Numerical Methods In Geotechnical Engineering

Proceedings Of The Seventh European Conference

On Numerical Methods In Geotechnical Engineering
Trondheim Norway 24 June 2010 Thomas Benz
Steinar Nordal download
https://ebookbell.com/product/numerical-methods-in-geotechnical-
engineering-proceedings-of-the-seventh-european-conference-on-
numerical-methods-in-geotechnical-engineering-trondheim-
norway-24-june-2010-thomas-benz-steinar-nordal-4421992

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Numerical Methods In Geotechnical Engineering Sixth European

Conference On Numerical Methods In Geotechnical Engineering Graz
Austria 68 In Engineering Water And Earth Sciences Helmut F Schweiger

https://ebookbell.com/product/numerical-methods-in-geotechnical-
engineering-sixth-european-conference-on-numerical-methods-in-
geotechnical-engineering-graz-austria-68-in-engineering-water-and-
earth-sciences-helmut-f-schweiger-2501226

Numerical Methods In Geotechnical Engineering M A Hicks R B J

Brinkgreve Alexander Rohe

https://ebookbell.com/product/numerical-methods-in-geotechnical-
engineering-m-a-hicks-r-b-j-brinkgreve-alexander-rohe-4732472

Numerical Methods In Geotechnical Engineering Ix Volume 1 Antnio S

Cardoso

https://ebookbell.com/product/numerical-methods-in-geotechnical-
engineering-ix-volume-1-antnio-s-cardoso-7160998

Advances In Numerical Methods In Geotechnical Engineering 1st Ed Hany

Shehata

https://ebookbell.com/product/advances-in-numerical-methods-in-
geotechnical-engineering-1st-ed-hany-shehata-7320718
Numerical Methods And Implementation In Geotechnical Engineering Part
1 Ym Cheng

https://ebookbell.com/product/numerical-methods-and-implementation-in-
geotechnical-engineering-part-1-ym-cheng-48746238

Numerical Methods And Implementation In Geotechnical Engineering Part

2 Ym Cheng

https://ebookbell.com/product/numerical-methods-and-implementation-in-
geotechnical-engineering-part-2-ym-cheng-48746240

Numerical Methods In Environmental Data Analysis Moses Eterigho

Emetere

https://ebookbell.com/product/numerical-methods-in-environmental-data-
analysis-moses-eterigho-emetere-46454868

Numerical Methods In Physics With Python 2nd Edition 2nd Alex Gezerlis

https://ebookbell.com/product/numerical-methods-in-physics-with-
python-2nd-edition-2nd-alex-gezerlis-51984566

Numerical Methods In Scientific Computing Volume 1 Germund Dahlquist

https://ebookbell.com/product/numerical-methods-in-scientific-
computing-volume-1-germund-dahlquist-2031168
NUMERICAL METHODS IN GEOTECHNICAL ENGINEERING
PROCEEDINGS OF THE SEVENTH EUROPEAN CONFERENCE ON NUMERICAL METHODS IN
GEOTECHNICAL ENGINEERING, TRONDHEIM, NORWAY, 2–4 JUNE 2010

Numerical Methods in Geotechnical

Engineering
Edited by
Thomas Benz & Steinar Nordal
Department of Civil and Transport Engineering, Norwegian University of
Science and Technology (NTNU), Trondheim, Norway
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2010 Taylor & Francis Group, London, UK

Typeset by MPS Ltd. (A Macmillan Company), Chennai, India

Printed and bound in Great Britain by Antony Rowe (A CPI Group Company), Chippenham, Wiltshire

All rights reserved. No part of this publication or the information contained herein may be reproduced,
stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by
photocopying, recording or otherwise, without written prior permission from the publisher.

Although all care is taken to ensure integrity and the quality of this publication and the information herein,
no responsibility is assumed by the publishers nor the author for any damage to the property or
persons as a result of operation or use of this publication and/or the information contained herein.

Published by: CRC Press/Balkema

P.O. Box 447, 2300 AK Leiden, The Netherlands
e-mail: [email protected]
www.crcpress.com – www.taylorandfrancis.co.uk – www.balkema.nl

ISBN: 978-0-415-59239-0 (Hbk)

ISBN: 978-0-203-84236-2 (eBook)
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

Table of Contents

Preface XIII
Scientific Committee (ERTC 7) XV

Constitutive modelling

A non-associated creep model for structured anisotropic clay (n-SAC) 3

G. Grimstad & S.A. Degago
A state dependent constitutive model for sand-structure interfaces 9
A. Lashkari
Adaptive integration of hypoplasticity 15
W. Fellin, M. Mittendorfer & A. Ostermann
An anisotropic bubble model for soft clays 21
N. Sivasithamparam, D. Kamrat-Pietraszewska & M. Karstunen
An anisotropic model for structured soils 27
G. Belokas & M. Kavvadas
An examination of strain space versus stress space for the formulation of
elastoplastic constitutive models 33
K.C. Ellison, K. Soga & B. Simpson
Anisotropic small strain stiffness within the multilaminate framework 39
B. Schädlich & H.F. Schweiger
Application of discontinuity layout optimization to problems involving non-associative friction 45
A.F. Babiker, C.C. Smith & M. Gilbert
Associated plasticity for nonassociated frictional materials 51
K. Krabbenhoft, A.V. Lyamin & S.W. Sloan
Comparison of methods for calculation of settlements of soft clay 57
H.P. Jostad & S.A. Degago
Effect of yield surface shape on the simulated elasto-plastic response of cohesive soils 63
A.G. Papadimitriou, A.D. Vranna, Y.F. Dafalias & M.T. Manzari
Impact of input data on soil model calibration using Genetic Algorithms 69
D. Taborda, A. Pedro, P.A.L.F. Coelho & D. Antunes
Influence of destructuration of soft clay on time-dependant settlements 75
D.F.T. Nash
Modeling liquefaction behavior of sands by means of hypoplastic model 81
A.B. Tsegaye, F. Molenkamp, R.B.J. Brinkgreve, P.G. Bonnier, R. de Jager & V. Galavi
Modeling of creep mechanism and damage of rock salt 89
B. Leuger, K. Staudtmeister, S. Yıldırım & D. Zapf
Modeling static liquefaction within multilaminate framework 95
A.B. Tsegaye, V. Galavi, R.B.J. Brinkgreve, R. de Jager, F. Molenkamp & P.G. Bonnier
On the differences between the Drucker-Prager criterion and exact implementation of
the Mohr-Coulomb criterion in FEM calculations 101
J. Clausen, L. Andersen & L. Damkilde

V
Simulation of mechanical behaviour of Toyoura sand using Severn Trent constitutive model 107
S. Miliziano, G.M. Rotisciani & F.M. Soccodato
Soil parameter identification for cyclic loading 113
A. Papon, Z.-Y. Yin, K. Moreau, Y. Riou & P.-Y. Hicher
Study of tensorial damage in a porous geomaterial 119
M. Mozayan Kharazi, C. Arson & B. Gatmiri
Time- and stress-compressibility of clays during primary consolidation 125
S.A. Degago, H.P. Jostad, M. Olsson, G. Grimstad & S. Nordal
Uncertainty and sensitivity analysis of laboratory test simulations using an elastoplastic model 131
F. Lopez-Caballero & A. Modaressi-Farahmand-Razavi
Validation of empirical formulas to derive model parameters for sands 137
R.B.J. Brinkgreve, E. Engin & H.K. Engin

Computer codes and algorithms

3D parallel computing FEA in offshore foundation design 145
L. Andresen, H. Sturm, M. Vöge & K. Skau
70-line 3D finite deformation elastoplastic finite-element code 151
W.M. Coombs, R.S. Crouch & C.E. Augarde
A simple time stepping algorithm for material point method 157
W.T. Sołowski & D. Sheng
Analysis of the stability of sheet pile walls using Discontinuity Layout Optimization 163
S.D. Clarke, C.C. Smith & M. Gilbert
Application of Discontinuity Layout Optimization to geotechnical limit analysis problems 169
M. Gilbert, C.C. Smith, I.W. Haslam & T.J. Pritchard
Enhancing solution procedures of a new numerical scheme for dynamic analysis
of soil-structure interaction problems 175
M.H. Bazyar & Ch. Song
Numerical bearing capacity computation and load-displacement behavior of shallow
foundations by stress level based ZEL method 181
M. Jahanandish, M. Veiskarami & A. Ghahramani
Simple quality indicators for FE analysis based on stress maps for Gauss points 187
C. Vulpe, N. Droniuc, E. Bourgeois & Ph. Mestat
The upper bound limit analysis of bearing capacity problems using the finite element method 193
A.I.M. AL-Janabi, A.A.R. Orabi & A.Y.A. Baqir

Discontinuum and particulate modelling

A numerical simulation on centrifuge liquefaction model using microscopic fluid
coupling scheme with Discrete Element Method 201
Y. Shimizu & Y. Inagawa
Discrete element modeling of low strength rock 207
N.B. Yenigül & M. Alvarez Grima
Effect of drying on a granular slope physical model analysed by Discrete Element Method (DEM) 213
F. Gabrieli, S. Cola, P. Simonini & F. Calvetti
Isotropic compression of cohesive-frictional particles with rolling resistance 219
S. Luding
Size effects on a virtual calibration chamber 225
J. Butlanska, M. Arroyo & A. Gens

VI
Large deformation – large strain analysis

A Coupled Eulerian-Lagrangian approach to solve geotechnical problems involving

large deformations 233
S. Henke, G. Qiu & J. Grabe
Advances in meshless methods with application to geotechnics 239
C.E. Heaney, C.E. Augarde, A.J. Deeks, W.M. Coombs & R.S. Crouch
An ALE Finite Element Method for cohesionless soil at large strains:
Computational aspects and applications 245
D. Aubram, F. Rackwitz & S.A. Savidis
Analysis of dynamic penetration of objects into soil layers 251
J.P. Carter & M. Nazem
Large deformation analysis of the installation of Dynamic Anchor 255
H. Sturm & L. Andresen
Modelling of installation effects of driven piles using hypoplasticity 261
H.D. Pham, H.K. Engin, R.B.J. Brinkgreve, & A.F. van Tol

Flow and consolidation

A multiscale approach for the consideration of spatial groundwater flow in the stability
analysis of a large excavation pit 269
H. Montenegro & R. Kauther
A numerical model for the electrokinetic treatment of natural soils with calcite 275
F. Cattaneo, C. Jommi & G. Musso
Analysis of artificial ground freezing in the Pari-Duomo platform tunnel of the Naples metro 281
S. Papakonstantinou, E. Pimentel & G. Anagnostou
Large scale hydraulic conductivity of the soil deposits of the Venezia Lagoon from
numerical back-analysis 285
E. Giacomini, F. Colleselli, F. Cattaneo, C. Jommi & G. Mayerle
Numerical analyses of granulometric stability of moraine dam cores 291
F. Federico & A. Montanaro
Numerical prediction of time-dependent rock swelling based on an example of
a major tunnel project in Ontario/Canada 297
A. Kirsch & T. Marcher
Some features of the coupled consolidation models used for the evaluation of the dissipation test 303
E. Imre & P. Rózsa
Steady state seepage flow through zoned earth structures affected by permeability defects 311
F. Federico, F. Calzoletti & A. Montanaro

Unsaturated soil mechanics

A comparison between numerical integration algorithms for unsaturated soils constitutive models 319
F. Cattaneo, G. Della Vecchia, C. Jommi & G. Maffioli
Comparison of stress update algorithms for partially saturated soil models 325
M. Hofmann, G. Hofstetter & A. Ostermann
Modelling of the hysteretic soil–water retention curve for unsaturated soils 331
A. Tsiampousi, L. Zdravkovic & D.M. Potts
Numerical integration and analysis of equilibrium in unsaturated multiphase media 337
R. Tamagnini, M. Mavroulidou & M.J. Gunn

VII
Artificial intelligence

A genetic algorithm for solving slope stability problems: From Bishop to a free slip plane 345
R. van der Meij & J.B. Sellmeijer
Simulation of the mechanical behavior of railway ballast by intelligent computing 351
M.A. Shahin
Three dimensional site characterization model of Suurpelto (Finland) using support vector machine 355
A. Pijush Samui & T. Länsivaara

Reliability and probability analysis

Evaluation of soil variability and its consequences 363
M. Huber, P.A. Vermeer & A. Bárdossy
Inverse modelling including spatial variability applied to the construction of a road embankment 369
A. Hommels, F. Molenkamp, M. Huber & P.A. Vermeer
Reliability analysis of piping in embankment dam 375
A. Noorzad & M. Rohaninejad
Spatial variability of soil parameters in an analysis of a strip footing using hypoplastic model 383
R. Suchomel & D. Mašín
Validating models against experience in foundation engineering, using the ROC curve 389
A.M.J. Mens & A.F. van Tol

Dynamic problems and Geohazards

A 2.5D finite element model for simulation of unbounded domains under dynamic loading 397
P. Alves Costa, R. Calçada, J. Couto Marques & A. Silva Cardoso
A comparison of different approaches for the modelling of shallow foundations in
seismic soil-structure interaction problems 405
S. Grange, D. Salciarini, P. Kotronis & C. Tamagnini
A finite element approach for dynamic seepage flows 411
R. Stucchi, A. Cividini & G. Gioda
A method to solve Biot’s u-U formulation for soil dynamics applications using
the ABAQUS/explicit platform 417
F.J. Ye, S.H. Goh & F.H. Lee
Alternative formulations for cyclic nonlinear elastic models: Parametric study and
comparative analyses 423
D. Taborda, L. Zdravkovic, S. Kontoe & D.M. Potts
Analysis of the effect of pile length in a pile group on the transfer and impedance functions
in soil-pile interaction models 429
A. Mahboubi & K. Panaghi
Dynamic fragmentation in rock avalanches: A numerical model of micromechanical behaviour 435
K.L. Rait & E.T. Bowman
Evaluation of the efficiency of a model of rockfall protection structures based
on real-scale experiments 441
F. Bourrier, Ph. Gotteland, A. Heymann & S. Lambert
Evaluation of viscous damping due to solid-fluid interaction in a poroelastic layer subjected
to shear dynamic actions 447
J. Grazina, P.L. Pinto & D. Taborda
Non linear numerical modeling of slopes stability under seismic loading – reinforcement effect 453
F. Hage Chehade, M. Sadek & I. Shahrour

VIII
Numerical analysis of blast impact on sealings of neighbouring structures 459
W. Krajewski, O. Reul & L. te Kamp
Numerical analysis of the seismic behavior of vertical shaft 465
S. Jeong, Y. Kim, S. Lee, J. Jang & Y. Lee
Numerical and experimental study of the detection of underground heterogeneities 471
P. Alfonsi, E. Bourgeois, F. Rocher-Lacoste, L. Lenti, & M. Froumentin
Numerical modelling of impacts on granular materials with a combined
discrete – continuum approach 477
A. Breugnot, Ph. Gotteland & P. Villard
Numerical simulations of the dynamic impact force of fluidized debris flows onto structures 483
F. Federico & A. Amoruso
Three dimensional analysis of seismic performance of an earthfill dam in Ethiopia 489
B.G. Tensay & W. Wu

Slopes and cuts

Effect of updated geometry in analyses of progressive failure 497

A.S. Gylland & H.P. Jostad
Evaluation of the effective width method for strip footings on slopes under undrained loading 503
K. Georgiadis & E. Skoufaki
Failure induced pore pressure by simple procedure in LEM 509
T. Länsivaara
Investigation of soil property sensitivity in progressive failure 515
A.S. Gylland, M.S. Sayd, H.P. Jostad & S. Bernander
Short-term slope stability calculation according to Eurocode 7 521
V. Thakur, S. Nordal & S. Hove

Embankments, shallow foundations, and settlements

3D settlement analysis using GIS and FEM: A case study in Sliedrecht area, the Netherlands 529
N.B. Yenigül & A.S. Elkadi
A comparison of 1D, 2D, and 3D settlement analyses of the Tower of Pisa 535
A.J. Klettke & L. Edgers
Analysis of a full scale failure test on old railway embankment 541
J. Mansikkamäki & T. Länsivaara
Analysis of ground movements induced by diaphragm wall installation 547
B. Garitte, M. Arroyo & A. Gens
Bearing capacity of a surface footing founded on a layered clay subsoil 553
Z. Bournta & L. Zdravkovic
Finite element analysis of the main embankment at Empingham dam 557
A. Grammatikopoulou, N. Kovacevic, L. Zdravkovic & D.M. Potts
Forecasting of the stability of the tailing dam in permafrost region on the basis
of numerical methods 563
A.B. Lolaev, A.P. Akopov, A.Kh. Oganesian, M.N. Sumin & V.V. Butygin
Numerical modeling of the mechanical response of recycled materials in embankments 569
M.M. Villani, X. Liu, A. Scarpas & A. D’Andrea
Rail track structural analysis using three-dimensional numerical models 575
A. Paixão & E. Fortunato

IX
Three dimensional analyses of ring foundations 581
M. Laman, A. Yildiz, M. Ornek & A. Demir

Piles

A back analysis of vertical load tests on bored piles in granular soil 589
L. Tosini, A. Cividini & G.Gioda
A numerical study on the effects of time on the axial load capacity of piles in soft clays 595
K.P. Giannopoulos, L. Zdravkovic & D.M. Potts
Analysis of foundation solution of new building in built-up area 601
Ž. Arbanas, V. Jagodnik & S. Dugonjić
Collapse of thin-walled model piles during hard driving 607
J. Bergan, S. Øren Holo & S. Nordal
Dynamic analysis of large diameter piles Statnamic load test 613
K.J. Bakker, F.J.M. Hoefsloot & E. de Jong
Finite difference analysis of pile on sloping ground under passive loading 619
K. Muthukkumaran & M. Gokul Khrishnan
Ground displacements due to pile driving in Gothenburg clay 625
T. Edstam & A. Kullingsjö
Lateral loading of pile foundations due to embankment construction 631
A. Feddema, J. Breedeveld & A.F. van Tol
Modelling of piled rafts with different pile models 637
S.W. Lee, W.W.L. Cheang, W.M. Swolfs & R.B.J. Brinkgreve
Modelling performance of jack-in piles 643
S. Jie & S.-A. Tan
Numerical analyses of axial load capacity of rock socketed piles in Turkey 649
M. Kirkit, H. Kılıç & C. Akgüner
Numerical simulation of low-strain integrity tests on model piles 655
J. Fischer, C. Missal, M. Breustedt & J. Stahlmann
Response of pile groups in clays under lateral loading based on 3-D numerical experiments 661
E.M. Comodromos, M.C. Papadopoulou & I.K. Rentzeperis
Selection of the proper hammer in pile driving and estimation of the total driving time 667
A. Afshani, A. Fakher & M. Palassi
Settlement analysis of a large piled raft foundation 673
M. Wehnert, T. Benz, P. Gollub & T. Cubaleski
Study of a complex deep foundation system using 3D Finite Element analysis 679
F. Tschuchnigg & H.F. Schweiger
The influence of pile displacement on soil plug capacity of open-ended pipe pile in sand 685
L. Sa, L. Grande, H. Jianchuan & L. Guohui

Deep excavations and retaining walls

3D modelling of a deep excavation in a sloping site for the assessment of induced
ground movements 693
O.J. Gastebled & S. Baghery
Analysis of an excavation in asymmetrical soil conditions: The Marquês station 699
A. Pedro, J. Almeida e Sousa, D. Taborda & P. França
Comparison of finite element predictions with results from a centrifuge test representing
a double anchor wall in sand 705
P.J. Bourne-Webb, D.M. Potts & D. König

X
Crane monopile foundation analysis 711
A. Mar
Influence of excavation and wall geometry on the base stability of excavations in soft clays 717
T. Akhlaghi, H. Norouzi & P. Hamidi
Numerical modelling of a steel sheet-pile quay wall for the harbour of Ravenna, Italy 723
D. Segato, V.M.E. Fruzzetti, P. Ruggeri, E. Sakellariadi & G. Scarpelli
Numerical modelling of spatial passive earth pressure in sand 729
M. Achmus, S. Ghassoun & K. Abdel-Rahman
Practical numerical modelling for very high reinforced earth walls 735
A. Mar, D.M. Tonks & D.A. Gorman
Short term three dimensional back-analysis of the One New Change basement in London 741
R. Fuentes, A. Pillai & M. Devriendt

