2007 - Mitteilung 55

Thomas Benz

Small-Strain Stiffness of Soils and

its Numerical Consequences

Thomas Benz

2007 - Mitteilung 55_PDFV2

des Instituts fr Geotechnik
Herausgeber P. A. Vermeer

IGS

Small-Strain Stiffness of Soils

and its Numerical Consequences

Bau und Umweltingenieurwissenschaften

Von der Fakultat fur
der Universitat Stuttgart

zur Erlangung der Wurde

eines Doktors der Ingenieurwissenschaften (Dr.-Ing.)
genehmigte Abhandlung,

vorgelegt von

T HOMAS B ENZ

aus Tubingen

Hauptberichter:

Prof. Dr.-Ing. Pieter A. Vermeer

Mitberichter:

Prof. Guy T. Houlsby, MA DSc FREng FICE

Prof. Dr.-Ing. habil. Christian Miehe

Tag der mundlichen

Prufung:

20. September 2006

Geotechnik der Universitat Stuttgart

Institut fur
2007

Mitteilung 55
Geotechnik
des Instituts fur
Universitat Stuttgart, Germany, 2007

Editor:
Prof. Dr.-Ing. P. A. Vermeer
c
Thomas
Benz
Geotechnik
Institut fur
Universitat Stuttgart
Pfaffenwaldring 35
70569 Stuttgart
All rights reserved. No part of this publication may be reproduced, stored in a
retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, without the permission in
writing of the author.
Keywords: Small-strain stiffness, constitutive soil models
Printed by e.kurz + co, Stuttgart, Germany, 2007
ISBN 978-3-921837-55-9
(D93 - Dissertation, Universitat Stuttgart)

Preface
In the seventies when I began to do research it was realised that soil behaviour was
highly non-linear, but at that time we still underestimated the extent of it. At that
time I should have had the book by Tsytovich (1973)1 , who taught soil mechanics at
the Moscow Civil Engineering Institute. In 1976 this book was translated into English
and I read it some ten years later. In this book Professor Tsytovich introduces a structural soil strength for stress increments as induced by external loads. Up to its structural
strength soil behaves extremely stiff as there is no rearrangement of particles. Once stress
increments exceed this threshold value, soils show the nowadays well-known stressdependent stiffness, as for instance expressed by hyperbolic and/or exponential rules.
In practical analysis of foundation settlements the high initial small-strain stiffness has
always been taken into account by the introduction of a so-called limit depth; below this
depths strains were simply assumed to be negligibly small. Tsytovich does not use the
name limit depth, but calls it depth of the active compression zone. In finite element
calculations I always accounted for this limit depth by using relatively shallow meshes.
From a mathematical point of view the subsoil should obviously be modelled by infinite
boundaries, but because of the small-strain stiffness only relatively shallow finite layer
had to be discretised. On using the new constitutive model by Dr. Thomas Benz this
dilemma on the depth of finite element meshes is non-existent. Boundaries can now
be chosen so for away that they do not influence the computed stress distribution and
the constitutive model ensures that the deep part of the mesh behaves virtually incompressible. The new HS-Small model is obviously not only relevant for the analysis of
foundations, but also for the analysis of settlements due to tunnelling or deep excavations, as shown in this dissertation study.
The focus of this study is on the small-strain stiffness of non-cemented soils, but the
model may also be applied to somewhat bonded soils. In recent years the behaviour of
sensitive and/or structured clays has got much attention from researches and it would
seem that their findings can also be modelled by the new HS-small model. In order to
do so one may have to choose relatively large values for the strain range parameter 0.7 ,
as used in the HS-Small model.
This thesis on a new computational model for the behaviour of soils at small strains
is a scholarly piece of work, covering a very wide scope of material. The new model
is developed, and calibrated, and in my opinion extremely useful for engineering problems. I am also indebted to the Federal Waterways Engineering and Research Institute
in Karlsruhe who have sponsored this study. Within this institute Dr. Radu Schwab has
advised Thomas Benz and it has been a real pleasure for me to work with them.
Pieter A. Vermeer
Stuttgart, March 2007

N. Tsytovich. Soil Mechanics. Mir Publishers, Moscow, 1976. (translation from the 1973 Russian edition)

Acknowledgments
This Dissertation would not have been completed without the support and the patience
of Professor Pieter Vermeer and Radu Schwab. I thank Pieter for giving me the opportunity to join his research team and for his extensive help. Likewise, I thank Radu for the
unlimited help, advice and support he provided at the Federal Waterways Engineering
and Research Institute (BAW). This Dissertation emanated from a research project that
was initiated and funded by the BAW.
I thank all colleagues at the BAW in Karlsruhe and at the Institute of Geotechnical
Engineering in Stuttgart for their friendly help and for the pleasant years. In particular,
I am grateful to the following persons (in random order) for the exchange of thoughts,
critical reviews, beta testing, proof reading, etc: Markus Herten, Marcin Cudny, Markus

Wehnert, Sven Moller,

Ayman Abed, Martino Leoni, Heiko Neher, Josef Hintner, Annette Lachler, Andrej Mey, Regina Kauther, Oliver Stelzer, Bernd Zweschper, Florian
Scharinger, Professor Helmut Schweiger, Professor Dimitrios Kolymbas, Professor Christian Miehe, Professor Guy Houlsby and Grainne McCloskey. Further, I thank Paul Bonnier from Plaxis B.V. for his valuable assistance with the numerical implementation.
I am very grateful to my family, in particular to my parents and to Marlis. Thanks to
their support and love I was able to actually write this Dissertation.
Thomas Benz
Stuttgart, March 2007

Contents

Introduction

Terminology and definitions

Experimental evidence for small-strain stiffness

3.1 Small-strain stiffness measurements . . . . . . . . . . . . . . . . . . . . .
3.1.1 Laboratory tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.1.1 Local measurements . . . . . . . . . . . . . . . . . . . . .
3.1.1.2 Bender elements . . . . . . . . . . . . . . . . . . . . . . .
3.1.1.3 Resonant column and torsional shear . . . . . . . . . . .
3.1.2 In-situ tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.2.1 Cross hole seismic . . . . . . . . . . . . . . . . . . . . . .
3.1.2.2 Down hole seismics . . . . . . . . . . . . . . . . . . . . .
3.1.2.3 Suspension logging . . . . . . . . . . . . . . . . . . . . .
3.1.2.4 Seismic cone and seismic flat dilatometer . . . . . . . . .
3.1.2.5 Spectral Analysis of Surface Waves (SASW) . . . . . . .
3.2 Parameters that affect small-strain stiffness . . . . . . . . . . . . . . . . .
3.2.1 The influence of shear strains and volumetric strain . . . . . . . .
3.2.2 The influence of confining stress . . . . . . . . . . . . . . . . . . .
3.2.3 The influence of void ratio . . . . . . . . . . . . . . . . . . . . . . .
3.2.4 The influence of soil plasticity . . . . . . . . . . . . . . . . . . . . .
3.2.5 The influence of the overconsolidation ratio (OCR) . . . . . . . .
3.2.6 The influence of diagenesis . . . . . . . . . . . . . . . . . . . . . .
3.2.7 The influence of loading history . . . . . . . . . . . . . . . . . . .
3.2.8 The influence of strain rate and inertia effects . . . . . . . . . . . .
3.2.9 Other factors that influence small-strain stiffness . . . . . . . . . .
3.3 Correlations for small-strain stiffness . . . . . . . . . . . . . . . . . . . . .
3.3.1 Correlations between G0 and e, p0 and OCR . . . . . . . . . . . . .
3.3.2 Correlations between G0 and CPT, SPT and cu . . . . . . . . . . .
3.3.2.1 Cone penetration (CPT) and standard penetration (SPT)
3.3.2.2 Estimating G0 from conventional tests - Chart by Alpan
3.3.3 Correlations for the stiffness modulus reduction curve . . . . . .
3.3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

9
10
12
12
12
13
14
14
15
15
15
15
16
16
19
21
23
23
24
25
28
30
31
31
34
34
35
35
38

Contents
4

ii

Small-strain stiffness at the soil particle level

4.1 Soil fabric and soil structure . . . . . . . . . . . . . . . .
4.2 Micromechanical considerations . . . . . . . . . . . . .
4.2.1 The influence of strain amplitude . . . . . . . . .
4.2.2 The influence of confining stress and void ratio
4.2.3 The influence of cementation . . . . . . . . . . .
4.2.4 Summary . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

Existing small-strain stiffness constitutive models

5.1 A brief history of small strain modeling . . . . . . . . . . . .
5.2 The Simpson Brick model . . . . . . . . . . . . . . . . . . . .
5.3 Models known from soil dynamics . . . . . . . . . . . . . . .
5.4 The Jardine model . . . . . . . . . . . . . . . . . . . . . . . . .
5.5 Multi (or Infinite) surface models . . . . . . . . . . . . . . . .
5.6 Intergranular Strain for the Hypoplastic model . . . . . . . .
5.7 A critical review of small-strain stiffness models regarding
routine design . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Small-Strain Overlay model
6.1 Model formulation . . . . . . . . . . . . . . . . . .
6.1.1 Material history and history mapping . . .
6.1.2 From strain history to stiffness . . . . . . .
6.1.3 Inital loading and reloading . . . . . . . . .
6.1.4 Effects of mean stress, void ratio, and OCR
6.2 A first model validation in element tests . . . . . .
6.2.1 Triaxial tests . . . . . . . . . . . . . . . . . .
6.2.2 Biaxial test . . . . . . . . . . . . . . . . . . .
6.2.3 Strain response envelopes . . . . . . . . . .
6.2.4 Cyclic mobility . . . . . . . . . . . . . . . .
6.3 Thermodynamic considerations . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
their use in
. . . . . . . .
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

HS-Small, a small-strain extension of the Hardening Soil model

7.1 Constitutive relations for infinitesimal plasticity . . . . . . . . .
7.2 The Hardening Soil model . . . . . . . . . . . . . . . . . . . . . .
7.3 The HS-Small model . . . . . . . . . . . . . . . . . . . . . . . . .
7.3.1 Small-Strain formulation . . . . . . . . . . . . . . . . . . .
7.3.2 Mobilized dilatancy . . . . . . . . . . . . . . . . . . . . .
7.3.3 Yield surface and plastic potential - HS-Small(MN) only
7.3.4 Generalized model formulation - HS-Small(MN) only . .
7.4 Local integration of the constitutive equations . . . . . . . . . .
7.4.1 Return mapping onto the yield surface . . . . . . . . . .
7.4.2 The closest point projection algorithm . . . . . . . . . . .
7.4.3 The consistent algorithmic tangent stiffness tensor . . . .
7.4.4 Corner and apex problems . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

39
39
41
42
44
46
47
49
49
50
51
53
55
56
58

.
.
.
.
.
.
.
.
.
.
.

59
59
59
62
63
65
66
66
66
69
72
72

.
.
.
.
.
.
.
.
.
.
.
.

75
75
77
82
83
85
86
91
94
97
98
100
102

Contents
8

Validation and verification of the HS-Small model

8.1 Element tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2 Boundary value problems . . . . . . . . . . . . . . . . . . . . . .
8.2.1 Tunnels . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.1.1 Steinhaldenfeld NATM tunnel . . . . . . . . . .
8.2.1.2 Heinenoord Tunnel . . . . . . . . . . . . . . . .
8.2.2 Excavations . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.2.1 Excavation in Berlin sand . . . . . . . . . . . . .
8.2.2.2 Excavation in Ruple clay (Offenbach) . . . . . .
8.2.3 Spread Foundations . . . . . . . . . . . . . . . . . . . . .
8.2.3.1 Texas A&M University spread footing on sand
8.2.4 Initialization of the HS-Small model . . . . . . . . . . . .
8.2.5 Capabilities and limitations of the HS-Small model . . .
3D case study Sulfeld

Lock

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

105
105
115
118
118
120
121
123
126
130
130
134
136
139

10 Conclusions

147

Bibliography

150

A Small-Strain Overly Fortran Code

167

B Return mapping in Fortran code

173

C Two surface return strategy

177

D Material data

179

iii

Abstract
This thesis is concerned with the stiffness of soils at small strains and constitutive models that can be applied to simulate this. The strain range in which soils can be considered
truly elastic, i.e. where they recover from applied straining almost completely, is very
small. With increasing strain amplitude, soil stiffness decays non-linearly: Plotting soil
stiffness against log(strain) yields characteristic S-shaped stiffness reduction curves. At
the minimum strain which can be reliably measured in classical laboratory tests, i.e. triaxial tests and oedometer tests without special instrumentation, soil stiffness is often
less than half its initial value. If this non-linear variation of soil stiffness at small strains
is considered in the analysis of soil-structure interaction, analysis results improve considerably: The width and shape of settlement troughs are more accurately modeled,
excavation heave is reduced to a more realistic value, etc. The importance of smallstrain stiffness in engineering practice is therefore well recognized. As the few existing
small-strain stiffness models are either research orientated, and/or applicable to specific
loading paths only, the objective of this thesis is to develop a small-strain stiffness model
for the engineering community.
The newly developed Small-Strain Overlay model is based on the Hardin-Drnevich
model [52]. Similar to the Hardin-Drnevich model, the Small-Strain Overlay model has
two input parameters. Both input parameters have a clear physical meaning. For strains
below the limit of classical laboratory testing, the Small-Strain Overlay model calculates
the actual strain dependent stiffness. This stiffness can subsequently be deployed in
many existing elastoplastic constitutive models. One of them is the well known Hardening Soil (HS) model as implemented in the finite element code P LAXIS V8 [21]. In this
thesis, the Small-Strain Overlay model is combined with the HS model. The resulting
HS-Small model is additionally enhanced by a Matsuoka-Nakai failure criterion. The
HS-Small model is validated in element tests and various 2D and 3D boundary value
problems. In the validation sequence, all effects that are commonly attributed to smallstrain stiffness in soil-structure interaction can be recognized.
Although the thesis main concern is with the development of a small-strain stiffness
model, it also provides much information and data that may be helpful in using it. This
information and data comprise an introduction to available small-strain stiffness testing
methods, many empirical small-strain stiffness correlations and sample model applications.
After a short introduction to the topic in Chapter 1 and clarification of important terminology in Chapter 2, the thesis is organized thematically in three parts:

Abstract
Tests and correlations - Chapter 3 Firstly, small-strain stiffness is discussed from
an experimental point of view. The most important small-strain stiffness testing
methods are introduced. Different parameters that affect small-strain stiffness are
identified. The influence of the most significant of these parameters, regarding
small-strain stiffness, is quantified in a collection of empirical correlations. In the
absence of experimental data, these correlations may prove very helpful in using
the new model.
Model building and formulation - Chapter 4 to 7 Secondly, the new Small-Strain
Overlay model and its combination with the HS model is formulated. This main
part of the thesis also contains a critical review of existing small-strain models.
Important numerical aspects, for example the local integration of the incrementally
formulated HS-Small equations, are presented at the end of the model building and
formulation section.
Code verification and model validation - Chapter 8 to 9 Finally, the HS-Small
model is validated in element tests, several 2D boundary value problems, and a
state-of-the-art 3D boundary value problem. Initialization of the HS-Small model
and a new possibility for verifying the proper extent of boundary conditions are
discussed within the validation section as well.
Results from the study presented in this thesis are concluded in Chapter 10.

vi

Zusammenfassung
Die vorliegende Arbeit beschaftigt sich mit der Steifigkeit von Geomaterialien unter
kleinen Dehnungen und deren Beschreibung in Stoffmodellen. Der Dehnungsbereich, in

dem Geomaterialien als elastisch angesehen werden konnen

- in dem sich also nach einer

Entlastung der ursprungliche Dehnungszustand nahezu wieder einstellt - ist sehr klein.
Auerhalb dieses sehr kleinen Bereiches zeigen Versuche eine nicht lineare Steifigkeits
abnahme mit zunehmender Dehnung. Aufgetragen uber
den Logarithmus der aufge die Abnahmefunktion der Steifigkeit eine charakterbrachten Dehnung, ergibt sich fur

istische S-formige
Kurve. Bereits bei der kleinsten Dehnung, die noch zuverlassig in
klassischen Laborversuchen (d.h. Dreiaxiale Versuche und Oedometer Versuche ohne
spezielle Instrumentierung) gemessen werden kann, hat sich die Steifigkeit, bezogen

auf ihren Ausgangswert, oftmals schon um mehr als die Halfte reduziert. Bei Berucksichtigung dieses nicht-linearen Verhaltens in Boden-Bauwerk Interaktionsproblemen
sich die Genauigkeit von berechneten Setzungen und Verschiebungen erheblich:
erhoht
Die laterale Ausdehnung und Form von Setzungsmulden wird erheblich besser abgebildet, Hebungen in Baugrubensohlen realistischer berechnet, etc. Die Bedeutung von
nicht-linearem Bodenverhalten bei kleinen Dehnungen wird daher auch in der praktischen geotechnischen Entwurfs- und Bemessungsarbeit allgemein als hoch eingeschatzt.

Da die wenigen bisher verfugbaren

Stoffmodelle zur Abbildung dieses Verhaltens, aber
vorgegebene Belastungsarten anwendhauptsachlich forschungsorientiert, oder nur fur
bar sind, wird in der vorliegenden Arbeit ein neues, praxisorientiertes Modell entwickelt.
Das neu entwickelte Small-Strain Overlay Model basiert auf einem hyperbolischen
Ansatz nach Hardin und Drnevich [52]. In Analogie zu diesem hat das neu entwickelte
Modell lediglich zwei Eingabeparameter. Beiden Parametern ist eine klare physikalische
Dehnungen unterhalb der Genauigkeitsgrenze von klassisBedeutung zugeordnet. Fur
chen Laborversuchen ermittelt das Small-Strain Overlay Modell eine dehnungsabhangige, isotrope Steifigkeit, die dann als quasi-elastische Steifigkeit in einem beliebigen
elastoplastischen Stoffmodell weiter verwendet werden kann. Ein solches elastoplastisches Modell ist z.B. das Hardening Soil Modell, welches in das FE Programm P LAXIS
V8 [21] implementiert ist. Mit diesem wird das Small-Strain Overlay Modell in der vorliegenden Arbeit kombiniert. Die als HS-Small Modell bezeichnete Kombination wird
zusatzlich durch ein optionales Matsuoka-Nakai Versagenskriterium erweitert. Das HSSmall Modell wird in einer Reihe von Element-Tests und einer Vielzahl von 2D und

3D Randwertproblemen validiert. In den Randwertproblemen konnen

im Vergleich
mit dem HS Modell, all die Phanomene beobachtet werden, die in der Regel der nichtlinearen Abnahme der Steifigkeit unter kleinen Dehnungen zugeschrieben werden.

vii

Zusammenfassung
Obwohl das primare Ziel der vorliegenden Arbeit die Entwicklung eines Stoffmodells
ist, enthalt diese auch zahlreiche Informationen und Daten, die sich bei der Nutzung

des neuen Modells als hilfreich herausstellen konnen.

Diese Informationen und Daten

bestehen unter anderem aus einer Ubersicht uber

vorhandene Testverfahren zur Ermittlung der Bodensteifigkeit unter kleinen Dehnungen und einer Sammlung empirischer
Korrelationen.

Nach einer kurzen Einfuhrung

in das Thema in Kapitel 1 und einiger Begriffsklarungen in Kapitel 2, untergliedert sich die Arbeit in drei thematische Teile:
Versuche und Korrelationen - Kapitel 3
Zunachst wird die Bodensteifigkeit unter kleinen Dehnungen unter einem experimentellen Gesichtspunkt betrachtet. Zum einen werden die wichtigsten Versuche
zu deren Bestimmung vorgestellt, zum anderen wird der Einfluss verschiedener
Faktoren, wie z.B. bodenmechanische Parameter und Belastungsgeschichte, auf

die Versuchsergebnisse untersucht. Diese Untersuchung mundet

in einer Sammlung empirischer Korrelationen, welche insbesondere dann hilfreich sind, wenn
das Small-Strain Overlay Modell (oder des HS-Small Modell) angewendet werden
soll und keine Versuchsdaten vorliegen.
Modellbildung und Formulierung - Kapitel 4 bis 7
Im zweiten und umfangreichsten Teil dieser Arbeit wird das Small-Strain Overlay Model und das HS-Small Model entwickelt. Neben den konstituierenden Gle

ichungen umfasst dieser Teil auch einen kritischen Uberblick

uber
bereits vorhan
dene Stoffgesetze, die Steifigkeit von Boden
unter kleinen Dehnungen prinzip
iell beschreiben konnen.
Numerische Aspekte, wie z.B. die Integration der inkrementell formulierten konstitutiven Gleichungen werden am Ende des zweiten Teils
der Arbeit behandelt.
Kontrolle der Implementierung und Modell Validierung - Kapitel 8 bis 9
Letztlich wird das HS-Small Modell in einer Reihe von Elementversuchen sowie
einiger 2D Randwertprobleme und eines umfangreichen 3D Randwertproblems

und validiert. An Hand der Randwertprobleme wird des Weiteren die

uberpr
uft

Moglichkeiten
der Modellinitialisierung diskutiert sowie eine Bewertungsmethode
die Randabstande von Berechnungsausschnitten vorgeschlagen.
fur
Die Schlufolgerungen aus der vorliegenden Arbeit sind in Kapitel 10 zusammengefat.

viii

Chapter 1
Introduction
One of the major problems in ground engineering in the 1970s and earlier was the
apparent difference between the stiffness of soils measured in laboratory tests and those
back-calculated from observations of ground movements (e.g. Cole & Burland [29], St
John [84], Wroth [195], Burland [23]). These differences have now largely been reconciled
through the understanding of the principal features of soil stiffness and, in particular,
the very important influence of non-linearity. This is one of the major achievements of
geotechnical engineering research over the past 30 years. (Atkinson [7])
In particular the non-linear influence of strain on soil stiffness has been extensively investigated over the past decades. The maximum strain at which soils exhibit almost fully
recoverable behavior is found to be very small. The very small-strain stiffness associated
with this strain range, i.e. shear strains s 1 106 , is believed to be a fundamental
property of all types of geotechnical materials including clays, silts, sands, gravels, and
rocks (Tatsuoka et al., 2001) under static and dynamic loading (Burland [24]) and for
drained and undrained loading conditions (Lo Presti et al. [140]). With increasing strain,
soil stiffness decays non-linearly. On a logarithmic scale, stiffness reduction curves exhibit a characteristic S-shape, see Figure 1.1.
The smallest shear strain that can be reliably measured in conventional soil testing,
e.g. triaxial or oedometer tests without special instrumentation, is s 1 103 . By
definition Atkinson [7] terms strains smaller than the limit of classical laboratory testing
(s < 1 103 ), small strains. Strains, s > 1 103 are termed large or larger strains. The
limit of classical laboratory testing coincides at the same time with characteristic shear
strains that can be measured near geotechnical structures (Figure 1.1). However, the soil
stiffness that should be used in the analysis of geotechnical structures is not the one that
relates to these final strains. Instead, very small-strain soil stiffness and its non-linear
dependency on strain amplitude should be properly taken into account in all analysis
that strive for reliable predictions of displacements. The Rankine lectures by Simpson
[168] and Atkinson [7], or the Bjerrum Memorial lecture by Burland [24] are just a few
occasions where this has been highlighted. As yet, small-strain stiffness has not been
widely implemented in engineering practice. Considering numerical analysis, this may
be due to a lack of capable, yet user-friendly constitutive models. The main objective
of this thesis is to provide such a capable and sufficiently simple small-strain stiffness
model for engineering practice.
The user-friendliness of a constitutive model largely depends on the input parameters. They should be limited in their number, easy to understand in their physical meaning, and easy to quantify based on test data or experience. The level of sophistication

Chapter 1 Introduction

Shear modulus G/G0 [-]

Retaining walls
Foundations
Tunnels

Very
small
strains

Conventional soil testing

Small strains

Larger strains
0
-6

10

10

-5

10

-4

10

-3

10

-2

10

-1

Shear strain gS [-]

Dynamic methods
Local gauges

Figure 1.1: Characteristic stiffness-strain behavior of soil with typical strain ranges for
laboratory tests and structures (after Atkinson & Sallfors [9] and Mair [109])

of a model is tied to the extent to which it is able to reproduce the experimentally observed functional relationships. The premise of fulfilling the objective defined above is
therefore a thorough description of experimental observations. It is only through this
that the models parameters and mechanisms can be decided upon, or in the words of
Ernst Mach: The description of functional relationships is an explanation in itself. The
methodology applied in this thesis is mostly inductive reasoning. Figure 1.2 explains the
terminology used to describe the different stages within the model building and validation process. The model is formulated incrementally. Therefore the code verification and
model validation ar regarded as separate processes.
Although the main objective of this thesis is the development of a simple and capable small-strain stiffness constitutive model, it also acts as a compilation of information
and data that might be helpful in using it. This information and data comprise empirical correlations that can be useful in quantifying small-strain stiffness, an introduction
to available testing methods and sample applications in 2- and 3-dimensional finite element analysis.
The outline of this thesis is as follows:
Chapter 2 introduces some definitions and conventions used throughout the thesis.
Chapter 3 concentrates on the experimental aspects and the quantitative description of small-strain stiffness. After introducing the experimental concepts of laboratory and in-situ testing, the influence of various parameters on small-strain stiffness is studied. Readily available test data and correlations from literature are
presented at the end of this chapter.

Experiment
Model
Validation

Simulation

Numerical Model

Abstraction

Programming

Model
Building

Model

Code Verification

Figure 1.2: Terminology used in the model building and validation process.

Chapter 4 looks at the small-strain stiffness phenomena at a micromechanical level.

This deductive part of the thesis is mainly aimed at providing a more fundamental
explanation of some observations made in Chapter 3.
Chapter 5 summarizes the best known existing small-strain stiffness models. These
range from simple 1D models in stress or strain space to complex formulations, for
example the Intergranular Strain [129] concept. All models are briefly appraised
regarding their value in practical applications.
Chapter 6 introduces a new small-strain stiffness model, the Small-Strain Overlay
model. The Small-Strain Overlay model is a straight-forward model that is largely
based on the well known Hardin-Drnevich relation [52], which is introduced in
Chapter 5. Its dependency on the materials strain history, however, is formulated
in multi-axial strain space. A first model validation is undertaken in a number of
element tests on sandy soils and clays.
In Chapter 7 the Small-Strain Overlay model is combined with an existing elastoplastic model. The elastoplastic model chosen for this purpose is the Hardening
Soil (HS) model as implemented in the finite element code P LAXIS V8. The resulting HS-Small model is further enhanced with the Matsuoka-Nakai [114] failure
criterion and a modified flow rule. A possible implicit integration algorithm for
the new model is presented at the end of the chapter where other numerical issues
are briefly discussed as well.
Chapter 8 covers the verification process of the HS-Small model. This includes
the evaluation of numerous element tests and boundary value problems from the
following problem categories: deep excavations, tunnels, and foundations. In analyzing the boundary value problems, the impact of the HS-Small model on the

Chapter 1 Introduction
different problem categories is quantified. More general issues, as for example
possible initialization procedures for the HS-Small model, are discussed at the end
of this chapter.

Chapter 9 contains a case study of the large navigable lock, Sulfeld.

The lock

Sulfeld
is currently being reconstructed right next to the high-speed railway link
between Hannover and Berlin. A state of the art 3D finite element analysis of the
deep excavation is conducted in order to predict the construction activitys influence on the railway tracks. It is shown how the HS-Small model could significantly
enhance the reliability of this analysis.
Finally, Chapter 10 presents the main conclusions of this study.

Chapter 2
Terminology and definitions
This thesis is mainly concerned with pre-failure deformation characteristics of soils.
Continuum mechanics is generally employed as the analysis tool. In soil mechanics
however, there is not a clear consensus in how to define some stress and strain properties in this framework (e.g. sign convention). This chapter briefly introduces the stress
and strain definitions, notations, and invariants used in the following chapters.
Tensor notation Tensorial quantities are generally expressed in indicial notation.
The order of a tensor is indicated by the number of unrepeated (free) subscripts.
Whenever a subscript appears exactly twice in a product, that subscript will take
on the values 1, 2, 3 successively, and the resulting terms are summed (Einsteins
summation convention). The Kronecker delta ij takes the value 1 or 0 if i = j or
i 6= j respectively. In the rare occasions where indicial notation is not used in this
thesis, first-order tensors (vectors) are denoted by lowercase Latin or Greek bold
letters, second-order tensors are denoted by uppercase Latin or Greek bold letters,
and fourth-order tensors are denoted by Calligraphic letters.
Stress and strain Infinitesimal deformation theory is applied. Cauchy stress is
related to linearized infinitesimal strain. Eigenvalues of stress and strain tensors
(principal stresses and strains) are denoted by one subscript only, e.g. i and i
with i = 1, 2, 3. Without loss of generality, let
1 2 3 .

(2.1)

The Roscoe stress invariants p (mean stress) and q (deviatoric stress), are defined
as:
r
ii
3
1
1
p=
and
q=
(ij ij kk )(ij ij kk ),
(2.2)
3
2
3
3
In triaxial compression with 1 2 = 3 , the Roscoe invariants simplify to
1
(axial + 2lateral )
3
q = (axial lateral ).

p =

(2.3)
(2.4)

In analogy to the stress invariants, volumetric strain v and shear strain s are defined as:
r
1
1
3
v = ii
and
s =
(ij ij kk )(ij ij kk ),
(2.5)
2
3
3

Chapter 2 Terminology and definitions

q

ETan(A)

E0

E0
ESec(A)
1

E0
eaxial

Figure 2.1: Definition of secant and tangent moduli in triaxial stress-strain space.

which simplify to
v = axial + 2lateral
s = (axial lateral ),

(2.6)
(2.7)

in triaxial compression. Shear strain relates to the deviatoric strain invariant q as

follows:
r
3
3 2
1
1
s = q =
(ij ij kk )(ij ij kk ).
(2.8)
2
2 3
3
3
Other invariants of the stress tensor are later introduced in Chapter 7, where they
are deployed to describe yield criteria.
Sign convention The sign convention of soil mechanics is used: Compressive stress
and strain is taken as positive. Tensile stress and strain is taken as negative.
Total and effective quantities Effective stress is generally denoted by a prime, e.g.
p0 . An exception to this rule is made in Chapter 7 where effective stresses are considered exclusively. Friction angle and cohesion are always taken to be effective
values without any special indication by a prime.
Tangent and secant moduli Tangent stiffness is used in numerical calculations. Secant stiffness is generally used to describe experimental results. Figure 2.1 depicts
the definition and difference of tangent and secant stiffness moduli.
Very small, small, and larger strains Very small strains are defined as strains at
which a soil exhibits almost fully recoverable, or almost elastic behavior. Considering that there can be no true elasticity in soils, the very small strain range is difficult
to quantify. For the purpose of this thesis, the transition between very small strains
and small strains is assumed in the range 1 106 s 1 105 . The borderline
between small strains and larger strains is commonly drawn at the limit of classical

laboratory testing, which is according to Atkinson [7] s 1 103 (see Figure 1.1).
In this thesis however, the limit of classical laboratory testing is taken as the shear
strain where the stiffness modulus reduction curve equals the unloading-reloading
stiffness observed in classical laboratory testing.
Small-strain stiffness Small-strain stiffness is the stiffness of soils at small shear
strains. The maximum soil stiffness found in stiffness reduction curves is in this
thesis referred to as initial, or maximum soil stiffness. Initial shear and Youngs
moduli are denoted as G0 and E0 respectively. Often, initial soil stiffness is also
called very small-strain stiffness, as it is commonly associated to the very-smallstrain range as defined above. In this thesis, small-strain stiffness is always discussed along with the respective stiffness reduction curve in the range of very
small and small strains. The term small-strain stiffness model consequently refers
in this thesis to a constitutive soil model that is applicable to very small and small
strains; Small-strain stiffness measurements include stiffness measurements at very
small and small strains; etc.

Chapter 3
Experimental evidence for small-strain stiffness
From dynamic response analysis, it has been found that most soils have curvilinear
stress-strain relationships as shown in Figure 3.1. The shear modulus is usually expressed as the secant modulus determined by the extreme points on the hysteresis loop
while the damping factor is proportional to the area inside the hysteresis loop. It is
readily apparent, that each of this properties will depend on the magnitude of the strain
for which the hysteresis loop is determined and thus, both shear moduli and damping
factors must be determined as functions of the induced strain in a soil specimen or soil
deposit. (Seed & Idriss [162])
In soil dynamics, small-strain stiffness has been a well known phenomena for a long
time. The above conclusions by Seed & Idris for example were drawn more than 35
years ago. In static analysis however, the findings from soil dynamics have long been
disregarded, or even considered not to be applicable. Seemingly differences between
static and dynamic soil stiffness have been attributed to the nature of loading (e.g. inertia
forces and strain rate effects) rather than to the magnitude of applied strain, which is
generally small in dynamic conditions (earthquakes excluded).
Nowadays, it is commonly accepted that inertia forces and strain rate have little influStress
G1

e1

G2
1

e2

Strain

Figure 3.1: Stress-strain hysteresis loops presented by Seed & Idris [162]. The secant
shear modulus of a closed stress-strain loop decays monotonically with its
strain amplitude: G2 < G1 for 2 > 1 .

Chapter 3 Experimental evidence for small-strain stiffness

ence on small-strain stiffness. Experimental evidence for the influence of strain rate, but
also other parameters that more significantly affect small-strain stiffness is presented in
this chapter. This chapter is exclusively dedicated to the experimental aspects of smallstrain stiffness. It provides a brief summary of laboratory and in-situ small-strain stiffness testing methods and a discussion of quantitative experimental findings.

3.1 Small-strain stiffness measurements

Small-strain soil stiffness can be measured in laboratory and/or field tests. Applicable
laboratory tests are triaxial tests with local strain measurements (e.g. Jardine et al. [80]);
bender elements (e.g. Shirley & Hampton [165], Dyvik & Madshus [37], Brignoli et al.
[20]); resonant column (e.g. Hardin & Drnevich [53]); and torsional shear tests. Resonant
column and torsional shear devices are both available for hollow cylindrical samples, as
well (Broms & Casbarian [22]). Field or in-situ tests for indirect identification of very
small-strain stiffness generally rely on geophysical principles: cross hole seismic (Stokoe
& Woods [175]), down hole seismic (e.g. Woods [194]), suspension logging (e.g. Nigbor
& Imai, [130]), seismic cone (e.g. Robertson et al. [150]), seismic flat dilatometer (e.g.
Hepton [56]), and spectral analysis of surface waves (e.g. Stokoe et al. [174]) are most
commonly used.
Surface based seismic reflection and refraction methods are nowadays less often used
to acquire very small-strain stiffness data. The impedance contrasts in shallow soil layers
are often not pronounced enough to allow these methods to compete with the aforementioned in-situ tests in terms of data resolution and reliability. From a historical perspective however, these were the first geophysical methods, which Engineers and Geologists
used for subsurface characterization, from the beginning of the 20th century. Though it
was not until the 1960s and 1970s, that geophysics also became popular in geotechnical
engineering. Back then, borehole seismic techniques were first employed for site characterization. The next major development of seismic techniques for use in geotechnical
engineering occured in 1984 when the seismic cone penetration test was presented. The
latest development is the spectral analysis of surface waves, which took off at the end of
the 1990s.
Some of the laboratory tests and all of the in-situ tests mentioned above are indirect
tests. In contrast to direct tests, indirect tests measure other quantities than those desired
and relate them through mathematical relationships. Seismic techniques, as indirect testing methods for very small-strain stiffness, yield wave propagation velocity profiles.
Assuming linear elastic material behavior, elastic stiffness relates to wave propagation
velocity as follows:
s
vp =
s
vs =

10

+ 2G

and

(3.1)
(3.2)

3.1 Small-strain stiffness measurements

P-wave (compression wave)

Travel direction

S-wave (shear wave)

Travel direction

Figure 3.2: P-wave (top) and S-wave (bottom) particle motion: The particle motion in Pwaves is longitudinal whereas it is transverse in S-waves. The particle motion
vector in the plane perpendicular to the direction of propagation, is referred
to as polarization (only S-waves).

where vp is the propagation velocity of pressure, or primary (P-) waves, vs is the propagation velocity of shear or secondary (S-) waves, and , and G are Lames constants (G
is also termed the shear modulus). The expression of Lames constants as a function of
Youngs modulus and Poisons ratio is generally more convenient for engineers:
E
2(1 + )
E
=
.
(1 + )(1 2)

G =

(3.3)
(3.4)

Figure 3.2 illustrates the particle motion in P- and S-waves.

Seismic techniques are said to belong to the class of dynamic tests because they operate at higher frequencies than static, or quasi-static tests as for example triaxial tests with
local strain gauges. In static and quasi-static tests, effects that are due to inertia forces
are considered negligible. It is common practice to distinguish between quasi-static and
dynamic tests based on the frequency of loading. The AK 1.4 of the German Geotechnical Association (DGGT) for example, suggests to consider loading frequencies below 10
Hz as quasi-static [156]. Unfortunately, this definition is useless in monotonic loading
situations, e.g. impact situations. In repeated loading its shortcoming is that it does not
consider load amplitudes. Assuming a harmonic excitation of the form a = amax sin t,
where amax is the amplitude of displacement, stress gradients due to inertia effects are
proportional to amax 2 . Only when these stress gradients are small compared to the
overall stress level divided by some characteristic length, one can assume quasi-static
loading. Stress gradients can in this way serve as an indicator for quasi-static or dynamic
loading in tests with no load reversals (monotonic tests) and tests with at least one load
reversal (cyclic tests). From this, it is clear that generally but by no means necessarily,
dynamic tests are at the same time cyclic tests.
What follows in the remaining part of this section, is a more detailed discussion of the

11

Chapter 3 Experimental evidence for small-strain stiffness

laboratory and in-situ tests mentioned above, excluding surface based seismic refraction
and reflection. The latter is covered in more detailed in Forel et al. [42], but will not be
repeated here due to its limited applicability to geotechnical engineering.

3.1.1 Laboratory tests

3.1.1.1 Local measurements
Conventionally, axial strain in triaxial testing is calculated from the relative movement
between the apparatus top cap and its fixed base pedestal. Besides possible small compliances of the apparatus loading system, sample bedding is a major problem in deriving reliable small-strain stiffness data from such tests: The sample bedding at the
beginning of the test is neither perfect, nor guarantees uniform straining.
Imperfections in sample bedding are generally due to the sample preparation process. It is very likely that the sample does not have perfectly parallel and smooth ends,
so that the top cap may not have full contact instantaneously. In this case there is a
rapid deformation at the beginning of the test until the cap is bedded properly. At this
point, however, the small-strain range may already have been exceeded. On top of that,
the restraints in the bedding planes cause the sample to strain non-uniformly over its
height. Therefore, most conventional triaxial tests tend to give apparent soil stiffnesses
much lower than those inferred from dynamic testing with small displacement amplitudes (Jardine et al. [80]). One solution is to equip the test specimen with local strain
transducers.
Local strain transducers are not sensitive to compliances of the testing equipment
(other than the transducer itself) and imperfect sample bedding. They cannot completely
eliminate the effects of bedding restraints though. An appropriate ratio of sample height
to diameter is therefore still recommended. The weight of the instrumentation is to be
compensated by an adequate suspension. For very small-strain stiffness measurements,
transducer resolution should be 5 105 . Transducers that can achieve this high
resolution and which can at the same time be integrated in triaxial cells are e.g. Linear Variable Differential Transformers (LVDT), Digital Displacement Transducers, Hall
Effect Transducers, etc.
3.1.1.2 Bender elements
Carrying out small-strain stiffness measurements with local strain transducers on a regular basis is expensive. Local measurements are therefore generally confined to research
projects. The measurement of wave velocities in triaxial samples is less costly. Low voltage piezo-ceramic transducers named bender elements have been commonly used for
this purpose since their introduction at the end of the 1970s (Shirley & Hampton [165]).
Bender elements can both transmit and receive signals, and thus can readily measure
wave velocities in a sample when supplied on both sides of it. The element itself is a
thin piezo-ceramic plate that makes contact to the sample (Figure 3.3). Bender elements
can be clamped to samples for measuring horizontal traveltimes, as well.

12

3.1 Small-strain stiffness measurements

horizontal
benders

axial
benders

25mm

hall effect gauge

Figure 3.3: Bender elements mounted in a top plate and to a radial belt. Photograph by
Pennington et al. [136]

Originally, bender elements could transmit and receive S-waves only. Providing a
samples density is known, the knowledge of shear wave velocity is sufficient to calculate its shear modulus G0 from Equation 3.2. Unfortunately, S-waves are slower than
P-waves (vs 12 vp ) so that they always have noisy first arrivals (in this context, reflected
P-waves and surface waves are considered noise). Less valuable is the isolated knowledge of P-wave velocity as this is governed by both elastic constants. Ideal is the knowledge of both, S and P wave velocities, which can be simultaneously measured using
newly developed bender-extender elements (Lings & Greening [104]).
Compared to local strain transducers, the obvious disadvantage of bender, and bender-extender elements is their restriction to the very small-strain range, e.g. s < 1
106 . The geophysical nature of the indirect bender element stiffness measurements can
sometimes be troublesome too: Due to the low signal to noise ratio, the time at which
the first wave arrives is usually subject to interpretation. Sample preparation on the
other hand is simple. Commercially offered bender element mounts for top cap and
base pedestal along with user-friendly software promote bender elements as a feasible
option for routine lab testing. Figure 3.3 shows the integration of bender elements into a
top cap and into a radial belt.
3.1.1.3 Resonant column and torsional shear
Resonant column and torsional shear devices can load soil samples not only triaxially
but also torsionally. The basic difference between resonant column and torsional shear
testing is the frequency and amplitude of loading.
Torsional shear tests are static, or quasi-static cyclic tests where an axially confined

13

Chapter 3 Experimental evidence for small-strain stiffness

cylindrical sample is sheared through rotating one of the apparatus end plates. The
basic advantage of torsional shear testing over triaxial testing is that the bedding has a
minimum effect on the test result. Resonant column tests are cyclic tests, in which an
axially confined cylindrical soil specimen is set in a fundamental mode of vibration by
means of torsional or longitudinal excitation of one of its ends. Once the fundamental mode of resonance frequency is established, the measured resonant frequency can
be related to the columns stiffness using a theoretical elastic solution, which provides
satisfactory results in the very small-strain range.
Nowadays, resonant column testing and torsional shear testing can usually be accomplished within a single device. Various dynamic boundary conditions introduced in different test devices have only negligible effects on test results ([87]). A further refinement
of resonant column and torsional shear tests is the hollow cylinder apparatus. Obeying
the same principles, the hollow cylinder apparatus allows for pressurizing an additional
inner cell in the now tubular or hollow sample. Torsional shear stress acting on the sample is now well defined. Additionally, an independent inner cell pressure allows rotation
of the major principal stress axis, at any angle from the horizontal to the vertical. Unfortunately, all of these tests, resonant column, torsional shear, and hollow cylinder are
expensive and are therefore rarely used in routine design.

3.1.2 In-situ tests

3.1.2.1 Cross hole seismic
Cross hole seismic surveys require a minimum of two vertically drilled boreholes. In
one of the boreholes an energy source is lowered to the target soil layers depth. In
the neighboring borehole(s), at least one receiver is placed at the same depth. From the
source signals first arrival at the receiver station(s), it is possible to calculate horizontal
wave propagation velocities. When readings are taken at different source and receiver
depths, classical cross hole seismic tests can provide propagation velocities for all, ideally horizontal, soil layers. From these velocities, maximum soil stiffness is calculated in
the same way as in the bender element test, but often without knowing the soils exact
density.
Nowadays, cross hole tomography usually replaces conventional cross hole seismic
tests. Cross hole tomography uses a string of receivers instead of just one receiver, so that
multiple ray paths can be recorded for a single source signal. The additional information
recorded can then, through tomographic techniques, be inverted to velocity and stiffness
profiles with improved spatial resolution.
Cross hole surveying is probably the most reliable in-situ small-strain stiffness testing
method, but also the most expensive one: Shear wave borehole sources generate too little
energy to obtain economical borehole spacings. On top of that, the distances between
the boreholes must be known exactly, which typically demands inclinometer readings in
each borehole.

