during small-amplitude cyclic

movements.

195

References

Coulomb yield criterion. Comput Struct 54:427441.

Abelev, A., J. Simeonov and P. J. Valent. 2009. Numerical investigation of dynamic free-fall

penetrometers in soft cohesive marine sediments using a finite element approach. Oceans

2009, IEEE/MTS conference, Biloxi, MS.

company.

Akiyoshi, T., K. Fuchida and H. L. Fang. 1994. Absorbing boundary conditions for dynamic

analysis of fluid-saturated porous media. Soil Dyn Earthq Eng 13(6):387397.

Akiyoshi, T., X. Sun and K. Fuchida. 1998. General absorbing boundary conditions for

dynamic analysis of fluid-saturated porous media. Soil Dyn Earthq Eng 17(6):397406.

API RP 2A. 2002. Recommended practice for planning, designing and constructing fixed

offshore platforms. 21st ed., American Petroleum Institute, Washington, D.C.

Ateshian, G. A., S. Maas and J. A. Weiss. 2012. Solute transport across a contact interface in

deformable porous media. J Biomech 45(6):10231027.

clays. J Geotech Geoenv Eng ASCE 132, No. 6, 770777.

Laboratory and field research. Int J Offshore Polar, San Francisco, Paper No. 2006-PCW-03.

Baligh, M. M. 1985. The strain path method. J Geotec Eng, ASCE, 111:11081136.

196

Benson, D. J. 1989. An efficient, accurate and simple ALE method for nonlinear finite

element programs. Comput Meth Appl Mech Eng 72:305350.

Bernardi, C., Y. Maday and A. Patera. 1992. A new nonconforming approach to domain

decomposition: the mortar element method. in: (eds, Lions, J., Brezia, H) Nonlinear Partial

Differential Equations and their Applications, Pitman and Wiley, pp. 1351.

Bettess, P. and O. C. Zienkiewicz. 1977. Diffraction and refraction of surface waves using

finite and infinite elements. Int J Numer Meth Eng 11:12711290.

Biot, M. A. 1941. General theory of three dimensional consolidation. J Appl Phys 12:155

164.

solid. J Acous Soc Amer 28:168191.

Acous Soc Amer 34:12541264.

Biots equations of consolidation. Int J Solids Struct 11:907917

Computations 6:163168.

Borja, R. I. 1991. CamClay plasticity, Part II: Implicit integration of constitutive equations

based on a non-linear elastic stress predictor. Comp. Meth. Appl. Mech. Eng 88:225240.

Bowen, R. M. 1980. Incompressible porous media models by use of the theory of mixtures.

Int J Eng Sci 18:11291148.

197

Bowen, R. M. 1982. Compressible porous media models by use of the theory of mixtures. Int

J Eng Sci 20:697735.

Brezzi, F. and M. Fortin. 1991. Mixed and hybrid finite element methods. Springer Verlag,

New York.

Brown, P. T. 1969. Numerical analyses of uniformly loaded circular rafts on deep elastic

foundations. Geotechnique 19(3):399404.

interaction behaviour during lateral buckling, including large amplitude cyclic displacement

tests by the Safebuck JIP. In the proceedings of the offshore technology conference, Houston,

OTC17944.

Budhu, M. and C. S. Wu. 1992. Numerical analysis of sampling disturbances in clay soils. Int

J Num Anal Meth Geomech 16:467492.

Carter, J. P., M. F. Randolph and C. P. Wroth. 1979. Stress and pore pressure changes in clay

during and after the expansion of a cylindrical cavity. Int J Numer Anal Meth Geomech

3:305322.

Carter, J. P., H. Sabetamal, M. Nazem and S. W. Sloan. 2015. One-dimensional test problems

for dynamic consolidation. Acta Geotechnica 10(1): 173-178.

Carter, J. P., J. R. Booker and J. C. Small. 1979. The analysis of finite elasto-plastic

consolidation. Int J Numer Anal Meth Geomech 3:107129.

Carter, J. P., M. Nazem, D. W. Airey and S. W. Chow. 2010. Dynamic analysis of free falling

penetrometers in soil deposits. Geotechnical Special Pub. 199. GeoFlorida 2010Advances

in Analysis, Modeling and design. American Society of Civil Engineers, West Palm Beach,

Florida, 5368.

205.

198

Chan, A. H. C. 1988. A unified finite element solution to static and dynamic geomechanics

problems. PhD thesis, University College Swansea.

Chen, X., Y. Chen and T. Hisada. 2005. Development of a finite element procedure of

contact analysis for articular cartilage with large deformation based on the biphasic theory.

JSME Int J Ser C 48(4):537546.

interaction during large amplitude cyclic movements of partially embedded pipelines. Can

Geotech J 44:977996.

Chow, Y. K. and I. M. Smith. 1981. Static and periodic infinite solid elements. Int J Numer

Meth Eng 17:503526.

Chung, J. and G. M. Hulbert. 1993. A time integration algorithm for structural dynamics with

improved numerical dissipation: the generalized- method. J Appl Mech 60:371375.

Clayton, R. and B. Engquist. 1977. Absorbing boundary conditions for acoustic and elastic

wave equations. Bulletin of the Seismological Society of America 67:15291540.

Advances in porous media, Vol 3, Elsevier, New York. 361440.

Coussy, O. 1995. Mechanics of porous continua. John Wiley & Sons, Chichester, UK.

de Boer, R. 2000. Theory of porous media: Highlights in the historical development and

current state. Springer Verlag, Berlin.

de Boer, R., W. Ehlers and Z. Liu. 1993. One-dimensional transient wave propagation in

fluid-saturated incompressible porous media. Arch Appl Mech 63:5972.

de Borst, R. 1982. Calculation of collapse loads using higher order elements. In the

proceedings of the IUTAM symposium. Deformation and Failure of Granular Materials (eds.

P.A. Vermeer and H.J. Luger) 503513.

199

Degrande, G. and De Roeck. 1993. An absorbing boundary condition for wave propagation in

saturated poroelastic mediaPart I: Formulation and efficiency evaluation. Soil Dyn Earthq

Eng 12:411421.

De Josselin de Jong, G. 1956. Wat gebeurt er in de grond tijdens het heien? De Ingenieur,

68:B77-B88.

Dingle, H. R. C., D. J. White and C. Gaudin. Mechanisms of pipe embedment and lateral

breakout on soft clay. Can Geotech J 45:636652.

Donzelli, P. S. and R. L. Spilker. 1998. A contact finite element formulation for biological

soft hydrated tissues. Comput Meth Appl Mech Eng 153(1):6379.

Eide, O., J. N. Hutchinson and A. Landva. 1961. Short and long term test loading of a friction

pile in clay. In the proceedings of the 5th international conference on soil mechanics and

foundation engineering. Paris, 4554.

deformations based on the mortar method. Comput Mech 36:22644.

Gadala, M. S. 2004. Recent trends in ALE formulation and its applications in solid

mechanics. Comput Meth Appl Mech Eng 193:42474275.

elastic solids. J Eng Mech Div ASCE 98(4):947963.

Giannakopoulos, A. E. 1989. The return mapping method for the integration of friction

constitutive relations. Comput Struct 32(1):157167.

Gilbert, R. B., M. Morvant and J. Audibert. 2008. Torpedo piles joint industry projectModel

torpedo pile tests in kaolinite test beds. Final Report to the Mineral Management Service,

Project No. 575.

