Numerical Modelling of Shotcrete

for Tunnelling

A thesis submitted to Imperial College London in partial fulfilment of the

requirements for the degree of Doctor of Philosophy in the Faculty of Engineering

by

Dipl. Ing.

Reinhard Schütz

Department of Civil and Environmental Engineering

Imperial College London
London SW7 2AZ

February 2010
This work is dedicated with much affection and
gratitude to my beloved parents Maria and
Siegfried

2
 Galileo Galilei (15.02.1564 - 08.01.1642)
Italian physicist, mathematician, astronomer and philosopher

“The laws of Nature are written in the language of mathematics . . . the

symbols are triangles, circles and other geometrical figures, without
whose help it is impossible to comprehend a single word.”

3
Declaration

The work presented in this dissertation was carried out in the Department of Civil and
Environmental Engineering at Imperial College London from August 2006. This thesis is
the result of my own work and any quotation from, or description of the work of others is
acknowledged herein by reference to the sources, whether published or unpublished.

This dissertation is not the same as any that I have submitted for any degree, diploma or other
qualification at any other university. No part of this thesis has been or is being concurrently
submitted for any such degree, diploma or other qualification.

This document is available on-line at www.imperial.ac.uk/geotechnics/publications, it is less

than 100,000 words long, contains less than 265 figures and less than 65 tables.

Reinhard Schütz
London, 05/02/2010

4
Abstract

Shotcrete is a special type of concrete which was invented at the beginning of the 20th century
and is nowadays an important support element for tunnels constructed with the New Austrian
Tunnelling Method (NATM). Immediately after tunnel excavation shotcrete is sprayed onto
the tunnel walls at high pressure in order to provide temporary support. Unfortunately, in the
past very little attention has been given to the development of sophisticated material laws
for shotcrete, since its design and application was mainly based on experience. However,
current design practice, such as that applied in the design of the new Crossrail tunnels,
requires sophisticated modelling of shotcrete behaviour in numerical analysis of tunnel-soil
interaction. Such models are not readily found in the literature, as they have been developed
mainly for structural, rather than geotechnical applications.

In this thesis, a constitutive model for the time-dependent behaviour of shotcrete has been
developed within the framework of elasto-plasticity. Two independent yield surfaces control
the behaviour of shotcrete in multiaxial loading conditions for both compression and tension.
The model formulation is based on strain hardening/softening plasticity, where the expansion
and contraction of the yield surfaces are governed by normalised plastic strains. Cracking
of the shotcrete is considered within the smeared crack concept. Furthermore, the proposed
material law includes the time-dependency of stiffness and strength behaviour. The reducing
deformability of the shotcrete during cement hydration has been taken into account. The
model has been extended to account for creep, shrinkage and hydration temperature induced
deformations at early shotcrete ages. After a robust implementation into the Imperial College
Finite Element Program (ICFEP), model calibration and validation have been performed for
shotcrete experiments taken from the literature. The developed constitutive model has then
been applied to the analysis of a typical tunnel construction in London Clay, modelling the
early age material properties of the shotcrete tunnel lining in detail for various excavation
schemes. Finally, it has been shown that the proposed constitutive model is capable of
reproducing the complex behaviour of young shotcrete at early ages and can be applied
successfully to boundary value problems in geotechnical engineering.

5
Acknowledgements

At first I would like to thank my supervisors Prof. D. Potts and Dr. L. Zdravković for giving
me the great opportunity to do research at the Geotechnics Section of Imperial College
London. Their confidence and endless support helped and motivated me to successfully
carry out this challenging work. Their guidance through hard stages of my thesis is highly
appreciated.

This research project was funded by the Geotechnical Consulting Group (GCG) London.
Their patience and financial support is gratefully acknowledged.

I would also like to express my gratitude to the College staff for their help, support and
contributions in several fruitful discussions. They showed me that science can be a very
interdisciplinary field where collaboration and social skills are of high importance. Many
thanks are due to all my friends and colleagues that I have made during my period of research
for making the experience of staying in London an unforgettable one. I do not want to forget
to thank all my friends back home in Austria, who visited me various times and therefore
made our bond of friendship grow stronger despite the distance.

Finally, my last thanks go to my family in Austria for encouraging and supporting me in

everything that I have done in life. Without them I would never have completed this thesis.

6
Contents

Abstract 5

Acknowledgements 6

Contents 7

List of figures 14

List of tables 25

1 Introduction 28
1.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.2 Tunnel collapse at Heathrow Airport . . . . . . . . . . . . . . . . . . . . . . . 29
1.3 Aims of research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.4 Outline of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.5 Definition of stress and strain variables . . . . . . . . . . . . . . . . . . . . . . 33
1.6 Terminology and symbols used in thesis . . . . . . . . . . . . . . . . . . . . . 39

2 Tunnelling in urban areas 44

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.2 The New Austrian Tunnelling Method . . . . . . . . . . . . . . . . . . . . . . 44
2.3 Typical tunnel construction process in urban areas . . . . . . . . . . . . . . . 48
2.4 The importance of monitoring in tunnelling . . . . . . . . . . . . . . . . . . . 49
2.4.1 Measuring and interpretation of displacements . . . . . . . . . . . . . 50
2.4.2 Measuring and interpretation of stresses in tunnel linings . . . . . . . 51
2.5 A review of tunnel lining design . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.5.1 Empirical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.5.2 Closed-form analytical methods . . . . . . . . . . . . . . . . . . . . . . 54
2.5.2.1 Continuum analytical methods . . . . . . . . . . . . . . . . . 55
2.5.2.2 Convergence-confinement method . . . . . . . . . . . . . . . 55
2.5.2.3 Bedded-beam-spring models . . . . . . . . . . . . . . . . . . 56
2.5.3 Numerical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.5.3.1 Bearing-capacity-diagram . . . . . . . . . . . . . . . . . . . . 58

7
 2.6 Single-shell method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.6.1 Innovative shotcrete tunnelling . . . . . . . . . . . . . . . . . . . . . . 61
2.7 Modelling 2D tunnel excavation in a numerical analysis . . . . . . . . . . . . 63
2.7.1 Stress relief method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.7.2 Stiffness reduction method . . . . . . . . . . . . . . . . . . . . . . . . 64
2.7.3 Gap method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.7.4 Volume loss control method . . . . . . . . . . . . . . . . . . . . . . . . 65
2.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3 The Finite Element Method 69

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.2 Finite element theory for linear materials . . . . . . . . . . . . . . . . . . . . 70
3.2.1 Element discretisation . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2.2 Approximation of primary variable . . . . . . . . . . . . . . . . . . . . 70
3.2.3 Formulation of element equations . . . . . . . . . . . . . . . . . . . . . 72
3.2.3.1 Numerical integration . . . . . . . . . . . . . . . . . . . . . . 74
3.2.4 Formulation of global equations . . . . . . . . . . . . . . . . . . . . . . 75
3.2.5 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.2.6 Solution of global equations . . . . . . . . . . . . . . . . . . . . . . . . 76
3.3 Geotechnical considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.3.0.1 Undrained analysis . . . . . . . . . . . . . . . . . . . . . . . 77
3.4 Finite element theory for non-linear materials . . . . . . . . . . . . . . . . . . 79
3.4.1 The modified Newton-Raphson method . . . . . . . . . . . . . . . . . 79
3.4.2 Stress point algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

4 Shotcrete Technology 85
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.2 Shotcrete as an engineering material . . . . . . . . . . . . . . . . . . . . . . . 85
4.3 Mix design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.3.1 Aggregates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.3.2 Cement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.3.3 Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.3.4 Additives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.3.4.1 Silica fume . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.3.4.2 Accelerators . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.3.4.3 Plasticisers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3.5 Fibres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3.5.1 Steel fibres . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3.5.2 Polypropylene fibres . . . . . . . . . . . . . . . . . . . . . . . 95
4.3.5.3 Glass fibres . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

8
 4.4 Cement hydration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.4.1 Reaction process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.4.2 Degree of hydration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.5 Shotcrete installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.5.1 Dry-mix process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.5.2 Wet-mix process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.5.3 Spraying technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.5.4 Rebound behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.6 Shotcrete testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

5 Mechanical and time-dependent behaviour of concrete and shotcrete 108

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.2 Uniaxial stress conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.2.1 Mechanical behaviour in compression . . . . . . . . . . . . . . . . . . . 108
5.2.2 Mechanical behaviour in tension . . . . . . . . . . . . . . . . . . . . . 111
5.3 Biaxial stress conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.4 Triaxial stress conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.5 Anisotropy of shotcrete . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.6 Mechanical behaviour of steel fibre reinforced concrete and shotcrete . . . . . 117
5.7 Time-dependent mechanical behaviour of young shotcrete . . . . . . . . . . . 120
5.7.1 Increase in stiffness and strength . . . . . . . . . . . . . . . . . . . . . 121
5.7.2 Deformability at early ages . . . . . . . . . . . . . . . . . . . . . . . . 132
5.7.3 Damage effects due to early-age loading . . . . . . . . . . . . . . . . . 136
5.8 Creep and relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
5.8.1 Mechanisms behind creep . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.8.2 Influencing factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.8.3 Creep in tunnelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.9 Shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.9.1 Mechanisms behind shrinkage . . . . . . . . . . . . . . . . . . . . . . . 148
5.9.1.1 Plastic shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.9.1.2 Drying shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . 148
5.9.1.3 Autogeneous shrinkage . . . . . . . . . . . . . . . . . . . . . 149
5.9.1.4 Carbonation shrinkage . . . . . . . . . . . . . . . . . . . . . 149
5.9.2 Factors influencing shrinkage . . . . . . . . . . . . . . . . . . . . . . . 150
5.10 Temperature induced deformation . . . . . . . . . . . . . . . . . . . . . . . . 153
5.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

6 Modelling of concrete and shotcrete 158

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
6.2 Empirical models for concrete in compression . . . . . . . . . . . . . . . . . . 158

9
6.3 Constitutive models based on elasticity . . . . . . . . . . . . . . . . . . . . . . 163
6.3.1 Linear elastic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.3.2 Non-linear elastic models . . . . . . . . . . . . . . . . . . . . . . . . . 164
6.4 Classical theory of elasto-plasticity . . . . . . . . . . . . . . . . . . . . . . . . 165
6.4.1 Basic concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
6.4.2 Formulation of the elasto-plastic constitutive matrix . . . . . . . . . . 168
6.4.3 Multi-surface plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . 169
6.5 Basic plasticity models for concrete . . . . . . . . . . . . . . . . . . . . . . . . 171
6.5.1 Rankine criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
6.5.2 Mohr-Coulomb criterion . . . . . . . . . . . . . . . . . . . . . . . . . . 172
6.5.3 Drucker-Prager criterion . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6.5.4 Chen & Chen criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
6.5.5 Further sophisticated constitutive models for concrete . . . . . . . . . 180
6.6 Modelling post-peak behaviour of concrete . . . . . . . . . . . . . . . . . . . . 180
6.6.1 Post-peak behaviour in tension . . . . . . . . . . . . . . . . . . . . . . 180
6.6.1.1 Discrete crack model . . . . . . . . . . . . . . . . . . . . . . 181
6.6.1.2 Smeared crack model . . . . . . . . . . . . . . . . . . . . . . 181
6.6.1.3 Effect of reinforcement . . . . . . . . . . . . . . . . . . . . . 187
6.6.1.4 Modelling steel fibre reinforced concrete (SFRC) . . . . . . . 190
6.6.2 Post-peak behaviour in compression . . . . . . . . . . . . . . . . . . . 193
6.7 Modelling of creep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
6.7.1 Basic expressions for creep . . . . . . . . . . . . . . . . . . . . . . . . 194
6.7.2 Rheological units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
6.7.2.1 Kelvin model . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
6.7.2.2 Maxwell model . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.7.2.3 Burger model . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
6.7.2.4 Rheological models used for modelling shotcrete behaviour . 198
6.7.3 Creep according to EC 2 (2004) . . . . . . . . . . . . . . . . . . . . . . 203
6.7.4 Rate of flow method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.7.4.1 Rate of flow method after Schubert (1988) . . . . . . . . . . 204
6.7.4.2 Rate of flow method after Golser et al. (1990) . . . . . . . . 206
6.7.4.3 Rate of flow method after Aldrian (1991) . . . . . . . . . . . 206
6.8 Modelling of shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
6.8.1 Shrinkage after American Concrete Institute (ACI) . . . . . . . . . . . 207
6.8.2 Shrinkage after Bazant Model B3 . . . . . . . . . . . . . . . . . . . . . 208
6.8.3 Shrinkage after Gardner & Lockman (2001) . . . . . . . . . . . . . . . 209
6.9 Modelling hydration temperature in concrete structures . . . . . . . . . . . . 209
6.10 A review of different constitutive models applied for tunnel lining design . . . 210
6.10.1 Hypothetical Modulus of Elasticity (HME) . . . . . . . . . . . . . . . 210
6.10.2 Non-linear beam model by Moussa (1993) . . . . . . . . . . . . . . . . 212

10
 6.10.3 Post-failure modelling of shotcrete by Hafez (1995) . . . . . . . . . . . 213
6.10.4 Viscoplastic material model for shotcrete by Meschke (1996) . . . . . 214
6.10.5 Shotcrete model based on chemoplasticity by Hellmich et al. (1999a) . 215
6.10.6 Assessment of constitutive models for shotcrete by Thomas (2003) . . 217
6.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

7 A constitutive Model for Shotcrete 221

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
7.2 Formulation of the constitutive model . . . . . . . . . . . . . . . . . . . . . . 222
7.2.1 Structure of the model . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
7.2.2 Yield Function and Plastic Potential in compression . . . . . . . . . . 223
7.2.3 Yield Function and Plastic Potential in tension . . . . . . . . . . . . . 224
7.3 Hardening and softening rules . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
7.3.1 Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
7.3.2 Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
7.3.2.1 Option A - Linear softening . . . . . . . . . . . . . . . . . . . 233
7.3.2.2 Option B - Exponential softening . . . . . . . . . . . . . . . 235
7.4 Time-dependent material behaviour . . . . . . . . . . . . . . . . . . . . . . . 237
7.4.1 Increase in stiffness and strength with time . . . . . . . . . . . . . . . 237
7.4.2 Shotcrete deformability at early ages . . . . . . . . . . . . . . . . . . . 240
7.4.3 Creep and relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
7.4.4 Shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
7.4.5 Temperature induced strains . . . . . . . . . . . . . . . . . . . . . . . 245
7.5 Summary of parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
7.6 Formulation of the elasto-plastic constitutive matrix for time-dependent ma-
terials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
7.7 Correction of the right hand side load vector . . . . . . . . . . . . . . . . . . 252
7.8 Multi-surface plasticity for time-dependent materials . . . . . . . . . . . . . . 253
7.9 Model implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
7.9.1 Selection of the active yield surface . . . . . . . . . . . . . . . . . . . . 258
7.9.2 Derivatives of the stress invariants . . . . . . . . . . . . . . . . . . . . 258
7.9.3 Partial derivatives for the compression yield surface . . . . . . . . . . 259
7.9.4 Partial derivatives for the tension yield surface . . . . . . . . . . . . . 260
7.9.5 Calculation of the hardening modulus for the compression surface . . 262
7.9.6 Calculation of the hardening modulus for the tension surface . . . . . 263
7.9.7 Calculation of the elastic constitutive matrix . . . . . . . . . . . . . . 265
7.9.8 Calculation of the changes in the compressive hardening parameter . . 265
7.9.9 Calculation of the changes in the tensile hardening parameter . . . . . 266
7.10 Model calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
7.10.1 Increase in stiffness and compressive strength . . . . . . . . . . . . . . 267

11
 7.10.2 Complete uniaxial compression tests at different shotcrete ages . . . . 269
7.10.3 Creep deformation in uniaxial compression . . . . . . . . . . . . . . . 272
7.10.4 Shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
7.10.5 Temperature development during cement hydration . . . . . . . . . . 275
7.11 Model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
7.11.1 Increase in stiffness with time . . . . . . . . . . . . . . . . . . . . . . . 276
7.11.2 Uniaxial loading in compression and tension . . . . . . . . . . . . . . . 277
7.11.3 Biaxial loading in compression . . . . . . . . . . . . . . . . . . . . . . 281
7.11.4 Creep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
7.11.5 Shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
7.11.6 Thermal strains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
7.12 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286

8 Analysis of tunnel excavation in London Clay 287

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
8.2 Geometry of the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
8.3 Ground conditions in London area and initial stresses . . . . . . . . . . . . . 289
8.4 Finite element model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
8.4.1 Mesh and boundary conditions . . . . . . . . . . . . . . . . . . . . . . 290
8.4.2 Modelling of construction sequence . . . . . . . . . . . . . . . . . . . . 294
8.4.2.1 Excavation Type 1: Full face excavation . . . . . . . . . . . . 294
8.4.2.2 Excavation Type 2: Bench-invert excavation . . . . . . . . . 295
8.4.2.3 Excavation Type 3: Sidewall drift . . . . . . . . . . . . . . . 296
8.5 Adopted constitutive models and material parameters . . . . . . . . . . . . . 298
8.5.1 Tunnel lining - Shotcrete . . . . . . . . . . . . . . . . . . . . . . . . . 298
8.5.2 Soil - London Clay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
8.6 Tunnel analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
8.6.1 Parametric study for excavation Type 1 . . . . . . . . . . . . . . . . . 303
8.6.2 Parametric study for excavation Type 2 . . . . . . . . . . . . . . . . . 306
8.6.3 Parametric study for excavation Type 3 . . . . . . . . . . . . . . . . . 308
8.7 Discussion of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
8.7.1 Surface settlements and volume loss . . . . . . . . . . . . . . . . . . . 312
8.7.1.1 Excavation Type 1 . . . . . . . . . . . . . . . . . . . . . . . . 312
8.7.1.2 Excavation Type 2 . . . . . . . . . . . . . . . . . . . . . . . . 318
8.7.1.3 Excavation Type 3 . . . . . . . . . . . . . . . . . . . . . . . . 322
8.7.2 Hoop stress distribution and utilisation factor . . . . . . . . . . . . . . 328
8.7.2.1 Excavation Type 1 . . . . . . . . . . . . . . . . . . . . . . . . 328
8.7.2.2 Excavation Type 2 . . . . . . . . . . . . . . . . . . . . . . . . 349
8.7.2.3 Excavation Type 3 . . . . . . . . . . . . . . . . . . . . . . . . 360
8.7.3 Lining displacements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371

12
 8.7.3.1 Excavation Type 1 . . . . . . . . . . . . . . . . . . . . . . . . 371
8.7.3.2 Excavation Type 2 . . . . . . . . . . . . . . . . . . . . . . . . 380
8.7.3.3 Excavation Type 3 . . . . . . . . . . . . . . . . . . . . . . . . 386
8.7.4 Soil stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
8.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405

9 Summary and conclusions 407

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
9.2 General remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
9.3 A constitutive model for shotcrete . . . . . . . . . . . . . . . . . . . . . . . . 410
9.4 Conclusions from tunnel analyses . . . . . . . . . . . . . . . . . . . . . . . . . 411
9.5 Recommendations for future research . . . . . . . . . . . . . . . . . . . . . . . 414

Appendix 417

References 419

13
List of Figures

1.1 CTA concourse tunnel eye after the collapse (from HSE, 2000) . . . . . . . . 30
1.2 a) Principal stress space and b) deviatoric plane (from Potts & Zdravković,
1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.1 New design philosophy for tunnels according to NATM (from Müller & Fecker,
1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2 Simplified definition of principles and effects of NATM (from Sauer, 1988) . . 46
2.3 Different excavation sequences for NATM tunnel construction . . . . . . . . . 48
2.4 Longitudinal tunnel section showing excavation rounds for a bench-invert con-
struction scheme (from ICE, 2004) . . . . . . . . . . . . . . . . . . . . . . . . 49
2.5 Typical arrangement of convergence targets for tunnel lining (from Jones et al.,
2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.6 Deformation examples of a roof lining (from Rokahr et al., 2005) . . . . . . . 51
2.7 Radial and tangential pressure cells (from Clayton et al., 2002 and Jones et al.,
2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.8 Plane strain continuum model and charactersitic distribution of radial displace-
ments u, radial stresses on the lining σr , hoop forces N and bending moments
M (from Duddeck & Erdmann, 1982) . . . . . . . . . . . . . . . . . . . . . . 55
2.9 The convergence-confinement method (after Oreste, 2003) . . . . . . . . . . . 56
2.10 Bedded ring model without bedding near the tunnel crown (from Duddeck &
Erdmann, 1982) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.11 Bearing-capacity-diagram for tunnel lining (from Hoek, 2007) . . . . . . . . . 59
2.12 Composition of tunnel lining with the single-shell method (from Kusterle &
Lukas, 1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
TM
2.13 The LaserShell method (from Jones et al., 2008) . . . . . . . . . . . . . . . 62
TM
2.14 TunnelBeamer system integration (from Eddie & Neumann, 2003) . . . . . 63
2.15 Principle of the stress relief method (from Potts & Zdravković, 2001) . . . . . 64
2.16 Principle of the stiffness reduction method (from Potts & Zdravković, 2001) . 65
2.17 Principle of the gap method (from Potts & Zdravković, 2001) . . . . . . . . . 65
2.18 Principle of the volume loss control method (from Potts & Zdravković, 2001) 66

3.1 8-noded isoparametric element used in ICFEP (from Potts & Zdravković, 1999) 71

14
3.2 Location of Gauss points for 2 x 2 and 3 x 3 integration order (from Potts &
Zdravković, 1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3 Basic principle of the modified Newton-Raphson scheme (after Potts & Zdravković,
1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.4 Basic principle of the substepping approach (from Potts & Zdravković, 1999) 81

4.1 Carl Akeley and his first cement-gun . . . . . . . . . . . . . . . . . . . . . . . 86

4.2 Recommended aggregate gradation zone from EFNARC (1996) . . . . . . . . 89
4.3 Recommended cement content of shotcrete mix before installation depending
on maximum aggregate size (from Aldrian, 1991) . . . . . . . . . . . . . . . . 90
4.4 Relationship between compressive strength and water-cement ratio for concrete
(from Karrer, 1986) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.5 Time dependent influence of accelerator dosage on compressive strength (from
Kusterle, 1985) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.6 Different types of steel fibres used in sprayed concretes (from Banthia et al.,
1992) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.7 Polypropylene fibres used in wet-mix sprayed concrete (from Austin & Robins,
1995) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.8 Schematic representation of the cement hydration (after Labahn, 1982) . . . . 97
4.9 Process of dry-mix technology for sprayed concrete (from Austin & Robins,
1995) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.10 Process of wet-mix technology for sprayed concrete (from Austin & Robins,
1995) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.11 Different types of conveying systems for shotcrete (from Girmscheid, 2000) . 101
4.12 Preferred application procedure for sprayed concrete (from Girmscheid, 2000) 102
4.13 Effect of layer thickness on material rebound (from Parker et al., 1977) . . . . 103
4.14 Effect of surface orientation on material rebound (from Girmscheid, 2000) . . 104
4.15 Test methods for compressive strength of young shotcrete according to ASCCT
(2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.16 Proposed test method for uniaxial compressive strength and Young’s modulus
by Chang (1994) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

5.1 Typical plots of stress-strain curves for uniaxial compression (from Chen, 1982)
(fc = fcp ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.2 Influence of specimen length on post peak behaviour in uniaxial compression
(from CEB-FIP Model Code, 1990) . . . . . . . . . . . . . . . . . . . . . . . . 110
5.3 Relation between stress level and Poisson’s ratio μ (from Chen, 1982) (fc = fcp )111
5.4 Stress-strain curves in tension for different concrete types (from Chen, 1982) . 111
5.5 Relation between uniaxial tensile and compressive strength (from Byfors, 1980)
(fct = ftp , fcube = fcp ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

15
5.6 Stress-strain relationships for concrete under biaxial compression (from Chen,
1982) (fc = fcp ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.7 Biaxial failure envelope (from Kupfer & Gerstle, 1973) (βp = fcp ) . . . . . . . 115
5.8 Triaxial yield and failure surface for concrete (from Chen, 1982) . . . . . . . . 116
5.9 Spraying direction and loading conditions in a real tunnel lining . . . . . . . . 117
5.10 Failure mechanism for steel fibre renforced shotcrete and concrete (from Ding
& Kusterle, 2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.11 Uniaxial stress-deformation characteristics of fibre reinforced concrete (from
Thomée, 2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.12 Compressive stress-strain curves for fibre reinforced concrete (from Strack, 2007)119
5.13 Stress-strain curves for plain concrete and steel fibre reinforced concrete with
60 kg/m3 at an age of 9 h (from Ding & Kusterle, 2000) . . . . . . . . . . . . 120
5.14 Stress-strain curves in compression at different shotcrete ages (from Sezaki
et al., 1989) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.15 Development of Young’s modulus E with shotcrete age obtained from various
tests (from Chang, 1994) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.16 Relationship between stiffness and strength (from Byfors, 1980) (Ecc = E,
fcc = fcp ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.17 Increase in stiffness with time according to various equations from the literature125
5.18 Variation of Poisson’s ratio with time (from Aydan et al., 1992) (ν = μ) . . . 126
5.19 Relation between Poisson’s ratio and compressive strength (from Byfors, 1980)
(fcc = fcp, νcc = μ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.20 Development of uniaxial compressive strength, fcp , with shotcrete age obtained
from various tests (from Chang, 1994) . . . . . . . . . . . . . . . . . . . . . . 127
5.21 Development of uniaxial compressive strength with time according to various
equations from the literature . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.22 Early-age strength classes for young shotcrete (after ASCCT, 2004) . . . . . . 130
5.23 Ratio between tensile and compressive strength at early concrete ages (from
Kasai et al., 1971) (Ft = ftp , Fc = fcp) . . . . . . . . . . . . . . . . . . . . . . 131
5.24 Development of compressive strains at peak strength with uniaxial compressive
strength and time (from Sezaki et al., 1989 and Thomas, 2003) . . . . . . . . 133
5.25 Development of compressive peak strains with shotcrete age (from Wierig, 1971)133
5.26 Development of tensile peak strains with shotcrete age (from Kusterle & Lukas,
1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.27 Ratio between tensile and compressive peak strain with compressive strength
(from Byfors, 1980) (fcc = fcp , εcto = εtp , εcco = εcp ) . . . . . . . . . . . . . . 136
5.28 Reduction in gain of uniaxial compressive strength due to preloading (from
Moussa, 1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.29 General form of strain development with time due to creep (from Neville, 1970)138
5.30 Creep recovery at unloading (from Bosnjak, 2000) . . . . . . . . . . . . . . . 139

16
5.31 Possible values of creep Poisson’s ratio μcr under uniaxial stress (from Neville,
1970) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.32 Factors influencing creep (from Byfors, 1980) . . . . . . . . . . . . . . . . . . 142
5.33 Relative creep with age at loading (from Byfors, 1980) . . . . . . . . . . . . . 143
5.34 Creep for different stress levels (from Neville, 1970) . . . . . . . . . . . . . . . 144
5.35 Creep for different relative humidity (from Neville, 1981) . . . . . . . . . . . . 145
5.36 Development of lining stresses during tunnel advance (from Pöttler, 1993) . . 146
5.37 Creep and relaxation during tunnel advance (from Rokahr & Lux, 1987) . . . 147
5.38 Concept of shrinkage induced stresses (from Malmgrem et al., 2005) . . . . . 148
5.39 Influence of water content on absolute values of shrinkage (from Neville, 1981) 150
5.40 Shrinkage of shotcrete (from Neubert & Manns, 1993) . . . . . . . . . . . . . 151
5.41 Influence of aggregate content in concrete on the ratio of shrinkage of concrete
to the shrinkage of neat cement paste (from Neville, 1981) . . . . . . . . . . . 152
5.42 Influence of relative humidity on absolute values for shrinkage (from Neville,
1981) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.43 Temperature development in hardening concrete (from Mehta & Monteiro, 1985)154
5.44 Age dependency for the coefficient of thermal expansion αt (-/ C)(from Byfors,
1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

6.1 Stress-strain diagram for uniaxial compression (from CEB-FIP Model Code,
1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.2 Uniaxial compressive stress strain curves for concrete (from Desayi & Krishnan,
1964) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
6.3 Compressive stress-strain curves for different lateral confining pressures (from
Attard & Setunge, 1996) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6.4 Uniaxial compressive stress-strain curves for a) non-linear structural analysis
and b) for the design of cross-sections (after EC 2, 2004) . . . . . . . . . . . . 162
6.5 Variation of bulk and shear moduli with σoct and τoct (from Kotsovos & New-
man, 1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
6.6 Yield surface and stress states for elastic and elasto-plastic material behaviour 166
6.7 Rankine criterion in J-p (left) and principal stress space (right) . . . . . . . . 172
6.8 Mohr-Coulomb failure criterion . . . . . . . . . . . . . . . . . . . . . . . . . . 172
6.9 Mohr-Coulomb criterion in J-p (left) and principal stress space (right) . . . . 174
6.10 Drucker-Prager criterion in J-p (left) and principal stress space (right) . . . . 174
6.11 Yield and failure surfaces of Chen & Chen criterion in the principal (left) and
biaxial stress space (right) (from Chen, 1982) . . . . . . . . . . . . . . . . . . 175
6.12 Zoning of the stress state (from Chen, 1982) (fc = fcp , ft = ftp ) . . . . . . . . 178
6.13 Idealised uniaxial stress-strain curve for concrete (from Chen, 1982) (fc = fcy ,
fc = fcp, ft = fty , ft = ftp ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
6.14 Early discrete crack modelling (from De Borst et al., 2004) . . . . . . . . . . 181

17
6.15 Deformation characteristics of tensile post-peak behaviour . . . . . . . . . . . 182
6.16 Intrinsic material function for fracture energy (from Lackner & Mang, 2004) . 184
6.17 Stress-strain and stress-crack-opening diagram for uniaxial tension (from CEB-
FIP Model Code, 1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
6.18 Some proposed stress-crack-opening relations (from Moussa, 1993) . . . . . . 186
6.19 Models for reinforcement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
6.20 Principal effect of tension-stiffening on post-peak behaviour of concrete . . . 188
6.21 Stress-strain diagram for reinforced concrete in tension (from Gilbert & Warner,
1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
6.22 Material function for γ depending on degree of hydration ξ (from Lackner &
Mang, 2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
6.23 Different stress-crack-opening relations for steel fibre reinforced concrete (from
Thomée, 2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
6.24 Uniaxial tensile stress-crack-opening relation (from Thomée, 2005) . . . . . . 192
6.25 Increase in fracture energy with fibre content (from Kazemi et al., 2007) . . . 193
6.26 Kelvin body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
6.27 Deformation response of Kelvin model . . . . . . . . . . . . . . . . . . . . . . 196
6.28 Maxwell body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.29 Deformation response of Maxwell model . . . . . . . . . . . . . . . . . . . . . 197
6.30 Burger body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
6.31 Deformation response of Burger model . . . . . . . . . . . . . . . . . . . . . . 198
6.32 Non-linear Kelvin model after Rokahr & Lux (1987) (from Hesser, 2000) . . . 199
6.33 Constitutive behaviour of shotcrete after Zheng (1989) (from Hesser, 2000) . 200
6.34 Non-linear Burger model (from Hauggaard et al., 1997) . . . . . . . . . . . . 202
6.35 Functions for viscosity and stiffness in non-linear Burger model (from Haug-
gaard et al., 1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
6.36 Deformation history with time(from England & Illston, 1965) . . . . . . . . . 204
6.37 Deformation components for young shotcrete (from Schubert, 1988) . . . . . 205
6.38 Development of shrinkage strains with time (after ACI, 1992) . . . . . . . . . 208
6.39 Typical shrinkage curves given by Model B3 (from Bažant & Baweja, 1995) . 208
6.40 Principle of the HME-approach (from Karakus & Fowell, 2003) . . . . . . . . 211
6.41 Stress-strain behaviour of young and old shotcrete (from Watson et al., 1999) 212
6.42 Proposed model for stress-strain history of shotcrete (from Moussa, 1993) . . 213
6.43 Loading surfaces of the viscoplastic material model for shotcrete (from Meschke
et al., 1996) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
6.44 Intrinsic material functions of shotcrete (from Lackner et al., 2002) . . . . . . 216
6.45 Yield surfaces and hardening law for chemoplasticity model for shotcrete (from
Lackner et al., 2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

18
7.1 Compression and tension yield surfaces in a) the biaxial stress space and b)
the J-p space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
7.2 Apex rounding for tension surface in a) the biaxial stress space and b) the J-p
space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
7.3 Tip tolerance for apex region of tension surface . . . . . . . . . . . . . . . . . 226
7.4 Strain hardening concept for time-dependent matrials . . . . . . . . . . . . . 227
7.5 Normalised stress-strain curve for shotcrete in compression . . . . . . . . . . 228
7.6 Fracture energy concept for shotcrete in compression . . . . . . . . . . . . . . 230
7.7 a) Snap back behaviour for total stress-strain curve, b) Limited ultimate plastic
compressive strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
7.8 Critical value of the ultimate compressive strength for hardened concrete at
28 days . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
7.9 Normalised hardening and softening curve in tension for option A . . . . . . . 233
7.10 Fracture energy concept for shotcrete in tension with linear softening . . . . . 234
7.11 Normalised hardening and softening curve in tension for option B . . . . . . . 235
7.12 Fracture energy concept for shotcrete in tension with exponential softening . 236
7.13 Increase in stiffness with time for various values of cement parameter s . . . . 238
7.14 Qualitative increase in compressive yield, peak, failure and ultimate strength
with time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
7.15 Development of plastic uniaxial compressive peak strains with time . . . . . . 240
7.16 Time-dependent rheological body for creep . . . . . . . . . . . . . . . . . . . 241
7.17 Development of creep strains with time upon loading at time to . . . . . . . . 242
7.18 Increase in shotcrete viscosity with time . . . . . . . . . . . . . . . . . . . . . 242
7.19 Creep potential surfaces and incremental creep strain vectors following a creep
flow rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
7.20 Adopted shrinkage model and the influence of parameter B . . . . . . . . . . 245
7.21 Development of ΔT with time . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
7.22 Parameter fitting for increase in stiffness with time (after Huber, 1991) . . . . 268
7.23 Parameter fitting for increase in compressive strength with time (after Huber,
1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
7.24 Parameter fitting for increase in stiffness with time (after Sezaki et al., 1989) 270
7.25 Parameter fitting for increase in compressive strength with time (after Sezaki
et al., 1989) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
7.26 Parameter fitting for development of total peak and failure strains (after Sezaki
et al., 1989) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
7.27 Parameter fitting for complete stress strain curves in compression for shotcrete
at various ages (after Sezaki et al., 1989) . . . . . . . . . . . . . . . . . . . . . 272
7.28 Parameter fitting for development of creep deformation at various stages of
hydration (after Kusterle, 1999) . . . . . . . . . . . . . . . . . . . . . . . . . . 273
7.29 Parameter fitting for development of shrinkage deformation . . . . . . . . . . 274

19
7.30 Parameter fitting for development of hydration temperature with time (from
Huber, 1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
7.31 Comparison of model predictions with analytical solutions for increase in stiff-
ness with time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
7.32 Comparison of model predictions with analytical solutions for stress-strain
curves at different shotcrete ages . . . . . . . . . . . . . . . . . . . . . . . . . 278
7.33 Comparison of model predictions with analytical solutions for complete stress-
strain curves at two different shotcrete ages . . . . . . . . . . . . . . . . . . . 279
7.34 Comparison of model predictions with analytical solutions for stress-strain
curves in tension for various shotcrete ages (Option A) . . . . . . . . . . . . . 280
7.35 Comparison of model predictions with analytical solutions for stress-strain
curves in tension for various shotcrete ages (Option B) . . . . . . . . . . . . . 280
7.36 Comparison of model predictions with analytical solutions for uniaxial and
biaxial compressive loading at a shotcrete age of 6 h and 28 d . . . . . . . . . 281
7.37 Comparison of model predictions with analytical solutions for creep deforma-
tion (n = 0.98) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
7.38 Comparison of model predictions with analytical solutions for creep deforma-
tion with n > 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
7.39 Comparison of model predictions with analytical solutions for shrinkage defor-
mation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
7.40 Comparison of model predictions with analytical solutions for hydration tem-
perature induced strains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285

8.1 Layout of tunnel geometry and ground conditions . . . . . . . . . . . . . . . . 288

8.2 Three different excavation sequences for tunnel analyses . . . . . . . . . . . . 289
8.3 Adopted Ko -profiles for performed analyses . . . . . . . . . . . . . . . . . . . 290
8.4 General layout of the 2D finite element mesh (without inner core) . . . . . . 291
8.5 Two types of blocks forming the finite element mesh . . . . . . . . . . . . . . 291
8.6 Zoom of mesh 1 for excavation Type 1: full face excavation . . . . . . . . . . 292
8.7 Zoom of mesh 2 for excavation Type 2: bench-invert excavation . . . . . . . . 293
8.8 Zoom of mesh 3 for excavation Type 3: sidewall drift . . . . . . . . . . . . . . 293
8.9 Graphical representation of construction sequence for excavation Type 1 . . . 295
8.10 Graphical representation of construction sequence for excavation Type 2 . . . 296
8.11 Graphical representation of construction sequence for excavation Type 3 . . . 297
8.12 Development of total uniaxial compressive peak strains with time . . . . . . . 300
8.13 Variation of the shear and bulk moduli with strain level (from Potts & Zdravković,
1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
8.14 Variation of the compressive peak strains with time . . . . . . . . . . . . . . . 306
8.15 Analysed cross sections of tunnel lining for all three excavation sequences . . 310

20
8.16 Zoom into connection point of temporary invert with tunnel lining for excava-
tion Type 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
8.17 Investigated lining displacements for all three excavation sequences . . . . . . 311
8.18 Surface settlements at end of tunnel excavation (Excavation Type 1, Ko = 0.5) 312
8.19 Surface settlements at end of tunnel excavation (Excavation Type 1, Ko = 1.5) 313
8.20 Predicted volume loss for both Ko -profiles adopting different shotcrete ages at
the end of tunnel excavation (Excavation Type 1) . . . . . . . . . . . . . . . . 314
8.21 Surface settlements for different developments of stiffness and strength of
shotcrete with time (Excavation Type 1, Ko = 0.5) . . . . . . . . . . . . . . . 315
8.22 Surface settlements for different developments of stiffness and strength of
shotcrete with time (Excavation Type 1, Ko = 1.5) . . . . . . . . . . . . . . . 315
8.23 Surface settlements for different shotcrete deformabilities (Excavation Type 1,
Ko = 0.5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
8.24 Surface settlements for analysis of creep, shrinkage and hydration temperature
of shotcrete (Excavation Type 1, Ko = 0.5) . . . . . . . . . . . . . . . . . . . 317
8.25 Surface settlements for analysis of creep, shrinkage and hydration temperature
of shotcrete (Excavation Type 1, Ko = 1.5) . . . . . . . . . . . . . . . . . . . 317
8.26 Surface settlements for various time-runs at the end of bench excavation (Inc
12, top) and end of invert excavation (Inc 25, bottom) . . . . . . . . . . . . . 319
8.27 Surface settlements after complete excavation for runs TPM 1 to 4 . . . . . . 320
8.28 Surface settlements after complete excavation for various developments in stiff-
ness and strength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
8.29 Surface settlements after complete excavation for various shotcrete deforma-
bilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
8.30 Surface settlements for various time-runs at the end of excavation of the left
side gallery (Inc 12, top) and end of complete excavation (Inc 25, bottom) for
excavation Type 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
8.31 Surface settlements after complete excavation for various developments of stiff-
ness and strength for excavation Type 3 . . . . . . . . . . . . . . . . . . . . . 325
8.32 Surface settlements after complete excavation for various shotcrete deforma-
bilities for excavation Type 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
8.33 Horizontal hoop stress distribution for crown at end of excavation (Inc 10) for
Ko = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
8.34 Horizontal hoop stress distribution for crown at end of loading (Inc 50) for
Ko = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
8.35 Vertical hoop stress distribution in right sidewall at end of excavation (Inc 10)
for various developments of stiffness and strength adopting Ko = 0.5 . . . . . 332
8.36 Horizontal hoop stress distribution in invert at end of loading (Inc 50) for
various developments of stiffness and strength adopting Ko = 0.5 . . . . . . . 333

21
8.37 Horizontal hoop stress distribution in crown at end of loading (Inc 50) for
various developments of shotcrete deformability adopting Ko = 0.5 . . . . . . 334
8.38 Vertical hoop stress distribution in right tunnel sidewall for analysis of creep,
shrinkage and hydration temperature adopting Ko = 0.5 . . . . . . . . . . . . 335
8.39 Horizontal hoop stress distribution in tunnel crown for analysis of creep, shrink-
age and hydration temperature adopting Ko = 0.5 . . . . . . . . . . . . . . . 335
8.40 Vertical hoop stress distribution for right sidewall at end of excavation (Inc
10) for Ko = 1.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
8.41 Change in bending direction of tunnel lining due to tunnel excavation with
initial stresses for Ko = 1.5 and surface loading . . . . . . . . . . . . . . . . . 338
8.42 Vertical hoop stress distribution for right sidewall at end of loading (Inc 50)
for Ko = 1.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
8.43 Vertical hoop stress distribution in right sidewall at end of excavation (Inc 10)
for various developments of stiffness and strength adopting Ko = 1.5 . . . . . 340
8.44 Horizontal hoop stress distribution in invert at end of loading (Inc 50) for
various developments of stiffness and strength adopting Ko = 1.5 . . . . . . . 341
8.45 Horizontal hoop stress distribution in crown at end of loading (Inc 50) for
various developments of shotcrete deformability Ko = 1.5 . . . . . . . . . . . 342
8.46 Vertical hoop stress distribution in right tunnel sidewall for analysis of creep,
shrinkage and hydration temperature adopting Ko = 1.5 (Inc 100) . . . . . . 343
8.47 Horizontal hoop stress distribution in tunnel crown for analysis of creep, shrink-
age and hydration temperature adopting Ko = 1.5 (Inc 100) . . . . . . . . . . 344
8.48 Stress level for different shotcrete ages at intrados of tunnel crown for both
Ko -profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
8.49 Stress level for different shotcrete ages at intrados of tunnel sidewall for both
Ko -profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
8.50 Stress level for different shotcrete ages at intrados of tunnel invert for both
Ko -profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
8.51 Horizontal hoop stresses in tunnel invert for various time-runs at the end of
invert excavation (Inc 25, top) and end of surface loading (Inc 65, bottom) . 351
8.52 Horizontal hoop stresses in temporary invert for various time-runs at the end
of bench excavation (Inc 12) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
8.53 Horizontal hoop stresses in tunnel crown for time-runs TPM 1 to 4 at the end
of invert excavation (Inc 25) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
8.54 Horizontal hoop stresses in tunnel crown for time-runs TPM 5 to 8 at the end
of invert excavation (Inc 25) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
8.55 Horizontal hoop stresses in tunnel crown for time-runs TPM 9 to 12 at the end
of invert excavation (Inc 25) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
8.56 Horizontal hoop stresses in tunnel invert for various developments of stiffness
and strength at the end of surface loading (Inc 65) . . . . . . . . . . . . . . . 355

22
8.57 Horizontal hoop stresses in tunnel crown for various shotcrete deformabilty at
the end of surface loading (Inc 65) . . . . . . . . . . . . . . . . . . . . . . . . 356
8.58 Horizontal hoop stresses in tunnel crown for analysis of creep, shrinkage and
hydration temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
8.59 Horizontal hoop stresses in tunnel invert for analysis of creep, shrinkage and
hydration temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
8.60 Vertical hoop stresses in left sidewall for time-runs TPM 1 to 4 at the end of
surface loading (Inc 65) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
8.61 Vertical hoop stresses in left sidewall for time-runs TPM 5 to 8 at the end of
surface loading (Inc 65) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
8.62 Vertical hoop stresses in left sidewall for time-runs TPM 9 to 12 at the end of
surface loading (Inc 65) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
8.63 Vertical hoop stresses in temporary sidewall at the end of excavation of the
left side gallery (Inc 12) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
8.64 Vertical hoop stresses in right tunnel sidewall at the end of tunnel excavation
(Inc 25) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
8.65 Vertical hoop stresses in right tunnel sidewall at the end of surface loading
(Inc 65) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
8.66 Vertical hoop stresses in right sidewall at the end of surface loading (Inc 65)
for various developments of stiffness and strength . . . . . . . . . . . . . . . . 366
8.67 Vertical hoop stresses in left sidewall at the end of surface loading (Inc 25) for
various shotcrete deformabilities . . . . . . . . . . . . . . . . . . . . . . . . . 367
8.68 Vertical hoop stresses in left tunnel sidewall for analysis of creep, shrinkage
and hydration temperature (Inc 100) . . . . . . . . . . . . . . . . . . . . . . . 368
8.69 Vertical hoop stresses in right tunnel sidewall for analysis of creep, shrinkage
and hydration temperature (Inc 100) . . . . . . . . . . . . . . . . . . . . . . . 369
8.70 Vertical crown displacement for various time runs adopting Ko = 0.5 (Exca-
vation Type 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
8.71 Vertical invert displacement for variation in development of stiffness and strength
with Ko = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
8.72 Vertical crown displacement for variation in shotcrete deformability with Ko =
0.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
8.73 Horizontal displacement of right tunnel sidewall for various time runs adopting
Ko = 1.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
8.74 Horizontal sidewall displacement for variation in development of stiffness and
strength with Ko = 1.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
8.75 Vertical invert displacement for variation in shotcrete deformability with Ko =
1.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
8.76 Comparison of vertical crown displacements v and horizontal sidewall displace-
ments u for both Ko -profiles for time run TPM 6 . . . . . . . . . . . . . . . . 380

23
8.77 Vertical crown displacement for various time runs adopting Ko = 1.5 . . . . . 381
8.78 Vertical crown displacement for runs TPM 1 to 4 adopting Ko = 1.5 . . . . . 382
8.79 Horizontal displacement at the intersection of temporary invert with tunnel
lining for various time runs adopting Ko = 1.5 . . . . . . . . . . . . . . . . . . 383
8.80 Horizontal sidewall displacement for runs TPM 1 to 4 adopting Ko = 1.5 . . . 383
8.81 Vertical crown displacement for various developments of stiffness and strength
adopting Ko = 1.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
8.82 Horizontal displacements of left sidewall . . . . . . . . . . . . . . . . . . . . . 387
8.83 Horizontal displacements of right sidewall . . . . . . . . . . . . . . . . . . . . 388
8.84 Vertical displacement of intersection point of temporary sidewall with tunnel
lining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
8.85 Vertical displacement of intersection point of temporary sidewall with tunnel
lining for variousn developments of stiffness and strength . . . . . . . . . . . . 390
8.86 Horizontal displacement of left sidewall for various shotcrete deformabilities . 391
8.87 Effective horizontal stresses in ground after excavation of tunnel (Inc 10) for
linear elastic model (top) and run TPM 1 (bottom) adopting Ko = 0.5 . . . . 394
8.88 Effective vertical stresses in ground after excavation of tunnel (Inc 10) for
linear elastic model LE (top) and run TPM 1 (bottom) adopting Ko = 0.5 . . 396
8.89 Sub-accumulated pore water pressures in ground after excavation of tunnel (Inc
10) for linear elastic model (top) and run TPM 1 (bottom) adopting Ko = 0.5 398
8.90 Effective horizontal stresses in ground after excavation of tunnel (Inc 10) for
linear elastic model (top) and run TPM 1 (bottom) adopting Ko = 1.5 . . . . 400
8.91 Effective vertical stresses in ground after excavation of tunnel (Inc 10) for
linear elastic model LE (top) and run TPM 1 (bottom) adopting Ko = 1.5 . . 402
8.92 Sub-accumulated pore water pressures in ground after excavation of tunnel (Inc
10) for linear elastic model (top) and run TPM 1 (bottom) adopting Ko = 1.5 404

24
List of Tables

4.1 Influencing factors on rebound behaviour . . . . . . . . . . . . . . . . . . . . 103

5.1 Parameters a and c for increase in Young’s modulus E with time (after Weber,
1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.2 Parameters a and b for development of Poisson’s ratio μ (after Byfors, 1980 . 127
5.3 Parameters a and c for increase in uniaxial compressive strength, fcp , with
time (after Weber, 1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.4 Parameters a and b for development of peak strains with uniaxial compressive
strength (after Byfors, 1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.5 Absolute values for plastic shrinkage (Relative humidity 50 %, Temperature
20 C) (from Neville, 1981) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
5.6 Drying shrinkage after Hills, 1982 . . . . . . . . . . . . . . . . . . . . . . . . . 151
5.7 Maximum temperature rise as a function of layer thickness (after Kusterle &
Lukas, 1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154

6.1 Eci , Ec1 and εc,lim for various concrete grades after CEB-FIP Model Code (1990)160
6.2 Material parameters for concrete taken from EC 2 (2004) . . . . . . . . . . . 163
6.3 Base values of fracture energy Gf o from CEB-FIP Model Code (1990) . . . . 183
6.4 Coefficient αf to estimate wc (from CEB-FIP Model Code, 1990) . . . . . . . 186

7.1 Summary of parameters for time-dependent plasticity model (all the strength
and strain parameters are uniaxial parameters) . . . . . . . . . . . . . . . . . 248
7.2 Summary of model parameters for creep, shrinkage and thermal strains . . . 249
7.3 Summary of parameters for time-dependent elasticity model . . . . . . . . . . 249
7.4 Mix design for tested shotcrete (from Huber, 1991) . . . . . . . . . . . . . . . 267
7.5 Cement parameter for increase in stiffness and strength fitted to test data from
Huber (1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
7.6 Mix design for tested shotcrete (from Sezaki et al., 1989) . . . . . . . . . . . . 269
7.7 Fitted stiffness and strength parameters . . . . . . . . . . . . . . . . . . . . . 272
7.8 Test series performed by Kusterle (1999) . . . . . . . . . . . . . . . . . . . . . 273
7.9 Fitted model parameters for predicting creep deformation . . . . . . . . . . . 274
7.10 Fitted model parameters for shrinkage deformation . . . . . . . . . . . . . . . 275
7.11 Fitted model parameters for temperature history . . . . . . . . . . . . . . . . 276

25
7.12 Adopted model parameters for tensile behaviour . . . . . . . . . . . . . . . . 279
7.13 Initial shotcrete ages for performed creep tests . . . . . . . . . . . . . . . . . 282
7.14 Model parameters for second test series of temperature history . . . . . . . . 285

8.1 Summary for mesh 1, 2 and 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 294

8.2 Construction sequence for excavation Type 1 . . . . . . . . . . . . . . . . . . 294
8.3 Construction sequence for excavation Type 2 . . . . . . . . . . . . . . . . . . 295
8.4 Construction sequence for excavation Type 3 . . . . . . . . . . . . . . . . . . 297
8.5 Basic strength parameters for shotcrete . . . . . . . . . . . . . . . . . . . . . 298
8.6 Numerical parameters for tension yield surface . . . . . . . . . . . . . . . . . 299
8.7 Basic deformability parameters for shotcrete . . . . . . . . . . . . . . . . . . . 299
8.8 Basic stiffness parameters for shotcrete . . . . . . . . . . . . . . . . . . . . . . 300
8.9 Basic material parameters for creep, shrinkage and hydration temperature . . 301
8.10 Strength parameters for London Clay . . . . . . . . . . . . . . . . . . . . . . 302
8.11 Stiffness parameters for London Clay . . . . . . . . . . . . . . . . . . . . . . . 302
8.12 Shotcrete ages for different time-runs for excavation Type 1 . . . . . . . . . . 304
8.13 Tunnel runs simulating variation in stiffness and strength . . . . . . . . . . . 305
8.14 Tunnel runs simulating the variation in the development of shotcrete deforma-
bility with time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
8.15 Tunnel runs simulating creep, shrinkage and hydration temperature . . . . . 306
8.16 Shotcrete ages for different time-runs for excavation Type 2 . . . . . . . . . . 307
8.17 Tunnel runs simulating the variation of increase in stiffness and strength and
shotcrete deformability with time . . . . . . . . . . . . . . . . . . . . . . . . 308
8.18 Tunnel runs simulating creep, shrinkage and hydration temperature for exca-
vation Type 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
8.19 Shotcrete ages for different time-runs for excavation Type 3 . . . . . . . . . . 309
8.20 Volume loss and maximum surface settlement for various time-runs at comple-
tion of tunnel excavation (Inc 25) for excavation Type 2 . . . . . . . . . . . . 322
8.21 Volume loss and maximum surface settlement for various time-runs at comple-
tion of tunnel excavation (Inc 25) for excavation Type 3 . . . . . . . . . . . . 327
8.22 Utilisation factor for variation in development of stiffness and strength for
excavation Type 1 at end of surface loading (Ko = 0.5) . . . . . . . . . . . . . 348
8.23 Utilisation factor for variation in development of stiffness and strength for
excavation Type 1 at end of surface loading (Ko = 1.5) . . . . . . . . . . . . . 348
8.24 Utilisation factor for variation in shotcrete deformability for excavation Type
1 at end of surface loading (Ko = 0.5) . . . . . . . . . . . . . . . . . . . . . . 349
8.25 Utilisation factor for variation in shotcrete deformability for excavation Type
1 at end of surface loading (Ko = 1.5) . . . . . . . . . . . . . . . . . . . . . . 349
8.26 Minimum and maximum principal stresses occurring in the intersection of the
temporary invert with the tunnel lining at end of bench excavation (Inc 12) . 359

26
8.27 Utilisation factor of lining intrados at tunnel crown and invert for excavation
Type 2 (Ko = 1.5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
8.28 Minimum and maximum principal stresses occurring in the top intersection
of the temporary sidewall with the tunnel lining at end of left side gallery
excavation (Inc 12) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
8.29 Utilisation factor of lining intrados at left and right tunnel sidewall for exca-
vation Type 3 (Ko = 1.5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
8.30 Lining displacements of performed analyses for creep, shrinkage and hydration
temperature adopting Ko = 0.5 . . . . . . . . . . . . . . . . . . . . . . . . . . 375
8.31 Lining displacements of performed analyses for creep, shrinkage and hydration
temperature adopting Ko = 1.5 . . . . . . . . . . . . . . . . . . . . . . . . . . 379
8.32 Lining displacements of performed analyses for creep, shrinkage and hydration
temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
8.33 Maximum vertical displacement v of temporary invert at the end of bench
excavation (Inc 12) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386
8.34 Lining displacements of performed analyses for creep, shrinkage and hydration
temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
8.35 Maximum horizontal displacement u at the centre of the temporary sidewall
at the end of the left side gallery excavation (Inc 12) . . . . . . . . . . . . . . 392

27
Chapter 1

Introduction

1.1 General
Nowadays, major cities around the world are faced with an ever increasing amount of traffic.
However, their relatively old infrastructure systems are often not capable of dealing with this
problem due to limited space. Therefore, the construction of new underground infrastructure
projects provides a possible solution and is becoming more important in urban planning,
where the design of new tunnels plays a key role. One of the tunnelling techniques that has
been applied successfully around the world is the New Austrian Tunnelling Method, usually
abbreviated as NATM. The basic principle of this method is a sequent tunnel advance charac-
terised by excavation and installation of supporting elements in order to stabilise the ground.
One of the main support elements of NATM is sprayed concrete, often called shotcrete. Tun-
nel construction in an urban environment requires a safe and economic design, where the
accurate calculation of stresses and ground movements that occur around geotechnical struc-
tures is of crucial importance. Often engineers are not only interested in the stresses and
movements at failure, but also in the mechanical behaviour of the tunnel at different stages of
construction and under working loads. Another big concern is the impact of a tunnel advance
on existing structures either near to the ground surface or in the subsurface vicinity of the
tunnel.

The finite element method as a particular type of numerical analysis is a technique,

which is capable of dealing with complex soil-structure interaction problems and has been
increasingly used in engineering practice over the last twenty years. However, although
numerical analysis is a very powerful tool to analyse geotechnical structures such as tunnels,
the interpretation of the results depends on the knowledge and experience of the user and
the adopted constitutive laws for modelling the involved materials. For the simulation of soil
behaviour significant advances have been made recently and sophisticated models exist that
are able to reproduce the non-linear soil behaviour reasonably well. In the open literature it
was observed that the situation seems to be completely different for modelling the behaviour
of the tunnel lining. In most of the cases a tunnel shell made of cast concrete or shotcrete is

28
modelled as a linear elastic material, sometimes coupled with a step-wise increase in stiffness
with time, which is basically a very crude assumption. In particular shotcrete shows a
highly non-linear stress-strain behaviour at all stages of the hardening process during cement
hydration.

Important developments have been made over the last decade in shotcrete technology by
improving the quality of the constituent materials and the equipment used for installation.
However, in the past the design of a shotcrete tunnel lining was mainly based on the experience
of tunnel engineers - an approach which appeared to work well for deep tunnels excavated
in rocks. When considering tunnel construction in soft ground conditions such as London
Clay, a safe design of the whole structure can only be achieved by a realistic simulation of
all the materials involved. Some elaborate sprayed concrete models have been developed
recently, but a consistent framework for modelling shotcrete behaviour at early ages does
not exist and there is a need for further developments including mathematical formulations
and specific laboratory testing to validate models and fully understand the time-dependent
material response of this special type of concrete.

1.2 Tunnel collapse at Heathrow Airport

Worldwide there have been a number of collapses and failures of NATM tunnels that led to
serious damage to public buildings and infrastructure and even to human fatalities. One of
the worst civil engineering disasters in the United Kingdom happened during the night of
20th to the 21st October 1994 (HSE, 2000). During this nightshift several tunnels undergoing
construction beneath Heathrow Airport´s Central Terminal Area (CTA) collapsed completely.
This incident caused severe damage to neighbouring buildings and structures, but luckily no
one was injured. As a consequence all construction work came to a stop and some major
short-term disruption to the airport followed. The tunnels were part of the Heathrow Express
Rail Link Project, which is a high speed passenger service from central London (Paddington)
directly to Heathrow Airport. Tunnel construction was being carried out according to the
principles of the New Austrian Tunnelling Method (NATM) and compensation grouting had
been used as additional ground treatment.

Extensive investigations were carried after the tunnel collapse by the Health and Safety
Executive (UK) and their complete report can be found in HSE (2000). It was concluded that
the direct cause of the incident was a chain of events which involved the following principal
aspects:

 Poor design and planning

 Lack of quality during construction

 Lack of engineering control

29
  Lack of safety management

Kolymbas (1998) states, that a successful tunnel advance according to NATM requires the
installation of a highly complex construction material, which is shotcrete. It was found that,
in some parts of the collapsed tunnels at Heathrow Airport, the required material quality of
the shotcrete lining was not achieved, mainly due to inexperienced and not properly trained
workmen. Panels of the tunnel lining that were removed after the incident showed that the
poor construction quality of the invert panels including joints was one of the key factors
leading to failure. Fig. 1.1 shows the CTA concourse tunnel eye after the collapse. The
complete lining has dropped leading to large settlements of the tunnel crown. Furthermore,
the broken invert has rotated almost 90 and parts stand on end.

Figure 1.1: CTA concourse tunnel eye after the collapse (from HSE, 2000)

Visual inspection of the investigated panels provided evidence of some severe problems re-
garding the material quality of the shotcrete, which involved:

 Exposed reinforcement mesh in the surface of the shotcrete

 Thin sections of shotcrete where the design thickness of the lining was not achieved

 Inclusion of rebound material in the structure

After the failure, the question as to whether NATM and the use of shotcrete is a safe and
proper construction method for tunnels in soft ground conditions beneath highly populated
areas such as central London was debated at length. In HSE (2000) some lessons to be learned
from this tunnel collapse are mentioned. However, the main conclusion is that all the parties
involved in such a huge infrastructure project have to ensure that they have in place the
culture, commitment, competence and health and saftey management systems to secure the
risk control and the safe completion of the construction work (HSE, 2000; Kolymbas, 1998).

30
1.3 Aims of research
Following the discussion from the tunnel collapse at Heathrow Airport, the aim of this thesis
was to get a better insight into the behaviour of shotcrete when used for tunnel construction
in urban areas by applying the finite element method. To achieve this, the following steps
have been involved:

 To perform an extensive literature review on the mechanical and time-dependent be-

haviour of concrete and in particular on young shotcrete at early ages. Furthermore, to
get an overview of different approaches for simulating concrete and shotcrete behaviour
in a numerical analysis and detect the key aspects that have to be considered when
modelling such a material.

 Development of a sophisticated constitutive model for shotcrete that is capable of re-

producing the main characteristics of sprayed concrete particularly at early ages. This
includes a robust numerical implementation into a finite element code in order to apply
the model to complex boundary value problems.

 Numerical investigation of the mechanical behaviour of shotcrete in tunnelling for var-

ious types of tunnel construction, with a particular focus on the lining behaviour.

1.4 Outline of thesis

The outline of the work presented in this thesis is the following:

The remaining part of this chapter aims to introduce the definitions of stress and strain
variables used within this thesis. Furthermore, the applied terminology and a list of symbols
are given to enable the reader to fully understand the presented theoretical background in
Chapters 2 to 9.

In Chapter 2 the main principles of the New Austrian Tunnelling Method (NATM) are pre-
sented with a particular focus on tunnel construction in soft ground conditions. An overview
about different tunnel lining design methods should enable the reader to understand the dif-
ficulties in achieving a safe design. Furthermore, some innovative tunnelling techniques using
fibre-reinforced shotcrete are highlighted.

Chapter 3 presents a basic introduction into the finite element method applied for geotechni-
cal engineering and describes the numerical algorithms that are used for the implementation
of a constitutive model into the Imperial College Finite Element Program (ICFEP).

Chapter 4 deals with shotcrete technology and important topics regarding the appropriate
mix design of sprayed concrete are presented. The components of this particular type of con-
crete are discussed individually with a reference to national standards and recommendations

31
that are available in the open literature. The impact of each of the concrete constituents on
the mechanical behaviour of shotcrete is highlighted. A key point for describing the behaviour
of shotcrete is to understand the process of cement hydration during the hardening phase of
the material, which is explained briefly. Two existing methods of shotcrete installation are
introduced (wet- and dry-mix techniques) with a particular focus on the use of sprayed con-
crete in tunnelling. Finally, the chapter gives a brief overview of different testing procedures
of shotcrete at early ages in order to estimate values of the material properties needed for
the design of shotcrete structures.

In Chapter 5 the main topic is the mechanical and time-dependent behaviour of concrete and
shotcrete. Starting with simple uniaxial stress conditions both in compression and tension,
it is then further explained, how the material behaves under biaxial and triaxial loading
conditions, leading to smooth and convex failure surfaces in the principal stress space. Of
great interest for designing a shotcrete tunnel lining is the mechanical behaviour of sprayed
concrete at early ages. It is shown that with curing time a transition from a very ductile and
plastic material at early ages to a relatively brittle material response of hardened shotcrete
after 28 days can be expected. This transition is controlled by the increase in stiffness and
strength and a reduction in material deformability with time. Furthermore, the chapter deals
with important aspects related to time such as creep, shrinkage and hydration temperature
induced deformation. The origin of the mechanisms for creep and shrinkage is believed to lie
within the cement paste of the shotcrete and the influencing factors are discussed in detail.
From the information available in the literature it can be expected that creep, shrinkage
and thermal deformations can have an important impact on the mechanical behaviour of a
shotcrete lining during tunnel advance. In particular, it is highlighted that young shotcrete is
a very creep active medium, a fact which can lead to relatively low stresses within the tunnel
lining.

Chapter 6 aims to introduce the reader to the numerical modelling of a quasi-brittle mate-
rial such as concrete or shotcrete. At the beginning, a couple of mathematical functions for
uniaxial stress-strain curves in compression fitted to experimental data are given. After a
brief introduction into the theoretical background of elasto-plasticity, some simple constitu-
tive models that are commonly used for modelling concrete in a finite element analysis are
presented. One of these models is the Chen & Chen (1975) concrete model, which serves later
in Chapter 7 as a starting point for the development of a constitutive model for shotcrete.
Furthermore, it is discussed how to incorporate steel reinforcement in a numerical analysis
and the so called “tension-stiffening effect” describing the interaction of concrete and steel
is explained. The problems associated with modelling of cracking of concrete under tensile
stresses are emphasized and an elegant fracture energy based approach to capture the post-
peak behaviour is suggested. This chapter also deals with the modelling of creep, shrinkage
and hydration temperature induced deformations. Some rheological models that are com-
monly used for describing creep effects are presented including complex viscosity functions

32
to realistically describe the creep behaviour of young shotcrete at early ages. However, it
is shown that up to now no consistent framework exists for a simple extension of these uni-
axial creep laws into 3D. Finally, a literature review about different constitutive models for
shotcrete and their application to tunnelling available in the literature concludes this chapter.

In Chapter 7 a new sophisticated constitutive model for the behaviour of shotcrete that has
been developed in this research project is presented. The model is based on an advanced
elasto-plasticity model for hardened concrete, the so called Chen & Chen (1975) concrete
model, and includes some important modifications in order to achieve a more realistic de-
scription of the material behaviour of young shotcrete. A detailed formulation of the model
is given, discussing the involved independent yield functions for compression and tension,
plastic potentials, hardening and softening rules and material parameters needed. Further-
more, it is shown that it is possible to calibrate this developed constitutive model for sprayed
concrete against experimental data available in the literature in order to reproduce the main
features of shotcrete behaviour at early ages. The necessary equations for a robust numerical
implementation into ICFEP are given in this chapter and the model is validated through
single element runs.

In Chapter 8 the above constitutive model for shotcrete is applied to a boundary value prob-
lem simulating a typical tunnel construction in London Clay. The results for three sets of
analyses modelling different excavation sequences, i.e. full face, bench-invert and sidewall
drift, are presented. Within this thesis the focus is mainly on the mechanical performance of
the tunnel lining and through an extensive parametric study it has been possible to establish
the main influencing parameters during tunnel construction. The presented results include
surface settlements, stresses in the tunnel lining, displacements of selected points at the lining
intrados and the utilisation factor (= stress level) within the shotcrete shell.

Finally, Chapter 9 provides a summary of the whole research project and the conclusions
reached from the numerical work carried out in this thesis. It is complemented by some
recommendations for future work to be done on the topic of shotcrete used for tunnelling.

1.5 Definition of stress and strain variables

In this part of the thesis the definitions of stress and strain variables and their notation
are given. They are used for the presentation of the theoretical background of the different
constitutive models for concrete and shotcrete in Chapter 6 and 7. It should be noted that
throughout this work a tension positive sign convention has been applied.

Stress is a second order tensor, which is defined by six components and is given by the

33
following equation: ⎡ ⎤
σx τxy τxz
⎢ ⎥
σij = ⎣ τyx σy τyz ⎦ (1.1)
τzx τzy σz

In this equation σ represents a normal component of stress and τ a shearing component,

where τxy = τyx , τxz = τzx and τzy = τyz . The stress tensor can be divided into a volumetric
and a deviatoric component:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
σx τxy τxz p 0 0 σx − p τxy τxz
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
σij = ⎣ τyx σy τyz ⎦ = ⎣ 0 p 0 ⎦ + ⎣ τyx σy − p τyz ⎦ (1.2)
τzx τzy σz 0 0 p τzx τzy σz − p

 

Volumetric component Deviatoric component

In this equation p is the mean stress given as:

1
p= (σx + σy + σz ) (1.3)
3

Equation 1.2 can be rewritten in a compact form as:

σij = p δij + sij (1.4)

where δij = δji is the Kronecker delta, which is defined as being equal to +1 if i and j are
the same numbers and 0 otherwise. The deviatoric stress tensor sij represents a state of pure
shear and is expressed as:
⎡ ⎤
σx − p τxy τxz
⎢ ⎥
sij = ⎣ τyx σy − p τyz ⎦ (1.5)
τzx τzy σz − p

For materials which undergo isotropic behaviour, i.e. whose material properties are the same
in all directions, it is often convenient to work with stress invariants, which are combinations
of the different stress components. In geotechnical engineering one of these stress invariants
is the mean stress p, which has already been introduced in equation 1.3. Furthermore, the
following two stress invariants are commonly in use. The deviatoric stress invariant J is given
as:  
1
J= (σx − p)2 + (σy − p)2 + (σz − p)2 + 2τxy
2 + 2τ 2 + 2τ 2
yz zx (1.6)
2

The Lode’s angle θ is: ⎡ ⎤

⎢ √ det(sij ) ⎥
1 −1 ⎢ 3 3 ⎥
θ = − sin ⎢ 3 ⎥ (1.7)
3 ⎣ 2 1 2 ⎦
(sij sij )
2

34
where det(sij ) is the determinant of the deviatoric stress tensor and is obtained as:
 
 σx − p τxy τxz 
 
 
det (sij ) =  τyx σy − p τyz  (1.8)
 
 τzx τzy σz − p 

If the coordinate axes are chosen to coincide with the principal stress axes, the stress tensor
reduces to: ⎡ ⎤
σ1 0 0
⎢ ⎥
σij = ⎣ 0 σ2 0 ⎦ (1.9)
0 0 σ3
and the three invariants are given as:

1
p = (σ1 + σ2 + σ3 ) (1.10)
3

1
J=√ (σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2 (1.11)
6
  
−1 1 (σ2 − σ3 )
θ = tan √ 2 −1 (1.12)
3 (σ1 − σ3 )

The geometrical significance of these stress invariants p, J and θ, is illustrated for the principal
stress space in Fig. 1.2. Furthermore, the space diagonal (σ1 = σ2 = σ3 ) and a deviatoric
plane, defined as any plane perpendicular to the space diagonal, are indicated. It can be
seen that the mean stress p is a measure of the distance of the current deviatoric plane
from the origin along the space diagonal. The deviatoric stress invariant J measures the
distance of the current stress state from the space diagonal in the deviatoric plane. Finally,
the Lode’s angle θ defines the orientation of the stress state within the deviatoric plane. It
varies between +30 , which corresponds to triaxial extension (σ1 = σ2 ≥ σ3 ), and -30 for
triaxial compression (σ1 ≥ σ2 = σ3 ).

35
Figure 1.2: a) Principal stress space and b) deviatoric plane (from Potts & Zdravković, 1999)

It should be noted, that usually, when dealing with soil modelling the above introduced
stresses and invariants are given as effective stresses in order to consider the pore water
pressure u by:

σij = σij + σf (1.13)
 the effective stress tensor and σ is the vector of pore
where σij is the total stress tensor, σij f
fluid (water) pressure given as:

σf = (u u u 0 0 0 )T (1.14)

In structural engineering a different set of stress invariants can often be encountered (Chen,
1982). For the general case the three invariants of the stress tensor can be written as:

I1 = σ x + σ y + σ z (1.15)
I2 = (σx σy + σy σz + σz σx ) − τxy − τyz − τzx
2 2 2
(1.16)
 
 σx τxy τxz 
 
 
I3 = det (σij ) =  τyx σy τyz  (1.17)
 
 τzx τzy σz 

which take the following form for the principal stress space

I1 = σ 1 + σ 2 + σ 3 (1.18)
I2 = (σ1 σ2 + σ2 σ3 + σ3 σ1 ) (1.19)
I3 = σ 1 σ 2 σ 3 (1.20)

36
In a similar way, the invariants of the deviatoric stress tensor sij can be derived as:

J1 = sii = sx + sy + sz = 0 (1.21)
1 
J2 = (σx − σy )2 + (σy − σz )2 + (σz − σx )2 + τxy 2 + τyz 2 + τzx 2 (1.22)
6
 
 sx τxy τxz 
 
 
J3 = det (sij ) =  τyx sy τyz  (1.23)
 
 τzx τzy sz 

If the cartesian coordinate axes x, y and z coincide with the principal directions, the deviatoric
invariants can be written as follows:

J1 = s1 + s2 + s3 (1.24)
1 2  1 
J2 = s1 + s22 + s23 = (σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2 (1.25)
2 6
1 3 
J3 = s + s32 + s33 = s1 s2 s3 (1.26)
3 1

From these invariants the Lode’s angle used in structural engineering can be obtained as:
 √ 
1 3 3 J3
θst = arccos √ (1.27)
3 2 ( J2 )3

For the mathematical description of a failure criterion for concrete, special attention is usually
given to the three independent invariants I1 , J2 and J3 (or θst ), which are of first, second
and third degree in stress, respectively.

Kotsovos & Newman (1978) used in the formulation of their non-linear constitutive model
for concrete (see Chapter 6) octahedral stresses which are defined as follows:

1
σoct = (σ1 + σ2 + σ3 ) (1.28)
3

and
1 
2
τoct = (σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2 (1.29)
9
From the above definitions for the different stress invariants it can be summarized that:

1
p = σoct = I1 (1.30)
3

and 
3 
J= τoct = J2 (1.31)
2

37
 Strain is like stress a second order tensor defined as well by six components:
⎡ 1 1
⎤
εx 2 γxy 2 γxz
⎢ ⎥
εij = ⎣ 1
2 γyx εy 1
2 γyz ⎦ (1.32)
1 1
2 γzx 2 γzy εz

where γxy = γyx , γxz = γzx and γyz = γzy . The strain tensor can be divided into two compo-
nents. The volumetric and the deviatoric components are given in the following equation:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
εx 1
2 γxy
1
2 γxz ev 0 0 εx − ev 1
2 γxy
1
2 γxz
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
εij = ⎣ 1
2 γyx εy 1
2 γyz ⎦ = ⎣ 0 ev 0 ⎦ + ⎣ 1
2 γyx εy − ev 1
2 γyz ⎦
1
2 γzx
1
2 γzy εz 0 0 ev 1
2 γzx
1
2 γzy εz − ev

 

Volumetric component Deviatoric component
(1.33)
In the above equation ev is equal to:

1 1
ev = (εx + εy + εz ) = εvol (1.34)
3 3

with εvol being the volumetric strain. Equation 1.33 can expressed in a different form as:

1
εij = εvol δij + eij (1.35)
3

where eij is the deviatoric strain tensor. It equals:

⎡ ⎤
εx − ev 1
2 γxy
1
2 γxz
⎢ ⎥
eij = ⎣ 1
2 γyx εy − ev 1
2 γyz ⎦ (1.36)
1
2 γzx
1
2 γzy εz − ev

The correspondent strain invariants to the earlier introduced stress invariants are the volu-
metric strain εvol , presented in equation 1.34, and the deviatoric strain Ed , which is given
as:   
1 1 1
Ed = 2 (εx − ev )2 + (εy − ev )2 + (εz − ev )2 + γxy
2 + γ2 + γ2 (1.37)
2 2 yz 2 zx
In terms of principal strains ε1 , ε2 and ε3 the strain invariants are given as follows:

εvol = ε1 + ε2 + ε3 (1.38)

and
2 
Ed = √ (ε1 − ε2 )2 + (ε2 − ε3 )2 + (ε3 − ε1 )2 (1.39)
6

38
1.6 Terminology and symbols used in thesis
When dealing with the academic literature concerning the material behaviour of concrete
and shotcrete and its modelling for tunnelling one comes across many distinct approaches
that differ not only in their basic assumptions and ideas but as well in terms of terminology,
used symbols and graphical representations. Hence, when writing this thesis, from a scientific
point of view, it was relatively difficult to find a consistent and clear formulation of all the
concepts, models and principles that are presented within this work.

In this section the aim is to introduce the adopted terminology used throughout this
thesis and provide some further information if a different notation or important symbols are
encountered, a fact which is almost unavoidable. It should enable the reader to understand
all the text, figures and tables that are presented in the following chapters.

à Chemical affinity
Ae Area of finite element
c Cohesion
D Damage parameter
[D] Constitutive matrix
d Displacement (Chapter 3)
dmax Maximum aggregate size
eij Deviatoric strain tensor
ΔE Incremental total potential energy
E Young’s modulus of elasticity
E28 Young’s modulus of elasticity for hardened shotcrete at 28 days
Ec Young’s modulus of elasticity (Chapter 4)
Eci Young’s modulus of elasticity (= E)
Ecc Young’s modulus of elasticity (= E)
Ecm Secant Young’s modulus of elasticity (at 0.4 fcm ) (CEB-FIP Model Code,
1990)
Ec1 Secant Young’s modulus of elasticity from origin to the peak compressive
strength (CEB-FIP Model Code, 1990)
Ed Deviatoric strain
Ek Stiffness of Kelvin spring
Em Stiffness of Maxwell spring
fc Uniaxial compressive stress
f Uniaxial compressive stress (=fc )
fcp Uniaxial compressive strength
fo Uniaxial compressive strength (=fcp )
fcp,28 Uniaxial compressive strength of hardened shotcrete at 28 days
fcp,1 Uniaxial compressive strength of shotcrete at 1 day

39
fc Uniaxial compressive strength (= fcp )
fcc Uniaxial compressive strength (= fcp )
fcube Uniaxial compressive cube strength (= fcp )
fck Characteristic uniaxial compressive strength, i.e. strength below which 5 % of
all possible strength measurements for a specified concrete may be expected
to fall (CEB-FIP Model Code, 1990)
fck,c Characteristic confined compressive strength (EC 2, 2004)
fcd Uniaxial compressive design strength (EC 2, 2004)
fcm Mean value of the uniaxial compressive strength, fcm = fck +8 MPa (CEB-FIP
Model Code, 1990)
fcy Uniaxial compressive yield stress
28
fcy Uniaxial compressive yield stress for hardened shotcrete at 28 days
fbc Biaxial compressive strength
fbcy Biaxial compressive yield strength
ft Unixial tensile stress
ftp Uniaxial tensile strength
fct Uniaxial tensile strength (= ftp )
fctm Mean value of the uniaxial tensile strength (= ftp ) (CEB-FIP Model Code,
1990)
fct,sp Mean value of the splitting tensile strength (CEB-FIP Model Code, 1990)
fct,f l Mean value of the flexural tensile strength (CEB-FIP Model Code, 1990)
Gf Fracture energy
Gf 1 Fracture energy of plain concrete
Gf 2 Fracture energy related to the contribution of steel fibres
Gc Fracture energy in compression
Gf o Base value of fracture energy
Grc
f Fracture energy of reinforced concrete
Gff r Fracture energy of fibre reinforced concrete
GSF RS Fracture energy of fibre reinforced shotcrete
G Shear modulus
HM E Hypothetical Modulus of Elasticity
I1 , I2 , I3 Stress invariants in structural engineering
J Deviatoric stress invariant
J1 , J2 , J3 Invariants of deviatoric stress tensor in structural engineering
[J] Jacobian matrix
[KE ] Element stiffness matrix
[KG ] Global stiffness matrix
K Bulk modulus
Kf Bulk modulus of the pore fluid
Ko Coefficient of earth pressure at rest

40
Kr Spring stiffness
Kskel Bulk modulus of the soil skeleton
ΔL Incremental work done by applied loads (Chapter 3)
leq Equivalent length of finite element
rc
leq Equivalent length of finite element for reinforced concrete
ls Average crack spacing
M Bending moment
N Hoop force
[N ] Matrix of shape functions
nint Number of integration points per element (= nIP )
p Mean stress
p Internal wall pressure (Chapter 2)
po In-situ isotropic stress in the ground before tunnel excavation
peq Equilibrium pressure acting on tunnel lining
r Degree of hydration (= ξ)
{ΔR} Right hand side load vector
RH Relative humidity
Rdf c Reduction of compressive strength to be gained between t1 and t2 due to
preloading
sij Deviatoric stress tensor
S, T Natural coordinates (Chapter 3)
T Temperature
u Radial displacement (Chapter 2)
u, v Horizontal and vertical displacements (Chapter 3)
uin Initial radial wall displacement
V (σ, t) Non-linear deformation modulus
Ve Volume of finite element (Chapter 5)
Ve Excavated tunnel volume (Chapter 2)
VL Volume loss
VS Volume of the surface settlement trough
Wf Fibre content
ΔW Incremental strain energy
w Crack opening
wu Complete crack opening
wc Complete crack opening (= wu )
x, y Coordinates
αt Coefficient of thermal expansion
β Stiffness reduction factor
βp Uniaxial compressive strength (= fcp )
δij Kronecker delta

41
εc Uniaxial compressive strain
εcp Uniaxial compressive peak strain
εcr Creep strain
εo Uniaxial compressive peak strain (=εcp )
εc1 Uniaxial compressive peak strain (=εcp ) (EC 2, 2004)
εc2 Uniaxial compressive peak strain (=εcp ) (EC 2, 2004)
εc2,c Compressive peak strain for confined stress conditions (EC 2, 2004)
εu Ultimate uniaxial compressive strain
εcu1 Ultimate uniaxial compressive strain (EC 2, 2004)
εcu2 Ultimate uniaxial compressive strain (EC 2, 2004)
εcu2,c Ultimate compressive strain for confined stress conditions (EC 2, 2004)
εcco Uniaxial compressive peak strain (=εcp )
εij Strain tensor
εtp Uniaxial tensile peak strain
εcto Uniaxial tensile peak strain (=εtp )
εp Plastic strain
εv Effective strain
εvol Volumetric strain
η Safety factors
ηk Viscosity of Kelvin dashpot
ηm Viscosity of Maxwell dashpot
κ Internal damage parameter (Chapter 5)
λ Stress relief factor
μ Poisson’s ratio
μcr Creep Poisson’s ratio
ν Poisson’s ratio (= μ)
νcc Poisson’s ratio (= μ)
φ Friction angle
φ Surface orientation (Chapter 4)
φ(t, to ) Creep coefficient
ρc Compressive utilization factor
ρt Tensile utilization factor
σf Vector of pore water pressure
σij Stress tensor
σc Uniaxial compressive stress (= fc )
σc Uniaxial compressive strength (Chapter 4) (= fcp)
σct Uniaxial tensile stress (= ft )
σoct Octahedral stress
σr Radial stress on lining

42
σt Uniaxial tensile stress (= ft )
σ̄Re Uniaxial tensile stress (= ft )
σv Effective stress
τoct Octahedral stress
θ Lode’s angle
θst Lode’s angle in structural engineering
{ψ} Residual load vector (Chapter 3)
ξ Degree of hydration

43
Chapter 2

Tunnelling in urban areas

2.1 Introduction
This chapter aims to introduce the reader to various aspects related to the challenging field
of tunnelling in soft ground conditions which can be regularly encountered in urban areas.
Since shotcrete is one of the main support elements of the New Austrian Tunnelling Method
(NATM), the historical background and the main concepts behind this special tunnelling tech-
nique will be briefly highlighted. Although NATM should not be regarded as a particular
construction method for tunnels, the main steps of the construction procedure for tunnels in
urban areas applying shotcrete for temporary support are presented. Furthermore, geotech-
nical observation and monitoring are of great importance for a safe assessment of the system
behaviour during tunnel advance. Several techniques for measuring displacements of and
stresses in a shotcrete tunnel lining and the difficulties associated with the correct interpreta-
tion of the obtained results are discussed. A review of tunnel lining design approaches that are
common in engineering practice, including empirical, analytical and numerical procedures,
are given. The single-shell method as a new economic tunnelling technology is introduced,
describing briefly some innovative trends for sprayed concrete lined tunnels that have been
developed recently. Finally, the presentation of some modelling techniques to account for 3D
effects in a 2D plane strain analysis concludes this chapter.

2.2 The New Austrian Tunnelling Method

The New Austrian Tunnelling Method (NATM) is a worldwide recognized tunnelling tech-
nique developed during 1957 - 1965, where Austrian engineers (Rabcewicz, Pacher, Müller)
have taken a decisive part in its development. It has its origin in rock tunnelling, where
tunnel construction under heavy rock pressure could successfully be handled by employing
the surrounding rock as a part of the support system, which is believed to be one of the main
concepts of the NATM. However, in the tunnelling industry confusion still exists regarding a
successful and safe application of NATM for tunnnelling in soft ground conditions since there

44
is a lack of understanding of the essential features and concepts behind NATM. In their paper
Karakus & Fowell (2004), try to give an insight into the New Austrian Tunnelling Method by
investigating its historical background, main concepts and design philosophy. Furthermore,
they raise important questions such as “What is NATM?” and “Is NATM a tunnelling tech-
nique or a design philosophy?”. Some definitions and ideas found in the literature describing
the NATM are presented below.

The original explanation of the method, given by one of the principal inventors of NATM,
is stated in Rabcewicz (1964) as:

. . . a new method consisting of a thin sprayed concrete lining, closed at the earliest possible
moment by an invert to a complete ring - called an “auxiliary arch” - the deformation of
which is measured as a function of time until equilibrium is obtained.

In this publication the use of shotcrete as a ground support element is emphasised with its fea-
ture against loosening and stress-rearrangement pressures through an immediate application
after opening and the interaction with the neighbouring rock highlighted (see Fig. 2.1).

Figure 2.1: New design philosophy for tunnels according to NATM (from Müller & Fecker,
1978)

As the popularity of the technique grew and it began to be practised more widely, some
confusion developed as to what was meant by NATM. Therefore the ÖIAV (1980) published a
definition in cooperation with the International Tunnelling Association (ITA) in 10 languages,
saying:

The New Austrian Tunnelling Method constitutes a method where the surrounding rock or
soil formations of a tunnel are integrated into an overall ring-like support structure. Thus
the formations will themselves be part of this supporting structure.

