ETH Library

On the delayed failure of

geotechnical structures in low
permeability ground

Doctoral Thesis

Author(s):
Schuerch, Roberto

Publication date:
2016

Permanent link:
https://doi.org/10.3929/ethz-a-010872874

Rights / license:
In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection.
For more information, please consult the Terms of use.
 DISS. ETH NO 23681

ON THE DELAYED FAILURE OF GEOTECHNICAL STRUCTURES

IN LOW PERMEABILITY GROUND

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH

(Dr. sc. ETH Zurich)

presented by

ROBERTO SCHUERCH

MSc, ETH Zurich

Born on 08.10.1985

Citizen of

Lugano

accepted on the recommendation of

Prof. Dr. Georg Anagnostou

Prof. Dr. Donatella Sterpi

2016
2
 To the man who inspired me.
To my grandfather, Ing. Luigi Pini (1923-2014).

3
4
Acknowledgments
I would like to express my gratitude to Prof. Dr. G. Anagnostou (ETH Zurich), who has
supervised this PhD thesis and guided me through the work.
I am deeply grateful to Prof. Dr. D. Sterpi (Politecnico di Milano) for agreeing to be co-
examiner of the Thesis. Her comments were essential for the realisation of the experimental
tests.
I am grateful to Dr. P. Perazzelli (ETH Zurich) for his continuous support and
encouragement.
I am thankful to Dr. A. Vrakas (ETH Zurich) and Prof. F. Casini (University of Rome Tor
Vergata) for their detailed and constructive comments and to Mr. R. Poggiati (ETH Zurich),
Mr. P. Maspoli (ETH Zurich) and Mr. R. Iten (former student ETH Zurich) for their
contribution in the elaboration of the numerical analyses and in the execution of the
laboratory tests.
I would like to thanks Prof. Dr. J. Laue (Luleå University of Technologye) and Dr. E.
Pimentel (ETH Zurich). Their experience helped me in the conceptual development of the
experimental apparatuses. I acknowledge also Mr. M. Iten, Mr. A. Kieper and Mr. E. Bleiker
(IGT workshop) for their technical support in the realisation of the testing devices.
Finally, the Thesis evolved within the framework of the research project "Tunnel face
stability and tunnelling induced settlements under transient conditions". The support given to
this project by the Swiss Tunnelling Society (STS) and the Federal Road Office of
Switzerland (FEDRO) is greatly appreciated.

Zurich, December 2016

Roberto Schuerch

5
6
Abstract
This Thesis investigates the problem of time-dependent stability of geotechnical structures (such as
trenches or tunnels) in medium- to low-permeability water-bearing grounds, typically clayey or silty
soils. The peculiarity of these soils is that they respond to excavation with a delay. The time-
dependency can be traced back to the swelling process triggered by the dissipation of the excavation-
induced negative excess pore pressures. Unstable conditions may necessitate improvement or
reinforcement of the ground or the application of a support (e.g. by compressed air or pressurized
bentonite slurry in the case of tunnel face). As such measures may present economical and operational
disadvantages, the question of whether and for how long the excavation can remain stable without
support is of great practical relevance. The stand-up time (time lapsing between end of the excavation
and the occurrence of failure), and thus the feasibility of refraining from ground reinforcement,
improvement or support, depends essentially on soil strength and permeability.
The goal of the Thesis is to develop a computational method that allows the estimation of the stand-up
time, and thus improve construction safety and economy. The main objectives towards this goal are: (i)
analysis of the mechanism of delayed failure by means of fully coupled hydraulic-mechanical
continuum-mechanical simulations, investigation into the role of the constitutive behaviour of the
ground (particularly that of plastic dilation), and development of a practical method of dealing with the
numerical problem of mesh-sensitivity which occurs, due to the localization of deformations when
assuming non-associated plastic flow, in any geotechnical structure at failure; (ii) planning and
performing experiments and validation of the computational method and assumptions; (iii) systematic
investigation of the stand-up time of the tunnel face and working-out of design charts.
The equations governing the coupled hydraulic-mechanical process of delayed failure are solved
numerically with the finite element method. Emphasis is placed on the numerical manifestation of
failure, on the role of the plastic dilatancy and, finally, on the influence of the spatial discretization.
The Thesis shows that under the frequently made simplifying assumption of a constant positive dilation
angle, coupled analyses inevitably lead to a constant deformation rate at failure, while failure is
commonly associated with accelerating displacements. Models that allow for shearing under constant
volume lead to accelerating displacements, but inherently exhibit numerical stability problems close to
the failure state, because a solution satisfying the two balance equations does not exist. Moreover, the
assumption of non-associated plastic flow leads to a dependency of the stand-up time to the coarseness
of the mesh: the finer the latter, the lower the stand-up time. The Thesis shows, based upon analytical
computations, that the mesh-dependency of the stand-up time is mainly due to the structural softening
that occurs even in a perfectly plastic material if the flow rule is non-associated. In numerical
simulations, the shear strains become localized in a band which is about 1–2 finite elements thick. The
coarser the finite element mesh, the thicker the numerically predicted shear band, the smaller the shear
strains, the less pronounced the structural softening and the higher the predicted stand-up time. For
very coarse finite element meshes (which practically prevent structural softening) the stand-up time
tends to a constant value. Vice-versa, for sufficiently fine spatial discretization (allowing for
considerable structural softening up to that corresponding to purely plastic shearing), the stand-up time
depends almost linearly on the element size. This finding is valuable from the practical viewpoint, as it
allows determining stand-up time by performing a few computations with a relatively fine meshes and
extrapolating their results to the mesh size that corresponds to the expected, grain-size-dependent
thickness of the shear band (practically zero considering the dimensions of typical geotechnical
structures). In case of relatively simple problems an upper and lower bound of the numerically
computed stand-up time can be determined analytically.

7
The experimental part of the Thesis aims to observe delayed failure under controlled conditions and
consists of uniaxial loading tests and centrifuge tests (the latter simulate the excavation of an
underwater vertical cut in over-consolidated clay). The experimental results were interpreted also
numerically, using material constants which were determined by means of laboratory tests or based
upon theoretical considerations, independently from the delayed failure tests ("class A" model
validation). The numerical analyses show that, the simple, but widely used, linearly plastic, perfectly
plastic model with Mohr-Coulomb yield condition (MC model) and isochoric plastic flow generally
underestimates the stand-up time. The reason is that it does not account for the plastic volumetric
strains accompanying shearing, which temporarily perpetuate negative excess pore pressures, thus
delaying failure. The Modified Cam Clay model (MCC), which is known to overestimate strength on
the dry side of the yield surface, generally overestimates the stand-up time when used in combination
with computationally manageable spatial discretization (i.e. fine, but still not as much as it would be
necessary on account of the experimentally observed very thin shear bands). However, stand-up times
very close to the observed ones are obtained by extrapolating the numerical results to the size of the
shear band. In conclusion, the MCC model allows for a reasonably accurate prediction of the stand-up
time in a typical over-consolidated clay, while the MC model in combination with a fine (but still
manageable) discretization provides a conservative estimate. The conclusions are based upon the
numerical results obtained for the material constants proposed in the literature. To this respect it is
important to note that an inconsistency exists between the proposed Young’s modulus of the MC
model and the MCC parameters. The effect of the assumption about the Young’s modulus on the
results, and so on the conclusions of the present study, is significant (taking the value of the Young’s
modulus consistently with the MCC parameters would have led to a remarkably longer stand-up time).
Additional laboratory tests would be required in order to validate the proposed constants.
The findings arising from the comparison between numerical and experimental results made it possible
to investigate systematically the effect of the decisive geotechnical parameters on the time-dependency
of tunnel face stability. Based upon the results of a comprehensive parametric study (covering a wide
range of conditions), dimensionless design charts were worked-out which allow to estimate stand-up
time of shallow tunnels for a given geotechnical situation, thus assisting decision-making during
planning and execution in soft ground tunnelling.

8
Zusammenfassung
Diese Dissertation untersucht die zeitabhängige Stabilität von Geländeeinschnitten und Tunneln in
wasserführenden, tonigen oder siltigen Lockergesteinen mittlerer bis niedriger Durchlässigkeit. Die
Besonderheit dieser Lockergesteine besteht darin, dass sie auf den Ausbruch erst mit einer
Verzögerung reagieren. Die Zeitabhängigkeit ist durch den Konsolidationsvorgang bedingt, welcher
durch den ausbruchsbedingten, negativen Porenwasserüberdruck ausgelöst wird. Zur Vermeidung einer
Instabilität kann eine Bodenverbesserung oder -verstärkung bzw. das Aufbringen eines Stützdrucks
erforderlich sein (z.B. mittels Druckluft oder einer druckbeaufschlagter Bentonit-Stützflüssigkeit im
Falle der Tunnelortsbrust). Da derartige Massnahmen wirtschaftliche oder baubetriebliche Nachteile
mit sich bringen können, ist die Frage der Standzeit – definiert als die Zeit zwischen Ausbruch und
Instabilität – für die Praxis von grosser Relevanz. Die Standzeit und somit die Möglichkeit eines
Verzichts auf Verstärkungs-, Verbesserungs- oder Stützmassnahmen hängt im Wesentlichen von der
Festigkeit und der Durchlässigkeit des Lockergesteins ab.
Ziel der Doktorarbeit ist die Entwicklung eines Berechnungsverfahrens, mit dem sich die Standzeit
abschätzen, und somit die Sicherheit und Wirtschaftlichkeit der Bauausführung erhöhen lässt. Zu
diesem Zweck wurde zunächst das Phänomen verzögert auftretender Instabilitäten mittels hydraulisch-
mechanisch gekoppelter, kontinuumsmechanischer Simulationen rechnerisch interpretiert, der Einfluss
des Stoffgesetzes (insbesondere der plastischen Volumendilatanz) auf das Modellverhalten untersucht
und eine praktikable Methode zur Behandlung des numerischen Problems der Netzabhängigkeit bei der
Verformungslokalisierung, die bei jeder geotechnischen Struktur im Versagensfall auftritt, entwickelt.
Anschliessend wurden Experimente zur Validierung der Berechnungsverfahren und Annahmen geplant
und durchgeführt. Nach diesen grundlegenden theoretischen und experimentellen Arbeiten wurde das
Problem der Standzeit der Ortsbrust systematisch untersucht und Bemessungsdiagramme ausgearbeitet.
Die dem gekoppelten, hydraulisch-mechanischen Prozess der verzögerten Instabilität zu Grunde
liegenden Gleichungen werden numerisch mit der Methode der finiten Elemente gelöst. Besonderes
Augenmerk wird dabei auf die numerische Äusserung einer Instabilität, auf die Rolle der plastischen
Dilatanz und schliesslich auf den Einfluss der räumlichen Diskretisierung gelegt. In der Dissertation
wird gezeigt, dass unter der häufig getroffenen, vereinfachenden Annahme eines konstanten positiven
Dilatanzwinkels die gekoppelten Analysen unvermeidlich zu einer konstanten Verformungsrate beim
Versagen führen, wohingegen Instabilitäten gemeinhin mit sich beschleunigenden Verschiebungen
assoziiert werden. Modelle, die ein volumentreues Abscheren erfassen, ergeben sich beschleunigende
Verschiebungen, weisen jedoch inhärenterweise numerische Stabilitätsprobleme in der Nähe des
Versagenszustands auf, weil es keine Lösung gibt, die gleichzeitig die Gleichgewichts- und
Massenerhaltungsbedingungen erfüllt. Des Weiteren führt die Annahme eines nichtassoziierten
plastischen Fliessgesetzes zu einer Netzabhängigkeit der Standzeit: je feiner das Netz, desto geringer
ist die Standzeit. Auf der Grundlage analytischer Überlegungen wird nachgewiesen, dass die
Netzabhängigkeit der Standzeit hauptsächlich auf die sogenannte strukturelle Entfestigung
zurückzuführen ist, die selbst beim idealplastischen Materialverhalten auftritt, wenn die Fliessregel
nichtassoziiert ist. In numerischen Simulationen bilden sich die Scherverformungen in einem Band mit
einer Stärke von 1–2 finiten Elementen aus. Je gröber das Netz, umso breiter ist die numerisch
prognostizierte Scherfuge, umso geringer sind die Scherverformungen, umso weniger ausgeprägt ist
die strukturelle Entfestigung und umso höher ist die prognostizierte Standzeit. Bei sehr groben Netzen,
die eine strukturelle Entfestigung praktisch verunmöglichen, tendiert die Standzeit zu einer oberen
Grenze. Umgekehrt erlauben sehr feine Netze eine beträchtliche strukturelle Entfestigung bis hin zu
Werten, die rein-plastischen Scherverfomungen entsprechen, und die Standzeit tendiert zu einer
unteren Grenze. Bei einer relativ einfachen Problemstellung kann der obere und der untere Grenzwert

9
der numerisch ermittelten Standzeit analytisch bestimmt werden. Die numerischen Simulationen
zeigen, dass bei einer hinreichend feinen räumlichen Diskretisierung die Standzeit praktisch linear mit
der Elementgrösse abnimmt. Dieses Ergebnis ist nützlich, weil es die Extrapolation der Ergebnisse von
Berechnungen mit relativ groben Netzen erlaubt.
Der experimentelle Teil der Doktorarbeit umfasst einachsige Druckversuche mit verzögertem Bruch
sowie Zentrifugenversuche. Letztere simulieren den Aushub einer Unterwasserbaugrube in
überkonsolidiertem Ton. Die Versuchsergebnisse werden unter Verwendung von Materialkonstanten
numerisch interpretiert, die anhand von Laborversuchen Dritter oder basierend auf theoretischen
Überlegungen unabhängig von den Versuchen der vorliegenden Arbeit ermittelt wurden. Die
numerischen Analysen zeigen, dass das in der Praxis häufig verwendete, linearelastische
idealplastische Modell mit Mohr-Coulomb-Fliessflächenbedingung (MC-Modell) ohne plastische
Volumendilatanz die Standzeit im Allgemeinen unterschätzt. Der Grund hierfür ist, dass die plastischen
Volumendehnungen, die durch dieses Modell nicht berücksichtigt werden, den negativen
Porenwasserüberdruck temporär aufrechthalten und somit die Instabilität verzögern. Das Modified
Cam Clay-Modell (MCC-Modell), das bekanntermassen die Grösse der Fliessfläche auf der trockenen
Seite überschätzt, überschätzt im Allgemeinen die Standzeit, wenn es in Verbindung mit einer
rechnerisch handhabbaren räumlichen Diskretisierung eingesetzt wird (d.h. zwar fein, aber noch nicht
in dem Masse, wie es aufgrund der experimentell beobachteten, sehr dünnen Scherbänder erforderlich
wäre). Standzeiten, die den beobachteten Werten sehr nahe kommen, lassen sich jedoch durch
Extrapolieren der numerischen Ergebnisse auf die Dicke der Scherfuge erzielen. Zusammenfassend
erlaubt das MCC-Modell eine ausreichend genaue Vorhersage der Standzeit in typischen
überkonsolidierten Tonen, während das MC-Modell in Verbindung mit einer feinen (aber noch
handhabbaren) Diskretisierung eine konservative Schätzung liefert. Diese Schlussfolgerungen basieren
auf den Ergebnissen der vergleichenden numerischen Berechnungen, die mit in der Literatur
angegebenen Materialkonstanten durchgeführt wurden. Allerdings besteht diesbezüglich eine
Inkonsistenz zwischen dem vorgeschlagenen Elastizitätsmodul des MC Stoffegesetzes und den
Parametern des MCC Modells. Die Annahme über das Elastizitätsmodul beeinflusst erheblich die
Voraussagen über das Verhalten des MC Modells und somit auch die aus den vergleichenden
Berechnungen gezogenen Schlussfolgerungen; würde man statt des in der Literatur vorgeschlagenen
Elastizitätsmoduls einen Wert annehmen, der konsistent mit den Parametern des MCC Modells ist, so
würde man eine viel längere Standzeit erhalten. Zur Klärung der erwähnten Parameterinkonsistenz sind
weiterführende Laboruntersuchungen erforderlich.
Ausgehend von diesen Erkenntnissen wurde der Effekt der entscheidenden geotechnischen Parameter
auf die Standzeit der Ortsbrust systematisch untersucht. Aufgrund einer umfassenden Parameterstudie
wurden dimensionslose Bemessungsdiagramme erstellt, die eine rasche Abschätzung der Standzeit von
oberflächennahen Tunnels für gegebene geotechnische Bedingungen ermöglichen und somit eine
nützliche Entscheidungshilfe bei Planung und Durchführung von Tunnelvortrieben im Lockergestein
darstellen.

10
Table of contents
1. Introduction......................................................................................................... 13
1.1 Problem definition ....................................................................................................... 13
1.2 Current state of research .............................................................................................. 17
1.3 Objectives and structure of the Thesis ......................................................................... 19
1.4 Computational method ................................................................................................ 21

2. Structural softening ............................................................................................ 23

2.1 Introduction ................................................................................................................. 23
2.2 Reduction factor η ....................................................................................................... 25

3. Manifestations of delayed failure ...................................................................... 29

3.1 Introduction ................................................................................................................. 29
3.2 Problem setup and computational model..................................................................... 30
3.3 Mohr-Coulomb material .............................................................................................. 30
3.4 Coupled analysis for Modified Cam Clay material ..................................................... 48
3.5 Conclusions ................................................................................................................. 56

4. Mesh dependency ................................................................................................ 57

4.1 Introduction ................................................................................................................. 57
4.2 Underwater vertical cut ............................................................................................... 58
4.3 Biaxial problem ........................................................................................................... 64
4.4 Conclusions ................................................................................................................. 91

5. Uniaxial loading tests .......................................................................................... 93

5.1 Introduction ................................................................................................................. 93
5.2 Experimental study ...................................................................................................... 94
5.3 Numerical study......................................................................................................... 113
5.4 Conclusions ............................................................................................................... 128

6. Centrifuge tests ................................................................................................. 128

6.1 Introduction ............................................................................................................... 129
6.2 Centrifuge modelling ................................................................................................. 131
6.3 Limit equilibrium analysis ......................................................................................... 146
6.4 Numerical modelling ................................................................................................. 149
6.5 Mesh dependency ...................................................................................................... 159
6.6 Conclusions ............................................................................................................... 161

11
7. Stand-up time of tunnel face ............................................................................ 163
7.1 Introduction ............................................................................................................... 163
7.2 Computational model ................................................................................................ 163
7.3 Undrained and drained tunnel face stability .............................................................. 166
7.4 Tunnel face stability under transient conditions ........................................................ 170
7.5 On the transient tensile failure of the ground ............................................................ 174
7.6 On some factors influencing stand-up time ............................................................... 177
7.7 Design diagrams ........................................................................................................ 185
7.8 Application example .................................................................................................. 192
7.9 The influence of the support pressure on the stand-up time ...................................... 193
7.10 Critical advance rate .................................................................................................. 194
7.11 Conclusions ............................................................................................................... 197

8. Conclusions and Outlook ................................................................................. 199

Appendix A. Publications from the present Thesis............................................... 201

Appendix B. Water retention curve ....................................................................... 203

Appendix C. List of symbols ................................................................................... 207

9. References .......................................................................................................... 211

Curriculum Vitae ..................................................................................................... 221

12
 Chapter 1 – Introduction

1. Introduction

1.1 Problem definition

1.1.1 Delayed failure

Saturated, water-bearing ground responds to excavation with a longer or shorter delay

depending on its permeability and on the rate of loading (Atkinson and Mair, 1981). In soft
ground, this time-dependency can almost always be traced back to consolidation. The
response of high-permeability soils (e.g. sandy or gravelly soil) is practically instantaneous
(reaching steady state practically immediately), while low-permeability soils (e.g. clay)
exhibit a pronounced time-dependency, which is due to the transient seepage flow process.
The latter is triggered by excavation and develops slowly over the course of time: The long-
term deformations of the ground include, in general, changes to its water content, which take
more or less time, depending on the seepage flow velocity and thus on the soil permeability.
For problems involving “unloading” of the soil, the short-term behaviour (i.e. the behaviour
under the undrained1 conditions prevailing instantaneously after unloading) is more
favourable than the long-term one (i.e. the behaviour under the drained conditions prevailing
at the end of the consolidation process). The resistance of the soil to shearing is higher in the
short term than in the long term because excavation would lead to an increase in the volume
and in the water content of the soil, which cannot happen instantaneously; in the short term
negative excess pressures develop and temporarily maintain an approximately constant mean
effective stress.
This Thesis deals with the stability of excavations under transient conditions (lying
somewhere between undrained and drained conditions), where the dissipation of excess pore
pressure is not so fast that steady state is achieved already during excavation, but on the other
hand also not so slow that the conditions may be considered as practically undrained within
the relevant time period (which depends on the question under investigation). Transient
conditions are particularly relevant for soils of medium-low permeability (e.g. the widely-
distributed moraines in Switzerland). In soils of this type the excavation may be stable only
for a limited, not easily quantifiable time period; failure occurs with a delay. (A “delayed
failure” may be also induced by other events, like intense raining, surcharge at the surface or
an earthquake. However, these aspects fall outside the aim of the present work.)
The dominant role of the progressive dissipation of pore pressure in delayed failure in low-
permeability soils has been widely reported in the literature. Sevaldson (1956), Skempton,
(1964, 1977), Vaughan and Walbancke (1973), Hughes et al. ( 2007), Potts et al. (1997),

1
The notions of “undrained / drained conditions” are used here in the classic soil mechanics sense.
Drained conditions are established when the changes in stress are so slow with respect to the ability
of the soil to drain, that nearly no excess pore pressures develop. Undrained conditions are
characterized by a rapid stress change with respect to the ability of soil to drain (the water is
trapped in the soil pores).

13
Lollino et al. (2011) and Ellis and O’Brien ( 2007) – among others – present several examples
of delayed failure of cuttings, while Davis et al. (1980) remarked that the assessment of time
dependent face stability is a major concern for tunnelling engineers.
The behaviour of high- or low-permeability soils is well understood and can be analysed by
considering the borderline cases of drained or undrained conditions, respectively. For medium
permeability soils, however, very little research work has been done and there are hardly any
systematic investigations available. The simplifying assumptions involving drained or
undrained conditions may be uneconomical or unsafe, respectively, for such soils.

(a) (b)

Figure 1.1. Stable unsupported vertical cut (a) and tunnel face (b) in low permeability soil
(pictures after blog.geotechpedia.com and www.dr-sauer.com, respectively).

(a) (b)

Figure 1.2. Collapse of excavation (a) (after cliffs.lboro.ac.uk/downloads/kovacevic.pdf) and

of a tunnel face (b) in weak ground ( Lausanne subway, Switzerland; courtesy of
H. Stadelmann, Implenia AG).

14
 Chapter 1 – Introduction

1.1.2 Stand-up time

The understanding and the assessment of the time dependent stability of excavations in
medium-low permeability soils is relevant for both temporary (e.g. open-cut excavations or
tunnel face) and permanent geotechnical structures (e.g. road or railway cuttings). A major
point of interest is the assessment of the so-called stand-up time (i.e., the time lapsing
between the completion of the excavation, Fig. 1.1, and failure, Fig. 1.2): a long stand-up time
allows to refrain from face support or ground improvement works during temporary
excavations, thus saving time and cost, while a short stand-up time requires the
implementation of costly and time-consuming auxiliary measures (Fig. 1.3). Understanding
and quantifying the behaviour of medium-low permeability soils under transient conditions is
therefore essential for construction safety and economy.
The stand-up time of an excavation increases with the shear strength of the ground (all other
parameters being fixed). Figure 1.4 shows qualitatively the dependency of the stand-up time
on cohesion. The two values (cU,lim, cD,lim) mark the boundary of the transient failure regime. If
the cohesion is lower than cU,lim, then the excavation would be unstable even under undrained
conditions; stand-up time in this case is equal to zero. If, on the other hand, the cohesion is
higher than cD,lim, then the excavation would be stable also in the long term; for c  cD ,lim
stand-up time tends to infinity. The limit cohesion cD,lim represents, therefore, a singularity
point. The asymptotic increase in stand-up time indicates a considerable sensitivity of the
ground response with respect to cohesion at the upper end of the transient failure regime:
close to the limit cohesion cD,lim a relatively small variation ∆c may result in an infinite
change in stand-up time (from a finite value to infinity or vice-versa; Fig. 1.4).

(a) (b)

Figure 1.3. Tunnel face support, (a), by compressed air (entrance to the cutterhead for
hyperbaric interventions during the construction of the St Petersburg subway,
Russia; courtesy of Salini-Impregilo) and, (b), face bolting (bypass Patras,
Greece; courtesy of Prof. G. Anagnostou).

15
A cohesion change ∆c can be due to intrinsically softening behaviour (i.e. strength parameters
decreasing with shear deformation) or, in the case of non-associated plastic flow, due to the
so-called structural softening (the definition of structural softening was introduced by Sterpi,
1999, while this aspect of mechanical behaviour was already described by Davis, 1968).
Structural softening means an apparent reduction of the shear strength (to the so-called
“operational strength”) caused by the progressive rotation of the principal stress axes inside
the shear band during shearing. It occurs even if the material is perfectly plastic or even
slightly hardening (Rice, 1975). As in FEM analyses the shear deformation depends on the
element size (the smaller the element size, the larger the possible rotation), the change in
cohesion ∆c caused by structural softening is also a function of the element size. This means,
in conclusion, that for FEM transient analyses assuming a non-associated plastic flow (and so
affected by structural softening), the stand-up time will be mesh dependent (decreasing with
the element size). As discussed above on the basis of Fig. 1.4, a significant mesh sensitivity is
to be expected at cohesion values just above the limit cohesion cD,lim.

ts

∆c

cU,lim cD,lim c

collapse occurs during collapse occurs during collapse never occurs

undrained unloading (unstable swelling process (stable under drained
under undrained conditions) (unstable under transient conditions) conditions)

Figure 1.4. Qualitative dependency of the stand-up time ts on cohesion c.

16
 Chapter 1 – Introduction

1.2 Current state of research

Over the course of the last decades the topic of delayed failure has been investigated
numerically and by means of field and physical tests.

1.2.1 Numerical studies

Stability of geotechnical structures is traditionally investigated by means of limit equilibrium

models (e.g. Culmann, 1875, Taylor, 1937, Bishop, 1955, Michalowski, 2002 for the stability
of slopes or trench excavations and Anagnostou and Kovári, 1994, Davis et al., 1980 for the
tunnel face stability). These models are simple but fail to consider soil deformations, which
are essential to the time-dependency of ground behaviour in general and to the problem of
transient failure in particular. Einstein and Samarasekera (1992) investigated the time-
dependent stability of a tunnel cross-section in water-bearing soil by estimating the pore
pressure field first (using transient seepage flow analyses that do not consider soil
deformation) and incorporating it into a numerical stress analysis. This approach is adequate
for drained conditions, where the problem – although physically coupled – can also be
analysed as uncoupled because the steady state hydraulic head field does not depend on the
deformations. For transient problems, however, fully coupled hydraulic-mechanical analyses
are essential. Although the need for a coupled approach was pointed out already by Bjerrum
and Kjaernsli (1957), very few works investigate the issue of delayed failure by means of
coupled transient numerical analysis.

Table 1.1. Investigations into the issue of delayed failure of cutting or slopes employing
coupled transient numerical analyses.
Reference Problem Constitutive model
Holt and Griffiths (1992) Vertical cut Linearly elastic, perfectly plastic material; Mohr-
Griffiths and Li (1993) Vertical cut Coulomb failure criterion; zero plastic volumetric
strains
Kirkebø (1994) Slope Linearly elastic, perfectly plastic; smoothened
Mohr-Coulomb failure criterion; dilatant, isochoric
and contractant plastic flow
Potts et al. (1997) (selected Cut slope Linearly elastic, perfectly plastic or strain-softening
results are presented also in material; Mohr-Coulomb failure criterion; dilatant
Vaughan 1994 and Potts and isochoric plastic flow
and Zdravković 2001)
Banerjee et al. (1988) (1) Vertical cut Extension of the Modified Cam-Clay model for
anisotropic behaviour
Kovacevic et al. (2007) Cut slope Strain-softening, nonlinearly elastic Mohr-Coulomb
material; isochoric plastic behaviour
Borges (2008) (1) Cut slope Modified Cam-Clay model
Kovacevic et al. (2008) Cut slope Critical state model of the Cam-Clay family
Kovacevic et al. (2012) Underwater slope
Sanavia (2009) Slope Three-phasic, elasto-plastic material; Drucker-
Prager failure criterion, isotropic linear softening;
non-associated plastic flow rule

(1)
Without investigation of the numerical behaviour close to failure.

17
Table 1.1 provides an overview of numerical studies on the transient stability of slopes or
trenches. With the exception of Banerjee et al. (1988) (who did not investigate numerical
behaviour close to failure) and Borges (2008) (whose simulations did not reach the ultimate
state for the material constants adopted) the works of Table 1.1 allow the following
conclusions to be made: numerical simulations can reproduce a distinct failure under the
assumption of isochoric plastic flow, i.e. zero plastic volumetric deformation (Holt and
Griffiths 1992, Griffiths and Li 1993, Kovacevic et al. 2007); the failure occurs more
suddenly in the case of a isochoric or contractant flow (Kirkebø, 1994); and stand-up time
decreases with a decreasing dilation angle (Potts et al., 1997; Sanavia, 2009).
Failure of a structure over time is associated with accelerating displacements close to the
ultimate state (Vermeer and Van Langen 1989, Kovacevic 1994, Griffiths 1989; Griffiths and
Lane, 1999). However, as mentioned already by Griffiths et al. (1991), Kirkebø (1994) and
Potts et al. (1997), coupled transient analysis models do not exhibit such behaviour under the
assumption of a constant positive dilation angle: the numerical solution fulfils equilibrium,
while the displacements increase over time at a constant rate. The present Thesis, aims to
improve the understanding of the causes of this model behaviour and to show which
parameters must be evaluated in order to determine stand-up time.
Concerning tunnelling, to the author's knowledge, no systematic study of the stand-up time of
the tunnel face during standstills exists in the literature. Only Ng and Lee (2002) and Callari
(2015) addressed the issue of the time dependent face stability. Ng and Lee (2002) estimated
the necessary face reinforcement as a function of the consolidation time. The authors showed
that the dissipation of the negative excess pore pressure induced by the excavation leads to an
increase of the bolting density requirement. Callari (2015) discussed the influence of the
plastic dilation and on the hardening modulus on the stand-up time. The author showed that
positive plastic dilation and hardening behaviour are favourable with respect to the stand-up
time. Moreover, Callari (2015) observed that FEM analyses of tunnel face instability have
difficulties to capture the localization of the strains ahead of the tunnel face near to the
ultimate state.
Another issue which is strictly related to the stand-up time of the tunnel face is the minimum
(“critical”) advance rate for which the face would remain stable during ongoing excavation.
Qualitatively, the lower the stand-up time, the higher the advance rate required for stability.
The issue of face stability during ongoing excavation was addressed by Höfle et al. (2008,
2009) and by Sitarenios and Kavvadas (2016). Höfle et al. (2008) showed that the
deformations increase with increasing permeability but pointed out that quantitative
statements about face stability cannot be made and mentioned the difficulty of applying the
so-called “phi-c-reduction” method for calculating the safety factor under transient
conditions. Höfle et al. (2009) could demonstrate, however, by means of one parametric study
with permeability values k = 10-6 to 10-9 m/s (the advance rate is not given in the paper) that
the safety factor of the tunnel face decreases with increasing permeability. Sitarenios and
Kavvadas (2016) also concluded that the ground permeability has a major relevance on the
stability of the tunnel face (more specifically on the required support pressure). The authors
showed that for a fixed advance rate of 36 m/day, favourable undrained conditions around the
tunnel face occur when the permeability of the soil is lower than 5 10-7 m/s, while
unfavourable drained conditions prevail for a permeability higher than 10-6 m/s.

18
 Chapter 1 – Introduction

1.2.2 Field tests and physical modelling

Two major field tests were executed by Banerjee et al. (1988) and Cooper et al. (1998). The
authors, investigated the manifestation of delayed failure of a 9.75 m high vertical cut and of
a 9 m high slope with inclination 1:2, respectively, by increasing artificially the pore pressure
in the ground close to the potential sliding surface (by means of wells). The tests were fully
instrumented in order to monitor the evolution over time of the pore pressure (by means of
piezometers and/or tensiometers) and of the deformations (by means of inclinometers and
extensometers).
Field tests are useful for the understanding of the problem, but at the same time, are expensive
and time-consuming (requiring several months before reaching collapse) and, on account of
the big number of involved and partially unknown parameters, not reproducible. These issues
can be overcome using small scale physical models.
The only works dealing with the physical modelling of time dependent stability of slopes are
those of Davies (1985) and Deepa and Viswanadham (2009). Both authors investigated the
delayed failure of slopes in the centrifuge. Davies (1985) investigated the time dependent
stability of a slope due to an increase of the surface loading, while Deepa and Viswanadham
(2009) investigated the stability over time of a nail reinforced slope induced by an imposed
increase of pore pressure inside the model. To the author’s knowledge, no physical models
reproducing a delayed failure as defined in Section 1.1 (delayed failure under constant
boundary conditions and loads) exist.

1.3 Objectives and structure of the Thesis

The first objective of the Thesis is to investigate the influence of the constitutive behaviour of
the ground (particularly that of plastic dilation) on the mechanism of delayed failure by means
of fully coupled hydraulic-mechanical continuum-mechanical simulations, and the
development of a practical method of dealing with the numerical problem of mesh-sensitivity
which occurs in any geotechnical structure at failure due to the structural softening.
As structural softening is important for the topic under investigation, the Thesis starts with a
brief review of the literature about the influence of the non-associativity of the plastic
potential on the shear resistance. Furthermore, a reduction factor is derived, which expresses
the decrease in resistance as a function of the rotation angle of the principal axes inside the
shear band (Chapter 2).
Subsequently, Chapter 3 investigates how delayed failure manifests itself by considering the
time dependent stability of an underwater vertical cut. Although the problem is relatively
simple, it provides valuable insights into the mechanism of delayed failure. The results
presented are thus also relevant to other geotechnical problems, such as the stand-up time of
tunnel headings. Emphasis is placed on the effects of plastic dilatancy as it affects ground
response not only quantitatively but also qualitatively. Under the frequently made simplifying
assumption of a constant positive dilation angle, coupled analyses inevitably lead to a
constant deformation rate at failure, while engineers associate failure commonly with
accelerating displacements. Models that allow for shearing under constant volume lead to
accelerating displacements but inherently exhibit numerical stability problems close to the

19
failure state. In order to determine stand-up time it is essential to evaluate the numerical
results in their entirety, including the time-development of the displacement, stress and pore
pressure fields.
Chapter 4 investigates the issue of mesh dependency starting with the problem of the
underwater vertical cut. The results show that when the flow rule is non-associated the stand-
up time decreases with decreasing finite element size. As expected according to the
qualitative considerations of Section 1.1.2, the mesh dependency becomes particularly
relevant when the assumed cohesion approaches the critical value for which the excavation
would be stable under drained cohesion; depending on cohesion, the underwater vertical cut
may remain stable in the long term for a coarse mesh, but experience delayed failure when
adopting a finer discretization. For lower cohesion values the mesh-dependency is less
pronounced and the stand-up time varies almost linearly with the size of the mesh. Finally, it
is shown that the mesh coarseness also affects the inclination of the shear band: the average
inclination of the shear band corresponds to the Roscoe’s inclination for a coarse mesh, while
for a fine mesh the inclination tends to the Coulomb’s one.
The remaining sections of Chapter 4 investigate the influence the mesh size on the
propagation speed of the shear band (and thus of the stand-up time) considering the simple
unloading problem of a rectangular domain under biaxial strain conditions. Based upon
theoretical considerations and analytical computations it is shown that (for given problem
layout and fixed material constants) the stand-up time must vary between an upper and a
lower limit depending on the mesh size. The two limits correspond to two borderline cases
with respect to structural softening: for a coarse mesh the stand-up time tends to an upper
limit because the amount of the shear deformation is so small that no softening occurs, while
for the fine mesh the stand-up time tends to a lower limit because the large shear
deformations cause a great decrease in shear strength.
Finally, Chapter 4 suggests a simple practicable way of estimating a lower upper limit of
stand-up time without the need to implement regularization techniques. As the stand-up time
depends almost linearly on element size for fine meshes, stand-up time can be determined by
performing few computations with relatively fine, but computationally manageable meshes
and extrapolating their results to the mesh size that corresponds to the expected, grain-size-
dependent thickness of the shear band.
The second objective of the Thesis is the planning and performing of experiments and the
validation of the computational method and assumptions. The experiments allow observing
the phenomenon of delayed failure under controlled conditions and consist of uniaxial loading
tests (Chapter 5) and centrifuge tests (Chapter 6). The experimental results were interpreted
also numerically, using material constants which were determined independently (mainly by
means of laboratory tests, partially based upon theoretical considerations).
The numerical analyses of Chapter 5 and 6 show that the simple, but widely used, linearly
elastic, perfectly plastic model with Mohr-Coulomb yield condition (MC model) and
isochoric flow rule generally (i.e. for a sufficiently fine mesh, which is able to reproduce a
thin shear band) underestimates the stand-up time. The reason is that it does not account for
the plastic volumetric strains accompanying shearing, which temporarily perpetuate negative
excess pore pressures, thus delaying failure. The Modified Cam Clay model (MCC), which is
known to overestimate strength on the dry side of the yield surface, generally overestimates

20
 Chapter 1 – Introduction

the stand-up time when used in combination with computationally manageable spatial
discretization (i.e. fine, but still not as much as it would be necessary on account of the
experimentally observed very thin shear bands). However, stand-up times very close to the
observed ones are obtained by extrapolating the numerical results to the size of the shear
band. In conclusion, the MCC model allows for a reasonably accurate prediction of the stand-
up time in a typical over-consolidated clay, while the MC model in combination with a fine
(but still manageable) discretization provides a conservative estimate.
The findings of Chapters 3–6 made it possible to investigate systematically the effect of the
decisive geotechnical parameters on the time-dependency of tunnel face stability (Chapter 7).
Based upon the results of a comprehensive numerical parametric study, dimensionless design
charts were worked-out. The diagrams cover a wide range of ground parameters and apply to
shallow tunnels (overburden less than four diameters). The results presented are important for
tunnel engineering practice, e.g. when estimating the feasibility of atmospheric interventions
in the working chamber of a closed shield. Finally, Chapter 7 presents a simplified approach
for the estimation of the critical advanced speed required in order to maintain the face stable
during ongoing excavation.
Preliminary results of this Thesis have been presented in international conferences
(Appendix A).

1.4 Computational method

The numerical computations of the Thesis are carried-out by means of 2D and 3D fully
coupled hydraulic-mechanical continuum-mechanical FEM analyses. All analyses are
executed employing the commercial finite element code Abaqus 6.12 (Dassault Systèmes
2012). The solution of the non-linear equations is carried out by means of the non-linear
iteration method after Newton.
The integration of the elasto-plastic incremental equations adopts the procedure after Clausen
et al. (2007) (for linearly elastic, perfectly plastic Mohr-Coulomb material) or the build-in
Abaqus procedure (for the Modified Cam Clay material).
Undrained stability is investigated by reducing practically instantaneously the total stresses
n normal to the excavation surface (e.g. tunnel face) from their initial value n0 (in situ
stress) to zero ("practically instantaneously" because a strictly instantaneous unloading, i.e. in
a single time step Δt of zero length, cannot be simulated with the Abaqus software; undrained
conditions are simulated by means of a transient analysis of a fast – compared to the ground
permeability – loading or unloading).
Drained stability is analysed as follows. In a first step, a steady state analysis is carried-out
fixing the pore pressure at the excavation boundary (e.g. tunnel face) and the effective stress
normal to the excavation ("effective support pressure") to its final value (e.g. atmospheric
conditions in case of an unsupported tunnel face) and to the initial normal effective stress 'n0,
respectively. Subsequently, the effective support pressure is reduced stepwise, while the pore
pressure field is fixed to the steady-state field obtained in the first step. In the drained
analyses, the solution of the equilibrium equations (time integration) is carried-out for fixed
hydrostatic pore pressure distribution.

21
Transient stability is investigated by starting a transient analysis after the undrained
excavation. In the transient analyses, the equilibrium and mass balance equations are solved
(time integration) simultaneously. The time stepping in the analyses is automatic with a
minimum step size defined after Vermeer and Verruijt (1981).

22
 Chapter 2 – Structural softening

2. Structural softening

2.1 Introduction
Structural softening occurs when the plastic potential is non-associated and the deformations
localize in a shear band. The structural softening is caused by the rotation of the principal
stresses inside shear band: the larger the rotation, the higher the amount of softening. The
rotation of the principal stresses depends on the shear deformations. In FEM analyses, the
latter depend on the elements size. As qualitatively explained in Section 1.1.2, this causes a
dependency of the stand-up time to the coarseness of the spatial discretization: the stand-up
time decreases with decreasing mesh size because the rotation of the principal axes increases
with decreasing mesh size (due to the occurrence of larger shear deformation).
On the basis of theoretical considerations, it is possible to define a reduction factor η of the
shear strength that takes into account the decrease in resistance (softening) caused by the
rotation of the principal stresses inside the shear band. In a generic way, when structural
softening occurs, the shear resistance reduces to

 n   c '  'n tan  ' , (2.1)

where c' and ' are the effective strength parameters. The reduced shear strength parameters
ηc' and ηtan φ' will be referred hereafter as "operational strength parameters".
As it will be shown in Section 2.2, the reduction factor is bounded between an upper and a
lower value. The maximum reduction corresponds to a full rotation of the principal axis,
while the minimum reduction corresponds to no-rotation of the axes. The two bounds ideally
correspond to infinitely small and very large elements size, respectively.
The origin of the structural softening can be understood as instability in the macroscopic
constitutive equation that leads to a bifurcation point (e.g. plastic loading inside the shear
band and elastic unloading outside it), at which non-uniform deformation occurs. De Borst
and Vermeer (1984) observed a clear softening in simulations of trap door problems and
concluded that its effect increases with the degree of non-associativeness of the plastic
potential. The fundamental role of the structural softening is also shown in Griffiths (1989),
Nordal (2008) and Krabbenhoft et al. (2012) for the bearing capacity of a footing and in
Houslby (1991) for the load capacity of piles. Vermeer (1990) observed the same kind of
softening in the response of bi-axial tests of purely frictional materials. Vermeer and de Borst
(1984) concluded on the basis of a numerical simulation of a simple shear test on dry sand
that for non-associated plasticity, the limit load is not unique and that this depends on the
initial stress state. Drescher and Detournay (1993) analytically proved (using the kinematic
method of limit analysis) that the limit load derived with a statically determinate translational
failure mechanism for a non-associative material is lower than the load for an associative
material. In addition, they solved the stability problem of an unsupported vertical cut
assuming non-associated plastic flow and showed that the formulation for the limit pressure is
equivalent to the one obtained from the static solution of the active earth pressure problem,

23
the only difference being that the “operational” strength parameters for the non-associated
flow rule are lower than the effective shear strength parameters c' and ' (the reduction being
a function of the dilation angle). For the associated plastic flow rule, the kinematic method of
limit analysis and the static solution lead to the same result.
In order to avoid the mesh dependency of the numerical predictions, several regularization
techniques were proposed in the literature (see Table 2.1 for an overview and Thakur 2006 for
a detailed review). In spite of the theoretical robustness of the regularization techniques, their
engineering application is – up to the present day – only limited because, to the knowledge of
the author, almost all regularization techniques (with the exception of those using enhanced
elements) require an extremely fine mesh in the vicinity of the shear band (with element size
in the order of the expected shear band thickness, i.e. few millimetres) and so are hardly
applicable to real geotechnical problems (the fine discretization would imply an extremely
high number of degrees of freedom and, consequently, an extremely long
computational time – even in combination with re-meshing techniques).

Table 2.1. Overview of the regularization techniques.