Tunnels and caverns

3D analysis of a micropile umbrella for stabilizing the tunnel face of a NATM tunnel 749
F. Schmidt, C. Sagaseta & H. Konietzky
Analysis and design of a two span arch cut & cover structure 755
S. Kumar, T. Suckling, L. Macdonald & H.C. Yeow
Analysis of a bolt-reinforced tunnel face using a homogenized model 761
E. Bourgeois & E. Seyedi Hosseininia
Class A prediction of the effects induced by the Metro C construction on a preexisting
building, in Rome 767
F. Buselli, A. Logarzo, S. Miliziano & A. Zechini
Estimated settlements during the Brescia Metrobus tunnel excavation 773
A. Sanzeni, L. Zinelli & F. Colleselli
Numerical investigation of the face stability of shallow tunnels in sand 779
A. Kirsch
Numerical modeling of a bolt-reinforced tunnel in a fractured ground 785
E. Seyedi Hosseininia, E. Bourgeois & A. Pouya
On the effects of modelling gap closure and assumed soil behavior on the FE predictions of
ground movements induced by tunneling in soft clay 789
C. Miriano & C. Tamagnini
Role of numerical modelling in the current practice of tunnel and cavern design
for hydroelectric projects 795
C. Vibert, G. Colombet & O.J. Gastebled
Some modeling techniques for deep tunnels in rock with FE-continuum models 801
T. Marcher
Stress-strain behaviour of a soft-rock pillar acted upon vertical loads 807
F. Federico, S. Screpanti & G. Rastiello
Tunnel face stability with groundwater flow 813
P.M. Ströhle & P.A. Vermeer
Viscoplastic models for the analysis of tunnel reinforcement in squeezing rock conditions 819
G. Barla, D. Debernardi & D. Sterpi

Ground improvement modelling

3D FEM analysis of soil improving resin injections underneath a mediaeval tower in Italy 827
M. Gabassi, A. Pasquetto, G. Vinco & F. Mansueto

XI
A numerical study of factors governing the performance of stone columns supporting rigid
footings on soft clay 833
M.M. Killeen & B.A. McCabe
Calibration and verification of numerical model of ground improved by dynamic replacement 839
S. Kwiecien
Identification and quantification of the mechanical response of soil-wall structures
in soft ground improvement 845
X. Liu, Y. Zhao, A. Scarpas & A. de Bondt
Modelling embankments on floating stone columns 851
D. Kamrat-Pietraszewska & M. Karstunen
Numerical investigation of the mechanical behaviour of Vibro Replacement stone columns
in soft soils 857
T. Meier, E. Nacke, I. Herle & W. Wehr
Numerical modelling of consolidation around stone columns 863
J. Castro & C. Sagaseta
Numerical modeling of inertial soil-inclusion interaction 869
X. Zhang, Ph. Gotteland, P. Foray, S. Lambert & A. Hatem
Performance of geogrid-encased stone columns as a reinforcement of soft ground 875
M. Elsawy, K. Lesny & W. Richwien

Offshore geotechnical engineering

A new elasto-plastic spring element for cyclic loading of piles using the p-y-curve concept 883
O. Hededal & R. Klinkvort
Behaviour of cyclic laterally loaded large diameter monopiles in saturated sand 889
H. Ercan Taşan, F. Rackwitz & S.A. Savidis
Caisson movement caused by wave slamming—a comparison of ABAQUS and FLAC analyses 895
L. Andersen, H.F. Burcharth, T. Lykke Andersen & A.H. Augustesen
Comparison of calculation approaches for monopiles for offshore wind turbines 901
A.H. Augustesen, S.P.H. Sørensen, L.B. Ibsen, L. Andersen, M. Møller & K.T. Brødbæk
Effects of diameter on initial stiffness of p-y curves for large-diameter piles in sand 907
S.P.H. Sørensen, L.B. Ibsen & A.H. Augustesen
Numerical investigations for the pile foundation of an offshore wind turbine under
transient lateral load 913
P. Cuéllar, M. Pastor, P. Mira, J.A. Fernández-Merodo, M. Baeßler & W. Rücker
Numerical study of piping limits for installation of large diameter buckets in layered sand 921
L.B. Ibsen & C.L. Thilsted
Shallow circular foundations under undrained general combined loading in three-dimensional space 927
B. Bienen
Undrained ultimate capacity of suction anchors using an advanced constitutive model 933
S. Panayides & M. Rouainia

Numerical methods and Eurocode

Embedded cantilever retaining wall ULS design by FEA in accordance with EN 1997-1 941
A.S. Lees & S. Perdikou
Ultimate Limit State Design to Eurocode 7 using numerical methods 947
C.C. Smith & M. Gilbert

Author index 953

XII
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

Preface

These proceedings present 154 scientific papers written for the 7th European Conference on Numerical Methods
in Geotechnical Engineering, NUMGE 2010, held at Norwegian University of Science and Technology (NTNU)
in Trondheim, Norway from 2nd to 4th June 2010.
NUMGE 2010 is the seventh conference in a series of conferences organized by the ERTC7 (Numerical
Methods in Geotechnical Engineering) under the auspices of the International Society for Soil Mechanics
and Geotechnical Engineering (ISSMGE). The first conference in this series was held in Germany in 1986 in
Stuttgart and was followed by conferences every fourth year, every time in a new country in Europe: Spain in
1990 (Santander), United Kingdom in 1994 (Manchester), Italy in 1998 (Udine); France in 2002 (Paris) and
Austria in 2006 (Graz).
Following the traditions of the preceding conferences, NUMGE 2010 provides a forum for exchange of
ideas and discussion on topics related to geotechnical numerical modeling. Both senior and young researchers,
scientists and engineers from Europe and overseas countries have met at NUMGE 2010 to share and exchange
their knowledge.
The papers for NUMGE 2010 cover topics from emerging research to engineering practice. For the proceed-
ings the contributions are organized into the following sections:

Constitutive modelling
Computer codes and algorithms
Discontinuum and particulate modelling
Large deformation – large strain analysis
Flow and consolidation
Unsaturated soil mechanics
Artificial intelligence
Reliability and probability analysis
Dynamic problems and Geohazards
Slopes and cuts
Embankments, shallow foundations, and settlements
Piles
Deep excavations and retaining walls
Tunnels and caverns
Ground improvement modelling
Offshore geotechnical engineering
Numerical methods and Eurocode

The editors would like to thank all authors for their contributions, for their cooperation during the review
process and for participating in the conference. Each paper has been reviewed by a minimum of two reviewers and
the editors are grateful for help from the reviewers in achieving quality. The national representatives in ERTC7
are thanked for promoting the conference in their respective home countries. Special thanks go to Professor
Cesar Sagaseta for keeping up the work within ERTC7.
This conference is jointly organized by NTNU, NGI/ICG, and SINTEF. These institutions and all conference
sponsors are gratefully acknowledged for their generous support. Sincere thanks go to the staff at the Geotech-
nical Division at NTNU and at the Conference Secretariat, NTNU Videre for all help in organizing NUMGE 2010.

Thomas Benz and Steinar Nordal

XIII
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

Scientific Committee (ERTC 7)

Chairman
C. Sagaseta, Spain

Core Members
I. Vanicek – ISSMGE Vice President Europe
P. Mestat, France
S. Nordal, Norway
M. Pastor, Spain
J. Pestana, U.S.A.
D. Potts, U.K.
H. Schweiger, Austria
S. Sloan, Australia

National Representatives
S. Aleynikov, Russia
K. Bagi, Hungary
R. Brinkgreve, The Netherlands
I. Bojtár, Hungary
A. Bolle, Belgium
H. Burd, U.K.
A. Cividini, Italy
G. Dounias, Greece
T. Edstam, Sweden
P. Fritz, Switzerland
M. Gryczmanski, Poland
O. Hededal, Denmark
I. Herle, Czech Republic
F. Kopf, Austria
T. Länsivaara, Finland
J.C. Marques, Portugal
T. Schanz, Germany
H. Walter, Austria

Local Organizing Committee at Geotechnical Division NTNU, Norway

S. Nordal, Conference chairman
T. Benz, Editor in chief conference proceedings
A. Bihs
A. Emdal
L. Grande
M. Skjåk Bræk
P. Paniagua Lopez

Reviewers not being member of ERTC 7 or the local organizing committee

L. Andresen
C. Athanasiu
M.G. Bæverfjord
B. Bostrøm
S. Degago
G. Eiksund

XV
G. Grimstad
A. Gylland
F. Hage Chehade
H.P. Jostad
S. Kirkebo
M. Leoni
R. van der Meij
R. Schwab
C. Tamagnini
V. Thakur
D. Unteregger
B.V. Vangelsten
M. de Vries
M. Wehnert

Conference secretariat NTNU Videre

A. Bye, Coordinator

XVI
Constitutive modelling
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

A non-associated creep model for structured anisotropic clay (n-SAC)

G. Grimstad
Norwegian Geotechnical Institute, NGI, Oslo, Norway

S.A. Degago
Norwegian University of Science and Technology, NTNU, Trondheim, Norway

ABSTRACT: Elastoplastic models, based on experiments on reconstituted clays, tend to adapt an associated
flow rule. This assumption is then included in models for natural clay. However, laboratory experiments indicate
that the idea of associated flow in natural clay is insufficient. Therefore a new model, abbreviated as n-SAC,
is proposed in this paper. The model incorporates creep, using the time resistance concept, with a single creep
parameter determined from oedometer tests. Two different cap surfaces are defined in the model, i.e. the refer-
ence surface (or alternatively surface of equivalent stress measure, peq ) and the potential surface, Q. Different
(kinematic) rotational hardening rules for the two surfaces are defined along with two hardening rules for the
size of the reference surface. The two size hardening rules consists of one for the decrease in compressibility
for equivalent reconstituted material and one for loss of unstable structure. A fully implicit backward Euler
implementation scheme for the n-SAC model is used for the simulations shown in this paper.

1 BACKGROUND big consequence of wrongly predicting the horizontal

to vertical compression ratio is “bad” predictions of
Grimstad et. al. (2008) proposed a model based on settlements close to fillings.
the S-CLAY1S model (Karstunen et. al. 2005). The
model proposed in this paper has some of the elements
found in this extended S-CLAY1S model. However, 2 MODEL FORMULATION
the model differs considerably in the formulation for
flow direction. In S-CLAY1 the yield surface rota- 2.1 The reference and plastic potential surface
tion is determined from the earth pressure coefficient
Grimstad et. al. (2008) presented rate of the plastic
under virgin loading, K0NC , and the friction angle,
multiplier as a function of an equivalent stress ratio,
ϕ, (Wheeler et. al. 2003). However, Grimstad (2009)
as given by equation (1). This expression is derived
showed that this imposes unnatural limitations on the
from the time resistance concept (Janbu, 1969). The
input parameters, i.e. a “narrow” band of “good” com-
full derivation can be found in Grimstad (2009).
binations of ϕ and K0NC . The n-SAC model proposed
in this paper allows a wider range of input parameters.
Prediction of the strain behavior under various stress
paths, based on experimental evidence from e.g. Feng
(1991), indicates that the non-associated flow rule
to be reasonable. Feng conducted a series of experi- where ζi = intrinsic viscoplastic compressibility coef-
ments, on natural anisotropic clays, where both volume ficient; rsi = intrinsic time resistance number; peq
and axial deformation were measured in isotropic equivalent effective stress; pmi = intrinsic reference
compression. A decrease in both volume and height stress; τ = reference time; x = amount of unstable
were measured when considerably permanent defor- structure; and
mations were experienced. This leads to a positive
ratio between the permanent vertical and horizontal
strain increments in isotropic compression test. How-
ever, for friction angles less than about 40◦ and K0NC
from Jaky’s formula (suggested used in Wheeler et. al. where αK0NC = rotation of the potential surface in K0NC
2003) the potential surface used in S-CLAY1 gives loading (virgin oedometer loading); ηK0NC = mobi-
a negative ratio. The conclusion is that the rotational lization in K0NC loading; and MfC = critical state line
hardening rule for the potential surface of S-CLAY1 in compression loading.
and/or the shape of the potential surface needs to be The equivalent stress, peq , is calculated from the
adjusted in order to describe general soil behavior. One shape of the reference surface, which in Grimstad et. al.

3
(2008) is assumed to be associated to the potential (2002), Dafalias et. al. (2006) etc.. Wheeler et. al.
surface, Q. The shapes of these surfaces are identical (2003) argues that the rotation is dependent on the
to that of the elastoplastic Anisotropic Modified Cam deviatoric part of the plastic strain and not only the vol-
Clay Model (Dafalias, 1986), later used in S-CLAY1S umetric part as suggested by Dafalias (1986). Dafalias
(Karstunen et. al. 2005) and SANICLAY (Dafalias et. al. (2006) states three requirements for the rota-
et. al. 2006). However, unlike Dafalias (1986) and tional rules. The rotational rule proposed by Wheeler
Karstunen et. al. (2005), the n-SAC model takes a et. al. (2003), unlike Dafalias (1986), fulfills all three
similar approach as Dafalias et. al. (2006) where requirements for certain limits of input parameters
non-associated yield and potential surfaces are used. (Grimstad 2009). In the n-SAC model, the rotation
This non-association allows simulation of “softening” (rotational vector αd ) of the potential surface is depen-
response in undrained shearing without including dent solely on the volumetric strain, while the rotation
deviatoric strain dependent destructuration or spe- of the reference surface (βd ) depends generally both
cial features in the rotational hardening rule of the on volumetric and deviatoric strain. The two rotational
yield surface. The equivalent stress is calculated from hardening rules are given in equation (7) and (8).
equation (3) while the plastic potential is given in
equation (4).

where p = mean stress; σ d = deviatoric stress vec-

tor; βd = deviatoric rotational vector; M = Lode angle
dependent peak of the reference surface of in p -q
space

ζi , MC , αK0NC and ηK0NC are internal model param-

eters determined via the user input parameters given
where Mf = Lode angle dependent critical state line later in this paper.
in p -q space; αd = deviatoric rotational vector. Note Note that peq is used rather than Q for the rotation
that the precise definitions of the deviatoric vectors of the reference surface. This is introduced along with
are given in the appendix of this paper together with the Macaulay brackets to fulfill the three requirements
the Mohr-Coulomb criterion. posted by Dafalias et. al. (2006). The dependency on
In addition to introducing the non-associated flow the deviatoric strain in the rotational rule of the refer-
rule, a “creep” limit is introduced by setting the param- ence surface is included to ensure a unique rotation at
eter tmax or alternatively OCRmax below which creep “critical state”.This quality has previously been argued
should not occur. Equation (1) is then modified to that for by e.g. Wheeler et. al. (2003) and it is also used by
of equation (5). e.g. Pestana and Whittle (1999).
Modeling loss of “unstable structure” (destructura-
tion) was in a general form proposed by Gens and
Nova (1993). The same formulation was later used
in a simpler form, in e.g. Karstunen et. al. (2005)
and Grimstad et. al. (2008), in which loss of attrac-
where < > is the Macaulay brackets tion was not included. In the multilaminate framework,
Alternatively tmax could be expressed in term of Cudny and Vermeer (2004) limited destructuration to
a maximum creep induced over-consolidation ratio, be dependent on solely “normal” strain. The equivalent
OCRmax , as: approach, in standard continuum models, is visco-
plastic volume strain dependent destructuration. In
such formulations, volumetric softening can be pro-
hibited by selecting a destructuration parameter under
where OCR is defined according to the reference a certain limit. In n-SAC such requirement is auto-
time τ. matically obtained when ω = 0 in equation (9) and
rsmin > 0. However, among others Grimstad et. al.
(2008) argues that the destructuration should also be
2.2 The various hardening rules
dependent on the visco-plastic deviatoric strain, such
The precise form of the rotational hardening rules has that the “true” softening response in undrained shear-
widely been discussed by e.g. Karstunen and Wheeler ing of natural clays can be modeled. Hence equation

4
(9) is proposed as a destructuration rule for the n-SAC
model.

Isotropic hardening of the intrinsic reference stress,

pmi , is given by equation (10).

Figure 1. concept of destructuration – effect on time resis-

where ζi = a hardening parameter determined via tance number.
ref
{Eoed }i , Eref , ν and pref as given in equation (11).

The mobilization under virgin loading in oedome-

ter, ηK0NC is given by:

where λi and κ are alternative input parameters to

the model, which is in accordance with the notation
commonly used for the modified cam clay model
(MCCM), Roscoe & Burland (1968)
The value for the internal parameters βK0NC and Table 1 gives a summary of the user input parameters
αK0NC , determined from steady state rotation under to the model. In addition to these parameters the initial
K0NC loading, are given by equation (12) and (13). state variables must be given. This includes the stress
vector, σ  , the initial value of the intrinsic reference
stress, pmi0 and the initial rotation vectors α0 and β0 .
Note that pmi0 can be found from the OCR and α0
and β0 from αK0NC and βK0NC . The initial structure is
defined in equation (17).

where rsi = the intrinsic time resistance number and

rsmin = the minimum measured time resistance num-
ber, see sketch in Figure 1.
In practice 11 external parameters are then left to
be determined. This can be accomplished from only
two laboratory tests. First an incremental oedometer
tests to determine rsmin and rsi has to be executed (rsi
can also be found from an incremental oedometer test
on a remolded sample). The normalized oedometer
stiffness in the NC region for a remolded sample,
ref
Equation (14)–(16) gives more details on the calcula- {E oed }i /pref , and the normalized stress dependent
tion of some of the other internal parameters. isotropic Young’s modulus, Eref /pref , should also be
established from this test.
The second test is one undrained triaxial compres-
sions test at a OCR in the interval of 1.0-OCRmax to
determine ω (deviatoric destructuration parameter), ϕp
(friction angle at peak of the reference curve) and ϕcs
where ϕcs is the critical state friction angle
(the critical state friction angle). Note that the normal-
ized stiffness, Eref /pref , could also be found from this
undrained test.
The Poisson ratio, ν, is set by default to 0.15. How-
ever, if one for instance has radial stress measurement
where ϕp is the friction angle at the peak of the in the oedometer test it can be determined from the
reference curve. stress path in the OC region. The reference time, τ, is

5
usually 1 day, since it is common to determine OCR Table 1. Model input parameters.
for the 24 h load duration in the incremental oedometer
ref
test. ν K0NC Eref /pref {E oed }i /pref rsmin rsi

0.15 0.55 160 16 100 800

3 IMPLEMENTATION SCHEME
ω ϕp ϕcs τ tmax
The n-SAC model is implemented in an implicit back-
ward Euler integration scheme. Procedures for implicit
0.3 25◦ 33◦ 1 day 1e3yr
backward Euler implementation of general viscoplac-
tic models may be found in for instance de Borst and
Heeres (2002). In this particular case the following
residuals are defined (equation (18) to (23)):

where Dn+1 is the mean stress dependent isotropic

elasticity matrix under the assumption of a constant
Poisson ratio at the end of the calculation step

Figure 2. Result of test 1 in p − q space.

Typically the scheme will converge in a few iterations.

However, to improve performance sub-stepping will
be initiated if the number of iteration becomes higher
than 50 or if the estimated reciprocal condition number
of ∂rin+1 /∂vin+1 becomes less than 1E-12.

The 21 state variables, including the effective stresses,

are stored in a vector v:

4 NUMERICAL PERFORMANCE

In order to speed up the implementation process the In order to test the performance of the model and
number of state variables could be reduced by making implementation three different tests where ran with
use of the dependencies, i.e. αd,xx + αd,yy + αd,zz = 0, the input parameters found in Table 1. The initial state
βd,xx + βd,yy + βd,zz = 0. variables were generated from an initial vertical stress
To find the new state, a standard iterative Newton- of 72.3 kPa and an OCR of 1.383.
Raphson scheme is used as given in equation (25) Test 1 contained 181 radial strain paths of 30 steps
to (27). The iteration is ran until r∗n+1 T · r∗n+1 < TOL2 . with time increments of 0.1 day for each step given
Where r∗n+1 is a normalized version of the residual vec- in εv − εq space under the condition that dε2v + dε2q =
tors, rn+1 . The normalization is done in such a way that 1e-6. The result of test 1 is given in p − q space in
the tolerance check is irrespective of the magnitude Figure 2.
and dimension of the state variables. TOL = 1E-6 is Test 2 and test 3 are both undrained tests, consist-
used in the particular simulations shown in this paper. ing of 91 paths of 100 steps with time increments

6
 Figure 5. Comparing 100 and 10 steps for a simulation of
an undrained triaxial compression test.

peak or at 6% shear strain) are 0.51 and 0.70 respec-

tively. This is in the range reported by Bjerrum (1973)
Figure 3. Result of test 2 in σd1 − σd2 − σd3 space. and by Whittle (1993).