One of the other Austrian advocates, Prof. Müller, contributed to the international discussion
trying to overcome misunderstandings of the NATM by publishing the following definition in
Müller (1990):

45
 The NATM is rather a tunnelling concept with a set of scientifically established principles
and ideas which the tunneller tries to follow and should not even be called a construction
method, since this implies a method of driving a tunnel.

Furthermore, he summarized the most characteristic features of the NATM in a list of 22

principles. The most important of these are as follows:

1. The main load bearing component of a tunnel is the surrounding rock mass. Preliminary
support and final lining have the function of establishing a load-bearing ring or a three-
dimensional spherical bearing shell in the rock mass.

2. Additional support elements should be used to preserve the support resistance of the
rock mass and therefore prevent loosening and extensive rock deformations.

3. The thickness of the shotcrete layer should be as thin and flexible as possible and
additional strengthening should be obtained by using mesh reinforcement, tunnel ribs
and anchors, rather than thickening the lining.

4. The ring closure time is of crucial importance from a mechanical point of view and
should be done as soon as possible.

5. In order to optimize the formation of the ground ring, preliminary laboratory tests and
deformation measurements in the tunnel should be carried out.

However, his conclusion about the rapid ring closure time in deep tunnels to minimise dis-
placements was not agreed by other engineers (Rabcewicz, Golser). A graphical simplified
representation of the principles and effects of NATM can be seen in Fig. 2.2.

Figure 2.2: Simplified definition of principles and effects of NATM (from Sauer, 1988)

46
Summarising, the following major principles of NATM can be derived from ICE (1996) and
HSE (1996):
1. The inherent strength of the soil or rock around the tunnel domain should be preserved
and deliberately mobilised to the maximum extent possible.

2. This mobilisation can be achieved by controlled deformation of the ground although

excessive deformations which will result in loss of strength or high surface settlements
must be avoided.

3. Initial and primary support elements consisting of systematic rock bolting or anchoring
and thin flexible shotcrete linings are used to achieve the particular purposes given in
(2). Permanent support works are usually carried out at a later stage.

4. Laboratory tests and monitoring of the deformations of supports and ground should be
carried out before and during construction.

5. The responsable engineers involved in the execution, design and supervision of NATM
construction must understand and accept the NATM approach and react co-operatively
on resolving problems.

6. The length of the unsupported span should be left as short as possible.

Having these main principles of NATM in mind and involving the boundary conditions of
each tunnel project (ground conditions, geometrical conditions, settlement restrictions, etc.),
one can perform the design of the structural elements and a specific construction process.
Nevertheless, it is always very important to distinguish between NATM tunnelling in rock
and in soft ground formations, as some points of the above mentioned concept may not be
applicable for one or the other ground condition. For this reason, tunnels in soft ground
conditions constructed with shotcrete as a main support element are often referred to as
“sprayed concrete lined tunnels”, or shortly SCL tunnels (Thomas, 2009).

For tunnelling in urban areas limiting surface settlements is one of the main concerns in
order to avoid damage to overlying structures. For achieving this, according to ICE (1996)
the following principal measures must be undertaken:
 Excavation stages must be sufficiently short, both in terms of dimensions and duration.

 The closure of the sprayed concrete ring must not be delayed.

The flexibility of NATM in soft ground conditions is limited to a certain extent, since design
details about the primary support are determined by the designer and then not usually
varied. The only aim of tunnel instrumentation and monitoring is therefore to validate the
anticipated design without changing it during tunnel advance. It can be concluded that
tunnelling in urban areas often follows the construction techniques usually associated with
NATM, but does not necessarily employ the entire concept and principles of the New Austrian
Tunnelling Method.

47
2.3 Typical tunnel construction process in urban areas
For NATM tunnels in urban areas, construction usually starts from a previously built vertical
shaft, which serves as access for personnel and machines and for the removal of excavated
material. If ground conditions permit, the full tunnel face is excavated, which is suitable
for tunnels up to 30 m2 of face area (ICE, 1996). For tunnelling in overconsolidated fissured
clays, such as London Clay, vertical face excavation is usually limited to tunnel diameters of
4 m. If necessary, a full face advance with an inclined face of 60 to 70 is applied. A similar
stabilising effect can be achieved by dividing the cross section into a number of small faces
- typically crown, bench and invert. Different partial face excavation techniques can be seen
in Fig. 2.3.

Figure 2.3: Different excavation sequences for NATM tunnel construction

Excavation of the tunnel is incrementally advanced in rounds with a variable length of

about 0.5 to 1.5 m, 1 m being a common target (see longitudinal tunnel section in Fig. 2.4). It
has been shown that the advance length has a significant influence on stability and settlement
behaviour and keeping the support close to the face results in an improvement of settlement
control. Directly after excavation shotcrete is sprayed at a high pressure on to the tunnel
walls. Generally a 50 mm sealing layer is first applied followed by two layers, each reinforced
by steel mesh and forming a typical NATM lining thickness of 20 to 40 cm. As an additional
support element, steel lattice girders can be incorporated into the lining to provide some
extra stiffness to the support system. They are also used for the correct profiling of the cross
section and to achieve the correct shotcrete thickness. Another important factor for soft
ground tunnelling at shallow depths is the advance rate and the ring closure times, which
appear to vary between 8 and 24 hours from completion of an excavation advance (ICE,
1996). The secondary or final tunnel lining formed of conventional cast concrete is usually
installed at a later stage of construction.

48
Figure 2.4: Longitudinal tunnel section showing excavation rounds for a bench-invert con-
struction scheme (from ICE, 2004)

2.4 The importance of monitoring in tunnelling

As mentioned earlier, observation of the tunnel behaviour during construction using geotech-
nical measurements is a key element of the NATM and serves as an assessment of the stability
of the whole structure. Therefore, extensive instrumentation is nowadays common practice
for underground projects, but difficulties remain with the right interpretation of all the mea-
sured field data (Rokahr et al., 2005).

Clayton et al. (2000) highlight four principal purposes of field instrumentation and mon-
itoring for shallow shotcrete tunnels in soft ground conditions, which are:

1. Examination of the actual performance and safety of the soil-primary lining system and
to assess whether the design is valid.

2. Detection of anomalous behaviour and to provide information upon which remediation

in response to unforeseen conditions or behvaiour can be reliably and safely based.

3. To determine whether design modifications, particularly strengthening works or con-

struction details such as advance length, are necessary.

4. To control the day-to-day application of compensation grouting which is commonly

used to reduce surface settlements.

The type of instrumentation used during the construction of a tunnel depends highly on the
objectives of the monitoring programme, which can be different for each individual project.

49
Therefore, a detailed monitoring scheme should be set up that meets the prescribed require-
ments. For the correct choice of an instrument it is important to have in mind its performance
under site conditions and the demanded accuracy of certain parameters to be measured. An
example of such a detailed monitoring programme at Heathrow Express Terminal 4 can be
found in Clayton et al. (2006). The following two sections of this chapter focus mainly on the
monitoring of the shotcrete tunnel lining and the right interpretation of the measured data.

2.4.1 Measuring and interpretation of displacements

Tunnel wall convergence between reference points bolted on the tunnel lining is usually mea-
sured with standard metal tape extensometers having an accuracy of ±0.2 mm for distances
of up to 10 to 15 m (Kavvadas, 2003). However, in most tunnelling projects deformations
of the lining are measured in three dimensions by the use of geodetic surveying with total
stations and integrated distance measurement. For such applications, optical targets are in-
stalled, as illustrated in Fig. 2.5, at regular distances along the tunnel axis, i.e. at sections
every 5 to 15 m, showing an accuracy of ±2 − 3 mm (Jones et al., 2008). One big advantage
of this method is that readings of absolute, three dimensional movements can be made from
an observation station located outside the area of intensive construction activity (Clayton
et al., 2000).

Figure 2.5: Typical arrangement of convergence targets for tunnel lining (from Jones et al.,
2008)

One big problem in measuring lining displacements is the start of monitoring, which
usually happens a certain amount of time after installation of the shotcrete. However, most
of the stress rearrangement and tunnel lining displacements occur before ring closure, since
the incomplete ring made of young shotcrete represents a much weaker bearing system than
the closed ring, which acts from the structural point of view as a tube. Therefore, some
practical diffulties still exist in gaining satisfactory access to the congested working area at
the face in order to obtain early displacement measurements that would give a considerable
insight into the behaviour of the tunnel.

Once the displacements of the lining are monitored, engineers on site are left with a large

50
amount of information that has to be carefully interpreted. Usually displacement histories,
deflection curves or cross-sectional displacement vectors should enable the responsible engi-
neers to understand the system behaviour during tunnel advance, and react accordingly if
critical target values are reached (Schubert & Grossauer, 2004). The assessment of the stabil-
ity of a tunnel with the help of displacement monitoring is a difficult task, as highlighted by
Rokahr et al. (2005). Fig. 2.6 illustrates the need for a proper understanding of the complete
system behaviour by showing various possible displacement patterns of a roof lining. In case
A, the tunnel crown and the left and right footings indicate the same amount of displacement
and therefore a translation of the lining takes place which leaves the shotcrete unstressed.
In case B, the sprayed concrete lining is only subjected to pure hoop forces. In case C,
only the tunnel crown settles a certain amount with the footings being unchanged and this
movement introduces a combination of hoop forces and bending in the lining. Finally, case
D is dominated mainly by bending stress. It can be concluded that the same degree of crown
settlement can be associated with at least four different stress states within the lining (Rokahr
et al., 2005). The displacement pattern of one single point on the tunnel lining is not decisive
and it is crucial to consider the displacement combinations of various target points along the
tunnel cross section.

Figure 2.6: Deformation examples of a roof lining (from Rokahr et al., 2005)

2.4.2 Measuring and interpretation of stresses in tunnel linings

With the increasing use of numerical methods for the design of tunnel linings relatively good
agreement can be achieved between predicted and measured displacements. However, this
fact often leads to the misleading assumption that stresses within the shotcrete lining have
also been well predicted. In reality, due to the complexity of the loading conditions and
the material behaviour of shotcrete at early ages, very little reliable information is available
regarding the stress state during tunnel advance. The correct determination of stresses in a
tunnel lining is of great importance for estimating the factor of safety and in order to verify
that design predictions are reasonable (Thomas et al., 2004).

In the past, several intrusive techniques, such a slot cutting and over- or undercoring, have
been successfully used to evaluate stress conditions in shotcrete linings (Clayton et al., 2000).
However, these methods provide just one-off measurements and are disruptive to tunnelling

51
operations and damaging to the lining. A continuous temporal stress variation in the lining,
especially in the first 2 to 3 days, is of much greater value for engineers in order to investigate
the safety of the tunnel. Two common approaches for achieving this are:

 Pressure cells

 Stress back-calculation from deformation measurements

Usually back-calculation of the stress history in a shotcrete shell is based on the use of strain
gauges and rheological models for the mathematical description of the material response. A
large number of assumptions has to be introduced in order to capture the complete behaviour
of shotcrete at early ages. One widely used technique is the rate of flow method proposed
by England & Illston (1965), which will be described later in Chapter 6. However, in the
literature it is often stated that an accurate estimation of the stresses within the tunnel lining
by back-calculation is not possible due to the large number of errors introduced in the course
of deformation measurements and the calculation procedure.

The performance of pressure cells for sprayed concrete linings was assessed in detail in
Clayton et al. (2002). Basically, they can be embedded within the shotcrete layer in two
different orientations: Radial pressure cells record the stress between the sprayed concrete
and the surrounding ground of the tunnel and tangential pressure cells measure the hoop
stress within the tunnel lining itself. Examples of such pressure cells can be seen in Fig. 2.7.

Figure 2.7: Radial and tangential pressure cells (from Clayton et al., 2002 and Jones et al.,
2005)

However, a correct recording of the stress history in any medium, such as shotcrete in
tunnelling, is difficult due to various factors that influence the performance of a pressure cell.
They can be summarized as:

 Installation effects (i.e. cavities around the cell during shotcreting, unwanted rotations,
etc.)

 Temperature sensitivity during hydration of the cement paste

 Offsets due to crimping

 Cell properties (cell fluids such as mercury or oil)

52
The so called Cell Action Factor (CAF) representing the ratio between the recorded pressure
and the actual stress in the shotcrete can be used to evaluate how the cell properties affect
the interaction of the pressure cell with the surrounding medium. Usually the CAF takes
a value close to unity, with an average value of 0.95 (Jones et al., 2005). Some researchers
claim that embedded pressure cells are not very reliable for monitoring the actual stress in a
tunnel lining. However, they still represent a valuable source of information that can be used
to assess whether the anticipated tunnel design assumptions are reasonably justified.

2.5 A review of tunnel lining design

The design of a tunnel is a complex process that involves an assessment of the mechanisms
of behaviour of the whole structure during tunnel advance, the principal risks and serves as a
basis for interpreting results from monitoring (ICE, 2004). However, geological uncertainties
in the material properties of the surrounding ground make this process a difficult task to
perform. One of the main difficulties is to estimate the loads that act on a tunnel lining in
terms of earth pressures and ground water. Given these uncertainties, a design analysis should
never be accepted as a definite solution and the sensitivity of the ground-support interaction
should be investigated in detail. Furthermore, it should be noted that any design analysis
is just an approximation of reality, including several assumptions and probably errors too.
Woods & Clayton (1993) mention six sources of errors in modelling of a complex structure
such as a tunnel, which are:

 Modelling the geometry of the problem

 Modelling the construction method

 Constitutive modelling of the involved materials and parameter selection

 Theoretical basis of the solution method

 Interpretation of results

 Human errors

Up to now various design methods for tunnels exist in engineering practice, each of them hav-
ing its advantages and disadvantages. Two main groups of approaches can be distuingished -
design methods for “continua” such as encountered in soft ground and/or massive rock, and
design methods for “discontinua” when dealing with a jointed rock mass. The most impor-
tant tunnel design methods will be introduced briefly in the following sections. For further
detailed information on tunnel design see ICE (2004).

2.5.1 Empirical methods

Empirical methods of tunnel design have been successfully applied mainly in rock tunnelling
and to a minor extent in soft ground conditions. They are usually based on assessments of

53
precedent practice and their support recommendations have been calibrated for a wide range
of tunnelling conditions. The two most frequently used empirical design methods that can
be found in the literature are:

 Rock Mass Rating (RMR) by Bieniawski (1994)

 Q - Systems by Barton et al. (1974)

Both systems have in common that with the help of certain rock mass parameters, such as
strength of the rock, rock quality (weathering), joints, number of sets, frequency, spacing
and ground water conditions, a rock mass classification can be established. A combination
of these parameters then leads to the support measures being determined from design charts
or tables. Obviously, such methods depend highly on the experience of the tunnel engineer
or engineering geologist responsible for the rock mass classification and should be used with
caution. Some other disadvantages of these empirical methods have been highlighted in the
literature (no guidance on the timing of support installation, no consideration of the effects
to adjacent structures, factor of safety unknown, etc.). However, the main advantage lies in
their simplicity and such empirical methods are therefore suitable for feasibility studies at
the concept design stage.

2.5.2 Closed-form analytical methods

Analytical or structural design models based on closed-form solutions were widely in use in
the past for the planning of tunnels in soft ground conditions and allowed the dimensioning
of tunnel linings in a simple and robust way. A comprehensive review of analytical design
models for tunnels can be found in Duddeck & Erdmann (1982) and Duddeck & Erdmann
(1985). However, most of these models are not capable of modelling the full complexity of
a tunnel during construction such as soil deformations ahead of the face, stress relief prior
to the installation of support elements or the soil-sructure interaction. The most common
assumptions that are made in such an analytical analysis are:

 Circular geometry of tunnel

 2D plane strain conditions

 Stresses acting on tunnel lining are equal to primary stresses in the undisturbed ground

 Bond between lining and ground for radial and tangential deformations

 Tunnel lining and ground (as a homogeneous material) behave linear elastically

In some cases the principle of superposition is applied when investigating adjacent structures
or tunnels, but obviously these analytical models fail to realistically capture the interaction
problem of twin-tunnel construction. However, despite these shortcomings, structural models
for tunnel lining design can still be a useful tool for a rough design of the involved structural
elements at the concept design stage.

54
2.5.2.1 Continuum analytical methods

Continuum analytical models are commonly based on excavation and lining of a hole in a
stressed continuum (ICE, 2004). Fig. 2.8 shows such a continuum model found in Duddeck &
Erdmann (1982) for which Engelbreth (1961) derived a closed form solution for the internal
forces and deformations of the lining. Further examples can be found in Muir Wood (1975),
Einstein & Schwartz (1979) and Duddeck & Erdmann (1985).

Figure 2.8: Plane strain continuum model and charactersitic distribution of radial displace-
ments u, radial stresses on the lining σr , hoop forces N and bending moments M (from
Duddeck & Erdmann, 1982)

Since the lining is assumed to be installed immediately before excavation, this fact tends
to overestimate the loads acting onto the lining. Some modifications of these models some-
times include a stress relief factor in order to account for a certain stress relaxation before
lining installation. Another important fact is that continuum analytical models treat the
surrounding ground as a semi-infinite medium and therefore they should only be used for
tunnels where the axis is deeper than two tunnel diameters below the surface (ICE, 2004).
This restriction implies some limitations for the use of such models for the design of very shal-
low tunnels in urban areas. Their benefit is obviously their simplicity and they provide the
designer quickly with information on the maximum deformations, normal forces and bending
moments in the tunnel lining.

2.5.2.2 Convergence-confinement method

The convergence-confinement method for a circular tunnel in an isotropic axisymmetric stress

field is graphically illustrated in Fig. 2.9, which contains a graph of the internal wall pressure
p versus the radial displacement u. From this figure, the displacement and the load acting on
the tunnel support are obtained in principle through the intersection of the ground reaction
curve of the tunnel and the support reaction line (Oreste, 2003).

55
 Figure 2.9: The convergence-confinement method (after Oreste, 2003)

po is the in-situ isotropic stress in the ground before tunnel excavation. The reaction line of the
support element (i.e. shotcrete) is defined by its stiffness k and the initial wall displacement
on installation of the support uin , which can be seen as a measure of the distance from the
tunnel face where the support is installed. In the case of a linear elastic perfectly plastic
support material a maximum stress pmax and displacement umax can be introduced. At the
above mentioned intersection point of the ground reaction curve and the support reaction
line (i.e. point A in the above figure), the system is assumed to be in equilibrium and the
pressure acting onto the lining peq and the displacement ueq can be obtained. Due to the
assumed axisymmetry for ground behaviour and tunnel geometry, this method is valid only
for ground conditions where the coefficient of earth pressure at rest Ko is close to 1.0 and
for deep tunnels. Another drawback is that no information is given on the distribution of
bending moments in the lining.

2.5.2.3 Bedded-beam-spring models

When designing a tunnel with a bedded-beam-spring model the tunnel lining is simulated
as a structural beam element. Furthermore, this beam is attached to the ground which is
represented by radial and/or tangential springs, as can be seen in Fig. 2.10. These springs
are usually limited to acting only in compression, allowing separation of the lining from the
soil formation in the expected area of tensile stresses (tunnel crown). Some uncertainties
exist regarding an appropriate estimation of the spring stiffness Kr , which in theory can be
derived from standard ground investigation tests (ICE, 2004). A critical point in the use of
a bedded-beam-spring model is the correct application of the loads at each spring position
which may be estimated as a percentage of the vertical and horizontal overburden pressure.
Complete closed-form solutions for this model can be found in Schulze & Duddeck (1964).

56
Figure 2.10: Bedded ring model without bedding near the tunnel crown (from Duddeck &
Erdmann, 1982)

Recently, bedded-beam-spring models have been widely superseded by more complex

numerical methods for tunnel lining design and are rarely in use for the analysis of temporary
lining support made of shotcrete. However, they can be useful in the design of final linings
where long-term full overburden loading conditions are appropriate (ICE, 1996).

2.5.3 Numerical methods

Over the last few decades significant progress has been made in the development of numerical
methods for the design of underground structures such as tunnels, replacing almost entirely
the previously described simple design methods. The finite element method as one particular
type of numerical analysis is capable of simulating complex tunnel construction by taking
into account the following aspects:

 Complex tunnel geometry

 Different geological strata

 Advanced constitutive behaviour of ground and supporting materials

 Construction sequences

 Complex boundary conditions

 Ground-support interaction

 Incorporation of adjacent structures (tunnels, buildings, etc.)

Sophisticated 2D and 3D analyses of tunnel construction can be performed capturing the

stress-rearrangement in the ground due to tunnel excavation in a realistic way. However,
despite all these advantages, results from a numerical analysis should be assessed in the con-
text of the quality of the site investigation, the estimated range of geomechanical properties

57
and the solution algorithms in the adopted code (ICE, 2004). Numerical methods can be
seen as a very powerful tool in the design process providing a detailed understanding of the
mechanical behaviour of the tunnel during construction.

2.5.3.1 Bearing-capacity-diagram

Once a numerical analysis has been carried out (i.e. finite element analysis), tunnel design
engineers are left with a large number of stresses or structural forces and displacements of
the structural elements. In the next step these results can be used to dimension the applied
support elements, i.e. shotcrete lining, anchors and steel arches. However, this process is
not straightforward, since the consideration of safety factors is not very clear in the case of
underground construction and standard codes for surface structures are often not applicable.
Therefore, a consistent design code for shotcrete tunnel linings is still lacking in the tunnelling
industry.

Sauer et al. (1994) and Wu & Roony (2001) proposed a method to dimension the primary
shotcrete lining with the help of the so called “bearing-capacity-diagram”. It is based on
the following equilibrium condition between the external forces (action) and internal forces
(resistance) for bending moment and thrust:

ηn Ne = ηs Nis + ηc Nic (2.1)

and
ηm Me = ηs Mis + ηc Mic (2.2)

where Ne and Me are the (external) thrust and bending moment obtained from the finite
element analysis and Nis , Mis , Nic and Mic are the (internal) thrust and moment in the
ultimate limit state for the steel and concrete. The subscripts s and c refer to steel and
concrete respectively. The possible safety factors for the loads (ηn and ηm ) and material side
(ηs and ηc ) depend on the applied code and can be taken as an example according to EC 2
or the German standard DIN 1045.

From the above equations the allowable axial thrust Na and bending moment Ma can be
obtained as:
1
Na = (ηs Nis + ηc Nic ) (2.3)
ηn
and
1
Ma = (ηs Mis + ηc Mic ) (2.4)
ηm
The safety of a shotcrete lining can therefore be judged by a comparison of the external with
the allowable forces. A graphical representation of all possible combinations of the allowable
forces Na and Ma leads to the bearing-capacity-diagram, as illustrated in Fig. 2.11. The
main assumption behind this diagram is a linear strain distribution across the thickness of

58
the shotcrete layer. By applying appropriate constitutive laws for the shotcrete and steel
reinforcement, the internal allowable forces of resistance (Nis , Nic , Mis and Mic ) can be
computed taking into account the following cross section properties:

 Lining thickness

 Concrete cover

 Reinforcement area

Figure 2.11: Bearing-capacity-diagram for tunnel lining (from Hoek, 2007)

The example presented in Fig. 2.11 has been calculated in Hoek (2007) for a final unreinforced
concrete lining with a thickness of 50 cm and a uniaxial compressive strength of 35 MPa. The
plotted points refer to axial thrust and bending moment combinations obtained from a finite
element analysis. It can be seen that almost all the points fall within the capacity envelope
and for these cross sections the design can be regarded as safe. One big drawback of such an
approach is the underlying simplicity of the adopted constitutive laws in particular for young
shotcrete, which does not include the gradual increase in stiffness and strength with time.

2.6 Single-shell method

Conventional tunnel construction according to the principles of the NATM consists of two
tunnel shells, each of them having an individual purpose. Directly after excavation, a tempo-
rary shotcrete lining is installed to stabilise the opening and to contain short to medium-term

59
loads. When this lining has fully stabilised and the system has reached equilibrium, which
can be months after the completion of the outer shotcrete layer, a relatively thick permanent
cast in-situ concrete lining is installed with the aim of bearing long-term loads. In practice
this permanent concrete shell is often overdesigned and capable of bearing much higher loads
than needed. Furthermore, this second concrete shell provides durability and watertightness
either by the use of a waterproof membrane between the temporary and the permanent lining
or by the use of appropriate steel reinforcement to reduce crack widths to an acceptable value.
In the literature this method of tunnel construction is commonly referred to as the “double
shell method”.

However, over the last twenty years significant progress has been made in the tunnelling
industry and in particular in shotcrete technology. Constituent materials and equipment
have improved dramatically and it is nowadays possible to produce a high quality shotcrete
that meets the requirements of national standards in terms of durability and watertightness.
Therefore, the idea of using a single permanent shotcrete shell as tunnel support has been
introduced, replacing the traditional double shell method and reducing construction costs sig-
nificantly. In ASCCT (2004) a definition for the so called “single-shell construction method”
can be found, which states:

The single-shell construction method is characterised by the fact that all static and
structural requirements are fulfilled by a single shell structure. This shell can be produced in
one or several operations. The shell structure has to be designed not only to meet the need
for support during tunnel driving, but also to fulfil requirements of the finished structure.

Basically, two forms of execution of such a single shell using shotcrete exist, which are:

 A single-layer sprayed concrete shell, installed directly in the course of tunnel driving
having the final thickness of the tunnel lining.

 A composite multi-layer sprayed concrete shell, installed during tunnel driving and
subsequent operations to achieve the final thickness of the tunnel lining.

If the shotcrete shell is applied in several steps, it has to be ensured that the shear bonding
between the sprayed concrete layers is activated by adequate measures. This can be achieved
by cleaning of the surfaces with a mixture of compressed air and water, or preferably by
means of a high-pressure water jet (ASCCT, 2004). Fig. 2.12 illustrates a typical tunnel
profile for the single-shell method consisting of several individual shotcrete layers.

60
Figure 2.12: Composition of tunnel lining with the single-shell method (from Kusterle &
Lukas, 1993)

Despite the economic and time-saving advantages of the single-shell method, according
to Pöttler & Klapperich (1999) certain limitations exist, when tunnels are driven under the
following conditions:

 Great ingress of water

 Tunnel is located far below ground water level

 Conditions with water corrosive to concrete

 Geological conditions causing large bending moments of the lining leading to cracking
and the need for heavy reinforcement

A quantification of the single-shell method from the structural point of view with respect to
safety of the overall load-bearing system can be found in Pöttler & Schweiger (1999).

2.6.1 Innovative shotcrete tunnelling

One of the biggest infrastructure projects in the UK over the last five years was the new Ter-
minal 5 at Heathrow Airport in London, consisting of large underground structures including
14 km of tunnels for rail and road. The majority of these tunnels were constructed with
tunnel boring machines (TBMs), but for more complex elements, such as shafts and tunnel
connections, a new innovative single-shell shotcrete-lined method called “LaserShellTM ” has

61
been adopted. This particular tunnelling method uses fibre reinforced sprayed concrete and
was developed by a collaboration between the two companies, Morgan Est (UK) and Beton-
und Monierbau (Austria), with the initial aim of improving the safety of workers at the tunnel
face. Its principles are illustrated in Fig. 2.13 and described in detail in Williams (2008) and
Eddie & Neumann (2003).

Figure 2.13: The LaserShellTM method (from Jones et al., 2008)

The key features of the LaserShellTM method can be summarized according to Hilar et al.
(2005) as follows:

1. The tunnel is constructed full face (up to 5 m diameter in London Clay) in order to
minimise the number of construction joints and to increase productivity. Furthermore,
an inclined and domed face improves the stability in comparison to a conventional
vertical face and tends to reduce surface settlements.

2. Steel-fibre reinforced shotcrete is used exclusively without any conventional wire mesh
or lattice girders. This fact improves safety significantly since access to the unsupported
face is not necessary anymore. Consequently, the quality of the lining is remarkably
better as problems of shadowing behind the lattice girders and the wire mesh can be
avoided completely.

3. For the control of the excavation and lining geometry the “TunnelBeamerTM ” laser

62
 distometer is used, which takes spot readings of the distance to the excavated face or
to the lining (see Fig. 2.14).

4. The tunnel lining is constructed as a single-shell, where almost all the sprayed concrete
forms part of the watertight permanent lining.

Figure 2.14: TunnelBeamerTM system integration (from Eddie & Neumann, 2003)

As can be seen in Fig. 2.13, the complete shotcrete lining is built up in three individual layers.
The initial steel-fibre reinforced layer with a thickness of 75 mm provides instant structural
support directly after excavation of the ground and enhances the watertightness of the lining.
However, this layer is considered to be sacrificial for design purposes as its resistance might
deteriorate with time due to sulphate attack. The primary steel-fibre reinforced shotcrete
layer has a thickness of 200 mm and is considered as the permanent load bearing structure.
A finishing layer of 50 mm thickness is sprayed onto the existing layers after completion of
construction works to provide a smooth lining profile. This material is unreinforced and
with low accelerator dosage. Another major improvement of the LaserShellTM method is that
robotic spraying techniques are used for the shotcrete application, which guarantees a better
material quality during construction.

Despite all these advantages of the LaserShellTM method, some problems regarding the
increased amount of reinforcement for zones where high bending can be expected remain
unsolved. However, further developments leading to the so called “UltraShellTM ” method are
in progress under the supervision of Morgan Est and Beton- und Monierbau (Eddie et al.,
2009).

2.7 Modelling 2D tunnel excavation in a numerical analysis

From an extensive literature review it can be concluded that tunnelling is commonly regarded
as a highly complex three-dimensional process where stress redistributions take place near the
tunnel face. Ground movements not only occur in the radial direction, but also occur in the

63
longitudinal direction, parallel to the tunnel axis. Furthermore, a significant component of
these deformations in the soil mass occur well ahead of the tunnel face before the installation
of the tunnel lining, leading to a certain stress relief. A full 3D numerical analysis requires
enormous computational resources and is still not frequently performed in engineering prac-
tice. Hence, when tunnel construction is simulated in 2D, certain assumptions have to be
introduced in order to account for these 3D effects. The available approaches in the literature
that are regularly used for 2D analyses are discussed in detail in Potts & Zdravković (2001)
and Karakus (2007) and will be described briefly in the following sections.

2.7.1 Stress relief method

The stress relief method, or often termed the “convergence-confinement method”, was intro-
duced by Panet & Guenot (1982) and is based on the principle of unloading at the tunnel
boundaries before construction of the tunnel lining. In Fig. 2.15 the concept of the stress relief
method is shown, where a factor λ governs the proportion of unloading, which is progressively
increasing from 0 to 1 during the analysis. At a prescribed value λd the lining is installed. The
load acting onto the system (ground + lining) after completion of the numerical excavation
is then given as:
{σd } = (1 − λd ) {σo } (2.5)

Figure 2.15: Principle of the stress relief method (from Potts & Zdravković, 2001)

2.7.2 Stiffness reduction method

The stiffness reduction method was developed by Swoboda (1979) for the modelling of NATM
tunnels and uses a support core with a reduced modulus of elasticity. The soil formation at
the tunnel face is softened in a controlled way by a reduction factor β so that the stiffness of
the support core is given as:
Ee = β Eo (2.6)

64
Fig. 2.16 illustrates the steps of the stiffness reduction method. The tunnel lining is installed
before the numerical excavation of the soil is completed. If a staged tunnel construction
is applied, i.e. a bench-invert excavation, the procedure can be carried out for each of the
excavation areas individually.

Figure 2.16: Principle of the stiffness reduction method (from Potts & Zdravković, 2001)

2.7.3 Gap method

Another method for taking into account 3D effects in a 2D tunnel analysis was presented by
Rowe et al. (1983) and is suitable for various tunnelling techniques (NATM, TBM, etc.). The
expected ground loss due to tunnel excavation is introduced in the finite element mesh as
a predefined void, which is placed around the final tunnel position as indicated in Fig. 2.17.
The invert of the tunnel is rested on the underlying soil boundary and a gap parameter is
introduced at the crown, representing the vertical distance between the tunnel crown and the
initial position before tunnel excavation. As the soil within the tunnel is excavated, the nodal
displacements at the tunnel opening are monitored until the void has been closed and the soil
is in contact with the predefined position of the tunnel lining. At this stage the soil-lining
interaction is activated at this node.

Figure 2.17: Principle of the gap method (from Potts & Zdravković, 2001)

2.7.4 Volume loss control method

The principal idea behind the volume loss control method is very similar to that for the stress
relief method, with the difference of prescribing a volume loss that will result on completion of

65
excavation instead of a load reduction factor. Therefore this method is suitable for tunnelling
in ground conditions where the volume loss can be estimated before tunnel construction with
a high level of confidence, as is the case for London Clay. In the London area the volume loss
usually varies between 1 % and 2 % based on field measurements (Potts & Zdravković, 2001).
Fig. 2.18 depicts the basic concept of the volume loss control method in a numerical analysis.

Figure 2.18: Principle of the volume loss control method (from Potts & Zdravković, 2001)

When tunnel excavation is started, the initial equivalent nodal forces {Fo } at the tun-
nel boundary are calculated and reduced in a step-wise manner over a certain number of
increments n. This incremental reduction equals:

{Fo }
{ΔF } = (2.7)
n

The equal and opposite nodal force vector −{ΔF } is applied at the excavation boundary
for each of the n increments. The volume loss is monitored continuously throughout this
procedure and the tunnel lining is installed on the increment, at which the desired volume
loss is achieved. When simulating tunnel excavation in undrained conditions, the volume loss
VL can be calculated from:
VS
VL = (2.8)
Ve
where VS is the volume of the surface settlement trough and Ve the excavated tunnel volume.

66
The remaining nodal forces act onto the tunnel lining and introduce stresses and displace-
ments. If the tunnel lining has a relatively low stiffness, which can be the case for a shotcrete
shell at early ages, some further increase in the volume loss has to be expected and should be
taken into account. Finally, it should be noted that the volume loss control method is adopted
in this thesis for the analysis of tunnel construction in London Clay, as will be explained later
in Chapter 8.

2.8 Summary
This chapter dealt with several aspects related to tunnel construction and its design in urban
areas. The following key facts can be summarized:

 Shotcrete is one of the main support elements of the New Austrian Tunnelling Method
(NATM) and provides temporary stability to the tunnel opening. The main idea behind
NATM is to achieve equilibrium of the system after excavation by actively integrating
the surrounding ground formation into an overall ring-like support structure. However,
the entire range of principles of the NATM is not applicable for tunnelling in soft
ground conditions, where a rapid ring closure and surface settlement control are the
main concerns during construction.

 Geotechnical observation and monitoring are key elements in any underground struc-
ture. Particularly for tunnelling, the shotcrete lining behaviour during tunnel advance
is of crucial importance for engineers. Measurements of displacements and stresses
serve as a source of information to assess the safety of the structure and to validate the
anticipated lining design. However, difficulties exist regarding the applicability of the
possible measuring techniques (i.e. optical surveying, strain gauges, pressure cells) and
the correct interpretation of the obtained results.

 A literature review has shown that several lining design techniques exist that are widely
used in engineering. Starting from simple methods based on empirical observations,
some closed-form analytical methods have been developed that are useful at the pre-
liminary design stage (continuum method, convergence-confinement method, bedded-
beam-spring models). Common assumptions for these methods are a circular geometry
of the tunnel and linear elastic behaviour of the involved materials. Numerical methods
are a powerful tool for analysing more complex tunnel geometries, realistic soil-structure
interaction and advanced constitutive behaviour.

 The single-shell method as a relatively new technology assumes that all the static and
structural requirements are fulfiled by a single shotcrete shell applied in several layers.
It replaces the costly double-shell method where a permanent cast in-situ final lining is
usually installed months after the completion of tunnel construction. However, certain
concerns exist for the application of the single-shell method regarding its durability

67
 and watertightness. New innovations combining safe excavation schemes and the use
of fibre-reinforced sprayed concrete layers that act statically as a single shell are in
progress.

 When performing a 2D numerical analysis of tunnel construction, certain 3D effects

such as stress relief and pre-deformation ahead of the tunnel face should be taken into
account. Several techniques for achieving this have been presented, among them the
volume loss control method, which will be applied in Chapter 8.

68
Chapter 3

The Finite Element Method

3.1 Introduction
The finite element method is a sophisticated form of numerical analysis that is capable of
dealing with complex structures in engineering practice. It has been used in many engineering
fields over the last thirty years and has more recently been introduced for analysing geotech-
nical problems. In contrast to simple solution methods, such as limit equilibrium, stress field
and limit analysis, the finite element method is capable of satisfying all the four fundamen-
tal requirements for a complete theoretical solution: equilibrium of forces, compatibility of
displacements, material constitutive behaviour and boundary conditions. It can therefore
provide advanced information for a safe design of subsurface structures.

The numerical research group at Imperial College London, led by Prof. D. Potts and
Dr. L. Zdravković, has been working at the leading edge of the development and applica-
tion of the finite element method for practical geotechnical problems. As a result of their
research, the Imperial College Finite Element Program (ICFEP) has been developed over
the last three decades. This sophisticated code makes use of the displacement based finite
element method and is capable of performing both 2D and 3D analysis. All the analyses
presented in this thesis have been carried out using ICFEP and 2D plane strain conditions
have been applied exclusively.

This chapter aims to give an introduction into the main concept of the finite element
method following in principle a procedure presented in Potts & Zdravković (1999). First,
the focus is on the basic theory for linear materials and highlighting all the involved steps
of a finite element analysis. Furthermore, some modifications that are necessary to enable
this theory to be used in geotechnical engineering are presented. However, concrete and
shotcrete show a highly non-linear material behaviour under all kinds of loading conditions
and therefore the finite element theory for linear materials is extended in such a way, that ad-
vanced non-linear elasto-plastic constitutive models can be successfully applied. The solution
strategies and algorithms implemented into ICFEP will be discussed in detail.

69
3.2 Finite element theory for linear materials
The following sections describe briefly each of the steps involved in the finite element method,
which are:

1. Element discretisation

2. Primary variable approximation

3. Formulation of element equations

4. Boundary conditions

5. Solution of global equations

3.2.1 Element discretisation

As a first step in analysing an engineering problem with the finite element method, the
geometrical domain to be investigated must be defined and then divided into small regions,
the so called finite elements. For a 2D analysis, the shape of these elements is often triangular
or quadrilateral and they are connected together at their nodes, forming a mesh over the whole
area under consideration. The nodes are usually situated at the corners of the mesh, although
additional nodes can be encountered at the midpoints of the straight or curved element sides.
This results in 6-noded triangular or 8-noded quadrilateral finite elements. An appropriate
mesh design is of crucial importance for a realistic modelling of a structure and the size
and number of elements strongly influences the accuracy of the obtained solution. For zones
where rapid changes in the unknown variables can be expected it is necessary to refine the
mesh by the use of smaller elements. However, an increasing refinement of the mesh leads to
high computational costs, consequently an optimum mesh design is recommended.

3.2.2 Approximation of primary variable

In the finite element method a primary variable has to be selected. Furthermore, some rules
have to be established to define how this primary quantity varies over the finite element.
Other quantities are then treated as secondary variables and can be determined once the
primary variable has been calculated. In geotechnical engineering it is common to adopt
displacements as the primary unknown quantity and other quantities, such as stresses and
strains, are treated as secondary variables, resulting from displacements

In the displacement based finite element method the displacement field varies over the
problem domain. This method assumes further that the displacements within a finite element
can be estimated from the nodal displacements by mathematical functions. These functions
are commonly termed “shape functions”. When considering a 2D analysis, the displacement

70
field is characterised by the two displacements u and v, in the horizontal x- and vertical y-
direction respectively. In this case, the displacement field within an element can be described
by:    
u u
= [N ] (3.1)
v v nodes

In this equation [N ] is the matrix of the shape functions, which are of a quadratic nature for
the higher order 8-noded elements. It is now possible to describe the displacements u and v
in terms of their values at the element nodes. These nodal displacements are often referred
to as the unknown degrees of freedom. Equation 3.1 reduces the problem of determining the
displacement field over the entire geometrical domain to the determination of the displacement
components at a finite number of nodes of the finite element mesh.

The type and shape of finite elements to be adopted depends largely on the geometry
being modelled and the type of analysis required. Best results are obtained if the elements
have reasonable shapes and do not get too distorted during the analysis. In ICFEP, 8-noded,
quadrilateral, isoparametric elements are used, as illustrated in Fig. 3.1. It can be seen, that
the global element, as it appears in the finite element mesh, is derived from a parent element,
which is defined with respect to a natural coordinate system. These natural coordinates S
and T vary from -1 to +1. This element is called “isoparametric”, since the same shape
functions [N ], that are used to describe the displacement field within the element, are also
adopted to map the geometry of the element from the global to the natural coordinate system.

Figure 3.1: 8-noded isoparametric element used in ICFEP (from Potts & Zdravković, 1999)

The global coordinates of a point within the element can be expressed as:


8 
8
x= Ni xi and y= Ni yi (3.2)
i=1 i=1

xi and yi are the global coordinates of the 8 nodes in the element and Ni are the so called
“interpolation functions”, which are expressed in terms of the natural coordinates S and T
varying from -1 to +1. For each node in the element such an interpolation function exists
and takes a value of +1 at the correspondent node. Examples for the quadratic interpolation

71
functions are given in Potts & Zdravković (1999).

3.2.3 Formulation of element equations

For the description of the deformational behaviour of each finite element a set of global
equations is needed, combining compatibility, equilibrium and constitutive conditions. Their
derivation will be explained here briefly, assuming 2D plane strain conditions.

As introduced earlier, the incremental displacements within a finite element are given through
the incremental nodal displacements as:

{Δd} = [N ] {Δd}n (3.3)

where    
Δu Δu
{Δd} = and {Δd}n = (3.4)
Δv Δv nodes

Assuming a compression positive sign convention, the definition of strains as an expression for
the mathematical compatibility can be written according to Timoshenko & Goodier (1951)
as:

∂Δu ∂Δv ∂Δu ∂Δv

Δεx = − Δεy = − Δγxy = − − Δεz = Δγxz = Δγzy = 0
∂x ∂y ∂y ∂x
(3.5)
By replacing the displacements in the above equations with the approximations given in equa-
tion 3.3, one obtains the following definition of strains with respect to nodal displacements:
 
Δu
{Δε} = [B] = [B] {Δd}n (3.6)
Δv n

The matrix [B] only contains derivatives of the shape functions Ni and {Δd}n represents
the list of nodal displacements for a single element. As mentioned earlier, for isoparametric
elements the shape functions are identical to the interpolation functions and depend only on
the natural coordinates S and T . As a consequence, the derivatives of the shape functions
with respect to the global coordinates x and y in the matrix [B] cannot be determined directly.
By applying the chain rule of differentiation it can be written that:
!T !T
∂Ni ∂Ni ∂Ni ∂Ni
= [J] (3.7)
∂S ∂T ∂x ∂y

where [J] is the Jacobian matrix given as follows:

⎡ ⎤
∂x ∂y
⎢ ∂S ∂S ⎥
[J] = ⎢
⎣ ∂x
⎥ (3.8)
∂x ⎦
∂T ∂T

72
Hence, by making use of the Jacobian matrix the global derivatives of the shape functions
can be obtained as:
⎡ ⎤ ⎡ ⎤⎧ ⎫
∂Ni ∂y ∂y ⎪ ∂Ni ⎪
− ⎪
⎨ ⎪
⎬
⎢ ∂x ⎥ 1 ⎢ ∂T ∂S ⎥ ∂S
⎢ ⎥ ⎢ ⎥
⎣ ∂N ⎦ = | J | ⎣ ∂x ∂x ⎦ ⎪ ⎪
(3.9)
i ⎪
⎩ ∂Ni ⎪
⎭
−
∂y ∂T ∂S ∂T

where | J | is the Jacobian determinant calculated as:

∂x ∂y ∂y ∂x
| J |= − (3.10)
∂S ∂T ∂S ∂T

In the general case, the constitutive behaviour of a material can be written as:

{Δσ} = [D] {Δε} (3.11)

where [D] is the constitutive matrix. This equation provides a relationship between the
stresses and strains and therefore links equilibrium and compatibility.

The element equations can now be obtained by applying the principle of minimum potential
energy, which states that the static equilibrium position of a loaded linear elastic body is the
one which minimises the total potential energy. The incremental total potential energy of a
body ΔE is defined as:
ΔE = ΔW − ΔL (3.12)

where ΔW is the incremental strain energy and ΔL the incremental work done by the applied
loads. These two quantities are given as:
)
1
ΔW = {Δε}T {Δσ} d V ol (3.13)
2
V ol

and ) )
T
ΔL = {Δd} {ΔF } d V ol + {Δd}T {ΔT } d Srf (3.14)
V ol Srf

In these two equations {Δd}T = {Δu, Δv} is the displacement vector, {ΔF }T = {ΔFx , ΔFy }
the body force vector and {ΔT }T = {ΔTx , ΔTy } the surface tractions vector (i.e. for line
loads or surcharge pressures).

The principle of minimum potential energy can now be expressed mathematically as:

δΔE = δΔW − δΔL = 0 (3.15)

Substitution of equations 3.13 and 3.14 into equation 3.15 and making use of equations 3.3,
3.6 and 3.11 allows the establishment of the equilibrium equations for a single finite element

73
which take the following form:

[KE ] {Δd}n = {ΔRE } (3.16)

[KE ] is the element stiffness matrix and is given as:

)
[KE ] = [B]T [D][B]d V ol (3.17)
V ol

{ΔRE } represents the right hand side load vector and is obtained from:
) )
{ΔRE } = [N ] {ΔF } d V ol +
T
[N ]T {ΔT } d Srf (3.18)
V ol Srf

The volume and surface integrals in equations 3.17 and 3.18 are evaluated for isoparametric
elements using the natural coordinate system of the parent element S and T . As an example,
the coordinate transformation for the volume integral can be written as:

d V ol = t dx dy = t | J | dS dT (3.19)

In the case of plane strain conditions the thickness t is taken equal to unity. From the
procedure presented above it can be observed that the main advantage of the isoparametric
element formulation is that the element equations need only be evaluated in the parent
coordinate system. The integrals can be carried out over a square, with S and T varying
between -1 and +1 and the stiffness matrix for each element in the mesh can be established
using a standard procedure.

3.2.3.1 Numerical integration

For the establishment of the stiffness matrix and the right hand side load vector, integrations
must be performed (see equations 3.17 and 3.18). Usually an explicit calculation of these
integrals is not possible and therefore numerical integration schemes have to be adopted. In
ICFEP a Gaussian integration scheme is used, which replaces the integral of a function by a
weighted sum of the function evaluated at a number of integration points. When considering
a 1D integral of a function f (x) over the domain −1 < x < +1 it can be written:

)+1 
n
f (x) dx ≈ Wi f (xi ) (3.20)
−1 i=1

where f (xi ) are the values of the function at the n integration points and Wi are the corre-
sponding values of the weighting coefficients. The number of integration points determines
the integration order. The accuracy of the integration obviously increases with a higher inte-

74
gration order, but at the same time this leads to longer computational time since the number
of function evaluations also increases.

When adopting the Gaussian integration scheme, the integration points are often referred
to as “Gauss points”. In most of the cases a 2 x 2 or a 3 x 3 integration order is used for
an 8-noded isoparametric element, which are termed reduced or full integration respectively.
The location of the Gauss points for such an 8-noded isoparamtric element in the global and
the parent coordinate system is shown in Fig. 3.2.

Figure 3.2: Location of Gauss points for 2 x 2 and 3 x 3 integration order (from Potts &
Zdravković, 1999)

3.2.4 Formulation of global equations

Once the element equations have been formulated it is necessary to assembly these separate
element equations into a set of global equations, which are of a similar form and can be
written as:
[KG ] {Δd}n,G = {ΔRG } (3.21)

where [KG ] is the global stiffness matrix, {Δd}n,G the displacement vector of all the nodes
for the entire mesh and {ΔRG } the global right hand side load vector.

The assembly process of transforming the element stiffness matrix into a global stiffness
matrix is called “direct stiffness method”. The basic principle of this method is that each term
of the global stiffness matrix is obtained by summing the individual element contributions
whilst taking into account the degrees of freedom which are common between the elements.

75
A similar technique is applied to the global right hand side load vector, which contains the
sum of the individual loads acting on each node. A more detailed description can be found
in Potts & Zdravković (1999).

3.2.5 Boundary conditions

Appropriate boundary conditions are of great importance in the finite element formulation.
These are usually displacement and load conditions that fully define the boundary value
problem to be analysed.

Loading conditions include the application of line loads or surcharge pressures and they
affect the right hand side vector {ΔRG } of the global system of equations. Other examples
for loading conditions are body forces from excavated or constructed elements. Displacement
boundary conditions influence the global nodal displacement vector {Δd}n,G . Sufficient dis-
placement conditions have to be specified at the mesh boundaries in order to avoid any rigid
body motion of the whole mesh. If this requirement is not satisfied, the global stiffness matrix
becomes singular and the equations can not be solved.

3.2.6 Solution of global equations

The last step in the finite element formulation is the solution of the global set of equations
in order to obtain the unknown values of the nodal displacements {Δd}n,G . In principle, this
global set of equations forms a large system of simultaneous equations which can be solved
by several mathematical techniques. One of the most commonly applied solution strategy in
finite element analysis is the Gaussian elimination method. For a more detailed description
of this method see Potts & Zdravković (1999). Once the nodal displacements of the entire
mesh are calculated, secondary quantities such as stresses and strains can be evaluated from
equations 3.6 and 3.11.

3.3 Geotechnical considerations

The finite element theory presented in the previous sections is applicable to the analysis
of any linear elastic continuum. However, several limitations exist regarding a successful
application of the theory for the realistic analysis of boundary value problems in geotechnical
engineering. Hence, certain modifications have to be made which will be introduced here
briefly.

In general, the constitutive behaviour of a material is formulated in terms of total stresses

and strains. It is well known that soil exhibits a highly non-linear stress-strain behaviour
and can be considered as a two phase material. It consists of a soil skeleton and water
that fills the voids of the soil skeleton. In some particular cases, the soil is only partially
saturated and air is included in the system. This type of soil is usually treated as a three

76
phase material, but is not considered within this thesis. In Chapter 1 it was highlighted that
in geotechnical engineering the total stress tensor is usually split into effective stresses and
pore water pressures due to the two phase nature of soil. This principle of effective stress
was formulated mathematically in equation 1.13. For any geotechnical analysis the flow of
water through the soil skeleton has to be taken into account and is further coupled with the
deformational behaviour of the soil. Therefore, the pore water pressure is introduced as an
extra degree of freedom.

In reality two extreme cases can be encountered: When loading is applied on a soil with a
relatively high permeability at a low rate, no excess pore water pressures develop. A steady
state drainage pattern exists in the soil and fully drained conditions can be assumed. On
the contrary, if a soil with a relatively low permeability is loaded rapidly, no flow occurs
within the soil and pore water pressures build up, which are linked in a unique way to the
deformation of the soil. This condition is termed to be undrained. As will be explained in
Chapter 8, undrained conditions are assumed for all the tunnel analyses carried out in this
thesis. Therefore, some theoretical background of how to consider pore water pressures in
the finite element method will be presented in the following section.

3.3.0.1 Undrained analysis

When performing an undrained analysis in terms of total stresses, no information can be

obtained regarding the changes in pore water pressures due to excavation or construction
of the soil mass, which sometimes is an important requirement for the design of a geotech-
nical structure. To overcome this lack of information it is more convenient to describe the
constitutive soil behaviour with effective stress parameters. Hence, the principle of effective
stresses can be applied and as in equation 1.13 it can be written:

{Δσ} = {Δσ  } + {Δσf } (3.22)

where the vector of pore pressures is given as:

{Δσf } = {Δu Δu Δu 0 0 0}T (3.23)

For undrained conditions no flow of water occurs in the soil and the two phases are deforming
together. Hence, the strains are the same in each phase and the constitutive behaviour can
be written as:
{Δσ  } = [D ] {Δε} (3.24)

and
{Δσf } = [Df ] {Δε} (3.25)

77
Making use of these two equations and equation 3.11 gives:

[D] = [D ] + [Df ] (3.26)

where [D] and [D ] are the stiffness matrices in terms of total and effective stresses respectively.
[Df ] is the pore fluid stiffness and can be related to the bulk modulus of the pore fluid Kf
for the general three dimensional case in the following way:

[Df ] = Kf {b} {b}T (3.27)

with
{b}T = {1 1 1 0 0 0 } (3.28)

and the equivalent bulk modulus of the pore fluid Ke as:

Δu
Ke = (3.29)
Δεvol

Usually it is convenient to set Ke = Kf (see Potts & Zdravković, 1999) and therefore the
total constitutive matrix can be expressed as:

[D] = [D ] + Kf {b} {b}T (3.30)

It is now straightforward to incorporate the principle of effective stresses into a finite element
analysis by replacing the total stress constitutive matrix [D] by equation 3.30. This requires
the specification of the effective stress constitutive matrix [D ] and the bulk modulus of the
pore fluid Kf . The determination of unknown nodal displacements follows the standard
procedure described in the previous sections. However, when calculating stresses the bulk
modulus of the pore fluid is used to determine the changes in pore water pressures. In a similar
way, the changes in effective soil stresses are obtained by applying the effective constitutive
matrix [D ]. The changes in total stresses can be calculated according to equation 3.22.

If the bulk modulus of water Kf is set to zero, the analysis reduces to drained conditions
and pore fluid pressures do not change during loading. In order to achieve undrained condi-
tions, a relatively high value for Kf has to be specified. In ICFEP it is common practice to
simulate undrained conditions by establishing the bulk modulus of the pore fluid from the
bulk modulus of the soil skeleton Kskel via the following equation:

Kf = β Kskel (3.31)

where β has a value between 100 and 1000. Kskel can be evaluated from the effective stress
parameters forming the [D ] matrix (see Potts & Zdravković, 1999).

78
3.4 Finite element theory for non-linear materials
When dealing with materials that exhibit non-linear constitutive behaviour, the finite ele-
ment theory presented in the previous sections has to be adapted. In the literature, several
mathematical strategies are available for dealing with non-linearity, having in common that
the boundary conditions are applied in a series of increments. In the case of non-linear elasto-
plastic models that are often used for simulating the material response of concrete and soil,
the constitutive matrix [D] is replaced by the elasto-plastic constitutive matrix [Dep ] and
varies with stress and/or strain. Therefore, the global stiffness matrix [KG ] is not constant
any more but changes during the finite element analysis, a fact which makes the solution of
the problem more difficult. The set of global equations earlier introduced in equation 3.21
can now be rewritten in an iterative form as:

[KG ]i {Δd}in,G = {ΔRG }i (3.32)

where [KG ]i is the incremental global stiffness matrix, {Δd}in,G the vector of incremental
nodal displacements, {ΔRG }i is the vector of incremental nodal forces and i the increment
number. This global set of equations must be solved for each increment during the change
in boundary conditions. The final solution is then obtained by summing up the results from
each increment. Some popular solution strategies that are commonly in use are the tangent
stiffness method, the visco-plastic method and the modified Newton-Raphson scheme. The
latter is known as a robust and economic solution technique and was therefore applied for all
the analyses performed throughout this thesis. Its basic principle is explained briefly in the
next section.

3.4.1 The modified Newton-Raphson method

When applying the modifed Newton-Raphson method each increment of equation 3.32 is
solved in a number of iterations. Fig. 3.3 illustrates graphically the basic principle of the
method for a simple uniaxial loading condition. At the beginning of the first iteration a
global stiffness matrix [KG ]o is calculated from the current stresses at the beginning of the
increment. However, during the first iteration the technique recognizes that the solution
according to equation 3.32 is likely to be in error due to the non-linear variation of the
stiffness matrix over the increment. The predicted displacements Δd1 are used to calculate
a residual load vector ψ 1 , which is a measure of the error in the analysis. Equation 3.32 is
then solved again in a second iteration, using the residual load vector ψ 1 as the right hand
side vector. The residual load vector ψ 2 at the end of the second iteration is evaluated from
the predicted displacements (Δd1 + Δd2 ).

79
Figure 3.3: Basic principle of the modified Newton-Raphson scheme (after Potts &
Zdravković, 1999)

The complete iterative process can be written in a compact mathematical form as:
* +j
[KG ]i {Δd}in,G = {ψ}j−1 (3.33)

The newly introduced superscript j refers to the iteration number and for the first iteration
{ψ}o = {ΔRG }i . This procedure continues until the prescribed convergence criteria are
satisfied. The incremental solution of the unknown displacement quantities is then evaluated
by summing up the displacements of each iteration. In ICFEP these convergence limits are
* +j
set on the size of both the iterative displacements {Δd}in,G and the residual loads {ψ}j .
They are given as:

,* 
, +j ,
, * +j T * +j
, {Δd}i ,= {Δd}i
{Δd}i
(3.34)
, n,G , n,G n,G

, 
and
,
,{ψ}j , = ({ψ}j )T {ψ}j (3.35)

The iterative displacement norm of equation 3.34 is usually checked against the norms of
the incremental and accumulated displacements. Likewise, the norm of the residual loads
of equation 3.35 is compared to the norms of the incremental and accumulated global right
hand side load vector. For the convergence criteria in ICFEP, 2 % is set as a default value
and has been used for the analyses performed within this thesis.

When adopting the modified Newton-Raphson method the incremental global stiffness
matrix [KG ]i is usually calculated and inverted at the beginning of an increment and used
for all the further iterations within this increment. This helps to reduce the amount of
computation compared to the original Newton-Raphson scheme, where the incremental global
stiffness matrix is updated for each iteration. ICFEP enables the user to specify the stiffness
matrix to be recalculated every certain number of iterations. Furthermore, sometimes the

80
global stiffness matrix is determined using the elastic constitutive matrix [D] rather than
the elasto-plastic constitutive matrix [Dep ], since the predicted solution by equation 3.32 is
expected to be in error anyway.

One key step in the modified Newton-Raphson solution technique is the evaluation of the
residual load vector {ψ}j for each iteration. It can be determined from the following proce-
dure: First, the current incremental strains at each integration point are calculated from the
incremental displacements obtained at the end of the previous iteration. The constitutive
model is then integrated along these strain paths in order to estimate the change in stresses,
which is added to the stresses at the beginning of the increment. From this current accumu-
lated stress state the equivalent nodal forces can be evaluated. The residual load vector is the
difference between these nodal forces and the externally applied loads. For the integration of
the constitutive equations along the incremental strain path a method termed “stress point
algorithm” is necessary. Several of these algorithms are available in the literature and the
one used in ICFEP will be presented in the next section of this chapter.

3.4.2 Stress point algorithm

The adopted stress point algorithm for the analyses presented in this thesis is a so called semi-
explicit “substepping algorithm” combined with a modified Euler integration scheme, which
integrates the constitutive equations over each substep. As mentioned earlier, the objective of
such a stress point algorithm is to estimate the incremental stresses {Δσ} from the previously
calculated incremental strains {Δε} in order to obtain the accumulated stresses at the end
of the increment as:
{σ} = {σo } + {Δσ} (3.36)

where {σo } are the stresses at the beginning of the increment. The size of each substep
is usually determined by the error in the numerical integration compared to a user defined
tolerance. The complete stress change at the end of an iteration is then obtained by summing
the stress changes from each substep. A simple graphical representation of the substepping
approach for an elasto-plastic material in the J-p space can be seen in Fig. 3.4

Figure 3.4: Basic principle of the substepping approach (from Potts & Zdravković, 1999)

The following general procedure based on a scheme presented by Sloan (1987) is carried

81
out: At the beginning, the elastic proportion α of an increment has to be determined by
checking the value of the yield function F and adopting some simple interpolation functions
(see Potts & Zdravković, 1999). The remaining strain increment associated with elasto-plastic
behaviour is therefore equal to (1 − α){Δε}. The initial stress and strain components of the
remaining elasto-plastic increment are given as:

{σ} = {σo } + {Δσ e } (3.37)

and
{Δεs } = (1 − α) {Δε} (3.38)

where {σo } are the stresses at the beginning of the increment and {Δσ e } the elastic proportion
of the stress increment. {Δεs } represents the elasto-plastic part of the total strain increment
{Δε} and will be split into a number of substeps in the course of the algorithm. It is assumed
that such a substep strain {Δεss } writes as:

{Δεss } = ΔT {Δεs } (3.39)

with the proportion ΔT varying between 0 and 1. Initially it is assumed that just one
substep is necessary and hence ΔT is set to unity. A first estimate of the associated changes
in stresses, plastic strains and hardening parameters is obtained by using a first order Euler
integration approximation, namely:

{Δσ1 } = [Dep ({σ}, {k})] {Δεss } (3.40)

∂P ({σ}, {m1 })
{Δεp1 } = Λ ({σ}, {k}, {Δεss }) (3.41)
∂σ
{Δk1 } = {Δk ({Δεp1 })} (3.42)

The above quantities are used to calculate the stresses and hardening parameters at the end
of the substep according to {σ} + {Δσ1 } and {k} + {Δk1 } respectively. Consequently, a
second estimate for the changes in stresses, plastic strains and hardening parameters over the
substep is carried out and is obtained as:

{Δσ2 } = [Dep ({σ + Δσ1 }, {k + Δk1 })] {Δεss } (3.43)

∂P ({σ + Δσ1 }, {m2 })

{Δεp2 } = Λ ({σ + Δσ1 }, {k + Δk1 }, {Δεss }) (3.44)
∂σ
{Δk2 } = {Δk ({Δεp2 })} (3.45)

Now a more accurate modified Euler estimate of the changes in stresses, plastic strains and
hardening parameters can be established as:

1
{Δσ} = ({Δσ1 } + {Δσ2 }) (3.46)
2

82
 1
{Δεp } = ({Δεp1 } + {Δεp2 }) (3.47)
2
1
{Δk} = ({Δk1 } + {Δk2 }) (3.48)
2
By subtracting equation 3.40 from 3.46 one obtains an estimate of the local error in the
predicted stresses as:
1
E≈ ({Δσ2 } − {Δσ1 }) (3.49)
2
The relative error in stress for the substep can be expressed as:

E
R= (3.50)
{σ + Δσ}

This error is checked against a user defined tolerance SST OL and if the error is unacceptable,
i.e. R > SST OL, the initial substep size is reduced further. The above presented procedure
is then repeated with a new value of ΔT in equation 3.39 until the calculated error E is
smaller than the defined tolerance and the substep size is accepted. This is followed by an
update in the accumulated stresses, plastic strains and hardening parameters and the next
substep can be carried out. For further detailed information about the error control and the
estimation of the size of the substep see Potts & Zdravković (1999).

3.5 Summary
This chapter intended to provide an introduction into the theory of the finite element method.
First, the theoretical background for modelling linear elastic materials was discussed in de-
tail, highlighting the principal steps that are involved in such a complex numerical analysis.
Second, in order to apply the finite element method to geotechnical engineering problems
some modifications for a realistic modelling of soil as a two phase material have been pre-
sented, including the principle of effective stresses. However, concrete and shotcrete show
highly non-linear stress-strain behaviour in various loading conditions. Therefore the basic
finite element theory has to be extended to account for non-linear material behaviour and
the necessary numerical algorithms for such a non-linear finite element analysis have been
presented. The following aspects can be summarized:

 The finite element method is a powerful tool for analysing complex engineering problems
and follows in general the following procedure: element discretisation, approximation
of the primary variable, formulation of element equations, boundary conditions and
solution of global equations.

 The displacement based finite element method is the most common approach for geotech-
nical engineering purposes. The variation of the displacement field over the element is
the key aspect of this method and can be described by shape functions.

 The formulation of the element equations couples compatibility, equilibrium and con-

83
 stitutive conditions leading to a set of equations that describe the relationship between
loads and displacements via the element stiffness. These element equations are then
assembled into a global set of equations.

 Boundary conditions with respect to loads and displacements are of crucial importance
for a finite element analysis for a full definition of the boundary value problem to be
analysed.

 For geotechnical applications soil as a two phase medium has to be considered in a finite
element analysis. The incorporation of the principle of effective stresses into the basic
theory of the finite element method allows the calculation of pore water pressures and
effective soil stresses in an undrained analysis, as presented later in Chapter 8.

 When performing a non-linear finite element analysis some iterative techniques have to
be used in order to solve the set of global equations. The modified Newton-Raphson
method combined with a semi-explicit substepping algorithm have been adopted as a
solution strategy.

84
Chapter 4

Shotcrete Technology

4.1 Introduction
Although the history of shotcrete goes back to the beginning of the 20th century, shotcrete
technology is a relatively young field of research, where many new innovations regarding
material composition and installation equipment have been achieved over the past decades.
However, in reality engineers have to face many problems when confronted with the design or
construction of sprayed concrete structures due to their limited knowledge about the funda-
mental principles with respect to material technology and mechanical behaviour. Therefore,
the aim of this chapter is to introduce the basic background information about shotcrete
technology with a particular focus on material ingredients and installation techniques. It
should enable the reader to understand in a better way the complex mechanical behaviour of
young shotcrete, that will be discussed in detail in Chapter 5. In the first part of this chap-
ter, the main material components of sprayed concrete will be presented, followed by a short
introduction into the chemical reaction of cement hydration. Furthermore, the two different
ways of spraying shotcrete in place, dry- and wet-mix, will be explained in detail. Finally, a
quick review about different testing techniques of shotcrete at early ages will complete this
chapter.

4.2 Shotcrete as an engineering material

Shotcrete is a special type of concrete, which was invented at the beginning of the 20th
century by Carl Akeley (1864 - 1926), a famous American taxidermist (see Fig. 4.1). He
experimented with the pneumatic application of first plasters and then cement mortars to
make models of animals for the Field Museum of Natural History in Chicago (Austin &
Robins, 1995). Akeley used the simple method of blowing a dry mixture of sand and cement
out of a hose with compressed air and wetting it as it was released. In 1911 he obtained
patents for both the equipment and method and the material was called “Gunite”. One
year later, the Cement Gun Company of Allentown, Pennsylvania, took over the patents and

85
expanded to Europe, bringing this new technology to Germany in 1921 and the UK in 1924.
Fig. 4.1 shows the original cement gun designed for the dry-mix process.

Figure 4.1: Carl Akeley and his first cement-gun

Nowadays several names for the sprayed mixture of aggregates and cement are in use
worldwide, among them “Sprayed Concrete” or “Shot Concrete”. They all have in common
to define this material as mortar or concrete which is pneumatically projected into place at
high pressures. The term “Shotcrete” is mainly used in the US when describing a mix whose
maximum aggregate size is no more than 10 mm, whereas “Sprayed Concrete” is the more
widely used terminology in Europe (SCA, 1999).

Shotcrete has a wide range of applications in all kinds of engineering fields, and each makes
use of the uniquely flexible nature of the application technique, which requires minimum
formwork and access space to produce flat or curved surfaces (Austin & Robins, 1995). They
include:

 New construction and repair of reinforced concrete structures

 Thin arches, domes and shells

 Rock stabilization

 Underground rock support and tunnelling

 Fire protection of steel framed buildings

In 1914, shotcrete was proposed for the first time in underground structures to protect
mine drifts against the atmosphere and to make them fireproof (Moussa, 1993). The first
application of shotcrete for tunnel construction in the USA is recorded in the early 1920s. In
Europe, shotcrete was introduced successfully in 1951 for the Ladano-Mosagno tunnel of the
Maggia Hydro-Electric scheme in Switzerland. With the rapid development of tunnelling,
shotcrete became more and more popular in underground structures and is nowadays an im-
portant support element of the New Austrian Tunnelling Method (NATM). The early sprayed
concrete did not show high material quality and large quantities of aggressive additives were

86
needed to make the material adhere to the ground in order to spray reasonably thick layers
(Thomas, 2009). It has largely been associated with rock tunnelling and with the treatment
of closely jointed or shattered rocks, such as those found in Alpine regions. Nevertheless, in
the 1970s shallow tunnels in soft ground conditions, as part of metro projects in Frankfurt
and Munich, were successfully constructed using shotcrete for ground stabilization. In the
UK shotcrete technology is relatively new in tunnelling industry and has only become widely
used within the last 15 years. Over the last 10 years significant progress has been made
regarding the constituent materials, mix formulations and equipment design for shotcrete,
but there is still a need for more research and development of design methods, standard test
methods and specifications (Austin & Robins, 1995).

Up to now, no internationally accepted framework exists for practical applications, cov-

ering the above mentioned aspects for all the different structures involving shotcrete. In the
literature the following guidelines and recommendations for shotcrete technology have been
found:

1. Austrian Society for Concrete- and Construction Technology, ASCCT (2004): Sprayed
concrete guideline

2. German Standard DIN 18551 (2005): Shotcrete - specification, production, design and
conformity

3. AFTES work group Nr. 20 (2000): Recommendations for the design of sprayed concrete
for underground support

4. American Concrete Institute ACI 506R-05 (2005): Guide to shotcrete

5. EFNARC (1996): European specification for sprayed concrete

6. European Standard EN 14487-1 (2005): Sprayed concrete

7. Sprayed Concrete Association UK (1999): Introduction to sprayed concrete

In ASCCT (2004) three different classes of shotcrete are defined according to its structural
use in engineering construction:

SpC I: Shotcrete without structural use for the sealing of surfaces and short temporary
support during construction stages.

SpC II: Shotcrete as a structural element for the support of underground excavations,
shaft constructions and retaining walls.

SpC III: Shotcrete with very special structural use for the construction of single-shell
tunnel linings.

87
For each individual sprayed concrete class certain requirements have to be fulfilled in terms of
material properties such as early strength, density of the material, uniformity, permeability
and leaching behaviour, frost resistance and chemical exposure. The test requirements to be
met differ accordingly and can be found in ASCCT (2004).

4.3 Mix design

Like any other ordinary cast structural concrete, shotcrete consists of aggregates, cement
and water. A variety of additives and admixtures is often added to the sprayed concrete mix
in order to improve or change certain material properties. In the following sections of this
chapter each shotcrete component should be discussed briefly.

4.3.1 Aggregates
The selection of the maximum aggregate size for shotcrete depends heavily on its structural
use, but is normally between 4 and 16 mm (ASCCT, 2004). For sprayed concrete of class
SpC II and SpC III the maximum diameter of aggregate should be limited to 11 mm because
of its rebound behaviour: the larger the pieces of aggregate, the more of them are lost in
rebound, specially when wire mesh is used as reinforcement. This also means, that the in situ
material is more finely graded than before spraying, a fact that has to be taken into account
in mix design to ensure a balanced in situ material. Both crushed and round aggregates can
be used and the grading curve should be smooth, slightly oversanded and be based usually
towards the finer end for reasons of pumpability. The moisture content of aggregates is an
important factor in the dry mix process. For a moisture content of less than 3 % the dust
production will be very high and therefore the optimum value is between 3 and 6 %. Fig. 4.2
shows a typical recommended grading zone for sprayed concrete taken from EFNARC (1996).

88
 m aterialpassing (% by w eight)

Figure 4.2: Recommended aggregate gradation zone from EFNARC (1996)

4.3.2 Cement
As a binder for shotcrete ordinary Portland cements can be used with a dosage between
325 and 450 kg/m3 before casting (Girmscheid, 2000). The amount of cement included in the
shotcrete mix depends also on the adopted maximum aggregate size, as can be seen in Fig. 4.3.
Due to the rebound of aggregates during spraying, the in situ material is slightly richer in
cement than the starting mix. The commencement of cement setting for Portland cement is
usually in the range of 1 to 3 hours. Special rapid hardening tunnel cements are used, when
high early strengths are required during construction. Such cements are an alternative to an
ordinary sprayed concrete with the use of an accelerator, in order to achieve similar effects on
the setting time. Concerning the compressive strength of cement the following requirements
are reported in Hafez (1995):

after 1 day ≥ 7 N/mm2

after 28 days ≥ 39 N/mm2
after 90 days ≥ 50 N/mm2

Nowadays Portland cement is increasingly being used with supplementary cementing mate-
rials, such as fly ash and silica fume, in both the dry and wet mix processes.

89
Figure 4.3: Recommended cement content of shotcrete mix before installation depending on
maximum aggregate size (from Aldrian, 1991)

4.3.3 Water
The water content of a certain mix design plays an important role in the mechanical be-
haviour of shotcrete. The higher the water-cement ratio, the lower the 28-days compressive
strength of concrete, as indicated in Fig. 4.4. This effect is due to the higher porosity which is
obtained after hydration when too much water is used for the mixing. However, pumpability
requirements dictate that the water-cement ratio for shotcrete is slightly higher than the one
for conventional cast concrete (0.3 - 0.4).

90
Figure 4.4: Relationship between compressive strength and water-cement ratio for concrete
(from Karrer, 1986)

4.3.4 Additives
As mentioned before, a variety of additives and admixtures can be added to sprayed concrete
to improve certain properties, such as strength, adhesiveness, durability, permeability and
rebound behaviour. Higher material costs are often balanced by savings in the volume of
shotcrete used and labour.

4.3.4.1 Silica fume

Silica Fume is a highly puzzolanic mineral admixture that has been used mainly to improve
concrete durability and as a Portland cement replacement (Wolsiefer & Morgan, 2003). It is
a waste product from the metal industry and has been primarily used in the United States,
Canada and the Scandinavian countries, but is now finding increasing use elsewhere in the
world. In the 1970s it was first used in shotcrete for a tunnel lining construction in Norway.
Due to its extreme fineness, the particles fill the microscopic voids between the cement par-
ticles, which leads to an increased density and very low permeability of shotcrete. Further,
through the use of silica fume high compressive and flexural strengths are achieved with the
elimination of the need for accelerators for early strengths. Material cost effectiveness is
improved by the reduction of rebound in the dry mix process by up to 50 %. For vertical
and overhead spraying, increased one-pass layer thicknesses are possible without using accel-
erators, thus improving productivity. As a rule of thumb, the dosage of silica fume is given

91
in various international guidelines for shotcrete as ranging from 5 to 15 % of mass of cement
(Austin & Robins, 1995).

4.3.4.2 Accelerators

The setting process of cement can be speeded up by the use of accelerators, which leads to high
early strengths of shotcrete in the first 12 hours after spraying. They are chemical products
usually available in powdery and liquid form and can be used for both the dry- and the wet-
mix processes. In the early days, various water soluble salts of the alkali metals were used
to accelerate the setting of cement, most of them based on alkali aluminates in combination
with carbonates and hydroxides. However, these chemicals were very caustic and exposed
the workmen to a significant danger. The dosage of such an accelerator and its fast reaction
with cement has to be proven before spraying and is normally in the range of 2 to 7 % of
weight of cement (Girmscheid, 2000). Another advantage of using accelerators is the improved
adhesiveness especially for wet-mix shotcrete, which leads to larger application thicknesses.
Despite these facts, there were two main drawbacks concerning the use of accelerators for
shotcrete. As can be seen in Fig. 4.5, the addition of such an accelerator in the concrete mix
can decrease the target 28-day compressive strength of 40 MPa significantly. Secondly, if the
accelerator dosage is too high, an existing shotcrete layer will stiffen and harden too fast and
therefore aggregates of later applied shotcrete cannot penetrate properly into the existing
layer and bounce off. This fact leads to an increased rebound behaviour and loss of material.

Figure 4.5: Time dependent influence of accelerator dosage on compressive strength (from
Kusterle, 1985)

However, in Thomas (2009) it is stated that accelerators of the new generation do not
show a reduction of the 28-day strength of shotcrete. They are safer to use and almost
alkali-free, which reduces the risk of an alkali reaction in the shotcrete significantly. Jodl

92
& Kusterle (1998) report new developments for completely alkali-free accelerators of an alu-
minium hydroxide base, which develop sufficient early strength for even low dosages (4 %)
and no difference in the final compressive strength compared to ordinary shotcrete.

4.3.4.3 Plasticisers

Especially for the wet mix process plasticisers are used for a better workability and transport
of the fresh shotcrete into the tunnel (Austin & Robins, 1995). Therefore, a lower water-
cement ratio can be applied and higher strengths achieved. These types of additives should
not be used for the dry-mix process, as the wetting time at the nozzle is too short and the
reaction of the plasticisers would only start when the shotcrete is already sprayed on the wall.
This results in a running off of shotcrete from the tunnel wall (Girmscheid, 2000).

4.3.5 Fibres
The use of fibre reinforced concrete and shotcrete has advanced substantially in the last
few years and significant progress has been made regarding the quality of involved materials
and application techniques. Due to its advantage of improving certain material properties,
fibre reinforced shotcrete has been widely accepted for the support of underground openings,
rock slope stabilization or thin shell dome constructions. Three different types of fibres are
commonly in use for reinforcing concrete and shotcrete and they will be discussed briefly in
the following sections.

4.3.5.1 Steel fibres

The conventional construction of reinforced shotcrete in underground structures involves the

application of ordinary wire mesh, which is a very time-consuming and heavy manual process.
In the early 1970s, steel fibres were proposed as a new technology for reinforcing shotcrete
tunnel linings, where pioneering developments have been achieved by the Scandinavian coun-
tries and at the Ruhr University at Bochum in Germany (Franzen, 1992). By replacing
welded wire mesh with steel fibres some major economical benefits regarding material and
labour costs can be achieved. However, one of the main reasons for using steel fibres for
shotcrete is the increase in material ductility and flexural strength, which depends heavily
on the amount and type of fibres adopted. Fig. 4.6 contains a compilation of commonly used
types of steel fibres available on the market, showing typical fibre cross-sections on the right
with dimensions in [mm].

93
Figure 4.6: Different types of steel fibres used in sprayed concretes (from Banthia et al., 1992)

Long fibres with a length greater than 25 mm and a rather high dosage between 40 and
75 kg/m3 are preferable, with 60 kg/m3 being commonly used in tunnelling and mining ap-
plications. Some of the critical and important parameters of steel fibres are:

 Geometry

 Length

 Aspect ratio (L/D)

 Steel quality

For a proper mix design the rebound of the steel fibres during the spraying process should
be taken into account and therefore a minimum dosage of 30 kg/m3 is proposed in ASCCT
(2004). Both installation techniques, dry- and wet-mix, are suitable for the application of
steel fibre reinforced concrete. For anchoring reasons, the fibres should be at least twice as
long as the largest aggregate. ASCCT (2004) recommends therefore to limit the maximum
aggregate size for steel fibre reinforced sprayed concrete to 8 mm. Some negative effects of
the steel fibres have to be expected on the pumping and spraying of the mix. Therefore,
steel fibre reinforced shotcrete requires the use of microsilica and admixtures that enhance
pumpability. The fibre length should not exceed 50 to 60 % of the pumping hose diameter.
This results in a maximum fibre length of about 25 mm for manual spraying and 40 mm for
robotic application of the shotcrete.

As mentioned earlier, the aspect ratio, defined as the ratio between the length and the
equivalent fibre diameter (L/D), is of crucial importance for the performance of steel fibres
in sprayed concrete, usually ranging from 40 to 60 (Austin & Robins, 1995). From the

94
mechanical point of view, a higher aspect ratio and volume concentration would be preferable,
but unfortunately this leads to difficulties for the mixing, conveying and shooting of the
shotcrete mixture. It can be observed that there are some practical and economical limitations
for the type and amount of fibres used for reinforcement. In the early days of steel fibre
technology most of the fibres were straight and did not show a good testing response regarding
pull-out resistance. Consequently, manufacturers focused on improving the aspect ratio of
steel fibres by methods such as crimping, introducing bent or enlarged ends or making special
deformations on the fibres.

Finally, it can be concluded that steel fibre reinforced shotcrete is a relatively new but
powerful technology in the field of tunnelling, where major developments have been achieved
over the last 20 years. The continuously increasing use of fibre reinforcement implies enhanced
saftey of underground structures and substantial cost savings during construction (Franzen,
1992).

4.3.5.2 Polypropylene fibres

In the late 1980s polypropylene fibres started to find use in concrete engineering, initially at
very low addition rates of about 1 kg/m3 (Austin & Robins, 1995). Their mechanical proper-
ties were similar to those of plain concrete and therefore they could not enhance the mechan-
ical behaviour significantly. The benefits were limited mainly to providing some resistance
against plastic shrinkage cracking and improved strength of the freshly applied shotcrete.
Another advantage was the reduced amount of rebound when adopting the wet-mix instal-
lation technique. However, recent developments led to new types of plastic fibres made of
high quality materials. If moderately dosed, at about 10 to 13 kg/m3 , it has been shown
that these higher fibre additions in the wet-mix process result in a more ductile material
behaviour, similar to the one obtained with steel fibre reinforced concrete or shotcrete. In
Fig. 4.7 typical fibrillated polypropylene fibres with a length of 20 to 40 mm are illustrated.

Figure 4.7: Polypropylene fibres used in wet-mix sprayed concrete (from Austin & Robins,
1995)

For dry-mix concrete, the incorporation of higher addition rates is still difficult to achieve,

95
since the wetting of the shotcrete mixture at the nozzle is simply not capable of adequately
coating and embedding the plastic fibres in the cementitious mix before the impact at the
surface. As a result the actual in-place fibre content remains relatively low (Austin & Robins,
1995).

4.3.5.3 Glass fibres

A certain concern exists for the use of glass fibres over their long-term durability due to
the alkali attack of the glass within the concrete matrix. Therefore, glass fibres are less
common for permanent shotcrete structures. However, Austin & Robins (1995) mention one
application of the repair to a navigation lock in Washington State.

4.4 Cement hydration

When cement is dispersed in water a chemical reaction is started between the clinker compo-
nents of the cement and the water, which causes hardening of the cement paste. This reaction
progresses at different rates for each of the involved components and the reaction products
also vary accordingly. In Neville (1981) the main clinker compounds of ordinary Portland
cement are given as:

 Tricalcium silicate (C3 S)

 Dicalcium silicate (C2 S)

 Tricalcium aluminate (C3 A)

 Tetracalcium aluminate ferrite (C4 AF )

The following part of this chapter describes briefly the main stages of cement hydration and
their hydration products.

96
4.4.1 Reaction process
According to Fig. 4.8 three stages of cement hydration can be identified.

Stage I Within the first minutes after mixing of cement and water, first needle-shaped
crystals of a calcium sulfoaluminate hydrate called ettringite appear on the
surface of the cement particles. This phase is still dominated by low strengths
of the formed solids and the structure is termed unstable.

Stage II A few hours after starting hydration, large prismatic crystals of calcium hy-
droxide and very small fibrous crystals of calcium silicate hydrates begin to
fill the empty space that was formerly occupied by water and the dissolving
cement particles. The cement paste is setting and a basic structure is built up.

Stage III After some days, the ettringite may become unstable and decompose to form
monosulfate hydrate, which has a morphology of hexagonal plates. The
strength development of the material is advanced and a stable structure has
been established.

Figure 4.8: Schematic representation of the cement hydration (after Labahn, 1982)

Further information about the complex process of cement hydration can be found in
Neville (1981), Byfors (1980) and Mehta & Monteiro (1985).

97
4.4.2 Degree of hydration
In the literature, the degree of hydration is a term often used to quantify how far the reactions
between cement and water have developed, varying between 0 (no reactions have occurred)
and 1 (cement hydration complete). It is also a useful parameter to describe the development
of certain material parameters of the young concrete. However, cement hydration is a very
complex process where various chemical reactions are taking place simultaneously and one
single parameter might sometimes not be enough to describe the complete hydration process
in detail. In Byfors (1980) an overview of different definitions for the degree of hydration ξ
available in the literature is given. Among them are:

Quantity of cement gel formed

ξ= (4.1)
Quantity of cement gel formed at complete hydration

Quantity of hydrated cement

ξ= (4.2)
Original quantity of cement

Quantity of heat developed

ξ= (4.3)
Quantity of heat developed at complete hydration

wn
ξ= (4.4)
0.25 c
where wn is the quantity of bound water and c the cement content. As one can imagine, the
values of these definitions can differ significantly from each other, particularly at an early age
and should therefore be used with caution. The adequate choice of the definition depends
to a considerable extent on the method used for determining the degree of hydration. For
further information on this topic see Byfors (1980).

4.5 Shotcrete installation

Shotcrete can be seen as a special type of concrete, not necessarily because of the mix design
but because of the nature of its installation - casting and compacting are done in one single
step. In this part of the thesis the reader will be introduced to the different installation
techniques of shotcrete, including material preparation, choice of the proper site equipment
and the spraying itself. As mentioned before, sprayed concrete finds many applications in
engineering but the focus here will be on the use of shotcrete in tunnelling, which involves
high volume output. Various production processes have been developed in recent years and
the distinguishing features are the preparation of the starting materials and the method
of transporting and casting the sprayed concrete in the tunnel. The two main dominating
techniques are:

 Dry-mix process

98
  Wet-mix process

There are essential differences between these two methods and the choice of one or the other
depends on many factors of each tunnelling project. The production of shotcrete is mainly
equipment orientated, whereas installation depends heavily on the technique and skills of
the operatives. The dry-mix process represents the original application method, but recently
significant developments and advances have been made to wet-mix technology.