Regularization technique Description / Comments References

Enrichment of the Non-local approaches (also known as Bazant and Cedolin
continuum adding deformation gradients or imbricate (1979), Bazant et al.
higher-order terms continuum): the stress depends on the average (1984)
(incorporating thus a of the strains within a characteristic distance
characteristic length -
Higher-order continuum model (also known as Mühlhaus (1986), De
i.e., the shear band
micropolar continuum) based upon the Borst (1991b)
thickness)
Cosserat theory: introduction of an additional
rotational degree of freedom
Modification of the The mesh dependency is reduced by artificially Pietruszckzak and Mroz,
constitutive law of the keeping the product between shear strength (1981), Sterpi et al.
element in which decrease (i.e., the softening of the material) (1995), de Borst and
localization is detected and elements size constant. Although this Sluys (1991)
simple method was successfully implemented
in order to regularize intrinsic softening (e.g.
Sterpi, 1995), this is not applicable to problems
in which the softening is due to non-
associativity
Introduction of enhanced The interpolation of the strain inside the Ortiz et al. (1987),
shape functions in order elements is enriched by means of additional Leroy and Ortiz (1989)
to allow the shape functions which reproduce the localized
concentration of the deformation patterns
shear strain in elements
for which localization is
detected
Introduction of a critical - Raniecki and Bruhns
amount of hardening (1981), Runesson and
Mroz (1989), Bigoni and
Hueckel (1991)
Viscoplastic Introduction of strain rate dependency Needleman (1988),
regularization Loret and Prevost (1990)

24
 Chapter 2 – Structural softening

2.2 Reduction factor η

It is assumed that the plastic deformations are localized within a shear band; the stress state
outside the shear band is elastic. The stress field is taken as homogeneous within the shear
band as well as outside the shear band. The shear band forms an angle to the direction of
the minimum principal stress prevailing outside the shear band (Fig. 2.1). The directions of
the principal axes in the shear band (σ1s, σ3s) form an angle ω to the principal axes outside the
shear band.
The normal and shear stresses (σn, τn) at the interface of the shear band with the elastic zone
must fulfil continuity (but the normal stress parallel to the shear band may be discontinuous
across the interface).

(a) w 1
a 1,s

3,s n
a90 –  + w n

3
1

3

(b)
MC-line
τ
(σn , τ n)

Rs R
Rs

θ
θ -w
σ3s σ3 σms σm σ1s σ1 σ
Figure 2.1. (a) shear band orientation and principal stress directions and, (b), graphical
representation of the stress state outside and inside the shear band.

25
From the stress state prevailing outside the shear band the following relationships are
obtained:

1   3
n  sin 2  R sin 2 , (2.2)
2

1   3 1   3
n   cos 2   m  R cos 2 , (2.3)
2 2

where R and m denote the radius and the centre, respectively, of Mohr's circle of the stress
state outside the shear band. Analogously, considering the stress state in the shear band,

1s   3s
n  sin 2   w   Rs sin 2   w  , (2.4)
2

 1s   3s  1s   3s
n   cos 2   w    ms  Rs cos 2   w  , (2.5)
2 2

where Rs and ms are the radius and the centre, respectively, of Mohr’s circle inside the shear
band. The stress state inside the shear band must additionally fulfil the MC yield criterion:

Rs cos  c tan  ( ms  Rs sin  ) . (2.6)

From Eq. (2.6) it follows

Rs c
 ms   , (2.7)
sin  tan 

which inserted into Eq. (2.5) leads to:

Rs c
n    Rs cos 2   w  . (2.8)
sin  tan 

Solving this equation with respect to Rs and introducing the result to Eq. (2.4) leads to the
following relationship between the shear and normal stress at the interface of the shear band
with the elastic zone:

 n    c   n tan   , (2.9)

where

cos  sin 2   w 
   ( , w , )  . (2.10)
1  sin  cos 2   w 

or

26
 Chapter 2 – Structural softening

cos  sin 2a
   (a , )  , (2.11)
1  sin  cos 2a

where α is the angle between the maximum principal stress (inside the shear band) and the
shear band (α = 90°-+ω, Fig. 2.1). To the author's knowledge, Equation (2.11) is derived for
the first time in the present Thesis. However, as it will be shown in the next paragraphs, under
the assumption of purely plastic flow and coaxiality between strain and stress rate, the generic
Equation (2.11) reduces to the well-known expression derived by Davis (1968). Figure 2.2
exemplary shows (for a friction angle of 30° and dilation angle of 0°) the relationship
between the reduction factor and the angle ω described by Equation (2.11).

1.02
η [-]

1.00 η = 1.0

0.98

0.96

ω
0.94 α

0.92

0.90

0.88
Davis (1968)’ s solution
φ' = 30 η = 0.867
0.86 =0
 = 45 + φ' /2 = 60
α = 90 -  + ω
ω[ ]
0.84
0 5 10 15
α[ ]

30 32 34 36 38 40 42 44 46
α = 45 - /2
Figure 2.2. Reduction factor η as a function of the principal axes rotation angle ω and of the
angle α between maximum principal stress and shear band.

27
The magnitude of the reduction factor depends on the amount of principal stress rotation ω
that occurs inside the shear band and, consequently, on the amount of shear deformation. The
relation between shear deformation and rotation of the principal stress is well described in Le
Pourhiet (2012): The rotation of the principal axis occurs because the direction of the plastic
flow dpl within the shear band (oriented at 45°-/2 to the maximum effective stress direction)
differs from the velocity v of the sliding body (oriented parallel to the shear band): taking into
account that the direction of the maximum principal stress at the onset of yielding is the same
inside and outside the shear band, the direction of dpl is obviously different than the velocity
v (the angle between dpl and v is larger than ). Consequently, elastic strains (de) are
generated in order to fulfil the kinematic compatibility along the shear band. The elastic strain
increments are not parallel to the cumulated elastic strain tensor (initially having the same
orientation with the principal stress outside the shear band), and causes therefore a rotation of
the principal directions. The difference between the inclination of elastic strain increments
and the cumulated elastic strain is maximal at the onset of localization and progressively
decreases with shear deformations until the plastic flow becomes parallel to the shear band
(i.e. until shearing is purely plastic). At this time the maximum principal stress forms an angle
of 45°-/2 with the shear band.
At the beginning of the shearing, the principal stresses inside the shear band have the same
directions as in the elastic zone outside the shear band (ω = 0°, α = 45°-φ/2 for the Coulomb
inclination θ = 45° + φ/2) and, according to Eq. (2.11), the reduction factor amounts to η = 1,
which means that the operational strength parameters correspond to the Mohr-Coulomb ones.
Assuming, on the other hand, that all points along the shear band have experienced a large
amount of shearing so that the plastic flow is fully developed (i.e.,  n   tan and  t  0 ),
the principal strain axes form an angle 45°±/2 with the shear band. Assuming coaxiality of
the principal stresses and strains, the angle α becomes 45°-/2 and the reduction factor

cos  cos
 . (2.12)
1  sin  sin
This equation was firstly obtained by Davis (1968). As pointed-out by Housbly (1991) and
Drescher and Detournay (1993), the reduction of the strength parameters expressed by
Equation (2.12), represents the limit of structural softening, after which the material behaves
as a perfectly plastic material.

28
 Chapter 3 – Manifestation of failure

3. Manifestations of delayed failure

3.1 Introduction
The present Chapter investigates how failure manifests itself in coupled FEM analyses by
studying in detail the relatively simple plane strain problem of an underwater vertical trench
excavation (Fig. 3.1), which, on account of the chosen material constants, remains stable in
the short-term, but collapses before reaching drained conditions. The reason for considering
an underwater excavation is to keep the problem as simple as possible. More specifically, as
the stationary hydraulic head field is hydrostatic in the case of an underwater excavation, it
avoids the additional complexity introduced by long-term seepage forces and by the resulting
tensile stresses (the influence of the tensile stresses on the manifestation of failure will be
discussed in Chapter 7). Although the problem considered is simple and is analysed assuming
conventional constitutive models, it provides an illustration of features that are indeed
relevant to many practical geotechnical problems, such the face stability of tunnels or the
behaviour of retained excavations in low-permeability soils.
Chapter 3 starts with the setup of the problem and an outline of the computational model, and
continues with an investigation of the borderline cases of undrained stability and drained
stability. The results of this preliminary analysis serve to identify a set of shear strength
parameters for which the excavation is unstable in the long term (hereafter referred to as
"critical parameters"). Afterwards, delayed failure is investigated for this parameter set by
means of fully coupled numerical analyses considering a linearly elastic and perfectly plastic
material obeying Mohr-Coulomb yield criterion (hereafter referred to as MC model); it will
be shown that under the simplifying assumption of a constant non-zero dilation angle the
system reaches a steady state that is characterized by constant non-zero rates of displacement
and shear strain. The reason for this model behaviour is that an acceleration in the
deformations would imply an accelerating increase in the water content in the shear band,
which, as explained later, is physically impossible.

10 m

x = pw0 h = h0
H=5m y = y0

h = h0 h0 = 20 m
h = h0

y
x qy = 0

30 m

Figure 3.1. Computational model and boundary conditions.

29
Accelerating displacements presuppose that shearing can occur at a constant volume, which is
true for most soils (after a certain amount of shearing deformation has taken place) and can be
mapped by the appropriate constitutive models. This is illustrated by numerical results for the
simplest possible model (the MC model with an isochoric flow rule) as well as for the
Modified Cam Clay model (Roscoe and Burland, 1968; hereafter referred to as MCC model).
Such models exhibit numerical stability problems close to the ultimate state, however,
because it is inherently impossible to satisfy the equilibrium and the mass balance equations
simultaneously.

3.2 Problem setup and computational model

Figure 3.1 shows the geometry and the boundary conditions of the plane strain problem under
investigation. The initial effective stresses increase linearly with depth depending on the
submerged unit weight γ' and a lateral earth pressure coefficient equal to unity. Excavation is
simulated by reducing the horizontal effective stress at the vertical excavation boundary from
its initial value to zero. The vertical stress at the bottom boundary of the excavation is fixed to
its initial value in order to avoid swelling of the bottom boundary, and to focus on the stability
of the vertical cut (as shown it will be shown in Section 3.3.3.5, this assumption is not
essential for the problem under consideration).
The soil is modelled either as an isotropic, linearly elastic and perfectly plastic material
obeying the Mohr-Coulomb yield criterion with a non-associated flow rule or as a modified
Cam Clay material. Seepage flow follows Darcy’s law; the effect of porosity-dependent
permeability (after Kozeny-Carman) was also investigated. Tables 3.1 and 3.2 give the
material constants and computational details of the two constitutive models considered (MC,
MCC), respectively.

3.3 Mohr-Coulomb material

3.3.1 Stability under undrained or drained conditions

The stability (or lack thereof) is evaluated from the displacement of the crest of the
excavation (point A in the inset of Fig. 3.2b). Figure 3.2b shows the computed displacements
of point A as a function of the effective support pressure 'x (normalized by the initial
effective stress 'x0) for a specific parameter set.
Under undrained conditions (dashed lines) the displacements reach a finite value after
complete unloading, which indicates that the excavation is stable. The dilation angle has no
influence on the displacements, which is trivial considering that the soil remains practically
elastic even after complete unloading (a plastic zone develops only locally at the toe of the
vertical cut and at the soil surface, Fig. 3.2a).
Under drained conditions the displacements become asymptotically infinite at about
'x/'x0 = 13% (solid lines in Figure 3.2b), while a continuous plastic zone develops from the
toe of the excavation to the soil surface (Figure 3.2a, solid lines), which indicates that the
system has reached the ultimate state. Therefore, for the shear strength parameters in Figure

30
 Chapter 3 – Manifestation of failure

3.2, the unsupported excavation, that is stable in the short term would be unstable in the long
term. The failure geometry agrees well with the known sliding wedge mechanism assumed in
limit equilibrium analyses (e.g. Culmann, 1875).
According to Figure 3.2b, the plastic dilation angle has a negligible effect on the magnitude
of the displacements. The reason is that plastic dilation occurs only inside a relatively narrow
shear zone.
Figure 3.3a shows the results of a parametric study into stability conditions in terms of the
effective shear strength parameters c' and '.
The straight lines were determined analytically according to the active earth pressure theory.
The drained stability was computed in terms of effective stresses considering an unsupported
vertical cut. The dash-dot line is based on the analytical solution of Drescher and Detournay
(1993), according to which the limit load for ψ = 0° is higher than for ψ = φ'. More
specifically, they showed that the formulation for the limit pressure is similar to the one
obtained from the active earth pressure theory, the only difference being that the
“operational” strength parameters (see Chapter 2) for the non-associated flow rule are lower
than the effective shear strength parameters c' and φ' (the reduction being a function of the
dilation angle). This is the reason why the cohesion required for drained stability of an
unsupported excavation is higher in the case of zero dilation.

Table 3.1. Assumed material constants.

Seepage flow parameters

Unit weight of water w [kN/m3] 10
Permeability k [m/s] 10-9
Water compressibility cw [MPa] 0

Parameters for Mohr-Coulomb material

Saturated unit weight  [kN/m3] 20
Young’s modulus E' [MPa] 15
Poisson’s ratio  [-] 0.3
Plasticity parameters c', ', ψ see Figures

Parameters for Modified Cam Clay material

Saturated unit weight  [kN/m3] 20
Gradient of swelling line [-] 0.010
Gradient of compression line [-] 0.085
Specific volume at a unit reference pressure on the [-] 2.0
compression line
Corresponding specific volume on the CSL  1.948
Poisson’s ratio  [-] 0.3
Slope of the critical state line M [-] 0.866

31
Table 3.2. Computational details.
Size of computational domain After Wehnert (2006) – see Fig. 3.1
Spatial discretization 8-node elements with biquadratic displacement and
bilinear pore pressure shape functions
Number of nodes 19,101
Number of degrees of freedom 44,627

(a)
Undrained analysis
Complete unloading
 eq
pl
0.01%

Drained analysis
Ultimate state
 eq
pl
10%

 = '

15
Ultimate state uA,y
(b)
uA,x [mm]

 = ' uA,x

=0 A

10
=0

 = '

0
0% 25% 50% 75% 100%
x'/'x0
uA,y [mm]

 = '
=0 Undrained analysis
=0  = ' Drained analysis
-5

Figure 3.2. (a) Contour lines of the equivalent plastic strain (defined as )

and, (b), displacements at the excavation crest (point A) as a function of the

normalized effective support pressure (c' = 4.5 kPa, ' = 30°, other parameters
according to Table 3.1).

32
 Chapter 3 – Manifestation of failure

The solid line applies to undrained stability and was computed in terms of total stresses,
taking account of the support force exerted by the water upon the vertical cut. It is a simple
matter to verify (Lambe and Whitman, 1979) that for equilibrium the minimum undrained
shear strength
1
su   ' H
4 , (3.1)

where H denotes the height of the cut. Under plane strain conditions and certain simplifying
assumptions (MC model with ψ = 0 , no out-of-plane plastic flow, K0 = 1), the actual
undrained shear strength is related to the effective shear strength parameters and to the in situ
effective stresses at the mid-height of the vertical cut:

H
su  c 'cos  '  ' sin  '
2 . (3.2)
Eqs. (3.1) and (3.2) lead to following undrained stability condition in terms of the effective
shear strength parameters (solid line in Fig. 3.3a):
1  2sin  '
c'   'H
4 cos  ' . (3.3)

The marked points in Figure 3.3a were obtained numerically and agree well with the
analytical solutions (lines). The points show combinations of the minimum shear strength
parameters under which the unsupported excavation remains stable under drained conditions
(white and black circles) or under undrained conditions for isochoric plastic behaviour (black
diamonds). In the case of dilatant plastic behaviour, the excavation always remains stable
under undrained conditions (due to the unloading-induced suction); for this case (dilatant
behaviour, undrained conditions), Figure 3.3a (white diamonds) shows the minimum shear
strength parameters that are necessary to prevent the development of a continuous plastic
zone from the bottom to the crest of the excavation.
The numerically computed points agree reasonably well with the analytical solutions (lines);
the numerical analyses also predict higher critical shear strength parameters in the case of
non-associated flow rule.

33
 (a) (b)

12 12

c' [kPa]
c' [kPa] 1 – always stable
' = 30
10 10 =0

8 8

3 – delayed failure
6 6

(' = 30 ;
c' = 4.5 kPa) 4
4

2 2
2 – immediately unstable

0 0
10 15 20 25 30 35 0 250 500 750 1000
' [°] ts [h]

Undrained (analytical solution) Drained (Drescher & Detournay 1993,  = 0 )

Undrained FEM ( = 0 ) Drained (analytical solution,  = ')
Undrained FEM ( = ') Drained FEM ( = 0 )
Drained FEM ( = ')

Figure 3.3. (a) Critical shear strength parameters (c', φ') for undrained and drained
stability and, (b), stand-up time ts as a function of the cohesion c' (parameters
according to Table 3.1).

As expected (Section 2), the assumption of an associated flow rule (dilation angle ψ = φ',
white markers) leads to lower critical shear strength parameters than that of isochoric plastic
flow (ψ = 0°, black markers), both under drained conditions and under undrained conditions
(due to structural softening).
The solid and dashed lines in Fig. 3.3a delimit three regions: for shear strength pairs (φ', c')
within region 1 the excavation is stable in both the short term and the long term (i.e. the
stand-up time is equal to infinity); in region 2 the excavation would fail instantaneously (i.e.
the stand-up time is equal to zero); region 3 exhibits delayed failure (excavation is stable in
the short term, but unstable in the long term). As region 3 is quite narrow, Figure 3.3a implies
that relatively small variations in the cohesion will result in variations in the stand-up time by
orders of magnitude. (The curve in Figure 3.3b was determined numerically after Section
3.3.3 and illustrates the sensitivity of stand-up time with respect to cohesion for a friction
angle of 30°).
The transient analyses of the next sections will be carried out for a parameter pair in the
delayed failure region (c' = 4.5 kPa, ' = 30°, marked by a triangle in Fig. 3.3a), for which the
question of stand-up time arises. The chosen parameter set is representative of soils exhibiting
a pronounced time dependency (the shear strength parameters of London Clay after
Kovacevic et al., 2007).

34
 Chapter 3 – Manifestation of failure

3.3.2 Coupled analysis for dilatant plastic behaviour

3.3.2.1 Displacement rate

Let first take a material with the associated flow rule ( = φ'). As can be seen from Figure
3.4a, a plastic zone develops progressively over the course of time from the bottom of the
vertical cut towards the soil surface, which is reached at about 250 h. As in many other
similar problems some early yielding also occurs close to the soil surface. It is associated with
the development of tensile stresses at the surface of the model, where the initial stresses are
very low (Kovacevic et al., 2008.) However, although the plastic zone extends up to the
surface, thus separating the unstable and stable soil regions, the displacements of the
excavation crest (Fig. 3.4b, lines for ψ = 30°) do not exhibit the acceleration that one would
expect close to the ultimate state, but instead continue to increase at a constant rate.
This happens not only under the assumption of an associated flow rule, but also for any ideal
material that does not cease to experience plastic volumetric strains during shearing, i.e. under
any simplifying constitutive law involving a positive constant non-vanishing dilation angle
(see lines for 0 < ψ ≤ φ' in Fig. 3.4b). The only difference is that the displacement rate
increases with decreasing dilation angle: As shearing is accompanied by plastic volumetric
strains, any specific value of the crest displacement is associated with a certain increase in the
water content in the plastic zone. The lower the dilation angle, the less the water content must
increase in order that the displacement reaches a specific value and, consequently, the earlier
the displacement will reach this value or, in other words, the higher the displacement rate will
be.

35
 t [h] = 0+ (a)

150

300

 = '
 eq
pl
 0.01%

8
(b)
uA,x [mm]

4
=5
10 20
30 (=')
2
constant rate

0 +
0 100 200 300 400 t [h]
30
-2
20

-4 uA,y 10
uA,x
5
uA,y [mm]

-6 A

-8

0.20
xy [%]

 =' (c)
constant rate
0.15
G E
F
0.10 E
F

0.05
G

0.00
+
0 100 200 300 400 t [h]

Figure 3.4. Time-development of , (a), the 0.01% contour line of the equivalent plastic
strain, (b), of the post-excavation crest displacements and, (c), of the shear
strain xy at three points in the plastic zone
(' = 30°, c' = 4.5 kPa, other parameters according to Table 3.1).

36
 Chapter 3 – Manifestation of failure

35 35
(a) (e)

pw [kPa]

pw [kPa]
B
30 30
Initial state
t [h] =
25 25 C
250 300
100
20 20
50

15 15
0+
10 10
D

5 5
2m 1 2 3 4 x [m]
0 0
30 30
Initial state (b) (f)

'x [kPa]
'x [kPa]

25 25
D 1m
t [h] =
20 20
B C
0+
300 2m
15 15 1 2 3 4 x [m]
250
50 100
10 10
C
5 5
D
B
0 0
50 50
'y [kPa]

(c) (g)
'y [kPa]

t [h] =
40 0+ 40
250
50 300 C
100 Initial state
30 30

20 20
B
D
10 10

0 0
0.30 0.30
(d) (h)
vol [%]

vol [%]

B
0.25 0.25

C
0.20 t [h] = 0.20

300
0.15 0.15
250

0.10 0.10

100
0.05 50 0.05
D
0+
0.00 0.00
0 1 2 3 4 0+ 100 200 300 400
x [m] t [h]

Figure 3.5. Spatial distribution along a horizontal line (see inset in diagram (a)): (a) pore
pressure pw; (b) effective horizontal stress 'x; (c) effective vertical stress 'y;
(d) volumetric strain εvol. Time-development at three points (see inset in
diagram (e)): (e) pore pressure pw; (f) effective horizontal stress 'x; (g) effective
vertical stress 'y; (h) volumetric strain εvol ( = ' = 30°, c’ = 4.5 kPa, other
parameters according to Table 3.1).

37
3.3.2.2 Pore pressure, stress and strain development

In order to gain more insight into the model behaviour, let study the time-development of
pore pressure pw, the effective stresses ('x, 'y) and the volumetric strain εvol. Figure 3.5
shows the distribution and the time-development of these variables along a horizontal line
(see inset in Fig. 3.5a; note that the stresses are taken from the Gauss integration points
located just above the line) and at selected points (see inset in Fig. 3.5e; points B and C are
located in the plastic zone, point D is located in the elastic region). Rapid excavation leads to
an instantaneous decrease in the pore pressure (Fig. 3.5a). The negative excess pore pressures
partially counterbalance the excavation-induced decrease in the horizontal effective stress
(compare lines for t = 0+ in Fig. 3.5a and 3.5b) and therefore have a stabilizing effect, which
manifests itself in the relatively high vertical effective stresses (line for t = 0+ in Fig.3.5c).
The stabilizing effect of the negative excess pore pressures is only temporary because they
dissipate over time (Fig. 3.5a) according to the imposed boundary condition (hydrostatic
pressure), thus leading to a progressive decrease in the horizontal and vertical effective
stresses (Fig. 3.5b and 3.5c). After about 250 hours, pore pressures and effective stresses
reach a stationary state (Fig. 3.5a to 3.5c, Fig. 3.5e to 3.5g). On the other hand, the volumetric
strain in the plastic zone continues to increase at a constant rate (see Fig. 3.5d and the lines
for points B and C in Fig. 3.5h). This is particularly evident in the middle of the plastic zone
(point C in Fig. 3.5h). As the effective stresses remain constant for t > 250 h, the volumetric
strain rate is purely plastic and is associated (due to the assumption of a constant dilation
angle) with a constant plastic shear strain rate, which manifests itself macroscopically as a
constant velocity of the sliding wedge (Fig. 3.4b).

3.3.2.3 Comments on the steady state

It is remarkable that the dissipation of the excess pore pressures ceases before the pore
pressures reach the hydrostatic distribution that would prevail under drained conditions. The
steady state is characterized by a non-hydrostatic pore pressure distribution or, in other
words, by a non-uniform hydraulic head field. The latter fulfils the mass balance equation:

k h  ncw pw   vol

el
  vol
pl
, (3.4)
where h, k, n and cw denote the hydraulic head, permeability, porosity and water
compressibility, respectively. At the steady state (i.e., practically, for t > 250 h), the first and
the second r.h.s. term disappear (the elastic strain rate is equal to zero because the effective
stresses no longer change), while the plastic volumetric strain rate is constant and non-zero.
Consequently, Eq. (3.4) becomes

k h   vol
pl
. (3.5)
According to this equation, the hydraulic head field remains stationary but, rather than being
uniform, is consistent with the r.h.s. term, which is imposed by the shearing of the soil at a
constant volumetric strain rate. In other words, the shear zone represents a sink. The drop in
water pressure in the plastic zone of dilatant materials is known in the literature (e.g. Rice,
1975) and has also been observed in laboratory and field tests (e.g. Han and Vardoulakis
1991, Cooper et al. 1998). The gradient of the non-uniform stationary head field causes a

38
 Chapter 3 – Manifestation of failure

permanent seepage flow towards the dilating shear zone, the rate of which controls the rate of
displacement. The displacement rate cannot increase continuously, because this would
presuppose an accelerating shearing rate and, due to the plastic flow rule, an accelerating
water uptake by the plastic zone. The latter would presuppose an accelerating seepage flow
rate and, consequently, a continuously increasing hydraulic gradient towards the plastic zone.
As the hydraulic boundary conditions are fixed, a continuously increasing hydraulic gradient
would be possible only if the hydraulic head in the plastic zone decreased continuously, i.e. if
the pore pressure in the shear band continued to become more and more negative. This is
impossible as it would lead to a continuous increase in shear resistance, thus halting the
failure process and contradicting the initial hypothesis of accelerating displacements.
In conclusion, the frequently made, simplifying but nevertheless unrealistic constitutive
assumption of a constant non-zero dilation angle not only delays failure in the model, but
actually prevents it from reaching a state with rapidly accelerating displacements, which is
commonly associated with collapse.
It should be noted here that the effect of dilatancy on the displacement rate at failure was first
observed by Potts et al. (1997) in their numerical investigation into the delayed collapse of
cut slopes in stiff clay. They intuitively explained the gradual (rather than sudden) increase in
displacement over time as the combined effect of two compensating mechanisms: the
decrease in pore water pressure in the rupture zone due to dilation and its increase due to
swelling (such that the hydraulic head field becomes stationary, which is the case). Thus, they
concluded that ‘the drop in pore water pressures due to dilation acts as a partial brake on the
rate of post-collapse movement on the slip surface’.

3.3.2.4 On the possible effect of an increase in permeability

Another simplifying constitutive assumption is that of a constant permeability coefficient. As

the permeability of soils depends on the void ratio e which increases considerably in the
plastic zone, the strain-dependency of permeability might be relevant for the problem under
investigation. In order to investigate the extent, to which an increase in permeability with
volumetric strain might cause accelerating displacements, comparative analyses were carried-
out assuming a permeability variation according to the Kozeny-Carman’s equation in the
form given by Chapuis and Aubertin (2003):

e3
log k  A  log
D S 1  e  ,
2
R
2 (3.6)

where k is the permeability (in m/s); e is the current void ratio; DR denotes the relative density
of the solids (DR ≅ 2.7 for most minerals); S is the specific surface of the solids (in m2/kg);
and the parameter A depends on the shape and tortuosity of the flow channels inside the soil.
Typical values for medium-low permeability soils are S = 1 – 30 x 103 m2/kg and A = 0.5
(Chapuis and Aubertin, 2003). In order to compare the results obtained for a variable
permeability with those obtained under the assumption of a constant permeability, an in situ
void ratio was chosen such that the initial permeability corresponds to k = 10-9 m/s. The dry
unit weight of the soil was adjusted accordingly. As the time-marching scheme of Abaqus is
explicit with respect to the permeability-induced non-linearity (i.e. k is updated only at the

39
end of every time interval), the time-integration of the coupled equations was performed in
small time-steps.
Figure 3.6a shows permeability as a function of the void ratio according to Eq. (3.6) for two
soils, while Figure 3.6b presents the numerically computed time-development of the crest
displacement taking account of the permeability curves in Figure 3.6a. Under the assumption
of Kozeny-Carman’s equation the displacement rate is slightly higher but reaches again a
constant non-zero value. The increase in permeability with the void ratio is not decisive.

(a)
e [-]
0 2 4 6 8 10
1.E+00

1.E-02

k after Kozeny-Carman
1.E-04 (silty soil parameters)

1.E-06 k after Kozeny-Carman

(clayey soil parameters)

1.E-08
k [m/s]

k = 10 -9 m/s

1.E-10

(b)
4
uA,x [mm]

k after Kozeny-Carman
(silty soil parameters)

3
k after Kozeny-Carman
(clayey soil parameters)

2
k = 10 -9 m/s

1 u A,x
A

0
+
0 100 200 300 400
t [h]
Figure 3.6. (a) Dependency of the permeability k on the void ratio e according to the
Kozeny-Carman equation for a silty soil (S = 1 m2/g) and for a clayey soil (S = 11
m2/g) and, (b), time-development of the crest displacement for the constant or
variable permeability coefficient k ( = ' = 30°, c' = 4.5 kPa, DR = 2.7, A = 0.5,
other parameters according to Table 3.1).

40
 Chapter 3 – Manifestation of failure

(a)

t [h] = 96

93

90
 =0
 eq
pl
1%

60
(b)
uA,x [mm]

40 Rapid increase of
the displacements

20

0
+
0 20 40 60 80 100 120 140
t [h]
-20
uA,y
uA,x
-40
uA,y [mm]

-60

Figure 3.7. Time-development of (a), the 1% contour line of the equivalent plastic strain
and, (b), the post-excavation crest displacements ( = 0°, ' = 30°, c' = 4.5 kPa,
other parameters according to Table 3.1).

3.3.2.5 Estimating stand-up time for ψ > 0

Considering the nature of the above limitation in the constitutive assumption (unlimited
plastic volumetric strains), stand-up time might be determined by checking when the
computed shear strains reach the values at which the soil approaches critical state and,
consequently, the assumption of a constant dilation angle becomes absolutely unrealistic.
Such an approach does not appear promising because – at least in the computational example
under consideration – the displacement rate is already constant at shear strains of less than a
few per cent (Fig. 3.4c), which are considerably smaller than the values at which the soil
would approach a critical state (10 – 40%, cf., e.g. Atkinson and Bransby, 1978).
On the other hand, considering the prominent role of plastic volumetric behaviour, it may be
an oversimplification to define stand-up time as the time lapsing between unloading and the
time at which the displacement rate becomes constant. A realistic flow rule, e.g. within the
framework of critical state soil mechanics, is indispensable in this respect. Before studying
such a model, it is instructive first to discuss the behaviour of the simplest possible model, i.e.
that of a perfectly-plastic material with isochoric behaviour ( = 0) right from the start of
shearing.

41
 (a)

(b)

Figure 3.8. Deformed mesh for associated (a) and isochoric (b) plastic flow at t = 250 and t
= 100 h, respectively (scaled by factor 20).

3.3.3 Coupled analysis for isochoric plastic behaviour

3.3.3.1 Displacement rate

In contrast to the case of dilatant behaviour (Fig. 3.4b), crest displacements increase rapidly
to infinity (after about 90 hours, Fig. 3.7b). The constitutive assumption concerning plastic
dilatancy thus affects the model behaviour not only quantitatively but also qualitatively.
The asymptotic increase in the displacement rate indicates that the system is approaching the
ultimate state, and it occurs as soon as a continuous plastic zone forms, delimiting the stable
part of the ground from the unstable part (i.e., t = 85 h, Fig. 3.7a). Once the plastic zone
reaches the soil surface, no further stress re-distribution is possible and the excavation
collapses.
Failure occurs considerably earlier than the onset of a constant displacement rate in the case
of an associated plastic flow (90 h vs. 250 h, Fig. 3.7b and 3.4b). These results agree
qualitatively with those of Kirkebø (1994) and Potts et al. (1997), who also found that
dilatancy delays the process of pore pressure dissipation. Another remarkable difference
between dilatant and isochoric behaviour is that in the latter case shearing occurs within a thin
band (see deformed meshes in Fig. 3.8) as the plastic deformations are localised in a narrow
zone due to the non-associativity of the plastic potential. The localization is not particularly
evident in Figure 3.7a because the plotted contour delimits a zone of non-zero, but very small
plastic strains. Figure 3.9 shows that at the ultimate state, the plastic deformations
(represented by the equivalent plastic strain

42
 Chapter 3 – Manifestation of failure

t
2 p p
 eq , pl    ij  ij dt
0
3 , (3.7)

as defined by Berg 1972) outside the shear band amount to less than 0.1%. The concentration
of the plastic deformations occurs between 80 h and 90 h (Fig. 3.9). Before this, plastic strains
larger than 0.1% occur only at the toe of the excavation. This behaviour clearly differs from
that of a dilatant material, for which plastic deformations remain smaller than 0.01% (with the
exception of the toe) even 250 h after excavation (Fig. 3.10).

 =0

t = 0+ t = 80 h

t = 30 h t = 90 h

t = 60 h t = 95 h

pl, eq = 0.01
%
pl, eq = 0.1 %

pl, eq = 1.0 %

Figure 3.9. Contour of the equivalent plastic strain eq,pl at several time points
( = 0°, c' = 4.5 kPa, other parameters according to Table 3.1).

43
  = ’

t = 0+ t = 150 h

t = 50 h t = 250 h

pl, eq = 0.01
%
pl, eq = 0.1 %

pl, eq = 1.0 %

Figure 3.10. Contour of the equivalent plastic strain eq,pl at several time points( = ' = 30°,
c' = 4.5 kPa, other parameters according to Table 3.1).

3.3.3.2 Development of pore pressures, strains and stresses

Figure 3.11 shows, in a similar way to Figure 3.5, the spatial distribution and time-
development of pore pressure, effective stresses and volumetric strain. The instantaneous
response is practically the same as for the case of an associated flow rule. The evolution of
the pore pressure (Fig. 3.11a) is similar to the case of an associated flow rule (Fig. 3.5a), but
the effective stresses decrease faster for  = 0° (Fig. 3.11b and 3.11c) than for  = ' (Fig.
3.5b and 3.5c) and, most remarkably, their spatial distribution exhibits oscillations after about
95 hours (Fig. 3.11b and 3.11c), while the pore pressure distribution remains almost constant
(Fig. 3.11a). Kirkebø (1994) experienced similar numerical stability problems close to failure.
The oscillations are particularly evident in the time-development of the stresses and the
volumetric strain in the shear zone after t = 95 h (Fig. 3.11f to 3.11h, point C).

44
 Chapter 3 – Manifestation of failure

35 35

pw [kPa]
pw [kPa]
(a) (e)
30 B
30
t [h] = Initial state

25 110 25
100
90
20 60 20

30 C
15 15
0+
10 10

5 5 D
2m 1 2 3 4 x [m]
0 0
30 30
Initial state
'x [kPa]

(b)

'x [kPa]
(f)
25 25
D 1m
t [h] =
20 20 B C
0+ 2m 1 2 3 4 x [m]
15 15
30 C
10 10
60
90
5 100 5 D
110
B
0 0
60 60
(c) (g)
'y [kPa]
'y [kPa]

50 50
t [h] = C
40 0+ 40
30
60 Initial state
30 30
90
100
20 110 20 B

D
10 10

0 0
0.20 0.20
vol [%]

(d) (h)
vol [%]

B
0.15 0.15
t [h] =
110
100
0.10 90 0.10

60
0.05 30 0.05
C
0+
D
0.00 0.00
+
0 1 2 3 4 0 15 30 45 60 75 90 105
x [m] t [h]
Figure 3.11. Spatial distribution along a horizontal line (see inset in diagram (a)): (a) pore
pressure pw; (b) effective horizontal stress 'x; (c) effective vertical stress 'y; (d)
volumetric strain εvol. Time-development at three points (see inset in diagram
(e)): (e) pore pressure pw; (f) effective horizontal stress 'x; (g) effective vertical
stress 'y; (h) volumetric strain εvol ( = 0°, ' = 30°, c' = 4.5 kPa, other
parameters according to Table 3.1).

45
3.3.3.3 Reason of numerical instability

The occurrence of numerical oscillations is related to the inherent impossibility of satisfying

both mass balance and equilibrium at the ultimate state for materials exhibiting shearing at
constant volume. At the ultimate state, the plastic flow should occur under constant effective
stresses and, the elastic volumetric strain rate should therefore become equal to zero. As the
plastic volumetric strains are equal to zero as well (ψ = 0°), the source term in the mass
balance equation should disappear, i.e.
k h  0
. (3.8)
Eq. (3.8) in combination with the given homogeneous hydraulic boundary conditions implies
a uniform hydraulic head field, under which the equilibrium condition cannot be satisfied
(Section 3.3.1). This is also the reason why the computed pore pressures (Fig. 3.11a and
3.11e) cannot reach the final hydrostatic distribution according to Eq. (3.8); the source term in
the mass balance equation (Eq. 3.4) remains non-zero.

3.3.3.4 Estimating stand-up time for ψ = 0

In conclusion, coupled transient numerical analyses that assume non-dilatant plastic flow can
approach, but not reach, the ultimate state. This makes a determination of failure time
difficult. In order to estimate stand-up time, one should simultaneously evaluate the time-
development and the distribution of displacements, effective stresses and volumetric strains.
The ultimate state is characterized by, (i), an increase in the displacement rate of points
located within the unstable region of the system, (ii), the development of a continuous plastic
zone and, (iii), the onset of spurious oscillations in volumetric strain and effective stresses.

3.3.3.5 Influence of the loading on the bottom boundary

Comparative analyses with and without unloading of the bottom boundary of the excavation
show that the assumption about the condition at the bottom boundary of the excavation is not
essential for the question under investigation, i.e. the manifestation of delayed failure (Fig.
3.12). Specifically,
 Under the assumption of isochoric plastic flow (Fig. 3.12a to c), the bottom boundary
condition has practically no influence on the results; the displacement evolution over
time (Fig. 3.12a) and the extent of the plastic zone close to the ultimate state (Fig.
3.12b, c) are similar.
 Under the assumption of dilatant plastic flow (Fig. 3.12d to e), the results obtained
with and without bottom boundary pressure are quantitatively different but
qualitatively the same. In both cases, the displacement rate becomes constant
approximately 200–300 h after unloading (Fig. 3.12d). This happens slightly later
without bottom pressure. At t = 250 h, the plastic zone has completely developed
through the model if there is a bottom pressure (Fig. 3.12e), but is still propagating
from the bottom to the ground surface in the case without pressure (Fig. 3.12f).
In conclusion, independently of whether the bottom boundary of the excavation is unloaded
or not, the displacements accelerate at failure if the plastic flow is isochoric (ψ = 0°), but
increase with a constant rate if ψ > 0°.

46
 Chapter 3 – Manifestation of failure

=0 (d)  = 30
(a)

20 4

uA [mm]
uA [mm] uA,x uA,x
With bottom eff. pressure
15 3
Without bottom eff. pressure

10 2

5 1

0 0
+ +
0 20 40 60 80 100 120 0 100 200 300 400
t [h]
-5 -1 t [h]

-10 u A,y -2
u A,x

-15 A -3

uA,y uA,y
-20 -4

(b) (e)

’y,0 ’y,0

t = 85 h t = 250 h
With bottom eff. pressure With bottom eff. pressure

(c) (f)

t = 85 h t = 250 h
Without bottom eff. pressure Without bottom eff. pressure

Figure 3.12. (a) Evolution over time of the vertical crest displacements taking place after
undrained excavation for ψ = 0°; (b) plastic zone close to the ultimate state for ψ
= 0° without unloading of bottom boundary (solid line in Fig. 3.12a); (c) plastic
zone close to the ultimate state for ψ = 0° with unloading of bottom boundary
(dashed line in Fig. 3.12a); (d) Evolution over time of the vertical crest
displacements taking place after undrained excavation for ψ = 30°; (e) plastic
zone close to the ultimate state for ψ = 30° without unloading of bottom
boundary (solid line in Fig. 3.12d); (f) plastic zone close to the ultimate state for
ψ = 30° with unloading of bottom boundary (dashed line in Fig. 3.12d).

47
3.4 Coupled analysis for Modified Cam Clay material
An accelerating increase in the displacement rate at failure presupposes that the material is
able to shear at a constant volume, which is true for most soils after the occurrence of large
shear strains. The Modified Cam Clay model (Roscoe and Burland, 1968) is able to reproduce
the typical behaviour of normally or slightly over-consolidated clay, including deformations
at constant volume after a certain amount of shearing has taken place (i.e. at the critical state).

3.4.1 Assumptions

The chosen material constants (Table 3.2) are representative for a stiff, slightly over-
consolidated clay. The yield, critical state and plastic potential surfaces assume a circular
shape in the deviatoric plane, with the critical stress ratio M taken equal to 0.866, which
corresponds to a friction angle of 30° at zero Lode angle. As explained later, the stress state at
failure and thus in the shear zone reaches a Lode angle of zero and this is why the critical
stress ratio M was adjusted to the hexagonal MC surface (with φ' = 30°) there.
In order to determine an initial state (void ratio, in situ effective stress and OCR distribution)
that is consistent with the constitutive model and the loading history of the soil, the numerical
analysis starts with four preliminary steps (see inset of Fig. 3.13b):
i) model initialization considering a uniform and isotropic initial stress field (1 kPa),
hydrostatic pore pressure distribution and homogenous soil with void ratio equal to 1
(start of sedimentation);
ii) gravity loading under drained conditions;

iii) pre-loading by a surface pressure 'OC (50 kPa in the example of Fig. 3.13); and,
iv) complete unloading of the surface.

(a) (b) (c)

20
y [m]

iv
iii i
i, ii, iii
18 ii, iii

i. e0 = 1, K0 = 1
iv
ii iv
i
16
'
ii.
'OC = 50 kPa
14

iii.
'
12

'
iv.
10
0.5 0.6 0.7 0.8 0.9 1 1.1 0 2 4 6 8 0 0.5 1 1.5 2 2.5 3
e OCR K0

Figure 3.13. MCC model: vertical distribution of void ratio e, OCR and K0 at pre-excavation
stages i–iv (parameters according to Table 3.1).

48
 Chapter 3 – Manifestation of failure

20

y [m]
19
'OC [kPa] =
40
50
18 60 70
80 90
100

17

5m
16

y
x
15
0 2 4 6 8 10

OCR
Figure 3.14. MCC model: OCR profiles for various pre-consolidation pressures (parameters
according to Table 3.1).

50
p', pw [kPa]

'OC = 50 kPa
B
2.5 m
40

1.8 m

30

p'

20
pw

10

0
0+ 300 600 900 1200 1500
t [h]

Figure 3.15. MCC behaviour: Time-development of the pore pressure pw and of the mean
effective stress p' at point B (parameters according to Table 3.1).

Figures 3.13a, 3.13b and 3.13c show the vertical distribution of the void ratio, the OCR and
the coefficient of lateral pressure in these four stages. Gravity loading (step ii) causes a
compaction of the ground; the void ratio decreases from 1.0 (at the ground surface) to about
0.65 at a depth of 10 m (Fig. 3.13a). Pre-loading (step iii) increases the vertical effective
stresses uniformly, thus causing further compaction of the soil (Fig. 3.13a). After unloading
(step iv) OCR varies from more than 8 close to the ground surface to about 1.5 at a depth of
10 m (Fig. 3.13b). Consistently with the distribution of the OCR, the coefficient of lateral
pressure varies from more than 3 close to the surface to about 1 at a depth of 10 m (Fig.
3.13c).

49
This simulation procedure was repeated for several values of over-consolidation surface
pressure 'OC between 30 kPa and 100 kPa, resulting in the OCR-distributions of Figure 3.14.
The effect of the OCR on stand-up time will be discussed later. First, the model behaviour
will be analysed by taking the numerical results obtained for an over-consolidation surface
pressure 'OC of 50 kPa, for which the current stress, the over-consolidation stress and the
OCR at the mid-height of the excavation are equal to 25 kPa, 75 kPa and 3, respectively.

3.4.2 Development of pore pressures, strains and stresses

After the excavation-induced instantaneous drop, the pore pressure increases and the mean
effective stress p' decreases (Fig. 3.15). As it will detailed explained in Chapter 4, the
progression of the shear band is governed by the dissipation of the excess pore pressures:
Figure 3.16a shows the evolution of the plastic zone and the shear stress along the sliding
plane qualitatively over a time interval during which the tip of the plastic zone moves from
point L to point M. The increase in pore pressure leads to a gradual reduction in the effective
stress normal to the sliding plane both inside and ahead of the plastic zone (Fig. 3.16b).
Consequently, the shear stresses decrease inside the plastic zone (point L, Fig. 3.16b) and, in
order to satisfy equilibrium, increase ahead of its tip (point M, Fig. 3.16b), thus causing
failure propagation. This stress redistribution process continues until the shear band reaches
the soil surface.
Fig. 3.17a shows the stress path corresponding to the redistribution process in the (p', q)-space
at point B of our numerical example.
During the undrained excavation (time lapse between 0 and 0+), the mean effective stress
remains constant, but the mean total stress decreases, while the deviatoric stress initially
decreases and then increases again. This non-monotonous change is associated with the
higher than unity in situ coefficient of lateral stress (the horizontal unloading initially
decreases and later increases the stress anisotropy).

