Information documents

TUNNEL FACE STABILITY

Collapse mechanisms, design
and construction principles

MINISTÈRE
CHARGÉ
DES TRANSPORTS
December 2022 Centre d’Études des Tunnels (Centre for Tunnel Studies)
www.cetu.developpement-durable.gouv.fr
DISCLAIMER
The purpose of the information documents is to provide information on a new or insufficiently covered technique or problem. The reader
may find it useful in his or her work. The content and any conclusions presented should not be considered as recommendations by
the CETU. Although every effort has been made to ensure the reliability of the sources used, neither CETU nor the authors of
the document can be held responsible.

2
 Information documents

TUNNEL FACE STABILITY

Collapse mechanisms, design
and construction principles

December 2022

Centre d’Études des Tunnels (Centre for Tunnel Studies)

25 Avenue François Mitterrand
69500, Bron, France
Tel. +33 (0)4 72 14 34 00
Fax. +33 (0)4 72 14 34 30
[email protected]
www.cetu.developpement-durable.gouv.fr

3
 TABLE OF CONTENTS
1 CONTEXT AND PURPOSE OF THIS DOCUMENT 5

2 GENERAL PRINCIPLES FOR STABILITY STUDIES 7

2.1 Problem statement 7
2.2 Hypothesis and notations 8
2.3 Calculation configurations 9
2.4 Safety factor approach 10
2.5 Modelling approaches 11

3 ASSESSMENT OF STABILITY CONDITIONS 13

3.1 Discontinuous medium 13
3.2 Continuous undrained medium 14
3.3 Continuous drained medium without seepage 17
3.4 Continuous drained medium with seepage 22

4 ON-SITE OBSERVATIONS AND CONSTRUCTION MEASURES 26

4.1 Construction methods in conventional tunnelling 26
4.2 Monitoring methods in conventional tunnelling 28
4.3 Construction methods in conventional tunnelling 28
4.4 Monitoring methods for pressurized face tunnelling 29

5 SUMMARY 30

6 GLOSSARY 32

7 REFERENCES 33

8 APPENDIX: FACE STABILITY EXAMPLES 36

4
1 CONTEXT AND PURPOSE
OF THIS DOCUMENT
The tunnel face exerts its immediate influence on the Consequently, a good understanding of the mechanisms
“active” zone of tunnelling in front of and behind the face by which the ground is deformed is necessary in order
(AFTES WG16, 2020). Pushed back as the tunnelling process to avoid initiating collapse mechanisms or to predict the
advances, the majority of shifts in ground stresses occur in consequences when they can not be avoided. This under-
this zone. standing will:

For deep tunnels, stress redistribtion (which may or may not • ensure the safety of the site (primarily that of underground
require interventions at the face, depending on the quality of the workers) and its surroundings (especially neighbouring
ground), enhances the formation of the arch, which helps to restore structures on the surface),
the overall balance throughout the zone. For shallow tunnels, • allow the installation of pre-support structures at the
these arching effects are obtained with greater difficulty and front and support structures at the back of the working face,
stresses that are not correctly dealt with can give rise to surface • help the project comply with time and budget requirements
settlements, causing damage to neighbouring structures. by limiting overbreaks and avoiding a situation with
The considerable deformation of poor ground at great unpredicted and uncontrolled stresses,
depth can lead to a decompressed zone at the edge of the • contribute to the long term stability of the construction,
excavation area. Gravity-related phenomena can occur in by limiting deterioration of the surrounding ground at the
this decompressed zone as well as near the surface. edge of the excavation zone and therefore the stresses
on the final lining structures.
Numerous instabilities have already been observed in different
geotechnical contexts. CEDD (2015) draw up, for example, The aim of this information document is to provide a general
a (non-exhaustive) list of more than sixty instabilities that overview of failure mechanisms, available modelling approaches
appeared between 1977 and 2014, at the face or at the wall. and their scope of application, as well as suitable calculation tools
The sheets available in the appendix provide some additional, for each project. Indeed, this topic has not yet been the subject
fairly recent examples. The following findings result from of a standard (Eurocode 7 standard for example) nor a French
these examples: (AFTES) or international (ITA-AITES) design recommendation.
The document that comes closest is the recommendation
• face instabilities are possible in a wide variety of from the DAUB (German counterpart of the AFTES) published
geotechnical contexts (homogeneous or mixed faces, in 2016, but which remains specific to the case of TBMs and
fractured rockmass, soft grounds, etc.), and for all boring introduces elements specific to the German context.
methods (conventional tunnelling, open face shield,
earth pressure balance or slurry shield), The present document is mainly intended for engineers in design
• underground, the volumes involved can range from a offices, project managers and contractors. By providing a summary
few cubic decimetres for localized instabilities to several of a broad litterature review, the various instability mechanisms are
hundred cubic metres for global instabilities affecting the presented and structured according to the nature of the ground
entire front or even the overburden, (rockmass with varying degrees of fracturing, loose ground such
• the shape of the mechanisms depends on the nature as gravelly soil or fine soil, etc.) and the excavation method
of the ground: blocks delimited by the pre-existing (conventional method with or without reinforcement of the face
discontinuities in rocks, mechanisms localized near or pre-support, pressurised or unpressurised TBM). It can also
the face in powdery grounds (with regressive evolution be useful to engineers in charge of monitoring the works (support
towards the surface) and more voluminous mechanisms manager on the contractor’s side, geotechnical supervision on
in cohesive clayey grounds, the project manager’s side) in order to understand the behaviour
• instabilities induced by partial or inappropriate control of the ground during excavation, then choose and adapt the best
of the face stability can progress to the surface with a construction measures in response to the encountered mechanisms.
delayed effect in time and space,
• the consequences of face instabilities are very variable, For each configuration, the different methods for calculating face
ranging from being “almost negligible” to “very significant” stability are described and an explicit empirical formulation for
with delays (up to several months) or significant “operational” purposes is proposed. This document aims to
additional costs (up to several million euros), and casualties be easily usable by engineers by giving an overview of the
(as underground workers are exposed to risks). different existing methods and their field of application.

5
1
This document does not address the essential issue of the If studies highlight unstable conditions at the face or if instability is
choice of calculation parameters, in particular the shear strength encountered during the works, different construction processes
parameters of the ground. In accordance with Eurocode 7-1 can be used to improve the stability of the face or work safety.
(AFNOR, 2005), this choice must be based on the measured The main methods used in both conventional and mechanized
values, values obtained from on-site tests and laboratoty tests, tunnelling are thus outlined, as well as necessary monitoring
supplemented by the lessons learned from experience. For stability actions during works, in order to confirm the geotechnical
issues, estimates need to be cautious, in particular as regards assumptions made during the studies and to determine the
effective cohesion. In case of significant uncertainty, calculations actual stability conditions.
must be carried out to provide values within various ranges.

6
2 GENERAL PRINCIPLES
FOR STABILITY STUDIES

2.1 PROBLEM STATEMENT

Consider the ideal case of a tunnel perfectly supported in its Figure 1 gives a schematic view and also makes the link with surface
interior zone, with only the face free to move and on which a movements in the case of a shallow tunnel above the water table:
certain pressure is exerted (which can be zero in the case of • if the pressure exerted on the face is less than the horizontal
an unsupported face). geostatic stress (σT < σh0), an extrusion movement of the
face appears and may lead to surface settlements. The
In the case of a TBM with a pressurized face, this pressure face may collapse with the appearance of a sinkhole for
is actually exerted by compressed air (air pressure), excavated extreme cases,
material (earth pressure) or bentonite (slurry pressure, with • on the other hand, if the pressure exerted on the face is
greater than the horizontal geostatic stress (σT > σh0), an
conventional or “heavy” slurry). In conventional tunnelling, an
upheaval movement of the tunnel face appears, and can
“equivalent” confinement pressure can be considered to take
lead to a surface upheaval in front of the face (this case,
into account face bolting. however, is only possible with an earth pressure balance
tunnel boring machine under a very low overburden).
Depending on the extent of this pressure, extrusion
In a global approach to evaluate the acceptable range of
movements (towards the tunnel cavity) or upheaval movements
pressure within the scope of a tunnelling project, a stability range
(towards the front of the face when viewed from inside
(shown in red in Figure 1) is evaluated, which prevents the
the tunnel in the advancement direction) may occur. Their occurence of failure mechanisms by collapse or blow-out. In
amplitude notably depends on the geometry of the tunnel, the addition, the evaluation of a range of “admissible displacements”
initial state of stresses and the geotechnical characteristics (in green in Figure 1) is needed for shallow tunnels, depending
of the ground. on the sensitivity of neighbouring constructions.

Figure 1: Links between face pressure, instability mechanisms at the face and displacements generated in proximity to neighbouring
structures in the case of a shallow tunnel.

