Benoît Jones - Soft Ground Tunnel Design-CRC Press (2022)

Soft Ground
Tunnel Design
Soft Ground
Tunnel Design
Benoît Jones
First edition published 2022
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
and by CRC Press
2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN
© 2022 Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of
their use. The authors and publishers have attempted to trace the copyright holders of all material
reproduced in this publication and apologize to copyright holders if permission to publish in this form
has not been obtained. If any copyright material has not been acknowledged please write and let us
know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information storage
or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.com
or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. For works that are not available on CCC please contact mpkbookspermissions@
tandf.co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks and
are used only for identification and explanation without intent to infringe.
Library of Congress Cataloging‑in‑Publication Data
Names: Jones, Benoît, author.

Title: Soft ground tunnel design / Benoît Jones.
Description: First edition. | Boca Raton, FL : CRC Press, [2022] | Includes index.
Identifiers: LCCN 2021013258 (print) | LCCN 2021013259 (ebook) | ISBN
9780367419592 (hbk) | ISBN 9781482254679 (pbk) | ISBN 9780429470387 (ebk)
Subjects: LCSH: Tunnels--Design and construction. | Tunneling. | Soil stabilization.
Classification: LCC TA815 .J66 2022 (print) | LCC TA815 (ebook) | DDC
624.1/93--dc23
LC record available at https://lccn.loc.gov/2021013258
LC ebook record available at https://lccn.loc.gov/2021013259
ISBN: 978-0-367-41959-2 (hbk)

ISBN: 978-1-4822-5467-9 (pbk)
ISBN: 978-0-429-47038-7 (ebk)
DOI: 10.1201/9780429470387
Typeset in Sabon
by KnowledgeWorks Global Ltd.
Contents
Prefacexv
Books on tunnel construction methods xvii
Acknowledgementsxix
Authorxxi
1 Real tunnel behaviour 1

1.1 In situ stress states 2
1.2 Overview of tunnel behaviour 3
1.2.1 Undrained soil behaviour 7
1.2.2 Drained soil behaviour 8
1.3 Movements of the ground surface 8
1.3.1 Transverse vertical settlements 9
1.3.2 Transient settlements 10
1.3.3 Horizontal surface movements 13
1.3.4 Long-term settlements 15
1.4 Subsurface ground movements 19
1.5 Stability 19
1.5.1 The consequences of instability 20
1.5.2 The causes of instability 22
1.6 Tunnel lining movements 24
1.6.1 Case studies of tunnel lining
movements 25
1.7 Tunnel lining stresses 28
1.7.1 How do tunnel lining stresses develop
over time? 28
1.7.2 Design based on precedent practice 33
1.7.3 Heathrow Express Terminal 4 Station
concourse tunnel case study 34
1.8 Summary 41
References 41
v
vi Contents
2 Undrained stability 45
2.1 Overview of stability theory 46
2.2 Undrained stability 47
2.2.1 Heading stability in homogeneous clay 49
2.2.2 Heading stability in clay with undrained
shear strength increasing with depth 59
2.2.3 Heading stability in clay with
overlying coarse-grained soils 61
2.2.4 Numerical modelling of heading stability
in clay 63
2.2.5 Summary of undrained stability 65
2.3 Blow-out failure in clay 66
2.3.1 Softening and erosion 66
2.3.2 Hydraulic fracturing in clay 67
2.3.3 Passive failure in clay 69
2.3.4 Uplift failure of a tunnel heading invert
in clay 76
2.3.5 Summary of undrained blow-out failure 76
2.4 Problems 77
References 80
3 Drained stability 83
3.1 Drained stability without seepage 83
3.1.1 Dry cohesionless soils 84
3.1.2 Dry drained soils with cohesion 85
3.1.3 Comparison of analytical methods with
centrifuge tests and finite element models 90
3.1.4 Summary of drained stability theory 92
3.2 Application of drained stability to closed-face TBMs 93
3.2.1 Application to slurry TBMs 93
3.2.2 Slurry infiltration during TBM standstills 96
3.2.3 Slurry infiltration during excavation 98
3.2.4 Application to earth pressure balance
TBMs 100
3.3 Blow-out failure in drained soils 103
3.3.1 Passive failure in drained soils 103
3.3.2 Blow-outs caused by hydraulic fracturing 111
3.3.3 Summary of blow-outs in drained soils 113
3.4 Piping 114
3.1 Problems 115
References 120
Contents vii
4 Stability of shafts123
4.1 Hydraulic failure in a shaft during excavation 123
4.2 Base heave failure of a shaft in clay 131
4.3 Uplift failure in a shaft during excavation 134
4.3.1 Verification of the uplift ultimate
limit state using Eurocode 7 134
4.3.2 Geometry of uplift failure during
excavation 138
4.4 Uplift failure of a shaft after base
slab construction 139
4.5 Long-term heave under a shaft base slab 145
4.6 Summary of shaft stability 146
4.7 Problems 146
References 148
5 Stability and Eurocode 7 151

5.1 Size of the zone of ground governing
the occurrence of the limit state 152
5.2 Correcting for confidence in the site
investigation 153
5.3 Modelling spatial variability of soil
parameters explicitly 155
5.4 Applying partial factors 157
References 158
6 Global design using analytical solutions 159

6.1 Simple wished-in-place equilibrium 160
6.2 Empirical methods 166
6.3 Soil-structure interaction using the
Curtis-Muir Wood solution 166
6.3.1 Notation used in the Curtis-Muir
Wood solution 167
6.3.2 Boundary conditions and ground
stresses 167
6.3.3 Elliptical deformation of a circular
opening 171
6.3.4 Elliptical deformation of a thin
inextensible lining 175
6.3.5 ‘Full slip’ - no shear between
lining and ground 177
viii Contents
6.3.6 ‘No slip’ - full shear interaction

between lining and ground 178
6.3.7 Direct compression of the lining
due to uniform load 178
6.4 Global design of shafts 183
6.5 Bedded beam models 184
6.6 Summary 186
6.7 Problems 187
References 189
7 Global design using numerical modelling 191

7.1 Boundary conditions at the tunnel perimeter 193
7.1.1 Wished-in-place tunnel lining 194
7.1.2 The convergence-confinement method 195
7.1.2.1 The β-factor method 198
7.1.2.2 The target volume loss
method 199
7.1.3 Gap method 202
7.1.4 The grout pressure method 202
7.1.5 Surface contraction 204
7.1.6 Core softening 204
7.1.7 Summary 204
7.2 Boundary conditions at the edges of the model 205
7.3 Boundary distances 208
7.4 Element types for the lining and the ground 209
7.5 Mesh density and refinement 210
7.6 Modelling groundwater 211
7.6.1 Undrained behaviour 212
7.6.2 Long-term effects 215
7.7 Validation and error checking 216
7.7.1 Comparison with an analytical
solution 217
7.7.2 Validation by comparison with a
laboratory test or experiment 219
7.7.3 Validation by comparison with
a case history 222
7.8 Constitutive models 223
7.9 Interpretation and presentation of results 224
7.10 3D numerical analysis 224
7.10.1 Modelling an advancing tunnel in 3D 224
7.10.2 Modelling the tunnel lining 228
Contents ix
7.10.3 Modelling junctions 231

7.10.3.1 Kirsch solution 231
7.10.3.2 Wished-in-place 3D numerical
model 235
7.10.3.3 3D numerical model with
sequential construction 236
7.11 Summary 240
7.12 Problems 241
References 244
8 Lining materials247
8.1 Reinforced concrete 248
8.2 Steel fibre-reinforced concrete 248
8.2.1 Codes of practice and sources of design
guidance 249
8.2.2 Material behaviour 250
8.2.3 Design assisted by testing 252
8.2.4 Determination of characteristic strength
values 254
8.2.5 Determination of the characteristic
mean strength values 257
8.2.6 Durability 258
8.2.7 Watertightness 259
8.2.8 Fire resistance 261
8.3 Concrete reinforced with other fibres 262
8.4 Plain concrete 262
8.5 Cast iron 263
8.6 Summary 265
References 265
9 Segmental lining design269

9.1 Taking account of the effect of joints
in 2D plane strain analyses 270
9.2 Rotational rigidity - flat joints 272
9.2.1 Linear elastic packers 274
9.2.2 Nonlinear packers 281
9.2.3 Comparison of different packer types 283
9.3 Estimating local forces due to joint rotation 286
9.3.1 Calculating joint rotation from
a specified ovalisation 286
x Contents
9.3.2 Using eccentricity of hoop force

to check joint capacity in crushing
and shear 294
9.4 Bursting stresses 299
9.5 Curved joints 301
9.6 Modelling joints in bedded beam analyses 304
9.7 Modelling joints in 2D or 3D numerical analysis 305
9.8 In service loads – ultimate limit state design 307
9.8.1 Application of partial factors to the lining
forces 308
9.8.2 Constructing a moment-axial
force interaction diagram 309
9.8.3 Design for shear 315
9.9 In service loads - serviceability limit state design 315
9.9.1 Estimating crack widths 315
9.9.2 Constructing a moment-axial force
interaction diagram for the SLS 316
9.10 Summary 317
9.11 Problems 318
References 320
10 Segment design for transient loads323

10.1 Demoulding 324
10.2 Storage 328
10.2.1 Actions 328
10.2.2 SFRC segment ultimate limit state
design for bending moment 333
10.2.3 SFRC segment ULS design for shear force 339
10.2.4 SFRC segment SLS design 339
10.3 Transportation and handling 348
10.4 Erection 349
10.5 Installation loads 349
10.5.1 Temporary TBM jacking loads 349
10.5.2 Longitudinal effects of TBM jacking loads 353
10.5.3 Grout pressures and tailseal brushes 355
10.5.4 Design for permanent effects of installation
loads 356
10.6 Gasket compression and bolt loads 357
10.7 Summary 359
10.8 Problems 360
References 361
Contents xi
11 Sprayed concrete lining design363

11.1 Primary and secondary linings 364
11.1.1 Single-pass lining 365
11.1.2 Single-shell lining 365
11.1.3 Traditional approach 366
11.1.4 Double-shell lining system 366
11.1.5 Composite shell lining system 367
11.1.6 Initial layer 369
11.1.7 Regulating layer 370
11.1.8 Finishing layer 370
11.2 Designing the profile 370
11.2.1 Three arc profile 371
11.2.2 Two arc profiles 372
11.2.3 Other profile shapes 373
11.2.4 Optimising the profile 373
11.3 Dividing the face 379
11.3.1 Full-face excavation 380
11.3.2 Top heading–invert (TH-I) excavation 381
11.3.3 Top heading–bench–invert (TH-B-I)
excavation 383
11.3.4 Single sidewall drift and enlargement 384
11.3.5 Twin sidewall drift and enlargement 384
11.3.6 Pilot tunnels 385
11.3.7 Binocular caverns 386
11.3.8 Trinocular caverns 387
11.4 Toolbox measures 387
11.4.1 Pocket excavation 388
11.4.2 Reduction of advance length
or ring closure distance 389
11.4.3 Ground improvement 390
11.4.4 Support ahead of the face 390
11.4.5 Face dowels 393
11.4.6 Dewatering/depressurisation 393
11.5 Construction details and tolerances 393
11.5.1 Circumferential construction joints 394
11.5.2 Radial construction joints 394
11.5.3 Tolerances 395
11.6 3D numerical modelling 396
11.6.1 Linear elastic sprayed concrete constitutive
model 396
11.6.2 Ageing sprayed concrete properties 397
xii Contents
11.6.3 Linear elastic – perfectly plastic sprayed

concrete constitutive model 399
11.6.4 More complex constitutive models 400
11.7 2D numerical modelling 401
11.8 Summary 403
11.9 Problems 404
References 407
12 Estimating ground movements411

12.1 Transverse surface settlements 411
12.2 Estimating volume loss 418
12.2.1 Estimating short-term volume loss
in stiff clays 419
12.2.2 Estimating volume loss in other
soil types 427
12.3 Longitudinal and transient surface settlements 427
12.4 Long-term ground movements 431
12.4.1 Long-term ground movements
due to excess pore pressures 432
12.4.2 Long-term ground movements due to
the tunnel lining acting as a drain 432
12.5 Subsurface ground movements 435
12.5.1 Subsurface ground movements
in undrained soils 435
in drained soils 441
12.6 Horizontal ground movements 443
12.7 Strains in the ground 446
12.8 Ground movements due to shaft construction 450
12.9 Summary 460
12.1 Problems 461
0
References 463
13 Estimating building damage467

13.1 Stage 1 assessment 469
13.2.1 Generic Stage 2 assessment 490
13.3 Stage 2b assessment 491
13.3.1 Buildings with discrete pad foundations 500
13.3.2 Steel or reinforced concrete frame
buildings 500
Contents xiii

13.4.1 Tunnel excavation 509
13.4.2 Soil-structure interaction 509
13.4.3 Modelling the building 510
13.5 Summary 511
13.6 Problems 512
References 514
Appendix A: Derivation of wedge-prism method519

Finding the vertical pressure on surface CDEF 520
Finding the sliding friction on surfaces ADE and BCF 525
Finding the sliding friction on ABEF 526
Calculating values for the F0 nomogram 529
Calculating values for the F1 nomogram 530
References 532
Appendix B: Details from derivation of Curtis–Muir

Wood equations 533
Integration by parts 533
Appendix C: Derivation of the deflection of a rectangular

simply-supported beam under a point load537
Bending deflection 538
Shear stress 539
Unit load method to calculate shear deflection – part 1 541
Form factor for shear for a rectangular section 544
Unit load method to calculate shear deflection – part 2 545
Comparison with Burland & Wroth (1974) equation 546
The effect of varying E/G ratio 549
References 551
Index 553
Preface
Dear Reader,
This textbook covers the design of tunnels and other underground spaces
in soft ground. ‘Soft ground’ means soil, as opposed to rock. The philoso-
phies of rock and soft ground tunnel design are different enough to merit
separate treatment.
In 2011, when I was setting up the MSc in Tunnelling and Underground
Space at the University of Warwick, I found that there was no comprehen-
sive textbook available on soft ground tunnel design. So I had to synthesise
the literature and my own experience to find the best ways to present this
subject to my students. Over several years the lectures and example prob-
lems were developed and improved. This book aims to extend and enhance
these teaching materials and make them more widely available.
Readers who will gain most from this book already have a civil engineer-
ing degree or similar and understand the basic principles of soil mechan-
ics and structural engineering, but have little or no experience of tunnel
design. My target audience is MSc students, graduate tunnel engineers and
experienced engineers who are new to soft ground tunnelling but I hope
that even for experienced tunnel design experts, there is something new or
interesting for you here and that you will find it a useful reference work,
compilation of methods and summary of the state of the art.
There are already several reasonably good (and very detailed) books
available on tunnel construction methods so I have not unnecessarily dupli-
cated that subject matter here. If you do not already know the basics of how
TBMs, shotcrete or segmental linings work, there is a book list below that I
would urge you to make use of as preparation for and in combination with
this textbook. There are also plenty of videos and other resources available
online.
There are many aspects of design that are not covered in this book,
including sustainability (encompassing health, safety and the environ-
ment), design for watertightness, fire, durability, spaceproofing and opera-
tional considerations. Site investigation and geological characterisation are
also not covered explicitly. These are all very important subjects and their
absence is purely due to space constraints.
xv
xvi Preface
This is first and foremost a textbook: you won’t just know a lot at the end
but will be able to do a lot. Each chapter begins by listing the knowledge,
understanding and skills you should have gained by the end of it. The use
of worked examples throughout this book and exercises at the end of most
chapters will give you an in-depth understanding and the confidence to
apply that understanding.
The three principal themes are stability, prediction of ground move-
ments, and structural design of the tunnel lining. These themes are the
key to understanding how tunnels work and are the basis for choices of
construction method and lining type. The reader is guided through the
basic principles of soil-structure interaction, the three-dimensional effects
of construction sequence and the effects of construction on other surface
or subsurface structures, in steps of gradually increasing complexity from
basic principles to sophisticated design. As the calculations become more
complex, it is more and more important that they remain grounded in an
in-depth understanding of how real tunnels behave. For this reason, the
very first chapter starts with case studies of real tunnel behaviour, to pro-
vide a conceptual framework of what happens to the tunnel lining and
what goes on in the ground around a tunnel in terms of displacements,
stresses and strains.
My final aim in this book is to collate and clearly present all the most
commonly used design methods from first principles. Even in the original
papers where these methods were first introduced, and more so in subse-
quent papers that cite and expand on them, there are sometimes missing
steps in the derivations, missing information, implicit assumptions that are
not explained, and even errors. It is alarming to consider how often these
errors have been inherited, used and disseminated in the intervening years.
I have made every effort to ensure that in this book these design methods
and their derivations are now set out in full; transparently and correctly. I
hope I’ve made it as easy as possible for you to spot any errors that remain
and draw them to my attention. Contact me, we will add them to the cor-
rigendum online at www.inbye.co.uk/soft-ground-tunnel-design and get
them fixed in the second edition.
Warmest Regards,
Benoît Jones
Books on tunnel
construction methods
The following books may be useful to gain an appreciation of tunnel con-

struction methods.
INTRODUCTION TO TUNNEL CONSTRUCTION
This is the only book intended as an introductory text for tunnel construc-
tion and as such is probably the one best suited to beginners.
Chapman, D., Metje, N. & Stärk, A. (2010). Introduction to tunnel

construction. Abingdon: Spon Press.
Another excellent reference that all tunnelling engineers should read is the
British Standard on health and safety in tunnelling.
BS 6164: 2019. Health and safety in tunnelling in the construction

industry – Code of practice. London: British Standards Institution.
TBM TUNNELLING
‘Mechanized tunnelling’ means tunnelling using a tunnel boring machine

(TBM). This book covers all the different types of TBMs and how to man-
age and control their use, particularly in urban areas:
Guglielmetti, V., Grasso, P., Mahtab, A. & Xu, S. (2008). Mechanized

tunnelling in urban areas – design methodology and construction
control. London: Taylor & Francis Group.
The following are essentially handbooks for practitioners. They are big and
heavy and cover almost every aspect of TBM use, but have very little to say
about design.
xvii
xviii Books on tunnel construction methods
Maidl, B., Herrenknecht, M., Maidl, U. & Wehrmeyer, G. (2012).

Mechanised shield tunnelling, 2nd Edition. Berlin: Ernst & Sohn.
Maidl, B., Thewes, M. & Maidl, U. (2013). Handbook of tunnel engi-
neering I – structures and methods, English edition translated by
David Sturge. Berlin: Ernst & Sohn.
Maidl, B., Thewes, M. & Maidl, U. (2014). Handbook of tunnel engi-
neering II – basics and additional services for design and construc-
tion, English edition translated by David Sturge. Berlin: Ernst &
Sohn.
SPRAYED CONCRETE
Sprayed Concrete for Ground Support is published by a product supplier

but is nevertheless a very good explanation of the use of sprayed concrete
in tunnelling, covering equipment, chemical admixtures, mix design and
application.
BASF (2014). Sprayed concrete for ground support. BASF Construction

Chemicals Europe Ltd.
Alun Thomas’s book is both an introduction to sprayed concrete tunnelling

and a reference guide for experienced practitioners. The second edition has
recently been released.
Thomas, A. H. (2019). Sprayed concrete lined tunnels, 2nd Edition.

Abingdon: Taylor & Francis.
Acknowledgements
First of all I would like to thank my civil partner Anna, who read the
entire first draft with a ruthless eye for anything unnecessary, confusing or
inelegant. She proposed myriad improvements, the vast majority of which
I adopted. Her patience and craftsmanship improved the book on almost
every page.
Secondly, I would like to thank everyone at Taylor & Francis, particu-
larly Frazer Merritt and Tony Moore, who have been incredibly patient as
deadline after deadline flew by.
Many friends and colleagues provided constructive comments on draft
chapters, in particular Ben Swatton, James Boswell, Dr Giorgia Giardina
and Sotiris Psomas.
My derivation of the Curtis-Muir Wood equations benefited from access
to handwritten notes from John Curtis himself, which explained some
(but not all!) the intermediate steps between equations in the published
papers. These were given to me by Nick Tucker when I was working at
Mott MacDonald.
My derivation of the wedge-prism method was aided by discussion with
and assistance from Dr Tiago Dias of the University of Southampton, and
correspondence with Prof. Dr Georgios Anagnostou of ETH Zurich. Prof.
Vicky Henderson of the University of Warwick helped me solve the differ-
ential equations.
Dr Giorgia Giardina of Delft University of Technology gave advice and
made many useful comments on the building damage assessments chapter.
Dr Stana Zivanovic of the University of Exeter helped me to understand
shear deflection and warping. Jamie Bradley showed me the importance of
the unit width assumption.
To all those others that corresponded with me to provide papers, data
or clarifications, a sincere thank you. These include Dr Alec Marshall of
the University of Nottingham, Prof. Daniel Dias of Polytech’ Grenoble,
Prof. Adam Bezuijen of Ghent University, Dr Jiang Su of Ramboll and John
Greenhalgh and Benoît de Rivaz of Bekaert.
xix
xx Acknowledgements
I tried for several years to write this book while working full-time in very
demanding jobs, but progress was slow. It has only been finished because
I was able to work part-time on a contract basis for an extended period. I
would like to thank London Bridge Associates and OTB Engineering for
providing me with the work to stay solvent during this time, in particular
Ken Spiby and Phil Astle. It was a bonus that much of the work they gave
me was very interesting and I learnt a lot while doing it. This book literally
wouldn’t have been possible without them.
Much of the bringing together of disparate sources to achieve the coher-
ent synthesis of the state-of-the-art presented in this book was done
between 2011 and 2015 to prepare teaching materials for the MSc in
Tunnelling and Underground Space at the University of Warwick, where I
was Course Director. I would like to thank all my students for their enthu-
siasm, patience and feedback.
This synthesis was developed further in a series of 32 articles for
Tunnelling Journal between 2012 and 2017. Thank you to Tris Thomas for
giving me that platform, and all his readers who gave me comments and
encouragement.
The British Tunnelling Society and the International Tunnelling
Association do so much to further the education and skills of their members,
and also to provide a forum for exchanging ideas and building friendships
and support networks. It is hard to imagine what a career in tunnelling
would be like if they didn’t exist. The writing of a book like this would
certainly be much harder.
I have benefited, over a career of more than 20 years now, from the
friendship, knowledge, experience and guidance of a large number of man-
agers, supervisors and colleagues – too many to name.
Finally I would like to thank the rest of my family – Christophe, Mum,
Dad, Damien and all the rest of you – for your love and support.
With heartfelt gratitude to all,
Benoît
Coventry, August 2021
Author
Dr Benoît Jones has two decades of experience in tunnelling as a designer,

researcher, contractor, academic, teacher, inventor and consultant. He is
a Chartered Civil Engineer and is the founder and Managing Director
of Inbye Engineering, a technology and consulting company established
in 2014.
Benoît graduated from the University of Bristol in 2000 with a First Class
Honours degree in Civil Engineering then joined Mott MacDonald as a
Graduate Tunnel Engineer. Between 2002-2006 he undertook research for
an EngD at the University of Southampton supervised by Professor Chris
Clayton. After working for Mott MacDonald on Crossrail MDC3, he moved
in 2007 to Morgan Est (now Morgan Sindall) as Section Engineer on the
King’s Cross Underground Station Redevelopment and later as Engineering
Manager on the Stoke Newington to New River Head Thames Water Ring
Main Extension Tunnel. In Autumn 2010 he joined OTB Engineering, a
small consultancy specialising in tunnelling and geotechnics, working on
temporary works designs for Tottenham Court Road Station Upgrade and
for the Crossrail Eastern Running Tunnels (C305). From 2011 to 2014 he
set up and then ran the MSc in Tunnelling and Underground Space at the
University of Warwick. Since 2015 Benoît has devoted his time to Inbye
Engineering, a company he set up in 2014 to deliver Strength Monitoring
Using Thermal Imaging (SMUTI) and provide engineering/tunnelling
design and consultancy services.
Benoît has led multiple research projects from within both industry and
academia and is the author of more than 60 articles and papers. He won
the ICE Telford Premium Prize for outstanding contribution to the civil
engineering literature in 2009 and was invited to give the ICE Geotechnical
Engineering Lecture in 2010 and 2012. He won two NCE Tunnelling
Awards and an ITA Austria Innovation Prize for SMUTI in 2016, followed
by an International Tunnelling Association award in 2017.
xxi
Chapter 1
Real tunnel behaviour
Even the most sophisticated design model is an idealised representation

of reality. In a model, it is neither possible nor desirable to include all the
complexity and uncertainty of ground and tunnel lining behaviour and
their interactions. Therefore, it is important to understand reality before
attempting to model it.
At different depths and in different geological and hydrogeological con-
ditions, and with different sizes of tunnels constructed using different
methods, different aspects of real behaviour will be more or less impor-
tant. It is a matter of experienced judgement, and of trial and error, to
decide which aspects of real behaviour are important enough to include
in the design model. The aim is to remove superfluous complexity so that
the design model outputs can be more easily interpreted and validated,
while retaining a design that is realistic. A designer who understands what
reality looks like will be better equipped to interrogate their model’s out-
puts and produce a reliable design.
This chapter describes what happens when real tunnels are constructed: the
ground movements, the tunnel lining movements, the effective stresses and
pore pressures in the ground, and the stresses in the tunnel lining, using data
from real projects. The objective of this chapter is to ‘upload a conceptual
model to your necktop’ (to paraphrase Daniel Dennett), so you know what
results to expect and you are able to achieve efficient and realistic designs.
After working through this chapter, you will understand:
• how the ground moves towards the face of an advancing tunnel

• what happens to pore pressures and effective stresses within the
ground as the tunnel excavation passes
• the factors affecting stability and what causes collapse or blow-out
failure
• how the ground and groundwater apply loads to the tunnel lining,
how these loads develop as the tunnel advances and what happens in
the long-term
• how all these aspects depend on the ground conditions, the depth of
the tunnel and the tunnel construction method
DOI: 10.1201/9780429470387-1 1
2 Soft Ground Tunnel Design
1.1 IN SITU STRESS STATES
Before a tunnel is excavated, there are existing ‘in situ’ stresses in the
ground. The principal stresses are usually in the vertical and horizontal
directions because they depend on gravity and the history of geological
deposition and erosion. Where there are steep slopes or tectonic stresses,
this may no longer be true, but this would rarely be the case for a tunnel in
soft ground.
Total stress in the ground is made up of two components: the pore pres-
sure and the effective stress, such that:
σ = σ ′ + u (1.1)
σ is the total stress in kPa

σ ′ is the effective stress in kPa
u is the pore pressure in kPa
The vertical stress in the ground is determined by the bulk unit weight of
the soil using the following equation:
σ v = γ z (1.2)
σ v is the vertical total stress in kPa

γ is the bulk unit weight of the soil, which is the weight of the soil
grains and pore water in kN/m3
z is the depth in m
Therefore, we can easily calculate the vertical total stress from basic knowl-
edge of the stratigraphy of the site and the bulk unit weight of the soil. If
the ground is made up of several layers with different values of bulk unit
weight, then they can be added together as follows:
σ v = γ 1z1 + γ 2 z2 + γ 3z3 + … + γ n zn (1.3)
γ 1, γ 2 , γ 3 … γ n are the bulk unit weights of soil layers 1, 2, 3 … in

kN/m3
z1, z2 , z3 … zn are the thicknesses of soil layers 1, 2, 3 … in m
The vertical total stress at the tunnel axis level is sometimes referred to as
the ‘full overburden pressure’.
In situ stresses are rarely equal in the vertical and horizontal directions.
The vertical stress is determined by the weight of the soil, but horizontal
stress depends on other factors as well. The relationship between horizontal
Real tunnel behaviour 3
and vertical effective stress is described by the coefficient of earth pressure

at rest, K0, which is given by:
σ h′
K0 = (1.4)
σ v′
K0 is the coefficient of earth pressure at rest

σ h′ is the horizontal effective stress in kPa
σ v′ is the vertical effective stress in kPa
In ‘normally consolidated’ soils, where the maximum vertical effective

stress in the soil’s history is the current value, the horizontal stress will be
lower than the vertical stress. In this case, the value of K0 may be given by
‘Jaky’s formula’:
K0 = 1 − sin φ ′ (1.5)
φ ′ is the soil’s angle of friction in a drained condition
Overconsolidated soils, where the soil has experienced a higher effective

stress at some point in its past than at present, perhaps because overlying
soils have subsequently been eroded or the water table has risen, can have
a higher horizontal effective stress than Jaky’s formula would predict and
they can even be significantly larger than the vertical effective stress, i.e. K0
may be greater than unity.
1.2 OVERVIEW OF TUNNEL BEHAVIOUR
During excavation of a tunnel, a void is formed in the ground. In an open-

face tunnel, the stress normal to the exposed surface of the excavation has
to be zero for there to be equilibrium. When a closed face tunnel boring
machine (TBM), such as an earth pressure balance (EPB) or slurry TBM, is
used, or when the tunnel face is pressurised using compressed air, the stress
normal to the excavated surface can be greater than zero and will be equal
to the applied pressure.
If the pressure applied to the surface of the excavation is less than the in
situ stress, then the ground will be unloaded, and will relax towards the
excavation. Therefore, tunnel construction nearly always induces ground
movements towards the tunnel, both radial displacements around the tun-
nel and longitudinal displacements towards the face. There will also be a
tendency for there to be more ground movement downwards than upwards,
due to gravity. An example of displacements at two different transverse sec-
tions of a sprayed concrete tunnel in London Clay is shown in Figure 1.1,
from a paper by Deane & Bassett (1995). London Clay is a stiff to very stiff
Figure 1.1 Subsurface displacement vectors inferred by Deane & Bassett (1995) from
inclinometers and extensometers installed at two transverse sections about
30 m apart to monitor the Heathrow Express Trial Tunnel construction.
Displacement vectors are exaggerated. (a) ‘Type 2’ side drift and enlarge-
ment, (b) ‘Type 3’ top heading, bench, invert.
Figure 1.2 Patterns of ground displacements above and ahead of the advancing
Heathrow Express Terminal 4 Station Concourse Tunnel (from van der Berg
et al., 2003). (a) Displacements on a vertical plane through the tunnel
centreline, and (b) Plan view of longitudinal displacements towards the top
heading, bench and invert.
overconsolidated clay. Figure 1.1 also shows the foci of the displacements,
marked by circles ‘C2’ and ‘C3’. In these cases, the foci lie somewhere
between the axis and the invert of the tunnel.
Figure 1.2 shows displacement vectors around an advancing tunnel face
from van der Berg et al. (2003). This was a shotcrete-lined tunnel excavated
in a top heading, bench, invert sequence, also in London Clay. Advances
were 1 m long for the top heading and bench, and 2 m for the invert.
Diagram (a) in Figure 1.2 is a long section through the tunnel and shows
how displacements developed in front of and then behind the face. Behind
the face, displacements continued to occur, but once the full ring had been
sprayed by closing the invert, displacements did not visibly increase further.
In diagram (b), plan sections show horizontal displacements at different
levels – the top deflectometer was in the top heading 1 m above axis level,
Figure 1.3 Principal stress tensors showing rotation of principal stress around the
tunnel and around the face (background image from numerical modelling by
Thomas, 2003).
the middle deflectometer was at axis level and the bottom deflectometer was
2.3 m below axis level. The strongly three-dimensional (3D) nature of ground
movements is evident – the longitudinal component of displacements is as
important as the radial. This is important to remember when we approxi-
mate tunnel construction using two-dimensional (2D) models.
The displacements in Figures 1.1 and 1.2 occur because the ground
around the tunnel heading is trying to find a new equilibrium. If it cannot,
then there will be a collapse. When a new equilibrium is found, it will look
something like Figure 1.3, which is a visualisation of principal stress magni-
tude and orientation in the ground around an advancing tunnel, in a block
cut from a larger 3D numerical model. In the transverse section on the right
hand side, we can see that as the ground has moved radially towards the
tunnel, the principal stresses have rotated and the major principal stresses
are now circumferential. This is referred to as ‘arching’. We can also see, in
the longitudinal section on the left hand side of the figure, that arching also
occurs around the face, which I like to call ‘front-to-back arching’.
The redistribution of ground stresses around a tunnel during its construc-
tion means that ground movements are inevitable. These ground move-
ments will have an impact on existing surface and subsurface structures.
By minimising the redistribution of stresses, either by applying an internal
pressure to the face or installing a stiff lining as early as possible, we can
minimise ground movements.
1.2.1 Undrained soil behaviour

The displacements shown in Figures 1.1 and 1.2 are ‘constant volume’.
This means that although the ground has deformed, the volume of the soil
has not changed. This has happened because these displacements are in
the short-term and the London Clay is behaving in an undrained manner.
‘Undrained’ means that pore water in the clay cannot flow during this time-
frame (that would take years or decades, whereas these readings have been
taken during construction over days or weeks). Therefore, virtually all the
changes in mean total stress caused by excavation are experienced as pore
water pressure changes, because water has a much higher bulk modulus
(i.e. it is much less compressible) than the soil particles. Since the pore water
is experiencing the change in stress, and water can be considered incom-
pressible at these stress levels, volume change will be negligible.
This undrained behaviour may be illustrated by the conceptual model in
Figure 1.4, representing Terzaghi’s principle of effective stress (Terzaghi,
1936). The spring represents the effective stress (the average of contact
stresses between soil particles), the water represents the pore water, and the
thin tube represents the permeability of the soil. As the loading changes, the
change in stress is immediately applied to the pore water. Over time, depend-
ing on the diameter of the tube, water will flow in or out of the container
and gradually the change in stress will transfer from the water to the spring.
As we have said, water has a high bulk modulus and is very difficult to
compress, but on the other hand, it cannot resist shear. Therefore, it is use-
ful to think of soil behaviour in terms of separate ‘shear deformation’ and
‘volumetric deformation’, governed by a shear modulus and a bulk modulus
respectively. For an undrained soil, the shear modulus will depend on the
soil particles and the bulk modulus will depend on the pore water.
Figure 1.4 Conceptual model illustrating undrained behaviour.

Figure 1.5 Volumetric strain contours from a centrifuge test by Marshall (2009),

showing zones of contraction and dilation. Note no data available in white
area around tunnel.
1.2.2 Drained soil behaviour

When a tunnel is constructed in a drained material, such as a sand or
gravel, which allows pore water to flow and pore water pressures to
reach either equilibrium or a steady-state within the timescale of con-
struction, then the soil can experience volumetric strains known as
‘contraction’ (negative volume change) or ‘dilation’ (positive volume
change, getting bigger). Sands that are not already in a dense state will
first contract at low shear strains, become denser, and then begin to
dilate at higher shear strains (Marshall et al., 2012). Therefore, differ-
ent zones of ground around a tunnel will contract or dilate, as shown in
Figure 1.5. This means that the ground displacement vectors will not all
point to one focus.
Another difference between drained and undrained behaviour is that
for undrained soils, the pattern of ground deformations is approximately
the same, regardless of the magnitude of deformations. For drained
soils, the pattern of ground movements, including their extent and the
positions of zones of dilation and contraction, varies as the magnitude
of deformations increases. We will examine this in more detail in the
following sections on surface and subsurface ground movements.
1.3 MOVEMENTS OF THE GROUND SURFACE
The ground movements towards the face of the tunnel we saw in

Section 1.2 cause settlements of the ground surface, as shown in Figure 1.6.
Transverse to the direction of tunnelling (in the y-axis direction), we
can see a surface settlement trough forming and along the centreline (the
x-axis direction), we can see an ‘S’-shaped longitudinal surface settlement
Figure 1.6 Movements of the ground surface (based on Attewell et al., 1986).
curve forming. The transverse surface settlement curve (or ‘trough’) has
a shape that may be described by an inverted normal distribution curve,
sometimes called a bell curve or a Gaussian curve, with the frequency of
the mean representing the maximum settlement Smax over the centreline of
the tunnel. The longitudinal surface settlement curve has a shape that may
be described by an inverted cumulative normal distribution curve.
1.3.1 Transverse vertical settlements

An example of transverse surface settlements measured above a tunnel
in London Clay (Standing et al., 1996; Nyren et al., 2001) is shown in
Figure 1.7. This was a 4.85 m diameter open-face TBM with a precast con-
crete expanded wedgeblock segmental lining installed behind the shield.
The depth to the tunnel axis was 31 m.
The settlements shown in Figure 1.7 were measured in the middle of
St James’s Park. These are referred to as ‘greenfield’ settlements because
they are not affected by anything human-made, such as buildings, con-
crete slabs, piles or underground structures. Only measurements on one
side of this particular tunnel are shown, because the settlement array only
extended about 5 m in the other direction. Usually surface settlements are
Figure 1.7 Short-term surface settlements measured after TBM passage at St James’s

Park, London (data from Grammatikopoulou et al., 2008).
symmetrical about the tunnel centreline, as long as the geology, surface

levels and excavation process are symmetrical.
Settlements in the middle of a park are unlikely to cause problems, but
greenfield settlements are very useful because our design models often
assume the tunnel is built in a greenfield situation, and we can use this data
to calibrate them.
1.3.2 Transient settlements

As a tunnel approaches and then passes a transverse settlement monitor-
ing array, the settlements gradually increase, as illustrated in Figure 1.6. A
real example from Crossrail on the North side of Hyde Park is shown in
Figure 1.8 (Wan et al., 2017). The depth to the tunnel axis was 34.5 m.
The TBM was 7.1 m diameter and used ‘EPB’ technology to apply a sup-
port pressure to the face, so even though this tunnel was much larger, the
settlements were smaller in magnitude than at St James’s Park. Face pres-
sures were approximately 200 kPa, which corresponds to about 30% of the
‘full overburden pressure’, i.e. the vertical total stress at tunnel axis level
(Wan et al., 2019). Tailskin grouting pressures were approximately 100 kPa
(Wan et al., 2019). The first set of measurements shown are from when the
Figure 1.8
Surface settlements as an EPB approaches and passes a settlement
monitoring array on the North side of Hyde Park (Wan et al., 2017).
face of the TBM was at −39.9 m approaching the monitoring array and the
last set are from when the face was at +42.7 m beyond it.
Figure 1.8 shows that settlement initially increases slowly, then as the
face gets closer to the array, the settlement increases more quickly. The
rate then decreases again as the TBM moves further away. Note that
the settlements were not monitored at regular TBM distance intervals,
but increase in frequency as the tunnel passed directly underneath. The
development of settlements as the tunnel advances can be demonstrated
more clearly by plotting the settlements against face position, as shown in
Figure 1.9. The largest magnitude settlements are the ones that are closest
to the centreline, i.e. YSMP10 and YSMP12 (labelled as ‘Y10’ and ‘Y12’)
in Figure 1.8.
In this case, when the face was under the array, the settlement was only
about 30% of the final short-term settlement. This was because more
ground movement could occur behind the face around the shield and
during the grouting process than in front of the face, because the face
pressure was higher than the grouting pressure. Based on several case
histories in firm to stiff clays, Attewell & Woodman (1982) found that
when tunnelling with no support pressure, between 30% and 50% of the
final short-term settlement occurs ahead of the face with an average value
at about 40%.
In other types of soil, Craig (1975) suggested the following percentages
of final short-term settlement would occur, shown in Table 1.1.
Figure 1.9 Longitudinal settlements on the Y-line array plotted against distance of TBM
cutterhead from the array (Wan et al., 2017).
Another example, from the Second Heinenoord Tunnel in the

Netherlands, is given in Figure 1.10 (van Jaarsveld et al., 2006). The
ground was loose and medium dense sand and soft clay, and two bores
were excavated by a large slurry TBM of 8.3 m diameter. Due to the soft
soils and the high groundwater pressure, face pressures closer to the in
situ stress were applied compared to the Crossrail EPB TBM that was
shown in Figure 1.9, resulting in very little settlement ahead of the face.
The reason that the second passage of the TBM resulted in less settle-
ment than the first was that more grout was injected, presumably at a
higher pressure. Figure 1.10 shows the same ‘S’-shaped longitudinal set-
tlement curve, illustrating that this general pattern of behaviour occurs
in all types of soils and groundwater conditions.
Table 1.1 Percentages of final short-term settlement above the face

and the tail of the shield (Craig, 1975).
Type of ground Face Tail of shield
Sand above water table 30–50% 60–80%
Sand below water table 0–25% 50–75%
Silts and soft clays 0–25% 30–50%
Figure 1.10 Longitudinal settlement profiles at various locations over the first and
second TBM passages at the Second Heinenoord Tunnel in the Netherlands
(van Jaarsveld et al., 2006).
1.3.3 Horizontal surface movements

Horizontal ground surface movements are much more difficult to measure.
This is because using a precise level to obtain vertical displacements is more
precise and accurate than using a total station to obtain 3D displacements.
There are more accurate methods to measure horizontal displacements
between fixed points, for example using micrometer sticks (Nyren et al.,
2001), but these are relatively cumbersome and require monitoring points
that are above ground, which cannot be easily achieved in public areas. For
these reasons, measurements of horizontal ground surface movements are
much less common than vertical movements.
Nyren et al. (2001) measured horizontal transverse and horizontal lon-
gitudinal movements as the tunnel face passed the monitoring array at
St James’s Park, shown in Figures 1.11 and 1.12, respectively.
In Figure 1.11, positive displacements are in the positive offset direction.
Therefore, horizontal movements were mostly towards the tunnel. The
point at which there was zero horizontal displacement should be directly
above the tunnel centreline, but in this case it was 2.3 m to the right of the
centreline. This asymmetry may have been due to a number of factors –
geology, excavation method or measurement accuracy (the quoted accuracy
of the micrometer stick measurements was ±1 mm).
Figure 1.12 shows the longitudinal horizontal displacements as the TBM
passed the array. These were measured using a total station. They are base-
lined to measurements made before the TBM was in the zone of influence.
Figure 1.11 Transverse horizontal displacements after passage of the Westbound

tunnel at St James’s Park (Nyren et al., 2001).
Figure 1.12 Longitudinal horizontal displacements during passage of the Westbound

TBM at St James’s Park (Nyren et al., 2001).
When the face was 7 m before the array, the ground surface had already
moved towards the face, with a maximum of around 5 mm of movement
above the tunnel centreline. The next survey was made when the face was
2.5 m past the array and the ground had moved even more. This was then
reversed as the face moved past the array, because increments of ground
movement were still orientated towards the face and shield of the TBM.
By the time the face was 32 m past the array, longitudinal movements had
returned to approximately zero, with some positive values close to the
centreline.
1.3.4 Long-term settlements

Long-term settlements are caused by pore pressure changes in the soil. For
drained soils, there will be no long-term pore pressure changes because
these will have all occurred in the short-term (this is, after all, the defini-
tion of drained behaviour – that pore pressures will find a steady-state or
equilibrium during the timescale of construction). An exception to this rule
would be if pore pressures were expected to change after construction,
perhaps due to switching off dewatering pumps, or any other foreseeable
change in the groundwater table. If the pore pressure increases, then effec-
tive stress will decrease, causing heave. If the pore pressure decreases, then
effective stress will increase, causing settlement.
For undrained soils, there will always be long-term settlements caused by
the dissipation of excess pore pressures over time. This dissipation is driven
by pore water flow away from areas where pore pressures are higher than
they should be (due to an increase in mean total stress during construc-
tion) and into areas where pore pressures are lower than they should be
(due to a decrease in mean total stress during construction). This dissipa-
tion continues until pore pressures stabilise at a steady-state or hydrostatic
equilibrium. In a soil with low permeability, this can take many years. As
the excess pore pressures dissipate, the effective stresses in the soil will also
change, and the soil will contract or dilate.
Referring back to Figure 1.4, if the piston is loaded and this causes an
increase in water pressure in the short-term, then over the long-term water
will be squeezed out of the tube until hydrostatic equilibrium is attained
and the load has been transferred to the spring. The spring has a much
lower stiffness than the water (which is virtually incompressible) and so the
piston will move downwards as the spring takes the load. This is analogous
to consolidation settlement. If unloading has occurred, then the opposite
will occur and over the long-term water will be drawn down the tube until
hydrostatic equilibrium has been attained. In this case as the load is trans-
ferred to the spring, the piston will be pushed upwards, which is analogous
to heave.
During this period of long-term settlement due to dissipation of excess
pore pressures, ground behaviour will not be constant volume and
Figure 1.13 Total stress (S) and pore pressure (PP) measured by a spade cell and
horizontal movement (H) measured by an inclinometer, as the Heathrow
Trial Tunnel passes (redrawn from New & Bowers, 1994).
displacement vectors will not be directed to a single point. Wherever there

has been a decrease in mean total stress, which in the short-term will cause
a decrease in pore pressure, there will be swelling in the long-term as water
flows from surrounding areas to equilibrate the difference. Wherever there
has been an increase in mean total stress, which in the short-term causes
an increase in pore pressure, there will be consolidation in the long-term as
water is squeezed out and flows away. If you are unclear about this, refer to
Section 1.2.1 again where undrained behaviour is explained.
Figure 1.13 shows an example of undrained behaviour close to an advanc-
ing shotcrete tunnel at Heathrow, UK (New & Bowers, 1994). As the side
drift and then enlargement passes, the ground moves towards the excava-
tion, causing a drop in total stress and pore pressure.
Pore pressures in undrained soils are affected by changes in mean total
stress. In addition, tunnels in soft normally consolidated clay may gener-
ate positive excess pore pressures, as soft clays will try to contract when
sheared, whereas tunnels in stiff overconsolidated clay will predominantly
generate negative excess pore pressures in the surrounding ground as they
try to dilate when sheared.
Therefore, as a tunnel advances, different zones of the ground are loaded
or unloaded, principal stresses rotate (see Figure 1.3), and pore pressures
change due to changes in both mean total stress and shear. This occurs in a
complex 3D manner that will depend strongly on the construction method,
the stratigraphy and the soil properties.
To minimise these long-term movements, the construction process should

minimise changes in the stress state of the ground. However, even with a
closed-face TBM applying support pressure very close to the in situ stress,
there will always be changes to the stress state of the ground, and so some
long-term movements are inevitable.
Long-term settlements due to the construction of the two running tun-
nels for the Jubilee Line Extension under St James’s Park in London Clay
are shown in Figure 1.14. Both tunnels were excavated in 1995/1996 by
a 4.85 m diameter open-face TBM with a backhoe. These values were
obtained from extensometers 5 m below the ground surface, since sur-
face settlement levelling points were removed before 1999 (Avgerinos
et al., 2016). The first tunnel constructed was the Westbound (WB), passing
this location from 27th to 28th April 1995 (Nyren et al., 2001), and a typi-
cal transverse settlement trough was created. The Eastbound (EB) tunnel
passed through 8 months later, from 8th to 10th January 1996. During this
8 month period (compare ‘After WB’ with ‘Before EB’ in Figure 1.14), the
settlement increased, and also appeared to spread over a wider area. The
next set of readings, on 1st March 1997, 14 months on from the ‘After
EB’ data set, the settlement increased even more. The next readings, on
1st August 2006 more than 10 years after the Eastbound tunnel was con-
structed, and on 6th August 2011 more than 15 years after the Eastbound
Figure 1.14 Long-term settlements due to the construction of Westbound (WB) and

Eastbound (EB) Jubilee Line Extension running tunnels under St James’s Park
(data from Avgerinos et al., 2016).
tunnel was constructed, show the settlements continue to increase to quite

large values. There is evidence in these last three readings that the trend has
been slowing to a long-term equilibrium or steady-state value.
For tunnels in stiff overconsolidated clay, the predominantly negative
excess pore pressures generated during tunnel construction should cause
heave in the long term as the clay draws in water from surrounding areas
and swells. So what is causing the long-term settlements we see in stiff
clays? Numerical modelling by Shin et al. (2002) shows that if the tunnel is
acting as a drain, then there will be long-term settlements. This is because
a permanent reduction in pore pressure caused by the tunnel draining the
ground around it will cause consolidation, in the same way as tree root
suctions will cause settlement of a building. If the tunnel lining is imperme-
able, then there will be a slight heave of the ground in the long term.
In soft, normally consolidated or lightly consolidated clays, we can expect
the magnitude of settlements to be greater than for stiff over-consolidated
clays in both the short and long term. Belshaw & Palmer (1978) presented
results from extensive monitoring of the Thunder Bay Tunnel in Ontario,
Canada. The measurements at one settlement point above the tunnel cen-
treline are shown in Figure 1.15. At the monitoring location, the stratigra-
phy consisted of about 3 m of silty sand underlain by 3.1 m of silt and then
17.7 m of silty clay and clay extending to bedrock. The consistency of the
Figure 1.15 Long-term settlement monitoring of a surface settlement pin over the

Thunder Bay Tunnel centreline (Belshaw & Palmer, 1978).
clay was soft above the tunnel, becoming firm at tunnel invert level. The
tunnel was constructed by an open-face TBM with an excavated diameter
of 2.47 m at a depth to axis of about 11 m. Although the area of the face
was only about a quarter of the size of the St James’s Park tunnels, the
short-term settlements at Thunder Bay were nearly double the magnitude,
due to the softer ground.
Figure 1.15 shows that over the first year after construction, settlements
continued to increase and it is likely the trend would have continued had
measurements been made over a longer period of time.
1.4 SUBSURFACE GROUND MOVEMENTS
Measuring subsurface ground movements is more difficult than measuring

surface settlements, and so there are fewer data available. Usually, these
movements have to be measured by installing instruments in boreholes to
measure relative movements or rotations at different depths. Some exam-
ples of this can be found in Section 1.2.
Vertical subsurface ground movements can be measured by extensometers.
A casing containing magnets is grouted into a borehole. An extensometer can
then be lowered into the casing to measure the relative distances between the
magnets. The level of the top of the casing must be measured as well at the
same time as every set of extensometer readings, usually with a precise level.
Therefore, to obtain a subsurface settlement trough, a number of boreholes
are required. This is expensive, and difficult to achieve in an urban area, so
most case studies of subsurface settlements tend to be in parks or other open
areas and they are far less common than surface settlement monitoring.
Horizontal subsurface ground movements can be measured by inclinom-
eters. A special casing is grouted into a borehole. The inclinometer is then
lowered into the borehole, or it could be left in place. The inclinometer has
several articulating sections and is capable of measuring changes in rota-
tion between each section. The changes in rotation can be used to calculate
horizontal displacements in two orthogonal directions.
It is possible to install a combined extensometer/inclinometer in a single
borehole. An example of this is the proprietary system called ‘shape accel
array’ (Lipscombe et al., 2015).
The vectors shown previously in Figures 1.1 and 1.2 were inferred from
inclinometers and extensometers installed in boreholes.
1.5 STABILITY
Stability in geotechnical engineering is about avoiding failure or collapse.

Instability is where dry soil, or soil and water fall or flow in an uncontrolled
manner.
As we have seen in the preceding sections, as the tunnel is advanced,

ground movements occur until the stresses in the ground and the support
find a new equilibrium. If a new equilibrium cannot be found, because the
destabilising forces are larger than the resistance provided by soil strength
and support forces, then the ground movements will continue increasing
until we have a collapse.
Chapters 2–5 will describe how to calculate and quantify stability as a
factor of safety. When stability is quantified in this way, the factor of safety
is a measure of how close we are to a global failure and we can expect that
when the factor of safety is large, ground movements will be small, and
when the factor of safety is approaching 1, ground movements will be large.
This relationship between the magnitude of ground movements and stabil-
ity will be further developed in Chapter 12.
1.5.1 The consequences of instability

In an open-face tunnel, instability may cause collapse, endangering the
underground workforce and equipment, and surface infrastructure and
people, who may fall into the hole or have a building collapse on top of
them (Figure 1.16). In a tunnel being excavated by a closed-face machine,
Figure 1.16 Collapse of the Munich Metro (Construction Today, 1994).

workers may be protected, but instability of the face can cause overexcava-
tion, excessive settlements of nearby structures and large holes opening up
at the surface. These can all be very dangerous to the general public and
are best avoided.
Cohesionless soils below the water table can be so unstable that they
will flow through a small hole if even a small hydraulic gradient is present
(Figure 1.17). In cases like these, the integrity of the lining and its gaskets
is very important, as any flow of soil into the tunnel means a void forming
outside the tunnel, which leads to unequal pressures acting on the lining
and possibly structural failure of the ring. It has been known for tunnels
to be flooded by silt and/or sand and water flowing through small gaps or
holes in the lining, leading to total abandonment of the TBM and a com-
plete restart (e.g. in Preston, UK: Thomas, 2011) or costly rescue works
(e.g. Hull wastewater flow transfer tunnel, UK: Brown, 2004; Grose &
Benton, 2005, 2006). Even in well-sealed tunnels being constructed using
a closed-face TBM, massive overexcavation can occur if the support pres-
sure is insufficient, as the ground will flow towards the cutterhead of the
machine (e.g. Storebælt Eastern Railway Tunnel, Denmark: Biggart &
Sternath, 1996).
Stability can also be a problem if there is too much support pressure in
the tunnel. For instance, if compressed air is used and the pressure is too
high, it can blow out to the surface as it did on the Docklands Light Railway
Lewisham Extension in 1998 (New Civil Engineer, 1998). A crater 22 m
wide and 7 m deep was created, and all the windows of the adjacent school
building were broken. Fortunately, it was the weekend and no one was hurt.
Heading stability is the single most important aspect of soft ground tun-
nelling. It is often what determines the choice of construction method, for
both conventional tunnelling and mechanised tunnelling, as shown in the
lists below. These decisions must be based on a thorough understanding of
the geology and the geotechnical behaviour.
Figure 1.17 Sand and water flowing through an opened grout plug in an old cast iron
lining.
Mechanised tunnelling (i.e. TBM) decisions:
• choice of TBM type, i.e. open-face, slurry, EPB

• choice of segmental lining type, i.e. bolted with gaskets or expanded
wedgeblock
• method for head interventions, i.e. atmospheric, compressed air or
use of divers, or use of ground improvement such as grouting or
dewatering
Conventional tunnelling (e.g. backactor or roadheader followed by shot-

crete lining) decisions:
• choice of construction sequence, how face is divided, when invert is

closed
• choice of available contingency measures
• choice of support types and toolbox measures, such as face shotcrete,
temporary inverts, canopy tubes, spiles and face dowels
• whether ground improvement is needed, e.g. permeation grouting, jet
grouting, ground freezing, dewatering
1.5.2 The causes of instability

Four main factors influence stability:
• Cohesion and support pressure, which improve stability

• Gravity and seepage forces, which worsen stability
Without either cohesion to hold the soil grains together or a support pres-
sure applied to the face, a vertical face will fall down due to gravity or
seepage forces. This is illustrated by Figure 1.18 which shows images from
a 1 g model test (i.e. a scale model but not in a centrifuge) of a heading in
dry sand by Messerli et al. (2010). As the plunger is retracted the sand falls
into the tunnel ending up with an angle of repose approximately equal to
its angle of friction.
Gravity is always present, but seepage forces only occur when there is
a hydraulic gradient (also called a head difference) between the ground
and the tunnel. As groundwater seeps through the ground towards the
face it pushes the soil grains apart with a ‘seepage force’ proportional to
the hydraulic gradient, in the direction of flow. This decreases stability.
Conversely, if the head of a TBM were filled with water or slurry with
a higher pressure than the groundwater pressure, it would flow into the
ground, and this would aid stability.
A cohesionless soil needs only a very small hydraulic gradient for it to
fail due to seepage, just a few centimetres of head is enough, as anyone who
has tried to build a dam on a sandy beach will know. On the other hand, a
Figure 1.18 Gravity failure of dry sand in a 1 g model (from Messerli et al., 2010).
moist (the technical term is ‘partially saturated’) sand above the water table
can have a small amount of apparent cohesion caused by capillary suction
in the pores, as anyone who has built a sandcastle will also know. This may
be just enough for a drained material to remain standing in small exposures
for a short time. However, even a small amount of perched groundwater
could cause local instability, so an open-face tunnel in these ground condi-
tions needs to be planned and executed with great care.
Clays can often be observed standing in vertical faces, sometimes in
large diameter open-face tunnels with no face support. This is because
clays have cohesion. This cohesion is largely due to the clay’s very
low permeability. During the timescale of construction, as the soil is
unloaded by removal of the soil that used to be next to it, the soil grains
relax towards the face and this causes the pore water pressure between
the soil grains to decrease. This drop in pore water pressure, known as
‘suction’ or ‘negative excess pore pressure’, holds the grains together. The
low permeability of clay means that positive or negative excess pore pres-
sures can exist for a long time because it takes so long for groundwater
to flow to or from the surrounding ground to dissipate them. This is why
it is called ‘undrained’ behaviour.
When these negative excess pore pressures are eventually dissipated, the
clay will behave in a drained manner, i.e. more like a sand. Drained cohe-
sion in clays is usually very small or zero. As shown above, all headings will
fail without either cohesion or support pressure; it’s just that in clays, it may
take a very very long time.
For all these reasons, drained and undrained stability are quite different.
The geometry of failure is different, and the calculations are done in a dif-
ferent way, so in this book they have been split into two separate chapters –
Chapter 2 for undrained stability and Chapter 3 for drained stability.
A rule of thumb is that a soil with a permeability less than 10 −7 to 10 −8 m/s
will behave in an undrained manner during the timescale of a typical tunnel
construction, and a soil with higher permeability will behave in a drained
manner (Anagnostou & Kovári, 1996). If it is unclear whether a soil will
Figure 1.19 Geometry of stability failure in clay and sand (redrawn from Mair & Taylor,
1997), in centrifuge tests on clay by Mair (1979) and on sand by Chambon &
Corté (1994).
behave in a drained or undrained manner, stability calculations need to be

done for both cases and the worst case used for design. It is important to
remember that real soils are variable, heterogeneous and rarely isotropic,
and that ground mass behaviour can be different to the behaviour of small
samples in geotechnical laboratory tests.
Interestingly, undrained stability depends on the depth of the tunnel,
whereas drained stability is independent of depth and depends only on tun-
nel diameter. Another difference is the shape of the stability failure. As
shown in Figure 1.19, in the drained case, the soil fails in a steep-sided
chimney (see also Figure 1.18), which may or may not extend to the sur-
face, whereas in the undrained case, the failure geometry has a much wider
extent and is more of a cone shape.
1.6 TUNNEL LINING MOVEMENTS
After installing a tunnel lining, the lining’s self-weight and the ground and
groundwater pressures acting on the outside of the lining will cause it to
deform. Since in soft ground the lining is installed either close to the face
or at the back of the shield, but cannot be installed ahead of the face, it will
not be directly affected by ground movements ahead of the face of the tun-
nel, but only those behind the face or shield. This means that the lining will
deform much less than the ground around the tunnel and thus monitoring
of lining movements tells us only part of the story of ground deformation,
but rather more about the interaction between the lining and the ground.
Once installed, the lining will interact with the ground until an equilib-
rium state is found. This will depend on the stiffness and thickness of the
lining and how many joints it has, as well as the ground behaviour. In soft
ground, it is usual for inward movements of the tunnel lining, sometimes
referred to as ‘convergence’, to be larger in the vertical direction than in
the horizontal direction. This type of deformation is called ‘squatting’ and
occurs because the vertical ground pressure applied to the lining is usually
higher than the horizontal.
Even in overconsolidated soils with K0 greater than 1, it is usual for tun-
nel linings to squat, as evidenced by large numbers of convergence mea-
surements made in London Underground tunnels by Wright (2013), where
average squat was between 0.5% and 1.0% of diameter. This may be due
at least in part to the tunnel acting as a drain on the ground around it.
The higher horizontal than vertical permeability in the London Clay (and
indeed all horizontally bedded sedimentary soils, which is most of the
soft ground on the planet) results in an asymmetric draw-down and hence
larger consolidation strains in the vertical than the horizontal direction.
This results in a higher vertical ground pressure applied to the lining than
horizontal, causing squatting.
1.6.1 Case studies of tunnel lining movements

Usually, lining deformations in soft ground are small. If a total station is used,
which is common practice because it is easy, convenient and already needed
for setting out, then the accuracy of the measurements is approximately
±2–3 mm (Bock, 2003). For a concrete lining, either precast or sprayed, we
expect strains at working loads to be less than 0.1% (e.g. Muir Wood, 2000).
This strain corresponds to 1 mm change in diameter per metre diameter, thus
4 mm convergence of opposite points on the lining for a 4 m diameter tunnel,
or 10 mm convergence for a 10 m diameter tunnel (Jones, 2007). Therefore, in
most cases, at expected working loads, the magnitude of the lining displace-
ments will not be much more than the accuracy of measurement with a total
station.
The purpose of lining displacement monitoring in a soft ground tunnel
is not to understand what is going on in any detail, but to give warning
of something catastrophic occurring. Some typical data from lining dis-
placement monitoring at Heathrow Terminal 4 station are given by Clayton
et al. (2003) and shown in Figure 1.20.
In Figure 1.20, we can see that the vertical lining displacements were larger
than the horizontal, despite this tunnel being constructed at 18 m depth in
London Clay at Heathrow Terminal 4, where the site investigation indicated the
K0 value was 1.5. This was partly due to redistribution of stresses in the ground
around the heading, and partly due to prior construction of adjacent platform
tunnels, which would have reduced the horizontal stress in the ground.
An extreme example of lining displacements, from monitoring of the
Heathrow Express Concourse Tunnel at the Central Terminal Area leading
up to its collapse, is shown in Figure 1.21 (HSE, 2000).
Figure 1.20 Typical lining displacements (exaggerated scale) for a sprayed concrete

lining in soft ground (data from Table 2 in Clayton et al., 2003).
Figure 1.21 Lining displacements measured in Heathrow Express CTA concourse tun-

nel Ch.30 leading up to collapse (redrawn from HSE, 2000; contains public
sector information published by the Health and Safety Executive and
licensed under the Open Government Licence).
Figure 1.22 Vertical and horizontal lining displacements in the Heathrow Express

CTA Station Concourse tunnel Ch.30 plotted against time leading up to
collapse on 20 th/21st October 1994 (redrawn from HSE, 2000; contains
public sector information published by the Health and Safety Executive and
licensed under the Open Government Licence).
The lining displacements shown as vector plots in Figure 1.21 can

also be viewed in graph form against time in Figure 1.22. You can see
that displacements never really stabilised and continued increasing over
4 months before collapse occurred. The magnitude of the lining displace-
ments (compare to Figure 1.20) was much larger than should have been
expected based on the trial tunnel constructed beforehand (Deane &
Bassett, 1995).
There were many factors involved in the collapse, which were dis-
cussed at length in the Health and Safety Executive report (HSE, 1996),
but even so, the lining displacements should have provided sufficient
warning that the tunnel lining was suffering distress, but the monitor-
ing data was ignored. The lining suffered from poor workmanship and
quality control, but the beneficent ground conditions meant this did
not result in immediate collapse. However, as well as time-dependent
behaviour of the London Clay, there were construction operations that
would have increased the load on the lining with time, such as sub-
sequent construction of the adjacent platform tunnels, compensation
grouting, and remedial works to the concourse tunnel invert. In less
favourable ground conditions, fewer of these factors could have caused
earlier and more sudden failure.
1.7 TUNNEL LINING STRESSES
It is difficult to measure tunnel lining stresses, and so there are not many
case studies available compared to surface settlements or even subsurface
ground movements.
Firstly, we will look at how tunnel lining stresses develop over time after
installation and as the tunnel continues to advance.
1.7.1 How do tunnel lining stresses

develop over time?
We would expect, based on the conceptual model we have developed so
far in this chapter, that the stress should increase after installation of the
tunnel lining, as the tunnel advances, until a stable value is obtained some
distance behind the face when a short-term equilibrium has been achieved.
Usually, face pressures and grouting pressures applied during construction
are below the in situ stress; therefore, we would expect the stable value to
be lower than the in situ stress in the ground. In effect, the in situ stress that
was present before construction has been shared between the lining and
arching in the ground around the tunnel.
How much of the ground load is taken by the lining depends on the tim-
ing of lining installation, the axial and flexural stiffness of the lining, and
the soil behaviour. In soft ground, we would expect the lining to be signifi-
cantly stiffer than the ground, and so it should attract more of the load. In
hard rock, we would almost always expect the rock to be stiffer than the
lining such that the rock itself is the main supporting element in the system.
In the long term, equilibrium of the lining-ground system may change
due to a variety of factors, such as:
• the lining may creep and/or shrink

• the lining may deteriorate, perhaps due to chemical attack, corrosion,
age or damage
• there will be temperature changes within the tunnel that cause the
lining to expand or contract, increasing or decreasing the stress in
the lining
• surface excavation, e.g. for a basement, may reduce the vertical
ground load on the tunnel and reduce the lining stress, and may cause
distortion of the tunnel
• surface loading, e.g. due to building construction, may have a similar
but opposite effect on the lining stress
• nearby tunnel excavation may change the stress state in the ground
and hence the lining stresses will change
• the pore pressures in the ground may change due to long-term changes
to the water table or flow regime, changing in turn the water pressure
applied to the lining
• in undrained soils, dissipation of excess pore pressures will cause

consolidation or swelling of the soil, which may change the stresses
applied to the lining
A comparison is made in Figures 1.23–1.25 of previous stress measure-

ments in London Clay by Skempton (1943), Cooling & Ward (1953), Ward
& Thomas (1965), Muir Wood (1969), Barratt et al. (1994) and Bowers
& Redgers (1996). These were all made using load cells or strain gauges
in precast concrete and cast iron segmental linings. These data could have
Figure 1.23 Tunnel lining stress measurements in London Clay, up to 50 days

(Jones, 2007).
Figure 1.24 Tunnel lining stress measurements in London Clay, up to 600 days

(Jones, 2007).
been presented in a single chart with log time on the horizontal axis, but
this would give a distorted perception of how the stresses change in the
long term.
Figure 1.23 shows that in the short term, lining stresses quickly increase
and a relatively stable equilibrium is found within a few days. By this time,
the face of the tunnel has probably advanced sufficiently far ahead that it
is no longer affecting the lining-ground system at the monitoring location.
Figures 1.24 and 1.25 show that in all cases, it then takes much longer to
achieve a proper equilibrium, and this is most likely due to long-term pore
pressure and temperature changes (Jones, 2007).
Figure 1.25 Tunnel lining stress measurements in London Clay, up to 19.5 years

(Jones, 2007).
For segmental linings, the maximum load has always been found to
occur in the long-term. Skempton (1943) found the maximum load to be
approximately equal to that corresponding to the hydrostatic full overbur-
den pressure (that is, the initial in situ stress with K0 = 1.0). Ward & Thomas
(1965) found that one of the tunnels they studied reached full overbur-
den pressure, while the second one did not but was continuing to increase
when measurements ceased. They therefore arrived at the conclusion that
hydrostatic full overburden pressure would eventually act on the lining in
the long-term. Since then, measurements by Muir Wood (1969), Barratt
Table 1.2 M
aximum stresses in tunnels in London Clay, expressed as a percentage of
the lining stress if full overburden pressure were applied (Jones, 2007).
Publication Tunnel Maximum load (% overburden)
Skempton (1943) Unknown, London 102–108%
Cooling & Ward (1953) 9′ diameter water 53–64%
tunnel, London
Ward & Thomas (1965) ‘Site O’, Victoria Line, 71% (not stabilised)
at 3.5 years
Ward & Thomas (1965) ‘Site V’, Victoria Line, 105%
at 6.5 years
Muir Wood (1969) Heathrow Cargo 60–80%
Tunnel, Heathrow
Barratt et al. (1994) Jubilee Line, Regent’s 40–64%
Park, at 19.5 years
Bowers & Redgers (1996) Jubilee Line Extension, 43–62%
St James’s Park, at
4 months
et al. (1994), Bowers & Redgers (1996) and Jones (2007) have all shown
that load can stabilise at a value well below that corresponding to full over-
burden pressure. Values of maximum load as a percentage of hydrostatic
full overburden pressure are shown in Table 1.2.
It is possible that the amount of stress in the tunnel lining is related to the
amount of ground deformation that was allowed to occur during construc-
tion, which may explain the higher loads in the older tunnels that probably
caused larger ground deformations. The exception to this rule would be the
Heathrow Cargo tunnel, which was constructed with an unusually high
degree of face support at shallow cover and had a volume loss of only 0.2%
(this low value of ‘volume loss’ means that ground movements would have
been very small – approximately an order of magnitude smaller than the
other tunnels presented in Table 1.2).
Therefore, it seems reasonable to assume that so long as the tunnelling
process is well-controlled and the ground movements are minimised, then
in the long-term the tunnel lining will be supporting 40–70% of the full
overburden load in London Clay. If, however, the ground movements are
large and the tunnel lining is relatively permeable, there is a potential for
up to 100% of full overburden to act on the tunnel lining in the long-term.
In other types of soil there are very few data, but long-term lining stresses
in all kinds of soft ground tend to be in the range of 25–100% of full over-
burden. Some case studies are summarised in Table 1.3.
Suzuki et al.’s (1996) two tunnels were of similar construction and were
in similar geology, but they were at different depths, and different grout-
ing pressures were applied. At the Yodo River, grouting pressures were
approximately 50% of overburden pressure, whereas at the Shigino Route,
they were 75–105%. So at Shigino Route, where higher grouting pressures
Table 1.3 Case studies of lining stress measurements not in London Clay.
Publication Tunnel Maximum load (% overburden)
Belshaw & Palmer Thunder Bay trunk sewer, 25–57%
(1978) Ontario, Canada, 2.38 m
OD TBM tunnel, in soft
to firm clay, at 1 year
Palmer & Belshaw Thunder Bay (as above but 24–61%
(1980) second array of
instrumentation), in soft
to firm clay, at 1 year
Suzuki et al. (1996) Yodo River cable tunnel, 40–82%
Osaka, Japan, 3.7 m OD
TBM tunnel, in clay/silt/
sand/gravel, at 10 months
Shigino Route cable tunnel, 35–60%
Osaka, Japan, 3.55 m OD
TBM tunnel, in clay/silt/
sand/gravel, at 10 months
Sakurai & Izunami Kobe Municipal Subway, 25% average
(1988) Japan, approx. 10 m wide ×
8 m high shotcrete tunnel
with rockbolts, in hard clay
and dense gravel layers,
time of measurement
unknown
were applied and ground deformation was minimised, the eventual ground
pressure acting on the tunnel was lower. This is perhaps counterintuitive, but
if there is any pattern in lining stress measurements in soft ground, it seems
to be that the smaller the ground deformations, the lower the long-term lin-
ing stress. This runs counter to what we expect from simple models so it is
important to remember this as you work through later chapters of this book.
1.7.2 Design based on precedent practice

If we had more measurements of lining stress, they could be very useful
for dimensioning the lining at a preliminary design stage and for calibra-
tion of numerical models during detailed design. However, even in London
Clay, which is probably the most measured and studied geomaterial, there
is a large amount of variation in the measured stresses. This is partly
because of the different construction methods employed but is also due to
other factors. For instance, none of the published work presented so far
in this chapter takes account of temperature, but it has been shown that
temperature variations in the tunnel will not only affect the instrumenta-
tion but will also cause changes in the tunnel-ground system equilibrium
(Jones, 2007). As the temperature in the tunnel increases, the lining will
expand, and this will increase the ground pressure acting on it. Similarly,
as the temperature in the tunnel decreases, the lining will contract away
from the ground, reducing the ground pressure. This factor was identified
and defined as ‘ground reaction temperature sensitivity’ in my thesis (Jones,
2007). It is likely that the long-term increases in lining stress in the London
Underground tunnels measured by Ward & Thomas (1965) and Barratt
et al. (1994) were caused, at least in part, by increasing tunnel temperatures
over those years.
Usually, it is not the average load that is important to lining design, but
how much the ground load varies around the tunnel lining. This is because
in relatively shallow soft ground tunnels, we are not worried about hoop
(axial) loads in the lining, which are usually well below the axial capacity,
but bending moments. In London Clay, vertical ground stresses after con-
struction (as measured in the lining at tunnel axis level in Figures 1.23–1.25)
have always been found to be higher than the horizontal ground stresses
(as measured at the tunnel crown or invert in Figures 1.23–1.25). This is
despite the fact that in situ horizontal stress is higher than the in situ verti-
cal stress prior to construction. However, this does agree with the lining
displacements that are typically observed, which were presented in Section
1.6, where tunnel linings in soft ground invariably squat.
Another aspect that is particularly important for shotcrete lining
design is the rate at which the lining stresses increase as the face con-
tinues to advance. This is because the shotcrete is gaining strength with
time and we need to know that it has sufficient strength at early age to
withstand these stresses. The published work we have looked at so far has
not provided much detail in terms of relating lining stress to the position
of the face. The following unique case study from the Heathrow Express
Terminal 4 Station concourse tunnel will demonstrate the kind of detailed
information that can be obtained with carefully installed and interpreted
instrumentation.
1.7.3 Heathrow Express Terminal 4 Station

concourse tunnel case study
The Heathrow Terminal 4 station concourse tunnel was constructed after
the Heathrow Express collapse at the Central Terminal Area (HSE, 2000).
A lot of instrumentation was installed to verify the design, some of which
was presented earlier in this chapter in Figures 1.2 and 1.20. More details
can be found in Clayton et al. (2002), van der Berg et al. (2003), Clayton
et al. (2006) and Jones (2007). In this section, we will focus on the radial
pressures applied by the ground onto the lining, as measured by radial pres-
sure cells installed before spraying the shotcrete lining. These allowed the
ground pressure to be measured from lining installation onwards.
The layout of the Heathrow Express Terminal 4 station is shown in
Figure 1.26. It consists of two platform tunnels with a central concourse
Figure 1.26 Plan of tunnels at Heathrow Express Terminal 4 station, showing location

of concourse tunnel and layout of monitoring points and instruments (from
van der Berg et al., 2003).
tunnel at the northeastern end. These three tunnels are connected by a

series of cross-passages and to the north and south ventilation tunnels at
either end, which were constructed last.
The platform tunnels are over 220 m long with a cross-sectional area
of 62 m 2 , and the concourse tunnel is 64 m long with a cross-sectional
area of 49 m 2 . A cross-section of the concourse and platform tunnels is
shown in Figure 1.27, which also shows the surface level and geological
strata. The concourse tunnel axis is at a depth of 17.2 m below ground
Figure 1.27 Cross-section of concourse and platform tunnels.
level and the tunnel is entirely within the London Clay. Piezometers
across the site and at different depths indicated a piezometric level in
the Terrace Gravels at approximately ground level with a hydrostatic
distribution from there down to the basal beds of the London Clay,
well below the tunnel horizon (van der Berg et al., 2003). The cen-
treline spacing between the concourse tunnel and the platform tunnels
is 13.5 m.
The platform tunnels were constructed first. Construction of the con-
course tunnel commenced in September 1996 after completion of the adja-
cent sections of the permanent secondary lining in the upline and downline
platform tunnels. The concourse tunnel headwall was completed on
7th November 1996.
The construction sequence for the concourse tunnel was top heading,
bench, top heading, bench, double-invert, schematically illustrated in
Figure 1.28. The invert was closed five rounds from the face. The advance
length varied from 0.8 m to 1.2 m depending on ground conditions and
design requirements, including the proximity of sensitive structures.
The primary support consisted of 350 mm of sprayed concrete (shot-
crete), reinforced with two layers of welded wire mesh (8 mm diameter at
150 mm centres) and full-section lattice girders ‘Type 110 ROM E3’. The
exposed ground was supported by a 50–100 mm shotcrete sealing layer
applied immediately after each advance. The 350 mm total thickness
included the sealing layer.
We will focus on the pressure cells installed in Main Monitoring Section
VIII (MMS VIII) of the concourse tunnel. The location of MMS VIII is
shown in Figure 1.29 and in the location plan (Figure 1.26).
Figure 1.28 Concourse tunnel construction sequence (from van der Berg et al., 2003).
At each section, 12 tangential pressure cells and 12 radial pressure cells

were installed. The locations are shown in Figure 1.30.
Table 1.4 lists construction events that may have affected stress measure-
ments at MMS I and MMS VIII.
A selection of the recorded radial pressures for MMS VIII is shown in
Figure 1.31. The positions of the radial pressure cells are marked by radial
lines normal to the extrados of the lining. The outer perimeter represents
the in situ stress normal to the lining calculated at the positions of the
radial pressure cells, based on a bulk unit weight for the Made Ground,
Terrace Gravel and London Clay of 19.5 kN/m3 and a coefficient of earth
pressure at rest (K0) of 1.5 (Powell et al., 1997).
Figure 1.29 Long section of the concourse tunnel showing locations of MMS I and
MMS VIII.
Figure 1.30 Main monitoring section schematically showing locations of pressure

cells and strain gauges embedded in the sprayed concrete primary lining.
Table 1.4 Construction events during concourse tunnel construction.

26th October 1996 12:00 Top heading excavated
26th October 1996 12:00 Top heading primary lining sprayed at positions 1–3
27th October 1996 05:30 Top heading primary lining sprayed at positions 4 and 5
27th October 1996 17:00 Bench excavated
28th October 1996 19:00 Bench primary lining sprayed
28th October 1996 15:30 Invert excavated and primary lining sprayed
7th–14th November 1996 Crosspassage construction (c.f. Figure 1.29)
18th November 1996 Compensation grouting
12th December 1996 (approx.) Invert secondary lining cast
6th January 1997 Full round sheet waterproofing membrane installed,
followed by casting of the secondary lining
approximately 6 weeks later
Figure 1.31 MMS VIII radial stresses (L-R, top-bottom: top heading and bench, invert,
6–19 days, 20–47 days, 72 days to 2 years, 3.1 years to 18.6 years).
The first diagram in Figure 1.31 shows the initial readings of the top
heading radial pressure cells. At the time of spraying the bench, 13.5 hours
after the top heading was sprayed, the average radial pressure on the top
heading was 23% of the in situ radial stress.
The second diagram in Figure 1.31 shows that upon invert excavation
the top heading radial pressures increased from 23% to 31% of in situ
radial stress and the bench radial pressures had climbed to 16%. As the
invert sprayed concrete became stiffer and gained strength, the lining began
to act as a complete ring and pressure built up gradually at all positions
except PCR1. A distinctive radial pressure distribution began to develop at
the invert, with higher pressures measured at PCR10 and 11 compared to
PCR12. This was also evident in the MMS I radial pressure cells and can
be attributed to the non-circular shape of the tunnel lining; at position 12,
the lower curvature makes the structural response more flexible relative to
the high curvature at positions 10 and 11. This means that there is more
unloading of the London Clay, and hence lower radial pressure in the short-
term at position 12, whereas the stiffer structural response at positions 10
and 11 will tend to attract more radial pressure in the short term. In the
long term, dissipation of negative excess pore pressures and the associated
swelling of the clay at position 12 in the centre of the invert will tend to
even out the radial pressure distribution. At 31/10/96 11:00, almost 5 days
after top heading excavation and nearly 3 days after invert closure, the
average radial pressure was 38% of the in situ radial stress or 43% of the
full overburden pressure.
The next diagram in Figure 1.31 shows very little change over the fol-
lowing week, except at PCR8 and 10 at 06/11/96 11:00, where there is a
noticeable increase in pressure. Although the dates given for cross-passage
construction from the construction records were 07/11/96 to 14/11/96, it
may be that some breaking out of the concourse tunnel lining occurred on
06/11/96, which would explain the change in pressure. The cross-passage
construction caused a noticeable increase in the radial pressures at the sides
of the tunnel but had much less effect near the crown and invert of the tun-
nel, as would be expected. By 14/11/96 12:45 the radial pressure changes
had stabilised, with the overall effect of the cross-passage construction
being an increase in average horizontal ground pressure from 45% of the in
situ radial stress at 04/11/96 10:00 to 53% at 14/11/96 12:45.
The fourth diagram in Figure 1.31 shows that from 20 to 47 days, ground
pressures had stabilised with only small changes evident. Compensation
grouting on the 18/11/96 did not have any effect on the radial pressures.
Casting the secondary lining invert on the 12/12/96 caused a significant
increase at PCR12.
From 72 days to 18.6 years (last two diagrams in Figure 1.31), there were
only gradual changes in radial pressure, tending to increase and even out
the pressure distribution. The higher level of unloading of the ground in
the area around PCR12 evident soon after closure of the invert resulted in
negative excess pore pressures, which over time dissipated due to ground-
water flow, leading to gradually increasing radial pressure due to swelling
of the London Clay.
In conclusion, although the ground this tunnel was excavated through
would already have been disturbed by construction of the parallel platform
tunnels and therefore this was not really a ‘greenfield’ situation, the trend
in the long term showed very little increase in load after the construction
period was over. This may have been due to the control of ground deforma-
tion during construction, reducing the magnitude and extent of excess pore
pressure, or due to the waterproof membrane making the tunnel imperme-
able so it did not act as a drain on the ground around it. Or indeed both of
these effects may have been important.
1.8 SUMMARY
Using case studies, this chapter has described the evolution of ground
movements, pore pressures, effective stresses and lining stresses as a tunnel
advances, and what happens to them in the long-term. Drained and und-
rained behaviour has been explained and the factors affecting stability have
been described. You should now have a conceptual model in your head of
ground and tunnel lining behaviour and how they change with face posi-
tion and time.
REFERENCES
Anagnostou, G. & Kovári, K. (1996). Face stability conditions with earth-

pressure-balanced shields. Tunn. Undergr. Space Technol. 11, No. 2,
165–173.
Attewell, P. B. & Woodman, J. P. (1982). Predicting the dynamics of ground settle-
ment and its derivatives caused by tunnelling in soil. Ground Eng. 15, No. 8,
pp. 13–18, 22, 36.
Attewell, P. B., Yeates, J. & Selby, A. R. (1986). Soil movements induced by tunnel-
ling and their effects on pipelines and structures. Glasgow: Blackie.
Avgerinos, V., Potts, D. M. & Standing, J. R. (2016). The use of kinematic harden-
ing models for predicting tunnelling-induced ground movements in London
Clay. Géotechnique 66, No. 2, 106–120.
Barratt, D. A., O’Reilly, M. P. & Temporal, J. (1994). Long-term measurements
of loads on tunnel linings in overconsolidated clay. Proc. Tunnelling ’94,
pp. 469–481. London: IMM.
Belshaw, D. J. & Palmer, J. H. L. (1978). Results of a program of instrumentation
involving a precast segmented concrete-lined tunnel in clay. Can. Geotech. J.
15, 573–583.
Biggart, A. R. & Sternath, R. (1996). Storebaelt eastern railway tunnel: construc-
tion. Proc. Inst. Civ. Engrs – Civ. Engrg 114, Special issue 1, 20–39.
Bock, H. (2003). Geotechnical instrumentation of tunnels. Summerschool on

rational tunnelling (ed Kolymbas, D.), Innsbruck, Austria, pp. 187–224.
Berlin: Logos Verlag.
Bowers, K. H. & Redgers, J. D. (1996). Discussion: observations of lining load
in a London Clay tunnel. Proc. Int. Symp. on Geotechnical Aspects of
Underground Construction in Soft Ground (eds Mair, R. J. & Taylor, R.
N.), London, UK. Rotterdam: Balkema.
Brown, D. A. (2004). Hull wastewater flow transfer tunnel: recovery of tunnel col-
lapse by ground freezing. Proc. Inst. Civ. Engrs – Geotech. Engrg 157, April,
issue GE2, 77–83.
Chambon, J. F. & Corté, J. F. (1994). Shallow tunnels in cohesionless soil: stability
of tunnel face. J. Geotech. Eng. ASCE 120, No. 7, 1150–1163.
Clayton, C. R. I., Thomas, A. H. & van der Berg, J. P. (2003). SCL tunnel
design in soft ground – insights from monitoring and numerical modelling.
Summerschool on rational tunnelling (ed Kolymbas, D.), Innsbruck, Austria,
pp. 61–92. Berlin: Logos Verlag.
Clayton, C. R. I., van der Berg, J. P., Heymann, G., Bica, A. V. D. & Hope, V. S.
(2002). The performance of pressure cells for sprayed concrete tunnel linings.
Géotechnique 52, No. 2, 107–115.
Clayton, C. R. I., van der Berg, J. P. & Thomas, A. H. (2006). Monitoring and dis-
placements at Heathrow Express Terminal 4 station tunnels. Géotechnique
56, No. 5, 323–334.
Construction Today (1994). Police probe repeat Munich tunnel breach.
Construction Today, October issue, pp. 4–5.
Cooling, L. F. & Ward, W. H. (1953). Measurements of loads and strains in earth
supporting structures. Proc. 3rd ICSMFE, Session 7/3, Vol. 2, pp. 162–166.
Craig, R. (1975). Discussion at meeting of British Tunnelling Society. Tunnels and
Tunnelling 7, 61–65.
Deane, A. P. & Bassett, R. H. (1995). The Heathrow Express trial tunnel. Proc.
Inst. Civ. Engrs – Civ. Engrg 113, July, 144–156.
Grammatikopoulou, A., Zdravkovic, L. & Potts, D. M. (2008). The influence of
previous stress history and stress path direction on the surface settlement
trough induced by tunnelling. Géotechnique 58, No. 4, 269–281.
Grose, B. & Benton, L. (2005). Hull wastewater flow transfer tunnel: tunnel col-
lapse and causation investigation. Proc. Inst. Civ. Engrs – Geotech. Engrg
158, October, Issue GE4, 179–185.
Grose, B. & Benton, L. (2006). Discussion: Hull wastewater flow transfer tunnel:
tunnel collapse and causation investigation. Proc. Inst. Civ. Engrs – Geotech.
Engrg 159, April, Issue GE2, 125–128.
HSE (1996). Safety of New Austrian Tunnelling Method (NATM) tunnels, Health
and Safety Executive, p. 86. London: HSE Books, HMSO.
HSE (2000). The collapse of NATM tunnels at Heathrow Airport – A report on
the investigation by the Health and Safety Executive into the collapse of
New Austrian Tunnelling Method (NATM) tunnels at the Central Terminal
Area (CTA) of Heathrow Airport on 20/21 October 1994. Norwich: Her
Majesty’s Stationery Office/HSE Books.
Jones, B. D. (2007). Stresses in sprayed concrete tunnel junctions. EngD thesis,
University of Southampton.
Lipscombe, R., Carter, C., Perkins, O., Thurlow, P. & Guerrero, S. (2015). The
use of Shape Accel Arrays (SAA) for measuring retaining wall deflection.
Crossrail Project – Infrastructure design and construction, Vol. 1. London:
ICE Publishing. Also available at: https://learninglegacy.crossrail.co.uk/
documents/use-shape-accel-arrays-saa-measuring-retaining-wall-deflection/
[last accessed: 23rd June 2020].
Mair, R. J. (1979). Centrifugal modelling of tunnel construction in soft clay. PhD
thesis, University of Cambridge.
Mair, R. J. & Taylor, R. N. (1997). Bored tunnelling in the urban environment.
Theme Lecture, Plenary Session 4. Proc. 14th Int. Conf. Soil Mechanics and
Foundation Engineering, Hamburg, Vol. 4, pp. 2353–2385.
Marshall, A. M. (2009). Tunnelling in sand and its effect on pipelines and piles.
PhD thesis, University of Cambridge.
Marshall, A. M., Farrell, R., Klar, A. & Mair, R. (2012). Tunnels in sands: the effect
of size, depth and volume loss on greenfield displacements. Géotechnique 62,
No. 5, 385–399.
Messerli, J., Pimentel, E. & Anagnostou, G. (2010). Experimental study into tun-
nel face collapse in sand. Physical Modelling in Geotechnics (eds Springman,
S., Laue, J. & Seward, L.), pp. 575–580. London: Taylor & Francis.
Muir Wood, A. M. (1969). Written contribution, plenary session 4. Proc. 7th
Int. Conf. Soil Mechanics and Foundation Engineering, Vol. 3, Mexico,
pp. 363–365.
Muir Wood, A. M. (2000). Tunnelling: management by design. London: E & FN
Spon.
New, B. M. & Bowers, K. H. (1994). Ground movement model validation at the
Heathrow Express trial tunnel. Tunnelling ’94, Proc. 7th Int. Symp. IMM
and BTS, London, UK, pp. 310–329. London: Chapman and Hall.
New Civil Engineer (1998). Docklands Light Railway Lewisham Extension blow-
out. February 1998.
Nyren, R. J., Standing, J. R. & Burland, J. B. (2001). Surface displacements at
St James’s Park greenfield reference site above twin tunnels through the London
Clay. Building Response to Tunnelling: Case Studies from Construction of
the Jubilee Line Extension, London – Vol. 2: Case Studies (eds Burland, J.
B., Standing, J. R. & Jardine, F. M.), CIRIA Special Publication 200, Chapter
25, pp. 387–400. London: CIRIA.
Palmer, J. H. L. & Belshaw, D. J. (1980). Deformations and pore pressures in the
vicinity of a precast, segmented, concrete-lined tunnel in clay. Can. Geot. J.
17, 174–184.
Powell, D. B., Sigl, O. & Beveridge, J. P. (1997). Heathrow Express – design
and performance of platform tunnels at Terminal 4. Proc. Tunnelling ’97,
London, pp. 565–593. London: IMM.
Sakurai, S. & Izunami, R. (1988). Field measurements of the Kobe Municipal
Subway Tunnel excavated in soil ground by NATM. Proc. 2nd Int. Symp.
on Field Measurements in Geomechanics (ed. Sakurai, S.), Kobe, Japan,
6th–9th April 1987, pp. 861–869. Rotterdam: Balkema.
Shin, J. H., Addenbrooke, T. I. & Potts, D. M. (2002). A numerical study of the effect
of groundwater movement on long-term tunnel behaviour. Géotechnique 52,
No. 6, 391–403.
Skempton, A. W. (1943). Discussion: Tunnel linings with special reference to a new

form of reinforced concrete lining, by G. L. Groves. J. Inst. Civ. Eng. 20,
No. 5, March, 53–56.
Standing, J. R., Nyren, R. J., Longworth, T. I. & Burland, J. B. (1996). The
measurement of ground movements due to tunnelling at two control sites
along the Jubilee Line Extension. Proc. Int. Symp. Geotechnical Aspects of
Underground Construction in Soft Ground (eds Mair, R. J. & Taylor, R. N.),
London, pp. 659–664. Rotterdam: Balkema.
Suzuki, M., Kamada, T., Nakagawa, H., Hashimoto, T. & Satsukawa, Y. (1996).
Measurement of earth and water pressures acting on the great depth shield seg-
ments. Proc. Int. Symp. Geotechnical Aspects of Underground Construction
in Soft Ground (eds Mair, R. J. & Taylor, R. N.), London, pp. 613–619.
Rotterdam: Balkema.
Terzaghi, K. (1936). The shearing resistance of saturated soils. Proc. 1st Int. Conf.
Soil Mechanics, Vol. 1, pp. 54–56.
Thomas, A. H. (2003). Numerical modelling of sprayed concrete lined (SCL) tun-
nels. Ph.D. thesis, University of Southampton.
Thomas, T. (2011). Silt inflow stops Preston EPBM. Tunnelling Journal, March
issue, p. 6. Also available at: http://tunnellingjournal.com/news/silt-stops-
epbm-in-preston/?doing_wp_cron [last accessed 27th March 2017].
van der Berg, J. P., Clayton, C. R. I. & Powell, D. B. (2003). Displacements ahead
of an advancing NATM tunnel in the London Clay. Géotechnique 53, No.
9, 767–784.
van Jaarsveld, E. P., Plekkenpol, J. W. & Messemaeckers van de Graaf, C. A.
(2006). Ground deformations due to the boring of the Second Heinenoord
Tunnel. Tunnelling. A Decade of Progress – GeoDelft 1995–2005, pp. 11–17.
London: Taylor & Francis Group.
Wan, M. S. P., Standing, J. R., Potts, D. M. & Burland, J. B. (2017). Measured
short-term ground surface response to EPBM tunnelling in London Clay.
Wan, M. S. P., Standing, J. R., Potts, D. M. & Burland, J. B. (2019). Pore water
pressure and total horizontal stress response to EPBM tunnelling in London
Clay. Géotechnique 69, No. 5, 434–457.
Ward, W. H. & Thomas, H. S. H. (1965). The development of earth loading and
deformation in tunnel linings in London Clay. Proc. 6th ICSMFE, Vol. 2,
Divisions 3–6, Montreal, Canada, 8th–15th September, pp. 432–436.
Wright, P. J. (2013). Validation of soil parameters for deep tube tunnel assessment.
Proc. Inst. Civ. Engrs – Geotech. Engrg 166, GE1, February, 18–30.
Chapter 2
Undrained stability
As explained in Section 1.5, drained and undrained stability are quite differ-
ent. The geometry of failure is different, and the calculations are done in a
different way. Therefore, they have been split into two separate chapters – this
chapter covers undrained stability and Chapter 3 will cover drained stability.
A rule of thumb is that a soil with a permeability less than 10 −7 to 10 −8 m/s
will behave in an undrained manner during the timescale of a typical tunnel
construction, and a soil with higher permeability will behave in a drained
manner (Anagnostou & Kovári, 1996). If it is unclear whether a soil will
behave in a drained or undrained manner, stability calculations need to be
done for both cases and the worst case used for design.
It is important to remember that real soils are variable, heterogeneous
and rarely isotropic, and that ground mass behaviour can be different to the
behaviour of small samples in geotechnical laboratory tests. Many of the
calculations presented here are simple models and will assume the ground
is homogeneous and isotropic and follows simple behavioural rules, so it
is important to think about how the true ground behaviour may affect the
potential failure mechanisms. For instance, a permeable layer of soil within
an open face with groundwater pressure present could cause an uncon-
trolled failure, even if 90% of the ground is clay.
After working through this chapter you will understand:
• how tunnel headings collapse or blow-out

• how upper and lower bound limit states, limit equilibrium, numeri-
cal modelling and empirical data help us to estimate the true collapse
load
• undrained stability ratio and critical stability number
After working through this chapter, you will be able to:
• calculate the factor of safety for undrained stability of a heading in a

variety of situations
• calculate blow-out pressures in clay for hydraulic fracturing and pas-
sive failure
DOI: 10.1201/9780429470387-2 45
2.1 OVERVIEW OF STABILITY THEORY
Stability is an ultimate limit state. This means that we are designing and
planning the construction to avoid a catastrophic failure. We want to try to
predict when or how that failure will occur, and then ensure that we have
a factor of safety that makes it very unlikely.
Since at failure the strength of the ground will be fully mobilised, head-
ing stability lends itself well to plasticity solutions. A plasticity solution
ignores what happens before failure and looks only at the fully mobilised
failure geometry to calculate the load required to get there.
A heading may be geometrically simplified as shown in Figure 2.1, where:
D is the excavated diameter in m

C is the cover from the crown of the excavation to the ground surface
in m
P is the unsupported length in m
In the case of a closed-face tunnel boring machine (TBM), P may be equal

to zero. For a sequentially excavated tunnel lined with shotcrete, some
assumptions about the values of P, C and D may need to be made.
For non-circular tunnels, Pound (2005) used numerical analysis to show
that even for elliptical or rectangular tunnels, with a width three times the
height or a height three times the width, or for semi-circular top headings,
stability may be approximated by an equivalent circular tunnel with the
same face area, where the cover C is taken as the distance from the actual
crown to the surface and the depth to axis is taken as the depth to the cen-
troid of the face. Errors were no more than 1.5%. So the precise geometry
is not important, as long as a value of D is chosen that gives the same face
area. Pound’s results were for undrained cohesive soils, and as we will see
Figure 2.1 Simplified geometry of a tunnel heading.

Undrained stability 47
they may not apply to drained non-cohesive soils – no one has done that
research yet.
There are several ways we can determine stability:
1. Assuming a ‘kinematically admissible mechanism’, such that if a

structure is loaded to this value it must collapse. This usually involves
assuming the ground is made up of several large blocks that slide into
the face. It demonstrates that failure must occur at this load, but there
may be situations (e.g. other geometries or configurations of blocks)
where failure may occur at a lower load. For this reason this is also
known as an ‘upper bound solution’.
2. Assuming a ‘statically admissible stress field’, such that if a structure
is loaded to this value it cannot collapse. This is done by determining
a set of stresses in the ground that are in equilibrium with the external
loads and do not exceed the strength of the ground (Atkinson, 2007),
thus demonstrating that failure cannot occur in this set of circum-
stances. The true failure load must therefore be higher, so this may be
overconservative. This is known as a ‘lower bound solution’.
3. These upper and lower bounds bracket the true collapse load.
Combining them to find a solution that is kinematically and stati-
cally admissible is known as a ‘limit equilibrium solution’. However, a
limit equilibrium solution may not represent the true collapse load, as
it relies on assumptions of geometry of failure, ignores stress–strain
behaviour and ignores compatibility of strains.
4. Empirical data from heading stability failures in the field and in cen-
trifuge tests can be used to develop relationships that may help predict
the true collapse load.
5. Numerical modelling.
2.2 UNDRAINED STABILITY
The study of undrained stability began because of a collapse near Stockholm

in 1964, recounted by Broms & Bennermark (1967):
On Friday, November 20, 1964, a slide took place at Edsådalen, near

Stockholm. The slide occurred in a soft clay when a 6.5 ft diameter hole
was cut at the base of a vertical sheet pile wall that supported a 35 ft
deep excavation. The slide took place approximately 1-1/2 hr after the
clay behind the sheet pile wall had been exposed. The clay surface had
been inspected approximately 1 hr before the slide occurred, revealing
that the exposed surface was “hard and dry”. The first indication that
failure was imminent was the observation (made a few seconds before
the actual slide by one of the three men present in the excavation) that
the exposed clay surface started to move. Before any warning could be
given, all three men were buried under about 1,500 cu ft of clay, and
one of the men lost his life.
In order to try to predict failures of this kind and prevent similar disas-
ters, Broms and Bennermark used laboratory tests where clay was extruded
under pressure through vertical circular openings, and also field observa-
tions of both stable and collapsed openings, to try to characterise stability,
and they published the results in a seminal paper in 1967. They defined a
stability ratio N, which is the ratio of full overburden pressure at the axis
of the opening minus any support pressure divided by the undrained shear
strength of the soil cu, given by:
γ (C + D/2) + σ s − σ t
N= (2.1)
cu
N is the stability ratio and is dimensionless

γ is the bulk unit weight of the soil in kN/m3
C is the cover in m
D is the diameter of the opening as defined in Figure 2.1 in m
σ s is the effect at tunnel level of a surcharge at the surface (for instance
consisting of a stockpile of bulk materials, traffic loads, a flexible
raft foundation or a body of water) in kPa
σt is an internal face pressure provided by compressed air, slurry in the
head of a slurry TBM or earth pressure in an earth pressure bal-
ance (EPB) machine in kPa
cu is the undrained shear strength of the soil in kPa
For headings where the unsupported length P = 0, as defined in Figure 2.1,

Broms & Bennermark (1967) found that the face should be stable if N is
less than 6 and may collapse if N is greater than between 6 and 8, for C /D
ratios greater than 3.5.
The soft normally consolidated clays Broms and Bennermark tested
tended to generate positive excess pore pressures due to shearing of the soil
near the hole. This reduced the effective stress and hence the strength of
the soil, resulting in a short-term failure. As the pore water migrated away
from the hole, consolidation increased the strength of the clay somewhat,
perhaps giving the hard and dry surface appearance observed before the
failure at Edsådalen.
For overconsolidated clays, there is a tendency for negative excess pore
pressures to be generated during shearing, which increases the effective
stress in the short-term. As pore water moves towards these areas over time,
the clay will swell and the strength of the clay will decrease. Therefore, it
is likely that the strength of a clay, its overconsolidation ratio and its mass
permeability will have an effect on the true value of the stability ratio, and
will also determine the stand-up time.
Further examination of Equation 2.1, and remembering that a low value

of N is more stable and a higher value of N is less stable, can lead us to the
following conclusions:
• Increasing the depth of the tunnel will make the heading less stable.
• Increasing the surcharge pressure will make the heading less
stable.
• Decreasing the support pressure will make the heading less stable.
• Decreasing the undrained shear strength will make the heading less
stable.
2.2.1 Heading stability in homogeneous clay

Davis et al. (1980) published both upper and lower bound solutions for an
undrained soil. They are reasonably close together and so the true collapse
load may be determined with reasonable accuracy.
Mair (1979, cited in Kimura & Mair, 1981) used centrifuge tests and
case histories of tunnel heading failures to develop relationships to help
predict the true collapse load. The data lay between the upper and lower
bound solutions of Davis et al. (1980). The most common method used in
practice for undrained stability calculations are Mair’s design charts based
on centrifuge testing (Kimura & Mair, 1981).
Mair (1979) performed a suite of centrifuge tests of a tunnel heading in
clay at different C /D and P /D ratios, reported in Kimura & Mair (1981).
The design curves based on these centrifuge tests are shown in Figure 2.2,
as well as some of the centrifuge test data. P /D ratios between zero and
infinity were also tested and used to interpolate the design curves for P /D =
0.5, 1 and 2. These are not shown in Figure 2.2, because the values of P /D
did not correspond to any of the curves.
A ‘critical stability ratio’ Nc is defined, which is the value of stability ratio
at which failure occurs. Another way of saying this is if N > Nc, failure
occurs. The ratio Nc/N can also be thought of as the factor of safety for
heading stability.
Figure 2.2 shows that as C /D increases, i.e. as the depth of the tunnel
increases or the diameter decreases, the heading becomes more stable.
This is because there is more space for arching in the ground around the
tunnel as it gets deeper (drained stability, on the other hand, is indepen-
dent of depth, as we will see in the following chapter). But remember that
as the depth of the tunnel increases, the value of N in Equation 2.1 also
increases. Therefore, moving the tunnel deeper will not necessarily improve
stability – this will depend on the diameter of the tunnel and whether the
undrained shear strength increases with depth.
Another interesting attribute of this chart is that as P /D increases, the
heading becomes less stable. Therefore, reducing the unsupported length
will improve stability.
Figure 2.2 Design chart based on Mair’s (1979) centrifuge tests (from Kimura & Mair,
1981), and Davis et al. (1980) lower bound envelope for P /D = 0. Values for
Brook Green sewer collapse and Singapore tunnel collapse from Mair &
Taylor (1997).
Also shown on the chart is the lower-bound plasticity solution provided

by Davis et al. (1980). It has a kink in it because the critical stress field that
causes collapse is different below and above C /D = 0.86. If the stability
ratio is below the lower bound, then the heading cannot fail; therefore, it
is reassuring that the P /D = 0 line from Mair’s centrifuge tests lies above
it. The undrained plasticity solutions will not be presented in detail in this
book, as the centrifuge tests give a better estimate of the true collapse load.
WORKED EXAMPLE 2.1 OPEN-FACE TBM

TUNNEL IN HOMOGENEOUS CLAY
An open-face shield mounted with a roadheader is to be used to excavate
a 7.5 m diameter tunnel in stiff, overconsolidated clay that can safely
be assumed to behave in an undrained manner during the timescale of
construction. The roadheader can only reach a maximum of 1.5 m ahead
of the shield, the shield is 5 m long, the tailskin is 2 m long and grouting
is done through grout ports in the tailskin as the machine advances. The
tunnel axis is 18 m below the ground surface, the bulk unit weight of the
clay is 20 kN/m3 and the undrained shear strength is 100 kPa.
The geometry may be idealised as shown in Figure 2.3.
a. With no surcharge or internal tunnel pressure, calculate the factor

of safety on undrained stability.
b. If a surcharge were applied, what value would cause collapse?
c. With no surcharge applied, at what value of undrained shear
strength would collapse occur?
d. What is the factor of safety on undrained shear strength, i.e. the
undrained shear strength divided by the undrained shear strength
at which collapse would occur? Compare this to the value calcu-
lated in part a.
Figure 2.3 Worked Example 2.1 cross-section.
a. We first need to calculate the critical stability number Nc at

the ultimate limit state. We can assume that deformations are
large enough at the ultimate limit state to cause closure of the
ground around the shield and tailskin; therefore, the unsup-
ported length P is 1.5 m.
Using Mair’s design chart based on centrifuge tests with the fol-
lowing parameters:
P /D = 1.5/7.5 = 0.2
C /D = (18 − 7.5/2) /7.5 = 14.25/7.5 = 1.9

A line is drawn vertically on the design chart in Figure 2.4 at

C/D = 1.9. Then, the value of Nc at P/D = 0.2 may be found
by interpolating between the Nc values at P/D = 0 (7.8) and
P/D = 0.5 (6.8), giving Nc = 7.4.
If you are using a ruler to interpolate, watch the y-axis gridlines,
which go up in steps of 2.
Figure 2.4 Worked Example 2.1 using the design chart to find the critical stability
number.
Now the value of stability ratio N needs to be calculated. This

is given by:
γ (C + D / 2) + σ s − σ t 20 (18) + 0 − 0 360
N= = = = 3.6
cu 100 100
The factor of safety may be given by:
N c 7.4
= = 2.06
N 3.6
b. Since collapse occurs when Nc = N, we can use the stability ratio
equation to find the value of surcharge σ s that would cause collapse:
360 + σ s
N= = N c = 7.4
100
Rearranging to solve for σ s:
σ s = 7.4 × 100 − 360 = 380 kPa
c. The stability ratio equation can be used in a similar way to calcu-

late the value of undrained shear strength that would cause collapse
with no surcharge:
360
N= = N c = 7.4
cu
Therefore:
360
cu = = 48.6 kPa
7.4
d. The factor of safety on undrained shear strength (i.e. the design
value divided by the collapse value) is 100/48.6 = 2.06. The previ-
ously calculated factor of safety was exactly the same. Therefore
the factors of safety on actions and on undrained shear strength
are the same, which will become important later when we look at
Eurocode 7 design approaches.
WORKED EXAMPLE 2.2 SPRAYED CONCRETE

A tunnel with an 85 m 2 cross-sectional area is to be constructed in
clay in an urban area using a sprayed concrete lining, at a depth to the
centroid of the cross-sectional area of 15 m. A cross-section is shown in
Figure 2.5.
The ground is silty clay, with a permeability k = 1 × 10 −9 m/s. The
ground has a bulk unit weight of 19 kN/m3, an undrained shear strength
of 60 kPa, a drained cohesion of 5 kPa and a drained angle of friction of
22°. The water table is at the surface. Assume there is no surcharge.
a. Assuming the tunnel is excavated full-face in 1 m advances, calcu-

late the factor of safety on stability. Is this sufficient for Eurocode
7 or GEO Report 249?
b. Describe ways of improving the factor of safety by inspection of the
stability ratio equation and the critical stability ratio design chart.
Try some of them and see what happens.

a. The permeability indicates this is an undrained stability problem.
First calculate stability ratio N:
γ z0 + σ s − σ t
N=
cu
Now, σt = 0 because there is no support pressure and σ s = 0 because
the question says to assume there is no surcharge. So, N is given by:
γ z0 19 × 15 285
N= = = = 4.75
cu 60 60
We need to calculate an equivalent diameter D and calculate C:
4A 4 × 85
D= = = 10.40 m
π π
The cover value C should be the actual depth to the crown, not the
value calculated using the depth to axis and the equivalent radius.
For design, a cautious value of P should be used to take account of
possible overexcavation of the face. It is also common to dome the
face rather than leaving it flat. In this case we will assume P = 1.5 m.
Now calculate C /D and P /D :
C 9.8
= = 0.94
D 10.4
P 1.5
= = 0.14
D 10.4
These values are used in the chart in Figure 2.6. Note that the curve
for P /D = 0.5 does not plot below C /D = 1, but it is reasonable
to extrapolate such a small distance, especially as all the curves
appear to be converging on Nc = 2 at C /D = 0.
Figure 2.6 Worked Example 2.2 using the design chart to find the critical stabil-
ity number.
Reading from the chart, Nc = 5.28. Therefore the factor of safety
may be given by:
N c 5.28
= = 1.11
N 4.75
Eurocode 7 design approach 1 (EN 1997-1: 2004, 2009) would
require a factor of safety greater than 1.4. GEO Report 249 (Golder
Associates, 2009) recommends a factor of safety greater than 1.5.
b. In order to improve the factor of safety on stability, we can either

try to increase the critical stability number Nc or decrease the sta-
bility ratio N. By inspection of the stability ratio equation, several
possible strategies are available to minimise N:
γ z0 + σ s − σ t
N=
cu
• increase the undrained shear strength cu of the soil
• decrease the depth z0 of the tunnel
• decrease surcharge σ s
• increase internal support pressure σt
By looking at the critical stability number chart, we can identify

some more possible strategies to maximise the value of critical
stability number Nc:
• increase the cover C
• decrease the unsupported length P
• decrease the diameter D
Note that by decreasing the depth z0 of the tunnel, we can reduce

the stability ratio N in the equation (which is good), but this will
also decrease the cover C (which is bad – refer to the stability chart
x-axis, where a lower value of C will result in a lower value for
critical stability number Nc). Also, in most natural soils, undrained
shear strength will increase with depth, so reducing the cover will
also decrease the value of undrained shear strength cu used to cal-
culate the stability ratio N, which will result in a higher value of
N (again bad). Therefore, undrained stability is usually improved
by increasing the depth of the tunnel, at least for shallow tunnels
where C /D < 1. If undrained shear strength increases with depth,
even by a small amount, then the factor of safety is nearly always
improved by increasing the depth of the tunnel.
For example, if the tunnel were raised by 4.6 m, this would make
C = 5.2 m and C /D = 0.5. From the undrained stability chart
Nc = 3.80, which is a lower (worse) value than before. The depth
to axis is now z0 = 10.4, so assuming undrained shear strength is
the same as before (60 kPa), then stability ratio N = 3.29, which
is a lower (better) value than before. Now, the factor of safety
Nc /N = 3.80/3.29 = 1.15, which is slightly better overall than the
previous value of 1.11. However, one would expect the undrained
shear strength to be lower nearer to the surface, so on balance the
factor of safety will probably be worse. In fact, the value of und-
rained shear strength would only need to be less than 57.7 kPa for
the overall factor of safety to be worse.
Alternatively, if the tunnel were lowered 5.8 m making C = 15.6 m,
then C /D = 1.5, and from the stability chart Nc = 6.77. The depth
to axis is now z0 = 20.8 m, so assuming undrained shear strength
is constant, then the stability ratio N = 6.59. The factor of safety
given by Nc /N = 6.77/6.59 = 1.03, which is slightly worse than
before. However, if the undrained shear strength increased with
depth at a rate of 8 kPa/m (a typical value) at 5.8 m deeper it would

increase from 60 kPa to 106.4 kPa, and the stability ratio would now
be N = 3.71, making the factor of safety Nc/N = 6.77/3.71 = 1.82.
In a clay, ground treatment options to increase the undrained shear
strength are limited and probably only jet grouting will work, but in
some other types of soils compaction or permeation grouting or ground
freezing may improve stability by increasing the undrained shear
strength.
For a sprayed concrete tunnel in clay, the most common methods
of improving stability are to reduce the unsupported length or to
divide the face. For a tunnel of this size (85 m 2 , or 10.4 m equiva-
lent diameter), there would be logistical benefits to dividing the face
as well, for instance allowing the excavator to reach the crown by
constructing a top heading first. Although beyond the scope of this
book, designers need to understand the interaction between size
of plant, construction logistics and stability, and how to optimise
production while mitigating the risk of instability.
One possible method of dividing the face is shown in Figure 2.7:
Figure 2.7 Worked Example 2.2 – Cross-section showing possible method of face
division using a top heading–bench–invert sequence.
This is a top heading–bench–invert sequence. It is staggered such

that although the equivalent diameter is reduced, the unsupported
length P is increased, as shown in Figure 2.8.
It is often assumed that the face area is reduced to the size of
the partial heading. This is difficult to justify but often the reason
given is because the weight of the bench and invert yet to be exca-
vated provides a kind of berm to support the lower part of the face,
and because the failure mechanism, at least for shallow tunnels, is
soil coming down from above and in front, not from below.
Figure 2.8 Worked Example 2.2 – long section showing top heading–bench–

invert face division.
Ascertaining a value of unsupported length is not straightfor-

ward either. Usually, the maximum distance from the excavated
face to full ring closure is taken as the unsupported length, because
in soft ground effective support is only really achieved when the
ring is closed.
In Figure 2.8, the previous invert ‘0’ is excavated and about to be
sprayed, then stage 1 will begin. This represents the maximum unsup-
ported length, which for 1 m top heading and bench advances and a
2 m invert advance will be between 4 and 6 m, depending on whether
it can be assumed that the recently sprayed invert is strong enough
to provide effective support when top heading 1 is excavated, and
what allowance for overexcavation ahead of the leading edge of the
top heading is assumed. A reasonable assumption could be that the
unsupported length is 5 m. If a temporary invert is sprayed in the top
heading, we could assume the unsupported length is 1.5 m.
For this example, we will assume P is 1.5 m, and that the face
area of the top heading is 35 m 2 .
We need to calculate a new equivalent diameter D:
4A 4 × 35
D= = = 6.68 m
π π
The cover C has not changed.
The depth to axis z0 has reduced slightly and could be found by

assuming it is at the centre of the equivalent circle.
D 6.68
z0 = C + = 9.8 + = 13.14 m
2 2
Using P /D = 1.5/6.68 = 0.22, and C /D = 9.8/6.68 = 1.47, the criti-
cal stability ratio Nc from the design chart in Figure 2.2 is now
approximately 6.5. The decrease in D caused by subdividing the
face led to an increase in C /D and a significant increase in Nc.
Since z0 has changed, N is now given by:
γ z0 19 × 13.14 250
N= = = = 4.16
cu 60 60
Now, the factor of safety Nc /N = 6.5/4.16 = 1.56, which is a sig-

nificant improvement on 1.11 and would be considered sufficient
according to Eurocode 7 (EN 1997-1: 2004, 2009) and GEO
Report 249 (Golder Associates, 2009).
2.2.2 Heading stability in clay with undrained

shear strength increasing with depth
When calculating stability ratio N to compare with design charts for criti-
cal stability number Nc, a single value of undrained shear strength must be
used, but it is unclear what this value should be if, as is often the case, the
undrained shear strength is not constant, but increases with depth. Mair
& Taylor (1997) recommended using the value of undrained shear strength
at tunnel axis level, but this value may underestimate stability ratio N and
hence be unsafe. Atkinson & Mair (1981) recommended using the average
value of undrained shear strength between the surface and the tunnel axis,
but this may overestimate N and be too conservative.
Pound (2005) ran 3D numerical analyses with a linear increase of
undrained shear strength with depth, for P /D = 0, and compared the
results to models with a constant undrained shear strength. The varia-
tion of undrained shear strength with depth was defined by the follow-
ing equation:
cu = A ( B + z ) (2.2)
cu is the undrained shear strength in kPa

A and B are constants
z is the depth in m
Figure 2.9 Relative depth below ground level at which undrained shear strength
is equal to the critical value in constant undrained shear strength analysis
(from Pound, 2005).
To encompass the range of soil strengths from soft normally consolidated

clays to heavily overconsolidated clays, B was given values of 0, 2, 5 and 10.
A was reduced in the model until failure occurred.
Pound (2005) defined an equivalent depth zc at which the value of und-
rained shear strength was equal to the value in the constant strength anal-
ysis with the same C /D ratio. Figure 2.9 shows this equivalent depth zc
divided by depth to axis z0 plotted against C /D ratio.
Figure 2.9 shows that for lower values of B, i.e. when the value of
undrained shear strength at the ground surface was lower, the equivalent
depth was nearer the surface (zc /z0 had a lower value). For higher val-
ues of B, the equivalent depth was nearer the tunnel (zc /z0 had a higher
value). But for any particular value of C /D, the relative depth did not
vary that much.
It also shows that the relative depth increases for higher values of C /D,
meaning that failure is governed by a value of undrained shear strength
located closer to the tunnel face for deeper tunnels. This may be because
the failure mechanism is more localised at depth.
Overall, within a range of C /D from 0.5 to 6, the relative depth zc /z0 var-
ies from 0.675 to 0.825.
Müller (2015) found that for P /D = ∞, the equivalent undrained shear
strength at failure was about 0.8 times the value at axis level, when it
increases with depth using the equation cu = 30 + 4.5z. This would result in
zc /z0 values between approximately 0.6 and 0.7.
Therefore, contrary to the recommendation of Mair & Taylor (1997),
the value at tunnel axis level should not be used. Rather, for a linearly
increasing undrained shear strength, the value at 0.6–0.8 times the depth
to axis should be used as a first estimate. If you have a nonlinear undrained
shear strength variation with depth, or layered soils, you will need to use
conservative assumptions or numerical modelling.
2.2.3 Heading stability in clay with

overlying coarse-grained soils
A special case, but one that is very common, is when the tunnel is in clay,
but the overlying deposits consist of coarse-grained cohesionless soils.
Grant & Taylor (2000) conducted ‘plane strain’ centrifuge tests to inves-
tigate this; the tunnel extends from one end of the model to the other and
contains a latex bag filled with compressed air that is slowly decompressed,
simulating gradual reduction of support pressure over an infinitely long
tunnel (P /D = ∞).
They found that applying a surcharge to the top of the model simulated
by water in a latex membrane or loose sand resulted in critical stability
number values that plotted very close to the design line for P /D = ∞ in
Figure 2.2 if the cover C was assumed to include the clay only. Therefore,
the loose sand appeared to act as though it were an applied surcharge just
like the water.
On the other hand, when dense sand was overlying the clay there was
a noticeable improvement in critical stability number for tunnels with
small depths of clay cover above the crown of Cclay /D < 2, enhancing the
stability. Grant & Taylor (2000) recommended that “if a significant
layer of relatively dense coarse grained material is present above a tun-
nel in clay, an Nc value of 4 (for P /D = ∞) would seem to be appropriate
regardless of the depth of clay cover above the tunnel crown”. This is
of limited use in practice, since P /D in soft ground is always a finite
distance.
More research is required to explore the effect of the density of the
coarse-grained material and further testing at shallow cover, as well as
testing at other values of P /D. At present, therefore, it is recommended to
include overlying coarse-grained soils in the value for surcharge and to take
the cover C as the distance from the crown to the top of the clay.
WORKED EXAMPLE 2.3 A TUNNEL IN CLAY

WITH OVERLYING COARSE-GRAINED SOILS
A 6 m diameter tunnel is to be built in soft clay with overlying sand using
an EPB TBM as shown in Figure 2.10. What is the minimum support
pressure needed to prevent collapse with a factor of safety of 1.5?
Assume that P /D = 0 because it is an EPB machine and unsupported

length P is zero.
Calculate C /D based on the cover of clay:
C 6
= =1
D 6
Now look up the critical stability number Nc in Mair’s design chart (c.f.
Figure 2.2; also see Figure 2.11):
N c = 5.64.
Rearrange the stability ratio from Equation 2.1:
γ (C + D/2) + σ s − σ t
N=
cu
σ t = γ (C + D/2) + σ s − Ncu
To avoid collapse with a factor of safety of 1.5, N ≤ N c /1.5, therefore:
N c cu
σ t = γ (C + D/2) + σ s −
1.5
Now taking the sand as surcharge, and assuming no other surcharge is
applied at the surface:
σ s = γ sand zsand = 17 × 6 = 102 kPa

Figure 2.11 Worked Example 2.3 using the design chart to find the critical stability
number.
Therefore the required minimum support pressure is:
5.64 × 30
σ t = 18 × 9 + 102 − = 264 − 112.8 = 151.2 kPa
1.5
2.2.4 Numerical modelling of heading

stability in clay
Except for the special case where P /D = ∞, which can be modelled as a 2D
plane strain problem, heading stability must be modelled in 3D. There are
a number of ways to bring a tunnel heading to failure in a numerical model,
which are all based on the stability ratio equation (Equation 2.1):
• Reduce the undrained shear strength until failure occurs

• Increase the surcharge until failure occurs
• Decrease the tunnel support pressure until failure occurs.
Increasing the surcharge or decreasing the support pressure are the two
ways in which 1 g and centrifuge models of heading failure are controlled
(Casarin & Mair, 1981; Kimura & Mair, 1981).
Müller (2015) found that each of these three methods gave similar results
and are therefore interchangeable (Figure 2.12). Bradley (2013) also got
Figure 2.12 Comparison of Kimura & Mair (1981) design chart for P /D = 1 and P /D = ∞
with numerical modelling by Müller (2015) and Bradley (2013). ‘Safety’ is
reducing undrained shear strength to failure, ‘Surcharge’ is increasing sur-
charge to failure, and ‘Support pressure’ is reducing support pressure to
failure.
very similar results at P /D = ∞ using the reduction of support pressure

method (and applying surcharge if failure was still not achieved). Müller’s
results for P /D = 1, however, did not agree with the design chart.
Pound (2005), Müller (2015) and Bradley (2013) all modelled a tunnel
heading with P /D = 0, and their results are shown in Figure 2.13. Again,
similar to P /D = 1 in Figure 2.12, the numerical model appeared to consis-
tently overestimate the value of critical stability ratio Nc at failure compared
to the design charts based on centrifuge tests and case histories produced
by Kimura & Mair (1981) and Mair & Taylor (1997). This may be because
for the 2D plane strain situation (P /D = ∞) the failure is well-defined in
the numerical model, with a sharp increase in displacement, whereas for
other values of P /D Bradley (2013) found that the displacement increased
more gradually, making it difficult to define a precise point where failure
occurred. Pound (2005) defined failure as an increase of maximum face
displacement of more than 0.1D for a 1% decrease in undrained shear
strength (or approximately a 1% increase in Nc). If this were applied to
Bradley’s models then the Nc values at higher C /D ratios would be higher
than shown in Figure 2.13. It should also be noted that Kimura & Mair
(1981) did not specify how failure was defined in their centrifuge tests, so
this may also be a source of discrepancy.
Figure 2.13 Comparison of Kimura & Mair (1981) design chart and Davis et al. (1980)
lower bound with numerical modelling by Pound (2005), Müller (2015) and
Bradley (2013), all for P /D = 0 .
Therefore, numerical modelling of typical headings in clay will not

be straightforward. However, in most practical situations, maximum
face areas, maximum unsupported lengths, or minimum support
pressures will be limited to a significantly lower value by allowable
deformations.
2.2.5 Summary of undrained stability

Undrained stability may be assessed by calculating the stability ratio N,
and comparing it to the critical stability number Nc in Mair’s design chart
in Figure 2.2. The factor of safety is given by N c /N .
Undrained stability depends on the depth and diameter of the tunnel, the
value of undrained shear strength, the unsupported length, support pres-
sure applied from within the tunnel and surcharge pressure applied at the
surface. Stability can be improved either by increasing Nc or by reducing N,
which can be achieved by:
• reducing the diameter (or reducing the equivalent diameter by reduc-

ing the face area)
• increasing the cover
• decreasing the unsupported length
• decreasing the depth or surcharge pressure
• increasing the support pressure applied within the heading
• increasing the undrained shear strength, either by ground improve-

ment or by relocation of the tunnel (usually if the tunnel is moved
deeper the undrained shear strength will be higher)
• presupport measures such as canopy tubes, spiling or face dowels
(these can be thought of as effectively either reducing the unsupported
length or applying a support pressure)
Where undrained shear strength increases with depth the value at 0.6–0.8
times the depth to axis should be used.
Where a tunnel is in clay overlain by coarse-grained cohesionless soils,
treat the cohesionless soils as a surcharge and assume the cover C used in
the design chart is the cover of clay only.
Collapse of a tunnel heading in clay does not usually occur suddenly but
is a progressive failure. Therefore, it can be difficult to define when failure
occurs in a laboratory test or numerical model.
2.3 BLOW-OUT FAILURE IN CLAY
Blow-out failure only occurs in tunnels with a pressurised face, which can
be compressed air, slurry in a slurry TBM or earth pressure in an EPB
TBM. There are several ways in which a blow-out can occur:
• softening and erosion

• hydraulic fracturing
• passive failure
Passive failure is like a stability failure in reverse: a large cone of soil is

forced upwards, failing along large shear surfaces. The main difference to
stability failure is that the weight of the soil is a favourable rather than
an unfavourable load, and the volume of soil involved is usually larger.
Therefore, passive failure requires high face pressures.
2.3.1 Softening and erosion

Softening and erosion appears to be most common for pressurised water
conveyance tunnels during operation where the head of water is greater
than the distance to the top of the clay or the surface. A leak out of the lin-
ing can, over time, gradually soften and erode the clay, eventually creating
a path out of the clay to the surface or into a more permeable soil, resulting
in flooding and/or excessive loss of water from the tunnel. This happened
to a raw water tunnel near the village of Datchet in 2006, where water
eroded a path through the London Clay and suddenly burst out, spout-
ing 15 ft into the air and producing flood waters 3 ft deep (BBC, 2006).
14 properties were flooded, most only in their gardens, before the tunnel
could be isolated. This was an unbolted concrete wedgeblock tunnel com-

missioned in 1976 (Pawsey & Humphrey, 1976), so it would have relied on
the ground pressure from the clay, and the impermeability of the clay itself,
to remain watertight (Wood, 2008). The tunnel was at approximately 30 m
depth below the ground surface, but the reservoir level was approximately
18 m higher than the ground surface level above the tunnel (Pawsey &
Humphrey, 1976). Therefore, even though the overburden pressure at the
tunnel depth, about 600 kPa, was significantly higher than the internal
water pressure of about 480 kPa, a blow-out still occurred.
2.3.2 Hydraulic fracturing in clay

Hydraulic fracturing is a localised effect that can happen in the short term,
or may occur after some softening and erosion has reduced the effective
cover. This is where the clay is fractured by a high localised fluid pressure
that exceeds the tensile or shear strength of the clay (Marchi et al., 2014),
creating a path for the escape of support fluid. This will usually occur at
the crown of the tunnel. In a tunnel with compressed air, the loss of air
pressure could be followed by flooding of the tunnel or collapse of the face.
In a slurry TBM, a sudden loss of slurry can be followed by overexcavation
due to failure of the face caused by loss of support pressure. Also, release
of the slurry at the surface or into an overlying lake or river could have
environmental consequences.
Looking back at the Docklands Light Railway tunnel blow-out described
in Section 1.5.1 and shown in Figure 2.14, compressed air was needed for
construction of a crosspassage between the two running tunnels near the
deepest point under the River Thames, where the water pressure was about
2.7 bar at axis level (270 kPa). One bulkhead was 60 m beyond the cross-
passage and the other one was much further outbye towards the portal,
where the cover was only 8 m. The blow-out occurred near the outbye
bulkhead through the tunnel lining as the bulkheads were being tested up
to 3 bar and pressure had reached only 2.1 bar (210 kPa). Eight of the
Figure 2.14 Details of the Docklands Light Railway Extension blow-out – long section
of the South drive (based on details in New Civil Engineer, 1998).
tunnel’s 1.2 m long 5.2 m diameter precast concrete rings were blown apart
by the blast (New Civil Engineer, 1998). The full overburden pressure at
the crown at this location was less than 160 kPa. The large crater caused
by the blast was probably due to the large volume of air that was released.
For hydraulic fracturing to occur, simple models assume the fluid pres-
sure usually has to exceed the full overburden pressure of soil and water
(Holzhäuser, 2003), and generally accepted practice is to limit support
fluid pressures to this value (e.g. Guglielmetti et al., 2008). This is usu-
ally significantly lower than the pressure needed for passive failure to
occur.
Bezuijen & Brassinga (2006) show from field data and centrifuge tests
that bentonite slurry blow-outs caused by hydraulic fracturing can occur at
much lower face pressures than those needed for passive failure predicted
by finite element or kinematic analysis methods. This is because these
methods do not take account of the fact that slurry is a fluid. Bezuijen &
Brassinga found this limit to be approximately the pore pressure plus 2 to
3 times the effective stress for their case.
Marchi et al. (2014) show that fracture initiation in clay may be caused
by either tension or shear. The most important factor is the confining pres-
sure, otherwise known as the minor principal stress. As fluid pressure
increases, the radial stress in the surrounding clay increases but the circum-
ferential stress decreases. Failure occurs either when the circumferential
stress reaches the tensile strength or when the difference between the radial
and circumferential stresses causes shear failure.
The pressure at tensile fracture is given by (Mitchell & Soga, 2005):
Pf = 2σ 0 − u0 + σ t′ (2.3)
Pf is the fracture pressure

σ 0 is the initial confining pressure (the minor principal total stress)
u0 is the initial pore pressure
σ t′ is the effective tensile strength (tension positive)
This equation means that fracturing pressure increases with initial confin-
ing pressure with a gradient of 2. However, this assumes the soil is linear
elastic, and in reality the gradient may be less than 2 (Marchi et al., 2014).
The pressure at shear fracture is given by (Soga et al., 2005):
Pf = σ 0 + ncu (2.4)
Pf is the fracture pressure

σ 0 is the initial confining pressure (the minor principal total stress)
n is a constant
cu is the undrained shear strength
Figure 2.15 Fracturing pressure as a function of initial confining pressure of different

clay soils, from Marchi et al. (2014).
For a clay with a positive liquidity index (i.e. the water content is above the
plastic limit), n = 1, but for a clay with a negative liquidity index n = 1.5 to 2.
From a large number of experimental tests and case histories of fracture
grouting, Marchi et al. (2014) showed that the lower bound to all the mea-
sured fracture pressures was approximately the initial confining pressure,
as shown in Figure 2.15. Therefore, although the undrained shear strength
or effective tensile strength of the soil may result in a higher value of frac-
turing pressure, it seems the fracture pressure cannot be lower than the
minor principal total stress.
2.3.3 Passive failure in clay

Since hydraulic fracturing is the critical mechanism when support fluid is
used, passive failure in clay is only likely to be the limiting case for EPB
TBMs. Most studies of passive failure in clay, involving numerical mod-
els or kinematic limit state analysis, have assumed a uniform face pres-
sure (Mollon et al., 2013), whereas in an EPB machine we would expect a
face pressure that increases with depth. Despite this shortcoming, a design
Figure 2.16 Design chart for critical passive failure blow-out (from Mollon et al., 2013).
chart by Mollon et al. (2013) based on their asymmetric ‘M2’ velocity field
method is shown in Figure 2.16, with values given in Table 2.1 to aid inter-
polation for design purposes. The values of N γ and N coh from the design
chart should be used in the following equation:
σ b = γ DNγ − cu N coh + σ s (2.5)
σ b is the critical face pressure for a passive failure blow-out

γ is the bulk unit weight of the clay
D is the tunnel diameter
N γ is a stability number taking account of soil weight
N coh is a stability number taking account of soil cohesion
σ s is the surcharge pressure in kPa
This equation is derived from the more general stability equation, which is:
σ t = γ DNγ − cN coh + σ s N s (2.6)
N γ, N coh and Ns are stability numbers for the effect of soil weight, cohe-
sion and other effects, respectively, which are dimensionless
σt is the required support pressure in kPa
c is Mohr–Coulomb cohesion in kPa
Table 2.1 Values used to produce Figure 2.16 for critical

passive failure blow-out (from Mollon et al., 2013).
Nγ (passive failure N coh (passive failure
C /D blow-out) blow-out)
0.6 0.98 −7.02
0.8 1.22 −8.47
1.0 1.42 −9.43
1.3 1.71 −10.44
1.6 2.01 −11.40
2.0 2.40 −12.53
2.5 2.90 −13.75
3.0 3.39 −14.80
For undrained constant volume behaviour, Ns = 1 and c = cu. In order to

make Equation 2.1 and Equation 2.6 equivalent, all that is needed is to
substitute Nγ = C /D + 0.5 and N coh = N . But in Mollon et al.’s ‘M2’ velocity
field method, Nγ ≠ C /D + 0.5, because the velocity field is asymmetric (the
maximum velocity is set 0.4D above the centre of the face), and therefore
the results cannot be used with Equation 2.1, but have to be used with the
more general Equation 2.6.
Unfortunately, no centrifuge modelling of passive failure in clay has been
published, but Mollon et al.’s M2 method is corroborated to some degree
by their numerical modelling of a heading using FLAC3D, for which a com-
parison is shown in Figure 2.17 for D= 10 m, γ = 18 kN/m3 and cu = 20 kPa.
Figure 2.17 shows that the kinematic analysis overestimates the critical
blow-out pressures compared to the numerical modelling results. This is
unsurprising because it is an upper bound and defines when the ground
must fail, and therefore will always be on the unsafe side. However, it must
be close to the true collapse geometry and conditions, as it is not too far
above the FLAC3D results.
Figure 2.17 also gives us a feel for the magnitude of face pressure
required to cause passive failure. Even for this low value of undrained
shear strength (20 kPa), the critical face pressure is about 1.5 times the
full overburden pressure at axis level. At an undrained shear strength of
30 kPa the critical face pressure is double the full overburden pressure
(Mollon et al., 2013).
Also shown on Figure 2.17 is the minimum face pressure required to avoid
collapse, from the design charts based on centrifuge testing produced by Kimura
& Mair (1981), for the same geometry D = 10 m and P = 0, soil bulk unit
weight γ = 18 kN/m3 and undrained shear strength cu = 20 kPa. Also shown,
for comparison, is the full overburden pressure at axis level. The aim of design
would be to find a safe zone of face pressures between collapse and blow-out,
allowing for factors of safety and for variability of the applied pressure.
Figure 2.17 Comparison of critical blow-out pressures using the design chart based on
‘M2’ velocity field kinematic analysis by Mollon et al. (2013) and a 3D finite
difference model in FLAC3D, also by Mollon et al. (2013), for D = 10 m,
γ = 18 kN/m3 and cu = 20 kPa.
In summary, to estimate the critical face pressure that would cause a

passive failure blow-out, the design chart in Figure 2.16 and Equation 2.5
can be used to provide an initial estimate, and if accuracy is of critical
importance, a 3D numerical model can be used. Where a fluid such as
slurry or compressed air is used for face support, a hydraulic fracturing
blow-out is more likely, and face pressures should be limited to below the
minor principal total stress, calculated at the crown of the tunnel, or to a
limit based on the worst case of shear or tensile hydraulic fracturing using
Equation 2.3 and Equation 2.4.
WORKED EXAMPLE 2.4 CRITICAL BLOW-

OUT PRESSURE FOR A TUNNEL IN CLAY
A 5 m diameter tunnel is to be built in clay with shallow cover of only
3 m. The clay has bulk unit weight γ = 18 kN/m3, undrained shear strength
cu = 25 kPa, it has a positive liquidity index, and effective tensile strength
σt′ = 0 kPa. The coefficient of earth pressure at rest K0 = 0.6 and the water
table is at the surface. What is the critical blow-out pressure for an EPB
machine?
For the face pressure, hydraulic fracturing should not normally be

possible in an EPB machine unless a large amount of soil condition-
ing is pumped into the ground. Therefore we will do a calculation for
passive failure.
C /D = 3/5 = 0.6
From Table 2.1, N γ = 0.98 and N coh = −7.02 for passive failure.
Using Equation 2.5:
σ b = γ DNγ − cu N coh + σ s = 18 × 5 × 0.98 − 25 × ( −7.02) + 0

= 88.2 + 175.5 = 263.7 kPa
While grouting the annulus between the segmental lining and the ground
behind the shield and tailskin, hydraulic fracturing may cause grout to
escape to the surface. The limiting pressure may be given by either a tensile
(Equation 2.3) or shear (Equation 2.4) mechanism, and the critical loca-
tion will be at the crown of the tunnel.
The initial confining pressure σ0 is the minor principal total stress,
which in this case is the horizontal total stress, which is given by:
σ 0 = K0 (σ v − u0 ) + u0
because K0 is the ratio of horizontal to vertical effective stress.