6 CONCLUSIONS AND RECOMMENDATIONS

This paper describes a constitutive model which

includes a variety of effects observed in clay behav-
ior. It has relatively few additional input parameters,
to that of for instance the Anisotropic Creep model,
ACM (Leoni et. al. 2008). The few “extra” input can
easily be calibrated from standard laboratory tests.
The paper presents the implementation scheme and
a variety of model simulations, i.e. constant rate of
Figure 4. Result of test 3 in plane strain deviatoric space. strain tests under various strain paths. The implemen-
tation is under these test conditions is illustrated to be
of 0.001 day for each step under the condition that sufficiently robust.
dε1 − dε3 = 6e-4. Test 2 involves only εz , εy and εz , The model and the implementation will go through
while test 3 were ran under plane strain condition (i.e. further testing, specifically trying to reproduce actual
involving εz , εy and γxy ). Results are shown in Figure 3 measured behavior in laboratory and in the field. Such
and Figure 4 in the deviatoric stress space correspond- cases will involve both settlement predictions of actual
ing to the respective strain space in which increments or hypothetical cases, compared to that of measure-
were applied. ments and along with the response predicted by other
In order to compare accuracy in the integration similar models.
method, the three tests were then repeated with 1/10 The model will hopefully be a candidate for fur-
of the number of steps used for previous simulations. ther studies on the modeling of clay behavior. Further
Result of one such comparison, for undrained triax- development to the n-SAC model could be to improve
ial compression) is shown in 5. This particular case is the behavior at small strains by incorporating non-
showing satisfactory results. linear response at small strain and elastic anisotropy.

5 DISCUSSION ACKNOWLEDGEMENTS

The particular input shown in Table 1 gives an ini- The work presented was partly carried out as a part
tial value of αNC
K0 of 0.1122 (equation (12)). This value of project 5, Geomechanical modeling, at the Inter-
for α gives a maximum ratio of horizontal to verti- national Centre of Geohazards, ICG, a Centre of
vp vp
cal visco-plastic strain (εh /εv ) of 6.7 in an isotropic Excellence (CoE) with funding from The Research
consolidation test. This is close to that measured in Council of Norway. Most of the work was finished
for instance Batiscan clay by Feng (1991). Associated when both the authors were PhD students at NTNU
anisotropic critical state models for clays do not nec- under the supervision of Professor Steinar Nordal.
essary guarantee a positive ratio for this case, or they Nordal is acknowledged for his contributions in dis-
fail to reproduce the measured K0NC − ϕ combination. cussing the content of this paper. The Marie Curie
The suP /suA and the suDSS /suA ratios (found in Figure 4 at Research Training Network “Advanced Modeling of

7
Ground Improvement on Soft Soils (AMGISS)” (Con- APPENDIX
tract No MRTN-CT-2004-512120) supported by the
European Community through the program “Human Definition of the deviatoric stress and rotational vec-
Resource and Mobility” is also acknowledged. tors:

REFERENCES
Bjerrum, L., 1973. Problems of Soil Mechanics and Con-
struction on Soft Clays, State of the Art Report to Session
IV, 8th International Conference on Soil Mechanics and
Foundation Engineering, Moscow, also in NGI report 100
(1974).
Cudny, M. and Vermeer, P. A. 2004. On the modelling of
anisotropy and destructuration of soft clays within the
multi-laminate framework, Computers and Geotechnics
31: 1–22.
Dafalias, Y. F. 1986. An anisotropic critical state soil plas-
ticity model, Mechanics research communications 13(6):
341–347.
Dafalias,Y. F., Manzari, M. T. and Papadimitriou, A. G. 2006.
SANICLAY: simple anisotropic clay plasticity model, Int.
J. Numer. Anal. Meth. Geomech. 30: 1231–1257.
de Borst, R. and Heeres, O. M. 2002,A unified approach to the
implicit integration of standard, non-standard and viscous
plasticity models, Int. J. Numer. Anal. Meth. Geomech. 26:
1059–1070.
Feng, T.W. 1991. Compressibility and permeability of natu-
ral soft clays and surcharging to reduce settlements. PhD
diss., University of Illinois at Urbana-Champaign, Urbana
Illinois.
Gens, A. and Nova, R. 1993. Conceptual bases for a constitu-
tive model for bonded soils and weak rocks, Geotechnical
Engineering of Hard Soils - Soft Rocks, Anagnostopoulos
et. al. (eds) Balkema, Rotterdam.
Grimstad, G. 2009. Development of effective stress based
anisotropic models for soft clays, PhD diss., Norwegian
University of Science andTechnology, NTNU,Trondheim.
Grimstad, G., Degago, S., Nordal, S. and Karstunen, M. 2008. The two Mohr Coulomb criteria in p − q space are
Modelling creep and rate effects using the time resis-
tance concept in a model for anisotropy and destructura-
given for the critical state (potential surface):
tion, Nordic Geotechnical Meeting, Sandefjord, Norway,
195–202.
Janbu, N. 1969. The resistance concept applied to deforma-
tions of soils. Proc. 7th Int. Conf. on Soil Mech. & Found.
Eng, Mexico city 1: 191–196.
and for the peak of the reference surface:
Karstunen, M. and Wheeler, S. 2002. Discussion of “Finite
Strain, Anisotropic Modified Cam Clay Model with Plas-
tic Spin. I: Theory” by George Z. Voyiadjis and Chung R.
Song. Journal of Engineering Mechanics, ASCE 128:
497–498.
Karstunen, M., Krenn, H., Wheeler, S. J., Koskinen, M. and
Zentar, R . 2005. The effect of anisotropy and destructura- where θ α = Modified lode angle (function of σ d and
tion on the behaviour of Murro test embankment. ASCE αd ) and θ β = Modified lode angle (function of σ d and
International Journal of Geomechanics 5(2): 87–97. βd )
Leoni, M., Karstunen, M. & Vermeer, P. A. 2008. Anisotropic The two modified Lode angles are calculated from
creep model for soft soils, Géotechnique 58(3): 215–226. the middle eigenvalues S α and S β of the tensors, sα2 and
Pestana, J. M. and Whittle, A. J. 1999. Formulation of a uni- β
fied constitutive model for clays and sands, Int. J. Numer. s2 , such that: sin (θ α ) = 3/2 · S2α /qα and sin (θ β ) = 3/2 ·
β
Anal. Meth. Geomech. 23: 1215–1243. S2 /qβ , where:
Roscoe, K. H. and Burland, J. B. 1968. On the generalized
stress-strain behavior of wet clay, Engineering plasticity,
535–609, Cambridge university press.
Wheeler, S. J., Näätänen, A., Karstunen, M. and Lojander, M.
2003. An anisotropic elastoplastic model for natural soft
clays, Canadian Geotechnical Journal 40: 403–418.
Whittle, A. J. 1993. Evaluation of a constitutive model for
overconsolidated clays, Géotechnique 43(2): 289–313.

8
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

A state dependent constitutive model for sand-structure interfaces

A. Lashkari
School of Civil and Environmental Engineering, Shiraz University of Technology, Shiraz, Iran

ABSTRACT: Numerical simulations of geotechnical engineering problems depend strongly on predictive

capacity of constitutive models used for modeling of soil-structure interface behavior. It has been shown that the
mechanical behavior of sand-structure interfaces is significantly influenced by the combined effect of density
and normal stress which is called the effect of interface state. In this paper, a two-surface interface model within
the context of bounding surface plasticity is presented. Several ingredients of the model are defined as direct
functions of interface state. It is shown that the model is capable of distinguishing interfaces in dense state from
loose ones and providing reasonable predictions.

1 INTRODUCTION later introduced by Gajo & Wood (1999). Hereto-

fore, the constitutive model of Manzari & Dafalias
The mechanical behavior of soil-structure interfaces (1997) has been widely developed and applied to var-
is an influential factor in the load-deformation and the ious problems in sands constitutive modeling includ-
bearing capacity of almost all of geotechnical engi- ing liquefaction under cyclic loadings (Dafalias and
neering problems such as retaining structures, rein- Manzari, 2004), inherent anisotropy (Dafalias et al.,
forced soils, piles, under ground and offshore gravity 2004), Non-coaxiality and liquefaction under rota-
structures. A number of experimental techniques have tional shear (Li & Dafalias, 2004; Lashkari &
been suggested in the literature to study the mechani- Latifi, 2007; Lashkari, 2009a), and unsaturated sands
cal behavior of soil-structure interfaces under different (Chiu & Ng, 2003). In the following lines, an interface
stress paths and stiffness boundary conditions. Among model based on the Manzari & Dafalias (1997) plat-
them, Ghionna & Mortara (2002) and Hu & Pu form is suggested. Comparisons between the model
(2004) used direct shear tests, Zeghal & Edil (2002) predictions and experimental results are presented to
employed ring shear apparatus, and Evgin & Fakhar- demonstrate the predictive capability of the model
ian (1996) used simple shear apparatus. On the other for simulation of the state dependent behavior of
hand, considerable developments have been achieved interfaces.
in the field of constitutive modeling of interfaces. In
this regard, Clough & Duncan (1971) suggested an
interface model based on hyperbolic elasticity theory. 2 GENERAL FORMULATION OF THE MODEL
Ghaboussi et al. (1973) suggested an elasto-plastic
cap model for soil-structure interfaces. De Gennaro & 2.1 Definitions of spaces
Frank (2002) considered phase transformation and
In this work, stress vector is defined as
ultimate state in their elasto-plastic interface model.
Ghionna & Mortara (2002) introduced a Cam-Clay
type model for sand-structure interfaces. Recently,
Liu et al. (2006) suggested that the Critical State where σn and τ are, respectively, the compressive nor-
Soil Mechanics concepts can be extended to include mal and tangential components of the stress vector.
the mechanical behavior of interfaces. They also pro- Corresponding to Equation (1), relative displacement
posed a state dependent generalized plasticity inter- vector becomes
face model using the state parameter of Been and
Jefferies (1985). More recently, Lashakari (2010) sug-
gested a generalized plasticity interface model for where v and u are normal and tangential displace-
sand-structure interfaces subjected to rotational shear. ments, respectively.
Following the impressive suggestion of Wood et al. Experimental studies have revealed that the thick-
(1994), Manzari & Dafalias (1997) introduced a ness of the interface zone is 5–10 times of mean grain
two-surface critical state compatible bounding sur- diameter of soil. In this study it is assumed that
face model for state dependent behavior of sands.
A similar constitutive model, Severn-Trent, has been

9
where t is the interface thickness and d50 is mean grain 3 THE MODEL SPECIAL ELEMENTS
diameter of grains in contact with structure. Assuming
that strains are uniformly distributed in the interface 3.1 Yield and plastic strain rate direction vectors
zone, strain vector can be defined as
Similar to the original platform of Manzari & Dafalias
(1997), a narrow wedge-shape yield function is
adopted here

where εn and εt are normal and tangential strains with

respect to the interface plane.
where η = τ/σn is stress ratio. In τ − σn plane, α is
back stress ratio which is the slope of the bisector of
the yield criterion with respect to the positive direction
2.2 Constitutive equation between stress of the σn -axis (Fig. 1). Finally, m indicates the yield
and relative displacement rate vectors surface size. In practice, m = 0.01 M is a reasonable
selection, where M is the slope of critical state line in
In the elasto-plasticity theory, the strain rate vector can τ-σn plane. Based on the yield surface introduced in
be decomposed into elastic and plastic parts Equation (10), {n} becomes

where superscripts “e” and “p” stand for the elastic where s = +1 when η > α , and s = −1 when η < α.
and plastic parts of the strain rate vector, respectively. Analogous to the work of Manzari & Dafalias (1997),
Hereafter, dot sign, . , on each parameter defines the a non-associated flow is introduced through the fol-
rate of corresponding parameter with time. In addition, lowing definition for {R}
one has the following constitutive equation between
the rates of stress and relative displacement vectors in
the elasto-plasticity theory

where Rn plays the role of dilatancy, d, in the proposed

ep
model flow rule.
where [D] is the elasto-plastic stiffness matrix which
is calculated by
3.2 State dependent peak and phase transformation
stress ratios
Liu et al. (2006) have suggested that the Critical State
Soil Mechanics concepts can also be applied to rough
interfaces. To this aim, the application of some proper
In Equation (7), [D]e is the elastic stiffness matrix

where Kn and Kt are respectively the interface normal

and tangential elastic moduli. Similar to sands, it is
assumed that these moduli are pressure dependent

Kn0 and Kt0 are model parameters and pref = 101 kPa
is the atmospheric pressure which plays the role of a
reference stress.
Kp is plastic hardening modulus. {n} and {R} are
two vectors which defines yield and plastic strain rate
directions. Particular definitions for these terms are
presented in sequel. Figure 1. Illustration of the model constitutive surfaces.

10
state parameters into the model formulation is essen- where
tial. State parameters are commonly used in order to
describe the current state of a soil or soil-structure
interface uniquely. Up-to-date reviews on a number of
state parameters can be found in Dafalias and Man-
zari (2004) and Lashkari (2009b). According to the
In above equation, A0 and A1 are model parameters,
latter work, the following constitutive equations are
where it is worthy to note that A0 is usually larger than
suggested here for state dependent peak and phase
A1 . It is observed that when a new tangential loading
transformation stress ratios
starts, the mechanical behavior of both loose and dense
interfaces is significantly contractive. Subsequently,
the mentioned contraction decreases and may turns
into dilation at moderate and large tangential displace-
ments. Considering Equation (19), α = αin < αd at the

Table 1. Model parameters used in model predictions shown

in Figs. 2–7.

where nb and nd are model parameters, and Parameter M-M-G * E-F ** S-R ***

Kt 0 (MPa) 1.57 5.0 3.0

Kn 0 (MPa) 1.85 5.85 3.60
A0 2.0 8.0 5.0
A1 0.7 1.4 0.4
In the above equation, emax , emin , and e are respec- h0 0.38 0.25 0.6
tively the maximum possible, minimum, and current M 0.63 0.638 0.64
amounts of the interface void ratio. In Equations (13) e0 0.787 1.01 0.85
and (14), σnc is normal stress corresponding to the cur- λ 0.0557 0.09 0.074
rent amount of void ratio. The location of critical state nb 0.110 0.12 0.35
line in e-σnc plane is defined by nd 0.055 0.0674 0.10

*: experiments reported by Mortara et al. (2007).

**: experiments reported by Evgin & Fakharian (1996).
***: experiments reported by Shahrour & Rezaie (1997).

In equation (16), e0 and λ are model parameters.

3.3 Dilatancy function and plastic hardening

modulus
The model constitutive surfaces are demonstrated in
Fig. 1. Similar to the yield surface, one can define back
stress ratios corresponding to bounding and dilatancy
surfaces

where αb and αd are respectively the bounding and

dilatancy back stress ratios.
Now, the model plastic hardening modulus is

where h0 is a model parameter and αin is the initial

amount of α when the most recent tangential loading
has started.
Dilatancy function is defined in the following form
Figure 2. The model predictions compared with three con-
stant normal stress tests [experimental data taken from
Mortara et al. 2007].

11
 Figure 4. The model predictions compared with three con-
stant normal stress tests [experimental data taken from
Evgin & Fakharian 1996].

Mortara et al. (2007) published the results of a

series of Gioia Tauro sand-steel interface tests car-
ried out with direct shear apparatus. Samples were
prepared by tamping method with initial relative den-
sity ID = 60%. The physical properties of the GT40
fraction of Gioia Tauro sand considered here are:
Figure 3. The model predictions compared with three con- Gs = 2.69, emax = 0.96, and emin = 0.60.
stant normal stiffness tests [experimental data taken from The model of this study is calibrated versus exper-
Mortara et al. 2007]. imental data of Mortara et al. (2007). Amounts of the
model parameters are given in Table 1. The model
predictions on tangential Stress vs. horizontal dis-
start of√a tangential loading and therefore, one has placement and normal displacement vs. horizontal
A = A0 pref /σn . As the loading proceeds, the ratio displacement for three constant normal stress tests
(α − αin )/(sαb − αin ) approaches toward 1 and leads to with σn = 25, 150, and 300 kPa are illustrated against
A = A0 at large tangential loading. The particular def- experimental data in Fig. 2. In addition, for three
inition of A given in Equation (19) enables the model constant normal stiffness tests with the same initial
for proper prediction of volume change behavior. stress conditions and K = 1.0 GPa/m, the model pre-
dictions for tangential stress vs. horizontal displace-
ment, tangential stress vs. normal stress, and normal
4 THE MODEL EVALUATION displacement vs. horizontal displacement are shown in
Fig. 3.
In experimental studies, stiffness boundary condition Evgin & Fakharian (1996) reported the results of a
in the direction normal to the interface plane is usually series of Ottawa sand-rough steel interface tests car-
defined by ried out using modified simple shear apparatus. The
physical properties of Ottawa sand are: Gs = 2.65,
emax = 1.024, and emin = 0.651. Grains are angular and
mainly made of quartz. The initial density of samples
In experiments, two types of stiffness conditions are is 84%. Using parameters presented in Table 1, simi-
usual lar comparisons are shown for evaluation of the model
versus this set of experiments. The model predictions
1. K = 0 (constant normal stress condition) in which are compared with experimental data in Fig. 4 for
σ̇n = 0 and v̇ = 0 three constant normal stress tests σn = 100, 300, and
2. K = constant (constant normal stiffness condition) 500 kPa. For three other tests carried out under con-
in which σ̇n = 0 and v̇  = 0 stant normal stiffness condition K = 800 kPa/m and

12
 Figure 6. The model predictions versus experimental
results for two dense (ID = 90%) and loose (ID = 15%)
Hostun sand-steel interfaces subjected to constant normal
stress σn = 100 kPa condition [experimental data taken from
Shahrour & Rezaie 1997].

Figure 5. The model predictions compared with three con-

stant normal stiffness tests [experimental data taken from
Evgin & Fakharian 1996].

σn = 100, 200, and 300 kPa, the model simulations

are depicted against corresponding experiments in
Fig. 5.
Using direct shear apparatus, Shahrour & Rezaie
(1997) studied the mechanical behavior of Hos-
tun sand-steel interfaces with different initial den-
sities (ID0 = 15, and 90%). The physical properties
of the Hostun sand are: Gs = 2.65, emax = 1.00, and
emin = 0.653. In all simulations, the interface thickness
is assumed 7 mm. For two dense and loose interfaces
subjected to constant normal stress σn = 100 kPa con-
dition, the model predictions are compared with exper-
imental data in Fig. 6. Similar comparisons are made
for samples with σn = 300k̇Pa in Fig. 7. For predictions
presented, the model parameters are given in Table 1. Figure 7. The model predictions versus experimental
The model has predicted a peak in tangential strength results for two dense (ID = 90%) and loose (ID = 15%)
for interfaces in dense state. In addition, the model Hostun sand-steel interfaces subjected to constant normal
is correctly predicted that the maximum tangential stress σn = 300 kPa condition [experimental data taken from
strength of loose samples is attained at large horizontal Shahrour & Rezaie 1997].

