Carneiro DL TM 15 PDF

A TIME-DEPENDENT AXIAL PIPE-SOIL INTERACTION MODEL FOR GLOBAL
ANALYSES OF SUBSEA PIPELINES
Daniel Lyrio Carneiro
Tese de Doutorado apresentada ao Programa de

Pós-graduação em Engenharia Civil, COPPE, da
Universidade Federal do Rio de Janeiro, como
parte dos requisitos necessários à obtenção do
título de Doutor em Engenharia Civil.
Orientadores: Gilberto Bruno Ellwanger

Fernando Artur Brasil Danziger
Rio de Janeiro
Março de 2015
TESE SUBMETIDA AO CORPO DOCENTE DO INSTITUTO ALBERTO LUIZ

COIMBRA DE PÓS-GRADUAÇÃO E PESQUISA DE ENGENHARIA (COPPE) DA
UNIVERSIDADE FEDERAL DO RIO DE JANEIRO COMO PARTE DOS
REQUISITOS NECESSÁRIOS PARA A OBTENÇÃO DO GRAU DE DOUTOR EM
CIÊNCIAS EM ENGENHARIA CIVIL.
Examinada por:
________________________________________________
Prof. Gilberto Bruno Ellwanger, D.Sc.
________________________________________________
Prof. Fernando Artur Brasil Danziger, D.Sc.
________________________________________________
Prof. Ian Schumann Marques Martins, D.Sc.
________________________________________________
Prof. José Renato Mendes de Sousa, D.Sc.
________________________________________________
Prof. Murilo Augusto Vaz, Ph.D.
________________________________________________
Prof. Fernando Schnaid, Ph.D.
RIO DE JANEIRO, RJ - BRASIL

MARÇO DE 2015
Carneiro, Daniel Lyrio
A Time-Dependent Axial Pipe-Soil Interaction Model
for Global Analyses of Subsea Pipelines/ Daniel Lyrio
Carneiro. – Rio de Janeiro: UFRJ/COPPE, 2015.
XXV, 152 p.: il.; 29,7 cm.
Tese (doutorado) – UFRJ/ COPPE/ Programa de
Engenharia Civil, 2015.
Referências Bibliográficas: p. 139-145.
1. Pipe-soil interaction. 2. Pipeline walking. 3. Finite
element analysis. I. Ellwanger, Gilberto Bruno et al. II.
Universidade Federal do Rio de Janeiro, COPPE, Programa
de Engenharia Civil. III. Título.
iii
For my loved Roberta, Rafa & Leo
iv
ACKNOWLEDGEMENTS
I would like to thank my advisors Professor Gilberto Ellwanger and Professor

Fernando Danziger for the encouragement, support and assistance, and for never giving
up on this often absent part-time student. I would also like to express my gratitude to my
“unofficial advisor” Professor David White, for the invaluable discussions, and for
always pushing me to go back and dig a bit deeper into my results. I also thank Professor
Ian Martins, whose inspiring lectures have spurred my interest for geotechnical
engineering.
I would not be able to complete this thesis while working without the time
flexibility and support given by my employers, and for that I thank in particular Liz
Lindahl and John Willcocks in Wood Group Kenny and Rafael Parrilha in Bureau Veritas.
I would also like to thank all my colleagues in Wood Group Kenny in Perth and ex-
colleagues in Bureau Veritas in Rio who, in any ways, assisted or cheered for me. In
particular, I thank Graham, Andy and Terry for the numerous discussions and ideas.
I am grateful to my extended family and my friends – the old friends back home,
as well as the new ones across the world. The words of support and incentive, regardless
of me often not being as available and close as a friend should be, are priceless. I thank
my new friends in Perth for filling in for family away from home. And for direct help in
the thesis I thank Lorena and Adriano for the proof reading, and Letícia and Luis for
reaching reference [74].
Finally, I am grateful to my loved wife Roberta, for her infinite support,

encouragement, patience and serenity. I would not have reached the completion of this
painful process without her strength and complicity, pushing me up every time I lost my
focus or courage. I also owe her my precious kids Rafael and Leonardo (and the third one
on the way). Their smiles when I arrived at home would “recharge my batteries” every
time.
v
Resumo da Tese apresentada à COPPE/UFRJ como parte dos requisitos necessários para
a obtenção do grau de Doutor em Ciências (D.Sc.)
UM MODELO DE INTERAÇÃO AXIAL DUTO-SOLO DEPENDENTE DO TEMPO

PARA ANÁLISES GLOBAIS DE DUTOS SUBMARINOS
Março/2015

Fernando Artur Brazil Danziger
Programa: Engenharia Civil
Este trabalho tem como foco a interação duto-solo no projeto de dutos submarinos
sujeitos a deslocamentos axiais cíclicos, em particular os sujeito ao fenômeno pelo qual
o deslocamento axial se acumula, conhecido como pipeline walking. O recente salto no
entendimento dos aspectos geotécnicos da complexa interação duto-solo (que envolve
drenagem parcial e ciclos de deformações plásticas) não foi acompanhado pelas
ferramentas disponíveis para análise global de dutos.
Esta tese apresenta um modelo novo, com o intuito de preencher esta lacuna. Este
consiste de um elemento finito formulado especificamente para esta aplicação,
permitindo capturar alguns dos principais aspectos do comportamento do solo, embora
sendo simples o suficiente para ser incluído em modelos globais de dutos (de maneira
similar a molas não-lineares). Dado o reduzido número de graus de liberdade, espera-se
que este modelo possa ser implementado com baixo custo computacional, apresentando
assim tempos de análise compatíveis com a prática de projeto.
Durante o desenvolvimento da tese, foram elaborados modelos em elementos

finitos do comportamento local do solo. Estes modelos não só foram úteis na validação
do novo modelo proposto, como também permitiram a observação de aspectos relevantes
relacionados à bidimensionalidade do adensamento e à redistribuição de poro-pressão
durante carregamentos parcialmente drenados. Tais aspectos são também apresentados e
discutidos.
vi
Abstract of Thesis presented to COPPE/UFRJ as a partial fulfillment of the requirements
for the degree of Doctor of Science (D.Sc.)

March/2015
Advisors: Gilberto Bruno Ellwanger

Department: Civil Engineering
This work is concerned with the pipe-soil interaction in the design of subsea
pipelines subjected to cyclic axial displacements, in particular those subjected to the
ratcheting process known as pipeline walking. Significant recent industry-led research
effort has improved the understanding of the complex soil response (involving partial
drainage and cyclic plasticity). The available pipeline analysis tools, however, are lagging
behind this evolution. The large scale difference between the local soil response and the
global pipeline behaviour hampers the direct cross-application of the latest knowledge
and models from both fields.
This thesis presents a novel approach to model the time-dependent pipe-soil

interaction, which intends to bridge this gap. It consists of bespoke finite elements which
capture key aspects of the complex soil response, whilst being simplified enough that they
can be included in global pipeline FE models (similar to non-linear springs). Given the
small number of degrees of freedom required, it is expected that this model can be
implemented with low computational cost, thus permitting global analyses within
timeframes that are compatible with design practice.
In the development of the thesis, local soil FE models were developed, which not
only served for validation of the proposed model, but also provided useful insight into
particular aspects of the soil response. Relevant observations concerning the two-
dimensionality of the consolidation process and the pore pressure migration under
partially drained loading are also presented and discussed.
vii
Table of Contents
1. Introduction ................................................................................................................ 1

1.1. Background ..................................................................................................... 1
1.2. Motivation ...................................................................................................... 4
1.3. Objectives ....................................................................................................... 5
1.4. Thesis Organisation ........................................................................................ 7
2. A Brief Review of Pipeline Walking .......................................................................... 8

2.1. Background ..................................................................................................... 8
2.2. Basic Mechanics of Pipeline Expansion and Contraction .............................. 9
2.3. Effect of Global Buckling and Other Discontinuities .................................. 14
2.4. Pipeline Walking .......................................................................................... 17
3. A Brief Review of Couple Consolidation FE Analysis ............................................ 24

3.1. Background ................................................................................................... 24
3.2. General Formulation ..................................................................................... 25
3.3. Formulation in FE ......................................................................................... 27
3.4. The MCC Constitutive Model ...................................................................... 31
4. Review of the Existing Literature on Axial PSI ....................................................... 39

4.1. Background ................................................................................................... 39
4.2. Time-Independent Axial PSI ........................................................................ 39
4.2.1. Drained and Undrained Resistance...................................................... 39
4.2.2. Mobilisation and Response Curve ....................................................... 41
4.3. Time-Dependent Axial PSI .......................................................................... 43
4.3.1. Set-up ................................................................................................... 43
4.3.2. Rate Dependency ................................................................................. 45
4.3.3. Consolidation Hardening ..................................................................... 47
4.3.4. Overarching Frameworks .................................................................... 50
4.3.5. Continuum 3D Solid FE Models ......................................................... 53
5. The Problem of Properly Modelling Time-Dependent Axial PSI in Global Pipeline
Analyses ......................................................................................................................... 55
5.1. Background ................................................................................................... 55
viii
5.2. The Reasons for Pursuing a Coupled Model ................................................ 55
5.3. The Reasons for Introducing Some Extra Complexity ................................. 58
6. Two-Dimensional Consolidation beneath a Pipeline ............................................... 63

6.1. Background ................................................................................................... 63
6.2. FE Model, Results and Discussion ............................................................... 67
7. Pore Pressure Migration ........................................................................................... 77

7.1. Background ................................................................................................... 77
7.2. Model Details and Run Cases ....................................................................... 79
7.3. Results for Uniform Cases ............................................................................ 85
7.4. Selected Results for Cases with an Overconsolidated Layer ....................... 91
8. Novel Model ........................................................................................................... 103

8.1. Background ................................................................................................. 103
8.2. 3D Model Reduced to 1D FE Mesh ........................................................... 107
8.3. Circumferential Drainage ........................................................................... 113
8.4. Application to Partially Embedded Pipelines ............................................. 118
9. Final Remarks ......................................................................................................... 130

9.1. Conclusions ................................................................................................ 130
9.2. Suggestions for Future Research ................................................................ 134
10. References .............................................................................................................. 139
Appendix A. Additional Results from the Axisymmetric Model of Section 7 ............ 146
ix
List of Figures
Figure 1.1 – Example of offshore oil and gas production systems and nomenclature
(nomenclature indicated on figure reproduced from [3]†). ....................................... 2
Figure 1.2 – Different cross sections: (a) single wall pipe, (b) thermal-insulated pipe, (c)
concrete-coated pipe (reproduced from [4]); (d) pipe-in-pipe (reproduced from [5]);
(e) pipe bundle (reproduced from [6]); (f) flexible pipe, (g) umbilical cables
(reproduced from [7]). .............................................................................................. 3
Figure 2.1 – Diagrams of effective axial force along a pipeline on horizontal seabed for
four different temperature increments. ................................................................... 10
Figure 2.2 – Diagrams of axial displacements along a pipeline on horizontal seabed for
four different temperature increments. ................................................................... 12
Figure 2.3 – Diagrams of axial strain and thermal expansion along a pipeline on horizontal
seabed for four temperature increments. ................................................................ 13
Figure 2.4 – Diagrams of effective axial force along a pipeline on horizontal seabed when
temperature is reduced back to the initial value. .................................................... 14
Figure 2.5 – Diagram of effective axial force along a pipeline with two global buckles
when the temperature is elevated (expansion); and then brought back down
(contraction). .......................................................................................................... 16
Figure 2.6 – Temperature transient obtained from a dynamic flow simulation

benchmarked against recorded process temperatures (reproduced from [9], courtesy
ASME). ................................................................................................................... 19
Figure 2.7 – Simplified temperature transient during pipeline heat-up. ......................... 19
Figure 2.8 – Axial force diagrams during heat-up with temperature transient............... 20
Figure 2.9 – Axial force diagrams after expansion and contraction for a sloping seabed.
................................................................................................................................ 22
x
Figure 3.1 – Critical state framework: e–p'–q space and the yield surface (reproduced†
from [27], courtesy Cambridge University Press). ................................................. 33
Figure 4.1 – Axial resistance-relative displacement curve and idealised linear elastic
mobilisation (reproduced† from [40]). .................................................................... 42
Figure 4.2 – Project-specific axial PSI response curve proposed by BRENNODDEN &
STOKKELAND [50] (reproduced from [50]†). ..................................................... 43
Figure 4.3 – Design chart for axial pipe-soil interaction proposed by JEWELL &
BALLARD [59] (reproduced from [59]†). ............................................................. 46
Figure 4.4 – Cycles of loading intervened by consolidation periods illustrating the

consolidation hardening process (reproduced from [64]†). .................................... 48
Figure 4.5 – Shear zone migration away from the pipe-soil interface (reproduced from
[8]). ......................................................................................................................... 49
Figure 4.6 – Profiles of shear strength after steps of consolidation hardening obtained by
YAN et al. [67] (reproduced from [67], courtesy ICE Publishing)........................ 50
Figure 4.7 – Mechanisms affecting axial PSI according to HILL et al. [61] and WHITE
et al. [62] (reproduced from [61]). ......................................................................... 51
Figure 4.8 – Results of back-analysis of the model test in [14] using the analytical model
proposed by RANDOLPH et al. [68] (reproduced from [68], courtesy ICE
Publishing). ............................................................................................................. 52
Figure 4.9 – Infinite pipe modelling scheme adopted by RANDOLPH et al. [68]
(reproduced from [68], courtesy ICE Publishing). ................................................. 53
Figure 4.10 – Stress–volume path including pipe embedment, set-up and twenty cycles
of axial displacements intervened by full consolidation, obtained by YAN et al. [67]
(reproduced from [67], courtesy ICE Publishing). ................................................. 54
Figure 5.1 – Effective axial force diagram for coupled (“v50 = vk”) and uncoupled
(“F50”) analyses (reproduced from [16], courtesy ASME). .................................. 56
xi
Figure 5.2 – Hybrid modelling scheme using 2D soil slices proposed by MARTIN et al.
[70] for lateral PSI (reproduced from [70], courtesy ICE Publishing). .................. 58
Figure 5.3 – Curves of excess pore pressure distribution at particular instants (isochrones)
along a vertical line beneath the pipe invert, obtained from plane strain set-up
analyses by JEWELL & BALLARD [59] (reproduced from [59]†). ..................... 61
Figure 5.4 – Schematic of the planar analogy by RANDOLPH et al. [68], indicating the
parabolic distribution of excess pore pressure distribution below the shear band next
to the pipe-soil interface (reproduced from [68], courtesy ICE Publishing). ......... 61
Figure 6.1 – Results by GOURVENEC et al. [49] of excess pore pressure u at the pipe
invert and degree of consolidation U, (obtained from their FE model and from
Equation ( 6-1 )) (reproduced from [49]†). ............................................................. 66
Figure 6.2 – FE model geometry and boundary conditions. .......................................... 68
Figure 6.3 – Vectors of apparent pore water flow velocity at three distinct instants:
(a) cv t / D2 = 0.02, (b) cv t / D2 = 0.1 and (c) cv t / D2 = 0.9. Illustrative arrows
highlight the two components (radial and circumferential) of the flow next to the
pipe-soil interface. .................................................................................................. 71
Figure 6.4 – Excess pore pressure along a vertical line beneath the pipe invert at five
distinct instants: (i) cv t / D2 = 0.001, (a) cv t / D2 = 0.02, (b) cv t / D2 = 0.1,
(c) cv t / D2 = 0.9 and (f) cv t / D2 = 10. Dashed lines indicate approximate gradient
next to pipe-soil interface. ...................................................................................... 73
Figure 6.5 – Normalised results of excess pore pressure along arcs of constant  at
particular instants t. ................................................................................................ 74
Figure 6.6 – Degree of consolidation (average along pipe-soil interface) over time
obtained from the 2D FE analysis results, as well as from equations ( 6-1 ), ( 6-2 )
and ( 6-4 ). .............................................................................................................. 76
Figure 7.1 – Section through bank slip studied by WARD et al. [71]. Top: post failure
geometry, indicating piezometric heads two months after the slip; Bottom: original
xii
profile, indicating active and passive thrust wedges considered in their back-analysis
(reproduced from WARD et al. [71], courtesy ICE Publishing). ........................... 78
Figure 7.2 – Isochrones along the peat layer in the back-analysis by WARD et al. [71],
indicating significant excess pore pressure redistribution (reproduced from WARD
et al. [71], courtesy ICE Publishing). ..................................................................... 79
Figure 7.3 – Axisymmetric model schematics, in particular a case with overconsolidated

soil adjacent to the pipe and normally consolidated soil elsewhere. ...................... 80
Figure 7.4 – Maximum observed values of (a) deviator stress qmax and (b) excess pore
pressure umax at the nodes at the pipe-soil interface as these moved with the
prescribed constant velocities v. ............................................................................. 86
Figure 7.5 – Results of the analyses with vD/cv = 0.1, for R = 1.0 and R = 1.4: (a) stress–
volume paths (for both total stress p and effective stress p') in the e–ln p / e–ln p'
space, including the SRL corresponding to the initial conditions; (b) stress paths in
the p–q / p'–q space, including the initial projected yield functions; (c) q and u for
axial displacement . .............................................................................................. 88
Figure 7.6 – Results of the analyses with vD/cv = 1, for R = 2.0 and R = 2.4: (a) stress–
volume paths (for both total stress p and effective stress p') in the e–ln p / e–ln p'
space, including the SRL corresponding to the initial conditions; (b) stress paths in
the p–q / p'–q space, including the initial projected yield functions; (c) q and u for
axial displacement . .............................................................................................. 90
Figure 7.7 – Maximum values, observed at the pipe-soil interface, of (a) deviator stress
qmax (including expected values calculated using Equation ( 7-2 ) and curve-fit from
Figure 7.4) and (b) excess pore pressure umax at the nodes at the pipe-soil interface
as these moved with the prescribed constant velocities v – analyses with initial
R = 2.4 in overconsolidated layer and R = 1.0 elsewhere (including benchmarking
cases for these two values of R).............................................................................. 94
Figure 7.8 – Results of the analyses with vD/cv = 1, for R = 2.4 within overconsolidated
layer of hoc/D = 0.4 and R = 1.0 elsewhere: (a) stress–volume paths (for both total
stress p and effective stress p') in the e–ln p / e–ln p' space, including the SRL
xiii
corresponding to the initial conditions; (b) stress paths in the p–q / p'–q space,
including the initial projected yield functions; (c) q and u for axial displacement .
................................................................................................................................ 98
Figure 7.9 – Results of the analyses with vD/cv = 0.1, for R = 2.4 within overconsolidated
.............................................................................................................................. 100
Figure 7.10 – Results of the analyses with vD/cv = 3, for R = 2.4 within overconsolidated
.............................................................................................................................. 102
Figure 8.1 – Cylindrical coordinate system (, , x) and vertical direction z. ............. 105
Figure 8.2 – Results of the implemented 1D bespoke element (first order and second order
elements; linear and backward interpolation) compared to ABAQUS results using
the model of Section 7: (a) volume–stress paths in the e–p' (normalised by the initial
mean effective stress p'0) space; (b) stress paths in the q–p' space; (c) q and u versus
axial displacement  normalised by the pipe diameter D. .................................... 112
Figure 8.3 – Cross-section of (a) axisymmetric and (b) modified FE models. ............ 113
Figure 8.4 – Angles θ' and θ* for any given arc of constant ρ in the modified FE model.
.............................................................................................................................. 114
Figure 8.5 – Thin ring arc results obtained with the implemented model and analytical
expression of the degree of consolidation for initial distribution of excess pore
pressure following a cosine distribution. .............................................................. 117
xiv
Figure 8.6 – Results of average excess pore pressure along arcs of constant  over time t,
from the implemented model (including the modification for considering
circumferential drainage) and from the ABAQUS model in Section 6................ 119
Figure 8.7 – Isochrones along a vertical line beneath the pipe invert, obtained from the
implemented model (including the modification for considering circumferential
drainage) and from the ABAQUS model in Section 6. ........................................ 120
Figure 8.8 – Cyclic axial displacements results: (a) axial resistance F normalised by
pipeline submerged unit weight W; and (b) excess pore pressure ratio ru. Time t
restarted at the beginning of every new sweep for clarity. ................................... 123
Figure 8.9 – Results from Figure 8.8 (plus axial resistance for one fully drained sweep)
overlaid (in light-green) on corresponding figures reproduced from YAN et al. [67]
(background figures courtesy ICE Publishing). ................................................... 124
Figure 8.10 – Maximum excess pore pressure ratio ru observed in each axial sweep in the
implemented model, and in the 3D FE analysis by YAN et al. [67]. ................... 127
Figure 8.11 – Results of drainage index  over cycles, compared to the results of the
expression proposed by YAN et al. [67] – Equation ( 8-28 ) – and WHITE et al. [84]
– Equation ( 8-29 ). ............................................................................................... 127
Figure 8.12 – Results from the implemented model: (a) volume–stress path in the e–p'
space; (b) stress paths in the q–p' space. .............................................................. 129
Figure A.1 – Results of the analyses with vD/cv = 0.1, for R = 2.0 within overconsolidated
.............................................................................................................................. 148
xv
.............................................................................................................................. 149
.............................................................................................................................. 150
Figure A.4 – Results of the analyses with vD/cv = 3, for R = 2.4 within overconsolidated
.............................................................................................................................. 151
Figure A.5 – Results of the analyses with vD/cv = 1, for R = 2.0 within overconsolidated
.............................................................................................................................. 152
xvi
List of Symbols
A pipeline cross-sectional area [L2]†
A first of four parameters in the explicit expression for D
B strain – displacement matrix
Bv volumetric strain – displacement matrix
C integration constant
D pipeline external diameter (including any coating if applicable) [L]
D50 mass median grain diameter of the soil [L]
D tangent elastoplastic stress – strain matrix
Del elastic parcel of the tangent stress – strain matrix
E Young’s modulus of pipeline (or equivalent modulus over A) [FL–2]
F, Fmax soil resistance (per unit length of pipeline) to pipeline axial movement and
its maximum value [FL–1]
G shear modulus of soil [FL–2]
H burial depth at pipeline centreline [L]
H, Hi interpolation matrix and its partial derivatives (i = ρ, θ, x, y or z)
J scalar Jacobian
K bulk modulus of soil [FL–2]
K0 coefficient of lateral earth pressure at rest
†
Dimensions of scalar quantities, when applicable, are presented between brackets. For simplicity, the
derived quantity force [F] = [LMT–2] is used along with the SI base quantities length [L], mass [M], time [T]
and temperature [Θ].
xvii
K, Ke tangent elastoplastic stiffness matrix of an FE model and of an element
Ki submatrix of K relating δfi to δξi (i = ρ or x)
Kpl submatrix of K relating δfi to δξj (i = ρ and j = x; or vice versa)
Ku, Kue transmissivity matrix of an FE model and of an element
K *u modified transmissivity matrix accounting for radial drainage
Kv, Kve coupling matrix of an FE model and of an element
L pipeline length [L]
P Effective axial force in a pipeline (positive in compression) [F]
Q matrix of artificial transmissivity accounting for radial drainage
R overconsolidation ratio in terms of mean effective stress
S ratio of total to effective average normal stress in first sweep
T dimensionless time
U degree of consolidation
W pipeline submerged unit weight [FL–1]
a second of four parameters in the explicit expression for D
b third of four parameters in the explicit expression for D
cv coefficient of consolidation [L2T–1]
d prefix “infinitesimal increment of-”
d fourth of four parameters in the explicit expression for D
e void ratio
ek intercept of the SRL in the e–ln p' space, for a given R
xviii
f ratio of interface friction to internal soil friction (angle or coefficient)
f vector of nodal external forces
fi vector of component of nodal external forces (i = ρ or x)
g vector of nodal external flow of pore water
h characteristic element size [L]
hoc thickness of overconsolidated layer [L]
i, j indices
k soil permeability [LT–1]
m exponent in hyperbolic curve fits or index of summation
mv coefficient of volume change [L2F–1]
n sweep number
nΓ unit vector normal to the domain surface
nne number of nodes in one element
nn total number of nodes in the model
p mean total stress (positive in compression) [FL–2]
p' mean effective stress (positive in compression) [FL–2]
p'c pre-consolidation stress [FL–2]
p'0 initial value of p' [FL–2]
q deviator invariant of the stress tensor [FL–2]
qmax maximum observed value of q [FL–2]
qmaxD, qmaxU qmax in fully drained and fully undrained conditions [FL–2]
xix
qmaxA, qmaxB qmax at locations A and B from benchmark cases [FL–2]
r local natural coordinate in isoparametric mapping
ru excess pore pressure ratio
t time [T]
t50 time required for dissipation of 50% of u0 [T]
u excess pore pressure [FL–2]
ui excess pore pressure beneath pipe invert [FL–2]
umax maximum (absolute value) observed u [FL–2]
u average excess pore pressure [FL–2]
u0 initial value of u [FL–2]
u* average excess pore pressure along interval [–*, *] [FL–2]
u nodal excess pore pressure vector
v pipe velocity [LT–1]
vw pore water flow velocity
vi component of pore water flow velocity (i = ρ, θ, x, y or z) [LT–1]
vθ* circumferential component of pore water flow velocity at * [LT–1]
w pipeline embedment [L]
x axial coordinate [L]
y coordinate [L]
z vertical coordinate [L]
Γ domain surface area [L2]
xx
 temperature difference from the initial ambient temperature [Θ
 stress ratio on the CSL
 intercept of the NCL in the e–ln p' space
 walking driving force to available resistance ratio
 yield function
p projected yield function
 drainage index
, e volume of FE model domain and of an element [L3]
 coefficient of linear thermal expansion [Θ–1]
u adhesion factor
 time integration parameter
ij shear strain component (i or j = ρ, θ, x, y or z)
w unit weight of pore water [FL–3]
' effective unit weight of soil [FL–3]
 prefix “small increment of-”
 pipeline axial strain (average over cross section, positive shortening)
i normal strain component (i = ρ, θ, x, y or z, positive shortening)
v volumetric strain (positive contracting)
vp, sp plastic parcel of volumetric and (octahedral) shear strain
, p strain tensor and its plastic parcel
 wedging factor
xxi
 stress ratio
 angular coordinate
' angular coordinate of the seabed surface (measured from the pipe bottom
invert)
* angular coordinate defining arbitrary soil wedge
 slope of the SRL in the e–ln p' space
 slope of the CSL and the NCL in the e–ln p' space
 Poisson’s ratio
ξ pipeline axial displacement [L]
ξF pipeline axial displacement at Fmax [L]
ξmob mobilisation distance [L]
ξi displacement component (i = ρ, θ, x, y or z) [L]
ξij nodal displacement component (i corresponds to the node number and

j = ρ, θ, x, y or z) [L]
ξ, ξe nodal displacement vector of an FE model and of an element
ξi vector of component of nodal displacement (i = ρ or x)
 radial coordinate [L]
e  i
element nodal (radial) coordinate (i = 1, 2, 3 corresponds to the element
node number) [L]
e vector of element nodal (radial) coordinates
i normal components of total stress (i = ρ, θ, x, y or z)† [FL–2]
†
All normal stress components are positive in compression.
xxii
n total normal stress† [FL–2]
v total vertical stress† [FL–2]
n average normal stress† [FL–2]
 total stress tensor
'i normal components of effective stress (i = ρ, θ, x, y or z)† [FL–2]
'n effective normal stress† [FL–2]
'v effective vertical stress† [FL–2]
 n average effective normal stress† [FL–2]
' effective stress tensor
 proportionality factor in the associated plasticity
ij shear stress component (i or j = ρ, θ, x, y or z) [FL–2]
 x i
shear stress component ρx at node i [FL–2]
' soil effective angle of internal friction
 seabed slope angle
, e a generic quantity and the vector of its nodal values within an element
 gradient operator
2 Laplace operator
 partial derivative prefix
†
All normal stress components are positive in compression.
xxiii
List of Acronyms
1D, 2D, 3D one-, two- and three-dimensional
ALA American Lifelines Alliance
API American Petroleum Institute
ASME American Society of Mechanical Engineers
BS British standard
COPPE Alberto Luiz Coimbra Institute – Graduate School and Research in