The initial pore pressure u0 at the crown is 30 kPa and the vertical total
stress is given by:
σ v = γ zcrown = 18 × 3 = 54 kPa
Therefore, the initial confining pressure is:
σ 0 = K0 (σ v − u0 ) + u0 = 0.6 ( 54 − 30) + 30 = 44.4 kPa
Assuming the effective tensile strength σ t′ is zero, the pressure at tensile

fracture is therefore:
Pf ,tension = 2σ 0 − u0 + σ t′ = 2 × 44.4 − 30 + 0 = 58.8 kPa
and the pressure at shear fracture is:
Pf ,shear = σ 0 + ncu = 44.4 + 1 × 25 = 69.4 kPa

The maximum grouting pressure at the crown will not cause hydraulic
fracturing if it is kept below the lower value of 58.8 kPa, or more conserva-
tively the initial confining pressure (or in situ minor principal total stress) of
44.4 kPa could be used. The grouting pressure needs to be higher than the
pore pressure of 30 kPa at the crown, giving a fairly narrow band of allow-
able pressures, especially if factors of safety and variability are accounted for.
The passing of the TBM will of course change the stress state in the
ground, so the minor principal total stress will not be this initial undisturbed
value. Stresses calculated by 3D numerical modelling could be used but there
will still be considerable uncertainty in the results of this calculation.
WORKED EXAMPLE 2.5 CRITICAL COLLAPSE AND

BLOW-OUT PRESSURES FOR A TUNNEL UNDER A RIVER
A 12 m diameter bored tunnel is to be constructed using a closed-face
TBM (either a slurry or EPB) beneath a river with clay cover of 6 m, as
shown in the section in Figure 2.18. The highest credible river level dur-
ing construction is 12 m and the lowest is 6 m.
a. Calculate the face pressure required to avoid collapse, with factor

of safety of 1.5 and allowance for variability of ±25 kPa.
b. Calculate the hydraulic fracturing pressure for slurry in the face of
a slurry TBM.
c. Calculate the blow-out pressure for an EPB TBM.

a. First calculate the face pressure required to avoid collapse, with

factor of safety of 1.5 and allowance for variability of ±25 kPa:
C /D = 6/12 = 0.5
P /D = 0
From Mair’s design chart for undrained stability, Nc = 3.94.
To achieve a factor of safety (Nc /N) of 1.5, requires:
Nc Nc 3.94
= 1.5 = =
N  γ (C + D/2) + σ s − σ t   18 × 12 + 120 − σ t 
 cu   40

 
Rearranging for σt gives:
N c cu 3.94 × 40
σ t = γ (C + D/2) + σ s − = 18 × 12 + 120 − = 231 kPa
1.5 1.5
Allowing for variability, the target minimum support pressure is
given by:
Pst ,collapse = 231 + 25 = 256 kPa
b. If a slurry TBM were used, the fluid pressure that would cause hydrau-
lic fracture would be above the minor principal total stress at the
crown, which for K0 < 1 is the horizontal total stress σ h . In this case,
the critical case is when the river level is at its lowest and initial pore
pressure at the crown u0 = 120 kPa (ignoring the effect of the TBM
on pore pressure, and assuming the ground is affected by tide level).
σ v = (6 × 10 + 6 × 18) = 168 kPa
σ v′ = σ v − u0 = 168 − (12 × 10) = 48 kPa
σ h′ = K0σ v′ = 0.6 × 48 = 28.8 kPa
σ h = σ h′ + u0 = 28.8 + 120 = 148.8 kPa
Note that no factor of safety or allowance for variation has yet been
applied, but the pressure at which hydraulic fracturing may occur
is much lower than the target minimum support pressure required
to prevent collapse. Therefore, it is not possible to use a slurry TBM
safely in this situation without risk of escape of slurry into the river,
a decrease of support pressure to the hydrostatic value and continu-
ous failure of the face resulting in overexcavation.
c. If an EPB machine were used, hydraulic fracturing could be avoided
and the upper limit on face pressure would be defined by passive failure.
Extrapolating from the curves in Figure 2.16, Nγ = 0.86 and
N coh = –6.3.
σ b = γ DNγ − cu N coh + σ s = 18 × 12 × 0.86 − 40 × ( −6.3) + 6 × 10

= 185.76 + 252 + 60 = 498 kPa
If a factor of safety of 1.5 were placed on the undrained shear

strength, the passive failure blow out pressure would decrease to
414 kPa. Also taking away the 25 kPa allowance for variability,
this would give a target maximum face pressure of 389 kPa.
Therefore, there is a safe range of operating pressures between
256 kPa and 389 kPa.
2.3.4 Uplift failure of a tunnel heading invert in clay

Very little is known about this, and there are no records of it ever occur-
ring in practice. However, it is theoretically possible that a tunnel in clay,
or with clay in the invert, could fail if there is a permeable layer beneath the
clay with a high enough water pressure to cause uplift failure of the clay. A
similar failure mechanism is known to occur in shafts or other deep exca-
vations. As excavation unloads the ground, failure is only resisted by the
weight of the soil and its shear resistance. Either passive failure or hydraulic
fracturing of the clay could occur, resulting in inundation of the heading
by water and soil.
For more background information on this type of failure, refer to
Section 4.2 on shaft uplift failure.
2.3.5 Summary of undrained blow-out failure

Blow-out failure only occurs in tunnels with a pressurised face, which can
be compressed air, slurry in a slurry TBM or earth pressure in an EPB TBM.
There are several ways in which a blow-out can occur:
• softening and erosion

• hydraulic fracturing
• passive failure
Passive failure is like a stability failure in reverse: a large cone of soil is

forced upwards, failing along large shear surfaces. However, passive failure
is different in that the weight of the soil is a favourable rather than an unfa-
vourable load, and the volume of soil involved is usually larger. Therefore,
passive failure requires high face pressures.
Softening and erosion appears to be most common for pressurised water
conveyance tunnels during operation where the head of water is greater
than the distance to the top of the clay or the surface.
Hydraulic fracturing is a localised effect that can happen in the short-
term, or may occur after some softening and erosion has reduced the
effective cover. This is where the clay is fractured by a high localised fluid
pressure (the fluid can be air, water, soil conditioning, slurry or grout) that
exceeds the tensile or shear strength of the clay, creating a path for the
escape of support fluid. Hydraulic fracturing can occur at much lower face
pressures than those needed for passive failure.
Uplift failure may also be a risk where a permeable layer with high pore
water pressure lies beneath a tunnel in clay, causing failure of the clay
upwards into the heading, potentially followed by flooding of water and
soil into the tunnel.
2.4 PROBLEMS
The following questions should each take about 45 minutes.

Q2.1. A 6.3 m excavated diameter tunnel is to be driven using an
EPB TBM through soft clay, at a depth to axis of 27.4 m (see
Figure 2.19). The tunnel is below a lake with maximum water
depth of 7.5 m. Assume that the soft clay has a bulk unit weight of
18.5 kN/m3 and a characteristic value of undrained shear strength
of 50 kPa.
i. Calculate the factor of safety on undrained stability (Nc/N)
for an EPB face pressure of 0 kPa. State whether the tunnel
face is stable.
ii. Calculate the target minimum EPB face pressure required to
provide a factor of safety >1.5, with an allowance for variabil-
ity of ±25 kPa.
iii. Calculate the value of undrained shear strength at which
the face would be stable during a head intervention (i.e.
with no EPB face pressure applied) with a factor of safety
of 1.5.
iv. Assuming that the undrained shear strength is 50 kPa
again and does not vary with depth, at what depth to axis
would the tunnel not require application of an EPB face
pressure?
Figure 2.19 Cross-section of EPB TBM tunnel drive beneath lake.
Q2.2. A 9.5 m excavated diameter tunnel is to be driven using an open-

face TBM through stiff clay, at a depth to axis of 12 m, as shown
in Figure 2.20. Assume that the bucket excavators in the shield
can reach a maximum of 1.5 m ahead of the leading edge of the
shield. The stiff clay has a bulk unit weight of 20 kN/m3 and
is overlain by 3 m of sandy gravel with a bulk unit weight of
18 kN/m3. The undrained shear strength of the stiff clay is 50 kPa
at the interface with the sandy gravel and increases with depth by
8 kPa per metre depth.
Figure 2.20 Cross-section of open-face TBM tunnel in stiff clay.

i. Using the undrained shear strength at the tunnel axis level,

calculate the factor of safety on undrained stability (Nc / N).
Assume there is no surcharge.
ii. Now use the undrained shear strength at 0.6 times the
depth to axis (refer to Section 2.2.2) to calculate the factor
of safety on undrained stability (Nc / N). Assume there is no
surcharge.
iii. Now repeating the same calculation as in (ii), but with a sur-
charge of 75 kPa, calculate the factor of safety on undrained
stability (Nc / N).
iv. What can you conclude from these calculations? If a factor
of safety of 1.4 were required for undrained stability, is it
stable in situations (ii) and (iii)? If not, what can be done
about it?
Q2.3. The same tunnel in Q2.2 may encounter silt and sand layers
within the stiff clay stratum that have pore pressures within them,
in the bottom half of the face. At this location the tunnel is 1.5 m
deeper, as shown in Figure 2.21. Assume the pore pressure is
hydrostatic with a piezometric level at the ground surface, and
that seepage in the bottom 1 m of the face is acceptable. Assume
that K0 = 1.0.
Figure 2.21 Cross-section of open-face TBM tunnel in stiff clay with silt and sand
partings below axis level.
i. To find the compressed air pressure that could cause hydrau-

lic fracturing of the stiff clay and air loss, calculate the hori-
zontal stress at the crown of the tunnel, assuming this could
reduce during excavation to a value equal to either twice the
undrained shear strength (since the vertical total stress will be
zero, Mohr’s circle for undrained behaviour sets this condi-

tion) or the initial horizontal total stress, whichever is lower.
ii. Calculate the pore pressure 1 m above the invert level. Is the
allowable compressed air pressure calculated in (i) sufficient
to counteract this?
iii. Calculate the compressed air pressure required to cause pas-
sive failure. Which mode of blow-out – hydraulic fracturing or
passive failure, is more likely to occur?
REFERENCES

165–173.
Atkinson, J. H. (2007). The mechanics of soils and foundations, 2nd Edition.
London: Taylor & Francis.
Atkinson, J. H. & Mair, R. J. (1981). Soil mechanics aspects of soft ground tunnel-
ling. Ground Eng., July, 20–38.
BBC (2006). Homes flooded by reservoir leak. BBC News, Saturday 8th April
2006. Available at: http://news.bbc.co.uk/1/hi/england/berkshire/4892522.
stm [last accessed 9th March 2017].
Bezuijen, A. & Brassinga, H. E. (2006). Blow-out pressures measured in a cen-
trifuge model and in the field. Tunnelling: A decade of progress – GeoDelft
1995–2005 (eds Bezuijen, A. & van Lottum, H.), pp. 143–148. Leiden:
Taylor & Francis/Balkema.
Bradley, J. (2013). Numerical analysis of tunnel heading stability. MSc Tunnelling
and Underground Space, Finite Element Methods Project, University of
Warwick, UK.
Broms, B. B. & Bennermark, H. (1967). Stability of clay at vertical openings.
ASCE J. Soil Mech. Found. Eng. Div. SM1 93, 71–94.
Casarin, C. & Mair, R. J. (1981). The assessment of tunnel stability in clay by
model tests. Soft ground tunnelling: Failures and displacements, pp. 33–44.
Rotterdam: Balkema.
Davis, E. H., Gunn, M. J., Mair, R. J. & Seneviratne, H. N. (1980). The stability
of underground openings in cohesive materials. Géotechnique 30, 397–416.
EN 1997-1:2004 (2009). Eurocode 7: Geotechnical design – Part 1: General rules,
incorporating corrigendum February 2009. Brussels: European Committee
for Standardization.
Golder Associates (2009). Ground control for slurry TBM tunnelling. GEO Report
249. Hong Kong: The Government of the Hong Kong Special Administrative
Region.
Grant, R. J. & Taylor, R. N. (2000). Stability of tunnels in clay with overlying
layers of coarse grained soil. Proc. ISRM Int. Symp., 19th–24th November,
Melbourne, Australia. Published by CRC Press/ISRM.
Guglielmetti, V., Grasso, P., Mahtab, A. & Xu, S. (2008). Mechanized tunnelling
in urban areas. London: Taylor & Francis Group.
Holzhäuser, J. (2003). Geotechnical aspects of compressed air support on TBM

tunnelling. Engineering and health in compressed air working, pp. 359–371.
London: Thomas Telford.
Kimura, T. & Mair, R. J. (1981). Centrifugal testing of model tunnels in soft
clay. Proc. 10th Int. Conf. Soil Mech. & Found. Engrg, Stockholm, Vol. 1,
pp. 319–322.
Foundation Engineering, Hamburg, Vol. 4.
Marchi, M., Gottardi, G. & Soga, K. (2014). Fracturing pressure in clay.
ASCE J. Geotech. Geoenviron. Eng. 140 (2). https://doi.org/10.1061/(ASCE)
GT.1943-5606.0001019
Mitchell, J. K. & Soga, K. (2005). Fundamentals of soil behavior, 3rd Edition.
London: Wiley.
Mollon, G., Dias, D. & Soubra, A.-H. (2013). Continuous velocity fields for col-
lapse and blowout of a pressurized tunnel face in purely cohesive soil. Int. J.
Numer. Anal. Meth. Geomech. 37, 2061–2083.
Müller, R. (2015). Design Study Report. MSc Tunnelling and Underground Space,
Finite Element Methods Project, University of Warwick, UK.
New Civil Engineer (1998). Docklands Light Railway Lewisham Extension blow-
out. February 1998.
Pawsey, D. B. H. & Humphrey, A. W. (1976). The Queen Mother Reservoir, Datchet
– some aspects of its design and construction. Ground Eng., October, 27–30.
Pound, C. (2005). Improved solutions for the stability of a tunnel heading in cohe-
sive soil. Proc. Underground Construction 2005. London: Brintex.
Soga, K., Ng, M. Y. A. & Gafar, K. (2005). Soil fractures in grouting. Proc. 11th
Int. Congress on Computer Methods and Advances in Geomechanics, Turin,
Italy, pp. 397–406.
Wood, R. (2008). Wraysbury Reservoir Inlet Tunnel – in situ structural
secondary lining of existing tunnel. Water Treatment and Supply, 159–161.
Available at: http://waterprojectsonline.com/case_studies/2008/Thames%20
Wraysbury%20Res%20Inlet%20tunnel%202008.pdf [last accessed 9th March
2017].
Chapter 3
Drained stability
The most common methods used in practice for stability calculations are
Mair’s design charts based on centrifuge testing for the undrained case
(Kimura & Mair, 1981), and Anagnostou & Kovári’s method for the
drained case (Anagnostou & Kovári, 1994, 1996a and 1996b). This chapter
will first provide some background and will then describe how to use
Anagnostou & Kovári’s method to solve drained stability problems.
• how tunnel headings collapse or blow-out in the drained case

• how upper bound and lower bound limit states, limit equilibrium
solutions, numerical modelling and empirical data help us to estimate
the required support pressure
• the role of groundwater pressure and seepage forces in drained
stability
• how slurry and earth pressure balance (EPB) tunnel boring machines
(TBMs) support the face
• slurry penetration and stand-up times
• calculate the minimum and maximum support pressures for tun-

nelling in drained soils to avoid collapse, overmucking or blow-out
failures
3.1 DRAINED STABILITY WITHOUT SEEPAGE
As discussed in the previous chapters, four main factors influence stability:
• Cohesion and support pressure, which improve stability

• Gravity and seepage forces, which worsen stability
DOI: 10.1201/9780429470387-3 83
Without either cohesion to hold the soil grains together or a support pres-
sure applied to the face, a vertical face will fall down due to gravity. In the
case of a tunnel below the water table, seepage forces will also contrib-
ute to instability. Seepage forces occur when there is a hydraulic gradient
between the ground and the tunnel. As groundwater seeps through the
ground towards the face, it pushes the soil grains apart with a ‘seepage
force’ proportional to the hydraulic gradient, in the direction of flow.
‘Drained’ soils are permeable enough that excess pore pressures are dis-
sipated during the timescale of construction. For tunnel heading stability,
this means that any negative excess pore pressures that may have helped
hold the soil grains together cannot be counted on, as water will have time
to flow in from the surrounding ground. Therefore, the failure of drained
soils depends on the drained cohesion c′ and the internal angle of friction
ϕ, rather than the undrained shear strength.
Note that in drained stability it has become common practice to refer to the
cover using the letter ‘H’ rather than the letter ‘C’, although some papers
still use ‘C’. These will be used interchangeably in this chapter.
3.1.1 Dry cohesionless soils

The simplest situation to consider is that of dry, cohesionless soils, i.e.
where cohesion c′ = 0. Failure is governed by the geometry of the heading,
the unit weight of the soil and the angle of friction ϕ. It is certain that a
support pressure is required. This may be provided by slurry pressure in a
slurry machine, compressed air in an open face, or by effective stress and
fluid pressure in an EPB machine.
Atkinson & Potts (1977) performed 1 g (i.e. scale models on the labora-
tory floor) and centrifuge tests (scale models accelerated in a centrifuge) on
dry cohesionless Leighton Buzzard sand. They also developed upper bound
and lower bound limit state solutions for a plane strain tunnel, i.e. where
the unsupported length P = ∞.
If you remember from Chapter 2, the upper bound is any kinematically
admissible mechanism, and at this load, the heading must fail. The lower
bound is a statically admissible stress field, which nowhere violates the fail-
ure criterion for the soil, and at this load, the heading cannot fail. The
upper and lower bounds bracket the true collapse load, and ideally they are
close enough together that a target support pressure may be set.
Figure 3.1 shows the upper and lower bound solutions for a dry cohe-
sionless soil from Atkinson & Potts’s paper (1977). The vertical axis is the
support pressure σ t , normalised by the unit weight of the soil γ and the
diameter D. The horizontal axis is the ratio of cover H to diameter D, and
it can be seen that the depth of the tunnel does not affect either the upper
or lower bound limit state solutions. Remember this is different to an und-
rained stability problem, where depth is a factor (c.f. Figure 2.2).
Drained stability 85
Figure 3.1 Upper and lower bound solutions for a dry cohesionless soil (Atkinson &
Potts, 1977) with angle of friction ϕ = 35°.
The upper bound is based on an infinitely long triangular wedge of soil

falling from the crown of the tunnel. The geometry of this wedge depends
on the angle of friction ϕ. At low values of H/D, the wedge intersects the
surface, so no values can be determined.
Although intended for dry cohesionless soils, it may be possible to use
this solution for cohesionless soils below the water table if it can be assumed
that there is no flow of groundwater and hence no seepage forces. In this
case one would use the submerged unit weight, which is given by:
γ ′ = γ − γ w (3.1)
γ ′ is the submerged unit weight in kN/m3

γ is the bulk unit weight in kN/m3
γ w is the unit weight of water, usually assumed to be 10 kN/m3
3.1.2 Dry drained soils with cohesion

Leca & Dormieux (1990) found upper and lower bound limit state solu-
tions for c′ – ϕ soils, that is, drained soils with cohesion. This was a great
step forward because many natural soils have some cohesion. The main
difference compared to Atkinson & Potts (1977) was that the kinematic
mechanisms (which provided the upper bound) were in three dimensions
and P/D was zero rather than infinity, which was more realistic. Also, the
Figure 3.2 Upper bound mechanisms from Leca & Dormieux (1990).
lower bound included three different stress states, whereas Atkinson &
Potts assumed one plane strain stress state. The geometry of the upper
bound is either one or two truncated cones, as shown in Figure 3.2. The
value of the angle α is varied to find the worst case.
Figure 3.3 shows the effect of increasing the value of cohesion from 0 to
5 to 10 kPa for a soil with ϕ = 35°. Both the upper and lower bounds move
downwards, meaning that the required support pressure to prevent instabil-
ity decreases as cohesion increases. At c′ = 10 kPa, Leca & Dormieux’s upper
bound is at zero. If the upper bound represented the true collapse mecha-
nism, which we don’t yet know for sure, then no support pressure would be
required to maintain stability. This shows that tunnel headings in drained soil
above the water table may be stable at quite small values of cohesion. This
can be provided by moisture in the soil (like a sandcastle), a low clay content
or cementing of the soil grains. However, driving a tunnel in such a situation
must be done carefully, because even a small amount of perched water can
cause instability due to seepage forces, and low values of cohesion cannot
always be relied on given the inherent variability of geological materials.
Figure 3.3 Effect of drained cohesion c ′ on required support pressure σ t in a dry c′ – ϕ

soil with ϕ = 35°. (‘A&P’ = Atkinson & Potts, ‘L&D’ = Leca & Dormieux, ‘L/B’ =
lower bound, ‘U/B’ = upper bound).
Unfortunately, although the geometry of Leca & Dormieux’s model is

closer to reality than Atkinson & Potts’s, the upper and lower bounds are
quite far apart and this makes it difficult to use these solutions with any cer-
tainty about the value of the true collapse load. Atkinson & Potts assumed
a plane strain tunnel at both the upper and lower bounds (because this was
what they were modelling in the 1 g and centrifuge tests), essentially assum-
ing an unlined infinitely long tunnel, whereas Leca & Dormieux assumed
a three-dimensional failure at the face with unsupported length P = 0 for
the upper bound.
Anagnostou & Kovári (1994, 1996a, 1996b) used a limit equilibrium
solution they attributed to Horn (1961) and developed it to allow direct
calculation of the required support pressure in any situation above or below
the water table, with or without seepage, for tunnels where P = 0. The
geometry is shown in Figure 3.4. This solution is potentially much more
useful for practical situations involving closed-face tunnelling machines,
although it cannot be guaranteed to represent the true collapse load because
the geometry is a simplification. Messerli et al. (2010) extended the equa-
tions to allow for short unsupported lengths and Anagnostou & Perazzelli
(2013) investigated the effect of non-uniform support pressures.
Figure 3.5 shows a comparison of Anagnostou & Kovári’s limit equi-
librium solution with the limit state solutions, for a dry cohesionless soil
with ϕ = 35°. Those of you familiar with Mair & Taylor’s excellent ‘Theme
Lecture’ paper at the 14th ICSMFE in 1997 will recognise Figure 3.5 as
Figure 3.4 Wedge and prism model (from Anagnostou & Kovári, 1996b, after
Horn, 1961).
very similar to their Figure 9 (Mair & Taylor, 1997). Interestingly, the sup-
port pressure calculated by the limit equilibrium solution is not completely
constant and as well as being lower for H/D < 0.5, it also does increase
very, very slightly with increasing H/D after that. The limit equilibrium
solution is consistent with the limit state solutions in that the upper bounds
Figure 3.5 Comparison of Anagnostou & Kovári’s (1994) limit equilibrium solution with
limit state solutions by Leca & Dormieux (1990) (‘L&D’) and Atkinson & Potts
(1977) (‘A&P’) for a dry cohesionless soil with ϕ = 35°.
Figure 3.6 Dimensionless coefficients F0, F1, F2 and F3 based on the wedge-prism model
shown in Figure 3.4 for use in Equation 3.2 (Anagnostou & Kovári, 1996b).
Note that the labels on the F0 nomogram have been switched around com-
pared to those incorrectly printed in Anagnostou & Kovári (1996a & 1996b),
as explained in Appendix A.
(where the heading must fail) are below it, and the lower bounds (where the
heading cannot fail) are above it.
A derivation of the wedge and prism limit equilibrium solution used by
Anagnostou & Kovári is given in Appendix A. Based on these equations,
Anagnostou & Kovári (1996a, 1996b) defined four dimensionless coeffi-
cients F0, F1, F2 and F3, which may be read from nomograms shown in
Figure 3.6 and used in Equation 3.2. The derivation is not included here as
it is more important that you use Figure 3.6 and Equation 3.2 and under-
stand their implications first. Once you understand the principles, feel free
to go to Appendix A for the full derivation.
∆h
s′ = F0γ ′D − F1c′ + F2γ ′ ∆h − F3c (3.2)
D
s′ is the effective support pressure required to prevent collapse
F0, F1, F2 and F3 are dimensionless coefficients
γ ′ is the submerged unit weight, which is the bulk unit weight minus
the unit weight of water
c′ is the drained cohesion
D is the diameter of the tunnel excavation

∆h is the head difference between the in situ groundwater head h0 (c.f.
Figure 3.4) and the head in the excavation chamber hf , given by:
∆h = h0 − hf (3.3)
The F0 term in Equation 3.2 is related to gravity, the F1 term is related to

cohesion, the F2 term is related to seepage via the head difference ∆h , and
the F3 term is related to cohesion and seepage combined.
Therefore, for a tunnel in cohesionless soil where there is no seepage,
either because the ground is above the water table or because the ground-
water pressure in the soil is balanced by a fluid pressure in the TBM’s exca-
vation chamber, only the F0 term would be used. For a tunnel in c′ − ϕ
soil where there is no seepage towards the face, only the first two terms
would be used. Below the water table, these conditions occur in the front
of a slurry TBM, where the slurry pressure is usually set higher than the
groundwater pressure. An EPB TBM could in theory create these condi-
tions if there were a perfectly impermeable plug of conditioned soil in the
screw conveyor and enough soil conditioning were injected. In reality it is
not possible to guarantee this, but it is useful to consider as a limiting case
and will be used in Section 3.2.4.
For a tunnel in cohesionless soil where there is seepage, only the F0 and
F2 terms are used. For a tunnel in c′ – ϕ soil where there is seepage, all the
terms are used.
3.1.3 Comparison of analytical methods with

centrifuge tests and finite element models
As mentioned previously, the wedge and prism limit equilibrium method
makes assumptions about the geometry of failure, so it would give us con-
fidence if it could be validated by field measurements, laboratory testing
or numerical analysis. Figures 3.7 and 3.8 show comparisons between 1 g
model tests by Messerli et al. (2010) and Chambon & Corté (1994) with
the limit state upper bound and the wedge-prism limit equilibrium solution,
and to finite element models by Vermeer et al. (2002).
Messerli et al.’s tests were on a uniform fine sand with angle of friction
ϕ = 33° and unit weight γ = 17 kN/m3. Analytical solutions are presented
for comparison, with ϕ = 33° and γ = 17 kN/m3 used as input parameters.
Vermeer et al. (2002) performed 25 finite element calculations and deter-
mined that the relationship between the soil weight stability number N γ
(= σ t /γ D) and angle of friction ϕ is given by:
1 σ
Nγ = − 0.05 = t (3.4)
9 tan φ γD
This relationship was found to hold true as long as H/D > 1.
Figure 3.7 Comparison of centrifuge tests by Messerli et al. (2010) on dry cohesionless

sand with limit state solutions (Atkinson & Potts (1977) upper bound and Leca
& Dormieux (1990) upper bound), a limit equilibrium solution (Anagnostou
& Kovári, 1996a) and finite element models (‘FEM’ = FE models by Vermeer
et al., 2002).
Figure 3.8 Comparison of centrifuge tests by Chambon & Corté (1994) on dry cohe-
sionless sand with Leca & Dormieux’s (1990) upper bound (‘L&D U/B’),
Anagnostou & Kovári’s (1994) limit equilibrium solution (‘A&K’) and finite
element models (‘FEM’ = FE models by Vermeer et al., 2002).
For Messerli et al.’s tests, Anagnostou & Kovári’s limit equilibrium solu-
tion provides a good, though slightly conservative, estimate of the support
pressure at failure. Atkinson & Potts’s upper bound (below which the head-
ing must fail) is in effect on the wrong side of the data points, but perhaps
this is to be expected because the failure is three-dimensional and their
upper bound solution is for a two-dimensional unlined plane strain tunnel.
Chambon & Corté’s centrifuge tests were on Fontainebleau sand. They
estimated the angle of friction ϕ was between 38° and 42° and the unit
weight of the sand for the tests shown in Figure 3.8 was between 16.0 and
16.2 kN/m3.
For Chambon & Corté’s tests, Leca & Dormieux’s upper bound with
c′ = 0 is on the wrong side. Remember that an upper bound should be
on the unsafe side. Chambon & Corté suggest in their paper that the
Fontainebleau sand has a value of cohesion somewhere between 0 and
5 kPa. If c′ were closer to 5 kPa, Leca & Dormieux’s upper bound would
drop below Chambon & Corté’s data. The relationship based on finite ele-
ment models (Vermeer et al., 2002) overestimates the support pressure at
failure when compared to Chambon & Corté’s tests, but would be closer if
a value of c′ ≠ 0 were used. Anagnostou & Kovári’s solution would require
a higher minimum support pressure than Chambon & Corté’s tests indi-
cate is necessary, even if c′ = 5 kPa is assumed, so it is on the safe side.
3.1.4 Summary of drained stability theory

The analytical solutions show that support pressure at failure for a drained
soil is very sensitive to the value of cohesion c′, but not so sensitive to the
angle of friction ϕ, although it does have an effect. The upper bound of
Leca & Dormieux (1990) appears to provide a reasonable estimate of the
true collapse load, but it is usually on the unsafe side.
All methods suggest that the depth of the tunnel has negligible effect on
the support pressure at failure. Vermeer et al. (2002) have found that the
only exception to this is when a surcharge is applied to the surface above a
shallow tunnel with C/D < 2 and ϕ < 25°.
The support pressures required to prevent failure in drained c′ – ϕ soils
may be normalised by the unit weight of the soil γ and the diameter D. The
proportion of γ D required is of the order of 0.05–0.2 for most sands. In
order to decide what face pressure to apply, the groundwater pressure must
also be added to this value, and a factor of safety introduced. How this is
done will be covered in the next section when we will go into more detail
on the effect of groundwater pressure and seepage on stability calculations
for closed-face TBMs, and what the results tell us about how to operate
these machines.
Both the relationships based on parametric finite element studies by
Vermeer et al. (2002) and the limit equilibrium solution proposed by
Anagnostou & Kovári (1994, 1996a & 1996b) seem to provide a reason-
able, and slightly conservative, estimate of the collapse load when com-
pared to centrifuge and 1 g laboratory tests.
3.2 APPLICATION OF DRAINED STABILITY

TO CLOSED-FACE TBMS
As demonstrated in the previous sections, drained soils with little or no

cohesion require a support pressure to be applied to maintain stability.
Closed-face tunnelling machines are often used in these soils, but it is
difficult to know what support pressure to specify. If there is instability,
the uncontrolled flow of ground into the face will lead to overexcavation,
which may not be easily detected, and this in turn may cause excessive
surface settlements or collapse. This has occurred all over the world, for
example the ‘Lavender Street incident’ during EPB tunnelling for the CTRL
in London (Lovelace, 2003), or the 37 incidents, some of which reached
the surface, during construction of the SMART tunnel in Malaysia using a
slurry TBM (CEDD, 2012).
3.2.1 Application to slurry TBMs

A slurry TBM supports the ground by filling the head with bentonite slurry
under pressure. The slurry pressure must be greater than the groundwater
pressure because otherwise it would not be possible to pump it in, but also
because we need the excess slurry pressure to apply a support pressure to
the soil grains and to include an allowance for variability. This is shown in
Figure 3.9.
All slurry machines will experience fluctuations in the slurry pressure,
due to dynamic effects such as rotation of the cutterhead, excavation and
pumping, and due to the difficulty in balancing the slurry feed in and the
slurry extraction rate out. Since a drop in the slurry pressure could result in
instability and hence overexcavation, an allowance for variability is added
to the target minimum face pressure. Sometimes a factor of safety of some
kind is also included, as well as an allowance for the effects of surcharge on
the surface (Golder Associates, 2009), as shown in the following equation:
Pst = s′ + u + v + q (3.5)
Pst is the slurry pressure in the excavation chamber in kPa

s′ is the effective support pressure in kPa
u is the pore pressure in the ground in kPa
v is the allowance for variability of slurry pressure (a function of the
TBM and how it controls slurry pressure) in kPa
q is an average surcharge pressure applied to the surface in kPa
Figure 3.9 Slurry pressure diagram for a slurry TBM (after Anagnostou & Kovári, 1996b).
In Figure 3.9, the slurry pressure increases at a faster rate with depth than the
groundwater pressure. This is because slurry, when mixed with excavated
soil, has a higher unit weight, typically assumed to be around 12 kN/m3,
whereas water is just under 10 kN/m3. This means that the minimum excess
slurry pressure is always at the crown. Therefore, stability calculations for
slurry TBMs are always performed at the crown level as this is the worst
case (Golder Associates, 2009).
Because the slurry pressure is always greater than the groundwater pres-
sure, the slurry will flow into the ground. As it displaces the groundwater in
the soil’s pore spaces, the slurry flow is slowed down and soil particles sus-
pended in the slurry are filtered out and block the pores. Bentonite slurry,
at the right concentrations, also has thixotropic properties – this means
that it forms a gel when not agitated. Therefore, a ‘filter cake’ is formed,
which acts as a kind of membrane that allows the excess slurry pressure to
be applied to the soil grains, counteracting instability.
The ideal situation is where the filter cake forms close to the excavation
surface, and Anagnostou & Kovári refer to this as the ‘membrane model’.
As mentioned earlier in this chapter, drained stability can be approximated
using a wedge and prism model (Anagnostou & Kovári, 1994, who attri-
bute the model to Horn, 1961). The geometry of the model was shown in
Figure 3.4. As the slurry penetrates into the soil, the excess slurry pressure
acting to support the soil grains is spread over the penetration distance, so
as the slurry penetrates through the ‘wedge’, the resultant support force it
applies to the wedge and hence the prisms above is gradually compromised.
For higher permeability soils, penetration occurs at a faster rate.
One way to reduce the slurry penetration rate is to use a higher vis-
cosity slurry, but this increases slurry pumping and treatment costs. For
continuous TBM tunnelling, usually penetration rate is not a problem
unless the soil permeability is very high (such as in open gravels), because
the TBM is constantly advancing and excavating the ground. It is usually
only when there is a standstill of the TBM that penetration is a prob-
lem, and often this can be mitigated by temporarily increasing the slurry
viscosity. Anagnostou & Kovári (1994) found that increasing the slurry
pressure also helps.
WORKED EXAMPLE 3.1 CALCULATING THE

TARGET MINIMUM SLURRY PRESSURE
A 9 m diameter tunnel is to be built in sand with cover H = 15 m. The
sand has bulk unit weight γ = 18 kN/m3, angle of friction ϕ = 30° and
effective cohesion c′ = 0 kPa. The variability of slurry pressure achiev-
able by this machine is ±25 kPa. Surcharge q may be up to 75 kPa. The
groundwater table is at the ground surface.
Assume the membrane model applies, i.e. a filter cake forms and the
slurry does not penetrate into the soil a significant distance.
Calculate the pore pressure u:
u = Hγ w = 15 × 10 = 150 kPa
Find value of F0 from nomograms, for H/D = 1.67 (need to interpolate

between H/D = 1 and H/D = 2), and h0 = H + D (i.e. use the dashed lines).
This gives F0 = 0.22.
The other coefficients F1, F2 and F3 are not required because c′ = 0, and
because Pst will be higher than the pore pressure in the ground, so it is a
safe assumption that there will be no seepage towards the face.
The submerged unit weight is given by:
γ ′ = γ − γ w = 18 − 10 = 8 kN/m3
Therefore, the effective support pressure is given by:
s′ = F0γ ′D = 0.22 × 8 × 9 = 15.84 kPa

The target minimum slurry pressure at the crown should therefore be:
Pst ,crown = s′ + u + v + q = 15.84 + 150 + 25 + 75 = 265.84 kPa
Note that the effective support pressure, which is the force applied by the
slurry to the soil grains needed to maintain stability of the face, is very
small relative to the target minimum slurry pressure – it only makes up
6% of the total in this case.
Note also that the effect of surcharge q has been added to the target
minimum slurry pressure without any reduction. This assumption is very
conservative and could be refined by use of a numerical model or by using
the extension to the wedge-prism method described in Appendix A.
No factor of safety has been applied to this calculation as yet. GEO
Report 249 (Golder Associates, 2009) recommends applying a partial
factor of 1.2 to c′ and tan ϕ. Eurocode 7 (EN 1997-1:2004, 2009) would
require a partial factor of 1.25 on these strength parameters.
3.2.2 Slurry infiltration during TBM standstills

As the worked example shows, the target minimum slurry pressure will
be higher than the pore pressure in a cohesionless soil. Therefore, during
a standstill of the TBM, the slurry will flow into the ground. The support
pressure is no longer applied to a thin membrane at the tunnel face, but is
spread over the depth of infiltration.
Slurry will penetrate the ground to a maximum distance e max, which
depends on the permeability of the ground, the viscosity of the slurry and
the excess pressure. Another way of thinking about this is that for a given
ground permeability and slurry viscosity, there will be a ‘stagnation gradi-
ent’ at which flow will stop. The stagnation gradient is the excess pressure
Δp divided by the maximum penetration emax.
The maximum penetration of slurry e max into a soil can be given by:
∆ps d10
emax = (3.6)
ατ f
∆ps is the total pressure difference across the filter cake (the differ-
ence between the slurry pressure and the pore pressure in the
ground)
d10 is the characteristic grain diameter at which 10% of the soil passes
through a sieve (in other words, 90% of the soil grains have a
larger characteristic diameter)
τ f is the shear strength of the slurry
α is a parameter that describes the relationship between grain size and

the effective radius of a flow channel, and normally takes a value
between 2 and 4 (see Krause, 1987, and Kilchert & Karstedt,
1984). Anagnostou & Kovári (1994) take α = 2, following the
German standard for diaphragm walling DIN 4126 (1986).
Anagnostou & Kovári (1994) modelled a TBM standstill by integrat-

ing the support pressure over the maximum slurry penetration distance in
the wedge-silo model. Figure 3.10 shows that face stability decreases with
increasing characteristic grain size. Curve B shows the effect of increasing
the excess slurry pressure compared to curve A. Stability is improved, but
the face quickly becomes unstable at around d10 = 2 mm. For finer soils the
excess pressure applied in curve A would have been sufficient so there is
no real benefit gained. Curve C shows the effect of increasing the bentonite
concentration from 4% to 7% (increasing the slurry’s yield strength from 15
Pa to 80 Pa). This allows a face to be stable even in very coarse-grained soils.
Since we may want to know how the penetration of the slurry increases
with time, the following hyperbolic equation proposed by Krause (1987)
may be used:
et t
= (3.7)
emax a+t
t is the time
et is the penetration distance at time t
Figure 3.10 Reduction of safety factor with increasing characteristic grain size d10,
due to slurry penetration in a soil with c′ = 0 and ϕ = 37.5° (redrawn from
Anagnostou & Kovári, 1996b).
a is the time at which half the penetration distance has been reached,
which can be determined by laboratory testing using an infiltra-
tion column
Anagnostou & Kovári (1994) use a more complicated model for slurry pen-
etration than shown in Equation 3.7, but the results were similar. Again,
by integrating the support pressure over the penetration distance in the
wedge-silo model, they showed that stability decreases over time as slurry
penetrates into the ground. This gives a ‘stand-up time’ after which the
face will be unstable, shown as tcr in Figure 3.11. Stand-up time can be
increased by increasing the slurry pressure, as demonstrated by the curves
for ∆p = 40 kPa and ∆p = 80 kPa.
Stand-up time has a near-reciprocal relationship with permeability, so
if the permeability is reduced by an order of magnitude, then the stand-up
time is increased by an order of magnitude.
3.2.3 Slurry infiltration during excavation

Anagnostou & Kovári (1994) gave a rule of thumb that if the soil d10 value is
smaller than 0.6 mm, then infiltration will be small and the membrane model
can be assumed to apply. However, they only looked into the effect of slurry
penetration on the effective zone of application of support pressure. Broere
& van Tol (2000) identified an additional effect which is the generation of
positive excess pore pressures in the ground ahead of the face, which has a
tendency to occur when fine or medium sands are being excavated by a TBM
in a confined aquifer (i.e. with an impermeable layer above).
Figure 3.11 Safety factor as a function of time for a 4% bentonite slurry with yield
strength 15 Pa, in a coarse-grained soil with d10 = 6 mm, k = 10 −4 m/s, c′ = 0
and ϕ = 37.5° (from Anagnostou & Kovári, 1996b).
Positive excess pore pressures can be generated in the sand due to repeated
excavation of the filter cake by the cutter bits followed by infiltration of the
slurry, driving filtrate water into the ground and elevating the pore pres-
sure. These positive excess pore pressures cause a reduction in effective
stress and hence a reduction in the shear resistance of the soil. Also, higher
pore pressures in the soil effectively cancel out some of the effective support
pressure that can be used to support the soil grains.
Using a modified wedge-silo limit equilibrium model developed by Broere
(1998), Broere & van Tol (2000) modelled this effect and found that the
required minimum support pressure was significantly higher, at 80 kPa
above the in situ pore pressure compared to only 16 kPa above the in situ
pore pressure when using Anagnostou & Kovári’s (1994) full membrane
model. This was corroborated by measurements made by piezometers
installed in the alignment of the Second Heinenoord Tunnel. A different
tunnel in Rotterdam in somewhat coarser sand and with no clearly defin-
able impermeable overlying stratum had a maximum positive excess pore
pressure of only 5 kPa, demonstrating that the effect depends on the pres-
ence of sand lenses or confined sand layers.
A graph from Broere & van Tol (2000) is shown in Figure 3.12. This
shows measurements of pore pressure as the cutterhead approaches the
piezometer. During excavation, significant positive excess pore pressures
are generated ahead of the face. During ringbuilding every 1.5 m these
positive excess pore pressures dissipate and the pore pressure returns
Figure 3.12 Piezometer measurements of pore pressure ahead of a slurry TBM at the

Second Heinenoord Tunnel (Broere & van Tol, 2000).
to the initial in situ value of approximately 120 kPa (with some tidal varia-
tion evident).
A further paper by Broere & van Tol (2001) says that if the soil perme-
ability is between 10−3 and 10−5 m/s, then the effect of positive excess pore
pressures ahead of the face will be significant. This paper also extends the
model to less permeable silty sands by including a transient groundwater flow
model to replace the steady-state model of the previous paper. This allowed
for a slower dissipation of excess pore pressures during the ringbuilding cycle.
3.2.4 Application to earth pressure balance TBMs

Face stability calculations for EPB machines are more complicated. The cut-
terhead breaks up the soil and the muck moves through openings in the cut-
terhead into the ‘excavation chamber’. The excavation chamber is kept filled
with muck under pressure (‘EPB pressure’), and it is this pressure that pro-
vides the face stability. Muck is removed in a controlled manner from the
excavation chamber using an Archimedes screw, and the EPB pressure is con-
trolled by balancing the screw speed with the penetration rate of the TBM
into the ground. It is important to note that EPB pressure is usually measured
by load cells in the excavation chamber, which actually measure total stress,
i.e. the total of both the effective support pressure and the pore pressure.
At the back end of the screw, the muck falls out onto a belt conveyor at
atmospheric pressure, so there is a gradient from the EPB pressure to atmo-
spheric pressure along the screw. The gradient in effective stress is maintained
by the tortuosity of the route the soil takes up the screw and the cohesion and
angle of friction of the muck. A pore pressure gradient in the screw can only
be maintained by filling the screw with a low permeability material. If the
muck has too high a permeability, soil conditioning can be used to reduce it.
It is helpful, as we did for slurry machines, to consider the effective support
pressure and the pore pressure separately. The ‘effective support pressure’ is
the stress in the soil grains transferred from the machine to the soil through
grain-to-grain contacts. The ‘pore pressure’ is the pressure of the fluid (water,
soil conditioning and air) between the soil grains. At the back end of the
screw, both of these are zero. In the ground, far away from the influence of
the TBM, they have undisturbed in situ values dependent on their weight
and, in the case of horizontal effective stress, stress history and other effects.
In an EPB machine, unlike a slurry machine, the fluid pressure in the
excavation chamber cannot reliably be higher than the pore pressure in the
ground. With a perfect plug of impermeable muck in the screw, it could be
close to the in situ pore pressure, but it cannot be higher unless soil condi-
tioning is injected at a high rate. Therefore, there is nearly always seepage
of groundwater towards the face, which, as we know, is bad for stability.
The effect can be calculated using the method described in Anagnostou
& Kovári (1996a). It can be illuminating to calculate the two extremes:
in the first case (A) with zero pore pressure in the excavation chamber,
giving the maximum destabilising hydraulic gradient in the ground, and

in the second case (B) with the pore pressure in the excavation cham-
ber equal to the in situ pore pressure in the ground. In the first case A,
the destabilising seepage force needs to be counteracted by an increased
effective support pressure, but the effective support pressure is equal to
the total EPB pressure because there is no pore pressure to drive against.
In the second case B, there is no destabilising seepage pressure, so the
effective support pressure is only what is required to maintain stability
due to gravity, but we have to add the pore pressure to get the EPB pres-
sure. It turns out, if you run the numbers, that the EPB pressure required
for Case A is always less than in Case B. In other words, the extra effec-
tive support pressure required to resist the destabilising seepage forces is
less than the in situ pore pressure.
WORKED EXAMPLE 3.2 TARGET MINIMUM EPB PRESSURE

An 8 m diameter tunnel is to be built in sandy silt with cover H = 16 m and
the water table located at the ground surface, i.e. Hw = 16 m and h0 = 24 m.
The sandy silt has saturated unit weight γ = 18 kN/m3, angle of friction
ϕ = 20° and effective cohesion c′ = 0 kPa.
The variability of EPB pressure achievable by this machine is ±25 kPa.
Surcharge q may be up to 25 kPa.
We want to use Equation 3.2:

∆h
s′ = F0γ ′ D − F1c + F2γ ′ ∆h − F3c
D
Since c = 0 kPa, we need to find F0 and F2 only.
Find value of F0 and F2 from the nomograms, for H/D = 2, and h0 = H + D
(i.e. use the dashed lines). As shown in Figure 3.13, this gives F0 = 0.50 and
F2 = 0.60.
Figure 3.13 Worked Example 3.2 – reading nomograms to find values of F0 and F2
(corrected versions of nomograms from Anagnostou & Kovári, 1996b).
The submerged unit weight is given by:
γ ′ = γ − γ w = 18 − 10 = 8 kN/m3
Case A: maximum seepage

For Case A, where pore pressure in the chamber hf = 0, the head
difference ∆h between the ground and the TBM that is driving seep-
age is given by:
∆h = h0 − hf = 24 − 0 = 24 m
The effective support pressure required to maintain stability is given by:
s′ = F0γ ′ D + F2γ ′ ∆h = 0.50 × 8 × 8 + 0.60 × 8 × 24 = 32 + 115.2 = 147.2 kPa
The pore pressure the TBM is pushing against is zero.

The target minimum EPB pressure at the crown should therefore be:
PEPB,crown = s′ + u + v + q = 147.2 + 0 + 25 + 25 = 197.2 kPa
No factor of safety has been applied to this calculation as yet. GEO Report
249 (Golder Associates, 2009) recommends applying a partial factor of
1.2 to c′ and tan ϕ. Eurocode 7 (EN 1997-1:2004, 2009) requires a partial
factor of 1.25.
Case B: no seepage
For Case B, a perfect impermeable plug is maintained in the screw and
pore pressure in the chamber is equal to the in situ pore pressure in the
ground. Therefore, there is no seepage towards the TBM, and ∆h = 0.
The effective support pressure required to maintain stability is
given by:
s′ = F0γ ′D + F2γ ′ ∆h = 0.50 × 8 × 8 + 0.60 × 8 × 0 = 32 + 0 = 32.0 kPa
Calculate the pore pressure u at the crown:
u = H wγ w = 16 × 10 = 160 kPa
The target minimum EPB pressure at the crown should therefore be:
PEPB,crown = s′ + u + v + q = 32 + 160 + 25 + 25 = 242.0 kPa
No factor of safety has been applied to this calculation.

In Case A the effective support pressure is much higher than in Case B, at

147.2 kPa versus 32.0 kPa, but in Case B, a higher EPB pressure is required
to maintain stability. Therefore, one would think, surely it is better to drain
the ground and run the machine with a lower EPB pressure? Why do we
go to great lengths to get the soil conditioning right so we can ‘maintain a
plug in the screw’? One reason is that, particularly in high permeability soils,
uncontrolled water flow through the screw could flood the tunnel. But there is
another very good reason: the higher the effective support pressure, the higher
the effective stress in the soil’s grain to grain contacts. The shear strength of
a soil is proportional to the effective stress, so this means the soil will behave
like a stronger material. This is undesirable because we want it to flow nicely
through openings in the cutterhead, fill the excavation chamber and move
along the screw, not arch around openings and block the screw. The increased
friction between soil grains will also generate a lot of heat and increase wear
of metal parts. This means an increase in maintenance stops and slower over-
all progress. In some soils, the machine cannot be advanced at all.
The same principles govern the stability calculations for both slurry and
EPB machines, even though they work in different ways. The main differ-
ence is that in a slurry machine, the slurry pressure is always greater than
the groundwater pressure, but we need a filter cake to form to also apply
an effective support pressure to support the soil grains. In an EPB machine,
the machine is applying an effective support pressure to the soil grains, but
we need to maintain a plug in the screw to counteract the groundwater
pressure and minimise the hydraulic gradient.
Soil conditioning is critical to making the muck in an EPB machine
behave the way we need it to. Ideally we turn it into a low-permeability,
cohesive and plastic material to minimise the work required to get it to
flow through the machine, minimise friction and wear of metal parts, and
create a good plug in the screw so that the pore pressure in the excavation
chamber is close to the in situ pore pressure. Often, the spoil needs to be
suitable for disposal, so we compromise between creating the perfect EPB
material and producing spoil that can be reused as fill or transported and
disposed of safely.
3.3 BLOW-OUT FAILURE IN DRAINED SOILS
For drained soils, as for undrained, there can be two types of blow-out: a
large-scale passive failure of a block of soil moving upwards towards the
surface (as described in Leca & Dormieux, 1990), or an escape of fluid
(slurry, grout or compressed air) through a fracture.
3.3.1 Passive failure in drained soils

To understand the shape of a passive failure in drained soils, we have to
look at numerical or physical models, because real-world failures are so
Figure 3.14 Blow-out passive failure mechanism in a reduced-scale model of an earth

pressure balance TBM (redrawn from Berthoz et al., 2012).
hard to observe. Berthoz et al. (2012) performed a series of reduced-scale

1 g laboratory tests in moist sand with bulk unit weight γ = 13 kN/m3,
drained cohesion c′ = 0.5 kPa and angle of friction ϕ = 36°. They used
a scale model 0.55 m diameter EPB TBM. It advanced 0.95 m into the
ground and then the face pressure was increased by reducing the extraction
rate of the Archimedes screw while continuing to advance the machine. By
measuring displacements at the surface and within the ground, they found
that the passive failure geometry consisted of rigid blocks as shown in
Figure 3.14. Displacements within the blocks were similar, while displace-
ments outside the blocks were small.
It is clear from Figure 3.14 that we cannot simply adapt Anagnostou &
Kovári’s (1994) wedge and prism model to passive failure as we no longer
have a vertical prism.
Wong et al. (2010, 2012) performed centrifuge tests of a tunnel in satu-
rated sand. The tunnel was impermeable, therefore there was no ground-
water flow. A rigid piston simulated the increase of face pressure. This is
unrealistic in some respects as the face displacement caused by a closed-
face tunnelling machine or an open face with compressed air applied will
always be pressure-controlled, not displacement-controlled. Figure 3.15
shows displacement vectors from their tests.
Figure 3.15 shows very different displacement vectors at C/D = 2.2 and
C/D = 4.3. A very similar pattern of vectors was found in numerical mod-
els of the centrifuge tests (Wong et al., 2012). At C/D = 2.2, the failure is
directed upwards towards the nearby ground surface, whereas at C/D = 4.3,
the failure is localised in front of the tunnel face, with very little effect
on the more distant ground surface except some slight settlement. Berthoz
et al. (2012) found that at C/D = 2, when surcharge of 50 kPa was added
to the ground surface, there was no heave of the surface at all during pas-
sive failure, and in fact there was settlement. In this case, the passive failure
was localised to horizontal displacements away from the tunnel face in the
Figure 3.15 Displacement vectors at face displacement of 0.5D in a centrifuge test for

C/D = 2.2 and C/D = 4.3, from Wong et al. (2012).
direction of tunnelling, the same pattern found by Wong et al. (2012) for
C/D = 4.3 on the right-hand side of Figure 3.15. One could consider the
surcharge as, in effect, making the tunnel seem to be deeper (as though C
were increased). In terms of vertical stress, 50 kPa surcharge in Berthoz
et al.’s reduced-scale model is the equivalent of increasing the cover from
1.1D to approximately 8D. This indicates that passive failure can have dif-
ferent forms with shallow and deep cover or with varying levels of sur-
charge pressure.
Figure 3.15 also shows an upper-bound kinematic mechanism proposed
by Soubra (2000). The experimental and numerical studies of Berthoz et al.
(2012) and Wong et al. (2010, 2012) both show that the geometry of failure
found in kinematic analyses of passive failure often includes a much greater
volume of soil than that found in practice. Berthoz et al. (2012) argue that
this may be because the kinematic analysis assumes a failure at much higher
displacements than is practical. Alternatively, Wong et al. (2012) suggest
that the assumption of associative plasticity (sometimes called the ‘normal-
ity condition’, where the angle of dilation is equal to the critical state angle
of friction ϕcs) in the kinematic analysis may be the cause of the difference.
As described earlier in this chapter in Section 3.1.1, there are two plastic-
ity limit states, the upper bound and the lower bound. The lower bound
is based on a statically admissible stress state and gives a limiting value
of face pressure at which the face cannot fail, that is definitely safe. The
upper bound is based on a kinematic mechanism, and gives a value of face
pressure at which the face will definitely fail, but it is an unsafe predic-
tion because it could fail at a lower value. A handy diagram is provided by
Berthoz et al. (2012), reproduced in Figure 3.16.
Kinematic analysis usually requires partial definition of the geometry
of failure, usually with one or two parameters that need to be optimised.
Many researchers have attempted to provide realistic kinematic mecha-
nisms for collapse and blow-out of tunnels in drained soils, though there
has been less success with blow-out than with collapse. The aim is to make
Figure 3.16 Diagram explaining limit states for a domain of safe face pressures K (redrawn
from Berthoz et al., 2012). σt is the tunnel support pressure, γ is the unit
weight of the soil and D is the tunnel diameter.
the geometry of the kinematic analysis as realistic as possible, so it gives a

close approximation of the tunnel support pressure needed to prevent col-
lapse or blow-out. Then it can become a useful tool for determining the safe
range of face pressures between collapse and blow-out.
Leca & Dormieux (1990) proposed a 3D upper-bound mechanism based
on a single truncated cone, shown in long section in Figure 3.17.
Kinematic analysis assumes the presence of a ‘velocity discontinuity sur-
face’. Within the surface a rigid block is moving and outside of the surface
there is no movement. This seems a reasonable assumption for drained soils
given the displacements presented by Berthoz et al. (2012), shown earlier
in Figure 3.14. The velocity discontinuity surface is at an angle ϕ to the
direction of motion of the block, because this is the angle that offers the
lowest shear resistance, and this is why the shape is conical. The direction
of motion has to be along the axis of the cone. The angle α is then found
at which blow-out occurs most easily, i.e. with the lowest tunnel support
pressure.
The problem with modelling the velocity discontinuity surface as a cone
of circular section is that when it intersects the tunnel face at an angle, the
intersection is an ellipse. This means that there are zones on either side of
Figure 3.17 Long section through single cone kinematic mechanism for blow-out
(redrawn from Leca & Dormieux, 1990). V is the resultant blow-out force
acting on the cone, ϕ is the angle of friction and α is the cone projection
angle.
Figure 3.18 Intersection of a circular cone at an angle with a circular tunnel face (redrawn
from Leca & Dormieux, 1990).
the face which do not move at all, as shown in Figure 3.18. It also may not
reflect the true shape of the failure.
A principle of kinematic analysis is that if the geometry can be changed
to show that failure can occur at a lower tunnel support pressure, then that
new geometry must be more realistic, because we assume that in the real
world failure occurs in the easiest way possible.
In order to better approximate the true geometry of failure, and hence
to move the upper bound closer to the true failure load, several researchers
have added complexity to Leca & Dormieux’s model. Soubra (2000) tried
two truncated cones with a log spiral in between, as shown by the dashed
lines in Figure 3.15. Mollon et al. (2009) used multiple truncated cones,
as shown schematically in Figure 3.19. Each time, results were improved,
i.e. the tunnel support pressure at failure was reduced and could therefore
be assumed to be closer to the true blow-out pressure that might occur in
a real situation (have another look at Figure 3.16). This may be, at least
in part, because the volume of the failure has been reduced (compare the
volume within the velocity discontinuity surface in Figure 3.19 with that in
Figure 3.17), but also because the support pressure is applied to the whole
face area and therefore will provide the same force (load V in Figures 3.17
and 3.19) at a lower tunnel support pressure.
To get over the problem illustrated by Figure 3.18, Mollon et al. (2010)
improved on the multiblock mechanism by applying a spatial discretisation
technique to ensure the intersection of the moving blocks with the tunnel
face matched the whole circular face. Mollon et al. (2011) then abandoned
multiple blocks in favour of a log spiral shape, again using spatial discre-
tisation to ensure the failure intersected with the whole face. Each time,
an incremental improvement in the upper bound was achieved, indicating
that the failure mechanisms were becoming more realistic. Looking back at
the vectors in Figure 3.15 or the sketch in Figure 3.14, and comparing the
Figure 3.19 Multiple truncated cone mechanism for passive failure (redrawn from
Mollon et al., 2009 with permission from ASCE).
shape of the failure with Figure 3.20, we can see that the log spiral seems
to fit the shape far better than the earlier mechanisms proposed by Leca &
Dormieux (1990) or Soubra (2000) (see Figures 3.17 and 3.15 respectively).
A design chart for critical blow-out pressures based on the work of Mollon
et al. (2011) is shown in Figure 3.21 for a cohesionless soil with ϕ = 20°
and ϕ = 40°. As C/D increases, the critical blow-out pressure increases
rapidly, indicating that blow-out to the surface may be a risk for shallow
Figure 3.20 Log spiral mechanism for passive failure (redrawn from Mollon et al., 2011).
Figure 3.21 Design chart based on the log-spiral passive failure geometry for purely
frictional cohesionless soils (redrawn from Mollon et al., 2011). σb is the
blow-out pressure.
tunnels only. This should be compared to collapse of a heading in frictional

soil, which does not depend on C/D.
There does not seem to be a consensus on whether kinematic analysis
gives the right answers. Wong et al. (2010) for some reason do not achieve
passive failure in their centrifuge tests, as a continuous increase of displace-
ment at constant face pressure is not found despite attaining very large
displacements, nor is it found in their numerical model (Wong et al., 2012),
so no firm conclusions can be drawn from their work.
Berthoz et al. (2012) reached passive failure in their experiments at face
pressures 1–2 orders of magnitude lower than predicted using a 2-cone
kinematic mechanism proposed by Subrin (2002). They ascribe this to
the smaller volume of failure and the fact that the very large deforma-
tions needed to mobilise yield on all surfaces could not be achieved in the
experiment. There also must be an error in their kinematic analysis because
Mollon et al.’s (2011) design charts give much lower values than the Subrin
(2002) values, but still higher than Berthoz et al.’s, as shown in Table 3.1.
Using numerical models in FLAC3D, Dias et al. (2008) found that a
blow-out failure occurred only in the upper part of the face (Figure 3.22)
for purely frictional soils, and at a face pressure 40% lower than predicted
using kinematic analysis methods of Leca & Dormieux (1990) and Soubra
(2000). They hypothesised that this mode of failure can occur when the
Table 3.1 Comparison of critical blow-out pressures σ b from Berthoz et al.

(2012) experiments and kinematic analysis using the method
of Subrin (2002) with values from design charts by Mollon et al. (2011).
Berthoz et al. (2012)
Experiment Subrin (2002) Mollon et al. (2011)
Experiment ID σb (kPa) σb (kPa) σb (kPa)
MC1-B1 34
612 168
MC1-B2 10
MC3-B2 21 612 174
MC5-B1 47
515 115
MC5-B3 21
face pressure is uniform, and presumably therefore would be less likely

in an EPB or slurry machine, where face pressure increases from crown
to invert. This would explain the full-face failure found in Berthoz et al.’s
(2012) experiments with a model EPB machine in purely frictional soil.
On the other hand, Mollon et al. (2013b) compare FLAC3D numerical
analyses with their log spiral kinematic mechanism, and find very close
agreement, as shown in Table 3.2, for cohesive-frictional soils. They state
that similar agreement was found for “several cases of frictional soils, with
or without cohesion (not shown in this paper)”. It is unfortunate that these
are not shown in their paper as it would have been interesting to see if the
failure for the purely frictional soils was only in the top half of the face as
found by Dias et al. (2008), and what impact this may have had on the criti-
cal blow-out pressure.
Figure 3.22 Displacement velocity field from FLAC3D model of passive failure, for
C/D = 0.5 in Fontainebleau sand, drained cohesion c′ = 0 kPa, angle of fric-
tion ϕ = 42°, bulk unit weight γ = 15.7 kN/m3 (from Dias et al., 2008 with
permission from ASCE).
Table 3.2 Critical blow-out pressures as found by the M1 kinematic analyses and the
FLAC3D numerical models for D = 10 m and γ = 18 kN/m3, from Mollon
et al. (2013b) with permission from ASCE.
ϕ = 17°, c′ = 7 kPa ϕ = 25°, c′ = 10 kPa
Blow-out pressure Blow-out pressure
(kPa) Difference (kPa) Difference
(M1-FLAC3D)/ (M1-FLAC3D)/
C/D M1 FLAC3D FLAC3D M1 FLAC3D FLAC3D
0.6 682.4 635 7.46% 1112.2 1091 1.94%
0.8 878.6 864 1.69% 1487.3 1521 −2.22%
1 1096.6 1113 −1.47% 1903.5 2004 −5.01%
1.5 1777.1 1842 −3.52% 3301.2 3488 −5.36%
2 2637.7 2740 −3.73% 5213.3 5337 −2.32%
3 5243.1 5253 −0.19% – – –
3.3.2 Blow-outs caused by hydraulic fracturing

Bezuijen & Brassinga (2006) show from field data and centrifuge tests that
bentonite slurry blow-outs can occur at much lower face pressures than
those predicted by finite element or kinematic analysis methods. This is
because these methods do not take account of the fact that slurry is a fluid.
The limit then becomes the pressure at which hydraulic fracturing of the
ground can occur. Generally accepted practice is to limit support fluid pres-
sures to the pore pressure plus the vertical effective stress at the crown of
the tunnel (e.g. Guglielmetti et al., 2008). Bezuijen & Brassinga found the
limit to be approximately the pore pressure plus two to three times the
effective stress for their case.
Broere (2015) also highlights the importance of pore pressures. If the
slurry is infiltrating the pores of the ground, then pore pressures will be
elevated in front of or above the face. This causes a drop in the effective
stress and hence the shear resistance of the ground. Although the infiltra-
tion will apply stabilising seepage forces to the soil grains, and in theory a
vertical face of cohesionless soil can be stable as long as the hydraulic gradi-
ent is greater than 2 (van Rhee & Bezuijen, 1992), a continuous infiltration
cannot be relied on in practice.
Holzhäuser (2003) discusses several types of compressed air blow-outs.
The air pressure is almost always set higher than the pore pressure in the
ground, at least at the crown level. The difference between air pressure
and slurry pressure is that air pressure does not vary with height, because
its density is only 1.225 kg/m3. Therefore, wherever the air can flow to, it
applies approximately the same pressure. The air tends to displace the pore
water, which can dry out and erode flow channels, leading to a sudden
blow-out and loss of air pressure, followed by collapse.
If a less permeable layer is some distance above the crown, the pore pres-
sure in the ground in between can become equal to that in the working
chamber. Thus, the compressed air pressure may act at a level where the
overburden is insufficient to contain it. In this case, blow-out can occur
due to hydraulic fracturing even though the applied compressed air pres-
sure is lower than the full overburden pressure at the crown level. This is
referred to as a ‘gasometer’ blow-out (Holzhäuser, 2003). This happened
on the Blackwall Tunnel construction under the Thames, reported by Moir
(1897), resulting in flooding of the tunnel by river water. So calculations
of maximum allowable air pressure need to take account of the geology of
the site.
As a minimum, compressed air pressure should be kept below the
overburden pressure. The British Standard for health and safety in tun-
nelling BS 6164:2019 is even more conservative and states that only the
dry density of the soil should be used in the calculation of limiting air
pressure, as the compressed air tends to drive away groundwater and
dry out the soil.
Slurry blow-outs can also occur, for example during construction of the
Second Heinenoord Tunnel where large quantities of bentonite escaped into
the Old Meuse River above the tunnel. Bezuijen & Brassinga (2006) inves-
tigated this event and using TBM records, finite element modelling and
centrifuge modelling they established that it was caused by hydraulic frac-
turing of the sand above the crown of the tunnel, possibly exacerbated by
slurry penetration causing positive excess pore pressures and hence reduc-
ing the effective stress in the soil. Bezuijen & Brassinga came to the conclu-
sion that a fluid escaping through a fracture caused by a strain localisation
needs a smaller face pressure than a passive failure. They found that the
fluid pressure required for a blow-out minus the pore pressure is two to
three times the effective in situ stress in the soil.
Although bentonite slurry produces a filter cake, this is destroyed when
the TBM starts advancing, and then positive pressures can be generated in
the soil in front of and just above the face, reducing the soil’s effective stress
and making a blow-out more likely.
There is also the possibility that a slurry TBM could meet an abandoned
well, a poorly backfilled borehole, or some other underground void, or the
permeability of the ground could suddenly increase. This could cause a
sudden loss of pressure, leading to collapse and/or overexcavation. Also,
if there is a route to the surface, the slurry only needs enough head to
reach the surface and it will flood the local area. After many incidents
of local streets being flooded with bentonite slurry and sinkholes appear-
ing along the route of the Singapore Circle Line and the SMART tunnel
in Kuala Lumpur, Malaysia (CEDD, 2012), ‘variable density’ TBMs have
been developed, which allow a higher density slurry to be used in the head
during excavation (Bäppler et al., 2017). It is then diluted to allow it to be
pumped out of the tunnel for treatment. The higher density means that a
Figure 3.23 Normal density and higher density slurry pressure diagrams.
higher slurry pressure can be used at shallow cover without risk of slurry
reaching the surface, as illustrated in Figure 3.23.
3.3.3 Summary of blow-outs in drained soils

For all types of blow-outs, increasing cover will generally reduce the risk.
An understanding of the ground and the groundwater, and how they inter-
act with the support method, will help predict blow-out scenarios.
Numerical and physical modelling techniques appear to be fairly reliable
methods for analysing collapse or passive failure of a heading. However,
the focus has been on collapse and more research is needed to improve our
knowledge of passive failure and blow-outs in general.
The kinematic analysis methods that have been applied to passive failure
of drained soils have not taken account of groundwater flow, and neither has
much of the numerical or physical modelling. Most theoretical approaches
suffer from the inability to model the actual interaction between the sup-
port fluid and the soil grains. Hydraulic fracturing, and soil layers, lenses
or channels with varying permeability, are all difficult to model, and all
may result in blow-outs at lower face pressures than predicted using cur-
rently available simple models. More research is needed to investigate all
these scenarios and further improve our predictions. Pan & Dias (2016a,
2016b) developed methods of calculating pore pressures due to steady-state
seepage in a finite element model and used them in a kinematic analysis
of collapse. They also looked at the effect of anisotropic permeability. If
they or others apply this method to blow-out in future the results will be
interesting.
Although there seem to be large discrepancies between physical and

analytical models, the results show that passive failure requires quite
large face pressures for tunnels where C/D > 1. On the other hand,
hydraulic fracturing, or pressurised fluid finding an escape route by some
other means, can occur at relatively low face pressures. If the support
fluid pressure at the crown is kept to below the pore pressure plus the
minor principal effective stress, then it is fairly certain that hydraulic
fracturing cannot occur.
To be completely safe from bentonite slurry escaping to the surface,
the face pressure should be kept below the hydrostatic value needed for
a column of slurry to reach the surface. For example, with a slurry unit
weight of 11 kN/m 3, and with a face pressure at the crown of 200 kPa,
the slurry will not reach the surface if the cover is more than 18.2 m,
regardless of whether there is an open route via a poorly backfilled bore-
hole or an abandoned well. Likewise, for a slurry TBM passing beneath a
river, slurry cannot escape into the river if the face pressure is kept below
the hydrostatic value needed for a column of slurry to reach the riverbed
with a pressure higher than the water pressure at the riverbed level. This
is hard, if not impossible, to achieve, since the slurry pressure usually
needs to be set higher than the pore pressure in the ground. In this case,
the properties of the ground between the crown and the riverbed become
of crucial importance as it is the only thing stopping a pollution incident
similar to the one we looked at from the Second Heinenoord Tunnel
(Bezuijen & Brassinga, 2006).
3.4 PIPING
In cohesionless soils below the water table, it is possible for soil and
water to flow into the tunnel due to destabilising seepage forces. This
was discussed in Chapter 1 and is known as ‘piping’. Tunnels have been
flooded by silt and/or sand and water flowing through small gaps or
holes in the lining, leading to total abandonment of the TBM and a com-
plete restart (e.g. in Preston, UK: Thomas, 2011) or costly rescue works
(e.g. Hull wastewater flow transfer tunnel, UK: Grose & Benton, 2005,
2006; Brown, 2004).
The only way to avoid this risk is to pay close attention to the quality of
gaskets, bolts and grommets in segments, and ensure that rings are built
to the specified tolerances and grouted through the tailskin immediately
behind the brush seals. Any evidence of water inflow bringing soil with it
should be dealt with immediately by grouting or sealing. At Hull, joints
between segments may have been opened up by movements of the tunnel
rings relative to the shaft, so any post-construction deformations, caused
by adjacent tunnel construction for example, should be assessed and moni-
tored carefully.
3.5 PROBLEMS
Q3.1. The following question is about slurry TBM face stability.

i. According to the method described in GEO Report 249, what
four values need to be calculated and added together to give
the target minimum slurry pressure at the face? Write down
the equation and explain each term.
ii. Describe qualitatively what the ‘effective support pressure’
is and explain how it is applied to the face of a slurry TBM
during the excavation cycle and, with reference to the ‘wedge
and prism model’, explain how its effect may change as time
elapses during stoppages.
iii. A slurry TBM is to be driven through sandy silt below a lake as
shown in Figure 3.24. Calculate the effective support pressure
required if the saturated unit weight γ = 19 kN/m3, drained
cohesion c′ = 5 kPa and angle of friction ϕ = 25°.
iv. Calculate the target minimum slurry pressure at the crown,
assuming the slurry pressure can be controlled to ±20 kPa and
ignoring the effect of surcharge.
v. From your answer to (iv), what parameters are the most impor-
tant to know accurately before and during the tunnel drive?
vi. If this were a mixshield TBM, what air pressure would be
needed if the slurry level behind the submerged wall were 2.5 m
below the crown? Assume the slurry has a unit weight of
12 kN/m3.
Figure 3.24 Q3.1 Cross-section of slurry TBM tunnel below lake.

Q3.2. A contractor has proposed using an EPB TBM for a 2 km section

of a 5 m ID metro tunnel between two construction shafts. The
bolted precast concrete segmental lining is 250 mm thick. The
overcut diameter of the TBM is 5.7 m. The soil has been char-
acterised as varying from sandy silt to silty clay with an angle of
friction between ϕmin = 20° and ϕmax = 30°, and drained cohesion
c′min = 0 kPa and c′max = 10 kPa. The depth to axis is 20 m and the
water table is at a depth of 5 m below ground level (Figure 3.25).
The saturated unit weight of the soil γ = 20 kN/m3.
i. Calculate the minimum face pressure required to maintain
stability in the drained case, using the method of Anagnostou
& Kovári. Ignore all factors of safety, any allowance for vari-
ability of face pressure and any surface surcharge.
ii. If it were possible to identify an area with the most favourable
ground conditions to be used for a head intervention, show
that it would still not be possible to enter the head without
compressed air. Demonstrate this using calculations for the
drained case.
iii. If dewatering were used to avoid the need for compressed air,
calculate the level the water table needs to be drawn down to.
Figure 3.25 Q3.2 Section through metro tunnel.
Q3.3. A tunnel with a 70 m 2 cross-sectional area is to be constructed

in an urban area using a sprayed concrete lining, at a depth to
the centroid of the cross-sectional area of 19 m as shown in
Figure 3.26. The ground is silty sand with some gravel, with a
bulk unit weight of 18 kN/m3 and a permeability of 10−7 m/s.
The ground has an angle of friction of 25°. The tunnel face is well
above the water table. Assume there is no surcharge.
Figure 3.26 Q3.3 Cross-section of sprayed concrete lined tunnel.
i. If the tunnel is excavated full-face, what value of drained

cohesion is required to provide a factor of safety of 1.0? Use
Anagnostou & Kovári’s nomograms, but explain any assump-
tion you had to make that was unrealistic compared to the real
situation described in the question.
ii. What other analytical (i.e. not numerical) method could be
used to obtain a better answer to (i)?
iii. Describe, with the aid of sketches, two different sprayed con-
crete construction sequences that could be used instead of full-
face, and say how they improve face stability.
iv. Describe, with the aid of sketches, two other toolbox measures
that could be used to improve the face stability, and explain
how they work.
Q3.4. A contractor has proposed using an EPB TBM for a section of
a 7 m ID rail tunnel between two shafts. The bolted segmental
lining is 300 mm thick. The overcut diameter of the TBM is
7.8 m. The soil has been characterised as varying from sandy
silt to silty clay with an angle of friction ϕ = 25° and drained
cohesion c′ = 0 kPa, and an undrained shear strength of
50 kPa. The depth to axis is 16 m and the water table is at a
depth of 3 m below ground level (Figure 3.27). The bulk unit
weight of the soil γ = 20 kN/m 3 both above and below the
water table.
Figure 3.27 Q3.4 Cross-section through rail tunnel.
i. Calculate the theoretical minimum face pressure required to

maintain face stability in the drained case, using the method
of Anagnostou & Kovári. Assume that the piezometric head
of water pressure in the chamber (hf ) is equal to the pre-
existing piezometric head in the ground (h0). Ignore all factors
of safety, any allowance for variability of face pressure and
any surface surcharge.
ii. Demonstrate, with the aid of calculations, the effect of drain-
ing the ground such that the piezometric head of water pres-
sure in the chamber (hf ) is equal to zero. Why might it be
desirable to maintain hf as close to h0 as possible, and how
might this be achieved?
iii. What value of cohesion would the soil need for the face to be
stable without applying any support pressure?
iv. Calculate the theoretical minimum face pressure required to
maintain stability in the undrained case, using Mair’s heading
stability chart. Ignore all factors of safety and ignore any allow-
ance for variability of face pressure and any surface surcharge.
v. How are factors of safety introduced in GEO Report 249
(Golder Associates, 2009) for drained soils at the ultimate
limit state? How would this approach affect the target min-
imum face pressures calculated in (i) and (iv)? Which is the
critical case for design – drained or undrained?
Q3.5. It is proposed to use an EPB TBM for an 8 m ID flood alleviation
sewer between two shafts. The bolted segmental lining is 300 mm
thick. The overcut diameter of the TBM is 8.7 m. The soil has
been characterised as sandy silt with an angle of friction ϕ = 23°
and drained cohesion c′ = 0 kPa. The depth to axis is 16 m and the
water table is at a depth of 2 m below ground level (Figure 3.28).
Assume the bulk unit weight of the soil γ = 20 kN/m3 both above
and below the water table.
Figure 3.28 Q3.5 Cross-section through proposed flood alleviation sewer.
i. Calculate the theoretical minimum face pressure required to

maintain face stability. Assume that the piezometric head of
water pressure in the chamber (hf ) is equal to the pre-existing
piezometric head in the ground (h0). Ignore all factors of safety
and ignore any allowance for variability of face pressure and
any surface surcharge.
ii. Demonstrate, with the aid of calculations, the effect of drain-
ing the ground such that the piezometric head of water pres-
sure in the chamber (hf ) is equal to zero. Why might it be
desirable to maintain hf as close to h0 as possible, and how
might this be achieved?
iii. What value of cohesion would the soil need for the face to be
stable without applying any support pressure?
iv. Assuming that the piezometric head of water pressure in the
chamber (hf ) is equal to the pre-existing piezometric head in
the ground (h0), as calculated in part (i) for sandy silt with
an angle of friction ϕ = 23° and drained cohesion c′ = 0 kPa,
calculate the target minimum face pressure required for a
factored surcharge of 75 kPa and a maximum variability of

face pressure of ±40 kPa.
v. How are factors of safety introduced in GEO Report 249 for
drained soils at the ultimate limit state? How would this approach
affect the target minimum face pressure calculated in (iv)?
vi. How could a head intervention be achieved in this tunnel?
REFERENCES
Anagnostou, G. & Kovári, K. (1994). The face stability of slurry shield-driven tun-
nels. Tunnels and Deep Space 9, No. 2, 165–174.
Anagnostou, G. & Kovári, K. (1996a). Face stability conditions with earth-
165–173.
Anagnostou, G. & Kovári, K. (1996b). Face stability in slurry and EPB shield tun-
nelling. Geotechnical Aspects of Underground Construction in Soft Ground
(eds Mair, R. J. & Taylor, R. N.), pp. 453–458. Rotterdam: Balkema.
Anagnostou, G. & Perazzelli, P. (2013). The stability of a tunnel face with a free
span and a non-uniform support. Geotechnik 36, Heft 1, 40–50.
Atkinson, J. H. & Potts, D. M. (1977). Subsidence above shallow tunnels in soft
ground. Proc. ASCE Geot. Engrg Div. 103, GT4, 307–325.
Bäppler, K., Battistoni, F. & Burger, W. (2017). Variable density TBM – combining
two soft ground TBM technologies. Proc. AFTES International Congress,
Paris, France, 13th–15th November 2017, Paper C3-2.
Berthoz, N., Branque, D., Subrin, D., Wong, H. & Humbert, E. (2012). Face fail-
ure in homogeneous and stratified soft ground: theoretical and experimental
approaches on 1g EPBS reduced scale model. Tunn. Undergr. Space Technol.
30, 25–37. Figures 3.14 and 3.16 reprinted with permission from Elsevier.
Bezuijen, A. & Brassinga, H. E. (2006). Blow-out pressures measured in a cen-
trifuge model and in the field. Tunnelling: A decade of progress – GeoDelft
1995–2005 (eds Bezuijen, A. & van Lottum, H.), pp. 143–148. Leiden:
Taylor & Francis/Balkema.
Broere, W. (1998). Face stability calculation for a slurry shield in heterogeneous
soft soils. Proc. Tunnels and Metropolises (eds Negro Jr., A. & Ferreira, A.
A.), Sao Paulo, Brazil, pp. 215–218. Rotterdam: Balkema.
Broere, W. (2015). On the face support of microtunnelling TBMs. Tunn. Undergr.
Space Technol. 46, 12–17.
Broere, W. & van Tol, A. F. (2000). Influence of infiltration and groundwater
flow on tunnel face stability. Proc. Geotechnical Aspects of Underground
Construction in Soft Ground (eds Kusakabe, O., Fujita, K. & Miyazaki, Y.),
Tokyo, Japan, pp. 339–344. Rotterdam: Balkema.
Broere, W. & van Tol, A. F. (2001). Time-dependant infiltration and groundwa-
ter flow in a face stability analysis. Proc. Modern Tunneling Science and
Technology (eds Adachi, T., Tateyama, K. & Kimura, M.), Kyoto, Japan,
pp. 629–634. Lisse: Swets & Zeitlinger.
BS 6164:2019. Health and safety in tunnelling in the construction industry – code
of practice. London, UK: British Standards Institution.
CEDD (2012). Catalogue of notable tunnel failure case histories (up to November
2012). Prepared by the Mainland East Division Geotechnical Engineering
Office, Civil Engineering and Development Department, The Government
of the Hong Kong Special Administrative Region. Available at: http://www.
cedd.gov.hk/eng/publications/geo/doc/HK%20NotableTunnel%20Cat.pdf
[last accessed 3rd June 2014].
Dias, D., Janin, J.-P., Soubra, A.-H. & Kastner, R. (2008). Three-dimensional
face stability analysis of circular tunnels by numerical simulations. Proc.
ASCE Geo Congress 2008: Characterization, Monitoring and Modeling of
Geosystems, pp. 886–893.
DIN 4126 (1986). Ortbeton-Schlitzwände Konstruktion und Ausführung [in
German – now withdrawn].
EN 1997-1:2004 (2009). Eurocode 7: Geotechnical design – Part 1: General rules,
incorporating corrigendum February 2009. Brussels: European Committee
Golder Associates (2009). Ground control for slurry TBM tunnelling. GEO Report
249. Hong Kong: The Government of the Hong Kong Special Administrative
Region.
Guglielmetti, V., Grasso, P., Mahtab, A. & Xu, S. (2008). Mechanized tunnelling
in urban areas. London: Taylor & Francis Group.
Holzhäuser, J. (2003). Geotechnical aspects of compressed air support on TBM
tunnelling. Engineering and health in compressed air working, pp. 359–371.
Horn, M. (1961). Horizontaler Erddruck auf senkrechte Abschlussflächen von
Tunnelröhren. In Landeskonferenz der Ungarischen Tiefbauindustrie,
Horizontal earth pressure on perpendicular tunnel face. Proceedings of
the Hungarian National Conference of the Foundation Engineer Industry,
Budapest, Hungary, pp. 7–16.
Kilchert, M. & J. Karstedt (1984). Schlitzwände als Tragund Dichtungwände,
Band 2, Standsicherheitberechnung von Schlitzwänden, pp. 28–34. Berlin:
DIN.
clay. Proc. 10th Int. Conf. Soil Mech. & Found. Engrg, Stockholm, Vol. 1,
pp. 319–322.
Krause, T. (1987). Schildvortrieb mit flüssigkeits- und erdgestützter Ortsbrust.
PhD thesis, Technischen Universität Carolo-Wilhelmina, Braunschweig.
Leca, E. & Dormieux, L. (1990). Upper and lower bound solutions for the face
stability of shallow circular tunnels in frictional material. Géotechnique 40,
No. 4, 581–605.
Lovelace, N. (2003). LUL fears for Central Line tunnels after CTRL ground col-
lapse. New Civil Engineer, 13th February 2003. Available at: http://www.nce.
co.uk/lul-fears-for-central-line-tunnels-after-ctrl-ground-collapse/792589.
article [last accessed 3rd June 2014].
nel face collapse in sand. Physical Modelling in Geotechnics (eds Springman,
Moir, E. W. (1897). Contribution to discussion on the Blackwall Tunnel. Minutes of
the Proceedings of the Institution of Civil Engineers vol. CXXX, pp. 80–96.
Mollon, G., Dias, D. & Soubra, A.-H. (2009). Probabilistic analysis and design of
circular tunnels against face stability. Int. J. Geomech. 9, No. 6, November,
237–249.
Mollon, G., Dias, D. & Soubra, A.-H. (2010). Face stability analysis of circular
tunnels driven by a pressurized shield. J. Geotech. Geoenviron. Eng. 136,
No. 1, 215–229.
Mollon, G., Dias, D. & Soubra, A.-H. (2011). Rotational failure mechanisms for
the face stability analysis of tunnels driven by a pressurized shield. Int. J.
Mollon, G., Dias, D. & Soubra, A.-H. (2013b). Range of the safe retaining pres-
sures of a pressurized tunnel face by a probabilistic approach. J. Geotech.
Geoenviron. Eng. 139, No. 11, 1954–1967.
Pan, Q. & Dias, D. (2016a). Face stability analysis for a shield-driven tunnel in
anisotropic and nonhomogeneous soils by the kinematical approach. Int. J.
Geomech. 16, No. 3.
Pan, Q. & Dias, D. (2016b). The effect of pore water pressure on tunnel face stabil-
ity. Int. J. Numer. Anal. Meth. Geomech. 40, 2123–2136.
Soubra, A.-H. (2000). Three-dimensional face stability analysis of shallow circu-
lar tunnels. Proc. Int. Conf. on Geotechnical and Geological Engineering,
Melbourne, Australia, November 19–24, pp. 1–6.
Subrin, D. (2002). Études théoriques sur la stabilité et le comportement des tunnels
renforcés par boulonnage. thèse de doctorat soutenue à l’Institut National
des Sciences Appliquées de Lyon, France.
van Rhee, C. & Bezuijen, A. (1992). Influence of seepage on stability of sandy
slope. ASCE J. Geotech. Eng. 8, 1236–1240.
Vermeer, P. A., Ruse, N. & Marcher, T. (2002). Tunnel heading stability in drained
ground. Felsbau 20, No. 6, 8–18.
Wong, K. S., Ng, C. W. W., Chen, Y. M. & Bian, X. C. (2010). Centrifuge model-
ling of passive failure of tunnel face in saturated sand. Proc. 7th Int. Conf.
Physical Modelling in Geotechnics (eds Springman, S., Laue, J. & Seward,
L.), pp. 599–604. London: Taylor & Francis Group.
Wong, K. S., Ng, C. W. W., Chen, Y. M. & Bian, X. C. (2012). Centrifuge and
numerical investigation of passive failure of tunnel face in sand. Tunn.
Undergr. Space Technol. 28, 297–303. Figure 3.15 Reprinted with permis-
sion from Elsevier.
Chapter 4
Stability of shafts
Unlike a tunnel heading, at the base of a shaft gravity is on our side.

However, seepage forces or constrained water pressures can still cause
instability, as can heave pressures due to time-dependent swelling
of clay.
Shafts can experience instability in four main ways. The first is hydrau-
lic failure of the soil in the base of the shaft due to an upwards seepage of
water pushing the soil grains apart. The second is a base heave failure. The
third is a geotechnical uplift failure of the base of the shaft during excava-
tion caused by water pressure in a confined aquifer acting on the underside
of an impermeable layer of soil. Finally, the fourth is a buoyancy uplift
failure of a shaft after base slab construction, where the whole shaft is
pushed out of the ground by water pressure. These will be described in the
following sections, along with worked examples.
• the types of stability failures that can occur in a shaft and what causes
them
• the importance of groundwater and permeability to shaft stability
• calculate the ultimate limit state for each type of shaft stability failure
according to Eurocode 7
4.1 HYDRAULIC FAILURE IN A SHAFT

DURING EXCAVATION
This is a well-known geotechnical problem for excavations in cohesionless

soils. If the hydraulic gradient is too large at the base of the shaft, then the
upwards seepage of water will cause hydraulic failure to occur, colloquially
known as ‘boiling’ or ‘quicksand’. Hydraulic failure occurs when the seep-
age forces reduce the effective stresses to zero and the soil grains begin to
DOI: 10.1201/9780429470387-4 123
Figure 4.1 Section through a circular dry caisson or underpinned shaft showing ground-
water flow.
move apart. The result is as horrifyingly dangerous as it sounds: a complete

loss of bearing capacity at the base of the shaft.
For a dry caisson or underpinned shaft, there will be one or more sumps
in the base of the excavation to pump out the water, maintaining the
groundwater level just below the excavation level, as shown in Figure 4.1.
This will cause a groundwater flow from the surrounding ground, shown
by the arrows, generating large upwards seepage forces through the soil in
the base of the shaft.
To verify whether hydraulic failure will be an issue, a flow calculation
needs to be made to determine the pore pressures in the ground and/or the
hydraulic gradients. This cannot be done on paper by sketching a flownet
as we might do for a trench supported by sheet piles or an embedded retain-
ing wall, because a circular shaft is not a 2D plane strain problem, but is
either axisymmetric for a circular shaft, or a 3D problem if the shaft is
non-circular (Polubarinova-Kochina, 1962) and requires calculation by a
numerical method.
Consider a unit volume of soil in the base of a shaft excavation, shown
in section in Figure 4.2.
Soil is a porous medium made up of soil particles with water filling the
voids in between. The porosity n is given by:
Vv
n= (4.1)
V
n is the porosity, the fraction of the total volume that is made up of
voids. Therefore, (1 − n) is the fraction of the total volume that is
made up of solid particles.
Stability of shafts 125
Vv is the volume of voids

V is the total volume
There are three forces acting on the soil particles in the unit volume
(Polubarinova-Kochina, 1962), all in units of kN/m3 because they are a
force per unit volume. The first is the weight of the soil particles, acting
downwards. Taking the downwards direction as positive, we get:
F1 = (1 − n ) γ s j (4.2)
F1 is the weight of soil particles in kN/m3

γs is the unit weight of the soil particles in kN/m3
j is the unit vector (0, 0, 1)
The second force is the effect of water pressure on the solid particles:
F2 = (1 − n ) grad u (4.3)
F2 is the water pressure force on the soil particles in kN/m3

u is the pore pressure
grad u is the pore pressure gradient, where ‘grad’ is the gradient of a
( )
scalar field. It is itself a vector, given by: grad u = ∂∂ ux , ∂∂ uy , ∂∂uz
For vertical flow, we could replace grad u with (u 2 – u1)/Δz, where Δz is the
vertical distance across the unit volume and u1 and u 2 are the pore pres-
sures, as shown in Figure 4.2.
Figure 4.2 A volume of soil experiencing an upwards vertical seepage of water.

The third force is the seepage force, the drag effect of the water moving
between the soil particles. It can also be thought of as the resistance of the
soil to the passage of water, and it is this resistance that gives the value of
permeability. It is given by:
F3 = n ( grad u + γ w j) (4.4)
F3 is the seepage force in kN/m3

γw is the unit weight of water in kN/m3
Summing the three forces gives us the resultant seepage force Sz. If positive,
then the soil is stable, if negative then it is boiling.
Sz = F1 + F2 + F3 = (1 − n ) γ s j + (1 − n ) grad u + n grad u + nγ w j (4.5)
S z is the resultant seepage force in kN/m3
Rearranging this gives:
Sz = (1 − n ) γ s j + grad u + nγ w j (4.6)
Now the bulk unit weight can be expressed as:
γ b = (1 − n ) γ s + nγ w (4.7)
Therefore:
Sz = γ b j + grad u (4.8)
For an upwards flow, the value of grad u will be negative, because we

are taking the downwards direction as positive. Therefore, Equation 4.8
says that as long as the bulk unit weight is greater than the upwards pore
pressure gradient, the ground will remain stable. If the resultant force on
the soil particles S z becomes negative, they will move upwards and ‘boil’.
Therefore, it is the pore pressure gradient that is the critical factor.
For design we should apply partial factors according to Eurocode 7 (EN
1997-1:2004+A1:2013) Equation (2.9a), where:
udst ;d ≤ σ stb;d (4.9)
udst ;d is the design value of the pore pressure at the bottom of a column
of soil at the base of the excavation
σ stb;d is the design value of stabilising total vertical stress at the bottom
of a column of soil at the base of the excavation
The ‘design value’ means a characteristic value modified by a partial factor
The Eurocode 7 approach is simplified, in that it assumes vertical flow of

groundwater and that the pore pressure at the base of the excavation is
zero. Since grad u is the gradient of pore pressure and γb is effectively the
gradient of total stress, then we can see that Equations 4.8 and 4.9 are simi-
lar. The difference is that Equation 4.8 is expressed in terms of gradients
and Equation 4.9 is expressed in terms of values of pore pressure and total
stress an arbitrary distance below the base of the excavation, and assuming
they are zero at the base of the excavation.
In Eurocode 7, partial factors need to be applied according to Table A.17
to obtain the design values of actions and resistance. These may be altered
by the National Annex. In the case of the UK they are the same as the rec-
ommended values, as shown in Table 4.1.
Therefore, we should multiply the pore pressure gradient by 1.35 and
multiply the bulk unit weight by 0.9 to obtain the design values. If the
design value of pore pressure gradient is lower than the design value
of the bulk unit weight, then there is sufficient factor of safety against
the ultimate limit state. This can be expressed as the following
inequality:
1.35 × grad u ≤ 0.9 × γ b j (4.10)
It may in some cases be easier to deal with hydraulic gradient rather than
pore pressure gradient. Hydraulic gradient is related to pore pressure gradi-
ent by the equation:
1
grad u = grad h − j (4.11)
γw
grad h is the hydraulic gradient, where h is the groundwater head.

For vertical flow parallel to the z-axis this would be equal to
∆h /∆z .
Table 4.1 Partial factors for use in the hydraulic heave ultimate limit state (HYD)
from Table A.17 of Eurocode 7 (EN 1997-1:2004+A1:2013) and from
Table A.NA.17 of the UK National Annex to Eurocode 7 (NA+A1:2014 to
BS EN 1997-1:2004+A1:2013).
Symbol Table A.17 Table A.NA.17
Permanent Unfavourable γG,dst 1.35 1.35
Permanent Favourable γG,stb 0.9 0.9
Variable Unfavourable γQ,dst 1.5 1.5
Variable Favourable γQ,stb 0 0
The seepage force F3, expressed in terms of hydraulic gradient, is now

given by:
F3 = nγ w grad h (4.12)
The pore pressure force F2 may be given by:
F2 = (1 − n ) [γ w grad h − γ w j] (4.13)
The F1 force remains the same as in Equation 4.2. Using Equations 4.12
and 4.13 instead of Equations 4.3 and 4.4, the expression for resultant
seepage force is:
Sz = F1 + F2 + F3 = (1 − n ) γ s j + (1 − n ) [γ w grad h − γ w j] + nγ w grad h (4.14)
Using Equation 4.7 to simplify gives:
Sz = (γ b − γ w ) j + γ w grad h (4.15)
Equation 4.15 says that as long as the upwards hydraulic gradient multi-
plied by the unit weight of water is less than the submerged unit weight
(γb − γw), then the ground will remain stable.
For design we should apply partial factors according to Eurocode 7 (EN
1997-1:2004, 2009) Equation (2.9b), where:
Sdst ;d ≤ Gstb
′ ;d (4.16)
Sdst ;d is the design value of seepage force, which is the second term on
the right-hand side of Equation 4.15
′ ;d is the design value of submerged weight, which is the first term on
Gstb
the right-hand side of Equation 4.15
The ‘design value’ means a characteristic value modified by a partial factor
The same partial factors in Table 4.1 should be used. Thus, the ultimate
limit state is defined by the following inequality:
1.35 × γ w grad h ≤ 0.9 × (γ b − γ w ) j (4.17)
For soils with cohesion, the design value of cohesion (the characteristic
value reduced by a partial factor) may be added as a stabilising action to the
right-hand side of Equation 4.17.
The risk of hydraulic failure can be mitigated in many different ways, for
example:
• Lengthening the seepage path; this may be achieved by construct-

ing the shaft walls using secant piles or diaphragm walls that extend
a significant distance below the formation level of the excavation.
Increasing the distance over which the head difference acts will reduce
the hydraulic gradient.
• Alternatively, these walls could extend into an impermeable stratum,

providing a complete cut-off. Then the interior of the shaft can be
dewatered prior to or during excavation such that there is no ground-
water flow.
• Dewater the ground in and around the shaft using deep wells. This will
reduce or remove any upwards hydraulic gradient within the shaft.
• Use grouting or ground freezing to increase the cohesion and decrease
the permeability of the soil.
• Keep the shaft flooded so that there is no hydraulic gradient and use
the caisson-sinking method along with a long-reach excavator or grab
from the surface to excavate. This is known as a ‘wet caisson’. The
water can be pumped out once the caisson is well embedded into an
impermeable soil, providing cut-off. If the soil is cohesionless all the
way down, the base slab can be poured underwater using a tremie
and guided by divers, then the water pumped out when it has gained
sufficient strength. Note that using divers introduces new risks and it
is difficult to assure quality of the base slab.
WORKED EXAMPLE 4.1 HYDRAULIC FAILURE IN A SHAFT

A 12 m diameter shaft is to be built using a dry caisson in sand. The
sand has bulk unit weight γb = 18 kN/m3, angle of friction ϕ = 30° and
effective cohesion c′ = 0 kPa. A numerical seepage analysis finds that the
maximum hydraulic gradient near the surface of the excavation in the
shaft is 1.5, in the upwards vertical direction.
Calculate the design value of both destabilising and stabilising forces
and say whether the shaft excavation is within the ultimate limit state for
hydraulic failure according to Eurocode 7.
The design value of the upwards destabilising seepage force is given by:
Sdst ;d = γ G,dst × γ w grad h = 1.35 × 10 × 1.5 = 20.25 kN/m3
The design value of the downwards stabilising force due to submerged

unit weight is given by:
′ ;d = γ G,stb × (γ b − γ w ) j = 0.9 × (18 − 10) × 1 = 7.2 kN/m3

Gstb
Since the bulk unit weight γb and the unit weight of water γw both act ver-
tically downwards, then they are multiplied by the resultant of the unit
vector j in the vertical downwards direction, which is equal to 1.
The design value of the destabilising seepage force is greater than the
design value of the stabilising force. Therefore, the ultimate limit state
is exceeded and the shaft cannot be built this way. Either a wet caisson,
embedded walls, dewatering or ground improvement is needed.
WORKED EXAMPLE 4.2 HYDRAULIC

FAILURE IN A DIAPHRAGM WALL BOX
A 10 m deep excavation is made in sand within two parallel diaphragm walls
20 m deep. The sand has bulk unit weight γ b = 17 kN/m3, angle of friction
ϕ = 32° and effective cohesion c′ = 0 kPa. The diaphragm walls are assumed
impermeable and the flownet shown in Figure 4.3 has been created.
Calculate the maximum hydraulic gradient. Then calculate the design
values of destabilising and stabilising forces. Does the design pass the
ultimate limit state for hydraulic failure according to Eurocode 7?
Figure 4.3 Worked Example 4.2 diaphragm wall flownet (based on an example in

Powrie, 1997).
From Figure 4.3, the maximum hydraulic gradient will be close to the
diaphragm wall and is a drop in head of 4 m over a vertical distance of
10 m. Therefore, the hydraulic gradient is 0.4.
The design value of the upwards destabilising seepage force is given by:
Sdst ;d = γ G,dst × γ w grad h = 1.35 × 10 × 0.4 = 5.4 kN/m3
The design value of the downwards stabilising force due to submerged

unit weight is given by:
′ ;d = γ G,stb × (γ b − γ w ) j = 0.9 × (17 − 10) × 1 = 6.3 kN/m3

Gstb
′ ;d ≥ Sdst ;d , it is within the ultimate limit state and the excavation

Since Gstb
will not fail.
4.2 BASE HEAVE FAILURE OF A SHAFT IN CLAY
In a similar manner to tunnel heading stability in clay, covered in

Chapter 2, the base of a shaft may fail if the shear stresses along failure
planes exceed the undrained shear strength. The ground around the shaft
moves downwards and the ground within the shaft moves upwards, with a
slip circle around the bottom of the shaft lining. This failure is resisted by
the undrained shear strength of the ground. Any surcharge on the ground
surface around the shaft will be unfavourable, and any surcharge within
the shaft, for example provided by compressed air, will be favourable.
Terzaghi (1943) considered the base heave stability of excavations, but
only in plane strain, i.e. assuming an infinitely long excavation. Bjerrum &
Eide (1956) were the first to consider rectangular, square and circular exca-
vations. Although not presented in this way in their paper, we can define
stability ratio in the same way as for heading stability:
γ H + q − σ ca
N= (4.18)
cu
N is the stability ratio, which is dimensionless
γ is the bulk unit weight of the clay in kN/m3
H is the depth of the shaft in m
q is a surcharge on the ground surface around the outside of the shaft
in kPa
σca is compressed air pressure within the shaft, if provided, or any per-
manent surcharge within the shaft, in kPa. This is favourable and
so should only be counted if permanently applied.
For a given ratio of shaft depth H to shaft diameter D, there is a critical

stability number Nc at which the base will fail. Note that Bjerrum & Eide
(1956) also gave Nc values for strutted excavations that are rectangular in
plan, in which case Nc also depends on the ratio of width to length. In this
section we will only consider circular shafts.
Values of Nc may be read from a design chart (Figure 4.4) and are iden-
tical to those provided by Skempton (1951) for the bearing capacity of
footings, because the situations are analogous – the deep excavation is
considered to be like a negatively loaded perfectly smooth footing (Bolton
et al., 2008). Bjerrum & Eide (1956) proved this method to be sufficiently
accurate for design purposes by reference to 14 case studies of failures or
partial failures. For the seven complete failures they studied, the factor of
safety was estimated to be between 0.82 and 1.16.
In exactly the same way as for tunnel heading stability, the factor of
safety is the ratio N c /N . For design according to Eurocode 7 (EN 1997-
1:2004+A1:2013), shaft base stability is a ‘GEO’ limit state. Therefore, the
factor of safety needs to be greater than 1.4.
Figure 4.4 Critical stability number for shaft base stability in clay.
WORKED EXAMPLE 4.3 – BASE HEAVE

FAILURE OF A SHAFT IN CLAY
A 10 m deep 6 m diameter shaft is to be excavated in soft clay with an
undrained shear strength of 30 kPa and a bulk unit weight of 19 kN/m3.
What is the factor of safety?
What is the maximum depth the shaft can be excavated with a factor
of safety of 1.4?
First, we need to calculate the stability ratio:
γ H + q 19 × 10 + 0 190
N= = = = 6.33
cu 30 30
The ratio of depth to diameter, H /D = 10/6 = 1.67, therefore from the
design chart in Figure 4.4, N c = 8.2.
Therefore, the factor of safety is given by:
Nc 8.2
= = 1.30
N 6.33
This is insufficient for a design to Eurocode 7, which requires a factor
of safety of 1.4. The maximum depth may only be found by trial and
error, because each time the depth is changed, both N and Nc change.
For a depth of 9 m:
19 × 9
N= = 5.7
30
And for H /D = 9/6 = 1.5, N c = 8.1. Therefore, the factor of safety is given by:
N c 8.1
= = 1.42
N 5.7
Thus, a depth of 9 m is an acceptable design.
Cai et al. (2002) used axisymmetric finite element modelling to investigate

base heave failure of circular shafts in soft clays. They found that results
were not affected by the thickness of the clay below the excavation, as long
as the thickness was greater than 2R, where R is the radius of the shaft. If
a hard stratum were closer to the base of the excavation than this, stability
would be improved.
Cai et al.’s finite element analysis results are shown in Figure 4.4 for stiff
walls with no embedment below excavation level, in a homogeneous clay
layer with constant undrained shear strength. We can see that Bjerrum &
Eide’s method gives more conservative, i.e. lower, values of Nc.
Cai et al. (2002) also looked at embedment of diaphragm walls or secant
piles below the excavation level. As the ratio of embedment depth to exca-
vation depth increased, there was a linear increase in critical stability ratio
Nc. If we define a modification factor λw based on their results to modify Nc
to account for wall embedment, it can be given by:
hw
λw = 1 + 0.435 (4.19)
H
λw is the modification factor for wall embedment below excavation level
hw is the wall embedment depth below excavation level in m
H is the excavation depth in m
WORKED EXAMPLE 4.4 – BASE HEAVE FAILURE

OF A DIAPHRAGM WALL SHAFT IN CLAY
A 20 m deep 20 m diameter shaft is to be excavated in clay with an und-
rained shear strength of 70 kPa and a bulk unit weight of 20 kN/m3 using
diaphragm walls with an embedment depth of 10 m below the excavation
level. An allowance for crane loading of 10 kPa surcharge should be made.
What is the factor of safety?
First we need to calculate the stability ratio:
γ H + q 20 × 20 + 10 410
N= = = = 5.86
cu 70 70
The ratio of depth to diameter, H /D = 20/20 = 1.0, therefore from the
design chart in Figure 4.4, N c = 7.6.
If the shaft walls had zero embedment, the factor of safety would be
7.6/5.86 = 1.30. This would not be sufficient if we were designing to
Eurocode 7.
Using Equation 4.19 to calculate the modification factor for wall
embedment depth:
hw 10
λw = 1 + 0.435 = 1 + 0.435 = 1.22
H 20
Therefore, the modified critical stability ratio is:
N c* = λw N c = 1.22 × 7.6 = 9.3
Thus the factor of safety against base heave failure is:

N c* 9.3
= = 1.59
N 5.86
4.3 UPLIFT FAILURE IN A SHAFT DURING

EXCAVATION
If there is an impermeable layer of soil overlying a permeable layer with

pore pressure present, it is possible that as excavation progresses, there
may come a point where the weight of the overlying impermeable layer is
less than the pore pressure in the permeable layer beneath. If this happens,
then the water will cause an uplift failure, pushing the impermeable soil
upwards.
The risk of uplift failure may be mitigated by reducing the pore pressure
in the underlying permeable soil by pumping from wells drilled into it.
Sometimes pumping may not be necessary and passive relief wells may be
used with water channelled to the excavation sump.
4.3.1 Verification of the uplift ultimate limit

state using Eurocode 7
Eurocode 7 (EN 1997-1:2004+A1:2013) provides partial factors for
uplift failure, which are given in Table 4.2 in the third column headed
‘Table A.15’. National Annexes to Eurocodes may give different guidance,
and the partial factors given by the UK National Annex (NA+A1:2014 to
Table 4.2 Partial factors for use in the uplift ultimate limit state (UPL) from
Table A.15 of Eurocode 7 (EN 1997-1:2004+A1:2013) and from
Table A.NA.15 of the UK National Annex to Eurocode 7 (NA+A1:2014 to
BS EN 1997-1:2004+A1:2013).
Symbol Table A.15 Table A.NA.15
Permanent Unfavourable γG,dst 1.0 1.1
Permanent Favourable γG,stb 0.9 0.9
Variable Unfavourable γQ,dst 1.5 1.5
Variable Favourable γQ,stb 0 0
BS EN 1997-1:2004+A1:2013) are given in the fourth column of Table 4.2,

headed ‘Table A.NA.15’. Note that for this limit state, the UK National
Annex requires a different value for the permanent unfavourable actions.
The ultimate limit state is verified by comparing design values of stabilis-
ing and destabilising forces, using Equation (2.8) in Eurocode 7. The design
value of stabilising forces needs to be equal to or higher than the design
value of the destabilising forces, as follows:
Vdst ;d ≤ Gstb;d + Rd (4.20)
Vdst;d is the design value of the destabilising vertical force

Gstb;d is the design value of the stabilising vertical force
Rd is the design value of any additional resistance to uplift
The stabilising force in this case is the weight of the lower permeability
layer plus any surcharge acting at the base of the excavation. The design
value is calculated by multiplying the total vertical stress at the base of
the lower permeability layer by the partial factor for permanent favourable
actions in Table 4.2.
The destabilising force is the groundwater pressure at the base of the
lower permeability layer. The design value is calculated by multiplying the
groundwater pressure by the partial factor for permanent unfavourable
actions in Table 4.2.
WORKED EXAMPLE 4.5 UPLIFT FAILURE

DURING SHAFT EXCAVATION
A 20 m deep 10 m ID shaft is excavated using the underpinning method,
entirely in clay, as shown in Figure 4.5. The shaft is to be built in the UK
according to the UK National Annex to Eurocode 7. The clay stratum
extends 5 m below the formation level of the excavation and below that
the ground is sandy gravel. The groundwater pressure is hydrostatic with
a piezometric level 10 m below the ground surface. The clay has a bulk
unit weight of 20 kN/m3 and a characteristic value of undrained shear
strength of 60 kPa.
Calculate the pore pressure at the base of the clay. Then calculate the
design values of destabilising and stabilising forces. Does the design pass
the ultimate limit state for uplift failure according to Eurocode 7?
Figure 4.5 Worked Example 4.5 underpinned shaft at risk of uplift failure during
excavation.
The destabilising force is the groundwater pressure at the base of the

clay. From Figure 4.5, the groundwater head at the base of the clay is
15 m. Taking the unit weight of water γw = 10 kN/m3, the groundwater
pressure is given by:
u = γ w h = 10 × 15 = 150 kPa
The design value of the destabilising pressure is given by:
Vdst ;d = γ G,dst × u = 1.1 × 150 = 165 kPa
The stabilising force is the weight of the clay. The total vertical stress at
the base of the clay below the shaft excavation is given by:
σ v = γ b,clay × tclay = 20 × 5 = 100 kPa
The design value of the stabilising stress is given by:
Gstb;d = γ G,stb × σ v = 0.9 × 100 = 90 kPa
Since Vdst;d > Gstb;d the shaft fails in uplift.