13
displacements. Moreover, the volume change behav- Evgin, E. & Fakharian, K. 1996. Effect of stress path on the
ior of loose interfaces is always contractive. On the behavior of sand-steel interface. Canadian Geotechnical
other hand, dense interfaces demonstrate contraction Journal 33: 853–865.
initially which turns into dilation in moderate and large Gajo, A. & Wood, D.M. 1999. Severn-Trent sand: a
kinematic-hardening constitutive model: the q-p formu-
horizontal displacements. From both figures, it can be lation. Géotechnique 49(5):595–614.
observed that the model can capture the fundamental Ghaboussi, J. Wilson, E. L. & Isenberg, J. 1973. Finite element
aspects of interfaces behaviors. for rock joints and interfaces. J. Soil Mech. & Found. Div.
ASCE 99 (SM10): 833–848.
Ghionna, V.N. & Mortara, G. 2002. An elastoplastic model
5 CONCLUSIONS for sand-structure interface behavior. Géotechnique 52(1):
41–50.
Within the frameworks of bounding surface plasticity Hu, L. & Pu, J. 2004. Testing and modeling of soil-structure
and Critical State Soil Mechanics, a state dependent interface. ASCE Journal of Geotechnical and Geoenviron-
mental Engineering 130(8): 851–860.
sand-structure interface model was presented. The Lashkari, A. 2009a. A constitutive model for sand liquefac-
constitutive model of Manzari and Dafalias (1997) was tion under rotational shear. Iranian Journal of Science &
selected as platform. Technology, Transaction B, Engineering 33 (B1): 31–48.
New elements for state dependency and dilatancy Lashkari, A. 2009b. On the modeling of the state depen-
were suggested. Employing the data reported by three dency of granular soils. Computers and Geotechnics 36:
independent research teams, the model predictions 1237–1245.
were compared with experimental results under var- Lashkari, A. 2010. Modeling sand-structure interfaces under
ious stiffness boundary conditions. Using a unique rotational shear. Mechanics Research Communications
set of parameters for each type of soil-structure inter- 37: 32–37.
Lashkari, A. & Latifi, M. 2007. A constitutive model for
face, it has been shown that the model is capable of non-coaxial flow of sand. Mechanics Research Commu-
providing reasonable predictions for samples of dif- nications 34: 191–200.
ferent initial states subjected to loading under various Li, X.S. & Dafalias, Y.F. 2004. A constitutive framework
stiffness boundary conditions. for anisotropic sand including non-proportional loading.
Géotechnique 54 (1): 41–55.
Liu, H., Song, E. & Ling, H. I. 2006. Constitutive modeling of
REFERENCES soil-structure interface through the concept of critical state
soil mechanics. Mechanics Research Communications 33:
Been, K. & Jefferies, M. G. 1985. A state parameter for sands. 515–531.
Géotechnique 35(2): 99–112. Manzari, M.T. & Dafalias, Y.F. 1997. A critical state two
Dafalias, Y.F. & Manzari, M.T. 2004. Simple plasticity sand surface plasticity model for sands. Géotechnique 47(2):
model accounting for fabric change effects. ASCE Journal 255–272.
of Engineering Mechanics 130(6): 622–634. Mortara, G. Mangiola,A. & Ghionna,V. N. 2007. Cyclic shear
Dafalias, Y.F. Papadimitriou, A.G. & Li, X.S. 2004. Sand stress degradation and post-cyclic behaviour from sand-
plasticity model accounting for inherent fabric anisotropy. steel interface direct shear tests. Canadian Geotechnical
ASCE Journal of Engineering Mechanics 130(11): Journal 44: 739–752.
1319–1333. Shahrour, I. & Rezaie, F. 1997. An elastoplastic constitu-
De Gennaro, V. & Frank, R. 2002. Elasto-plastic analysis tive relation for the soil-structure interface under cyclic
of the interface behavior between granular media and loading. Computers and Geotechnics 21(1): 21–39.
structure. Computers and Geotechnics 29: 547–572. Wood, D.M. Belkheir, K. & Liu D. F. 1994. Strain soften-
Chiu, C.F. & Ng, C.W.W. 2003. A state dependent elastoplas- ing and state parameter for sand modeling. Géotechnique
tic model for saturated and unsaturated soils. Géotech- 44(2):335–339.
nique 53 (9): 809–829. Zeghal, M. & Edil,T. 2002. Soil structure interaction analysis:
Clough, G.W. & Duncan, J.M. 1971. Finite element analysis modeling the interface. Canadian Geotechnical Journal
of retaining wall behavior. J. Soil Mech. & Found. Div. 39: 620–628.
ASCE 97 (SM12): 1657–1672.

14
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

Adaptive integration of hypoplasticity

W. Fellin
Division of Geotechnical and Tunnel Engineering, Department of Infrastructure,
University of Innsbruck, Austria

M. Mittendorfer & A. Ostermann

Department of Mathematics, University of Innsbruck, Austria

ABSTRACT: Evaluating the stress response of a constitutive relation of the rate type for a given strain increment
can be seen as a time integration. The question whether explicit or implicit methods should be used for this
integration is controversially discussed in the literature. In our previous paper (Fellin et al. 2009), we have
analysed two adaptive second order methods, constructed by extrapolation of the explicit and a semi-implicit
Euler method, respectively. Here, we compare their numerical behaviour on two geotechnical finite element tests.
As constitutive relation, we use hypoplasticity with the intergranular strain concept.

1 INTRODUCTION 2 ADAPTIVE INTEGRATION

The reliable computation of the stress response for a The constitutive rate equations of hypoplasticity
given strain increment is an important issue in compu-
tational geotechnics. Hypoplasticity (Kolymbas 1985)
is a framework for constitutive models of the rate
type specialised for soil behaviour. For the compar- form a system of ordinary differential equations,
ison of the numerical behaviour of an adaptive second see (Fellin et al. 2009) and references therein. Here, D
order semi-implicit method (Fellin et al. 2009) and denotes the Eulerian stretching, T the effective Cauchy
an adaptive second order explicit method (Fellin and stress and Q the additional state or internal variables.
Ostermann 2002), we choose the hypoplastic model Collecting the components of T, Q, and their deriva-
with the intergranular strain concept (Niemunis and tives with respect to the stretching in a vector y, we
Herle 1997). For single element tests the solutions obtain a nonlinear initial value problem
show a stiff behaviour all over the computational
domain (Mittendorfer 2010). Stiffness in the mathe-
matical sense means that certain implicit integrators
perform much more efficiently than explicit ones. For
adaptive explicit methods applied to stiff problems the Its efficient and reliable numerical solution is an
product of the time step size with the dominant eigen- essential step in solving the equilibrium equations.
value of the linearised system lies near the border of In our recent article (Fellin et al. 2009), we have
the stability domain (Hairer and Wanner 1996). This discussed in detail how (2) can be solved efficiently.
can enforce very small step sizes, already observed in Here, we resume briefly two attractive second order
the single element tests in (Fellin et al. 2009). methods that are both endowed with an error estimate
Motivated by these element tests, we compare here and an adaptive step size strategy.
the performance of the proposed integration schemes
for typical geotechnical problems. For this purpose, 2.1 A second order explicit method
we implement a user subroutine for the finite ele-
ment package Abaqus and we compare the numerical Starting from a numerical approximation yn ≈ y(tn ) at
behaviour on two typical finite element problems from time tn , the explicit Euler method
geotechnics: a biaxial test and a sheet pile wall exam-
ple. For these examples not only the behaviour of the
solution is important but also the structure of the spec-
imen. If the stresses in an integration point remain yields a numerical approximation yn+1 at time
roughly constant, only few time steps are required and tn+1 = tn + τn . Due to its simplicity, the method is still
the implicit method cannot exploit its advantages. much in use for integrating (2). Its main drawbacks,

15
however, are its low accuracy and the lacking error 2.3 Error estimation and step size control
control.
Next, we will treat the problem of step size selection.
A simple combination of two consecutive Euler
Our approach is that of (Fellin and Ostermann 2002;
steps, combined with a local extrapolation procedure,
Fellin et al. 2009; Hairer et al. 1993). The difference
avoids both of these drawbacks without destroying the
of the auxiliary values (4) and (7), respectively,
simplicity of the method. In the following, we briefly
describe this method.
Starting from yn , we first perform an Euler step of
size τn
is an asymptotically correct estimate for the local error
of w. For a user-supplied tolerance TOL, we obtain
an optimal step size τopt . We use this for controlling
as well as two Euler steps of size τn /2 the step size. If the estimated error EST is below the
tolerance TOL, the step is accepted and a new larger
step size is chosen for the next step. If the estimated
error EST is larger than TOL, however, we reject the
step and redo it with a smaller step size.
In order to obtain a reliable error estimate, it is
common to use the maximum norm in (9)
Taylor expansions show that the combination

is a second order approximation to the solution. with the scaling factors

The resulting method (5) is called Richardson
extrapolation of the explicit Euler method. In this
paper, we will use the name ERK2 henceforth.
The parameters ai and ri are used to fine-tune the error
2.2 A second order semi-implicit method estimate concerning the absolute and relative error
For stiff problems, explicit methods like ERK2 become tolerances for each entry of y.
inefficient. As a remedy, we proposed in (Fellin et al.
2009) to replace the Euler steps in the construction of
(5) by semi-implicit Euler steps of the form 3 FINITE ELEMENT EXAMPLES

The performance of the proposed time integration

schemes is shown with finite element calculations car-
ried out with Abaqus. The integration schemes were
In contrast to the fully implicit Euler method, this
implemented together with the constitutive equations
method only requires the solution of linear systems
and the calculation of the consistent tangent stiffness in
of equations.
a user defined material subroutine, the so called Umat.
To obtain a second order integration scheme, we
We use the default convergence criteria and load incre-
define the auxiliary variables v and w as before. Start-
mentation of Abaqus. The material parameters for the
ing from the approximation y ≈ y(tn ), we compute a
calculations are listed in Table 1.
step of size τ
Table 1. Parameters for the extended hypoplastic model.

Parameter Values
as well as two steps of size τn /2
ϕc [◦ ] 33
hs [kPa] 1 × 106
n 0.25
ed0 0.55
ec0 0.95
ei0 1.05
α 0.25
The extrapolated value β 1.50
R 1 × 10−4
mR 5.0
mT 2.0
βr 0.5
is the searched second order approximation which will χ 6.0
be called SIRK2 in the remainder of this paper.

16
 The following values for the tolerances are used in large strains and large rotations, so the lateral pressure
all numerical experiments: TOL = 10−3 and remains perpendicular to the edge of the specimen.
The calculations were performed without gravity. In
this case, the problem is symmetric with respect to the
horizontal direction.
Figure 2 is a quilt plot for the void ratio e at the end
with AERRi being the lowest resolution of the compo- of the test. The void ratio increased from the initial
nent yi . We set this value for the state variables to: 0.1 value 0.569 in the dashed area up to maximum 0.708
for the stress, 0.01 for the void ratio and 10−6 for the in the shear bands, which are formed during the test.
intergranular strain. The derivatives of the state vari- Both numerical integrations, ERK2 and SIRK2, give
ables are needed to calculate the consistent tangent the same plot.
stiffness (Fellin and Ostermann 2002). These deriva- The time steps used at the end of the test for
tives are included in the error estimation and step size ERK2 and SIRK2 are shown in Figures 3 and 4.
control, with AERR equal to 0.1. The weighting fac- For ERK2 small time step sizes are required in the
tors ri are set to one for state variables and to 100 for shear bands, whereas in other regions the time steps
the derivatives of them. are much larger. This indicates the adaptivity of our
Abaqus uses the time t as parameter throughout the method, which accounts for the stiff behaviour of
calculation. To distinguish between this time and the the constitutive equations. Due to the comparatively
time steps τ used in the constitutive time integration, large deformations in the shear bands combined with
we shall call changes of loads or boundary conditions the stiff behaviour of the hypoplastic equations the
load steps and increments of them load increments t. explicit integrator requires small time steps to meet
the accuracy requirements. It is worth to note that the
constitutive equations show stiff behaviour outside the
3.1 Biaxial test shear band as well. As the strains are rather small in
these regions, the stress remains roughly constant, and
We start with a biaxial test as standard geotechnical therefore larger steps can be accepted.
benchmark example (Hügel 1995). A soil specimen of The semi-implicit method, however, shows a rather
0.04 m width and 0.14 m height is laterally confined balanced allocation of the step sizes over the whole
with a constant stress of size 400 kN/m2 with plain structure, see Figure 4. There are barely elements in
strain condition in the other horizontal direction. The the shear bands where the step sizes are considerably
specimen is compressed vertically by a prescribed dis- smaller than in other regions. However, they are much
placement u = 0.01 m. The material in the dashed area larger than that used by the ERK2 method in the shear
of Figure 1 is given an initially higher void ratio of band.
0.569, whereas the void ratio is 0.506 elsewhere. In An “exact” solution of the load displacement curve,
this way, an initial imperfection in the dashed area is which was obtained with 200 load increments, and
simulated. The dashed area is of the size 0.02 m by the behaviour of the automatic load incrementation
0.02 m. strategy of Abaqus are shown in Figure 5. The total
The biaxial test is modelled with 8 by 28 linear
plane strain elements. The calculation accounts for

Figure 2. Biaxial compression test without gravity: void

Figure 1. Biaxial test. ratio at the end of the test.

17
 Figure 5. Biaxial compression test without gravity: load
displacement curve; the continuous line is an “exact” solu-
tion obtained with 200 load increments, the circles denote
the increments of the automatic load incrementation strategy,
which are the same for both methods ERK2 and SIRK2.

Figure 3. Biaxial compression test without gravity: time

step size τ of ERK2 at the end of the test.

Figure 6. Biaxial compression test without gravity: total

number of time steps.

post peak behaviour is not studied here. The solu-

tion obtained by the automatic load incrementation
matches quite well with the exact solution over the
whole time window.
Figure 4. Biaxial compression test without gravity: time The mean value of the conducted time steps in the
step size τ of SIRK2 at the end of the test.
four integration points of the plain strain per element
per increment are summed over all Abaqus load incre-
loading time is t = 1. The initial and maximum load ments, see Figure 6. The minimum number of required
increments are t = 0.1. steps is nearly equal for both methods. The time inte-
In this test Abaqus requires 20 load increments gration with ERK2 integration needs in the shear band
with the same sequence for both time integration meth- much more time steps in total than the integration with
ods. There is some decreasing of the load increments SIRK2. Switching from ERK2 to SIRK2 in those ele-
around the peak due to the inherent mechanical diffi- ments would make sense. However, the number of
culties of material softening. However, the increments elements where the implicit method needs consider-
are increased by Abaqus in the softening branch due to ably fewer time steps than the explicit method is rather
rapid convergence of the equilibrium iterations. This small.
indicates the good performance of the implemented Significant deformations take part only in small
consistent tangent stiffness operator (Fellin and Oster- regions of the computational domain of the biaxial
mann 2002). Note that the mesh dependence of the test. In large regions of the specimen, the imposed load

18
Table 2. Comparison of computational costs in the biaxial
test without gravity. Abaqus: automatic load incrementation
with tstart = tmax = 0.1.

Int.method No. of Inc. No. of It. CPU [s]

ERK2 20 79 200.3
SIRK2 20 83 621.8

Figure 8. Sheet pile wall example.

Table 3. Comparison of computational costs in the sheet

pile wall test. Abaqus: automatic load incrementation with
tstart = tmax = 0.1.

Int.method No. of Inc. No. of It. CPU [s]

ERK2 21 56 28.3
SIRK2 21 52 98.3

Figure 7. Biaxial compression test without gravity: relative

number of time steps per integration call.

increments are small. For such elements, the time step

size of ERK2 and SIRK2 are similar and the advan-
tages of SIRK2 for stiff equations cannot be exploited,
which is shown by a comparison of the computational
costs in Table 2. Using ERK2 in the whole domain is
about three times faster than using SIRK2.
As just mentioned, the fact that significant deforma-
tions take part in small regions of the computational
domain only has a considerable effect on the efficiency Figure 9. Sheet pile wall example: horizontal displacement
at the end of the calculation; deformations are scaled by a
of the integrators. Figure 7 displays the relative number factor 5.
of conducted times steps, taken by Umat. For instance,
in about 10% of the Umat calls SIRK2 takes between
31 and 40 time steps. To exploit the advantages of section area A = 1.2 × 10−2 m2 , moment of iner-
SIRK2, ERK2 should require considerably more time tia I = 3.84 × 10−5 m4 , Young’s modulus E = 2.1 ×
steps. This, however, is not the case for the consid- 108 kN/m2 .
ered example. Nevertheless, we have shown a good After an initial static step to impose the geostatic
performance of 2SIRK2 for our constitutive equations stress state, the second step is the excavation with fixed
in single element tests in (Fellin et al. 2009). bottom of the pit. The third load step is releasing the
bottom of the pit. For the second and third step the
initial and maximum load increments are t = 0.1, i.e.
3.2 Sheet pile wall example
a minimum of 10 load increments per step is required.
As a further typical geotechnical application we The automatic load incrementation strategy ofAbaqus
choose a 3.0 meter deep excavation with a 4.5 meter uses 10 load increments per load step for both methods
cantilever sheet pile wall. The geometry of the model only, i.e. one increment plus two times 10 increments in
is shown in Figure 8. The soil continuum is modelled total, see Table 3. Although the calculation with ERK2
with plane strain linear triangular finite elements. The requires a little bit more iterations, its computational
left and right boundary are fixed in horizontal direc- effort in terms of CPU time is considerably smaller
tion, the bottom boundary in vertical direction. The than the calculation with SIRK2S.
initial void ratio is set to e = 0.735 and the unit weight Figure 9 is a filled contour plot for the horizon-
to γ = 14.99 kN/m3 . The initial stress state is geostatic tal displacement u1 at the end of the calculation.
with the coefficient of earth pressure K0 = 0.5. The resulting horizontal displacement at the top of
The sheet pile wall is modelled with 2-node the wall is u1 = 3.819 × 10−2 m. Significant displace-
plane linear beam elements with the properties: cross ments occur in the area around the sheet pile wall,

19
 large discretisation areas around the explored problem,
i.e. the regions where considerable deformations take
part are comparatively small. Due to this fact, in most
of the elements only few time steps with an arbitrary
adaptive integration method have to be conducted. A
time step of an implicit or semi-implicit method con-
sumes more computing time than a time step of an
explicit integrator. As a consequence, implicit or semi-
implicit methods can exploit their advantages only
in regions where an explicit method needs far more
time steps. Such regions are rather small in typical
geotechnical problems as exemplified here, and adap-
tive explicit methods turn out to be the superior choice
for integrating hypoplasticity with intergranular strain
in geotechnical applications. Switching from ERK2
to SIRK2 in regions with many explicit time steps is
worth to think about. However, as these regions are typ-
ically small and any switch algorithm will take some
extra time, the effect on the overall performance is
assumed to be small.

REFERENCES
Fellin, W., M. Mittendorfer, and A. Ostermann (2009). Adap-
tive integration of constitutive rate equations. Computers
Figure 10. Sheet pile wall example: total number of time and Geotechnics 36, 698–708.
steps in constitutive time integration. Fellin, W. and A. Ostermann (2002). Consistent tangent oper-
ators for constitutive rate equations. International Journal
whereas the displacements in the rest of the structure for Numerical and Analytical Methods in Geomechan-
are rather small. ics 26, 1213–1233.
Hairer, E., S. Nørsett, and G. Wanner (1993). Solving Ordi-
Figure 10 shows the total number time steps in each nary Differential Equations I. Nonstiff Problems (2nd ed.).
element, compare Section 3.1. The minimum number Berlin: Springer.
of required steps is nearly equal for both methods. In Hairer, E. and G. Wanner (1996). Solving Ordinary Dif-
the elements around the sheet pile wall more steps ferential Equations II: Stiff and Differential-Algebraic
have to be conducted. The explicit method needs at Problems. Berlin: Springer.
most twice as much steps than the semi-implicit one. Hügel, H. (1995). Prognose von Bodenverformungen, Vol-
However, there are only few elements where ERK2 ume 136 of Veröffentlichung des Institutes für Boden-
requires significantly more steps than SIRK2. Thus, mechanik und Felsmechanik. Universität Fridericiana in
the explicit method is more efficient for this test. Karlsruhe.
Kolymbas, D. (1985). A generalized hypoelastic constitutive
It is worth to note that hypoplasticity cannot handle law. In Proc. XI Int. Conf. Soil Mechanics and Foundation
stress states with tr T > 0. Such stress states could be Engineering, San Francisco, Volume 5, Rotterdam, pp.
predicted if the integrator chooses too large time steps 2626. Balkema.
in regions where the stresses are near to zero, which Mittendorfer, M. (2010). A modular finite element setting for
is the case directly below the ground surfaces. Our nonlinear constitutive models: design and implementa-
adaptive integrator rejects such steps into the unde- tion. Ph. D. thesis, University of Innsbruck.
fined area automatically, without extra checks for the Niemunis, A. and I. Herle (1997). Hypoplastic model for
admissibility of the solution. cohesionsless soils with elastic strain range. Mechanics
of Cohesive-frictional Materials 2, 279–299.