Engineering
CS critical state
CSL critical state line
DOF degrees of freedom
DNV Det Norske Veritas
EN European standard
FE finite element
ISO International Organization for Standardization
JIP joint industry project
MCC modified Cam clay
MPC multi-point constraint
N/A not applicable
NC normally consolidated
NCL normal consolidation line
xxiv
OC overconsolidated
OCR overconsolidation ratio in terms of effective vertical stress
OS offshore standard
PSI pipe-soil interaction
RP recommended practice
SI International System of Units
SRL swelling and reconsolidation line
UFRJ Federal University of Rio de Janeiro
UK United Kingdom
xxv
1. Introduction
1.1. Background
Subsea pipelines are the arteries of offshore oil and gas production systems [1, 2],
conveying produced hydrocarbons and other fluids between subsea wells and processing
facilities (on land or offshore). Pipelines are also one of the two options for transporting
processed product to the final costumer (the alternative being moving batches, for
example stored in tanks). As oil and gas production moves to deeper water and harsher
environments, having processing facilities just above every well becomes impractical or
uneconomic. Increasingly complex networks of subsea infrastructure – such as the one
illustrated in Figure 1.1 – are hence being adopted for offshore oil and gas production, in
which pipelines play a major role.
Pipelines are slender structures, typically with dimensions in the order of

magnitude of 0.1 m to 1 m in their cross-section, and of 1 km up to 1000 km in length.
They run at about the seabed elevation: on the seabed (often partially embedded) or
shallowly buried; and eventually spanning just above the seabed over limited lengths.
Exceptions are risers (which are sections of a pipeline between the seabed and the topside
of an offshore facility) or other short suspended elements (e.g. buoyant offloading hoses
or freestanding jumpers). Hence, their structural response to any loading is highly
dependent on the interaction with the seabed soil.
1
Floating
Tanker platform
Offloading hose
Midline
structure
Risers
Infield
flowlines Electro-hydraulic
Tie-in umbilicals
spool
Well
Manifold Export End
pipeline structure
Figure 1.1 – Example of offshore oil and gas production systems and nomenclature (nomenclature
indicated on figure reproduced from [3]†).
Pipelines exist in a diversity of types, ranging from the simplest steel tubular
through to composite flexible structures or bundle configurations, some of which are
illustrated in Figure 1.2. All these different structures have, however, the same essential
function: to maintain an open channel for the products to flow, while ensuring that no
product is released. In performing their main function, pipelines must also withstand
loadings to which they might be subjected over their design life, which may be induced
by the carried products (e.g. pressure and temperature), the environment (e.g.
hydrodynamic drag) or other external agents (e.g. contact with fishing gear).†
†
Copyright 2010, Offshore Technology Conference. Reproduced with permission of OTC. Further
reproduction prohibited without permission.
2
(a) (b) (c)
(d) (e) (f) (g)
Figure 1.2 – Different cross sections: (a) single wall pipe, (b) thermal-insulated pipe, (c) concrete-coated
pipe (reproduced from [4]); (d) pipe-in-pipe (reproduced from [5]); (e) pipe bundle (reproduced from [6]);
(f) flexible pipe, (g) umbilical cables (reproduced from [7]).
When the temperature and / or pressure of the carried fluid increase, a pipeline
will tend to expand. When not fully free to expand, it will attract reactions – both locally
at appurtenances (e.g. end or midline structures) and distributed along the pipe-soil
interface – and develop axial compression. Expansion will also cause curved sections to
bend further. Determining the reactions and internal forces each region of a pipeline will
endure, is a complex hyperstatic problem that needs to be carefully addressed in design,
as excessive local strains could cause the pipeline to fail in performing its key function,
either due to high deformations, reducing the available cross section, or – more
dangerously – causing it to release product.
The pipe-soil interaction (PSI), which plays a major role in the hyperestatic
problem, is beyond the limits of conventional foundation engineering [8]. Within the
expansion process, the pipeline and appurtenances are expected to remain in the elastic
range (or to undergo limited local plastic deformation). The soil, however, shall allow the
pipe to move (axially as it expands, and laterally at bends), in processes that, in
conventional foundation engineering, would be classified as a “failure”. Most of the
knowledge developed within this almost one century of modern soil mechanics focused
3
on designing structures sufficiently far from the “failure”, whereas the design of subsea
pipelines requires understanding the soil response way past this point.
Pipelines are eventually shutdown. Under these circumstances, the internal fluids
will gradually cool down as they lose temperature to the environment. Their pressure
might eventually be reduced as well. Under these conditions, a pipeline would tend to
contract back to its original length. As the expansion had caused irrecoverable plastic
deformation at the pipe-soil interface and / or within the soil adjacent to it, the soil will
respond in a hysteretic way, resisting to this contraction. As a result, the “unloaded”
pipeline will present axial tension. A similar hysteretic response will be observed when
the pipeline is brought back into operation, presenting axial compression again.
Asymmetries in pipeline expansion and contraction can induce a ratcheting

process, causing it to present a net axial movement towards one end after each
shutdown / restart cycle. Hence the original load sharing – when the pipeline first
expanded – will change over time, as the solution of the hyperstatic problem will differ
each time the pipeline expands (having accumulated a different amount of net axial
movement each time).
1.2. Motivation
The ratcheting process introduced in the previous section has caused a tie-in spool
to rupture after a few years of operation, releasing hydrocarbons to the environment in
the late 1990’s, in the UK sector of the North Sea [9]. Whilst it is known for a long time
that many pipelines present increasing axial displacement over time [10], the mechanism
proposed by TØRNES et al. in 2000 [9] was the first to provide a sound explanation for
the phenomenon (being later complemented with the introduction of additional driving
mechanisms [11–13]). They originally called it “pipeline creep”, but the industry later
more widely adopted the term “pipeline walking”.
The rationale supporting the mechanisms described in [9, 11–13], while sound,
makes use of a rudimentary axial PSI model: a simple Coulomb-like friction (often
combined with a linear-elastic pre-sliding compliancy). Physical model tests – in
laboratory using reconstituted natural soil; as well as in situ – have proven however that
4
time-dependent aspects (e.g. pipe velocity and intervals between movements) can cause
the resistance to span an order of magnitude within a single test [14].
The use of conservative “friction factors” to cover such a scatter in axial resistance
results in most (if not all) of the pipelines subjected to high temperature and high pressure
being identified as prone to some degree of walking. This, combined with the fact that
pipelines typically undergo hundreds of cycles over their design life (and hence even very
small values of net displacement per cycle can become a major design issue), have made
pipeline walking a costly hassle in the design, installation and monitoring of most subsea
pipelines.
Still, whilst evidences of increasing axial displacements continue to be observed

in many pipelines (including those designed before the awareness of the existence of this
phenomenon), the failure that instigated the work of TØRNES et al. [9] is the only failure
ever attributed to pipeline walking. This suggests that the current design practice may
lead to overconservative (and overcostly) pipelines.
Of course pipeline walking is not the only design aspect in which conservatisms
may conceal potential cost savings. Yet, whilst seldom a major fraction of the total cost
of a pipeline, pipeline walking mitigation costs can often be significant. As oil and gas
production progresses to less lucrative reserves under harsher conditions and at remoter
locations, improving the design to pipeline walking can help in making currently
uneconomic projects viable.
1.3. Objectives
The understanding of the geotechnical aspects of axial pipe-soil interaction has

significantly progressed over the last few years, as evident from the literature review
presented in Section 4. Still many questions remain unanswered and many aspects require
further investigation. The first objective of this thesis is to contribute to this
understanding. To this purpose, two aspects of the time-dependent soil response are
scrutinized using detailed local finite element (FE) analyses. These are: the two-
dimensional consolidation beneath a pipeline; and the redistribution of excess pore
pressure between regions of soil in different states (due to the loading history). Some of
5
the results and conclusions obtained from these numerical investigations have been
accepted for publication in [15].
The significant ongoing progress in understanding the geotechnical processes and

the local soil behaviour (mentioned earlier), however, has not yet been followed by
similar evolution in the available pipeline analysis tools. Recent projects have been
assessing complex aspects of the time-dependent PSI, to then lump them all into
“equivalent” constant frictional resistances. A recent paper published by the candidate
and other co-authors [16] argues that this approach is inadequate, as different sections of
the same pipeline will see different displacements, under different velocities, as the pipe
expands and contracts. Results of pipeline walking analyses, using velocity-dependent
resistance only, demonstrate the importance of assessing the soil response at each
location, in a coupled global model.
It is recognised, however, that extrapolating from the local geotechnical FE

models to global pipeline FE models is impractical. This is due to the large scale
difference between the mesh required to model the continuum underlying soil; and the
pipeline slender bar behaviour (with respective dimensions in the aforementioned orders
of magnitude). On the other hand, analytical PSI models – such as the velocity-dependent
one employed in [16] – have limited capabilities, not being able to handle the complex
aspects involved in the time-dependent soil response.
The final objective of the present work is thus to develop a new axial PSI model,
capable of properly reproducing the time-dependent soil response whilst being
computationally cheap enough so that it could be implemented in a pipeline global FE
analysis. The development of the proposed model and its validation against results of
detailed local FE models – both developed within the present work and published by
others – are presented (having also been submitted for publication in [17]). The final
implementation into a global analysis tool, however, requires further refinement in the
numerical techniques, which were deemed beyond the scope of the thesis, and left for
future research.
6
1.4. Thesis Organisation
Following the present introductory section, the thesis is organised as follows:
Section 2 presents a brief review of the mechanics of pipeline walking, providing

the reader with a basic understanding of the global pipeline behaviour under cyclic
temperature / pressure loading (whilst not focusing on the PSI).
Section 3 briefly reviews key aspects of soil mechanics that are later employed
along the thesis – namely: consolidation; and critical state constitutive model – focusing
on their formulation in FE. These two initial sections give background information on the
two fields across which the present work spans.
Section 4 presents a review of the existing axial PSI models, contextualising the
remarkable progress over the last few years.
Section 5 formally states the ultimate objective of the thesis, providing

justification for the need and discussing the value of developing the proposed model.
The sections then following capture the two contributions to the understanding of
the geotechnical processes involved in the time-dependent axial PSI, presenting the local
FE analyses performed and discussing their relevant results. Section 6 discusses aspects
of the two-dimensional consolidation beneath a pipe.
Section 7 discusses the radial pore pressure migration between regions of soil in
different states, and its potential effects in the late development of excess pore pressure
at the pipe-soil interface.
Section 8 presents the novel model, describing its development, formulation and
implementation in symbolic programming. It also presents the validation exercises
performed, against the FE models developed for Sections 6 and 7, as well as published
results by others.
Section 9 presents the final remarks: conclusions of the present work, a summary
of the key contributions and suggestions for future research.
7
2. A Brief Review of Pipeline Walking
2.1. Background
This section presents a brief review of the mechanics of pipeline walking,

introducing basic concepts required to understand the thesis. The intention is to provide
a complete didactic description, which is based on the original works of TØRNES et al.
[9] and CARR et al. [12]. In order to focus only on the most relevant aspects, this section
makes use of a number of key simplifications.
The simplifications adopted throughout Section 2 are:
 The pipeline is essentially straight, initially unloaded, on a non-deformable

flat seabed (either horizontal or presenting a constant slope along the
pipeline length).
 The soil resistance to pipeline axial movement is constant, with no pre-slip

mobilisation (i.e. axial displacement is zero until a maximum soil
resistance – per unit length of pipeline – of F is reached, and, after that,
the pipe will move under constant resistance F – with direction always
opposed to the local increment of displacement).
 Changes in internal pressure are negligible, pipeline expansion and

contraction is due to change in temperature only (for the effects of internal
pressure in the so-called “effective axial force” refer to [18]).
 Temperature is constant along the pipeline length, except during “non-

uniform heat-up” (as discussed later, in which the transient is
approximated by linear variation along the pipeline length). Temperature
Θ is measured as change in relation to the initial ambient temperature (i.e.
before operation, Θ = 0).
 No local reactions from appurtenances, hence pipeline ends are free to

move under negligible end reaction.
8
2.2. Basic Mechanics of Pipeline Expansion and Contraction
As the temperature Θ gradually increases, a pipeline will tend to expand. If the

pipeline were fully free, it would expand to a total length of L(1+) where L is its initial
length and  its coefficient of linear thermal expansion. If on the other hand it were fully
fixed, it would maintain its original length, presenting a compressive axial force P. The
compressive (mechanical) axial strain to which the pipeline would be subjected is given
by:
P
 ( 2-1 )
EA
where EA is the axial stiffness of the pipeline cross section. As the pipe is not allowed to
expand, L(1+–) = L such that:
P  EA ( 2-2 )
If the pipeline is on a horizontal seabed, considering all the simplifications set in

Section 2.1, it will expand, mobilising soil resistance. As illustrated in Figure 2.1, the
compressive axial force will build up from zero at x = 0. Because the soil resistance F is
constant, the compressive force increases linearly with x (i.e. P = Fx for small values of
x/L). The soil resistance in this initial region point towards the +x direction, opposed to
the pipeline axial displacement  which is negative (i.e. towards the –x direction).
Symmetrical forces and displacements are observed next to the other pipeline end, at
x = L.
9
>0
compressive axial force EA4
<0 >0
EA3
P(3) = P(4)
EA2
P(2)
EA1
P(1)
F
0
0 0.25 L 0.5 LL
0.5 0.75LL
0.75 1. LL
x
Figure 2.1 – Diagrams of effective axial force along a pipeline on horizontal seabed for four different
temperature increments.
For low temperature increments, the increasing axial compression will reach a
maximum value, equivalent to the fully constrained condition in which the compressive
force is given by Equation ( 2-2 ). At this point, the compressive mechanical axial strain
 will be equal to the thermal expansion . Just past this point, the pipeline will present
no net elongation, and thus no tendency to move axially. This corresponds to the
horizontal regions of the compressive force diagrams for the small temperatures 1 and
2 in Figure 2.1. For higher temperatures, however, the linear axial force building up
from one end might intercept that from the other end before the “fully restrained” axial
compression is achieved, such as observed in Figure 2.1 for 4.
TØRNES et al. [9] have proposed calling pipelines “long” if they are long enough
for the “fully restrained” force to be reached, and “short” otherwise. Some confusion often
arises from the use of these labels. Although they are length adjectives, they categorize a
pipeline not only by its length, but by the relationship between its length, stiffness and
coefficient of thermal expansion, the soil resistance and changes in temperature (not to
mention the influence of internal pressure, global buckles, and axial resistance at end
appurtenances). Hence, the quotation marks will be used in this thesis every time the
labels refer to the classification proposed in [9]. A “long” pipeline is a pipeline where:
10
2EA
L ( 2-3 )
F
The categorisation can be done (with no confusion introduced by semantics) using

a dimensionless parameter measuring the ratio of the walking driving force to available
soil resistance:
EA
 ( 2-4 )
LF
A “short” pipeline will present   0.5 while a “long” pipeline  < 0.5. This is analogous
to the “constraint friction” proposed by CARR et al. [12], and will permit introducing a
third category, as they did, which is discussed later in this section.
The displacements along a pipeline can be obtained by integrating the net

expansion:
      dx ( 2-5 )
which is zero for  = . For pipelines with   0.5 (i.e. “short” pipelines) and for the
moving end regions of those with  < 0.5 (i.e. “long” pipelines), the integral gives:
Fx 2
  x  C ( 2-6 )
2 EA
where the  sign depends on the direction of the movement (and hence the direction
towards which the soil resistance F will act), and the integration constant provides
continuity with the regions where  = 0. For “long” pipelines,  = 0 over all the central
region where  = . For “short” pipelines, compatibility requires that  = 0 at the apex
of the axial force diagram, as the change in direction of the soil resistance (and thus in the
slope of the axial force diagram) implies change in direction of movement.
Figure 2.2 presents the axial displacement diagrams corresponding to the axial
compression diagrams in Figure 2.1. The parabolic form (with direction reversal) of
Equation ( 2-6 ) – due to the constant soil resistance F – can be observed where the
displacements are non-zero. Another interesting observation is the difference between the
curves for 3 and 4: while both are non-zero over the entire length except at x/L = 0.5,
11
the curve for 3 also presents d/dx = 0 at this point (which is not the case for 4). Figure
2.3 presents the diagram of  and  (which are proportional to the diagrams in Figure
2.1 by the factor 1/EA). The shaded areas provide a graphic illustration of the integration
of Equation ( 2-5 ), with the axial displacements corresponding to each of the two shaded
areas indicated in Figure 2.2.
>0
axial displacement
 (1)
0
 (2)
 (4)
 (3)
<0 >0
<0
0 0.25
0.25 L 0.5
0.5 L 0.75
0.75L 1L
x
Figure 2.2 – Diagrams of axial displacements along a pipeline on horizontal seabed for four different
temperature increments.
12
>0
axial strain & thermal expansion 4
3
 (3) =  (4)
2
 (2)
1
 (1)
0
0 0.25L
0.25 0.5L
0.5 0.75 L
0.75 1L
x
Figure 2.3 – Diagrams of axial strain and thermal expansion along a pipeline on horizontal seabed for
four temperature increments.
When the temperature is reduced back to zero, the pipeline will tend to contract
back. In reversing its axial displacements, it will mobilise reversed soil resistance, which
will induce axial tension. “Long” pipelines will still observe a central region with no net
expansion – and hence no axial displacement. As the thermal expansion is reduced back
to zero, this region becomes unloaded, as observed in Figure 2.4. “Short” pipelines will
be fully in tension, induced by the soil resistance to pipeline contraction. However, if 
is just above 0.5, the change in temperature is not enough to mobilise the reversal in axial
displacement over the entire pipeline. A central region will thus not move during
temperature reduction, maintaining its position at the end of the heat-up process. For
  1.0 the reversal in movement is complete and the tension force profile presents a
single peak in the middle.
13
0
axial force
 < 0.5
0.5 <  < 1.0
 > 1.0
 > 0  < 0
<0
0 0.25L
0.25 0.5L
0.5 0.75 L
0.75 1L
x
Figure 2.4 – Diagrams of effective axial force along a pipeline on horizontal seabed when temperature is
reduced back to the initial value.
The parameter  can thus indicate not only if the change in temperature is enough
to mobilise the soil resistance (and pipeline slip) over the entire pipeline length during
loading, but also if soil resistance reversal can be fully mobilised during unloading. This
will be of relevance in assessing susceptibility to walking as discussed in Section 2.4.
2.3. Effect of Global Buckling and Other Discontinuities
The traditional way to control the levels of compression in piping is to use

expansion loops, where deliberate changes in direction cause the pipe to bend. As the
flexural stiffness of a slender structure is much lower than its axial stiffness, flexural
strains can perform the same internal work under much lower loads – although with larger
displacements.
Conventional expansion loops are not compatible with the established methods
for installing subsea pipelines, as all involve laying the pipe from a vessel under
significant top tension. Post installing these loops is uneconomical. The alternative
frequently used in early days, of restraining the pipe by burying it, can become equally
uneconomic for long and remote pipelines.
14
A clever solution, which has been successfully used for thirty years now [19], is
to permit pipelines to buckle in a controlled manner. The bends formed as the pipelines
buckle laterally will act in the same way as the conventional expansion loops, providing
low stiffness regions to dissipate energy under lower axial loads, with larger
displacements.
It is worth highlighting that these buckles are “global buckles” of the pipeline as
a slender structure, not a “local buckle” of the plate forming the pipe (which in most cases
would lead to pipeline failure). Furthermore, unlike a column – for which buckling under
sustained load will cause the structure to collapse – the lateral displacements and bending
will be governed by the thermal expansion of the pipe – with the axial load reducing as
the pipeline is allowed to expand. If carefully engineered, buckles can absorb the thermal
expansion under safe levels of stresses and strains.
Pipeline global buckling (also commonly referred to as lateral buckling) is not the
focus of the present work. As such, the intention was to keep the reference to it to a
minimum. However, the two aspects (walking and buckling) interact closely. The buckles
change the longitudinal distribution of axial force and axial displacements. The gradual
changes in axial displacements (over many cycles), on the other hand, change the amount
of expansion to be absorbed by each buckle, which is directly linked to the pipeline
integrity at the buckle.
With regards to axial force and axial displacements along the pipeline length,
buckles introduce “discontinuities”†. Pipelines will expand into buckles – as they expand
to their ends – mobilising the soil resistance. Figure 2.5 illustrates the longitudinal profile
of axial force of a pipeline with two buckles, both in compression under elevated
temperature; and in tension when the temperature is reduced back. Buckles in general
provide low axial resistance to pipeline expansion, as well as to contraction, so that the
axial force diagram in Figure 2.5 is essentially equivalent to a concatenation of that of
three “short” pipelines.
†
The term “discontinuities” was used as the global distributions of axial force and axial displacements give
the impression that the sections of a pipeline either side of a buckle are axially disconnected. This
impression is due to the loss of axial stiffness at the buckle location. While the global distributions look
like the pipelines were discontinuous, of course they should remain continuous (and a closer look at their
force and displacements distributions around the buckle location would confirm it).
15
>0
axial force <0 >0 <0 >0 <0 >0
expansion
buckle buckle F
contraction
 > 0  < 0  > 0  < 0  > 0  < 0

<0
0 0.25L
0.25 0.5L
0.5 0.75L
0.75 1L
x
Figure 2.5 – Diagram of effective axial force along a pipeline with two global buckles when the
temperature is elevated (expansion); and then brought back down (contraction).
The interaction between the “short” pipelines between the buckles and the buckles
themselves is quite complex. However, for qualitatively assessing pipeline walking, this
complex behaviour can be disregarded. Independent assessment of each section between
buckles will give a very good indication of the tendency of a pipeline to walk, providing
the designer with a good understanding of the pipeline global behaviour before assessing
the interaction in details using a complex FE model. The parameter , for example, shall
thus be calculated using the length between “discontinuities” rather than the total pipeline
length.
16
2.4. Pipeline Walking
Pipeline walking is an axial ratcheting process induced by asymmetries in the

expansion and contraction development†. Key sources of asymmetry – usually called
walking driving mechanisms – are: [20][21]
 Temperature transients during heat-up [9].
 Seabed slope [12].
 Sustained tension at one end (e.g. from a riser) [11].
 Uneven weight distribution of multiphase fluid during shutdown [13].
The present review is limited to the first two mechanisms. The mechanics of the
third one are essentially the same as that of the second mechanism. The last mechanism
involves change in soil resistance along the pipeline length during contraction, and is thus
not compatible with the present simplification of constant soil resistance. Its derivation
in [13] assumes immediate fluid phase separation, and that the soil resistance (to axial
pipe movement) is directly proportional to the pipeline unit weight, whereas the process
is probably much more complex (and for that reason neglected herein).
The two driving mechanisms are discussed independently (i.e. temperature

transients with no seabed slope; then slope with no temperature transient) for the sake of
simplicity, but of course they interact in real life. Interestingly, in many cases they induce
walking tendency in opposite directions, and it is not uncommon to see designs (in
particular those covering a wide range of soil resistances) recommending walking
mitigation in both directions.
If a “short” pipeline is gradually heated-up, whilst always presenting uniform

temperature along its length, its axial force diagram will develop with intermediate steps
similar to the four cases of Figure 2.1. It will initially resemble that of a “long” pipeline
– with a central stationary region – up until the temperature change reaches that giving
†
Note that other mechanisms for accumulated axial displacements have been proposed earlier: LOEKEN
[20] suggested that the increasing end expansion in a pipeline could be due to creep of the anti-corrosion
coating between the pipe wall and the concrete weight coating, with relative slipping between these layers;
KONUK [21] described a ratcheting process at the end of “long” pipelines that would cause ratcheting,
however, being self-contained (i.e. ceasing after a small number of cycles) and usually causing negligible
changes to the initial expansion.
17
Ξ = 0.5. At this moment, only one point in the pipeline (halfway along its length) is
stationary, and will remain stationary as the temperature increases further (while the rest
of the pipe continues to expand).
A similar process is observed during contraction – except that the condition of

only one point being stationary is reached for Ξ = 1.0. Under the current simplified
conditions, the location of this single stationary point is always the same: at x = L/2. As
this point is always stationary, it is considered a “virtual anchor”. In this process, the
pipeline will not accumulate axial displacements over cycles of expansion and
contraction.
The thermal transients observed during pipeline heat-up, however, are often much
more complex than the assumed uniform temperature profile, as indicated by Figure 2.6.
When hot product starts to flow through a pipeline, it will lose heat through the pipe wall
to the ambient. The observed temperature profile shows a temperature decay along the
pipeline length. Although this is usually a complex curve (close to a logarithmic or
hyperbolic distribution), for the purposes of this simplified discussion it will be assumed
that this decay can be approximated by a linear temperature variation along x. As the hot
product continues to flows, the temperature “front” is pushed further, with the
temperature profiles showing longer length of hot pipeline. Within the present
simplification, the linear temperature profile moves along the pipeline (whilst
maintaining its constant gradient), as illustrated by Figure 2.7. At some stage, a steady
state profile will be reached (herein simplified by a constant maximum temperature 
along the pipeline length).
18
Figure 2.6 – Temperature transient obtained from a dynamic flow simulation benchmarked against
recorded process temperatures (reproduced from [9], courtesy ASME).
heat-up steps

temperature
0
0 0.25L
0.25 0.5L
0.5 0.75L
0.75 1L
x
Figure 2.7 – Simplified temperature transient during pipeline heat-up.
19
As a “short” pipeline is heated up by the transient represented by Figure 2.7, it
will gradually expand. In the beginning (for an intermediate temperature profile a) only
the upstream end (left-hand side of Figure 2.8) will move, mobilising the soil resistance
and generating compressive axial forces. At the other side, a region of pipeline presents
no expansion and no displacements, thus not mobilising any soil resistance. In between,
the compression built up as the upstream end expanded needs to be balanced. To mobilise
the reversed soil resistance, a length of pipe needs to slip towards +x. Within the
simplifications herein, the length of pipe moving towards +x needs to be equal to the
length of pipe moving towards –x, so that equilibrium is achieved in the transition, under
compressive axial force equal to this given length times F.
>0
compressive axial force
P()
P(c)
P(b)
F P(a)
0
0 0.25
0.25L 0.5
0.5L 0.75
0.75L 1L
x
Figure 2.8 – Axial force diagrams during heat-up with temperature transient.
At some point, half of the pipeline will be mobilised, the first quarter moving
towards –x being balanced by the second quarter moving towards +x., with a maximum
compressive force of FL/4 at x = L/4. This corresponds to the axial force diagram for a
temperature profile b in Figure 2.8. The downstream half of the pipeline will still be
stationary. The point at x = L/2 which was stationary up until this instant, will then be
pushed towards +x as more pipe is mobilised (e.g. at the intermediate step with a
temperature profile c in Figure 2.8). This point will continue to be pushed until the whole
pipeline length is mobilised, when the compressive force at this point will reach FL/2.
20
After that (as discussed before for uniform heat-up) a further increase in temperature will
cause further pipeline expansion, whilst with no change in axial force (and no further
displacements at the point x = L/2).
When the pipeline is shutdown (and the flow stops) the pipeline (and internal
fluid) will cool-down as it loses heat to the ambient. The steep temperature gradients are
not observed in this process, and the changes in axial displacement and temperature occur
as previously described for the uniform temperature case. The point at x = L/2 will hence
not move – thus maintaining the axial displacement accrued during heat-up (from the
moment it was first mobilised up to the moment the whole pipeline length was mobilised).
As the pipeline heats up again, a very similar process will occur – with the only
difference being the tension profile observed after contraction, as opposed to the unloaded
initial condition – and the point at x = L/2 will accumulate some more displacement
towards +x. This point was the focus of the description as it remains stationary over part
of the expansion and all the contraction processes. The accumulated displacement at this
point, however, is seen to translate in a global pipeline shift towards the downstream end,
which accumulates every cycle.
It is not possible to identify the tendency of a pipeline to walk driven by the

temperature transient by looking only at the final profiles of axial force during expansion
and contraction. As observed, it requires understanding how the pipeline heats up, and
how the axial compression (and thus axial displacements) behave before reaching the
final configuration. This is not the case for pipeline walking driven by seabed slope, as
follows, in which the load asymmetry can be perceived at the final condition (and for
which the intermediate conditions are not relevant).
The axial force diagram for a pipeline on a sloping seabed is not symmetrical, as
observed in Figure 2.9. A (usually small) part of the constant soil resistance F is required
to resist the axial component of the pipeline self-weight. If the seabed is sloping up along
the direction +x by an angle  (i.e. the end at x = 0 is at a lower elevation than the end at
x = L), the axial component of the self-weight is –W sin  (being negative as it points
towards –x), where W is the pipeline submerged unit weight.
21
>0
compressiveaxial force <0 >0
Expansion
+F –F
–W sin  –W sin 
F – W sin  F + W sin 
F + W sin  F – W sin 
 > 0  < 0
–F +F
–W sin  Contraction –W sin 