Further question: At what excavation depth will the shaft fail the uplift
ultimate limit state?
Let z be the depth of excavation. The thickness of the clay will be
given by:
tclay = 25 − z
The total vertical stress will therefore be given by:
σ v = γ b,clay × tclay = 20 × ( 25 − z ) = 500 − 20z
The groundwater pressure at the base of the clay is unchanged and so the
destabilising force is also unchanged. The ultimate limit state is reached
when the design value of the stabilising force equals the design value of
the destabilising force:
Vdst ;d = Gstb;d = 165 kPa
Therefore:
165 = 0.9 × ( 500 − 20z )
And:
z = 15.83 m
So the shaft will fail the ultimate limit state for uplift when the excava-
tion has reached 15.83 m depth.
It seems intuitively unlikely that the shaft would fail in uplift with
a 9 m thickness of reasonably strong clay below the base of the exca-
vation. This is because we have neglected the fact that we would need
to mobilise quite large shear surfaces in the clay to make this failure
actually happen. Eurocode 7 allows for including this by using the
term R d in Equation 4.20.
We do not know the geometry of a large-scale uplift failure. A
numerical or physical model could be used to find the most likely
failure geometry, but to my knowledge nothing has been published on
this subject. To start with we could assume a cylindrical failure with
vertical sides equal to the thickness of clay and a diameter equal to the
internal diameter of the shaft.
The value of shear resistance on this surface would be equal to the
characteristic value of undrained shear strength multiplied by the surface
area of the cylinder. This could be added to the weight of the cylinder to
give the total stabilising force, as shown in Figure 4.6.
Figure 4.6 Worked Example 4.5 Shaft section showing forces for uplift failure.
The additional resistance due to shear is given by:
cu 60
Rd = π Dtclay = π × 10 × 5 × = 6732 kN
γ cu 1.4
where γcu is the partial factor for undrained shear strength resistance
from Table A.16 of Eurocode 7, and is equal to 1.4.
In this case we need to calculate Vdst;d and Gstb;d as forces rather than
pressures.
 π D2 
Vdst ;d = γ G,dst × u  = 1.1 × 150 × 78.54 = 12959 kN
 4 
 π D2 
Gstb;d = γ G,stb × σ v  = 0.9 × 100 × 78.54 = 7069 kN
 4 
The combined value of Rd + Gstb;d is now greater than Vdst;d, so the design
passes the ultimate limit state and is acceptable.
4.3.2 Geometry of uplift failure during excavation

As was mentioned in Worked Example 4.5, the geometry of uplift failure
is unknown. The simplest shape would be a cylinder, and this would be
consistent with the ‘normality condition’, where for an undrained soil with
ϕ = 0, the failure surfaces should be parallel to the direction of motion.
However, it is known that for tunnel heading collapse, the failure does
not involve the motion of rigid blocks bounded by a velocity discontinuity
surface, but is best described as a continuous deformation of the soil.

Mollon et al. (2013a) modelled collapse and blow-out of a tunnel and
found that the geometry of the plastic zone was not cylindrical, but torus
shaped. Therefore, the assumption of a cylindrical block failure is probably
not accurate. More research is needed to investigate this.
4.4 UPLIFT FAILURE OF A SHAFT AFTER

BASE SLAB CONSTRUCTION
Sometimes groundwater is drained to a sump and pumped away during

construction, or the groundwater level is lowered by deep wells in or around
the shaft. Then a watertight reinforced concrete base slab is cast, and once
the concrete has gained sufficient strength, the pumping is stopped or the
pipes are grouted and water pressure will build up under the base slab. If
the upwards force due to the water pressure is larger than the weight of the
shaft structure plus the friction between the shaft lining and the ground,
then the entire shaft could quite simply be pushed out of the ground like a
cork out of a bottle.
In this section we are assuming that the base slab has sufficient structural
capacity to withstand the shear forces and bending moments induced by
the water pressure acting upwards, and that there is a shear connection
between the base slab and the shaft lining.
Similar to the mechanism described in Section 4.2, this is also an uplift
limit state according to Eurocode 7 (EN 1997-1:2004+A1:2013). Partial
factors will be applied as listed in Table 4.2.
The ultimate limit state is verified by comparing design values of stabilis-
ing and destabilising forces, using Equation (2.8) in Eurocode 7. The design
value of stabilising forces needs to be equal to or higher than the design
value of the destabilising forces, as follows:
Vdst ;d ≤ Gstb;d + Rd (4.21)
Vdst;d is the design value of the destabilising vertical force

Gstb;d is the design value of the stabilising vertical force
Rd is the design value of any additional resistance to uplift
Remember that ‘design values’ are determined by applying the partial fac-
tors from Table 4.2 to characteristic values of the forces.
The stabilising force in this case is the weight of the shaft lining and
base slab plus any additional permanent weight within the shaft already
installed at the time of base slab construction, such as intermediate slabs,
roof slab or secondary lining. The design value is calculated by multiplying
the total weight by the partial factor for permanent favourable actions in
Table 4.2.
The destabilising force is the groundwater pressure under the base slab.
The design value is calculated by multiplying the groundwater pressure by
the partial factor for permanent unfavourable actions in Table 4.2.
The additional resistance to uplift is the friction between the shaft lining
and the surrounding ground. This friction is similar to shaft friction of a
bored pile, and so a similar methodology can be used.
Eurocode 7 (EN 1997-1:2004+A1:2013) only allows pile design based
on static pile tests, or on the results of static pile tests ‘in comparable situ-
ations’. Unless we have pile test data available, the best we can do for most
shaft designs is to make a cautious estimate of the friction based on empir-
ical knowledge of bored pile shaft friction in tension in similar ground
conditions and depths, and the soil parameters obtained from site inves-
tigation. This is a large subject area, and depends greatly on the type and
quality of site investigation data available and local experience of pile test-
ing. You will need to do some research. The simplest method is described
in this chapter, based on soil shear strength data from laboratory tests, but
this should be used only for preliminary design.
Depending on the construction method of the shaft, the interface between
the lining and the ground will be disturbed or softened to a greater or
lesser degree. A review of shaft sinking methods can be found in Allenby
& Kilburn (2015). Underpinned shafts may cause more unloading of the
ground, but are usually grouted after installation of each ring, providing
a good and rough contact between the cut ground and the grout. Shafts
underpinned using cast concrete or shotcrete will have a similar interface.
Caisson-sinking usually entails keeping an annulus filled with benton-
ite slurry between the lining and the ground until the required depth is
reached, at which point the annulus is grouted. Therefore, the ground at
the perimeter may have been remoulded by the cutting edge and softened
by the slurry. Note that shafts are larger and shaft sinking invariably takes
longer than installation of a single pile, and so soil softening and distur-
bance will likely be more significant.
In clay soils in the short-term, the shaft friction for bored piles can be
estimated from the following equation:
fs = α cu (4.22)
fs is the shaft friction in kPa

α is a coefficient that depends on the properties of the shaft lining-
ground interface
The value of the interface coefficient α is empirical and can vary from
around 0.3 to 1.0 (Craig, 1997). A typical value for a bored pile in stiff fis-
sured clays is 0.45 (Skempton, 1959; Fleming, 1997).
An effective stress approach can also be used in clay, assuming that

excess pore pressures in the ground close to the shaft lining may dissipate
in the timescale of construction, and is always used in drained soils such as
silts, sands and gravels.
fs = Ksσ v′ tanδ (4.23)
K s is the coefficient of earth pressure at the shaft-ground interface, i.e.

Ksσ v′ is the effective horizontal stress
σ v′ is the effective vertical stress
δ is the angle of friction of the shaft-ground interface
The effective vertical stress is usually assumed to be the in situ value prior
to construction. If the groundwater level has changed, this should be taken
into account.
The coefficient of earth pressure at the shaft-ground interface can be
calculated using Jaky’s formula for normally consolidated soils:
Ks = 1 − sinφ ′ (4.24)
ϕ′ is the soil’s drained angle of friction
For overconsolidated soils, a modified version of this formula may be used:
Ks = (1 − sinφ ′ ) OCRsinφ ′ (4.25)
OCR is the soil’s overconsolidation ratio, which is the maximum value

of effective stress experienced by the soil in the past divided by
its present value
Sometimes pile designers use the normally consolidated value even for over-
consolidated soils (Powrie, 1997). This provides a conservative value for
overconsolidated soils and would allow for stress relief and remoulding of
the soil at the pile perimeter.
The angle of friction of the shaft-ground interface δ will depend on
the properties of the interface, and some judgement may be required
to take into account the method of shaft construction. For most bored
piles in cohesionless soil, δ can be assumed to be approximately equal to
the angle of friction ϕ ′ of the soil. Powrie (1997) recommends that this
should be the critical state value rather than the peak value. In clays, δ
may be lower, and has been taken as 12° to 16° for London Clay, for
example (Fleming, 1997).
WORKED EXAMPLE 4.6 UPLIFT FAILURE OF A

SHAFT AFTER BASE SLAB CONSTRUCTION
A 12 m deep 15 m ID shaft is excavated using the caisson-sinking method
in sandy gravel, as shown in Figure 4.7. The shaft is to be built in the UK
according to the UK National Annex to Eurocode 7. The shaft lining is
350 mm thick and the base slab is 2 m thick and both are reinforced con-
crete. The unit weight of the reinforced concrete is 24 kN/m3.
The groundwater pressure is hydrostatic with a long-term maximum
expected groundwater level at 2 m below the ground surface. During
caisson sinking, the sandy gravel is dewatered using deep wells around
the shaft perimeter, to a level below the base slab formation level. A 2 m
thick reinforced concrete base slab is then cast and allowed to gain suf-
ficient strength before the dewatering wells are turned off.
The sandy gravel is normally consolidated, has a bulk unit weight of
19 kN/m3 and effective shear strength parameters c′ = 0 and ϕ′ = 35°. These
are the characteristic mean values determined from a series of tests from
0 to 12 m depth, in three boreholes outside the shaft perimeter.
Calculate the pore pressure under the base slab. Then calculate the
design values of destabilising forces, stabilising forces and additional
resistance to uplift due to friction. Does the design pass the ultimate limit
state for uplift failure according to Eurocode 7?
Figure 4.7 Worked Example 4.6 shaft at risk of uplift failure after base slab
construction.
Note that for this design to work, the base slab needs sufficient bend-
ing moment and shear capacity to withstand the groundwater pressure,
and there needs to be shear transfer between the base slab and the shaft
lining, provided by shear keys, dowels, or by excavating the base slab
under the last ring.
The destabilising force is the groundwater pressure at the formation

level of the base slab. From Figure 4.7, the groundwater head at this level
is 10 m. Taking the unit weight of water γw = 10 kN/m3, the groundwater
pressure is given by:
u = γ w h = 10 × 10 = 100 kPa
This time we will use forces rather than stresses. Let A be the area of the
shaft base. The design value of the destabilising force is given by:
2
 15.7 
Vdst ;d = γ G,dst × u × A = 1.1 × 100 × π  = 21295.2 kN
 2 
The stabilising force is the weight of the shaft lining and base slab. This
is given by:
  15  2 

( )
Gstb = γ conc × Vbase + Vlining = 24 ×  π   × 2 + π × 15.35 × 0.35 × 12
  2 
= 24 × (353.4 + 202.5) = 13343 kN
The design value of the stabilising force is given by:
Gstb;d = γ G,stb × Gstb = 0.9 × 13341.6 = 12009 kN
Since Vdst ;d > Gstb;d , we need to consider the additional resistance provided
by friction between the lining and the ground.
It is usual practice to fill an annulus between the shaft lining and the
ground with bentonite slurry during caisson sinking. Once the required
depth has been reached, the bentonite is replaced by cementitious grout.
Therefore, the friction between the grouted lining and the ground should
be similar to a bored pile.
Eurocode 7 (EN 1997-1:2004+A1:2013) only allows pile design based
on static pile tests, or on the results of static pile tests ‘in comparable
situations’ (Frank et al., 2005). To estimate the friction between the shaft
lining and the ground, the best we can do for most shaft designs is to
make a cautious estimate of the friction based on empirical knowledge of
bored pile shaft friction in tension in similar soil, and the soil parameters
obtained from site investigation.
Table A.NA.16 of the UK National Annex to Eurocode 7 (NA+A1:2014
to BS EN 1997-1:2004+A1:2013) specifies partial factors for uplift on
soil parameters and resistances.
According to this table, the tangent of the soil’s angle of friction should
be divided by 1.25.
Also, the rules of A.3.3.2 and A.3.3.3 relating to piles should be
followed, where correlation factors are specified to account for the
number of tests. For mean characteristic strengths based on three pro-
files, the correlation factor ξ3 = 1.42 . This is the UK value, and you
should look up the National Annex of the country the project is in for
the appropriate value.
A.3.3.2 of the UK National Annex also says that where ground test
results are used to calculate characteristic resistances, a model factor
should be applied, equal to 1.4.
The interface angle of friction, for sand and gravel, we will assume is
equal to the soil’s internal angle of friction. Therefore:
tanφ ′ tan35°
tanδ = = = 0.394
ξ3γ s;tγ φ ′ 1.42 × 1.0 × 1.25
The characteristic value of shaft friction qs;i ;k is given by:
qs;i ;k = Ksσ v′ tanδ = (1 − sinφ ′ ) σ v′ tanδ = (1 − sin35° ) × (19 − 10) × 6 × 0.394
= 9.08 kPa
The effective stress σ v′ is taken as the average value over the depth of the
shaft, i.e. at 6 m depth. In a more complex geological situation you may
need to calculate values of shaft friction for slices of the shaft and add
them together.
The design value for stabilising resistance due to shaft friction Rd is
therefore:
As qs;i ;k π × 15.7 × 12 × 9.08

Rd = = = 3840.3 kN
1.4 1.4
The total of design stabilising force and resistance is now:
Gstb;d + Rd = 12009 + 3840.3 = 15849 kN
This is still insufficient to prevent floatation of the shaft, because the

design destabilising force is 21295.2 kN. The weight of the shaft could be
increased by increasing the thickness of the base slab. Alternatively, the
design could be refined by increasing the number of soil tests, and thereby
increasing the characteristic strength and reducing the correlation factor,
or by performing tension pile tests to determine shaft resistance.
If the base slab thickness were increased to 4 m thick, then the ground-
water pressure acting on the underside of the base slab increases to
120 kPa and the design value of destabilising force becomes:
2
 15.7 
Vdst ;d = γ G,dst × u × A = 1.1 × 120 × π  = 25554 kN
 2 
The stabilising force is increased as the weight of the base slab is larger,
and is now:
  15  2 

( )
Gstb = γ conc × Vbase + Vlining = 24 ×  π   × 4 + π × 15.35 × 0.35 × 14
  2  
= 24 × ( 706.9 + 236.3) = 22636 kN
The design value of stabilising force is now:
Gstb;d = γ G,stb × Gstb = 0.9 × 22636 = 20372 kN
The shaft is deeper, so the average vertical effective stress is larger, which
increases the shaft friction:
qs;i ;k = Ksσ v′ tanδ = (1 − sinφ ′ ) σ v′ tanδ = (1 − sin35° ) × (19 − 10) × 7 × 0.394

= 10.60 kPa
The design value for stabilising resistance due to shaft friction Rd is also
increased as the shaft depth is increased to 14 m:
As qs;i ;k π × 15.7 × 14 × 10.60

Rd = = = 5227.1 kN
1.4 1.4
So, now the total of design stabilising force and resistance is:
Gstb;d + Rd = 20372 + 5227 = 25599 kN
This is now greater than the design destabilising force of 25554 kN and
so the design is now adequate.
4.5 LONG-TERM HEAVE UNDER A SHAFT BASE SLAB
During excavation, clay soils in the base of a shaft will be unloaded, i.e.
there will be a decrease in total stress. This unloading will be experienced in
the short-term as a decrease in pore pressure. In the long-term, pore water
will gradually flow from surrounding areas until these negative excess pore
pressures are dissipated. When this happens, the clay will increase in vol-
ume. Under the base of a shaft, this swelling will generate heave pressures,
trying to push the base of the shaft upwards.
There is more than one failure scenario to consider. The base slab may
fail in bending or shear due to the heave pressure. If the base slab does not
fail, it is also possible that the whole shaft could be pushed upwards out
of the ground, which could cause serviceability or structural problems for
utility connections or tunnel junctions, or indeed for any headhouse or
other structure constructed on top of the shaft.
The risk of heave may be mitigated by laying compressible void formers
on the ground before casting the base slab. Then the clay can swell without
applying pressure to the base slab. In large shafts, tension piles may be used
to anchor the base slab.
4.6 SUMMARY OF SHAFT STABILITY
During construction, shafts can experience instability in several ways:
• In drained soils, upwards seepage forces can push the soil particles
apart causing hydraulic failure (also known as ‘boiling’ or ‘quicksand’
or ‘hydraulic heave’), resulting in a complete loss of bearing capacity.
• In undrained soils, the base can fail if the shear stresses exceed the
undrained shear strength along failure surfaces, in a similar manner
to undrained stability failure of a tunnel heading.
• As excavation progresses, a low permeability soil overlying a higher
permeability soil can be pushed upwards due to trapped groundwater
pressure, causing uplift failure.
• If groundwater pressure can build up under a shaft base slab, a shaft may
be pushed up out of the ground. This is another form of uplift failure.
• Long-term heave pressures may build up beneath a base slab due to
swelling of the clay, which may push the shaft upwards. It is likely
that this will be a serviceability issue, but it could in some cases cause
a structural failure.
4.7 PROBLEMS
Q4.1. A sheet pile circular shaft is to be excavated in cohesionless silt

and sand with a bulk unit weight of 17 kN/m3. A seepage analysis
determines that the maximum hydraulic gradient will be 0.45.
i. Calculate the upwards destabilising seepage force.
ii. Calculate the downwards stabilising force.
iii. Apply the appropriate partial factors to the stabilising and

destabilising forces according to Eurocode 7. Is the design
within the ultimate state?
Q4.2. A secant pile wall shaft is to be excavated to 15 m depth, below
the water table, in sand with no cohesion. The sand has a bulk
unit weight of 18 kN/m3.
i. Ignoring the need for factors of safety, what is the maximum
hydraulic gradient that can exist at the base of the shaft before
hydraulic failure (i.e. boiling or piping) will occur?
ii. Including factors of safety for a HYD limit state according to
Eurocode 7, what is the maximum hydraulic gradient allowed?
iii. Describe ways in which the design of the shaft or the construc-
tion method could be altered to reduce the hydraulic gradient.
Q4.3. A 10 m deep 5 m diameter shaft is to be excavated in soft clay with
an undrained shear strength of 35 kPa and a bulk unit weight of
18 kN/m3.
i. Calculate the undrained stability ratio at the maximum depth
of 10 m.
ii. From the design chart in Figure 4.4, find the value of critical
stability number Nc.
iii. Calculate the factor of safety. Does the shaft design meet the
requirements of Eurocode 7 for base heave stability?
iv. Can a 30 kPa surcharge be safely applied around the outside
of the shaft?
Q4.4. A 15 m deep 7.5 m diameter shaft is to be excavated in soft clay
with an undrained shear strength of 30 kPa and a bulk unit weight
of 17 kN/m3.
i. Calculate the factor of safety. Does the shaft design meet the
requirements of Eurocode 7 for base heave stability?
ii. At what depth is a factor of safety of 1.4 exceeded?
iii. The design is changed to a circular sheet pile shaft. What embed-
ment of the sheet piles is required to give a factor of safety of 1.4?
Q4.5. A 12 m diameter 32 m deep underpinned shaft is to be excavated
in stiff clay with an undrained shear strength of 150 kPa and a
bulk unit weight of 20 kN/m3. At 36 m below ground level, the
geology changes from stiff clay to sandy gravel. The pore pressure
distribution is hydrostatic, with a piezometric level 2 m below
ground level.
i. Calculate the pore pressure at the base of the clay.
ii. Calculate the design values of destabilising and stabilising
forces, ignoring the resistance provided by the undrained shear
strength of the clay. Show that the design does not pass the
ultimate limit state for uplift failure according to Eurocode 7.
iii. Calculate the additional resistance to uplift provided by the

undrained shear strength of the clay. Show that the design
still does not pass the ultimate limit state for uplift failure
according to Eurocode 7. Remember that for this calculation
Rd, Vdst;d and Gstb;d need to be calculated as forces rather than
pressures.
iv. Since the shaft is to be used as a storage tank, it can be made
shallower and with a larger diameter, as long as the volume
remains the same. Find the depth at which the design will pass
the UPL ultimate limit state.
Q4.6. A 10 m OD shaft is to be excavated to 40 m depth in stiff fis-
sured clay, at which point a 2 m thick reinforced concrete base
slab will be cast. The ground is dewatered during construction
using deep wells to avoid uplift failure during construction caused
by a permeable stratum below formation level, but the pumps
will be switched off once the base slab concrete has gained its
design strength. The long-term maximum water table will be
at 5 m below ground level with hydrostatic distribution to well
below the shaft bottom. The shaft lining is 400 mm thick. The
clay has an undrained shear strength of 200 kPa and drained
parameters are c′ = 0 and ϕ′ = 22°, obtained from three boreholes.
The overconsolidation ratio, OCR = 2.0. The bulk unit weight is
21 kN/m3.
i. Show that the shaft fails the uplift ultimate limit state in the
long-term, if friction between the shaft lining and the ground
is ignored.
ii. Estimate the shaft friction that is available in the short-term
(using undrained shear strength) and in the long-term (using
drained parameters) and show that the shaft now passes the
uplift ultimate limit state.
REFERENCES
Allenby, D. & Kilburn, D. (2015). Overview of underpinning and caisson shaft-

sinking techniques. Proc. ICE Geotech. Engrg 168, issue GE1, 3–15.
Bjerrum, L. & Eide, O. (1956). Stability of strutted excavations in clay. Géotechnique
6, No. 1, 32–47.
Bolton, M. D., Lam, S. Y. & Osman, A. S. (2008). Keynote Lecture: Supporting
excavations in clay – from analysis to decision-making. Proc. 6th International
Symposium on Geotechnical Aspects of Underground Construction in Soft
Ground, IS-Shanghai, Vol. 1, pp. 12–25. London: CRC Press.
Cai, F., Ugai, K. & Hagiwara, T. (2002). Base stability of circular excavations in
soft clay. ASCE J. Geotech. Geoenviron. Eng. 128, No. 8, 702–706.
Craig, R. F. (1997). Soil mechanics, 6th edition. London: Chapman & Hall.
EN 1997-1:2004+A1:2013. Eurocode 7: Geotechnical design – Part 1: General

rules, incorporating corrigendum February 2009. Brussels: European
Committee for Standardization.
Fleming, W. G. K., Weltman, A. J., Randolph, M. F. & Elson, W. K. (1985). Piling
engineering. Glasgow and London: Surrey University Press.
Fleming, W. G. K. (1997). Pile design and practice notes. Kvaerner Cementation
Foundations Limited, March 1997.
Frank, R., Bauduin, C., Driscoll, R., Kavvadas, M., Krebs Ovesen, N., Orr, T. &
Schuppener, B. (2005). Designers’ guide to Eurocode 7: geotechnical design.
London: Thomas Telford Limited.
Mollon, G., Dias, D. & Soubra, A.-H. (2013a). Continuous velocity fields for col-
lapse and blowout of a pressurized tunnel face in purely cohesive soil. Int. J.
NA+A1:2014 to BS EN 1997-1:2004+A1:2013. UK National Annex to Eurocode
7: Geotechnical design – part 1: general rules, incorporating corrigendum
February 2009. British Standards Institution.
Polubarinova-Kochina, P. Ya. (1962). Theory of ground water movement, trans-
lated from the 1952 Russian edition by J. M. R. De Wiest. Princeton, NJ:
Princeton University Press.
Powrie, W. (1997). Soil mechanics – concepts and applications. London: E & FN
Spon.
Skempton, A. W. (1951). The bearing capacity of clays. Proc. Building Research
Congress, London, Division 1, pp. 180–189.
Skempton, A. W. (1959). Cast-in-situ bored piles in London Clay. Géotechnique 9,
No. 4, 153–173.
Terzaghi, K. (1943). Theoretical soil mechanics. New York, London: John Wiley
& Sons.
Chapter 5
Stability and Eurocode 7
Eurocode 7 (EN 1997-1:2004+A1:2013) does not have a specific section

on the design of tunnels, though there are sections on retaining struc-
tures, spread foundations, pile foundations, anchorages and embankments.
Perhaps a section on tunnels will be added in a future revision.
Face stability is a ‘GEO’ ultimate limit state according to Clause 2.4.7.1
(1)P of Eurocode 7 (EN 1997-1:2004+A1:2013), defined as “failure or
excessive deformation of the ground, in which the strength of soil or rock
is significant in providing resistance”. As we saw in Chapter 4, there are
other forms of instability, such as uplift, which is the ‘UPL’ limit state, and
hydraulic failure, which is the ‘HYD’ limit state, both of which can apply to
shafts or to tunnels. In this chapter we will focus on large-scale face stabil-
ity to pick up on issues relating to the selection of characteristic values and
the selection of partial factors.
The first thing we need to do is decide on the ‘characteristic value’ of each
soil parameter we are using, based on the results of laboratory or in situ
tests on a number of soil samples. There isn’t just one characteristic value
of each geotechnical parameter, each limit state will have its own. So the
characteristic value of a soil parameter selected for design of the tunnel lin-
ing may be different to the characteristic value used for assessing stability.
• how Eurocode 7 ensures stability through the use of characteristic

values and partial factors
• how to select characteristic values of parameters
• the use of statistics and probability theory in geotechnical design
• select an appropriate characteristic value of a geotechnical parameter

for use in design
• calculate the ultimate limit state for each type of stability failure
DOI: 10.1201/9780429470387-5 151
5.1 SIZE OF THE ZONE OF GROUND GOVERNING

THE OCCURRENCE OF THE LIMIT STATE
Characteristic values are defined by Eurocode 0 (EN 1990:2002 +A1:2005)

as the 5th percentile value. This means that the probability of the true
value being lower than the characteristic value is 5% or 0.05. This
applies to all kinds of construction materials, such as concrete or timber.
However, in Eurocode 7, a characteristic value (clause 2.4.5.2, EN 1997-
1:2004+A1:2013) for a geotechnical parameter is defined as a cautious
estimate of the true mean value over the zone of ground governing the
behaviour at the limit state. If statistical methods are used, this means we
need to have a probability of less than 5% that the true mean value over
the zone of ground governing the occurrence of the limit state is lower than
the characteristic value. Therefore, we need to know something about the
probability distribution of the geotechnical parameter, and how it may vary
over the zone of ground we are interested in.
If the limit state involves hydraulic fracturing, or failure of a discrete
block in a fissured clay, then the characteristic value needs to be much more
cautious because failure may depend on a small zone of ground. In effect,
failure may depend on a single value. In this case, the characteristic value
should be the 5th percentile value of the probability distribution for that
parameter. Note that the probability distribution for a soil parameter is not
just a normal distribution defined by the mean and standard deviation of
the test data, it needs to be adjusted to take account of the level of confi-
dence, which is based on the number of data points; this will be covered in
the next section.
The size of the shear surfaces that need to be mobilised at failure due
to undrained or drained stability are usually large and therefore the
value chosen should be somewhere between the 5th percentile value and
the mean of the individual test data, because failure is not governed by
a single value of undrained shear strength or drained shear strength
parameters. If the zone of ground governing the occurrence of the limit
state can be estimated to be very large compared to the ‘scale of fluctua-
tion’, then the mean value with 95% confidence level can be used (see
next section), and we are in effect assuming that the variability within
that zone of ground is similar to the variability in the site investigation
data (Hicks & Nuttall, 2012).
The ‘scale of fluctuation’ may be interpreted as the distance within which
soil properties are largely correlated (Schneider & Fitze, 2011). The deter-
mination of the scale of fluctuation is complex and outside the scope of
routine site investigation campaigns, but there is some research available.
For instance, Phoon & Kulhawy (1999a & 1999b) quote a range of values
of vertical scale of fluctuation between 1 m and 6 m and horizontal scale of
fluctuation between 3 m and 80 m for strength parameters in a variety of
different natural soil types.
Stability and Eurocode 7 153
5.2 CORRECTING FOR CONFIDENCE

IN THE SITE INVESTIGATION
In order to determine the characteristic value of a geotechnical parameter

for a particular limit state, first we need to calculate the mean and stan-
dard deviation of the site investigation test data. Since the site investiga-
tion test data may in some cases be based on a small number of samples,
we need to make allowances for the degree of confidence we have that
our sample size reflects the true variability of the ground we are tunnel-
ling through. Statistical methods exist to quantify this uncertainty and can
make a powerful case for spending more money on site investigation to
reduce uncertainty.
In the following, we will assume that the geotechnical parameters follow
a normal distribution. However, the normal distribution is not always the
best fit to geotechnical data and you should consider whether other dis-
tributions may be more appropriate, for example a Weibull or Lognormal
distribution (Masoudian et al., 2019).
Eurocode 7 (EN 1997-1:2004+A1:2013, clause 2.4.5.2 (4)P) requires
the designer to take account of the type and number of samples used
in the site investigation. The number of samples will affect the level of
confidence in the estimate of the mean and standard deviation. With a
small number of samples of a given soil stratum, we have less confidence
in the calculated mean and standard deviation, and so we need to correct
for this lack of confidence. The ‘Student’ or ‘t’ distribution allows us to
do this. Tables of ‘t’ distribution values may be found quite easily on the
World Wide Web.
For hydraulic fracturing or the case of a localised block failure, we want
to find the 5th percentile value, i.e. we want to know that 95% of the ground
is stronger than the characteristic value. If there were an infinite number of
samples taken in the site investigation (i.e. perfect knowledge of the prob-
ability distribution), the ‘t’ distribution becomes the same as the normal
distribution, and 95% of values will be higher than Xm – 1.645S, where Xm
is the mean and S is the standard deviation of the test data. However, if the
number of samples was 5, effectively the ‘t’ distribution becomes flatter and
wider than the normal distribution and 95% of values will be higher than
Xm – 2.015S. Therefore, with a small number of samples, the characteristic
value of a soil strength parameter will be further away from the mean of the
test data, and the design will therefore be less efficient.
For the case of a large-scale stability failure, we need a 95% confidence
that the true mean value will be higher than the characteristic mean value.
This characteristic mean value may be calculated using the following
expression from Schneider (1999):
f
Xk = X m − S (5.1)
n
Xk is the characteristic mean value

Xm is the mean of the test data
f is a coefficient related to the type of distribution, confidence limits
and number of test values
n is the number of test values (also called ‘sample size’)
S is the standard deviation of the test data
The coefficient f may be obtained from statistical tables of the ‘Student’ or

‘t’ distribution using a one-tail confidence limit of 5%. f has a higher value
when n is small, and decreases exponentially as n increases. Figure 5.1
shows how the value of f / n decreases with the number of test values n.
Schneider (1999) explains how this calculation of the true mean value
can be difficult to achieve using a statistical calculation in practice. If the
number of samples is small, then the high value of f means that the char-
acteristic value is very low – generally much lower than prior knowledge
and experience would suggest is reasonable, and sometimes lower than is
theoretically possible.
Schneider (1999) proposes a simple method, which is widely used in
practice, where the characteristic value is equal to the mean minus half a
standard deviation (this is the grey horizontal line in Figure 5.1):
Xk = Xm − 0.5S (5.2)
with notation as for Equation 5.1
Figure 5.1 Improvement in confidence with increasing no. of samples (Schneider, 1999).

For parameters that vary with depth, multivariate statistics should be

used. This is explained by Bond & Harris (2008), where worked examples
are also provided.
There is often a theoretical lower limit to soil strength parameters, which
can be calculated. The shear strength cannot be exceeded by the shear stress
present in the in situ stress state, usually defined by the coefficient of earth
pressure at rest, K0. If the characteristic value calculated using statistical
methods were found to be lower than the theoretical minimum, then the
theoretical minimum should be used instead. This anomaly is often caused
by assuming that the soil parameter follows a normal distribution, when in
fact it is bounded by a lower limit and would probably be better modelled
by a Weibull or Lognormal distribution. An example of undrained shear
strength test data from a real project is shown in Figure 5.2, where values
below the theoretical minimum can be seen.
Another anomaly caused by the normal distribution assumption is that as
sample size increases, the probability of a rare very high value being found
increases, but because there is a lower limit (e.g. the laboratory or in situ
test cannot provide negative strength values), this causes a trend where as
the sample size increases, the mean also increases (see Figure 5.2 where rare
high values can be seen that if mirrored about the mean would probably be
negative). The normal distribution is symmetric and its standard deviation
and mean are calculated based on all the data, regardless of which side of
the mean they are on. The lognormal distribution, as a counter-example,
is not symmetric. The ‘geometric’ mean and standard deviation provided
by the lognormal distribution will not, on average, vary with sample size.
Therefore, where the site investigation is extensive and there is a relatively
large amount of data, consideration should be given to what probability
distribution functions are used.
5.3 MODELLING SPATIAL VARIABILITY

OF SOIL PARAMETERS EXPLICITLY
Where the zone of ground governing the occurrence of the limit state is
more than 20 times larger than the scale of fluctuation, then a cautious
estimate of the true mean value can be used. Where the zone of ground gov-
erning the occurrence of the limit state is smaller or of similar magnitude to
the scale of fluctuation, then a cautious estimate of the 5th percentile value
can be used. If we are between these two cases, or where we suspect that
failure may follow a path of least resistance, then we may wish to model
spatial variation of parameters explicitly.
Spatial variation can be modelled explicitly by using a Monte Carlo
method to assign random values of the parameters to different zones
within a model. The model is run many times with different randomly
generated variations of parameters. This is sometimes called the ‘random
Figure 5.2 An example of real undrained shear strength test data, sample size n = 66.
field method’. The size of the zones is key – an ‘auto-correlation distance’ is

defined, which is related to the scale of fluctuation, over which parameters
can be considered to remain constant. For slope stability, Hicks & Nuttall
(2012) and Griffiths & Fenton (2007) found that the mean factor of safety
was lower when spatial variation was modelled, as failure surfaces followed
a path of least resistance. This may be due to the different scales of problem
and different scales of fluctuation modelled, and also how the areas mod-
elled lined up with failure surfaces. They used the random field method
to apply values of parameters to a finite element mesh that were spatially
correlated, that is, where the distance between midpoints of the finite ele-
ments determined how much the parameters could vary. This is sometimes
referred to as the ‘random finite element method’ or ‘RFEM’ (Fenton &
Griffiths, 2007). Papaioannou et al. (2009) also found that ignoring spatial
variation led to an overestimate of probability of failure, this time for a
rock tunnel with 1100 m of overburden.
If you are interested in finding out more about probabilistic methods,
start with the overview in Jones (2015). It is inevitable that as the cost of
computing continues to decrease, these methods will become part of rou-
tine design practice.
5.4 APPLYING PARTIAL FACTORS
The design approach and partial factors used depend on the country of
application, but generally speaking, partial factors are applied either to the
strength parameters or to the effect of actions. In the United Kingdom,
Design Approach 1 must be used, as required by the UK National Annex
to Eurocode 7 (NA+A1:2014 to BS EN 1997-1:2004+A1:2013). Design
Approach 1 has two combinations as described in Eurocode 7 clause
2.4.7.3.4.2 (EN 1997-1:2004+A1:2013).
Applying Design Approach 1 to undrained tunnel stability means that the
factor of safety Nc/N must be greater than 1.35 (Combination 1) and the
tunnel face must be stable when the characteristic value of undrained shear
strength is divided by 1.4 (Combination 2). By inspection of Equation 2.1 in
Chapter 2, it is clear that both these combinations will have the same effect,
except Combination 1 requires a factor of safety of 1.35 and Combination 2
requires 1.4. Therefore, Combination 2 will always be the critical case.
Applying Design Approach 1 to drained stability is a little more tricky.
The in situ pore pressure should be well known from the ground investi-
gation, and as long as a cautious estimate of the piezometric level is made
based on seasonal and, where appropriate, tidal variations relevant to
the construction period, no safety factor is required. Therefore, only the
drained shear strength parameters c and ϕ should be factored by dividing
c and tan ϕ by 1.25 according to Design Approach 1 Combination 2 (EN
1997-1:2004+A1:2013).
REFERENCES
Bond, A. & Harris, A. (2008). Decoding Eurocode 7. Abingdon: Taylor & Francis.
EN 1990:2002 +A1:2005 (2010). Eurocode – Basis of structural design, incor-
porating corrigenda December 2008 and April 2010. Brussels: European
EN 1997-1:2004+A1:2013. Eurocode 7: Geotechnical design – part 1: gen-
eral rules, incorporating corrigendum February 2009. Brussels: European
Fenton, G. A. & Griffiths, D. V. (2007). Review of probability theory, random
variables and random fields. Probabilistic methods in geotechnical engi-
neering (eds Griffiths, D. V. & Fenton, G. A.), CISM Courses and Lectures
No. 491, pp. 1–69. New York/Wien: Springer.
Griffiths, D. V. & Fenton, G. A. (2007). The Random Finite Element Method
(RFEM) in slope stability. Probabilistic methods in geotechnical engineering
(eds Griffiths, D. V. & Fenton, G. A.), CISM Courses and Lectures No. 491,
pp. 317–346. New York/Wien: Springer.
Hicks, M. A. & Nuttall, J. D. (2012). Influence of soil heterogeneity on geotechni-
cal performance and uncertainty: a stochastic view on EC7. Proc. 10th Int.
Probabilistic Workshop, Stuttgart, Germany (eds Moorman, C., Huber, M.
& Proske, D.), Mitteilung 67, pp. 215–228. Stuttgart: Institut für Geotechnik
der Universität Stuttgart.
Jones, B. D. (2015). Probabilistic methods. Tunnelling Journal, October/November
issue, 26–29.
Masoudian, M. S., Afrapoli, M. A. H., Tasalloti, A. & Marshall, A. M. (2019).
A general framework for coupled hydro-mechanical modelling of rainfall-
induced instability in unsaturated slopes with multivariate random fields.
Comput. Geotech. 115, paper 103162, 1–12.
NA+A1:2014 to BS EN 1997-1:2004+A1:2013. UK National Annex to Eurocode
7: Geotechnical design – Part 1: General rules, incorporating corrigendum
February 2009. British Standards Institution.
Papaioannou, I., Heidkamp, H., Düster, A., Rank, E. & Katz, C. (2009). Random
field reliability analysis as a means for risk assessment in tunnelling. Proc.
2nd Int. Conf. on Computational Methods in Tunnelling, Ruhr Universität
Bochum, Germany. Aedificatio Publishers.
Phoon, K. -K. & Kulhawy, F. H. (1999a). Characterization of geotechnical vari-
ability. Can. Geot. J. 36, 612–624.
Phoon, K. -K. & Kulhawy, F. H. (1999b). Evaluation of geotechnical property vari-
ability. Can. Geot. J. 36, 625–639.
Schneider, H. R. (1999). Determination of characteristic soil properties. Proc. XII
European Conf. on Soil Mech. & Geot. Engrg – Geotechnical Engineering for
Transportation Infrastructure: Theory and Practice, Planning and Design,
Construction and Maintenance (eds Barends, F. B. J. et al.), Amsterdam, The
Netherlands, pp. 273–281. Rotterdam: Balkema.
Schneider, H. R. & Fitze, P. (2011). Characteristic shear strength values for EC7:
guidelines based on a statistical framework. Proc. XV European Conf. on
Soil Mech. & Geot. Engrg, Athens, Greece, Vol. 4.
Chapter 6
Global design using

analytical solutions
This chapter describes the global design of tunnels and shafts, taking
account of soil-structure interaction, to give estimates of forces in the tun-
nel lining. Details, particularly pertaining to segmental linings, such as
joints, bolts, gaskets and reinforcement, will be dealt with in Chapters 9
and 10, and Chapter 11 will focus on sprayed concrete. You should be
aware though, that details such as joints can have an important effect on
the flexibility of the lining, so this current chapter should not be used in
isolation.
A robust, rational and efficient design process should start with very sim-
ple calculations and gradually introduce layers of complexity. In this way,
the sensitivity of the design to each layer of complexity may be assessed,
errors or bugs in spreadsheets or numerical models easily found and gross
errors avoided.
For example, a simple wished-in-place equilibrium calculation allows
an upper bound to the hoop force to be obtained very quickly. This
may be followed by an analytical solution that includes soil-structure
interaction and will calculate bending moments, and a slightly lower
hoop force. The next stage may be to introduce 3D effects via the con-
vergence-confinement method. All these calculations can be easily done
by hand or in a spreadsheet. Only after all these simple methods have
been exhausted and you have a very good feel for the problem should
you start using numerical analysis in 2D and 3D to introduce fur-
ther layers of complexity, such as nonlinear soil and lining behaviour,
non-circular tunnel shapes and interaction with other tunnels or
structures.
• the limitations of analytical solutions for tunnel design

• the relative importance of soil stiffness, lining stiffness and lining
flexibility to the axial forces and bending moments
• the principles of continuum and bedded beam analytical solutions
DOI: 10.1201/9780429470387-6 159
• calculate simple wished-in-place equilibrium for a circular tunnel

• use the Curtis-Muir Wood solution to calculate axial forces and
bending moments in circular tunnel and shaft linings
6.1 SIMPLE WISHED-IN-PLACE EQUILIBRIUM
As a real tunnel is constructed, there is redistribution of stresses in the

ground around and ahead of the face, and then interaction between the
lining and the ground, such that the stresses transmitted to the lining
are not equal to the initial in situ stresses. If we assume in a design
model that the tunnel lining is ‘wished-in-place’, we ignore this redis-
tribution and interaction and just assume that we can magically mate-
rialise a perfectly stiff and infinitely long tunnel lining into existence
without disturbing the soil. This will always provide an upper bound
to hoop forces in the tunnel lining compared to more sophisticated
analytical or numerical methods, so it can provide a useful ‘reality
check’.
Hoop forces (sometimes referred to as ‘tangential’ or ‘axial’ forces) at any
position in the tunnel lining may be found by (theoretically) cutting the ring
and resolving the internal and external forces in one direction, as shown in
Figure 6.1, using the following notation:
N is the hoop force in kN/m (kN per metre length of tunnel)

σ v , axis is the initial in situ vertical total stress at axis level in kPa
σ h, axis is the initial in situ horizontal total stress at axis level in kPa
Figure 6.1 Wished-in-place equilibrium to calculate (a) hoop force at axis or (b) hoop
force at crown and invert.
Global design using analytical solutions 161
To calculate the hoop force at axis in Figure 6.1(a), the vertical total stress
at axis level should be used, because although the initial in situ stress at
the invert will be higher than at the crown, the tunnel must be in equilib-
rium, so the tunnel will be pushed upwards until the ground pressure on
the crown equals the ground pressure on the invert. This is a simplifica-
tion and not strictly true, because shear stresses between the lining and
the ground can also be generated. Therefore, the vertical stress on the top
of the tunnel can be different to the vertical stress on the bottom of the
tunnel. It is also simple to add in the effect of the weight of the tunnel
lining if you wish.
Figure 6.1(b) could also be made slightly more sophisticated by making
the horizontal total stress increase with depth and applying the average
value above axis level to calculate the crown hoop force and the average
value below axis level to calculate the invert hoop force. Although it might
be tempting to make things more complicated at this point, the whole ratio-
nale for this kind of basic initial analysis is that it may easily be done on
the back of an envelope, or even in one’s head, so it is better to keep things
simple, unless the tunnel is particularly large and shallow. The equilibrium
equations are given by:
N axis = σ v , axis r0 (6.1)
N crown = N invert = σ h, axis r0 (6.2)
Naxis is the hoop force at axis in kN/m

σ v , axis is the vertical in situ stress in the ground at axis level in kPa (note
that 1 kPa = 1 kN/m 2)
Ncrown is the hoop force at the crown (the highest point in the tunnel
lining) in kN/m
Ninvert is the hoop force at the invert (the lowest point in the tunnel lin-
ing) in kN/m
σ h, axis is the horizontal in situ stress in the ground at axis level in kPa
r0 is the external radius of the tunnel (also referred to as the ‘radius to
extrados’) in m
To calculate the hoop stress, the hoop force, which is in kN/m, need only
be divided by the thickness of the lining.
Thus, a very simple calculation provides us with a conservative estimate
of the hoop force in the tunnel lining. Bearing in mind of course that we
have not yet introduced the relaxation of stress in the ground prior to instal-
lation of the lining, or the interaction between the lining and the ground
once the lining is installed, and we have no way of calculating bending
moments in the lining.
WORKED EXAMPLE 6.1 SIMPLE

WISHED-IN-PLACE EQUILIBRIUM, K 0 = 1.0
From the simplified geometry of a tunnel in Figure 6.2, calculate the
vertical and horizontal total ground pressure acting on the tunnel lining,
assuming K0 = 1.0. Note that OD means ‘outside diameter’, i.e. the diam-
eter to the extrados or outside surface of the lining, and ID means ‘inside
diameter’, i.e. the diameter to the intrados or inside face of the lining.
γb is the bulk unit weight of the soil, i.e. of the soil particles and pore
water combined.
Assume the tunnel is ‘wished-in-place’, i.e. there is no relaxation of
stress during construction, there is a perfectly stiff lining so the initial
stresses are applied directly to the tunnel lining, and assume the tunnel
lining is weightless. Calculate the hoop force in the lining.
If the tunnel lining thickness t = 0.3 m, what is the hoop stress per
metre length of tunnel?
Vertical in situ total stress at axis level σ v0 = γ b z0 = 20 × 20 = 400 kPa.

For K0 = 1.0, horizontal in situ total stress at axis level σ h0 = σ v0 = 400 kPa.
To calculate the thrust in the lining at axis, assume buoyancy means
that vertical ground pressures above and below the tunnel equilibrate so
the average is the vertical stress at axis level σ v0 .
Therefore, for static equilibrium, the hoop force at axis is given by:
N axis = σ v0 r0 = 400 × 4 = 1600 kN/m
where r0 is the radius to the extrados of the tunnel.

In the horizontal direction, since σ h0 = σ v0 = 400 kPa at axis level,
N crown = N invert = 1600 kN/m. Alternatively, you could calculate the
average horizontal ground pressure on the top half as 360 kPa and on
the bottom half as 440 kPa, resulting in a hoop force at the crown of
1440 kN/m and at the invert of 1760 kN/m.
At axis, hoop stress is given by the hoop thrust divided by the cross-
sectional area. Since the hoop thrust is already expressed in units per
metre length of tunnel, this is given by:
N axis 1600
σ axis = = = 5333 kPa = 5.33 MPa
t 0.3
For varying horizontal in situ stress with depth, at the crown the hoop
stress would be:
N crown 1440
σ crown = = = 4800 kPa = 4.80 MPa
t 0.3
and at the invert it would be:
N invert 1760 kN/m
σ invert = = = 5867 kPa = 5.87MPa
t 0.3 m
EXTENSION TO WORKED EXAMPLE 6.1

INCLUDING LINING WEIGHT
If the unit weight of the concrete lining γc = 24 kN/m3, what is the radial
ground pressure acting on the crown and invert?
Total weight of concrete ring is given by:

W = 2π rmtγ c = 2 × π × 3.85 × 0.3 × 24 = 174.2 kN/m
where r m is the radius to the centroid of the lining = 3.85 m.

For equilibrium in the vertical direction, the force due to vertical
ground pressure on the top half of the ring, 2r0σ v(crown) , plus the weight
of the ring W is equal to the force due to vertical ground pressure on the
bottom half of the ring, 2r0σ v( invert ):
2r0σ v(crown) + W = 2r0σ v( invert )
Therefore the vertical component of ground pressure on the ring will

always be larger at the invert by the weight of the ring. Assuming the
average σ v0 is the same, then:
W 174.2
σ v(crown) = σ v0 − = 400 − = 389.1 kPa
4r0 16
W 174.2
σ v( invert ) = σ v0 + = 400 + = 410.9 kPa
4r0 16
So, the effect of the ring weight is quite small, and generally, if one
assumes it is weightless (as many more sophisticated methods such as the
Curtis-Muir Wood equations do), the error is not large unless the tunnel
is very large and shallow.
WORKED EXAMPLE 6.2 – SIMPLE

WISHED-IN-PLACE EQUILIBRIUM, K0 ≠ 1.0
From the simplified geometry of a tunnel in Figure 6.3, calculate the ver-
tical and horizontal total stress at tunnel axis level, assuming K0 = 0.7.
Remember that K0 is the ratio of horizontal to vertical effective stress.
The unit weight of the soil above the water table is denoted by γunsat and
below the water table is denoted by γsat.
Assuming the tunnel is wished-in-place and perfectly stiff, i.e. the full
overburden load is acting on the tunnel lining, calculate the hoop force
in the lining at crown and invert level.
The tunnel lining is 200 mm thick concrete. Calculate the hoop stress
at crown/invert and axis.
Vertical total stress at axis level σ v0 is given by:
σ v0 = 5 × 15 + 30 × 20 = 675 kPa
The pore pressure at axis level u0 is given by:
u0 = 30 × 10 = 300 kPa
The vertical effective stress at axis level σ v0

′ is therefore:
σ v′0 = σ v0 − u0 = 675 − 300 = 375 kPa
The horizontal effective stress at axis level σ h0

′ is:
σ h′ 0 = K0σ v′0 = 0.7 × 375 = 262.5 kPa
Therefore the horizontal total stress at axis level σ h0 is:
σ h0 = σ h′ 0 + u0 = 262.5 + 300 = 562.5 kPa
Remember that K0 ≠ σ h0 /σ v0 . It is always K0 = σ h′ 0 /σ v′0.

Using the horizontal total stress at axis level for both the crown and
invert hoop thrust calculation, because r0 is small and z0 is large:
N cr ,inv = σ h0 r0 = 562.5 × 1.5 = 843.75 kN/m
Now calculate the hoop force at axis level:
N axis = σ v0 r0 = 675 × 1.5 = 1012.5 kN/m
At crown/invert the hoop stress is given by:
N cr ,inv 843.75
= = 4.22 N/mm 2
0.2 0.2
At axis level the hoop stress is given by:
N axis 1012.5
= = 5.06 N/mm 2
0.2 0.2
Note that when K0 < 1.0, the hoop force at axis is larger than the hoop
force at the crown or invert. When K0 > 1.0, the hoop force at axis is
smaller than the hoop force at the crown or invert.
6.2 EMPIRICAL METHODS
A first estimate of the expected load in a tunnel lining may be obtained

by considering case histories of lining forces where similar construction
methods have been used in similar geology. Some data were provided in
Section 1.7. Often these data are normalised by dividing by the full overbur-
den pressure (the vertical total stress at the tunnel axis level). Even so, there
are large differences between lining forces measured on different projects,

even in similar geology, and this is because they are heavily dependent on
the construction method.
It may be possible in the future to develop empirical relationships
to allow preliminary estimates of lining forces to be made, but at
present there is insufficient data because lining forces are so rarely
measured.
6.3 SOIL-STRUCTURE INTERACTION USING

THE CURTIS-MUIR WOOD SOLUTION
In order to make the calculation more realistic, soil-structure interaction

may be introduced by using an analytical solution. These solutions allow
the tunnel lining to deform elastically due to axial compression, flexure
and shear stresses between the lining and the ground. This means that
some load sharing with the ground is allowed and bending moments in the
lining may be calculated. However, relaxation of the ground stresses prior
to installation of the lining is not taken into account, unless of course we
were to make some assumptions and reduce the in situ stresses we input to
the model.
An ‘analytical solution’, or ‘closed-form solution’, is a calculation that
may be performed by hand (or in a simple spreadsheet) without iteration.
These methods are usually simple and based on elastic constitutive models,
homogeneous materials and simple boundary conditions. Even if numerical
modelling is going to be employed, an analytical solution may be used to
verify that the numerical modelling program is functioning correctly (see
Section 7.7 in the following chapter).
In order to reduce the number of variables, several assumptions are made.
The tunnel is circular. The ground is homogeneous and isotropic, meaning
that layered soils cannot be considered, and the elastic soil parameters must
be the same in the horizontal and vertical directions.
In this book, we are going to look at the Curtis-Muir Wood solution
(Curtis, 1974; Muir Wood, 1975; Curtis, 1975). Other solutions exist,
most of which vary only slightly in their composition or notation, and
either give exactly the same results or a negligible difference. A summary
of solutions available at the time was provided by Duddeck & Erdmann
(1985), noting their differences. Einstein & Schwartz (1979) later pro-
vided a ‘simplified analysis’ to achieve the same result as the Curtis-Muir
Wood solution, which is not really any simpler (if you look this one up,
be sure to also read the lively discussion!). Ahrens et al. (1982) did a
thorough study of both the continuum solution and a bedded beam solu-
tion, and these methods were adopted in the German recommendations
at the time (Duddeck, 1980).
The original analytical solution published by Sir Alan Muir Wood in

1975 (Muir Wood, 1975) contained some errors and an unnecessary sim-
plification, which were corrected by John Curtis soon after (Curtis, 1975),
who also extended the method to include time-dependent deformation of
the ground and lining (i.e. viscoelasticity; Curtis, 1974). Hence the cor-
rected analytical solution is commonly referred to as the ‘Curtis-Muir
Wood’ solution.
6.3.1 Notation used in the Curtis-Muir

Wood solution
Table 6.1 lists the notation used in the Curtis-Muir Wood solution. I have
mostly used the same notation as Curtis (1974), but please be aware that
it differs in some ways from notation used in the rest of this book. It is
intended to be optimal for those of you who wish to refer to the original
papers.
Table 6.1 Notation used in the Curtis-Muir Wood analytical solution.

c As suffix to denote ‘of the ground’
d As suffix to denote ‘due to distortional loading’
E Young’s modulus for the lining
Ec Young’s modulus for the ground
I Second moment of area of lining per unit length of tunnel
Ie Effective value of I for a jointed ring (i.e. a segmental lining)
l As suffix to denote ‘of the lining’
M Bending moment in lining per unit length of tunnel
N Hoop thrust in lining per unit length of tunnel
n Number of segments in a ring of lining (also as suffix to denote
‘normal’)
P0 Distortional stress (see Equation 6.3)
Pu Uniform stress (see Equation 6.60)
r Radius
rm Radius to the centroid of the tunnel lining
r0 Radius to extrados of tunnel lining
Rc or Q1 Compressibility factor (see Equation 6.61)
Sn , St Normal and tangential (shear) stresses
t Effective thickness of lining (also as suffix to denote ‘tangential’)
u Radial movement of the ground (also as suffix to denote ‘due to
uniform loading’)
u0 Deformation of the ground in the absence of a lining at r = r0
u Deformation of the ground inhibited by the presence of the lining
ul Deformation of the lining
u0θ Circumferential movement of ground at r = r0
(Continued)
Table 6.1 Notation used in the Curtis-Muir Wood analytical solution (Continued).
η Ratio of radius of lining centroid to that of extrados
θ Angle from crown (also as suffix to denote circumferential
direction)
ν Poisson’s ratio for the ground
νl Poisson’s ratio for the lining
σ v , σ h Pre-existing vertical and horizontal stress in undisturbed ground
σr , σθ , σz , τrθ Radial, tangential and out-of-plane normal stress and shear stress
in the ground using polar coordinates
εr , εθ , εz Radial, tangential and out-of-plane strain in the ground using
polar coordinates
6.3.2 Boundary conditions and ground stresses

The method is based on assuming the ground extends infinitely around the
tunnel. The surface is not modelled, so this method is not appropriate for
very shallow tunnels where arching around the crown will be limited by
the cover available.
The method assumes the ground is homogeneous – that it has the same
properties everywhere. This assumption means that layered soils or soils
whose properties vary with depth or from one side of the tunnel to the
other cannot be modelled.
It also assumes ‘plane strain’. This means that the tunnel and the ground
extend infinitely into the page, and there is zero strain in that direction –
hence all strains occur within the 2D plane.
Pre-existing vertical and horizontal total stresses in the ground σ v and σ h
are allowed to be different, but they do not vary with depth, i.e. we use the
values at tunnel axis level.
The tunnel-ground boundary is at the radius to the extrados, r0,
and this is where ground stresses are calculated and interact with the
tunnel lining. Where the axial or bending stiffness of the tunnel lining
is being calculated, the radius to the centroid of the tunnel lining rm is
used.
Figure 6.4 shows the boundary conditions and ground stresses used in
the Curtis-Muir Wood solution.
The ground and the lining are assumed elastic, therefore there is no non-
linearity, no plasticity, no failure criterion, no groundwater flow and no
soil consolidation. Groundwater pressure is taken into account by the use
of total stress rather than effective stress plus pore pressure. An undrained
soil response may be modelled by using the undrained Young’s modulus
and a Poisson’s ratio of 0.5.
If σ v ≠ σ h, then one can imagine separating the stress into two compo-
nents, one which is uniform in all directions and one which is exactly oppo-
site in the vertical and horizontal directions.
Figure 6.4
Boundary conditions and ground stresses in the Curtis-Muir Wood
solution.
The distortional stress P0 is defined, which is given by:
P0 = σ v − σ h (6.3)
Therefore, the distortional unloading of the ground is:
P0
cos2θ (6.4)
2
which varies with angle θ from the crown as shown in Figure 6.5.
So at the crown, where θ = 0°, the distortional stress is (σ v − σ h ) /2, and
at axis level, where θ = 90°, the distortional stress is − (σ v − σ h ) /2. Try it for
different values of θ.
It follows from the definition of distortional stress that the
uniform stress Pu is the average of the vertical and horizontal stresses,
i.e. (σ v + σ h ) /2.
Figure 6.5 Distortional stress in the Curtis-Muir Wood solution from θ = 0 to θ = 90°.

Only distortional stress in the first quadrant is shown.
If stresses Snc and Stc are the normal and shear stresses in the ground
around the opening, such that:
P0
Snc = Stc = (6.5)
2
then the normal stress Snc cos2θ acts normally around the opening
and − Stc sin2θ acts tangentially around the opening. The minus sign is
because for θ = 0 at the crown and positive clockwise, Mohr’s circle gives
negative shear acting on the lining for 0 < θ < π /2. The arrow in Figure 6.6
shows the magnitude and direction of the shear.
Figure 6.6 Magnitude and direction of shear in the Curtis-Muir Wood solution.

6.3.3 Elliptical deformation of a circular opening

First, elliptical deformation due to distortional stress of a circular opening
in a homogeneous elastic material is considered, i.e. deformation of the
ground with no tunnel lining. Airy’s stress function is used for this purpose
(Timoshenko & Goodier, 1970), with the form:
φ = ( ar 2 + br 4 + cr −2 + d ) cos2θ (6.6)
ϕ is a stress field
a, b, c and d are constants
We hypothesise that the solution has the form of Airy’s stress function
and we calculate the values of the constants a, b, c and d using equations
derived from the boundary conditions and Hooke’s Law. This will give us
expressions for ground displacements around a circular opening.
In polar coordinates the radial, tangential and shear stresses in the
ground are given by:
1 ∂φ 1 ∂2 φ
σr = + (6.7)
r ∂r r 2 ∂θ 2
∂2 φ
σθ = (6.8)
∂r 2
∂  1 ∂φ 
τ rθ = −   (6.9)
∂r  r ∂θ 
Substituting Equation 6.6 into Equations 6.7–6.9 gives:
σ r = ( −2a − 6cr −4 − 4dr −2 ) cos2θ (6.10)
σ θ = ( 2a + 12br 2 + 6cr −4 ) cos2θ (6.11)
τ rθ = ( 2a + 6br 2 − 6cr −4 − 2dr −2 ) sin2θ (6.12)
At the tunnel boundary (r = r0), we have σ r = τ rθ = 0, so from

Equations 6.10 and 6.12, we get:
Sn
a + 3cr0−4 + 2dr0−2 = − (6.13)
2
St
a + 3br02 − 3cr0−4 − dr0−2 = − (6.14)
2
At the infinite boundary (r = ∞), we have σ r = Sn cos2θ and τ rθ = − St sin2θ ,

giving:
a + 3cr∞−4 + 2dr∞−2 = 0 (6.15)
a + 3br∞2 − 3cr∞−4 − dr∞−2 = 0 (6.16)
Since r∞ is very large, Equation 6.15 tells us that a = 0.

For the same reason, and because a = 0, Equation 6.16 tells us that b = 0.
Rearranging Equation 6.13, and using a = 0, gives:
− Sn r04 2dr02
c= − (6.17)
6 3
Inserting Equation 6.17 into 6.14 gives:
 S + St  2
d = − n r (6.18)
 2  0
Substituting for d in Equation 6.17 gives:
c=
( Sn + 2St ) r04 (6.19)
6
Substituting for a = 0, c and d in Equation 6.10 gives:
 r4 r2 
σ r = − ( 2St + Sn ) 04 + 2 ( Sn + St ) 02  cos2θ (6.20)
 r r 
Substituting for a = 0, b = 0, and c in Equation 6.11 gives:
 r4 
σ θ = ( Sn + 2St ) 04  cos2θ (6.21)
 r 
And substituting for a = 0, b = 0, c and d in Equation 6.12 gives:
 r4 r2 
τ rθ = − ( Sn + 2St ) 04 + ( Sn + St ) 02  sin2θ (6.22)
 r r 
Now, for plane strain, where the strain in the longitudinal tunnel axis
direction (here denoted with subscript z) is assumed to be zero, Hooke’s
Law in polar cylindrical coordinates gives us the following three equations:
1
εr = σ r − ν (σ θ + σ z )  (6.23)
E
1
εθ = σ θ − ν (σ r + σ z )  (6.24)
E
1
εz = σ z − ν (σ r + σ θ )  = 0 (6.25)
E
Rearranging Equation 6.25 for σ z gives:
σ z = ν (σ r + σ θ ) (6.26)
Substituting for σ z in Equations 6.23 and 6.24 gives:
(1 + ν )  1 − ν σ − νσ  (6.27)
εr =
E ( ) r θ
(1 + ν )  1 − ν σ − νσ  (6.28)
εθ =
E ( ) θ r
Now, substituting values of σθ and σr from Equations 6.20 and 6.21 into
Equations 6.27 and 6.28, and then simplifying the expressions, gives:
(1 + ν )  1 − ν − 2S + S r04 + 2 S + S r02  − ν  S + 2S r04  cos2θ 

εr = ( )  ( t n ) 4 ( n t ) 2  ( n t) 4  
E   r r   r  

=
(1+ν)  r02 r04 
2 (1 − ν ) ( Sn + St ) 2 − ( 2St + Sn ) 4  cos2θ
E  r r 
(6.29)
(1 + ν )  1 − ν  S + 2S r04  − ν − 2S + S r04 + 2 S + S r02   cos2θ

εθ = ( ) ( n t) 4   ( t n) 4 ( n t ) 2 
E   r   r r 

(1+ν)  r04 r02 
= ( Sn + 2St ) 4 − 2ν ( Sn + St ) 2  cos2θ
E  r r 
(6.30)
For a polar cylindrical coordinate system, the following relationships

between strains and displacements apply:
∂ur
εr = (6.31)
∂r
1  ∂uθ 
εθ =  + ur  (6.32)
r  ∂θ 
Therefore:
(1 + ν )  (2St + Sn ) r04 − 2 1 − ν S + S r02  cos2θ (6.33)

ur =
∫ ε r dr =
E

 33
( )( n t ) 
r r 
And:
(  1+ν)  r04 r02 

uθ =
∫ ( rεθ − ur )dθ = ∫   E 
 ( Sn + 2 St )
r3
− 2ν ( Sn + St )
r 
cos 2θ
(1 + ν )  (2St + Sn ) r04 − 2 1 − ν S + S r02  cos 2θ  dθ

−  ( )( n t )  
E  3 3 r r  
(1 + ν )  ( 2Sn + 4St ) r04 + r02 

=  ( 2 − 4ν ) ( Sn + St )  sin 2θ
2E  3 r3 r 
(1 + ν )  ( Sn + 2St ) r04 + r02 

=  (1 − 2ν ) ( Sn + St )  sin 2θ (6.34)
E  3 r 3 r 
Equations 6.33 and 6.34, when combined with the displacements due to
uniform compression (coming up later), are identical to the ‘Kirsch prob-
lem’ for an infinite elastic plate with a hole, adapted to plane strain (see e.g.
Jaeger & Cook, 1976 or Hoek & Brown, 1980).
When r = r0, i.e. at the periphery of the opening or the extrados of the
lining, the deformations are obtained.
Radial deformation at the periphery of the opening, when r = r0:
− r0 (1 + ν )
u0r = ( 5 − 6ν ) Snc + ( 4 − 6ν ) Stc  cos2θ (6.35)
3Ec
Tangential deformation at the periphery of the opening, when r = r0:
r0 (1 + ν )
u0θ = ( 4 − 6ν ) Snc + ( 5 − 6ν ) Stc  sin2θ (6.36)
3Ec 
By substituting Equation 6.5 into 6.35 and 6.36, we get Equations 6.37 and
6.38, which describe the elastic elliptical deformation of the ground due to
distortional stress:
− r0 (1 + ν ) P
u0rc = (3 − 4ν ) 0 cos2θ (6.37)
Ec 2
r0 (1 + ν ) P
u0θ c = (3 − 4ν ) 0 sin2θ (6.38)
Ec 2
6.3.4 Elliptical deformation of a thin

inextensible lining
Since this analytical solution considers elasticity, the principle of superposi-
tion holds. This means we can consider elliptical distortion and uniform
compression of the lining separately, then add the displacements, strains or
forces together at the end. In this section we are considering the effect of
distortional load on a thin inextensible circular lining. We can then equate
the displacements of the lining with the displacements of the ground found
in the previous section, giving us the soil-structure interaction.
Looking back at Figure 6.6, the radial ground pressure is given by
Sncos2 θ and the tangential stress at the boundary is given by St sin2θ .
At the crown, where θ = 0, the internal forces in a thin inextensible lining
will be given by integration along the arc shown in Figure 6.7.
To find N at the crown, horizontal force equilibrium is used, integrating
Sn and St over small arc lengths r0dθ from θ = 0 to θ = π/2.
To find M at the crown, moment equilibrium is used, where the stresses
Sn and St over small arc lengths r0dθ are integrated from θ = 0 to θ = π/4 with
lever arms j = r0sinθ and k = r0 (1 − cosθ ), respectively, shown in Figure 6.7.
The reason the integration is only from 0 to π/4 is because the deformation is
elliptical and the moment M is equal to zero at θ = π/4, and Mθ =0 = − Mθ = 90° .
For this purpose we are assuming that the elliptical deformation will have
a negligible effect on the geometry of the lever arms.
π
2
Nθ =0 = r0
∫ ( S cos2θ sinθ − S sin2θ cosθ ) dθ (6.39)
0
n t
π
4
Mθ =0 = r02
∫ ( S cos2θ sinθ + S sin2θ (1 − cosθ )) dθ (6.40)
0
n t
Figure 6.7 Resolving equilibrium to find internal forces at the crown of a thin

inextensible lining subject to a distortional stress field.
Using the following equations (Equations 6.41–6.44 are found by ‘integra-

tion by parts’ – see Appendix B) helps us with the integration:
π
2 π
 cos2θ cosθ + 2sin2θ sinθ  2 1

∫
0
cos2θ sinθ dθ = 
 3  = − 3 (6.41)
0
π π
 − ( sin2θ sinθ + 2cos2θ cosθ )  2 2
2

∫
0
sin2θ cosθ dθ = 
 3
 = (6.42)
0 3
π
4 π
 cos2θ cosθ + 2sin2θ sinθ  2 1

∫
0
cos2θ sinθ dθ = 
 3  = − 3 (6.43)
0
π π
 − ( sin2θ sinθ + 2cos2θ cosθ )  2 2
4

∫
0
sin2θ cosθ dθ = 
 3
 = (6.44)
0 3
π
4
1

∫sin2θ dθ = 2 (6.45)
0
Therefore:
− r0
Nθ =0 = ( Sn + 2St ) (6.46)
3
r02
Mθ =0 = (2Sn + St ) (6.47)
6
And it follows that:
− r0
N = Nθ =0 cos2θ = ( Sn + 2St ) cos2θ (6.48)
3
r02
M = Mθ =0 cos2θ = (2Sn + St ) cos2θ (6.49)
6
Morgan (1961) showed that the moment M is induced by a change in the

radius of curvature ∆r from a circle to an ellipse, where M = EI /∆r. The
ellipse major axis radius is (r0 + δ) and the minor axis radius is (r0 − δ).
The largest change in curvature happens at these points, and hence the
largest moments.
Morgan (1961) also showed that the radial deformation of a thin inex-
tensible lining deforming elliptically is given by:
− Mr02
ul = (6.50)
3EI
Substituting for M in Equation 6.50 using the expression in Equation 6.49
gives radial deformation of the lining:
− r04
ul = (2Sn + St ) cos2θ (6.51)
18EI
6.3.5 ‘Full slip’ – no shear between lining and ground

In order to equate ground deformations and lining deformations, we need
to first consider the interface between the lining and the ground. There are
two limiting conditions, ‘full slip’ and ‘no slip’.
‘Full slip’ means there is no shear between the lining and the ground
and the interface is assumed to slip without resistance. ‘No slip’ means
a perfect bond between the lining and the ground, and shear defor-
mations are exactly equal. These limiting conditions may seem odd or
unrealistic, but we can look at both ‘full slip’ and ‘no slip’, and the true
situation is somewhere in between, so by using both we can bound the
true solution.
‘Full slip’ means St = 0, and we may proceed as follows:
a. The ground will deform radially according to Equation 6.37.

b. The lining will deform radially under the load Sn cos2θ in
Equation 6.52:
ul = r04 2Snl cos2θ /18EI (6.52)
c. The ground will be restrained from deforming radially under the

action of a load Snc cos2θ = − Sn cos2θ according to Equation 6.35:
− r0 (1 + ν )
u0 = (5 − 6ν ) Sn cos2θ (6.53)
3Ec
d. Since u0 + u0 = ul , then we combine Equations 6.35, 6.52 and 6.53 to

find an expression for Sn:
3 (3 − 4ν ) P0 /2
Sn = (6.54)
5 − 6ν + 4Q2
where:
Ec 1 r03
Q2 = (6.55)
E (1 + ν ) 12I
Q2 is just a shortcut to make the equation simpler; it doesn’t appear to

mean anything on its own. It appears in several of the following equations.
What we have done is calculate the radial deformation of an unlined
tunnel due to distortional load, and how the lining will deform under
such a load. Ensuring compatibility of radial deformations (by equat-
ing the two), we have used Equation 6.54 to calculate directly the
load Sn that is required for equilibrium.
e. From Equations 6.48 and 6.49, we can calculate the moment and
hoop force in the lining using the value of Sn calculated in Equation
6.54, or we can get there directly by substitution:
Nd =
− P0 r0 (3 − 4ν ) cos2θ (6.56)
2 ( 5 − 6ν + 4Q2 )
M=
P0 r02 (3 − 4ν ) cos2θ (6.57)
2 ( 5 − 6ν + 4Q2 )
6.3.6 ‘No slip’ – full shear interaction between

lining and ground
In this case we need to equate both radial and tangential deformations.
Repeating a similar procedure to Section 6.3.5, we find:
Sn =
(1 − Q2 ) P0 /2
 3 − 2ν  (6.58)
1 + Q2 
 3 − 4ν 
St =
(1 + 2Q2 ) P0 /2
 3 − 2ν  (6.59)
1 + Q2 
 3 − 4ν 
Putting Sn and St into Equations 6.48 and 6.49, we can calculate thrust and
moment due to distortional load when there is no slip allowed between the
lining and the ground.
6.3.7 Direct compression of the lining

due to uniform load
In Sections 6.3.5 and 6.3.6 we calculated the effects of distortional load
with either ‘full slip’ or ‘no slip’ while ignoring the uniform load. In this
section we will calculate the effects of uniform load only. Since the analysis
is elastic, we can simply superpose the effects of distortional load and uni-
form load to get the final result. The uniform load will only affect the axial
thrust N. It will not affect the moment M, because a uniform compression
of a circular lining will not generate any bending moments.
First, calculate the uniform loading from the ground:
σv +σh
Pu = (6.60)
2
There is an interaction, and in effect the ground will share some of this load
through arching around the opening, so the thrust due to uniform loading
is not equal to the uniform loading multiplied by the radius of the tunnel,
but less than this.
The relative stiffness of the ground and lining determines how much
of the load is taken by the lining. This was called the ‘compressibility
factor’, Rc, by Muir Wood (1975) and called Q1 by Curtis (1974), and is
given by:
ηr0 Ec (1 − ν l2 )
Rc = (6.61)
tE (1 + ν )
ν l is the Poisson’s ratio of the lining

η is the ratio of the radius to the lining centroid to that of the extrados
Note that there is an error in Muir Wood (1975: equation 36),

which had η in the denominator (bottom part) of the quotient in
Equation 6.61.
For the case of a rectangular section ‘smoothbore’ lining or a sprayed
concrete lining, the centroid is positioned at the mid-thickness of the lining,
denoted r m, so Equation 6.61 can be rewritten:
Ec (1 − ν l2 ) rm
Rc = (6.62)
E (1 + ν ) t
The axial hoop force N u due to uniform loading can be calculated using the
following equation:
r0 Pu
Nu = (6.63)
1 + Rc
The equations are now all derived. The following worked example will
show how they are used to calculate the hoop force and bending moment
in a tunnel lining.
WORKED EXAMPLE 6.3

The following example is for a rectangular section lining in undrained
stiff clay with the parameters listed in Table 6.2.
A calculation of a wished-in-place, perfectly stiff lining, where the pre-
existing ground stresses are applied directly to the lining with no interac-
tion, would result in the following hoop forces:
630 kN/m at the crown and invert

900 kN/m at axis
765 kN/m average
Calculate expressions for hoop force and bending moment as a function

of angle 2 θ for both the full slip and no slip cases.
Then program the Curtis-Muir Wood solution into a spreadsheet.
Calculate hoop forces and bending moments around the ring for both
the full slip and no slip cases. Plot these on a graph.
Table 6.2 Soil and lining parameters for Worked Example 6.3.
Parameter Value Units Description
Ec 80 MPa Young’s modulus of the soil
ν 0.5 - Poisson’s ratio of the soil
σv 300 kPa Vertical in situ total stress
σh 210 kPa Horizontal in situ total stress
r0 3 m Radius to extrados
E 20 GPa Young’s modulus of the lining
t 0.3 m Thickness of the lining
I 0.00225 m4 Second moment of area of the lining
νl 0.2 - Poisson’s ratio of the lining
Uniform compression:
First, calculate the direct compression of the lining due to uniform load
(see Section 6.3.7):
σ v + σ h 300 + 210
Pu = = = 255 kPa
2 2
Ec (1 − ν l2 ) rm 80000 × (1 − 0.22 ) 2.85

Rc = = = 0.02432
E (1 + ν ) t 20 × 106 × (1 + 0.5) 0.3
r0 Pu 3 × 255
Nu = = = 746.84 kN/m
1 + Rc 1 + 0.02432
Note that compared to the average hoop force calculated by simple

wished-in-place equilibrium, N u is lower, by a factor of 1/ (1 + Rc ). This
occurs because the lining compresses and the resulting radial deformation
of the ground results in the ground arching around the tunnel. Increasing
Rc will increase load sharing with the ground and reduce the uniform
direct compression of the lining N u . This occurs when, for instance, the
Young’s modulus of the soil Ec is increased, or when the Young’s modulus
of the lining E is reduced.
Distortion – ‘full slip’ case:
Next we need to calculate the effect of distortional load Sn in the ‘full slip’
case:
P0 = σ v − σ h = 300 − 210 = 90 kPa
Ec 1 r03 80 × 103 1 33
Q2 = = = 2.667 m −1
E (1 + ν ) 12I 20 × 10 (1 + 0.5) 12 × 0.00225
6
The hoop force due to distortional load is given by Equation 6.56:
− P0 r0 (3 − 4ν ) −90 × 3 (3 − 4 × 0.5)
Nd = cos2θ = cos2θ
2 5 − 6ν + 4Q2 2 5 − 6 × 0.5 + 4 × 2.667
= −10.658cos2θ
Now we can calculate the hoop force N, which is the sum of the distor-
tional load (Equation 6.56) and the uniform load (Equation 6.63):
N = N u + N d = 746.84 − 10.658cos2θ
The bending moment M is given by Equation 6.57:

P0r02 (3 − 4ν ) 90 × 32 (3 − 4 × 0.5)
M= cos2θ = cos2θ = 31.974cos2θ
2 5 − 6ν + 4Q2 2 5 − 6 × 0.5 + 4 × 2.667
Distortion – ‘no slip’ case:

Using Equation 6.58:
Sn =
(1 − Q2 ) P0 /2 =
(1 − 2.667 ) × 90/2 = −11.842 kN/m
 3 − 2ν   3 − 2 × 0.5 
1 + Q2  1 + 2.667 × 
 3 − 4ν   3 − 4 × 0.5 
St =
(1 + 2Q2 ) P0 /2 =
(1 + 2 × 2.667 ) × 90/2 = 45 kN/m
 3 − 2ν   3 − 2 × 0.5 
1 + Q2  1 + Q2 
 3 − 4ν   3 − 4 × 0.5 
Inserting values for Sn and St into Equations 6.48 and 6.49 gives:
− r0
Nd =
3
( Sn + 2St ) cos2θ = −33 ( −11.842 + 2 × 45) cos2θ = −78.158cos2θ kN/m
r02 32
M=
6
( (
2Sn + St ) cos2θ =
6
)
2 × ( −11.842) + 45 cos2θ = 31.974cos2θ kN/m
The total hoop force N for the ‘no slip’ case is given by:
N = N u + N d = 746.84 − 78.158cos2θ kN/m
Results from spreadsheet calculation varying θ:

The results of the Curtis-Muir Wood calculation are shown in
Figure 6.8. Note that the hoop force is consistently lower than the
average hoop force in the simple wished-in-place calculation (given in
the question), and varies much less around the lining. This is because
the lining is flexible and hence will deform elliptically with the ground,
reducing radial stress where the in situ stress was highest to begin with
(i.e. in the vertical direction) and increasing radial stress where the in
situ stress was lowest (in the horizontal direction).
We can also see that the ‘no slip’ calculation gives a larger variation of
hoop force around the tunnel than the ‘full slip’ calculation.
Figure 6.8 Worked Example 6.3 spreadsheet-generated graph showing hoop

forces and bending moments in the lining.
Note that the bending moments are the same in both the ‘full slip’
and ‘no slip’ cases. This only happens when the Poisson’s ratio of the
ground is equal to 0.5.
The bending moment is low compared to the hoop force. This is nearly
always the case and the section is likely to be mainly in compression in
most cases, making design of tunnel linings in lightly reinforced or unre-
inforced concrete possible.
The variation of hoop force around the lining is higher when the no
slip condition is in force. This is because the higher vertical stress is trans-
ferred to the lining via interface shear even where the angle of incidence
is high, whereas in the full slip condition, the stress is redistributed more.
If the ground stiffness were increased, the effect would be to reduce
both hoop forces and bending moments in the lining. Interestingly, if
the ground stiffness were very low, and the no slip condition were in
force, the solution would approach the wished-in-place perfectly stiff lin-
ing situation described in the question with 900 kN/m hoop force at axis
and 630 kN/m hoop force at the crown and invert.
Figure 6.9 shows the effect of ground stiffness, which serves to redis-
tribute loads and, with higher ground stiffness, to reduce them. A ground
stiffness of 80 MPa would be typical of a stiff overconsolidated clay such
as London Clay at strains of around 0.1%. A ground stiffness of 800 MPa
would be typical for a soft rock such as chalk or weak sandstone.
Figure 6.9 Worked Example 6.3 hoop forces at different values of ground stiffness Ec.
6.4 GLOBAL DESIGN OF SHAFTS
If we were to apply the Curtis-Muir Wood solution to a shaft, this would

allow some soil-structure interaction to occur in our design, result-
ing in a more efficient lining design. However, on most soft ground
tunnelling projects the horizontal principal stresses will be assumed
equal, so no bending moments will be calculated. This would result
in an unconservative design, because it is reasonable to expect some
uneven loading to induce distortion and bending moments in the shaft
lining. Therefore, it is recommended that the horizontal total stress in
one of the principal directions is reduced by a factor of 0.8 to induce
bending moments and ensure that the lining has sufficient structural
capacity.
It will also become clear in later sections of this book that in many
cases hoop forces are beneficial, because they increase the moment
capacity of a reinforced concrete lining. With this in mind, careful atten-
tion needs to be paid to the shaft lining near to the surface where the in
situ ground stresses are lowest, not just the deepest point where they are
highest. It is also near to the surface that uneven loading is most likely
to occur, perhaps unloading on one side due to adjacent excavation, or
surcharges caused by structures, stockpiled materials, spoil heaps or
construction plant such as a crane. Using the Curtis-Muir Wood solu-
tion with an unequal stress ratio of 0.8 makes the bending moments
proportional to the ground stresses, and so close to the surface it may
underestimate the true bending moments. Therefore, it is important to
understand the potential for uneven loading close to the surface and to
calculate its effect.
6.5 BEDDED BEAM MODELS
‘Bedded beam’ or ‘beam-spring’ models are analytical or numerical models

consisting of beam elements representing the lining, and spring elements
representing the interaction with the surrounding ground. Springs are
usually in radial and tangential directions to represent radial and shear
stresses, respectively. They are sometimes also referred to as ‘hyperstatic
reaction method’ models or ‘HRM’ models.
Bedded beam models are generally thought of as less sophisticated than
continuum analytical solutions such as the Curtis-Muir Wood solution
(Duddeck & Erdmann, 1985). This is because the springs are independent
of each other and hence they do not properly model soil-structure interac-
tion, where it would be expected that deformation of the lining and hence
the ground adjacent to it in one location would result in a redistribution of
stress in the ground, affecting neighbouring areas (e.g. Tomlinson, 1995:
p. 199). This phenomenon was observed for a tunnel by Do et al. (2018), in

a comparison of the HRM with a numerical continuum model in FLAC3D.
For this reason, bedded beam models are not examined in much detail in
this book.
A comparison of the boundary conditions and the resulting deforma-
tions and stresses for a bedded beam model (Schulze & Duddeck’s method,
1964) and a continuum analytical solution (such as the Curtis-Muir Wood
solution) is shown in Figure 6.10.
In Figure 6.10, the spring constant Kr is related to the ground stiffness
and simulates load sharing between the ground and the tunnel lining. The
force in the spring increases as the lining moves inwards, reducing the net
load on the lining, and vice versa. It is often necessary to remove springs
in locations where they would experience too much tension, and so springs
Figure 6.10 Comparison of bedded beam (top) vs continuum (bottom) analysis (redrawn

from Duddeck & Erdmann, 1985).
may be omitted in the crown of the tunnel, as Schulze & Duddeck (1964)
did, and as shown in Figure 6.10.
Bedded beam models may also be implemented in a structural frame
analysis software package. One advantage of bedded beam models is
that they do allow non-circular linings to be quickly and easily modelled
and are therefore still used in preliminary design of sprayed concrete
linings.
In 2D, these models can be used to simulate a solid monolithic ring with
full rigidity or with a reduced moment of inertia to take account of joints
in a segmental lining. Alternatively, joints can be modelled by introducing
a rotational spring and tangential shear springs at joint locations. This will
be discussed in Chapter 9.
In 3D bedded beam models, springs can also be added in the longitu-
dinal direction to simulate longitudinal shear between the lining and the
ground, and circumferential joints may also be added. It is usual to stagger
the radial joints in adjacent rings in real tunnels, and so this should also be
modelled. More details may be found in Chapter 9.
6.6 SUMMARY
In this chapter we have started with a very simple wished-in-place model

of a tunnel and then looked at the effect of soil-structure interaction using
the Curtis-Muir Wood solution. Finally, we looked at the global design of
shafts and briefly touched upon bedded beam models and seismic design.
This will be built upon in the following chapter where numerical model-
ling will be used to introduce more complexity and make the design more
realistic.
You should now understand the boundary conditions and assumptions
inherent in analytical solutions of a tunnel, and know how to use the Curtis-
Muir Wood solution to estimate the hoop forces and bending moments in
a tunnel or shaft lining.
Hoop forces and bending moments calculated in a tunnel using the
Curtis-Muir Wood solution or other similar 2D plane strain analytical
solutions do not account for the 3D nature of tunnel excavation, where
extrusion of the face and radial ground movements change the stress state
of the ground before the lining is installed. There are also important lon-
gitudinal bending moments and axial forces in tunnel linings that are not
considered in 2D.
Shotcrete tunnel linings can generally be considered to be structurally
continuous, but segmental linings have joints that allow some rotation.
The design of segmental linings is covered in some detail in Chapter 9,
where methods for calculating the rotational stiffness of joints will be
explained.
6.7 PROBLEMS
Q6.1. A tunnel is to be constructed with the geometry and geology as

shown in Figure 6.11. Assume the pore pressures below the water
table are hydrostatic.
Figure 6.11 Q6.1 tunnel cross-section.
i. What is the in situ vertical total stress at tunnel axis level?

ii. Assuming K0 = 1.0, what is the in situ horizontal total stress
at tunnel axis level?
iii. Using the basic method of applying the in situ total stress directly
to the lining (assuming the lining is wished-in-place and perfectly
stiff), what is the hoop thrust at axis in the tunnel lining, per
metre length of tunnel? Ignore the self-weight of the lining.
iv. Using the basic method of applying the in situ total stress
directly to the lining, what is the hoop thrust at crown and
invert per metre length of tunnel, using the horizontal in situ
stress calculated at axis level?
v. Using the Curtis-Muir Wood solution, calculate the hoop
thrust in a smoothbore precast concrete segmental lining
using the following parameters:
Radius to extrados of tunnel lining r0 = 3.5 m
Lining thickness t = 300 mm
Young’s modulus of the lining E = 30 GPa
Poisson’s ratio of the lining νl = 0.2
Undrained Young’s modulus of the stiff clay Ec = 48 MPa
Undrained Poisson’s ratio of the stiff clay ν = 0.5
Number of segments n=8
Q6.2. Use the same tunnel geometry as in Question Q6.1, except with
the groundwater level 3 m below the ground surface and K0 = 0.6.
Assume the bulk unit weight of the sand and gravel is unchanged.
Calculate the following:
i. What is the in situ vertical effective stress at tunnel axis level?
ii. What is the in situ horizontal effective stress at tunnel axis level?
iii. What is the in situ horizontal total stress at tunnel axis level?
iv. Using the basic method of applying the in situ total stress
directly to the lining (assuming the lining is wished-in-place
and perfectly stiff), what is the hoop thrust at axis per metre
length of tunnel? Ignore the self-weight of the lining.
v. Using the basic method of applying the in situ total stress
directly to the lining, what is the hoop thrust at crown and
invert per metre length of tunnel, using the in situ stress cal-
culated at axis level?
vi. Using the Curtis-Muir Wood solution, calculate the hoop
thrust per metre length of lining as a function of the form
A + Bcos2 θ, using the elastic parameters and tunnel lining
described in Question 6.1(v), for the no slip case.
vii. Using the Curtis-Muir Wood solution, calculate the bending
moment per metre length of lining as a function of cos2θ,
using the elastic parameters and tunnel lining described in
Question 6.1(v), for the no slip case.
viii. Sketch a graph plotting hoop thrust per metre length of lining
on the y-axis and polar angle θ (clockwise from the crown
of the tunnel) on the x-axis from 0 to 180° (crown to invert),
marking the maximum and minimum values of hoop thrust.
There should be no need to calculate intermediate points.
Q6.3. A 12 m internal diameter shaft 40 m deep is to be constructed
using a shotcrete lining 0.35 m thick in stiff overconsolidated
clay. Assume the Young’s modulus of the shotcrete is 30 GPa
and the Poisson’s ratio is 0.2. The London Clay has a Young’s
modulus that increases with depth according to the relation-
ship Ec = 20 + 3.2z [GPa]. It has a unit weight of 20 kN/m3 and
K0 = 0.8. Pore pressures are hydrostatic with the water table 5 m
below the surface.
i. Set up a spreadsheet to calculate hoop force and bending
moment using the Curtis-Muir Wood solution.
ii. Calculate the horizontal total stress at 10, 20, 30 and 40 m
depth.
iii. Assuming that the horizontal stress in the minor principal
direction is 0.8 times the value in the major principal direc-
tion, calculate the hoop forces and bending moments at 10,
20, 30 and 40 m depth in the short-term undrained case.
Remember that the soil stiffness is increasing with depth.
iv. To assess the sensitivity of the design to lower shotcrete

stiffness at early age, calculate the hoop forces and bending
moments at 40 m depth using a shotcrete Young’s modulus of
5 and 15 GPa. Plot on a graph along with the 30 GPa values
and describe the effect.
v. Using a shotcrete Young’s modulus of 30 GPa, by how much
would you need to reduce the initial in situ stresses in the
ground to get the same hoop force you get with a shotcrete
Young’s modulus of 5 GPa? Is the difference related to the
parameter Rc in Equation 6.62?
REFERENCES
Ahrens, H., Lindner, E. & Lux, K.-H. (1982). Zur Dimensionierung von
Tunnelausbauten nach den ‘Empfehlungen zur Berechnung von Tunneln
im Lockergestein 1980’ [Dimensioning of tunnel linings by applying the
recommendations of 1980]. Bautechnik 59, No. 8, 260–273 and 303–311.
Curtis, D. J. (1974). Visco-elastic tunnel analysis. Tunnels and Tunnelling,
November, 38–39.
Curtis, D. J. (1975). Discussion: The circular tunnel in elastic ground. Géotechnique
25, No. 1, 115–127.
Do, N. A., Dias, D. & Oreste, P. (2018). Numerical investigation of segmental tun-
nel linings-comparison between the hyperstatic reaction method and a 3D
numerical model. Geomech. Eng. 14, No. 3, 293–299.
Duddeck, H. (1980). Empfehlungen zur Berechnung von Tunneln im Lockergestein,
Deutsche Gesellschaft für Erd- und Grundbau. Bautechnik 57, No. 10,
349–356.
Duddeck, H. & Erdmann, J. (1985). On structural design models for tunnels in
soft soil. Undergr. Space 9, 246–259.
Einstein, H. H. & Schwartz, C. W. (1979). Simplified analysis for tunnel supports.
J. Geotech. Engng Div. ASCE 105, No.4, April, 499–518.
Hoek, E. & Brown, E. T. (1980). Underground excavations in rock. London:
IMM.
Jaeger J. C. & Cook, N. G. W. (1976). Fundamentals of rock mechanics,
2nd Edition. London: Chapman & Hall.
Morgan, H. D. (1961). A contribution to the analysis of stress in a circular tunnel.
Géotechnique 11, 37–46.
Muir Wood, A. M. (1975). The circular tunnel in elastic ground. Géotechnique 25,
No. 1, 115–117.
Schulze, H. & Duddeck, H. (1964). Spannungen in Schildvorgetriebenen Tunneln.
Beton- und Stahlbeton 59, 169–175.
Timoshenko, S. P. & Goodier, J. N. (1970). Theory of elasticity, 3rd Edition –
International Student Edition. New York: McGraw-Hill Book Company.
Tomlinson, M. J. (1995). Foundation design and construction, 6th Edition.
Harlow: Addison Wesley Longman.
Chapter 7
Global design using numerical

modelling
Numerical models are similar to the analytical models we used in

Chapter 6. Both are defined by geometry and boundary conditions and
have material properties that govern the constitutive behaviour of the
ground and the tunnel lining. They are just more complex and so special
software is needed to build the model and to solve the equations.
Numerical models allow a problem to be discretised into a mesh of ele-
ments that are ‘finite’. This is to differentiate them from the infinitesimal
elements one might find in calculus. Across each element, field quantities,
such as stress or displacement, are only allowed to vary in a simple manner,
typically following a linear or quadratic equation. This allows a solution
to be found by ensuring that the number of unknowns is matched by the
number of equations. If the real variation of field quantities is more com-
plex, which it often is, then this means that the outputs will be approxi-
mate. Accuracy may be improved by making the elements smaller or more
complex. Smaller elements mean more elements, and hence more equations
to solve. More complex elements (for instance, switching from linear to
quadratic) will also mean more equations. Consequently, there is always a
trade-off between accuracy and the amount of time the computer takes to
run the model.
• the principles of numerical modelling in 2D and 3D

• the limitations of numerical modelling
• the effect of using different boundary conditions at the edges of the
model and at the tunnel excavation boundary
• how groundwater is modelled in a soil in the short- and long-term
• build a 2D or 3D numerical model of a tunnel

• set suitable boundary conditions and distances
• select suitable element types and mesh density
• debug your model
DOI: 10.1201/9780429470387-7 191
• validate your model

• interpret and present results
There are two main types of numerical analysis that are suitable for the
routine design of tunnels in soft ground: the finite element method and the
finite difference method. The finite element method, broadly speaking, cre-
ates a large number of equations from considerations of equilibrium, com-
patibility, constitutive behaviour and boundary conditions. It puts all the
equations into a big matrix and inverts the matrix to find a solution to the
unknowns, which are the stresses, strains and displacements. The finite dif-
ference method, on the other hand, finds equilibrium gradually by stepping
towards the solution in increments of displacement or velocity. This is done
on an element-by-element basis and there is no need to solve a large matrix
of equations. The stepwise solution method means that the materials can
follow almost any stress path to failure and makes it easy to incorporate
complex constitutive behaviour.
Information on specific programs may be found online or in their manu-
als. The theoretical background to the finite element method may be found
in textbooks such as the one by Cook et al. (2002), or more specifically, for
geotechnical problems, see Potts & Zdravković (1999, 2001). For the finite
difference method, the FLAC or FLAC3D theoretical background manuals
(Itasca, 2006) are probably the best source of information.
Once the problem has been defined and a preliminary design obtained by
using simple calculations and analytical solutions, the next stage is to begin
introducing further layers of complexity to the model. This might involve:
• Geometry – non-circular tunnel lining, multiple soil layers, non-

horizontal ground surface, other tunnels, basements, piles or surface
structures, groundwater, asymmetric surface surcharge or excava-
tion, tunnel excavation sequence
• Constitutive models – soil constitutive model (e.g. effective stress
analysis, plasticity, nonlinearity, stress path direction, contraction
or dilation under shear, consolidation and swelling), lining constitu-
tive model (e.g. stress-strain nonlinearity, plasticity, shrinkage, creep,
thermal effects, crack formation, joints, plastic hinges)
• Dynamic analysis for earthquake resistant design
• Long-term effects – degradation of temporary linings or sacrificial
zones due to chemical attack or because they are not designed for
long-term durability, long-term groundwater level changes or steady-
state flow scenarios, consolidation or swelling of clays due to perma-
nent changes in pore pressure
Most of these effects can only be included in a model by using numerical

analysis, using either a finite element or a finite difference method, but
would be difficult or impossible to achieve in an analytical model.
Global design using numerical modelling 193
I am again going to emphasise how important it is to understand that the

aim here is not to immediately produce a model that is as complex and as
close to reality as the design budget allows. The aim is to understand the
influence of each layer of complexity. This is a subtle but crucial difference.
You must resist all temptation to launch straight into developing a single
very complex model from the beginning, because this will make it difficult,
if not impossible, to identify modelling errors and properly understand the
relative importance of the different factors.
Another way to think of it is that the objective of any campaign of mod-
elling is to better understand the problem by using a suite of numerical
experiments. Each experiment is an idealised version of reality where some
boundary conditions and parameters are simplified or controlled and a
small number of parameters are varied, one at a time, within realistic lim-
its, to isolate their effect and understand which parameters are important
and which are not. The output will be a set of lining loads that represent the
range of expected values in a range of possible scenarios.
This chapter will cover the following steps as part of a rational approach
to numerical modelling:
7.1 Boundary conditions at the tunnel perimeter

7.2 Boundary conditions at the edges of the model
7.3 Boundary distances
7.4 Element types for the lining and the ground
7.5 Mesh density and refinement
7.6 Modelling groundwater
7.7 Validation and error checking
7.8 Constitutive models
7.9 Interpretation and presentation of results
7.10 3D numerical analysis
This will be followed by a summary and then some example problems.
7.1 BOUNDARY CONDITIONS
AT THE TUNNEL PERIMETER
In a 2D plane strain model, it is not possible to model deformations in the

ground ahead of the face due to excavation and prior to lining placement.
There are a number of ways to simulate the excavation and installation of
the tunnel lining in a 2D model:
7.1.1 Wished-in-place tunnel lining

7.1.2 The convergence-confinement method
7.1.3 Gap method
7.1.4 The grout pressure method
7.1.5 Surface contraction

7.1.6 Core softening
These will be described in the following sub-sections.
7.1.1 Wished-in-place tunnel lining

In a 2D plane strain numerical model, a rectangular block of soil is mod-
elled, as shown in Figure 7.1.
In Step 1, the initial in situ stresses are imposed on the ground, and usu-
ally this is solved to equilibrium to set up the stress matrices needed for
subsequent steps. In some programs this is called the ‘Initial’ or ‘Geostatic’
step. After this step has solved, the displacements and strains are usually
reset to zero. In Step 2, the tunnel is excavated and lined and then solved
to equilibrium.
In this case, the analysis is what is called ‘wished-in-place’. This is
because the ground only notices what has happened when the changes are
calculated. So it is as though the whole length of the tunnel has been exca-
vated and lined instantaneously, as if by magic. This is similar to what hap-
pens in the Curtis-Muir Wood solution, in that the initial in situ stresses
Figure 7.1 Geometry of a 2D plane strain numerical model of a tunnel.

in the ground are applied to a ‘wished-in-place’ tunnel, and only then does
some deformation and redistribution of stress occur due to soil-structure
interaction.
7.1.2 The convergence-confinement method

The main problem with analytical solutions, such as the Curtis-Muir Wood
solution, is that they are 2D and thus cannot explicitly account for redis-
tribution of stresses in the ground prior to lining installation. The same
can be said of a wished-in-place 2D numerical analysis. One way to take
account of this effect would be to use a 3D numerical analysis that mod-
els the advancing tunnel. Since 3D numerical models are complex and
time-consuming, an approximate method was developed that allows some
relaxation of the in situ stresses before inputting them to the 2D analyti-
cal solution or 2D numerical model. This is known as the ‘convergence-
confinement’ method (Panet, 2001). It is based on the concept of a ‘ground
reaction curve’ (Fenner, 1938; Pacher, 1964) and the definition of a
‘confinement loss factor’ by Panet & Guellec (1974).
A ground reaction curve is shown in Figure 7.2. Ground pressure curve ‘I’
shows what would happen if a tunnel were excavated in a very stiff and
strong rock mass – the ground pressure can reduce from the in situ pres-
sure p0 to zero without any support installation. Ground pressure curve ‘II’
is a ground mass that undergoes plastic deformation – the ground pres-
sure would reduce somewhat with increasing radial deformation, but at
some point failure would occur. Theoretically, in an elastic-perfectly plastic
model, the ground pressure should remain constant when plastic failure has
occurred. But some softening may lead to an increase in ground pressure
after failure begins, as indicated by the dashed line. If a tunnel lining were
Figure 7.2 Conceptual model of the ground reaction curve.

installed at ‘A’, then it will be loaded and will undergo its own deformation
until it meets the ground pressure curve ‘II’ at ‘B’. A lining installed earlier,
i.e. closer to the face, would have to support a higher load.
The convergence-confinement method assumes that a certain amount of
convergence has occurred prior to installation of the tunnel lining, which
has reduced the ground pressure that will act on the lining. Convergence
depends on the distance of the section in question from the working face,
on the unsupported length of the heading, and on the stiffness of the sup-
port. It may also be time-dependent as most soils and rocks exhibit time-
dependent creep or consolidation behaviour.
It should be remembered that in using the convergence-confinement
method we are ignoring stresses and strains in the longitudinal direction,
which may be important. In fact, convergence occurs not only because
there is an unsupported length of heading that allows radial movement
but also because extrusion of the face into the tunnel allows radial move-
ment to occur in the ground ahead of the face (Janin, 2017). Therefore, the
convergence-confinement method is always inductive; we tell it how much
we think the ground will converge based on empirical evidence, simple
plasticity solutions or judgement. Convergence (or confinement loss factor)
is always an input, not an output.
The convergence-confinement approach can be used with the Curtis-
Muir Wood or any other 2D plane strain analytical method. It is also used
with 2D plane strain numerical models. In principle, the stress σ applied to
the tunnel lining is given by:
σ = (1 − λ ) σ 0 (7.1)
σ0 is the initial in situ stress value

λ is the ‘confinement loss factor’, between 0 and 1, which simulates
the relaxation of the ground stresses prior to installation of the
tunnel lining
This is, in principle, a simple and straightforward idea. However, it is

important to remember that the convergence-confinement method is based
on the following assumptions:
• the tunnel is deep and the stresses in the ground are isotropic, i.e.
radial stress and radial deformation are uniform all around the tunnel
and do not vary with depth
• the ground is homogeneous (e.g. no layering or variation with depth)
and all aspects of behaviour (e.g. stiffness) are isotropic
• excavation is full-face and the tunnel is circular
The most common approach to using the convergence-confinement method

in a 2D numerical model is shown in Figure 7.3. The initial stresses are set
Figure 7.3 The convergence-confinement method applied in a 2D numerical model.
up, then the elements of soil within the tunnel boundary are removed and
an internal pressure is applied to the boundary equal to the initial stresses
at each node multiplied by (1 − λ). This is solved to equilibrium. In the final
step, the internal pressure is removed, the tunnel lining is installed and the
model is again solved to equilibrium.
The value of the confinement loss factor λ needs to be carefully cali-

brated. It will depend heavily on the excavation method: for example, an
open-face tunnel will have a much higher value of λ than a closed-face tun-
nel boring machine (TBM). This λ value is usually either based on empiri-
cal data and judgement or on calibration to a 3D model of an advancing
tunnel. The λ value is varied until the maximum surface settlement matches
the value expected from case histories of similar tunnels or the maximum
settlement in a 3D model.
Panet (1995) proposed equations based on an axisymmetric elastoplastic
model for determining λ in a deep, homogeneous, isotropic medium:
  mr  2 
λ = α 0 + (1 − α 0 ) 1 −  0
 (7.2)
  mr + d  
0

d is the distance between the point considered and the face

r0 is the radius to the extrados
α0 and m are dimensionless parameters where α0 is the value of λ at
d = 0. Panet (1995) recommends α0 = 0.25 and m = 0.75.
This equation is only appropriate for deep tunnels in homogeneous, iso-

tropic soil. The convergence-confinement method is nevertheless often
used for shallow tunnels, anisotropic ground, heterogeneous ground and
for non-circular tunnels, often inappropriately according to Janin (2017).
Gilleron et al. (2017) found that the value of λ found by calibration to a
3D model could be significantly different to the value of λ from Panet’s
equation for shallow tunnels in heterogeneous, anisotropic soil. Therefore,
Panet’s equation is of limited use for soft ground tunnelling, which nearly
always occurs near to the surface in heterogeneous soil where K0 is not
equal to 1.0.
7.1.2.1 The β-factor method

An alternative to λ is to use what is known as the ‘ β-factor method’ (e.g.
Möller & Vermeer, 2008). β is defined differently to λ in that it repre-
sents the proportion of initial stresses that remain after relaxation, rather
than the proportion lost during deconfinement (compare Equation 7.3
with Equation 7.1). It is also expressed in terms of effective stress rather
than total stress, which makes sense because the pore pressure cannot arch
around the tunnel.
σ ′ = βσ 0′ (7.3)
σ ′ is the effective stress

β is the stress reduction factor
σ0 ′ is the initial in situ effective stress before excavation
7.1.2.2 The target volume loss method

Another variant of the convergence-confinement method of numerical
analysis is where, based on field data from past projects, a target volume
loss is set (e.g. Addenbrooke et al., 1997). The volume loss in the model can
be measured at the tunnel boundary, or in the case of undrained analysis,
it can be measured at the surface, which is a bit simpler. Using this method,
the internal pressure is reduced and then the lining is installed in the same
way as for the convergence-confinement method, but the aim is to achieve
a target volume loss rather than a maximum settlement. This is some-
times called the ‘volume loss method’. When calibrating the convergence-
confinement method to case histories, it makes more sense to use volume
loss, which should be fairly constant for tunnels constructed using simi-
lar methods in similar geology, rather than maximum settlement, which is
strongly dependent on tunnel diameter.
An example of the volume loss method is shown in Figure 7.4. The aver-
age stress in the tunnel lining, as a percentage of the full overburden pres-
sure, is plotted on the y-axis and the volume loss is plotted on the x-axis.
An undrained analysis was used with an elastoplastic soil model. The und-
rained shear strength was set so that the stability number was equal to 2.
In the 2D analysis, several models were run with different values of con-
finement loss factor λ. In the 3D analysis, several models were run with dif-
ferent values of unsupported length P. As P or λ increases, the volume loss
Figure 7.4 2D numerical analysis using the convergence-confinement method compared

with 3D numerical analysis.
increases, and the trends are virtually collinear. Therefore, by using a 3D

numerical model to simulate the construction sequence, a suitable value of
confinement loss factor can be selected for use in 2D analysis. For example,
if the unsupported length were 3 m, then an appropriate value of confine-
ment loss factor λ would be 0.75.
There is evidence to suggest that using the convergence-confinement
method to predict ground movements and lining stresses in a 2D analysis
will be inaccurate. Gilleron et al. (2017) showed that stress paths and plastic
zones in the ground around the tunnel were completely different in a 2D
numerical model when compared to a 3D numerical model of an advancing
tunnel, even when the confinement loss factor λ in the 2D model was cali-
brated to give the same maximum settlement as in the 3D model. Stress paths
are shown in Figure 7.5. The difference in stress paths should be expected to
result in very different patterns of ground movements, because the deforma-
tion behaviour of soils depends strongly on the values of deviatoric stress q
and mean effective stress p′. This criticism probably applies to all methods of
2D tunnel analysis, not just the convergence-confinement method.
Mean effective stress p′ is given by:
σ 1′ + σ 2′ + σ 3′
p′ = (7.4)
3
σ 1′ , σ 2′ and σ 3′ are the principal effective stresses
Figure 7.5 Stress paths in (p′, q) space in the ground 1 m from the extrados of the tunnel
lining versus confinement loss factor in 2D and longitudinal distance from the
face in 3D (redrawn from Gilleron et al., 2017).
Deviatoric stress q is given by:
q=
(σ 1 − σ 2 )2 + (σ 2 − σ 3 )2 + (σ 3 − σ 1 )2 (7.5)
2
σ 1 , σ 2 and σ 3 are the principal total stresses
Jones (2012) found that the stress reduction required to give a specified
value of volume loss depended on the constitutive model and parameters
used for the ground. These different values of β and different constitutive
models and parameters would result in a large variation of predicted lining
forces at the same value of volume loss. Möller & Vermeer (2008) similarly
found that a different value of β was needed to match surface settlements
(typically 0.3–0.4) and lining forces (typically 0.5–0.7) in the same 2D
model. Janin et al. (2015) also warned against calibrating the convergence-
confinement method to a single type of measurement, and identified the
application of a single value of confinement loss factor to all points on the
tunnel perimeter as the cause of significant differences between 2D analysis
and field measurements, whereas a 3D analysis gave much better agree-
ment. Therefore, the sensitivity of the convergence-confinement method to
its inputs should be carefully assessed, particularly if it is to be used for
tunnel lining design. Just because a 2D model can match surface settle-
ments does not mean that the lining forces it predicts will also be reliable,
and vice versa.
The convergence-confinement method generally assumes that the ratio
of horizontal to vertical stress that will be applied to the tunnel lining is
the same as the initial in situ stress ratio. This will lead to strange results
when K0 > 1.0. After installation of the tunnel lining, it will converge
more horizontally than vertically because the horizontal stress will be
larger than the vertical. We know from large numbers of tunnel circu-
larity measurements in London Clay by Wright (2013), that even when
K0 > 1.0, the tunnels almost always squat (the vertical convergence is
larger than the horizontal convergence). Therefore, it is unrealistic, in
soft ground at least, to apply initial stresses with K0 > 1.0 in a 2D model,
since this will predict a mode of deformation completely opposed to real-
ity. For London Clay, Wright (2013) suggests using K0 = 0.7 in simple
models. In a 3D numerical model, it may be possible to use K0 > 1.0,
because the redistribution of stresses ahead of the face can occur in the
model as they do in reality. It may also be possible in a 2D numerical
model to effectively use K0 > 1.0 if the full stress history is modelled
and if long-term drainage of the ground around the tunnel is modelled
(Avgerinos et al., 2018), as squatting may be partly caused by the tunnel
acting as a drain with a higher horizontal than vertical permeability in
the ground around it (Wright, 2013).
In all cases, it is wise to perform sensitivity analyses. For instance, using

K0 = 1.0 will give higher values of axial force in the tunnel lining, but will
only generate bending moments due to increase of stress with depth and
non-circularity of the lining. Therefore, a lower bound value of K0 should
also be used to find a maximum bending moment. Only in very special cir-
cumstances, for example, when tectonic stresses are important and will re-
establish high horizontal stresses during the design life, will it make sense
to use K0 > 1.0 in a 2D analysis.
7.1.3 Gap method

The gap parameter method was introduced by Rowe et al. (1983), though
it is perhaps better described in Rowe & Lee (1992). The ‘gap parameter’
may be thought of as the distance between the crown of the tunnel and
the original position of the ground at that location prior to tunnelling, and
seeks to represent radial deformation and unloading of the ground ahead of
the face and before lining installation in a 2D numerical model. The invert
of the tunnel is assumed to rest on the ground beneath, with no gap.
To use this method in a 2D numerical model, the perimeter of the excava-
tion must be gradually unloaded, in a similar manner to the convergence-
confinement method, by applying nodal forces equal to the initial in situ
stresses and gradually decreasing them until the gap is closed. At the point
that the gap is closed, then the ground and the lining interact in the normal
way. Applying contact rules in this way is actually not as straightforward
as it sounds and most geotechnical numerical modelling programs do not
have a built-in ability to do this.
For a TBM tunnel, the lower limit of the gap parameter will be the differ-
ence between the excavated diameter and the outside diameter of the lining
plus grouted annulus. An allowance should also be made for deformation
ahead of the face.
For a tunnel lined with shotcrete, the gap parameter will represent defor-
mation of the ground ahead of lining installation. In a very simple model
with elastic shotcrete an allowance may also be made for creep and shrink-
age of the shotcrete.
Rowe et al. (1983) also looked at the effect of applying a grout pressure,
though they did this after installation of the lining. A uniform pressure was
applied and it was found to reduce surface settlements and to slightly nar-
row the settlement trough.
7.1.4 The grout pressure method