7
2
If the air, earth or slurry pressure at the face do not counteract here as it is an issue in its own right. The recommendations
pore pressures in the ground, two related effects may occur: of AFTES WG16 (2020) provide information on this subject:
(i) seepage forces that increase the instability of the face, and evaluation of the intrinsic sensitivity of neighbouring constructions,
(ii) potentially large volumes of water that can lead to the evaluation of the displacements induced by underground works
flooding of the works site. and evaluation of the associated level of damage.

This information document is mainly devoted to the lower The case of face pressures higher than the existing horizontal
limit of the stability range, i.e. the minimum pressure to be stresses is easier to deal with from an operational point of view.
exerted to prevent collapse. Indeed, the theoretical approaches (limit equilibrium or yield
design theory) dedicated to face blow-out imply a total
The lower limit of the “acceptable” range with respect to the mobilization of the shear stresses along the collapsed surfaces,
sensitivity of any neighbouring structures in the case of a shallow which always leads to maximum face pressures much higher
tunnel is based on semi-empirical approaches (so-called “volume than those leading to significant upheaval movements on the
loss” method) or numerical calculations. This point is not addressed surface (Berthoz et al., 2012).

2.2 HYPOTHESIS AND NOTATIONS

The study of face stability is usually based on the following • Excavation method and confinement pressure:
assumptions with the notations summarised in Figure 2: – Excavation can be carried out with a TBM or by
conventional tunnelling. In both cases, the average
• Geometry: The face is equated to a disc with a diameter D. advancement rate of the face (integrating the excavation
In conventional tunnelling, the actual excavation is not circular and support phases) is noted vavct,
and a disc, (or even a rectangle), equivalement to the area – a constant pressure σT is considered in order to represent
covered by the face is generally used. If the excavation is very the boundary conditions on the surface of the face and
narrow (horizontally or vertically), the choice may be more along the unsupported span. In practice, we consider
difficult, and is to be made with regard to the geometry of the the value exerted in the tunnel axis. For tunnels with
mechanism involved. In addition, C is the overburden thick- very large diameters (greater than about ten meters),
ness, H is the depth of the tunnel axis, and d is the length of it is important to take into account the face pressure
the unsupported span (area not supported during excavation). gradient over the height of the face, or at least, ensure
• Hydrogeology: We consider the groundwater table to be that the pressure retained in the axis is representative
located at a height Hw above the tunnel axis. In the case of the real boring conditions.
of a confined aquifer, an equivalent height Hw deducted from
the hydrostatic pressures can be estimated in order to The minimum value of this pressure σT is calculated so that
include it in the same case study. the stability of the face is ensured. If this value is negative
• Geology and geotechnical engineering: the face, whether (with the sign conventions of soil and rock mechanics, i.e. positive
homogeneous or composed of different facies, may comprise: stresses in compression), the face is considered theoretically
– “mildly fractured” rockmass with few discontinuities at the self-stable. If it is positive, the face is not self-stable and suitable
scale of the face, and behaving as discontinuous media construction processes must be implemented to ensure stability.
at the scale of the construction. In this case, the stability These processes are designed on the basis of this pressure
is conditioned by the density γ of the unstable dihedra, the value with a safety margin to be defined.
orientation of the discontinuities θ and the shear strength
along the discontinuities, in particular the friction angle φ’disc,
– “loose” ground (soils, highly fractured rocks), which
can be considered as continuous at the scale of the
construction, or as intact rock, continuous by definition.
In this case, each facies is generally assimilated to a
homogeneous medium of density γ, of permeability k, C
of perfectly plastic rigid behaviour parametrised by a H
)
Mohr-Coulomb criterion (c′, φ′) under drained conditions, H OU
Tresca (Cu) under non-drained conditions, or possibly )
Hoek & Brown for rocky soils. TBM OU
• Surface load: σs denotes the constant pressure exerted T

at the surface, so as to simulate the weight of the OU

buildings on the surface (considering 12 kPa per building
floor for example), excess site loads (generally around
20 kPa) or excess operating loads (around 10 kPa for
a roadway for example). Figure 2: Problem position and notations.

8
2.3 CALCULATION CONFIGURATIONS
Different calculation configurations are derived from the CASE C
previous assumptions. These four cases are detailed in Continuous drained medium without seepage
chapter 3.
In the case of more permeable ground or slower boring (with
the above example values, k > 10-6 m/s and vavct < 0.1 m/h),
CASE A the ground is solicited in a “drained” state (i.e. without pore
Discontinuous medium overpressures). In this case, the stability of the face is conditioned
by the effective stresses exerted in the ground.
In mildly fractured rockmasses, the network of discontinuities
cuts potentially unstable dihedrals. Blocks falls or slides can
There is no seepage of pore water into the ground (water
appear at the face and/or in the unsupported span area. The
pressures remain equal to the initial hydrostatic pressures)
geometry of the dihedra that may become detached depends
in the following cases:
on the geometry of the face and the orientation of the main
families of discontinuities. Without support, the stability of • boring outside the water table, naturally, or induced by
these dihedra depends on their geometry, on the density of the work (preliminary water table lowering, drainage
the rock matrix and on the shear strength which can be while advancing leading to the absence of hydrostatic
mobilised on the discontinuities surfaces. pressures up to at least one diameter in front of the face),
• pressurized TBM tunnelling under the water table in
In rocky grounds with little overall fracturing, more fractured relatively coarse soils, apart from uncontrolled conditions
zones or fault zones with degraded, altered materials may corresponding to the “drained with seepage” case.
exist, in which it may be relevant to retain mechanisms of
type B/C/D below. The face pressure to be exerted is then the sum of the pressure
exerted by the ground (calculated as effective stresses,
with the assumption of a shear strength isotropy) and the
CASE B (hydrostatic) pressure of the water, i.e.:
Continuous undrained medium
In the case of TBM or conventional tunnelling progressing
at a high speed in ground with low permeability, it may be
reasonable to consider that excavation is carried out under [2]
“undrained” conditions.
Outside the water table, the effective stresses are equal to the
In this context, the calculation is carried out in total stresses total stresses and the solution is reduced to:
with the total density γ of the ground and the undrained shear
strength of the ground (parametrized by an “undrained cohesion”
Cu, and a zero friction angle φu). The solution to the problem
(minimum face pressure preventing face collapse σT-eff) is [3]
written as follows:
CASE D
Continuous drained medium with seepage
[1] In conventional tunnelling (or possibly with an open face TBM)
below the water table, the hydraulic imbalance between the
The choice between “undrained” and “drained” conditions inside of the tunnel (zero pressure) and the ground (initial pore
must be considered specifically for each geotechnical pressure equal to the hydrostatic pressure) leads to seepage
context and each actual construction phase. Based on directed towards the face.
hydro-mechanical calculations, Anagnostou & Kovari (1996)
give the following limit to this calculation case: permeability k With a pressurized face TBM below the water table, the seepage
less than 10-7 to 10-6 m/s and penetration advance speed vavct towards the face is prevented: (a) due to the low permeability
greater than 0.1 to 1 m/h. It is important to take into account any of the cake, control of the mucking rate and control of the air
stoppages in excavation works (due to hyperbaric operations, bubble pressure in the crown of the working chamber with a
weekends, site incidents...), as by slowing the advancement slurry pressure balance shield TBM, (b) thanks to the right
of the face, these stoppages may mean that the "undrained" sizing of the screw conveyor (length and diameter) coupled
assumption is unrealistic. If there is any doubt about the reality with the addition of possible admixtures reducing permeability
of undrained conditions, both calculations (drained and in the working chamber with an earth pressure balance shield
undrained) can be carried out to judge the sensitivity of TBM. Pore overpressures can be generated at the face if these
the face stability conditions with regard to this assumption. conditions are insufficiently controlled.