<0
0 0.25
0.25L 0.5
0.5L 0.75
0.75L 1L
x
Figure 2.9 – Axial force diagrams after expansion and contraction for a sloping seabed.
As the pipeline expands, the region close to x = 0 will move towards –x, attracting
a net soil resistance of F – W sin . At the other end, the movement towards +x will
mobilise a soil resistance of –F – W sin , thus larger in absolute value. If the pipeline is
“short”, the two linear segments of the compressive axial force diagram (building up from
the ends) will intercept, but not at x = L/2. The interception will occur a bit further, at
x = L (F + W sin ) / (2F) instead.
During contraction, the soil resistance reverses, but not the weight component. As
a result, the asymmetry in the gradients of the axial force diagram reverses, with peak
now at x = L (F – W sin ) / (2F). The regions of pipe outside the central section between
the two peak force locations – at x = L (F  W sin ) / (2F) – will alternate positive and
negative increments of displacements as the pipeline expands and contracts. The pipe
within this interval, however, will see negative displacements during expansion and
negative displacements again during contraction.
22
The stationary point will alternate between these two locations (i.e. the pipe will
be stationary at one point during expansion, and at another point during contraction). The
net shift per cycle can thus be calculated as either the increment in displacement at
x = L (F + W sin ) / (2F) at the end of the expansion (as this point will remain stationary
during contraction) or vice versa, the increment in displacement at
x = L (F – W sin ) / (2F) at the end of the contraction (as this point would have remained
stationary during expansion).
It is worth mentioning a couple of differences between the two driving

mechanisms. As discussed, the walking due to temperature happens during heat-up, in
particular during the mobilisation of the second half of the “short” pipeline. How much
the temperature is further increased beyond that point is not relevant, as it will not cause
any more movement at the stationary central point. For walking due to seabed slope, the
total change in temperature is the governing load, as the stationary point in contraction
will continue moving until the expansion is completed (and vice versa).
TØRNES et al. [9] claimed that “long” pipelines (i.e. with Ξ < 0.5) would not
walk, as they would be “anchored” by the central stationary region. CARR et al. [12] later
showed that they can be driven to walk by temperature transients – which is consistent
with the fact that this mode of walking is not governed by the final form of the axial force
distribution, but instead by its transition as the pipe heats up. Walking due to seabed slope,
on the other hand, requires full mobilisation in contraction (i.e. with Ξ > 1.0), or to be
more precise, that at least part of the central section between the two peak force locations
at x = L (F  W sin ) / (2F) is mobilised. This is observed when Ξ > (F – W sin )/F
(which is usually just below the unity for typically small slopes).
23
3. A Brief Review of Couple Consolidation FE Analysis
3.1. Background
For a civil engineer, soil is a mix of granular material, in which the grains are
small enough that the inter-granular mechanics are not relevant for the structures
interacting with it. As such, it can be treated as a continuous, porous medium. Seabed soil
is saturated (except in very unusual conditions), which means that the voids between
grains are totally filled with water (called pore water). For most engineering purposes,
both the soil grains and the pore water can be considered incompressible, so that change
in volume can only occur if water is expelled from or absorbed into the soil voids.
In geotechnical engineering, it is common practice to address the effect of any

loading on the soil as one of two extreme cases: “drained” or “undrained”. In “drained”
conditions, the water imposes no restriction to change in soil volume. If the load induces
a given change in volume, this will only occur after a certain quantity of water flows
within the soil and in or out the soil voids. In “drained” conditions, the time required for
this water flow to occur is negligible, hence the drainage (and the corresponding change
in volume) occurs at the same time as the load develops. On the other hand, in “undrained”
conditions, the time required for the load to develop is much smaller than the time it
would take for the water to flow. A negligible flow of water occurs while the load is
applied, and as such, only a negligible change in volume occurs. In reality, most of the
times (if not always), loading will be “partially-drained” (i.e. somewhere between the two
extremes), but the idealisation of the two extremes is exceptionally useful in engineering
practice.
If a loading is not “drained” (i.e. it is “undrained” or “partially-drained”), excess

pore pressure – which is the pore water gauge pressure at a given point subtracted by the
water head at this point – will develop. Gradients of excess pore pressure will induce
seepage, and the soil will gradually change in volume over time as the excess pore
pressure dissipates. This process is called consolidation. Consolidation will change the
effective stresses at the pipe-soil interface and throughout the soil mass over time. It will
also change the soil state, usually making it harder with cycles of excess pore pressure
24
generation intervened with consolidation intervals. These aspects are fundamental for the
pipe-soil interaction processes discussed hereafter.
This section briefly reviews the classic formulation of the consolidation process,
followed by its formulation in coupled FE. A brief description of the “critical state”
framework and the “modified Cam clay” constitutive model is then presented, followed
by its implementation within the FE stiffness matrix. The introductory formulations are
intended to provide a basis for the readers which are not familiar with geotechnical
engineering. Although it might read too basic for geotechnical engineers, this section also
introduces the FE formulation, which is fundamental reading for geotechnical engineers
not familiar with FE application.
3.2. General Formulation
The equation governing the soil consolidation is the continuity equation, ensuring
local mass balance. As previously mentioned, the soil grains and the pore water are
considered incompressible such that any change in soil volume is due to water being
expelled or absorbed. The volume of water expelled from an infinitesimal rectangular
hexahedron of soil, of sides dx, dy and dz, over a small time increment t is given by:
 vx   v 
  vx  dx  vx  dy dz   v y  y dy  v y  dx dz 
 x   y 
( 3-1 )
 v  
  vz  z dz  vz  dx dy   t  vw  t dx dy dz
 z  
where vw = [vx vy vz]T is the pore water apparent velocity (Darcy velocity). This has to
be equal to the change in volume due to the soil deformation, which (neglecting the higher
order terms) is given by:
  x    y 
  x  dx   x  dy dz    y  dy   y  dx dz 
 x   y 
( 3-2 )
  z  
   z  dz   z  dx dy    v dx dy dz
 z  
25
with x, y and z being the components of displacement increment and v the
increment in volumetric strain observed over that time increment. Combining equations
( 3-1 ) and ( 3-2 ) yields:
 v
 vw  0 ( 3-3 )
t
The pore water flow velocity is assumed governed by Darcy’s law, such that:
k
vw   u ( 3-4 )
w
where w is the unit weight of the pore water and u the excess pore pressure. For the sake
of simplicity, the soil permeability k is assumed to be constant and isotropic (scalar).
Equation ( 3-3 ) can thus be rewritten as:
 v k 2
  u0 ( 3-5 )
t  w
which is the consolidation equation.
The classical solution proposed by TERZAGHI [22] considers one-dimensional

condition and constant total vertical stress v, for which v = mv'v = –mvu, where mv
is the coefficient of volume change and 'v the effective vertical stress. For these
particular conditions, Equation ( 3-5 ) can be rewritten as:
u  2u
 cv 2  0 ( 3-6 )
t z
with the coefficient of consolidation given by:
k
cv  ( 3-7 )
 w mv
If the soil is linear elastic, for which mv (and thus cv) is constant, Equation ( 3-6 )
has the same form as the classical thermodynamics equation for one-dimensional heat
dissipation, so TERZAGHI [22] has proposed adopting the same solution. BIOT [23] has
later shown that when the particular conditions of one-dimension and constant total stress
26
are not applicable, the continuity equation (Equation ( 3-5 )) has to be solved together
with the equilibrium equations, which in Cartesian coordinates (with z being the vertical
pointing down) are:
 x  xy  xz  x  xy  xz u
      0
x y z x y z x
 xy  y  yz  xy  y  yz u
      0 ( 3-8 )
x y z x y z y
 xz  yz  z  xz  yz  z u
       w
x y z x y z z
3.3. Formulation in FE
For the majority of real conditions, there is no analytical solution of the system of
equations encompassing equations ( 3-5 ) and ( 3-8 ), which thus require numerical
methods for obtaining approximate solutions. One of such methods is the finite element
(FE) method, in which the domain is discretised in small elements connected by nodes.
The FE method uses interpolation functions to interpolate calculated values at the

nodes to any point within each element. For example, let  be any given quantity (e.g. a
displacement component, the excess pore pressure, a stress or strain component, etc.) and
e the vector containing the values of  calculated at every node of a given element (nodal
values). The value of  at any point (x, y, z) within the element is interpolated as:
  x, y , z   H  x, y , z  χ e ( 3-9 )
where H is the interpolation function, a 1nne matrix, where nne is the number of nodes in
the element.
The partial derivatives of  at (x, y, z) are given by the functions Hx, Hy and
Hz, also 1nne matrices, such that:
27

 H x  x, y , z  χ e
x  x, y, z 

 H y  x , y , z  χ e ( 3-10 )
y  x, y, z 

 H z  x , y , z  χ e
z  x, y, z 
The relationships between increments of strain  and displacement  in the soil

skeleton can be written, in Cartesian coordinates, as:
 x  y  z

 x   yz  
x z y
 y  x  z
 y   xz   ( 3-11 )
y z x
 z  x  y
 z   xy  
z y x
or in matrix notation:
 ε  B ξ e ( 3-12 )
T
where  ε   x  y  z  yz  xz  xy  is the Cauchy engineering† strain
increment tensor (a second order tensor, thus a 61 vector in Voigt notation),  ξ e is the
3nne1 vector encompassing the nodal values of the three components of displacements,
given by:
T
 ξ e  1x  2 x ...  n ne x
1 y ...  n ne y
1z ...  n z 
ne
( 3-13 )
and the 63nne matrix B is given by:
†
Engineering notation is used, with engineering shear strain components . Rigorous tensor analysis would
require using shear strain components /2, but these are not employed in the present work.
28
 H x 0 0 
 0 H y 0 

 0 0 H z 
B    ( 3-14 )
 0 H z H y 
 H z 0 H x 
 
H y H x 0 
The increment of volumetric strain may similarly be written as:
 v  B v ξ e ( 3-15 )
where the 13nne matrix Bv is given by:
B v   H x H y H z  ( 3-16 )
The relationship between increments of effective stress and strain is governed by
 σ  D ε ( 3-17 )
T
where  σ   x  y  z  yz  xz  xy  is the effective stress increment
tensor and D is the 66 tangent effective stress-strain stiffness matrix (in fact a fourth-
order tensor) of the soil skeleton, presented in Section 3.4.
Employing the principle of virtual work, the FE method permits calculating the
increments in nodal displacements from the equilibrium equations (for a detailed and
rigorous derivation of the FE formulation for coupled consolidation, refer to the Appendix
A of BENTLER [24]) as:
K  ξ  K Tv  u   f ( 3-18 )
where  and f are the 3nn1 vectors containing, respectively, the increments of
displacements and forces for the three degrees of freedom, for all nodes in the model –
with nn being the total number of nodes in the model – and u the nn1 vector of
increments of excess pore pressure in all model nodes. The 3nn3nn matrix K is the model
global tangent stiffness matrix; and the nn3nn matrix Kv the model global coupling
matrix.
29
The global matrices are obtained by superposing the contribution of each model
element. The 3nne3nne stiffness matrix of each element is given by:
Ke   B D B d
T
e ( 3-19 )
e
where e is the element volume. Likewise, the nne3nne element coupling matrix is:
K ve   H B d
T
v e ( 3-20 )
e
Note that the system of equations in Equation ( 3-18 ) is undetermined, as it has

4nn unknowns (3nn increments of displacements and nn increments of excess pore
pressure) but only 3nn equations. A determined system is obtained if the continuity
equation – the consolidation equation, Equation ( 3-5 ) – is also considered. In FE (again
reference is made to [24] for a thorough derivation), Equation ( 3-5 ) is solved as:
ξ
Kv  Ku u    u  g    g ( 3-21 )
t
where g is the vector of nodal external flow of pore water (a model boundary condition)
and the nnnn matrix Ku is the model global transmissivity matrix. The time integration
parameter  determines the type of approximation for u and g during the time integration
step t, for example:  = 0 correspond to forward interpolation;  = 1 to backward
interpolation; and  = 0.5 to linear interpolation. BOOKER & SMALL [25] have shown
that, to ensure numerical stability, it is necessary that   0.5.
The global transmissivity matrix is also obtained by adding the contribution of

each element, with the element transmissivity matrix being given by:
k
K ue    HT H d e ( 3-22 )
e
w
with H = Hx+Hy+Hz in Cartesian coordinates.
Rearranging Equation ( 3-21 ) and combining it with Equation ( 3-18 ), a single

equation – in matrix notation – can be written:
30
K K Tv   ξ   f 
      t g    g  K u  ( 3-23 )
K v   t K u   u    u 
3.4. The MCC Constitutive Model
The soil constitutive model adopted in this work was “modified Cam clay” (MCC)
[26, 27], which builds upon the “critical state soil mechanics” concept, first proposed by
ROSCOE, SCHOFIELD & WROTH [28]. This concept has provided probably the first
framework for consistently considering soil stability and deformation together.
The adoption of critical state soil mechanics is key for a number of aspects
considered in the present work. Constitutive models in which strength and stability are
not directly linked to soil deformation would not permit, for example, accounting for the
consolidation hardening process later described in Section 4.3.3. From the available
constitutive models based on critical state soil mechanics, MCC was selected for use in
the present work for ease of comparison with published FE analysis results, as well as FE
models developed during the present research. The main reason for that is that the MCC
model is implemented and available as an in-built constitutive model in the commercially
available FE package ABAQUS [29].
The soil state is described by three key parameters: the void ratio e (the ratio
between the volume of pore water and the volume of the solid soil grains, following the
assumption of full saturation) and the mean and deviator invariants of the effective stress
tensor, denoted respectively p' and q. For a general effective stress tensor
T
σ   x  y  z  yz  xz  xy  , the invariants are given by:
 x   y   z
p  ( 3-24 )
3
and:
                     6  yz2   xz2   xy2 

2 2 2
q
y z x z x y ( 3-25 )
2
31
The soil behaviour is elastoplastic, limited by a curved yield surface defined by
(e, p', q) = 0, illustrated in Figure 3.1. The intersection between the yield surface and
the e–p' plane is called “normal consolidation line” (NCL) – for being the locus of the
normal consolidated soil states – and is given by:
e     ln p ( 3-26 )
which is a straight line in the e–ln p' space. The slope  and intercept  of this line are
soil properties† assumed constant (although the value of  depends on the selected unity
for p'). As a normally consolidated soil sample consolidates under hydrostatic conditions
(i.e. with q = 0), it sees increase in p' and decrease in e along the NCL (e.g. from point C
to point A in Figure 3.1). [30]
†
For guidance on how to obtain each of the MCC soil parameters from soil tests, refer for example to
ATKINSON & BRANSBY [30].
32
Figure 3.1 – Critical state framework: e–p'–q space and the yield surface (reproduced† from [27], courtesy
Cambridge University Press).
Swelling and reconsolidation under hydrostatic conditions, from point C, will take
place along the curve CBR in Figure 3.1, a “swelling and reconsolidation line” (SRL),
given by:†
e  e   ln p ( 3-27 )
which is also a straight line in the e–ln p' space. Its slope  is also a (constant) soil
property, but not the intercept e, which depends on the current pre-consolidation stress
p'c (i.e. the location of point C along the NCL). The SRL is thus not a unique line, but
parallel lines in the e–ln p' space, each one made unique by its intercept:
e         ln pc ( 3-28 )
†
Minor changes in notation were made in figure to suit that adopted in the present work.
33
The soil response on the SRL is non-linear elastic, with a stress-dependent bulk
modulus K:
K
1  e  p ( 3-29 )

Within the elastic range, p'c will remain constant. If the yield surface (the NCL if q = 0)
is reached, further change in e and p' will cause p'c to shift, and a new (elastic) SRL to be
applicable. This is similar to the plasticity in metals, in which accumulated plastic strains
will shift the uniaxial linear relationship x = Ex across the x axis (without changing the
slope E).
For q  0, soil will still present this non-linear elastic response – with the
relationship between e and p' still following equations ( 3-27 ) and ( 3-28 ) – while beneath
the yield surface (i.e. for  < 0), which corresponds to the “elastic wall” illustrated in
Figure 3.1. As the SRL, the elastic wall is not unique, but depends on the current p'c. The
curve CXF in Figure 3.1 is called the “elastic limit”, the intercept between the elastic wall
and the yield surface. Its projection on the p'–q plane, the curve C'X'F', is the projected
yield function p(p', q) = 0. While each elastic wall will have its corresponding projected
yield function, they are proportional to p'c, such that p is a single curve in the normalised
(p'/p'c)–(q/p'c) space.
Plasticity in the critical state framework is assumed to follow the associated flow
rule. As such, the plastic potential function in the vp–sp space (where the former is the
plastic volumetric strain, and the last the plastic shear strain) is proportional to the
projected yield function. As plastic strain increments are perpendicular to the plastic
potential function, the ratio between their volumetric and shear components can be
inferred from the projected yield function, as illustrated in Figure 3.1. As the soil state
reaches the point X in the e–p'–q space, the plastic strain increment will be perpendicular
to the projected yield function at X'.
Plastic strains as illustrated will cause hardening, corresponding to the yield

function projected on the p'–q plane increasing. Note that this hardening does not change
the yield surface in the e–p'–q space, only the projection of the elastic limit is enlarged as
p'c increases.
34
The soil state may eventually reach the line EF in Figure 3.1, at which the
projected yield function is at its maximum (e.g. at F'), and p/p' = 0. At this point, the
plastic strain vector will be vertical, with increments of plastic shear strain occurring with
no change in plastic volumetric strain, and hence no tendency in changing soil state. This
state is called “critical state”, and the locus of the critical state for different values of p'c
is the “critical state line” (CSL). When an element of soil undergoing uniform shear
distortion eventually reaches the CSL, it will continue distorting with no change in void
ratio e or effective stresses p' and q.
The projection of the CSL on the p'–q plane is a straight line crossing the origin.
Its slope  is a soil property also assumed constant. Projected on the e–ln p' plane, the
CSL is a straight line parallel to the NCL (i.e. with slope ). Its intercept is also constant
and, for a given yield surface, it can be calculated from  (the intercept of the NCL) – or
vice versa.
The overconsolidation ratio in critical state soil mechanics is defined in terms of

mean effective stress invariant (rather than in terms of effective vertical stress like OCR
in classical soil mechanics), and is given by:
pc
R ( 3-30 )
p
for which R = 1.0 at the NCL and R > 1.0 anywhere else in the e–p'–q space. Note that
while  < 0 corresponds to the elastic range and  = 0 to the yield surface,  > 0 is not
a possible soil state.
Following the introduction of the critical state concept by ROSCOE,

SCHOFIELD & WROTH [28], ROSCOE & SCHOFIELD [31] proposed the following
equation for the projected yield function:
   ln R ( 3-31 )
where the stress ratio  = q/p'. The soil constitutive model following the critical state
framework and using this yield function became known as “Cam clay” model [32].
Another yield function was proposed by BURLAND [26] (and then more thoroughly
presented by ROSCOE & BURLAND [27]):
35
   R 1 ( 3-32 )
The constitutive model using this projected yield function was called by ROSCOE &
BURLAND [27] “modified Cam clay” (MCC) – and after that the model using the yield
function proposed by ROSCOE & SCHOFIELD [31] is often called “original Cam clay”
model, to distinguish it from the MCC.
Within the elastic range, the effective stress-strain relationship within the soil FE
model is given by the elastic matrix:
 1 1 1 
 m  2G  2G 00 0
mv mv
 v

1 1 1 
 m  2G mv mv
 2G 0 0 0 
 v 
Del   1 1 1  ( 3-33 )
  2G  2G 0 0 0
 mv mv mv 
 0 0 0 G 0 0
 
 0 0 0 0 G 0
 0 0 0 0 0 G 
where G is the soil shear modulus. Within the present work, it was chosen to consider the
Poisson’s ratio  as a constant†, and to calculate mv and G from the constant  and the
stress-dependent K as per Equation ( 3-29 ), such that:
1 
mv  ( 3-34 )
3K 1  
and
3K 1  2 
G ( 3-35 )
2 1   
On the yield surface, the associated plasticity rule implies that:
†
Other approaches have been used elsewhere, such as considering G as a constant or as proportional to K.
While small differences in the numerical results are expected, changing the approach is not expected to
have any qualitative impact on the conclusions of the present work.
36
 p
 vp   ( 3-36 )
p
and:
 p
 sp   ( 3-37 )
q
where  is a proportionality factor. This should be valid for all components of the strain
tensor, such that:
εp  a ( 3-38 )
with
T
  p  p  p  p  p  p 
a  ( 3-39 )
  x  y  z  yz  xz  xy 
As presented by ZIENKIEWICZ & NAYLOR [33], the elastoplastic relationship

between increments of effective stress and increments of strain can be written as:
 ε   Del a   σ
1
  T   ( 3-40 )
 0   a A    
where, for the MCC projected yield function, the parameter A is given by:
       2  2 
A  1  e    2 2  ( 3-41 )
 p     
and the partial derivatives of p are:
 p      9  p  9 i
2 2
 ( 3-42 )
 i 3 p   2   2 
for i = x, y and z; and:
37
 p 6 i
 ( 3-43 )
 i p   2   2 
for i = yz, xz and xy.
Finally, an explicit expression for D – which can be directly employed in the

element stiffness matrix of Equation ( 3-19 ) during yielding (i.e. when  = 0) – is derived
by ZIENKIEWICZ & NAYLOR [33]:
1
D  Del  d dT ( 3-44 )
b
where d = Del a and b = A + aTd.
38
4. Review of the Existing Literature on Axial PSI
4.1. Background
Soil resistance to axial pipeline movement has for long been observed to be time-
dependent. Still, only within the last five years have quantitative means for assessing this
dependency been proposed. Up until the last decade, the soil resistance to axial pipeline
movement has only been assessed as either a drained or undrained response – which (as
previously mentioned) is in line with the traditional way of dealing with soil mechanics
– or as a sum of “frictional” and “cohesive” components.
The internationally relevant subsea pipeline design standards do not provide much
guidance. ISO 13623 [34], EN 14161 [35] and DNV-OS-F101 [36] provide no specific
guidance†. API RP 1111 [37] makes reference to confidential joint industry project (JIP)
reports‡. The old BS 8010-3 [38] (now superseded by EN 14161 [35]) recommended
particular ranges of “friction factors” to be used in the North Sea, for “non-cohesive” and
“cohesive” soils. ASME B31.4 [39] is the only to make reference to the publically
available ALA design guideline [40], which provides equations for calculating soil axial
resistance to movement of buried pipelines. [41] [42] [43] [44] [45] [46]
4.2. Time-Independent Axial PSI
4.2.1. Drained and Undrained Resistance

The American Lifelines Alliance’s (ALA) “Guidelines for the Design of Buried
Steel Pipe” [40] recommends calculating the maximum soil axial resistance Fmax as:
†
The recommended practice DNV-RP-F105 [41] (associated with [36]) mentions that “For sands with low
content of fines, the frictional component of the axial and lateral resistance is proportional to the vertical
force at any time. For clays, the resistance is proportional to the undrained shear strength.”
‡
Reference is made to the design guidelines from HOTPIPE and SAFEBUCK JIPs. The candidate had no
access to the first one, but as it has served as the basis for DNV-RP-F110 [42] (which provides no guidance
on axial pipe soil interaction), not much information is expected. The candidate did have access to the
second JIP guideline while working for one of its sponsors. While this JIP has fostered a significant portion
of the latest developments in axial PSI, the guideline version referenced by API RP 1111 [37], CARR &
BRUTON [43], provide very limited information and a single equation for undrained response – very
similar to that by SCHAMINÉE et al. [44], including the adhesion-sensitivity relationship suggested by
FINCH [45] (mentioned hereafter) and a reduction factor based on the tilt table test results published by
NAJJAR et al. [46].
39
 1  K0 
Fmax   D   u su  H   tan  f     ( 4-1 )
 2 
which corresponds to an average shear stress – around the pipe circumference D –

encompassing two distinct parcels: a “cohesive” and a “frictional”. The first is obtained
by multiplying the “representative cohesion” su by an “adhesion factor” u. The last from
the average normal effective stress – approximated by H'(1+K0)/2 where H is the burial
depth at the pipeline centreline – multiplied by the tangent of an interface friction angle,
which is the soil internal friction angle ' multiplied by a coating dependent factor f.
SCHAMINÉE et al. [44] proposed a different approximation for the average

effective normal stress, taking into account the pipeline submerged unit weight W:
1  D W 
 n    H      H  K 0   ( 4-2 )
2  2 2D 
Note that [44] have not specified which “coefficient of lateral earth pressure” to use. The
present interpretation, above (that it should be the coefficient at rest K0), is different from
that by FINCH et al. [47], who adopted Equation ( 4-2 ) but using the coefficient of active
earth pressure instead.
SCHAMINÉE et al. [44] have not provided any indication of what the interface
friction angle should be. FINCH et al. [47] adopted a factor multiplying the tangent of
the internal friction angle, and with proposed values dependant not only on the coating
roughness, but also on the soil grains size – represented by the median value D50. Other
difference is that FINCH et al. [47] did not suggest adding the two parcels, but instead to
adopt the relevant one, depending on the expected – drained or undrained – response.
In assessing the drained resistance for unburied pipelines, FINCH et al. [47]
suggested replacing the normal force – the average normal effective stress  n multiplied
by the circumference D – by the pipeline submerged unit weight W. They highlighted,

however, that while this approach provided good correlations with laboratory tests, larger
values of “apparent frictional resistance” are observed in real pipelines. They attributed
this to the fact that real pipelines are not perfectly straight, so that transversal components
of soil resistance would contribute to the overall axial resistance.
40
WHITE & RANDOLPH [48] highlighted that, for partially buried pipes, the
integral of the normal effective stress modulus along the contact perimeter is more than
W due to wedging effects, and that this total contact force is what should be multiplied by
the interface friction. They proposed an expression for a “wedging factor”  based on the
analytical solution for a line load on an elastic semi-space:
2sin  
  ( 4-3 )
   sin   cos  
where ' = arccos(1–2w/D) is the angle from the pipe bottom invert to the seabed surface
(along the pipe-soil interface, for the embedment w), such that 'D is the pipe-soil contact
perimeter. Another expression, curve fitted from the results of elastic coupled
consolidation FE analyses, was later proposed by GOURVENEC & WHITE [49].
For determining the undrained resistance for unburied pipes, BRENNODDEN &
STOKKELAND [50] suggested multiplying  u su by the contact perimeter 'D – which
is analogous to the  D u su resistance for buried pipelines in ALA [40] and
SCHAMINÉE et al. [44]. They also suggested that u – called “activation factor” by
SCHAMINÉE et al. [44], then “adhesion factor” by most of the references that followed
– should have a “maximum” and a “residual” value. FINCH [45] suggested that u could
be taken as the unity if the “remoulded undrained shear strength” was used for the
“residual” condition.
BRUTON et al. [51] recommended using the effective stress approach – of

multiplying the total (effective) contact force (including wedging effects) by the interface
friction – for both drained and undrained (or partially drained) conditions. In the absence
of means for calculating the excess pore pressure generated during undrained pipe
movements, they have used correlations with excess pore pressure measured in model
tests. An additional aspect considered by BRUTON et al. [51] was the increase in
interface friction observed for very low levels of interface normal effective stress [46, 48,
52].
4.2.2. Mobilisation and Response Curve

For use in FE analyses, common practice is to adopt a constant maximum
resistance preceded by a linear elastic pre-slip response characterized by a “mobilization
41
distance” mob. ALA [40] specifies a “displacement at Fmax”, F, ranging between 3 mm
and 10 mm (for soil categories ranging from “dense sand” to “soft clay”, respectively),
but then presents an illustration of the idealized curve (reproduced in Figure 4.1) which
suggests that mob should be much less than F (apparently about half of it, although no
guidance is provided).
Figure 4.1 – Axial resistance-relative displacement curve and idealised linear elastic mobilisation
(reproduced† from [40]).
CATHIE et al. [53] reiterated the use of a “linear elastic-perfectly plastic”

response curve for “granular soils”, mentioning that tests with sand have indicated
mob = 2 mm, thus consistent with the 0.1 inch mobilisation distance for axial friction on
piles recommended by API RP 2A [54]. †
BRUTON et al. [55] – based on results of tests using soft clay – mentioned that
the displacement observed when the full axial resistance is mobilised is typically over
10 mm, and can exceed 100 mm. They highlighted, however, that only “a small part” of
this displacement “is elastically recovered when the pipe is unloaded”. As such, they
recommended, for FE modelling purposes, mob < 5 mm or 0.01D for initial loading, and
0.0025D for unload-reload conditions.
A much more complex, multi-linear response curve had earlier been proposed for
a specific project, based on model tests using soft clay, by BRENNODDEN &
†
Minor changes in notation were made in figure to suit that adopted in the present work.
42
STOKKELAND [50]. This response, capturing an early peak response followed by a
lower residual resistance, is described by the diagram reproduced in Figure 4.2. †
Figure 4.2 – Project-specific axial PSI response curve proposed by BRENNODDEN &
STOKKELAND [50] (reproduced from [50]†).
4.3. Time-Dependent Axial PSI
4.3.1. Set-up
As previously mentioned, for many years axial pipe-soil interaction is observed to
be time-dependent. PALMER & LING [56], back in 1981, mentioned that: “Experience
in bottom pull installation of (pipelines) indicates that if a pipeline has been in place for
some time, a coefficient (of friction) of between 0.8 and 1 should be used to find the force
needed to move it.” Eleven years later, BRENNODDEN & STOKKELAND [50]
published the results of real scale tests which focused on the effect of consolidation‡ on
the soil resistance to pipeline axial – and lateral – movement. The following key
†
reproduction prohibited without permission. Minor changes in notation were made in figure to suit that
adopted in the present work.
‡
BRENNODDEN & STOKKELAND [50] mentioned that “the term consolidation”, in their paper, “was
chosen to represent all time effects on soil resistance” and that “the mechanisms involved may not solely
be related to dissipation of pore water in clay”. They also indicated their belief that “adhesion bonds on the
contact area may (also) play a role”.
43
conclusions were obtained from their test results, and applicable for both peak and
residual soil resistance:
 Resistance increased with time of consolidation (both due to increase in soil

resistance but also due to increase in embedment, and hence contact area).
 Resistance increased with pipe weight during consolidation – but was not
sensitive to the pipe weight during axial movement.
 Resistance increased with pipe velocity (which suggests viscous effects).
CATHIE et al. [53], in their review of the state of the art in pipeline geotechnics,
in 2005, stated that the axial soil resistance is influenced by the pipeline embedment and
“time dependent factors”, namely:
 For sands: further embedment or densification due to environmental loading; and

build-up of sediment against the pipe.
 For clays: “set-up”, with strength increase arising from thixotropy and
consolidation; and development of adhesion between the pipe and the soil.
Again, an analogy with piles is made: the term “set-up”, commonly used in piling, refers
to the consolidation of the soil surrounding a pile after excess pore pressure is generated
during the pile installation, which translates into a gain in resistance.
Probably the first papers to quantitatively address the time-dependency of axial

pipe-soil interaction, GOURVENEC & WHITE [49] and KROST et al. [57] link the “set-
up” of the pipeline (the consolidation after excess pore pressure is generated during pipe
lay) with the availability of axial pipe-soil resistance. The set-up is analysed using elastic
coupled consolidation FE analysis, and the results compared to excess pore pressure
measurements in an in situ model test. CHATTERJEE et al. [58] later reproduced these
set-up analyses using a more sophisticated FE model, including large deformation and
plasticity (employing the MCC constitutive model).
JEWELL & BALLARD [59], however, questioned the high cv adopted by KROST
et al. [57] to match the in situ tests. They suggested instead that the quick dissipation
observed could have been due to “hydraulic leaks” at the pipe-soil interface – either due
44
to “piping and / or regressive erosion” (due to the “relatively high hydraulic gradient,
short drainage path length and low normal stress”) or due to preferential drainage caused
by the model pipe coating roughness.
4.3.2. Rate Dependency