200

Graff, K. F. 1975. Wave motions in elastic solids. In: L. C. Woods, W. H. Wittrick and A. L.

Cullen editors. Oxford engineering science series.

Griffiths, D. V. 1982. Elasto-plastic analysis of deep foundations in cohesive soil. Int J Num

Anal Meth Geomech 6:211218.

Hallquist, J. O., G. L. Goudreau and D. J. Benson. 1985. Sliding interfaces with contactimpact in large-scale Lagrangian computations. Comput Meth Appl Mech Eng 107137.

expansion design and implementation. In the proceedings of the 35th offshore technology

conference, Houston, Tex., 58 May 2003, pp. 111

Hilber, H. M., T. J. R. Hughes and R. L. Taylor. 1977. Improved numerical dissipation for

time integration algorithms in structural dynamics. Earthquake Eng Struct Dyn 5:283292.

behavior. In Computational Methods for Transient Analysis, T. Belytschko and T.J.R.

Hughes, Editors, North-Holland, Amsterdam, The Netherlands. 67155.

Hughes, T. J. R and H. M. Hilber. 1978. Collocation, dissipation and overshoot for time

integration schemes in structural dynamics. Earthquake Eng Struct Dyn 6:99117.

Jaky, J. 1948. Pressure in soils. 2nd ICSMFE, London, 1:103107.

Science and Technology, Faculty of Information Technology, Mathematics, and Electrical

Engineering, Department of Engineering Cybernetics.

Kanok-Nukulchal, W. and V. W. Suaris. 1982. An efficient finite element scheme for elastic

porous media. Int J Sol Struct 48:3749.

Kaye, D., D. Marchand, P. Blondin and M. Carr. 1995. Lateral buckling design of the

DunbarNorth Alwyn double wall insulated pipeline. In the proceedings of the 9th offshore

pipeline technology conference, Copenhagen.

201

Kellezi, L. 2000. Local transmitting boundaries for transient elastic analysis. Soil Dyn Earthq

Eng 19 (7):533547

Kellezi. 1998. Dynamic soil-structure interaction transmitting boundary for transient analysis.

PhD Thesis, Department of structural engineering and materials, Technical University of

Denmark.

Khalili, N., S. Valliappan, J. T. Yazdi and M. Yazdchi. 1997. 1D infinite element for

dynamic problems in saturated porous media. Commun Numer Meth Eng 13:727738.

Khalili, N., M. Yazdchi and S. Valliappan. 1999. Wave propagation analysis of two-phase

saturated porous media using coupled finite-infinite element method. Soil Dyn Earthq Eng

18(8):533553.

Kiousis, P. D., G. Z. Voyiadjis and M. T. Tumay. 1988. A large strain theory and its

application in the analysis of the cone penetration mechanism. Int J Num Anal Meth Geomech

12:4560.

Kontoe, S., L. Zdravkovic and D. M. Potts. 2008. An assessment of time integration schemes

for dynamic geotechnical problems. Comput Geotech 35:253264.

Springer Verlag.

interaction. Springer-Verlag Berlin Heidelberg.

Lieng, J. T., F. Hove and T. I. Tjelta. 1999. Deep penetrating anchor: Subseabed deepwater

anchor concept for floaters and other installations. Proc. 9th International offshore and polar

engineering conference, Brest, France 1:613619.

Lieng, J. T., A. Kavli, F. Hove and T. I. Tjelta. 2000. Deep penetrating anchor: Further

development, optimisation and capacity clarification. In the proceedings of the 10th

international offshore & polar engineering conference. Seattle, US 2: 410416.

202

cone penetration in soft clay. Comput Geotech 36:851860.

Lysmer, J. and R. L. Kuhlemeyer. 1969. Finite dynamic model for infinite media. J Eng

Mech Div 5:859877.

Struct 54:10051021.

nonlinear problems. Comput Geotech 18(4):739751.

108:851-872.

Medeiros, C. J. 2001. Torpedo anchor for deep water. In the proceedings of the deepwater

offshore technology conference. Rio de Janeiro, Brazil.

Medeiros, C. J. 2002. Low cost anchor system for flexible risers in deep waters. In the

proceedings of the offshore technology conference. Houston, Texas, US.

Medina, F. and R. L. Taylor. 1983. Finite element techniques for problems of unbounded

domains. Int J Numer Meth Eng 19:12091226.

Meroi, E. A., B. A. Schrefler and O. C. Zienkiewicz. 1995. Large strain static and dynamic

semi saturated soil behavior. Int J Num Anal Meth Geomech 19:81106.

Mirza, U. A. A. 1999. Pile short-term capacity in clays. In the proceedings of the 9th

international offshore and polar engineering conference. 693699.

Modaressi, H. 1995. A note on absorbing boundary conditions for dynamic analysis of fluidsaturated porous media by Akiyoshi et al. Soil Dyn Earthq Eng 14(5):397397.

Modaressi, H. and I. Benzenati. 1994. Paraxial approximation for poroelastic media. Soil Dyn

Earthq Eng 13(2):117129.

203

Morland, L. W. 1972. A simple constitutive theory for a fluid-saturated porous solid. J Geoph

Res 77:890900.

Nazem, M., J. P. Carter and D. W. Airey. 2009. Arbitrary LagrangianEulerian method for

dynamic analysis of geotechnical problems. Comput Geotech 36(4):549557.

Nazem, M., J. P. Carter, D. W. Airey and S. H. Chow. 2012. Dynamic analysis of a smooth

penetrometer free-falling into uniform clay. Geotechnique 62(10):893905.

Nazem, M., D. Sheng and J. P. Carter. 2006. Stress integration and mesh refinement in

numerical solutions to large deformations in geomechanics. Int J Numer Meth Eng 65:1002

1027.

method for large-deformation consolidation problems in geomechanics. Int J Num Anal Meth

Geom 32(9):10231050.

Newmark, N. M. 1959. A method of computation for structural dynamics. J Eng Mech Div

ASCE 85:6794.

Nystrom, P. R., K. Tornes, J. S. Karlsen, G. Endal and E. Levold. 2002. Design for thermal

buckling of Asgard transport gas trunkline. Int J offshore Polar Eng 12:271279.

installed anchors in clay. Gotechnique 63:909919.

studies of deep penetrating anchors. Offshore Technology Conference, Houston, Paper No.

OTC 16841.

Oritz, M. and J. C. Simo. 1986. An analysis of a new class of integration algorithms for

elasto-plastic constitutive relations. Int J Num Meth Eng 23:353366.

204

Papadopoulos, P. and R. L. Taylor. 1990. A mixed formulation for the finite element solution

of contact problems. UCB/SEMM Report 90/18, University of California at Berkeley.

disposal. Can Geotech J 26:640652.

Theory. Thomas Telford, London.

Prvost, J. H. 1980. Mechanics of continuous porous media. Int J Eng Sci 18:787800.

Puso, M. and T. Laursen. 2004. A mortar segment-to-segment contact method for large

deformation solid mechanics. Comp Meth Appl Mech Eng 193(68):601629.

Raie, M. 2009. A computational procedure for simulation of torpedo anchor installation, setup and pull-out. Ph.D. Thesis, The University of Texas at Austin.

Geotech Geoenviron Eng 135(12), 18051813.