4.5.1 Dry-mix process

In the dry spraying method, a dry mixture of aggregates, cement and occasionally powdery
additives is transported from the spraying machine into the tunnel by compressed air. At
the spraying nozzle, water is added and the mixture can be applied at velocities of about
20 to 30 m/s in order to compact it (Girmscheid, 2000). Modern systems also use liquid
accelerators which are added to the shotcrete upstream of the spraying nozzle. The basic
concept of the dry mix technology in tunnelling is illustrated in Fig. 4.9.

Figure 4.9: Process of dry-mix technology for sprayed concrete (from Austin & Robins, 1995)

In this process, the nozzle operator, or the so called “nozzleman”, has a great influence on
the quality of the shotcrete since the water content is controlled by him manually. Therefore
the water-cement ratio for the dry-mix shotcrete is not a well defined value and normally
lies in the range of 0.35 - 0.5 (Austin & Robins, 1995). If the water content is too low, then
rebound and dust production increases. On the contrary, if the water-cement ratio is too high,
the shotcrete will not stick on the tunnel wall but will run off. A severe problem with dry-
mix shotcrete is the high dust production during spraying which leads to very dirty working
conditions. This high dust level is due to the compressed air and the relatively high percentage
of rebound. A big advantage of the dry-mix technology is its great flexibility. For example,

99
small quantities of sprayed concrete can be produced easily and necessary breaks in spraying
can be bridged without cleaning the equipment. Furthermore, this method ensures transport
over long distances (up to 350 m) without problems, thus eliminating the need for frequent
movement of the equipment which would be very time consuming. Because of producing
shotcrete with higher early strengths due to the lower water-cement ratio compared to the
wet-mix process, the dry-mix technology has been preferred in the past and some countries,
notably Austria, retain this preference (Thomas, 2009).

4.5.2 Wet-mix process

In the wet spraying method, as illustrated in Fig. 4.10, the ready mixed fresh concrete (con-
taining aggregates, cement, water and additives) is transported to the spraying nozzle by
compressed air or pumps. This results in a homogeneous concrete mass and causes less dust,
because the water is already mixed to the concrete at the very beginning. For this reason,
the big advantage of a well defined water-cement ratio can be obtained. The accelerator for
increasing the early strength of the shotcrete can be added at the nozzle, as its setting time
is just a couple of minutes.

Figure 4.10: Process of wet-mix technology for sprayed concrete (from Austin & Robins,
1995)

An important distinction is made in the literature between three methods of transporting

the shotcrete into the tunnel, as shown in Fig. 4.11:

 Thin stream conveying method

 Plug-flow conveying method

 Thick stream conveying method

100
In the thin stream method, similar to the dry-mix, compressed air is admitted to the fresh
concrete in the wet mix machine in order to transport it pneumatically and to spray the
shotcrete with a certain velocity. The difference between this method and the plug-flow
process is that for the second, only the compressed air necessary for transporting the concrete
into the tunnel is admitted to the concrete at the wet machine whereas the air for placement is
supplied at the nozzle. During thick stream transport a compact material stream is pumped
hydraulically from the wet-mix machine to the nozzle, where compressed air is used again to
spray the concrete.

Figure 4.11: Different types of conveying systems for shotcrete (from Girmscheid, 2000)

Some disadvantages of the wet-mix technology are its lack of flexibility, the shorter pos-
sible transport distances and the difficulties in handling the filled spraying hose and nozzle.
Furthermore, the method is not suitable for processing smaller quantities of shotcrete with
numerous breaks in placement, as such stops require the removal of the hardening concrete
and a proper cleaning of the complete spraying unit.

4.5.3 Spraying technique

It is well accepted practice, that shotcrete should be sprayed at right angles to the target sur-
face at a distance of about 0.5 - 2 m in order to get good compaction and minimum overspray
(Austin & Robins, 1995). For overhead spraying the nozzle is kept normally slightly closer
to the surface at about 1 m of distance, because gravity reduces the velocity of the particle
stream and leads to increased rebound. Girmscheid (2000) highlighted in his publication
some key aspects regarding the optimum spraying technique, as illustrated in Fig. 4.12. In
this figure d denotes the shotcrete layer thickness, gSB the weight of the installed shotcrete
layer, h the adhesion between shotcrete and target surface and τ the induced shear force
acting at the interface. As a general rule material should be built up from the bottom, in
order to activate the supporting behaviour of the already applied shotcrete below.

101
Figure 4.12: Preferred application procedure for sprayed concrete (from Girmscheid, 2000)

Furthermore, the movement of the nozzle has a great influence on the quality of shotcrete.
Therefore the nozzle tip should be rotated in such a way that the stream moves in loops across
the surface. Special care should be taken to fully encase reinforcement in order to ensure
full bond between concrete and steel and guarantee long term durability of the shotcrete
structure.

4.5.4 Rebound behaviour

When shotcrete is applied to a tunnel wall some of the material is lost in a number of ways
and is not part of the finished in situ material any more. Austin & Robins (1995) distinguish
three different types of losses that are:

 Overspray = Material missing the target surface

 Rebound = Material which touches the surface but does not adhere

 Cutback = Material that adheres but is subsequently removed whilst plastic to keep
the correct profile

The biggest concern is the loss due to rebound, which is in the order of 20 - 30 % for the
dry-mix process, but only 10 - 20 % for the wet-mix process. The bouncing off of material
is not just an economic loss, but it changes also the in situ proportions of the mix design,
because the individual constituents of shotcrete rebound by different amounts. Obviously
this has an important effect on the mechanical behaviour of shotcrete. The following factors
summarized in Tab. 4.1 play a keyrole in the rebound behaviour of shotcrete:

102
 Concrete mix design Installation technology Boundary conditions

Water-cement ratio Spraying angle Consistency of surface

Cement content Nozzle distance Reinforcement

Grading curve of aggregates Dry- or wet-mix Layer thickness

Dosage of additives Transport system Orientation of surface

Table 4.1: Influencing factors on rebound behaviour

Parker et al. (1977) carried out some experiments on the rebound behaviour of a dry-mix
sprayed concrete with a maximum aggregate size of 13 mm. From their results they concluded
that it is much more economical to spray a single thick layer than two separate thinner layers,
as shown in Fig. 4.13.

Figure 4.13: Effect of layer thickness on material rebound (from Parker et al., 1977)

The influence of the orientation of the target surface on the rebound behaviour was
studied by many researchers, among them Ryan (1975), Gullan (1975) and Kobler (1966). It is
generally accepted that the rebound of shotcrete material increases with increasing orientation
of the target surface compared to the horizontal direction, as illustrated in Fig. 4.14.

103
 Figure 4.14: Effect of surface orientation on material rebound (from Girmscheid, 2000)

No clear trend has been observed in the literature whether the inclusion of fibres in the
shotcrete mix can reduce material rebound during spraying or not. Some researchers report a
reduction in rebound of up to 45 % with steel fibre reinforced sprayed concrete (Ryan, 1975).
However, it is stated that these enormous reductions might be due to other factors in the
spraying process.

Finally, Armelin & Banthia (1998) present in their work a mechanical approach for the
simulation of an aggregate particle rebounding from a fresh concrete surface. This mechanical
model is based on the theory of rebound and aims to provide some fundamental knowledge
on the mechanics of the rebound process. By analysing in detail the particle penetration, the
reaction phase and contact time, they intend to predict the amount and composition of the
aggregate rebound for an actual dry-mix shotcrete.

4.6 Shotcrete testing

The quality control of sprayed concrete during construction is of crucial importance for a
safe tunnel lining design, since the shotcrete quality is highly operative dependent. However,
it is reported in the literature that testing of young shotcrete is a difficult task to perform,
as it is impossible to drill cores until the strength of the material has reached a certain
level - for example a uniaxial compressive strength of 10 MPa according to Chang (1994).
Sometimes shotcrete samples are sprayed into test panels under field conditions, which also
has its shortcomings. In EFNARC (1996) the required tested material properties of sprayed
concrete are listed as follows:

 Uniaxial compressive strength and density

 Flexural and residual strength

 Energy absorption class

 Modulus of elasticity

 Bond strength

104
  Permeability and frost resistance

 Determination of fibre content of shotcrete

Among these material properties, the uniaxial strength in compression is one of the most
important ones for describing material quality of shotcrete and is frequently tested during the
progress of tunnel construction. ASCCT (2004) provides a range of different test methods for
the compressive strength at very early ages, as can be seen in Fig. 4.15. In the first couple of
hours after spraying up to a compressive strength of about 1 MPa, penetration needles with a
diameter of 3 and 9 mm are used. The measured penetration resistance or depth for a constant
applied force allows the estimation of the compressive strength at this early stage of cement
hydration. For shotcrete ages up to 2 days the pull-out test finds its use as a testing technique
for the uniaxial compressive strength. However, it is stated that the measured strength is
somehow a combination of the uniaxial compressive and shear strength with a precision of
± 25 % (Kusterle, 1985). For more or less hardened shotcrete cylindrical core samples taken
from the sprayed concrete structue are tested in order to quantifiy material quality. According
to EFNARC (1996) the minimum diameter should be 50 mm and the height/diameter ratio
should be in the range of 1.0 to 2.0. Alternatively, the uniaxial compressive strength can be
determined from cubes cut from sprayed test panels and their strength values converted to
the cylindrical strength.

Figure 4.15: Test methods for compressive strength of young shotcrete according to ASCCT
(2004)

Chang (1994) proposed a new field testing method that allows the measurement of the
uniaxial compressive strength and the Young’s modulus at a certain time after casting. A

105
steel bar with square cross section is inserted into a shotcrete layer immediately after in-
stallation. Fig. 4.16 illustrates the basic principle of the method, where the two quantities
under investigation can be estimated from the applied torque and the registered angular
displacement. The uniaxial compressive strength σc and the Young’s modulus Ec are then
calculated from the maximum torque Mmax , the stiffness kt and the two laboratory calibrated
parameters α and β.

Figure 4.16: Proposed test method for uniaxial compressive strength and Young’s modulus
by Chang (1994)

4.7 Summary
This chapter started with an introduction into the development of sprayed concrete from a
historical point of view. An overview of different applications of shotcrete in civil engineering
was then given, focusing on the use of shotcrete in subsurface construction such as mining and
tunnelling. Regarding the material technology the following key aspects can be summarised:

 As any conventional cast concrete, shotcrete consists of cement, aggregates and water.
Each of these components has an important impact on the mechanical behaviour of the
shotcrete and has to be addressed in a proper mix design. Additives such as accelerators,
plasticisers and silica fume are commonly used to change certain material properties,
but they have to be treated with care. Advanced developments have been recently
achieved for using steel and polypropylene fibres as a reinforcement replacement for
ordinary steel wire meshes.

 Cement hydration is a complex chemical reaction between the clinker components of the
cement and water and leads to setting and hardening of the cement paste. Occasionally
the so called “degree of hydration” is used in the literature to describe the development
of certain material parameters such as stiffness and strength.

 For the installation of shotcrete two different techniques exist - the dry-mix and the
wet-mix process. Both technologies differ significantly in material preparation, but have

106
 in common that placing and compacting of the shotcrete is done in one single step by
spraying the material at high pressures onto the target surface. Certain advantages and
disadvantages exist for each technique, but the tendency in tunnelling is clearly towards
the wet-mix process due to its better control of the shotcrete quality (in particular of
the water-cement ratio).

 The spraying technique is of great importance for achieving good material quality and
this requires excellent skills and training of the nozzle-operator. Rebound of shotcrete
is not only an economical loss but influences as well the mechanical behaviour of the
material. It leads to a slightly richer cement mix of the in-situ material and this has to
be taken into account in the mix design.

 Testing of young sprayed concrete at early ages is a difficult task to perform and no
commonly agreed techniques exist in the literature. Usually, shotcrete samples are
taken from the tunnel lining or are sprayed into boxes under field conditions, where the
compressive strength of the material is the main parameter of interest in the design.
For fresh concrete penetration needles and pull-out tests are often used to estimate the
compressive strength of the material at early ages of cement hydration.

107
Chapter 5

Mechanical and time-dependent

behaviour of concrete and shotcrete

5.1 Introduction
In the previous chapter shotcrete as an engineering material has been introduced, highlighting
the influence of material technology, such as mix design, installation technique and rebound
behaviour, on the mechanical behaviour of sprayed concrete. Therefore, the current chapter
deals at first with the fundamental description of the mechanical behaviour of hardened con-
crete, starting with simple uniaxial conditions in compression and tension. For completeness,
it is further explained how concrete behaves mechanically under multiaxial stress states that
are nearly always encountered in engineering practice. After a brief introduction into the be-
haviour of steel fibre reinforced concrete, the mechanical description of concrete is extended to
the time-dependent behaviour of shotcrete, emphasising the increase in stiffness and strength
with time, the deformability at early ages and wether or not early-age loading has any damage
effects on the young shotcrete. Other important time aspects such, as creep and relaxation,
shrinkage and temperature induced strains due to cement hydration, are explained in detail
and their influence on the mechanical behaviour of a tunnel lining shell discussed.

5.2 Uniaxial stress conditions

5.2.1 Mechanical behaviour in compression
In a typical uniaxial compression test on cylindrical or cubical concrete samples the following
stress-strain relationship, illustrated in Fig. 5.1 as normalised stress versus axial strain, is
usually observed: up to about 30 % of its maximum compressive strength, fcp, the behaviour
of concrete can be assumed nearly linear elastic and the microcracks already existing in the
concrete before loading remain unchanged (Chen, 1982). This indicates that the available
internal energy is less than the energy required to create new microcrack surfaces. For
stresses above this proportionality limit, the shape of the stress-strain curve is closely related

108
to the mechanism of internal progressive microcracking. With further loading bond cracks
start to extend due to stress concentrations at the crack tips and the stress-strain curve
becomes more and more non-linear. Between 30 to 50 % of fcp the available internal energy is
approximately balanced by the required crack-release energy and crack propagation is stable.
For compressive stresses above 70 % of fcp the system is unstable and crack propagation will
increase. In Mehta & Monteiro (1985) the special stress level of about 75 % of fcp is termed
“critical stress” since it corresponds to the minimum value of volumetric strain, as seen in
Fig. 5.1 (assuming a compression positive sign convention).

Figure 5.1: Typical plots of stress-strain curves for uniaxial compression (from Chen, 1982)
(fc = fcp)

If loading is continued, the stress-strain curve will bend more sharply and will approach
the peak point at fcp. Axial compressive peak strains are usually close to a value of 0.2 %.
Furthermore, the sign of the volume change is reversed, resulting in a volumetric expansion
(dilatancy) near the compressive strength fcp (Chen, 1982). After reaching the peak stress,
the stress-strain curve shows a descending part, which is very much dependent on the test
conditions and the equipment used. The influence of the length of the tested concrete sample
is shown in Fig. 5.2, where an increasing specimen length leads to a more brittle material
behaviour after peak. Crushing failure of concrete occurs finally at some ultimate axial
compressive strain εu , where the stresses abruptly reduce to zero. Swoboda & Moussa (1992)
suggested a value for the compressive failure strength of shotcrete, just prior to crushing,
between 0.5 and 1.0 times the uniaxial compressive strength fcp.

109
Figure 5.2: Influence of specimen length on post peak behaviour in uniaxial compression
(from CEB-FIP Model Code, 1990)

i) Modulus of elasticity
The initial Young’s modulus of elasticity, E, in compression is highly dependent on the
uniaxial compressive strength fcp and can be estimated for normal weight concrete by some
empirical equations published in the literature (Chen, 1982). The CEB-FIP Model Code
(1990) provides the following expression:
 1
fck + Δf 3
Eci = Eco (5.1)
fcmo

where Eci is the modulus of elasticity (MPa) at a concrete age of 28 days, Δf = 8 MPa,
fcmo = 10 MPa and Eco = 2.15 × 104 MPa. The characteristic uniaxial compressive strength
fck is defined as that strength below which 5 % of all possible strength measurements for
the specified concrete may be expected to fall. Generally, the Young’s modulus E and the
uniaxial compressive strength fcp are well defined parameters in various national standards
and can be taken according to a certain strength class, see EC 2 (2004).

ii) Poisson’s ratio

According to Neville (1981), for concrete under uniaxial compressive loading the Poisson’s
ratio μ takes a value between 0.11 and 0.21. As illustrated in Fig. 5.3, during uniaxial
compression loading μ remains constant until approximately 75 to 80 % of the compressive
strength fcp . At this stress level μ starts to increase and in the unstable crushing phase can
even become larger than 0.5 (Chen, 1982).

110
Figure 5.3: Relation between stress level and Poisson’s ratio μ (from Chen, 1982) (fc = fcp )

5.2.2 Mechanical behaviour in tension

Typical stress-strain curves for concrete in uniaxial tension are given in Fig. 5.4.

Figure 5.4: Stress-strain curves in tension for different concrete types (from Chen, 1982)

Most of the curves (which represent different concrete types) show an almost linear elastic
behaviour up to 70 % of the uniaxial tensile strength ftp . In this stress region the creation
of new microcracks within the concrete sample is negligible. With further tensile loading

111
bond microcracks start to occur and the tensile behaviour becomes non-linear. At about
75 % of the tensile strength, ftp , the crack propagation becomes unstable and the initiation
and growth of new microcracks is very fast (Chen, 1982). On reaching the uniaxial tensile
strength ftp , much larger and visible cracks at the surface start to appear and the concrete
sample is termed to be “cracked”. Peak strains for uniaxial tension are considerably smaller
than those under uniaxial compression and lie in the range of 0.005 to 0.02 %. Due to the very
brittle behaviour of concrete under tensile stresses, it is difficult to follow the descending part
of the stress-strain curve in an experiment after reaching the peak stress ftp . It is somehow
obvious that cracking in concrete is a very discrete problem and deformations localise in a
small fracture zone. Outside this fracture zone the material exhibits elastic unloading upon
further crack opening. Another interesting phenomenon during crack formation is the fact
that, due to the rough crack surfaces, opposite faces will be able to interlock and to transfer
a certain amount of shear force, an effect which is often termed “aggregate interlocking”.

i) Tensile strength
Usually the ratio between uniaxial tensile and compressive strengths is in the range of 0.07
to 0.11 (Mehta & Monteiro, 1985), as can be seen in Fig. 5.5, which contains a compilation
of various uniaxial experimental test data.

Figure 5.5: Relation between uniaxial tensile and compressive strength (from Byfors, 1980)
(fct = ftp , fcube = fcp )

Some of the equations found in the literature relating the uniaxial tensile strength, ftp ,
to the uniaxial compressive strength, fcp , are given as follows:

The CEB-FIP Model Code (1990) provides the following expression for the mean tensile
strength, fctm , associated with the characteristic uniaxial compressive strength, fck :
 2
fck 3
fctm = fctko,m (5.2)
fcko

112
with fctko,m = 1.40 MPa and fcko = 10 MPa. The idea of defining a “mean” value of the
tensile strength is to provide an estimate for the uniaxial tensile strength of concrete, which
usually shows a great variability due to various influencing factors.
Swoboda & Moussa (1992) used in their work the following equation for calculating the
uniaxial tensile strength of shotcrete:
2
3
ftp = 0.21 fcp (5.3)

Finally, Meschke (1996) adopted in his material model for shotcrete an empirical equation
published by Oluokun et al. (1991b), which is of the form:

0.79
ftp = 0.876 fcp (5.4)

The dimension of the strength parameters in this equation is kN/m2 . As an example, for a
given uniaxial compressive strength of 30 MPa, the results for the uniaxial tensile strength
obtained from the above introduced equations 5.2 to 5.4 range from 2.03 MPa to 3.01 MPa.
However, the uniaxial tensile strength of concrete is a parameter which is difficult to measure
from the experimental point of view. The three methods that are commonly in use for
determining the tensile strength are:
 Splitting test

 Ring test

 Flexural beam test

CEB-FIP Model Code (1990) allows the tensile strength obtained from a splitting test to be
converted into the uniaxial tensile strength in the following way:

fctm = 0.9 fct,sp (5.5)

where fctm is the mean axial tensile strength and fct,sp is the mean splitting tensile strength.
Likewise, if the tensile strength is obtained from a flexural beam test, the equation below can
be used:
1.5 (hb /ho )0.7
fctm = fct,f l (5.6)
1 + 1.5 (hb /ho )0.7
where fct,f l is the mean flexural tensile strength, hb is the depth of the beam and ho = 100 mm.

According to Chen (1982), the modulus of elasticity E in uniaxial tension is slightly higher
and the Poisson’s ratio μ somewhat lower than in uniaxial compression.

5.3 Biaxial stress conditions

In a real structure a purely uniaxial stress state is very rare and most commonly multiaxial
stress states are encountered. In the 1970s many investigations were performed on the me-

113
chanical properties of concrete under biaxial loading conditions, in order to obtain a better
understanding of strength, deformational characteristics and cracking behaviour of concrete
under such a biaxial stress state. A typical experimental stress-strain curve for concrete under
various biaxial stress states is shown in Fig. 5.6 (strains are given in absolute values).

Figure 5.6: Stress-strain relationships for concrete under biaxial compression (from Chen,
1982) (fc = fcp )

It can be seen clearly that, if some lateral pressure is applied, the maximum compressive
strength increases. For the special stress ratio of σ1 /σ2 ∼
= 0.5 a maximum increase in strength
of about 25 % is achieved (Huber, 2006). This value is reduced to about 16 % at an equal bi-
axial compression state with σ1 /σ2 = 1.0. Under the combination of compression and tension
the compressive strength decreases almost linearily as the tensile stress is increased. For a
biaxial tensile stress state the tensile strength remains almost unchanged and is independent
of the applied stress ratio. A range of biaxial tests performed by Kupfer & Gerstle (1973) at
different stress ratios (σ1 /σ2 ) lead to a biaxial failure envelope, as can be seen in Fig. 5.7.

114
 Figure 5.7: Biaxial failure envelope (from Kupfer & Gerstle, 1973) (βp = fcp )

Some information about different fracture modes of concrete in biaxial stress conditions
can be found in Nelissen (1972). Furthermore, Chen (1982) points out that the stress-strain
curves under biaxial loading conditions are different depending on whether the stress states
are compressive or tensile.

According to EC 2 (2004), confinement of concrete results in a modification of the stress-

strain relationship, where higher strengths and higher critical strains are achieved, whereas
the other material characteristics may be considered unaffected. The characteristic confined
compressive strength, fck,c, can be calculated from:
 
σ2
fck,c = fck 1.125 + 2.50 (5.7)
fck

where σ2 (= σ3 ) is the effective lateral compressive stress and fck the characteristic uniaxial
compressive strength. The peak strain for confined stress conditions, εc2,c , i.e. the strain
when the characteristic confined compressive strength is fully mobilised, is given as:
 2
fck,c
εc2,c = εc2 (5.8)
fck

with εc2 being the uniaxial compressive peak strain. In a similar way, the ultimate confined
compressive strain εcu2,c , where the concrete is assumed to be crushed, can be estimated
from:
σ2
εcu2,c = εcu2 + 0.2 (5.9)
fck

115
5.4 Triaxial stress conditions
The behaviour of concrete under triaxial stress conditions was investigated, for instance, by
Kotsovos & Newman (1978) and van Mier et al. (1987). Similar to biaxial tests, their results
showed an increase in compressive strength with increasing confining pressure. Balmer (1949)
conducted some triaxial tests at very high confining stress levels and reported compressive
strengths of up to 560 MPa. All these experiments indicate that concrete has a fairly constant
shape of the failure surface which is a function of the three principal stresses. With the
assumption of isotropic material behaviour, the elastic limit (i.e. yield surface) and the
failure criterion can be represented as smooth surfaces in principal stress space as shown
in Fig. 5.8. For small hydrostatic pressures, the deviatoric cross sections of the surfaces are
convex and have the shape of rounded triangles. With increasing hydrostatic compression
the deviatoric cross sections become more and more circular, which means that failure in this
region is independent of the Lode angle θ. The intersection curves between the yield and
failure surfaces and a plane containing the hydrostatic axes with θ = const. are often termed
“meridians”. The general character of these meridians for the failure of concrete is that they
are smooth, curved and convex lines, depending on the hydrostatic stress component I1 or p.

Figure 5.8: Triaxial yield and failure surface for concrete (from Chen, 1982)

5.5 Anisotropy of shotcrete

Generally speaking, concrete can be considered as a homogeneous mixture of aggregates,
cement and water and is therefore commonly treated as an isotropic material. Anisotropy in
shotcrete can be expected due to the spraying process and various uncertainties in material
technology and curing conditions (Aldrian, 1991). Furthermore, imperfections of the tunnel
shell can introduce a certain anisotropy in the material behaviour that can lead to unex-
pected loading conditions (Stelzer & Golser, 2002). For experiments regarding the strength
development of shotcrete, it is very important to take into account the spraying direction
for sample preparation. In standard tests on shotcrete cores the loading direction would

116
be the same as the spraying direction, which is the opposite in a real tunnel structure, as
can be seen in Fig. 5.9. Within the tunnel lining the major compressive stresses would act
perpendicular to the spraying direction. According to Thomas (2003), the quantification of
this anisotropy for shotcrete in the literature seems to be difficult to establish. Huber (1991)
and Fischnaller (1992) tested the early strength of shotcrete with two loading directions and
came to the conclusion that the samples that where tested parallel to the spraying direction
had a reduced strength of 20 % compared to the samples tested perpendicular to the spraying
direction. Steel fibre reinforced concrete exhibits pronounced anisotropy in its behaviour for
both compression and tension (Thomas, 2003). However, anisotropy effects of shotcrete are
usually ignored in a numerical analysis for the purpose of simplicity.

Figure 5.9: Spraying direction and loading conditions in a real tunnel lining

5.6 Mechanical behaviour of steel fibre reinforced concrete

and shotcrete
As already mentioned in the previous chapter 4, the mechanical properties of concrete can
be greatly improved by the addition of steel fibres (Strack, 2007). It is therefore possible to
convert concrete from a quasi-brittle material behaviour to a more ductile one. However, it
is well known that in an experiment the type, mechanical properties, shape and distribution
of the fibres within the concrete matrix play an important role in the obtained material
response, as can be seen in Fig. 5.10.

117
Figure 5.10: Failure mechanism for steel fibre renforced shotcrete and concrete (from Ding
& Kusterle, 2000)

For shotcrete, (Fig. 5.10 on the left) the fibre distribution is mainly two-dimensional due
to the spraying process and in a tunnel lining the main compressive loads occur parallel
to the fibre orientation. For this particular failure mode steel fibres can only improve the
cracking behaviour in zones h1 and h3 , whereas in zone h2 the fibres do not affect the material
behaviour. The situation is different for conventional cast concrete reinforced with steel fibres,
where a homogeneous three-dimensional fibre distribution can be expected.

The uniaxial behaviour of hardened fibre reinforced concrete in tension can be best ex-
plained by the load-deformation curve in Fig. 5.11.

Figure 5.11: Uniaxial stress-deformation characteristics of fibre reinforced concrete (from

Thomée, 2005)

For tensile loading up to point A (about 60 to 90 % of the uniaxial peak strength ftp )
linear elastic material behaviour can be observed, governed by the properties of the plain
concrete matrix. With the low fibre content common in engineering practice due to economic
reasons, there is no noticeable influence on the Young’s modulus. After a non-linear loading
phase where micro crackbands start to develop, a small increase in the uniaxial tensile peak
strength, ftp , might be possible (point B). In the post peak regime firstly a steep drop in

118
stresses occurs with increasing deformation, controlled by the behaviour of the plain concrete
matrix. However, when crack opening continues to develop, the steel fibres start to contribute
to the load-carrying capacity by transferring tensile stresses across the crack. At point C the
reduction in tensile stresses starts to stabilise and a residual tensile strength can be identified.
Therefore a more ductile material behaviour in tension is obtained.

The mechanical behaviour of steel fibre reinforced concrete in uniaxial compression shows
a similar pattern, as illustrated in Fig. 5.12. The elastic limit appears to be unchanged,
located in the range of up to 30 % of the uniaxial compressive peak strength, fcp. For
hardened concrete the incorporation of steel fibres seems to have almost no influence on
the uniaxial compressive peak strength, fcp , compared to plain concrete. However, the strain
when the peak strength is reached appears to be increased for an increased fibre content. The
biggest difference is now noticeable for the post peak behaviour, where a significant increase
in the absorbed energy leads to a fairly ductile material behaviour. The reason behind this
phenomenon might be that the steel fibres prohibit the lateral elongation which results in a
more controlled crack formation in the post peak zone.

Figure 5.12: Compressive stress-strain curves for fibre reinforced concrete (from Strack, 2007)

Little information is available in the literature about the influence of steel fibres on the
behaviour of shotcrete at early ages. Ding & Kusterle (2000) performed uniaxial compression
tests on concrete samples with different fibre content (20, 40 and 60 kg/m3 ) at a starting age
ranging from 8 to 72 h. They observed that steel fibres can improve the compressive strength
of young concrete, as can be seen in Fig. 5.13, although this improvement in strength does
not necessarily increase with a larger dosage of fibres.

119
Figure 5.13: Stress-strain curves for plain concrete and steel fibre reinforced concrete with
60 kg/m3 at an age of 9 h (from Ding & Kusterle, 2000)

For concrete older than 30 h this influence on the compressive strength seems to cease.
The softening part in the post peak regime descends for both concrete types (plain and
steel-fibre reinforced) at the early age in a similar manner. Regarding the energy absorption
capacity, it can be mentioned that at early ages steel fibres have a great influence not only on
the post peak behaviour but improve as well the ascending part of the stress-strain curve. In
Fig. 5.13 an increase of energy absorption until peak of up to 25 % was observed. A similar
trend was obtained by Ding & Kusterle (1999) for early age concrete panel tests reinforced
with steel fibres and conventional wire mesh.

Camps et al. (2008) investigated the effects of steel fibre reinforcement during cement
hydration by uniaxial tension tests at a concrete age of 2, 7 and 28 days. As expected, the
plain concrete specimens demonstrated very brittle material behaviour in the tensile post-
peak regime for all the different stages of cement hardening. The material response of the
concrete samples reinforced with 85 kg/m3 of steel fibres followed the basic behaviour for
hardened steel fibre reinforced concrete explained above. Much higher strain levels were
achieved than for plain concrete and an increase in ductility was observed. Furthermore, the
residual tensile capacity appears to be increasing with hydration development. This post-
peak strengthening effect can probably be explained by the improvement of the bond between
fibre and concrete matrix during cement hardening.

5.7 Time-dependent mechanical behaviour of young shotcrete

For conventional concrete after 28 days, most of the material parameters mentioned in the
previous sections for describing the mechanical behaviour are assumed to be constant and
well defined values in national standards, which is not the case for shotcrete. It can be
expected that as a consequence of the hydration process within the cement paste, the material
properties of young shotcrete gradually change and these early age properties control the
mechanical interaction between the shotcrete lining of a tunnel and the surrounding soil. In

120
the hardening cement paste a gradual increase of stiffness and strength is accompanied by a
rise in temperature caused by the generated hydration energy. Furthermore, the mechanisms
of creep, relaxation and shrinkage can lead to complex stress states within a tunnel lining.
Unfortunately, all these time-dependent effects of young shotcrete should not be treated
seperately from each other, as they occur as a coupled mechanical process, which is not
straightforward to understand and/or quantify.

Complete stress-strain curves for shotcrete under uniaxial compression at different ages
were presented by Sezaki et al. (1989). In Fig. 5.14 it can be seen clearly that, in the first 6
hours after installation, shotcrete behaves as a very soft, plastic material with low strength
and stiffness. Nevertheless, it is able to resist very high peak and ultimate strains, which
are both in the range of per cent. This means that young shotcrete has a high ability to
accommodate tunnel deformations that occur after excavation. Shotcrete samples at an age
of 12 hours up to 3 days still show a ductile material behaviour, but stiffness and strength
have already increased significantly. With further curing time the behaviour becomes more
and more brittle because of the cement hardening, till the shotcrete has its highest stiffness
and strength after 28 days.

Figure 5.14: Stress-strain curves in compression at different shotcrete ages (from Sezaki et al.,
1989)

5.7.1 Increase in stiffness and strength

The development of stiffness and strength with time due to hardening of the shotcrete is
well reported and investigated in the literature by many researchers. Furthermore, this time-
dependency is one of the most influential factors on the axial forces and bending moments in
a shotcrete shell and therefore of crucial importance for tunnel lining design (Thomas, 2003).
It depends highly on the mix design of the particular shotcrete used for a tunnelling project

121
and the use of accelerators, speeding up the hardening process, plays an important role.

i) Young’s modulus
A typical development of the Young’s modulus, E, with time is shown in Fig. 5.15, including
results from Huber (1991) and Fischnaller (1992) who investigated the mechanical properties
of shotcrete at very early ages. In such an experimental work, usually the water-cement ratio,
temperature and cement type are varied in order to study the growth of the Young’s modulus
E with time (Byfors, 1980).

Figure 5.15: Development of Young’s modulus E with shotcrete age obtained from various
tests (from Chang, 1994)

A quantitative estimation of the increase in stiffness with time is possible with the help
of various empirical equations found in the literature. Weber (1979) published an equation
of the following form: * c +
E (t) = a E28 exp (5.10)
t0.6
with E28 being the stiffness for hardened shotcrete at 28 days and t the time in days. The
material parameters a and c depend on the type of cement used in the mix design (i.e.
cement classes) and can be estimated from Tab. 5.1. Schubert (1988) describes the variation

Cement class Z 25 Z 35 F Z 55
Z 35 L Z 45 F
Z 45 L
a 1.132 1.084 1.062
c -0.915 -0.596 -0.445

Table 5.1: Parameters a and c for increase in Young’s modulus E with time (after Weber,
1979)

122
of stiffness with time, using the stiffness at 28 days E28 as a single input parameter, as:

t
E (t) = E28 (5.11)
4.2 + 0.85 t

Here again t is the age of shotcrete in days. Another empirical formulation based on the test
results from Huber (1991) and Fischnaller (1992) and published by Chang (1994) is:
 
−0.446
E (t) = 1.062 E28 exp (5.12)
t0.7

with E28 being the Young’s modulus of shotrete in GPa at 28 days and t the time in days.
In the study of Aydan et al. (1992) the following relationship is adopted:
 
E (t) = 5000 1 − exp−0.42 t (5.13)

The CEB-FIP Model Code (1990) provides the following equation to estimate the increase
in stiffness with time: -   .0.5
t28
E (t) = E28 exp s 1 − (5.14)
t

with t28 being the time at 28 days. The cement parameter s governs the increase in stiffness
and can be taken according to the type of cement as:

 s = 0.2 for rapid hardening high strength cements RS

 s = 0.25 for normal and rapid hardening cements N and R

 s = 0.38 for slowly hardening cements SL

Byfors (1980) related the increase in stiffness to the development of the uniaxial compressive
strength, fcp , with time via the mathematical expression given as:
a1
Eo fcp
E (t) = (a −a2 )
(5.15)
1 + A fcp 1

where Eo , A, a1 and a2 are material constants fitted to experimental results. Fig. 5.16 shows
the experimental relation between Young’s modulus E and uniaxial compressive strength fcp
on a linear scale. These test data show that the Young’s modulus grows more rapidly than the
compressive strength. This is an important phenomenon in tunnelling since a higher increase
in stiffness may lead to a higher loading rate of the shotcrete shell and higher strengths would
be needed to sustain these stresses (Chang, 1994).

123
Figure 5.16: Relationship between stiffness and strength (from Byfors, 1980) (Ecc = E,
fcc = fcp )

Fig. 5.17 compares graphically some of the earlier presented equations found in the literature
for the prediction of the increase of the shotcrete stiffness with time. A value for the Young’s
modulus at 28 days of 30 GPa has been adopted. Although relatively similar at first sight,
some major differences are obtained, particularly at early ages. At a shotcrete age of 12
hours, the Young’s modulus according to Weber (1979) and CEB-FIP Model Code (1990) is
relatively high, taking a value of 16.2 and 15.7 GPa respectively. This implies that the stiffness
of the shotcrete has reached already 50 % of its final stiffness at 28 days. The predictions from
Schubert (1988) and Chang (1994) are significantly lower for this stage of cement hydration
and values of 9.7 and 9.5 GPa are obtained. However, it should be pointed out that the
adopted cement class in equation 5.10 and 5.14 plays an important role in the development
of the shotcrete stiffness with time.

124
 32

28

24

Young's modulus [GPa]

20

16

12

8 Weber (Z 55)
Schubert
4
Chang
CEB-FIP Model Code (s=0.2)
0
0 4 8 12 16 20 24 28
Shotcrete age [d]

Figure 5.17: Increase in stiffness with time according to various equations from the literature

ii) Poisson’s ratio

The Poisson’s ratio μ of early age concrete and shotcrete has received very little attention
in the literature and test results are therefore rare. Byfors (1980) states that the influencing
factors for the development of the Poisson’s ratio with time are the type and quantity of
aggregates used in the concrete mix. Oluokun et al. (1991a) performed an experimental
study on the early age behaviour of concrete and for various concrete mixes, with a water-
cement ratio ranging from 0.388 to 0.763, they came to the conclusion that the Poisson’s ratio
appears to be insensitive to steam curing. Furthermore, the obtained values were generally
the same for all ages and conditions of curing. However, Aydan et al. (1992) showed some
results for the variation of the Poisson’s ratio of shotcrete with time, which are of a different
nature, as can be seen in Fig. 5.18. At early ages μ seems to be very high, with values of
about 0.45. Within the first 3 days after spraying, this value reduces significantly to an
almost constant value of 0.18, which is also predicted by the equation given by Aydan et al.
(1992):
μ (t) = 0.18 + 0.32 exp−5.6 t (5.16)

where the time t is in days.

125
 Figure 5.18: Variation of Poisson’s ratio with time (from Aydan et al., 1992) (ν = μ)

Test results from Byfors (1980) indicate that the variation of the Poisson’s ratio is linked
to the development of the uniaxial compressive strength, fcp . Fig. 5.19 shows that μ decreases
very rapidly during the early stage when the strength is still small, starting from a value of
0.48. At a uniaxial compressive strength of around 1 N/mm2 μ has its minimum value of
about 0.15 and increases afterwards, when hardening of the cement paste continues.

Figure 5.19: Relation between Poisson’s ratio and compressive strength (from Byfors, 1980)
(fcc = fcp , νcc = μ)

126
 A mathematical expression of the form given below is proposed, that couples the devel-
opment of the Poisson’s ratio with the increase in uniaxial compressive strength fcp.

b
μ (t) = a fcp (5.17)

where a and b are material constants. For the range of tests considered by Byfors (1980), the
proposed values for these parameters are summarized in Tab. 5.2:

Parameter fcp ≤ 1 MPa fcp > 1 MPa

a 0.148 0.098 to 0.164
b -0.486 0.102 to 0.281

Table 5.2: Parameters a and b for development of Poisson’s ratio μ (after Byfors, 1980

iii) Compressive strength

Similar to the Young’s modulus, the growth of the uniaxial compressive strength, fcp , with
time has been widely studied in the literature for early age concrete, including both dry- and
wet-mix shotcrete. Fig. 5.20 shows some data from the literature regarding the development
of the uniaxial compressive strength with time.

Figure 5.20: Development of uniaxial compressive strength, fcp, with shotcrete age obtained
from various tests (from Chang, 1994)

Again, the large influence of mix design and installation technique results in a certain
scatter of published test data, that makes it difficult to calculate the strength at a certain
shotcrete age with accuracy. However, a variety of equations can be found in the literature
that describe the increase in strength with time. In general, these expressions should be
treated with care, not to be taken as an absolute value, and always validated against actual
test data for each tunnelling project.

127
Weber (1979) proposed the following equation:
* c +
fcp (t) = a fcp,28 exp (5.18)
t0.55

where t is the shotcrete age in days and fcp,28 the uniaxial compressive strength for hardened
shotcrete at 28 days. The parameters a and c are material constants depending on the cement
type (i.e. cement class) and the water-cement ratio. For different cement types their values
are given in Tab. 5.3.

Cement class Z 25 Z 35 F Z 55
Z 35 L Z 45 F
Z 45 L
a 1.45 1.27 1.20
c -2.32 -1.49 -1.14

Table 5.3: Parameters a and c for increase in uniaxial compressive strength, fcp , with time
(after Weber, 1979)

In the constitutive model for shotrete developed by Meschke et al. (1996), two different
functions for describing the increase in uniaxial compressive strength with time were applied:
 0.72453
t + 0.12
fcp (t) = fcp,1 for t < 24 h (5.19)
24

where fcp,1 is the uniaxial compressive strength after 24 hours and t the time in hours. For
t > 24 h the following empirical relation is adopted:
 
bc
fcp (t) = ac exp − for t > 24 h (5.20)
t

The determination of the constants ac and bc is based on the values for the uniaxial compres-
sive strength at 24 hours and at 28 days. They can be obtained as:

fcp,28 672
ac = and bc = − ln(κ) (5.21)
exp (ln(κ)/27) 27

with κ = fcp,1/fcp,28 and fcp,28 is the uniaxial compressive strength at 28 days. Research from
Chang (1994) includes a curve to match published data from the literature. The following
expression is proposed:  
−0.743
fcp = 1.105 fcp,28 exp (5.22)
t0.7
Golser et al. (1990) adopted an equation that uses just one single input parameter, the

128
uniaxial compressive strength at 28 days. It can be written as:

t
fcp(t) = fcp,28 (5.23)
101 + 0.85 t

The CEB-FIP Model Code (1990) uses the following expression:

-   .
t28
fcp(t) = fcp,28 exp s 1 − (5.24)
t

where s is a cement parameter controlling how fast the strength increases with time. s
depends on the cement type and takes the same values as used to determine the Young’s
modulus (see equation 5.14), namely:

 s = 0.2 for rapid hardening high strength cements RS

 s = 0.25 for normal and rapid hardening cements N and R

 s = 0.38 for slowly hardening cements SL

A graphical representation of various of the above presented equations for predicting the
development of the compressive strength with time can be seen in Fig. 5.21. In this example,
the compressive strength of hardened shotcrete at 28 days was assumed to be 30 MPa, which
is even exceeded by the proposed equation from Chang (1994).

32

28
Compressive strength [MPa]

24

20

16

12

8 Weber (Z 25)
Golser
Chang
4
CEB-FIP Model Code (s=0.2)
0
0 4 8 12 16 20 24 28
Shotcrete age [d]

Figure 5.21: Development of uniaxial compressive strength with time according to various
equations from the literature

For the stability of a shotcrete tunnel lining during construction, the gain in shotcrete
strength with time is of crucial importance. In ASCCT (2004) three early strength classes,

129
J1 , J2 and J3 , are defined for young shotcrete, that is concrete at an age less than 24 h.
As mentioned in chapter 4, for vertical and overhead application in a tunnel, the increase
in strength controls its ability to stick to the surface. For spraying shotcrete to the tunnel
crown, a strength of 0.1 to 0.2 MPa after 2 min is required. At the same time, to minimise
rebound and dust, the early strength of shotcrete should not be bigger than 0.2 MPa after
2 min. However, in the case of a large amount of groundwater penetrating into the tunnel,
an increased early strength is necessary (ASCCT, 2004). For each of these early-age strength
classes certain requirements exist and they are defined as follows:

Class J1 : Shotcrete of this early-strength class is used for the application of thin layers
on dry surfaces. It is not meant for structural use and has the advantage of
low dust production.

Class J2 : For the spraying of thick layers on vertical or overhead surfaces with struc-
tural use, shotcrete has to fulfill the requirements of this early-strength class.

Class J3 : Due to the high dust production and rebound, shotcrete of Class J3 should
just be used when absolutely necessary (high ground water pressures, very
fast tunnel advance, etc.)

In Fig. 5.22 the minimum strengths of each class J1 , J2 and J3 (limited by the lines A, B and
C) are defined graphically for young shotcrete up to an age of 24 hours.

Figure 5.22: Early-age strength classes for young shotcrete (after ASCCT, 2004)

iv) Tensile strength

The tensile strength in general is rarely considered in the design of a tunnel lining according to
national standards (Thomas, 2009) and therefore test data for young shotcrete is hardly avail-
able in the literature. Byfors (1980) states that the growth of the tensile strength with time
is influenced basically by the same factors as those which govern the compressive strength.

130
Kasai et al. (1971) presented results showing how the relationship between compressive and
tensile strength changes at an early age, as can be seen in Fig. 5.23.

Figure 5.23: Ratio between tensile and compressive strength at early concrete ages (from
Kasai et al., 1971) (Ft = ftp , Fc = fcp )

These observations are in conflict with experiments by Weigler (1974), where the tensile
strength grows linearily with the compressive strength at a constant strength ratio of about
0.08.

Byfors (1980) performed some tests on young concrete and described the relationship
between the uniaxial tensile strength, ftp , and the uniaxial compressive strength, fcp , at
early ages with the following equation:

ftp (t) = 0.082 fcp (t)1.09 (5.25)

This expression has an exponent close to 1.0 and consequently an almost linear relation
appears to occur for young concrete, resulting in a similar increase in uniaxial tensile strength
with time as for the uniaxial compressive strength. Meschke et al. (1996) adopted in their
work an empirical relation which can be written as:

ftp (t) = 0.876 fcp (t)0.79 (5.26)

v) Yield stress
The time-dependency of the elastic limit in either compression or tension is difficult to quan-
tify as there is hardly any information available in the literature. Thomas (2003) investigated
the ratio of the yield stress to the peak stress in compression by visual estimation from pub-
lished stress-strain curves. Some data suggested a value of 0.70 to 0.85 of the peak stress,
whereas others tended to be in the generally accepted range of 0.25 to 0.4. Due to this
considerable scatter it was not possible to detect any particular time-dependency of the yield
stress ratio. Meschke et al. (1996) assumed in their constitutive model for shotcrete a time-

131
dependent yield stress fcy which is linked to the uniaxial compressive strength fcp via a
constant ratio, as given in the following equation:
28
fcy
fcy (t) = f (t)
28 c
(5.27)
fcp

28 and f 28 are the uniaxial yield stress and the uniaxial compressive peak strength
where fcy cp
for hardened shotcrete at 28 days. No information has been found in the literature regarding
the elastic limit of young concrete in tensile stress conditions.

5.7.2 Deformability at early ages

For the full description of a complete stress-strain curve for concrete under uniaxial stress
conditions the strains at peak and failure strength might be the necessary input parame-
ters. For conventional cast concrete these are constant values and do not change with time.
However, as mentioned before, shotcrete at early ages shows a very soft and ductile material
behaviour and can withstand large strains without being completely damaged (ICE, 2004).
This means that a young shotcrete lining has the ability to accommodate tunnel deformations
that occur immediately after shotcrete installation close to the tunnel face (Chang, 1994).

i) Compressive peak strains

For hardened concrete the uniaxial compressive strains at peak strength are in the range of
0.15 to 0.25 % and up to 0.35 % at failure strength. However, in Golser (1999) it is reported
that in tests performed on 2-hour old shotcrete the material failed at strains of about 2 %
when loaded in compression. Graziani et al. (2005) mentioned that freshly applied shotcrete
typically exhibits plastic behaviour with strains at failure strength even as high as 3 to 4 %.
This fact is supported by test data for the uniaxial compressive peak strains from Sezaki et al.
(1989) developing with the uniaxial compressive strength for different accelerator dosages, as
can be seen on the left in Fig. 5.24. It should be mentioned that Sezaki et al. (1989) termed
the peak strains as “failure” strains in this plot. This figure shows on the right a data col-
lection for young shotcrete from the literature performed by Thomas (2003). Although a
considerable scatter exists, it appears that there is a decreasing trend for the compressive
peak strains with time.

132
Figure 5.24: Development of compressive strains at peak strength with uniaxial compressive
strength and time (from Sezaki et al., 1989 and Thomas, 2003)

Wierig (1971) observed in his experiments that compressive peak strains for young con-
crete are also much higher than those of hardened concrete, as indicated in Fig. 5.25. During
the first hours these strains are reducing rapidly and reach a minimum at about 1 day of
concrete age. Afterwards a slight increase in the peak strains with concrete ageing can be
identified. Byfors (1980) performed an extensive experimental parametric study investigating
the deformational behaviour of young concrete by varying the water-cement ratio, tempera-
ture and cement types. His results show a similar trend with high compressive peak strains
of up to 1.5 % for samples at an age of 6 h.

Figure 5.25: Development of compressive peak strains with shotcrete age (from Wierig, 1971)

Regarding the deformability of shotcrete at early ages Rokahr et al. (2005) are of a
completely different opinion to the researchers introduced above. They state that a slow
build-up load only causes failure at a “measured” strain which far exceeds the straining limit
of hardened concrete. According to them, this measured large strain includes a considerable
proportion of creep strains which already occur during the loading phase. Consequently, they

133
conclude that for quick loading conditions the compressive peak strain of young shotcrete
is far below that of hardened concrete, a fact which is clearly the opposite to the earlier
presented information. However, the work presented within this thesis follows the route of
larger deformability of shotcrete at early ages.

For the mathematical description of the compressive peak strains for early age concrete
and shotcrete various equations can be found in the literature. Thomas (2003) fitted a
regression curve to his data collection in Fig. 5.24, which is given as:

εcp (t) = −0.4142 ln(t) + 3.1213 (5.28)

where t is the shotcrete age in hours and εcp the compressive peak strain in per cent. On the
basis of experimental tests conducted by Byfors (1980), a trilinear relation for the compressive
peak strain is proposed by Meschke et al. (1996), which can be written as:
⎧
⎪ 0.0575
⎪
⎪ 0.06 − t for 0 ≤ t < 8h
⎪
⎪ 8
⎨
0.0008
εcp (t) = 0.0025 − (t − 8) for 8 h ≤ t < 16 h (5.29)
⎪
⎪ 8
⎪
⎪
⎪
⎩0.0017 + 0.0003 (t − 16) for t ≥ 16 h
654

where t is the shotcrete age in hours. For their sprayed concrete model Jones et al. (2008)
fitted a curve to the compressive strains at peak strength data from tests and the following
relationship was found:
−0.47
εcp = 0.0136 fcp (5.30)

where fcp is the time-dependent uniaxial compressive peak strength. De Schutter (1999)
presented some data for young concrete, where high peak strains of early age concrete were
reducing to a minimum value at a compressive strength of about 10 MPa. Based on this data
set, he described mathematically the development of compressive peak strains with the mean
compressive peak strength fcm by the expression given below:
 1
fcm (t) 3 0.000664
εcp = 0.000948 +  (5.31)
fcmo fcm (t)
fcmo

where t is the time in days and fcmo = 10 MPa. Finally, as mentioned earlier, some ex-
perimental work on the early age deformability of conventional concrete was conducted by
Byfors (1980). The compressive peak strain is presented as a function of the time-dependent
uniaxial compressive peak strength in the following way:

εcp = a fcp (t)b (5.32)

where a and b are two correlation coefficients that can be taken from Tab. 5.4.

134
 Compressive strength a b
fcp ≤ 5 MPa 0.244 -0.652
fcp > 5 MPa 0.0467 0.331

Table 5.4: Parameters a and b for development of peak strains with uniaxial compressive
strength (after Byfors, 1980)

ii) Tensile peak strain

Although the early age deformability of shotcrete in tension is of crucial importance regarding
the cracking behaviour of tunnel linings due to shrinkage and temperature effects, hardly any
information can be found in the literature. From the few test results available, a similar
pattern as for the compressive peak strains can be observed, as indicated in Fig. 5.26.

Figure 5.26: Development of tensile peak strains with shotcrete age (from Kusterle & Lukas,
1990)

At early ages the tensile peak strains (or termed “failure” strains in Fig. 5.26) appear to
be in the range of 0.02 to 0.05 %, reducing significantly within the first hours after spraying.
At a shotcrete age of about 12 hours this drop seems to stop and even a slight increase in the
magnitude of the tensile peak strain may be identified. Byfors (1980) conducted some tension
tests on young concrete and established a relation between the tensile and compressive peak
strain depending on the development of the compressive strength, as can be seen in Fig. 5.27.
At low strength levels, i.e. when the concrete is still very young, the tensile peak strain is
approximately just about 3 % of the compressive peak strain. Starting from the period when
the tensile and the compressive peak strain are at a possible minimum, the tensile peak strain
grows far more rapidly than the compressive peak strain. In the case of hardened concrete
at 28 days, the tensile peak strain is finally about 8 to 9 % of the compressive peak strain.

135
Figure 5.27: Ratio between tensile and compressive peak strain with compressive strength
(from Byfors, 1980) (fcc = fcp , εcto = εtp , εcco = εcp )

5.7.3 Damage effects due to early-age loading

During tunnel construction, the very young shotcrete is loaded immediately after being
sprayed onto the free tunnel walls, leading to possible stress levels within the lining that
are relatively high or can even be close to failure. The question arises whether this early-age
loading causes some damage to the internal structure of the shotcrete and whether it influ-
ences the further increase in stiffness and strength and mechanical behaviour at later stages
of construction. From the published information in the literature this aspect is not clear and
various opinions exist among researchers.

Moussa (1993) performed some experimental work investigating the effect of early age
loading on cylindrical shotcrete core samples. He introduced a damage parameter, Rdf c ,
which represents the reduction of the compressive strength to be gained in the mean time of
two loading cycles (t1 to t2 ). By fitting a linear function to the obtained test data it can be
written that:  
σo
Rdf c = 2.532 − 0.691 (5.33)
fcp,t1
where σo is the applied stress for the first loading at time t1 and fcp,t1 is the respective
compressive strength. According to this equation, the reduction of the gain of strength due
to damage of the material can be neglected for shotcrete preloaded by a stress less than 69.1 %
of the uniaxial compressive strength at the loading age t1 . If the sample was preloaded up
to the peak strength and immediately unloaded (σo /fcp,t1 = 1.0), the reduction of the peak
strength at a later age would increase up to 78 %. However, it can be seen in Fig. 5.28 that
there is a considerable scatter in the obtained results from which this relationship has been
obtained.

136
Figure 5.28: Reduction in gain of uniaxial compressive strength due to preloading (from
Moussa, 1993)

On the other hand, Aldrian (1991) reported on some experimental tests performed by
Golser et al. (1990) that indicated a different behaviour. In some particular creep tests
shotcrete samples were loaded at an age ranging from 8 to 672 h for a duration of 3 weeks.
By comparing the compressive strength of the samples at the end of the loading period with
unloaded samples, they found that the preloaded samples had a compressive strength that
was up to 50 % higher than the one of the unloaded samples. Two possible explanations were
given by Aldrian (1991). First, the increase in compressive strength might be due to the fact
that the chemical process of cement hydration changes under a sustained stress. Second, a
compressive stress applied for a long period leads to a certain mechanical compaction within
the structure of the shotcrete sample that strengthens the material.

Abdel-Jawad & Haddad (1992) carried out some experimental work in order to study
the healing capacity of concrete subjected to overloading within the first three days after
casting. Concrete samples were first loaded in compression at an age of 8, 16, 24 and 72
hours and reloaded at 7, 28 and 90 days. It was observed that loading the concrete at an
age beyond 8 hours and up to 90 % of its compressive failure load had no effect on the later
strength development. Nevertheless, samples loaded firstly in compression at an age of 8
hours showed a loss in strength of about 25 %. They concluded that moist curing is an
essential requirement for the healing capacity of young concrete, although complete healing
of cracks in damaged samples does not mean complete strength regain. In the retesting of
healed specimens the failure mode always initiated from existing cracks.

Finally, Lu et al. (2004) conducted extensive research into the damage of concrete due to
triaxial loading history. Their damage parameter D, representing the compressive strength
reduction after a certain loading history, is linked to the maximum axial strain by the following

137
expression:
D = Do + A e−(ε1 −εo )/B (5.34)

where D is the damage parameter in both the axial and lateral directions, ε1 the maximum
axial strain due to the applied stress σ, and Do , εo , A and B are parameters correlated to
the lateral confining pressure. It was concluded that the lateral pressure is the main factor
that affects the damage development in concrete.

5.8 Creep and relaxation

Creep is the terminology for defining the gradual increase in strain with time under a sustained
stress and can be of considerable importance in structural mechanics. The reverse process
of creep is called relaxation and describes the progressive decrease in stress with time for a
specimen that is subjected to a constant strain. For concrete structures both phenomena
are related to moisture movement, when such a stress or strain is applied. Neville (1970)
distuingishes between creep of concrete under conditions of no moisture movement to or from
the ambient medium which is often referred to as “basic creep”. The additional creep is
termed “drying creep”, caused by the concurrent drying process of the material. The general
form of a strain-time curve for a material subjected to creep can be seen in Fig. 5.29.

Figure 5.29: General form of strain development with time due to creep (from Neville, 1970)

At the start of loading at time zero an instantaneous deformation occurs, consisting

primarily of an elastic component but may include as well an irreversible plastic part. With
further development of time, three stages of creep can be identified, these are:

 Primary creep

138
  Secondary creep

 Tertiary creep

In the primary creep phase, the creep rate decreases with time followed by the range of
steady state creep where a minimum constant creep rate can be identified. A straight line
for secondary creep may be a convenient approximation. The tertiary creep phase is dom-
inated by an increasing creep rate leading to creep failure due to microcracking within the
concrete sample. This phase may or may not exist, depending on the applied stress level. On
unloading, together with the instantaneous elastic recovery, there will be a gradual recovery
of a portion of creep, meaning that creep components can be divided into a reversible and
irreversible part, see Fig. 5.30.

Figure 5.30: Creep recovery at unloading (from Bosnjak, 2000)

Practically, almost all the creep data available in the literature refer to experimental
work for concrete in uniaxial compression. This might be due to the fact that concrete
structures are designed for utilising the high compressive strength of concrete rather than
the low tensile capacity. Furthermore, creep tests in uniaxial compression are much easier to
perform than under other complex stress states. However, when the cracking behaviour of
early age concrete is the purpose of an analysis, the creep behaviour under tensile stresses
might be of considerable importance. Atrushi (2003) studied the creep capacity of young
concrete in compression and tension performing some experimental tests. From the obtained
results it was observed that the creep rates within the first 24 hours after loading where
higher for compression than for tension. Afterwards, both creep rates decreased continually
with time, but the decrease in tensile creep was much less pronounced than in compressive
creep. Generally, it was concluded that the magnitude of creep in tension for equal stress
levels is higher than in compression and this is in accordance with investigations performed
by Illston (1965).

In a uniaxial compression test, creep not only occurs in the axial direction but also in
the normal directions, which is referred to as lateral creep. Similar to the approach used in
elasticity, the ratio of the lateral creep strain to the axial creep strain is often termed the
creep Poisson’s ratio, μcr . In the literature there is no clear agreement on the magnitude of

139
the creep Poisson’s ratio (Neville, 1970). Some researchers found it to be zero and others
reported the same value as the elastic Poisson’s ratio. Fig. 5.31 shows the possible theoretical
values of lateral creep predicted by the creep Poisson’s ratio.

Figure 5.31: Possible values of creep Poisson’s ratio μcr under uniaxial stress (from Neville,
1970)

For creep in multiaxial stress conditions it follows that there is creep induced by the
applied stress in a particular direction and also creep due to the Poisson’s ratio effect of creep
in the two normal directions. The question arises whether these strains occur independently
of each other and the principle of superposition is applicable, or whether the behaviour is
more complex. Neville (1970) presented some results from multiaxial compression tests which
indicated that creep in a multiaxial loading system is significantly lower than under a uniaxial
stress. Furthermore, the value of the creep Poisson’s ratio varies in the three principal stress
directions and it follows that creep strains under multiaxial compression can not be simply
superposed. Experimental work by Gopalakrishnan et al. (1969) showed that even under
isotropic compression considerable creep deformation takes place. Hence, the principle of
superposition does not hold for creep and creep in multiaxial loading conditions is a complex
phenomenon that cannot be simply estimated from uniaxial creep measurements.

5.8.1 Mechanisms behind creep

A number of theories exist in the literature which try to reveal the cause of creep, but it has
to be mentioned that the mechanisms behind creep are not fully understood (Thomas, 2003).
However, it is recognised that, since creep can be related to the removal of absorbed water, its
origin lies within the cement paste (Mehta & Monteiro, 1985). In contrary to shrinkage, for
creep an applied stress is the driving force leading to a loss of physically absorbed water. A
minor cause of the contraction of the cement-aggregate system could be the removal of water
held by hydrostatic tension in small capillaries of the hydrated cement paste. Another theory
states that creep is the result of a rearrangement of the capillary structure of the cement
paste due to the applied load. Flow theories assume that the hydrated cement paste acts as
a highly viscous liquid whose viscosity increases with time as a result of chemical changes
within the structure. The occurrence of delayed elastic response in the aggregates could be

140
another cause of creep in concrete. Due to the bond between the cement paste and the
aggregates, an applied stress is gradually transformed from the former concrete component
to the latter, which consequently starts to deform elastically. Hence, the delayed elastic
strain in the aggregates contributes to the total creep. Further information about different
mechanisms behind creep is given in Neville (1970).

5.8.2 Influencing factors

As already pointed out, creep is a very complex process where various factors influence the
mechanism behind it. Through his experimental work Byfors (1980) summarised them as:

 Concrete age at loading (or degree of hydration)

 Loading time

 Concrete composition and mix design

 Temperature

 Moisture conditions (Relative Humidity RH)

 Applied stress level

Fig. 5.32 provides a schematic presentation of the influence of these factors.

141
 Figure 5.32: Factors influencing creep (from Byfors, 1980)

At very early ages concrete and shotcrete are very creep active media since the developed
strength is still at a low level and hence the applied stress levels are high. For instance, Davis
et al. (1934) compared the creep of water-stored concrete samples subjected to a constant
stress during 80 months. The ratio of the creep deformations of the specimens loaded at
7 days, 28 days and 3 months averaged at 3 : 2 : 1. They noted further that the rate of
creep during the first few weeks under load was much higher for concrete loaded at an early
age than for older age. However, Kuwajima (1999) reported that for young shotcrete this
early age dependency of creep is just relevant for an age of up to 40 hours, since the strength
development happens faster compared to conventional cast concrete. Parrott (1978) compiled
the test results on early age creep from various researchers as can be seen in Fig. 5.33. The
graph illustrates the creep as a relative ratio to the creep deformation at an age of loading
of 28 days. The age dependency of creep is clearly visible and can be directly related to the
process of cement hydration. After a rapid decrease in creep at early ages, the influence of
age seems to stabilise from a concrete age of 100 days onwards. The considerable scatter in
the test data around 1 day age may be linked also to the differences in the degree of hydration
which is more marked at an early age.

142
 Figure 5.33: Relative creep with age at loading (from Byfors, 1980)

The quantification of creep for young shotcrete is not straightforward, but Thomas (2003)
states that, according to the test data by Huber (1991), a sample loaded at an age of 8 days
may creep around 25 % more than a similar sample loaded at 28 days.

In the literature it is well accepted knowledge that in uniaxial compression, up to a

stress level of 0.4, creep is proportional to the applied stress (Neville, 1970), as can be seen
in Fig. 5.34. Above this level the creep behaviour becomes non-linear and is believed to
increase at an increasing rate. This fact is strongly linked to the internal microcracking that
takes place in a concrete specimen during loading at high stress levels. A sufficiently high
sustained stress can produce failure in the sample, which is also referred to as tertiary creep
or static fatigue failure. This stress level dependency explains further why young concrete
and shotcrete exhibit higher creep rates than at older ages. Due to the low concrete strength
at early ages the stress level for a constant applied load is relatively high at the beginning
of loading and decreases with time as the strength development continues. Hence, creep is
greater for more slowly hydrating concretes (Thomas, 2003).

143
 Figure 5.34: Creep for different stress levels (from Neville, 1970)

Since the origin of creep is believed to lie within the cement paste, the type and amount
of cement adopted in the mix design is another factor influencing creep (Neville, 1981). It
is reported that creep increases with increasing cement content, a fact which is of great
importance for shotcrete since its composition is richer in cement than conventional cast
concrete due to the rebound behaviour (see Chapter 4). The type of cement influences the
creep behaviour in so far as it has an impact on the concrete strength development with
time. The same rule applies to the water-cement ratio and to the fineness of the cement.
However, studies with different cement types have failed to show a simple direct correlation
between creep and the chemical composition of the cement. Cement replacements, such as
microsilica, reduce the porosity and therefore may reduce as well the drying creep, since they
restrict water movement within the concrete (Thomas, 2003).

Certain physical properties of the aggregate also have an influence on the creep of concrete,
where it appears that the modulus of elasticity is the most important factor (Neville, 1981).
A higher modulus leads to a decrease in creep since a greater restraint is offered by the
aggregate to the potential creep of the cement paste. Furthermore, it may be expected that
the porosity of the aggregate plays a direct role in the moisture movement within the concrete
and therefore has an influence on the creep behaviour. According to Neville (1970), other
parameters, such as grading, maximum size and shape of the aggregate, only have an indirect
impact on the creep.

144
 One of the most influencing factors on creep is the relative humidity of the air surrounding
the concrete. As a general rule it can be assumed that creep is higher the lower the relative
humidity (Neville, 1981), as indicated in Fig. 5.35. This influence, which is strongly linked
to the process of drying, is much smaller for samples that are in hygral equilibrium with
the surrounding medium prior to the application of the load. Hence, it can be said that
concrete which exhibits high shrinkage shows generally also high creep. Therefore, these
two phenomena appear to be coupled and linked to the same aspect of the structure of the
hydrated cement paste and it is very difficult to separate them completely.

Figure 5.35: Creep for different relative humidity (from Neville, 1981)

Regarding the impact of temperature on the creep behaviour of concrete it is important

to mention that the rate of creep increases with an increase in temperature. In Mehta
& Monteiro (1985) it is reported that the creep rate at a temperature of about 70 C is
approximately 3.5 times higher than at 21 C. However, Thomas (2003) states that these
temperature effects on creep can be ignored for tunnel linings since the increase in temperature
is relatively small and short-lived.

Finally, Ding (1998) investigated the creep behaviour of fibre reinforced concrete and
shotcrete at early ages of 11 h. For the first 24 hours after loading the concrete samples
with a fibre content of 20 and 40 kg/m3 showed a significant reduction in creep deformation.
In contrary, the specimens with 60 kg/m3 fibre content developed almost the same creep
deformation as the plain concrete. With increasing age the influence of steel fibres on the
creep behaviour becomes more apparent. The biggest reduction in creep of 45 % was obtained
for the concrete samples with 40 kg/m3 at 240 h after loading, whereas the reduction for the
60 kg/m3 samples was approximately 12 %. It can be concluded that steel reinforcement
(fibres or wire mesh) has the potential for reducing creep deformation at early ages, but
there is no direct quantifiable relation between the amount of reinforcement and the creep
reduction.

145
5.8.3 Creep in tunnelling
The importance of shotcrete creep in tunnelling has been highlighted by many researchers,
among them Rokahr & Lux (1987), Pöttler (1990), Schubert (1988), Golser et al. (1990),
Thomas (2003), Aldrian (1991), Hesser (2000), Yin (1996), Huber (2006), Fischnaller (1992)
and Walter (1997). As mentioned above, young shotcrete is a very creep active medium and
this property is of crucial importance in tunnelling, where a shotcrete shell is immediately
loaded to relatively high stress levels directly after installation. In Fig. 5.36 Pöttler (1993)
illustrated schematically the development of stresses in a certain cross section of the lining
during tunnel advance.

Figure 5.36: Development of lining stresses during tunnel advance (from Pöttler, 1993)

This figure shows that due to every excavation step the stresses in the lining increase,
where the amount of loading depends heavily on the advance length and rate. But between
every two rounds relaxation takes place in the very young shotcrete and stresses decrease
by a certain amount. These relaxation phases are influenced by the stress level within the
shotcrete shell, the creep capacity of the young shotcrete and the tunnel advance. Therefore
the maximum stresses that occur in a lining are normally smaller than expected. Further,
it is clearly visible that this maximum stress does not occur directly at the tunnel face, but
a couple of excavation rounds behind, within a distance of 2 to 3 tunnel diameters, because
of three dimensional arching of the ground. Once the maximum stress is reached, no further
loading and influence of the working face takes place and the stresses reduce to a fairly
constant stress level. In a tunnel analysis one should take into account the three dimensional
stress field up to the point of maximum stress. For bigger distances behind the tunnel face a

146
two dimensional analysis would be sufficient.

According to Rokahr & Lux (1987) Fig. 5.37 shows the development of hoop stresses in
a certain cross section of the tunnel lining with radial displacements (marked as w in the
figure). Similar to the previous figure, due to tunnel advance both the radial displacements
and the hoop stresses increase. During two excavation rounds the shotcrete is creep active
and deformations continue to increase. As can be inferred, creep and relaxation are processes
that are taking place within the material simultaneously, leading normally to the phenomenon
of relatively large displacements and lower than expected stresses (Rokahr & Lux, 1987).
However, Pöttler (1993) states that stress relaxation is not of importance for shallow tunnels
with an overburden of less than 2 to 3 tunnel diameters, but additional deformations due to
creep occur.

Figure 5.37: Creep and relaxation during tunnel advance (from Rokahr & Lux, 1987)

5.9 Shrinkage
During the process of cement hydration at early ages, concrete or shotcrete undergoes a
certain change in volume which is mainly due to shrinkage or swelling. These are generally
stress-independent deformations caused by drying or wetting and therefore proper curing
conditions are of great importance for the hydration of concrete (Neville, 1981). Kuwajima
(1999) states that in tunnelling the moisture provided by the soil plays a major role in
the shrinkage process, resulting in a swelling at the invert region of the lining and reduced
shrinkage in other areas. Furthermore, in the area of the tunnel crown shrinkage strains are
highly related to the presence of air ventilation. If part of a concrete structure is restrained,
the shrinkage of concrete can introduce stresses in the structure that may lead to cracking,
as can be seen in Fig. 5.38.

147
 Figure 5.38: Concept of shrinkage induced stresses (from Malmgrem et al., 2005)

When the bending stress σz reaches the maximum tensile strength, which is very low
for young shotcrete, cracks start to occur perpendicular to the interface between the new
shotcrete layer and the soil and concerns may arise regarding the watertightness of the lining.
In a similar manner failure takes place when the normal tensile stress σE exceeds the bond
strength of the interface. However, when such shrinkage induced stresses are to be expected,
it has to be taken into account that these stresses might be reduced significantly by the
mechanism of creep (Malmgrem et al., 2005).

5.9.1 Mechanisms behind shrinkage

Thomas (2009) distinguishes 4 basic types of shrinkage that have different origins and are
described here briefly:

5.9.1.1 Plastic shrinkage

While the concrete is still in a plastic state at a very early age, it can happen that in the
case of insufficient protection some volumetric contraction takes place caused by the loss
of water from the surface due to evaporation, or by the suction from adjacent dry soil or
an existing concrete (Mehta & Monteiro, 1985). This phenomenon is referred to as plastic
shrinkage and can lead to surface cracking, where cracks are usually parallel to one another
and spaced about 0.3 to 1.0 m apart. In Tab. 5.5 Neville (1981) gives some absolute values for
the magnitude of plastic shrinkage at 8 hours after casting, depending on the wind velocity.

5.9.1.2 Drying shrinkage

Contrary to plastic shrinkage, drying shrinkage takes place in the hardened cement paste as
water is lost to the air (Thomas, 2009). The removal of free water in larger voids and capillary

148
 Wind velocity Shrinkage
(m/s) (10−6 )
0 1 700
0.6 6 000
1.0 7 300
7.0 to 8.0 14 000

Table 5.5: Absolute values for plastic shrinkage (Relative humidity 50 %, Temperature 20 C)
(from Neville, 1981)

pores causes little or no shrinkage, but as drying continues, absorbed water in the cement
paste is removed and leads to irreversible volume changes. In general, sprayed concrete
shows higher drying shrinkage than conventional cast concrete due to the slightly different
mix design, installation technique and curing conditions in a tunnel (Austin & Robins, 1995).
However, Hills (1982) states that shrinkage values for sprayed concrete lie broadly within the
expected range for cast concretes.

5.9.1.3 Autogeneous shrinkage

Autogeneous shrinkage of concrete and shotcrete is defined as the macroscopic volume change
occurring with no moisture transferred to the exterior surrounding environment (Holt, 2001).
It is associated with the chemical hydration process of the cement particles, where water is
drawn from the capillary pores and localised drying of these pores takes place. This process is
often termed “self-desiccation” and causes the cement matrix to contract (Thomas, 2009). In
the literature it is agreed that autogeneous shrinkage cannot be prevented by casting, placing
or curing methods, but must be addressed by the correct mix design.

5.9.1.4 Carbonation shrinkage

This type of shrinkage occurs mostly in the surface layers of concrete limited to a depth of
2 cm (Holt, 2001). Carbonation is the chemical process where the hardened cement paste
reacts with moisture and carbon dioxide in the air. This leads to a certain volume change
and a reduction in the pH of the concrete, which is of big concern regarding the corrosion
of reinforcing steel. Therefore carbonation of concrete and shotcrete is a durabilty issue that
takes a long time to affect a structure. Houst (1997) reported that carbonation shrinkage
is highly dependent on the concrete density and quality, but typical values for the highest
carbonation shrinkage at a relative humidity of 70 to 80 % after 100 days would reach 3 to
4 mm/m.

149
5.9.2 Factors influencing shrinkage
Shrinkage of concrete and shotcrete is a complex process where many factors are influencing
the early volume changes. Most of them are related to mix design, but also the installa-
tion technique in the case of shotcrete, environmental conditions and curing, and storage
conditions can have a great impact on shrinkage.

Since the origin of shrinkage lies in general in the cement paste, it is sensitive to the water
content of the concrete. Fig. 5.39 shows the general pattern where a higher water content
leads to an increased drying shrinkage (Neville, 1981).

Figure 5.39: Influence of water content on absolute values of shrinkage (from Neville, 1981)

The properties of the adopted cement type have little influence on the shrinkage be-
haviour, although for a given water-cement ratio shrinkage increases with higher cement
content (Malmgrem et al., 2005). Tab 5.6 contains some values for the drying shrinkage of
shotcrete after six months, applied either as wet-mix or dry-mix. It can be seen that the
wet-mix method with its higher water-cement ratio indicated higher shrinkage deformation
than the dry-mix technique.

150
 Wet-mix Dry-mix
Water-cement ratio 0.54 - 0.61 0.31 - 0.42
Cement content (kg/m3 ) 375 - 475 550 - 700
Shrinkage after 6 months (%) 0.08 - 0.1 0.045 - 0.07

Table 5.6: Drying shrinkage after Hills, 1982

This fact is also reported by Neubert & Manns (1993), who investigated the effect of ac-
celerating admixtures on the mechanical properties of shotcrete. Their test results presented
in Fig. 5.40 illustrate that an increased amount of accelerators leads to an increase in shrink-
age, in particular for the wet-mix method. This influence seems not to be so large for the
dry-mix shotcrete. Further experimental work on shrinkage of wet- and dry-mix shotcrete
can be found in Cornejo-Malm (1995).

Figure 5.40: Shrinkage of shotcrete (from Neubert & Manns, 1993)

The aggregate used in the concrete mix restrains the amount of shrinkage that can actually
be realised and therefore shrinkage can be reduced by an increased aggregate content, as
shown in Fig. 5.41.

151
Figure 5.41: Influence of aggregate content in concrete on the ratio of shrinkage of concrete
to the shrinkage of neat cement paste (from Neville, 1981)

The same effect is achieved by increasing the maximum aggregate size, that leads to a
leaner mix and hence results in a lower shrinkage (Neville, 1981). Furthermore, Wolfsier
& Morgan (1993) concluded that the addition of silica fume in the mix design (about 45 -
50 kg/m3 ) can reduce slightly the shrinkage of shotcrete. In contrast, Thomas (2009) reported
that fine materials added to the concrete, such as silica fume and fly ash, are known to increase
shrinkage strains.

Ding (1998) worked on the technological properties of young steel fibre reinforced concrete
and shotcrete, investigating also their shrinkage behaviour. For conventional cast concrete he
observed that the inclusion of 60 kg/m3 steel fibres led to a reduction in shrinkage of about
30 % after 320 h. At early ages within the first 24 h this reduction rises even up to 60 %. A
similar trend was obtained for steel fibre reinforced shotcrete, although a higher fibre content
(90 kg/m3 ) was necessary to minimise shrinkage deformation. In addition, Malmberg (1977)
stated that steel fibres were effective in distributing the cracking that occurs for restrained
shrinkage and in restricting the crack widths.

Another important influencing factor for shrinkage is the relative humidity of the medium
surrounding the concrete structure, as can be seen in Fig. 5.42. This figure illustrates that
for a relative humidty of 100 % concrete starts to swell rather than to reduce its volume.
Thomas (2003) reports that in a tunnel the relative humidity is around 50 %, with a fairly
constant temperature within the range from 12 to 24 C.

152
Figure 5.42: Influence of relative humidity on absolute values for shrinkage (from Neville,
1981)

It can be summarised that sprayed concrete exhibits higher drying shrinkage than conven-
tional cast concrete. However, Austin & Robins (1995) relate this effect to the typical lower
aggregate-cement ratio and higher cement contents due to rebound for shotcrete, rather than
the spraying process itself.

5.10 Temperature induced deformation

The chemical reaction of the cement hydration in early-age concrete and shotcrete is an
exothermic process, where energy is released in form of heat. Due to this temperature changes
the tunnel lining expands and contracts within the first few days after shotcrete installation.
Similar to shrinkage, if this thermal deformations are restrained stresses are introduced within
the concrete structure, leading to a compressive stress in the case of expansion, and a tensile
stress for the cooling phase when the temperature is decreasing again. Obviously, the thermal
boundary conditions, such as ventilation, temperature in the tunnel and surface temperature
of the excavated ground, highly influence the temperature distribution within the shotcrete
layer, but a typical profile of temperature development over time can be seen in Fig. 5.43.

153
Figure 5.43: Temperature development in hardening concrete (from Mehta & Monteiro, 1985)

According to Thomas (2003) the maximum temperature rise depends on the following
factors:

 Thickness of the sprayed concrete layer

 Initial temperature of the concrete mix

 Rate of cement hydration

In Kusterle & Lukas (1993) it is reported that, typically, the maximum temperature rise
occurs around the centre of the lining and can reach values of up to 25 C for a 30 cm thick
shotcrete layer. Tab 5.7 gives a guideline to estimate the maximum temperature rise in the
shotcrete depending on the layer thickness.

Shotcrete layer thickness ΔT from initial concrete Temperature drop

temperature to Tmax for 35 h after Tmax
2 - 3 cm 2-4 C negligible
5 - 10 cm 6-9 C 4-7 C
10 - 15 cm 10 - 15 C 8 - 12 C
30 cm 25 C 17 C

Table 5.7: Maximum temperature rise as a function of layer thickness (after Kusterle &
Lukas, 1993)

It can be concluded that in order to keep the hydration temperature to a low level,
the application of a shotcrete lining as a layered system is preferable (Kusterle & Lukas,
1993). Cornejo-Malm (1995) measured the development of hydration temperature for both
dry- and wet-mix shotcrete samples and observed that for the dry-mix the peak temperature
Tmax occurred between 8 to 10 h after casting, whereas for the wet-mix shotcrete Tmax was
between 10 to 14 h. This is in general in accordance with Huber (1991) who detected the
maximum temperature rise in his shotcrete samples between 9 to 12 h after spraying. His

154
experimental results showed further that, due to the addition of accelerators in the mix
design, the development of hydration temperature is speeded up.

As mentioned earlier, the initial tendency of the lining to expand can introduce compres-
sive stresses. However, since the Young’s modulus is still very low at this early age, these
compressive stresses are expected to be of no major importance for the structural behaviour
of the lining. The situation is different for the cooling phase where, due to contraction, sig-
nificant tensile stresses can develop that may lead to cracking of the shotcrete shell. Another
important effect of hydration temperature is reported by Jones et al. (2008). Radial pressure
cells installed at the front-shunt tunnel at Heathrow Terminal 5 indicated that, due to the
temperature expansion of the lining at early ages, the ground pressure acting on the tunnel
can reach up to 100 % of the overburden pressure, exposing the lining to the highest potential
stress in the first 24 h after spraying. Nevertheless, in most cases researchers do not take into
account the thermal effects of shotcrete at early ages.

For the mathematical description of the early age deformation of shotcrete due to hydra-
tion heat, the coefficient of thermal expansion is of great importance and it can be written:

εth = αt ΔT (5.35)

where εth is the thermal strain and ΔT the increase in temperature. αt is the coefficient of
thermal expansion for shotcrete and is generally assumed to be the same as for conventional
cast concrete (Thomas, 2003). Here again, the magnitude of αt depends on the thermal
properties of the cement, aggregates and their proportions in the mix (Neville, 1981). Ac-
cording to CEB-FIP Model Code (1990) for the purpose of structural analysis the coefficient
of thermal expansion for hardened concrete at 28 days may be taken as αt = 10 × 10−6 C−1 .
In Byfors (1980) a literature review on the thermal properties of young concrete at early
ages can be found. From the collected data it appears that αt is not a constant material
parameter, but decreases with increasing concrete age, as can be seen in Fig. 5.44.

155
Figure 5.44: Age dependency for the coefficient of thermal expansion αt (-/ C)(from Byfors,
1980)

5.11 Summary
The purpose of this chapter was to provide an introduction into the mechanical and time-
dependent behaviour of concrete and shotcrete under various loading conditions. The follow-
ing topics have been discussed in detail:

 Hardened concrete in uniaxial compression and tension shows a highly non-linear ma-
terial behaviour, where crushing and cracking of the concrete govern the post-peak
behaviour. Increased compressive strengths can be expected for concrete in biaxial and
triaxial loading conditions, leading to smooth yield and failure surfaces.

 From the available data in the literature it is difficult to quantify any anisotropy of
shotcrete that might be expected due to the spraying process.

 The incorporation of steel fibres in the concrete mix design results in a more ductile
material behaviour, particularly for the post peak stress regime in tension.

 Shotcrete at early ages shows a relatively plastic and ductile material response with
low stiffness and strength. Furthermore, higher strain limits can be achieved due to
the increased deformability of shotcrete at early stages of cement hydration. With

156
 curing time material behaviour becomes more and more brittle, caused by the increase
in stiffness and strength with time. However, compressive and tensile strains at peak
and failure strengths appear to reduce during cement hydration.

 The effect of compressive preloading of shotcrete samples at early ages on the develop-
ment of the compressive strength at later stages is not very clear. In the literature both
types of results can be found, i.e. an increase and a reduction in the strength values.

 Creep and relaxation are important aspects in shotcrete behaviour related to “time”,
depending on various factors, such as loading age, stress level, temperature, moisture
conditions and concrete composition. It is believed that creep in tunnelling leads to a
certain reduction in the expected lining stresses and is therefore of a beneficial nature.

 Deformations of young shotcrete that occur due to shrinkage and increased hydration
temperature might be important to consider in a realistic tunnel lining design, since
these two phenomena can lead to cracking of the shotcrete shell. Their origin lies within
the cement paste of the concrete and can be addressed in the mix design.

157
Chapter 6

Modelling of concrete and shotcrete

6.1 Introduction
After having introduced the main characteristics of the time-dependent mechanical behaviour
of concrete and shotcrete in Chapter 5, the aim of the current chapter is to give the reader
an overview about certain aspects of modelling a quasi-brittle material such as concrete
and shotcrete both in compression and tension. It starts with presenting simple approaches
based on empirical observations and elasticity, moving then on to more complex constitutive
models that can be used in a finite element analysis. A short introduction into the classical
theory of elasto-plasticity serves as a background for presenting some basic elasto-plasticity
models that are commonly applied for the modelling of hardened concrete. The simulation
of the post-peak and cracking behaviour of unreinforced and reinforced concrete and its
associated localization problems will be discussed briefly. Furthermore, some attention will
be given to the modelling of the time-dependent phenomena of creep, shrinkage and hydration
temperature induced strains, with a special focus on shotcrete behaviour. Finally, a review
of different constitutive models adopted for tunnel lining design that can be found in the
literature will conclude this chapter.

6.2 Empirical models for concrete in compression

A starting point for developing an empirical constitutive law is experimental data from various
tests focusing mainly on uniaxial stress conditions in either compression or tension. Through
results obtained from a series of experiments a curve fitting process is normally performed,
that aims to derive the material parameters used in the proposed model functions. Although
this approach seems quite simple and straightforward, some difficulties exist associated with
the testing techniques of concrete, in particular for the post-peak behaviour. Other aspects
that highly influence the result of such a material test can be the precision of the testing
machine, the rate of loading, size of the tested specimen and statistical variations of the
material properties from sample to sample (Babu et al., 2005). Popovics (1970) provides a

158
review of different empirical stress-strain relationships for concrete under compression and a
few of them are presented below.