9
2
In all of these conditions, the distribution of pore pressures is also a factor, and the duration after which the face becomes
(thus also of effective stresses) depends on the conditions at unstable can be assessed. The general shape of the minimum
the hydraulic boundaries of the model (function of Hw) and on pressure preventing face collapse is therefore as follows:
the ground permeability (k). The pressure to be exerted on the
face to ensure its stability then depends on the seepage
forces. In steady state, these depend only on the hydraulic
gradient (function of Hw and k). In transient conditions, time (t) [4]

2.4 SAFETY FACTOR APPROACH

Three main approaches are possible in absolute terms: In soft ground, the German recommendation (DAUB, 2016),
which deals more specifically with pressurised TBMs,
• with partial safety coefficients on the effects of the actions: recommends weighting the effects of the actions, with partial
in this case, the calculations are carried out without safety safety factors of 1.5 on the earth pressure σT-eff and 1.05 on
coefficients, then the forces obtained related to the
the water pressure. This approach may be relevant when
stability of the ground (or the stresses σT-eff if the face is
σT-eff >> 0, but is questionable when the face is at the limit of
homogeneous and totally below the water table) and to
stability, the condition σT-eff = 0 being true regardless of the
the water pressure are multiplied by safety coefficients,
safety coefficient considered.
• with partial safety coefficients applied before calculation,
directly on the ground shear strength parameters,
By analogy with the problems of slope stability, an overall
• by seeking an overall safety coefficient, thanks to the
safety factor can be estimated by relating the driving forces
calculation of the ratio between the driving forces and the
to the resisting forces. A calculation approach with partial
resisting forces without weighting the ground parameters
coefficients on the resistance of the ground can also be
or the effects of the actions.
considered in soft ground (Paternesi et al., 2017) since
As “category 3 geotechnical structures” (§2.1 of Eurocode 7-1), the degradation of the ground parameters (in particular the
tunnels are currently outside the scope of Eurocode 7, however friction angle) increases the geometry of the mechanism
the approaches used in practice are compatible with its principles. but does not change the overall behaviour, which remains
For the design of tunnel support and lining structures using a continuous medium. This is the approach adopted in this
“displacement” approaches (finite element type), AFTES has document for soft ground with the following partial safety
taken a position by recommending the weighting of the effects coefficients:
of actions: normal forces and bending moments in the structures
• 1.25 on c’ and tan(φ’) in drained conditions,
(AFTES WG29, 2007). This choice is motivated by the desire
• 1.4 on Cu in undrained conditions.
not to induce plastic behaviour by taking into account degraded
parameters for the ground, which would be likely to artificially
In the case of pressurised TBM excavation, the calculated
disturb the distribution of forces between the ground and the lining.
minimum pressures σT can be increased by 10 kPa in the
For the study of face stability in a poorly fractured rock case of a slurry shield (SS) and 30 kPa in the case of an earth
mass (dihedral falls and slides) by “failure” methods (limit pressure balanced shield (EPB) in order to take into account
equilibrium and yield design methods), a “global safety factor” the difficulties in maintaining and controlling the applied face
type approach is recommended. Indeed, a degradation of the pressure (DAUB, 2016). It is also reasonable to increase the
rock mass parameters (level of fracturing, state of alteration, water pressures by 5% in the case of a continuous drained
shear strength on discontinuities, etc.) could lead to a significant medium without flow (Case C, §.2.3).
modification of the geometry of the blocks, or even to a radical
modification of the mechanisms in play, the ground being able However, the recommended approach of applying a partial
to pass from a behaviour of discontinuous medium type to that weighting on the shear strength parameters in soft ground
of a continuous medium. By analogy with radial bolting in the is clearly not recommended for the design of support and
interior zone (AFTES WG30, 2021), an overall safety factor of lining structures, due to the modification of the soil/structure
the order of 1.6 to 1.8 can be sought at the face, for example. interaction phenomena.

10
2.5 MODELLING APPROACHES
All the existing models for studying the stability of the working face of materials somewhat complex. Validating calculation models
can be grouped into four main categories, the main principles can prove difficult, as data obtained from feedback on instabilities
of which are given below, with some examples in Illustration 3. encountered in real full-scale structures is often incomplete and
in particular data on deformation and failure mechanisms, as well as
data on material properties. Some information from face instabilities
Physical models and experiments (Exp) observed in situ during excavation work using conventional
Several test campaigns have been carried out on centrifugal methods and tunnel boring machines is given in the appendix.
and non-centrifugal models (in England and France, and more
recently in Asia), in purely cohesive, purely frictional, and then
Models based on the Limit Equilibrium
frictional cohesive and stratified soils. The failure of the working
face is generally caused by the displacement of a rigid wall or the
approach (LE)
deflation of a flexible membrane at the working face (Figure 3a). This approach is the one most commonly used in current practice:
it consists of directly assessing the forces in play (Figure 3b).
These models are valuable since on the one hand, they help Its implementation in soil masses, or materials that can be
to identify the kinematics of the fracture mechanism that considered as continuous at the scale of the structure, requires
should be considered in semi-empirical and analytical models strong assumptions regarding:
(Limit Equilibrium (LE) and Yield Design (YD) described below),
and on the other hand, they help to evaluate the face collapse • the geometry of the mechanism: the face being assumed
pressures under controlled conditions, although they do not cover rectangular, the most classical mechanism (Horn, 1961)
the whole range of variations of the different parameters. is made up of a wedge, delimited by an inclined plane in
front of the face and two triangular lateral faces, as well
Respecting similarities in conditions and minimizing scale effects as a vertical right prism with a rectangular base rising to
makes experimental set-ups and the characterisation and installation the surface (3D extension of the Terzaghi mechanism),

(a) Example of a reduced scale model (b) Force balance on the wedge, mobilised at the front
based on cohesionless soil in a centrifuge of the face in limit equilibrium approaches
(Chambon & Corté, 1994) (according to Anagnostou & Kovari, 1996)

(c) An example of a 3D mechanism considered in (d) Example of a 3D finite elements numerical model
the yield design kinematic approach (Wong & Subrin, 2006) (Vermeer et al., 2002)

Figure 3: Examples of face stability models.

11
2
• the profile of the stresses (or their global resultant) applied This approach, although not widely used, has the merit of
on each face of the mechanism: the shear strength is defined proposing possible value brackets and enabling collapse
by the Coulomb strength criterion (possibly Tresca) and conditions to be determined by means of a rigorous approach,
the normal stresses on the fracture surfaces presuppose hence enabling the relevance of limit equilibrium approaches
a value of the ratio between the horizontal and vertical (based on more numerous assumptions) to be verified.
stresses in the ground. The virtual failure mechanisms constructed “by hand” in
the framework of the kinematic approach can be refined by
The principle consists in comparing the driving forces (weight
numerical approaches.
of the wedge and vertical stress on its upper surface, possible
surface overload, flow forces) with the mobilisable resistant forces
(shear resistance of the ground on the different faces of the wedge Numerical models (Num)
and face pressure simulating the effect of a confinement).
The numerical models (Figure 3d) considered consist in
It is possible to take into account horizontally stratified ground (e.g. solving the system with partial derivatives resulting from
in the context of a sedimentary basin) and to consider the effect of equilibrium equations, the ground’s behaviour law, boundary
a water table without changing the failure mechanism considered. conditions and initial conditions, by rewriting the problem in
discrete form (in finite elements or explicit finite differences).
The finer the mesh used, the more precise the solution.
Models developed within the framework Given the three-dimensional nature of the mechanisms
of the Yield Design (YD) theory involved at the face, the calculation times required to obtain
Face stability conditions can be estimated by the yield design acceptable precision may be significant. Special attention should
theory (Salençon, 1990), the results of which are similar to be paid to the representativeness of the assumptions made:
those of the limit analysis, if it is assumed that the associated radial blocking of nodes at the periphery of the face, vertical
ground has perfectly plastic rigid behaviour. face pressure gradient (especially in the case of large diameter
tunnels), mechanical characteristics of the ground, etc.
Based on the determination of statically admissible stress fields, the
static approach, known as the interior approach, leads to a default This type of deformation approach is used to estimate
estimate, therefore from a safety point of view, of the domain of the collapse conditions, either by simulating the progressive
bearable loads. In other words, if it is possible to find a stress reduction of the pressure at the face, or by using a method
field in the ground that verifies the equilibrium equations, the stress consisting in reducing the resistance parameters. Compared to
boundary conditions and the strength criterion, then the stability of the external approach of the Yield Design Theory, the search for
the working face is assured. In the case of a collapse mechanism, the deformation and fracture kinematics is carried out among
as the face pressure is a loading parameter that opposes instability, an “infinite” number of possible geometries, with optimisation
this approach leads to an upper bound of the face collapse pressure. resulting from the numerical calculation without fixing the
fracture geometry a priori.
In contrast, the kinematic approach, known as the external
approach, is derived from the proposal of kinematically admissible
Starting from an initial state of geostatic stress in the ground,
virtual collapse mechanisms (Figure 3c), i.e. verifying the velocity
face collapse can be sought by progressively decreasing the
boundary conditions. For a family of mechanisms characterised by
face pressure, simulating the test procedure usually practised
a set of geometrical parameters, the power of the external forces is
in laboratory experiments. The isovals of the plastic shear
compared to the power of dissipation that the ground can mobilise
strains then allow the geometry of the fracture mechanism
given its resistance capacities. Optimisation with respect to the
to be shown, and in particular the fracture surfaces to be
geometry of the mechanism leads to a lower bound of the face
highlighted in the case of rigid block mechanisms. The limit
collapse mechanism : for the calculated value, face collapse is certain.
face pressure leading to collapse can also be estimated by
Compared to limit equilibrium approaches, the optimal geometry of representing the evolution of the extrusion of certain points
the failure mechanism results from the calculation by optimisation of the face as a function of the reduction in face pressure.
of the geometric parameters of the considered mechanism family.
The result does not depend on the initial conditions regarding If the face is self-stable (i.e. for zero face pressure), the
the stress state. The models developed in homogeneous media associated safety margin can be estimated by gradually
are not directly applicable in the case of heterogeneous grounds decreasing the shear strength of the ground until collapse
since the stress and velocity fields require the verification of (so-called “c-phi reduction” method), and adopting an approach
continuity conditions between materials of different characteristics. where ground parameters are weighted.