Randolph, M. F. and C. P. Wroth. 1979. An analytical solution for the consolidation around a

driven pile. Int J Numer Anal Meth Geomech 3(3):217230.

Randolph, M. F., J. P. Carter and C. P. Wroth. 1979. Driven piles in clay: The effects of

installation and subsequent consolidation. Geotechnique 29(4):361393.

finite element analysis. Int J Num Anal Meth Geomech 8: 243257.

installation of dynamic anchors in normally consolidated Clay. J Geotech Geoenviron Eng

135(4):487496.

205

performance of deep penetrating anchors in calcareous sand. In the proceedings of the

frontiers in offshore geotechnics. pp. 357363.

Sabetamal, H., M. Nazem, J. P. Carter and S. W. Sloan. 2014. Large deformation dynamic

analysis of saturated porous media with applications to penetration problems. Comput

Geotech 55:117131

Sandhu, R. S., H. Lui and K. J. Singh. 1977. Numerical performance of some finite element

schemes for analysis of seepage in porous elastic media. Int J Num Anal Meth Geomech

1:177194.

Seed, H. B. and L. C. Reese. 1955. The action of soft clay along friction piles. Proc Am Sot

Civ Engrs 81, Paper 842.

unbounded saturated soil media. J Geotech Eng 115(11):16331646.

Sheng, D., S. W. Sloan and H. S. Yu. 2000. Aspects of finite element implementation of

critical state models. Comp Mech 26(2):185196.

Sheng, D., P. Wriggers and S. W. Sloan. 2006. Numerical algorithms for frictional contact in

pile penetration analysis. Comput Geotech 33(67):341354.

Sheng, D., P. Wriggers and S. W. Sloan. 2006. Improved numerical algorithms for frictional

contact in pile penetration analysis. Comput Geotech 33(67):341354.

geotechnical engineering. Int J Geomech 7(3):176185.

Simo, J. C. and G. Meschke. 1993. A new class of algorithms for classical plasticity extended

to finite strains: Application to geomaterials. Comput Mech 11(4):253278.

206

elastoplasticity. Comput Meth Appl Mech Eng 48:101118.

Simoni, L. and B. A. Schrefler. 1987. Mapped infinite elements in soil consolidation. Int J

Numer Meth Eng 24:513527.

Sloan, S. W. and M. F. Randolph. 1982. Numerical prediction of collapse loads using finite

elements methods. Int J Num Anal Meth Geomech 6:4776.

Sloan, S. W. and A. J. Abbo. 1999. Biot consolidation analysis with automatic time stepping

and error control Part 1: Theory and implementation. Int J Num Anal Meth Geom 23:467

492.

Small, J. C., J. R. Booker and E. H. Davis. 1976. Elasto-plastic consolidation of soil. Int J

Solids Struct 12: 431448.

Sousa, J., S. Cristiano and G. B. de Aguiar. 2010. Undrained load capacity of torpedo anchors

embedded in cohesive soils. J Offshore Mech Arct Eng 133(2):021102-1.

construction. Ph.D. Thesis, University of Western Australia.

Strang, G. and G. J. Fix. 1973. An analysis of the finite element method. Prentice-Hall, series

in Automatic Computation (ed. G. Forsythe)

Straughan, B. 2008. Stability and wave motion in porous media. Applied mathematical

sciences. Vol. 165. Springer, New York.

Sturm, H. and L. Andresen. 2010. Large deformation analysis of the installation of dynamic

anchor. In the proceedings of the 7th European conference on Numerical Methods in

Geotechnical Engineering (NUMGE), pp. 255260.

207

Talbot, A. 1979. The accurate numerical inversion of Laplace transforms. J. Inst. Math.

Applic. 23:97-120.

Taylor, R. L. and P. Papadopoulos. 1991. On a patch test for contact problems in two

dimensions. Nonlinear Computational Mechanics (ed. P. Wriggers and W. Wagner) pp. 690

702.

Teh, C. I. and G. T. Houlsby. 1991. An analytical study of the cone penetration test in clay.

Gotechnique 41:1734.

Terzaghi, K. 1923. Theoretical soil mechanics, Third edition, John Wiley and Sons, Inc.,

New York, N.Y.

Terzaghi, K. 1936. The shearing resistance of saturated soils and the angle between the

planes of shear. In the proceedings of the international conference on soil mechanics and

foundation engineering, Cambridge, Mass, Vol. I: 5456

True, D. G. 1976. Undrained Vertical Penetration into Ocean Bottom Soils. PhD Thesis,

University of California, Berkeley, California.

Truesdell, C. and R. Toupin. 1960. The classical field theories. in: Handbuch der Phsik (ed.

S. Flugge) Vol. III/1, Springer-Verlag Berlin.

Vancouver.

van den Berg, P. 1991. Numerical model for cone penetration. In the proceedings of the 7th

Int. Conference of International Association for Computer Methods and Recent Advances in

Geomechanics, IACMAG, Cairns, Australia. 3:17771782.

Verley, R. and K. M. Lund. 1995. A soil resistance model for pipelines placed on clay soils.

In the proceedings of the 14th international conference on Offshore Mechanics and Arctic

Engineering (OMAE), Copenhagen, Denmark, Vol. 5, pp. 225232.

208

elements. Int J Num Anal Meth Geomech 5:114.

Voyiadjis, G. Z. and Y. M. Abu-Farsakh. 1997. Coupled theory of mixtures for clayey soils.

Comput Geotech 20:195222.

Wan, J. 2002. Stabilized finite element methods for coupled geomechanics and multiphase

flow. PhD thesis, Stanford university.

calcareous sediments. Appl Ocean Res 20: 8394.

White, D. J. and C. Y. Cheuk. 2008. Modelling the soil resistance on seabed pipelines during

large cycles of lateral movement. Marine Structures 21(1):5979.

Englewood Cliffs, New Jersey.

approach, Elsevier.

method. Int J Numer Meth Eng 15:15621566.

Worth, C.P. 1984. The interpretation of in situ soil tests. Gotechnique, 34(4): 449489. DOI:

10.1680/geot.1984.34.4.449

problems using a moving friction cone. Commun Numer Meth Eng 19:285295.

209

Wriggers, P. and J. C. Simo. 1985. A note on tangent stiffness for fully nonlinear contact

problems. Commun Appl Numer Methods 1:199203.

Xia, K. and Z. Zhang. 2006. Three-dimensional finite/infinite elements analysis of fluid flow

in porous media. Appl Math Model 30(9):904919.

Yu, H. S. and J. K. Mitchell. 1998. Analysis of cone resistance, review of methods. J Geotech

Geoenviron Eng 124(2):140149.

Yu, H. S., L. R. Herrmann and R. W. Boulanger. 2000. Analysis of steady state cone

penetration in clay. J Geotech Geoenviron Eng 126:594605.

contact patch test. Int J Numer Meth Eng 79:379416

Zerfa, Z. and B. A. Loret. 2004. Viscous boundary for transient analyses of saturated porous

media. Earthq Eng Struct Dyn 33:89110.

Zhao, C. 2009. Dynamic and transient infinite elements: Theory and geophysical,

geotechnical and geoenvironmental applications. Springer dordrecht Heidelberg, London

New York.