Möller & Vermeer (2008) introduced the ‘grout pressure method’ for the
simplified analysis of TBM-driven tunnels. In a TBM tunnel in soft ground,
the ground will close around the shield and tailskin, but the boundary con-
dition at the point where the lining is installed is determined by the grout
pressure. Earlier in the same paper Möller & Vermeer had demonstrated
how the convergence-confinement method did not give a good estimate of
surface settlements and by introducing this method they were attempting
to overcome most of the limitations of 2D numerical analysis, at least for
TBM-driven tunnels, without resorting to a full 3D analysis.
Figure 7.6 shows the in situ radial stress distribution applied to a tunnel
when K0 = 0.47 (solid circles), based on a diagram in Möller & Vermeer (2008).
The radial stress is high at the crown, then it decreases around the axis
level as horizontal stresses dominate, then it increases to a maximum
at the invert. When the β-factor method is applied, the distribution is
still strongly influenced by K0 (open circles). When compared to sur-
face settlements from the Second Heinenoord Tunnel, this resulted in an
underprediction of maximum settlement and a settlement trough that
was too wide. However, when the radial stresses around the perimeter
of the excavation were relaxed to the value of the grout pressure (solid
squares), the agreement with the field measurements was exceptionally
good. The agreement was also reasonably good for horizontal ground
movements to the sides of the tunnel, as measured by inclinometers
Figure 7.6 Radial stresses using the convergence-confinement (stress reduction to

β = 0.69) and grout pressure method (redrawn based on Möller &
Vermeer, 2008).
installed in vertical boreholes, and for lining forces and moments back-
calculated from strain gauges installed in the lining.
The grout pressure method seems to be a promising simplified 2D
approach to the design of TBM tunnels, perhaps to be used at an early
design stage when 3D modelling is considered too expensive. Möller &
Vermeer (2008) actually used a combination of the gap parameter method
and grout pressure method, where the overcut and conicity of the shield
were simulated by a gap. Once the gap was closed, the grout pressure was
applied.
7.1.5 Surface contraction

Another modelling method is to allow a ‘surface contraction’
(Brinkgreve et al., 2018). The tunnel perimeter is allowed to reduce by a
specified amount, usually to meet a target volume loss. In this method,
the stresses applied to the lining in the final stage are not necessarily
predetermined.
7.1.6 Core softening

In this modelling method the soil within the tunnel perimeter (the ‘core’)
is softened by reducing its stiffness and/or strength, which allows stresses
to redistribute in the ground around the excavation. In a further step the
lining is installed and the core removed.
7.1.7 Summary
In summary, when analysing a tunnel in 2D, assumptions need to be made
about how to simulate 3D effects. These might include excavation, lining
installation, face pressure and grouting pressure and toolbox items such as
grouted pipe arches or face dowels.
For tunnels in overconsolidated soils with K0 > 1.0 it will rarely make
sense to apply this stress state to the tunnel lining. In these cases sensitivity
analyses should be performed with an upper limit of K0 = 1.0 and a lower
limit of K0 = 0.7.
For open-face tunnels, the best approach seems to be to use the
convergence-confinement method, but it can be misused. There are two
ways to avoid this:
1. Find one or more case studies of real tunnels constructed with similar
methods in similar geology to the one you are designing and try to
match deformations and lining stresses in the model to the measured
deformations and lining stresses in the real tunnel by varying the con-
finement loss factor λ. The paucity of lining stress measurements will
make this difficult to achieve.
2. In the absence of empirical data from real tunnels, use a 3D numeri-

cal analysis of an advancing tunnel to determine the confinement loss
factor to be used in 2D analysis.
Studies of open-face tunnels have also shown that the convergence-

confinement method can either be calibrated to give the correct maximum
settlement, the correct volume loss or the correct lining stresses, but rarely
all at the same time (Möller & Vermeer, 2008; Jones, 2012). Therefore, the
results from 2D design models should always be treated with caution and
with an awareness of their limitations.
For TBM-driven tunnels, the grout pressure method proposed by
Möller & Vermeer (2008) appears to provide a reasonable means of simpli-
fying the TBM excavation process in a 2D model.
7.2 BOUNDARY CONDITIONS AT THE EDGES

OF THE MODEL
In a model the top surface is always left free of any constraints. The lateral
(vertical) boundaries are usually fixed in the horizontal direction but allow
movement in the vertical direction. The bottom (horizontal) boundary is
usually fixed in both the horizontal and vertical directions. These condi-
tions are shown in Figure 7.7.
There is actually no particular reason why the model’s bottom boundary
should be fixed in the horizontal direction (after all the lateral boundar-
ies are not fixed in the vertical direction). In reality the soil on the other
side of these arbitrary boundaries will allow some movement but this will
be resisted by shear stresses, so it will be somewhere between fixed and
free. Möller (2006) argues that due to the tendency for soils to increase in
Figure 7.7 Boundary conditions at the edges of a 2D numerical model showing fixity

in the horizontal direction only for the lateral boundaries and fixity in both
horizontal and vertical directions for the bottom boundary.
stiffness with depth, modelling the bottom boundary as fully fixed, i.e. not
allowing any lateral movement and generating shear stresses, is closer to
reality.
Another possibility is to have a stress applied at the boundary rather than
fixing displacements. These stresses would be set to be equal to the initial in
situ stress. So a stress boundary condition keeps stress constant but allows
displacement, and a displacement boundary condition keeps displacements
at zero but allows stresses to change. Neither of these conditions is more
realistic than the other, as the reality is somewhere in between. But if the
boundary is far enough away from the tunnel, then stresses should remain
approximately constant and displacements should remain at approximately
zero, regardless of whether a stress boundary condition or a displacement
boundary condition is selected.
Symmetry may also be exploited to reduce the size of the model and
hence reduce the computation time. Often, there is a vertical plane of sym-
metry down the centreline of the tunnel, which allows the size of the model
to be halved, as shown in Figure 7.8. This also will make handling the
model and visualising results easier. Whether the left- or right-hand side is
modelled is up to you. For convenience you may wish to model the side that
gives you positive coordinates in your coordinate system.
For deep and/or small diameter tunnels, the change in in situ stress with
depth may be small relative to the magnitude of the stress. In this case, a
quarter of the tunnel could be modelled, and the model may not extend
to the surface, but have a stress applied to it at a boundary, as shown in
Figure 7.9. This is not a common situation for tunnels in soft ground, which
tend to be relatively shallow, but we will do a variant of this in the valida-
tion section below to mimic the boundary conditions of the Curtis-Muir
Wood solution.
Figure 7.8 2D numerical model boundary conditions with vertical plane of symmetry
on tunnel centreline.
Figure 7.9 2D numerical model boundary conditions exploiting two planes of symmetry.
Where lining elements meet a symmetry boundary, they should have dis-
placement fixity normal to the boundary, but be allowed to move parallel
to the boundary. They should also have rotational fixity about the tunnel
longitudinal axis direction to ensure that bending moments are mirrored.
This is shown in Figure 7.10. The condition that rotation is prevented about
Figure 7.10 Boundary conditions for a tunnel lining on a symmetry boundary in a 2D

numerical model. Displacements in the x, y, z directions are denoted by ux,
uy, uz and rotations about the x, y and z axes are denoted by θx, θ y and θz.
the y-axis will ensure that the lining at this point will remain perpendicular
to the symmetry axis, thereby ensuring that the correct deformation mode
is produced and the correct moment is generated (My in Figure 7.10).
Note that when beam or shell elements are used, moments and forces in
each beam or shell element will be defined by local axes that will almost
always have a different orientation to the global axes. In some programs this is
made clearer by labelling the local axes ‘1’, ‘2’ and ‘3’ instead of ‘x’, ‘y’ and ‘z’.
Also, in some software programs, ‘Mx’ or ‘My’ is not the moment about
the local x-axis or y-axis, but the moment along it (i.e. it would require
reinforcement parallel to the x-axis or y-axis). The only way to be sure is
to read the manual and it is always good practice when starting to use new
software to run some simple tests with single elements or small numbers
of elements. This is all part of building your understanding of the soft-
ware and the problem being modelled by beginning with simple models and
incrementally adding more complexity.
7.3 BOUNDARY DISTANCES
In reality, of course there are no boundaries in the ground, and the lateral
extent of the soil can be considered, to all intents and purposes, infinite.
There are ways to use infinite boundary elements on the outside of a model
(e.g. Reichl et al., 2003), but this is not common practice and these elements
are not available in most commercially available numerical analysis pack-
ages. Therefore, in a standard continuum model, the boundaries must be
finite and the aim is to make the model as small as possible without intro-
ducing significant errors. This can be achieved by varying the boundary
distance and using the shortest distance that has an acceptable error. This
will depend on the focus of the modelling, the geometry and depth of the
tunnel, and the constitutive models used.
Möller (2006) recommended using a lateral boundary distance that
resulted in surface settlements at the model boundary that were less than
1% of the maximum surface settlement above the tunnel centreline in the
model. Based on this criterion he developed Equation 7.6 for his models:
 H
w = 2D  1 +  (7.6)
 D
w is the distance from the tunnel centreline to the lateral boundary

H is the cover above the tunnel (the vertical distance from the tunnel
crown to the ground surface)
It is important to remember that the optimum boundary distance will

depend on the problem geometry and the constitutive model of the soil,
and so it should be verified for each soil model used and each geometric
configuration (i.e. for different tunnel depths or tunnel sizes). For example,
Franzius et al. (2005) found that the lateral boundary distance needed to be
larger for higher values of K0.
If there is a change to a much stiffer geological stratum below the tun-
nel, this is often taken as the bottom boundary, because it will be rela-
tively rigid. For situations where soft ground extends well below the tunnel,
Möller (2006) recommended a distance (from tunnel invert to the bottom
boundary) that resulted in less than 2.5° rotation of principal stresses at the
boundary after tunnel construction, which for his models was between 1.3
and 2.2 times the diameter D.
Möller’s boundary distance relationships can be used as a first approxi-
mation. It is recommended that this first approximation is then tested by
varying the boundary distance to higher and lower values to find the dis-
tance that gives an acceptable error in the required outputs. If the only
concern is lining stresses, the boundaries may not need to be as far away as
for ground movements or surface settlements (Jones, 2007).
Sometimes, it is found that increasing the boundary distance beyond a
certain point will actually begin to increase the error. For example, the
surface settlement may be zero at the boundary when the lateral boundary
distance w = 30 m, but may be non-zero at the boundary when w = 50 m.
This often happens if the number of elements is remaining constant, such
that increasing the boundary distance is increasing the size of the elements,
thereby reducing accuracy. Therefore, while varying the boundary distance
it is important to try to keep the mesh density constant. This problem also
occurs due to a sort of ‘mirror effect’ where displacements appear to be
reflected in the bottom boundary and then have an effect on the lateral
boundary. It can then be quite frustrating to find a compatible pair of lat-
eral and bottom boundary distances as changes to the bottom boundary
distance will affect the acceptability of the lateral boundary distance and
vice versa. This effect usually is only a problem when linear elastic soil
models are used, and when soil stiffness is not increasing with depth. If the
bottom boundary is fixed in the horizontal direction, the principal stresses
may be rotating at the boundary (Möller, 2006) – if this appears to be the
case, consider increasing the bottom boundary distance and/or allowing
displacements in the horizontal direction at the bottom boundary.
Some practitioners fix the lateral boundary in all directions, and this of
course results in zero settlement at the boundary. However, if the bound-
ary is too close to the tunnel it will be reducing settlements throughout the
model and there is no way of knowing this is happening.
7.4 ELEMENT TYPES FOR THE LINING

AND THE GROUND
Modern finite element software packages, particularly those that special-

ise in geotechnics, usually have elements suitable for modelling soil. Often
there is a choice between more accurate higher order elements and less
accurate lower order elements. ‘Higher order’ means that the elements have
more nodes and more degrees of freedom, and therefore can adopt higher
order deformation modes (e.g. the side of an element can follow a quadratic
curve rather than a straight line). This may be important where there are
high stress gradients, where a larger number of lower order elements would
be needed to get the same accuracy as a smaller number of higher order
elements. There is a trade-off here between accuracy and simulation time,
because higher order elements mean more equations in the matrix that need
to be solved, but it may be more efficient than having a larger number of
lower order elements.
The way forward is to test the different elements you plan to use in a
validation exercise similar to the problem you wish to solve. An excellent
example of this kind of validation may be found in Pound (2006), where
the number of solid elements needed to model a sprayed concrete lining is
optimised by modelling a beam in flexure and comparing the stresses and
displacements to simple hand-calculations.
When modelling the tunnel lining, either 1D beam elements or 2D solid
elements can be used. Beam elements are usually better at modelling bend-
ing, but the constitutive models may be limited to simple elastic or elasto-
plastic behaviour. There is also the issue of where to place beam elements. To
simulate the soil-structure interaction, they are usually placed at the perim-
eter of the excavation; the extrados. But this will result in an overestimate of
bending moments, as effectively the radius to the centroid of the beam has
been increased by half the lining thickness, as though you had increased the
span of a bridge while keeping the overall load the same. For an elastic tun-
nel lining the overprediction can be estimated using the following equation:
M* r02
= (7.7)
M r0 ( r0 − t )
M* is the bending moment calculated by the numerical model

M is the ‘true’ bending moment
t is the thickness of the lining
7.5 MESH DENSITY AND REFINEMENT
Using a larger number of smaller size elements will generally be more accu-
rate. However, the model will use more computer memory and will take
longer to solve. As with the choice of element types and boundary dis-
tances, this can be optimised by varying the mesh density to obtain an
acceptable error for the outputs we are interested in.
A higher mesh density will improve accuracy in areas with high stress
gradients, which is usually in a zone around the tunnel. Therefore, it is not
necessary to use the same mesh density everywhere in the model, and where
stress gradients are small the mesh density can be reduced substantially.
This is known as ‘mesh refinement’.
For obtaining accurate lining stresses, Möller (2006) found that higher
mesh density was required in 3D models than in 2D models. This was
because there was significant arching in the front-to-back direction close to
the face of the tunnel, resulting in high stress gradients in the longitudinal
direction. This is a real phenomenon but cannot occur in a 2D model.
7.6 MODELLING GROUNDWATER
Soil is made up of solid soil particles, water and air. In tunnel modelling
we normally ignore the air and model the soil as either ‘fully saturated’ or
‘dry’, though there may be situations where partial saturation is important.
In permeable soils, we can assume that below the groundwater table the
soil is fully saturated. Therefore, unless there is groundwater flow, the pore
pressures will increase hydrostatically from the groundwater table down.
Above the water table the soil will probably be moist (‘partially saturated’),
and this can be taken into account in the unit weight of the soil, but essen-
tially we will assume it is dry in terms of its pore pressure, which we will
assume to be zero (e.g. Itasca, 2006).
In permeable soils, any excess pore pressures generated due to stress
changes will dissipate quickly relative to the speed of construction.
Therefore, with no excess pore pressures, behaviour is governed by the
interaction between soil grains, i.e. the effective stress. We can model this
by using an analysis option that specifies, rather than calculates, the pore
pressures. If we want to include pore water in the model explicitly, we can
set the fluid bulk modulus of the pore water Kf to zero. This will ensure that
any changes in pore volume do not induce changes in pore pressure.
Due to capillary suction in low-permeability soils (clays), we generally
assume that the clay is saturated both above and below the water table and
will therefore have the same unit weight. However, as with permeable soils,
unless there is a groundwater flow, the pore pressures will increase hydro-
statically from the groundwater table down, and will be assumed equal to
zero above the water table.
In low-permeability soils, excess pore pressures generated due to stress
changes will not dissipate during construction, but may take months,
years or decades. Therefore, there is a short-term ‘undrained’ behaviour,
followed in the long term by a ‘drained’ behaviour. It is important that
the software prevents the generation of negative pore pressures below
−100 kPa, because in reality at this pressure a vacuum will form. Even
though it applies to all geotechnical problems, in some software programs
this cap on negative pore pressure needs to be specified by the user. This is
because checking for this criterion takes computational effort.
Note that it is standard practice to express pore pressures relative to

atmospheric pressure, which at sea level on Earth is approximately 100 kPa.
Therefore, when pore pressure is expressed as −100 kPa, it is really 0 kPa and
a vacuum has formed, an empty space with nothing in it, and the pressure
cannot go any lower. This is called the ‘cavitation pressure’. Incidentally, this
is why no water pumps can ever suck more than a 10 m head, but they can
push a lot higher if the pump is near the bottom of the pipe.
7.6.1 Undrained behaviour

During the timescale of construction, most soils with sufficient clay content
can be considered to be undrained. Whether a soil behaves in a drained
or undrained manner depends on the permeability of the soil, as well as
the scale and speed of construction. A rule of thumb is that a soil with a
permeability lower than 10 −7 to 10 −8 m/s will behave in an undrained man-
ner, and soils with a higher permeability will behave in a drained manner
(Anagnostou & Kovári, 1996).
If groundwater cannot flow significantly during the timescale of con-
struction, then pore pressure changes are temporarily locked in and we call
this state ‘undrained’. It is often useful to model the construction phase
as undrained and the long term as drained, where all pore pressures have
reached either a hydrostatic equilibrium or a steady-state.
During tunnel excavation, the ground around a tunnel experiences a
stress decrease in the radial direction, and a stress increase in the circum-
ferential direction. This results in both temporary and permanent changes
to the total stresses. If excess pore pressures do not have time to dissipate,
the change in total stresses will mostly be experienced by the pore water,
because although water can support very little shear, it is relatively incom-
pressible compared to the soil particles and will therefore resist changes
in volume. Undrained behaviour is often referred to as ‘constant volume’,
because although an undrained soil may undergo shear deformations,
the volume change depends on the pore water, which may be assumed
to be incompressible and hence the overall volume of the soil will remain
constant.
Elastic behaviour can be fully described by bulk modulus K and shear
modulus G. Shear modulus is the stiffness in shear and is equal to the shear
stress divided by the shear strain. Bulk modulus is the volumetric stiffness
and is equal to the pressure divided by the volumetric strain. Alternatively,
Young’s modulus E and Poisson’s ratio ν may be used instead of K and
G, but they do not separate volumetric and shear behaviour in the same
convenient way. Bulk modulus and shear modulus are related to Young’s
modulus and Poisson’s ratio by the following expressions:
E
K= (7.8)
3 (1 − 2ν )
E
G= (7.9)
2 (1 + ν )
K is the bulk modulus

G is the shear modulus
E is the Young’ modulus
ν is the Poisson’s ratio
Shear modulus G is the same for both drained and undrained behaviour,
but bulk modulus K can have different values for drained and undrained
behaviour. This is because pore pressures are only affected by volumetric
changes, which is why separating shear and volumetric behaviour using
G and K is so elegant. Because drained and undrained shear modulus are
equal, the following equation results:
Eu E′
G= = (7.10)
2 (1 + ν u ) 2 (1 + ν ′ )
E′ and Eu are the drained and undrained values of Young’s modulus,

respectively
ν ′ and ν u are the drained and undrained values of Poisson’s ratio,
respectively
If we were to assume that undrained Poisson’s ratio is equal to 0.5, it fol-

lows from Equation 7.10 that:
3E′
Eu = (7.11)
2 (1 + ν ′ )
Numerical modelling programs cannot use a Poisson’s ratio of 0.5, as

it results in numerical difficulties when solving the matrix equations (it
tries to divide by zero), but usually a value slightly lower than 0.5, such
as 0.495 or 0.499, can be used instead without introducing significant
error.
Now we are going to look in more detail at how to model undrained
behaviour. The notation we will use is as follows:
K′ is the effective or drained bulk modulus, i.e. the bulk modulus of the
soil particle assembly or ‘soil skeleton’
Kf is the bulk modulus of the pore fluid, often taken as the value for pure
water at room temperature, which is 2.0 GPa (note that this value can
be significantly reduced by the presence of dissolved air or air bubbles
in the pore water)
Ke is an equivalent bulk modulus for the soil skeleton and pore fluid
combined (more explanation to follow)
There are two ways of modelling undrained behaviour:
Option 1 Pore water can be ignored and the equivalent bulk modulus Ke
set to a high value relative to the drained bulk modulus K′,
such that the Poisson’s ratio is close to 0.5. With Ke = 100 K′,
the Poisson’s ratio will be greater than 0.495, as recommended
by Potts & Zdravković (1999). This is known as a ‘total stress
analysis’ because we are not treating the effective stresses and
pore pressures separately, but lumping them together.
Option 2 Pore water can be modelled by setting a fluid bulk modulus Kf
to a realistic value (e.g. 2 GPa, which is the value for pure
water at room temperature) and no groundwater flow is
allowed. This allows the explicit use of effective stress
parameters, but increases the runtime of the model.
Option 1 aims to make the Poisson’s ratio as close to 0.5 as possible

without introducing numerical instability. However, in reality, undrained
behaviour is never strictly constant volume and undrained volume change
does depend on K′ for most soils.
Bishop & Hight (1977) found that the undrained Poisson’s ratio
approaches 0.5, only in the case where K′ is very low (as, for example, in a
normally consolidated soil under low effective stresses) and the soil is fully
saturated. Bishop & Hight (1977) reported the undrained Poisson’s ratio of
several natural soils to be between 0.470 and 0.487. Therefore, it is simpler,
and more consistent, to use Option 2 and to estimate the values of fluid
bulk modulus Kf and drained bulk modulus of the soil K′ and to use those
in a model incorporating pore pressures, rather than trying to input values
that have no physical meaning.
Option 1 also runs into difficulties because it does not calculate pore
pressures. When this option is selected in a finite element program, it may
be possible to plot pore pressures and effective stresses as outputs, but these
pore pressures are based on the assumption that the final long-term equi-
librium state after all excess pore pressures are dissipated will be the same
as the initial pore pressure distribution. If the tunnel acts as a drain on the
ground around it, or if the long-term groundwater situation consists of a
steady-state groundwater flow, then the final pore pressure distribution is
not the same as the initial and the results will be wrong. ‘Excess pore pres-
sure’ should be defined as excess to the final state, not the initial state; it is
what will be dissipated over time by groundwater flow, causing consolida-
tion or swelling. A detailed explanation of this may be found in Tan et al.
(2005).
Option 2 allows effective stiffness parameters to be used, for which the
results of laboratory tests are often better suited, and it also allows long-
term effects of groundwater flow and dissipation of excess pore pressures to
be modelled. Volume changes induce stresses in both the soil skeleton and
the pore water, but because the bulk modulus of the pore water is much
higher than the bulk modulus of the soil skeleton, most of the stress change
will be represented by a change in pore pressure. Most of the more sophis-
ticated soil models use effective stress parameters.
The soil’s failure criterion can also be expressed in terms of either
drained (c′ and ϕ ′) or undrained (cu) parameters. This will be covered in
Section 7.8.
7.6.2 Long-term effects

When a geotechnical problem is assumed to be undrained during con-
struction, this means that in the long term we can expect the excess pore
pressures to dissipate, causing the soil to experience changes in effec-
tive stresses. This will also result in strains and hence displacements.
Therefore, as designers, we need to calculate these long-term effects to
ensure that our design remains safe and serviceable for the full design
life of the structure.
For soils that are considered undrained in the short term, we need to
also calculate what happens in the long term by allowing pore water to
flow in the model. This will dissipate the excess pore pressures and allow
calculation of long-term stresses, strains and displacements. There are two
ways this can be done: using either an ‘uncoupled’ or a ‘coupled’ fluid-
mechanical analysis.
There are also two main types of problem: where the perturbation is
mechanical, and where the perturbation is driven by a pore pressure change
(Itasca, 2006). Tunnel construction is the former type, whereas draw-down
from a well would be an example of the latter.
Most finite element modelling programs only allow ‘uncoupled’ fluid-
mechanical analysis. ‘Uncoupled’ means that modelling pore water flow
and modelling mechanical behaviour are two separate calculation stages
performed one after the other. ‘Coupled’ analysis is where pore water
flow and mechanical behaviour are calculated together. Therefore, as pore
water flows, pore pressures change and this causes the effective stresses,
strains and displacements to change, which in turn have an effect on pore
pressures.
In some cases, particularly where the problem is driven by a pore pressure
change, an uncoupled analysis may be considered sufficient. An uncoupled
analysis will have the following stages:
1. An undrained calculation stage using effective stress parameters

(Option 2 above). No fluid flow is allowed. This gives the short-term
stresses and displacements for design.
2. A fluid flow calculation stage, with Kf set at a realistic value. No
mechanical calculation is performed, so the effective stresses do not
change. Flow continues either to simulate a period of time, or for a
long-term analysis, flow continues until either equilibrium or a steady

state is reached.
3. A further mechanical calculation stage to calculate the effect of the
change of pore pressures on the soil skeleton. No fluid flow is com-
puted and Kf is set to zero. This provides the long-term stresses and
displacements for design.
Uncoupled analysis is only reasonably accurate for elastic materials because

plastic materials are path-dependent (Itasca, 2006). ‘Path-dependent’
means that the stress path the soil takes is important, not just its final stress
state. Therefore, to calculate long-term behaviour in soft ground, a coupled
analysis should be used, with the following stages:
1. An undrained calculation stage using effective stress parameters

(Option 2 above). No fluid flow is allowed. This gives the short-term
stresses and displacements for design.
2. A coupled flow-mechanical calculation stage. Depending on the soft-
ware used, a lower value of Kf may be used to reduce the computa-
tion time without a significant impact on the accuracy (Itasca, 2006).
Guidance should be given in the manual for the software you are
using. This stage provides the long-term stresses and displacements
for design.
7.7 VALIDATION AND ERROR CHECKING
Numerical models are based on many assumptions and are approximate by

nature. Validation is a crucial part of the modelling process to ensure that
the approximation is good enough and the results can be trusted. There are
three main ways to validate a numerical model, which will be described in
the following sections:
7.7.1 Comparison with an analytical solution

7.7.2 Validation by comparison with a laboratory test or experiment
7.7.3 Validation by comparison with a case history
It is always best to do as much validation as possible and of all three dif-

ferent types. Comparison with an analytical solution will tell us that cer-
tain aspects of the numerical model are making correct predictions, but
since analytical solutions are very simple by nature, this cannot validate
all aspects of the model. Comparison with a laboratory soil test is use-
ful for validating the constitutive models, but may not represent the true
ground mass behaviour. Only comparison with a case history will validate
all aspects, but the case histories that are available may not be very similar
to the tunnel we are designing.
Even when we are designing a tunnel in well-known geology with con-

struction methods we have used many times before, we can never com-
pletely validate a design model, as there are always uncertainties and
every design is a unique prototype. We cannot build a tunnel in exactly
the same place as we have built one before. Therefore, we need monitor-
ing during construction to verify the assumptions and simplifications
made in design, and also to provide a case study for future designs to be
validated against.
7.7.1 Comparison with an analytical solution

To compare a numerical model with an analytical solution, it is possible
to get very good agreement if the boundary conditions and constitutive
models are replicated as closely as possible. To replicate the Curtis-Muir
Wood solution, for instance, we need to simulate a biaxial stress field.
This is best done by building a model of a quarter of the tunnel with two
planes of symmetry and applying the axis level vertical and horizontal in
situ stress values (σ v and σ h) to the top and side of the model, as shown
in Figure 7.11. To replicate the analytical solution more closely, set the
unit weight of the soil and the lining to zero. This should result in close
to an elliptical deformation with hoop forces and bending moments that
are quite similar to the analytical solution.
Validation using an analytical solution can also be used to check the ele-
ment types and mesh density. The following worked example shows how
this could be done using the Curtis-Muir Wood solution.
Figure 7.11 Boundary conditions for validation of a numerical modelling program using

the Curtis-Muir Wood analytical solution.
WORKED EXAMPLE 7.1 VALIDATION

USING THE CURTIS-MUIR WOOD ANALYTICAL SOLUTION
This Worked Example goes through the process of validation for a
2D numerical model of a 6 m diameter tunnel in soft ground by mim-
icking the boundary conditions of the Curtis-Muir Wood analytical
solution.
A 2D plane strain model of a block of soil 50 m by 50 m is used. The
water table is set to the bottom of the model so that no water pressures
are calculated.
A linear elastic soil constitutive model is used with Young’s modulus
E = 80 MPa and Poisson’s ratio ν = 0.2. The unit weight of the soil is set
to zero.
A tunnel lining is placed in one corner of the model, as shown in
Figure 7.11, as a quarter circle arc of radius 3 m.
The lining is represented by beam elements with Young’s modulus
E = 20 GPa and Poisson’s ratio ν = 0.3. The unit weight is set to zero.
Boundary conditions are set as shown in Figure 7.11.
Apply σ v = σ h = 500 kPa to the free top and side boundaries as shown
in Figure 7.11.
Set up an initial construction phase, where the 50 m × 50 m block of
soil is intact. Then set up an excavation phase, where the soil elements
within the tunnel boundary are deactivated and the tunnel lining beam
elements are activated.
Run the numerical model. See Section 6.2 and use the Curtis-Muir
Wood equations to hand-calculate the hoop force and bending moment
in the lining. If the numerical model is working well, its results should be
close to our Curtis-Muir Wood solution values, which are:
• 1457.99 kN/m uniform hoop force around lining

• 0 kNm/m uniform bending moment around lining
Now change the horizontal stress σ h at the boundary to 350 kPa. Then
solve the numerical model and the Curtis-Muir Wood solution again.
Compare the numerical model results to the Curtis-Muir Wood solution:
• Maximum hoop force (at axis) 1436.68 kN/m

• Minimum hoop force (at crown) 1041.90 kN/m
• Maximum bending moment (at crown) 68.33 kNm/m
• Minimum bending moment (at axis) −68.33 kNm/m
In this case, the validation is checking that the numerical modelling pro-
gramme is configured correctly, and that the element types and mesh den-
sity are appropriate for modelling a tunnelling problem. It is not possible to
use a model such as this to verify that boundary distances are sufficient for
accurate modelling of ground movements. Also, only linear elastic behav-
iour of the ground and tunnel lining can be validated – as soon as we
make these constitutive models more sophisticated, the numerical model
will diverge from the analytical solution.
A further worked example describing the validation of a 3D numeri-
cal model of a shaft-tunnel junction using the Kirsch analytical solution is
shown later in Section 7.10.3.
7.7.2 Validation by comparison
with a laboratory test or experiment
Constitutive models describe the stress-strain behaviour of soil or the tun-
nel lining material. They are simple mathematical rules in the form of equa-
tions and algorithms.
Usually, the first step is to compare the constitutive model with dif-
ferent laboratory tests to ensure that the most important aspects of
behaviour are modelled well. The laboratory tests should involve differ-
ent stress paths, and sometimes different loading durations and stress
histories, and should be relevant to the stress levels, stress paths, stress
histories and magnitudes of strain expected in the actual soil during and
after construction.
To validate the use of a constitutive model for the ground or tunnel lining
in a numerical model, a laboratory test can be simulated in the numerical
modelling program and compared to the behaviour observed in a real labo-
ratory test. If the constitutive model is formulated correctly and adequately
describes the soil behaviour, then the agreement between the two should
be good.
Some numerical modelling programs allow this to be done using a built-
in laboratory test module. If not, then in any case this is easy to model as
laboratory tests have simple geometry. A single element could be used, but
for a triaxial test Simpson et al. (1979) found that this did not work well
because a triaxial specimen does not fail in either a plane strain or axi-
symmetric manner. Therefore, a 3D cylindrical model geometry should be
used with the same dimensions as the test. The boundary conditions can
be replicated in the numerical model. For example, a triaxial test can have
a rigid top or bottom boundary, a cell pressure applied around the model,
and either a pressure (for a stress path triaxial) or a displacement (for a
standard triaxial) applied to the top or bottom, whichever is not fixed in
the vertical direction.
Another method of validation for constitutive models is to compare
their theoretical stress-strain behaviour, calculated by hand, with the
outputs of a numerical model. This validates the implementation of the

constitutive model in the software. As long as we know that the theoreti-
cal behaviour, as expressed by the constitutive model’s equations, repre-
sents the behaviour of the soil in the laboratory tests, then this can be an
effective validation.
WORKED EXAMPLE 7.2 VALIDATION

USING A NUMERICAL MODEL
OF A LABORATORY TEST
This is a case study, rather than a worked example, showing the proce-
dure and results of validation of a nonlinear ‘Jardine’ constitutive model
in FLAC3D using a cylindrical model of a triaxial test. The aims of the
validation were:
• to ensure that the stress increment size was not important (i.e. that
FLAC3D could follow the correct stress path to failure even when
the stress increment was large)
• to reduce the calculation time without affecting accuracy by opti-
mising the solution control method, such as varying the number of
steps between stiffness updates or varying the convergence crite-
rion (the ‘mech. ratio’)
• to check that the constitutive model was working correctly
Stress increment size

The results are shown in Figure 7.12. Graph (a) shows how the stiffness
decreases as strain increases and Graph (b) shows the deviatoric stress ver-
sus axial strain relationship. The solid lines show the theoretical relation-
ship from the Jardine constitutive model equations and the dashed lines
show the results of a cylindrical triaxial test model in FLAC3D. The agree-
ment is very good.
In order to check the effect of the stress increment size on the
nonlinearity of the stress path, a deviatoric stress of 160 kPa was
approached in various increments of stress. The difference between
the value of strain at 160 kPa using 4 kPa increments and 40 kPa
increments was negligible. There was a slight difference between these
values and the strain calculated in one increment of 160 kPa, but this
error was still very small, less than 5%. It seems that due to the way
FLAC3D approaches a solution by using the velocity of gridpoints, the
nonlinear stress path was followed almost perfectly.
Figure 7.12 Comparison between theoretical Jardine curve and results from

a FLAC3D model of a triaxial test using a Jardine constitutive model
(Jones, 2007). Graph (a) shows how the stiffness decreases as strain
increases and Graph (b) shows the deviatoric stress versus axial strain
relationship.
Frequency of stiffness updates

The Jardine constitutive model gives a value for the bulk and shear stiff-
ness based on the current value of the strain tensor at each point in the
model. Therefore, the stiffness must be updated regularly as the strain
tensor values change in the model. Each time the stiffness is updated,
the constitutive model’s algorithm must cycle through all the zones in
the model calculating the stiffness from the strain increment vector. This
takes 2–3 seconds, whereas each calculation step takes about 1 second.
So, if the stiffness must be updated at every step, it will have a significant
impact on the total runtime.
Figure 7.13 shows how the number of steps between updates was var-
ied with the aim of optimising the process, for a model of a laboratory
triaxial test. Updating the stiffness every ten steps results in an error of
approximately 5%. When applied to the much larger 3D model of an
advancing tunnel, the runtime went from 3 days when updating every
10 steps to 10 days when updating every step, so this really was a critical
aspect to optimise.
Figure 7.13 Optimisation of number of steps between soil stiffness updates

(Jones, 2007).
7.7.3 Validation by comparison with a case history

Comparing a numerical model with a case history enables us to bring
together all the aspects of the model and validate them together. This
should only be done once boundary conditions and distances, element
types and mesh density, and constitutive models, have been validated and
checked individually. If the model can replicate those aspects of the case
history that we are interested in, this will give us confidence that when we
use the same modelling approach to design a new tunnel, the results should
be reasonably accurate.
Validation to a case history also allows a model to be calibrated – in
other words, for the key parameters used in the model, such as soil param-
eters used in the soil constitutive model or the deconfinement factor (see
Section 7.1.2 for an explanation of deconfinement factor as part of the
convergence-confinement method) used in a 2D plane strain model, to
be varied within realistic limits to find the best value to use. This can only
be done if the case history is in similar geology and involves a similar con-
struction method to the new design situation.
Well-documented case histories are immensely valuable for validation of
numerical models. This means that some case histories are used a lot and
we can have some fun comparing different numerical modelling approaches
to the same set of data (see for example Jones, 2013).
There are many papers that describe a numerical model, present
results, and then compare these modelling results to field measurements
during construction. Invariably the predictions are very good. Partly
Table 7.1 Classes of prediction (Lambe, 1973).

Prediction type When prediction made Results at time prediction made
A Before event -
B During event Not known
B1 During event Known
C After event Not known
C1 After event Known
this is ‘publication bias’, where positive results are more likely to be

written up and more likely to be accepted for publication. But often it
is because the model has been calibrated after construction to give as
good a match to the field measurements as possible. In this case it is
not a prediction at all, but a back-analysis. This is a crucial difference,
easily missed.
Lambe (1973) proposed a classification system for types of predictions,
which is shown in Table 7.1. Every time we are designing a new structure,
we are doing a Class A prediction. However, most papers written about
‘predictions’ are actually Class C1 predictions. Lambe describes Class C1
predictions as ‘autopsies’. He wrote, “… one must be suspicious when an
author uses type C1 predictions to ‘prove’ that any prediction technique is
correct”.
The Jubilee Line Extension’s running tunnels under St James’s Park in
London, UK, have been the subject of a lot of Class C1 back-analyses. This
is because a large amount of high-quality measurements of surface and sub-
surface ground movements was gathered by a team from Imperial College,
both during construction and into the long term. A review of many of these
attempts to obtain good agreement with the field measurements is given in
Jones (2013).
7.8 CONSTITUTIVE MODELS
So far in this book we have focussed on modelling using linear elastic soils
with a Mohr-Coulomb failure criterion and linear elastic tunnel linings,
because this simple behaviour is widely understood by civil engineering
graduates (the target audience of this book) and allows analytical solutions
to be used and the first steps of numerical modelling to be understood.
It should always be remembered that these constitutive models are gross
approximations. True soil behaviour is very complex and nonlinear (e.g.
see Chapter 4 of Potts & Zdravković, 1999). The same can be said of true
shotcrete or concrete behaviour. To have any hope of accurately predicting
ground movements and tunnel lining stresses, we need to use more sophis-
ticated constitutive models.
7.9 INTERPRETATION AND PRESENTATION

OF RESULTS
First of all, think about the intended audience of your report. Is it the client?
An internal/external checker? Is this a scientific paper on numerical model-
ling? It is important to always provide sufficient information so that some-
one else can replicate your work independently, understand the reasons for
all the steps you have taken, have confidence that you have approached the
problem in a rational and efficient manner, and have confidence in your
results.
Try to plot graphs when presenting results rather than hundreds of con-
tour plots. They are pretty, but they only give the reader a qualitative view
of the pattern of behaviour. This can be helpful in moderation but think
carefully about what is the purpose of each table/graph/plot and ensure
they are all discussed in the text and are important to whatever point you
are making. Working out how best to present the results to make the points
you want to make will take some time and experimentation, and is an itera-
tive process.
7.10 3D NUMERICAL ANALYSIS
There are two main types of 3D numerical models used for soft ground tun-
nel design; a 3D numerical model of an advancing tunnel, and any other 3D
model, for example of a tunnel-tunnel junction or a shaft-tunnel junction.
3D numerical analysis is also used for modelling complex details in the
design of segmental linings. This will be covered in Chapter 9.
A big advantage of 3D analysis over 2D is that we do not need to make
any assumptions about what happens ahead of the face, because it is mod-
elled explicitly. On the other hand, a big disadvantage of 3D analysis is
the larger model size and hence increased time to build, debug and run the
model, and increased time to interpret and present the results.
As always, start with the simplest methods first and gradually increase
complexity. Thus, simple hand calculations followed by analytical solu-
tions, followed by 2D numerical analysis, followed by 3D numerical analy-
sis. Thus, by the time we are ready to begin the 3D numerical analysis, we
have understood the problem, have confidence in the methodology, and
have a good idea what the results will tell us.
7.10.1 Modelling an advancing tunnel in 3D

In 3D, each advance of the tunnel can be modelled by an excavation stage
followed by a lining stage. In fact, since the lining will not experience any
increment of load until the next advance is excavated, these two stages can
often be grouped together in one calculation step. Exceptions to this rule
would be if you are modelling any time-dependent processes that occur

between installing the lining and excavating the next step, such as consoli-
dation or creep of the soil.
Regarding the boundary conditions, these will be similar in principle to
the 2D plane strain model described in Section 7.1.7. The bottom bound-
ary will be fixed in all directions and the vertical boundaries (comprising
the lateral boundary, a lateral symmetry boundary if used, and the start
and end boundaries) will be fixed in the direction normal to each plane, as
shown in Figure 7.14.
Regarding the lateral and bottom boundaries, their distance should be
set in a similar way to a 2D plane strain model (see previous Section 7.3). As
a starting point, Möller (2006) found the lateral boundary in a 3D model
needed to be the same distance as for the 2D model (c.f. Equation 7.6 in
Section 7.3). The bottom boundary, however, could be closer at 1.1–1.45D
below the invert of the tunnel. These distances should be checked and fine-
tuned in each new model by varying the distances to achieve an acceptable
level of accuracy.
Unlike a 2D numerical model, in a 3D model the tunnel must start some-
where. Usually it begins at a fixed boundary and is incrementally advanced
through the model. This means that the first advances will be affected
Figure 7.14 Boundary conditions for a 3D model of an advancing tunnel with a vertical

symmetry boundary on the tunnel centreline. Displacements in x, y and z
directions are denoted by ux, uy and uz, respectively.
Figure 7.15 Achieving a steady state in a 3D tunnel model in terms of centreline surface

settlement (redrawn based on Möller, 2006).
by the presence of the starting boundary behind, as shown in Figure 7.15.

The tunnel face will also be affected as it approaches the end boundary.
Therefore, the model needs to be long enough that a kind of steady-state
situation occurs in a central zone of the model where the start and end
boundaries are not affecting the results and the tunnel face is sufficiently
far ahead that it is no longer affecting the lining forces, stresses, strains and
pore pressures. This central steady-state zone needs to be long enough to
allow a normal ground behaviour ahead of and behind the face, and long
enough to demonstrate that a steady-state has been achieved.
Figure 7.15 shows an example of achieving a steady-state in a 3D tunnel
model in terms of centreline surface settlement. As the position of the face
of the tunnel advances beyond 80 m, the settlements in the steady-state
zone should be unaffected, and these can be taken as the final short-term
settlements. If we are also interested in lining stresses, then we need to plot
a graph of bending moments or axial forces in the lining in the same way,
to ensure a steady-state is achieved.
Möller (2006) also found that the first calculation stage from the start
boundary has a big effect on the model before reaching steady state. In
effect the start boundary has the same properties as a symmetry boundary
and acts like a mirror. Therefore, starting with an excavation step and then
solving to equilibrium can result in large settlements because we effectively
have a two advances-long unlined cylinder with two open faces – this is the
lower curve on the left-hand side of Figure 7.15. In some types of soil, this
may even result in a mass failure and the calculation may not converge. If
the first advance is excavated and lined before solving to equilibrium, the
settlements are smaller – this is the upper curve in Figure 7.15.
Determining the model length is something that can only be achieved
by trial and error. Various tricks can be employed to try to reduce the time
needed to reach the central steady-state zone, particularly when departing
from the start boundary. For example, you could model a 15 m length of
tunnel excavated all at once and an internal pressure applied in a pseudo-
convergence-confinement approach. The whole length could then be lined,
and the internal pressure removed. This will get the model away from the
start boundary quickly without causing ground movements that are too
unrealistic and can replace 15 calculation steps with just 2.
In a tunnel model, we are often not just concerned with the final short-
term settlements and lining loads well behind the face, but also what hap-
pens near the face in terms of transient settlements or possibly early age
loading of shotcrete. So we need the whole response from ahead of the
face to well behind the face to be in a steady-state, such that each advance
resembles the one before. This can be verified by plotting a series of curves
similar to the one shown in Figure 7.15, at different face positions. If they
all align when plotted against distance from face position, the model has
reached a steady-state.
The total model length needs to be determined for every new situation
depending on the soil and tunnel lining constitutive behaviour, the tunnel
geometry and depth, and the tunnel construction method. For a reasonable
first estimate Möller (2006) found the following equation described the
model length l needed to achieve steady-state:
 11 H 
l = D  13 + ⋅  (7.12)
 3 D
l is the length of the model needed to achieve steady-state

H is the cover from tunnel crown to ground surface
Franzius et al. (2005) found that when coefficient of earth pressure at rest
K0 = 1.5, a steady-state was not achieved in their 3D model of an advanc-
ing tunnel in London Clay, even with sophisticated constitutive soil mod-
els and a total model length of 33D. Using Equation 7.12 with Franzius
et al.’s model geometry of H = 28.125 m and D = 4.75 m, gives l = 35D, so
it was slightly less than Möller would have used. However, when K0 = 0.5
and an unrealistically high degree of anisotropy was used, a steady-state
was achieved, though the settlements were far too large. Therefore, the
constitutive model and the initial in situ stress regime have an important
influence on whether a model can reach a steady-state.
Demagh et al. (2013) modelled an advancing TBM used for the Toulouse
Metro with face pressures, conicity and grout pressures included. The
Toulouse ‘Molasse Argileuse’ is a very stiff overconsolidated clay soil with
K0 = 1.7. In their case, settlements do stabilise behind the face using the
same model length of 33D that Franzius et al. used. Perhaps the reason for
the difference has something to do with the width of the model – Demagh
et al. used a width of 19D whereas Franzius et al. used 16D, or perhaps
it is because the Toulouse soil is stiffer and has a higher undrained shear
strength than London Clay.
When modelling an advancing tunnel in 3D, quite large longitudinal ten-
sile stresses may be generated in the lining. This is because at each advance,
ground movements towards the face will pull the lining towards the face, and
this tends to continue to increase with distance, as noted by Jones (2007) and
Thomas (2003). It is not known whether these longitudinal tensile stresses
occur in real tunnels, but it is common practice to ignore them on the basis
that in a shotcrete tunnel lining creep and cracking would tend to dissipate
them, and in a segmentally lined tunnel the TBM jacks will ensure the lin-
ing is in longitudinal compression. To date there is no known study of the
effect of these longitudinal tensile stresses, but they may have a Poisson’s
ratio effect on hoop stresses in the model. This effect can be mitigated to
some degree by detaching the elements representing one ring from another,
in effect having two nodes at each node location on the circumferential joints
and allowing them to displace independently. This, however, can cause prob-
lems with how the elements representing the soil are attached to the lining,
and how they behave across any gap that is formed, as discussed for radial
segment joints by Potts & Zdravković (2001: 49–50).
7.10.2 Modelling the tunnel lining

The tunnel lining may be modelled by solid 3D elements, or by 2D shell
or plate elements. The same arguments discussed in Section 7.4 regarding
1D beam elements or 2D solid elements in a 2D model apply to 3D solid
elements and 2D shell elements in a 3D model. The use of shell elements is
far more common because fewer elements are needed to achieve the same
accuracy and the outputs are in the form of forces and moments. With
solid elements far more sophisticated constitutive models may be used, but
outputs will be in stresses that are more difficult to interpret, and quite a
large number of small elements may be needed to achieve sufficient accu-
racy (Pound, 2006).
The same rules regarding boundary conditions on a symmetry boundary
apply to shell elements in a 3D model as for beam elements in a 2D model.
In addition, at the start boundary of a 3D model of an advancing tunnel,
the lining is usually fixed in the direction of tunnelling but free in the other
directions.
When using shell elements to model a tunnel lining, there are usually
three bending moments and five forces given in the output. The moments
are usually denoted ‘Mx’, ‘My’ and ‘Mxy’, the shear forces ‘Qx’ and ‘Qy’,
and the membrane forces ‘Nx’, ‘Ny’ and ‘Nxy’. These are illustrated in
Figure 7.16. In some programs, the numbers 1, 2 and 3 may be used for the
local axes instead of x, y and z. Remember that in most numerical model-
ling programs, Mx is not the moment about the x-axis, but the moment
along it. If this is the case then when we use the outputs to look at the
moment-axial force interaction, then Mx is coupled with Nx and My is
coupled with Ny.
Nxy is an in-plane shear force and Mxy is a twisting moment. Nxy and
Mxy need to be combined with Nx, Ny, Mx and My in order to arrive at
Figure 7.16 Local axis directions and the resultant forces in a shell element.
the design forces and moments in the orientations we are interested in (for
reinforced concrete this will be the primary and transverse reinforcement
directions). This usually has to be done in a post-processing stage outside
of the numerical modelling program. For straight tunnels modelled in 3D,
Nxy and Mxy are usually small, but not insignificant, relative to the values
of Nx and Ny, and Mx and My. Close to junctions Nxy and Mxy can be
quite large, for instance Jones (2007) found that Mxy moments close to a
shaft-tunnel junction could represent bending stresses that were equal to
40% of the hoop stress. Therefore, they always need to be considered.
Assuming the primary and transverse reinforcement directions are the
same as local axes directions x and y, the design moments Mx* and My*
according to the Wood and Armer method (Wood, 1968) are given by the
following equations.
Assuming a positive bending moment in the model is defined such that it

results in tension on the bottom surface, for the bottom layer of reinforce-
ment use:
Mx* = Mx + Mxy (7.13)
My* = My + Mxy (7.14)
If Equation 7.13 results in Mx* < 0, then set Mx* = 0 and now use the fol-
lowing equation to calculate My*:
Mxy 2
My* = My + (7.15)
Mx
If Equation 7.14 results in My* < 0, then set My* = 0 and now use the fol-
lowing equation to calculate Mx*:
Mxy 2
Mx* = Mx + (7.16)
My
If both Mx* and My* are negative, then no bottom reinforcement is

required.
For the top layer of reinforcement, use the following equations:
Mx* = Mx − Mxy (7.17)
My* = My − Mxy (7.18)
If Equation 7.17 results in Mx* > 0, then set Mx* = 0 and now use the fol-
lowing equation to calculate My*:
Mxy 2
My* = My − (7.19)
Mx
If Equation 7.18 results in My* > 0, then set My* = 0 and now use the fol-
lowing equation to calculate Mx*:
Mxy 2
Mx* = Mx − (7.20)
My
If both Mx* and My* are positive, then no top reinforcement is required.
For the design axial forces Nx* and Ny* the same procedure should be
followed to include the effect of Nxy. In Equations 7.13–7.20, replace every
instance of ‘M’ with ‘N’.
7.10.3 Modelling junctions

In the past, junctions have been modelled in all sorts of approximate ways,
due to the difficulty and expense of 3D numerical modelling. The most
common methods are:
• using the Kirsch plane stress analytical solution to estimate axial

stresses around an opening
• 3D numerical modelling of a wished-in-place junction using shell ele-
ments for the lining, applying a pressure to the outside to simulate
ground loading and using model springs to provide ground reaction
• 3D numerical modelling of a wished-in-place junction using shell ele-
ments for the lining and modelling the ground explicitly using solid
elements
• 3D numerical modelling of a junction constructed sequentially using
shell elements for the lining and modelling the ground explicitly using
solid elements
7.10.3.1 Kirsch solution

The Kirsch plane stress analytical solution is given in the following equa-
tions (Hoek & Brown, 1980; Kirsch, 1898):
1   a2   4 a2 3a4  
σr = pz (1 + k)  1 − 2  + (1 − k)  1 − 2 + 4  cos 2θ  (7.21)
2   r   r r  
1   a2   3a4  
σθ = pz (1 + k)  1 + 2  − (1 − k)  1 + 4  cos 2θ  (7.22)
2   r   r  
1   2a2 3a4  
τ rθ = pz  − (1 − k)  1 + 2 − 4  sin 2θ  (7.23)
2   r r  
σ r is the radial stress
σ θ is the circumferential stress
τ rθ is the shear stress
r is the radial distance from the centre of the hole
a is the hole radius
θ is the angle from the vertical centred on the hole
pz is the vertical applied stress in the plate
ph is the horizontal applied stress in the plate
k is the ratio of horizontal to vertical stress, k = ph /pz
The above parameters are also defined in Figure 7.17.

Usually we will assume that pz is in the direction of the hoop stress and
that k = 0. The flat sheet or ‘plate’ that has the hole in it is assumed to extend
Figure 7.17 Kirsch solution diagram with notation.
an infinite distance in the two plane directions. The plate has finite thickness
but this is only used to calculate the in-plane stresses. The plate is assumed
to have zero stress in the out of plane direction (into or out of the page in
Figure 7.17) and therefore the thickness does not feature in the equations.
There are two main types of tunnel junctions: shaft-tunnel junctions and
tunnel-tunnel junctions, where we create a hole in the side of the ‘parent’
tunnel or shaft and begin to drive a new ‘child’ tunnel away from the junc-
tion. If we unwrapped and laid flat the parent tunnel lining, it would look
rather like Figure 7.17, with pz equal to the parent tunnel hoop stress prior
to creating the opening, and usually we would assume that the longitudinal
stress in the parent tunnel lining is zero (i.e. ph = 0 in Figure 7.17).
For a shaft-tunnel junction, the hoop stress is still circumferential. So if a
shaft lining were unwrapped and laid flat, it would also resemble a flat sheet
or plate with a hole in it, only we would probably visualise it with the hoop
stress (pz) shown horizontally on the page, i.e. like Figure 7.17 but rotated 90°.
For a shaft-tunnel junction the Kirsch solution predicts the stress dis-
tribution shown in Figure 7.18. When the new tunnel opening is made,
Figure 7.18 Stress concentrations predicted by the Kirsch analytical solution. The

applied hoop stress is in the horizontal direction (Jones, 2007).
the hoop stress that is going circumferentially around the shaft lining has
to divert around it, resulting in a higher hoop stress above and below the
opening. Exactly three times higher in fact, as shown on the graph above
the opening in Figure 7.18. Due to this diversion, hoop stress in the shaft
lining decreases to zero at the axis level of the opening, and there is tension
in the vertical direction equal to −1 times the initial hoop stress (shown on
the graph to the right of the opening in Figure 7.18).
The Kirsch solution will predict axial stresses around an opening in a
flat plate, but because it does not model soil-structure interaction or the
curvature of the parent tunnel or shaft, these axial stresses will be very
inaccurate (Jones, 2007). In addition, the Kirsch solution will not calcu-
late any bending stresses, which have been found to be very significant
(Jones, 2007).
A comparison of hoop stresses in a shaft just after an opening has been
made in the side for a child tunnel is shown in Figure 7.19, along a vertical
Figure 7.19 Comparison of Kirsch solution (thick line) with 3D numerical modelling

(triangles), from Jones (2007).
line going through the centreline of the child tunnel (as shown in the inset
figure). The Kirsch solution gives a stress concentration factor increasing
from 1 far away from the opening up to a maximum value of 3 close to
the opening. The 3D numerical modelling shows that hoop (axial) stress
concentration can increase more rapidly and have a significantly higher
maximum close to the opening compared to the Kirsch solution. This is
due to curvature of the parent shaft and soil-structure interaction. The
numerical modelling in FLAC3D also predicts bending stresses, which
are significant in magnitude and not predicted by the Kirsch solution. More
plots for vertical stresses and for hoop stresses around the shaft may be
found in Jones (2007).
7.10.3.2 Wished-in-place 3D numerical model

Using a wished-in-place 3D model of the lining only, and applying a pres-
sure to its surface, is only appropriate in situations where most of the
ground or groundwater load is applied after construction. This could occur
where an impermeable secondary lining is installed while the ground is
dewatered, and then the pumps are switched off and the water pressure
is applied. Similarly, where tunnel linings resist internal pressures during
operation, such as water conveyance or sewage tunnels, then a wished-in-
place model may be suitable to model stresses at junctions. However, these
models will not model soil-structure interaction properly, even if springs are
used, because the springs are independent and will not redistribute stresses
from one part of the lining to another as the lining and the ground deform.
Hafez (1995) created a 3D finite element model of a skewed tunnel junc-
tion in the Heathrow Baggage tunnel, a shallow soft ground shotcrete-lined
tunnel in London Clay within Heathrow Airport. The shotcrete junction
was modelled wished-in-place without any surrounding ground, and a
pressure was applied to the lining. It was found that the highest stress con-
centrations were at axis level at the corners of the lining at the intersection
of the parent and child tunnels.
For design of primary linings in soft ground, a wished-in-place 3D
model will ignore the sequential nature of construction. In reality, the most
important step is when the opening is made in the parent tunnel. At this
point, hoop stress is redistributed in the lining around the opening, the
lining deforms leading to bending moments, and arching in the ground
occurs in front of the opening as it does at an open tunnel face, leading to
increased ground pressure applied to the parent tunnel around the open-
ing. Subsequent advances of the child tunnel only cause further incremental
changes to this major redistribution of stresses. Lining loads in the child
tunnel close to the junction are actually lower than they are further away,
as stresses in the ground arch onto the larger parent tunnel (Jones, 2007).
However, a wished-in-place 3D model will share the stress concentrations
between the parent and child tunnels at the junction, as Hafez (1995)
found. This has led to junction designs where both the parent and the child
tunnel linings have been unnecessarily thickened and reinforced close to
junctions, even though only the parent tunnels need it.
7.10.3.3 3D numerical model with sequential construction

This method is the most time-consuming, because the ground must be mod-
elled explicitly using solid elements, and many calculation steps are needed
to first model the construction of the parent tunnel or shaft and then the
child tunnel. Unfortunately, this is the only way to model a tunnel junction
to obtain realistic values of lining stresses for design. This is because:
• the 3D nature of junctions means the problem cannot be simplified to

2D without introducing gross unconservative errors
• 3D wished-in-place models will tend to be unconservative for design
of the parent tunnel or shaft and overconservative for design of the
child tunnel
• in most cases, the order in which the parent and child tunnel are con-
structed is critical to the stress distributions, and therefore sequential
excavation and lining of first the parent and then the child tunnel
must be modelled
• the soil must be modelled explicitly, i.e. modelling the ground as springs
or simply as an applied pressure is inaccurate and unconservative
For more detail, see Jones (2007).

Validation of a 3D numerical model using the Kirsch solution is shown in
the following worked example.
WORKED EXAMPLE 7.3 VALIDATION OF A

NUMERICAL MODEL USING THE KIRSCH SOLUTION
When a tunnel junction is constructed, there are large stress gradients
around the opening in the parent tunnel. Therefore, it is important to
check that the numerical method used to model the lining can model
these stresses.
An example is shown here from Jones (2007) of the validation of a
shaft-tunnel junction numerical model. The boundary conditions and
elastic constitutive model used in the Kirsch solution were replicated in
the numerical modelling program FLAC3D. The objective was to check
that the shell element type and mesh density around the opening could
model the high axial stress gradients. For the precise situation of a flat
elastic plate with a hole, the Kirsch solution provides an exact solution,
whereas the numerical model is approximate.
By far the biggest stress change occurs when the opening for the child
tunnel is made in the shaft lining. Hoop stresses in the shaft have to
divert around the opening, resulting in higher axial forces above and
below the opening, and tension in the vertical direction at axis level.
Taking advantage of symmetry on two axes, we only need to model ¼
of the child tunnel opening as shown in the inset figure in Figure 7.20.
The shaft lining is modelled by shell elements. The vertical stress in the
shaft lining is assumed to be zero.
An additional reason for using this kind of validation is that we can
compare the results to more complex models, for example introducing
3D geometry, soil-structure interaction or using more sophisticated con-
stitutive behaviour, and better understand what is driving the design.
A desirable mesh density around the perimeter of the shaft lining,
in terms of calculation time and detail of results, is approximately
1.5 zones/m. ‘Zones’ in FLAC3D are analogous to elements in a finite
element calculation. Here we will look at a model in FLAC3D of a
plane stress plate with a hole in it, with mesh densities of 1.5, 2, 4 and
8 zones/m. The shell elements are linear elastic. The radius of the hole is
2.4 m, the same as the external radius of the child tunnel. The boundary
Figure 7.20 Effect of mesh density on the stress concentration factor (N x /P) for
a plane stress plate with a hole, in the direction of the applied stress
along a line parallel to the direction of the applied stress (dashed line
on inset figure), compared with the Kirsch analytical solution.
is set at 20 m from the edge of the hole. The agreement between the Kirsch
analytical solution and the numerical models is very good, with the largest
differences occurring within half a radius distance from the hole. The finer
the mesh, the closer the agreement with the analytical solution.
‘Stress concentration factor’, which is the stress Nx divided by the
applied stress P at the boundary of the model, is plotted in Figure 7.20
for the axial force in the x-direction along a line parallel to the direction
of applied stress (as indicated in the location key overlaid on Figure 7.20)
up to a 5 radius distance (12 m) from the edge of the hole. Also plotted
on the same figure is the stress concentration factor calculated using the
Kirsch analytical solution. The stress concentration factor decreases to
zero as it approaches the edge of the hole. The FLAC3D models repli-
cated this behaviour with varying degrees of accuracy, dependent on the
fineness of the mesh. The stress at the edge of the hole was predicted by
the model with a mesh density of 1.5 zones/m with less than 5% error.
Figure 7.21 shows the stress concentration factor for the y-component
of the membrane force Ny along a line parallel to the direction of
applied stress. The Kirsch solution shows that the stress concentra-
tion factor should decrease from 0 to −1 as the edge of the hole is
Figure 7.21 Effect of mesh density on the stress concentration factor (N y /P ) for a

plane stress plate with a hole, in the direction transverse to the applied
stress along a line parallel to the direction of the applied stress (dashed
line on inset figure), compared with the Kirsch analytical solution.
approached. Again the FLAC3D models replicated this behaviour with

varying degrees of accuracy, generally dependent on the fineness of the
mesh. The mesh density of 1.5 zones/m was actually the closest to the
Kirsch solution adjacent to the hole.
Figure 7.22 shows the stress concentration factor for the x-component
of the axial force N x along a line transverse to the direction of applied
stress. The Kirsch solution shows that the maximum stress concen-
tration factor of 3.0 occurs on this section, adjacent to the hole, and
decreases with distance from the hole, eventually dropping towards
1.0. The FLAC3D models replicated the Kirsch solution with varying
degrees of accuracy, generally dependent on the fineness of the mesh.
The mesh density of 1.5 zones/m was actually the most accurate adja-
cent to the hole with an error of less than 0.5%.
Figure 7.22 Effect of mesh density on the stress concentration factor (N x /P) for a
plane stress plate with a hole, in the direction parallel to the applied stress
along a line transverse to the direction of the applied stress (dashed line
on inset figure), compared with the Kirsch analytical solution.
In conclusion, the numerical model with mesh density of 1.5 zones/m

predicted the maximum and minimum axial stresses to within 5% of the
Kirsch analytical solution. The generally good agreement between the
FLAC3D models and the Kirsch solution showed that the FLAC3D shell
elements could be used to predict axial stresses in shell structures with
high stress gradients with reasonable accuracy.
7.11 SUMMARY
Numerical modelling must be approached in a methodical and rational

manner. Remember to always begin with simple analytical models, then
proceed to simple numerical models, then add layers of complexity. Only
add as much complexity as is needed to characterise the problem. In this
way you will gain confidence and understanding of the problem and what is
driving the design and be in a better position to assess the outputs.
We must accept that modelling a tunnel in 2D is always a gross approxi-
mation, and whether we use the convergence-confinement method or any
other method of accounting for the 3D effect in a 2D model, we are fixing
as an input something that is really an unknown and should be an out-
put. In effect, although some soil-structure interaction will occur in the
final stage, we are deciding beforehand what distribution of stress will be
applied to the tunnel lining at the start of that final stage, which has a very
big influence on the results. Therefore, 2D analysis of a tunnel is at best
semi-empirical and at worst we are telling it the answer we want before we
start.
2D numerical models can be calibrated to give the correct maximum
settlement, the correct volume loss or the correct lining stresses, but
rarely all at the same time. Therefore, the results from 2D design models
should always be treated with caution and with an awareness of their
limitations.
3D numerical models of advancing tunnels avoid the need to make
assumptions about ground deformation prior to lining installation.
However, they can still be wrong and validation is just as important as for
2D models. As long as the geometry, boundary conditions, element types,
mesh density and initial stresses are all good, the main difference between
a 3D numerical model and reality will be the constitutive models used for
the ground and the lining.
It is important to understand the boundary conditions at the perimeter
of the tunnel excavation and at the edges of the model, and how they affect
the results. Boundary distances, element types and mesh density are also
important. Every effort should be made to ensure that these aspects are
not affecting the reliability of the results, using validation techniques and
sensitivity analyses.
Numerical models must be validated in as many ways as possible. This
can be achieved by comparison to analytical solutions, laboratory tests
and case histories. Well-documented case histories are immensely valuable
for the validation of numerical models, but care must be taken because a
numerical model calibrated to give the correct surface settlements will not
necessarily give the correct lining forces, and vice versa.
The validation does not end when the drawings are issued for construc-
tion. Monitoring during construction should be reviewed and design mod-
els updated to ensure that the assumptions used in design are still valid.
7.12 PROBLEMS
Q7.1. This question may be answered without needing access to numer-

ical modelling software.
i. Sketch the boundary conditions you would use for a 2D plane
strain continuum numerical model of a circular tunnel, taking
advantage of symmetry.
ii. For a 3D numerical model of a 7 m diameter tunnel at a depth
to axis z0 = 20 m, estimate appropriate boundary distances to
the lateral and bottom boundary using the relationships pro-
posed by Möller (2006).
iii. For the model in (ii), using the approach of Möller (2006) esti-
mate the total length of the model needed to achieve steady-
state surface settlements.
iv. If the tunnel modelled in (ii) and (iii) were to be excavated
by TBM in 1 m advances, estimate the number of calculation
stages needed to achieve a steady-state.
v. Using a series of sketches, show how you would model in 3D
the construction of parallel 7 m diameter tunnels followed by
a perpendicular 4 m diameter cross passage excavated from
one tunnel to the other.
vi. The 3D numerical model in (ii) and (iii) provides the forces in
one of the shell elements modelling the lining in the steady-
state zone listed in Table 7.2. ‘x’ is in the circumferential direc-
tion, ‘y’ is in the longitudinal direction, and ‘z’ is the direction
normal to the shell. Calculate the values of Nx∗ and Mx∗ to
use in design using the Wood and Armer method. Since the
shell elements are placed at the excavation perimeter, estimate
the errors introduced by their centroid not being in the correct
position.
Table 7.2 Q7.1(vi) shell element internal forces.

Nx Ny Nxy Qxz Qyz Mx My Mxy
kN/m kN/m kNm/m
1500 −300 120 800 200 250 70 30
Q7.2. This question may be answered without needing access to numer-

ical modelling software.
i. A 4.5 m diameter metro tunnel will be excavated by a TBM.
There are no case histories of TBM tunnelling in this soil.
Describe how you would determine the value of deconfine-
ment factor λ to use in a 2D plane strain continuum model.
ii. Describe the calculation stages 0, 1 and 2 needed to apply the

convergence-confinement method to a 2D plane strain contin-
uum numerical model.
iii. At the deepest point, the tunnel is 25 m deep to axis level, the
water table is at the surface and the unit weight of the soil is
20 kN/m3. For a coefficient of earth pressure at rest K0 = 0.5,
what should the initial in situ stresses be in Stage 0 of the
model (i.e. before any excavation takes place), at axis level?
iv. Applying a deconfinement factor λ = 0.4, what in situ stresses
will be applied to the tunnel lining at the start of Stage 2?
v. Estimate the tunnel lining hoop forces and bending moments
you should get at the end of Stage 2 using the Curtis-Muir
Wood solution with lining stiffness parameters E = 34 GPa and
ν = 0.2, lining thickness 0.25 m, and soil stiffness parameters
Ec = 50 MPa and ν = 0.25. If you have access to a numerical
modelling program, build and run a 2D model and compare
the results with the Curtis-Muir Wood solution.
Q7.3. This question requires access to 2D numerical modelling software.

i. For the tunnel in Q7.1, build a numerical model with the rec-
ommended boundary conditions and distances that you calcu-
lated based on Möller (2006).
ii. Give the soil drained elastic material properties E′ = 30 MPa
and ν ′ = 0.3, and a Mohr-Coulomb failure criterion ϕ ′ = 35°
and c′ = 0 kPa. The soil has a unit weight of 18 kN/m3. The
tunnel is well above the water table. If the program requires
definition of a water table, locate it at the bottom of the model.
Use K0 = 0.43 to establish the initial stresses. Give the lin-
ing a thickness of 0.3 m, Young’s modulus E = 30 GPa and
Poisson’s ratio ν ′ = 0.2. Using the convergence-confinement
method with λ = 0.5, run the model.
iii. Check the in situ stresses are calculated correctly in Stage 0
and that the radial stresses are reduced to the correct values
in Stage 1. Plot the final surface settlements in Stage 2 in a
spreadsheet graph. Create a graph of the lining hoop force
and bending moment around the tunnel, plotting them against
angle from the crown.
iv. Vary the lateral boundary distance by −20 m and +20 m and
compare the surface settlements and lining forces. What is the
maximum error, expressed as a percentage?
v. Vary the bottom boundary distance by −10 m and +10 m and
compare the surface settlements and lining forces. What is the
maximum error, expressed as a percentage?
vi. Increase and decrease the mesh density and compare the
results for surface settlements and lining forces with those
from before. If possible, also change the soil and lining ele-
ment types and examine any differences in the results.
Q7.4. This is a suggested project/exploration and requires access to 2D

numerical modelling software.
i. Use a 2D plane strain model to simulate the construction of one
tunnel, followed by construction of a second, identical, tunnel
parallel to it. Allow 50% relaxation of overburden pressure
before installation of the lining for both tunnels. Use soil and
lining parameters from a previous question, or use your own.
ii. Vary the size of the tunnels and the spacing between them and
analyse the effect on surface settlements and lining deforma-
tions and stresses. What is the difference in behaviour between
the first and second tunnel?
iii. For the surface settlements, to what extent does the principle
of superposition hold true (i.e. could the surface settlements
due to construction of the first tunnel be superposed for the
second tunnel, or does the second tunnel cause more/less set-
tlement than the first?). How does using different soil param-
eters or constitutive models affect this?
iv. Compare your results to numerical modelling or field meas-
urements of twin tunnels in the literature.

i. Use both 2D plane strain and 3D numerical models to replicate
Robert Mair’s stability charts (Refer to Chapter 2, Section 2.2,
or to Kimura & Mair, 1981; Mair, 1979; or Mair & Taylor, 1997)
based on centrifuge tests in clay. It is probably easier not to
use centrifuge scale and acceleration, but to use full-scale
tunnel dimensions in your model. This can be achieved by
reducing undrained shear strength to failure (some programs
have a function that allows this), or by increasing surcharge or
decreasing an internal tunnel pressure.
ii. Compare the results to upper bound plasticity solutions pre-
sented in Chapter 2.
iii. Compare the results to numerical modelling of undrained sta-
bility in the literature.
iv. Compare the results to actual reported undrained stability
failures (see e.g. in Mair & Taylor, 1997).

i. Build a 3D numerical model of an advancing tunnel to inves-
tigate the effect of unsupported length (TBM tunnel) or ring
closure distance (sprayed concrete-lined tunnel). Use soil and

lining parameters from a previous question, or use your own,
but be aware that if the strength parameters for the soil are
too low the heading may be unstable, and certainly if you
are using drained parameters with no cohesion you will need
to apply a support pressure – if in doubt, use the methods
in Chapters 2 and 3 to check the parameters will result in a
stable tunnel. Establish boundary conditions, boundary dis-
tances, element types and mesh density to demonstrate that a
steady state has been achieved and that errors will be accept-
able, using sensitivity analyses.
ii. Check and validate the model in as many ways as you can
think of.
iii. Plot steady-state surface settlements transverse to the tunnel.
Analyse how the settlement trough changes from ahead of the
face all the way back to the steady-state zone.
iv. Maintaining the advance length constant at e.g. 1 m, vary the
distance to lining installation. Create a ground reaction curve
similar to the one shown in Figure 7.2, plotting average radial
convergence or volume loss on the x-axis and average hoop
force in the lining on the y-axis. Add the results to the ground
reaction curve, showing how as distance to lining installation
increases, the lining forces decrease and the convergence or
volume loss increases.
v. Compare the effect of using different soil constitutive mod-
els on the surface settlements and the lining forces. Does the
shape of the ground reaction curve change?
REFERENCES
Addenbrooke, T. I., Potts, D. M. & Puzrin, A. M. (1997). The influence of prefailure

soil stiffness on the numerical analysis of tunnel construction. Géotechnique
47, No. 3, 693–712.
pressure-balanced shields. Tunn. Underg. Space Technol. 11, No. 2,
165–173.
Avgerinos, V., Potts, D. M., Standing, J. R. & Wan, M. S. P. (2018). Predicting
tunnelling-induced ground movements and interpreting field measurements
using numerical analysis: Crossrail case study at Hyde Park. Géotechnique
68, No. 1, 31–49.
Bishop, A. W. & Hight, D. W. (1977). The value of Poisson’s ratio in saturated
soils and rocks stressed under undrained conditions. Géotechnique 27,
No. 3, 369–384.
Brinkgreve, R. B. J., Kumarswamy, S. & Swolfs, W. M. (2018). PLAXIS 3D refer-
ence manual. The Netherlands: Plaxis bv.
Cook, R. D., Malkus, D. S., Plesha, M. E. & Witt, R. J. (2002). Concepts

and applications of finite element analysis, 4th Edition. New York: John
Wiley & Sons.
Demagh, R., Emeriault, F. & Kastner, R. (2013). Modélisation 3D du creusement
de tunnel par tunnelier à front pressurisé dans les sols surconsolidés. Revue
Française de Géotechnique, No. 142, 17–26.
Fenner, R. (1938). Untersuchungen zur Erkenntnis des Gebirgsdruckes. Glückauf
74, No. 32, 681–695.
Franzius, J. N., Potts, D. M. & Burland, J. B. (2005). The influence of soil anistot-
ropy and K0 on ground surface movements resulting from tunnel excavation.
Gilleron, N., Bourgeois, E. & Saïtta, A. (2017). Limites de la modélisation bidi-
mensionnelle des tunnels urbains pour la prévision des tassements. Revue
Française de Géotechnique, No. 150, Paper 2.
Hafez, N. M. (1995). Post-failure modelling of three-dimensional shotcrete lining
for tunnelling. PhD thesis, University of Innsbruck.
Hoek, E. & Brown, E. T. (1980). Underground excavations in rock. London:
IMM.
Itasca. (2006). FLAC3D manual, version 3.1. Minneapolis, MN: Itasca Consulting
Group Inc.
Janin, J. -P. (2017). Apports de la simulation numérique tridimensionnelle dans les
études de tunnels. Revue Française de Géotechnique, No. 150, Paper 3.
Janin, J. -P., Dias, D., Emeriault, F., Kastner, R., Le Bissonnais, H. & Guilloux,
A. (2015). Numerical back-analysis of the southern Toulon tunnel measure-
ments: A comparison of 3D and 2D approaches. Eng. Geol. 195, 42–52.
Jones, B. D. (2007). Stresses in sprayed concrete tunnel junctions. EngD thesis,
University of Southampton.
Jones, B. D. (2012). A loaded question. Tunnelling Journal, April/May issue,
26–28.
Jones, B. D. (2013). The most difficult question? Tunnelling Journal, June/July
issue, 33–35.
clay. Proc. 10th Int. Conf. Soil Mech. Found. Eng., Stockholm, Vol. 1,
pp. 319–322.
Kirsch, G. (1898). Die theorie der Elastizität und die bedürfnisse der festigkeit-
slehre. Zeitschrift des Vereines deutscher Ingenieure 42, 797–807.
Lambe, T. W. (1973). Predictions in soil engineering (Rankine Lecture).
Theme Lecture, Plenary Session 4. Proc. 14th Int. Conf. Soil Mech. Found.
Eng., Hamburg, Vol. 4.
Möller, S. (2006). Tunnel induced settlements and structural forces in lin-
ings. Doctoral Thesis, Universität Stuttgart. Mitteilung 54 des Instituts für
Geotechnik, Herausgeber P. A. Vermeer.
Möller, S. C. & Vermeer, P. A. (2008). On numerical simulation of tunnel installa-
tion. Tunn. Undergr. Space Technol. 23, 461–475. Figure 8-6 reprinted with
permission from Elsevier.
Pacher, F. (1964). Deformationsmessungen im Versuchsstollen als Mittel zur

Erforschung des Gebirgsverhaltens und zur Bemessung des Ausbaues.
Grundfragen auf dem Gebiete der Geomechanik (ed Müller, L.),
pp. 149–161. Wien: Springer-Verlag.
Panet, M. (1995). Le calcul des tunnels par la méthode convergence-confinement.
Paris: Presses de l’ENPC.
Panet, M. (2001). AFTES recommendations on the convergence-confinement
method. Paris: AFTES.
Panet, M. & Guellec, P. (1974). Contribution à l’étude du soutènement derrière le
front de taille. Proc. 3rd Cong. ISRM, Denver, Vol. 2.
Potts, D. M. & Zdravković, L. (1999). Finite element analysis in geotechnical engi-
neering, Vol. 1: Theory. London: Thomas Telford.
Potts, D. M. & Zdravković, L. (2001). Finite element analysis in geotechnical engi-
neering, Vol. 2: Application. London: Thomas Telford.
Pound, C. (2006). The performance of FLAC zones in bending. Proceedings of
4th International FLAC Symposium FLAC and Numerical Modeling in
Geomechnics 2006 (eds Varona, P., Hart, R.), Madrid, Spain, pp. 351–357.
Minneapolis: Itasca Consulting Group, Inc.
Reichl, T., Mathis, B. & Beer, G. (2003). Advanced pre-processing methods.
Numerical Simulation in Tunnelling, Chapter 4. Wien/New York: Springer.
Rowe, R. K. & Lee, K. M. (1992). Subsidence owing to tunnelling. II. Evaluation
of a prediction technique. Can. Geot. J. 29, 941–954.
Rowe, R. K., Lo, K. Y. & Kack, G. J. (1983). A method of estimating surface settle-
ment above tunnels constructed in soft ground. Can. Geot. J. 20, 11–22.
Simpson, B., O’Riordan, N. J. & Croft, D. D. (1979). A computer model for the
analysis of ground movements in London Clay. Géotechnique 29, No. 2,
149–175.
Tan, T. -S., Setiaji, R. R. & Hight, D. W. (2005). Numerical analyses using commer-
cial software – a black box? Proc. Underground Singapore 2005, Singapore,
pp. 250–258.
Thomas, A. H. (2003). Numerical modelling of sprayed concrete lined (SCL)
tunnels. PhD thesis, University of Southampton.
Wood, R. H. (1968). The reinforcement of slabs in accordance with a pre-
determined field of moments. Concrete 2, No. 2, February, 69–76.
Wright, P. J. (2013). Validation of soil parameters for deep tube tunnel assessment.
Proc. ICE Geot. Engrg 166, GE1, 18–30.
Chapter 8
Lining materials
This chapter describes the most common lining materials and their
constitutive behaviours. It covers segmental linings, cast-in-place (CIP)
linings and sprayed concrete linings, and both primary and secondary
linings.
The United Kingdom in particular has a large stock of cast iron
segmentally-lined tunnels. Many of these are over 100 years old and are
mostly still in very good condition. Due to the high cost of materials and
casting, cast iron is rarely used today for new construction, except occa-
sionally where local clients require them or where other types of lining
would be difficult to apply, for example, for the step-plate junction of the
Northern Line extension in London (FLO/TfL, 2017).
There are many brick and masonry tunnels all over the world, which can
be more than 250 years old. These materials are not used for new construc-
tion so they will not be dealt with in this book.
Concrete is strong in compression and relatively inexpensive, and so it
is nowadays by far the most common material used for tunnel lining. The
majority are made from conventional reinforced concrete or steel fibre-
reinforced concrete (SFRC). Plain concrete is still occasionally used for
smaller diameter tunnels. Due to its fluidity in its fresh state, concrete can
be applied in four main ways:
• precasting segments in moulds in a factory to allow assembly of a

segmental lining in the tunnel
• casting behind a shutter to form the tunnel lining in situ
• slipforming or extruding a lining, i.e. continuously casting behind a
moving shutter
• spraying directly onto the tunnel wall
In all four methods of application, the concrete can be conventional steel

bar-reinforced concrete, fibre-reinforced concrete (FRC; steel fibres or
fibres made from other materials) or plain concrete.
DOI: 10.1201/9780429470387-8 247
• the different materials used in tunnel linings and their advantages and
disadvantages
• durability, watertightness and fire considerations for lining design
• material behaviour of SFRC
• determine the mean and characteristic residual strength values of

SFRC from a series of beam tests
8.1 REINFORCED CONCRETE
This has been the dominant material of the last 60–70 years, though steel
fibre-reinforced segmental linings are becoming more and more com-
mon. In Europe and many other countries around the world, Eurocode 2
(EN 1992-1-1: 2004) is used for design.
Since graduates of civil engineering courses should already be able to
design a reinforced concrete section, the examples in this book will focus
more on SFRC.
There are limits to the flexural capacity of FRC, and so bar-reinforced
concrete is still used for all types of concrete tunnel linings. Sometimes
FRC may be augmented by reinforcement bars at high tensile stress loca-
tions, for example, around junctions in a sprayed concrete-lined tunnel or
near to joints in segmental linings.
8.2 STEEL FIBRE-REINFORCED CONCRETE
SFRC segmental linings have become increasingly common over the last
30 years, as have hybrid steel fibre plus conventional bar-reinforced concrete
segmental linings. Likewise, steel fibre-reinforced shotcrete has also become
more popular, as this removes the need to fix reinforcement in the tunnel.
More recently, SFRC has been used for CIP secondary linings for both tun-
nels and shafts on the Lee Tunnel and Tideway projects (Hover et al., 2017;
Psomas, 2019). This is due to the development of codes of practice and design
guidance, improvements in fibre and concrete technology, as well as increasing
confidence and acceptance of the use of SFRC through accumulated industry
experience.
The use of SFRC has clear benefits for CIP, slipformed and sprayed con-
crete tunnel linings, some of which are:
• SFRC can have improved post-crack behaviour, resulting in narrower

crack widths than in conventional reinforced concrete, and hence bet-
ter durability (fib Bulletin 83, 2017)
• there is no need to fix reinforcement bars, mesh or lattice girders
Lining materials 249
Table 8.1 Embodied CO2 of different reinforcement types per m3 of concrete

(ITAtech, 2016).
Embodied CO2
Embodied CO2 Typical dosage per m3 of concrete
Reinforcement type (kgCO2/tonne) (kg/m3) (kgCO2/m3)
Conventional steel bars 1932 60–160 116–309
Steel fibres 2425 25–40 61–97
Polypropylene synthetic 260 8–10 2–2.6
fibres
The use of SFRC also has clear benefits for segmental linings, some of
which are:
• SFRC can have improved post-crack behaviour, resulting in narrower

crack widths than in conventional reinforced concrete, and hence bet-
ter durability (fib Bulletin 83, 2017).
• Since the steel fibres are mixed into the concrete prior to pouring into
the segment moulds, there is no need to manufacture and place rein-
forcement cages. This saves on factory space, labour and materials.
• Due to the dispersion of the fibres throughout the concrete, includ-
ing close to segment faces, SFRC can better resist bursting and
spalling stresses caused by the application of tunnel boring machine
(TBM) jacks, radial joint stresses (de Waal, 2000; Schnütgen, 2003),
or accidental impacts during transportation and handling (fib
Bulletin 83, 2017).
For all types of tunnel linings and methods of application, the total mass of
steel required for SFRC is usually far less than for conventional reinforced
concrete, resulting in a saving in the overall carbon footprint (see Table 8.1).
8.2.1 Codes of practice and sources

of design guidance
SFRC is not covered by the current revision of Eurocode 2 (EN 1992-1-1:
2004) or the American concrete standard ACI 318-19 (2019). At time of
writing, the best sources of design guidance are:
1. fib Model Code 2010 (2013): ‘fib’ is the Fédération Internationale du

Béton, or the International Federation for Structural Concrete. fib
is an international organisation aiming to synthesise international
research and experience to produce practical documents for design,
so that national code commissions can take advantage of it. Model
Code 1990 served as an important basis for Eurocode 2 (EN 1992-1-1:
2004) and it is likely that Model Code 2010 will feed into the
Eurocode’s next revision. Until Eurocode 2 is updated to include
the design of concrete reinforced by fibres, Model Code 2010 is the

most authoritative and internationally recognised guidance avail-
able. Section 5.6 describes FRC material behaviour and Section 7.7
describes structural design.
2. fib Bulletin 83 (2017): This is a state-of-the-art report on the design
of precast tunnel segments in FRC, and provides relevant load cases,
guidance and example calculations. Unfortunately there is a lack of
detail in some areas.
3. ITAtech Guidance for Precast Fibre-Reinforced Concrete Segments –
Vol.1: Design Aspects (ITAtech, 2016). ITAtech is a committee of the
International Tunnelling Association (ITA-AITES) dedicated to new
technologies. This report is relatively brief but provides specific guid-
ance on the benefits and limitations of the use of SFRC in segmental
linings. Unfortunately it also contains many text formatting errors
and unclear figures.
4. ACI 544.7R-16 (2016) Report on Design and Construction of Fiber-
Reinforced Precast Concrete Tunnel Segments: This report presents the
history of FRC precast segments in tunnelling projects throughout the
world, sets out the load cases and structural design, and describes
the material parameters, tests and analyses required to complete the
design.
5. AFTES Recommendation GT38R1A1 (2013) Design, dimensioning
and execution of precast steel fibre reinforced concrete arch segments:
AFTES is the French tunnelling society.
8.2.2 Material behaviour

The presence of fibres does not significantly influence the behaviour of
uncracked concrete. The point where cracking begins is known as the ‘first
crack’, as shown in Figure 8.1. Usually SFRC exhibits deflection-softening
behaviour post-crack, as the fibres bridge across the cracks and are gradu-
ally drawn out. Using an increased dosage and/or high-performance fibres
can produce deflection-hardening behaviour (e.g. Hover et al., 2017), and
this is sometimes referred to as ‘high performance fibre-reinforced con-
crete’ or ‘HPFRC’. ‘Deflection hardening’ means that there is a peak stress
after first crack that is higher than the first crack stress, and ‘deflection
softening’ means that the first crack stress is the peak.
Steel fibres provide ductility in SFRC by failing in ‘pull-out’, rather than
breaking. Therefore, the anchorage of the ends of the fibres is crucial to
the post-crack behaviour. Steel fibres usually have what are called ‘hooked’
ends, although they are not so much hooked as stepped, as shown in
Figure 8.2. Resistance to pull-out is initially provided by the bond to the
concrete matrix. The fibre is gradually debonded from the crack to the tip.
Once the bond has been overcome, resistance is provided by a combination
of friction and the work done in continuously deforming the fibre as it is
Figure 8.1 Load-deflection curves for plain concrete, strain-softening fibre-reinforced

concrete (‘FRC’) and strain-hardening (high-performance) fibre-reinforced
concrete (‘FRC & HPFRC’) (ITAtech, 2016).
drawn along its stepped path. This is completely different from bar rein-
forcement, which is expected to fail in elongation, not anchorage.
The ductile post-crack behaviour of SFRC depends not just on the type
and dosage of fibres, but also on the concrete itself. A high-strength con-
crete will require higher strength or a higher dosage of fibres to ensure a
ductile response, because otherwise the fibres will break before the bond is
overcome. This means that at mature ages, when the concrete has a higher
strength, the behaviour may switch from ductile to brittle. This has been
called ‘embrittlement’ (Bernard, 2008), though this term is somewhat mis-
leading because it seems to imply a chemical degradation (Jones, 2014).
The solution is to ensure that the fibres are compatible with the concrete
matrix at all ages and within the full range of expected strengths, as shown
by a large number of panel tests by Bjøntegaard et al. (2014), where embrit-
tlement was avoided by using a slightly lower dosage of a higher strength
steel fibre. ITAtech (2016) recommends a fibre yield strength greater than
800 N/mm 2 when the concrete class is less than or equal to C40/50, but for
Figure 8.2 Progressive failure of a hooked-end steel fibre under pull-out load P: (a) elon-
gation and partial debonding, (b) full debonding, (c) deformation of the fibre
as it is pulled out of the bend in the concrete, (d) pull-out of straightened
fibre.
higher strength concretes a fibre yield strength greater than 1500 N/mm 2
may be required.
In a deflection-softening SFRC, it is likely that the first crack to occur
will continue to widen. SFRC tunnel linings usually rely on compressive
hoop force in the ring to limit crack widths at the serviceability limit state
(SLS). Deflection-hardening SFRC or hybrid SFRC with conventional bar
reinforcement can develop multiple small cracks, but this can only be veri-
fied by a testing programme (Hover et al., 2017; Psomas, 2019).
8.2.3 Design assisted by testing

Testing is required to determine the flexural tensile strength parameters of
SFRC for design. To design to ACI 544.FR requires four-point loading of
un-notched beams. To design to the fib Model Code 2010 (2013) requires
three-point loading of notched beams according to EN 14651: 2005.
The EN 14651: 2005 beam test set-up is shown in simplified form in
Figure 8.3. The load is displacement-controlled, meaning that the load is
increased or decreased by a computer to achieve a predefined displacement
rate. The crack mouth opening distance (‘CMOD’) is measured continu-
ously across the notch on the underside of the beam, producing a load vs
CMOD curve.
SFRC may be characterised by strength classes according to the fib
Model Code 2010 (2013). Using beam tests to EN 14651: 2005, we get a
value for the ‘Limit of Proportionality’ (LOP), which is the stress at the tip
of the notch assumed to act in an uncracked mid-span section. The ‘LOP’
is defined as the maximum stress at a CMOD between 0 and 0.05 mm,
denoted f L , which is assumed to be the point when the first crack occurs. We
also get values for ‘residual flexural tensile strengths’, which are fictitious
Figure 8.3 Notched beam under three-point loading test set-up according to EN 14651:
2005.
stresses at the tip of the notch assumed to act in an uncracked mid-span

section with linear elastic stress distribution. These residual flexural tensile
strengths are determined at CMOD values of 0.5, 1.5, 2.5 and 3.5 mm and
are denoted f R1, f R2 , f R3 and f R4, respectively.
These stresses are all calculated based on assuming uncracked linear
elastic behaviour. The bending moment M at mid-span is given by the fol-
lowing equation:
Fl
M= (8.1)
4
M is the mid-span bending moment in Nmm

F is the applied force in N
l is the span between the supports (nominally 500 mm) in mm
Using Euler-Bernoulli bending theory, we get the following relationship:
M σ
= (8.2)
I y
I is the second moment of area in mm4

σ is the stress in N/mm 2
y is the distance from the neutral axis to the extreme fibre in mm
For a linear elastic rectangular section, the second moment of area is given
by:
bhsp
3
I= (8.3)
12
b is the width of the beam in mm (nominally 150 mm)

hsp is the section depth at mid-span from the top of the beam to the
notch (nominally 125 mm)
And:
hsp
y= (8.4)
2
Thus, the stress σ is given by:
3 Fl
σ = (8.5)
2 bhsp
2
Using Equation 8.5 gives us values for LOP and residual flexural strengths
at different values of CMOD in the test, as shown in Table 8.2.
Table 8.2 Strength parameters determined by beam tests to EN 14651: 2005

for use in the fib Model Code 2010 (2013).
Strength parameter CMOD (mm)
Limit of proportionality fL 0 ≤ CMOD ≤ 0.05
Residual flexural tensile strength fR1 0.5
fR2 1.5
fR3 2.5
fR4 3.5
8.2.4 Determination of characteristic
strength values
The strength parameters from multiple beam tests are statistically anal-
ysed to determine the characteristic values of these parameters, which are
denoted by a subscript ‘k’. When designing to the fib Model Code 2010
(2013), we use f R1k for SLS and f R3k for ultimate limit states (ULS). These
are used to define a stress block that can be either linear elastic or rigid
plastic. This is similar in principle to how we design conventional rein-
forced concrete.
For specification of SFRC, strength classes can be used where the com-
pressive strength, the value of f R1k and the f R3k /f R1k ratio are defined as
shown in the following example:
FRC 40/50 – 5.0c
This shorthand notation means FRC with a characteristic compressive

cylinder strength fck of 40 N/mm 2 , a characteristic compressive cube
strength of 50 N/mm 2 , f R1k = 5.0 N/mm 2 and the letter ‘c’ means that
0.9 ≤ f R3k /f R1k ≤ 1.1. Therefore, f R3k will be taken as equal to 0.9 × f R1k =
4.5 N/mm 2 .
The strength interval for f R1k must be two consecutive numbers from the
series 1.0, 1.5, 2.0, 2.5, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0. The lower value in the
interval is used in the classification, so in the example above f R1k is between
5.0 and 6.0 N/mm 2 .
The letters defining the f R3k /f R1k ratio represent the intervals shown in
Table 8.3. The letter ‘a’ represents the minimum level of ductility required
for the SFRC to be considered ‘reinforced’, otherwise the design rules appli-
cable to plain concrete must be used. The letter ‘c’ represents the minimum
level of ductility to achieve crack width control.
More precise minimum characteristic values may be specified by the
designer if desired, the strength classes do not have to be followed strictly
(ITAtech, 2016).
Table 8.3 Residual strength ratio classification according to the

Model Code 2010 (2013).
Residual strength ratio interval Corresponding letter
0.5 ≤ fR3k/fR1k ≤ 0.7 a
0.7 ≤ fR3k/fR1k ≤ 0.9 b
0.9 ≤ fR3k/fR1k ≤ 1.1 c
1.1 ≤ fR3k/fR1k ≤ 1.3 d
1.3 ≤ fR3k/fR1k e
Typical values used in the design of precast SFRC segments are given in
ITAtech (2016) and are as follows:
• Characteristic compressive cylinder strength fck:

• at early age for demoulding, handling and storage: fck ≥ 12 N/mm2
• at age ≥ 28 days: fck ≥ 40 N/mm 2
• at age ≥ 90 days: fck ≥ 50 N/mm 2
• Residual flexural tensile strength:
• at early age for demoulding, handling and storage: f R1k ≥ 1.2 N/mm2
• at age ≥ 28 days: f R1k ≥ 2.2 N/mm 2 , f R3k ≥ 1.8 N/mm 2
These values are indicative and, depending on the project, specified

strengths may be higher.
An example of mean and characteristic flexural tensile strengths
derived from more than 30 EN 14651: 2005 beam tests is shown in
Figure 8.4, from Psomas & Eddie (2016). The testing was for a 7.8 m
ID water tunnel segmental lining in London, UK, designed for transient
design situations only, not for long-term operational conditions. We can
see that the variability of the test results must be quite high for the char-
acteristic value to be so far below the mean. The coefficient of variation
of these tests was 10% for the LOP (the peak at CMOD ≤ 0.05 mm), and
varies between 21% and 27% for the residual flexural tensile strengths.
This is in line with expectations for this kind of testing according to
ITAtech (2016), which gives typical values of 10% for the LOP and
25% for residual strengths. Once these characteristic values have been
reduced to the design value by dividing by the partial material factor for
FRC in tension γF = 1.5, the design values will be around 2.5 times lower
than the mean values.
ITAtech (2016) recommends using the approach described in EN 1990:
2002+A1: 2005 Annex D “Design assisted by testing”, clause D7.1. This is
for the determination of a characteristic value based on a limited number
of test results. The 5th percentile characteristic value f Rjk, where f Rjk may
represent the LOP or any of the residual flexural tensile strengths, is given
Figure 8.4 An example of mean and characteristic flexural tensile strengths derived
from testing (ITAtech, 2016).
by the following equation, on the assumption that the test data follow a
normal distribution:
fRjk = mx (1 − knVx ) (8.6)
f Rjk is the 5th percentile characteristic value for the LOP (in which case
replace Rj with L) or any of the residual flexural tensile strengths
(in which case replace Rj with R1, R2, R3 or R4)
mx is the mean of the flexural tensile strengths at the same CMOD
measured in the tests
kn is a statistical factor, in this case, values may be taken from Table D1
of EN 1990: 2002 (reproduced in Table 8.4)
Vx is the coefficient of variation, which is the standard deviation divided
by the mean
Either the coefficient of variation Vx is ‘known’, i.e. a conservative upper

estimate is chosen based on engineering judgement and prior experience, or
Table 8.4 Values of statistical factor kn for use in Equation 8.6 to calculate 5th percentile
characteristic values (from Table D1, EN 1990: 2002+A1: 2005).
No. of tests n 1 2 3 4 5 6 8 10 20 30 ∞
‘Vx known’ 2.31 2.01 1.89 1.83 1.80 1.77 1.74 1.72 1.68 1.67 1.64
‘Vx unknown’ - - 3.37 2.63 2.33 2.18 2.00 1.92 1.76 1.73 1.64
it is ‘unknown’ and Vx is calculated directly from the test data. According

to the ‘Designers’ Guide to Eurocode 0’ (Gulvanessian et al., 2012), and
ITAtech (2016), it is preferable to assume Vx is ‘known’ because even
though the mean and standard deviation may be unknown, in most cases,
particularly when the number of tests is small, a better estimate of the coef-
ficient of variation may be estimated based on experience than calculated.
For ‘Vx known’, the statistical factor kn in Equation 8.6 takes account of
uncertainty in estimation of the mean related to sample size, and provides
the 5th percentile value of the strength parameter based on an assumed nor-
mal distribution with coefficient of variation Vx, all with 95% confidence
level (i.e. there is a 5% probability of a further test giving a lower strength
than the characteristic value). This can be expressed by the following equa-
tion (Gulvanessian et al., 2012):
1
 1 2
kn = up  + 1 (8.7)
n 
up is the p fractile of the standard normal distribution. For the char-

acteristic value, we are taking p = 0.05, i.e. the 5th percentile,
therefore, up = 1.645
n is the sample size, i.e. the number of tests
For ‘Vx unknown’, kn also takes into account uncertainty related to the coef-
ficient of variation that has been calculated from the test results, hence the kn
values are larger, but as n approaches infinity, kn converges on the 5th percen-
tile value of a standard normal distribution (1.645), as shown in Table 8.4.
This can be expressed by the following equation (Gulvanessian et al., 2012):
1
 1 2
kn = t p s  + 1 (8.8)
n 
t p is the p fractile from the Student t distribution with (n – 1) degrees of

freedom. It is different for every value of n and can be looked up
in a statistical table.
s is the standard deviation calculated from the test data
n is the sample size, i.e. the number of tests
8.2.5 Determination of the characteristic

mean strength values
To determine the characteristic value of the mean f R,jm, which is used for
the SLS, we only need to take account of uncertainty in estimation of the
mean related to sample size. The following equation can be used:
fR, jm = mx (1 − knVx ) (8.9)

Table 8.5 Values of statistical factor kn for use in Equation 8.9 to calculate
characteristic mean values from test data (EN 1990: 2002+A1: 2005).
No. of tests n 1 2 3 4 5 6 8 10 20 30 ∞
‘Vx known’ 1.64 1.16 0.95 0.82 0.74 0.67 0.58 0.52 0.37 0.30 0
‘Vx unknown’ - 4.46 1.69 1.18 0.95 0.82 0.67 0.58 0.39 0.31 0
mx is the mean of the flexural tensile strengths at the same CMOD

measured in the tests
Vx is the coefficient of variation, which is the standard deviation divided
by the mean
If Vx is known, then the lower bound value of the mean with 95% confi-
dence level is given by Equation 8.9 using the following value for kn:
up
kn = (8.10)
n
up is the p fractile of the standard normal distribution. We are taking
p = 0.95, i.e. the 95th percentile, therefore, up = 1.645
If Vx is unknown, then the lower bound value of the mean with 95% confi-
dence level is given by Equation 8.9 using the following value for kn:
tp
kn = (8.11)
n
Values for kn are given in Table 8.5 based on Equations 8.10 and 8.11.
For a more detailed explanation of how the statistics are calculated,
see the ‘Designers’ Guide to Eurocode 0’, Appendix C (Gulvanessian
et al., 2012). For 5th percentile characteristic values, see the section on
the ‘Prediction method’ on p. 175, and for mean characteristic values, see
‘Estimation of the mean’ on p. 172 of Gulvanessian et al. (2012).
8.2.6 Durability
There are several potential durability concerns regarding SFRC:
• chloride-induced corrosion of the steel fibres

• carbonation of the concrete
• alkali-aggregate reaction in the concrete
• freeze-thaw attack
• stray current-induced corrosion
In a literature review of chloride corrosion of SFRC, fib Bulletin 83 (2017)

found that SFRC could resist chloride penetration much better than
conventional reinforced concrete. For uncracked high-quality SFRC, any

corrosion of steel fibres is limited to the outer 2–5 mm of uncracked con-
crete, and therefore has no effect on the long-term flexural capacity. In
addition, steel fibres do not cause spalling of the concrete when they cor-
rode, unlike steel bar reinforcement.
For cracked SFRC, fib Bulletin 83 (2017) found much more variation in
the results of tests by different researchers. Recommended crack width lim-
its proposed in the literature varied between 0 and 0.5 mm. Furthermore,
it is difficult to extrapolate results of accelerated laboratory tests on such a
wide variety of different concretes and steel fibres to real structures that are
usually in far less onerous exposure conditions.
Tests by Nordström (2005) found that galvanised steel fibres had a
delayed onset of corrosion and less loss of fibre diameter compared to low-
carbon steel fibres, and that stainless steel fibres were completely resistant
to corrosion.
In order to design an SFRC tunnel lining for durability, an assumption
needs to be made about the maximum allowable crack width for the SLS.
There does not seem to be a scientific consensus on what this value should
be (fib Bulletin 83, 2017) and examples of values used on real projects
are scarce. The SLS crack width used on the Lee Tunnel project for the
300 mm thick CIP secondary lining was 0.15 mm (Hover et al., 2017). For
the Tideway project SFRC shotcrete shaft primary linings, the SLS crack
width was 0.3 mm, and for the CIP SFRC secondary lining, it was 0.2 mm
(Psomas, 2019).
8.2.7 Watertightness
Virtually all tunnels and shafts are required to limit leakage from the
ground into the structure. Tunnels and shafts that are designed to contain
liquid need to be designed to limit leakage out into the ground.
EN 1992-3: 2006 classifies all concrete structures according to the
amount of leakage that is allowed:
• ‘Tightness class 0’: Some leakage acceptable, or leakage irrelevant.

Durability considerations will govern the design. The exposure class
according to Section 4 of EN 206-1: 2000 and the loading type will
determine the maximum crack width according to Section 7.3 of
EN 1992-1-1: 2004. In most tunnel linings this will be 0.2 mm.
• ‘Tightness class 1’: Leakage to be limited to a small amount, some
surface staining or damp patches acceptable. Any cracks expected
to pass through the full thickness of the lining should be limited to a
value wk1. If cracks are not expected to penetrate the full thickness,
then durability will govern the design as above.
• ‘Tightness class 2’: Leakage to be minimal, no staining. If cracks
are expected to pass through the whole thickness of the lining, then
additional measures such as waterproof membranes or waterbars will

need to be specified.
• ‘Tightness class 3’: No leakage permitted. Waterproof membranes or
other special measures will be needed.
The limiting crack width wk1 for a watertight concrete structure without
additional measures such as a waterproof membrane is determined by the
hydraulic gradient across the lining thickness, i.e. the hydraulic head dif-
ference hD divided by the lining thickness h. Values of wk1 are given by
National Annexes. The recommended values are 0.2 mm for hD /h ≤ 5 and
0.05 mm for hD /h ≥ 35. Interpolation is used for intermediate values of hD /h .
For ease of reference:
wk1 is the limiting crack width in mm

hD is the hydraulic head difference from one side of the lining to the other
in m
h is the lining thickness in m
hD /h is the hydraulic gradient across the lining thickness
For most tunnel linings, we do not expect cracks to pass through the full
thickness because some part of the section will usually be in compression.
EN 1992-3: 2006 specifies that a minimum depth of the section needs to be
in compression otherwise cracks are assumed to pass all the way through.
The recommended values are the lesser of 50 mm or 0.2h, where h is the lin-
ing thickness in mm. Therefore, for SFRC tunnel linings, it may be impor-
tant to ensure that the compression zone at the SLS meets this criterion, for
characteristic design situations at least.
In some situations, tunnel linings will have cracks passing all the way
through. One example is where a secondary lining is resisting an internal
liquid pressure, which may result in the full thickness being in tension.
Another example is the construction joints between bays of a CIP tun-
nel lining, which are usually circumferential. These will need additional
measures such as a waterbar installed if a waterproof membrane is not
provided.
Remember that serviceability of the structure is not the only con-
sideration when specifying watertightness. For example, a sewer pass-
ing through a zone where groundwater is abstracted as drinking water
may have strict requirements to prevent sewage leaking out of the
tunnel. A potable water tunnel may have strict requirements on leak-
age to avoid wasting precious drinking water, or to prevent untreated
groundwater entering the system. Also, a tunnel acting as a drain on the
ground around it may cause settlement due to consolidation, may incur
expensive pumping costs over many years or may cause environmental
damage by lowering the groundwater level and draining peat bogs or
ponds.
8.2.8 Fire resistance

When sufficient heat is applied to the surface of a tunnel lining, water in
the concrete will be turned into steam. The pressure that is built up by the
steam in the voids in the concrete near the surface breaks chunks off and
this is known as ‘explosive spalling’. This then exposes a new surface and
it happens again and again until the lining collapses, the full thickness of
the lining is gone, the fire is put out or the fire has consumed all the avail-
able fuel.
In most tunnel linings it is sufficient to include a low dosage of microfila-
ment polypropylene fibres in the concrete mix. These very fine fibres melt
at approximately 160°C, leaving space for steam to expand and in this way
preventing explosive spalling. Full-scale fire tests in a testing laboratory are
normally needed to prove the design is adequate.
The material properties of concrete are known to deteriorate with tem-
perature, and it is possible that this could cause structural failure. The
thermal properties of SFRC, i.e. the thermal conductivity and specific heat
capacity, are primarily determined by the type of coarse aggregates and
are not affected by the presence of steel fibres (Li et al., 2016) and so the
research and experience related to concrete and reinforced concrete behav-
iour in a fire can be applied.
EN 1992-1-2: 2004 gives the following equation for the design value of
concrete compressive strength during a fire:
kθ fck
fcd , fi = (8.12)
γ c , fi
fcd,fi is the design value of compressive strength during a fire in

N/mm 2
kθ is a reduction factor that is dependent on the temperature θ and the
type of aggregates. Values are given in Table 3.1 of EN 1992-1-2:
2004.
fck is the characteristic value of compressive strength in N/mm 2
γ c , fi is the material partial factor for concrete in compression during a
fire. Values for γ c , fi are given in the National Annexes. The recom-
mended value is 1.0.
The relationship between the residual flexural tensile strength f R3k and
temperature also needs to be known so that the flexural capacity of the
section can be calculated at different temperatures. This is not in the fib
Model Code 2010 (2013), but fib Bulletin 83 (2017) recommends the use
of a relationship in the Italian guideline (CNR-DT 204, 2006), which
gives values of a reduction factor k θt that are 1.0 at 20°C, 0.5 at 200°C,
0.25 at 400°C and 0 at 600°C, with linear interpolation between these
values.
8.3 CONCRETE REINFORCED WITH OTHER FIBRES
Although the use of structural synthetic fibres, colloquially referred to as

‘plastic fibres’, is common in shotcrete, particularly in deep mines where
large deformations are expected, it is not common in segmental linings,
where serviceability requirements usually require small crack widths that
are difficult to achieve with structural synthetic fibres.
The fib Model Code 2010 (2013) may also be used for concrete rein-
forced with structural synthetic fibres, as long as their Young’s modu-
lus is not “significantly affected by time and/or thermo-hygrometrical
phenomena”. This means that the fibres should not experience significant
creep, particularly if subjected to humidity or temperature. The reason for
this caveat is that the Model Code is mainly based on experience and test-
ing of SFRC.
Structural synthetic fibres tend to fail in elongation, not in pull-out. Since
they can generally elongate to high strains before breaking, this does not
in itself cause brittle behaviour, and embrittlement has not been known to
occur (Bjøntegaard et al., 2014).
Since structural synthetic fibres have a much lower density than steel,
about seven to eight times less mass is required to achieve the same vol-
ume of fibres; 6 kg/m3 of synthetic fibres is equivalent in volume to about
45 kg/m3 of steel fibres.
There are other types of fibres under development, from nanofibres of
graphene (Papanikolaou et al., 2019) to alkali-resistant glass or basalt
(Sandbakk et al., 2018). These are not yet widely used and are not covered
by existing guidelines or standards.
8.4 PLAIN CONCRETE
Historically, segmental linings have been made from plain concrete, and
there are many still in service today. These are generally small diameter
tunnels that were excavated by open-face TBMs, ensuring that jacking
loads were relatively light. The segments were small and so handling
loads and bending moments were reduced. As long as the design showed
that the section was mainly in compression under service loads, prob-
lems only arose at the joints, where bursting or shear stresses may have
caused cracking or spalling. Radial and circumferential joint capacity
was often verified by full-scale testing, though sadly much of this was
not published.
In conventional reinforced concrete design, it is usual to ignore the
tensile strength of the concrete itself. Concrete does have a tensile
strength, but without reinforcement by bars or fibres, failure will be
brittle and sudden (c.f. Figure 8.1), so a more generous safety margin
is required.
Design values of tensile and compressive strength of plain concrete may

be determined using Section 12 of EN 1992-1-1: 2004. The plain concrete
design strength in tension fctd is given by:
α ct , pl fctk
fctd = (8.13)
γc
fctd is the design value of concrete tensile strength in N/mm 2 .
α ct , pl is a coefficient to take account of “long-term effects on the ten-
sile strength and of unfavourable effects resulting from the way
the load is applied” (EN 1992-1-1: 2004). For plain concrete,
EN 1992-1-1: 2004 recommends α ct , pl = 0.8 and the UK National
Annex requires α ct , pl = 0.6 (NA to BS EN 1992-1-1: 2004).
fctk is the characteristic value of concrete tensile strength in N/mm 2 .
For plain concrete, the characteristic value fctk can be obtained by
splitting tests to EN 12390-6: 2009. For preliminary design, the
value from Table 3.1 of EN 1992-1-1: 2004 can be used.
γc is the partial material factor for concrete and the recommended value
is 1.5.
The plain concrete design strength in compression fcd is given by:
α cc , pl fck
fcd = (8.14)
γc
fcd is the design value of concrete compressive (cylinder) strength in

N/mm 2 .
α cc , pl is a coefficient to take account of “long-term effects on the com-
pressive strength and of unfavourable effects resulting from the
way the load is applied” (EN 1992-1-1: 2004). For plain concrete,
EN 1992-1-1: 2004 recommends α cc , pl = 0.8 and the UK National
Annex requires α cc , pl = 0.6 (NA to BS EN 1992-1-1: 2004).
fck is the characteristic value of concrete compressive strength in
N/mm 2 .
8.5 CAST IRON
Cast iron has good strength in compression and good corrosion resistance
and is therefore a good choice of material for tunnel linings. However, it
is more expensive than the various types of concrete linings, can be time-
consuming to install, and temporary works are required to support the
exposed ground at the face (unless a TBM is used). Cast iron tunnel linings
consist of segments that are always bolted together at both the radial and
circumferential joints. Radial joints are usually staggered. Any void outside
the rings is usually grouted with cementitious grout through grout ports
built into the segments. Special tapered segments may be used to go round
curves, sometimes with very tight radii.
Older cast iron is known as ‘grey cast iron’. Properties can vary for his-
toric grey cast iron before 1928, but for grey cast iron made after 1928,
several Grades are defined in BS EN 1561: 2011, from which compres-
sive strength, tensile strength, Young’s modulus and Poisson’s ratio may be
derived (LUL, 2007). Strengths are usually expressed as permissible design
stresses, with no need for partial factors.
Figure 8.5 shows an example of relatively modern cast iron rings built
in 2007. Note the staggered radial joints and the solid key segments in the
crown. These were a modern form of cast iron called ‘spheroidal graphite
iron’ (SGI), for which there is a European Standard BS EN 1563: 2018.
Since cast iron segments are not solid rectangular sections, the second
moment of area must be calculated using the parallel axis theorem. Also,
the position of the neutral axis must be found by taking moments of area.
A good overview of the properties of cast iron tunnel linings, com-
mon defects and common durability issues may be found in the ‘London
Underground Manual of Good Practice for Deep Tube Tunnels and Shafts’
(LUL, 2007).
Figure 8.5 Cast iron rings in the Piccadilly Line Access passageway at King’s Cross under
construction.
Laboratory testing of both grey cast iron and SGI was reported in detail
by Thomas (1977). More recent testing and finite element modelling were
described in Tsiampousi et al. (2017). 3D numerical modelling of cast iron
tunnel linings may be found in Li et al. (2014, 2015).
8.6 SUMMARY
The most common materials used as tunnel linings were described, with
particular focus on SFRC.
The procedure for deriving characteristic residual flexural strength val-
ues for SFRC was described in detail, as well as durability, watertightness
and fire resistance.
REFERENCES
ACI 318-19 (2019). Building code requirements for structural concrete. Farmington
Hills: American Concrete Institute.
ACI 544.7R-16 (2016). Report on design and construction of fiber-reinforced
precast concrete tunnel segments, Reported by ACI Committee 544.
Farmington Hills: American Concrete Institute.
AFTES Recommendation GT38R1A1 (2013). Design, dimensioning and execution
of precast steel fibre reinforced concrete arch segments. Tunnels et Espace
Souterrain, No. 238, 312–324.
Bernard, E. S. (2008). Embrittlement of fibre-reinforced shotcrete. Shotcrete,
Summer issue, 16–20.
Bjøntegaard, Ø., Myren, S., Klemetsrud, K. & Kompen, R. (2014). Fibre reinforced
sprayed concrete (FRSC): Energy absorption capacity from 2 days age to one
year. Proceedings of 7th International Symposium on Sprayed Concrete –
Modern Use of Wet Mix Sprayed Concrete for Underground Support (eds
Beck T., Woldmo, O. & Engen, S.), Sandefjørd, Norway, 16th–19th June,
pp. 88–97. Oslo, Norway: Norwegian Concrete Society/Norsk Betongforening.
BS EN 1561: 2011. Founding. Grey cast irons. London, UK: British Standards
Institution.
BS EN 1563: 2018. Founding. Spheroidal graphite cast irons. London, UK: British
Standards Institution.
CNR-DT 204 (2006). Guidelines for design, construction and production control
of fiber reinforced concrete structures. Rome: National Research Council of
Italy.
de Waal, R. G. A. (2000). Steel fibre reinforced tunnel segments for the application
in shield driven tunnel linings. PhD Thesis. Technische Universiteit Delft.
EN 14651: 2005. Test method for metallic fibered concrete – Measuring the flex-
ural tensile strength (limit of proportionality (LOP), residual). Brussels:
European Committee for Standardization.
EN 12390-6: 2009. Testing hardened concrete – Tensile splitting strength of test
specimens. Brussels: European Committee for Standardization.
EN 1990: 2002+A1: 2005. Eurocode – Basis of structural design, incorporating

corrigenda December 2008 and April 2010. Brussels: European Committee
EN 1992-1-1: 2004. Eurocode 2: Design of concrete structures – Part 1-1:
General rules and rules for buildings. Brussels: European Committee for
Standardization.
EN 1992-1-2: 2004. Eurocode 2: Design of concrete structures – Part 1-2:
General rules – Structural fire design. Brussels: European Committee for
Standardization.
EN 1992-3: 2006. Design of concrete structures – Part 3: Liquid retaining and
containment structures. Brussels: European Committee for Standardization.
EN 206-1: 2000. Concrete – Part 1: Specification, performance, production and
conformity, incorporating Corrigendum No. 1 and Amendment No. 1.
Brussels: European Committee for Standardization.
fib Bulletin 83 (2017). Precast tunnel segments in fibre-reinforced concrete, State-
of-the-art Report, fib WP 1.4.1, October. Lausanne: Fédération internatio-
nale du béton (fib).
fib Model Code 2010 (2013). fib Model Code for concrete structures 2010.
Lausanne: Fédération internationale du béton (fib).
FLO/TfL (2017). Northern Line Extension Kennington Step-Plate Junction
Update. Available at: http://content.tfl.gov.uk/step-plate-junction-surface-
works-possession.pdf [last accessed 17th September 2019].
Gulvanessian, H., Calgaro, J. -A. & Holický, M. (2012). Designers’ guide to
Eurocode: Basis of structural design, 2nd Edition. London: ICE Publishing.
Hover, E., Psomas, S. & Eddie, C. (2017). Estimating crack widths in steel fibre-
reinforced concrete. Proc. Inst. Civ. Engrs Construction Materials 170,
No. 3, 141–152.
ITAtech (2016). Design guidance for precast fibre reinforced concrete segments –
Vol. 1: Design aspects, ITAtech Report No. 7. Lausanne: ITA-AITES.
Reproduction of figures with permission of the Committee on new technolo-
gies of the International Tunnelling and Underground Space Association –
ITAtech.
Jones, B. D. (2014). Roundup of the 7th International Symposium on Sprayed
Concrete. Tunnelling Journal, August/September issue, 25–29.
Li, S., Jones, B., Thorpe, R. & Davis, M. (2016). An investigation into the ther-
mal conductivity of hydrating sprayed concrete. Constr. Build. Mater. 124,
363–372.
Li, Z., Soga, K., Wang, F., Wright, P. & Tsuno, K. (2014). Behaviour of cast-iron
tunnel segmental joint from the 3D FE analyses and development of a new
bolt-spring model. Tunn. Undergr. Space Technol. 41, 176–192.
Li, Z., Soga, K. & Wright, P. (2015). Behaviour of cast-iron bolted tunnels and
their modelling. Tunn. Undergr. Space Technol. 50, 250–269.
LUL (2007). Manual of good practice – Civil Engineering – Deep tube tunnels and
shafts, G-055 A1. London, UK: London Underground Limited.
NA to BS EN 1992-1-1: 2004. UK National Annex to Eurocode 2: Design of con-
crete structures – Part 1-1: General rules and rules for buildings. London:
British Standards Institution.
Nordström, E. (2005). Durability of sprayed concrete – Steel fibre corrosion in
cracks. Doctoral thesis, Luleå University of Technology, Sweden.
Papanikolaou, I., Al-Tabbaa, A. & Goisis, M. (2019). An industry survey on the use
of graphene-reinforced concrete for self-sensing applications. Proceedings of
International Conference on Smart Infrastructure and Construction (eds De
Jong, M. J., Schooling, J. M. & Viggiani, G. M. B.), pp. 613–622. London:
ICE Publishing.
Psomas, S. (2019). Service limit state design for pressurised steel fibre reinforced con-
crete tunnel linings. Proceedings of WTC2019 – Tunnels and Underground
Cities: Engineering and Innovation meet Archaeology, Architecture and Art
(eds Peila, D., Viggiani, G. & Celestino, T.), Naples, Italy, pp. 2898–2908.
London, UK: Taylor & Francis Group.
Psomas, S. & Eddie, C. M. (2016). SFRC segmental lining design for a pres-
surised tunnel. Proceedings of WTC2016, San Francisco, USA, Paper 0118.
Englewood, CO: SME.
Sandbakk, S., Miller, L. W. & Standal, P. C. (2018). MiniBarsTM – A new dura-
ble composite mineral macro fiber for shotcrete, meeting the energy absorp-
tion criteria for the industry. Proceedings of 8th International Symposium
on Sprayed Concrete – Modern Use of Wet Mix Sprayed Concrete for
Underground Support, Trondheim, Norway, 11th–14th June (eds Beck, T.,
Myren, S. & Engen, S.), pp. 272–282. Oslo, Norway: Norwegian Concrete
Society/Norsk Betongforening.
Schnütgen, B. (2003). Design of precast steel fibre reinforced tunnel elements.
International RILEM Workshop on Test and Design Methods for Steel
Fibre Reinforced Concrete (eds Schnütgen, B. & Vandewalle, L.), Bochum,
Germany, pp. 145–152. Paris, France: RILEM Publications SARL.
Thomas, H. S. H. (1977). Measuring the structural performance of cast iron tunnel
linings in the laboratory. Ground Engineering, July, 29–36.
Tsiampousi, A., Yu, J., Standing, J., Vollum, R. & Potts, D. (2017). Behaviour of
bolted cast iron joints. Tunn. Undergr. Space Technol. 68, 113–129.
Chapter 9
Segmental lining design
In Chapters 6 and 7 we learnt how to undertake simple 2D plane strain

analysis of a tunnel lining and the basic principles of numerical modelling
in 2D and 3D. We can use these methods to design segmental linings for
ground and water loads if we adjust the lining stiffness to take account of
the joints. The presence of joints also means there will be contact eccen-
tricities, which themselves cause bending moments and shear forces, as well
as bursting and spalling stresses.
This chapter will first deal with joints in a global analysis of a segmental
lining. Then it will go through the design of a segmental lining for service
loads from the ground and groundwater. It is assumed that most readers
of this book will already be familiar with reinforced concrete design, and
so the focus in the worked examples will be on fibre-reinforced concrete
design.
• how flat and curved joints behave when rotated and the effect of
packers
• contact stress concentrations caused by geometrical inaccuracies
• how global analysis of a tunnel lining can incorporate joint rotation
behaviour
• include joint stiffness in an analytical or numerical model of a seg-

mental lining
• design for bursting stresses near to joints
• estimate the effect of joint rotation on axial force eccentricity and
hence on bending moments
• design segmental linings for ground and groundwater loads in ser-
vice, with a particular focus on steel fibre-reinforced concrete (SFRC)
This chapter is not intended to be a comprehensive manual for design of

segmental tunnel linings, but a teaching aid to demonstrate the principles
DOI: 10.1201/9780429470387-9 269
and most important aspects of design. Every project is unique and has dif-
ferent load cases and performance requirements, some of which may not
be covered here. But the understanding gained from working through this
chapter, and careful reading of the papers, guidelines and standards cited,
will give you a solid foundation.
In this chapter we will look at how to design a segmental lining taking
account of the effect of the joints. Having joints in a ring will reduce its
overall flexural stiffness compared to a monolithic (solid, jointless) ring.
This is beneficial, as it will reduce the bending moments in the ring. You
can demonstrate this yourself using the Curtis-Muir Wood method (see
Chapter 6) in a spreadsheet – reduce the moment of inertia of the lining and
see what happens. There are different methods available to take account of
this effect, depending on the type of analysis used. These will be presented
and critiqued in the following sections.
As usual in design there is a trade-off; increasing the number of segments
may slow down the rate of tunnelling, as it takes longer to build a ring with
more segments. It also increases the number of joints, and thus the poten-
tial for defects to allow water ingress, which could be an issue for operation
and maintenance. On the other hand, reducing the number of segments
will increase the size and weight of each segment, increasing the handling
loads the segment needs to resist during demoulding, storage, transporta-
tion, handling and erection. This will be covered in Chapter 10, where we
will cover transient loads during the whole life of the segments before they
are in their final position.
9.1 TAKING ACCOUNT OF THE EFFECT

OF JOINTS IN 2D PLANE STRAIN ANALYSES
In preliminary analysis of ground and groundwater loads on a segmental

tunnel lining, Muir Wood (1975) proposed that the effective moment of
inertia Ie of the tunnel lining should be reduced to take account of the pres-
ence of radial joints using the following equation:
2
 4
Ie = I j + I   with Ie  I , n > 4 (9.1)
 n
Ie is the effective moment of inertia of the tunnel lining

Ij is the moment of inertia of the joints
I is the moment of inertia of a segment
n is the number of equal segments in a ring. Where a key segment is
used, it is usually counted as one joint, not two, since the two top
segments plus key usually have the same arc as two ordinary seg-
ments. So a ring with six ordinary segments, two top segments
and a key would have n = 8.
Segmental lining design 271
Note that this equation was incorrectly printed in Muir Wood’s paper,
though it is clear from his example in the text that Equation 9.1 above is
what was intended.
The equation is only applicable if there are more than four joints. In
reality a ring can deform with four joints if the joints are aligned with
the principal stress directions in the ground (i.e. joints at 3, 6, 9 and
12 o’clock positions). But if they are rotated 45°, then the joints have no
effect. When a ring has four joints, their orientation makes a huge differ-
ence, but this effect decreases as the number of joints increases. Hefny &
Chua (2006) compared Muir Wood’s equation to analyses using contin-
uum finite element models and found that as the number of joints increases
beyond six, the influence of joint position relative to the principal stress
directions becomes negligible.
Judgement is required to determine Ij. We know that Ij cannot be less
than zero, and that Ie must be less than I of a segment section. If Ij = 0,
this is the same as assuming that the joints are completely free to rotate. If
Ie = I, this is the same as assuming the lining is monolithic. Free rotation is
virtually impossible in practice, as rotation will always result in an eccen-
tricity of thrust across a radial joint, resulting in a bending moment that
resists further rotation. It also assumes that an individual ring’s deforma-
tion is unimpeded by the adjacent rings, which would only be the case if
the circumferential joint had no shear keys, dowels or friction, or if the
radial joints were all aligned down the length of the tunnel (Klappers
et al., 2006). It is usual practice to stagger joints and to encourage shear
resistance, so this assumption is unrealistic. The difference between
aligned and staggered radial joints, and monolithic rings, is illustrated in
Figure 9.1.
Our objective is to estimate a realistic but conservative value of Ie so
that we can design a more efficient tunnel lining. The two limiting situa-
tions are either that the radial joints are pin joints, providing no rotational
Figure 9.1 From left to right: monolithic rings, aligned radial joints, staggered radial
joints.
stiffness, or that the lining is monolithic (i.e. as though there were no joints
at all). de Waal (2000) showed, and perhaps it is intuitively obvious, that
the truth always lies somewhere between these two limiting situations. Fei
et al. (2014) also proved this using scale models. Kavvadas et al. (2017)
found that when the radial joints are staggered and have a high rotational
stiffness, and the shear stiffness at the circumferential joint is also high, the
behaviour may be similar to a continuous monolithic lining with no joints
at all. Their work suggests that for a conservative first approximation it is
probably best to ignore the effect of joints altogether.
Moreover, Equation 9.1 is effectively smearing the effect of the joints
around the whole ring. This allows us to calculate ‘average’ bending
moments, but in reality, bending moments near to the joints will have
a different value (usually higher) than bending moments mid-segment
and will depend on the joint geometry, bolt details, packer and gasket.
Also, in staggered rings with friction and/or shear keys across the cir-
cumferential joints, forces are diverted around joints in a complex 3D
manner.
One way to conservatively estimate the local bending moment
(described in greater detail in Section 9.2) is to assume a joint rotation
based on the maximum specified ovalisation and then to use the joint
geometry and packer properties to calculate the eccentricity of the hoop
force and hence a bending moment. The designer can then use the greater
of the ‘average’ and this local value for design, and this approach should
be conservative. From experience, however, I know that this approach is
not commonplace and the usually less conservative ‘average’ values are
used for design.
9.2 ROTATIONAL RIGIDITY – FLAT JOINTS
In reality, bending moments at the joints depend on the geometry of the

joint and the properties of the packer, if present. As a joint rotates, the
line of thrust across it moves and this eccentricity of the thrust transfers
moment to the adjacent segment (Figure 9.2). The moment is equal to
the hoop force multiplied by the eccentricity. In addition, as eccentric-
ity increases and the hoop force gets closer to the edge, the shearing
resistance of the corner reduces as the potential shear plane reduces
in length. This can result in damage to the edges of the joints and if it
occurs on the intrados, requires repair. If a shear failure occurs on the
extrados, it cannot be seen or repaired, and it may also compromise the
watertightness of the gasket. Therefore, designing to minimise eccen-
tricity is important.
The properties of the packer have a big influence on the eccentricity of
the hoop force. A packer that is too thin or too compressible will result
Figure 9.2 Eccentricity of hoop thrust caused by joint rotation.
in a concrete-to-concrete point contact and a high stress concentration.