4 CONCLUSION

For sophisticated geotechnical problems one chooses

usually a finite element approach. Generally, one has

20
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

An anisotropic bubble model for soft clays

Nallathamby Sivasithamparam
Department of Civil Engineering, University of Strathclyde, UK
Plaxis B.V, Delft, The Netherlands

Daniela Kamrat-Pietraszewska & Minna Karstunen

Department of Civil Engineering, University of Strathclyde, UK

ABSTRACT: This paper describes the principles behind a new anisotropic bubble model for natural soils. The
model is a hierarchical extension of the anisotropic S-CLAY1 model. The kinematic yield surface of S-CLAY1
model is treated as a bounding surface and a bubble surface is introduced within the bounding surface. The
bubble surface is similar in shape to the S-CLAY1 yield surface, and assumes an isotropic elastic behaviour
and an associated flow rule. A translation rule of the bubble is used to control the movement of the bubble. The
implementation of the model is first verified by simulating slow cyclic loading with constant deviator stress
on Kaolin clay, and secondly, simulations of undrained triaxial shear tests (in compression and extension) are
made to highlight the effect of evolution of anisotropy, and finally, simulations of high number of loading cycles
performed to examine ratcheting feature of the model.

1 INTRODUCTION for Bubble S-CLAY1) and its isotropic version, which

is very similar to the Al-Tabbaa model, is referred
During past few decades several modifications have as BMCC. With the introduction of the bubble, the
been proposed to enhance elasto-plastic models devel- model allows the simulation of important features of
oped within the framework of kinematic hardening soil behaviour not realized by the S-CLAY1 model,
plasticity. One of the most successful approaches is to such as non-linearity and plasticity from early stages
introduce one or two kinematic surfaces within a con- of loading and hysteretic behaviour during cyclic load-
ventionally defined yield surface (Mroz et al. 1978, ing. BSCLAY model would be ideal for simulating the
1979). Models of this type are often termed kinematic behaviour of overconsolidated soils and/or the cyclic
hardening “bubble” models (Al-Tabbaa et al. 1987, response of soils.
1989). This paper presents a new constitutive model
that is capable of representing anisotropic and cyclic
behaviour of clay. The model implementation is first 2 DESCRIPTION OF THE BSCLAY MODEL
partially verified by comparisons with the Al-Tabbaa
(1987) model simulations for Kaolin clay considering The bounding surface in the general space of model is
isotropic material. Secondly, simulations of undrained formulated based on S-CLAY1 yield surface, which in
triaxial shear tests in compression and extension high- the simplified case of triaxial space considering cross-
light the effect of evolution of anisotropy on the anisotropic sample simplifies as follows:
predicted soil response.
The proposed constitutive model is developed
within the framework of the critical state theory and
bounding surface plasticity. The model is an extension where M is the value of the stress ratio η = q/p (devi-
of the S-CLAY1 model (Wheeler et al. 1999, 2003). ator stress over mean effective stress) at critical state,
The kinematic yield surface of S-CLAY1 is treated pm defines the size of the bounding surface (see Fig-
as bounding surface, and a bubble surface (kinematic ure 1) and α defines the orientation of the bounding
yield surface) is introduced within bounding surface surface analogously to the S-CLAY1 formulation. For
to enclose a truly elastic region. The anisotropic S- the finite element implementation of the model as a
CLAY1 bounding surface can describe the effect of user-defined model in PLAXIS, the equations have
initial anisotropy caused by one-dimensional deposi- been reformulated in terms of deviatoric stress vec-
tion and K0 -consolidation process, and the subsequent tor, a deviatoric fabric tensor (in vector form) and
evolution of anisotropy due to large strains is described mean effective stress, as explained in detail by Wheeler
by a kinematic hardening law of the S-CLAY1 model. et al. (2003). The bubble (kinematic yield) surface,
The new model is called the BSCLAY model (in short based on the ideas by Al-Tabbaa (1987), encloses the

21
 the size of the bounding surface is controlled by the
change of plastic strain as follows:

where λ and κ are the slopes of the normal compression

line and swelling line in the e-lnp space and e is the
void ratio. When the bubble touches the bounding sur-
face the BSCLAY model becomes S-CLAY1 model,
and the bubble is dragged with the bounding surface.
Hardening rules and other details for the mathematical
formulation of the S-CLAY1 model can be found in
Wheeler et al. (1999, 2003).

2.2.2 Translation rules of the bubble surface

The translation rules are formulated based on transla-
tion rules proposed byAl-Tabbaa (1987). Two different
translations rules are needed, one for the case when the
bubble surface moves within the bounding surface, and
Figure 1. Schematic illustration of BSCLAY model. one for the case when the two surfaces are in contact.
The translation rule of the bubble surface is formu-
lated in a manner that guarantees that the two surfaces
truly elastic region, and it has the same shape as the (bubble & the bounding surface) can come in contact
bounding surface, but is smaller in size (see Figure 1). at a common normal but never intersect, similarly to
Although in the simplified form shown in this paper, Al-Tabbaa (1987). The centre of the kinematic yield
M is assumed the same in compression and extension, surface moves always along a vector, β, which joins the
in the FE implementation of the model, Lode angle current stress state, C, to its conjugate point, D, on the
dependency has been taken into account. bounding surface, see Figure 1. When bubble moves
within the bounding surface no rotation is allowed.
The translation rule is divided into two components.
2.1 Flow rule One is associated with the change in size of the bub-
ble surface due to isotropic expansion or contraction
Experimental evidence suggests that the assumption of the bounding surface, the other is associated with
of an associated flow rule is a reasonable approxima- the movement of the bubble surface along the vector
tion of natural clays when combined with an inclined β. When the two surfaces are in contact at the cur-
yield curve (Wheeler et al. 1999, 2003, Karstunen et al. rent stress state, the vector β is equal to zero and the
2005, 2008). Therefore, the plastic strain increment translation rule reduces to:
vector is assumed to be normal to the kinematic yield
surface at the current stress state. Consequently, the
flow rule of the model is associated and the plastic
potential is given by Equation 1.

2.2.3 Hardening modulus

2.2 Hardening rules The hardening modulus is defined in such a way that
The evolution of the bounding surface is described by when the two surfaces are in touch, and the yielding
anisotropic hardening; this means the surface rotates is continuous, the model predicts the same behaviour
as a function of large plastic strains (corresponding to as the S-CLAY1 model. It is initially formulated for
stress states at the bounding surface). The evolution of special case when two surfaces are in contact, and then
the bubble surface is described by a combination of modified for the general case when two surfaces are
kinematic and isotropic hardening, in which the sur- not in contact and the stress state is within the bounding
face translates in the stress space following the current surface.
stress point and changes size simultaneously. When the For general case when the two surfaces are not
bubble is not in touch with the bounding surface, it is in contact the hardening function h0 is given by the
not allowed to rotate. This combination of kinematic following equation:
and isotropic hardening forms the translation rule of
the kinematic yield surface.

2.2.1 Isotropic hardening of the bounding surface

The model adopts the volumetric hardening rule of In order to calculate the plastic strains whenever
the Modified Cam Clay (MCC) model. The change in they occur, whether or not the bubble surface and

22
the bounding surface are in contact, Al-Tabbaa (1987) Table 1. Model parameters for Kaolin clay.
replaced h0 with more general expression:
λ κ ν e0 M

0.187 0.017 0.3 1.2 0.9

where h0 is given by Equation 4 and H is a scalar α0 β µ R ψ
quantity which is function of the stress state.Al-Tabbaa 0.0 0.37 60 0.2 1.5
(1987) assumed, after Hashiguchi (1985), that H is a
function of a measure of the proximity of the bubble
surface to the bounding surface. The measure of the
proximity used in this model is the scalar product of to zero, the model simplifies to the isotropic MCC
the vector β and the vector n normal to the bubble model.
surface at the current stress state (see Fig. 1), divided
by the measure of the size of the bubble surface.
4 COMPARISON OF RESULTS

3 REQUIRED MODEL PARAMETERS The implementation of proposed model was first ver-
ified in a case of isotropic soil by comparing against
The proposed formulation of the model in general the results of Al-Tabbaa (1987) model predictions for
stress space requires values for 8 soil constants and slow cyclic triaxial test at constant deviator stress q.
3 state variables. These are: This simulation was initially started from normally
Soil constants: consolidated state corresponding to one-dimensional
loading, and the deviator stress q was kept constant
κ Initial slope of swelling/recompression line in e-
when cyclic loading cycles (unloading/reloading) were
lnp -space (see Al-Tabbaa 1987)
applied by changing p’. Secondly, the proposed model
ν Poisson’s ratio
was used to simulate soil behaviour under undrained
λ Slope of post yield compression line in e−lnp -
triaxial shearing following isotropic and anisotropic
space
triaxial consolidation. Table 1 summarizes the model
M Stress ratio at critical state (in triaxial compres-
parameters which obtained from Al-Tabbaa (1987),
sion)
and the additional soil constant and state variables were
µ Absolute effectiveness of rotational hardening
determined based on the suggestions by Wheeler et al.
β Relative effectiveness of rotational hardening (cal-
(1999, 2003) for β and Karstunen et al. 2005, 2008 and
culated based on M , see Wheeler et al. 2003)
Zentar et al. (2002) for µ. Given all tests by Al-Tabbaa
R Ratio of the size of the bubble surface to that of
were done for reconstituted Kaolin, the initial value
the bounding surface
for anisotropy (α0 ) has been assumed zero.
 Exponent in the hardening function H (see Al-
Figure 2 compares the simulations of the isotropic
Tabbaa 1987)
version of the proposed model (BMCC) with the sim-
State variables: ulations of Al-Tabbaa (1987). The initial values of
p and q are 300 kPa and 80 kPa respectively, and
e0 Initial void ratio
q is kept constant while cyclic changes of p are
pm Initial size of the bounding surface (calculated
applied. As mentioned above, initial anisotropy has
based on vertical preconsolidation stress)
been switched off (α0 = 0) and additionally, the evolu-
α0 Initial inclination of the yield surface (calculated
tion of anisotropy was switched off by setting µ equal
based on M, see Wheeler et al. 2003)
to zero. In reality, anisotropy would have been cre-
The soil constants of the BSCLAY model include ated though the initial K0 consolidation, resulting in a
four parameters from the MCC model (κ, λ, M and theoretical value of α0 = 0.35. The mach between the
Poisson’s ratio ν ) that can be determined from conven- two model predictions is overall very good. Although
tional laboratory tests. Two additional parameters (R, BMCC is very similar to the Al-Tabbaa (1987) model,
) are required for introduction of the bubble surface she used the modified compression and swelling
into the S-CLAY1 model. Al-Tabbaa (1987) explains indices instead of λ and κ, and hence small differ-
how these six model parameters can be obtained from ences would be expected. These results suggest that
simple standard tests or multi-stage test using the triax- the proposed model has been implemented correctly.
ial apparatus. Two additional soil constants (µ and β) A corresponding simulation with the anisotropic ver-
and additional state variable (α0 ) govern the evolution sion of the model (BSCLAY), which has not been
of anisotropy and the initial anisotropy, respectively. included in the paper, suggests that for this type of
Wheeler et al. (1999, 2003) discussed the determina- cycling loading and amplitude, anisotropy does not
tion of these three parameters in detail and generally have significant influence in the volumetric response,
no non-standard tests are needed to get reasonable esti- but nevertheless it has major impact on the predicted
mates for these values. The model is hierarchical, so deviatoric straining. Just like the isotropic BMCC ver-
it is possible to reduce the model to the S-CLAY1 sion of the model, the BSCLAY model seems to be
model, by setting R equal to one. Furthermore, if initial able to reproduce well the soil response under slow
anisotropy is switched off, by setting α0 and µ equal cyclic loading.

23
Figure 2. Slow cyclic isotropic constant q triaxial simulation: a) After Al-Tabbaa (1987) model predictions b) BSCLAY
Model simulation.

Figure 3. Simulation of undrained stress path a) after an isotropic stress history b) after a one dimensional stress history.

In Figure 3, thick solid lines represent the pre- initially isotropic and during the initial isotropic con-
dictions of the anisotropic BSCLAY model and the solidation and unloading it stays isotropic according to
dashed lines represent the equivalent results by the both models, as even BCSLAY model predicts soil to
isotropic BMCC model. In both cases the soil is stay isotropic under isotropic loading. The isotropic

24
Figure 4. High number (100 cycles) of cyclic simulation of BSCLAY model a) q/p’ versus εs and b) q/p’ versus εv .

loading is followed by isotropic unloading, corre- A high number of cyclic loading constant q triaxial
sponding to overconsolidation ratios (OCR) of 1, 1.3, simulation was performed with BSCLAY model after
2, 4 and 8. Due to the initial isotropic compression, as a one dimensional stress history. The simulation was
seen in Figure 3(a), both models predict similar stress initially one-dimensionally compressed to σv = 20 kPa
paths for undrained shearing after the isotropic com- then cyclically (100 cycles) loaded between stress
pression at early stages of the simulations, but once ratios of η = 0.45 and η = 0.23, see Figure 4. Shear
the bounding surface is reached, the prediction devi- strain continues to accumulate with increasing num-
ate. The anisotropic version of the model (BSCLAY) ber of cycles. The ratcheting feature of the model
predicts lower excess pore pressures and higher values may over-predict the shear strain after large number
of deviator stresses at failure than BMCC. of cycles. To avoid ratcheting feature of the model, the
Differences between the two model predictions are size of the bubble, R, could be made to function of
very striking in Figure 3b relating to the simulations number of cycles so that the soil will behave elasti-
of anisotropically consolidated undrained shearing in cally after large number of cycles. However, this will
compression and extension. Again, the soil is assumed require further investigations.
initially isotropic, but during the initial K0 consolida-
tion anisotropy evolves in the case of BSCLAY model,
5 CONCLUSIONS
resulting in an α-value of 0.35 at the start of undrained
shearing. Due to the associated flow rule, K0 - load-
A new constitutive model, BSCLAY, which is a hier-
ing results in different predicted stress paths, both for
archical extension of the S-CLAY1 model, has been
loading and unloading. BMCC gives just like the MCC
developed to simulate cyclic loading of anisotropic
model a very poor K0 prediction, and consequently
clays. The model is based on the principles of bound-
in the cases of high OCR the shearing starts close
ing surface plasticity. A bubble surface is introduced
to failure. Overall, during compression the BSCLAY
within S-CLAY1 model to enhance the performance
model predicts higher undrained strength than BMCC,
of the model to describe soil behaviour in over-
and the predicted undrained strength in extension is
consolidated region and under cyclic loading. The
notably lower than in compression. Once Lode angle
comparisons of the model predictions with the Al-
dependency is included, the difference is even more
Tabbaa (1987) model simulations of Kaolin clay under
significant than in the case of Drucker-Prager assumed
different stress paths, considering slow cycling load-
in this paper. In contrast the isotropic BMCC model
ing and shearing under compression and extension,
predicts almost the same value of undrained shear
revealed the predictive capability of the proposed
strength in compression and extension.
model. Ratcheting feature of the model is also verified.
In order to have a unique critical state, the rotational
hardening law of the BSCLAY model is formulated
in such as way that at reaching critical state the ACKNOWLEDGEMENTS
bounding surface keeps rotating until a unique orien-
tation is reached (see Wheeler et al. 2003 for details). The research was carried out as part of a “GEO-
Because of this feature, the results for triaxial exten- INSTALL” (Modelling Installation Effects in Geotech-
sion have strange looking curvature when approaching nical Engineering), supported by the European
critical state. This may require some modification Community through the programme “Marie Curie
when considering finite element applications, such as Industry-Academia Partnerships and Pathways” (Con-
excavations. tract No PIAP-GA-2009-230638).

25
REFERENCES Mróz, Z., Norris, V.A. & Zienkiewicz, O. C. 1979. Applica-
tion of an anisotropic hardening model in the analysis of
Al-Tabbaa, A. 1987. Permeability and stress-strain response elasto-plastic deformation of soils. Géotechnique 29, No.
of Speswhite kaolin. PhD dissertation. University of 1, 1–37.
Cambridge. Wheeler, S.J., Karstunen, M. & Näätänen, A. 1999.
Al Tabbaa & Wood, D.M. 1989 An experimentally based bub- Anisotropic hardening model for normally consolidated
ble model for clay. In: Proc. 3rd Int. Conf. on Numerical soft clay. In G.N. Pande, S. Pietruszczak & H.F. Schweiger
Models in Geomechanics. Niagara Falls, pp. 91–99. (ed.), Proc. 7th Int. Symp. on Numerical Models in
Hashiguchi, K. 1985. Two- and three-surface models of Geomechanics (NUMOG VII), Graz : 33–40. A.A.
plasticity. Proceedings of 5th International Confer- Balkema.
ence on Numerical Methods in Geomechanics, Nagoya, Wheeler, S.J., Näätänen, A., Karstunen, M. & Lojander, M.
pp. 285–292. 2003. An anisotropic elastoplastic model for soft clays.
Karstunen, M & Koskinen, M. 2008. Plastic anisotropy of Canadian Geotechnical Journal 40: 403–418.
soft reconstituted clays. Canadian Geotechnical Journal Zentar, R., Karstunen, M. & Wheeler, S.J. 2002. Influence
45: 314–328. of anisotropy and destructuration on undrained shearing
Karstunen, M.; Krenn, H.; Wheeler, S.J.; Koskinen, M., Zen- of natural clays. In P. Mestat (ed.), Proc. 5th European
tar, R. 2005. The effect of anisotropy and destructuration Conf. on Numerical Methods in Geotechnical Engineering
on the behaviour of Murro test embankment. International (NUMGE 2002), Paris: 21–26. Presses de l’ENPC.
Journal of Geomechanics (ASCE); 5(2): p. 87–97. Zentar, R., Karstunen, M., Wiltafsky, C., Schweiger, H.F. &
Mróz, Z., Norris, V.A. & Zienkiewicz, O. C. 1978. An Koskinen, M. 2002. Comparison of two approaches for
anisotropic hardening model soils and its application modelling anisotropy of soft clays. In G.N. Pande &
to cyclic loading. International Journal for Numer- S. Pietruszczak (ed.), Proc. 8th Int. Symp. on Numer-
ical and Analytical methods in Geomechanics. 2, ical Models in Geomechanics (NUMOG VIII), Rome:
203–221. 115–121. A.A. Balkema.

26
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

An anisotropic model for structured soils

G. Belokas & M. Kavvadas

National Technical University of Athens, Greece

ABSTRACT: The paper presents the mathematical formulation of the recently developed constitutive Model
for Structured Soils – 2, which is a two surface anisotropic bounding surface plasticity model. This model is able
to reproduce the engineering effects of the structure inducing mechanisms, including the effect of anisotropy
by incorporating: a) distorted and rotated ellipsoids for the Structure Strength Envelope (bounding surface) and
the Plastic Yield Envelope (elastic region – inner surface) to describe bond and stress induced anisotropy, b)
the Intrinsic Strength Envelope as a reference locus that delimits all possible unbonded states, representing a
lower bound of the bounding surface, c) the Intrinsic Compressibility Framework that describes all structureless
states, d) a damage-type mechanism to model bond degradation and e) a non-associated flow rule depending
on structure. The proposed model is modular, its features can be activated simultaneously or selectively, and the
3-D tensorial formulation facilitates direct implementation in finite elements codes.