50
 Chapter 3 – Manifestation of failure

(a)

(b)

Figure 3.16. (a) Qualitative evolution of the plastic zone and the shear stress n along the
sliding plane over a fictitious time interval [t, t + t], (b) stress path at points L
and M in the (’n, n)-space.

(a) (b)
50 50
'OC = 50 kPa 'OC = 100 kPa
q [kPa]

q [kPa]

40 40
Initial yield curve ∞

30 330 30 t [h] = 0
0+

0+ p' p
20 1400 20
p' (0+)
1500 p

t [h] = 0
10 10
(0+)

0 0
0 20 40 60 80 0 20 40 60 80
p', p [kPa] p', p [kPa]
Point (0+): undrained, before complete unloading Point (0+): undrained, before complete unloading
(x'/'x0 = 40%) (x'/'x0 = 60%)

Figure 3.17. MCC behaviour: Effective and total stress path at point B (see inset in Fig. 3.15)
in the (p', q)-space and (p, q)-space, respectively, for an over-consolidation
surface pressure of, (a) 'OC = 50 kPa and, (b), 'OC = 100 kPa
(parameters according to Table 3.1).

51
 35

angle [ ]
Mobilized friction angle
30

25

20

15
Lode’s angle

10
B
2.5 m
5

1.8 m
0
0 400 800 1200 1600
t [h]

Figure 3.18 MCC behaviour: Lode’s angle and corresponding mobilized friction angle at
point B (see inset) for a pre-consolidation pressure of 50 kPa (material
parameters according to Table 3.1).

During the swelling process (t > 0+), in a similar way to the qualitative explanation in the
(’n – n)-space (Fig. 3.16), the decrease in the mean effective stress p' in combination with
the increase in the deviatoric stress q leads to yielding on the dry side of the Critical State
Line at t = 300 h. During this time period the total stress p remains almost constant (Fig.
3.17a). The subsequent shearing of the soil (t > 300 h) consequently involves softening (the
yield surface shrinks), which continues until the stress path reaches the CSL and causes a
decrease in the total stress p. At the same time, the void ratio (Fig. 3.21) increases slightly
with decreasing mean effective stress as long as the stress state is within the elastic domain (t
< 330 h, Fig. 3.17), moving then towards the CSL. Moreover, the stress state rotates on the
yield surface in the deviatoric plane until it reaches the zero Lode angle (Fig. 3.18; this is due
to the von Mises plastic potential at critical state, see Potts and Gens, 1984; Lagioia and
Panteghini, 2014). Clearly, the mobilized friction angle prior to failure (and thus outside the
shear zone) is higher than 30° (it reaches 32° at the end of undrained excavation; Fig. 3.18). It
should be noted, however, that during pre-consolidation loading (step iii in Fig. 3.13), the
stress state is at the triaxial compression corner (φ' = 22.2°), while during the subsequent
unloading (step iv in Fig. 3.13) it moves to the triaxial extension corner (φ' = 30.4°). The
plastic dilation that occurs after yielding (t > 750 h) causes a slight decrease in the pore
pressure (Fig. 3.15). At about t = 1200 h, the plastic zone reaches the soil surface (Fig. 3.19a)
and the displacements of crest point A increase rapidly (solid line in Fig. 3.19c), indicating
the onset of failure. The plastic strain localizes in a narrow band (Fig. 3.19b), as in the non-
dilatant MC problem (Fig. 3.7a).
For reasons similar to those set out in the last section, the results exhibit a spatial oscillation
approaching the ultimate state (Fig. 3.20), thus making it difficult to determine the point of
failure precisely.

52
 Chapter 3 – Manifestation of failure

'OC = 50 kPa (a)

t [h] = 1450

1150
850

 xypl  2%

Nearly ultimate state

60
(b)
uA,x [mm]

'OC [kPa] =
40 40 50 60 70 80

90
20
100

0
+
0 2000 4000 6000 8000
t [h]

-20

uA,y
uA,x
-40
uA,y [mm]

-60

Figure 3.19. MCC behaviour: Time-development of of, (a), the 2% contour line of the shear
plastic strain for a pre-consolidation pressure of 50 kPa and, (b), of the post-
excavation crest displacements (parameters according to Table 3.1).

53
 50 (a)
'OC = 50 kPa

pw [kPa]
40
2m 1 2 3 4 x [m]
30
t [h] =
1400
750
20
150
0+

10

30 (b)
'x [kPa]

25 1400
t [h] =
20
0+

15

10
150

5
750

0
0 1 2 3 4
x [m]

Figure 3.20. MCC behaviour: Distribution along a horizontal line (see inset in lower
diagram), (a), of the pore pressure pw and, (b), of the effective horizontal stress
'x (parameters according to Table 3.1).

0.75
Void ratio e

'OC = 50 kPa

0.72
1500

0.69
1400

0.66

B
330 t [h] = 0+
2.5 m
0.63

1.8 m
20 25 30 35 p' [kPa]
0.6
2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 ln p'

Figure 3.21. MCC behaviour: Evolution of the void ratio e and of the mean effective stress p'
at point B for a pre-consolidation pressure of 50 kPa (material parameters
according to Table 3.1).

54
 Chapter 3 – Manifestation of failure

OCR
1 2 3 4 5
15'000

ts [h]
12'000

9'000

6'000

3'000

0
0 20 40 60 80 100
'OC [kPa]

Figure 3.22. MCC behaviour: Stand-up time ts as a function of the over-consolidation

pressure 'OC at the soil surface and of the OCR at the cut mid-height,
respectively (parameters according to Table 3.1).

3.4.3 Effect of OCR on stand-up time

A higher OCR would result in a larger initial yield surface and would necessitate a larger
decrease in mean effective stress and a longer swelling time up to the onset of yielding. As
the rate of pore pressure dissipation decreases over time, even a small difference in the initial
yield surface may have a major influence on the time to failure. For the sake of comparison,
Figure 3.17b shows the total and effective stress paths at point B for the case of a higher over-
consolidation surface pressure. In this case, the effective stress path is located entirely inside
the elastic domain, which means that the plastic zone does not propagate to point B and thus
the excavation remains stable (see also dashed curves for 'OC = 100 kPa in Fig. 3.19b). The
sensitivity of stand-up time with respect to the past surface surcharge 'OC is remarkable (Fig.
3.22).

55
3.5 Conclusions
In order to improve the understanding of the mechanism behind delayed failure and to
identify the criteria to be used for determining the stand-up time of a geotechnical structure,
Chapter 3 investigated the relatively simple problem of an underwater vertical cut by means
of hydraulic-mechanical coupled finite element analyses.
Starting with the borderline cases of undrained and drained conditions Chapter 3 showed that,
depending on the effective shear strength parameters, a vertical cut may either fail during
excavation, experience delayed failure (i.e. be stable in the short-term but fail in the long
term) or be stable in the long term. Stability can be determined by observing the evolution of
the displacements at the crest of the vertical cut. It does not depend significantly on the plastic
flow rule.
Successively, Chapter 3 studied the transient process under the assumption of dilatant
behaviour and a parameter set for which the excavation should experience delayed failure.
According to the presented results, the displacements at the ultimate state, rather than
accelerating (as in the common conception of a system at failure), continue to increase
linearly over time, while the hydraulic head field remains stationary but non-uniform.
This is due to the simplifying constitutive assumption of a constant, non-zero dilatancy angle,
which implies that accelerated shearing must be accompanied by an accelerating increase in
the water content in the shear zone and thus by negative excess pore pressures of
continuously increasing magnitude, which is physically impossible. This behaviour was
illustrated here with reference to a simple and widely-used linearly elastic, perfectly plastic
constitutive model. However, on account of the underlying mechanism outlined above,
softening constitutive models with constant dilation angle would exhibit the same behaviour
after reaching residual strength.
In conclusion, the widely made, simplifying but nevertheless unrealistic constitutive
assumption of a constant non-zero dilation angle not only delays failure, but actually prevents
it from reaching the state of accelerating displacements, which is commonly associated with
the notion of failure. The constitutive assumption about plastic dilatancy affects the model
behaviour not only quantitatively, but also qualitatively.
In order to obtain accelerating displacements at failure, it is essential to adopt constitutive
models that map the soil property to experience shearing at a constant volume (e.g. elastic,
perfectly plastic models with variable dilatancy or models from the critical state family).
Isochoric shearing leads, however, to numerical stability problems close to failure because it
is inherently impossible to satisfy both the equilibrium and the mass balance conditions. In
order to estimate the failure time, it is therefore essential to evaluate the evolution of
displacements and effective stresses simultaneously, as well as the stability behaviour of the
iterative numerical solution algorithm.

56
 Chapter 4 – Mesh dependency

4. Mesh dependency

4.1 Introduction
According to the qualitative explanations of Section 1.1.2 and Chapter 2 the structural
softening (which occurs if the plastic flow rule is non-associated and if the plastic
deformations localize in shear bands) causes a mesh-dependency of the stand-up time.
Chapter 4 investigates the influence of the mesh coarseness on the stand-up time starting with
the underwater cut problem of Chapter 3 (Section 4.2). Subsequently, the very simple
problem of an axially loaded rectangular specimen under bi-axial strain conditions will be
considered (hereafter referred to as "biaxial problem"; Section 4.3). The simplicity of the
biaxial problem allows, (a), estimating stand-up time also analytically and, (b), performing
comparative computations with extremely fine meshes.
The analysis of the underwater cut problem (Section 4.2) shows that the spatial discretization
affects the stand-up time when assuming a material obeying to a non-associated flow rule.
According to the numerical results of Sections 4.2 and 4.3 the occurrence of failure is
associated with the formation of a shear band (i.e. a region with concentrated intense shear
deformation), which propagates from the bottom of the model towards its top over time.
Failure occurs when the shear band fully develops through the model. The propagation of the
shear band is mainly due to a combination of (i) a reduction of the shear resistance caused by
the dissipation of the excess pore pressure (leading to a reduction of the effective stresses)
and (ii) a progressive reduction of the operational shear strength caused by structural
softening. According to the numerical results, the propagation speed of the shear band (and
thus the stand-up time as well) is mesh-dependent. The finer the mesh, the rapider the
propagation speed and the shorter the stand-up time will be.
An analytical solution is derived for the stand-up time of the biaxial problem, considering two
borderline cases concerning amount of structural softening: no softening at all (i.e., no
rotation of the principal axes and, consequently, no reduction of the shear strength
parameters, and, maximum softening (i.e., rotation of the principal axes up to the
development of purely plastic shear strains). These two borderline cases provide an upper and
a lower limit of the stand-up time, respectively. The numerically predicted stand-up times are
between these two bounds.
In order to practically deal with the mesh dependency of the stand-up time, Chapter 4
suggests a way of estimating a lower limit of stand-up time without the need to implement
regularization technique: It can be determined by performing few computations with a
relatively fine mesh and extrapolating their results to the mesh size that corresponds to the
expected, grain-size-dependent thickness of the shear band. The proposed method is based
upon the observation that dependency of the stand-up time on the size of the elements
becomes almost linear for relatively fine meshes and it is supported by the analytical solution
(Chapter 4), which shows that there exist a non-zero bound of the stand-up time.

57
4.2 Underwater vertical cut
The purpose of the present Section is to investigate the role of the discretization on the
manifestation of failure and on the stand-up time on the basis of the same problem of Chapter
3. In order to check the plausibility of the numerical results, the present Section finally
compares the numerically predicted inclination of the shear band with the theoretically
expected inclination.
For linearly elastic, perfectly plastic behaviour obeying Mohr-Coulomb yield criterion and the
Modified Cam-Clay model (assuming the material parameters of Table 3.1), the element size
has no influence on the model behaviour (failure manifests itself by an acceleration of the
displacements rate to infinity when assuming materials that allow for shearing under constant
volume) but influences the stand-up time quantitatively when considering a non-associated
plastic flow: the stand-up time decreases with decreasing element size; the finer the
discretization, the earlier the system fails (Fig. 4.1).
(a) (b)
60 10
uA,y [mm]
uA,y [mm]

MCC MC
'OC = 50 kPa c' = 4.5 kPa
50 φ' = 30
8
=0
40
6
30
4
20
A
2
10 u A,y

0 0
0 2000 4000 6000 8000 0 30 60 90 120 150
t [h] t [h]

(c) (d)
4 2
uA,y [mm]
uA,y [mm]

MC MC
3.5 1.8
c' = 4.5 kPa c' = 4.5 kPa
φ' = 30 1.6 φ' = 30
3  = 15  = 30
1.4
2.5 1.2
2 1

1.5 0.8
0.6
1
0.4
0.5 0.2
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
t [h] t [h]

Mesh size s:
5 cm (s/ H = 1%)
10 cm (s/ H = 2%)
50 cm (s/ H = 10%)

Figure 4.1. Time-development of the vertical crest displacement for several values of mesh
size (parameters according to Table 3.1).

58
 Chapter 4 – Mesh dependency

4.2.1 Inclination of the sliding surface

Figures 4.2 and 4.3 compare theoretical predictions with the results of a parametric study into
the effect of the mesh size on the average inclination of the sliding surface at ultimate state
(note that the inclination at the ground surface always correspond to the Roscoe’s solution).
Mesh sensitivity was investigated both for drained and for transient analyses (Figs. 4.2 and
4.3).
The inclination of the shear band at the ground surface, where the principal axes are fixed,
agrees with the Roscoe’s plasticity solution (see Davis, 1968; Roscoe, 1970), while the
average inclination lies between the Coulomb’s and Roscoe’s solutions (circular marks in Fig.
4.2).
As shown by Vermeer and De Borst (1984) and Vermeer (1990), the admissible inclination of
the shear band is between Roscoe's and Coulomb's solution (i.e., between 45° + ψ/2 and 45° +
φ'/2). Vardoulakis (1980) explained theoretically that Coulomb's and Roscoe's solutions
represent upper and lower bounds, respectively, for the shear band inclination, showing by
means of bifurcation analysis that the inclination of the sliding surface should be equal to the
mean of these solutions, i.e. to 45° + (φ' + ψ)/4, and finding that this agrees well with
experimental observations (e.g. Arthur et al., 1977). In the special case of associated plastic
flow the three solutions coincide.
The numerically predicted sliding surface becomes steeper with decreasing mesh size and the
obtained range of angles fall within the theoretical range bounded by the Coulomb’s and
Roscoe’s solution (Fig. 4.2). For associated flow-rule the inclination amounts to the
Coulomb’s solution for all the considered mesh sizes. According to Figure 4.3, the influence
of the discretization on the average inclination of the sliding plane is less pronounced in the
case of the MCC model (for a surface over-consolidation pressure of 50 kPa).

59
 65

Inclination of the sliding plane [ ]

c' = 4.5 kPa, ' = 30

Coloumb’s solution
60

55

50
drained FEM – surface
drained FEM – average
transient FEM – surface
transient FEM – average
45
0 5 10 15 20 25 30

Dilation angle  [ ]
Figure 4.2. MC behaviour: Inclination of the sliding plane as a function of the dilation angle
for drained and transient FEM analyses (parameters according to Table 3.1).

65
T ransient analyses
MCC, 'OC = 50 kPa
Inclination of the sliding plane  [ ]

Coloumb’s solution
60

55

50

Roscoe’s solution
45

40
0 10 20 30 40 50 60 Mesh size s [cm]

0% 2% 4% 6% 8% 10% 12% s / H [-]

Figure 4.3. MCC behaviour: Inclination of the sliding plane as a function of the mesh size
for transient FEM analyses (parameters according to Table 3.1).

60
 Chapter 4 – Mesh dependency

4.2.2 Critical strength parameters

Figure 4.4a shows the analytical relationships between the friction angle and the cohesion that
are required for stability under undrained (solid line) or drained conditions (two upper lines)
as well as numerical results for drained stability, ' = 30° and three mesh sizes (black dots).
The dashed and the dot-dashed lines in Figure 4.4 are obtained with and without taking into
account the reduction of the strength parameters caused by structural softening (solution
derived by Drescher and Detournay, 1993), respectively.
With a coarse mesh (50 cm) a critical strength is numerically predicted which is close to the
analytical solution obtained disregarding the reduction of the operational strength due to
principal axes rotation. The finer the mesh, the closer the numerical prediction to the solution
considering the maximum reduction of the shear strength caused by structural softening will
be.

(a) φ' = 30 (b)

10 stable under 10
drained conditions mesh size mesh size
c' [kPa]

= =
9 9
drained stability 5 cm
depends on the 10 cm 5 cm (s/H = 1%)
8 mesh size 8
10 cm (s/H = 2%)
50 cm
7 7
50 cm (s/H = 2%)
unstable under drained
6 conditions 6

5 5

4 4

3 unstable under undrained 3

conditions
2 2

1 1 φ' = 30

0 0
5 10 15 20 25 30 35 0 500 1'000 1'500
' [ ] ts [h]

Undrained (analytical solution) T ransient FEM (  = 0 )

Drained (analytical solution obtained considering full rotation
of the principal stresses inside the shear band for  = 0 )
Drained (analytical solution obtained without considering
the rotation of the principal stresses inside the shear band)
Drained FEM (  = 0 )

Figure 4.4. MC model. (a) Critical strength parameters for undrained and drained stability
and influence of the mesh size for drained FEM stability analyses; (b) Stand-up
time as a function of the cohesion for mesh size 5 cm, 10 cm and 50 cm
(parameters according to Table 3.1).

61
4.2.3 Stand-up time

Figure 4.4b shows the stand-up time as a function of the cohesion for the three mesh sizes. In
accordance with the qualitative explanation of Section 1.1.2, the influence of the spatial
discretization on the stand-up time is particularly pronounced when the cohesion is close to
the critical value for drained stability. For cohesion values of Figure 4.4a in the range
bounded by the analytical solutions obtained taking into account or disregarding the effect of
structural softening, the stand-up time can be finite or infinite depending on the mesh size.
Consider, for example, a cohesion of 7.5 kPa. The stand-up time corresponds to the
intersection point (white dots) between the dashed line of Figure 4.4b and the curves obtained
for three mesh sizes. The stand-up time amounts to 440 hours and to 670 hours for mesh size
5 cm and 10 cm, respectively. Thus, the increase of mesh size from 5 cm to 10 cm causes a
finite increase of stand-up time of about 230 hours. The curve obtained for mesh size 50 cm,
however, does not intersect the dashed line corresponding to 7.5 kPa because, the stand-
uptime increases asymptotically to infinity approaching a cohesion value of about 7 kPa. This
means that for mesh size 50 cm and cohesion 7.5 kPa, the stand-up time is infinite, so that the
change of stand-up time caused by the increase of mesh size from 10 cm to 50 cm is infinite
as well.
The dependency of the stand-up time on the mesh size is additionally illustrated by the
diagrams in Fig. 4.5. For a cohesion c of 4.5 kPa the stand-up time depends almost linearly on
the mesh size; for c = 7 kPa the mesh dependency of the stand-up time becomes strongly non-
linear; and, for c = 7.5 kPa, the stand-up time becomes infinite for mesh sizes larger than 20–
30 cm, which means that the spatial discretisation influences the predicted response
qualitatively. This observation can be explained taking into account the reduction factor
defined in Section 2.2. For a friction angle of 30° and dilation angle of zero, the reduction
factor varies between 1.0 and 0.87. The strength parameters corresponding to a cohesion of
7.5 kPa and a friction angle of 30 ° reduces consequently to a minimum of 6.5 kPa and 26.7°.
The vertical cut is stable for large element size because the corresponding strength parameters
(7.5 kPa and 30 °) are high enough in order to satisfy the stability under drained conditions
(strength parameters located above the dashed line of Fig. 4.4a), while for smaller mesh size
the mobilized strength parameters are such that, the excavation collapses before reaching
drained conditions (minimum strength parameters 6.5 kPa and 26.7° are located below the
dashed line of Fig. 4.4a).

62
 Chapter 4 – Mesh dependency

(a)
150

ts [h]
c' = 4.5 kPa
φ' = 30
125 =0

100

75

50

25

0
(b)
3'000
c' = 7.0 kPa
ts [h]

φ' = 30
2'500 =0

2'000

1'500

1'000

500

0
(c)
2'000
c' = 7.5 kPa ts = 
ts [h]

1'750 φ' = 30 (vertical cut stable

=0 under drained
1'500 conditions)

1'250

1'000

750

500

250

0
0 10 20 30 40 50
mesh size [cm]

0% 2% 4% 6% 8% 10%
s / H [-]
Figure 4.5. MC model. Stand-up time as a function of the mesh size for cohesion (a) 4.5 kPa,
(b), 7 kPa and, (c), 7.5 kPa (other parameters according to Table 3.1).

63
4.3 Biaxial problem

4.3.1 Problem definition

The problem under consideration is a rectangular specimen under biaxial conditions (Fig.
4.6). The specimen is initially under an isotropic total stress of σx0 = σy0 = 300 kPa and a pore
pressure of pw0 = 100 kPa; subsequently, it is subjected to undrained unloading along its right
boundary PQ by reducing the boundary traction sx from 300 kPa to 100 kPa rapidly (sx = -
200 kPa), while keeping the boundary traction sy along the upper boundary QR constant to
300 kPa.
The undrained unloading of the model generates a uniform negative excess pore pressure,
which can be calculated based upon Hooke's equation and imposing the restriction of no
volumetric deformation:

sx
pw (y,0 )  , (4.1)
2

Subsequently, a pore pressure of 100 kPa (as in the initial state) is prescribed at the drained
boundary OP (the remaining boundary PQRO is considered as being impervious), thus
initiating an excess pore pressure dissipation process.
The material behaviour is taken linearly elastic, perfectly plastic according to Mohr-Coulomb
(MC) yield criterion with isochoric plastic flow ( = 0°) and the material constants after
Table 4.1. The material parameters and the boundary conditions are such that the model is
stable immediately after unloading but unstable under drained conditions.

y0 = sy = 300 kPa

R Q
qn = 0

x0 = 300 kPa,

H = 100 mm qn = 0 p w0 = 100 kPa qn = 0
sx = 100 kPa
L
y s

x p w = 100 kPa 
O P

200 mm

Figure 4.6. Biaxial problem layout, initial and boundary conditions.

64
 Chapter 4 – Mesh dependency

Table 4.1. Assumed material constants.

Unit weight of water w [kN/m3] 10

Water compressibility cw [MPa] 0

Young’s modulus E [MPa] 100
Poisson’s ratio  [-] 0.3
Angle of internal friction  [°] 30
Dilation angle  [°] 0
Cohesion c [kPa] 50
Permeability k [m/s] 10-9

4.3.2 Analytical solution for the estimation of the stand-up time

4.3.2.1 Derivation

The analytical solution is based upon the limit equilibrium of a sliding wedge (delimited by
the dashed line in Fig. 4.6). The stand-up time is defined as the time at which the pore
pressures along the sliding plane have increased so much that the considered wedge reaches
limit equilibrium.
The equilibrium conditions of the sliding wedge parallel and perpendicular to the sliding
plane are

  s, t  ds  H cos  s
0
n y  sx  (4.2)

and

L
H
   s, t  ds  s
0
n y
sin 
cos2   sx H sin  , (4.3)

wheren is the shear stress along the plane, n is the total stress normal to the plane, H the
height of the model, L the length of the sliding plane, sy and sx the vertical and the horizontal
boundary traction after unloading and  the inclination of the sliding plane.
At ultimate state the shear stress along the sliding plane fulfil the MC- failure criterion

L L

 n  s, ts  ds    c   n  s, ts   pw  s, ts  tan    s, ts  ds ,
0 0
(4.4)

where c and  are the effective strength parameters of the ground, η(s, ts) is the reduction
factor after Section 2.2 and pw(s, ts) is the pore pressure at the ultimate state (ts is the stand-up
time).

65
Introducing τn from Eq. (4.2) into Eq. (4.4), assuming that the reduction factor at the ultimate
state is approximately constant over the shear band and considering the integral of σn after Eq.
(4.3) leads to

sin  cos
1
c
tan 
 sx sin 2   s y cos2  
 (ts ) tan 
 sy  sx    pw ( , ts ) d , (4.5)
0

where ξ = s/L is the normalized local coordinate.

The r.h.s. term of Eq. (4.5) represents the average pore pressure in the model at the ultimate
state and can be written as the sum between the initial pore pressure pw0 (which is uniform)
and the excess pore pressure ∆pw at time ts:

1 1

 pw ( , ts ) d  pw0   pw ( , ts ) d .
0 0
(4.6)

Equation (4.5) requires thus the determination of the excess pore pressure at time ts.
For the given hydraulic boundary conditions (drained bottom boundary and no-flow condition
applied to the other boundaries) and taking into account that the shear band is thin, it is
reasonable to assume that the stress and strain field are uniform in the horizontal direction and
the water flow is one-dimensional (from the bottom boundary to the top). This allows to
express the Darcy’s law and the mass balance condition as:

k  2 pw
 vol   , (4.7)
 w y 2

where the volumetric strain rate increment  vol   x   y , while k and γw denote the hydraulic
conductivity and the unit weight of water, respectively.
Under the assumption of elastic behaviour (which is reasonable considering (i) the small
thickness of the shear band compared to the model size and (ii) isochoric plastic flow), the
increments of the effective stresses are related to the strain increments by the Hooke’s law:

 'x    x   y   2 x , (4.8)

 ' y    x   y   2 y , (4.9)

where  and μ are the Lamé’s constants.

Assuming that the total stresses remain constant over time (which is reasonable under the
assumption of constant boundary conditions, loads and for the assumption of elastic material
behaviour), the change of effective stress is due to the change of pore pressure only:

 'x    x   y   2 x   pw , (4.10)

66
 Chapter 4 – Mesh dependency

 ' y    x   y   2 y   pw . (4.11)

From Eqs. (4.10) and (4.11) it follows

pw
x  y   , (4.12)


which inserted into Equation (4.7) gives

 2 pw
pw  cv , (4.13)
y 2

where cv is the coefficient of consolidation for the considered problem:

k k E
cv       , (4.14)
w  w 2 1   1  2 

where E is the Young’s modulus and  the Poison’s number.

For the initial and boundary conditions of the considered problem, the solution of Equation
(4.13) gives the excess pore pressure at elevation y and time t (Taylor, 1948):


2pw (y,0 )  My   M 2Tv
pw ( y, t )   sin  e , (4.15)
m 0 M  H 

where Tv is a dimensionless time,

cv t
Tv (t )  (4.16)
H2 ,

and


M  2m  1 . (4.17)
2

From Equations (4.6) and (4.15):

p w ( , ts ) d  pw0U  ts  , (4.18)
0

where


2
U (ts )  1   2 
cos M  1  e M Tv
2
(4.19)
m0 M .

67
It can easily be verified that U represents the consolidation degree. For the computation of the
consolidation degree (Eq. 4.19) it is sufficient to consider the first 15 terms of the series (m =
15). Inserting the average pore pressure from Eq. (4.18) into Eq. (4.5) leads to the following
non-linear equation for the stand-up time:

1  c sin  cos 
U (ts )  
pw0  tan 
 sin 2  sx  cos2  s y 
 tan 
 s y  sx   . (4.20)


Eq. (4.20) has a solution only if its r.h.s. – which is constant – is between 0 and 1. One can
easily verify that if the r.h.s. term is negative, then the system would be unstable already at t =
0+ (i.e., under undrained conditions), while if it is higher than 1, then the system would be
stable even under drained conditions. In order to avoid the numerical solution of the non-
linear Equation (4.20), it is advantageous to solve it with respect to c and compute the
cohesion that is required for given stand-up time ts.
The evaluation of Eq. (4.20) requires an assumption concerning the inclination of the
sliding surface. The latter corresponds to the angle which minimizes the stand-up time (and so
maximizes the required cohesion).

4.3.2.2 Upper and lower limit of the stand-up time

For given material parameters and boundary stresses (sx, sy), the stand-up time (resulting from
the solution of Eq. 4.20) depends only on the reduction factor η. Taking the upper and lower
bound of the reduction factor derived in Section 2.2 (η = 1 and according to Eq. 2.12,
respectively), Equation (4.20) provides an upper and a lower bound of the stand-up time,
respectively.
1500
upper limit of the
ts [s]

stand-up time

1250 lower limit of the

stand-up time

1000

750

φ = 30
500
30
20

250 10

0
0 20 40 60 80 100
c [kPa]

Figure 4.7. Upper and lower limit of the stand-up time (other parameters according to
Tab. 4.1).

68
 Chapter 4 – Mesh dependency

Figure 4.7 shows these bounds for three values of friction angle. The influence of the
structural softening on the stand-up time increases with the friction angle. As discussed in
Section 4.2.3, this difference is particularly pronounced when the cohesion approaches the
critical value for which the problem is stable under drained conditions. For such cohesion, the
stand-up time is infinite when the rotation of the principal axes is neglected, but reduces to a
finite value when taking into account the full rotation of the axes. In conclusion, for cohesion
values close to the critical one, the structural softening influences the stand-up time
extremely.
Finally, for the parameters given in Table 4.1, the upper limit of the stand-up time is 759 s
and the lower one 335 s.

4.3.3 Numerical estimation of the stand-up time

4.3.3.1 Computational model

Section 4.3.3 solves the problem presented in Section 4.3.1 numerically. The width of the
model (Fig. 4.6) is chosen sufficiently big in order to allow the formation of low inclination
shear bands.
The spatial discretization of the problem varies from an extremely fine mesh consisting
400x200 elements to a very coarse one composed by 16x8 elements. The shear band
propagation mechanism will be discussed in detail for the 200x100 mesh. The analyses are
carried-out with 8-node square elements (plane strain quadrilateral, biquadratic displacement,
bilinear pore pressure).

4.3.3.2 Shear band propagation mechanism

4.3.3.2.1 Elastic behaviour at early excess pore pressure dissipation stage

Over the first 215 s the model behaves elastically, which is unexpected on account of the
prescribed boundary conditions. More specifically, one would expect that yielding would
occur immediately after prescribing the boundary pore pressure of 100 kPa because at the
bottom-right corner of the model the expected effective stress state would almost correspond
to uniaxial loading (horizontal and vertical stresses would instantaneously become equal to
about 0 and 200 kPa, respectively) higher than the compressive strength of the material
(approximately 173 kPa).

69
 (a)
210

 'y [kPa]
t [s]
=
200 2
25
200

190

180

170

160

150
(b)
60
'x [kPa]

2
50

25
40

30
200

20

10

0
(c)
0.4
'xy [kPa]

0.4 y
100 mm
0.3
x
0.3

200 mm
0.2

0.2

0.1

0.1
200
25
0.0 2
0 0.05 0.1 0.15 0.2
x [m]

Figure 4.8. Spatial distribution of the vertical effective stresses (a), horizontal effective
stress (b) and shear stresses (c) along the axis y = 0.2 mm.

70
 Chapter 4 – Mesh dependency

(a) (b)
0.10

y [m]
0.09 100 mm y

0.08 x

0.07
200 mm
0.06 200
200
0.05 150 150
100 100
0.04 50
50 2
0+
0.03

0.02 t [s]
=
0.01 2
0+
0.00
-20 0 20 40 60 80 100 0 1 2 3 4 5
p w [kPa] σxy [kPa]

(c) (d)
0.10
y [m]

0.09

0.08

0.07

0.06
200
0.05 200
150
150
0.04 100 100

0.03 50 50

0.02

0.01 2 2
0+ 0+
0.00
0 20 40 60 80 100 120 200 220 240 260 280 300 320
σ'x [kPa] σ’ y [kPa]

Figure 4.9. Distribution along the y-axis at x = 0.1 m of the, (a), pore pressure, (b), shear
stresses, (c), horizontal and, (d) vertical effective stresses.

What actually happens, however, is an elastic stress redistribution, which results to a lower
than 200 kPa effective vertical stress in the vicinity of right boundary (150 kPa at t = 2 s,
rising to 170 kPa at t = 200 s; see distribution of effective vertical stress along the bottom
boundary in Fig. 4.8a). The stress field satisfies equilibrium in the y-direction because the
effective vertical stresses away from the right boundary (for x < 160 mm) are slightly higher
than 200 kPa. The horizontal effective stress is zero at the right boundary (in accordance with
the imposed boundary conditions), increases toward the impervious left boundary and
decreases over time due to the excess pore pressure dissipation (Fig. 4.8b). Shear stresses
occur only at the right bottom corner (Fig. 4.8c), but are very low relatively to the normal
stresses, which means that the principal stress axes remain practically vertical and horizontal.

71
 t =2 s
(a) Pore pressure, p w

(b) Horizontal effective stress, ’x

(c) Vertical effective stress, ’ y

(d) Deformed mesh (scaled-up 200 times)

Figure 4.10. Spatial distribution at t = 2 s of the, (a), pore pressure, (b), horizontal effective
stresses and, (c), vertical effective stresses; (d) deformed mesh at
t = 2 s.

72
 Chapter 4 – Mesh dependency

t = 200 s
(a) Pore pressure, p w

(b) Horizontal effective stress, ’x

(c) Vertical effective stress, ’ y

(d) Deformed mesh (scaled-up 200 times)

Figure 4.11. Spatial distribution at t = 200 s of the, (a), pore pressure, (b), horizontal effective
stresses and, (c), vertical effective stresses; (d) deformed mesh at
t = 200 s.

73
Fig. 4.9, 4.10 and 4.11 provide a more complete picture of the stress- and displacements fields
at the early process stage. The pore pressure of 100 kPa imposed to the bottom boundary
causes the increase of the pore pressure in the model over time (Fig. 4.9a) and, consequently,
a decrease of the effective stresses in the horizontal (Fig. 4.9c) and vertical (Fig. 4.9d)
direction. It should be noted that the pore pressure becomes slightly negative in the upper part
of the model after the application of the boundary pore pressure of 100 kPa at the lower
boundary. This causes a slight increase of the effective stresses during the early consolidation
stage (according to Fig. 4.9, for t < 100 s). The reason of the slight increase in the pore
pressure is not clear and may be purely numerical.

4.3.3.2.2 Onset of yielding

Yielding starts to occur at t = 215 s, close to (but not exactly at) the bottom-right corner of the
model. Figure 4.12 shows the effective stress paths in the principal stress space at two
integration point – the lower- and right-most point (point E) and the point that experiences
first yielding (point Y). The latter is the starting point of the progressive shear band
propagation. The latter does not start exactly at the corner because of the mentioned stress re-
distribution (a similar observation was made by De Borst, 1989, Ehlers and Volk, 1998 and
Collins et al., 2006 in the numerical simulation of bi-axial tests). The location of the
initialization remains the same even by imposing an imperfection (slightly lower cohesion to
the element at the right bottom corner).
With ongoing time, the plastic zone progresses towards the upper (impervious) boundary of
the model (Fig. 4.13).

74
 Chapter 4 – Mesh dependency

350

 'max [kPa]
MC-line
0+
300

250

2s t =0
200 Y
215 s
215 s E
150
2s

100

50

0
0 50 100 150 200 250
 'min [kPa]

Plastic zone at t = 215 s:

y
5 mm
100 mm

x
Y E
(192.8 mm, 0.2 mm) (199.8 mm, 0.2 mm)
200 mm

12 mm

Figure 4.12. Principal effective stress path at points Y and E for t ≤ 215 s.

530 s

100 mm 500 s

B (172.8 mm, 28.2 mm)

400 s
A (10.2 mm, 12.5 mm)

300 s

215 s

200 mm
Figure 4.13. Boundary of the plastic zone (the shear band) at different time periods.

75
4.3.3.2.3 The numerical identification of structural softening

The plastic deformations localize in a narrow, 1 – 2 elements wide shear band (Fig. 4.13). As
analysed by Leroy and Ortiz (1989) and shown below for the specific problem under
consideration, the onset of localization is associated with that of the structural softening. The
latter is numerically characterized by a non-positive value of determinant of the so-called
localization (or acoustic) tensor Aep,

det  Aep   det  N T Dep N   0 , (4.21)

where N collects the direction cosines of the unit vector perpendicular to the shear band and
Dep is the elasto-plastic stiffness matrix. As explained by Leroy and Ortiz (1989), Equation
(4.21) represents the condition for the existence of a discontinuity in the velocity field and
results from the incremental elastoplastic stress-strain relationships and the conditions of
strain compatibility across the discontinuity (i.e., zero incremental tangential strain),
equilibrium of the normal stress and of the shear stress acting upon the interface of the shear
band with the adjacent elastic domain.
Under plane strain conditions, Eq. (4.21) can be written as follows (Leroy and Ortiz,1989):

det  N T Dep N   a0 n14  a1n13n2  a2 n12 n22  a3n1 n23  a4 n24 , (4.22)

where

a0  D11ep D33ep  D13ep D31ep , (4.23)

a1  D11ep D23
ep
 D11ep D32ep  D13ep D21ep  D12ep D31ep , (4.24)

a2  D11ep D22
ep
 D13ep D32ep  D31ep D23ep  D12ep D21ep  D12ep D33ep  D33ep D21ep , (4.25)

a3  D13ep D22
ep
 D31ep D22
ep
 D12ep D23ep  D32ep D21ep , (4.26)

a4  D33ep D22
ep
 D32ep D23
ep
, (4.27)

where n1 and n2 are the components of the normal vector (n1 = cos() and n2 = sin() with 
denoting the angle between the normal to the shear band n and the x-axis; inset of Fig. 4.14a).
Figure 4.14a shows the determinant of the localization tensor (normalized by the determinant
of the elastic localization tensor) at point B at several times. For t < 430 s, the normalized
determinant amounts to 1 (as the material behaves elastically). At 430 s, the normalized
determinant becomes negative for 47° < < 62°, which means that the stress state at the
considered point B allows localization. The minimum of the determinant (Fig. 4.14b) occurs
at an angle  of about 54°, which is approximately equal to the numerically predicted shear
band inclination (about 54.5° according to Fig. 4.13).

76
 Chapter 4 – Mesh dependency

(a)

1.2

de t(Aep)/det(Ae)
1 t < 430 s
n

0.8   90 

0.6

0.4

t = 430 s
0.2

0
30 40 50 60 70 80 90
 []
-0.2

(b)
1.2
min(det(Aep)/det(Ae))

0.8

0.6

0.4
onset of localization at
0.2 point B (t = 430 s)

0
0 100 200 300 400 500
t [s]
-0.2

Figure 4.14. (a) Normalized determinant of the elasto-plastic localization tensor at point B of
Figure 4.13 as a function of the angle ; (b) Minimum of the normalized
determinant at point B over time.

77
4.3.3.2.4 The stress re-distribution process

The stress re-distribution process will be discussed considering the stress paths and history of
two points over a specific time interval (400 s, 430 s). At t = 400 s point A is located already
on the shear band, while point B is ahead of its tip (Fig. 4.15). The shear band reaches point B
at the end of the considered interval.
The initial negative excess pore pressure dissipates over time (Fig. 4.16). Under the
assumption of approximately constant effective stresses, the increase of pore pressure cause a
decrease in the effective stress 'n normal to the MC plane. For simplicity, the couple of
stresses ’n and τn in the present Chapter always refer to a plane inclined according to
Coulomb's solution (i.e. MC = 45+φ/2). Although the numerically computed inclination of the
shear band is slightly lower than Coulomb's solution (54.5° vs 60°, see Section 4.3.3.2.3), the
results obtained under the simplifying assumption of  = MC are qualitatively the same of the
one that would be obtained assuming the inclination computed numerically.
At t = 430 s (exactly when the shear band reaches point B), the effective stress over the time
curve (Fig. 4.16) exhibits an inflexion point, i.e. 'n rate accelerates suddenly, although the
pore pressure increases smoothly over time. The sudden change of the stress rate is related to
the occurrence of structural softening.
During the shear band propagation over the considered time interval, both point A and point
B (inside and ahead the shear band, respectively) experience a reduction in the effective
normal stress (Fig. 4.17) due to the ongoing excess pore pressure dissipation. As the stress
state at point A is on the yield surface, the decrease in the effective normal stress ’n implies
that the shear stress n decreases as well. As this happens everywhere along the shear band, its
resultant shear resistance decreases with the consequence that the shear stresses in the soil
ahead of its tip must increase in order to satisfy equilibrium. The stress state at point B
therefore moves upwards and comes closer to the yield condition.

B
R Q
430 s
40 mm
400 s

O P P
40 mm

Figure 4.15. Shear band evolution between t = 400 s and 430 s.

78
 Chapter 4 – Mesh dependency

Initial effective normal stress

200

p w ,  n' [kPa]
Point B
turns
plastic
150

Initial pore pressure

100 pw

’n

50

Pore pressure drop due to

undrained unloading
0
0 100 200 300 400 500
t [s]

Figure 4.16. Time-development of the pore pressure pw and of the effective normal stress ’n
at point B of Figure 4.13.

100
 n [kPa]

95

430 s

B
90
400 s

85
B
A 400 s
430 s
80 430 s 400 s

75
50 60 70 80 90 100
 n' [kPa]

Figure 4.17. Effective stress path at points A and B during the time interval 400 s – 430 s.

79
In conclusion, the progression of the shear band is caused by the combined effect of: (i) the
gradual reduction of the effective normal stress both inside and ahead of the plastic zone due
to excess pore pressure dissipation2; (ii) the gradual reduction of the shear resistance inside
the shear band due to (i) and structural softening; (iii) the gradual increase of the shear stress
ahead of shear band due to (ii).

4.3.3.3 Stand-up time

The process of stress re-distribution continues until the shear band reaches the top boundary
of the system; this happens at t = 530 s (see inset in Fig. 4.18). Afterwards, a stress re-
distribution, which would be necessary for equilibrium, cannot occur anymore and the system
collapses. As isochoric plastic flow is assumed, collapse manifests itself by an asymptotic
increase to infinity of the displacements of the unstable part of the model (Fig. 4.18; cf.
Chapter 3).
0.30
ux , uy [mm]

0.15 ux

0.00
0 100 200 300 400 500 600
t [s]

uy uy
-0.15
ux

-0.30

Figure 4.18. Time-development of the displacements at the top-right corner.

2
The role of the hydraulic-mechanical coupling was pointed-out already in the 1976 by Rice and
Simons. The authors investigated the steady crack propagation in a linear elastic fluid-saturated porous
solid under plane strain conditions. As explained later in Rudnicki (1983), the coupling between
deformation and diffusion results in a delay in the onset of failure for dilatant materials: when dilation
occurs more rapidly than fluid mass can diffuse into the newly created pore space, the pore pressure is
reduced and the effective compressive stress is increased. Due to the dependency of the frictional
resistance on the effective stresses, the increase inhibits further inelastic deformation. At the contrary
the inverse process (pore pressure increase) causes the reduction of the frictional resistance and has a
destabilizing effect in the sense that the pore pressure increases the effective driving forces (Rudnicki
and Koutsibelas, 1991).

80
 Chapter 4 – Mesh dependency

4.3.3.4 Influence of the element size on the stand-up time

4.3.3.4.1 Numerical results

The model behaviour was investigated for a series of mesh sizes varying between 0.5 mm and
12.5 mm (which is equal to 0.5 – 12.5% of the specimen height and results to a spatial
discretization by 400x200 – 16x8 elements, respectively). The numerical results (Fig. 4.19a)
allow drawing the following conclusions: The stand-up time increases from 455 s to 685 s
with increasing element size; the numerically predicted stand-up times are between by the
two analytically determined bounds after Section 4.3.2.2. The stand-up time trends to a
constant value for large elements, while it depends almost linearly on the element size for
small elements. As the element size approaches zero, stand-up time tends a non-zero value,
which is higher than the analytically determined lower bound, because the latter assumes that
the strength parameters decrease to the operational residual ones (η according to Eq. 2.12)
instantaneously over the entire shear band.
Furthermore, as observed in Section 4.2.1, the numerically predicted shear band inclination
increases from Roscoe’s solution (45º+ψ/2) for coarse elements to Coloumb’s solution
(45º+φ/2) for fine meshes (Fig. 4.19b).
Comparative coupled analyses with other material parameters allow drawing the same
conclusions (Figs. 4.20 and 4.21).
Pure stress analyses (without hydro-mechanical coupling) with coarse meshes predict, too, a
Roscoe inclination (Fig. 4.22). This was observed already by Kaus (2010) based upon
numerical stress analyses of shear localization around an inclusion of fixed length. The author
concluded that the main factor controlling shear band inclination is the ratio of inclusion to
finite element size; Coulomb angles are predicted only with fine meshes, while larger element
sizes result either to Arthur or Roscoe orientations. (Less inclined, Roscoe-type shear bands
are observed also experimentally in the case of coarse soils; Vermeer, 1990.)
A finer mesh allows for a bigger rotation of the principal axes inside the shear band (Fig.
4.23); rotation is negligible for the coarse mesh. (The principal directions in the elastic zone
outside the shear band remain anyway vertical and horizontal.)