A packer that is too stiff will also generate a large eccentricity as stress is
mobilised at the edge of the packer first. Understanding the stress-strain
behaviour of a packer and how it interacts with joint rotation is absolutely
essential.
Sometimes radial joints are designed with curved surfaces, and in the-
ory, curved joints have a lower rotational stiffness than flat joints. This
is because when flat joint surfaces rotate relative to each other, they will
quickly generate a larger eccentricity by tending to hinge about the edge
of the contact area. As curved joints rotate, the line of thrust stays closer
to the centreline of the segment. This will be discussed later in this chap-
ter. Notice also that as flat joints in a ring rotate, the perimeter of the
ring increases. As far as I know, no one has researched this effect, but it
should be expected to slightly increase the hoop force. If curved joints have
a radius approximately equal to one half the arc length of a segment, then
the perimeter does not change.
Eccentricity in radial joints is also caused by ‘lipping’, where segments
are misaligned and have a step between them. In this case, the eccentricity
is equal to half the misalignment, as shown in Figure 9.3.
What we want to estimate is a value for rotational stiffness. This is the
bending moment per unit length (in the tunnel longitudinal direction)
required to cause a unit rotation angle along a radial joint. According
to full-scale tests by Teachavorasinskun & Chub-uppakarn (2010), typi-
cal values are 1000–3000 kNm/rad. In the following sections we will
Figure 9.3 Effect of misalignment on eccentricity.
look at the following different packer types and their effect on rotational
stiffness:
• linear elastic packers – these do not exist in the real world, but the
assumption of linear elastic behaviour allows for a simple calculation
of the contact stress distribution, which helps us to understand the
behaviour of a rotating joint and the calculation steps
• nonlinear bituminous and plastic packers, which exhibit an increase
in stiffness with strain
• nonlinear plywood and medium density fibreboard (MDF) packers,
which exhibit a decrease in stiffness with strain
This will be followed by a comparison of the different types and a summary.
9.2.1 Linear elastic packers

Figure 9.4 shows a contact stress distribution that one might expect from a
linear elastic packer, and this is sometimes assumed in calculations because
it is easy to solve, even though linear elastic packers do not exist in the real
world. Nowadays, packers are most often made from bituminous or plastic
(polyethylene or polyurethane based) materials. MDF or plywood is becom-
ing less popular due to durability concerns (Cavalaro & Aguado, 2012).
None of these materials behave in a linear elastic manner so to get a rea-
sonable estimate of rotational stiffness we will then need to do something
a bit more sophisticated, but let’s start with a simple linear elastic packer
Figure 9.4 Calculation of contact stress distribution – linear elastic packer with triangu-
lar stress distribution.
model to explain the steps. Figure 9.4 shows the geometry and symbols we
are going to use.
For simplicity it is assumed that the packer is symmetrically positioned
on the centreline of the joint. The packer is linear elastic, with Young’s
modulus Ep, so it follows from Hooke’s Law that the maximum compres-
sion displacement of the packer δmax is given by:
σ max t p
δ max = (9.2)
Ep
δmax is the maximum compression displacement of the packer in mm

σmax is the maximum stress as shown in Figure 9.4 in N/mm 2
t p is the thickness of the packer in mm
Ep is the Young’s modulus of the packer in N/mm 2
Also, the area of the triangular contact stress distribution in Figure 9.4
must equal the hoop force N across the joint for there to be equilibrium:
σ max Lc Lp
NLs = (9.3)
2
N is the hoop force per metre length of tunnel in N/m
L s is the segment length in the tunnel longitudinal direction in mm
Lc is the contact length as shown in Figure 9.4 in mm
Lp is the length of the packer in the tunnel longitudinal direction (as it
won’t be covering the full length of the segment) in mm
Rearranging Equation 9.2 for σmax and substituting it into Equation 9.3
gives:
δ max Ep Lc Lp
N= (9.4)
2t p Ls
Now we have two unknowns, δmax and Lc, but they are related geometri-
cally by the following Equation 9.5:
δ max
Lc = (9.5)
ψ
ψ is the angle of joint rotation, as defined in Figure 9.4, in radians (rad)
Substituting Equation 9.5 into Equation 9.4 gives:
δ max
2 E L
p p
N= (9.6)
2t pψ Ls
Rearranging for δmax gives us:
2t pψ Ls N
δ max = (9.7)
Ep Lp
Using Equation 9.7, we can find the value of δmax that gives us equilibrium
for a given value of hoop force N, packer properties Ep, Lp and t p, and joint
rotation ψ. We can then use this value of δmax in Equation 9.5 to find Lc.
Once Lc is known, the position of the resultant hoop force can be found,
in this case, a triangular contact stress distribution, it will be Lc /3 from the
position of σmax.
If the calculated contact length Lc is greater than the packer width a, then
it means that the contact pressure distribution is trapezoidal rather than
triangular, as shown in Figure 9.5.
Equation 9.2 still stands but we now reason as follows:
The hoop force per segment is now in equilibrium with a trapezoidal pres-
sure distribution varying from σmax to σmin over the full width of the packer a:
σ + σ min 
NLs =  max  Lp a (9.8)
 2
Now δmin has a similar relationship to σmin as δmax has to σmax in Equation 9.2,
so we end up with:
Ep Lp a
NLs =
2t p
(2δ max − aψ ) (9.9)
Figure 9.5 Calculation of contact stress distribution – linear elastic packer with trap-
ezoidal stress distribution.
Solving for δmax gives us:
NLs t p aψ
δ max = + (9.10)
Ep Lp a 2
Because we know δmax and the rotation angle ψ, δmin can be calculated:
δ min = δ max − aψ (9.11)

By taking moments about the centreline, we can calculate the eccentricity:
NLs e =
(σ max − σ min ) aLp ×  a − a  (9.12)
 
2 2 3
The moment due to eccentricity of hoop force is given by:
M = Ne (9.13)
M is the moment in Nmm
e is the eccentricity of the hoop force N from the centreline of the seg-
ments in mm
The rotational stiffness is defined as the bending moment per unit length
(in the tunnel longitudinal direction) required to cause a unit rotation angle
along a radial joint, and is given by:
M
k= (9.14)
ψ
k is the rotational stiffness in kNm/rad
Thus we have calculated the eccentricity of hoop force due to joint rotation,
and from that we have calculated the moment induced by that eccentricity
and the rotational stiffness of the joint.
WORKED EXAMPLE 9.1

Use a spreadsheet to calculate the eccentricity of hoop force for joint
rotations from ψ = 0° to 1° and for hoop forces N = 500, 1000, 1500,
2000 and 2500 kN/m.
The linear elastic packer has the following properties: width a =
150 mm, thickness t p = 3 mm, packer length in tunnel longitudinal direc-
tion Lp = 900 mm and Young’s modulus Ep = 40 MPa.
The segment has the following properties: internal diameter ID = 6000 mm,
segment thickness ts = 250 mm, length in tunnel longitudinal direction
L s = 1000 mm.
From the values of eccentricity, calculate the bending moments. Plot a
graph of bending moment (on the y-axis) vs joint rotation (on the x-axis).
The relationship between bending moment and joint rotation is not
linear, even though the packer is linear elastic. Investigate this by plotting
a graph of contact stress (on the y-axis) vs distance across the packer (on
the x-axis) for hoop force N = 1000 kN/m and explain what is going on.
An example of the spreadsheet calculation for ψ = 0.5° (= 0.008727 rad)
and N = 1500 kN/m (= 1500 N/mm) is presented below.
Assuming a triangular contact stress distribution, we can use
Equation 9.7:
2t pψ Ls N 2 × 3 × 0.008727 × 1000 × 1500

δ max = = = 1.477 mm
Ep Lp 40 × 900
Then using Equation 9.5:
δ max 1.477
Lc = = = 169.257 mm
ψ 0.008727
Now Lc > a, so the contact stress distribution must be trapezoidal.

Start again using Equations 9.10 and 9.11 instead:
NLs t p ψ a 1500 × 1000 × 3 0.008727 × 150

δ max = + = + = 1.488 mm
Ep Lp a 2 40 × 900 × 150 2
δ min = δ max − aψ = 1.488 − 150 × 0.008727 = 0.179 mm

Using Equation 9.2 rearranged for σmax and σmin:
δ max Ep 1.488 × 40
σ max = = = 19.838 N/mm 2
tp 3
δ min Ep 0.179 × 40
σ min = = = 2.384 N/mm 2
tp 3
To check the answer is correct, we can insert σmax and σmin into the equi-
librium Equation 9.8:
σ + σ min  Lp a  19.838 + 2.384  900 × 150

N =  max  =  × = 1500 kN/m
 2 Ls  2 1000
To find the eccentricity e of the resultant hoop force N from the joint
centreline, we resolve moments about the centreline as follows:
NLs e =
(σ max − σ min ) aLp ×  a − a 
 
2 2 3
Therefore:
e=
(19.838 − 2.384 ) × 150 × 900 ×  150 − 150  = 19.635 mm
 
2 × 1500 × 1000 2 3 
So in this case the eccentricity e = 19.64 mm, and using Equation 9.13 the
bending moment M resisting rotation is M = Ne = 29.45 kNm/m.
We can then calculate that the (secant) rotational stiffness, using
Equation 9.14, is M/ α = 3375 (kNm/m)/rad.
Figure 9.6 shows the relationship between joint rotation and bend-
ing moment for a linear elastic packer as calculated in a spreadsheet. At
higher values of joint rotation, the line changes from a straight line to
a curve, and this corresponds exactly to the point at which the contact
stress switches from a trapezoidal distribution to a triangular distribu-
tion. As one might expect, this transition from trapezoidal to triangular
stress distribution occurs later at higher values of hoop force.
The rotational stiffness, which is the gradient of the curves in
Figure 9.6, therefore is only a constant while there is a trapezoidal stress
distribution, but beyond the transition to a triangular distribution, it
decreases (or becomes less stiff) as rotation increases.
In order to illustrate the transition, the contact stress distributions for the
hoop force of 1000 kN/m, as rotation is increased, are shown in Figure 9.7.
Figure 9.6 Bending moment – joint rotation relationship for a linear elastic packer
with Young’s modulus 40 MPa at different levels of hoop force.
Figure 9.7 Contact stress distribution across a 150 mm wide linear elastic packer
as joint rotates under 1000 kN/m hoop force (note transition from trap-
ezoidal to triangular stress distribution at approximately 0.43°).
At approximately 0.43 degrees, shown by the thick black dashed line,

the stress distribution switches from a trapezoidal shape to a triangular
shape. Perhaps counterintuitively, the rate of change of the eccentricity
(i.e. how far the resultant hoop force is from the centreline) decreases
after the transition to a triangular distribution, and this results in the
shape of the curves in Figure 9.6.
9.2.2 Nonlinear packers

Linear elastic packers do not exist in the real world. All real packers are
nonlinear.
Cavalaro & Aguado (2012) performed tests on several types of packers
used in projects in Spain. They presented stress-strain data for a 1.92 mm
thick bituminous packer and a 2.15 mm thick plastic packer, both used for
Barcelona Metro Line 9 tunnels. I have fitted a stress-strain relationship to
their data, which has the form:
σ = A exp ( ε n ) − 1 (9.15)
σ is the stress in the packer in N/mm 2

ε is the strain in the packer
A and n are curve-fitting constants
(
Cavalaro & Aguado (2012) did propose an equation σ = A 1 − e − ε n  for )
the stress-strain relationship but it did not fit their data.
An important effect, particularly for bituminous packers, is that the first
load cycle has a much softer response and on subsequent load cycles the
material has a much stiffer behaviour. This is known as the ‘Mullins effect’
and is due to densification of the material.
The stress-strain curves based on Equation 9.15 and using the curve-
fitting parameters in Table 9.1 are shown in Figure 9.8.
Table 9.1 Parameters from curve-fitting of tests on plastic and bituminous packers
(based on test data from Cavalaro & Aguado, 2012).
Code Thickness tp A n
Plastic 1st loading cycle P1 2.15 mm 0.028 0.398
Plastic 2nd loading cycle P2 0.047 0.394
Plastic 3rd loading cycle P3 0.048 0.394
Bituminous 1st loading cycle B1 1.92 mm 0.013 0.366
Bituminous 2nd loading cycle B2 0.079 0.425
Bituminous 3rd loading cycle B3 0.063 0.434
Figure 9.8 Stress-strain curves of plastic and bituminous packers plotted using

Equation 9.15.
As an example, we will use the parameters for only the first loading cycle
of the plastic packer ‘P1’ as given in Table 9.1.
This calculation is now a bit more complicated than the one in
Section 9.2.1. Nevertheless, it can still be done in a spreadsheet by dividing
the width of the packer into slices and calculating the stress in each slice.
The rotation angle is fixed and the maximum compression displacement
δmax is varied using a ‘goal seek’ function until the integral of the stresses is
in equilibrium with the hoop force.
The contact stress distributions as rotation is increased for the plastic
packer with parameters P1 is shown in Figure 9.9. Compared to the linear
elastic packer in Figure 9.7 the stress moves much more quickly to the edge
of the packer and attains much higher levels, but once there it changes less
and less. This is because as the strain increases, the plastic packer becomes
stiffer and stiffer, as was shown in Figure 9.8.
Other types of packer become less stiff as strain increases. This is the
opposite of the plastic or bituminous materials we have just seen, but is
typical behaviour of most engineering materials such as steel and concrete,
and indeed other materials used as packers, such as plywood and MDF.
Based on a compression test on a 3 mm plywood packer, the following
bilinear relationship could be assumed:
E = 170 MPa for 0 ≤ ε ≤ 0.05
E = 23 MPa for 0.05 ≤ ε ≤ 0.5
Figure 9.9 Contact stress distribution for nonlinear plastic packer with parameters P1.
The contact stress distribution for a 3 mm thick plywood packer is shown

in Figure 9.10 for different values of joint rotation. The stress distribution
here is completely different from Figure 9.9, being much more spread out
and avoiding the very high values found at the edge of the plastic packer.
Maximum compressive stresses are lower than for the linear elastic packer
Figure 9.10 Contact stress distribution for nonlinear 3 mm plywood packer.

and the potential for crushing or shearing of concrete at the edge of the
joint is therefore much reduced.
9.2.3 Comparison of different packer types

Figure 9.11 shows how the bending moment induced by eccentricity of con-
tact stress develops as the joint rotates for three different packer materials.
The joint with a plastic packer develops much higher bending moments
more quickly, as perhaps one might expect given the stress-strain relation-
ship in Figure 9.8 and the contact stress distribution in Figure 9.9. The
plywood packer also exhibits a nonlinear relationship but with a lower
magnitude. The linear elastic packer (remember, this type of packer does
not exist in the real world!) begins with a linear relationship between
moment and rotation angle, but as the contact stress distribution transi-
tions from a trapezoidal to a triangular shape, and the contact area begins
to shrink, it also exhibits a nonlinear relationship.
Since in all the cases the bending moment-joint rotation relation-
ship is not linear (except at low values of rotation for the linear elastic
packer), there is not a single value of rotational stiffness that can be
used for a particular packer type. In Figure 9.12 the rotational stiff-
ness is plotted against rotation angle for the segment geometry in
Worked Example 9.1, with the three different packers, and with a hoop
Figure 9.11 Bending moment vs joint rotation for plastic, plywood and linear elastic
packers, at a hoop force of 1000 kN/m.
Figure 9.12 Variation of rotational stiffness of joints with plastic, plywood and linear
elastic packers with rotation angle, at a hoop force of 1000 kN/m.
force of 1000 kN/m. All packer types settle into a rotational stiffness of
around 1000–3000 kNm/rad as found by Teachavorasinskun & Chub-
uppakarn (2010), but at lower levels of joint rotation, rotational stiff-
ness can be much higher.
It is difficult to briefly summarise the overall effect of these very dif-
ferent packer behaviours on how a segmental lining ring will behave. For
instance, a higher radial joint rotational stiffness will result in a higher
overall ring stiffness, which will result in less deformation but higher bend-
ing moments. A lower rotational stiffness will result in a more flexible ring,
with larger deformations but lower bending moments.
On the other hand, the high stress concentrations at the edge of the
packer that result from using plastic or bituminous packers that become
stiffer at higher strain levels, may have a knock-on effect on the design of
the concrete segment itself in the vicinity of the joint, as bursting, crush-
ing or shear failure may occur at the intrados or extrados due to the larger
eccentricity of the hoop force.
For detailed design it may be appropriate to feed these relationships
between rotational stiffness and rotation angle into a model of a tun-
nel (either bedded beam or finite element) to investigate the effect. It is
clear, however, that a material that has a stiffness that decreases with
strain level, such as concrete (i.e. no packer) or plywood, is preferable to
a material whose stiffness increases with strain level, such as bituminous
or plastic packers.
9.3 ESTIMATING LOCAL FORCES DUE

TO JOINT ROTATION
The simplest way to estimate local bending moments induced by joint rota-
tion is to specify the maximum rotation as an input and then to calculate
the eccentricity of axial force that will be generated. Joint rotation is not
usually specified, but ovalisation (the percentage change of diameter) is in
most cases explicitly specified in the contract as part of the build toler-
ances. For instance, the British Tunnelling Society (BTS) Specification for
Tunnelling (BTS, 2010) says:
The maximum and minimum measured diameters in any one ring shall
be within 1% of the theoretical design diameter of the ring measured on
completion of ring build and grouting, or such other tolerance stated in
the Particular Specification. This tolerance includes all building errors.
The first thing we need to do is to convert this specified ovalisation into a

joint rotation angle, then we can analyse the joint to calculate the eccentric-
ity of the hoop force.
As an aside, the specification of a maximum ovalisation as 1% of the
diameter results in rather large tolerances for tunnels larger than approxi-
mately 6.5 m diameter. Therefore, for larger tunnels, a value rather than a
percentage should be given.
9.3.1 Calculating joint rotation from

a specified ovalisation
The ovalisation is assumed to be elliptical (as we assumed in Chapter 6 for
the Curtis-Muir Wood analytical solution), as shown in Figure 9.13.
Figure 9.13 Elliptical distortion or ‘ovalisation’, shown for half a tunnel ring in squatting
mode.
The change in radius in Figure 9.13 is given by:
δ = ρ rm (9.16)
δ is the change in radius in mm

ρ is the percentage ovalisation/100, so for BTS (2010) this would be 0.01
r m is the radius to the centroid of the lining in mm
The worst case, in terms of joint rotation, would be for a joint to be located
at either the axis or the crown. Due to the symmetry of an ellipse, the rota-
tion in both these locations will be equal and opposite, so we only need to
calculate for one.
A segment to one side of a joint at the axis is shown in Figure 9.14,
and a further simplification of the geometry is shown in Figure 9.15.
In Figure 9.15, AB represents the original segment chord and DE represents
the new segment chord position. The segment is assumed to be rigid and
hence the chord length does not change.
The segment arc angle θ shown in Figures 9.14 and 9.15 is given by:
360°
θ= (9.17)
n
θ is the angle of the segment arc in °
n is the number of segments in the ring, assuming they are of equal size
From Figure 9.15, the joint rotation ψ will be given by the following equa-
tion, which is the difference between angles α and β, multiplied by 2 because
of symmetry (the adjacent segment below axis level will rotate as well):
ψ = 2 (α − β ) (9.18)
ψ is the joint rotation

α and β are the angles defined in Figure 9.15
Figure 9.14 Geometrical representation of a segment when the ring undergoes elliptical

distortion.
Figure 9.15 Geometrical simplification of segment rotation.
It can be seen from Figure 9.14 that the distance OG is given by:
OG = rm − δ (9.19)
For the undeformed perfectly circular lining we get the following

relationship:
OA = OB = rm (9.20)
The triangle OAB is an isosceles triangle, so the angle α in Figure 9.15 is

given by:
180° − θ
α= (9.21)
2
The length AC is given by:
AC = OAsinθ (9.22)
The length OC is given by:
OC = OAcosθ (9.23)
The length CB is given by:
CB = OB − OC (9.24)
The chord lengths AB and DE are assumed to be equal and are given by:
AB = DE = AC 2 + CB2 (9.25)
The ellipse major axis OE is:
OE = OB + δ (9.26)
The equation for an ellipse gives us:
DF 2 OF 2
+ = 1 (9.27)
OG2 OE2
Rearranged this gives:
 OF 2 
DF 2 =  1 −  OG (9.28)
2
 OE2 
Now Pythagoras’s equation for right-angled triangle DEF gives us:
(OE − OF )2 + DF 2 = DE2 (9.29)

Combining Equations 9.28 and 9.29 gives:
 OG2 
 1 −  OF − 2 × OE × OF + (OE + OG − DE ) = 0 (9.30)
2 2 2 2
OE2 
This is a quadratic equation of the form:
a × OF 2 + b × OF + c (9.31)
a, b, c are coefficients
To find OF, we first find the coefficients a, b and c:
 OG2 
a = 1 −  (9.32)
 OE2 
b = −2 × OE (9.33)
c = (OE2 + OG2 − DE2 ) (9.34)
Now to solve the quadratic equation:
b2 − 4 ac
OF = −b ± (9.35)
2a
There are two solutions to the quadratic equation, one being a spurious
result that will be much too large and can be discarded.
Now EF is given by:
EF = OE − OF (9.36)
And using a trigonometric identity for triangle DEF:
 EF 
β = arccos  (9.37)
 DE 
arccos is the inverse cosine
Once the angle β has been calculated, the joint rotation may be calculated
using Equation 9.18. Then we can calculate the eccentricity of the hoop
thrust across the joint, taking account of the joint geometry and packer
properties by using the procedure described in the previous Section 9.2.
Once the eccentricity is calculated, several aspects then need to be checked
including local compression (crushing of the concrete), bending stresses,
shear and bursting stresses.
WORKED EXAMPLE 9.2

Calculate the joint rotation for the precast concrete segmental lining
described in Worked Example 9.1, then calculate the eccentricity of the
hoop force.
The segmental lining has eight segments plus a key. It is to be built with
an ovalisation tolerance of 1%. The ring geometry is:
Internal diameter, ID = 6.0 m

Outside diameter, OD = 6.5 m
Lining thickness, t = 0.25 m
Usually the hoop force would be estimated using an analytical solution,

such as the Curtis-Muir Wood solution described in Section 6.2, or a
bedded beam model. For this example we will assume the design value of
the hoop force N Ed = 1000 kN/m.
A plywood packer is used in the joint, of thickness 3 mm and a bilinear
stress-strain relationship as described in Section 9.2, i.e. Ep = 170 MPa
for 0 ≤ ε ≤ 0.05 and Ep = 23 MPa for 0.05 ≤ ε ≤ 0.5.
First we calculate the radius to the lining centroid:
ID + t 6 + 0.25
rm = = = 3.125 m
2 2
Each segment arc angle θ = 360°/8 = 45°
Maximum change in radius, δ = 0.01r m = 0.01 × 3.125 = 0.03125 m
A segment to one side of a joint at the axis (with arc of 45°) is shown
in Figure 9.16, and a further simplification of the geometry is shown in
Figure 9.17. In Figure 9.17, AB represents the original segment chord and
DE represents the new segment chord position. The segment is assumed
to be rigid and hence the chord length does not change.
Figure 9.16 Worked Example 9.2 – Geometrical representation of a segment

when the ring undergoes elliptical distortion.
Figure 9.17 Worked Example 9.2 – Geometrical simplification of segment rotation.

It can be seen from Figure 9.16 that:
OG = rm − δ = 3.125 − 0.03125 = 3.09375 m
For the undeformed lining:
OA = OB = rm = 3.125 m
The triangle OAB is an isosceles triangle, so the angle α is given by:
180° − 45°
α= = 67.5°
2
The length AC is given by:
AC = OAsin45° = 3.125 × sin45° = 2.209709 m
The length OC is given by:
OC = OAcos45° = 3.125 × cos45° = 2.209709 m
Note we are carrying quite a few decimal places because we are calculat-
ing a small change in angle and the precision is needed.
The length CB is given by:
CB = OB − OC = 3.125 − 2.209709 = 0.915291 m
The chord lengths AB and DE are equal and are given by:
AB = DE = AC 2 + CB2 = 2.2097092 + 0.9152912 = 2.391771 m
The ellipse major axis OE is:
OE = OB + δ = 3.125 + 0.03125 = 3.15625 m
The equation for an ellipse gives us:
DF 2 OF 2
+ =1
OG2 OE2
Rearranged this gives:
 OF 2 
DF 2 =  1 − OG2
 OE2 
Now Pythagoras’s equation for right-angled triangle DEF gives us:
(OE − OF )2 + DF 2 = DE2
Combining these two equations gives:
 OG2 
 1 − OE2  OF − 2 × OE × OF + (OE + OG − DE ) = 0
2 2 2 2
This is a quadratic equation of the form:
a × OF 2 + b × OF + c
To find OF, we first find the coefficients a, b and c:
 OG2   3.093752 
a = 1 −  = 1 − = 0.039212
 OE2   3.156252 
b = −2 × OE = −2 × 3.15625 = −6.3125
c = (OE2 + OG2 − DE2 ) = (3.156252 + 3.093752 − 2.3917712 ) = 13.812632
Now to solve the quadratic equation for OF:
b2 − 4 ac ( −6.3125)2 − 4 × 0.039212 × 13.812632

−b ± = 6.3125 ±
2a 2 × 0.039212
= 2.218719 m
The second solution to the quadratic equation (158.7658 m) is spurious

and can be discarded.
Now EF is given by:
EF = OE − OF = 3.15625 − 2.218719 = 0.937531 m
And using a trigonometric identity for triangle DEF:
EF 0.937531
cosβ = = = 0.391982
DE 2.391771
And therefore the angle β is 66.922117°, and the joint rotation

ψ = 2(67.5° − β) = 1.156°.
Now we need to use this joint rotation to calculate the eccentricity of
the hoop thrust.
Figure 9.18 Worked Example 9.2 – Plywood packer contact stress distribution at

joint rotation of 1.156° corresponding to an ovalisation of 1%.
The contact stress distribution may be calculated using the method

described in Section 9.2, and the result of the spreadsheet calculation is
By finding the centroid of the stress distribution, the eccentricity
of the hoop thrust across the joint is 43.7 mm from the segment cen-
treline, and therefore the bending moment induced by this eccentricity is
43.7 kNm/m (the hoop thrust of 1000 kN/m multiplied by the eccentricity).
This ‘local’ bending moment can be compared to the ‘average’ bending
moment obtained by a global analysis. The larger of the two should be
used for design of the segment body.
Note that this calculation assumes that the packer is centred on the
joint centreline. This may not always be the case, and adjustments need
to be made to the methodology to allow for this.
9.3.2 Using eccentricity of hoop force to check

joint capacity in crushing and shear
The next step is to use the contact stress distribution and eccentricity of
the hoop force at the maximum ovalisation to check the joint for concrete
crushing or shear.
Figure 9.19 Shear plane across a segment joint.
From the contact stress distribution we know the maximum compressive

stress applied to the concrete joint surface through the packer. Eurocode 2
for concrete design, EN 1992-1-1: 2004 Cl. 5.10.2.2(5), requires this to be
limited to less than 0.6fck to avoid local crushing, where fck is the character-
istic cylinder compressive strength.
Knowing the eccentricity of the hoop force, we can also calculate whether
the corner of the segment could be sheared off. This kind of damage is quite
common in segmentally lined tunnels.
Figure 9.19 shows the shear plane. We need to calculate the shear stress
on the shear plane, then check that it is within the shear resistance capac-
ity of the segment. The shear stress will depend on the angle β assumed
for shear failure and the position of shear plane initiation on the joint
surface Lsh .
We could use the contact stress distribution rather than the resultant
hoop force. Then if the shear plane begins closer to the edge of the segment,
it will be shorter but the force acting on it will be lower. The position of the
shear plane Lsh could be varied to find the worst case. However, it is simpler
and more conservative to assume the shear plane begins under the resultant
hoop force, such that:
t
Lsh = − e (9.38)
2
The shear stress along the shear plane is given by:
N Ed sinβ cosβ
τ= (9.39)
Lsh × Ls
τ is the design value of shear stress in N/mm 2

N Ed is the design value of hoop force in N
β is the angle of the shear plane (see Figure 9.19)
Lsh is the distance from the beginning of the shear plane to the edge of
the segment in mm
Ls is the length of the segment in the tunnel longitudinal direction in mm
The maximum shear stress always occurs when β = 45°, because sin βcosβ
is at a maximum when β = 45°. Therefore, since sin45°cos45° = 0.5,
Equation 9.39 can be rewritten as:
N Ed
τ= (9.40)
2Lsh Ls
For a conventional steel bar-reinforced concrete, EN 1992-1-1: 2004 can be

used to check the shear capacity of the section using the following equation
to calculate shear resistance:
vRd ,c = CRd ,c k (100 ρl fck )

1/3
+ k1σ cp (9.41)
vRd ,c is the shear resistance as a stress in N/mm 2

CRd ,c has a recommended value of 0.18/ γc or a value defined in the coun-
try’s National Annex
γc is the partial material factor for concrete in compression and the
recommended value is 1.5
k = 1 + 200/d ≤ 2.0
d is the depth of the section in shear

ρl is the proportion of the shear area that is longitudinal steel bar rein-
forcement, where ρl = As /bw d
bw is the width of the member, in this case, it is the segment length in
the longitudinal tunnel direction, which we have denoted Ls
fck is the characteristic cylinder compressive strength of the concrete in
N/mm 2
k1 = 0.15 or a value defined in the country’s National Annex
σ cp = N Ed /bw d , which is the axial stress in the member in N/mm 2
For plain concrete, lightly reinforced concrete or SFRC, Section 12.6.3 of

EN 1992-1-1: 2004 can be used.
The axial stress σ cp is defined as for reinforced concrete. The shear stress
calculated in Equation 9.40 must be less than the ‘concrete design strength
in shear and compression’ fcvd.
The concrete design strength in tension fctd is needed for the calculation
of fcvd. It is given by:
α ct fctk
fctd = (9.42)
γc
fctd is the design value of concrete tensile strength in N/mm 2

αct is a coefficient to take account of “long-term effects on the tensile
strength and of unfavourable effects resulting from the way the
load is applied” (EN 1992-1-1: 2004). For reinforced concrete,
EN 1992-1-1: 2004 recommends α ct = 1.0 and this is also the value
in the UK National Annex (UK NA to BS EN 1992-1-1: 2004).
For plain concrete, EN 1992-1-1: 2004 recommends α ct , pl = 0.8
and the UK National Annex requires α ct , pl = 0.6. The reinforced
concrete value could be used for SFRC if ductility requirements of
the fib Model Code 2010 (2013) are met.
fctk is the characteristic value of concrete tensile strength in N/mm 2 .
For plain or SFRC concrete, the characteristic value fctk can be
obtained by splitting tests to EN 12390-6: 2009. For preliminary
design, the value from Table 3.1 of EN 1992-1-1: 2004 can be
used, and we will use that in the worked examples here.
γc is the partial material factor for concrete in tension and the recom-
mended value is 1.5
The concrete design strength in compression fcd is needed for the calcula-
tion of fcvd. It is given by:
α cc fck
fcd = (9.43)
γc
fcd is the design value of concrete compressive (cylinder) strength in

N/mm 2 .
αcc is a coefficient to take account of “long-term effects on the compres-
sive strength and of unfavourable effects resulting from the way
the load is applied” (EN 1992-1-1: 2004). For reinforced con-
crete, EN 1992-1-1: 2004 recommends α cc = 1.0, but the value
in the UK National Annex (UK NA to BS EN 1992-1-1: 2004) is
α cc = 0.85. For plain concrete, EN 1992-1-1: 2004 recommends
α cc , pl = 0.8 and the UK National Annex requires α cc , pl = 0.6. The
reinforced concrete value could be used for SFRC if ductility
requirements of the fib Model Code 2010 (2013) are met.
fck is the characteristic value of concrete compressive strength in N/mm2.
One of the following two expressions is used to obtain fcvd:
If σ cp ≤ σ c ,lim ; fcvd = fctd

2
+ σ cp fctd (9.44)
2
 σ cp − σ c ,lim 
If σ cp > σ c ,lim ; fcvd = fctd
2
+ σ cp fctd −   (9.45)
 2
fcvd is the concrete design strength in shear and compression in N/mm 2

σ cp = N Ed /bw d , which is the axial stress in the member in N/mm 2
σ c ,lim is a limiting compressive stress, where σ c ,lim = fcd − 2 fctd ( fctd + fcd ) ,
in N/mm 2
fcd is the concrete design strength in compression in N/mm 2
The following Worked Example will continue from where Worked Example 9.2
left off, using the previously calculated hoop force and eccentricity to check
for local crushing and shear of the concrete local to the joint.
WORKED EXAMPLE 9.3

Check for local crushing and shear local to the joint face for an SFRC
segment. The same geometry as in Worked Examples 9.1 and 9.2 is used.
The SFRC segment is FRC 50/60 5.0c according to the fib Model Code
2010 (2013).
Checking for local crushing:
The maximum contact stress in Figure 9.18 is 20.15 MPa. The
European Standard EN 1992-1-1: 2004 Cl. 5.10.2.2(5) requires this to
be limited to less than 0.6fck to avoid local crushing, where fck is the char-
acteristic cylinder compressive strength, in this case 50 MPa. Therefore
the contact stress is OK and there will be no local crushing failure.
Checking for local shear:
This kind of situation is not covered by the fib Model Code 2010
(2013), so the provisions of EN 1992-1-1: 2004 for plain concrete
should be used. For a concrete with a characteristic compressive
strength fck = 50 N/mm 2 , Table 3.1 of EN 1992-1-1: 2004 gives a char-
acteristic tensile strength fctk = 2.9 N/mm 2 .
For the values of design tensile and compressive strength of concrete,
we will use the values of αcc and αct for reinforced concrete, not plain
concrete, because the SFRC should prevent a brittle failure.
Taking α ct = 1.0, the EN 1992-1-1: 2004 ‘recommended value’ for

reinforced concrete in tension, and γ ct = 1.5, the design value of tensile
strength is given by:
α ct fctk 1.0 × 2.9

fctd = = = 1.93 N/mm 2
γ ct 1.5
Taking α cc = 0.85, the UK National Annex (UK NA to BS EN 1992-

1-1: 2004) value for reinforced concrete in compression, and γ cc = 1.5, the
design value of compressive strength is given by:
α cc fck 0.85 × 50
fcd = = = 28.33 N/mm 2
γ cc 1.5
Now, according to Section 12.6.3 of EN 1992-1-1: 2004:
σ c ,lim = fcd − 2 fctd ( fctd + fcd ) = 28.33 − 2 1.93 (1.93 + 28.33) = 13.05 N/mm 2
Similarly to a reinforced concrete calculation, the axial stress in the

member is given by:
σ cp = N Ed /bw d = 1000 × 103 /1000 × 81.3 = 12.3 N/mm 2
Therefore, σ cp ≤ σ c ,lim and we use Equation 9.44:
fcvd = fctd
2
+ σ cp fctd = 1.932 + 12.3 × 1.93 = 5.24 N/mm 2
Using units of N and mm, the shear stress is given by Equation 9.40:
N Ed 1000 × 103
τ= = = 6.15 N/mm 2
2Lsh Ls 2 × 81.3 × 1000
This is more than the value of fcvd, so the segment corner will fail in
shear. There may be ways to improve the geometry of the joint or a
different geometry or type of packer could be used. A thicker lining is
also a possibility, or alternatively, bar reinforcement could be provided
close to the radial joints.
9.4 BURSTING STRESSES
When ram loads are applied to the circumferential joint by the tunnel
boring machine (TBM) jacks, and when hoop force is transferred across
a radial joint, the force is not applied evenly across the whole segment
Figure 9.20 Bursting stress visualisation. (a) Stress trajectories, and (b) Horizontal stress.
thickness, but concentrated on a loading area defined by the contact width

of the packer or the size of the ram shoe.
For TBM jacks, usually pairs of hydraulic rams have a ‘shoe’ on the end,
which pushes on the circumferential joint. There are typically two of these
shoes on each segment with a gap between them, and so the loading area
also does not cover the full segment arc and bursting stresses can develop
in both directions, radially and circumferentially.
Whenever a compressive stress is applied to an area smaller than the sur-
face area of a segment face, then as this compressive stress spreads out with
depth into the section, orthogonal tensile bursting stresses are induced.
This is illustrated in Figure 9.20.
We can calculate bursting force Fbst using Leonhardt’s equations (repro-
duced in BS EN 1992-1-1: 2004 as Equation 9.14):
 a
Fbst = 0.25  1 −  N Ed (9.46)
 b
Fbst is the bursting force in kN, which can be assumed to act between
a depth of 0.1b and b.
a is the bearing width in mm, as shown in Figure 9.20.
b is the width of the segment face in mm, as shown in Figure 9.20 for a
centrally applied load. For an eccentric load, it is 2× the distance
to the nearest segment edge.
N Ed is the design value of axial force in kN, which could be the TBM
jacking force for a circumferential joint or the hoop force for a
radial joint.
This solution assumes the segment material is elastic.

For reinforced concrete, bars need to be detailed to resist this bursting

force. This is usually done by providing U-shaped returning bars at some
distance from the joint face.
For SFRC, it is more difficult because the SFRC is not elastic in tension,
except at very low stresses. Therefore, if the design stress exceeds the char-
acteristic value of the residual tensile strength at the limit of proportionality
divided by its partial material factor, then nonlinear numerical modelling
and/or full-scale laboratory testing will be needed. For example, Nogales &
de la Fuente (2020) describe laboratory testing and 3D numerical model-
ling of bursting due to TBM ram load application on an SFRC segment,
with the load applied both centrally and with an eccentricity.
For plain concrete, the bursting stress cannot exceed the design tensile
strength fctd.
9.5 CURVED JOINTS
Curved joints can be either convex-convex or convex-concave. Convex-

convex joints usually have the same radius on both joint faces. Convex-
concave, or ‘knuckle’ joints, will always have a larger radius on the concave
face compared to the convex face.
Usually it is only the radial joints that are curved, and the circumferential
joints are flat. An example of a curved radial joint is shown in Figure 9.21,
which uses joint dimensions from the Storebælt Eastern Railway Tunnel
(Elliott et al., 1996). The 7.7 m ID ring was made up of six segments plus
a key. The reinforced concrete segments were 400 mm thick and 1650 mm
long and were bolted and gasketed. There was no packer in the convex-
convex radial joints but there was a bituminous packer in the flat circum-
ferential joints.
Figure 9.21 Convex-convex radial joint for the Storebælt Eastern Railway Tunnel (joint
dimensions from Elliott et al., 1996).
It has been known for segments to have curved circumferential joints,

or even radial joints that are curved in the longitudinal direction as well
as across the thickness, but these are thankfully rare. These doubly curved
joints were used on the Jubilee Line in London in the 1970s and were very
difficult to build within tolerance. They also gradually deteriorated over
time and began moving and cracking, until they eventually needed to be
replaced with steel segments (TfL, 2015).
It is sometimes said (e.g. Luttikholt, 2007) that curved joints do not
transfer any bending moment. This is not true. When curved joints rotate,
there is an eccentricity produced, but it is smaller than for a flat joint. The
eccentricity due to curved joint rotation may be calculated using the fol-
lowing equation:
 RR 
e j =  1 2  θ (9.47)
 R1 + R2 
ej is the eccentricity of thrust due to joint rotation in mm

θ is the joint rotation in radians
R1 and R 2 are the joint radii, assuming they are convex-convex, in mm
The eccentricity of a curved joint e m due to a misalignment s may be calcu-

lated using the following equation:
 R1 
em =  s (9.48)
 R1 + R2 
e m is the eccentricity of hoop force across a curved radial joint in mm

s is a misalignment (or ‘lip’) between one segment and another in mm,
at a radial joint
As for flat joints, the total eccentricity is the sum of the eccentricity due to
joint rotation and the eccentricity due to misalignment.
Curved joints usually do not have packers, or bolts, but they can have a
gasket and a caulking groove (Elliott et al., 1996; Psomas & Eddie, 2016;
Nirmal, 2019). If we assume the two segments are elastic, then the contact
width b across a convex-convex radial joint with radii R1 and R 2 may be
calculated using elastic theory (Young, 1989), illustrated in Figure 9.22.
The equation for contact width is as follows:
N K C 
b = 1.6  d d E  (9.49)
 Lc 
b is the contact width in mm

Nd is the design hoop force in N
Lc is the contact length in mm
Kd is given by:
2R1R2
Kd = (9.50)
R1 + R2
R1 and R 2 are the radii of the curved radial joint surfaces of segment 1
and segment 2, respectively, in mm
C E is given by:
1 − ν12 1 − ν 22
CE = + (9.51)
E1 E2
E1 and v1 are the Young’s modulus in N/mm 2 and Poisson’s ratio of

segment 1
E 2 and v2 are the Young’s modulus in N/mm 2 and Poisson’s ratio of
segment 2
The contact length Lc will not necessarily be the length of the segment, as
there are usually chamfers at the ends to take account of.
The contact width allows us to calculate a contact stress that we can
check against concrete crushing, and also determines the bursting stress
that will be generated deeper in the segments below the contact. The con-
tact width and eccentricity allow us to check for local shear and bending
in the segment.
Figure 9.22 Contact width for a convex-convex radial joint.

9.6 MODELLING JOINTS IN BEDDED BEAM ANALYSES
When a bedded beam model is used, the radial joints may be modelled by
rotational hinges, which is a slight improvement on the 2D plane strain
analytical solutions, where the effect of joints is smeared or averaged
around the whole ring. If a finite element or frame analysis program is used,
then these hinges can be assigned a rotational stiffness. However, knowing
what the value of this rotational stiffness should be is not straightforward.
de Waal (2000) showed that the results will always be between the two
limiting situations of a monolithic ring with no joints and a ring with free
(or pin) joints. Lee & Ge (2001) estimated that it is between 1/10 and 1/4
of the segment stiffness. An analysis like that described in Section 9.2 could
be used to estimate rotational stiffness, or the rotational stiffness could
be calibrated to measured deformations of previously constructed tunnels
(similar to what Lee & Ge (2001) did for effective bending rigidity).
A 3D bedded beam model can be used to take account of the shear stiff-
ness of circumferential joints by introducing ‘shear springs’ into the model
between one ring and another, and the staggering of radial joints in adja-
cent rings. An example is shown in Figure 9.23 for a tunnel lining with
six segments and staggered radial joints. Two rings have been modelled to
allow the coupling effect between adjacent rings to be modelled, but only
half of the segment width is considered in each ring. The boundary condi-
tions do not allow displacement in the tunnel longitudinal direction, so this
Figure 9.23 3D bedded beam (or ‘beam-spring’) model of two six-segment half-rings
with staggered radial joints.
means the model behaves as though the structure were infinitely mirrored
in the longitudinal direction, forwards and backwards, and this is why it
makes sense to only consider half the segment width. The segments are
modelled by curved 1D beams. This kind of model can be easily built in
most 3D frame analysis programs.
In this kind of model, ground and water loads are applied to the beams as
distributed loads. The ground reaction is modelled by radial and tangential
springs, which are not shown in Figure 9.23. This is an approximation of
the soil-structure interaction, but the ground reaction springs are indepen-
dent and will not allow redistribution of stresses caused by arching within
the ground. For instance, a more flexible zone of lining would result in more
deformation in the ground locally and a local reduction in ground pressure,
which would result in arching and hence an increase in the ground pres-
sure either side of this zone. This becomes increasingly important as stress
gradients in the ground become larger, for example, in situations where the
tunnel is passing near to an existing tunnel, or perhaps when local grouting
pressures are applied.
Guidance on spring stiffness values to use can be found in Bakhshi &
Nasri (2013a, 2013b and 2013c). Kavvadas et al. (2017) found by paramet-
ric study that for the radial joints only the rotation about the longitudinal
tunnel axis direction is important and for the circumferential joints only
shear displacement is important. For the other degrees of freedom, high
stiffness values were used. For the radial joint rotational stiffness, values
could be based on full-scale testing or the methods described in Section 9.2.
For the circumferential joints, Gijsbers & Hordijk (1997) found from full-
scale tests that the shear stiffness value in the elastic zone could be assumed
to be 106 kN/m. The peak coefficient of friction varied between 0.4 and 0.7,
with higher values tending to be for lower normal forces and vice versa.
9.7 MODELLING JOINTS IN 2D OR 3D
NUMERICAL ANALYSIS
Using the finite element or finite difference method in 2D or 3D allows

the ground to be modelled by continuum elements, which is an improve-
ment over bedded beam models as it allows soil-structure interaction to be
modelled properly, although there will be a cost in terms of computation
time. It also allows complete flexibility of geometry in terms of modelling
nearby structures, varying surface or geological strata levels, fault zones or
otherwise nonuniform soil properties, as well as nonlinear soil or structure
behaviour and staged construction.
The simplest method of modelling a segmental tunnel lining is to use
beam elements in 2D, or shell elements in 3D. This means that the thick-
ness is not modelled explicitly but is taken account of implicitly in the axial
and flexural stiffness parameters. Therefore, joints cannot be modelled
explicitly, but can be represented by introducing rotational and shear

springs between beam or shell elements, in the same way as for a bedded
beam model. When these kinds of models are used, it is usually assumed
that the lining shell elements are placed at the extrados. This will model
the ground pressures more accurately, but the lining will behave as though
its centroid is at the extrados. This means it has a radius that is half the
segment thickness larger than it should be. This will slightly reduce bend-
ing stiffness, and this effect can be estimated using the following equation:
M* r02
= (9.52)
M r0 ( r0 − t )
M* is the bending moment calculated by the model

M is the bending moment in the real situation
t is the thickness of the lining
Kavvadas et al. (2017) used 3D numerical modelling of an advancing earth

pressure balance (EPB) TBM to compare the effect of aligned radial joints,
staggered radial joints and a continuous monolithic lining without radial or
circumferential joints. The results, in terms of bending moments and axial
forces in the lining, are shown in Figure 9.24. They showed that aligned
radial joints resulted in significantly lower bending moments but similar
axial forces to the continuous monolithic lining in their models. For stag-
gered radial joints, the internal forces in the lining varied from ring to ring
and so the results were presented as an envelope of forces that encompassed
the aligned radial joints and continuous lining forces.
Kavvadas et al. (2017) did not explicitly model the joint contacts but used
joint rules in ABAQUS to apply a rotational stiffness. The rotational stiff-
ness they used was based on concrete-to-concrete contact with no packer
Figure 9.24 Lining forces from a 3D finite element analysis with staggered joints, con-
tinuous lining and aligned joints (redrawn from Kavvadas et al., 2017). For
the axial force, compression is negative.
(described in detail in Litsas et al., 2015), using the method of Blom (2002)
and was therefore more than an order of magnitude higher than values pre-
sented earlier in this chapter where the properties of the packer were used.
This may be why the envelope for staggered joints has an upper bound of
bending moment close to the value for a continuous lining.
It is theoretically possible to model joints explicitly in a 3D numerical
model, by using continuum elements for the tunnel lining, and paying
close attention to the contact rules and geometry of the joints. It would
very likely be necessary to have a much finer mesh for the segments, par-
ticularly in the vicinity of the joints, than for the ground. Very few com-
mercially available software packages will allow sudden changes of mesh
size, or even contact rules (it is possible in ABAQUS and to some extent in
FLAC3D), while also being able to model the ground explicitly and with
a realistic constitutive model. A model built in this way will be very large
and complex. It is unlikely to be feasible at present for anything other than
an advanced research project. It is possible to model the lining with solid
elements in 3D in many structural finite element analysis programs with
nonlinear contacts modelled, but there will be limitations in the way the
ground is modelled.
Attempts have been made to model joints using finite elements and con-
tact rules, but without modelling the ground, in order to obtain rotational
stiffness values. These kinds of models are usually used to back-analyse
full-scale segment tests, to check on stress concentrations around details,
such as rebates, shear keys, gasket grooves, grout holes or bolt holes, or to
find values of joint rotational and shear stiffness for use in simpler models.
9.8 IN SERVICE LOADS – ULTIMATE LIMIT

STATE DESIGN
The ground and water loads applied to the segmental lining should be the
most onerous loads in terms of design of the lining thickness and mate-
rial strength. In preliminary design, the lining bending moment and hoop
force may be obtained from an analytical solution, such as the Curtis-Muir
Wood solution (see Chapter 6), whereas in detailed design it is common
practice to use numerical modelling in 2D or 3D to obtain the lining forces
(see Chapter 7).
Usually, the design models will have various load cases and combina-
tions, which may include:
• a representative sample of cross-sections along the route, representing

the highest and lowest overburden or other geometric or geological
variations
• short-term and long-term groundwater levels
• undrained and drained soil behaviour
• for twin tunnels, the distance between them or their level relative to
each other may vary along the route
• sensitivity analyses, e.g. of a realistic range of coefficient of earth
pressure at rest, soil strength or stiffness parameters, or confinement
loss factor (see Section 7.1.2)
These load cases and combinations will each result in lining forces, usu-
ally expressed as a hoop thrust N, a bending moment M and shear force V,
which will vary around the lining. Therefore, we will have a large number
of points at which we have calculated a set of N, M and V and we need
an efficient way of dealing with all of these sets. We cannot just take the
maximum N and pair it with the maximum M and maximum V, because
they may not have occurred at the same point in the lining or in the same
design model.
An important aspect to bear in mind is that a high value of N is not nec-
essarily a bad thing, because it can increase the moment capacity at low to
moderate load levels. This is why we should check the lowest overburden
and the lowest groundwater table level as well as the highest.
9.8.1 Application of partial factors

to the lining forces
Each set of lining forces N, M and V must have a partial factor of 1.35
applied to them according to EN 1997-1: 2004 +A1: 2013, if they are unfa-
vourable loads. Since it is difficult to ascertain whether applying the partial
factor will be beneficial or not, the lining should be checked for both fac-
tored and unfactored loads.
To illustrate this, the moment-hoop force interaction diagram for a
200 mm thick, 1000 mm long SFRC segment with a characteristic compres-
sive cylinder strength fck = 40 N/mm 2 and a characteristic value of flexural
tensile residual strength at crack mouth opening distance (CMOD) = 2.5 mm,
fR3k = 4.5 N/mm 2 is shown in Figure 9.25. The eye-shaped line is the capac-
ity envelope and points anywhere inside this line are within the capacity of
the section. The unfactored loads are shown by the solid marker. In this case,
if the hoop force is factored it is favourable because the moment capacity is
increased. The worst case would be to factor M but not N.
Most practitioners argue that N and M are interdependent because
each set is from the same point in the lining in the same design model and
therefore constitutes a ‘single source’. Therefore, both N and M should be
factored or unfactored together. However, the partial factor on actions or
the effects of actions is to take account of “the possibility of unfavour-
able deviations of the action values from the representative values” (EN
1990: 2002 +A1: 2005, Section 6.3.1), and it is possible to imagine that
M could increase without also affecting N. Situations that could cause this
include: unforeseen heterogeneity of ground or drainage conditions, a value
Figure 9.25 An example moment-hoop force interaction diagram showing the effect of
application of partial factors.
of K0 outside the range assumed, soil strength or stiffness anisotropy not

included in the model, 3D effects not included in the design model, geo-
metrical imperfections, adjacent construction, variations in grouting proce-
dures. Therefore, it is my personal view that as well as applying the partial
factor to both M and N, the combination where a partial factor is applied
to M but not to N should also be checked. Alternatively, the traditional
method of applying a partial factor of 1.0 or 1.35 to each pair of M and
N could be used, as long as consideration is given to design situations that
may cause additional bending moments to arise.
9.8.2 Constructing a moment-axial
force interaction diagram
The key part of segmental lining design for geotechnical actions is to con-
struct a moment-axial force interaction diagram. This is commonplace for
the design of reinforced concrete columns, but for SFRC segments it is not
described in the fib Model Code 2010 (2013) nor in detail in any of the
supporting guidance for segment design such as fib Bulletin 83 (2017) or
ITAtech (2016). The procedure is described here, for SFRC segments with-
out bar reinforcement.
The inputs are:
f R3k is the characteristic value of the SFRC flexural tensile residual

strength at CMOD = 2.5 mm, obtained from a suite of EN 14651:
2005 beam tests, in N/mm 2
fck is the characteristic value of cylinder compressive strength in

N/mm 2
αcc is a coefficient that reduces the compressive strength to take account
of long-term effects and the application of load (EN 1992-1-1: 2004,
Section 3.1.6). The value depends on the National Annex. The recom-
mended value is 1.0 but in the UK the value is 0.85 (UK NA to BS EN
1992-1-1: 2004).
γc is the partial material factor for concrete in compression, usually
1.5
γF is the partial material factor for SFRC, usually 1.5 (see Section 8.2)
λ is the neutral axis depth factor, usually taken as 0.8
b is the member width, which in this case is the segment length in the
tunnel longitudinal direction, in mm
h is the member depth, which in this case is the segment thickness,
in mm
The first step is to calculate the design values of the strength parameters.
The design compressive strength is given by:
α cc fck
fcd = (9.53)
γc
fcd is the design compressive strength in N/mm 2
The design value of flexural tensile residual strength at CMOD = 2.5 mm

is given by:
fR3k
fR3d = (9.54)
γF
f R3d is the design value of flexural tensile residual strength at CMOD =

2.5 mm, in N/mm 2
The design value of the ultimate post-cracking tensile strength of the SFRC
is given by:
fR3d
fFtud = (9.55)
3
f Ftud is the design value of the ultimate post-cracking tensile strength of

the SFRC, in N/mm 2
The behaviour in bending can be simplified to two rigid plastic stress

blocks, as shown in Figure 9.26. We will not include bar reinforcement in
this procedure, though it is fairly straightforward to add it in.
Figure 9.26 Rigid-plastic stress block approach for ultimate limit state (ULS) design of a
fibre-reinforced concrete segment.
The symbols used in Figure 9.26 that we have not already defined are:
xn is the distance from the compressive extreme fibre to the neutral axis
in mm.
η is a factor that is used to reduce the design compressive strength at
high strength values. For fck ≤ 50 MPa, η = 1.
A s is the area of steel bar reinforcement (if present) in mm 2 .
f yd is the design yield strength of the steel bar reinforcement (if present)
in N/mm 2 .
If we have no axial force, then we can directly calculate the depth to the
neutral axis xn using horizontal force equilibrium:
λ xnη fcd = ( h − xn ) fFtud (9.56)
Rearranging gives:
hfFtud
xn = (9.57)
λη fcd + fFtud
Now we know xn we can take moments about the neutral axis to calculate
the resistance moment of the section at the ULS Mu:
b ( h − xn )
2
 λx 
Mu = fFtud + η fcd bλ xn  xn − n  (9.58)
2  2 
Mu is the ULS moment in Nmm
In order to construct a moment-axial force interaction diagram, set up

a spreadsheet with the first column containing values of axial force.
We will assume compression is positive and tension is negative. To
find the minimum axial force N min when Mu = 0 , use the following
equation:
N min = − fFtud bh (9.59)
Nmin is the minimum axial force in N
The answer will be in units of N, so divide by 1000 if you want kN.

To find the maximum axial force Nmax when Mu = 0, use this equation:
N max = fcd bh (9.60)
Nmax is the maximum axial force in N
Set the values of axial force to be separated by equal increments between

the minimum and maximum values. Approximately 20 increments will
give sufficient resolution in the diagram.
In the second column, divide the axial force in the first column by the
cross-sectional area bh to give the axial stress σ a as follows:
N
σa = (9.61)
bh
In a third column, calculate the stress capacity available for tension in
bending σ t , avail :
σ t , avail = σ a + fFtud (9.62)
And in a fourth column, calculate the stress capacity available for compres-
sion in bending σ c , avail :
σ c , avail = η fcd − σ a (9.63)
Now, in a fifth column, calculate the depth to the neutral axis xn using
Equation 9.57, with f Ftud replaced with σ t , avail and ηfcd replaced with σ c , avail ,
as follows:
hσ t , avail
xn = (9.64)
λσ c , avail + σ t , avail
Now, in a sixth column, Mu may be calculated using Equation 9.58, again

with f Ftud replaced with σ t , avail and ηfcd replaced with σ c , avail , as follows:
b ( h − xn )
2
 λx 
Mu = σ t , avail + σ c , avail bλ xn  xn − n  (9.65)
2  2 
Now plot a graph with values of axial force N from the first column on
the x-axis and values of bending moment Mu from the sixth column on the
y-axis. It should have a similar shape to Figure 9.25.
WORKED EXAMPLE 9.4

A precast steel fibre-reinforced concrete segmental lining is to be
designed for a 4.5 m internal diameter tunnel. The concrete is of grade
FRC 40/50 – 5.0c. Take α cc = 0.85, λ = 0.8, γ c = 1.5 and γ F = 1.5.
Construct a moment-axial force interaction diagram for the ultimate
limit state, taking the lining thickness t = 250 mm and the segment length
b = 1000 mm.
The concrete is grade FRC 40/50 – 5.0c, which means that:
fck = 40 N/mm 2
fR1k = 5.0 N/mm 2
The letter ‘c’ means that 0.9 ≤ f R3k /f R1k ≤ 1.1, so we will take the lower value:
fR3k = 0.9fR1k = 0.9 × 5.0 = 4.5 N/mm 2
Now γ F = 1.5, so:
fR3k 4.5
f R3 d = = = 3.0 N/mm 2
γF 1.5
And:
fR3d 3.0
fFtud = = = 1.0 N/mm 2
3 3
Also, α cc = 0.85 and γ c = 1.5, so:
α cc fck 0.85 × 40
fcd = = = 22.67 N/mm 2
γc 1.5
The range of axial force we need to calculate is bounded by:
N min = − fFtud bh = −1.0 × 1000 × 250 = −250 × 103 N = −250 kN
N max = fcd bh = 22.67 × 1000 × 250 = 5667 × 103 N = 5667 kN
This calculation shows ten equal intervals of axial force N, but you can
do more if you want to. The calculation results are shown in Table 9.2.
Table 9.2 Calculation of values to construct a moment-axial force interaction

diagram.
Available Available Depth to
Axial Axial for for neutral Ultimate
force stress tension compression axis moment
N σa σ t , avail σ c , avail xn Mu
(kN) (N/mm2) (N/mm2) (N/mm2) (mm) (kNm)
−250.00 −1.000 0.000 23.667 0.00 0.00
341.67 1.367 2.367 21.300 30.49 66.52
933.33 3.733 4.733 18.933 59.52 118.07
1525.00 6.100 7.100 16.567 87.21 154.56
2116.67 8.467 9.467 14.200 113.64 176.03
2708.33 10.833 11.833 11.833 138.89 182.61
3300.00 13.200 14.200 9.467 163.04 174.48
3891.67 15.567 16.567 7.100 186.17 151.87
4483.33 17.933 18.933 4.733 208.33 115.05
5075.00 20.300 21.300 2.367 229.59 64.32
5666.67 22.667 23.667 0.000 250.00 0.00
The values can be plotted on a graph, as shown in Figure 9.27. Note

that because we have a symmetrical section, the negative moments (hog-
ging) are the same but with the sign reversed.
Figure 9.27 Worked Example 9.4 moment-axial force interaction diagram.

9.8.3 Design for shear

Coccia et al. (2015) proposed a method of including the effect of shear in
the ULS check for moment and axial force, by reducing the design value of
the ultimate post-cracking tensile strength f Ftud as follows:
  τ 2
fFtud
red
= fFtud 1 −    (9.66)
  fFtud  
fFtud
red
is the reduced value of f Ftud in N/mm 2
f Ftud is the design value of the ultimate post-cracking tensile strength
in N/mm 2
τ is the shear stress, given by τ = V /bh, in N/mm 2
The moment-axial force interaction diagram should then be constructed

with the reduced value fFtud
red
in place of f Ftud.
9.9 IN SERVICE LOADS – SERVICEABILITY

LIMIT STATE DESIGN
So far we have only looked at the ULS design under geotechnical loads.
We also need to design the tunnel for the serviceability limit state (SLS), to
demonstrate that it will meet whatever criteria are needed for it to remain
serviceable for the whole design life.
9.9.1 Estimating crack widths

Specifying the maximum allowable crack width for the SLS was discussed
in Sections 8.2.6 and 8.2.7. Here we are concerned with how to estimate
crack widths during design of a tunnel lining to ensure they are smaller
than this specified maximum allowable value.
Estimating crack widths in reinforced concrete is a subject that is well
covered in design guides and codes, such as EN 1992-1-1: 2004. For SFRC
it is still an area that requires further research, and really we can only
be sure of any estimate by undertaking full-scale laboratory tests. This
is because for a given strain in an SFRC segment, the size of cracks will
depend on the spacing between the cracks. In other words, a given tensile
strain could result in either multiple small cracks or one big one.
Methods of crack width estimation are therefore based on simplifying
assumptions that require verification by testing (e.g. Johnson et al., 2017).
Without testing, calculations can only provide an average crack width
based on a mean tensile strain.
Since most tunnel linings have significant hoop forces and the section is
therefore mainly in compression, some tunnel designers advocate a simple
method where the lining is assumed to be elastic in the analysis and any ten-
sile stresses are then checked to ensure they are below the first crack value.
This avoids the need to model the tunnel lining as a nonlinear material.
9.9.2 Constructing a moment-axial force

interaction diagram for the SLS
It can be useful to define a capacity envelope for moment-axial force inter-
action at the SLS. This is much more complicated than for the ULS and
guidance is scarce.
fib Bulletin 83 (2017) proposes one method, but the explanation is very
brief and does not explain all the steps. The strain distribution is assumed
to be linear, with a maximum compressive stress equal to 0.6fck at the com-
pressive extreme fibre and the maximum tensile strain defined by the speci-
fied maximum crack width for serviceability.
To construct a moment-axial force interaction diagram, we must calcu-
late the strain distribution, convert this into a stress distribution and then
calculate the moment and axial force.
To calculate the maximum compressive strain, we need to estimate the
Young’s modulus of the concrete to convert 0.6fck into a compressive strain.
This should probably include an allowance for plasticity and creep.
σ c ,max 0.6fck
ε c ,max = = (9.67)
Ec Ec
ε c ,max is the maximum compressive strain

σ c ,max is the maximum compressive stress in N/mm 2
Ec is the Young’s modulus of the concrete in compression in N/mm 2
fck is the characteristic compressive strength of the concrete N/mm 2
To calculate the maximum tensile strain, we need to use the following

equation to convert crack width to strain:
wd
ε t ,max = (9.68)
s
ε t ,max is the maximum tensile strain

wd is the maximum allowable crack width in mm
s is the crack spacing in mm
The key assumption here is the crack spacing s. There is currently no

clear guidance available on what this should be, since the fib Model Code
2010 (2013) does not consider the influence of axial force. The charac-
teristic crack spacing could be determined by testing (for an example, see
Hover et al., 2017). A potential design method is proposed by Johnson
et al. (2017), where crack spacing is assumed to be equal to the depth of

the crack, but this needs validation by testing. The fib Bulletin 83 (2017)
states that crack spacing may be assumed to be equal to the segment
thickness.
9.10 SUMMARY
Joints in segmental linings have an effect on the overall stiffness of the

lining. Muir Wood (1975) proposed a simple expression to reduce the
moment of inertia of the lining to account for the number of radial
joints in a ring. However, studies have shown that tunnel linings with
staggered radial joints have a behaviour that is closer to a monolithic
ring.
The behaviour of joints undergoing rotation may be analysed using rela-
tively simple methods to determine the rotational stiffness. This could then
be used in a bedded beam analysis or finite element analysis program to
model the interaction of joints in adjacent rings.
On most projects, a maximum ovalisation is specified in the contract.
This can be used to calculate the maximum joint rotation, which can in
turn be used to calculate the eccentricity of axial force across the joint. We
can then use this eccentricity of axial force to calculate a bending moment,
and to check for local crushing, local shear and bursting stress at or near
the joint.
Curved radial joints are also used, and generally speaking they tend to
reduce eccentricity of hoop force across the joint. The bearing area and
eccentricity may be calculated, and from these values the same design
checks on bending moment, local crushing, local shear and bursting stress
may be performed as for flat joints with packers.
In service, the tunnel lining is loaded by the ground and groundwater
pressures. This imposes hoop forces, bending moments and shear forces
on the lining. For concrete segments, the capacity at the ULS may be veri-
fied by plotting a moment-axial force interaction diagram. For plain or
reinforced concrete, EN 1992-1-1: 2004 may be used and for SFRC, fib
Model Code 2010 (2013) and fib Bulletin 83 may be used to construct the
interaction diagram. To take account of shear force for SFRC segments, a
reduction of the design value of the ultimate post-cracking tensile strength
f Ftud was proposed by Coccia et al. (2015).
For the in service loading of SFRC segments at the SLS, the limiting
parameter is crack width. There is much research still to be done before
design rules can be codified, and the current methods are exceedingly com-
plex. It was decided that this was beyond the scope of this textbook, which
only seeks to cover the basic principles of design using well-established and
proven methods.
9.11 PROBLEMS
Q9.1. An 8 m ID, 400 mm thick segmental lining is to be designed with

ten precast concrete segments per ring. The ring length is 2 m in
the tunnel longitudinal direction. The design value of hoop force
is 2200 kN/m. The radial joints are flat joints with dimensions as
Figure 9.28 Q9.1 radial joint dimensions.
i. The specified ovalisation is 1%. Calculate the maximum radial

joint rotation.
ii. Calculate the eccentricity of hoop force caused by a misalign-
ment of the radial joint of 10 mm.
iii. Assuming the packer is linear elastic, with Young’s modulus
Ep = 40 MPa, length Lp = 1900 mm and thickness t p = 4 mm,
calculate the eccentricity of hoop force caused by the maxi-
mum radial joint rotation from part (i).
iv. Calculate the total eccentricity of hoop force.
v. Calculate the bending moment induced by the total eccentricity.
vi. Calculate the bursting force at the total eccentricity.
vii. [Advanced] Calculate the eccentricity if the packer is bitumi-
nous, using the parameter set ‘B2’ from Table 9.1 and with
the same length Lp = 1900 mm and thickness t p = 4 mm as for
the linear elastic packer. Then calculate the bending moment
induced by the total eccentricity.
Q9.2. A 6.6 m ID, 300 mm thick segmental lining has eight precast
concrete segments per ring. The ring length is 1.5 m long in the
tunnel longitudinal direction. The radial joints are flat with 3 mm

thick plywood packers that are 1400 mm long and 200 mm wide,
positioned centrally. The plywood packers are 4 mm thick and
have a bilinear stress-strain relationship with E = 170 MPa for
0 ≤ ε ≤ 0.05 and E = 23 MPa for 0.05 ≤ ε ≤ 0.5. The design value
of hoop force is 1500 kN/m.
i. The specified ovalisation is 0.7%. Calculate the maximum
joint rotation.
ii. Using a spreadsheet, plot a graph of the contact stress distri-
bution at the maximum joint rotation. Check it by integrating
the stresses and comparing the result with the hoop force.
iii. Calculate the eccentricity of hoop force caused by the maxi-
mum joint rotation by taking moments of the stress distribu-
tion about the centre of the segment thickness.
iv. Calculate the bending moment induced by the eccentricity of
hoop force.
v. The precast concrete is of class FRC 50/60 – 5.0d according
the fib Model Code 2010 (2013). Check the concrete bearing
stress under the packer using the maximum stress calculated in
the contact stress distribution found in part (ii). If you skipped
the earlier parts, then use a maximum stress of 18 N/mm 2 .
vi. Check also that the eccentricity at the maximum joint rotation
will not cause local shear failure of the corner of the segment.
If it fails, what would you do to improve the design?
Q9.3. A segmental lining is designed with FRC 50/60 – 5.0d of thick-
ness 300 mm.
i. Construct a moment-axial force interaction diagram at the
ULS. Use α cc = 0.85, λ = 0.8, γ c = 1.5 and γ F = 1.5. Do the
calculations per metre length of tunnel, i.e. take the segment
length in the tunnel longitudinal direction as 1000 mm.
ii. The results of numerical modelling of the tunnel are in
Table 9.3. These pairs of actions have not been factored.
Plot them on the moment-axial force interaction diagram
created in (i) as both unfactored and factored values. Do
they exceed the ULS?
Table 9.3 Q9.3 Axial force and bending moment
values from numerical modelling.
N (kN/m) M (kNm/m)
1997 185
1912 130
1657 163
2339 205
1321 144
2027 233
iii. The last pair of values in Table 9.3 should have exceeded the
capacity of the segment. Change the segment design so that all
the pairs of values are within the capacity curve.
REFERENCES
Bakhshi, M. & Nasri, V. (2013a). Structural design of segmental tunnel linings.

3rd International Conference on Computational Methods in Tunnelling
and Subsurface Engineering: EURO:TUN 2013. Ruhr University Bochum,
17–19 April.
Bakhshi, M. & Nasri, V. (2013b). Latest developments in design of segmental tun-
nel linings. Canadian Society for Civil Engineering General Conference.
Montréal, Québec, 29 May–1 June.
Bakhshi, M. & Nasri, V. (2013c). Practical aspects of segmental tunnel lining
design. Underground – the way to the future, Proc. World Tunnel Congress
2013, Geneva, Switzerland (eds Anagnostou, G. & Ehrbar, H.). London:
Taylor & Francis Group.
Blom, C. B. M. (2002). Design philosophy of concrete linings for tunnels in soft
soils. PhD thesis, Technical University of Delft.
BTS (2010). Specification for tunnelling, 3rd Edition, including 2012 errata.
Cavalaro, S. H. P. & Aguado, A. (2012). Packer behavior under simple and coupled
stresses. Tunn. Undergr. Space Technol. 28, 159–173.
Coccia, S., Meda, A. & Rinaldi, Z. (2015). On shear verification according to fib
Model Code 2010 in FRC elements without traditional reinforcement. Struct.
Concr. 16, No. 4, 518–523.
Elliott, I. H., Odgård, A. S. & Curtis, D. J. (1996). Storebælt eastern railway tun-
nel: design. Proc. ICE Civil Eng. Supp. 114, Special Issue 1, 9–19.
EN 12390-6: 2009. Testing hardened concrete – Tensile splitting strength of test
specimens. Brussels: European Committee for Standardization.
EN 14651: 2005 (2005). Test method for metallic fibered concrete – Measuring the
flexural tensile strength (limit of proportionality (LOP), residual). Brussels:
EN 1990: 2002 +A1: 2005 (2010). Eurocode – Basis of structural design, incor-
porating corrigenda December 2008 and April 2010. Brussels: European
EN 1992-1-1: 2004 (2004). Eurocode 2: Design of concrete structures – Part 1-1:
Standardization.
EN 1997-1: 2004 +A1: 2013. Eurocode 7: Geotechnical design – Part 1: General
rules, incorporating corrigendum February 2009. Brussels: European
EN 206-1: 2000 (2004). Concrete – Part 1: Specification, performance, produc-
tion and conformity, incorporating Corrigendum No. 1 and Amendment
No. 1. Brussels: European Committee for Standardization.
Fei, Y., Chang-fei, G., Hai-dong, S., Yan-peng, L., Yong-xu, X. & Zhuo, Z. (2014).
Model test study on effective ratio of segment transverse bending rigidity of
shield tunnel. Tunn. Undergr. Space Technol. 41, 193–205.
fib Bulletin 83 (2017). Precast tunnel segments in fibre-reinforced concrete, State-
of-the-art Report, fib WP 1.4.1, October. Lausanne: Fédération internatio-
nale du béton (fib).
fib Model Code 2010 (2013). fib Model Code for concrete structures 2010.
Gijsbers, F. B. J. & Hordijk, D. A. (1997). Experimenteel onderzoek naar
bet afschuifgedrag van ringvoegen, 1st November, TNO Bouw No.97-
CON-R1337, COB Reference K111-W-001, Delft, The Netherlands.
Hefny, A. M. & Chua, H. -C. (2006). An investigation into the behaviour of jointed
tunnel lining. Proceedings of the ITA-AITES 2006 World Tunnel Congress –
Safety in the Underground Space, Seoul, Korea, 22–27 April.
Hover, E., Psomas, S. & Eddie, C. (2017). Estimating crack widths in steel fibre-
reinforced concrete. Proc. Inst. Civil Eng. Constr. Mater. 170, No. 3,
141–152.
ITAtech (2016). Design guidance for precast fibre reinforced concrete segments –
Vol. 1: Design aspects, ITAtech Report No. 7. Lausanne: ITA-AITES.
Reproduction of figures with permission of the Committee on new technolo-
gies of the International Tunnelling and Underground Space Association –
ITAtech.
Johnson, R. P., Psomas, S. & Eddie, C. M. (2017). Design of steel fibre rein-
forced concrete tunnel linings. Proc. Inst. Civil Eng. Struct. Build. 170, SB2,
February, 115–130.
Kavvadas, M., Litsas, D., Vazaios, I. & Fortsakis, P. (2017). Development of a
3D finite element model for shield EPB tunnelling. Tunn. Undergr. Space
Technol. 65, 22–34. Figure 10-24 reprinted with permission from Elsevier.
Klappers, C., Grübl, F. & Ostermeier, B. (2006). Structural analyses of segmen-
tal lining – coupled beam and spring analyses versus 3D-FEM calculations
with shell elements. Proceedings of the ITA-AITES 2006 World Tunnel
Congress – Safety in the Underground Space, Seoul, Korea, 22–27 April.
Lee, K. M. & Ge, X. W. (2001). The equivalence of a jointed shield-driven tunnel
lining to a continuous ring structure. Can. Geotech. J. 38, 461–483.
Litsas, D., Paterianaki, G. & Kavvadas, M. (2015). Investigation of the influence of
cracking on the stiffness and capacity of segmental tunnel lining. Proceedings
of the World Tunnel Congress (ed. Kolić, D.), Dubrovnik, Croatia, 22–28
May. Zagreb, Croatia: HUBITG.
Luttikholt, A. (2007). Ultimate limit state analysis of a segmented tunnel lining –
Results of full-scale tests compared to finite element analysis. Master’s the-
sis, Delft University of Technology.
Muir Wood, A. M. (1975). The circular tunnel in elastic ground. Géotechnique 25,
No. 1, 115–117.
Nirmal, S. S. (2019). Design of steel fibre reinforced concrete segment with curved
radial joints. Proceedings of the WTC 2019 – Tunnels and Underground
Cities: Engineering and Innovation meet Archaeology, Architecture and Art
(eds Peila, D., Viggiani, G. M. B. & Celestino, T.), pp. 2804–2809. London:
Nogales, A. & de la Fuente, A. (2020). Crack width design approach for fibre rein-
forced concrete tunnel segments for TBM thrust loads. Tunn. Undergr. Space
Technol. 98, Paper 103342.
Psomas, S. & Eddie, C. M. (2016). SFRC segmental lining design for a pressurised
tunnel. Proceedings of the WTC2016, San Francisco, USA, Paper 0118.
Englewood, CO: SME.
Teachavorasinskun, S. & Chub-uppakarn, T. (2010). Influence of segmental joints
on tunnel lining. Tunn. Undergr. Space Technol. 25, 490–494.
TfL (2015). Jubilee Line Tunnel Works, Rail and Underground Panel paper, 16th
July. Available at: http://content.tfl.gov.uk/rup-20150716-part-1-item06-
jubilee-line-tunnel-works.pdf (last accessed 11th August 2019).
UK NA to EN 1992-1-1: 2004. UK National Annex to Eurocode 2: Design of
concrete structures – Part 1-1: General rules and rules for buildings. London:
Young, W. C. (1989). Roark’s formula for stress and strain, 6th Edition. New York:
McGraw-Hill.
Chapter 10
Segment design
for transient loads
As well as designing segments for their long-term condition as part of a

tunnel lining, we also need to consider the transient load cases that occur
before then.
This chapter will describe how segments may be designed for demould-
ing, storage, transportation, handling and erection. These are different to
the in-service loading, because a single segment with external loads applied
to it in these ways is a structurally determinate load case. This means that
the internal forces in the segment may be determined by equilibrium of the
external forces. Another difference is that there is no hoop force and the
segments must be able to resist bending moments and shear forces without
the (usually) beneficial effect of axial force.
The chapter will then discuss installation loads, such as gasket compres-
sion, bolt loads, application of the tunnel boring machine (TBM) jacks and
grouting pressures.
• the main load cases that apply to segments from casting in a factory
to installation in a tunnel
• the complex interaction of TBM jacking loads and grouting pressures
and their effect on longitudinal bending of the tunnel lining
After working through this chapter you will be able to:
• design a concrete segment for transient loads, made from reinforced

concrete, plain concrete or steel fibre reinforced concrete

tunnel segments, but a teaching aid to demonstrate the principles and most
important aspects of design. Every project is unique and has different load
cases and performance requirements, some of which may not be included
here. But with the understanding gained from working through this chap-
ter, and with careful reading of the papers, guidelines and standards that
are cited, you will have a solid foundation.
DOI: 10.1201/9780429470387-10 323
10.1 DEMOULDING
Assuming that the moulds and curing process of the precast concrete seg-
ments are designed to minimise adhesion or trapping of the segment in
the mould, the load case here is the lifting of a segment against its self-
weight. The concrete is young, perhaps with an age between 3 and 12 hours,
depending on the concrete mix and the curing process (fib Bulletin 83,
2017). Cycle times in the factory and hence the size of the factory and
the number of segment moulds required depend on the time from casting
to demoulding, so this could have an impact on cost and programme.
There are a number of factors at play and the rate of strength gain also
depends on the concrete mix ingredients and proportions. This will have
an impact on durability and carbon footprint, so it is important that the
designer communicates with the segment manufacturer to achieve the
optimum solution.
Usually the designer will specify a concrete strength required and
it will be up to the segment factory to ensure the concrete has attained
this strength before demoulding, usually by using standard cube or cylin-
der compressive strength tests as an index test. For SFRC, a relationship
between the flexural tensile strength required to avoid cracking and the
compressive strength needs to be established at early ages to determine a
value of compressive strength at which it is safe to demould the segment
(fib Bulletin 83, 2017).
Designing for demoulding requires knowledge of the lifting equipment.
For smaller segments, lifting points may be cast into the segment. In most
cases nowadays, and particularly for larger segments, demoulding will be
done by vacuum lifting. Vacuum lifters usually have one to three areas
where a vacuum is applied. Segments are usually cast intrados-down, so the
vacuum lifter is applied to the extrados as shown in Figure 10.1.
The calculation of bending moments and shear forces due to lifting are
a simple case of equilibrium of vertical loads. Vacuum lifters will spread
the load more and bending moments will be reduced because the length of
cantilevering segment outside the vacuum pads will be smaller compared to
using one or more lifting points.
This is a statically determinate load case, i.e. there is no need to consider
deflections or compatibility of strains. The self-weight is the same regard-
less of how the segment deforms under load. When we looked at a segmen-
tal lining being loaded by the ground and groundwater this was a statically
indeterminate load case and was treated slightly differently.
The self-weight of a segment as a function of its arc length, is given by:
w arc = tLsγ seg (10.1)
warc is the self-weight per unit arc length in kN/m

t is the segment thickness in m
Segment design for transient loads 325
Figure 10.1 Segment demoulding idealisation for design.
L s is the segment length in the tunnel longitudinal direction in m (the

dimension into the page in Figure 10.1)
γseg is the unit weight of the segment concrete in kN/m3 (usually
assumed to be 25 kN/m3 but a more exact value may be known
from testing)
If we simplify the problem to a straight beam, as shown in the middle dia-

gram of Figure 10.1, the self-weight may be expressed as:
Larc
wchord = w arc (10.2)
Lchord
wchord is the self-weight per unit length of the segment simplified as a

straight beam in kN/m
Larc is the arc length of the segment in m
Lchord is the chord length, i.e. the length of the segment idealised as a
straight beam, in m
The design value of self-weight is the self weight multiplied by the partial
factor for unfavourable permanent actions from EN 1992-1-1: 2004:
wd = wchord γ G (10.3)
wd is the design value of the self-weight for the segment idealised as a

straight beam, in kN/m
γ G is the partial factor for unfavourable permanent actions from
EN 1992-1-1: 2004, where γ G = 1.35
The ends of the segment are cantilevering as shown in the idealised struc-
tural model at the bottom of Figure 10.1. The shear force and bending
moment may be calculated as shown in Figure 10.2 with the following
notation:
Lcant is the cantilever length in m

Vmax is the maximum shear force in kN
M max is the maximum moment in kNm
The design value of shear force is given by:
Vmax = wd Lcant (10.4)
Figure 10.2 Shear force and bending moment diagrams for segment demoulding load
case.
The design value of bending moment is given by:
wd L2cant
Mmax = (10.5)
2
We can check the design moment and shear force against the capacity
of the segment if we know the compressive strength and flexural tensile
strength development at early age. It is not acceptable to assume that the
relationship between compressive strength and flexural tensile strength
will be the same at early age as it will be at 28 days (ACI 544.7R-16,
2016; fib Bulletin 83, 2017). Therefore, specific early age unconfined
compressive strength and EN 14651: 2005 flexural beam tests will be
required.
WORKED EXAMPLE 10.1

During preliminary design, the designer assumes a segment is to be
demoulded using a vacuum lifter that covers the central third of the
segment extrados. The geometry of the problem is similar to that
shown in Figure 10.1. There are six segments in the ring, so the seg-
ment arc is 60°. The length of the ring L s (in the tunnel longitudinal
direction) is 1 m, the thickness of the segment t is 200 mm and the
internal diameter of the ring is 3.0 m.
Calculate the maximum shear force and bending moment due to
demoulding.
The arc length along the centroid of the segment (mid-thickness) is
given by:
π (3 + 0.2)
Larc = = 1.676 m
6
Because the segment arc is 60°, the chord length in this case is one side
of an equilateral triangle with the other two sides being the radius to the
centroid. Therefore the chord length is:
Lchord = 1.6 m
The segment has a uniform thickness along the arc length and there-
fore assuming a unit weight of 25 kN/m3 the segment has a self weight
warc = 5.0 kN per metre of arc length, hence 5.0 kN/m.
If we simplify the problem to a straight beam, the self-weight per metre

of chord length wchord is given by:
Larc 1.676
wchord = w arc = × 5.0 = 5.236 kN/m
Lchord 1.6
The design value of self-weight wd is the self-weight multiplied by the

partial factor for permanent actions γG:
wd = wchord γ G = 5.236 × 1.35 = 7.069 kN/m
The vacuum lifter in this example covers the central third of the segment.
The ends of the segment are therefore cantilevering as shown in the ide-
alised structural model at the bottom of Figure 10.1. The shear force and
bending moment may be calculated as shown in Figure 10.2. The canti-
lever length Lcant = Lchord /3 = 1.6/3 = 0.533 m.
The design value of shear force is given by:
VEd = wd Lcant = 7.069 × 0.533 = 3.768 kN
The design value of bending moment is given by:
wd L2cant
MEd = = 1.005 kNm
2
For an SFRC segment, we can check the design moment and shear force
against the capacity of the segment if we know the compressive strength
and flexural tensile strength development at early age. The procedure will
be similar to that shown in the next section for segment stacking, which
is also a statically determinate load case with no axial force.
For a reinforced concrete or plain concrete segment, we can check
the design moment and shear force against the capacity of the segment
using EN 1992-1-1: 2004.
10.2 STORAGE
Segments need to be stacked to minimise space requirements at the factory

and site. Usually timber battens are used between segments to avoid dam-
age. A photograph of segments stacked on site is shown in Figure 10.3.
In this case there were six trapezoidal segments per ring with a universal
taper, so each stack is one ring. If the rings have a key, this will be at the top
of the stack. Sometimes with large diameter tunnels, rings are divided into
two or more stacks to reduce stacking loads and lifting weight.
Figure 10.3 Segments stacked on site.
10.2.1 Actions
The positioning of the timber battens should be specified by the designer on
the segment drawings. If they were positioned perfectly then there would
be no bending moments induced in the segments except for those induced
by self-weight of an individual segment, because loads from segments
above travel straight down through the battens, as shown in Figure 10.4.
However, when the battens are not aligned vertically, additional bending
moments and shear forces will be induced. Segment stacking is a statically
determinate load case.
In the case of Figure 10.4, bending moments and shear forces in the seg-
ments will depend on self-weight with two supports, as shown in Figure 10.5.
Similar to the demoulding load case, we have simplified each segment to a
straight beam.
The notation in Figure 10.5 is as follows:
Lchord is the chord length, i.e. the length of the segment idealised as a
straight beam, in m
Lb is the distance between the timber battens in m
L s is the segment length in the tunnel longitudinal direction in m
W is the weight of one segment in kN
Figure 10.4 Perfect segment stacking using timber battens.
Figure 10.5 Shear force and bending moment diagrams for perfect segment stacking.
w is the self-weight of one segment as a uniformly distributed load per

unit chord length, in kN/m
V1 to V6 are the shear forces per unit segment length in the tunnel longi-
tudinal direction, in kN, at the locations shown in Figure 10.5
M1 to M5 are the bending moments per unit segment length in the tunnel
longitudinal direction, in kNm, at the locations shown in Figure 10.5
The design shear force and moment will be the maximum values calculated
in Figure 10.5, multiplied by the partial safety factor for permanent unfa-
vourable actions, γG = 1.35.
The designer also needs to take account of tolerances in the placement
of battens. The worst case is if a pair of battens is placed as far apart as
the tolerances allow, with the battens of the segment above being as close
together as the tolerances allow (fib Bulletin 83, 2017). The tolerance is
usually taken to be 100 mm (fib Bulletin 83, 2017; ACI 544.7R-16, 2016),
so the adjacent battens could be up to 200 mm apart.
The moment due to eccentric batten placement in the bottom two seg-
ments can be added to the self-weight moment calculated above. Since
poor batten placement is an accidental action, a partial factor of 1.00
can be used for this action (fib Bulletin 83, 2017). Figure 10.6 shows
Figure 10.6 Forces due to eccentric placement of battens in segment stack.

Figure 10.7 Shear force and bending moment diagrams for eccentric batten placement
in a segment stack.
the forces due to eccentric batten placement on the fifth segment in a

segment stack.
The symbols in Figure 10.6 that are not already defined are:
ei is the batten eccentricity to the inside in m

eo is the batten eccentricity to the outside in m
The shear force and bending moment diagrams for eccentric batten place-
ment are shown in Figure 10.7. Note that because we are separating the
effect of eccentric batten placement from the self-weight, the reaction from
the battens beneath is 2W.
The procedure that has been described is the most basic method of
design. More sophisticated numerical modelling of the segment geometry
could be used, along with a more sophisticated constitutive model for
the concrete behaviour, whether it is SFRC, reinforced concrete or plain
concrete.
10.2.2 SFRC segment ultimate limit state design

for bending moment
In the previous section, we calculated the loads, so now we need to calcu-
late the capacity of the segment at the ultimate limit state. If the loads are
within the capacity, then the design is safe.
Ultimate limit state (ULS) and serviceability limit state (SLS) design to
the fib Model Code 2010 (2013) is based on the load-displacement curve
Figure 10.8 uses the following notation:
Pcr is the load at first crack

Pser is the load based on service actions calculated from a linear elastic
analysis with the assumption of uncracked concrete and initial elastic
Young’s modulus
Pu is the ultimate load, which is often driven by the maximum deforma-
tion requirement of the structure
Pmax is the maximum load
δser is the displacement at Pser
δpeak is the displacement at Pmax
δu is the displacement at Pu
Note that Pu and δu are not the ULS load and delection, but are the ‘ulti-
mate load’ and the ‘ultimate deflection’, respectively. The ULS load is the
ultimate load divided by a partial factor, and represents the resistance of
the structure.
Figure 10.8 Idealised load-displacement curve with definitions according to the fib

Model Code 2010 (2013) (based on a diagram in ITAtech, 2016).
Note also that Pser and δser are written as PSLS and δSLS in the fib Model Code
2010 (2013). However, this is misleading as they are not the SLS load and
deflection, but are the load and deflection due to service actions (as defined
in Section 7.7.2 of the fib Model Code 2010). The subscript ‘SLS’ is not in the
list of notation of the fib Model Code 2010, and appears to have been used in
error, so here we will use Pser and δser, which are in the list of notation.
If fibres are used as the only reinforcement or the minimum conventional
reinforcement to achieve ductility is not present, the fib Model Code 2010
(2013) Section 7.7.2 requires that one of the following conditions is satisfied:
1. δ u ≥ 20δ ser
2. δ peak ≥ 5δ ser
fib Bulletin 83 (2017) says that δu need only be greater than δser, rather
than at least 20 times greater as the Model Code requires. In this case, the
designer needs to ensure that the segment has sufficient ductility and rota-
tion capacity.
These conditions impose a requirement for minimum deformability in
statically indeterminate structures to allow bending moment redistribution
to occur. They do not apply to segment demoulding or stacking, which are
statically determinate situations. However, they will apply to ground and
water loads on the tunnel lining later in this chapter.
A further condition is that the ultimate load Pu shall be higher than both
the cracking load Pcr and the service load Pser. This is to avoid a brittle
failure after cracking. The load cases we have seen so far for demoulding
and stacking are ‘statically determinate’ situations. In other words, static
equilibrium equations are sufficient to determine the internal forces in the
segment. Therefore, the requirement that Pu ≥ Pcr can also be expressed in
terms of bending moments, such that Mu ≥ Mcr.
We also want to ensure that Pcr ≥ Pser (or Mcr ≥ Mser ) so that the segment is
not cracked at the service load, as recommended by fib Bulletin 83 (2017).
This is different to how we design the segmental lining once installed,
where we allow some cracking when the service loads are applied but try
to keep crack widths below 0.1–0.3 mm (depending on the specification) to
allow autogenous healing to occur.
First we need to calculate what bending moment the section can resist.
This may be done using a sectional analysis approach. To do this we need
to calculate the design values of resistance.
The segment concrete is specified by a grade, for example ‘FRC 40/50 – 5.0c’
(see Chapter 8 Section 8.2 for more details). This means fibre reinforced
concrete with a compressive cylinder strength of 40 N/mm 2 , a compres-
sive cube strength of 50 N/mm 2 , fR1k = 5.0 N/mm 2 and 0.9 ≤ fR3k /fR1k ≤ 1.1.
Therefore f R3k will be taken as equal to 0.9 × fR1k = 4.5 N/mm 2 .
The segment concrete will also be required to meet the requirement
fR1k /fLk ≥ 0.4 (Equation 5.6-2, fib Model Code 2010, 2013).
Figure 10.9 Rigid-plastic stress block approach for ultimate limit state design of a fibre-
reinforced concrete segment.
To check the ultimate limit state, a rigid-plastic stress block approach

may be used, as shown in Figure 10.9.
fcd is the design compressive strength in N/mm 2

f Ftud is the design value of the ultimate post-cracking tensile strength in
N/mm 2
xn is the distance from the compressive extreme fibre to the neutral axis in mm
λ defines the depth of the compressive stress block as a factor of xn
η is a factor that is used to reduce the design compressive strength at
high strength values
A s is the area of steel bar reinforcement (if present) in mm 2
f yd is the design yield strength of the steel bar reinforcement (if present)
in N/mm 2
We are only going to consider SFRC segments here, the bar reinforcement
in Figure 10.9 is only shown for completeness as it is possible to use both
fibre and bar reinforcement together and this is included in the fib Model
Code 2010 (2013) design methodology.
The moment that can be resisted at the ultimate limit state will depend
on the values of the design compressive strength fcd and the design value of
the ultimate post-cracking tensile strength f Ftud, as well as the distance to
the neutral axis xn.
If the characteristic cylinder compressive strength fck ≤ 50 MPa, then
λ = 0.8 and η = 1.0. If fck > 50 MPa then η and λ need to be determined
by equations 7.2–15 to 7.2–18 in the fib Model Code 2010 (2013). As an
example, for FRC 40/50 – 5.0c, fck = 40 MPa, so η = 1.0 and λ = 0.8.
The design compressive strength fcd is given by Section 3.1.6 of EN 1992-
1-1: 2004, and the value for FRC 40/50 – 5.0c concrete is calculated like
this for our numerical example:
α cc fck 0.85 × 40
fcd = = = 22.67 N/mm 2 (10.6)
γc 1.5
fcd is the design compressive strength in N/mm 2 .

αcc is a coefficient to take account of long-term loading and other
factors and is between 0.8 and 1.0, with its value given in the
National Annex to EN 1992-1-1. In the UK the value is 0.85
(UK NA to BS EN 1992-1-1: 2004).
fck is the characteristic value of the cylinder compressive strength in N/mm2.
γc is the partial material factor for concrete in compression, and is 1.5.
The design value of ultimate post-cracking tensile strength f Ftud is given by

the following equation:
fR3d
fFtud = (10.7)
3
f Ftud is the design value of the ultimate post-cracking tensile strength

in N/mm 2 .
f R3d is the design value of the residual tensile strength at CMOD = 2.5 mm.
The design value is the characteristic value f R3k divided by the material
partial factor, therefore fR3d = fR3k /γ F , where γF is the material partial
factor for fibre reinforced concrete in tension, usually taken as 1.5 (fib
Bulletin 83, 2017) but can be as low as 1.3 if this can be justified by
‘improved control procedures’ (Section 5.5.6, fib Model Code 2010,
2013). For example, Psomas & Eddie (2016) used γF = 1.35.
The following equation demonstrates a numerical calculation for the exam-

ple of FRC 40/50 – 5.0c concrete:
fR3d fR3k /γ F 4.5/1.5

fFtud = = = = 1.0 N/mm 2 (10.8)
3 3 3
From the values of f Ftud = 1.0 N/mm2 and fcd = 22.67 N/mm2 we can imme-
diately see from Figure 10.9 that, for our numerical example, without con-
ventional bar reinforcement the height of the compression stress block will
be 22.67 times smaller than the height of the tension stress block in order
for there to be horizontal force equilibrium. Therefore the neutral axis at the
ultimate limit state will be very close to the upper surface of the segment.
This can also be expressed by the equation for horizontal force
equilibrium:
λ xn fcd = ( h − xn ) fFtud (10.9)
Which can be rearranged such that:
hfFtud
xn = (10.10)
( λ fcd + fFtud )
Using Equation 10.10 we can calculate the size of the stress blocks, and
from that we can calculate the ULS resistance moment M Rd, which is the
design value of the ultimate moment. The ULS resistance moment may be
obtained by taking moments about the neutral axis of the design values of
compressive and tensile strength (c.f. Figure 10.9):
 h − xn   λx 
MRd = fFtud b ( h − xn )  + f bλ xn  xn − n  (10.11)
 2  cd  2 
M Rd is the ULS resistance moment in Nmm
The ULS resistance moment can be either positive or negative. Unless bar
reinforcement is present and not the same in the top and bottom layers,
moment capacity is always symmetrical in SFRC members if the section
geometry is symmetrical.
If the ULS resistance moment M Rd is greater than the design moment
M Ed, the segment passes the ultimate limit state check for this load case.
The design moment M Ed is calculated from the characteristic actions mul-
tiplied by a partial factor.
To ensure there is no brittle failure after cracking, we need to check that
Mu ≥ Mcr. This is because this is a statically determinate load case, and after
cracking the self-weight load cannot be redistributed. The fib Bulletin 83
(2017) states that unfactored mean strengths should be used for this check,
though this is not mentioned in the fib Model Code 2010 (2013). If design
values are used then we can check that M Rd ≥ Mcrd instead.
Mcr is the cracking bending moment in Nmm

Mcrd is the design value of cracking bending moment in Nmm
The design value of cracking bending moment Mcrd may be calculated as fol-
lows, assuming an elastic stress-strain relationship (fib Bulletin 83, 2017):
bh2
Mcrd = fctd , fl (10.12)
6
Mcrd is the design value of cracking bending moment in Nmm

b is the width in mm
h is the segment thickness in mm
fctd , fl is the design flexural tensile strength at first crack in N/mm 2
In the absence of experimental data, the design flexural tensile strength at

first crack fctd , fl can be calculated using the equations in EN 1992-1-1: 2004
for concrete, as follows.
First we get the characteristic mean value of axial tensile strength fctm
from Table 3.1 in EN 1992-1-1: 2004. For our example FRC 40/50 – 5.0c,
which has a characteristic compressive cylinder strength fck = 40 MPa,

fctm = 3.5 MPa.
Next we can calculate the mean flexural tensile strength using
Equation (3.23) from EN1992-1-1: 2004, as follows:
 h  
fctm, fl = max  1.6 −  fctm ; fctm  (10.13)
 1000  
fctm, fl is the mean flexural tensile strength at first crack in N/mm 2

fctm is the mean axial tensile strength in N/mm 2
A very similar value may be obtained using equations 5.1–3a, 5.1–8a and
5.1–8b in the fib Model Code 2010 (2013).
Equation 10.13 may also be used to obtain the characteristic flex-
ural tensile strength fctk, fl from a value of characteristic axial tensile
strength fctk.
Alternatively, if EN 14651: 2005 flexural beam test data exist (and this is
preferred for SFRC segment design), then the mean flexural tensile strength
fctm, fl may be taken as the characteristic mean value of flexural tensile
strength at the limit of proportionality f Lm:
fctm, fl = fLm (10.14)
f Lm is the characteristic mean value of flexural tensile strength at the

limit of proportionality in N/mm 2
Then the design value may be calculated by dividing the mean value by the
partial material factor for fibre-reinforced concrete in tension γF, usually
taken as 1.5.
fctm, fl
fctd , fl = (10.15)
γF
γF is the partial material factor for fibre-reinforced concrete
Thus we can now calculate Mcrd using Equation 10.12, and if M Rd ≥ Mcrd
then there will not be a brittle failure after cracking. If M Rd ≤ Mcrd then
the fib Bulletin 83 (2017) recommends applying an additional partial fac-
tor of 1.2 to reduce M Rd, to provide additional safety against a brittle
failure.
Note that if the segments are stacked before they have achieved their
specified 28 day strength, then a design check will need to be made for an
earlier age strength.
10.2.3 SFRC segment ULS design for shear force

The fib Model Code 2010 (2013) allows a simplified method of checking
for tensile-hardening FRC, that the design value of the ultimate residual
tensile strength fFtud ( = fFtuk /γ F ) must be greater than the principal tensile
stress in the segment. However, we have a complex situation where we have
both bending stresses and shear stresses. If a continuum numerical model
of the segment were used, then the outputs could be in the form of principal
stresses, and then this check would be straightforward.
Coccia et al. (2015) proposed a method for including the effect of shear
in the ULS check for moment and axial force, by reducing the design value
of the ultimate post-cracking tensile strength f Ftud as follows:
  τ 2 
fFtud
red
= fFtud 1 −    (10.16)
  fFtud  
fFtud
red
is the reduced value of f Ftud in N/mm 2
f Ftud is the design value of the ultimate post-cracking tensile strength in
N/mm 2 and is given by fFtud = fR3d /3 (see Equation 10.8)
τ is the shear stress, given by τ = V /bh, in N/mm 2
The moment capacity check described in the previous section should then
be repeated with the reduced value fFtud
red
in place of f Ftud.
Note that if the segments are stacked before they have achieved their
specified 28 day strength, then a design check will need to be made for an
earlier age strength.
10.2.4 SFRC segment SLS design

We must also check the SLS, because even though we have the opportunity to
discard segments that are damaged by storage, we do not want to throw away
lots of segments either. For transient situations such as demoulding, storage
and transportation it is common to impose a condition that no cracking should
occur (fib Bulletin 83, 2017). This means that the minimum moment at which
cracking occurs must be greater than the moment due to service actions, i.e.:
Mcr ,min ≥ Mser (10.17)
Mcr ,min is the minimum moment at which cracking may occur in kNm
M ser is the moment due to service actions in kNm
The minimum cracking moment Mcr ,min is calculated assuming an elastic

stress-strain distribution, as shown in Figure 10.10:
bh2
Mcr ,min = fctk, fl (10.18)
6
Figure 10.10 Elastic stress blocks for SLS design of a fibre-reinforced concrete segment.
b is the segment length in the tunnel longitudinal direction in mm

fctk, fl is the characteristic flexural tensile strength in N/mm 2
Note that for the SLS the partial material factor is 1.0.
The characteristic flexural tensile strength is given by Equation 10.13:
 h  
fctk, fl = max  1.6 −  fctk ; fctk  (10.19)
  1000 
fctk is the characteristic (5th percentile) axial tensile strength in N/mm 2
The characteristic axial tensile strength, in the absence of experimental data,

may be obtained from Table 3.1 of EN 1992-1-1: 2004, and is given by:
fctk = 0.7 fctm (10.20)
fctm is the characteristic mean value of axial tensile strength in N/mm 2 ,

which may be found in Table 3.1 of EN 1992-1-1: 2004
The moment due to service actions M ser is calculated in the same way as the
design moment M Ed, but with a partial factor on actions of 1.0.
WORKED EXAMPLE 10.2

We will use the same segment dimensions as in Worked Example 10.1,
i.e. the length of the ring (in the tunnel longitudinal direction) L s = 1 m,
the thickness of the segment t = 200 mm and the internal diameter of
the ring is 3.0 m. There are six segments in a ring and all six will form a
single stack. The unit weight of the segment concrete γseg = 25 kN/m3 and
consists of SFRC classified as FRC 40/50 – 5.0c.
The design drawings specify a batten placement at quarter points mea-

sured horizontally, with a maximum batten eccentricity ei = eo = 100 mm.
The lowest level of support battens between the 6th segment and the
ground are joined together and so cannot be out of tolerance. The geom-
etry is shown in Figure 10.11.
a. Calculate the design values of the maximum shear force and bend-
ing moment.
b. Check the capacity of the segment at the ULS.
c. Check for the effect of shear on the resistance moment at the ULS.
d. Check the capacity of the segment at the SLS.
Figure 10.11 Forces due to eccentric placement of battens in segment stack.
a. Calculate the design values of the maximum shear force and bend-
ing moment.
The arc length L arc of a segment is given by:
Larc = rmθ
r m is the radius to the lining centroid in m, which is 1.6 m

θ is the segment angle in radians, which is 2π /6 rad
Therefore:
2π
Larc = rmθ = 1.6 = 1.676 m
6
Now we will calculate the self weight of one segment W:
W = γ seg Ls Larc t = 25 × 1.0 × 1.676 × 0.2 = 8.38 kN
Since the segment arc is 60°, then the segment chord and two radii
make an equilateral triangle and the segment chord length is 1.6 m.
Therefore the battens should be placed at 0.4 m from the edges of
the segment measured horizontally.
Therefore, the worst case is shown for the 5th segment where the
battens between the 4th and 5th are as close together as the toler-
ance ei allows and the battens between the 5th and 6th segments are
as far apart as the tolerance eo allows. This will generate a sagging
moment in the central part of the segment’s span.
The tolerance assumed on batten placement in this example is
0.1 m for both inside and outside tolerances ei and eo. Therefore, we
will assume that the battens between the 4th and 5th segments are
0.5 m from the segment edges and the battens between the 5th and
6th segments are 0.3 m from the segment edges.
First we will calculate the shear forces and bending moments
due to self-weight assuming perfect batten placement, according to
Figure 10.5 (reproduced below as Figure 10.12) using Lchord = 1.6 m
and Lb = 0.8 m.
The forces are:
V1 = V6 = 0
 1.6 − 0.8 
V2 = −   × 7.069 = −2.828 kN
 2
0.8
V3 = × 7.069 = 2.828 kN
2
0.8
V4 = − × 7.069 = −2.828 kN
2
 0.8 − 1.6 
V5 = −   × 7.069 = 2.828 kN
 2
Figure 10.12 Shear force and bending moment diagrams for perfect segment stacking.
M1 = M5 = 0
M2 = M4 =
(1.6 − 0.8)
2
× 7.069 = 0.566 kNm
8
1.62 − 2 × 1.6 × 0.8
M3 = × 7.069 = 0 kNm
8
Next we can look at the eccentric batten placement. The shear force
and bending moment diagrams for this load case were shown in
Figure 10.7, reproduced here as Figure 10.13.
In Figure 10.13, the maximum shear force in the 5th segment
due to eccentric batten placement is equal to half the weight of the
segments above. The segment weight will have a partial factor of
1.0, because eccentric batten placement is considered an accidental
action and so the design value of the action is given by:
Wd = W × 1.0 = 8.38 × 1.0 = 8.38 kN
The maximum design value of shear force due to batten eccentricity

is therefore:
VEd ,ecc = ±2 × Wd = ±16.76 kN
Figure 10.13 Shear force and bending moment diagrams for eccentric batten
placement in a segment stack.
The maximum design value of bending moment due to batten

eccentricity is given by:
MEd ,ecc = −2 × Wd ( eo + ei ) = −2 × 8.38 × 0.2 = −3.352 kNm
The shear force and bending moment induced by eccentric

batten placement are quite a lot higher than those induced in a
perfect stack, and this illustrates the importance of considering
tolerances.
Adding together the self-weight value multiplied by a load factor
of 1.35 and the eccentric batten placement design value, the design
value of shear force is:
VEd = 16.76 + 2.828 = 19.59 kN

Superposing the self-weight and eccentric batten placement, the

worst combination in sagging will occur at mid-span, where the
self-weight with perfect batten placement moment is zero.
MEd ,sag = M3 + MEd ,ecc = −3.352 kNm
The worst combination in hogging will be over the supports due

to self-weight. We will ignore the sagging moment provided by the
battens at these locations as it is favourable.
MEd ,hog = M2 = 0.566 kNm
b. Check the capacity of the segment at the ultimate limit state.