14

3.1 Small-strain stiffness measurements

3.1.2.2 Down hole seismics
Down hole seismic surveys require the drilling of only one borehole, in which a string of
receivers is placed. The energy source, generally a hammer blow against a steel plank,
is now located at the surface. Compared to cross hole seismic, the troublesome borehole source and the costs of multiple boreholes are eliminated. The main drawback of
down hole seismic tests are the almost vertical and very lengthy raypaths of sometimes
additionally refracted waves. Down hole surveys can thus be considered an integral
measurement over different soil layers. However, having receiver recordings from different depths, the initial stiffness of different soil layers can be back-calculated.
3.1.2.3 Suspension logging
Similar to down hole seismic surveys, suspension logging provides velocity data from
a single borehole. In suspension logging, source and receivers are placed in the same
borehole where they are separated by a few meters drilling suspension only. Unlike
in all other seismic techniques reviewed so far, the focus in suspension logging is not
on wave propagation velocity of direct waves, but rather on the propagation velocities
of waves that travel along the boreholes walls. Suspension logging can generate an
approximate very small-strain stiffness profile of the boreholes vicinity.
3.1.2.4 Seismic cone and seismic flat dilatometer
The commercially available seismic cone and seismic flat dilatometer are hybrid tests
that combine down hole seismic surveying with penetration, and dilatometer testing respectively. Instead of placing seismic receivers in a predrilled hole, they are now pushed
in. The energy source is located at the surface like in conventional down hole surveys.
Standard- or cone penetration, pressuremeter, and flat dilatometer testing devices without integrated seismic receivers cannot deliver very small-strain soil stiffness directly.
Still, they can provide some less reliable small-strain information by means of integral
deformation measurements or empirical relationships.
3.1.2.5 Spectral Analysis of Surface Waves (SASW)
Spectral Analysis of Surface Waves (SASW) is a non-intrusive geophysical technique for
evaluating subsurface shear wave velocity profiles. Using this technique, source and
receivers are both located at the surface. Instead of analyzing traveltimes of P- and Sbody waves as in most other seismic techniques, SASW uses the dispersive character of
Rayleigh surface waves. Long wavelength, low frequency Rayleigh waves penetrate the
subsurface deeper than high frequency Rayleigh waves. Since the propagation velocity
of waves increases with increasing confining pressure and hence depth, longer wavelength (low frequency) waves travel faster than shorter wavelength (high frequency)
waves. The ease and speed of SASW measurements in the field combined with automated data processing and inversion techniques probably makes SASW the most efficient in-situ small-strain stiffness exploration technique.

15

Chapter 3 Experimental evidence for small-strain stiffness

3.2 Parameters that affect small-strain stiffness

Based on a literature review, the influence of various parameters on small-strain stiffness
is identified and quantified here. In order to ease the quantitative description of smallstrain stiffness, and in particular its decay, the following notation is introduced:
G0 and E0 denote the maximum small-strain shear modulus and Youngs modulus
respectively.
0.7 denotes the shear strain, at which the shear modulus G is decayed to 70 percent
of its initial value G0 .
The tuples (G0 , 0) and (G0.7 , 0.7 ) mark two points of the small-strain stiffness degradation curve. The entire degradation curve can be reasonably well extrapolated from these
two points, for example by using the Hardin-Drnevich [52] relationship, which is introduced later in this chapter. In soil dynamics, the decay of small-strain stiffness with
applied strain is usually quantified as damping. Damping is a measure for energy dissipation in closed load cycles. With the Hardin-Drnevich model, the shear strain 0.7 can
be related to damping: The larger the value of 0.7 , the less the damping. The specific
threshold value of 70 percent is chosen here following a recommendation of Santos &
Correia [155] which is introduced later in this chapter, as well.
Strain amplitude, void ratio, confining stress, and the amount of in-situ interparticle
bonding turn out to be the most important parameters that affect the stiffness of soils at
small strains. This result was basically published by Seed& Idris [162] as early as 1970,
particle bonding excluded. Their results are illustrated in Figure 3.4. A complete list of
parameters that will be discussed in the following section is given in Table 3.1, which
is mainly based on the works by Hardin & Drnevich [53]. In Table 3.1, the original
classification of parameter importance by Hardin & Drnevich was updated wherever
more recent research results suggested it.

3.2.1 The influence of shear strains and volumetric strain

Soils conserve their initial stiffness G0 only at very small strains. With increasing strain,
soil stiffness decreases. Soils instantaneously recover their initial stiffness upon load
reversals. Therefore, the accumulated strain since the last load reversal is the main variable in small-strain stiffness modulus reduction. With the help of scalar valued strain
invariants, the accumulated strain since the last load reversal is usually expressed as
strain amplitude. In primary or virgin loading, strain amplitude refers to the reference
configuration at the onset of loading.
In literature, small-strain stiffness data is almost exclusively treated as a function of deviatoric strain amplitudes, e.g. shear strain s . Although small-strain stiffness has first
been recognized and analyzed in soil dynamics, the background of this is not only historical. In deviatoric loading, damping in the small strain rage is less during unloadingreloading than in primary loading, but still of the same magnitude (see Masings rule
later in this chapter). In isotropic loading, damping in small strain unloading-reloading

16

3.2 Parameters that affect small-strain stiffness

Figure 3.4: Small-strain stiffness decay as a function of (a) effective friction angle 0 , (b)
vertical effective stress v0 , (c) void ratio e, and (d) K0 [162].

17

Chapter 3 Experimental evidence for small-strain stiffness

Table 3.1: Parameters affecting the stiffness of soils at small-strains (modified after
Hardin & Drnevich [53]).
Parameter

Importance to a
G0
Clean
Cohesive
sands
soils
V
V
V
V
V
V
V
R
L

V
V
R
R
R
R
L
L
L
L

0.7
Clean
Cohesive
sands
soils
V
V
V
V
R
V
V
R
L

R
R
V
V
R
R
L
L
R
R

Strain amplitude
Confining stress
Void ratio
Plasticity index (PI)
Overconsolidation ratio
Diagenesis
Strain history
Strain rate
Effective material strength
Grain Characteristics
(size,shape,gradation)
Degree of saturation
R
V
L
L
Dilatancy
R
R
R
R
a
V means Very Important, L means Less Important, and R means
Relatively Unimportant

Modified from the original table presented in Hardin & Drnevich[53]

18

3.2 Parameters that affect small-strain stiffness

2.0

Volume strain eV [%]

1.4
1.2
0.8
0.4
0.0

400

800

1200

1600

Isotropic Pressure p' [kPa]

Figure 3.5: Isotropic compression test interrupted by small-strain cycles (after Lade &
Abelev [98]).

is much less than in primary loading: Figure 3.5 shows a recent test result by Lade &
Abelev [98]. In their test, Lade & Abelev compared the stiffness in isotropic loading
and unloading to that obtained when interrupting the continuous loading process by
small load cycles. In primary loading, soil stiffness within the load cycles is found to be
higher than the soil stiffness in continuous loading. Load cycling during unloading on
the other hand did not show any significant stiffness increase. This is also in agreement
with the findings of Zdravkovic & Jardine [202]: The secant and tangent bulk moduli
curves developed during swelling fall relatively gently with strain, remaining far above
the tangent compression value.
The remaining part of this chapter is hence focused on the reduction of soil stiffness
with shear strain. The terms strain and shear strain are sometimes used as synonyms.
Small-strain stiffness in volumetric loading is again discussed in the next chapter.

3.2.2 The influence of confining stress

Similar to the well known Ohde [131] or Janbu [77] type power laws for larger strains
(see Chapter 7), Hardin & Richard [54] proposed the following relationship between the
initial modulus G0 and the effective confining stress p0 :
G0 (p0 )m ,

(3.5)

which has not yet been superseded. Hardin & Richard themselves used the power law
exponent m = 0.5 for both, cohesive and non-cohesive soils. Today, their exponent is
widely confirmed for non-cohesive soils: All recent correlations use exponents in the
range of 0.40 m 0.55. For cohesive soils, the exponent m = 0.5 is controversial.
Many researchers confirmed it, others found exponents as high as m = 1.0.
At this point, it has to be considered that confining stress influences the degradation
of small-strain stiffness as well: Damping decreases with increasing confining stress.

19

1.0

1.0

0.9

0.9

Exponent m

Exponent m

Chapter 3 Experimental evidence for small-strain stiffness

0.8
0.7
0.6
0.5

0.8
0.7
0.6

10

20

30

40

50

Plasticity Index (IP) [-]

60

0.5

20

40

60

80

100

120

140

160

180

Liquid Limit wL [%]

Figure 3.6: The power law exponent m as a function of plasticity index (PI) and liquid
limit wL (after Viggiani & Attkinson [187], and Hicher [57]).

Correlating secant stiffnesses at greater strain amplitudes (s > 1 106 ) yields therefore higher m values than strictly using low strain bender element measurements (e.g.
s < 1 106 ). Additionally, it has to be taken into account that Hardin & Richard, like
many others, use a void ratio term in their relationship (see Equation 3.8), which also
relates to confining stress. Considering that non-cohesive soils are typically less compressible than cohesive soils, the scatter in the exponent m for cohesive soils can readily
be explained. Without taking into account void ratio in their relationship, Viggiani &
Attkinson [187] compiled the exponents m for different clays at very small strains as a
function of plasticity index; Hicher [57] compiled them as a function of Liquid Limit.
Both charts are shown in Figure 3.6. Small-strain stiffness data for a Kaolin clay with
PI 20 and LL 43 acquired by Rammah et al. [146] even correlated well, with m = 1.0.
This suggests that whenever void ratio is assumed constant in the corresponding relationship, the small-strain power law exponents m are equal (sands) or only a little less
(clays) than the exponents generally obtained in large strain oedometer and triaxial tests,
e.g. m = 0.5 for sands and m = 0.7 1.0 for clays.
Normalized modulus reduction curves, as for example the ones shown in Figure 3.8,
indicate that the threshold shear strain 0.7 is also mean stress dependent. Only a few relationships between confining stress and reduction of small-strain stiffness are proposed
in literature though. That by Ishibashi & Zhang [70] varies the exponent m in Equation
3.5 non-linearly with strain, as proposed earlier by Iwasaki et al. [76], [75]. A general recommendation for practical application can hardly be given though from their analysis.
Darendeli & Stokoe [30] developed normalized modulus reduction curves from an extensive study on over 100 undisturbed specimens from depths of 3 to 263 m. From their
data, the threshold shear strain 0.7 correlates very well to confining pressure p0 where:
0 m
p
,
(3.6)
0.7 = (0.7 )ref
pref
and pref = 100 kPa is a reference pressure, (0.7 )ref is the threshold shear strain at

20

Ratio g0.7/(g0.7)ref [-]

2.5

Darandeli & Stokoe

m=
0.6
5
m=
0.5
0

Iwasaki et al.

m
=0
.3
5

3.2 Parameters that affect small-strain stiffness

Kallioglou et al.
2

Wichtmann & Triantafyllidis

1.5
1

g0.7(p'):=(g0.7)ref*(p'/pref)

0.5
0
10

100

1000

Confining stress p' [kPa]

Figure 3.7: Correlation between confining stress p0 and shear strain 0.7 .

p0 = pref , and m
= 0.35. In Figure 3.7, the result from Darendeli & Stokoe is presented
together with other test data on non-plastic soils, i.e. data by Iwasaki et al. [76] that
was later used by Ishibashi & Zhang [70] for their correlation. From this representation
it can be concluded that a) the power law works reasonably well also for the threshold
shear strain and b) the power law exponent for non-plastic soils under moderate confining pressures is typically 0.35 < m
< 0.65. According to Stokoe et al. [89], Equation 3.6
can be used in combination with m
= 0.35 in plastic soils, as well. Other test data does
not always support this statement. Biarez & Hicher [16], for example, find the threshold
shear strain 0.7 of Kaolinite (PI = 30) largely unaffected by confining stresses in the range
of p0 = 100 300 kPa.

3.2.3 The influence of void ratio

The most frequently applied relationship between void ratio and initial soil stiffness
dates back to Hardin & Richart [54], as well. Based on their measurements of wave propagation velocities in Ottawa sand, they proposed a linear dependency between propagation velocity v, and void ratio e of the form:
n

v = a(b e)p0 2 .

(3.7)

From this linear dependency, Hardin & Richart derived their well known formula for
void ratio dependency of G0 to:
(2.17 e)2
1+e
(2.97 e)2

1+e

G0
G0

and

(3.8)
(3.9)

21

Chapter 3 Experimental evidence for small-strain stiffness

1.1

1.1

1.0

1.0
0.9

G / G0 [-]

G / G0 [-]

0.9
0.8

ID=0.95
ID=0.85
0.7
ID=0.70
ID=0.63
0.6
ID=0.50
0.5 all tests:
p'=80 kPa
0.4
10 -6

10 -5

0.8
0.7
0.6

10 -4

10 -3

0.5
10 -6

p'=400 kPa
p'=200 kPa
p'=100 kPa
p'= 50 kPa
all tests:
ID=0.63-0.66

Shear strain amplitude gs [-]

10 -5

10 -4

10 -3

Shear strain amplitude gs [-]

Figure 3.8: Influence of void ratio e (left) and confining stress p0 (right) on the decay of
small-strain stiffness (after Wichtmann & Triantafyllidis [193]). The influence
eemin
of void ratio is expressed by the density index ID = emax
.
emin

for round-grained sands (e < 0.80), and angular-grained sands (e > 0.60) respectively.
Hardin & Black [50], [51] later indicated that Equation 3.9 also correlates reasonably
well for clays with low surface activity. For clays with higher surface activity, the basic
structure of relationship 3.9 is often maintained, only the coefficient 2.97 is replaced by a
somewhat increased one (see Table 3.3).
Other relationships between void ratio and initial stiffness found in literature are typically of the form:
G0 ex
(3.10)
where the exponent x is quantified for instance as:
x = 0.8 (for sand - Fioravante [40])
x = 1.0 (for sand and clay - Biarez & Hicher [16])
x = 1.3 (for cemented sands, fine-grained soils - Lo Presti [141])
1.1 x 1.5 (for various clays - Lo Presti & Jamiolkowski [139]).
The influence of void ratio on the threshold shear strain 0.7 is apparently very limited in non-cohesive soils. This can be observed for example in normalizing the results
of Figure 3.4.c or in the more recent results presented by Wichtmann & Triantafyllidis
[193] (Figure 3.8). Damping in cohesive soils on the other hand is linked to void ratio.
In cohesive soils, void ratio also correlates to plasticity index (PI), as higher plasticity is
generally a prerequisite for a more open soil structure and thus a higher void ratio. Yet,
void ratio is also related to the confining stress and overconsolidation ratio. Therefore,
the influence of void ratio on the threshold shear strain 0.7 is not discussed in this section. Instead the effects of PI, OCR, and p0 on 0.7 are discussed in the following sections.

22

1.1

1.0

1.0

0.9

0.9

G / G0 [-]

1.1

0.8

PI
=1
5

0.8

PI

=5

PI
=1
00

0.7

=3
0

0.6

=0
PI

PI=0
0.7
PI=1-7
PI=8-20
0.6
PI=21-40
PI=41-44
0.5 47 soils tesed:
sV'=24-1700 kPa
0.4
10 -6
10 -5

PI=200

PI

G / G0 [-]

3.2 Parameters that affect small-strain stiffness

0.5
10 -4

10 -3

0.4
10 -6

Shear strain amplitude gs [-]

OCR=1-15
10 -5

10 -4

10 -3

Shear strain amplitude gs [-]

Figure 3.9: Influence of plasticity index (PI) on stiffness reduction: Left database for soils
with different PI; Right: PI-chart by Vucetic & Dobry (after Hsu & Vucetic
[64], and Vucetic & Dobry [188] respectively).

3.2.4 The influence of soil plasticity

In relating small-strain stiffness reduction curves to the plasticity index (PI or IP ), Vucetic
& Dobry [188] proposed the modulus reduction chart (PI-chart) shown in Figure 3.9.
These well known curves have been compiled from the results of 16 different publications of 12 or more research groups. The original data showed considerable scatter so
the PI-chart should be used with care, especially for PI 30. For lower plasticity clays,
the PI-chart is in reasonable agreement with many recently published test results. The
threshold shear strain 0.7 at p0 = 100 kPa of clean sands for example, is tested in the limits 8105 0.7 2104 , where the chart by Vucetic & Dobry suggests 0.7 1104 .
Stokoe et al. [89] also agree reasonably well with the PI-chart by Vucetic & Dobry up
to PI = 15. Subsequently, they suggest lower threshold shear strains. Generally, Stokoe
et al. propose a linear increase of 0.7 from 0.7 1 104 for PI = 0 up to 0.7 6 104
for PI = 100.

3.2.5 The influence of the overconsolidation ratio (OCR)

In cohesive soils G0 increases with OCR. The amount of increase depends upon the soils
plasticity. Hardin & Black [49] proposed an empirical relationship of the form:
G0 OCRk

(3.11)

where they define OCR as the ratio of maximum past vertical effective stress to the cur0

), and k is a parameter that is varying between 0

rent vertical effective stress (OCR = v,OC
v0
for sands and 0.5 for high plasticity clays. Atkinson & Little [8] came up with a different
relationship for tests on undisturbed London Clay:
G0 m log R0 ,

(3.12)

23

Chapter 3 Experimental evidence for small-strain stiffness

where R0 is defined as the ratio of maximum past effective mean stress to the current
p0
). Houlsby & Wroth [63] then combined the power law
effective mean stress (R0 = OC
p0
proposed by Hardin & Black [49] with the overconsolidation ratio R0 to give:
G0 R0k .

(3.13)

Again the empirical parameter k increases with clay plasticity. For clays with 10 < PI <
40, Atkinson & Viggiani found 0.20 < k < 0.25. For practical applications this seems
a relatively small variation, such that it is sometimes proposed to neglect the effect of
OCR on G0 completely, e.g. in Lo Presti & Jamiolkowski [139]. However, overconsolidation cannot be neglected in the inherent and stress-induced anisotropy of small-strain
stiffness, which is discussed later in Section 3.2.7.
In non-plastic soils, overconsolidation has only a small effect on the normalized modulus reduction curve. In plastic soils, overconsolidation markablly increases the threshold
shear strain 0.7 . The increase of 0.7 with overconsolidation OCR can after Stokoe et al.
[89] be approximated as:
0.7 = (0.7 )ref + 5 106 PI(OCR)0.3 ,

(3.14)

where (0.7 )ref is here a reference threshold shear strain for a non-overconsolidated, nonplastic soil, e.g. (0.7 )P I=0,OCR=1 1e4. For overconsolidated soils, the findings by
Stokoe et al. [89] and Vucetic & Dobry [188] are thus also in better agreement for soils
with higher plasticity (Note that the PI-chart by Vucetic & Dobry is independent of overconsolidation).

3.2.6 The influence of diagenesis

Diagenesis refers to postdepositional processes, involving seawater, meteoric water, or
subsurface brines, that alter sediments or sedimentary rocks up to the point of metamorphism, and include burial, compaction, dissolution, and precipitation, (Morse and
Mackenzie [124]). Diagenetic processes alter the stiffness of soils with time: Generally,
any process that alters the existing interparticle structure within a soil will also alter its
stiffness. Diagenetic processes that have a considerable effect are cementation and aging, which in the classical sources (Lambe & Whitman [182], Mitchell [120], Terzaghi et
al. [179]) are defined as a change in various mechanical properties resulting from secondary compression under a constant external load. Although traditionally linked to
clays (e.g. Bjerrum [17]), aging is also known to occur in sands, sandstones, and clayey
sands (e.g. Mesri et al. [117], Schmertmann [159]). Cementation, which is not necessarily
linked to secondary compression, is particularly important to the stiffness of sandy soils.
Fernandez & Santamarina [39] for example, found that small-strain stiffness of sands
can increase by an order of magnitude or more due to cementation.
The time-dependent increase of G0 due to aging in clays may be linked to engineering
properties, e.g. the secondary consolidation coefficient C with empirical expressions of
the type proposed by Anderson & Stokoe [4]

t
G0 (t)
= 1 + NG,1 log
(3.15)
G0 (tp )
tp

24

3.2 Parameters that affect small-strain stiffness

or the one proposed in Shibuya et al. [164]
NG,2
G0 (t)
t
=
G0 (tp )
tp

(3.16)

where tp is the time required to reach the end of primary consolidation, t is any time
(t > tp ), G0 (tp ) is the maximum stiffness at time tp , and NG,x is an empirical material
factor. Lo Presti et al. [140] proposed the following relationship:
NG,1 = C0.5 ,

(3.17)

which was confirmed by additional test results of Lohani et al. [107].

Diagenetic effects of a soil can be readily lost upon changing its state of stress. Disturbed soil samples may therefore show considerably different small-strain behavior
than undisturbed samples. Toki et al. [180] prepared a database from case studies in
Japan (Figure 3.10) that show a clear correlation between sampling method and the reliability of very small-strain stiffness laboratory measurements. Employing in-situ freezing
methods yields the best agreement between in-situ G0,F ield and laboratory G0,Lab determined stiffness values. Thin wall sampling also gives reliable laboratory results, except
for sands, which are densified during sampling. More commonly used sampling methods for sands and soft rocks, however, often disturb the samples so much that the laboratory determined G0,Lab is as low as 0.25G0,F ield . Results from the ROSRINE (Resolution
Of Site Response Issues from the Northridge Earthquake) study, which is an on-going
study set up after the 1994 Northridge earthquake, show similar reduction factors. Here,
G0,Lab
a general trend is depicted between the ratio G0,F
and the in-situ shear wave velocity
ield
(Figure 3.10).

3.2.7 The influence of loading history

Masing [111] described the hysteresis in stress-strain behavior in the form of the following two rules:
The shear modulus in unloading is equal to the initial tangent modulus for the
initial loading curve.
The shape of the unloading and reloading curves is equal to the initial loading
curve, except that its scale is enlarged by a factor of two.
Although Masings original work was concerned with the value of the proportional limit
of brass under cyclic loading, his rules appear to describe the actual behavior of soil
under cyclic loading reasonably well. In irregular cyclic loading however, the above
rules have to be extended. A detailed discussion of possible extensions is for example
given in Pyke [143], which is also summarized in Section 6.1.3 of this thesis.
In terms of the above introduced threshold shear strain 0.7 , Masings second rule is
fulfilled by writing:
1
(0.7 )initial loading = (0.7 )reloading
2

(3.18)

25

Chapter 3 Experimental evidence for small-strain stiffness

In-situ shear wave velocity vS,Field[m/s]

Modulus ratio G0,Lab/G0,Field

Modulus ratio G0,Lab/G0,Field

2
Holocene sand
(thin-wall sampling)
In-situ frezzing
method
1 Holocene clay

Pleistocene clay Gravel

Soft rock
(block)

Pl
e
(tu isto
Soft rock
be ce
sa ne
(rotary core)
m sa
pl
n
in d
g)
0

10

100

0.10

0.25

1.50

2.00

150
300

Remolded
cemented
sandy soils

450

Range from
ROSRINE
study

600
750

General trend

900

Range often found

with rock cores

1050
0.0

1000

0.50 0.80 1.00

0.5

G0,Field [MPa]

1.0

1.5

Velocity ratio vS,Lab/vS,Field

Figure 3.10: Differences in the ratio of labratory-to-field stiffness. Left: Data from
Japanese case studies (after Toki et al. [180]). Right: Results from the US
American ROSRINE study (after Stokoe & Santamarina [173]).

Shear modulus G [MPa]

200

first increase of g
180

re-increase of g
160

140

reduction of g

plateau
increase of
gS > gprestarin

prestraining
amplitude
-4

gprestarin=10
120
10 -6

10 -5

10 -4

10 -3

Shear strain amplitude gs [-]

Figure 3.11: Influence of cyclic prestraining on small-strain stiffness (after Wichtmann &
Triantafyllidis [193]).

Although experimental data often confirms Masings rules, the threshold shear strain
0.7 in unloading-reloading increases not always to the extent described by Masing. An
example presented in Wichtmann & Triantafyllidis [193] is shown in Figure 3.11. Here,
a sample was cyclic loaded to investigate the effects of cyclic prestraining. In this particular example, the modulus G0 slightly decayed in cyclic loading. In other samples,
which are not shown here, it also slightly increased so that no correlation between the
number of applied prestraining cycles and G0 is found. Part one of Masings rule on the
other hand is confirmed. Normalizing all load cycles for their individual maximum also
confirms the second part.

26

3.2 Parameters that affect small-strain stiffness

Wave velocity [m/s]

Wave velocity [m/s]

400

400

200
S

VX,v
-400

-200

200

200

-200

VX,h

VX

VX,v
-400

400

VX

-200

200

VX,h

VX

400

-200

VX,h
-400
sV=100 kPa; sH=100 kPa

VX,v

-400
sV=150 kPa; sH=75 kPa

Figure 3.12: Polar plot of wave speeds as a function of their inclination ( = 0 := vertically travelling waves, = 90 := horizontally travelling waves). Isotropic
loading yields almost isotropic wave speeds (left); axial unloading to a stress
ratio of 2 decreases axial stiffness (right). Dotted lines give the results from
a cross anisotropic model after Bellotti et al. [14].

A second, and probably as important an effect of loading history is the formation of

stress or strain induced anisotropy. This has been examined i.a. by Hoque & Tatsuoka
[61], Bellotti et al. [14], and Yu & Richart [200]. In summary these studies conclude
0
that principal stress ratios ( 10 > 1) increase the maximum small-strain stiffness in the
3
direction of the highest principal stress (10 ). The stiffness in the direction of the minor
0
principal stress (30 ) is almost unaffected up to stress ratios of ( 10 = 3). For higher stress
3
ratios, stiffness decreases in the minor principal stress direction. Figure 3.12 illustrates
these findings by a comparison of wave speeds at different stress ratios. The stress ratio in normally consolidated soils is K0 = 0.5. When back-calculated from cross hole
measurements, the vertical stiffness of normally consolidated soils can thus be underestimated up to 30%.
Even though Figure 3.12 indicates near isotropic stiffness in isotropic loading conditions, the issue of inherent anisotropy cannot be neglected in many soils, either. Inherent anisotropy is attributed to the genesis of soils, for example depositional processes, as
well as to their diagenesis (see Section 3.2.6). Often, the stiffness in the horizontal (bedding) plane appears to be 20 30% higher than the one in the vertical plane (Bellotti et al.
[14], Chaudhary et al. [28]). In a way, isotropic stiffness might therefore be a reasonable
assumption for practical applications, considering that inherent anisotropy compensates
somewhat for stress induced anisotropy in normally consolidated soils.

27

gS,1

GHigh,1
aG,1
1

gS,2 > gS,1

GHigh,2
1

GLow,1

aG,2

GLow,2
0

gS,Low

gS,High

Shear modulus parameter aG [MPa]

Secant shear modulus G [MPa]

Chapter 3 Experimental evidence for small-strain stiffness

-5

-5

gS

.4
<1

-5

e
2.5

-5

e
9.7

<

e
6.0
g S<

-4

< gS

.2e
<1

Average shear strain rate log g [-]

10

20

30

40

50

Plasticity Index IP [-]

Figure 3.13: The strain-rate shear modulus parameter and its dependence on strain amplitude, and plasticity Index (PI). After Vucetic et al. [190].

3.2.8 The influence of strain rate and inertia effects

Systematic experimental investigations revealed that in many soils the small-strain modulus G0 increases with the rate of loading (e.g. Whitman [192], Matesic & Vucetic [112],
and Vucetic & Tabata [189]). This so called strain rate effect can be attributed to viscosity
and thus to soil plasticity as there is virtually no such strain rate effect in sands. The
strain rate effect in cohesive soils increases with plasticity index. As a measure for the
increase, Yong & Japp [198] define the stress or strain rate shear modulus parameter to:
G =

(GS )
.
log

(3.19)

In this way, the strain rate shear modulus parameter G defines the slope of the stress
versus strain rate line in a semi-logarithmic format. In plotting the parameter G versus
strain rate and amplitude (Figure 3.13), the following correlation becomes clear: The
strain rate effect increases with strain rate and soil plasticity, but decreases with shear
strain amplitude. Vucetic et al. [190] first quantified the strain rate effect as a function of
plasticity index (PI) and liquid limit wL but with significant uncertainties.
From Figure 3.13 and 3.14 it can be concluded that small-strain stiffness is only a little
biased by common straining rates. This is in agreement with findings from dynamic
tests, where inertia effects have to be considered on top of strain rate effects. Stokoe et
al. [172] conclude from a recent study that for excitation frequencies changing from 1 to
about 100 Hz, G0 increases by about 5% to 30%, with the effect generally increasing with
increasing PI. Many other researchers found similar trends. Georgiannou et al. [44], for
example, report a 10% increase of G0 in different clays and marl due to dynamic loading.
In sands, only a very small or no increase is found (Hicher [57]). The Japanese working
hypothesis that soil parameters for earthquake analysis can be derived from static tests
is therefore validated from a practical point of view.

28

3.2 Parameters that affect small-strain stiffness

10

Hardrock core

Concrete
Mortar
Ultrasonic
Resonant-column
4

Sagamihara soft rock (U)

E0 [MPa]

10

Metramo silty sand (U)

10

OAP clay (U)

Sandy gravel (D)

Wet Chiba
gravel (D)
Saturated Toyoura
sand (U)
Hostun sand (D)

10

NC Kaolin (CU)

Vallerica clay

10

-5

10

-4

10

-3

10

-2

10

-1

10

10

10

10

10

Axial strain rate de/dt [%/min]

Figure 3.14: Strain rate effect on the Youngs modulus E0 after Tatsuoka [177].

29

Chapter 3 Experimental evidence for small-strain stiffness

3.2.9 Other factors that influence small-strain stiffness

Small-strain stiffness is only slightly influenced by the effective angle of friction for granular soils. An example is given in Figure 3.4, where the threshold shear strain 0.7 is
somewhat higher in the sample with a higher angle of friction. As the effective angle
of friction can be related to the density index, this observation is in good agreement to
the one discussed in Section 3.2.3. Grain shape, also has little affect on the small-strain
stiffness. Dilatancy is not activated in the range of small strains at all.
Mean grain diameter and grain size distribution, on the other hand, are reported to
influence small-strain stiffness of poorly graded soils considerably. Iwasaki & Tatsuoka
[74] found the shear modulus, G0 , to be approximately 10 25% reduced in well graded
sands compared to poorly graded sands. The addition of 2% 5% fines then showed
another decrease in G0 of about 20%. Fine contents is also important for the effect of
saturation on small-strain stiffness.
In fully saturated samples, pore water has an indirect impact on small-strain stiffness:
Dry and saturated clean sands show exactly the same G0 modulus, at equal effective
confining pressures. Therefore pore water can be ruled out as a parameter that affects
the shear modulus G0 of clean sands, except for its influence on mean effective stress.
The water contents in clay play a major role, if it is close to the liquid limit. Then, the
shear modulus G0 decays to a minimum threshold value, which can be related to the
undrained shear strength cu by:
G0 = IR cu

(3.20)

Anderson & Woods [5] determined the dimensionless soil parameter IR for many clays
within the band 300 < IR < 1800. Considering that the undrained shear strength of
most clays decays to cu 1.7 kPa when they approach the liquid limit [199], their shear
modulus can be as low as 500 kPa. Such low shear moduli at the liquid limit have, for

example, been verified by Holzlohner

[59].
In partially saturated samples, capillary effects have to be considered in the presence
of fines (Stokoe & Santamarina [173]). Capillary forces at interparticle contacts increase
the shear modulus G0 with decreasing level of saturation and decreasing grain diameter
of fines. With decreasing saturation, fine clay particles increasingly migrate to interparticle contacts of sand and silt grains. There, they form clay buttresses and bridges that
increase small-strain stiffness (see also Section 3.2.6).
When discussing the effect of saturation, it should be finally noted that for temperatures below 0 C, the role of pore water is obviously different. In saturated clays, for
example, a 3000% to 5000% increase of G0 is to be expected for a temperature decrease
from 20 C to 10 C.

30

3.3 Correlations for small-strain stiffness

3.3 Correlations for small-strain stiffness

3.3.1 Correlations between G0 and e, p0 and OCR
Tables 3.2 to 3.4 provide a number of very small-strain stiffness correlations on the basis
of the modified Hardin & Black [49] equation:
0 m
p
k
G0 = A f (e) OCR
,
(3.21)
pref
where G0 is the maximum small-strain shear modulus in MPa, p0 is the mean effective
stress in kPa, pref = 100 kPa is a reference pressure equal to the atmospheric pressure,
OCR is the overconsolidation ratio, and A, f (e), k, m are the correlated functions and
parameters given in Tables 3.2 to 3.4. The type of OCR used in a specific relationship is
denoted by the subscript of k: ()1 indicates the use of vertical stress based OCR values,
()2 indicates the use of mean pressure based OCR values (see Section 3.2.5). The symbols
emin , and emax stand for minimum and maximum void ratio respectively. D50 is the mean
grain diameter and UC is the uniformity coefficient.
None of the relationships given in Tables 3.2 to 3.4 addresses explicitly the second
independent elastic constant: It is a common simplification to assume Poissons ratio
a constant in the small-strain range, although it increases slightly with applied strain.
At least in granular materials, Poissons ratio is found to be almost independent of the
mean effective stress and is only slightly influenced by void ratio. Poissons ratio instead
appears to depend on grain-size distribution, e.g. Biarez & Hicher found = 0.18 for
poorly graded sand and = 0.23 for well-graded sand. As a general guideline, it can be
assumed that 0.10 0.30 for the small-strain range. Finally it should be acknowledged that the use of a pressure dependent shear modulus as introduced in Equation
3.21 together with a constant void ratio can violate thermodynamical principles.
The tables are sorted for the grain size D50, or PI instead of for author or reference.
Ranges in the parameters A, and m refer to the extreme limits due to inherent anisotropy,
if anisotropic measurements are available. All relationships given in Tables 3.2 to 3.4 are
illustrated in Figure 3.15 for the respective range of void ratios.
As a rule of thumb, either the simple relationship by Biarez & Hicher [16]
s
p0
140
E0 [MPa] =
(3.22)
e
pref
or the relationship by Hardin & Black [51]
(2.97 e)2
G0 [MPa] = 33
1+e

s
p0
pref

(3.23)

can be used to estimate E0 and G0 in various soils. Here again, e is void ratio, p0 is the
mean effective stress in kPa, and pref = 100 kPa is the reference pressure equal to the
atmospheric pressure. The relationship by Biarez & Hicher was proposed for all soils
with wl < 50%, the one by Hardin & Black was derived for undisturbed clayey soils and
crushed sands.

31

Chapter 3 Experimental evidence for small-strain stiffness

Table 3.2: Relations for the shear modulus G0 of CLEAN SANDS and GRAVELS.
Soil tested
Kenya carbonate sand

D50
[mm]

UC
[-]

A
[-]

f (e)
[-]

k
[-]

m
[-]

Ref.

0.13

1.86

101-129

e0.8

0.45-0.52

[40]

(2.17e)2

0.41-0.51

[61]

0.50-0.57

[28]

0.50

[88]

0.50

[88]

0.50

[88]

Toyoura sand (subangular)

0.16

1.46

71-87

Toyoura sand (subangular)

0.19

1.56

84-104

Silica sand (subangular)

0.20

1.10

80

Silica sand (subangular)

0.20

1.70

62

Silica sand (subangular)

0.20

1.10

62

1+e
(2.17e)2
1+e
(2.17e)2
1+e
(2.17e)2
1+e
(2.17e)2
1+e

Toyoura sand (subangular)

0.22

1.35

72

e1.3

0.45

[141]

H.River sand (subangular)

0.27

1.67

72-81

0.50-0.52

[96]

Glass ballotini (spheres)

0.27

1.28

64-69

0.55-0.56

[96]

Hostun sand (angular)

0.31

1.94

80

0.47

[60]

Silica sand (angular)

0.32

2.80

48

0.50

[88]

Ticino sand (subangular)

0.50

1.33

61-64

0.44-0.53

[61]

Ticino sand (subangular)

0.54

1.50

71

0.43

[141]

Silica sand

0.55

1.80

275

(2.17e)2
1+e
(2.17e)2
1+e
(2.17e)2
1+e
(2.17e)2
1+e
(2.17e)2
1+e
(2.27e)2
1+e
(1.46e)2
1+e

0.42

[193]

Ticino sand (subangular)

0.55

1.66

79-90

e0.8

0.43-0.48

[40]

(2.17e)2

0.44-0.53

[61]

0.50

[54]

SLB sand (subround)

0.62

1.11

82-130

Ottawa sand No. 20-30

0.72

1.20

69

1+e
(2.17e)2
1+e

Quiou carbonate sand

0.75

4.40

71

e1.3

0.62

[141]

Decomposed granite

1.30

75

45

0.88

[86]

Hime gravel (subround)

1.73

1.33

53-94

0.45-0.51

[61]

Chiba gravel

7.90

10

76

(2.17e)2
1+e
(2.17e)2
1+e

0.50

[122]

32

3.3 Correlations for small-strain stiffness

Table 3.3: Relations for the shear modulus G0 of CLAYS.

Soil tested

A
[-]

f (e)
[-]

k
[-]

m
[-]

Ref.

Avezzano clay (Holocene-Pleistocene) 10-30

74

e1.27

NA

0.46

[139]

Garigliano clay (Holocene)

10-40

44

e1.11

NA

0.58

[139]

Montaldo di Castro clay (Pleistocene)

15-34

50

e1.33

NA

0.40

[139]

Speswhite caolin clay (recon.)

24

40

0.202

0.65 [187]

Recon. Vallericca clay (Pleistocene)

27

44

...

0.85

[147]

Kaolin clay

35

45

(2.97e)2
1+e

NA

0.50

[110]

23-46

50

e1.43

NA

0.44

[139]

London clay (reconstituted)

41

13

0.252

0.76 [187]

Panigaglia clay (Holocene)

44

52

e1.30

NA

0.50

[139]

45-75

64

e1.52

NA

0.40

[139]

4.5

(4.40e)2

NA

0.50

[110]

Pisa clay (Pleistocene)

Fucino clay (Holocene-Pleistocene)

PI
[%]

Bentonite

60

1+e

Table 3.4: Proposed relationships for the shear modulus G0 of entire soil groups.
Soil tested

emin
[-]

emax
[-]

A
[-]

f (e)
[-]

m
[-]

Ref.

Clean sands with Cu < 1.8

0.5

1.1

57

(2.17e)2
1+e

0.40 [74]1

All soils with wl < 50%

0.4

1.8

59

0.50 [16]2

Undisturbed clayey soils & crushed sand

0.6

1.5

33

Undisturbed cohesive soils

0.6

1.5

16

Loess

1.4

4.0

1.4

1
e
(2.97e)2
1+e
(2.97e)2
1+e
(7.32e)2
1+e

0.50 [51]3
0.50 [90]4
0.60 [92]5

Iwasaki & Tatsuoka

Biarez & Hicher (assumes = 0.2)
3 Hardin & Black
4 Kim et al.
5 Kokusho et al.
2

33

Chapter 3 Experimental evidence for small-strain stiffness

p' = 100 kPa

250

p' = 400 kPa

500

Hardin & Black

Hardin & Black

Biarez & Hicher

200

Kim et al.

150

Kokusho et al.
Sand &
Gravel

100

Kenya
sand
Clay

50

0
0.5

1.5

Iwasaki & Tatsuoka

Shear modulus G0 [MPa]

Shear modulus G0 [MPa]

Iwasaki & Tatsuoka

Biarez & Hicher

400

Kim et al.
Kokusho et al.

300

200

100

2.5

0
0.5

Void ratio e [-]

1.5

2.5

Void ratio e [-]

Figure 3.15: Shear modulus G0 versus void ratio e, after Table 3.2, 3.3, and 3.4.

3.3.2 Correlations between G0 and CPT, SPT and cu

3.3.2.1 Cone penetration (CPT) and standard penetration (SPT)
Relationships between the small-strain modulus G0 and the corrected cone penetration
tip resistance qt have been proposed for clean quartz sands (Baldi et al. [11]), clays
(Mayne & Rix [116]), and various soil types (Lunne et al. [108]). The correction of tip
resistance due to pore water pressure effects is described in Lunne et al. [108].
For clays, Mayne & Rix [116] give the expression:
G0 [MPa] = 49.4

qt0.695
e1.13

(3.24)

where qt is the corrected tip resistance in units of MPa. The correlation by Mayne & Rix
[116] and that by Lunne et al. [108] are shown in Figure 3.16.
Since G0 is related to CPT tip resistance, which can be correlated to the undrained
shear strength, G0 should also be directly related to the undrained shear strength. Larsson & Mulabdic [100] proved this for Scandinavian clays by formulating:

A
G0 =
+ B cu
(3.25)
PI
where PI is the plasticity index, cu is undrained shear strength in kPa, and A, B are
empirical soil parameters. For Scandinavian clays they found a reasonable fit for A =
208, and B = 250.

34

3.3 Correlations for small-strain stiffness

1000

Shear Modulus G0 [MPa]

Gravelly
sands
100
Sand

nte
me
nce
gu

n
You
Sand mixtures

10

oils
ds

Normalized tip stress: (qt-s'vo)/s'vo [-]

1000

Silt mixtures

100

.
e=0

10

Multiple
Regression
n=481
r=0.901

.
e=1

.0
e=2 0
.
e=3

Clays

Clay soil type:

Residual Clay (Gneiss)
Intact Clay (Sediment)
Fissured Clays
Residual Clays

G0=49.4 qt

0.695

-1.130

1
1.0

10.0

Ratio G0/qt [-]

100.0

0.1

1.0

10.0

Cone tip stress qt [MPa]

Figure 3.16: Correlation between CPT results and very small-strain stiffness after Lunne
et al. [108] (left) and Mayne & Rix [116] (right).

Most correlations between standard penetration test (SPT) results and very smallstrain stiffness take the form:
G0 = A((N1 )60 )B
(3.26)
which was first proposed in Ohsaki & Iwasaki [132]. In Equation 3.26, (N1 )60 is the number of blows per 0.3 m, and A and B are again empirical soil parameters. The relationship
is generally limited to a maximum blow count of (N1 )60 = 30. For cohesive soils, Imai &
Tonouchi [69] found a best fit to test data for A = 15.56 and B = 0.68 (G0 in MPa units).
3.3.2.2 Estimating G0 from conventional tests - Chart by Alpan
Alpan [3] published the chart shown in Figure 3.17 which, as he stated, relates static to
dynamic soil stiffness (see axis labels Es , Ed ). Actually, this chart relates stiffness from
conventional laboratory tests to very small-strain stiffness: The dynamic modulus is the
initial, or very small-strain modulus E0 , and the static modulus is the apparent elastic
Youngs modulus in conventional soil testing, e.g. at axial strains of a 1 103 in
triaxial testing. For a soil with known Youngs modulus in triaxial unloading-reloading,
the Alpan chart can thus provide an estimate for its very small-strain modulus E0 .

3.3.3 Correlations for the stiffness modulus reduction curve

In contrast to the numerous relationships that have been proposed for the shear modulus
G0 , only a few correlations are available for its reduction with strain amplitude. Some of
these modulus reduction schemes rely on parameters that can be derived from test data

35

Chapter 3 Experimental evidence for small-strain stiffness

Figure 3.17: Correlation between very small-strain stiffness and stiffness at larger strains
from conventional laboratory tests after Alpan (10 kg/cm2 1 MPa).

directly, others use more complex fitting procedures. Only the former are discussed in
this section. The latter are discussed in detail in Chapter 5.
From test data, sufficient agreement is found that the stress-strain curve for small
strains can be adequately described by a simple hyperbolic law. An analogy to the hyperbolic law for larger strains by Kondner [94] (see Section 7) was proposed by Hardin
& Drnevich [52]:
G
1
.
=
(3.27)

G0
1+
r

where the threshold shear strain r is quantified as:

r =

max
G0

(3.28)

with max being the shear stress at failure. Essentially, Equation 3.28 relates large (failure)
strain to small-strain properties, which astonishingly often works well. More straightforward, and less prone to error is probably the use of a smaller threshold shear strain.
Stokoe et al. [89] for example, suggest to use the shear strain r = 0.5 at which the shear
modulus G0 is reduced to its half. Additionally they modify Equation 3.27 by introducing an additional exponent = 0.92:
1
G
.
=
G0
1 + r

36

(3.29)

3.3 Correlations for small-strain stiffness

1.0
0
Secant modulus
G/G0 [-]

0
Secant modulus
G/G0 [-]

1.0
0.8
0.6
0.4

Hardin &
Drnevich

0.2
0.0 -3
1e

1e

-2

1e

-1

1e

1e

1e

1e

0.8

0.6
0.4

Santos &
Correia

0.2
0.0 -3
1e

Normalized shear strain gS/g0.7 [-]

-2

1e

-1

1e

1e

1e

1e

1e

Normalized shear strain gS/g0.7 [-]

Figure 3.18: The original Hardin & Drnevich [52] relationship (left) and its modification
by Santos & Correia [155] (right) compared to test data.