81
 (a)
1000
c = 50 kPa,  = 30 ,  = 0

ts [s]
Analytical solution without considering principal
800 axes rotation (  = 1, ts = 759 s)

600 FEM

400 Analytical solution considering maximum principal

axes rotation (  = 0.87, ts = 310 s)

200

0
0 2 4 6 8 10 12 14
Element size [mm]

(b)
65
She ar band inclination [ ]

Coloumb’s inclination
60

 = 20

Roscoe’s inclination (  = 20 )
55

50 =0

45 Roscoe’s inclination (  = 0 )
0 2 4 6 8 10 12 14
Ele ment size [mm]

Figure 4.19. (a) Stand-up time and, (b), shear band inclination as a function of element size.

82
 Chapter 4 – Mesh dependency

(a)
800
c = 40 kPa,  = 35

ts [s]
700 Analytical solution without considering principal
axes rotation (  = 1, ts = 653 s)

600

500
FEM
400

300 Analytical solution considering maximum principal

axes rotation (  = 0.82, ts = 259 s)

200

100

0
0 2 4 6 8 10 12 14
Element size [mm]

(b)

1200
c = 60 kPa,  = 25
ts [s]

Analytical solution without considering principal

1000 axes rotation (  = 1, ts = 979 s)

800 FEM

600
Analytical solution considering maximum principal
axes rotation (  = 0.91, ts = 445 s)
400

200

0
0 2 4 6 8 10 12 14

Ele ment size [mm]

Figure 4.20. Stand-up time as a function of the mesh size for (a) lower and (b) higher
strength parameters.

83
 (a)

c = 40 kPa,  = 35
63
Shear band inclination [ ]
Coloumb’s inclination

61

59  = 20

57

55 Roscoe’s inclination ( = 20 )

53

51
=0
49

47

45 Roscoe’s inclination ( = 0 )
0 2 4 6 8 10 12 14
Ele ment size [mm]

(b)

c = 60 kPa,  = 25
Shear band inclination [ ]

63

61

59

Coloumb’s inclination
57  = 20

55 Roscoe’s inclination ( = 20 )

53

51

49
=0
47

45 Roscoe’s inclination ( = 0 )
0 2 4 6 8 10 12 14
Element size [mm]

Figure 4.21. Numerically computed inclination of the shear band for (a) lower and (b) higher
strength parameters.

84
 Chapter 4 – Mesh dependency

=0  = 20
45 55

100 mm

120 mm

Figure 4.22. Inclination of the shear band in a pure stress analysis (model geometry after
Fig. 4.6; c = 50 kPa; φ = 30º; discretisation by 10x5 elements; isotropic and
uniform initial stress of 100 kPa; shear band formation induced by imposing a
vertical displacement of 30 mm to the upper boundary).

mesh 20x10 mesh 400x200

 = 45 (170 mm, 70 mm)  = 56.5 (170 mm, 53 mm)

Plastic zone (shear band) Plastic zone (shear band)

ω = 14
ω=1
max principal stress
30 mm max principal stress 7 mm

(140 mm, 40 mm) t = 455 s (163 mm, 46 mm) t = 685 s

30 mm 7 mm

Figure 4.23. Shear band inclination θ and principal stress rotation ω in the centre of the
model obtained with the coupled hydraulic-mechanical analysis (model
geometry after Fig. 4.6; c = 50 kPa; φ = 30°,  = 0°).

85
4.3.3.4.2 Interpretation

The main cause of mesh-dependency of stand-up time is the structural softening. The amount
of structural softening increases with increasing shear deformation (Chapter 2). Due to
localization, the shear deformation increases with decreasing element size. The larger
deformation causes a decrease of the operational shear strength, and so a reduction of the
stand-up time.
Localization occurs also when using large elements, but in this case the shear deformation
and the principal axes rotation are so small (Fig. 4.23) that softening has almost no influence
on the stand-up time (η practically equal to 1).
In addition, coarse meshes result (as mentioned above) to less inclined shear bands. This, too,
suggests a higher stand-up time, because the wedge-like sliding body (Fig. 4.6) is more stable,
if the shear band is less inclined, and, therefore, excess pore pressure dissipation must
progress more in order to reach ultimate state. In addition, a less inclined shear band is longer
and needs therefore more time to reach the top boundary of the model – unless the shear band
propagation speed is higher in the case of the coarse mesh (which is not true as it will be
shown in the following)

4.3.3.4.3 The effect of the discretization on the propagation speed

The present section will show by comparing the results obtained with a fine (200x100) and a
coarse mesh (200x19) that a fine discretization results to a higher propagation speed because
it captures the occurrence of yielding earlier.
Figure 4.24 zooms to the bottom-right corner of the two numerical models. In order to avoid
boundary effects in the initialization phase of failure (Section 4.3.3.2.2), the same (fine)
discretization was chosen for both meshes within a 10 mm wide strip at the bottom boundary
of the system (hereafter referred to as the “boundary strip”, Fig. 4.24).
(a) (b)

B B

Element D
355 s
10 mm
Element D

310 s 310 s,
355 s
A A
10 mm
(boundary strip)

Figure 4.24. Extent of the shear band t = 310 s (solid line) and t = 355 s (dashed line), (a), for
a fine mesh (200x100) and, (b), for a coarse mesh (200x19).

86
 Chapter 4 – Mesh dependency

0.1

ytip,shear band [m]

Fine mesh

0.08 Coarse mesh

0.06

0.04

0.02

10 mm
(boundary strip)
0
0 100 200 300 400 500 600 700
t [s]

Figure 4.25. Evolution over time of the y-coordinate of the shear band tip.

Points A and B (Fig. 4.24), which are used for making comparisons, have exactly the same
location in both meshes. They are located close to the top of the boundary strip (point A) and
to the top of the first row of coarse elements (point B). The difference between the two
meshes is that the tip of the shear band must cross 10 (fine) elements in order to reach point B
starting from point A in the fine mesh, while only one (coarse) element-row between A and B
has to be crossed in the coarse mesh.
According to the numerical results (Fig. 4.25), the shear band reaches the top of the boundary
strip (point A) at t = 310 s in both models, which means that, in the early pore pressure
dissipation stage, the shear band propagation speed does not depend on the spatial
discretization of the upper part. Beyond the top of the boundary strip, however, the shear band
propagates much quicker through the fine mesh. 45 s later its tip is close to point B in the fine
mesh, but remains stuck in the vicinity of point A in the coarse mesh (see also Fig. 4.24) –
and this although the excess pore pressure dissipation rate is the same for both models (Fig.
4.26).
In search of the reason for the different propagation speeds, the stress paths at points A and B
over the time-interval (310 s, 355 s) will be considered next (Fig. 4.27): As point A is inside
the shear band already at the start of the time interval (in both meshes), the stress state at
point A moves downwards along the yield condition for the reason explained in Section
4.3.3.2.4 Point A experiences exactly the same stress change in the two meshes (the two
stress paths overlap).

87
 100

p w [kPa]
80
A

B
60

40

20
Fine mesh

Coarse mesh

0
0 100 200 300 400 500
t [s]

Figure 4.26. Time-development of the pore pressure at points A and B of Fig. 4.24.

87
 n [kPa]

355 s

86 B
355 s

85 310 s

84

83

A 310 s

82 Fine mesh
355 s
Coarse mesh

81
54 58 62 66 70
 'n [kPa]

Figure 4.27. Effective stress path at points A and B of Fig. 4.24 between 310 s and 355 s.

88
 Chapter 4 – Mesh dependency

Due to the excess pore pressure dissipation, the effective normal stress 'n decreases also at
point B. However, as B is within the elastic zone ahead of the tip of the shear band (in both
meshes), its shear stress n increases (in order to ensure equilibrium; cf. Section 4.3.3.2.4).
More specifically, the effective normal stress 'n decreases by exactly the same amount in
both meshes (due to the identical time-development of pore pressure; Fig. 4.27), but the shear
stress n increases considerably less in the case of the coarse mesh (compare dashed with solid
path in Fig. 4.27). This suggests (due to equilibrium) that the shear stresses underneath point
B must be higher in the coarse mesh than in the fine mesh. Therefore, next the stress changes
will be studied, which occur in the centre of the first element above the boundary strip
(element D in Fig. 4.24a and 4.24b) during the early stage of shear band propagation beyond
point A (from t = 305 s to t = 310 s). Figure 4.28 shows the stress path at the centre C of
element D of the coarse mesh as well as the stress path at the centre F of element D of the fine
mesh.
At t = 305 s the stress states of points C and F are still within the elastic domain (Fig. 4.28),
because the shear band has not reached these points yet. However, as point F is located closer
to the drained boundary, pore pressure increases earlier at F and, therefore, the effective stress
at point F is lower than at point C.
During the short time interval (305 s, 310 s) the shear stresses decrease slightly along the
shear band (due to structural softening) and increase slightly ahead of the shear band (in order
to ensure equilibrium; Section 4.3.3.2.4). As the stress state of point F at t = 305 s is close to
the yield condition, the small increase in shear stress brings point F to the yield surface, while
point C remains inside the elastic domain: at t = 310 s the shear band has reached point F, but
not point C.

84.5
 n [kPa]

Coarse mesh
310 s (Element D, Fig. 23b):

84.0 305 s 310 s

C C
305 s 10 mm
Fine mesh
(Element D,
Fig. 23a):

83.5
58.0 60.0 62.0 64.0
'n [kPa]

Figure 4.28. Effective stress path at the integration point in the centre of element D (see Fig.
4.24) between 305 s and 310 s.

89
In conclusion, the shear band propagates faster through the fine mesh because the fine
discretization allows to capture the stress changes induced by the pore pressure dissipation
earlier. The development of a longer shear band at the early stage of failure accelerates further
propagation: Remember that shear band propagation is associated with a gradual reduction of
the shear stresses inside the shear band and, in order to satisfy equilibrium, a gradual increase
of the shear stresses in the soil ahead of the tip of the shear band. As the shear band in the fine
mesh is longer than in the coarse mesh already at the early stage of failure, the shear stresses
decrease over a greater length and, consequently, the shear stresses ahead of the tip of the
shear band increase more than in the coarse mesh, thus leading to an earlier yielding, i.e. to a
faster failure propagation. One may say that in the considered problem the process of shear
band propagation is self-reinforcing.

4.3.3.4.4 Stand-up time estimation for very thin shear bands

The linear dependency of the stand-up time to on the mesh size observed for fine meshes in
(Sections 4.2.3 and 4.3.3.4.1) and the demonstration that the stand-up time tends to a non-zero
value for extremely small element sizes, suggest that a practicable way of estimating the
stand-up time is to perform few computations with relatively fine, but computationally
manageable meshes and extrapolate their results to the mesh size that corresponds to the
expected, grain-size-dependent thickness of the shear band (which is practically zero).

c' = 7.5 kPa ts = ∞

ts [h]

1'800 (vertical cut stable

φ' = 30
=0 under drained
1'550 conditions)

1'300

1'050

800

550

300
ts > 0
50

ts < 0 0 5 10 15 20 25 30
-200
mesh size [cm]

0% 2% 4% 6%
s / H [-]

Figure 4.29. Linear extrapolation of the stand-up time in the example of the underwater
vertical cut of Fig. 4.5c.

90
 Chapter 4 – Mesh dependency

As shown in Section 4.2.3 and as depicted exemplary by Figure 4.29, there are situations for
which the strength parameters are such that the stand-up time varies very sensitively with the
element size (from a finite value to an infinite one). For these cases the linear extrapolation
should be carried-out only between the stand-up times obtained with sufficiently small
elements (such that the relationship between the stand-up times and the mesh size is almost
linear, see dashed line in Fig. 4.29). Otherwise the inclination of the extrapolated line could
result to a negative extrapolated stand-up time, which would be obviously erroneous (see dot-
dashed line of Fig. 4.29).

4.4 Conclusions
The excess pore pressure dissipation causes a gradual reduction of the effective stress both
inside and ahead of the shear band. The decrease in the effective stress along the shear band
results to a decrease in its shear resistance, which is additionally assisted by structural
softening induced by the non-associativity of the flow rule. As the shear stresses decrease
inside the shear band, the shear stresses must increase ahead of the shear band in order to
satisfy equilibrium, thus causing failure propagation. This stress redistribution process
continues until the shear band has progressed through the entire system. Afterwards, the
system collapses.
Chapter 4 showed that the coarseness of the spatial discretization generally influences the
model behaviour quantitatively but not qualitatively. As explained in Chapter 2, the
assumption of non-associated plastic flow leads to mesh-dependency of the stand-up time: the
finer the FE mesh, the earlier the system will fail. An exception occurs when the strength
parameters are close to the critical values for drained stability. In this case the occurrence or
lack of instability during the swelling process depends on the size of the mesh.
The main reason for the mesh dependency of the stand-up time is that the amount of softening
increases with the shear deformation, and so with decreasing element size. Chapter 4
analytically showed that, the numerically estimated stand-up time is bounded by an upper and
a lower limit, which correspond to zero and maximum softening (as derived in Chapter 2),
respectively .
Finally, the numerical results showed that the stand-up time depends almost linearly to the
mesh size for relative fine meshes. This is a valuable result, as it suggests a practicable way of
estimating the stand-up time: It can be determined by performing a few computations with
relatively fine, but computationally manageable meshes and extrapolating their results to the
mesh size that corresponds to the expected, grain-size-dependent thickness of the shear band.

91
 Chapter 5 – Uniaxial loading tests

5. Uniaxial loading tests

5.1 Introduction
The goal of the present Chapter is to study the delayed failure induced by the transient
process of excess pore pressure dissipation by means of a simple experiment and asses the
capability of hydraulic-mechanical coupled FEM analyses to reproduce the observed
behaviour.
The test consists in loading small prismatic (50 mm x 50 mm x 100 mm) specimens of over-
consolidated clay by a constant uniform vertical pressure and leaving the specimens
consolidate or swell under constant boundary conditions until failure occurs. The tests are
executed either underwater or under water vapour at very high relative humidity (average HR
about 97%). The latter allows to show the influence of the hydraulic boundary condition on
the stand-up time.
The short-term stability of the specimen is due to the negative excess pore pressures that
prevail in the specimen at the moment at which the vertical load is applied. The negative
excess pore pressures are induced by quickly unloading the sample from the oedometer-cell at
the end of the preparation phase.
The specimen behaviour is monitored by recording the vertical displacement of the top of the
specimen and the pore pressure inside the specimen over time, and, by taking pictures at
regular time intervals.
Section 5.2 focuses on the experimental study. During testing shear bands develop starting
from the specimen corners and propagating over time towards the centre of the specimen. The
progression of the shear band is due to the dissipation of the negative excess pore pressure
that are generated during the pre-test phase and partially perpetuated during uniaxial testing
due to the plastic dilation in the shear bands. Failure occurs shortly after the full development
of the shear bands throughout the specimen. It manifests itself by an accelerating
displacement. The time lapsing from loading to collapse is defined as stand-up time; it
averagely amounts to 6 minutes in the underwater tests and is by a factor of about 4 higher in
the tests under water vapour.
Section 5.3 presents the computational assumptions and results of the performed three-
dimensional coupled FEM analyses. The MCC model reproduces qualitatively the observed
behaviour but it overestimates stand-up time. The MC model (with isochoric flow rule)
underestimates the stand-up time because it does not capture the dilatant behaviour of the
overconsolidated material.

93
5.2 Experimental study

5.2.1 Material and methods

5.2.1.1 Specimen preparation

Specimen preparation consists of 4 main phases: over-consolidation of the clay, extraction of

the consolidated material, cutting of the samples and, finally, sample storage (Fig. 5.1).
Birmensdorf Clay – a reconstituted natural high plasticity clay (Weber 2007) – was used in
the experiments. The clay was initially mixed with water (initial water content of the clay
mixture = 100%) under vacuum and then poured into a large oedometer cell with diameter
250 mm (filling height about 217 mm). In order to avoid sticking, the internal wall of the
oedometer cell was covered with a thin plastic sheet lubricated with paraffin oil.
Afterwards the clay was loaded to 100 kPa in 5 loading steps (4 kPa, 15 kPa, 25 kPa, 50 kPa
and 100 kPa), then unloaded (in one step) and, finally, left to consolidate at 50 kPa, thus
obtaining an uniform OCR of 2. The total consolidation time (loading up to 100 kPa and
successive unloading to 50 kPa) took about 30 days.
After the consolidation phase, the material was quickly unloaded and removed from the
oedometer cell (Fig. 5.1b). The undrained unloading generates a negative excess pore
pressure (Section 5.2.2.1).
The material was subsequently cut in 12 prismatic samples, 10 cm high with a base 5 cm x 5
cm (cutting scheme in Fig. 5.1c). In order to obtain a precise geometry and limit sample
disturbance, the cutting was executed using a thin cutting wire, which was vertically guided
by an aluminium frame (Fig. 5.1d). Sand was inserted in the open cuts during cutting of the
consolidated material in order to avoid the re-sticking of the clay (this also facilitated the
handling of the samples in the subsequent preparation phases).
For the period between sample preparation and uniaxial testing, the samples were sealed by a
cellophane film and closed in an aluminium formwork (Fig. 5.1e) in order to protect them and
to limit premature dissipation of the excess negative pore pressure. This sealing system
proved to be effective (Section 5.2.2.1).

5.2.1.2 Installation of pore pressure transducers

Pore pressure was measured during consolidation and uniaxial testing by means of micro pore
pressure transducers (ppts). The ppts are of type Keller (type 2MIX/2bar/81840.1) and consist
of a small sensor (diameter 7 mm, length 11.6 mm) and a data transfer cable (diameter 0.6
mm).
In order to measure pore pressure as reliably as possible, the consolidated material should be
disturbed as less as possible. For this reason, the ppts were installed in the clay right from the
start (i.e. in the initial water - clay mixture). Two installation systems were developed for this
purpose and are described below.

94
 Chapter 5 – Uniaxial loading tests

(a)

(b)

(c)

(d)

Legend:

1 Load piston

2 Load plate
(e)
3 Filter plate

4 T extile and paper filters

5 Oedometer cell

6 Clay

7 Drainage plate

8 Cutting device

9 Cutting wire

10 Cutting guide

11 Cellophane film

12 Formwork

Figure 5.1. Sample preparation (pre-test phase). (a) oedometer cell, (b) consolidated clay
cylinder, (c) cutting scheme, (d) cutting device, (e) sample storage.

95
The first system consists of removable aluminium supports installed in the bottom drainage
plate (Fig. 5.2a). The supports are composed by a small tube (diameter 6 mm) and a
supporting base. The tube can slide inside the base and can be fixed to the target height by
means of one horizontal screw. The base is fixed to the drainage plate by means of four
vertical screws that are removed at the end of the consolidation phase. The sealing between
base and drainage plate during consolidation is achieved by a sealing ring, while a rubber
plug (fixed at the bottom of the aluminium tube by a screwed cylinder) seals the tube. In order
to fix the ppt at the top of the cylinder two flat washers with a diameter of 7.5 mm were
inserted below the sensor. Figure 5.2b shows the 8 supports after extraction of the
consolidated material from the oedometer cell.

(a) (b)
10 cm

1.8 cm

Legend:

1 Ppt sensor

2 Aluminium tube

3 Support base
5 cm
4 Sealing ring

5 Ppt cable
Figure 5.2. Installation of ppts with removable aluminium support (system 1). (a) section of
a single system, (b) consolidated clay cylinder with installed aluminium
supports.

Figure 5.3. Detachment of the support from the sample during the pre-test phase.

96
 Chapter 5 – Uniaxial loading tests

The main advantage of the system is that the position of the ppts remains unchanged during
the consolidation. Moreover, the ppts experience compressive stresses during all phases of the
test while its cable is stress free. This is favourable in respect to the fragile connection
between cable and sensor. The disadvantages are the stiffness and the weight of the support
which causes a disturbance of the material (particularly in the sample preparation phase, i.e.
when the sample lies horizontally, Fig. 5.3). Moreover, the rigidity of the support influences
(locally) the stress distribution inside the sample during the test, as its stiffness is much higher
than that of the material (the loading of the sample may cause a punching of the material
around the sensor). These aspects will be discussed later in the interpretation of the
experimental results.
The second system refrains from the fixed supports; only the sensors and their cables remain
inside the material at the end of the consolidation phase. In this system, the ppts are fixed by
means of vertical auxiliary wires (Fig. 5.4a), which ensure that the ppts maintain their
position during the filling of the oedomenter cell with the clay - water mixture and avoid
tension between sensors and cables.

(a)
1

(b)
2
Legend:
3
1 Support plate
1.6 cm

2 Ppt cable
4
3 Auxiliary hanging wire
5
4 Ppt sensor
ca. 1 cm

5 Aluminium casing
6
6 Suture wire

0.9 cm
Figure 5.4. Installation of ppts by means of auxiliary hanging and suture wire (system 2).
(a) hanging structure and, (b), detail of ppt fixation.

97
Small diameter mono-braid wires (0.4 mm) composed by Dyneema fibers (BR8 by
CLIMAX) were used to hang the ppts to an upper support plate (Fig. 5.4b). The used wires
have a high axial stiffness and tensile strength. The ppts were kept fixed in their initial
position until the end of the consolidation step at 25 kPa. After this step the vertical hanging
wires were cut, thus leaving the sensors to move freely together with the consolidating soil.
The initial elevation of the ppts was chosen such that they reach the centre of the sample at
the end of consolidation.
The ppts were connected to the bottom plate by means of 0.48 mm diameter synthetic
absorbable surgical suture wires (type Vicryl Rapide by ETHICON) which dissolve over time
(after about 40 days and 90 days in water and clay, respectively). The advantage of the
selected wires is that they offer a high short-term resistance (thus ensuring a fixed position of
the ppts during filling the oedometer) but, as they dissolve, allow an easy separation of the
material from the bottom drainage plate after the consolidation.
A special loading/drainage plate was constructed which allowed:
i) the cables and the wires to pass through it;
ii) an almost friction-free movement in the vertical direction;
iii) to avoid the loss of fresh slurry during the early stages of the consolidation (sealed
system).
The plate was constructed with 12 grease chambers (one per ppt). Figure 5.5a shows the
detail of one chamber. After placing the cables and the wires, the chamber was filled with
high viscosity copper paste and closed by means of two half covers fixed to the plate with
vertical screws. Two small circular holes were bored in the cover with diameter slightly larger
than the auxiliary wire and ppt cable, respectively. The small tolerances let the plate move
freely in the vertical direction but avoid the grease to escape. The sealing of the system was
tested in advance up to a water pressure of 100 kPa.

(a) (b)
Load plate

Clay

Legend:

1 Special washer

2 Grease chamber

3 Filter plate

4 Filter

5 Ppt cable

6 Auxiliary hanging wire

Figure 5.5. Filter plate with grease chambers. (a) section of the chamber and, (b), top view
of a grease chamber.

98
 Chapter 5 – Uniaxial loading tests

(a)

0-13.5 cm
16.5 cm
5 6

43 cm
29 cm
8

16 cm

22.5 cm

(b)

Legend:

1 Laser

2 Weight

3 Steel cylinder

4 Cylindrical guide
10
5 Load plate

6 Water container

7 Sample

8 Plexiglas box

9 Cylinder support

10 Humidity sensor

Figure 5.6. Test device. (a) underwater and, (b), under water vapour.

99
5.2.1.3 Testing device and procedure

The test consisted of a rapid uniaxial loading by a uniform vertical pressure of 33.3 kPa
(applied within about 2 s), followed by a swelling stage under constant load and hydraulic
boundary conditions. The test ended after certain time with the failure of the specimen.
The specimen behaviour was monitored by measuring the vertical displacement of the load
cylinder and the pore pressure in the specimen centre (by laser and ppts, respectively). In
addition, pictures of the deformed specimen were taken at regular time intervals (about every
5 min).
Figure 5.6 shows the apparatus specifically developed for the tests. In order to control the
hydraulic boundary conditions, the tests were executed within a chamber, which was
composed by an aluminium frame and removable Plexiglas walls. The loading device was
outside the chamber; the load was transferred to the sample by means a loading plate that was
connected to a vertical steel cylinder. The latter was vertically guided by a tube which was
fixed to the upper plate of the chamber (Fig. 5.6). The tube was equipped with two miniature
metric ball bushing bearings (MM12WW by Thomson) at its extremities in order to eliminate
friction.
The tests were executed either underwater or under water vapour. In the first case the
specimen was put in a cylindrical transparent cylinder filled with water (Fig. 5.6a). In the
second case, water vapour (generated by an external ultrasonic vapour generator) was
introduced into the (closed) chamber via a plastic tube (Fig. 5.6b). The relative humidity of
the air in the chamber was monitored by a sensor and varied between 92% and 100% over
time (Fig. 5.7). The lower values occurred when the chamber was opened in order to take
pictures of the deformed specimen. The average HR amounts to 97.2%.
The top and bottom boundaries of the specimens were blocked with a 3 mm thick Teflon
frames during testing (Fig. 5.8). The mentioned mechanical boundary conditions were chosen
in the conception phase of the test because they have proven to lead to the highest
reproducibility of the results.

100
 Chapter 5 – Uniaxial loading tests

100%
HR [-]

98%
avg. HR = 97.2%

96%

94%

92%

90%
0 10 20 30 40
t [min]
Figure 5.7. Evolution over the test time of the relative air humidity HR in the test chamber.
10 cm
5 cm

0.3 cm
0.3 cm

Legend:

1 Plexiglas plate
5 cm
2 T eflon frame

3 Sample

4 Ppt sensor

5 Ppt cable

6 Auxiliary hanging wire

Figure 5.8. Sample fixation and ppt position (for system 2).

101
5.2.1.4 Material parameters

Table 5.1 gives the parameters of the Birmensdorf clay in terms of the Mohr-Coulomb (MC)
and Modified Cam Clay (MCC) models. With the exception of the effective cohesion of the
MC model, the values were obtained by means of laboratory tests (Table 5.2).
As later shown in Section 6.4.3.1, there is an inconsistency between the Young's modulus E
assumed for the MC analyses (value derived by Küng, 2003, and later used by Weber, 2007)
and the MCC parameters (derived by Küng, 2003, and Weber, 2007, respectively): for the
problem under consideration, a consistent Young's modulus would be considerably lower
than the one given by Küng (2003). The inconsistency makes a careful interpretation of the
numerical results necessary (as a lower Young's modulus leads to longer stand-up times).
The effective cohesion was estimated assuming that the soil behaves as MCC and considering
the theoretical effective stress path expected during testing. More specifically, the cohesion is
taken such that yielding in the MC model occurs at the same point (p', q) as for the MCC
model. The latter corresponds to the intersection point between the assumed effective stress
path and the yield surface according to the MCC model. The slope of the MC yield surface in
the p'-q plane is obtained from the friction angle measured in the laboratory at critical state.
Figure 5.9 shows exemplarily the expected effective stress path in the stress space p'-q (where
p' and q denote the mean effective and deviatoric stress, respectively). The vertical loading
causes an increase of the deviatoric stress from zero to the applied pressure of 33 kPa. As the
material is over-consolidated and the process is assumed to be undrained, the mean effective
stress remains constant during the loading (elastic behaviour). During subsequent swelling the
mean effective stress decreases under constant deviatoric stress. Yielding occurs when the
stress path hits the yield surface (point A in Fig. 5.9).

50
q [kPa]

Yield surface after

consolidation (MCC model)
40

A t = 0+
q’lim = 33 kPa
30

20

10

t =0
0
0 20 40 60 80
p' [kPa]
p’lim = 14 kPa
Figure 5.9. Theoretical stress path (in the stress space p’-q) expected during the uniaxial
tests.

102
 Chapter 5 – Uniaxial loading tests

Table 5.1. Material constants.

Saturated unit weight  [kN/m3] 18.2

Unit weight of water w [kN/m3] 10

Liquid limit wl [%] 58

Plastic limit wp [%] 19

Plasticity index Ip [%] 39

Void ratio at v = 1 kPa e0 [-] 2

Seepage flow parameters

Permeability k [m/s] 1.5 10-9
Water compressibility cw [MPa] 0

Parameters for linearly elastic, perfectly plastic MC material

Young’s modulus E [MPa] 18
Poisson’s ratio  [-] 0.3
Cohesion c' [kPa] 9 see Section 5.2.1.4
Friction angle ' [°] 24.5
Dilation angle ψ [-] 0

Parameters for Modified Cam Clay material

Gradient of swelling line [-] 0.01 (set 1, see Table 5.2)
Gradient of compression line [-] 0.21 (set 1, see Table 5.2)
Poisson’s ratio  [-] 0.3

103
Table 5.2. Laboratory tests executed for the determination of the material parameters 3.

Parameter Test Reference

φ CU Triaxial Effective friction angle at critical state, Küng (2003)
based upon 12 tests executed at p' = 100
kPa and pre-consolidation pressure 100
kPa, 200 kPa and 400 kPa
λ (set 1) Large diameter (250 Based upon 5 tests, first loading curve Küng (2003)
mm) oedometer tests
λ (set 2) Oedometer tests Based upon 10 tests executed on clay Weber (2007)
samples extracted from the material
after centrifuge test
κ (set 1) Large diameter (250 Based upon 5 tests, loading-reloading Küng (2003)
mm) oedometer tests curve
κ (set 2) Oedometer tests Based upon 10 tests executed on clay Weber (2007)
samples extracted from the material
after centrifuge test (executed by
Weber, 2007), loading-reloading curve
E Triaxial tests Calculated from unloading-reloading Küng (2003)
curve at a reference pressure of 100 kPa
and OCR = 2
k Large diameter (250 Pre-consoldiation pressure of 100 kPa Küng (2003)
mm) oedometer tests
ψ - Assumed equal to 0 Weber (2007)
ν - Assumed equal to 0 Weber (2007)

3
Remark: Two sets of material parameters λ and κ are considered in the present study. Parameter set 1 was
derived from soil samples, which were consolidated under 1g, while parameter set 2 was obtained from
samples of the same material, after consolidation in the centrifuge. The reason for the difference between
parameter sets 1 and 2 is not clear to the author and would require additional experimental investigations (as
concluded by Weber, 2007). In the present work, set 1 is used for the tests, which have been executed under
1g (uniaxial tests of Section 5), while – for sake of comparison – the numerical simulations of Section 6
(centrifuge tests) have been executed for both sets of parameters.

104
 Chapter 5 – Uniaxial loading tests

The shape of the yield surface according to the MCC Model is described by:

2
 p' 
q  a0 M 1    1 , (5.1)
 a0 

where M is the slope of the CSL and a0 is the centre of the yield surface.
For triaxial compression (which is a reasonable assumption for the problem under
consideration) the critical state line has the slope

6 sin  '
M  0.96 , (5.2)
3  sin  '

where φ' denotes the effective friction angle at critical state.

The centre of the yield surface, which determines its size, corresponds to the maximum
stresses
1  2 K0,NCL
p 'NCL   'v ,max (5.3)
3
and

qNCL   'v,max (1  K0, NCL ) (5.4)

experienced by the material, where σ'v,max is the maximum over-consolidation pressure (100
kPa), while K0,NCL denotes the coefficient of lateral pressure at the end of the consolidation.
As the material after loading at 100 kPa is a normally consolidated, K0,NCL can be determined
after Jacky (1948):

K0, NCL  1  sin  '  0.59 . (5.5)

Solving Eq. (5.1) with respect to a0 and considering Eqs. (5.2) to (5.5) results to

q 2NCL   M p 'NCL 
2

a0   49.0 kPa . (5.6)

2M 2 p 'NCL

The stress state at the intersection point A is given by qlim (= 33 kPa) and p'lim, whereby the
latter is obtained by solving Eq. (5.1) with respect to p':

2
q 
p 'lim  a0  a 02   lim   14.0 kPa . (5.7)
 M 

105
Finally, the effective cohesion c' at point A is determined by expressing the MC failure
surface in terms of principal stresses (p’, q) and applying the Caquot’s transformation to all
normal effective stresses (see Vrakas and Anagnostou, 2015), so that

q 
c '   u ,lim  p 'lim  tan  ' , (5.8)
 M 

which in combination with Eq. (5.2) leads to:

3qlim   6 p 'lim  qlim  sin  '

c'   9.2 kPa . (5.9)
6cos  '

The numerical analyses in Section 5.3.4 were carried-out (for simplicity) assuming a cohesion
of 9 kPa.

5.2.2 Results

5.2.2.1 Pore pressure evolution during the pre-test phase

Figure 5.10 shows the time-development of the pore pressure measured during the pre-test
phase using, (a), the ppt-support system 1 (later used for the tests executed under water
vapour) and, (b), the ppt-support system 2 (later used for the underwater tests). The
preparation took about 34 days for system 1 and 123 days for system 2. The longer period for
system 2 was due to the time needed in order that the suture wires dissolve. Due to the
fragility of the ppts and to the long duration of the pre-test phase, most transducers stopped
measuring during the consolidation phase. The problem was maybe additionally caused by
the dissolution of the suture wires which resulted to the formation of a fine layer of crystals
on the sensor surface; this was observed on some sensors after disassembling the specimens at
the end of the uniaxial loading test.
The pre-test phase consists, (i), in the consolidation of the material (0 – 100 kPa – 50 kPa),
(ii), the extraction of the material from the oedomenter cell and the preparation of the samples
and, (iii), the storage of the samples in the aluminium framework.
The sudden increases and decreases in the pore pressure observed in Figure 5.10 (phase i)
correspond to the increase of the pre-consolidation pressure and to the subsequent unloading,
respectively. The cumulative increase in pore pressure (sum of all positive increments)
measured by the ppts amounts to approximately 90 kPa and 100 kPa for ppt-support system 1
(Fig. 5.10a) and 2 (Fig. 5.10b), respectively, while the decrease caused by the unloading from
100 kPa to 50 kPa amounts to -30 kPa and -40 kPa, respectively.

106
 Chapter 5 – Uniaxial loading tests

(a)
50

p w [kPa]
40

30

20

10
t [days]
0
0 5 10 15 20 25 30 35 40
-10 -11 kPa
(pre-test pore pressure)
-20

-30

-40
Ppt support system 1 (used for tests under water vapour)
-50

i. ii. iii.
(b)
50
p w [kPa]

40

30

20

10
t [days]
0
0 20 40 60 80 100 120 140
-10

-20

-30 -32 kPa

(pre-test pore pressure)
-40
Ppt support system 2 (used for tests underwater)
-50

i. ii., iii.

Pre-test phase:
i. Consolidation 0 – 100 kPa – 50 kPa
ii. Specimen preparation
iii. Storage
Figure 5.10. Pore pressure evolution measured in the pre-test phase using (a) aluminium
tubes and (b) hanging wires as support for the ppts.

107
The unloading and the extraction of the material caused a decrease of the pore pressure by 46
kPa and 37 kPa for system 1 and 2, respectively. Subsequently, during the specimen
preparation time, the pore pressure increased to -9 kPa and -28 kPa for system 1 and 2,
respectively. The further change of pore pressure during the storage time is in the order of
few kPa, which indicates that the sealing of the samples by means of the cellophane film was
proper.
Finally, the pore pressure measured in the middle of the specimen just before the uniaxial
loading was equal to -11 kPa and -32 kPa for system 1 and system 2, respectively.
These values are lower than expected theoretically: Assuming undrained behaviour during
quick loading or unloading in the oedometer cell, the change in pore pressure should
correspond to the change in the applied pressure. Consequently, the sum of the sudden
increases of pressure occurred during the pre-consolidation phase should amount to the over-
consolidation pressure (100 kPa), while the sudden decrease of pore pressure when unloading
the material should amount to 50 kPa. This means that the expected excess pore pressure at
the end of the pre-test phase should be equal to -50 kPa.
According to Figure 5.10 the measured excess pore pressure are lower than expected
particularly at unloading. Possible reasons for the observed difference are: erroneous
measurement by the ppt; local disturbance of the pore pressure field close to the sensor (for
reasons explained in Section 5.2.1.2); partial dissipation of the excess pore pressure during
the preparation of the specimens (or due to loading during handling of the samples);
differences between applied pressure and imposed pressure. (The latter is rather improbable
because the oedometer cell was tested in advance with water by measuring the change of
water pressure that occurred when changing the applied oedometer pressure.) The influence
of the lower than expected initial excess pore pressure on the behaviour during testing will be
discussed in Section 5.3 based upon numerical calculations.

5.2.2.2 Underwater tests

Figures 5.11a and 5.11b show the time-development of the vertical displacement at the top of
the specimen and of the pore pressure in its centre, respectively. Figure 5.11c shows images
of the specimen at several times (t = 0 is the time immediately before loading).
The vertical loading of the specimen, which occurs within about 2 s, causes an instantaneous
settlement by 0.5 mm, followed by a time-dependent settlement. The displacement rate is
almost constant over the first 3.5 minutes (0.5 mm/min) and increases continuously
afterwards (Fig. 5.11a). The displacement starts to accelerate when shear bands develop at the
corners of the specimen (Fig. 5.11c). The shear bands are about 0.1 mm thick and extend over
the entire width of the specimen. On the specimen side perpendicular to the plane shown in
Fig. 5.11c, no shear bands develop.
The shear bands progress over time toward the specimen centre. The displacement rate
increases with increasing shear bands length (compare Fig. 5.11a and c). At t = 5.7 min the
rate increases rapidly and the specimen fails. This happens shortly after the shear bands
completely develop through the model.

108
 Chapter 5 – Uniaxial loading tests

(a) (c)
10
t = 4 min

uz [mm]
5.7 min

uz

6 5.5 min

5 min
4

4 min
0.5 mm/min
2
t = 5 min
0.4 min
2s
0
0 1 2 3 4 5 6
t [min]
(b)
0
0 1 2 3 4 5 6
t [min]
-5

-10
5 cm

-15
t = 5.5 min

-20

-25
p w [kPa]

-30

-35

t = 5.7 min

Figure 5.11. Typical results of underwater tests: Time-development of, (a), vertical
displacement of the top specimen boundary and, (b) of the pore pressure in the
centre of the specimen (ppt installed with auxiliary hanging and suture wire);
(c) lateral view of the deformed sample at several times (the black line in the
Figure is the ppt cable).

109
 (a) (c)
14
t = 20 min

uz [mm]
uz
12

10
35 min

30 min
6
25 min
20 min
4
0.17 mm/min

2 t = 25 min

2s
0
0 10 20 30 40
t [min]
(b)
15
p w [kPa]

10

5 3 cm

t = 30 min
0
0 10 20 30 40
t [min]

-5

-10

-15

t = 35 min

Figure 5.12. Typical results of tests with water vapour: Time-development of, (a), vertical
displacement of the top specimen boundary and, (b) of the pore pressure in the
centre of the specimen (ppt installed with removable aluminium support); (c)
lateral view of the deformed sample at several times.

110
 Chapter 5 – Uniaxial loading tests

Figure 5.11b shows that the pore pressure in the centre of the model at t = 0 amounts to -32
kPa. As explained in Section 5.2.2.1, this value is higher than the pore pressure expected
theoretically after unloading in the pre-load phase (i.e. -50 kPa). The higher value indicates
that the excess negative pore pressure partially dissipated during the time lapsing from the
unloading and the test execution.
Due to the loading of the specimen by 33 kPa, pore pressure increases to -21 kPa (Fig. 5.11b,
t = 2 s). The loading-induced excess pore pressure of 11 kPa agrees with the value expected
for fully undrained loading (pw = 33 kPa/3=11 kPa).
As the test is executed underwater (i.e. the boundary pore pressure is atmospheric), one might
expect that the pore pressure would increase continuously to zero. The observed time-
development of the pore pressure (in the centre of the specimen) is nevertheless complex:
Pore pressure starts to increase, reaches a maximum at t = 0.4 min, then decreases with almost
constant rate (-2.1 kPa/min) for about 5 min and shortly before failure starts again to increase.
The decrease in pore pressure after t = 0.4 min is probably due to the partially constrained
plastic dilation in the middle of the specimen. This hypothesis is supported by the
photographs (Fig. 5.11c), which show a slight horizontal expansion of the specimen, and will
be validated by means of numerical analyses in Section 5.3. Also the zero rate of pore
pressure observed briefly before to failure seems to be related with the plastic shearing: when
the material at the centre of the specimen reaches critical state, the shearing continues at
constant plastic volume so that no further negative excess of pore pressure is generated.

5.2.2.3 Tests executed under water vapour

Figure 5.12 presents, similarly to Figure 5.11, the typical behaviour observed in the tests that
were executed under water vapour.
According to Figure 5.12b, the measured initial negative excess pore pressure (11 kPa) is
considerably lower than theoretically expected (50 kPa). In addition, the axial loading causes
an increase in pore pressure by 22 kPa, which is by factor 2 higher than expected (11 kPa).
This may be due to the ppt-support. In the tests under water vapour, the ppts are supported by
stiff aluminium tubes, which probably influence the pore pressure distribution around the
sensor, thus making the interpretation of the measured pore pressure evolution impossible.
The instantaneous loading-induced settlement (Fig. 5.12a) is about equal to the one observed
in the underwater test (Fig. 5.11a). This agrees with the expected behaviour under undrained
conditions (the hydraulic boundary conditions are meaningless for the undrained response).
The time-development of the displacement is qualitatively the same as in the underwater test,
but much slower. The specimen fails 38 minutes after loading; the stand-up time is by factor 6
to 7 higher than in the test discussed before.
This was observed in all performed tests (Fig. 5.13). Although the results are characterized by
a variability (which is probably due to small disturbances during specimen preparation),
failure occurred systematically earlier in the underwater tests than in the tests that were
performed under water vapour. This suggests that the effect of the hydraulic boundary
condition is significant although the relative humidity of the air was close to 100%. This
hypothesis is supported also by the numerical analyses of Section 5.3.

111
 12

uz - uz0 [mm]
10

uz

2
water vapour
underwater
0 +
0 10 20 30 40
t [min]

Figure 5.13. Evolution over time of the displacement for all executed tests.

(a) (b) (c)

σv σv = 33 kPa

E y D A D A
H A
D Load plate q=0 10 mm
x 10 mm
z Soil

pw = 0
(underwater) pw = 0
q=0 q=0 or or
100 mm p w = -3 kPa p w = -3 kPa 100 mm
(water vapour)

G pw = 0 q=0
B
C B C B

b C
50 mm
b

Figure 5.14. (a) Geometry of the 3D numerical model; (b) boundary conditions and load
applied during the consolidation phase; (c) boundary conditions and load
applied during the test.

112
 Chapter 5 – Uniaxial loading tests

5.3 Numerical study

5.3.1 Computational model

The performed tests were simulated by means of three-dimensional analyses. The model was
discretized by means of 8-node cubic elements with side length 5 mm and trilinear
displacement and pore pressure shape functions (Figure 5.14a).
Figures 5.14b and 5.14c show the boundary conditions and the load applied to the model
during the simulation of the consolidation phase in the oedometer and during the uniaxial
loading test, respectively. Considering the testing setup, the nodes of the bottom boundary of
the model were fixed, while a very high stiffness, simulating the load plate, was assigned to
the two uppermost element rows. This modelling of the load plate allows the transfer of the
vertical load σv (acting on the upper boundary of the numerical model) to the soil, ensuring at
the same time a uniform vertical settlement of the upper boundary of the specimen.
The clay is modelled either as a saturated porous medium according to the Modified Cam
Clay (MCC) model or as an isotropic, linearly elastic and perfectly plastic material obeying
the Mohr-Coulomb yield criterion with isochoric plastic flow (MC model). The seepage flow
follows the Darcy’s law.
The material constants were determined by laboratory tests with the exception of the cohesion
of the MC model (Section 5.2.1.4, Table 5.1). Due to the small size of the specimen, the
weight of the soil is not considered in the analyses.

5.3.2 Computational steps

5.3.2.1 MCC model

In the case of the MCC model one must take into account the stress path and the volumetric
changes occurring in the oedometer. This is done in 4 simulation steps:
i) Liquid clay in the oedometer: Initialization of the numerical model considering a
uniform and hydrostatic stress of 1 kPa and atmospheric pore pressure;
ii) Consolidation to 100 kPa: Stepwise drained loading of the numerical model up to a
vertical pressure σv of 100 kPa;
iii) Unloading and swelling to 50 kPa: Stepwise drained unloading to 50 kPa.
iv) Extraction of the consolidated clay from the oedometer, cutting of the specimen and
storage of the samples: Deactivation of the rollers along the lateral boundaries and
complete unloading of the model in one time step of 2 hours (transient analysis with
no-flow condition at the model boundaries).