12
3
ASSESSMENT OF STABILITY CONDITIONS
In this section and for each of the four calculation configurations under the keyword “tunnel face stability” in the international scientific
(A, B, C, D) identified in §2.3, the different existing models are journals “Tunnelling and Underground Space Technology” and
presented successively according to the four approaches (Exp, LE, “Computers and Geotechnics”, 60 of which were for the period
YD, Num) explained in §2.5. 2018-2020 (3% of the articles published in these journals).

Many scientific articles have been devoted to the study of face The references cited are not exhaustive and do not retrace the
stability over the last forty years. By way of illustration, it is worth men- history of the work that led to the construction of all these models:
tioning that, at the time of writing, nearly 100 articles are referenced all the “intermediate” models have not been mentioned.

3.1 DISCONTINUOUS MEDIUM

In moderately fractured rock masses, comparable to discontinuous
mediums at the scale of the structure, face instabilities lead to
dihedral falls or slides isolated by the network of discontinuities
(Figure 4).

The geometry of the collapse mechanisms is dictated by the

orientation and dip of the discontinuities with respect to the tunnel
axis, their extension, their space and the shape of the face.
The natural stability of these dihedra depends on the density
of the rock and on the shear strength that can be mobilized on
the discontinuities (function of the roughness of the wall rocks,
their opening, possible filling, the presence of water, etc. The
phenomenon of prevented dilatancy is an essential factor in
the natural stability of underground excavations in rock).

As such, the design process can be summarised in the form of Figure 4: Example of instability in rock mass in the Chavannes tunnel
the following steps: (France).

• Identification of geometric and geomechanical data • Calculation of the safety factor for each block and
on discontinuities: calculation of the bolting scheme required to obtain
The main families of discontinuities, their orientation an acceptable safety factor:
and average spacing are assessed using the structural The safety factor is calculated, in the limit equilibrium
analysis of the massif deduced mainly from information approach, by the ratio between the driving forces
from field surveys and core drilling. (weight of the block) and the strengths (shear strength
The shear strength of discontinuities can be measured along the discontinuities and the contribution of the
from shear tests on discontinuities in the laboratory, or from bolting). The bolts have two effects: on the one hand,
Barton’s (1977) work on surface condition (JRC) and wall rock they directly take on part of the weight of the dihedron,
strength (JCS). In the absence of bolts or with unprestressed and on the other they increase the shear strength that
bolts, it is recommended by AFTES WG30 (2021) not to is available along the faces of the dihedron, by increasing
retain cohesion on the discontinuities (residual strength). the normal stress there if they are prestressed. These
• Identification of potentially unstable blocks: two effects are, of course, to be integrated into the
The geometry of the largest blocks that can slide at the face stability calculations.
is assessed with a stereographic projection-based approach, A safety factor of the order of 1.6 to 1.8 can be sought,
or through 3D numerical modelling. A bolting set-up to ensure by analogy with the values recommended in AFTES
general stability can be deduced from this (see the following WG30 (2021) for radial bolting in the interior zone.
point). The stability of smaller sized blocks (local stability) For situations where the excavation is conducted for
can be provided thanks to the shear strength of sprayed a long time, the required coefficient can increase up
concrete, or by releasing these small unstable blocks. to 2.0.

13
3

3.2 CONTINUOUS UNDRAINED MEDIUM

When tunnelling at high speed in isotropic homogeneous with a 2-cm diameter hole modelling the face. The assembly
low-permeability ground, it can be legitimately assumed that was placed on a triaxial test frame. First, a vertical stress σ0
the ground behaves as undrained, whatever the boring method was imposed on the sample, and then the normal stress on
(conventional or TBM tunnelling), even if the necessary the hole p0 was gradually reduced until the face collapsed
conditions (see §2.3) are rarely combined in practice. In (Figure 6).
this context, the ground strength criterion can be considered
to be of the “Tresca” type (purely cohesive soils) caracterised These analyses lead the authors to write the stability condition
by the undrained shear strength (or in other words “undrained of the face, in purely cohesive ground, in the form of equation [5],
cohesion”) Cu. Figure 5 summarises the main references where N is called the “stability number” and in fact expresses
given over to this subject. a load factor relative to the strength of the ground.

Broms & Bennermark (1967) are the first known authors to have
taken an interest in this issue. They performed laboratory tests
using a cylindrical sample of clay contained in a steel cylinder, [5]

Case Type Reference Brief description

First laboratory tests carried out in Sweden with an experimental device similar to a triaxial cell
Broms & Bennermark (1967) with a hole allowing the extrusion of the soil ; Analysis of face instabilities on 14 excavation sites ;
Homogeneous continuous undrained medium

Introduction of the "stability number".

Exp Centrifuge tests carried out in Cambridge, England, with a tunnel modelled as a 60 mm pressurised tube.
Kimura & Mair (1981) Different lengths of unsupported span (zero to "infinite", i.e. plane deformations) and overburden
thicknesses studied.
Bezuijen & Van Seters (2006) Centrifuge tests in Delft, the Netherlands, with a tunnel modelled as a 150 mm pressurised tube.
Adaptation of the model of Anagnostou (2012) carried out in cohesive-frictional ground, to the purely
Perazelli & Anagnostou (2017) cohesive case. The novelty lies in the calculation of the vertical force V acting on the dihedral, deduced
from the static approaches of Gunn (1980), instead of a "Terzaghi" type discharge vault.
LE
Extension of the model of Perazzelli & Anagnostou (2017) by considering a circular geometry of the
Champagne de Labriolle (2018) working face (instead of rectangular) and a more realistic distribution of horizontal stresses
along the sliding surfaces.
Kinematic approach: several cases studied including a 3D mechanism with two blocks (elliptical cylinders).
Davis et al. (1980) Static approach: spherical and cylindrical stress fields, and comparison with experimental results in
YD
the case of a plane strain collapse mechansm in the cross section.
Mollon et al. (2013) Kinematic approach: 3D mechanism with a torus shape.
Num Ukrichton et al. (2017) 3D finite element calculations taking into account a linear increase of Cu with depth.

Figure 5: Summary of existing approaches in "undrained" conditions.

(a) (b)

Figure 6: Broms & Bennermark Laboratory Tests (1967): (a) experimental device, (b) summary of results.

14
Broms & Bennermark (1967) also summarized the stability Bezuijen & Van Seters (2006) also conducted two centrifugal tests
conditions observed at the face during eleven sewerage pipe at Delft University for shallow overburden thicknesses (C/D equal
jacking operations (diameters between approximately 0.8 to 0.6 and 0.8). The geometry of the collapse mechanism observed
and 2.3 m) carried out in Sweden, as well as three tunnels is not described by the authors, but the values of the stability
approximately 7 m in diameter in Norway and the United numbers associated with the collapse are provided (see Figure 8).
States, for overburden thicknesses of between 1.5D and 4D
Different theoretical approaches have been developped in parallel
approximately. Most have stability numbers less than 6,
by different authors. In the field of yield design, mention may be
without major face instabilities. The experience of the Tyholt made of the work of Davis et al. (1980) and Mollon et al. (2013),
Tunnel (Norway) in 1950 is interesting however, with two concerning the kinematic approach (approach from the outside).
face instabilities, for C/D (overburden thickness/diameter) ratios Davis et al. (1980) considered different mechanisms in the interior
equal to 2.0 and 2.2 respectively, and stability numbers of the zone of the tunnel, but also a mechanism at the face (zero unsup-
order of 6.2 and 7.8. The length of the unsupported span during ported span), consisting of two elliptical cylinders. Mollon et al. (2013)
the collapses is not known. Compressed air pressure was then improved it by considering a torus-shaped collapse mechanism.
imposed in the tunnel, reducing the stability numbers by 2 units, Regarding the static approach, Davis et al. (2013) studied different
thus ensuring stability. stress fields with cylindrical or spherical geometry at the face.