This is an SFRC design calculation according to the fib Model
Code 2010 (2013).
Structural and serviceability design is based on the load-dis-
placement curve shown in Figure 10.8. The ultimate moment M u
shall be higher than the cracking moment M cr to avoid a brittle
failure after cracking. We also want to ensure that Mcr ≥ Mser so
that the segment is not cracked at the service load, as recom-
mended by fib Bulletin 83 (2017).
We need to calculate the ultimate resistance moment of the sec-
tion, M u. To do this we first need to calculate the design values of
resistance.
The segment concrete is FRC 40/50 – 5.0c. This means fibre rein-
forced concrete with a compressive cylinder strength of 40 N/mm2,
a compressive cube strength of 50 N/mm2, fR1k = 5.0 N/mm 2
and 0.9 ≤ fR3k /fR1k ≤ 1.1. Therefore f R3k will be taken as equal to
0.9 × fR1k = 4.5 N/mm 2.
The segment concrete will also be required to meet the require-
ment fR1k /fLk ≥ 0.4 (Equation 5.6-2, fib Model Code 2010, 2013).
To check the ultimate limit state, a rigid-plastic stress block
approach may be used, as shown in Figure 10.9.
The characteristic cylinder compressive strength fck ≤ 50 MPa, so
we will use λ = 0.8 and η = 1.0. The design compressive strength fcd
is given by Equation 10.6:
α cc fck 0.85 × 40
fcd = = = 22.67 N/mm 2
γc 1.5
Ultimate post-cracking tensile strength f Ftud is given by Equation 10.8:
fR3d fR3k /γ c 4.5/1.5

fFtud = = = = 1.0 N/mm 2
3 3 3
Considering horizontal force equilibrium of the stress blocks
(Equation 10.10), for h = 200 mm we get:
hfFtud 200 × 1.0

xn = = = 10.45 mm
( λ fcd + fFtud ) (0.8 × 22.67 + 1.0)
The resistance moment at the ultimate limit state M Rd is calculated
by taking moments about the neutral axis (Equation 10.11):
 h − xn   λ xn 
MRd = fFtud b ( h − xn )  + f bλ xn  xn − 
 2  cd  2 
 200 − 10.45 
= 1.0 × 1000 × ( 200 − 10.45) ×   + 22.67 × 1000
 2
 0.8 × 10.45 
× 0.8 × 10.45 ×  10.45 − 
 2
= 19.15 × 106 Nmm = 19.15 kNm
Since MRd ≥ MEd in both hogging and sagging, the segment passes
the ultimate limit state check for storage.
To ensure there is no brittle failure after cracking, we need to
check that M Rd ≥ Mcrd.
First we get the value of mean axial tensile strength fctm from
Table 3.1 in EN 1992-1-1: 2004. For a characteristic compressive
cylinder strength fck = 40 MPa, fctm = 3.5 MPa.
Then we calculate the mean flexural tensile strength using
Equation 10.13:
 h    200  
fctm, fl = max  1.6 −  fctm ; fctm  = max  1.6 −  × 3.5;3.5
 1000     1000  

= 4.9 N/mm 2
Then the design value may be calculated by dividing the mean value
by the partial material factor for fibre reinforced concrete in ten-
sion γF, usually taken as 1.5.
fctm, fl 4.9
fctd , fl = = = 3.27 MPa
γF 1.5
Thus the design value of the cracking bending moment is given by:
bh2 1000 × 2002

Mcrd = fctd , fl = × 3.27 = 21.78 kNm
6 6
Thus M Rd ≤ Mcrd and so there could be a brittle failure after crack-
ing. fib Bulletin 83 (2017) argues that in this situation an addi-
tional partial factor of 1.2 should be applied to reduce the risk of
cracking and brittle failure, and strongly recommends full-scale
bending tests on segments.
Applying the additional partial factor:
19.15
MRd = = 15.96 kNm
1.2
Since MRd ≥ MEd and Mcrd ≥ MEd in both hogging and sagging, the
segment still passes the ultimate limit state check for storage.
c. Check for the effect of shear on the resistance moment at the ULS.
Using Coccia et al. (2015)’s method of including the effect of
shear in the ULS check for moment and axial force, by reducing
the design value of the ultimate post-cracking tensile strength f Ftud.
The shear stress is given by:
VEd 19.59 × 103

τ= = = 0.098 N/mm 2
bh 1000 × 300
And the reduced value of f Ftud is given by Equation 10.16:
  τ 2   0.098  2 
fFtud
red
= fFtud 1 −    = 1.0 1 −    = 0.990 N/mm
2
  fFtud     1.0  
The ULS check in (c) and (d) should be repeated with the reduced
value fFtud
red
in place of f Ftud. We can see that in this case the effect will
be negligible.
d. Check the capacity of the segment at the serviceability limit state.
For transient situations such as demoulding, storage and transporta-
tion it is usual practice to impose a condition that no cracking should
occur.
The characteristic axial tensile strength, in the absence of exper-
imental data, may be obtained from Table 3.1 of EN 1992-1-1:
2004, and is given by:
fctk = 0.7 fctm = 0.7 × 3.5 = 2.45 N/mm 2

And the characteristic flexural tensile strength is given by

Equation 10.13:
 h    200  
fctk, fl = max  1.6 −  fctk ; fctk  = max  1.6 − 1000  × 2.45; 2.45
  1000   

= 3.43 N/mm 2
The minimum cracking moment Mcr ,min is given by:
bh2 1000 × 2002

Mcr ,min = fctk, fl = × 3.43 = 22.87 kNm
6 6
The moment due to service actions M ser is calculated in the same
way as the design moment M Ed, but with a partial factor on actions
of 1.0. Therefore the maximum values are:
1.0 1.0
Mser ,sag = M3 × + MEd ,ecc × = −3.352 kNm
1.35 1.0
1.0 1.0
Mser ,hog = M2 × = 0.566 × = 0.419 kNm
1.35 1.35
We need to check that:
Mser ≤ Mcr ,min
So for both the sagging and hogging service moments, the SLS
check is therefore passed.
10.3 TRANSPORTATION AND HANDLING
The segments must be moved from the factory to storage yard, then to site
on a railway wagon, barge or lorry, and then to the TBM on a multi-service
vehicle or rail wagon. There are also all the crane and telehandler lifts
between these modes of transport to consider. For each of these, the load-
ing on the segments or a stack of segments must be considered.
There are also situations where the segments need to be turned over.
Most segments are cast extrados-upwards, whereas it is usual to stack seg-
ments extrados-down. Sometimes segments are cured or stored standing on
their circumferential joint. All these manipulations and the loads they may
induce must be considered.
In most cases, the bending moments are important, but the axial and
shear forces are relatively small (fib Bulletin 83, 2017).
fib Bulletin 83 (2017) recommends that where dynamic loads are
expected, for example during road or rail transport, an additional partial
factor of γd = 2.0 should be applied to the self-weight (as well as the usual
partial factor of 1.35).
In all these cases, the design model will be of lifting points or areas and
cantilevering sections, with a design procedure similar to that for demould-
ing and storage.
If the segments are lifted using cast-in lifting pins, or by using a device
screwed into the grout plug, then these lifting devices need to be checked.
Also, pull-out of a cone of concrete or edge failure around the lifting device
also needs to be considered.
10.4 ERECTION
Within the TBM, segments may be lifted off the transport by a hoist or roll-
ers and moved through to the front. Then they will be lifted by the segment
erector, which could have lifting points or use vacuum lifting.
In all these cases, the design model will be of lifting points or areas
and cantilevering sections, with a design procedure similar to those for
demoulding, storage, transportation and handling.
10.5 INSTALLATION LOADS
TBM jacks shove the machine forward by pushing against the circumferen-
tial joint of the last ring built. These jacks are usually equally spaced around
the circumference and are often in pairs with a shared ram shoe that applies
the load to the circumferential joint. Although the aim is to spread the load as
evenly as possible, bending moments and other kinds of stress concentrations
may result from imperfections in ring build and segment geometry.
10.5.1 Temporary TBM jacking loads

Burgers et al. (2007) used 3D numerical modelling to investigate the effect
of jacking forces on SFRC segments for the Barcelona Line 9 tunnels. Each
ring consisted of seven segments plus a key segment, the tunnel was about
12 m diameter, and each segment had an arc length of 4.70 m. Each segment
was loaded by two pairs of thrust rams. The trailing edge of each segment
had four bearing pads to help transfer jacking loads to the previous ring.
As jacking loads were increased in the model, the first cracks were formed
between the loading surfaces (the ram shoes) due to spalling. At higher loads,
the model showed splitting cracks developing under the loading surfaces.
The steel fibres ensured that cracks propagated in a controlled manner and
increased load resulted in more cracks rather than widening of a single crack.
This numerical model ignored the effect of imperfections in the segments
or geometrical tolerances in ring assembly. In real life the ring joint may
Figure 10.14 Cracking induced by eccentricity of jacking load towards the extrados

turning segment outwards.
not be plane and the jacking forces may not be applied perfectly. Therefore,
Burgers et al. (2007) went on to analyse the effects of eccentric placement
of a thrust jack in the radial direction by applying a jacking force in the
model with a triangular pressure distribution across the segment thickness.
This caused a significant reduction in the failure load from 30.7 MN to
23.6 MN and cracking began much earlier. The (exaggerated) deformation
of a segment due to eccentricity of the jacking load towards the extrados
of the circumferential joint is shown in Figure 10.14. The eccentricity causes
the segment to bend outwards about a tangential axis, which concentrates
the bearing stress on the previous ring into the centre and induces bending
about a radial axis causing a longitudinal tensile crack on the leading edge
of the segment between the two pairs of jacks, due to the uneven support
on the trailing edge. Cracks exactly like this one were observed in real seg-
ments during the construction of Line 9.
Another possible cause of longitudinal cracks between the ram shoes is
uneven support due to steps in the circumferential joint plane (de Waal, 2000).
Cavalaro et al. (2011) say that this is the most common cause of this type
of cracking. An illustration of this is shown in Figure 10.15.
The effect of this uneven support depends on its magnitude, on the thick-
ness and material properties of the segments, and on the thickness and
material properties of the packing material. For a small step between seg-
ments in the previous ring, the new segment will deform until contact is
made across the gap, with only a small reduction in capacity. For a larger
step critical damage may occur before contact is made. Cavalaro et al.
(2011) found that thinner and longer rings could cope with larger steps.
Figure 10.15 Longitudinal cracks caused by uneven support.
Thinner rings have a lower stiffness, allowing more deformation to occur

to close the gap before critical tensile stresses are induced. Longer rings,
although stiffer, allow more load to be transferred to the first contact and
therefore result in lower tensile stresses between the ram shoes. A lower
packer stiffness was found to help too. Heijmans & Jansen (1999) looked
at this load case for the design of the Pannerdensch Kanaal Tunnel and
found that the gap needed to be less than a few tenths of a millimetre to
avoid cracking.
Bilotta & Russo (2012) installed strain gauges in segments used for the
Line 1 tunnels in Naples. Each ring consisted of six segments plus a key.
Four independent rings were instrumented with a large number of strain
gauges. In each ring, five vibrating wire strain gauges were cast into each
segment, all on the circumferential centreline, i.e. equidistant from both
circumferential joints. A pair of strain gauges, one near the intrados and
one near the extrados, was installed near to each radial joint of every seg-
ment (except the key) to measure hoop strain, and one was installed in the
centre of each segment to measure longitudinal strain. They found that
the jacking forces caused a high longitudinal compression, as one would
expect, but also an extension in the circumferential direction, which they
attributed to a Poisson’s ratio effect and the lack of significant restraint
to the ring within the tailskin. This circumferential extension was later
reversed by grouting and ground load.
By using a back-calculation method to convert the strains into stresses,
Bilotta & Russo (2012) also calculated internal forces around the lining.
A further calculation, which they did not explain in the paper, enabled
Figure 10.16 Hoop forces around a segmental lining back-calculated from strain gauges
(from Bilotta & Russo, 2012).
the forces across the radial joints to be estimated. The assembly of the lin-
ing, jacking forces and grout pressures resulted in a highly variable hoop
axial force and bending moment distribution, particularly if the longitudinal
strains were taken into account in the calculation, as shown in Figure 10.16.
Hoop axial forces were lower at joints than in the body of the segment.
Bilotta & Russo explained that the segments were placed quite smoothly, but
didn’t explain where the hoop force was going. It is possible that the hoop
force could have been transferred via dowels and friction in the circumferen-
tial joints to the adjacent ring’s segments rather than across the joints.
If the results of Bilotta & Russo’s back-calculation are taken at face value,
the most interesting aspect is how much real hoop thrusts vary within seg-
ments and around the ring compared to conventional calculations using
analytical solutions. The effects of small geometric misalignments, asperi-
ties, variable and/or eccentric jacking forces and grout pressures in the real
situation clearly have a large impact on the forces in the segments in the
medium- to long-term.
We need to be able to design a segmental lining for jacking forces, when
small geometrical misalignments can cause such large changes in stress.
These effects can only really be modelled using 3D numerical analysis, with
careful attention paid to modelling the expected geometrical tolerances and
the contact between segments. It is also wise to consider full-scale testing of
segments in a laboratory at the tolerance limits.
Usual practice in design is to make assumptions, usually in consultation
with the TBM manufacturer and the contractor, about the worst case dis-
tribution of jacking forces around the ring and their potential eccentricities.
Groeneweg (2007) recommends a partial factor of 2.0 on jacking loads
for comparison with the ultimate limit state. Alternatively, fib Bulletin 83
(2017) suggests using the maximum capacity of the rams, based on the
maximum pressure of the hydraulic system.
Since we do not want to cause excessive cracking of the segments, fib
Bulletin 83 (2017) recommends applying a partial factor of 1.5 on the nom-
inal working load of the thrust rams for comparison with the SLS.
The following load cases should be considered:
• local crushing of the concrete under the ram shoes

• bursting stress within the segment beneath the ram shoes, in both the
radial and circumferential directions (see Section 9.4), also consider-
ing potential eccentricity of the ram shoe
• shear failure of the circumferential joint due to ram load
• spalling near the surface of the circumferential joint between the ram
shoes
• geometric imperfections, such as steps in the circumferential joint
plane
10.5.2 Longitudinal effects of TBM jacking loads

Hoefsloot (2009) used an analytical model and also strain gauge measure-
ments from the Groene Hart Tunnel to show that jacking forces remain
in the tunnel lining permanently. At a distance of about 40 m behind
the TBM, average longitudinal axial force in the lining had decreased to
approximately 70% of the jacking load, and this stayed in the lining perma-
nently. Talmon et al. (2009), in an accompanying paper, calculated that the
residual axial force acted at 1.5m above the axis level of the Groene Hart
Tunnel, meaning a significant residual longitudinal bending moment is also
left in the tunnel permanently.
Figure 10.17, taken from Hoefsloot (2009), shows the longitudinal bend-
ing moments (about the horizontal axis of the tunnel) generated by jacking
loads. These bending moments are caused by the fact that jacking forces
tend to be higher at the invert to counteract the tendency for TBMs to dive,
perhaps due to the weight of the cutterhead and higher friction around
the bottom part of the shield. The rings within the TBM are also to some
extent cantilevering off the rings further back where the grout has hard-
ened (Hoefsloot & Verweij, 2006). With distance behind the TBM, these
bending moments are gradually reversed by grout pressures and the weight
of the backup equipment.
Predicting these longitudinal bending moments is complex, but can be
done analytically or numerically, as shown by Hoefsloot (2009), Hoefsloot &
Verweij (2006) and Talmon et al. (2009). They depend mainly on the dis-
tribution of jacking forces, restraint provided by the shield/tailseal brushes,
the properties of the grout and the position and weight of the backup.
Hoefsloot (2009) also found that the bending moment distribution in
Figure 10.17 means that the last ring built nearly always has overhang,
Figure 10.17 Strain gauge data and analytical model of the Groene Hart Tunnel (from
Hoefsloot, 2009).
i.e. the crown is slightly further inbye than the invert. The new ring has
to be placed to the design alignment, hence there will always be a change
in plane of the rings and perhaps stepping. This can cause eccentric loads
across the circumferential joint and increases the risk of cracking. Another
problem is that the TBM itself often has ‘lookup’ (i.e. the axis of the TBM
is pointing slightly upwards) while the rings in the tailskin have overhang,
which results in wear to the tailseal brushes or excessive forces applied to
the segments at the crown and invert, similar to what happens on a tight
curve but rotated 90°.
The effect of ringbuilding tolerances, stiffness of tailseal brushes and
TBM attitude were investigated in detail by Mo & Chen (2008) using a 3D
numerical model. They found that the key segment was the most vulner-
able to dislocation or overstressing and this meant that the TBM diving
downwards was the worst case (assuming the key is positioned at or near
the crown, which is not always the case).
As anyone who has seen a bent pipe will know, tubes flatten in bending,
and so longitudinal bending of a tunnel causes it to squat, as shown by
Huang et al. (2012).
Koyama (2003) used pressure cells on the extrados of segments to
measure the pressure applied by the tailseal brushes and found that it
could be twice as high as the pressure later exerted by the ground and
that tight curves could result in permanent loads and deformations in
the lining.
10.5.3 Grout pressures and tailseal brushes

Initially, when grout is pumped into the tail void using grout ports through
the tailskin, the principal direction of flow is circumferential (Talmon et al.,
2006). Further back from the injection point, the velocity is slower and the
grout flow is longitudinal. By modelling the flow using a Bingham model
for the grout based on laboratory tests, Talmon et al. found that the grout
pressures exerted on the tunnel lining depended on the grout ports used.
When all six grout ports were used uniformly (Figure 10.18), the pressure
distribution in the model increased linearly with depth, whereas when only
the top three grout ports were used, the downwards flow made the pressure
distribution approximately uniform (Figure 10.19). Talmon et al. (2009)
later showed that the ‘centre of gravity’ of the grout injection relative to
the axis of the tunnel could be used to calculate the uplift force for situ-
ations between these two limiting cases. The type of grout used in these
cases, as is common practice in the Netherlands, was an inert grout. Where
semi-active or active grouts are used (i.e. grouts that set within seconds or
minutes of injection), the patterns of behaviour may be different.
The linear increase of grout pressure with depth in Figure 10.18 is not
hydrostatic, but increases at a lower rate than the unit weight of the grout.
This is due to buoyancy of the tunnel lining, which applies an uplift force
on the grout. By taking this into account, Talmon et al. (2006) found good
agreement with field measurements of grout pressure.
As fresh grout exerts an uplift force on the tunnel lining, this is resisted
by the rings within the TBM, and by the rings further back where the grout
has already hardened (Hoefsloot & Verweij, 2006). It is also resisted by the
self-weight of the rings and the weight of the TBM backup and depends
on the properties of the grout. The effect is obviously worse if inert or
Figure 10.18 Calculated grout pressures at 0 m and 4.1 m from the rear of the TBM using
all six injection ports uniformly (from Talmon et al., 2006).
Figure 10.19 Calculated grout pressures at 0 m and 4.1 m from the rear of the TBM using
the top three injection ports only (from Talmon et al., 2006).
slow-setting grouts are used. The tunnel is therefore like a pipe supported
at both ends with an upwards pressure in the middle, which induces lon-
gitudinal bending moments in the tunnel, exacerbating the deformations
due to jacking loads. A basic conclusion from this Dutch research is that by
injecting more grout through the upper injection ports, uplift forces on the
lining, and hence longitudinal bending moments, can be reduced.
10.5.4 Design for permanent effects

of installation loads
It is clear from Figure 10.16, as well as the work of Bakker & Bezuijen
(2009), Bilotta & Russo (2012), Blom et al. (1999) and Molins & Arnau
(2011), that the installation of segments, the TBM jacking forces and the
injection of grout have a permanent effect on the medium- and long-term
loads in the tunnel lining. The use of simple 2D analytical solutions or 2D
bedded beam models will not predict these loads.
As well as the segmental lining material properties, geometry and tol-
erances, detailed knowledge of the TBM is required in order to properly
design a segmental tunnel lining and this may include:
• precise geometry of the TBM jacks, their shoes, the tailskin and the
tailseal brushes
• the location of grout ports, the grout rheology and hardening param-
eters, and grouting strategy (i.e. how much to inject through each port)
• likely range of advance rates
• segment manufacturing tolerances
• ringbuilding tolerances
• jacking forces and their potential eccentricity
This information is not always available at the design stage, so conservative

assumptions may need to be made.
Current practice usually considers TBM jacking forces as a transient load
only, i.e. in isolation and only for a short duration. The maximum jacking
force, the ram shoe geometry and the maximum eccentricity of jacking force
are assumed, and segments are checked for crushing, bursting stresses, shear
and bending, in a similar way to how the radial joints are designed. Designing
for tensile stresses between ram shoes (c.f. Figure 10.14 and Figure 10.15) can
only be done using 3D numerical analysis.
10.6 GASKET COMPRESSION AND BOLT LOADS
Gaskets consist of strips of extruded rubber that are placed all around each
segment on the radial and circumferential joints. Therefore, each joint will
have two gaskets in contact. Their purpose is to prevent water ingress to the
tunnel. There are two ways that watertightness may be achieved. One is by
compression; when a segment is pushed against another segment, the gaskets
are compressed until the concrete faces of the segments or their packers make
contact and are compressed. Compression of the gaskets prevents water from
pushing its way in between them. The other way is by using a compound that
swells in contact with water, known as a ‘hydrophilic’ gasket. Hydrophilic
gaskets for precast concrete segmental linings usually consist of a compres-
sion gasket with a superficial layer or bead of hydrophilic rubber.
Gaskets are either glued into a preformed gasket groove (Figure 10.20), or
cast-in to the segment by attaching them inside the moulds (Figure 10.21).
Gaskets may be needed to keep groundwater and grout from leaking
into the tunnel, or to keep a fluid that is in the tunnel from leaking out.
Figure 10.20 Glued-in segment gasket (based on VIP028 type gasket manufactured by

VIP-Polymers Ltd, 2020).
Figure 10.21 Cast-in segment gasket (based on VIP028CI type gasket manufactured by

VIP-Polymers Ltd, 2020).
Sometimes, they need to achieve both, for instance in a potable water tun-
nel below the groundwater table with fluctuating internal water pressure.
We do not want potable water to leak into the ground, because this is
wasteful, and we do not want untreated groundwater entering the tunnel.
In this case we also have the requirement that the gasket materials are safe
to use in contact with drinking water. In the case of a sewage tunnel, we
do not want sewage leaking into the ground, but we may be less concerned
with groundwater infiltration into the tunnel, except that pumping and
treatment costs may be increased because of the additional flow. The gas-
kets will need to be chemically resistant to the groundwater and to any fluid
within the tunnel. It is not possible to replace segment gaskets, so they are
required to last for the full design life of the tunnel.
The applied load for segment design is the force required to compress the
gasket. Laboratory tests can determine what this closing force should be.
An example is shown in Figure 10.22. When the gap reaches zero, the con-
crete faces of the segments are in contact and very little further deformation
is possible. Therefore, this can be taken as the design load. The maximum
gasket compression force is applied to the gasket groove and the concrete
can be checked for bearing capacity, bursting stress and, since the gasket
is usually close to the edge of the segment, shearing off of the corner of the
segment.
Usually the bolts are designed to maintain gasket compression until the
ring is grouted using a simple tensile capacity calculation, with the load
calculated from the required gasket closure force.
Radial bolts may be curved, such that they have a pocket on both sides
of the joint, or they are ‘spear bolts’, which consist of a straight bolt that is
screwed into a plastic cast-in socket.
For some tunnels, it may be necessary to design the bolts and connectors
to hold the weight of the segment, if the TBM jacks are to be removed for
Figure 10.22 An example of gasket compression force versus joint closure (VIP-
Polymers Ltd, 2020).
some reason. In this case, the bolts and connectors will have to resist ten-
sion and shear forces, and sometimes bending.
It is common nowadays to use plastic dowels across the circumferential
joints and steel bolts across the radial joints. The plastic dowels are also
helpful in aligning the new segments with the previous ring.
In recent years, designers have begun to question the need for radial bolts,
because the TBM jacks and the confinement provided by the tailskin brushes
should be sufficient to keep the segments in position and the joint faces in con-
tact until the ring is grouted. On some projects bolts or dowels across the radial
joints have been omitted and guide rods have been used to aid positioning.
Concrete around bolt pockets or connector sockets may need to be
checked for local bearing stresses, because there is a possibility of local
crushing or shear failure. For SFRC segments, more detailed guidance is
available in fib Bulletin 83 (2017).
10.7 SUMMARY
Segments undergo various load cases during their life, including demould-
ing, storage, transportation, erection and then installation loads from the
TBM jacks and grouting pressures. The most common load cases were
described, with worked examples used to demonstrate the ultimate limit
state and SLS calculations.
When a segment is loaded by its self-weight or external forces, such that

the internal forces may be calculated from consideration of force equilib-
rium, then this is ‘structurally determinate’ behaviour. When a ring of
segments is in its final position and loaded by the ground and groundwa-
ter, then this is ‘structurally indeterminate’ because as the ring deforms,
stresses will redistribute. The requirements for SFRC design are different
for structurally determinate or indeterminate load cases.
segmental tunnel linings, but a teaching aid to demonstrate the principles
and most important aspects of design. Every project is unique and has dif-
ferent load cases and performance requirements, some of which may not be
included here.
10.8 PROBLEMS
Q10.1. A 6.5 m ID tunnel lining consists of six ordinary segments, two

top segments and a key. The segments are 325 mm thick, 1.2 m
long, and are made from steel fibre reinforced concrete of grade
FRC 40/50 – 5.0d and of unit weight 25 kN/m3.
i. The segments are to be stacked in the yard in a single stack for
each ring, using timber battens between each segment. The six
ordinary segments are at the bottom of each stack and the key
is at the top. The battens are to be placed at ¼ of the segment
chord length from each end, with a maximum eccentricity of
100 mm, except for the ground level battens, which are joined
together. Calculate the bending moments and shear forces in
the second segment up from the bottom, assuming the worst
possible eccentricities of batten placement. If the stack does
not work, try a smaller stack.
ii. Check the capacity of the segment at the ULS for these loads
according to the fib Model Code 2010 (2013), taking account
of shear using the method of Coccia et al. (2015).
iii. Check the capacity of the segment at the SLS.
iv. Each ring is transported to site on a lorry in three stacks of
three segments. For a stack of three ordinary segments, cal-
culate the design values of bending moments and shear forces
using the same method as in part (i), and applying the addi-
tional partial factor for dynamic loads recommended by fib
Bulletin 83 (2017).
Q10.2. An 8.5 m ID tunnel lining consists of ten trapezoidal segments.
The segments are 400 mm thick, 2 m long, and are made from
steel fibre reinforced concrete of grade FRC 50/60 – 5.0d and of
unit weight 25 kN/m3.
i. The segments are to be erected in the TBM using a vacuum
lifter that covers the central 2/3 of the segment intrados.
Calculate the bending moments and shear forces when the

segment is being lifted vertically.
ii. Check the capacity of the segment at the ULS for these loads
according to the fib Model Code 2010 (2013), taking account
of shear using the method of Coccia et al. (2015).
iii. Check the capacity of the segment at the SLS.
iv. [Advanced] What is the worst case orientation for the segment
erector to be holding the segment?
REFERENCES
ACI 544.7R-16 (2016). Report on Design and Construction of Fiber-Reinforced

Precast Concrete Tunnel Segments, Reported by ACI Committee 544.
Farmington Hills: American Concrete Institute.
Bakker, K. J. & Bezuijen, A. (2009). Ten years of bored tunnels in The Netherlands:
Part II, structural issues. Geotechnical aspects of underground construction
in soft ground (eds Ng, C. W. W., Huang, H. W. & Liu, G. B.), pp. 249–254.
Bilotta, E. & Russo, G. (2012). Backcalculation of internal forces in the segmental
lining of a tunnel: The experience of Line 1 in Naples. Geotechnical aspects
of underground construction in soft ground (ed. Viggiani, G.), pp. 213–221.
Blom, C. B. M., van der Horst, E. J. & Jovanovic, P. S. (1999). Three-dimensional
structural analyses of the shield-driven green heart tunnel of the high-speed
line south. Tunn. Undergr. Space Technol. 14, No. 2, 217–224.
Burgers, R., Walraven, J., Plizzari, G. A. & Tiberti, G. (2007). Structural behavior
of SFRC tunnel segments during TBM operations. Underground space – the
4th dimension of metropolises (eds Barták, J., Hrdina, I., Romancov, G. &
Zlámal, J.), pp. 1461–1467. London: Taylor & Francis Group.
Cavalaro, S. H. P., Blom, C. B. M., Walraven, J. C. & Aguado, A. (2011). Structural
analysis of contact deficiencies in segmented lining. Tunn. Undergr. Space
Technol. 26, 734–749.
Coccia, S., Meda, A. & Rinaldi, Z. (2015). On shear verification according to fib
Model Code 2010 in FRC elements without traditional reinforcement. Struct.
Concret. 16, No. 4, 518–523.
EN 14651: 2005 (2005). Test method for metallic fibered concrete - Measuring the
flexural tensile strength (limit of proportionality (LOP), residual). Brussels:
EN 1992-1-1: 2004 (2004). Eurocode 2: Design of concrete structures – Part 1-1:
Standardization.
fib Bulletin 83 (2017). Precast Tunnel Segments in Fibre-Reinforced Concrete,
State-of-the-art Report, fib WP 1.4.1, October 2017. Lausanne: Fédération
internationale du béton (fib).
fib Model Code 2010 (2013). fib Model Code for Concrete Structures 2010.
Groeneweg, T. W. (2007). Shield driven tunnels in ultra high strength concrete –

Reduction of the tunnel lining thickness. Ph.D. Thesis, Technical University
of Delft.
Heijmans, R. W. M. G. & Jansen, J. A. G. (1999). Design features of the
Pannerdensch Kanaal Tunnel in the Betuweroute. Tunn. Undergr. Space
Technol. 14, No. 2, 151–160.
Hoefsloot, F. J. M. (2009). Analytical solution of longitudinal behaviour of tun-
nel lining. Geotechnical aspects of underground construction in soft ground
(eds Ng, C. W. W., Huang, H. W. & Liu, G. B.), pp. 775–780. London: Taylor &
Francis Group.
Hoefsloot, F. J. M. & Verweij, A. (2006). 4D grouting pressure model PLAXIS.
Geotechnical aspects of underground construction in soft ground (eds
Bakker, K. J., Bezuijen, A., Broere, W. & Kwast, E. A.), pp. 529–541. London:
Huang, X., Huang, H. & Zhang, J. (2012). Flattening of jointed shield-driven tun-
nel induced by longitudinal differential settlements. Tunn. Undergr. Space
Technol. 31, 20–32.
ITAtech (2016). Design Guidance for Precast Fibre Reinforced Concrete
Segments – Vol.1: Design Aspects, ITAtech Report No.7. Lausanne: ITA-
AITES. Reproduction of figures with permission of the Committee on
new technologies of the International Tunnelling and Underground Space
Association – ITAtech.
Koyama, Y. (2003). Present status and technology of shield tunneling method in
Japan. Tunn. Undergr. Space Technol. 18, 145–159.
Mo, H. H. & Chen, J. S. (2008). Study on inner force and dislocation of segments
caused by shield machine attitude. Tunn. Undergr. Space Technol. 23, 281–291.
Molins, C. & Arnau, O. (2011). Experimental and analytical study of the struc-
tural response of segmental tunnel linings based on an in situ loading test.
Part 1: Test configuration and execution. Tunn. Undergr. Space Technol. 26,
764–777.
Psomas, S. & Eddie, C. M. (2016). SFRC Segmental Lining Design for a Pressurised
Tunnel. Proc. WTC2016, San Francisco, USA, Paper 0118, p. 13. Englewood,
Colorado, USA: SME.
Talmon, A. M., Aanen, L., Bezuijen, A. & van der Zon, W. H. (2006). Grout pres-
sures around a tunnel lining. Geotechnical aspects of underground construc-
tion in soft ground (eds Bakker, K. J., Bezuijen, A., Broere, W. & Kwast, E. A.),
pp. 77–82. London: Taylor & Francis Group.
Talmon, A. M., Bezuijen, A. & Hoefsloot, F. J. M. (2009). Longitudinal tube bend-
ing due to grout pressures. Geotechnical aspects of underground construction
in soft ground (eds Ng, C. W. W., Huang, H. W. & Liu, G. B.), pp. 357–362.
UK NA to EN 1992-1-1: 2004. UK National Annex to Eurocode 2: Design of
concrete structures – part 1-1: general rules and rules for buildings. London:
VIP-Polymers Ltd (2020). Tunnel Segment Gaskets. Brochure available at: https://
www.vip-polymers.com/wp-content/uploads/2020/01/VIP-TSG-Brochure-
2017-Lo.pdf (last accessed 3rd November 2020).
Chapter 11
Sprayed concrete lining design
Sprayed concrete linings differ from the segmental linings covered in the
previous two chapters in the following ways:
• Precast or prefabricated linings already have their long-term design

strength and stiffness at the time of installation, while sprayed con-
crete gains strength and stiffness while also being increasingly loaded
by the ground as the tunnel advances.
• Sprayed concrete quality cannot be verified prior to installation as it
depends on workmanship, variability of materials and environmental
conditions.
• Unlike segmental linings, there are no intentional articulating joints
in a sprayed concrete lining. There may be construction joints, but
usually every effort is made to ensure these behave as monolithically
as possible.
• Sprayed concrete linings do not have to be circular.
Sprayed concrete allows much more flexibility in terms of geometry than

other lining types. However, there are still constraints on the shapes we can
use. In soft ground we still need the lining to act as an arch in order to have
an efficient structural form, but the profile does not have to be circular.
Depending on the space envelope required inside the tunnel, the profile can
be modified to reduce the volume of excavation.
There is a trade-off between volume of excavation, lining thickness and
reinforcement. The less circular the profile of a sprayed concrete lining is,
the lower the volume of excavation will be, but the bending moments will
be larger, and the lining will need to either be thicker or contain more rein-
forcement. If we make the lining thicker, it will be stiffer, and will therefore
attract higher bending moments, possibly requiring an even thicker lining.
Therefore, it is really important to find the right balance between minimis-
ing volume of excavation and minimising lining materials to get the most
sustainable solution.
This chapter is not intended to be a comprehensive manual for design
of sprayed concrete linings, but a teaching aid to demonstrate the most
DOI: 10.1201/9780429470387-11 363
important principles. For sprayed concrete, even more than for other con-
struction methods, the design and construction are inextricably interlinked.
The design is not finished when construction starts, but must be verified by
monitoring and back-analysis. For brevity, this book focuses on design and
includes very little discussion of construction, but make no mistake: you
should find out as much as you can about construction before embarking
on the design of a real sprayed concrete tunnel. Designers who do not take
the trouble to understand construction are inefficient and dangerous, and
this could equally be said of site engineers who do not understand design.
Diverse teams of people with construction and design backgrounds, who
are willing to learn from each other, produce the best and safest solutions.
• the principles of design for sprayed concrete tunnel linings

• the importance of collaboration between design and construction teams
• the complex factors that will affect the real stresses and strains in a
sprayed concrete lining
• how basic toolbox measures work
After working through this chapter you will be able to:
• optimise the profile and face division of a sprayed concrete lining to

fit a required space envelope
• design toolbox measures to improve stability and reduce ground
movements
• design a sprayed concrete lining using 2D or 3D numerical analysis
and a simple ageing elastic or elastoplastic constitutive model
11.1 PRIMARY AND SECONDARY LININGS
The traditional approach to design of a sprayed concrete lining was to

assume that the primary lining is ‘temporary works’ only and does not
contribute any load-carrying capacity in the long term. A waterproof mem-
brane is then installed followed by a cast-in-place (CIP) reinforced concrete
secondary lining. This was because of concerns about the long-term dura-
bility of sprayed concrete, but was also due to the traditional contractual
separation of temporary and permanent works. The secondary lining was
designed for the full long-term ground and groundwater pressures, which
were assumed to act on the outside of the waterproof membrane.
As quality control of sprayed concrete improved and confidence about
its long-term durability increased, it became common practice to assume
the primary lining had at least some long-term load-bearing capacity (ITA-
AITES, 2020). This can make the overall design of the lining system more
efficient. It is also possible to spray the secondary lining, which in some
situations may be more cost-effective than a CIP lining.
Sprayed concrete lining design 365
There are many variants, which depend on the design philosophy regard-
ing water ingress, durability and the final finish at the intrados of the lining
system. These will be described in the following sub-sections. The main
types are summarised in Table 11.1.
11.1.1 Single-pass lining

If the primary lining is durable, has sufficient capacity for the long-term
loads, is considered sufficiently watertight for the ground conditions and
intended usage of the tunnel, and an adequate internal finish can be pro-
vided, then no secondary lining is required.
11.1.2 Single-shell lining

In soft ground, a primary lining that is designed for the temporary
loads during construction is often sufficient, or almost sufficient, for the
Table 11.1 Sprayed concrete lining system types.

Waterproofing
Lining type system Primary lining Secondary lining
Single-pass No waterproof Permanent, takes all None
lining membrane long-term ground and
groundwater loads
Single-shell Sprayed Permanent, takes all For protection of the
lining waterproof long-term ground and membrane and to take
membrane groundwater loads fixings only
Traditional Sheet Temporary loads only CIP, designed for full
approach waterproof long-term ground and
membrane groundwater loads
Traditional No waterproof Temporary loads only CIP or sprayed,
approach, no membrane designed for full
membrane long-term ground and
groundwater loads
Double-shell Sheet Permanent CIP, designed for a share
lining – waterproof of long-term ground
permanent membrane pressures and 100% of
primary groundwater loads
Composite shell No waterproof Permanent CIP or sprayed,
lining – no membrane designed for a share of
membrane long-term ground and
groundwater loads
utilising partial or full
composite action
Composite shell Sprayed Permanent CIP or sprayed,
lining – waterproof designed for a share of
sprayed membrane long-term ground and
membrane groundwater loads
utilising partial or full
composite action
long-term loads as well. If a sprayed waterproof membrane is used to pro-

vide enhanced watertightness, and is bonded to the primary lining, then in
theory as long as the bond strength is higher than the expected long-term
groundwater pressure, no secondary lining is needed, except to protect the
membrane from damage and to provide a substrate for internal fixings.
Therefore, there is little or no additional cost to designing the primary
lining for the long-term loads, and providing a 100–150 mm thick sec-
ondary lining that is not designed for any ground or groundwater loads.
The secondary lining is there only to protect the waterproof membrane
(whether sprayed or sheet) and to allow penetration of fixings for things
like cladding, services or equipment without puncturing the membrane. In
this case, the design loads on the secondary lining are only its self-weight
and equipment or cladding loads on the fixings. The secondary lining may
also be designed as a fire protection layer.
11.1.3 Traditional approach

The traditional method of design of a primary and secondary lining system
is to assume that the primary lining is only temporary, and the secondary
lining takes all the loads in the long term. In a numerical model, the primary
lining is there for the short-term construction period, then for the long-term
analysis it is switched off and the secondary lining is switched on.
For this case, it does not matter whether there is a waterproof mem-
brane between the primary and secondary linings, or not. The full long-
term ground and groundwater loads will be conservatively applied to the
secondary lining in the design.
11.1.4 Double-shell lining system

When the primary lining is considered permanent and there is also a sec-
ondary lining inside it, the simplest assumption to make is that there is
no shear interaction between the primary and secondary linings. This is
referred to as a ‘double-shell lining system’. This is a valid assumption when
a sheet waterproof membrane is used between the two linings, but may not
be realistic when a bonded sprayed membrane is used.
If the primary lining is considered permanent and we wish to take advan-
tage of its ability to share the load in the long-term, then we need to know
about the ultimate limit state. In other words, what is the ultimate resis-
tance of the double-shell lining system? Will the primary lining fail before
load can be shared with the secondary lining? This seems unlikely, but it
all comes down to compatibility of strains. All the strains – elastic, plastic,
creep, shrinkage and thermal, and how they evolve over time – must be
considered to find the answer to this question.
As we saw in Chapter 1, the primary lining will experience elastic and
plastic strains due to ground and groundwater loads, as well as creep,
shrinkage and thermal strains, early in its life. When the secondary lining is
installed, it also experiences shrinkage and thermal strains. The secondary
lining contracts and moves away from the primary lining, and this means
that a significant increment of load is needed before the secondary lining
will even begin to share the load applied to the primary lining. Jones (2018)
estimated for a case study that this load increment would need to be at least
10 MPa, and therefore it is unrealistic to assume any load-sharing for most
tunnels in soft ground. Exceptions to this may include when another tunnel
is excavated nearby or where a junction is broken out through the primary
lining after the secondary lining has been installed.
The groundwater pressure may be assumed to be acting on the second-
ary lining, particularly if a sheet waterproof membrane is present between
the primary and secondary linings. Some bending moment will be induced
by the hydrostatic increase in groundwater pressure with depth around the
membrane, and if the secondary lining is non-circular.
If there is the potential for chemical attack from the ground, it is some-
times assumed that in the long term a certain thickness of the primary
lining is degraded, and is not considered in the long-term analysis. This is
called a ‘sacrificial layer’. In these ways, even with a permanent primary
lining, the secondary lining can be designed for some load.
11.1.5 Composite shell lining system

In some cases, the designer may consider the primary and secondary lining
to be sufficiently watertight so that no waterproof membrane is required.
Then the secondary lining may be either CIP or sprayed, and should have
good bond and interlock with the primary lining, thus providing interface
shear and some composite action.
If a sprayed waterproof membrane is used between the primary and sec-
ondary lining and it is bonded to both of them, then to some degree the two
linings may act together as a composite.
The importance of composite action to flexural capacity may be illustrated
by considering an elastic beam, as shown in Figure 11.1. If the interface is
perfectly smooth and frictionless, then there will be no composite action,
Figure 11.1 Varying degrees of composite action of a primary and secondary lining

assuming elastic behaviour.
and the primary and secondary lining act independently. If the interface has
a perfect bond, then there will be full composite action, where the primary
and secondary lining act as though they were one monolithic lining. If there
is some bond and/or friction between the primary and secondary linings,
providing an interface shear stiffness, then some degree of partial composite
action will occur, which could be anywhere between the two limiting cases
of no composite action or full composite action (Jäger, 2016).
Figure 11.1 shows, and it can be demonstrated by the calculation that
follows, that the same bending moment will induce higher bending stresses
in the system with no composite action. Therefore, partial or full composite
action will increase the moment capacity.
Assuming the primary and secondary linings have the same thickness t,
the moment with no composite action may be given by:
Iσ max 4t 3σ max t 2σ max

M = 2× = = (11.1)
y 12t 3
M is the bending moment in Nmm

σmax is the maximum stress at the extreme fibre in N/mm 2
t is the thickness of each lining in mm
I is the second moment of area in mm4, which per metre length is t 3 /12
y is the distance from the neutral axis to the extreme fibre, which is
equal to t /2
For full composite action, the primary and secondary linings act like a
monolithic beam of thickness 2t, and thus the moment is given by:
Iσ max 8t 3σ max 2t 2σ max

M= = = (11.2)
y 12t 3
I is the second moment of area in mm4, which per metre length is

(2t )3 /12
y is the distance from the neutral axis to the extreme fibre, which is
equal to t
Comparing Equation 11.1 and Equation 11.2, if the moment were the same,
then the value of σmax when there is full composite action would be half the
magnitude of the value for no composite action, as shown in Figure 11.1. If
the maximum stress σmax were the same in both cases, i.e. if it represented
the strength of the lining material, then the resistance moment would be
twice as large when there is full composite action.
Jäger (2016) noted that composite action means a higher overall stiffness
(compare the values of I in the two cases above) and therefore the bending
moments obtained from a soil-structure interaction analysis will be higher
as a result.
Su et al. (2019) showed that if fully composite behaviour is assumed,

where there is a hogging moment (tension at the intrados), the secondary
lining will have a lower bending moment compared to a double-shell lining
system, but it will also have a lower hoop force. When the moment – axial
force interaction was plotted, they found that the secondary lining was out-
side the capacity curve when full composite action was applied, but inside it
when there was no composite action. Therefore, fully composite behaviour
does not necessarily result in a thinner overall lining system.
The simple model in Figure 11.1 assumes the two linings are elastic and
have no pre-existing strains. However, in reality they are not elastic and
there are plastic, creep, shrinkage and thermal strains. These non-elastic
strains will affect the primary lining first, before the secondary lining is
installed. Shrinkage and thermal strains in the secondary lining will be
restrained by bond with the primary lining, and this may lead to signifi-
cant tensile hoop strains in the secondary lining. Alternatively, the second-
ary lining may debond from the primary lining if the bond strength is not
high enough. Su et al. (2019) included only shrinkage and self-weight of a
secondary lining in a numerical model and found that the fully composite
lining developed significant tensile stresses and induced compression in the
primary lining, whereas a double-shell lining system did not. Jäger (2016)
considered pre-loading of the primary lining and shrinkage of the second-
ary lining and showed that the primary lining failed before the axial load
capacity of the secondary lining could be reached.
The ability of sprayed membranes to bond to the primary lining means
that the water pressure can be assumed to be applied to the primary lining,
and as long as the water pressure is not so high that this bond strength is
exceeded, this can be beneficial because in most cases the primary lining
will have sufficient capacity for the long-term loads. It is therefore possible,
and this approach has been used on recent projects, to design the secondary
lining only to support its self-weight and to confine and protect the sprayed
membrane. Nasekhian & Feiersinger (2017) proposed a minimum second-
ary lining thickness of 150 mm to allow bolt penetrations for services and
equipment.
11.1.6 Initial layer

It is common practice in soft ground to spray an initial or ‘sealing’ layer
of shotcrete onto the exposed ground immediately after excavation.
This is typically 50–75 mm thick and covers the exposed ground of the
tunnel profile and the face. Because it is thin, it can be applied relatively
quickly. The purpose of the initial layer is to prevent low cohesion soils
from drying out or ravelling, or to prevent local block failures in clay. It
cannot do much to prevent global instability, though the face is typically
domed to provide some structural arching effect in the ground and the
initial layer.
One advantage of applying an initial layer to the face is that if there is

significant movement indicating the beginning of a failure, cracks in the
initial layer will be easily seen. Without the initial layer, the early signs of
face collapse may go unnoticed.
The initial layer at the excavation profile is usually included as part of
the primary lining for structural analysis purposes, but the initial layer on
the face is usually ignored in 3D numerical models (e.g. Jones et al., 2008).
11.1.7 Regulating layer

When a sprayed membrane is used, it is usual practice to spray a ‘regu-
lating layer’ over the primary lining first. The regulating layer consists of
a sprayed concrete without fibre reinforcement and without larger aggre-
gates greater than 4 mm, and is typically 30–50 mm thick. This creates a
smoother surface for the sprayed membrane and means that less volume of
the sprayed membrane product is needed to achieve the coverage required.
It also covers any fibres that are protruding from the primary lining, which
may be difficult to spray the membrane around without shadowing.
11.1.8 Finishing layer

If the sprayed concrete forms the internal surface of the tunnel (i.e. there
is no secondary lining, or the secondary lining is sprayed concrete), then
a finishing layer may be applied. This is a sprayed concrete without large
aggregates or fibres, and with a reduced accelerator dosage.
Immediately after spraying, the surface is trowel-finished by hand to give
a smooth internal surface. This requires a lot of manual labour, but may
be cost-effective for non-public areas such as equipment rooms, emergency
intervention/escape passages or ventilation tunnels that do not require clad-
ding, but need a smooth internal surface that can be washed down. A typi-
cal sprayed concrete lining without a finishing layer will be quite bumpy
and will collect dust, and if it has steel fibre reinforcement, protruding
fibres can cause injuries to people who accidentally lean against or brush
past the lining.
Concrete tunnel linings often have microfilament polypropylene fibres
added to prevent explosive spalling during a fire. In some cases, these fibres
are added only to the finishing layer if it is of sufficient thickness to also act
as a fire protection layer.
11.2 DESIGNING THE PROFILE
Most sprayed concrete linings have a vertical line of symmetry through the
centreline. It is possible to design non-symmetrical linings, but for most
purposes, symmetry provides the right fit.
Usually we assume that the upper part of the lining (the ‘crown’) above
axis level is a single circular arc. This is because in the temporary case dur-
ing construction it tends to take more load and therefore needs the stron-
gest shape. This has practical benefits in highway or railway tunnels, where
above the kinematic envelope of the vehicles some space is often needed
for ventilation, lighting or overhead line electrification. For pedestrian pas-
sageway tunnels, sometimes the required space envelope is chamfered in
the top corners, making a semi-circular crown more efficient.
In many tunnels for vehicle or pedestrian use, the finished floor is hori-
zontal, and although there is some underfloor space needed for drainage
and cable ducts, much of the invert would be wasted space if the tunnel
were circular. Therefore, the invert may be flattened to reduce the vol-
ume of excavation. A flatter invert is also easier to excavate using mobile
excavators.
11.2.1 Three arc profile

The most commonly-used non-circular shape is formed by four circular
arcs with three different radii for the crown, bench (two arcs, one each
side) and invert, and we will call this a ‘three arc profile’ because there are
three types of arc, i.e. three different radii are used. An example is shown
in Figure 11.2 for a pedestrian passageway tunnel. To avoid attracting large
bending moments at the points where the circular arcs meet, we need to
ensure they have common tangents, and therefore no vertices. A common
tangent is achieved by ensuring the centres of adjacent arcs are on the same
radial line. For example, where the crown arc meets the bench arc, the cen-
tre C B of the bench arc with radius R B is on a radius RC of the crown arc,
Figure 11.2 Three arc sprayed concrete lining profile.

which has its centre at CC. Similarly, C B is also on an invert radius R I that
has its centre at C I.
It is important to understand the origin of the three arc profile.
Originally, lattice girders or steel ribs were used between two layers of
welded wire mesh within a sprayed concrete lining, and if set out cor-
rectly they enabled the sprayer to spray to the correct profile. Each lat-
tice girder or steel rib could be easily fabricated to a single radius, with
bolted joints at the interface between construction stages, i.e. between the
top heading, bench and invert. Therefore, it was convenient if the transi-
tions between arcs were at the levels of the construction stage interfaces.
Nowadays, in most soft ground tunnels, steel fibre reinforced concrete
has replaced lattice girders or steel ribs and welded wire mesh to provide
the necessary flexural strength and ductility, and reflectorless laser total
stations are used during spraying to get the correct profile. Therefore,
there is no real need to define the profile using three arc radii, or for the
arc transitions to be at the construction stage interfaces. However, using a
three arc profile does keep things simple and provides a reasonable shape
in most situations.
In Figure 11.2, the crown arc is semi-circular, the centre of the invert arc
is at the apex of the crown, and the transitions from the bench arc to the
invert arc occur at the vertices of the required rectangular space envelope.
These constraints enable a reasonable shape to be found algebraically to fit
the rectangular envelope, but none of them have to be fixed. For instance,
some designers prefer to make the crown arc less than 180°, particularly
for wide or tall caverns. Similarly, the positions of the bench and invert arc
centres can vary, as long as the arcs have common tangents at the points
where they meet to avoid vertices.
The profile shown in Figure 11.2 would be the intrados of the sprayed
concrete lining. In order to draw the extrados, simply offset the curves by
the thickness of the lining. If drawing by hand, use the same centres, but
increase the radii by the thickness of the lining. The common tangents of
the extrados arcs will be on the same radial lines as the common tangents
of the intrados arcs.
11.2.2 Two arc profiles

Another simple non-circular shape is the two arc profile. This can be useful
when the space envelope required is particularly wide or tall, as shown in
Figure 11.3. This profile is also quite common for noncircular shotcrete-
lined shafts.
The crown arc with centre at CC has the same radius RCI as the invert
arc with centre at C I. The side arcs have the same radius R S and centres at
C S1 for the left side and C S2 for the right side. The profile has two lines of
symmetry, as shown by the dash-dot lines in Figure 11.3. The profile can be
Figure 11.3 Two arc sprayed concrete lining profile.
made more circular by moving the centres closer to the profile centre (where
the two lines of symmetry cross).
11.2.3 Other profile shapes

There are many possible variants using a number of circular arcs joined at
common tangents. For instance, the invert could be flattened in the wider
profile in Figure 11.3 by continuing the side arcs further down and having
a larger invert radius, making it a three-arc profile (but unlike Figure 11.2 it
would not have a semi-circular crown arc). Four arc profiles are also possi-
ble, and could be used to flatten the invert of the taller profile in Figure 11.3
by introducing short radii arcs between the side arcs and the invert arc.
Ellipse profiles have been used in the past, but they are actually quite
mathematically complex to use for setting out, since the radius is continu-
ously changing around the perimeter.
11.2.4 Optimising the profile

With the exception of the need to avoid vertices by having common tan-
gents and avoid large differences in arc radii to minimise bending moments,
it is difficult to come up with generic rules or constraints for optimisation.
The optimal shape depends not just on the required space envelope, but
also on the lining forces. Usually a 2D bedded beam model or a 2D or
3D numerical model is used and the profile is refined iteratively. A more
efficient profile in terms of excavation volume often means larger differ-
ences between the arc radii and hence larger bending moments, which may
require a thicker lining or more reinforcement. A good rule of thumb to

use as a first estimate before optimisation is that the bench and invert radii
should not be more than double, or less than half, the crown radius.
The lining profile will determine the deformation mode, and hence the
bending moments. Assuming the ground pressure is uniform, it is possible
to sketch the radial displacements and the bending moments that will be
induced by them. This can help us to understand how best to optimise
the profile and can be a useful check on the validity of numerical models.
Where there is a short radius, the lining will behave in a stiffer manner than
where there is a long radius. Accordingly, where there is a flattened invert,
the midspan of the invert will deflect more towards the centre of the tunnel
than at the bench arcs where there is a shorter radius.
Figure 11.4 shows a sketch of the expected radial displacement pattern
for the three arc profile from Figure 11.2. Around the semi-circular crown
we expect the radial displacement to be uniform, at the benches the shorter
radius has a stiffer response and so radial displacement is reduced, and at
the invert the longer radius behaves more flexibly and so radial displace-
ment is increased.
This ignores plenty of complex effects, such as the construction sequence,
shear between the lining and the ground, and anisotropy of in situ stress
and ground stiffness, but it approximates well to the pattern we see in field
data and numerical models.
Figure 11.5 shows a sketch of the bending moments that might be induced
by the pattern of radial displacements in Figure 11.4. If the bending moments
calculated in a numerical model are too high in the invert, then we could
reduce the invert radius. If the bending moments are too high at the benches,
then we could increase the bench radius. The closer we make the profile to a
perfect circle, the lower the bending moments should be.
Figure 11.4 Sketch of radial displacement pattern for a three arc profile.

Figure 11.5 Sketch of bending moment pattern for a three arc profile.
WORKED EXAMPLE 11.1

A rectangular required space envelope 3.65 m high and 3.2 m wide is
required for a metro station passageway.
a. Using a spreadsheet or iterative hand-calculation find a best fit

three arc profile assuming the crown arc is semi-circular, the invert
arc centre is at the crown and the transition from bench to invert
occurs at the vertex of the rectangle.
b. Using a CAD program or drawing by hand using a compass, create
a four arc profile by adding a long radius arc between the crown
and bench arcs to see if the excavation volume can be reduced.
Some trial and error may be required.
c. Sketch the radial displacement and bending moment distributions
you would expect to see for this four arc profile, assuming it is
wished-in-place in a homogeneous isotropic soil with K0 = 1.0.
Comment on how the construction sequence might influence the
radial deformations and bending moments and what changes could
be made if this were a problem.
a. Three arc profile

The geometry of a three arc profile is shown in Figure 11.6. The
required space envelope has height H = 3.65 m and width B = 3.2 m.
We need to find the position of the crown centre CC below the cen-
treline of the required space envelope, denoted A in Figure 11.6,
then we can find the rest of the geometry.
Figure 11.6 Worked Example 11.1 three arc geometry.

Using Pythagoras’s theorem for RC:
2 2
H   B
RC =  + A +   (11.3)
 2   2
And again for R I:

2 2
 H    B
RI =  RC +  − A  +   (11.4)
  2   2
Let the distance between the centres CB and CC be denoted G. It is

given by:
RC B
G = RC − RB = ×
H  2
RC +  − A (11.5)
 2 
And
RB = RC − G (11.6)
The bench radius R B may also be expressed in terms of the invert

radius R I:
H
−A
RB = 2 × RI (11.7)
H 
RC +  − A
 2 
By putting these equations into a spreadsheet, we can vary A
until Equation 11.6 and Equation 11.7 give the same value of R B.
Alternatively, keep trying different values of A in a hand calcula-
tion. This is the optimal geometry and occurs when A = 0.293 m
and R B = 1.640 m.
From this value of A we can use Equation 11.3 and Equation 11.4
to calculate the other two radii.
2 2 2 2
H   B  3.65   3.2 
RC =  + A +   =  + 0.293 +  = 2.655 m
 2   2  2   2 
2 2
 H    B
RI =  RC +  − A  +  
  2   2
2 2
  3.65    3.2 
= 2.655 +  − 0.293  +  = 4.482 m
  2    2 
b. Four arc profile.

Using a CAD program and some trial and error, the four arc pro-
file shown in Figure 11.7 was obtained. It is not perfect, and does
not quite touch the required space envelope, but it is close. The three
arc profile is also shown for comparison. The four arc profile has a
smaller cross-sectional area and hence a reduced excavation volume.
c. Sketch of radial displacement and bending moment.
A sketch of the radial displacement (left side) and bending moment
(right side) is shown in Figure 11.8.
Note that there is very little bending moment in the crown,
because the ground load is assumed to be radial and uniform. If
a top heading–invert or top heading–bench–invert construction
sequence were used (these are described in the next section), then
we would expect bending moments to be induced in the crown arc
Figure 11.7 Worked Example 11.1 four arc profile geometry.
Figure 11.8 Worked Example 11.1 four arc profile sketch of radial displacement
and bending moment.
because the bottom parts of the arc near axis level will deform
inwards. This effect could be reduced by a temporary invert acting
as a strut across the base of the top heading.
Note also that if the coefficient of earth pressure at rest K0 ≠ 1
then the radial displacements will also be anisotropic even for a
perfect circle and bending moments will be induced, as was dis-
cussed in Chapter 6 for the Curtis-Muir Wood analytical solution.
11.3 DIVIDING THE FACE
For smaller tunnels, it may be possible to advance with a full-face excava-

tion. For tunnels larger than approximately 5–6 m, it is normal to divide
the face. This may be done to improve face stability, but is also dictated
by the size and type of plant used to excavate the tunnel. Even in very
good ground where face stability is not an issue, the reach of the excavator
will limit the height of the top heading. At the detailed design stage it is
very important to discuss face division with the contractor. For preliminary
design purposes, at the very least it is necessary to consult someone who
has construction expertise.
Where there are cohesionless soils exposed during excavation, par-
ticularly if they are very dry or saturated, toolbox measures will almost
certainly be needed to avoid uncontrolled loss of ground. Very dry sand
or saturated sand will both be unstable even when only small areas are
exposed, whereas moist sand may stand up in small exposures for a short
time. The sizes of partial headings need to have enough space to allow the
planned toolbox measures to be undertaken. For this it is necessary to con-
sult with the contractor or a construction expert.
As well as considering the limitations of construction plant, the maxi-
mum allowable size of a heading may be determined using an undrained
or drained stability calculation, according to the methods described in
Chapters 2 and 3. Minimising ground movements may also be important,
and if so, then the methods described in Chapter 12 should be used.
The size of a sprayed concrete tunnel, since the diameter is not constant,
may be expressed as the maximum width of the excavation profile. Typical
minimum and maximum sizes of tunnel for each face division method are
shown in Table 11.2.
It would be very difficult to construct a sprayed concrete tunnel smaller
than the minimum size of 3 m, because of the typical sizes of small excava-
tion and spraying plant. Once a 3 m tunnel is lined, it may be more like 2.5 m.
Even if excavation and spraying plant could access the space, it would be
impossible for them to pass each other and one would need to back all the
way out before the other could enter, and drive lengths would be limited.
Table 11.2 Sprayed concrete tunnel face division methods and typical sizes.
Typical minimum
Face division method Shorthand and maximum size Notes
Full-face 3–6 m
Top heading–invert TH-I 4–7 m
Top heading–bench–invert TH-B-I 5–10 m
Sidewall drift and 9–15 m Min. limited by width
enlargement* of plant
Twin sidewall drift* 14–18 m Min. limited by width
of plant
Pilot tunnel then 8–11 m
enlargement by top
heading–bench–invert
* Note that sidewall drifts and enlargements are usually advanced with a TH-B-I method, so that a
twin sidewall drift may be divided into nine partial faces.
Smaller tunnels are possible with hand excavation and/or hand spraying,
but below 3 m timber support may be preferable to sprayed concrete. In
these situations the design and construction team should consider whether
the tunnel should be made larger to allow mechanisation, and hence safer
and faster tunnelling.
The following sections discuss each of these face division methods in
more detail.
11.3.1 Full-face excavation

Full-face excavation means that an advance length of the whole face is exca-
vated then covered by a full ring of sprayed concrete lining, as shown in
Figure 11.9. In reality, the upper part of the face is usually excavated fol-
lowed by an initial layer of 50–75 mm sprayed concrete on the perimeter
and the face, then the invert is excavated followed by the full primary lining
sprayed from the invert up to the crown. Note also that the face is usually
domed, as shown in Figure 11.9.
Figure 11.9 Full-face excavation. An advance has just been excavated and is ready to be
sprayed.
Advance lengths are typically 1 m, but can be up to 1.5 m if plant, logis-

tics, stability and ground movements allow.
Full-face excavation can achieve advance rates up to 8 m/day. It is limited
in size by the maximum face area that will remain stable and by the reach
of construction plant. A big advantage is that there are no radial joints in
the lining.
Full-face excavation is, in some respects, the simplest to analyse in design.
For 3D models, there is only one advancing face and no longitudinal stag-
gering of the construction stages. For 2D models, we only have to make one
assumption about the proportion of confinement loss that occurs before
installation of the lining, whereas with more complex partial face excava-
tion methods we have to make several assumptions about the proportion of
confinement loss for each stage. Modelling will be discussed in more detail
later in this chapter.
Lining stresses measured by pressure cells installed in the Storm
Water Outfall Tunnel (SWOT) frontshunt tunnel at Heathrow Terminal
5, UK (Jones et al., 2008) showed that a full-face excavation can result
in high stresses at early age. This is because the ring was closed immedi-
ately, and so when the peak temperature of shotcrete hydration occurred
10–15 hours later, the thermal expansion induced a compressive stress
higher than the long-term stress in the lining. For this tunnel, the peak
stress was fortunately within the capacity of the lining, but for larger
full-face tunnels, or at higher advance rates, this effect could cause
overstressing.
The face can be inclined to improve safety by providing a canopy
of hardened shotcrete above someone standing near the face. This
is usually an inclination of about 20° from the vertical, as shown in
Figure 11.10.
11.3.2 Top heading–invert (TH-I) excavation

An example of a TH-I excavation is shown in Figure 11.11. In this example,
the excavator has just finished a second top heading advance and the pri-
mary lining is about to be sprayed. Then a double advance length of invert
will be excavated and sprayed. Then the invert needs to be re-covered with
Figure 11.10 Full-face excavation with an inclined face.

Figure 11.11 Top heading–invert (TH-I) excavation.
a ramp of muck so that the excavator can reach the next top heading. This
covering and uncovering of the invert is inefficient, and that is why a double
advance length of the invert is usually excavated and sprayed after every
two top heading advances.
There are many possible variations on this sequence. For instance, the
invert could be staggered, so that it always lags behind the top heading by
one or two advance lengths. Although this increases the ring closure dis-
tance and hence should increase ground movements, it means that the joint
between the top heading and the invert shotcrete can be cleaned and pre-
pared more safely because the operatives are under a canopy of hardened
shotcrete, rather than recently sprayed ‘green’ shotcrete.
The most recently sprayed top headings are overhanging the last com-
pleted invert. As illustrated by the case studies described in Chapter 1, the
overhanging top heading experiences some loading from the ground, which
will tend to make it move downwards. This downwards displacement is
arrested by the completion of the invert, from which point the lining begins
to behave like a ring in compression and further ground movements should
be relatively small. Sometimes the top heading is given ‘elephant’s feet’, a
thickening of the lining at the footings of the top heading, which has the
aim of spreading the lining load and hence increases the bearing capacity.
Another option is to give the top heading a ‘temporary invert’, which is a
shallow arch of shotcrete (Figure 11.12). This is removed when the invert
excavation advances.
Figure 11.12 TH-I excavation, with ‘temporary invert’ in the base of the top heading.
Similar to the full-face excavation method shown in Figure 11.10, the top
heading could be inclined. However, it is not common practice to incline
the invert excavation.
11.3.3 Top heading–bench–invert (TH-B-I)

excavation
An example of a TH-B-I excavation is shown in Figure 11.13. Stage 1 has
been excavated and sprayed, then 2, 3, 4 and 5 have each in turn been
excavated and sprayed, completing a cycle. The first stage of the next cycle
(Stage 6 in Figure 11.13) has just been excavated and is about to be sprayed.
This will be followed by a bench excavation, another top heading, another
bench, and then a double invert. These are not shown on the diagram but
follow the same pattern as Stages 2-5.
There are many variations of the TH-B-I sequence. Sometimes the
advance lengths themselves are staggered by a half advance length so that
the circumferential joints of the top heading, bench and invert do not
occur at the same chainage. Sometimes two top headings are followed by
two benches, rather than alternating. Sometimes the ring is closed earlier
by completing the invert closer to the face. The optimum method will be
the one that balances the competing demands of controlling stability and
ground movements with quality, productivity and safety.
Unlike in rock tunnels where the bench and invert excavations may be
hundreds of metres behind the top heading, in soft ground the invert is usu-
ally closed within approximately 5 m of the face. This is because the lining
is not very effective at supporting the ground until the ring is closed, at
which point continuing deformations should be small. Each time the invert
is excavated and sprayed, the backfill then has to be replaced on top of
the invert shotcrete to allow the excavator to advance and attain sufficient
height to excavate the next top heading advance. This adds time to each
cycle, so invert advances are usually twice as long as top heading and bench
advances so they only need to be done half as often.
Similar to the full-face excavation method shown in Figure 11.10, the top
heading and/or the bench advances could be inclined.
Figure 11.13 Top heading–bench–invert (TH-B-I) excavation.

Figure 11.14 Single sidewall drift and enlargement excavation.
11.3.4 Single sidewall drift and enlargement

An example of a single sidewall drift and enlargement is shown in
Figure 11.14. First the sidewall drift (the lenticular shape on the left of
the cross-section in Figure 11.14) is excavated, in a TH-B-I sequence.
This means there is a completed ring in the temporary case, limiting
ground movements, and meaning that the enlargement can follow some
distance behind the sidewall drift. The enlargement is also excavated in
a top heading – bench – invert sequence. The central wall is not removed
until the full ring is completed and has gained sufficient strength. Access
is not usually allowed to the sidewall drift while enlargement works are
ongoing.
The single sidewall drift method requires joints at the crown and invert
to be formed, which need to be designed carefully to ensure they have
sufficient rotation capacity. This is usually achieved using reinforcement
bars, and so miners need access to the joint to place these, which involves
working at height, where the risk of falls of ground or green shotcrete
must be managed. Alternatively, an inner shotcrete lining may be sprayed
after the sidewall is removed to ensure structural continuity without
requiring access.
The main advantage of sidewall drifts is that the width and area of the
open excavation are reduced, thereby improving stability in very wide cav-
erns. It is possible to construct large caverns with a TH-B-I sequence if a
pilot tunnel is constructed first, or toolbox measures are used to improve
stability, such as jet grouting, face dowels and pipe arches. In fact, this is
common practice in Italy and is known as the ‘Adeco-RS system’ (Lunardi,
2008).
11.3.5 Twin sidewall drift and enlargement

An example of a twin sidewall drift and enlargement is shown in Figure 11.15.
The principles are the same as for a single sidewall drift and enlargement. In
Figure 11.15 Twin sidewall drift and enlargement excavation.
this case the two sidewall drifts are constructed and then the enlargement, or
‘core excavation’ is done last.
11.3.6 Pilot tunnels

Pilot tunnels are another form of face division, and improve stability by
reducing the area of exposed ground during excavation. During excavation
of the enlargement, the remaining pilot tunnel will act as a large dowel in
the ground ahead of the face, potentially reducing ground movements and
improving stability. However, this does mean that the ground is disturbed
twice by an advancing tunnel, and incremental demolition of the pilot tunnel
can cause a lot of vibration, which may have a negative impact on stability.
A pilot tunnel is like a very large horizontal borehole, and can be used to
learn more about the ground conditions to be encountered along the route
of the tunnel. It can also be used for grouting or dewatering of the ground
ahead of the enlargement.
There may also be logistical advantages to constructing pilot tunnels.
Being full-face and of a relatively small area, they can be advanced much
more quickly than a larger tunnel. This may provide early access to con-
struct another part of the works that is critical to complete early in the
programme.
For metro systems where the running tunnels are to be constructed
by TBM and the stations by sprayed concrete, the TBMs may run right
through creating the pilots for the platform tunnels. This depends on the
programme, but has been shown to be safer and quicker than shotcrete
pilot tunnels and, if a closed-face TBM is used, the overall ground move-
ments should be smaller (St John et al., 2015). It is also easier to separate
concrete segments from the spoil than to separate fragments of sprayed
concrete. A potential disadvantage is that the enlargement cannot begin
Figure 11.16 Sprayed concrete tunnel enlarged from (a) a sprayed concrete pilot, and
(b) a TBM segmental lining pilot.
until the TBM has reached its final destination and all the backup systems
have been removed from the tunnel.
An example from the design of Crossrail Farringdon Station platform
tunnels is illustrated in Figure 11.16, which shows a comparison of two
pilot tunnels, one a sprayed concrete pilot and the other a TBM-driven pilot
tunnel. Note that because the TBM pilot was larger the enlargement also
had to be larger to ensure sufficient space for toolbox measures, such as
spiling, grouting and depressurisation of sand channels, and for excavating
around and breaking out the segmental lining. Breaking out is much easier
if the length of the segmental lining rings is the same as the advance length
of the sprayed concrete enlargement, and for Farringdon special shorter
rings were used through the station section of the TBM drives for this pur-
pose (St John et al., 2015).
11.3.7 Binocular caverns

A question is often asked by those planning underground infrastructure
projects about how close parallel tunnels can be constructed to each other.
The answer is that tunnels can be constructed as close as you like, as long
as the lining is designed for the higher forces and bending moments that
will result. At the extreme, a pair of tunnels may have no ground at all
between them, in which case they become essentially a wide cavern with a
wall down the middle.
There are two main ways in which such binocular caverns may be built.
One method is to drive a pilot tunnel and then to construct the strong cen-
tral wall within it, usually structural steel or reinforced concrete. The main
tunnels are then driven to each side, with the lining connected to the central
wall. This central wall may be continuous, or have openings, or be a series
of columns with a longitudinal beam along the top and bottom. The second
method is to construct one of the tunnels, then to cast an internal lining
with the strong central wall on one side of it. Then the second tunnel may
be driven, utilising the central wall for both the temporary and permanent
support.
11.3.8 Trinocular caverns

We do not have to stop at binocular; this method can be used to construct
three parallel caverns, which we could call ‘trinocular’. These are quite com-
mon in metro stations in Prague, Kiev, Moscow and elsewhere, where there
is a central concourse tunnel with platform tunnels either side (Figure 11.17).
There is no reason why we should stop at three tunnels joined together,
any number of parallel caverns could be constructed using similar methods.
In Budapest there is a metro station with five (Mecsi, 2002).
11.4 TOOLBOX MEASURES
Toolbox measures are additional measures that may be employed in order

to improve face stability, either to prevent collapse or excessive ground
movements. They are usually designed in advance to mitigate geological
risks that may or may not be realised, and the equipment and materials
needed to implement them must be ready to be used on site. For exam-
ple, if the ground is stiff clay but probing identifies a saturated sand lens
ahead of the face, then grouting, dewatering or depressurisation may be
employed to reduce the risk of running sand flowing uncontrollably into
the tunnel.
Figure 11.17 Trinocular cavern construction sequence.

Note that when using a sprayed concrete lining in soft ground, we do not
reduce the basic support design if the ground is better than foreseen, but we
will employ toolbox measures if a geological risk is realised.
The principles of undrained and drained stability in Chapters 2 and 3
can be used to predict the need for, and effectiveness of, different toolbox
measures.
11.4.1 Pocket excavation

In either very dry or saturated drained soils with very low cohesion, where
the ground has limited stand-up time and will fall down from the crown or
the face soon after exposure, pocket excavation may be used (e.g. Aagaard
et al., 2017; Gall et al., 2017). Two examples of pocket excavation are
shown in Figure 11.18. Each pocket is excavated and then sprayed with a
shotcrete sealing layer on all the exposed ground in the pocket until the top
heading arch has been excavated (1–3 on the left of Figure 11.18, 1–5 on
the right), then a primary lining layer is sprayed over the whole top head-
ing arch. Then the core may be excavated and the core face sprayed with a
sealing layer.
If the stand-up time of the ground at the crown is very short and a chim-
ney failure is possible, then pocket excavation should be used in conjunc-
tion with a means of preventing loss of ground at the perimeter of the
excavation, such as spiles, permeation grouting, jet grouting, ground freez-
ing or a pipe arch.
Note that pocket excavation still requires some stand-up time. Ground
that will fail immediately, particularly saturated cohesionless soils, cannot
be controlled using this method.
Ideally the spraying robot is positioned next to the excavator and each
pocket is sprayed with shotcrete as soon as it has been excavated. For this
reason, this method is usually only feasible in quite wide tunnels.
The size of the pockets can be determined using the stability calculation
methods in Chapters 2 and 3.
Figure 11.18 Two examples of pocket excavation.

11.4.2 Reduction of advance length or ring

closure distance
The advance length is usually set at 1 m in soft ground, but has been known
to be reduced to 0.8 m or increased to 1.2 or 1.5 m. If the advance length is
changed, then consideration will need to be given to logistics and the capac-
ity of spoil handling plant such as loaders, dumpers and skips, as well as the
capacity of the batching plant and shotcrete delivery trucks to provide the
correct amount of shotcrete.
As we saw in Chapter 2, the unsupported length P of a heading has an effect
on undrained stability. The smaller P is, the more stable the heading will be.
For a full-face excavation, the value of P is equal to the advance length,
plus an allowance for excavation of the face beyond the advance length,
which is usually assumed to be 0.5 m. For undrained soil, a change from
1 m advance length to 0.5 m will only make a small difference to the P/D
ratio (where D is the equivalent excavation diameter), even at the smallest
feasible sprayed concrete tunnel diameter of 3 m, so the improvement to
stability will be negligible.
For TH-I or TH-B-I excavation, the effective value of P is more difficult
to determine. A conservative assumption is that it is equal to the ring clo-
sure distance, which is the maximum distance from the face to a completed
invert. For the estimation of volume loss, Macklin (1999) recommended
using only the excavation ahead of the leading edge of top heading shot-
crete (this is covered in Chapter 12), because the unexcavated ground below
the top heading provides some weight to counter instability, like a berm.
For the sequence shown in Figure 11.13 (also shown on the left side of
Figure 11.19), and assuming that the advance length is 1 m, the maximum
distance occurs just after excavation of the invert (Stage 5) but before it is
sprayed, and is therefore 5 m. With an allowance for excavation of the face
ahead of the top heading lining of 0.5 m, the ring closure distance is 5.5 m.
Reducing the ring closure distance may be achieved by reducing the
advance length. However, this is likely to be inefficient in terms of produc-
tivity. Another solution is to move the bench and invert excavation closer to
the face, as shown in Figure 11.19, where on the left is shown the sequence
Figure 11.19 Reducing the ring closure distance in a TH-B-I excavation.

from Figure 11.13, and on the right is shown a modified sequence with a
shorter ring closure distance. In this example the ring closure distance has
been reduced from 5.5 to 4.5 m.
11.4.3 Ground improvement

Ground improvement means improving the properties of the ground, either
by grouting or ground freezing. The aim is to improve face stability by
increasing the soil’s cohesion. It will also reduce the soil’s permeability and
this will improve working conditions by reducing groundwater inflows.
Grouting can be targeted at zones of ground with low cohesion, for
instance sand lenses within an otherwise competent clay, or it can be a
systematic procedure if all the ground has low cohesion. It can be done
from the surface, from a shaft, or from the tunnel itself. The wedge-prism
method described in Chapter 3 can be used to determine the cohesion
required to make the face stable.
This book will not go into detail on grouting methods. These are covered
by EN 12715: 2000 ‘Execution of special geotechnical work – Grouting’,
and there is plenty of guidance in the CIRIA guide ‘Grouting for ground
engineering’ (Rawlings et al., 2000) or in the book ‘Chemical Grouting and
Soil Stabilization’ (Karol, 2007).
11.4.4 Support ahead of the face

In both drained and undrained soils, tunnel headings collapse in a wedge-
shaped mechanism that is loaded by the weight of the ground above.
Therefore, if the ground above the face can be supported before the face is
excavated, this gravity load will not be applied to the face wedge, improving
stability. Even if the face collapses, the pre-support may prevent upwards
migration of the collapse, where it might have an impact on buildings or
utilities. This is illustrated in Figure 11.20.
Pre-support may be in the form of spiles, pipes, or could be jet-grouted
or frozen soil columns. Typically this pre-support is only around the crown
over an arc of 120°, as shown in Figure 11.21 for spiles.
Spiles are usually steel or GRP reinforcement bars with a sharpened tip
to aid driving, but can also be hollow rods. They are typically 32 mm diam-
eter and 3–4 m long, but can be up to 6 m long (Aagaard et al., 2017). They
span between the sprayed concrete lining that will be sprayed over their
back end and the ground ahead of the face. They need to be embedded a
certain distance into the ground ahead to make sure they can support the
ground at the perimeter of the excavation, so they will stop being effective
when the face approaches and the end bearing capacity begins to reduce.
Therefore, they are driven into the ground at a slight outwards angle so that
arrays of spiles can overlap, as shown in Figure 11.21. A typical spacing
around the perimeter is 200–300 mm (Holzleitner et al., 2005).
Figure 11.20 How pre-support improves stability.
In soils with very low cohesion, support ahead of the face will help to
prevent the soil in the crown from falling down, since it only needs to arch
over the spacing between the spiles or pipes, rather than arching front-to-
back over the advance length, as shown in Figure 11.22. Usually the soil
will break back to the line of the spiles and may in some cases continue to
ravel back until an arch is formed.
Pipe arches typically consist of self-drilling steel pipes of 114 mm diameter
(they can be any size, though normally they are between 60 and 200 mm)
and 15–20 m long. If the soil around them is to be grouted, they can be fitted
with tube à manchette valves at intervals along their length. They can be self-
drilling, or installed in pre-drilled boreholes (Lopez et al., 2019). Because of
their length, pipe arches need to be installed at a shallower angle than spiles,
and so an enlargement of the tunnel of approximately 20% is required to
create a headwall to allow installation (Holzleitner et al., 2005). The tunnel
Figure 11.21 Spiles as pre-support.

Figure 11.22 Soil arching between spiles or pipes.
therefore has a sawtooth profile in long-section, as shown in Figure 11.23,

which means more secondary lining concrete is used. As with spiles, pipe
arches need to overlap and this is normally by at least 0.4 times the width
of the tunnel (Peila & Pelizza, 2003), but may be larger depending on the
ground conditions and potential failure mechanisms. Pipes are installed with
an inclination away from the tunnel of between 3–14° (Janin et al., 2015;
Lopez et al., 2019). The pipes are typically spaced at 250–500 mm centres.
Spiles do not help much with face stability and are mostly used to limit
overbreak at the perimeter of the excavation. Pipes, on the other hand, will
have a small effect and this can be quantified. To calculate the drained stability
using the wedge-prism method, we can calculate the silo pressure acting on the
wedge using the Janssen silo equation (see Appendix A, Equation 15.27), with
the width of the silo being the spacing of the pipes. Remember to take account
of the fact that because they are angled outwards, the spacing and the offset
from the theoretical excavation line increase with length.
To check the spiles or pipes have sufficient structural capacity, a num-
ber of design methods are available. The spiles or pipes could be put into
a 3D numerical model of the advancing tunnel as beam elements (e.g.
Yamamoto et al., 2005; Janin et al., 2015; Janin, 2017; Syomik et al., 2019).
Figure 11.23 Sprayed concrete tunnel long section showing pipe arch pre-support.
Alternatively, a simple model can be used with a pin support (or a series
of springs) at one end representing the shotcrete lining and spring supports
over the embedded length representing the ground, with an assumed ground
load applied over the span (Ischebeck, 2005; Volkmann & Schubert, 2010).
11.4.5 Face dowels

While pre-support consisting of spiles or a pipe arch do improve face stabil-
ity, they only have a limited effect, and will not prevent face collapse in soils
with low cohesion and/or where there is groundwater present, or where the
face area is large. To stabilise the face wedge, face dowels can be used to
anchor it to the undisturbed ground ahead. Face dowels usually consist of
steel or GRP bars, and should be grouted into place.
Anagnostou & Perazzelli (2015) provide an excellent literature review of
design methods for face dowels, and present design nomograms based on
the wedge-prism method adapted with a method of slices to calculate the
stabilising effect. Face dowels will act as a reinforcement to strengthen the
ground, and also will apply a support pressure to restrain the face defor-
mations, in a similar manner to a soil nailed wall restraining a slope. They
primarily act in tension, and so their capacity is dependent not only on their
tensile strength, but also the bond strength of the bar/grout interface and
the resistance provided by the grout/soil interface.
Design and installation guidance for face dowels may be found in CIRIA
Report C637 ‘Soil Nailing – best practice guidance’ and partial factors and
design rules may be found in BS 8006-2: 2011 +A1: 2017 ‘Code of practice
for strengthened/reinforced soils – Part 2: soil nail design’.
11.4.6 Dewatering/depressurisation
Dewatering or depressurisation may be used as a toolbox measure to reduce
the risk of seepage-induced instability. The further this is done from the
face of the tunnel, the more effective it will be and the less it will disrupt
excavation and spraying, so if it is possible to do this from the surface, a
shaft, another tunnel or from a pilot tunnel, then this may be more efficient.
The effect of dewatering or depressurisation on drained soils may be calcu-
lated by using a 3D numerical seepage model to calculate the pore pressures,
and then to use these in the wedge-prism method presented in Chapter 3, or
to use a 3D numerical model for both the seepage and stability analysis.
11.5 CONSTRUCTION DETAILS AND TOLERANCES
Construction joints between parts of the lining sprayed at different times

are important for structural stability and watertightness of the tunnel.
There are two main types of joints; circumferential joints and radial joints.
Tolerances for setting out and spraying of the lining, including an allow-
ance for deformation, need to be included in the design drawings so that the
lining does not encroach on the required space envelope.
11.5.1 Circumferential construction joints

Circumferential construction joints are not structurally important, but to
make the tunnel as durable and watertight as possible, the joints should be
carefully cleaned and prepared prior to spraying. The leading edge of the
previous advance is tapered to make it more difficult to trap rebound in a
corner. In some cases, the initial layer has been staggered so that it does
not coincide with the primary layer to make any potential water path more
tortuous (Jones et al., 2008).
11.5.2 Radial construction joints

Radial construction joints are structurally important, and can be a weak
point in the lining where a failure may occur. Where lattice girders or steel
ribs and welded wire mesh are used this is not a problem, because there will
always be connections in the lattice girders or steel ribs and lapping of the
mesh between construction stages.
In fibre-reinforced shotcrete linings there will be very little continuity of
reinforcement across the joint, and the bond strength of shotcrete across a
construction joint is not usually as good as within a part of the lining sprayed
in one stage. Typically, bar reinforcement is placed across the joint to ensure
shear may be transferred and to provide rotation capacity in bending. This
may be achieved by drilling and resin-grouting steel reinforcement bars into
the exposed joint before spraying the adjoining part of the lining, or by using
‘Kwikastrip’, which are bent reinforcement bars housed within a casing that
can be removed, allowing the bars to be straightened (Figure 11.24).
In some cases, designers and constructors working together have found
ways of removing the need for bar reinforcement across radial construction
joints, improving safety by avoiding the need for operatives to enter the face
of the tunnel. As long as joints are cleaned and spraying is done carefully,
the bond across the joints can be very good. This usually requires careful
attention to mix design and pre-construction testing of joints in cylinder
and flexural beam tests, followed by careful supervision and quality control
during construction (Eddie & Neumann, 2004). An alternative used on the
Bank Station Capacity Upgrade project was to use lapped joints, where the
lowest 1 m of the primary lining of the top heading was only sprayed to half
thickness. This was then sprayed to full thickness at the same time as the
bench primary lining (Anthony et al., 2020). Compressive testing of cores
through the joint showed no reduction of compressive strength.
For sidewall drifts, the joint at the crown and invert between the sidewall
drift and the enlargement is critical. Due to the size of these tunnels and the
Figure 11.24 ‘Kwikastrip’ installed at the axis level prior to spraying the top heading (left).
The bars hidden in the box are later exposed and bent downwards ready for
invert spraying (right). This tunnel was a 5.3 m diameter TH-I excavation.
high shear forces and bending moments that can be induced at these joints,
they typically require reinforcement. Installing this reinforcement in the
crown during excavation exposes operatives to the risk of falling ground
or shotcrete. It is possible to design this out by making the sidewall drift
slightly oversize and then, during enlargement excavation, spraying a sec-
ond layer of primary lining that is continuous across the sidewall drift and
enlargement. This will require careful 3D numerical modelling of the con-
struction stages to ensure that the shotcrete lining has sufficient capacity at
each stage of construction for the various temporary and permanent loads.
11.5.3 Tolerances
To ensure that the sprayed concrete lining does not impinge on the required
space envelope, a tolerance needs to be provided between the theoretical
intrados and the required space envelope. This will account for setting out
and spraying tolerances and for deformation of the lining, and will be based
on what the contractor believes is achievable. A lower tolerance means a
smaller overall volume of excavation, but it needs to be achievable most of
the time to avoid the need for milling back the lining.
On recent projects in the United Kingdom, a tolerance of 100 mm has
been used with approximately 75 mm being required as a spraying toler-
ance and 25 mm required as a deformation tolerance. For particularly large
caverns >10 m wide the deformation tolerance may be larger, and for areas
that are tricky to spray accurately, a larger tolerance may be necessary.
The tolerance is normally included in the drawings as a dashed line offset
inside the theoretical intrados line.
11.6 3D NUMERICAL MODELLING
For the basics of numerical modelling of tunnels, refer to Chapter 7. In this

section and the next we will discuss the principles of numerical modelling
of sprayed concrete tunnels, starting with 3D numerical modelling because
it involves fewer assumptions and is closer to the real situation, and is there-
fore easier to understand than 2D numerical modelling.
Since the excavation and lining sequence for a shotcrete-lined tunnel is usually
quite complex, 3D numerical modelling is often used. A 3D model allows the
explicit modelling of age-dependent shotcrete properties, such as its compressive
and flexural tensile strength, plasticity, creep, shrinkage and temperature effects.
The simplest constitutive model is to assume the shotcrete has a constant
elastic stiffness. This stiffness can be made to increase with age as the tunnel
advances. Assumptions about the speed of construction need to be made, which
should encompass a range because the actual advance rate on site will vary a lot.
In a numerical model, unless time is being explicitly modelled (which is
rare), the stress state only changes when the soil is excavated. Therefore,
each time an advance has been excavated (i.e. the elements representing the
soil are removed/deactivated in the model), there are out of balance forces,
so we need to perform a calculation to find the new equilibrium. Then the
lining is sprayed (i.e. lining elements are added/activated at the excavation
perimeter), but if a further calculation were performed after installing the
lining elements, nothing would happen because the model is still in equilib-
rium. Therefore, the model is solved after each excavation step but not after
each lining installation step.
11.6.1 Linear elastic sprayed concrete

constitutive model
The simplest constitutive model that can be used for a sprayed concrete
lining is a linear elastic model, and the simplest elements are shell elements
located at the perimeter of the excavation with no interface (i.e. the lin-
ing shell element nodes and soil solid element nodes cannot move indepen-
dently of each other). For a linear elastic model, only the lining thickness,
Young’s modulus and Poisson’s ratio need to be defined. As mentioned in
Chapter 7, positioning shell elements at the extrados results in an overesti-
mate of bending moments, which is conservative.
The outputs from any model, in terms of axial forces, shear forces and
bending moments, need to be checked against the capacity of the section.
If the constitutive model for the lining is linear elastic, then the model
itself will not fail if it is overstressed, so this check is essential. For fibre-
reinforced concrete, it may be simpler to output the maximum tensile and
compressive stresses and strains, and to compare these to the limiting values.
If the capacity of the section is exceeded at any point, then either the
lining thickness is increased, the degree of reinforcement (whether mesh or
fibres) is increased, or the profile may be changed to a more circular one.

Alternatively, a more sophisticated shotcrete constitutive model could be
used, because it is likely this will reduce peak stresses.
11.6.2 Ageing sprayed concrete properties

Each lining and excavation stage in a numerical model happens instanta-
neously, but represents a period of time in the real tunnel. If a full-face excava-
tion were proceeding at four advances per day, then each lining and excavation
stage would be six hours. During the time period of each advance, the stiffness
and strength of the sprayed concrete (and other properties, if modelled) will be
changing rapidly, therefore it is not obvious what single value to use for each
property to represent its behaviour during each time period. In reality, excava-
tion and spraying are not immediate and take some time. There are also pauses
between excavation and spraying activities as plant is moved around, invert
backfill is excavated or placed, hoses are connected, shotcrete is delivered,
joints in the lining are cleaned and prepared, etc. Common practice is to use
the value of the property at the midpoint of the time period of each advance.
An example of this is shown in Figure 11.25 for an advance rate of 4 m/day.
The designer should use the maximum advance rate the contractor believes
they can achieve at peak production, as this will be the worst case. If this
results in overstressing of the lining in the model, then either the lining needs
to be strengthened or a limit can be placed on the maximum advance rate.
The variation of properties with age is usually linked to the compressive
strength development. This should be obtained from testing of the shotcrete
used on site at different ages, but if this is not yet available, the designer
may use values from the literature for preliminary design, which can be
verified at a later stage when test data is available.
An example of compressive strength development used for numerical
analysis by Jones et al. (2008), based on shotcrete testing for the Heathrow
Terminal 5 works, is shown in Table 11.3.
On some projects, the design is based on one of the strength develop-
ment curves in EN 14887-1: 2005, known colloquially as the ‘J’ curves,
because they originate from the Austrian sprayed concrete guideline where
the curves were called ‘J1’, ‘J2’ and ‘J3’. However, the strengths given in the
standard are minimum requirements and do not follow the typical strength
development of sprayed concrete.
For a linear elastic constitutive model, we need to know how the Young’s
modulus develops with age. Chang & Stille (1993) proposed the following
equation that relates Young’s modulus with compressive strength:
Ec = 3.86fc0.6 (11.8)
Ec is the Young’s modulus of the sprayed concrete in GPa

fc is the compressive strength of the sprayed concrete in MPa
Figure 11.25 Ageing shotcrete in a 3D numerical model of an advancing full-face excava-

tion, at an advance rate of 4 m/day.
Table 11.3 E xample of compressive strength

development with age.
Sprayed concrete characteristic
Age compressive strength (MPa)
0 hours 0
0.1 hours 0.2
1 hours 0.5
3 hours 1.0
6 hours 3.0
12 hours 8.0
1 day 15.0
3 days 25.0
5 days 28.0
7 days 30.0
10 days 31.5
15 days 33.0
20 days 34.0
28 days 35.0
Since a higher stiffness will attract more ground load and will result
in higher bending moments, it is important to consider both the lower
and upper bound characteristic values. The lower bound value will
give a more conservative estimate of ground movements, whereas an
upper bound value will give a more conservative estimate of lining
forces.
Poisson’s ratio is usually taken as 0.2 for shotcrete or concrete.
11.6.3 Linear elastic – perfectly plastic sprayed

concrete constitutive model
This kind of constitutive model is particularly useful for fibre-reinforced
concrete, where it is very likely that the tensile strength will be exceeded
and a linear elastic model would therefore result in an unrealistic stress
distribution in the lining at these locations.
Not all numerical modelling programs are able to model plasticity
in shell elements, and it may be necessary to use solid elements for the
lining instead. Nasekhian & Feiersinger (2017) describe the use of a
‘concrete damaged plasticity model’ for shell elements in the finite ele-
ment program ABAQUS to model elastic – perfectly plastic behaviour in
tension. The yield stress was set at the design value of the ultimate post-
cracking tensile strength f Ftud (fib Model Code 2010, 2013), as shown in
Figure 11.26.
Figure 11.26 Elastic model with perfect plasticity at the design value of ultimate post-
cracking tensile strength.
fcd is the design value of compressive strength for the fibre-reinforced

sprayed concrete according to EN 1992-1-1: 2004 or the fib Model
Code 2010 (2013)
f Ftud is the design value of the ultimate post-cracking tensile strength for
the fibre reinforced sprayed concrete according to the fib Model Code
2010 (2013)
After running a numerical model using the constitutive relationship shown

in Figure 11.26, the tensile and compressive strains were checked against
the strain limits to verify that all parts of the lining were within the ultimate
limit state. As long as tensile strains in the model were less than 1%, and
compressive strains less than 0.2%, the lining was assumed to be within its
capacity. These strain limit values were determined based on the properties
of the sprayed concrete used on this particular project and included partial
factors on material strengths and the partial factor for actions.
Geotechnical numerical modelling programs often have limited shell ele-
ment constitutive models, in which case using solid elements for the lining
may be the only way of including shotcrete plasticity. An example is given in
Jones et al. (2008), where solid elements were used with a Mohr-Coulomb
model, with the parameters fit to pre-construction shotcrete testing data.
11.6.4 More complex constitutive models

Thomas (2003) used a series of 3D numerical models to investigate the
effect of different ageing shotcrete constitutive models, including linear elas-
tic, nonlinear elastic, hypothetical modulus of elasticity, elastoplastic, creep
and shrinkage. Replacing linear elastic behaviour with ‘softer’ models, such
as the model including creep, resulted in reduced stresses in the lining, with
bending stresses reduced more than axial stresses because creep is stress-
dependent and there was more stress relaxation where stresses were larger.
Schütz (2010) and Schütz et al. (2011) proposed an elastoplastic model
for shotcrete that takes account of creep, shrinkage and thermal deforma-
tion. It can also allow for cracking using a smeared crack approach.
All of the models we have discussed so far use the age of the shotcrete
as the basis for the development of all parameters with time. However, the
properties of all concretes including shotcrete actually depend on maturity
rather than age. Maturity can be thought of as the proportion of total
hydration (the chemical reactions between cement and water) that have
occurred. The rate of hydration is highly dependent on temperature, and
approximately doubles every 10°C. Therefore, at higher temperatures
concrete strength will increase much faster with time than at lower tem-
peratures. This has led to the development of shotcrete models based on
maturity rather than age by Hellmich et al. (2001), heavily based on the
experimental work of Byfors (1980) and the theoretical framework of

Ulm & Coussy (1995).
The problem for the designer in using these more complex models is that
at the design stage there is considerable uncertainty about the shotcrete
behaviour, because every shotcrete is different, and some aspects of these
models are based on experiments performed more than 20 years ago when
shotcrete technology was very different. Even if pre-construction testing
allows the use of parameters specific to the shotcrete being used, the envi-
ronmental conditions on site, and in particular the temperature, will have a
huge impact on how the properties of the shotcrete vary with time. It is also
very difficult to predict the rate that a tunnel will be advanced.
For all these reasons, it is important to undertake sensitivity analysis,
where all the parameters used in the constitutive model are varied within
realistic limits, so that the sensitivity of the design to each of them can be
understood. It is also important to monitor the deformations and stresses
within the tunnel during construction, so that the design assumptions can
be verified.
11.7 2D NUMERICAL MODELLING
In a 2D numerical model, we need some way to take account of the 3D

nature of tunnel excavation, with stress redistribution occurring as arch-
ing develops in the ground ahead of and around the face and unsupported
perimeter. In Chapter 7, various methods for simulating the relaxation of
the ground ahead of lining installation in a 2D numerical model were dis-
cussed. The most common method used for sprayed concrete tunnels is the
convergence-confinement method.
Essentially, the convergence-confinement method involves removing the
elements representing the ground within the tunnel perimeter and replacing
them with nodal forces at the perimeter equal to some proportion of the in
situ stress in the ground. The lining is then installed and the nodal forces
are removed. For sprayed concrete tunnels, however, (except when full-face
excavation is used) this can become quite complex, as for each excavation
stage an assumption needs to be made about the amount of relaxation that
will occur.
An example for a 2D numerical model of a TH-B-I excavation is shown
in Figure 11.27. A total of six values of either confinement loss factor or β
factor (see Chapter 7) need to be specified to obtain the nodal forces pTH1,
pTH2 , pTH3, pB1, pB2 and pI. For a sidewall drift and enlargement excavation,
12 or more values may be needed.
Designers will often assume that the nodal forces are removed when the
lining is installed, i.e. that pTH2 = pTH3 = pB2 = 0 in Figure 11.27. However,
it is important to remember that assuming they are equal to zero is still an
assumption.
Figure 11.27 2D numerical model of a TH-B-I excavation.
The values of these nodal forces may be estimated using engineering

judgement, or preferably calibrated to a case study of a similar construction
method in similar ground conditions. Another option is to calibrate the
nodal forces using a 3D numerical model. This is not as silly as it sounds,
because the 2D model can be used more easily for sensitivity analyses, or
for slightly different sizes or shapes of tunnels, and it takes much less time
to run a series of 2D numerical models than to do the same in 3D.
Whichever calibration method is used, there will always be a large
amount of judgement required. In addition, calibration to a case study
is likely to only be a comparison with ground movements and not lining
stresses, because there are very few case studies that include measure-
ments of lining stresses. As discussed in Chapter 7, a model calibrated to
ground movements will not necessarily predict accurate lining forces. For
this reason, a sensitivity analysis varying the amount of relaxation prior to

lining installation should be performed, and the effect on the lining forces
understood. This is also one of the reasons why it is important to verify the
design during construction using monitoring, and why designers should
always be given access to the site and the monitoring data.
11.8 SUMMARY
Sprayed concrete linings allow much more flexibility in terms of geometry

than other lining types. However, there are constraints, and we cannot use
just any shape. In soft ground we still need the lining to act as an arch in
order to have an efficient structural form, but the shape does not have to
be a circle. Depending on the space envelope required inside the tunnel, the
shape can be modified to reduce the volume of excavation, though this must
be balanced against the thickness and reinforcement requirements of the
lining. The less circular the profile of a sprayed concrete lining is, the lower
the volume of excavation will be, but the bending moments will be larger,
and the lining will need to either be thicker or contain more reinforcement.
As the tunnel advances, the sprayed concrete lining is gaining strength
and stiffness while at the same time being increasingly loaded by the
ground, unlike precast or prefabricated linings, which have their long-term
design strength and stiffness at the time of installation.
Sprayed concrete quality cannot be fully verified prior to installation and
is dependent on workmanship, variability of materials and environmental
conditions. In particular, the temperature of the shotcrete will affect its rate
of strength development.
3D numerical modelling is essential for accurately modelling the shape
and construction sequence of a sprayed concrete lining. 2D numerical mod-
els can be useful and efficient if carefully calibrated, but the results should
be used with caution.
Decisions about face division, toolbox measures, construction details
and tolerances require the input of the construction team. Ideally, sprayed
concrete tunnels are designed and planned by an integrated design and
construction team.
The traditional contractual separation of temporary and permanent
works is inappropriate for any tunnelling project, but especially so for
sprayed concrete tunnels where the temporary and permanent works are
essentially the same structure, and the way that they are built affects the
loads they have to withstand. Even if the (inefficient) traditional view is
taken that the primary lining is temporary and the secondary lining is per-
manent, the construction methods used for the primary lining will affect
the loads in the secondary lining and it cannot be designed in isolation.
Constitutive models for sprayed concrete are available that simulate the
most important aspects of behaviour: the variation of properties with age
or maturity, elastoplasticity, creep, shrinkage and thermal effects. Every

effort should be made to base the model parameters on the actual shot-
crete being used, but even so, there will remain considerable uncertainties
due to the inherent variability of shotcrete as a material, the variability of
environmental conditions, the actual rate of advance of the tunnel, and
potential for variability in both profile and thickness. For these reasons,
as with other tunnelling methods but even more so; for sprayed concrete
the design and construction are inextricably interlinked. The design is not
finished when construction starts, but must be verified by monitoring and
back-analysis.
11.9 PROBLEMS
Q11.1. The required space envelope for a large pedestrian travellator pas-
sageway is shown in Figure 11.28. The bottom of the required
space envelope is 18 m below ground level. The ground is soft clay
with undrained shear strength cu = 40 kPa and bulk unit weight
γ = 18 kN/m3.
i. Fit a sprayed concrete lining profile around this required space
envelope.
ii. Allowing 100 mm construction tolerance and allowing 300 mm
for a secondary lining, draw a primary lining 300 mm thick.
iii. Calculate the excavation area.
Figure 11.28 Problem Q11.1 required space envelope for a large pedestrian travellator
passageway.
iv. From the excavation area in (iii), calculate an equivalent diam-

eter. Use this to estimate the undrained stability for a full-face
excavation, remembering to use a factor of safety of 1.4.
v. Assuming the maximum reach of the excavator is 5 m, and
taking account of undrained stability, divide the face in an
appropriate manner. Draw a cross-section and long-section
showing the construction sequence.
Q11.2. The required space envelope for an upper lift lobby in a metro
station is rectangular, 5.3 m high and 3.6 m wide. The perma-
nent structure will be placed inside this envelope. The ground is
slightly clayey medium dense fine and medium sand with drained
cohesion c′ = 5 kPa, drained angle of friction ϕ ′ = 35° and unit
weight γ = 17 kN/m3. The bottom of the required space envelope
is 18 m below ground level. The tunnel is above the water table.
i. Fit a sprayed concrete lining profile around this required space
envelope.
ii. Allowing 100 mm construction tolerance, draw a primary lin-
ing 250 mm thick.
iii. Sketch the likely radial deformation assuming a uniform
radial ground pressure applied to the lining.
iv. From the radial deformation in (iii), sketch the likely bending
moment distribution in the lining.
v. Taking account of drained stability using the wedge-prism
method, and remembering to apply a partial factor of 1.25
on tan ϕ ′ and c′, design a viable face division and construction
sequence.
vi. Design toolbox measures in case the soil cohesion is lower
than expected.
Q11.3. The required space envelope for a road tunnel is given in Figure 11.29.
It is to be constructed using a sprayed concrete lining. The bot-
tom of the required space envelope is at 20 m below ground level.
Assume the tunnel is in stiff clay with a unit weight of 20 kN/m3,
a characteristic value of undrained shear strength of 140 kPa and
an undrained Young’s modulus of 84 MPa. The pore pressure is
hydrostatic with a piezometric level at the surface. Coefficient of
earth pressure at rest K0 = 0.8.
i. Fit a sprayed concrete lining profile around the required space
envelope.
ii. Allowing 100 mm construction tolerance and 300 mm for a
secondary lining, draw the primary lining assuming it is
300 mm thick.
iii. Sketch the likely radial deformation assuming a uniform
radial ground pressure applied to the lining.
Figure 11.29 Problem Q11.3 required space envelope for a road tunnel.
iv. From the radial deformation in (iii), sketch the likely bending
moment distribution in the lining.
v. Assuming the maximum vertical reach of the excavator is 5 m,
and taking account of undrained stability, divide the face in
an appropriate manner. Draw a cross-section and long-section
showing the construction sequence.
vi. Using the compressive strength development in Table 11.3 and
Chang & Stille’s relationship between compressive strength
and Young’s modulus, calculate the stiffness values that need
to be used in a 3D numerical analysis, if the advance rate is 4
advances (whether they are top headings, benches or inverts)
per day.
vii. (Advanced) Build and run a 2D numerical model of the tunnel pri-
mary lining. Experiment with different assumptions about nodal
forces for each calculation stage and lining thickness. Based on
your answer to (vi), you will need to estimate the Young’s modu-
lus of the sprayed concrete at each calculation stage.
viii. (Advanced) Build and run a 3D numerical model of the tunnel
primary lining. Compare the results in terms of ground move-
ments and lining forces with the 2D numerical modelling for (vii).
ix. (Advanced) Assuming the lining is steel fibre reinforced shot-
crete grade FRC 40/50 – 4.0c, plot the results from your
numerical analysis in (vii) and/or (viii) on a moment-force
interaction diagram, remembering to apply partial factors.
Are the lining forces within the ultimate limit state capacity
curve? If not, adjust the profile or the thickness of the lining
and re-run the numerical model.
REFERENCES
Aagaard, B., Gylland, A. S., Schubert, P. & Løne, B. (2017). The Joberg Tunnel –
Successful tunnelling in moraine. Proc. World Tunnel Congress 2017,
Surface Challenges – Underground Solutions (eds Nilsen, B., Holter, K. G. &
Jakobsen, P. D.), Bergen, Norway, pp. 1038–1047. Norwegian Tunnelling
Society NFF.
Anagnostou, G. & Perazzelli, P. (2015). Analysis method and design charts for bolt
reinforcement of the tunnel face in cohesive-frictional soils. Tunn. Undergr.
Space Technol. 47, 162–181.
Anthony, C., Kumpfmüller, S., Feiersinger, A. & Ares, J. (2020). Improving safety
through design at London Underground’s Bank station capacity upgrade.
Proc. Instn of Civ. Engrs – Civil Engrg 173, No. 5, 41–47.
BS 8006-2: 2011 +A1: 2017. Code of practice for strengthened/reinforced soils –
Part 2: Soil nail design. London, UK: British Standards Institution.
Byfors, J. (1980). Plain concrete at early ages, CBI Forskning Fo 3:80, p. 464.
Stockholm: Swedish Cement and Concrete Research Institute.
Chang, Y. & Stille, H. (1993). Influence of early age properties of shotcrete on tun-
nel construction sequences. Proc. shotcrete for underground support VI (eds.
Wood, D. F. & Morgan, D. R.), Niagara-on-the-lake, Canada, pp. 110–117.
USA: ASCE.
CIRIA (2005). Soil nailing – best practice guidance, C637. London, UK: CIRIA.
Eddie, C. and Neumann, C. (2004). Development of LaserShell™ Method of
Tunneling. Proc. of the North American Tunneling Conference 2004,
Atlanta. Rotterdam: Balkema.
EN 12715: 2000. Execution of special geotechnical work – Grouting. Brussels:
EN 14887-1: 2005. Sprayed concrete – Part 1: definitions, specifications and con-
formity. Brussels: European Committee for Standardization.
EN 1992-1-1: 2004. Eurocode 2: Design of concrete structures – Part 1-1: gen-
eral rules and rules for buildings. Brussels: European Committee for
Standardization.
fib Model Code 2010 (2013). fib Model Code for Concrete Structures 2010.
Gall, V., Pyakurel, S. & Munfa, N. (2017). Conventional tunnelling in diffi-
cult grounds. Proc. World Tunnel Congress 2017, Surface Challenges –
Underground Solutions (eds Nilsen, B., Holter, K. G. & Jakobsen, P. D.),
Bergen, Norway, pp. 2519–2528. Norwegian Tunnelling Society NFF.
Hellmich, C., Macht, J., Lackner, R., Mang, H. A. & Ulm, F.-J. (2001). Phase
transitions in shotcrete: from material modelling to structural safety assess-
ment. Shotcrete: Engineering Developments, Proc. Int. Conf. Engineering
Developments in Shotcrete, Hobart, Tasmania, Australia (ed. Bernard, E. S.),
pp. 173–184. Rotterdam, The Netherlands: Balkema.
Holzleitner, W., Deisl, F. & Holzer, W. (2005). Tunnel support ahead of the face
with pipe roofs or forepoling. Rational Tunnelling 2nd Summerschool (eds
Kolymbas, D. & Laudahn, A.), Innsbruck, Austria, pp. 147–164. Berlin,
Germany: Logos Verlag.
Ischebeck, E. F. (2005). New approaches in tunnelling with composite canopies –
installation-design-monitoring. Rational Tunnelling 2nd Summerschool
(eds Kolymbas, D. & Laudahn, A.), Innsbruck, Austria, pp. 165–171. Berlin,
Germany: Logos Verlag.
ITA-AITES (2020). Permanent sprayed concrete linings. ITA Report No. 24, ITA
Working Group no. 12 and ITAtech. Lausanne, Switzerland: ITA-AITES.
Jäger, J. (2016). Structural design of composite shell linings. Proc. World Tunnel
Congress 2016, 22nd –28th April, San Francisco, USA. Englewood, CO, USA:
Society for Mining, Metallurgy and Exploration.
Janin, J. P. (2017). Apports de la simulation numérique tridimensionnelle dans les
études de tunnels. Revue Française de Géotechnique 150, No. 3, 1–13.
Janin, J. P., Dias, D., Emeriault, F., Kastner, R., Le Bissonnais, H. & Guilloux,
A. (2015). Numerical back-analysis of the southern Toulon tunnel measure-
ments: A comparison of 3D and 2D approaches. Eng. Geol. 195, 42–52.
Jones, B. D. (2018). A 20 year history of stress and strain in a shotcrete primary lin-
ing. Proc. 8th Int. Symp. Sprayed Concrete - Modern use of wet mix sprayed
concrete for underground support, Trondheim, Norway, 11th - 14th June (eds
Beck, T., Myren, S. A. & Engen, S.), pp.206-222. Oslo, Norway: Norwegian
Concrete Society/Norsk Betongforening.
Jones, B. D., Thomas, A. H., Hsu, Y. S. & Hilar, M. (2008). Evaluation of innova-
tive sprayed-concrete-lined tunnelling. Proc. Instn of Civ. Engrs – Geotech.
Engrg 161, GE3, 137–149.
Karol, R. H. (2007). Chemical grouting and soil stabilization, 3rd revised edition.
New York: Dekker.
Lopez, F., von Havranek, F. & Severi, G. (2019). Alternative umbrella arches: The
use of composite pile roofs. Proc. World Tunnel Congress 2019, Tunnels
and Underground Cities: Engineering and Innovation meet Archaeology,
Architecture and Art (eds Peila, D., Viggiani, G. & Celestino, T.) Naples,
Italy, pp. 2527–2535. London, UK: Taylor & Francis Group.
Lunardi, P. (2008). Design and construction of tunnels. Analysis of controlled
deformation in rocks and soils (ADECO-RS). Springer.
Macklin, S. R. (1999). The prediction of volume loss due to tunnelling in overcon-
solidated clay based on heading geometry and stability number. Ground Eng.
No. 4, 30–33.
Mecsi, J. (2002). Ground surface movements resulting from metro station and
tunnel construction in Budapest (Hungary). Proc. Int. Symp. Geotechnical
Aspects of Underground Construction in Soft Ground, Toulouse, France
(eds Emeriault, F., Kastner, R., Dias, D. & Guilloux, A.), pp. 349–354. Lyon,
France: Spécifique JLP.
Nasekhian, A. & Feiersinger, A. (2017). SCL Design Optimisation at Bank - A
combined lining approach. Proc. World Tunnel Congress 2017, Surface
Challenges – Underground Solutions (eds Nilsen, B., Holter, K. G. & Jakobsen,
P. D.), Bergen, Norway, pp. 658–665. Norwegian Tunnelling Society NFF.
Peila, D. & Pelizza, S. (2003). Ground reinforcing and steel pipe umbrella sys-
tem in tunnelling. Rational Tunnelling Summerschool (ed. Kolymbas, D.),
Innsbruck, Austria, pp. 93–132. Berlin, Germany: Logos Verlag.
Rawlings, C. G., Hellawell, E. E. & Kilkenny, W. M. (2000). Grouting for ground
engineering, CIRIA C514. London: CIRIA.
Schütz, R. (2010). Numerical modelling of shotcrete for tunnelling. PhD Thesis,
Imperial College London.
Schütz, R., Potts, D. M. & Zdravkovic, L. (2011). Advanced constitutive model-

ling of shotcrete: Model formulation and calibration. Comput. Geotech. 38,
834–845.
St John, A., Potts, V., Perkins, O. & Balogh, Z. (2015). Use of TBM pilots for large
diameter SCL caverns: Crossrail C300/C410. Proc. World Tunnel Congress
2015, SEE Tunnel: Promoting tunnelling in SEE Region (ed. Kolić, D.), 22nd–
28th May, Dubrovnik, Croatia, p. 15. Zagreb, Croatia: HUBITG.
Su, J., Bedi, A. & Bloodworth, A. G. (2019). On composite action and tunnel lining
design; is composite action beneficial in reducing lining thickness? Tunnelling
Journal, April/May, 38–46.
Syomik, A., Rex, R., Zimmermann, A. & Anagnostou, G. (2019). Evaluation and
numerical interpretation of measured pipe umbrella deformations. Proc.
World Tunnel Congress 2019, Tunnels and Underground Cities: Engineering
and Innovation meet Archaeology, Architecture and Art (eds Peila, D.,
Viggiani, G. & Celestino, T.) Naples, Italy, pp. 6249–6257. London, UK:
Thomas, A. H. (2003). Numerical modelling of sprayed concrete lined (SCL) tunnels.
PhD thesis, University of Southampton.
Ulm, F. -J. & Coussy, O. (1995). Modeling of thermochemomechanical couplings
of concrete at early ages. J. Engrg Mech. ASCE 121, No.7, pp. 785–794.
Volkmann, G. M. & Schubert, W. (2010). A load and load transfer model for
pipe umbrella support. Proc. Eurock 2010, Rock Mechanics in Civil and
Environmental Engineering (eds. Zhao, J., Labiouse, V. Dudt, J. -P. &
Mathier, J. -F.), pp. 379–382. London, UK: CRC Press.
Yamamoto, T., Murakami, K., Shirasagi, S., Negishi, K., Nishioka, K. &
Kobayashi, T. (2005). Application of the tunnel face reinforcement and fore-
piling. Proc. World Tunnel Congress 2005, Istanbul, Turkey, pp. 391–396.
Chapter 12
Estimating ground movements
In Chapter 1, we looked at how the ground moves when a tunnel is con-

structed, and how this depends on the construction method, soil type and
groundwater. In this chapter empirical equations and numbers will be intro-
duced to characterise these patterns of ground movements, and to allow pre-
dictions to be made for future tunnelling projects. The word ‘prediction’ is
used here advisedly, as it is not possible to predict ground movements with
any degree of accuracy, even in greenfield conditions. The word ‘estimate’
should perhaps be used, as suggested by Burland (2001) and Pound (2003).
After working through this chapter you will be able to estimate:
• transient and short-term transverse and longitudinal surface settle-

ments in different soil types
• volume loss
• long-term surface settlements in clay soils
• subsurface settlements
• surface and subsurface horizontal ground movements
• ground slopes, curvatures and strains
• ground movements due to shaft construction
‘Transient’ means the settlement is changing as the tunnel face is passing.

It does not mean the ground movements are temporary, their cumulative
effect is permanent.
‘Short-term’ settlement means after the tunnel face has advanced sufficiently
far beyond the point in question that construction is no longer having an effect
and settlement has stabilised to a relatively constant value. This short-term
value may change in the long-term due to creep or consolidation effects.
12.1 TRANSVERSE SURFACE SETTLEMENTS
Peck (1969) and Schmidt (1969) found that ground surface settlements trans-
verse to the tunnel may be represented by a Gaussian curve. Surface settle-
ments above real tunnels have repeatedly been found to take approximately
DOI: 10.1201/9780429470387-12 411
Figure 12.1 Gaussian surface settlement trough.
this shape, though as we will see the approximation tends to be better for
clays than for coarse-grained soils such as sands and gravels. This Gaussian
settlement trough is mathematically an ‘error function’ or Gaussian ‘normal
distribution function’ upside-down. It is shown in Figure 12.1.
The distance from the centreline to the point of inflexion of the curve,
which in a normal distribution would be the standard deviation, is known
as the ‘trough width’, i. The maximum settlement over the centreline of
the tunnel S max is analogous to the frequency of the mean of a normal dis-
tribution. Settlement S at offset y from the tunnel centreline is therefore
given by:
 − y2 
S = Smax exp  2  (12.1)
 2i 
S is the settlement at offset y, in m

Smax is the maximum settlement at y = 0, in m
y is the offset, or transverse distance, from the tunnel centreline, in m
i is the ‘trough width’, or the offset to the point of inflexion, in m
The area under the curve is defined as the ‘volume loss’, Vs. This may be
found by integrating Equation 12.1, as shown in the following equation:
∞ ∞
 − y2 
Vs =
∫
−∞
Sdy =
∫
−∞
Smax exp  2  dy = 2π .iSmax (12.2)
 2i 
Vs is the volume loss per unit face advance, in m3/m or m 2
Using Equation 12.1, Equation 12.2 or combinations thereof, the curve

may be defined by any two of the parameters Vs, Smax or i.
Estimating ground movements 413
O’Reilly & New (1982) did a big meta-analysis of settlements induced

by tunnelling in the United Kingdom, and found that the trough width may
be estimated by multiplying the height above the tunnel by a ‘trough width
parameter’ K, which is estimated by consideration of case histories. Trough
width parameter K is related to trough width i by the following equation:
i = K ( z0 − z ) (12.3)
K is the trough width parameter

z0 is the depth to axis of the tunnel under construction in m
z is the depth to the point under consideration, where z ≤ z0, in m. At
the surface, z = 0 and i = Kz0.
Note that Equation 12.3 can also be used to estimate the subsurface settle-
ment trough width, and we will look into this in the following Section 12.5.
The Gaussian curve is a useful tool, as it allows comparison of settle-
ments at different locations and on different projects by a small number
of practically relevant parameters. This then facilitates the empirical pre-
diction of the magnitude of settlement (as described by volume loss Vs or
maximum settlement Smax) and its extent (described by trough width i).
Empirical prediction is usually based on a meta-analysis of case histories.
This meta-analysis may be refined by restricting the list of case histories
considered to consist of, for example, only those tunnels using the same
excavation and support method at similar depth and with similar geology,
or by using correlations that take some or all of these factors into account.
O’Reilly & New (1982) provided meta-analyses for most soil types in the
UK, and Lake et al. (1996) and then Mair & Taylor (1997) extended this
to tunnels in a wide range of soil types around the world. All these studies
validated the use of the Gaussian curve.
In order to better make comparisons between tunnels of different sizes,
the volume loss is often expressed as a percentage of the excavation volume.
In other words, Vs is divided by the cross-sectional area of the excavation.
This is also referred to as the ‘volume loss’ but is given the symbol Vl .
Vs
Vl = % (12.4)
A
Vl is the volume loss as a percentage of face area, in %
Vs is the volume loss per unit face advance, in m 2
A is the face area, in m 2
Based on their meta-analysis of tunnels in the United Kingdom, O’Reilly &

New (1982) gave typical ranges of values of trough width parameter K and
volume loss Vl in different soil types as shown in Table 12.1.
Table 12.1 does not include coarser, granular deposits such as sand or gravel.
For cohesionless granular soils above the water table, O’Reilly & New (1982)
Table 12.1 Trough width parameter and volume loss in different soil types
(O’Reilly & New, 1982).
Soil type Construction method Value of K Value of Vl
Stiff fissured clay Shield or none 0.4–0.5 0.5–3%
Glacial deposits Shield in free air 0.5–0.6 2–2.5%
Shield in compressed air 1–1.25%
Recent silty clay deposits Shield in free air 0.6–0.7 30–45%
Shield in compressed air 5–20%
recommended a value of trough width parameter K between 0.2 and 0.3.

However, more recent work by Marshall (2009) and Marshall et al. (2012)
has shown that trough width parameter in sand is dependent on initial density
and tunnel volume loss. Also, we should remember that, unlike clays, sands
and gravels do not exhibit constant volume behaviour, and the volume loss at
the surface or subsurface will not be the same as the volume loss at the tunnel.
In a series of geotechnical centrifuge experiments investigating the effect
of tunnel size, depth and volume loss on the surface and subsurface settle-
ment trough in sand, Marshall et al. (2012) showed that the Gaussian curve
does not always provide a good fit to the settlement trough. As volume loss
at the tunnel increased, the trough width decreased, indicating the develop-
ment of a chimney-type failure mechanism above the tunnel.
In addition, they found that as volume loss at the tunnel increased, the
volume loss at the surface increased at a faster rate, which they ascribed to
the relative density of the sand. Under shear strain, the sand was found to
contract initially, followed by dilation at higher shear strains. They also
found that as depth increased or the size of the tunnel decreased, the trough
width became narrower.
WORKED EXAMPLE 12.1 OPEN FACE TBM

An open-face shield mounted with a roadheader is to be used to excavate a
10 m diameter tunnel in stiff clay. The tunnel axis is 20 m below the ground
surface. Based on case histories in similar geology using similar construc-
tion methods, the volume loss is predicted to be 1.5% and the trough width
parameter K is 0.45.
Calculate the maximum surface settlement Smax. Then calculate the
settlement at y = 5 m, 10 m, 15 m and 20 m.
We first need to calculate the face area A:
π D2 π × 102
A= = = 78.54 m 2
4 4
Then we can calculate the volume loss Vs using Equation 12.4:
Vs = Vl A = 0.015 × 78.54 = 1.178 m 2
The trough width is given by Equation 12.3:
i = Kz0 = 0.45 × 20 = 9 m
The maximum settlement Smax is given by Equation 12.2:
Vs 1.178
Smax = = = 0.05222 m = 52.2 mm
2π .i 2π × 9
The settlement at y = 5 m is given by Equation 12.1:
 − y2   − ( 52 ) 
S = Smax exp  2  = 0.052 × exp   = 0.04475 m = 44.8 mm
 2i   2 × 92 
Note that –y2 means –(y2), so if using a spreadsheet, pay particular atten-
tion to the use of brackets. Also take care with percentages and use con-
sistent units (m or mm, but not both).
Similarly, the settlements at the other offsets are given in Table 12.2.
Table 12.2 Worked Example 12.1 calculated settlements.