113
The simulation of the actual test phase (uniaxial loading) consists of 2 steps:
v) Loading of the sample: Increase in the vertical load from zero to 33 kPa over a time
period of 2 s (transient analysis).
vi) Swelling under constant load: transient analysis after instantaneous change of the
hydraulic boundary conditions at the lateral surface of the model (Fig. 5.14c).

5.3.2.2 MC model

The initial state considered in the simulations with the MC model are those prevailing in the
specimen just before uniaxial loading. Consequently, the initial total stresses are taken equal
to zero, while the initial pore pressure is considered as uniform and equal to -50 kPa
(theoretical suction induced by the undrained unloading after quick removal of the material
from the oedometer).
The uniaxial loading is simulated as for the MCC model.

5.3.3 MCC model results

5.3.3.1 Simulation results for the underwater test

Atmospheric pore pressure is imposed to the lateral boundaries in step (vi) (swelling phase
after uniaxial loading).
In order to have a fixed slope of the CSL, and so to easier interpret the results, the
computations assume a circular yield function and plastic potential in the deviatoric plane,
taking the slope of the critical state line M equal to 0.96 (Eq. 5.2), which assumes a Lode's
angle of -30° (triaxial compression). This simplifying assumption is reasonable because the
adoption of a more realistic shape of the yield surface in the deviatoric plane, would have
almost no influence on the numerical results for the considered boundary conditions.
Figure 5.15a shows the effective (solid line) and total (dashed line) stress paths in the centre
of the specimen in the plane p'-q and p-q, respectively. Figure 5.15b and 5.15c show the time-
development of the pore pressure in the centre of the specimen and of the vertical settlements
at the top boundary of the model, respectively.
The initial deviatoric stress q is equal to zero and the initial mean stress is equal to 1 kPa
(point i). The consolidation at 100 kPa causes hardening of the material and expansion of the
yield surface to its maximum size (yield surface passing through point ii characterized by p’ =
85 kPa, q = 22.3 kPa and K0 = 0.76). As the process is drained, the effective mean stress at
this stage is equal to the total mean stress. The subsequent drained unloading to 50 kPa is
elastic. At the end of the unloading (point iii), the over-consolidation ratio (defined by the
ratio between the maximum vertical effective stress experienced in the past and the current
vertical effective stress) amounts to 2.0. It is interesting, that the deviatoric stress decreases
considerably during unloading (from 22.3 kPa to 2 kPa; state ii and iii, respectively), while
(as expected on account of the over-consolidation) the coefficient of lateral pressure increases
to about 1.

114
 Chapter 5 – Uniaxial loading tests

(a)
50
(i) Initial conditions

q [kPa]
Shape of the
(ii) Drained loading at q v = 100 kPa
yield surface after
(iii) Drained unloading at q v = 50 kPa
step (ii)
40 (iv) Undrained complete unloading
23 min (v) Undrained loading with q v = 33 kPa
4 min (v) 2 s

30

(ii)

20 p'

10

(iii)

(i) (iv) t = 0
0
0 20 40 60 80 100
p', p [kPa]
(b)
0
0 5 10 15 20 25 30
t [min]
-10

-20

-30

-40 2s

-50
p w [kPa]

t =0

-60

(c)
7
uz [mm]

6 uz

2s
0
0 5 10 15 20 25 30
t [min]
-1
Figure 5.15. Results obtained for an underwater test with MCC model (a) effective (solid
line) and total (dashed line) stress path of a point located in the centre of the
model in p' – q and p – q diagram, (b) evolution over the test time of the pore
pressure in the centre of the specimen and (c) of the vertical settlements at the
top of the model.

115
The last step of the sample preparation is simulated by a complete unloading of the model
under undrained conditions (from state iii to state iv). During the undrained unloading the
effective mean stress remains constant, while the total mean stress and the deviatoric stress
decrease to zero (Fig. 5.15a). The unloading-induced suction is equal to -53 kPa (Fig. 5.15b).
This value is higher than the measured value (about -30 kPa) – the effect of this difference on
the model behaviour will be discussed later – and slightly lower than the theoretical expected
value of -50 kPa (the small difference may be due to the non-uniform stress state caused by
the mechanical boundary conditions, Fig. 5.14c).
The practically instantaneous loading of the specimen by 33 kPa (from state iv at t = 0 to state
v at t = 2 s) causes an increase in the deviatoric stress from zero to the applied pressure. As
the loading is undrained the mean effective stress remains equal to zero, while the pore
pressure and the mean total stress increase by 11 kPa (i.e., the suction decreases from 53 to 42
kPa). The settlement occurring at this stage is equal to 0.2 mm (Fig. 5.15c), i.e. about half as
measured experimentally.
During the swelling process (t > 2 s) the excess negative pore pressure dissipates and,
consequently, the effective stress state moves leftward toward the drained equilibrium point p
≈ qv/3 = 11 kPa, while the deviatoric stress remains almost constant (also in this case, the
small change in the deviatoric stress is probably due to the non-uniform stress state caused by
the mechanical boundary conditions). During this period the specimen slight expands in the
vertical direction (swelling due to the decrease in the mean effective stress). This heave was
not observed in the laboratory tests; it may be related to the inability of the used constitutive
model to capture the higher stiffness of the material at small strain (Jardine et al., 1986).
At t = 4 min the effective stress path reaches the yield surface. The plastic dilation
accompanying yielding results to decreasing pore pressure in the centre of the model (Fig.
5.15b), and consequently an increase of the mean effective stress (Fig. 5.15a), so that the
effective stress state moves back towards the critical state line. During this process the
deviatoric stress slightly increases.
Figure 5.16a shows the time-development of the plastic zone in the vertical symmetry plane
of the model. Yielding starts to occur at the model corners already at t = 1 min. After 4
minutes, a plastic zone develops in the middle of the model. The two plastic zones merge
after about 20 minutes. Figure 5.17 compares the deformed mesh 23 minutes after loading
with the undeformed one and shows that the model diameter increases over time (as observed
experimentally). Differently to the experimental results, no localization of the deformation
and of the plastic zone occurs in the numerical model. It should be noted that localization
does not occur even when imposing a lower strength to the corner elements.
After 23 min, the effective stress path reaches the critical state line, which means that further
plastic shearing would occur under constant plastic volumetric strain. However, as further
shearing does not produce a stabilizing decrease in pore pressure, the system reaches the
ultimate state. This is evident from the rapid displacement acceleration at t = 23 min (Fig.
5.15c).

116
 Chapter 5 – Uniaxial loading tests

(a) Underwater test

t = 1 min 5 min 20 min 30 min

(b) Test under water vapour

Plastic zone

Figure 5.16. Plastic zone on the mid plane of the sample at several times for test executed (a)
underwater and (b) under water vapour (MCC model).

t =0

t = 23 min

Figure 5.17. Initial shape of the sample deformed model shape after 23 min.

117
The time-development of the vertical displacement and of the pore pressure qualitatively
agrees with the experimental observations (cf. Fig. 5.15 with Fig. 5.11). This is true
particularly concerning the complex evolution of the suction over time (decreasing initially,
then increasing and finally, just before failure, decreasing again). However, the numerical
analysis predicts a much slower process than observed; the predicted stand-up time (23
minutes) is by factor 3.5 longer than the average of all 8 tests (6.5 minutes).
As mentioned above the measured initial suction (32 kPa) is lower than the theoretically
expected suction (50 kPa) and this indicates that some excess pore pressure dissipation might
have occurred before the uniaxial loading. So, the question arises whether the difference
between computed and measured stand-up time might be due to the different initial suction
(the lower the initial suction, the closer to failure is the stress state inside the specimen at the
beginning of the uniaxial loading phase). The following paragraphs investigate the influence
of the initial pore pressure inside the specimen numerically (taking the same parameter values
as in the numerical example above). The only difference to the previous computation is that
an additional simulation step (iv') is introduced between complete unloading (simulation step
iv) and undrained uniaxial loading (simulation step v). The additional step simulates the
partial suction dissipation by means of a transient analysis in which the pore pressure at the
model boundary is set to zero. Step (iv') ends when the pore pressure in the centre of the
specimen has reached -30 kPa (i.e., about to the experimental value measured in the centre of
the specimen just before uniaxial loading). The time required in order to let the suction
dissipate to -30 kPa amounts to about 2 minutes. The assumption of atmospheric pressure at
the model surface means that the suction at the model boundaries has completely dissipated.
Taking into account the low permeability of the material and the samples storage procedure,
the assumption of a complete dissipation of the pore pressure at the model boundary is not
likely to happen. However, this represents a conservative assumption for the purpose of the
investigation because it leads to the lowest average suction in the model, and so to the lowest
shear strength of the material.
Figure 5.18 compares the numerical results presented above without partial dissipation of the
initial excess pore pressure before uniaxial loading (dashed lines) with those obtained
considering a partial dissipation (solid lines). The latter results to earlier yielding (2 vs. 4 min)
and to a slightly shorter stand-up time (19 vs. 23 min; Fig. 5.18a). Consequently, a partial
dissipation of the initial pore pressure does not explain the observed difference between
predicted and observed behaviour.
It has been shown that, for the problem under investigation, the MCC model overestimates
the stand-up time even when considering a partial dissipation of the initial negative excess
pore pressure. The main reason for the overestimation must be related to the shape of the
yield surface of the MCC model; it is well known (Potts and Zdravković, 2001a) that the
MCC model overestimates yield strength on the dry side (the actual yield surface is less
expanded). This means that the stress path between undrained unloading and yield surface as
well as back from the yield surface to the CSL is longer, and so that the plastic dilative
shearing is overestimated. This causes a longer the stand-up time. In addition, the obtained
numerical results differ from the experimental observations in that they do not reproduce the
observed localization of the deformation. The absence of localization may be an additional
reason for the numerical overestimation of the stand-up time: as explained in Section 2, the

118
 Chapter 5 – Uniaxial loading tests

localization of deformation along a band would have caused a local increase of shear
deformation and so of softening.

(a)
50
(i) Initial conditions
q [kPa]

Shape of the
(ii) Drained loading at q v = 100 kPa
yield surface after
(iii) Drained unloading at q v = 50 kPa
step (ii)
(iv) Undrained complete unloading
40
(iv’) Partial dissipation of the negative excess
(v) 19 min pore pressure to -30 kPa
2s
(v) Undrained loading with q v = 33 kPa
3 min
30
underwater (with initial
(ii) partial pore pressure
dissipation)
20 underwater (with underwater (without
initial partial pore initial partial pore underwater (without initial
pressure dissipation) pressure dissipation) partial pore pressure
dissipation)

10

(iii)
(iv’)
(i) t =0 (iv)
0
0 20 40 60 80 100
p' [kPa]
(b)
0
0 5 10 15 20 25 30
t [min]
-10

-20 2s

-30 t =0

-40

-50
p w [kPa]

-60

(c)
7
uz [mm]

6 uz

1 2s

0
0 5 10 15 20 25 30
t [min]
-1
Figure 5.18. Results obtained considering or not an initial partial dissipation of the pore
pressure (solid and dashed lines, respectively) for an underwater test with MCC
model (a) effective stress path of a point located in the centre of the model in p' –
q diagram, (b) evolution over the test time of the pore pressure in the centre of
the specimen and (c) of the vertical settlements at the top of the model.

119
5.3.3.2 Simulation results for the test under water vapour

If the relative humidity at the specimen surface is higher than that of the air, then water
evaporates from the specimen and its saturation degree decreases. An initially saturated
specimen becomes partially saturated if the relative air humidity is lower than 100%. The
desaturation induces capillary suction. The relationship between the effective saturation Se
(which, in case of zero residual moisture content, corresponds to the degree of saturation) and
the suction  is represented by the so-called water retention curve (van Genuchten, 1980):

1
Se    , (5.10)
1  a  
1
n 1 n

where α and n are material constants.

As the permeability of the Birmensdorf clay is very low (1.5x10-9 m/s) and the present tests
take maximum 40 minutes, it can be assumed that the desaturation occurs only at the surface
of the specimens. Consequently, it is reasonable to consider the specimens as fully saturated
and to take into account the superficial desaturation by prescribing a negative pore pressure as
boundary condition. (A longer testing time, a higher permeability or a lower relative air
humidity would necessitate the consideration of a three-phase medium.) This assumption was
made also by Kovacevic et al. (2007), which investigated the effect of a surface suction
(induced by partial desaturation) – on the delayed failure and showed that stand-up time
increases considerably when imposing a surface suction of 25 kPa at the ground surface (the
surface suction corresponded to atmospheric conditions and was estimated by back-analysing
slope failures at Prospect Park).
The equivalent boundary pore pressure can be estimated making the simplifying assumption
that the relative air humidity is equal to the degree of saturation of the material. This allows to
estimate the surface suction based upon the water retention curve. Figure 5.19a shows the
experimentally determined water retention curve of the Birmensdorf clay (details are given in
Appendix B). The dots correspond to the experimental data, while the solid line is after Eq.
(5.10). The dots with Sr > 100% lack of course physical meaning and are due to measurement
inaccuracies (see Appendix B).
Figure 5.19b zooms to the range Sr = 0.9 – 1. Under the simplifying assumption HR ≈ Sr and
taking into account the relative humidity of the air that was measured during testing (Fig.
5.7), the suction varies in the range 0 – 13.5 kPa. For the average HR of 97.2%, the suction
amounts to about 3 kPa.
As it will be shown below, the assumption of a suction of 3 kPa is able to explain the
difference in stand-up time between the underwater tests and the tests executed under water
vapour.
The computation is carried-out as in the last section (underwater test without partial
dissipation of the initial excess pressure), the only difference being that a suction of 3 kPa
(rather than atmospheric pressure) is prescribed at the lateral model boundaries in simulation
step (vi).

120
 Chapter 5 – Uniaxial loading tests

(a)
1000

Suction [kPa]
100

10

Sample 1
Sample 2
van Genuchten’s model (α = 0.02 and n = 3)
1
0% 20% 40% 60% 80% 100%
Sr [-]

(b)

14
Suction [kPa]

12

10

0
90% 92% 94% 96% 98% 100% 102%
Sr (≈HR) [-]
avg. HR = 97.2%

range of HR measured during the

execution of the tests under water vapour
(see Fig. 5.7)
Figure 5.19. Water retention curve estimated for Birmensdorf clay represented over the
range (a) 0 < Sr < 0.1 and (b) 0.9 < Sr < 0.1.

121
 (a)
50
(i) Initial conditions

q [kPa]
Shape of the
(ii) Drained loading at q v = 100 kPa
yield surface after
(iii) Drained unloading at q v = 50 kPa
step (ii)
40 (iv) Undrained complete unloading
(iv’) Partial dissipation of the negative excess
53 min pore pressure to -30 kPa
6 min (v) 2 s
(v) Undrained loading with q v = 33 kPa

30
water vapour
(ii)
underwater (without initial
parital pore pressure
20
dissipation)

10

(iii)

(i) (iv) t = 0
0
0 20 40 60 80 100
p' [kPa]
(b)
0
0 10 20 30 40 50 60
t [min]
-10

-20

-30

-40 2s

-50
p w [kPa]

t =0

-60

(c)
7
uz [mm]

6 uz

2s
0
0 10 20 30 40 50 60
t [min]
-1
Figure 5.20. Comparison between underwater and under water vapour condition (dashed
and solid lines, respectively) with MCC model (a) effective stress path of a point
located in the centre of the model in p' – q diagram, (b) evolution over the test
time of the pore pressure in the centre of the specimen and (c) of the vertical
settlements at the top of the model.

122
 Chapter 5 – Uniaxial loading tests

Figure 5.20 compares the numerical results obtained assuming a boundary suction of 3 kPa
(solid lines) with those of the previous section (dashed lines).
The effective stress paths are qualitatively similar, the only difference being that when
considering the boundary suction of 3 kPa the deviatoric stress increases slightly less before
reaching the yield surface (Fig. 5.20a). Yielding in the centre of the specimen starts therefore
later (at t = 6 instead of 4 min) and at a lower mean effective stress (p' = 16 instead of 21
kPa). The earlier yielding is evident also from the plastic zones in Figure 5.16; the final shape
of the plastic zone, however, is similar for the two considered cases.
For the same reasons as in the simulation of the underwater test (last Section), failure occurs
when the stress state in the specimen centre reaches the CSL. This happens, however, much
later than in the simulation of the underwater test (t = 53 min since beginning of pore pressure
dissipation; Fig. 5.19a and 5.19c).
The hydraulic boundary condition does not influence the excess pore pressure dissipation in
the centre of the specimen until yielding (t < 4 min; Fig. 5.20b). After yielding (which as
mentioned above starts 2 min later) excess pore pressure dissipates considerably slower if
suction is prescribed at the model boundaries (Fig. 5.20b).
According to Figure 5.21 (which shows the pore pressure distribution along a horizontal axis
in the middle of the specimen), the pore pressure gradient is always lower in the case of a
boundary suction of 3 kPa (due to a smaller difference between the imposed pore pressure at
the model surface and the suction in the centre of the model), which means that the Darcy
velocity is lower and the excess pore pressure dissipation proceeds at lower rate.
x [mm] x [mm]
0
0 5 10 15 20 25 0 5 10 15 20 25

30 10, 20

-10 5
5
10 40

15

50
-20 20
t [min] = 1

-30
x
1

-40
2s 2s
p w [kPa]

-50
(a) (b)

Figure 5.21. Pore pressure distribution along the horizontal axis in middle of the specimen at
several time points for test executed (a) underwater and (b) under water vapour
(MCC model).

123
 (a)
50

q [kPa]
40
1.4 min
2s

30

20 p'

10

t =0
0
0 20 40 60 80 100
p', p [kPa]
(b)
0
0.0 0.5 1.0 1.5 2.0
t [min]
-10

-20

-30

-40 0+

-50 t =0
p w [kPa]

-60

(c)
0.5
uz [mm]

uz

0.4

0.3

0.2

0+
0.1

0.0
0.0 0.5 1.0 1.5 2.0
t [min]

Figure 5.22. Results obtained for an underwater test with MC model (a) Effective (solid line)
and total (dashed line) stress path of a point located in the centre of the model in
p' – q and p – q diagram, (b) evolution over the test time of the pore pressure in
the centre of the specimen and (c) of the vertical settlements at the top of the
model.

124
 Chapter 5 – Uniaxial loading tests

The computational results provide an explanation for the different stand-up time in the tests
executed underwater and in the tests under water vapour. The difference is due on one hand to
the lower pore pressure gradient and on the other hand to the stabilizing effect of the suction.
The latter increases the effective stresses to -3 kPa (about 10% of the applied load), which,
although small at a first glance, is able to extend stand-up time considerably as explained
below (based upon Mohr - Coulomb yield condition).
The specimen would remain stable under drained conditions (i.e. the stand-up time would be
infinite), if the uniaxial compressive strength is higher than the applied load (qv = 33 kPa).
According to Mohr-Coulomb yield condition, the critical cohesion (i.e., the one required for
drained stability) reads as follows:

qv 1  sin  
cD ,lim   10.6 kPa , (5.11)
2cos 

which is only slightly higher than the actual cohesion (estimated to 9 kPa, see Section
5.2.1.4). The close vicinity of c to cD,lim means that small variations in c would result to big
variations in stand-up time. In the presence of a boundary suction of 3 kPa, the steady-state
pore pressure inside the specimen would be -3 kPa as well. This suction would apparently
increase cohesion by Δc = tanφ 3 kPa = 1.4 kPa to c = 10.4 kPa,which is much closer to the
critical cohesion of 10.6 kPa (the distance to the critical cohesion would decrease from 1.6
kPa to 0.2 kPa), so that the stand-up time considerably increases.

5.3.4 Underwater test interpretation by MC model

Figure 5.22 presents computational results analogously to the previous Section (Fig. 5.15),
considering an effective cohesion of 9 kPa, which is theoretically consistent with the material
strength according to MCC model (see Section 5.2.1.4). The effect of cohesion on stand-up
time will be discussed at the end of the present Section.
As explained for the MCC model, the failure process is driven by the dissipation of the
negative initial excess pore pressure (Fig. 5.22b). The increase in pore pressure results to a
decrease in the effective stresses, which, for the given deviatoric stress, leads to yielding. The
latter occurs in the model centre at t = 1.4 min (Fig. 5.22a). At the same time, the model
collapses (Fig. 5.22c); the plastic zone extends over the whole specimen (Fig. 5.23).
The predicted stand-up time of 1.4 min is by about factor 15 shorter than predicted by the
MCC model and by about factor 4.5 shorter than observed in the laboratory tests. The main
difference to the MCC model is that the MC model (with isochoric flow rule) does not takes
account of the plastic dilatant shearing, which, as shown by the MCC simulations, delays
failure. In the case of MCC model (which considers dilatant behaviour), negative excess pore
pressure develop after yielding, i.e. the pore pressure decreases (Fig. 5.15b), thus increasing
the mean effective stress (Fig. 5.15a) and stabilizing the system until the stress state reaches
the CSL (Fig. 5.15a). This cannot happen in the isochoric MC model. This is shown clearly
by the time-development of pore pressure Fig. 5.22b: Contrary to the MCC model and to the
experimental observations, pore pressure increases monotonically until failure. The predicted
time-development of the vertical displacement, too, differs qualitatively from the MCC
predictions and from the observed behaviour: the MC model predicts that heave (swelling)

125
rather than settlement occurs during the pore pressure dissipation; this trend reverses to
accelerating settlement only very close to the ultimate state (Fig. 5.22c).
Fig. 5.24 shows the influence of the effective cohesion (the only parameter that was not
measured experimentally) on the stand-up time. The sensitivity of the stand-up time to the
cohesion is high: the model collapses immediately after loading for a cohesion of 2 kPa
(minimum cohesion required for short-term stability) while, it remains stable even in the long
term for cohesion higher than cD,lim = 10.6 kPa.
For cohesion values between 2 and 10 kPa, the predicted stand-up time is less than 4 minutes
(0 – 2.4 min with average 0.76 min), and so rather shorter than stand-up time of the physical
models (2.2 – 8.2 min with average about 4 min). This means that, considering a cohesion
slightly different than 9 kPa, the MC model with isochoric plastic behaviour will
underestimate the stand-up observed experimentally. The underestimation of the stand-up
time would be even higher if considering the formation of shear bands observed
experimentally, which – as shown by Figure 5.23 – could not be reproduced numerically (a
localization does not occur even when imposing a lower resistance to the element rows
located at upper and lower corners of the model). As explained in Chapters 2 to 4, localization
would lead to a decrease in the operational strength parameters and thus also in the stand-up
time.
In conclusion, based upon the comparison between the experimental data (pore pressure and
vertical settlement) and the numerical results, it can be concluded that, for the present
example and for the assumed material parameters, the MC model with isochoric plastic
behaviour underestimates the stand-up time when the cohesion is lower or slightly higher than
the one estimated based upon theoretical consideration. The underestimation occurs because
this model does not capture the plastic dilation which has a stabilizing effect because it
maintains negative excess pressures.

126
 Chapter 5 – Uniaxial loading tests

t = 1 min 1.2 min 1.4 min

Plastic zone

Figure 5.23. Plastic zone on the mid plane of the sample at several times for test executed
underwater (MC model).

9.0
ts [min]

8.5

8.0
4.0
7.5
3.5
Experimental range
7.0
3.0
6.5
2.5
6.0
2.0
5.5
1.5
5.0
1.0 0 2 4 6 8 10 12

0.5

0.0
0 2 4 6 8 10 12
c' [kPa]
cD,lim = 10.6 kPa
cohesion required for drained stability

c' = 9.2 kPa

cohesion consistent with the stress path and
the MCC model (see Section 5.2.1.4)

Figure 5.24. Influence of the effective cohesion on the stand-up time (MC model, underwater
tests).

127
5.4 Conclusions
Chapter 5 investigated the problem of delayed failure induced by the transient swelling
process by means of simple uniaxial tests.
The experimental tests showed that shear bands develop in the corners of the specimen and
progress toward its centre over the time; failure occurs shortly after a continuous shear band
develops through the specimen. The occurrence of failure manifests itself by a rapid
acceleration of the vertical displacement.
The numerical interpretation of the measured pore pressure in the centre of the specimen
showed that failure is driven by the dissipation of the negative initial excess pore pressure and
delayed by the plastic dilation that occurs after the onset of yielding.
The experimental results further showed that the hydraulic boundary conditions influence
stand-up time considerably. In the underwater tests considerably lower stand-up times were
measured than in the tests with water vapour. In the latter the relative humidity of the air was
slightly lower than 100%, thus inducing a partial desaturation and a small suction (in the
order of few kPa) at the specimen boundary. The numerical simulations showed that this
small suction suffices to delay the failure significantly. The stabilizing effect is mainly due to
the increase of the effective stresses at the model boundary, and so of the apparent drained
cohesion.
The numerical simulations showed, furthermore, that the use of the MCC model allows to
qualitatively reproduce the behaviour observed experimentally in terms of pore pressure and
settlements evolution over time (but not the observed localization of deformations). However,
the MCC simulations overestimate the stand-up time (mainly because of the overestimation of
the plastic dilation). For the assumed material parameters, the isochoric MC model besides
failing to reproduce the observations qualitatively underestimates the stand-up time, as it does
not consider dilatant shearing. This result is relevant under the practical point of view, as it
suggest that the simple and widely used MC model represents a simple and “safe” approach
for the estimation of the stand-up time for problems characterized by stress paths similar to
the one reproduced in the present laboratory test.
In the opinion of the author, a parametric study investigating the sensitivity of the stand-up
time on the cohesion (for the plausible cohesion range) should be always performed when
adopting the MC model and, as safe engineering assumption, the stand-up time should be
taken equal to the lowest one.

128
 Chapter 6 – Centrifuge tests

6. Centrifuge tests

6.1 Introduction
Numerical investigations of delayed failure of cuttings were presented among others by
Kirkebø (1994), Potts et al. (1997), Kovacevic et al. (2007), Ellis and O’Brien (2007) and
Lollino et al. (2011). The listed authors investigated the progressive evolution of deep-seated
failures in overconsolidated clays (or tills) due to a combination of pore pressure increase and
intrinsic softening behaviour (see e.g. Potts et al., 1997). In the mentioned literature works,
the verification of the models and of the constitutive laws is performed calibrating the
parameters and the boundary conditions (e.g. the seepage boundary condition on the
excavated surface; Kovacevic et al., 2007) in order to reproduce already occurred delayed
failures of man-made cuts in natural soils (intrinsically characterized by heterogeneity and
variability of the ground parameters). As described by Lambe (1973), this approach (so-called
C-type prediction) is useful in order to improve the understanding of a problem, but it should
not be used in order to prove that a prediction technique is correct; in other words the
methods are numerically sophisticated but at the same time require assumptions that are
uncertain at the field-scale, making thus them validation not robust.
In order to reduce the uncertainties (homogeneity of the soil, material parameters, boundary
conditions, …) of the analysed problem, small-scale (reproducible) physical models are
required. Physical models at laboratory scale allow to control the model parameters and the
boundary conditions during the execution of the test. For the purpose of delayed failure
investigation, centrifuge modelling is the most attractive method because it allows to
reproduce the stress level of real scale problem (so-called prototype) and, at the same time, to
reproduce the diffusion process driving the instability (i.e. the swelling process) much faster
than at the prototype scale (real time).
The present Chapter investigates the delayed failure of a 7 m high underwater vertical cut
(Fig. 6.1) in over-consolidated clay by means of physical tests executed under well-controlled
conditions and numerical coupled hydraulic-mechanical FEM analyses. Differently form the
works of Deepa and Viswanadham (2009) and Davies (1985) (in which transient failure was
induced by an artificial change of boundary conditions and of loading, respectively), the
delayed failure of the proposed physical model occurs due to swelling process only (i.e. under
constant boundary conditions and loads).
The experiments have been carried-out in the drum centrifuge of ETH (Springman et al.,
2001). The simulation of the excavation during the test was achieved retracting a support wall
installed – prior acceleration – against the vertical cut (Fig. 6.2). The soil used in the present
experiments is a natural reconstituted high plasticity clay coming from Birmensdorf
(Switzerland) overconsolidated at 100 kPa (same material used for the uniaxial tests presented
in Chapter 5).

129
The FEM analyses have been carried-out with material constants, which have been
determined independently on the centrifuge test results (with laboratory tests or based upon
theoretical considerations).

Figure 6.1. Problem layout (at prototype scale).

Section a-a

Figure 6.2. Model set-up.

r
Figure 6.3. Drum centrifuge with assembled model.

130
 Chapter 6 – Centrifuge tests

6.2 Centrifuge modelling

6.2.1 Material and methods

6.2.1.1 Scaling laws and model dimensions

The scaling laws (see Table 6.1 for an overview) define the relationship between the
prototype scale and the model scale (physical models in the centrifuge) as function of n
(where n expresses the centrifuge acceleration as a multiple of the earth gravity g).
The centrifuge acceleration is obtained by rotating the model with a constant angular speed
of:

ng 60
w , (6.1)
r 2

where r = 0.874 m is the radial distance between the center of rotation and the middle height
of the excavation (see Fig. 6.3).

Table 6.1. Scaling laws prototype/model (being n the number of g’s) according to
Schofield (1980).

Acceleration n

Stress and pressure 1

Density 1

Length 1/n

Velocity 1

Volume 1/n3

Mass 1/n3

Force 1/n2

Time (diffusion) 1/n2

Time (dynamic) 1/n

Table 6.2. Model height h and corresponding centrifugal acceleration n and rotational
speed ω.

h [cm] n [-] ω [rpm]

Test 1 (reference case) 11.0 63.5 255
Test 2 10.7 65.7 259
Test 3 10.2 68.6 265
Test 4 11.2 62.5 253

131
Table 6.2 shows the height of the vertical cut and the corresponding centrifuge acceleration –
required in order to reproduce a 7 m height vertical cut – for the four executed centrifuge
tests.

6.2.1.2 Model preparation

The preparation of the model consisted of 5 phases: consolidation of the clay, excavation of
the vertical cut at 1 g, installation of the supporting wall, drawing of lateral contrast lines and,
finally, installation in the centrifuge.
The natural clay was initially mixed under vacuum with water (in order to remove the air
contained in the mixture) until reaching an homogeneous slurry with 100% water content.
The slurry was then poured into the strongbox and consolidated up to 100 kPa (Fig. 6.4a).
The consolidation was carried-out by means of an hydraulic press in 5 steps: 4 kPa, 15 kPa,
25 kPa, 50 kPa and 100 kPa. The total consolidation time amounted to about 25 days. In order
to eliminate friction between soil and strongbox (during both the consolidation and the test
phase), the longitudinal walls of the box where initially lubricated by means of paraffin oil.

(a)

(b)

(c) Actuator

Figure 6.4. Model preparation: (a) consolidation, (b), excavation and, (c), insertion of the
retaining wall.

132
 Chapter 6 – Centrifuge tests

The material in front of the vertical cut was excavated at the end of the consolidation phase
(Fig. 6.4b). A sharped aluminum blade was temporarily installed in the middle of the model
in order to assure an accurate geometry of the cut and protect it during the excavation.
The support wall was installed in the model after excavation (Fig. 6.4c). The longitudinal
movement of the support wall was controlled by an electric actuator. In order to avoid
sticking of the material during the execution of the centrifuge test (particularly during the
retraction of the wall), the wall was covered by oily sand.
The strongbox has a removable Plexiglas window on one of its longer sides, which allows to
observe the overall model behavior during the centrifuge test (see photography in Fig. 6.4).
After the installation of the supporting wall, the window was temporarily removed and
horizontal white lines spaced 1 cm were drawn on the lateral boundary of the model.
Before installation in the centrifuge, small plastic pipes (4 mm in diameter) were installed in
the corners of the model. The plastic pipes were connected to the bottom plate of the
strongbox (drainage plate) in order to reduce the time required for the rise of the water level
inside the model during the initial phase of the centrifuge test. An overflow pipe was
connected to the drainage plate and fixed against the external wall of the strongbox. This
allowed – in combination to a constant water inflow through the drainage plate of the
strongbox – to keep a constant water level during the duration of the centrifuge test.
Finally, the strongbox was installed in the drum centrifuge (note that the model stands
vertically, see Fig. 6.3) and connected to the external water supply system.

6.2.1.3 Instrumentation

The model was instrumented with laser distance meter and micro pore pressure transducers
(ppts) measuring the vertical settlement and the pore pressure, respectively. Moreover, in
order to observe the overall model behavior during testing, two fast cameras were fixed on the
lateral and upper sides of the strongbox.
A total of seven ppts of type Druck (type PDCR81) and Keller (type 2MIX/2bar/81840.1)
were used. Six of them were placed into the clay and one was connected to the water supply
system, in order to monitor the hydraulic head in the drainage plate during the test.

[cm]

Section a-a:
4
8,5

Figure 6.5. Model instrumentation.

133
Four out of the six ppts were placed near the theoretical lowest inclination of the failure
surface (i.e. 45°) at a depth of 4 cm (2 ppts) and 8.5 cm (2 ppts), respectively (Fig. 6.5). The
selected position allowed to collect data near the potential sliding plane without influencing
the failure mechanism. The remaining two ppts were installed near the strongbox back wall
(same elevation as described above) as control of the pore pressure field in a larger distance
from the failure plane.
The ppts were installed into the clay at the end of the consolidation step of 50 kPa as follows:
drilling the clay by means of a borer through the strongbox back wall (special openings were
disposed in the wall); introduction of the sensor in the hole; and, finally, filling the hole with
liquid slurry by means of a syringe equipped with a long injection micro pipe. In order to
avoid the risk of trapping air bubbles inside the hole, the slurry was injected from the end of
the hole (similarly to the grouting procedure used for rock bolts grouted by mortar). After
filling the hole, the opening was sealed by means of a rubber ring and closed with a screwed
steel cylinder. Finally, the vertical pressure was increased to 100 kPa, allowing thus the fresh
slurry (injected around the sensor and the cable of the ppts) to consolidate (ensuring thus a
reliable pore pressure measurement during the tests in the centrifuge).
A short range distance sensor (high precision laser distance meter) was used for the
measurement of the vertical settlement of a fixed point located at the crest of the wall (Fig.
6.5). In order to avoid the measurement errors associated with the refraction of the laser on
the wet clay surface, the target point was located on an aluminum pin with flat circular head.
The pin was constructed with a ring at its mid-height, which stabilized the pin and prevented
its penetration into the soil during testing, thus ensuring that the pin displacement is equal to
that of the soil surface. Due to the small size of the pin and to the low unit weight of
aluminum, the pin weight is negligible compared to the stresses in the soil and does not
influence the failure mechanism.

6.2.1.4 Test procedure

The centrifuge test started with rising of the water level in the model. This was carried-out at
a rotational speed of 100 rpm (corresponding to a centrifugal force of 10 g). After the water
level reached the top of the vertical cut, the rotational speed was increased to its final value
(255 rpm for a 11 cm high wall; Table 6.2).
The acceleration of the model caused the generation of positive excess pore pressure in the
clay. Therefore, after the acceleration phase, a consolidation phase of about 8 hours was
required in order to reach hydrostatic pore pressure distribution. The latter was controlled by
means of the micro pore pressure transducers.
Finally, the supporting wall was quickly retracted (thus simulating a rapid excavation) and the
model behavior was monitored until failure.

134
 Chapter 6 – Centrifuge tests

6.2.1.5 Determination of the material parameters

Table 5.3 gives the parameters of the clay (Birmensdorf clay) used in the centrifuge tests for
the Mohr-Coulomb (MC) and the Modified Cam Clay (MCC) Model. With the exception of
the effective cohesion of the soil, the values were obtained by means of the laboratory tests
described in Chapter 5 (Table 5.2).
The remarks of Section 5.2.1 about the uncertainty related to the material parameters
(inconsistency between the Young’s modulus of the MC model and the MCC parameters; see
also Section 6.4.3.1) apply also to the present study.

Table 6.3. Material constants.

Saturated unit weight  [kN/m3] 18.2

Unit weight of water w [kN/m3] 10

Liquid limit wl [%] 58

Plastic limit wp [%] 19

Plasticity index Ip [%] 39

Void ratio at v = 1 kPa e0 [-] 2

Seepage flow parameters

Permeability k [m/s] 1.5 10-9
Water compressibility cw [MPa] 0

Parameters for linearly elastic, perfectly plastic MC material

Young’s modulus E [MPa] 18
Poisson’s ratio  [-] 0.3
Cohesion c' [kPa] 8 kPa, see Section 6.2.1.5.1
Friction angle ' [°] 24.5
Dilation angle ψ [-] 0

Parameters for Modified Cam Clay material

Gradient of swelling line [-] 0.01 (set 1) / 0.022 (set 2)
Gradient of compression line [-] 0.21 (set 1) / 0.152 (set 2)
Poisson’s ratio  [-] 0.3

135
6.2.1.5.1 Theoretical estimation of the effective cohesion

As done in Section 5.2.1.4, the effective cohesion is estimated assuming that the soil behaves
as MCC and considering the theoretical effective stress path expected during testing. More
specifically, the cohesion is taken such that yielding in the MC model occurs at the same
point (p', q) as for the MCC model. The latter corresponds to the intersection point between
the assumed effective stress path and the yield surface according to the MCC model. The
slope of the MC yield surface in the p'-q plane is obtained from the friction angle measured in
the laboratory at critical state. This procedure is applied on order to estimate two values of
cohesion: the one representative for the underwater excavation (at its mid-height), the other
for the T-bar test (Stewart and Randolph, 1994).
The purpose of the T-bar test (see Section 6.2.1.5.2) is to validate the method used in the
present Thesis for the estimation of the effective cohesion by comparing the measured
undrained shear resistance of the ground (by means of the T-bar test) with the one calculated
based upon the cohesion obtained from the theoretical effective stress path. Due to the
geometry of the problem (for both underwater excavation and T-bar test) plane strain
conditions are considered for the estimation of the cohesion.
The stress path considered for the T-bar test corresponds to undrained deviatoric loading up to
the yield surface, while the stress path for the excavation (considering a point located on the
sliding plane) corresponds to undrained deviatoric unloading followed by swelling under
constant deviatoric stress (similar to Section 5.2.1.4). The latter represents a simplifying
assumption because the deviatoric stress may increase over time due to stress re-distribution
(see Chapter 3).

50
q [kPa]

45
Size of the yield surface for triaxial compression:
CSL
40 6 sin  '
M  0.96
3  sin  '
35 A
q u,lim
30 B B' Size of the yield surface for plane strain conditions:
q lim
25 M  3sin  '  0.72

20

15

10

5
t =0 a0
0
0 20 40 60 80
p' [kPa]

Figure 6.6. Schematic representation of the effective stress path at the mid-height of the
excavation for T-bar test (0 – A) and for the underwater excavation (0 – B’ – B).

136
 Chapter 6 – Centrifuge tests

Figure 6.6 shows the effective stress paths of a point located at the elevation of the mid-height
of the vertical cut. As the material is over-consolidated, the undrained stress path in the p'-q
space is given by a vertical line (undrained elastic response) which starts from the stress state
just before the retraction of the supporting wall (t = 0). For the undrained deviatoric loading
up to the yield surface the stress path ends at point A (the intersection point A represents the
undrained resistance), while for the undrained deviatoric unloading (up to point B') followed
by swelling under constant deviatoric stress it ends on the yield surface at point B.
The state at t = 0 is taken as isotropic in Figure 6.6 because, as explained in the following
based upon the known over-consolidation ratio (OCR) and Mayne and Kulhawy’s (1982)
equation, the coefficient of lateral pressure K0 is theoretically equal to about 1 at mid-height
of the excavation at the state just before the retraction of the supporting wall. The over-
consolidation ratio (OCR) is obtained dividing the maximum effective vertical stress
experienced by the soil 'v,max to the actual effective vertical stress 'v. As the consolidation
(up to 100 kPa) was carried-out at 1g in the physical model, 'v,max is practically uniform and
equal to the surface loading of 100 kPa that was applied in the last consolidation step.
Consequently,

 v ,max ' 100 kPa

OCR( y )   , (6.2)
 v '(y)  v '(y)

where y is the depth in the model scale and σv' the effective vertical scale in the accelerated
model (63.6g in the case of Test 1, see Table 6.2). As the vertical stress in the accelerated
model at the state just before the retraction of the supporting wall increases linearly over
depth (according to the unit-weight of the soil), the OCR-distribution is non-uniform (Fig.
6.7). The OCR equals about 2 at the bottom of the excavation, 4 at the mid-height and tends
to infinity at the soil surface. For such OCR values the coefficient of lateral pressure K0 can
be estimated after Mayne and Kulhawy (1982):

K0  1  sin  ' OCRsin  ' . (6.3)

For the OCR-values after Figure 6.7 and the known friction angle (' = 24.5°), K0 increases
from about 0.8 at the bottom boundary to more than 2 at the soil surface (Fig. 6.8). The value
of K0 =1 in Figure 6.6 is approximately equal to the one prevailing at mid-height of the cut
(0.98). The mean effective stress p' at t = 0 in Figure 6.6 is thus equal to the effective vertical
stress at mid-height (28 kPa). Generally, for plane strain conditions,

1  K0
p'  v ' (6.4)
2 .

137
For point A, Equation (6.4) in combination with the yield surface of the MCC model provides
the undrained resistance

2
 p' 
qu ,lim  a0 M 1    1 , (6.5)
 a0 

where M  3 sin  '  0.72 is the slope of the CSL for plane strain conditions and a0 (= 49
kPa) is the centre of the yield surface after consolidation of the material under 100 kPa
(derived solving according to Eq. 5.6).

OCR [-]

0 2 4 6 8 10
0

20

40

60

80
y [mm]

y
110 mm
100

Figure 6.7. Vertical distribution of the OCR.

K0 [-]

0.0 0.5 1.0 1.5 2.0 2.5 3.0

0

20

40

60

80
y [mm]

y
110 mm

100

Figure 6.8. Vertical distribution of the coefficient of lateral pressure.

138
 Chapter 6 – Centrifuge tests

The point B' represents the undrained deviatoric stress change due to the excavation. Under
the simplifying assumption of a full lateral unloading at the mid-height of the excavation the
latter amounts to the effective vertical stress at mid-height qlim = 28 kPa (as the stress state is
isotropic). By solving Equation (6.5) with respect to p' and by inserting qlim instead of qu,lim, it
is possible to obtained the mean effective stress state at point B (p' = 18 kPa).
As done in Chapter 5, the effective cohesion at the mid-height is obtained by taking into
account the MC failure criterion formulated in the p'-q stress space (Vrakas and Anagnostou,
2015):

q 
c '   u ,lim  p '  tan  ' , (6.6)
 M 

which in combination with Eq. (6.5) (i.e., under isotropic plane strain conditions) leads to:

  
2 
p '
c '   a0 1    1  p '  tan  ' . (6.7)
  a0  
 

Equation (6.7) shows that for given friction angle, the effective cohesion depends only on the
mean effective stress. At the mid-height of the excavation the effective cohesion amounts to
7.5 kPa for point A (stress path assumed for the T-bar test) and to 9.1 kPa for point B (stress
path assumed for the underwater excavation).

6.2.1.5.2 Cohesion estimation based upon the results of T-bar penetrometer tests

As explained in Section 3.3.1, the undrained shear strength of the ground can be expressed as
follows:

su  c 'cos ' p 'sin  ' , (6.8)

where p' is taken according to Equation (6.4).

Solving Eq. (6.9) with respect to the effective cohesion gives:

su
c'   p ' tan  ' . (6.9)
cos  '

For given friction angle φ' and known mean effective stress p' (Eq. 6.4 with the K0-
distribution after Fig. 6.8), Equation (6.9) allows to estimate the effective cohesion from the
undrained shear strength. The latter is measured here by means of T-bar penetrometer tests.

139
 Section a-a:

Figure 6.9. Schematic representation of the T-bar tests execution.