Centrifugal tests on a scale model of a tunnel were then carried Among the limit equilibrium approaches, Perazzelli & Anagnostou
(2017) extended the approach developed in the case of frictional
out in Cambridge on clays (Schofield, 1980; Kimura & Mair, 1981).
cohesive ground (see §3.3) to the case of purely cohesive ground.
The tunnel was modelled in the form of a pressurized half-tube
The main modification made concerns the calculation of the vertical
60 mm in diameter, increased to an acceleration of the order
stress exerted on the upper face of the dihedron. Contrary to the
of 120g, i.e. modelling a real tunnel of around 7 m in diameter. cohesive-frictional case where it is deduced from a Terzaghi
Different overburden thicknesses (1.2 < C/D < 3.3) were studied, mechanism, based on the assumption of vertical sliding surfaces,
as well as different lengths of unsupported span d, between 0 here it is, deduced from the static approach to yield design in accor-
(displacements allowed only at the face), and infinity (study in dance with Gunn’s work (1980). Three forms of stress fields in ground
the interior zone of an unsupported tunnel). located above a rectangular trap are proposed by Gunn: a cylindrical
geometry stress field the axis of which is parallel to the length of the
These tests clearly demonstrate that: trapdoor (i.e. longitudinally to the tunnel in our case), a stress field
with a cylindrical geometry, the axis of which is parallel to the width
• the collapse mechanism affects the entire face (Figure 7a), of the trapdoor (i.e. transversely to the tunnel in our case), and a
• the lower sliding surface is inclined at less than 45° spherical geometry stress field encompassing the trapdoor. The mini-
horizontally, which leads to a mechanism extending mum vertical stress value derived from these three forms is used.
quite far in front of the face (almost a diameter at tunnel
Champagne de Labriolle (2018) improved the Perazzelli &
axis height),
Anagnostou (2017) approach by considering the cylindrical
• the initial mechanism spreads very rapidly to the surface
geometry of the tunnel. The integration of the stresses considers
when the overburden is shallow (1 to 2D) with a final slices (of tiny heights) of variable width over the height of the
sinkhole advancing to about 2D in front of the face, and face, contrary to the Perazzelli & Anagnoustou (2017) approach
0.5D to the rear, in which the face is considered rectangular. Finally, Ukrichton
• there is no clear shear surface inside this mechanism. et al. (2017) carried out numerical finite element calculations
by considering a circular tunnel and a zero unsupported span.
The numerical models carried out later by Ukrichton et al. (2017) Perazzelli & Anagnostou (2017) also performed two numerical
also confirmed these experimental observations (Figure 7b). calculations in finite differences with a square tunnel cross-section.

(a) (b)

Figure 7: Mechanism geometry: (a) Centrifugal tests by Kimura & Mair (1981), (b) numerical modelling by Ukrichton et al. (2017).

15
3
Figure 8 compares the stability numbers N obtained by these Note also the following limitations to the approaches mentioned
different approaches in the case of a zero unsupported span above:
and a constant "undrained cohesion". This summary firstly shows
that the experimental results, notably Kimura & Mair (1981), are • static approaches to yield design consider simple
stress fields, which it is difficult to imagine as being
based on:
fully representative of the stress state around the face,
• approaches from the outside, for which the collapse is • the limit equilibrium models are based on a wedge,
certain, whether they be of the “kinematic approach to delimited at the level of the cutting face by a flat sliding
yield design” type (Davis et al. (1980), improved by Mollon surface, contrary to experimental observations which
et al. (2013)), or of the numerical type (Ukrichton et al. (2017) tend to consider a spiral arc,
and Perazzelli & Anagnostou (2017)). It should be noted • it is surprising that such a large gap exists between
that the difference between the two previous numerical the numerical and experimental approaches.
approaches can be related to the geometry considered
(square or circular face), but also to the rather coarse In view of the previous developments, the face collapse limit
mesh considered by Perazzelli & Anagnostou (2017), pressure can thus be calculated from equation [7], integrating
• approaches from the inside, for which the face is the principle in §2.4 that takes safety into account. The
choice of the value of the critical stability number Ncrit is to
potentially stable, whether they are of the “static approach
be made by the engineer on the basis of Figure 8, within
to yield design” type (Davis et al., 1980) or “limit equilibrium
the illustrated bracket. Lower values within the bracket
+ lead u
+
of the dihedron taking into account a vertical stress on
to a safer design.
the dihedron deduced from the static approach to yield
design” (Perazzelli & Anagnoustou (2017), Champagne u
de Labriolle (2018)). + + , ,

u
The disparity between these two broad categories of approaches + + + u u
++ + on Figure 8,
chosen
remains quite large, around 30%. For example, for a tunnel 10 m
u
in diameter, with a 15-m thick overburden, bored in a ground with
30 kPa of "undrained cohesion", the stability number is between ,
6.5 and 9, which leads to a limit pressure at the face, without u u u
otherwise
safety factors, of between 130 and 200 kPa. [7]

COLLAPSE

Broms & Bennermark (1967) -

Tyholt tunnel construction site (Norway, 1950)
Kimura & Mair (1981) -
Centrifugal tunnel experiment (case P/D=0)
Bezuijen & Van Seters (2006) -
Centrifugal tunnel experiment

STABILITY
Davis et al. (1980) - Kinematic approach
(two elliptical cylinder mechanism)
Broms & Bennermark (1967) - Mollon et al. (2013) - Kinematic approach
Tyholt tunnel construction site (Norway, 1950) (torus mechanism)
Kimura & Mair (1981) - Ukrichton et al. (2017) -
Centrifugal tunnel experiment (case P/D=0) Numerical calculation (cylindrical face)
Bezuijen & Van Seters (2006) - Perazzelli & Anagnostou (2017) -
Centrifugal tunnel experiment Numerical calculations (square face)
Broms & Bennermark (1967) - Davis et al. (1980) - Kinematic approach Davis et al. (1980) - Static approach
Tyholt tunnel construction site (Norway, 1950) (two elliptical cylinder mechanism) (maximum cylindrical and spherical fields)
Kimura & Mair (1981) - Mollon et al. (2013) - Kinematic approach Perazzelli & Anagnostou (2017) -
Centrifugal tunnel experiment (case P/D=0) (torus mechanism) Boundary equilibrium - V-force deduced
Bezuijen & Van Seters (2006) - Ukrichton et al. (2017) - from the static approach of Gunn (1980)
Centrifugal tunnel experiment Numerical calculation (cylindrical face) Champagne de Labriolle (2018) -
Davis et al. (1980) - Kinematic approach Perazzelli & Anagnostou (2017) - Similar approach to P & A (2017),
(two elliptical cylinder mechanism) Numerical calculations (square face) but with 'cylindrical face'
Mollon et al. (2013) - Kinematic approach Davis et al. (1980) - Static approach
Figure 8: mechanism)
(torus Comparison of the stability numbers deduced from cylindrical
(maximum the different existing
and spherical models.
fields)
Ukrichton et al. (2017) - Perazzelli & Anagnostou (2017) -
Numerical calculation (cylindrical face) Boundary equilibrium - V-force deduced
Perazzelli & Anagnostou (2017) - from the static approach of Gunn (1980)

16 Numerical calculations (square face) Champagne de Labriolle (2018) -

Davis et al. (1980) - Static approach Similar approach to P & A (2017),
but with 'cylindrical face'
 The influence of the length of the unsupported span on the stability span 0.2.D long (see the author's Figure 24) in the case of
of the face has also been studied by various authors. Mention may an “undrained cohesion” greater than or equal to 50 kPa
be made, in particular, of the Kimura & Mair centrifugal tests (1981), (i.e. Cu / (γ.D) > 0.3).
carried out for a large unsupported span, equal to 0.5D, 1D, 2D
and infinite (i.e. equivalent to a stability calculation in the interior For very low “undrained cohesions” (Cu / (γ.D) < 0.3), the influence
zone of the tunnel). The results obtained reveal significantly lower of the unsupported span in the Champagne de Labriolle (2018)
stability numbers in "plane strains" (infinite unsupported span) than approach is greater than indicated in the previous paragraph.
in the case of a zero unsupported span (N divided by about two).
Some authors, notably Ukrichton et al. (2017) and Champagne
For an unsupported span limited to 0.2D, it can be estimated
de Labriolle (2018) have integrated a changing "undrained
that the stability number is reduced by about 0.5 according to the
cohesion" depending on the depth. The results obtained are not
conclusions obtained by the authors.
presented here since it is rare, in practice, to have sufficient test
The analytical calculations by Champagne de Labriolle (2018) results to demonstrate this change in shear strength depending
result in the same order of magnitude for an unsupported on the depth.

3.3 CONTINUOUS DRAINED MEDIUM WITHOUT SEEPAGE

3.3.1 Homogeneous ground

In the case of permeable soil and a low advance rate (as an
indication, as previously, for k > 10-6 m/s and vavct < 0.1 m/h),
the ground is solicited in a “drained” state. The stability of
the face is in this case conditioned by the effective stresses
exerted in the ground. Let's consider that there is no seepage
of pore water into the ground (water pressures remain equal to the
initial hydrostatic pressures). This condition is encountered if the
boring takes place outside the water table or if a pressurized TBM
is used (except for the conditions described later in §3.5 and 3.6).