1 INTRODUCTION the tensorial mathematical formulation of the model

allows its direct implementation in finite elements
Most natural soils are structured and the main structure codes and 3D analyses.
inducing mechanisms are stress history (e.g. preload- The tensorial formulation includes the stress ten-
ing, densification) and bonding (e.g. cementation, sor σ = s + σI (s: deviatoric stress, σ: mean effective
aging, thixotropy). These mechanisms can have a dom- stress) and the strain tensor ε = e + εI (s: deviatoric
inant role on natural soil behaviour affecting stiffness, strain, σ: volumetric strain), with I the unit tensor. In
dilatancy and strength, as well as their anisotropic the following, all stress is effective. Dots over a symbol
characteristics. Bonding results in components of stiff- denote incremental quantities.
ness, dilatancy, strength and anisotropy that cannot
be accounted solely from stress history. Therefore, it
is essential that modern constitutive modelling takes 2 STRUCTURELESS SOIL BEHAVIOUR
into account these mechanisms and their anisotropic
characteristics. This improves greatly the predictive Following Roscoe et al (1963), Lewin & Burland
capability of a constitutive model. (1970) and Leroueil & Vaughan (1990), structure-
The proposed Model for Structured Soils – 2 less materials can occur after thorough remoulding
(MSS-2) simulates the engineering effects of these (to eliminate all memory of original structure) and
structure inducing mechanisms. More specifically it subsequent radial consolidation, i.e., along a path of
advances present elastoplastic constitutive modelling constant stress ratio ησ = (s:s)0.5 / σ. This is supported
practice of structured soils by incorporating: a) dis- by various experimental evidence (see Belokas & Kav-
torted ellipsoids for the Structure Strength Envelope vadas, under review). It has also been assumed that
(bounding surface) and the Plastic Yield Envelope the behaviour of structureless materials is controlled
(elastic region) to describe bond and stress induced by current effective stress state (σ, s) and specific vol-
anisotropy, b) the Intrinsic Strength Envelope as a ume (v = 1 + e) only (e.g. Leroueil & Vaughan 1990),
reference locus that delimits all possible unbonded since remoulding has erased any pre-existing bonding
states, representing a lower bound of the bounding sur- and stress history effects.
face, c) the Intrinsic Compressibility Framework that During a radial consolidation, structureless materi-
describes all structureless states, d) a new damage-type als move along Intrinsic Compression Curves (ICC η ),
mechanism to model bond degradation and e) a plastic which are herein assumed to be parallel straight lines
dilatancy dependence on bonding and anisotropy. in the lnv–lnσ plane (Fig. 1, Equation 1) with slope
MSS-2 has been developed starting from the described by a Modified Cam-Clay compressibility
Kavvadas & Belokas (2001) model and the original parameter (ρc ). Equation 2 correlates the stress path
MSS model (Kavvadas & Amorosi, 2000). It is formu- orientation with the ICC η and is a variance of the
lated based on a hierarchical approach, which results equation proposed by Belokas & Kavvadas (in press).
in a modular and versatile model with features than Equations 1 and 2 are used for the mathematical
can be used simultaneously or selectively. Moreover, description of the structureless soil states. The lower

27
Figure 1. Structureless states and Intrinsic Compressibility
Framework (from Belokas & Kavvadas, 2010).

bound of all possible ICC η is the Critical State Curve

(CSC), which in Critical State Soil Mechanics sep-
arates contractant from dilatant behaviour, and the
upper bound is the Isotropic Intrinsic Comrpession
Curve.
Figure 2. Influence of structure in plastic dilatancy (based
on Belokas 2009 and Belokas & Kavvadas 2010b).

where σ PTC is given by Equation 4 if ICC η is described

by Equation 1.

3 STRUCTURE AND DILATANCY

The ICC η curves delimit all possible states under radial

compression. In Critical State Soil Mechanics, CSC
serves the purpose of a Phase Transition Curve (PTC). where σ o ≥ σ ∗o (see Fig. 2) is a measure of the available
To the left of PTC the behaviour is dilatant, while to structure and defines the Structure Strength Curve in
the right of PTC the behaviour is contractant (Fig. 2a). v – σ plane.
For structured soils, PTC moves to the right, therefore As loading proceeds, it generally results in a loss of
enlarging the dilatant domain (Fig. 2b). The larger the structure and bonding. Moreover, PTC shifts towards
structure. the larger the dilatant domain and the PTC the CSC and σ o tends to σ ∗o .
moves further to the right.
The distinction between dilatant and contractant
behaviour is controlled by the phase parameter, ψσ , 4 FORMULATION OF MSS-2 MODEL
which is defined by Equation 3 and is a variance of the
state parameter proposed by Been & Jefferies (1985). The formulation of the MSS-2 model is based on rate-
When the current state lies: a) to the left of PTC it independent incremental bounding-surface elasto-
is ψσ < 0 (i.e. dilatant behaviour), b) upon PTC it is plasticity. It incorporates the Critical State concepts
ψσ = 0 (e.g. critical state) and c) to the right of PTC it and the behavioural framework for structured soils
is ψσ > 0 (i.e. contractant behaviour). (e.g. Belokas, 2008). MSS-2 encompasses the impor-
tant aspects of soil behaviour described in paragraphs
2 and 3. Moreover, it is able to model structure and

28
 (Kavvadas 1983) and is the bounding surface:

Tensor σ K = σ K I + sK is the centre K of SSE and

the half-axes of the ellipsoid are equal to (α) along the
isotropic axis and (cα) along the deviatoric axes.
The above mathematical formulation can use dif-
ferent values of the c-parameter along the various
deviatoric stress components to model shear strength
anisotropy. The size (α) of the SSE is controlled by
the magnitude of structure, while the orientation vec-
tor (bK ≡ sK /σ K ) of the centre of the SSE is a direct
Figure 3. The characteristic surfaces scheme used in the measure of structure-induced anisotropy.
MSS-2 model. The PYE is, for mathematical simplicity, similar in
shape and has axes parallel to the SSE scaled down by
a factor ξ (<<1):

anisotropy evolution and degradation, as well as their

effect on soil strength, dilatancy and stiffness. It has
been developed in a modular way, i.e. its various
individual features can be activated or de-activated
depending on the amount of information or the type
of the problem to be solved.
The model presented herein has been based on the
Kavvadas & Belokas (2001) model and its complete Tensor σ L = σ L I + sL is the centre L of PYE. Since
formulation is presented in Belokas (2008). most soils behave elastically in a very limited strain
domain (order of strain ε ≈ 0.001% ÷ 0.2%), the
model typically uses a very small PYE (ξ ≈ 0.001)
which can also provide realistic modelling of cyclic
4.1 Characteristic Surfaces
loading.
MSS-2 employs three characteristic surfaces (Fig. 3): The ISE corresponds to an equivalent structureless
the Structure Strength Envelope (SSE), the Plastic state, which has the same specific volume (v) and con-
Yield Envelope (PYE) and the Intrinsic Strength Enve- solidation stress ratio (ησ ) as the natural material. This
lope (ISE), each one serving a different purpose. SSE envelope is also a rotated distorted ellipsoid centred at
delimits all possible states of a structured soil (it is the point K ∗ having coordinates: σ ∗K = σK∗ I + s∗K :
bounding surface), PYE bounds the elastic states (it
is the yield surface), while ISE is a reference surface
that corresponds to an equivalent structureless state.
These surfaces are formulated in a tensorial space
consisting of the isotropic axis (σ) and the deviatoric
hyperaxis (s).
The behaviour within the PYE has been assumed to The long axis of the ISE, oriented along the vec-
be hyperelastic. States on PYE represent the onset of tor b∗K ≡ s∗K /σK∗ , describes inherent anisotropy (con-
yielding, which correspond to limited structure degra- trolled by the consolidation stress ratio ησ ) while the
dation. For loadings directing outwards of PYE, the corresponding orientation of the SSE, bK ≡ sK /σ K ,
PYE moves together with current state (kinematic describes structure-induced anisotropy (which can be
hardening) until it comes in touch with SSE (the different). The orientation of the ISE can vary accord-
structure yield of Kavvadas 1998). This is the onset ing to a kinematic hardening rule, modelling the
of appreciable structure degradation and evolution of variation of inherent anisotropy with consolidation.
structure – induced anisotropy. The ISE represents a The size (α∗ ) of the ISE is controlled by classical Cam-
lower bound of the SSE, when all effects of structure Clay type isotropic hardening depending on the current
are eliminated (typically by intense straining). Thus, specific volume (v), the intrinsic compressibility (ρc )
in structureless soils, the SSE and ISE coincide and and the consolidation stress ratio (ησ ) expressed via
the model reduces to a Cam-Clay type twin-surface the parameter Nη (Equation 2 and Fig. 1):
model with a rotated bounding surface (SSE ≡ ISE)
and an internal bubble (PYE).
The geometrical representation of the SSE in the
stress space σ ≡ (σ, s) is a rotated distorted ellipsoid

29
 The difference in size between the SSE and ISE i.e., K moves along a radial path passing through the
(α − α∗ ) is a direct measure of the magnitude of origin. As the ratio bK ≡ sK /σK remains constant, pri-
structure. mary anisotropy does not change. The SSE reduces
to the Modified Cam-Clay yield surface if K lies on
4.2 Hardening rules the isotropic axis (σ K = αI), e.g. during an isotropic
consolidation path.
The isotropic and kinematic hardening rules control
For material states on the SSE:
the evolution of the characteristic surfaces during plas-
tic straining. Upon plastic straining, current stress state
(σ) is always on the PYE.

4.2.1 Isotropic hardening rule

MSS-2 model employs a structure degradation The second term in the above expression causes K to
isotropic hardening: deviate from the radial path during straining, altering
primary anisotropy (bK ), while the constant ψ controls
the rate of its evolution.
Motion of the centre L of the PYE
For material states on the SSE, surfaces PYE and
SSE remain in contact (at the current stress state σ)
and the position of L is dictated by the position (σ K )
of K:

For material states inside the SSE, the motion of

point L is such that the current state on the PYE (point
M on PYE in Fig. 3) moves towards a conjugate point
where Equations 11a and 11b give respectively the M  on the SSE (normal vectors at these points are paral-
p
plastic volumetric and deviatoric strain increment, (εv , lel). The geometric similarity of PYE and SSE defines
p
εq ) are the accumulated plastic volumetric and devia- the direction vector β ≡ (σ − σ L )/ξ − (σ − σ L ). There-
toric strains, (ρc , ρs ) are the intrinsic compressibility fore, the translation of the centre L is given by the
parameters during virgin compression and rebound, formula:
(ζ v , ηv ) are volumetric structure degradation parame-
ters and (θ q , ζ q , ηq ) are deviatoric structure degradation
parameters. Parameter θ q is used to increase the rate of
structure degradation by shearing, since shear-induced
This rule ensures that the characteristic surfaces do
structure degradation is usually dominant (compared
not intersect even for finite increments of the material
to volumetric structure degradation).
state (Kavvadas & Amorosi, 2000). The factor µ̇ is
In Equation 9 the component (α − α∗ ) represents
determined from the “consistency condition”, i.e. a
the magnitude of structure and α∗ is given by Equa-
requirement that during plastic deformation the stress
tion 8. Equation 9 ensures that the α − α∗ is constantly
decreasing down to the value of α − α∗ = 0, when all point remains on the PYE (f˙ = 0).
structure has been lost (i.e. SSE≡ISE), shifting to a
Cam – Clay type of isotropic hardening. Moreover, 4.3 Flow rule
by appropriately selecting the structure degradation
parameters, various rates of destructuring can be The plastic strain increment is determined by a non-
simulated, including collapse-type behaviour. associated incrementally linear plastic flow rule:

4.2.2 Kinematic hardening rules

The kinematic hardening rules describe the evolu-
tion of the structure-induced anisotropy during plastic
where the scalar ( ˙ f ) and the plastic potential ten-
straining by controlling the position of the centres
K and L of the SSE and PYE in the stress space. sor (Pf ) give the magnitude and direction of the
The MSS-2 model describes anisotropy by the primary plastic strain increment, (σ̇) is the corresponding effec-
(bK ≡ sK /σ K ) and secondary (bL ≡ sL /σ L ) anisotropy tive stress increment, (Hf ) is a “plastic modulus” as
tensors and uses the kinematic hardening rules of the described in a following section, and Qf ≡ ∂f /∂σ is
original MSS model (Kavvadas & Amorosi, 2000). the gradient of the PYE.
The plastic gradient (Pf ) has the following isotropic
Motion of the centre K of the SSE and deviatoric components:
For material states inside the SSE:

30
 where ρs , A and B are hyper-elastic constants.

6 PLASTIC MODULUS

For material states on the SSE, the plastic modulus,

Hf , is determined from the “consistency condition”,
which ensures that the stress point remains on the SSE
during plastic loading and results in equations 19, 20
and 21:

Figure 4. Failure envelopes (FE) and phase parameter (ψσ ).

where λ1 is a positive constant, h(σ) is a conical fail-

ure envelope (FE) given by Equation 16 and shown in
Figure 4, and λ2 is a positive constant.

For material states on the PYE but inside the SSE,

the plastic modulus is determined by an interpolation
mapping rule, which ensures a smooth and continuous
transition of Hf as PYE approaches SSE:
where kc and ξ are scalar and tensorial constants
respectively, related to the critical slope (M ), kd is
a positive constant and kp = kc − kd ψσ is a variable
(similar to the parameter used by Wood et al 1994). where Hfo is the plastic modulus at point M " (where
Improving the Kavvadas & Belokas (2001) model, −−→
the MSS-2 flow rule employs the phase parameter, ψσ vector OM intersects the SSE in Fig. 3), δ is the nor-
(Fig. 2), to distinguish dilatant from contractant states malized length of MM (M is the current state), δo is
and introduces a dependence on primary anisotropy the value of the parameter δ upon yield initiation and
for deviatoric strains. γ = γ 1 + γ 2 (α/α ∗ − 1) is a variable with γ 1 and γ 2
Constants λ1 and λ2 control the direction of the positive constants.
plastic strain increment for states on the SSE. The func- Equation 22 interpolates between the values:
tion hc serves the purpose of the critical state line in Hfo = ∞ (upon initiation of yielding) and: Hfo = Hfo
terms of stresses, while the function hp allows for stress (when the stress state reaches the SSE). Point M 
paths that intersect this critical state line during dilatant is calculated by the condition F(λσ; σ K , α) = 0 and
behaviour. parameter δ is defined by Equation 23 (see also
Kavvadas & Amorosi 2000). Parameter δo is reset to
the value of current δ each time yielding is re-initiated.
5 ELASTICITY
Thus, δ/δo = 1 upon initiation of yielding, δ/δo < 1 at
any later stage, and δ = 0 when the material state lies
A hyper-elastic formulation which results from an
on the SSE.
elastic strain potential V e has been incorporated:

A differentiation from previous model versions is that

variable γ includes the effect of structure on the plastic
hardening modulus.

31
 ψσ , has been employed, which controls dilatancy –
contractancy and depends on the magnitude of struc-
ture, d) the deviatoric component of the plastic flow
depends on structure anisotropy and e) the plastic
hardening modulus depends on the magnitude of
bonding.
Compared to the original MSS model (Kavvadas &
Amorosi, 2000) the major advances include the incor-
poration of: a) rotated distorted ellipsoids for the
bounding and the yield surfaces, b) a different damage-
type mechanism to model structure degradation and
c) the Intrinsic Strength Envelope as a reference
Figure 5. Influence volumetric degradation parameters on envelope.
compressibility.

REFERENCES
Been K and Jefferies MG. 1985. A state parameter for sands.
Géotechnique. 35(2):99–112.
Belokas G. 2008 Modelling of the Mechanical Behaviour
of Structured and Anisotropic Soil Materials. Ph.D The-
sis. National Technical University of Athens. pp695 (in
Greek).
Belokas G and Kavvadas M. An intrinsic compressibility
framework for clayey soils. Geotechnical and Geological
Engineering, under review.
Kavvadas M. 1983. A constitutive model for clays based
on non-associated anisotropic elasto-plasticity. Proc. of
the 2nd Int. Conf. on Constitutive Laws for Engineering
Figure 6. Influence of bonding on undrained shear response. Materials, in Tucson. p. 263–270.
Kavvadas M. 1998. Hard Soils – Soft Rocks: Modelling the
soil behaviour – Selection of soil parameters, General
Report. Proc. 2nd Int. Symp. on the Geotechnics of Hard
7 EXAMPLE SIMULATIONS Soils – Soft Rock, in Napoli. p. 1441–1482.
Kavvadas M and Amorosi A. 2000. A constitutive model for
Figure 5 shows the structure degradation during radial structured soils. Géotechnique. 50(1): 263–273.
compression and Figure 6 shows structure degrada- Kavvadas MJ and Belokas G. 2001. An anisotropic elasto-
tion for various degrees of bonding (α/α∗ ) during and plastic constitutive model for natural soils. Proc. 10th Int.
undrained shear. Conf. on Computer Methods andAdvances in Geomechan-
ics (IACMAG), in Tucson, Arizona,. p. 335–340.
Leroueil S and Vaughan PR. 1990. The general and congru-
ent effects of structure in natural soils and weak rocks.
8 CONCLUSIONS Géotechnique. 40(3):467–488.
Lewin P.I. and Burland J.B. 1970. Stress-probe experiments
The formulation of an anisotropic bounding sur- on saturated normally consolidated clay. Géotechnique.
face plasticity constitutive Model for Structured Soils 20(1):38–56.
Roscoe K.H., Schofield A.N. and Thurairajah A. 1963.Yield-
(MSS-2) has been presented. It has been based on the ing of clays in states wetter than critical. Geotechnique.
Kavvadas & Belokas (2001) the major advances being 13(3):211–240.
the following: a) a hyperelastic formulation has been Wood DM, Belkheir K and Liu DF. 1994. Strain softening
employed, b) the Intrinsic Compression Curves are and state parameter for sand modelling. Technical Note.
linear in the lnv-lnσ plane, c) the phase parameter, Géotechnique. 44(2):335–339.

32
 Numerical Methods in Geotechnical Engineering – Benz & Nordal (eds)
© 2010 Taylor & Francis Group, London, ISBN 978-0-415-59239-0

An examination of strain space versus stress space for the formulation of

elastoplastic constitutive models

K.C. Ellison & K. Soga

University of Cambridge, Cambridge, UK

B. Simpson
Arup Geotechnics, London, UK

ABSTRACT: Several elastoplastic soil models have been proposed over the years that are formulated in
strain space rather than stress space due to certain analytical and computational advantages. One such model,
BRICK (Simpson 1992), has been continuously utilized and developed for industrial applications within Arup
Geotechnics for more than two decades. This paper aims to describe the advantages and difficulties associated
with strain space modeling. In addition, it will show how recent advances in modeling the effects of stress
history, stiffness anisotropy, strength anisotropy and time-dependence in conventional stress space models can
be transferred to the BRICK model.