The T-bar penetrometer test was developed by Stewart and Randolph (1994) and allows to
indirectly measure the undrained shear resistance of the soil. The test consists in lowering a
bar with a T-shaped end with a constant low vertical speed (in the present case 0.5 mm/s) and
measure the force F resisting to penetration. Assuming the formation of a plastic plane strain
failure mechanism around the T-bar it is possible to relate the measured force F (acting on the
T-shaped end) to the undrained shear resistance su:

F
su  , (6.10)
Nb d L

where Nb is the dimensional resistance factor (10.5 in the case of steel surface and plane strain
mechanism, Stewart and Randolph, 1994), d is the diameter of the bar (7 mm) and L is the
length of the T-shaped end (28 mm).
Weber (2007) and Gautray (2014) reported about a successful application of the T-bar
method (with equipment developed at the ETH) during an on-flight centrifuge test. Here, the
T-bar penetrometer tests were executed after failure of the vertical cut. In order to test the
undisturbed material, the tests were executed at a distance of 8 cm from the failure zone (Fig.
6.9). Figure 6.10 shows the measured force F and the calculated undrained shear strength
over the depth of the vertical cut. The shear resistance increases with the depth up to a
maximum of about 25 kPa. At mid-height of the vertical cut (y = 55 mm), the shear resistance
amounts to approximately 18 kPa. It is to remark that the shear resistance calculated over the
first 14 mm (2 diameters of the T-bar) of penetration is not correct as the failure mechanism is
different than the assumed one. The local increase of resistance at around depth 50 mm and
90 mm may be related to the proximity of the ppts.
The solid lines in Figure 6.11 show the vertical distribution of the effective cohesion based
upon the measured undrained shear strength (Fig. 6.10). As exception of the first 14 mm
below the ground surface, over which the T-bar results are note reliable, the cohesion varies
between 7 and 11 kPa almost over the height of the excavation and decreases to 1.5 kPa or 5
kPa (depending on the considered T-bar test) at the bottom of the excavation.
Generally, the cohesion estimated in the last section (7.5 kPa) based upon theoretical
considerations for the undrained deviatoric loading up to the yield surface (dashed line)
agrees well (is only slightly lower than) the cohesion estimated experimentally (solid lines).
This result confirms the validity of the methodology used for the estimation of the effective
cohesion.

140
 Chapter 6 – Centrifuge tests

su [kPa]
0 5 10 15 20 25 30

F [N]
0 10 20 30 40 50 60
0
Unreliable range
y
110 mm
20

40

60

80
y [mm]

100

Figure 6.10. Measured (solid lines) and estimated (dashed line) force F (resisting to T-bar
penetration) and corresponding undrained shear resistance su as a function of
the penetration distance y.

c' [kPa]
0 2 4 6 8 10 12
0
Unreliable range

20

y
110 mm
40

60

80
y [mm]

100

Figure 6.11. Vertical distribution of the effective cohesion estimated based upon the
measured undrained shear strength (solid lines) and theoretical considerations
(dashed lines).

141
6.2.1.5.3 Selection of the effective cohesion to be used in the MC model

For the sake of simplicity, the MC effective strength parameters will be taken constant over
depth. Based upon Figure 6.11, the reasonable cohesion range to be considered by the MC
model (according to the results of the T-bar test) is 7 – 9 kPa while, according to the
theoretical considerations, the cohesion representative for the underwater excavation amounts
to about 9 kPa.
In the numerical example of Section 6.4, the cohesion will be 8 kPa. The assumption of a
constant cohesion basically implies a linear increase in undrained shear strength over depth
(represented by the dashed line in Fig. 6.10 for 8 kPa).

6.2.2 Experimental results

Figures 6.12, 6.13 and 6.14a show the time-development of the vertical settlement,
photographs of the model at two times and the time-development of the pore pressure,
respectively, for test No. 1. For sake of completeness, Figure 6.14b additionally shows the
pore pressure evolution over time of all four centrifuge tests (10 rather than 16 curves are
shown, because some ppt stopped measuring during execution of Tests 2, 3 and 4).
The reference state (t = 0) for the presented vertical settlement is the one prevailing just
before retracting the supporting wall. Pore pressure distribution at t = 0 is hydrostatic (28 kPa
and 52 kPa at the highest and lowest ppts elevation, respectively).
The quick retraction of the supporting wall (which simulates an undrained excavation) causes
a practically instantaneous settlement of the soil surface (by about 2 mm; Fig. 6.12) as well as
negative excess pore pressures (of about 55 kPa and 115 kPa, at points B and C, respectively;
Fig. 6.14a). The development of negative excess pore pressures can be explained as a
consequence of the excavation-induced reduction of the horizontal total stress along the
vertical cut; pore pressure decreases more at the elevation of the deeper ppt, because the in
situ stresses (and thus the excavation-induced decrease in total stress) increase over depth.
The negative excess pore pressures dissipate over time (Fig. 6.14a). As one might expect for a
dissipation process, its rate decreases over time initially, but reaches a constant value after
about 2 min (upper ppt) or 3 min (lower ppt).
At t = 17 min a tensile crack develops at the crest of the excavation in a distance of about half
the height of the cut (Fig. 6.13a). The development of such cracks during progressive failure
is observed also in the field (e.g. Banerjee et al. 1988, Cooper et al. 1998, Hughes et al. 2007,
Lollino et al. 2011). At t = 18.5 min a thin shear band starts to develop at the toe of the
excavation; it is probably associated with a high stress concentration over there.
The settlement increases at an approximately constant rate over a long period and starts to
accelerate significantly shortly after the above-mentioned first signs of failure (at about t = 22
min; Fig. 6.12). Cooper et al. (1998) defined this time as the end of the “apparent overall
stability”. The measured displacement evolution – characterized by an initial phase of low
displacement rate followed by a phase of acceleration – was also observed in the field by
Cooper et al. (1998) and Kovacevic et al. (2012).

142
 Chapter 6 – Centrifuge tests

10
A

uy [mm]
9 ts
uy
8
7
6
5
4
3
2
1
0
0 5 10 15 20 25 30
t [min]
Figure 6.12. Evolution over time of the vertical displacement after retraction of the
supporting wall (Test No.1).

(a)
tensile
crack

shear
band
t = 23.5 min

(b)

θ ≈ 57
t = 25.5 min
Figure 6.13. Deformed model at, (a), 23.5 minutes and, (b), 25.5 minutes (Test No.1).

143
At t = 25.5 min the tensile crack and the shear band join (Fig. 6.13b) and the cut collapses
(see settlement acceleration in Fig. 6.12); the stand-up time in this test is thus equal to 25.5
minutes. At ultimate state, a wedge slides on a plane inclined by 57°, which is close to
Coulomb’s inclination (45°+φ'/2 = 57.3°). Vermeer (1990), too, observed that the shear band
inclination in finely grained soils is close to the Coloumb’s inclination.
It is interesting to note that, shortly before failure, the pore pressure measured by the lower
ppt decreases (C; Fig. 6.14a), while the pore pressure at the upper ppt first decreases and then
increases. This may be due to an increase of the total stresses caused by stress re-distribution,
which occurs after tensile failure.

(a)
60
p w [kPa]

51.5 kPa (hydrostatic pressure at lower ppts elevation)

40 C, lower ppts
negative 28 kPa (hydrostatic pressure at upper ppts elevation)
excess pore
pressure 20 B, upper ppts
generated by
the retraction
of the wall: 115 kPa 55 kPa 0
0 5 10 15 20 25 30
t [min]
-20 B
C
-40

-60

(b)
60
p w [kPa]

C, lower ppts
40

20 B, upper ppts

0
0 5 10 15 20 25 30
t [min]
-20

-40

-60
Figure 6.14. Evolution over time of the pore pressure for, (a), the Test No.1 and, (b), all
centrifuge tests.

144
 Chapter 6 – Centrifuge tests

Finally, the model behaviour was both qualitatively and quantitatively reasonably good
reproducible, which indicates a high quality of the tests. The variability of the measured pore
pressures and displacements was small (Fig. 6.14b and 6.15, respectively). The average stand-
up time of the four centrifuge tests amounts to 24.5 minutes (Fig. 6.15a), which corresponds
(taking account of the scaling laws and centrifuge accelerations; Table 6.1 and 6.2) to 72.4
days at the prototype scale (Fig. 6.15b).

(a)
8
A
uy - uy0+ [mm]

7 uy T est No. 4
h 1
6 n∙g
2
3
5

1
tavg =
24.9 min
0
0 5 10 15 20 25 30
tmodel [min]

(b)
500
uy - uy0+ [mm]

A
T est No..4
uy
1
400 H= 7m
1∙g
2
3

300

200

100
tavg =
72 d
0
0 10 20 30 40 50 60 70 80 90
tprototype [d]
Figure 6.15. Evolution over time of the vertical displacement of all 4 tests at, (a), the model
and, (b), the prototype scale.

145
6.3 Limit equilibrium analysis
Before proceeding with the (more demanding) numerical analyses, a simple limit equilibrium
analysis is performed here in order to verify the validity of the assumed strength parameters.
The validation is important because, as explained at the end of Section 6.2.1.5.3, the cohesion
was the only parameter which was not estimated based upon laboratory tests.
This Section estimates the time-development of the safety factor (SF) of the cut (in the
prototype scale) considering the limit equilibrium of a sliding wedge with or without tension
crack (see inset in Fig. 6.16) and taking account of the measured pore pressures. The tension
crack considers the observed failure mode (Fig. 6.13) and is taken 1.5 m long.

(a) Without tension crack

pw [kPa]
-60 -40 -20 0 20 40 60 80
0

1
G
upper ppts row
2
l W1 H=7m
N 3
Position
lower ppts row upper ppts 4
l = 8.4 m
s 5 t=0
θ = 57
6 t = ts
Position
lower ppts 7
t = 0+
s [m] 8

(b) With tension crack

pw [kPa]
-60 -40 -20 0 20 40 60 80
0
W2 Assumed crack
1.5 m
length 1
G
upper ppts row
2
W1 H=7m
N l 3
Position
lower ppts row upper ppts 4

s l = 6.6 m
5
t=0
θ = 57
Position 6
t = ts
lower ppts
7
t = 0+
s [m] 8

Figure 6.16. Distribution of the pore pressure distribution along the

potential sliding surface at several times (Test No.1).

146
 Chapter 6 – Centrifuge tests

The safety factor is defined as the ratio of the shear resistance Tres along the sliding plane to
the shear force T, that is required for fulfilling equilibrium:

Tres
SF  . (6.11)
T

The forces acting upon the wedge (see insets in Figure 6.16) have to fulfil equilibrium parallel
and normal to the sliding plane:

T  G sin   W1  W2  cos , (6.12)

and

N  G cos  W1  W2  sin  , (6.13)

where N is the integral of the (total) normal stress along the sliding plane; W1 and W2 denote
the integral of the water pressure acting on the vertical wall and along the tension crack,
respectively (W2 = 0 if there is no tension crack); G is the total weight of the wedge and  is
the inclination of the sliding plane. Assuming that the shear strength reduction factor η is
constant, the shear resistance amounts to

  l

Tres    c ' l  tan  '  N   pw ds   , (6.14)
 
  0 

where η = 0.91-1.0 for φ' = 24.5° and  = 0° depending on the rotation angle of the principal
stresses and thus on the shear deformation (Section 2.2). For simplicity, a piecewise linear
pore pressure distribution is assumed between the ppts (where the pore pressure is measured
over time) and the integration boundaries, where the pressure is hydrostatic (see diagrams in
Fig. 6.16). Taking account of the time-development of the measured pore pressures, the time-
development of the safety factor can be determined with Equations (6.11), (6.12) and (6.14)
(Fig. 6.17). The assumptions underlying curve 1 are reasonable for the initial phase of the
pore pressure dissipation process, before the tension crack formation and before the
development of the shear strains that would cause significant rotations of the principal axes.
The safety factor follows the (inverse) trend of the excess pore pressure dissipation (Fig.
6.14): it decreases rapidly within the first 2 – 3 minutes and very slowly afterwards, and
increases slightly a few minutes before failure (perhaps due to the plastic dilation of the soil).
Curve 1 is obviously inadequate for the later dissipation stage (tension crack, probably large
shear strains with principal stress rotation and thus η < 1) and in fact ends at a safety factor of
1.25, which contradicts to the observed failure. With the formation of the crack (at about t =
17 min) the safety factor would move (more or less suddenly depending on the crack
propagation speed) from curve 1 to curve 2 and with increasing shear strains towards curve 3.
The latter applies to purely plastic shearing along the entire shear band and leads to a final
safety factor of about 1, which is an astonishing result in view of the very simplified
computational model. The simplified analysis provides evidence about the adequacy of the
assumed material constants, particularly of the cohesion.

147
 3.0

SF [-]
A 1: η = 1.0, without tension crack

2: η = 1.0, with tension crack

2.5 3: η = 0.91, with tension crack

B

2.0

1.5

A'
B'
1.0
C'

0.5
10 15 10
1 15
1 20
1 25
1 301
t [min]

t = 18.5 min, apparition shear band

t = 17 min, apparition of tension crack

Figure 6.17. Evolution over time of the safety factor against instability estimated by means of
limit equilibrium (Test No.1).

148
 Chapter 6 – Centrifuge tests

6.4 Numerical modelling

6.4.1 Computational model

The problem is formulated in plane strain at the prototype scale. Figure 6.18 shows the
geometry and the boundary conditions of the problem under investigation.
The soil is modelled either as an isotropic, linearly elastic and perfectly plastic material
obeying the Mohr-Coulomb yield criterion with isochoric plastic flow or as a Modified Cam
Clay material (Table 6.3). The seepage flow follows the Darcy’s law.
The shape of the yield function for the MCC model in the deviatoric plane is taken according
to the Gudehus (1973) expression:

2b
g  L   , (6.15)
1  b  (1  b)cos  3 L 

where L is the Lode’s angle and b is the convexity factor (assumed equal to 0.8). The present
example adopts the Gudehus formulation for the MCC model in order to capture the possible
small changes of the Lode’s angle during the stress re-distribution process, and so, reproduce
as best as possible the expected material response. As shown in Section 3.4.2 for a similar
problem, a circular yield surface (as in Chapter 5) with slope of the CSL matching plane
strain conditions would be also a reasonable assumption for the problem under consideration.
The spatial discretization employs 8-node square elements with biquadratic displacement and
bilinear pore pressure shape functions. The element side length is equal to 50 cm, which
means that the height of the vertical cut is discretized by 14 elements (Fig. 6.18); the effect of
the mesh size on the stand-up time is studied in Section 6.5.

Φ = h0

H= 7 m
Φ = h0

h 0 = 10.5 m No flow

Φ = h0

y No flow
x Φ = h0

25.4 m

Figure 6.18. Numerical model and boundary conditions imposed during the simulation of the
delayed failure.

149
As the ppts are fixed to the strongbox by means of micro-cables, they cannot move
horizontally towards the cut and this may have an influence on the pore pressure field. The
potential effect of the ppts on the model behaviour was investigated by comparative
computations in which the nodes located along the horizontal lines between the strongbox and
the ppts have been kept fixed in the horizontal direction. This represents a simplifying
assumption that probably overestimates the effect of the ppts.

6.4.2 Simulation procedure

6.4.2.1 MC model

The initial conditions represent the state just before excavation. The initial effective stresses
increase linearly with depth according to the submerged unit weight and a lateral earth
pressure coefficient equal to unity. The latter is consistent with the value at the mid-height of
the vertical cut (see Fig. 6.8).
The excavation is simulated by reducing the horizontal effective stresses at the vertical
excavation boundary and the effective vertical stress at the bottom boundary of the excavation
from its initial value to zero in one short time step of 3600 s, while the water pressure along
the boundaries of the model is kept equal to hydrostatic one. Afterwards, a transient analysis
starts.

6.4.2.2 MCC model

The MCC model requires assumptions on the spatial distribution of the initial void ratio and
of the initial effective stresses (particularly of the K0 coefficient) consistent with the
constitutive model itself and with the construction history of the physical model.
i) Initialization of the numerical model: The initial conditions represent the state at the
end of the consolidation at 100 kPa. As the latter was carried-out outside the
centrifuge and the effect of a gravity acceleration of 1g is negligible in the model
scale, gravity is disregarded in the model initialization, too, although the numerical
problem is formulated at the prototype scale. Therefore, uniform initial pore
pressure, vertical effective stress and void ratio e0 are considered (equal to zero, 100
kPa and 1.1, respectively). The coefficient of horizontal stress is taken after Jacky
(1948) (K0 = 1-sin' = 0.58), because the soil is normally consolidated at this state.
ii) This step simulates the complete vertical unloading of the physical model and the
excavation of the vertical cut under 1g. Therefore, the gravitational forces are not
activated yet in numerical model. The complete unloading is simulated by setting all
boundary tractions equal to zero and performing a steady state analysis (pore
pressure remains equal to zero). Subsequently, in view of the next simulation step,
the soil elements in front of the vertical cut are deactivated and nodes on the vertical
cut fixed in the horizontal direction in order to simulate the supporting wall.

150
 Chapter 6 – Centrifuge tests

The next steps simulate the test phases in the centrifuge:

iii) The acceleration in the centrifuge and the successive consolidation is simulated by
activating the gravitational forces, fixing the boundary pore pressures equal to the
hydrostatic pressure and performing a steady state analysis.
iv) The rapid excavation of the vertical cut is simulated by deactivating the horizontal
fixity of the nodes located along the vertical cut and performing a transient analysis
for one short time step (3600 s).
v) The phase of excess pore pressure dissipation is simulated by a transient analysis as
for the MC model.

6.4.3 Results

6.4.3.1 MC model

First, the results obtained without fixing the nodes at the elevation of the ppts will be
discussed. The solid line in Figure 6.19a represents the predicted time-development of the
vertical displacement at the location of the pin used for the laser measurement. The predicted
stand-up time is equal to approximately 32 days, which is in the same order of magnitude as
the actual one (66 – 77 days with an average of 72 days). However, differences exist between
observation and prediction with respect to the deformation pattern and to the time-
development of settlement. These differences are discussed below.
The predicted deformation pattern (Fig. 6.19b) is that of one sliding wedge – without the
observed tension crack (Fig. 6.13) – and this along a plane, which is less inclined than the
actual one (53° vs. 57°). The reason for the absence of a tension crack in the numerical
simulation is that tensile failure is not taken into account by the adopted MC model. Tension
cracks shorten drainage paths and, consequently, tentatively accelerate excess pore pressure
dissipation and failure. However, according to the experimental results, the tension crack
started to develop very late, which means that its effect is relevant only close to failure. The
results of the Chapters 3 and 4 show that, the inclination of the slip line increases with the
fineness of the mesh. This could explain the difference with respect to the inclination of the
sliding plane: the latter is slightly underestimated by the numerical analyses because the
chosen discretization is not fine enough.
According to the numerical simulation, the soil surface heaves at a low rate over a long period
of time (Fig. 6.19a), while the monitoring results show a settlement right from the start (Fig.
6.19c). The predicted heave is probably due to an overestimation of the soil swelling by the
MC model, which assumes a constant Young’s Modulus, thus underestimating the stiffness of
the soil in unloading.
Swelling occurs due to the continuous increase in the pore pressure until few days before
failure (solid lines in Fig. 6.20a). The predicted time-development of the pore pressure agrees
qualitatively well with the measurements (Fig. 6.20b), but the predicted excavation-induced
instantaneous drop in pore pressure (about 8 kPa and 20 kPa at point B and C, respectively) is
considerably less than measured (about 55 kPa and 115 kPa at point B and C, respectively).

151
 (a)

30

u y [mm]
A
ts
uy
25 without fixing the nodes
corresponding to the ppts

20 fixing the nodes

corresponding to the ppts

15

10

MC (c’ = 8 kPa)
0
0 5 10 15 20 25 30 35
t [d]

(b) 53

without fixing the nodes, t = 32 days

(c)

500
uy - uy0+ [mm]

400

300

200

100

0
0 10 20 30 40 50 60 70 80 90
tprototype [d]
Figure 6.19. MC model: (a) evolution over time of the vertical displacement at point A, (b),
displacement vectors at failure and, (c), monitoring results at prototype scale.

152
 Chapter 6 – Centrifuge tests

(a)

60
51.5 kPa (hydrostatic pressure at lower ppts elevation)

p w [kPa]
C, lower ppts
40
28 kPa (hydrostatic pressure at upper ppts elevation)
B, upper ppts
20

0
0 5 10 15 20 25 30 35
t [d]
-20 B
C without fixing the nodes
corresponding to the ppts
-40
fixing the nodes
corresponding to the ppts
MC (c’ = 8 kPa)
-60

(b)

60
p w [kPa]

C, lower ppts
40

20 B, upper ppts

0
0 20 40 60 80
t [d]
-20

-40

-60
Figure 6.20. MC model: (a) evolution over time of the pore pressure at the ppts locations
(points B and C) and, (b), monitoring results at prototype scale.

This difference seems to be due the fact that the ppts represent an obstacle to horizontal
displacements of the soil in their close vicinity. In fact, the comparative computations with
fixed nodes predict an excavation-induced instantaneous drop in pore pressure by about 39
kPa and 72.5 kPa (at point B and C, respectively; dashed lines in Fig. 6.20a) which is closer
to the monitored values.
After the initial phase of the excess pore pressure dissipation process (about 5–10 % of ts), the
pore pressures obtained with fixed nodes become quite similar to those obtained without
fixing the nodes (Fig. 6.20a).

153
The simulation with fixed nodes results to a stand-up time of about 17 days (Fig. 6.19a,
dashed line), i.e. the difference between observed and predicted stand-up time is bigger than
in the case of free nodes.
The computation with fixed nodes provides a possible interpretation of the excavation-
induced instantaneous drop in pore pressure measured experimentally. The comparison of the
results obtained with and without fixed nodes also indicates that the ppts locally influence the
pore pressure field (the effect of the ppts is limited to the vicinity of their tips). This means
that the measured pore pressure may not be representative of the pore pressure acting in the
model (the size of the ppts is small compared to the width of the model). For this reason, the
safety factor illustrated in Figure 6.17 may be – particularly for the initial phase of the
swelling process – overestimated, as the high negative excess pore pressure measured during
the test may occur only at the ppt position.
In search of possible reasons for the difference between predicted and observed stand-up
time, the sensitivity of the simulation results with respect to cohesion, Young’s modulus E
and permeability k was investigated, whereby only the product Ek is significant on account of
the structure of the consolidation equations (Fig. 6.21). For the reasonable cohesion range (7
– 9 kPa, Section 6.2.1.5.4) and Ek after Table 6.3, the stand-up time is equal to 12.5 – 100
days (4 – 130 d when considering a variation by ±30% in Ek – Note: the magnitude of the
variation has been chosen arbitrarily), which is in the order of the measured time (70 – 80 d),
but mostly lower. A simulation with a finer mesh discretization would predict an even lower
stand-up time (Section 6.5). As shown in Chapter 5, the reason for the underestimation of the
stand-up time may be due to the isochoric flow rule of the MC model, which does not
consider the plastic dilation delaying the excess pore pressure dissipation process.
On the other hand, it is important to remark, that there is an uncertainty concerning the actual
material parameters. Specifically, the Young’s modulus provided by Küng (2003) (18 MPa)
does not seem to be consistent with the MCC parameters  and . Taking into account the
mean effective stress prevailing at the mid-height of the excavation before excavation (about
29 kPa), a Poisson’s number of 0.3 and  and  after Table 6.3, the Young's modulus would
be equal to 0.5 – 1 MPa, i.e. by factor 18-36 lower than given by Küng (2003). Considering
such a low Young's modulus would result to a by factor 18-36 higher stand-up time, which is
considerably higher than the experimentally observed stand-up time.

6.4.3.2 MCC model

The numerical computations for the MCC model are carried-out for the two sets of the
parameters  and  (Table 6.3). As shown in Table 5.2, set 1 was obtained by means of
oedometer tests on the consolidated material before executing the centrifuge test (Küng,
2003), while set 2 was obtained by means of large diameter oedometer tests on the material
extracted from the model after the centrifuge tests (Weber, 2007). The reason for the
difference between the two parameters sets is not clear to the author and, as suggested by
Weber (2007), it would require additional experimental investigations.

154
 Chapter 6 – Centrifuge tests

200
Cohesion values c’ considered in the

ts [d]
study: 5.5, 7, 8, 8.5, 9, 9.2, 9.5 kPa
175

150

Ek by 30% lower than Table 6.3

125

100 Ek = 2.7 10-5 N/mms (Table 6.3)

75 Ek by 30% higher than Table 6.3

Experimental range

50

25

0
5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0
c' [kPa]

Reasonable range
Figure 6.21. MC model. Stand-up time as a function of the effective cohesion.

The higher values of κ and the lower value of λ of set 2 (compared to set 1) represent softer
elastic and stiffer plastic soil response, respectively. As the rate of pore pressure dissipation
increases with the stiffness of the material, the pore pressure will dissipate faster in the plastic
region (inside the shear band) and lower in the elastic one (ahead the shear band) when
assuming the material constants of set 2. As the main cause of shear band propagation is the
shear stress decrease caused by swelling inside the shear band, it is expected that the model
assuming material set 2 will collapse earlier than the one assuming the constants of set 1.
Selected computational results are presented in Figures 6.22 and 6.23 (analogously to Fig.
6.19 and 6.20, respectively). The time-development of settlement is qualitatively the same for
both parameter sets and agrees qualitatively well with the experimental results (Fig. 6.22).
Quantitative differences exist with respect to the initial instantaneous settlement and the
stand-up time. As expected according to the elastic stiffness of the material, the initial
settlement computed for set 2 is almost double of the one obtained for set 1. According to the
expectations, the stand-up time is higher for set 1 than for set 2 (410 days vs. 130 days). For
both material set the stand-up time is higher than the actual stand-up time (72 d at the
prototype scale). The settlement close to failure (about 100 mm for both parameter sets) is
about half the measured one (Fig. 6.22c). As in the MC simulations (and for the same
reasons), a wedge (without tension crack) develops at ultimate state (Fig. 6.22b) and the
sliding plane is less inclined than observed in the experiments (52.5° and 54° for set 1 and 2,
respectively, vs. 57° in the experiments).
In conclusion, for the assumed material parameters, the MCC model predicts the observed
settlement much better than the MC model. However, the stand-up time computed with the
MCC model is longer than the observed one.
As explained in Chapter 5, a first reason for the different stand-up time could be related the
shape of the MCC yield surface, which overestimates the resistance of the soil on the dry side
of the critical state line. A higher resistance means that a bigger stress change is required in
order to reach the yield surface, and successively to soften to the CSL. As the stress change in

155
the considered problem is mainly due to the swelling process, a higher resistance leads to a
longer swelling time. Another reason is the sensitivity of the stand-up time to the mesh size.
As it will show in Section 6.5, with finer meshes (that are able to map the development of a
thin shear band) the model predicts stand-up times very close to the actual ones.
As the soil at the ppt locations (points B and C) remains elastic immediately after unloading,
the excavation-induced instantaneous pore pressure drop is for both parameter sets close to
the one computed with the MC model. It is lower than the measured instantaneous negative
excess pore pressure for the reasons explained in Section 6.4.3.1. The predicted time-
development of the pore pressure is different from the observed time-development in that it
increases only at the beginning of the process; over the last two thirds of the process, pore
pressure decreases. The decrease in pore pressure is associated with the plastic dilation that
accompanies shearing and occurs over such a long period (much longer than observed)
because, as explained above, the MCC overestimates the size of the yield surface on the dry
side (and thus the amount of shearing and plastic dilation needed to reach the CSL). As
plastic dilation has a stabilizing effect, the overestimation of the plastic dilation results to an
overestimation of the stand-up time.
Summarizing, the MCC well captures qualitatively the deformation behaviour observed
experimentally. However, the pore pressure evolution (and thus of the settlement rate) is
affected by the simplified shape of the yield surface, which, for the considered problem
(unloading an over-consolidated soil) overestimates the resistance of the ground. This
inherently leads to an overestimation of the stand-up time.

156
 Chapter 6 – Centrifuge tests

(a) Experimental range

400

uy [mm]
ts ts
350

300

250

200

150
MCC (set 2;
100 κ = 0.022, λ = 0.152) MCC (set 1;
κ = 0.01, λ = 0.21)
50

0
0 50 100 150 200 250 300 350 400 450 500
t [d]

(b)

52.5 54

vector scale [mm]

0 400 600

t = 410 s t = 130 s
MCC (set 1) MCC (set 2)

(c)

500
uy - uy0+ [mm]

400

300

200

100

0
0 10 20 30 40 50 60 70 80 90
tprototype [d]
Figure 6.22. MCC model: (a) evolution over time of the vertical displacement at point A, (b),
displacement vectors at failure and, (c), monitoring results at prototype scale.

157
 (a)

60
51.5 kPa (hydrostatic pressure at lower ppts elevation)

p w [kPa]
40 C C
28 kPa (hydrostatic pressure at upper ppts elevation)

20
B B

0
0 100 200 300 400 500
t [d]
-20
B
C
-40
MCC (set 1; κ = 0.01, λ = 0.21)
MCC (set 2; κ = 0.022, λ = 0.152)
-60

(b)

60
p w [kPa]

C, lower ppts
40

20 B, upper ppts

0
0 20 40 60 80
t [d]
-20

-40

-60
Figure 6.23. MCC model: (a) evolution over time of the pore pressure at the ppts locations
(points B and C) and, (b), monitoring results at prototype scale.

158
 Chapter 6 – Centrifuge tests

6.5 Mesh dependency

As explained in Chapters 2, 3 and 4, the finer the mesh, the shorter the stand-up time. This is
confirmed also by a series of numerical simulations of the centrifuge test, which were
performed with different meshes. The mesh dependency is pronounced particularly for the
MCC model (Fig. 6.24). The finest considered discretization consists of square elements of
side length 0.20 m, which means that the vertical cut is 35 elements high. It is obvious that
selecting an element size in the order of few mm (the size of the shear band) is hardly feasible
as it would require an extremely long computational time (Fig. 6.25 exemplary shows the
non-linear increase of the computational time with decreasing element size). The stand-up
time for thin shear bands can be estimated by linear extrapolation (Fig. 6.24; see Section
4.3.3.4.4) and amounts to 9 – 11 days for the isochoric MC model (which is by factor 8 less
than observed) and to 43 – 80 days for the MCC model, which agrees reasonably well with
the observed stand-up times.
450

MCC (set 1; κ = 0.01, λ = 0.21)

400
ts [d]

350

300

MCC (set 2; κ = 0.022, λ = 0.152)

250

MC (c’ = 9 kPa)
200

150

100

80 d
Experimental range

50
43 d

11 d MC (c’ = 7 kPa)
9d
0
0.00 0.20 0.40 0.60 0.80 1.00 1.20
s [m]

very fine mesh fine mesh coarse mesh

(s/H = 3%) (s/H = 7%) (s/H = 14%)

linear extrapolation
to a zero mesh size
Figure 6.24. Stand-up time as function of the mesh size.

159
 (a)
18000

D.O .F [-]
16000

14000

12000

10000

8000

6000

4000

2000

0
(b)
600
CPU time [s]

500

400

300
MCC (set 1)

200

100
MC (8 kPa)

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
s [m]

very fine mesh fine mesh coarse mesh

(s/H = 3%) (s/H = 7%) (s/H = 14%)
Figure 6.25. (a) Degree of freedom as a function of the element size and, (b), total CPU time
required for two computations adopting MC and MCC models, respectively (64-
bit operating system, processor Intel® Xeon® CPU E3-1245 V2 3.4 GHz parallel
processing, RAM 16 GB).

160
 Chapter 6 – Centrifuge tests

6.6 Conclusions
Chapter 6 investigated experimentally and numerically the stand-up time of an underwater
vertical excavation in over-consolidated clay. Although the problem considered is simple, it
provides an illustration of features that are indeed relevant to many practical geotechnical
problems, such the face stability of tunnels or the behaviour of slope excavations in low-
permeability soils.
The experimental tests were executed in a drum centrifuge that allows to reproduce a
prototype height of the unsupported excavation of 7 m. The tests showed a very good
reproducibility, both with respect to settlement and pore pressure evolution.
These physical models showed that collapse occurs due to a combination of shear and tensile
failure. Failure manifests itself with rapid settlement acceleration. The average stand-up time
was 25 minutes at the model scale, which corresponds – according to the scaling laws – to 72
days at the prototype scale.
The numerical computations were carried-out by means of coupled FEM analyses. The
behaviour was studied for two constitutive models: the Modified Cam Clay model and the
linearly elastic perfectly plastic Mohr Coulomb model with isochoric plastic flow.
The material parameters used in the numerical analyses were mostly estimated or validated in
advance based upon laboratory tests or determined based upon theoretical considerations.
For the material parameters according to the literature, the MC model led to a stand-up time
generally shorter than the actual, and gives displacements that are much smaller than the
measured one. The MCC model generally led to an overestimation of the stand-up time, but
better reproduced the physical model behaviour. Taking into account the mesh sensitivity of
the results by extrapolating the stand-up time to the mesh size corresponding to the thickness
of the shear band the underestimation of the stand-up time increased for the MC model, while
the stand-up time computed with the MCC model got close to the experimental results. In
spite of the observed differences the simulation predictions were very good considering that
they were class A predictions.
In conclusion, Chapter 6 confirmed that the estimation of the stand-up time, although being a
highly complex phenomenon, can be approximated numerically by hydraulic-mechanical
coupled FEM analyses, even using a simple constitutive model (the stand-up time has the
same order of magnitude than the one observed experimentally). The numerical results
(obtained for the material parameters proposed in the literature), show that the selection of a
MC model with isochoric plastic flow generally leads to an underestimation of the stand-time
(particularly when extrapolating the results to the thickness of the shear band), while the
assumption of a MCC model leads to an overestimation of the stand-up time if the results are
not extrapolated to the size of the shear band.

161
162
 Chapter 7 – Stand-up time of tunnel face

7. Stand-up time of tunnel face

7.1 Introduction
The most serious hazards in soft ground tunnelling are associated with a collapse of the tunnel
face. In shallow tunnels the instability may propagate towards the surface, thereby creating a
crater. Experience shows that an unsupported tunnel face may remain stable for a certain
period after excavation. Under such conditions the question arises upon the stand-up time.
The assessment of the stand-up time is important for both mechanised and conventional
tunnelling, particularly under the economical point of view: a short stand-up time would
require the implementation of ground improvement or reinforcement or, in the case of
mechanized tunnelling, hyperbaric interventions, while a long stand-up time would allow to
refrain from such costly and time consuming measures.
Chapter 7 investigates numerically the time dependent stability of the tunnel face assuming
that all time effects are due to the excess pore pressure dissipation process in the ground. This
assumption is reasonable for shallow tunnels crossing water-bearing soils. The stand-up time
is estimated by means of coupled hydraulic-mechanical spatial analyses making use of the
findings of the fundamental investigations of the previous Chapters. The computational
results are presented in the form of easy to use dimensionless design charts for the assessment
of the face stability in shallow tunnelling. The diagrams cover a wide range of ground
parameters and allow to estimate quickly the stand-up time of the face, representing thus a
valuable decision-making aid in tunnel engineering practice.

7.2 Computational model

The ground is discretized by 8-node brick elements (Fig. 7.1). The element size varies from
0.5 m (close to the tunnel face) to 8 m (at the model boundary). The influence of spatial
discretization is investigated in Section 7.7.
The water table is taken equal to the elevation of the ground surface (Hw = H). The hydraulic
potential is fixed to its initial value at the far field boundaries as well as at the level of the
water table. The latter is true if there is no draw-down of the water table, which presupposes
sufficient ground water recharge. No-flow conditions are imposed at the symmetry plane as
well as at the tunnel wall, which is true for a practically impervious lining, i.e. a lining having
a permeability much lower than that of the soil (by a factor of at least 1000; Wongsaroj 2005).
Consider for example a shotcrete with permeability 10-12 m/s (Shin et al., 2005), the condition
of no-flow apply if the soil permeability is higher than 10-9 m/s (which corresponds to silty-
clay). For segmental lining with watertight joints or shotcrete combined with waterproof
membrane (double lining), the permeability of the support is such that the conditions of no-
flow apply for all type of soils.
In the case of a more permeable lining or of water table draw-down, the computational results
are on the safe side, because the hydraulic heads on the ground around the tunnel face are
tentatively lower.

163
 H = 20 m

H = 10 m D = 10 m

D = 10 m

30 m
30 m

46 m
35 m
30 m 40 m
20 m 30 m
H H
=1 =2
D D

H = 40 m

D = 10 m

40 m

60 m

H
=4
40 m 40 m D

Figure 7.1. Numerical models.

Mixed conditions are imposed at the tunnel face (i.e. no-flow in case of negative pore
pressure, and drainage-surface elsewhere). This condition allows the water to flow from the
ground toward the tunnel, but not vice-versa (which would be physically impossible). The
tunnel lining is simulated in a simplified way by fixing all nodal displacements at the
excavation boundary.
The nodes of the model located between one diameter below the tunnel and the bottom of the
model have been kept fixed in order to reduce the number of degrees of freedom of the
problem and thus the computational time.
The initial stress field corresponds to the overburden pressure at each point. The analyses
have been performed for two values of coefficient of lateral pressure (K0 = 0.5 and 1.0).

164
 Chapter 7 – Stand-up time of tunnel face

The ground is modelled as an isotropic, linearly elastic and perfectly plastic material obeying
the Mohr-Coulomb yield criterion with isochoric plastic behaviour without tension cut-off.
This assumption presupposes that the ground exhibits a tensile strength, and so the presence
of cementation bonds between the grains. Otherwise, tensile failure would occur in the form
of progressive ravelling of the ground at the face. In order to take into account the absence of
tensile strength, a tension cut-off has to be considered in the constitutive model. The influence
of the tension cut-off on the results is investigated in Section 7.5. The results show that a
tension cut-off prevents a distinct failure (characterized by an asymptotic increase in the
displacement rate). In the present Chapter, the stand-up time in a no-tension ground is taken
equal to the time at which tensile stresses start to develop in the model.
The reasons for selecting the MC model are: it is widely used in engineering practice; there is
considerable experience in practice about its parameters; it provides in general conservative
estimates of the stand-up time (see Chapters 5 and 6). Table 7.1 summarizes the parameters
considered in the analysis.

Table 7.1. Assumed material constant (in case of multiple values the number in
brackets is the assumed default parameter).

Saturated unit weight  [kN/m3] 20

Lateral pressure coefficient K0 [-] 0.5, (1.0)

Unit weight of water w [kN/m3] 10

Seepage flow parameters

Permeability k [m/s] 10-7
Water compressibility cw [MPa] 0

Parameters for linearly elastic, perfectly plastic material

Young’s modulus E' [MPa] 20
Poisson’s ratio  [-] 0.3
Cohesion c' [kPa] 5-40, (20)
Friction angle ' [°] 15, (25), 35
Dilation angle ψ [-] 0

165
7.3 Undrained and drained tunnel face stability

7.3.1 Numerical investigation

First the borderline cases of face stability under undrained and drained conditions will be
considered on the example of a 10 m diameter tunnel, 10 m deep. Since the goal of the later
Sections is to study the evolution of face stability over time, the ground parameters are
selected such that the unsupported tunnel face is stable under undrained conditions, but fails
under drained conditions.
Undrained stability is investigated by reducing practically instantaneously the total stresses
n normal to the face (hereafter referred to as "total face support pressure") from their initial
value n0 (horizontal in situ stress) to zero. Typical results are presented in Figures 7.2a and
7.3a. The plastic zone is limited to a small area in front of the tunnel face (Fig. 7.2a) and the
displacements stabilize to finite values even in the case of an unsupported tunnel face (Fig.
7.3a), thus indicating that the latter is stable in the short term.
As explained in Section 1.4, drained stability is analysed as follows. In a first step, a steady
state analysis is carried-out fixing the pore pressure at the tunnel face and the effective stress
normal to the face (hereafter referred to as "effective support pressure") equal to atmospheric
and to the initial horizontal effective stress 'n0, respectively. Subsequently, the effective face
support pressure is reduced stepwise, while the pore pressure field is fixed to the steady-state
field obtained in the first step. The results are presented in Figures 7.2b and 7.3b. The
depressurization of the pore water in the first step causes consolidation of the ground, face
extrusion and surface settlement although the effective face support pressure is kept fixed to
its initial value (Fig. 7.3b; lines for points A and B; displacements at 'n/'n0 =100%). With
decreasing effective face support pressure 'n, the displacements increase and the plastic zone
ahead of the face becomes larger (Figure 7.2b). Contrary to the undrained analysis, the
displacement of the face (point A) does not reach a finite value, but increases asymptotically
to infinity when the effective support pressure 'n approaches a critical value of about
25%'n0 (Fig. 7.3b; state marked by "(iii)"). This indicates that the system has reached its
ultimate state; re-distribution of the effective stresses is no longer possible and the face
becomes unstable. It is noteworthy that the plastic zone at the ultimate state does not reach the
ground surface (Fig. 7.2b, state (iii)). A localized failure ahead of the face presupposes the
development of tensile stresses above the unstable region of material. The latter develop in
the model because of the adoption of the MC model without tension cut-off.

166
 Chapter 7 – Stand-up time of tunnel face

(a)
Ground surface B
8 mm c' = 20 kPa
' = 25
K0 = 1.0
H = D = 10 m Other parameters
according to Table 7.1

Undrained analysis,
unsupported face

85 mm A
D
5 x [m]

(b)
Ground surface B

21 mm (iii)

H = D = 10 m
Drained analysis
(State (iii) of Fig. 7.3)

A
D
5 x [m]
204 mm (iii)

Drained analysis
(State (ii) of Fig. 7.3)
Figure 7.2. Boundary of the plastic zone and displacement vectors of the points A and B,
(a), at the end of the undrained excavation and, (b), short before failure under
drained conditions.

167
 (a)
100

ux [mm]
c' = 20 kPa
' = 25
K0 = 1.0
80 Other parameters
according to Table 7.1
u A,x
B
60

z
A x
40

20

u B,x
0 u A,z
100%
uz [mm]

0% 50%
u B,z
 n/ n0 [-]

-20

(b)
200 (iii) Ultimate state
ux [mm]

150

100
u A,x

(ii)

50 (i)

u B,x
0
0% 50% 100%
uz [mm]

u B,z
 'n / 'n0 [-]
u A,z
-50

Figure 7.3. Displacement of points A and B as a function of the normalized total face
support pressure or the normalized effective face support pressure for, (a)
undrained analysis and, (b), drained analysis, respectively.

168
 Chapter 7 – Stand-up time of tunnel face

7.3.2 Comparison with analytical solutions

The prediction of stable undrained conditions agrees with the results obtained by applying the
static solution of Davis et al. (1980), according to which the required face support pressure

ps   v,0 ( z  H  0.5D)  Ns su , (7.1)

where v,0 is the mean in situ total stress at the tunnel axis (= 300 kPa in the present example),
Ns is the so-called stability number (=4.5 for H/D = 1; Davis et al. 1980) and su is the
undrained shear strength. For a uniform undrained shear strength su of 82 kPa (after Eq. (3.2)
of Section 3, taking into account the mean effective stress at the tunnel axis elevation), the
required support pressure is negative (-69 kPa), which means that the face is stable without
support.
The stability of the tunnel face under drained conditions is investigated after Perazzelli et al.
(2014), according to which the required support pressure

ps  P1  w h  P2  ' H  P3 c ' , (7.2)

where ∆h denotes the difference in hydraulic head between water table and tunnel axis (15 m
in the present example), γ' (= 10 kN/m3) is the submerged unit weight of the soil, c' is the
effective cohesion and the coefficients P1, P2, P3 are functions of the friction angle and of the
inclination of the sliding wedge (determined such that it maximizes ps).
For the strength parameters of the present problem (c' = 20 kPa and ' = 25°), P1 = 0.47, P2 =
0.11, P3 = 2.07 and the required effective support pressure is equal to to 37.9 kPa, which is
very close to the numerically predicted limit pressure (25% x 15 m x 10 kN/m3 = 37.5 kPa).
The required support pressure was obtained for a critical inclination of the sliding wedge of
36°.
The critical cohesion, i.e. the cohesion required for drained stability of an unsupported face,
can be obtained from Eq. (7.2) by setting the support pressure equal to zero,

P1  w h  P2  ' H
c 'D ,lim  , (7.3)
P3

and is equal to 38 kPa in the present case.