In this configuration, calculations are performed in effective

stresses with:
• the density γ* of the ground equal to: γ’ = γsat – γw (soil unit
weight under buoyancy) if the tunnel is below the water
table, or γh (moist soil unit weight) if the tunnel is outside
the water table,
• the drained shear strength, expressed as effective cohesion Figure 9a: Geometry of the collapse mechanism in cohesive-
c’ and an internal friction angle φ’ in the context of a frictional ground (cross and longitudinal sections): Chambon
Mohr-Coulomb criterion. & Corte (1989) tests.

Figure 9b: Geometry of the collapse mechanism in cohesive-frictional ground (cross and longitudinal sections): Berthoz et al. (2012) tests.

17
3
The case of partially saturated soils is usually treated by Examples include:
considering an "apparent" effective cohesion and friction angle • Tests carried out by Chambon & Corté (1989) in a centrifuge
deduced from laboratory tests conducted on soil taken in a with dry sands, where the boring is generated by deflating
state of saturation identical to that observed on site. a flexible membrane at the face, with constant radial support
by means of a metal tube,
Different testing campaigns were carried out in the laboratory • Tests performed by Berthoz et al. (2012) on a scale model of
in this configuration. a non-centrifuged earth pressure TBM, with dry sands or
slightly moist sands, giving them a cohesion of a few kPa,
which corresponds to about ten kilopascals at full scale.

Case Type Reference Brief description

Centrifugal tests of rigid tubes with flexible faces on frictional cohesive ground
Chambon & Corté (1989) → Shape of the mechanisms, and face pressure value for some H/D values.

Non-centrifugal tests of rigid tubes with displaceable faces on dry sands

Exp Messerli et al. (2010) → Shape of the mechanisms, and limit face pressure values for some H/D values.

Non-centrifugal model tests in purely frictional and cohesive frictional soils

Berthoz et al. (2012) → Shape of the mechanisms, and limit face pressure values for some values of c'.

3D finite element models with perfectly plastic elastic ground behaviour governed by a Mohr-Coulomb
criterion, and displacement allowed only at the face (radial displacements blocked at the wall). Different
Vermeer et al. (2002)
Num values of dilatancy angle and earth pressure coefficient at rest were tested, with no significant effect
on collapse pressure.
Alagha & Chapman (2019) A model similar to that of Vermeer et al. (2002) was used, with some additional parametric studies.
Horn (1961) type mechanism. Equilibrium of the wedge at the face with these assumptions:
Homogeneous continuous medium drained without seepage

(1) rectangular front (B*H),

Anagnostou & Kovari (1996) (2) vertical force on the wedge deduced from a relief arch with σH = K1.σV0 where K1 = 0.8,
(3) shear force along the corner deduced by assuming that the vertical stress increases linearly with depth
over the height of the prism, with the horizontal stress proportional via a factor K2 = 0.4.
The model of Anagnostou & Kovari (1996) is modified by:
(1) deleting the previous assumption (3) and replacing it with a "discharge arch" type analysis in
Anagnostou (2012)
LE the horizontal plane,
(2) choosing σH =1.0 σV (instead of 0.8) for the block and the prism (assumption (2)).
Recommends using the method of Anagnostou & Kovari (1996) but with a coefficient K2 = (K0 + Ka)/2
DAUB (2016)
instead of 0.4.
Modifications to the model of Anagnostou & Kovari (1996) by: (i) expressing K from the assumption
that the criterion is verified on the boundaries of the mechanism (thus relationship between σV
Champagne de Labriolle (2018)
and σH , which are not necessarily principal constraints); (ii) considering the circular character of the
working face (instead of rectangular).
Leca & Dormieux (1990) Kinematic approach: initial model: one or two 2D blocks.
Kinematic approach: 3D one-block model defined by logarithmic spirals without internal discontinuity
Wong & Subrin (2006)
surface.
Kinematic approach: 3D model with a large (but variable) number of discontinuity surfaces cutting
Mollon et al. (2011)
blocks defined by logarithmic spiral arcs.
Kinematic approach: modifications to the model of Mollon et al (2011) by replacing the part of the
mechanism located above the tunnel crown by a pressure deduced from a "limit equilibrium" type
Zou et al. (2019)
mechanism with an assumption of normal stress on this mechanism (Ks) drawn from various approaches
(analytical, numerical, experimental).
YD
Kinematic approach: using the geometry of the Wong & Subrin (2006) model, considering a partially
Li et al. (2019)
saturated soil.
Kinematic approach: 3D model with two blocks (a tetrahedron at the face, and a prism with a
Quarmout et al. (2019) triangular base above). Resolution by the KEM method (Kinematic Element Method), which is not
exactly Yield Design, but is similar.
Kinematic approach: 3D one-block mechanism defined by logarithmic spirals with no internal
Li & Yang (2020)
discontinuity surface, and with truncation of the tensile strength criteria.
Kinematic approach: 2D 3-block mechanism, considering the existence of a unsupported span.
Senent et al. (2020)
Possible consideration of an umbrella arch.

Figure 10: Summary of existing approaches in continuous homogeneous medium drained without seepage.

18
These tests, the results of which are shown in Figure 9, reveal Ultimately, all the approaches cited here converge. Vermeer
that the mechanism: et al. (2002) also performed numerical calculations considering
a non-zero unsupported length d. They have shown that Nγ
• affects the entire face, increases quite significantly (i.e. the minimum necessary
• is not very far from the front of the face (less than 0.5 D) pressure to exert on the face and in the unsupported span
and rises slightly above the crown (approximately 0.1 D increases), when the unsupported length becomes greater
for an internal friction angle of around 35°), than about 0.3 D. Senent et al. (2020) also showed this via an
• can propagate to the surface by progressive collapse in analytical approach to yield design (2D longitudinal mechanism
the case of low overburden. with 3 blocks defined by logarithmic spirals).

Many numerical and analytical models have also been devoted Given the consistency between these different results, the
to this configuration. Figure 10 gives the main references. expression [10], deduced from Vermeer et al. (2002) can be
used for most of the cases encountered in practice (d/D ≤ 0.5,
The numerical modelling carried out by Vermeer et al. (2002) C/D ≥ 0.8 and φ’ ≥ 20°).
or Berthoz et al. (2012), confirm these main features of the
mechanisms observed in a frictional-cohesive medium, on
which various recent analytical models have been based
(Wong & Subrin (2006), Mollon et al. (2011), etc.).
with
The minimum face collapse pressure associated with this
type of mechanism can be expressed as the following
equation:
where: ,

with , ,

[9] (under the water table)

,
(outside the water table)
otherwise
Different authors expressed the Nγ function in equation [9]
via different approaches. Figure 11 summarises the results. [10]

Figure 11: Comparison between different existing models.

19
3
In the context of the kinematic approach for yield design,
the mechanism can intercept the surface of the natural
ground in the case of a very shallow tunnel in low friction
ground (typically for approximately C/D ≤ 0.3 (cot φ’ – 1)).
The minimum face stability pressure is then a function of
the surface overload σs and can be expressed in the form
of an equation [11] (Wong & Subrin 2006), analogous to
the classical yield design approaches for shallow or deep
foundations.

[11]

Numerical calculations by Alagha & Chapman (2019) also

show that the influence of the overburden thickness (C/D ratio)
becomes significant when the internal friction angle of
the soil is very small (φ’ < 20°). This hypothesis of a very Figure 12: Influence of tensile strength on face collapse limit
shallow tunnel and very low friction ground is however rare pressures (Li & Yang, 2020). Here, d is the tunnel diameter (D)
among projects. and ξ = 0 means that the criterion is not truncated at all.

All the numerical and analytical models presented consider a In purely frictional ground (dry or saturated sands), a seepage
Mohr-Coulomb strength criterion with a non-zero tensile strength mechanism originating at the crown and rising to the surface,
of the soil. Li & Yang (2020) performed analytical calculations like an “hourglass” (Figure 13), can be mobilized according
(with the kinematic approach to yield design) and numerical to the scale model tests carried out by Berthoz et al. (2012).
calculations with this hypothesis to evaluate the influence of In practice, this type of solid grain seepage mechanism could
this questionable hypothesis in soils. The calculations performed only be encountered in the case of tunnelling with an earth
show that conventional approaches (without truncation of the pressure TBM outside the water table in sands. Indeed, in
tensile criterion) tend to underestimate the confinement pres- conventional tunnelling methods, field treatment would be
sures at collapse. This underestimation remains low for φ’ ≥ 20° required, and bentonite would be injected into the slurry
and c’ ≤ 20 kPa (Figure 12) but becomes significant in the pressure TBM. The latter two cases would lead to the existence
case of low friction and highly cohesive ground (φ’ ≤ 20° and of cohesion over a short length of ground at the front of the
c’ ≥ 20 kPa). In practice, however, soils having such shear face, preventing the appearance of this seepage mechanism.
strengths may probably be considered as "undrained" and In all of these cases, using the equation [10] approach with
thus treated by the approach described in §3.2. zero cohesion is safe.