The CEB-FIP Model Code (1990) approximates the stress-strain curve for concrete under
uniaxial compression shown in Fig. 6.1 with the following equation:

Figure 6.1: Stress-strain diagram for uniaxial compression (from CEB-FIP Model Code, 1990)

 2
Eci εc εc
−
Ec1 εc1 εc1
σc = −   fcm for | εc | < | εc,lim | (6.1)
Eci εc
1+ −2
Ec1 εc1
where Eci is the initial tangent modulus, σc the uniaxial compressive stress, εc the uniaxial
compressive strain, εc1 is taken as an absolute value of −0.0022 (-) and Ec1 is the secant
modulus from the origin to the peak compressive stress fcm. The use of this equation is
limited to compressive strains smaller than the limit strain εc,lim . For further straining, the
descending branch of the stress-strain curve may be described by
-  2   .−1
1 2 εc 4 εc
σc = − ξ− + −ξ fcm (6.2)
εc,lim /εc1 (εc,lim /εc1 )2 εc1 εc,lim /εc1 εc1

with - 2   .
εc,lim Eci εc,lim Eci
4 −2 +2 −
εc1 Ec1 εc1 Ec1
ξ=    2 (6.3)
εc,lim Eci
−2 +1
εc1 Ec1

The values for Eci , Ec1 and εc,lim can be estimated for various concrete grades from the
following Tab. 6.1:

159
 Concrete grade C 12 C 20 C 30 C 40 C 50 C 60 C 70 C 80

Eci (GPa) 27 30.5 33.5 36.5 38.5 41 42.5 44.5

Ec1 (GPa) 9 12.5 17.5 22 26.5 31 35.5 40

εc,lim (%) -0.5 -0.42 -0.37 -0.33 -0.30 -0.28 -0.26 -0.24

Table 6.1: Eci , Ec1 and εc,lim for various concrete grades after CEB-FIP Model Code (1990)

For short-term loading tests in uniaxial compression Desayi & Krishnan (1964) proposed
a stress-strain relation given by:
Eε
f=  2 (6.4)
ε
1+
εo
where f is the uniaxial compressive stress at any compressive strain ε, εo the strain at
maximum compressive stress fo and E the tangent modulus that can be calculated from:

2fo
E= (6.5)
εo

The descending part of the above equation is valid up to a failure stress k fo at a maximum
strain εc . Fig. 6.2 shows a comparison of stress-strain curves of tested cylindrical concrete
samples with those plotted from equation 6.4.

160
Figure 6.2: Uniaxial compressive stress strain curves for concrete (from Desayi & Krishnan,
1964)

Another expression for the uniaxial compressive stress-strain behaviour which shows good
fit to experimental data is the one developed by Sargin (1968). It is of a non-dimensional
mathematical form and applicable for both the confined (= lateral applied stress) and uniaxial
case:
AX + BX 2
Y = (6.6)
1 + CX + DX 2
where
f ε
Y = and X= (6.7)
fo εo
f represents the compressive stress at strain ε, while fo is the peak stress at strain εo . Two
sets of constants A, B, C and D are required, one set for the ascending portion and a second
set for the descending portion of the curve depending on the applied boundary conditions
during the test. Fig. 6.3 compares model predictions with experimental data for different
lateral confining pressures f r.

161
Figure 6.3: Compressive stress-strain curves for different lateral confining pressures (from
Attard & Setunge, 1996)

EC 2 (2004) provides two different uniaxial compressive stress-strain curves, both shown
in Fig. 6.4.

Figure 6.4: Uniaxial compressive stress-strain curves for a) non-linear structural analysis and
b) for the design of cross-sections (after EC 2, 2004)

The first one is intended to be used for a non-linear structural analysis and the relation
between the uniaxial compressive stress σc and the uniaxial compressive strain εc is given as:

σc k η − η2
= (6.8)
fcm 1 + (k − 2) η

162
where η and k can be estimated through the strain at peak stress εc1 and the secant modulus
of elasticity Ecm (at 0.4 fcm )as:

εc 1.05 Ecm εc1

η= and k= (6.9)
εc1 fcm

In the case of the non-linear design of a concrete cross-section, the following stress-strain
curve may be adopted:
⎧   n 
⎪
⎨fcd 1 − 1 − ε c
for 0 ≤ εc ≤ εc2
σc = εc2 (6.10)
⎪
⎩f
cd for εc2 ≤ εc ≤ εcu2

where fcd is the uniaxial compressive design strength, n an exponential parameter ranging
from 1.4 to 2.0, εc2 the compressive strain at reaching the maximum strength and εcu2 the
ultimate compressive strain. For the most common strength classes C 12 to C 50 typical
values for the above concrete parameters according to EC 2 (2004) would be:

Ecm εc1 εcu1 εc2 εcu2

(GPa) (%) (%) (%) (%)

27 to 37 0.18 to 0.245 0.35 0.20 0.35

Table 6.2: Material parameters for concrete taken from EC 2 (2004)

6.3 Constitutive models based on elasticity

6.3.1 Linear elastic models
If a material behaves linear elastically the incremental stresses can be associated with the
strains via the following constitutive equation:

{Δσ} = [D] {Δε} (6.11)

where {Δσ} is the incremental stress vector, {Δε} the incremental strain vector and [D] is
the elastic stiffness matrix. Furthermore, in the case of an isotropic material, it can be shown
that only two independent elastic constants are necessary to describe the material behaviour,
i.e. the Young’s Modulus E and the Poisson’s ratio μ. Therefore the elastic stiffness matrix

163
takes the symmetric form shown in the following equation:
⎡ ⎤
1−μ μ μ 0 0 0
⎢ ⎥
⎢ μ 1−μ μ 0 0 0 ⎥
⎢ ⎥
⎢ μ μ 1−μ 0 0 0 ⎥
⎢ ⎥
E ⎢ 1 − 2μ ⎥
[D] = ⎢ 0 0 0 0 0 ⎥ (6.12)
(1 + μ) (1 − 2μ) ⎢ ⎥
⎢ 2
1 − 2μ ⎥
⎢ ⎥
⎢ 0 0 0 0 0 ⎥
⎣ 2 ⎦
1 − 2μ
0 0 0 0 0
2

For geotechnical analysis, the elastic soil behaviour is often more convenient to describe in
terms of the elastic shear modulus G and the bulk modulus K. They can be related to the
Young’s modulus E and the Poisson’s ratio μ via the following expressions:

E E
G= and K= (6.13)
2 (1 + μ) 3 (1 − 2μ)

In spite of its obvious shortcomings, apparently the linear elastic theory is still the most
commonly applied stress-strain relation in industry for modelling the complex non-linear
material behaviour of concrete and in particular of shotcrete.

6.3.2 Non-linear elastic models

A logical first step to improve the model predictions for concrete within the concept of
elasticity is to incorporate material parameters that depend on stress and/or strain level. In
this way some of the main characteristics of concrete behaviour presented in Chapter 5 can
be captured. In the case of an isotropic material there are only two parameters required, E
and μ or K and G, and therefore the inclusion of non-linearity in the model formulation is
relatively straightforward. In the following a non-linear elastic model developed by Kotsovos
& Newman (1978), based on experimental results obtained at Imperial College London as
part of an international cooperative programme, is presented. The model aims to describe
mathematically the ascending branch of the stress-strain relationship of concrete under any
stress state. Therefore, in the analysis of the experimental data, each stress and strain state
is decomposed into a hydrostatic and a deviatoric component. For the description of the
non-linear compressive stress-strain curve the following expressions of the secant bulk and
shear moduli, varying with the octahedral stresses σoct and τoct , are given:

Ks 1
=  1.09 (6.14)
Ko σoct
1 + 0.52
fc

164
and
Gs 1
=  1.7 (6.15)
Go τoct
1 + 3.57
fc
The tangent expressions are written in a similar form as:

Kt 1
=  1.09 (6.16)
Ko σoct
1 + 1.08
fc

and
Gt 1
=  1.7 (6.17)
Go τoct
1 + 9.63
fc
where Ko and Go are the initial values at the start of loading. However, it should be pointed
out that a simple derivation of the tangent expressions from the secant formulation is not
straightforward, as σoct and τoct are unknown functions of the volumetric and deviatoric strain
components. A graphical representation of the above equations is shown in Fig. 6.5 together
with experimental test data.

Figure 6.5: Variation of bulk and shear moduli with σoct and τoct (from Kotsovos & Newman,
1978)

6.4 Classical theory of elasto-plasticity

Over the last three decades considerable improvements have been achieved in simulating the
complex non-linear behaviour of concrete under multi-axial stress states. Nowadays, in the
literature a large variety of constitutive models based on elasto-plasticity can be found that
make it possible to analyse concrete structures by means of the finite element method. In
the next sections the four essential ingredients required to formulate such an elasto-plastic
constitutive model will be presented briefly. A more detailed description can be found in

165
Potts & Zdravković (1999). It should be noted that the procedure presented here focuses
only on the time-independent behaviour of a material such as hardened concrete at 28 days.
An extension to the basic theory of elasto-plasticity for the time-dependent behaviour of
shotcrete can be found in chapter 7.

6.4.1 Basic concepts

i) Coincidence of axes
Contrary to the elastic material behaviour, it is assumed that for an elasto-plastic material
law the principal directions of accumulated stress and incremental plastic strain coincide.

ii) Yield surface

In order to separate purely elastic behaviour from elasto-plastic behaviour a scalar function of
the stress state {σ} and the state parameters {k}, which can be related to hardening/softening
parameters, is needed. This yield function can be written as:

F ({σ}, {k}) = 0 (6.18)

The value of the yield function F is used to identify the type of material behaviour, as can
be seen in Fig. 6.6 for the biaxial stress space.

Figure 6.6: Yield surface and stress states for elastic and elasto-plastic material behaviour

Two different scenarios can be encountered.

1. Stress state within the yield surface (F < 0) −→ Purely elastic material behaviour

2. Stress state on the yield surface (F = 0) −→ Elasto-plastic material behaviour

iii) Flow rule and plastic potential function

A flow rule is required to determine the direction of plastic straining at every stress state.
This is usually achieved by assuming a plastic potential surface P , which is of the form:

P ({σ}, {m}) = 0 (6.19)

166
The outward vector normal to the plastic potential surface has components which provide an
indication of the relative magnitudes of the plastic strain increment components. Therefore
the flow rule can be expressed as:

∂P ({σ}, {m})
{dεp } = Λ (6.20)
∂σ

where {εp } is the plastic strain increment vector. The scalar parameter Λ controls the mag-
nitude of the plastic strain components and depends on the hardening/softening rules which
will be discussed later.

In the case of the plastic potential function being identical with the yield function
(F ({σ}, {k}) = P ({σ}, {m})), the flow rule is said to be “associated” and a normality
condition applies. The flow rule is defined as “non-associated” in the general case, where the
plastic potential function differs from the yield function (F ({σ}, {k}) = P ({σ}, {m})).

The assumption of an associated flow rule is a common simplification for the modelling
of concrete. However, experimental data indicate that associated flow rules might not be the
most appropriate assumption for predicting the material response of concrete (Babu et al.,
2005). Some researchers, among them Smith et al. (1989), Grassl et al. (2002), Vermeer &
de Borst (1984) and Frantziskonis et al. (1986), observed some dilatancy during shearing of
concrete, characterized by volume changes associated with shear distortion of the material.
Typical yield functions for concrete fail to predict this behaviour. In addition, data from
uniaxial stress-strain curves for concrete in compression show that concrete exhibits non-
linear volume change, displaying contraction at low load levels and dilation when the peak
strength is approached (Babu et al., 2005; Chen, 1982). Therefore in the literature various
forms of the plastic potential functions can be found with the aim of predicting the volumetric
behaviour of concrete in a more realistic way. Further information about appropriate plastic
potential functions for concrete is available in Han & Chen (1986), Dvorkin et al. (1989) and
Onate et al. (1988).

iv) Hardening and softening rule

A hardening/softening rule is required for specifying the evolution of the yield surface in
the course of plastic deformation. It defines how the state parameters {k} vary with plastic
straining and enables the calculation of the scalar parameter Λ given in equation 6.20. If
a material is perfectly plastic no hardening/softening occurs and the state parameters {k}
are constant and as a consequence no hardening/softening rule is required. Otherwise, the
hardening parameters control the change in size of the yield surface and it is common to
relate them to the components of the accumulated plastic strain. The material is assumed to
be “strain hardening/softening”. Alternatively, the motion of the yield surface in the stress
 / 
space can be related to the increase in plastic work W p = {σ}T {Δεp } , and the material
is said to follow a “work hardening/softening” concept.

167
 With the help of a hardening/softening rule the scalar multiplier Λ can be determined
through the following procedure: when the material is at yield the stress state must satisfy
the yield function so that F ({σ}, {k}) = 0; as a consequence, the total differential of the
yield function can be expressed as dF = 0, which is known as the “consistency condition”;
applying the chain rule of differentiation gives:
!T !T  
∂F ({σ}, {k}) ∂F ({σ}, {k}) ∂k
dF = {dσ} + {dεp } = 0 (6.21)
∂σ ∂k ∂εp

By substituting the flow rule (equation 6.20) into the above equation it becomes:
!T !T  
∂F ({σ}, {k}) ∂F ({σ}, {k}) ∂k ∂P ({σ}, {m})
dF = {dσ} + Λ =0 (6.22)
∂σ ∂k ∂εp ∂σ

Rearrangement of this equation leads to:

!T
1 ∂F ({σ}, {k})
Λ= {Δσ} (6.23)
A ∂σ

where A is a hardening modulus and can be expressed as:

!T   !
∂F ({σ}, {k}) ∂k ∂P ({σ}, {m})
A=− (6.24)
∂k ∂εp ∂σ

Finally, for simple elasto-plastic models, the elastic part of a constitutive model describes
the purely elastic material behaviour when the current stress state remains within the yield
surface. Furthermore, it defines as well the elastic deformations that occur as part of the
elasto-plastic behaviour, if the stress state is on the yield surface. However, advanced con-
stitutive models show non-linear elasto-plastic behaviour even before the conventional yield
surface is touched.

6.4.2 Formulation of the elasto-plastic constitutive matrix

In the following, the derivation of the elasto-plastic constitutive matrix needed for the de-
scription of the relationship between incremental stresses and incremental strains is presented.
This relationship can be written as:

{Δσ} = [Dep ] {Δε} (6.25)

where [D ep ] is used to distinguish that the constitutive behaviour is elasto-plastic. The

incremental total strains can be split into elastic and plastic components:

{Δε} = {Δεe } + {Δεp } (6.26)

168
The changes in stresses are related to the changes in elastic strains by the elastic constitutive
matrix [D] in the form:
{Δσ} = [D] {Δεe } (6.27)

Making use of equations 6.26 and 6.27 gives:

{Δσ} = [D] ({Δε} − {Δεp }) (6.28)

By substituting the flow rule (equation 6.20) into the above equation one obtains:
!
∂P ({σ}, {m})
{Δσ} = [D] {Δε} − Λ [D] (6.29)
∂σ

Combining equation 6.23 and 6.29 results in the following expression for the scalar multiplier:
!
∂F ({σ}, {k}) T
[D] {Δε}
∂σ
Λ= ! ! (6.30)
∂F ({σ}, {k}) T ∂P ({σ}, {m})
[D] +A
∂σ ∂σ

Substitution of this equation into equation 6.29 gives a relation between the incremental
stresses and the incremental strains:
! !
∂P ({σ}, {m}) ∂F ({σ}, {k}) T
[D] [D] {Δε}
∂σ ∂σ
{Δσ} = [D] {Δε} − ! ! (6.31)
∂F ({σ}, {k}) T ∂P ({σ}, {m})
[D] +A
∂σ ∂σ

Therefore, the elasto-plastic constitutive matrix is:

! !
∂P ({σ}, {m}) ∂F ({σ}, {k}) T
[D] [D]
∂σ ∂σ
[D ] = [D] −
ep
! ! (6.32)
∂F ({σ}, {k}) T ∂P ({σ}, {m})
[D] +A
∂σ ∂σ

6.4.3 Multi-surface plasticity

Sometimes the behaviour of concrete is described by two yield surfaces, controlling either the
compressive or the tensile behaviour in multiaxial loading conditions. In this section, the
theory for the elasto-plastic constitutive matrix of one single yield surface introduced in the
previous section is extended for the case when two yield surfaces F1 and F2 act simultaneously,
following a procedure from Potts & Zdravković (1999). As before, the incremental total
strains can be divided into elastic and plastic components. In addition, the plastic components
can be further split into plastic strains associated with each of the two yield surfaces. This
gives:
0 1 0 1
{Δε} = {Δεe } + Δεp1 + Δεp2 (6.33)

169
In a similar form as equation 6.28 it can be written that:
 0 1 0 1
{Δσ} = [D] {Δε} − Δεp1 − Δεp2 (6.34)
0 1 0 1
Two flow rules relate the incremental plastic strains Δεp1 and Δεp2 to the plastic po-
tential functions P1 and P2 . They can be expressed as:

∂P1 ({σ}, {m}) ∂P2 ({σ}, {m})

{dεp1 } = Λ1 and {dεp2 } = Λ2 (6.35)
∂σ ∂σ

where Λ1 and Λ2 are the two scalar multipliers. Substituting the two flow rules given in
equation 6.35 into equation 6.34 results in:
! !
∂P1 ∂P2
{Δσ} = [D] {Δε} − Λ1 [D] − Λ2 [D] (6.36)
∂σ ∂σ

When the material is plastic and both yield surfaces are active, the stress state must satisfy
both yield functions, so that F1 = 0 and F2 = 0. Using the chain rule of differentiation, the
consistency condition is obtained as:
!T !T  
∂F1 ∂F1 ∂k1 0 p1 1
dF1 = {dσ} + dε =0 (6.37)
∂σ ∂k1 ∂εp1

and !T !T  
∂F2 ∂F2 ∂k2 0 p2 1
dF2 = {dσ} + dε =0 (6.38)
∂σ ∂k2 ∂εp2
Substituting equation 6.36 into equations 6.37 and 6.38 leads to:
!T !T ! !T !
∂F1 ∂F1 ∂P1 ∂F1 ∂P2
dF1 = [D] {dε} − Λ1 [D] − Λ2 [D] − Λ1 A1 = 0
∂σ ∂σ ∂σ ∂σ ∂σ
(6.39)
and
!T !T ! !T !
∂F2 ∂F2 ∂P1 ∂F2 ∂P2
dF2 = [D] {dε} − Λ1 [D] − Λ2 [D] − Λ2 A2 = 0
∂σ ∂σ ∂σ ∂σ ∂σ
(6.40)
where the hardening moduli A1 and A2 are:
! 
1 ∂F1 ∂k1 0 p1 1
A1 = − dε (6.41)
Λ1 ∂k1 ∂εp1

and ! 
1 ∂F2 ∂k2 0 p2 1
A2 = − dε (6.42)
Λ2 ∂k2 ∂εp2
Equations 6.39 and 6.40 can be rewritten in a simpler form as:

Λ1 L11 + Λ2 L12 = T1 and Λ1 L21 + Λ2 L22 = T2 (6.43)

170
with:
!T !
∂F1 ∂P1
L11 = [D] + A1
∂σ ∂σ
!T !
∂F2 ∂P2
L22 = [D] + A2
∂σ ∂σ
!T !
∂F1 ∂P2
L12 = [D]
∂σ ∂σ
!T ! (6.44)
∂F2 ∂P1
L21 = [D]
∂σ ∂σ
!T
∂F1
T1 = [D] {dε}
∂σ
!T
∂F2
T2 = [D] {dε}
∂σ

The solution for Λ1 and Λ2 can be obtained from equation 6.43 as:

L22 T1 − L12 T2 L11 T2 − L21 T1

Λ1 = and Λ2 = (6.45)
L11 L22 − L12 L21 L11 L22 − L12 L21

Finally, combining these two equations with equation 6.36 gives the elasto-plastic constitutive
matrix in the case when the two yield surfaces are active simultaneously:
 ! ! 
[D] ∂P1 T ∂P2 T
[D ] = [D] −
ep
{b1 } + {b2 } [D] (6.46)
Ω ∂σ ∂σ

where
Ω = L11 L22 − L12 L21 (6.47)

and
! ! ! !
∂F1 ∂F2 ∂F2 ∂F1
{b1 } = L22 − L12 and {b2 } = L11 − L21 (6.48)
∂σ ∂σ ∂σ ∂σ

6.5 Basic plasticity models for concrete

6.5.1 Rankine criterion
To describe the brittle failure of concrete in tension a maximum tensile stress criterion, often
called the Rankine criterion or tension-cut-off (Chen, 1982), is generally used. It is a one
parameter model that assumes that failure takes place when the maximum principal stress at
any point inside the material reaches the tensile strength ftp , regardless of any other normal or
shear stresses that occur on other planes through this stress point. In terms of the maximum
principal stress this failure criterion can be expressed by the following equation (assuming a
tension positive sign convention):
σ1 = ftp (6.49)

171
Making use of the geotechnical stress invariants (see chapter 1) this equation can be written
within the range of −30◦ ≤ θ ≤ +30◦ :
 
2J 2π
F (p, J, θ) = p + √ sin θ + − ft = 0 (6.50)
3 3

Fig. 6.7 illustrates the cross sectional shape in the deviatoric plane and in the J − p space.

Figure 6.7: Rankine criterion in J-p (left) and principal stress space (right)

6.5.2 Mohr-Coulomb criterion

The well known Mohr-Coulomb criterion dates back to 1773 and can be expressed as a two
parameter model, where the limiting shear stress τ in a plane is only dependent on the normal
stress σ which is acting in the same plane. The failure envelope for the corresponding Mohr
circles is shown in Fig. 6.8 and given by the equation:

|τ | = c + σ tan φ (6.51)

where c is the cohesion and φ represents the internal friction angle of the material.

Figure 6.8: Mohr-Coulomb failure criterion

Failure of material will occur for all states of stress for which the largest Mohr circle touches

172
the failure envelope. This means that the intermediate principal stress has no influence on
the failure. The two material parameters c and φ are related to the uniaxial compressive and
tensile strength in the following way:

2c cos φ
| fcp |= (6.52)
1 − sin φ

and
2c cos φ
| ftp |= (6.53)
1 + sin φ
The friction angle φ is obtained from:

fc 1 + sin φ
= (6.54)
ft 1 − sin φ

Using the geotechnical stress invariants the Mohr-Coulomb criterion can be written as:
 
c
F (p, J, θ) = J − +p g (θ) = 0 (6.55)
tan φ

where
sin φ
g (θ) = (6.56)
sin θ sin φ
cos θ + √
3
In the principal stress space the deviatoric cross section plots as an irregular hexagon, which
can be seen on the right in Fig. 6.9. For moderate values of mean stress this model describes
reasonably well the failure of a brittle-ductile material like concrete, but is most of the
time combined with a Rankine tension cut-off when tensile stresses occur (Chen, 1982). As
mentioned before, one big disadvantage is that the intermediate stress σ2 is not taken into
account. This fact implies that the biaxial compressive strength fbc for concrete (when σ1 =
σ2 ) is the same as the uniaxial compressive strength fcp , which is contrary to experimental
results. When hydrostatic pressure is increased, the failure surfaces in the J-p space should
be curved lines, which is not the case for the Mohr-Coulomb criterion as indicated in Fig. 6.9.
Furthermore, when used in numerical analysis, the corners of the failure surface represent
singularities, which can be difficult do handle.

173
 Figure 6.9: Mohr-Coulomb criterion in J-p (left) and principal stress space (right)

6.5.3 Drucker-Prager criterion

To overcome the problem of the corners and associated singularities of the Mohr-Coulomb
criterion, the Drucker-Prager criterion was developed. The failure surface is given as:

F (p, J) = α p + J − k = 0 (6.57)

where α and k are material parameters that can be related to the uniaxial compressive and
tensile strength in several ways. This failure surface in principal stress space is clearly a
circular cone whose meridians in the p-J space and the deviatoric cross sections can be seen
in Fig. 6.10.

Figure 6.10: Drucker-Prager criterion in J-p (left) and principal stress space (right)

One of the shortcomings of this material model is the independence of the failure surface
on the Lode angle θ, which has been proved to be the contrary for concrete. Through the
parameters α and k it is possible to adjust the failure surface to the Mohr-Coulomb hexagon
resulting in either an inscribing, circumscribing or best fit circle (see Potts & Zdravković

174
(1999)).

6.5.4 Chen & Chen criterion

For the modelling of concrete, Chen & Chen (1975) proposed a constitutive model that
describes concrete as an elasto-plastic, isotropic material. Three material parameters are
used to define either the failure or the yield surface both in compression and tension in the
principal stress space. Furthermore, it is a plasticity model based on flow theory and includes
isotropic hardening of concrete for multiaxial loading conditions. Assuming that the failure
of concrete depends on the deviatoric stresses and the hydrostatic pressure, the following
surfaces can be established by using the structural stress invariants I1 and J2 , which were
earlier introduced in Chapter 1 (see Fig. 6.11).

.
Figure 6.11: Yield and failure surfaces of Chen & Chen criterion in the principal (left) and
biaxial stress space (right) (from Chen, 1982)

i) Failure surfaces
On the basis of experimental test data the failure surface for the compression-compression
region representing the peak stress state can be written as:

1
Fuc = J2 + Auc I1 − τuc
2
=0 (6.58)
3

In the tension-tension or tension-compression zone the failure surface is given as:

1 1
Fut = J2 − I1 2 + Aut I1 − τut
2
=0 (6.59)
6 3

ii) Yield surfaces

For the description of the onset of plastic deformation two first yield surfaces are intro-
duced. They enclose the elastic area and have a similar form as the failure surfaces. In the

175
compression-compression region the yield surface becomes:

1
Foc = J2 + Aoc I1 − τoc
2
=0 (6.60)
3

The equation of the yield surface for the tension-tension and tension-compression zone has
the following form:
1 1
Fot = J2 − I1 2 + Aot I1 − τot2
=0 (6.61)
6 3

In the above equations the variables Auc and Aut are material constants describing the ulti-
mate stress state at peak stress and can be determined as functions of the uniaxial compressive
strength fcp , the uniaxial tensile strength ftp and the biaxial compressive strength fbc (i.e.
when σ1 = σ2 > σ3 ). In a similar way, the parameters Aoc and Aot represent a yield stress
state and are obtained from the uniaxial compressive yield stress fcy , the uniaxial tensile
yield stress fty and the biaxial compressive yield stress fbcy . For a stress state on the yield
surface the parameter τ takes the yield value of τoc for the compression-compression region
and τot for the tension-tension and tension-compression region. In the same way, the ultimate
values of τ at peak stress are τuc and τut .

The involved material parameters for the compression-compression surfaces can be sum-
marised as:

Auc − Aoc Aoc · τuc

2 − A · τ2
uc oc
αc = 2 − τ2
βc = 2 − τ2
(6.62)
τuc oc τuc oc

∗2 − 1
fbc 2 − f 2
fbc
∗ − 1 · fcp  − f  · fcp
c
Auc = Aoc = (6.63)
2fbc 2fbc c

∗ − f∗2
2fbc fc fbc
 (2f  − f  )
2
τuc =  ∗ bc  · fcp
2 2
τoc =   c bc · fcp
2
(6.64)
3 2fbc − 1 3 2fbc − fc 

fcy fbcy fbc

fc = 
fbc = ∗
fbc = (6.65)
fcp fcp fcp

176
The material parameters for the tension-tension and tension-compression region may be cal-
culated from:

Aut − Aot Aot · τut

2 − A · τ2
ut ot
αt = 2 − τ2 βt = 2 − τ2 (6.66)
τut ot τut ot

1 − ft∗ fc − ft

Aut = · fcp Aot = · fcp (6.67)
2 2

ft∗ 2 fc ft 2

2
τut = · fcp 2
τot = · fcp (6.68)
6 6

ftp fty
ft∗ = ft = (6.69)
fcp fcp

These parameters give fixed values for the initial yield surfaces at the onset of plastic defor-
mations and for the surfaces at peak stress. However, for the general description of the yield
surfaces (termed “loading functions” later in this section), the parameters αc , αt , βc and βt
will be needed in equations 6.75 and 6.76.

iii) Zoning of the stress state

As the failure and yield surfaces are different for the purely compression region and the other
3 tension regions, it is important to determine the correct stress zones in an analysis. In the
biaxial stress plane this zoning is quite obvious as can be seen easily in Fig. 6.11. But for
the generalization to the triaxial stress space some linear functions have to be introduced for
seperating the stress states, which pass through the uniaxial compression and the uniaxial
tensile state. The following conditions have to be satisfied:

Compression-compression region:
 I1
I1 < 0 and J2 + √ < 0 (6.70)
3

Compression-tension region:
 I1
I1 < 0 and J2 + √ > 0 (6.71)
3

Tension-compression region:
 I1
I1 > 0 and J2 − √ > 0 (6.72)
3

Tension-tension region:
 I1
I1 > 0 and J2 − √ < 0 (6.73)
3

177
 √
Fig. 6.12 illustrates the four different stress zones in the I1 - J2 space together with the yield,
failure and subsequent loading functions.

.
Figure 6.12: Zoning of the stress state (from Chen, 1982) (fc = fcp , ft = ftp )

iv) Loading functions

Concrete is assumed to be a strain hardening material and therefore the variable τ is used
as a hardening parameter governed by the plastic strains εp . For isotropic strain hardening
a general loading function can be written as:

F (σ) − τ 2 (εp ) = 0 (6.74)

where τ is the hardening parameter, representing the so called “equivalent (effective) stress”.
When a stress state is on the yield surface, the function F (σ) is such that it equals the initial
yield stress τoc or τot and F reduces to the yield function Foc or Fot . For further loading
the yield surfaces expand and plastic deformations occur until they reach the failure surface
at peak strength Fuc or Fut . The development of the subsequent yield surfaces is governed
by the “equivalent (effective) plastic strains” εp . The equivalent stress-strain relation can be
directly obtained from a uniaxial compressive test and therefore it is possible to compare a
multiaxial stress state with a simple uniaxial stress state. For the compression-compression
region the following loading function is derived from equations 6.58 and 6.60:

βc
J2 + I1
F (σ) = 3 = τ2 (6.75)
αc
1 − I1
3

178
Similiarly, the loading function for the tension-tension and the tension-compression region
has the following form (obtained from equations 6.59 and 6.61):

1 βt
J2 − I12 + I1
F (σ) = 6 3 = τ2 (6.76)
αt
1 − I1
3

The two compressive and tensile material parameters α and β can be determined from equa-
tions 6.62 and 6.66.

v) Hardening rule
For the description of the isotropic expansion of the loading functions a hardening rule is
needed. As mentioned before, the hardening parameter τ depends on the equivalent plastic
strains εp and varies between τo < τ < τu . The components of plastic strains which are
obtained from an associated flow rule (F (σ) = P (σ)) are assumed to accumulate into an
equivalent plastic strain as: ) ) 
p
ε = p
dε = dεpij dεpij (6.77)

The following uniaxial stress-strain relation in Fig. 6.13 may be adopted.

.
Figure 6.13: Idealised uniaxial stress-strain curve for concrete (from Chen, 1982) (fc = fcy ,
fc = fcp , ft = fty , ft = ftp )

When a stress point touches the yield surface plastic deformations start to occur and
the hardening parameter takes the value of τo (associated with the yield stress). For further
loading, both surfaces for compression and tension expand simultaneously and plastic defor-
mations start to develop. The relation between τ and εp is assumed to be non-linear until it

179
reaches the peak point τu at peak strength for both compression and tension. In the post-peak
regime either perfectly plastic behaviour or linear softening of concrete takes place, which
would result in isotropic contraction of the surfaces. In Fig. 6.13 such a post-peak zone is
neglected. At an ultimate plastic strain εpu the material is assumed to be completely crushed
or cracked and the equivalent stress τ reduces to zero.

6.5.5 Further sophisticated constitutive models for concrete

An overview of other plasticity models for concrete of higher order can be found in Babu et al.
(2005). Some of these models are extensions or modifications of existing constitutive models
and by incorporating more model parameters it is possible to capture the main characteristics
of the failure surface of concrete more realistically. Therefore, sophisticated yield surfaces
have a smooth and convex shape and plot as non-circular cross-sections in the deviatoric
plane. The following models can be encountered in the literature:

 Three parameter models: Bresler-Pister criterion (1958), William-Warnke criterion

(1975)

 Four parameter models: Ottosen criterion (1977), Reimann criterion (1965), Hsieh-
Ting-Chen criterion (1979)

For limiting the failure of concrete under purely hydrostatic pressure in compression Hof-
stetter et al. (1993) proposed a so called cap-model. Huber (2006) took up this idea and
introduced a cap of spherical shape, coupled with a conventional Drucker-Prager criterion
for the behaviour in compression. More recently constitutive models based on fracture and
continuum damage have been developed, where the microstructural degradation process, i.e.
the degradation of the Young’s modulus upon unloading in the post-peak regime, is modelled
at a phenomenological level (Meschke et al., 1998). Finally, micromechanical models attempt
to develop the macroscopic stress-strain relationship from the mechanics of the microstruc-
ture (Babu et al., 2005). However, it has to be questioned if such models are from a practical
point of view more suitable for the modelling of shotcrete in a boundary value problem.

6.6 Modelling post-peak behaviour of concrete

6.6.1 Post-peak behaviour in tension
As already highlighted in Chapter 5, crack formation of concrete and its propagation is a
highly discrete problem and difficult to handle in a numerical analysis. According to Chen
(1982) two different types of approaches for the numerical simulation of concrete fracture can
be identified:

1. Discrete crack model

2. Smeared crack model

180
The selection of an adequate cracking model depends strongly on the purpose of the analysis
to be performed. If the aim is the general overall behaviour of a concrete structure, without
any particular need for realistic crack patterns and local stress concentrations, the smeared
crack approach is probably the best choice. On the contrary, if the area of interest to be
analysed is a very local one, a discrete crack model might be more useful to adopt. The basic
principles and the associated problems of each approach are discussed briefly below.

6.6.1.1 Discrete crack model

In the discrete crack approach, which started to be developed in the late 1960s, a crack is
introduced as a geometric entity. This is normally achieved by disconnecting the displacement
at nodal points of adjoining elements, when the nodal force ahead of the crack tip exceeds a
tensile strength criterion, as can be seen in Fig. 6.14. As a consequence cracks are forced to
propagate along element boundaries, introducing a certain mesh bias (De Borst et al., 2004).
Complex and time-consuming techniques, such as automatic remeshing, are needed in order
to overcome such problems. These limitations and difficulties involved with proper robust
three-dimensional implementations result in a very limited acceptance of the use of discrete
crack models (Chen, 1982).

Figure 6.14: Early discrete crack modelling (from De Borst et al., 2004)

6.6.1.2 Smeared crack model

In the smeared crack approach the cracked concrete is assumed to remain as a continuum
(Chen, 1982). A single crack is not discrete but implies an infinite number of parallel fissures
across the part of the finite element where the crack occurs. Generally, when the principal
stress at a certain integration point reaches the tensile strength, a crack is initiated leading to
orthotropic material behaviour and a deterioration of the current stiffness and strength at that
particular integration point. Experimental results lead to the conclusion that plain concrete
is not a perfectly brittle material, but that it has some residual load-carrying capacity after
reaching its tensile strength (De Borst, 2002). Therefore, the material exhibits tensile strain-
softening and a descending branch has to be introduced into a model to capture the gradually
diminishing tensile strength of concrete upon further crack opening. If the direction of the
crack remains orthogonal to the direction of the major principal stress, a rotating smeared

181
crack model is obtained, whereas a fixed direction of the crack at crack initiation leads to a
fixed smeared crack model.

However, it is well reported in the literature that the inclusion of strain-softening material
behaviour in a numerical analysis leads to certain difficulties (Cendón et al., 2000; Meschke,
1996; Pamin, 1994; Vermeer & Brinkgreve, 1994; Arsland & Sture, 2008; Zervos et al., 2007;
Mosler & Meschke, 2004; De Borst et al., 2004; Möller et al., 2004; Bićanić & Pearce, 1996;
De Borst, 1987). Initiation of strain localisation results in material instability, the loss of
ellipticity of the governing equations and spurious mesh sensitivity may appear (Mosler &
Meschke, 2004). According to Pamin (1994) these problems can be handled by either mod-
elling cracking in a discrete way (see above) or applying an enhanced continuum approach,
where three concepts have been proved to be successful so far:

1. Cosserat (micropolar) continuum

2. Non-local (integral) model

3. Higher-order gradient continuum

All these regularisation techniques have in common that they make use of an internal length
parameter related to the specific material. Due to their complexity and the requirement for a
sufficiently fine resolution of the localisation zone to guarantee mesh objectivity resulting in
high computational effort, these approaches are of limited value for analysing a real boundary
value problem in engineering practice.

An elegant way to eliminate mesh dependency of a strain-softening material in a numerical

analysis was presented by Hillerborg et al. (1976). In their work a method is proposed in
which fracture mechanics is incorporated into a cracking model based on an energy balance
approach, often termed “fictitious crack model”. Since cracking of concrete is a discrete
process, a stress-crack-opening relationship is adopted to describe the post-peak behaviour
in tension, as can be seen in Fig. 6.15.

Figure 6.15: Deformation characteristics of tensile post-peak behaviour

The governing parameter for this curve is the so called fracture energy Gf , which rep-
resents the energy required to propagate a tensile crack of unit area until complete crack

182
opening. The graphical interpretation of the fracture energy is the area below the stress-
crack-opening curve and it can be written as:
) wu
Gf = σ dw (6.78)
w=0

Therefore, the fracture energy is independent of the test method or size of the sample and
can be interpreted as a real material parameter. In the absence of experimental data, the
CEB-FIP Model Code (1990) allows to estimate the value for Gf for hardened concrete at
28 days from the following equation:
 0.7
fcm
Gf = Gf o (6.79)
fcmo

where fcm is the mean compressive strength and fcmo = 10 MPa. Gf o is the base value of
fracture energy and depends on the maximum aggregate size dmax . Table 6.3 provides the
following values:

dmax Gf o
(mm) (Nmm/mm2 )
8 0.025
16 0.030
32 0.058

Table 6.3: Base values of fracture energy Gf o from CEB-FIP Model Code (1990)

Although very few results are available in the literature concerning the fracture energy of
early age concrete and shotcrete, experiments show that the fracture energy is not constant,
but of an increasing nature with time or degree of hydration (Gutsch & Rostásy, 1994).
De Schutter & Taerwe (1997) proposed an equation based on the degree of hydration which
is of the following form:
 d
Gf (r) r − ro
= (6.80)
Gf (r = 1) 1 − ro
where r is the degree of hydration, Gf (r = 1) is the specific fracture energy at a degree of
hydration r = 1 and ro and d are parameters. Brameshuber & Hilsdorf (1989) derived an
equation for the fracture energy from their test results with different Portland cements, where
Gf increases with time. Lackner & Mang (2004) use the experimental work of Brameshuber
& Hilsdorf (1989) to establish a function for the linear increase in fracture energy with degree
of hydration:
Gf = 0.488 + 77.8 ξ (6.81)

with ξ being the degree of hydration ranging from 0 to 1 and Gf the fracture energy in
10−3 Nmm/mm2 . The data set for one particular cement together with the above expression

183
is shown in Fig. 6.16.

Figure 6.16: Intrinsic material function for fracture energy (from Lackner & Mang, 2004)

De Borst & van den Boogaard (1994) investigated the modelling of deformation and crack-
ing in early age concrete by applying a smeared crack model. Although they concentrate on
the issue that the fracture energy appears to be an increasing parameter with time, they
neglected this and adopted a much simpler approach with a constant strain input parameter.
Finally, De Schutter & Taerwe (1997) make the criticism that even at some major interna-
tional conferences regarding the behaviour of early-age concrete, no or very little attention is
paid to the experimental study of the evolution of the softening behaviour of concrete during
hardening.

For the modelling of crack formation in a finite element analysis within the smeared crack
concept, the crack-opening has to be transformed into a crack-strain via a length parameter:

w
ε= (6.82)
leq

In this equation, leq is the equivalent length of a finite element (or sometimes termed char-
acteristic length) and should correspond to a representative dimension of the mesh size.
Therefore it depends in general on the chosen element type, size, shape and integration order
(Feenstra, 1993; Thomée, 2005). In the literature a couple of approaches for estimating leq
can be found and most of them ignore the orientation of the crack within the element and
simply relate the equivalent length to the area of the element. Rots (1988) suggested the
following equation for a 2D element:

leq = α Ae (6.83)

where Ae is the area of the element and α a scalar factor, being 1.0 for quadratic elements
√
and 2 for linear elements. Feenstra (1993) mentions that this approximation is relatively
accurate if the mesh is not distorted too much and when the cracks are aligned with the
element boundaries. He states further, that the choice of the equivalent length depends also
on the problem to be analysed. Thomée (2005) and Haufe (2001) showed in their work
that this approach is suitable for most practical applications in engineering practice. Strack
(2007) applied in his research on the numerical modelling of steel fibre reinforced concrete an

184
equation suggested by Pölling (2000), taking into account the number of integration points
per element nint : 
Ae
leq = (6.84)
nint
Huber (2006) used in his analyses a similar equation for volumetric elements proposed by
Bićanić et al. (1994): 
3 Ve
leq = (6.85)
nIP
where Ve is the volume of the finite element and nIP the number of integration points.

With the introduction of the characteristic length and the fracture energy as additional
parameters, it is guaranteed that during material softening the same amount of energy is
released within the element until complete crack opening and hence the results obtained are
objective with regard to mesh refinement. A more sophisticated and consistent estimation
and discussion of the equivalent length can be found in Oliver (1989).

For an unreinforced cracked concrete section, the CEB-FIP Model Code (1990) provides
a bilinear stress-crack-opening relation as can be seen in Fig. 6.17.

Figure 6.17: Stress-strain and stress-crack-opening diagram for uniaxial tension (from CEB-
FIP Model Code, 1990)

The mathematical description is as follows:

 
w
σct = fctm 1 − 0.85 for 0.15fctm ≤ σct ≤ fctm (6.86)
w1

and
0.15fctm
σct = (wc − w) for 0 ≤ σct < 0.15fctm (6.87)
wc − w1
where σct is the uniaxial tensile stress, w the crack opening (mm), w1 the crack opening (mm)
at a tensile stress of 0.15fctm and wc the complete crack opening (mm) at zero stress. wc
may be estimated from:
Gf
wc = αf (6.88)
fctm

185
with Gf being the fracture energy in Nmm/mm2 and fctm the mean uniaxial tensile strength
in MPa. The coefficient αf depends on the maximum aggregate size dmax and is given in
Tab. 6.4.

dmax (mm) 8 16 32
αf (-) 8 7 5

Table 6.4: Coefficient αf to estimate wc (from CEB-FIP Model Code, 1990)

Haufe (2001), Meschke (1996), Feenstra (1993) and Huber (2006) describe the post-peak
behaviour of concrete in tension with an exponential function where an internal damage
parameter κ controls the decay in the uniaxial tensile stress σt with crack growth:
 
κ
σt = fctm exp (6.89)
κu

where the ultimate damage parameter is calculated as:

Gf
κu = (6.90)
leq fctm

König & Duda (1991) compared different stress-crack-opening relations suggested by various
researchers in the literature, as can be seen in Fig. 6.18, and they concluded that apart from
the linear softening, which is often chosen because of simplicity, all the curves have in common
that the initial branch is very steep, transforming into a flat part on approaching or reaching
zero. Furthermore, the most practical applications seem to be insensitive to the exact shape
of the softening curve adopted.

Figure 6.18: Some proposed stress-crack-opening relations (from Moussa, 1993)

186
6.6.1.3 Effect of reinforcement

Huber (2006) distinguishes three different types of modelling of steel reinforcement for ana-
lyzing reinforced concrete structures (see Fig. 6.19), which are:

1. Discrete modelling

2. Smeared modelling

3. Embedded modelling

Figure 6.19: Models for reinforcement

In the discrete modelling concept, reinforcement is simulated as individual elements, usually

aligned along the boundaries of the concrete elements. Therefore, a proper constitutive
model for steel has to be involved describing the uniaxial behaviour of the reinforcement
bars. However, one drawback of this approach is that a high level of mesh refinement is
required in the area near the reinforcement bars, which results in high computational effort.
Furthermore, complex interface elements have to be used in order to capture the real bond-
slip behaviour between concrete and steel (Huber, 2006). Examples for this type of modelling
reinforcement are given in Eierle & Schikora (1999).

Probably the most frequent method for analysing reinforced concrete structures is to
smear or evenly distribute the reinforcement over the concrete element (Feenstra, 1993). This
is achieved by modifying the constitutive equations taking into account the additional internal
forces and stiffness due to the reinforcement bars, leading to a certain anisotropic behaviour
of the concrete element. This methodology was used by Haufe (2001) for performing some
numerical studies on reinforced plates, shells and folded plates.

Finally, the embedded formulation of reinforced concrete considers steel bars as discrete
elements but with the difference that the elements are of the same type as the concrete
elements (number of nodes, degrees of freedom, shape functions). Therefore, a uniaxial steel
bar is transformed into a two- or three-dimensional element and then embedded into the
concrete element. A big advantage of this method is that any orientation of the reinforcement
bar within the element can be chosen and results of the overall structure are independent of

187
the alignement of reinforcement. For further information about the embedded reinforcement
concept seek Huber (2006).

Due to the complexity of the three approaches presented above regarding discretization
and real bond-slip behaviour, reinforcement is rarely considered in a finite element analysis
of tunnel construction (Thomas, 2003).

However, in the case of a smeared crack model it is possible to take into account the
effect of reinforcement by modifying the post-peak behaviour of concrete in tension and
the so called “tension-stiffening effect” is introduced. This term is used to describe the
interaction between steel reinforcement and concrete after crack formation (Moussa, 1993).
The existence of reinforcement improves the softening response of concrete by contributing
to the overall stiffness of the system. This is achieved by a gradual redistribution of internal
forces from concrete to reinforcement during crack propagation due to their bond behaviour
until a stabilised crack pattern has developed. The average stresses in the concrete would
not decrease with further crack opening as quickly as in the case of plain concrete but would
gradually reduce to zero, as indicated in Fig. 6.20.

Figure 6.20: Principal effect of tension-stiffening on post-peak behaviour of concrete

In the literature many attempts can be found to model the tension-stiffening effect for
reinforced concrete, among them Nayal & Rasheed (2006), Ebead & Marzouk (2005), Bischoff
& Paixao (2004) and Fields & Bischoff (2004). Fig. 6.21 shows a collection of some stress-
strain diagrams for concrete in tension by Gilbert & Warner (1978), including the tension-
stiffening effect.

188
Figure 6.21: Stress-strain diagram for reinforced concrete in tension (from Gilbert & Warner,
1978)

Feenstra (1993) critisises the fact that in most of the different formulations for tension-
stiffening that can be encountered in the literature no reference is made to the fracture energy,
which is actually released in the material. For a numerical analysis he suggested to estimate
the fracture energy for reinforced concrete Grc
f in the following way:

 
leq
Grc
f = Gf 1 + (6.91)
ls

where Gf is the fracture energy of a single crack and leq the equivalent length of the finite
element. ls is the average crack spacing which is in general a function of the bar diameter,
the concrete cover and the reinforcement ratio and can be calculated according to the CEB-
FIP Model Code (1990). Haufe (2001) states that in a finite element analysis of reinforced
concrete it is possible that several cracks occur within one element and therefore he proposed
to establish the length parameter used for regularisation as:

rc
leq = min (ls , leq ) (6.92)

In a similar way Bockhold (2005) evaluates the fracture energy for reinforced concrete through:
 
leq
Grc
f = max Gf , Gf (6.93)
ls

Lackner & Mang (2003) investigated the cracking of reinforced shotcrete tunnel shells by
adopting a smeared cracking concept. They take into account the tension-stiffening effect by

189
multiplying the fracture energy by a factor γ > 1:

Grc
f = γ Gf (6.94)

This factor γ is computed by means of a 1D composite model presented in Lackner & Mang
(2000) consisting of one steel bar embedded in shotcrete including a non-linear bond slip-
bond stress relation extended for ageing materials. From their calculations they derived a
linear function for γ (see Fig. 6.22), depending on the cement hydration degree ξ which can
be expressed as:
γ (ξ) = 22.8 − 15.0 ξ (6.95)

Figure 6.22: Material function for γ depending on degree of hydration ξ (from Lackner &
Mang, 2003)

However, it is to be questioned if such approaches are suitable for the description of

reinforced shotcrete, since the assessment of crack spacing, bond-slip behaviour and other
mechanical aspects for shotcrete at early ages are still not clear from a scientific point of
view.

6.6.1.4 Modelling steel fibre reinforced concrete (SFRC)

As already mentioned in Chapter 5, the inclusion of steel fibres in concrete or shotcrete

results in a more ductile material behaviour after peak for both compression and tension.
Thomée (2005) gives an overview about different qualitative stress-crack-opening relations
for tension from various researchers (Hillerborg, 1985; Kooiman, 2000; Matsuo et al., 1995;
Kützing, 2000), which can be seen in Fig. 6.23.

190
Figure 6.23: Different stress-crack-opening relations for steel fibre reinforced concrete (from
Thomée, 2005)

From these curves it can be observed that for the uniaxial behaviour, the softening branch
can be separated basically into two parts. Directly after reaching the maximum tensile stress
there is a steep drop in stresses, since the mechanical contribution of the steel fibres is still
low and the governing parameter is the fracture energy of plain concrete. In the second part
the stresses appear to stabilise with an almost constant or a slightly decreasing value. Stang
& Aarre (1992) describe the post peak behaviour with the following expression (curve f) in
Fig. 6.23):
fctm
σt = * w +p (6.96)
1+
w∗
Through an empirical variation of the involved parameters p and w∗ the curve can be fitted
to the post-peak behaviour of concrete with various steel fibre contents. Kazemi et al. (2007)
concluded that in principle the stress-crack-opening curve depends heavily on the type and
the amount of fibres adopted. Thomée (2005) developed in his research a stress-crack-opening
relation consisting of an exponential and a linear softening branch that can be seen in Fig. 6.24.

191
 Figure 6.24: Uniaxial tensile stress-crack-opening relation (from Thomée, 2005)

In order to use this model in a finite element analysis he converted the crack opening
w into a strain ε via the equivalent length leq (see previous sections). At crack opening the
exponential softening curve for plain concrete proposed by Feenstra (1993) controls the decay
in tensile stress and is given as:
 
ε
σt = fctm exp − for ε < ε1 (6.97)
εe

with  
Gf 1 fctm1
εe = and ε1 = −ln εe (6.98)
fctm leq fctm
where fctm is the tensile strength, fctm1 a residual tensile strength and Gf 1 the fracture
energy of plain concrete. The linear softening branch representing the impact of steel fibres
can be expressed as:
 
ε − ε1
σt = fctm1 − fctm1 for ε1 < ε < εu (6.99)
εu − ε1

where
* +
2 Gf 2 + Gf 1  
ε1 fctm leq
εu = ε1 + and Gf 1 = Gf 1 exp − (6.100)
fctm1 leq Gf 1

Gf 2 is the part of the total fracture energy that can be related to the contribution of the steel
fibres and may be estimated from experimental tests. In his numerical examples Thomée
(2005) used values for Gf 2 ranging from 2.10−3 to 9.10−3 MNm/m2 . Barros & Figuieiras
(1999) provide some empirical equations that allow the calculation of the increased total
fracture energy from the amount of fibres used in the concrete mix. Their equations are of
the type:
Gff r
= a + b Wf (6.101)
Gf o

where Gff r is the total fracture energy including fibres, Gf o the fracture energy of plain

192
concrete, Wf the fibre content in per cent of concrete weight and a and b two constants
depending on the type of fibre. Fig. 6.25 shows some test results for the total fracture energy
increasing with volumetric content of steel fibres. For further information about modelling
the effect of steel fibres on the behaviour of concrete and shotcrete see Strack (2007).

Figure 6.25: Increase in fracture energy with fibre content (from Kazemi et al., 2007)

6.6.2 Post-peak behaviour in compression

For a consistent formulation of the complete post-peak behaviour, the softening behaviour
of concrete in compression can be treated in a similar context as the tensile regime. Ac-
cording to Feenstra (1993) a compressive fracture energy Gc as a real material parameter
describes the softening behaviour of concrete in compression. Such an assumption can be
considered reasonable since the underlying failure mechanisms for compression and tension in
the post-peak regime are identical, i.e. continuous crack growth at micro level. According to
experiments performed by Vonk (1992), the compressive fracture energy Gc ranges from 10
to 25 Nmm/mm2 and is therefore roughly 50 -100 times the tensile fracture energy Gf . With
the help of the equivalent length leq regularisation of the softening behaviour is achieved and
the post peak behaviour can be described as mesh independent.

6.7 Modelling of creep

Since it was observed from experimental data that the progress of creep follows a definite
pattern, many attempts have been made to express creep in a mathematical way. It is well
known that initially, after the application of loading, a considerable amount of creep takes
place, resulting in a steep deformation curve at the beginning, followed by a flattening curve
controlled by reducing creep rates. However, Neville (1970) states that it is not clear whether
there is a finite limit to creep after infinite time or not. Therefore, two broad types of creep
laws exist: those which tend to a limiting value and those which increase indefinitely. In the
following sections a selection of different creep models available in the literature is presented,
with a particular focus on the attempts to model the complex creep behaviour of shotcrete.

193
6.7.1 Basic expressions for creep
According to Neville (1970) the existing basic creep expressions date back to the 1930s and
can be divided into the following groups:

 Power expressions (Shank, 1935)

 Exponential expressions (Thomas, 1933)

 Hyperbolic expressions (Ross, 1937)

 Logarithmic expressions (U.S. Bureau of Reclamation, 1955)

All the parameters involved in such creep models are of empirical or quasi-rational nature
and therefore their application for general cases has to be treated with caution. This fact
makes the application of a basic creep model for shotcrete relatively difficult, since the creep
behaviour of sprayed concrete appears to be highly non-linear, depending on age of loading,
stress level, temperature and other aspects related to material technology. One of the simplest
expressions for predicting uniaxial concrete creep was suggested by Straub (1930) and is of
the form:
εcr = Q σ A tB (6.102)

where t is the time, εcr are the creep strains, σ is the applied stress and A, B and Q are
parameters depending on the properties of concrete, the time of application of the load and
the units used in the analysis. Alkhiami (1995) and Probst (1999) suggested some values for
these parameters based on experimental results. Kuttner (1989) introduced this model in a
slightly modified form for the simulation of creep in shotcrete. It can be written as:

A
εcr = σ n tm+1 (6.103)
m+1

where A is the temperature-dependent material parameter and m and n are material con-
stants. As this expression is one of the standard creep models in the finite element code
ABAQUS, a couple of researchers adopted this model in their work for the analysis of shotcrete
as tunnel support, among them Kienberger (1999), Galler (1997) and Golser (1999).

6.7.2 Rheological units

The use of rheological models for describing the viscoelastic behaviour of concrete and
shotcrete is quite common in engineering practice due to their simplicity. Mostly an ar-
rangement of several rheological units (Hook spring, Newton dashpot and St. Venant plastic
element) serves for the prediction of uniaxial creep behaviour of concrete by fitting rheo-
logical properties to the observed behaviour from experimental tests. Therefore, the fitted
constants represent the material behaviour of a particular concrete or shotcrete under given
conditions and are not generally applicable. This fact is a severe drawback since it is the
actual data that predict the model and not vice versa. Furthermore, a detailed understanding

194
of the fundamental creep behaviour is lost as these models imply nothing about the molecular
mechanisms responsible for creep. However, they can be useful for visualizing the basic creep
effects and especially the superposition of deformations. A detailed overview of different
rheological models can be found in Neville (1970) and Flügge (1967) and the basic ones are
presented in the following sections.

6.7.2.1 Kelvin model

Fig. 6.26 shows the Kelvin model, where a spring and a dashpot are in parallel so that they
undergo the same displacement. The spring is characterised with a spring stiffness Ek , while
the dashpot is described by the viscosity coefficient ηk .

Figure 6.26: Kelvin body

The total applied stress is the sum of the forces on the individual elements and it can be
written as:

σ = σspring + σdashpot
(6.104)
ε = εspring = εdashpot

with

σspring = Ek ε
(6.105)
σdashpot = ηk ε̇

Substitution leads to a differential equation:

σ(t) = Ek ε + ηk ε̇ (6.106)

with the solution taking the form:

σ * − t
+
ε= 1 − exp τk (6.107)
Ek

where τk = ηk /Ek is the “retardation time”, representing the time required for the defor-
mation to attain a value equal to 1/ exp of its ultimate magnitude. Fig. 6.27 shows the
deformation under a sustained load and after its removal, approaching full recovery asymp-
totically. Therefore, the Kelvin model is suitable for predicting the phenomenon of delayed
elasticity.

195
 Figure 6.27: Deformation response of Kelvin model

6.7.2.2 Maxwell model

The Maxwell model consists of a spring and a dashpot in series (see Fig. 6.28), where the
load carried by both elements is the same:

Figure 6.28: Maxwell body

The mathematical formulation is given as:

σ = σspring = σdashpot
(6.108)
ε = εspring + εdashpot

Making use of equation 6.105 results in a differential equation which can be expressed as:

σ̇ σ
ε̇ = + (6.109)
Em ηm

Hence, the increase in deformation is given as:

σ σ
ε= + t (6.110)
Em ηm

Under a sustained load the development of the deformation is unlimited and consequently the
model represents a liquid. Furthermore, with a constant viscosity parameter ηm the deforma-

196
tion curve is a straight line, whereas a non-linear expression for ηm leads to a curved shape
(Vaishnav & Kesler, 1961). After removal of the load, a permanent or plastic deformation
remains, as can be seen in Fig. 6.29.

Figure 6.29: Deformation response of Maxwell model

If the Maxwell model is subjected to a constant deformation it exhibits the property of

relaxation and the reduction in stress can be written as:
t
σ = σo exp− τm (6.111)

where σo is the initial stress and τm = ηm /Em the so called “relaxation time” representing
the time in which the stress relaxes to 1/ exp of its original value.

6.7.2.3 Burger model

A more complex rheological body is the Burger model, which is a series combination of a
Kelvin model and a Maxwell model, as shown in Fig. 6.30.

Figure 6.30: Burger body

Consequently, the deformation response of a Burger model is the sum of its Kelvin and
Maxwell components and it can be written as:
 E

σ σ σ − ηk t
ε= + t+ 1 − exp k (6.112)
Em ηm Ek

197
where the subscripts k and m refer to the Kelvin and Maxwell components respectively.
Fig. 6.31 indicates schematically the deformation response of a Burger model. Directly after
the load is applied an instantaneous elastic deformation takes place followed by a time-
dependent increase in deformation at a decreasing rate, tending asymptotically to a straight
inclined line. On removal of the load, firstly the elastic recovery takes place followed by a
time-dependent recovery approaching a horizontal line. Hence, part of the time-dependent
deformation is irreversible and it is believed that the Burger model represents qualitatively
the creep behaviour of concrete reasonably well.

Figure 6.31: Deformation response of Burger model

6.7.2.4 Rheological models used for modelling shotcrete behaviour

In the literature a couple of models based on the combination of various rheological bodies
can be found for simulating creep deformation of young shotcrete. Most of them have in
common that they include sophisticated functions for their rheological parameters depending
on shotcrete age and stress level, in order to capture in a better way the complex behaviour
observed in experimental tests. The basic principles of these models are presented here with
the aim of highlighting the problems involved in their development.

On the basis of eight conventional short term creep tests, with loading ages varying
between 4 h to 168 h, Kuwajima (1999) proposed a linear Kelvin model for describing the
viscoelastic deformation of young shotcrete. It is of the form:
 
εc (t) 1 Et
= 1 − exp− η
t
(6.113)
σo Et

with 1/Et being the final specific creep and Et /η indicates how fast the creep deformation
would develop. It was concluded that, although creep presents a strong age dependency for
the early ages, after an age of 10 hours a time-independent creep model can be applied. The

198
suggested values of the creep parameters are the following:

1 Et
= 0.03 GPa−1 and = 0.003 min−1 (6.114)
Et η

Rokahr & Lux (1987) take into account that the creep rate of young shotcrete is decreasing
with time and depends highly on the applied stress level (see Fig. 6.32).

Figure 6.32: Non-linear Kelvin model after Rokahr & Lux (1987) (from Hesser, 2000)

Their proposed creep law is a modified Kelvin model based on strain-hardening theory
and the viscous strain rate can be written as:
  
1 εv
{ε̇v } = 1−3 Gk (σv ) [M2 ] {σ} (6.115)
2ηk (σv , t) σv

where {ε̇v } is the creep rate vector, {σ} the stress vector, σv the effective stress, εv the
effective strain and [M2 ] a filter matrix. The effective strain and stress respectively are given
in Hesser (2000) as:
√ 
2
εv = (ε1 − ε2 )2 + (ε2 − ε3 )2 + (ε3 − ε1 )2 (6.116)
3

and  
1
σv = (σ1 − σ2 )2 + (σ2 − σ3 )2 + (σ3 − σ1 )2 (6.117)
2
No information could be found about the formulation of the filter matrix [M2 ]. The usual
viscosity parameters are replaced by viscosity functions, which are expressed as:

Ḡk = 3 Gk = G∗k expk1 σv and η̄k = 3 ηk = ηk∗ tn expk2 σv (6.118)

Values for the above material parameters G∗k , ηk∗ , n, k1 and k2 , used in the time- and stress-
dependent viscosity functions, based on tests performed by Wierig & Gollasch (1967) can be
found in Hesser (2000) and Yin (1996).

Peterson (1989) performed long term creep tests over 10 days on shotcrete samples starting
at an age of 30 h. Based on his experimental results he published the following non-linear

199
creep model based on time-hardening theory:
Ḡ (σ )
1 −t k v 3
{ε̇v } = exp η̄k (σv ,t) [M2 ] {σ} (6.119)
3 η̄k (σv , t) 2

As in the previous model, the creep rate is dependent on time and stress-level. However,
one drawback of this formulation is that it is limited to the lower stress regime (smaller than
∼10 MPa), otherwise the creep rate would decrease with increasing stress, which does not
represent real creep behaviour (Yin, 1996). Pöttler (1990) tried to overcome this problem by
proposing a creep model of a polynomial form, given as:

 3
{ε̇v } = α (t) σv2 + β (t) σv3 [M2 ] {σ} (6.120)
2

with
 
α (t) = 0.02302 − 0.01803 t + 0.00501 t2 10−3
  (6.121)
β (t) = 0.03729 − 0.06656 t + 0.02396 t2 10−4

With the help of the above modification the applicable stress range for the correct modelling
of creep is extended. However, over the first 1.5 days the model predicts a creep rate that
is reducing as one would expect to observe in an experimental test. For creep tests longer
than that, the model simulates a creep rate that would suddenly increase again, which is
not realistic (Hesser, 2000). Zheng (1989) developed a non-linear creep model based on the
combination of various rheological bodies, as can be seen in Fig. 6.33.

Figure 6.33: Constitutive behaviour of shotcrete after Zheng (1989) (from Hesser, 2000)

The time-dependent Newton dashpot is intended to describe the creep deformation of

young concrete ε1 , whereas the behaviour of older concrete is simulated by two Kelvin units
which are placed in series (ε2 ). The creep rate for young concrete is estimated according to
Peterson (1989) and the creep rate for older concrete can be calculated from:

3  1
2 E
−t k,r
{ε̇2 } = exp ηk,r [M2 ] {σ} (6.122)
2 ηk,r
r=1

Yin (1996) provides some values for the material parameters but states also that some in-
consistency exists for this model regarding the parameter derivation and that the applicable
stress range is limited. Zachow (1995) modifies the model presented by Peterson (1989) by

200
proposing new viscosity functions in order to obtain better agreement with lab test data for
shotcrete. In his work the creep rate is given as:
 
tn2
σv + n1 exp − (n2 tn2 ln (t) − σv ) Ḡ
σv − η̄ k t
{ε̇v } = exp k (6.123)
η̄k

where the viscosity function equals:

* n +
n1 exp − tσ 2
η̄k = η̄k∗ exp k2 σv
t v (6.124)

where k2 , n1 and n2 are material parameters that can be derived from shotcrete creep tests.

Hesser (2000) agrees on the fact that with this model a better reproduction of creep behaviour
is possible, but shows as well that certain limitations exist regarding the applicable shotcrete
age (max. 14 days) and stress range (less than 35 MPa). Another empirical creep model is
published by Yin (1996) using an exponential function for the creep rate.

3
{ε̇} = a (t) σvb(t)−1 M2 {σ} (6.125)
2

with    
A2 B2
a (t) = A1 exp and b (t) = B1 exp (6.126)
A3 + t B3 + t
A1 , A2 , A3 , B1 , B2 and B3 are material parameters for a particular shotcrete type. On the
basis of published information in the open literature Thomas (2003) defined an age-dependent
Kelvin body for uniaxial short-term creep of shotcrete. He extended the model further for
the 3D case and it can be written as:
 G

Ṡij Sij 1 − ηk t
ėij = + 1 − exp k (6.127)
2G 2 ηk

where G is the elastic shear modulus, Sij is the deviatoric stress, Ṡij the deviatoric stress
rate and ėij the deviatoric strain rate. The viscosity functions are given as:
−1.5 −1.0
1.5 1011 exp t0.6 8.0 106 exp t0.4
ηk = and Gk = (6.128)
2 (1 + μ) 2 (1 + μ)

with μ being the Poisson’s ratio. Finally, the rheological model from Hauggaard et al. (1997)
used for modelling creep of early age concrete is a non-linear Burger model as shown in
Fig. 6.34.

201
 Figure 6.34: Non-linear Burger model (from Hauggaard et al., 1997)

The development of this model contains also a complex generalisation to 3D and the viscosity
functions are given as:
  c 
b
E (t) = a exp −
t (6.129)
η (t) = a (1 − exp (−b t)) (1 − c exp (−d | t − f | ))
e

where a, b, c, d, e and f are concrete dependent parameters. These equations take into
account that the viscosity increases only slightly initially and after about 200 hours the
major increase takes place, as can be seen in Fig. 6.35.

Figure 6.35: Functions for viscosity and stiffness in non-linear Burger model (from Hauggaard
et al., 1997)

Further expressions of the viscosity functions for a non-linear Burger model can be found

202
in Bosnjak (2000), Hagihara et al. (2002) and Atrushi (2003).

6.7.3 Creep according to EC 2 (2004)

EC 2 (2004) provides a procedure for the calculation of creep deformation for normal cast
concrete structures. Jones (2007) aimed to verify that these creep predictions are also appli-
cable for sprayed concrete and can be used in a numerical analysis of tunnel construction. His
investigations are based on comparisons with some numerical results from Thomas (2003),
who defined an age-dependent generalised Kelvin model for uniaxial creep and fitted the
involved parameters to creep tests on shotcrete samples in the literature (see section above).

According to EC 2 (2004) uniaxial compressive creep of concrete is considered via a creep

coefficient φ which relates the time-dependent behaviour to the elastic deformation:

σc
εc (t, to ) = φ (t, to ) (6.130)
E

where σc is the applied compressive stress and E the Young’s modulus. The creep coefficient
is a function of various influencing factors, such as:

 Relative humidity RH

 Mean compressive concrete strength fcm

 Concrete age at loading to

 Cross sectional area A and perimeter of member u in contact with the atmosphere

 Cement type α

 Temperature T

If the compressive stress of concrete at the age of loading to exceeds 45 % of the compressive
strength, then creep non-linearity should be considered and the creep coefficient φ(t, to ) in
equation 6.130 is replaced by:

φk (t, to ) = φ (t, to ) exp (1.5 (kσ − 0.45)) (6.131)

with kσ = σc /fcm being the stress level. Considering a shotcrete tunnel lining of 30 cm
thickness, cement class N, a compressive cylinder strength of 30 MPa and a relative humidity
of 80 %, Jones (2007) showed that reasonable agreement can be achieved by comparing the
procedure from EC 2 (2004) with experimental data for shotcrete. It is further stated to
divide the elastic modulus by 1.5 to allow for creep in a numerical model, taking into account
that much of the loading in tunnel construction would occur at early ages.

203
6.7.4 Rate of flow method
The rate of flow method was suggested by England & Illston (1965) and is based on the divi-
sion of creep into reversible and irreversible components. The principle is shown in Fig. 6.36,
where the total strains consist of instantaneous elastic strains, delayed elastic strains and irre-
versible “flow” strains. Several modifications of the method exist for the description of creep
of shotcrete developed by various researchers at the University for Mining and Metallurgy in
Leoben, Austria. They are presented here briefly.

Figure 6.36: Deformation history with time(from England & Illston, 1965)

6.7.4.1 Rate of flow method after Schubert (1988)

The deformation of shotcrete in uniaxial compression can be decomposed into 4 different

components, as illustrated in Fig. 6.37 and explained in the list below. Upon loading some
instantaneous elastic deformation occurs followed by a time-dependent creep deformation.
This creep deformation is made up of a reversible “delayed” elastic component and the irre-
versible viscous deformation. Furthermore, the total deformation includes some deformation
developing with time due to hydration temperature and shrinkage of the shotcrete.

204
 Figure 6.37: Deformation components for young shotcrete (from Schubert, 1988)

a) Instantaneous elastic deformation, governed by the time-dependent Young’s modulus

E(t)

b) Delayed elastic creep deformation, approaching a limit strain

c) Irreversible viscous deformation (or so called “flow”)

d) Deformation due to shrinkage and hydration temperature

For the calculation of each component, the stress-strain history is divided into time steps Δt
and the total strain is written as:

σ1 − σ2
ε2 = ε1 + + Δεv + Δεd + Δεsh + Δεt (6.132)
E (t)

where ε1 and σ1 are the strains and stresses at the beginning of the increment at time t1 ,
ε2 and σ2 the strains and stresses at the end of the increment at time t2 , E (t) the time-
dependent Young’s modulus, Δεsh the change in shrinkage deformation and Δεt the change
in thermal deformation. The irreversible viscous strains are given as:

Δεv = σ2 ΔC(t) (6.133)

The function ΔC (t) controls the change of the specific irreversible creep deformation and is
given by:
1
C (t) = A t 3 (6.134)

205
with A being a material constant and t the time in days. The change in delayed elastic strains
can be estimated from:
 
−Δ C(t)
Δεd = (σ2 Cd,∞ − εd1 ) 1 − exp Q (6.135)

where εd1 is the delayed elastic creep strain at the beginning of the time increment, Q a
material constant and Cd,∞ the limit deformation at infinity per unit stress. If the deformation
history of a tunnel shell is known, the stresses in the lining can be estimated by rearrangement
of equation 6.132:
 
σ1 − ΔC(t)
ε2 − ε1 + + εd1 1 − exp Q − Δεsh − Δεt
E (t)
σ2 =   (6.136)
1 ΔC(t)
+ ΔC (t) + Cd,∞ 1 − exp− Q
E (t)

For the calculation of shrinkage and temperature strains see sections 6.8 and 6.9.

6.7.4.2 Rate of flow method after Golser et al. (1990)

A further modification to the above method was proposed by Golser et al. (1990). Instead of
using the time-dependent Young’s modulus E (t) they suggested to apply a so called defor-
mation modulus V (t), which also includes the non-linear stress-dependency of the stiffness.
It can be expressed (in GPa) as:
  2 
σ σ
V (σ, t) = E28 bo (t) + b1 (t) + b2 (t) (6.137)
βD (t) βD (t)

with

t
bo (t) = 1.36
22 + 1.78 t

t
b1 (t) = 2.23 · 10−3 (6.138)
0.1 + 0.075 t

t
b2 (t) = 2.23 · 10−6
1810 + 0.1 t

The parameter t is the time in hours. The increase in the uniaxial compressive strength with
time is considered in the function βD (t).

6.7.4.3 Rate of flow method after Aldrian (1991)

In his work, Aldrian (1991) tried to improve the rate of flow method after Golser et al. (1990)
in order to facilitate the calibration of the involved model parameters and to obtain a better
prediction for the creep behaviour of young shotcrete during the first couple of days after
casting. Similar to Golser et al. (1990), a deformation modulus is used for calculating the

206
instantaneous strain, depending on time t and stress level α (= σ/βD ):
-   .
1 t t
V (t, α) = E28 (1 − α) 25 + 3α (6.139)
22.5 25 + 1.2 t 100 + 0.9 t

with α varying from 0 to 1. For describing the creep behaviour Aldrian (1991) introduced
the irreversible viscous deformation of the following form:
 
Δεv = σ2 ΔC exp8α−6 +1 (6.140)

taking into account the non-linear relation between irreversible creep and applied stress via
the stress level α. The increase in irreversible creep strains with time is defined through a
new function C(t) as:
0.2
C(t) = A (t − t1 )0.25 α (6.141)

where t1 (in days) is the age of shotcrete at first loading.

Summarising, the rate of flow method is a special approach for back-calculating stresses
from strain histories and was adopted in several two dimensional analyses of tunnel construc-
tion (i.e. Kienberger (1999)). However, Thomas (2003) states that, according to Rathmair
(1997), it has not been possible to implement this method in a three dimensional finite ele-
ment analysis. Furthermore, a certain criticism exists that the method relies on the principle
of superposition and that there is no allowance for plastic strains (Thomas, 2003).

6.8 Modelling of shrinkage

For the simulation of shrinkage strains Huo & Wong (2006) give an overview of five different
shrinkage models available in the literature based on empirical observations. Their published
research focuses on the early-age behaviour of high performance concrete (HPC) deck slabs
under different curing methods, comparing experimental test results with different model
predictions. The three most basic of the adopted models are presented in the following
sections with the aim of highlighting the main parameters involved in shrinkage behaviour.

6.8.1 Shrinkage after American Concrete Institute (ACI)

The following equation is suggested by ACI (1992) for the estimation of shrinkage strains at
any time:
t
εsh (t) = εsh,∞ (6.142)
B+t
where εsh,∞ is the ultimate shrinkage strain and B a parameter governing the increase in
shrinkage strains with age of drying t (see Fig. 6.38). The above equation was found to be
the most common one adopted by several authors for the prediction of shrinkage in shotcrete
(Kienberger, 1999; Schubert, 1988; Walter, 2003; Aldrian, 1991; Golser, 1999; Galler, 1997).

207
 Figure 6.38: Development of shrinkage strains with time (after ACI, 1992)

6.8.2 Shrinkage after Bazant Model B3

Bažant & Baweja (1995) proposed to calculate the so called “mean shrinkage strain” in a
cross section according to:
εsh (t, to ) = −εsh∞ kh S (t) (6.143)

with the time dependency given as:


t − to
S(t) = tanh (6.144)
τsh

εsh∞ is the time-dependent ultimate shrinkage and to is the age when drying begins. This
model takes further into account the humidity dependence (see Fig. 6.39, the coefficient kh can
be estimated from the relative humidity h), size and shape of the concrete member (via τsh ),
the concrete compressive strength at 28 days, the used cement type, curing conditions and the
water content of the concrete. Meschke (1996) included this shrinkage law in his viscoplastic
material model for a complete description of the time-dependent shotcrete behaviour.

Figure 6.39: Typical shrinkage curves given by Model B3 (from Bažant & Baweja, 1995)

208
6.8.3 Shrinkage after Gardner & Lockman (2001)
In their publication, a design-office procedure for calculating shrinkage is presented in the
following way:
εsh = εshu β (h) β (t) (6.145)

with  0.5
  t − tc
β(h) = 1 − 1.18 h 4
and β(t) = (6.146)
t − tc + 0.15 (V /S)2
where h is the relative humidity expressed as a decimal (not in %), t the age of concrete, tc
the age when drying commenced (end of moist curing) and V /S the volume-surface ratio.
Furthermore, the ultimate shrinkage εshu depends on the cement type (K) and the mean
compressive strength of the concrete at 28 days, fcm28 , and can be estimated from:
 0.5
30
εshu = 1000 K 10−6 (6.147)
fcm28

The model predictions are compared with experimental results for 115 data sets for shrinkage
and it was observed that agreement was within ± 30 %.

6.9 Modelling hydration temperature in concrete structures

The estimation of the increase in temperature with time in a concrete structure due to ce-
ment hydration is a complex process controlled by material technology and external boundary
conditions (i.e. adiabatic or semi-adiabatic curing). According to RILEM (1998) the numer-
ical models for the prediction of the temperature development in cement-based systems like
concrete and shotcrete can be subdivided into two different families of models, which are:

1. Models on a micro level, allowing for physico-chemical processes and mechanisms oc-
curring on the micro- or even nano-level material scale (Garboczi & Bentz, 1992)

2. Models on a macro level, developed for the prediction of temperature fields in real
concrete structures and used in engineering practise (Bosnjak, 2000)

Several analytical methods for the determination of a temperature field were proposed in
the late thirties (i.e. Carlson (1937)) and most of them are based on the Fourier differential
equation for heat conduction, which is given as:
 
∂T ∂2T ∂2T ∂2T
ρc =k + + + Q (t, x, y, z) (6.148)
∂t ∂x2 ∂y 2 ∂z 2

where k is the thermal conductivity, T the temperature, ρ the mass density, c the specific heat
and Q is the rate of internal heat generation per unit volume depending on the coordinates
x, y, z and time t. The correct estimation of the function Q is a central part in temperature

209
calculation of early age-concrete and several definitions can be found in the literature, among
them Reinhardt et al. (1982) and Freiesleben Hansen & Pederson (1977).
Such an approach is mostly suitable for concrete structures of a certain thickness and
mass and usually not performed in engineering practice for the temperature calculation in a
tunnel shotcrete shell, where heat boundary conditions and input parameters are difficult to
establish.

Schubert (1988) includes temperature induced deformations in his constitutive model for
early-age shotcrete and suggests the following mathematical function based on experimental
laboratory tests performed by Rabensteiner (1988):
   
εth = 30 − cos 250 t0.25 + 1 (6.149)

Nevertheless, Schubert (1988) states that the magnitude of these temperature strains εth
might differ considerably from those in a real shotcrete tunnel shell due to different mass and
boundary conditions, but that the basic shape of the development might be realistic. Meschke
(1996) considers temperature induced strains that are uniform throughout the shotcrete body
and independent of the acting stress through the following vector formulation:

εij = α (T − To ) δij (6.150)

where T is the hydration temperature, To a reference temperature, α the coefficient of thermal

expansion and δij the Kronecker-delta. However, no further information about the tempera-
ture development with time is given.

6.10 A review of different constitutive models applied for tun-

nel lining design
6.10.1 Hypothetical Modulus of Elasticity (HME)
The first attempt to realistically model the behaviour of shotcrete for tunnel lining design
was introduced by Pöttler (1985). In his work a Hypothetical Modulus of Elasticity (HME)
was derived treating shotcrete as a pseudo-elastic material. Essentially, the HME represents
a simple reduction of the Young’s modulus E at 28 days by various factors accounting for
the following aspects:

 3D effects

 Creep and shrinkage

 Ageing of shotcrete (i.e. increase of stiffness and strength with time)

 Time-dependent history of loading

210
Through an extensive parametric study published in Pöttler (1990), changing various param-
eters involved in tunnel construction, it was concluded that a value of HME=7 GPa gives the
highest stresses in the shotcrete lining. These stresses are referred to as short-term values
and occur at a distance of one tunnel diameter behind the tunnel face. Although this method
appears to be straightforward, Thomas (2003) states that there exists no established means
of determining the reduction factors for calculating the HME from the Young’s modulus E.
Powell et al. (1997) adopted the HME-approach in their analyses of the platform tunnels for
the Heathrow Express at Terminal 4, estimating a value of the reduced lining stiffness by back
analysis for the required volume loss at ground surface. Their values ranged from 0.75 GPa
to 2 GPa for young shotcrete immediately after excavation and reached the full stiffness of
25 GPa at later stages of construction.

Karakus & Fowell (2003) investigated different excavation types by using a reduced stiff-
ness depending on the distance to the tunnel face as shown in Fig. 6.40.

Figure 6.40: Principle of the HME-approach (from Karakus & Fowell, 2003)

Other influencing parameters were the ground strength, shape and size of the area to be
excavated and the number of excavation stages until ring closure. The values for the HME
ranged from 0.15 GPa to 5 GPa. As a result they suggested a correlation between the HME
and the maximum surface settlement smax of the form:

HM E = a sbmax (6.151)

where a and b are two constants depending on the face advance sequence.

Finally, in the analyses of the CTRL North Downs Tunnel performed by Watson et al.
(1999) a reduced stiffness of 7.5 GPa was adopted for young shotcrete not older than 10 days
and 15 GPa for old shotcrete. The maximum tunnel lining stress was further limited to a
compressive shotcrete strength of 5 MPa and 16.75 MPa as can be seen in Fig. 6.41.

211
Figure 6.41: Stress-strain behaviour of young and old shotcrete (from Watson et al., 1999)

From these observations in the literature it can be concluded that no common agreement
exists on suitable values for the HME that can be used for a realistic and safe design of
shotcrete tunnel linings.

6.10.2 Non-linear beam model by Moussa (1993)

Moussa (1993) developed a constitutive model for simulating the response of shotcrete under
incremental loading with time. The material behaviour includes non-linear hardening and
softening in compression and cracking of shotcrete in tension. In the case of reinforced
concrete, the tension-stiffening effect has been taken into account. Furthermore, shotcrete
was treated as a time-dependent material considering the increase in strength with time. One
important part in this research was the attempt to capture plastic deformations in shotcrete
due to early loading by applying an incremental stress-strain history based model as depicted
in Fig. 6.42.

212
 Figure 6.42: Proposed model for stress-strain history of shotcrete (from Moussa, 1993)

The effect of initial damage due to previous loading on the gain in compressive strength
with time has been considered. Moussa (1993) applied this model for analyzing a 2D tunnel
construction with a single permanent shotcrete shell. Excavation was performed in two stages
(top-heading and invert) with a temporary invert for the intermediate construction stage. The
lining was modelled by a layered beam element consisting of 3 layers, each of them having
different age and straining history. Results showed that after ring closure the axial forces in
the lining where reduced by about 25 % in the crown and 60 % for the invert. The assumed
shotcrete non-linearity and strain history seemed to have an important influence also on the
bending moments in the lining, which were reduced by about 70 % for corners and 50 % for
other parts of the lining. A general increase of deformation in the lining combined with a
change in the deformation mode has been observed. The impact of initial damage during
the increase of shotcrete strength on the structural lining behaviour is not very clear from
the results presented, but generally a lower load capacity could be expected. Finally, it was
concluded that the applied constitutive model had little influence on the behaviour of the
ground adjacent to the tunnel, where surface settlements depend mainly on the tunnel depth.
A slight increase in ground deformations and plastic soil regions has been obtained.

6.10.3 Post-failure modelling of shotcrete by Hafez (1995)

Hafez (1995) implemented in his work the three parameter model developed by Chen &
Chen (1975) into the finite element code FINAL, assuming constant material parameters and
therefore ignoring the time-dependent behaviour of shotcrete. In a 3D analysis of the Lambach
Tunnel in Austria he studied the behaviour of the tunnel shotcrete shell by adopting an elastic

213
and a non-linear elasto-plastic material model. In a comparison of those two approaches it was
noticed that the increase of surface settlements due to the elasto-plastic material response
is about 52 %, whereas the crown displacement showed an increased vertical displacement
of 26 % compared to the simple linear elastic constitutive law. Furthermore, Hafez (1995)
concluded that the elasto-plastic material behaviour has a significant influence on the tunnel
face, as the movement of the unsupported face in the elasto-plastic analysis was about twice
the movement in the elastic approach. Cross sectional forces within the tunnel shell were
generally significantly smaller in the elasto-plastic analysis.

6.10.4 Viscoplastic material model for shotcrete by Meschke (1996)

Meschke (1996) presented a sophisticated model for shotcrete which is formulated on the
basis of multi-surface viscoplasticity with time-adjusted material parameters. The ductile
behaviour of concrete in compression is accounted for by a Drucker-Prager yield surface, in-
cluding non-linear hardening and perfectly plastic material behaviour after peak. The brittle
behaviour of plain concrete in tension is modelled by a Rankine failure criterion considering
isotropic softening after peak in the context of the smeared crack concept (see Fig. 6.43).

Figure 6.43: Loading surfaces of the viscoplastic material model for shotcrete (from Meschke
et al., 1996)

For exponential softening, the involved softening parameters are related to the element
size via the fracture energy (see Section 6.6.1.2) in order to avoid strong mesh-dependency.
In the case of linear softening the softening modulus is related in a simple way to the Young’s
modulus. Furthermore, time-dependent functions for the increase in shotcrete stiffness and
strength and the development of peak strains with time are presented based on experimental
results. The essential phenomena of creep and relaxation of shotcrete have been taken into
account by adopting a viscoplastic rheological model based on the material law by Duvaut
& Lions (1972). A shortcoming of this creep model is the description of creep with only one
single creep parameter and the restriction of viscoplastic strains to an elasto-plastic stress

214
state. Therefore, viscous deformations in the elastic area are not accounted for by the model.
To complete the model the effects of shrinkage and hydration temperature are included in
the model formulation. In Meschke et al. (1996) this proposed constitutive law was used for
a 3D numerical simulation of the excavation of a single-track tunnel in a creep-active soil,
modelling the tunnel lining with solid elements. Construction was performed according to
the principles of NATM in three stages, i.e. top-heading, bench and invert. From the results
it was observed that the stress state within the shotcrete shell is almost biaxial. The load-
carrying behaviour of the shotcrete linings is dominated by the formation of a stress ring in
the circumferential direction of the shotcrete lining. Stress redistribution takes place from
the freshly sprayed concrete parts to the neighbouring parts, which have gained already a
sufficient larger strength and stiffness. Walter (1997) refined this material law for shotcrete
by including additional creep terms based on the rate of flow method (see Section 6.7.4) for
covering as well the elastic area. Further information about the presented constitutive model
can be found as well in Kropik & Mang (1995).