Zhao, C. and S. Valliappan. 1993. Transient infinite elements for seepage problems in infinite

media. Int J Num Anal Meth Geomech 17:323341

Zhao, C., S. Valliappan and Y. C. Wang. 1992. A numerical model for wave scattering

problems in infinite media due to p- and sv-wave incidences. Int J Numer Meth Eng 33:1661

1682.

Zienkiewicz, O. C., C. Emson and P. Bettess. 1983. A novel boundary infinite element. Int J

Numer Meth Eng 19:393404.

210

Zienkiewicz, O. C. and T. Shiomi. 1984. Dynamic behaviour of saturated porous media: The

generalized Biot formulation and its numerical solution. Int J Numer Anal Meth Geomech

8:7196.

Zytynski, M., M. F. Randolph, R. Nova and C. P. Wroth. 1978. On modelling the unloadingreloading behaviour of soils. Int J Numer Anal Meth Geomech 2:8793.

211

Appendix A.I

K =

NL

K=

K +

B L [ D] B L JdV

T

A.1

[] B NL JdV

NL T

+ B [ ] B L JdV +

L T

K sp =

B NL [ P ] B NL JdV

T

div ( N ) N

T

s

A.2

JdV

A.3

JdV

A.4

K rp =

div ( N )N

T

r

K pp =

T

p

div ( N p ) v r JdV

M ss = NTs N s JdV

A.6

M sr = NTs f N r JdV

A.7

M rs = NTr f N s JdV

A.8

M rr = NTr

=

Css

N r JdV

A.9

NTs

f (1 n0 )

+ NTs

A.5

div ( N s ) v r dV

A.10

grad ( n ) N s v r JdV

n

T

Derivation of Eq. A.2 and definition of B matrices are presented in Appendix A.II. In Eq.

A.10, c denotes Rayleigh damping matrix.

212

grad ( n ) v r N r JdV

n

Csr = NTs

Crr

=

NTr

+ NTr

+ NTr

f

n

f

n

k 1N r JdV + NTr

n2

grad ( v r ) N r JdV

f (1 n0 )

n2

A.11

div ( v s ) N r dV

A.12

n

n

T

[grad(N

C pr =

)]T N r JdV

A.13

C ps =

T

p

div ( N s ) JdV

A.14

C pp =

T

p

=

fs

N P JdV

N t ds + N

T

s

T

s

bJdV

A.15

A.16

f p = NTp qN ds

A.17

=

fr

N t

T p

r

ds +

213

T

r

f bJdV

A.18

Appendix A.II

The contributions of virtual work related to effective stress appearing in the solid momentum

balance equation (the forth term in Eq. 2.52) are obtained in this appendix.

The procedures according to Bathe (1996) are followed to obtain contributions to both the

tangent matrix and the residual vector for Newton iterations. This leads to an updated

Lagrangian (UL) formulation, in which the geometry and field variables are updated upon

convergence in a step, and for the following step variables are referred to the updated

configuration. According to Bathe (1996), the following equation for the UL method is

obtained:

d S

ij

R t ij d (d ij ) JdV

A.19

where Sij denotes the second PiolaKirchhoff stress tensor d ij and dij are linear and

nonlinear parts of the increment of GreenLagrange strain tensor, respectively, and Rt

denotes external forces. As the material constitutive equations are usually defined in terms of

the Cauchy stress and linear strain tensors, it is necessary to use an invariant stress rate with

respect to rigid-body rotation instead of the PiolaKirchhoff stress tensor. As an objective

measure of the stress rate, the Jaumann effective stress rate d ijj was adopted for the present

formulation as:

d ij j =

d ik + ik d kj + jk d ki

A.20

where ij is the effective Cauchy stress and ij is the antisymmetric spin tensor. The linear

relationship between the objective stress and deformation rates (i.e., the effective stressstrain

law) is:

=

d jij d ijj + d ij dpij = Dijkl d kl + d ij dpij

A.21

where d ij is the Kronecker-delta and Dijkl is the elasto-plastic tensor of the constitutive model.

It is worth noting that the Jaumann stress rate is an approximation of the Truesdell rate, as the

stretch component of deformation is ignored in the Jaumann stress rate, and the spin tensor

ij is assumed to be equal to the material rotation rate tensor. Therefore, the Jaumann stress

214

rate can be used properly if the strain increments are kept small enough in each step of a UL

analysis.

Introducing Eqs A.21 and A.20 into Eq. A.19, the following equilibrium equation for the UL

method is obtained:

ijkl

+ ( ik d jk + jk d ik ) d ( d ij ) JdV

+ dpij d ij d ( d ij ) JdV +

=R ij d ( d ij ) JdV

t

ij

A.22

ij d ( d ij ) JdV

Applying the standard finite element procedure to Eq. A.22, the following equation is

obtained by which the equilibrium of the body (in quasi static condition) is formulated:

K NL U + K sp P= F ext F int

A.23

where K sp was already defined by Eq. A.3, K NL now includes the nonlinear (NL) and linear

(L) terms of strain as:

K NL

=

[ D] B L JdV +

L T

+ B [ ] B JdV +

L T

NL

[] B JdV

NL T

[ P ] B JdV

NL T

A.24

NL

where the first term in the right-hand side of Eq. A.24 represent K whereas the other three

terms are related to the geometrical nonlinearity. F ext denotes external forces and F int

represents internal forces as:

=

F int

T

A.25

In Eq. A.24, the B matrices relate the displacements into the strain. For two dimensional

plane strain conditions, the B matrices for the ith node of an arbitrary element are defined as

215

N is

x1

B L = 0

N is

x

2

N i

x

1

B NL =

s

N i

x2

N is

x1

N i

x2

N i

x1

1 N i

0 0 2 x

1

B L =

1 N i

0 0

2 x1

A.26

N i

x2

A.27

A.28

For 2D plane strain conditions, the stress vectors and matrices in the above equations are

defined as:

0

0

11 12

0

0

21

22

[ ] = 0 0

11

12

22

0 21

0

A.29

0

2 12

2 11

=

[] 0 2 22 2 12

12

11

12 22

p

0

[ P ] = 0

=

{}

0

p

0

0

0

0

p

0

0

0

0

A.31

22

12 }

p11 p22 p12 } {m} {1

{=

{P} {=

11

T

A.30

1 0}

A.32

Note that the effective Cauchy stress tensor is obtained from the time integration of elastoplastic large strain constitutive model as elaborated in Nazem et al.(2008).

216

Appendix A.III

INTRODUCTION

In this section, the basic equations governing the dynamics of a saturated porous medium are

considered. They were first derived by De Josselin de Jong (1956) and Biot (1956) and a

clear exposition of them, together with some useful solutions, may be found in Chapter 5 of

the book by Verruijt (2010).

In particular, the basic equations will be presented for the one-dimensional case of

propagation of plane waves and the associated coupled consolidation. The solution for the

problem of step loading applied to a layer of saturated soil with a linear elastic skeleton and a

compressible pore fluid is presented. This solution may be useful in the validation of FE

codes developed for the solution of dynamic consolidation problems.

As indicated by Verruijt (2010), the governing differential equations for the one-dimensional

case of plane wave propagation in a soft soil, saturated with a compressible pore fluid, are as

follows:

1.

Mass conservation of the fluid and solid particles, i.e., total mass conservation:

2.

n ( v w )

w

p

+ Sp

=

x

t

x

mv

3.