Figure 13: Geometry of the collapse mechanism in purely frictional ground (Berthoz et al., 2012).

20
3.3.2 Layered grounds upper part of the face (height β.D) has a shear strength lower
than that of the lower part. The decrease in the associated
Some authors have extended the previous approaches to the factor Nγ (i.e. the increase in stability caused by the reduced
case of horizontally stratified grounds, with layers of different shear dimensions of the face) is given in Figure 16b.
strengths, but always with the assumption that they are individually
isotropic. Figure 14 provides a brief overview of these. Limit Equilibrium (LE) approaches make it quite easy to take
into account several layers of different shear strengths, over
In particular, we should note the physical models in stratified the height of the face and in the thickness of the overburden,
ground produced by Berthoz et al. (2012). These tests have
such as the developments presented in Broere (2001),
highlighed the associated collapse mechanisms, as shown in
adopted by Vu et al. (2015) and Champagne de Labriolle (2018).
Figure 15 in the case of a three-layer ground (lower part of
These two models consider a "wedge and vertical prism"
the face self-stable frictional cohesive, upper part of the face
purely frictional, with a frictional cohesive overburden). type mechanism similar to that of Horn (1961), adopted by
Anagnostou & Kovari (1996), but with some specificities,
Berthoz (2012) also extended the Wong & Subrin (2006) such as the consideration of the arching effects, via the K
mechanism to the case of a two-layer stratified mass where the ratio between vertical stress and horizontal stress.

Case Type Reference Brief description

Non-centrifugal tests with a reduced scale model in two-layered and three layered grounds
Exp Berthoz et al. (2012) → Shape of the mechanisms.

The face is divided into ten "slices", heightwise. For each slice, the pressure at the face is calculated
Seepage-free drained stratified

Piakowski & Kowalewski (1965) as a function of the active ground pressure (Rankine theory) and a shape coefficient allowing for
arch effects.
continuous medium

LE The approach is identical to that of Anagnostou & Kovari (1996) but with:
Broere (2001)
(1) the face split into slices, allowing application,
and Vu et al. (2015)
(2) horizontal arching effects taken into account via the choice of K1 = K2 = K0 and replacing R by R/(1+tanθ).

Champagne de Labriolle (2018) Homogeneous soil approach transposable by reasoning by horizontal layers of different shear strength.

Berthoz (2012) Extension of the model of Wong & Subrin (2006) to the case of a stratified bilayer medium.

Ibrahim et al. (2015) Extension of the model of Mollon et al. (2011) to the case of a stratified bilayer medium.

YD Extension of the model of Mollon et al. (2011) to the case of a non-circular tunnel. Finally, the case
Pan & Dias (2017)
studied is similar to that studied by Berthoz (2012) and Ibrahim et al. (2015).

Extension of the model of Mollon et al. (2011) to the case of a linear evolution with depth of the ground
Zou et al. (2019)
cohesion and friction angle.

Figure 14: Summary of existing approaches in a seepage-free stratified drained continuous medium.

Figure 15: Collapse mechanism in stratified ground (purely frictional layer between two cohesive frictional layers) (Berthoz et al., 2012).

21
3
However, his work remains rather difficult to use since there is no of a strong contrast, with one very strong layer in the upper part
direct expression of the result: a spreadsheet is required. When (rocky slab), the mechanism can simply be truncated at the
working on a project, if the strength contrasts between the lower (soft) part of the face, the vertical stress exerted on the
different layers are not too significant, then an initial approach wedge being automatically reduced due to the cohesion
can be to consider the shear strength of a homogeneous considered in the upper layer. If the very strong layer is in the lower
medium as equal to that of the weakest actual layer. In the case part of the face, a reduced face height can simply be considered.

Nγ

(a) (b)

Figure 16: Two-layer stratified ground (Berthoz, 2012): (a) Mechanism geometry, (b) nomograph.

3.4 CONTINUOUS DRAINED MEDIUM WITH SEEPAGE

3.4.1 Seepage towards the face This hypothesis is legitimate to the extent that the permeability
of the ground k is greater than or equal to 10 –7 to 10 –6 m/s
When boring a tunnel by the conventional method or with an and that the face advance rate vavct is less than or equal
open-face TBM in water-saturated ground, seepage towards to 0.1 to 1 m/h (as an indication according to Anagnostou
the face appears with seepage forces decreasing the stability & Kovari (1996)). It should be remembered that the
of the face. The main authors who have studied this problem assumption of an isotropic ground is made, both in terms of
are indicated in Figure 17. shear strength and permeability. Schuerch et al. (2019) have,
however, also studied the transient phase of the seepage,
The seepage force reaches its maximum when the hydraulic quantifying the length of time during which the face remains
steady state is reached with a maximum flow rate. Most calcu- stable in the short term. This question is posed, for example,
lations carried out in this case are done so in a conservative during hyperbaric maintenance operations on the cutterhead
manner and consider that the steady state has been reached. of pressurized TBMs.

Case Type Reference Brief description

Non-centrifugal tests on a reduced-scale model in dry, saturated no-flow, and saturated flowing sands.
Exp Lü et al. (2018)
Comparison of mechanisms and limit collapse pressures.
with seepage towards the face
Continuous drained medium

Num Schuerc et al. (2019) 3D finite element modelling with hydro-mechanical coupling for transient analysis.
Extension of the model of Anagnostou (2012) by taking into account the flow forces towards the tunnel.
LE Perazelli et al. (2014) For this purpose, the hydraulic load h is assumed to depend only on the distance to the face x for the wedge,
with an expression calculated through numerical finite element modelling of the flow.
Extension of the model of Mollon et al. (2011) by adding a component Nw modelling the flow
Pan & Dias (2016) forces (i.γw) to the classical effective stress calculation (function of Nγ'). Nw is deduced from a flow
calculation on Flac.
YD Extension of the Pan & Dias (2016) model to the case of a Hoek & Brown plasticity criterion
taken into account via an optimization of the mechanism geometry as a function of the internal
Pan & Dias (2018)
friction angle corresponding to the linearization of H&B and keeping a couple (c,φ) consistent with
the initial H&B criterion (if φ increases, c decreases and Nc increases).

Figure 17: Summary of approaches for a continuous drained medium with seepage towards the face.

22
Figure 18: Results of experiments by Lü et al. (2018): geometry of the mechanisms in the presence of a seepage (purple) for different
water table heights (H) relative to the tunnel crown. The mechanisms observed for dry (dotted blue line) and saturated grounds
without seepage (solid blue line) are also presented for comparison. The axes of the x and y coordinates correspond to the distances
with respect to the foot of the face. The colour scale corresponds to the plastic strain isovalues.

(a) (b)

Figure 19: Finite element calculations of seepage forces from Perazzelli et al. (2014): (a) mesh, (b) hydraulic load isovals.

Let us first consider the steady state reached. Lü et al. (2018) Figure 18) than the same dry (63°) or saturated material (57°)
conducted tests on an uncentrifuged model, with a tunnel of without seepage. The mechanism consequently extends further
D = 15 cm in diameter and an overburden thickness between 0.5 out in front of the face in the presence of seepage forces,
and 2.D. The ground is made up of sands. Given the d10 of this and mobilises a larger volume of earth at collapse.
sand (about 85 microns), its permeability is around 5.10–5 to
10–4 m/s according to the Hazen relationship (AFTES WG8, 2006). Different analytical models have been developed to evaluate the
The instability of the face is generated by pushing back a rigid plate. face pressure σ’T required in the presence of seepage. In these
Three test conditions are studied: dry sand, sand saturated with water models, the seepage forces are estimated using numerical
but without seepage ("watertight" plate at the face) and water- calculations (with finite elements or finite differences) as shown
saturated sand with seepage (totally permeable plate at the face). in Figure 19. These are then injected into the face stability
analytical models presented in the preceding paragraphs.
These tests show that the seepage forces lead to a lower Perazzelli et al. (2014) adapted Anagnostou & Kovari's limit
inclination of the mechanism sliding surface in the presence equilibrium model (1996) and Pan & Dias (2016, 2018) extended
of seepage (about 40° relative to the horizontal according to the yield design model of Mollon et al. (2011).