6.10.5 Shotcrete model based on chemoplasticity by Hellmich et al. (1999a)

Lackner et al. (2006) published some research about hybrid analysis for the quantification of
stress states in tunnel shells applying a highly complex material model for shotcrete devel-
oped by Hellmich et al. (1999a) and Sercombe et al. (2000). The thermo-chemo-mechanical
material model is formulated within the framework of thermodynamics of chemically reactive
porous media. Phenomena on the micro-level of the material are interpreted macroscopically
by means of state variables and energetically conjugated thermodynamic forces (Lackner
et al., 2002). The chemical reaction of hydration between cement and water is described by
a scalar variable, referred to as the degree of hydration ξ. It is obtained by relating the mass
of reaction product per unit volume m, to the mass of reaction products at the end of cement
hydration. Therefore it can be written as:

m
ξ= (6.152)
m∞

The evolution of ξ is an underlying intrinsic material function being independent of field

variables and boundary effects. It is of an Arrhenius type law and given as:
 
Ea
ξ̇ = Ã exp − (6.153)
RT

where T is the absolute temperature, Ea is the activation energy and R is the universal
constant for ideal gases with R=8315 J/(mol K). The chemical affinity à is the driving force
of the cement hydration and for shotcrete it depends mainly on ξ. Lackner et al. (2002)
established the following analytical expression:

1 − exp (−bξ)
à (ξ) = a (6.154)
1 + cξ d

215
with a, b, c and d being material constants. Therefore, the mechanical properties of shotcrete
are controlled by the degree of hydration ξ and six material functions are required within the
model formulation (see Fig. 6.44):

a) Normalised chemical affinity Ã

b) Compressive strength fc

c) Young’s modulus E

d) Chemical shrinkage strain εs

e) Charactersitic time for short term creep τw

v
f) Final viscous compliance J∞

Figure 6.44: Intrinsic material functions of shotcrete (from Lackner et al., 2002)

The mechanical behaviour of shotcrete is controlled further by two yield surfaces of a Drucker-
Prager type for compression and a Rankine criterion for the tensile behaviour including
isotropic hardening and softening, as shown in Fig. 6.45.

216
Figure 6.45: Yield surfaces and hardening law for chemoplasticity model for shotcrete (from
Lackner et al., 2006)

Applying this constitutive model in a 2D analysis of a circular NATM tunnel in a creep

active soil, Hellmich et al. (1999b) and Hellmich et al. (2000) observed that the bending
moments in a tunnel shell are strongly influenced by the stiffness of the shell, which depends
both on the chemical hardening characteristics of the shotcrete mixture and on the degree
of plasticising within the shell. Furthermore, they concluded that chemical shrinkage always
leads to a reduction of bending moments by the formation of cracking. The high creep
capacity of young shotcrete tends to increase the horizontal inward movements of the tunnel
section as a consequence of the redistribution of the stresses in the soil after excavation.
Generally, creep is the dominant factor in areas where high stresses are generated and can
lead to some changes in the overall structural behaviour of the tunnel lining (change of signs of
bending moments). According to their results, temperature also appears to be an important
factor for the behaviour of the shell. A higher temperature leads to higher bending moments
due to the faster hardening and stiffening of the shotcrete. Finally, tangential forces within
the tunnel lining seem to be practically independent of the material law used for shotcrete
and depend exclusively on the stress distribution that has taken place in the soil before the
hardening process of the shotcrete starts.

6.10.6 Assessment of constitutive models for shotcrete by Thomas (2003)

In his research, Thomas (2003) investigated the behaviour of shotcrete for tunnelling by
adopting a large variety of different constitutive laws found in the literature. The key as-
pects of his analyses were the ageing of the material properties, the non-linear stress-strain
behaviour, creep and shrinkage. At first instance, a large-scale laboratory load test on a ring
of sprayed concrete was simulated numerically, moving then on to the numerical modelling
of 3D tunnel construction. The set of analyses involved the following material behaviour for
shotcrete, implemented into the finite difference program FLAC:

 Linear elastic with constant and age-dependent stiffness

 Non-linear elastic model after Kotsovos & Newman (1978) including time-dependent
material parameters

217
  Hypothetical Modulus of Elasticity (HME)

 Simple plasticity models (Drucker-Prager, Mohr-Coulomb) adopting perfectly plastic

post-peak behaviour, strain-hardening and time-dependency

 Various creep models based on Kelvin model assuming different viscosity functions

 Shrinkage model

From the obtained results it was concluded that the choice of the constitutive model for
shotcrete affects the stress distribution within the tunnel lining. Lower lining stiffness results
in higher deformation coupled with lower stresses. Ageing of the elastic modulus, non-linearity
of the stress strain curve and creep all lead to higher strain and deformation levels. The
relative stiffness of the ground and the lining appears to be a governing factor for the loads
in the lining and its movements. Hence, the importance of the soil-structure interaction was
highlighted. Creep models are often said to represent “softer” shotcrete models, although
the input data for the involved creep functions are usually quite scattered. Essentially,
creep leads to a reduction in lining stresses and this effect is generally accepted as being
beneficial for lining design. However, this depends heavily on the ability of the ground to
sustain the remainder of the load. Bending moments were more strongly influenced than
hoop forces by the constitutive model adopted, depending also on the geometrical shape of
the tunnel cross section. Some details of the constitutive model may also be significant, for
example the assumed shape of the failure surface in the deviatoric plane (Drucker-Prager or
Mohr-Coulomb). Finally, it was stated that in a full 3D analysis the effect of an advanced
constitutive model for sprayed concrete was noticed mainly within the first few metres of the
lining behind the tunnel face.

6.11 Summary
The topic of the current chapter was to give an introduction into the modelling of a quasi-
brittle material such as concrete or shotcrete, covering the mechanical behaviour both in
compression and tension. Furthermore, an overview of different approaches available in the
literature to simulate the time-dependent phenomena of creep, shrinkage and hydration tem-
perature that occur in concrete structures were presented. The following conclusions can be
drawn:

 A large number of empirical models based on experimental data exist for modelling
concrete, taking into account the non-linear material behaviour before peak. However,
one big drawback is that these models focus mainly on uniaxial stress conditions in
compression and tension and can hardly be used in a numerical analysis. Nonetheless,
simple concrete models have been developed within the framework of elasto-plasticity,
that are capable of reproducing the main characteristics of the material behaviour in
multiaxial loading conditions.

218
 The Chen & Chen (1975) concrete model represents a more advanced constitutive
model for hardened concrete based on elasto-plasticity, where the material behaviour
in compression and tension is governed by two yield surfaces. On the onset of plastic
deformation these two surfaces move together simultaneously in the principal stress
space and isotropic strain-hardening controls the material behaviour before reaching
peak.

 Cracking of concrete in tension is a relatively difficult phenomenon to model, since in

reality it is a highly discrete and localised problem that occurs in a concrete structure.
For simulating cracking in a numerical analysis two different approaches exist, i e. the
discrete and the smeared crack model. However, when treating cracking of concrete
within the framework of continuum mechanics, one has to face severe problems that are
usually associated with strain-softening materials. A simple approach based on fracture
energy was presented to avoid spurious mesh dependency on the obtained results.

 For considering steel reinforcement of concrete or shotcrete three different approaches

exist in the literature (i.e. discrete, smeared and embedded reinforcement models).
It was highlighted that due to their complexity, reinforcement is hardly considered
in the analysis of tunnel construction. However, sometimes it is convenient to adopt
the stress-strain behaviour of concrete in tension by introducing the so called “tension-
stiffening effect”. This interaction of concrete and steel leads to a slightly softer material
response of concrete in the tensile post peak regime. When analysing steel-fibre rein-
forced concrete structures a fracture energy based approach might be a possible solution
for simulating the increase in ductility of concrete behaviour in tension.

 No well accepted framework exists in the literature for the modelling of concrete creep
and in particular for young shotcrete at early ages, which is a very creep active medium.
Apart from some basic mathematical functions, in most of the cases a combination of
rheological units serves for the calculation of creep deformation in concrete structures
(Kelvin-, Maxwell- and Burger-model). By establishing complex viscosity and stiffness
functions it is possible to fit model predictions to uniaxial compressive test data. How-
ever, very limited information is available for the extension of creep to 3D or how to
deal with creep of concrete in tensile stress conditions.

 Three simple models for predicting shrinkage deformation of concrete have been pre-
sented. They all have in common that they are based mainly on experimental data and
take into account factors such as relative humidity, concrete strength in compression,
geometry of the structure to be investigated and concrete composition.

 The modelling of the increase in temperature due to cement hydration in hardening

concrete or shotcrete is a complex process, mostly performed for the analysis of rel-
atively thick concrete structures. Little information is available in the literature on

219
how to incorporate these temperature effects in a constitutive model for the practical
analysis of shotcrete tunnel shells.

220
Chapter 7

A constitutive Model for Shotcrete

7.1 Introduction
In this chapter a non-linear constitutive model for the time-dependent behaviour of shotcrete
based on elasto-plasticity is presented, which was developed during this research project. The
well known Chen & Chen (1975) concrete model, described in Chapter 6, was selected as a
starting point because of its relative simplicity and capability to simulate the main character-
istics of the mechanical behaviour of hardened concrete. However, it was decided to modify
this constitutive model in such a way that a better reproduction of the complex mechanical
behaviour of sprayed concrete was possible. Therefore it was necessary to implement two
yield surfaces that move independently from each other in stress space, being governed by
two different hardening parameters - one for compression and one for tension. Cracking of
shotcrete is considered within the smeared-crack concept. Furthermore, the main material
parameters of the model are changing with time according to test data for shotcrete that
can be found in the literature. As a result it will be shown that the developed constitutive
model is capable of simulating the transition from a ductile and plastic material behaviour
at early ages to a quasi-brittle material response after curing at 28 days. Other mechanical
aspects related to time, such as creep, shrinkage or effects due to cement hydration tempera-
ture, are also considered within the current model formulation. Anisotropy effects that might
be observed due to the spraying process are not taken into account and therefore material
behaviour is assumed to be isotropic. After a robust numerical implementation into the Im-
perial College Finite Element Program (ICFEP, Potts & Zdravković (1999)) the outcome was
a sophisticated constitutive model for sprayed concrete that can be applied for the analysis
of complex boundary value problems in tunnelling. Finally, some model predictions for single
element runs presented at the end of this chapter show the model capabilities but also its
limitations.

221
7.2 Formulation of the constitutive model
As mentioned in the introduction of this chapter, the original Chen & Chen (1975) concrete
model contains some shortcomings for the correct reproduction of shotcrete behaviour that
are discussed here briefly.

Firstly, uniaxial tests of hardened concrete indicate that the stress-strain curves for com-
pression and tension are different in shape, in particular for the post-peak behaviour. The
original Chen & Chen (1975) concrete model does not account for this fact and the non-linear
hardening curves are assumed to be identical. Consequently, the two yield surfaces move to-
gether simultaneously in stress space, being governed by only one single hardening parameter
(see equation 6.77). In a first step it was tried to improve this drawback by implementing two
different hardening and softening rules for compression and tension controlled by two differ-
ent hardening parameters. Unfortunately, the problem that arose during some test analyses
was the following: the two yield surfaces given in equations 6.75 and 6.76 are supposed to
√
be connected to each other at their intersection point in the J-p or I1 - J2 space at any
time. However, their mathematical nature is such that some restrictions exist regarding the
minimum and maximum size of both yield surfaces, which is controlled by the strength val-
ues used for the description of the hardening and softening rules, otherwise convexity of the
yield surfaces is lost. Therefore it would not be possible to entirely fit these hardening and
softening curves to experimental data.

Secondly, during the hardening process of the cement paste the material parameters of
shotcrete are a function of time and hence are not constant. Furthermore, shotcrete behaviour
includes also creep, shrinkage and thermal deformation due to the cement hydration. The
original Chen & Chen (1975) concrete model was formulated for hardened concrete at 28
days and does not contain any time-dependency.

Due to these major shortcomings it was decided to incorporate some important modifi-
cations to the original model formulation. In the following sections of this chapter the new
structure of the modified Chen & Chen (1975) model for the behaviour of shotcrete is pre-
sented and discussed in detail. Throughout the complete model description a tension positive
sign convention is adopted in the model development and the yield functions for both com-
pression and tension are formulated in terms of the geotechnical stress invariants p, J and θ,
previously introduced in Chapter 1.

7.2.1 Structure of the model

In the present constitutive model for shotcrete, the classical theory of elasto-plasticity for
hardening and softening materials along with its algorithmic formulations is extended to
account for the effects of ageing, creep, shrinkage and thermal deformation. The increase in
the strength properties of shotcrete implies that the adopted yield surfaces are considered to

222
be time-dependent and the yield condition for a single surface can be expressed as:

F ({σ} , {k} , t) = 0 (7.1)

where {σ} is the current stress state, {k} the vector of state (or hardening) parameters and t
represents the time. Applying the principle of superposition the total strains are made up of
elastic strains εel , conventional plastic strains εp arising from the yield and plastic potential
surfaces, creep strains εcr , shrinkage strains εsh and thermal strains εth due to the generated
heat during cement hydration. It can be written that:

ε = εel + εp + εc (7.2)

with the irrecoverable strains εc given as:

εc = εcr + εsh + εth (7.3)

Each of those components will be described in detail in the following sections.

7.2.2 Yield Function and Plastic Potential in compression

The yield surface used to enclose the elastic region and to describe the onset of plastic
deformation of the shotcrete in multi-axial compression is based on the results of biaxial tests
performed by Kupfer & Gerstle (1973) and was originally formulated by Chen & Chen (1975).
In the current development, this yield surface is applied to describe shotcrete behaviour in
compression and is given by the following equation:

Fc = J 2 + Ac p − τc2 = 0 (7.4)

where
e2 − 1
Ac = fc (7.5)
2e − 1
and
2e − e2 2
τc2 = f (7.6)
3 (2e − 1) c
The parameter e is a dimensionless biaxial strength parameter and accounts for the fact that
the strength of concrete under biaxial compression (σ1 = σ2 ) is up to 25 % higher than under
uniaxial conditions. Meschke (1996) proposed a value of e = 1.16. For geometric reasons
regarding the elliptical shape of the adopted yield surface in the biaxial stress space, this
parameter e is limited to the following range:
√
1+ 3
1 < e < emax = (7.7)
2

223
Fig. 7.1 shows the two yield surfaces for both compression and tension, where the compression
surface takes an elliptical shape in the biaxial stress space (on the left) and a parabolic one
in the J-p space (on the right). Experimental data justify that these shapes are appropriate
to describe the behaviour of concrete reasonably well (Chen & Chen, 1975).

Figure 7.1: Compression and tension yield surfaces in a) the biaxial stress space and b) the
J-p space

Since equation 7.4 is independent of the Lode angle θ, the compression surface plots as
a circle in the deviatoric plane. Furthermore, the size of the yield surface is governed by
the parameter fc , which represents an “equivalent uniaxial compressive strength” and is a
function of the plastic strains and time. Hence, the surface expands and contracts isotropically
according to plastic strain hardening and softening rules that will be described in detail in
Section 7.3.1. Furthermore, the surface also moves due to the increase in strength with time,
as will be presented later in Section 7.4.1. In the absence of detailed experimental data for
a realistic characterisation of the volumetric behaviour of shotcrete, an associated flow rule
is adopted and therefore the yield surface Fc equals the plastic potential Pc , so it can be
written:
Fc (p, J) = Pc (p, J) (7.8)

7.2.3 Yield Function and Plastic Potential in tension

For the numerical description of the quasi-brittle material behaviour of shotcrete in tension, a
modified Rankine criterion after Thomée (2005) is used, see Fig. 7.1. Contrary to the original
Rankine failure criterion, the adopted yield surface includes hyperbolic rounding near the
apex in order to avoid numerical problems associated with the singularity present in the

224
original criterion. The tension yield surface can be written as:
  n 1
2J 2π n
Ft = p + √ sin θ + + (aft )n
− ft = 0 (7.9)
3 3

Fig. 7.2 represents a close up view of the apex zone of the yield function in the biaxial stress
space (on the left) and the J-p space (on the right), indicating the difference between the
original Rankine criterion and the modified tension yield surface adopted in the current model
formulation.

Figure 7.2: Apex rounding for tension surface in a) the biaxial stress space and b) the J-p
space

The parameter a is the apex tolerance and describes the distance between the two in-
tersection points of the original and the modified Rankine yield surface with the hydrostatic
axes, whereas the rounding of the surface is controlled by the parameter n. Furthermore,
the above presented yield function depends on the Lode angle θ and plots as a triangle in
the deviatoric plane. The corners of this triangle are treated in a similar manner to the
other yield surfaces implemented in ICFEP with corners in the deviatoric plane. A rounding
parameter CORT OL is introduced, which represents a corner tolerance for θ = +30 (i.e.
triaxial extension). Within this corner tolerance, the partial derivatives of the yield function
and the plastic potential are rounded in a linear way. Similar to the compression surface,
the parameter ft represents an “equivalent uniaxial tensile strength” and varies with plas-
tic strains and time, resulting in isotropic expansion and contraction of the surface in the
general stress space. The mathematical description of these hardening and softening rules is
presented in Section 7.3.2.

In order to avoid a stress path going beyond the yield surface during an analysis, which
would result in a forbidden stress state, another tolerance parameter had to be introduced
for the apex region. The parameter T IP T OL denotes a vertical limit, which is linked to the

225
current intersection point of the modified tension surface with the hydrostatic axes, as can
be seen in Fig. 7.3.

Figure 7.3: Tip tolerance for apex region of tension surface

For the tensile peak strength it can be written:

x = ftp (1 − a) T IP T OL (7.10)

Therefore, this tip limit is moving during plastic strain hardening/softening and the increase
in tensile strength. If during the sub-stepping algorithm of an analysis the mean stress p
appears to go past this limit, the stress point is forced to stick to the tip limit by setting any
further changes in mean stress Δp to zero.

As in the case of the compression surface, associated conditions are assumed and it can
be written:
Ft (p, J, θ) = Pt (p, J, θ) (7.11)

7.3 Hardening and softening rules

The basis for the non-linear and independent evolution of the two implemented yield surfaces
in the general stress space in the course of plastic deformation are uniaxial stress-strain curves
for concrete in compression and tension. Since the strain-hardening concept is applied, the
chosen hardening parameters are a function of the plastic strain history. If the compression
surface is the active yield surface, hardening and softening is controlled by the minor principal
plastic strain εp3 . In the case of the tension surface being the active surface, the motion
of the yield surface is governed by the major principal plastic strain εp1 . However, as the
current model also includes the time-dependent behaviour of shotcrete, the choice of a proper
hardening parameters used in the model formulation is not straightforward. Fig. 7.4 illustrates
the problem:

226
 Figure 7.4: Strain hardening concept for time-dependent matrials

From published data in the literature it appears that the material behaviour of sprayed
concrete becomes more and more brittle with curing time, resulting in a decrease in the
strain at peak strength. Assume a shotcrete sample is loaded up to a certain compressive
plastic strain (less than that required to reach peak strength, see Fig. 7.4) at an early age
t1 . After loading, a certain amount of time Δt elapses (without any further straining) and
the loading is then continued at time t2 . During the elapsed time Δt the compressive peak
strength increases but the compressive strain associated with this new peak strength would
have reduced such that, for the further stress path, it is not clear where the stress point would
touch the yield surface again beyond peak. To overcome this dilemma it was decided to adopt
normalised stress-strain curves for both compression and tension, where the normalisation
is performed in terms of the peak strengths and the plastic peak strains at time t. The
following two sections describe the mathematical functions of the hardening and softening
rules adopted in the current model.

7.3.1 Compression
As mentioned above, the hardening and softening rule in compression follows a uniaxial
stress-strain curve where the normalised equivalent uniaxial compressive strength is given as:

fc
fc,n = (7.12)
fcp (t)

Similarly, the minor principal plastic strain εp3 is normalised by the peak compressive plastic
strain at time t, εpcp (t), such that the compressive hardening parameter Hc can be written
as:
εp3
Hc = p (7.13)
εcp (t)

227
 Figure 7.5: Normalised stress-strain curve for shotcrete in compression

Fig. 7.5 shows the normalised stress-strain curve that is applied in the current model
formulation, consisting of 4 different zones for hardening and softening:

Zone I: 0 ≤ Hc ≤ 1
As soon as the current stress state touches the yield surface the behaviour becomes elasto-
plastic and the normalised stress-strain curve follows a parabolic hardening, starting from
the normalised compressive yield stress fcy,n up to peak at unity. The following equation is
adopted:
 
fc,n = fcy,n + (1 − fcy,n ) 2Hc − Hc2 (7.14)

Zone II: 1 < Hc ≤ Hcf

If any further straining occurs after peak, a linear softening branch is applied until failure of
the shotcrete is reached at the normalised compressive failure strength fcf,n . At this point, in
a real compression test the shotcrete is assumed to be completely crushed and stresses would
reduce to zero. The equation for the adopted softening part is given as:

fcf,n − 1
fc,n = 1 + (Hc − 1) (7.15)
Hcf − 1

Zone III: Hcf < Hc ≤ Hcu

For numerical purposes a normalised ultimate compressive strength fcu,n is introduced to
describe the linear reduction in stresses after failure given by:

fcu,n − fcf,n
fc,n = fcf,n + (Hc − Hcf ) (7.16)
Hcu − Hcf

Zone IV: Hc > Hcu

In this zone no further changes in stress occur with further straining and therefore the nor-
malised compressive stress remains constant.

fc,n = fcu,n (7.17)

228
In the equations 7.14 to 7.17 five dimensionless material parameters are needed for the me-
chanical description of the hardening and softening behaviour in compression, which are based
on the values of hardened shotcrete at the age of 28 days and are assumed to be constant
with time. They are the following:

Normalised compressive yield stress:

fcy,28
fcy,n = (7.18)
fcp,28

Normalised compressive failure strength:

fcf,28
fcf,n = (7.19)
fcp,28

Normalised compressive ultimate strength:

fcu,28
fcu,n = (7.20)
fcp,28

Normalised compressive failure strain:

εpcf,28
Hcf = (7.21)
εpcp,28

Normalised compressive ultimate strain:

εpcu,28
Hcu = (7.22)
εpcp,28

The material parameters for the implemented stress-strain curve depend on the type of
shotcrete used and can be estimated from experimental uniaxial tests or published data.
It should be emphasized that a plastic strain hardening concept is applied and the adopted
stress-strain curve is formulated for plastic strains. Therefore, the total peak strain at 28
days, εcp,28 , needs to be reduced by the elastic component in order to obtain the plastic peak
strain εpcp,28 , as written below:
fcp,28
εpcp,28 = εcp,28 − (7.23)
E28
where E28 is the Young’s modulus for hardened shotcrete at 28 days.

For a correct description of the softening part regarding mesh dependency it is recom-
mended to estimate the normalised compressive failure strain Hcf from the compressive
fracture energy Gc , as highlighted in the previous Chapter 6. Fig. 7.6 illustrates the basic
principle adopted for the current model:

229
 Figure 7.6: Fracture energy concept for shotcrete in compression

In a uniaxial compression test for hardened shotcrete at 28 days, it is assumed that the
complete compressive fracture energy Gc is the area below the stress-displacement curve
from peak until failure, where the shotcrete is believed to be completely crushed. With
the help of the equivalent length leq of the finite element it is possible to transform this
stress-displacement relation into a typical plastic stress-strain curve and it can be written as:

Gc fcp,28 + fcf,28 ∗
gc = = ε (7.24)
leq 2

Rearrangement of this equation leads to:

2 Gc
ε∗ = (7.25)
leq (fcp,28 + fcf,28 )

The plastic failure strain at 28 days as a softening parameter is therefore dependent on the
element size and can be obtained as:

εpcf,28 = εpcp,28 + ε∗ (7.26)

Finally, the normalised plastic failure strain Hcf is then calculated according to equation
7.21. Since the ultimate plastic compressive strain εpcu,28 is of purely numerical purpose, it
is convenient to choose the magnitude of this limit strain to be close to the plastic failure
strain εpcf,28 . However, in order to avoid the so called “snap-back behaviour” (Thomée, 2005)
in the total stress-strain curve which is caused by an excessive softening rate (see Fig. 7.7),
it is necessary that the following condition is fulfilled:

εpcu,28 ≥ εpcf,28 + ε̄ (7.27)

230
with
fcf,28 − fcu,28
ε̄ = (7.28)
E28

Figure 7.7: a) Snap back behaviour for total stress-strain curve, b) Limited ultimate plastic
compressive strain

A certain limitation exists for the choice of the ultimate compressive strength fcu,28 due
to numerical reasons, as illustrated for the biaxial stress space in Fig. 7.8.

Figure 7.8: Critical value of the ultimate compressive strength for hardened concrete at 28
days

Although the basic idea for introducing this ultimate strength parameter is to reduce
concrete stresses to a value close to zero after crushing failure, a critical value fcu,crit is
established in oder to avoid the contraction of the compressive yield surface below the tension

231
yield surface. Otherwise this would mean that even the purely tensile material behaviour
would be governed by the compression yield surface, which is an unacceptable scenario (see
dashed ellipse in Fig. 7.8). Therefore, the following condition derived from the compressive
yield surface given in equation 7.4 for J = 0 and depending on the maximum tensile stress,
which is the tensile peak strength at 28 days, ftp,28 , and the biaxial strength parameter e has
to be fulfilled:
e2 − 1
fcu,28 ≥ fcu,crit = 3 ftp,28 (7.29)
2e − e2
The same condition applies also to the compressive yield stress and it can be written in a
similar manner:
e2 − 1
fcy,28 ≥ fcy,crit = 3 ftp,28 (7.30)
2e − e2

7.3.2 Tension
For the tension surface, a similar approach as for the compression surface is adopted and the
normalised equivalent uniaxial tensile stress results in:

ft
ft,n = (7.31)
ftp (t)

After normalizing the major principal plastic strain εp1 the tensile hardening parameter is
given as:
εp1
Ht = p (7.32)
εtp (t)

The uniaxial tensile plastic peak (or cracking) strain εptp (t) is assumed to develop with time
in the same way as the compressive plastic peak strain and is given as:

εptp (t) = δ εpcp (t) (7.33)

where δ represents a constant proportion between the plastic tensile and compressive peak
strains for hardened shotcrete at 28 days:

εptp,28
δ= (7.34)
εpcp,28

Similar to the compression hardening and softening, the complete normalised stress-strain
curve in tension is divided into 3 different zones.

Zone I 0 ≤ Ht ≤ 1
Although concrete tests under tension show almost linear elastic behaviour before peak, a
yield stress of about 90% of the tensile peak strength can be identified (Feenstra, 1993).
Therefore a small parabolic hardening zone is introduced given by the following equation:
 
ft,n = fty,n + (1 − fty,n ) 2Ht − Ht2 (7.35)

232
One normalised material parameter is used to describe mathematically the onset of yield-
ing, based on the ratio between the yield stress and the tensile peak strength for hardened
shotcrete at 28 days. The normalised tensile yield stress can be written as:

fty,28
fty,n = (7.36)
ftp,28

For the post-peak behaviour of shotcrete two different options exist within the smeared crack
concept of the current constitutive model, which will be described in detail in the following
sections.

7.3.2.1 Option A - Linear softening

Zone II and III of option A are characterised by linear softening after peak and followed by
a constant stress in the ultimate stress range, as can be seen in Fig. 7.9.

Figure 7.9: Normalised hardening and softening curve in tension for option A

Zone II: 1 < Ht ≤ Htu

After initiation of cracking at the peak strength, tensile strength reduce with further straining
in a linear way until an ultimate tensile strength is reached. The equation for the linear
softening branch in tension can be written as:

ftu,n − 1
ft,n = 1 + (Ht − 1) (7.37)
Htu − 1

Zone III: Ht > Htu

Once the ultimate tensile stress is reached in Zone III, there is no further change in stresses
with continuing straining and therefore the normalised tensile strength remains constant.

ft,n = ftu,n (7.38)

Two normalised material parameters control this linear post peak behaviour, which are:

233
Normalised ultimate tensile strength:

ftu,28
ftu,n = (7.39)
ftp,28

Normalised tensile ultimate strain:

εptu,28
Htu = (7.40)
εptp,28
In order to avoid the spurious mesh dependency that is well known to occur for materials that
undergo tensile strain softening, similar to the compression behaviour, a fracture energy based
approach is proposed to estimate the softening parameter involved in this model. Fig. 7.10
shows the basic principle for a uniaxial tension test for hardened shotcrete at 28 days.

Figure 7.10: Fracture energy concept for shotcrete in tension with linear softening

The total fracture energy in tension Gf needed for complete crack opening is represented
as the area below the curve from peak until zero stress. Mesh dependency is now overcome by
regularising the tensile stress-crack-opening relation with the equivalent length of the finite
element leq . This procedure results in the following expression:
* +
Gf ftp,28 ε̄t − εptp,28
gf = = (7.41)
leq 2

where ε̄t is the plastic tensile strain at zero stress and can be obtained by rearrangement as:

ε̄t = εptp,28 + ε∗t (7.42)

234
The distance between the peak strain εptp,28 and ε̄t is given as:

2Gf
ε∗t = (7.43)
leq ftp,28

Consequently, the plastic ultimate strain for hardened shotcrete in tension can be estimated
as:
2Gf (ftp,28 − ftu,28 )
εptu,28 = εptp,28 + 2 (7.44)
leq ftp,28
Similar to the condition in compression, the plastic stress-strain curve must not be too steep
otherwise the total stress-strain relation would result in unstable material behaviour. For the
tensile linear softening rate the following condition has to be fulfilled:

εptu,28 ≥ εptp,28 + ε̂t (7.45)

with
ftp,28 − ftu,28
ε̂t = (7.46)
E28

7.3.2.2 Option B - Exponential softening

In contrast to the softening behaviour in the previous option, in option B the gradual decrease
in strength with crack propagation in Zone II is described by an exponential softening law
that is followed again by a constant strength regime in Zone III, as indicated in Fig. 7.11.

Figure 7.11: Normalised hardening and softening curve in tension for option B

Zone II: 1 < Ht ≤ Htu

The mathematical expression for the reduction in tensile strength with plastic straining in
this zone is given by the following equation:

ft,n = exp [−ψ (Ht − 1)] (7.47)

235
where ψ is the normalised softening parameter that governs the decay in tensile strength.
This exponential reduction in tensile strength continues until an ultimate tensile strength is
reached and hence the normalised tensile ultimate strain can be calculated from:
 
ftu,28
ln
ftp,28
Htu = 1 − (7.48)
ψ

Zone III: Ht > Htu

As in option A, Zone III of option B shows a material behaviour with constant strength
controlled by the ultimate tensile strength that can be chosen to be close to zero.

ft,n = ftu,n (7.49)

The normalised ultimate tensile strength ftu,n as a material parameter can be obtained from
equation 7.39.

Regarding the choice of the proper softening parameter ψ it is recommended to estimate

this parameter again from the fracture energy of hardened shotcrete at 28 days, Gf . By
integrating the plastic stress-strain curve from peak strain to infinity the area below the
curve, which represents the fracture energy, is obtained, as can be seen in Fig. 7.12.

Figure 7.12: Fracture energy concept for shotcrete in tension with exponential softening

It can be written:
)  +
Gf +∞
1* p p
gf = = ftp,28 exp − εt − εtp,28 dεpt (7.50)
leq p
εtp,28 κ

236
where κ is a softening parameter depending on the fracture energy Gf and the equivalent
length of the finite element leq . Solving the integral and rearranging results in:

Gf
κ= (7.51)
leq ftp,28

In the present constitutive model a normalised hardening and softening rule is adopted and
therefore the normalised softening parameter from equation 7.47 can be written as:

1 p leq ftp,28 εptp,28

ψ= εtp,28 = (7.52)
κ Gf

Snap back behaviour of the total stress strain curve can be avoided by the following condition
derived from the derivative of the softening curve in Zone II taking into account the elastic
deformation:
εptp,28 E28
ψ≤ (7.53)
ftp,28

7.4 Time-dependent material behaviour

As mentioned in the introduction to this chapter, the current constitutive model for shotcrete
accounts for time-dependent material behaviour. This means that the main material param-
eters used in the formulation of the elasto-plastic model are developing gradually with time
during cement hydration. Due to the increase in the strength parameters both in compres-
sion and tension the adopted yield surfaces are not only moving in stress space with plastic
strain hardening or softening, but also expand simultaneously with time. Although cement
hydration is a relatively long process, it is assumed that in an analysis, for hardened shotcrete
after 28 days no further change in material properties takes place. Therefore:

 Shotcrete age ≤ 28 days → Changing material properties

 Shotcrete age > 28 days → Constant material properties

The inclusion of time-dependent material behaviour in a numerical analysis requires the

specification of the initial age of the shotcrete at installation to . This input parameter may
be chosen by the user but a minimum age of to = 1 h has been adopted in this thesis. Finally,
constitutive laws that describe the evolution of creep, shrinkage and temperature deformation
were developed and will be discussed in detail in the following sections.

7.4.1 Increase in stiffness and strength with time

i) Young’s modulus
For the development of the stiffness with time a well established equation from CEB-FIP

237
Model Code (1990) has been adopted within the current model formulation. It is given by:
-    .0.5
t28
E(t) = E28 exp s 1 − (7.54)
t

where E28 is the Young’s modulus for hardened shotcrete at 28 days and t28 is the time
at 28 days in the corresponding time units used in the analysis. The cement parameter s
usually takes a value smaller than 1 and controls how fast the stiffness increases with time,
as indicated in Fig. 7.13.

Figure 7.13: Increase in stiffness with time for various values of cement parameter s

Furthermore, a smaller value of s results in a faster increase in stiffness with time

(s1 < s2 < s3 < s4 < 1). This curve predicts generally a high initial increase in stiffness for
the first hours after shotcrete installation, which becomes less pronounced with increasing
time until it reaches the full stiffness of hardened shotcrete E28 at t28 . In the case of absence
of any experimental data for shotcrete the following values for s are provided by the CEB-FIP
Model Code (1990):

 s = 0.2 for rapid hardening high strength cements RS

 s = 0.25 for normal and rapid hardening cements N and R

 s = 0.38 for slowly hardening cements SL

The incremental stiffness formulation needed for the numerical implementation is given as:
* *  ++0.5
d E(t) E t
28 28 s exp s 1 − t28
t
Ė(t) =
dt
= √ 3 (7.55)
t28 t 2

The Poisson’s ratio μ is taken as a constant value and any possible time-dependency is
therefore ignored.

238
ii) Compressive strength
The hardening and softening curve for compression presented in Section 7.3.1 requires for
its normalisation knowledge of the development of the compressive peak strength with time.
Similar to the increase in stiffness, an expression from the CEB-FIP Model Code (1990) was
chosen, which yields: -   .
t28
fcp(t) = fcp,28 exp s 1 − (7.56)
t

where fcp,28 is the compressive peak strength at 28 days and s a cement parameter, which is of
the same nature as the one introduced for the increase in stiffness with time. The development
of the compressive yield, failure and ultimate strength is linked to the compressive peak
strength via the constant ratios given in equations 7.18 to 7.20. Therefore, these strength
values develop in the same manner as the compressive peak strength, as can be seen in
Fig. 7.14.

Figure 7.14: Qualitative increase in compressive yield, peak, failure and ultimate strength
with time

iii) Tensile strength

A similar equation as presented above for the development of the compressive peak strength
with time applies also to the increase in the tensile peak strength with time. It can be written
as: -   .
t28
ftp (t) = ftp,28 exp s 1 − (7.57)
t

where s is the same cement parameter already adopted for the compressive strength param-
eters.

239
7.4.2 Shotcrete deformability at early ages
As explained in Chapter 5, young shotcrete appears to have the ability to withstand higher
deformations at early age loading than hardened shotcrete. This fact is taken into account
in the current constitutive model by making the plastic compressive peak strains vary with
time according to the following equation:
  B
t
εpcp (t) = A ln +C (7.58)
t1

where t1 is the time of 1 hour in the corresponding time unit. The constants A, B and C
should be chosen to fit the experimental test data. For the analysis presented in this thesis
this was achieved by using the plastic uniaxial compressive peak strain εpcp,t1 at time t1 = 1 h,
the plastic uniaxial compressive peak strain εpcp,t2 at an intermediate time t2 and the plastic
uniaxial compressive peak strain for hardened shotcrete at 28 days, εpcp,28 . A, B and C are
then obtained as:
 
εpcp,t2 − εpcp,t1
ln
εpcp,28 − εpcp,t εpcp,28 − εpcp,t1
A =   B1 and B= ⎛  ⎞ and C = εpcp,t1 (7.59)
t28 t2
ln ⎜ ln t1 ⎟
t1 ln ⎜
⎝  t28  ⎠
⎟
ln
t1

Fig. 7.15 illustrates equation 7.58 and the involved material parameters. The reason for using
the intermediate plastic uniaxial compressive peak strain at time t2 is that it allows the control
of the rate of decay in plastic uniaxial compressive peak strains with time. A convenient choice
for the intermediate time t2 could be for instance 24 h after shotcrete installation.

Figure 7.15: Development of plastic uniaxial compressive peak strains with time

Set A would represent a shotcrete having a very brittle type of material behaviour, where

240
the compressive peak strains reduce quickly within the first hours after casting to an almost
constant level, whereas, set C indicates a softer material response. It should be noted that
the above curve (equation 7.58) describes the development of the plastic uniaxial compressive
peak strains and for model calibration the time-dependent elastic strain component has to
be taken into account. In this respect it is noted that that:

fcp(t)
εpcp (t) = εcp (t) − (7.60)
E(t)

As mentioned in Section 7.3.2, the plastic uniaxial tensile peak strains εptp (t) are assumed to
develop with time in the same manner as the plastic uniaxial compressive peak strains, being
related to each other via the constant ratio δ (see equation 7.33).

7.4.3 Creep and relaxation

Chapter 5 highlighted the importance of the phenomena of creep and relaxation for young
shotcrete and its influence in tunnelling. In the current constitutive model a creep law is
proposed that is based on the rheological body illustrated in Fig. 7.16.

Figure 7.16: Time-dependent rheological body for creep

In the uniaxial case the deformation due to creep is described by a single dashpot, which
is characterised by a time-dependent viscosity η(t). The mathematical expression for the
irrecoverable creep strains (either compressive or tensile) based on Neville (1970) is:

t − to
εcr = σ (7.61)
η(t)

where to denotes the starting age of the shotcrete and σ is the applied compressive or tensile
stress. Fig. 7.17 shows the development of creep strains due to loading at time to . The in-
stantaneous deformation at loading consists of elastic and plastic strain components, followed
by the increasing creep strains with time. Equation 7.61 implies further that the loading of
shotcrete begins immediately after its installation at time to and creep strains start to occur,
which can be considered as realistic in the application for tunnel construction.

241
 Figure 7.17: Development of creep strains with time upon loading at time to

The time-dependent viscosity function proposed in the current model can be written as:
 n
t
η(t) = ηo (7.62)
t1

The parameter ηo is the starting viscosity at an age of t = 1 h and n is a viscosity exponent

that governs the increase in viscosity with time. As in equation 7.58, t1 refers to a time
value of 1 h in the corresponding time units used in the analysis. Fig. 7.18 illustrates the
development of the shotcrete viscosity with time for various values of n.

Figure 7.18: Increase in shotcrete viscosity with time

As can be seen, a value of n = 1 results in a linear increase in viscosity with time.

However, assuming a value of n > 1 has to be treated with care since this can lead to a
negative creep rate for shotcrete at very early ages. Sometimes it is convenient to give the
magnitude of creep, not for the actual applied stress, but per unit stress. This creep strain
is referred to as specific creep and can be written for the uniaxial case as:

εcr t − to
εcr
sp = = (7.63)
σ η(t)

Since in multiaxial loading conditions creep occurs in more than one direction, the above

242
introduced uniaxial creep law has to be extended for the general stress space. This is achieved
by assuming a so called creep potential P cr for both compression and tension. Similar to
the case of simple elasto-plasticity, a creep flow rule can be established which enables the
calculation of the incremental creep strain vector in a numerical analysis. This creep flow
rule can be written as:
∂P cr ({σ})
{dεcr } = λcr Δt (7.64)
∂σ
where λcr is the creep multiplier controlling the magnitude of the developing creep strains.
It is associated with the specific creep rate, which can be expressed as:

d εcr
sp t (1 − n) + n to
ε̇cr
sp = = * +n+1 (7.65)
dt
t1 ηo tt1

For simplicity reasons and in the absence of experimental data, the creep potential functions
P cr for compression and tension have been chosen to be the same as the plastic potential
for the establishment of the conventional plastic strains. Therefore, the creep potential for
compressive behaviour is:
Pccr = J 2 + Ac p − τc2 (7.66)

and the tensile creep potential can be written as:

  n 1
2J 2π n
Ptcr =p+ √ sin θ + + (aft )n
− ft (7.67)
3 3

Fig. 7.19 shows the plastic potential surfaces for both compression and tension in the biaxial
stress space and J-p space. Furthermore, the directions of the incremental creep strain
vectors, according to the creep flow rule given in equation 7.64, passing through the current
stress state are indicated.

243
Figure 7.19: Creep potential surfaces and incremental creep strain vectors following a creep
flow rule

In the case of a compressive stress state the minor principal stress σ3 is the governing
factor and the creep multiplier can be written as:

ε̇cr
sp Δσ3
λcr
c = (7.68)
∂Pccr
∂σ3

Similarly, if the current stress state is of a tensile nature, the creep multiplier is expressed as
a function of the major principal stress σ1 :

ε̇cr
sp Δσ1
λcr
t = (7.69)
∂Ptcr
∂σ1

A crucial step in a numerical analysis of creep is now the selection of the active creep potential
surface in the case of the current stress state being elastic. It was decided to base this decision
on the magnitude of the principal stresses σ1 and σ3 according to the following criteria:

If σ1 < DU M and σ3 < DU M −→ Compressive creep potential active

Else −→ Tensile creep potential active

with
DU M = M in[| σ1 |, | σ3 |] ∗ 0.001 + SM LT OL (7.70)

where SM LT OL is a user defined tolerance.

244
7.4.4 Shrinkage
Deformation of the shotcrete due to shrinkage caused by the hardening phase of the cement
paste and drying is accounted for in the present constitutive model by the following shrinkage
model according to ACI (1992) (see also Section 6.8.1).

t
εsh = −εsh
∞ (7.71)
B+t

where εsh
∞ is the ultimate shrinkage strain reached at time infinity. Since shotcrete tends to
reduce its volume during shrinkage, these strains are introduced as negative (compressive)
strain components in equal parts to the x-, y- and z-direction. The constant B is a param-
eter (in time units) that controls how fast this ultimate shrinkage strain is approached, as
can be seen in Fig. 7.20. As illustrated, for a lower B value (B1 < B2 < B3 ), in the corre-
sponding time units used in the analysis, the increase in shrinkage deformation is faster. The
incremental formulation of the above shrinkage law is obtained as:

dεsh B
ε̇sh = = −εsh
∞ (7.72)
dt (B + t)2

Finally, these induced shrinkage strains contribute to the plastic strain components and hence
are assumed to be irrecoverable.

Figure 7.20: Adopted shrinkage model and the influence of parameter B

7.4.5 Temperature induced strains

For the modelling of the expansion and contraction of the shotcrete at very early ages due to
the cement hydration, an approach based on the proposed equation by Schubert (1988) (see
Section 6.9) is adopted. The temperature distribution within a shotcrete layer is assumed to
be constant across the thickness of the shotcrete body and hence internal variations due to
thermal boundary conditions are ignored. The thermal deformations that develop during the

245
hardening of the cement paste can be estimated from:

εth = αth ΔT (t) (7.73)

where αth is the coefficient of thermal expansion and is assumed to be constant with time.
ΔT (t) represents the increase in hydration temperature over time and is given by the following
equation:
ΔTmax   
ΔT (t) = 1 − cos At tCt (7.74)
2
where ΔTmax is the maximum increase in temperature above the ambient environmental tem-
perature in the tunnel that occurs after shotcrete installation at the time tmax , see Fig. 7.21.

Figure 7.21: Development of ΔT with time

With increasing time, ΔT starts to reduce again and the shotcrete tends to contract until
there is no further influence of the hydration temperature at the time tzero . For a time t
larger than tzero no thermal strains occur. The two constants At and Ct in equation 7.74 can
be established from the above introduced parameters through the expressions given below:

π ln 2
At = and Ct =   (7.75)
tCt
max tzero
ln
tmax

For the numerical implementation of the current constitutive model the thermal strain rate
ε̇th is needed and can be calculated as:

dεth ΔTmax  
ε̇th = = αth At Ct tCt −1 sin At tCt (7.76)
dt 2

Furthermore, these thermal strains are treated as irrecoverable (or plastic) volumetric strains
and are added to the strain tensor in equal parts to the x-, y- and z-direction.

246
7.5 Summary of parameters
A constitutive model for the time-dependent behaviour of shotcrete has been presented and
discussed in detail in the previous sections. The following tables, Tab. 7.1 to 7.3, summarise
the model parameters that are required for its use in ICFEP. It should be noted that the
time-dependent plasticity model incorporating creep, shrinkage and thermal effects is imple-
mented as NON-LINEAR MODEL 55 in the series of constitutive models available in ICFEP.
However, the time-dependency of the stiffness has been treated as a separate model (MODEL
53), which belongs to the family of the SMALL-STRAIN-STIFFNESS models. Consequently
both SMALL-STRAIN-STIFFNESS MODEL 53 and NON-LINEAR MODEL 55 must be
used together in order to model both stiffness and strength varying with time.

247
 Time-dependent plasticity model
fcp,28 >0 Compressive peak strength at 28 days
fcy,28 fcy,crit < fcy,28 ≤ fcp,28 Compressive yield stress at 28 days
fcf,28 fcf,crit < fcf,28 ≤ fcp,28 Compressive failure strength at 28 days
fcu,28 0 < fcu,28 ≤ fcf,28 Compressive ultimate strength at 28 days
e 1 < e < emax Biaxial strength parameter
ftp,28 0 < ftp,28 ≤ fcp,28 Tensile peak strength at 28 days
fty,28 0 < fty,28 ≤ ftp,28 Tensile yield stress at 28 days
ftu,28 0 < ftu,28 ≤ ftp,28 Tensile ultimate strength at 28 days
s >0 Cement parameter for increase in strength
εpcp,28 >0 Plastic compressive peak strain at 28 days
εpcf,28 > εcp,28 Plastic compressive failure strain at 28 days
εpcu,28 > εcf,28 Plastic compressive ultimate strain at 28 days
εpcp,t2 > εcp,28 Plastic compressive peak strain at time t2
εpcp,t1 > εcp,t2 Plastic compressive peak strain at time t1
εptp,28 >0 Plastic tensile peak strain at 28 days
εptu,28 > εtp,28 Plastic tensile ultimate strain at 28 days
or
ψ <0 Exponential softening exponent
t1 >0 Time corresponding to 1 hour
t28 >0 Time corresponding to 28 days
t2 t1 < t2 < t28 Time corresponding to εpcp,t2
to ≥ t1 Initial shotcrete age
CORT OL 0◦ < CORT OL < 15◦ Tensile deviatoric corner tolerance
a 0<a≤1 Tensile apex tolerance tolerance
n n≥1 Tensile rounding parameter
T IP T OL ≥0 Tip tolerance

Table 7.1: Summary of parameters for time-dependent plasticity model (all the strength and
strain parameters are uniaxial parameters)

248
 Creep, shrinkage and temperature model
ηo >0 Initial creep viscosity
n >0 Creep viscosity exponent
εsh
∞ ≥0 Ultimate shrinkage strain
B >0 Shrinkage parameter
αth ≥0 Coefficient of thermal expansion
ΔTmax ≥0 Maximum increase in hydration temperature
tmax >0 Time t when ΔTmax occurs
tzero > tmax Time t when increase in hydration temperature ΔT is zero

Table 7.2: Summary of model parameters for creep, shrinkage and thermal strains

Time-dependent elasticity model

E28 >0 Young’s modulus of hardened shotcrete at 28 days
s >0 Cement parameter for increase in stiffness
μ −1.0 < μ < 0.5 Poisson’s ratio (constant)
t28 >0 Time corresponding to 28 days
to ≥ t1 Initial shotcrete age

Table 7.3: Summary of parameters for time-dependent elasticity model

7.6 Formulation of the elasto-plastic constitutive matrix for

time-dependent materials
In this section the formulation of the elasto-plastic constitutive matrix, [Dep ], for a time-
dependent material such as shotcrete is presented in detail. This is an extension of the theory
for conventional elasto-plastic materials introduced in Section 6.4.2, including the additional
factor “time” in the adopted yield surfaces and deformations due to creep, shrinkage and
hydration temperature.

In this time-dependent elasto-plastic constitutive model for shotcrete the change in stresses
is related to the change in total strains via the elasto-plastic constitutive matrix [Dep ]. This
relationship can be expressed as:

{Δσ} = [D ep ] {Δε} (7.77)

The changes in total strains can be divided into three components, the elastic strains, the

249
plastic strains and the irreversible strains due to creep, shrinkage and thermal effects. It can
be written:
{Δε} = {Δεel } + {Δεp } + {Δεc } (7.78)

with
{Δεc } = {Δεcr } + {Δεsh } + {Δεth } (7.79)

The changes in stresses are related to the elastic strains through the elastic constitutive
matrix [D] as following:

{Δσ} = [D] [{Δε} − {Δεp } − {Δεc }] (7.80)

The change in irreversible strains {Δεc } yields:

!
∂P cr
{Δε } = λ
c cr
Δt + δij ε̇sh Δt + δij ε̇th Δt (7.81)
∂σ 
 


 shrinkage temperature
creep

where δij = δji is the Kronecker delta, which is defined as being equal to +1 if i and j are
the same numbers and 0 otherwise. By applying the flow rule for the conventional plastic
strains given in equation 6.20, equation 7.80 can be written as:
 !
∂P
{Δσ} = [D] {Δε} − {Δε } − λ c
(7.82)
∂σ

As mentioned previously, the yield function depends on the stress state {σ}, the hardening
(or state) parameters {k} and the time t. In the case of the material being plastic the yield
condition results in:
F ({σ}, {k}, t) = 0 (7.83)

When the material is yielding, the consistency condition requires that the stress state remains
always on the yield surface, so that dF = 0. By applying the chain rule of differentiation for
a plastic strain hardening/softening material one obtains:
!T !T  
∂F ∂F ∂k ∂F
dF = {Δσ} + p
{Δεp } + Δt = 0 (7.84)
∂σ ∂k ∂ε ∂t

Rearrangement of this equation leads to:

!T  
∂F ∂k ∂F
− {Δεp } − Δt
∂k ∂εp ∂t
{Δσ} = ! (7.85)
∂F T
∂σ

250
Combining equations 7.82 and 7.85 results in:
!T  
∂F ∂k ∂F
! p
{Δεp } + Δt
∂P ∂k ∂ε ∂t
[D] [{Δε} − {Δεc }] − λ [D] =− ! (7.86)
∂σ ∂F T
∂σ

and further
!T ! !
∂F ∂P ∂F T
[D] [{Δε} − {Δε }] − λ [D]
c
=
∂σ ∂σ ∂σ
- !   .
∂F T ∂k ∂F
− {Δεp } + Δt (7.87)
∂k ∂εp ∂t

From this equation the plastic multiplier λ can be obtained as:

!T
∂F ∂F
[D] [{Δε} − {Δεc }] + Δt
∂σ ∂t
λ= ! ! (7.88)
∂P ∂F T
[D] +A
∂σ ∂σ

where A denotes the hardening modulus, which is given as:

!T  
1 ∂F ∂k
A=− {Δεp } (7.89)
λ ∂k ∂εp

Making use of equations 7.82 and 7.88 leads to:

!T
∂F ∂F
[D] [{Δε} − {Δεc }] + Δt !
∂σ ∂t ∂P
{Δσ} = [D] [{Δε} − {Δε }] − c
! ! [D] (7.90)
∂P ∂F T ∂σ
[D] +A
∂σ ∂σ

Further rearrangement of the above equation results in:

! !T
∂P ∂F
[D] [D] [{Δε} − {Δεc }]
∂σ ∂σ
{Δσ} = [D] [{Δε} − {Δεc }] − ! ! −
∂P ∂F T
[D] +A
∂σ ∂σ
!
∂P ∂F
[D] Δt
∂σ ∂t
− ! ! (7.91)
∂P ∂F T
[D] +A
∂σ ∂σ

Finally, a relationship between the changes in stresses and strains for the time-dependent

251
material shotcrete can be obtained as:

{Δσ} = [Dep ] [{Δε} − {Δεc }] − {Δσ c } (7.92)

where [Dep ] is the conventional elasto-plastic constitutive matrix given as:

! !T
∂P ∂F
[D] [D]
∂σ ∂σ
[Dep ] = [D] − ! !T (7.93)
∂P ∂F
[D] +A
∂σ ∂σ

and {Δσ c } is an additional generated stress vector that accounts for the time-dependent
behaviour of the yield surfaces. It can be written:
!
∂P ∂F
[D] Δt
∂σ ∂t
{Δσ c } = ! ! (7.94)
∂P ∂F T
[D] +A
∂σ ∂σ

In the case of purely elastic material behaviour this stress vector reduces to zero.

7.7 Correction of the right hand side load vector

Chapter 3 introduced the basic theory behind the establishment of the governing equations
that are necessary to describe the deformational behaviour of each element during a finite
element analysis. However, this procedure did not include the aspects affecting these govern-
ing element equations that arise due to the time-dependent nature of the material behaviour
of sprayed concrete. Therefore, this section is a continuation from Section 3.2.3 and the
previous Section 7.6 and explains in detail how the right hand side load vector is determined
in the case of a time-dependent material behaviour.

The principle of minimum potential energy requires the determination of the strain energy,
which can be written as: )
1
ΔW = {Δε}T {Δσ}dV ol (7.95)
2 V ol

by integrating over the volume of the body. This equation can be rearranged as:
)
1
ΔW = {Δε}T [[Dep ] [{Δε} − {Δεc }] − {Δσ c }] dV ol (7.96)
2 V ol

and further
)
1  
ΔW = {Δε}T [Dep ] [{Δε} − {Δεc }] − {Δε}T {Δσ c } dV ol (7.97)
2 V ol

252
The incremental total potential energy ΔE of a body is now defined as:

ΔE = ΔW − ΔL (7.98)

with ΔL being the work done by the applied loads, which is given in equation 3.14 of Chapter
3. It can be written as:
) )
1 1
ΔE = {Δε} [D ] [{Δε} − {Δε }]dV ol −
T ep c
{Δε}T {Δσ c }dV ol
2 V ol 2 V ol
) )
− {Δd} {ΔF }dV ol −
T
{Δd}T {ΔT }dSrf (7.99)
V ol Srf

This equation can be further expressed as:

N  )
 )
1 1
ΔE = {Δd}Tn [B]T [Dep ][B]{Δd}n dV ol − {Δd}Tn [B]T [Dep ]{Δεc }dV ol
2 V ol 2 V ol
i=1
) )
1
− {Δd}Tn [B]T {Δσ c }dV ol − {Δd}T {ΔF }dV ol
2
) V ol
 V ol

− {Δd} {ΔT }dSrf

T
(7.100)
Srf

After minimising the potential energy with respect to the incremental nodal displacements,
the right hand side load vector {ΔRE } gets an extra term φ due to the time-dependency of
the adopted yield functions and the strain components accounting for creep, shrinkage and
hydration temperature, which can be written as:
) )
1 1
φ=+ [B]T {Δσ c }dV ol + [B]T [Dep ]{Δεc }dV ol (7.101)
2 V ol 2 V ol

7.8 Multi-surface plasticity for time-dependent materials

Section 6.4.3 of the previous chapter dealt with the theoretical background of elasto-plasticity
in the case when several yield surfaces are active at the same time. Within the current section,
this theory is extended for the developed time-dependent constitutive model for shotcrete,
where during an analysis both the compression and the tension yield surface can be controlling
the mechanical behaviour simultaneously. The numerical formulation includes also the time-
dependent nature of the adopted yield surfaces which leads to the mathematical procedure
presented below:

If the stress state during an analysis is such that both the compression and the tension yield
surfaces are moving simultaneously in the stress space, the change in total strains can be
divided into the components given below:
0 1 0 1
{Δε} = {Δεe } + Δεp1 + Δεp2 + {Δεc } (7.102)

253
 0 1
where Δεp1 denotes the change in plastic strains due to the compression yield surface
0 1
being active and Δεp2 the change in plastic strains given by the tension yield surface. The
strain component {Δεc } describes the additional irreversible strains according to the adopted
constitutive behaviour for creep, shrinkage and thermal deformation and can be written in
the case of multiplasticity as:
! !
1 ∂Pccr 1 ∂Ptcr
{Δε } = λcr
c
Δt + λcr Δt + δij ε̇sh Δt + δij ε̇th Δt (7.103)
2 c ∂σ 2 t ∂σ

In this equation λcr cr

c and λt are the creep multipliers for the compressive and the tensile
creep behaviour and were previously introduced in equations 7.68 and 7.69 respectively. It
should be noted that the equal contribution of both creep potentials expressed by the factor
1/2 is an arbitrary assumption. The changes in stresses are related to the changes in elastic
strains through the elastic constitutive matrix [D] as:

{Δσ} = [D] {Δεe } (7.104)

It follows that:
0 1 0 1
{Δσ} = [D] [{Δε} − {Δεc } − Δεp1 − Δεp2 ] (7.105)

The plastic strain components for both yield surfaces can be obtained from the flow rules,
that read as: ! !
0 p1 1 ∂P1 0 1 ∂P2
Δε = λ1 and Δεp2 = λ2 (7.106)
∂σ ∂σ
Substitution into equation 7.105 results in:
! !
∂P1 ∂P2
{Δσ} = [D] [{Δε} − {Δε } − λ1 c
− λ2 ] (7.107)
∂σ ∂σ

As mentioned earlier, both yield functions depend on the stress state {σ}, the state (or
hardening) parameters {k} and time t. They can be expressed as:

F1 ({σ}, {k}, t) = 0 and F2 ({σ}, {k}, t) = 0 (7.108)

Applying the consistency condition to both yield functions results in the following total
differentials for both the compressive and the tensile behaviour:
!T !T  
∂F1 ∂F1 ∂k ∂F1
dF1 = {Δσ} + p {Δε } +
p
Δt = 0 (7.109)
∂σ ∂k ∂ε3 ∂t

and !T !T  
∂F2 ∂F2 ∂k ∂F2
dF2 = {Δσ} + p {Δε } +
p
Δt = 0 (7.110)
∂σ ∂k ∂ε1 ∂t

254
By substituting equation 7.107 into the above equations one obtains:
!T ! !
∂F1 ∂F1 T ∂P1
dF1 = [D]({Δε} − {Δε }) − λ1
c
[D] −
∂σ ∂σ ∂σ
! ! !  
∂F1 T ∂P2 ∂F1 T ∂k
−λ2 [D] + {Δεp } +
∂σ ∂σ ∂k ∂εp3
∂F1
+ Δt = 0 (7.111)
∂t

and
!T ! !
∂F2 ∂F2 T ∂P1
dF2 = [D]({Δε} − {Δε }) − λ1
c
[D] −
∂σ ∂σ ∂σ
! ! !  
∂F2 T ∂P2 ∂F2 T ∂k
−λ2 [D] + {Δεp } +
∂σ ∂σ ∂k ∂εp1
∂F2
+ Δt = 0 (7.112)
∂t

Furthermore, A1 and A2 are defined as:

!T  
1 ∂F1 ∂k
A1 = − {Δεp } (7.113)
λ1 ∂k ∂εp3

and !T  
1 ∂F2 ∂k
A2 = − {Δεp } (7.114)
λ2 ∂k ∂εp1
Combining these two equations with equations 7.111 and 7.112 leads to:
!T !T !
∂F1 ∂F1 ∂P1
dF1 = [D]({Δε} − {Δε }) − λ1
c
[D] −
∂σ ∂σ ∂σ
!T !
∂F1 ∂P2 ∂F1
−λ2 [D] − λ1 A1 + Δt = 0 (7.115)
∂σ ∂σ ∂t

and
!T !T !
∂F2 ∂F2 ∂P1
dF2 = [D]({Δε} − {Δε }) − λ1
c
[D] −
∂σ ∂σ ∂σ
!T !
∂F2 ∂P2 ∂F2
−λ2 [D] − λ2 A2 + Δt = 0 (7.116)
∂σ ∂σ ∂t

For simplification it can be rewritten:

λ1 L11 + λ2 L12 = T1 and λ1 L21 + λ2 L22 = T2 (7.117)

255
The coefficients L11 , L12 , L21 and L22 are now the same as for a material without the time-
dependent response and can therefore be written as:
!T !
∂F1 ∂P1
L11 = [D] + A1
∂σ ∂σ
!T !
∂F2 ∂P2
L22 = [D] + A2
∂σ ∂σ
!T ! (7.118)
∂F1 ∂P2
L12 = [D]
∂σ ∂σ
!T !
∂F2 ∂P1
L21 = [D]
∂σ ∂σ

It can be seen that the parameters T1 and T2 contain an additional term accounting for the
time-dependent nature of the yield functions and are obtained as:
!T
∂F1 ∂F1
T1 = [D] ({Δε} − {Δεc }) + Δt (7.119)
∂σ ∂t

and !T
∂F2 ∂F2
T2 = [D] ({Δε} − {Δεc }) + Δt (7.120)
∂σ ∂t
As a solution for the plastic strain multipliers λ1 and λ2 it can be written:

L22 T1 − L12 T2 L11 T2 − L21 T1

λ1 = and λ2 = (7.121)
L11 L22 − L12 L21 L11 L22 − L12 L21

By setting Ω = L11 L22 − L12 L21 these equations are expressed as:

L22 T1 − L12 T2 L11 T2 − L21 T1

λ1 = and λ2 = (7.122)
Ω Ω

Substituting these expressions for λ1 and λ2 into equation 7.105 leads to:
!  !
[D] ∂P1 ∂F1 T
{Δσ} = [D] ({Δε} − {Δε }) − c
L22 [D] ({Δε} − {Δεc }) +
Ω ∂σ ∂σ
  ! 
∂F1 ∂F2 T ∂F2
+ Δt −L12 [D] ({Δε} − {Δε }) +
c
Δt −
∂t ∂σ ∂t
!  ! 
[D] ∂P2 ∂F2 T ∂F2
L11 [D] ({Δε} − {Δε }) +
c
Δt −
Ω ∂σ ∂σ ∂t
 ! 
∂F1 T ∂F1
L21 [D] ({Δε} − {Δε }) +
c
Δt (7.123)
∂σ ∂t

256
Rearrangement of the above equation yields in:
!  !
[D] ∂P1 ∂F1 ∂F2 [D] ∂P2 ∂F2
{Δσ} + L22 Δt − L12 Δt + L11 Δt −
Ω ∂σ ∂t ∂t Ω ∂σ ∂t
  ! ! ! 
∂F1 [D] ∂P1 ∂F1 T ∂F2 T
−L21 Δt = [D] − L22 [D] − L12 [D] −
∂t Ω ∂σ ∂σ ∂σ
! ! ! 
[D] ∂P2 ∂F2 T ∂F1 T
− L11 [D] − L21 [D] ({Δε} − {Δεc }) (7.124)
Ω ∂σ ∂σ ∂σ

Finally, the change in stresses is given as:

!  !
[D] ∂P1 ∂F1 ∂F2 [D] ∂P2 ∂F2
{Δσ} + L22 Δt − L12 Δt + L11 Δt −
Ω ∂σ ∂t ∂t Ω ∂σ ∂t

∂F1
−L21 Δt +[Dep ]{Δεc } = [Dep ] {Δε} (7.125)
∂t

where the elasto-plastic constitutive matrix can be expressed as:

! ! ! 
[D] ∂P1 ∂F1 T ∂F2 T
[D ] = [D] −
ep
L22 [D] − L12 [D] −
Ω ∂σ ∂σ ∂σ
! ! ! 
[D] ∂P2 ∂F2 T ∂F1 T
− L11 [D] − L21 [D] (7.126)
Ω ∂σ ∂σ ∂σ

The additional generated stress vector accounting for the time-dependent nature of the
adopted yield surfaces can be obtained as:
! 
[D] ∂P1 ∂F1 ∂F2
{Δσ } =
c
L22 Δt − L12 Δt +
Ω ∂σ ∂t ∂t
! 
[D] ∂P2 ∂F2 ∂F1
+ L11 Δt − L21 Δt (7.127)
Ω ∂σ ∂t ∂t

7.9 Model implementation

The numerical implementation of the presented constitutive model for shotcrete into the finite
element program ICFEP involved the following steps:

1. Calculation of the elasto-plastic constitutive matrix [Dep ] according to Section 7.6 or

7.8, which allows the establishment of the element stiffness matrix [KE ]. This includes
the incorporation of the time-dependent behaviour of the adopted yield surfaces by
calculating the additional stress contribution {Δσ c }. The element stiffness matrix is
then assembled into the global stiffness matrix [KG ] in order to form the global set of
equations, where the right hand side load vector is corrected by the factor φ given in
Section 7.7.

2. Calculation of the residual load vector {ψ}j at the end of each iteration during the mod-
ified Newton-Raphson scheme which is used as the non-linear solver. This is achieved

257
 by integrating the constitutive equations along the incremental strain path by a sub-
stepping algorithm to estimate the stress changes (see Chapter 3). This procedure
requires the calculation of the changes in stresses, plastic strains and hardening param-
eters during each substep.

By inspecting equations 6.20, 7.93 and 7.94 it can be observed that the following terms are
required:

 Partial derivatives of the yield functions with respect to stress {∂F/∂σ}

 Partial derivatives of the plastic potential functions with respect to stress {∂P/∂σ}

 Partial derivatives of the yield functions with respect to time ∂F/∂t

 The elastic constitutive matrix [D]

 The hardening modulus A

The following sections present the necessary calculations neeeded for the robust numerical
implementation of the current constitutive model for shotcrete into ICFEP for each of the
two yield surfaces that are adopted.

7.9.1 Selection of the active yield surface

The selection of the active yield surface during an analysis is based on the values of the
compressive yield function Fc and the tensile yield function Ft calculated from the current
stress state. Three different scenarios are possible, which are the following:

1. Ft ≥ −Y T OL and Ft > Fc → Tension surface active

2. Fc ≥ −Y T OL and Fc > Ft → Compression surface active

3. Fc ≥ −Y T OL and Ft ≥ −Y T OL → Both surfaces active

where Y T OL is a user defined tolerance, specifying the numerical tolerance on the yield
function.

7.9.2 Derivatives of the stress invariants

The general derivatives of the stress invariants p, J and θ will be used in the following sections:
!
∂p 1
= {1 1 1 0 0 0}T (7.128)
∂σ 3

!
∂J 1
= {σx − p σy − p σz − p 2 τxy 2 τxz 2 τyz }T (7.129)
∂σ 2J

258
 ! √  ! !
∂θ 3 3 det s ∂J ∂(det s)
= − (7.130)
∂σ 2 cos (3θ) J 3 J ∂σ ∂σ
with

det s = (σx − p)(σy − p)(σz − p) − (σx − p) τyz

2
− (σy − p) τzx
2
−
−(σz − p) τxy
2
+ 2 τxy τyz τzx (7.131)

7.9.3 Partial derivatives for the compression yield surface

In this case, when only the compressive yield surface is active, the mechanical behaviour of
shotcrete is goverened by the yield function given in equation 7.4. The calculation of the
partial derivative of the yield function with respect to stress, {∂Fc /∂σ}, will be given first,
followed by the partial derivative of the yield function with respect to time, ∂Fc /∂t.

i) Calculation of {∂Fc /∂σ}

As Fc is independent of θ, it can be written:

! ! !
∂Fc ∂Fc ∂p ∂Fc ∂J
= + (7.132)
∂σ ∂p ∂σ ∂J ∂σ

with
∂Fc ∂Fc
= Ac and = 2J (7.133)
∂p ∂J
Since associated conditions are adopted for the current constitutive model, the derivative of
the compressive plastic potential with respect to stress, {∂Pc /∂σ}, equals the derivative of
the compressive yield function and this can be expressed as:
! !
∂Fc ∂Pc
= (7.134)
∂σ ∂σ

ii) Calculation of ∂Fc /∂t

The calculation of the derivative of the compressive yield function with respect to time
involves the following calculations:
 
∂Fc ∂Fc ∂Ac ∂Fc ∂τc ∂fc
= + (7.135)
∂t ∂Ac ∂fc ∂τc ∂fc ∂t

with
∂Fc ∂Ac
=p and = αc (7.136)
∂Ac ∂fc
In addition it can be written:

∂Fc ∂τc 
= −2τc and = βc (7.137)
∂τc ∂fc

259
Therefore equation 7.135 can be rewritten as:

∂Fc   ∂fc
= αc p − 2 τc βc (7.138)
∂t ∂t

The derivative of the equivalent uniaxial compressive strength with respect to time, ∂fc /∂t,
is now required for all four zones of the normalised hardening and softening curve, earlier
described in Section 7.3.1. This term can be expressed as:

Zone I:
√
∂fc    s t28
= fcy,n + (1 − fcy,n ) 2Hc − Hc2fcp (t) −
∂t 2 t 32
  B−1
Hc A B t
−2 fcp (t) (1 − fcy,n ) (1 − Hc ) p ln (7.139)
εcp (t) t t1

Zone II:
  √
∂fc fcf,n − 1 s t28
= 1+ (Hc − 1) fcp (t) −
∂t Hcf − 1 2 t 32
  B−1
fcf,n − 1 Hc A B t
fcp (t) p ln (7.140)
Hcf − 1 εcp (t) t t1

Zone III:
  √
∂fc fcu,n − fcf,n s t28
= fcf,n + (Hc − Hcf ) fcp(t) −
∂t Hcu − Hcf 2 t 23
  B−1
fcu,n − fcf,n Hc A B t
−fcp (t) p ln (7.141)
Hcu − Hcf εcp (t) t t1

Zone IV: √
∂fc s t28
= fcu,n fcp(t) (7.142)
∂t 2 t 32
As discussed previously, the material parameters for shotcrete are changing for a shotcrete
age less than 28 days but are constant afterwards. This implies that for the time t > t28
during an analysis the derivative ∂fc /∂t becomes zero.

7.9.4 Partial derivatives for the tension yield surface

The yield function given in equation 7.9 controls the material response of shotcrete in tension.
Similar to the compression surface, the derivative of the yield function with respect to stress,
{∂Ft /∂σ}, is established first, followed by the derivative of the yield function with respect to
time, ∂Ft /∂t.

i) Calculation of {∂Ft /∂σ}

260
It is given that: ! ! ! !
∂Ft ∂Ft ∂p ∂Ft ∂J ∂Ft ∂θ
= + + (7.143)
∂σ ∂p ∂σ ∂J ∂σ ∂θ ∂σ
with
∂Ft
=1 (7.144)
∂p
  n  1−n   n
∂Ft 2J 2π n 2 2π
= √ sin θ + + (a ft )n
J n−1
√ sin θ + (7.145)
∂J 3 3 3 3
and
  n  1−n  
∂Ft 2J 2π n 2J n
= √ sin θ + + (a ft ) n
√ ·
∂θ 3 3 3
  n−1  
2π 2π
· sin θ + cos θ + (7.146)
3 3

As in the case of the compression surface, associated conditions are applied for determining
the tensile plastic potential and therefore the following expression for the derivative of the
tensile plastic potential with respect to stress, {∂Pt /∂σ}, can be established:
! !
∂Ft ∂Pt
= (7.147)
∂σ ∂σ

ii) Calculation of ∂Ft /∂t

The following procedure applies to the determination of the derivative of the tensile yield
function with respect to time, ∂Ft /∂t:

∂Ft ∂Ft ∂ft

= (7.148)
∂t ∂ft ∂t

with
  n  1−n
∂Ft 2J 2π n
= √ sin θ + + (a ft )n
an ftn−1 − 1 (7.149)
∂ft 3 3
The derivative of the equivalent uniaxial tensile strength with respect to time, ∂ft /∂t, has to
be calculated for each of the 3 zones of the normalised hardening/softening curve for tension
presented in section 7.3.2.

Zone I:
√
∂ft   
s t28
= fty,n + (1 − fty,n ) 2Ht − Ht2
ftp (t) −
∂t 2 t 23
  B−1
Ht A B t
−2 ftp (t) (1 − fty,n ) (1 − Ht ) p ln (7.150)
εcp (t) t t1

261
Zone II:
In the case of the linear softening in option A it can be written as:
  √
∂ft ftu,n − 1 s t28
= 1+ (Ht − 1) ftp (t) −
∂t Htu − 1 2 t 32
  B−1
ftu,n − 1 Ht A B t
−ftp (t) ln (7.151)
Htu − 1 εpcp (t) t t1

If the softening is described by the exponential function given in option B, it can be expressed
as:
√
∂ft s t28
= exp (−ψ (Ht − 1)) ftp (t) +
∂t 2 t 32
  B−1
Ht A B t
+ψ ftp (t) exp (−ψ (Ht − 1)) p ln (7.152)
εcp (t) t t1

Zone III: √
∂ft s t28
= ftu,n ftp (t) (7.153)
∂t 2 t 32
Regarding the time-dependency of the yield function, the same rule applies for the tension
surface as for the compression surface: material parameters are considered to be constant for
a shotcrete age larger than 28 days and therefore ∂ft /∂t reduces to zero if t > t28 during the
analysis.

7.9.5 Calculation of the hardening modulus for the compression surface

In Section 7.3.1 it was mentioned that the minor principal plastic strain εp3 is controlling the
motion of the compression yield surface in the general stress space. Following equation 7.89,
the hardening modulus A for the compression surface being the active yield surface can be
written as:
1 ∂Fc p
A=− p Δε3 (7.154)
λ ∂ε3
The general flow rule given in equation 6.20 for an elasto-plastic material can now be specified
for the compression surface of the current constitutive model as:

∂Fc
Δεp3 = −λ (7.155)
∂σ3

Combining these two equations one obtains the hardening modulus as:

∂Fc ∂Fc
A= (7.156)
∂εp3 ∂σ3

The derivative of the compression surface with respect to the minor principal plastic strain,
∂Fc /∂εp3 , is given as:  
∂Fc ∂Fc ∂Ac ∂Fc ∂τc ∂fc
= + (7.157)
∂εp3 ∂Ac ∂fc ∂τc ∂fc ∂εp3

262
The terms ∂Fc /∂Ac , ∂Ac /∂fc , ∂Fc /∂τc and ∂τc /∂fc have been presented in equations 7.136
and 7.137. Furthermore, the derivative of the equivalent uniaxial compressive stress with
respect to the minor principal plastic strain, ∂fc /∂εp3 is required for all the four zones of the
normalised hardening/softening curve for compression introduced in Section 7.3.1.

Zone I:
∂fc 2 fcp (t) (1 − fcy,n ) (1 − Hc )
p = (7.158)
∂ε3 εpcp (t)
Zone II:
∂fc fcf,n − 1
= fcp (t) (7.159)
∂εp3 (Hcf − 1) εpcp (t)
Zone III:
∂fc fcu,n − fcf,n
= fcp (t) (7.160)
∂εp3 (Hcu − Hcf ) εpcp (t)
Zone IV:
∂fc
=0 (7.161)
∂εp3
In addition, the derivative of the compression surface with respect to the minor principal
stress, ∂Fc /∂σ3 , is required for the calculation of the hardening modulus A. It can be
written:
∂Fc ∂Fc ∂p ∂Fc ∂J
= + (7.162)
∂σ3 ∂p ∂σ3 ∂J ∂σ3
The derivatives ∂Fc /∂p and ∂Fc /∂J have already been established in equation 7.133. The
missing terms can be expressed as:

∂p 1 ∂J 1
= and = (σ3 − p) (7.163)
∂σ3 3 ∂σ3 2J

Finally, the hardening modulus A for the compression surface can be determined as:
*  
 + Ac ∂fc
A = αc p − 2 τc βc + σ3 − p (7.164)
3 ∂εp3

7.9.6 Calculation of the hardening modulus for the tension surface

In contrast to the compression surface, the governing parameter for the tensile plastic strain
hardening/softening is the major principal plastic strain εp1 , as highlighted in Section 7.3.2.
Therefore, the hardening modulus for the tensile material behaviour can be expressed as:

1 ∂Ft
A=− Δεp1 (7.165)
λ ∂εp1

The general flow rule from equation 6.20 yields for the tension surface:

∂Ft
Δεp1 = λ (7.166)
∂σ1

263
Rearrangement leads to:
∂Ft ∂Ft
A=− (7.167)
∂εp1 ∂σ1
The derivative of the tension surface with respect to the major principal plastic strain,
∂Ft /∂εp1 , can be written as:
∂Ft ∂Ft ∂ft
p = (7.168)
∂ε1 ∂ft ∂εp1
The term ∂Ft /∂ft was introduced in equation 7.149. Similar to the compression surface, the
derivative of the equivalent uniaxial tensile stress with respect to the major principal plastic
strain, ∂ft /∂εp1 , has to be calculated for each zone of the normalised hardening/softening
curve presented in Section 7.3.2.

Zone I:
∂ft 2 ftp (t) (1 − fty,n ) (1 − Ht )
p = (7.169)
∂ε1 εptp (t)
Zone II:
In the case of linear softening in option A it can be written:

∂ft ftu,n − 1
p = ftp (t) (7.170)
∂ε1 (Htu − 1) εptp (t)

For the exponential softening in option B it can be expressed:

∂ft exp (−ψ (Ht − 1))

p = −ψ ftp (t) (7.171)
∂ε1 εptp (t)

Zone III:
∂ft
=0 (7.172)
∂εp1
Furthermore, the derivative of the tensile yield function with respect to the major principal
stress, ∂Ft /∂σ1 is needed, which is given as:

∂Ft ∂Ft ∂p ∂Ft ∂J ∂Ft ∂θ

= + + (7.173)
∂σ1 ∂p ∂σ1 ∂J ∂σ1 ∂θ ∂σ1

The derivatives ∂Ft /∂p, ∂Ft /∂J and ∂Ft /∂θ were calculated in equations 7.144, 7.145 and
7.146 respectively. The missing terms can be established as:

∂p 1 ∂J 1
= and = (σ1 − p) (7.174)
∂σ1 3 ∂σ1 2J

furthermore
√  
∂θ 3 3 det s 1 1
= (σ1 − p) − (2 σ2 σ3 − σ1 σ3 − σ1 σ2 ) (7.175)
∂σ1 2 cos (3θ) J 3 J 2J 3

264
Combining the above equations, the hardening modulus for the tension surface can finally be
determined as:
 
  1 1 ∂ft
A = − Q an ftn−1 − 1 + (σ1 − p) R + (2 σ2 σ3 − σ1 σ3 − σ1 σ2 ) S (7.176)
3 3 ∂εp1

with  
2 2π
P = √ sin θ + (7.177)
3 3
1−n
Q = [(P J)n + (a ft )n ] n (7.178)

and n  
nQ Q (P J) cos θ + 2π
n−2 3 3 det s
R=P J + (7.179)
2 2 P cos (3θ) J 5
 
Q (P J)n cos θ + 2π
S=− 3
(7.180)
P cos (3θ) J 3

7.9.7 Calculation of the elastic constitutive matrix

The elastic constitutive matrix has been introduced in Section 6.3.1 of the previous chapter.
It can be expressed in terms of the time-dependent Young’s modulus E(t) and the constant
Poisson’s ratio μ as:
⎡ ⎤
⎢ 1−μ μ μ 0 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ μ 1−μ μ 0 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
E(t) ⎢ μ μ 1−μ 0 0 0 ⎥
[D] = ⎢ ⎥ (7.181)
(1 + μ) (1 − 2μ) ⎢
⎢ 0 1 − 2μ ⎥
⎥
⎢ 0 0 0 0 ⎥
⎢ 2 ⎥
⎢ 1 − 2μ ⎥
⎢ 0 0 0 0 0 ⎥
⎢ 2 ⎥
⎣ ⎦
1 − 2μ
0 0 0 0 0
2

with the development of the time-dependent Young’s modulus E(t) given in equation 7.54.

7.9.8 Calculation of the changes in the compressive hardening parameter

The normalised hardening parameter which controls the motion of the compression yield
surface in general stress space has been introduced in equation 7.13. The change of this
hardening parameter during an analysis can be expressed as:

∂Hc p ∂Hc ∂εpcp (t)

ΔHc = Δε + Δt (7.182)
∂εp3 3
∂εpcp (t) ∂t

265
The involved derivatives in the above equation can be calculated as:

∂Hc 1
p = p (7.183)
∂ε3 εcp (t)

∂Hc Hc
p =− p (7.184)
∂εcp (t) εcp (t)
  B−1
∂εpcp (t) AB t
= ln (7.185)
∂t t t1
Therefore, for a shotcrete age less than 28 days the change in the compressive hardening
parameter can be written as:
  B−1
1 Hc A B t
ΔHc = p Δεp3 − p ln Δt (7.186)
εcp (t) εcp (t) t t1

In the case of t > t28 this equation reduces to:

1
ΔHc = Δεp3 (7.187)
εpcp (t)

7.9.9 Calculation of the changes in the tensile hardening parameter

Equation 7.32 presented the normalised hardening parameter for the tension yield surface
where the major principal plastic strain, εp1 controls the hardening and softening of the
shotcrete. The change in the tensile hardening parameter during an analysis can be expressed
as: p
∂Ht p ∂Ht ∂εtp (t)
ΔHt = Δε + Δt (7.188)
∂εp1 1
∂εptp (t) ∂t
The following derivatives can be calculated:

∂Ht 1
p = p (7.189)
∂ε1 εtp (t)

∂Ht Ht
=− p (7.190)
∂εptp (t) εtp (t)
  B−1
∂εptp (t) δAB t
= ln (7.191)
∂t t t1
Combining these equations, the change in the normalised tensile hardening parameter be-
comes:   B−1
1 Ht A B t
ΔHt = p Δεp1 − p ln Δt (7.192)
εtp (t) εcp (t) t t1
For t > t28 the above equation becomes:

1
ΔHt = Δεp1 (7.193)
εptp (t)

266
7.10 Model calibration
The aim of the following sections is to derive the necessary material parameters needed for
the above presented constitutive model for shotcrete, so that the model can be validated
and used in a boundary value problem. Since any experimental work on shotcrete behaviour
was not part of this research, this model calibration is performed mainly with the help of
published data. It should be mentioned that a complete set of test data for one particular
shotcrete type at various ages is rare, if not impossible to find. Furthermore, tests on shotcrete
focus mainly on the uniaxial compression behaviour and no information could be found on
the early age material behaviour in tension. Test results for multiaxial loading conditions
are hardly considered in any experimental research programme. Unfortunately, since there
is a large amount of influencing factors on the material behaviour (mix design, spraying
technique, rebound behaviour, etc.) almost all of the experimental results available in the
literature show a considerable scatter, which makes a precise derivation of model parameters
difficult. Sample preparation seems to be another key factor, where the factor “time” plays
an important role. However, it will be shown that the developed constitutive model is capable
of predicting the main characteristics of shotcrete behaviour, including the transition from a
very ductile to a more brittle material response with curing time. Finally, some curve fitting
for published data on creep, shrinkage and hydration temperature is also presented.

7.10.1 Increase in stiffness and compressive strength

Huber (1991) performed a series of 5 tests on shotcrete samples, taken from the Inntal Tunnel
in Austria, with different accelerator contents. The specimens where sprayed in the dry-
mix method directly at the tunnel construction site into a sample-preparation box, avoiding
the inclusion of rebound material, and had a size of 10 × 10 × 40 cm. The shotcrete mix
proportions were the following (Tab. 7.4):

Aggregates (0/12 mm) Cement PZ 275 Water Accelerator

1800 kg/m3 350 kg/m3 160 l 5 and 7 %

Table 7.4: Mix design for tested shotcrete (from Huber, 1991)

During the load controlled compression tests the increase in stiffness and strength was
measured for different shotcrete ages of up to 168 h. Fig. 7.22 and Fig. 7.23 illustrate the
obtained results for all the 5 tests.