1 INTRODUCTION 2 BRIEF OVERVIEW OF THE BRICK MODEL

Conventional elastoplastic critical state based con- A simple plane-strain version of the BRICK model
stitutive models for soil behavior are formulated was described by Simpson (1992). The key features of
primarily in stress space with one or more yield this model will be reviewed briefly here.
surfaces defined in terms of principal stresses. Alter- In the original BRICK model, the current strain
natively, several constitutive models for geomaterials state was defined in a three-dimensional coor-
have been proposed over the years that are formu- dinate system with one volumetric strain axis,
lated entirely in strain space due to certain analytical εvol = (εx + εy )/2, and two shear strain axes, εx − εy
and computational advantages over the conventional (pure shear) and γxy (simple shear). In addition, sev-
models (e.g. Yoder & Iwan (1981) and Iwan & Chel- eral ‘bricks’ that represented a portion of the material
vakumar (1988)). While most of these have failed were attached to the strain point by ‘strings’of different
to gain traction outside of academic realms, the lengths in this strain space. When a string became taut,
BRICK model (Simpson 1992) has been continu- the corresponding brick would move toward the strain
ously utilized and developed for industrial applica- point as demonstrated by the sequence in Figure 1 and
tions within Arup Geotechnics for more than two plastic strain would develop along that path. The total
decades. plastic strain increment {ε̇p } was determined by sum-
Since the initial formulation of the BRICK model, ming the contributions of each brick while the elastic
many advances have been described in the literature strain increment was determined simply via the rela-
to capture specific features of soil behavior in stress tionship {ε̇e } = {ε̇} − {ε̇p }. Thus, when all the strings
space models; however, there is little in the litera- were loose, the soil behavior was fully elastic and when
ture to describe how these advances may be applied all the strings were taut and lined up behind the strain
in strain space. For example, recent constitutive mod- point in the orientation of the strain increment, the soil
els can explicitly consider the effects of stress history, behavior was perfectly plastic.
creep, rate-dependence, stiffness anisotropy, strength In addition to the BRICK effect described above, it
anisotropy and other factors in ways that the original was assumed that the soil’s capacity for elastic strains
single-surface models could not. During the ongoing increased as the mean effective stress increased. This
development of the BRICK model, several of these was achieved by discounting some of the volumetric
advances have been modified for use in strain space. plastic strains indicated by brick movements so that
This paper aims to describe the advantages and diffi- changes in volumetric strain followed the appropriate
culties of modeling in strain space as well as to outline normal consolidation or swelling line in εvol -ln s space
how recent advances can be applied to BRICK-type (when all the bricks were aligned) where s was the
models. mean in-plane effective stress. Increases in volumetric

33
Figure 1. Example sequence of strain point and brick movement during 1D consolidation (a→b) and undrained extension
(b→d).

strain were also associated with an increase in the by the accumulation of strain. Thus, strain space
capacity for shear strain. Therefore, shear plastic strain models may be better suited to explain the underly-
reductions were applied such that shear failure was ing mechanisms that govern the constitutive behavior
achieved at a constant stress ratio t/s = f ({Lb },{Rb }) of soil.
where t was the shear stress and {Lb } and {Rb } were the It has long been recognized that void ratio is a criti-
string lengths and material proportions, respectively, cal parameter to soil behavior. In stress space models,
for each brick ‘b’. the void ratio (or volumetric strain) is needed to iso-
The elastic stiffness in the BRICK model is late the critical state line. Moreover, plastic strains are
pressure-dependent, i.e. K e = s /ι where K e is the needed to compute the hardening of yield surfaces in
elastic bulk modulus and ι is a user-defined elastic stress space models and it seems more appropriate to
stiffness parameter. However, it is also presumed that derive these plastic strains directly from an increment
the strength and stiffness of the material will increase of strain rather than an increment of stress. Therefore,
with overconsolidation as measured by the distance strain space is simply a more consistent basis for a
of the current strain point from the normal consoli- constitutive model.
dation line in εvol -ln s space. Increased stiffness was As another example, consider the phenomenon of
introduced by decreasing the parameter ι from its user- stiffness anisotropy which arises from the preferred
specified value. This would also lead to an increase orientations of particles and particle contacts that make
in the failure stress ratio t/s which was modified by up a soil’s fabric. The accumulation of large strains
adjusting the string lengths {Lb }. might change this fabric and corresponding anisotropy
The three-dimensional (3D) BRICK model cur- even if the initial and final stress states are the same.
rently used by Arup is based on the same principles as Stress space models account for this by allowing the
the plane-strain version described by Simpson (1992) yield surface to expand, translate or rotate throughout
and summarized above.The major difference is that the the stress history; however, a more realistic descrip-
strain point and bricks are defined in a six-dimensional tion should examine strain history since this is a better
space comprised of one volumetric strain and five measure of the change in fabric.
shear strains. The details of this formulation have Furthermore, if a specimen is subjected to an abrupt
recently been described by Ellison (2009) and Clarke change in stress path, plastic strains will initially con-
(2009). tinue to develop in the direction of its recent strain
history (Atkinson et. al. 1990). If the initial stress path
is small, then its recent stress history can be ‘forgotten’
3 ADVANTAGES OF STRAIN SPACE
after a period of creep (Clayton & Heymann 2001).
However, more significant stress histories cannot be
3.1 Philosophical advantages
completely forgotten due to creep (Gasparre 2005).
As stated in the frontispiece of Professor John This observation is best described by an examination of
Burland’s PhD thesis (Burland 1967) and reiterated strain: small strains will result in a small change to the
during Brian Simpson’s Rankine Lecture (Simpson soil fabric that can be overcome by subsequent creep
1992): ‘Stress is a philosophical concept - deformation strains whereas larger strains may result in a signifi-
is the physical reality’. This quote encapsulates one of cant change to the soil fabric that cannot be overcome
the most compelling reasons that an examination of by creep. In contrast, the influence of stress changes
strain rather than stress might be more appropriate to on the soil fabric will be harder to gauge since this will
describe the evolution of soil behavior. While these vary with the current stress state.
two measures are inextricably linked, changes in soil It also makes intuitive sense to model both creep and
behavior are ultimately caused by micromechanical the related phenomenon of stress relaxation (whereby
changes in soil fabric that are reflected at the mesoscale stress decreases over time while strain remains

34
constant) in strain space since these phenomena are
thought to arise from the gradual rearrangement of par-
ticles due to bond failures at the molecular level. This
rearrangement would best be expressed directly by the
development of plastic strains rather than indirectly
through the propensity of a yield surface to reposition
itself in stress space.

3.2 Practical advantages

There are also some practical reasons to formulate con-
stitutive models in strain space. The primary reason
is that strain space models are naturally compatible
with the finite element method whereby an incre- Figure 2. Sample discretization of the S-shaped curve.
ment of strain is shuttled into the constitutive model
and an increment of stress must be produced, i.e. rate of stiffness degradation compared with conven-
{σ̇} = f({ε̇}, {k}) where {k} is comprised of the state tional flow rules. However, a finite number of bricks
variables (e.g. Chelvakumar & Iwan (1988)). On the must be used and, therefore, the stiffness curve will
other hand, stress space formulations must be rear- not necessarily be smooth. For practical purposes, 10
ranged to make strain the independent variable during bricks are normally used, but in principle the number
a numerical time step, often at the expense of computa- is not limited.
tional efficiency. This complication is apparent in the The greatest challenge in working with BRICK is
calculation of the scalar multiplier  which is used in the determination of the in situ state variables. At the
conjunction with the plastic potential ∂P/∂σ to obtain start of a BRICK simulation, the bricks and strain
the plastic strain increment. point must be located along the volumetric strain axis
A second practical advantage of strain space model- or else anisotropic strength will be predicted. There-
ing is that it eliminates the need to make assumptions fore, a best estimate of the brick locations at the in
about the intersection of yield surfaces. Multi-surface situ state is usually obtained by modeling a portion of
formulations in stress space with bilinear constitutive geological history considered to have influenced the
laws require that no two surfaces intersect or else the current state. For London clay, an arbitrary reference
uniqueness of solution could be destroyed (Puzrin & state of p = 2 kPa is typically used where p is the
Houlsby 2001). However, Yoder (1981) demonstrated mean effective stress. This is inconvenient because the
that no such problem exists for yield surfaces in strain geological strain history is not always well-known and
space. Moreover, in multisurface stress space models, there is no guarantee that its simulation will result in
it can be difficult to determine the proportion of a the same horizontal pressures observed in the field.
strain increment that takes place before a new surface By comparison, conventional stress space models are
is encountered. This is usually remedied via a com- attractive because their simulations may begin at the in
putationally expensive iteration scheme that would be situ state. As a result, the initial size, shape and posi-
unnecessary in a strain space formulation. tion of their yield surfaces may be selected to agree
Finally, Yoder (1981) asserts that stress space mod- with probed stress paths from laboratory tests. How-
els can become unstable for special cases (such as ever, this advantage will disappear once a yield surface
perfect plasticity) unless they are explicitly consid- is engaged.
ered within a computer code. On the other hand, strain The ability of the BRICK model to replicate
space models can seamlessly unify the cases of hard- undrained deformations has been well-established
ening, softening and ideally plastic behavior (Iwan & (e.g. Simpson (1992) and Pillai (1996)); however,
Chelvakumar 1988). comparison of the measured and simulated tests of
high quality undrained triaxial tests on London Clay in
Figure 3 demonstrates that the model is less successful
4 DIFFICULTIES ASSOCIATED WITH THE at capturing the effective stress paths. The orientations
ORIGINAL 3D BRICK MODEL of the undrained stress paths are incorrect in this sim-
ulation partly because the model does not account for
Despite the advantages described above, strain space elastic anisotropy. Moreover, the BRICK effect occa-
models exhibit a few difficulties relative to conven- sionally results in unrealistic ‘kinks’ or rapid changes
tional models that have prevented them from gaining in the stress path direction. These kinks occur when a
widespread popularity. complex recent strain history has changed the location
Rather than defining yield surfaces and a flow rule, of the bricks on short strings relative to the cur-
a user of the BRICK model must explicitly select rent strain point without significantly affecting those
the size and weight of many kinematic surfaces by attached to the longer strings. The resulting orienta-
defining points along an S-shaped curve comprised of tion of the plastic strain increment during a monotonic
material proportions and string lengths as shown in loading sequence can change rapidly once the bricks
Figure 2. This affords the user greater control over the on the longer strings become engaged.

35
Figure 3. Simulations of undrained triaxial compression and drained triaxial extension tests on samples from a depths of
11 m in Unit B2(c) of London Clay using the original 3D BRICK model (lab data from Gasparre (2005)).

As also shown by Figure 3, the model tends to under- and the orientation of a vector connecting the current
predict dilation during drained simulations. It can be stress point to its conjugate point on a larger surface.
seen from Figure 1d that continued shear straining will This ensures that yield surfaces will only intersect
lead to additional volumetric strain until the bricks line tangentially at conjugate points.
up parallel to the shear strain axis. In this manner, the Stress history is considered by the BRICK model
model computes some plastic dilation; however, there in a similar manner except that there may be multi-
is no explicit flow rule. ple active surfaces (i.e. taut strings) at a given time.
It will be shown in a subsequent section that the In fact, as mentioned in a previous section, the for-
incorporation of stiffness anisotropy in the BRICK mulation is even simpler in strain space since there is
model significantly improves the predictions of both no need to force subsequent yield surfaces to intersect
dilation and effective stress paths. However, this is tangentially.
much more difficult to accomplish in BRICK-type
models than in conventional models. In stress space,
it is a relatively trivial task to incorporate stiffness 5.2 Creep, ageing and rate effects
anisotropy by substituting the isotropic elastic stiff- Many studies have highlighted the roles of creep, age-
ness matrix with an anisotropic one. However, in ing and rate effects on soil behavior. It has been shown
the BRICK model, strength and elastic stiffness are that creep and ageing can cause the elastic region
closely intertwined. Therefore, one cannot simply to recenter itself about the current stress state (e.g.
introduce an anisotropic elastic stiffness matrix with- Clayton & Heymann (2001) and Gasparre (2005)).
out inducing an equivalent and undesired anisotropy A related phenomenon known as isotach behavior
of strength. describes how changes in the strain rate applied to
Lastly, unlike the conventional critical state based some soils can cause a jump between different isotach
models, BRICK does not necessarily approach a criti- stress-strain curves (e.g. Suklje (1969)).
cal state line in εvol -ln p space. As a result, its primary The most popular methods to capture creep and
applications are currently limited to stiff clays that isotach behavior due to viscoplastic time-dependent
undergo strain localization before such a line would effects in geomaterials are the nonstationary flow
ever be reached. surface (NSFS) and overstress theories (e.g. Perzyna
(1966) and Naghdi & Murch (1963)). The NSFS the-
5 INCORPORATING SPECIFIC FEATURES IN ory utilizes a variant of the classical elastoplastic yield
BRICK-TYPE MODELS surface that is a function of strain rate. The over-
stress theory postulates that a dynamic yield surface
5.1 Stress history effects exists beyond the static yield surface that depends upon
the strain rate and that these surfaces will gradually
Many constitutive models have employed multiple converge as the strain rate reduces to zero.
kinematically-translating yield surfaces to describe the The principles of the overstress theory readily lend
influence of stress history on the anisotropic hardening themselves for incorporation into BRICK-type mod-
of geomaterials (e.g. Dafalias & Herrmann (1982) and els. Rather than employing rate-dependent dynamic
Stallebrass & Taylor (1997)). It is straightforward to yield surfaces, Clarke (2009) has employed rate-
convert this type of formulation to strain space and this dependent string lengths that gradually converge
is the only extraordinary feature explicitly considered upon their reference values as the stress/strain rate
by the original BRICK model. decreases. Clarke’s strain rate dependent string lengths
In conventional multi-surface models, one or more are determined by the following equation:
surfaces are usually nested within an isotropically-
expanding bounding surface. The largest yield surface
engaged at any time is the active yield surface and the
translation of this surface is a function of normality

36
where β is a material constant and the superscripts shear failure. Therefore, the following equation is
‘tar’ and ‘ref ’ refer to ‘target’ and ‘reference’ values, used:
respectively. However, to avoid a large jump in string
length due to a sudden change in strain rate, a damping
function is introduced:

It is postulated that the same type of string length

correction could be applied to introduce strength
anisotropy; however, such a formulation has yet to be
where α is a material constant, t is time and the incorporated into the model.
superscript ‘n’ refers to the current step number.
Implications of the Clarke (2009) formulation are:
(i) Creep occurs due to the accumulation of plas- 5.4 Stiffness anisotropy
tic strains as the strings contract, (ii) stiffer behavior Numerous laboratory studies have revealed that the
is predicted after creep since all of the strings will small strain linear elastic behavior of soil is not
initially become loose when the strain rate increases isotropic and that the extent of anisotropy is relatively
again, and (iii) isotach behavior can be replicated. constant up to intermediate strains (e.g. Gasparre
(2005) and Yimsiri et al. (2009)). The conventional
isotropic elastic stiffness matrix can easily be substi-
5.3 Strength anisotropy and Lode angle effects tuted with an anisotropic one in stress space models.
It has long been observed that the size of the elastic It is not so simple to incorporate stiffness anisotropy
region and the residual stress ratio achieved by labora- in BRICK-type models for two reasons: 1) elastic
tory specimens are stress path dependent. Thus, many stiffness anisotropy would result in an equivalent
formulations have been proposed to alter the shape of anisotropy of strength; and 2) the BRICK model must
the yield surface in the π-plane with respect to the Lode simulate the entire geologic history over which the
angle θ (e.g. Matsuoka & Nakai (1977)). Such formu- anisotropy has likely developed. An appropriate treat-
lations can easily be incorporated into BRICK-type ment of stiffness anisotropy should account for the
models by modifying the string lengths to be strain evolution of anisotropy with strain without altering the
path dependent. residual strength condition.
For example, the Modified Drucker Prager surface The problem of strength anisotropy arising from
may be adopted, i.e.: elastic stiffness anisotropy can be mitigated by pass-
ing a modified strain increment {ε̇mod } through the
constitutive model, i.e.,

where µ is a constant that controls the shape of the where [Daniso ] and [Diso ] are the anisotropic and
yield surface in the π-plane, q* and r* are invari- isotropic stiffness matrices, respectively.
ants derived from the deviatoric stress tensor and α A matrix [M ] can be defined to convert between
is related to the effective friction angle and controls the real strain increment and the modified strain
the slope of the yield surface in the meridian plane. increment, i.e.,
This criterion is readily employed within models where
the failure surface is imposed as a discrete boundary
around pre-failure behaviour. However, it is not read-
ily adapted to the BRICK model in which stiffness and
strength are interrelated. In this formulation both the elastic and plastic strains
The 3D BRICK model used by Arup employs a will be anisotropic. As a result, there is no guaran-
revised version of Equation 3 rewritten using equiv- tee that the perfectly plastic behavior at critical state
alent strain terms to express the ratio of the polar will be volume-preserving. Thus, in order to achieve a
distance from the volumetric strain axis to the max- constant volume condition, the model must evolve to
imum polar distance (corresponding to triaxial com- become isotropic at critical state (i.e. [M ] → [I ] where
pression), therefore: [I ] is the identity matrix).
Ellison (2009) presents one possible formulation for
stiffness anisotropy in BRICK using the above frame-
work. In this model, direction-dependent anisotropy
increases with the amount of shear strain developed
since a reference value and decreases with the devel-
q
where γbr and γb are invariants of the deviatoric strain opment of shear stress such that the model becomes
along string ‘b’. However, the full string length correc- isotropic near the residual stress ratio. This formula-
tion need not be applied unless the string has reached tion significantly improves the predictions of effective

37
 Another Random Document on
Scribd Without Any Related Topics
 "'The reference to agencies for the uplifting of the drink-
victim would be sadly incomplete without a very definite
acknowledgment of the incalculable assistance which the
wise worker and unprejudiced physician may obtain by
bringing to bear upon the whole life of the patient that
Power, the majesty and mystery, the consolation and
inspiration of which it is the mission of religion to reveal.'"

"Then even the doctors are coming round?" Mary said. "And it
means exactly, you would say—?"

"I would tell you what has been proved without possibility of dispute
a thousand times. I would tell you that when all therapeutic agencies
have failed, the Holy Spirit has succeeded. The Power which is above
every other power can do this. No loving heart need despair.
However black the night that influence can enlighten it. Ask those
who work among the desolate and oppressed; the outcast and
forlorn, the drink-victims and criminals. Ask, here in England, old
General Booth or Prebendary Carlile. Ask the clergy of the Church in
the London Docks, ask the Nonconformist ministers, ask the Priests
of the Italian Mission who work in the slums.

"They will tell you of daily miracles of conversion and

transformations as marvellous and mystical as ever Jesus wrought
when He was visible on earth. Mary! It goes on to-day, it does go
on. There is the only cure, the only salvation. Jesus."

There was a passionate fervour in her voice, a divine light upon her
face. She also prophesied, and the Spirit of God was upon her as
upon the holy women of old.

And Mary caught that holy fire also. Her lips were parted, her eyes
shone. She re-echoed the sacred Name.

"I would give my life to save Gilbert," she said.

"I have no dear one to save, now," the other answered. "But I would
give a thousand lives if I had them to save America from Alcohol. I
love my land! There is much about my country that the ordinary
English man or woman has no glimmering of. Your papers are full of
the extravagances and divorces of wealthy vulgarians—champagne
corks floating on cess-pools. You read of trusts and political
corruption. These are the things that are given prominence by the
English newspapers. But of the deep true heart of America little is
known here. We are not really a race of money-grubbers and cheap
humourists. We are great, we shall be greater. The lamps of freedom
burn clearly in the hearts of millions of people of whom Europe
never hears. God is with us still! The Holy Spirit broods yet over the
forests and the prairies, the mountains and the rivers of my land.
Read the 'Choir Invisible' by James Lane Allen and learn of us who
are America."

"I will, dear Mrs. Daly. How you have comforted me to-night! God
sent you to me. I feel quite happy now about my darling sister. I feel
much happier about my husband. Whatever this life has in store,
there is always the hereafter. It seems very close to-night, the veil
wears thin."

"We will rest, Mary, while these good thoughts and hopes remain
within us. But before we go to bed, listen to this."

Julia Daly felt in her dressing bag and withdrew a small volume
bound in vermilion morocco.

"It's your best English novel," she said, "far and away the greatest—
Charles Reade's 'The Cloister and the Hearth,' I mean. I'm reading it
for the fifth time. For five years now I have done so each year."

"For ever?" she began in her beautiful voice, that voice which had
brought hope to so many weary hearts in the great Republic of the
West.
 "'For ever? Christians live "for ever," and love "for ever" but
they never part "for ever." They part, as part the earth and
sun, only to meet more brightly in a little while. You and I
part here for life. And what is our life? One line in the great
story of the Church, whose son and daughter we are; one
handful in the sand of time, one drop in the ocean of "For
ever." Adieu—for the little moment called "a life!" We part
in trouble, we shall meet in peace; we part in a world of
sin and sorrow, we shall meet where all is purity and love
divine; where no ill passions are, but Christ is, and His
Saints around Him clad in white. There, in the turning of
an hour-glass, in the breaking of a bubble, in the passing
of a cloud, she, and thou, and I shall meet again; and sit
at the feet of angels and archangels, apostles and saints,
and beam like them with joy unspeakable, in the light of
the shadow of God upon His throne, for ever—and ever—
and ever.'"

The two women undressed and said their prayers, making humble
supplication at the Throne of Grace for themselves, those they loved
and for all those from whom God was hidden.

And as the train bore them through Nimes and Arles, Avignon and
the old Roman cities of southern France, they slept as simple
children sleep.
 CHAPTER VI

GILBERT LOTHIAN'S DIARY

 "It comes very glibly off the tongue to say, 'Put yourself in
his position,'—'What would you have done under the
circumstances?' but if self-analysis is difficult, how much
more so is it to appreciate the 'Ego' of another, to
penetrate within the veil of the maimed and debased inner
temple of the debauched inebriate?"—"The Psychology of
the Alcoholic," by T. Claye Shawe, M.D., F.R.C.P., Lecturer
on psychological medicine. St. Bartholomew's Hospital,
London.

"Like one, that on a lonesome road,

Doth walk in fear and dread,
And having once turned round walks on,
And turns no more his head;
Because he knows, a frightful fiend
Doth close behind him tread."

—Coleridge.

When Mary Lothian returned home to Mortland Royal she was very
unwell. The strain of watching over Lady Davidson, and the wrench
of a parting which in this world was to be a final one, proved more
than she was able to endure.

She had been out of doors, imprudently, during that dangerous hour
on the Riviera between sunset and nine o'clock. Symptoms of that
curious light fever, with its sharp nervous pains, which is easily
contracted at such times along the Côte d'Azur, began to show
themselves.

Dr. Morton Sims was away in Paris for a few weeks upon a scientific
engagement he was unable to refuse, and Mary was attended by Dr.
Heywood, the general practitioner from Wordingham.
There was nothing very serious the matter, but the Riviera fever
brings collapse and great depression of spirits with it. Mary remained
in bed, lying there in a dreamy, depressed state of both physical and
mental faculties. She read but little, preferred to be alone as much
as possible, and found it hard to take a lively interest in anything at
all.