Using the threshold shear strain a = 0.7 , Santos & Correia [155] have put forward the
following modified Hardin & Drnevich relationship:
G
1
.
=
G0
1 + a a

(3.30)

The constant a in Equation 3.30 is problematic though: For a 6= 73 the relationship is inconsistent as the shear modulus G is not reduced exactly to 0.7G0 at shear strain = 0.7 .
This for example is the case for the best fit a = 0.385, Santos & Correia [155] found from
a correlation using many test results. The modification by Santos & Correia combined
with a constant of a = 73 on the other hand is identical to the Hardin-Drnevich relationship. Looking at both reduction curves (Figure 3.18), their deviation is < 3% and thus
negligible for practical purposes anyway. The approach by Santos & Correia gives a
slight upshift of the reduction curve which is indeed desirable in comparison to the not
truly hyperbolic test data collected by Santos & Correia [155] shown in Figure 3.18).
In the following, mostly the modified Hardin-Drnevich relationship (3.30) with a =
0.385 is used. For the original Hardin-Drnevich model, the reference shear strain a =

0.7
is then taken at the shear strain where G decayed to 0.722G0 . For the Santos & Correia
upshift, a = 0.7 where G decayed to 0.7G0 is to be selected. The exponent proposed
by Stokoe et al. is not adopted as it may sometimes flatten the shear modulus reduction
curve too much in the small-strain range.
In Equation 3.28 the threshold shear strain r is related to the shear stress at failure. In
the Mohr-Coulomb failure criterion [123] the shear stress at failure can be expressed as
follows:
1
(3.31)
max = (2c(1 + cos 2) + (10 + 30 ) sin 2) .
4

of normally
Remembering that a = amax /G0 , the threshold shear strains 0.7 , and 0.7

37

Chapter 3 Experimental evidence for small-strain stiffness

consolidated soils can be calculated as:
0.7 =

0.385
(2c(1 + cos 2) + 10 (1 + K0 ) sin 2) ,
4G0

(3.32)

0.7 =

3
(2c(1 + cos 2) + 10 (1 + K0 ) sin 2) ,
28G0

(3.33)

and

respectively. If no small-strain experimental data is available, these values should also

be crosschecked with the chart by Vucetic & Dobry [188] shown in Figure 3.9 and the
correlation by Stokoe et al. [89] given in Section 3.2.4.

3.3.4 Summary
The stiffness of soils at very small strains is affected by a number of parameters. Its
correct determination is not a trivial task: Test results on disturbed samples may give
very different results than in-situ tests. In-situ testing methods use geophysical methods,
which are not yet common in geotechnical engineering.
However, it has been shown that many of the factors which affect the shear modulus
G0 and its reduction can be regarded as being of minor importance. Neglecting these
factors leads to the simple relationships shown in Table 3.4. and illustrated in Figure
3.15. Empirical relationships for estimating the shear modulus reduction with strain
are available as well. From a practical point of view, small-strain stiffness can thus be
incorporated into routine design at minimum cost. The use of laboratory and/or in-situ
testing data is still preferably though.
If no in-situ test data are available, laboratory small-strain stiffness data can be extrapolated to the field. In the case of common tube or rotary core sampled cemented soils,
the actual soil stiffness in the field G0,F ield should be expected to be considerably higher
than that measured in the laboratory G0,Lab (Figure 3.10). The modulus reduction curve
found in the laboratory can be extrapolated to the field by assuming [173]:
G,F ield =

G0,F ield
G,Lab .
G0,Lab

(3.34)

A numerical model that allows for the incorporation of small-strain stiffness into routine design is developed in the following chapters. The input parameters of this model
will be exactly the ones quantified above: The shear modulus G0 and the threshold shear

.
strain 0.7 or 0.7

38

Chapter 4
Small-strain stiffness at the soil particle level
The most important parameters that affect small-strain stiffness have been identified in
the previous chapter. In summary, these are void ratio, confining stress, cementation
and strain amplitude. The observed influence of these parameters on material stiffness
are now interpreted at the soil particle level. However, it is not the aim of this chapter
to derive a complete small-strain stiffness model from micromechanical considerations.
It is aimed rather at providing a thorough understanding of the small-strain stiffness
phenomenon.
The material, which we commonly refer to as soil is usually composed of air, water,
and solid particles, which may vary in size, shape, and mineralogic composition. The
mechanical properties of soil depend directly on the interaction of these phases with
each other and with externally applied potentials such as stress and temperature. Some
mechanical properties of soil, for example suction, cannot be understood without explicitly considering multi-phase physics. Others, that are purely determined by the net
forces between particles, for example small-strain stiffness, can be sufficiently explained
by the micromechanical considerations presented in this chapter.

4.1 Soil fabric and soil structure

The term soil fabric refers to the arrangement of particles, particle groups, and pores in
a soil. Soil structure is made up of both, the fabric and the sum of inter-particle forces
within the fabric. Soil structure thus takes both fabric and stability into account and is
therefore a key feature in determining the engineering properties of a soil.
Sand has typically a single-grain fabric, which means that the particles behave as independent units in the overall soil structure. Possible forces between sand particles are
largely electrostatic forces in the form of the Born repulsion forces as well as primary
valence bonds in the presence of cementing agents. Born repulsion develops at contact points between particles to avoid the interpenetration of particles. Primary valence
bonds are due to cementation and hence the chemical bonding of particles. Silica or
carbonate cement is often found in aged sand deposits. Together with other diagenesis effects as for example densification, cementation leads to increased material stiffness
(see Chapter 3) and strength. The penetration resistance of aged sand, for example, is reported to be twice as high as the penetration resistance of freshly deposited sand (Mesri
et al. [117]). Figure 4.1 shows the bonded single-grain fabric of a Fontainebleau sandstone.

39

Chapter 4 Small-strain stiffness at the soil particle level

Figure 4.1: Scanning electron microscopy images of sand Particles: Single-grain sand
fabric (left); Close-up of amorphous silica and quartz bonds in a sandstone (right). Photographs by IWF, Universitat Marburg, and Tore Stendahl,
NTNU, Trondheim, Norway.

Clay matrix

Silt or
Sand

(a)

(b)

(c)

Figure 4.2: Particle arrangements after Mitchell [120]: (a) single-grain sand or silt fabric,
(b) multi-grain clay fabric, (c) mixed fabric.

In contrast to sand, clay has typically a multi-grain fabric. If not an artefact from a
preexisting rock, multi-grain clay fabrics develop due to high surface forces and chemical activity. Assemblages of clay particle groups or clusters can be arranged within clay
fabric in several ways, i.e. edge-to-face or face-to-face. The particle arrangement within
the clusters themselves depends mainly on the clays mineralogy. Possible inter-particle
and intercluster forces in clay, in addition to the aforementioned ones, include electrostatic forces in the form of ionic bonds, hydrogen bonds, and hydration forces, as well as
capillary stresses.
Although sand and clay structures are different in many ways, their qualitative smallstrain behavior is not (see Chapter 3). Both soil types are modeled as granular assemblies
here, which represents the level of single-grain sand fabric, and multi-grain clay fabric.
Figure 4.2 gives an illustration of the above introduced sand and clay fabrics and the
more common case of mixed fabric.

40

4.2 Micromechanical considerations

t
th

c -contact

branch vector li
x

Global coordinate system

Local coordinate system

Figure 4.3: The cth inter-particle contact in global and local coordinates.

4.2 Micromechanical considerations

Assuming the particle stiffness to be infinite, the global material stiffness of a granular
material can be derived from its inter-particle contact properties (e.g. [27]):
Cijkl

M
1 X c c c
=
l k l
V c=1 i jl k

(4.1)

where V is the representative volume of the granular assembly, M is the total contact
c
is the
number in the volume V, lic is the branch vector as introduced in Figure 4.3, and kjl
contact stiffness tensor in the global coordinate system. In a local coordinate system the
contact stiffness tensor can be written in the form:
c
kjl
= knc ncj ncl + ksc scj scl + ktc tcj tcl

(4.2)

where knc , ksc , and ktc are the local contact stiffnesses along directions nc , sc , and tc respectively. Several homogenization techniques have been proposed in literature to rewrite
Equation 4.1 in an integral form. However, all these homogenization techniques involve
the definition of density functions for directional contact frequency, branch lengths, and
the definition of local stiffness properties. Postulating these density functions is not an
easy task for a non-ideal particle assembly: The perturbation function of branch vector
lengths for example should reflect particle gradation, shape and preferred particle orientation. As an alternative to explicitly introducing these density functions, numerical
discrete element codes can be employed. At the cost of calculation time, these codes
keep track of all inter-particle contacts.
Either way, using homogenization techniques or discrete element codes, deriving the
global material stiffness from Equation 4.1 is an elaborate task. Unfortunately, not even
the constitutive law for the inter-particle contacts is known exactly. Often the HertzMindlin theory (see Section 4.5) is used in combination with correction factors, bonding,
viscous components etc. Relating local inter-particle parameters to global soil parameters from laboratory or field testing is also difficult. Largely due to these reasons, micromechanical models are not yet used in engineering practice. Nevertheless, they can

41

Chapter 4 Small-strain stiffness at the soil particle level

often sufficiently explain experimentally observed material behavior. In the following
sections, micromechanical considerations are used to explain the observed dependence
of small-strain stiffness on confining stress, void ratio, cementation and strain amplitude.

4.2.1 The influence of strain amplitude

Assuming that small-strain stiffness is due to intermolecular and surface forces within
the soil skeleton, as pointed out for example by Israelachvili [72], it can only be altered
by rearranging these forces. If a simple Coulomb-type frictional law is assumed to exist
inbetween the particles of a soil skeleton, the influence of strain amplitude on smallstrain stiffness can be explained as follows: Initially, all inter-particle contacts are in a
sticking mode. Due to a certain normal and tangential contact-elasticity, most contacts
remain in the sticking mode for a finite strain amplitude. This finite strain amplitude
can be considered to be equivalent to the quasi-elastic very small-strain regime. Further
increase of the shear strain will then cause an increasing number of particles to slip.
c
Assuming Coulomb friction, the local contact stiffness kjl
of all sliding contacts c reduces
upon slipping. The global material stiffness, being the sum of all local contact stiffnesses,
decays according to Equation 4.1. Strain reversals switch back the models inter-particle
contacts to sticking mode so that the maximum small-strain stiffness is recovered. The
basic features of small-strain stiffness can therefore be readily simulated with a simple
frictional inter-particle law.
Rearrangements of a soils inter-particle forces can be visualized in chains of contact
forces. The connection between chains of contact forces and soil stiffness has previously
been pointed out by Kuwano & Jardine [96]. As long as there exists no clear directional
chains of contact forces (very small strains), the intermolecular and surface forces are
undisturbed. The more directional and concentrated chains of contact forces that are
formed, the more that intermolecular and surface forces are distorted, which leads to a
decrease in stiffness. To illustrate this, Figure 4.4 shows a biaxial test on a soil sample
modeled within a 2D discrete element program, in which a simple frictional law between
the particles is assumed. Inter-particle stiffness is modeled with a Hertz-Mindlin type
law which will be discussed in more detail in the next section. An additional small
bond between the particles is introduced upon the onset of loading in order to visualize
sticking contacts. As long as the predefined bonding force is not exceeded, no relative
movement between the bonded particles is allowed: The contact sticks. In this case the
intermolecular and surface forces between the particles are undisturbed. In Figure 4.4,
inter-particle contact forces are identified by lines pointing in the force direction. Line
thickness represents the force magnitude. Intact bonds between particles are identified
by small black lines that are orientated tangentially to the particle contacts.
Inter-particle force rearrangements, which are reflected in the force-chains shown in
Figure 4.4 are related to the type of loading: Deviatoric loading distorts inter-particle
forces more than isotropic loading. In case of psammic behavior (rigid grains), isotropic
loading causes only an upscale and no rearrangement of existing inter-particle forces.
Although soil particles are neither rigid nor unbreakable, the experimental data shown

42

4.2 Micromechanical considerations

a1

b1

a2

b2

Figure 4.4: Disturbance of inter-particle forces in a biaxial test. A bonded discrete element model is first consolidated (a) and then biaxially sheared (b). From
hydrostatic loading equally distributed chains of contact forces (a1) become
concentrated during deviatoric loading (b1). At the same time, most of the
initially sticking contacts (a2) slide so that the bonds break up (b2).

43

Chapter 4 Small-strain stiffness at the soil particle level

in Section 3.2.1 can therefore be explained.

4.2.2 The influence of confining stress and void ratio

The classical Hertz-Mindlin contact theory (Mindlin [118], Mindlin & Deresiewicz [119])
c
is employed to constitute the stiffness tensor kjl
introduced in Equation 4.1. The HertzMindlin theory relates the contact stiffnesses of two equally sized elastic spheres to normal and tangential contact forces via their radii and elastic moduli. The elastic spheres
are assumed to be smooth and perfectly rounded.
The original Hertz-Mindlin equation for normal contact tangent stiffness kn reads:
kn =

2GS
a,
1 S

(4.3)

where GS , and S are the elastic constants of the sphere material, and a is the radius of
the circular contact area, which is calculated as:
s
3N R(1 S )
a= 3
.
(4.4)
8GS
The normal contact force N , the contact radius a, and the sphere radius R are illustrated
in Figure 4.5. Contact shear stress distributions caused by different amounts of slip xT
inbetween the spheres are shown in Figure 4.5. The partial slip introduced in Figure 4.5
is a refinement of the previously discussed simple Coulomb-type frictional law. Energy
dissipation and stiffness reduction start with the onset of partial slip. Johnson [85] gives
as the resulting tangential contact stiffness:
s
2(1 S ) 3
T
ks = kt = kn
1
,
(4.5)
2 S
N tan S
where S is the inter-particle friction angle, and T is the shear force in the inter-particle
contact. Both normal and shear stiffness therefore increase with the normal contact force
to the power of m = 31 .
The empirical power law by Hardin & Richard [54] that relates initial stiffness to confining pressure (Equation 3.5) can therefore be readily understood from a micromechanical point of view. The power law exponent observed in experiments is generally somewhat higher though. More sophisticated, and for soil particles probably more adequate,
contact theories, yield these higher exponents. Rough surface contacts for example yield
a power law exponent of m 0.5 0.6 [197]. Meta-stable cone to sphere contacts yield a
power law exponent of m = 0.5 [45], too. Higher power law exponents for large strains
can be explained by slippage between the particles.
Local contact stiffness is not an explicit function of void ratio as can be seen in Equation
4.3, and 4.5. Therefore, the effect of void ratio is studied here by comparing two different
cubical assemblies: simple cubic (SC) packing and face-centered cubic (FCC) packing of

44

4.2 Micromechanical considerations

Section and plan view of the contact for...

no slip
xT/2

xT/2

xT/2

T
T

N
T

T
T

xT/2

xT/2

gross slip

contact
stress

partial slip

xT/2

T
T

b
a

xT/2

xT/2
b

Gross
slippage

No slippage
Zone of
slippage

Zone of
no slippage

Figure 4.5: Two elastic and equally sized spheres in contact.

Simple cubic packing (SC)

2R

Unit cell SC

Face-centered cubic packing (FCC)

2 2R

Unit cell FCC

Figure 4.6: Simple cubic packing (SC) versus face-centered cubic packing (FCC).

45

Chapter 4 Small-strain stiffness at the soil particle level

equally sized spheres. Both assemblies are illustrated in Figure 4.6 together with their
unit cell. SC packing has a coordination number of 6, and a void ratio of 0.91; FCC has a
coordination number of 12, and a void ratio of 0.35.
2
The SC unit cell has 3 contacts
3 in a volume of 8R , where theFCC unit cell has 24
contacts in a volume of 16 2R . Per unit volume, this results in 2 2 more inter-particle
contacts in the FCC packing than in the
SC packing. The force per contact in isotropically
loaded unit cells on the other hand, is 2 2 times less in the FCC packing. From Equation
4.3, the bulk stiffness ratio between FCC and SC packing becomes:

(k)F CC
=2 2
(k)SC

2 2

13

= 2.

(4.6)

Assuming Poisons ratio to be constant, the empirical relationship given by Hardin &
Richart [54] presented in Equation 3.8, suggests a bulk stiffness increase of 65% in the
FCC packing. Compared to the empirical value, the above deduced stiffness increase of
100% is too high. The difference is probably due to the simplification of real soil structure
in SC and FCC assemblies. Nevertheless, the stiffness increase is of the same magnitude.
The number of contacts per particle and particle bracing can therefore be assumed to be
the main factors, which lead to higher stiffness in real soils.

4.2.3 The influence of cementation

Figure 4.7 shows a cement coated contact of two spherical particles. Again, a simple
Hertz contact between the spherical particles is assumed. The cementing agent shall
have uniform thickness, the same material properties as the soil particles, and compatible deformations with the particles upon loading. As bonding increases the contact area,
it is readily apparent from Equation 4.3, that bonding also increases contact stiffness. For
spherical particles the cement coat thickness, t relates to cement contents by weight CC
as follows:
1
t
(CC + 1) 3 = 1 + .
(4.7)
R
Using the geometrical relationship between coat thickness and cement contents, Fernandez & Santamarina [39] derived the following expressions for cementing before and after
contact loading respectively:
v
u
32
u
2
3 1
u
(CC
+
1)
3
GS t
3(1 S ) N
(kn )bef ore =
(4.8)
+
1
1 S
2
4R2 GS
(CC + 1) 3
s

32
GS
3(1 S ) N
2
(kn )af ter =
(CC + 1) 3 1 +
(4.9)
1 S
2
4R2 GS
where kn is the local normal contact stiffness as introduced in Equation 4.2. In both cases,
cementation is causing a stiffness increase independent of confining pressure. With in-

46

4.2 Micromechanical considerations

creasing confining pressure, the relative importance of bonding decreases. If the number of particle contacts increases with decreasing void ratio, like for example in the FCC
packing, the effect of cementation is more pronounced in the denser configuration.

R
N
Soil
particle

TN
a

R+t
a0

Cementation

Figure 4.7: Cemented inter-particle contact of two spheres after Fernandez & Santamarina [39].

4.2.4 Summary
The influences of parameters that appear in almost all very small-strain stiffness relationship (see Table 3.2, 3.3, and 3.4) were studied at the inter-particle contact level. Most
previously presented experimental findings were qualitatively reproduced. In order to
obtain a quantitative reliable constitutive response of a micromechanical model, a number of conditions have to be fulfilled: First, the inter-particle contact needs to be defined
adequately, second the soil grains and its structure have to be reasonably well abstracted
and finally the problem of homogenization or calculation time must be solved. For the
moment, a small-strain stiffness model aimed for practical use can hardly be formulated
from a particle contact level. The following chapters therefore concentrate on continuum
models. Existing continuum small-strain stiffness models are discussed in Chapter 5. A
new small-strain model is developed in Chapter 6.

47

Chapter 5
Existing small-strain stiffness constitutive
models
The first small-strain stiffness constitutive models that were intended for static soil analysis date from the late 1970s. This chapter gives an overview of the developments in
the field of small-strain stiffness since then. It begins with a historical review followed
by a discussion of the most well known models and modeling directions. Some of the
models that will be introduced in this chapter should be classified rather as relationships
because they lack a constitutive response in non-monotonic loading patterns. Nevertheless they are included in this chapter, since they have been extensively used in the past
despite this shortcoming.
The models and modeling directions covered in more detail are:
The Simpson Brick model
Models known from soil dynamics
The Jardine model
Multi surface models
Intergranular Strain
The order of the above list is chosen according to model complexity rather than according the date they were proposed. Only continuum models are considered in this
chapter. Although micromechanical concepts might be very helpful in the model abstraction process, they cannot yet be used in the analysis of boundary value problems.

5.1 A brief history of small strain modeling

The curves shown in Figure 3.4 were used in soil dynamics for many years before being
compiled and published by Seed & Idris in 1970 [162]. They were not however applied
to static problems until about ten years later. That was when the first small-strain models for static applications were introduced. Amongst them were models that use one
or more kinematic yield surfaces, either in strain space as the model by Burland et al.
published in 1979 [25] or stress space as the model by Mroz et al. [126] which dates

49

Chapter 5 Existing small-strain stiffness constitutive models

back to 1978. Limiting the number of kinematic yield surfaces to one for the purpose of
modeling small-strain stiffness only, Al-Tabbaa [1], and Al-Tabbaa & Wood [2] created
the class of bubble models in 1987. In 1990 Stallebrass [170] introduced an additional
(kinematic) history surface for the bubble within the Cam-Clay framework and created
the well known 3-SKH model. Parallel to these models with kinematic surfaces, a different approach was taken by Jardine et al. [79] in 1986. By fitting stiffness-strain curves,
they directly calculated soil stiffness as a mathematical function of applied strain. In
his 1992 Rankine lecture, Simpson [168] made the analogy between soil behavior and a
man pulling bricks behind him. The analogy in fact can be used to explain the concept
introduced by Mroz et al. [126], as well as many other models. Simpsons analogy together with his interpretation in strain space became well known by the term Simpson
brick model and shall be the first to be discussed in detail.
Many refined models based on the ones mentioned above have been developed in
subsequent years. Advances in finding mathematical descriptions of the stiffness-strain
curve in the sense of Jardine [79] are for example documented by Gunn [48] in 1993 and
by Tatsuoka [177] in 2000. The class of bubble models have mainly been enhanced by
additional mechanisms, as for example the use of a second yield loci as found in the
Milan Model published in 2002 [201]. However, most of these models still use a loadingunloading criterion similar to that of a one-brick Simpson brick model, as does the model
by Papadimitriou et al. [134] and the Intergranular Strain concept, which is the smallstrain extension of Hypoplasticity introduced by Niemunis & Herle in 1997 [129]. The
Intergranular Strain concept makes use of a pre-defined interpolation function for the
materials strain history dependent stiffness. For this reason, Intergranular Strain falls
outside the mainstream modeling directions presented.

5.2 The Simpson Brick model

Simpsons brick model explains soil behavior on the basis of a man dragging some bricks
behind him, each of them connected to a string of unique length. Some of these strings,
or perhaps all of them, may slacken when the man heads in a new direction. Then, after
continuously walking in the new direction, strings will be tensioned one by one again.
An illustration of this analogy is presented in Figure 5.1.
Although these bricks could be interpreted as kinematic yield surfaces in stress space,
Simpson favors the idea that the man is a point in strain space that represents a soil element. The bricks attached to his back represent quantified portions of the soil element.
Bricks being dragged while the man moves, correspond to plastic strains, bricks with
slackened strings correspond to elastic strains. Hence, fully elastic behavior is modeled
whenever the man is moving without tensioning any strings. By selecting appropriate
brick sizes and string lengths, an entire S-shaped stiffness-strain curve can be approximated as for example shown in Figure 5.1. Short strings represent the behavior of smallstrains, long strings that of large strains. Strain history is automatically accounted for
in the model by remembering the bricks positions in strain space. Although the original Simpson brick model is rarely found to be closely implemented as for example in

50

5.3 Models known from soil dynamics

Soil stiffness

log (strain)

Figure 5.1: Simpson brick model with four bricks (left). Stiffness decay after load reversal
is a discontinuous approximation of the real stiffness-strain curve (right).

the model by Lansivaara [105], its basic idea of history tracking has been widely used.
Doing so, the number of bricks are often reduced to one only.

5.3 Models known from soil dynamics

Typical models used in dynamic soil analysis are mainly the bilinear, Ramberg-Osgood,
and Hardin-Drnevich models. The characteristics of bilinear models were studied by
Caughey [26] and incorporated in dynamic analysis for example by Idriss & Seed [68].
The Ramberg-Osgood model was first introduced by Ramberg & Osgood [145] and further developed by Jennings [81]. The Hardin-Drnevich model [52] is an extension of
Kondners [94] monotonic stress-strain relationship to the cyclic loading case. The hyperbolic law by Kondner is explored in more detail in Chapter 7. The Hardin-Drnevich
model is probably the most frequently used model in soil dynamics.
In shear-strain shear-stress space these models are defined as follows:

Bilinear:
Ramberg-Osgood:
Hardin-Drnevich:

G0
for < y
y + G0 ( y ) for y

=
1 +
G0
y
G0

=

1 + r
=

(5.1)
(5.2)
(5.3)

where G0 is the initial shear modulus, , and 1 are material constants, and y , y ,
and r are user defined threshold values for shear- strain and stress respectively. Rewriting the Ramberg-Osgood, and Hardin-Drnevich models in terms of normalized secant

51

Chapter 5 Existing small-strain stiffness constitutive models

t

t
G0
1

t
G0

G0

ty

G0

tf

ty

gy

gy

gr

Figure 5.2: Skeleton curves in -space of a) the bilinear model, b) the RambergOsgood model, and c) the Hardin-Drnevich model.

stiffness,

G
G0

shows the similarity and basic difference in these two models:

G
1

=

G0
1 + y

(Ramberg-Osgood)

(5.4)

G
1

=

G0
1 + r

(Hardin-Drnevich).

(5.5)

The Ramberg-Osgood defines stiffness decay as a function of stress whereas the HardinDrnevich model defines it as a function of strain. Other than that, the two formulations are very similar. The hysteresis effect described in Masings second rule (see Section 3.2.7) can be easily incorporated in both models using the expression proposed in
Hashiguchi [55]:

0 R
0 R = f
(5.6)
L
with shape Factor L = 1 in primary loading and L = 2 in unloading and reloading.
In Equation 5.6, f () represents the monotonic stress-strain relationship = f (). The
strain and stress levels 0 and 0 denote the last load reversal point in stress-strain space,
where R , and R specify the actual strain and stress respectively. Figure 5.2 shows examples of hysteresis loops derived by the bilinear, Ramberg-Osgood and Hardin-Drnevich
models.
The Ramberg-Osgood model has the probably unintended quality that it can be easily used in combination with stress space based elastoplastic models, as shown in Papadimitriou et al. [134]. For the coupling of the Ramberg-Osgood formulation to their
elastoplastic model, Papadimitriou et al. defined two scalar valued functions for the

52

5.4 The Jardine model

materials stress and strain history:
r =

(r rlr ) : (r rlr ) and e =

(e elr ) : (e elr )

(5.7)

where r and e are deviatoric stress and strain ratio tensors respectively and superscript
() denotes tensors that are associated to the last memorized load reversal point. A load
reversal is assumed whenever the scalar valued function e looses its monotony, e.g.
e e < 0. In terms of a one-brick Simpson model in strain space, a load reversal is
thus assumed whenever the string between the man and the brick is slacking. Then the
position of the brick is memorized and stiffness is calculated as a function of the distance
to this memorized point in stress space, e.g. as the resulting stiffness of the RambergOsgood formulation 5.4.
Small-strain stiffness models similar to the one published by Papadimitriou et al. can
be found in Hueckel & Nova [65] or Pestana & Whittle [137]. The model of Hueckel &
Nova mainly differs from the one by Papadimitriou et al. in the stress memory. The
model by Hueckel & Nova memorizes not only one, but several load reversal points
and does not explicitly use the Ramberg-Osgood formulation. Likewise, the hysteretic
equations of the MIT-S1 model described in Pestana & Whittle [137] as:
(
1
for unloading
2G/K
1+r
=
(5.8)
1
for reloading
2G0 /K0
1+p r
lr

are modified from the Ramberg-Osgood formulation. Here, is a material parameter

and the scalar valued functions p and r give the volumetric and the deviatoric stress
in respect to the last load reversal point ()lr :

p/plr for p < plr

p =
.
(5.9)
plr /p for p plr
The deviatoric stress measure r used in the MIT-S1 model is the same as the one defined
in Equation 5.7. Another difference in the hysteretic equations of the MIT-S1 compared
to the model by Papadimitriou et al. is the definition of load reversal points. The MIT-S1
model assumes load reversals whenever the volumetric strain history looses monotony.
Only if the volumetric strain increment is zero (e.g. undrained loading), the MIT-S1
model checks for the monotony of deviatoric strains. Then, a load reversal is again
assumed whenever e e < 0.
Note that all of the above stress-strain laws are not continuous in strain (load reversals)
so that by definition they do not qualify as elastic stress-strain laws. Nevertheless, strain
on a closed stress cycle starting from a stress reversal point is recovered. In the following,
the above stress-strain laws are therefore denominated paraelastic (Hueckel & Nova [65]).

5.4 The Jardine model

Jardine et al. [79] proposed a periodic logarithmic function to express the non-linear
relationship between normalized secant Youngs modulus Eu and axial strain a in an

53

Chapter 5 Existing small-strain stiffness constitutive models

Eu/c u

Observed, or projected maximum

2B

Peak deviator (end of stiffness data)

A
Projected function

ea= C

log10 e a
Projected minimum

Figure 5.3: Jardine model - secant stiffness defined as a trigonometric function of strain
(projected curve).

undrained triaxial test:

n h
i o
Eu
a
= A + B cos log10
.
cu
C

(5.10)

Today, this type of non-linear elastic stress-strain law is commonly known as the Jardine
model. Its empirical model constants A, B, C, , and can be determined from (smallstrain) triaxial test data. The secant Youngs modulus Eu is normalized with undrained
shear strength cu . Due to the Jardine models trigonometric nature, it is mandatory to
specify the exact strain range in which the non-linear stress-strain law is to be applied.
When exceeding the upper or lower limit of the strain range specified, stiffness is set as
constant (Figure 5.3).
The normalized tangent Youngs modulus Eut corresponding to the secant Youngs
modulus Eu can be derived by differentiating and rearranging Equation 5.10:
1

BI
Eut
= A + B cos(I )
cu
2.303

sin(I ),

(5.11)

where I = log10 Ca . Since most numerical procedures make use of the tangent modulus Eut instead of the secant modulus Eu , Equation 5.11 represents the more commonly
used form of the Jardine model. Jardine et al. [79] also mention in their original work
the possibility of expanding the proposed nonlinear stress-strain law to bulk and shear
moduli. Potts & Zdravkovic [138] propose the following notation for such an expansion:

54

5.5 Multi (or Infinite) surface models

c1

3G
p0

2
= C1 + C2 cos(c1 X c2 ) C2 c1 c2 X2.303
sin(c1 X c2 )

K
p0

4
= C4 + C5 cos(c3 X c4 ) C5 c3 c4 X2.303
sin(c3 X c4 ) with

1

2 2 [(1 2 )2 (2 3 )2 (1 3 )2 ]
2d
= log10 3C3 = log10
,
3C3

= log10 Cv6 = log10 1 +C26+3

X
Y

c1

and

(5.12)

where G is the shear modulus, K is the bulk modulus, p0 is the mean effective stress,
i are the principal strains, and the number of material parameters (C1 C6 , c1 c4 ) is
increased to 10. In its expanded version, the Jardine model implicitly varies Poissons
ratio as a function of strain. When selecting the models parameters the user needs to
be aware of this in order to keep Poissons ratio within its physical limits. Since the
Jardine model is often used in Great Britain, model parameters for many British clays
are available in the literature.

5.5 Multi (or Infinite) surface models

The first multi surface models were independently proposed by Mroz [125], and by Iwan
[73]. Back then, these models where developed to model the hysteretic Bauschinger
effect in metals. Later, multi surface models were mainly developed for soils, starting
with the model by Prevost [142], and this by Mroz et al. [126]. The common idea of all
these early models is to have a set of similar, nested yield loci fn that are kinematically
(and optionally volumetrically) hardened:

fn
= f (
ij )n rn f (q?,n )
(
ij )n = ij (ij )n

with

(5.13)

where rn is the ratio of the nth yield surface to the outermost, or distinct yield surface,
(ij )n gives the internal backstress of the nth yield surface, and q?,n is a suitable set of
scalar internal variables.
One of the most important features of multi-surface models is their ability to capture
recent stress history in their internal variables (mainly backstress). Small-strain stiffness
models based on the multi-surface concept associate small-strain stiffness with the innermost yield surface. By defining an appropriate set of hardening laws for all yield surfaces, such multi-surface models can give realistic predictions over a wide strain range.
A special subclass of kinematic multi-surface small-strain stiffness models is the class
of bubble models (Al-Tabbaa [1], [2]). Bubble models have only one small kinematic yield
surface (bubble), with elastic stiffness equal to the soils small-strain stiffness. Increasing
the number of kinematic yield surfaces to two, Stallebrass & Taylor [171] extended the
bubble model by Al-Tabbaa & Wood [2] to the 3-SKH model. Although the 3-SKH model
might be considered sufficiently complex, the number of yield surfaces could also theoretically be increased to infinite.

55

Chapter 5 Existing small-strain stiffness constitutive models

q

q
D

40

40

20
0
-20

20

A
20

B
40

C
60

80

100

20

40

60

80

100

-20

Stress path: 0ACBDB

-40

-40

Figure 5.4: Position of 10 yield surfaces after stress path 0ACBDB. The inner surfaces
have been dragged by the stress point. The outermost yield surface has never
been reached because of volumetric hardening: Plastic volumetric strains on
inner yield surfaces increase the size of all yield surfaces.

An illustration of the multi-surface concept is given in Figure 5.4. Here the evolution of
10 yield surfaces for a given load path is shown, using a discrete form of Houlsbys [62]
continuous hyperplastic formulation. Note that this formulation abandons the concept
of nested and non-overlapping yield surfaces, which goes back originally to Prevost
[142].

5.6 Intergranular Strain for the Hypoplastic model

In 1997 Niemunis & Herle [129] introduced a new tensorial state variable to the hypoplastic theory. The new state variable has been named Intergranular Strain since it
is meant to account for intermolecular and surface forces within the soil skeleton. The
normalized magnitude of Intergranular Strain is defined as:
=

||||
R

(5.14)

where R is a material constant associated with the very small-strain range. Evolution of
Intergranular Strain is controlled by the loading direction; loading history is memorized
as Intergranular Strain. Again, a one-brick like loading-unloading criterion is used:

= (1 ) : D f or : D > 0
(5.15)
D
f or : D 0
where D is the stretch tensor and R is a material parameter that controls the evolution
of Intergranular Strain. At the same time, R controls the nonlinearity of the resulting
stiffness-strain curve. Numerically integrating Equation 5.15 over time gives continuous
Intergranular Strain with 0 1. For = 0 stiffness is increased in all loading directions; for = 1 stiffness is a function of loading direction. Minimum stiffness is obtained

56

5.6 Intergranular Strain for the Hypoplastic model

Figure 5.5: Interpolated stiffness M as a function of angular change in the loading path
according to Equation 5.16 with L = 1, N = 0, mR = 5, and mT = 2. Response
curves are drawn at Intergranular Strains of = 0 to 1 in steps of 0.1 for
= 6.0 on the left and = 2.0 on the right respectively.

for monotonic loading, neutral loading gives a somewhat increased stiffness defined by
the multiplier mT whereas reversed loading gives maximum stiffness by applying the
multiplier mR . Loading directions in between are interpolated according to:

M = [ mT + (1 )mR ]L +

(1 mT )L : + N f or
(mR mT )L :
f or

: D > 0
: D 0

(5.16)

where M is the interpolated stiffness and is the direction of Intergranular Strain defined as = /||||. Material parameters mR , mT and affect the interpolation, where
purely increases the impact of on the interpolation and so together with R controls
the non-linear stiffness decay. The hypoplastic operators L and N represent the linear
and the nonlinear terms in the original hypoplastic equation respectively [93]. Assuming a straight monotonic loading history in a plane problem, the directional dependency
of M on the loading direction can be visualized as shown in Figure 5.5. For increasing
Intergranular Strain the directional dependency grows to its maximum at = 1.
The high number of parameters and also the model abstraction might pose a serious
problem for routine application of Intergranular Strain. Engineers working in practise
may simply not understand the meaning of the parameters they should provide. Different aspects of the practical usability of all the models introduced so far are discussed in
more detail in the next section.

57

Chapter 5 Existing small-strain stiffness constitutive models

5.7 A critical review of small-strain stiffness models

regarding their use in routine design
All the small-strain models discussed in this chapter can be divided into three subclasses:
(i) Non-linear (para-)elastic stress-strain laws, (e.g. models from soil dynamics, Jardine
model), (ii) Kinematic hardening elastoplastic models formulated in stress space, (e.g.
multi-surface models) and (iii) Other concepts that might include strain space based
formulations, (e.g. Simpson Brick model, Intergranular Strain).
Non-linear paraelastic stress-strain laws are generally deduced for a specific loading
path. Prior to application in boundary value problems, they need to be generalized.
Without such a generalization, the above introduced non-linear stress-strain laws cannot
be recommended for use in practical applications. In combination with an elastoplastic
approach, they may prove to be very suitable to use in small-strain stiffness modeling.
The Hardin-Drnevich model particulary qualifies for such applications as it combines
minimum material parameter input with reasonable modeling qualities (see Chapter 3).
In this respect, the Jardine model seems less adequate.
Kinematic multi-surface elastoplastic models, for example the 3-SKH model, can provide reasonable analysis results while the computational costs remain moderate. The
main obstacle in using these models in practical applications is probably the material
parameter selection process. The recent implementation of the 3-SKH model by Grammatikopoulou [46], expects for example, user input for a) the ratio of the size of the history surface to that of the outermost surface, b) the ratio of the size of the yield surface
to that of the history surface, and c) the exponent in the hardening function. Without relating these parameters to test data or providing trustworthy correlations, such models
cannot be recommended for routine design. On the other hand, providing such correlations might be an elaborate or even impossible task.
The remaining concepts discussed in this chapter lack also some substantial features of
real soil behavior (Simpson Brick model) or they afford too many (abstract) user parameters (Intergranular Strain) to allow them to be recommended for use in routine design.
However, it should be noted that many parameters in the Intergranular Strain concept
can be set to default values. Determination of the full set of Hypoplastic material parameters is still not an easy task to undertake.

58

Chapter 6
The Small-Strain Overlay model
All isotropic hardening, and most kinematic hardening elastoplastic models use yield
loci of finite size. Strategies to implement small-strain stiffness in just one very small
elastic domain (bubble models), or in many yield loci (multi-surface plasticity), have
been discussed in the previous chapter. However, models that are used in engineering practice today, rarely have more than one elastic domain. The Cam-Clay, or the
Hardening Soil model are just two examples of such models. It is impossible for very
small-strain stiffness to be associated with the entire elastic domain in these models as
they would respond too stiff in larger unloading-reloading cycles as a consequence.
Therefore, their elastic stiffness is generally taken as secant stiffness of larger strain
unloading-reloading loops. Small-strain stiffness is neglected. Here, the Small-Strain
Overlay model which will be developed in this chapter, comes into play.
The Small-Strain Overlay model is, similar to some of the models discussed in the previous chapter, a paraelastic approach. The overlay model can be used in combination
with elastoplastic formulations: If it detects small or very small-strain amplitudes, it increases the stiffness of the elastoplastic model accordingly. For larger strain amplitudes,
the overlay model is inactive, but still monitors loading history.
In order to guarantee the new models compatibility with many existing constitutive
models, it was decided to keep its output isotropic. Its objective multi-axial formulation
though accounts for strain induced anisotropy. In this chapter, the small-strain stiffness
overlay model is first formulated, and secondly validated.

6.1 Model formulation

6.1.1 Material history and history mapping
As pointed out in Chapter 3 and 4, decay of small-strain soil stiffness with straining
can be associated with inter-particle forces within the soil skeleton. Once the direction
of loading is reversed, the stiffness regains a maximum recoverable value. Then, while
loading in the reversed direction is continued, the stiffness decreases again. From a
modeling point of view, it is therefore essential to keep track of the materials strain
history. It has already been mentioned that deviatoric loading has a higher impact on
the re-arrangement of inter-particle force than hydrostatic loading. Hydrostatic loading
mainly increases force magnitude, whereas deviatoric loading additionally causes the
directional rearrangement of forces. In order to simplify the model and its calibration,

59

Chapter 6 The Small-Strain Overlay model

effects of hydrostatic strain and strain history are neglected in the following. Hence, the
model presented relies completely on the deviatoric strain history.
A criterion has to be defined for whether the actual strain rate is to be considered
monotonic or not. The criterion has to distinguish whether the actual strain rate points
in the same direction of the previous applied loading or not. In order to have a realistic
3D model response, this criterion should not only rely on a scalar valued test variable, i.e.
length measured in the deviatoric plane. Hence, the criterion proposed here considers
all three principal deviatoric strain directions separately, which resembles three independent brick models. When there is no principal strain rotation, the criterion reduces
to two independent brick-models.
In order to account for principal strain rotations, the strain history is stored in general
strain space. Transformed to the principal system of the actual strain increment, it is
then interpreted as a 2nd order surface. To avoid any possible singularities of the 2nd
order surface, a volumetric 1 strain is superimposed onto the deviatoric history strain.
A geometrical interpretation of 2nd order surfaces in some basic tests is shown in Figure
6.1. Here, the small deviatoric strains have been amplified to the power of 1e3 in order to
visualize the distortion of the volumetric 1 strain component, which is a perfect sphere
in this interpretation.
In a mathematical framework, determination of the actual strain history is as follows.
Let Hkl be a strain tensor that memorizes the deviatoric strain history. The actual deviatoric strain rate is denoted as e kl . By solving the Eigenvalue problem:

(m)
e kl (m) kl Sl = 0

(6.1)

a set of orthogonal Eigenvectors Slm for e kl is found. Unlike the transformed deviatoric
strain increment
e kl = Skm e mn Snl ,
(6.2)
the transformed strain history
H kl = Skm Hmn Snl ,

(6.3)

typically is not diagonal. However, extending strain history by a volumetric 1 strain

makes it a positive definite symmetric tensor. Its deviation from the volumetric 1 sphere
in the principal directions is given by the diagonal terms of H ikl . Each principal direction
is checked for possible reversed loading separately: Reversed loading is assumed whenever the sign of a diagonal term does not equal that of the corresponding Eigenvalue .
To reset strain history in reversed loading directions, a diagonal transformation matrix
T kl is defined as:

1 + u((1) H11 )( H11 + 1 1)

1
1 + u((2) H22 )( H22 + 1 1)
= H22
+1

1
(3)
1
+
u(
H
)(
= H33
H
+
1

1)
33
33
+1

T 11 =
T 22
T 33

60

1
H11 +1

(6.4)

6.1 Model formulation

e1

e1

e2

e2

e3

e3

e1

e1

e2

e2
e3

e3

Figure 6.1: Strain history interpreted as 2nd order surface in strain space. From left to
right: (a) hydrostatic loading, (b) triaxial extension, (c) simple shear, (d) simple shear after triaxial extension.

61

Chapter 6 The Small-Strain Overlay model

where u(x) is the Heaviside step function, defined as:

u(x) =

0 f or x < 0
1 f or x 0

(6.5)

The updated strain history H kl is then calculated as:

H kl = T km (H mn + mn )T nl kl .

(6.6)

From the updated strain history a scalar valued shear strain measure Hist is defined
next:
ke H k
Hist = 3 km ml
(6.7)
ke kl k

where k...k denotes the Hilbert-Schmidt norm kAk = aij aij , which can be proven to
conserve objectivity. In a geometric context, Equation 6.7 is the projection of the strain
history onto the actual loading direction. The shear strain measure Hist can then subsequently be used to define an isotropic material stiffness as shown in the next section. In
an incremental form, the model formulation is summarized in Box 6.1.

6.1.2 From strain history to stiffness

The modified Hardin-Drnevich relationship introduced in Chapter 3 is used next in relating scalar valued strain amplitudes Hist to material stiffness. Stiffness is described
by two independent elastic moduli. However, it should be noted that a material with
non-linear strain dependent stiffness as introduced here, is neither an elastic nor a hypoelastic material. Definitions of these material types and the consequences of being
neither of these types is discussed in Section 6.3.
Shear modulus
As an alternative to the formulation by Jardine [78] [79] or the Ramberg-Osgood formulation [145], the modified Hardin-Drnevich [52] relationship is used here to relate
deviatoric strain history to shear stiffness:
G=

G0
,
1 + a Hist
0.7

(6.8)

where G is the actual secant shear modulus, G0 is the initial shear modulus at very smallstrains, and Hist is the monotonic shear strain defined above. The parameter a and
the threshold shear strain 0.7 is discussed in more detail in Section 3.3.3. In numerical
applications, the secant modulus given in Equation 6.8 must be converted to a tangent
modulus:

2
0.7
.
(6.9)
G = G0
0.7 + aHist

62

6.1 Model formulation

In an incremental formulation, the actual elastic tangent shear modulus is best calculated by integrating the stiffness modulus reduction curve over the actual shear strain
increment. In this way, large strain increments will not miss out small-strain stiffness:

Gi+1 =

i+1
Hist

G0

i
Hist
1

i+1
Hist
0.385 i+1
+ 0.7Hist

i
Hist
0.385 i
+ 0.7Hist

(6.10)

where i and i + 1 denote quantities of the previous and the actual calculation step respectively.
A complete numerical implementation of the Small-Strain Overlay model that integrates over the stiffness reduction curve is shown in Appendix A. Here, a lower cut-off
of the models stiffness reduction curve is introduced. The cut-off should be activated in
combination with hardening plasticity formulations as otherwise, material stiffness for
larger strains would become too small (see also Section 7.3.1).
Bulk modulus
Assuming Poissons ratio to be constant (see Section 3.3) the bulk modulus can be calculated from the actual shear modulus as:
K=G

2(1 + )
.
3(1 2)

(6.11)

6.1.3 Inital loading and reloading

To systematically describe the behavior of brass under cyclic loading, Masing [111] proposed the following two rules (see also Section 3.2.7):
1. The shear modulus in unloading is equal to the initial tangent modulus for the
initial loading curve.
2. The shape of the unloading and reloading curve is equal to the initial loading
curve, except that its scale is enlarged by a factor of two.
Given that a functional form exists that can describe the initial loading curve, the above
rules can likewise be applied to construct the hysteresis loops of soils for symmetrical or
periodic loadings. However, if the loading is irregular, i.a. not symmetrical or periodic,
two additional rules are commonly added to the original Masing rules:
3. Unloading and reloading curves should follow the initial curve in case the previous
maximum shear strain is exceeded.
4. If the current loading or unloading curve intersects a previous one, it should follow
the previous curve.