Offset from tunnel centreline y (m)
0 5 10 15 20
Settlement S (mm) 52 45 28 13 4
It is not considered good practice to predict surface settlements to better

than 1 mm accuracy. This is because of the considerable uncertainties
involved in the estimation of trough width, volume loss and the char-
acterisation of the geology. Therefore, final outputs of any settlement
calculation should be rounded to the nearest mm.
WORKED EXAMPLE 12.2 CLOSED

FACE TBM TUNNEL IN SAND
An earth pressure balance TBM is to be used to excavate a 7 m diameter
tunnel in sand. The tunnel axis is 15 m below the ground surface. Based
on case histories in similar geology using similar construction methods,
the volume loss is predicted to be 0.5% and the trough width parameter
K is 0.25.

settlement at y = 2.5 m, 5 m, 7.5 m and 10 m.
π D2 π × 7 2
A= = = 38.48 m 2
4 4
Vs = Vl A = 0.005 × 38.48 = 0.192 m 2
i = Kz0 = 0.25 × 15 = 3.75 m
Vs 0.192
Smax = = = 0.020 m = 20 mm
2π .i 2π × 3.75
 − y2   − ( 52 ) 
S = Smax exp  2  = 0.020 × exp   = 0.0082 m = 8.2 mm
 2i   2 × 3.752 

0 2.5 5 7.5 10
WORKED EXAMPLE 12.3 OPEN FACE TUNNEL IN CLAY

WITH OVERLYING COARSE-GRAINED SOIL LAYER
The same tunnel in Worked Example 12.1 is to be constructed in stiff
clay with an overlying sandy gravel layer.
The tunnel axis is 20 m below the ground surface. The sand layer is
8 m thick.
Based on case histories in similar geology using similar construction

methods, the volume loss is predicted to be 1.5% and the trough width
parameter K is 0.45 in the clay and 0.25 in the sandy gravel.
settlement at y = 5 m, 10 m, 15 m and 20 m.
π D2 π × 102
A= = = 78.54 m 2
4 4
Vs = Vl A = 0.015 × 78.54 = 1.178 m 2
The trough width is given by:
i = K1z1 + K2 z2 = 0.45 × 12 + 0.25 × 8 = 5.4 + 2 = 7.4 m
Vs 1.178
Smax = = = 0.06351 m = 63.5 mm
2π .i 2π × 7.4
 − y2   − ( 52 ) 
S = Smax exp  2  = 0.052 × exp   = 0.05055 m = 50.6 mm
 2i   2 × 7.42 

0 5 10 15 20
The effect of the overlying sand layer is to make the surface settlement
trough width narrower, from 9 m in Worked Example 12.1 to 7.4 m in
this example. This causes a higher maximum settlement and a steeper
trough, as shown in the comparison of the two troughs in Figure 12.2.
Figure 12.2 Settlement troughs from Worked Examples 12.1 and 12.3.
12.2 ESTIMATING VOLUME LOSS
Short-term volume loss arises from several sources in TBM-driven tun-

nels. There is ground movement towards the face of the tunnel, which will
depend on support pressure and unsupported length. There is always an
overcut at the front of the shield that allows the shield to steer, either cre-
ated by an overcutting bead when the profile is cut by the shield itself, or by
the position of the cutters on the cutterhead. This means that there will be
ground closure around the shield and tailskin, adding to the volume loss.
In closed face TBMs, this closure around the shield may be reduced if ben-
tonite slurry is pumped in and pressurised around the shield. As the TBM
is advanced, either the ring will be expanded against the ground using a
wedgeblock segment, or grout will be introduced into the annulus between
the segmental lining and the ground. The pressure of the ring expansion
or the grout pressure relative to the in situ stress in the ground will also
contribute to volume loss, as will any unfilled voids due to overbreak or
incomplete grouting. Finally, there is deformation of the lining under the
ground load, which allows more volume loss to occur.
For conventional soft ground tunnels lined with shotcrete, short-term
volume loss arises from ground movement towards the face and then by
deformation of the lining. Since the lining is loaded at early age, deforma-
tion of the lining will typically be greater than for a precast concrete seg-
mental lining.
12.2.1 Estimating short-term volume

loss in stiff clays
For most tunnels in clay there will be a long-term increase in volume loss.
This will be fully discussed in Section 12.4. In this section we will only
discuss ‘short-term’ volume loss, i.e. when the tunnel face has advanced
sufficiently far past the section in question that ground movements have
stabilised, but in a sufficiently short time period that undrained behaviour
can be assumed.
The magnitude of ground movements towards a tunnel heading, and
hence the volume loss, is dependent on the stability ratio N (also known as
the stability number). The stability ratio was introduced in Chapter 2, and
is given by:
γ z0 + σ s − σ t
N= (12.5)
cu

γ is the bulk unit weight of the soil in kN/m3
z0 is the depth to the tunnel axis in m
σ s is the effect at tunnel level of a surcharge at the surface (for instance
consisting of a stockpile of bulk materials, traffic loads, a flexible
raft foundation or a body of water) in kPa
σt is an internal face pressure provided by compressed air, slurry in the
head of a slurry TBM or earth pressure in an EPB machine in kPa
cu is the undrained shear strength in kPa (for guidance on what this
value should be, see Chapter 2)
The relationship between stability ratio and volume loss was first proposed
by Schmidt (1969). This was further developed by Mair et al. (1981) who,
based on centrifuge tests by Mair (1979), related volume loss to a param-
eter they called ‘Load Factor’ LF as follows:
N
LF = (12.6)
Nc
LF is the load factor, which is dimensionless

Nc is the critical stability ratio, which is the stability ratio at collapse,
also dimensionless
The critical stability ratio Nc was defined in Chapter 2 and is usually

determined from the empirical design chart produced by Mair (1979) and
Kimura & Mair (1981) based on centrifuge testing of tunnel headings in
overconsolidated kaolin clay. This is shown again in Figure 12.3.
Figure 12.3 Design chart based on Mair’s (1979) centrifuge tests (redrawn from
Kimura & Mair, 1981).
P is the unsupported length of the idealised tunnel heading in m

C is the cover from the crown of the tunnel to the top of the clay in m
D is the excavated diameter of the tunnel in m
The load factor LF is the inverse of factor of safety and is equal to zero
when the support pressure within the tunnel is exactly equal to the in situ
stress in the ground and equal to 1 when the tunnel heading is on the point
of collapse.
Case histories of volume losses measured above real tunnels in clay and
their estimated load factors were compiled by Macklin (1999), who pro-
posed an empirical formula to predict volume loss, as follows:
Vl (%) = 0.23e4.4 LF for LF ≥ 0.2 (12.7)
Vl is the volume loss as a percentage

LF is the load factor, defined in Equation 12.6
Load factors below 0.2 are not included because this would be outside the
range of the centrifuge data. It is also the value of load factor below which
the soil is essentially elastic.
The relationship between volume loss and load factor in Equation 12.7
is the central dashed line in Figure 12.4. Upper and lower bounds to the
Figure 12.4 Load factor chart for volume loss prediction in overconsolidated clays
(redrawn from Macklin, 1999).
case history data are also shown as solid lines. One of the outlying points
from Eden & Bozozuk (1969) is not in overconsolidated clay but extremely
sensitive Leda Clay, and the undrained shear strength was almost certainly
overestimated by Macklin (1999).
Since we are concerned here with ground movements that are well within
the ultimate limit state, then some judgement is required to determine the
P /D ratio used for calculating the critical stability ratio and hence the
load factor. Macklin (1999) recommended using the length of the shield
for unsupported length P, thereby assuming that the ground will not close
around it, stating that this would be a reasonable assumption for load fac-
tors less than 0.5. Where grouting does not occur immediately at the back
of the shield, unsupported length P could be even longer. For shotcrete-
lined tunnels, Macklin (1999) assumed unsupported length P was the
advance length plus overdig in front of the crown shotcrete.
Dimmock & Mair (2007) compared predictions using Macklin’s load fac-
tor approach with actual volume losses measured during the Jubilee Line
Extension in London. They showed that predictions for shotcrete-lined tun-
nels were reasonably good, but that predictions for TBM tunnels could be
underestimated and so they proposed a slightly modified approach where the
load factor approach was used only for volume loss due to ground movements
ahead of the face, using an unsupported length only based on excavation ahead
of the shield. Volume loss due to convergence around the shield and any void
between the shield and the lining not immediately filled with grout would be
calculated separately and added to the load factor volume loss.
The volume loss due to closure of ground around the shield is given by:
π Dδ 4δ
Vl (%) = × 100 = × 100 (12.8)
πD / 4
2 D
Vl is the volume loss as a percentage

D is the excavated diameter of the tunnel in m
δ is the average annular overcut (i.e. the theoretical average gap between
the ground and the shield before ground closure), plus any allow-
ance for a void between the shield and the lining, in m
Note that judgement is required to determine the average annular overcut δ,

because closure of the ground around the shield may only be partial, in which
case the volume loss will be less than the overcut. In some cases, for example
the Jubilee Line Extension running tunnels south of the lake in St James’s Park,
overbreak may result in a volume loss due to closure around the shield that is
greater than the theoretical overcut (Dimmock & Mair, 2007).
An alternative approach is to use a cylindrical cavity expansion method,
modelling the displacements of the soil around the shield. Mair & Taylor
(1993) proposed a cylindrical cavity expansion equation for estimating
displacements around a heading, which can be used for this purpose. It
assumes axisymmetry and that the soil is elastic-perfectly plastic.
δ c r 
= u  0  exp ( N * − 1) (12.9)
r0 2G  r 
δ is the radial displacement at radius r in m

r0 is the radius of the cylindrical cavity, i.e. the radius of the excavation
at the front of the shield, in m

G is the shear modulus of the soil in kPa
r is the radius where we wish to calculate radial displacement in m
N * is the stability ratio, where N * = σ 0 / cu
σ0 is the initial total stress at the cavity boundary in kPa, usually
assumed to be equal to the vertical total stress
To find the closure around the shield, we take r = r0 = D/2. So Equation 12.9
becomes:
cu D
δ= exp ( N * − 1) (12.10)
4G
This value of δ can be used in Equation 12.8 to calculate volume loss. If

Equation 12.10 gives a larger radial displacement than the average annu-
lar overcut, then take the average annular overcut as the value of δ in
Equation 12.8.
Dimmock & Mair (2007) also propose that an estimate be made of the
volume loss due to closure of the void between the shield and the lining.
This depends on the construction method, and they say that if the void is
grouted tight, or an expanded lining is expanded tightly against the ground
to the full diameter of the shield, then it should be negligible.
WORKED EXAMPLE 12.4 ESTIMATING

VOLUME LOSS FOR A SHOTCRETE TUNNEL
IN OVERCONSOLIDATED CLAY
A shotcrete-lined tunnel with a face area A of 87 m 2 is to be constructed
in stiff overconsolidated clay.
The centroid of the tunnel is 25 m below the ground surface. The excava-
tion level of the crown of the tunnel is at 20 m below the ground surface.
Excavation ahead of the leading edge of the last shotcrete sprayed is 1.5 m.
The undrained shear strength of the clay is given by cu = 50 + 8z kPa,
where z is the depth below the ground surface. The unit weight is
γ = 20 kN/m3
Estimate the volume loss this tunnel will experience using the load
factor approach.
We first need to calculate the equivalent diameter of the tunnel:
A 87
D=2 =2 = 10.525 m
π π
Cover C = 20 m, so C /D = 20/10.525 = 1.900 Unsupported length

P = 1.5 m, so P /D = 1.5/10.525 = 0.143 Reading from the design chart in
Figure 12.3, N c = 7.6
The undrained shear strength varies with depth so we will conserva-
tively take the value at 0.6z0 (see Section 2.2.2), i.e. at 15 m depth. Thus:
cu = 50 + 8 × 15 = 170 kPa
Stability ratio is calculated using Equation 12.5:
γ z0 + σ s − σ t 20 × 25 + 0 − 0
N= = = 2.94
cu 170
Therefore, load factor LF = N /N c = 2.94/7.6 = 0.39.

Vl (%) = 0.23e4.4 LF = 0.23exp ( 4.4 × 0.39) = 1.28%
WORKED EXAMPLE 12.5 ESTIMATING VOLUME

LOSS FOR A TBM IN OVERCONSOLIDATED CLAY
An EPB TBM-driven tunnel with an excavation diameter of 6.5 m is to
be constructed in very stiff overconsolidated clay overlain by a 5 m thick
layer of Made Ground.
The tunnel axis is 18 m below the ground surface. The average face
pressure to be applied (i.e. the value at axis level) is 100 kPa. Assume that
no support pressure is applied around the shield.
The average annular overcut of the TBM is 40 mm. The rings are
grouted immediately as they leave the tailskin, 7 m behind the cutter-
head, with a grout pressure equal to the overburden pressure.
The Made Ground has a unit weight of γ = 18 kN/m3 and is assumed
to have no cohesion.
The undrained shear strength of the clay is given by cu = 75 + 11z kPa,
where z is the depth below the top of the clay. The unit weight is
γ = 20 kN/m3 . The shear modulus G = 200cu .
Estimate the volume loss this tunnel will experience using the load factor
approach for the tunnel face, with an allowance for closure around the shield.
Cover of clay C = 18 − 6.5/2 − 5 = 9.75 m, so C /D = 9.75/6.5 = 1.5;

Unsupported length ahead of the shield P = 0 m, so P /D = 0; Reading
from the design chart in Figure 12.3, N c = 7.0.
The undrained shear strength varies with depth so we will conserva-

tively take the value at 0.6z from the top of the clay (see Section 2.2.2), i.e.
at 12.8 m depth, or 7.8 m depth from the top of the clay:
cu = 75 + 11 × 7.8 = 160.8 kPa
The stability ratio is calculated using Equation 12.5:
γ z0 + σ s − σ t (18 × 5 + 20 × 13) + 0 − 100 250

N= = = = 1.55
cu 160.8 160.8
Therefore, load factor LF = N /N c = 1.55/7.0 = 0.22.

Vl (%) = 0.23e4.4 LF = 0.23exp ( 4.4 × 0.22) = 0.61%
Next we will see if the ground closes around the shield due to elastoplas-
tic deformation, using the cavity expansion plasticity solution from Mair
& Taylor (1993), given in Equation 12.10:
cu D
δ= exp ( N * − 1)
4G
For this equation, it makes sense to use the undrained shear strength and
shear modulus at tunnel axis level (13 m from top of clay):
cu = 75 + 11z = 75 + 11 × 13 = 218 kPa
And:
G = 200cu = 200 × 218 = 43600 kPa
We also need to calculate N *:
N * = σ 0 /cu
Now the full overburden pressure at axis level is given by:
σ 0 = 18 × 5 + 20 × 13 = 350 kPa
Therefore:
N * = σ 0 /cu = 350/218 = 1.61

So closure (radial displacement) around the shield is given by:
cu D 218 × 6.5
δ= exp ( N * − 1) = exp (1.61 − 1) = 0.015 m = 15 mm
4G 4 × 43600
This radial displacement is less than the average annular gap of 40 mm,
so we will use it to calculate volume loss using Equation 12.8:
4δ 4 × 0.015
Vl (%) = × 100 = × 100 = 0.92%
D 6.5
Adding together the volume loss in front of the shield and around the
shield gives a total volume loss of 1.53%. This is probably an overes-
timate because we have assumed an infinitely long cylinder when esti-
mating ground closure around the shield, and also some of the radial
deformation to equilibrium will have already occurred in front of the
TBM so we are counting it twice. However, this is a better estimate than
using Dimmock & Mair (2007)’s method, which assumes full closure of
the annular overcut.
Since we know that the shield is not supporting the ground because
the plasticity solution told us that the ground closure around the shield
was less than the average annular gap, then we could also calculate the
volume loss by using an unsupported length P = 7 m.
Using P /D = 7/6.5 = 1.08 and C /D = 1.5 in the design chart in
Figure 12.3 gives N c = 5.2.
The stability ratio remains the same, so load factor is now:
LF = N /N c = 1.55/5.2 = 0.30
Vl (%) = 0.23e4.4 LF = 0.23exp ( 4.4 × 0.30) = 0.85%
This is significantly less than the first method we used, which estimated
a volume loss of 1.53%, but, as mentioned previously, is probably more
accurate because the first method assumed an infinitely long unlined cyl-
inder and is double-counting some of the radial deformation in the sepa-
rate estimates of volume loss ahead of the TBM and around the shield.
There will be no volume loss due to a void between the shield and the
lining, because grout is injected through the tailskin at a pressure equal
to the overburden pressure. Also, much, if not all, of the radial conver-
gence of the ground around the shield will be reversed by this grout pres-
sure, so the total volume loss is a conservative estimate.
This example is intended to show that only a rough estimation of vol-

ume loss is possible, but using a variety of different tools and methods
we can get a feel for what aspects of the construction method are impor-
tant. In this case, the application of face pressure reduces the volume loss
ahead of the cutterhead, but a significant proportion of the total volume
loss is still caused by the ground closing around the shield.
12.2.2 Estimating volume loss in other soil types

Vu et al. (2016) provide a state-of-the-art review of methods of estimating
volume loss at the face, around the shield, and at the tail due to grouting
pressure. For volume loss at the face, they use the load factor approach
described in the previous section for clays, but do not explain how to esti-
mate volume loss at the face in drained soils even though they later present
the results of this kind of calculation.
For volume loss along the shield, Vu et al. (2016) assume the annular
overcut is filled with either bentonite flowing from the face, or with grout
flowing from the tail. In the Netherlands, inert grouts are often used, so
it does not cause problems if the grout flows forwards around the shield.
For cementitious grouts this situation is strenuously avoided by employing
brushes or plates on the outside of the tailskin, otherwise grout can build
up around the shield and create a lot of friction.
For volume loss during grouting, Vu et al. (2016) recommend the use of
a cavity expansion method proposed by Mair & Taylor (1993) for clay soils
to model the cylindrical expansion or contraction caused by grout pressure
relative to the overburden pressure. Again, they do not explain how this
can be adapted from the undrained case to the drained case, though they
do present results for sand.
Generally speaking, settlements above tunnels in sands and gravels
with faces that are stable due to either cohesion or support pressure,
will be small. When instability occurs, the failure is ‘brittle’ and ground
movements increase quickly. Therefore, as long as face pressure is well
controlled to maintain stability, or the ground has sufficient cohesion,
volume loss in sand can be kept below 1.0% and in some cases reliably
below 0.5%.
12.3 LONGITUDINAL AND TRANSIENT

SURFACE SETTLEMENTS
In Section 12.1 we looked at short-term transverse settlements after con-

struction of the tunnel. In effect, the transverse section we considered was
well back from the face. In some situations, we may want to know the
transient settlements as the tunnel face approaches and passes a point of

interest. This was illustrated in Figure 1.6.
Attewell & Woodman (1982) showed how ground surface settlements at
all points relative to the face could be estimated using a combination of a
normal probability distribution function in the transverse direction and a
normal cumulative distribution function in the longitudinal direction.
Vs  − y 2    x − xi   x − xf  
S= exp  2  G   − G   (12.11)
2π .i  2i    i   i 
S is the vertical settlement at any point on the surface with coordinates

(x, y), in m
Vs is the volume loss in m 2
i is the trough width or the transverse distance to the point of inflexion
in m
y is the transverse distance from the centreline of the tunnel in m
x is the position along the centreline relative to the face of the tunnel
in m
xi is the tunnel starting point in m
xf is the tunnel face position in m
G is the standard normal cumulative distribution function. In partic-
ular, G(0) = 0.5 and G(∞) = 1. Tables of values may be found
online or in statistics textbooks. In Microsoft Excel, use the
NORMSDIST function.
Equation 12.11 can also be expressed in terms of the ‘short-term’ transverse

settlement at the same offset but well back from the face, now denoted S∞.
This may be easier to calculate because we can determine the transverse
settlements using Equation 12.1 and then multiply them by a function that
varies depending on position relative to the face.
  x − xi   x − xf  
S = S∞ G   − G   (12.12)
  i   i 
WORKED EXAMPLE 12.6 OPEN FACE SHOTCRETE-

LINED TUNNEL IN HOMOGENEOUS CLAY
A shotcrete-lined tunnel is to be excavated in clay with a face area of
54 m 2 . The tunnel axis is 15 m below the ground surface. Based on case
histories in similar geology using similar construction methods, the vol-
ume loss is predicted to be 1% and the trough width parameter K is 0.5.
Calculate the maximum surface settlement above the face. Then calcu-
late the settlement at y = 5 m, 10 m, 15 m and 20 m.
Calculate the same surface settlements at 10 m in front of the face, 5 m in
front of the face, 5 m behind the face and 10 m behind the face. Assume the
tunnel start position is xi = 0 m and the tunnel face position is xf = 100 m.
We will first calculate the volume loss Vs using Equation 12.4:
Vs = Vl A = 0.01 × 54 = 0.54 m 2
i = Kz0 = 0.5 × 15 = 7.5 m
Above the face, x = xf , therefore:
 x − xi   100 − 0 
  =  = 13.33
i   7.5 
 x − xf   100 − 100 
 i  =  7.5 
 =0
Calculating, or looking up the standard normal cumulative distribution

values gives:
 x − xi 
G = G (13.33) = 1
 i 
 x − xf 
G = G (0) = 0.5
 i 
The maximum settlement above the face is given by Equation 12.11:
Vs  − y 2    x − xi   x − xf  
S= exp  2  G   − G  
2π .i  2i    i   i 
 −0 
= 0.02872 × exp  {1 − 0.5}
 112.5 
= 0.02872 × 1{0.5} = 0.014 m = 14 mm
Now we can calculate the settlements along a transverse line above the face
using the maximum settlement we just calculated. Firstly, for y = 5 m:
 − y2   − ( 52 ) 
S = Smax exp  2  = 0.014 × exp   = 0.012 m = 12 mm
 2i   2 × 7.52 
0 5 10 15 20
The settlements at 5 m ahead of the face, 10 m ahead of the face, 5 m

behind the face and 10 m behind the face are calculated using the follow-
ing values in Table 12.6.
Table 12.6 Worked Example 12.6 longitudinal settlement factors.
x − xi  x − xi  x − xf  x − xf 
G G
i  i  i  i 
Above the face 13.33 1 0 0.5
5 m ahead of the face 14 1 0.67 0.75
10 m ahead of the face 14.67 1 1.33 0.91
5 m behind the face 12.67 1 −0.67 0.25
10 m behind the face 12 1 −1.33 0.09
Since the tunnel face is so far from the starting point, the function in the
third column of the table is always equal to 1. The position relative to
the face (the fifth column in the table) is going to be what modifies the
settlement values. We can see that if we subtract these values from 1, we
get smaller settlements ahead of the face, increasing as we pass the face
and move behind the face, as we would expect from our understanding
of ground movements as illustrated by Figure 1.6.
The full set of settlements is given in Table 12.7.
Table 12.7 Worked Example 12.6 longitudinal settlements.

10 m ahead of the face 3 2 1 0 0
5 m ahead of the face 7 6 3 1 0
Above the face 14 12 6 2 0
5 m behind the face 21 17 9 3 1
10 m behind the face 26 21 11 4 1
These transient surface settlements may also be presented graphically, as

Figure 12.5 Worked Example 12.6 transverse surface settlement curves relative to face
position.
It is usually assumed, as we have done so far, that the shape of the trans-
verse settlement trough remains constant and that only the magnitude of
the settlements increases as the tunnel face approaches and passes a trans-
verse array. However, data presented by Hill & Stärk (2015, 2016) from
Crossrail tunnels at Whitechapel, London, showed that the trough was
wider ahead of the face, becoming narrower as the face passed the array,
and then became wider again behind the face and into the long term. They
also found that when a pilot tunnel was constructed first followed by an
enlargement, the pilot tunnel would have a wider trough than the enlarge-
ment. The reasons for these effects are not known for certain, but may be
due to the stress history of the soil and the changes in stress path direction
the soil experiences as the tunnel advances.
12.4 LONG-TERM GROUND MOVEMENTS
As discussed in Chapter 1, long-term ground movements only occur in

undrained soils. They can occur in drained soils if the groundwater level
changes over the long term, but that is a straightforward 1D consolidation
problem that we will not cover here because it is not caused by the tunnel
construction.
There are two causes of long-term ground movements. The first is the
excess pore pressures induced by volumetric changes and shearing during
construction, which will dissipate over time to a final equilibrium or steady-

state value. The second is if the tunnel lining acts as a drain on the ground
around it, then this drainage will cause consolidation and thus settlement.
In low permeability clays, these long-term ground movements can take
years or decades to diminish to a negligible rate. Therefore, modern urban
tunnelling projects will often only allow settlement monitoring to stop
when the rate has reduced to below 2 mm/year.
12.4.1 Long-term ground movements

due to excess pore pressures
Since the stress changes that induce excess pore pressures are complex and
depend on the constitutive model of the soil, the only way to calculate them
is to use numerical modelling. This was covered in Section 7.6.2.
12.4.2 Long-term ground movements due

to the tunnel lining acting as a drain
Numerical models by Wongsaroj et al. (2007 & 2013) and Mair (2008)
demonstrated that in London Clay even slight changes in permeability of
the lining or the ground can have a significant effect on long-term settle-
ment magnitude and extent. Wongsaroj et al. (2007) also found that their
numerical model matched the piezometer and extensometer results from
the Jubilee Line Extension at St James’s Park better when the London Clay
was given a higher permeability in the horizontal direction compared to the
vertical, which increased the lateral extent of the drainage effect.
Wongsaroj et al. (2013) defined a ‘relative permeability index’ RP, given as:
kt C
RP = ⋅ (12.13)
ks t L
RP is the relative permeability index
kt is the permeability of the lining in m/s
ks is the permeability of the soil in m/s
C is the clay cover above the crown in m
tL is the lining thickness in m
Laver et al. (2017) improved the model of Wongsaroj et al. (2013) to include
a radial flow pattern, and derived the following equation for relative perme-
ability index:
Dkt  2C 
RP = ln  + 1 (12.14)
2ks t L  D 
ks is now the ‘average’ permeability of the soil in m/s, where ks = kv kh

D is the tunnel diameter in m
Figure 12.6 Relative permeability index and effect on long-term settlement due to con-
solidation (Laver et al., 2017 with permission from ASCE).
Results from Laver et al.’s 2D finite element modelling of long-term settle-

ment at various values of C/D ratio and volume loss are shown in Figure 12.6.
A curve with the following equation has been fitted to the data:
1
DS =
 1.4  (12.15)
 1 + 
RP 
DS is the ‘dimensionless settlement’, which is the long-term settlement

above the centreline of the tunnel, normalised so that it lies on a
scale of 0 to 1 where DS = 0 for a fully impermeable lining and
DS = 1 for a fully permeable lining
Figure 12.6 shows that when the relative permeability index is less than
0.01, the ground-lining system behaves as though the lining were imperme-
able, and no long-term settlement due to drainage consolidation occurs, and
when the relative permeability index is greater than 100, the ground-lining
system behaves as though the lining were fully permeable and the maximum
long-term settlement occurs. For relative permeability index values between
0.01 and 100, the ground is drained to a lesser extent and settlements will be
intermediate. Remember that changes in mean total stress, and shear-induced
dilation (in overconsolidated clays) and contraction (in normally consolidated
clays) may also cause long-term heave or settlement due to consolidation pro-
cesses, so even when RP < 0.01 there can still be long-term settlements.
The lining permeability required to achieve a relative permeability of

0.01 and hence exhibit impermeable behaviour can be calculated by rear-
ranging Equation 12.14 as follows:
kt 0.02t L
=
ks  2C  (12.16)
D.ln  + 1
 D 
Assuming a clay cover of 20 m, a tunnel diameter of 10 m and a lining
thickness of 0.3 m, values for lining permeability needed for the lining to
be considered impermeable are given in Table 12.8 for a range of different
soil permeabilities.
The values for lining permeability in Table 12.8 will vary if the cover,
diameter or thickness of the tunnel lining is changed, but they will not
change significantly and the orders of magnitude will remain the same.
Table 12.8 leads us to the slightly strange conclusion that the lower the soil
permeability is, the lower the lining permeability needs to be to prevent the
tunnel from acting like a drain and causing long-term settlements.
The permeability of clay is often less than 1 × 10 −9 m/s and London Clay,
for example, can have a permeability as low as 1 × 10−11 m/s (Wan & Standing,
2014), so the lining permeability needs to be at most 3.7 × 10 −13 m/s and
possibly as low as 3.7 × 10 −15 m/s, which is approximately the permeability
of a very good concrete without accounting for construction joints, shrink-
age cracks or other imperfections. Therefore, if long-term settlements are
to be avoided, then the tunnel probably also needs a waterproof membrane
or a watertight secondary lining.
This method can be used to predict long-term settlements if the permeabil-
ity of the lining is known. However, it is difficult to predict what the actual
permeability of a tunnel lining will be, because it depends at least as much on
joints, cracks and poor workmanship as on its inherent material properties.
If long-term settlements have been measured, then the lining permeabil-
ity may be estimated ‘a posteriori’ (after the fact) and a database of values
may be created for use in preliminary design on future projects. Laver et al.
(2017) did this for the Heathrow Express Trial Tunnel (Bowers et al., 1996)
and the Jubilee Line Extension at St James’s Park (Nyren, 1998), and the
Table 12.8 Values of lining permeability required for lining to be

considered impermeable, based on Wongsaroj et al. (2013)
equation for relative permeability.
Soil permeability Lining permeability required for RP = 0.01
1 × 10 m/s
−7 3.7 × 10−11 m/s
1 × 10−8 m/s 3.7 × 10−12 m/s
1 × 10−9 m/s 3.7 × 10−13 m/s
1 × 10−10 m/s 3.7 × 10−14 m/s
1 × 10−11 m/s 3.7 × 10−15 m/s
lining permeabilities were approximately 3 × 10 −11 m/s for the sprayed con-
crete lining of the Heathrow Express Trial Tunnel and 5 × 10 −10 m/s for the
expanded wedgeblock precast concrete segmental lining of the Jubilee Line
Extension. Both tunnels were in London Clay, and so far there have not
been any case studies analysed in this way in other types of clay.
12.5 SUBSURFACE GROUND MOVEMENTS
Subsurface settlements have also been found to follow a Gaussian settle-

ment trough, with the trough width becoming narrower with depth. In
undrained soils this can lead to some reasonably straightforward predic-
tion methods, but in drained soils it is a little more complicated.

in undrained soils
Due to the assumption of constant volume behaviour, the volume loss of a
subsurface settlement trough will be equal to the volume loss at the surface.
Therefore, we only need to know the trough width in order to be able to
calculate the subsurface settlements.
O’Reilly & New (1982) only wrote about surface settlements, but for tunnels
of different depths they found a linear correlation between tunnel depth and
trough width. Assuming that the presence of the ground surface itself has little
effect, then we could adapt their equation to calculate subsurface settlements:
i = K ( z0 − z ) (12.3)
i is the trough width or distance to the point of inflexion in m

K is the trough width parameter
z0 is the depth to axis of the tunnel under construction in m
z is the depth to the point under consideration in m, where z ≤ z0. At the
surface, z = 0 and therefore i = Kz0
By using this equation, we are effectively assuming that the subsurface

settlement trough width is equal to K multiplied by the height above the
tunnel, where K is a constant.
For tunnels in clays, Mair et al. (1993) looked at case history and cen-
trifuge test data of subsurface settlements and observed that the trough
width could be wider close to the tunnel than predicted by Equation 12.3.
Effectively, K could increase with depth. They plotted i/z0 against z/z0 and
fitted a straight line to the data. This is shown in Figure 12.7 with more
recent case studies added from Clayton et al. (2006), Cooper et al. (2002),
Jones (2010), New & Bowers (1994), Nyren (1998), Standing & Selman
(2001) and Wan et al. (2017b). The measurements by Cooper et al. (2002),
Jones (2010) and Standing & Selman (2001) are not from instruments in
Figure 12.7 Subsurface trough width variation with depth, normalised to tunnel axis
depth. Tunnel depths are given in square brackets.
boreholes, but from monitoring of third party tunnels above the tunnel
being constructed. These data appear to follow the same settlement pattern
as the other data, indicating that the tunnels are reasonably flexible (due to
the joints in the lining) and follow the ground deformations.
The equation of the straight dashed line in Figure 12.7 was proposed by
Mair et al. (1993), and is:
i  z
= 0.175 + 0.325  1 −  (12.17)
z0  z0 
Substituting for i using the relationship in Equation 12.3 gives:
K=
(
0.175 + 0.325 1 − z z
0
)
(12.18)
1− zz
0
Note that Equation 12.17 ensures that at the surface K = 0.5. Mair et al.
(1993) base this assumption on Rankin (1988)’s review of case histories of
surface settlements in the UK and overseas, where he stated this value was
appropriate for “initial practical estimation purposes”. This is not appro-
priate for soft clays, where K at the surface can be greater than 0.5.
It is easier to see what is going on if K is plotted on a graph against nor-
malised depth z/z0. This is shown in Figure 12.8.
An observation we can make about Figure 12.7 and Figure 12.8 is that
although Mair et al.’s relationship appears to fit the data reasonably well,
Figure 12.8 Subsurface trough width parameter K variation with normalised depth z/z0.
Tunnel depths are given in square brackets.
the deeper tunnels tend to be to the left and the shallower tunnels tend to
be to the right of it. Jones (2010) speculated that by using the normalised
depth z/z0 no account is made for the real distances involved. For deep tun-
nels, Mair et al.’s equation would predict high K values for a much larger
distance above the tunnel than for a shallow tunnel. Also, the value of K at
the surface must be equal to 0.5, regardless of the depth of the tunnel. Lake
et al. (1996) noted that for surface settlements the trough width parameter
may be smaller than might be predicted when the tunnel is deeper than 20 m,
and it does seem to be the case that deeper tunnels tend to have a K value
at the surface less than 0.5.
Jones (2010) therefore plotted K against height above the tunnel (z0 − z).
This is shown in Figure 12.9 with more recent data from Crossrail (Wan et al.,
2017b) added, and the logarithmic curve fit adjusted to take account of the
new data. The advantage of this method is that surface settlements may also be
plotted on the same graph and the value of K at the surface can vary depending
on the depth of the tunnel and does not always have to be constant.
The logarithmic relationship in Figure 12.9 is:
K = −0.261ln ( z0 − z ) + 1.286 (12.19)
In Figure 12.9 we can see that there is a lot of scatter in the data in the first
10 m above tunnel axis level and this is probably due to the effect of tunnel
diameter on the trough width so close to the tunnel. This is also present in
Figure 12.8 but is less evident due to the way the data is presented. Figure 12.10
presents the same data as Figure 12.9 but without the log scale.
Figure 12.9 Trough width parameter K variation with height above tunnel axis (z0 − z) on
a logarithmic scale. Subsurface settlements are from case histories shown in
the legend. Surface settlements are from Mair & Taylor (1997), Jones et al.
(2008), Clayton et al. (2006), Jones (2010) and Wan et al. (2017a).
Figure 12.10 Trough width parameter K variation with height above tunnel axis (z0 − z).
Subsurface settlements are from case histories shown in the legend.
Surface settlements from Mair & Taylor (1997), Jones et al. (2008), Clayton
et al. (2006), Jones (2010) and Wan et al. (2017a).
There is clearly more work to be done to improve the meta-analysis,

since there is still quite a large amount of scatter in the data. It is likely that
the cover, the tunnel diameter and the presence of non-cohesive superfi-
cial deposits above the clay are factors, and possibly also the construction
method and the volume loss.
WORKED EXAMPLE 12.7 SUBSURFACE SETTLEMENTS

ABOVE A TUNNEL IN HOMOGENEOUS CLAY
A TBM with an excavated diameter of 8.5 m is to be advanced in clay.
The tunnel axis is 35 m below the ground surface. Based on case histories
in similar geology using similar construction methods, the volume loss is
conservatively predicted to be 1%.
Calculate the trough width parameter K at 10 m, 15 m, 25 m and 35 m
above tunnel axis, using three different methods:
1. Assuming K has a constant value of 0.5.

2. Using Mair et al.’s empirical relationship in Equation 12.18.
3. Using the logarithmic empirical relationship in Equation 12.19.
Then calculate the maximum settlement at each level.
We will first calculate the trough width parameter K and then trough
width i using each of the methods:
Method 1:
K is constant, and i is given by i = K ( z0 − z ), giving the values in
Table 12.9.
Knowing that the percentage volume loss is 1%, we can calculate the
volume loss Vs from the excavated area as follows:
 D2  8.52
Vs = 1% × π   = 0.01 × π × = 0.567 m 2
 4  4
The maximum settlement can be calculated from the volume loss Vs and
the trough width i by rearranging Equation 12.2:
Vs
Smax =
2π .i
The Smax values are given in Table 12.9. They reduce as height above the
tunnel increases because the settlement trough is getting wider while the
volume loss is constant.
Table 12.9 Worked Example 12.7 Method 1 calculation.

z0 – z 10 m 15 m 25 m 35 m
K 0.5 0.5 0.5 0.5
i 5m 7.5 m 12.5 m 17.5 m
Smax 45.3 mm 30.2 mm 18.1 mm 12.9 mm
Method 2:
For this method we are using Mair et al.’s empirical relationship in
Equation 12.18:
K=
(
0.175 + 0.325 1 − z z
0
)
1− zz
0
Table 12.10 shows the steps in the calculation row-by-row. The question
gave the height above the tunnel axis of each point z0 – z, so we first con-
vert that to depth z, then we can calculate the relative depth z/z0. Then
K is calculated using Equation 12.18 (shown above). Then i and Smax are
calculated as for Method 1.

z0 – z 10 m 15 m 25 m 35 m
z 25 m 20 m 10 m 0m
z/z0 0.714 0.571 0.286 0
K 0.938 0.733 0.570 0.5
i 9.38 m 11.00 m 14.25 m 17.50 m
Smax 24.1 mm 20.6 mm 15.9 mm 12.9 mm
Compared to Method 1, using Mair et al.’s equation has resulted in

higher values of K, which increase closer to the tunnel, except for at
the surface where in both cases K = 0.5. The higher values of K result
in higher values of i in the subsurface. The wider trough width results
in lower maximum settlement Smax, because the same trough volume is
spread over a wider area.
Method 3:
For this method we are using the logarithmic empirical relationship in
Equation 12.19, which is:
K = −0.261ln ( z0 − z ) + 1.286
Table 12.11 shows the calculation row-by-row.

z0 – z 10 m 15 m 25 m 35 m
K 0.685 0.579 0.446 0.358
i 6.85 m 8.69 m 11.15 m 12.53 m
Smax 33.0 mm 26.1 mm 20.3 mm 18.1 mm
The subsurface trough width i values are greater than for Method 1, but
less than for Method 2, because this is a relatively deep tunnel. For a
shallow tunnel, it could be the other way round.
At the surface, the trough width is narrower than predicted using either
Method 1 or Method 2, because in this case K < 0.5, and for Method 1
and 2 K = 0.5 at the surface.
The logarithmic empirical relationship will give a surface trough
width parameter K < 0.5 for tunnel depths greater than about
20 m, but will give values of K greater than 0.5 for tunnel depths less
than 20 m.

in drained soils
Much less is known about subsurface ground movements in drained coarse-
grained soils. This is because the ground behaviour is much more complex.
It cannot be assumed to be constant volume, and different regions of the
ground will dilate or contract, depending on the soil’s relative density and
the current level of strain. Also, as we saw in Chapters 2 and 3, localised
shear bands can form in drained soils such that the behaviour can approxi-
mate to sliding blocks at limit state, whereas failure in undrained soils is
progressive and will be bounded by wide regions of shear that are better
described by velocity fields.
For these reasons, trough width in drained soils will depend on the level
of ground deformation that has occurred. In effect, for a given soil, trough
width varies with tunnel volume loss, tending to get narrower as tunnel
volume loss increases.
This behaviour is illustrated in Figure 12.11, which shows the variation
of trough width parameter K with depth ratio, where z is any depth below
the ground surface and z0 is the depth of the tunnel axis, so z/z0 = 0 is at the
ground surface and z/z0 = 1 is at tunnel axis level.
Figure 12.11 Variation of trough width parameter K with depth ratio z/z0 in dry sand
(redrawn from Marshall, 2009).
Figure 12.11 also shows that, for tunnels in sand, trough width param-
eter K tends to remain approximately constant or even decrease with depth,
which is the opposite of what occurs in clay.
Further work by Franza et al. (2019) investigated the influence of initial
density of the sand, and found that denser sand formed a ground arch
more readily, and closer to the tunnel, whereas loose sand, if it could form
one at all, formed a ground arch further away and was more prone to the
chimney-like vertical displacements above the tunnel crown that result in a
narrower settlement trough shape.
Another important thing to remember is that when we are discussing
ground movements in drained soils, because the behaviour is not constant
volume, the volume loss at any level in the ground can be different to the
volume loss at the tunnel. For loose and medium dense sands, it is usually the
case that the volume loss at the surface is significantly higher than volume
loss at the tunnel, whereas for dense sands the volume loss at the surface is
similar to that at depth, as shown in Figure 12.12 (Franza et al., 2020).
Predicting subsurface settlement in sand is therefore quite difficult.
However, as was explored in Chapter 3, only relatively small values of effec-
tive support pressure are required to keep deformations and hence volume
loss quite small. Therefore, as long as face pressure is well controlled to
maintain stability, volume loss in sand can be kept below 1.0% and in some
cases reliably below 0.5%. At these magnitudes of ground movements, in
Figure 12.12 Relationship between volume loss at the tunnel and volume loss at the
surface, at different C/D ratios and relative densities, in dry sand (redrawn
from Franza et al., 2020).
dense sands the behaviour may be close to constant volume, whereas in

looser sands the TBM driver may have to restrict ground movements at the
tunnel to a much greater degree because the volume loss at the surface will
be amplified, particularly for deeper tunnels (c.f. Figure 12.12).
Numerical modelling of dilatancy effects and shear band formation is
very challenging. Yang (2017) implemented a version of the ‘Norsand’
model, which could model stress dilatancy, in PLAXIS3D, and compared it
to the built-in ‘Hardening Soil Model’, which could not. He found that there
was very little difference at 0.5% volume loss, but at 1% volume loss and
higher the surface and subsurface settlement curves diverged. Therefore,
fairly standard models can be used to predict ground movements as long as
deformations are kept small, for example if a closed-face TBM is used with
good control of face pressure.
12.6 HORIZONTAL GROUND MOVEMENTS
Up to now we have focussed on vertical ground movements, but horizontal

ground movements also occur due to tunnelling and they can be just as
important to surface or subsurface structures.
If we assume that ground movements are constant volume and directed
towards the tunnel axis (O’Reilly & New, 1982), then it follows that hori-
zontal displacements are given by:
−y
H( y ,z ) = S( y ,z ) (12.20)
( z0 − z )
H( y ,z ) is the horizontal displacement at transverse offset y and depth z
S( y ,z ) is the vertical displacement (settlement) at transverse offset y and
depth z
In this way, horizontal displacements H( y ,z ) at any position ( y, z ) may be

calculated by first calculating a vertical displacement S( y ,z ) at the same posi-
tion and then using Equation 12.20.
Note that O’Reilly & New (1982) use z rather than (z0 − z) in Equation 12.20
(their equation (7)), but define z as ‘height above the axis’, therefore it is
equivalent to (z0 − z) here.
WORKED EXAMPLE 12.8 HORIZONTAL GROUND

DISPLACEMENTS ABOVE A TUNNEL IN STIFF CLAY
A TBM with an excavated diameter of 8 m is to be driven through stiff clay.
The tunnel axis is 20 m below the ground surface. Based on case histories in
similar geology using similar construction methods, the volume loss is con-
servatively predicted to be 1.2% and the trough width parameter K = 0.5.
Calculate the horizontal displacements at the ground surface.
We will first calculate the volume loss:
D2 82
Vs = Vl A = Vl π = 0.012 × π × = 0.603186 m 2
4 4
Next we need to calculate the trough width at the surface using
Equation 12.3:
i = Kz0 = 0.5 × 20 = 10 m
Then we can calculate the maximum settlement by rearranging

Equation 12.2:
Vs 0.603186
Smax = = = 0.024064 m = 24.1 mm
2π .i 2π × 10
Now we can calculate the vertical settlements using Equation 12.1:
 − y2 
S = Smax exp  2 
 2i 
The distance at which settlements usually become negligible is at a trans-
verse offset y = ±3i. Therefore, we will calculate vertical settlements S in
a spreadsheet with a range of y from −3i to 3i. An example calculation of
S is given below for y = −10 m:
 − ( −10)2 
S = 0.024064 exp  2 
= 0.024064 exp ( −0.5)
 2 × 10 

= 0.024064 × 0.606531 = 0.0146 m = 14.6 mm
A graph of S from the spreadsheet calculation is shown in Figure 12.13.
Figure 12.13 Worked Example 12.8 vertical settlements.
Next we will calculate horizontal displacements from the vertical

settlements, using Equation 12.20. An example, again for y = −10 m, is
given below:
−y
H( y ,z ) = S
z0 − z ( y ,z )
−10
H( −10,0) = × 0.0146 = 0.0073 m = 7.3 mm
20
The horizontal displacements calculated in the spreadsheet are shown in
Figure 12.14. Note that the displacements are positive on the left-hand
side and negative on the right-hand side, as the ground is moving towards
the tunnel centreline. Horizontal displacement is zero above the tunnel
centreline, as the ground is moving inwards from both sides and there
must be symmetry. The maximum horizontal displacement occurs at the
point of inflexion of the settlement trough, i.e. at y = − i and the mini-
mum likewise occurs at y = i.
Figure 12.14 Worked Example 12.8 horizontal displacements.

12.7 STRAINS IN THE GROUND
The Gaussian settlement trough may also be used to derive settlement gra-
dients, curvatures and strains, which are needed to predict the impact of
tunnelling on utilities pipelines or buildings. This section will focus on the
calculation of these derivatives. They will be used in Chapter 13 when esti-
mating building damage.
The horizontal ground strain may be found by differentiating Equation 12.20
with respect to y. The simplest way to do this is to manually calculate changes
in horizontal displacement along 1 m segments in a spreadsheet. The more ele-
gant method is to derive an equation, which is not as straightforward as it first
appears. We start by expanding Equation 12.20, because there are some hidden
functions of y we need to account for:
−y −y  − y2 
H( y ,z ) = S( y , z ) = S( max , z )exp  2  (12.21)
z0 − z z0 − z  2i 
H( y ,z ) is the horizontal displacement at transverse offset y and depth z

S( y , z ) is the vertical displacement (settlement) at transverse offset y and
depth z
S( max , z ) is the maximum settlement at depth z
y is the transverse offset from the tunnel centreline
z is the depth below the ground surface
z0 is the depth to the tunnel axis
Now we have to differentiate the following expression:
dH( y ,z ) d  −y  − y2  
εh = =  S( max,z ) exp  2   (12.22)
dy dy  z0 − z  2i  
Since Equation 12.22 contains a product of two terms with y in them, we

need to use the product rule, which states that:
duv dv du
=u + v (12.23)
dy dy dy
( )
Let ε h = d uv /dy , where:
−y
u= S (12.24)
z0 − z ( max , z )
and:
 − y2 
v = exp  2  (12.25)
 2i 
The derivative du /dy is given by:
du d  −y  − S( max,z )
=  S( max,z )  = (12.26)
dy dy  z0 − z  z0 − z
The derivative dv /dy is given by:
dv d   − y2   1  − y2 
=  exp  2   = − 2 ⋅ y ⋅ exp  2  (12.27)
dy dy   2i   i  2i 
Therefore, given that ε h = duv /dy , then:
 −y   1  − y2    − y 2   − S( max , z ) 
εh =  S( max , z )  ⋅  − 2 ⋅ y ⋅ exp  2   + exp  2  ⋅   (12.28)
 z0 − z   i  2i    2i   z0 − z 
Which simplifies to:
 − S( max,z )   − y2   y2 
εh =   ⋅ exp  2  ⋅ 1 − 2  (12.29)
 z0 − z   2i   i 
The horizontal strains calculated using Equation 12.29 for the tunnel in
Worked Example 12.8 are shown in Figure 12.15. As one would expect
from the distribution of horizontal displacements (c.f. Figure 12.14), there
is a maximum compressive strain above the tunnel centreline and the strain
becomes tensile outside the points of inflexion (the points of inflexion of
the Gaussian settlement trough, not of the horizontal strain). This latter
effect is because the maximum inward horizontal displacement occurs at
Figure 12.15 Worked Example 12.8 horizontal strains.

Figure 12.16 Worked Example 12.8 ground surface slope.
the points of inflexion; where horizontal displacement is increasing there is

tension, and where it is decreasing there is compression.
The change in slope θ of the ground surface, or of a subsurface trans-
verse line if we are considering the effect of tunnelling on a basement or
underground utility, can be found by differentiating the vertical settlement
(Equation 12.1) with respect to y.
dS d  − y2  − y  − y2 
θ= = Smax exp  2  = 2 Smax exp  2  (12.30)
dy dy  2i  i  2i 
The change in slope of the ground surface, calculated using Equation 12.30
for the tunnel in Worked Example 12.8, is shown in Figure 12.16. The
maximum and minimum slope occur at the points of inflexion of the
Gaussian settlement trough, i.e. at a transverse offset y = ± i. There is zero
slope over the tunnel centreline. Often slope is expressed as a fraction, for
example 0.002 would be written as 1/500.
Curvature of the ground surface may also be of interest. This is the sec-
ond derivative of vertical settlement with respect to y.
d2S d − y  − y2 
= S exp  2  (12.31)
dy 2 dy i 2 max  2i 
We need to use the product rule again (Equation 12.23). Let

d2 S /dy 2 = C.d ( uv ) /dy , where C is a constant given by:
− Smax
C= (12.32)
i2
Now let:
u = y (12.33)
and:
 − y2 
v = exp  2  (12.34)
 2i 
Therefore:
du
= 1 (12.35)
dy
and:
dv y  − y2 
= − 2 exp  2  (12.36)
dy i  2i 
Now we use the product rule equation (Equation 12.23) to find the curvature:
d2 S − Smax   y  − y2    − y 2  
= 2  y ×  − 2 exp  2   + 1 × exp  2   (12.37)
dy 2 i   i  2i    2i  
This simplifies to:
d2 S Smax  − y2   y2 
= ⋅ exp  ⋅ − 1 (12.38)
dy 2 i2  2i 2   i 2 
The curvature of the ground surface, calculated using Equation 12.38

for the tunnel in Worked Example 12.8, is shown in Figure 12.17.
Figure 12.17 Worked Example 12.8 ground surface curvature.

The maximum sagging curvature, which is perhaps obvious if you look

back at Figure 12.13 showing the settlement trough, occurs over the
tunnel centreline at y = 0. The maximum hogging curvature occurs at
y = ± 3.i .
In Figure 12.17, sagging curvature is negative because in the coordinate
system we have chosen to use the vertical axis z is pointing downwards
so it represents a negative change in slope. Hogging curvature is positive
because it represents a positive change in slope.
12.8 GROUND MOVEMENTS DUE

TO SHAFT CONSTRUCTION
Until recently, there were very few case studies of ground movements due to
shaft construction. This may be because shafts were rarely built close enough
to adjacent buildings for ground movements to be a concern. However,
nowadays it is increasingly common for deep, large diameter shafts to be
constructed in urban areas close to existing buildings or utilities.
New & Bowers (1994) analysed data from an access shaft at
Heathrow, near London, UK. Settlements and horizontal displace-
ments due to shaft excavation were measured along two radial lines
labelled ‘S’ and ‘T’, as shown in Figure 12.18. The 11 m diameter
shaft was constructed by caisson-sinking and then underpinning to
Figure 12.18 Settlements caused by Heathrow Express Trial Tunnel Access Shaft con-
struction (redrawn from New & Bowers, 1994).
16 m depth, and then excavated in 1 m steps and lined with shotcrete to

26 m below ground level.
Also shown on Figure 12.18 is an empirical relationship proposed by
New & Bowers (1994) between distance from the shaft wall d and the
settlement Sd, which has the following equation:
α .( H − d )
2
Sd = (12.39)
H
Sd is the settlement at distance d from the shaft wall in m

α is an empirical constant
H is the depth of the shaft excavation in m
d is the distance from the shaft wall in m
New & Bowers (1994) found the best fit for the empirical constant α was
0.0006. This corresponds to a maximum settlement at the shaft wall equal
to 0.06% of the depth, which they said was consistent with field data for
propped retaining structures in stiff fissured clays. However, this relation-
ship probably only holds true for shafts constructed in a similar manner
and in similar ground conditions to the one at Heathrow.
Faustin et al. (2018a) compared the settlements measured during
excavation of the 39 m deep Crossrail Limmo Peninsular Auxiliary
Shaft, constructed with sheet piles and then a shotcrete lining, with
Equation 12.39 and found that the settlements close to the shaft wall
were significantly underpredicted. They attributed this to the much
larger diameter, which was 28 m compared to the 11 m Heathrow Access
Shaft in New & Bowers (1994). Equation 12.39 does not account for
shaft diameter. In contrast, the Limmo Peninsular Main Shaft in East
London, of similar size to the Auxiliary Shaft but with a circular dia-
phragm wall installed before excavation began, had a maximum settle-
ment close to the shaft extrados of approximately 15 mm, approximately
10 mm of which was due to dewatering and only about 5 mm was due to
shaft excavation. Similarly, Schwamb et al. (2016) reported that Thames
Water’s Abbey Mills Shaft F in East London, 30 m diameter and 68 m
deep and also of diaphragm wall construction, caused maximum surface
settlements of less than 4 mm.
Therefore, it is likely that settlements due to shaft construction depend
on the following factors:
• shaft depth H
• shaft diameter D
• ground conditions
• construction method, in particular the stiffness of the lining and
whether it is installed before excavation (e.g. diaphragm walls) or
progressively (e.g. underpinning using a segmental lining or shotcrete)
Faustin et al. (2018b) showed that pre-installed linings, such as diaphragm

wall panels or secant piles, could induce as much or more settlement dur-
ing installation than the subsequent excavation. These settlements were
of the order of 0.02%H (0.02% of the depth of the shaft), which was a
similar magnitude to the settlements due to excavation in these cases.
These settlements due to installation are not always presented in case
studies, so it is important to remember this when using case studies to
make predictions.
Faustin et al. (2018b) analysed 26 case studies of circular shaft construc-
tion in London, and very considerately placed all the data on the journal’s
website. They separated the shafts into two main types; shafts where the
lining was pre-installed (named ‘SBE’, for ‘support before excavation’) and
shafts where the lining was installed concurrently with excavation (named
‘EBS’, for ‘excavation before support’). SBE shafts were further subdivided
into two groups: those that were a jacked caisson and those that were either
diaphragm wall or secant pile. Combined SBE/EBS shafts were also consid-
ered as two separate groups with the initial part of the shaft as a circular
sheet pile cofferdam or a secant pile wall, or the initial part as a jacked cais-
son. In both cases the EBS for the lower part consisted of a shotcrete lining.
This is summarised in Table 12.12.
When all this data was plotted by Faustin et al. (2018b), they found that
New & Bowers’ (1994) empirical relationship (Equation 12.39) did not fit
well to all shaft types. For some shaft types the maximum settlement at
the shaft wall could be substantially different to 0.06%H, and the extent
of the settlements could also be more or less than a distance H from the
shaft wall.
Developing this further, Equation 12.39 can be altered to the following
expression without changing its quadratic nature:
α .( β H − d )
2
Sd = (12.40)
β 2H
Table 12.12 Shaft types defined by Faustin et al. (2018b).

Faustin et al. (2018b) shaft types Description
‘SBE pre-installed’ Diaphragm wall or secant pile wall shaft
‘SBE jacked’ Precast concrete segments as a jacked caisson
‘EBS’ Precast concrete segments or shotcrete
underpinning
‘Combined SBE & EBS’ Pre-installed sheet pile or secant pile walls
through superficial gravels followed by
shotcrete lining in the London Clay
‘Combined SBE jacked & EBS’ Precast concrete segments as a jacked caisson
through superficial gravels followed by
Figure 12.19 Settlements caused by shaft construction – underpinning using precast con-

crete segments or shotcrete (EBS only).
The distance at which settlements are zero is given by β H , and the maximum
settlement at the shaft wall is still given by α H . I used regression to find the
best fit values of α and β for each shaft type. When β = 1, Equation 12.40 is
the same as Equation 12.39.
First let’s look at similar situations to New & Bowers’ case study; a
jacked caisson followed by underpinning using segments or shotcrete,
shown in Figure 12.19. Faustin et al. (2018b) put it into the EBS category,
even though the initial part of the shaft was a jacked caisson. This was
sensible because their settlements are at the higher end of the range. Also
on Figure 12.19 is a solid black line labelled ‘New & Bowers prediction’,
which is Equation 12.39, and a solid grey line labelled ‘Best fit curve’,
which is Equation 12.40. A least squares regression method was used to
find the best fit for α and β.
Note that Figure 12.19 has the axes normalised by dividing by the excavation
depth H, and the settlement is expressed as a percentage. This allows shafts of
different depths to be compared on the same graph and is a trick borrowed from
the analysis of retained cut and basement excavations (e.g. Peck, 1969).
Figure 12.20 shows the case studies for shafts constructed using a jacked
caisson of precast segments followed by shotcrete underpinning. There is a
lack of data close to the shaft wall that leaves some uncertainty about the
maximum settlements, but the best fit curve gives a better estimate of the
extent of the settlements, to 1.35H.
Figure 12.20 Settlements caused by shaft construction – jacked caisson followed by

underpinning.
Quigg (2019) presented a case study, also in London, of a jacked cais-

son followed by underpinning using shotcrete, and this has been added
to the dataset provided by Faustin et al. (2018b) in Figure 12.20. Quigg
found that by looking at the jacked caisson settlement on its own, there was
a maximum settlement of approximately 0.04%H, and for the shotcrete
underpinning it was 0.06%H.
Figure 12.21 shows the settlements for secant piles or sheet piles pre-
installed through the superficial gravels and made ground, followed by
shotcrete underpinning.
Figure 12.22 shows the settlements for shafts constructed using a jacked
caisson only. Data from Quigg (2019) for the initial jacked caisson part of
the shaft only has been added to the data from Faustin et al. (2018b).
Figure 12.23 shows the settlements for shafts constructed using a pre-
installed diaphragm wall or secant pile wall lining. Due to the high rigidity
of this type of shaft, as the ground moves towards the excavation it can
induce a small amount of heave close to the shaft wall. It was not pos-
sible to adequately fit a curve to this data. It should be borne in mind that
there will be settlements due to wall installation before excavation begins,
and these are of a similar magnitude to the settlements due to excavation
(Faustin et al., 2018b).
The values of α and β for the different types of shaft construction are sum-
marised in Table 12.13. As with any empirical relationship, consideration
Figure 12.21 Settlements caused by shaft construction – secant piles or sheet piles fol-
lowed by shotcrete underpinning.
Figure 12.22 Settlements caused by shaft construction – jacked caisson only.

Figure 12.23 Settlements caused by shaft construction – pre-installed secant piles or

diaphragm walls only.
has to be given to the case studies that have gone into producing it, the
variability or scatter of the data, and the applicability of the relationship to
any new situation.
All these case studies were in the London Basin. Unfortunately, case stud-
ies of shaft settlements from anywhere else in the world are quite scarce.
Table 12.13 Values of α and β found by curve-fitting for shaft types defined by Faustin
et al. (2018b).
Faustin et al. (2018b)
shaft types Description α β
‘SBE pre-installed’ Diaphragm wall or secant pile wall shaft N/A N/A
‘SBE jacked’ Precast concrete segments as a jacked 0.00035 1.7
caisson
‘EBS’ Precast concrete segments or shotcrete 0.00035 1.1
underpinning
‘Combined SBE & Pre-installed sheet pile or secant pile walls 0.00045 2.8
EBS’ through superficial gravels followed by
‘Combined SBE Precast concrete segments as a jacked 0.00045 1.4
jacked & EBS’ caisson through superficial gravels
followed by shotcrete lining in the
London Clay
Figure 12.24 Settlements due to diaphragm wall installation and shaft excavation in pre-
dominantly coarse-grained soils (data from Muramatsu & Abe, 1996).
Muramatsu & Abe (1996) presented a case study of a 28.2 m ID circular

diaphragm wall shaft in a coarse-grained soil with a maximum settlement
of 0.01%H. The diaphragm walls were 98 m deep and the excavation depth
was 60.3 m. Settlements were measured along three radial lines in differ-
ent directions. The excavation-induced settlements were very small, even
though the ground was very soft. Approximately half of the settlements
shown in Figure 12.24 were due to diaphragm wall installation. These
settlements are even smaller than most of those presented in Figure 12.23.
Based on what we know about shaft base heave stability in clay (c.f.
Section 4.2), and the relationship between undrained stability and volume
loss for tunnels demonstrated by the load factor approach in Section 12.2,
we can expect that the magnitude of settlements will be larger in clays that
have a lower undrained strength than London Clay, and will be smaller in
stronger clays. Unfortunately, at present, we do not have the field data to
confirm and quantify this.
Very few case studies of shaft construction include horizontal ground
movements. Muramatsu & Abe (1996) found that maximum horizontal
displacements 1.4 m from the shaft wall were between 6 and 10 mm, com-
pared to vertical settlements of between 4 and 7 mm. Schwamb et al. (2016)
installed instrumentation in and around a large diaphragm wall shaft in
East London, 29 m diameter and with an excavation depth of 73 m and a
diaphragm wall depth of 84 m. They also found that near-surface settle-
ments were between 4 and 7 mm, but unlike Muramatsu & Abe (1996), the
maximum horizontal movements at depth measured by inclinometers were
Figure 12.25 Horizontal displacements due to Heathrow Express Trial Tunnel Access

Shaft construction (redrawn from New & Bowers, 1994).
less than 4 mm, and at the surface were negligible. The differences may
have been due to the quality of the joints between diaphragm wall panels
or the stiffness of the ground.
New & Bowers (1994) measured horizontal ground surface displacements
along two radial lines ‘S’ and ‘T’, at the same locations as the vertical settle-
ments presented earlier in Figure 12.18. These are shown in Figure 12.25.
There are two anomalous displacements in the S line data, at 2.7 and 9.1 m,
that were due to the proximity of heavy plant movements and shallow exca-
vations for temporary works, so the T line data is more representative of the
greenfield situation.
At present there is insufficient empirical evidence to make good predic-
tions of horizontal ground movements due to shaft construction. More
measurements are needed. It is likely that numerical modelling would need
to be used if a shaft is to be constructed close to structures sensitive to
horizontal displacements, such as buildings, pipelines, sewers or tunnels.
WORKED EXAMPLE 12.9 SETTLEMENTS DUE TO

JACKED CAISSON SHAFT CONSTRUCTION
A 15 m diameter shaft is to be constructed using a jacked caisson seg-
mental lining to an excavation depth of 45 m.
Estimate the maximum settlement and the settlement at 10, 20 and

30 m from the shaft wall based on case studies of similar shafts.
Look at Figure 12.22 and Table 12.13. The best fit curve to the case study
data has parameters α = 0.00035 and β = 1.7. Using Equation 12.40 we get:
α.(β H − d ) 0.00035 (1.7 × 45 − d )

2 2
Sd = =
β H
2 1.7 2 × 45
At d = 0 the maximum settlement is Sd = α H = 0.00035 × 45 = 0.0158

m = 15.8 mm
At d = 10 m the settlement is given by:
0.00035 (1.7 × 45 − 10)

2
Sd = = 0.0119 m = 11.9 mm
1.7 2 × 45
Similarly, at d = 20 m the settlement is 8.6 mm, and at d = 30 m the set-
tlement is 5.8 mm. By putting Equation 12.40 into a spreadsheet, we can
plot a graph of the predicted settlement, as shown in Figure 12.26.
Figure 12.26 Worked Example 12.9 shaft settlements.
Further refinements to this methodology could be to take into account

the diameter of the shaft D. Newhouse (2018) cites a GCG report for
Crossrail that proposes the following relationship:
α
= 0.006%/m, D ≤ 25 m (12.41)
D
where for values of D greater than 25 m, a constant value of α = 0.15%

should be used.
Newhouse (2018) showed, with reference to the same case studies that
Faustin et al. (2018b) had access to, that Equation 12.41 is a conservative
upper bound to the data for underpinned shafts, but is very conservative
in relation to diaphragm wall or secant pile wall shafts. He then proposed
alternative values for α / D, depending on construction method. Although
the correlation to construction method is strong, the correlation between
shaft diameter and α is weak, and this may be because the range of shaft
depths and diameters is quite small for each construction method.
Newhouse (2018) then plotted all the data on a graph of α vs. D / H and
proposed design lines with slope Sv ,max / D. This method, however, does not
take into account the effect of construction method or shaft diameter on
the extent of settlements, represented by β in Equation 12.40.
An improvement could be to use Equation 12.40 with the recommended
values of α for the construction method, which determines the maximum
settlement at the shaft wall. Then we could relate β, which determines the
extent of the settlements as a multiple of shaft excavation depth H, to shaft
diameter. Setting β = 0.1D fits the data reasonably well, and means that a
shaft of 10 m diameter has an extent of settlement equal to H and a shaft
of 30 m diameter has an extent equal to 3H.
12.9 SUMMARY
Tunnelling will invariably cause ground movements, which cannot be pre-

dicted accurately in advance and can only be estimated.
This chapter has shown how to estimate transient and short-term surface
settlements and horizontal displacements using the Gaussian curve approx-
imation, both at the surface and subsurface.
The volume loss may be estimated using the load factor approach in clay
soils. In drained soils there is no analogous method, and an estimate must
be made based on case studies and what we know about the construction
method and the soil behaviour. If a face pressure is applied and we are
in dense sand then volume loss should be low and behaviour should be
approximately constant volume. In loose or medium sands, volume loss can
be significantly higher at the surface than at the tunnel.
The Gaussian curve approximation enables the calculation of ground
strain, slope and curvature. This will be needed in the following chapter on
building damage assessment.
Shafts will also cause ground movements. Ground movements due to shaft
construction generally follow a quadratic relationship with a maximum at
the shaft wall and settlements reducing to zero some distance away from the
shaft. The depth and diameter of the shaft are also important factors.
Unfortunately there are only sufficient case studies to develop empirical
relationships for shafts in the London Basin, which may not be applicable
elsewhere. Shafts where the support is provided before excavation, such
as secant pile or diaphragm wall shafts, cause very little ground move-
ment, whereas shafts where the support is provided after excavation, such
as underpinned shafts lined with precast concrete segments or shotcrete,

cause larger ground movements. The installation of secant piles or dia-
phragm walls will cause ground movements before excavation inside the
shaft has begun.
12.10 PROBLEMS
Q12.1. A 10 m diameter shotcrete-lined tunnel is to be excavated in stiff

clay at a depth to axis of 23 m. Assume the clay has a constant
value of undrained shear strength of 125 kPa and a bulk unit
weight of 20 kN/m3. Assume the unsupported length is 2 m.
i. Estimate the volume loss using the load factor approach.
ii. Estimate the trough width parameter K at the surface using
Table 12.1 and then calculate trough width i at the surface.
iii. Calculate maximum short-term surface settlement Smax using
the Gaussian curve equation.
iv. Calculate the short-term settlements at transverse offsets of 5,
10, 15, 20 and 25 m.
v. Assuming that 50% of the short-term settlement happens
ahead of the face and 50% behind the face, calculate settle-
ments along the tunnel centreline at 20, 15, 10, 5, −5, −10, −15
and −20 m relative to the face. Assume the tunnel is very far
from its start position.
Q12.2. A 5 m diameter EPB TBM tunnel is to be excavated in dense sand
at a depth to axis of 15 m. From prior experience of driving EPB
TBMs in these ground conditions, a conservative estimate of vol-
ume loss at the ground surface of 1% has been assumed.
i. Calculate the short-term value of maximum settlement Smax
and trough width i, for K = 0.25.
ii. Calculate at suitable intervals and then plot a graph of the
transverse short-term surface settlement trough.
iii. Calculate horizontal displacements at the same intervals and
then plot a graph, assuming constant volume behaviour.
iv. Calculate horizontal strains at the same intervals and then
plot a graph.
v. Calculate change in ground slope at the same intervals and
then plot a graph.
vi. Calculate ground curvature at the same intervals and then
plot a graph.
vii. Repeat (i) to (vi) for K = 0.5 and plot onto the same graphs.
What difference does it make?
Q12.3. A 7 m diameter EPB TBM tunnel is to be excavated in clay at
a depth to axis of 17 m. The expected value of minimum face
pressure at axis level is 100 kPa, with 100 kPa bentonite slurry
pressure also provided around the shield, which is 6 m long. The

undrained shear strength of the clay is given by cu = 25 + 3z , where
z is the depth below the ground surface, and the bulk unit weight
is 19 kN/m3.
i. Estimate the volume loss using the load factor approach, tak-
ing unsupported length P as the length of the shield.
ii. Calculate the face pressure required to limit the volume loss to 1%
iii. Using a trough width parameter K = 0.5, calculate trough
width i at the surface.
iv. Using a constant value of K = 0.5, calculate trough width i at
4, 8 and 12 m depth.
v. Using Mair et al. (1993)’s empirical relationship (Equation 12.18),
calculate trough width i at 0, 4, 8 and 12 m depth.
vi. Using the logarithmic empirical relationship (Equation 12.19),
calculate trough width i at 0, 4, 8 and 12 m depth.
Q12.4. An 8 m diameter open face TBM with a backhoe is to driven under
a railway embankment at a 45° skew. The embankment is 10 m
high and consists of non-cohesive fill with a bulk unit weight of 17
kN/m3. The natural ground below the embankment is stiff boulder
clay and the tunnel is 20 m below the natural ground surface level.
The undrained shear strength of the clay is given by cu = 50 + 8z,
where z is the depth below the natural ground surface, the shear
modulus G = 200cu and the bulk unit weight is 20 kN/m3.
i. Assuming trough width parameter K = 0.5 in the stiff boulder
clay and K = 0.25 in the fill and volume loss Vl = 1.5%, calcu-
late the surface settlements along the embankment centreline
when the tunnel face is at 20, 10, 0, −10 and −20 m from the
embankment centreline. Plot these five settlement troughs on
a graph against distance from the tunnel centreline along the
track chainage rather than transverse offset.
ii. Calculate the maximum change in track gradient for each of
the five settlement troughs.
iii. Estimate the volume loss at the face using Dimmock & Mair
(2007)’s load factor approach, assuming that excavation can
be no more than 1.5 m ahead of the front of the shield.
iv. The average annular overcut of the shield is 20 mm and the
length of the shield is 5 m. Will the ground close onto the
shield? What is the volume loss due to closure of the ground
around the shield?
Q12.5. A 15 m diameter shaft is to be constructed in stiff overconsoli-
dated clay by underpinning with precast concrete segments. The
final excavation depth is 30 m.
i. Estimate the extent of the settlement, i.e. the radial distance
from the shaft wall at which settlement is zero.
ii. Estimate the settlement adjacent to the shaft wall.

iii. Calculate the settlement profile at intervals of no more than
5 m from the shaft wall to the extent of settlement. Plot them
on a graph.
iv. A building is 11 m away from the shaft wall. At what excavation
depth would you expect settlement to begin affecting the building?
v. [Advanced] Referring back to Section 4.2 on base heave sta-
bility of shafts in clay, what factors would you expect to influ-
ence the magnitude of settlement in any type of clay soils?
REFERENCES
Atkinson, J. H. & Potts, D. M. (1977). Subsidence above shallow tunnels in soft

ground. Proc. ASCE Geot. Engrg Div. 103, GT4, 307–325.
Attewell, P. B. & Farmer, I. W. (1974). Ground deformations resulting from shield
tunnelling in London Clay. Canadian Geotechnical Journal 11, 380–395.
Attewell, P. B. & Woodman, J. P. (1982). Predicting the dynamics of ground settle-
ment and its derivatives caused by tunnelling in soil. Ground Engineering 15,
No. 8, 13–22 & 36.
Barratt, D. A. & Tyler, R. G. (1976). Measurements of ground movement and
lining behaviour on the London Underground at Regent’s Park, TRRL Lab
Report 684. Crowthorne, Berkshire: TRRL.
Bowers, K. H., Hiller, D. M., and New, B. M. (1996). Ground movement over
three years at the Heathrow Express Trial Tunnel. Proc. Int. Symp. on
Geotechnical Aspects of Underground Construction in Soft Ground,
pp. 557–562. Rotterdam: Balkema.
Burke, H. H. (1957). Garrison dam – tunnel test section investigation. J. Soil Mech.
Found. Div. A.S.C.E., 83, SM4, Paper No. 1438, 1–50.
Burland, J. B. (2001). Chapter 3: Assessment methods used in design. Building
Response to Tunnelling, Volume 1: Projects and Methods (eds Burland, J. B.,
Standing, J. R. & Jardine, F. M.), CIRIA Special Publication 200, pp. 23–43.
London: Thomas Telford Publishing.
Clayton, C. R. I., van der Berg, J. P. & Thomas, A. H. (2006). Monitoring and dis-
placements at Heathrow Express Terminal 4 station tunnels. Géotechnique
56, No. 5, 323–334.
Cooper, M. L., Chapman, D. N., Rogers, C. D. F. & Chan, A. H. C. (2002).
Movements in the Piccadilly Line tunnels due to the Heathrow Express con-
struction. Géotechnique 52, No. 4, 243–257.
Dimmock, P. S. & Mair, R. J. (2007). Estimating volume loss for open-face tun-
nels in London Clay. Proc. Instn Civ. Engrs – Geot. Engrg 160, GE1, 13–22.
Eden, W. J. & Bozozuk, M. (1969). Earth pressures on Ottawa outfall sewer tun-
nel. Canadian Geotech. J. 6, No. 17, 17–32.
Faustin, N. E., Mair, R. J., Elshafie, M. Z. E. B., Menkiti, C. O. & Black, M.
(2018a). Field measurements of ground movements associated with cir-
cular shaft construction. Proc. 9th Int. Symp. on Geotechnical Aspects of
Underground Construction in Soft Ground (eds Negro, A. & Cecilio Jr., M. O.),
IS-São Paulo, pp. 301–308. London: Taylor & Francis Group.
Faustin, N. E., Elshafie, M. Z. E. B. & Mair, R. J. (2018b). Case studies of circular

shaft construction in London. Proc. Instn Civ. Engrs – Geot. Engrg 171,
No. 5, 391–404.
Franza, A., Marshall, A. M. & Zhou, B. (2019). Greenfield tunnelling in sands: the
effects of soil density and relative depth. Géotechnique 69, No. 4, 297–307.
Franza, A., Marshall, A. M., Zhou, B., Shirlaw, N. & Boone, S. (2020). Discussion
of: Greenfield tunnelling in sands: the effects of soil density and relative
depth. Géotechnique 70, No. 7, 639–646.
Glossop, N. H. (1978). Ground movements caused by tunnelling in soft soils.
Ph.D. thesis, University of Durham.
Harris, D. I., Mair, R. J., Love, J. P., Taylor, R. N. & Henderson, T. O. (1994).
Observations of ground and structure movements for compensation grouting
during tunnel construction at Waterloo Station. Géotechnique 44, No. 4,
691–713.
Hill, N. & Stärk, A. (2015). Volume loss and long-term settlement at Kempton
Court, Whitechapel. Crossrail Project: Infrastructure design and construc-
tion – Volume 2(eds Black, M., Dodge, C. & Yu, J.), pp. 347–385. London:
ICE Publishing.
Hill, N. & Stärk, A. (2016). Long-term settlement following SCL-tunnel excava-
tion. Crossrail Project: Infrastructure design and construction – Volume 3
(ed. Black, M.), pp. 227–247. London: ICE Publishing.
Jones, B. D., Thomas, A. H., Hsu, Y. S. & Hilar, M. (2008). Evaluation of innova-
tive sprayed-concrete-lined tunnelling. Proc. Instn Civ. Engrs – Geot. Engrg
161, GE3, 137–149.
Jones, B. D. (2010). Low-volume-loss tunnelling for London Ring Main Extension.
Proc. Instn. Civ. Engrs – Geotech. Engrg 163, GE3, June, 167–185.
Kimura, T. & Mair, R. J. (1981). Centrifugal testing of model tunnels in soft clay. Proc.
10th Int. Conf. Soil Mech. & Found. Engrg, Stockholm, Vol. 1, pp. 319–322.
Lake, L. M., Rankin, W. J. & Hawley, J. (1996). Prediction and effects of ground
movements caused by tunnelling in soft ground beneath urban areas, CIRIA
Project Report 30. London: CIRIA.
Laver, R. G., Soga, K., Wright, P. & Jefferis, S. (2013). Permeability of aged grout
around tunnels in London. Géotechnique 63, No. 8, 651–660.
Laver, R., Li, Z. & Soga, K. (2017). Method to Evaluate the Long-Term Surface
Movements by Tunneling in London Clay. ASCE J. of Geotech. and Geoenv.
Engineering 143, No. 3, 1–7.
Macklin, S. R. (1999). The prediction of volume loss due to tunnelling in over-
consolidated clay based on heading geometry and stability number. Ground
Engineering, No. 4, 30–33.
Macklin, S. R. & Field, G. R. (1998). The response of London Clay to full-face
TBM tunnelling at West Ham, London. Proc. Int. Conf. on Urban Ground
Engineering, Hong Kong, November 1998.
Mair, R. J. (1979). Centrifuge modelling of tunnel construction in soft clay. Ph.D.
Mair, R. J. (2008). Tunnelling and geotechnics: new horizons. Géotechnique 58,
No. 9, 695–736.
Mair, R. J., Gunn, M. J. & O’Reilly, M. P. (1981). Ground movements around
shallow tunnels in soft clay. Proc. 10th Int. Conf. on Soil Mechanics and
Foundation Engrg, Stockholm, 1981, Vol. 1, pp. 323–328.
Mair, R. J. & Taylor, R. N. (1993). Prediction of clay behaviour around tunnels

using plasticity solutions. Predictive Soil Mechanics - Proc. of the Wroth
Memorial Symposium (eds Houlsby, G. T. & Schofield, A. N.), St Catherine’s
College, Oxford, 27th-29th July 1992, pp.449–463. London: Thomas Telford.
Mair, R. J., Taylor, R. N. & Bracegirdle, A. (1993). Subsurface settlement profiles
above tunnels in clays. Géotechnique 43, No. 2, 315–320.
Marshall, A. M. (2009). Tunnelling in sand and its effect on pipelines and piles.
Ph.D. thesis, University of Cambridge.
Marshall, A. M., Farrell, R., Klar, A. & Mair, R. (2012). Tunnels in sands: the effect
of size, depth and volume loss on greenfield displacements. Géotechnique 62,
No. 5, 385–399.
Muramatsu, M. & Abe, Y. (1996). Considerations in shaft excavation and periph-
eral ground deformation. Proc. Int. Symp. on Geotechnical Aspects of
Underground Construction in Soft Ground (eds Mair, R. J. & Taylor, R. N.),
London, UK, pp. 173–178. Rotterdam: Balkema.
New, B. M. & Bowers, K. H. (1994). Ground movement model validation at the
Heathrow Express trial tunnel. Tunnelling ’94, Proc. 7th Int. Symp. IMM
and BTS, London, UK, pp. 310–329. London: Chapman and Hall.
Newhouse, J. (2018). Ground movement due to shaft construction. Ground
Engineering, July 2018, 26–30.
Nyren, R. J. (1998). Field measurements above twin tunnels in clay. Ph.D. thesis,
Imperial College of Science, Technology and Medicine, London.
O’Reilly, M. P. (1988). Evaluating and predicting ground settlements caused by tun-
nelling in London Clay. Proc. Tunnelling ’88, pp. 231–241. London, UK: IMM.
O’Reilly, M. P. & New, B. M. (1982). Settlements above tunnels in the United
Kingdom – their magnitude and prediction. Proc. Tunnelling ’82, pp. 173–181.
London, UK: IMM
Peck, R. B. (1969). Deep excavations and tunnelling in soft ground. Proc. 7th Int.
Conf. Soil Mechanics and Foundation Engrg (7th ICSMFE), Mexico, State-
of-the-art report, pp. 225–290.
Pound, C. (2003). Session 1 Report: Prediction of damage to buildings and other
structures from tunnelling. Response of buildings to excavation induced
ground movements, Proc. Int. Conf. held at Imperial College, London, UK,
17th–18th July 2001 (ed. Jardine, F. M.), CIRIA Special Publication 199, pp.
27–36. London: CIRIA.
Quigg, G. (2019). Battersea cable tunnel: powering regeneration in central London,
UK. Proc. Instn Civ. Engrs – Civil Engrg 172, No. 2, 77–82, https://doi.
org/10.1680/jcien.18.00020.
Rankin, W. J. (1988). Ground movements resulting from urban tunnelling: predic-
tion and effects. Conf. on Engineering Geology of Underground Movements,
Nottingham, Geological Society Engineering Geology Special Publication
No. 5, pp. 79–92.
Sauer, G. & Lama, R. D. (1973). An application of New Austrian Tunnelling
Method in difficult builtover areas in Frankfurt/Main Metro. Proc. Symp. On
Rock Mechanics and Tunnelling Problems, Kurukshetra, India, December
1973, pp. 79–92. New Delhi: Indian Geotechnical Society.
Schmidt, B. (1969). Settlements and ground movements associated with tunnelling

in soils. Ph.D. thesis, University of Illinois, Urbana.
Schwamb, T., Elshafie, M. Z. E. B., Soga, K. & Mair, R. J. (2016). Considerations
for monitoring of deep circular excavations. Proc. Instn Civ. Engrs – Geot.
Engrg 169, GE6, 477–493.
Shirlaw, J. N., Doran, S. & Benjamin, B. (1988). A case study of two tunnels driven
in the Singapore ‘Boulder Bed’ and in grouted coral sands. Engineering
Geology of Underground Movements (eds Culshaw et al.), Geological
Society Engineering Geology Special Publications Volume 5, pp. 93–103.
Simic, D. & Craig, R. N. (1997). Lisbon Metro—settlement behaviour of large
diameter shield driven and NATM tunnels. Proc. Conf. Tunnelling ’97,
IMM, London.
Standing, J. R. & Selman, R. (2001). The response to tunnelling of existing tun-
nels at Waterloo and Westminster. Building Response to Tunnelling, Vol. 2
Case studies, CIRIA Special Publication 200, pp. 509–546. London: Thomas
Telford.
Umney, A. R. & Heath, G. R. (1996). Recorded settlements from the DLR tun-
nels to Bank. Proc. Int. Symp. on Geotechnical Aspects of Underground
Construction in Soft Ground (eds Mair, R. J. & Taylor, R. N.), London, UK,
Vu, M. N., Broere, W. & Bosch, J. (2016). Volume loss in shallow tunnelling.
Tunnelling and Underground Space Technology 59, 77–90.
Wan, M. S. P. & Standing, J. R. (2014). Field measurement by fully grouted vibrat-
ing wire piezometers. Proc. Instn Civ. Engrs – Geot. Engrg. 167, GE6,
547–564.
Wan, M. S. P., Standing, J. R., Potts, D. M. & Burland, J. B. (2017a). Measured
short-term ground surface response to EPBM tunnelling in London Clay.
Wan, M. S. P., Standing, J. R., Potts, D. M. & Burland, J. B. (2017b). Measured
short-term subsurface ground displacements from EPBM tunnelling in
London Clay. Géotechnique 67, No. 9, 748–779.
Wongsaroj, J., Soga, K. & Mair, R. J. (2007). Modelling of long-term ground
response to tunnelling under St James’s Park, London. Géotechnique 57,
No. 1, 75–90.
Wongsaroj, J., Soga, K. & Mair, R. J. (2013). Tunnelling-induced consolidation
settlements in London Clay. Géotechnique 63, No. 13, 1103–1115.
Yang, B. (2017). Numerical modelling of tunnelling in sand using a state param-
eter constitutive model. Ph.D. thesis, University of Warwick.
Chapter 13
Estimating building damage
Buildings may be affected by ground movements induced by tunnelling in

a variety of ways, and how they are affected will depend on the buildings
themselves. Reinforced concrete or steel frame buildings, with concrete
floor slabs and with or without infilled masonry wall panels, will behave
differently to houses consisting of load-bearing brick or stone walls and
timber floors. Both horizontal and vertical movements will affect build-
ings, as well as the change of slope of the ground and its curvature.
• the concept of limiting tensile strain and its correlation to building

damage categories
• the stages of building damage assessment
• deflection ratio and how to calculate it
• the idealisation of a building as a simple beam and how this helps
estimate maximum strains
• the combination of maximum bending and diagonal strains with hor-
izontal strain to determine the damage category
• the use of modification factors
• what may be included in a detailed Stage 3 assessment
• perform Stage 1, 2 and 2b building damage assessments for masonry

walls and frame structures
It is necessary to assess buildings along the route of a tunnel for the risk
of damage due to tunnelling settlements. This will usually follow a staged
process. At each stage buildings are classified into one of six damage catego-
ries, from 0 to 5, as shown in Table 13.1. Categories 0, 1 and 2 are aesthetic
damage, 3 and 4 are serviceability damage, and 5 is the most severe category
where stability of the building will be affected and it may collapse. At the pre-
liminary stage, very simple methods are used to conservatively assess the risk
of settlement damage to all the buildings, with the aim that many of them
DOI: 10.1201/9780429470387-13 467
Table 13.1 Classification system for visible damage to building walls (based on Burland
et al., 1977; Rankin, 1988; Burland, 2001).
Normal Limiting
Category degree of Description of typical damage (ease of tensile strain
of damage severity repair is in bold) and typical crack width ε lim (%)
0 Negligible Hairline cracks < 0.1 mm wide. 0–0.05
1 Very slight Fine cracks that are easily treated during 0.05–0.075
normal redecoration. Damage generally
restricted to internal wall finishes. Close
inspection may reveal some cracks in
external brickwork or masonry. Typical
cracks < 1 mm.
2 Slight Cracks easily filled. Redecoration probably 0.075–0.15
required. Recurrent cracks can be masked
by suitable linings. Cracks may be visible
externally and some repointing may be
required to ensure weather-tightness.
Doors and windows may stick slightly.
Typical cracks < 5 mm.
3 Moderate The cracks require some opening up and 0.15–0.3
can be patched by a mason. Repointing
of external brickwork and possibly a
small amount of brickwork to be
replaced. Doors and windows sticking.
Service pipes may fracture. Weather-
tightness often impaired. Typical cracks
5–15 mm or several cracks > 3 mm.
4 Severe Extensive repair work involving breaking- > 0.3
out and replacing sections of walls,
especially over doors and windows.
Windows and door frames distorted,
floor sloping noticeably. Walls leaning or
bulging noticeably, some loss of bearing
in beams. Service pipes disrupted. Typical
cracks 15–25 mm, but also depends on
the number of cracks.
5 Very severe This requires a major repair job involving > 0.3
partial or complete rebuilding. Beams
lose bearing, walls lean badly and
require shoring. Windows broken with
distortion. Danger of instability. Typical
cracks > 25 mm, but depends on the
number of cracks.
can be ruled out as being at low risk. The remaining buildings will proceed to
a second stage, where they will be analysed in more detail, again hoping that
many of them will be moved to the low risk category. Any remaining build-
ings will proceed to a third, more detailed stage. Usually, sensitive structures,
high rise buildings and heritage buildings will go straight to the third stage.
Estimating building damage 469
Structures may be considered sensitive if they contain sensitive equipment

or if even minor damage would be intolerable to the user/owner or for the
proper functioning of the structure (Rankin, 1988).
Building damage up to category 2 can be easily caused by a variety of
environmental phenomena, such as shrinkage or thermal effects on the
structure itself, or natural movements of the ground due to rising or low-
ering groundwater, tree root suctions, or other effects. This means that
identification of the cause of any category 1 or 2 damage is difficult and
could be a combination of causes, whereas category 3 damage is almost
certain to be associated with ground movements due to tunnelling if they
are occurring at that time. Therefore the division between category 2 and 3
is important (Burland, 2001). It is also the threshold beyond which repair
work starts to become expensive and it is common practice on tunnelling
projects to design underground construction methods and settlement miti-
gation measures so that all buildings are predicted to be in category 2 or
below and are hence ‘low risk’.
Building damage assessments usually follow a three stage process. Stage 1,
sometimes called ‘preliminary assessment’, Stage 2 or ‘second stage assess-
ment’ and Stage 3 or ‘detailed evaluation’ (Burland, 2001; Bowers & Moss,
2006). On major projects this process is often set out in the legislation used to
fund and enable the project to go ahead (e.g. for CTRL see Moss & Bowers,
2006, for Crossrail see Crossrail, 2008 or DeJong et al., 2019).
13.1 STAGE 1 ASSESSMENT
Stage 1, or ‘preliminary assessment’, is a very simple and conservative

approach. The aim is to identify buildings that are very likely to be in the low
risk category and to rule them out from further studies, so that the number of
buildings that require a more time-consuming Stage 2 assessment is reduced.
Two criteria are used, the predicted greenfield maximum slope and maxi-
mum settlement of the ground surface at the location of each building. If the
maximum slope is less than 1/500 and the maximum settlement is less than
10 mm, then the building has negligible risk of any damage (Rankin, 1988).
It is straightforward, for a given tunnel alignment, to plot contours of
surface settlement either side of the tunnel. Knowing the depth and diam-
eter of the tunnel, and estimating the value of maximum settlement Smax
and trough width i, the transverse offset y from the tunnel centreline to any
settlement contour value S may be found by rearranging Equation 12.1 to
obtain the following expression:
S 
y = 2i 2 .ln  max  (13.1)
 S 
y is the transverse offset from the tunnel centreline, i.e. the horizontal
distance perpendicular to the tunnel centreline, in m
i is the trough width, i.e. the horizontal distance from the tunnel cen-
treline to the point of inflexion of the Gaussian curve, in m
S is the settlement at transverse offset y in m
Smax is the maximum settlement, which occurs at the tunnel centreline,
in m
Equation 12.1 can also be differentiated to obtain an equation for slope

θ at any offset distance y (c.f. Equation 12.30 in Section 12.7 for the full
derivation):
dS y  − y2 
θ= = − 2 Smax exp  2  (13.2)
dy i  2i 
Equation 13.2 can only be solved for y iteratively, which can be achieved
using a solver or ‘goal seek’ function in a spreadsheet.
Equation 13.1 and Equation 13.2 can be used to define limits either side
of the tunnel on a set of plans of the alignment. All buildings that are out-
side both limits can then be excluded from further analysis.
WORKED EXAMPLE 13.1 STAGE 1

BUILDING DAMAGE ASSESSMENT
An 8 m diameter tunnel for a railway line is to be constructed using a
TBM through a dense urban area at 18 m depth to axis. The parameters
to be used for Stage 1 building damage assessment are a volume loss
Vl = 1.5% and a trough width parameter K = 0.5. Calculate the trans-
verse offset y at which surface settlement S = 10 mm. Calculate the
transverse offset y at which surface slope θ = 1/500.
First calculate volume loss Vs using:
π D2 π × 82
Vs = Vl = 0.015 × = 0.754 m 2
4 4
Now trough width i can be calculated at the surface (z = 0) using
Equation 12.3:
i = K ( z0 − z ) = 0.5 × (18 − 0) = 9 m
Smax can be calculated using Equation 12.2 rearranged:
Vs 0.754
Smax = = = 0.0334 m = 33.4 mm
2π .i 2π × 9
We can now calculate the transverse offset from the tunnel centreline to
the 10 mm settlement contour using Equation 13.1:
S   0.0334 
y = 2i 2 .ln  max  = 2 × 92 × ln  = 13.977 m
 S   0.01 
This can be checked by inserting this value for y into Equation 12.1 and
seeing that the settlement calculated is 10 mm.
A limiting slope of 1/500 = ± 0.002. Using Equation 13.2 in a spread-
sheet, values of slope θ can be calculated every 1 m and plotted on a graph
against transverse offset y, as shown in Figure 13.1. It can be seen that the
limiting slope value is exceeded in two discrete areas, and this is often the
case because the maximum and minimum slope values occur at the points
of inflexion and the slope is always zero above the tunnel centreline.
Figure 13.1 Worked Example 13.1 ground surface slope values.
In this case, since all buildings between y = −13.977 m and y = 13.977 m

will be exposed to a surface settlement greater than 10 mm, we are only
concerned with the outer limits of where the slope is greater than 1/500. By
using a goal seek function in a spreadsheet, this occurs at y = ±12.262 m.
This can be checked by inserting y = 12.262 into Equation 13.2:
y  − y2  12.262  −12.2622 
θ=− S exp   = − × 0.0334 × exp  = −0.002
i2
max
 2i 2  92  2 × 92 
Therefore, in this case the 10 mm settlement contour will be the limiting
criterion. All buildings between y = −13.977 m and y = 13.977 m will be
exposed to a surface settlement greater than 10 mm, and will need to
carried through to Stage 2 assessment.
In this stage, each building is assumed to deform to the greenfield settle-

ment profile at the level of the foundations. The deflection ratio is then
calculated in the hogging zone(s) and the sagging zone. The façade of the
building is then assumed to deform like a simple elastic beam under a point
load and this enables tensile strains to be calculated. These tensile strains
are then combined with the horizontal strains in the ground using the prin-
ciple of superposition to obtain resultant strains. These are then compared
to the limiting tensile strain values in Table 13.1 to determine the damage
category.
The deflection ratio is defined in Figure 13.2. The greenfield settlement
trough is an arbitrary one for illustrative purposes. The building, shown in
grey, is assumed to deform to the same shape as the greenfield settlement
trough. Between the points of inflexion the building is bending in a sagging
mode, and outside it is in a hogging mode.
The two parts of the building, of lengths Lhog and Lsag , are treated sepa-
rately and are known as ‘partitions’. In the hogging zone a straight line is
drawn from a point on the settlement trough at the edge of the building to
the point of inflexion. The maximum vertical distance between this line
and the settlement trough is the deflection ∆GFhog , where the superscript ‘GF’
indicates that this is the greenfield deflection. This deflection ∆GF
hog divided
by the length of the building in the hogging zone Lhog is the deflection ratio
DRhog
GF
. The same procedure is applied to the building partition in the sag-
ging zone. The equations are therefore:
∆GF
hog
DRhog
GF
= (13.3)
Lhog
Figure 13.2 Stage 2 building damage assessment – definition of deflection ratio.

GF =
∆GF
sag
DRsag (13.4)
Lsag
Note that a long building that spans beyond both points of inflexion may
have two hogging zones and therefore two values of hogging deflection
ratio will need to be calculated.
The maximum deflection values ∆GF hog and ∆ sag can be difficult to calculate and
GF
are not necessarily in the centre. They can be found graphically using a scale
ruler, or by using a spreadsheet to calculate the distance between the line and the
settlement trough curve at, for example, every 0.5 m along the length and then to
use the maximum of these values as the deflection. In practice, taking the value
at the centre of the zone will be approximately correct in most cases.
The next step is to impose these deflection ratios onto the building.
Burland & Wroth (1974) first proposed this should be done by modelling
the building as a simple beam under the action of a point load. This will not
give the exact deformation mode of the building, but this is only intended
to be an approximate method, with many gross simplifications.
Figure 13.3 shows how a bearing-wall building can be idealised as a
beam representing the load-bearing façade in bending and shear deforma-
tion according to the method of Burland & Wroth (1974).
Buildings will often have a height that is large relative to the ‘span’ in
our beam idealisation, at least compared to typical beams in structural
engineering. Therefore shear deflection may be significant and needs to be
added to the bending deflection. The midspan deflection of a centrally-
loaded simply-supported beam including both bending and shear deflection
is given by the following equation (Gere & Timoshenko, 1991: p.694; or
Bhatt, 1999: pp. 275–279):
PL3  72EI 
∆= 1+ 2 (13.5)

48EI  5L AG 
Δ is the midspan deflection of the beam in m

P is the applied line load in kN, uniformly applied across the width of
the beam
L is the length of the beam (for sagging this would be Lsag and for hog-
ging Lhog) in m
E is the Young’s modulus of the beam in kN/m 2
A is the shear area of the beam in m 2
I is the second moment of area in m4
G is the shear modulus of the beam in kN/m 2
The first term in Equation 13.5 is the deflection due to bending and the
second term is the deflection due to shear. Note that Equation 13.5 is dif-
ferent to the one proposed by Burland & Wroth (1974), which had 18 as
the coefficient of the second term in the brackets rather than 72/5. This
Figure 13.3 Stage 2 building damage assessment – idealisation of building as a beam

(based on Burland & Wroth, 1974).
is because they assumed that the shear deflection was equal to the sum of
the maximum shear strains along the centroid of the beam, but this is not
the case. In reality shear stress varies parabolically with the maximum
at the centroid and zero shear stress at the top and bottom of a rectangu-
lar cross-section, and shear strain varies with a cubic relationship. This is
called ‘warping’, and a ‘form factor for shear’ may be calculated to take
account of this, which for a rectangular section is equal to 6/5. Burland
& Wroth’s equation effectively assumes the form factor for shear is 3/2.
Gere & Timoshenko (1991) use the unit load method to derive Equation
13.5 and Bhatt (1999) uses a different method but arrives at the same
answer. Gere & Timoshenko’s method is given in Appendix C. This error in
Burland & Wroth’s deflection equation was first spotted by Netzel (2009).
The cross-sectional area A in Equation 13.5 is the shear area, and this is
not simply the building height multiplied by its width in the tunnel longitu-
dinal direction (unless the building consists of a solid isotropic rectangular
prism!). Burland & Wroth (1974) assume that the problem is plane strain
and the building properties may be expressed per unit width in the tunnel
longitudinal direction, and therefore that A = H in Equation 13.5, where
H is the height of the building from the foundation level to the eaves, and
second moment of area I is in units of m4/m. It is important to remember
this because when estimating the stiffness of more structurally complex
buildings with concrete slabs, columns, masonry infill walls, and other
structural features, the shear area and second moment of area will need to
be calculated in a more rigorous manner.
For masonry bearing-wall structures, a masonry façade perpendicular to
the tunnel alignment can be considered to act as a single plane stress rect-
angular section of length Lhog or Lsag , height H and of unit width. Note that
we are ignoring the effect of openings in the façade, such as windows and
doors. This is an unconservative assumption because stress concentrations
tend to occur near the corners of openings where cracks are more likely to
initiate and propagate from.
We need to rearrange Equation 13.5 so we can relate deflection ratio Δ/L
to extreme fibre bending strain, while removing the point load P from the
equation.
The idealised geometry of the beam is shown in Figure 13.4. The z axis
is out of the page. The beam has width in the z axis direction b, height in
the y axis direction H and length in the x axis direction L. A line load P is
applied at midspan.
Figure 13.4 Simply-supported rectangular beam with a centrally-applied line load.

The midspan bending moment is given by:

PL
M= (13.6)
4
M is the midspan bending moment in kNm
P is a centrally-applied line load in kN
L is the span of the beam in m
The extreme fibre bending stress is given by:

Md
σ bmax = (13.7)
I
σ bmax is the extreme fibre bending stress in kPa

d is the vertical distance from the neutral axis to the extreme fibre in m
I is the second moment of area of the beam in m4
PLd
σ bmax = (13.8)
4I
Therefore the extreme fibre bending strain may be given by:
σ bmax PLd
ε bmax = = (13.9)
E 4EI
ε bmax is the extreme fibre bending strain

E is the Young’s modulus of the beam in kPa
Rearranging Equation 13.9 for P gives:
4EI
P= ε bmax (13.10)
Ld
L2  72EI 
∆= 1+ 2 ε bmax (13.11)
12d  5L AG 

We want to calculate ε bmax from the deflection ratio we have already deter-
mined earlier, so rearranging Equation 13.11 gives:
∆ /L
ε bmax =
L  72EI  (13.12)
12d 1 + 5L2 AG 
In the sagging zone, the building’s neutral axis can be assumed to be at

mid-height, i.e. d = H /2. Also, assuming plane stress, i.e. a building of unit
width, such that shear area A = H and the second moment of area is also
per unit width, we get the following expression for second moment of area:
H3
Isag = (13.13)
12
For a masonry bearing-wall, rather than thinking of it as plane stress, we
can just as well use the thickness of the wall t. The shear area is then given
by A = tH and the second moment of area is given by Isag = tH 3 /12. When
inserted into Equation 13.12, thickness t cancels out and the resulting
equation is the same.
In the hogging zone, the building’s neutral axis is usually conservatively
assumed to be at the foundation level, i.e. d = H , because the ground-structure
interface may provide restraint (Burland & Wroth, 1974). Using the paral-
lel axis theorem, this gives:
H3
Ihog = (13.14)
3
It is strictly speaking impossible for the neutral axis to be at the foundation
level, because then the compressive stress required for equilibrium in bend-
ing is infinite (Netzel, 2009). However, many materials, such as masonry (or
steel fibre reinforced concrete – see Chapter 8) that have a very low tensile
strength relative to their compressive strength, will have a neutral axis very
close to the compressive extreme fibre in bending. Therefore, the assumption
is approximately correct in terms of estimating the strain distribution.
Now to calculate the maximum diagonal tensile strain ε dmax we first need
an expression relating it to the simply supported rectangular beam under
point load.
The shear strain is related to the shear force by the following expression:
αV
γ xy = (13.15)
AG
γ xy is the shear strain

α is the ratio of maximum shear stress to average shear stress, which
for a rectangular section is 3/2 (see Figure C.2 in Appendix C)
V is the shear force, where V = P / 2, in kN
A is the cross-sectional area, where A = bH , in m 2
G is the shear modulus in kPa
Substituting for V and α in Equation 13.15 gives:
3P
γ xy = (13.16)
4 AG
Figure 13.5 Definition of diagonal extension caused by shear strain.
Figure 13.5 defines diagonal extension, which we will call ∆ d . A rectangu-

lar element with sides of length dx and dy is deformed into a rhomboid by
the action of shear stresses on the sides τ xy and τ yx , where because of equi-
librium τ xy = τ yx. Angles and deformations are assumed to be small, and are
greatly exaggerated in the figure. The length of the sides of the element do
not change as we are considering pure shear.
To get the maximum diagonal tensile strain we use the following equation:
∆ d γ xy dycosθ
ε dmax = = (13.17)
ds ds
Now dy /ds = sinθ , therefore:
ε dmax = γ xy sinθ cosθ (13.18)
We also know that the direction θ of the maximum diagonal tensile strain
is 45°, because shear stresses τ xy = τ yx. Thus:
γ xy
ε dmax = (13.19)
2
And Equation 13.16 becomes:
3P
ε dmax = (13.20)
8AG
Rearranging Equation 13.20 for P gives:
8AG
P= ε dmax (13.21)
3
AGL3  72EI 
∆=
18EI 1 + 5L2 AG  ε dmax (13.22)
Similarly to the bending strain we want to calculate the maximum diagonal

tensile strain ε dmax from the deflection ratio we have already determined
earlier, so simplifying and rearranging Equation 13.22 gives:
∆ /L
ε dmax =
 AGL 2 4  (13.23)
 18EI + 5 
Poisson’s ratio v is often assumed to be equal to 0.3 for masonry (Burland

& Wroth, 1974). Although it doesn’t appear in Equation 13.12 or Equation
13.40, Poisson’s ratio is what relates Young’s modulus E and shear modulus
G in the following equation (e.g. Gere & Timoshenko, 1991):
E = 2G (1 + ν ) (13.24)
E is the Young’s modulus

G is the shear modulus
v is the Poisson’s ratio
Assuming ν = 0.3, then the ratio E / G, which appears in Equation 13.12 and
Equation 13.23, may be replaced with the number 2.6. The E / G ratio repre-
sents the relationship between bending and shear deformation of the building
as a whole. Remember that this does not take account of the openings in the
wall, we are considering it to be a solid rectangular prism, so the Poisson’s
ratio implied by the global E /G ratio value is the same as the Poisson’s ratio
of the material. If we were to use a different value of E /G ratio, for example
the suggested value of 12.5 for a frame structure (Burland & Wroth, 1974),
then the E/G ratio would cease to be directly related to the Poisson’s ratio of
the material. This will become important later in the chapter.
So now, from the hogging and sagging greenfield deflection ratios, we
can calculate the maximum bending tensile strain and the maximum diag-
onal tensile strain in the building in hogging or sagging (by applying the
appropriate value of second moment of area I). We now need to use the
principle of superposition to add in the effect of horizontal axial strain to
obtain a resultant bending strain and a resultant diagonal strain (Boscardin
& Cording, 1989; Mair et al., 1996a).
The expression for greenfield horizontal strain was derived in Section 12.7
and is given by:
 − S( max,z )   − y2   y2 
εh =   ⋅ exp  2  ⋅ 1 − 2  (13.25)
 z0 − z   2i   i 
ε h is the greenfield horizontal strain

S( max,z ) is the maximum settlement at depth z below the ground surface
in m
z is the depth below the ground surface in m
z0 is the depth to the tunnel axis in m
y is the transverse offset in m
i is the trough width in m
For Stage 2 assessment, it is usual practice to use the average horizontal strains
under the building’s foundation in the hogging and sagging zones, i.e. along
Lhog and Lsag. This was justified by Mair et al. (1996a and 1996b) by arguing
that these greenfield horizontal strains are applied to the building, where they
are added to tensile strains generated by shear and bending, and the precise
location of these strains is unknown. They also argued that the method is
effectively empirical and predicts building damage satisfactorily, also bearing
in mind that the horizontal strain induced in the building is in many cases con-
siderably less than the greenfield horizontal strain in the ground.
Therefore, Equation 13.25 should be used at, say, 1 m intervals along the
building partitions within the hogging or sagging zone to calculate hori-
zontal strains, which are then averaged within each zone to obtain the hog-
ging and sagging values of horizontal strain.
Since the maximum extreme fibre bending strain acts in the same direc-
tion as the horizontal axial strain, the resultant bending strain ε br is simply
given by:
ε br = ε bmax + ε h (13.26)
ε br is the resultant bending strain

ε bmax is the maximum extreme fibre bending tensile strain
ε h is the average horizontal strain along the building partition
Remember that in the sagging zone the horizontal strain will be compressive
and will therefore have a beneficial effect on the resultant bending strain, and
in the hogging zone it will be tensile and will have an adverse effect.
The resultant diagonal strain ε dr needs to be found using Mohr’s circle
of strain, because the maximum diagonal tensile strain and the horizontal
strain are not in the same direction and the direction and magnitude of the
resultant principal strain is therefore unknown.
Two systems of strain are superposed:
1. The horizontal (axial) strain ε h. This is in the beam’s x axis direction

(c.f. Figure 13.4), therefore we can say that ε x = ε h . In this system this
is the major principal strain, so γ xy = 0 and using Hooke’s Law we
also have ε y = −νε h.
2. The maximum diagonal tensile strain ε dmax . This is at 45° to the
beam’s x axis. This system is in pure shear, so the major principal
strain is ε dmax and the minor principal strain is −ε dmax . This system can
be represented by a shear strain γ xy = 2ε dmax.
The System 1 Mohr’s circle of strain is shown in Figure 13.6.

For System 1, the variation of strain with angle θ to the horizontal εθ ,h is
given by the average at the centre of its Mohr’s circle plus the radius multi-
plied by cos2θ :
ε h (1 − ν ) ε h (1 + ν )
εθ ,h = + cos2θ (13.27)
2 2
The System 2 Mohr’s circle of strain is shown in Figure 13.7.
For system 2, the variation of strain with angle ϕ from the direction of
principal strain ε dmax is given by the average at the centre of its Mohr’s circle
plus the radius multiplied by cos2φ :
εφ = ε dmax cos2φ (13.28)
Now since we know that the diagonal strain is acting at 45° to the horizon-
tal, we can say that:
φ = θ + 45° (13.29)
Figure 13.6 Mohr’s circle of strain for System 1: horizontal strain.

Figure 13.7 Mohr’s circle of strain for System 2: diagonal strain.
And therefore that:
cos2φ = cos ( 2θ + 90° ) = sin2θ (13.30)
So we can rewrite Equation 13.28 as:
εθ ,d = ε dmax sin2θ (13.31)
Superposition, i.e. adding together, of Equations 13.27 and 13.31 gives:
ε h (1 − ν ) ε h (1 + ν )
εθ = + cos2θ + ε dmax sin2θ (13.32)
2 2
Now we can draw a Mohr’s circle of strain for the superposed strains εθ ,
but we don’t yet know the direction or magnitude of the principal strain
when the two are combined. The Mohr’s circle is shown in Figure 13.8.
We can calculate the values of εθ at θ = 0, 45° and 90° using Equation 13.32,
and these are shown on Figure 13.8.
From Figure 13.8 we can see that the value of strain at the centre of the
circle is the average of the θ = 0 and θ = 90° values, and this is given in the
figure. If we can also calculate the radius of the circle, then we can find
the magnitude of the principal strain, which will be the resultant diagonal
strain ε dr that we are looking for.
Distance CB is the θ = 45° value of εθ minus the average strain, and dis-
tance CA is the θ = 0° value of εθ minus the average strain.
ε h (1 − ν ) ε (1 − ν )
CB = + ε dmax − h = ε dmax (13.33)
2 2
Figure 13.8 Mohr’s circle of strain for superposition of System 1 and System 2.
ε h (1 − ν ) ε h (1 + ν ) ε h (1 − ν ) ε h (1 + ν )
CA = + − = (13.34)
2 2 2 2
Triangles CBD and CAE are similar triangles, therefore the radius R of the
Mohr’s circle may be given by:
 ε (1 + ν ) 
2
+ ( ε dmax ) (13.35)
2
R = CB + CA =  h
2 2

 2 
The magnitude of the principal strain of the two superposed systems is

therefore the average strain plus the radius:
ε (1 − ν )  ε (1 + ν ) 
2
+ ( ε dmax ) (13.36)
2
ε dr = h +  h 
2  2 
Another way to derive the equation is to use the principal strain equation
for plane strain (e.g. Gere & Timoshenko, 1991: p.438), which also applies
to plane stress and is given by:
2 2
εx + εy  εx − εy   γ xy 
ε1,2 = ±   + (13.37)
2  2   2 
ε1,2 represents the major principal strain ε1 and the minor principal
strain ε 2
Since the diagonal strain is pure shear and the horizontal (axial)
strain is a principal strain (i.e. with no shear strain), we can say that
ε x = ε h , ε y = −νε h and γ xy / 2 = ε dmax, and insert these into Equation 13.37

such that:
ε (1 − ν )  ε (1 + ν ) 
2
+ ( ε dmax ) (13.38)
2
ε dr = h +  h 
2  2 
which is the same as Equation 13.36.

The maximum tensile value of either the resultant bending strain ε br from
Equation 13.26 or the resultant diagonal strain ε dr from Equation 13.36
will be used to determine the damage category for the building.
The maximum diagonal strain ε dmax and the maximum bending strain
ε bmax are not considered in combination with each other because they occur
at different locations. The maximum shear strain occurs close to the mid-
height of a rectangular beam and is zero at the extreme fibre, whereas the
maximum bending strain occurs at the extreme fibre. The distribution of
shear stress is discussed in Appendix C.
WORKED EXAMPLE 13.2 STAGE 2

This is the same tunnel as described in Worked Example 13.1 – Stage 1
building damage assessment.
to be used for Stage 2 building damage assessment are a volume loss Vl =
1.5% and a trough width parameter K = 0.5.
A row of terraced houses is situated with its front façade transverse to
the tunnel alignment, as shown in Figure 13.9. The façade is a solid dou-
ble skin brick wall, i.e. with a thickness of 9 inches or 23 cm. The height
from the foundation to the eaves is 5.6 m, and the total length is 31.5 m.
The settlement trough was calculated in Worked Example 13.1, where
it was found that the trough width i = 9 m, and the maximum settlement
Smax = 33.4 mm.
The building should be divided into partitions in the hogging and sagging
zones, and we will treat each partition as though they were separate build-
ings. This is shown in Figure 13.10, where Lhog = 16 m and Lsag = 15.5 m.
Hogging zone calculation

In the hogging zone, we can calculate the settlements at y = −25 m and at
y = −9 m and then calculate the values of settlement T caused by whole-
body tilt between those two points, i.e. we plot a straight line between them.
Figure 13.9 Worked Example 13.2 problem geometry for Stage 2 building damage
assessment.
Figure 13.10 Worked Example 13.2 hogging and sagging zones.
Then we can calculate the actual Gaussian curve settlements at 1 m inter-

vals, and for each one determine the deflection, i.e. the difference between
the actual settlement and the straight line value. This is shown in Table 13.2.
Table 13.2 also has values of horizontal strain ε h calculated using
Equation 13.25, which we will need later.
Table 13.2 Worked Example 13.2 calculation of hogging deflection.

Transverse Greenfield Settlement
offset settlement due to tilt Deflection Horizontal strain
y (mm) S (mm) T (mm) ∆ = S − T (mm) εh

-25 0.7055 0.7055 0 0.0002632
-24 0.9547 1.9284 -0.9737 0.0003241
-23 1.2761 3.1512 -1.8752 0.0003921
-22 1.6846 4.3741 -2.6895 0.0004656
-21 2.1968 5.5970 -3.4002 0.0005424
-20 2.8294 6.8198 -3.9904 0.0006191
-19 3.5996 8.0427 -4.4431 0.0006913
-18 4.5231 9.2655 -4.7424 0.0007539
-17 5.6139 10.4884 -4.8745 0.0008009
-16 6.8823 11.7113 -4.8289 0.0008261
-15 8.3338 12.9341 -4.6004 0.0008231
-14 9.9675 14.1570 -4.1895 0.0007862
-13 11.7752 15.3798 -3.6046 0.0007107
-12 13.7401 16.6027 -2.8626 0.0005937
-11 15.8361 17.8256 -1.9895 0.0004345
-10 18.0279 19.0484 -1.0205 0.0002349
-9 20.2713 20.2713 0 0.0000000
The maximum deflection in Table 13.2 is shown in grey shading. To

calculate the deflection ratio, we use Equation 13.3:
∆GF −4.8744536
DRhog
GF
=
hog
= = −0.000304653
Lhog 16000
The sign of the deflection ratio is usually ignored and the absolute value
is used.
Mair et al. (1996a) recommended cutting off the deflection ratio cal-
culation at a transverse offset of ±2.5i, presumably to avoid deflection
ratio being made smaller by considering a long building with a signifi-
cant portion of its length far away from the tunnel, experiencing rela-
tively little settlement. In this case, however, cutting off the calculation
at 2.5i (y = 22.5 m) would give a significantly lower value of deflection
ratio than the one calculated here.
Netzel (2009) also noted that considering the length of the building
outside the cut-off point often results in a larger deflection ratio. Also
if L /H is larger the maximum bending tensile strain will be larger (c.f.
Equation 13.12). In one example he shows that considering the length
of the building beyond the cut-off can increase maximum tensile strain
by 75%. However, Netzel (2009) did not consider that a longer building
will result in a lower average horizontal strain and potentially a lower
resultant strain. Usually, for a building that extends a long distance into
the hogging zone, the maximum resultant strain is at a maximum when
the building partition is cut-off somewhere between ±2.5i and ±3i.
We now need to know the properties of the building and idealise it as
a rectangular isotropic beam of unit width.
For brick or masonry, Burland & Wroth (1974) recommended using a
Poisson’s ratio of 0.3, such that E/G = 2.6.
The neutral axis in hogging is assumed to be at the foundation level of
the structure, and therefore the distance of the extreme fibre in tension
from the neutral axis d = H = 5.6 m.
The second moment of area in hogging is given by Equation 13.14:
H 3 5.63
I hog = = = 58.5387 m 4 /m
3 3
The cross-sectional area of the building modelled as a beam of unit width
is given by:
A = 1 × H = 1 × 5.6 = 5.6 m 2 /m
The maximum bending tensile strain may be calculated using

Equation 13.12:
∆ /L 0.000304653
ε bmax = = = 0.0005060 = 0.050%
L  72EI  16  72 × 58.5387 
1+ 2  1+ × 2.6 
12d  5L AG  12 × 5.6  5 × 162 × 5.6 
The maximum diagonal tensile strain may be calculated using

Equation 13.23:
∆ /L 0.000304653
ε dmax = = = 0.0002302 = 0.023%
 AGL 2 4   5.6 × 162 1 4
+ ⋅ +
 18EI 5   18 × 58.5387 2.6 5 
The average horizontal strain is the average of the values in the right-
hand column of Table 13.2, which is 0.0005448 or 0.054%.
The resultant bending strain is given by Equation 13.26:
ε br = ε bmax + ε h = 0.0005060 + 0.0005448 = 0.0010508 = 0.105%

The resultant diagonal strain is given by Equation 13.38:
ε (1 − ν )  ε (1 + ν )  0.0005448 (1 − 0.3)
2
+ ( ε dmax ) =
2
ε dr = h +  h 
2  2  2
 0.0005448 (1 + 0.3) 
2
+   + 0.0002302 = 0.0006130 = 0.061%

2
 2
The maximum of the resultant bending strain and the resultant diagonal
strain is used to determine the damage category. This is the resultant
bending strain, which is 0.105%. Referring to Table 13.1, the hogging
zone of this building is in damage category 2 – ‘slight’.
Sagging zone calculation
In the sagging zone, we can calculate the settlements at y = −9 m and at
y = 6.5 m and then calculate the values of settlement T caused by whole-
body tilt between those two points, i.e. we plot a straight line between them.
Then we can calculate the actual Gaussian curve settlements at 1 m inter-
vals, and for each one determine the deflection, i.e. the difference between
the actual settlement and the straight line value. This is shown in Table 13.3.
Table 13.3 Worked Example 13.2 Calculation of sagging deflection.