Another attempt to quantify the time-dependency of the axial pipe-soil interaction
is the correlation between resistance and pipeline velocity (rate of movement),
formalising what had already been suggested by BRUTON et al. [51]. JEWELL &
BALLARD [59] proposed a design chart, reproduced in Figure 4.3, in which the peak and
residual resistances vary with the pipeline velocity v. They suggested a gradual transition
from a (higher) drained resistance to a (lower) undrained resistance – although not
proposing any specific relationship –, with logarithmic gain in resistance for higher
velocities attributed to “rate effects”.
45
Figure 4.3 – Design chart for axial pipe-soil interaction proposed by JEWELL & BALLARD [59]
(reproduced from [59]†).
WHITE & CATHIE [8] suggested using a hyperbolic relationship for the drained-
undrained transition, mentioning a series of other applications in which this backbone-
shaped curve has been used. In all of these applications, the velocity is normalised by the
coefficient of consolidation and some relative drainage (as in Figure 4.3). WHITE &
CATHIE [8] questioned, however, whether the soil bulk stiffness is relevant for the axial
PSI, or whether the normalisation should be by the soil permeability instead (as in the
case of seabed ploughing studied by PALMER [60]). †
HILL et al. [61] and WHITE et al. [62] indicated that, for dilatant soils, the
transition could be from a lower drained resistance to a higher undrained resistance –
which was later demonstrated in the model test results by BOYLAN et al. [63].
†
46
4.3.3. Consolidation Hardening
Set-up will increase soil resistance to axial pipe movement, as the effective
stresses at the pipe-soil interface will gradually rise as the excess pore pressure dissipates.
Further positive excess pore pressure may be generated due to shear each time the pipeline
moves. In this case, resistance will increase not only as a consequence of the set-up after
each of these events, but also because of consolidation hardening.
Within the critical state framework (as reviewed in Section 3.4), consolidation
hardening will occur when the undrained (or partially-drained) loading events cause the
yield envelope to expand (increasing p'c). After consolidation (i.e. dissipation of the
excess pore pressures induced by the loading; set-up), the new soil state will present a
higher overconsolidation ratio (and thus higher resistance) than before the loading-
consolidation cycle.
Figure 4.4 illustrates the consolidation hardening process, with even points
indicating end of loading events and odd points end of consolidation periods. The figure
is reproduced from PALMER [64], who used it to explain the development of plough pan
(a compacted layer in cultivated soil resulting from repeated ploughing) and its analogy
with the effect of ice gouging on the seabed (and how the safe burial depth for pipelines
could be assessed based on geotechnical investigation). The same rationale was also used
by WHITE & HODDER [65] to predict the increase in soil strength after cyclic vertical
movements of penetrometers, spudcan foundations or touchdown point of risers or lateral
movements of pipelines; and more recently by DEEKS et al. [66] for directly on-seabed
sliding foundations.
47
Figure 4.4 – Cycles of loading intervened by consolidation periods illustrating the consolidation
hardening process (reproduced from [64]†).
WHITE & CATHIE [8] applied this concept to the axial pipe-soil interaction,
expanding the single-element behaviour of Figure 4.4 to the interaction between
consecutive layers beneath the pipe, as illustrated by Figure 4.5. The shear stresses
imposed to the soil decay approximately with the inverse of the radial position. Hence, in
a first movement, the soil resistance would be first exhausted immediately at the pipe-soil
interface (where the radius is minimum and thus the shear stress is maximum), as
illustrated by Figure 4.5a. As the limit resistance is reached, distortion in this region will
be much more noticeable than in others, thus characterising a “shear zone”. †
As this region hardens after this first event, a second movement may cause the
shear zone to occur elsewhere, further away from the pipe-soil interface (e.g. at zone B
in Figure 4.5d). Although the shear stresses in zone B are less than those in zone A, the
gain in resistance in zone A due to hardening is, in this example, in larger proportion than
this stress difference.
†
Copyright 2008 Canadian Science Publishing or its licensors. Reproduced with permission.
48
Figure 4.5 – Shear zone migration away from the pipe-soil interface (reproduced from [8]).
YAN et al. [67] presented results of 3D coupled consolidation FE analyses, using

the MCC constitutive model, of cyclic axial PSI. Their results of soil strength after each
cycle of consolidation hardening (undrained pipeline sweep followed by full
consolidation), reproduced in Figure 4.6, clearly show the strength gain next to the pipe,
although indicating a smooth variation rather than the process in steps as idealised by
WHITE & CATHIE [8]. The stress–volume path in the e–ln p' space, obtained in their
FE model is very similar to the idealised one in Figure 4.4.
49
Figure 4.6 – Profiles of shear strength after steps of consolidation hardening obtained by YAN et al. [67]
(reproduced from [67], courtesy ICE Publishing).
4.3.4. Overarching Frameworks

Following the compilation of a detailed axial PSI model test using (reconstituted)
natural seabed soil, along with the results of in situ tests, by WHITE et al. [14], a few
overarching frameworks have been proposed to explain the high variability observed and
provide useful input for pipeline design. Combining some of the already mentioned
aspects, combined with other specifics, these frameworks intended to provide a rational
way whereby the complete set of responses observed could be calculated and considered
in engineering practice.
JEWELL & BALLARD [59] proposed a framework which is in essence the rate-
dependent response summarised by Figure 4.3, combined with an in situ experimental
assessment of the set-up. The notable points in the response graph are obtained with
equations for drained and undrained resistance similar to those discussed in Section 4.2.1,
with the undrained accounting for an OCR based correction on the su. One interesting
point highlighted (whilst not accompanied by further indication on how to address) is the
effect of the soil temperature in the rate-dependent response. It is well known that the
temperature influences the viscosity of fluids, for which differences in temperature (either
between in situ and laboratory conditions or due to heat exchange between the operating
50
pipeline and the soil beneath it) can influence both the pore water drainage velocity and
the viscous response of clays.
A more complete framework, accounting for more mechanisms, was proposed by

HILL et al. [61] and WHITE et al. [62]. These include the rate-dependent transition
between drained and undrained response (for both contractile and dilatant soil states);
viscous effects; set-up; wedging; and low stress level effects; as well as consolidation
hardening (or softening in case of initially dilatant soil states); and surface roughness
dependence. Their illustration of how these mechanisms interact is reproduced in Figure
4.7.
Figure 4.7 – Mechanisms affecting axial PSI according to HILL et al. [61] and WHITE et al. [62]
(reproduced from [61]).
Still, the framework proposed in [61] and [62] is qualitative. It provides a basis
for understanding the different aspects involved, but relies fundamentally on testing to
establish each of the proposed relationships.
A significant step forward was taken by RANDOLPH et al. [68], who proposed a
complete set of equations, forming an analytical model encompassing all the mechanisms
in the framework proposed in [61] and [62] (illustrated by Figure 4.7). The model
combines:
51
 A critical state-based equation for the generation and dissipation of shear-
induced excess pore pressure (derived from the analogy of surface
shearing on an infinite semi-space).
 A damage model to reproduce the observation in [14] that “even after very
slow shearing where the pipe has mobilised its full drained resistance,
increasing the axial velocity can lead to renewed generation of excess pore
pressure and a significant reduction in axial resistance”.
 A hyperbolic pre-slip stress-displacement relationship
 A strain-rate enhancement function governed by the exponentiation of the

shear strain rate to an exponent dependent on the plasticity index.
 An empirical relationship lumping together the roughness and low stress

level effects.
 The analytical relationship for wedging effect by WHITE & RANDOLPH

[48].
The model was shown able to reproduce the model test results in [14] with
impressive accuracy, as illustrated by Figure 4.8
Figure 4.8 – Results of back-analysis of the model test in [14] using the analytical model proposed by
RANDOLPH et al. [68] (reproduced from [68], courtesy ICE Publishing).
52
4.3.5. Continuum 3D Solid FE Models
To support and validate their analytical model, RANDOLPH et al. [68] presented
the results of local continuum analyses, in which the soil beneath a pipe is modelled using
a 3D mesh of solid elements (eight node hexahedrons). The analyses include coupled
consolidation, and the soil constitutive model is modified Cam clay (MCC). The pipe is
modelled as a rigid surface.
Given the slender nature of the pipeline, the hypothesis of an infinite pipeline is
assumed valid. However, plane strain condition cannot be used as, although being infinite
along the axial direction, both pipe and soil are allowed (and required, for the analysis
purpose) to move in the axial direction. The infinite pipe assumption incurs that any two
points in the model that are aligned over a (imaginary) line parallel to the pipe axis will
present an identical response in all degrees of freedom. This was imposed on their model
using ABAQUS’ multi-point constraints, linking the opposing faces of their thin slice
model, as illustrated by Figure 4.9.
Figure 4.9 – Infinite pipe modelling scheme adopted by RANDOLPH et al. [68] (reproduced from [68],
courtesy ICE Publishing).
CHATTERJEE et al. [58] have improved the same thin sliced 3D models by
including large deformation and automatic model re-meshing techniques. While axial
displacements alone are not likely to cause enough deformation to justify these advanced
features, they were required for modelling the pipe embedment process. As such, the soil
53
response during set-up and subsequent pipe axial displacements will be influenced by the
soil load history, in a more realistic manner than if the pipe was “wished-in-place”, as
previously considered in set-up [49, 57] and axial PSI [68] analyses.
YAN et al. [67] recently used the same thin sliced 3D model to analyse the cyclic
PSI. Their model have shown in details the consolidation hardening occurring within
twenty cycles of axial displacement intervened by full consolidation intervals, as
illustrated by their stress–volume path reproduced in Figure 4.10. This kind of model,
whilst being computationally intensive, is the most accurate numerical tool available for
assessing the axial PSI.
Figure 4.10 – Stress–volume path including pipe embedment, set-up and twenty cycles of axial
displacements intervened by full consolidation, obtained by YAN et al. [67] (reproduced from [67],
courtesy ICE Publishing).
54
5. The Problem of Properly Modelling Time-Dependent
Axial PSI in Global Pipeline Analyses
5.1. Background
The impressive recent progress in understanding time-dependent axial PSI has not
yet been followed by a similar evolution in the available tools for pipeline design analysis.
The continuum 3D models mentioned in Section 4.3.5 are not compatible with global
pipeline models. The analytical model proposed by RANDOLPH et al. [68] could be
implemented into global pipeline models with relative ease. However, it requires
adjusting in an empirical parameter during a continuous cyclic analysis.
The model proposed in this thesis was conceived to stand somewhere in between
the two. It is still computationally cheap and therefore usable in global pipeline design
analyses. However, it is more complex than the analytical model proposed by
RANDOLPH et al. [68], capturing additional aspects that are not compatible with their
planar analogy. With this increased complexity, the proposed model can reproduce with
satisfactory accuracy the complete cyclic analyses presented by YAN et al. [67].
5.2. The Reasons for Pursuing a Coupled Model
None of the existing pipeline design analysis tools are capable of accounting for
all the aspects covered by the new framework proposed by HILL et al. [61] and WHITE
et al. [62]. As such, the candidate’s experience is that this framework has been used in
recent design practice in an “uncoupled” fashion. The soil response is assessed separately
from the pipeline analysis. All relevant time-dependent factors, such as pipe velocity,
consolidation periods, accumulated number of cycles, are taken into account, however,
merely through use of a “representative” value, assumed to account for the response of
the entire pipeline (or a long section of it). From this separate study, time-independent
response curves are obtained, which are used in the conventional tools for pipeline design
analysis. Eventually, curves are replaced between cycles to account for change in the
accumulated number of cycles.
55
However, it is the candidate’s understanding – and this argument was already
pursued in the recent paper by CARNEIRO et al. [16] – that the soil response needs to be
coupled to the pipeline global analysis. The use of a single value of velocity or
accumulated displacement which is “representative” of the complete length of a pipeline
is highly questionable, as different points along the pipeline see completely different
displacements (and thus different velocities) as the pipe expands – as illustrated by Figure
2.2.
CARNEIRO et al. [16], using a simple backbone curve to model a rate-dependent

axial PSI, show the effect of coupling the PSI model with the pipeline global model. The
non-linear distribution of effective axial force observed in Figure 5.1 provides clear
evidence as to different velocities at different locations along the pipeline. Uncoupling
the analyses (whilst using carefully selected “representative” velocities) was shown to
cause the walking rate to vary by up to 30% (up or down, in different analysis conditions)
for the particular cases analysed in [16].
Figure 5.1 – Effective axial force diagram for coupled (“v50 = vk”) and uncoupled (“F50”) analyses
(reproduced from [16], courtesy ASME).
56
This coupling, however, cannot straightforwardly extrapolate from the existing
continuum models, using 3D solid elements, such as those used in [58, 67, 68]. The
incompatibility is due to the large difference in scale between the two types of models, as
previously mentioned. The continuum analyses require thousands to hundreds of
thousands of elements across a pipeline cross section to suitably model the soil beneath
it. CHATTERJEE et al. [58] indicate the use of elements with side length as small as 2%
of the pipe diameter.
Pipelines, on the other hand, are modelled (for the purpose of assessing their
global response) as beam-column elements. This kind of element, combined with the
smoothness of typical seabed topographies, permit achieving very good results out of
coarse meshes. Typical element lengths in global pipeline design analyses may exceed 20
times the pipe diameter. Coarse meshes are required for modelling long pipelines, as well
as for assessing cyclic conditions, which may require running hundreds to many
thousands of load cases. In particular for pipeline walking analysis, it is not uncommon
that analyses may require more than 20 cycles to achieve a response resembling steady
state, even without considering time-dependent PSI (e.g. [69]). Note that each of these
cycles involves gradual heat up and gradual cool down, thus requiring, each, tens of load
steps.
This 1000-fold difference in mesh dimension cannot be accommodated by any of

the available FE formulations known to the candidate. Even with the enormous evolution
of computer technology over the last few years, it is not likely that a global analysis of a
typical pipeline, using a mesh compatible with the continuum soil model, can be analysed
with a run-time compatible with typical design schedules, anytime soon.
To illustrate, in an analogous application, for modelling the lateral PSI, MARTIN

et al. [70] proposed a hybrid model to overcome this scale difference. As indicated in
Figure 5.2, it consists of 2D soil slices, modelled considering plane strain, which are
attached to each node of a global pipe model – with no interconnection between the soil
slices but through the pipeline. The soil analysis did not include consolidation effects, and
used a much simpler soil constitutive model. The model was very short, only 50 m (or
125 diameters) – thus irrelevant if compared to the real length of typical pipelines. The
loading process was equally simplified, comprising a displacement controlled
embedment, followed by a displacement controlled single lateral movement of one
57
diameter only. Still, the run time (“on one node of a cluster with eight cores of 2.53 GHz
Xeon/Nehalem and 24 GB RAM” at the Oxford Supercomputing Centre) varied between
4.5 and 31 hours (for element length to diameter ratios of 25 and 3.1, respectively).
Figure 5.2 – Hybrid modelling scheme using 2D soil slices proposed by MARTIN et al. [70] for lateral
PSI (reproduced from [70], courtesy ICE Publishing).
In summary, means for capturing the recent development in understanding of the

time-dependent soil behaviour into global pipeline design analyses in a coupled manner
(i.e. with the PSI being assessed at every pipeline model node, and updated at every time
step) are needed. However, it cannot be done using the existing continuum models, as the
analyses (if at all possible) would require impracticable run times. The model proposed
in this thesis, as well as the analytical model proposed by RANDOLPH et al. [68], are
seen as viable options for this purpose.
5.3. The Reasons for Introducing Some Extra Complexity
The planar analogy used by RANDOLPH et al. [68] resulted in an extremely

elegant analytical model, which was shown capable of accurately reproducing the
complex response observed in model tests with real soil. Its simplicity, translated in its
straightforward analytical equations, offers an excellent solution for the problem of
modelling time-dependent axial PSI in global pipeline design analyses. If implemented
as a user subroutine in the FE packages commonly used for pipeline analysis, it would
probably result in a very efficient design tool.
58
A key drawback, however, is the need for adjusting the “damage parameter” over
cycles, which was required to achieve the impressive results reproduced in Figure 4.8. Its
initial value, calibrated for the first sweep shown in Figure 4.8(a), was doubled for the
back-analyses of Figure 4.8(b) and Figure 4.8(c). Then, to achieve the results in Figure
4.8(d) – which correspond to the cycles after deliberated overconsolidation under higher
pipe weight – it had to be increased again, this time by a factor of three (on top of the
previous two).
The damage parameter “is a proportionality constant between the potential

volumetric strain rate and a nominal shear strain rate”, where this proportionality itself is
a hypothesis put forward by RANDOLPH et al. [68]. It is clearly dependent on the history
of previous loading, although no way to predict it has been established yet.
Aspects of the soil behaviour that cannot be captured under the planar analogy
could explain some of the features observed in the experiments by WHITE et al. [14],
which have been framed into the damage hypothesis by RANDOLPH et al. [68]. It is
worthwhile transcribing in full two questions posed by WHITE et al. [14] when
highlighting the “key issues (that) remain to be resolved” “to fully understand the results
from (their) model test”:
(a) How is the state of positive ru sustained during continuous sweeping? If

excess pore pressure is generated in the interface zone, why does this not
dissipate during the sweep, leading to a change in the resistance until it
equals the drained value?
(b) Why does the interface appear dilatant initially, before reverting to a state
of positive ru during a single sweep? If the interface is initially dilatant,
why does it appear to pass through the constant volume (critical) state and
revert to a contractile condition, developing positive excess pore pressure
at the residual state? Is another mechanism responsible?†
†
59
While no final answer can be offered, the local FE analyses presented in the following
sections 6 and 7 describe ingredients that contribute in explaining these two initially
unexpected behaviours.
Using plane strain set-up models, Section 6 discusses the two-dimensionality of

the consolidation process. As the soil beneath a pipeline consolidates, pore water seeps
not only radially away from the pipe-soil interface, but also around it towards the draining
seabed surface. Furthermore, the results show that radial component of the flow, rather
than monotonically diminishing as pore pressure equalises, may reverse as a result of this
2D process.
This reversion in radial gradient was already observed on the plot of excess pore
pressure distribution below the pipe invert, obtained from similar plane strain set-up
analyses by JEWELL & BALLARD [59], reproduced in Figure 5.3. It causes, later in the
consolidation process, pore water to be fed back towards the pipe-soil interface, probably
contributing with the sustained excess pore pressure ratio ru, observed by WHITE et al.
[14]. This feature cannot be captured in the planar analogy by RANDOLPH et al. [68],
in which the maximum excess pore pressure always happens in the shear band next to the
pipe-soil interface, decaying away from it following a parabolic distribution, as illustrated
by Figure 5.4.
60
†
Figure 5.3 – Curves of excess pore pressure distribution at particular instants (isochrones) along a vertical
line beneath the pipe invert, obtained from plane strain set-up analyses by JEWELL & BALLARD [59]
(reproduced from [59]†).
Figure 5.4 – Schematic of the planar analogy by RANDOLPH et al. [68], indicating the parabolic
distribution of excess pore pressure distribution below the shear band next to the pipe-soil interface
(reproduced from [68], courtesy ICE Publishing).
†
61
Section 7 explores further aspects of the radial flow of pore water, using an
axisymmetric model to investigate the effects of differences in soil state, partial drainage
and shear induced excess pore pressure. It adopts the same artifice of using multi-point
constraints to simulate an infinite pipeline (whilst allowing axial displacements)
employed in the 3D models in [58, 67, 68]. As such, the 2D axisymmetric mesh is in
effect reduced to 1D. Initial soil state is arbitrarily varied along the model, imposing a
higher overconsolidation ratio next to the pipe-soil interface, thus representing an
idealisation of the effects of the history of previous loading.
Results indicate an excess pore pressure migration mechanism, in which excess

pore pressure generated at a soil region may influence the strength of another region (the
peak strength in some particular conditions, or the strength under sustained displacement
post peak). Of course this can only happen in partially drained conditions – in fully
undrained conditions the excess pore pressure would not have time to migrate, while in
fully drained conditions this transient migration would be an irrelevant stage before full
dissipation. An analogous excess pore pressure migration process was proposed much
earlier, in 1955 by WARD et al. [71], to explain the delayed failure of a bank. Such a
mechanism may also contribute to the observed sustained ru. It may also explain the
observation of excess pore pressure at regions where no excess pore pressure generation
was expected – or of positive ru where dilatant response was expected (or even observed,
initially).
The mechanisms observed from the two FE models used in the studies in sections
6 and 7 justify the pursuit for a more complex model, one which is able to mimic (even
if in a simplified manner) the interactions between the different regions of the soil beneath
a pipeline. A model capturing these extra features, whilst still not too computationally
intensive, is expected to provide subsea pipeline designers with means to reduce
conservatism and improve the reliability in design, in particular for pipelines sensitive to
axial PSI – which is often the case for high temperature, high pressure flowlines.
62
6. Two-Dimensional Consolidation beneath a Pipeline
6.1. Background
As previously discussed in Section 4.3, the set-up of a pipeline – the dissipation

of the excess pore pressure generated in the soil beneath it during its installation – has
been previously assessed via FE analysis. KROST et al. [57] studied this process using
coupled consolidation FE analyses of elastic soil, benchmarking the numerical results
against analytical solutions for consolidations beneath strip foundations and to field
measurements from an in situ model test. GOURVENEC & WHITE [49] proposed a
closed-form expression, curve-fitted to the results of average excess pore pressure
dissipation along the pipe-soil interface from the same FE analyses:
2
w  w
0.33 0.12  0.38 
 cvt  D D
 
u  D 2
 ( 6-1 )
U  1 
u0  cv t w w 
2
 2  0.04  1.7  1.7   

D D D 
Where the “degree of consolidation” U is defined in terms of the average excess pore
pressure along the pipe-soil interface u and its initial value u0 (which is obtained from
the undrained loading step, before consolidation is allowed, in the FE analysis).
CHATTERJEE et al. [58] studied the same problem using much more
sophisticated FE models, including MCC constitutive model and large deformation re-
meshing techniques, which permitted realistic modelling of the pipe embedment process
(as opposed to the “wished-in-place” initial condition adopted in the elastic models).
While relevant differences are observed in many results, those of average excess pore
pressure dissipation along the pipe-soil interface agree fairly well with the elastic models.
A closed-form hyperbolic expression for the degree of consolidation observed in these
elastoplastic consolidation models has later been proposed by CHATTERJEE et al. [72]:
1
U  1 ( 6-2 )
1   t / t50 
m
63
Where the time t50 required for dissipation of 50% of u0 and the exponent m are given,
for different values of embedment, in Table 6.1.
Table 6.1 – Values of t50 and m to use in Equation ( 6-2 ), proposed by CHATTERJEE et al. [72].
Embedment ratio cv t50

m
w/D D2
0.1 0.015 0.85
0.2 0.032 0.88
0.3 0.052 0.93
0.4 0.075 1.00
0.5 0.090 1.05
One of the aspects highlighted by KROST et al. [57] and GOURVENEC &
WHITE [49] – and previously by SCHIFFMAN et al. [73] in their analytical solutions
for strip foundations, used by [57] and [49] for benchmarking – is the so-called
MANDEL-CRYER effect. This effect is characterised by an increase in the excess pore
pressure at early times over the initial excess pore pressure. It was first identified by
MANDEL [74], in his analytical solution for a flat layer of soil compressed between two
impermeable frictionless rigid walls; and later by CRYER [75] in his analysis of a sphere
of soil subjected to a uniform hydrostatic pressure.
CRYER [75] and SCHIFFMAN et al. [73] showed that, in both their analyses, the
Mandel-Cryer effect is more relevant for smaller values† of Poisson’s ratio  (with no rise
in excess pore pressure for  = 0.5 in [75]), so different values of  were initially
considered in the analyses. KROST et al. [57] and GOURVENEC & WHITE [49]
indicated, however, that for the case of partially embedded pipes, this effect is more
relevant for small values of w/D. As the analyses herein have adopted w/D = 0.5, the
results presented minimum sensitivity to the Poisson’s ratio.
Another aspect to consider is that, as observed in the results of GOURVENEC et

al. [49], the rise in excess pore pressure is much more evident at the pipe invert, being
†
Only the interval 0   < 0.5 is considered.
64
smoothed out when average excess pore pressure along the pipe-soil interface is plotted.
This can be observed in Figure 6.1, which reproduces plots from GOURVENEC et al.
[49] of excess pore pressure u measured at the pipe invert (normalised by its maximum
value umax at the beginning of the consolidation process) and of degree of consolidation
U, (obtained from their FE model, as well as from Equation ( 6-1 )) over time t. For this
reason (i.e. to minimise any interference from the MANDEL-CRYER effect), similar
results hereafter are presented in terms of degree of consolidation.
65
†
Figure 6.1 – Results by GOURVENEC et al. [49] of excess pore pressure u at the pipe invert and degree
of consolidation U, (obtained from their FE model and from Equation ( 6-1 )) (reproduced from [49]†).
All of the mentioned studies focused on the excess pore pressure along the pipe-
soil interface. From similar elastic analyses, JEWELL & BALLARD [59] presented
instead the curves of excess pore pressure through depth along the symmetry axis at
†
66
particular instants (vertical isochrones), the plot reproduced in Figure 5.3. These curves
present an interesting feature, which is the reversion of the hydraulic gradient as the
excess pore pressure dissipates. JEWELL & BALLARD [59] mention that “the point of
maximum excess pore pressure moves progressively outward with time”. This statement,
while correct, does not highlight the main consequence, which is the change in direction
of the radial component of pore water flow next to the pipe-soil interface.
The consolidation beneath a pipeline can become a much more complex process
(than the monotonic set-up post installation) when a pipeline starts to move axially. If
consolidation is slow, further excess pore pressure can be generated before the set-up is
complete. On the other hand, if the soil can consolidate quicker, the dissipation during
the pipeline movement may not be negligible, so that the dissipation would need to be
assessed combined with continuous generation of further excess pore pressure (as the pipe
moves). As such, further understanding of the complex two-dimensional (2D) dissipation
process is considered of value.
This short section contributes to this understanding by presenting some further

interpretation of the results of elastic set-up FE analyses, which are in essence identical
to those in [49, 57, 59]. The key contribution offered consists in the cross interpretation
of the results in the two directions: circumferentially along the pipe-soil interface (as in
[49, 57]) and radially below the pipe invert (as in [59]). In addition, the shape of the
circumferential isochrones is investigated, which is of relevance for the proposed PSI
model, as further discussed in Section 8.
6.2. FE Model, Results and Discussion
The analyses were carried out using the commercially available FE software
ABAQUS [29]. The pipe was modelled as a rigid frictionless contact surface. The soil
medium was modelled in plane strain condition using second order elements with coupled
pore pressure (8-node biquadratic displacement, bilinear pore pressure, reduced
integration, CPE8RP elements), assigned with constant permeability and isotropic linear
elastic properties. The mesh was conveniently structured in the polar coordinate system
over one quadrant – assuming the pipeline embedment w equal to 50% of its diameter D,
67
and taking benefit of symmetry – on constant angular steps and increasing radial
increments, as illustrated by Figure 6.2.
Reference point on which vertical load is applied