63

Chapter 6 The Small-Strain Overlay model

1.0

1.0

t/ty

0.8

t/ty

C
1

0.8

A
0.6

0.6
3

0.4

0.4

0.2

0.2

g/ gy
-8.0

-6.0

-4.0

-2.0

2.0

4.0

6.0

8.0

g/ gy
-8.0

-6.0

-4.0

-2.0

2.0
4

-0.2

-0.4

-0.6

-0.6

-1.0

6.0

8.0

-0.2

-0.4

-0.8

4.0

-0.8

0
-1.0

Figure 6.2: Hysteresis loops in symmetric (a) and irregular (b) loading according to the
extended Massing rules (after Pyke [143]).

All four rules together are also referred to as the extended Masing rules. A discussion
of several possible extensions replacing the fourth of the above rules, which goes back
to Rosenblueth & Herrera [152], is given in Pyke [143]. Jennings [81], [82] for example
proposes that the upper and lower bounds of the stress-strain relationship could be given
by the reloading and unloading curves corresponding to the previous minimum and
maximum strain amplitudes. Richart [149] on the other hand, simply limits the shear
stress to a value no greater than the shear strength.
The Masing rule extensions by Rosenblueth & Herrera, Jennings and Richart are illustrated in Figure 6.2 for a one-dimensional simple shear problem. If point 2 is the previous minimum strain amplitude and curve 3 is the greatest previous reloading curve,
on reloading from point 4, the solutions suggested by Rosenblueth & Herrera and that
proposed by Jennings would follow path A. If, however, the dashed curve 0-1 is the
greatest previous reloading curve, the solution used by Jennings follows stress path B.
The solution used by Richart always follows path C.
The Small-Strain Overlay model adopts the first Masing rule by assuming G0 = constant in Equation 6.8. The second rule is obeyed by assuming a shape factor of 2 in the
relationship given by Hashiguchi [55] (see Equation 5.6) i.e. (0.7 )ur = 20.7 where (0.7 )ur
is the threshold shear strain in unloading and reloading. The third and fourth rule request an internal memory of previous changes in the stress-strain path. For all of the
above introduced Masing rule extensions, the memory i.a. requires loci in stress space.

64

6.1 Model formulation

A possible scheme of material memory rules is given in Hueckel & Nova [65]. However,
an internal memory in stress space is not a practicable constitutive ingredient to the
Small-Strain Overlay model, which is not intended to be used as a stand-alone model.
In combination with an elastoplastic model, the memory would be highly influenced by
plastic yielding, so that the response of the overlay model could only be determined in
an iterative scheme. The yield loci of an combined elastoplastic model, e.g. the later
discussed Hardening Soil model, can basically enforce the third Masing rule.
The material memory required to enforce the second Masing rule can on the other
hand be defined in strain space by memorizing the minimum stiffness in loading history
Gmin : For a given strain increment, it is always assumed first that the material is unloaded
or reloaded. If then
G0
G=
Gmin ,
(6.12)
Hist
1 + a (0.7
)ur
holds, the assumption is verified. Otherwise, primary loading is indicated.
A more convenient description of the minimum stiffness in the loading history can be
given by normalized shear moduli. The elastic stiffness Gur of a combined elastoplastic
model can serve as a normalization measure. The small-strain stiffness multiplier Gm is
defined as:
Gmin
Gm =
.
(6.13)
Gur
If the Small-Strain Overlay model is combined with an elastoplastic model that hardens from the onset of primary loading, the above described implementation of Masings
second rule could be accomplished differently. Hardening plasticity can readily account
for more rapidly decaying small-strain stiffness during initial loading. If the reloading
threshold shear strain is also used in initial loading, material hardening of the elastoplastic model should equal the difference between initial loading and reloading as shown in
Figure 6.3. This requires an elastoplastic model that would allow an adaption to its
hardening laws in the small-strain range. One possibility, how such a modification can
be accomplished is shown in the next chapter.

6.1.4 Effects of mean stress, void ratio, and OCR

Many of the influences on small-strain stiffness discussed in Chapter 3 can be easily incorporated into the Small-Strain Overlay model. The incorporation of mean stress, void
ratio, or OCR effects for example, can be obtained by deploying the given correlations
to the models parameters G0 and 0.7 . User input is necessary for a set of parameters at
well defined reference void ratio, reference mean stress, etc. A detailed example of the
incorporation of mean stress dependency into the Small-Strain Overlay model is given in
the next chapter. What follows in the next section is a verification of the basic model with
fixed input parameters. An exception is made only when cyclic mobility is considered.
Here, a power law of the form:
p0 m
)
(6.14)
G0 = Gref
(
0
pref

65

Chapter 6 The Small-Strain Overlay model

scales the initial shear modulus. The shear modulus Gref
0 is specified at the reference
ref
mean stress p .

6.2 A first model validation in element tests

For its first verification and validation, the Small-Strain Overlay model is not combined
with any elastoplastic model. Triaxial loading paths render the overlay models threebrick criterion for load reversals to a one-brick model. Applying the Small-Strain Overlay model in triaxial conditions can therefore be considered as code verification and validation of the modified Hardin-Drnevich relationship only. The Hardin-Drnevich relationship, however, can be assumed to be sufficiently validated for its use in a constitutive
model (see for example Section 3.3.3). Not many triaxial test data are therefore presented
here. Unfortunately, there are no biaxial, true triaxial, or hollow cylinder data known to
the author, which could be used in the validation of the models strain history mapping.
Numerical simulations of a biaxial test published in [129] are used instead.

6.2.1 Triaxial tests

Almost all available stiffness-strain curves are derived in triaxial extension or compression. Figures 6.4 and 6.5 present triaxial test results by Rivera & Bard [16] and Richardson [148] for dense sand and reconstituted London clay respectively. The tests by Rivera
& Bard comprise static and cyclic (repeated) triaxial loading. The test by Richardson
gives shear stiffness for different loading histories: four samples were consolidated to
point A (Figure 6.5 right), from where on the samples were brought to point O on different paths (AOX, (AO)BOX, (AO)COX, and (AO)DOX). Finally, the stiffness on the path
from point O to X was studied as shown in Figure 6.5. In both experiments it turns out
that the models ability to capture differences between initial or monotonic loading and
reloading is essential. As shown on the right hand side of Figure 6.4, Masings rule is
fulfilled. In Richardsons experiment, the differences in paths AOX and COX could not
be simulated by the Small-Strain Overlay model as it does neither account for the higher
pre-consolidation pressure in path AOX, nor for the decreased void ratio. For larger
strains, stiffness is too small in all loading paths. In the small-strain range, however,
both materials tested, could be simulated reasonably well with the Small-Strain Overlay
model .

6.2.2 Biaxial test

Figure 6.6 illustrates stiffness reduction curves in strain controlled plane strain biaxial
loading. First, equal strain increments are applied in both axial directions. From there
on, three different loading paths are examined: (1) Monotonic loading - strain is equally
increased in both directions; (2) Loading to the side - strain is equally increased in both
directions, but one of them is reversed; (3) Reversed loading - both loading directions
are reversed. As no small-strain biaxial laboratory test data are presently available, the

66

6.2 A first model validation in element tests

Box 6.1: Algorithmic setting for the Small-Strain Overlay model.
0
1. Initalize strain history. Default setting is Hkl
= 0.
ref
2. Adapt Gref
0 and 0.7 to the actual material state (optionally):
ref
G0 = f (Gref
0 , p, e, ...) and 0.7 = f (0.7 , p, e, ...)

3. Calculate shear strain Hist :

i
to the strain increments Eigensystem:
a) Transform Hkl
(m)

i
(m)
H ikl = Skm Hmn
Snl where (ei+1
kl )Sl
kl

=0

b) Calculate monotonic shear strain at time i:

kei+1 H iml k
i
km
= 3 ke
Hist
i+1
k
kl

c) Calculate transformation matrix T ikl :

1
(1)
T 11 = H11
1
+
u(
H
)(
H
+
1

1)
11
11
+1

1
(2)
H
+
1

1)
T 22 = H22
1
+
u(
H
)(
22
22
+1

1
T 33 = H33
1 + u((3) H33 )( H33 + 1 1)
+1
T ij

= 0 for i 6= j
(
0 f or x < 0
where u(x) =
1 f or x 0
d) Update strain history:
i
i
i
i+1
H i+1
kl = T km (H mn + mn )T nl + ekl kl

e) Calculate monotonic shear strain at time i + 1:

kei+1 H i+1 k
i+1
km
ml
Hist
= 3 ke
i+1
k
kl

4. Calculate tangentstiffness Gi+1 (unloading/

! reloading):
Gi+1 =

G0
i+1
i
Hist
Hist

i+1
Hist

0.385 i+1
1+ 2 Hist
0.7

i
Hist
0.385 i
1+ 2 Hist
0.7

5. IF Gi+1 < Min{G1 , G2 , ..., Gi } THEN (primary loading):

a) Set 0.7 = 0.7 /2
b) Recalculate stiffness (4)
END IF

67

Chapter 6 The Small-Strain Overlay model

Esec [kN/m2]

reloading

4E+5

primary loading

2E+5

plastic stiffness reduction

during primary loading

0
-5
10

-4

10

-3

10

e1[-]

Figure 6.3: Stiffness reduction in initial- or primary loading and in reloading.

Esec [kN/m2]

q [kN/m2]
100

4E+5
50

0
2E+5
-50
Calculated
Rivera/Bard (initial)

-100

Rivera/Bard (reloading)

0
-5
10

-4

10

-3

10

e1[-]

-4

-5x10

-4

-3x10

0x10

-4

3x10

-4

5x10

e1[-]

Figure 6.4: Standard drained triaxial test by Rivera & Bard (after Biarez & Hicher [16])
and simulation with the Small-Strain Overlay model (G0 = 185MPa, = 0.20,
0.7 = 2e4 ).

68

6.2 A first model validation in element tests

GTangent [MPa]
40

q [kPa]
DOX
AOX

150

100

COX

30

BOX

50

Calculated (HS)
20

0
0

100

O
200

A
300

400

p' [kPa]

-50
10
-100

0
10-4

10-3

10-2

e1[-]

-150

Figure 6.5: Triaxial test by Richardson [148] and its simulation with the Small-Strain
4
Overlay model (Gref
0 = 42 MPa, = 0.20, 0.7 = 5.5e ) Left: path dependent
shear stiffness. Right: loading paths in q-p space.

model response of the Small-Strain Overlay model is compared to that of the Intergranular Strain concept.
Both, the model proposed here and the Hypoplastic model with Intergranular Strain
show a similar behavior. The least stiffness is obtained during monotonic loading; highest stiffness is obtained for a full load reversal, which in the case of biaxial loading is
a reversal in both loading directions. Reversing only one biaxial loading direction and
keeping the other constant, results in a somewhat increased stiffness which lies between
the minimum and the maximum stiffness. Here the Small-Strain Overlay model clearly
shows its ability to capture deviatoric strain history. Loading to the side yields less stiffness than reversed loading.

6.2.3 Strain response envelopes

Strain response envelopes were originally introduced by Lewin & Burland [101] and
later used by Gudehus [47] as a tool to evaluate constitutive models. In their simplest
form, response envelopes for axisymmetric stress states are constructed by applying
stress increments of fixed magnitude but at different directions to a triaxial probe with
an initially predefined stress state. Connecting the single strain-responses of all tested
stress directions yields the strain response envelope.
Doanh [32] experimentally derived three strain response envelopes for Hostun sand,
for different initial stress states. A first response envelope is obtained after isotropic
consolidation to 100 kPa. A second envelope is obtained at an intermediate stress state,

69

Chapter 6 The Small-Strain Overlay model

e1 Biaxial test e3=0

3
1

2
e2

Figure 6.6: Stiffness-strain decay in a biaxial test on sand. Comparison of the tangent
stiffness generated by the Small-Strain Overlay model to that, generated by a
Hypoplastic model with Intergranular Strain range.

which is reached by isotropic consolidation to 100 kPa and application of 100 kPa vertical stress afterwards. The magnitude of the then applied stress increments is 10 kPa.
Experimentally tested stress increment directions are shown in Figure 6.7. Only the intermediate stress state is considered here.
Doanhs strain response envelopes are derived in the triaxial plane. In the triaxial
plane however, the unloading-reloading criterion of the Small-Strain Overlay model reduces to a one brick criterion. A sharp stiffness jump is obtained whenever the deviatoric
strain increment reverses its direction. In order to avoid sharp stiffness jumps around
hydrostatic loading, an alternative to the above model formulation is explored in Figure
6.7. Assuming Poissons ratio not to be constant, but varying from = 0.05 0.25 in the
case of the sand tested, smooth stiffness transitions could be obtained: The bulk modulus is assumed constant as long as Poissons ratio is within an admissible range, i.e.
min . In the experiment, the bulk modulus is therefore unaffected by the non-linear
small-strain shear stiffness decrease. However, as such large variations in Poissons ratio
within the small strain range is not supported by experimental evidence, it is not further
considered in this thesis.
In Figure 6.7 experimentally derived strain responses are compared to these derived
by the Small-Strain Overlay model. Strain responses after load reversals are modeled
reasonably well except for those which involve more soil plasticity. However, a combination of the Overlay model with an elastoplastic model can also reproduce monotonic
and almost monotonic strain responses.

70

6.2 A first model validation in element tests

6x10

15

-4

125

(a)

90

90

10

61

125

4x10-4
145

135
145

35

61

2x10-4
0

180

193

-5

180

e1 [-]

s1 [kN]

(b)

135

0x10

215

35
230

215
230

-15
-15

-2x10

305

243

-10

-4

270
-4

-10

-5

10

-4x10
-4
-6x10

15

-4

-4x10

0x10

-4

2x10

-4

4x10

e3 2 [-]

s3 2 [kN]
-4

6x10

-4

-2x10

125

Plot (b):

(c)

135

Test by Doanh
linear elastic

90

small-strain overlay

-4

4x10

145
61
-4

2x10

e1 [-]

180
215

35
230

0x10

Plot (c):
-4

-2x10

Test by Doanh
elastoplastic
elastoplastic with small-strain overlay

-4

-4x10
-4
-6x10

-4

-4x10

-2x10

-4

0x10

-4

2x10

4x10

-4

e3 2 [-]

Figure 6.7: Strain response envelopes experimentally derived by Doanh [32] and their
simulation with the Small-Strain Overlay model: (a) strain increment directions, (b) Small-Strain Overlay model, (c) Small-Strain Overlay combined
with hardening plasticity.

71

Chapter 6 The Small-Strain Overlay model

q [kPa]
300

200

100

-100

-200

Calculated
Rivera/Bard (Monotonic)

-300
-3
-3
-3x10 -2x10

10

-3

0x10

-3

10

-3

2x10

-3
3x10 e1[-]

Figure 6.8: Strain accumulation in closed cycles due to mean stress dependent elastic
stiffness.

6.2.4 Cyclic mobility

Extending the basic Small-Strain Overlay model with pressure or void ratio dependent
formulations leads to unrecoverable deformations in cyclic loading (Figure 6.8). Such a
possible extension is for example, the adaption of the models initial shear modulus G0 to
mean pressure as shown above. In a drained problem, the unrecoverable deformations
may lead to small, but constant ratcheting. In an undrained problem, cyclic mobility will
eventually be reached. This behavior, although close to experimental evidence, needs to
be considered for combining the overlay model with any elastoplastic model. Void ratio
dependent dilatancy concepts can stabilize the observed ratcheting. In case the SmallStrain Overlay model is not used explicitly in cyclic modeling, or if only a few cycles are
modeled, the aspect of cyclic mobility can be neglected.

6.3 Thermodynamic considerations

Elastic materials are notdependent on their stress or strain history. Thus the stress in an
elastic material at time t depends only on the local configuration at time t, and not on the
past stress or strain history. Hypo-elastic materials, although history dependent, assume
the stress rate to be linear and isotropic in stretching. Hyperelastic materials conserve
energy in closed cycles, which means that hyperelasticity can only be applied to either
perfect materials at uniform temperature or uniform specific entropy, or to simple materials in thermal equilibrium at uniform temperature or uniform specific entropy. For
a definition of perfect materials and simple materials see for example Truesdell & Noll
[181]. The material introduced here belongs to neither of these classes of elastic material.
For a rigorous mechanical description, a more complex theory has to be applied, for example Colemans general thermodynamics of simple materials. In this theory, attention

72

6.3 Thermodynamic considerations

is also paid to the Clausius-Duhem inequality. In a local form, this inequality can be
expressed as [181]:
1
1
q

divh 2 h grad +
(6.15)
%
%

where denotes specific entropy, % mass density, temperature, q heat absorption, and
h is the heat flux vector.
Constitutive equations have then not only to be defined for stress, but also for specific
entropy, internal energy, and heat flux. This, however, is not within the scope of the
Small-Strain Overlay model, which simulates near elastic material behavior with a relatively small change in specific entropy compared to the overall entropy change during
plastic loading of the background model. Nevertheless, a rigorously formulated constitutive model should obey thermodynamic principles. The model has not, however, been
demonstrated to obey such principles, although its constitutive parts have been chosen
to represent soil behavior realistically.
An example of the potential problems due to the non-rigorous thermodynamic formulation of the Small-Strain Overlay model is given next. Considered are isothermal
processes, starting from a state of equilibrium only. For such processes, the theorem on
isothermal cyclic processes hold (e.g. Truesdell & Noll [181]: In every isothermal process
starting from a state of equilibrium, the total stress work around a closed path is non-negative.
Partly as a consequence of not having implemented the entire set of Masing rules in
the Small-Strain Overlay model, the irregular cyclic loading situation shown on the left
hand side of Figure 6.9 may violate the theorem on isothermal cyclic loading processes.
The same loading situation, however, with less negative stress work release during unloading does not violate the above theorem (Figure 6.9 right hand side). The unloading
stiffness in the second case is calculated as follows:
Let t0 be the time at the onset of loading. For any time t1 > t0 the total stress work
associated with the overlay model is given by:
Z

t1

W (t1 ) =

(kl )ol (kl )ol dt

(6.16)

t0

The stiffness of a linear elastic material with

Poissons ratio identical to that of the overlay model
Loading history identical to that of the overlay model
Elastic energy identical to the stress work W (t1 )
calculates as:
el
(t1 ) = (ij )ol (t1 )(kl )ol (t1 )1 .
Cijkl

(6.17)

The shear modulus Gel (t1 ) of the iso-energetic linear elastic material can thus be explicitly
determined. In order to avoid negative stress work in a closed path simulated by the

73

Chapter 6 The Small-Strain Overlay model

q [kPa]

q [kPa]

250

250

200

200

150

150

100

100

50

50

-4

-2x10

0x10

-4

2x10

-4

4x10

-4

6x10

-4

8x10

e1[-]

-4

-2x10

0x10

-4

2x10

-4

4x10

-4

6x10

-4

8x10

e1[-]

Figure 6.9: Stress work in the Small-Strain Overlay model.

overlay model, its stiffness Gol (t) in unloading should at all times t > t0 be greater than
or equal to that of the corresponding linear elastic material:
(t) < 0.
Gol (t) Gel (t) t > t0 with W

(6.18)

In this way, it is ensured that the total negative stress work during unloading is not
less than that of a linear elastic material with identical internally stored strain energy at
the onset of unloading. Furthermore, as the positive stress work of the materials before
unloading are equal, the total stress work in a closed path cannot be negative.

74

Chapter 7
HS-Small, a small-strain extension of the
Hardening Soil model
In this chapter, the Small-Strain Overlay model is combined with an existing doublehardening elastoplastic model. As the main focus of this thesis is with the development
of a model for routine design, the combination is made with the Hardening Soil (HS)
model as implemented in the finite element code P LAXIS V8. The small-strain enhanced
version of the HS model resulting from the combination is referred to as the HS-Small
model. Compared to the HS model, the HS-Small model additionally offers a MatsuokaNakai [113] [114] failure criterion (optional) and a refined flow rule (mandatory). Following a brief introduction to the basic equations of infinitesimal elastoplasticity, both
models constitutive equations are developed within this chapter.
Because of the availability of two yield criteria, it is generally distinguished between
the HS-Small(MC) model with Mohr-Coulomb failure criterion, and the HS-Small(MN)
with Matsuoka-Nakai failure criterion. The terminology HS-Small is used whenever the
actual type of yield criterion is not of interest for a particular discussion. The term HSSmall is thus a collective term for HS-Small(MC) and HS-Small(MN).
Within this chapter, all stresses are effective although not explicitly indicated by a
prime. In triaxial conditions, 1 = axial , denotes axial strain and 2 = 3 = lateral , denotes
lateral strain. In the following, it is generally assumed axial /lateral 1. The reader can
easily verify all triaxial equations to hold likewise in triaxial extension (axial /lateral < 1)
by substituting 1 with 3 and vice versa.

7.1 Constitutive relations for infinitesimal plasticity

Elastoplastic materials can be described by means of the following set of constitutive
equations (e.g. Hill [58], Drucker [34], or Hughes [66]):
ij = eij + pij
ij = Dijkl ekl
g
dpij = d
ij
dq = dh (ij , q )

(7.1)
(7.2)
(7.3)
(7.4)

where ij , eij , and pij denote the total, elastic, and plastic strain tensor respectively, ij
is the Cauchy stress tensor, q is a suitable set of scalar or tensorial internal material

75

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

variables, and Dijkl is a fourth-order tensor describing the materials elastic stiffness. The
asterisk in place of indices in q replaces n indices of tensorial functions which vanish for
scalar functions.
From top to bottom, Equation 7.1 gives the additive decomposition of total strains in
an elastic and a plastic part, which is generally assumed in elastoplasticity. Equation 7.2
expresses the elastic stress-strain relationship by means of a Hookes law with the elastic
stiffness fourth-order tensor Dijkl . Equation 7.3 gives the flow rule for the plastic strain.
Depending on the choice of the plastic potential function g = g(ij , q ) in respect to the
yield function f = f (ij , q ), the flow rule is either associated f = g, or non-associated
f 6= g. The hardening laws expressed in Equation 7.4 complete the basic set of constitutive equations by introducing a set of evolution functions h for the internal variables.
In the last equation, d represents the plastic multiplier that is to be determined with the
aid of the Karush-Kuhn-Tucker conditions:
f (ij , q ) 0
d 0
f d = 0

(7.5)
(7.6)
(7.7)

The Karush-Kuhn-Tucker, must simultaneously hold up in any process of loading and

hence, distinguish between elastic and plastic loading processes. The yield function
f (ij , q ) encloses the elastic material domain. The material is elastic whenever f < 0
holds. In this case Equation 7.7 yields d = 0. Plastic flow is characterized by f = 0 and
d > 0, while neutral loading is obtained for f = d = 0. From the Karush-Kuhn-Tucker
conditions, the consistency condition for plastic flow can be derived as follows:
f
f
df =
dij +
dq = 0.
(7.8)
ij
q
Combining the consistency condition introduced above with the basic set of constitutive
equations (Equation 7.1 to 7.4), the plastic multiplier d can be calculated as:
d =

f
D dkl
ij ijkl
f
g
D
+
ij ijkl kl

with h =

f
h .
q

(7.9)

If all derivatives in Equation 7.9 are constant or linear functions in stress and the internal
variables, the plastic multiplier may be found analytically. However, commonly used
yield functions and plastic potentials do not satisfy these requirements. Calculation of
the plastic multiplier and at the same time the integration of the constitutive equations
therefore requires an efficient iterative scheme.
Many constitutive laws can be formulated within the framework of elastoplasticity
presented above. These laws differ only in their yield function, plastic potential, and
evolution laws for the internal variables. When functions of strain are applied the above
formulation can also be recast in strain space in terms of { ij , pij }, by eliminating ij
with the help of Equation 7.1 and Equation 7.2.
The implications of large strains are not considered explicitly in this thesis. However,
it should be noted that conserving objectivity, for example by the consideration of a
Jaumann term in the stress-rate, will not alter the following model description.

76

7.2 The Hardening Soil model

q
qa

asymptote

qa
qf

asymptote
1

E50

E50
qf

qa

Eur
1

e1

e1

Figure 7.1: Hyperbolic stress-strain law by Kondner & Zelasko (left) and its modification
after Duncan & Chang (right).

7.2 The Hardening Soil model

The original HS model was developed by Schanz [157] and Schanz et al. [158] on the
basis of the Double Hardening model by Vermeer [183]. Thus, the HS model also comprises ideas by Kondner [94], Duncan & Chang [36], Ohde [131] or Janbu [77], and Rowe
[153]. Standard lab tests, such as triaxial and oedometer tests provide the models basic
characteristics.
In drained triaxial primary loading, the experimentally observed relation between axial strain and deviatoric stress in soils can be well approximated by a hyperbolic function. Kondner & Zelasko [94] described the hyperbolic stress-strain relation for drained
triaxial loading as follows:
1 = 50

q
2 sin
q50
qa
with qa =
(3 + c cot ) and 50 =
=
qa q
1 sin
E50
2E50

(7.10)

Duncan & Chang [36] based their hypoelastic model on the above formulation by Kondner & Zelasko, additionally introducing the deviatoric measure qf in the form:
1 = 50

q
2 sin
qf
f or q < qf =
(3 + c cot ) and qa =
qf
qa q
1 sin
Rf

(7.11)

The conceptual difference in the formulations by Kondner & Zelasko and that by Duncan
& Chang is illustrated in Figure 7.1. Extending the hypoelastic Duncan-Chang model to
an elastoplastic formulation, Schanz [157] proposed the following yield function:
fs =

qa
q
2q

ps
E50 qa q Eur

(7.12)

where ps is an internal material variable for the accumulated plastic deviatoric strain,
q = 1 3 is defined for triaxial loading, and qa is the asymptotic deviatoric strain

77

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

700
600

re

1200

om
b fa
ilu

1000

300

600

400

200

p
on
g =c

st.

200

100
0

-Co
ul

800

Mo
hr

400

q (kN/m)

q (kN/m)

q (kN/m)

500

100

200

300

400

500

p (kN/m)

200

400

600

800

p (kN/m)

p (kN/m)

Figure 7.2: Experimentally derived shear strain contours versus HS yield loci. Left:
Shear strain contours of loose Fuji River sand [71]. Middle: Shear strain contours of dense Fuji River sand [178]. Right: Yield loci of the HS model for
stress independent stiffness (m=0).

as defined in the original Duncan-Chang model (Equation 7.11). As the stress-strain

relation of soils in unloading and reloading can be approximated by a linear function,
isotropic elasticity inside the yield function is another assumption of the HS model: The
elastic unloading-reloading stiffness Eur relates elastic stress to elastic strain.
For constant volumetric strain, the equivalence of Equation 7.12 with the approach by
Duncan & Chang is given by defining:
ps = p1 p2 p3 and thus ps = 2p1

(7.13)

as then the following relation holds:

1 = e1 + p1 =

q
1
qa
q
q50 q
q
fs =
=
= 50
.
Eur 2
2E50 qa q
E50 qa q
qa q

(7.14)

In accordance with the basic idea of Vermeer [185], the defined yield loci are lines
of constant plastic shear strain in p-q space. The striking similarity between experimentally derived shear strain contours and the yield loci defined by Equation 7.12 is depicted
in Figure 7.2. All the shear strain contours shown in Figure 7.2 (left and middle) were
experimentally derived by Tatsuoka & Ishihara [178], and Ishihara et al. [71]. Their
triaxial testing procedure was as follows: They first sheared a given sample at mean
stress p1 to shear stress q1 . After reducing shear stress (unloading) and increasing mean
stress to p2 , they sheared the sample again until it yielded at shear stress q2 . This procedure was repeated many times so that the regression line through the n-tuple of points
{(p1 , q1 ), (p2 , q2 ), , (pn , qn )}, called the shear strain contour, marks a line of constant
shear strain in p-q space.
When translating Equation 7.12 to principal stress space, the resulting cone-type yield

78

7.2 The Hardening Soil model

function can only be formulated in stages:
s
f12
=
s
f13 =

2qa (1 2 )
Ei qa (1 2 )
2qa (1 3 )
Ei qa (1 3 )

2(1 2 )
Eur
2(1 3 )
Eur

ps , and
ps .

(7.15)

The transition from E50 to Ei is made because of a second yield surface that will be introduced later in this section. The second yield surface will affect material stiffness such
that, the meaning of Ei in the full HS model is not as closely related to the hyperbolic
model by Kondner & Zelasko as the one of E50 .
Associated plasticity is an unrealistic assumption for most geotechnical materials.
Thus, an additional plastic potential of the form:
s
g12
=
s
g13 =

(1 2 )
2
(1 3 )
2

1 +2
2
1 +3
2

sin m , and
sin m

(7.16)

is introduced. The mobilized dilatancy angle m in the above equation is defined according to Rowes stress dilatancy theory [153]:
sin m =

sin m sin cs
1 sin m sin cs

(7.17)

where, cs is the critical state friction angle and the mobilized friction angle m is calculated as follows:
1 3
sin m =
.
(7.18)
1 + 3 + 2 c cot
The stiffness moduli Ei and Eur are scaled for their stress dependency with an Ohde
[131] or Janbu [77] power law:
cot m
) , and
Ei = Eiref ( pref3 +c
+c cot
ref 3 +c cot m
Eur = Eur ( pref +c cot )

(7.19)

where Eiref and Eurref are the material stiffness moduli at the reference pressure pref , and m
is the exponent of the power law. In Equation 7.19 the minor principal stress 3 is used
as an indicator of the actual stress state in the material instead of the mean stress p = 3ii .
The cone-type yield loci of the HS model mainly accounts for plastic deviatoric strains,
or shear hardening. A second cap-type yield surface is introduced next. The cap-type
yield surface accounts for plastic volumetric strains, or volumetric hardening. Volumetric hardening corrects too stiff primary oedometric or isotropic loading, obtained in pure
shear hardening models. The HS models cap-type yield surface is defined as follows:
q2
p2 p2p
(7.20)
2
where p = 3ii is the mean stress, is an internal material constant, controlling the steepness of the cap in p-q space as shown in Figure 7.4, pp is an internal material variable for
pre-consolidation stress, and q is a special stress measure, defined as:
fc =

q = 1 + ( 1 1)2 1 3 with =

3 sin
.
3 + sin

(7.21)

79

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

s1

s1

q
s2

s3

s2

s3

Figure 7.3: Yield surfaces of the HS model for cohesionless soil. Left: Cap and cone-type
yield surfaces in principal stress space with the cone being in its ultimate
Mohr-Coulomb failure criterion position. Right: p-q slice through the yield
surfaces.

The definition of the special stress measure q is necessary to adopt the cap-type yield
surfaces deviatoric shape to this of the cone-type yield surface as shown in Figure 7.3.
The plastic potential for the cap-type yield surface is chosen equal to its yield surface
(g c = f c ), so that plastic strain on the cap-type yield surface is associated, in contrast to
plastic strain, on the cone-type yield surface.
A total of two state variables are used within the HS models yield and potential functions: The plastic deviatoric strain ps , and the pre-consolidation stress pp . Their evolution laws are defined as follows:
d ps = ds h ps
dpp

= dc hpp

g
g
g


)=1
with h ps = (
1
2
3
m
3 +c cot
with hpp = 2H pref +c cot
p

(7.22)

where m represents the power law exponent, and H relates plastic volumetric strain
pv = p1 + p2 + p3 to pre-consolidation stress pp as follows:

m
3 + c cot
dpp = H
dpv .
(7.23)
pref + c cot
In decomposing volumetric strain in elastic and plastic contributions, H can be rewritten
as a function of the bulk stiffness in unloading-reloading, Ks and the bulk stiffness in
primary loading, Kc as:
1
Ks Kc
= Ks
Ks .
(7.24)
H=
Ks Kc
1
Kc
where due to the assumption of isotropic elasticity, the elastic bulk stiffness Ks relates to
Eurref as follows:
Eurref
Ks =
.
(7.25)
3(1 2)

80

7.2 The Hardening Soil model

q
a pp

en
str

h
gt

ite
cr

elastic region
c cotj

ria

pp

Figure 7.4: Evolution of the HS modelss cap and cone-type yield surfaces in p-q space.

The model parameter H can therefore be determined by the bulk stiffness ratio Ks /Kc .
As the physical meaning of the latter is more evident, it is often used to quantify H. An
illustration of yield surface evolution in p q space is given in Figure 7.4. Note that by
the above definition of d ps = ds , it is always assured that for zero volumetric strain,
d ps = 21 as required in Equation 7.13.
In order to formally complete the model description in the framework of infinitesimal
elastoplasticity, the elastic stiffness tensor Dijkl must be defined. The HS model assumes
isotropic elasticity inside the yield loci, so that Dijkl is assembled from the user defined
stiffness Eur and Poisons ratio, ur as follows:
Dijkl =

Eur
((1 2ur )ik jl + ur jk jl ).
(1 + ur )(1 2ur )

(7.26)

A summary of the HS models governing equations and parameters is presented in

Box 7.1 and Table 7.1 respectively. In Table 7.1, a differentiation is made between user
input and internal parameters because some model parameters cannot be quantified as
results of standard triaxial and oedometer tests directly.
Internal model parameters are the stiffness measures Eiref and H, and the cap-type
yield surfaces steepness . These internal parameters mainly relate to the user input
ref
ref
ref
is the tangent stiffness at 1 = pref
, and K0nc respectively, where Eoed
, Eoed
parameters E50
in K0 (oedometer) loading, and K0nc is the stress ratio of horizontal effective stress to
vertical effective stress in a normally consolidated state.
In double hardening situations, i.e. both yield loci are hardened simultaneously, analytical back-calculation of internal model parameters is impossible. Therefore, the internal parameters are solved for in an iterative scheme so that the HS model simulates
ref
ref
, and K0nc in oedometer element
in triaxial element tests and both, Eoed
the user input E50
tests, to within a tolerated error.

81

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

Box 7.1: Governing equations of the Hardening Soil model.
1. Elastic stiffness tensor:
Dijkl =

Eur
((1
(1+ur )(12ur )

2ur )ik jl + ur jk jl )

2. Yield functions:
(1 2 )
s
a
f12
= 2q

Ei qa (1 2 )
s
f13
=

fc

2(1 2 )
Eur
2(1 3 )
Eur

ps

2qa (1 3 )

Ei qa (1 3 )
q2
2
2
p pp where
2
1 + ( 1 1)2 1 3

ps and
with =

3sin
3+sin

3. Plastic potentials:
2)
s
g12
= (1

2
s
g13
=

gc

1 +2
sin m
2
(1 3 )
1 +3
2 sin m and
2
q2
2
p p2p where
2
1 + ( 1 1)2 1 3 with

3sin
3+sin

4. Mobilized dilatancy (modified Rowe):

sin m =
sin cs =

sin m sin cs
1sin m sin cs
sin sin
1sin sin

5. Hardening laws:
d ps = ds h ps
dpp

= dc hpp

Ks Kc
Ks Kc

0, where

g
g
g
with h ps = (


)=1
1
2
3
m
3 +c cot
p, where
with hpp = 2H pref +c cot

and Ks

ref
Eur
3(12)

6. Stress dependent stiffness:

cot m
Ei = Eiref ( pref3 +c
)
+c cot
cot m
Eur = Eurref ( pref3 +c
)
+c cot

7.3 The HS-Small model

The HS model is next combined with the Small-Strain Overlay model which was introduced in Chapter 6. Although probably the most important new feature of the resulting
HS-Small model is the Small-Strain Overlay model, two existing model mechanisms are
also modified. These are the HS models flow rule, and failure criterion (Figure 7.5).
While the incorporation of the Small-Strain Overlay model in the HS-Small model is
straight forward, the application of a new failure criterion is not. It requires the more
generalized model description developed at the end of this section.

82

7.3 The HS-Small model

s1

s1

q
s2

s2

s3

s3

Figure 7.5: Yield surfaces of the HS-Small model for cohesionless soil. Left: Cap- and
cone-type yield surfaces in principal stress space with the cone being in its
ultimate Matsuoka-Nakai failure criterion position. Right: p-q slice through
the yield surfaces.

7.3.1 Small-Strain formulation

The Small-Strain Overlay model primarily renders the HS models isotropic elastic stiffness dependant on strain history. Additionally, it attenuates plastic straining under initial loading when necessary, to preserve the materials small-strain stiffness. For its application within the HS-Small model, the Small-Strain Overlay models Poissons ratio is
set to = ur = const. so that the basic elastic relationships:
G =
K =

E
2(1+)
E
3(12)

(7.27)

hold. Two additional material parameters are needed to control the stress and strain history dependent stiffness of the HS-Small model. These are the initial shear modulus Gref
0
defined for the reference pressure pref and the shear strain 0.7 , at which the shear modulus has decayed to 70 percent of its initial value. The shear modulus G0 is calculated
from:

m
3 + c cot
ref
G0 = G0
,
(7.28)
pref + c cot
where m is the power law exponent that scales also the other stiffness parameters of the
HS-Small model (see Equation 7.19). The threshold shear strain 0.7 is taken independent
of mean stress.
A lower cut-off in the hyperbolic small-strain stiffness reduction curve is introduced at
the shear strain c where tangent stiffness is reduced to the unloading-reloading stiffness
Gur in larger strain cycles. The unloading-reloading shear modulus Gur relates to the HS
model parameter Eur as follows:
Gur =

Eur
.
2(1 + ur )

(7.29)

83

Tangent shear modulus

G [kN/m]

Secant shear modulus

G [kN/m]

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

g0.7
HS-Small
Hardin & Drnevich

Gur
gc

Shear strain amplitude gs [-]

Shear strain amplitude gs [-]

Figure 7.6: Cut-off in the tangent stiffness degradation curve as used in the HS-Small
model.

The shear strain c is then obtained from the modified Hardin-Drnevich relationship,
which for the tangent stiffness Gur yields:

2
ac
0.7
1

Gur = G0
= G0
(7.30)

2 +
c
1 a 0.7
0.7 + ac
c
0.7 1 a 0.7
and hence:
0.7
c =
a

!
G0
1 .
Gur

(7.31)

The resulting small-strain stiffness behavior of the HS-Small model is illustrated in Figure 7.6. Its isotropic elasticity tensor is calculated as:
Dijkl

2G
=
((1 2ur )ik jl + ur jk jl ) with G = G0
1 2ur

0.7
0.7 + aHist

2
.

(7.32)

The stiffness tensor given by Equation 7.32 is not a truly elastic, but a paraelastic one
(Hueckel & Nova [65]). Nevertheless, the terminology elastic stiffness is maintained for
the use of the above defined stiffness in the HS-Small model as a reminder that it applies
to all stress states, within as well as on the yield surface, similar to the elastic stiffness
known from traditional elastoplastic models.
By default, the HS model is hardened under initial loading. Without altering its hardening laws, plastic straining will reduce the HS-Small models small-strain stiffness. In
order to preserve the stiffness defined by the material parameters G0 and 0.7 , the HSSmall hardening laws are rewritten as:
d ps = ds hi h ps
dpp = dc hi hpp

84

with h ps
with hpp

as defined in Equation 7.22

as defined in Equation 7.22

(7.33)

7.3 The HS-Small model

where hi approximates plastic hardening under initial small-strain loading. In reasonably well agreement with initial modulus reduction curves, hi is defined as:
1+ EEur

hi := Gm

(7.34)

where Gm is the stiffness multiplier defined in Section 6.1.3. Note, that in combination
with the HS model, the threshold shear strain 0.7 in the Small-Strain Overlay model
is always doubled for reloading. If a primary loading condition exists, this will be accounted for by plastic hardening of the yield locus. For further details on the distinction
between initial and reloading, it is referred to Section 6.1.3.

7.3.2 Mobilized dilatancy

A main drawback of Rowes [153] approach to model mobilized dilatancy angles, is the
highly contractive behavior at low mobilized friction angles. In the original HS model
the mobilized dilatancy angle, sin m is therefore set to be greater or equal to zero overriding Rowes original equation (7.17):
sin m =

sin m sin cs
0.
1 sin m sin cs

(7.35)

Unfortunately, the dilatancy cut off introduced in Equation 7.17 typically yields too little
plastic volumetric contraction. Recently, several formulations have been proposed to
resolve this problem.
Wehnert [191] proposed a non-zero lower cut-off value, similar to Equation 7.35, for
the allowable contraction in loose sands:
sin m =

sin m sin cs
0.07 .
1 sin m sin cs

(7.36)

Even though the method by Wehnert is efficient, the lower cut-off value effectively represents an additional material parameter. Soreide [169] proposed to scale Rowes equation
as a function of mobilized friction:

sin m sin cs
sin m
sin m =
.
(7.37)
1 sin m sin cs
sin
From a quantitative point of view, the methods by Wehnert, and Soreide seem to be very
much alike as shown in Figure 7.7 and again in the model validation (Chapter 8). Soreide
additionally claims numerical benefits from his formulation: The likelihood of stress
return towards the apex is less when using big step sizes in the return mapping (see
Section 7.4 for further details on return mapping). However, having well defined initial
stresses and a suitable return mapping scheme this advantage should not be overrated.
In contrast to the Rowe trouble shooting strategies by Wehnert and Soreide, the modified flow rule of the HS-Small model makes use of critical state soil mechanics. Li
& Dafalias [103] proposed to treat dilatancy as a state-dependent quantity within the

85

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

framework of critical state soil mechanics. They defined dilatancy D, the ratio of plastic
volumetric strain increment to plastic deviatoric strain increment in triaxial space, as:
D=

d0
(M em ),
M

(7.38)

where d0 , and m are two material parameters, M is the critical stress ratio, = pq is the
actual stress ratio, and is the state parameter defined by Been & Jefferies [13]. The state
parameter is a scalar quantity that measures the difference between the actual and
the critical void ratio at the same mean stress p. Equation 7.38 reduces to the Cam-Clay
dilatancy (Roscoe & Schofield [151]) by specifying m = 0, and d0 = M .
Using the original formulation by Li & Dafalias in the HS-Small model would require
four additional material input parameters. These are the parameters d0 , and m introduced in Equation 7.38, and two parameters which define the critical state line in e ln p
space. The critical state line can for example, be defined by an initial void ratio e0 and
its slope . Unfortunately, the parameters d0 , and m can be obtained in numerical fitting
procedures only. Altogether, the input to the original approach by Li & Dafalias seems
too demanding for a model that is meant to be used in practical engineering applications. Therefore, the formulation by Li & Dafalias is simplified by a) calculating the state
parameter, based on constant void ratio, and b) providing default parameters for m,
and d0 . With these simplifications, Equation 7.38 can be rewritten as:
1

where

tan m = (M e 15 ln(pcs /p) )/10,

(7.39)

pcs
sin cs (1 sin m )
=
p
M sin m (1 sin cs )

(7.40)

1
, and d0 = M
. These default parameters are in good agreement to the
and m = 15
10
ones presented in Li & Dafalias [103]. As the default parameters cannot be guaranteed to
work in dilatant conditions, they are used for contraction only. Whenever dilatant material behavior is detected, the HS-Small model falls back to the original Rowe equation.
Figure 7.7 compares the mobilized dilatancy angles of all flow rules discussed above.
The comparison of mobilized dilatancy angles is supplemented by test data in Chapter 8.
However, it can already be noted that not one of the discussed approaches is considered
fully satisfactory. In future work, the HS-Small model could perhaps fully adopt the
approach by Li & Dafalias, therefore extending its use to dilatant behavior in a void
ratio dependent formulation (see Chapter 10).