267
 35000

30000
s = 0.1
25000

Young's modulus [MPa]

20000
s = 0.2 s = 0.4
15000

10000

5000 Test data

Fitted equations
0
0 20 40 60 80 100 120 140 160 180
Time [h]

Figure 7.22: Parameter fitting for increase in stiffness with time (after Huber, 1991)

28

24
s = 0.1
Compressive strength [MPa]

20

16

s = 0.2 s = 0.35
12

4 Test data
Fitted equations
0
0 20 40 60 80 100 120 140 160 180
Time [h]

Figure 7.23: Parameter fitting for increase in compressive strength with time (after Huber,
1991)

From the experimental data it can be seen that the development of stiffness with time
appears to grow somewhat faster than the increase in compressive strength with time, which is
consistent with the adopted model equations. Furthermore, by fitting the involved parameters
for equation 7.54 and 7.56, a possible scatter in the test data can be covered by establishing
a lower bound, upper bound and intermediate curve, which would represent an optimum fit.
As a reference model parameter for the Young’s modulus at 28 days a value of E28 = 32 GPa

268
has been chosen to be a reasonable value. The compressive strength for hardened shotcrete
at 28 days was estimated as fcp,28 = 30 MPa. Tab. 7.5 contains the possible values of the
cement parameter s for both the increase in stiffness and strength according to the model
equations used for the graphical illustration in Fig. 7.22 and Fig. 7.23.

s for stiffness s for strength

(-) (-)
Upper bound 0.1 0.1
Lower bound 0.4 0.35
Intermediate value 0.2 0.2

Table 7.5: Cement parameter for increase in stiffness and strength fitted to test data from
Huber (1991)

7.10.2 Complete uniaxial compression tests at different shotcrete ages

As mentioned in Section 5.7 of Chapter 5, Sezaki et al. (1989) performed an advanced experi-
mental study on the mechanical properties of early age shotcrete in compression. Their tests
involved the analysis of the development of the compressive strength, the Young’s modulus
and the failure strain with curing time for the following specified wet-mix, summarised in
Tab. 7.6:

Max. size W/C Water Cement Sand Gravel Accelerator

of aggregate concentration
(mm) (%) (kg/m3 ) (kg/m3 ) (kg/m3 ) (kg/m3 ) (%)

10 57 217 380 1115 633 6, 8, 10

Table 7.6: Mix design for tested shotcrete (from Sezaki et al., 1989)

The first step for calibrating the corresponding equations of the presented constitutive
model was to fit the increase in stiffness and compressive strength according to the adopted
model equations 7.54 and 7.56 to the obtained test data, as can be seen in Fig. 7.24 and
Fig. 7.25.

269
 6000
Test data
Fitted equation
5000

Young's modulus [MPa]

4000

3000

2000

1000

0
1 10 100 1000
log time [h]

Figure 7.24: Parameter fitting for increase in stiffness with time (after Sezaki et al., 1989)

20

Test data
Fitted equation
16
Compressive strength [MPa]

12

0
1 10 100 1000
log time [h]

Figure 7.25: Parameter fitting for increase in compressive strength with time (after Sezaki
et al., 1989)

The next step in the calibration procedure involved the parameter fitting of the total
peak and failure strains at various shotcrete ages according to the adopted model equation.
It was highlighted earlier that the presented constitutive model is formulated within the
concept of plastic strain hardening/softening. Furthermore, equation 7.58 is established for
the development of the plastic peak strains and therefore it is of crucial importance for a
correct model calibration to incorporate the elastic strain components via the stiffness and

270
strength (see Section 7.4.2). Fig. 7.26 illustrates the fitted model equations to the obtained
test data measured from the published uniaxial compressive stress strain curves by Sezaki
et al. (1989).

5,5

5,0 Test data peak strains

Test data failure strains
4,5 Fitted equations
4,0

3,5
Strain [%]

3,0

2,5

2,0

1,5

1,0

0,5

0,0
1 10 100 1000
log time [h]

Figure 7.26: Parameter fitting for development of total peak and failure strains (after Sezaki
et al., 1989)

Finally, with the estimated model parameters the complete uniaxial stress strain curves in
compression at different shotcrete ages can be established up to failure, where the concrete is
assumed to be crushed. In Fig. 7.27 it can be seen that with such a calibration procedure very
good agreement can be achieved with the test data and the complex behaviour of shotcrete
at early ages can be captured in a realistic way.

271
 20

Test data
28 d
Fitted equations
16

Compressive stress [MPa]

7d

12
3d

24 h
8

12 h

4
6h
3h

0
0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5
Axial strain [%]

Figure 7.27: Parameter fitting for complete stress strain curves in compression for shotcrete
at various ages (after Sezaki et al., 1989)

The material parameters used to obtain the fitted equations in the above plots are sum-
marized in Tab. 7.7.

E28 s fcp,28 fcy,28 fcf,28 s εpcp,28 εpcf,28 εpcp,t1 εpcp,t2 t2

(MPa) (-) (MPa) (MPa) (MPa) (-) (%) (%) (%) (%) (h)
5220 0.4 17.4 4.4 16.6 0.15 0.45 0.81 2.00 1.00 24

Table 7.7: Fitted stiffness and strength parameters

It should be noted that the obtained stiffness for hardened shotcrete at 28 days, E28 ,
appears to be very low compared to the usual stiffness quoted in the literature for concrete.
However, it was not possible to find any reason for this apparent anomaly in the paper by
Sezaki et al. (1989).

7.10.3 Creep deformation in uniaxial compression

High quality data for creep of shotcrete at various stages of hydration is difficult to find in the
literature. Although shotcrete is a very creep active medium in the first hours after casting,
it seems that systematic testing to understand the principal behaviour behind creep is still
lacking. Furthermore, from the available test results it is sometimes not possible to derive
appropiate model parameters since necessary additional information regarding stiffness and
strength parameters are often missing. However, Outterside (2003) worked on the behaviour
of young shotcrete by interpreting test data from an unpublished report by Kusterle (1999)

272
commissioned by Morgan Tunnelling. In this research project, creep tests started at shotcrete
ages of around 11 hours after spraying, which was the earliest possible time, given the fact
that the shotcrete had to harden to allow measurements to be taken. Three compressive tests
were performed on wet mix specimens of 15 × 15 × 30 cm with a constant applied stress of
3.2 MPa (Test 1), 5.2 MPa (Test 2) and 7.2 MPa (Test 3). The duration of all three tests was
roughly 250 hours, after which the load was removed. Tab. 7.8 summarizes the creep tests.

Shotcrete age at Applied stress

start of loading

Test 1 11 h 3.2 MPa

Test 2 24 h 5.2 MPa
Test 3 34 h 7.2 MPa

Table 7.8: Test series performed by Kusterle (1999)

As the compressive strength increases rapidly within the first 100 hours after casting,
the stress level within the shotcrete samples would reduce significantly. Fig. 7.28 shows the
obtained test results for creep including the instantaneous deformation at the onset of loading.

0,10

Test data
0,09
Fitted equations
Test 3
0,08
Creep deformation [%]

0,07
Test 2
0,06

0,05

0,04
Test 1

0,03

0,02

0 50 100 150 200 250 300

Time [h]

Figure 7.28: Parameter fitting for development of creep deformation at various stages of
hydration (after Kusterle, 1999)

The calibrated model parameters for the time-dependent viscosity function given in equation
7.62 are summarized in the following Tab. 7.9

273
 ηo n
(kPa × h) (-)
Test 1 12 × 106 0.98
Test 2 13 × 106 0.97
Test 3 16 × 106 0.96

Table 7.9: Fitted model parameters for predicting creep deformation

By comparing these experimental data with the fitted model equations it can be seen that
by assuming a time-dependent viscosity function the model is capable of predicting the main
characteristics of shotcrete creep highlighted in Chapter 5.

7.10.4 Shrinkage
Thomas (2003) presented in his work a data collection for shrinkage from several researchers,
among them Cornejo-Malm (1995), Ding (1998) and Golser et al. (1989). The available data
consist of experimental results from various types of shotcrete, including for example wet-
and dry-mix, steelfibre reinforced shotcrete of different mix proportions. Additional test
data from other researchers that could be found in the literature were incorporated resulting
in an extensive shrinkage data collection, as illustrated in Fig. 7.29. It can be seen that
a considerable scatter can be identified. Therefore, it was decided to calibrate the model
parameters given in equation 7.71 for an upper bound, lower bound and an intermediate
curve that represents approximately an average or optimum fit. Tab. 7.10 gives the estimated

2,0

1,8 Test data

Fitted equations
1,6

1,4
Set A
Set B
Shrinkage [mm/m]

1,2

1,0

0,8

0,6

0,4
Set C
0,2

0,0
0,1 1 10 100 1000
Time [d]

Figure 7.29: Parameter fitting for development of shrinkage deformation

274
model parameters for the three different sets.

εsh
∞ B
Upper bound - Set A 1.8  12.5 d
Lower bound - Set C 1.2  167 d
Intermediate curve - Set B 1.4  27 d

Table 7.10: Fitted model parameters for shrinkage deformation

7.10.5 Temperature development during cement hydration

For the same test series on shotcrete samples, introduced in Section 7.10.1, Huber (1991) also
measured the temperature development within the specimens during the cement hydration for
the first 24 hours. Fig. 7.30 summarizes the obtained results and shows a possible parameter
fitting according to equation 7.74.

16

Test data
Fitted equation
12
Delta T [°C]

0
0 4 8 12 16 20 24 28 32
Time [h]

Figure 7.30: Parameter fitting for development of hydration temperature with time (from
Huber, 1991)

As expected, for all the test samples the maximum increase in hydration temperature
ΔTmax occurs between 9 and 12 hours after spraying. The maximum temperature rise mea-
sured was about 14.5 C. This value might not be appropriate for a real tunnel shell, where
due to different thermal boundary and mass conditions this value could be higher than for a
single shotcrete sample. However, the general trend of the temperature and thermal defor-
mation development can be assumed to be of the same nature. The adopted model equation

275
7.74, involving the fitted model parameters given in Tab. 7.11, represents a good simulation
for the obtained test data and relatively good agreement is achieved.

ΔTmax tmax tzero

12 C 11 h 43 h

Table 7.11: Fitted model parameters for temperature history

7.11 Model validation

In the previous section it was shown that it is possible to calibrate the adopted equations of
the presented constitutive model against experimental test data for various types of shotcrete
found in the literature. These calibrations covered mainly uniaxial stress-strain curves in
compression including the increase in stiffness and strength, creep deformation at early ages,
volume changes due to shrinkage and the temperature history during cement hydration.
In the current section, the aim is to prove that the robust numerical implementation of
the constitutive model as presented in Section 7.9 was performed correctly and that model
predictions from ICFEP coincide with analytical solutions. This procedure is carried out by
taking up some of the obtained model parameters from the previous section and applying
them to single element runs in ICFEP.

7.11.1 Increase in stiffness with time

The development of the Young’s modulus E with time was numerically tested for ten different
values of the cement parameter s, which governs the increase in stiffness with time. The
chosen values for s ranged from 0.1 to 1.0 and the shotcrete had an initial age of to = 1 h.
Fig. 7.31 illustrates the curves from the analytical solution (i.e. equation 7.54) compared to
model predictions from ICFEP. The values of the stiffness at various shotcrete ages are
normalised by the stiffness of hardened shotcrete at 28 days, E28 , resulting in unity at t28
for all the curves. The influence of the cement parameter s can be clearly seen. Values for s
close to 1.0 result in a very slow increase of the Young’s modulus during the first hours after
spraying. In contrast, a value for s of 0.1 already predicts a stiffness of a bit less than 30 %
of the full stiffness E28 at the beginning of the test for to = 1 h.

276
 1,0

0,9

Normalised Young's modulus [-]

0,8 t28
s = 0.1
0,7

0,6

0,5

0,4
s = 1.0
0,3

0,2
Analytical solution
0,1
ICFEP
0,0
1 10 100 1000
log time [h]

Figure 7.31: Comparison of model predictions with analytical solutions for increase in stiffness
with time

7.11.2 Uniaxial loading in compression and tension

In Section 7.10.2 some experimental data for uniaxial compression tests at various shotcrete
ages carried out by Sezaki et al. (1989) were presented and appropriate model parameters
derived. During the testing of the numerical implementation these model parameters were
applied to single element runs for both compression and tension. Since for the tensile be-
haviour of young shotcrete no particular set of data could be found in the literature, the
necessary model parameters were estimated in such a way that they would represent realistic
material behaviour as could be expected for shotcrete. The simulations focused mainly on
uniaxial loading conditions performed in a deformation controlled way in order to capture
also the post peak behaviour of shotcrete.

Fig. 7.32 shows the uniaxial stress-strain curves for shotcrete in compression at different
stages of cement hydration. It is shown that the implemented equations for the increase in
the compressive strength parameters (yield, peak and failure) and for the development of the
plastic strain limits (peak and failure) are correct (i.e. they agree with the test data, see also
Fig. 7.27) and model predictions agree well with analytical solutions.

277
 20

Analytical solution
28 d ICFEP
16

Compressive stress [MPa]

7d

12
3d

24 h
8

12 h

4 6h

3h
0
0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5
Axial strain [%]

Figure 7.32: Comparison of model predictions with analytical solutions for stress-strain curves
at different shotcrete ages

For clarity, the drop in strength from the failure strength down to the residual strength
at ultimate level was omitted in the above figure. However, in Fig. 7.33 the complete stress
strain curve for a shotcrete at an age of 6 h and 28 d is illustrated, including the ultimate
stress regime of zone 4 of the normalised hardening/softening curve. It is shown that the
increase in ultimate strength and the reduction in the ultimate plastic strain with time follow
the same equation as for the other involved strength and strain limit values.

278
 20

Analytical solution
fcf
ICFEP
16

Compressive stress [MPa]

12

fcu
28 d fcf
4

fcu
6h
0
0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5 5,0
Axial strain [%]

Figure 7.33: Comparison of model predictions with analytical solutions for complete stress-
strain curves at two different shotcrete ages

As mentioned earlier, test data for shotcrete at early ages under tensile stresses are hard
to find in the literature. Therefore, some assumptions had to be made during the numerical
testing procedure of the implemented tensile behaviour. It is well reported that the ten-
sile strength of concrete is about 10 % of the compressive strength showing little hardening
behaviour before reaching peak (see Chapter 5). The post peak response during a tensile
test is a highly discrete problem mainly goverened by concrete composition and the testing
equipment. However, the following tensile model parameters summarised in Tab. 7.12 have
been adopted in the series of numerically performed tests.

fty ftp ftu εptp εptu ψ

(MPa) (MPa) (MPa) (%) (%) (-)
1.57 1.74 0.1 0.01 0.15 -0.2

Table 7.12: Adopted model parameters for tensile behaviour

The cement parameter s, which controls the development of the strength parameters, was
chosen to be the same as obtained previously in Section 7.10.2 (s = 0.15). For the post
peak regime, softening branches following either a linear or exponential function were tested.
Fig. 7.34 shows the tensile uniaxial stress-strain curves for shotcrete at various ages, starting
at an early age of 3 h adopting the linear reduction in tensile stresses after peak according to
option A.

279
 2,0

Analytical solution
28 d ICFEP
1,6

Tensile stress [MPa]

1,2

Option A - Linear softening

3d
0,8

0,4
12 h

3h
0,0
0,0 0,1 0,2 0,3 0,4 0,5 0,6
Axial strain [%]

Figure 7.34: Comparison of model predictions with analytical solutions for stress-strain curves
in tension for various shotcrete ages (Option A)

If an exponential softening branch is chosen according to option B, the softening parameter

ψ controls the decay in tensile stresses with further straining, as can be seen in Fig. 7.35.

2,0

Analytical solution
28 d
ICFEP
1,6
Tensile stress [MPa]

1,2
3d
Option B - Exponential softening
0,8

0,4
3h
12 h
0,0
0,0 0,1 0,2 0,3 0,4 0,5 0,6
Axial strain [%]

Figure 7.35: Comparison of model predictions with analytical solutions for stress-strain curves
in tension for various shotcrete ages (Option B)

From both figures it can be seen that, similar to the mechanical behaviour in compression,
shotcrete shows a transition from a relatively ductile to a stiff and quasi-brittle material
response at 28 days. During the increasing curing time, the fracture energy, interpreted as

280
the area below the stress-strain curve, starts to grow in a similar manner as the strength
parameters.

7.11.3 Biaxial loading in compression

One important characteristic of the current constitutive model is the fact that under biaxial
stress confinement shotcrete can exhibit higher compressive strength values than under uni-
axial conditions. This increase in compressive strength is governed by the biaxial strength
parameter e and is usually in the range of up to 25 % for stress conditions of σ1 = σ2 . In the
course of the numerical testing of the implemented constitutive model a value of e = 1.15
was adopted, which results in biaxial strength values that are 15 % higher than the uniaxial
strengths. It is assumed that this biaxial strength ratio remains constant during the cement
hydration and therefore the biaxial strength increases in the same manner as in the uniaxial
case. Fig. 7.36 illustrates this mechanical behaviour for two numerical tests for shotcrete at
an age of 6 h and 28 d. Both tests were performed with axisymmetric boundary conditions
and were deformation controlled, in order to investigate the post peak behaviour until the
ultimate stress regime is reached.

22

20 Analytical solution - uniaxial

Analytical solution - biaxial
18
ICFEP
Compressive stress [MPa]

16
28 d
14

12
Biaxial loading for V V2
10

6
6h
4

0
0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5 5,0
Axial strain [%]

Figure 7.36: Comparison of model predictions with analytical solutions for uniaxial and
biaxial compressive loading at a shotcrete age of 6 h and 28 d

By comparing the obtained results with uniaxial test runs it can be clearly seen that all
the strength values (yield, peak, failure and ultimate) are increased. Furthermore, strain
limits appear to be slightly larger than for the uniaxial stress strain curves, although the
same plastic strain parameters from the previous sections were adopted. This results from
the fact that the elastic proportion of the total strains is increased at all stages of cement
hydration. Again, the numerical results from ICFEP coincide with the analytical solutions

281
and it is shown that the numerical implementation was carried out correctly.

7.11.4 Creep
In Section 7.10.3 three uniaxial creep tests in compression, performed by Kusterle (1999),
were introduced and appropriate model parameters derived. For the numerical testing of the
implemented creep subroutine, the parameter set of Test 1 was taken and applied to further
creep simulations with ICFEP. Six numerical creep tests in compression were carried out with
a different shotcrete age at the beginning of loading, to . The following Tab. 7.13 summarizes
the test series:

Test 1 Test 2 Test 3 Test 4 Test 5 Test 6

Initial shotcrete age to 6h 24 h 48 h 96 h 168 h 336 h

Table 7.13: Initial shotcrete ages for performed creep tests

The duration of each creep test was 200 hours. Fig. 7.37 shows the specific creep per unit
stress (see equation 7.63), comparing the analytical solutions with the obtained numerical
results from ICFEP. The applied stress is not of importance for the specific creep and is
therefore not mentioned here. It can be observed that the young shotcrete in test T 1 is very
creep active, predicting relatively high creep rates. These creep rates reduce significantly with
increasing shotcrete age, resulting in a lower creep magnitude for test T 6. It should be pointed
out that with increasing shotcrete age, the creep curves become more and more straight, which
is probably not what would be expected in a real creep test for almost hardened shotcrete.
The limited number of creep parameters within the viscosity function given in equation 7.62
(ηo and n) seems not to be appropriate to cover the complete mechanical creep phenomenon
for the whole time range of the cement hydration process. However, the presented constitutive
model appears to be able to predict well the creep behaviour of relatively young shotcrete
within the first couple of days after spraying. For hardened shotcrete at 28 days, with a
different set of model parameters a better simulation of creep deformation is possible.

282
 0,014
Analytical solution
0,012 ICFEP
T1

Specific creep per MPa [%]

0,010
T2

T3
0,008
T4
T5
0,006
T6
0,004

0,002

0,000
0 50 100 150 200 250 300 350 400 450 500 550 600
Time [h]

Figure 7.37: Comparison of model predictions with analytical solutions for creep deformation
(n = 0.98)

In a second series of numerical creep simulations, a different set of model parameters was
adopted. The initial shotcrete ages were the same as in the first test series, but the value
of the viscosity exponent was now changed to n = 1.3. The initial viscosity ηo was kept the
same as in the previous test series. Fig. 7.38 shows the analytical solutions compared with
model predictions obtained from ICFEP.

0,0030
Analytical solution
ICFEP
0,0025
Specific creep per MPa [%]

T1
0,0020

0,0015
T2
T3
0,0010 T4
T5
T6
0,0005

0,0000
0 100 200 300 400 500 600
Time [h]

Figure 7.38: Comparison of model predictions with analytical solutions for creep deformation
with n > 1

It can be clearly seen that the viscosity exponent n is of crucial importance for creep

283
analysis. Tests T 1 to T 3 predict relatively high creep strains at the beginning of the tests
directly after loading, as one would expect. However, after this initial steep increase in creep,
deformations reach a maximum value and reduce with continuation of the test. This means
that negative creep would be induced which is not realistic at all. Creep curves for test T 4
and T 5 do not show this phenomenon and indicate a reasonable behaviour and development
of deformations. From these observations it can be concluded that during model calibration
the creep viscosity parameter n has to be treated with caution. A value of n > 1 might
be appropriate for analysing old shotcrete resulting in better agreement with test data, but
should not be adopted in an analysis where the shotcrete is given a very small initial age to .

7.11.5 Shrinkage
The numerical testing of the adopted shrinkage model involved a small parametric study for
the shrinkage parameter B. This parameter was varied between 10 and 200 d and an ultimate
shrinkage value of εsh
∞ = 0.1 % was chosen. Furthermore, the shotcrete was given an initial
age of to = 1 h and shrinkage was predicted for a time of up to 200 days. Fig. 7.39 illustrates
the different shrinkage curves obtained by ICFEP compared with analytical solutions.

1,1

1,0
Ultimate shrinkage
0,9
B = 10
0,8
Shrinkage [mm/m]

0,7

0,6

0,5

0,4

0,3 B = 200
0,2
Analytical solution
0,1
ICFEP
0,0
0 20 40 60 80 100 120 140 160 180 200
log time [d]

Figure 7.39: Comparison of model predictions with analytical solutions for shrinkage defor-
mation

It should be noted that in this figure, the shrinkage strain is plotted as positive, although
implementation into ICFEP was performed as a negative (i.e. compressive) strain resulting
in contraction. The lowest value of B = 10 d gives a relatively steep increase in shrinkage
deformation directly after spraying. This behaviour is not so pronounced for increasing values
of B (25, 50, 75, 100, 125, 150 and 175 d). As expected, the highest value of B = 200 d predicts
a slow, almost linear, development of shrinkage and volume reduction. One particular point

284
to mention is that, although a value of 0.1 % was adopted for the ultimate shrinkage value,
all the curves are approaching a slightly different value since shrinkage deformation starts
to develop at an initial age of to = 1 h. Therefore, the ultimate shrinkage for all the model
predictions is reduced by the corresponding shrinkage strain at the initial age, εsh (t = 1 h).

7.11.6 Thermal strains

Two different hydration temperature histories were tested in the course of the numerical
implementation of the present constitutive model into ICFEP. The first temperature devel-
opment follows the experimental data on shotcrete samples published by Huber (1991) and
the fitted model parameters were discussed in Section 7.10.5. For the second numerical test
run the following parameters in Tab. 7.14 were chosen:

ΔTmax tmax tzero

25 C 11 h 96 h

Table 7.14: Model parameters for second test series of temperature history

By specifying a value for the coefficient of thermal expansion of αth = 10 × 106 / C, the
temperature induced strain histories illustrated in Fig. 7.40 are obtained.

0,16

Analytical solution
0,12
ICFEP

0,08
Thermal strain [‰]

0,04

0,00
tmax Test 1
-0,04
tzero tzero
-0,08
Test 2
-0,12

-0,16
1 26 51 76 101 126 151
Time [h]

Figure 7.40: Comparison of model predictions with analytical solutions for hydration tem-
perature induced strains

As in the case of shrinkage, temperature deformations start to build up at an initial

age for shotcrete of to = 1 h increasing rapidly up to peak at a time of t = 11 h. After
reaching the maximum rise in temperature, the shotcrete starts to contract again, resulting

285
in negative (tensile) strains due to the fact that the strain at the initial age of 1 h, εth (t = 1 h)
has to be subtracted. The developed strains during this cooling phase of the shotcrete are
of crucial importance regarding the possible cracking behaviour of the material in the case
of restrained conditions. After reaching the time tzero there is no further influence of the
hydration temperature for both test runs and deformations remain constant.

7.12 Summary
In this chapter a sophisticated constitutive model for the behaviour of shotcrete has been
presented, which is a modification of the well known Chen & Chen (1975) concrete model.
After a robust numerical implementation into ICFEP, the model was calibrated and validated
against experimental data for young shotcrete available in the literature. The following key
features of the model can be summarised:

 The mechanical behaviour of shotcrete in compression and tension is controlled by two

yield surfaces that move independently from each other in stress space. This expansion
and contraction is controlled by plastic strain hardening/softening and the increase in
strength with time during cement hydration. Cracking of the shotcrete is considered
within the smeared crack concept.

 The hardening and softening of the material during an analysis follows uniaxial stress-
strain curves that are based on typical experimental tests. Due to the time-dependent
material behaviour, normalised hardening parameters for both compression and tension
have been adopted in order to model more realistically complex loading scenarios with
respect to time.

 The main material parameters of the model are time-dependent and account for the
increase in stiffness and strength with time. Furthermore, the increased deformability
of young shotcrete at early ages was taken into account by assuming a time-dependent
function for the plastic peak strains in compression and tension.

 Creep of shotcrete at various stages of cement hydration is simulated by a Newton

dashpot which is characterised by a non-linear function for the shotcrete viscosity.
This uniaxial creep law is extended into 3D with the help of creep potential functions
for both compression and tension.

 Shrinkage and temperature induced deformations during cement hydration are consid-
ered as irreversible volumetric components of the total strains. Their development with
time is based on standard functions that can be found in the literature.

286
Chapter 8

Analysis of tunnel excavation in

London Clay

8.1 Introduction
The following chapter investigates the influence of the constitutive model developed for
shotcrete on tunnel lining design in a boundary value problem. Although a realistic as-
sessment of tunnel construction would require a full 3D simulation, the analyses presented
in this study were performed in 2D due to computational limitations. In order to obtain a
detailed understanding of the behaviour of a shotcrete shell during tunnel construction it was
decided not to model a real case study, but to analyse a tunnel excavation that would be
typical for the London area. Furthermore, three different types of excavation sequences have
been investigated, which are the following:

 Type 1 - Full face excavation

 Type 2 - Bench and invert excavation

 Type 3 - Sidewall drift

The sophisticated elasto-plasticity model presented in Chapter 7 was applied for modelling
the tunnel lining and results are compared with much simpler approaches that still represent
the current state-of-the-art for the simulation of shotcrete tunnel linings, i.e. linear elastic
and the Hypothetical Modulus of Elasticity (HME). Through an extensive parametric study,
varying the main parameters related to construction sequence, material behaviour and initial
stress conditions, it was possible to detect critical stages during tunnel construction and to
understand the fundamental behaviour of a shotcrete shell in a better and more detailed
way. It should be noted that some of the parameters adopted in the analyses might appear
unrealistic at first sight, but given the fact that most tunnel failures occur because of un-
expected heavy loading conditions caused by the variability in material parameters (ground
and shotcrete), these assumptions are justifiable.

287
8.2 Geometry of the problem
Fig. 8.1 illustrates the adopted geometry and ground conditions for the chosen boundary value
problem, that will be discussed in detail in the following sections.

Figure 8.1: Layout of tunnel geometry and ground conditions

The circular tunnel to be analysed within this research has an outer diameter of 8 m and
its centre is located at a depth of 20 m below the ground surface, which leads to a tunnel
overburden of 16 m. As a temporary support element, a shotcrete layer of 20 cm thickness
will be installed immediately after excavation of the ground resulting in a inner clear profile
of the tunnel of 7.60 m. Since tunnel excavation in reality is performed relatively quickly,
undrained conditions are assumed and any consolidation effects that might occur are ignored
throughout the whole analysis. Furthermore, the whole tunnel is being excavated in a 60 m
thick layer of London Clay overlying a layer of chalk, which represents a typical ground
profile in the western London area. For the bulk unit weight of London Clay a value of
γ = 20 kN/m3 has been chosen. Although in many cases under-drained pore pressure profiles
can be encountered in London (Potts & Zdravković, 2001), the groundwater conditions are
assumed to be hydrostatic with the ground water table (GWT) situated at the ground surface
level. Any further subsurface structures in the vicinity of the tunnel are not taken into account
in order to avoid interaction effects, which are not the purpose of this study.

As mentioned above in the introduction, three different excavation sequences are adopted
within this investigation. These are shown in Fig. 8.2.

288
 Figure 8.2: Three different excavation sequences for tunnel analyses

For Type 1, the tunnel is to be constructed full face and the complete circular lining ring
is installed directly after excavation of the complete tunnel cross section. Type 2 represents
a bench-invert excavation, where the upper half of the tunnel cross section is excavated first,
followed by the invert region. For stability reasons and in order to minimize surface settle-
ments, a temporary invert made of sprayed concrete with a thickness of 15 cm is considered
in the intermediate construction stage. Furthermore, the geometry of this temporary invert
shows a curvature with a radius of 10 m. Finally, excavation of the tunnel according to Type 3
starts with the construction of the left side-gallery including a temporary sidewall consisting
of a 15 cm thick shotcrete layer. Similar to Type 2, this sidewall is curved with a radius of
10 m. After excavation of the right gallery in the second step, the tunnel profile is completed
and the lining installation can be finished.

8.3 Ground conditions in London area and initial stresses

London Clay is the predominant soil formation that can be encountered in the London area
in the course of subsurface construction and usually has a thickness in the range of 50 to
150 m (Gasparre, 2005). This clay is well known to be highly overconsolidated and it is a
typical example of a stiff plastic clay showing brittle behaviour and localization of strains.
Furthermore, London Clay is characterised by natural discontinuities, such as laminations,
backs and fissures, whose engineering importance are emphasized in the open literature.
Its mechanical properties are influenced by many geological aspects, such as weathering or
erosion. Eddie et al. (2009) state that for tunnelling according to the principles of NATM,
London Clay represents an ideal medium, which is relatively easy to handle. For further
information regarding an advanced laboratory characterisation of London Clay see Gasparre
(2005).

The heavy overconsolidation of the adopted soil formation gives rise to high horizontal
effective stresses, resulting in Ko values that are usually greater than 1. It is reported that
in the upper 10 m of the London Clay Ko varies between 2 and 2.5 and this value tends to
decrease with increasing depth, falling to 1.5 at about 30 m depth. Within this research, two
different Ko -profiles were established, which are both constant with depth as can be seen in
Fig. 8.3.

289
 Figure 8.3: Adopted Ko -profiles for performed analyses

For all the performed analyses adopting excavation Type 1, the influence of both Ko -
profiles on the behaviour of the tunnel shell was investigated. However, for the tunnel exca-
vation following the construction scheme of Type 2 and Type 3, one single Ko -profile with a
value of 1.5 was adopted for simplicity reasons. The initial stresses at the beginning of each
analysis were therefore calculated in terms of effective stresses as:

σv = (γ − γw ) z and σh = Ko σv (8.1)

where σv and σh are the vertical and horizontal effective stresses respectively. z is the depth
below ground surface in metres and γw the bulk unit weight of water, which is taken as
9.81 kN/m3 .

8.4 Finite element model

All the analyses of this study were performed using ICFEP (Imperial College Finite Element
Program), the in house finite element code of the Geotechnics Section at Imperial College in
London, written and developed by Prof. Potts and recently by Dr. Zdravković. It is a sophis-
ticated and powerful finite element program for the analysis of geotechnical problems and
operates on Sun UNIX workstations and servers. As mentioned earlier, tunnel construction
was investigated by a 2D plane strain model. The undrained conditions of the London Clay
were simulated by prescribing high bulk compressibility of the pore fluid.

8.4.1 Mesh and boundary conditions

Fig. 8.4 shows the general layout of the finite element mesh used for the numerical calculations
with its boundary conditions concerning displacements.

290
 Figure 8.4: General layout of the 2D finite element mesh (without inner core)

The mesh has a total size of 200 m × 60 m and the bottom boundary is placed at the top of
the chalk layer. Along this mesh boundary, both displacements in the vertical and horizontal
directions are restricted to zero. On the right and left hand sides of the mesh vertical
displacements are permitted, but horizontal movements are not allowed to take place. The
ground surface is a stress free boundary. In general, the applied mesh consists of eight-noded
quadrilateral isoparametric elements, which are constructed in blocks. Two types of blocks
can be distuingished in ICFEP, as indicated in Fig. 8.5. As one can imagine, transitional
blocks are used for the transition of smaller to bigger elements. The Gauss Integration
points, for which the main equations are solved, are located inside the element. Reduced
(2 × 2) integration is applied in all analyses presented here.

Figure 8.5: Two types of blocks forming the finite element mesh

From Fig. 8.4 it can be seen that the model includes a surface surcharge of 200 kPa placed
over a distance of 2.5 tunnel diameters. This additional load was applied in order to inves-

291
tigate the effect of the construction of a possible surface structure on the tunnel shell and
therefore to study in detail the behaviour of the young shotcrete lining under heavy loading
conditions. It is appreciated that this is a rather hypothetical situation compared with what
might happen in reality.

Obviously, for all the three excavation Types 1 to 3, individual meshes had to be designed
to account for differences in the geometry and construction sequence. For simplicity reasons
it was decided to adapt just the area in the vicinity of the tunnel lining and keep the rest of
the finite element mesh the same for all the three excavation types. Fig. 8.6 to 8.8 illustrate
a zoom into the core of the applied meshes 1 to 3 respectively.

Figure 8.6: Zoom of mesh 1 for excavation Type 1: full face excavation

292
 Figure 8.7: Zoom of mesh 2 for excavation Type 2: bench-invert excavation

Figure 8.8: Zoom of mesh 3 for excavation Type 3: sidewall drift

Both the lining and soil were modelled by solid elements and refinement of the mesh was
performed in the zones of high and low changes in stresses and strains. Since the mechanical
behaviour of the shotcrete shell was of special interest during this research, the tunnel lining
was simulated by four elements across its thickness resulting in a quadratic element shape
with a size of roughly 5 cm × 5 cm. Therefore, a more accurate stress distribution over the
lining thickness was achieved with the help of 8 Gauss Intergration points. For the temporary

293
invert and sidewall of excavation Types 2 and 3, three elements were used to represent the
shotcrete thickness of 15 cm. It should be noted that the need for a relatively fine mesh
when simulating the tunnel lining with solid elements is a major drawback of the developed
finite element model, resulting in high computational effort. Finally, the following Tab. 8.1
summarizes the important information for all the three meshes that were used within this
study.

Number of blocks Number of elements Number of nodes

Mesh 1 1006 6840 20611

Mesh 2 + 3 1361 7804 23503

Table 8.1: Summary for mesh 1, 2 and 3

8.4.2 Modelling of construction sequence

The simulation of a particular construction sequence in a finite element code like ICFEP in-
volves the elimination and/or activation of certain elements within the finite element mesh. In
this study, tunnel construction was modelled in two dimensions and the volume loss method,
described earlier in Chapter 2, was applied in order to account for some 3D effects, such as
pre-deformation ahead of the tunnel face. Typical field measurements of tunnel construction
in the London area show that the volume loss VL usually lies in the range of 1 to 2 % (Potts
& Zdravković, 2001). Therefore, a target of 1.5 % was chosen for the current study and the
construction sequence of the three different excavation types was adapted for the linear elastic
shotcrete model in order to fufill this criterion.

8.4.2.1 Excavation Type 1: Full face excavation

Tab. 8.2 gives an overview of the construction steps adopted for excavation Type 1 (full
face). Since the Ko -profile, and therefore the initial stress conditions, highly influence the
development of the volume loss as the tunnel is excavated, the increment at which the lining
is installed is different for both sets of analyses.

Ko = 0.5 Ko = 1.5
1) Excavation Increments 1 to 10
2) Installation of lining Increment 3 Increment 5
3) Surface loading Increments 11 to 50

Table 8.2: Construction sequence for excavation Type 1

During excavation, all the elements within the tunnel, within the outer diameter of 8 m,

294
are removed and equivalent nodal forces, simulating excavation, applied at the tunnel bound-
ary. Furthermore, these forces are stepwise reduced over 10 increments. At the particular
increment where the desired volume loss is achieved, the complete lining ring is installed by
activating (i.e. constructing) the respective lining elements in one single step. The remain-
ing equivalent nodal forces are released until excavation is completed. Surface loading takes
place over 40 increments, resulting in an additional loading of 5 kPa per increment. Fig. 8.9
illustrates the adopted construction sequence.

Figure 8.9: Graphical representation of construction sequence for excavation Type 1

8.4.2.2 Excavation Type 2: Bench-invert excavation

Construction of the tunnel according to excavation Type 2 is performed numerically in several

steps as can be seen in Tab. 8.3. It should be noted that just one single Ko -profile with a
value of 1.5 is adopted for analysing this particular tunnel advance.

1) Excavation of bench Increments 1 to 12

2) Installation of upper shell Increment 7
3) Excavation of invert Increments 14 to 25
4) Installation of lower shell Increment 14
5) Surface loading Increments 26 to 65

Table 8.3: Construction sequence for excavation Type 2

The soil elements of the tunnel bench are removed over twelve increments and again the
equivalent nodal forces on the excavation boundary are reduced in a stepwise manner in or-
der to allow for some pre-relaxation ahead of the tunnel face. The installation of the upper
lining shell, including the temporary invert, takes place in one single step by constructing
the respective lining elements in increment 7. The temporary invert is then removed and
excavation of the invert region continued from increment 14 to 25. The lining is completed
by activating the lining elements belonging to the lower shotcrete shell immediately in incre-
ment 14 and the surface surcharge is applied incrementally until construction is finished in

295
increment 65. It should be noted that in increment 13 no particular numerical construction
or excavation is performed. This increment is therefore used to simulate the distance between
the tunnel face and the completion of the lining ring, that usually happens in reality a couple
of metres behind the front of the tunnel. Since it is not possible to capture a longitudinal
distance along the tunnel axis in a 2D analysis, this means that a certain amount of time Δt
will elapse during this increment, leading to a different shotcrete age of the upper and lower
shell at later construction stages. The desired volume loss is measured on the whole tunnel
face. Finally, Fig. 8.10 shows graphically the above described construction sequence for the
bench-invert excavation.

Figure 8.10: Graphical representation of construction sequence for excavation Type 2

8.4.2.3 Excavation Type 3: Sidewall drift

The numerical construction sequence for excavation Type 3 is summarized in Tab. 8.4. As in
the previous case in Section 8.4.2.2, one constant Ko -profile with a value of 1.5 was adopted
for the tunnel runs of this particular excavation type.

296
 1) Excavation of left gallery Increment 1 to 12
2) Installation of left shell Increment 6
3) Excavation of right gallery Increment 14 to 25
4) Installation of right shell Increment 14
5) Surface loading Increment 26 to 65

Table 8.4: Construction sequence for excavation Type 3

In the first step of construction the soil elements of the left side gallery are removed and
replaced by equivalent nodal forces. The excavation process, simulated by the removal of
these nodal forces, takes place over 12 increments, while the left shotcrete shell is placed in
increment 6 together with the temporary sidewall. Tunnel construction is then continued in
increment 14 by eliminating the temporary sidewall and the remaining soil elements of the
right side gallery. Here again, the shotcrete ring is completed in increment 14 by activating
the corresponding lining elements. After the complete release of the equivalent nodal forces
in increment 25, the surface surcharge of 200 kPa is placed on top of the tunnel over 40
increments. Similar to the approach adopted for excavation Type 2, increment 13 has the
purpose of simulating the longitudinal distance between the left and right side gallery and
therefore a certain amount of time Δt elapses in the adopted 2D model. Fig. 8.11 illustrates
the described construction procedure.

Figure 8.11: Graphical representation of construction sequence for excavation Type 3

297
8.5 Adopted constitutive models and material parameters
8.5.1 Tunnel lining - Shotcrete
As already highlighted in the introduction to this chapter, the aim of the performed study is to
investigate the behaviour of a shotcrete tunnel lining for various excavation types by applying
the sophisticated constitutive model for young shotcrete, that was developed Chapter 7.
Furthermore, the obtained results from these analyses are compared with other approaches
commonly used in engineering practice due to their simplicity. Hence, the following four
constitutive laws for shotcrete are adopted:
 Linear elastic (LE)

 Hypothetical Modulus of Elasticity (HME)

 Non-linear elasto-plasticity model (NL)

 Time-dependent non-linear elasto-plasticity model (TPM)

When modelling the tunnel lining as a linear elastic isotropic material, a value for the Young’s
modulus of hardened shotcrete at 28 days of E = 30 GPa has been chosen. In the case of
the Hypothetical Modulus of Elasticity (HME) a reduced stiffness of E = 7 GPa is applied,
which represents a reasonable value taken from the literature (see Pöttler (1990)). For both
types of analysis a constant Poisson’s ratio of μ = 0.2 has been used. For the non-linear
elasto-plastic model (NL) and the time-dependent non-linear elasto-plastic model (TPM) the
constitutive model described in the previous chapter was used.

The establishment of appropiate material parameters for these models is not straightfor-
ward and is discussed here in detail. Basically, two different types of shotcrete have been
used to model the tunnel lining, which are:
 Excavation Type 1 −→ Unreinforced shotcrete

 Excavation Type 2 and 3 −→ Steel-fibre reinforced shotcrete (SFRS)

With the developed constitutive model it is possible to incorporate the effect of steel fibres on
the material response by a slight modification of the normalised stress-strain curve in tension,
whereas the compressive behaviour remains unaffected.

Firstly, the following strength parameters, summarised in Tab. 8.5, have been chosen to
represent shotcrete behaviour (unreinforced and SFRS) both in compression and tension.

fcp,28 fcy,28 fcf,28 fcu,28 e ftp,28 fty,28 ftu,28 s t28

(MPa) (MPa) (MPa) (MPa) (-) (MPa) (MPa) (MPa) (-) (h)

30.0 7.5 24.0 5.0 1.15 3.0 2.4 0.1 0.2 672.0

Table 8.5: Basic strength parameters for shotcrete

298
 Furthermore, Tab. 8.6 contains the applied numerical parameters needed for the descrip-
tion of the tensile yield surface:

a n T IP T OL CORT OL
(-) (-) (-) ()
3.0 0.01 0.01 1.0

Table 8.6: Numerical parameters for tension yield surface

Secondly, for the complete description of the normalised stress-strain curve both in com-
pression and tension the deformability parameters listed in Tab. 8.7 have been adopted.

εpcp,28 εpcf,28 εpcu,28 εpcp,t2 εpcp,t1 εptp,28 ψ ψSF RS t1 t2

(%) (%) (%) (%) (%) (%) (-) (-) (h) (h)
0.5 0.97 1.3 1.6 3.5 0.02 -0.47 -0.0043 1.0 24.0

Table 8.7: Basic deformability parameters for shotcrete

The choice of the compressive plastic peak strain for hardened shotcrete at 28 days, εpcp,28 ,
is in accordance with the suggested values given by Swoboda & Moussa (1992). Taking into
account the elastic strain component, the total cracking (or peak) strain of hardened shotcrete
in tension has been selected to be 5 % of the total compressive peak strain (see Byfors (1980))
throughout the whole hardening process of the cement paste. Furthermore, it is important
to note that all the above mentioned softening parameters are estimated by applying the
fracture energy approach earlier presented in Sections 6.6.1.2, 6.6.1.4, 7.3.1 and 7.3.2. The
following input parameters were involved in the mathematical calibration procedure:

 Fracture energy in compression: Gc = 6.4 Nmm/mm2 (see Vonk (1992))

 Fracture energy of unreinforced shotcrete in tension: Gf 1 = 0.064 Nmm/mm2 for

dmax = 8 mm (CEB-FIP Model Code, 1990)

 Fracture energy for SFRS in tension: GSF RS = Gf 1 +Gf 2 with Gf 2 = 6.9 Nmm/mm2
(see Thomée (2005))
√
 Equivalent length of lining elements: leq = Ae = 50 mm (see equation 6.83 after
Rots (1988))

The selection of the appropriate material parameters regarding the development of the plastic
peak strains with time is based on published data by Thomas (2003) and Sezaki et al. (1989),
as can be seen in Fig. 8.12.

299
 6

Data collection from Thomas (2003)

5 Test data from Sezaki et al. (1989)

Total compressive peak strains [%]

Adopted equation for analyses

0
1 10 100 1000
Time [h]

Figure 8.12: Development of total uniaxial compressive peak strains with time

In this figure, the variation of the total compressive peak strains is established by including
the elastic strain components for this particular shotcrete with the given strength and stiffness
parameters. It can be seen that the obtained curve represents a good fit to the test data from
the literature.

The increase in stiffness with time is accounted for by the basic shotcrete parameters
summarized in Tab. 8.8

E28 s Poisson’s ratio μ

(MPa) (-) (-)
30000 0.2 0.2

Table 8.8: Basic stiffness parameters for shotcrete

The treatment of the time-dependent material behaviour of shotcrete in this numerical

analysis requires the specification of the initial age of the young shotcrete to . In order to
capture the mechanical properties at very early ages, this parameter to was set to 1 hour. In
the case of the non-linear elasto-plasticity model (NL), where no time-dependency is included
during the analyses, the initial age is chosen to be to = 28 days and therefore all the material
properties are constant and do not change with time.

For analysing the impact of creep, shrinkage and hydration temperature on the mechanical
behaviour of the constructed shotcrete tunnel lining, the parameters given below in Tab. 8.9
were adopted for all the three excavation Types 1 to 3.

300
 ηo n εsh
∞ B ΔTmax tmax tzero
(kPa h) (-) (%) (h) ( C) (h) (h)
12 × 106 0.97 0.14 648.0 35.0 12.0 96.0

Table 8.9: Basic material parameters for creep, shrinkage and hydration temperature

These parameters were mainly taken and/or estimated from data available in the literature
(see Section 7.10 about model calibration) and should represent realistic values that one could
expect for shotcrete.

Concluding, all the above introduced material parameters for shotcrete were used as a
basic set in the numerical analysis of tunnel construction according to the three different con-
struction sequences. As already highlighted in the introduction to this chapter, a parametric
study on the main material parameters involved was performed in order to investigate the
impact of each of the components on the mechanical behaviour of the shotcrete lining during
tunnel advance. These variations in material parameters will be discussed in detail in Section
8.6 of this chapter.

8.5.2 Soil - London Clay

For modelling the mechanical behaviour of London Clay a relatively simple Mohr-Coulomb
model was adopted. However, since this constitutive model describes soil as a linear elastic,
perfectly plastic material it was decided to couple the Mohr-Coulomb plastic criterion with
a small-strain-stiffness model of the Jardine et al. (1986) type. By doing this, it is possible
to capture the highly non-linear material response of London Clay before yielding. In the
following, a brief introduction into the structure of the model is given.

The adopted small-strain-stiffness model assumes that soil behaviour is isotropic and
that the bulk and shear moduli vary with mean effective stress p and strains. The equations
defining the model behaviour can be written as:

Gtan A B B α X γ−1 sin (α X γ )

= + cos (α X γ
) − (8.2)
p 3 3 6.909

with  
E
X = log10 √d (8.3)
3C
where Gtan is the tangent shear modulus and Ed the deviatoric strain. A, B, C, α and γ
are model parameters. A similar expression for the variation of the bulk modulus is adopted,
namely:
Ktan S δ μ Y μ−1 sin (δ Y μ )
= R + S cos (δ Y μ
) − (8.4)
p 2.303

301
with  
| εvol |
Y = log10 (8.5)
T
In this equation Ktan is the tangent bulk modulus and εvol the volumetric strain. R, S, T ,
δ and μ are model parameters. In Fig. 8.13 typical variations of the secant shear and bulk
moduli for London Clay are shown, indicating the decay in stiffness with increasing straining.

Figure 8.13: Variation of the shear and bulk moduli with strain level (from Potts &
Zdravković, 1999)

For the present study, the strength parameters in Tab. 8.10 have been adopted for the
Mohr-Coulomb criterion.

Cohesion Angle of shearing resistance Angle of dilatancy

c  = 5 kPa φ  = 25 ν = 12.5

Table 8.10: Strength parameters for London Clay

The stiffness parameters for describing the non-linear small-strain-stiffness model are
summarized in Tab. 8.11. It is common to set minimum (Ed,min , εvol,min ) and maximum
(Ed,max , εvol,max ) strain limits below which and above which the moduli vary with p alone,
and not with strain. Furthermore, minimum values for G and K can be specified for the soil.

A B C α γ Ed,min Ed,max Gmin

970.0 890.0 0.001 % 1.470 0.700 1.7321 × 10−3 % 0.17321 % 3333.3 kPa

R S T δ μ εvol,min εvol,max Kmin

772.5 712.5 0.001 % 2.069 0.420 5.0 × 10−3 % 0.15 % 4000.0 kPa

Table 8.11: Stiffness parameters for London Clay

302
8.6 Tunnel analyses
In Chapter 5 it was emphasized, that experimental data on shotcrete behaviour at early
ages show a considerable scatter mainly due to the large number of influencing parameters
regarding mix design, application technique and testing procedures. Furthermore, it is even
more difficult to interpret field data from tunnel monitoring, since the adopted construction
process plays an important role. Within this research, the aim was to gain a better insight
into the behaviour of shotcrete in tunnelling through an extensive parametric study. The
parameter variation for the three excavation types focused on the following aspects:

 Initial stress conditions in the ground (for excavation Type 1)

 Elapsed time during tunnel excavation and surface loading

 Increase in stiffness and strength of shotcrete with time

 Development of shotcrete deformability with time

 Influence of creep, shrinkage and hydration temperature

8.6.1 Parametric study for excavation Type 1

Tab. 8.12 summarises the performed tunnel runs for excavation Type 1 for both Ko -profiles
simulating different shotcrete ages at the end of excavation and loading.

303
 Run Shotcrete age at Shotcrete age at
end of excavation (Inc 10) end of loading (Inc 50)
(h) (h)
TPM 1 2.0 4.0
TPM 2 4.0 8.0
TPM 3 8.0 16.0
TPM 4 12.0 24.0
TPM 5 16.0 32.0
TPM 6 24.0 48.0
TPM 7 36.0 72.0
TPM 8 48.0 96.0
TPM 9 72.0 144.0
TPM 10 96.0 192.0
TPM 11 144.0 288.0
TPM 12 168.0 336.0
TPM 13 336.0 672.0
TPM 14 672.0 1344.0

Table 8.12: Shotcrete ages for different time-runs for excavation Type 1

Regarding the possible effects of the increase in stiffness and strength with time, the
analyses listed in Tab. 8.13 were carried out. It should be noted that for these analyses time-
run TPM 6 was chosen as a reference run, since it represents a realistic tunnelling scenario.
This results in a shotcrete age of 24 h after excavation and 48 h at the end of surface loading.

304
 Run Description
STRENGTH 1 Cement parameter s = 0.1
STRENGTH 2 Cement parameter s = 0.3
STRENGTH 3 Cement parameter s = 0.4
STRENGTH 4 Cement parameter s = 0.5
STRENGTH 5 Stiffness and strength reduced by 25 %
STRENGTH 6 Stiffness and strength reduced by 50 %

Table 8.13: Tunnel runs simulating variation in stiffness and strength

The estimated reductions in stiffness and strength for the runs STRENGTH 5 and 6
represent a realistic scenario, when the design quality of shotcrete is not achieved at the con-
struction site due to rebound behaviour or other uncertainties regarding material installation.

The parametric study investigating the influence of the deformability of the shotcrete at
an early age involved the analyses summarized in Tab. 8.14 and illustrated in Fig. 8.14. Here
again, run TPM 6 serves as a reference run to compare the obtained results, and therefore
the shotcrete has an age of 24 h at the end of excavation and 48 h at the end of the surface
loading.

Run Description εpcp,28 εpcp,t1 εpcp,t2

(%) (%) (%)

STRAIN 1 Upper bound 1.20 4.50 2.50

STRAIN 2 Lower bound 0.10 2.50 0.8
STRAIN 3 Brittle behaviour 0.50 3.50 1.00
STRAIN 4 Constant strain 0.10 0.20 0.15

Table 8.14: Tunnel runs simulating the variation in the development of shotcrete deforma-
bility with time

305
 6
Data collection Thomas (2003)
Test data Sezaki et al. (1989)
5
Upper bound

Total compressive peak strain [%]

Lower bound
Brittle behaviour
4
Constant strain

0
1 10 100 1000
Time [h]

Figure 8.14: Variation of the compressive peak strains with time

Tab. 8.15 contains the main information about the performed calculations with respect to
creep, shrinkage and hydration temperature.

Run Additional elapsed time after loading (Inc 51 to 100)

(h)

CREEP (C) 672.0

SHRINKAGE (SH) 672.0
TEMPERATURE (T) 48.0
C + SH + T 672.0

Table 8.15: Tunnel runs simulating creep, shrinkage and hydration temperature

8.6.2 Parametric study for excavation Type 2

Tab. 8.16 lists the performed time-runs applying different shotcrete ages for tunnel construc-
tion following the Type 2 excavation scheme. Similar to the previous Type 1 excavation
scheme, a number of parametric studies were carried out, investigating the influence of the
rate of increase of stiffness and strength with time and of the deformability of the shotcrete.
A description of these analyses is summarised in Tab. 8.17.

306
 Run End of excavation bench (Inc 12) Increment 13 End of excavation invert (Inc 25) End of surface loading (Inc 65)
Shotcrete age of Δt Shotcrete age of Shotcrete age of Shotcrete age of Shotcrete age of
upper shell upper shell lower shell upper shell lower shell
(h) (h) (h) (h) (h) (h)
TPM 1 4.0 0.0 8.0 4.0 12.0 8.0
TPM 2 4.0 12.0 20.0 4.0 24.0 8.0
TPM 3 4.0 24.0 32.0 4.0 36.0 8.0
TPM 4 4.0 48.0 56.0 4.0 60.0 8.0
TPM 5 12.0 0.0 24.0 12.0 36.0 24.0

307
TPM 6 12.0 12.0 36.0 12.0 48.0 24.0
TPM 7 12.0 24.0 48.0 12.0 60.0 24.0
TPM 8 12.0 48.0 72.0 12.0 84.0 24.0
TPM 9 24.0 0.0 48.0 24.0 72.0 48.0
TPM 10 24.0 12.0 60.0 24.0 84.0 48.0
TPM 11 24.0 24.0 72.0 24.0 96.0 48.0
TPM 12 24.0 48.0 96.0 24.0 120.0 48.0

Table 8.16: Shotcrete ages for different time-runs for excavation Type 2
 Run Description
STRENGTH 1 Cement parameter s = 0.1
STRENGTH 3 Cement parameter s = 0.4
STRENGTH 5 Stiffness and strength reduced by 25 %
STRAIN 3 Brittle behaviour
STRAIN 4 Constant strain

Table 8.17: Tunnel runs simulating the variation of increase in stiffness and strength and
shotcrete deformability with time

It is important to note that for these parametric studies, TPM 7 of Tab. 8.16 has been
chosen as the reference run to compare results and therefore the same amount of time elapses
during the process of excavation and surface loading. The respective model parameters for
STRAIN 3 and 4 can be found in the previous Section 8.6.1.

For analysing the effect of creep, shrinkage and hydration temperature, the tunnel runs
listed in Tab. 8.18 have been performed.

Run Correspondent Additional elapsed time

time-run after surface loading

CREEP (C) TPM 7 672.0

SHRINKAGE (SH) TPM 7 672.0
TEMPERATURE (T) TPM 7 72.0
C + SH + T TPM 7 672.0

Table 8.18: Tunnel runs simulating creep, shrinkage and hydration temperature for excavation
Type 2

8.6.3 Parametric study for excavation Type 3

Since the construction sequence of excavation Type 3 is, from a numerical point of view, very
similar to the one of excavation Type 2, most of the parameters in the current investigation
are varied in the same way as presented in the previous Section 8.6.2. Tab. 8.19 contains a
review of the adopted shotcrete ages for the two staged tunnel construction with sidewall
drift.

Regarding the possible effects of the development of stiffness, strength and shotcrete de-
formability with time, creep, shrinkage and hydration temperature, the same tunnel runs

308
 Run End of excavation Increment 13 End of excavation right gallery (Inc 25) End of surface loading (Inc 65)
left gallery (Inc 12) Δt
Shotcrete age of Shotcrete age of Shotcrete age of Shotcrete age of Shotcrete age of
left shell left shell right shell left shell right shell
(h) (h) (h) (h) (h) (h)
TPM 1 4.0 0.0 8.0 4.0 12.0 8.0
TPM 2 4.0 12.0 20.0 4.0 24.0 8.0
TPM 3 4.0 24.0 32.0 4.0 36.0 8.0
TPM 4 4.0 48.0 56.0 4.0 60.0 8.0
TPM 5 12.0 0.0 24.0 12.0 36.0 24.0

309
TPM 6 12.0 12.0 36.0 12.0 48.0 24.0
TPM 7 12.0 24.0 48.0 12.0 60.0 24.0
TPM 8 12.0 48.0 72.0 12.0 84.0 24.0
TPM 9 24.0 0.0 48.0 24.0 72.0 48.0
TPM 10 24.0 12.0 60.0 24.0 84.0 48.0
TPM 11 24.0 24.0 72.0 24.0 96.0 48.0
TPM 12 24.0 48.0 96.0 24.0 120.0 48.0

Table 8.19: Shotcrete ages for different time-runs for excavation Type 3
already introduced in Tab. 8.17 and 8.18 were carried out and compared with the reference
run TPM 7.

8.7 Discussion of results

The following part of this thesis deals with the presentation and interpretation of the results
obtained from the performed analyses that were described in detail in the previous sections.
For a detailed understanding of the effects of tunnel construction the focus was therefore on
the following aspects for all three excavation sequences:

 Surface settlements and volume loss

 Hoop stress distribution and utilisation factor within shotcrete tunnel lining

 Displacements of tunnel shell

 Effective stresses and pore water pressures in the ground surrounding the tunnel

In order to work out possible critical stages during tunnel construction, it was decided to
analyse the results at various stages of construction, i.e. at the end of excavation, end of
surface loading or 28 days after completion of the tunnel. The stress distributions for the
particular cross sections of the lining indicated in Fig. 8.15 and the utilisation factor for lining
extrados or intrados were investigated.

Figure 8.15: Analysed cross sections of tunnel lining for all three excavation sequences

The intersection of the temporary invert or sidewall of excavation Type 2 and 3 with
the circular lining is of special interest, since they represent areas where high stresses could
be expected. Hence, the maximum and minimum stresses in the following area indicated in
Fig. 8.16 were investigated in detail for an intermediate construction stage.

310
Figure 8.16: Zoom into connection point of temporary invert with tunnel lining for excavation
Type 2

Horizontal and vertical displacements of particular points of the tunnel lining intrados,
as shown in Fig. 8.17, have been processed over the whole construction and loading sequence
so that a detailed study of the deformational behaviour of the lining was possible.

Figure 8.17: Investigated lining displacements for all three excavation sequences

The formulation of the constitutive model for shotcrete presented in Chapter 7 allows
the utilisation factor for a particular Gauss point to be established with the help of the
normalised hardening parameters Hc for compression and Ht for tension. Based on those two
parameters, the utilisation factor ρc can be written for a compressive stress state as:
⎧
⎪
⎨fc,n for Hc ≤ 1
ρc = (8.6)
⎪
⎩1 for Hc > 1

In a similar way, the utilisation factor ρt is given for a tensile stress state as:
⎧
⎪
⎨ft,n for Ht ≤ 1
ρt = (8.7)
⎪
⎩1 for Ht > 1

A value of unity indicates that the particular stress state is already beyond peak and that
shotcrete behaviour is goverened by strain-softening. If ρt = 1 the shotcrete is likely to crack
in tension.

311
8.7.1 Surface settlements and volume loss
8.7.1.1 Excavation Type 1

Fig. 8.18 contains the obtained results regarding the surface settlement troughs for runs LE,
HME and TPM 1 to 6 at the end of tunnel excavation, applying a Ko -profile of 0.5. It
can be seen that all runs show a similar shape resembling a typical Gaussian distribution
curve often referred to in the literature. However, due to the early age behaviour of young
shotcrete, the maximum settlements that occur directly above the axis of the tunnel differ
significantly in magnitude. As expected, due to the high lining stiffness of the linear elastic
approach (LE), settlements are relatively small, with a maximum value of 21 mm. Although
the adopted stiffness for the HME model is just 25 % of the LE-stiffness, results are relatively
close, showing a maximum settlement of 23 mm. The largest surface settlement was obtained
for run TPM 1 with almost 82 mm. This results from the low age of the young shotcrete
at the end of excavation (2 h). With increasing shotcrete age (TPM 2 to 6) the behaviour
of the lining becomes stiffer and settlements reduce drastically, from 55 to 25 mm. For the
runs TPM 7 to 14 no difference compared to the linear elastic approach was noticeable and
therefore these runs are not included in Fig. 8.18. The same observation was made for the
non-linear analysis (NL) where the material properties of shotcrete at an age of 28 days were
applied and no time-dependency exists anymore. Therefore, it can be concluded that for
hardened shotcrete the non-linear pre-yield behaviour is not affecting the surface settlement
predictions.

20
Settlement [mm]

40
LE
HME
TPM 1
60 TPM 2
TPM 3
TPM 4
TPM 5
80
TPM 6

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]

Figure 8.18: Surface settlements at end of tunnel excavation (Excavation Type 1, Ko = 0.5)

For a Ko -profile of 1.5, settlements are in general smaller in magnitude due to the high
horizontal stresses in the ground before tunnel excavation. However, Fig. 8.19 shows a similar

312
pattern for the surface settlements for the runs LE, HME and TPM 1 to 6. The linear elastic
and HME approach represent a very stiff lining behaviour with maximum surface settlements
of just 14 and 15 mm respectively. Again, run TPM 1 is governed by the low lining stiffness
at the age of 2 h and settlements increase up to 39 mm. These settlements reduce with
an increase in shotcrete age from 30 mm down to 14 mm. No noticeable difference in the
settlement behaviour has been detected for runs NL and TPM 7 to 14.
0

10
Settlement [mm]

20
LE
HME
TPM 1
30 TPM 2
TPM 3
TPM 4
TPM 5
40
TPM 6

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]

Figure 8.19: Surface settlements at end of tunnel excavation (Excavation Type 1, Ko = 1.5)

Fig. 8.20 compares the development of the volume loss VL calculated from the above
presented surface settlement troughs for both Ko -profiles. A clear general trend can be
identified, which is a reducing volume loss with increasing shotcrete age at the end of tunnel
excavation (Inc 10). Run TPM 1 predicts for both Ko -profiles extremely high values of
volume loss with 5.63 % for Ko = 0.5 and 3.84 % for Ko = 1.5. These values are clearly
above the target value of 1.5 %. If construction is performed slower, i.e. the elapsed time
during excavation increases, the volume loss reduces until no further significant difference is
obtained for runs TPM 6 and above. For runs TPM 7 to 14 the volume loss appears to be
constant, which is the same for the non-linear run NL.

313
 6

Ko = 0.5
Ko = 1.5
5

TPM 1

Volume loss [%]

4

TPM 6
2
Target value 1.5

1
1 10 100 1000
Shotcrete age at end of excavation [h]

Figure 8.20: Predicted volume loss for both Ko -profiles adopting different shotcrete ages at
the end of tunnel excavation (Excavation Type 1)

Fig. 8.21 contains a compilation of settlement troughs for analyses where the increase in
stiffness and strength with time was the main parameter under investigation. Results are com-
pared with the reference run TPM 6 showing a maximum settlement of 23 mm. STRENGTH
2 to 4 represent tunnel runs with a different cement parameter s, which controls the hardening
of the cement paste. It can be seen that an increase in s leads to higher settlements, as one
would expect, since the development of the stiffness is retarded significantly. STRENGTH
2 to 4 lead to a maximum surface settlement of 27, 35 and 44 mm respectively. As a conse-
quence, the volume loss VL for those runs is clearly above the target value, reaching 2.95 %
for run STRENGTH 4. For runs STRENGTH 5 and 6, the 28 days stiffness and strength
was reduced by 25 and 50 % respectively. This scenario results in maximum settlements that
are 10 and 32 % higher than in run TPM 6 (25 and 30 mm). Results from run STRENGTH
1 are not included in the figure, as no significant difference in the results compared to TPM
6 was noticed.

314
 0

10

Settlement [mm]
20

30
TPM 6
STRENGTH 2
STRENGTH 3
40 STRENGTH 4
STRENGTH 5
STRENGTH 6
50
-100 -75 -50 -25 0 25 50 75 100
Distance from centre [m]

Figure 8.21: Surface settlements for different developments of stiffness and strength of
shotcrete with time (Excavation Type 1, Ko = 0.5)

A similar picture is obtained for the same type of investigation adopting a Ko -profile of
1.5 as can be seen in Fig. 8.22

5
Settlement [mm]

10

15

20 TPM 6
STRENGTH 2
STRENGTH 3
25 STRENGTH 4
STRENGTH 5
STRENGTH 6
30
-100 -75 -50 -25 0 25 50 75 100
Distance from centre [m]

Figure 8.22: Surface settlements for different developments of stiffness and strength of
shotcrete with time (Excavation Type 1, Ko = 1.5)

The influence of the early age deformability of the shotcrete on the surface settlement
behaviour, adopting a Ko -profile of 0.5, is illustrated in Fig. 8.23. Maximum settlements above
the tunnel axis vary between 22.3 and 23.2 mm. This difference is even less pronounced for

315
Ko = 1.5 and can therefore be neglected.

Settlement [mm]
10

15

TPM 6
STRAIN 1
20 STRAIN 2
STRAIN 3
STRAIN 4
25
-100 -75 -50 -25 0 25 50 75 100
Distance from centre [m]

Figure 8.23: Surface settlements for different shotcrete deformabilities (Excavation Type 1,
Ko = 0.5)

Very little influence has been obtained of the time-dependent aspects of creep, shrinkage
and hydration temperature on the surface settlement behaviour during tunnel construction,
as can be seen in Fig. 8.24 for a Ko -profile of 0.5. The inclusion of the development of
shotcrete hydration temperature leads to a small reduction in volume loss and maximum
surface settlement (22.6 mm) compared to reference run TPM 6, since the lining tends to
expand slightly during the first 12 h after installation. The opposite effect has been noticed
for the run SHRINKAGE, where due to the decrease in shotcrete volume with time a surface
volume loss of 1.57 % can be expected at the end of excavation in Inc 10. Taking into account
the creep of early age shotcrete in run CREEP, does not make any difference at all for
the settlement behaviour compared to run TPM 6 and volume loss and maximum surface
settlements are identical.

316
 0

Settlement [mm]
12

16
TPM 6
CREEP
20 TEMPERATURE
SHRINKAGE
C+SH+T
24

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]

Figure 8.24: Surface settlements for analysis of creep, shrinkage and hydration temperature
of shotcrete (Excavation Type 1, Ko = 0.5)

Results are similar when adopting a Ko -profile of 1.5 with maximum surface settlements
ranging from 14.0 to 14.6 mm, as shown in Fig. 8.25.

4
Setlement [mm]

TPM 6
12 CREEP
TEMPERATURE
SHRINKAGE
C+SH+T
16
-100 -75 -50 -25 0 25 50 75 100
Distance from centre [m]

Figure 8.25: Surface settlements for analysis of creep, shrinkage and hydration temperature
of shotcrete (Excavation Type 1, Ko = 1.5)

A complete set of results for the settlement behaviour of excavation Type 1 for both
Ko -profiles can be found in the Appendix.

317
8.7.1.2 Excavation Type 2

The following Fig. 8.26 illustrates the settlement troughs obtained for the tunnel construction
according to excavation Type 2. At the end of bench excavation (Inc 12), which represents
an intermediate construction stage, the linear elastic model (LE) for shotcrete predicts a
maximum settlement of 20.5 mm. Slightly higher values are obtained when adopting non-
linearity for hardened shotcrete (NL) and the HME approach. The analysis for TPM 1 shows
the highest value of maximum settlements of 26.8 mm due to the early age properties of
shotcrete. When construction of the tunnel is continued by excavating the invert region of
the tunnel cross section, the settlement troughs for all the performed analyses except TPM
1 seem to move upwards resulting in smaller maximum surface settlements at the end of
excavation than at the intermediate stage. In addition, the shape of these settlement troughs
tends to change slightly by getting wider in such a way that the aimed volume loss of 1.5 %
is achieved for the linear elastic model (LE). For TPM 1, the maximum settlements continue
to increase during invert excavation up to a final value of 29.2 mm. TPM 5 and 9, which
represent analyses with older and consequently stiffer shotcrete, predict maximum settlements
of 16.1 and 18.2 mm respectively.

318
 0

Settlement [mm]
10

15
LE
HME
20 NL
TPM 1
TPM 5
25 TPM 9

30
-100 -75 -50 -25 0 25 50 75 100
Distance from centre
0

8
Settlement [mm]

12

16

20 LE
HME
24 NL
TPM 1
28 TPM 5
TPM 9
32
-100 -75 -50 -25 0 25 50 75 100
Distance from centre [m]

Figure 8.26: Surface settlements for various time-runs at the end of bench excavation (Inc
12, top) and end of invert excavation (Inc 25, bottom)

The reason behind performing the time runs TPM 1 to 4 was to simulate the longitudinal
distance between bench and invert excavation by adopting some amount of elapsed time Δt
(Inc 13) in the present 2D analysis. As a consequence, the upper tunnel shell has a different
shotcrete age and material properties than the lower one, which is installed at a later time of
construction (Inc 14). However, Fig. 8.27 shows that there is no influence of the intermediate
time Δt on the predicted settlement behaviour since the settlement troughs for TPM 1 to 4
are almost identical or show negligable differences. It can be concluded that it is not possible
to capture the real 3D effect of staged tunnel excavation on the settlement profiles in a 2D
analysis.

319
 0

Settlement [mm]
12

16

20

24 TPM 1
TPM 2
28 TPM 3
TPM 4
32
-100 -75 -50 -25 0 25 50 75 100
Distance from centre [m]

Figure 8.27: Surface settlements after complete excavation for runs TPM 1 to 4

The runs STRENGTH 1, 3 and 5 model different developments of the increase in shotcrete
stiffness and strength with time as summarised in Tab. 8.17. The reference run TPM 7 in
Fig. 8.28 shows a maximumm surface settlement of 18.2 mm, whereas the maximum value
of STRENGTH 1 is reduced to 15.6 mm since the stiffness and strength of the shotcrete
increase more quickly with time (s = 0.1). Maximum settlements of STRENGTH 3 reach
up to 30.1 mm due to the retarded development of the lining stiffness (s = 0.4). This results
in a relatively high volume loss of 2.78 %, which is far above the target value of 1.5 %. If
the stiffness and strength of shotcrete are reduced by 25 % (STRENGTH 5), an increase in
maximum settlements of 11 % can be expected.

320
 0

12

Settlement [mm]
16

20

24
TPM 7
28 STRENGTH 1
STRENGTH 3
32 STRENGTH 5

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]

Figure 8.28: Surface settlements after complete excavation for various developments in stiff-
ness and strength

By looking at the settlement troughs shown in Fig. 8.29, it can be concluded that the
shotcrete deformability at early ages plays only a minor role in the realistic prediction of the
settlement behaviour. Maximum settlements vary from 16.2 mm for run STRAIN 4 (constant
peak strains) to 18.2 mm for run TPM 7.

6
Settlement [mm]

12

15
TPM 7
STRAIN 3
18 STRAIN 4

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]

Figure 8.29: Surface settlements after complete excavation for various shotcrete deformabili-
ties

321
 Finally, Tab. 8.20 summarises the obtained results regarding volume loss and maximum
surface settlements for all the performed analyses for excavation Type 2. It should be noted
that only the linear elastic (LE) and the non-linear (NL) run fulfill the volume loss criterion
of 1.5 %. For very young shotcrete the volume loss can increase up to almost 2.8 %, which is
likely to be an unacceptable value.

Run Volume loss VL Maximum settlement δ

(%) (mm)
Linear elastic (LE) 1.48 14.8
HME 1.61 16.1
Non-linear (NL) 1.51 15.1
TPM 1 2.71 29.2
TPM 5 1.80 18.2
TPM 9 1.62 16.1
STRENGTH 1 1.57 15.6
STRENGTH 3 2.78 30.1
STRENGTH 5 1.97 20.2
STRAIN 3 1.67 16.8
STRAIN 4 1.62 16.2
CREEP 1.80 18.1
SHRINKAGE 1.80 18.2
TEMPERATURE 1.78 17.8
C+SH+T 1.80 18.1

Table 8.20: Volume loss and maximum surface settlement for various time-runs at completion
of tunnel excavation (Inc 25) for excavation Type 2

8.7.1.3 Excavation Type 3

Fig. 8.30 illustrates the development of surface settlements during tunnel construction ac-
cording to excavation Type 3 for various runs adopting different constitutive behaviour for
shotcrete. At the end of excavation of the left side gallery (Inc 12), all the predicted surface
settlement troughs have a relatively irregular shape with the maximum settlement occurring
to the sides of the tunnel centre line. This effect occurs due to the temporary sidewall that
is installed as an intermediate support element. At this stage of tunnel construction the
predicted settlements are in the range of 8 to 10 mm for the runs LE, HME, TPM 5 and
9. No difference to LE has been observed when adopting the non-linear model (NL). This
irregularity is not so pronounced for run TPM 1, which shows a relatively flat area of the size

322
of double the tunnel diameter directly above the tunnel axis. The maximum settlements for
TPM 1 reach up to almost 16 mm. When construction of the tunnel continues by excavating
the right side gallery, surface settlement predictions seem to follow again a distribution similar
to a Gaussian distribution curve and the maximum settlements tend to increase, which was
not the case for excavation Type 2. Due to the high lining stiffness, the linear elastic model
(LE) shows the smallest settlements with 12.6 mm at completion of tunnel excavation (Inc
25). The HME approach with its softer lining stiffness predicts slightly larger settlements
with a maximum of 14.6 mm. This result is almost identical with run TPM 9, where the
lower tunnel shell has a shotcrete age of 24 h and the upper shell 48 h. As might be expected,
TPM 1 simulating a relatively high loading rate on young shotcrete predicts the largest sur-
face settlements of up to 27.4 mm. From these analyses it can be observed again that with
increasing shotcrete age settlements tend to reduce because of the further developed lining
stiffness.

As for the the case of excavation Type 2, the elapsed time Δt in Inc 13, simulating the
distance between the face of the left and right side gallery, has no influence on the settlement
behaviour (runs TPM 2, 3, 4, 6, 7, 8, 10, 11 and 12). It is therefore not possible to capture
the full stress rearrangement by adopting different shotcrete ages for the left and right tunnel
shell in a 2D analysis.

323
 0

Settlement [mm]
8

12 LE
HME
TPM 1
TPM 5
16 TPM 9

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]
0

8
Settlement [mm]

12

16

LE
20
HME
NL
24 TPM 1
TPM 5
28 TPM 9

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]

Figure 8.30: Surface settlements for various time-runs at the end of excavation of the left side
gallery (Inc 12, top) and end of complete excavation (Inc 25, bottom) for excavation Type 3

Results differ significantly when adopting different developments of stiffness and strength
for shotcrete in runs STRENGTH 1, 3, and 5, as can be seen in Fig. 8.31. When simulating
a fast increase in lining stiffness for run STRENGTH 1 (s = 0.1), the maximum settlements
reduce by 2.7 mm compared to run TPM 7 to an absolute value of 13.9 mm. A relatively high
cement parameter of s = 0.4 for run STRENGTH 3 leads to a slow increase in stiffness and
therefore surface settlements increase significantly of up to 28.1 mm. By reducing the stiffness
and strength by 25 %, the obtained maximum surface settlements for run STRENGTH 5 are
13 % higher (18.7 mm) compared to run TPM 7.

324
 0

Settlement [mm]
12

16

20

TPM 7
24 STRENGTH 1
STRENGTH 3
28 STRENGTH 5

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]

Figure 8.31: Surface settlements after complete excavation for various developments of stiff-
ness and strength for excavation Type 3

Fig. 8.32 investigates the influence of early age shotcrete deformability on the surface
settlement behaviour at the end of tunnel excavation (Inc 25). The run STRAIN 3 represents
a very brittle material behaviour, where the compressive peak strains reduce quickly with
increasing shotcrete age (see Tab. 8.17). As a result, the maximum settlements are slightly
reduced compared to reference run TPM 7 and reach a value of 15.2 mm. When adopting a
constant compressive and tensile peak strain in run STRAIN 4, the surface settlements tend
to be smaller with a maximum value of 14.6 mm.

325
 0

Settlement [mm]
8

12

TPM 7
STRAIN 3
16
STRAIN 4

-100 -75 -50 -25 0 25 50 75 100

Distance from centre [m]

Figure 8.32: Surface settlements after complete excavation for various shotcrete deformabili-
ties for excavation Type 3

Tab. 8.21 contains a summary of the results for all the performed analyses of excavation
Type 3 regarding surface settlement behaviour. It can be seen clearly, that most of the
runs adopting young shotcrete behaviour show a relatively high volume loss of up to 2.74 %.
Runs LE, HME, NL and STRENGTH 1 predict a volume loss close to 1.5 % reaching the
target value. Maximum surface settlements vary for all runs from 12.6 mm (LE) to 28.1 mm
(STRENGTH 4).

326
 Run Volume loss VL Maximum settlement δ
(%) (mm)
Linear elastic (LE) 1.37 12.6
HME 1.53 14.6
Non-linear (NL) 1.45 13.2
TPM 1 2.66 27.4
TPM 5 1.73 16.6
TPM 9 1.54 14.4
STRENGTH 1 1.51 13.9
STRENGTH 3 2.74 28.1
STRENGTH 5 1.91 18.7
STRAIN 3 1.62 15.2
STRAIN 4 1.57 14.6
CREEP 1.72 16.4
SHRINKAGE 1.72 16.5
TEMPERATURE 1.70 16.2
C+SH+T 1.73 16.5

Table 8.21: Volume loss and maximum surface settlement for various time-runs at completion
of tunnel excavation (Inc 25) for excavation Type 3

From these observations it can be concluded that the stiffness of the shotcrete at early
ages is of crucial importance for a realistic prediction of surface settlements and volume loss.
Relatively fast tunnel excavation can lead to unacceptable surface movements during the first
24 hours, where the stiffness properties of shotcrete have not yet developed enough to form
a stiff ring. Modelling the tunnel shell as a linear elastic material (LE) leads to relatively
small surface settlements and volume loss due to the very high lining stiffness from the very
beginning of tunnel construction. These results can be slightly improved by adopting the
HME approach and a reduced stiffness. When simulating shotcrete behaviour with the time-
dependent elasto-plasticity model the correct development of the stiffness with time is a key
parameter for the realistic prediction of surface settlement troughs. Therefore, the cement
parameter s should be treated with caution. The deformability of very young shotcrete at
early ages plays a minor role in the surface settlement behaviour, where results vary within a
relatively small range for excavation Types 2 and 3. Creep, shrinkage and shotcrete hydration
temperature seem to have very little or almost no influence on the predicted settlement
behaviour during tunnel excavation. However, due to the expansion of the lining during
cement hydration the surface settlements tend to reduce slightly.

327
8.7.2 Hoop stress distribution and utilisation factor
8.7.2.1 Excavation Type 1

As explained in the introduction to this section, the stress distribution within the shotcrete
shell was investigated for several cross sections of the tunnel lining. Fig. 8.33 shows how
the obtained horizontal stresses vary across the lining thickness at the tunnel crown for
various adopted constitutive models at the end of excavation (Inc 10), assuming a Ko -profile
of 0.5. It should be noted that the lining intrados is located at 0.0 m, whereas the lining
extrados is situated at a distance of 0.20 m. It can be seen that the linear elastic model (LE)
predicts large bending of the lining with a maximum compressive stresses of -8.3 MPa and
a tensile stress of 0.58 MPa. This bending is significantly reduced for the HME approach
due to the reduced elastic stiffness with maximum and minimum compressive stresses of -
5.1 and -2.7 MPa respectively. However, the time-dependent runs TPM 1 to 6 appear to
predict almost no bending at all, and the mechanical behaviour is mainly governed by hoop
compression. Since the compressive strength is not fully developed at these early stages of
shotcrete hydration, these compressive stresses are at a relatively low level ranging from -1.3
(TPM 1) to about -4.0 MPa (TPM 6). With further increase in shotcrete age (TPM 7 to 14),
the material response of the shotcrete becomes stiffer and significant bending of the lining is
introduced, approaching the linear elastic model (LE). Non-linearity of hardened shotcrete
at 28 days (NL) does not play an important role at the end of excavation and no difference
to the linear elastic model has been identified.

328
 0,20

0,15

Cross section [m]

0,10
LE
HME
TPM 1
TPM 2
0,05 TPM 3
TPM 4
TPM 5
TPM 6
0,00
-8 -6 -4 -2 0 2
Horizontal stress [MPa]
0,20

0,15
Cross section [m]

HME
0,10 TPM 7
TPM 8
TPM 9
TPM 10
0,05 TPM 11
TPM 12
TPM 13
TPM 14
0,00
-8 -6 -4 -2 0
Horizontal stress [MPa]

Figure 8.33: Horizontal hoop stress distribution for crown at end of excavation (Inc 10) for
Ko = 0.5

These differences in lining behaviour for different constitutive models and elapsed time
become even more pronounced when loading of the shotcrete shell is continued by placing the
surface surcharge on top of the tunnel, see Fig. 8.34. The linear elastic model (LE) predicts
compressive stresses of up to -36.5 MPa and tensile stresses of 27.2 MPa, which result in a
huge and probably unrealistic bending moment. Furthermore, the shotcrete is obviously not
able to sustain such a large tensile stress and therefore cracking of the lining would occur
in the region of the crown. This large bending is reduced by applying the HME approach,
but still unrealistic tensile stresses would be induced at the lining intrados (5.0 MPa). When
adopting the non-linear approach (NL) it can be observed that stresses are changed due to the

329
non-linear hardening and softening of shotcrete at the onset of plastic deformation. On the
lining extrados compressive stresses reach a level of -20.1 MPa and are far below the uniaxial
strength. Tensile stresses on the lining intrados are restricted to 3.0 MPa by the adopted
tensile yield surface and cracking of the shotcrete takes place for almost half the width of the
cross section. The runs TPM 1 to 6, still present relatively low bending compared to the LE
and HME approach and maximum compressive stresses vary between -3.6 and -10.2 MPa. All
these runs also predict low tensile stresses at the lining intrados which are in the range of 0.2
to 1.7 MPa. Here again, with increasing shotcrete age, the material behaviour becomes more
and more brittle and a highly non-linear stress distribution can be observed, where cracking
and consequently tensile strain softening play an important role at the lining intrados (TPM
7 to 14).

330
 0,20

0,15

Cross section [m]

LE
0,10
HME
NL
TPM 1
TPM 2
0,05 TPM 3
TPM 4
TPM 5
TPM 6
0,00
-40 -30 -20 -10 0 10 20 30
Horizontal stress [MPa]
0,20

0,15
Cross section [m]

HME
0,10 TPM 7
TPM 8
TPM 9
TPM 10
0,05 TPM 11
TPM 12
TPM 13
TPM 14
0,00
-21 -18 -15 -12 -9 -6 -3 0 3 6
Horizontal stress [MPa]

Figure 8.34: Horizontal hoop stress distribution for crown at end of loading (Inc 50) for
Ko = 0.5

The effect of the development of stiffness and strength on the stress distribution within
the tunnel lining can be studied in Fig. 8.35, which illustrates the vertical stresses at the
right tunnel sidewall at the end of excavation, adopting a Ko -profile of 0.5. In this plot,
the lining intrados is again located at 0.0 m, whereas the lining extrados is situated at a
distance of 0.20 m. An interesting fact is that almost all of the performed analyses predict
mainly hoop compression without any bending at all, apart from run STRENGTH 1. For
this particular run an accelerated stiffness and strength increase is assumed (s = 0.1) and
therefore stresses are at a higher level compared to the reference run TPM 6 (-4.7 MPa). The
lowest compressive stress is predicted by run STRENGTH 4 (-3.7 MPa) due to the retarded

331
stiffness and strength development as a result of adopting the cement parameter s = 0.5. A
reduction of the stiffness and strength by 25 and 50 % (runs STRENGTH 5 and 6) leads to
a slightly smaller compressive stress ranging between -4.0 and -5.0 MPa.