Rigid surface
Contact interface (frictionless)
Free, free-draining top boundary
Vertical axis of symmetry
Figure 6.2 – FE model geometry and boundary conditions.
The analyses assumed small displacements (i.e. linear geometric, with equilibrium
assessed considering the original geometry), and were conducted in two consecutive
stages. First, an arbitrary vertical load was applied to the reference node at the pipeline
centroid over a very short time step. Then, the “drainage” boundary condition (i.e. zero
68
excess pore pressure) was specified to the seabed surface and the model was allowed to
consolidate with no change in vertical load.
To minimise instability (which may cause spurious oscillations in excess pore

pressure along a drainage path), the initial time steps for the consolidation stage followed
the accuracy condition derived by VERMEER & VERRUIJT [76]:
1 h2
t  ( 6-3 )
6  cv
Where  is the time integration parameter (refer to Section 3.3) – which shall be taken as
unity as ABAQUS employs forward interpolation – and h is the characteristic element
size – which was taken as the distance between integration points in the elements at the
pipe soil interface. Although the element sizes increase away from the pipe, the excess
pore pressure gradients are typically higher next to this interface. Adopting this h has
resulted in limited instability only at the intersection between the pipe-soil interface and
the seabed surface, which is already a numerical discontinuity point – thus already
problematic in FE. These limited inaccuracies, however, seem not to compromise the
overall model results, which all appear quite stable, as shown hereafter.
Figure 6.3 presents vector plots of apparent pore water flow velocity (Darcy
velocity) at three selected instants. This graphic output, while not suitable for extracting
quantitative results, permit a very interesting qualitative visualisation of the different
stages of the consolidation process, providing valuable insight on how the two directions
of flow (radial and circumferential) are interdependent. The structured mesh in the
background provides a convenient reference grid in the polar coordinate system, assisting
in clearly identifying the two directions. The three instants are identified by their
consolidation time t – measured from the moment when the drainage boundary condition
is applied – normalised by the soil cv (which is constant, as the soil model is linear elastic)
and the pipe diameter D.
In all cases, the observed flow is predominantly circumferential towards the

draining boundary. Still, the radial component of the flow vectors seems relevant.
Illustrative arrows in Figure 6.3 highlight the two components (radial and circumferential)
of the flow next to the pipe, as follows. Figure 1(a) present significant outward flow;
Figure 1(b) shows essentially circumferential flow next to the pipe whilst a less
69
considerable outward component further away from it; while a reversed radial flow
component (now inwards) is observed in Figure 1(c).
70
(a)
(b)
(c)
Figure 6.3 – Vectors of apparent pore water flow velocity at three distinct instants: (a) cv t / D2 = 0.02,
(b) cv t / D2 = 0.1 and (c) cv t / D2 = 0.9. Illustrative arrows highlight the two components (radial and
circumferential) of the flow next to the pipe-soil interface.
71
As discussed, the graphical appeal of the vector plots makes this effect clear, but
such a reversal in flow direction can also be inferred from the vertical isochrones
presented by JEWELL & BALLARD [59], which are reproduced in Figure 5.3. Figure
6.4 presents these same vertical isochrones (i.e. profiles of excess pore pressure along a
vertical line beneath the pipe invert at particular instants) for the FE results from which
the plots in Figure 6.3 were extracted. The excess pore pressure u is normalized by the
maximum observed value umax (at the pipe invert, just after the vertical load is applied).
The depth z is normalised by the pipe diameter D, with origin at the pipe centreline (for
which the results start at z/D = 0.5).
The isochrones (a), (b) and (c) correspond to the same instants of Figure 6.3(a),
(b) and (c), respectively. Curves (i) and (f) correspond to the boundaries (minimum and
maximum values of the horizontal axis) of Figure 6.6. The reversal of hydraulic gradient
(and consequently the reversal of flow) next to the pipe is highlighted by the indicative
dashed lines. Initial positive gradient (pushing the pore water downwards) at instant (a)
goes to zero (vertical line next to the pipe) at instant (b), then reversing (again, next to the
pipe) at instant (c). At this last instant, the point of maximum excess pore pressure is at
about z/D = 2. Above that point, the gradient is negative, causing the flow to present (for
z/D < 2) a radial component towards the pipe.
72
u / umax
0.0 0.2 0.4 0.6 0.8 1.0
0
(f) (c) (b) (a) (i)
Zero
gradient Positive gradient,
water flows down
2
Negative gradient,
water flows up
z/D
5
Figure 6.4 – Excess pore pressure along a vertical line beneath the pipe invert at five distinct instants:
(i) cv t / D2 = 0.001, (a) cv t / D2 = 0.02, (b) cv t / D2 = 0.1, (c) cv t / D2 = 0.9 and (f) cv t / D2 = 10. Dashed
lines indicate approximate gradient next to pipe-soil interface.
As opposed to the changing vertical isochrones shown in Figure 6.4, the shape of
the circumferential isochrones – curves of excess pore pressure along arcs of constant
radius  (concentric with the pipe) at particular instants – remained approximately
unchanged over the analysis. Figure 6.5 presents circumferential isochrones at three
different arcs (identified by their radial coordinate ) and three particular instants. The
horizontal axis indicates the position around the circumference – given by the angular
coordinate , which is measured from the bottom invert of the pipe (i.e.  = 0 at the invert
of each arc). Each one of the nine isochrones was normalised by the excess pore pressure
at the arc invert, at the particular instant, ui. As such all isochrones intercept the vertical
axis at u/ui = 1, for  = 0.
It can be observed that the normalised isochrones are all very close to a cosine
curve, being slightly off for large  (i.e. away from the pipe-soil interface) and small t
73
(i.e. earlier in the consolidation process). This proximity is later used in the proposed PSI
model, which assumes that the circumferential isochrones follow the cosine distribution.
u/ui
cos 
0.8
0.6
cos 
cvt/D2 = 0.01
cvt/D2 = 0.1
cvt/D2 = 1
0.4
/D = 0.5
0.2
/D = 1.0
/D = 2.0
cos 
0
0 /4
0.5 1/2

Figure 6.5 – Normalised results of excess pore pressure along arcs of constant  at particular instants t.
Figure 6.6 shows the average excess pore pressure along the pipe-soil interface,
represented in terms of degree of consolidation U over time. The three intermediate steps
depicted in Figure 6.3 are indicated with the corresponding label. For illustration
purposes, three other curves are superimposed. As expected, very good agreement is
observed with the curve obtained using the expression proposed by GOURVENEC &
WHITE [49], reproduced as Equation ( 6-1 ) – since (as aforementioned) it was fitted to
results of models essentially identical to the present one. A poorer agreement is observed
with the curve obtained using the expression proposed by CHATTERJEE et al. [72],
reproduced as Equation ( 6-2 ). This was also expected, as they have used different
(elastoplastic) models. This difference between the elastic and elastoplastic set-up results
has already been discussed by CHATTERJEE et al. [58].
74
Finally, to qualitatively assess the effect of the radial component of flow on the
circumferential flow, the obtained results were compared with the dissipation curve from
the classical TERZAGHI’s one-dimensional consolidation theory. The vector plots in
Figure 6.3 indicate that the flow is predominantly circumferential. It could be thus
considered that the differences between the FE results and the results of 1D consolidation
along a drainage path length of D/4 – which corresponds to one quarter of the pipe
circumference – are due to the influence of the radial component. Based on the
circumferential isochrones shown in Figure 6.5, the initial distribution of excess pore
pressure following a cosine curve was considered. The degree of consolidation in the
classical consolidation theory, for initial distribution following a cosine curve, is given
by (e.g. TAYLOR [77]):
 2 
U  1  exp   T  ( 6-4 )
 4 
where the dimensionless time T, for a drainage path length of D/4, is:
cv t
T ( 6-5 )
 D / 4 
2
The FE results (solid line) show initially a substantially quicker dissipation than
the 1D results (dashed line), which is consistent with the water draining in both directions
(radial outwards and circumferential towards the surface), followed by a raised tail, being
fed by the inwards radial draining component as the hydraulic gradient in this direction
reverses.
At instant (a), the solid line is below the dashed line. This is consistent with the
pore water flow pattern in Figure 6.3(a), draining away from the pipe. The circumferential
drainage next to the pipe is enhanced, in this early phase, by the radial component and as
a result, U is lower than the 1D solution would indicate. At instant (b) the two curves are
close to parallel (i.e. presenting a similar rate of consolidation). Figure 6.3(b) shows flow
essentially tangential in the vicinity of the pipe. Figure 6.3(c) shows reversed radial
components, now draining inwards (and up) throughout the soil medium, again consistent
with Figure 6.6, which shows endurance in the FE results of excess pore pressure.
75
0.0
(a) (b) (c)
U
TERZAGHI 1D
Eq. ( 6–4 )
0.2
FE analysis results
0.4
GOURVENEC & WHITE

Eq. ( 6–1 )
0.6
CHATTERJEE et al.
Eq. ( 6–2 )
0.8
1.0
0.001 0.01 0.1 1 cv t / D2 10
Figure 6.6 – Degree of consolidation (average along pipe-soil interface) over time obtained from the 2D
FE analysis results, as well as from equations ( 6-1 ), ( 6-2 ) and ( 6-4 ).
While this interaction between circumferential and radial components of the pore
water flow is found insightful, the presented results are still limited to the set-up following
an initial charge of excess pore pressure (due to a vertical load increment). The continuous
generation of excess pore pressure during pipeline axial movement cannot be captured
within plane strain models.
76
7. Pore Pressure Migration
7.1. Background
Owing to the plane strain assumption, the models used in Section 6 cannot account
for the axial displacements of the pipeline. As a straightforward extension of the plane
strain models to account for the pipe axial displacements, RANDOLPH et al. [68]
developed the 3D “slice” models mentioned in Section 4.3.5, which were also used later
by CHATTERJEE et al. [58] and YAN et al. [67]. As previously described, these models
consist in continuum solid FE models, using 3D mesh, in which multi-point constraints
(MPC) impose the infinite pipe condition.
The MPC are kinematic constraints between nodes, in which degrees of freedom
of a given node are eliminated from the FE equilibrium equations, being submitted to the
corresponding degrees of freedom of another reference node. In the 3D slice mode, all
the degrees of freedom (DOF) of one side of the slice are eliminated, being forced to
present the same value calculated at the corresponding DOF at the opposite side of the
slice, as illustrated by Figure 4.9.
The study presented herein does not follow this straightforward extension to the
3D slice. A step back was taken, thus resulting in simpler models. The models consider
the same type of slice of infinite pipe. However, they do not include a complete 2D cross
section. Instead, they assume symmetry around the pipeline axis. The intention was to
decouple the effects of the axial pipe movement from the 2D consolidation studied in
Section 6. By assuming the more simple axisymmetric condition, two-dimensional effects
are disregarded, and the study can focus on the implications of the pipeline axial
displacements. Of course this is an idealisation, not expected to give realistic results.
However, looking at one aspect at a time permits a clearer understanding of particular
mechanisms. The understanding of the complete 3D geometry shall of course account for
the complex interaction between all the mechanisms involved.
One key feature studied is the effect of the soil state adjacent to the pipe being
different from that further away in the soil mass. The intention was to investigate the
build-up of positive excess pore pressure at highly overconsolidated soil next to the pipe
interface, as observed by WHITE et al. [14] (and highlighted as one of the key unexpected
77
behaviours, as indicated by their question (b) transcribed in Section 5.3). The hypothesis
proposed is that the positive excess pore pressure measured in the tests by WHITE et al.
[14] was in fact not generated at the pipe-soil interface. It would have been generated
elsewhere in the soil mass, then migrated towards the pipe wall.
Literature review indicates that an analogous process of excess pore pressure

migration (with excess pore pressure generated at one location influencing the response
of another location) may have caused the failure of a bank built on a thin peat layer, as
suggested by WARD et al. [71]. A cross section of the bank after and before the slip is
reproduced in Figure 7.1. Their investigation lead to the conclusion that the excess pore
pressure generated beneath the centre of the bank would have migrated horizontally
through the peat layer, eventually reducing the effective vertical stress beneath the berm
(between the toe of the bank and the edge of the diversion channel) to a level which caused
instability.
Figure 7.1 – Section through bank slip studied by WARD et al. [71]. Top: post failure geometry,
indicating piezometric heads two months after the slip; Bottom: original profile, indicating active and
passive thrust wedges considered in their back-analysis (reproduced from WARD et al. [71], courtesy
ICE Publishing).
Their back-analysis considered an idealised triangular initial excess pore pressure

distribution, with maximum value at the centreline and zero at x=a, as indicated in Figure
7.2. Isochrones in Figure 7.2 show that significant excess pore pressure is observed at
regions with initially no excess pore pressure, in particular for x/a just above unity. It
should be highlighted, however, that this excess pore is transient, and will eventually
dissipate. Figure 7.2 shows that, for x/a > 1, the excess pore pressure increases
78
monotonically up until cvt/a2 = 0.25. After that, the excess pore pressure will decay
(although further isochrones are not plotted). A plot of excess pore pressure over time at
a point with x/a > 1 would indicate that the zero initial excess pore pressure would
gradually increase, but then dissipate (thus presenting both initial and final values nil,
interceded by positive transients).
Figure 7.2 – Isochrones along the peat layer in the back-analysis by WARD et al. [71], indicating
significant excess pore pressure redistribution (reproduced from WARD et al. [71], courtesy ICE
Publishing).
Within the tests reported by WHITE et al. [14], if the soil adjacent to the pipe was
in critical state (thus presenting no tendency to generate excess pore pressure) or even in
a dilatant state (i.e. presenting a tendency to generate negative excess pore pressure when
sheared), but positive excess pore pressure was generated elsewhere in the soil as the
model pipe moved, a similar redistribution process could eventually lead to positive
excess pore pressure being measured at the pipe-soil interface. The axisymmetric models
used in this section were run under a range of conditions, trying to reproduce such
behaviour.
7.2. Model Details and Run Cases
As in the previous model, the analyses were carried out using ABAQUS [29]. A
2D mesh of first order, coupled pore pressure, axisymmetric elements (CAX4P elements)
79
was laid out on the axial-radial place, as illustrated in Figure 7.3 (note that the element
sizes are exaggerated for visualisation purposes only). A single row of elements was used,
with MPC tying the nodes of same radial coordinate  at opposite sides of each element
(thus representing the infinite length assumption).
Figure 7.3 – Axisymmetric model schematics, in particular a case with overconsolidated soil adjacent to
the pipe and normally consolidated soil elsewhere.
The soil constitutive model used was the ABAQUS implementation of modified
Cam clay (called “clay plasticity”). The soil parameters used in the model (which were
arbitrarily taken from [78]) are listed in Table 7.1. For simplicity, the model initial stress
state is uniform, hydrostatic and fully consolidated (i.e. the initial deviator stress and the
initial excess pore pressure are q = 0 and u = 0, and the initial mean effective stress p' = p'0,
80
which was arbitrarily selected, is constant over the entire model). Initial void ratios were
selected to impose the desired initial overconsolidation ratios under the arbitrary initial
stress state.
Table 7.1 – Soil parameters selected for the axisymmetric model.
Parameter Symbol Value

Slope of the CSL and the NCL in the e–ln p' space  0.15
Slope of the SRL in the e–ln p' space  0.03
Stress ratio on the CSL  1.20
Intercept of the NCL in the e–ln p' space (1)  1.48
Poisson’s ratio(2)  0.30
Notes:
(1) Since all the stresses were normalised by the arbitrarily selected value of p'0, this was considered as
the unity (thus the origin of the ln p' axis). The intercept  thus correspond to the void ratio for p' = p'0
and R = 1.
(2) As discussed in Section 3.4 it was chosen to consider the Poisson’s ratio  as a constant, so that the
elastic coefficients mv and G are calculated (automatically by the software at each analysis step) from
the constant  and the stress-dependent bulk modulus K as per Equation ( 3-29 ).
The pipe was not modelled. The shearing imposed by the pipeline axial movement
was considered by prescribing a constant velocity to the pair of nodes corresponding to
the pipe-soil interface. This implies that no relative sliding is permitted. The total
prescribed displacement  (gradually applied under constant velocity) was 10% of the
pipe diameter, with some models experiencing numerical convergence issues. These were
not worked further as, in general, they occurred beyond the point of maximum strength
(either a peak value or close to an asymptotic response), which were observed for  / D
in the range of approximately 2% to 4%.
The pair of nodes at the pipe soil interface was fixed in the radial direction (thus
preventing any change in diameter of the pipe-soil interface as the soil contracts or
expands). As a result, the total radial stress is not constant, but reduces as the soil tends
to contract or increases as the soil tends to expand. The initial stress p'0 was set to a value
large enough that this effect is minor, although it can still be observed as discussed
hereafter. The opposite pair of nodes (farther from the pipe) were fixed in both (radial
and axial) directions, and set as a draining boundary (i.e. boundary condition u = 0). These
81
nodes were placed suitably far from the pipe-soil interface. The analyses were non-linear
geometric (i.e. with equilibrium assessed considering the deformed geometry).
The soil initial overconsolidation ratio was varied for the different analysis cases,
which considered R = 1.0, 1.4, 2.0 and 2.4. These were obtained by setting the initial
voids ratio to e = 1.48, 1.44, 1.40 and 1.38, respectively. The first case corresponds to a
normally consolidated initial state, the other three to increasing levels of
overconsolidation. In the case where R = 2.0, the stress path will start from right below
the apex of the projected yield surface. If the loading is such that q increases from the
initial q = 0 with no change in p', it will reach the yield surface at the CSL. If any reduction
in p' is observed, yielding will occur with stress ratio  > . This is also likely to occur
for the most overconsolidated case of R = 2.4.
It is recognised that the MCC, as proposed by BURLAND [26] and ROSCOE &
BURLAND [27], was meant to be used for “wet” clays, in which yielding occurs with
  . The MCC yield surface is not suitable for yielding at the “dry” side (i.e. with
 > ), for which other yield functions have been proposed (e.g. the HVORSLEV surface
as discussed by ATKINSON & BRANSBY [30]). The ABAQUS implementation of
MCC, however, makes no such distinction. While a better representation of yielding at
the “dry” side would lead to different results, it is not expected that it would change the
qualitative conclusions of the present study.
Four benchmark cases were run with a single value of R over the entire model.
Then the initial soil state was modified over a layer of thickness hoc as indicated in Figure
7.3. The initial soil state is uniform throughout this layer, between the pipe-soil interface
(indicated as location A in Figure 7.3) and the interface at location B. At this location, a
step change in initial soil state is imposed, with soil beyond this point being also uniform
(although with a different initial soil state). Figure 7.3 illustrates the case in which the soil
beyond B is normally consolidated (NC), whilst being overconsolidated (OC) between A
and B.
The soil beneath a pipeline is sheared during pipe lay and during subsequent axial
displacement cycles. The effect of any load would decay with the distance from the pipe
– being inversely proportional to the radius, in the case of axisymmetric conditions. It is
hence plausible that the effect of previous loads would concentrate next to the pipe
82
interface. In the present study, the load history is simplified by a layer, adjacent to the
pipe wall, presenting a higher degree of overconsolidation. Rather than trying to
reproduce the gradual decay, a homogeneous soil with a step change in initial void ratio
at an arbitrary distance from the pipe wall was adopted.
Three different values were used for the thickness of the layer between A and B,
namely hoc/D = 0.1, 0.2 and 0.4. These were arranged with different combinations of
initial R within this layer and beyond it, as detailed in Table 7.2. The first four cases in
Table 7.2 correspond to the benchmark cases (i.e. with uniform initial state throughout
the model). In all other cases, combination is such that initial R is always higher in the
layer between A and B, hence called “overconsolidated layer”. Note that in many cases
both parts of the model are over consolidated, but in those cases the “overconsolidated
layer” always present a higher degree of overconsolidation (i.e. a higher initial R).
83
Table 7.2 – Analysed cases.
Initial R in layer between A and B Initial R beyond B Layer thickness hoc/D

1.0 1.0 N/A
1.4 1.4 N/A
2.0 2.0 N/A
2.4 2.4 N/A
2.4 1.0 0.1
2.4 1.0 0.2
2.4 1.0 0.4
2.4 1.4 0.1
2.4 1.4 0.2
2.4 1.4 0.4
2.0 1.0 0.1
2.0 1.0 0.2
2.0 1.0 0.4
2.0 1.4 0.1
2.0 1.4 0.2
2.0 1.4 0.4
1.4 1.0 0.1
1.4 1.0 0.2
1.4 1.0 0.4
For each of the cases listed in Table 7.2, the imposed constant velocity v was
varied over five orders of magnitude, from vD/cv = 10–3.5 to vD/cv = 101.5 in exponential
steps of 0.5 (i.e. vD/cv = 0.0003, 0.001, 0.003, 0.01, 0.03, 0.1, 0.3, 1, 3, 10 and 30), thus
resulting in a total of 209 analyses. For each case, the cv was calculated using the constant
k and w, and the initial values of e and p'. For models with the “overconsolidated layer”,
the initial e at this layer was used. The differences in e for the considered range of R lead
to differences in cv between models, but this is of only 2% from the average cv.
The velocities were set considering the average cv. This way, for each of the 11
aforementioned nominal values of vD/cv, a single constant velocity was used across all
84
models, with the actual values of vD/cv thus varying by 2%. As observed hereafter, the
scatter around the nominal values (in plots where vD/cv is the horizontal axis) is
negligible. Such a small difference makes this study (with the particular ranges of input
data considered) insensitive to the discussion on whether vD/cv is the relevant
normalisation or should v/k be considered instead. For information, the considered input
data results in v/k  830 vD/cv.
7.3. Results for Uniform Cases
Before analysing the results of the models with the overconsolidated layer, the
results from the benchmark cases – those with uniform soil, with the same initial state
(same initial R) everywhere – are discussed. Figure 7.4(a) presents the maximum
observed values of q at the nodes at the pipe-soil interface as these moved with the
prescribed constant velocities†. This was used as a measure of the soil strength, or
resistance to pipeline axial movement.
The obvious observation is that the two most overconsolidated cases presented
constant strength, insensitive to the applied velocity, whereas the two cases with lower R
presented the drained to undrained transition in the usual form of a backbone curve. The
curve-fits presented follow the usual expression:
  1   v  m 
qmax  qmaxD   qmaxD  qmaxU  exp  ln      ( 7-1 )
  2   v50  
 
Which is essentially the same as that proposed by RANDOLPH et al. [68]. The terms
qmaxD and qmaxU are, respectively, the value of qmax for the fully drained and fully undrained
conditions. The curves overlaid in Figure 7.4(a) were obtained with v50D/cv = 0.2 and
m = 0.6 for R = 1.0, and v50D/cv = 0.4 and m = 0.5 for R = 1.4.
†
Note that one point (for R = 1.0 and vD/cv = 30) is missing. This is due to numerical convergence issues.
This missing condition, however, does not compromise the overall conclusions, and no further attempts to
improve convergence were made.
85
1.5
qmax (a)
p'0
R = 2.4
1.2
R = 2.0
R = 1.4
0.9
Curve-fits R = 1.0
0.6
0.0001 0.001 0.01 0.1 1 10 100
vD/cv
(b)
0.4
umax
R = 1.0
p'0.3
0
0.2
R = 1.4
0.1
R = 2.0
0.0
R = 2.4
-0.1
0.0001 0.001 0.01 0.1 1 10 100
vD/cv
Figure 7.4 – Maximum observed values of (a) deviator stress qmax and (b) excess pore pressure umax at the
nodes at the pipe-soil interface as these moved with the prescribed constant velocities v.
Figure 7.4(b) presents the results of maximum (absolute value) excess pore
pressure umax for the same cases. The correspondence between the positive excess pore
86
pressure observed in the two less overconsolidated cases is evident. The development of
positive excess pore pressure is also obvious in Figure 7.5, which illustrate the models’
behaviour by detailing the results for one particular intermediate velocity (vD/cv = 0.1,
for which the effect of partial drainage is well noticeable). Figure 7.5(c) shows that, for
these cases, the strength profile presents no peak. This was the case for all the runs with
R = 1.0 and 1.4.
For R = 1.4, Figure 7.5(b) and (c) indicate zero excess pore pressure up until
yielding. This is a consequence of the MCC constitutive model. As the imposed velocity
is tangential, it should (initially) impose shear strains only. These would be associated
with shear stresses, which would induce changes to q (but not to p or p'). As such, the
stress state representation in the e–p'–q space would move along the q direction within
the elastic wall, with no change in p' or e. It is only after yielding that the associated
plastic flow hypothesis couples the volumetric strains with the shear strains. With this
coupling, changes in q may induce changes in p' and e, which are translated in generation
of excess pore pressure (except if full drainage is allowed).
A relevant aspect observed in Figure 7.5(a) is that, as soon as yielding occurs, the
reduction in e is followed by a reduction in total stress p. This is due to the confined
boundary condition, as aforementioned. As the model cannot change its total volume, a
soil tendency to shrink causes p to reduce.
An accentuated turn in the effective stress paths, just before critical state is
reached, can also be observed in Figure 7.5(a) and (b). This is a consequence of the late
reduction in u observed in Figure 7.5(c). As the soil approaches critical state, the rate at
which positive excess pore pressure is generated reduces. At some point, this rate
becomes lower than the rate at which excess pore pressure dissipates (as pore water drains
radially away from the pipe-soil interface). At this point, u starts to fall, which causes p'
to rise. This effect is not observed (or is observed, although with lower importance) for
higher velocities – as the rate of dissipation becomes proportionally lower, not overtaking
the rate of generation before critical state is reached.
87
1.50
(a)
e SRL
R = 1.0
1.45
p'/p'0
p/p'0
R = 1.4
1.40
CSL NCL
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0 p'/p'0
1.0
p/p'0
p
0.5
p'/p'0 R = 1.4
R = 1.0
CSL
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.2 0.4
(c) q/p'0 u/p'0
q/p'0
0.9 0.3
R = 1.4 u/p'0
0.6 0.2
R = 1.0
0.3 0.1
0.0 0.0
0 1 2 3  /D (%) 4
Figure 7.5 – Results of the analyses with vD/cv = 0.1, for R = 1.0 and R = 1.4: (a) stress–volume paths (for
both total stress p and effective stress p') in the e–ln p / e–ln p' space, including the SRL corresponding to
the initial conditions; (b) stress paths in the p–q / p'–q space, including the initial projected yield
functions; (c) q and u for axial displacement .
88
Figure 7.4 also shows that negligible excess pore pressure is generated for R = 2.0,
while the case with R = 2.4 presents negative excess pore pressure for the less drained
cases. This negative excess pore pressure, however, does not interfere in the observed
values of qmax as the excess pore pressure in these cases is only generated in the post peak
response (i.e. after qmax is observed), as observed in Figure 7.6(c)†.
Similarly to the results for R = 1.4, q increases with no change in p' or e up until
yielding. However, for the two most overconsolidated cases, the maximum strength
corresponds to yielding. For R = 2.0, yielding occurs at critical state already, and hence
no change in p' or e is at all observed. The deviator stress q also maintains its maximum
value qmax for further displacements.
For R = 2.4, however, yield occurs at the “dry” side of the yield surface, with
 > . Post yield, q suddenly drops below qmax and negative excess pore pressure is, also
suddenly, generated. At this same time, an initial increase in e is followed by an increase
in total stress p. After that, the negative excess pore pressure gradually dissipates – which
(similarly to the low overconsolidation cases) causes a bend in the stress path – and the
final results, including q, tend to the same observed for R = 2.0.
†
Results of umax for the two highest velocities are not shown in Figure 7.4(b), although the corresponding
results of qmax are presented in Figure 7.4(a). This is because numerical convergence issues were
experienced just after peak strength was reached. Again, these missing points do not compromise the overall
results and conclusions, not being further pursued.
89
1.50
(a)
e
CSL NCL
1.45
SRL R = 2.0
1.40
p/p'0
R = 2.4 p'/p'0
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0 R = 2.4 p'/p'0
p/p'0
1.0
p
R = 2.0
0.5
CSL
0.0
0.0 0.5 1.0 p/p'0, p'/p'0 1.5
1.5 0.25
q/p'0 (c) u/p'0
R = 2.4
1.2 0.20
R = 2.0
0.9 0.15
q/p'0
0.6 0.10
0.3 0.05
u/p'0 R = 2.0
0.0 0.00
R = 2.4
-0.3 -0.05
0 1 2 3  /D (%) 4
Figure 7.6 – Results of the analyses with vD/cv = 1, for R = 2.0 and R = 2.4: (a) stress–volume paths (for
both total stress p and effective stress p') in the e–ln p / e–ln p' space, including the SRL corresponding to
the initial conditions; (b) stress paths in the p–q / p'–q space, including the initial projected yield
functions; (c) q and u for axial displacement .
90
7.4. Selected Results for Cases with an Overconsolidated Layer
All the 209 analyses derived from the combinations in Table 7.2 were run and
analysed. This section (complemented by Appendix A) presents just a small selection of
those, in particular the cases in which the excess pore pressure migration process being
proposed is more noticeable. Firstly, results of qmax and umax at the pipe-soil interface for
the complete range of velocities, for the particular cases where the initial R = 2.4 within
the overconsolidated (OC) layer and R = 1.0 elsewhere, are presented in Figure 7.7. The
benchmark cases with these same values (i.e. uniform R = 2.4 and uniform R = 1.0) are
also plotted. Their results are used to establish expected values, as follows.
The strength in the models with the OC layer is limited by either:
(i) the strength of the harder OC layer, with soil local strength being
exhausted next to the pipe-soil interface (labelled location A in Figure 7.3);
or
(ii) the strength of the less overconsolidated soil just beyond the OC layer,
with soil local strength being exhausted at location B.
While the benchmark cases indicate a much lower soil strength for the normally
consolidated (NC) soil (i.e. that with R = 1.0) than for the OC soil (in this case with
R = 2.4), this difference is measured in terms of unit strength (qmax). The resistant area at
B is larger than that in A – proportionally to the radii ratio (0.5D + hoc) / (0.5D), or
equivalently (D + 2hoc) / D. The total resistance to pipeline axial movement is given by
the unit strength integrated over the resistant area. The limiting condition (i) or (ii) will
thus depend, from case to case, on the ratio between the unit strength of the two soil states
(the stronger soil within the OC layer and the weaker soil beyond that), as well as on the
value of hoc.
The maximum deviator stress qmax, measured at the pipe-soil interface could thus
be estimated by:
 D  2hoc 
qmax  min  qmaxA , qmaxB  ( 7-2 )
 D 
91
Where qmaxA and qmaxB are, respectively, the expected unit strength at locations A and B.
These can be taken from the benchmark cases, as being either the constant resistance
observed on the two most overconsolidated cases, or the curve-fit of Equation ( 7-1 ) for
the two less overconsolidated cases. Note that if qmax is limited by qmaxA, soil unit strength
would be exhausted at A, as described in (i), whereas if it is limited instead by qmaxB (with
the total resistance enhanced by the radii ratio), soil unit strength would be exhausted at
B, as described in (ii).
Expected strengths for the three values of hoc are plotted in Figure 7.7(a). It can
be observed that, for hoc/D = 0.1 and hoc/D = 0.2, the expected curves are simply the curve-
fit for the uniform NC case scaled by the radii ratio. This means that, for all those cases,
model strength is expected to be limited by the unit strength of the soil beyond the OC
layer (i.e. at B). It is just for hoc/D = 0.4 that this scaled curve-fit reaches qmaxB, thus being
capped by this limiting resistance (the continuation of the scaled curve-fit is illustrated in
dashed line). It is thus expected that analyses with vD/cv < 0.3 would present model
strength being limited by the soil strength within the OC layer (i.e. at A), while analyses
with vD/cv > 0.3 would present model strength being limited by the soil strength beyond
the OC layer (i.e. at B).
The results obtained from FE analyses with an OC layer deviate from the expected
values of qmax (based on the benchmark uniform soil results, using Equation ( 7-2 )),
presenting higher strength for vD/cv > 0.1 (or vD/cv > 1 for the particular case of
hoc/D = 0.4) with maximum differences observed for vD/cv = 3. This is due to excess pore
pressure redistribution between the two regions. The OC layer is dilatant, (potentially)
generating negative excess pore pressure while the NC region is contractile generating
positive excess pore pressure. A hydraulic gradient is formed causing water to flow from
the NC region into the OC region (eventually reaching the pipe-soil interface). This
drainage strengthens the NC soil, raising qmax.
Figure 7.7(b) present the values of umax observed at the pipe-soil interface. Note
that, despite the fact that the OC layer is generating no excess pore pressure (or some
negative excess pore pressure, as discussed next), significant levels of positive excess
pore pressure can be observed in many cases. For the cases with lower velocities, umax is
comparable with that for the uniform NC case, which means that the positive excess pore
pressure generated at the NC soil tends equalise into the OC layer. For higher velocities,
92
excess pore pressure would take longer to migrate, and thus equalisation was not
expected. Still, relevant levels of excess pore pressure can be observed for
0.1  vD/cv  10, indicating that the migration is still of importance.
For hoc/D = 0.4, the results for low velocities were expected to present qmax very
close to those obtained for the uniform OC cases. The small reduction observed in all
cases is due to the model’s volumetric constraint. While the strength in these cases is
governed by the OC layer, the NC soil beyond it still undergoes relevant levels of shear
stress. As such, yielding, with tendency of volume reduction, is observed. A consequence
of the volumetric constraint is that this tendency of volume reduction is translated in
reduction in total stress p (and consequently in effective stress p'). The reduction in p'
incurs in the observed reduction in qmax.
An interesting outcome of this rise in resistance in the NC soil for velocities at the
right-hand side of Figure 7.7, is observed for hoc/D = 0.4 and vD/cv = 1. Equation ( 7-2 )
would predict that, for this condition, soil strength would be exhausted at B. However,
detailed results presented in Figure 7.8 indicate otherwise. The stress paths in Figure
7.8(b) show that critical state is almost reached at B, but just before that, A yields, then
reaching CS as the peak q drops. This is highlighted in Figure 7.7(a) by the annotation
“CS at A” (as critical state is reached when the soil strength is exhausted, or just after that
in case of peak resistance). This is the case for all runs with hoc/D = 0.4 and vD/cv  1.
For all other cases, soil strength is exhausted at B as annotated.
93
1.5
hoc/D = 0.4 Uniform OC CS at A CS at B
qmax
p'0
hoc/D = 0.2 FE results