7.3.3 Yield surface and plastic potential - HS-Small(MN) only

The Mohr-Coulomb failure criterion [123] for soils, which is implemented in the original
HS model, is one of the earliest and most trusted failure criteria. It is experimentally
verified in triaxial compression and extension and is of striking simplicity. However,
the Mohr-Coulomb criterion is very conservative for intermediate principal stress states

86

7.3 The HS-Small model

j = 35, y = 10
10

ym []

ym []

j = 35, y = 0
10

-10

-20

Rowe

-10

-20

Rowe

Wehnert
Soreide

-30

Wehnert
Soreide

-30

HS-Small
0

10

20

jm []

30

HS-Small
0

10

20

30

jm []

Figure 7.7: Mobilized dilatancy angle versus mobilized friction angle. Results from the
original Rowe equation, its modification by Wehnert, and Soreide, and the
modified Li & Dafalias formulation.

between triaxial compression and extension. Figure 7.8, for example, shows experimental results for reconstituted Osaka alluvial clay. Similar findings for sand and sandstone
have been for example reported by Lade [97], Ramamurthy & Rawat [144], Sutherland
& Mesdary [176], and Yamada & Ishihara [196].
Matsuoka & Nakai [113], [114] (MN) proposed a failure criterion that is in better agreement with the experimental surface shown in Figure 7.8. They propose the concept of
a Spatial Mobilized Plane (SMP), which defines the plane of maximum spatial, averaged
particle mobilization in principal stress space. The SMP is geometrically constructed
by deriving the mobilized (Mohr-Coulomb) friction angles for each principal stress pair
separately (Figure 7.9, left) and sketching the respective mobilized planes in principal
stress space (Figure 7.9, right). Matsuoka & Nakai derive their failure criterion by limiting the averaged ratio of spatial normal stress to averaged spatial shear stress on this
plane. Their failure stress ratio can be expressed as a simple function of the first, second,
and third stress invariant, I1 , I2 , andI3 as shown in Equation 7.41. With the SMP concept, the Matsuoka-Nakai criterion automatically retains the well established material
strength of the Mohr-Coulomb criterion in triaxial compression and extension.
The likewise well-known failure criterion by Lade & Duncan [99] (Lade) appears compared to the Mohr-Coulomb criterion and the Matsuoka-Nakai criterion rather optimistic in plane strain conditions and triaxial extension. Therefore, it has not (unlike
the Matsuoka-Nakai criterion), been considered as an option for the HS-Small model,
though it is mentioned here for the sake of completeness. It should also be mentioned
that using bifurcation analysis, progressive failures would most likely correct for the

87

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

s1

Experimental surface
Mohr-Coulomb surface
(j = 33.7, c = 2 kPa)

s3

s2

Legend:

sc = 100 kPa
sc = 200 kPa
sc = 300 kPa

Figure 7.8: Test results by Shibata & Karube [163] on reconstituted Osaka alluvial clay.

sI
s1
SMP
Mobilized
Planes

fmob23
fmob13

45+

fmob12
s3

s2

s1

45+

fmob12
2

s3
45+

s2

fmob13
2
sIII

fmob23
2

sII

Figure 7.9: The SMP concept of Matsuoka & Nakai. Left: Three mobilized planes where
the maximum shear stress to normal stress ratio is reached for the respective
principal stresses. Right: SMP in principal stress space.

88

7.3 The HS-Small model

Lade criterions overly optimistic, ultimate material strength estimate.
Both failure criteria, Matsuoka-Nakai, and Lade, are functions of the first, second, and
third stress invariants, I1 , I2 , andI3 respectively:
fLade =
fMN
where

I13
c1
I3
I1 I2
c2
I3

=0

with

c1 =

=0

with

c2 =

(3+sin )3
(1sin )(1+sin )2
9sin2
1+sin2

I1 = ii
I2 = 12 (ij ij ii jj )
I3 = 61 (ii jj kk + 2ij jk ki 3ij ji kk ).

(7.41)

(7.42)

In principal stress, the stress invariants simplify to:

I1 = 1 + 2 + 3
I2 = 1 2 2 3 3 1
I3 = 1 2 3 .

(7.43)

The constants c1 and c2 in Equation 7.41 are defined so that both failure criteria are
identical to the Mohr-Coulomb criterion in triaxial compression.
The HS-Small model optionally uses a Matsuoka-Nakai yield surface and criterion instead of the original HS models Mohr-Coulomb criterion. In the HS-Small(MN) model
the non-associated Mohr-Coulomb potential is substituted by a Drucker-Prager [35] potential. The cap-type yield function remains unchanged, only its potential is adjusted to
the new deviatoric flow of the cone-type yield surface.
The true triaxial experiments by Yamada & Ishihara [196] shown in Figure 7.10 not
only support the failure hypothesis by Matsuoka & Nakai but at the same time proof
that the Drucker-Prager potentials radial deviatoric flow direction is a reasonable assumption for low mobilized friction angles. For higher mobilized friction angles, flow
directions increasingly tend towards an associated deviatoric flow, the solid curve in
Figure 7.10 represent the Matsuoka-Nakai failure criterion ( = 30 ). However, in the
example presented in Figure 7.10, the error in assuming radial flow for high mobilized
friction angles is of the same magnitude or less than that made in assuming deviatoric
associated flow. A more detailed discussion of the test results by Yamada & Ishihara is

[127], a summary of true triaxial test results of several researchers

given in Mullersch
on
is given in Biarez & Hicher [16].
Replacing the Mohr-Coulomb yield surface and plastic potential by smooth surfaces
is also an advantage from a numerical point of view. Other than the Mohr-Coulomb
criterion, these have well defined gradients at all points except the apex. Therefore,
they allow for a more rigorous numerical formulation. The singular intersections of the
Mohr-Coulomb surface (Figure 7.3) on the other hand require a Koiter [91] algorithm
which will be explored in more detail in Section 7.4.4.
When using a Drucker-Prager plastic potential, from an numerical point of view it
is beneficial to formulate the yield criterion, Lode angle dependent. This will become
more obvious later in this Chapter when discussing the cap-type yield surface of the

89

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

s1
Q=15
Q=0

Direction of strain
increment

Q=-15
Failure stress
Q=-30
Q=-15
Q=0

Q=15

s3

s2
Q=15
Q=-30 Q=-15

Q=0

Figure 7.10: Experimentally derived shear strain increment vectors projected on the octahedral plane (after Nakai & Matsuoka [128]).

HS-Small(MN) model or its return mapping. As the Lode dependent formulation of the
original Matsuoka-Nakai yield criterion is bulky, several other simplified Lode dependent yield criterion were also proposed in the literature.
The Lode angle [106] in this thesis is defined as:
=

1
3

arcsin( 3

3J3
3/2 )
2J2

(7.44)

where J2 and J3 are the second and third deviatoric stress invariants. These are defined
for the deviatoric stress tensor, ij = ij 31 ii analog to Equation 7.42, substituting by
. In principal stress space, J2 and J3 can be expressed as:
J2 =
J3 =

1
((1 2 )2 + (2 3 )2 + (3 1 )2 )
6
1
(21 2 3 )(22 3 1 )(23
27

1 2 ).

(7.45)

The Lode angle in triaxial compression and extension is evaluated as = 6 and = 6

respectively. In the literature, Lode angles are sometimes defined in an arccos() function,
which implies a radial shift of 6 . All Lode dependent relationships presented in this
section can be easily adapted to the alternatively defined Lode angle as well.
The Lode dependent formulation of the Matsuoka-Nakai criterion is known as LMN

90

7.3 The HS-Small model

dependence [12]:
()

=
with
and

1
3
2 2 +1 cos
(

1
6

arccos(1 +

= 3 16
3sin
3+sin

arccos(1

27 2 (1)2
sin2 (3))
2( 2 +1)
2
(1)2
+ 27
sin2 (3))
2
2( +1)

for 0
for > 0

(7.46)

Unfortunately, the derivatives of the LMN dependence are bulky compared to those
of other Lode dependent strength criteria proposed in the literature. Here the criteria
by Argyris [6], Papadimitriou & Bouckovalas (PB) [133], and Wang [102] shall be mentioned:
2
(7.47)
(1 + ) (1 ) sin(3)
4
1
PB: () =
((1 + ) + (1 ) sin(3)) (7.48)
(1 + ) (1 ) sin(3) 2
p
(1 + 2 )2 + 4(1 2 ) sin(3) (1 + 2 )
Wang: () =
(7.49)
2(1 ) sin(3)

Argyris: () =

Results of all the above strength criteria are plotted in Figures 7.11, 7.12, and 7.13 for
triaxial compression friction angles of 20 , 30 , and 40 respectively. Obviously the
less complex formulations 7.47 to 7.49 tradeoff convexity at high friction angles for their
simplicity. The Matsuoka-Nakai (MN), and Lade yield surfaces stay convex even at high
friction angles and hence, fulfill Druckers convexity postulate [33].
Due to the Drucker-Prager plastic potential of the HS-Small(MN) model, the Lode
angle is constant during return mapping (see Section 7.4). This favors the use of the
LMN dependence in the HS-Small(MN) model. The less complex and very similar criterions by Wang or Papadimitriou & Bouckovalas would seem more appropriate in case

the condition
= 0, does not hold while higher order derivatives are needed in the
i
algorithmic setting.

7.3.4 Generalized model formulation - HS-Small(MN) only

The original formulation of the HS model is closely related to triaxial stress conditions.
In order to apply the Matsuoka-Nakai yield criterion, the basic yield function (7.12) is
reformulated in terms of mobilized friction m . The Mohr-Coulomb criterion in triaxial
conditions yields:
sin m =

1 3
1 + 3 + 2c cot

and
qf =

q = sin m (1 + 3 + 2c cot )

2 sin
(3 + c cot )
1 sin

(7.50)

(7.51)

91

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

1.2

30

s1

MN
Lade

0.8

Argyris

28

PB
0.4

Wang

j []

26
0

24
-0.4

s3

s2

22

-0.8

-1.2
-1.2

20
-0.8

-0.4

0.4

0.8

1.2

0.2

0.4

0.6

0.8

b = (s2-s3)/(s2-s1) [-]

deviatoric plane (p = 1)

Figure 7.11: Different failure criteria for a friction angle of = 20 in triaxial compression. Left: Deviatoric shape of the yield criteria. Right: Friction angle as a
function of the intermediate stress parameter b.

1.2

40

s1

MN
Lade

0.8

Argyris

38

PB
0.4

Wang

j []

36
0

34
-0.4

s3

s2
-0.8

-1.2
-1.2

32

30
-0.8

-0.4

0.4

deviatoric plane (p = 1)

0.8

1.2

0.2

0.4

0.6

0.8

b = (s2-s3)/(s2-s1) [-]

Figure 7.12: Different failure criteria for a friction angle of = 30 in triaxial compression. Left: Deviatoric shape of the yield criteria. Right: Friction angle as a
function of the intermediate stress parameter b.

92

7.3 The HS-Small model

1.2

50

s1

MN
Lade

0.8

Argyris

48

PB
0.4

Wang

j []

46
0

44
-0.4

s3

s2

42

-0.8

-1.2
-1.2

40
-0.8

-0.4

0.4

0.8

1.2

0.2

0.4

0.6

0.8

b = (s2-s3)/(s2-s1) [-]

deviatoric plane (p = 1)

Figure 7.13: Different failure criteria for a friction angle of = 40 in triaxial compression. Left: Deviatoric shape of the yield criteria. Right: Friction angle as a
function of the intermediate stress parameter b.

and thus:

q
= Rf
qa

1 sin
sin

sin m
1 sin m

With the objective shear strain measure:

r
1
s =
((1 2 )2 + (2 3 )2 + (3 1 )2 )
2

(7.52)

(7.53)

which reduces to s = 1 3 in triaxial conditions, one obtains s = 23 1 for zero volumetric strain. Adopting the original HS models yield function (Equation 7.12) to shear
strain s and considering Equation 7.52 results in:

1sin m
sin m
3 q
3 q

sps
(7.54)
fs =
1sin

1sin

m
2 Ei
2 Eur
R
sin m

sin

where sin m is the mobilized friction angle in triaxial compression. In mobilized friction,
the Matsuoka-Nakai yield criterion defined in Equation 7.41, can be written as:
sin2 m

9
1

I1 I2
I3
I1 I2
I3

(7.55)

This definition yields deviatoric isolines of mobilized friction that are similar to the shape
of the Matsuoka-Nakai yield criterion. Alternatively, Equation 7.55 can be expressed in

93

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

the LMN dependency = () as:
sin m =

3q
.
6(p + c cot ) + q

(7.56)

The plastic potential to the cone-type yield surface is defined as:

g s = (p + c cot )

6 sin m
3 sin m

(7.57)

Due to the cone-type yield surfaces new shape, the cap-type yield surface is also to
be reformulated. The specialstress measure defined in Equation 7.21 is now replaced
by the Roscoe invariant q = 3J2 . At the same time the Lode angle dependency of the
cone-type yield surface is translated to the cap by scaling its steepness:
q2
f =
p2 p2p .
2
()
c

(7.58)

The cap-type yield surface uses the non-associated plastic potential

gc =

q2
p2 p2p
2
()

(7.59)

where = (T rial ). The caps deviatoric plastic flow direction is thus consistent with
the cones radial Drucker-Prager potential.
Although the HS-Small model should respond slightly different under general (plastic) loading conditions, its triaxial behavior should be equal to the HS model. The HSSmall evolution laws are therefore adapted to the objective shear strain measure as follows:
q
g 2
g
g 2
g
g 2
1
(( g
) + (

) + (

) ) = 32
dsps = ds hi hsps with hsps =
2 1
2
2
3
3
1

m
cot
dpp = dc hi hpp
with hpp = 2H pref3 +c
p
+c cot
(7.60)
The evolution laws finally round up the formulation of the HS-Small model, which is
summarized for the HS-Small(MN) model in Box 7.2. A complete list of model parameters can be found in Table 7.1. The integration of the constitutive equations is discussed
in the next section.

7.4 Local integration of the constitutive equations

In the context of finite element analysis, integration of the constitutive equations (Section 7.1) is carried out within a finite set of integration points. The integration points
spatial coordinates are defined by the element meshs geometry and the element type.
In case of commonly used isoparametric elements, the integration points generally are
Gauss (quadrature) points. By balancing internal and external loads within the finite

94

7.4 Local integration of the constitutive equations

Box 7.2: Governing equations of the HS-Small(MN) model.
1. (Para-)Elastic stiffness tensor:
2G
Dijkl =
((1 2ur )ik jl + ur jk jl ) where
12ur

2
G
0.7
for s < c
0 0.7 +aHist
G
=
E

ur
for s c
2(1+ur )
2. Yield functions:
fs
= 32 Eqi

m
( 1sin
sin m )

(
)Rf ( 1sin
sin )
q2
2
2
= ()
2 p pp where

fc

sin2 m =

1sin m
sin m

3 q
2 Eur

sps

I1 I2
I3
I1 I2
1 I3

3. Plastic potentials:
6 sin m
g s = q (p + c cot ) 3sin
m
q2
()
2

gc =

p2 p2p where

= (T rial )

4. Mobilized
(dilatancy (modified Rowe):
sin m sin cs
for sin m sin cs 0
1sin m sin cs
sin m =
1
1
ln(p
/p)
cs
(M e 15
) for sin m sin cs < 0
10
sin cs =

sin sin
1sin sin

5. Hardening laws:
dsps = ds hsps
dpp
H

= dc hpp
=

Ks Kc
Ks Kc

and

pcs
p

sin cs (1sin m )
M sin m (1sin cs )

with hsps =
with hpp
and Ks

3
2

= 2H
=

3 +c cot
pref +c cot

ref
Eur

3(12)

6. Stress dependent stiffness:

cot m
Ei = Eiref ( pref3 +c
)
+c cot
cot m
Eur = Eurref ( pref3 +c
)
+c cot
3 +c cot m
G0 = Gref
0 ( pref +c cot )

7. Lode Matsuoka-Nakai
dependency:

3
1
() = 22 +1
with
cos
(
2 (1)2
1
arccos(1 + 27
sin2 (3))
for 0
6
2( 2 +1)
() =
2
2
(1)

16 arccos(1 + 27
sin2 (3)) for > 0
3
2( 2 +1)

3sin
3+sin

95

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

Table 7.1: Parameters for the HS and the HS-Small models

Parameter
I. User defind parameters
Triaxial secant stiffness
Oedometric tangent stiffness
Unloading/ reloading stiffness
Power of stress dependency
Cohesion (effective)
Friction angle (effective)
Dilatancy angle
Poissons ratio
Reference stress for stiffness
K0 -value (normal consolidation)
Failure ratio
Tensile strength
Small strain Stiffness
Threshold shear strain
II. Internal parameters
Initial secant stiffness
Cap parameter (steepness)
Cap parameter (stiffness ratio)
III. State parameters
Plastic deviatoric strain
Plastic shear strain
Pre-consolidation pressure
Deviatoric strain history

96

Symbol
ref
E50
ref
Eoed
Eurref
m
c

ur
pref
K0nc
Rf
Tension
E0ref
0.7

Eiref

Ks /Kc
ps
sps
pp
Hij

Unit
kN
m2
kN
m2
kN
m2

]
[ kN
m2

[ ]
[ ]
[ ]
kN
m2

[ ]
[ ]
kN

m2
kN
m2

[ ]
kN
m2

[ ]
[ ]
[ ]
[ ]
kN
m2

[ ]

HS

HS-Small

X
X
X
X
X
X
X
X
X
X
X
X
-

X
X
X
X
X
X
X
X
X
X
X
X
X
X

X
X
X

X
X
X

X
X
-

X
X
X

7.4 Local integration of the constitutive equations

element mesh, the global iterative procedure supplies strain increments for each integration point. Supplying constitutive stresses for the given strain increments is the task of
the constitutive model. The process of finding an admissible stress state and its corresponding internal variables for each integration point is called local integration.
In this section the local integration of the HS-Small model is discussed. The algorithm
chosen for this purpose is a member of the class of generalized midpoint algorithms that
were originally introduced by Simo & Hughes [166] as closest point projection algorithm.
The global integration is here not discussed in more detail since only standard finite
element procedures are deployed. These are documented in many books, for example
those by Hughes [66] or Belytschko [15].
The following abbreviations are used in the algorithmic formulations:

f
n+1
nij =

ij n+1

g
n+1
mij =

(7.61)
ij n+1

f
n+1

= q

n+1

where again the asterisk denotes n indices or vanishes for scalar functions.

7.4.1 Return mapping onto the yield surface

Local integration of the constitutive equations can be geometrically interpreted as the
projection of an elastic trial stress onto the updated yield surface. Numerical schemes
for stress return onto the yield surface are therefore known as return mapping algorithms. The elastic predictor or trial stress T rialij that functions as the mapping object is
initially obtained by freezing plastic flow. The flow rule then defines the mapping direction back onto the yield surface. The point in time when the flow rule is evaluated is set
by the time integration scheme. The class of generalized midpoint algorithms contains
all possible evaluation times. Using Vermeers [184] notation, the class of generalized
midpoint algorithms can be expressed as:

n+1
ij = Dijkl n+1kl n+1 pkl
(7.62)
n+1 p
n p
n+
kl = kl + mij
(7.63)
n+1
n
n+
q = q + h
(7.64)
Fn+1 = 0
(7.65)
where

mij = mij (1 ) nij + n+1ij , (1 ) nq + n+1q

n+
h = h (1 ) nij + n+1ij , (1 ) nq + n+1q

n+

and

(7.66)
(7.67)

and the superscripts n and n + 1 denote two subsequent steps in the integration procedure and 0 1, defines the time when the flow rule is evaluated. For = 0 the flow
rule is evaluated at the initial stress or, if this is elastic, at the intersection with the yield

97

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

s11

n+1

cross

n+a

s
n+1

n+1

n+1
n+a

n+a

sTrial
n
m

f
s33

s22

Figure 7.14: Deviatoric return mapping using the generalized midpoint algorithm with
different values. The chosen yield surface (Matsuoka-Nakai) and plastic
potential (Drucker-Prager) correspond to the HS-Small(MN) model.

surface. Hence, = 0 corresponds to an explicit Forward Euler approach. Similarly, the

generalized midpoint algorithm corresponds to an implicit Backward Euler scheme for
= 1 and to a Crank-Nicholson scheme for = 12 .
Figure 7.14 illustrates different possible results of applying these integration schemes
to the deviatoric return mapping in the HS-Small(MN) model. Obviously, the Forward
Euler approach could not converge for the rather large excursion outside the yield surface. The implicit and midpoint schemes would both converge in the example given, the
implicit one with less accuracy than the midpoint scheme (first order only). However,
the implicit scheme has several advantages over the midpoint scheme. There is no need
to calculate the intersection with the yield surface in the implicit scheme, its solution is
sought by evaluating the flow rule at the final stress. By implicitly assuming that such
a stress state exists, the Backward Euler algorithm is guaranteed to provide a solution.
Another important advantage of the implicit algorithm lies in the fact that it can be linearized in a closed form. From linearization, consistent elastoplastic tangent moduli can
be obtained that preserve quadratic convergence in the global equilibrium iterations, if
a Newton-Raphson scheme is used.
In the following, an implicit closest point projection algorithm for = 1 is introduced.
The consistent algorithmic tangent moduli for a single yield surface is discussed as well.
The algorithmic setting presented, is based on works of Jeremic & Sture [83]. The implications of multiple yield surfaces is discussed in Section 7.4.4.

7.4.2 The closest point projection algorithm

Let the load step n (n+1) be a plastic load step so that > 0. With the strain increment
n+1
d the elastic trial stress calculates for this load step as:
T rial

ij =n ij + Dijkl

98

n+1

dkl .

(7.68)

7.4 Local integration of the constitutive equations

The implicit form of the general midpoint algorithm (Equation 7.62) can now be rewritten as:
n+1
ij =T rial ij Dijkl n+1mkl
(7.69)
where n+1mkl is the derivative of the plastic potential for the implicitly assumed stress
at the end of the load step. A algorithmic solution for this equation can be found by
defining the tensor of residuals:

rij = ij T rialij Dijkl n+1mkl

(7.70)
and enforcing rij = 0 for the final stress ij =n+1 ij in the linearized form of Equation
7.70:

mkl
mkl
n+1
dmn + Dijkl
dq = 0.
(7.71)
rij + dij + Dijkl
mkl + Dijkl
mn n+1
q n+1
By algebraic manipulation the above equation can be solved to dij :

dmn = rij + dDijkl n+1kl ijmn with

1

mkl
ijmn =
im nj + Dijkl
and
mn n+1

mkl
n+1
kl =
mkl +
h
q n+1

(7.72)
(7.73)
(7.74)

and applied to the linearized yield function:

n+1

f +n+1 nmn dmn +n+1 dq = 0

(7.75)

which is derived by a first order Taylor expansion around the final stress state n+1f = 0.
Solving the above equation for d, the solution for the iterative update of the plastic
multiplier is readily found as:
n+1

d =

f rij ijmn n+1nmn

Dijkl kl ijmn n+1nmn n+1 h

(7.76)

where again the conventions defined in Equation 7.61 have been used.
In order to successfully apply the closest point projection as an iterative procedure, a
suitable starting point is necessary. A good guess for the starting point can typically be
found by using the normal to the elastic trial point [83]:
T rial

f +T rial nmn dmn +T rial dq = 0,

(7.77)

and assuming that any stress relaxation from the trial stress is due to plastic straining
and thus:

(7.78)
dT rialmn = Dmnpq d(0) T rialmpq .

99

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

Solving the above equations for d
d(0) =

T rial

T rialn

T rialm
nm Dnmpq
pq

T rial h

(7.79)

(0)

the starting point mn is calculated as:

(0)
mn
=T rial mn d(0) Dmnpq

T rial

mpq .

(7.80)

The algorithmic setting for the closest point projection algorithmic can now be summarized as follows (Box 7.3):

7.4.3 The consistent algorithmic tangent stiffness tensor

When the Newton-Raphson scheme is used for solving the global equilibrium equations,
only the consistent tangent stiffness can preserve its quadratic convergence. Using other
stiffness tensors, for example the traditional tangent stiffness obtained by a Forward
Euler method ( = 0), the quadratic convergence is lost.
The consistent elastoplastic algorithmic tangent stiffness was originally introduced by
Hughes & Taylor [67] for viscoplastic materials. Simo & Taylor [167] and Runesson &
Samuelsson [154] later adopted it for elastoplastic materials.
In any case the consistent elastoplastic stiffness is obtained by enforcing the consistency condition on the discrete algorithmic problem instead of the continuum problem,
which would result in the continuum elastoplastic tangent. Enforcing the consistency
condition at the final converged stress position n+1ij yields :

f
f
dn+1f =
dq
d
+

ij
q
ij n+1
(7.81)
n+1
= n+1nij dij +n+1 dq = 0.
The unknown stress increment dij is found by linearization or differentiation of the
implicit Backward Euler equation (7.69) to:

mkl
mkl
n+1
dij = Dijkl dkl dDijkl
mkl Dijkl
dmn Dijkl
dq (7.82)
mn n+1
q n+1
and isolation through simple algebraic manipulations:
dmn = ijmn Dijkl (dkl dkl )

(7.83)

where and are defined in Equation 7.73 and 7.74 respectively. Finally substituting
7.83 into 7.81 gives an expression for the consistent elastoplastic tangent:
ep
dmn
dpq = Dpqmn
ep
Dpqmn
= (rspq Drsmn )

100

with
(rspq Drskl )kl n+1nij (rsij Drsmn )
n+1n (
n+1 h
ot
rsot Drspq )pq +

(7.84)
(7.85)

7.4 Local integration of the constitutive equations

Box 7.3: Closest point projection algorithm.

1. Initalize: k = 0; n+1q (0) =n q; d(0) = 0.
2. Calculate trial stress and yield function
T rial

ij =n ij + Dijkl

T rial (0)

n+1

dkl
(0)
f (T rialij ,T rial q (0) )

3. IF n+1f (0) 0 THEN EXIT.

4. Calculate starting point and update state variables
T rialf
T rialm T rial h
D
nm nmpq
pq

T rial
n+1 (0)
mn d(0) Dmnpq T rialmpq
mn =
n+1 (0)
q =n q + d(0) h

d(0) =

T rialn

5. DO WHILE n+1f (k) T OL

a) Calculate yield function n+1f (k)
n+1 (k)

(k)

= f (n+1ij ,n+1 q (k) )

(k)

(k)

b) Calculate derivatives n+1nkl , n+1mkl ,

(k)

mkl
mn

n+1

, and

(k)

mkl
q

n+1

(k+1)

c) Update plastic multiplier d

(k)
n+1f r
n+1n
mn
ij ijmn
(k+1)
(k)
d
= d + Dijkl kl ijmn n+1nmn n+1 h
d) Update stress and state variables
(k)
n+1 (k+1)
mn =T rial mn d(k+1) Dmnpq n+1mpq
n+1 (k+1)
q
=n q + d(k+1) h

e) SET k = k + 1
END DO

101

Chapter 7 HS-Small, a small-strain extension of the Hardening Soil model

corner region
n+1 s

n ij

n+1

msij

sTrial
ij

n+1 c n+1

n ij = mcij

n+1

sij
s

c cotj

sij
c

f =g
elastic region pp

Figure 7.15: Corner problem at the intersection of cone- and cap-type yield surface.

When using an implicit return mapping scheme as the one presented in the previous
section, it is only a small effort to calculate the consistent elastoplastic tangent. However, the main drawback associated with the implicit closest point projection algorithm
and the calculation of the consistent tangent is their need for the plastic potentials second order gradients. This may prove exceedingly laborious for more complex plastic
potentials, but is straightforward in the HS-Small model.

7.4.4 Corner and apex problems

Corner problems arise at non-smooth intersections of yield surfaces, if not a single surface criterion is used (e.g. Ehlers [38]). By substituting the Mohr-Coulomb yield surface
of the original HS model, by the smooth Matsuoka-Nakai criterion, some singularities
are eliminated in the HS-Small(MN) model. Yet, the intersection of cone and cap-type
yield surface still causes a corner problem.
Koiter [91], additively decomposes plastic strain rates in such a corner problem as
follows:
dpij = dcone cone mij (ij , q ) + dcap cap mij (ij , q ).
(7.86)
The two vector cone-cap return problem is illustrated in Figure 7.15. Return mapping
in corner regions is slightly more tricky than regular return mapping to one yield surface. Yet, the concepts explained above remain the same. Appendix B gives the updated
algorithm for two surface closest point projection.
Another difficulty in corner problems is the estimate of the corner gray region. If the
Lode angle is frozen during return mapping (Drucker-Prager potential), and there is
no hardening (cone in final position), a two-dimensional analysis can be conducted as
shown in Figure 7.15. Otherwise, a strategy that determines to which surface the trial
stress is to be returned, is required. The strategy proposed by Bonnier [18] shown in
Appendix C is used in the HS-Small implementation.

102

7.4 Local integration of the constitutive equations

q
corner region
f

msij

n+1
n+1

mtij

apex gray
region

sTrial
ij

f =g
n

sij

n+1

sij

apex c cotj st

elastic region pp

Figure 7.16: Apex gray region and tension cut-off illustrated in a section with 2 = 0.

If not smoothly curved, the apex is another singularity in the yield surface. The apex
gray region is defined by the gradient to the cone-type potential surface as shown in Figure 7.16. If the apex corresponds to an admissible tensile stress, the trial stress is returned
to the apex. If the apex point violates the user defined maximum allowable tensile stress,
a two vector return mapping scheme to the respective tension cut-off planes is evoked.
The HS-Small model uses a tension cut-off criterion based on minimum principal stress,
which implies three (fixed) orthogonal tension cut-off planes in principal stress space:
fit = Tension i ,

(7.87)

where Tension is the user defined maximum allowable tensile stress.

103

Chapter 8
Validation and verification of the HS-Small
model
The HS-Small model is validated by comparing numerically simulated experimental
data to real (measured) data. Model validation is divided into two main parts: Element
tests and boundary value problems.
The element tests section strictly speaking also contains partly code verification. Iteratively derived stiffness decay curves are checked against the closed form model by
Hardin & Drnevich in the version of Santos & Correia. This is necessary due to the hardening plasticitys impact on total elastoplastic stiffness. For clarification of the differences
implied by the terms validation and verification see also Figure 1.2.
All element tests are calculated on the Gau-Point level. All boundary value problems
are analyzed using the finite element code P LAXIS V 8.2 [21]. As the HS-Small model was
implemented in the calculation kernel of this code, standard interface definitions and
calculation settings could be used. The help and cooperation of Plaxis B.V, especially the
help of Dr. Paul Bonnier, during model implementation is greatly acknowledged here.
All calculations are performed with triangular elements (6 nodes, 3 integration points)
and default settings in the iteration controls of the finite element code P LAXIS.

8.1 Element tests

The element test section as the first part of the verification and validation process of the
HS-Small model makes use of triaxial, biaxial and oedometer test data readily available
in the literature. Materials considered include dense sand, medium dense to loose sand,
and reconstituted Kaolin clay.
Most test data on sand presented in this section are taken from the database compiled
by Desrues et al. [31]. This database contains many well documented experiments on
Hostun-RF-Sand (Hostun sand), an industrial sand quarried near the village of Hostun,

Drome,
Rhone-Alpes,
France. Hostun sand results from the specification of its sieving
process (Flavigny et al. [41]) uniformly graded. Its coefficient of uniformity is U =
D60
= 1.70, where its medium grain size is D50 = 0.35mm. The minimum and maximum
D10
void ratio are experimentally derived as emin = 0.63 and emax = 1.00. The experimental
database distinguishes between dense and loose Hostun sand, based on its initial void
ratio e. For initial void ratios in the range of e = 0.63...0.68 the sand is classified as
dense, whereas in the range of e = 0.85...0.92, it is classified as loose. The corresponding

105

Chapter 8 Validation and verification of the HS-Small model

emax e
density index, ID = emax
calculates to ID = 0.87...1.00 and ID = 0.22...0.40 for the
emin
dense and loose sand respectively. According to DIN EN ISO 14688-2:2004 [43] the sand
should therefore rather be classified very dense and loose to medium dense respectively.
However, for the sake of consistency with the database used, its classification as dense
and loose is adopted here, too. Beyond the data available in the database by Desrues,
test data from the book by Biarez & Hicher [16] is used. In excess of conventional lab
data, Biarez & Hicher provide also some small-strain stiffness data in their book.
The Kaolin clay test presented in this section is also taken from the data collection by
Biarez & Hicher [16]. Its plasticity index is specified as, PI = 30, its liquid limit is wl = 70.
Only normally consolidated samples are considered. Unfortunately the test data provided for drained and undrained tests are obtained at very different confining pressures
(and hence void ratios). Drained test data are available at confining pressures up to 800
kPa whereas those for undrained tests are available only for confining pressures above
2000 kPa. Here the HS as well as the HS-Small model, both have the shortcoming that
the materials void ratio is not considered in the model formulation.
Two separate material data sets have been defined for Hostun sand. In Table D.1,
Sand(D) and Sand(L) denote dense Hostun sand and loose Hostun sand respectively.
The clay data in Table D.1 describes normally consolidated Kaolin clay at low confining
pressures. The undrained Kaolin tests given in Biarez & Hicher [16] are not considered in
the following due to their high confining pressures. If not indicated otherwise, the material parameter sets given in Table D.1 are used exclusively in all element tests presented
in this section.
Oedometer, or K0 compression test data for loose and dense Hostun sand are shown
in Figure 8.1. The HS as well as the HS-Small model can simulate these tests reasonably
well. As expected, the HS-Small model gives a somewhat stiffer response than the original HS model in both tests. The difference is emphasized more for the dense sample due
to its higher threshold shear strain 0.7 . Since both models, HS and HS-Small, cannot predict the accumulation of plastic strain within closed cycles correctly, both models give
overall, less deformation compared to the experiment. This effect is more pronounced
for the loose sample, which accumulates more plastic strain in closed cycles.
Tracing of the unloading-reloading cycles is nevertheless improved by the HS-Small
model, which gives a small amount of hysteresis as also seen in the experiment. From the
unloading-reloading loops it can be concluded that the power law exponent m = 0.55
chosen for dense Hostun sand is rather high. However, the chosen exponent is also
attributed to the triaxial test data shown next.
Drained triaxial test data is shown in Figure 8.2, 8.3, and 8.4 for dense Hostun sand,
loose Hostun sand, and normally consolidated Kaolin clay respectively. For the sand
tests little experimental small-strain data are available. For the Kaolin clay there is no
pq test data available at these confining pressures where small-strain data are available.
From the right hand side of Figures 8.2, 8.3, and 8.4 it can be concluded that the incrementally (HS-Small) and analytically derived (Santos & Correia) stiffness decay curves
match up pretty well. The need for the modification of the HS hardening laws is also
illustrated in Figures 8.2, 8.3, and 8.4: A pure combination of the HS model with the
Small-Strain Overlay model without modification of the HS hardening laws (HS + Small-

106

8.1 Element tests

s1[kN/m2]
400
300

s1[kN/m2]

Dense Hostun Sand

Loose Hostun Sand

200

HS-Small
HS (original)

150

Experiments
200

100

100

50

0
0.000

0.004

0.008

0.012

e1[-]

0
0.000

0.005

0.010

0.015

0.020

e1[-]

Figure 8.1: Oedometer test on dense (left) and loose (right) Hostun sand.

Strain Overlay) yields too low small-strain stiffness. Here, plasticity is over-dominating
the initial elastoplastic stiffness.
It can be noted that there is a small numerical discrepancy between the material input
parameter 0.7 and the computed shear strain (HS-Small), at which small-strain stiffness
is decreased to 70% of its initial value. For practical applications this discrepancy can be
easily neglected since it is probably well below the experimental precision and repeatability. Alternatively, the internal model parameter hi (see Section 7.3.1), could be fitted in
an iterative procedure by including it with the internal parameter conversion discussed
in Chapter 7. In either case, a perfect match between the HS-Small decay curves and the
analytically derived curves is obtained for zero plastic straining. Then, the elastoplastic
model degenerates to the Small-Strain Overlay model as presented in Chapter 6.
Altogether, the simulated triaxial tests match the test data presented, reasonably well.
It turns out that the modified flow rule has a vanishing effect on the volumetric behavior
in drained tests. This will be proven to be different for undrained tests later in this
section.
The implementation of the Small-Strain Overlay model into the HS-Small model does
not overly affect the overall stress-strain curve. There is no over-pronounced shift or
bump in the stress-strain curve. Combining the Small-Strain Overlay model with a bubble model, or any other model that has an initial elastic range would probably not give
such smooth results.
In oedometer and triaxial tests the intermediate stress parameter b always equals 0.
Both failure criteria used in the HS-Small model, Mohr-Coulomb (MC) and MatsuokaNakai (MN), are coincident in this case. For this reason, there has so far been no distinction made towards the HS-Small models failure criteria in the validation process.
In biaxial tests (plane strain), however, the non-zero intermediate stress parameter increases the material strength estimate of the Matsuoka-Nakai criterion compared to that
of the Mohr-Coulomb criterion.
Figures 8.5, and 8.6 show biaxial test results for dense and loose Hostun sand respectively. As expected, the HS-Small model with the Matsuoka-Nakai yield criterion

107

Chapter 8 Validation and verification of the HS-Small model

s3 = 100 kPa CD

s1/s3

GSecant [kN/m2]

eVol[-]
-0.20

-0.16

80000

-0.12
2

-0.08

40000

-0.04
0
0.00

0.00
0.02

0.04

0.06

0.08

s3 = 300 kPa CD

s1/s3

e1[-]

0.0001

0.01

e1-e3[-]

0.01

e1-e3[-]

0.01

e1-e3[-]

GSecant [kN/m2]

eVol[-]
-0.20

0.001

160000

-0.16
120000
-0.12
80000

-0.08
-0.04

0
0.00

40000

0.00
0.02

0.04

0.06

0.08

s3 = 600 kPa CD

s1/s3

e1[-]

0.0001

0.001

GSecant [kN/m2]

eVol[-]
-0.20

-0.16

200000

-0.12
2

-0.08

100000

-0.04
0
0.00

0.00
0.02

0.04

0.06

0.08

e1[-]

0.0001

Experiment

HS (original)

Santos & Correia

HS-Small

0.001

HS + Small-Strain Overlay

Figure 8.2: Drained triaxial tests on dense Hostun sand at confining pressures of 100,
300, and 600 kPa. Left: Stress-strain data. Right: Shear modulus reduction.

108

8.1 Element tests

s3 = 100 kPa CD

s1/s3

GSecant [kN/m2]

eVol[-]
-0.20
60000

-0.16
-0.12

40000

-0.08
20000
-0.04

0
0.00

0.00
0.04

0.08

s3 = 300 kPa CD

s1/s3

e1[-]

0.0001

0.001

0.01

e1-e3[-]

0.01

e1-e3[-]

0.01

e1-e3[-]

GSecant [kN/m2]

eVol[-]
-0.20
120000

-0.16
-0.12

80000

-0.08
40000
-0.04

0.00

0.00

0.04

0.08

s3 = 600 kPa CD

s1/s3

0
0.0001

e1[-]

0.001

GSecant [kN/m2]

eVol[-]
-0.20

-0.16

200000

-0.12
2

-0.08

100000

-0.04
0
0.00

0.00
0.04

0.08

0
0.0001

e1[-]

Experiment

HS (original)

Santos & Correia

HS-Small

0.001

HS + Small-Strain Overlay

Figure 8.3: Drained triaxial tests on loose Hostun sand at confining pressures of 100, 300,
and 600 kPa. Left: Stress-strain data. Right: Shear modulus reduction.

109

Chapter 8 Validation and verification of the HS-Small model

s3 = 100 kPa CD

s1/s3

GSecant [kN/m2]

eVol[-]
-0.20
30000

-0.16
-0.12

20000

-0.08
10000
-0.04

0
0.00

0.00
0.05

0.10

0.15

0.20

s3 = 300 kPa CD

s1/s3

e1[-]

0.0001

0.001

0.01

e1-e3[-]

0.01

e1-e3[-]

0.01

e1-e3[-]

GSecant [kN/m2]

eVol[-]
-0.20
60000

-0.16
-0.12

40000

-0.08
20000
-0.04

0
0.00

0.00
0.05

0.10

0.15

0.20

s3 = 600 kPa CD

s1/s3

e1[-]

0.0001

0.001

GSecant [kN/m2]

eVol[-]
-0.20
120000

-0.16
-0.12

80000

-0.08
40000
-0.04

0
0.00

0.00
0.05

0.10

0.15

0.20

e1[-]

0.0001

Experiment

HS (original)

Santos & Correia

HS-Small

0.001

HS + Small-Strain Overlay

Figure 8.4: Drained triaxial tests on reconstituted Kaolinite clay at confining pressures of
100, 300, and 600 kPa. Left: Stress-strain data. Right: Shear modulus reduction.

110

8.1 Element tests

s3 = 100 kPa CD

s1/s3

eVol[-]

-0.20

s3 = 200 kPa CD

s1/s3

eVol[-]

-0.20

-0.16
4

-0.16
4

-0.12
-0.08

-0.12
-0.08

-0.04
0
0.00

0.00
0.02

e1[-]

0.04

s3 = 400 kPa CD

s1/s3

-0.04
0
0.00

eVol[-]

-0.20

0.00
0.02

0.04

0.06

e1[-]

s3 = 800 kPa CD

s1/s3

eVol[-]

-0.20

-0.16
4

-0.16
4

-0.12
-0.08

-0.12
-0.08

-0.04
0
0.00

0.00
0.02

0.04

0.06

Experiment

0.08

e1[-]

-0.04
0
0.00

HS-Small(MN)

0.04

0.08

0.00
0.12 e1[-]

HS-Small(MC)

Figure 8.5: Drained biaxial tests on dense Hostun sand at confining pressures of 100, 200,
400, and 800 kPa.

111

Chapter 8 Validation and verification of the HS-Small model

s3 = 100 kPa CD

s1/s3

eVol[-]
-0.20

s3 = 200 kPa CD

s1/s3

eVol[-]
-0.20

-0.16
3

-0.16
3

-0.12
2

-0.12
2

-0.08
1
0
0.00

0.02

0.04

0.06

s3 = 400 kPa CD

s1/s3

0.08

-0.04

0.00

0
0.00

e1[-]

eVol[-]
-0.20

-0.08
-0.04
0.00
0.04

0.08

e1[-]

s3 = 800 kPa CD

s1/s3

eVol[-]
-0.20

-0.16
3

-0.16
3

-0.12
2

-0.12
2

-0.08

-0.08

-0.04

-0.04

0.00

0.00

0.00

0.04

0.08

Experiment

0.12 e1[-]

0.00

HS-Small(MN)

0.04

0.08

0.12 e1[-]

HS-Small(MC)

Figure 8.6: Drained biaxial tests on loose Hostun sand at confining pressures of 100, 200,
400, and 800 kPa.

112

8.1 Element tests

Esec [kN/m2]

q [kN/m2]
100

4E+5
50

0
2E+5
-50
Calculated
Rivera/Bard (initial)

-100

Rivera/Bard (reloading)

0
-5
10

-4

10

-3

10

e1[-]

-4

-5x10

-4

-3x10

0x10

-4

3x10

-4

5x10

e1[-]

Figure 8.7: Hysteresis loop in a drained triaxial test on dense Hostun sand.

(HS-Small(MN)) always predicts higher peak strengths than the model with the MohrCoulomb yield criterion (HS-Small(MC)). The test data fully validates the MatsuokaNakai approach for dense sand whereas for loose sand the strength is slightly overpredicted. Here it seems that bifurcations set in even before the limit state according to
the Matsuoka-Nakai criterion is reached.
The element tests presented in Figures 8.7, 8.8, and 8.9 recapitulate some of the tests
already presented and introduced in Chapter 6. Now, the HS-Small and the original HS
model are evaluated in these tests instead of the pure Small-Strain Overlay model.
The initial stiffness in Figure 8.7 is slightly over-predicted since the standard material
set for dense Hostun sand given in Table D.1 is used. The calculated hysteresis loop is not
closed anymore because of too much unloading plasticity. Here, as for other cyclic phenomena as well, kinematic hardening formulations should be used rather than purely
isotropic hardening models. Nevertheless, the hysteresis loop is simulated not too badly
by the (isotropic hardening) HS-Small model.
Figures 8.8, and 8.9 show numerical simulations of the tests by Richardson [148] on
reconstituted London clay. The London clay could be reasonably modeled with the data
set defined for Kaolin clay by slightly reducing the initial stiffness to Gref
0 = 26 MPa and
ref
ref
ref
= 7 MPa).
increasing the unloading-reloading stiffness to Eur = 21 MPa (E50 = Eoed
Again, there is no difference between the Matsuoka-Nakai and the Mohr-Coulomb flavor of the HS-Small model as b = 0. The improvement achieved by using the HS-Small
model is obvious. However, a pronounced distinction between stress paths AOX and
COX is missing. The HS-Small model tracks the difference in these stress paths by means
of different overconsolidation pressures pp only. It does not track the second main difference between these stress paths, which is the materials void ratio at the start of devi-

113

Chapter 8 Validation and verification of the HS-Small model

GTangent [MPa]
40

q [kPa]
DOX
AOX

150

100

COX

30

BOX

50

Calculated (HS)
20

0
0

100

O
200

A
300

400

p' [kPa]

-50
10
B

-100

0
10-4

10-3

10-2

e1[-]

-150

Figure 8.8: HS model predictions for Richardsons [148] tests.