A.33

w

=

t

x

217

A.34

n f

4.

v

w

p

+ (1 n ) s

=

t

t

x

x

A.35

Conservation of momentum of the pore fluid, i.e., the generalization of Darcys law to

( v w)

v

p n 2

n f

+ n s

=

n

( v w)

t

t

x

A.36

These are four governing equations in the four basic field quantities, defined as follows:

v

The symbols x and t represent the one-dimensional spatial coordinate and time, respectively.

The other symbols appearing in these equations represent the material properties, as follows:

n

mv

conditions,

Sp

= a tortuosity factor, describing the added mass due to the tortuosity of the fluid flow

path.

In developing these equations, it was assumed that the total stress, , can be decomposed into

the isotropic effective stress, , and the pore pressure, p, as follows:

= + p

It can also be shown that the storativity of the pore space can be written as:

218

A.37

S p = nC f + ( n ) Cs

A.38

= 1

Cs

Cm

A.39

where Cf, Cs and Cm are the compressibility of the pore fluid, the solid particle material and

the porous medium, respectively.

If the soil skeleton can be represented by an ideal isotropic linear elastic material, then the

one-dimensional compressibility, mv, can be expressed in terms of elasticity coefficients as:

mv =

1

4

K+ G

3

A.40

where K and G represent the elastic bulk modulus and shear modulus, respectively.

SPECIAL CASE

Consider the special case where = 0 and = 1. This corresponds to a soil where the

tortuosity is insignificant and the compressibility of the solid particles is much less than that

of the saturated soil overall. These are reasonable approximations of many cases of soils

encountered in engineering practice. It is also reasonable to assume that variations in the

porosity of the soil are of second order importance, so that the porosity n may be assumed as

approximately constant.

With these assumptions, the governing equations for this special case simplify to the

following:

w

p

v w

+ Sp

=

n

x

t

x x

mv

w

=

t

x

219

A.41

A.42

n f

v

w

p

+ (1 n ) s

=

t

t

x x

A.43

v

p n 2 w

=

n

( v w)

k

t

x

A.44

n f

In Eq. A.44 account has also been taken of the following relationship:

f g

=

k

w

k

A.45

where g is the acceleration due to gravity and k is the hydraulic conductivity of the soil that is

familiar from Darcys law.

Solutions of the governing equation, A.41A.44, can be obtained by a variety of means. For

example, analytical solutions can be pursued using the method described by Verruijt (2010),

in which the fundamental solutions for harmonic variations in the field quantities applied at

the boundaries can be combined appropriately as Fourier series to represent the required

boundary conditions.

Alternatively, closed form solutions may also be obtained using the technique that involves

taking Laplace transforms of the governing equations, solving these equations in Laplace

transform space, and then inverting the solution for the Laplace transforms, numerically if

necessary, to recover the original field quantities.

A third option is to apply the numerical technique of finite differences to solve Eqs A.41 to

A.44 directly, subject to the appropriate boundary conditions.

In this study, the Laplace transform method will be used and the obtained solutions will be

checked using an independent finite difference approach.

Consider first the problem of a layer of saturated porous soil subjected to a sudden increase in

pore water pressure applied at the soil surface.

220

The problem of an infinitely deep layer was considered previously by Verruijt (2010) who

solved it both numerically and using the Fourier series technique. As indicated, the method

of taking Laplace transforms will be used here.

Taking Laplace transforms of the governing equations, A.41 toA.44, provides:

( n 1)

w

v

=n + S p ps

x

x

mv s =

w

x

A.46

A.47

x x

A.48

p

n n 2 w ( v w )

n f vs =

x

k

A.49

n f vs + (1 n ) s ws =

where the superior bar indicates a Laplace transform quantity and s is the Laplace transform

variable.

If Eqs A.48 and A.49 are both differentiated with respect to the coordinate x, and appropriate

substitutions are made, making use of Eqs A.46 and A.47, these four governing equations

expressed in terms of Laplace transforms can be reduced to the following two equations in

terms of the transforms of the effective stress and pore pressure, i.e.,

Ap + B +

2 2 p

+

=

0

x 2 x 2

A.50

2 p

=

0

x 2

A.51

Cp + D +

where

A = f S p s2

A.52

B =( f s ) (1 n ) mv s 2

A.53

221

S p s

C=

f s + n w

k n

A.54

w 1 n w

D=

f s + n

+ n mv s

k

n

A.55

Further simplification provides the following governing equation for the transform of the

pore water pressure:

4 p

2 p

0

+ X 2 + Yp =

x 4

x

A.56

where

X = B+C D

A.57

=

Y BC AD

A.58

and

The solution of Eq. A.56 is well known and in general it takes the form:

p = E1e 1x + E2 e1x + F1e 2 x + F2 e 2 x

A.59

where the coefficients E1, E2, F1, F2 must be determined from the boundary conditions of the

problem. It can also be shown that the terms 1 and 2 are given by:

1

=

X 2 4Y X

2

2

X 2 4Y X

2

2

2 =

222

A.60

A.61

First consider the case of an infinitely deep layer. In such cases, the solution must remain

bounded as x approaches , which means that E2 = F2 = 0. The boundary condition at x = 0

corresponds to a step loading in the pore pressure p, which in turn implies the following:

p=

po

s

A.62

where po is the magnitude of the step rise in pore pressure. Also at x = 0 the effective stress

boundary condition is expressed as:

= 0

A.63

Applying these boundary conditions provides the following solutions for the non-zero

constants E1 and F1:

2

p + C

E1 = o 22

2

s 1 2

A.64

2

p + C

F1 = o 21

2

s 1 2

A.65

The solution for the Laplace transform of the pore pressure is given by the combination of

Eqs A.59, A.64 and A.65. It then remains to invert this transform to recover values of the

pore pressure p. For this problem analytical inversion of the transform is difficult, if not

impossible, so that numerical inversion is required.

problem, the transforms have been inverted numerically using the algorithm suggested by

Talbot (1979).

Solution Evaluation

Solutions have been evaluated for the case of an infinitely deep layer of saturated soil to

which a step loading in pore water pressure (and total stress) of magnitude po is applied at the

surface x = 0. These solutions are presented in Figures A.1 and A.2 and correspond to the

223

material properties listed in Table A.1. They show the pore water pressure as a function of

time at a location given by x = 0.2 m.

Symbol

n

Property

Porosity of the soil (-)

Biots coefficient for a saturated soil (-)

Tortuosity (-)

Density of the pore fluid (kg/m3)

Density of the solid particles (kg/m3)

Hydraulic conductivity of soil (m/s)

g

mv

Cf

Cs

Compressibility of soil (m2/N)

Compressibility of pore fluid (m2/N)

Compressibility of solid particles (m2/N)

f

s

Value

0.4

1

0

1000

2650

0.001 and

0.0005

10

2x10-10

5x10-10

0

The solutions plotted in Figure A.1 correspond to the case of a soil with k = 0.001 m/s. 1

In Figure A.1(a), corresponding to small values of time, it can be clearly seen that two waves

of dynamic pore pressure are developed and pass through the given location.