169
7.4 Tunnel face stability under transient conditions
In the example of Section 7.3 the unsupported tunnel face is stable in short-term (undrained
conditions) but unstable in long-term (drained conditions). Next, the evolution of face
stability over time will be studied and the stand-up time ts of the tunnel face will be estimated.
The time-development of face stability is simulated by means of a transient analysis
following the undrained excavation (as described in the previous Section). The assumption of
undrained conditions – prevailing at the beginning of the transient analysis – represents a
reasonable assumption in the following cases: (i) rapid advance rate compared to the soil
permeability (ratio between advance rate and ground permeability of about 5’000 – 10’000
for a tunnel of 10 m diameter; Anagnostou 1995 – e.g. for a clayey soil with permeability 10-9
m/s, undrained conditions would prevail around the tunnel face, if the advance rate is higher
than 0.5–1 m/day); (ii) closed shield tunnelling with full compensation of the initial stresses
(i.e., where the stress and hydraulic fields around the tunnel during advance correspond to the
in situ one) followed by rapid depressurization of the face (a rather theoretical case). The
influence of the advance rate on the face stability is discussed in Section 7.10.
As explained in Chapter 3, stand-up time is estimated co-evaluating the time-development of
the stresses and pore pressures (Figs. 7.4 and 7.7), of the plastic zone (Fig. 7.5a), of the
equivalent plastic and total volumetric strain (Fig. 7.5b and Fig. 7.6b, respectively) and of the
displacements of selected points (7.6a).
In spite of the excavation-induced unloading of the tunnel face, the axial effective stress 'x at
the face is compressive (Fig. 7.4b), which is favourable for stability. The reason is that
negative excess pore pressures develop due to the excavation (Fig. 7.4a). As the horizontal
effective stress practically coincides with the minor principal stress, the effective stresses in
the axial direction govern the magnitude of the effective stresses in the vertical and in the
transverse directions (Fig. 7.4c and 7.4d), which explains the similarity in the distribution of
'z and 'y.
During the dissipation of the negative excess pore pressures (Fig. 7.4a) the effective stresses
(and thus also the shear resistance of the ground) decrease (Figs. 7.4b, c and d). A plastic
zone develops ahead of the tunnel face, progresses upwards and reaches the ground surface at
about t = 11.3 h (Fig. 7.5a). The shape of the failure zone agrees reasonably well with the
wedge and prism mechanism that is often assumed by limit equilibrium analyses (e.g.
Anagnostou and Kovári, 1994), nevertheless with a curved slip surface. This is visible also
from the contour lines of the total equivalent plastic strain eq,pl (Fig. 7.5b). The latter is
defined after Berg (1972),

t
2 p p
 eq , pl    ij  ij dt , (7.4)
0
3

and provides an overall measure of the shearing of the ground. The shape of the contour lines
indicates an average inclination of the sliding surface (assumed as the surface with the
maximum equivalent plastic strain, i.e. of the most intense shearing) of about 39° (which is
only 3° higher than the angle predicted by the limit equilibrium analysis of Section 7.3.2).

170
 Chapter 7 – Stand-up time of tunnel face

It interesting to note that at the ultimate state the ground remain elastic close to the tunnel
face (around point A) as well as close to the ground surface above the central part of the
failure mechanism (around point C); shearing localizes along the boundaries of a wedge and
prism failure mechanism. The ground remains elastic around point A due to the negative
excess pore pressures that developed in the proximity of the tunnel face during undrained
unloading but have not dissipated yet (thus keeping the effective stresses sufficiently high)
and because of the small amount of shear deformation. Point B is located in the elastic region
delimited by the shear bands, which – at the ultimate the failure – develop from the tunnel
face up to the ground surface.

(a) 200 (b)

c' = 20 kPa
' = 25 0 5 10 15
150 K0 = 1.0 x [m]
-30
Other parameters
according to
100 Table 7.1

-80
50

0
-130
0 5 10 15
x [m]
-50 z
t = 00++ t1
POR
POR t <htst3(t = 2.5
0 + <5.25 h) -180
 'x [kPa]
p w [kPa]

x
-100 t = t11.3
POR s h t5 Tunnel
face

-150
-230

(c) (d)
0 5 10 15 0 5 10 15
x [m] x [m]
-30 -30

-80 -80

-130 -130

-180 -180
 'y [kPa]
 'z [kPa]

-230 -230

Figure 7.4. Spatial distribution (along the tunnel axis ahead of the face) of: (a) pore
pressure pw (nodal values); (b) effective horizontal stress 'x; (c) effective
vertical stress 'z; (d) effective out of plane stress 'y (effective stress values
averaged at the element centroid).

171
Failure occurs at about t = 11.3 hrs. The latter is indicated by (Section 3.3.3.2):
 the time-development of displacements, which start to increase rapidly at t = 11.3 h
and tend to infinity at about t = 12.3 h (Fig. 7.6a);
 the extent of the plastic zone (it reaches the ground surface at t = 11.3 hrs, Fig. 7.5a);
 the almost constant rate of the volumetric strain and the effective stresses (observed
particularly in points D, C and A' at about 11.3 hours, see Fig. 7.6b and 7.7,
respectively);
 numerical convergence problems shortly before collapse (for t > 11.3 hrs, the
volumetric strains at point D as well the horizontal effective stress 'y and the vertical
stress 'z decrease instead of remaining constant; Fig. 7.6b and 7.7).

Elastic region
(a) B
Ground surface
153 mm
c' = 20 kPa
C ' = 25
t [h] = 11.3 11.3 158 mm K0 = 1.0
7.7 Other parameters
H= D 5.3 according to Table 7.1

2.3
0+

Elastic z
region 0+
A D
D = 10 m x [m]
5 10
492 mm 334 mm

(b)
Ground surface

H= D

A A'
D = 10 m x [m]
5 10

max value eq,pl = 1.1

Figure 7.5. Contour of, (a), the plastic zone at different time and displacement vectors of
points A, B, C and D at t=ts=11.3 h (ultimate state) and, (b), the total equivalent
plastic strain eq,pl at failure. The dashed lines indicate the shape of the failure
mechanism.

172
 Chapter 7 – Stand-up time of tunnel face

(a)
4

ux [m]
C
Nearly ultimate state u A,x
3
z
D
2 A x
u D,x

1
u C,x c' = 20 kPa
0 ' = 25
0+ 5 10 t [h] K0 = 1.0
-1 Other parameters
according to Table 7.1

-2

-3
uz [m]

u D,z
u C,z
-4 u A,z
(b)
0.8%
ε vol [-]

A
0.6%
D
0.4%

0.2%

ts
+
-0.1% 0 5
C
10
t [h]

Figure 7.6. (a) Evolution of the displacements (ux, uz) and of (b) the volumetric strain (vol)
at points A, C and D over time. The dashed vertical line crossing the time axis at
t = ts = 11.3 h corresponds to the near ultimate state.

t [h]
0
ts c' = 20 kPa
0+ 5 10
' = 25
K0 = 1.0
Other parameters
-40
according to Table 7.1
'x
p z
-80 'z A'
x
'y

-120

Initial value
', p w [kPa]

-160
Nearly ultimatestate

-200
Figure 7.7. Evolution of the effective stresses ('x, 'y, 'z) and of the pore pressure (pw) at
point A over time. The dashed vertical line crossing the time axis at t=ts=11.3 h
indicates the near ultimate state.

173
7.5 On the transient tensile failure of the ground
High hydraulic head gradients ahead of the tunnel face may induce tensile stresses in the
ground. The latter are statically admissible only if the ground exhibits a tensile strength,
which presupposes the presence of cementation bonds between the grains. Otherwise, tensile
failure would occur in the form of progressive ravelling of the ground at the face. Under the
assumption of the MC-failure criterion the tensile strength of the ground is a function of the
friction angle and of the cohesion (see Fig. 7.8a). In order to take into account the absence of
tensile strength, a tension cut-off has to be considered in the constitutive model (see Fig.
7.8b).
(a) without tension cut-off (b) with tension cut-off

 

 
ft ft = 0
Figure 7.8. MC-yield surface, (a), without and, (b), with tension cut-off.

1
ux [m]

MC without tension
H = D = 10 m cut-off
c' = 35 kPa
0.8 = 25
ux
K0 = 1.0

0.6

(a)

0.4

0.2 (b) MC with tension

cut-off (ft = 0)
constant
displacement rate
0
0 25 50 75 100 125 150
t [h]
Figure 7.9. Time-development of the horizontal extrusion of the tunnel face ux computed
without and with tension cut-off in the constitutive model (other parameters
according to Table 7.1).

174
 Chapter 7 – Stand-up time of tunnel face

without tension cut-off with tension cut-off

(a) (d)
Ground surface Ground surface

shear failure
shear failure
tensile failure
T unnel face
x x

(b) 150 (e) 150

p W [kPa] p W [kPa]

100 100

50 50

0 0
0 5 10 15 20 0 5 10 15 20
x [m] x [m]

-50 -50

(c) 100 (f) 100

 'x ,  'max  'x ,  'max
[kPa] [kPa]
 'max 50
50
 'x
 'max
0 0
0 5 10 15 20 0 'x 5 10 15 20
x [m] x [m]

-50 -50

-100 -100

-150 -150

Figure 7.10. (a), (d) Plastic zone, (b), (e) spatial distribution of the pore pressure and (c), (f) )
spatial distribution of the maximum and horizontal effective stresses 70 hours
after undrained excavation computed (a) without and (b) with tension cut-off in
the constitutive model, respectively (' = 25°, c' = 35 kPa, K0 = 1.0, H/D = 1,
other parameters according to Table 7.1).

175
This section investigates numerically the influence of tension cut-off on the manifestation of
failure. With the exception of tension cut-off (which is considered in the constitutive model
after Clausen et al., 2005) and of cohesion (which is taken here equal to 35 kPa in order that
tensile rather than shear failure occurs), computational model and parameters are the same as
in Section 7.4.
Without tension cut-off, failure occurs at about t = 70 h and is characterized – as in all
numerical computations with isochoric plastic flow until now – by a rapidly accelerating face
extrusion (Fig. 7.9, upper line) and a fully developed plastic zone up to the soil surface (Fig.
7.10a). Due to the distribution of the pore pressure ahead of the tunnel face (Fig. 7.10b) close
to failure, seepage forces act towards the tunnel face, thus necessitating tensile effective
stresses (Fig. 7.10c) in order to fulfil equilibrium in the horizontal direction.
With tension cut-off, however, the model behaviour is qualitatively different with respect
both to the time-development of the face extrusion and to the shape of the failure zone: after a
certain period of time (about at t = 30 h) the displacement rate becomes constant rather than
accelerating (Fig. 7.9, lower line). Within about 70 hours after excavation, a tensile failure
zone has developed ahead of the tunnel face (Fig. 7.10d); it is characterized by zero
maximum principal effective stress’max (Fig. 7.10f) and enclosed by a shear failure zone,
which – differently from Fig 7.10a – does not extend to the overburden. In addition, the pore
pressure distribution (Fig. 7.10e) is slightly different than in Fig. 7.10b, with negative values
close to the tunnel face. The difference between the results with and without tension cut-off
(which is particularly evident in the time-development of the displacements, Fig. 7.9) is due
to the irreversible volumetric strains accompanying tensile failure (separation). Their effect is
the same as that of plastic dilation in the case of shear failure described in Section 3.3.2, the
difference being that, in the present case, the volumetric strains are due to the associativity of
the flow rule at tensile failure (Fig. 7.8b).
Physically, the occurrence of tensile failure corresponds to the formation of tensile cracks.
The latter, would (on account of the negative pore pressure, Fig. 7.10e) desaturate as soon as
they reach a free surface (the tunnel face); the stabilizing negative pore pressure would then
drop to zero and the face would collapse.
As desaturation is not taken into account and all computations of the present Chapter are
carried-out with the MC model without tension cut-off (Fig. 7.8a), the stand-up time in a no-
tension ground will be taken equal to the time at which tensile stresses start to develop in the
model.

176
 Chapter 7 – Stand-up time of tunnel face

7.6 On some factors influencing stand-up time

7.6.1 Cohesion

The stand-up time depends sensitively on the cohesion of the ground; in the example of Fig.
7.11, the tunnel face will collapse shortly after undrained excavation for a cohesion of 5 kPa
and will remain stable even at long term (i.e., under drained conditions) for a cohesion higher
than about 37 kPa, which agrees well with the limit equilibrium prediction after Perazzelli et
al. (2014) (see Section 7.3.2).
Figure 7.12 shows from top to down the boundary of the plastic zone and the distribution of
the horizontal effective stresses 'x and of the pore pressure pw along the horizontal axis ahead
of the tunnel face immediately after the undrained excavation (on the l.h.s.) and at the
ultimate state (on the r.h.s.) for three cohesion values.
After undrained excavation (Fig. 7.12a), the plastic zone extends to the surface for c' = 5 kPa,
develops around the entire tunnel face for c' = 20 kPa and is limited to the lower and upper
part of the face if c' = 35 kPa. At failure the shape of the plastic zone is similar for all the
considered cohesion values (Fig. 7.12b).
For a cohesion of c' = 5 kPa, failure occurs shortly after the beginning of the swelling process.
Therefore, the effective stresses and the pore pressures at failure are quite close to the
undrained values and far away from the steady state distribution (Fig. 7.12d and f).
For c' = 20 kPa, the pore pressure ahead of the tunnel face remains negative at ultimate state
(Fig. 7.12f), but close to the tunnel face is higher than after undrained excavation.

70
ts [h]

H=H w= 10 m

60 D = 10 m

’ = 25
50 K0 = 1
k = 10 -7 m/s
E = 20 Mpa
40  = 20 kN/m3

30 Face stable
under drained
conditions
20

10

0
5 10 15 20 25
35 30
40
c' [kPa]
Figure 7.11. Stand-up time of the tunnel face as a function of the cohesion c'.

177
 (a) (b)
t = 0+ t = ts

Ground surface Ground surface

c' [kPa] =5 5
(b) 20

H = D = 10 m c' [kPa] =35

20

z 35 z

D
5 10 15 x [m] 5 10 15 x [m]

(c) (d)
90 90
'x [kPa]

'x [kPa]
40 40

-10 0 5 10 15 -10 0 5 10 15
x [m] x [m]

-60 -60

-110 -110

Initial state Initial state

-160 -160

(e) Initial state (f)

Steady state distribution
p w [kPa]

p w [kPa]

140 140

90 90

40 40

-10 -10
0 5 10 15 0 5 10 15
x [m] x [m]
-60 -60
c' = 5 kPa ' = 25
-110 -110 K0 = 1.0
c' = 20 kPa
Other parameters according
c' = 35 kPa to Table 7.1
-160 -160

Figure 7.12. Boundary of the plastic zone (a) immediately after undrained excavation and (b)
at ultimate state for c' = 5 kPa, 20 kPa and 35 kPa, and spatial distributions
along the x-axis immediately after undrained excavation and at ultimate state of
the effective horizontal stress 'x (diagrams (c) and (d), respectively) and of the
pore pressure pw (diagrams (e) and (f), respectively) for c' = 5 kPa, 20 kPa and
35 kPa.

178
 Chapter 7 – Stand-up time of tunnel face

70

ts [h]
H=H w= 10 m
60 D = 10 m

’ = 25
50 K0 = 1
k = 10 -7 m/s
40 E = 20 Mpa
 = 20 kN/m3

30

20
occurrence of tensile stress

10

0
5 10 15 20 25 30 35 40
c' [kPa]
Figure 7.13. Stand-up time of the tunnel face as a function of the cohesion c' for a ground
with (solid line) and without (dashed line) tensile strength.

For c' = 35 kPa, tensile stresses develop ahead the tunnel face close to failure (Fig. 7.12d).
The development of tensile stresses is due to the pore pressure distribution (Fig. 7.12f), which
during the dissipation process approaches the steady state distribution and, therefore, causes
seepage forces directed towards the face. In the considered example tensile stresses ahead of
the tunnel face appear only if the cohesion is higher than about 20 kPa (Fig. 7.13). This
happens – independently on the cohesion – at approximately 15 hours. As mentioned in
Section 7.5, this time is taken as stand-up time for no-tension soils.

7.6.2 Friction angle

Stand-up time increases with friction angle provided that cohesion is less than 30 kPa (Fig.
7.14). For higher cohesion values, however, stand-up time decreases with increasing friction
angle. The reason is that tensile strength (which is important due to the development of tensile
stresses) decreases with the friction angle (the value of cohesion being fixed; Fig. 7.15).
For the three values of friction angle, tensile stresses appear in the model about 10 to 15 hours
after excavation. This time corresponds, according to the definition given in Section 7.5, to
the stand-up time for a non-tension soil.
Furthermore, the critical cohesion (i.e. the cohesion for which stand-up time becomes infinite)
practically does not depend on friction angle (Fig. 7.14).
Finally, the lower the friction angle, the more extended the failure zone both short-term and at
failure (Fig. 7.16).

179
 70

ts [h]
H=H w= 10 m

60 D = 10 m

K0 = 1
50 k = 10 -7 m/s
E = 20 Mpa
 = 20 kN/m3
40

30

20

10 ' = 35 25
15
0
5 10 15 20 25 30 35 40
c‘ [kPa]
Figure 7.14. Stand-up time of the tunnel face as a function of the cohesion c' for a friction
angle ' = 15°, 25° and 35°.


 '2 > '1

 '1

c'1 = c'2

 '1  '2
ft1 ft2 

Figure 7.15. Tensile strength ft decreasing with increasing friction angle φ'
(the cohesion c' being fixed).

t = 0+ t = ts

Ground surface Ground surface

H = D = 10 m ' =15

' = 35

25
25
15
z 35 z

D
x [m] x [m]
5 10 5 10

c' = 20 kPa
K0 = 1.0
Other parameters
(a) (b)
according to Table 7.1

Figure 7.16. Boundary of the plastic zone, (a), immediately after undrained excavation and,
(b), at ultimate state for ' = 15°, 25° and 35°.

180
 Chapter 7 – Stand-up time of tunnel face

7.6.3 In situ horizontal effective stress

The coefficient of lateral pressure K0 does not have any influence on stability in the
uncoupled problem (Vermeer and Ruse, 2001), but stand-up time increases with in situ
horizontal effective stress (Fig. 7.17).
For K0 = 0.5, the short-term plastic zone (i.e., the one developing under undrained conditions)
is more extended (ahead of the face and toward the surface) than for K0 = 1.0 (Fig. 7.18a).
This is because under undrained conditions the shear resistance of the ground depends
essentially on the mean initial effective stress (Broms and Bennermark, 1967), which is
higher in the case of K0 = 1. Consequently, in spite of the smaller stress change caused by the
excavation in the case of a low coefficient of lateral pressure, the latter is unfavourable with
respect to the time-development of stability. This is evident considering the stress paths at a
point ahead of the face (Fig. 7.19). As the initial effective stresses are closer to the MC-line in
the case of the low K0, the effective stress change that is needed to reach yielding condition is
smaller, necessitates less excess pore pressure dissipation and, consequently, yielding starts
earlier. At failure, however, the extent of the plastic zone is practically independent from K0
(Fig. 7.18b).

70
ts [h]

H=H w= 10 m

60 D = 10 m

’ = 25
50 k = 10 -7 m/s
E = 20 Mpa
 = 20 kN/m3
40

30
K0= 1
20 0.5

10

0
5 10 15 20 25 30 35 40
c‘ [kPa]
Figure 7.17. Stand-up time of the tunnel face as a function of the cohesion c' for
a coefficient of lateral pressure K0 of 0.5 and 1.0.

181
 t = 0+ t = ts

Ground surface Ground surface

H = D = 10 m
K0 = 1
0.5

K0 = 1
z z
0.5

D
5 10 x [m] 5 10 15 x [m]

c' = 20 kPa
' = 25
Other parameters
(a) (b) according to Table 7.1

Figure 7.18. Boundary of the plastic zone, (a), immediately after undrained excavation and,
(b), at ultimate state for K0 of 0.5 and 1.0.

-250 c' = 20 kPa

'1 [kPa]

' = 25
Other parameters
according to Table 7.1
-200
z
E
x
-150
Initial state Initial state
(K0 = 0.5) (K0 = 1)
ts

-100
ts

-50
0 -50 -100 -150
'3 [kPa]
Figure 7.19. Stress paths in the principal stress plane at point E (see inset in the figure) for
K0 of 0.5 and 1.0.

182
 Chapter 7 – Stand-up time of tunnel face

7.6.4 Depth of cover

The stand-up time increases with overburden in the case of soils exhibiting low to medium
cohesion (c' < about 30 kPa, Fig. 7.20). This is due to the longer drainage path and to the
longer propagation distance of the plastic zone. At higher cohesion values (c' > about 30 kPa),
however, these effects are overweighted by the effect of the (destabilizing) pore pressure
gradients (which increase with the overburden; Fig. 7.21a and 7.21b) so that the stand-up time
decreases.
It should be noted that numerical problems did occur in the case the highest considered
overburden and for a friction angle of 35°. The numerical analyses aborted long before
reaching ultimate state. This is probably due to the combination of high seepage forces and
low tensile strength.
Figure 7.22 shows the boundary of the plastic zone after undrained excavation (continuous
lines) and at failure (dashed lines) for three depths of cover; the extent of the plastic zone
around the tunnel at t = 0+ does not change much with depth. At the ultimate state, however,
the extent of the plastic zone ahead of the face increases with depth.

70 H=D
ts [h]

D = 10 m

60

50
H = 2D

H/D = 4 D
40

30
2
1
20 H = 4D
' = 25
K0 = 1
10 k = 10 -7 m/s D
E = 20 Mpa
 = 20 kN/m 3
0
5 10 15 20 25 30 35 40
c' [kPa]
Figure 7.20. Stand-up time of the tunnel face as a function of the cohesion c' for a
depth of cover H/D = 1, 2 and 4.

183
 c' = 20 kPa c' = 35 kPa

p w [kPa]

p w [kPa]
450 450
z

x
350 350
4
4
250 250

2 2
150 150
H/D = 1
t = ts ' = 25 Other parameters
50 H/D = 1 50 K0 = 1.0 according to
Steady state Table 7.1
distribution

0 10 20 30 0 10 20 30
-50 x [m] -50 x [m]

(a) (b)
Figure 7.21. Spatial distributions along the x-axis (see inset in diagram (a)) of the pore
pressure p at ultimate state for a ground cohesion c' of 20 kPa and 35 kPa.

(a) (b) (c)

Ground surface Ground surface Ground surface

ts
0+

z ts
ts
x [m]
5 10 15
0+

z
x [m]
5 10 15

0+
c' = 20 kPa z
' = 25
K0 = 1.0 x [m]
Other parameters 5 10 15 20 25
according to Table 7.1

Figure 7.22. Boundary of the plastic zone immediately after undrained excavation and at
ultimate state for H/D = 1, 2 and 4.

184
 Chapter 7 – Stand-up time of tunnel face

7.7 Design diagrams

A parametric study was performed with the aim to prepare deign charts for the estimation of
the stand-up time of the tunnel face. The latter depends, in general, on the geometric
parameters of the problem (D, Hw, H), the material constants of the ground (submerged unit
weight ', Young’s modulus E, Poisson’s ratio , ground cohesion c', friction angle ',
dilation angle  and coefficient of permeability k), of the unit weight of water w and the
coefficient of lateral pressure K0:

ts  f  E, , c ', ', ,  ',  w ,k, K0 ,D, H , H w  . (7.5)

For dimensional reasons and due to the structure of the equations underlying consolidation
theory, a dimensionless stand-up time Ts can be defined (cf. Eisenstein and Samarasekera,
1992) and Eq. (7.5) can be written in normalized form as follows:

ts E k  c' w H H 
Ts   f , , ', , K 0 , , ,  . (7.6)
 'D 2
  'D  ' Hw D 

The l.h.s. term of Eq. (7.6) shows that if the permeability or the Young’s modulus is higher by
a factor of ten, then the stand-up time will be ten times shorter.
For working-out a manageable number of design charts, the following values are kept fixed:
 = 0, 0, Hw/H = 1, w/' = 1. Consequently, Eq. (7.6) reduces to

ts E k  c' H
Ts  f , ', K0 ,  . (7.7)
 'D 2
  'D D

On account of the expected mesh dependency (see Chapter 4), the stand-up time was
additionally computed for a spatial discretization with 1 and 2 m big elements. As presented
in Chapter 4 and as validated in Chapter 6, the obtained results allow to linearly extrapolate
the results to a mesh size of zero (remark that the assumption is reasonable considering that
the thickness of the shear band, for soil of low permeability, amounts to few millimetres,
only).

185
 K0 = 1 K0 = 0.5

60 60

ts [h]

ts [h]
50 50

40 40
35
c' [kPa]
φ' = 15 30 = 30 30

20 20

10 25 10 30
20
25
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
s [m] s [m]

60 60
ts [h]

ts [h]
50 50

30
40 40
35
φ' = 25 30 30

25
20 20 30

20
10 15 10 25
10 20
5 15
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
s [m] s [m]

60 60
35
ts [h]

ts [h]

50 50
30
40 40

φ' = 35 35
30 30
25
30
20 20 20
15 25
10 10 10
5 20
15
10
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
s [m] s [m]

Figure 7.23. Stand-up time as a function of the mesh size for H/D = 1
(= 0°,  = 0.3, Hw / H = 1, w / '=1).

186
 Chapter 7 – Stand-up time of tunnel face

K0 = 1 K0 = 0.5

60 60
c' [kPa]

ts [h]

ts [h]
= 40
50 50

40 40
40
φ' = 15 30 30
30

20 20
25
10 20 10
30
15
10 25
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
s [m] s [m]

60 60
ts [h]

ts [h]
40

50 50

30 40
40 40

25
φ' = 25 30 30
20
20 20
15 30
10
25
10 5 10
20
15
10
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
s [m] s [m]

60 60
ts [h]

ts [h]

40
50 50
30
40 40 40
25

φ' = 35 30 20 30
15
30
20 10 20
5 25
20
10 10 15
10
5
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
s [m] s [m]

Figure 7.24. Stand-up time as a function of the mesh size for H/D = 2
(= 0°,  = 0.3, Hw / H = 1, w / '=1).

187
 K0 = 1 K0 = 0.5

100 60
c' [kPa]

ts [h]

ts [h]
90 = 35
50
80
30 40
70
25 40
35
60
20
φ' = 15 50 30 30
15
40 10 25
5 20 20
30
15
20 10
10 5
10

0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
s [m] s [m]

100 60
ts [h]

ts [h]
90
40
50
80

70
40
60 35

φ' = 25 50 30
30
40 40
25 20
30
20 35
20 15
10 30
10
10 5 25
0 0 20
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
s [m] s [m]

Figure 7.25. Stand-up time as a function of the mesh size for H/D = 4
(= 0°,  = 0.3, Hw / H = 1, w / '=1).

Figures 7.23 to 7.25 present the stand-up time ts as a function of element size for H/D = 1, 2
and 4, respectively. As expected based upon the previous Chapters, the stand-up time
increases with increasing element size. The relationship between mesh size and stand-up time
is almost linear with the exception of the results obtained for H/D = 1 and cohesion values
higher than about 20 – 30 kPa. In the latter case, the mesh dependency of stand-up time is
strongly non-linear; for c' = 35 kPa, for example, the stand-up time is finite for the finer mesh
(0.5 m), but becomes infinite (meaning stable face under drained conditions) for the coarser
meshes. This model behaviour was already observed on the example of an underwater vertical
cut in Section 4.2.3 for cohesion values close to the critical one (leading to drained stability).
Callari (2015) also observed that there are combinations of strength parameters for which the
tunnel face collapses or not depending on the size of the spatial discretization. Consider for
example the case H/D = 1 and φ' = 25°. As shown in Section 7.3.2, the theoretical cohesion
required for drained stability amounts to about 38 kPa. When the cohesion approaches this
value the stand-up time increases asymptotically to infinity. Consequently, a small change of
cohesion leads to a large change of stand-up time. As the mobilized shear strength depends on
the element size, a change of element size strongly influences the stand-up time. In the
extreme case, the mobilized strength parameters are such that the tunnel face remains stable
under drained conditions for large elements (not allowing the rotation of the principal axes,

188
 Chapter 7 – Stand-up time of tunnel face

and so structural softening), but fails when choosing a finer discretization (which allows the
rotation of the principal axes, and so structural softening).
The numerical results of Fig. 7.23 to 7.25 were post-processed in order to obtain design
diagrams giving the normalized stand-up time Ts as a function of the normalized cohesion
(Fig. 7.26 and 7.27). Figure 7.26 is based upon the finite element computations with the finer
mesh (s/D = 0.05) and partially revises the results presented in Schuerch et al. (2016) because
it considers additional computations carried-out during this PhD research. On one hand side,
the design diagrams of Figure 7.26 overestimate the stand-up time because they are obtained
for a mesh size which leads to a shear band thickness larger than the expected one. On the
other hand, they underestimate the stand-up time because, they do not consider the plastic
dilation, which as shown in Chapters 5 and 6, delays the occurrence of failure. As these two
effects are opposite, the design diagrams of Figure 7.26 give a stand-up time which may be
close to the real one.
Figure 7.27 was obtained by a linear extrapolation of the numerical results for 0.5 m and 1.0
m to an element size of zero (dashed lines in Figs 7.23 – 7.25). The extrapolation has been
performed only for results exhibiting an almost linear relationship between stand-up time and
mesh size. The design nomograms of Fig. 7.27 give a lower limit of the stand-up time – thus
representing a conservative version of the charts of Figure 7.26 – because, (i), they do not
consider the plastic dilation, and, (ii), assume a very thin thickness of the shear band.
Generally, considering the uncertainties related to the ground parameters (particularly that of
the permeability), the influence of the mesh size on the stand-up time (illustrated by the
differences between Figs. 7.26 and 7.27) is not relevant from the practical engineering point
of view. Indeed, the mesh-dependency becomes relevant, when the strength is close to the
critical cohesion for drained face stability.

189
 K0 = 1 K0 = 0.5
(a) (b)

 -
0.50 0.50

 -
H=H w=D H=H w=D

ts k E
ts k E

 D 2
 D 2
0.45 0.45
D D
0.40 0.40
H/H w = 1 H/H w = 1
0.35  w/  ’ = 1 0.35  w/  ’ = 1
= 0 H/D = 1 = 0 H/D = 1
0.30 = 0.3 K0 = 1 0.30 = 0.3 K0 = 0.5

0.25 0.25
H/D = 1
0.20 0.20

0.15 occurrence of 0.15

tensile stress occurrence of
0.10 0.10 tensile stress
' = 35 25
0.05 0.05
15 ' = 35 25 15
0.00 0.00
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
c  - 
c  - 
γ D γ D  
(c) (d)
 -

0.50 0.50

 -
ts k E

ts k E
 D 2

 D 2
0.45 H=H w=2D 0.45 H=H w=2D

0.40 0.40
D D

0.35 0.35
H/H w = 1 H/H w = 1
0.30  w/  ’ = 1 0.30  w/  ’ = 1
= 0 H/D = 2 = 0 H/D = 2
= 0.3 K0 = 1 = 0.3 K0 = 0.5
0.25 0.25
H/D = 2
0.20 0.20
' = 35 25
0.15 15 0.15 occurrence of
occurrence of tensile stress
0.10 tensile stress 0.10
' = 35 25
0.05 0.05
15
0.00 0.00
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
c  - 
c  - 
γ D γ D  
(e) (f)
 -

 -

0.50 0.50
H/H w = 1
ts k E

ts k E
 D 2

 D 2

0.45 0.45
 w/  ’ = 1
= 0 H=H w=4D
0.40 0.40 = 0.3
' = 25
0.35 0.35
15
D
0.30 0.30 H/D = 4
occurrence of K0 = 0.5
0.25 tensile stress 0.25
H/D = 4 H/H w = 1
0.20  w/  ’ = 1
0.20 occurrenceof
= 0 tensile stress
H=H w=4D
0.15 = 0.3 0.15
' = 25
0.10 0.10
D 15
0.05 H/D = 4 0.05
K0 = 1
0.00 0.00
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
c  - 
c  - 
γ D γ D  

Figure 7.26. Dimensionless diagrams for the determination of the stand-up time of the tunnel
face (= 0°,  = 0.3, Hw / H = 1, w / '=1) obtained for an element size of 0.05 D.

190
 Chapter 7 – Stand-up time of tunnel face

K0 = 1 K0 = 0.5
(a) (b)

 -
 -
0.15 0.05
H=H w=D H=H w=D

ts k E
ts k E

 D 2
 D 2
D D
0.12 0.04
H/H w = 1 H/H w = 1
 w/  ’ = 1  w/  ’ = 1
= 0 H/D = 1 = 0 H/D = 1
0.09 = 0.3 K0 = 1 0.03 = 0.3 K0 = 0.5

H/D = 1
0.06 0.02

' = 35
0.03 25 0.01
15
' = 35
25 15
0.00 0.00
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
c  -  c  - 
γ D   γ D  
(c) (d)
 -

0.40 0.20

 -
ts k E

ts k E
 D 2

 D 2
H=H w=2D H=H w=2D
0.35
0.16
D D
0.30
H/H w = 1 H/H w = 1
0.25  w/  ’ = 1  w/  ’ = 1
H/D = 2 0.12 H/D = 2
= 0 = 0
= 0.3 K0 = 1 = 0.3 K0 = 0.5
0.20
H/D = 2
0.08
0.15
' = 35
25
0.10
0.04 ' = 35
15 25
15
0.05

0.00 0.00
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
c  -  c  - 
γ D   γ D  
(e) (f)
 -

 -

0.70 0.30
H/H w = 1 H/H w = 1
ts k E

ts k E
 D 2

 D 2

0.60  w/  ’ = 1  w/  ’ = 1
= 0 H=H w=4D
0.25 = 0 H=H w=4D
= 0.3 = 0.3
0.50
0.20
D D
0.40 H/D = 4 H/D = 4
K0 = 1 K0 = 0.5
0.15
H/D = 4 0.30
' = 25
15 0.10
0.20 ' = 25

15
0.05
0.10

0.00 0.00
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
c  -  c  - 
γ D   γ D  

Figure 7.27. Dimensionless diagrams for the determination of the stand-up time of the tunnel
face obtained based upon linear extrapolation to a zero element size
(= 0°,  = 0.3, Hw / H = 1, w / '=1).

191
7.8 Application example
In order to illustrate the utility of the design diagrams let consider the following example,
based upon the construction of the tunnel of Line 3 of the Athens Metro. The tunnel is under
construction with an EPB machine and runs in the south-west of the city. Under the
operational point of view it is advantageous to carry-out inspection and/or maintenance of the
cutterhead under atmospheric conditions. For this reason the stability of the unsupported
tunnel face is investigated in a specific tunnel location. The ground consists of weak Lower
Athens Schists at tunnel elevation, overlayed by a layer of the more competent Upper Athens
Schist. The water table is located at the ground surface (Fig. 7.28a).
The design nomograms are applied making the conservative assumption of homogenous
ground, with the mechanical and hydraulic properties of the particularly weak Lower Athens
Schists. Two models are considered: one extending up to the ground surface (Fig. 7.28c) and,
one extending up to the boundary between the two geological units (Fig. 7.28b).
For the parameter set given in Figure 7.28c (normalized cohesion c'/('D) = 0.24, friction
angle ' = 25°) the dimensionless stand-up time obtained with Figure 7.26 is equal to 0.05 –
0.075, while tensile stresses would start developing at 0.065 – 0.1. With a simple
transformation taking into account the permeability of the ground k, the Modulus of Young
E, the submerged unit weight of the ground ' and the tunnel diameter D the stand-up time
amounts to 10 – 15 hours. The design diagrams of Figure 7.27 provide a stand-up time of 2 –
8 hours. This period is sufficient for depressurizing and emptying the working chamber and
carrying-out routine inspection and maintenance works.
The field experience agreed well with the computational predictions: after 6 hours of works
inside the excavation chamber under atmospheric conditions the first indications of instability
were observed (in form of some ravelling in the upper part of the tunnel face). As a
consequence the excavation chamber was evacuated and pressurized again.
This simple application example shows the usefulness of using the design diagrams: they
assist operational decision-making – in the present example related to the feasibility of men
entries in the cutterhead chamber under atmospheric conditions.

considered tunnel
location

Surface deposits
' = 13 kN/m3
9.5 m Upper Athens Schists K0 = 0.5
E = 160 MPa
c' = 30 kPa
9.5 m Lower Athens Schists ' = 21
k = 10-8 m/s

TBM
D = 9.5 m

(a) (b) (c)

Figure 7.28. (a) Geological profile; (b, c) simplified models assumed for an estimation of the
stand-up time from the design nomograms.

192
 Chapter 7 – Stand-up time of tunnel face

7.9 The influence of the support pressure on the stand-up time

The design diagrams apply to an unsupported tunnel face, i.e. to the worst case with respect to
stability. The application of a slight support pressure (for example, a low air pressure,
partially compensating the in situ hydrostatic pressure), however, is often operationally not
problematic, particularly considering that short stand-up times are disadvantageous, too.
Figure 7.29 shows the influence of the support pressure on the stand-up time for the example
of a 20 m deep tunnel with a diameter of 10 m. The solid line are obtained assuming that the
applied support pressure acts as effective support, while the dashed lines assumes that the
support pressure partially compensate the water pressure. The two conditions are simulated
by applying the support pressure as uniform total stress and imposing as hydraulic boundary
conditions at the tunnel face either an atmospheric pressure or the support pressure.
The partial compensation of the water pressure is representative for hyperbaric interventions
(dashed lines), while the simulation of effective support is representative for a mechanical
support of the tunnel face, e.g. by reinforcing the tunnel face with bolts (note that a support
pressure of 0.5 to 1 bar amounts – assuming sufficient bolt strength and overlapping length –
to a reinforcement density of about 0.25 to 0.5 bolts per square meter, respectively;
Anagnostou and Perazzelli, 2015). The effect of a light support pressure is remarkable:
applying an effective support pressure of 0.8 bar, the tunnel face would remain stable even
long-term, while in a hyperbaric intervention under 1 bar (40% of the in situ hydrostatic
pressure) the stand-up time would be twice as high as in the case of an atmospheric
intervention.

80
ts [h]

H=H w=2D
70
ps D =10 m
60
' = 25
50 c' = 20 kP a
K0 = 1
40 k = 10 -7 m/s
E = 20 MP a
 = 20 kN/m3
30

20 Effective support pressure p s' = p s

P artial compensation of the water
10 pressure p w= p s

0
0 25 50 75 100
p s [kPa]
Figure 7.29. Stand-up time of the tunnel face as a function of the applied support pressure ps.

193
7.10 Critical advance rate
The previous Sections investigated stand-up time during an excavation stand-still, assuming
that undrained conditions prevail at the beginning of the stand-still. As explained in Section
7.4, this assumption is reasonable in the case of very fast (compared to the ground
permeability) tunnel advance. Otherwise, partially drained conditions would prevail during
advance and thus at the beginning of the standstill.
With respect to the question of face stability, “partially drained conditions” means that the
permeability of the ground is such that dissipation of excess pore pressures occurs to a certain
(practically relevant but nevertheless incomplete) degree simultaneously with the advance of
excavation or during a standstill. As explained by Anagnostou (2007), a simple dimensional
analysis shows that the ratio of soil permeability k to advance rate v will govern whether
steady, partially drained or undrained conditions prevail during ongoing excavation. The
higher the advance rate v and the lower the permeability k, the less will the pore pressures
dissipate in the vicinity of the face and, consequently, the higher will be the shear resistance.
If, on the other hand, the permeability is high or the advance slow, drained conditions, which
are less favourable for stability, will already prevail around the working face.
Under such conditions, the relevant important question is the estimation of the minimum
(“critical”) advance rate for which the face would remain stable during excavation. The
question of face stability during ongoing excavation was addressed only by Höfle et al. (2008,
2009) and by Sitarenios and Kavvadas (2016) (Section 1.2.1).
The design charts presented in Section 7.7 can be used for a simplified estimation of the
critical advance rate. Noting that the failure zone ahead is extended up to one diameter ahead
of the tunnel face (Figs. 7.12 and 7.16) and that during advance this zone is continuously
excavated (Fig. 7.30), the critical advance rate vcr can be taken equal to the ratio of tunnel
diameter D to the stand-up time ts :

D
vcr  , (7.8)
ts

where ts can be obtained from Ts given by the design charts of Fig. 7.26 or 7.27.
where ts can be obtained from Ts given by the design charts of Fig. 7.26 or 7.27. Figure 7.31
shows the critical advance rate as a function of cohesion for a circular tunnel with diameter
10 m located at depth 10 m (same geometry and ground parameters as in Section 7.4). In this
example, assuming a cohesion of the ground of 25 kPa, the face would be stable during
ongoing excavation, if the advance rate amounts to at least to 12 m/day (typical range) or
39 m/day (extremely high value) with Ts after Fig. 7.26 and Fig. 7.27, respectively.

194
 Chapter 7 – Stand-up time of tunnel face

vcr vcr = D / ts
D t0 ts

D
Figure 7.30. Geometrical definition of critical advance speed.

50 70
ts [h]
vcr [m/day]

H=H w= 10 m
45
60 D = 10 m

40 ’ = 25
50 K0 = 1
35 k = 10 -7 m/s
E = 20 Mpa
30
40  = 20 kN/m3

25 30 Face stable
under drained
20 vcr = D/ts (ts according to Fig. 7.26) conditions
20
vcr = D/ts (ts according to Fig. 7.27)
15
10 Reference
vcr obtained from case computation
FEM step-by-step
10 (element size 0.05 D)

5 0
5 10 15 20 25 30 35 40
0 c' [kPa]
5 10 15 20 25 30 35
c’ [kPa]

Figure 7.31. Critical advance rate as a function of the cohesion c'.

For the same example, a comparative computation was carried-out, simulating numerically
the tunnel advance step by step. Figure 7.32 shows the computational model. The excavation
is simulated by successively deactivating the ground elements over a total of 15 successive
steps of 2 m length. The elements size amounts to 2 m over steps 1 – 9 and to 0.5 m over
steps 10 – 15 (being thus the same, i.e. equal to 0.05D, as in the design charts of Fig. 7.26).
Each step consists in a transient analysis of duration consistent with the imposed advance
rate, during which the nodal forces at the tunnel face are progressively reduced to zero (e.g.
for advance rate of 10 m/day and step length 2 m, the deactivation of the elements is carried-
out over a time step of 0.2 days). The tunnel lining is simulated in a simplified way by fixing
the displacements at the excavation boundary just prior to the deactivation of the ground
elements. No-flow conditions are imposed to the tunnel wall, while mixed conditions where
applied to the tunnel face.

195
Figure 7.32. Computational model used for the step-by-step FEM analyses.

v = 10.5 m/d (0.5D/ts)

v = 15.8 m/d (0.75D/ts)
v v = 21.1 m/d (D/ts)

c ' = 20 kPa
φ' = 25
K0 = 1

Figure 7.33. Boundary of the plastic zone for several values of advance rate (other material
parameters according to Table 7.1).

196
 Chapter 7 – Stand-up time of tunnel face

Figure 7.33 shows the boundary of the plastic zones at the end of step 15 obtained for three
value of advance speed and for cohesion 20 kPa. The extent of the plastic zone increases with
decreasing advance rate. The critical advance speed for the presents example amounts to 10.5
m/day. For lower advance rates the numerical algorithm fails to find an equilibrated solution.
Taking into account the stand-up time according to Figure 7.26 (which was based upon the
same mesh size as used in the step-by-step calculations) the critical advance speed
corresponds to a ratio D/ts of 0.5.
The black diamonds in Figure 7.31 corresponds to the critical advance rate determined by
means of the step-by step FEM analyses for several cohesion values. For the considered
example, the numerically computed critical advance rate varies between about 0.5 – 0.6 D/ts.
This means that the critical advance rate after Equation (7.8) overestimates the one computed
numerically, thus providing a conservative estimation which can be used in the engineering
practice.

7.11 Conclusions
Experience shows that the tunnel face may be stable in the short term, but collapse after a
certain time period. The stand-up time of the tunnel face is important in engineering practice,
especially in medium- and low-permeability water-bearing ground.
Numerical models are able to explain the phenomenon of delayed failure; they show that
instability occurs more or less rapidly depending on the permeability, stiffness and shear
strength of the ground, as well as on the coefficient of lateral pressure.
On the basis of a parametric study, Chapter 7 presented design charts allowing the stand-up
time to be estimated for a given geotechnical situation, thus assisting the engineers on-site
decision-making. The design diagrams are elaborated for a fixed relatively fine mesh size and
for a mesh size of zero (obtained by means of a linear extrapolation of the results to zero,
taking into account the linear dependency of the stand-up time to the mesh size). The latter
represent a lower limit of the stand-up time expected for an infinitely thin shear band.
In view of the underlying simplifying assumptions (e.g. homogenous ground and undrained
conditions up to the beginning of the standstill period) their use presupposes a measure of
engineering judgment. Due to the uncertainties related to the computation method and to the
heterogeneity of the ground (concerning both structure and material parameters), a
sufficiently high safety factor should be applied to the estimated stand-up time, and the face
behaviour should be monitored during standstills in order to timely detect the onset of any
instability (cracks, extrusion, increasing water inflows, ravelling). In the case of
unsatisfactory ground behaviour, re-pressurization of the excavation chamber (mechanized
tunnelling) or reinforcement of the face (conventional tunnelling) will be required. Chapter 7
showed that even a small support pressure (0.5 to 1 bar) increase the stand-up time of the
tunnel face considerably.
Finally, Chapter 7 gave an estimation of the minimum average advance rate, which allows to
excavate without support/reinforcement of the tunnel face. The minimum advance rate can be
taken approximately equal to the ratio of tunnel diameter to stand-up time.