GTangent [MPa]
40

q [kPa]
DOX

150

100

AOX
COX

30

BOX

50

Calculated
(HS-Small)

20

0
0

100

O
200

A
300

400

-50
10
-100

0
10-4

10-3

10-2

e1[-]

-150

Figure 8.9: HS-Small model predictions for Richardsons [148] tests.

114

p' [kPa]

8.2 Boundary value problems

atoric loading. The effect of pp in p0 = constant loading is negligible.
After having validated the small-strain formulation and the yield criteria of the HSSmall model it remains to evaluate its modified flow rule. As mentioned before, the
effects of the modified flow rule in drained conditions are negligible. However, they
are not negligible in undrained conditions. Undrained triaxial tests on dense and loose
Hostun sand are shown in Figures 8.10, and 8.11 respectively. These Figures include
results from all dilatancy modifications discussed in Section 7.3.2.
The undrained test on dense Hostun sand (Figure 8.10) clearly points out the difficulties in using Rowes formulation for dilatant materials in undrained conditions. Rowes
decoupling of dilatancy and void ratio may lead to notable overestimates in undrained
shear strength. Such calculations should thus include a dilatancy cut-off (as optionally
available in the HS-Small model), if not a more refined flow rule is available.
Figure 8.11 shows undrained triaxial tests on loose Hostun sand. Here the advantage of the modified flow rule is clearly visible. The overestimated undrained strength
from the original HS model is now reduced to the experimentally derived results. The
modified flow rule of the HS-Small model does not incorporate additional material parameters. It rather uses a default set of material parameters. Therefore, the results for
other materials may not be always in such close agreement to experimentally derived
strength values. Finally it should be noted that none of the flow rules considered reproduces the experimentally derived stress path particularly well. Reason for that is again
their void ratio independent formulation.
In summary, the validation of the HS-Small model in element tests gave mostly positive results. In comparison to the original HS model, only improvements were found.
The lack of void ratio dependency and kinematic hardening have been identified as the
main drawbacks of the HS-Small model as well as of the HS model.

8.2 Boundary value problems

In the second part of the model validation process, a set of five boundary value problems is examined. The set of boundary value problems includes two tunnels, two deep
excavations and one spread footing foundation. Though model validation, through the
comparison of simulated results with experimentally derived results is the main focus
of this section, more general issues will also be addressed. These include general finite
element issues (i.e. What section of the problem should be meshed?), as well as more
specific small-strain stiffness matters (i.e. What impact has small-strain stiffness on different problem classes?).
In the literature, tunnels are probably the most popular examples for validating smallstrain stiffness models. In appreciation of that fact, the first problems discussed in this
section are tunnelling problems. Secondly, deep excavations will be looked at before the
section is closed with a very basic spread footing foundation example.

115

Chapter 8 Validation and verification of the HS-Small model

q [kN/m2]

(s1-s3)/2 [kN/m2]

s3 = 200 kPa CU

2000

300

1500
200
1000
100

500
0
0.00

0.02

q [kN/m2]

0.04

0.06

e1[-]

200

300

400

400

1000

200

(s1+s3)/2 [kN/m2]

0
0.02

q [kN/m2]

0.04

0.06

500

(s1-s3)/2 [kN/m2]

s3 = 300 kPa CU

2000

0
0.00

(s1+s3)/2 [kN/m2]

e1[-]

400

600

800

(s1-s3)/2 [kN/m2]

s3 = 600 kPa CU

3000
800
2000
400
1000

0
0.00

(s1+s3)/2 [kN/m2]

0
0.02

0.04

Experiment
HS (original)

0.06

e1[-]

HS-Small

800

1200

1600

HS-Small with flow by Wehnert

HS-Small with flow by Soreide

Figure 8.10: Undrained triaxial tests on dense Hostun sand at confining pressures of 200,
300, and 600 kPa. Left: Stress-strain data. Right: s-t diagram.

116

8.2 Boundary value problems

q [kN/m2]

(s1-s3)/2 [kN/m2]

s3 = 200 kPa CU
160

200

120
80

100

40
0
0.00

0
0.02

0.04

0.06

e1[-]

100

200

(s1+s3)/2 [kN/m2]
q [kN/m2]

(s1-s3)/2 [kN/m2]

s3 = 300 kPa CU
160

300

120
200
80
100

40

0
0.00

0
0.02

0.04

0.06

e1[-]

100

200

300

(s1+s3)/2 [kN/m2]
q [kN/m2]

(s1-s3)/2 [kN/m2]

s3 = 600 kPa CU
400

600

300
400
200
200

0
0.00

100
0
0.02

0.04

0.06

e1[-]

200

400

600

(s1+s3)/2 [kN/m2]

Experiment
HS (original)

HS-Small

HS-Small with flow by Wehnert

HS-Small with flow by Soreide

Figure 8.11: Undrained triaxial tests on loose Hostun sand at confining pressures of 200,
300, and 600 kPa. Left: Stress-strain data. Right: s-t diagram.

117

Chapter 8 Validation and verification of the HS-Small model

Figure 8.12: Steinhaldenfeld NATM tunnel before concreting the inner liner.

8.2.1 Tunnels
Schmidt [160] and Peck [135] were the first to propose that green field settlement troughs
above tunnels can be approximated by a (Gaussian) distribution function of the form:
y2
s(y) = smax exp( 2 )
2i

(8.1)

where smax is the maximum vertical settlement above the tunnel axis, y is the horizontal
distance to the tunnel axis, and i is the horizontal distance of the settlement troughs
inflection point from the tunnel axis.
The empirical relationship expressed in Equation 8.1 holds very well, independent of
the applied tunnelling method. Numerical methods often fail in predicting the correct
shape of settlement troughs. Especially, if the applied soil model does not account for
small-strain stiffness, numerical methods commonly predict too flat and wide settlement
troughs. This can be observed for instance in the following two case studies, where the
HS-Small model yields significantly improved results compared to the HS model.
8.2.1.1 Steinhaldenfeld NATM tunnel
Figure 8.12 shows Steinhaldenfeld tunnel, a tunnel near Stuttgart that is constructed using the new Austrian tunnelling method (NATM). The soil layers together with the FE
mesh used in the plane strain analysis of Steinhaldenfeld tunnel are shown in Figure
8.13. The lower soil layers (Keuper marl and Limestone) are overconslidated. The overconsolidation ratio is assumed to be 2.0. The material behavior is drained in all the soil
layers. A complete list of the soil layers material parameters used in the calculation is
given in Table D.2.
The weak marl around the tunnel excavation was improved by rock bolts during the
NATM construction process. The strength increase due to rock bolting is modeled by

118

8.2 Boundary value problems

0.00
Fill

Upper Keuper Marl

10.00

Lower Keuper Marl

6.20 m

20.00
Lacustrine Limestone

0.00

10.00

20.00

30.00

40.00

Figure 8.13: Steinhaldenfeld NATM tunnel: 2D mesh and soil layers.

introducing additional cohesion of 25 kPa in the bolted area. The tunnels linear elastic
lining plate elements have a stiffness of EA = 3.75 106 , a flexural rigidity of EI =
1.95 104 , a Poissons ratio of = 0.2, and a unit weight of C = 25 kN
.
m3
The meshed problem area below the tunnel is relatively small since the Limestone is
much stiffer than the overlaying marl. For the mesh boundaries it is generally proposed
here to require the stiffness multiplier:
Gm 0.8

G0
Gur

(8.2)

in the meshs outermost stress points (at boundaries). For a definition of the stiffness
multiplier Gm refer to Section 7.3.1.
For the numerical analysis of Steinhaldenfeld tunnel, the stress reduction method is
adopted. Using this method, the tunnel excavation remains unsupported by the lining
till, an artificial support pressure p0 is reached, where p0 is the initial rock mass pressure before excavation. Then, the tunnel lining is activated and the support pressure is
removed completely. The factor 0 < < 1 is called the load reduction factor.
Figure 8.14 shows results of the numerical analysis. The settlement trough calculated
by the HS-Small model is clearly steeper and more accurate compared to that calculated
by the original HS model. The remaining discrepancy to the measured settlement trough
can possibly be explained by the inadequateness of 2D analysis for 3D problems. Also
the small-strain stiffness parameters could be underestimated because they are based
solely on literature data and correlations. Set 2 in Figure 8.14 therefore indicates the
changes in settlement from a 40 percent increase of the initial stiffness, displayed in Table
D.2 (Set 1). The failure criterion (Mohr-Coulomb or Matsuoka-Nakai) has almost no
influence on the shape of the settlement trough. However, it has an influence on the load
reduction factor. This is chosen so that in all calculations the settlement directly above

119

Chapter 8 Validation and verification of the HS-Small model

Distance to tunnel axis [m]
0

10

20

30

Distance to tunnel axis [m]

40

-0.01

-0.01

Measurement
-0.02

HS (original)

10

20

30

40

Measurement
-0.02

HS (original)

HS-Small (MC) Set 1

HS-Small (MN) Set 1

HS-Small (MC) Set 2

HS-Small (MN) Set 2

Settlement [m]

Settlement [m]

Figure 8.14: Green field settlement trough due to construction of Steinhaldenfeld NATM
tunnel. Left: HS-Small(MC) prediction. Right: HS-Small(MN) prediction.

the tunnel axis agrees with the measurement. It is found that the load reduction factor is
highest for the HS model = 0.36, followed by the HS-Small(MC) model = 0.32, and
the HS-Small(MN) model = 0.25. Unfortunately the load reduction factor cannot be
experimentally derived. The difference in the calculated factors is however reasonable:
The Matsuoka-Nakai criterion yields the highest strength and stiffness in the given plane
strain conditions; its load reduction factor should therefore be least.
8.2.1.2 Heinenoord Tunnel
The second Heinenoord tunnel, south of Rotterdam is a shield tunnel project in largely
sandy soils. Figure 8.15 shows the finite element mesh, used in the following plane strain
analysis. Figure 8.15 depicts also the different soil layers, for which Table D.3 provides
HS material parameters based on the work by Bakker [10]. Drained material behavior
is assumed in all soil layers. The lining of the second Heinennord tunnel is modeled by
linear elastic plate elements with EA = 3.75 106 , EI = 1.95 104 , = 0.2, and = 25 kN
.
m3
Several numerical simulation methods for the analysis of shield tunnels can be found

in literature. An overview of these methods is given in Moller

[121]. Moller
concludes
that the pressure method is the most suited for shield tunnels. Here, the tunnel excavation is supported by a predefined grout pressure between soil and lining. This pressure
is lowest at the top of excavation proof and increases linearly with depth by the grouts
.
unit weight, which is in the following assumed to = 15 kN
m3
Figure 8.17, and 8.16 depict numerical analysis results for the second Heinenoord tunnel employing the pressure method. Again, there is no markable difference in the settlement trough for the two available yield criteria (Mohr-Coulomb, and Matsuoka-Nakai),
although there is a difference in material strength. In all analyses, the roof pressure proof
is chosen so that the maximum settlement is in close agreement with the measurement
in the tunnel axis. Doing so, the least roof pressure of 125 kPa is obtained from the HS-

120

8.2 Boundary value problems

0.00

GW

Fill

10.00
Sand
8.30 m

20.00
Clay
0.00

10.00

20.00

30.00

40.00

50.00

Figure 8.15: Second Heinenoord tunnel: 2D mesh and soil layers.

Small(MN) analysis for its highest strength and stiffness. The highest roof pressure of
130 kPa is obtained in the analysis using the HS model, in between is the HS-Small(MC)
analysis with a pressure of 129 kPa. All pressures seem very reasonable as they are of
the same size as the overburden pressure.
The Inclinometer measurements (Figure 8.16) at a distance of 6, 10, and 16 meters from
the tunnel axis, as well as the surface settlement troughs (Figure 8.17) underline again the
HS-Small models improved precision in displacement analysis. The HS-Small model is
not perfect though. The Inclinometer measurement at a 6 meter distance from the tunnel
axis show a somewhat different characteristic to the calculated one. Nevertheless, the
decay of displacement with distance from the tunnel axis is modeled nicely. This is
entirely due to the new small-strain formulation. The choice of failure criterion, besides
its influence on the total material stiffness (see roof pressure), has a negligible effect on
the calculated strain distribution.
Due to missing experimental data, the two tunnel examples presented above are not
suited to validate the HS-Small model completely. However, it can be concluded that the
HS-Small model can improve displacement analysis considerably by causing a diffuse

strain localization around the tunnel excavation. Moller

[121] comes to a similar conclusion in his work, in which he additionally considers different simulation methods and
3D approaches in the two tunnel examples.

8.2.2 Excavations
Simpson introduced his brick model in a case study of the British Library deep excavation. The House of Parliament underground car park case study is another deep excavation study, that is often cited along with small-strain stiffness. The tilt of Big Ben clock
tower right next to this excavation illustrates nicely the effects of advanced (small-strain

121

Chapter 8 Validation and verification of the HS-Small model

Horizontal displacement [m]

-0.01

Horizontal displacement [m]

0.01

-0.01

Horizontal displacement [m]

0.01

-0.01

-10

0.01
0

-10

-10

Experiment
HS
HS-Small (MC)

-20

HS-Small (MN)

-20

-20

b
Depth [m]

Depth [m]

Depth [m]

Figure 8.16: Results from Inclinometer measurements next to the second Heinenoord
tunnel. Horizontal displacements are displayed at a distance of (a) 6 m,
(b) 10 m, and (c) 16 m from the tunnel axis.

Distance to tunnel axis [m]

0

10

20

30

Distance to tunnel axis [m]

40

-0.01

Settlement [m]

20

30

40

-0.01

Measurement
-0.02

10

HS (original)

Measurement
-0.02

HS (original)

HS-Small (MC) Set 1

HS-Small (MN) Set 1

HS-Small (MC) Set 2

HS-Small (MN) Set 2

Settlement [m]

Figure 8.17: Green field settlement trough due to construction of the second Heinenoord
tunnel. Left: HS-Small(MC) prediction. Right: HS-Small(MN) prediction.

122

8.2 Boundary value problems

0.00

30.00

excavation step 1

Sand (Layer 1)

-3.00 (GW)

-4.80

excavation step 2

-9.30

excavation step 3

-14.35

excavation step 4

-16.80

27
27

20
Sand (Layer 2)
40

27

8.0

0
A1

8.0

8.0

-17.90 (GW)

A2
A3

hydraulic barrier

60

-30.00

-32.00
80 diaphragm wall

Anchor Prestress

80

A1
A2
A3

Sand (Layer 3)
100
-20

20

40

60

80

100

120

768 kN
945 kN
980 kN

Distance
2.30 m
1.35 m
1.35 m

Length l
15 cm
15 cm
15 cm

19.80 m
23.30 m
23.80 m

Excavation and anchor detail (1:1000)

Figure 8.18: Excavation in Berlin sand: 2D mesh and geometry detail (right).

stiffness) models. These can predict the total excavation heave more correctly and thus
do not suggest that the clock tower tilts away from the pit, which less refined models
may do. Beyond the constitutive model however, there are many more things to consider in the analysis of deep excavations, for example; the interface between soil and
retaining structure, initial stress conditions, structural elements, time effects, and pore
water pressures. Some of these issues should always be thoroughly considered in excavation analysis (interface, initial stresses, structural elements), others are less critical in
less difficult soil conditions. Two examples with less critical soil conditions are chosen
for the validation process of the HS-Small model: An excavation in Berlin sand and an
excavation in Rupel boom clay.
8.2.2.1 Excavation in Berlin sand
The working group 1.6 Numerical methods in Geotechnics of the German Geotechnical
Society (DGGT) has organized several comparative finite element studies (benchmarks).
One of these benchmark examples is the installation of a triple anchored deep excavation
wall in Berlin sand. The reference solution by Schweiger [161] is used here as the starting
point for the next validation example: Both, the mesh shown in Figure 8.18, and the
soil parameters given in Table D.3 are taken from this reference solution. However, the
bottom soil layer defined by Schweiger could be omitted in the analysis when using the
HS-Small model. In the reference solution this layers only purpose is the simulation of
small-strain stiffness due to a lack of small-strain stiffness constitutive models back then.
Figure 8.19, and 8.20 show results from the finite element calculation using the HSSmall(MC), and HS-Small(MN) model respectively. The original reference solution is
calculated with the HS model (Figure 8.19, Table D.4). A second reference solution utilizing the HS model with Matsuoka-Nakai failure criterion, but no small-strain stiffness

123

Chapter 8 Validation and verification of the HS-Small model

Distance to diaphragm wall [m]
0

20

40

Wall deflection [m]

60

-0.04 -0.03 -0.02 -0.01

0
0

-0.004
-0.008

-10

-0.012
-0.016

-20

Settlement [m]
Measurement
HS (Reference)

-30

HS-Small (MC)
Depth below surface [m]

Figure 8.19: Excavation in Berlin sand. HS-Small(MC) model predictions versus measured displacements after the final excavation step. Left: Surface settlement
trough. Right: Lateral wall deflection.

formulation (HS(MN)) is shown in Figure 8.20. From the comparison of HS-Small results
with the reference solutions, the impact, which a) the small-strain stiffness, and b) the
failure criteria have can now be derived.
The small-strain stiffness formulation accumulates more settlements right next to the
wall, whereas the settlement trough is smaller. The triple anchored retaining wall is generally deflected less when using the two HS-Small models. The deflection calculated
by the HS-Small(MC) almost fits that measured. The Matsuoka-Nakai criterion in the
HS-Small(MN) model gives less displacements compared to the Mohr-Coulomb criterion. In comparison to the measured wall deflection, the HS-Small(MN) model even
gives lower displacements. However, this might be related to the problems original reference solution. The chosen parameters are most likely partly back-analyzed from HS
model calculations. In this case the Mohr-Coulomb strength parameters already include
an increase for plane strain conditions. Additionally increasing the strength by using the
HS-Small(MN) model may thus overdo things.
Calculated excavation heaves at the end of excavation are shown in Figure 8.21. The
heave, which is due to elastic unloading, is roughly halved when using the HS-Small
models, independent of the failure criterion chosen. From Figure 8.21 it can be further
concluded that the meshed problem area is not quiet sufficient in the example since the
heave is not asymptotically decaying to zero. Figure 8.22 verifies in this respect the
criterion proposed in Equation 8.2. The monotonic stiffness multiplier Gm at the end of
excavation is well below 3 at the lower mesh boundary.
However, defining the mesh boundary in a depth of 2-3 times the excavation width as

124

8.2 Boundary value problems

Distance to diaphragm wall [m]

0

20

Wall deflection [m]

40

60

-0.04 -0.03 -0.02 -0.01

-0.004
-0.008

-10

-0.012
-0.016

-20

Settlement [m]
Measurement
-30

HS (MN)
HS-Small (MN)

Depth below surface [m]

Figure 8.20: Excavation in Berlin sand. HS-Small(MN) model predictions versus measured displacements after the final excavation step. Left: Surface settlement
trough. Right: Lateral wall deflection.

Section A-A
20.00 m

0.01

0.02

0.03

0.04

Vertical heave [m]

-20

-40

-60

-80

HS (Reference)
HS (MN)
HS-Small (MC)

-100

HS-Small (MN)

Depth below surface [m]

Figure 8.21: Excavation in Berlin sand: Vertical displacements in the excavation pit at a
distance of 10 m from the retaining wall.

125

Chapter 8 Validation and verification of the HS-Small model

0

3.0
20

G m [-]

40

2.0

60
80

1.0
100
-20

20

40

60

80

100

120

Figure 8.22: Excavation in Berlin sand: Factorized increase of unloading/ reloading stiffness at the end of excavation.

proposed by the AK 1.6, and seen in the reference solution by Schweiger, seems reasonable, too. The error in the excavation heave is relatively small. A mesh according to the
requirements of the criterion defined in Equation 8.2 is used in the next example of an
excavation in Ruple clay.
8.2.2.2 Excavation in Ruple clay (Offenbach)
Figure 8.23 shows the geometry of an excavation in Offenbach. Similar to the well known
subsoil conditions of the neighboring town Frankfurt, the predominant soil here is here
a highly overconsolidated clay. Different to Frankfurt clay, the Rupel clay is a oligocene
clay, with no solid rock intrusions. Besides the excavations geometry and measurement
data, Krajewski [95] also published a material parameter set for Rupel clay that can be
used together with the HS model. The published material parameters have been obtained by back-analysis of lab experiments and a large hopper, built at a neighboring
chemical plant. Due to the axisymmetric conditions of all back-analyzed experiments, it
can be assumed that the parameter set given is not adjusted for plane strain conditions
(as it might have been the case in the Berlin sand excavation). Again, there is no smallstrain stiffness testing data available. The parameter set given in Table D.5 is therefore
supplemented by correlated small-strain stiffness data. Figure 8.24 shows the 2D plane
strain mesh used in the numerical analysis of the excavation.
The retaining wall in the example is a bored pile wall with pile diameter r = 0.90m
and axial distance d = 1.30m. The struts are IPBs 340 and IBPs 360 with a clearance of
2.45 m, 1.95 m, and 1.80 m in the top, second, and third strut layer respectively.
The Rupel boom clay has an overconsolidation ratio (OCR) well above 2. Using the
relationship
(8.3)
K0oc = (1 sin )(OCR)sin

126

8.2 Boundary value problems

p =30 kN/m

12.00

0.00

p = 39 kN/m

-2.50

Fill
Rupel boom clay

-2.80 (GW)

IPB 340
-7.60
IPB 360
-13.00
IPB 360
Bored pile wall
Pile diameter: 90 cm
Axial distance: 1.20 m

-19.30
-22.70

Inclinometer

-30.00

Figure 8.23: Excavation in Rupel clay: Geometry of the Offenbach excavation for a railway tunnel.

0
20
40
60
Rupel boom clay
80

-100

-80

-60

-40

-20

20

40

60

80

100

Figure 8.24: Excavation in Rupel clay: 2D mesh used in the plane strain analysis. Neighboring buildings to the excavation are modeled by a line load at the top
ground surface.

127

Chapter 8 Validation and verification of the HS-Small model

Distance to center axis excavation [m]

-40

Wall deflection [m]

40

-0.02

-5

-0.04

0.04

0.08

-10

-0.06
-15

Settlement [m]
Measurement
HS (original)

-20

HS-Small (MC)
HS-Small (MN)

Depth below surface [m]

Figure 8.25: Excavation in Rupel clay. HS-Small model predictions versus measured displacements after the final excavation step. Left: Surface settlement trough.
Right: Lateral deflection of the left retaining wall.

presented by Mayne & Kulhawy [115], the K0 value for generating the initial stress field
is calculated as 0.95, for an assumed OCR of 2.8. Because of stress relaxation, no initial
elastic clay behavior is assumed. Consolidation is considered only to the extend that no
excess pore pressures built up behind the wall.
Figure 8.25 shows the calculated results in comparison to the available measurements
at the end of excavation. The original HS model overall gives too much surface settlements accompanied by too much lateral deformations in the retaining structure. The
HS-Small(MC) model improves results significantly, the HS-Small(MN) model gives almost perfect results. The settlements to the right hand side of the excavation are still
too big, probably due to a slightly overestimated surcharge at the ground surface. Analogous to the previous example, the settlements right next to the wall increase slightly
when using the HS-Small model. This is again related to the reduced excavation heave,
which is shown in Figure 8.26. Excavation heave is calculated to be lowest for the HSSmall(MN) model and highest for the HS model.
Figure 8.27 shows the elastic material stiffness at the end of excavation. Here, stiffness
is plotted as a factorized increase of the unloading-reloading stiffness (see Equation 6.13).
Stiffness increases due to strain reversals are not considered in this representation. In
agreement with the criterion defined in Equation 8.2, all calculated displacements (see
Figure 8.25, and 8.26) are negligible in areas where Gm > 0.8 GGur0 .
In conclusion, the HS-Small model(s) performed very well in both excavation examples, although there has been no small-strain stiffness test data available at all. All small-

128

8.2 Boundary value problems

0.04

0.08

0.12

0.16

0.2 Vertical heave [m]

-20

-40

-60

HS (Reference)
HS (MN)
-80

HS-Small (MC)
HS-Small (MN)

Depth below surface [m]

Figure 8.26: Excavation in Rupel clay: Vertical displacements in the excavation pits symmetry.

4.0

0
20

G m [-]

3.0
40

2.0

60
80

1.0
-100

-80

-60

-40

-20

20

40

60

80

100

Figure 8.27: Excavation in Rupel clay: Factorized increase of unloading-reloading stiffness at the end of excavation.

129

Chapter 8 Validation and verification of the HS-Small model

strain stiffness parameters have been solely picked on the basis of data and correlations
presented in Chapter 3 of this thesis. In order to account for diagenetic effects, smallstrain stiffness parameters have been generally chosen in the upper range of possible
values. The threshold shear strain 0.7 in sand, for example, was generally assumed at
2 104 .
Two main tendencies could be identified in this section: a) The HS-Small model gives
less excavation heave than the HS model and hence, less uplift of the retaining structure
and the adjoining soil, and b) the HS-Small model generally yields displacements that
are in noticeable better agreement with the experimentally derived results than those of
the HS model.

8.2.3 Spread Foundations

Spread foundations seem to be simple problems compared to the analysis conducted before. Most spread footings can be numerically analyzed in a plane strain or axisymmetric
approach. Axisymmetric conditions may also be suitable for quadratic footings under
service loads. A quadratic footing is also the last verification example presented in this
chapter. A small note on the initialization procedure of the HS-Small model is included
at the end of this section.
8.2.3.1 Texas A&M University spread footing on sand
Five load tests on spread footings ranging from 1x1 to 3x3 m in size were conducted
at Texas A&M Universitys National Geotechnical Experimental Site [19]. Vertical loads
were applied at the center of the footings which rested on flat ground. Load-settlement
curves as well as inclinometer and extensometer measurements were recorded for each
load test. In particular, the North 3x3 m footing has been chosen for the following HSSmall verification. The footing has a thickness of 1.22 m and an embedded depth of 0.76
m. Figure 8.28 shows the test site as well as the 2D axisymmetric finite element mesh
used in the following axisymmetric calculation. In this, the quadratic footing is modeled
by a circular one with an equivalent loaded area.
The HS-Small model parameters shown in Table D.6 have been determined from several available triaxial and resonant column tests. Figure 8.29 shows the comparison of
experimental data to model simulations using this parameter set. The triaxial tests could
be modeled nicely. The resonant column tests scattered too much to allow them to be
reasonably well simulated by any model. Nevertheless, the resonant column tests give
a first idea of the small-strain stiffness parameters to be used for the sandy site.
Ignoring the available field test data in [19], but strictly applying the parameters determined from laboratory tests, yields the load settlement curves shown in Figure 8.30.a.
Note that since the HS-Small(MC) and the HS-Small(MN) models yield almost exactly
the same settlement curves in the axisymmetric problem, only one curve is drawn for
both models. The well improved load settlement curves shown in Figure 8.30.b are
obtained by additionally accounting for the soils actual initial stress state. From the

130

8.2 Boundary value problems

A
0.00

5.00

10.00

15.00

0.00

5.00

10.00

15.00

20.00

Figure 8.28: Load test on a spread footing on sand. Left: Texas A&M Universitys National Geotechnical Experimental Site. Right: 2D axisymmetric mesh used
in the finite element analysis.

available CPT and Pressuremeter field tests it can be derived that the sandy site is overconsolidated. The amount of overconsolidation is here back-analyzed from the load settlement curves. Yet, the stiffness in initial loading and in unloading-reloading cycles is
still underestimated. Increasing the initial stiffness derived from laboratory test data by
40% yields the results presented in Figure 8.30.c. Now, the unloading-reloading stiffness
is simulated almost perfectly. However, the overall load settlement curve overestimates
soil stiffness slightly, which may be due to the axisymmetric simplification of the given
3D problem.
As shown in Chapter 3, stiffness discrepancies in the magnitude found in the A&M
footing example is not very uncommon. Nevertheless, it should be noted that underestimating small-strain stiffness by using laboratory data or correlations is always better
than not accounting for small-strain stiffness at all. This, for example, can also be concluded from Figure 8.30: The HS-Small model gives more realistic simulation results
than the original HS model regardless of initial stress and stiffness is used.
From a practical point of view, the most important difference in the load settlement
curves is probably their initial slope (assuming that ultimate load design can also be
handled by other methods). In the given example, service load settlements are overestimated by more than 80% when not using the HS-Small model.
The Texas A&M spread footing example is a nice demonstration of the benefits of
small-strain stiffness on the one hand. On the other hand, it demonstrates the sensitivity
of numerical calculations to initial stress conditions. As there is also a link between
initial stress conditions and a materials strain history, the initialization of strain history
is briefly discussed in the next section.

131

Chapter 8 Validation and verification of the HS-Small model

s3 = 34.5 kPa CD

s1-s3

GSecant [kN/m2]

eVol[-]
-0.03

80

80000

All experimental data

acquired at s3 = 50 kPa

60000
-0.02
40000

40
-0.01
0.04

0.08

20000

0.12

0.00

e1[-]

s3 = 138 kPa CD

s1-s3

0.0001

0.01

e1-e3[-]

GSecant [kN/m2]

eVol[-]

400

0.001

-0.03

120000

-0.02

80000

-0.01

40000

All experimental data

acquired at s3 = 150 kPa

300
200
100
0.04

0.08

0.12

0.00

e1[-]

s3 = 345 kPa CD

s1-s3

0.0001

800

0.01

e1-e3[-]

GSecant [kN/m2]

eVol[-]
-0.03

0.001

200000

All experimental data

acquired at s3 = 300 kPa

150000
-0.02
100000

400
-0.01

0
0.00

50000

0.00
0.04

0.08

0.12

e1[-]

0
0.0001

0.001

0.01

e1-e3[-]

HS (original)

Experiment

Note: Resonant column tests for

HS-Small

Santos & Correia

various sampling depths from 0 to 6 m.

Figure 8.29: Triaxial tests on sand samples from the A&M test site (0.6 m sampling
depth). Small-strain stiffness parameters are obtained by averaging the results from several resonant column tests as shown on the right hand side.
132

8.2 Boundary value problems

4000

8000

4000

8000

0
Load [kN]

-0.04

-0.04

-0.08

-0.08

-0.12

-0.12

1a

Load [kN]

4000

1b

Settlement [m]

8000

4000

Settlement [m]

8000

Load [kN]

-0.04

-0.04

-0.08

-0.08

-0.08

-0.12

-0.12

-0.12

2b

Settlement [m]

Experiment

8000
Load [kN]

Load [kN]

-0.04

2a

4000

Settlement [m]

HS (original)

2c

Settlement [m]

HS-Small

Figure 8.30: Calculated and experimentally derived load-settlement curves for the A&M
3x3 m footing on sand (HS - top row; HS-Small - bottom row): a) NC sand,
b) OC sand with laboratory small-strain stiffness, c) OC sand with increased
(in-situ) small-strain stiffness.

133

Chapter 8 Validation and verification of the HS-Small model

A

0.00

0.00
110.00

110.00

98.00

5.00

98.00

5.00

86.00

86.00

74.00

74.00

10.00

10.00
62.00

62.00

50.00

15.00

50.00

15.00

G [kN/m]

0.00

5.00

10.00

15.00

20.00

G [kN/m]

0.00

5.00

10.00

15.00

20.00

Figure 8.31: Initial elastic stiffness of the HS-Small model obtained by different initialization procedures in the A&M footing example. Left: No strain history
reset between removing the surcharge used to model the materials geologic
history and construction of the footing. Right: Full strain history reset by
applying a small reverse load step before introducing the footing.

8.2.4 Initialization of the HS-Small model

Stress relaxation erases a soils memory of previous applied stress. Soil ageing in the
form of particle (or assembly) reorganization during stress relaxation and formation of
bonds between them can erase a soils strain history. Considering that the second process
in a naturally deposited soil develops relatively fast, the strain history should start from
zero in most boundary value problems. This is the default setting in the HS-Small model.
However, sometimes an initial strain history may be desired. In this case the strain
history can be triggered by applying an extra load step before starting the actual analysis. Such an additional load step might also be used to model overconsolidation. Usually
the overconsolidations cause has vanished long before the start of calculation, so that
the strain history should be reset afterwards. However, by adding and removing a surcharge the HS-Small models strain history is already triggered. In this case the strain
history can be reset manually, either by replacing the material or code manipulation.
More convenient is sometimes the application of a small reverse load step.
Figure 8.31 shows the A&M footing example after two different initialization runs: On
the left hand side the (overconsolidation) surcharge is removed just before the footing is
built. By introducing the footing, the previous uplift is reversed below the footing but
nowhere else. On the right hand side, a small surcharge is applied after removing the
overburden loading and before introducing the footing. This small reverse loading step
(in the example 1 kPa is used) resets the entire strain history of the problem. However,
the problem is not reset in all of its aspects. The distinction between initial, or virgin
loading, and secondary loading is made on the basis of the monotonic strain history.

134

8.2 Boundary value problems

A = 4500 kN (initial loading)

A = 4500 kN (reloading)

0.00

0.00
110.00

110.00

98.00

5.00

98.00

5.00

86.00

86.00

74.00

74.00

10.00

10.00
62.00

62.00

50.00

15.00

50.00

15.00

G [kN/m]

0.00

5.00

10.00

15.00

20.00

G [kN/m]

0.00

A = 8000 kN

5.00

10.00

15.00

20.00

A = 12000 kN

0.00

0.00
110.00

110.00

98.00

5.00

98.00

5.00

86.00

86.00

74.00

74.00

10.00

10.00
62.00

62.00

50.00

15.00

50.00

15.00

G [kN/m]

0.00

5.00

10.00

15.00

20.00

G [kN/m]

0.00

5.00

10.00

15.00

20.00

Figure 8.32: Development of elastic stiffness in the A&M footing example.

An overburden surcharge will therefore trigger secondary loading although the strain
history is reset. If that is not desired, the monotonic strain history can be manually
reset or an initial stress procedure as offered for example in the finite element program
P LAXIS, can be used.
In discussing the strain historys reset, loading-unloading-reloading cycles as seen in
the A&M footing example, are also worth a closer look. Strain history is fully reset in 180
degree load reversals. Due to isotropic hardening, the accumulated strains are usually
less in reloading than in primary loading. Strain history is hence more salient in virgin
loading than in reloading. This can be observed for example in the decay of elastic
material stiffness shown in Figure 8.32.

135

Chapter 8 Validation and verification of the HS-Small model

8.2.5 Capabilities and limitations of the HS-Small model

The HS-Small model passed all of the above validation tests reasonably well. Its predictions are without exception more reliable than those of the original HS model, even
if default small-strain stiffness parameters are used. The HS-Small model with MohrCoulomb criterion can therefore be readily recommended for all applications where the
HS model could also be used. The HS-Small model with Matsuoka-Nakai failure criterion presumes additional material strength and stiffness. Stiffness is particularly increased in unloading problems due to the extended size of the isotropically hardened
yield surface. Therefore, the HS-Small model with Matsuoka-Nakai failure criterion
needs to be further validated before it is used in ultimate load design applications.
The main objective of this thesis is to develop a capable small-strain stiffness model
for practical applications. This objective has obviously been met by the introduction of
the HS-Small model. However, the HS-Small model also has some limitations. As the
user of a model should always be aware of its limitations, those of the HS-Small model
are summarized here.
Most of the HS-Small models limitations are purely inherited from the original HS
model. As already pointed out in the element test section, these are mainly its void ratio independent formulation and a lack of kinematic hardening. The HS-Small model
is therefore not suited for unified modeling (modeling of material at different void ratios with a single set of material parameters) and cyclic modeling. Despite its modified flow rule and as a consequence of the void ratio independent formulation, the HSSmall model should be used with the same care in undrained analysis as the HS model.
Undrained calculations with the HS model are explored in more detail in Wehnert [191].
A limitation unique to the HS-Small model is its exaggerated stiffness in isotropic
loading conditions (Figure 8.33). The models assumption that soil grains or assemblies
have infinite strength is not validated. This was to be expected. Still, this is not a serious
model limitation as pure isotropic loading is unlikely in boundary value problems. In
K0 loading, the models response correlates very well to experimental data (Figure 8.1).

136

8.2 Boundary value problems

s1[kN/m2]

s1[kN/m2]

Dense Hostun Sand

800

800

600

600

400

400

200

200

0
0.000

0.004

0.008

Experiment

0.012

ev[-]

0
0.000

HS (original)

0.004

Loose Hostun Sand

0.008

0.012

ev[-]

HS-Small

Figure 8.33: Isotropic compression test on dense (left) and loose (right) Hostun sand.

137

Chapter 9
3D case study Sulfeld

Lock
The Mittellandkanal (MLK) links the rivers Rhine, Ems, Weser, and Elbe to Berlin and to

eastern Europes waterways. Eastbound ship traffic is lowered at the Sulfeld

Lock from

the channels peak altitude of 65 m to 56 m height above sea level. The existing Sulfeld
Lock was constructed between 1934-37 and was opened in 1938. The locks southern
chamber is currently being rebuilt in order to adapt it to raise ship traffic and generations
of new ships. The northern lock chamber including its six economizing basins is to be
kept fully operational during the 4 year construction period. The high-speed railway
link between Hanover and Berlin lies close to the construction site and also needs to be

kept operational at all times during construction. Figures 9.1 and 9.2 show the Sulfeld
lock and the high-speed railway link passing nearby.
Investigation of the site shows that overconsolidated glacial deposits overlie Lias bedrock. The glacial deposits are composed of clayey silt, sand, and glacial till (Figure 9.3).
The cohesive layers (silt and glacial till) were characterized in laboratory tests. A two-

Figure 9.1: Birds eye view of Sulfeld

Lock before excavation. The northern lock chamber
(left) is still in operation. The southern lock chamber (right) is currently being
rebuilt.

139


Chapter 9 3D case study Sulfeld
Lock

Figure 9.2: Strutted and back-anchored Sulfeld

excavation near the railway bridge. The
section left hand side of the channels centerline is shown in Figure 9.3.

step model calibration procedure was employed: First, all test results were statistically
averaged before secondly, selected tests were back analyzed in numerical element tests.
Properties of the sand layer were estimated from in-situ heavy dynamic penetration test
results. The obtained material parameters were finally calibrated in a 2D finite element
model by comparing calculated displacements of the existing lock during operation to
measured ones. In the calibration procedure, the confidence in the correlated smallstrain stiffness parameters could be specifically increased. An overview of the final material sets used in the 3D analysis is given in Table D.7.
The excavation for the new lock chamber is fully housed by vertical retaining walls.
The selected type of retaining wall is closely related to the type and presence of adjacent
structures. A tied back diaphragm wall is built in sections, close to structures that are
sensitive to settlements. The most sensitive part of the excavation pit next to the railway
link is strutted. A cut-off wall is built in all sections where settlements are tolerable. If
required by design, the cut-off walls are reinforced by sheet piles.
The HS-Small model is deployed here in a 3D finite element analysis of the interaction
between the excavation pit and railway bridge abutments. As solely the Mohr-Coulomb
yield criterion is used in the following, the term HS-Small refers to HS-Small(MC) within
this chapter. Figure 9.3 shows the geometry, structural features, and excavation stages
modeled within the finite element code ABAQUS. The structural features include two
abutments on pile foundations, back-anchored and strutted retaining walls, sheet pile
walls, and a floor slab in the strutted section. Embedded elements are used to model the
floor slab, sheet pile walls and piles. All other structural features and the subsoil are dis-

140

Abutment backfill (not removed)

Fill (removed in pre-excavation step)

3.0 m (step 0)

Ship channel
Abutments with piled found.

80 m

Excavation
+57 mNN

Fill
Glacial till
27 m

2.0 m (step 1)

Sand
Clayey silt

4.0 m (step 2)

+46 mNN

5.0 m (step 3)

17.7 m

70 m

Figure 9.3: Geometry and Geology of the Sueldfeld excavation in the 3D FE model. Lias
bedrock marks the lower boundary of the model.

cretized by full integrated volume elements. Four excavation stages are introduced: One
before wall construction (pre-excavation), and three after wall construction (excavation
step 1 to 3). Construction of the retaining walls is not modeled accurately. Instead, they
are placed as a whole. The phreatic level is adjusted several times in the calculation due
to groundwater discharge in the bedrock or groundwater lowering in the fill. After the
final excavation step, the groundwater table is lowered due to works near the pumping
station. This change in phreatic level is the final load step considered in the analysis. A
list of all calculation stages is given in Table 9.1. As pore water measurements indicate
a relatively fast dissipation of excess pore pressures in the vicinity of the excavation pit,
drained conditions are assumed.

The Sulfeld
analysis main focus is on displacements of the railway bridge abutments
and their backfill. Due the rather unprecise numerical retaining wall placement method,
measured and calculated abutment displacements are compared for excavation stages 1

141


Chapter 9 3D case study Sulfeld
Lock

Table 9.1: Stages in the Sulfeld

3D calculation
Stage

Action

1
2
3
4
5
6
7
8
9
10

Initialization
Pre-excavation (Excavation to +57 m NN)
Installation cut-off wall
Phreatic level at +53 m NN (local vacuum)
Installation diaphragm wall
Phreatic level at +57 m NN
Excavation to +55 m NN (excavation step 1)
Excavation to +51 m NN (excavation step 2)
Excavation to +46 m NN (excavation step 3)
Phreatic level at +53 m NN (channel section only)

Bottom slab

vertical heave [mm]

Excavation step 3
8
4
0
-4
-8

06.2005

07.2005

08.2005

09.2005

Figure 9.4: Heave measured in the excavation pit during excavation step 3.

to 3 and the final change in phreatic level only. The well equipped site allows for various
other comparisons, as well. The excavation heave for example, is presented in Figure 9.4.
From extensometer readings in the excavation pit, the final heave during excavation step
3 can be quantified as 23 mm. In the analysis, heave 4 mm is obtained when using
the HS-Small model. The HS analysis over-predicts the excavation heave by more than
8 mm. The overall vertical displacements from the two analysis are compared in Figure
9.5. The measured settlement (extensometer) of the abutment closest to the excavation
pit is shown in Figure 9.6. From geodetic measurements it can be additionally derived
that the abutments are not tilting. Exactly the same behavior is found in the HS-Small
analysis: Almost no tilt of the abutments occurs with a total settlement of 2 mm. The
HS calculation on the other hand, shows settlements of up to 10 mm in combination with
tilt.
Figure 9.8 shows a comparison of measured and calculated horizontal displacements
of both bridge abutments. Measured displacements were separated in displacements
due to pre-excavation and wall construction (not shown) and displacements due to ex-

142

a.) Vertical displacement uz - HS

uz [cm]
6.5
4.5
2.5

b.) Vertical displacement uz - HS-Small

0.5
-1.5
-3.5

Figure 9.5: Vertical displacements due to excavation (step 1 to 3) and groundwater lowering.

settlement [mm]

Pre-excavation
(wall constr. etc.)

Excavation Lowering
step 1-3
GW

0
1
2
3
4

01.2004

01.2005

01.2006

Figure 9.6: Settlement of the abutment closest to the excavation pit.

143


Chapter 9 3D case study Sulfeld
Lock
a.) Vertical displacement uz - HS

b.) Vertical displacement uz - HS-Small

uz [mm]
0.0
-5.0

-10.0

-15.0

Figure 9.7: Vertical displacements of the abutments due to excavation and groundwater
lowering.

cavation step 1 to 3 and final groundwater lowering. The latter can be directly compared
to the deformed abutments drawn to scale for both, HS and HS-Small analysis. The overall trend in the HS-Small calculation is in better agreement with the measured displacements compared to those of the HS model. Right next to the abutment front walls, the
calculated displacements resemble reasonably well the measured displacements. However, the wing walls do not deform extensively enough, which could be partly due to the
linear elastic concrete material model chosen in the calculation.
Figure 9.9 finally depicts the factorized increase of the unloading-reloading stiffness
at several calculation stages. For illustration purposes, the factorized stiffness increase is
shown in the clayey silt only. It is only towards the end of excavation, that small-strain
stiffness disappears with greater distance to the excavation pit. Remarkably is also the
effectiveness of the pile foundation. Within the reinforced part of the soil, almost no
stiffness reduction can be observed.

In summary, the HS-Small model could be successfully deployed in the Sulfeld

3D finite element analysis. It significantly improved all results obtained from the HS model.
Same as in the previous calculations, the improvement is achieved with zero strain history at the onset of loading. For complex geometries however, the numerical initialization of the HS-Small model can be challenging. In the ABAQUS code for example, initial
stress fields are not calculated with reference to surface geometry. Gravity loading on
the other hand accumulates much deformation and affects strain history. For users of
the ABAQUS code, a remedy to the initialization difficulty is to restart the analysis after
initialization.