As indicated by Verruijt (2010), the first wave arrives at a time of approximately 0.00009 s

(moving with a velocity of 2242 m/s) and is what is known as an undrained wave because

the soil skeleton and the pore fluid move in phase with each other, i.e., for this type of wave

the velocities of the solid particles (w) and the pore fluid (v) are the same. The second wave

observed in Figure A.1(a) corresponds to the case where the velocities of the solid particles

and the fluid are equal in magnitude but opposite in direction. For this type of wave, the

velocity is much slower, i.e., at approximately 1180 m/s which is about one-half of the

velocity of the undrained wave.

Figure A.1(b) shows the solution for the same case as depicted in Figure A.1(a), but for larger

values of time. It can be observed that with the passage of time, after the initial shock due to

the arrival of the dynamic waves at x = 0.2, the pore pressure gradually increases and

approaches the value po applied at the boundary x = 0. The mechanism causing this increase

is consolidation, as the pore fluid flows through the solid skeleton of the soil. Evidence for

1

The minor oscillations in the plotted solution are simply an artifice of the numerical algorithm used to invert

the Laplace transforms and are not physically real.

224

pseudo static consolidation can be found in the predicted consolidation curve, but this is best

illustrated by considering a layer of finite thickness rather than an infinitely thick layer. The

latter problem will be further elaborated in due course.

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

x=0.2m

0.1

0.0

0.0E+00

1.0E-04

2.0E-04

3.0E-04

4.0E-04

5.0E-04

Time (s)

(a)

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

x=0.2

0.1

0.0

0.0E+00

2.0E-02

4.0E-02

6.0E-02

8.0E-02

1.0E-01

Time (s)

(b)

Figure A.1: Pore pressure at x = 0.2 m in infinitely deep layer with k = 0.001 m/s

Meanwhile, as observed by Verruijt (2010), the second type of wave detected in this problem

attenuates reasonably quickly. This attenuation or damping arises principally because the

water must flow through the solid skeleton (i.e., v and w are different) and in doing so it

225

meets resistance. The undrained wave is not attenuated in the same way because the soil and

water move together.

As also noted by Verruijt (2010), this attenuation is a function of the hydraulic conductivity

of the soil; the lower the value of hydraulic conductivity the more quickly the wave is

damped. An example of this effect may be seen in Figure A.2, which shows results plotted

for the case where k = 0.0005 m/s, i.e., a soil only one-half as permeable as that shown in

Figure A.1.

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

x=0.2

0.1

0.0

0.0E+00

1.0E-04

2.0E-04

3.0E-04

4.0E-04

5.0E-04

Time (sec)

Figure A.2: Pore pressure at x = 0.2 m in infinitely deep layer with k = 0.0005 m/s

Comparison of Figure A.1(a) and Figure A.2 reveals that by the time the second wave of pore

pressure arrives at x = 0.2, it already has a smaller amplitude, while the first wave type

appears to have the same magnitude in each case. Note that the overall pressure immediately

after the arrival of the second wave is slightly more than 0.8po in Figure A.1(a), while it is

lower at approximately 0.7po in Figure A.2.

Finite Layer

Consider now the case of a layer of finite thickness, H. Conceptually, this case is no more

challenging to solve than the infinitely deep layer, involving only slightly different yet

significant boundary conditions. As it shall be seen revealed in the evaluated solution, the

226

presence of a rigid impermeable boundary at the bottom of the layer produces some very

interesting effects.

For this case, there are two additional boundary conditions that must be applied at x = H.

These are:

p

x H

= = 0 ,=

x x

A.66

provides four equations allowing solutions to be obtained for the coefficients E1, E2, F1 and

F2 in the general solution expressed as Eq. A.59. These equations can be written as:

1

1

2

2

1 + C

1 + C

1H

1e

1e1H

1H

2

1 (12 + C ) e1H

1 (1 + C ) e

1

2

2 + C

2 e 2 H

2 ( 22 + C ) e 2 H

E1

E2 =

F

1

2 ( 22 + C ) e 2 H F2

1

2

2 + C

2 e 2 H

p0

s

0 A.67

0

Values of these coefficients are required in the Laplace transform solution of the problem of a

finite layer. Otherwise, inversion of the transforms proceeds as for the infinitely deep layer.

Solution Evaluation

Solutions have been evaluated 2 for the case of a 1 m deep layer of saturated soil to which a

step loading in pore water pressure (and total stress) of magnitude po is applied at the surface

x = 0. These solutions are presented in Figures A.3 and A.4, and they correspond to the

material properties listed in Table A.1, with the smaller hydraulic conductivity,

k = 0.0005 m/s, being adopted. Figure A.3 shows the variation of the pore pressure at

x = 0.2 m, while Figure A.4 shows the pore pressure at the bottom of the layer, x = 1 m.

There are several interesting features depicted in the plots shown in Figures A.3 and A.4,

which are now described.

2

For convenience the results in Figures A.3 and A.4 were computed using the finite difference approach, rather

than Laplace transforms. This explains the slight numerical overshoot when a wave arrives. It should also be

noted that the Talbot method of inversion of the Laplace transforms proved to be problematic for times greater

than about 0.0008 s at x = 0.2 m, presumably due to singularities in the transform of pore water pressure. It is

curious that this occurred at about the time the first reflected wave arrived at x = 0.2 m. This issue requires

further investigation but is beyond the scope of the present study.

227

First, a comparison of Figures A.3(a) and A.4 further illustrates the point about the damping

of the second type of wave. Closer to the source of the disturbance, at x = 0.2 m, two distinct

types of wave can be seen arriving at different times, as previously discussed. However,

further from the source, at x = 1m, the undrained wave type clearly arrives at a time of

approximately 0.00045 s, corresponding to a velocity of 2242 m/s.

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

x=0.2m

0.1

0.0

0.0E+00

1.0E-04

2.0E-04

3.0E-04

4.0E-04

5.0E-04

Time (s)

(a)

1.8

1.6

1.4

1.2

1.0

0.8

0.6

0.4

x=0.2

0.2

0.0

0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.010

Time (s)

(b)

Figure A.3: Pore pressure at x = 0.2 m in finite (1 m thick) layer with k = 0.0005 m/s

Only a very weak second pulse can be observed in the time trace shown in Figure A.4, i.e., at

a time of about 0.00085 s, corresponding to the speed of a wave of the second type of

228

1180 m/s. So although the second type of wave can just be observed it has almost completely

attenuated by the time it reaches the bottom of the 1 m deep layer. Thereafter, it should play

no significant part in the on-going pore pressure history of the finite layer of saturated soil.

1.1

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0

0.0E+00

x=1.0m

2.0E-04

4.0E-04

6.0E-04

8.0E-04

1.0E-03

Time (s)

Figure A.4: Pore pressure at x = 1 m in finite (1 m thick) layer with k = 0.0005 m/s

Figure A.3(b) shows the variation of the pore pressure at x = 0.2 m for a longer period than

depicted in Figure A.3(a). The series of square pulses traced out in this plot correspond to a

sequence of undrained waves reflected from the boundaries of the layer, both top and bottom.

The first reflection arrives at a total elapsed time of approximately 0.0008 s, which is

consistent with a wave travelling at 2242 m/s generated at the surface at t = 0 and travelling

1 m to the bottom of the layer and being reflected to arrive back at the location x =0.2 m after

having travelled a total distance of 1.8 m in about 0.0008 s. Note that this first reflected wave

causes an increase in the pore water pressure. In other words, the reflection from the fixed

boundary has caused the reflected wave pulse to have the same sign as the incoming wave.