TPM 6
-1 STRENGTH 1
STRENGTH 2
STRENGTH 3
-2
Vertical stress [MPa] STRENGTH 4
STRENGTH 5
-3 STRENGTH 6

-4

-5

-6

-7
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.35: Vertical hoop stress distribution in right sidewall at end of excavation (Inc 10)
for various developments of stiffness and strength adopting Ko = 0.5

A similar observation can be made when looking at the horizontal hoop stresses in the
tunnel invert at the end of surface loading (Inc 50), adopting a Ko -profile of 0.5, as illustrated
in Fig. 8.36. Here, the lining extrados is located at 0.0 m and the lining intrados at a cross-
sectional distance of 0.2 m. At the end of the whole construction process (excavation and
surface loading) the stiffness of reference run TPM 6 is already developed such that some
bending has been introduced, predicting compressive stresses of -7.8 MPa and -1.2 MPa. The
higher stiffness characterising run STRENGTH 1 leads to an increased bending of the invert
with compressive stresses of -9.3 MPa and tensile stresses of 1.8 MPa, that could result in
cracking. It is worth mentioning that even runs STRENGTH 3 and 5, which represent a
retarded increase in stiffness and strength, show some tensile stresses of 0.3 and 0.6 MPa
respectively at the lining intrados. Bending of the invert is significantly decreased for run
STRENGTH 6, where the shotcrete stiffness and strength were reduced by 50 %.

332
 0,20

0,15

Cross section [m]

0,10
TPM 6
STRENGTH 1
STRENGTH 2
0,05 STRENGTH 3
STRENGTH 4
STRENGTH 5
STRENGTH 6
0,00
-10 -8 -6 -4 -2 0 2
Horizontal stress [MPa]

Figure 8.36: Horizontal hoop stress distribution in invert at end of loading (Inc 50) for various
developments of stiffness and strength adopting Ko = 0.5

The impact of early age shotcrete deformability on the stress formation within a tunnel
lining has been investigated in Fig. 8.37, where the horizontal stresses in the tunnel crown
at the end of surface loading (Inc 50) are illustrated, adopting a Ko -profile of 0.5. The
compressive stresses at the lining extrados vary over a wide range from -9.0 MPa for the
run STRAIN 1 (upper bound of compressive peak strains) to -18.9 MPa for run STRAIN 4.
Even for the tensile stresses on the lining intrados a significant difference can be identified.
Due to the constant low strain level of the peak strains for run STRAIN 4, half of the cross
section appears to be cracked and stresses drop down to the ultimate tensile strength. Strain
softening occurs also for run STRAIN 3, representing very brittle material behaviour of the
shotcrete. Due to the early age of the shotcrete in runs STRAIN 1 and 2, the full tensile
strength of 3.0 MPa is not reached completely and tensile stresses take a value of 1.60 MPa.
It can be concluded that, for a tunnel lining under heavy loading conditions, the admissible
strain levels at early ages play an important role particularly for the cracking behaviour and
lead to highly non-linear stress distributions over the cross section.

333
 0,20

0,15

Cross section [m]

0,10

TPM 6
STRAIN 1
0,05
STRAIN 2
STRAIN 3
STRAIN 4

0,00
-18 -15 -12 -9 -6 -3 0 3
Horizontal stress [MPa]

Figure 8.37: Horizontal hoop stress distribution in crown at end of loading (Inc 50) for various
developments of shotcrete deformability adopting Ko = 0.5

Fig. 8.38 and 8.39 contain the results for the analysis of shotcrete creep, shrinkage and
hydration temperature during tunnel construction, adopting a Ko -profile of 0.5. Here again,
run TPM 6 was taken as a reference run, but an additional amount of elapsed time was
added after the end of surface loading from Inc 51 to 100 (see Tab. 8.15). Fig. 8.38 shows the
vertical stress distribution in the right tunnel sidewall at the end of Inc 100 and results are
compared with those from run TPM 6 after Inc 50. It can be seen that, due to shotcrete
creep at relatively early ages, the bending of the cross section is slightly reduced with max-
imum compressive stresses of -9.9 MPa at the lining intrados (0.0 m) and -5.9 MPa at the
lining extrados (0.2 m). This result is surprising, since creep of young shotcrete is usually
considered to be of significant influence on the stress distribution within a shotcrete tunnel
lining. Due to the development of heat during cement hydration, compressive stresses at
the lining extrados reduce to -4.6 MPa and slightly increase up to -10.6 MPa at the lining
intrados. The incorporation of shotcrete shrinkage shows almost no effect and a very similar
stress distribution compared to run TPM 6 was obtained.

334
 -4,5 TPM 6
CREEP
TEMPERATURE
-6,0
SHRINKAGE
C+SH+T

Vertical stress [MPa]

-7,5

-9,0

-10,5

0,00 0,05 0,10 0,15 0,20

Cross section [m]

Figure 8.38: Vertical hoop stress distribution in right tunnel sidewall for analysis of creep,
shrinkage and hydration temperature adopting Ko = 0.5

In Fig. 8.39 the horizontal hoop stresses at the tunnel crown are illustrated for the same
type of analyses as before. Once again, shotcrete creep leads to a slight reduction in bending
of the cross section with a compressive stress of -9.2 MPa at the lining extrados (0.2 m) and a
tensile stress of 0.5 MPa at the lining intrados (0.0 m). One important observation that can
be made is that the early age shrinkage of shotcrete results in cracking of the shotcrete shell
at the lining intrados and strain softening is introduced.

0,20

TPM 6
CREEP
TEMPERATURE
0,15 SHRINKAGE
C+SH+T
Cross section [m]

0,10

0,05

0,00
-12 -10 -8 -6 -4 -2 0 2
Horizontal stress [MPa]

Figure 8.39: Horizontal hoop stress distribution in tunnel crown for analysis of creep, shrink-
age and hydration temperature adopting Ko = 0.5

335
 In Fig. 8.40 the vertical stresses within the right tunnel sidewall are plotted for various
amounts of elapsed time during excavation and loading and compared with the simple analyses
LE and HME, adopting a Ko -profile of 1.5. Once again, the linear elastic model (LE) predicts
large bending of the cross section with compressive stresses of -0.90 MPa at the lining intrados
and -6.70 MPa at the lining extrados. This bending mechanism is significantly reduced when
simulating shotcrete with the HME approach where the compressive stresses range from -
3.2 to -4.4 MPa. It can be seen that in this case the HME approach represents a good
approximation for the time-dependent runs TPM 3 to 6. Run TPM 1 is characterised by a
very low stiffness and strength and therefore the mechanical behaviour of the tunnel sidewall is
goverened mainly by hoop compression without any bending (-1.0 MPa). The same conclusion
applies for run TPM 2 although compressive stresses are already slightly larger with a value
of -2.3 MPa. Runs TPM 7 to 14 show a relatively stiff behaviour of the shotcrete due to their
increased shotcrete age and overestimate the bending of the cross section compared to the
HME approach. Compressive stresses appear to vary linearily over the cross section from
-1.1 to -6.5 MPa for TPM 14.

336
 0

-2

Vertical stress [MPa]

-4
LE
HME
TPM 1
TPM 2
-6
TPM 3
TPM 4
TPM 5
TPM 6
-8
0,00 0,05 0,10 0,15 0,20
Cross section [m]
0

-2
Vertical stress [MPa]

-4
HME
TPM 7
TPM 8
TPM 9
-6 TPM 10
TPM 11
TPM 12
TPM 13
TPM 14
-8
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.40: Vertical hoop stress distribution for right sidewall at end of excavation (Inc 10)
for Ko = 1.5

When looking at the vertical hoop stress distribution in the tunnel sidewall at Inc 50, it
is very important to highlight that during the surface loading a change in loading direction
takes place. This mechanism is illustrated in Fig. 8.41.

337
Figure 8.41: Change in bending direction of tunnel lining due to tunnel excavation with
initial stresses for Ko = 1.5 and surface loading

In Fig. 8.42 the compressive stresses for the linear elastic model (LE) vary from -13.3 MPa
on the lining intrados to -0.7 MPa at the lining extrados. Including non-linearity for hard-
ened shotcrete at 28 days (NL) changes this stress distribution slightly and leads to smaller
compressive stresses at the lining intrados due to plastic deformation. However, runs TPM 1
to 6 still show hardly any bending of the cross section at all and stresses vary from -3.5 MPa
(TPM 1) to -6.4 MPa (TPM 6). The HME approach predicts some reduced stresses compared
to the linear elastic model (LE) due to its lower stiffness. The non-linear distribution of the
lining stresses becomes much more pronounced for runs TPM 7 to 14, where results differ
significantly from each other with compressive stresses ranging from -7.7 to -10.8 MPa at the
lining intrados and from -1.1 to -4.5 MPa at the lining extrados. The HME approach can be
seen as a good approximation for runs TPM 7 and 8, but TPM 14 clearly overpredicts the
bending of the cross section.

338
 0

-3

Vertical stress [MPa]

-6

LE
HME
-9 NL
TPM 1
TPM 2
-12 TPM 3
TPM 4
TPM 5
TPM 6
-15
0,00 0,05 0,10 0,15 0,20
Cross section [m]
0

HME
TPM 7
-2
TPM 8
TPM 9
TPM 10
Vertical stress [MPa]

-4
TPM 11
TPM 12
-6 TPM 13
TPM 14

-8

-10

-12
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.42: Vertical hoop stress distribution for right sidewall at end of loading (Inc 50) for
Ko = 1.5

Once again, the influence of the development of the shotcrete stiffness and strength with
time has been studied in detail for the analyses adopting a Ko -profile of 1.5, as shown in
Fig. 8.43. Results from various analyses have been compared with reference run TPM 6.
Run STRENGTH 1 predicts relatively large bending of the right tunnel sidewall at the end
of excavation in Inc 10, due to the accelerated stiffness and strength increase controlled by
the cement parameter s = 0.1. For this analysis stresses vary from -1.8 MPa at the lining
intrados (0.0 m) to -5.8 MPa at the lining extrados. Runs STRENGTH 2 to 6 show almost
no bending of the cross section and stresses are generated mainly as hoop compression. Run
STRENGTH 4, which has a cement parameter of s = 0.5 predicts very small compressive

339
stresses of -2.8 MPa. These stresses are slightly larger for runs STRENGTH 2 and 3, since
their development of shotcrete stiffness and strength is happening faster due to their cement
parameters s = 0.3 and s = 0.4. Reducing both the stiffness and strength of the shotcrete by
25 % in run STRENGTH 5 leads to a reduction in compressive stresses of 10 % at the lining
extrados, whereas this result is reversed at the lining intrados. A very similar trend applies
also to run STRENGTH 6, where a stiffness and strength reduction of 50 % was adopted.

-1

-2
Vertical stress [MPa]

-3

-4
TPM 6
STRENGTH 1
-5 STRENGTH 2
STRENGTH 3
-6
STRENGTH 4
STRENGTH 5
STRENGTH 6
-7
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.43: Vertical hoop stress distribution in right sidewall at end of excavation (Inc 10)
for various developments of stiffness and strength adopting Ko = 1.5

Fig. 8.44 focuses on the horizontal stress distribution in the tunnel invert at the end of
surface loading (Inc 50), for analysis adopting Ko = 1.5. The same conclusions can be
drawn as before: a retarded or reduced increase in shotcrete stiffness and strength leads to a
softer lining behaviour where bending of the cross section is reduced and smaller compressive
stresses are generated. In this figure the lining extrados is located at 0.0 m and the lining
intrados is situated at a cross-sectional distance of 0.2 m. Run STRENGTH 4 predicts the
lowest compressive stresses that vary almost linearily over the cross section with a value of
-1.4 MPa at the lining intrados and -6.4 MPa at the lining extrados. The lowest bending of the
cross section is obtained by run STRENGTH 6, which has a reduced stiffness and strength
by 50 %. On the lining intrados, results for the runs TPM 6 and STRENGTH 1, 5 and 6 are
almost identical reaching about -3.0 MPa.

340
 0,20

0,15

Cross section [m]

0,10

TPM 6
STRENGTH 1
STRENGTH 2
0,05
STRENGTH 3
STRENGTH 4
STRENGTH 5
STRENGTH 6
0,00
-8 -6 -4 -2 0
Horizontal stress [MPa]

Figure 8.44: Horizontal hoop stress distribution in invert at end of loading (Inc 50) for various
developments of stiffness and strength adopting Ko = 1.5

For a Ko -profile of 1.5, the difference in results regarding stress distributions over the
lining thickness due to early age deformability of the shotcrete is not so pronounced as for
Ko = 0.5 (see Fig. 8.37). Fig. 8.45 contains the cross-sectional stress distributions for the
horizontal stresses at the tunnel crown for various analyses at the end of surface loading
in Inc 50. Compared to reference run TPM 6, runs STRAIN 2 and 3, which represent a
relatively brittle shotcrete behaviour, show slightly incrased bending of the tunnel crown,
with tensile stresses at the lining intrados (0.0 m) of 1.1 MPa and compressive stresses at the
lining extrados (0.2 m) of -8.0 MPa. The increased straining capacity of run STRAIN 1 (upper
bound of peak strains) makes almost no difference and results are close to those of run TPM
6. When adopting a constant low strain limit for the compressive and tensile peak strains in
run STRAIN 4, stresses rise up significantly and may lead to cracking on the lining intrados
with a maximum tensile stress of 1.7 MPa. On the lining extrados, compressive stresses reach
a value of almost -10.0 MPa and are therefore about 40 % larger than for TPM 6.

341
 0,20

TPM 6
STRAIN 1
STRAIN 2
0,15
STRAIN 3
STRAIN 4

Cross section [m]

0,10

0,05

0,00
-10 -8 -6 -4 -2 0 2
Horizontal stress [MPa]

Figure 8.45: Horizontal hoop stress distribution in crown at end of loading (Inc 50) for various
developments of shotcrete deformability Ko = 1.5

The influence of creep, shrinkage and hydration temperature of early age shotcrete on the
mechanical behaviour of the tunnel lining can be studied in Fig. 8.46 and 8.47, for analysis
adopting a Ko -profile of 1.5. When looking at the vertical stress distribution in the right
tunnel sidewall at Inc 100 (28 days after completion of the surface loading) in Fig. 8.46, it
can be observed that in this case creep hardly affects the generated stresses compared to
reference run TPM 6. However, the generation of hydration temperature leads to a larger
bending of the cross section with compressive stresses of -7.7 MPa at the lining intrados (0.0 m)
and -5.2 MPa at the lining extrados (0.2 m). When combining all the three components of
creep, shrinkage and hydration temperature together (run C+SH+T), an almost linear stress
distribution across the lining section with an increased bending compared to run TPM 6 is
obtained.

342
 -5,0

TPM 6
CREEP
-5,5
TEMPERATURE
SHRINKAGE
C+SH+T

Vertical stress [MPa]

-6,0

-6,5

-7,0

-7,5

-8,0
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.46: Vertical hoop stress distribution in right tunnel sidewall for analysis of creep,
shrinkage and hydration temperature adopting Ko = 1.5 (Inc 100)

The situation is different at the tunnel crown, as illustrated in Fig. 8.47. The consideration
of shotcrete creep slightly reduces the bending of the cross section, with stresses located
exclusively in the compressive stress regime. The same mechanism is obtained by including
shrinkage in the analyses, where compressive stresses of -6.2 MPa are obtained at the lining
extrados (0.2 m) and -0.5 MPa at the lining intrados (0.0 m). Run C+SH+T leads to a
significant reduction in bending of the tunnel crown compared to reference run TPM 6,
which indicates, that the combination of these time-dependent aspects can have an important
impact on the lining behaviour during tunnel construction.

343
 0,20

TPM 6
CREEP
TEMPERATURE
0,15 SHRINKAGE
C+SH+T

Cross section [m]

0,10

0,05

0,00
-8 -6 -4 -2 0 2
Horizontal stress [MPa]

Figure 8.47: Horizontal hoop stress distribution in tunnel crown for analysis of creep, shrink-
age and hydration temperature adopting Ko = 1.5 (Inc 100)

Fig. 8.48 shows the maximum stress levels (or utilisation factors) that occur in the course
of tunnel excavation and surface loading at the intrados of the lining at the tunnel crown
for both Ko -profiles. First, it can be seen that the run TPM 1 predicts for both types of
analyses a very high stress level of almost ρc = 0.8 for Ko = 1.5 and even failure (ρc = 1.0)
for Ko = 0.5. The material response is governed exclusively by compression due to the
very low stiffness and strength level at a shotcrete age of 2 to 4 h and the compressive yield
surface controls the mechanical behaviour. For Ko = 1.5 this stress level reduces quickly
with increasing shotcrete age until it reaches a minimum of about ρc = 0.25 for run TPM
6, where the shotcrete has an age of 48 h at the end of surface loading. For older shotcrete,
the behaviour of the lining intrados switches to tensile stresses and stress levels immediately
rise up to a high value of about ρt = 0.85. The situation is somewhat similar when adopting
a Ko -profile of 0.5. For runs TPM 2 to 5 the compressive stress level slightly reduces to
ρc = 0.97 since the compressive strength of the shotcrete becomes a bit more developed with
increasing shotcrete age. However, for runs TPM 6 to 14, the mechanical behaviour of the
lining intrados is goverened by tensile stresses which go beyond peak and experience strain
softening. Consequently, cracking of the shotcrete shell occurs and the stress level ρt reaches
unity.

344
 1,0
TPM 14
TPM 1
Cracking
0,8

Stress level [-]

0,6

0,4 TPM 6

0,2
Compression Tension Ko = 0.5
Ko = 1.5
0,0
1 10 100 1000 10000
Shotcrete age at end of loading [h]

Figure 8.48: Stress level for different shotcrete ages at intrados of tunnel crown for both
Ko -profiles

In Fig. 8.49 the maximum stress levels of the shotcrete at the intrados of the right tunnel
sidewall are investigated for both Ko -profiles. Here again, the compressive stress levels for
run TPM 1 are very high, taking a value of ρc = 0.98 for Ko = 1.5 and ρc = 1.0 for Ko = 0.5.
For older shotcrete of runs TPM 2 to 6, these stress levels reduce relatively quickly and most
importantly remain of a compressive nature, where the compressive yield function controls
the mechanical behaviour. When adopting Ko = 1.5 the compressive stress levels seem to
stabilise for runs TPM 6 to 14 on a constant level of ρc = 0.29. This value is slightly higher
(ρc = 0.43) for Ko = 0.5 and even increases a small amount for runs TPM 9 to 14.

345
 1,1

1,0
Ko = 0.5
TPM 1
0,9 Ko = 1.5

0,8

0,7

Stress level [-]

0,6

0,5

0,4 TPM 14
0,3

0,2

0,1

0,0
1 10 100 1000 10000
Shotcrete age at end of loading [h]

Figure 8.49: Stress level for different shotcrete ages at intrados of tunnel sidewall for both
Ko -profiles

For completion, the stress levels at the lining intrados of the tunnel invert are presented
for both Ko -profiles in Fig. 8.50. As in the previous case of the tunnel crown, the mechanical
behaviour for runs TPM 1 to 6 is dominated by compressive stresses and stress levels reduce
from ρc = 0.75 down to a constant level of ρc = 0.25 for Ko = 1.5. In the case of Ko = 0.5,
the mechanism is different. The stress level for TPM 1 is very close to failure but reduces in
the same way as for Ko = 1.5 down to a relatively low level of ρc = 0.26. With increasing
age, the governing behaviour is of tensile nature and the tensile stress levels rise up for runs
TPM 7 to 13, until the shotcrete even cracks during run TPM 14.

346
 1,1

1,0 TPM 1 TPM 14

0,9

0,8
TPM 1
0,7

Stress level [-]

0,6

0,5

0,4

0,3
TPM 6 TPM 14
0,2
Compression Tension Ko = 0.5
0,1
Ko = 1.5
0,0
1 10 100 1000 10000
Shotcrete age at end of loading [h]

Figure 8.50: Stress level for different shotcrete ages at intrados of tunnel invert for both
Ko -profiles

Tables 8.22 and 8.23 summarize the results regarding the maximum stress level in the
lining intrados at the tunnel crown, right sidewall and invert for various developments of
stiffness and strength. For both Ko -profiles it can be observed that the faster increase in
stiffness and strength for run STRENGTH 1 (s = 0.1) leads to a small reduction in the
maximum stress level. For Ko = 1.5 there is even no plastic deformation noticeable for the
intrados at the tunnel crown and invert and the stress state remains below the compression
yield surface. On the other hand, when the increase of shotcrete stiffness and strength is
retarded as in the case of runs STRENGTH 2 to 4, the stress levels rise up to almost failure
(ρc = 0.94) for Ko = 0.5 and ρc = 0.72 for Ko = 1.5 at the intrados of the right tunnel sidewall.
The reduction of the stiffness and strength by 25 % leads to a compressive stress level that
is about 40 % higher than in the reference run TMP 6. Run STRENGTH 6 represents a
reduction of the shotcrete stiffness and strength by 50 % and consequently the compressive
stress levels increase of up to 90 % in the case of the tunnel invert. The utilisation factor in
the tunnel crown seems to be unchanged by the variation of stiffness and strength for the
Ko -profile of 0.5. In this case cracking of the shotcrete takes place throughout all analyses
indicated by a stress level of unity.

347
 Run Intrados sidewall Intrados crown Intrados invert
ρc ρt ρc
(-) (-) (-)
TPM 6 0.46 1.00 0.26
STRENGTH 1 0.41 1.00 0.25
STRENGTH 2 0.59 1.00 0.39
STRENGTH 3 0.76 1.00 0.58
STRENGTH 4 0.94 1.00 0.76
STRENGTH 5 0.57 1.00 0.35
STRENGTH 6 0.76 1.00 0.50

Table 8.22: Utilisation factor for variation in development of stiffness and strength for exca-
vation Type 1 at end of surface loading (Ko = 0.5)

Run Intrados sidewall Intrados crown Intrados invert

ρc ρc ρc
(-) (-) (-)
TPM 6 0.32 0.26 0.29
STRENGTH 1 0.28 elastic elastic
STRENGTH 2 0.41 0.37 0.41
STRENGTH 3 0.56 0.51 0.54
STRENGTH 4 0.72 0.65 0.67
STRENGTH 5 0.41 0.32 0.37
STRENGTH 6 0.82 0.43 0.48

Table 8.23: Utilisation factor for variation in development of stiffness and strength for exca-
vation Type 1 at end of surface loading (Ko = 1.5)

The effect of shotcrete deformability at early ages on the stress level within the tunnel
lining can be analysed with the help of Tab. 8.24 and 8.25. Run STRAIN 1, which represents
a softer material behaviour through larger peak strains for both compression and tension,
shows a slightly decreased stress level compared to reference run TPM 6 for both adopted
Ko -profiles. However, when a more brittle material response of shotcrete is assumed, as
in the case of runs STRAIN 2 and 3, the stress level at the intrados of the right tunnel
sidewall tends to increase significantly for Ko = 0.5, reaching values of ρc = 0.58. When a
constant strain level is applied in run STRAIN 4, the stresses in the lining intrados of the
tunnel crown suddenly switch from compressive behaviour to a tensile one and consequently

348
the stress level rises rapidly. A value of unity for both Ko -profiles indicates cracking of the
shotcrete. A similar mechanism is observed for the tunnel invert when adopting Ko = 0.5
although the stress level is slightly below peak. From these results it can be concluded that the
deformability of early age shotcrete has an important influence on the utilisation factor of a
tunnel lining, leading in some cases to cracking and strain softening behaviour. Furthermore,
it appears that cracking of the lining intrados of the tunnel crown is very likely to happen
when adopting a Ko -profile of 0.5 independently of the applied deformability values of runs
STRAIN 1 to 4.

Run Intrados sidewall Intrados crown Intrados invert

ρc ρt ρc
(-) (-) (-)
TPM 6 0.46 1.00 0.26
STRAIN 1 0.42 0.91 0.26
STRAIN 2 0.56 1.00 0.28
STRAIN 3 0.58 1.00 0.27
STRAIN 4 0.80 1.00 ρt = 0.98

Table 8.24: Utilisation factor for variation in shotcrete deformability for excavation Type 1
at end of surface loading (Ko = 0.5)

Run Intrados sidewall Intrados crown Intrados invert

ρc ρc ρc
(-) (-) (-)
TPM 6 0.32 0.26 0.29
STRAIN 1 0.31 0.26 0.28
STRAIN 2 0.37 0.28 0.31
STRAIN 3 0.37 0.29 0.31
STRAIN 4 0.47 ρt = 1.00 0.37

Table 8.25: Utilisation factor for variation in shotcrete deformability for excavation Type 1
at end of surface loading (Ko = 1.5)

8.7.2.2 Excavation Type 2

In this section, the hoop stress generation within the shotcrete shell for various tunnel cross
sections due to tunnel construction according to excavation Type 2, following a bench and
invert excavation scheme, will be discussed in detail. For all analyses a constant Ko -profile

349
of 1.5 was adopted.

Fig. 8.51 illustrates the horizontal hoop stresses in the tunnel invert for various analyses
regarding the elapsed time during construction at the end of the complete tunnel excavation
(Inc 25) and at the end of surface loading (Inc 65). The lining extrados is located at 0.0 m
and the lining inrados is situated at a cross-sectional distance of 0.2 m. At the end of invert
excavation in Inc 25, the linear elastic model (LE) predicts a linear stress distribution over
the cross section that leads to large bending of the invert. The non-linear run simulating
shotcrete behaviour at 28 days (NL) follows almost the same curve, except that the non-
linearity leads to a reduced compressive stress at the lining extrados. As expected, the HME
approach shows a much lower bending due to its relatively low constant stiffness compared
to the run LE. Stresses are remarkably reduced for the time-dependent runs TPM 1, 5 and 9,
where no bending of the invert is predicted and the mechanical behaviour is governed mainly
by hoop compression. Run TPM 1 simulates a very young shotcrete and therefore stiffness
and strength are at a low level. This leads to a constant compressive stress of -2.5 MPa.
The shotcrete of the analyses TPM 5 and 9 is slightly older due to the elapsed time during
excavation and compressive stresses rise up to -4.9 MPa and -5.6 MPa respectively. However,
during the placing of the surface surcharge from Inc 26 to 65, the loading direction is changed
and the bending of the invert is reversed, as can be seen in Fig. 8.51. Although heavily loaded,
all analyses still predict compressive stresses over the whole lining cross section, which range
from -6.0 MPa to -8.6 MPa for the linear elastic model (LE). The non-linear plasticity of the
hardened shotcrete at 28 days (NL) has a minor effect at the lining intrados leading to a
smaller stress of -4.7 MPa. One interesting fact is that the time-dependent analyses TPM 1,
5 and 9 show a very similar bending as run NL, since the curves appear to be almost parallel
to each other, but differ in magnitude of the compressive stress. This indicates that for those
time-runs, the compressive hoop stress is reduced for young shotcrete (TPM 1), but bending
remains constant with cement hydration. The missing time-runs simulating the longitudinal
distance between the bench and invert face by adopting a certain amount of elapsed time Δt
in Inc 13, are not included in these two figures since there is obviously no effect on the invert
behaviour, which is installed in Inc 14.

350
 0,20

0,15

Cross section [m]

0,10

LE
HME
0,05 NL
TPM 1
TPM 5
TPM 9
0,00
-12 -10 -8 -6 -4 -2 0
Horizontal stress [MPa]
0,20

0,15
Cross section [m]

0,10

LE
HME
0,05 NL
TPM 1
TPM 5
TPM 9
0,00
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0
Horizontal stress

Figure 8.51: Horizontal hoop stresses in tunnel invert for various time-runs at the end of
invert excavation (Inc 25, top) and end of surface loading (Inc 65, bottom)

The horizontal stresses in the centre of the temporary invert at the end of bench exca-
vation (Inc 12) are presented in Fig. 8.52. The soil below this temporary support element
pushes the temporary invert upwards leading to large bending of the cross section when ap-
plying a linear elastic model (LE). Tensile stresses at the intrados (0.15 m) are of a value of
3.5 MPa and slightly above the tensile strength of hardened shotcrete (3.0 MPa). Compressive
stresses at the extrados (0.0 m) rise up to -9.1 MPa. Both stress values are slightly reduced
when including non-linear plasticity of the shotcrete in run NL. The bending is significantly
decreased when simulating shotcrete behaviour with the HME approach, where no tensile
stresses are predicted and the compressive stresses vary from -2.0 to almost -4.0 MPa. Once

351
again, the time dependent runs TPM 1, 5 and 9 indicate that the mechanical behaviour of
the temporary invert is governed mainly by hoop compression with little bending of the cross
section. Due to the early age material properties of the shotcrete compressive stresses develop
at a low level ranging from -1.4 MPa (TPM 1) to -3.9 MPa (TPM 9).

0,15

0,12
Cross section [m]

0,09

0,06
LE
HME
NL
0,03 TPM 1
TPM 5
TPM 9
0,00
-10 -8 -6 -4 -2 0 2 4
Horizontal stress [MPa]

Figure 8.52: Horizontal hoop stresses in temporary invert for various time-runs at the end of
bench excavation (Inc 12)

The adopted construction process of excavation Type 2 implies that the upper shotcrete
shell of the tunnel for the bench is of an older age than the lower shell and experiences therefore
a different mechanical behaviour due to the advanced progress of cement hydration within
the shotcrete. When looking at the stress distribution in the tunnel crown at the completion
of tunnel excavation (Inc 25), the elapsed time Δt at the intermediate construction stage in
Inc 13 seems to have an impact on the stresses that are generated due to the soil excavation,
as can be seen in Fig. 8.53 to 8.55. In these figures, the lining intrados is located at 0.0 m,
whereas the lining extrados is situated at a 0.2 m. From Fig. 8.53 it can be observed that the
runs HME and TPM 1 predict a relatively low bending of the cross section, where stresses
vary in between -2.0 and -4.0 MPa for TPM 1 and -2.4 to 5.6 MPa for the HME approach.
The situation is very different for runs TPM 2, 3, and 4, which already show a very stiff lining
behaviour due to the advanced shotcrete age. These analyses appear to approach the result
from the linear elastic model (LE), introducing tensile stresses at the lining extrados ranging
from 0.6 to 1.6 MPa. Differences are slightly larger at the lining intrados, where compressive
stresses lie between -7.1 and -9.8 MPa.

352
 0,20

0,15

Cross section [m]

0,10

LE
HME
0,05 TPM 1
TPM 2
TPM 3
TPM 4
0,00
-10 -8 -6 -4 -2 0 2
Horizontal stress [m]

Figure 8.53: Horizontal hoop stresses in tunnel crown for time-runs TPM 1 to 4 at the end
of invert excavation (Inc 25)

This effect of the difference in shotcrete age becomes slightly less pronounced for the
time-dependent runs TPM 5 to 8, as can be seen in Fig. 8.54. Similar to the previous case,
runs TPM 5 and HME are relatively close together predicting slight bending of the tunnel
crown. However, stresses from the analyses TPM 6, 7 and 8 locate themselves in between
the linear elastic model (LE) and the HME approach, showing compressive stresses over the
whole lining cross section. Consequently no cracking at the lining extrados can be expected.

0,20

0,15
Cross section [m]

0,10

LE
HME
0,05 TPM 5
TPM 6
TPM 7
TPM 8
0,00
-10 -8 -6 -4 -2 0 2
Horizontal stress [MPa]

Figure 8.54: Horizontal hoop stresses in tunnel crown for time-runs TPM 5 to 8 at the end
of invert excavation (Inc 25)

353
 Finally, in Fig. 8.55 the horizontal stress predictions at the tunnel crown for runs TPM
9 to 12 are presented and compared to the simple models LE and HME. In the upper half
of the cross section (0.1 to 0.2 m), elastic behaviour seems to be the governing factor of
the mechanical behaviour since the stress distribution curves for all time runs tend to be
parallel to the linear elastic model (LE). Cracking at the lining extrados is likely to happen
upon further loading since tensile stresses start to develop for runs TPM 11 and 12. At the
lining intrados, the non-linear behaviour of shotcrete during plastic deformation seems to be
dominant and compressive stresses vary from -6.0 MPa (TPM 9) to -7.2 MPa (TPM 12). It
appears, that for a more advanced age of shotcrete at the end of bench excavation, i.e. 24
hours for runs TPM 9 to 12, the influence of the elapsed time Δt in Inc 13, simulating the
longitudinal distance between bench and invert face, diminishes. A relatively stiff tunnel
lining can lead to cracking at the lining extrados of the tunnel crown during excavation of
the tunnel.
0,20

0,15
Cross section [m]

0,10

LE
HME
0,05 TPM 9
TPM 10
TPM 11
TPM 12
0,00
-10 -8 -6 -4 -2 0 2
Horizontal stress [MPa]

Figure 8.55: Horizontal hoop stresses in tunnel crown for time-runs TPM 9 to 12 at the end
of invert excavation (Inc 25)

The influence of the increase of stiffness and strength with time on the mechanical lining
behaviour of excavation Type 2 was investigated for the tunnel invert at the end of surface
loading (Inc 65), as illustrated in Fig. 8.56. Run TPM 7 serves as a reference run and results
are compared with three different runs with various stiffness and strength developments. Run
STRENGTH 1 (s = 0.1) predicts almost identical compressive stresses as run TPM 7 at the
lining extrados (0.0 m) of up to -7.9 MPa. The bending of run STRENGTH 5, having a
stiffness and strength reduced by 25 %, appears to be very similar to the one obtained by
run TPM 7. The largest difference in stresses is observed for run STRENGTH 3, where
a cement parameter s = 0.4 simulates a retarded increase in stiffness and strength. At the
lining extrados compressive stresses rise up to -5.0 Mpa, whereas at the lining intrados (0.2 m)

354
stresses of -0.9 MPa can be expected. However, when comparing these results with run TPM
7, it can be concluded that the major difference is the decreased hoop stress and bending
seems to be unchanged. All runs show exclusively compressive behaviour and there is no risk
of shotcrete cracking due to any tensile stresses.

0,20

Cross section [m] 0,15

0,10

0,05 TPM 7
STRENGTH 1
STRENGTH 3
STRENGTH 5
0,00
-8 -6 -4 -2 0
Horizontal stress [MPa]

Figure 8.56: Horizontal hoop stresses in tunnel invert for various developments of stiffness
and strength at the end of surface loading (Inc 65)

When analysing the tunnel construction with different admissible strain levels of the young
shotcrete, it can be seen that results of runs STRAIN 3 and 4 are very similar to those of
reference run TPM 7. In Fig. 8.57 the horizontal stresses at the tunnel crown are plotted at
the end of surface loading in Inc 65. For all runs tensile stresses are generated at the tunnel
intrados (0.0 m) that might lead to cracking and tensile strain softening of the crown. At the
lining extrados (0.2 m), the compressive stresses vary between -7.4 and -9.0 MPa. No major
change in hoop stress or bending of the cross section can be identified.

355
 0,20

TPM 7
STRAIN 3
0,15
STRAIN 4

Cross section [m]

0,10

0,05

0,00
-10 -8 -6 -4 -2 0 2
Horizontal stress [MPa]

Figure 8.57: Horizontal hoop stresses in tunnel crown for various shotcrete deformabilty at
the end of surface loading (Inc 65)

The influence of creep, shrinkage and hydration temperature of early age shotcrete on
the mechanical behaviour of the lining is shown in Fig. 8.58 and 8.59, where the results 28
days after completion of the surface loading (Inc 100) are compared with reference run TPM
7 (Inc 65). In Fig. 8.58 the horizontal stresses in the tunnel crown are illustrated. It can
be seen that the stresses for the run CREEP reduce slightly on both the lining intrados
(0.0 m) and the lining extrados (0.2 m), giving values of 0.3 and -6.9 MPa respectively. This
difference on the lining extrados is even more pronounced for run SHRINKAGE, where the
compressive stresses decrease to -6.4 MPa. When combining all three effects (i.e. creep,
shrinkage and hydration temperature) together in run C+SH+T, the bending of the tunnel
crown is reduced in such a way, that even compressive stresses of -0.60 MPa occur at the
lining intrados, preventing therefore the risk of cracking.

356
 0,20

TPM 7
CREEP
0,15 SHRINKAGE
TEMPERATURE
C+SH+T

Cross section [m]

0,10

0,05

0,00
-7,5 -6,0 -4,5 -3,0 -1,5 0,0 1,5
Horizontal stress [MPa]

Figure 8.58: Horizontal hoop stresses in tunnel crown for analysis of creep, shrinkage and
hydration temperature

Differences in the obtained stresses are of a similar magnitude for the horizontal stresses
in the tunnel invert, as can be seen in Fig. 8.59. For this cross section of the tunnel lining,
the run CREEP shows the biggest reduction in compressive stresses at the lining intrados
(0.2 m) taking a value of -3.5 MPa. Results are almost unchanged for runs C+SH+T and
TEMPERATURE compared to reference run TPM 7 (around -2.7 MPa). The compressive
stresses at the lining intrados (0.0 m) vary from -6.9 to -7.8 MPa for all runs, with the run
CREEP predicting the smallest bending of the cross section.

357
 0,20

0,15

Cross section [m]

0,10

TPM 7
0,05 CREEP
SHRINKAGE
TEMPERATURE
C+SH+T
0,00
-8 -7 -6 -5 -4 -3 -2
Horizontal stress [MPa]

Figure 8.59: Horizontal hoop stresses in tunnel invert for analysis of creep, shrinkage and
hydration temperature

The connection point of the temporary invert to the tunnel lining is of great importance
regarding the stability of the excavated bench of the tunnel at an intermediate construction
stage. From a numerical point of view, this connection area represents a zone where conver-
gence problems could be expected because of the corners of the geometry (see Fig. 8.16) and
since large changes in stresses can be expected.

The following Tab. 8.26 summarises the maximum tensile stress σ1 and the maximum
compressive stress σ3 in this connection area for the different constitutive models for shotcrete.
It can be seen immediately that the large maximum tensile stress of 25.5 MPa, predicted by
the linear elastic model (LE), is unrealistic and would not occur in reality. An improvement in
model predictions is achieved by adopting the HME approach with a reduced constant lining
stiffness. However, the obtained tensile stress of 10.9 MPa still exceeds the tensile strength of
3.0 MPa of hardened shotcrete at 28 days, which is the limit stress for the non-linear plasticity
model (NL). Incorporating the time-dependent behaviour of shotcrete in runs TPM 1, 5 and
9 leads to relatively low tensile stresses ranging from 0.3 to 1.3 MPa, since for these runs the
shotcrete stiffness and strength are not fully developed yet. The same conclusion applies to
the maximum compressive stresses, that are relatively large (-41.9 MPa) for the linear elastic
model (LE) and are reduced by including non-linear plasticity (-30.5 MPa) or by adopting
the HME approach (-22.0 MPa). For the time-dependent runs these stresses vary between
-4.4 and -14.2 MPa. Retarding and reducing the increase in stiffness and strength (runs
STRENGTH 3 and 5) leads to slightly reduced maximum tensile and maximum compressive
stresses. The opposite effect has been noticed when changing the shotcrete deformabilty to
a very brittle material behaviour (runs STRAIN 3 and 4), where the maximum compressive

358
stresses are increased up to -14.3 and -15.0 MPa compared to run TPM 7 (which equals TPM
5).

Run Maximum σ1 Minimum σ3

(MPa) (MPa)
Linear elastic (LE) + 25.5 - 41.9
HME + 10.9 - 22.0
Non-linear (NL) + 3.0 - 30.5
TPM 1 + 0.3 - 4.4
TPM 5 + 0.8 - 10.1
TPM 9 + 1.3 - 14.2
STRENGTH 1 + 1.6 - 15.8
STRENGTH 3 + 0.2 - 4.0
STRENGTH 5 + 0.6 - 8.3
STRAIN 3 + 0.8 - 14.3
STRAIN 4 + 0.8 - 15.0

Table 8.26: Minimum and maximum principal stresses occurring in the intersection of the
temporary invert with the tunnel lining at end of bench excavation (Inc 12)

The maximum stress levels (or utilisation factors) that occur at the intrados of the tunnel
crown and invert in the course of tunnel excavation and surface loading are listed in the
following Tab. 8.27 and will be discussed here briefly. From the results it can be concluded,
that non-linear plasticity of shotcrete at early ages plays an important role in the mechanical
behaviour of the shotcrete shell. For very young shotcrete, represented by runs TPM 1 to
7, the compressive stress levels range from ρc = 0.28 to ρc = 0.53 at the tunnel crown and
from ρc = 0.46 to ρc = 0.70 at the tunnel invert. For these runs the compressive yield
surface controls the material response during the formation of plastic deformation. However,
when the cement hydration within the shotcrete is progressed at later ages (TPM 8 to 14),
the behaviour of the lining intrados at the tunnel crown becomes tensile and the utilization
factor rises quickly up to ρt = 0.84 which is already close to peak. In contrary, the stress
states at the lining intrados of the tunnel invert remain compressive and the stress levels
reduce to ρc = 0.34, since the compressive peak surface has moved away in stress space
with time due to the advanced strength levels at later shotcrete ages. When applying an
accelerated increase in stiffness and strength for run STRENGTH 1 (s = 0.1), the stress
state at the tunnel crown does not show any plastic deformation and remains in the elastic
area below the compressive yield surface. At the lining intrados of the tunnel invert the
stress level is reduced to ρc = 0.29 compared to ρc = 0.46 for reference run TPM 7. The

359
utilization factor rises again in the case of a retarded development of stiffness and strength
for run STRENGTH 3 (s = 0.4) and STRENGTH 5 (reduced stiffness and strength by 25 %).
Maximum stress levels can reach ρc = 0.58 for the tunnel crown and ρc = 0.80 for the tunnel
invert. A similar trend is observed when the early age shotcrete deformability is changed to
a very brittle or even constant development of peak strains with time. For run STRAIN 4,
the obtained utilization factor of the lining intrados at the tunnel invert goes up to ρc = 0.62
compared to ρc = 0.46 for reference run TPM 7.

Run Intrados crown Intrados invert

ρc ρc
(-) (-)
Non-linear (NL) elastic 0.30
TPM 1 0.53 0.70
TPM 2 0.51 0.70
TPM 3 0.50 0.70
TPM 4 0.48 0.70
TPM 5 0.29 0.46
TPM 6 0.28 0.46
TPM 7 0.28 0.46
TPM 8 ρt = 0.82 0.46
TPM 9 ρt = 0.81 0.34
TPM 10 ρt = 0.82 0.34
TPM 11 ρt = 0.83 0.34
TPM 12 ρt = 0.83 0.34
STRENGTH 1 elastic 0.29
STRENGTH 3 0.58 0.80
STRENGTH 5 0.34 0.54
STRAIN 3 0.31 0.53
STRAIN 4 0.37 0.62

Table 8.27: Utilisation factor of lining intrados at tunnel crown and invert for excavation
Type 2 (Ko = 1.5)

8.7.2.3 Excavation Type 3

For a detailed understanding of the staged tunnel construction according to excavation Type
3, the figures presented in this section illustrate the hoop stresses within the tunnel lining

360
for various types of analyses at different construction and loading stages. As in the previous
case of excavation Type 2, a constant Ko -profile of 1.5 was adopted.

In Fig. 8.60 to 8.62 the distributions of the vertical hoop stresses in the left tunnel sidewall
are illustrated for several analyses investigating the influence of the elapsed time during
excavation of the tunnel and surface loading. The lining extrados is located at 0.0 m, whereas
the lining intrados is situated at a cross-sectional distance of 0.2 m. From Fig. 8.60 it can be
observed, that at the end of surface loading in Inc 65 the linear elastic model (LE) predicts
large bending of the cross section with tensile stresses at the lining extrados of 2.0 MPa and
compressive stresses rising up to -17.1 MPa at the lining intrados. Due to the softer material
behaviour of the HME approach, the bending is significantly reduced and the mechanical
behaviour of the tunnel shell is mainly governed by compressive stresses across the section,
ranging from -4.7 to -9.8 MPa. At this final stage of construction, non-linear plasticity of
hardened shotcrete at 28 days (NL) seems to play an important rule, reducing mainly the
compressive stresses at the lining intrados (-12.6 MPa) compared to the LE model. Results
from run TPM 1 show the typical mechanical behaviour of very young shotcrete, where
almost no bending of the cross section is obtained due to the low stiffness at early ages.
Compressive stresses vary from -6.0 to -7.2 MPa. Runs TPM 2 to 4, which are controlled by
the intermediate elapsed time at Inc 13, represent a transition from the HME approach to
the linear elastic model. With an increasing amount of elapsed time in Inc 13, the material
behaviour of shotcrete becomes stiffer and therefore a significant bending is introduced in the
left tunnel sidewall, leading to critical tensile stresses at the tunnel extrados, ranging from
0.4 to 1.9 MPa.

0
Vertical stress [MPa]

-3

-6

-9 LE
HME
-12 NL
TPM 1
TPM 2
-15
TPM 3
TPM 4
-18

0,00 0,05 0,10 0,15 0,20

Cross section [m]

Figure 8.60: Vertical hoop stresses in left sidewall for time-runs TPM 1 to 4 at the end of
surface loading (Inc 65)

361
 A slightly different situation can be observed in Fig. 8.61, where the time-runs TPM 5
to 8 predict a very similar behaviour of the left tunnel sidewall. Stress distributions for all
runs are located exclusively in the area of compressive stresses, showing very little differences
at the lining intrados (-7.8 to -9.5 MPa). The stress distribution becomes more non-linear
towards the extrados of the lining, where compressive stresses range from -1.7 to -6.6 MPa.
It can be concluded, that the HME approach represents a fairly good approximation to these
time-dependent runs, where bending of the cross section is much smaller compared to the
linear elastic model (LE). The difference through the elapsed time in Inc 13 seems not so
pronounced as in the previous case for runs TPM 1 to 4.

0
Vertical stress [MPa]

-3

-6

-9
LE
-12 HME
TPM 5
TPM 6
-15
TPM 7
TPM 8
-18

0,00 0,05 0,10 0,15 0,20

Cross section [m]

Figure 8.61: Vertical hoop stresses in left sidewall for time-runs TPM 5 to 8 at the end of
surface loading (Inc 65)

By comparing Fig. 8.62 with Fig. 8.61 it can be seen that the compressive stresses at the
lining intrados hardly change at all for runs TPM 9 to 12 and are located around -9.0 MPa.
However, due to the older age of the shotcrete for these runs, the behaviour on the lining
extrados becomes more brittle and stresses tend to move towards the tensile stress region. The
smallest compressive stresses are obtained for run TPM 12 (-2.3 MPa), where the shotcrete
of the left tunnel shell is the oldest with an age of 120 h.

362
 3

Vertical stress [MPa]

-3

-6

-9
LE
-12 HME
TPM 9
-15
TPM 10
TPM 11
TPM 12
-18

0,00 0,05 0,10 0,15 0,20

Cross section [m]

Figure 8.62: Vertical hoop stresses in left sidewall for time-runs TPM 9 to 12 at the end of
surface loading (Inc 65)

Fig. 8.63 depicts the vertical stress distribution within the temporary sidewall at the end
of the left side gallery excavation in Inc 12. High horizontal stresses in the surrounding ground
(Ko = 1.5) push this temporary support element to the left towards the centre of the tunnel,
introducing high tensile stresses at the intrados of the sidewall (0.0 m) reaching 7.4 MPa for
the linear elastic model (LE). Compressive stresses at the sidewall extrados (0.15 m) rise up
to -12.9 MPa. When adopting the HME approach, this large bending is reduced significantly
and represents a fairly good approximation for the obtained curves from the time-dependent
runs TPM 1, 5 and 9. Particularly runs TPM 1 and 5 do not show any bending of the
temporary sidewall and the governing behaviour is hoop compression at relatively low stresses
of -1.5 MPa for TPM 1 and -2.8 MPa for TPM 5. Due to the slightly progressed shotcrete age
of run TPM 9 (24 h at the end of left side gallery excavation) a somewhat stiffer behaviour
can be expected introducing a little bending. The non-linear plasticity model for hardened
shotcrete at 28 days (NL) appears to be an average result with compressive stresses of -
8.7 MPa. Tensile stresses go up to 2.5 MPa. Since they are already relatively close to the
tensile capacity of hardened shotcrete (3.0 MPa), there is a potential risk of cracking at the
intrados of the temporary sidewall.

363
 8

Vertical stress [MPa]

0

-4

LE
-8 HME
NL
TPM 1
-12 TPM 5
TPM 9

0,00 0,03 0,06 0,09 0,12 0,15

Cross section [m]

Figure 8.63: Vertical hoop stresses in temporary sidewall at the end of excavation of the left
side gallery (Inc 12)

Fig. 8.64 and 8.65 show the vertical stresses in the right tunnel sidewall at the end of
excavation in Inc 25 and at the end of surface loading in Inc 65. The lining intrados is located
at 0.0 m and the lining extrados at a cross-sectional distance of 0.2 m. One interesting fact
that can be observed from these figures is, that at the end of tunnel excavation the bending
predicted by all applied constitutive models, even the linear elastic model (LE), is relatively
small. Furthermore, although high initial horizontal stresses are anticipated (Ko = 1.5), the
bending direction seems not to change for this particular cross section during the phase of
surface loading, unlike the behaviour observed for excavation Type 1 and 2. Results from
runs TPM 1, 5 and 9 show low compressive stresses with hoop compression as the main
mechanical behaviour of the tunnel shell even when heavily loaded in Inc 65. During the
phase of surface loading from Inc 25 to 65, the bending obtained from the linear elastic
model (LE) appears to increase enormously, leading to tensile stresses at the lining extrados
of 2.1 MPa and compressive stresses at the lining intrados of -17.2 MPa. These stresses are
reduced by incorporating non-linear plasticity for hardened shotcrete at 28 days (NL).

364
 0

-1

Vertical stress [MPa]

-2

-3

-4

LE
-5 HME
NL
-6 TPM 1
TPM 5
TPM 9
-7
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.64: Vertical hoop stresses in right tunnel sidewall at the end of tunnel excavation
(Inc 25)

0
Vertical stress [MPa]

-3

-6

-9
LE
-12 HME
NL
TPM 1
-15
TPM 5
TPM 9
-18

0,00 0,05 0,10 0,15 0,20

Cross section [m]

Figure 8.65: Vertical hoop stresses in right tunnel sidewall at the end of surface loading (Inc
65)

The influence of the increase in shotcrete stiffness and strength with time on the stress
generation in the tunnel lining for excavation Type 3 can be studied in Fig. 8.66. The vertical
stresses in the right tunnel sidewall are plotted at the end of surface loading in Inc 65 and
results are compared with reference run TPM 7. It can be seen, that assuming a stiffer
behaviour of the shotcrete for run STRENGTH 1 (s = 0.1) results in a slight increase of

365
the bending of the cross section, where stresses at the lining intrados (0.0 m) increase up to
-7.8 MPa and reduce to -6.0 MPa at the lining extrados. Runs STRENGTH 3 and 5 simulate
a softer lining behaviour through a retarded and reduced increase in stiffness and strength
with time. Their results indicate that the bending of the right tunnel sidewall remains almost
unchanged, since the curves appear to be almost parallel to each other. However, there is
a noticeable difference for the hoop compression that occurs in the cross section, which is
significantly reduced for run STRENGTH 3 (s = 0.4). Run STRENGTH 5, representing
a reduction of the shotcrete stiffness and strength of 25 %, leads to a small reduction in
compressive stresses on both the lining intrados and extrados, ranging from -6.1 to -6.7 MPa.

-4

-5
Vertical stress [MPa]

-6

-7

TPM 7
-8 STRENGTH 1
STRENGTH 3
STRENGTH 5
-9
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.66: Vertical hoop stresses in right sidewall at the end of surface loading (Inc 65) for
various developments of stiffness and strength

The application of different shotcrete deformabilities at early ages has been analysed in
runs STRAIN 3 and 4, which represent a more brittle mechanical behaviour of shotcrete than
reference run TPM 7. Fig. 8.67 shows the results for the vertical stress distribution at the
left tunnel sidewall at the end of surface loading in Inc 65. As a general trend, it can be
observed that reduced peak strains for both compression and tension lead to an increase in
bending of the cross section. Run STRAIN 4, representing a low constant possible straining
level of the uniaxial stress-strain curve for shotcrete, predicts small tensile stresses at the
lining extrados (0.0 m) of 0.2 MPa. However, on the lining intrados (0.2 m) the compressive
stresses increase from -8.7 MPa for run TPM 7 up to -13.0 MPa. Run STRAIN 4, governed
by a sharp decrease in peak strains with time, can be seen to give an intermediate stress
distribution, where stresses remain compressive at the lining extrados (-1.8 MPa), but rise to
-10.1 MPa at the lining intrados.

366
 TPM 7
0
STRAIN 3
STRAIN 4
-3

Vertical stress [MPa]

-6

-9

-12

-15
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.67: Vertical hoop stresses in left sidewall at the end of surface loading (Inc 25) for
various shotcrete deformabilities

Fig. 8.68 and 8.69 investigate the influence of creep, shrinkage and hydration temperature
of young shotcrete on the generated stress distributions within the tunnel lining. Fig. 8.68
plots the vertical stresses in the left tunnel sidewall at 28 days after the end of surface
loading (Inc 100) and compares the results with reference run TPM 7 (Inc 65). It can
be observed, that very little change occurs when incorporating creep of shotcrete, with the
analysis giving a very similar stress distribution pattern as for run TPM 7. Shrinkage increases
the compressive stresses at the lining extrados (0.0 m) from -3.6 MPa for run TPM 7 up to
-5.1 MPa. On the contrary, at the lining intrados (0.2 m) run SHRINKAGE predicts slightly
reduced compressive stresses of -8.4 MPa. Run TEMPERATURE tends to follow an almost
linear stress distribution across the lining cross section with compressive stresses of -4.5 MPa
at the lining extrados and relatively large stresses of -9.1 MPa at the lining intrados. These
results are very close to those from run C+SH+T. The incorporation of all the time-dependent
effects of creep, shrinkage and hydration temperature leads to a fairly linear stress distribution
for the left tunnel sidewall.

367
 -3

TPM 7
-4 CREEP
SHRINKAGE
-5 TEMPERATURE
C+SH+T

Vertical stress [MPa]

-6

-7

-8

-9

-10
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.68: Vertical hoop stresses in left tunnel sidewall for analysis of creep, shrinkage and
hydration temperature (Inc 100)

In Fig. 8.69 the vertical stresses in the right tunnel sidewall are illustrated at Inc 100.
It can be seen, that there are only minor differences in the obtained stress results for the
investigated runs taking into account creep, shrinkage and hydration temperature compared
to reference run TPM 7. At the lining intrados (0.0 m) compressive stresses range from
-7.1 MPa (SHRINKAGE) to -7.3 MPa (TEMPERATURE) and from -5.6 MPa (C+SH+T)
to -6.3 MPa (CREEP). In this particular case, the bending of the cross section is slightly
increased for run C+SH+T, compared to reference run TPM 7.

368
 -5,0

TPM 7
-5,5 CREEP
SHRINKAGE
TEMPERATURE

Vertical stress [MPa]

-6,0 C+SH+T

-6,5

-7,0

-7,5

-8,0
0,00 0,05 0,10 0,15 0,20
Cross section [m]

Figure 8.69: Vertical hoop stresses in right tunnel sidewall for analysis of creep, shrinkage
and hydration temperature (Inc 100)

Similar to the analysis of excavation Type 2, the connection of the temporary sidewall to
the tunnel lining was of special interest regarding the generated stresses due to excavation
of the left side gallery. Tab. 8.28 summarises the results for the maximum principal stress
σ1 and the minimum principal stress σ3 that occur in this top intersection area near the
tunnel crown for all the analyses carried out. Firstly it is noted, that the linear elastic
model predicts unrealistic high tensile stresses of up to 41.1 MPa. These stresses are reduced
by adopting the softer HME approach, but the stresses still exceed the tensile capacity of
hardened shotcrete at 28 days, as indicated for run NL (3.0 MPa). Runs TPM 1, 5 and 9
represent young shotcrete with low developed strengths and therefore, the tensile stresses
are reduced significantly ranging from 0.3 to 1.3 MPa. The same observation can be made
when retarding or reducing the development of the shotcrete stiffness and strength for runs
STRENGTH 3 and 5. Tensile stresses remain unchanged for runs STRAIN 3 and 5, but
compressive stresses increase slightly compared to reference run TPM 7 (-12.9 MPa). The
compressive stress of -63.2 MPa predicted by the linear elastic model exceeds as well the
adopted uniaxial compressive strength of -30 MPa. However, it should be mentioned that for
a triaxial stress state such high compressive stresses for concrete and shotcrete are possible to
achieve. A fact which is taken into account by the compressive yield surface of the presented
constitutive model for shotcrete in the previous Chapter 7.

369
 Run Maximum σ1 Minimum σ3
(MPa) (MPa)
Linear elastic (LE) + 41.1 - 63.2
HME + 18.1 - 32.3
Non-linear (NL) + 3.0 - 29.7
TPM 1 + 0.3 - 4.9
TPM 5 + 0.8 - 12.9
TPM 9 + 1.3 - 19.3
STRENGTH 1 + 1.6 - 21.6
STRENGTH 3 + 0.2 - 4.8
STRENGTH 5 + 0.6 - 10.4
STRAIN 3 + 0.8 - 16.1
STRAIN 4 + 0.8 - 17.2

Table 8.28: Minimum and maximum principal stresses occurring in the top intersection of
the temporary sidewall with the tunnel lining at end of left side gallery excavation (Inc 12)

Finally, Tab. 8.29 contains information about the maximum utilisation factors (or stress
levels) that occur in the lining intrados of the left and right tunnel sidewall throughout the
whole construction process. It can be seen, that for very young shotcrete (TPM 1), the stress
level is mainly of a compressive nature reaching levels of up to ρc = 0.67 for the left sidewall
and ρc = 0.74 for the right sidewall. These utilization factors tend to decrease with increasing
age of the shotcrete during the analysis (TPM 2 to 12) and appear to stabilize at a relatively
low level of around ρc = 0.32 to ρc = 0.34 for both cross sections. The maximum utilisation
factor was achieved for run STRENGTH 3, where the increase in stiffness and strength is
retarded by the cement parameter s = 0.4. Stress levels rise up to ρc = 0.70 for the left
sidewall and ρc = 0.82 for the right sidewall. A faster development of stiffness and strength,
as simulated in run STRENGTH 1 (s = 0.1), leads to a slightly reduced stress level for both
cross sections (ρc = 0.31) compared to reference run TPM 7. Another conclusion that can
be drawn is, that a reduced deformability of shotcrete at early ages implies an increased
stress level as predicted by runs STRAIN 3 and 4. For the intrados of the right sidewall, the
utilization factor rises up to ρc = 0.64 for run STRAIN 4. Finally, as already highlighted
in the previous section for excavation Type 2, the connection of the temporary sidewall to
the tunnel lining near the crown, is a heavily loaded zone, where cracking and crushing of
the shotcrete is dominating the mechanical behaviour. This results in utilisation factors for
both compression and tension that have a value of unity, indicating the occurance of strain
softening.

370
 Run Intrados left sidewall Intrados right sidewall
ρc ρc
(-) (-)
Non-linear (NL) 0.32 0.32
TPM 1 0.67 0.74
TPM 2 0.60 0.74
TPM 3 0.59 0.74
TPM 4 0.58 0.73
TPM 5 0.38 0.43
TPM 6 0.37 0.43
TPM 7 0.35 0.43
TPM 8 0.35 0.43
TPM 9 0.32 0.34
TPM 10 0.32 0.34
TPM 11 0.32 0.34
TPM 12 0.32 0.34
STRENGTH 1 0.31 0.31
STRENGTH 3 0.70 0.82
STRENGTH 5 0.43 0.28
STRAIN 3 0.42 0.51
STRAIN 4 0.56 0.64

Table 8.29: Utilisation factor of lining intrados at left and right tunnel sidewall for excavation
Type 3 (Ko = 1.5)

8.7.3 Lining displacements

8.7.3.1 Excavation Type 1

Fig. 8.70 shows the development of the vertical displacements of the lining intrados at the
tunnel crown during the whole construction process starting at lining installation (Inc 3)
until the end of surface loading (Inc 50) adopting a Ko -profile of 0.5. Furthermore, the
results from various constitutive models for simulating shotcrete behaviour are included. It
can be seen that all runs indicate a negative movement of the crown towards the centre of
the tunnel during the excavation process. This vertical displacement is very small for the
linear elastic model (LE) with a value of -1.5 mm. The HME approach predicts a vertical
displacement four times larger at -6.1 mm. The run TPM 1, where the shotcrete has just an

371
age of 2 h at the end of excavation, results in an extremely large lining displacement of up to
-155.8 mm due to the low stiffness of the shotcrete at the early age. However, these vertical
displacements reduce relatively quickly with increasing shotcrete age, showing values within
a range of -77.6 to -5.8 mm for runs TPM 2 to 6. For runs TPM 7 to TPM 14 no significant
change in vertical displacement with shotcrete age can be noted, with values close to the LE
and HME approach. No difference has been obtained by including non-linear plasticity of
the hardened shotcrete at 28 days (NL) meaning that with the full shotcrete stiffness the
tunnel lining behaves mainly elastically. If the loading of the tunnel shell is continued by
placing the surface surcharge from Inc 11 to Inc 50, the vertical displacements of the crown
increase further, resulting in a final value of -105.9 mm for the linear elastic model (LE).
Here again, the HME approach predicts a slightly larger value of -126.9 mm due to the lower
constant stiffness of the lining. The model predictions of run TPM 1 reach a vertical crown
displacement of -284.0 mm, followed by -195.3 mm for TPM 2. One interesting fact is, that
all the displacement curves appear to be almost parallel to each other during the surface
loading phase which indicates that the loading-displacement ratio must be more or less equal
for all performed runs. Therefore the material properties of the young shotcrete age have a
significant influence on the deformational behaviour of the tunnel lining particularly during
tunnel excavation.

372
 0
End of surface
-30
loading

Vertical displacement [mm]

-60

-90

-120
LE
-150
HME
-180 NL
TPM 1
-210 TPM 2
End of excavation
TPM 3
-240
TPM 4
-270 TPM 5
TPM 6
-300
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Increment

0
HME
TPM 7
TPM 8
Vertical displacement [mm]

-20
TPM 9
End of excavation TPM 10
-40
TPM 11
TPM 12
-60
TPM 13
TPM 14
-80
End of surface
loading
-100

-120

-140
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Increment

Figure 8.70: Vertical crown displacement for various time runs adopting Ko = 0.5 (Excavation
Type 1)

The displacement pattern is of a similar nature for the behaviour of the lining intrados
at the tunnel invert, as can be seen in Fig. 8.71. During excavation of the tunnel the lining
tends to move inwards resulting in positive displacement values from Inc 3 to 10 adopting
a Ko -profile of 0.5. In this graph, the influence of the development of stiffness and strength
of the shotcrete on the vertical invert displacement was studied by performing six differ-
ent runs (STRENGTH 1 to 6) and comparing results with the reference run TPM 6. The
vertical displacement of TPM 6 reaches a value of 15.8 mm at the end of excavation. The
run STRENGTH 1, which represents a stiffer tunnel lining by applying a cement param-
eter s = 0.1, shows a slightly reduced displacement of 14.5 mm. The retarded increase in
shotcrete stiffness of run STRENGTH 4 (s = 0.5) leads to a large heave of the invert moving

373
up 30.5 mm. A reduction of the lining stiffness of 25 % (STRENGTH 5) leads to a displace-
ment of 18.1 mm, whereas run STRENGTH 6 has a 41 % higher displacement (22.3 mm)
than run TPM 6. When the surface surcharge is placed on top the tunnel is pushed further
into the ground and the vertical invert displacements reduce for all analyses. Here again,
the displacement curves tend to follow a parallel pattern resulting in a more or less constant
loading-displacement ratio where the lining stiffness does not play an important role any-
more. The final displacement values are in the range of -12.5 mm (for run STRENGTH 1)
and 1.5 mm (for run STRENGTH 4).

30 TPM 6
STRENGTH 1
STRENGTH 2
Vertical displacement [mm]

STRENGTH 3
20
STRENGTH 4
STRENGTH 5
STRENGTH 6
10

End of excavation
0

End of surface
loading
-10

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Increment

Figure 8.71: Vertical invert displacement for variation in development of stiffness and strength
with Ko = 0.5

Another investigation focused on the impact of the shotcrete deformability on the vertical
displacement of the lining intrados at the tunnel crown adopting a Ko -profile of 0.5, as
illustrated in Fig. 8.72. No significant difference can be distuingished during the first loading
phase of the tunnel in the course of excavation (Inc 3 to 10). However, the difference becomes
a bit more pronounced during the surcharge loading with final values ranging from -134.3 mm
for run STRAIN 1, representing a relatively high deformability of shotcrete at early ages, to
-123.8 mm for run STRAIN 4, where the deformability is kept at a constant low level.

374
 0
TPM 6
-20 STRAIN 1
STRAIN 2

Vertical displacement [mm]

-40
STRAIN 3
End of excavation STRAIN 4
-60

-80

-100

-120
End of surface
-140
loading

0 5 10 15 20 25 30 35 40 45 50 55
Increment

Figure 8.72: Vertical crown displacement for variation in shotcrete deformability with Ko =
0.5

Tab. 8.30 summarizes the vertical and horizontal displacements of various points on the
lining intrados for different analyses concerning creep, shrinkage and hydration temperature
(Inc 100) and compares the results with reference run TPM 6 (Inc 50). According to these
results, the vertical crown displacements are increased for all analyses, but the inclusion of
creep shows the smallest impact. Taking into account all the time-dependent aspects in
run C+SH+T, leads to the largest absolute displacement of -139.2 mm. This trend is the
opposite for the horizontal sidewall and vertical invert displacements, which tend do decrease
compared to reference run TPM 6.

Displacement TPM 6 CREEP SHRINKAGE TEMPERATURE C+SH+T

Crown - v (mm) -131.4 -132.5 -136.8 -132.9 -139.2
Sidewall - u (mm) 50.7 50.5 47.9 49.4 46.3
Invert - v (mm) -11.6 -11.2 -9.6 -10.6 -8.5

Table 8.30: Lining displacements of performed analyses for creep, shrinkage and hydration
temperature adopting Ko = 0.5

Fig. 8.73 shows the development of the horizontal displacements of the lining intrados at
the right sidewall of the tunnel during the excavation and surface loading phase from Inc 5 to
Inc 50 for an adopted Ko -profile of 1.5. Due to the high horizontal stresses all the illustrated
time runs indicate a movement of the right sidewall towards the centre of the tunnel during
the excavation phase. At the end of tunnel excavation (Inc 10), the linear elastic model (LE)
predicts a horizontal displacement of -6.0 mm, which is the same for the non-linear time-

375
independent run (NL). The predictions from the HME approach are very similar to the LE
run although the lining stiffness is much smaller, showing a horizontal movement of -8.0 mm.
Here again, the largest horizontal displacement is predicted by the time-dependent run TPM
1, where the shotcrete is very young. Displacements reach up to -52.6 mm at Inc 10. With
ageing of the shotcrete these displacements reduce significantly, having a value of -32.7 mm
for run TPM 2 and -9.3 mm for run TPM 5. As illustrated in the figure, for the time runs
TPM 7 to 14 there is no significant difference in the horizontal displacements due to the
fact that the lining stiffness had developed to a relatively high level. On the onset of the
surface loading, the general movement of the tunnel at the right sidewall is outwards and
horizontal displacements tend to become positive again around Inc 30 to 40. At the final stage
of the construction sequence the linear elastic model (LE) predicts horizontal movements of
10.1 mm, whereas run TPM 1 still shows a negative value of -43.7 mm.

376
 10
End of excavation
End of surface
loading

Horizontal displacement [mm]

0

-10

-20 LE
HME
-30
NL
TPM 1
TPM 2
-40
TPM 3
TPM 4
-50 TPM 5
TPM 6
-60
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Increment
12

End of surface
loading
8
Horizontal displacement [mm]

HME
End of excavation TPM 7
0
TPM 8
TPM 9
TPM 10
-4 TPM 11
TPM 12
TPM 13
-8 TPM 14

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Increment

Figure 8.73: Horizontal displacement of right tunnel sidewall for various time runs adopting
Ko = 1.5

As explained earlier, the development of the lining stiffness has a big influence on the
deformational behaviour of the tunnel shell, as can be seen in Fig. 8.74. For a retarded
increase in stiffness, represented by runs STRENGTH 3 and 4 (s = 0.4 and s = 0.5), the
horizontal displacements of the lining intrados of the right sidewall increase up to -23.2 mm
at the end of excavation in Inc 10 and are therefore 2 to 3 times larger than for the reference
run TPM 6 (-7.5 mm). By reducing the stiffness and strength of the shotcrete by 25 and
50 %, horizontal displacements increase up to -9.7 and -14.6 mm respetcively. After placing
the surface surcharge during Inc 11 to 50, the horizontal displacements of the right sidewall
become positive for all investigated runs apart from run STRENGTH 4. These displacements

377
are in the range of 0.9 to 10.6 mm (STRENGTH 1).

12

8
End of excavation End of surface

Horizontal displacement [mm]

4
loading
0

-4

-8

-12
TPM 6
STRENGTH 1
-16 STRENGTH 2
STRENGTH 3
-20
STRENGTH 4
-24 STRENGTH 5
STRENGTH 6
-28
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Increment

Figure 8.74: Horizontal sidewall displacement for variation in development of stiffness and
strength with Ko = 1.5

A slight difference in the vertical displacements of the tunnel invert can be identified
when applying different shotcrete deformabilities, as can be seen in Fig. 8.75. When adopting
a Ko -profile of 1.5, a very small upwards movement of the tunnel invert can be expected
for all performed analyses. Reference run TPM 6 shows a value of 0.7 mm at the end of
excavation. For runs STRAIN 1 to 4 this value varies from 0.3 to 0.9 mm depending on the
adopted development of peak strains representing either relatively soft or very brittle material
behaviour regarding the possible strain levels of the shotcrete. This displacement range seems
to keep constant during the surface loading phase, where the tunnel is pushed downwards. At
the end of the construction process in Inc 50, the vertical invert displacements show negative
values ranging from -6.0 to -5.3 mm.

378
 2

End of excavation TPM 6

STRAIN 1
0 STRAIN 2

Vertical displacement [mm]

STRAIN 3
STRAIN 4

-2

-4

-6 End of surface
loading

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Increment

Figure 8.75: Vertical invert displacement for variation in shotcrete deformability with Ko =
1.5

In Tab. 8.31 the vertical and horizontal lining displacements of the tunnel crown, sidewall
and invert are listed for the performed analyses which take into account creep, shrinkage and
hydration temperature of the shotcrete, adopting a Ko -profile of 1.5. For the run C+SH+T a
significant impact of these time-dependent aspects can be identified for the vertical displace-
ments of the tunnel invert, which are reduced by 50 %. As in the case of Ko = 0.5, creep
surprisingly seems to be of minor importance.

Displacement TPM 6 CREEP SHRINKAGE TEMPERATURE C+SH+T

Crown - v (mm) -34.7 -35.4 -40.1 -36.9 -42.9
Sidewall - u (mm) 9.9 9.7 7.0 8.4 5.4
Invert - v (mm) -5.5 -5.3 -3.9 -4.5 -2.7

Table 8.31: Lining displacements of performed analyses for creep, shrinkage and hydration
temperature adopting Ko = 1.5

In Fig. 8.76 a comparison of different lining displacements for both Ko -profiles are pre-
sented, in order to show the effect of the initial stress conditions on the deformational be-
haviour of the tunnel lining. Run TPM 6 serves as a reference run and displacements are
plotted for the intrados of the tunnel crown and the right sidewall. At the end of excavation
in Inc 10, the vertical crown displacement for Ko = 0.5 shows a value of -5.8 mm, whereas for
Ko = 1.5 the tunnel crown moves upwards resulting in a displacement of 9.5 mm. However,
at the end of the surface loading in Inc 50 this effect is even more pronounced, with a dif-
ference in displacements of 97 mm, although both types of analyses predict negative values

379
meaning that the tunnel was pushed into the ground when placing the surface surcharge. A
similar mechanism can be identified for the horizontal displacements of the right sidewall of
the tunnel. In the case of Ko = 0.5, this movement goes up to 6.1 mm. The movement is
contrary for Ko = 1.5, where the sidewall moves -7.5 mm towards the centre of the tunnel
due to the high horizontal stresses in the ground. After loading the tunnel from the surface
a total difference in horizontal displacements of 41 mm is indicated.

60
End of excavation
40
41 mm
20

0
End of surface
Displacement [mm]

-20 loading

-40

-60
All runs TPM 6
-80
97 mm
Crown, v with Ko=0.5
-100
Sidewall, u with Ko=0.5
-120 Crown, v with Ko=1.5
Sidewall, u with Ko=1.5
-140

0 10 20 30 40 50 60 70
Increment

Figure 8.76: Comparison of vertical crown displacements v and horizontal sidewall displace-
ments u for both Ko -profiles for time run TPM 6

8.7.3.2 Excavation Type 2

The development of the vertical displacement of the lining intrados at the tunnel crown is
shown for excavation Type 2 in Fig. 8.77, starting at the installation of the upper shell in Inc
7. Furthermore, the results from various time runs are compared with the linear elastic (LE)
and HME approach. It can be seen that during the excavation of the bench, the tunnel crown
tends to move downwards towards the centre of the tunnel showing negative displacement
values for all runs. The runs LE, HME, TPM 5 and 9 predict displacements that are very
close together of about -10.0 mm. However, run TPM 1 with its low stiffness of the young
shotcrete, predicts movements of up to -29.2 mm. This trend of deformation is reversed
during the excavation of the invert region of the tunnel from Inc 14 to 25, where the tunnel
crown starts to move upwards resulting in a positive displacement of about 10 mm for the
linear elastic run (LE). This crown movement is not so pronounced for run TPM 1, which
still shows a negative displacement of -23.2 mm at the end of the complete tunnel excavation
in Inc 25. The HME approach and run TPM 5 predict almost the same displacement value
of about 5.7 mm. When the surface surcharge is applied from Inc 26 to Inc 65, the whole
tunnel is pushed into the ground and therefore displacements start to reduce again. Run

380
TPM 1 shows relatively large crown displacements of -70.2 mm, since the stiffness is still at
a very low level. The model predictions of the remaining analyses vary from -30.0 mm for
the linear elastic model (LE) to -40.7 mm for run TPM 5. Here again, it appears that the
load-displacement ratio during the surface loading phase seems to be of a similar magnitude
for all the performed runs since the curves show a parallel position to each other.

10
LE
Vertical displacement [mm] 0 HME
TPM 1
-10 TPM 5
End of excavation TPM 9
-20

-30

-40
Installation of invert
-50 End of surface
loading
-60

-70

-80
0 10 20 30 40 50 60 70 80
Increment

Figure 8.77: Vertical crown displacement for various time runs adopting Ko = 1.5

To study the effect of the difference in age of shotcrete of the upper and lower tunnel
shell on the displacement behaviour, results for the time runs TPM 1 to 4 are plotted in
Fig. 8.78. During the excavation of the tunnel from Inc 7 to 25 no significant difference is
obtained. However, during the loading phase of the tunnel from the surface surcharge, a
slight difference in vertical crown displacements with a maximum value of 4.4 mm can be
identified. All curves for the time-dependent model remain far below the linear elastic model
(LE) and the HME approach, as already discussed for the previous figure.

381
 10
LE
0
HME
TPM 1

Vertical displacement [mm]

-10 TPM 2
End of excavation TPM 3
-20 TPM 4

-30

-40
Installation of
End of surface
-50 invert
loading
-60

-70

-80
0 10 20 30 40 50 60 70 80
Increment

Figure 8.78: Vertical crown displacement for runs TPM 1 to 4 adopting Ko = 1.5

Fig. 8.79 illustrates the development of the horizontal displacements of the lining intrados
at the intersection point of the tunnel shell with the temporary invert (see also Fig. 8.17).
Here a different mechanism is obtained for time run TPM 1 due to the low stiffness of the
young shotcrete. The intersection point tends to move towards the centre of the tunnel,
resulting in a negative displacement of -3.4 mm at the end of the bench excavation in Inc 12.
Conversely, the linear elastic model (LE), the HME approach and the time runs TPM 5 and
9 predict a positive movement of this point to the right with a maximum value of 5.5 mm for
run TPM 9. During excavation of the tunnel invert from Inc 14 to 25, all models indicate
a negative increase in horizontal displacements leading to movement towards the inside of
the tunnel. The linear elastic model (LE) shows a value of -13.8 mm and the HME approach
-16.9 mm in Inc 25. Here again, the run TPM 1 shows the largest displacement value of
-32.8 mm. When placing the surface surcharge, the tunnel lining intrados starts to move to
the right again with a maximum final displacement of 2.1 mm for the linear elastic model
(LE), almost zero displacement (0.4 mm) for the HME approach and -17.4 mm for run TPM
1. The analyses TPM 5 and 9 show final displacements which are very close to zero (-1.3 and
1.3 mm respectively).

382
 10

Horizontal displacement [mm]

0

-5 End of excavation

-10 Installation
of invert
-15

-20

-25 LE
HME End of surface
-30 TPM 1 loading
TPM 5
-35
TPM 9
-40
0 10 20 30 40 50 60 70
Increment

Figure 8.79: Horizontal displacement at the intersection of temporary invert with tunnel
lining for various time runs adopting Ko = 1.5

As shown in Fig. 8.80, a small difference in the displacement curves can be identified for
runs TPM 1 to 4 during the surface loading of the tunnel. Similar to the case for the vertical
crown displacement, these differences lie in the range of 1.0 to 4.0 mm.

-4 Installation of invert
Horizontal displacement [mm]

-8
End of surface
-12 loading

-16

-20
End of excavation
-24
TPM 1
-28 TPM 2
TPM 3
-32
TPM 4
-36
0 10 20 30 40 50 60 70
Increment

Figure 8.80: Horizontal sidewall displacement for runs TPM 1 to 4 adopting Ko = 1.5

The impact of the development of the shotcrete stiffness and strength with time on the
deformational behaviour of the tunnel lining for excavation Type 2 can be studied in Fig. 8.81,
where results for the vertical displacement of the lining intrados at the tunnel crown are

383
plotted. Run STRENGTH 1 has a slightly stiffer material response of the shotcrete due
to its cement parameter s = 0.1. Vertical movements seem to be very close to those from
the reference run TPM 7, with a maximum difference of 3.6 mm at the end of the complete
tunnel excavation in Inc 25. Run STRENGTH 3 simulates a retarded increase in stiffness
and strength (s = 0.4) and therefore displacements remain negative at Inc 25 with a value
of -25.7 mm. A reduction in the shotcrete stiffness of 25 % leads to a softer lining behaviour,
where the crown displacements at the end of excavation are almost zero (0.4 mm). However,
during the construction phase of surface loading, the tunnel crown moves down vertically
towards the centre of the tunnel resulting in negative displacements which lie in the range of
-33.4 (STRENGTH 1) to -69.3 mm (STRENGTH 3).

10
TPM 7
0 STRENGTH 1
STRENGTH 3
Vertical displacement [mm]

-10 STRENGTH 5
End of excavation
-20

-30

-40
Installation of invert
-50
End of surface
loading
-60

-70

-80
0 10 20 30 40 50 60 70 80
Increment

Figure 8.81: Vertical crown displacement for various developments of stiffness and strength
adopting Ko = 1.5

Tab. 8.32 summarises the lining displacements of various points at the lining intrados at
28 days after the end of surface loading (Inc 100), taking into account creep, shrinkage and
hydration temperature of the shotcrete. Results for run CREEP are very close to the ones
from reference run TPM 7. An increase of about 3 to 5 mm can be expected for the tunnel
crown and sidewall when considering shrinkage of shotcrete (run SHRINKAGE). However,
for this run the invert displacement is reduced by 1.4 mm. A similar picture is obtained
when investigating the effect of hydration temperature on the deformational behaviour of the
lining (run TEMPERATURE). For the tunnel crown and sidewall an increase of about 1 to
2 mm has been calculated. The reduction for the lining displacement of the invert reaches
0.5 mm and is therefore negligible. In general, the biggest difference in lining displacement
can be expected when coupling all the three time-dependent mechanisms in run C+SH+T.
The vertical crown displacement reaches a value of 46.1 mm, which is 7 mm larger than for

384
run TPM 7. At the tunnel sidewall, the horizontal displacement increases from 0.7 to 4.8 mm.
Once again, the vertical displacement of the lining invert is reduced by 2.2 mm taking a final
value of 4.4 mm.

Displacement TPM 7 CREEP SHRINKAGE TEMPERATURE C+SH+T

Crown - v (mm) -39.1 -39.8 -44.4 -40.4 -46.1
Sidewall - u (mm) -0.7 -1.0 -3.7 -1.8 -4.8
Invert - v (mm) -6.6 -6.4 -5.2 -6.1 -4.4

Table 8.32: Lining displacements of performed analyses for creep, shrinkage and hydration
temperature

For completeness, the vertical displacement of the temporary invert, which is installed in
Inc 7, should be discussed here briefly. Tab. 8.33 summarises the results for the performed
analyses with different constitutive behaviour for shotcrete. For the stiff linear elastic model
(LE) the maximum displacements at the end of the bench excavation (Inc 12) are the smallest
with 14.0 mm, followed by the non-linear approach (NL) with a value of 15.0 mm. Runs TPM
1 and STRENGTH 3 represent a very soft lining behaviour with a low stiffness and therefore
displacements increase up to 20.1 mm. The shotcrete deformability (runs STRAIN 3 and 4)
seems to have a minor influence on the behaviour of the temporary invert, compared to their
reference run TPM 7 (15.6 mm).

385
 Run Maximum vertical displacement v
of temporary invert
(mm)
Linear elastic (LE) + 14.0
HME + 15.5
Non-linear (NL) + 15.0
TPM 1 + 19.6
TPM 5 + 16.0
TPM 9 + 15.5
STRENGTH 1 + 15.4
STRENGTH 3 + 20.1
STRENGTH 5 + 16.6
STRAIN 3 + 15.7
STRAIN 4 + 15.8

Table 8.33: Maximum vertical displacement v of temporary invert at the end of bench exca-
vation (Inc 12)

8.7.3.3 Excavation Type 3

In Fig. 8.82 the development of the horizontal displacements of the lining intrados at the left
tunnel sidewall are shown for various analyses comparing simple elastic models with sophis-
ticated time-dependent approaches. All runs have in common that during the excavation
of the left side gallery the lining intrados is moving to the right towards the centre of the
tunnel resulting in positive deformations. As expected, the stiff linear elastic model (LE)
predicts the smallest displacement with a value of 14.4 mm in Inc 12. The HME approach is
slightly softer and leads to a displacement of 16.4 mm. Including non-linearity of hardened
shotcrete at 28 days makes a difference of about 1.6 mm, resulting in a total displacement of
16.0 mm. However, the time-dependent run TPM 1 shows a relatively large lining movement
of up to 31.1 mm, governed by the low stiffness of the shotcrete during excavation. Runs
TPM 9 tends to follow a similar behaviour as the HME approach, whereas TPM 5 predicts a
slightly larger displacement of 18.0 mm. During the excavation of the right side gallery, this
mechanism is reversed with the tunnel lining moving to the left and displacements reducing
for all runs. In Inc 25 the linear elastic model results in a horizontal displacement of just
5.6 mm, whereas, the value for run TPM 1 is 25.8 mm. When loaded from the surface during
Inc 26 to 66, the loading-displacement curves for all runs tend to be parallel to each other
leading to a negative displacement of about -11.0 mm for runs LE, HME, NL and TPM 9.
TPM 5 has a slightly smaller displacement in Inc 65 with a value of -8.5 mm. Run TPM

386
1 predicts positive displacement after placing the surface surcharge and leads to a horizon-
tal movement of 8.2 mm. As noted in the previous case of excavation Type 2, the elapsed
time Δt in Inc 13, simulating the longitudinal distance between the left and right tunnel
face and therefore different shotcrete ages of the tunnel shell, is of minor importance to the
deformational behaviour of the lining and results vary within ±4.0 mm.

40
End of left side
LE
gallery excavation
Horizontal displacement [mm] HME
NL
30 TPM 1
TPM 5
TPM 9
20

10

End of surface
0 loading
End of right side
-10
gallery excavation

0 10 20 30 40 50 60 70 80
Increment

Figure 8.82: Horizontal displacements of left sidewall

Fig. 8.83 illustrates the development of the horizontal displacements of the lining intrados
at the right tunnel sidewall, starting at the installation of the right shell in Inc 14. An
interesting fact is that all runs, apart from run TPM 1, predict a positive lining movement
to the right, where the linear elastic model (LE) and the HME approach show an almost
identical displacement pattern, with a value of 5.2 to 5.6 mm in Inc 25 and 21.5 to 22.4 mm
at the end of surface loading in Inc 65. Non-linear behaviour of hardened shotcrete (NL) leads
to a larger displacement of 7.4 mm in Inc 25 and 24.5 mm in Inc 65. The displacement curve
for run TPM 5 represents a somewhat softer lining behaviour. However, the deformational
mechanism for run TPM 1 appears to be different than for the other analyses. During the
excavation of the right side gallery, the right sidewall tends to move to the left towards the
centre of the tunnel leading to a horizontal displacement of -16.0 mm at the end of excavation
(Inc 25). During the surface loading the lining is pushed to the right again compensating
almost all the previous deformation and resulting in a total displacement of 0.25 mm in Inc
65.