144

5 mm
4 mm

11 mm

5 mm 4 mm

18 mm

8 mm

a.) Lateral displacement - HS

5 mm
4 mm

11 mm

5 mm 4 mm

18 mm

8 mm

b.) Lateral displacement - HS-Small

Figure 9.8: Horizontal displacements of the railway bridge abutments.

145


Chapter 9 3D case study Sulfeld
Lock

Gm [-]

6.0

3.5

1.0

Legend:
Construction stages
1: Pre-excavation
2: Excavation step 1
3: Excavation step 2
4: Excavation step 3
5: Lowering GW

Figure 9.9: Modulus reduction during excavation and GW lowering.

146

Chapter 10
Conclusions
The strain range in which soils can be considered truly elastic, i.e where they recover
from applied straining almost completely, is very small. With increasing strain amplitude, soil stiffness decays non-linearly: Plotting soil stiffness against log(strain) yields
characteristic S-shaped stiffness reduction curves. Hardin & Drnevich [52] first described
these small-strain stiffness reduction curves with a simple hyperbolic formula. Since
then, many additional experimental findings have been reported in the literature. These
findings concern mainly the qualitative and quantitative influence of various parameters on small-strain stiffness and stiffness reduction curves. Void ratio and mean stress,
for example, are two parameters that strongly affect small-strain stiffness. The influence
of many other parameters, such as cementation, grain size distribution, plasticity index
etc. is likewise appreciated, but is less straightforward to quantify. Recently, inherent
and stress induced anisotropic features of small-strain stiffness have became important
themes in research. At the same time, engineers working in practice have only very
limited possibilities to include small-strain stiffness in design. There is a lack of simple
and capable small-strain stiffness models. Those small-strain stiffness models available
(Chapter 5) are either research orientated, and/or applicable to specific loading paths
only. The objective of this thesis was therefore, to develop a capable small-strain stiffness model suitable for the engineering community.
The Small-Strain Overlay model proposed in this thesis generalizes the hyperbolic
Hardin-Drnevich model to multi-axial strain space. Although it takes into account strain
induced anisotropy, the overlay model renders small-strain stiffness isotropic. It can
therefore be easily combined with many existing elastoplastic constitutive models that
are based on isotropic elasticity. In fact, these are the kind of models most commonly
used in engineering practice today. One of these is the Hardening Soil (HS) model, as
implemented in the finite element code P LAXIS V8 [21]. The combination of the HS
model with the small-strain overly model introduced in this thesis, is named the HSSmall model. The HS-Small model was validated in a single stress point and various
boundary value problems. In these, all effects that are commonly attributed to smallstrain stiffness in soil-structure interaction could be recognized: The width and shape
of settlement troughs, for example, are more precisely modeled and excavation heave
is reduced to a realistic value. The overall reliability of numerical displacement analysis is considerably increased. Analysis results are also less sensitive to the choice of
proper boundary conditions. Large meshes no longer cause extensive accumulation of
displacements, because marginally strained mesh parts are very stiff. Conserved smallstrain stiffness at the meshs boundaries is an indicator for its proper extent.

147

Chapter 10 Conclusions
The HS-Small model with the Mohr-Coulomb failure criterion is recommended for
all applications where the HS model could also be used. The HS-Small model with the
Matsuoka-Nakai failure criterion seems to perform a little better than its Mohr-Coulomb
counterpart. Its increased strength and stiffness should however be further validated
before it is applied in design. If, for a certain design situation the original HS model is
already at hand, it may be beneficial to also stay with the HS-Small(MC) model: Plane
strain calibrated strength and stiffness parameters yield overly high material stiffness
and strength using the HS-Small(MN) model.
Application of the HS-Small model is straightforward. Its Small-Strain Overlay extension requires only two additional parameters. These are the very small-strain stiffness
G0 and the threshold shear strain 0.7 at which stiffness is reduced to 0.7G0 . In case no
small-strain stiffness in-situ or laboratory test data are available, these two parameters
can be correlated with various other soil or test parameters. In the validation process
of the HS-Small model, the correlations summarized in Chapter 3 of this thesis could
be successfully applied. Although these correlations bear uncertainties, it seems advisable to use them if the only alternative is to neglect small-strain stiffness. Neglecting
small-strain stiffness certainly produces higher analysis errors than using correlations
with slight uncertainties. Most correlations are at the lower end of in-situ determined
small-strain stiffness anyway.
The adaptation of G0 and 0.7 to parameters that affect them within the Small-Strain
Overlay model could be improved in future work. In the present implementation, no
void ratio dependency is assumed. This is in clear contradiction to the findings presented in Chapter 3. However, as most models that are currently used in engineering
practice assume a constant void ratio, this concept is also adopted in the present implementation. The main drawback of void ratio independent constitutive models is that
their material parameters have to be adjusted to the actual void ratio via manual user
input. Mean stress dependency of small-strain stiffness is taken into account using the
same power law as for large strains. In this way, the number of additional material parameters is kept at a minimum. For clays, the large strain power law exponent might
be slightly too high, though. The threshold shear strain on the other hand, is not explicitly corrected for mean stress in the present implementation. The latter compensates
for overly high power law exponents. Future work on the Small-Strain Overlay model
should also include further validation of the strain history mapping presented in Chapter 6. This can presently not be undertaken due to a lack of experimental data.
Future work on the HS-Small model might include the formulation of a cap-type yield
loci similar to that of the Cam-Clay model. This would allow for degenerating the HSSmall model to a full Cam-Clay model. Kinematic hardening should be introduced to
open up the HS-Small model to cyclic loading applications as well. Within a preliminary
implementation, the possibilities of introducing bounding surface plasticity to the HSSmall model are shown in [186]. The resulting model is illustrated in Figure 10.1.
In summary, the HS-Small model is a suitable tool to incorporate small-strain stiffness
into routine design. Testing procedures and correlations for determining its small-strain
parameters are described in Chapter 3. From experience so far, no situations are known
where the HS-Small model gives less reliable results than the often validated HS model.

148

Figure 10.1: Bounding surface model with cap in principal strain space. Only the highlighted cap portion is active. Right: View inside the model onto yield-,
dilatancy-, critical-, and bounding surface.

In contrary, analysis results are throughout, in better agreement with experimental findings when the HS-Small model is deployed. Combining the Small-Strain Overlay model
with other existing elastoplastic models should turn out to be equally beneficial.

149

Bibliography
[1] A. Al-Tabbaa. Permeability and stress strain response of speswhite kaolin. PhD thesis,
University of Cambridge, 1987.
[2] A. Al-Tabbaa and D. Muir Wood. An experimentally based bubble model for
clay. In S. Pietruszczak and G. N. Pande, editors, Numerical Models in Geomechanics:
NUMOG III, pages 9199. London: Elsevier Applied Science, 1989.
[3] I. Alpan. The geotechnical properties of soils. Earth-Science Reviews, 6:549, 1970.
[4] D.G. Anderson and K.H. Stokoe. Shear modulus: A time-dependent soil property,
chapter Dyn. Geotech. Test., pages 6689. Number STP 654. ASTM Spec. Tech.
Publ., Philadelphia, 1978.
[5] D.G. Anderson and R.D. Woods. Time-dependent increase in shear modulus of
clay. Proc. ASCE: Journal of the Geotechnical Engineering Division, 102(GT5):525537,
1976.
[6] J.H. Argyris, G Faust, J Szimmat, P Warnke, and K William. Recent developments
in finite element analyses of prestressed concrete reactor vessels. Nuclear Engineering and Design, 28:4275, 1974.
[7] J. H. Atkinson. Non-linear soil stiffness in routine design. Geotechnique, 50(5):
487508, 2000.
[8] J.H. Atkinson and J.A. Little. Undrained triaxial strength and stress-strain characteristics of a glacial till soil. Can. Geotech. J., 25(3):428439, 1988.
[9] J.H. Atkinson and G. Sallfors. Experimental determination of soil properties. In
Proc. 10th ECSMFE, volume 3, pages 915956, Florence, 1991. General Report to
Session 1.
[10] J.K. Bakker. Soil retaining structures - development of models for structural analysis.
PhD thesis, Technical Univesity of Delft, 2000.
[11] G. Baldi, R. Bellotti, V.N. Ghionna, M. Jamiolkowski, and D.C.F. Lo Presti. Modulus of sands from cpts and dmts. In Proc. 12th Int. Conf. SMFE, volume 1, pages
165170, Rio de Janeiro, 1989.
[12] J.P. Bardet. Lode dependences for isotropic pressure-sensitive elastoplastic materials. Transactions of the ASME, 57(9):498506, 1990.

151

Bibliography
[13] K. Been and M.G. Jefferies. A state parameter for sands. Geotechnique, 35(2):99112,
1985.
[14] R. Bellotti, M. Jamiolkowski, D.C.F. Lo Presti, and D.A. ONeill. Anisotropy of
small strain stiffness in ticino sand. Geotechnique, 46(1):115131, 1996.
[15] T. Belytschko, W.K. Liu, and B. Moran. Nonlinear Finite Elements for Continua and
Structures. John Wiley & Sons Ltd., Chichester, 2000.
[16] J. Biarez and P.-Y. Hicher. Elemantary Mechanics of Soil Behaviour. Balkema, 1994.
[17] L. Bjerrum. Problems of soil mechanics and construction on soft clays. In Proc. 8th
Int. Conf. SMFE, volume 3, pages 111159, Moscow, 1973.
[18] P.G. Bonnier. Implementational aspects of constitutive modelling. SCMEP Workshop No 1 at NTNU, August 2000.
[19] J.-L. Briaud and R. Gibbens. Large scale load tests and data base of spread footings on sand. Final Report FHWA-RD-97-068, U.S. Department of Transportation.
Federal Highway Administration, 11 1997.
[20] E.G.M. Brignoli, M. Gotti, and K.H. Stokoe. Measurement of shear waves in laboratory specimens by means of piezoelectric transducers. Geotechnical Testing Journal,
19(4):384397, 1996.
[21] R.B.J. Brinkgreve, editor. P LAXIS, 2D Version 8. AA. Balkema, 2002.
[22] B.B. Broms and A.O. Casbarian. Effects of rotation of the principal stress axes and
of the intermediate principal stress on the shear strength. In Proc. 6th Int. Conf.
SMFE, volume 1, pages 179183, 1965.
[23] J.B. Burland. Contribution to discussion on session 4. In Proc. 7th Eur. Conf. Soil
Mech., volume 4, page 137, Brighton, 1979. SMFE.
[24] J.B. Burland. Small is beautiful - the stiffness of soils at small strains, Ninth Laurits
Bjerrum memorial lecture. Canadian Geotechnical Journal, 26:499516, 1989.
[25] J.B. Burland, B. Simpson, and H.D. St. John. Movements around excavations in
London clay. In Proc. 7th Eur. Conf. Soil Mech., volume 1, pages 1330, Brighton,
1979. SMFE.
[26] T.K. Caughey. Sinusoidal excitation of a system with bilinear hysteresis. Journal of
Applied Mechanics, ASME, 27(4):640643, 1960.
[27] C.S. Chang. Micromechanical modelling of constitutive relations for granular material. In M. Satake and J.T. Jenkins, editors, Micromechanics of Granular Materials,
pages 271278, Amsterdam, The Netherlands, 1978. Elsevier.

152

Bibliography
[28] S.K. Chaudhary, J. Kuwano, and Y. Hayano. Measurement of quasi-elastic stiffness parameters of dense Toyoura sand in hollow cylinder apparatus and triaxial
apparatus with bender elements. Geotechnical Testing Journal, 27(1):113, 2004.
[29] K.W. Cole and J.B. Burland. Observation of retaining wall movements associated
with a large excavation. In Proc. 5th ECSMFE, volume 1, page 445453, Madrid,
1972.
[30] M.B. Darendeli and K.H. Stokoe. Development of a new family of normalized
modulus reduction and material damping curves. Engrg. Rpt. GD01-1, University
of Texas, Austin, Texas, 2001.
[31] J. Desrues, P.A. Vermeer, and B. Zweschper. Database for tests on Hostun RF sand.
Institutsbericht 13, Universitat Stuttgart, 2000.
[32] T. Doanh. Constitutive Modelling of Granular Materials, chapter Strain resonse envelope: a complementary tool for evaluating hypoplastic constitutive equations,
pages 375396. Berlin Heidelberg: Springer, 2000.
[33] D.C. Drucker. On uniqueness in the theory of plasticity. Quart. Appl. Math., XIV:
3542, 1956.
[34] D.C. Drucker. Introduction to Mechanics of Deformable Solids. McGraw-Hill, 1967.
[35] D.C. Drucker and W. Prager. Soil mechanics and plastic analysis or limit design.
Quart. Appl. Math., 10(2):157165, 1952.
[36] J.M. Duncan and C.-Y. Chang. Nonlinear analysis of stress and strain in soil. Proc.
ASCE: Journal of the Soil Mechanics and Foundation Division, 96:16291653, 1970.
[37] R. Dyvik and C. Madshus. Lab measurements of Gmax using bender elements. In
Advances in the Art of Testing Soils Under Cyclic Conditions, pages 186196., Detroit,
Michigan, 1985. ASCE. ASCE convention.
[38] W. Ehlers. A single-surface yield function for geomaterials. Arch. Appl. Mech., 65:
246259, 1995.
[39] A.L. Fernandez and J.C. Santamarina. Effect of cementation on the small-strain
parameters of sands. Can. Geotech. J., 38:191199, 2001.
[40] V. Fioravante. Anisotropy of small-strain stiffness of Ticino and Kenya sands from
seismic wave propagation measured in triaxial testing. Soils and Foundations, 40(4):
129142, 2000.
[41] E. Flavigny, J. Desrues, and B. Palayer. Le sable dHostun RF. Rev. Franc. Geotechn.,
53:6770, 1990.

153

Bibliography
[42] D. Forel, T. Benz, and W.D. Pennington. Seismic Data Processing with Seismic Un*x.
Number 12 in Course Notes. Society of Exploration Geophysicists, Houston, TX,
2005.
Normung (DIN). DIN EN ISO 14688. Berlin: Beuth Verlag,
[43] Deutsches Institut fur
11 2004. Geotechnical investigation and testing. Identification and classification of
soil.
[44] V.N. Georgiannou, S. Rampello, and F. Silvestri. Static and dynamic measurements
of undrained stiffness on natural over consolidated clays. In Proc. 10th ECSMFE,
volume 1, pages 9195, Florence, 1991.
[45] J.D. Goddard. Nonlinear elasticity and pressure-dependent wave speeds in granular media. Proceedings of the Royal Society London, (430):105131, 1990.
[46] A. Grammatikopoulou. Development, implementation and application of kinematic
hardening models for overconsolidated clays. PhD thesis, University of London (Imperial College), 2004.
[47] G. Gudehus. A comparison of some constitutive laws for soils under radially symmetric loading and unloading. In Proc. of 3rd Num. Meth. in Geomechanics, volume 4,
pages 13091323. A.A. Balkema, 1979.
[48] M.J. Gunn. The prediction of surface settlement profiles due to tunnelling. In
G.T. Houlsby and A.N. Schofield, editors, Predictive Soil Mechanics, pages 304316.
Thomas Telford, 1993. Proc. Wroth Memorial Symposium.
[49] B.O. Hardin. The nature of stress-strain behaviour of soils. In Proc. Earthquake
Engineering and Soil Dynamics, volume 1, pages 390, Pasadena, CA, 1978. ASCE,
New York. State-of-the-art report.
[50] B.O. Hardin and W.L. Black. Vibration modulus of normally consolidated clays.
Proc. ASCE: Journal of the Soil Mechanics and Foundations Division, 94(SM2):353369,
1968.
[51] B.O. Hardin and W.L. Black. Closure to vibration modulus of normally consolidated clays. Proc. ASCE: Journal of the Soil Mechanics and Foundations Division, 95
(SM6):15311537, 1969.
[52] B.O. Hardin and V.P. Drnevich. Shear modulus and damping in soils: Design
equations and curves. Proc. ASCE: Journal of the Soil Mechanics and Foundations
Division, 98(SM7):667692, 1972.
[53] B.O. Hardin and V.P. Drnevich.
Shear modulus and damping in soils:
Measurement and parameter effects. Proc. ASCE: Journal of the Soil Mechanics and
Foundations Division, 98(SM6):603624, 1972.

154

Bibliography
[54] B.O. Hardin and F.E. Richart Jr. Elastic wave velocities in granular soils. Proc.
ASCE: Journal of the Soil Mechanics and Foundations Division, 89(SM1):3365, 1963.
[55] K. Hashiguchi. Fundamental requirements and formulations of elastoplastic constitutive equations with tangential plasticity. Int. J. Plasticity, 9(5):525549, 1993.
[56] P. Hepton. Shear wave velocity measurements during penetration testing. In Penetration Testing in the UK, pages 275283, Birmingham, UK, 1989. Thomas Telford,
London. Proceedings of the Geotechnology Conference on Penetration Testing in
the United Kingdom (PTUK).
[57] P.-Y. Hicher. Elastic properties of soils. Journal of Geotechnical Engineering, 122(8):
641648, 1996.
[58] R. Hill. The Mathematical Theory of Plasticity. Oxford University Press, London,
1950.

[59] U. Holzlohner.
Dynamische Bodenkennwerte - Meergebnisse und Zusammenhange. Bautechnik, 65(9):306312, 1988.
[60] E. Hoque and F. Tatsuoka. Kinematic elasticity of a granular material. In Proc.
GeoEng2000, Melbourne, 2000. GEE0085.PDF.
[61] E. Hoque and F. Tatsuoka. Effects of stress ratio on small-strain stiffness during
triaxial shearing. Geotechnique, 54(7):429439, 2004.
[62] G.T. Houlsby. Critical state models and small-strain stiffness. In Developments in
Theoretical Geomechanics, pages 295312. Balkema, 2000. Proc. of the Booker Memorial Symposium (Sydney).
[63] G.T. Houlsby and C.P. Wroth. The variation of shear modulus of a clay with pressure and overconsolidation ratio. Soils and Foundations, 31(3):138134, 1991.
[64] C-C. Hsu and M. Vucetic. Dynamic and cyclic behavior of soils over a wide range
of shear strains in ngi-type simple shear testing device. UCLA Reserarch Report
ENG-02-228, Civil and Environmental Engineering Department, University of California, Los Angeles, 2002.
[65] T. Hueckel and R. Nova. Some hysteresis effects of the behavior of geological
media. International Journal of Solids and Structures, 15(8):625642, 1979.
[66] T.J.R. Hughes. The Finite Element Method; Linear Static and Dynamic Finite Element
Analysis. Prentice-Hall, 1987.
[67] T.J.R. Hughes and R.L. Taylor. Unconditionally stable algorithms for quasi-static
elasto/viscoplastic finite element analysis. Computers and Structures, 8:169173,
1978.

155

Bibliography
[68] I.M. Idriss and H.B. Seed. Seismic response of horizontal soil layers. Proc. ASCE:
Journal of the Soil Mechanics and Foundation Division, 94(SM4):10031031, 1968.
[69] T. Imai and K. Tonouchi. Correlation of N-nalue with S-wave velocity and shear
modulus. In Proc. 2nd European Symposium on Penetration Testing, pages 6672,
Amsterdam, 1982.
[70] I. Ishibashi and X. Zhang. Unified dynamic shear moduli and damping ratios of
sand and clay. Soils and Foundations, 33(1):182191, 1993.
[71] K. Ishihara, F. Tatsuoka, and S. Yasuda. Undrained deformation and liquefaction
of sand under cyclic stresses. Soils and Foundations, 15(1):2944, 1975.
[72] J.N. Israelachvili. Intermolecular and surface forces. Academic Press, 1992.
[73] W.D. Iwan. On a class of models for the yielding behaviour of continuous and
composite systems. Journal of Applied Mechanics, ASME, 34(3):612617, 1967.
[74] T. Iwasaki and F. Tatsuoka. Effects of grain size and grading on dynamic shear
moduli of sands. Soils and Foundations, 17(3):1935, 1977.
[75] T. Iwasaki, F. Tatsuoka, and Y. Takagi. Hysteretic damping of sands under cyclic
loading and its relation to shear modulus. Soils and Foundations, 18(2):2540, 1978.
[76] T. Iwasaki, F. Tatsuoka, and Y. Takagi. Shear moduli of sands under cyclic torsional
shear loading. Soils and Foundations, 18(1):3956, 1978.
[77] N. Janbu. Soil compressibility as determined by oedometer and triaxial tests. In
Proc. 3rd ECSMFE, volume 1, pages 1925, Wiesbaden, 1963.
[78] R.J. Jardine, A.B. Fourie, J. Maswoswe, and J.B. Burland. Field and labaratory
measurement of soil stiffness. In Proc. 11th Int. Conf. SMFE, volume 2, pages 511
514, San Fransisco, 1985.
[79] R.J. Jardine, D.M. Potts, A.B. Fourie, and J.B. Burland. Studies of the influence of
non-linear stress-strain characteristics in soil-structure interaction. Geotechnique,
36(3):377396, 1986.
[80] R.J. Jardine, M.J. Symes, and J.B. Burland. The measurement of soil stiffness in the
triaxial apparatus. Geotechnique, 34(3):323340, 1984.
[81] P.C. Jennings. Periodic response of a general yielding structure. Proc. ASCE: Journal
of the Engineering Mechanics Division, 90(EM2):131166, 1964.
[82] P.C. Jennings. Earthquake response of a yielding structure. Proc. ASCE: Journal of
the Engineering Mechanics Division, 91(EM4):4167, 1965.
[83] B. Jeremic and S. Sture. Implicit integrations in elastoplastic geotechnics. Mechanics
of Cohesive-Frictional Materials, 2:165183, 1997.

156

Bibliography
[84] H.D. St John. Field and theoretical studies of behaviour of ground around deep excavations
in London Clay. PhD thesis, University of Cambridge, 1975.
[85] K.L. Johnson. Contact Mechanics. Cambridge University Press, 1985.
[86] V. Jovicic and M.R. Coop. Stiffness of coarse-grained soils at small strains.
Geotechnique, 47(3):545561, 1997.
[87] F.E. Richart Jr., R.D. Woods, and J.R. Hall Jr. Vibrations of Soils and Foundations.
Civil Engineering and Engineering Mechanics Series. Prentice-Hall, Inc., Englewood Cliffs, N.J., 1970.
[88] T.T. Kallioglou, A. Papadopoulou, and K. Pitilakis. Shear modulus and damping
of natural sands. In H. Di Benedetto, T. Doanh, H. Geoffroy, and C. Sauzeat, editors, Deformation Characteristics of Geomaterials, pages 401407. A.A. Balkema, 2003.
Proc. IS-Lyon03.
[89] Stokoe K.H., M.B. Darendeli, R.B. Gilbert1, F.-Y. Menq, and W.K. Choi. Development of a new family of normalized modulus reduction and material damping
curves. In International Workshop on Uncertainties in Nonlinear Soil Properties and
their Impact on Modeling Dynamic Soil Response, UC Berkeley, CA, 2004.
[90] T.C. Kim and M. Novak. Dynamic properties of some cohesive soils of ontario.
Canadian Geotech. Journal, 18(3):371389, 1981.
[91] W.T. Koiter. General theorems for elastic-plastic solids. In I.N. Sneddon and R. Hill,
editors, Progress in Solid Mechanics, volume 1, pages 165221, North-Holland, Amsterdam, 1960.
[92] T. Kokusho, Y. Yoshida, and E. Yasuyuki. Dynamic properties of soft clay for wide
strain range. Soils and Foundations, 22(4):118, 1982.
[93] D. Kolymbas. An outline of hypoplasticity. Archive of Applied Mechanics, 61:143
151, 1991.
[94] R.L. Kondner and J.S. Zelasko. A hyperbolic stress-strain formulation for sands.
In 2nd Pan. Am. Conf. Soil Mech. Found. Eng., volume 1, pages 289394, Brazil, 1963.
[95] W. Krajewski and O. Reul. Erfahrungen zur numerischen Berechnung von

Baugruben und Grundungen

im Rhein-Main-Gebiet. In T. Schanz, editor, Workshop AK 1.6 Numerik in der Geotechnik, number 11 in Schriftenreihe Geotechnik,
pages 103117. Bauhaus-Universitat Weimar, 2003.
[96] R. Kuwano and R.J. Jardine. On the application of cross-anisotropic elasticity to
granular materials at very small strains. Geotechnique, 52(10):727749, 2002.
[97] P.V. Lade. Modelling the strengths of engineering materials in three dimensions.
Mech. Cohes. Frict. Mat., 2:339356, 1997.

157

Bibliography
[98] P.V. Lade and V. Abelev. Characterization of cross-anisotropic soil deposits from
isotropic compression tests. Soils and Foundations, 45(5):89102, 2005.
[99] P.V. Lade and J.M. Duncan. Elasto-plastic stress-strain theory for cohesionless soil.
Proc. ASCE: Journal of the Geotechnical Engineering Division, 101(GT10):10371053,
1975.
[100] R. Larsson and M. Mulabdic. Piezocone tests in clay. Technical Report 42, Swedish

Geotechnical Institute, Linkoping,

1991.
[101] P.I. Lewin and J.B. Burland. Stress-probe experiments on saturated normally consolidated clay. Geotechnique, 20(1):3856, 1970.
[102] X.S. Li. A sand model with state-dependent dilatancy. Geotechnique, 52(3):173186,
2002.
[103] X.S. Li and Y.F. Dafalias. Dilatancy for cohesionless soils. Geotechnique, 50(4):449
460, 2000.
[104] M.L. Lings and P.D. Greening. A novel bender extender element for soil testing.
Geotechnique, 51(8):713717, 2001.
[105] T. Lansivaara and S. Nordal. Loading history model for small strain behavior of
soils. In 15th Int. Conf. SMGE, Istanbul, 2001. Rotterdam: Balkema.

[106] W. Lode. Versuche uber

den Einflu der mittleren Hauptspannung auf das Flieen
der Metalle Eisen Kupfer und Nickel. Physik, 36:913939, 1926.
[107] T.N. Lohani, G. Imai, K. Tani, and S. Shibuya. Gmax of fine grained soils at wide
void ratio range focusing on time-dependent behavior. Soils and Foundations, 41
(5):87102, 2001.
[108] T. Lunne, P.K. Robertson, and J.J.M. Powell. Cone Penetration Testing in Geotechnical
Practice. E & FN Spon, London, 1997.
[109] R.J. Mair. Developments in geotechnical engineering research: application to tunnels and deep excavations. Proceedings of Institution of Civil Engineers, Civil Engineering, pages 2741, 1993. Unwin Memorial Lecture 1992.
[110] W.F. Marcuson and H.E. Wahls. Time effects on dynamic shear modulus of clays.
Proc. ASCE: Journal of the Soil Mechanics and Foundation Division, 98(SM12):1359
1373, 1972.
[111] G. Masing. Eigenspannungen und Verfestigung beim Messing. In Proc. 2nd Int.
Congr. Appl. Mech., Zurich, 1926.
[112] L. Matesic and M. Vucetic. Strain-rate effect on soil secant shear modulus at small
cyclic strains. Journal of Geotechnical and Geoenvironmental Engineering, 129(6):536
549, 2003.

158

Bibliography
[113] H. Matsuoka. Stress-strain relationships of sands based on the mobilized plane.
Soils and Foundations, 14(2):4761, 1974.
[114] H. Matsuoka and T. Nakai. A new failue criterion for soils in three dimensional
stresses. In IUTAM Conference on Deformation and Failure of Granular Materials,
pages 253263, Delft, 1982.
[115] P.W. Mayne and F.H. Kulhawy. K0 -OCR relationships in soil. Journal of Geotechnical
Engineering, 108(6):851872, 1982.
[116] P.W. Mayne and G.J. Rix. Gmax qc relationships for clays. Geotechnical Testing
Journal, 16(1):5460, 1993.
[117] G. Mesri, T.W. Feng, and J.M. Benak. Postdensification penetration resistance of
clean sands. Journal of Geotechnical Engineering, ASCE, 116(7):10951115, 1990.
[118] R.D. Mindlin. Compliance of elastic bodies in contact. Journal of Applied Mechanics,
ASME, 16:259268, 1949.
[119] R.D. Mindlin and H. Deresiewicz. Elastic spheres in contact under varying oblique
forces. Journal of Applied Mechanics, ASME, 20(3):327344, 1953.
[120] J.K. Mitchell. Fundamentals of Soil Behavior. John Wiley & Sons, 2 edition, 1993.

[121] S.C. Moller.

Tunnel induced settlements and structural forces in linings. PhD thesis,
Universitat Stuttgart, 2006.
[122] G. Modoni, A. Flora, L.Q. Anh Dan, C. Mancuso, J. Koseki, K. Balakrishnaier, and
F. Tatsuoka. A simple experimental procedure for the complete characterization of
small strain stiffness of gravels. In M. Jamiolkowski, R. Lancellotta, and D.C.F. Lo
Presti, editors, Pre-Failure Deformation Characteristics of Geomaterials, pages 123130.
A.A. Balkema, 1999. Proc. IS-Torino99.
[123] O. Mohr. Welche Umstande bedingen die Elastizitatsgrenze und den Bruch eines
Materials? VDI-Zeitschrift, 44:1524, 1900.
[124] J.W. Morse and F.T. Mackenzie. Geochemistry of sedimentary carbonates, volume 48
of Developments in sedimentology. New York, Elsevier, 1982.
[125] Z. Mroz. On the description of anisotropic workhardening. J. Mech. Phys. Solids,
15:163175, 1967.
[126] Z. Mroz, V.A. Norris, and O.C. Zienkiewicz. An anisotropic hardening model for
soils and its application to cyclic loading. Int. J. Numer. Anal. Meth. Geomech., 2:
203221, 1978.

Spannungs-Verformungsverhalten granularer Materialien am Beispiel

[127] H. Mullersch
on.
von Berliner Sand. PhD thesis, Universitat Stuttgart, 2000.

159

Bibliography
[128] T. Nakai and H. Matsuoka. Deformations of soils in three-dimensional stresses. In
IUTAM Conference on Deformation and Failure of Granular Materials, pages 275285,
Delft, 1982.
[129] A. Niemunis and I. Herle. Hypoplastic model for cohesionless soils with elastic
strain range. Mechanics of Cohesive-Frictional Materials, 2(4):279299, 1997.
[130] R.L. Nigbor and T. Imai. The suspension p-s velocity logging method. In Proc. 13th
Int. Conf. SMFE, volume 1, pages 5761, New Delhi, India, 1994. TC10.

[131] J. Ohde. Grundbaumechanik, volume 3. Hutte,

1951.
[132] Y. Ohsaki and R. Iwasaki. On dynamic shear moduli and Poissons ratio of soil
deposits. Soils and Foundations, 13(4):6173, 1973.
[133] A.G. Papadimitriou and G.D. Bouckovalas. Plasticity model for sand under small
and large cyclic strains: a multiaxial formulation. Soil Dynamics and Earthquake
Engineering, 22:191204, 2002.
[134] A.G. Papadimitriou, G.D. Bouckovalas, and Y. F. Dafalias. Plasticity model for
sand under small and large cyclic strains. Journal of Geotechnical and Geoenvironmental Engineering, 127(11):973983, 2001.
[135] R. B. Peck. Deep excavations and tunneling in soft ground. In 7th Int. Conf. SMFE.
Sociedad Mexican de Mecanica de Suelos, A. C., 1969.
[136] D.S. Pennington, D.F.T. Nash, and M.L. Lings. Anisotropy of G0 shear stiffness in
Gault clay. Geotechnique, 47(3):391398, 1997.
[137] J.M. Pestana and A.J. Whittle. Formulation of a unified constitutive model for clays
and sands. Int. J. Numer. Anal. Meth. Geomech., 23:12151243, 1999.
[138] D.M. Potts and L. Zdravkovic. Finite element analysis in geotechnical engineering:
theory. Thomas Telford, 1999.
[139] D.C.F. Lo Presti and M. Jamiolkowski. Discussion: Estimate of elastic shear modulus in holocene soil deposits. Soils and Foundations, 38(263-265), 1998.
[140] D.C.F. Lo Presti, M. Jamiolkowski, O. Pallara, and A. Cavallaro. Rate and creep
effect on the stiffness of soils. In T.C. Sheahan and V.N. Kaliakin, editors, Measuring and Modeling Time Dependent Soil Behavior, number 61 in Geotechnical Special
Publication, pages 166180. ASCE, 1996.
[141] D.C.F. Lo Presti, O. Pallara, R. Lancellotta, M. Armandi, and R. Maniscalco. Monotonic and cyclic loading behaviour of two sands at small strains. Geotechnical Testing Journal, 16(4):409424, December 1993.
[142] J.-H. Prevost. Mathematical modelling of monotonic and cyclic undrained clay
behavior. Int. J. Numer. Anal. Meth. Geomech., 1(2):195216, 1977.

160

Bibliography
[143] R. Pyke. Nonlinear soil models for irregular cyclic loadings. Proc. ASCE: Journal of
the Geotechnical Engineering Division, 105(GT6):715726, 1979.
[144] T. Ramamurthy and P.C. Rawat. Shear strength of sand under general stress system. In Proc. 8th Int. Conf. SMFE, volume 1, pages 339342, Moscow, 1973.
[145] W. Ramberg and W. R. Osgood. Description of stress-strain curve by three parameters. Technical Note 902, National Advisory Committee for Aeronautics, Washington, DC, 1943.
[146] K.I. Rammah, D.V. Val, and A.M. Puzrin. Effects of ageing on small-strain stiffness
of overconsolidated clays. Geotechnique, 54(5):319322, 2004.
[147] S. Rampello, G.M.B. Viggiani, and A. Amorosi. Small-strain stiffness of reconstituted clay compressed along constant triaxial effective stress ratio paths.
Geotechnique, 47(3):475489, 1997.
[148] D. Richardson. Investigations of threshold effects in soil deformation. PhD thesis, City
University, London, 1988.
[149] F.E. Jr. Richart. Some effects of dynamic soil properties on soil structure interaction.
Proc. ASCE: Journal of the Geotechnical Engineering Division, 101(GT12):11931240,
1975.
[150] P.K. Robertson, R.G. Campanella, D. Gillespie, and A. Rice. Seismic cpt to measure in-situ shear wave velocity. Journal of Geotechnical Engineering, 112(8):791803,
1986.
[151] K.H. Roscoe and A.N. Schofield. Mechanical behaviour of an idealised wet clay.
In Proc. 2nd European Conf. SMFE, volume 1, pages 4754, Wiesbaden, 1963.
[152] E. Rosenblueth and I. Herrera. On a kind of hysteretic damping. Proc. ASCE:
Journal of the Engineering Mechanics Division, 90(EM4):3747, 1964.
[153] P.W. Rowe. The stress-dilatancy relation for static equilibrium of an assembly of
particles in contact. In Proc. of the Royal Society of London. Series A, Mathematical and
Physical Sciences, volume 269, pages 500527, 1962.
[154] K. Runesson and A. Samuelsson. Numeta 85 Numerical Methods in Engineering,
Theory and Applications, chapter Aspects on numerical techniques in small deformation plasticity, pages 337347. A.A. Balkema, 1985.
[155] J.A. Santos and A.G. Correia. Reference threshold shear strain of soil. Its application to obtain a unique strain-dependent shear modulus curve for soil. In 15th Int.
Conf. SMGE, volume 1, pages 267270, Istanbul, 2001. A.A. Balkema.
[156] S.A. Savidis. Empfehlungen des Arbeitskreises Baugrunddynamik. Grundbauinstitut der Technischen Universitat Berlin, 2002.

161

Bibliography
[157] T. Schanz. Zur Modellierung des mechanischen Verhaltens von Reibungs Geotechnik 45, Universitat Stuttgart, 1998.
materialien. Mitt. Inst. fur
[158] T. Schanz, P.A. Vermeer, and P.G. Bonnier. Beyond 2000 in Computational Geotechnics,
chapter Formulation and verification of the Hardening-Soil Model, pages 281290.
Balkema, Rotterdam, 1999.
[159] J.H. Schmertmann. The mechanical aging of soils. Journal of Geotechnical Engineering, 117(9):12881330, 1991.
[160] B.F. Schmidt. Settlements and ground movements associated with tunnelling in soils.
PhD thesis, University of Illinois, Urbana, 1969.
[161] H.F. Schweiger. Results from numerical benchmark exercises in geotechnics. In
P. Mestat, editor, Numge 2002. 5th European Conference Numerical Methods in Geotechnical Engineering, volume 1, pages 305314, Paris, 2002. LCPC.
[162] H.B. Seed and I.M. Idriss. Soil moduli and damping factors for dynamic response
analysis. Report 70-10, EERC, Berkeley, CA, 1970.
[163] T. Shibata and D. Karube. Influence of the variation of the intermediate principal
stress on mechanical properties of normally consolidated clay. In Proc. 6th Int. Conf.
SMFE, volume 1, pages 359363, Paris, 1985.
[164] S. Shibuya, T. Mitachi, S. Yamashita, and H. Tanaka. Effect of sample disturbance
on Gmax of soils - a case study. In S. Shibuya T. Mitachi and S. Miura, editors,
Prefailure Deformations of Geomaterials, volume 2, pages 7782. Balkema, Rotterdam,
1995.
[165] D.J. Shirley and L.D. Hampton. Shear-wave measurements in laboratory sediments. The Journal of the Acoustical Society of America, 63(2):607613, 1978.
[166] J.C. Simo and T.J.R. Hughes. Computational Inelasticity, volume 7 of Interdisciplinary
Applied Mathematics. Springer, 1998.
[167] J.C. Simo and R.L. Taylor. Consistent tangent operators for rate-independent
elastoplasticity. Computer Methods in Applied Mechanics and Engineering, 48:101
118, 1985.
[168] B. Simpson. Retaining structures: displacement and design, the 32nd Rankine
lecture. Geotechnique, 42(4):541576, 1992.
[169] O.K. Soreide. Mixed hardening models for frictional soils. PhD thesis, Norwegian
University of Science and Technology (NTNU), Trondheim, 1990.
[170] S.A. Stallebrass. Modelling the effect of recent stress history on the deformation of overconsolidated soils. PhD thesis, City University, London, 1990.

162

Bibliography
[171] S.E. Stallebrass and R.N. Taylor. The development and evaluation of a constitutive model for the prediction of ground movements in overconsolidated clay.
Geotechnique, 47(2):235253, 1997.
[172] K.H. Stokoe, M.B. Darendeli, R.D. Andrus, and L.T. Brown. Dynamic soil properties: laboratory, field and correlation studies. In Proc. 2nd Int. Conf. on Earthquake
Geotech. Eng., volume 3, pages 811845. A.A. Balkema, 1999.
[173] K.H. Stokoe and J.C. Santamarina. Seismic-wave-based testing in geotechnical engineering. In GeoEng 2000: An International Conference on Geotechnical and Geological
Engineering, volume 1, pages 14901536, Melbourne, Austtralia, 2000. Technomic
Publishing Company.
[174] K.H. Stokoe, S.G. Wright, A.B. James, and M.R. Jose. Geophysical characterization of
sites, chapter Characterization of geotechnical sites by SASW method, pages 1525.
Balkema, A.A., Rotterdam, 1994.
[175] K.H.I. Stokoe and R.D. Woods. In situ shear wave velocity by cross-hole method.
Proc. ASCE: Journal of the Soil Mechanics and Foundations Division, 98(SM5):443460,
1972.
[176] H.B. Sutherland and M.S. Mesdary. The influence of the intermediate principal
stress on the strength of sand. In Proc. 7th Int. Conf. SMFE, volume 1, pages 391
399, Mexico City, 1969.
[177] F. Tatsuoka. Impacts on geotechnical engineering of several recent findings from
laboratory stress-strain tests on geomaterials. Department of Civil Engineering
Mechanics, Columbia University, 2000. The 2000 Burmister Lecture.
[178] F. Tatsuoka and K. Ishihara. Yielding of sand in triaxial compression. Soils and
Foundations, 14(2):6376, 1974.
[179] K. Terzaghi, R.B. Peck, and G. Mesri. Soil Mechanics in Engineering Practice. WileyInterscience; 3 edition (January 1, 1996), 3 edition, 1996.
[180] S. Toki, S. Shibuya, and S. Yamashita. Standardization of laboratory test methods to determine the cyclic deformation properties of geomaterials in Japan. In
S. Shibuya T. Mitachi and S. Miura, editors, Pre-failure Deformations of Geomaterials,
volume 2, pages 741784. Balkema, Rotterdam, 1995.
[181] C. Truesdell and W. Noll. The Non-Linear Field Theories of Mechanics, volume III of
Encyclopedia of Physics. Berlin, Heidelberg: Springer, 1965.
[182] T.W. and R.V. Whitman. Soil Mechanics. John Wiley & Sons, New York, NY, 1969.
[183] P.A. Vermeer. A double hardening model for sand. Geotechnique, 28(4):413433,
1978.

163

Bibliography
[184] P.A. Vermeer. A modified initial strain method for plasticity problems. In Proc. 3th
Int. Conf. Numer. Meth. Geomech., pages 377387. Balkema, Rotterdam, 1979.
[185] P.A. Vermeer. Formulation and analysis of sand deformation problems. PhD thesis,
Delft University of Science and Technology, 1980.
[186] P.A. Vermeer, T. Benz, and R. Schwab. Modern Trends in Geomechanics, chapter Two
Elastoplastic Models for Small and Large Strains and Their Use in Engineering
Practise, pages 175190. Springer, 2006.
[187] G. Viggiani and J.H. Atkinson. Stiffness of fine grained soils at very small strains.
Geotechnique, 45(2):249265, 1995.
[188] M. Vucetic and R. Dobry. Effect of soil plasticity on cyclic response. Journal of
Geotechnical Engineering, ASCE, 117(1):89107, 1991.
[189] M. Vucetic and K. Tabata. Influence of soil type on the effect of strain rate on
small-strain cyclic shear modulus. Soils and Foundations, 43(5):161173, 2003.
[190] M. Vucetic, K. Tabata, and L. Matesic. Effect of average straining rate on shear
modulus at small cyclic strains. In H. Di Benedetto, T. Doanh, H. Geoffroy, and
C. Sauzeat, editors, Deformation Characteristics of Geomaterials, pages 321328. A.A.
Balkema, 2003. Proc IS-Lyon03.
[191] M. Wehnert. Ein Beitrag zur drainierten und undrainierten Analyse in der Geotechnik.
PhD thesis, Universitat Stuttgart, 2006.
[192] R.V. Whitman. The behaviour of soils under transient loadings. In Proc. 4th Int.
Conf. SMFE, volume 1, pages 207210, London, 1957.
[193] T. Wichtmann and T. Triantafyllidis. Influence of a cyclic and dynamic loading
history on dynamic properties of dry sand, Part i: cyclic and dynamic torsional
prestraining. Soil Dynamics and Earthquake Engineering, 24(2):127147, 2004.
[194] R.D. Woods. Measurement of dynamic soil properties. In Proc. Earthquake Engineering and Soil Dynamics, volume 1, pages 9178, Pasadena, CA, 1978. ASCE, New
York.
[195] C.P. Wroth. In-situ measurement of initial stresses and deformation characteristics.
In Proc. Geot. Engng Div. Speciality Conf. on In-situ Measurement of Soil Properties,
pages 181230, Rayleigh, NC, 1975. ASCE.
[196] Y. Yamada and K. Ishihara. Anisotropic deformation characteristic of sand under
three dimensional stress conditions. Soils and Foundations, 19(1):7994, 1979.
[197] S. Yimsiri and K. Soga. Effect of surface roughness on small-strain modulus: Micromechanics view. In M. Jamiolkowski, R. Lancellotta, and D. Lo Presti, editors,
Pre-failure Deformation Characteristics of Geomaterials, volume 1, pages 597602. A.A.
Balkema, Rotterdam, 1999.

164

Bibliography
[198] R.N. Yong and R.D. Japp. A flow law for clays in dynamic compression. In
Proc. Int. Symp. Wave Propagation and Dynam. Prop. of Earth Materials, Albuquerque,
N.M., 1967. University of New Mexico Press.
[199] M.S. Youssef, A.H. El Ramli, and M. El Demery. Relationships between shear
strength, consolidation, liquid limit, and plastic limit for remoulded clays. In Proc.
6th Int. Conf. SMFE, volume 1, pages 126129, Montreal, 1965.
[200] P. Yu and Jr. F.E. Richart. Stress ratio effects on shear modulus of dry sands. Journal
of Geotechnical Engineering, 110(3):331345, 1984.
[201] C. Zambelli. Modellazione constitutiva del comportamento meccanico in campo dinamico
di material granulari. Tesi di laurea, Politecnico di Milano, 2002.
[202] L. Zdravkovic and R.J. Jardine. Some anisotropic stiffness characteristics of a silt
under general stress conditions. Geotechnique, 47(3):407437, 1997.