The reflected wave then continues to travel back towards the surface where it again is

reflected, in this case from the free boundary. Travelling at a speed of 2242 m/s it arrives

back at x = 0.2 m after a further period of approximately 0.0002 s, corresponding to the time

required to traverse a distance of 2 0.2 m = 0.4 m. It passes through x = 0.2 m again at a

total elapsed time of approximately 0.001 s. On this occasion it causes a reduction in the

pore water pressure, having been reflected from a free surface. In other words, reflection

from the free surface has caused a change in sign of the reflected pulse.

229

This process of sequential reflection from the fixed and free surfaces of the layer continues,

as is evidenced by the series of regular spiked pulses in the pore pressure history.

Meanwhile, the mean pore pressure, ignoring the pulsing, rises consistently with time, driven

by the underlying consolidation process taking place in the saturated soil. It is worth noting

that in Terzaghis theory of consolidation for static loading applied to a finite layer with oneway (surface) drainage, the non-dimensional time for about 90% consolidation is

approximately 1. For the example studied here it can be shown that coefficient of

consolidation of the soil, cv = k/(mvw) = 250 m2/s, which implies a real time for about 90%

consolidation in a 1 m thick layer of approximately 0.004 s. The curve in Figure A.3(b)

indicates that a mean pore water pressure of about 90% of the applied pressure occurs around

t = 0.004 s, supporting the contention that the rise in mean pore pressure is driven by

consolidation.

VALIDATION

The example solutions plotted in Figures A.3 and A.4 might be useful for validating FE codes

for dynamic consolidation. While they consider only elasticity and small strains, they do

allow a check on the concurrent wave transmission and consolidation processes.

230

Appendix A.IV

1 m

1 m

1 m

Fns+1 = f ns+1 f + M ss

Un +

1 U n + M sr

1 ( Vr )

n

t

2

(1 f )

int

1 U n + (1 f )

1 tU

+ Css

n C sr ( Vr ) n f n

A.68

where=

1 m

1 m

1 m

1 U

1 ( V

Fnr+1 = f nr+1 f + M rs

Un +

n + M rr

r )n

2

A.69

(1 f )

p

p

+ (1 ) 1 tU

+ C (V )

1

F=

f

C

U

n +1

n +1 f

ps

f

n

pr

r n

A.70

( Crr (Vr ) n + K rp Pn )

(1 f )

+ C pp

1 Pn + (1 f )

1 tP

n K PP Pn

where f n +1 f =

231

Appendix B

The following expressions represent the necessary changes to be included in the residual

contact vectors and stiffness matrices when an augmented Lagrangian scheme is adopted,

based on the scheme presented in Table 2.1.

Pore-pressure contributions

nq

R p = qN B p L q wq

B.1

q =1

K pu=

nq

q =1

a guN cn

qN B p , + p B p BTp , P ( ) Bu aT + guN Bu , nT L q wq

B.2

nq

R v = B v ntvN L q wq

B.3

q =1

K vu

2

2

nq

a

a

L q wq

q =1

1

T T

T T

T T

2 v B v na B v Vr n B v , + tvN B v an B v ,

232

B.4

Appendix C.I

(i ) g u

=

where =

n+1

a n +1 (i )

0

vN n+1

a n +1 (i )

u

C.1

v

. Multiplying Eq. C.1 by ai results in:

u

(i ) =

g u ai sign( guT ) sign( guN )g uN

n+1

n+1

C.2

+ sign( guNi ) ( v

nm

r

m

r

) n i ai

C.3

x,m ai + ( x nm x m ) ai , sign( guT ) + sign( guN )x,m ni ai

i , =

sign( guNi ) v ni ai + sign( guNi ) ( v

m

,

nm

)n

C.4

i ,

ai

where:

m

i ,

=a ,

m

i

m

i

am am

= i m i = aim aim

ai

m

i ,

aim aim,

aim

C.5

and:

nimaim =

0 nim, aim + nimaim, =

0 nim, aim =

nimaim,

Using these results in Eq. C.4 and doing some algebraic manipulation yields:

233

C.6

2

=

i , sign( guTi ) ( x nm x m (i ) ) aim, aim

aim aim,

m

m

guTi g vTi ni ai , + g vNi guNi

aim

+ sign( guNi )

m

m

m

v , ni ai

C.7

where:

g vTi =

(v

nm

aim

v (i ) ) m

ai

m

234

C.8

Appendix C.II

can be evaluated from the linearisation of the slip function (Eq. 378 ) as:

a

g un+1

sign( guTn+1 ) sign( guNn+1 )( guNn+1 v g vNn+1 ) a =

0

a

C.9

g un+1 a sign( guTn+1 ) + g un+1 a sign( guTn+1 ) sign( guNn+1 ) guNn+1 a + guNn+1 a

0

where =

C.10

v

. Expanding Eq C.10 based on the definition of gap vectors provides:

u

n+1

n+1

+ ( x nm x m ) U ,m sign( guTn+1 )

( U nm U m ) n a a n a + ( x nm x m ) n a

sign( guNn+1 )

+ ( x nm x m ) n a

m

nm

m

( v nm

r v r ) n a v , n a + ( v r v r ) n a

+ sign( guNn+1 )

m

+ ( v nm

r v r ) n a

C.11

=0

where:

a a

a =

,

a

a =(x, ) =U , + x,

(na) = na + na na = na

Using Eqs C.5 and C.12 in Eq. C.11 yields:

235

C.12

sign( g ) ( x nm x m ) x m a 2

,

uTn+1

m

m

+ sign( guNn+1 ) ( x nm x m ) ( v nm

r v r ) ( a n n a ) x ,

a

v m n a sign( g

r ,

uN n+1 )

nm

m

( U U ) sign( guTn+1 )a sign( guNn+1 )n a

n

m

m

nm

m

U , a + sign( guNn+1 ) ( v r v r ) n a

U , n

236

C.13

Appendix C.III

Linearisation of contact virtual work arising from Darcy velocity ( ccv slip )

r

ccvr slip =

(g vNd g vN + g vN d g vN )d

C.14

( v

g vN =

nm

r

v mr v mr , ) n m + ( v rnm v rm ) n m

d g vN = d v rm, n m + (d v rnm d v mr ) n m

C.15

=K u U K v Vr

C.16

where:

n+1

n+1

sign(

g

)

N

u

n+1

K u = (G Tu G Tv )

anT naT ) + T

(

R

a

Bu ,

G T sign( g )

uTn+1

K v =

R

n+1

C.17

C.18

For each Gauss point, Eq. C.14 can be expressed based on the definition of gap functions as:

slip

m

nm

m

ccvr =

v ( v rnm v rm ) n v mr , n + ( v nm

r v r ) n ( d v r d v r ) n

+ v g vN (d v rnm d v rm ) n d v mr , n

C.19

n =

a

a

( u

+ x, ) n

Different terms of Eq. C.19 can be written in matrix form, such as:

237

C.20

(v

(d v

nm

r

nm

r

v rm ) n =

1 T T

a B v Vr nT ( BTv , U + c )

2

a

d v rm ) n =d VrT

1

B v anT ( BTv , U + c )

2

a

C.21

C.22

B v n nT BTv , Vr + 2 aT BTv Vr nT c +

vr slip

T

T T

d Vr u B v nn B v +

=

cc

K v Vr

1

T

g

B

n

B

an

c

vN v ,

v

2

1

B v n nT BTv , Vr + 2 aT BTv Vr nT c +

K u

+d VrT u g B n + 1 B anT c

vN

v ,

v

2

1

T T

T T

T T

a 2 ( B v na B v Vr n B v , + g vN B v an B v , )

C.23

slip

=

U

ccvr slip d VrT K vslip Vr + K vu

slip

with K vslip and K vu

defined in Eqs 3.101 and 3.102.