387
 30

LE
24
HME

Horizontal displacement [mm]

NL
18
TPM 1
TPM 5
12 End of surface
TPM 9
loading
6

0
End of right side
-6 gallery excavation

-12

-18

0 10 20 30 40 50 60 70 80
Increment

Figure 8.83: Horizontal displacements of right sidewall

The vertical displacements of the top intersection point of the temporary sidewall with the
tunnel lining (see Fig. 8.17) during the course of excavation and surface loading are plotted
in Fig. 8.84. As in the previous case in Fig. 8.83, almost all model predictions show a positive
upwards movement of this particular point during the excavation of the left side gallery.
However, this movement is the opposite for run TPM 1, which has a vertical displacement of
-5.1 mm in Inc 12. The linear elastic model and the HME approach predict very similar results
with 12.2 and 13.5 mm respectively. The largest displacement is obtained by the non-linear
model (NL) with a value of 15.1 mm. Displacement values for TPM 5 and 9 lie in the range of
13.7 to 15.0 mm. During the excavation of the right side gallery, the upper part of the tunnel
shell tends to move downwards towards the centre of the tunnel, where displacement values
are located within -4.7 (NL) and -12.7 mm (TPM 5). Displacements according to run TPM
1 increase up to -42.6 mm. During the surface loading phase, these displacements increase
further as the tunnel is pushed into the ground. In Inc 65 the linear elastic model (LE)
predicts a vertical lining displacement of -44.1 mm and the HME approach -53.8 mm. The
time-dependent runs TPM 5 and 9 show values of -57.4 and -51.4 mm respectively. The low
lining stiffness of TPM 1 leads to the largest displacement value of -86.3 mm.

388
 45
End of left side LE
30
gallery excavation HME

Vertical displacement [mm]

15 NL
TPM 1
0 TPM 5
TPM 9
-15

-30

-45

-60
End of surface
End of right side loading
-75 gallery ecavation
-90

0 10 20 30 40 50 60 70 80
Increment

Figure 8.84: Vertical displacement of intersection point of temporary sidewall with tunnel
lining

Fig. 8.85 investigates the influence of the development of the shotcrete stiffness and
strength on the vertical displacements of the lining intrados located at the top intersec-
tion point between the temporary sidewall and the tunnel shell. It can be seen that run
STRENGTH 3, with its retarded increase in stiffness (s = 0.4), shows a similar behaviour
as run TPM 1 in Fig. 8.84. Reference run TPM 7, STRENGTH 1 and 5 tend to move up-
wards during the first excavation for the left side gallery leading to vertical displacements
between 10.6 and 15.4 mm in Inc 12. At this stage of construction, STRENGTH 3 predicts
a negative value of -9.8 mm. When excavation of the right side gallery is continued, the
lining point moves vertically downwards for all the analyses. During the surface loading this
movement is continued in the negative direction, leading to -49.4 mm for the slightly stiffer
run STRENGTH 1 (s = 0.1) in Inc 65. The reduction of the stiffness by 25 % results in a
vertical displacement that is 12 % larger (-62.2 mm) than the one from reference run TPM 7
(-55.4 mm) at the end of surface loading.

389
 45
End of left side
30 TPM 7
gallery excavation
STRENGTH 1
15 STRENGTH 3
STRENGTH 5

Vertical displacement [mm]

0

End of surface
-15
loading
-30

-45

-60

-75 End of right side

gallery excavation
-90

0 10 20 30 40 50 60 70 80
Increment

Figure 8.85: Vertical displacement of intersection point of temporary sidewall with tunnel
lining for variousn developments of stiffness and strength

The aim of Fig. 8.86 is to show briefly the effect of the shotcrete deformability on the hor-
izontal displacements of the lining intrados of the left tunnel sidewall. During the excavation
of the left side gallery, no significant difference in the analyses is noticeable. Run TPM 7
shows a vertical displacement of 18.0 mm, whereas run STRAIN 4, representing a constant
low peak strain, predicts a displacement of 16.9 mm. This difference becomes slightly more
pronounced during the excavation of the right side gallery and the surface loading leading to
a maximum difference of 1.4 mm in Inc 65.

390
 30

End of left side TPM 7

25
gallery excavation STRAIN 3
20 STRAIN 4

Horizontal displacement [mm]

15

10

5
End of surface
0 loading
End of right side
-5 gallery excavation

-10

0 10 20 30 40 50 60 70 80
Increment

Figure 8.86: Horizontal displacement of left sidewall for various shotcrete deformabilities

Tab. 8.34 provides some information about the effect of creep, shrinkage and hydration
temperature of the shotcrete on the lining displacement at various points of the tunnel shell.
Results at 28 days after the end of surface loading (Inc 100) are compared with reference run
TPM 7 (Inc 65). Surprisingly, the results taking into account creep of the young shotcrete
show very little difference compared to run TPM 7. When considering shrinkage, the hor-
izontal displacements at the left and right tunnel sidewall are reduced by 2.6 and 2.4 mm
respectively. The vertical crown displacement is slightly larger, taking a value of 60.4 mm.
The influence of hydration temperature on the deformational lining behaviour is slightly less
pronounced. Horizontal displacements at both sidewalls reduce by roughly 1 mm and vertical
displacements increase by 1.5 mm for the tunnel crown. Here again, the incorporation of all
the three time-dependent mechanisms, i.e. creep, shrinkage and hydration temperature (run
C+SH+T), has a bigger impact on the displacements for all the three points. At the left and
right sidewall these displacements are reduced by 3.6 mm. An increase of 7.1 mm is predicted
for the vertical displacement at the tunnel crown.

Displacement TPM 7 CREEP SHRINKAGE TEMP. C+SH+T

Left sidewall - u (mm) -9.2 -9.0 -6.6 -8.2 -5.4
Right sidewall - u (mm) 19.9 19.7 17.5 19.1 16.3
Crown - v (mm) -55.4 -56.1 -60.4 -56.9 -62.5

Table 8.34: Lining displacements of performed analyses for creep, shrinkage and hydration
temperature

391
 Finally, Tab. 8.35 contains a summary regarding the maximum horizontal displacement u
at the centre of the temporary sidewall at the intermediate construction stage Inc 12. The
obtained values range from -29.0 (LE) to -45.0 mm (STRENGTH 3).

Run Maximum horizontal displacement u

of temporary sidewall
(mm)
Linear elastic (LE) - 29.0
HME - 32.9
Non-linear (NL) - 32.3
TPM 1 - 43.8
TPM 5 - 35.3
TPM 9 - 34.0
STRENGTH 1 - 33.6
STRENGTH 3 - 45.0
STRENGTH 5 - 36.8
STRAIN 3 - 34.9
STRAIN 4 - 34.5

Table 8.35: Maximum horizontal displacement u at the centre of the temporary sidewall at
the end of the left side gallery excavation (Inc 12)

8.7.4 Soil stresses

Tunnel excavation leads to subsurface movements and rearrangement of stresses in the ground
adjacent to the tunnel. These changes in effective stresses and pore water pressures will be
discussed in detail in the course of this section for excavation Type 1, adopting both Ko -
profiles. Furthermore, results will be shown for two sets of analyses, where the mechanical
behaviour of the lining is of crucial importance: the linear elastic model LE, representing a
very stiff shotcrete shell, and run TPM 1, where the behaviour of the tunnel lining is mainly
governed by the material properties of very young shotcrete.

Fig. 8.87 shows the contours of horizontal effective stresses in the ground near the tunnel
opening at the end of excavation (Inc 10) for the linear elastic model LE (top) and run TPM
1 (bottom), assuming Ko = 0.5. For the linear elastic approach, horizontal stresses near the
tunnel crown rise up to -130 kPa (compared to -82 kPa for the initial stress state), which is
almost the same value as for the tunnel sidewalls (-140 kPa). Near the invert of the tunnel
effective stresses increase up to -200 kPa. Due to the large displacement of the lining for run
TPM 1, the stress conditions after excavation are different. Effective horizontal stresses at

392
the tunnel crown and invert increase significantly showing values of -650 kPa. A similar value
of -580 kPa is obtained for the tunnel sidewall, compared to -102 kPa for the initial stress
conditions.

393
Figure 8.87: Effective horizontal stresses in ground after excavation of tunnel (Inc 10) for
linear elastic model (top) and run TPM 1 (bottom) adopting Ko = 0.5

394
 In Fig. 8.88 the contours of the vertical effective stresses in the ground after tunnel exca-
vation (Inc 10) are illustrated for both lining models. For the linear elastic model LE (top),
the vertical stresses near the tunnel crown reduce to -110 kPa compared to an initial vertical
stress of -163 kPa. A similar reduction is achieved for the tunnel invert, with a vertical effec-
tive stress of -180 kPa after excavation. As expected, the stresses near the tunnel sidewalls
show an increase from -204 kPa up to -380 kPa. Once again, the stress rearrangement for run
TPM 1 is highly influenced by the large displacements of the lining due to the soft behaviour
of shotcrete at early ages. Vertical effective stresses go up to -260 kPa for the tunnel crown
and invert. A high stress value is obtained for the tunnel sidewalls reaching almost -1300 kPa,
with the soil being close to failure.

395
Figure 8.88: Effective vertical stresses in ground after excavation of tunnel (Inc 10) for linear
elastic model LE (top) and run TPM 1 (bottom) adopting Ko = 0.5

396
 The contours of the sub-accumulated pore water pressures induced by tunnel excavation
(i.e. the pore water pressures at Inc 0 subtracted from the pore water pressures at Inc 10)
are plotted for the linear elastic lining model LE (top) and run TPM 1 (bottom) in Fig. 8.89,
assuming Ko = 0.5. It can be observed that due to the inwards movement of the tunnel
lining, positive changes in pore water pressures (i.e. suctions) are induced into the ground
in the close vicinity of the tunnel. This pressure change is of about 50 kPa near the tunnel
crown assuming linear elastic shotcrete behaviour (LE) and 162 kPa for the tunnel invert.
An intermediate pressure change of 105 kPa can be expected in the vicinity of the tunnel
sidewall. The results from run TPM 1 show in general a similar trend as the LE approach
by introducing positive changes in pore water pressures in the ground due to the movement
of the tunnel lining. However, since these lining movements are relatively large, the changes
in pore water pressure are of a much larger magnitude and appear to be unrealistic. Near
the tunnel crown a pressure change of 380 kPa is predicted by run TPM 1, which is a similar
value as for the tunnel invert (400 kPa). However, in the ground near the tunnel sidewall a
larger change in pore water pressure of 690 kPa is obtained for run TPM 1. As can be seen
in Fig. 8.89, these larges pressure changes reduce quickly within the first 2 to 4 metres of
distance from the tunnel opening and tend to stabilise to a smaller value of about 70 kPa.

397
Figure 8.89: Sub-accumulated pore water pressures in ground after excavation of tunnel (Inc
10) for linear elastic model (top) and run TPM 1 (bottom) adopting Ko = 0.5

398
 Fig. 8.90 contains the contours of the horizontal effective stresses at the end of tunnel
excavation (Inc 10) for the linear elastic model LE (top) and run TPM 1 (bottom), adopting
a Ko -profile of 1.5. Applying a relatively stiff lining behaviour in the linear elastic approach
LE leads to an increased horizontal effective stress in the ground near the tunnel crown of
-510 kPa. In contrast, the horizontal stresses near the tunnel sidewall are reduced to -190 kPa
compared to -306 kPa for the initial stress state before excavation. Results show an increase in
horizontal stresses near the tunnel invert with a predicted value of -610 kPa, compared to an
initial stress of -368 kPa. Ground stresses are of a much larger magnitude when simulating the
tunnel lining with very young shotcrete in run TPM 1. For the tunnel crown, the horizontal
effective stresses rise up to -1150 kPa and to -1300 kPa in the vicinity of the tunnel invert.
Furthermore, run TPM 1 predicts an increase in horizontal stresses for the tunnel sidewall
with a value of -420 kPa after excavation in Inc 10.

399
Figure 8.90: Effective horizontal stresses in ground after excavation of tunnel (Inc 10) for
linear elastic model (top) and run TPM 1 (bottom) adopting Ko = 1.5

400
 The contours of the vertical effective stresses for Ko = 1.5 are illustrated in Fig. 8.91.
In the case of modelling the tunnel lining linear elastic (LE), the vertical stresses near the
tunnel crown increase somewhat from -163 kPa for the initial stress conditions to -200 kPa at
the end of excavation. A slightly smaller increase occurs at the tunnel invert, where vertical
effective stresses of -260 kPa are predicted. At the tunnel sidewall, these stresses reach a
value of -350 kPa. Stress conditions are significantly different for run TPM 1 caused by the
extensive lining movement at early shotcrete ages. Near to the tunnel crown the vertical
effective stresses increase up to -460 kPa. A similar stress rise occurs for the tunnel invert
with a stress value of -550 kPa at the end of excavation. A substantial increase in vertical
effective stresses with run TPM 1 occurs adjacent to the tunnel sidewall, reaching a value of
-910 kPa.

401
Figure 8.91: Effective vertical stresses in ground after excavation of tunnel (Inc 10) for linear
elastic model LE (top) and run TPM 1 (bottom) adopting Ko = 1.5

402
 Finally, the changes in pore water pressures in the ground from Inc 0 to Inc 10, adopting
Ko = 1.5, are depicted in Fig. 8.92. Once more, the contours of the sub-accumulated pore
water pressures for the linear elastic lining model (top) are compared with results from run
TPM 1 (bottom). As in the previous case of K0 = 0.5, positive changes in the pore water
pressures (i.e. suctions) in the ground near the tunnel opening are to be expected due to the
inwards movement of the tunnel lining. Applying the linear elastic model (LE) leads to a
pressure change of 150 kPa for the tunnel crown and 120 kPa for the tunnel sidewalls. Near
the invert of the tunnel these changes are larger in magnitude and reach a value of 230 kPa.
A dramatic change in pore water pressure was obtained for run TPM 1. Near the tunnel
crown the pressure change rises up to 540 kPa and 460 kPa for the tunnel sidewall. However,
an extreme value was predicted for the pore water pressures in the vicinity of the tunnel
invert. A positive pressure change of up to 700 kPa is introduced to the ground. In general,
this high induced suctions tend to decrease with increasing distance from the tunnel opening
and stabilise to a more realistic value of 50 kPa.

403
Figure 8.92: Sub-accumulated pore water pressures in ground after excavation of tunnel (Inc
10) for linear elastic model (top) and run TPM 1 (bottom) adopting Ko = 1.5

404
 From these observations it can be concluded that the deformational behaviour of the
tunnel lining highly affects the rearrangement of stresses and pore water pressures in the
ground in the vicinity of the excavated tunnel. Since the movement of a tunnel shell is
strongly influenced by the material properties of the early age shotcrete, the elapsed time
during tunnel construction plays an important role. However, it was shown that for very
young shotcrete and quick loading conditions (run TPM 1) quite unrealistic stresses and
pore water pressure changes were predicted as a consequence of the associated high volume
loss. For further analyses it is therefore recommended, to simulate this interaction between
soil and tunnel lining in a more sophisticated way in 3D (and probably with the use of
interface elements) in order to predict realistic results regarding the effective stresses and
pore water pressures in the ground.

8.8 Summary
This chapter presented results of the 2D finite element analyses of a typical tunnel to be
excavated under undrained conditions in London Clay. In this investigation, the mechanical
behaviour of the installed shotcrete tunnel lining was of special interest and therefore the
newly developed constitutive model for sprayed concrete, presented in Chapter 7, was adopted
to simulate the complex material behaviour of shotcrete. Results from these analyses were
compared with much simpler approaches for the simulation of sprayed concrete that are
still state-of-the-art in engineering practice, i.e. linear elastic material behaviour and the
Hypothetical Modulus of Elasticity (HME). Furthermore, the simulation of tunnel excavation
followed three different excavation schemes (full face, bench-invert and sidewall drift), each
of them leading to different stress generations within the tunnel lining. Construction of
a possible structure above the tunnel was considered by applying some surface surcharge,
which results in heavy loading of the tunnel shell. The following aspects were adressed in the
extensive parametric study performed:

 In the case of full face excavation of the tunnel two Ko -profiles with a value of 0.5 and
1.5 have been adopted leading to different initial stress conditions in the soil before
the start of excavation. It was shown, how this difference influences the mechanical
behaviour of the tunnel lining.

 Two different types of shotcrete have been adopted: plain shotcrete in the case of full
face excavation and steel-fibre reinforced sprayed concrete for staged tunnel excavation.
The difference in material behaviour is accounted for by a modification of the tensile
post-peak behaviour of each individual shotcrete type.

 Material parameters for young shotcrete were taken from various experimental data
available in the open literature. This model calibration involved the realistic repre-
sentation of the increase in stiffness and strength and the development of shotcrete
deformability with time.

405
 The elapsed time during tunnel excavation and surface loading, and therefore the load-
ing rate of the shotcrete, was one of the main parameters to be investigated. In the
calculations, shotcrete ages at the end of the complete construction process ranged from
only 2 hours up to 2 months, simulating the transition from fast to slow construction.

 The model parameters were varied in a controlled manner in order to capture the
large scatter usually present in most of the experimental data for young shotcrete.
Furthermore, an attempt was made to account for uncertainties regarding the material
quality of shotcrete that can often be encountered in engineering practice.

 Additional analyses focusing on creep, shrinkage and hydration temperature induced

deformation of sprayed concrete have been carried out for all of the three excavation
types. Unfortunately, material parameters for these calculations had to be estimated
from the very limited amount of experimental data available in the literature.

 The mechanical behaviour of the tunnel lining has been analysed extensively by dis-
cussing in detail the obtained results with respect to stress distributions within the
shotcrete shell for selected cross sections, the maximum utilization factor (= stress
level) during construction as a measure of safety against failure and the displacements
of particular points on the lining intrados.

 Possible damage to adjacent buildings and surface structures due to tunnel excavation
has been investigated by studying the predicted surface settlement troughs and the
volume loss obtained at the end of excavation.

 Stress rearrangements and changes in pore water pressures in the ground as a conse-
quence of tunnel construction have been discussed for the soil immediately adjacent to
the tunnel and the interaction problem of soil-lining has been addressed.

406
Chapter 9

Summary and conclusions

9.1 Introduction
The main aim of this research project was to obtain a better understanding of NATM tun-
nelling in soft ground conditions by applying the finite element method, with a particular
focus on the time-dependent, mechanical behaviour of the shotcrete tunnel lining. There-
fore, it was necessary at first to perform an extensive literature review in order to study
the fundamental characteristics of shotcrete and how material properties influence its be-
haviour. The next step in the research involved the choice of a proper constitutive model for
sprayed concrete for a realistic description of the mechanical behaviour in multiaxial loading
conditions. The methodology followed was to choose a well-known failure criterion that is
capable of capturing the important features of the material response of hardened concrete,
the Chen & Chen (1975) concrete model. This model was then improved by adopting two
independent yield surfaces for compression and tension and introducing time-dependent ma-
terial parameters for stiffness, strength and deformability. Furthermore, the model accounts
for the complex aspects of creep, shrinkage and hydration temperature of the cement paste
at early ages. Although published data for young shotcrete covering a complete set of tests
is very hard to find in the literature, it was possible to calibrate the model in a reasonable
and satisfactory way. The developed constitutive model for shotcrete was implemented into
the Imperial College Finite Element Program (ICFEP) adopting robust numerical algorithms
and was successfully used in the numerical analysis of different tunnel excavations in London
Clay. The main conclusions that can be drawn from this research project are summarised in
the following sections.

9.2 General remarks

A basic knowledge about material technology, installation techniques and the main mechan-
ical characterstics of shotcrete are of great importance for a detailed understanding of the
behaviour of shotcrete in tunnelling. Based on current literature the following key facts can

407
be summarised:

 Sprayed concrete is one of the main support elements for tunnels driven according to the
principles of the New Austrian Tunnelling Method (NATM). Directly after excavation,
shotcrete is sprayed at high pressure onto the tunnel walls providing temporary stability
to the opening. It is usually reinforced with conventional wire mesh or lattice girders,
but there is a tendency towards the innovative use of steel-fibres as reinforcement for
mechanical, practical and economical reasons.

 Shotcrete can be seen as a special type of concrete since casting and compacting are
performed in one single step. As any conventional cast concrete, shotcrete consists of
cement, aggregates and water. Each of these components has an important impact on
the mechanical behaviour of the shotcrete and has to be addressed in a proper mix
design. Additives such as accelerators, plasticisers and silica fume are commonly used
to change certain material properties, but they have to be treated with care.

 For the installation of shotcrete two different techniques exist - the dry-mix and the
wet-mix process. Both technologies differ significantly in material preparation, each of
them having its advantages and disadvantages. However, the tendency in tunnelling is
clearly towards the wet-mix process due to its better control of the shotcrete quality
(in particular of the water-cement ratio).

 The spraying technique is of great importance for achieving good material quality and
this requires excellent skills and training of the nozzle-operator. Rebound of shotcrete
is not only an economical loss but influences as well the mechanical behaviour of the
material. It leads to a slightly richer cement mix of the in-situ material and this has to
be taken into account in the mix design.

 Testing of young sprayed concrete at early ages is a difficult task to perform and no
commonly agreed techniques exist in the literature. Usually, shotcrete samples are
taken from the tunnel lining or are sprayed into boxes under field conditions, where the
compressive strength of the material is the main parameter of interest in the design.
For fresh shotcrete penetration needles and pull-out tests are often used to estimate the
compressive strength of the material at early ages of cement hydration.

 Hardened shotcrete in uniaxial compression and tension shows a highly non-linear ma-
terial behaviour, where crushing and cracking of the concrete govern the post-peak
behaviour. Increased compressive strengths can be expected for concrete in biaxial
and triaxial loading conditions, leading to smooth yield and failure surfaces. From the
available data in the literature it is difficult to quantify any anisotropy of shotcrete that
might be expected due to the spraying process.

 Shotcrete at early ages shows a relatively plastic and ductile material response with
low stiffness and strength. Furthermore, higher strain limits can be achieved due to the

408
 increased deformability of shotcrete at early stages of cement hydration. With curing
time, material behaviour becomes more and more brittle associated with an increase
in stiffness and strength. However, compressive and tensile strains at peak and failure
strengths reduce during cement hydration. The effect of compressive preloading of
shotcrete samples at early ages on the development of the compressive strength at later
stages is not clear. In the literature both types of results can be found, i.e. an increase
and a reduction in the strength values.

 Creep and relaxation are important aspects in shotcrete behaviour related to “time”,
depending on various factors, such as loading age, stress level, temperature, moisture
conditions and concrete composition. It is believed that creep in tunneling leads to a
reduction in the expected lining stresses and is therefore of a beneficial nature. De-
formations of young shotcrete that occur due to shrinkage and increased hydration
temperature might be important to consider in a realistic tunnel lining design, since
these two phenomena can lead to cracking of the shotcrete shell. Their origin lies within
the cement paste of the concrete and can be addressed in the mix design.

 A large number of empirical models based on experimental data exist for modelling
concrete taking into account the non-linear material behaviour before peak. However,
one big drawback is, that these models focus mainly on uniaxial stress conditions in
compression and tension and are difficult to use in a numerical analysis. Nonetheless,
simple concrete models have been developed within the framework of elasto-plasticity
that are capable of reproducing the main characteristics of the material behaviour in
multiaxial loading conditions. One of these models is the Chen & Chen (1975) concrete
model, where the material behaviour in compression and tension is governed by two
yield surfaces. At the onset of plastic deformation these two surfaces move together
simultaneously in the principal stress space and isotropic strain-hardening controls the
material behaviour before reaching peak.

 Cracking of concrete in tension is a relatively difficult phenomenon to model, since in

reality it is a highly discrete and localized problem that occurs in a concrete structure.
For simulating cracking in a numerical analysis two different approaches exist, i e. the
discrete and the smeared crack model. However, when treating cracking of concrete
within the framework of continuum mechanics, severe problems that are usually associ-
ated with strain-softening materials arise. A simple approach based on fracture energy
was presented to avoid spurious mesh dependency of the results obtained in this thesis.

 No well accepted framework exists in the literature for the modelling of concrete creep
and in particular for young shotcrete at early ages, which is a very creep active medium.
Apart from some basic mathematical functions, in most of the cases, a combination of
rheological units serves for the calculation of creep deformation in concrete structures
(Kelvin-, Maxwell- and Burger-model). By establishing complex viscosity and stiffness

409
 functions it is possible to fit model predictions to uniaxial compressive test data. How-
ever, very limited information is available for the extension of creep to 3D or how to
deal with creep of concrete in tensile stress conditions.

 Three simple models for predicting shrinkage deformation of concrete have been pre-
sented. They are all based mainly on experimental data and take into account factors
such as relative humidity, concrete strength in compression, geometry of the structure
to be investigated and concrete composition. The modelling of the increase in temper-
ature due to cement hydration in hardening concrete or shotcrete is a complex process
mostly performed for the analysis of relatively thick concrete structures. Little infor-
mation is available in the literature on how to incorporate these temperature effects in
a constitutive model for the practical analysis of shotcrete tunnel shells.

9.3 A constitutive model for shotcrete

The constitutive model for sprayed concrete developed in this research project is formulated
within the framework of elasto-plasticity and is generalised for the three-dimensional stress
space. Crack formation of the shotcrete under tensile stresses is taken into account by ap-
plying the smeared crack concept and therefore cracks are treated as plastic strains. Two
independent yield surfaces govern the mechanical behaviour under compression and tension
and are described mathematically with the geotechnical stress invariants p, J and θ. These
yield surfaces expand and contract during loading, which is controlled by non-linear plas-
tic strain hardening and softening in the post peak regime, following normalised uniaxial
stress-strain curves. This normalisation is performed in terms of the peak plastic strains and
strengths since the material properties of the shotcrete are gradually changing with time. In
order to avoid the strong mesh dependency typical for strain softening materials, the softening
parameters are related to the element size by applying a fracture energy concept. With the
help of a characteristic length of the finite element and the tensile and compressive fracture
energy it is ensured, that the same amount of energy is released within the element upon
complete crack opening or crushing.

As mentioned earlier, the main material parameters governing stiffness and strength
are time-dependent and vary according to well established equations taken from widely ac-
cepted concrete standards. Making the compressive and tensile strain limits of the hard-
ening/softening curves time-dependent accounts for the fact, that shotcrete at early ages
appears to have the ability to withstand large strains without being completely damaged.
Furthermore, the model takes into account that young shotcrete at early ages is a very creep
active material. A uniaxial creep law, based on a Newton dashpot and including a time-
dependent function for the shotcrete viscosity, was extended for the three-dimensional stress
and strain space by assuming a creep plastic potential. Shrinkage of sprayed concrete repre-
sents a certain risk for early age cracking and is considered within the developed constitutive

410
model. Its mathematical formulation is given according to a standard shrinkage law taken
from the American concrete code. Finally, the model formulation includes temperature in-
duced deformations that occur due to the increase in temperature during cement hydration.
In a series of numerical tests on a single element it was shown that the model is capable
of reproducing various types of experimental tests such as uniaxial stress-strain curves in
compression at various shotcrete ages, creep tests or shrinkage measurements.

9.4 Conclusions from tunnel analyses

A typical tunnel construction in London Clay (undrained conditions) following three different
excavation schemes has been simulated within this thesis. The focus was on the mechanical
and time-dependent behaviour of the shotcrete shell, which was modelled with the newly
developed constitutive model for sprayed concrete. The obtained results were then compared
with much simpler approaches for modelling the tunnel lining behaviour (linear elastic and
HME). The following conclusions can be drawn:

 The overall system behaviour during tunnel construction and surface loading strongly
depends on the initial stress conditions in the ground before excavation. Therefore, the
adoption of a correct Ko -profile is of crucial importance as it controls the deformation
of the tunnel lining and the predicted surface settlements. With a Ko value greater
than 1, typical for overconsolidated clays, the tunnel lining tends to oval to the vertical
during excavation, but the bending direction changes during the surcharge loading. As
a consequence, a relatively large amount of elastic deformation is introduced in the
shotcrete shell since the stress path is reversed and travels through the elastic area of
the adopted yield surfaces.

 The factor “time” plays a key role in tunnel construction and should be considered
realistically in a finite element analysis. The loading rate of the shotcrete is controlled
by the elapsed time during excavation and surface loading, leading to a complex stress-
strain history within the tunnel lining. From a numerical point of view it is important to
investigate how plastic strain hardening and softening of the material occurs in combi-
nation with the development of the shotcrete parameters with time. It can be concluded
that the proper choice of a hardening parameter for a time-dependent material such as
shotcrete is of crucial importance.

 Surface settlements and volume loss seem to be highly influenced by the constitutive
model adopted for simulating the tunnel lining. The stiffness of the shotcrete shell is
one of the main parameters that controls the surface settlement behaviour. Sprayed
concrete at early ages represents a very soft material and hence the predicted surface
settlements can increase resulting in a volume loss that is above a certain target value.
However, with increasing shotcrete age the developed stiffness is sufficient to reduce

411
 surface settlements to an acceptable limit. A rapid ring-closure in the case of staged
tunnel construction should be aimed for as highlighted in the literature.

 Regarding the stress distribution within the lining it can be concluded that young
shotcrete does not develop any appreciable bending stiffness and the overall mechani-
cal behaviour is governed mainly by hoop compression. This fact is particularly pro-
nounced for the full face excavation, where the lining acts as a perfect ring without
any imperfections or geometrical irregularities. Furthermore, the absolute values of
these compressive stress states are at a relatively low level, since the strength of the
shotcrete has not fully developed at early stages of cement hydration. However, with
increasing curing time (or elapsed time during construction), the material behaviour
becomes more and more brittle and significant bending will be introduced. This can
lead to cracking (tensile softening) of the shotcrete shell when heavily loaded.

 It was observed that plasticity of the shotcrete plays an important role when modelling
young shotcrete. Already during the first increments of excavation most of the Gauss
points across a lining section touch the current yield surface and plastic deformations
are introduced. Plasticity becomes much less pronounced for older shotcrete, occuring
mainly under heavy surface loading conditions.

 The formulation of normalised hardening parameters allows the stress level (or utiliza-
tion factor) of a particular Gauss point to be established relatively easily. When tunnel
construction is performed fast, relatively high compressive stress levels have been ob-
served in all the analyses, being close to peak or even slightly beyond. However, with
increasing time during excavation and loading these stress levels reduce significantly to
a fairly low level. Some of the observed mechanisms of behaviour of the tunnel lining
were such, that a switch from a compressive to a tensile stress state took place, which
causes a risk of cracking.

 Non-linearity of the stress-strain behaviour of the young shotcrete represents an impor-

tant factor when simulating tunnel construction. Particularly the non-linear hardening
in compression and the tensile softening are of great importance. However, when mod-
elling the tunnel lining with the material properties of hardened shotcrete at 28 days,
almost no difference has been observed during excavation when comparing results with
the linear elastic approach. The difference is more pronounced after placing the surface
loading when tensile cracking was introduced in some of the tunnel cross sections due
to the heaviy loading conditions.

 When performing a staged tunnel construction the connections between the tunnel lin-
ing and some temporary structures such as invert or sidewall are critical areas from the
numerical point of view. Sharp corners and irregular shapes introduce high stresses and
can be seen as weak points of the structural system. Compressive crushing and tensile
cracking has to be expected with utilization factors far beyond peak. The associated

412
 problems with such weak zones could be avoided or at least improved by a smooth
transition or corner rounding.

 Modelling the tunnel lining as a linear elastic material with the stiffness of hardened
shotcrete at 28 days overpredicts stresses that develop during loading and unrealistic
high bending moments are introduced. Furthermore, the obtained lining deformations
are very small and do not represent realistic values. Therefore, it is highly recommended
not to adopt a linear elastic constitutive behaviour for shotcrete when the focus of the
analysis is mainly on the tunnel lining behaviour. Assuming a reduced lining stiffness as
in the HME approach is a huge improvement. It was shown, that with an appropriate
reduction in the Young’s modulus results tend to give a relatively good approximation
of the overall system behaviour compared to more sophisticated models. However,
it is this particular estimation of the stiffness reduction that has to be treated with
caution, since it is purely based on experience. When analysing complex structures and
geometries the HME approach will still overpredict stresses in the lining.

 The analyses carried out indicate, that for shotcrete at early ages, large differences can
be expected regarding the lining displacements, where the stiffness of the material is
a governing factor. In some cases these differences were pronounced such, that during
excavation of the tunnel a completely different system behaviour was observed, when
modelling a very young shotcrete response.

 The development of shotcrete stiffness and strength with time has been investigated
in this thesis by performing a parametric study. It was shown, that variations in the
development of these material properties highly influence the obtained results. A slower
increase in stiffness and strength with time results in larger surface settlements, lower
lining stresses but higher stress levels and larger lining displacements particularly during
excavation of the tunnel.

 The shotcrete deformability at early ages has been analysed by adopting different de-
velopments of the plastic strain limits with time both in compression and tension.
Results indicated that shotcrete deformability at early ages has a minor impact on the
settlement behaviour. In contrast, some major differences in the stress distributions
and stress levels have been obtained, with the main conclusion that a reduced strain-
ing capacity can increase the risk of cracking significantly due to the brittle material
response.

 The consideration of creep and relaxation of the sprayed concrete has the ability to
reduce stresses and bending moments in the tunnel lining. However, this reduction was
much less than expected and less than that usually emphasised in the literature. One
reason might be the choice of the adopted creep parameters. It is therefore recom-
mended to investigate the influence of creep on the tunnel lining behaviour further by
applying different sets of creep parameters and varying the elapsed time during tunnel

413
 construction. From the performed analyses it was not possible to detect new mechanical
mechanisms that might be caused by creep.

 Small changes in the predicted surface settlement troughs have been detected when
considering the hydration temperature of the cement paste. An expansion of the lining
during the first 10 to 12 hours after installation causes a relatively small reduction of
the volume loss. For some cross sections of the lining a slightly increased bending has
been observed when taking into account the temperature induced deformation. The
reason for this is supposed to be the brittle material behaviour as a consequence of the
cooling phase after reaching the peak temperature rise.

 Some minor differences have been observed in the results when accounting for shrinkage
of the shotcrete, causing cracking in some particular cases. However, the elapsed time
for these tunnel runs might not have been sufficient to reveal the complete impact of
the decrease in volume on the behaviour of the tunnel lining. Shrinkage of shotcrete can
be considered as a long-term problem and further investigations should be performed
on this topic.

 Due to the high volume loss occuring during tunnel construction with very young
shotcrete unrealistic high stress rearrangements in the soil and relatively high suctions
in the pore water pressures were introduced.

 Finally, it can be concluded that the consideration of the material response of young
shotcrete during the first 2 to 3 days after installation is a key element when simulat-
ing tunnel construction. During this time, the overall system behaviour is exclusively
governed by the material properties of the young shotcrete. The combination of exter-
nal loading and the transition from a ductile to a brittle type of material controls the
development of stresses and displacements in the tunnel lining. Shotcrete behaviour in
tunnelling can be regarded as a highly non-linear and complex problem being influenced
by many factors.

9.5 Recommendations for future research

Shotcrete technology and its application in tunnel construction is a very innovative and
challenging field of research. Enormous improvements have been made over the last few
years with respect to the constituent materials, quality control, installation techniques and
equipment. Nowadays, safe construction of shotcrete tunnel linings can be achieved with well
trained workmen who are aware of the involved difficulties and their possible consequences.
However, in a tunnelling project some uncertainties regarding varying ground conditions
or unexpected loading scenarios of the lining will always exist, which can lead to critical
system behaviour and even failure of the structure. It is the purpose of an appropriate design
procedure to account for such situations and guarantee safety of the tunnel at all construction

414
stages. As highlighted throughout this thesis, advanced constitutive modelling of shotcrete
was usually missing in the past and the design of a shotcrete shell was mainly based on the
experience of the tunnel engineers. Furthermore, from the extensive literature review it can
be observed that there is a large gap between experimental testing and numerical modelling
of the material behaviour of sprayed concrete. Therefore, from the authors point of view, the
following aspects and ideas should be taken into account for future research on the topic of
shotcrete in tunnelling:
 For design purposes, experimental testing of shotcrete should be carried out in accor-
dance with a particular constitutive model following a predefined testing scheme in
such a way, that the necessary model parameters can be determined reliably from the
obtained test results. For a correct modelling it is of crucial importance to calibrate a
constitutive model using a complete set of tests for one specific type of shotcrete.

 In the past, testing of shotcrete has usually been restricted to uniaxial compression tests.
This stress range should be extended in order to investigate the material behaviour
under various loading conditions, such as biaxial or triaxial stress paths. Furthermore,
the focus of these tests should be on the early age stages of cement hydration in order
to capture the complete development of material properties up to an age of 28 days.

 Very little information is available in the literature about the tensile capacity of sprayed
concrete due to various testing difficulties. However, it must be noted that these tensile
material properties can be of great importance for a tunnel lining design, since cracking
of the shotcrete should be avoided.

 The large scatter in almost any available shotcrete test data due to the large number of
influencing factors represents a huge problem for a realistic calibration of appropriate
model parameters. Future research should therefore shift its focus on new innovative
testing methods that are already in use in other scientific fields (i.e. fibre optics, ultra-
sound, laser technology, etc.), since material science in general is a very interdisciplinary
field. Collaboration with different areas of engineering or science could have a fruitful
outcome for further experimental work on early age shotcrete.

 The current design philosophy for tunnel construction has to open up to new ideas and
scientific trends. With respect to the design of a tunnel lining, it is not enough to
guarantee safety of a structure by applying an approach that is almost entirely based
on experience, since experience can lead to fatal errors. Simple design methods based
on linear elasticity are easy to use and have their warranty indeed, but engineers should
know their limits and should be aware of when there is the need for a more sophisticated
constitutive model. The realistic modelling of shotcrete behaviour gives room for more
developments and progress and will enable engineers to validate the simpler approaches.

 No claim is made for the entire completeness of the new constitutive model for shotcrete
presented within this thesis. Several simplifications and problems remain unsolved,

415
 such as the question of the correct hardening parameter for the realistic description of
material behaviour for various loading conditions with respect to time. The applied nor-
malisation of the principal plastic strains ε1 and ε3 is a first step in the right direction,
but further progress has to be made on this topic.

 The generalisation of creep deformation into 3D with the help of a creep plastic potential
is a powerful way of modelling creep. However, the obtained creep deformations have
to be validated against experimental data in order to consider possible changes of the
involved potential functions.

 One of the drawbacks of the current constitutive model for shotcrete is the use of
solid elements for the simulation of the tunnel lining, which requires an extreme mesh
refinement in order to obtain reasonable results for stress distributions across the lining
thickness. Obviously, certain computational limitations exist and therefore, it might
be better to focus in the future more on the development of advanced beam elements
including time-dependent material behaviour.

 Due to time limitations all the analyses within this thesis were performed in 2D. How-
ever, it is recommended for the future to perform full 3D analyses in order to capture
the complete process of stress rearrangements and deformations both in the soil and
the tunnel lining close to the tunnel face in a realistic way.

 Finally, when analysing tunnel construction with the finite element method it is im-
portant to investigate extreme cases that might happen in reality. Case histories have
proven that failures of tunnel linings occur mostly due to unexpected ground condi-
tions, abnormalities in material quality, construction defects and irregular geometries.
Analysing simple circular tunnels without imperfections in 2D will not detect criti-
cal stages during tunnel advance. The finite element method is a powerful tool for
investigating different types of tunnel failures and this potential should be utilised.

416
Appendix
Volume loss and maximum surface settlement for excavation Type 1 and Ko = 0.5

Run Volume loss VL Maximum settlement δ

(%) (mm)
Linear elastic (LE) 1.48 21.3
HME 1.57 23.0
Non-linear (NL) 1.48 21.3
TPM 1 5.63 81.9
TPM 2 3.65 54.5
TPM 3 2.26 33.3
TPM 4 1.87 27.4
TPM 5 1.70 25.0
TPM 6 1.56 23.0
TPM 7 1.51 22.2
TPM 8 1.50 22.0
TPM 9 1.50 21.8
TPM 10 1.49 21.7
TPM 11 1.49 21.5
TPM 12 1.48 21.5
TPM 13 1.48 21.3
TPM 14 1.48 21.3
STRENGTH 1 1.49 21.7
STRENGTH 2 1.86 27.4
STRENGTH 3 2.33 34.7
STRENGTH 4 2.95 44.1
STRENGTH 5 1.72 25.3
STRENGTH 6 2.07 30.3
STRAIN 1 1.58 23.2
STRAIN 2 1.54 22.7
STRAIN 3 1.54 22.6
STRAIN 4 1.52 22.3
CREEP (C) 1.56 23.0
SHRINKAGE (SH) 1.57 23.1
TEMPERATURE (T) 1.53 22.6
C + SH + T 1.55 22.9

417
Volume loss and maximum surface settlement for excavation Type 1 and Ko = 1.5

Run Volume loss VL Maximum settlement δ

(%) (mm)
Linear elastic (LE) 1.52 14.1
HME 1.61 14.9
Non-linear (NL) 1.52 14.1
TPM 1 3.84 38.5
TPM 2 2.87 29.5
TPM 3 2.07 20.4
TPM 4 1.79 17.1
TPM 5 1.67 15.4
TPM 6 1.58 14.4
TPM 7 1.55 14.2
TPM 8 1.54 14.2
TPM 9 1.54 14.1
TPM 10 1.53 14.1
TPM 11 1.53 14.1
TPM 12 1.53 14.1
TPM 13 1.53 14.1
TPM 14 1.53 14.1
STRENGTH 1 1.54 14.1
STRENGTH 2 1.78 16.8
STRENGTH 3 2.08 20.4
STRENGTH 4 2.46 24.8
STRENGTH 5 1.69 15.7
STRENGTH 6 1.95 19.0
STRAIN 1 1.58 14.5
STRAIN 2 1.57 14.4
STRAIN 3 1.57 14.4
STRAIN 4 1.56 14.3
CREEP (C) 1.58 14.5
SHRINKAGE (SH) 1.59 14.6
TEMPERATURE (T) 1.55 14.0
C + SH + T 1.57 14.3

418
References

Abdel-Jawad, Y. & Haddad, R. (1992), ‘Effect of early overloading of concrete on strength

at later ages’, Cement and Concrete Research 22, pp. 927–936.

ACI 209R (1992), Prediction of creep, shrinkage and temperature effects in concrete struc-
tures, ACI Committee 209.

Aldrian, W. (1991), Beitrag zum Materialverhalten von früh belastetem Spritzbeton, PhD
thesis, University for Mining and Metallurgy, Leoben, Austria.

Alkhiami, H. (1995), Ein Näherungsverfahren zur Abschätzung einer Spritzbetonkalotten-

schale auf der Grundlage von in-situ Messungen, PhD thesis, Technical University Han-
nover, Germany.

Armelin, H. & Banthia, N. (1998), ‘Mechanics of aggregate rebound in shotcrete - Part I’,
Materials and Structures 31, pp. 91–98.

Arsland, H. & Sture, S. (2008), ‘Finite element simulation of localization in granular materials
by micropolar continuum approach’, Computers and Geotechnics 35, pp. 548–562.

ASCCT - Austrian Society for Concrete- and Construction Technology (2004), Guideline
sprayed concrete, Csöngei GmbH.

Atrushi, D. (2003), Tensile and compressive creep of early-age concrete: testing and mod-
elling, PhD thesis, Norwegian University of Science and Technology, Trondheim, Norway.

Attard, M. & Setunge, S. (1996), ‘Stress-strain relationship of confined and unconfined con-
crete’, ACI Materials Journal 93(5), pp. 432–442.

Austin, S. & Robins, P. (1995), Sprayed concrete - properties, design and application, Whittles
Publishing.

Aydan, O., Sezaki, M. & Kawamoto, T. (1992), Mechanical and numerical modelling of
shotcrete, in Pande & Pietruszczak, eds, ‘Numerical Models in Geomechanics’, pp. 757–
764.

Babu, R., Benipal, G. & Singh, A. (2005), ‘Constitutive modelling of concrete: an overview’,
Asian Journal of Civil Engineering (Building and Housing) 6(4), pp. 211–247.

419
Balmer, G. (1949), Shearing strength of concrete under high triaxial stress - computation
of Mohr’s envelope as a curve, Technical Report SP-23, Structural Research Laboratory,
Branch of Design and Construction, Denver.

Banthia, N., Trottier, J., Wood, D. & Beaupré, D. (1992), ‘Influence of fibre geometry on
steel fibre reinforced dry-mix shotcrete’, Concrete International 14(5), pp. 24–28.

Barros, J. & Figuieiras, J. (1999), ‘Flexural behaviour of SFRC: testing and modelling’,
Journal of Materials in Civil Engineering 11, pp. 331–339.

Barton, N., Lien, R. & Lunde, J. (1974), ‘Engineering classification of rock masses for the
design of tunnel support’, Rock Mechanics 6(4), pp. 189–239.

Bažant, Z. & Baweja, S. (1995), Creep and shrinkage prediction model for analysis and design
of concrete structures: Model B3, a report submitted to ACI Committee 209.

Bićanić, N. & Pearce, C. (1996), ‘Computational aspects of a softening plasticity model for
plain concrete’, Mechanics of Cohesive-Frictional Materials 1, pp. 75–94.

Bićanić, N., Pearce, C. & Owen, D. (1994), Failure predictions of concrete like materials using
softening Hoffman plasticity model, in ‘Euro-C 1994 International Conference, Computer
Modelling of Concrete Structures’, pp. 185–198.

Bieniawski, Z. (1994), Rock mechanics design in mining and tunnelling, Balkema, Rotterdam.

Bischoff, P. & Paixao, R. (2004), ‘Tension stiffening and cracking of concrete reinforced with
glass fibre reinforced polymer (GFRP) bars’, Canadian Journal of Civil Engineering 31, pp.
579–588.

Bockhold, J. (2005), Modellbildung und numerische Analyse nichtlinearer Kriechprozesse

in Stahlbetonkonstruktionen unter Schädigungsaspekten, PhD thesis, Ruhr-University
Bochum, Germany.

Bosnjak, D. (2000), Self-induced cracking problems in hardening concrete structures, PhD

thesis, Norwegian University of Science and Technology, Trondheim.

Brameshuber, W. & Hilsdorf, H. (1989), Development of strength and deformability of very

young concrete, in Shah & Swartz, eds, ‘Fracture of Concrete and Rock’, pp. 409–421.

Byfors, J. (1980), Plain concrete at early ages, Technical report, Swedish Cement and Con-
crete Research Institute.

Camps, G., Turatsinze, A., Sellier, A., Escadeillas, G. & Bourbon, X. (2008), ‘Steel-fibre
reinforcement and hydration coupled effects on concrete tensile behaviour’, Engineering
Fracture Mechanics 75, pp. 5207–5216.

420
Carlson, R. (1937), ‘A simple method for the computation of temperatures in concrete struc-
tures’, ACI Journal 34, pp. 89–102.

CEB-FIP Model Code (1990), Design Code - Comite Euro-International du Beton, Thomas
Telford, London.

Cendón, D., Gálvez, J., Elices, M. & Planas, J. (2000), ‘Modelling the fracture of concrete
under mixed loading’, International Journal of Fracture 103, pp. 293–310.

Chang, Y. (1994), Tunnel support with shotcrete in weak rock - A rock mechanics study,
PhD thesis, Royal Institute of Technology, Stockholm, Sweden.

Chen, A. & Chen, W. (1975), ‘Constitutive relations for concrete’, Journal of the Engineering
Mechanics Division, ASCE 101(4), pp. 465–481.

Chen, W. (1982), Plasticity in reinforced concrete, McGraw-Hill Book Company.

Clayton, C., Hope, V., Heymann, G., van der Berg, J. & Bica, A. (2000), ‘Instrumentation
for monitoring sprayed concrete lined soft ground tunnels’, Proceedings of the Institution
of Civil Engineers, Geotechnical Engineering 143(July), pp. 119–130.

Clayton, C., van der Berg, J., Heymann, G., Bica, A. & Hope, V. (2002), ‘The performance
of pressure cells for sprayed concrete linings’, Géotechnique 52(2), pp. 107–115.

Clayton, C., van der Berg, J. & Thomas, A. (2006), ‘Monitoring and displacements at
Heathrow Express Terminal 4 station tunnels’, Géotechnique 56(5), pp. 323–334.

Cornejo-Malm, G. (1995), Spritzbeton und seine Eigenschaften - Schwinden von Spritzbeton,

Technical report, Institut für Bauplanung und Baubetrieb, ETH Zürich, Switzerland.

Davis, R., Davis, H. & Hamilton, J. (1934), ‘Plastic flow of concrete under sustained stress’,
ASTM Proceedings 34(2), pp. 354–386.

De Borst, R. (1987), ‘Computation of post-bifurcation and post-failure behaviour of strain-

softening solids’, Computers & Structures 25(2), pp. 211–224.

De Borst, R. (2002), ‘Fracture in quasi-brittle materials: a review of continuum damage-based

approaches’, Engineering Fracture Mechanics 69, pp. 95–112.

De Borst, R., Remmers, J., Needleman, A. & Abellan, M. (2004), ‘Discrete vs. smeared crack
models for concrete failure: bridging the gap’, International Journal for Numerical and
Analytical Methods in Geomechanics 28, pp. 583–607.

De Borst, R. & van den Boogaard, A. (1994), ‘Finite-element modelling of deformation and
cracking in early-age concrete’, Journal of Engineering Mechanics 120(12), pp. 2519–2534.

421
De Schutter, G. (1999), ‘Extension towards early age concrete of CEB-FIP Model Code 1990
stress-strain relation for short-term compressive loading’, ACI Materials Journal 96(1), pp.
95–100.

De Schutter, G. & Taerwe, L. (1997), ‘Fracture energy of concrete at early ages’, Materials
and Structures 30, pp. 67–71.

Desayi, P. & Krishnan, S. (1964), ‘Equation for the stress-strain curve of concrete’, ACI
Journal 61(22), pp. 345–350.

Ding, Y. (1998), Technologische Eigenschaften von jungem Stahlfaserbeton und Stahlfaser-

spritzbeton, PhD thesis, University of Innsbruck, Austria.

Ding, Y. & Kusterle, W. (1999), ‘Comparative study of steel-fibre reinforced concrete and
steel mesh-reinforced concrete at early ages in panel tests’, Cement and Concrete Research
29, pp. 1827–1834.

Ding, Y. & Kusterle, W. (2000), ‘Compressive stress-strain relationship of steel fibre-

reinforced concrete at early ages’, Cement and Concrete Research 30, pp. 1573–1579.

Duddeck, H. & Erdmann, J. (1982), Structural design models for tunnels, in ‘Tunnelling ’82,
Institute of Mining and Metallurgy, London’, pp. 83–91.

Duddeck, H. & Erdmann, J. (1985), ‘On structural design models for tunnels in soft soil’,
Underground Space 9, pp. 246–259.

Duvaut, G. & Lions, J. (1972), Les inequations en mechanique et en physique, Dunod, Paris.

Dvorkin, E., Cuitiño, A. & Gioia, G. (1989), ‘A concrete material model based on non-
associated plasticity and fracture’, Engineering Computations 6, pp. 281–294.

Ebead, U. & Marzouk, H. (2005), ‘Tension-stiffening model for FRP-strengthened RC con-

crete two-way slabs’, Materials and Structures 38, pp. 193–200.

Eddie, C. & Neumann, C. (2003), ‘LaserShell leads the way for SCL tunnels’, Tunnels &
Tunnelling International (June), pp. 38–42.

Eddie, C., Neumann, C. & Jäger, J. (2009), Innovative permanent shotcrete tunnel linings
in London Clay, in Kusterle, ed., ‘Internationale Spritzbetontagung, Alpbach, Austria’,
pp. 1–40.

EFNARC (1996), European specification for sprayed concrete.

Eierle, B. & Schikora, K. (1999), ‘Computational modelling of concrete at early ages’, Diana
World 2.

Einstein, H. & Schwartz, W. (1979), ‘Simplified analysis for tunnel supports’, Journal of
Geotechnical Engineers, ASCE .

422
EN 1992-1-1 (2004), Eurocode 2: Design of concrete structures - Part 1-1: General rules and
rules for buildings, European Committee for Standardization.

Engelbreth, K. (1961), ‘Beregning av tunnel eller rør med sirkulaert tverrsnitt gjennom ho-
mogen jordmasse’, Teknisk Ukeblad 108, pp. 625–627.

England, G. & Illston, J. (1965), ‘Methods of computing stress in concrete from a history
of measured strain’, Civil Engineering and Public Works Review 60(1,2,3), pp. 513–517,
692–694,846–847.

Feenstra, P. (1993), Computational aspects of biaxial stress in plain and reinforced concrete,
PhD thesis, Delft University, Netherlands.

Fields, K. & Bischoff, P. (2004), ‘Tension stiffening and cracking of high-strength reinforced
concrete tension members’, ACI Structural Journal 101(4), pp. 447–456.

Fischnaller, G. (1992), Untersuchungen zum Verformungsverhalten von jungem Spritzbeton

im Tunnelbau, Grundlagen und Versuche, Master’s thesis, University of Innsbruck, Austria.

Flügge, W. (1967), Viscoelasticity, Blaisdell Publishing Company.

Frantziskonis, G., Desai, C. & Somasundaram, S. (1986), ‘Constitutive model for non-
associative behaviour’, Journal of the Engineering Mechanics Division, ASCE 112, pp.
932–946.

Franzen, T. (1992), ‘Shotcrete for underground support: a state-of-the-art report with focus
on steel-fibre reinforcement’, Tunnelling and Underground Space Technology 7(4), pp. 383–
391.

Freiesleben Hansen, P. & Pederson, E. (1977), ‘Måleinstrument til kontrol af betons hærd-
ning’, Nordisk Betong 21, pp. 21–25.

Galler, R. (1997), Shotcrete - realistic modelling for the purpose of economical tunnel design,
in Yuan, ed., ‘Computer Methods and Advances in Geomechanics’, pp. 1383–1387.

Garboczi, E. & Bentz, D. (1992), Computer-based models of the micro structure and proper-
ties of cement-based materials, in ‘Proceedings of the 9th International Conference on the
Chemistry of Cement’, Vol. VI, pp. 3–15.

Gardner, N. & Lockman, M. (2001), ‘Design provisions for drying shrinkage and creep of
normal-strength concrete’, ACI Materials Journal 98(2), pp. 159–167.

Gasparre, A. (2005), Advanced laboratory characterisation of London Clay, PhD thesis, Im-
perial College, London, UK.

Gilbert, R. & Warner, R. (1978), ‘Tension stiffening in reinforced concrete slabs’, Journal of
the Structural Division, ASCE 104, pp. 1885–1899.

423
Girmscheid, G. (2000), Baubetrieb und Bauverfahren im Tunnelbau, Ernst & Sohn.

Golser, J. (1999), Behaviour of early-age shotcrete, in T. Celestino & H. Parker, eds,

‘Shotcrete for Underground Support VIII, Sao Paulo, Brazil’, pp. 83–92.

Golser, J., Rabensteiner, K., Sigl, O. & Aldrian, W. (1990), Materialgesetz für Spritzbeton,
Straßenforschung 696, Technical report, Bundesministerium für Wirtschaftliche Angelegen-
heiten, Austria.

Golser, J., Schubert, P. & Rabensteiner, K. (1989), A new concept for evaluation of loading
in shotcrete linings, in ‘International Congress on Progress and Innovation in Tunnelling’,
pp. 79–85.

Gopalakrishnan, K., Neville, A. & Ghali, A. (1969), ‘Creep Poisson’s ratio of concrete under
multiaxial compression’, ACI Journal 66, pp. 1008–1020.

Grassl, P., Lundgren, K. & Gyltoft, K. (2002), ‘Concrete in compression: A plasticity theory
with novel hardening law’, International Journal of Solids and Structures 39(20), pp. 5205–
5223.

Graziani, A., Boldini, D. & Ribacchi, R. (2005), ‘Practical estimate of deformations and
stress relief factors for deep tunnels supported by shotcrete’, Rock Mechanics and Rock
Engineering 38(5), pp. 345–372.

Gullan, G. (1975), ‘Shotcrete for tunnel linings’, Tunnels and Tunnelling 7(9), pp. 37–47.

Gutsch, A. & Rostásy, F. (1994), Young concrete under high tensile stresses - creep, relax-
ation and cracking, in Springenschmid, ed., ‘Thermal Cracking in Concrete at Early Ages’,
pp. 111–118.

Hafez, N. (1995), Post-failure modelling of three-dimensional shotcrete lining for tunnelling,

PhD thesis, University of Innsbruck, Austria.

Hagihara, S., Masuda, Y. & Nakamura, S. (2002), Creep behaviour of high-strength con-
crete at early age, in ‘6th International Symposium on High Strength/High Performance
Concrete’.

Han, D. & Chen, W. (1986), ‘Strain space plasticity formulation for hardening-softening ma-
terials with elastoplastic coupling’, International Journal of Solids and Structures 22, pp.
935–950.

Haufe, A. (2001), Dreidimensionale Simulation bewehrter Flächentragwerke aus Beton mit

der Plastizitätstheorie, PhD thesis, Universit Stuttgart, Germany.

Hauggaard, A., Damkilde, L., Freisleben Hansen, P., Pederson, E. & Nielsen, A. (1997), Hetek
- control of early age cracking in concrete - phase 4 and 5: Material modelling, continuum
approach, Technical report, Ministry of Transport, Denmark.

424
Hellmich, C., Sercombe, J., Ulm, F. & H.A., M. (2000), ‘Modelling of early-age creep of
shotcrete. II: Application to tunnelling’, Journal of Engineering Mechanics 126(3), pp.
292–299.

Hellmich, C., Ulm, F. & Mang, H. (1999a), ‘Multisurface chemoplasticity. I: Material model
for shotcrete’, Journal of Engineering Mechanics 125(6), pp. 692–701.

Hellmich, C., Ulm, F. & Mang, H. (1999b), ‘Multisurface chemoplasticity. II: Numerical
studies on NATM tunneling’, Journal of Engineering Mechanics 125(6), pp. 702–713.

Hesser, J. (2000), Zum Einfluß unterschiedlicher Spritzbetonqualitäten auf das Tragverhalten

tiefliegender Gesteinsstrecken - Laborative Untersuchungen und numerische Analysen, PhD
thesis, Technical University Clausthal, Germany.

Hilar, M., Thomas, A. & Falkner, L. (2005), ‘The latest innovation in sprayed concrete lining
- the LaserShell method’, Tunel (Magazine of the Czech Tunnelling Committee and the
Slovak Tunnelling Committee ITA/AITES) 4, pp. 11–19.

Hillerborg, A. (1985), Determination and significance of the fracture toughness of steel fibre
concrete, in Shah & Skarendahl, eds, ‘Steel Fibre Concrete, US-Sweden joint seminar’.

Hillerborg, A., Modéer, M. & Petersson, P. (1976), ‘Analysis of crack formation and crack
growth in concrete by means of fracture mechanics and finite elements’, Cement and Con-
crete Research 6, pp. 773–782.

Hills, D. (1982), ‘Site-produced sprayed concrete’, Concrete 16(12), pp. 44–50.

Hoek, E. (2007), ‘Integration of geotechnical and structural design in weak rock tunnels’,
RocNews (Spring), pp. 1–15.

Hofstetter, G., Simo, J. & Taylor, R. (1993), ‘A modified cap model: closest point solution
algorithms’, Computers & Structures 46, pp. 203–214.

Holt, E. (2001), Early age autogeneous shrinkage of concrete, Technical Report 446, Technical
Research Centre of Finland.

Houst, Y. (1997), Carbonation shrinkage of hydrated cement paste, in ‘4th CANMET/ACI

International Conference on Durability of Concrete, Ottawa, Canada’, pp. 481–491.

HSE (2000), The collapse of NATM tunnels at Heathrow Airport, Technical report, Health
and Safety Executive.

HSE - Health & Saftey Executive (1996), Safety of the New Austrian Tunnelling Method
(NATM) Tunnels, HSE Books, Sudbury.

Huber, F. (2006), Nichtlineare dreidimensionale Modellierung von Beton- und Stahlbeton-

tragwerken, PhD thesis, Universit Stuttgart, Germany.

425
Huber, H. (1991), Untersuchungen zum Verformungsverhalten von jungem Spritzbeton im
Tunnelbau, Master’s thesis, University of Innsbruck, Austria.

Huo, X. & Wong, L. (2006), ‘Experimental study of early-age behaviour of high performance
concrete deck slabs under different curing methods’, Construction and Building Materials
20, pp. 1049–1056.

ÖIAV (1980), Neue Österreichische Tunnelbaumethode, Definition und Grundsätze, Technical

Report Heft 74, Österreichischer Ingenieur- und Architektenverein.

ICE - Institution of Civil Engineers (1996), Sprayed concrete linings (NATM) for tunnels in
soft ground, Thomas Telford, London.

ICE - Institution of Civil Engineers (2004), Tunnel lining design guide, Thomas Telford,
London.

Illston, J. (1965), ‘The creep of concrete under uniaxial tension’, Magazine of Concrete Re-
search 17(51), pp. 77–84.

Jardine, R., Potts, D., Fourie, A. & Burland, J. (1986), ‘Studies of the influence of non-linear
stress-strain charactersitics in soil-structure interaction’, Géotechnique 36(3), pp. 377–396.

Jodl, H. & Kusterle, W. (1998), Experiences with austrian sprayed concrete technology in
tunnelling, in Bergdahl & Nordmark, eds, ‘Underground Construction in Modern Infras-
tructure’, pp. 367–373.

Jones, B. (2007), Design of SCL tunnels in soft ground using Eurocodes, in Eberhard-
steiner, ed., ‘ECCOMAS Thematic Conference on Computational Methods in Tunnelling
(EURO:TUN 2007)’, pp. 1–12.

Jones, B., Thomas, A., Hsu, Y. & Hilar, M. (2008), ‘Evaluation of innovative sprayed-
concrete-lined tunnelling’, Proceedings of the Institution of Civil Engineers, Geotechnical
Engineering 161, pp. 137–149.

Jones, B., Thomas, A. & Stärk, A. (2005), The importance of stress measurement in a holistic
sprayed concrete tunnel design process, in Chambery, ed., ‘Int. Congress on Tunnelling for
a Sustainable Europe, Lyon’.

Karakus, M. (2007), ‘Appraising the methods accounting for 3D tunnelling effects in 2D plane
strain FE analysis’, Tunnelling and Underground Space Technology 22(1), pp. 47–56.

Karakus, M. & Fowell, R. (2003), ‘Effects of different tunnel face advance excavation on the
settlement by FEM’, Tunnelling and Underground Space Technology 18, pp. 513–523.

Karakus, M. & Fowell, R. (2004), An insight into the New Austrian Tunnelling Method
(NATM), in ‘ROCKMEC - VIIth Regional Rock Mechanics Symposium, Sivas, Turkey’.

426
Karrer, J. (1986), Spritzbeton im Tunnelbau: Technologie, Festigkeitsentwicklung, Tragver-
halten, Master’s thesis, University for Mining and Metallurgy, Leoben, Austria.

Kasai, Y., Yokoyama, K. & Matsui, I. (1971), Tensile properties of early-age concrete, in
‘Proc. of the International Conference on Mechanical Behaviour of Materials’, pp. 288–
299.

Kavvadas, M. (2003), Monitoring and modelling ground deformations during tunnelling, in

‘11th Int. Symposium on Deformation Measurements, Santorini, Greece’, pp. 371–390.

Kazemi, M., Fazileh, F. & Ebrahiminezhad (2007), ‘Cohesive crack model and fracture energy
of steel-fibre-reinforced-concrete notched cylindrical specimens’, Journal of Materials in
Civil Engineering 19(10), pp. 884–890.

Kienberger, G. (1999), Einschaliger Tunnelbau - Einfluß des zeitabhängigen Materialverhal-

tens auf die Ausbaubeanspruchung, PhD thesis, University for Mining and Metallurgy,
Leoben, Austria.

König, G. & Duda, H. (1991), Basic concept for using concrete tensile strength, in ‘IABSE
Colloquium on Structural Concrete’, pp. 605–621.

Kobler, H. (1966), Dry-mix coarse-aggregate shotcrete as underground support, in ACI, ed.,

‘Shotcreting (SP-14)’, pp. 33–58.

Kolymbas, D. (1998), Geotechnik - Tunnelbau und Tunnelmechanik, Springer.

Kooiman (2000), Modelling steel fibre reinforced concrete for structural design, PhD thesis,
Delft University of Technology, Netherlands.

Kotsovos, M. & Newman, J. (1978), ‘Generalized stress-strain relations for concrete’, Journal
of the Engineering Mechanics Division, ASCE 104(4), pp. 845–857.

Kropik, C. & Mang, H. (1995), ‘Computational mechanics of the excavation of tunnels’,

Engineering Computations 13(7), pp. 49–69.

Kützing, L. (2000), Tragfähigkeitsermittlung stahlfaserverstärkter Betone, B.G. Teubner.

Kupfer, H. & Gerstle, K. (1973), ‘Behaviour of concrete under biaxial stresses’, Journal of
the Engineering Mechanics Division, ASCE 99(4), pp. 853–866.

Kusterle, W. (1985), Frühfestigkeiten des Spritzbetons, in Lukas, ed., ‘Internationale Fach-

tagung Spritzbetontechnologie’, pp. 35–38.

Kusterle, W. (1999), Comparison of shrinkage behaviour and creep properties under different
compressive stress levels for wet mix sprayed concrete from ten hours up to two weeks,
Technical report, Morgan Tunnelling.

427
Kusterle, W. & Lukas, W. (1990), Spritzbeton hoher Güte für die einschalige Spritzbeton-
bauweise, in ‘Tagungsband der 3. Int. Fachtagung Spritzbeton-Technologie, Innsbruck,
Austria’, pp. 29–40.

Kusterle, W. & Lukas, W. (1993), Single-shell shotcrete method, in Wood & Morgan, eds,
‘Shotcrete for Underground Support VI, Canada’, pp. 118–129.

Kuttner, C. (1989), Numerische Simulation einer Druckmeßdose, eingebaut in einen Spritz-

betonkörper, Master’s thesis, University for Mining and Metallurgy, Leoben, Austria.

Kuwajima, F. M. (1999), Early age properties of shotcrete, in T. Celestino & H. Parker, eds,
‘Shotcrete for Underground Support VIII, Sao Paulo, Brazil’, pp. 153–173.

Labahn, O. (1982), Ratgeber für Zementingenieure, Bauverlag.

Lackner, R., Hellmich, C. & Mang, H. (2002), ‘Constitutive modelling of cementitious ma-
terials in the framework of chemoplasticity’, International Journal for Numerical Methods
in Engineering 53, pp. 2357–2388.

Lackner, R., Macht, J. & Mang, H. (2006), ‘Hybrid analysis method for on-line quantification
of stress states in tunnel shells’, Computer Methods in Applied Mechanics and Engineering
195, pp. 5361–5376.

Lackner, R. & Mang, H. (2000), Material modelling and computational strategies in the
analysis of concrete shells, in ‘ECCOMAS European Congress on Computational Methods
in Applied Sciences and Engineering’, pp. 1–23.

Lackner, R. & Mang, H. (2003), ‘Cracking in shotcrete tunnel shells’, Engineering Fracture
Mechanics 70, pp. 1047–1068.

Lackner, R. & Mang, H. (2004), ‘Chemoplastic material model for the simulation of early-age
cracking: From the constitutive law to numerical analyses of massive concrete structures’,
Cement & Concrete Composites 26, pp. 551–562.

Lu, J., Lin, G., Wang, Z. & Xiao, S. (2004), ‘Reduction of compressive strength of concrete
due to triaxial compressive loading history’, Magazine of Concrete Research 56(3), pp.
139–149.

Malmberg, B. (1977), Steel fibre reinforced concrete under free and restrained shrinkage,
Technical report, Fiberbetong, Nordforsks projektkommitte för FRC-material, Sweden.

Malmgrem, L., Nordlund, E. & Rolund, S. (2005), ‘Adhesion strength and shrinkage of
shotcrete’, Tunnelling and Underground Space Technology 20, pp. 33–48.

Matsuo, S., Matsuoka, S., Masuda, A. & Yanagi, H. (1995), A study on approximation method
of tension softening curve of steel fibre reinforced concrete, in Wittmann, ed., ‘Second

428
 International Conference on Fracture Mechanics of Concrete Structures (FRAMCOS II)’,
pp. 745–754.

Mehta, P. & Monteiro, P. (1985), Concrete - Structure, Properties and Materials.

Meschke, G. (1996), ‘Consideration of aging of shotcrete in the context of a 3-D viscoplastic

material model’, International Journal for Numerical Methods in Engineering 39, pp. 3123–
3143.

Meschke, G., Kropik, C. & Mang, H. (1996), ‘Numerical analysis of tunnel linings by means of
a viscoplastic material model for shotcrete’, International Journal for Numerical Methods
in Engineering 39, pp. 3145–3162.

Meschke, G., Lackner, R. & H.A., M. (1998), ‘An isotropic elastoplastic-damage model for
plain concrete’, International Journal for Numerical Methods in Engineering 42, pp. 703–
727.

Müller, L. (1990), ‘Removing misconceptions an the New Austrian Tunnelling Method’, Tun-
nels & Tunnelling (Special Issue), pp. 15–18.

Müller, L. & Fecker, E. (1978), Grundgedanken und Grundsätze der Neuen Österreichischen
Tunnelbauweise. Grundlagen und Anwendung der Felsmechanik, in ‘Felsmechanik Kollo-
quium Karlsruhe’, pp. 247–262.

Möller, S., Vermeer, P. & Marcher, T. (2004), NATM-tunnelling in softening stiff clays and
weak rocks, in ‘Proceedings of the 9th Symposium on Numerical Models in Geomechanics
(NUMOG IX), Ottawa, Canada’, pp. 404–413.

Mosler, J. & Meschke, G. (2004), ‘Embedded crack vs. smeared crack models: a comparison of
elementwise discontinuous crack path approaches with emphasis on mesh bias’, Computer
Methods in Applied Mechanics and Engineering 193, pp. 3351–3375.

Moussa, A. M. (1993), Finite element modelling of shotcrete in tunnelling, PhD thesis, Uni-
versity of Innsbruck, Austria.

Muir Wood, A. (1975), ‘The circular tunnel in elastic ground’, Géotechnique 25(1), pp. 115–
127.

Nayal, R. & Rasheed, H. (2006), ‘Tension stiffening model for concrete beams reinforced with
steel and FRP bars’, Journal of Materials in Civil Engineering 18(6), pp. 831–841.

Nelissen, L. (1972), ‘Biaxial testing of normal concrete’, Heron 18(1).

Neubert, B. & Manns, W. (1993), Mechanical-technological properties of shotcrete with ac-

celerating admixtures, in ‘International Symposium on Sprayed Concrete, Norwegian Con-
crete Association’, pp. 258–270.

429
Neville, A. (1970), Creep of Concrete: Plain, Reinforced and Prestressed, North Holland
Publishing Company, Amsterdam.

Neville, A. (1981), Properties of concrete, Longman Scientific & Technical.

Oliver, J. (1989), ‘A consistent characteristic length for smeared cracking models’, Interna-
tional Journal for Numerical Methods in Engineering 28, pp. 461–474.

Oluokun, F., Burdette, E. & Deatherage, J. (1991a), ‘Elastic modulus, Poisson’s ratio and
compressive strength relationships at early ages’, ACI Materials Journal 88(1), pp. 3–10.

Oluokun, F., Burdette, E. & Deatherage, J. (1991b), ‘Splitting tensile strength and compres-
sive strength relationship at early ages’, ACI Materials Journal 88, pp. 115–121.

Onate, E., Oller, S., Oliver, J. & Lubliner, J. (1988), ‘A constitutive model for cracking of
concrete based on the incremental theory of plasticity’, Engineering Computations 5, pp.
309–319.

Oreste, P. (2003), ‘Analysis of structural interaction in tunnels using the convergence-

confinement approach’, Tunnelling and Underground Space Technology 18, pp. 347–363.

Outterside, J. (2003), Sprayed concrete tunnel liners, Technical report, Department of Engi-
neering Science, Oxford University.

Pamin, J. K. (1994), Gradient-dependent plasticity in numerical simulation of localization

phenomena, PhD thesis, Delft University of Technology, Netherlands.

Panet, M. & Guenot, A. (1982), Analysis of convergence behind the face of a tunnel, in
‘Tunnelling ’82, Institute of Mining and Metallurgy, London’, pp. 197–203.

Parker, H., Fernandez-Delgado, G. & Lorig, L. (1977), A practical new approach to rebound
loss, in ACI, ed., ‘Shotcrete for Ground Support (SP-54)’, pp. 149–187.

Parrott, L. (1978), Effect of loading at early age upon creep and relaxation of concrete,
Technical Report UK5, RILEM Committee 42-CEA.

Peterson, A. (1989), Geostatische Untersuchungen für tiefliegende Regionalbahnen am

Beispiel Hannover, Technical report, University of Hannover, Germany.

Pölling, R. (2000), Eine praxisnahe, schädigungsorientierte Materialbeschreibung von

Stahlbeton für Strukturanalysen, PhD thesis, Ruhr-University Bochum, Germany.

Popovics, S. (1970), ‘A review of stress-strain relationships for concrete’, ACI Journal

67(14), pp. 243–248.

Potts, D. & Zdravković, L. (1999), Finite element analysis in geotechnical engineering -

Theory, Thomas Telford, London.

430
Potts, D. & Zdravković, L. (2001), Finite element analysis in geotechnical engineering -
Application, Thomas Telford, London.

Powell, D., Sigl, O. & Beveridge, J. (1997), Heathrow express - design and performance of
platform tunnels at Terminal 4, in ‘Tunnelling ’97’, pp. 565–593.

Probst, B. (1999), Entwicklung einer Langzeitdruckversuchsanlage für den Baustellenbetrieb

zur Bestimmung des Materialverhaltens von jungem Spritzbeton, Master’s thesis, Univer-
sity for Mining and Metallurgy, Leoben, Austria.

Pöttler, R. (1985), ‘Ideeller Elastizitätsmodul zur Abschätzung der Spritzebeton-

beanspruchung bei Felshohlraumbauten’, Felsbau 3(3), pp. 136–139.

Pöttler, R. (1990), ‘Time-dependent rock-shotcrete interaction - a numerical shortcut’, Com-

puters and Geotechnics 9, pp. 149–169.

Pöttler, R. (1993), To the limits of shotcrete linings, in Wood & Morgan, eds, ‘Shotcrete for
Underground Support VI, Niagara-on-the-Lake, Canada’, pp. 83–90.

Pöttler, R. & Klapperich, H. (1999), Single-shelled shotcrete lining aspects and application
in Central Europe, in Celestino & Parker, eds, ‘Shotcrete for Underground Support VIII,
Sao Paulo, Brasil’, pp. 174–192.

Pöttler, R. & Schweiger, H. (1999), Single-shell permanent lining reflections from the struc-
tural point of view, in Celestino & Parker, eds, ‘Shotcrete for Underground Support VIII,
Sao Paulo, Brasil’, pp. 57–66.

Rabcewicz, L. (1964), ‘The New Austrian Tunnelling Method’, Water Power (Nov.), pp.
453–457.

Rabensteiner, K. (1988), ‘Research project on the constitutive behaviour of shotcrete (per-

sonal communication with Schubert P.)’.

Rathmair, F. (1997), Numerische Simulation des Langzeitverhaltens von Spritzbeton und

Salzgestein mit der im FE-Programm Abaqus implementierten Routine, Master’s thesis,
University for Mining and Metallurgy, Leoben, Austria.

Reinhardt, H., Blaaunwendraad, J. & Jongedijk, J. (1982), Temperature development in

concrete structures taking account of the state dependent properties, in ‘International
Conference of Concrete at Early Ages’.

RILEM (1998), Prevention of thermal cracking in concrete at early ages, Technical Report 15.

Rokahr, R. B. & Lux, K. H. (1987), ‘Einfluß des rheologischen Verhaltens des Spritzbetons
auf den Ausbauwiderstand’, Felsbau 5(1), pp. 11–18.

431
Rokahr, R., Zachow, R. & Zander-Schiebenhöfer, D. (2005), Calculation of the stress intensity
index of a sprayed concrete lining, in E. . Solak, ed., ‘Underground Space Use: Analysis of
the Past and Lessons for the Future’, pp. 985–989.

Ross, A. (1937), ‘Concrete creep data’, The Structural Engineer 15(8), pp. 314–326.

Rots, J. (1988), Computational modelling of concrete fracture, PhD thesis, Delft University
of Technology, Netherlands.

Rowe, R., Lo, K. & Kack, K. (1983), ‘A method of estimating surface settlement above
shallow tunnels constructed in soft ground’, Canadian Geotechnical Journal 20, pp. 11–22.

Ryan, T. (1975), ‘Steel fibres in gunite: an appraisal’, Tunnels and Tunnelling 7(7), pp.
74–75.

Sargin, M. (1968), Stress-strain relationship for concrete and the analysis of structural con-
crete sections, PhD thesis, University of Waterloo, Ontario, Canada.

Sauer, G. (1988), ‘Further insights into the NATM’, Tunnels & Tunnelling (July), pp. 35–39.

Sauer, G., Gall, V., Bauer, E. & Dietmaier, P. (1994), Design of tunnel concrete linings using
capacity limit curves, in Siriwardane & Zaman, eds, ‘Computer Methods and Advances in
Geomechanics’, pp. 2621–2626.

SCA - Sprayed Concrete Association (1999), Introduction to sprayed concrete.

Schubert, P. (1988), ‘Beitrag zum rheologischen Verhalten von Spritzbeton’, Felsbau 6(3), pp.
150–153.

Schubert, W. & Grossauer, K. (2004), Evaluation and interpretation of displacements in

tunnels, in ‘14th Int. Conference on Engineering Surveying, Zürich, Switzerland’.

Schulze, H. & Duddeck, H. (1964), ‘Spannungen im schildvorgetriebenen Tunnel’, Beton und

Stahlbetonbau 59, pp. 169–175.

Sercombe, J., Hellmich, C., Ulm, F. & Mang, H. (2000), ‘Modelling of early-age creep of
shotcrete. I: Model and model parameters’, Journal of Engineering Mechanics 126(3), pp.
284–291.

Sezaki, M., Kibe, T., Ichikawa, Y. & Kawamoto, T. (1989), ‘An experimental study on the
mechanical properties of shotcrete’, Journal of the Society of Materials Science, Japan
38, pp. 106–110.

Shank, J. (1935), The plastic flow of concrete, Technical Report 91, Engineering Experiment
Station, Ohio State University.

432
Sloan, S. (1987), ‘Substepping algorithms for numerical integration of elasto-plastic stress-
strain relations’, International Journal for Numerical Methods in Engineering 24, pp. 893–
911.

Smith, S., William, K., Gerstle, K. & Sture, S. (1989), ‘Concrete over the top, or is there life
after peak?’, ACI Materials Journal 86(5), pp. 491–497.

Stang, H. & Aarre, T. (1992), ‘Evaluation of crack width in FRC with conventional reinforce-
ment’, Cement and Concrete Composites 14, pp. 143–154.

Stelzer, G. & Golser, J. (2002), Investigations regarding geometrical imperfections of

shotcrete linings - results from laboratory and from numerical calculations, in ‘Spritzbeton-
Technologie, Alpbach, Austria’, pp. 105–111.

Strack, M. (2007), Modellbildung zum rissbreitenabhängigen Tragverhalten von Stahlfaser-

beton unter Biegebeanspruchung, PhD thesis, Ruhr-University Bochum, Germany.

Straub, L. (1930), ‘Plastic flow in concrete arches’, Proceedings of the American Society of
Civil Engineers 56, pp. 49–114.

Swoboda, G. (1979), Finite element analysis of the New Austrian Tunnelling Method
(NATM), in ‘Proc. of the 3rd International Conference on Numerical Methods in Ge-
omechanics, Aachen, Germany’, pp. 581–586.

Swoboda, G. & Moussa, A. (1992), Numerical modelling of shotcrete in tunnelling, in Pande

& Pietruszczak, eds, ‘Numerical Models in Geomechanics’, pp. 717–727.

Thomas, A. (2003), Numerical modelling of sprayed concrete lined (SCL) tunnels, PhD thesis,
University of Southampton, UK.

Thomas, A. (2009), Sprayed concrete lined tunnels, Taylor and Francis.

Thomas, A., Legge, N. & Powell, D. (2004), The development of sprayed concrete lined (SCL)
tunnelling in the UK, in Schubert, ed., ‘53rd Geomechanik Colloquium, Salzburg, Austria’.

Thomas, F. (1933), ‘A conception of the creep of unreinforced concrete and an estimation of

the limiting values’, The Structural Engineer 11(2), pp. 69–73.

Thomée, B. (2005), Physikalische nichtlineare Berechnung von Stahlfaserkonstruktionen, PhD

thesis, Technical University Munich, Germany.

Timoshenko, S. & Goodier, J. (1951), Theory of elasticity, McGraw Hill, New York.

U.S. Bureau of Reclamation (1955), Investigation of creep in concrete: review of literature on

creep in concrete, Technical Report 1, U.S. Army Engineer Waterways Experiment Station,
Corps of Engineers, Vicksburg.

433
Vaishnav, R. & Kesler, C. (1961), Correlation of creep of concrete with its dynamic properties,
Technical Report 603, University of Illinois.

van Mier, J., Reinhardt, H. & van der Vlugt, B. (1987), ‘Ergebnisse dreiachsiger verformungs-
gesteuerter Belastungsversuche an Beton’, Bauingenieur 62, pp. 353–361.

Vermeer, P. & Brinkgreve, R. (1994), A new effective non-local strain-measure for softening
plasticity, in V. Chambon, Desrues, ed., ‘Localisation and Bifurcation Theory for Soils and
Rocks’, pp. 89–100.

Vermeer, P. & de Borst, R. (1984), ‘Non-associated plasticity for soils, concrete and rock’,
Heron 29(23), pp. 1–64.

Vonk, R. (1992), Softening of concrete loaded in compression, PhD thesis, Eindhoven Uni-
versity of Technology, Netherlands.

Walter, H. (1997), Application of a new shotcrete model in a 3-D FE-analysis of a tunnel

excavation, in Pande & Pietruszczak, eds, ‘Numerical Models in Geomechanics (NUMOG
VI)’, pp. 455–460.

Walter, H. (2003), Design of the shotcrete tunnel lining of a metro station - safety considera-
tions, in ‘Proceedings of the Euro-C Conference on Computational Modelling of Concrete
Structures’, pp. 855–865.

Watson, P., Warren, C., Eddie, C. & Jäger, J. (1999), CTRL north downs tunnel, in ‘Tunnel
Construction and Piling ’99’, pp. 301–323.

Weber, J. (1979), ‘Empirische Formeln zur Beschreibung der Festigkeitsentwicklung und des
E-Moduls von Beton’, Betonwerk- und Fertigteiltechnik 12, pp. 753–756.

Weigler, K. (1974), Junger Beton - Beanspruchung - Festigkeit - Verformung, Technical Re-

port 20, Institut für Massivbau, University Darmstadt, Germany.

Wierig, H. (1971), ‘Einige Beziehungen zwischen den Eigenschaften von “grünen” und “jun-
gen” Betonen und denen des Festbetons’, Beton 11+12, pp. 445–490.

Wierig, H. & Gollasch, E. (1967), Untersuchung über das Verformungsverhalten von jungem
Beton, Technical Report 47, University of Hannover.

Williams, I. (2008), ‘Heathrow Terminal 5: tunnelled underground infrastructure’, Civil En-

gineering, Proceedings of ICE 161(May), pp. 30–37.

Wolfsier, J. & Morgan, D. (1993), ‘Silica fume in shotcrete’, Concrete International 15(4), pp.
34–39.

Wolsiefer, J. & Morgan, D. (2003), ‘Silica fume in shotcrete’, Shotcrete Magazine 5(1), pp.
28–33.

434
Woods, R. & Clayton, C. (1993), The application of CRISP finite element program to prac-
tical retaining wall problems, in ‘ICE Conference on Retaining Structures, Cambridge’,
pp. 102–111.

Wu, W. & Roony, P. (2001), The role of numerical analysis in tunnel design, in D. Kolymbas,
ed., ‘Tunnelling Mechanics, Eurosummerschool, Innsbruck’, pp. 87–168.

Yin, J. (1996), Untersuchungen zum zeitabhängigen Tragverhalten von tiefliegenden

Hohlräumen im Fels mit Spritzbetonausbau, PhD thesis, Technical University Clausthal,
Germany.

Zachow, R. (1995), Dimensionierung zweischaliger Tunnel im Fels auf Grundlage von in situ
Messungen, PhD thesis, Technical University Hannover, Germany.

Zervos, A., Vardoulakis, I. & Papanastasiou, P. (2007), ‘Influence of nonassociativity on

localization and failure in geomechanics based on gradient plasticity’, International Journal
of Geomechanics 7(1), pp. 63–74.

Zheng, H. (1989), Beanspruchungen des Tunnelausbaus bei zeitabhängigem Materialverhalten

von Beton und Gebirge, PhD thesis, Technical University Braunschweig, Germany.

435