1.2 All reach CS at B
Eq. ( 7–2 )
hoc/D = 0.1
All reach CS at B
0.9
Uniform NC
Curve-fit
(a)
0.6
0.0001 0.001 0.01 0.1 1 10 100
vD/cv
(b)
0.4
Uniform NC
umax
p'0.3
0
hoc/D = 0.1
0.2
hoc/D = 0.2
0.1
hoc/D = 0.4
0.0
Uniform OC
-0.1
0.0001 0.001 0.01 0.1 1 10 100
vD/cv
Figure 7.7 – Maximum values, observed at the pipe-soil interface, of (a) deviator stress qmax (including
expected values calculated using Equation ( 7-2 ) and curve-fit from Figure 7.4) and (b) excess pore
pressure umax at the nodes at the pipe-soil interface as these moved with the prescribed constant velocities
v – analyses with initial R = 2.4 in overconsolidated layer and R = 1.0 elsewhere (including benchmarking
cases for these two values of R).
94
The following figures, Figure 7.8 to Figure 7.10, present detailed results for three
points from Figure 7.7, which were considered illustrative of the excess pore pressure
migration process. A few additional plots, for analyses with other values of R (thus not
captured in Figure 7.7), are enclosed in Appendix A. Stress–volume paths are presented,
for each case, at both locations A and B (with the results in B obtained for the soil state
outside of the OC layer, immediately after the discontinuity). Excess pore pressure u is
also presented at both locations. Table 7.3 summarises the analysis parameters for all the
presented detailed results figures (both herein and in Appendix A).
Table 7.3 – Cases selected for detailed results.
Figure presenting vD OC layer Initial R within Initial R

detailed results cv thickness hoc/D OC layer elsewhere
Figure 7.8 1 0.4 2.4 1.0
Figure 7.9 0.1 0.4 2.4 1.0
Figure 7.10 3 0.2 2.4 1.0
Figure A.1 0.1 0.1 2.0 1.0
Figure A.2 0.1 0.2 2.0 1.0
Figure A.3 0.3 0.4 2.0 1.0
Figure A.4 3 0.2 2.4 1.4
Figure A.5 1 0.1 2.0 1.4
As previously mentioned, Figure 7.8 shows that, for hoc/D = 0.4 and vD/cv = 1,
soil strength is exhausted (and subsequently critical state is reached) at A, while the
critical path for B gets very close to CS – albeit not reaching it. Other interesting aspects
are observed, the key one being the development of positive excess pore pressure at the
pipe-soil interface after an initially dilatant response – just as unexpectedly observed by
WHITE et al. [14] (and mentioned in their question (b) transcribed in Section 5.3). Three
distinct stages can be observed in the curve of u at A, in Figure 7.8(c):
 Initially, some small negative excess pore pressure develops. This is not
due to the dilatant nature of the highly overconsolidated soil in the OC
layer, since, as previously discussed, the implementation of MCC in use
would only induce negative excess pore pressure after yielding. Instead,
95
this is caused by the reduction in total stress induced by the confined
boundary condition. This negative excess pore pressure reaches a local
maximum around /D = 2.5%. At this point, the excess pore pressure
migrating – due to the large hydraulic gradient built up between B and A
– reverses the curve. The negative excess pore pressure then starts to
reduce (in absolute value).
 Around /D = 3.3%, soil at A yields, causing a sharp reduction in q and a

sudden generation of negative excess pore pressure – this time due to the
dilatant nature of the highly overconsolidated soil in the OC layer.
 Post peak strength, excess pore pressure migration continues, causing A to

present increasing levels of positive excess pore pressure. At the end of
the total prescribed displacement, excess pore pressure at A is almost the
same as observed at B.
Another noticeable feature in Figure 7.8(b) is the roundhouse shape curve

observed for q before its peak. Two aspects are believed to contribute to this reduction in
the rate in which the shear strength increases with pipe displacement – as opposed to the
constant rate (before peak) observed for uniform highly overconsolidated cases, as
illustrated in Figure 7.6:
 Firstly, as the NC soil yields, its shear stiffness reduces (with constant
increments of shear stress inducing larger increments of shear strain; or
vice versa, constant increments of shear strain inducing smaller increments
of shear stress). This reduction in stiffness causes a reduction in the ratio
of shear strain increments at A to increments of pipeline displacement.
 In addition, the reduction in total stress p (due to the model boundary

conditions) induces reduction in effective stress p' (slightly later, as seen
in Figure 7.8(b), as this is initially compensated by the negative excess
pore pressure).
After this roundhouse curve, yielding at A occurs at the “dry” side (i.e. with  > ), for
which the aforementioned reservations apply.
96
Finally, generation of positive excess pore pressure at B ceases as q drops post
peak. From this point, excess pore pressure at B start to decay as the soil reconsolidates.
This reconsolidation is clear from the last portion of the effective stress–volume path at
B shown in Figure 7.8(a), which follows a SRL (parallel to the two initial SRL depicted).
Interestingly, the excess pore pressure migration causes soil at A to swell. This swelling,
however, is not due to unloading – the shearing continues to mobilise all the available
resistance. It is cause by the increase in water content (and thus increase in e) as more
water migrates into this region. The result is a swelling along the critical state line, with
gradual reduction in available resistance – which is translated in the gradual decay of q
post peak, observed in Figure 7.8(c).
97
1.50
(a)
e SRL
CSL
B
1.45
p'/p'0
p/p'0
1.40
SRL
NCL
A
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0 A
CSL
p'/p'0
1.0
p p/p'0
0.5
B
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.5 0.5
(c)
q/p'
1.20 u/p'
0.4 0
q/p'0
0.9 0.3
B
A
0.6 0.2
u/p'0
0.3 0.1
0.0 0.0
-0.3 -0.1
0 2 4 6 8  /D (%) 10
Figure 7.8 – Results of the analyses with vD/cv = 1, for R = 2.4 within overconsolidated layer of
hoc/D = 0.4 and R = 1.0 elsewhere: (a) stress–volume paths (for both total stress p and effective stress p')
in the e–ln p / e–ln p' space, including the SRL corresponding to the initial conditions; (b) stress paths in
the p–q / p'–q space, including the initial projected yield functions; (c) q and u for axial displacement .
98
Figure 7.9 presents the results for a very similar condition albeit with velocity one
order of magnitude lower (i.e. hoc/D = 0.4 and vD/cv = 0.1). While most of the features
are similar, in this condition the positive excess pore is proportionally quicker (as, in fact,
the loading is slower). As a result, positive excess pore pressure is observed earlier at A.
Figure 7.9(a) shows that this early increase in water content causes forced swelling
along the SRL, with Figure 7.9(b) showing reduction in p' – with no (or very small)
reduction in p within the elastic range. Later, some forced swelling on the CSL is observed
just after peak q, up until the point where the excess pore pressure is approximately
equalised between A and B – as observed in Figure 7.9(c). As in the previous case,
generation of positive excess pore pressure at B ceases at peak q, after which the soil in
B reconsolidated along a SRL. After equalisation, A also starts to reconsolidate
(maintaining approximately equalised condition). However, this reconsolidation occurs
along the CSL. Again, this is due to the constant imposed velocity mobilising whatever
shear resistance is available (thus always pushing the stress state back to the CSL). As a
result of this post peak consolidation, q gradually rises again.
99
1.50
(a)
e SRL
CSL
B
1.45
p'/p'0 p/p'0
1.40
SRL
NCL
A
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0
A CSL
p'/p'0
1.0
p/p'0
p
0.5
B
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.5 0.5
(c)
q/p'
1.20 u/p'
0.4 0
q/p'0
0.9 0.3
A
0.6 0.2
B
0.3 0.1
0.0 0.0
u/p'0
-0.3 -0.1
0 2 4 6 8  /D (%) 10
Figure 7.9 – Results of the analyses with vD/cv = 0.1, for R = 2.4 within overconsolidated layer of
100
As opposed to the two previous cases, Figure 7.10 presents detailed results for a
case in which the soil at B reaches critical state before yielding at A, for hoc/D = 0.2 and
vD/cv = 3. Again, some initial negative excess pore pressure is observed in A – due to the
confining boundary conditions – reversing later to positive excess pore pressure as the
migration process develops. Again the increase in water content at A causes forced
swelling along the SRL. However, as opposed to the previous examples, the stress path
at A never reaches the yield surface.
The tendency for equalising excess pore pressure between A and B is observed,
nonetheless this equalisation is not reached. Numerical convergence issues were
experienced, which caused the analysis to stop just after /D = 6%. It is expected that,
after equalisation, soil at A would start to reconsolidate along the same SRL it swelled,
while soil at B would reconsolidate along the CSL, with some gradual increase in
strength. Similar convergence issues were experienced in all analyses with CS reached at
B. Figure A.1 in Appendix A shows the case in which the soil at B reconsolidated along
the CSL for longer (amongst all the analyses), still just for a very short period before the
analysis stoped.
Further work in improving this convergence could have been done if this feature
was observed before. However, it was only noticed late in the interpretation of the results.
Significant effort was put in stabilising the runs with peak resistance, which are
intrinsically numerically unstable, whereas the analyses with the roundhouse q curves,
typical of normally consolidated or slightly overconsolidated soils, deserved not much
attention as they all easily converged to what appeared to be sufficient. Still, while being
a relevant feature, this reconsolidation along the CSL is not directly connected to the main
objective of these analyses, which was to demonstrate the excess pore pressure migration
process. For that, the present results are considered to have presented enough evidence to
indicate that this process may occur.
101
1.50
(a)
e SRL
CSL
B
1.45
p'/p'0
p/p'0
1.40
SRL
NCL
A
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0 A
CSL
p'/p'0
1.0
p p/p'0
0.5
B
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.5 0.5
(c)
q/p'0
q/p'
1.20 u/p'
0.4 0
0.9 0.3
u/p'0 B A
0.6 0.2
0.3 0.1
0.0 0.0
-0.3 -0.1
0 2 4 6 8  /D (%) 10
Figure 7.10 – Results of the analyses with vD/cv = 3, for R = 2.4 within overconsolidated layer of
102
8. Novel Model
8.1. Background
As previously discussed, current industry practice – from early design stage

analytical calculations through to detailed three-dimensional (3D) finite element (FE)
analyses – is still to represent the axial soil resistance by a Coulomb-like friction. The
significant improvement in understanding the local soil behaviour over the last few years
has not yet been followed by a similar evolution in the available pipeline analysis tools.
As a consequence, recent projects have tried to translate complex time-dependent soil
behaviour into equivalent constant frictional resistances. This may incur misleading
results, as different sections of the same pipeline will see different displacements, under
different velocities, as the pipe expands [16].
Analyses of a short local length of pipeline employing detailed continuum models

using 3D solid elements have been published [67, 68]. The difficulty in extrapolating
them to a global pipeline analysis lies in the large scale difference between the mesh
required to model the continuum of underlying soil (beneath a pipe cross-section with
typical diameter of a fraction of a metre) and the pipeline slender bar behaviour (over
several kilometres). Hybrid 2D/3D approaches have been proposed for lateral soil
resistance, but either with a simplistic representation of the soil continuum [79] or with a
computational cost for complete global analyses that is still not compatible with typical
design schedules [70].
The first model to properly address time dependent aspects of axial pipe-soil
interaction whilst not depending on time consuming FE analyses was proposed by
RANDOLPH et al. [68]. In their model, analytical expressions describe the development
and dissipation of excess pore pressure at the pipe-soil interface. These are based on
extrapolation from the response of a finite thickness, infinite plane shear band at the
surface of a semi-space. They use critical-state framework supplemented by a damage
mechanism to reproduce late excess pore pressure generation observed in physical model
tests.
Two aspects of the soil behaviour that cannot be captured by the planar
idealisation have been discussed in sections 6 and 7, respectively:
103
(a) As the soil beneath a pipeline consolidates, pore water seeps not only
radially away from the pipe-soil interface, but also around it towards the
draining seabed surface. Furthermore, the radial component of the flow,
rather than monotonically diminishing as pore pressure equalises, may
reverse as a result of this 2D process.
(b) As the soil hardens next to the pipe, the shear band could migrate,
depending on the trade-off between gain in unit strength and increased
resistant area as it moves away from the pipe (as suggested by WHITE &
CATHIE [8]). In partially drained conditions, excess pore pressure
generated away from the pipe may drain towards it, eventually reducing
the effective normal stress at the pipe-soil interface.
A plane strain model, similar to those used by GOURVENEC et al. [49] and
KROST et al. [57], was used in Section 6 for discussion (a). It cannot capture axial
displacements as all degrees of freedom are limited to the plane – (see Figure 8.1). In
Section 7, for discussion (b), an axisymmetric model, with a mesh of 2D solid elements
in the plane –x, was employed. Multi-point constraints were used to force nodes of same
coordinate  but different x to have identical responses in all degrees of freedom. As such,
the model represents boundary conditions around a 2D slice perpendicular to an infinite
pipeline with axial displacements permitted. This same artifice was previously used in
the detailed continuum models using 3D solid elements by RANDOLPH et al. [68] and
YAN et al. [67].
104
Pipeline

 or z
Figure 8.1 – Cylindrical coordinate system (, , x) and vertical direction z.
The aim of the present model is to realistically capture the soil response to pipeline
axial movement, while maintaining low computational costs to permit its use in global
pipeline analyses. The model was conceived to bridge between the two models in [68]:
the detailed soil continuum model, which captures in details the complexity of the pipe-
soil interaction but is impractical for design purposes; and the elegant planar model,
which provides essentially instantaneous results but does not capture the two-
dimensionality of the drainage behaviour. In particular, the new model was intended to
capture the two aspects of behaviour raised in sections 6 and 7 and noted above.
As derived hereafter, the model is based on a 1D mesh of bespoke finite elements,

which is equivalent to the axisymmetric model used in Section 7. This uses a 2D mesh of
axisymmetric elements in the plane –x, having the dimension x eliminated using multi-
point constraints. It is then supplemented by an approximate analytical solution for the
circumferential drainage of the excess pore pressure (along ), thus reproducing the 2D
consolidation process observed in the plane strain set-up continuum FE model used in
Section 6.
A succinct comparison between these four models (two by RANDOLPH et al.

[68], and two used in sections 6 and 7) is presented in Table 8.1. Within it, the low
computation cost is obtained by minimising the “soil mesh dimensions”, while the
realistic response is achieved by maximising the other two parameters.
105
Table 8.1 – Comparison between existing models and proposed one.
Soil mesh Pipeline degrees Pore water flow

Model Reference
dimensions of freedom directions
Detailed 3(1) 2 2 RANDOLPH

continuum (,  and x) (z and x) ( and ) et al. [68]
1 1(2) RANDOLPH
Planar 0
(x) (z) et al. [68]
2 1 2
Plane strain Section 6
( and ) (z) ( and )
2(1) 1 1
Axisymmetric Section 7
( and x) (x) ()
1 2 2(3) Present
Proposed model
() (z and x) ( and ) Section 8
Notes:
(1) Dimension x is eliminated by the use of multi-point constraints (refer to Section 7.2 for details).
(2) Analytical approximation assuming parabolic isochrones.
(3) FE solution for the radial () direction combined with analytical approximation for circumferential
() direction.
The model was written using symbolic programming in MATHCAD [80], which
provided the flexibility required for development and sufficient accuracy for proof of
concept. Its implementation as a user-programmed element into a commercial FE package
should be straightforward, and would provide the model with the numerical robustness
and solving optimisation required of a reliable design tool.
Section 8.2 describes the 1D bespoke element, relevant aspects of its

implementation and its validation against the results of an axisymmetric, elastoplastic,
partially drained analysis from Section 7. As discussed, the models are equivalent and
thus the validation exercise is solely to ensure that the symbolic programming
implementation is adequate. Section 8.3 presents the modification to account for
circumferential drainage, in which the analytic approximation for excess pore pressure
dissipation along  is superposed on the 1D bespoke element. A brief validation exercise
compares the results of the implemented circumferential drainage against the analytical
solution for 1D consolidation. Section 8.4 then presents the application of the
implemented model to assess partially embedded pipelines, thus combining the radial and
106
circumferential drainage directions. The first example reproduces the plane strain set-up
analysis results in Section 6. Finally, the implemented model is used to reproduce the
complete analysis of cyclic axial displacements intervened by consolidation periods,
performed by YAN et al. [67] using detailed continuum 3D FE analyses.
8.2. 3D Model Reduced to 1D FE Mesh
The proposed model is based on coupled consolidation FE analysis, which is

presented in detail in Section 3.3. As presented therein, it consists of solving the system
of equations:
K K Tv   ξ   f 
      t g    g  K u  ( 3-23 bis )
K v   t K u   u    u 
where K, Kv and Ku are, respectively, the stiffness, coupling and transmissivity matrices
and the unknowns  and u are the nodal displacement and pore pressure increment
vectors. The vector u contains the nodal values of excess pore pressure at the beginning
of the integration step. The increment of external nodal forces over the time step t is f;
while g is the vector of flux of pore water away from the model domain, and g its
increment.
Equation ( 3-12 ) expresses the relationships between increments of strain  and

displacement  in the soil skeleton, in Cartesian coordinates. Adopting instead a
cylindrical coordinate system (, , x), these relationships become:
   1  x

     x  
 x  
  1     x
     x   ( 8-1 )
   x 
 x  1  
 x     
x   
By assuming axial symmetry around x, conveniently coincident with the pipeline

axis (as depicted in Figure 8.1) and assuming that the pipeline is straight and infinite,
Equation ( 8-1 ) is reduced to:
107
   x
      x  
 
( 8-2 )

      x    x     0

in which the normal strain increment components (and consequently the volumetric strain
increment) depend on the radial displacement increment only, while the shear strain
depends on the axial displacement increment only. Equation ( 3-23 ) can be rewritten as:
 Kx K Tpl 0   ξ x    fx 
    
K pl K K Tv   ξ      f  ( 8-3 )

 0 Kv   t K u    u   t  g    g  K u u  
   
If the soil is within the elastic range, Kpl = 0. In the plastic range however, the
volumetric and distortional strains are interrelated by the associated flow rule, as
discussed in Section 3.4, and this matrix is non-nil.
Applying the FE interpolation functions as discussed in Section 3.3 to the strain–

displacement relationships in Equation ( 8-2 ), the strain increment components and
volumetric strain increment can be obtained by:
 0 H  
    
    0 H 
    ξ xe   
   B  B   ( 8-4 )
  x   ξ e   0 0 
  x  
H  0 
and
 H
 v  Bv ξ  Bv    H    ( 8-5 )
e
 
where H is the interpolation function and H its derivative with respect to  – with the
axial symmetry imposing H = Hx = 0.
A solution for the system of equations in Equation ( 8-3 ), with the MCC
constitutive model as presented in Section 3.4, was implemented using symbolic
108
programming in MATHCAD [80]. The usual polynomial interpolation with isoparametric
mapping and Gaussian quadrature rule for numerical integration were employed (e.g. [81,
82]). Comparisons between first order (two-node) elements with a single integration point
and second order (three-node) with three integration points (for both displacements and
pore pressure) were performed in the validation exercises hereafter. Results indicate no
benefit in adopting higher order elements, however a thorough comparison including
mesh density sensitivity was not pursued further. One of the potential benefits of a second
order element would be the 40% reduction in the minimum time step required to prevent
spurious pore pressure oscillation along the model length [76]. Minimum time step
restrictions are of relevance for partially drained analyses close to the undrained
condition. A more elegant and efficient solution however would be to implement a
numerical stabilisation technique (e.g. [83]).
T
For the first order element, with nodal coordinates ρe    e1 e  , the
2
interpolation matrix is:
1  r 1 r 
H r    ( 8-6 )
 2 2 
where r is the local natural coordinate of the isoparametric element; and its derivative –
which is constant over each element, thus independent of r – is given by:
1 1 1
H    ( 8-7 )
J  2 2 
where the (scalar) Jacobian is:
e  e
J  ρe   2 1
( 8-8 )
2
T
For the second order element, with nodal coordinates ρe    e1 e
2
 e  , the
3
interpolation matrix is:
 r2  r r2  r 
H r    1 r2  ( 8-9 )
 2 2 
109
and its derivative:
1 1 1
H   r   r  2r r  ( 8-10 )
J  2 2
with:
 1 1
J  r , ρe    r   2r r    ρe ( 8-11 )
 2 2
The bespoke 1D finite element just described – in its two (first and second order)
versions – is equivalent to the axisymmetric model in Section 7, which was modelled
using a 2D mesh of axisymmetric elements, with the axial dimension eliminated with the
use of multi-point constraints. The only difference is that, instead of using a 2D mesh and
then eliminating one of the dimension, the bespoke element is implemented already into
a 1D mesh, in which each node is described by its only relevant coordinate .
To ensure the symbolic implementation in MATHCAD is correct (and accurate

enough), one analysis case from Section 7 was reproduced using the implemented model.
The case reproduced was the one with uniform normally consolidated soil and pipe
velocity of vD/cv = 0.1, for which detailed results have already been presented in Figure
7.5. A uniform case was selected for convenience, as the model as implemented cannot
handle the step change in soil state adopted in models with overconsolidated layer in
Section 7. It is able, though, to handle smooth variations of soil state, either as an initial
condition or as a result of changes in soil state as the analysis progresses.
Both ends of the model (the pipe-soil interface and the far end of the model,
sufficiently far to approximate well the infinite soil mass) were considered impervious.
As such, g and g are zero (i.e. no flux of pore water away from the model domain – only
redistribution within it). A change in boundary conditions was, however, adopted, in
which the total radial force was maintained constant, so that the force increment vectors
fx and f are zero (as no change in load is applied).
Figure 8.2 compares the results obtained from the implementation using both first
and second order elements and both linear and backward interpolation with those obtained
using the ABAQUS model from Section 7 (re-run with the aforementioned change in
110
boundary condition). The agreement was considered satisfactory, with some of the small
differences observed attributed to the simplicity of the symbolic programming
implementation (concerning convergence tolerances and round-off handling) compared
with the commercial software. For example, a continuous, single iteration march was
adopted, in which the matrices K, Kv and Ku are updated at the beginning of every time
step (but not within the time step as would occur in a Newton-Raphson scheme for
example).
As previously mentioned, the results show very small differences between the
implementations using first order elements (with one integration point) and second order
elements (with three integration points). However, the drainage was found to be slower
when using linear interpolation (i.e.  = 0.5), with the implemented model in this case
presenting higher levels of excess pore pressure. Better agreement with the ABAQUS
results was obtained using backward interpolation (i.e.  = 1), which is expected as this
is the interpolation scheme adopted by ABAQUS [29]. These trials show that the
symbolic implementation of the bespoke 1D element, and in particular that using
backward interpolation, is valid.
111
1.50
SRL NCL
e
1.48
 0.5 1.0
1.46 1st order
2nd order
Abaqus
(a) CSL
1.44
0.6 0.8 1.0 p'/p'0
1.2
q/p'0 CSL
0.8
0.4
(b)
0.0
0.6 0.7 0.8 0.9 p'/p'0 1.0
0.9 0.45
(c)
q/p'0  = 0.5 u/p'0
0.6 0.30
1st order
q/p'0
=1
2nd order ABAQUS
0.3 0.15
u/p'0
0.0 0.00
0.0 0.5 1.0 1.5  /D (%) 2.0
Figure 8.2 – Results of the implemented 1D bespoke element (first order and second order elements;
linear and backward interpolation) compared to ABAQUS results using the model of Section 7:
(a) volume–stress paths in the e–p' (normalised by the initial mean effective stress p'0) space; (b) stress
paths in the q–p' space; (c) q and u versus axial displacement  normalised by the pipe diameter D.
112
8.3. Circumferential Drainage
The previous section presented the formulation of coupled consolidation analysis

of an axisymmetric, infinite model, as illustrated in Figure 8.3(a), in which the 3D domain
is reduced to a 1D FE mesh. While the infinite length approximation is acceptable, the
axial symmetry assumption is not applicable to the majority of subsea pipelines, which
are laid onto the seabed and left partially embedded. Assuming a semi-space, as in Figure
8.3(b), is more appropriate, but would require a 2D FE mesh, leading to a much higher
computational cost.
(a) (b)
Water
Pipeline u=0
 
Nodes
 
Infinite soil domain Semi-infinite soil domain
Figure 8.3 – Cross-section of (a) axisymmetric and (b) modified FE models.
The proposed model assumes that the 1D bespoke element, as formulated in the
previous section, can well approximate the stresses and strains along a vertical line
beneath the pipe invert, as long as the circumferential component of pore water drainage
is artificially accounted for. This is done by means of an approximate analytical solution,
assuming that the distribution of excess pore pressure along the circumferential (hoop)
direction (for any constant ) can be described by a semi-period sinusoidal curve, with
maximum value at the bottom invert and nil at the seabed surface either side.
The simplifications of axial symmetry and infinite length imply that any quantity
will present a constant value along any point in space with same coordinate . The values
at each node are hence representative of a cylindrical locus, of circular cross section
113
concentric with the pipeline, as illustrated by dashed lines in Figure 8.3(a). For the
partially embedded pipe geometry, quantities (such as stresses, strains, displacements and
pore pressure) will vary along the interrupted loci – dashed arcs in Figure 8.3(b).
The proposed model assumes thus that, at any particular time, the excess pore
pressure at any point in the soil domain u(,) is given by:
  
u (  ,  )  ui (  )  cos   ( 8-12 )
 2 
where ui() is the excess pore pressure along the vertical line beneath the pipe invert; and
the cylindrical loci intercept the seabed surface at ', as indicated in Figure 8.4. This
assumption is supported by the results of the plane strain analysis in Section 6, in
particular by the isochrones (curves of excess pore pressure distribution at particular
instants) along arcs of constant  shown in Figure 6.5.
'
* 

Figure 8.4 – Angles θ' and θ* for any given arc of constant ρ in the modified FE model.
If u * is the average excess pore pressure over an arc of constant ρ in the interval
[–*, *], where * is a small angle from the vertical, as indicated in Figure 8.4, then:
*
0
u d  u *  * ( 8-13 )
Integrating the approximate expression of Equation ( 8-12 ) gives:
114
*
ui  2
 ( 8-14 )
u*  *  
sin  
  2 
If the pipe-soil interface is impervious, the flux of pore water seeping away from
the infinite soil wedge in the interval [–*, *] through its surface area Γ is:
 
v n

 d   2  v * d 

( 8-15 )
where the circumferential component of flow velocity at * is given by:
k u s  2 k u * *
v *    ( 8-16 )
w     *
4  w   2
Equation ( 8-15 ) can be rewritten using the coupled consolidation FE notation,

presented in Section 3.3. It gives hence the vector of nodal flux g at the boundary of the
limited domain – the infinite soil wedge in the interval [–*, *]:
  2 k u * *
g   H T v  n d    HT d   Q u ( 8-17 )