238

C.24

The End

- Catalogo de difusores serie 5000Uploaded byÁngeles Rodríguez
- Ch.08 FrictionUploaded byCK_85_3
- Advanced Mathematics &Mechanics Applications Using MATLABUploaded byHéctor Paprika
- Finite element method 2007 DecUploaded byBinoy Varghese
- Hughes Multiscale 1995Uploaded by'alvHie' Boudcjah Gathel II
- [Doi 10.1016%2Fj.finel.2004.07.001] Bao-Lin Wang; Zhen-Hui Tian -- Application of Finite Element–Finite Difference Method to the Determination of TraUploaded bySaiful Ahmad
- The influence of surface roughness and the contact pressure distribution on friction in rolling/sliding contactsUploaded bydoruoctaviandumitru
- Dormit Warangal DesignsUploaded byDaya Kiran
- CHAPTER 4.pdfUploaded byHani Barjok
- 2015_JCAM_LinLiuSadreUploaded byHassan Zmour
- WheelsUploaded byAlmadovar Alfonso
- 1-s2.0-S0043164899003154-mainUploaded bychawlakomal
- Undrained Stability of a Square Tunnel in a Soil Whose Strength Increases Linearly With DepthUploaded byHumberto Jorge
- Contoh tugas fatigue-fracture mech 2017.xlsUploaded bybagasferis
- Vizag Steel Plant Written Test Will Comprise General AwarenessUploaded byKishore Gv
- 1. Are Totally Recycled Hot Mix Asphalts a Sustainable Alternative for Road PavingUploaded byDiegoMartinVV
- PavementUploaded bySedano Jaim
- 45GTCUploaded byNaim
- UBC_1990_A1 K66Uploaded byRavi Khandelwal
- 28944_01Uploaded byJason Oliver
- SNM SyllabusUploaded bytml tml
- Local Time SteppingUploaded byroyb_84
- 001738Uploaded byAtul Khatri
- Force analysisUploaded bykürşat babul
- Result and Calculation PvtUploaded byMuhamad Baihakhi Shamsudin
- Paper or report.docxUploaded byJorge Iglesias
- BBA Certificate 08-4609 ACEGrid Geogrids for Basal Reinforcement (BBA, Third Issue, Nov 2013)Uploaded bysandycastle
- 1. Introductory pages (1).pdfUploaded bykankirajesh
- Ijege 11 Bs Deganutti Et AliiUploaded bytkujun
- Doc1Uploaded byTesfamichael Abathun

- steel tank designUploaded byJanasheen Bond
- 2012_Elect Des DetailsUploaded byArnab Das
- 2012 API Storage Tank Conference - Tank Standards UpdateUploaded bymortaza7094
- A GIS-based borehole data management and 3D visualization systemUploaded bymortaza7094
- Mohr CircleUploaded bymortaza7094
- Applied Soil Mechanics With ABAQUS ApplicationsUploaded bySoumya Bhattacharyya
- Response of Circular Flexible Foundations Subjected to HorizontalUploaded bymortaza7094
- 322 Settlement of Large Piled FoundationsUploaded byalvaedison00
- Boehm-Foundation Solutions for New TanksUploaded bymortaza7094
- Ens 110323 en Jz Rm Lecture 2011 Part 4Uploaded bymortaza7094
- ENS_080312_EN_JZ_RM_Lecture_2008_Part_6.pdfUploaded bymortaza7094
- Arzyabiye Asib Paziri Va Behsazi.pdfUploaded bymortaza7094
- Notes of Dr. DesaiUploaded byRed Orange
- NUMERICAL STUDY ON DAMAGE TO BURIED OIL PIPELINE UNDER LARGE FAULT DISPLACEMENTUploaded bymortaza7094
- Ppr10Uploaded byParamita Bhattacharya
- Soil Damping Formulation in Nonlinear Time Domain Site Response Analysis (Park, Hashash, 2003)Uploaded byEderEnzoTapiaBañez
- Soil Improvment(in persian)Uploaded bymortaza7094
- Planning and Site Investigation in TunnellingUploaded byJean Daly
- PART 05 - Rock Foundations and SlopesUploaded byattaelmula
- Tunnel Design FireProtectionSystemUploaded byhiimruju
- calculus of variationsUploaded bymortaza7094
- Solutions Manual Continuum Mechanics Lai 4th EdittionUploaded bydiego_dave
- Introduction to the Mechanics of a Continuous MediumUploaded byVictoria Maciel
- Total-and-Effective-Stress.pptUploaded bymortaza7094

- Chapter02-1.pdfUploaded byzerpthederp
- lecture490_ch6.pdfUploaded byTrần Tín
- Rafter Splice DesignUploaded bykalpanaadhi
- XmTRP-1703010Uploaded byAditya Whisnu Heryudhanto
- ME 205 Thermodynamics - CopyUploaded bynandan144
- (Developments in Petroleum Science 3) A.P. Szilas (Eds.)-Production and Transport of Oil and Gas-Academic Press, Elsevier (1975).pdfUploaded byanellbmc
- Chapter 2Uploaded byMarco Luigi
- Bearing Capacity ComparisonUploaded bylinamohdzhor4815
- 02_Scour and ProtectionUploaded byAlejandro Urquieta Quiroga
- Optimization of Intake System and Filter of an Automobile Using CFD AnalysisUploaded bykiran.76100
- Ch05-Distributed Forces (Centroids and Centers of Gravity)Uploaded bymantapto
- Effect Of A Circular Cylinder On Separated Forced Convection At A Backward-Facing StepUploaded byGo Go
- QUIZ 2 (2)Uploaded byAzraaei Azaharuddin
- Busbar CalculationUploaded byd_rosemberg
- TANAREKEMberly- Comprehensive Examination in Stattics of Rigid BodiesUploaded byjonathan
- DESIGN OF REINFORCED CONCRETE COLUMNS.pdfUploaded byYoshua Yang
- STAAD 2.14.10 Design ParametersUploaded byFWICIP
- Ansys Analysis-Trolley Beam 3T1Uploaded byYe Wint Thu
- Helicopterdynamics Chapter1Uploaded byAlexandreSidant
- bab 5 hal 71Uploaded bylimadua
- UBC_1987_A7 S53.pdfUploaded bywidayat81
- The_Comparison_of_the_Classical_Keplerian_Orbit_El.pdfUploaded byalvaroroman92
- D638Uploaded byNeville Camilleri
- Bolt Loading NotesUploaded byTarek Abulail
- Agitator Power Requirement and Mixing Intensity CalculationUploaded byBalaji Krishnan
- Eec 125 QuestionUploaded by27051977
- sir yakobUploaded byilham
- 2.1 Vectors and Scaler Distance and DisplacementUploaded bypixelhobo
- Project Navier StokesUploaded byYves Garnard
- TesUploaded byChandra Lutfi