165

Appendix A
Small-Strain Overly Fortran Code
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42

!+
! 23456789012345678901234567890123456789012345678901234567890123456789|
!+
! Small S t r a i n O v e r l a y
|
!+
! P u r p o s e : Add s m a l l s t r a i n s t i f f n e s s t o e l a s t o p l a s t i c m o d e l s
|
!
|
!
I /O Type
|
! Eps0
I R ( 6 ) : t o t a l train at s t a r t of increment
|
! dEps
I R ( 6 ) : s t r a i n increment
|
! UsrSta0 I R ( ) : s t a t e v a r i a b l e s a t s t a r t o f i n c ; u s e s # 2 t o # 1 0 |
! U s r S t a 0 O R ( ) : s t a t e v a r i a b l e s a t end o f i n c ; u s e s # 2 t o # 1 0
|
! Sig0
I R ( 6 ) : s t r e s s at s t a r t of increment
|
! G
I R
: e l a s t i c modulus o f b a c k g r o u n d m o d e l ( l a r g e r s t r . ) |
! GFac
I R
: r a t i o G0 /G w h e r e G0 i s t h e s m a l l s t r a i n modulus |
! Gam07
I R
: 70% threshold shear strain
|
! GAct
O R
: A c t u a l e l a s t i c modulus t o b e u s e d f o r c a l c u l a t i o n |
!
|
! The s t a t e v a r i a b l e s a r e o r g a n i z e d a s f o l l o w s
|
! UsrSta ( 1 ) : not used
|
! UsrSta ( 2 ) : not used
|
! U s r S t a ( 3 ) : S t r a i n h i s t o r y component 1
|
! U s r S t a ( 4 ) : S t r a i n h i s t o r y component 2
|
! U s r S t a ( 5 ) : S t r a i n h i s t o r y component 3
|
! U s r S t a ( 6 ) : S t r a i n h i s t o r y component 4
|
! U s r S t a ( 7 ) : S t r a i n h i s t o r y component 5
|
! U s r S t a ( 8 ) : S t r a i n h i s t o r y component 6
|
! U s r S t a ( 9 ) : A c t u a l s t i f f n e s s GAct f o r o u t p u t e t c . / r e d u n d a n t
|
! U s r S t a ( 1 0 ) : Monoton ic s t i f f n e s s ; r e l o a d i n g f o r GAct>G UsrSt ( 1 0 )
|
!
|
! Needs : S u b r o u t i n e P r n S i g ( dEpsD , xN1e , xN2e , xN3e , dEpsD1 , dEpsD2 , dEpsD3 ) |
!
t h a t f i n d s t h e E i g e n s y s t e m ( xN1e , xN2e , xN3e ) and E i g e n v a l u e s |
!
( dEpsD1 , dEpsD2 , dEpsD3 ) o f t h e s t r a i n i n c r e m e n t dEpsD
|
!+
! Author : Thomas Benz < t h o m a s . benz@wewinumge . de >
|
!+
!
SUBROUTINE S m a l l S t r a i n ( Eps0 , dEps , UsrSta , UsrSta0 , Sig0 ,

G, GFac , Gam07 , GAct )

!
IMPLICIT DOUBLE PRECISION ( AH, OZ)
!

167

Appendix A Small-Strain Overly Fortran Code

DIMENSION Eps0 ( 6 ) , dEps ( 6 ) , UsrSta ( ) , UsrSta0 ( ) , S i g 0 ( 6 ) ,
2
Trafo ( 3 , 3 ) , Hist1 ( 3 , 3 ) , Hist2 ( 3 , 3 ) ,
3
Chi ( 3 , 3 ) , GamHis ( 2 ) , GBct ( 2 ) ,
4
xT ( 3 ) , xN1e ( 3 ) , xN2e ( 3 ) , xN3e ( 3 ) ,
5
dEpsD ( 6 ) , Eps ( 6 ) , DEl ( 6 , 6 ) , S i g E l ( 6 )
! Initialization
Eps = 0 .
Chi = 0 .
I f ( UsrSta0 ( 1 0 ) . EQ . 0 . ) UsrSta0 ( 1 0 ) = 1 0 . ! G F a c t o r m a x : = 1 0
! D e t e r m i n e new s t i f f n e s s i n t h e p r i n c i p a l s y s t e m o f dEpsD
dEpsVol =( dEps ( 1 ) + dEps ( 2 ) + dEps ( 3 ) )
dEpsD ( 1 ) = dEps ( 1 ) dEpsVol/3d0
dEpsD ( 2 ) = dEps ( 2 ) dEpsVol/3d0
dEpsD ( 3 ) = dEps ( 3 ) dEpsVol/3d0
dEpsD ( 4 ) = dEps ( 4 )
dEpsD ( 5 ) = dEps ( 5 )
dEpsD ( 6 ) = dEps ( 6 )
CALL PrnSig ( dEpsD , xN1e , xN2e , xN3e , dEpsD1 , dEpsD2 , dEpsD3 )
! Build f i e l d s
dEpsD ( 1 ) = dEpsD1
dEpsD ( 2 ) = dEpsD2
dEpsD ( 3 ) = dEpsD3
dEpsD ( 4 ) = 0 d0
dEpsD ( 5 ) = 0 d0
dEpsD ( 6 ) = 0 d0
IF ( UsrSta0 ( 3 ) . LE . 1 ) THEN ! I n i t H i s t o r y
H i s t 1 ( 1 , 1 ) = 1 0 0 d0
Hist1 ( 2 , 1 ) =
0d0
Hist1 ( 3 , 1 ) =
0d0
Hist1 ( 1 , 2 ) =
0d0
H i s t 1 ( 2 , 2 ) = 1 0 0 d0
Hist1 ( 3 , 2 ) =
0d0
Hist1 ( 1 , 3 ) =
0d0
Hist1 ( 2 , 3 ) =
0d0
H i s t 1 ( 3 , 3 ) = 1 0 0 d0
ELSE
H i s t 1 ( 1 , 1 ) = UsrSta0 ( 3 )
H i s t 1 ( 2 , 1 ) = UsrSta0 ( 6 ) / 2 d0
H i s t 1 ( 3 , 1 ) = UsrSta0 ( 8 ) / 2 d0
H i s t 1 ( 1 , 2 ) = UsrSta0 ( 6 ) / 2 d0
H i s t 1 ( 2 , 2 ) = UsrSta0 ( 4 )
H i s t 1 ( 3 , 2 ) = UsrSta0 ( 7 ) / 2 d0
H i s t 1 ( 1 , 3 ) = UsrSta0 ( 8 ) / 2 d0
H i s t 1 ( 2 , 3 ) = UsrSta0 ( 7 ) / 2 d0
H i s t 1 ( 3 , 3 ) = UsrSta0 ( 5 )
END IF
T r a f o ( 1 , 1 ) = xN1e ( 1 )
T r a f o ( 2 , 1 ) = xN1e ( 2 )
T r a f o ( 3 , 1 ) = xN1e ( 3 )
T r a f o ( 1 , 2 ) = xN2e ( 1 )
T r a f o ( 2 , 2 ) = xN2e ( 2 )
T r a f o ( 3 , 2 ) = xN2e ( 3 )

43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94

168

95
96
97

98
99
100
101
102
103
104
105
106
107
108
109
110
111

112
113
114
115

116
117
118
119
120
121
122
123
124

125
126
127
128
129
130
131
132
133

134
135
136
137
138
139
140

2
3
4

141
142
143
144
145
146

T r a f o ( 1 , 3 ) = xN3e ( 1 )
T r a f o ( 2 , 3 ) = xN3e ( 2 )
T r a f o ( 3 , 3 ) = xN3e ( 3 )
T r a n s f o r m s t a r i n h i s t o r y i n p r i n c i p a l s y s t e m o f dEpsD
DO i = 1 , 3
DO j = 1 , 3
X1 = 0 d0
DO k = 1 , 3
X2 = 0 d0
DO l = 1 , 3
X2 = X2 + H i s t 1 ( k , l ) T r a f o ( l , j )
END DO
X1 = X1 + X2 T r a f o ( k , i )
END DO
H i s t 2 ( i , j ) = X1
END DO
END DO
Elongation of uncorrected history
DO i = 1 , 3
Chi ( 1 , i ) = H i s t 2 ( i , i ) 1 0 0 d0
END DO
Build transformation matrix
nCountXT = 0
DO i = 1 , 3
xT ( i ) = 1 d0
IF ( c h i ( 1 , i ) dEpsD ( i ) . LT . 0 d0 ) THEN
xT ( i ) = 1 0 d0/DSQRT( c h i ( 1 , i ) + 1 0 0 d0 )
nCountXT = nCountXT + 1
END IF
END DO
C a l c u l a t e new h i s t o r y
DO i = 1 , 3
DO j = 1 , 3
H i s t 2 ( i , j ) = H i s t 2 ( i , j ) xT ( i ) xT ( j )
END DO
Chi ( 2 , i )
= H i s t 2 ( i , i ) 1 0 0 d0
H i s t 2 ( i , i ) = H i s t 2 ( i , i ) + dEpsD ( i )
Chi ( 3 , i )
= H i s t 2 ( i , i ) 1 0 0 d0
END DO
Calculate history strain ( iso )
dEps2 = ( dEpsD ( 1 ) 2 + dEpsD ( 2 ) 2 + dEpsD ( 3 ) 2 )
DO i = 1 , 2
GamHis ( i ) = DSQRT( 3 d0
( ( dEpsD ( 1 ) Chi ( i + 1 , 1 ) ) 2 +
( dEpsD ( 2 ) Chi ( i + 1 , 2 ) ) 2 +
( dEpsD ( 3 ) Chi ( i + 1 , 3 ) ) 2 )/ dEps2 )
END DO
From s t r a i n h i s t o r y t o s t i f f n e s s
Fac0 = GamHis(2) GamHis ( 1 )
Fac1 = Gam07/ 0 . 3 8 5 d0 ! u s e 0 . 4 2 9 f o r HD
DO i = 1 , 2 ! i =1:= l o a d i n g ; i = 2 : = r e l o a d i n g
IF ( i . EQ . 2 ) Fac1 = 2 d0 Fac1

169

Appendix A Small-Strain Overly Fortran Code

147
148
149
150

151
152
153
154

155
156
157
158

2
3

159
160
161
162
163
164
165
166
167

168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187

!
!

188
189
190

191
192
193

194
195
196
197
198

IF ( ( UsrSta0 ( 1 0 ) . NE . 1 ) . OR . ( i . EQ . 2 ) ) THEN
GamOff = ( DSQRT( GFac)1d0 ) Fac1
IF ( Fac0 . LE . 0 ) THEN
GBct ( i ) = GGFac ( 1 / ( 1 +GamHis ( 1 ) / Fac1 )
GamHis ( 1 ) / Fac1 /(1+GamHis ( 1 ) / Fac1 ) 2 )
ELSE
IF ( GamHis ( 1 ) . LT . GamOff .AND. GamHis ( 2 ) . LT . GamOff ) THEN
GBct ( i ) = GGFac ( GamHis ( 2 ) / ( 1 d0+GamHis ( 2 ) / Fac1 )
GamHis ( 1 ) / ( 1 d0+GamHis ( 1 ) / Fac1 ) ) / Fac0
ELSE
IF ( GamHis ( 1 ) . LT . GamOff .AND. GamHis ( 2 ) . GT . GamOff ) THEN
GBct ( i ) = ( GGFac ( GamOff/(1 d0+GamOff/Fac1 )
GamHis ( 1 ) / ( 1 d0+GamHis ( 1 ) / Fac1 ) ) +
G ( GamHis(2) GamOff ) ) / Fac0
ELSE
GBct ( i ) = G
END IF
END IF
END IF
END IF
END DO
Update S t a t e V a r w i t h new h i s t o r y ( B a c k T r a f o )
DO i = 1 , 3
DO j = 1 , 3
X1 = 0 d0
DO k = 1 , 3
X2 = 0 d0
DO l = 1 , 3
X2 = X2 + H i s t 2 ( k , l ) T r a f o ( j , l )
END DO
X1 = X1 + X2 T r a f o ( i , k )
END DO
H i s t 1 ( i , j ) = X1
END DO
END DO
UsrSta ( 3 ) = H i s t 1 ( 1 , 1 )
UsrSta ( 4 ) = H i s t 1 ( 2 , 2 )
UsrSta ( 5 ) = H i s t 1 ( 3 , 3 )
UsrSta ( 6 ) = H i s t 1 ( 1 , 2 ) + H i s t 1 ( 2 , 1 )
UsrSta ( 7 ) = H i s t 1 ( 2 , 3 ) + H i s t 1 ( 3 , 2 )
UsrSta ( 8 ) = H i s t 1 ( 1 , 3 ) + H i s t 1 ( 3 , 1 )
L o a d i n g / r e l o a d i n g i n c o m b i n a t i o n w i t h an e l a s t o p l a s t i c m o d e l :
Update o f U s r S t a ( 1 0 )
IF ( UsrSta0 ( 1 0 ) .NE . 1 . ) UsrSta ( 1 0 ) =MIN( GBct ( 1 ) /G, UsrSta0 ( 1 0 ) )
Copy s t i f f n e s s ( a l w a y s r e l o a d i n g i n c o m b i n a t i o n w i t h p l a s t i c i t y )
GAct = GBct ( 2 )
END IF
Plateau avoids numerical problems in i s o t r o p i c loading
I f ( GamHis ( 2 ) . Lt . 1 d 6 ) GAct = GGFac

!
!
! E i t h e r RETURN h e r e o r c h e c k s t r e s s work . . .

170

199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238

!
Do i =1 ,6
Eps ( i ) = Eps0 ( i ) + dEps ( i )
End Do
EpsV0 =( Eps0 ( 1 ) + Eps0 ( 2 ) + Eps0 ( 3 ) )
EpsD0 = ( ( Eps0 (1) Eps0 ( 2 ) ) 2
2
+( Eps0 (2) Eps0 ( 3 ) ) 2
3
+( Eps0 (3) Eps0 ( 1 ) ) 2 ) / 6 d0
4
+( Eps0 ( 4 ) / 2 d0 ) 2 + ( Eps0 ( 5 ) / 2 d0 ) 2 + ( Eps0 ( 6 ) / 2 d0 ) 2
IF ( DABS( EpsV0 ) . Lt . 1 d 1 2 .OR.
2
DABS( EpsD0 ) . Lt . 1 d12
) THEN
! No Check o r a l t e r n a t i v e c o d e
ELSE
SigV0 =( S i g 0 ( 1 ) + S i g 0 ( 2 ) + S i g 0 ( 3 ) ) / 3 d0
SigD0 = ( ( S i g 0 (1) S i g 0 ( 2 ) ) 2
2
+( S i g 0 (2) S i g 0 ( 3 ) ) 2
3
+( S i g 0 (3) S i g 0 ( 1 ) ) 2 ) / 6 d0
4
+( S i g 0 ( 4 ) ) 2 + ( S i g 0 ( 5 ) ) 2 + ( S i g 0 ( 6 ) ) 2
xxK=SigV0/EpsV0
xxG=DSQRT( SigD0/EpsD0 )/2 d0
xxF1=xxK + 4 . d0xxG / 3 . d0
xxF2=xxK2.d0xxG / 3 . d0
DEl=0
DO i =1 ,3
DO j =1 ,3
DEl ( i , j ) = xxF2
END DO
DEl ( i , i ) = xxF1
DEl ( i +3 , i + 3 ) = xxG
END DO
xWElasticD2 = DOT PRODUCT(MATMUL( DEl , Eps ) , Eps )
2
DOT PRODUCT(MATMUL( DEl , Eps0 ) , Eps0 )
END IF
IF ( xWElasticD2 . LT . 0 d0 ) THEN ! U n l o a d i n g
GAct=MAX( GAct , xxG )
! Correction
END IF
! For output e t c .
UsrSta ( 9 ) = GAct
RETURN
END SUBROUTINE S m a l l S t r a i n

171

Appendix B
Return mapping in Fortran code
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

!+
! 23456789012345678901234567890123456789012345678901234567890123456789|
!+
! A cutting plane algotithm a f t e r Boris Jeremic
|
!+
! P u r p o s e : I m p l i c i t r e t u r n mapping
|
!
|
!
I /O Type
|
! D
I R ( 6 , 6 ) : e l a s t i c s t i f f n e s s matrix
|
! Sig0
I R(6)
: s t r e s s at s t a r t of increment
|
! SigE
I R(6)
: elastic trial stress
|
! dEps
I R(6)
: s t r a i n increment
|
! UsrSta0 I R ( )
: s t a t e v a r i a b l e s at s t a r t of increment
|
! UsrSta O R ( )
: s t a t e v a r i a b l e s a t end o f i n c r e m e n t
|
! Sig
O R(6)
: s t r e s s a t end o f i n c r e m e n t
|
! R0 , Q0
I R
: i n t e r n a l v a r i a b l e s at s t a r t of increment
|
! R ,Q
O R
: i n t e r n a l v a r i a b l e s a t end o f i n c r e m e n t
|
! ...
I R or I : model s p e c i f i c p a r a m e t e r s
|
!
|
! Needs : D e r i v a t i v e s o f t h e y i e l d f u n c t i o n ( s ) and p l a s t i c p o t e n t i a l ( s )
! Surface1 ( yield surface ) :
dfdSig :=n
|
! S u r f a c e 1 ( p l a s t i c p o t e n t i a l ) : d g d S i g : =m , d 2 g d S i g 2 : =m2
|
! Surface1 ( int . varaibles ) :
R
|
! Surface2 ( yield surface ) :
dfdSig :=u
|
! S u r f a c e 2 ( p l a s t i c p o t e n t i a l ) : d g d S i g : =w , d 2 g d S i g 2 : = w2
|
! Surface2 ( int . variables ) :
Q
|
! Assumes : dmdR = 0 and dwdQ= 0
|
!+
! Thomas Benz < t h o m a s . benz@wewinumge . de >
|
!+

31
32
33
34

!+
! Case 1 : S i n g l e s u r f a c e r e t u r n
|
!+

35
36

! 1. Initialize

37
38
39
40
41
42

i t =0
itmax =20
xTol = 1 d9
Unity = 0 . 0 d0
Unity ( 1 , 1 ) = 1 . 0 d0

173

Appendix B Return mapping in Fortran code

Unity ( 2 , 2 ) = 1 . 0 d0
Unity ( 3 , 3 ) = 1 . 0 d
R = R0

43
44
45
46

! 2 . Starting Point

47
48

S i g = SigE
xF1 = f 1 ( Sig , R . . . )
xF1Tol = xTol DAbs ( xF1 )
CALL C a l c D e r i v a t i v e s 1 ( Sig , R , n , n2 ,m, . . . )
xLI1 = xF /(DOT PRODUCT( n ,MATMUL(D,m) ) ( dF1dRdR ) )
S i g = SigExLI1 MATMUL(D,m)
R
= R0 + dR xLI1
xF1 = f 1 ( Sig , R . . . )

49
50
51
52
53
54
55
56
57

! 3 . E u l e r Backward . . . ( I t e r a t i o n )

58
59

DO WHILE ( ABS ( xF1 ) . GT . xF1Tol .AND. i t . LE . itmax )

CALL C a l c D e r i v a t i v e s 1 ( Sig , R , n , n2 ,m, . . . )
Res = Sig (SigExLI1 MATMUL(D,m) )
T=Unity+xLI1 MATMUL(D, m2)
CALL MatInv33 ( T , Ti )
dLL1 = ( xF1DOT PRODUCT( n ,MATMUL( Res , Ti ) ) )
dLI1 = dLI1 /(DOT PRODUCT( n ,MATMUL(MATMUL(D,m) , Ti ) ) ( dF1dRdR ) )
xLI1 = xLI1 + dxLI1
S i g = SigMATMUL( Res+dxLI1 MATMUL(D,m) , Ti )
R
= R0 + dR xLI1
xF1 = f 1 ( Sig , R . . . )
i t = i t +1
END DO

60
61
62
63
64
65
66
67
68
69
70
71
72
73

IF ( xLI1 . Gt . 0 .AND. i t . LE . itmax ) THEN

! R e t u r n mapping OK
ELSE
! Numerical problem ( s t a r t i n g p o i n t )?
END IF

74
75
76
77
78
79

RETURN
END

80
81
82
83
84
85

!+
! C a s e 2 : Two s u r f a c e r e t u r n
|
!+

86

! 1. Initialize

87
88

i t =0
itmax =20
xTol = 1 . 0 d9
Unity = 0 . 0 d0
Unity ( 1 , 1 ) = 1 . 0 d0
Unity ( 2 , 2 ) = 1 . 0 d0

89
90
91
92
93
94

174

95
96
97

Unity ( 3 , 3 ) = 1 . 0 d0
R = R0
Q = Q0

98
99

! 2 . Starting Point

100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120

S i g = SigE
xF1 = f 1 ( Sig , R . . . )
xF2 = f 2 ( Sig ,Q . . . )
xF1Tol = xTol DAbs ( xF1 )
xF2Tol = xTol DAbs ( xF2 )
CALL C a l c D e r i v a t i v e s 1 ( Sig , R , n , n2 ,m, . . . )
CALL C a l c D e r i v a t i v e s 2 ( Sig , Q, u , u2 ,w , . . . )
xOmega11 = dF1dRdR + DOT PRODUCT( n ,MATMUL(D,m) )
xOmega12 = DOT PRODUCT( n ,MATMUL(D,w) )
xOmega21 = DOT PRODUCT( u ,MATMUL(D,m) )
xOmega22 = dF2dQdQ + DOT PRODUCT( u ,MATMUL(D,w) )
xLI1
= ( xF1
xOmega22 xF2
xOmega12)/
2
( xOmega11 xOmega22 xOmega12 xOmega21 )
xLI2
= ( xF2
xOmega11 xF1
xOmega21)/
2
( xOmega11 xOmega22 xOmega12 xOmega21 )
S i g = SigE ( xLI1 MATMUL(D,m) + xLI2 MATMUL(D,w) )
R = R0 + dR xLI1
Q = Q0 + dQ xLI2
xF1 = f 1 ( Sig , R . . . )
xF2 = f 2 ( Sig ,Q . . . )

121
122

! 3 . E u l e r Backward . . . ( I t e r a t i o n )

123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146

DO WHILE ( ( ABS ( xF1 ) . GT . xF1Tol . OR.

2
ABS ( xF2 ) . GT . xF2Tol . OR . )
3
.AND. i t . LE . itmax )
CALL C a l c D e r i v a t i v e s 1 ( Sig , R , n , n2 ,m, . . . )
CALL C a l c D e r i v a t i v e s 2 ( Sig , Q, u , u2 ,w , . . . )
Res = Sig (SigExLI1 MATMUL(D,m) xLI2 MATMUL(D,w) )
T=Unity+xLI1 MATMUL(D, m2)+ xLI2 MATMUL(D, w2)
CALL MatInv33 ( T , Ti )
xOmega11 = DOT PRODUCT( n ,MATMUL(MATMUL(D,m) , Ti ) )
2
dF1dRdR
xOmega12 = DOT PRODUCT( n ,MATMUL(MATMUL(D,w) , Ti ) )
xOmega21 = DOT PRODUCT( u ,MATMUL(MATMUL(D,m) , Ti ) )
xOmega22 = DOT PRODUCT( u ,MATMUL(MATMUL(D,w) , Ti ) )
2
dF2dQdQ
xOmegaF1 = F1 DOT PRODUCT( Res ,MATMUL( Ti , n ) )
xOmegaF2 = F2 DOT PRODUCT( Res ,MATMUL( Ti , u ) )
dxLI1 = ( xOmegaCo xOmega22 xOmegaCa xOmega12)/
2
( xOmega11 xOmega22 xOmega12 xOmega21 )
dxLI2 = ( xOmegaCa xOmega11 xOmegaCo xOmega21)/
2
( xOmega11 xOmega22 xOmega12 xOmega21 )
xLI1 = xLI1 + dxLI1
xLI2 = xLI2 + dxLI2
S i g = SigMATMUL( Res+dxLI1 MATMUL(D,m)

175

Appendix B Return mapping in Fortran code

2

147

+dxLI2 MATMUL(D,w) , Ti )

R = R0 + dR xLI1
Q = Q0 + dQ xLI2
xF1 = f 1 ( Sig , R . . . )
xF2 = f 2 ( Sig ,Q . . . )
i t = i t +1
END DO

148
149
150
151
152
153
154

IF ( xLICone . Gt . 0 .AND. xLICap . GT . 0 .AND. i t . LE . itmax ) THEN

! R e t u r n mapping OK
ELSE
! Numerical problem ( s t a r t i n g p o i n t )?
END IF

155
156
157
158
159
160

RETURN
END

161
162

176

Appendix C
Two surface return strategy
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

!+
! 23456789012345678901234567890123456789012345678901234567890123456789|
!+
! HS g r a y c o r n e r s t r a t e g y o r when t o a c t i v a t e c o n e , cap , o r b o t h
|
!+
! After P .G . Bonnier ( 2 0 0 0 )
|
!+
!
1 . I n i t ( Small s t r a i n s t i f f n e s s , D i l a t a n c y , e t c . )
!
2. Trial stress
!
3 . Check y i e l d s u r f a c e s and r e t u r n s t r e s s
IF ( Cone y i e l d s u r f a c e i s v i o l a t e d ) THEN
CALL Cone Return ! R e t u r n and h a r d e n c o n e
IF ( F a i l u r e c r i t e r i a i s v i o l a t e d ) THEN
CALL C o n e F a i l u r e ! R e t u r n t o u l t i m a t e c o n e
IF ( Tension c r i t e r i a i s exceeded ) THEN
CALL Tension CutOff ! Do a 3 s u r f a c e t e n s i o n r e t u r n
END IF
END IF
IF ( Cap y i e l d s u r f a c e i s v i o l a t e d ) THEN
CALL Cap Return ! R e t u r n and h a r d e n c a p
IF ( Cone y i e l d s u r f a c e i s v i o l a t e d ) THEN
CALL Mixed Return ! R e t u r n and h a r d e n c o n e and c a p
IF ( F a i l u r e c r i t e r i a i s v i o l a t e d ) THEN
CALL C a p F a i l u r e ! R e t u r n t o c a p and u l t i m a t e c o n e
END IF
END IF
END IF
ELSE ! No f r i c t i o n h a r d e n i n g ( y e t ) : c h e c k Cap o n l y
IF ( Cap y i e l d s u r f a c e i s v i o l a t e d ) THEN
CALL Cap Return ! R e t u r n and h a r d e n c a p
IF ( Cone y i e l d s u r f a c e i s v i o l a t e d ) THEN
CALL Mixed Return ! R e t u r n and h a r d e n c o n e and c a p
IF ( F a i l u r e c r i t e r i a i s v i o l a t e d ) THEN
CALL C a p F a i l u r e ! R e t u r n t o c a p and u l t i m a t e c o n e
END IF
END IF
END IF
END IF
END IF
!
4 . Check Apex ; Update s t r e s s and s t a t e v a r i a b l e s ; E x i t

177

Appendix D
Material data

Table D.1: Element tests - HS-Small parameters

Parameter (Symbol)

Unit

I. User defind parameters

ref
E50
ref
Eoed
Eurref
m
c

ur
pref
K0nc
Rf
Tension
E0ref
0.7
II. Internal parameters
Eiref

KS /KC

kN
m2
kN
m2
kN
m2

]
[ kN
m2

[ ]
[ ]
[ ]
kN
m2

[ ]
[ ]
kN

m2
kN
m2

[ ]
kN
m2

[ ]
[ ]

Sand(D) Sand(L)
30000
30000
90000
0.55
0.00
42.0
16.0
0.25
100
0.40
0.90
0.00
270000
0.0002
65488
1.47
1.84

Clay

12000
2150
16000
1050
60000 11500
0.75
0.80
0.00
0.00
34.0
20.0
0.0
0.0
0.25
0.20
100
100
0.44
0.66
0.90
0.90
0.00
0.00
168000 80000
0.0001 0.0002
23800
1.56
2.01

14050
0.78
4.76

179

Appendix D Material data

Table D.2: Steinhaldenfeld NATM tunnel - HS-Small parameters

Parameter (Symbol)

Unit

I. User defined parameters

unsat
sat
ref
E50
ref
Eoed
Eurref
m
c

ur
pref
K0nc
Rf
Tension
E0ref
0.7
II. Internal parameters
Eiref

KS /KC

kN
m3
kN
m3
kN
m2
kN
m2
kN

180

m2

]
[ kN
m2

[ ]
[ ]
[ ]
kN
m2

[ ]
[ ]
kN

m2
kN
m2

[ ]
kN
m2

[ ]
[ ]

Fill

Keuper (up)

Keuper (lo)

Limestone

20.0
10000
10000
30000
0.50
10.00
25.0
0.0
0.20
100
0.58
0.90
0.00
120000
0.0002

24.0
33000
33000
100000
0.40
25.00
25.0
0.0
0.20
100
0.58
0.90
0.00
400000
0.0002

23.0
16000
16000
48000
0.40
25.00
25.0
0.0
0.20
100
0.58
0.90
0.00
192000
0.0002

23.0
190000
190000
575000
0.30
200.00
35.00
0.0
0.20
100
0.43
0.90
0.00
2400000
0.00005

24206
0.80
1.62

81216
0.83
1.72

39353
0.83
1.71

415649
1.52
1.95

Table D.3: Heinenoord tunnel - HS-Small parameters

Parameter (Symbol)

Unit

I. User defined parameters

unsat
sat
ref
E50
ref
Eoed
Eurref
m
c

ur
pref
K0nc
Rf
Tension
E0ref
0.7
II. Internal parameters
Eiref

KS /KC

kN
m3
kN
m3
kN
m2
kN
m2
kN
m2

]
[ kN
m2

[ ]
[ ]
[ ]
kN
m2

[ ]
[ ]
kN

m2
kN
m2

[ ]
kN
m2

[ ]
[ ]

Fill

Sand

Clay

17.5
17.5
20.0
20.0
14000
35000
12000
14000
35000
7000
42000 105000
35000
0.50
0.50
0.90
3.00
0.00
7.00
27.0
35.0
31.0
0.0
0.0
0.0
0.20
0.20
0.20
100
100
100
0.55
0.43
0.48
0.90
0.90
0.90
0.00
0.00
0.00
126000 240000 210000
0.0002 0.0002 0.0002
32786
0.88
1.62

72888
1.43
1.64

34582
0.93
1.93

181

Appendix D Material data

Table D.4: Deep excavation in sand (Berlin) - HS-Small parameters

182

Parameter (Symbol)

Unit

I. User defined parameters

unsat
sat
ref
E50
ref
Eoed
Eurref
m
c

ur
pref
K0nc
Rf
Tension
E0ref
0.7
II. Internal parameters
Eiref

KS /KC

kN
m3
kN
m3
kN
m2
kN
m2
kN
m2

]
[ kN
m2

[ ]
[ ]
[ ]
kN
m2

[ ]
[ ]
kN

m2
kN
m2

[ ]
kN
m2

[ ]
[ ]

Sand (L1)

Sand (L2)

Sand (L3)

19.0
20.0
45000
45000
180000
0.55
1.00
35.0
5.0
0.20
100
0.43
0.90
0.00
405000
0.0002

19.0
20.0
75000
75000
300000
0.55
1.00
38.0
6.0
0.20
100
0.38
0.90
0.00
675000
0.0002

19.0
20.0
105000
105000
315000
0.55
1.00
38.0
6.0
0.20
100
0.38
0.90
0.00
na
na

96662
1.48
2.15

154447
1.87
2.07

208642
1.88
1.59

Table D.5: Excavation in Rupel clay - HS-Small parameters

Parameter (Symbol)

Unit

I. User defined parameters

unsat
sat
ref
E50
ref
Eoed
Eurref
m
c

ur
pref
K0nc
Rf
Tension
E0ref
0.7
II. Internal parameters
Eiref

KS /KC

kN
m3
kN
m3
kN
m2
kN
m2
kN
m2

]
[ kN
m2

[ ]
[ ]
[ ]
kN
m2

[ ]
[ ]
kN

m2
kN
m2

[ ]
kN
m2

[ ]
[ ]

Fill

Rupel Clay

18.0
20.0
20000
20000
60000
0.50
1.00
32.0
2.0
0.20
100
0.43
0.90
0.00
180000
0.00015

19.0
20.0
15000
15000
50000
0.70
25.00
20.0
0.0
0.15
100
0.66
0.90
0.00
207000
0.00025

43421
1.16
1.63

39992
5.31
1.36

183

Appendix D Material data

Table D.6: Load test on a spread footing on sand - HS-Small parameters

184

Parameter (Symbol)

Unit

I. User defined parameters

unsat
sat
ref
E50
ref
Eoed
Eurref
m
c

ur
pref
K0nc
Rf
Tension
E0ref
0.7
II. Internal parameters
Eiref

KS /KC

kN
m3
kN
m3
kN
m2
kN
m2
kN
m2

]
[ kN
m2

[ ]
[ ]
[ ]
kN
m2

[ ]
[ ]
kN

m2
kN
m2

[ ]
kN
m2

[ ]
[ ]

Sand (Triaxial) Sand (Site)

18.0
20.0
16000
22000
90000
0.50
0.00
37.0
1.5
0.20
100
0.41
0.90
0.00
260000
0.0002

18.0
20.0
16000
22000
90000
0.50
0.00
37.0
1.5
0.20
100
0.41
0.90
0.00
390000
0.0004

30785
5.08
2.38

30785
5.08
2.38


Table D.7: Sulfeld
excavation - HS-Small parameters
Parameter (Symbol)

Unit

I. User defined parameters

unsat
sat
ref
E50
ref
Eoed
Eurref
m
c

ur
pref
K0nc
Rf
Tension
E0ref
0.7
II. Internal parameters
Eiref

KS /KC

kN
m2
kN
m2
kN
m2
kN
m2
kN
m2

]
[ kN
m2

[ ]
[ ]
[ ]
kN
m2

[ ]
[ ]
kN

m2
kN
m2

[ ]
kN
m2

[ ]
[ ]

Sand(D) Glacial Till Clayey silt

18.5
20.5
40000
38000
120000
0.50
0.10
37.5
5.0
0.20
100
0.80
0.90
0.00
480000
0.0002

18.5
21.5
8500
6150
25750
0.70
6.00
28.0
6.0
0.29
100
0.80
0.90
0.00
154500
0.0003

18.5
21.5
8500
6000
23000
0.90
30.00
32.0
10.0
0.30
100
0.80
0.90
0.00
138000
0.0003

82037
1.62
1.65

27783
0.90
2.78

20715
1.10
2.22

185

Curriculum Vit

14.05.1971 Geboren in Tubingen

1991-1993 Universitat Tubingen

Studium der Physik (Vordiplom)
1993-1997 Universitat Stuttgart
Studium des Bauingenieurwesens (Dipl.-Ing.)
1997-1999 Michigan Technological University
Studium der Geophysik (M.Sc.)

1998-2000 Schlumberger Geco-Prakla, Nord- und Sudamerika

Field Geophysicist
2000-2002 Leonhardt, Andra und Partner, Stuttgart

Projektingenieur Bruckenbau
Wasserbau, Karlsruhe, Referat Grundbau
2002-2006 Bundesanstalt fur
Wissenschaftlicher Angestellter
Geotechnik
2002-2006 Universitat Stuttgart, Institut fur
Wissenschaftlicher Mitarbeiter
20.09.2006 Promotion

187

Mitteilungen des Instituts fur

Geotechnik

Nr. 01

Thamm, B. R.

(1974)

Nr. 02

Gumann, P.

(1975)

Nr. 03

Feeser, V.

(1975)

Nr. 04

Du Thin, K.

(1976)

Nr. 05

Smoltczyk, U. /
Pertschi, O. /
Hilmer, K.

(1976)

Nr. 06

Hilmer, K.

(1976)

Nr. 07

Laumans, Q.

(1977)

Nr. 08

Lachler, W.

(1977)

Nr. 09

Spotka, H.

(1977)

Anfangssetzungen und Anfangsporenwasser

uberdr
ucke
eines normalverdichteten wassergesattigten Tones.
e 5,11
Einheitliche Berechnung von Grundbruch und

Boschungsbruch.
e 2,56
die
Die Bedeutung des Kalziumkarbonats fur

bodenphysikalischen Eigenschaften vom Lo.

Vergriffen

Standsicherheit von Boschungen:

ProgrammDokumentation.
Vergriffen
Messungen an Schleusen in der UDSSR.
Schleusennorm der UDSSR (SN 30365).
Vergriffen
Erddruck auf Schleusenkammerwande.
e 9,20
Verhalten einer ebenen, in Sand eingespannten
Wand bei nichtlinearen Stoffeigenschaften des
Bodens.
e 9,20
Beitrag zum Problem der Teilflachenpressung

bei Beton am Beispiel der Pfahlkopfanschlusse.

Vergriffen
Einflu der Bodenverdichtung mittels Ober
flachenruttelger
aten auf den Erddruck einer

Stutzwand
bei Sand.
Vergriffen

189

 Geotechnik
Mitteilungen des Instituts fur
Nr. 10

Schad, H.

(1979)

Nr. 11

Ulrich, G.

(1980)

Gumann, P.
Nr. 12

Salden, D.

(1980)

Nr. 13

Seeger, H.

(1980)

Nr. 14

Schmidt, H.H.

(1981)

Nr. 15

Smoltczyk, U. /
Schweikert, O.

(1981)

Nr. 16

Malcharek, K. /
Smoltczyk, U.

(1981)

Nr. 17

Gruhle, H.D.

(1981)

Nr. 18

Kobler, W.

(1982)

Nr. 19

Lutz, W.

(1983)

Nr. 20

Smoltczyk, U.

(1983)

190

Boden

Nichtlineare Stoffgleichungen fur

und
ihre Verwendung bei der numerischen Analyse
von Grundbauaufgaben.
Vergriffen
Verschiebungs- und kraftgesteuerte Platten
druckversuche auf konsolidierenden Boden.
Zum Modellgesetz der Konsolidation.
e 10,23
Der Einflu der Sohlenform auf die Traglast
von Fundamenten.
e 12,78
Beitrag zur Ermittlung des horizontalen Bet
tungsmoduls von Boden
durch Seitendruckversuche im Bohrloch.
e 12,78
Beitrag zur Ermittlung des Erddrucks auf

Stutzw
ande bei nachgiebigem Baugrund.
e 12,78

Vorstudie uber
bauliche Alternativen fur
Durchgangsstraen in Siedlungen.
e 6,14
die BerechVergleich nationaler Richtlinien fur
nung von Fundamenten.
e 7,67
Das Verhalten des Baugrundes unter Einwirkung vertikal gezogener Ankerplatten als
raumliches Problem des Erdwiderstandes.
Vergriffen

Untersuchungen uber
Boschungsund Grundbruch bei begrenzten Lastflachen.
e 12,78
Tragfahigkeit des geschlitzten Baugrunds
neben Linienlasten.
e 12,78
Studienunterlagen
Bodenmechanik
und

Grundbau; uberarbeitete
Ausgabe 1983.
e 20,45

 Geotechnik
Mitteilungen des Instituts fur
Nr. 21

Schweikert, O.

(1984)

Nr. 22

Vogt, N.

(1984)

Nr. 23

Buchmaier, R.

(1985)

Nr. 24

Schad, H.

(1985)

Smoltczyk, U. /
Schad, H. /
Zoller, P.
Nr. 25

Gumann, P.

(1986)

Nr. 26

Steinmann, B.

(1985)

Nr. 27

Lee, S.D.

(1987)

Nr. 28

Kolb, H.

(1988)

Nr. 29

Ochmann, H.

(1988)

Nr. 30

Breinlinger, F.

(1989)

Nr. 31

Smoltczyk, U. /
Breinlinger, F. /
Schad, H. /
Wittlinger, M.

(1989)

Der Einflu des Boschungswinkels

auf die
Berechnung des aktiven Erddrucks.
e 10,23
Erdwiderstandsermittlung bei monotonen und
wiederholten Wandbewegungen in Sand.
Vergriffen
Zur Berechnung von Konsolidationsproblemen bei nichtlinearem Stoffverhalten.
e 12,78

Moglichkeiten
der Boschungssicherung
bei
kleinen Baugruben.

Sonderkonstruktionen der Boschungssicherung.

e 17,90
Die Methode der Kinematischen Elemente.
e 10,23

Zum Verhalten bindiger Boden

bei monotoner
einaxialer Beanspruchung.
Vergriffen
Untersuchungen zur Standsicherheit von
Schlitzen im Sand neben Einzelfundamenten.
Vergriffen

Ermittlung der Sohlreibung von Grundungs

korpern
unter horizontalem kinematischen
Zwang.
e 12,78

Ebene Grenzzustande von Erdboschungen

im
stochastischen Sicherheitskonzept.
e 12,78
Bodenmechanische
Stoffgleichungen
bei
groen Deformationen sowie Be- und
Entlastungsvorgangen.
e 15,34
Beitrag zur Bemessung von Tunneln in offener
Bauweise.

e 12,78

191

 Geotechnik
Mitteilungen des Instituts fur
Nr. 32

Gumann, P. /
Schanz, T. /
Smoltczyk, U. /
Willand, E.

(1990)

Nr. 33

Gruhle, H.D.

(1990)

Nr. 34

Henne, J.

(1995)

Nr. 35

Wittlinger, M.

(1994)

Nr. 36

Schad, H.

(1992)

Nr. 37

Belz, I.

(1992)

Nr. 38

Ma, J.

(1994)

Nr. 39

Smoltczyk, U.

(1994)

Nr. 40

Rilling, B.

(1994)

Nr. 41

Vermeer, P.A.

(1996)

Nr. 42

Vermeer, P.A.

(1997)

Nr. 43

Brinkmann, C.

(1998)

192

Beitrage zur Anwendung der KEM (Erddruck,

Grundbuch, Standsicherheit von Boschungen)

Vergriffen

Der raumliche Erdwiderstand vor uberwiegend horizontal belasteten Ankerplatten.

Vergriffen
Zur Bewehrung von verformten Bodenschichten durch Einsatz zugfester Geokunststoffe.
e 15,34
Ebene Verformungsuntersuchungen zur Weck
ung des Erdwiderstandes bindiger Boden.
e 15,34
Zeit- und geschwindigkeitsabhangiges Materialverhalten in der Geotechnik Experimentelle
Erfassung und numerische Analyse.
e 15,34
Zur Ermittlung dynamischer Bodenkennwerte
in situ aus der Systemantwort des Erregers.
e 15,34
Untersuchungen zur Standsicherheit der durch

Stutzscheiben
stabilisierten Boschungen.
e 15,34
Sonderheft: 25 Jahre Lehre und Forschung in
der Geotechnik.
e 15,34
Untersuchungen zur Grenztragfahigkeit bindi

ger Schuttstoffe
am Beispiel von Lolehm.
e 17,90
Deponiebau und Geotechnik.
e 17,90
Baugruben in Locker- und Festgestein.
e 17,90
Untersuchungen zum Verhalten von Dich
tungsuberg
angen im Staudammbau.
e 17,90

 Geotechnik
Mitteilungen des Instituts fur
Nr. 44

Fiechter-Scharr,
I.

(1998)

Nr. 45

Schanz, T.

(1998)

Nr. 46

Akinrogunde,
A.E.

(1999)

Nr. 47

Vogt-Breyer, C.

(1999)

Nr. 48

Vermeer, P.A.

(1999)

Nr. 49

Marcher, T.

(2002)

Nr. 50

Marcher, T.

(2003)

Nr. 51

Ruse, N.M.

(2004)

Nr. 52

Beutinger, P.H.

(2005)

Nr. 53

Wehnert, M.

(2006)

Nr. 54

Moller,
S.C.

(2006)

Nr. 55

Benz, T.

(2007)

Beeinflussung von Erdbaustoffen durch

Beimischen eines organophilen Bentonits.
e 17,90
Zur Modellierung des mechanischen Verhaltens von Reibungsmaterialien.
e 17,90
Propagation of Cement Grout in Rock Discontinuities Under Injection Conditions.
e 17,90
Experimentelle und numerische Untersuchungen zum Tragverhalten und zur Bemessung
horizontaler Schraubanker.
e 17,90
Neue Entwicklungen in der Geotechnik.
e 17,90
Resultate eines Versuchsprogramms an Beaucaire-Mergel.
e 17,90
Nichtlokale Modellierung der Entfestigung
dichter Sande und steifer Tone.
e 17,90
Raumliche Betrachtung der Standsicherheit
der Ortsbrust beim Tunnelvortrieb.
e 17,90
Ein geotechnischer Beitrag zur Standsicherheit
mobiler Baumaschinen.
e 17,90
Ein Beitrag zur drainierten und undrainierten
Analyse in der Geotechnik.
e 17,90
Tunnel induced settlements and forces in linings
e 17,90
Small-Strain Stiffness of Soils and its Numerical Consequences.
e 17,90

193