2   w   2
where:
2 k *
Q  H H d
T
( 8-18 )
2  w   2
The continuity equation, Equation ( 3-21 ) – which is also the last equation of the
system in Equation ( 3-23 ) –, can be rewritten incorporating Equation ( 8-17 ) as:
K v  ξ    t K u  u   f   t  Q u   Q u  K u u  ( 8-19 )
and rearranging to keep the unknown u in one term only:
K v ξ    t  K u  Q   u   f   t  K u  Q  u ( 8-20 )
115
Note that Equation ( 8-20 ) is identical to the original continuity equation – in the
form of last equation of the system in Equation ( 3-23 ) – with g = g = 0 – which is one
of the assumptions adopted in the bespoke 1D element – if Ku is replaced by:
K *u  K u  Q ( 8-21 )
One very convenient outcome from choosing Equation ( 8-12 ) to describe the
circumferential distribution of excess pore pressure is that the resulting matrix Q is
directly proportional to * – see Equation ( 8-18 ). The same applies for each of the other
terms of Equation ( 3-23 ), following the integration of K, Kv and Ku, if a single
integration point is considered along  (which is consistent with the selection of a small
*). This makes the problem independent of the selected value of *.
By adopting an infinitesimal *  0, all calculated values correspond to the actual

value along the vertical line (or plane) beneath the pipe invert, irrespective of how each
of the quantities vary away from it (over the rest of the soil semi-space). The average
excess pore pressure u * thus tends to ui (and within the model will be called simply u).
Moreover, if the plane –x for  = 0 is a plane of symmetry, the general –
relationships in Equation ( 8-1 ) can be reduced to Equation ( 8-2 ) – and hence Equation
( 3-23 ) can be reduced to Equation ( 8-3 ) – except for the elimination of the term in
∂δξθ/∂θ from δεθ. For simplicity, so that the same constitutive equations previously
derived can be used in this new model, ∂δξθ/∂θ is arbitrarily assumed zero at  = 0.
Again, to ensure the symbolic implementation in MATHCAD is correct, a first

validation exercise compares the result for circumferential drainage only with the
analytical solution of TERZAGHI’S one-dimensional consolidation. A single element
model, representing a very thin ring arc, with large mean radius  and arbitrary angle '
was run. A mean radius to thickness ratio of 1000 and an arbitrary ' = 1 radian were
adopted – although the dimensionless results are invariant with '. By making both nodes
impervious, excess pore pressure can only flow around the  direction (i.e. with
dissipation only through Q). A constant total radial force was maintained by restraining
only one of the nodes in the  direction. Linear elastic soil behaviour was obtained by
simply setting p'c to a high value (see Section 3.4).
116
The result of this simple model coincides with the analytical solution of
TERZAGHI’S 1D consolidation for initial distribution of excess pore pressure following
the same cosine distribution adopted by the modified FE model. For this initial condition,
the degree of consolidation U is given by Equation ( 6-4 ) with dimensionless time T,
considering the drainage path length of the ring model (which is   ), given by:
cv t
T ( 8-22 )
   
2
As expected, Figure 8.5 shows that the thin ring arc results obtained with the
implemented model match closely those from the analytical expression of the degree of
consolidation for initial distribution of excess pore pressure following the same initial
cosine distribution.
0.5
Implemented model
TERZAGHI 1D – Eq. ( 6–4 )
1
0.01 0.1 1 T
Figure 8.5 – Thin ring arc results obtained with the implemented model and analytical expression of the
degree of consolidation for initial distribution of excess pore pressure following a cosine distribution.
117
8.4. Application to Partially Embedded Pipelines
Fundamental validation exercises, comparing the implemented model with

equivalent results, were presented at the end of each of the previous sections 8.2 and 8.3.
In the present section, the previously implemented 1D bespoke element – modified to
include the matrix Q into K u – is then used in trying to reproduce FE analyses of
embedded pipelines. Note that the results presented herein were directly obtained from
the implemented model, with no curve fitting or arbitrary adjustment of additional
parameters.
Firstly, the implemented model was used to reproduce the results of the elastic
set-up analysis in Section 6, in which the soil beneath a partially embedded (“wished-in-
place”) pipeline is allowed to consolidate under a constant vertical load, after this load is
instantaneously applied to the pipe. The set-up analysis assumes plane strain condition,
which is equivalent to the implemented model with zero axial displacement at every node.
As in the last validation exercise, linear elastic soil response was obtained by setting p'c
to a high value.
Figure 8.6 presents calculated average excess pore pressures along arcs of constant
, normalised by u1,1 – the first index indicating “1st node”, and the second “1st time step”,
thus the initial average value (just after the load application) at the pipe-soil interface (i.e.
/D = 0.5). The abscissa cvt/D2 in Figure 8.6 considers a constant coefficient of
consolidation, calculated with the initial value of the coefficient of volume change mv,
obtained from Equation ( 3-34 ) using the initial values of p' and e. Note that the ABAQUS
results’ curve for /D = 0.5 (i.e. at the pipe-soil interface) is equivalent to the degree of
consolidation previously presented in Figure 6.6.
Only the results of the first order, one integration point, backward interpolation
implementation are shown. The difference to those using higher order with more
integration points or different time integration schemes are not noticeable at the scale of
the figure. Results indicate very good agreement, with the implemented model presenting
slightly faster excess pore pressure dissipation.
118
1.0
__ /D = 0.5
u/u1,1
0.8
/D = 0.6
/D = 0.7
0.6
/D = 1.0
0.4
ABAQUS
0.2
/D = 2.0
Implemented model
0.0
0.001 0.01 0.1 1 10
cvt/D2
Figure 8.6 – Results of average excess pore pressure along arcs of constant  over time t, from the
implemented model (including the modification for considering circumferential drainage) and from the
ABAQUS model in Section 6.
In particular for the arcs with larger radius (i.e. /D = 1.0 and /D = 2.0 in Figure
8.6), the curves obtained from ABAQUS present a longer horizontal length followed by
a slight rise just after the two sets of results diverge. This rise may be related to the
MANDEL-CRYER effect [74, 75], which was discussed in Section 6.1. This stress
redistribution phenomenon is likely to be influenced by the assumption that ∂δξθ/∂θ could
be arbitrarily taken as zero for  = 0. Another aspect also discussed in Section 6.1 – and
highlighted in the results of GOURVENEC et al. [49] reproduced in Figure 6.1 – is that
the rise in excess pore pressure due to the MANDEL-CRYER effect is much more evident
in results for a point beneath the pipe invert, being smoothed out when average excess
pore pressure along the arc is plotted. This implies that the isochrones deviate from the
cosine distribution (i.e. taking a more protruding shape, as the values next to the invert
rise more than the average).
119
Figure 8.7 presents curves of excess pore pressure along a vertical line beneath
the pipe invert at particular instants, which is similar to those presented by JEWELL &
BALLARD [59] (reproduced in Figure 5.3) and in Section 6 (Figure 6.4). The faster
dissipation in the implemented model, as previously observed in Figure 8.6, is noticeable.
The longer horizontal lengths in Figure 8.6 translate in all curves but the last, from the
ABAQUS results, being superposed for larger values of , as opposed to the earlier
consolidation observed in results from the implemented model. Still, the reversion in
gradient, which causes pore water to flow towards the pipe-soil interface later in the set-
up process (as highlighted in Section 6) can be clearly observed.
0.0 0.2 0.4 0.6 0.8 u/u1,1 1.0

0
cvt/D2 = 10 cvt/D2 = 0.1 cvt/D2 = 0.01 cvt/D2 = 0.001
cvt/D2 = 1
ABAQUS
/D Implemented model
Figure 8.7 – Isochrones along a vertical line beneath the pipe invert, obtained from the implemented
model (including the modification for considering circumferential drainage) and from the ABAQUS
model in Section 6.
120
Despite the small differences discussed, these results were considered satisfactory
and were obtained at a far smaller computational cost than the 2D analyses against which
they were benchmarked. This confirms that the hybrid numerical / analytical approach
proposed is adequately accurate whilst being more practical for assessing the axial PSI of
partially embedded pipelines in full-length pipeline structural models.
Finally, the implemented model – the 1D bespoke element including the

modification for circumferential drainage – was tested by trying to reproduce the detailed
continuum 3D FE analysis of pipeline cyclic axial movements intervened by
consolidation periods presented by YAN et al. [67]. Only the first order, one integration
point implementation with backward integration scheme was used. The total soil
resistance to pipeline axial movement was obtained by integrating the interface shear
stress x along the pipe-soil interface. The interface shear stress was assumed to vary
with cos . This is consistent with the assumption adopted by WHITE & RANDOLPH
[48] for quantifying the wedging effect due to embedment, that the normal stress at the
pipe wall follows this same cosine distribution. Note that, unlike the adopted excess pore
pressure distribution, the interface shear stress (or the normal stress in [48]) does not
necessarily reduce to zero at the seabed surface (only for the particular case where
' = /2). The maximum value of x is simply the shear stress calculated (by the
implemented model) in each time step at the first node,   x . The total soil resistance F
1
is then given by:
1
F     x1 cos  1 d    x1 D sin 1 ( 8-23 )
1
A key difference of the cyclic loading process compared to the analysis reported
by YAN et al. [67] is that each sweep was interrupted and the axial load released
immediately after the critical state was reached at the pipe-soil interface, rather than after
a prescribed distance. As observed in Figure 8.8(a), the axial resistance profile obtained
using the proposed model reached a maximum value sharply, not presenting the more
progressive loss of tangent stiffness observed in the results in [67]. Figure 8.9(a)
reproduces the plot by YAN et al. [67], overlaying the results obtained with the
implemented model, making this difference evident. A fully drained sweep is also
included, presenting a consistent difference in curve shape.
121
The more progressive loss in stiffness is likely due to the soil around the pipe
gradually reaching critical state – i.e. first at the pipe invert (most loaded point) and
gradually spreading on either side until the whole interface is in critical state. This cannot
be captured in the proposed model, as it represents the entire interface by a single node
(hence under a single representative condition, either at the critical state or not).
The sudden loss in stiffness after reaching critical state, combined with the limited
numerical robustness in the simple symbolic programming implementation, has caused
the model to present significant numerical convergence issues for further axial
displacements past this point. This was the reason for interrupting the sweeps. The major
effect, observed in Figure 8.8(b) and Figure 8.9(b), is the earlier dissipation of excess
pore pressure (as excess pore pressure generation ceases, as opposed to the continuum
model in which dissipation is contrasted with sustained generation of excess pore
pressure).
The excess pore pressure ratio ru expressed in these figures is the ratio of the
average excess pore pressure along the pipe-soil interface u to the average (total) normal
stress n , thus:
u
ru  ( 8-24 )
n
Assuming that, as does x, 'n varies with cos , the average normal effective
stress is then:
sin 1
 n   ( 8-25 )
1 1
The assumption that the excess pore pressure follows the cosine distribution of
Equation ( 8-12 ) makes u = 2u/ , for which ru becomes:
u u1
ru  
u   n u   sin 1   ( 8-26 )
1
2 1
1
122
0.6
(a)
F/W Sweep 13
0.4
Sweep 1  undrained resistance
0.2
Unloading immediately after peak resistance
0
0.0000001 0.00001 0.001 cvt/D2 0.1
0.5
(b)
ru
Sweep 1
0.4
0.3
0.2
Sweep 13
0.1
0
0.0000001 0.00001 0.001 0.1
cvt/D2
Figure 8.8 – Cyclic axial displacements results: (a) axial resistance F normalised by pipeline submerged
unit weight W; and (b) excess pore pressure ratio ru. Time t restarted at the beginning of every new sweep
for clarity.
123
Figure 8.9 – Results from Figure 8.8 (plus axial resistance for one fully drained sweep) overlaid (in light-
green) on corresponding figures reproduced from YAN et al. [67] (background figures courtesy ICE
Publishing).
124
The progressive loss of stiffness in the resistance curve obtained from the 3D FE
analysis by YAN et al. [67] is matched by the ru response, as regions of soil reaching
critical state will cease to generate excess pore pressure. After that, for continuous
displacement, the 3D FE results show a sustained level of ru up until the end of the sweep,
as observed in their plot reproduced in Figure 8.9(b). This is due to the reconsolidation
along the CSL as partial drainage occurs, similarly to what was observed in the results in
Section 7. This reconsolidation along the CSL can be observed in their stress volume
path, previously reproduced in Figure 4.10 (segments D–E and I–J for the end of sweeps
1 and 2, respectively). In the present results, the sharp peak in resistance is accompanied
by a sharp peak in ru.
Despite the different curve shapes, the maximum ru observed in the first sweeps
match quite well with those observed in [67]. This can be observed in Figure 8.9(b), and
more clearly in Figure 8.10. However, as the dissipation starts immediately after this peak
(i.e. no period of sustained ru), full dissipation is always observed earlier in the
implemented model. The implemented presents a cvt50/D (i.e. the dimensionless time at
which ru has reduced to 50% of the maximum ru observed at the point of maximum
resistance) of approximately 0.0005, as observed in Figure 8.8(b). This is an order of
magnitude quicker than the cvt50/D = 0.005 observed by YAN et al. [67] (and can be
observed in Figure 8.9(b)).
It is worth highlighting, however, that this result observed in [67] is itself an order
of magnitude quicker than that for the set-up of the same 3D models after the application
of a vertical load. According to CHATTERJEE et al. [72] (and their table reproduced in
Table 6.1), the vertical load set-up presented cvt50/D = 0.05 (for the applicable w/D = 0.3),
indicating that the dissipation of excess pore pressure induced by axial pipeline
displacement is an order of magnitude quicker than the dissipation of excess pore pressure
induced by vertical load. This is of particular relevance for pipeline design, as the results
in [67] and [72] suggest that the relative gain in strength due to set-up after a pipeline
moves axially is an order of magnitude quicker than for example the gain in strength post
pipe lay. However, the model implemented herein appears to overestimate this rate of set-
up, so should be treated with caution.
125
Figure 8.10 indicates that the rate at which the maximum ru reduces over cycles is
slower in the implemented model (in comparison with the results in [67]). Figure 8.11
plots, for every cycle, the “drainage index” proposed by YAN et al. [67]:
( F / W )  ( F / W )undrained
 ( 8-27 )
( F / W )drained  ( F / W )undrained
along with results of the expression for  curve-fitted by YAN et al. [67]:
 
 n 1 
 ( n )  1  exp  0.7  ( 8-28 )
  0.9  
  
  /   
where n is the sweep number. It also present another curve, geometrically derived from
the theoretical critical state volume–stress path by WHITE et al. [84]:
    n1 
11  
    
S 1 ( 8-29 )
(n) 
S 1
where S is the ratio of average total normal stress n to average effective normal stress
n in the first sweep:
   1 
S  n    ( 8-30 )
  n 1st sweep  1  ru 1st sweep
Results show that the gain in resistance due to consolidation hardening is slightly
slower in the present model (which is consistent with the slower rate at which the
maximum ru decreases over cycles observed in Figure 8.10). However, it appears to tend
to unity at about the same number of cycles predicted by the expressions proposed by
YAN et al. [67] and WHITE et al. [84].
126
0.5
ru
0.4
0.3
Implemented model
0.2
0.1
3D FE results by YAN et al.

0.0
0 5 10 15 20
Sweep number
Figure 8.10 – Maximum excess pore pressure ratio ru observed in each axial sweep in the implemented
model, and in the 3D FE analysis by YAN et al. [67].
1.0
WHITE et al.

YAN et al.
0.5
Implemented model
0.0
0 5 10 15
Sweep number 20
Figure 8.11 – Results of drainage index  over cycles, compared to the results of the expression proposed
by YAN et al. [67] – Equation ( 8-28 ) – and WHITE et al. [84] – Equation ( 8-29 ).
127
The number of cycles was limited to 13 due solely to numerical convergence
issues experienced beyond that. However, inappropriate results are observed after sweep
7, in which the consolidation progresses beyond p'/p'0 = 1, as observed in the volume–
stress paths in Figure 8.12. This is probably an error accumulation problem, consequence
of the single iteration scheme, and adopting a Newton-Raphson scheme (thus balancing
the internal and external forces within every integration step) would prevent this anomaly.
It also seems to be further prejudiced by the limitation in minimum time step (as observed
by VERMEER & VERRUIJT [76]). As aforementioned, adopting a numerical
stabilisation scheme (such as that proposed by WHITE & BORJA [83]) would permit
using smaller time steps, probably improving the obtained results. Such an error
accumulation could also contribute to the divergence in the cycle by cycle results of
excess pore pressure ratio ru and drainage index .
The obtained results were still considered very satisfactory as a proof of concept
for the proposed model, justifying its implementation in a more robust platform. A more
thorough validation would benefit from it, thus permitting isolation of numerical issues
(such as error accumulation) from the actual limitations of the assumptions made in the
simplified model.
128
2.35
SRL NCL (a)
e
Sweep 1
2.30
Intervening
consolidation
periods
2.25
Sweep 13
CSL
2.20
0.6 0.8 1.0 p'/p'0 1.2
1.2
Sweep 1 Sweep 13
q/p'0
Analysis starting point CSL
End of each consolidation
period
Sweep
0.8
Unloading
0.4
Consolidation
(b)
0.0
0.6 0.7 0.8 0.9 1.0 1.1 p'/p'0 1.2
Figure 8.12 – Results from the implemented model: (a) volume–stress path in the e–p' space; (b) stress
paths in the q–p' space.
129
9. Final Remarks
9.1. Conclusions
The foremost objective of this thesis was to develop a model for axial pipe-soil
interaction capable of capturing key aspects of the soil response, while with modest
computational cost, so that it could be used in global pipeline design analysis. As
previously discussed, the understanding of the geotechnical mechanisms involved in this
interaction have significantly evolved over the last few years. The existing tools for global
design analysis of subsea pipelines, however, have not followed this evolution yet. Owing
to the large difference in scale between the local soil response and the global pipeline
behaviour, existing local models of the soil response cannot be straightforwardly
extended to global pipeline analyses. The PSI model proposed herein is seen as having
the potential to bridge this gap.
The proposed axial PSI model was developed to a proof of concept level, being
implemented using symbolic programming in MATHCAD [80], then validated and tested
against results of FE analysis of different complexities. Yet, to achieve the ultimate
objective of having a design tool, implementation within a complete FE package is still
required. A number of improvements which are considered relevant in this future
implementation have been mentioned, and are summarised in Section 9.2.
Some particular aspects of the soil response were focused through the
development of the axial PSI model, namely:
 Two-dimensional consolidation
 Partially drained response
 Cyclic soil plasticity
The current state of the knowledge in axial pipe-soil interaction was reviewed in Section
4. Studies in Sections 6 and 7 present contributions to this knowledge, in particular to the
understanding of mechanisms which influence the particular aspects just highlighted.
130
Section 6 scrutinised the results of elastic set-up analyses of partially embedded
pipelines. The analyses are essentially identical to those in [49, 57, 59]. However, further
interpretation of the results provided useful insight. GOURVENEC & WHITE [49] and
KROST et al. [57] focused on dissipation curves for the excess pore pressure at the pipe
invert or average around the pipe-soil interface, while JEWELL & BALLARD [59]
presented isochrones along a vertical line beneath the pipe invert. Section 6 shows that
both results are interrelated in a two-dimensional consolidation process.
Isochrones taken around the circumferential direction were observed to present a

cosine shape – which proved convenient when developing the proposed PSI model. The
curve of degree of consolidation over time for one-dimensional consolidation with initial
cosine-shaped distribution of excess pore pressure was compared with the dissipation
curves for the average excess pore pressure around the pipe-soil interface obtained in the
elastic set-up FE model. The quicker initial consolidation followed by sustained excess
pore pressure at late stages observed in the FE results is consistent with the hydraulic
gradient reversion observed in the vertical isochrones. Initial outward drainage enhances
the rate of dissipation at the pipe-soil interface in early stages; then, reversed drainage
inwards sustains the excess pore pressure (at this interface) for longer.
These results suggest that the two-dimensional consolidation can be broken down
into two orthogonal directions of pore water flow:
 Circumferentially, draining to the seabed surface; and
 Radially migrating through the seabed soil mass.
Each of these two interacting components present particular features. The circumferential
drainage takes place maintaining approximately constant cosine-shaped isochrones. The
radial migration only redistributes excess pore pressure within the soil mass (as the soil
is assumed an infinite semi-space and the pipe-soil interface assumed impervious).
This breakdown in two distinct components was convenient in the development

of the proposed model, which consists of a coupled consolidation FE model for the radial
direction (plus the axial direction, not relevant for the plane strain set-up analyses in
Section 6) combined with an analytical solution for the circumferential drainage.
131
Section 7 ignored the circumferential drainage, using axisymmetric models to
focus on the partially drained response to axial pipe displacement. The models also
considered soil plasticity, adopting modified Cam clay constitutive model in particular.
Cyclic loading was not applied to the model. However, the effect of previous loading in
changing the soil state was considered, by adopting an idealised overconsolidated (or
more overconsolidated) layer adjacent to the pipe-soil interface. The idealised step change
in soil state, while not realistic, was considered more effective for the understanding of
some features of the soil response than a (more realistic) smooth transition model.
The analyses showed an interesting pore pressure migration mechanism, whereby

excess pore pressure generated in one region of the soil migrates, thus influencing the
response and strength of another soil region. In particular, positive excess pore pressure
generated at the less overconsolidated soil farther away from the pipe-soil interface may
migrate towards this interface, thus reducing the soil resistance to pipeline axial
movement. Of course, this effect is only relevant if this migration occurs within a period
of time comparable to the loading (pipe axial sweep) period. This depends on:
 Pipeline velocity
 Soil permeability
 Hydraulic gradient formed when pipe moves
 Drainage path length (i.e. distance between the region where the excess
pore pressure is prominent and the pipe-soil interface)
In particular for the last bullet point, having a well-defined boundary between soil regions
permitted clearer distinction than more realistic models would.
The effect of the pore pressure migration on the soil resistance to pipeline axial
movement was observed for a number of cases (under the particular conditions
considered, typically for the velocity range of approximately 0.1  vD/cv  10). This
effect had different consequences, at different conditions, such as:
 Change in maximum soil resistance (either increase or reduction, however

an increase was observed in most cases)
132
 Variable post-peak resistance (rather than being constant, resistance
gradually changes as pipe moves, either increasing or reducing)
 Reduced stiffness, with longer distances required to mobilise maximum

resistance
These aspects are neglected when the response is framed as drained or undrained.
Moreover, the common assumption of undrained sweeps intervened by full consolidation
periods neglects the effects of any residual excess pore pressure distribution that can exist
when the pipeline moves again. For real pipeline design, full consolidation between
sweeps may often be an unrealistic assumption. The interaction of partially drained
pipeline displacements with existing residual excess pore pressures may further
complicate the soil response.
Another interesting aspect observed on the stress–volume paths obtained (which

was unexpected to the candidate, although slightly noticeable also on the stress–volume
paths by YAN et al. [67]) was the swelling and reconsolidation along the critical state
line. The continuous pipeline movement is considered a key cause of this behaviour. By
continually mobilising all the available shear resistance (as the pipe moves) while the soil
begin to swell or reconsolidate, the stress–volume path is continuously “pushed back”
onto the critical state line.
Section 8 presented the conception, implementation and validation of the novel

model proposed for axial pipe-soil interaction. It extrapolates from the axisymmetric,
infinite length geometry from the models in Section 7, in which the 3D continuum
problem is reduced to a 1D FE mesh. The second (circumferential) dimension is
accounted for by means of artificial drainage governed by an analytical approximation –
following the breakdown suggested by the results in Section 6. Details of the formulation
and validation of each aspect of the model are presented. The computational cost saving
achieved by reducing the two-dimensional problem into a one-dimensional mesh is
enormous, which makes the proposed model potentially viable for use in global design
analyses of subsea pipelines.
The analytical approximation for circumferential drainage adopted semi-period

sinusoidal isochrones, as the isochrones observed in the set-up models in Section 6
presented approximately this shape (as observed in Figure 6.5). This resulted in very
133
convenient expressions, permitting accounting for the circumferential drainage only by
adopting a small modification in the transmissivity matrix.
Despite some small difference (attributed to a MANDEL-CRYER-like stress

redistribution), the proposed model presented very good agreement with the results of the
detailed local FE analyses of 2D set-up in Section 6. Comparison with the complete 3D
FE analysis of cyclic axial displacements by YAN et al. [67] presented very promising
results, with the very good agreement in first cycles, however followed by some small
deviation as cycles accumulated.
As discussed, implementation using a robust platform is still required (e.g. as a

user-programmed element within an FE package) for use in design practice. This will also
permit assessing how much of the observed deviation over cycles is due to model
limitations (as opposed to numerical error accumulation). Still, the presented results are
considered very satisfactory as proof of concept. The proposed model is considered a
promising design tool, with a very good balance between complexity and efficiency, thus
bridging the gap between the local modelling of the geotechnical processes and the global
analyses required in the practice of subsea pipeline design.
9.2. Suggestions for Future Research
As discussed, the proposed model was only developed to a proof of concept level,
implemented using symbolic programming. The obvious first suggestion is thus to
implement it using a robust platform. For use in design practice, the most effective option
would be to implement it into a FE package already used in global design analysis of
pipelines, for example as a user-programmed element in ABAQUS [29].
A number of issues faced during the proof of concept implementation are raised,
which should be addressed in particular before (or during) this new implementation. The
key ones are:
 To include a proper Newton-Raphson algorithm:
The results in Section 8.4 (in particular for the cyclic pipe displacement)
seem sensitive to accumulated numerical error. As such, the simple
134
continuous, single iteration march adopted (while acceptable for proof of
concept), is inadequate in a robust implementation.
 To include numerical stabilisation:
As any other coupled consolidation application, analyses close to

undrained condition present a very high volumetric stiffness, which
attracts numerical convergence issues. Efficient ways around this issue
have been implemented in other applications, one example being the
stabilised elements proposed by WHITE & BORJA [83] for assessing fault
zone transients. A robust implementation would greatly benefit from such
an artifice.
 Stabilisation post peak:
As any other non-linear analysis presenting peak strength, static stability

post peak was a difficulty faced in the models presenting peak soil
resistance. While this requires attention, it might not be an issue after
implementation in a commercial FE package (as the existing stabilisation
tools in the package would be available).
Other aspects not related to numerical convergence, but instead to the realism of
the soil response, are also raised for further investigation:
 Yielding with  > :
As discussed in Section 7.2, the ABAQUS [29] implementation of

modified Cam clay constitutive model uses the MCC yielding surface
irrespective of whether the yielding happens on the “wet” or “dry” side,
even though MCC has been originally proposed only for “wet” soils [26,
27]. This same assumption was repeated in the proposed model. Another
yield surface (e.g. the HVORSLEV surface as discussed by ATKINSON
& BRANSBY [30]) may be more suitable.
135
 Yielding beneath the state boundary surface:
The elastic soil response before yielding may conceal relevant aspects of
the cyclic response, in particular the soil volume reduction accumulated
for (reversing) cyclic shear. For an analogous application, sliding
foundations for pipeline end structures, DEEKS et al. discuss the
“shakedown” observed in physical model tests, suggesting the use of a
“bubble” model such as that proposed by STALLEBRASS & TAYLOR
[85] (which is similar to that previously proposed by AL-TABBAA &
MUIR WOOD [86]). Such a feature is likely to be required for reproducing
physical model test observations.
 Viscous effects:
Viscous effects can be of relevance, in particular for clayey soils typically

observed in deep-water. Viscous resistance of the pore water, and in
particular of that adsorbed around the clay particles, causes the soil local
shear strength to vary with the rate of shear strain. On the other hand, this
viscosity varies with temperature. As the pipeline temperature varies, so
will that of the soil beneath it, thus introducing an additional complex
interaction. These effects have been totally ignored in the development of
the present model. Its constitutive model could be enhanced to account for
viscous effects, for example following the behaviour model developed by
the COPPE/UFRJ Soil Rheology Group [87–91]. While the effects on the
FE formulation in terms of mechanical response would be limited,
accounting for the temperature dependency would require a completely
new model.
 Interface sliding:
The models in Section 7, as well as the analyses using the implemented

model in Section 8, considered no interface sliding, with all the pipe
displacement translated into soil distortion. In reality, interface sliding
may occur. Means for accounting for this feature, and to realistically
model this sliding, should be pursued.
136
Lastly, the following aspects would likely benefit from further investigations (e.g.
using the 3D slice model proposed by RANDOLPH et al. [68]):
 Large difference in consolidation time:
As discussed in Section 8.4, a difference of an order of magnitude is

observed between the set-up time after the application of a vertical load,
and after a pipeline axial displacement. Further understanding of the
reasons for this discrepancy is of relevance. This result is potentially of
great benefit for pipeline design, so it is prudent to better understand the
reasons behind it before blindly adopting their outcome in design practice.
After that, the extra order of magnitude difference between the 3D FE
model and the proposed model (as discussed in Section 8.4) should also
be clarified.
 Effect of incomplete drainage:
The effect of residual excess pore pressure distribution, as a result of

incomplete consolidation between pipeline axial sweeps, may affect the
soil response. Further analyses can help understand to what extent this is
of relevance.
 Excess pore pressure migration:
The excess pore pressure migration process discussed in Section 7 was

only studied with the idealised conditions of axial symmetry and step
change in soil state. Following the observations herein, this study could be
extended to 3D, as well as to consider more realistic geometric
distributions of the soil’s initial state. Complex large deformation analyses
such as those presented by CHATTERJEE et al. [58] could be used for
obtaining realistic initial conditions.
 Swelling and reconsolidation along the CSL:
This feature deserves further investigation, in particular for the conditions

not well captured in the models in Section 7, and combined with the last
bullet point, accounting for the 3D geometry.
137
 Velocity normalisation:
Finally, it does not seem too complicated to clarify the polemic point (as
mentioned in Section 4.3.2, discussed by WHITE & CATHIE [8]
following concern previously raised by PALMER [60]) of whether the soil
compressibility is relevant for normalising the velocity of shear loading
(i.e. whether permeability or coefficient of consolidation should be used).
138
10. References
[1] PALMER, A. C., KING, R. A., Subsea Pipeline Engineering, 1st ed. Tulsa:
Pennwell Books, 2004.
[2] RANDOLPH, M. F., GOURVENEC, S. M., Offshore Geotechnical Engineering,

1st ed. Milton Park, Oxon: Spoon Press, 2011.
[3] STINGL K. H., PAARDEKAM, A. H. M., “Parque das Conchas (BC-10) - An

Ultra-Deepwater Heavy Oil Development Offshore Brazil,” In: Proceedings of
the Offshore Technology Conference, 2010.
[4] BREDERO SHAW, “Offshore Products,” 2012. [Online]. Available:

http://www.brederoshaw.com/solutions/offshore/offshore_products.html.
[Accessed: 25-Jan-2015].
[5] JAMES WALKER DEVOL, “Photo Gallery.” [Online]. Available:

http://www.devol.co.uk/about/gallery.asp. [Accessed: 25-Jan-2015].
[6] ANTHROPOLOGY IN THE WIND, “An Industry Perspective,” 2013. [Online].