197
198
 Chapter 8 – Conclusions and Outlook

8. Conclusions and outlook

This Thesis investigated the phenomenon of delayed failure of excavations, which is
particularly relevant in soils of medium-low permeability.
The first part of the Thesis (Chapters 3 and 4) consisted in fundamental numerical research on
to the manifestation of failure and on the influence of the discretization on the results. The
results can be summarized as follows:
 In order to obtain accelerating displacements at failure, it is essential to adopt constitutive
models that map the soil property to experience shearing at a constant volume (MC model
or models of the critical state family).
 Isochoric plastic shearing leads, however, to numerical stability problems close to failure
because it is inherently impossible to satisfy both the equilibrium and the mass balance
condition, thus making the determination of the stand-up time difficult. The stand-up time
can be only identified by evaluating the entirety of the numerical results; the occurrence of
failure is indicated by: the increase in the displacement rate over time, the development of
a plastic zone extending from the tunnel face to the ground surface, and, finally, the
occurrence of numerical stability problems.
 The assumption of a non-associated plastic flow leads to structural softening and so to a
localization of the plastic deformation in a shear band and to mesh-dependency of the
stand-up time (it decreases with mesh fineness).
 For fine meshes the observed mesh dependency becomes almost linear, which suggests a
practicable way of stand-up time estimation (linear extrapolation the results to the
expected, grain-size-dependent thickness of the shear band).
The second part of the Thesis (Chapters 5 and 6) presented the physical models developed for
the study of the delayed failure and described the obtained experimental results as well their
numerical interpretation. For the assumed material parameters, the latter showed that:
 When using a computationally manageable fine spatial discretization, the assumption of a
material obeying to a simple linearly plastic perfectly plastic MC model generally
underestimates stand-up time as it does not capture the plastic dilative behaviour, while the
MCC model generally overestimates the stand-up time because of the overestimation of
the shear resistance for stress paths typical of delayed failure (yielding on the dry side of
the yield surface). However, the adoption of the MCC model and the extrapolation of the
numerical results to the size of the shear band give stand-up times very close to the
observed one.
The last part of the Thesis (Chapter 7) presented a systematic numerical study on the time
dependent stability of the tunnel face and presented design nomograms for the estimation of
stand-up time, which apply for a wide range of geotechnical conditions (soil properties, depth
of cover and water table). The elaboration of the nomograms will assist tunnel engineers in
their decision-making.

199
In the opinion of the author, future research on delayed failure of geotechnical structures in
low permeability soil should focus on four aspects.
1. Validation of the material parameters (elimination of the uncertainty regarding the
inconsistency between the MC and MCC parameters proposed in the literature) – and so of
the numerical results – by means of additional laboratory tests (e.g. triaxial tests). The tests
should be carried-out at stress states close to the one prevailing in the analysed problems.
2. Better understanding of the evolution of tensile failure by means of experiments and
numerical studies. The present Thesis showed that in soils characterized by high cohesion
(being thus less prone to shear failure) tensile stresses develop behind the tunnel face due
to hydraulic head gradients. Constitutive models considering a tension cut-off predict a
qualitatively different behaviour (constant displacement rate at failure), which contradict
the expected physical behaviour (accelerating displacement rate at failure).
3. Validation of FEM results on the stand-up time of the tunnel face by means of a
comparison with centrifuge tests (compared to in situ observations, centrifuge tests are
more suitable for the validation of the results because of the controlled boundary
conditions and material parameters; moreover, in situ observation would necessitate the
occurrence of a collapse, which is usually avoided by timely increasing the amount of the
reinforcement or the support pressure in the case of a conventional or mechanized
pressurized excavation, respectively). Although, the ability of the numerical methods has
been validated in the present research work, a specific validation for the tunnel face
problem should be carried-out in order to take into account the complex mechanism of
pore pressure dissipation and stress-redistribution that occurs in the ground around the
consolidating tunnel face.
4. Use of other numerical methods or element types (e.g. enhanced shape elements) dealing
with the mesh-dependency of the results in order to validate the stand-up time obtained
based upon linear extrapolation.
5. More thorough and detailed evaluation of the critical advance rate and of the effect of a
support pressure on stand-up time by means of FEM analyses.

200
 Appendix A – Publications from the present Thesis

Appendix A. Publications from the present Thesis

Preliminary results of Chapter 3 have been presented in:

Schuerch, R., Anagnostou, G. (2015a). Delayed failure identification by coupled hydraulic-

mechanical numerical analyses. In: Proc. 14th IACMAG, 1061–1066, Kyoto.

Schuerch, R., Vrakas, A., Anagnostou, G. (2017). On manifestations of delayed failure and
the effect of dilatancy in transient poro-elasto-plastic analyses of slopes and excavations.
Géotechnique (http://dx.doi.org/10.1680/jgeot.15.P.238]).

Preliminary results of Chapter 4 have been presented in:

Schuerch, R., Anagnostou, G. (2015b). Failure propagation and mesh-dependency in coupled

hydraulic-mechanical transient problems. In: Proc. 14th IACMAG, 329–334, Kyoto.

Schuerch, R., Vrakas, A., Anagnostou, G. (2017). On manifestations of delayed failure and
the effect of dilatancy in transient poro-elasto-plastic analyses of slopes and excavations.
Géotechnique (accepted for publication).

Preliminary results of Chapter 7 have been presented in:

Schuerch, R., Anagnostou, G. (2012). Tunnel face stability under transient conditions: stand-
up time in low permeability ground. In: Proc. of the 22nd European Young Geotechnical
Engineers Conference, Gothenburg.

Schuerch, R., Anagnostou, G. (2013a). Analysis of the stand-up time of the tunnel face. In
Proc. ITA-AITES World Tunnel Congress, WTC 2013, Geneva.

Schuerch, R., Anagnostou, G. (2013b). The influence of the shear strength of the ground on
the stand-up time of the tunnel face. In: Proc. TU-Seoul 2013 International Symposium on
Tunnelling and Underground Space Construction for Sustainable Development, Seoul.

Schuerch, R., Anagnostou, G. (2015b). Failure propagation and mesh-dependency in coupled

hydraulic-mechanical transient problems. In: Proc. 14th IACMAG, 329–334, Kyoto.

Schuerch, R., Poggiati, R., Anagnostou, G. (2016) Design diagrams for estimating tunnel face
stand-up time in water-bearing ground. In: Proc. WTC16, San Francisco.

Schuerch, R., Poggiati, R., Maspolo, P., Anagnostou, G. (2016) Design charts for estimating
tunnel face stand-up time in soft ground tunnelling. In: Proc. 2nd International Conference on
Tunnel Boring Machines in Difficult Grounds, Istanbul.

201
202
 Appendix B – Water retention curve

Appendix B. Water retention curve

The water retention curve relates the suction induced by the partial desaturation (capillary
suction)  to the effective degree of saturation Se.
As firstly proposed by Wind (1968), the water retention curve can be estimated by letting a
soil specimen drying (or wetting) and, at the same time, measuring its change of mass and the
generated pore pressure. In the test proposed by Wind (1968), water evaporates from the top
boundary of a vertical soil column, which is laterally sealed by a plastic film. The change in
weight is controlled by means of an electronic scale, while the pore pressure is measured with
tensiometers.
Equations (B.1–B.5) derive the relationship between the effective degree of saturation and the
mass of the specimen.
The effective degree of saturation Se is defined as

  r
Se  , (B.1)
s  r

where  is the volumetric water content, r is the residual water content and s is the saturated
water content. Under the assumption zero residual water content, Equation (B.1) reduces to


Se   Sr , (B.2)
s

where Sr is the degree of saturation. The latter is be defined as

w s
Sr  , (B.3)
e

where ρs is the relative density of the solids e is the void ratio and w is the gravimetric water
content. The latter is the ratio between the mass of water inside the specimen mw and to its dry
mass mdry:

mw mtot  mdry
w  , (B.4)
mdry mdry

where mtot is the total (wet) mass of the specimen. Eqs. (B.3) and (B.4) lead to

 s mtot  mdry
Sr  . (B.5)
e mdry

As explained recently by Lourenço et al. (2011), two procedures exist with respect to the
method proposed by Wind (1968): the continuous and the discrete one. In the continuous
procedure, the sample is exposed to the atmosphere and the soil is let to dry continuously. In

203
the discrete procedure, the drying (or wetting) process is halted at different stages to ensure
equalization of the pore pressure within the sample before measuring suction.
Lourenço et al. (2011) showed that the water retention curves obtained by the continuous
drying procedure are similar to those obtained by the discrete drying procedure when using
small specimens (30 mm high, with a diameter of 100 mm or 35 mm in the given reference)
that are allowed to dry (or wetted) from the top and lateral surfaces. In the mentioned work
the ppt was inserted from the top surface and pushed into the soil to a depth of 3 mm.
Lourenço et al. (2011) further proposed a new procedure, which eliminates the inaccuracy in
the mass measurement caused by the weight and the stiffness of the ppt cable. The procedure
consists in testing, at the same time, two samples of the same size: one is used for the measure
of the weight and the second one for the measure of the pore pressure.
In the present work the water retention curve for the Birmesndorf clay was estimated by
adopting the continuous procedure. The tests were executed on two prismatic samples (Fig.
B.1a) with dimensions 100 mm x 100 mm x 30 mm (sample 1) and 60 mm x 60 mm x 25 mm
(sample 2), respectively. The material used for the test was the same as the one used for the
uniaxial tests: Birmensodorf clay, over-consolidated at 100 kPa and successively consolidated
at 50 kPa. Similarly to the test proposed by Lourenço et al. (2011), the specimens were placed
on an aluminium plate and could evaporate from the upper and the lateral surfaces. The pore
pressure change was measured by means a miniature pore pressure transducer (type
2MIX/2bar/81840.1) pushed in the centre of the specimen up to 7 mm below its upper
surface. The change of weight was recorded with an electronic scale.
The specimens were submitted to successive cycles of wetting (by submerging the sample
into water) and drying (by exposing the specimens to the laboratory conditions; relative
humidity 45% and temperature of 18.5° C) during which the mass and the pore pressure were
recorder at several times. Table B.1 shows the data measured (mtot and w) and derived (mw,
Sr). The degree of saturation expressed by Equation (B.5) was calculated assuming ρs = 2.75
(Weber, 2007), e = 1.1 (Weber, 2007) and mdry = 395.21 g and 137.97 g for sample 1 and 2,
respectively.
Figure B.1 shows the water retention curve obtained for the Birmensdorf clay. The results
leading to Sr > 100%, which lack physical meaning, are due to inaccuracy in mass
measurements (as explained above, the error is due to the weight and the stiffness of the ppt
cable).

204
 Appendix B – Water retention curve

(a)
100 mm

Sample 1 7 mm

ppt 30 mm

impervious layer

60 mm

Sample 2 7 mm

25 mm
ppt

impervious layer

(b)

1000
Suction [kPa]

100

10

Sample 1
Sample 2

1
0% 20% 40% 60% 80% 100%
Sr [%]

Figure B.1. (a) Ppt position and sample size; (b) measured suction as a function of the
degree of saturation in the samples. The points indicated the water retention
curve for the Birmensdorf clay.

205
Table B.1. Measured and derived data for the estimation of the water-retention curve.

mtot mw w Sr 
[g] [g] [-] [-] [kPa]
Sample 1 (100 mm x 100 mm x 30 mm)
Drying 562.8 167.6 0.42 1.06 2.2
548.4 153.2 0.39 0.97 13.4
532.9 137.7 0.35 0.87 52.8
Wetting 558.4 163.2 0.41 1.03 2.6
556.0 160.8 0.41 1.02 2.8
553.3 158.1 0.40 1.00 2.4
552.6 157.4 0.40 1.00 2.3
552.3 157.1 0.40 0.99 2.4
550.1 154.9 0.39 0.98 3.0
549.5 154.3 0.39 0.98 4.2
548.6 153.4 0.39 0.97 4.2
Drying 546.5 151.3 0.38 0.96 5.8
487.4 92.2 0.23 0.58 42.7
461.7 66.5 0.17 0.42 59.3
473.5 78.3 0.20 0.50 54.4
Sample 2 (60 mm x 60 mm x 25 mm)
Wetting 189.3 51.3 0.37 0.93 27.4
189.1 51.1 0.37 0.93 11.1
193.3 55.3 0.40 1.00 1.9
Drying 192.6 54.6 0.40 0.99 1.8
178.8 40.8 0.30 0.74 78.8
Wetting 177.4 39.4 0.29 0.71 79.3
193.6 55.6 0.40 1.01 0.0
193.6 55.6 0.40 1.01 0.8

206
 Appendix C – List of symbols

Appendix C. List of symbols

A Factor of the Kozeny-Carman equation
a0 Centre of the initial yield surface
Ae, Aep Elastic and elasto-plastic acoustic (or localization) tensor
ai, i=0,4 Coefficients for the computation of the determinant of the acoustic tensor
b Width of the sample
c, c' Cohesion
cu Undrained shear strength
cv Coefficient of consolidation
cw Water compressibility
d Diameter of the T-bar
Dep Elasto-plastic stiffness matrix
DR Relative density of solids
e Void ratio
E, E' Young’s Modulus
e0 Initial void ratio
F Measured force during T-Bar test
ft Tensile resistance
G Total weight of the wedge
g, n∙g Ground acceleration, centrifugal acceleration
h, H Height
HR Relative air humidity
Hw Water table above the tunnel crown
i, j, m Dummy index

Dijep
i,j =1,3 Components of the elasto-plastic stiffness matrix
IP Plasticity index
k Coefficient of permeability
K Third invariant ratio
K0 Lateral pressure coefficient
L Length
l Local coordinate along the sliding plane
M Slope of the critical state line
n Porosity
N Specific volume at a unit reference pressure on the NCL
n1, n2 Direction cosines of the unit vector normal to the shear band

207
Nb Dimensional resistance factor
OCR Over-consolidation ratio
p, p' Mean effective and total stress
ps Support pressure

p Average pore pressure

w

Pw,0 Initial pore pressure

pw Pore pressure
pw0 Initial pore pressure
q Deviatoric stress
qn Normal flow
qv Vertical load
qx Water flow
qy Vertical flow
s Mesh size
sx, sy Boundary traction
S Specific surface of the solids
su Undrained shear strength
T Tangential component of the resultant of the forces acting along the sliding plane
t Time
t∞ Time at the end of the consolidation process
Tres Integral of the shear resistance along the sliding plane
ts Stand-up time
Tv Dimensionless time factor

U Average consolidation degree in the model

uA,x Horizontal displacement at point A
uA,y Vertical displacement at point A
ux Horizontal displacement
u y, u z Vertical displacement
uz,0 Initial vertical displacement
v Velocity vector of the sliding wedge (tangential to the shear band)
W Resultant of the water pressure acting on the vertical cut
wl Liquid limit
wp Plastic limit
x, y, z Coordinates
ytip,shear band Vertical position of the tip of the shear band
t Time increment

208
 Appendix C – List of symbols

Greek symbols

Φ Water potential
H Lateral unloading of the sample
elvol Elastic component of the volumetric strain
 pl
eq Equivalent plastic strain
 pl
vol Plastic component of the volumetric strain
vol Volumetric strain
x Axial strain
xy Shear strain
del Elastic strain increment
dpl Plastic strain increment
  Saturated unit weight
' Submerged unit weight of the ground
s Unit weight of the solid substance
w Unit weight of water
 Reduction factor
' Effective friction angle
  Slope of the swelling line in the semi-logarithmic compression plane 
  Slope of the compression line in the semi-logarithmic compression plane
  Poisson's ratio
 Angle between the unit vector normal to the shear band and the horizontal axis
 Angle between x-axis and the shear band
'v Vertical effective stress
x'x Total and effective horizontal stress (2D, 3D)
y'y Total and effective vertical stress (2D) / horizontal transversal stress (3D)
'z Total and effective vertical stress (3D)
'1'3 Effective maximal and minimal principal stress
n0’n0 Initial total and effective stress normal to the tunnel face
n’n Total and effective normal stress
'OC Overconsolidation pressure
x0,'x0 Initial horizontal total and effective stress
xy Shear stress
y0, 'y0 Initial total and effective vertical stress
n Tangential stress to the shear band
w Angle between the shear band and the horizontal axis
 Dilation angle

209
210
 Chapter 9 – References

9. References

Anagnostou, G., Kovári, K. (1994). The face stability of slurry-shield driven tunnels. Tunn.
Undergr. Space Technol. 9, No. 2, 65–174.

Anagnostou, G. (1995). The influence of tunnel excavation on the hydraulic head. Int. J.
Num. Anal. Meth. Geomech. 19, No. 10, 725–746.

Anagnostou, G. (2007). Practical consequences of the time-dependency of ground behaviour

in tunnelling. In: Proc. RETC, 255–265, Toronto.

Anagnostou, G., Perazzelli, P. (2015). Analysis method and design charts for bolt
reinforcement of the tunnel face in cohesive-frictional soils. Tunnelling and Underground
Space Technology 47, No. 1, 162–181.

Arthur, J.R.F., Dunstan, T., Al-Ani, Q.A.J.L., Assadi, A. (1977). Plastic deformation and
failure in granular media. Géotechnique 27, No. 1, 53–74.

Atkinson, J.H., Bransby, P.L. (1978). The mechanics of soils. McGraw-Hill Book Company
(UK).

Atkinson, J.H., Mair, R.J. (1981). Soil mechanics aspects of soft ground tunnelling. Ground
Eng. 14, No. 5, 20–28.

Banerjee, P.K., Kumbhojkar, A.S., Yousif, N.B. (1988). Finite element analysis of the
stability of a vertical cut using an anisotropic soil model. Can. Geotech. J. 25, 119–
127.

Bazant, Z.P., Cedolin, L. (1979). Blunt crack propagation in finite element analysis. Journal
of the Engineering Mechanics Division, ASCE 105, 297–315.

Bazant, Z.P., Belytschko, T., Chang, T.P. (1984). Continuum theory for strain-softening.
Journal of Engineering Mechanics 110, No. 2, 666–1692.

Bazant, Z.P., Lin, F.B. (1989). Stability against localization of softening into ellipsoids and
bands: Parameter Study. Int. J. Solids Struct. 28, 1483–1498.

Berg, C.A. (1972). A note on construction of the equivalent plastic strain increment. Journal
of Research 76, 197–213.

Bigoni, D., Hueckel, T. (1991). Uniqueness and localization-I. Associative and

nonassociative elastoplasticity. International Journal of Solids and Structures 28, 197–
213.

Bishop, A.W. (1955). The use of the slip circle in the stability analysis of slopes.
Géotechnique 5, No. 1, 7–17.

211
Bjerrum, L., Kjaernsli, B. (1957). Analysis of the stability of some Norwegian natural clay
slopes. Géotechnique 7, No. 1, 1–16.

Borges, J.L. (2008). Cut slopes in clayey soils: consolidation and overall stability by finite
element method. Geotech. Geol. Engng 26, 479–491.

Borja, R.I., Regueiro, R.A. (2001). Strain localization in frictional materials exhibiting
displacement jumps. Comput. Meth. Appl. Mech. Engng. 191, 2555–2580.

Broms, B.B., Bennermark, H. (1967). Stability of vertical openings. Journal of the Soil
Mechanics and Foundations Division 93, 71–94.

Callari, C., Armero, F. (2002). Finite Element Methods for the Analysis of Strong
Discontinuities in Coupled Poroplastic Media. Computer Methods in Applied
Mechanics and Engineering 19, 4371–4400.

Callari, C. (2015). Numerical assessment of tunnel face stability below the water table. In:
Proc. 14th IACMAG, 2007–2010, Kyoto.

Chandler, R.J. (1984). Recent European experience of landslides in over-consolidated clays

and soft rocks. In: Proc. 4th Int. Symp. Landslides, 61–81, Toronto.

Chapuis, R.P., Aubertin, M. (2003). On the use of the Kozeny-Carman equation to predict the
hydraulic conductivity of soils. Can. Geotech. J. 40, No. 3, 616–628.

Clausen, J., Damkilde, L., Andersen, L., (2005). An efficient return algorithm for non-
associated Mohr-Coulomb plasticity. In: Proc.10th Int. Conf. Civil, Structural and
Environmental Engineering Computing, United Kingdom.

Clausen, J., Damkilde, L., Andersen, L. (2007). An efficient return algorithm for non-
associated plasticity with linear yield criteria in principal stress space. Computers and
Structures 85, No. 23–24, 1795–1807.

Cleary, M.P. (1976). Continuously distributed dislocation model for shear bands in softening
materials, International Journal for numerical methods in engineering 10, 679–702.

Collin, F., Chambon, R., Charlier, R. (2006). A finite element method for poro mechanical
modelling of geotechnical problems using local second gradient models. Int. J. Num.
Meth. Engng. 65, 1749–1772.

Cooper, M.R., Bromhead, E.N., Petley, D.J., Grant, D.I. (1998). The Selborne cutting stability
experiment. Géotechnique 48, No. 1, 83–101.

Culmann, C. (1875). Die graphische Statik. Meyer & Zeller, Zurich.

Dassault Systèmes (2012). Abaqus 6.12, Theory and Analysis User’s Manual.

Davis, E.H. (1968). Theories of plasticity and the failure of soil masses. In: Soil Mechanics:
Selected Topics, (Editor: I. K. Lee), Butterworths, London, 341–380.

212
 Chapter 9 – References

Davis, E.H., Gunn, M.J., Mair, R.J., Seneviratne, H.N. (1980). The stability of shallow
tunnels and underground openings in cohesive material, Géotechnique 30, No. 4, 397–416.

Davies, M.C.R. (1985). Centrifuge model of an embankment failure. In: Proc. Failure in
earthworks. Thomas Telford Ltd, 451–453, London.

de Borst, R., Vermeer, P.A. (1984). Possibilities and limitations of finite elements for limit
analysis. Géotechnique 34, No. 2, 199–210.

de Borst, R. (1988). Bifurcations in finite element models with a non-associated flow law. Int
J Numer Anal Meth Geomech, 12, 99–116.

de Borst, R. (1989). Numerical methods for bifurcation analysis in geomechanics. Ingenieur-

Archiv 59, 160–174.

de Borst, R., Sluys, L.J. (1991). Localisation in a Cosserat continuum under static and
dynamic loading conditions. Computer Methods in Applied Mechanics and
Engineering 90, No. 1, 805–827.

de Borst, R. (1991a). Numerical modelling of bifurcation and localisation in cohesive-

frictional materials. pure and applied geophysics 137, No. 4, 367–390.

de Borst, R. (1991b). Simulation of strain localization: a reappraisal of the Cosserat

continuum. Eng. Comput. 8, No. 4, 317–332.

de Borst, R., Sluys, L.J., Mühlhaus, H.B., Pamin, J. (1993). Fundamental issues in finite
element analyses of localization of deformation. Engineering computations 10, No. 2,
99–121.

Deepa, V., Viswanadham, V.S. (2009). Centrifuge model tests on soil-nailed slopes subjected
to seepage. Ground improvement 162, No. 13, 133–144.

Drescher, A., Detournay, E. (1993). Limit load in translational failure mechanisms for
associative and non-associative materials. Géotechnique 43, No. 3, 443–456.

Drucker, D.C. (1951). A more fundamental approach to plastic stress-strain relations. In:
Proc. of the First U.S. National Congress of Applied Mechanics, ASME, 487–491.

Ehlers, W., Volk, W. (1998). On theoretical and numerical methods in the theory of porous
media based on polar and non-polar elasto-plastic solid materials. Int J Solids
Structures 35, 4597–4617.

Eisenstein, Z., Samarasekera, L. (1992). Stability of unsupported tunnels in clay. Can.

Geotech. J. 29, 609–613.

Ellis, E.A., O'Brien, A.S. (2007). Effect of height on delayed collapse of cuttings in stiff clay.
Geotechnical Engineering 160, No. 2, 73–84.

213
Griffiths, D.V. (1989). Computation of collapse loads in geomechanics by finite elements.
Ing. Archiv 59, 237–244.

Griffiths, D.V., Hicks, M.A., Li, C.O. (1991). Transient passive earth pressure analyses.
Géotechnique 41, No. 4, 615–620.

Griffiths, D.V., Li, C.O. (1993). Analysis of delayed failure in sloping excavations. J.
Geotech. Eng. 119, No. 9, 1360–1378.

Griffiths, D.V., Lane, P.A. (1999). Slope stability analysis by finite elements. Géotechnique
49, No. 3, 387–403.

Gudehus, G. (1973). Elastoplastische Stoffgleichungen für trockenen Sand. Ingenieur Archiv

42, 151–169.

Guglielmetti, V., Grasso, P., Mahtab, A., Xu, S. (2008). Mechanized tunnelling in urban
areas. Taylor & Francis, London.

Han, C., Vardoulakis, I. (1991). Plane-strain compression experiments on water-saturated

fine-grained sand. Géotechnique 41, No. 1, 49–78.

Höfle, R., Fillibeck, J., Vogt, N. (2008) Time dependent deformations during tunnelling and
stability of tunnel faces in fine-grained soils under groundwater. Acta Geotech 3, 309–316.

Höfle, R., Fillibeck, J., Vogt, N. (2009). Time depending stability of tunnel face. In: Proc.
35th ITA–AITES General Assembly, Budapest.

Holt, D.A., Griffiths, D.V. (1992). Transient analysis of excavation in soil. Comput. Geotech.
13, No. 2, 159–174.

Houslby, G.T. (1991). How the dilatancy of soils affects their behaviour. Soil mechanics
report number 121/91, University of Oxford, UK.

Hueckel, T., Maier, G. (1979). Incremental boundary value problems in the presence of
coupling of elastic and plastic deformations: a rock mechanics oriented theory. Int. J.
Solids Struct. 13, I.

Hughes, D., Sivakumar, V., Glynn, D., Clarke, G. (2007). A case study: delayed failure of
deep cutting in lodgement till. Geotechnical Engineering 160, No. 4, 193–202.

Jardine, R.J. Potts, D.M., Fourie, A.B., Burland, J.B. (1986). Studies of the influence of non-
linear stress-strain characteristics in soil-structure interaction. Géotechnique 36, No. 3,
377–396.

Kirkebø, S. (1994). A numerical study of excavation in low permeable soils. Dr. Ing. thesis
Dept. of geotechnical Eng., The Norwegian Institute of Technology.

Kovacevic, N. (1994). Numerical analyses of rock fill dams, cut slopes and road
embankments. Ph.D. thesis, University of London.

214
 Chapter 9 – References

Kovacevic, N., Hight, D.W., Potts, D.M. (2007). Predicting the stand-up time of temporary
London Clay slopes at Terminal 5, Heathrow Airport. Géotechnique 57, No. 1, 63–74.

Kovacevic, N., Miligan, G.W.E., Menkiti, C.O., Long, M., Potts, D.M. (2008). Finite element
analyses of steep man-made cuts in Dublin boulder clay. Can. Geotech. J. 45, No. 4,
549–559.

Kovacevic, N., Jardine, R.J., Potts, D.M., Clukey, C.E., Brand, R.J., Spikula, D.R. (2012). A
numerical simulation of underwater slope failures generated by salt diapirism
combined with active sedimentation. Géotechnique 62, No. 9, 777–786.

Krabbenhoft, K., Karim, M.R., Lyamin, A.V., Sloan, S.W. (2012). Associated computational
plasticity schemes for nonassocited frictional materials. International Journal for
Numerical Methods in Engineering 89, 1089–1117.

Lagioia, R., Panteghini, A. (2014). The influence of the plastic potential on plane strain
failure. Int J Numer Anal Meth Geomech 38, 844–862.

Lambe, T.W. (1973). Predictions in soil engineering. Géotechnique 23, No. 2, 149–202.

Lambe, T.W., Whitman, R.V. (1979). Soil mechanics - SI version. Chichester: John Wiley.

Larsson, J., Larsson, R. (2000). Finite-element analysis of localization of deformation and

fluid pressure in elastoplastic porous medium. Int. J. Solids Structures 37, 7231–7257.

Le Pourhiet, L. (2012). Strain localization due to structural softening during pressure sensitive
rate independent yielding. Bull. Soc. Geol. Fr. 9, 1503–1518.

Lee, G,T,K,, Ng, C.W.W. (2005). Effects of advancing open face tunnelling on an existing
loaded pile. J Geotech. Geoenv. Eng. 131, No. 2, 193–201.

Lee, I.M., Lee, J.S., Nam, S.W (2004). Effect of seepage force on tunnel face stability
reinforced with multi-step pipe grouting. Tunn. Undergr. Space Technol. 19, No. 6, 551–
565.

Leroy, Y., Ortiz, M., (1989). Finite element analysis of strain localization in frictional
materials. International Journal for Numerical and Analytical Methods in
Geomechanics 13, 53–74.

Li, Y., Emeriault, F., Kastner, R., Zhang, Z.X. (2009). Stability analysis of large slurry shield-
driven tunnel in soft clay. Tunn. Undergr. Space Technol. 24, No. 4, 472–481.

Liu, X., Scarpas, A., Blaauwendraad, J. (2005). Numerical modelling of nonlinear response of
soil. Part 2: Strain localization investigation on sand. International Journal of Solids
and Structures 42, 1883–1907.

Lollino, P., Santaloia, F., Amorosi, A., Cotecchia, F. (2011). Delayed failure of quarry slopes
in stiff clays. Géotechnique 61, No. 10, 861–874.

215
Loret, B., Prevost, J.H. (1990). Dynamic strain localization in elasto-(visco-) plastic solids,
Part 1. General formulation and one-dimensional examples. Computer Methods in
Applied Mechanics and Engineering 83, No. 3, 247–273.

Loret, B., Prevost, J.H. (1991). Dynamic strain localization on fluid-saturated porous media.
J. Eng. Mech. 117, No. 4, 907–922.

Lourenço, S.D.N., Gallipoli, D., Toll, D.G., Augarde, C.E., Evans, F.D. (2011). A new
procedure for the determination of soil-water retention curves by continuous drying
using high-suction tensiometers. Can. Geotech. J. 48, 327–335.

Mair, R.J. (1979). Centrifugal modelling of tunnel construction in soft clay. PhD thesis,
University of Cambridge.

Michalowski, R.L. (2002). Stability charts for uniform slopes. J. Geotech. Geoenviron.
Engng. 128, No. 4, 351–355.

Muhlhaus, H.B. (1986). Shear Band Analysis in Granular Materials by Cosserat Theory. Ing.-
Arch. 56, 389–399.

Mühlhaus, H.B., Vardoulakis, I. (1987). The thickness of shear bands in granular materials.
Géotechnique 37, No. 3, 271–283.

Needleman, A. (1988). Material rate dependence and mesh sensitivity in localization

problems. Comput. Meth. Appl. Mech. Engrg. 67, 69–85.

Ng, C.W.W., Lee, G.T.K. (2002). A three-dimensional parametric study of the use of soil
nails for stabilizing tunnel faces. Computers and Geotechnics 29, 673–697.

Nordal, S. (2008). Can we trust numerical collapse load simulations using non-associated
flow rules? In: Proc. International Association for Computer Methods and Advances in
Geomechanics, Goa, 892–899.

Nova, R. (2004). The role of non-normality in soil mechanics and some of its mathematical
consequences. Computers and Geotechnics 31, No. 3, 185–191.

Ortiz, M., Leroy, Y., Needlemann, A. (1987). A finite element method for localized failure
analysis. Computer Methods in Applied Mechanics and Engineering 61, No. 2, 189–
214.

Perazzelli, P., Leone, T., Anagnostou, G. (2014). Tunnel face stability under seepage flow
conditions. Tunnelling and Underground Space Technology 43, 459–469.

Pietruszczak, S., Mroz, Z. (1981). Finite element analysis of deformation of strain softening
materials. International Journal for Numerical Methods in Engineering 17, 327–334.

Potts, D.M., Gens, A. (1984). The effect of the plastic potential in boundary value problems
involving plane strain deformation. Int J Numer Anal Meth Geomech 8, 259–286.Potts,

216
 Chapter 9 – References

D.M., Kovacevic, N., Vaughan, P.R. (1997). Delayed collapse of cut slopes in stiff clay.
Géotechnique 47, No. 5, 953–982.

Potts, D.M., Zdravković, L.T. (2001a). Finite element analysis in geotechnical engineering:
Theory. Thomas Telford Limited, London.

Potts, D.M., Zdravković, L.T. (2001b). Finite element analysis in geotechnical engineering:
Application. Thomas Telford Limited, London.

Raniecki, B., Bruhns, O.T. (1981). Bounds to bifurcation stresses in solids with non-
associated plastic flow law at finite strain. J. Mech. Phys. Solids 29, No. 2, 153–172.

Rice, J.R (1968). Mathematical analysis in the mechanics of fracture. In: Fracture: an
advanced treatise Liebowitz H Mathematical Fundamentals, (ed. New York: Academic
Press), vol. 2, 191–311.

Rice, J.R. (1975). On the stability of dilatant hardening for saturated rock masses. J. Geophys.
Res. 80, 1531–1536.

Rice, J.R. (1976). The localization of plastic deformation. In: Theoretical and Applied
Mechanics, (ed. W. T. Koiter), NorthHolland, Amsterdam, 207–220.

Rice J.R., Simons D.A. (1976). The stabilization of spreading shear faults by coupled
deformation–diffusion effects in fluid-infiltrated porous materials. J. Geophys. Res. 81,
5322–5334.

Roscoe, K.H., Burland, J.B. (1968). On the generalised stress-strain behaviour of ‘wet’ clay,
Engng Plasticity, Cambridge Univ. Press, 535–609.

Roscoe, K.H. (1970). The influence of strains in soil mechanics. Géotechnique 20, No. 2,
129–170.

Rowe, P.W. (1969). The relation between the shear strength of sands in triaxial compression,
plane strain, and direct shear. Géotechnique 19, No. 1, 75–86.

Rudnicki, J.W., Rice, J.R. (1975). Conditions for the localization of deformation in pressure-
sensitive dilatant materials. Journal of Mechanics and Physics of Solids 23, 371–394.

Rudnicki, J.W. (1983). A formulation for studying coupled deformation - Pore fluid diffusion
effects on localization of deformation. In: Proceedings of The Symposium on the
Mechanics of Rocks, Soils, and Ice (ed. S. Nemat-Nasser), American Society of
Mechanical Engineers 35–44.

Rudnicki, J.W. (1991). Coupled deformation-diffusion effects in the mechanics of faulting

and failure of geomaterials. Applied Mechanics Reviews 54, No. 6, 483–502.

Rudnicki, J.W., Koutsibelas, D.A. (1991). Steady propagation of plane strain shear cracks on an
impermeable plane in an elastic diffusive solid. International Journal of Solids and
Structures 27, 205–225.

217
Runesson, K., Mroz, Z. (1989). A note on nonassociated plastic flow rules. International
Journal of Plasticity 5, No. 6, 639–58.

Ruse, N.M. (2004). Räumliche Betrachtung der Standsicherheit der Ortsbrust beim
Tunnelvortrieb. Mitteilung 51 Universität Stuttgart, Institut für Geotechnik.

Sanavia, L. (2009). Numerical modelling of a slope stability test by means of porous media
mechanics. Engng Computation. 26, No. 3, 245–266.

Schofield, A.N., Wroth, C.P. (1968). Critical state soil mechanics. McGraw-Hill Book
Company (UK).

Schofield, A.N. (1980). Cambridge geotechnical centrifuge operations. Géotechnique 30, No.
3, 227–268.

Schrefler, B.A., Sanavia, L., Majorana, C.E. (1996). A multiphase medium model for
localization and post-localization simulation in geomaterials. Mechanics of Cohesive-
Frictional Materials and Structures 1, 95–114.

Schrefler, B.A., Zhang, H.W., Pastor, M., Zienkiewicz, O.C. (1998). Strain localisation
modelling and pore pressure in saturated sand samples. Computational Mechanics 22,
No. 3, 266–280.

Schuerch, R., Anagnostou, G. (2012). Tunnel face stability under transient conditions: stand-
up time in low permeability ground. In: Proceedings of the 22nd European Young
Geotechnical Engineers Conference, Gothenburg.

Schuerch, R., Anagnostou, G. (2013a). Analysis of the stand-up time of the tunnel face. In
Proc. ITA-AITES World Tunnel Congress, WTC 2013, Geneva.

Schuerch, R., Anagnostou, G. (2013b). The influence of the shear strength of the ground on
the stand-up time of the tunnel face. In: Proc. TU-Seoul 2013 International Symposium on
Tunnelling and Underground Space Construction for Sustainable Development, Seoul.

Schuerch, R., Anagnostou, G. (2015a). Delayed failure identification by coupled hydraulic-

mechanical numerical analyses. In: Proc. 14th IACMAG, 1061–1066, Kyoto.

Schuerch, R., Anagnostou, G. (2015b). Failure propagation and mesh-dependency in coupled

hydraulic-mechanical transient problems. In: Proc. 14th IACMAG, 329–334, Kyoto.

Schuerch, R., Poggiati, R., Anagnostou, G. (2016). Design diagrams for estimating tunnel
face stand-up time in water-bearing ground. In: Proc. WTC16, San Francisco.

Schuerch, R., Poggiati, R., Maspolo, P., Anagnostou, G. (2016) Design charts for estimating
tunnel face stand-up time in soft ground tunnelling. In: Proc. 2nd International
Conference on Tunnel Boring Machines in Difficult Grounds, Istanbul.

218
 Chapter 9 – References

Schuerch, R., Vrakas, A., Anagnostou, G. (2017). On manifestations of delayed failure and
the effect of dilatancy in transient poro-elasto-plastic analyses of slopes and excavations.
Géotechnique (http://dx.doi.org/10.1680/jgeot.15.P.238]).

Shin, J.H., Potts, D.M., Zdravković, L.T. (2005). The effect of pore-water pressure on NATM
tunnel linings in decomposed granite soil. Canadian Geotechnical Journal 42, 1585-1599.

Sitarenios, D.L., Kavvadas, M. (2016). The interplay of face support pressure and soil
permeability on face stability in EPB tunneling. In: Proc. WTC16, San Francisco.

Sevaldson, R.A. (1956). The slide in Lodalen, October 6th, 1954. Géotechnique 6, No. 4,
167–182.

Skempton, A.W. (1964). Long-term stability of clay slopes. Géotechnique 14, No. 2, 77–101.

Skempton, A.W. (1977). Slope stability of cuttings in brown London Clays. Proc. 9th int.
conf. on Soil Mechanics and Foundation Engineering, 261–270, Tokyo.

Sterpi, D., Cividini, A., Donelli, M. (1995). Numerical analysis of a shallow excavation in
strain softening rock. In: Proc. 8th Int. Congr. On Rock Mechanics, Balkema,
Rotterdam, Vol. 2, 545–549.

Sterpi, D. (1999). An analysis of geotechnical problems involving strain softening effects. Int
J. Numer. Anal. Meth. Geomech. 23, 1427–1454.

Ströhle, P.M., Vermeer, P.A. (2010). Tunnel face stability with groundwater flow. In: Proc
7th European Conference on Numerical Methods in Geotechnical Engineering,
Trondheim.

Stewart, D.P. and Randolph, M.F. (1994). T-bar penetration testing in soft clay. ASCE
Journal of Geotechnical Engineering 120, No. 12, 2230–2235.

Taylor, D.W. (1937). Stability of earth slopes. J. Boston Soc. Civ. Engng 24, No. 3, 197–246.

Taylor, D.W. (1948). Soil mechanics. John Wiley & Sons, New York.

Thakur, V. (2011). Numerically observed shear bands in soft sensitive clays. International
Journal Geomechanics and Geoengineering 6, No. 2, 131–146.

Vaid, Y.P., Eliadorani, A. (1998). Instability and liquefaction of granular soils under
undrained and partially drained states. Canadian Geotechnical Journal 35, 1053–1062.

van Genuchten, M.T. (1980). A closed-form equation for predicting the hydraulic
conductivity of unsaturated soils. Soil Science Society of America Journal 44, No. 5, 892–
898.

Vardoulakis, I. (1980). Shear band inclination and shear modulus in biaxial tests. Int. J. Num.
Anal. Meth. Geomech. 4, 103–119.

219
Vaughan, P.R., Walbancke, H.J. (1973). Pore pressure changes and the delayed failure of
cutting slopes in overconsolidated clay. Géotechnique 23, No. 4, 531–539.

Vaughan, P.R. (1994). Assumption, prediction and reality in geotechnical engineering.

Géotechnique 44, No. 4, 573–609.

Vermeer, P.A., Verruijt, A. (1981). An accuracy condition for consolidation by finite

elements. Int. J. Num. Anal. Meth. Geomech. 5, No. 1, 1–14.

Vermeer, P.A. (1982). A simple shear-band analysis using compliances. In: Proc. IUTAM
Conf. Deform. Failure Granular Matls., Delft, Netherlands, 493–499.

Vermeer, P.A., De Borst, R. (1984). Non-associated plasticity for soils, concrete and rock.
Heron 29, No. 3, 3–64.

Vermeer, P.A., Van Langen, H. (1989). Soil collapse computations with finite elements. Ing.
Archiv 59, No. 3, 221–236.

Vermeer, P.A. (1990). The orientation of shear bands in biaxial tests. Géotechnique 40, No. 2,
323–336.

Vermeer, P.A. , Ruse, N. (2001). Die Stabilität der Tunnelortsbrust in homogenem Baugrund.
Geotechnik 24, No. 3, 186–193.

Vermeer, P.A., Ruse, N., Marcher, T. (2002). Tunnel heading stability in drained ground.
Felsbau 20, No. 6, 8–18.

Vrakas, A., Anagnostou, G. (2015). Finite strain elastoplastic solutions for the undrained
ground response curve in tunnelling. Int. J. Num. Anal. Meth. Geomech. 39, No. 7, 738–
761.

Weber, T.M. (2007). Modellierung der Baugrundverbesserung mit Schottersäulen. PhD

Thesis Nr. 17321, ETH Zurich.

Wind, G.P. (1968). Capillary conductivity data estimated by a simple method. In: Proc.
Wageningen Symp., Paris.

Wehnert, M. (2006). Ein Beitrag zur drainierten und undrainierten Analyse in der
Geotechnik. Universität Stuttgart, Mitteilung 53 des Instituts für Geotechnik.

Wongsaroj, J. (2005). Three-dimensional finite analysis of short- and long-term ground

response to open face tunnelling in stiff clay. PhD Thesis, Cambridge University.

Zienkiewicz, O.C., Zhu, J.Z. (1991). Adaptively and mesh generation. International Journal
for Numerical Methods in Engineering 32, No. 4, 783–810.

220
 Curriculum Vitae

Curriculum Vitae

Personal Information
First Name Roberto
Family Name Schuerch
Date of Birth 8.10.1985
Place of Birth Lugano, Switzerland
Citizenship Swiss

Education
2012 – 2016 PhD, ETH Zurich, Switzerland
2007 – 2009 Master of Science, ETH Zurich, Zurich, Switzerland
2004 – 2007 Bachelor of Science, EPFL, Lausanne, Switzerland
2000 – 2004 High School, Lugano, Switzerland

Work Experience
2016 – * Pini Swiss Engineers, Zurich, Switzerland
2009 – 2016 Teaching assistant, research activities and scientific services,
Chair of Underground Construction, Prof. Dr. G. Anagnostou,
ETH Zurich, Switzerland
2008 – 2009 Teaching assistant, Chair of Geotechnical Engineering,
Prof. Dr. S. Springman, ETH Zurich, Switzerland
2008 – 2009 Teaching assistant, Institute of Structural Engineering,
Prof. Dr. M. Fontana, ETH Zurich, Switzerland

Memberships
2016 – * STSym (Swiss Tunnelling Society, young members)
2015 – * ITA Working Group 17, Long Tunnels at Great Depth
2012 – 2015 Strategic Group, Pini Swiss Engineers, Lugano, Switzerland

221