Gilbert was attentive enough. He saw that every possible thing was
done for her comfort. But his manner was nervous and staccato,
though he made great efforts at calm. He was assiduous, eager to
help and suggest, but there was no repose about him. In her great
longing for rest and solitude—a necessary physical craving resulting
upon her illness—Mary hardly wanted to see very much even of
Gilbert. She was too weak and dispirited to remonstrate with him,
but it was quite obvious to her experienced eyes that he was
drinking heavily again.

His quite unasked-for references to the fact that he was taking

nothing but a bottle of beer in the middle of the morning, a little
claret at meals and a single whiskey and soda before going to bed,
betrayed him at once. His tremulous anxiety, his furtive manner, the
really horrible arrogation of gaiety and ease made upon a most
anxious hope that he was deceiving her, told their own tale.

So did the heavy puffed face, yellowish red and with spots appearing
upon it. His eyes seemed smaller as the surrounding tissues were
dilated, they were yellowish, streaked with little veins of blood at the
corners, and dull in expression.

His head jerked, his hands trembled and when he touched her they
were hot and damp.

Her depression of mind, her sense of hopelessness, were greatly

increased. Darkness seemed to be closing round her, and prayer—for
it happens thus at times with even the most saintly souls—gave little
relief.
"I shall be better soon," she kept repeating to herself. "The doctor
says so. Then, when I am well, I shall be able to take poor Gillie
really in hand. It won't be long now. Then I will save him with God's
help."

In her present feebleness she knew that it was useless to attempt to

do anything in this direction. So she pretended to believe her
husband, said nothing at all, and prayed earnestly to recover her
health that she might set about the task of succour.

She did not know, had not the very slightest idea, of Lothian's real
state. Nobody knew, nobody could know.

On his part, freed of all restraint, his mind a cave of horror, a

chamber of torture, he drank with lonely and systematic persistence.

It was about this time that he began to make these notes in the
form of a diary which long afterwards passed into the hands of Dr.
Morton Sims. The record of heated horror, the extraordinary glimpse
into an inferno incredible to the sane man, has proved of immense
value to those who are engaged in studying the psychology of the
inebriate.

From much that they contain, it is obvious that the author had no
intention of letting them be seen by any other eyes than his own, at
the time of writing them. Dr. Morton Sims had certainly suggested
the idea in the first place, but there can be no doubt whatever that
Lothian soon abandoned his original plan and wrote for the mere
relief of doing so, and doubtless with a sinister fascination at the
spectacle of his own mind thus revealed by subtle analysis and the
record of a skilled pen. Alcoholised and impaired as his mind was, it
was nevertheless quite capable of doing this accurately and forcibly,
and there are many corroborative instances of such an occurrence.
More than one medical man during the progress of a protracted
death agony has left minute statements of his sensations for the
good of Society.
Such papers as these, for use in a book which has an appeal to all
sorts of people, cannot, of course, be printed entire. There are
things which it would serve no good purpose for the layman to
know, valuable as they are to the patient students of morbid states.
And what can be given is horrible enough.

The selected passages follow herewith, and with only such comment
as is necessary to elucidate the text.

. . . Last night a letter came from a stranger, one of the

many that I get, thanking me for some of the poems in
"Surgit Amari" which he said had greatly solaced and
helped him throughout a period of mental distress. When I
opened the letter it was after dinner, and I had dined well
—my appetite keeps good at any rate, and while that is so
there is no fear of it—according to the doctors and the
medical books. I opened the letter and read it without
much interest. I am not so touched and pleased by these
letters as I used to be. Then, after I had said good-night to
my wife, I went into the library. After two or three whiskies
and a lot of cigarettes the usual delusion of greatness and
power came over me. I know, of course, that I have great
power and am in a way celebrated, but at ordinary times I
have no overmastering consciousness and bland, suave
pride in this. When I am recovering from the effects of too
much alcohol I doubt everything. My own work seems to
me trivial and worthless, void of life and imitations of
greater work.

Well, I had the usual quickening, but vague and incoherent

sense of greatness, and I picked up the letter again. I
walked up and down the room smoking furiously, and then
I had some more whiskey. The constant walking up and
down the room, by the way, is a well-marked symptom of
my state. The nerves refuse me calm. I can't sit down for
long, even with the most alluring book. Some thought
comes into my mind like a stone thrown suddenly into a
pool, and before I am aware of it I am marching up and
down the room like a forest beast in a cage. When I had
read the letter twice more I sat down and wrote a most
effusive reply to my correspondent. I almost wept as I read
it. I went into high things, I revealed myself and my
innermost thoughts with the grave kindness and wish to be
of help that a great and good man; intimate with a lesser
and struggling man; might use.

In the morning I read the letter which I had thought so

wonderful. As usual, I tore it up. It was written in a
handwriting which might have betrayed drunkenness to a
child. Long words lacked a syllable, words ending in "ing"
were concluded by a single stroke, the letter "l" was the
same size as the letter "e" and could not be distinguished
from it. But what was worse, was the sickly sentiment,
expressed in the most feeble sloppy prose.

It was sort of educated Chadband or Stiggins and there

was an appalling lack of reticence.

It is a marked symptom of my state, that when I am drunk

I always want to write effusive letters to strangers or mere
acquaintances. Sometimes, if I have been reading a book
that I liked, I sit down and turn out pages of gush to the
unknown author, hailing him as a brother and a master.
Thank goodness I always tear the wretched things up next
day. It is a good thing I live in the country. In London
these wretched letters, which I am impelled to write,
would be in some adjacent pillar box before I realised what
I had done.
 Oh, to be a sane man, a member of the usual sane army
of the world who never do these things!

The above passage must have been re-read some time after it was
written and been the raison d'être of what follows. The various
passages are only occasionally dated, but their chronological order
can be determined with some certainty by these few dates, changes
of handwriting, and above all by the progress and interplay of
thought.

It had not occurred to me before, with any strength that is,

how very far my inner life diverges now from ordinary
paths! It is, I see in a moment such as the present when I
am able to contemplate it, utterly abnormal. I am glad to
realise this for a time. It is so intensely interesting from the
psychologist's point of view. I can so very, very rarely
realise it. Immediately that I slip back into the abnormal
life, long custom and habit reassert themselves and I
become quite unaware that it is abnormal. I live
mechanically according to the bizarre and fantastic rules
imposed upon me by drink. Now, for a time, I have a
breathing space. I have left the dim green places under
the sea and my head is above water. I see the blue sky
and feel the winds of the upper world upon my face. I
used to belong up there, now I am an inhabitant of the
under world, where the krakens and the polyps batten in
their sleep and no light comes.

I will therefore use my little visit to "glimpse the moon" like

the Prince of Denmark's sepulchral father. I will catalogue
the ritual of the under world which has me fast.

I will, that is, write as much as I can. Before very long my

eyes will be tired and little black specks will dance in front
 of them. The dull pain in my side—cirrhosis of course—
which is quiet and feeding now—will begin again.
Something in my head, at the back of the skull on the left
hand side—so it seems—will begin to throb and ache. Little
shooting pains will come in my knees and round about my
ankles and drops of perspiration which taste bitter as brine
will roll down my face. And, worse than all, the fear of It
will commence. Slight "alcoholic tremors" will hint of what
might be. After a few minutes I shall feel that it is going to
be.

I will define all that I mean by "It" another time.

Well, then I shall send "It" and all the smaller "Its" to the
right about. I shall have two or three strong pegs. Then
physical pains, all mental horrors, will disappear at once.
But I shall be back again under the sea nevertheless. I
shan't realise, as I am realising now, the abnormality of my
life. But I should say that I have an hour at least before I
need have any more whiskey, before that becomes
imperative. So here goes for a revelation more real and
minute than de Quincey, though, lamentable fact! in most
inferior prose!

Here this passage ends. It is obvious from what follows that the
period of expected freedom came to an end long before the author
expected. Excited by what he proposed to do, he had spent too
much of his brief energy in explaining it. Mechanically he had taken
more drink to preserve himself upon the surface—the poisoned mind
entirely forgetting what it had just set down—and with mathematic
certainty the alcohol had plunged the poet once more beneath the
ruining waters.

The next entry, undated, is written in a more precise and firmer

handwriting. It recalls the small and beautiful caligraphy of the old
days. There is no preamble to the bald and hideous confession of
mental torture.

I wish that my imagination was not so horribly acute and

vivid when it is directed towards horrors—as indeed it
always seems to be now. I wish, too, that I had never
talked curiously to loquacious medical friends and read so
many medical books.

I am always making amateur, and probably perfectly

ridiculous, tests for Locomotor Ataxy and General Paralysis
—always shrinking in nameless fear from what so often
seems the inevitable onslaught of "It."

Meanwhile, with these fears never leaving me for a

moment, to what an infinity of mad superstitions I am
slave! How I strive, by a bitter, and (really) hideously
comic, ritual to stave off the inevitable.

Oh, I used to love God and trust in Him. I used to pray to

Jesus. Now, like any aborigine I only seek to ward off evil,
to propitiate the Devil and the Powers of the Air, to drag
the Holy Trinity into a forced compliance with my conjuring
tricks. I can hardly distinguish the devil from God. Both
seem my antagonists. Hardly able to distinguish Light from
dark, I employ myself with dirty little conjuring tricks. I
well know that all these are the phantasms of a disordered
brain! I am not really fool enough to believe that God can
be propitiated or Satan kept at bay by movements:
touchings and charms.

But I obey my demon.

These things are a foolish network round my every action

and thought. I can't get out of the net.
Touching, I do not so much mind. In me it is a symptom of
alcoholism, but greater people have known it as a mere
nervous affection quite apart from drink. Dr. Johnson used
to stop and return to touch lamp-posts. In "Lavengro,"
Borrow has words to say about this impulse—I think it is in
Lavengro or it may be in the Spanish book. Borrow used to
"touch wood." I began it a long time ago, in jest at
something young Ingworth said. I did it as one throws spilt
salt over one's shoulder or avoids seeing the new moon
through glass. Together with the other things I have to do
now, it has become an obsession. I carry little stumps of
pencil in all my pockets. Whenever a thought of coming
evil, a radiation from the awful cloud of Apprehension
comes to me, then I can thrust a finger into the nearest
pocket and touch wood. Only a fortnight ago I was
frightened out of my senses by the thought that I had
never been really touching wood at all. The pencil stumps
were all varnished. I had been touching varnish! It took me
an hour to scrape all the varnish off with a pocket knife. I
must have about twenty stumps in constant use. At night I
always put one in the pocket of my pyjama coat—one
wakes up with some fear—but, half asleep and lying as I
do upon my left side, the pocket is often under me and I
can't get to the wood quickly. So I keep my arm stretched
out all night and my hand can touch the wooden top of a
chair by the bed in a second. I made Tumpany sand-paper
all the varnish off the top of the chair too. He thought I
was mad. I suppose I am, as a matter of fact. But though I
am perfectly aware of the damnable foolishness of it, these
things are more real to me than the money-market to a
business man.
If it were only this compulsion to touch wood I should not
mind. But there are other tyrannies coincident which are
more urgent and compelling. My whole mind—at times—
seems taken up by the necessity for ritual actions. I have
no time for quiet thought. Everything is broken in upon.
There is the Sign of the Cross. I have linked even that in
the chain of my terrors. I touch wood and then I make this
sign. I do it so often that I have invented all sorts of
methods of doing it secretly in public, and quickly when I
am alone. I do it in a sort of imaginary way. For instance, I
bend my head and in so doing draw an imaginary line with
my right eye upon the nearest wall, or upon the page of
the book that I am reading. Then I move my head from
side to side and make another fictitious line to complete
the cross. A propos of making the sign, the imaginary lines
nearly always go crooked in my brain. This especially so
when I am doing it on a book. I follow two lines of type on
both pages and use the seam of the binding between them
to make the down strokes. But it hardly ever comes right
the first time. I begin to notice people looking at me
curiously as I try to get it right and my head moves about.
If they only knew!

Then another and more satisfactory way—for the

imaginary method always makes my head ache for a
second or two—I accomplish with the thumb of my right
hand moving vertically down the first joint of the index
finger, and then laterally. I can do this as often as I like
and no one can possibly see me. I have a little copper
Cross too, with "In hoc vinces" graved upon it. But I don't
like using this much. It is too concrete. It reminds me of
the use I am making of the symbol of salvation. "In hoc
vinces"! Not I. There are times when I think that I am
surely doomed.
But I think that the worst of all the foul, senseless, and yet
imperative petty lordships I endure, is the dominion of the
two numbers. The Dominion of The Two Numbers!—capital
letters shall indicate this! For some reason or other I have
for years imagined mystical virtue in the number 7 and
some maleficent influence in the number 13. These, of
course, are old superstitions, but they, and all the others,
ride me to a weariness of spirit which is near death.

Although I got my first in "Lit. Hum." at Oxford, have read

almost everything, and can certainly say that I am a man
of wide culture and knowledge, Figures always gave me
aversion and distaste. I got an open scholarship at my
college and was as near as nothing ploughed in the almost
formal preliminary exam of Responsions by Arithmetic. I
can't add up my bank-book correctly even now, and I have
no sense whatever of financial amounts and affairs.

But I am a slave to the good but stern fairy 7 and the hell-
hag 13.

I attempt lightness and the picturesque. There is really

nothing of the sort about my unreasoning and mad
servitude. It's bitter, naked, grinning truth.

In my bath I sponge myself seven times—first. Then I

begin again, but I stop at six in the second series and
cross myself upon the breast with the bath sponge. Seven
and six make thirteen. If I did not cancel out that thirteen
by the sign of the Cross I should walk in fear of some
dreadful thing all day.

Every time I drink I sip seven times first and then again
seven times. When six times comes in the second seven, I
make the Cross with my head. My right hand is holding the
glass so that the thumb and finger joint method won't
work. It would be disastrous to make the sign with the left
hand.

That is another thing. . . . I use my left hand as little as I

can. It frightens me. I always raise a glass to my lips with
the right hand. If I use the left hand owing to momentary
thoughtlessness, I have to go through a lengthy
purification of wood-touching, crossing, and counting
numbers.

All my habits re-act one upon the other and the rules are
added to daily until they have become appallingly intricate.
A failure in one piece of ritual entails all sorts of protracted
mental and physical gestures in order to put it right.

I wonder if other men who drink know this heavy,

unceasing slavery which makes the commonest actions of
life a burden?

I suppose so. It must be so. All drugs have specific actions.

Men don't tell, of course. Neither do I! Sometimes, though,
when I have gone to some place like the Café Royal, or
perhaps one of the clubs which are used by fast men, I
have had a disgusting glee when I met men whom I knew
drank heavily to think that they had their secrets—must
have them—as well as I.

On reading through these notes that I have been making

now and then, I am, of course, horrified at what they
really seem to mean. Put down in black and white they
convey—or at least they would convey to anyone who saw
them—nothing but an assurance of the fact that I am mad.
Yet I am not really mad. I have two lives. . . . I see that I
have referred constantly to "It." I have promised myself to
define exactly what I mean by "IT."
I am writing this immediately after lunch. I didn't get up till
eleven o'clock. I am under the influence of twenty-five
grains of ammonium bromide. I had a few oysters for
lunch and nothing else. I am just about as normal as any
man in my state can hope to be.

Nevertheless when I come to try and define "It" for myself

I am conscious of a deep horror and distrust. My head is
above water, I am sane, but so powerful is the influence of
the continual FEAR under which I live my days and nights,
that even now I am afraid.

"It" is a protean thing. More often than not it is a horrible

dread of that Delirium Tremens which I have never had,
but ought to have had long ago. I have read up the
symptoms until I know each one of them. When I am in a
very nervous and excited condition—when, for example, I
could not face anybody at all and must be alone in my
room with my bottle of whiskey—I stare at the wall to see
if rats or serpents are running up it. I peer into the corners
of the library to detect sheeted corpses standing there. I
do not see anything of the sort. Even the imaginings of my
fear cannot create them. I am, possibly, personally immune
from Delirium Tremens, some people are. All the same, the
fear of it racks me and tears me a hundred times a day. If
it really seized me it surely would be almost enjoyable!
Nothing, at any rate, can be more utterly dreadful than the
continual apprehension.

Then I have another and always constant fear—these

fears, I want to insist, are fantastically intermingled with all
the crossings, wood-touchings and frantic calculations I
have to do each minute of my life. The other fear is that of
Prison.
Now I know perfectly well that I have done nothing in my
life that could ever bring me near prison. All the same I
cannot now hear a strange voice without a start of dread.
A knock at the front door of my house unnerves me
horribly. I open the door of whatever room I am in and
listen with strained, furtive attention, slinking back and
closing the door with a sob of relief when I realise that it is
nothing more than the postman or the butcher's boy. I can
hardly bear to read a novel now, because I so constantly
meet with the word "arrest."

"He was arrested in the middle of his conversation,"—"She

placed an arresting hand upon his arm." . . . These
phrases which constantly occur in every book I read fill me
with horror. A wild phantasmagoria of pictures passes
through my mind. I see myself being led out of my house
with gyves upon my wrists like the beastly poem Hood
made upon "Eugene Aram." Then there is the drive into
Wordingham in a cab. All the officials at the station who
know me so well cluster round. I am put into a third class
carriage and the blinds are pulled down. At St. Pancras,
where I am also known, it is worse. The next day there is
the Magistrate's Court and all the papers full of my affair. I
know it is all fantastic nonsense—moonshine, wild dream.
But it is so appallingly real to me that I sometimes long to
have got the trial over and to be sitting with shaven head,
wearing coarse prison clothes, in a lonely cell.

Then, I think to myself, I should really have peace. The

worst would have happened and there would be an end of
it all. There would be an end of deadly Fear.

I remember "——" telling me at Bruges, where so many

mauvais sujets go to kill themselves with alcohol, that
wherever he went, night and day, he was always afraid of
a tiger that would suddenly appear. He had never
experienced Delirium Tremens either. He knew how mad
and fantastic this apprehension was but he was quite
unable to get rid of it.

At other times I have the Folie de Grandeur.

My reading has told me that this is the sure sign of

approaching General Paralysis. General paralysis means
that one's brain goes, that one loses control of one's limbs
and all acts of volition go. One is simply alive, that is all.
One is alive and yet one is fed and pushed about, and put
into this place or that as the entomologist would use a
snail. So, in all my wild imaginings the grisly fear is never
far away.

The imaginings are, in themselves, not without interest to

a student of the dreadful thing I have become.

I always start from one point. That is that I have become

suddenly enormously rich. I have invented all sorts of ways
in which this might happen, but lately, in order to save
trouble, and to have a base to start from I have arranged
that Rockefeller, the American oil person, has been so
intrigued by something that I have written that he presents
me with two million pounds.

I start in the possession of two million pounds. I buy

myself a baronetcy at once and I also purchase some
historic estate. I live the life of the most sporting and
beneficent country gentleman that ever was! I see myself
correcting the bucolic errors of my colleagues on the Bench
at Quarter Sessions. I am a Providence to all the labourers
and small farmers. My name is acclaimed throughout the
county of which I am almost immediately made Lord
Lieutenant.

After about five minutes of this prospect I get heartily sick

of it.

I buy a yacht then. It is as big as an Atlantic liner. I fit it up

and make it the most perfect travelling palace the world
has ever seen. I go off in it to sail round the globe—to see
all the most beautiful things in the world, to suck the last
drop of honey that the beauty of unknown seas, fairy
continents, fortunate islands can yield. During this progress
I am accompanied by charming and beautiful women.
Some are intellectual, some are artistic—all are beautiful
and charming. I, I myself, am the central star around
which all this assiduous charm and loveliness revolve.

Another, and very favourite set of pictures, is the one in

which I receive the two millions from Mr. Rockefeller—or
whoever he is—and immediately make a public
renunciation of it. With wise fore-thought I found great
pensions for underpaid clergy. I inaugurate societies by
means of which authors who could do really artistic work,
but are forced to pot-boil in order to live, may take a
cheque and work out their great thoughts without any
worldly embarrassments. I myself reserve one hundred
and fifty or two hundred pounds a year and go and work
among the poor in an East-end slum. At the same time I
am most anxious that this great renunciation should be
widely spoken of. I must be interviewed in all the papers.
The disdainful nobility of my sacrifice for Christ's sake must
be well advertised.
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com