Available: https://anthropologyinthewind.wordpress.com/2013/06/30/an-
industry-perspective/. [Accessed: 25-Jan-2015].
[7] TECHNIP, “Subsea,” 2014. [Online]. Available: http://www.technip.com/en/our-

business/subsea. [Accessed: 25-Jan-2015].
[8] WHITE D. J., CATHIE, D. N., “Geotechnics for Subsea Pipelines,” in

Proceedings of the International Symposium on Frontiers in Offshore
Geotechnics II, 2011, pp. 87–123.
[9] TØRNES, K., OSE, B. A., JURY, J., THOMSON, P., “Axial Creeping of High
Temperature Flowlines Caused by Soil Ratcheting,” In: Proceedings of the
ETCE/OMAE joint conference, Energy for the new millennium, v. 2, pp. 1229–
1240, 2000.
[10] DRISKILL, ,“Review of Expansion and Thermal Growth Problems in Subsea

Pipelines,” In: Offshore Oil & Gas Pipeline Technology: Proceedings of a
European Seminar, 1981.
[11] CARR, M., BRUTON, D. A. S., LESLIE, D., “Lateral Buckling and Pipeline
Walking, a Challenge for Hot Pipelines,” In: Proceedings of the Offshore
Pipeline Technology Conference, 2003.
[12] CARR, M., SINCLAIR, F., BRUTON, D. A. S., “Pipeline Walking –

Understanding the Field Layout Challenges, and Analytical Solutions Developed
for the SAFEBUCK JIP,” In: Proceedings of the Offshore Technology
Conference, 2006.
[13] BRUTON, D. A. S., SINCLAIR, F., CARR, M., “Lessons Learned from
Observing Walking of Pipelines with Lateral Buckles, Including New Driving
139
Mechanisms and Updated Analysis Models,” In: Proceedings of the Offshore
Technology Conference, 2010.
[14] WHITE, D. J., GANESAN, S. A., BOLTON, M. D., BRUTON, D. A. S.,

BALLARD, J.-C., LANGFORD, T., “SAFEBUCK JIP - Observations of Axial
Pipe-Soil Interaction from Testing on Soft Natural Clays,” In: Proceedings of the
Offshore Technology Conference, 2011.
[15] CARNEIRO, D., WHITE, D. J., DANZIGER, F. A. B., ELLWANGER, G. B.,

“Excess Pore Pressure Redistribution beneath Pipelines: FEA Investigation and
Effects on Axial Pipe-Soil Interaction,” In: Proceedings of the 3rd International
Symposium on Frontiers in Offshore Geotechnics, 2015 (in press).
[16] CARNEIRO, D., RATHBONE, A. D., SOON, K. S., VIECELLI, G., “Rate
Dependent Soil Resistance in FE Analysis of Pipeline Walking,” In: Proceedings
of the ASME 2014 33rd International Conference on Ocean, Offshore and Arctic
Engineering, 2014.
[17] CARNEIRO, D., WHITE, D. J., DANZIGER, F. A. B., ELLWANGER, G. B.,

“A Novel Approach for Time-Dependent Axial Soil Resistance in the Analysis of
Subsea Pipelines,” Comput. Geotech., 2015 (submitted).
[18] SPARKS, C. P., “The Influence of Tension, Pressure and Weight on Pipe and
Riser Deformations and Stresses,” J. Energy Resour. Technol., v. 106, n. 1, p. 46,
1984.
[19] ELLINAS, C. P., SUPPLE, W. J., VASTENHOLT, H., “Prevention of Upheaval

Buckling of Hot Submarine Pipelines by Means of Intermittent Rock Dumping,”
In: Proceedings of the Offshore Technology Conference, 1990.
[20] LOEKEN, P. A., “The ‘CREEP’ on the Ekofisk - EMDEN 36" Gas Pipeline,” In:
Proceedings of the Offshore Technology Conference, 1980.
[21] KONUK, I., “Expansion of Pipelines Under Cyclic Operational Conditions:

Formulation of Problem and Development of Solution Algorithm,” In:
Proceedings of the Seventeenth International Conference on Offshore Mechanics
and Arctic Engineering (OMAE 1998), 1998.
[22] TERZAGHI, K., “Principles of Soil Mechanics: IV-Settlement and Consolidation

of Clay,” Eng. News Rec., v. 95, n. 22, pp. 874–878, 1925.
[23] BIOT, M. A., “General Theory of Three-Dimensional Consolidation,” J. Appl.

Phys., v. 12, n. 2, pp. 155–164, 1941.
[24] BENTLER, D. J., “Finite Element Analysis of Deep Excavations,” PhD Thesis,
Virginia Polytechnic Institute and State University, 1998.
[25] BOOKER J. R., SMALL, J. C., “An Investigation of the Stability of Numerical
Solutions of Biot’s Equations of Consolidation,” Int. J. Solids Struct., v. 11, n. 7–
8, pp. 907–917, Jul. 1975.
140
[26] BURLAND, J. B., “The Yielding and Dilation of Clay (correspondence),”
Géotechnique, v. 15, n. 2, pp. 211–214, Jan. 1965.
[27] ROSCOE, K. H., BURLAND, J. B., “On the Generalised Stress– Strain
Behaviour of ‘Wet’ Clay,” in Engineering Plasticity, J. Heyman and F. A.
Leckie, Eds. Cambridge, UK: Cambridge University Press, pp. 535–609, 1968.
[28] ROSCOE, K. H., SCHOFIELD, A. N., WROTH, C. P., “On the Yielding of
Soils,” Géotechnique, v. 8, n. 1, pp. 22–53, Jan. 1958.
[29] DASSAULT SYSTÈMES, “Abaqus 6.10 Analysis User’s Manual.” Dassault

Systèmes Simulia Corp., Providence, RI, USA, 2010.
[30] ATKINSON, J. H., BRANSBY, P. L., The Mechanics of Soils - An Introduction

to Critical State Soil Mechanics, 1st ed. Maidenhead, Berkshire, England:
McGraw-Hill, 1977.
[31] ROSCOE, K. H., SCHOFIELD, A. N., “Mechanical Behaviour of an Idealised

‘Wet Clay,’” In: Proceedings of the 2nd European Conference on Soil Mechanics
and Foundation Engineering, v. 1, pp. 47–54, 1963.
[32] SCHOFIELD, A. N., WROTH, C. P., Critical State Soil Mechanics. London:
McGraw-Hill, 1968.
[33] ZIENKIEWICZ, O. C., NAYLOR, D. J., “The Adaptation of Critical State Soil
Mechanics Theory for Use in Finite Elements,” in Stress-Strain Behaviour of
Soils - Proceedings of the Roscoe Memorial Symposium, Parry, R. H. G., Ed. G.
T. Foulis & Co Ltd, 1972, pp. 537–547.
[34] ISO, “ISO 13623: Petroleum and natural gas industries — Pipeline transportation
systems,” 2009.
[35] EN, “EN 14161: Petroleum and natural gas industries — Pipeline transportation
systems (ISO 13623:2009 modified),” 2011.
[36] DNV, “DNV-OS-F101: Submarine Pipeline Systems,” 2013.
[37] API, “API Recommended Practice 1111: Design, Construction, Operation, and
Maintenance of Offshore Hydrocarbon Pipelines (Limit State Design),” 2009.
[38] BSI, “BS 8010: Code of practice for Pipelines — Part 3: Pipelines subsea:
design, construction and installation,” 1993.
[39] ASME, “ASME B31.4 Pipeline Transportation Systems for Liquids and
Slurries,” 2012.
[40] ALA, “Guidelines for the Design of Buried Steel Pipe,” 2001.
[41] DNV, “DNV-RP-F105: Free Spanning Pipelines,” 2006.
141
[42] DNV, “DNV-RP-F110: Global Buckling of Submarine Pipelines Structural
Design due to High Temperature / High Pressure,” 2007.
[43] CARR, M., BRUTON, D. A. S., “SAFEBUCK JIP - Safe Design of Pipelines
with Lateral Buckling - Design Guideline,” 2004.
[44] SCHAMINEE, P. E. L., ZORN, N. F., SCHOTMAN, G. J. M., “Soil Response

for Pipeline Upheaval Buckling Analyses: Full-Scale Laboratory Tests and
Modelling,” In: Proceedings of the Offshore Technology Conference, pp. 563–
572, 1990.
[45] FINCH, M.”, “Upheaval Buckling and Floatation of Rigid Pipelines : The
Influence of Recent Geotechnical Research on the Current State of the Art.,” In:
[46] NAJJAR, S. S., GILBERT, R. B., HALL, C., MCCARRON, B., “Tilt Table Test
for Interface Shear Resistance Between Flowlines And Soils,” In: Proceedings of
the 22nd International Conference on Offshore Mechanics and Arctic
Engineering, 2003.
[47] FINCH, M., FISHER, R., PALMER, A. C., BAUMGARD, A., “An Integrated
Approach to Pipeline Burial in the 21st Century,” In: Proceedings of the Deep
Offshore Technology 2000, 2000.
[48] WHITE, D. J., RANDOLPH, M. F., “Seabed Characterisation and Models for
Pipeline-soil Interaction,” Int. J. Offshore Polar Eng., v. 17, n. 3, pp. 193–204,
2007.
[49] GOURVENEC, S. M., WHITE, D. J., “Elastic Solutions for Consolidation

Around Seabed Pipelines,” In: Proceedings of the Offshore Technology
Conference, 2010.
[50] BRENNODDEN, H., STOKKELAND, A., “Time-Dependent Pipe-Soil

Resistance for Soft Clay,” in Proceedings of the Offshore Technology
Conference, 1992.
[51] BRUTON, D. A. S., WHITE, D. J., LANGFORD, T., HILL, A. J., “Techniques
for the Assessment of Pipe-Soil Interaction Forces for Future Deepwater
Developments,” In: Proceedings of the Offshore Technology Conference, 2009.
[52] HILL, A. J., JACOB, H., “In-Situ Measurement of Pipe-Soil Interaction in Deep
Water,” In: Proceedings of the Offshore Technology Conference, 2008.
[53] CATHIE, D. N., JAECK, C., BALLARD, J.-C., WINTGENS, J.-F., “Pipeline
Geotechnics – State-of-the-Art,” in Proceedings of the International Symposium
on Frontiers in Offshore Geotechnics, pp. 95–114, 2005.
[54] API, “API Recommended Practice 2A-WSD: Recommended Practice for

Planning , Designing and Constructing Fixed OffshorePlatforms — Working
Stress Design,” 2010.
142
[55] BRUTON, D. A. S., BOLTON, M. D., CARR, M., WHITE, D. J., “Pipe-Soil
Interaction With Flowlines During Lateral Buckling and Pipeline Walking - The
SAFEBUCK JIP,” In: Proceedings of the Offshore Technology Conference,
2008.
[56] PALMER, A. C., LING, M. T. S., “Movements Of Submarine Pipelines Close To

Platforms,” In: Proceedings of the Offshore Technology Conference, 1981.
[57] KROST, K., WHITE, D. J., GOURVENEC, S. M., “Consolidation Around

Partially Embedded Seabed Pipelines,” Géotechnique, v. 61, n. 2, pp. 167–173,
Feb. 2011.
[58] CHATTERJEE, S., WHITE, D. J., YAN, Y., RANDOLPH, M. F., “Elastoplastic
Consolidation beneath Shallowly Embedded Offshore Pipelines,” Géotechnique
Lett., v. 2, n. 2, pp. 73–79, Apr. 2012.
[59] JEWELL, R. A., BALLARD, J.-C., “Axial Pipe-Soil Interaction: a Suggested

Framework,” In: Proceedings of the Offshore Technology Conference, 2011.
[60] PALMER, A. C., “Speed Effects in Cutting and Ploughing,” Géotechnique, v. 49,
n. 3, pp. 285–294, Jan. 1999.
[61] HILL, A. J., WHITE, D. J., BRUTON, D. A. S., LANGFORD, T., MEYER, V.,
JEWELL, R. A., BALLARD, J.-C., “A New Framework for Axial Pipe-Soil
Resistance, Illustrated by a Range of Marine Clay Datasets,” In: Proceedings of
the International Conference on Offshore Site Investigation and Geotechnics,
2012.
[62] WHITE, D. J., CAMPBELL, M. E., BOYLAN, N. P., BRANSBY, M. F., “A

New Framework for Axial Pipe-Soil Interaction, Illustrated by Shear Box Tests
on Carbonate Soils,” In: Proceedings of the International Conference on
Offshore Site Investigation and Geotechnics, 2012.
[63] BOYLAN, N. P., WHITE, D. J., BRUNNING, P., “Seabed Friction On

Carbonate Soils: Physical Modelling of Axial Pipe-Soil Friction,” In:
[64] PALMER, A. C. “Geotechnical Evidence of Ice Scour as a Guide to Pipeline

Burial Depth,” Can. Geotech. J., v. 34, n. 6, pp. 1002–1003, Dec. 1997.
[65] WHITE, D. J., HODDER, M. S., “A Simple Model for the Effect on Soil
Strength of Episodes of Remoulding and Reconsolidation,” Can. Geotech. J., v.
47, n. 7, pp. 821–826, Jul. 2010.
[66] DEEKS, A., ZHOU, H., KRISDANI, H., BRANSBY, M. F., WATSON, P.,
“Design of Direct On-Seabed Sliding Foundations,” In: Proceedings of the ASME
2014 33rd International Conference on Ocean, Offshore and Arctic Engineering,
2014.
143
[67] YAN, Y., WHITE, D. J., RANDOLPH, M. F., “Cyclic Consolidation and Axial
Friction for Seabed Pipelines,” Géotechnique Lett., v. 4, n. July – September, pp.
165–169, Jul. 2014.
[68] RANDOLPH, M. F., WHITE, D. J., YAN, Y., “Modelling the Axial Soil
Resistance on Deep-Water Pipelines,” Géotechnique, v. 62, n. 9, pp. 837–846,
Sep. 2012.
[69] CARNEIRO, D., GOUVEIA, J., PARRILHA, R., CARDOSO, C. de O., “Buckle
Initiation and Walking Mitigation for HP / HT Pipelines,” in Proceedings of the
Deep Offshore Technology International Conference, 2009.
[70] MARTIN, C. M., KONG, D., BYRNE, B. W., “3D Analysis of Transverse Pipe–
Soil Interaction Using 2D Soil Slices,” Géotechnique Lett., v. 3, n. July -
September, pp. 119–123, Sep. 2013.
[71] WARD, W. H., PENMAN, A., GIBSON, R. E., “Stability of a Bank on a Thin
Peat Layer,” Géotechnique, v. 5, n. 2, pp. 154–163, Jan. 1955.
[72] CHATTERJEE, S., WHITE, D. J., RANDOLPH, M. F., “Coupled Consolidation

Analysis of Pipe–Soil Interactions,” Can. Geotech. J., v. 50, n. 6, pp. 609–619,
Jun. 2013.
[73] SCHIFFMAN, R. L., CHEN, A. T.-F., JORDAN, J. C., “An Analysis of

Consolidation Theories,” J. Soil Mech. Found. Div., v. 95, n. 1, pp. 285–309,
1969.
[74] MANDEL, J., “Étude Mathématique de la Consolidation des Sols,” In: Acts du
Colloque International de Mécanique, vol. 4, pp. 9–19, 1950 (in French).
[75] CRYER, C. W., “A Comparison of the Three-Dimensional Consolidation

Theories of Biot and Terzaghi,” Q. J. Mech. Appl. Math., v. 16, n. 4, pp. 401–
412, 1963.
[76] VERMEER, P. A., VERRUIJT, A., “An Accuracy Condition for Consolidation
by Finite Elements,” Int. J. Numer. Anal. Methods Geomech., v. 5, n. 1, pp. 1–14,
Jan. 1981.
[77] TAYLOR, D. W., Fundamentals of Soil Mechanics. New York: Wiley, 1948.
[78] RANDOLPH, M. F., CARTER, J. P., WROTH, C. P., “Driven piles in clay—the
effects of installation and subsequent consolidation,” Géotechnique, v. 29, n. 4,
pp. 361–393, Jan. 1979.
[79] GRIFFITHS, T. J., SHEN, W., “Further Development and Benchmarking of a

Novel Pipe-Soil Interaction Model for Subsea Pipeline Design,” In: Proceedings
of the ASME 2013 32nd International Conference on Ocean, Offshore and Arctic
Engineering, 2013.
144
[80] PARAMETRIC TECHNOLOGY CORPORATION, “User’s Guide - Mathcad
15.0.” Parametric Technology Corporation, Needham, MA, USA, 2011.
[81] VAZ, L. E., Método dos Elementos Finitos em Análise de Estruturas. Rio de
Janeiro, Brazil: Elsevier, 2011 (in Portuguese).
[82] ZIENKIEWICZ, O. C., TAYLOR, R. L., The Finite Element Method for Solid
and Structural Mechanics, 6th ed. Oxford: Elsevier Ltd, 2005.
[83] WHITE, J. A., BORJA, R. I., “Stabilized Low-Order Finite Elements for
Coupled Solid-Deformation/Fluid-Diffusion and their Application to Fault Zone
Transients,” Comput. Methods Appl. Mech. Eng., v. 197, n. 49–50, pp. 4353–
4366, Sep. 2008.
[84] WHITE, D. J., LECKIE, S. H. F., DRAPER, S., ZAKARIAN, E., “Temporal
Changes in Pipeline-Seabed Condition, and their Effect on Operating
Behaviour,” In: Proceedings of the ASME 2015 34th International Conference on
Ocean, Offshore and Arctic Engineering, 2015 (in press).
[85] STALLEBRASS, S. E., TAYLOR, R. N., “The Development and Evaluation of a

Constitutive Model for the Prediction of Ground Movements in Overconsolidated
Clay,” Géotechnique, v. 47, n. 2, pp. 235–253, Apr. 1997.
[86] AL-TABBAA, A., MUIR WOOD, D., “Experimentally Based ‘Bubble’ Model
for Clay,” In: Proceedings of the 3rd International Symposium on Numerical
Models in Geomechanics, no. 6, pp. 90–99, 1989.
[87] MARTINS, I. S. M., “Fundamentos de um Modelo de Comportamento de Solos

Argilosos Saturados,” DSc Thesis, Federal University of Rio de Janeiro, 1992 (in
Portuguese).
[88] ALEXANDRE, G. F., “Contribuição ao Entendimento da Fluência Não-

Drenada,” DSc Thesis, Federal University of Rio de Janeiro, 2006 (in
Portuguese).
[89] SANTA MARIA, P. E. L., MARTINS, I. S. M., SANTA MARIA, F. C. M.,

“Rheological Behaviour of Soft Clays,” in Proceedings of the International
Symposium on Frontiers in Offshore Geotechnics II, 2011, pp. 335–340.
[90] AGUIAR, V. N., “Contribuição ao Estudo das Relações Tensão-Deformação-

Resistência-Tempo das Argilas Moles,” DSc Thesis, Federal University of Rio de
Janeiro, 2014 (in Portuguese).
[91] ANDRADE, M. do E. S., “O Adensamento Unidimensional Considerando a

Resistência Viscosa à Compressão,” DSc Thesis, Federal University of Rio de
Janeiro, 2014 (in Portuguese).
145
Appendix A.
Additional Results from the Axisymmetric Model of Section 7
The figures enclosed in this appendix complement those presented in Section 7.4,
as per Table 7.3 reproduced below:
Table 7.3(bis) – Cases selected for detailed results.
Figure presenting vD OC layer Initial R within Initial R

detailed results cv thickness hoc/D OC layer elsewhere
Figure 7.8 1 0.4 2.4 1.0
Figure 7.9 0.1 0.4 2.4 1.0
Figure 7.10 3 0.2 2.4 1.0
Figure A.1 0.1 0.1 2.0 1.0
Figure A.2 0.1 0.2 2.0 1.0
Figure A.3 0.3 0.4 2.0 1.0
Figure A.4 3 0.2 2.4 1.4
Figure A.5 1 0.1 2.0 1.4
Relevant aspects of each figure are:
Figure A.1: Early excess pore pressure migration, with approximate equalisation
achieved immediately, and early forced swelling at A. Critical state is reached at B,
followed by reconsolidation at both A and B. Reconsolidation at A is along the same SRL,
while reconsolidation at B is along the CSL.
Figure A.2: Again, early excess pore pressure migration, with approximate
equalisation achieved immediately, and early forced swelling at A. Strength however is
exhausted at A first, with yielding and sudden drop in q. Stress path at B very close to the
CSL. Post peak q, both soils reconsolidate, B along a SRL and A along the CSL, which
146
is translated in gradual increase in q. Initial forced swelling along CSL at A (as observed
in Figure 7.8 and Figure 7.9) not present, with reconsolidation starting immediately after
critical state is reached.
Figure A.3: Yield at A at the “dry” side, however very close to the CSL, resulting
in a very modest peak in q. Reduction in strength post peak appears to be more due to the
increase in water content and forced swelling along the CSL than due to sudden softening
typical of “dry” side response. Gradual reduction in q as soil at A swells along the CSL,
up until excess pore pressure equalisation. After that, reconsolidation (A along CSL and
B along SRL) with gradual increase in q – although very small as total displacement is
soon achieved.
Figure A.4: Critical state appears to have been reached at B, followed by yielding
at A as excess pore pressure migration develops – hence with soil strength exhausted at
B, then at A (as strength reduces at A). After that, B reconsolidates, with stress path at B
moving away from the CSL along a SRL, while A swells along the CSL.
Figure A.5: Again stress path at B very close to CSL, but A yields slightly earlier.
Stress–volume path at A shows early forced swelling along initial SRL, forced swelling
along the CSL immediately post peak q, then reconsolidation along CSL.
147
1.50
(a) B
e SRL
1.45
p'/p'0 p/p'0
SRL
1.40
A
CSL NCL
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0 CSL
A
1.0
p p'/p'0
p/p'0
0.5
B
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.5 0.5
(c)
q/p'
1.20 u/p'
0.4 0
q/p'0
0.9 0.3
u/p'0
0.6 0.2
A
B
0.3 0.1
0.0 0.0
-0.3 -0.1
0 2 4 6 8  /D (%) 10
Figure A.1 – Results of the analyses with vD/cv = 0.1, for R = 2.0 within overconsolidated layer of
148
1.50
(a) B
e SRL
p'/p'0
1.45
p/p'0
SRL
1.40
A
CSL NCL
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0 A CSL
1.0
p p'/p'0
p/p'0
0.5
B
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.5 0.5
(c)
q/p'
1.20 u/p'
0.4 0
q/p'0
0.9 0.3
0.6 0.2
A
B
0.3 0.1
u/p'0
0.0 0.0
-0.3 -0.1
0 2 4 6 8  /D (%) 10
149
1.50
(a)
e SRL
B
1.45
p'/p'0 p/p'0
SRL
1.40
A
CSL NCL
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0 A CSL
p'/p'0
1.0
p p/p'0
0.5
B
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.5 0.5
(c)
q/p'
1.20 u/p'
0.4 0
0.9 0.3
q/p'0
A
0.6 0.2
B
0.3 0.1
u/p'0
0.0 0.0
-0.3 -0.1
0 2 4 6 8  /D (%) 10
150
1.50
(a)
e CSL NCL
1.45
SRL
B
p'/p'0 p/p'0
1.40
SRL
A
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b)
q/p'0 A
CSL
p'/p'0
1.0
p/p'0
p
B
0.5
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.5 0.5
(c)
q/p'
1.20 u/p'
0.4 0
q/p'0
0.9 0.3
0.6 0.2
B A
0.3 0.1
u/p'0
0.0 0.0
-0.3 -0.1
0 2 4 6 8  /D (%) 10
Figure A.4 – Results of the analyses with vD/cv = 3, for R = 2.4 within overconsolidated layer of
151
1.50
(a)
e
1.45
B
SRL
p'/p'0 p/p'0
1.40
A
CSL NCL
1.35
0.3 1 p/p'0 , p'/p'0 3
1.5
(b) A
q/p'0 CSL
p'/p'0
1.0
p/p'0
p
B
0.5
0.0
0.0 0.5 1.0 p/p'0 , p'/p'0 1.5
1.5 0.5
(c)
q/p'0
q/p'
1.20 u/p'
0.4 0
0.9 0.3
0.6
A 0.2
B
u/p'0
0.3 0.1
0.0 0.0
-0.3 -0.1
0 2 4 6 8  /D (%) 10
Figure A.5 – Results of the analyses with vD/cv = 1, for R = 2.0 within overconsolidated layer of
152

Carneiro DL TM 15 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Carneiro DL TM 15 PDF

Uploaded by

Copyright:

Available Formats

A TIME-DEPENDENT AXIAL PIPE-SOIL INTERACTION MODEL FOR GLOBAL

ANALYSES OF SUBSEA PIPELINES

Daniel Lyrio Carneiro

Tese de Doutorado apresentada ao Programa de

Orientadores: Gilberto Bruno Ellwanger

Daniel Lyrio Carneiro

TESE SUBMETIDA AO CORPO DOCENTE DO INSTITUTO ALBERTO LUIZ

RIO DE JANEIRO, RJ - BRASIL

I would like to thank my advisors Professor Gilberto Ellwanger and Professor

Finally, I am grateful to my loved wife Roberta, for her infinite support,

UM MODELO DE INTERAÇÃO AXIAL DUTO-SOLO DEPENDENTE DO TEMPO

Daniel Lyrio Carneiro

Orientadores: Gilberto Bruno Ellwanger

Programa: Engenharia Civil

Durante o desenvolvimento da tese, foram elaborados modelos em elementos

A TIME-DEPENDENT AXIAL PIPE-SOIL INTERACTION MODEL FOR GLOBAL

Daniel Lyrio Carneiro

Advisors: Gilberto Bruno Ellwanger

Department: Civil Engineering

This thesis presents a novel approach to model the time-dependent pipe-soil

1. Introduction ................................................................................................................ 1

2. A Brief Review of Pipeline Walking .......................................................................... 8

3. A Brief Review of Couple Consolidation FE Analysis ............................................ 24

4. Review of the Existing Literature on Axial PSI ....................................................... 39

6. Two-Dimensional Consolidation beneath a Pipeline ............................................... 63

7. Pore Pressure Migration ........................................................................................... 77

8. Novel Model ........................................................................................................... 103

9. Final Remarks ......................................................................................................... 130

10. References .............................................................................................................. 139

Figure 2.6 – Temperature transient obtained from a dynamic flow simulation

Figure 4.4 – Cycles of loading intervened by consolidation periods illustrating the

Figure 6.2 – FE model geometry and boundary conditions. .......................................... 68

Figure 7.3 – Axisymmetric model schematics, in particular a case with overconsolidated

A pipeline cross-sectional area [L2]†

A first of four parameters in the explicit expression for D

B strain – displacement matrix

Bv volumetric strain – displacement matrix

D pipeline external diameter (including any coating if applicable) [L]

D50 mass median grain diameter of the soil [L]

D tangent elastoplastic stress – strain matrix

Del elastic parcel of the tangent stress – strain matrix

E Young’s modulus of pipeline (or equivalent modulus over A) [FL–2]

G shear modulus of soil [FL–2]

H burial depth at pipeline centreline [L]

H, Hi interpolation matrix and its partial derivatives (i = ρ, θ, x, y or z)

K bulk modulus of soil [FL–2]

K0 coefficient of lateral earth pressure at rest

Ki submatrix of K relating δfi to δξi (i = ρ or x)

Kpl submatrix of K relating δfi to δξj (i = ρ and j = x; or vice versa)

Ku, Kue transmissivity matrix of an FE model and of an element

K *u modified transmissivity matrix accounting for radial drainage

Kv, Kve coupling matrix of an FE model and of an element

L pipeline length [L]

P Effective axial force in a pipeline (positive in compression) [F]

Q matrix of artificial transmissivity accounting for radial drainage

R overconsolidation ratio in terms of mean effective stress

S ratio of total to effective average normal stress in first sweep

W pipeline submerged unit weight [FL–1]

a second of four parameters in the explicit expression for D

b third of four parameters in the explicit expression for D

cv coefficient of consolidation [L2T–1]

d prefix “infinitesimal increment of-”

d fourth of four parameters in the explicit expression for D

u* average excess pore pressure along interval [–, ] [FL–2]