Transverse Greenfield Settlement Horizontal
offset settlement due to tilt Deflection strain
y S (mm) T (mm) ∆ = S − T (mm) εh

-9 20.2713 20.2713 0.0000 0.0000000
-8 22.5141 20.6247 1.8894 -0.0002625
-7 24.6984 20.9781 3.7202 -0.0005421
-6 26.7620 21.3315 5.4305 -0.0008260
-5 28.6423 21.6849 6.9574 -0.0011001
-4 30.2786 22.0384 8.2402 -0.0013499
-3 31.6156 22.3918 9.2238 -0.0015613
-2 32.6066 22.7452 9.8614 -0.0017220
-1 33.2160 23.0986 10.1174 -0.0018226
0 33.4217 23.4520 9.9697 -0.0018568
1 33.2160 23.8054 9.4106 -0.0018226
2 32.6066 24.1588 8.4478 -0.0017220
3 31.6156 24.5122 7.1033 -0.0015613
4 30.2786 24.8657 5.4129 -0.0013499
5 28.6423 25.2191 3.4232 -0.0011001
6 26.7620 25.5725 1.1895 -0.0008260
6.5 25.7492 25.7492 0.0000 -0.0006843
Table 13.3 also has values of horizontal strain calculated using

Equation 13.25, which we will need later.
The maximum deflection in Table 13.3 is highlighted in grey shading.
To calculate the deflection ratio, we use Equation 13.4:
∆GF 10.1174
GF =
DRsag
sag
= = 0.000652738
Lsag 15500
We now need to know the properties of the building and idealise it as a

rectangular isotropic beam of unit width.
We will again use a Poisson’s ratio of 0.3, such that E/G = 2.6 .
The neutral axis in sagging is assumed to be at the mid-height of the
structure, and therefore the distance of the extreme fibre in tension from
the neutral axis d = H /2 = 5.6/2 = 2.8 m.
The second moment of area in sagging is given by Equation 13.13:
H 3 5.63
I sag = = = 14.6347 m 4 /m
12 12
The cross-sectional area of the building modelled as a beam of unit width

is again given by:
A = 1 × H = 1 × 5.6 = 5.6 m
The maximum bending tensile strain may be calculated using
Equation 13.12:
∆ /L 0.000652738
ε bmax = = = 0.0010055 = 0.101%
L  72EI  15.5  72 × 14.6347 
1 + 1 + × 2.6
12d  5L2 AG  12 × 2.8  5 × 15.52 × 5.6 

The maximum diagonal tensile strain may be calculated using

Equation 13.23:
∆ /L 0.000652738
ε dmax = = = 0.0002361 = 0.024%
 AGL 2 4   5.6 × 15.52 1 4
+ ⋅ +
 18EI 5   18 × 14.6347 2.6 5 
The average horizontal strain is the average of the values in the right-
hand column of Table 13.3, which is –0.0011829 or –0.118%. Note
that it is negative because the horizontal ground strain is compressive
in the sagging zone.
The resultant bending strain is given by Equation 13.26:
ε br = ε bmax + ε h = 0.0010055 − 0.0011829 = −0.0001774 = −0.018%

ε h (1 − ν )  ε (1 + ν )  −0.0011829 (1 − 0.3)
2
 + ( ε dmax ) =
2
ε dr = +  h
2  2 2
 −0.0011829 (1 + 0.3) 
2
+   + 0.0002361 = 0.0003903 = 0.039%

2
 2
The maximum of the resultant bending strain and the resultant diagonal
strain is used to determine the damage category. The resultant bending
strain is compressive, so it is the resultant diagonal strain that is critical,
at 0.039%. Referring to Table 13.1, the sagging zone of this building is in
damage category 1 – ‘very slight’.
Since both the hogging and the sagging zone are in damage category
2 or below, the building can be considered at low risk of damage due to
settlement and no further analysis will be required.
13.2.1 Generic Stage 2 assessment

Harris & Franzius (2006) proposed a generic form of Stage 2 assessment as
a more efficient means of assessing large numbers of buildings. Major urban
tunnelling projects may have hundreds or even thousands of buildings that
progress from Stage 1 to Stage 2 and it is costly and time-consuming to
produce a Stage 2 assessment and report for each one. In this method, no
real buildings are considered but instead a wide range of generic building
lengths and positions are analysed.
Representative transverse sections are produced, each with a representa-
tive surface settlement trough. Along each section, a large number of build-
ing geometries are analysed using the standard Stage 2 assessment method
and the highest maximum tensile strain is determined.
This method is limited to buildings with foundations at or close to the
surface, since it is the surface settlement trough that is used in the analysis.
Harris & Franzius (2006) suggest a limit of z f /z0 < 0.2, where zf is the foun-
dation depth and z0 is the tunnel axis depth, which in most cases means a
single basement, or no basement, is acceptable. Buildings with two or more
basement levels or piled foundations have to be assessed separately.
Using this method, many ordinary buildings over smaller diameter bored
tunnels may be dismissed, and it is usually buildings close to large cav-
erns, for instance crossovers, turnouts or underground stations, that will
be taken forward for more detailed analysis.
13.3 STAGE 2B ASSESSMENT
A further refinement to the Stage 2 assessment may be added to filter out

even more buildings prior to the relatively expensive Stage 3, and I have
called this ‘Stage 2b’. It has been known, at least since the experiences
of the Jubilee Line Extension in London, that applying the greenfield
deflection ratios and horizontal strains to a building is very conservative,
particularly with respect to the horizontal strain (Viggiani & Standing,
2001), and particularly when dealing with buildings on continuous con-
crete strip or raft foundations. For example, a recent set of detailed field
measurements in Bologna, Italy by Farrell et al. (2014) showed that hor-
izontal strains measured in two buildings of different stiffnesses were
negligible, even though the settlements were quite large with a volume
loss of 5.1% causing a maximum settlement of approximately 200 mm.
Numerical modelling by Goh & Mair (2012) found that if a building’s
axial stiffness EA is greater than 105 kN/m, which is less than the stiff-
ness of most buildings, then the horizontal strain induced in the building
is effectively zero.
Based on finite element modelling of tunnelling in a greenfield situation
and of tunnelling under a building modelled as a simple beam, Potts &
Addenbrooke (1997) proposed modification factors to reduce the deflection
ratios and horizontal strains. These were defined as follows:
DRsag
M DRsag = (13.39)
DRsag
GF
DRhog
M DRhog = (13.40)
DRhog
GF
ε hc
M ε hc = (13.41)
ε hc
GF
ε ht
M ε ht = (13.42)
ε ht
GF
M DRsag and M DRhog are the modification factors for deflection ratio in
sagging and hogging, respectively
DRsag and DRhog are the deflection ratios at the foundation level of the
building in the sagging and hogging zone, respectively, including
the effect of the building on the ground movements
DRsagGF and DRGF are the greenfield deflection ratios of the ground at the
hog
foundation level of the building in the sagging and hogging zone,
respectively
M ε hc and M ε ht are the modification factors for horizontal compressive
strain and horizontal tensile strain, respectively
ε hc and ε ht are the maximum horizontal compressive and tensile strains,

respectively, at the foundation level of the building, including the
effect of the building on ground movements
ε hc
GF
and ε ht
GF
are the maximum greenfield horizontal compressive and
tensile strains, respectively, of the ground at the foundation level
of the building
The relative stiffness of the structure-soil system was characterised by two

relative stiffness factors, one for bending stiffness and one for axial stiffness.
By calculating the relative stiffness values for the structure-soil system, the
four modification factors could be read off from a design chart. They could
then be used to convert greenfield deflection ratios and greenfield horizontal
strains into values that take account of the structure-soil interaction.
In theory, the deflection ratio of a building will depend on both its bending
stiffness and its axial stiffness. However, Goh & Mair (2008) showed that when
the axial stiffness is high and in the range corresponding to actual buildings,
the modification factor for deflection ratio depends mainly on bending stiffness.
After performing a large number of further analyses taking into account
building length in the longitudinal direction, the effect of tunnel depth, the
influence of interface roughness between the building and the ground, the
number of floor slabs in the building, 3D tunnel construction and the effect of
building weight, Franzius et al. (2006a) modified the relative stiffness expres-
sions, and these are the ones we will use here. If you would like to see a compar-
ison of the relative stiffness methods of Potts & Addenbrooke (1997), Franzius
et al. (2006a) and Goh & Mair (2008), then refer to DeJong et al. (2019).
The relative bending stiffness is given by:
EI
ρmod
* = (13.43)
Es z0 B2 L
ρmod
* is the modified relative bending stiffness (modified compared
to the original proposed by Potts & Addenbrooke, 1997). It is
dimensionless.
EI is the bending stiffness of the building, where E is the Young’s mod-
ulus of the building material and I is the second moment of area
of the building idealised as a beam.
E s is the secant Young’s modulus of the soil at depth z0/2, at 0.01%
axial strain in a triaxial compression test.
z0 is the depth to the tunnel axis.
B is the total building width perpendicular to tunnelling.
L is the building length in the direction of tunnelling.
Note that the definitions of B and L are different to what we have used so
far in this chapter, but this is the notation used in Potts & Addenbrooke
(1997) and Franzius et al. (2006a).
The relative axial stiffness is given by:
EA
α mod
* = (13.44)
Es BL
α mod
* is the modified relative axial stiffness. It is dimensionless.
EA is the axial stiffness of the building, where E is the Young’s modu-
lus of the building material and A is the cross-sectional area of
the building structure.
L is the length of the building in the tunnel longitudinal direction.
Note that Goh & Mair (2012) suggested new expressions for the relative bend-
ing and axial stiffness that included the total length of the building in the trans-
verse direction B, such that the charts had only one design curve that was not
dependent on e /B. However, this was for deep multipropped excavations and it
is unclear whether this method would be appropriate for tunnels.
The design procedure for a Stage 2b assessment is as follows:
1. Predict greenfield settlement at the building foundation level using a

Gaussian curve with assumed values of volume loss and trough width.
2. From the Gaussian curve, calculate greenfield deflection ratios in sag-
ging and hogging (DRsag GF and DRGF ), and maximum greenfield hori-
hog
zontal tensile and compressive strains (ε hc GF
and ε ht
GF
).
3. Evaluate the EI and EA values for the building and obtain modifica-
tion factors from the design charts (Figures 13.11 and 13.12). The
e / B ratio is needed. Eccentricity e is the perpendicular horizontal dis-
tance from the centreline of the tunnel to the midpoint of the building
width B.
4. Multiply the greenfield values of deflection ratios and maximum hori-
zontal strains to obtain the modified values.
5. Use the modified values of deflection ratios and maximum horizon-
tal strains to calculate resultant diagonal and bending strains and
thereby obtain the damage category for the building using the limit-
ing tensile strains in Table 13.1. If it falls within damage category 0, 1
or 2 then it will not have to be carried forward to Stage 3 assessment.
A limitation of this approach is that the modelling by Potts & Addenbrooke

(1997) and Franzius et al. (2006a) used the same small strain stiffness
model for London Clay, with a target volume loss of 1.5%. It is not known
if these modification factors are appropriate for soils with different proper-
ties, e.g. drained cohesionless coarse-grained soils, or normally-consolidated
clays or silts, or when the tunnelling method achieves significantly smaller
or larger volume losses. Farrell et al. (2014) found that the design lines
of Franzius et al. (2006a) provided an upper bound to actual observed
modification factors for deflection ratio in hogging and sagging, both for
Figure 13.11 Design charts relating relative bending stiffness ρ mod

* to modification factor
in (a) sagging MDRsag and in (b) hogging MDRhog (Franzius et al., 2006a).
centrifuge model experiments in dry silica sand and for field measurements
in Bologna. The ground conditions in Bologna consisted of overconsolidated
highly stratified fluvial deposits with layers of silty clays and clayey silts,
interbedded with lenses of sandy silt and silty sand. As mentioned previously,
the volume loss for this 12 m diameter tunnel was very high at 5.1%.
Although compared to 3D finite element analyses of an advancing tunnel
in Franzius et al. (2006a) as well as 2D plane strain finite element analy-
ses, the modification factors only produce modified deflection ratios and
horizontal strains for after the tunnel has passed, i.e. the short-term plane
strain situation. As the tunnel face approaches, one would expect the nearer
parts of the building to move downwards and towards the face more than
the further away parts of the building. This will cause the building to twist,
Figure 13.12 Design charts relating relative axial stiffness α mod

* to modification factor in
(a) compression Mε hc and (b) tension M ε ht (Franzius et al., 2006a).
first towards the tunnel as it approaches, and then in the opposite direction
as it passes beyond (i.e. the incremental ground and structure movements
will be predominantly directed towards the tunnel face at every advance
step). This was investigated by Franzius et al. (2006b), where modification
factors for twist were proposed.
A further limitation is that the modelling by Potts & Addenbrooke
(1997) and Franzius et al. (2006a and 2006b) only examined buildings
that are square to the tunnel centreline, i.e. not skewed at an arbitrary
angle. In reality, most buildings will not be perpendicular or parallel to
the tunnel centreline. Skew should be expected to cause permanent twist
of the building. This is discussed to some degree by Franzius et al. (2006b)
and field data of skewed building twist can be found in Standing (2001)
for Elizabeth House in London, the data being later analysed by Franzius
(2004).
Giardina et al. (2018) performed a large parametric study comparing the
relative stiffness methods of Potts & Addenbrooke (1997), Franzius et al.
(2006a) and Goh & Mair (2012) for a large number of building configura-
tions, and found they gave very different values of modification factors in
both hogging and sagging, and hence very different predictions of strain.
Therefore these methods should be used with some caution.
At the end of the day, after all these sophisticated numerical and labo-
ratory models and field measurements have been used to calculate these
modification factors, we are going to apply them to greenfield deflection
ratio and horizontal strain. We will then use these modified values of
deflection ratio and horizontal strain in the simple beam model, with all
its limitations. There is no way that the simple beam model can calculate
the true strains in a building or accurately predict where or when damage
will occur. It is, however, a simple well-established method of assessing
risk that is very widely used and accepted. Perhaps one day, if enough field
measurements are available, we can move to a purely empirical method of
Stage 2 building damage assessment. Alternatively, with increasing com-
puter power it may be possible to undertake a generic Stage 2 assessment
using large numbers of numerical models.
WORKED EXAMPLE 13.3 STAGE 2B

building damage assessment, that we also used for Worked Example 13.2
– Stage 2 building damage assessment.
to be used for Stage 2 building damage assessment are a volume loss Vl =
1.5% and a trough width parameter K = 0.5.
A row of terraced houses is situated with its front façade transverse to
the tunnel alignment, as shown in Figure 13.9. The façade is a solid dou-
ble skin brick wall, i.e. with a thickness of 9 inches or 23 cm. The height
from the foundation to the eaves is 5.6 m, and the total length is 31.5 m.
The secant ground stiffness determined at 0.01% axial strain in a
triaxial compression test from a sample taken from a depth midway
between the surface and the tunnel axis is 100 MPa.
The Young’s modulus of the brick wall may be assumed to be 10 GPa
and Poisson’s ratio 0.3.
The greenfield hogging and sagging deflection ratios were calculated in

Worked Example 13.2 and had the following values:
∆GF
DRhog
GF
=
hog
= 0.000304653
Lhog
∆GF
GF =
DRsag
sag
= 0.000652738
Lsag

Using the design charts in Figure 13.11 requires the calculation of the
relative bending stiffness using Equation 13.43:
EI
ρ*mod =
Es z0 B2 L
For unit wall thickness, the length of the building in the tunnel longitu-
dinal direction L = 1.
The building’s length in the transverse direction B = 31.5 m , and we will
use I = I hog = 58.5387 m 4 /m that we calculated in Worked Example 13.2.
E I hog 10 × 103 58.5387

ρ*mod = ⋅ = ⋅ = 0.3278
Es z0 B2 100 18 × 31.52
We will use the lower chart in Figure 13.11, for which we need to know
the value of e /B. Now e is the transverse distance of the midpoint of the
building from the tunnel centreline, which is 9.25 m. Therefore:
e 9.25
= = 0.29
B 31.5
Reading off from the design chart gives M DRhog = 0.27.
Therefore the hogging deflection ratio may be calculated using
Equation 13.40:
DRhog = M DRhog × DRhog

GF
= 0.27 × 0.000304653 = 0.00008226
The maximum bending tensile strain and the maximum diagonal tensile
strain can now be calculated using the modified value of deflection ratio.
By inspection of the equations, they will be multiplied by 0.27 giving:
ε bmax = 0.27 × 0.0005060 = 0.0001366 = 0.014%
ε dmax = 0.27 × 0.0002302 = 0.00006215 = 0.006%

The horizontal strain also needs to modified, and first we need to calcu-
late the relative axial stiffness using Equation 13.44:
EA
α *mod =
Es BL
Again we assume unit wall thickness so A = H and L = 1, such that:
EA 10 × 103 5.6
α *mod = = ⋅ = 17.78
Es BL 100 31.5 × 1
Reading off from the lower design chart in Figure 13.12 gives M ε ht = 0.005.
This is applied to the maximum greenfield horizontal tensile strain
(remember that for the standard Stage 2 analysis we used the average
horizontal strain). The maximum greenfield horizontal tensile strain we
previously calculated in Table 13.2 is ε ht
GF
= 0.0008261.
ε ht = M ε ht × ε ht
GF
= 0.005 × 0.0008261 = 4.13 × 10−6 = 4.13 × 10−4 %
The resultant bending strain may now be calculated using Equation 13.26:
ε br = ε bmax + ε h = 0.0001366 + 4.13 × 10−6 = 0.000141 = 0.0141%
ε h (1 − ν )  ε (1 + ν )  4.13 × 10−6 (1 − 0.3)

2
+ ( ε dmax ) =
2
ε dr = +  h 
2  2  2
 4.13 × 10−6 (1 + 0.3) 

2
+   + 0.00006215 = 0.00006366 = 0.0064%

2
 2
The maximum tensile strain in the building is therefore the resultant bend-
ing strain of 0.0141%. Referring to Table 13.1, the hogging zone of this
building is in damage category 0 – ‘negligible’. In Worked Example 13.2,
without the modification factors, the damage category in the hogging
zone was 2 – ‘slight’, so this is a considerable improvement.

Everything is as for hogging, but we will use I = I sag = 14.6347 m 4 /m that
we calculated in Worked Example 13.2.
E I sag 10 × 103 14.6347

ρ*mod = ⋅ = ⋅ = 0.0819
Es z0 B2 100 18 × 31.52
We will use the upper chart in Figure 13.11, and as for the hogging cal-
culation, e /B = 0.29.
Reading off from the design chart gives M DRsag = 0.29.
Therefore the sagging deflection ratio may be calculated using
Equation 13.39:
DRsag = M DRsag × DRsag

GF = 0.29 × 0.000652738 = 0.000189294
The maximum bending tensile strain and the maximum diagonal tensile
strain can now be calculated using the modified value of deflection ratio.
By inspection of the equations, they will be multiplied by 0.29 giving:
ε bmax = 0.29 × 0.0010055 = 0.00029160 = 0.029%
ε dmax = 0.29 × 0.0002361 = 0.00006847 = 0.007%
The horizontal strain also needs to modified. The relative axial stiffness
is the same as for hogging, i.e. α *mod = 17.78.
Reading off from the upper design chart in Figure 13.12 gives
M ε hc = 0.002. This is applied to the maximum greenfield horizontal com-
pressive strain (remember that for the standard Stage 2 analysis we used
the average horizontal strain). The maximum greenfield horizontal com-
pressive strain we previously calculated in Table 13.3 is ε hc
GF
= −0.0018568.
ε hc = M ε hc × ε hc
GF
= 0.002 × ( −0.0018568) = −3.714 × 10−6 = −3.714 × 10−4%
The resultant bending strain may now be calculated using

Equation 13.26:
ε br = ε bmax + ε h = 0.00029160 − 3.714 × 10−6 = 0.0002879 = 0.029%
ε h (1 − ν )  ε (1 + ν )  ( −3.714 × 10−6 ) (1 − 0.3)

2
+ ( ε dmax ) =
2
ε dr = +  h 
2  2  2
 ( −3.714 × 10−6 ) (1 + 0.3) 

2
+   + 0.00006847 = 0.0000672 = 0.007%

2
 2 
The maximum tensile strain in the sagging zone of the building is therefore
the resultant bending strain of 0.029%. Referring to Table 13.1, the sag-
ging zone of this building is therefore in damage category 0 – ‘negligible’.
In Worked Example 13.2, without the modification factors, the dam-

age category in the sagging zone was 1 – ‘very slight’, so this is an
improvement.
Note that in the sagging zone, the horizontal strain is compressive and
is therefore beneficial, so using the modification factor for horizontal
strain will be reducing this beneficial effect.
13.3.1 Buildings with discrete pad foundations

So far we have assumed that the buildings have a continuous strip or raft
foundation in contact with the ground. It is not clear how buildings with
discrete footings will respond to ground movements induced by tunnelling.
Goh & Mair (2014) presented a case study of two 2-storey reinforced
concrete frame structures in Singapore, with individual footings founded
on short timber piles estimated to be 6 to 9 m long. They found that the
horizontal strains in the building were much higher than for structures with
continuous footings. In addition, the highest horizontal strains occurred
where the columns were not connected together at ground floor level.
Where ground beams were present between columns, horizontal strains
were much smaller. This was corroborated by finite element modelling.
13.3.2 Steel or reinforced concrete frame buildings

So far we have only considered buildings whose response is dominated by
load-bearing wall behaviour, which are typically masonry houses. Another
common form is a frame structure. These typically consist of columns and
beams made from steel or reinforced concrete, with reinforced concrete floor
slabs. The façades may have brick infill or cladding. Therefore, the building
stiffness predominantly comes from the floor slabs, rather than the walls.
The same procedure may be applied as for masonry walls, but when the
axial and bending stiffness are calculated, the floor slabs are considered.
We need to find a way to add together the stiffness of various structural
elements to obtain the stiffness of the building as a whole.
Potts & Addenbrooke (1997) suggested that the bending stiffness of a
framed structure may be estimated by using the parallel axis theorem, namely:
n +1
∑(I
( EI )structure = E
i =1
slab + Aslab dslab
2
) (13.45)
( EI )structure is the bending stiffness of the overall structure modelled as

a single beam
Islab is the second moment of area of an individual slab i
Aslab is the cross-sectional area of an individual slab i

dslab is the vertical distance from the overall structure’s neutral axis to
the individual slab’s neutral axis
n is the number of storeys, such that the building has n + 1 slabs
The axial stiffness is likewise given by:

n +1
( EA)structure = E∑ ( Aslab ) (13.46)

i =1
( EA)structure is the axial stiffness of the overall structure modelled as a

single beam
Potts & Addenbrooke (1997) acknowledged that the parallel axis theo-
rem would tend to overestimate stiffness, as it is assumed that the slabs
act compositely with transfer of shear between each slab. However, it is
unlikely that there would be sufficient shear walls and moment connec-
tions between columns and beams to make this happen, except perhaps in
a partial manner.
In making predictions of building response for Elizabeth House,
Waterloo, during tunnelling for the Jubilee Line Extension, Mair & Taylor
(2001) assumed that the floor slabs and 1.4 m thick raft foundation all
contributed to overall bending stiffness, but did not act compositely. They
assumed that all the slabs, including the raft foundation, were the same
thickness, i.e. 0.3 m thick, using the following equation:
n +1
( EI )structure = E∑ ( Islab ) (13.47)

i =1
Since this was a 10-storey structure with two basement levels, it probably
seemed unreasonable to use all the slabs in the building for the axial stiffness,
so only the lower basement, basement and ground floor slabs were included
with 1.4 m, 0.3 m and 0.3 m thickness, respectively (though the reason for
doing this was never explained by Mair & Taylor, 2001). The predictions
agreed well with actual measured building movements (Standing, 2001).
Goh & Mair (2014) tried both methods of estimating the bending stiff-
ness of frame structures modelled using finite element analysis and com-
pared the computed modification factors to the simple beam method, as
shown in Figure 13.13. Using the parallel axis theorem (Equation 13.46)
significantly underestimated the modification factor, while summing the
bending stiffness (Equation 13.47) gave a realistic upper bound estimate of
modification factor.
Based on finite element modelling, Goh & Mair (2014) also proposed
correction factors for the stiffening effect of columns on beams, which can
Figure 13.13 Modification factor from a simple beam model compared with finite ele-
ment modelling of various configurations of frame structure with bending
stiffness calculated either using the summation method or the parallel axis
theorem (redrawn from Goh & Mair, 2014).
be used to make a better estimate of the modification factor when using the
simple beam model. Meyerhof (1953) put forward an approach to estimate
the bending stiffness of a building taking account of columns, infill panels
and load-bearing walls without openings. Melis & Rodriguez Ortiz (2001)
developed a method to take account of the contribution of walls and their
proportion of openings, floors, and foundation slabs with either a rigid or
hinged connection to the superstructure. The second moment of area is cal-
culated relative to the neutral axis of the basement. Giardina et al. (2018)
provided a useful review of all these methods.
In summary, these methods are ways of estimating the stiffness of a
framed building, i.e. the second moment of area I and the cross-sectional
area A, for use in either a Stage 2 or 2b analysis:
• In the case of a Stage 2 analysis, the calculated values of second

moment of area I and cross-sectional area A are assigned to the beam,
and the greenfield deflection ratio and horizontal strain are imposed
to obtain a value of maximum tensile strain, which can then be used
to determine the damage category.
• In a Stage 2b analysis, the stiffness of the building is one of the
inputs needed to calculate the relative stiffness and hence obtain the
modification factors. Then, the greenfield deflection ratio and hori-

zontal strain, multiplied by the modification factors, are applied to
the beam to obtain a value of maximum tensile strain. The second
moment of area I and cross-sectional area A of the beam will be the
same as that determined in the relative stiffness calculation.
In both types of analysis, we also need to know the position of the neu-
tral axis in order to use the bending strain equation (c.f. Equation 13.12),
for both hogging and sagging. Since reinforced concrete or steel frame
structures can support tension forces as well as compression forces (unlike
masonry), then it seems sensible that the neutral axis should be the neutral
axis of the structure, which is likely to be close to the mid-height, or a bit
below, since basement and ground floor slabs are likely to be thicker. Often
it is assumed to be at mid-height because of the difficulty in accounting for
all structural details. Unlike for masonry walls, for frame structures the
same neutral axis position should be used in both hogging and sagging.
Yet another aspect to bear in mind is that a frame structure cannot be
assumed to be isotropic in the way we assumed a load-bearing masonry wall
was isotropic, as it is an assembly of structural elements acting together in
a partially composite manner. Therefore, the E /G ratio will not be directly
dependent on the Poisson’s ratio. Burland & Wroth (1974) argued that
because a frame structure has a relatively low stiffness in shear and a reason-
able degree of tensile restraint, then an E /G ratio of 12.5 should be used.
This has a large influence on the maximum tensile strains calculated using
Equation 13.12 and Equation 13.23. In most cases the maximum diagonal
tensile strain will be larger than the maximum bending tensile strain and
they will both be smaller compared to values obtained when E/G = 2.6.
It shouldn’t be assumed that the behaviour of a framed building will be
the same as a simple beam or that the damage to a framed building can be
described using this method (Giardina et al., 2018). It is a very simple method
used to make a conservative estimate of the level of risk of potential damage.
WORKED EXAMPLE 13.4 STAGE 2B BUILDING

DAMAGE ASSESSMENT FOR A FRAME STRUCTURE
building damage assessment, that we also used for Worked Example 13.2 –
Stage 2 building damage assessment and Worked Example 13.3 – Stage
2b building damage assessment.
to be used for Stage 2b building damage assessment are a volume loss

Vl = 1.5% and a trough width parameter K = 0.5.
A reinforced concrete frame building is over the tunnel alignment, as
shown in Figure 13.14. It has 4 storeys with no basement. All the storeys
are 3.5 m high. The ground floor has a reinforced concrete raft founda-
tion 1.0 m thick. The other floor slabs are all 0.3 m thick.
The secant ground stiffness determined at 0.01% axial strain in a
triaxial compression test from a sample taken from a depth midway
between the surface and the tunnel axis is 100 MPa.
The Young’s modulus of the concrete is 30 GPa and Poisson’s ratio 0.2.
Figure 13.14 Worked Example 13.4 – Cross-section through building and tunnel.
Since the building has the same length and position relative to the tun-
nel as Worked Example 13.2, the greenfield hogging and sagging deflec-
tion ratios were calculated previously and had the following values:
∆GF
DRhog
GF
=
hog
= 0.000304653
Lhog
∆GF
GF =
DRsag
sag
= 0.000652738
Lsag
We will assume that the building stiffness depends only on the reinforced
concrete slabs, and ignore the contribution of columns, shear walls and
other structural features.
Assuming that the overall second moment of area of the structure may
be approximated by summing the individual slabs, using Equation 13.47
for a unit width gives:
n +1
∑(I ) = 1 ×12
1.0  1 × 0.33 
3
I structure = + 4 = 0.0833 + 0.0090 = 0.0923 m 4 /m
slab
 12 
i =1
Note that the contribution of the raft is much more than the sum of the
four floor slabs above. Comparing this value of second moment of area
with those calculated in Worked Example 13.2, this value is much lower
than for a solid brick wall and here we are not calculating separate values
of second moment of area for hogging and sagging.
We also need to know the cross-sectional area, again per unit
width. We will make the same assumption as Mair & Taylor (2001)
and count only the slabs at ground level or below, which in our case
means just the raft:
A = 1.0 × 1.0 = 1.0 m 2 /m

Using the design charts in Figure 13.11 requires the calculation of the
relative bending stiffness using Equation 13.43:
EI
ρ*mod =
Es z0 B2 L
For unit thickness, the length of the building in the tunnel longitudinal
direction L = 1.
The building’s length in the transverse direction B = 31.5 m, so we get:
E I structure 30 × 103 0.0923

ρ*mod = ⋅ = ⋅ = 1.550 × 10−3
Es z0 B2 100 18 × 31.52
We will use the charts in Figure 13.11, for which we need to know the
value of e /B. Now e is the transverse distance of the midpoint of the
building from the tunnel centreline, which is 9.25 m. Therefore:
e 9.25
= = 0.29
B 31.5
Reading off from the design charts gives M DRhog = 1.26 and M DRsag = 1.02.
The hogging deflection ratio may be calculated using Equation 13.40:
DRhog = M DRhog × DRhog

GF
= 1.26 × 0.000304653 = 3.839 × 10−4
And the sagging deflection ratio may be calculated using Equation 13.39:
DRsag = M DRsag × DRsag

GF = 1.02 × 0.000652738 = 6.658 × 10−4
The distance from the structure’s neutral axis to the extreme fibre
in bending is assumed to be the distance from the overall structure’s
neutral axis to the top floor slab in hogging and to the ground floor
slab in sagging.
To find the height of the neutral axis above ground level dsag, take
moments of area about ground level:
0.3 × 3.5 + 0.3 × 7 + 0.3 × 10.5 + 0.3 × 14 = (1 + 4 × 0.3) dsag
∴ dsag = 4.77 m
Now the vertical distance from the neutral axis to the extreme fibre in
hogging will be given by:
dhog = H − dsag = 14 − 4.77 = 9.23 m
Similarly, the relative axial stiffness is calculated using Equation 13.44:
EA
α *mod =
Es BL
Again we assume unit thickness L = 1, such that:
EA 30 × 103 1
α *mod = = ⋅ = 9.524
Es BL 100 31.5 × 1
Reading off from the design charts in Figure 13.12 gives M ε ht = 0.006
and M ε hc = 0.002. These modification factors will be applied to the maxi-
mum greenfield horizontal strain in the hogging and sagging zone.
The modified horizontal strain in the hogging zone is given by:
ε ht = M ε ht × ε ht
GF
= 0.006 × 0.0008261 = 4.957 × 10−6 = 4.957 × 10−4%
And the modified horizontal strain in the sagging zone is given by:
ε hc = M ε hc × ε hc
GF
= 0.002 × ( −0.0018568) = −3.714 × 10−6 = −3.714 × 10−4%

The maximum bending tensile strain and the maximum diagonal
tensile strain can now be calculated using the modified value of
deflection ratio. Taking E/G = 12.5, the maximum bending tensile

strain in hogging is given by:
DRhog 3.839 × 10−4

ε bmax = =
Lhog  72EI  16  72 × 0.0923 
12dhog 1 + 5L2 AG  12 × 9.23 1 + 5 × 162 × 1.0 × 12.5
= 0.002496 = 0.250%
The maximum diagonal tensile strain in hogging is given by:
DRhog 3.839 × 10−4

ε dmax = = = 0.000029 = 0.003%
 AGLhog 2 4   1.0 × 162 1 4
 18EI + ⋅ +
 5   18 × 0.0923 12.5 5 
ε br = ε bmax + ε ht = 0.002496 + 4.957 × 10−6 = 0.002501 = 0.250%
ε ht (1 − ν )  ε (1 + ν ) 
2
 + ( ε dmax )
2
ε dr = +  ht
2  2
It may be unclear what the value of Poisson’s ratio v should be – the value
for concrete or the value for the structure as a whole. If we calculate
the value for the structure as a whole, then based on E/G = 12.5 we get
ν = 5.25. A typical value for concrete is 0.2.
Although maximum diagonal tensile strain ε dmax and horizontal strain
ε ht have been calculated based on the strain distribution of the structure
as a whole, when we are calculating the resultant we are effectively super-
posing the strains at a single point in the structure so we should use the
Poisson’s ratio of the material, which is 0.2 in this case.
ε (1 − ν )  ε (1 + ν )  4.957 × 10−6 (1 − 0.2)

2
+ ( ε dmax ) =
2
ε dr = ht +  ht 
2  2  2
 4.957 × 10−6 (1 + 0.2) 

2
+   + 0.000029 = 0.00003113 = 0.003%

2
 2
The maximum tensile strain in the building is therefore the resultant

bending strain of 0.250%. Referring to Table 13.1, the hogging zone of
this building is in damage category 3 – ‘moderate’.

Taking E/G = 12.5, the maximum bending tensile strain in sagging is
given by:
DRsag 6.658 × 10−4

ε bmax = = = 0.002228 = 0.223%
Lsag  72EI  16  72 × 0.0923 
12dsag 1 + 5L2 AG  12 × 4.77 1 + 5 × 15.52 × 1.0 × 12.5
   
The maximum diagonal tensile strain in sagging is given by:
DRsag 6.658 × 10−4

ε dmax = = = 0.000054 = 0.005%
 AGLsag 2
4   1.0 × 15.52 1 4
 18EI + 5   18 × 0.0923 ⋅ 12.5 + 5 
 
ε br = ε bmax + ε hc = 0.002228 + −3.714 × 10−6 = 0.002224 = 0.222%
ε (1 − ν )  ε (1 + ν )  −3.714 × 10−6 (1 − 0.2)

2
+ ( ε dmax ) =
2
ε dr = hc +  hc 
2  2  2
 −3.714 × 10−6 (1 + 0.2) 

2
+   + 0.000054 = 0.00005256 = 0.005%

2
 2
The maximum tensile strain in the building is therefore the resultant

bending strain of 0.222%. Referring to Table 13.1, the sagging zone par-
tition of this building is in damage category 3 – ‘moderate’.
We categorise the building according to the worst damage category
of the hogging and sagging zone partitions. Therefore, the building is in
damage category 3 – ‘moderate’ and needs to be taken forward to Stage 3
assessment for more detailed survey and analysis.
A Stage 3 assessment is a refinement of the preceding analysis, and must

take account of the individual building’s structural details and the tun-
nelling method (Burland, 2001), as opposed to Stage 2, which may be
approached generically (Harris & Franzius, 2006). Sometimes Stage 3 is
approached as a series of sub-steps, each sub-step increasing the sophistica-
tion of the building and tunnel model (Crossrail, 2008). It is assumed that
at each sub-step the maximum tensile strain will decrease and the damage
category may also decrease. If the damage category reduces to 2 or below,

then the assessment is stopped.
If deemed necessary, a building condition survey is undertaken by a
structural engineer to determine the structural form and condition of the
building (Crossrail, 2008). They will look for evidence of previous move-
ments and damage, the type of foundations, and the form of the structure.
Buildings that have already suffered previous damage may be less able to
cope with additional movements. Buildings on continuous foundations
such as strip footings or concrete raft foundations will be less vulnerable to
differential movements. Buildings with structural continuity such as steel
or reinforced concrete frame structures are less likely to suffer damage than
load-bearing brick or masonry wall structures (Burland, 2001).
Giardina et al. (2020) defined four main areas where the tunnel-soil-building
model may be refined: excavation, propagation of displacements from the
tunnel boundary to the ground surface, the effect of soil-structure interac-
tion, and the vulnerability of the building to the imposed displacements.
13.4.1 Tunnel excavation

Up to now, we have imposed empirical greenfield settlements on the build-
ing. By looking into the tunnel excavation method in detail, a lower value
of volume loss may be found to be justified. Or perhaps by looking at the
particular ground conditions at the location of the building, a lower value
of volume loss or a larger trough width may be justified.
If a numerical model is used, then the geometry of the problem may be
made more realistic by including soil properties that vary with stratum and
with depth, by modelling the depth and skew of the building’s foundation
relative to the tunnel, and by including details of the tunnel excavation,
such as face pressure, shield gap or grout pressure. Refer to Chapter 7, on
numerical modelling, for more details.
13.4.2 Soil-structure interaction

As we have seen in Stage 2b, where finite element modelling by Potts &
Addenbrooke (1997) and Franzius et al. (2006a) was used to determine
modification factors to take account of soil-structure interaction, a numeri-
cal model that includes the building could be used to calculate building
strains and thereby determine the damage category. Usually the building
is modelled as a beam in 2D or using shell elements in 3D with equivalent
elastic properties (Goh & Mair, 2012). This enables the interaction of the
bending, shear and axial stiffness of the building with the ground settle-
ments to be taken into account, as well as the building’s self weight.
In a slight variation, Pickhaver et al. (2010) used a mesh of equivalent
elastic Timoshenko beams to model the walls of a building in a 3D finite
element analysis, including internal walls. They found that, compared to
nonlinear numerical models of masonry panels with and without openings,
the beams replicated the behaviour well in sagging, but required the use of
a reduced flexural stiffness to replicate behaviour in hogging.
Using a centrifuge model with a building modelled by two layers of tiny
mortared bricks, Farrell & Mair (2012) found that the horizontal strain
transferred to buildings through their foundations is negligible, as has been
found by others. When the results were compared to the modification fac-
tors produced by Franzius et al. (2006a), agreement was poor but the modi-
fication factors did provide an upper bound. This means that buildings that
proceed from Stage 2b analysis are likely to have reduced strains predicted
in Stage 3 analyses when more sophisticated methods are used.
Farrell & Mair (2012) also found that model buildings that respond rig-
idly do so by redistributing their weight. Therefore, corners or edges of the
building may embed into the ground, as has been found during monitoring
of real buildings affected by tunnelling (Farrell et al., 2014), and by centri-
fuge models of buildings on sand by Ritter et al. (2017). Although this is
likely to reduce building distortion, it could cause serious problems for util-
ity connections entering the building, which may experience a rapid change
of displacement between the adjacent ground and the building.
In their numerical analyses with the building modelled by an elastic
beam, Goh & Mair (2012) noted that when a building’s axial stiffness is
low, the transfer of horizontal strains from the ground to the building is
determined by the interface properties between the ground and the build-
ing in the model. When the interface coefficient was 0.5, the modification
factor also tended towards 0.5 as building axial stiffness decreased. If the
interface coefficient were 1.0, meaning that the nodes of the building ele-
ments in the model were tied rigidly to the ground elements, then the modi-
fication factor would tend towards 1.0.
13.4.3 Modelling the building

There are a large number of numerical modelling studies of building
response to settlement, but since very few of them have been compared to
real or scale model buildings, they will not all be discussed here.
Giardina et al. (2012) built a physical 1/10th scale model of a masonry
façade and applied a hogging settlement profile to its base. They found that
the presence of door and window openings had an influence on the crack
distribution. Corners of openings create stress concentrations and tend to
be where cracks initiate. This cannot be taken into account in the Stage
2 analysis, and a building façade with large openings may end up with a
higher damage category when modelled in more detail in Stage 3. Giardina
et al. (2013) built a 2D continuum numerical model of the physical scale
model in Giardina et al. (2012), and found it was able to accurately repro-
duce the façade deformation and cracking pattern. For masonry buildings,
therefore, it is of crucial importance to model the initiation and propaga-
tion of cracking in the structure.
Giardina et al. (2019) presented the results of physical and numerical

modelling of another masonry façade, this time of a 1/20th scale palace
in Brescia, Italy. Two types of numerical models were compared with the
experimental results: a continuum finite element model and a discrete
element model. The discrete element model allowed the mortared joints
between masonry blocks to be modelled explicitly, with crack patterns fol-
lowing the mortar joints. In the continuum model the joints could not be
modelled, so a total strain rotating crack model was used, which distributes
the localised damage over a certain crack bandwidth of the finite elements.
The stress-strain relationship for the façade was elastic in compression
and linear softening in tension, determined by the fracture energy. The
crack patterns and deformations of both the models were similar to the
experiment.
Giardina et al. (2015) used finite element analyses to perform a sensitiv-
ity study to investigate the importance of the interface between a building
and the soil, the building’s stiffness and the building’s weight, and com-
pared the results with centrifuge models. They found that it was important
to enable a gap to form between the building and the soil. The results were
also sensitive to building stiffness, and also it was found to be important to
model the building’s weight.
Ritter et al. (2017) investigated the effect of openings in a façade in cen-
trifuge models of 3D printed buildings on sand. They found that the area of
openings affected the structure response to settlement, especially in the hog-
ging zone, where the building behaved more flexibly. In addition, the area of
the openings reduced the axial stiffness such that more horizontal ground
strain was transferred to the building. It therefore seems important that
openings in façades should be included in any physical or numerical model.
In summary, numerical models of masonry structures affected by settle-
ment need to include cracking, whether this is approximated in a contin-
uum model, or done explicitly in a discrete element model. The layout and
size of openings also needs to be included, as well as the building’s weight.
Interface elements between the building and the ground are needed to allow
a gap to form if it wants to.
13.5 SUMMARY
The concept of limiting tensile strain and its correlation to building damage
categories has been described.
There are three stages to building damage assessment. Each stage is
intended to be progressively more sophisticated and less conservative, with
the aim that structures at low risk of damage can be ruled out, reducing the
number of buildings that require detailed analysis.
Stage 1 analysis involves calculation of contours of 10 mm settlement
and 1:500 slope and classifying all buildings outside of both contours as at
low risk of damage. All buildings within either of the contours will go on
to Stage 2 assessment.
Stage 2 assessment takes the greenfield deflection ratio and horizontal strain
and imposes them on each building using the simple beam method. The build-
ing is separated into partitions in the hogging and sagging zones, and cut-off
at a distance of 2.5i or 3i from the tunnel centreline. The simple beam method
models the building as a rectangular isotropic beam and allows the calculation
of resultant bending strain and resultant diagonal strain. The greater of these
in the hogging and sagging partitions of the building are taken as the limiting
tensile strain and give us the damage category for the building.
Stage 2b assessment involves the use of a relative stiffness method to
allow for soil-structure interaction between the building and the ground.
Modification factors from Franzius et al. (2006a) can be read from design
charts based on numerical modelling, and are applied to the greenfield
deflection ratios and horizontal strains. This generally provides a less con-
servative estimate of damage category compared to Stage 2.
Stage 3 assessment must take account of the individual building’s struc-
tural details and the tunnelling method. It is usually approached in a series
of sub-steps that introduce layers of complexity and refine the prediction.
As well as desk studies and surveys of the building itself, Stage 3 analysis
may include numerical modelling of the structure, with structural details,
façade openings and material behaviour characterised. For masonry struc-
tures the model should include cracking. Generally speaking, Stage 3
should reduce the predicted damage, but can in some cases result in greater
predicted damage.
13.6 PROBLEMS
Q13.1. This question is about Stage 1 building damage assessment. A 6.5 m

diameter tunnel will be driven under an urban area at a depth to
axis of 17 m. Assume the volume loss is 2% and the trough width
parameter K is 0.3.
i. Calculate the trough width i and the maximum settlement
Smax.
ii. Calculate the transverse offset from the tunnel centreline at
which the settlement is 10 mm.
iii. Calculate the transverse offset from the tunnel centreline at
which the slope is 1:500.
iv. Give the transverse offset within which buildings would pro-
ceed to Stage 2 assessment.
Q13.2. This question is about Stage 1 building damage assessment. A

72 m 2 face area shotcrete tunnel will be driven in a straight line
under an urban area at a constant level for 120 m length. Due to
the surface level changing, the depth to axis varies between 17

and 25 m from one end of the tunnel to the other. Assume the
volume loss is 1.5% and the trough width parameter K is 0.45.
Smax at 17 and 25 m.
ii. Calculate the transverse offset from the tunnel centreline at
which the settlement is 10 mm for each case.
iii. Calculate the transverse offset from the tunnel centreline at
which the slope is 1:500 for each case.
iv. Draw the transverse offset within which buildings would pro-
ceed to Stage 2 assessment as a contour on a plan of the tunnel.
Q13.3. This question is about Stage 2 building damage assessment. A 4.5 m

diameter tunnel is to be driven under a brick masonry building at a
depth to axis of 12 m. The building façade is perpendicular to the
tunnel centreline and it has strip foundations at 1 m below ground
level. The façade is 8 m high from the foundation level to the eaves,
and extends 25 m either side of the tunnel centreline. Assume the
volume loss is 1.5% and the trough width parameter K is 0.5.
Smax.
ii. Separate the building into hogging and sagging partitions,
with the hogging partitions cut-off at 2.5i.
iii. Calculate the hogging and sagging deflection ratios.
iv. Calculate the average horizontal strain in the hogging parti-
tion and the sagging partition.
v. Calculate the maximum bending tensile strain and the maxi-
mum diagonal tensile strain in the hogging partition.
vi. Calculate the resultant bending strain and resultant diagonal
strain in the hogging partition and assign the building to a
damage category.
vii. Calculate the maximum bending tensile strain and the maxi-
mum diagonal tensile strain in the sagging partition.
viii. Calculate the resultant bending strain and resultant diagonal
strain in the sagging partition and assign the building to a
damage category. What is the overall damage category for the
building?
Q13.4. This question is about Stage 2 building damage assessment. A

7 m diameter tunnel is to be driven under a reinforced concrete
frame building at a depth to axis of 20 m. The building is per-
pendicular to the tunnel centreline, is 30 m long and is centred
over the tunnel alignment. It has a single level basement with a
1.2 m thick reinforced concrete raft foundation slab at 5 m below
ground level. The building has a 0.4 m thick ground floor slab at
ground level and three more 0.3 m thick floor slabs at 3.5 m cen-
tres above. Assume the volume loss is 1.5% and the trough width
parameter K is 0.5.
Smax at the foundation level.
ii. Separate the building into hogging and sagging partitions,
with the hogging partitions cut-off at 2.5i if necessary, and
calculate the hogging and sagging deflection ratios.
iii. Calculate the average horizontal strain in the hogging parti-
tion and the sagging partition.
iv. Calculate the maximum bending tensile strain and the maxi-
mum diagonal tensile strain in the hogging partition.
v. Calculate the resultant bending strain and resultant diagonal
strain in the hogging partition.
vi. Calculate the maximum bending tensile strain and the maxi-
mum diagonal tensile strain in the sagging partition.
vii. Calculate the resultant bending strain and resultant diagonal
strain in the sagging partition. What is the overall damage
category for the building?
Q13.5. This question is about Stage 2b building damage assessment and uses
the same geometry of tunnel and building described in Q13.4. Assume
the reinforced concrete slabs have a Young’s modulus E = 30 GPa and
assume the secant ground stiffness determined at 0.01% axial strain in
a triaxial compression test from a sample taken from a depth midway
between the surface and the tunnel axis Es = 150 MPa.
i. For the tunnel and building described in Q13.4, calculate the rel-
ative bending stiffness and relative axial stiffness of the building.
ii. Using the relative stiffness values, use the design charts from
Franzius et al. (2006a) to find values for modification factors.
iii. Apply the modification factors to the deflection ratios and max-
imum horizontal strains in the hogging and sagging partitions.
iv. Calculate the maximum bending tensile strain and the max-
imum diagonal tensile strain in the hogging and sagging
partitions.
v. Calculate the resultant bending strain and resultant diagonal
strain in the hogging and sagging partitions. What is the over-
all damage category for the building?
REFERENCES
Bhatt, P. (1999). Structures. New Jersey: Prentice Hall.

Boscardin, M. D. & Cording, E. J. (1989). Building response to excavation-induced
settlement. ASCE J. Geot. Eng. 115, No. 1, 1–21.
Bowers, K. H. & Moss, N. A. (2006). Settlement due to tunnelling on the

CTRL London Tunnels. Proc. 5th Int. Symp. on Geotechnical Aspects of
Underground Construction in Soft Ground (eds Bakker, K. J., Bezuijen, A.,
Broere, W. & Kwast, E. A.), Amsterdam, The Netherlands, 15th–17th June
2005, pp.203–209. London: Taylor & Francis.
Burland, J. B. (2001). Assessment methods used in design. Building response to
tunnelling, volume 1: projects and methods (eds Burland, J. B., Standing,
J. R. & Jardine, F. M.), CIRIA Special Publication 200, pp.23–43. London:
Thomas Telford Publishing.
Burland, J. B., Broms, B. B. & de Mello, V. F. B. (1977). Behaviour of founda-
tions and structures. Proc. 9th Int. Conf. Soil Mechanics and Foundation
Engineering, Tokyo, Japan, Session 2, pp. 495–546.
Burland, J. B. & Wroth, C. P. (1974). Settlement of buildings and associated dam-
age. State of the art review. Conf. on Settlement of Structures, Cambridge,
pp. 611–654. London: Pentech Press.
Crossrail (2008). Crossrail Information Paper D12 – Ground Settlement. London,
UK: CLRL.
DeJong, M. J., Giardina, G., Chalmers, B., Lazarus, D., Ashworth, D. & Mair, R.
J. (2019). Impact of the Crossrail tunnelling project on masonry buildings
with shallow foundations. Proc. Instn Civ. Engrs – Geot. Engrg 172, No. 5,
402–416.
Farrell, R. P. & Mair, R. J. (2012). Centrifuge modelling of the response of buildings
to tunnelling. Proc. 7th Int. Symp. on Geotechnical Aspects of Underground
Construction in Soft Ground (ed. Viggiani, G.), Rome, Italy, 17th–19th May
2011, pp. 343–351. London: Taylor & Francis Group.
Farrell, R., Mair, R., Sciotti, A. & Pigorini, A. (2014). Building response to tunnel-
ling. Soils Found. 54, No. 3, 269–279.
Franzius, J. N. (2004). The behaviour of buildings due to tunnel induced subsid-
ence. PhD thesis, Imperial College, University of London, 2004.
Franzius, J. N., Potts, D. M. & Burland, J. B. (2006a). The response of surface
structures to tunnel construction. Proc. Instn Civ. Engrs – Geot. Engrg 159,
GE1, 3–17.
Franzius, J. N., Potts, D. M. & Burland, J. B. (2006b). Twist behaviour of buildings
due to tunnel induced ground movement. Proc. 5th Int. Symp. on Geotechnical
Aspects of Underground Construction in Soft Ground (eds Bakker, K. J.,
Bezuijen, A., Broere, W. & Kwast, E. A.), Amsterdam, The Netherlands,
15th–17th June 2005, pp. 107–113. London: Taylor & Francis Group.
Gere, J. M. & Timoshenko, S. P. (1991). Mechanics of materials, 3rd S. I. edition.
London: Chapman & Hall.
Giardina, G., DeJong, M. J., Chalmers, B., Ormond, B. & Mair, R. J. (2018). A
comparison of current analytical methods for predicting soil-structure inter-
action due to tunnelling. Tunn. Undergr. Space Technol. 79, 319–335.
Giardina, G., DeJong, M. J. & Mair, R. J. (2015). Interaction between surface
structures and tunnelling in sand: Centrifuge and computational modelling.
Tunn. Undergr. Space Technol. 50, 465–478.
Giardina, G., Losacco, N., DeJong, M. J., Viggiani, G. M. B. & Mair, R. J.
(2020). Effect of soil models on the prediction of tunnelling-induced defor-
mations of structures. Proc. Instn Civ. Engrs – Geot. Engrg 173, No.5,
379–397.
Giardina, G., Marini, A., Hendriks, M. A. N., Rots, J. G., Rizzardini, F. &
Giuriani, E. (2012). Experimental analysis of a masonry façade subject to
tunnelling-induced settlement. Eng. Struct. 45, 421–434.
Giardina, G., Marini, A., Riva, P. & Giuriani, E. (2019). Analysis of a scaled
stone masonry facade subjected to differential settlements. Int. J. Architect.
Heritage, published online, DOI: 10.1080/15583058.2019.1617911.
Giardina, G., van de Graaf, A. V., Hendriks, M. A. N., Rots, J. G. & Marini, A.
(2013). Numerical analysis of a masonry façade subject to tunnelling-induced
settlements. Eng. Struct. 54, 234–247.
Goh, K.H. & Mair, R.J. (2008). Response of a building under excavation-induced
ground movements, Proc. Int. Conf. on Deep Excavations, 10th–12th Nov
2008, Singapore.
Goh, K.H. & Mair, R.J. (2012). The response of buildings to movements induced by
deep excavations. Proc. 7th Int. Symp. on Geotechnical aspects of underground
construction in soft ground (ed. Viggiani, G.), Rome, Italy, 17th–19th May 2011,
Goh, K. H. & Mair, R. J. (2014). The response of framed buildings to excavation-
induced movements. Soils Found. 54, No. 3, 250–268.
Harris, D. I. & Franzius, J. N. (2006). Settlement assessment of running tunnels
– a generic approach. Proc. 5th Int. Symp. on geotechnical aspects of under-
ground construction in soft ground (eds Bakker, K. J., Bezuijen, A., Broere,
W. & Kwast, E. A.), Amsterdam, The Netherlands, 15th –17th June 2005,
Mair, R. J. & Taylor, R. N. (2001). Elizabeth House: settlement predictions.
Building response to tunnelling, vol. 1 projects and methods, CIRIA Special
Publication 200, pp. 195–215. London: Thomas Telford.
Mair, R. J., Taylor, R. N. & Burland, J. N. (1996a). Prediction of ground move-
ments and assessment of risk of building damage due to bored tunnelling.
Proc. Int. Symp. on geotechnical aspects of underground construction in
soft ground (eds Mair, R. J. & Taylor, R. N.), London, UK, pp. 713–718.
Rotterdam: Balkema.
Mair, R. J., Taylor, R. N. & Burland, J. N. (1996b). Discussion: Reply to discussion
by M. P. O’Reilly. Proc. Int. Symp. on geotechnical aspects of underground
construction in soft ground (eds Mair, R. J. & Taylor, R. N.), London, UK,
Melis, M. & Rodriguez Ortiz, J. (2001). Consideration of the stiffness of buildings in
the estimation of subsidence damage by EPB tunnelling in the Madrid subway.
Response of buildings to excavation induced ground movements, Proc. Int.
Conf. held at Imperial College, London, UK, 17th–18th July 2001 (ed. Jardine,
F. M.), CIRIA Special Publication 199. London: CIRIA.
Meyerhof, G. G. (1953). Some recent foundation research and its application to
design. Struct. Eng. 31, 151–167.
Moss, N. A. & Bowers, K. H. (2006). The effect of new tunnel construction under
existing metro tunnels. Proc. 5th Int. Symp. on Geotechnical Aspects of
Underground Construction in Soft Ground (eds Bakker, K. J., Bezuijen, A.,
Broere, W. & Kwast, E. A.), Amsterdam, The Netherlands, 15th–17th June
2005, pp. 151–157. London: Taylor & Francis.
Netzel, H. D. (2009). Building response due to ground movements. PhD thesis,
Technische Universiteit Delft. Delft, The Netherlands: Delft University Press.
Pickhaver, J. A., Burd, H. J. & Houlsby, G. T. (2010). An equivalent beam method

to model masonry buildings in 3D finite element analysis. Comput. Struct.
88, 1049–1063.
Potts, D. M. & Addenbrooke, T. I. (1997). A structure’s influence on tunnelling-
induced ground movements. Proc. Instn Civ. Engrs – Geot. Engrg 125, April,
109–125.
Rankin, W. J. (1988). Ground movements resulting from urban tunnelling: predic-
tion and effects. Conf. on Engineering Geology of Underground Movements,
Nottingham, Geological Society Engineering Geology Special Publication
No. 5, pp. 79–92.
Ritter, S., Giardina, G., DeJong, M. J. & Mair, R. J. (2017). Influence of building
characteristics on tunnelling-induced ground movements. Géotechnique 67,
No. 10, 926–937.
Standing, J. R. (2001). Elizabeth House, Waterloo. Building response to tunnelling,
vol. 2 case studies, CIRIA Special Publication 200, pp. 547–612. London:
Thomas Telford.
Viggiani, G. & Standing, J. R. (2001). The Treasury. Building response to tun-
nelling, vol. 2 case studies, CIRIA Special Publication 200, pp. 401–432.
Appendix A: Derivation
of wedge-prism method
Anagnostou & Kovári (1994, 1996a, 1996b) used a limit equilibrium solu-
tion they attributed to Horn (1961) and developed it to allow direct calcu-
lation of the required support pressure in any situation above or below the
water table, with or without seepage, for tunnels where the unsupported
length P = 0. The geometry is shown in Figure A.1. This solution is poten-
tially much more useful for practical situations involving closed-face tun-
nelling machines, although whether it represents the true collapse load is
not guaranteed because the geometry is a simplification. Jancsecz & Steiner
(1994) stated that use on several slurry tunnel boring machine (TBM)
projects has validated the wedge-prism method, as does Broere (1998).
Anagnostou & Kovári (1996a) compared results from the wedge-prism
method with centrifuge tests in dry sand by Chambon & Corté (1994), and
Messerli et al. (2010) ran their own 1 g model tests of a tunnel in dry sand,
and in both cases a good agreement was found.
The wedge ABCDEF is acted on by:
1. its own weight

2. resultant normal and shear forces along failure surfaces BCF, ADE
and ABFE
3. the resulting support force acting on surface ABCD
4. a vertical silo pressure produced by the rectangular prism above the
surface CDEF
The angle of the wedge, labelled ω in Figure A.1, is unknown. It may be

found by iteration, either by finding the value that results in the maximum
support force, or by fixing the support force and finding the value that
results in the lowest factor of safety.
Groundwater pressure is assumed to be hydrostatic, and effective stress
is used throughout the calculations.
519
520 Appendix A: Derivation of wedge-prism method
Figure A.1 Wedge and prism model (Anagnostou & Kovári, 1996b).
In reality, the shear stresses on the failure surfaces depend on the hori-
zontal effective stress acting normal to each surface, and this cannot be
calculated easily. Anagnostou & Kovári (1994) recommended using a con-
stant ratio of horizontal to vertical effective stress of 0.8.
FINDING THE VERTICAL PRESSURE

ON SURFACE CDEF
The vertical pressure on CDEF is found by applying Janssen’s silo formula

to the prism above the water table (HIJGKLMN), and then to the prism
below the water table (CDEFHIJG).
Using Jancsecz & Steiner (1994)’s definition initially as it is simpler:
Perimeter of CDEF = perimeter of JGHI = U
Area of CDEF = area of JGHI = F
The forces on a slice of a prism silo, excluding weight, are shown in

Figure A.2.
The weight of each slice is given by:
dW = Fγ dz (A.1)
Appendix A: Derivation of wedge-prism method 521
Figure A.2 Prism silo representation.
where γ is the unit weight of the soil, which would be the dry unit weight
above the water table and the submerged unit weight below the water table,
and λ is the horizontal to vertical effective stress ratio. Note that the symbol λ
is used rather than K0 because K0 refers specifically to the ratio of horizontal
to vertical in situ effective stress at rest. In this situation we are finding the
point of mass failure of the soil, so it is very disturbed and K0 no longer exists.
The vertical force on the top of the slice is given by Fσ v . The upwards
vertical force on the bottom of the slice is given by F (σ v + dσ v ) .
The resultant friction resistance force around the sides of each slice is
given by:
dS = Udz ( c + λσ v′ tan φ ) (A.2)
Vertical equilibrium of the slice gives:
dW = F (σ v′ + dσ v′ ) − Fσ v′ + dS (A.3)
This can be written in full as:
Fγ dz = F (σ v′ + dσ v′ ) − Fσ v′ + Udz ( c + λσ v′ tan φ ) (A.4)
Which simplifies to:
Fγ dz = Fdσ v′ + Udz ( c + λσ v′ tan φ ) (A.5)
Dividing across by dz gives:
dσ v′
Fγ = F + U ( c + λσ v′ tan φ ) (A.6)
dz
Dividing across by F gives:

dσ v′ U
γ = + ( c + λσ v′ tan φ ) (A.7)
dz F
Rearranging:
dσ v′  Uλ tanφ  U
+  σ v′ = γ − c (A.8)
dz  F  F
Since σ v′ is a function of z, this is a linear first-order ordinary differential

equation (ODE) of the form:
dy
+ Py = Q (A.9)
dx
(
where, y = σ v′, x = z, P = Uλ tan
F )
φ
, and Q = γ − UF c.
We need to integrate to solve this ODE, but this is not straightforward.
The solution to this form of ODE may be found if we can find a function
F ( x ) such that dF /dx = PF ( x ). Or, using the names of variables and con-
stants in our specific situation, a function F ( z ) such that dF /dz = PF ( z ) ,
with P defined as above. Try using:
 Uλ tanφ z 
F ( z ) = exp   (A.10)
 F
Then:
1  d 
(
F ( z )  dz
)
F ( z ) σ v′ ( z )  = Q (A.11)

Substituting F ( z ) and Q according to the definitions in Equation A.9 gives:
1  d   Uλ tanφ z    U
λ φ
  exp   σ v′   = γ − c
 U tan z   dz  F   F (A.12)
exp  
 F
Solving this gives:
1   Uλ tanφ z  dσ v′ Uλ tanφ  Uλ tanφ z   U

λ φ  exp   + exp 
  σ v′  = γ − c
 U tan z  F dz F F  F
exp  
 F (A.13)
Which can be simplified to:
dσ v′  Uλ tanφ  U
+ exp 
  σ v′ = γ − c (A.14)
dz F  F
This is our original ODE, proving that Equation A.12 is equivalent to

Equation A.8.
Equation A.12 is useful to us because we can integrate it, as follows. First
rearrange Equation A.12:
d   Uλ tanφ z    U   Uλ tanφ z 
exp   σ v′  =  γ − c  exp   (A.15)
dz  F  F F
Now integrate with respect to dz:
 Uλ tanφ z   U   Uλ tanφ z 
exp 
 F
 σ v′ =  γ − c 
F ∫ exp  F
 dz (A.16)
This gives:
 U   Uλ tanφ z 
λ φ  γ − c exp  
 U tan z  F F
exp   σ v′ = + C (A.17)
 F  Uλ tanφ
F
where C is an integration constant. To find C we use the boundary condi-
tion σ v′ = q at z = 0:
 U 
 γ − c 
F
q= + C (A.18)
Uλ tanφ
F
Therefore:
 U 
 γ − c 
F (A.19)
C =q−
Uλ tanφ
F
Putting this expression for C into Equation A.17 gives:
 U   Uλ tanφ z   U 
 γ − c  exp    γ − c 
 Uλ tanφ z  F F F
exp   σ v′ = +q− (A.20)
 F  Uλ tanφ Uλ tanφ
F F
Rearranging:
 U 
 γ − c  
 −Uλ tanφ z  F  −Uλ tanφ z  
σ v′ = qexp   +  1 − exp   
 F  U λ tanφ F  (A.21)
F
Simplifying:
F
γ −c
U  −Uλ tanφ z    −Uλ tanφ z 
σ v′ =  1 − exp    + qexp   (A.22)
λ tanφ   F   F
Equation A.22 is the Janssen silo equation used by Jancsecz & Steiner
(1994) in their paper to calculate the vertical effective stress on the
top of the wedge. Anagnostou & Kovári (1994) use a slightly different
notation, where the ratio of area to perimeter F /U = r , which may also
be expressed in terms of the wedge slope angle ω (cf. Figure A.1) as
follows:
0.5Dtanω
r= (A.23)
1 + tanω
Substituting F /U for r in Equation A.22 and applying z = ( H − H w )

for the prism above the water table (HIJGKLMN) gives the following
expression:
σ v′ =
γ dr − c
λ tanφ
( )
1 − exp ( − λ tanφ ( H − H w ) /r ) + qexp ( − λ tanφ ( H − H w ) /r ) (A.24)
Anagnostou & Kovári (1994) do not include a surface surcharge in their

equations, so for now we will assume that q = 0. Therefore the vertical
effective stress on surface GHIJ may be given by:
σ v′ ,GHIJ =
λ tanφ
(
γ dr − c
)
1 − exp ( − λ tanφ ( H − H w ) /r ) (A.25)
The prism below the water table has the following equation for the vertical
effective stress on surface CDEF:
γ ′r − c
σ v′ ,CDEF =
λ tanφ
(1 − exp ( −λ tanφHw /r )) + qexp ( −λ tanφHw /r ) (A.26)
For the lower prism, q = σ v′ ,GHIJ , which gives:
γ ′r − c
σ v′ ,CDEF =
λ tanφ
(1 − exp ( −λ tanφHw /r ))
(A.27)
γ r−c
+ d
λ tanφ
( )
1 − exp ( − λ tanφ ( H − H w ) /r ) ( exp ( − λ tanφ H w /r ))
This simplifies to:
γ ′r − c
σ v′ ,CDEF =
λ tanφ
(A.28)
γ r−c
+ d
λ tanφ
(exp ( −λ tanφHw /r ) − exp ( −λ tanφH /r ))
Equation A.28 is exactly the same as Equation (2) in Anagnostou & Kovári
(1994).
If we wanted to include a surface surcharge, we could repeat this process
using Equation A.24 instead of Equation A.25 and we would get:
γ ′r − c
σ v′ ,CDEF =
λ tanφ
γ dr − c
+
λ tanφ
(exp ( −λ tanφHw /r ) − exp ( −λ tanφH /r )) (A.29)
+qexp ( − λ tanφ H /r )
FINDING THE SLIDING FRICTION

ON SURFACES ADE AND BCF
Anagnostou & Kovári (1994) assume that the vertical effective stress on
these surfaces is made up of two components, which are illustrated in
Figure A.3, and is according to the German standard for slurry walls (DIN
4126, 1986). The first component is the interface vertical effective stress
σ v′ ,CDEF , which is assumed to decrease linearly from σ v′ ,CDEF at the top, to
zero at the bottom. The second component is the self-weight of the soil,
which is assumed to increase linearly from zero at the top to γ ′D at the
bottom.
Note that later papers by the same authors (Anagnostou & Kovári,
1996a, 1996b) assume the square ABCD is of equal area to the circular
tunnel face, and so is not of side D but is slightly smaller. Broere (1998)
states that this makes the required support force about 3% smaller, so is
not significant for most practical purposes. When using these calculations,
just make D2 equal to the face area for a more accurate solution.
The shear stress τ ω is given by:
τ ω = c + λω σ z′ tanφ (A.30)
To compensate for uncertainties in the distribution of vertical effective

stress, which some studies have found to be overestimated by the DIN
4126 method, Anagnostou & Kovári (1994) recommended using a reduced
Figure A.3 Anagnostou & Kovári (1994, 1996a & 1996b)’s assumptions for vertical effec-
tive stress on wedge surfaces ADE and BCF.
value of the horizontal to vertical effective stress ratio λω = 0.4. A later

paper (Anagnostou, 2012) used a method of slices to investigate what
the true value of λω should be, and found that λω = 0.4 was reasonably
conservative.
The mean shear stress is found by using the assumption of the triangular
distribution of σ v′ and D γ ′ in Figure A.3, as well as the triangular shape
of ADE or BCF. This triangular distribution is an assumption and alterna-
tive assumptions are investigated by Broere (2001), and Anagnostou (2012)
shows how there is no need to make an assumption of the distribution of
vertical stress if a method of slices is used, dividing the wedge into infini-
tesimal horizontal slices.
Because of the triangular shape of ADE/BCF the average occurs at dis-
tance D/3 from the top. For σ v′ the distribution decreases linearly from
σ v′ ,CDEF at the top of the wedge, to zero at the bottom, such that the aver-
age will be 2/3 × σ v′ ,CDEF. Similarly, for D γ′ the average will be 1/3 × Dγ ′ .
Therefore:
1 2  tanφ
τ ω = c + λω  γ ′D + σ v′ ,CDEF  (A.31)
3 3  v
where v is the factor of safety. When calculating the minimum support

pressure required to prevent failure, use v = 1.
FINDING THE SLIDING FRICTION ON ABEF
Anagnostou & Kovári (1994) do not describe how the sliding friction on
the wedge slope surface ABEF is calculated, but a very sensible way of
doing this is set out by Dias & Bezuijen (2015) and this will be explained
Figure A.4 Forces on the wedge viewed in 3D (left) and in 2D cross-section (right).
in the following diagrams and equations. The forces are shown in

Figure A.4,1 where:
S is the effective support force required to maintain equilibrium (this is

what we want to calculate)
T are the friction and cohesion forces on the sides of the wedge, assumed
parallel to the wedge slope ABEF
Gs is the effective vertical force due to effective silo pressure on CDEF
Q is the effective reaction force due to friction on the wedge slope ABEF
Cw is the resistance due to cohesion on the wedge slope ABEF
Gw is the effective weight of the wedge
Kolymbas (2008) points out that the horizontal component of friction

on the surface CDEF is ignored by “most authors”. Certainly Jancsecz &
Steiner (1994), Anagnostou & Kovári (1994, 1996a, 1996b), Broere (1998)
and Dias & Bezuijen (2015) do not include it, though none of them explain
why it is neglected. The reason for neglecting it is that the wedge and prism
are not really separate moving rigid blocks, even though we may consider
them separately to calculate the vertical stress, but they should be consid-
ered as a single rigid plastic block moving downwards. If friction on CDEF
1 Note that both Jancsecz & Steiner (1994) and Broere (1998) have errors in their wedge
forces diagrams. The forces on the wedge slope ABEF can either be represented by a shear
force parallel to the slope and a normal force perpendicular to the slope, or by a resultant
force at an angle ϕ to the normal. They both show a resultant at an angle ϕ to the normal
and a shear force parallel to the slope.
Figure A.5 Polygon of forces on the wedge.
were included then faces would be stable with no cohesion or support pres-
sure, which is clearly unrealistic.
The effective reaction force Q and the effective support force S are
unknown. Using the method described by Dias & Bezuijen (2015), if we
solve for force equilibrium in the direction perpendicular to Q, then it
will not appear in the equation and we can more easily calculate S. Q is
orientated at an angle ϕ to the slope normal. The best way to visualise
the force equilibrium is to use a ‘polygon of forces’, where the forces
have to create a closed polygon if they are in equilibrium, as shown in
Figure A.5.
The equilibrium equation perpendicular to Q may be expressed as:
(Gw + Gs ) cos (φ + ω ) = (2T + Cw ) cos φ + S sin (φ + ω ) (A.32)

Rearranging for S gives:
S=
Gw + Gs
−
(2T + Cw ) cos φ (A.33)
tan (φ + ω ) sin (φ + ω )
T may be calculated by multiplying the mean shear stress calculated in

Equation A.31 by the area of ADE or BCF, which is given by 0.5D2 tanω :
1 2  tanφ
T = τ ω × (0.5D2 tan ω ) = λω  γ ′D + σ v′ ,CDEF  (0.5D2 tan ω ) (A.34)
3 3  v
Gs may be calculated by multiplying σ v′ ,CDEF calculated in Equation A.28 or
Equation A.29 by the area of CDEF as follows:
Gs = σ v′ ,CDEF D2 tan ω (A.35)
Gw is simply the effective weight of the wedge, given by:
Gw = 0.5γ ′D3 tan ω (A.36)
Cw is the resistance force due to cohesion on the wedge slope, which is the
cohesion c multiplied by the surface area of ABEF, given by:
cD2
Cw = (A.37)
cos ω
S is the effective support force required to maintain equilibrium of the
wedge and prism. The required minimum effective support pressure, usually
denoted by s′, may be found by dividing S by the area of ABCD, which is D2 .
CALCULATING VALUES FOR THE F 0 NOMOGRAM
If the same assumptions are used that Anagnostou & Kovári (1996a, 1996b)
used to create their nomograms, i.e. γ d /γ ′ = 1.6, λ = 0.8 and λω = 0.4,
then we can calculate values of s ′ and compare these to the nomograms.
For H = 8 m, D = 8 m, H w = 4 m, γ d = 16 kN/m3, γ ′ = 10 kN/m3, ϕ = 15°
and c = 0 kPa:
H /D = 1
h0 = 1.5D
The maximum value of S is found when ω = 40°. At this value:
r = 1.82502 [Equation A.23]

σ v′ above the water table = 51.06769 kPa [Equation A.25]
σ v′ ,CDEF = 63.84038 kPa [Equation A.28]
τ ω = 7.41972 kPa [Equation A.31]
T = 199.23 kN [Equation A.34]
Gs = 3428.38 kN [Equation A.35]
Gw = 2148.10 kN [Equation A.36]
S = 3434.84 kN [Equation A.33]
s′ = S /D2 = 53.6694 kPa
F0 = s′ /γ ′D = 0.671
Further values can be found in the same way, for ϕ = 20, 25, 30, 35°, for
H /D = 1, 2, 5, and for h0 = 1.5D and h0 = H + D, and exact agreement with
the nomograms can be found, except that the values found for h0 = 1.5D
Figure A.6 F0 nomogram based on wedge-prism calculations presented in this appendix.
exactly match those in the nomograms for h0 = H + D, and vice-versa. The

lines were mislabelled in the papers by Anagnostou & Kovári (1996a &
1996b), as confirmed by Anagnostou (2019). The results are shown in
Figure A.6.
CALCULATING VALUES FOR THE F1 NOMOGRAM
If the same assumptions are used that Anagnostou & Kovári (1996a, 1996b)
used to create their nomograms, i.e. γ d /γ ′ = 1.6, λ = 0.8 and λω = 0.4,
then we can calculate values of s ′ and from Equation 3.2 we can calculate
F1. We are assuming no groundwater flow, so only the F0 and F1 terms are
active. Therefore:
s′ = F0γ ′D − F1c (A.38)
For the same situation we used to calculate F0 above, i.e. H = 8 m, D = 8 m,

H w = 4 m, γ d = 16 kN/m3, γ ′ = 10 kN/m3, ϕ = 15° and c = 5 kPa:
H /D = 1
h0 = 1.5D
The maximum value of S is found when ω = 40°. At this value:
r = 1.82502 [Equation A.23]

σ v′ above the water table = 42.32333 kPa [Equation A.25]
σ v′ ,CDEF = 49.62981 kPa [Equation A.28]
τ ω = 11.40433 kPa [Equation A.31]
T = 306.22 kN [Equation A.34]
Gs = 3428.38 kN [Equation A.35]
Gw = 2148.10 kN [Equation A.36]
Cw = 417.73 kN [Equation A.37]
S = 2155.58 kN [Equation A.33]
s′ = S /D = 33.6809 kPa
2
Now from the definition of F1 (i.e. Equation A.38):
F1 =
( F0γ ′D − s′ ) = (53.6694 − 33.6809) = 3.9977 (A.39)
c 5
Further values can be found in the same way, for ϕ = 20, 25, 30, 35°, for
H /D = 1, 2, 5, and for h0 = 1.5D and h0 = H + D, and exact agreement with
Figure A.7 F1 nomogram based on wedge-prism calculations presented in this appendix.

the nomograms can be found. The values for h0 = 1.5D are the same as for
h0 = H + D, indicating that the position of the water table does not affect F1.
The results are shown in Figure A.7 and are the same as the values in the
papers by Anagnostou & Kovári (1996a, 1996b).
REFERENCES
Anagnostou, G. (2012). The contribution of horizontal arching to tunnel face sta-

bility. geotechnik 35, Heft 1, 34–44.
Anagnostou, G. (2019). Personal communication.
Anagnostou, G. & Kovári, K. (1994). The face stability of slurry shield-driven tun-
nels. Tunn. Deep Space 9, No. 2, 165–174.
Anagnostou, G. & Kovári, K. (1996a). Face stability conditions with earth-
165–173.
Anagnostou, G. & Kovári, K. (1996b). Face stability in slurry and EPB shield tun-
nelling. Geotechnical aspects of underground construction in soft ground
(eds Mair, R. J. & Taylor, R. N.), pp. 453–458. Rotterdam, The Netherlands:
Balkema.
Broere, W. (1998). Face stability calculation for a slurry shield in heterogeneous
soft soils. Proc. Tunnels and Metropolises (eds Negro Jr, A. & Ferreira, A.
A.), Sao Paulo, Brazil, pp. 215–218. Rotterdam: Balkema.
Broere, W. (2001). Tunnel face stability & new CPT applications. Ph.D. thesis,
Technical University of Delft.
Dias, T. G. S. & Bezuijen, A. (2015). TBM pressure models: calculation tools. Proc.
WTC2015 – SEE Tunnel: Promoting Tunneling in SEE Region, Dubrovnik,
Croatia, 22nd–28th May 2015.
Horn, M. (1961). Horizontaler Erddruck auf senkrechte Abschlussflächen
von Tunnelröhren. Landeskonferenz der Ungarischen Tiefbauindustrie,
pp. 7–16. Horizontal earth pressure on perpendicular tunnel face. Proceedings
of the Hungarian National Conference of the Foundation Engineer Industry,
Budapest, Hungary, pp. 7–16.
Jancsecz, S., Steiner, W., 1994.Face support for a large mix-shield in heterogeneous
ground conditions. Proc. 7th Int. Symp. Tunnelling ’94, 5th–7th July 1994,
London, UK, pp. 531–549. London, UK: IMM.
Kolymbas, D. (2008). Tunnelling and tunnel mechanics. Berlin, Germany: Springer
Verlag.
nel face collapse in sand. Physical modelling in geotechnics (eds Springman,
Appendix B: Details from derivation
of Curtis–Muir Wood equations
INTEGRATION BY PARTS
For calculation of N, the following trigonometric expressions need to

integrated:
π
2
A=
∫ cos 2θ sinθ dθ (B.1)
0
π
2
B=
∫ sin 2θ cosθ dθ (B.2)
0
Use the following integration by parts rule:
du  

∫ uv dθ = u ∫ v dθ − ∫ dθ  ∫ v dθ  dθ (B.3)
Starting with A, let u = cos2θ and v = sinθ . First calculate:
du d
= cos 2θ = −2sin 2θ (B.4)
dθ dθ
And:

∫ v dθ = ∫ sinθ dθ = − cosθ (B.5)
533
534 Appendix B: Details from derivation of Curtis–Muir Wood equations
Now using Equation (B.3):
π
2 π
∫
A = cos2θ sinθ dθ =  cos 2θ ( − cos θ ) −
∫ ( −2sin 2θ )( − cos θ ) dθ  (B.6)
2

 0
0
This simplifies to:
∫
A =  − cos 2θ cos θ − 2 sin 2θ cos θ dθ  (B.7)
2

 0
Now let’s calculate B. Let u = sin2θ and v = cosθ . First calculate:
du d
= sin 2θ = 2 cos 2θ (B.8)
dθ dθ
And:

∫ v dθ = ∫ cosθ dθ = sinθ (B.9)
Now using Equation (B.3):
π
2 π
∫
B = sin2θ cosθ dθ = sin 2θ sin θ −
∫ 2 cos 2θ sin θ dθ  (B.10)
2

 0
0
This simplifies to:
∫
B = sin 2θ sin θ − 2 cos 2θ sin θ dθ  (B.11)
2

 0
And can be expressed as:
π
 2
B = sin 2θ sin θ − 2A  (B.12)
 0
Similarly Equation (B.7) can be expressed as:

π
 2
A =  − cos 2θ cos θ − 2B  (B.13)
 0
Appendix B: Details from derivation of Curtis–Muir Wood equations 535
If we substitute for B in Equation (B.13), we get:
π
 2
A =  − cos 2θ cos θ − 2 ( sin 2θ sin θ − 2A)  (B.14)
 0
This simplifies to:
π
 2
A =  − cos 2θ cos θ − 2sin 2θ sin θ + 4 A  (B.15)
 0
Then:
π
1 2
A =  cos 2θ cos θ + 2sin 2θ sin θ  (B.16)
3 0
Applying the integration limits gives:
1 1
A=
3
{ }
(0 + 0) − (1 + 0) = − (B.17)
3
Substituting this value for A into Equation (B.12):
π
2 2


B = sin 2θ sin θ  + =
 0 3
{(0) − (0)} + 23 = 23 (B.18)
For the calculation of M, the following trigonometric expressions need to

be integrated:
π
4

∫
A = cos2θ sinθ dθ (B.19)
0
π
4

∫
B = sin2θ cosθ dθ (B.20)
0
π
4

∫
C = sin2θ dθ (B.21)
0
536 Appendix B: Details from derivation of Curtis–Muir Wood equations
A and B turn out to have the same values as before when the integration
was between 0 and π /2, i.e. −1/3 and 2/3 respectively. The value for C is
given by:
π
4 π
 1 4   1  1

0
∫
C = sin2θ dθ =  − cos 2θ  = (0) −  −   = (B.22)
 2 0   2  2
Quod erat demonstrandum.

Return to Section 6.3.4.
Appendix C: Derivation of the
deflection of a rectangular simply-
supported beam under a point load
This is the derivation of Equation 13.5 in Chapter 13, which gives the mid-
span deflection of a simply supported rectangular beam with a centrally-
applied point load. The equation is reproduced here:
PL3  72EI 
∆= 1 + 5L2 AG  (13.5)
48EI
The idealised geometry is shown in Figure C.1. The z axis is out of the
page. The beam has width in the z axis direction b, height in the y axis
direction H and length in the x axis direction L. A point load P is applied
at midspan.
The cross-sectional area A of the beam is given by:
A = bH (C.1)
The shear force V from x = 0 to x = L /2 is given by:
P
V= (C.2)
2
The bending moment M from x = 0 to x = L /2 is given by:
Px
M= (C.3)
2
537
538 Appendix C: Derivation of deflection under a point load
Figure C.1 Simply-supported rectangular beam with a centrally-applied point load.
BENDING DEFLECTION
The deflection at midspan due to bending moment may be obtained from

simple beam theory (also known as ‘Euler-Bernoulli beam theory’). If u, v,
w denote displacements in the x, y, z directions then the slope θ is given by:
dv 1
θ=
dx
=−
EI ∫ M dx (C.4)
E is the Young’s modulus of the beam

I is the second moment of area
Substituting for M in Equation C.4 using the expression in Equation C.3

gives:
1 Px 1  Px2 
θ=−
EI ∫ 2
d x = −
EI  4
+ C1 

(C.5)
C1 is an integration constant
At midspan x = L /2 and due to symmetry the slope θ = 0. Hence

C1 = − PL2 /16 and we get:
dv 1  Px2 PL2 
θ= =− − (C.6)
dx EI  4 16 
The slope θ, for small angles, is equal to dv /dx. If we integrate dv /dx from
x = 0 to x = L / 2 with respect to x we will get the midspan deflection due
to bending ∆ bmax .
L /2 L /2
dv 1  Px2 PL2  1  Px3 PL2 x 
∆ bmax =
∫ dx
dx = −
EI ∫0
 4 − 16  dx = − EI  12 − 16 + C2  (C.7)
0
C2 is an integration constant
Appendix C: Derivation of deflection under a point load 539
At the support, when x = 0 , ∆ bmax = 0, ∴ C2 = 0. Therefore:
1   PL3 PL3   PL3

∆ bmax = −   − + 0  − {0}  = (C.8)
EI   96 32   48EI
SHEAR STRESS
Burland & Wroth (1974), based on Timoshenko (1957), assumed that sim-
ple beam theory could also be applied to shear deflection. However, this
will not be accurate because the ‘shear area’, the effective area of a section
participating in the shear deformation, is not equal to the cross-section area
(e.g. Bhatt, 1999: pp. 275–279). This effect is largely to do with warping,
caused by the parabolic shear stress distribution, which induces a cubic
displacement profile as shown in Figure C.2.
When simple beam theory is applied to shear deformations, the shear
stress is determined on the basis of equilibrium only and compatibility is
not considered (Iyer, 2005). Simple beam theory assumes that plane sec-
tions remain plane, but they do not.
The shear stress distribution, as shown in Figure C.2, is parabolic. This
is because τ xy = τ yx and the shear stress on the top and bottom surfaces of
the beam must be zero. The maximum shear stress occurs at the centroid
of the section and is equal to 1.5 times the average shear stress. This causes
the shear strain distribution to also be parabolic, and this leads to a cubic
displacement profile, i.e. warping, as seen in the last diagram of Figure C.2.
The average shear stress is given by:
V V
τ ave = = (C.9)
A bH
τ ave is the average shear stress in a cross-section
V is the shear force acting at the cross-section
b is the width of the section (in the z axis direction)
H is the height of the section (in the y axis direction)
A is the area of the cross-section, where A = bH
By considering equilibrium of an element in the cross-section of the beam,

the following formula for shear stress, known as the ‘shear formula’ may
be obtained:
VQ
τ= (C.10)
Ib
τ is the shear stress

Q is the first moment of area
I is the second moment of area
Figure C.2 Shear warping in a rectangular cross-section beam.
The shear stress at distance y in the cross-section may be calculated using

Figure C.3.
The shaded area A1 in Figure C.3 is given by:
H 
A1 = b  − y  (C.11)
 2 
And the distance of the centroid of the shaded area from the neutral
axis is:
H 
 + y  H y (C.12)
2
y1 = = +
2 4 2
Figure C.3 First moment of area calculation.

The first moment of area Q for the shaded area is its area multiplied by the
distance of its centroid from the neutral axis:
H   H y  b  H2 
Q = A1y1 = b  − y   +  =  − y2  (C.13)
 2   4 2 2  4 
Substituting for Q in the shear formula (Equation C.10) gives:
V  b  H2   V  H2 
τ=   − y2   =  − y2  (C.14)
Ib  2 4   2I 4 
Equation C.14 is quadratic and defines the parabolic distribution of shear

stress shown in Figure C.2.
UNIT LOAD METHOD TO CALCULATE

SHEAR DEFLECTION – PART 1
The simple beam method assumes that the maximum shear stress is related
to shear strain via the shear modulus, and hence deflection may be calcu-
lated based on integrating shear strains along the centroid. Therefore, the
form factor for shear is effectively 1.5. However, using closed form solu-
tions and finite element analysis, Renton (1991), Schramm et al. (1994),
Pilkey (2003) and Iyer (2005) have all found that form factor should be
approximately 1.2 for rectangular beams where b < H. Therefore, the sim-
ple beam method is not accurate.
Gere & Timoshenko (1991, pp. 691–694) provided a derivation for shear
deflection for a uniformly loaded beam with simple supports, using the unit
load method. This is reproduced here but with the load changed to a point
load to obtain the correct equation for shear deflection for our case. Bhatt
(1999) used a different method of derivation but arrived at the same result.
The unit load method is based on the principle of virtual work, where
the external work is equal to the internal work. A unit load is applied to the
structure, in our case as a point load at the midspan, such that:
Wext = 1.∆ (C.15)
Wext is the external work

1 is the unit load
Δ is the midspan deflection due to the actual loads, which is what we
want to calculate
The unit load produces reactions at the supports and stress resultants within
the beam, in our case a moment MU and a shear force VU . These stress resul-
tants perform the internal virtual work due to this virtual deformation.
Figure C.4 Virtual flexural and shearing deformations of an element of the beam.
However, and this is the clever bit, we choose the virtual deformations
to be the same as the actual deformations that occur in the real beam with
the real point load. The virtual deformations of an internal element of the
beam are shown in Figure C.4.
The normal stress σ may be obtained from the flexure formula:
MU y
σ = (C.16)
I
And the shear stress τ may be obtained from the shear formula:
VU Q
τ= (C.17)
Ib
The extension strain ε due to the actual moment in the structure ML is

given by:
ML y
ε= (C.18)
EI
And the shear strain γ due to the actual shear force in the structure VL is
given by:
VLQ
γ = (C.19)
GIb
The internal virtual work of the stresses σ and τ acting on the sides of the
differential element is:
dWint = (σ dydz )( ε dx ) + (τ dydz )(γ dx ) (C.20)
Substituting for σ , τ, ε and γ from Equations C.16 to C.19 gives:
MU ML y 2 V V Q2
dWint = dxdydz + U 2L 2 dxdydz (C.21)
EI 2 GI b
Integrating Equation C.21 through the whole volume of the beam

gives:
MU ML y 2 VUVLQ2
Wint =
∫ EI 2
d x d y dz +
∫ GI 2b2
d x d y dz (C.22)
At any given cross-section of the beam, MU , ML , VU , VL , E, G and I are
constants, so we can simplify the integrals to an integration over the cross-
sectional area (denoted A) and an integration over the length of the beam
(denoted L):
MU ML  2  VUVL  Q2 
Wint =
∫
L
EI 2 
A
∫
 y dydz  dx +


∫
L

∫
GI 2  b2
A
dydz  dx


(C.23)
By the definition of the second moment of area I, we know that:
 
∫
 y 2dydz  = I
A


(C.24)
The second bracketed integral in Equation C.23 is also solely a property of

the cross-section dimensions of the beam. We can therefore define a prop-
erty called the ‘form factor for shear’ fs as follows:
A Q2
fs =
I2 ∫b
A
2
dydz (C.25)
The form factor for shear is dimensionless. Replacing the two bracketed
integrals in Equation C.23 with I and fs I 2 /A respectively we get:
MU ML fsVUVL
Wint =
∫
L
EI
dx +
∫
L
GA
dx (C.26)
By equating internal work in Equation C.26 with external work in Equation

C.15 we obtain the equation of the unit load method:
MU ML fsVUVL
∆=
∫ EI
dx +
∫ GA
dx (C.27)
FORM FACTOR FOR SHEAR FOR

A RECTANGULAR SECTION
The form factor for shear fs must be evaluated for every shape of cross-
section, using Equation C.25. For a rectangular section of width b and
height H, the first moment of area Q is given by Equation C.13 as:
b  H2 
Q=  − y2  (C.28)
2 4 
We can also calculate A /I 2 as:

A bH 144
= 2 =
I 2
 bH 
3 bH 5 (C.29)
 12 
And Q2 may be obtained by squaring Equation C.28:

2 2
 b  H2   bH 2 by 2  b2 H 4 b2 H 2 y 2 b2 y 4
Q = 
2 − y2   =  − = − + (C.30)
2  4   8 2  64 8 4
And Q2 /b2 is given by:
Q2 H 4 H 2 y 2 y 4
= − + (C.31)
b2 64 8 4
Now inserting these values into Equation C.25 gives:
H /2
A Q2 144  H 4 H 2 y2 y4 
fs =
I2 ∫
A
b2
dydz =
bH 5 ∫
− H /2

64
−
8
+  b.dy
4
(C.32)
Integrating:
H /2
144b  H 4 y H 2 y3 y 5 
fs = − + (C.33)
bH 5  64 24 20  − H /2
Now applying the limits:
144b  H 5 H 5 H 5   H 5 H 5 H 5  
fs =  − +  − − + −  (C.34)
bH 5  128 192 640   128 192 640  
Simplifying:
144  H 5 H 5 H 5  6
fs =  − + = (C.35)
H 5  64 96 320  5
As mentioned previously, form factor for shear is dimensionless, and for

a rectangular section we have just proved it is equal to 1.2. This is the
same value as found by other sources using a variety of different meth-
ods (Renton, 1991; Schramm et al., 1994; Bhatt, 1999; Pilkey, 2003;
and Iyer, 2005). Some of those methods also considered the variation of
shear stress across the width of the beam due to Poisson’s ratio effects,
but for b < H the effect is negligible. Since for buildings the thickness
of the wall will always be less than the height, this is a satisfactory
assumption.
UNIT LOAD METHOD TO CALCULATE

SHEAR DEFLECTION – PART 2
Continuing with the unit load method, the actual stress resultants from
x = 0 to x = L /2 are given by:
Px
ML = (C.36)
2
P
VL = (C.37)
2
For a unit load, i.e. P = 1, the stress resultants are:
1x
MU = (C.38)
2
1
VL = (C.39)
2
Inserting these stress resultants into the unit load equation gives:
L /2 L /2
MU ML fsVUVL 2 Px2 2f P
∆=
∫ EI
dx +
∫ GA
dx =
EI ∫
0
4
dx + s
GA ∫ 4 dx (C.40)
0
Note that the terms are multiplied by 2 because the integration is only from
0 to L /2, i.e. only half the beam. We did this because the moment and shear
force equations we have defined are only valid from 0 to L /2.
Integrating:
L /2 L /2
2  Px3  2f  P  2  PL3  2fs  PL 
∆= + s   =  +   (C.41)
EI  12 0 GA  4 0 EI  96  GA  8 
Simplifying:
PL3  12fs EI 
∆= 1+ (C.42)
48EI  GAL2 
Equation C.42 incorporates both bending deflection in the first term,

and shear deflection in the second term and is appropriate for any cross-
section simply-supported beam with a centrally-applied point load. If
we substitute for fs = 6/5, which is the value calculated for rectangular
section beams, we get the following equation, which is Equation 13.5 in
Chapter 13:
PL3  72EI 
∆=  1 +  (13.5)
48EI 5GAL2 
Quod erat demonstrandum.
COMPARISON WITH BURLAND

& WROTH (1974)
If we compare Equation C.42 with the following equation from Burland

& Wroth (1974), we can see that the correct shear deflection we have just
derived has a lower coefficient of 72/5 (= 14.4) compared to 18 in Burland
& Wroth (1974).
PL3  18EI 
∆= 1 + 2  (C.43)
48EI  L HG 
Burland & Wroth (1974) also assumed A = H, i.e. that the calculation is
per unit width. This is fine as long as the second moment of area I is also
per unit width (i.e. b = 1 m), and the point load P is also expressed per unit
width. This is useful because it means we do not have to know the thickness
of the wall. Alternatively, we can include the wall thickness and Equation
16.43 can be expressed as:
PL3  18EI 
∆=  1 + 2  (C.44)
48EI L AG 
Effectively, the form factor for shear fs is equal to 1.5 in Burland &
Wroth’s equation (Equation C.43). Shear deflections calculated using
Equation 13.5 will be 1.2/1.5 = 0.8 times smaller than those calculated
using Burland & Wroth’s equation. This means that higher strains will
be generated as a higher point load is needed to achieve the same green-
field deflection ratio.
Our new, correct version of this equation (Equation 13.5) was used in
Chapter 13 to derive the following equations for the maximum bending
strain and the maximum diagonal strain:
∆/L
ε bmax =
L  72EI  (13.12)
12d 1 + 5L2 AG 
∆ /L
ε dmax =
 AGL2 4  (13.23)
 18EI + 5 
As L /H varies, the relative importance of maximum diagonal and bending

strains will vary. Since the maximum tensile strain will be the greater of
either the maximum diagonal strain or the maximum bending strain, it can
be of interest to know which is greater. If we assume the following so that
the graphs we will plot will be similar to those of Burland & Wroth (1974):
Unit width, b = 1 and A = H.

Neutral axis at foundation level for hogging, so that I = H 3 /3 and d = H.
Take Poisson’s ratio as 0.3, such that E /G = 2.6.
Then we can simplify the equations to:
∆ /L L H
= + 1.04 (C.45)
ε bmax 12H L
∆ /L L2 4
= + (C.46)
ε dmax 15.6H 2 5
Burland & Wroth (1974) also did this and derived the following equations:
∆ /L L H
= + 1.3 (C.47)
ε bmax 12H L
∆ /L L2
= +1 (C.48)
ε dmax 15.6H 2
By setting either ε bmax = ε lim or ε dmax = ε lim we can now plot these equa-
tions against L /H , as Burland & Wroth (1974) did. This is shown in
Figure C.5.
Figure C.5 shows that the corrected equations give lower values of
( ∆ /L) /ε lim at all values of L /H and for both bending strains and diagonal
strains. This is because at the same deflection ratio ∆ /L the corrected equa-
tions will give higher values of ε bmax and ε dmax .
Figure C.5 Relationship between ( ∆ /L ) /ε lim and L /H for buildings modelled as isotropic

rectangular cross-section beams in hogging with the neutral axis at founda-
tion level.
Figure C.5 also shows that the corrected equations do not change the
point, at exactly L /H = 1.3, where diagonal strains cease to dominate and
bending strains begin to dominate.
We can also look at sagging if we assume the following:
Unit width, b = 1 and A = H.

Neutral axis at centroid, so that I = H 3 /12 and d = H /2.
Take Poisson’s ratio as 0.3, such that E /G = 2.6.
Then we can simplify the equations to:
∆/L L H
= + 052 (C.49)
ε bmax 6H L
∆/L L2 4
= + (C.50)
ε dmax 3.9H 2 5
Burland & Wroth (1974) also did this and derived the following equations:
∆ /L L H
= 0.167 + 0.65 (C.51)
ε bmax H L
Figure C.6 Relationship between ( ∆/L ) /ε lim and L /H for buildings modelled as iso-
tropic rectangular cross-section beams in sagging with the neutral axis at
mid-height.
∆/L L2
= 0.25 2 + 1 (C.52)
ε dmax H
These are plotted on Figure C.6, which shows that the corrected equa-
tions give lower values of ( ∆ / L ) /ε lim at all values of L /H and for both
bending strains and diagonal strains. This was also the case for hogging,
but here in sagging the difference is much more marked. The points at
which diagonal strains cease to dominate and bending strains begin to
dominate are further left than for the hogging case and are not at the
same value of L /H for the corrected equations and Burland & Wroth
(1974).
THE EFFECT OF VARYING E/G RATIO
In Figures C.5 and C.6, we assumed that Poisson’s ratio ν = 0.3 and
therefore E/G = 2.6. However, the effective E /G ratio can vary a lot in
real buildings. Burland & Wroth (1974) describe situations where the
longitudinal stiffness of the ‘building as a beam’ can be much higher
than the shear stiffness, for instance a building with lots of openings in
Figure C.7 The effect of varying E /G ratio for buildings in hogging.
it, and situations where the shear stiffness can be higher than the longi-
tudinal stiffness, for instance a wall made of precast concrete elements
connected with dowels.
Burland & Wroth (1974) used limiting values of E/G = 12.5 for
a building with high longitudinal stiffness but low shear stiffness,
and E/G = 0.5 for a building with low longitudinal stiffness but high
shear stiffness. These values have been used with our corrected equa-
tions, and the limiting tensile strains are shown in Figure C.7 for
hogging only.
Figure C.7 shows that the best situation for minimising tensile strains
is to have a building that is flexible in shear, i.e. to have a high E /G ratio.
Also of note in Figure C.7 is that as E /G increases, the diagonal strains
become more important than bending strains over a larger range of L /H,
up to around L /H = 6.3.
Figure C.8 shows the effect of varying E /G ratio for buildings in sag-
ging. Similar to Figure C6.7, it shows that the best situation for minimising
tensile strains is to have a building that is flexible in shear, i.e. to have a
high E/G ratio. When E /G = 0.5, bending dominates at all except very low
values of L /H. When E /G = 12.5, diagonal strains dominate up to around
L /H = 3.2.
Figure C.8 The effect of varying E /G ratio for buildings in sagging.
REFERENCES
Bhatt, P. (1999). Structures. London, UK: Addison-Wesley Longman Ltd.

Burland, J. B. & Wroth, C. P. (1974). Settlement of buildings and associated dam-
age – state of the art review. Conf. on Settlement of Structures, Cambridge,
pp. 611–654. London, UK: Pentech Press.
Gere, J. M. & Timoshenko, S. P. (1991). Mechanics of materials, 3rd S. I. Edition.
London, UK: Chapman & Hall.
Iyer, H. (2005). The effects of shear deformation in rectangular and wide flange
sections. MS thesis, Virginia Polytechnic Institute and State University.
Pilkey, W. D. (2003). Analysis and design of elastic beams. New York, USA: John
Wiley and Sons.
Renton, J. D. (1991). Generalized beam theory applied to shear stiffness. Int. J.
Solids Struct. 27, 1955–1967.
Schramm, U., Kitis, L., Kang, W., & Pilkey, W. D. (1994). On the shear deforma-
tion coefficient in beam theory. Finite Elem. Anal. Des. 16, 141–162.
Timoshenko, S. (1957). Strength of materials – Part I, 3rd edition. London, UK:
D van Nostrand.
Index
3D effects 5–6 C
numerical modelling of 224–228
canopy tubes, see pre-support
carbon footprint 249, 324
A cast-in-place (CIP) lining 247–248
advance length 389–390 cast iron 247, 263–5
characteristic grain size/diameter
96–97
B
characteristic values
Battersea UKPN Cable Tunnel 454 for soil strength parameters
bedded beam models 184–186, 151–157
304–305 for steel fibre-reinforced concrete
beta factor method 198 254–258
binocular caverns 386–387 classes of prediction 222–223
Blackwall Tunnel 112 Coccia et al. (2015) method
blow-out 21; see also drained blow-out to account for shear in steel
undrained blow-out fibre-reinforced concrete
boiling see stability of shafts hydraulic design 315
failure during excavation coefficient of earth pressure at rest 3,
bolts in segmental linings, 25, 155, 165, 198, 201–204,
see installation loads 209, 227, 308–309, 521
in segmental linings collapse, see stability; see also
boundary conditions, see numerical Heathrow Express collapse at
modelling Central Terminal Area
building damage assessment composite sprayed concrete lining, see
categories/classification system sprayed concrete composite
467–468 shell lining system
stage 1 assessment 469–471 constant volume behaviour, see
stage 2 assessment 472–490, undrained soil behaviour
537–551 constitutive models, see numerical
stage 2b assessment 491–508 modelling
stage 3 assessment 508–511 contraction 8, 433
stages 467–469 convergence, see lining displacements
bulk modulus 7–8, 211–215 convergence-confinement method
bursting stresses, see segmental 159, 195–203, 222,
lining joints; see also 401–403
jacking loads core softening 204
553
554 Index
crack mouth opening distance E

(CMOD) 252–254
crack widths earth pressure balance (EPB) tunnel
in buildings 468 boring machines (TBMs),
in steel fibre-reinforced concrete see drained stability,
(SFRC) 248–249, 252, 254, for EPB TBMs
259–260, 262, 315–317, 334 Edsåsdalen 47–48
critical stability ratio 49–50 effective stress principle 2–3,
Crossrail 10–12, 386, 431, 436–438, 7–8, 15
451, 455–456, 459, 469, elephant’s feet 382
508–509 elliptical distortion, see ovalisation
Curtis-Muir Wood solution xix, embedment depth of shaft lining
166–183, 533–536 133–134
applied to shafts 184 embrittlement 251
effect of lining stiffness 270 empirical design, see precedent
summary 186 practice
use in validation of numerical erection of segmental linings 349
modelling 217–219 errors in established methods xvi
curved joints see segmental lining Muir Wood (1975) analytical
joints solution 167, 179, 271
simple beam method of Burland &
Wroth (1974) 473–475
D wedge-prism method error in wedge
deflection hardening/softening forces diagram 527
250–251 wedge-prism method mislabelling
demoulding segments 324–328 of F0 nomogram 89, 529–530
depressurisation, see dewatering Eurocode 7 151–158; see also stability
design model 1, 45, 191–193, 240, of shafts
307–309 excess pore pressure, see undrained soil
dewatering 15, 22, 385, 387, 393, 451 behaviour; see also effective
dilation 8, 414, 433 stress principle; numerical
Docklands Light Railway 21, 67–68 modelling groundwater
drained blow-out 21, 103 explosive spalling 261
due to hydraulic fracturing 111–113
due to passive failure 103–111 F
summary 113–114
drained soil behaviour 8, 15, 23–24, face dowels 393
83–84, 212–215, 388, finite difference method 192;
390–393, 427, 441–443 see also numerical modelling
drained stability finite element method 192;
centrifuge tests and finite element see also numerical modelling
modelling 91–92 fire resistance 261
for closed-face TBMs 93–103 flat joints, see segmental lining joints
in dry cohesionless soils 84–85 forepoling, see pre-support
in dry soils with cohesion 85–87 four arc profile 377–379
for earth pressure balance (EPB) front-to-back arching 6
tunnel boring machines full-face excavation 380–381
(TBMs) 100–103
with groundwater 87–90, 93–103 G
overview 22–24, 83
for slurry tunnel boring machines gap method 202
(TBMs) 93–100 gasket compression, see installation
summary 92–93 loads in segmental linings
Index 555
gasometer blow-out 112 gasket compression and bolt loads

ground improvement 390 357–359
ground movements due to shaft grout pressures 355–357
construction 450–461 overview 349
ground movements due to tunnelling, permanent effects 356–357
see surface settlements; TBM jacking loads 349–354,
see also subsurface ground 356–357
movementsgrout pressure
method 202–204 J
grout pressures
effect on ground movements jacking loads 349–354
10–13, 32–33 bursting stresses at circumferential
for numerical modelling of, joints 299–301
see grout pressure method Jaky’s formula 3, 141
for segmental lining design, joints
see installation loads in in segmental linings, see segmental
segmental linings lining joints
in sprayed concrete, see sprayed
H concrete, construction joints
Jubilee Line 29–32, 302
Handling, see transportation and Jubilee Line Extension at St James’s
handling of segments Park 9–10, 13–15, 17–18,
Heathrow Baggage Tunnel 235 223, 422–423, 432–435,
Heathrow Cargo Tunnel 29–32 491, 501
Heathrow Express Junctions, see numerical modelling
collapse at Central Terminal Area
25–27 K
Terminal 4 Station 5–6, 25–26,
34–41 K0, see coefficient of earth
Trial Tunnel 3–4, 16, 27, 434–435, pressure at rest
450–453, 458 Kirsch solution 174, 231–239
Heathrow Terminal 5 project 397–398 Kobe Municipal Subway 33
Storm Water Outfall Tunnel Kwikastrip, see sprayed concrete,
(SWOT) 381 construction joints
horizontal surface displacements
13–15, 443–445 L
Hull wastewater flow transfer tunnel
21, 114 Lee Tunnel 248, 259
hydraulic failure, see stability of shafts, Leonhardt’s equations 300
hydraulic failure during limit equilibrium solutions 45, 47, 83;
excavation see also wedge-prism method
hydraulic fracturing, see blow-out limit of proportionality (LOP)
hyperstatic reaction method, see 252–256, 301, 338
bedded beam models lining
displacements of 24–25
I stresses in 28–41; see also sprayed
concrete
in situ concrete lining, see cast-in-place lipping, see misalignment of segment
(CIP) lining joints
in situ stress 2, 24–25, 28, 34, 39–41, load factor approach
99–101, 155, 160–161, 184, 418–427, 457
194–195 London Power Tunnels (LPT)
installation loads in segmental linings 453–456
556 Index
M overview 272–274
plywood and MDF 282–283
mesh density, see numerical modelling partial factors 139–140, 157
misalignment of segment joints for concrete in fire 261
circumferential joints 352 for drained stability 96
curved joints 302 for dynamic loads 348–349
radial joints 273–274 for fibre-reinforced concrete in
moment-axial force interaction tension 336, 338
diagram 308–309 for hydraulic heave 126–128
for steel fibre-reinforced concrete for jacking loads 352–353
309–314 for lining internal forces 308–309
for uplift 134–135
N passive failure, see blow-out
permeability
Northern Line Extension project 455 of lining 18, 432–435
numerical modelling of soil 15, 45, 432–435
advancing tunnel in 3D 224–228 pilot tunnels 380, 384–386, 393, 431
boundary conditions at edges of the pipe arch, see pre-support
model 205–208 piping 21, 114
boundary conditions at the tunnel plain concrete 262–263
perimeter 193–205 plasticity solutions
boundary distances 208–209 for convergence (cavity expansion)
building response 509–511 196, 422–426
constitutive models 192, 223 for stability 46–47, 49–50
element types 209–210 pocket excavation 388
mesh density and refinement pore pressure
210–211 in drained stability 84, 93–94,
modelling groundwater 211–216 96, 98–103
modelling junctions 231–239 in hydraulic failure of a shaft
modelling segment joints 305–307 124–128
modelling the tunnel lining in hydraulic fracturing 68, 111–112
228–230 in long-term ground movements
philosophy 191–193, 240 431–432
sprayed concrete 396–403 in long-term heave under a shaft
summary 240 base slab 145–146
validation and error checking in soil behaviour 2, 15–18, 23,
216–223, 402–403 28–30, 40–41
in undrained stability 48
O in uplift failure of a shaft 134;
see also numerical modelling
ovalisation 286; see also squatting groundwater
overconsolidated soils 3, 16, 18, 25, pre-support 390–393
48, 141, 433 precedent practice, design based on
33–34, 47, 165–166
P primary lining, see sprayed concrete,
primary and secondary
packers linings
bituminous and plastic
281–282
R
comparison of different types
283–285 random field method 155–157
linear elastic 274–281 random finite element method,
nonlinear 281–283 see random field method
Index 557
reinforced concrete ground movements, see ground

in framed buildings 465, 500–503 movements due to shaft
shaft base slab 139 construction
tunnel lining 184, 248, 301, stability, see stability of shafts
317, 364 shear modulus 7–8, 212–214
residual flexural tensile strength shell elements 228–230, 305–307
252–257 Shigino Route cable tunnel 32–33
in fire 261 shotcrete see sprayed concrete
rigid-plastic stress block approach 335 sidewall drift 380, 384–385
ring closure distance 58, 382, simple wished-in-place equilibrium
389–390 160–165
ring roll, see stagger of radial joints Singapore Circle Line 112
rotational rigidity, see segmental slurry infiltration
lining joints during excavation 98–100
during tunnel boring machine
S (TBM) standstills 96–98
slurry tunnel boring machines (TBMs),
scale of fluctuation 155–157 see drained stability, for
Second Heinenoord Tunnel 12–13, slurry TBMs
99–100, 111–112, 114, 203 SMART tunnel 93, 112
secondary lining, see sprayed concrete, softening and erosion, see blow-out
primary and secondary spheroidal graphite iron (SGI),
linings see cast iron
seepage force in tunnel heading spiles, see pre-support
stability 22–23, 83–84, sprayed concrete
86–87, 90, 100–103, 114 composite shell lining system 365,
in dewatering for sprayed concrete 367–369
tunnels 393 construction joints 393–395
in shaft stability 123–128 double-shell lining system 365–367
segmental lining joints face division 379–386
in bedded beam analysis 304–305 finishing layer 370
bursting stresses 299–301 initial layer 369–370
crushing and shear 294–299 lining design 363–364
curved joints 301–303 lining material 247
effect of number of joints 270–271 numerical modelling 396–403
effect of radial joint stagger primary and secondary linings
271–272 364–370
local forces due to joint rotation profile 369–379
286–299 reference books xvii–xviii
in numerical analysis 305–307 regulating layer 370
in plane strain analyses 270–272 single-pass lining 365
rotational rigidity of flat joints single-shell lining 365–366
272–285 tolerances 395
summary 317 toolbox measures 387–393
segmentation 270–2 traditional approach 364–6
serviceability limit state (SLS) design squatting 24–5, 34, 201, 354;
for in-service loads 315–317 see also lining displacements;
for transient loads 339–340 ovalisation
settlements, see surface settlements stability
shafts causes of instability 22–24
global design, see Curtis-Muir consequences of instability 20–22
Wood solution applied to decisions dependent on stability 22
shafts geometry of failure 23–24
558 Index
numerical modelling of stability overview 8–9, 411

63–65; see also drained transient 10–11, 427–431
stability transverse 9–10, 411–418; see also
overview 19–20 ground movements due to
stability of shafts shaft construction;
undrained stability volume loss
stability of shafts
base heave failure 131–134 T
hydraulic failure during excavation
123–130 target volume loss method
long-term heave 145–146 199–202
overview 123 TBM jacking loads, see installation loads
summary 146 temporary invert 382
uplift failure after base slab three arc profile 371–372, 376–377
construction 139–145 Thunder Bay Tunnel 18–19, 32–33
uplift failure during excavation top heading–bench–invert excavation
134–139 57–59, 380, 383–384
stability ratio 48–49 top heading–invert excavation
stacking segments, see storage of 380–383
segments transportation and handling of
stagger of radial joints 271–272, 306 segments 348–349
steel fibre-reinforced concrete trinocular caverns 387
benefits 248–249 tunnel construction xvii–xviii
design assisted by testing 252–254 twin sidewall drift see sidewall drift
design for shear 315 two arc profile 372–373
determining characteristic mean
strength values 257–258 U
determining characteristic strength
values 254–257 ultimate limit state (ULS) design for
durability 258–259 in-service loads (statically
fire resistance 261 indeterminate load cases)
material behaviour 250–252 307–315
sources of design guidance for transient loads (statically
249–250 determinate load cases)
watertightness 259–60 333–339
stepping, see misalignment of undrained blow-out 21
segment joints due to hydraulic fracturing
storage of segments 328–348 67–69
Storebælt Eastern Railway Tunnel 21, due to passive failure 69–74
301–302 due to softening and erosion
strains in the ground 66–67
curvature 448–450 summary 76–77
horizontal 446–447 undrained soil behaviour 7–8, 23–24,
slope 448 212–215
structural synthetic fibre undrained stability
reinforcement 262 in homogeneous clay 49–59
subsurface ground movements 3–6, 19 numerical modelling of 63–65
in drained soils 441–443 with overlying coarse-grained
in undrained soils 435–441 soils 61–63
surface contraction 204 overview 22–24, 45–47
surface settlements with undrained shear strength
long-term 15–19, 431–435 increasing with depth
longitudinal 11–13 59–61
Index 559
uplift failure W
of shaft, see stability of shafts
of tunnel heading invert 76 watertightness 259–260
wedge-prism method 83,
87–92, 104, 519–532
V Wood and Armer method
vacuum lifting 324–325 229–230
variable density slurry tunnel
boring machine Y
(TBM) 112–113
volume loss 418–427 Yodo River cable tunnel 32–33

Benoît Jones - Soft Ground Tunnel Design-CRC Press (2022)

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Benoît Jones - Soft Ground Tunnel Design-CRC Press (2022)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Benoît Jones - Soft Ground Tunnel Design-CRC Press (2022)

Uploaded by

Copyright:

Available Formats

Soft Ground

Library of Congress Cataloging‑in‑Publication Data

Names: Jones, Benoît, author.

ISBN: 978-0-367-41959-2 (hbk)

1 Real tunnel behaviour 1

5 Stability and Eurocode 7 151

6 Global design using analytical solutions 159

6.3.6 ‘No slip’ - full shear interaction

7 Global design using numerical modelling 191

7.10.3 Modelling junctions 231

9 Segmental lining design269

9.3.2 Using eccentricity of hoop force

10 Segment design for transient loads323

11 Sprayed concrete lining design363

11.6.3 Linear elastic – perfectly plastic sprayed

12 Estimating ground movements411

13 Estimating building damage467

13.4 Stage 3 assessment 508

Appendix A: Derivation of wedge-prism method519

Appendix B: Details from derivation of Curtis–Muir

Appendix C: Derivation of the deflection of a rectangular

The following books may be useful to gain an appreciation of tunnel con-

INTRODUCTION TO TUNNEL CONSTRUCTION

Chapman, D., Metje, N. & Stärk, A. (2010). Introduction to tunnel

BS 6164: 2019. Health and safety in tunnelling in the construction

‘Mechanized tunnelling’ means tunnelling using a tunnel boring machine

Guglielmetti, V., Grasso, P., Mahtab, A. & Xu, S. (2008). Mechanized

Maidl, B., Herrenknecht, M., Maidl, U. & Wehrmeyer, G. (2012).

Sprayed Concrete for Ground Support is published by a product supplier

BASF (2014). Sprayed concrete for ground support. BASF Construction

Alun Thomas’s book is both an introduction to sprayed concrete tunnelling

Thomas, A. H. (2019). Sprayed concrete lined tunnels, 2nd Edition.

With heartfelt gratitude to all,

Dr Benoît Jones has two decades of experience in tunnelling as a designer,

Real tunnel behaviour

Even the most sophisticated design model is an idealised representation

• how the ground moves towards the face of an advancing tunnel

1.1 IN SITU STRESS STATES

σ is the total stress in kPa

σ v is the vertical total stress in kPa

σ v = γ 1z1 + γ 2 z2 + γ 3z3 + … + γ n zn (1.3)

γ 1, γ 2 , γ 3 … γ n are the bulk unit weights of soil layers 1, 2, 3 … in

and vertical effective stress is described by the coefficient of earth pressure

K0 is the coefficient of earth pressure at rest

In ‘normally consolidated’ soils, where the maximum vertical effective

φ ′ is the soil’s angle of friction in a drained condition

Overconsolidated soils, where the soil has experienced a higher effective

1.2 OVERVIEW OF TUNNEL BEHAVIOUR

During excavation of a tunnel, a void is formed in the ground. In an open-

1.2.1 Undrained soil behaviour

Figure 1.4 Conceptual model illustrating undrained behaviour.

Figure 1.5 Volumetric strain contours from a centrifuge test by Marshall (2009),

1.2.2 Drained soil behaviour

1.3 MOVEMENTS OF THE GROUND SURFACE

The ground movements towards the face of the tunnel we saw in

Figure 1.6 Movements of the ground surface (based on Attewell et al., 1986).

1.3.1 Transverse vertical settlements

Figure 1.7 Short-term surface settlements measured after TBM passage at St James’s

1 Real tunnel behaviour 1

9 Segmental lining design269

10 Segment design for transient loads323

11 Sprayed concrete lining design363

12 Estimating ground movements411

13 Estimating building damage467

Appendix A: Derivation of wedge-prism method519

Figure 1.4 Conceptual model illustrating undrained behaviour.

Figure 1.5 Volumetric strain contours from a centrifuge test by Marshall (2009),

Figure 1.6 Movements of the ground surface (based on Attewell et al., 1986).

Figure 1.7 Short-term surface settlements measured after TBM passage at St James’s

Table 1.1 Percentages of final short-term settlement above the face

Figure 1.11 Transverse horizontal displacements after passage of the Westbound

Figure 1.12 Longitudinal horizontal displacements during passage of the Westbound

Figure 1.14 Long-term settlements due to the construction of Westbound (WB) and

Figure 1.15 Long-term settlement monitoring of a surface settlement pin over the

Figure 1.16 Collapse of the Munich Metro (Construction Today, 1994).

Figure 1.20 Typical lining displacements (exaggerated scale) for a sprayed concrete

Figure 1.21 Lining displacements measured in Heathrow Express CTA concourse tun-

Figure 1.22 Vertical and horizontal lining displacements in the Heathrow Express

1.7.1 How do tunnel lining stresses

Figure 1.23 Tunnel lining stress measurements in London Clay, up to 50 days

Figure 1.24 Tunnel lining stress measurements in London Clay, up to 600 days

Figure 1.25 Tunnel lining stress measurements in London Clay, up to 19.5 years

1.7.3 Heathrow Express Terminal 4 Station

Figure 1.26 Plan of tunnels at Heathrow Express Terminal 4 station, showing location

Figure 1.27 Cross-section of concourse and platform tunnels.

Figure 1.30 Main monitoring section schematically showing locations of pressure

Figure 2.1 Simplified geometry of a tunnel heading.