23
3
The calculations performed by the various authors show that: state (right) in a particular case. These figures are deduced
from Schuerch et al. (2019) and Perazzelli et al. (2014), after
• the instability of the face increases significantly when
standardising the notations used. The red dots correspond
the water height increases (see Figure 20), which
to the following assumptions with the notations in Figure 2:
is explained by the increase in the amplitude of the
D = 10 m, Hw = H = 15 m, γ = 20 kN/m³, K0 = 1, k = 10–8 m/s,
(driving) seepage forces when the hydraulic gradient
E = 20 MPa, c’ = 25 kPa and φ’ = 25°. With these assumptions,
increases,
these calculations show that the critical duration ts of auto-
• the instability of the face increases when the density
stability of the face is equal to 700 s, and that a pressure σ’T
of the ground decreases (dotted lines compared to the
equal to 15 kPa must be applied to the face in steady state
solid lines in Figure 21b),
to ensure its stability.
• the trends are similar, but there is a disparity between
the results obtained by Perazzelli et al. (2014) and Pan
& Dias (2016, 2018). This difference depends in particular
on the amplitude of the hydraulic load (Figure 20),
• horizontal permeability of the ground greater than the
vertical permeability (anisotropy) improves the stability
of the face compared to the isotropic case of minimum
permeability (Pan & Dias, 2016).

Analyses in transient conditions by Schuerch et al. (2019)

on the basis of the coupled hydro-mechanical finite element
calculations, made it possible to assess the duration during
which the seepage forces remain sufficiently low for the
front to remain stable. Different values for the hydraulic load,
overburden thickness and earth pressure at rest coefficients
were studied. Nomographs were deduced from it. For Figure 20: Comparison of existing solutions with the following
example, Figure 21 shows the maximum duration of self- assumptions: circular tunnel of diameter D, γ w / γ’ = 1.0 (i.e.
stability (left) and the confinement pressure required in steady γ = 20,0 kN/m³), c’ = 0 kPa, φ’ = 30°.

Figure 21: (a) Assessment of the stability limit duration, according to Schuerch et al. (2019), and (b) assessment of the containment pressure
necessary to ensure steady-state stability, according to Perazzelli et al. (2014). In these two figures, the notations have been modified to be
consistent with those in Figure 2. They correspond to the following case: Hw = H = 1.5 D, K0 = 1. The two red dots correspond to the case
described in the body of the text.

24
3.4.2 Seepage towards the ground Broere (2001) conducted analytical developments to express
the amplitude of pore overpressures Δu as a function of the
Boring with a pressurized face TBM can generate pore particle size (d10) and of the permeability (k) of the ground, the
overpressures in the ground in front of the face, thereby shear strength of the bentonite (τF), and the initial hydrostatic
reducing the effective stresses, including the shear strength pressure (u0). The field of effective stresses at any point
on the collapse surfaces of the face instability mechanism. in the collapse mechanism can be deduced from this and
applied to the models described in §3.4.
Measurements made by Xu & Bezuijen (2018) in the
Netherlands during boring with a slurry pressure TBM in sands However, the parametric studies carried out by Broere (2001)
(10–5 < k < 10–3 m/s) show that these reach 50 kPa in the show that taking into account pore overpressures when
immediate vicinity of the face and 10 kPa 10 m in front of the studying face stability leads to a small increase in face pressure
face (Figure 22a). During the segment laying phases (boring (less than 20 kPa) in the vast majority of the cases studied.
is stopped), these overpressures dissipate quickly until they
cancel each other out (return to hydrostatic pressure). This Consequently, a first approach in this case (boring with a
phenomenon has also been measured by the same authors slurry pressure TBM or an earth pressure TBM with many
in the case of earth pressure tunnelling using admixtures fine admixtures in sandy soils with a permeability of between
(Figure 22b), with overpressure peaks which are even greater in 10–5 and 10–3 m/s) could be to calculate the minimum face
the case presented (100 kPa in the immediate vicinity of the face, collapse pressure without taking pore overpressures into
and 20 to 30 kPa 10 m ahead of the face). The associated boring account, and then increase this pressure by a lump-sum of
conditions are not, however, described precisely by the authors. 30 kPa.

(a) (b)

Figure 22: Pore overpressures measured in front of a slurry pressure TBM (a) and earth pressure TBM (b) by Xu & Bezuijen (2018).

25
4 ON-SITE OBSERVATIONS
AND CONSTRUCTION MEASURES
This section deals with construction principles that ensure conventional tunnelling, and (ii) the case of pressurized face
face stability, with recommendations in terms of face stability TBMs. The case of open tunnelling machines is partly covered
monitoring. Two cases must be distinguished: (i) the case of by the information given for conventional method tunnelling.

4.1 CONSTRUCTION METHODS IN CONVENTIONAL TUNNELLING

In conventional tunnelling, different processes may be applied to Drainage and dewatering
control face stability, notably cross-section reduction, drainage,
The calculations presented in section 3.4 have demonstrated, if
face bolting and ground treatment. The following paragraphs
necessary, the very negative impact of water on the stability of the
outline some main principles for each of these process.
working face. Pre-draining the water table or draining the ground
as the face advances is therefore an excellent method of reducing
Cross-section reduction pore pressures at the front of the face and preventing the
The lower the height of the face, the greater its stability. Boring development of significant seepage forces that are detrimental to
sections separately, (top heading, then bench), or maintaining stability. This drainage should preferably be peripheral in order to
a central merlon in place can therefore increase its stability. avoid local instabilities at the centre of the advancing tunnel face.
However, this is not without consequences on the use of wall
supports and more generally the tunnelling advance rate. Zingg & Anagnostou (2016) performed analytical and numerical
calculations of the stability of the face in soils (0 < c' < 400 kPa
and φ' = 30°) by varying the number, length and position of
Sprayed concrete layer these boreholes. The authors conclude, on the basis of charts
Although in practice calculations do not justify it, purging must quantifying the effect of drainage on the stability of the face,
be carried out, followed by the application of a thin layer of that a very good efficiency is obtained with 4 to 6 boreholes of
fibre-reinforced sprayed concrete on the face after each 10 cm diameter and 1.5 D in length located near the tunnel wall,
excavation sequence (3-5 cm thick), primarily to ensure the in the upper part of the face.
safety of workers and protect them from small block falls.
However, implementing dewatering and drainage as the tunnel
Indeed the effectiveness of the sprayed concrete comes from face advances can lead to consolidation settlements over an
two actions. On the one hand, it protects the ground against extended area around the structure.
surface alteration (by limiting desication caused by exposure
of the ground to air and limiting water flow that may lead
to out-wash and seepage forces). On the other, this layer
Face bolting
of sprayed concrete has a very local mechanical action (of a When the previous processes are insufficient, face bolting
centimetre magnitude), but sufficient to prevent micro- can be used.
displacements, and consequently micro-ruptures, which
would be likely to lead to a degradation of the characteristics of In the case of discontinuous mediums, where the mechanisms
the mass in depth, by a phenomenon of "chain discohesion: are of the “dihedral fall” type, a few continuously anchored
the grain of sand that escapes releases the stone that wedged bolts of moderate length (up to 6 m) oriented according to
the block, and so on” (“type 1” sprayed concrete within the the discontinuities are generally sufficient. The design of bolts
meaning of AFTES WG20, 2001). is based on limit equilibrium methods in accordance with §3.1.

26
In the case of continuous or continuous equivalent mediums, However, the effect of the umbrella vault on the stability of
long horizontal fibreglass bolts (generally 1.5 to 2.0 D, never the face is limited. As an example, the 2D numerical and
less than 0.5 D) are required. The bolts must be long enough analytical calculations (yield design) carried out by Senent
and changed regularly (generally by thirds) to ensure the et al. (2020) in cohesive frictional grounds without seepage
stability and control of face strains at all times. In particular, estimate that the presence of the umbrella vault reduces the
the calculation should be made considering the shortest length confinement pressure required to ensure face stability by
of the bolts, i.e. just before their renewal. about 15%.

There are several different approaches that can be used It should also be reminded here that installing an umbrella
to calculate the length, cross-section and number of these vault is a major construction process. Given the large diameter
face bolts. The simplest method consists in calculating the of the boreholes to be made, it is essential that the tubes be
stress σT required to ensure the stability of the face using the sealed one after the other, after the drilling of each hole, in order
approaches described in §2.5 and then, from this, deducing to avoid a "pre-splitting" of the ground and the appearance of
the total confinement force to be supported by the bolting. The very large settlements on the surface related to the installation
number and cross-section of the bolts required to support this phase prior to the actual excavation.
force without breaking the bolts, depends on the minimum
anchoring length required given the lateral friction (qs) that can Ground treatments
be mobilised around each bolt, taking into account the fact
that the anchoring must be verified over the "active" length Various ground treatment methods can be used with the
(at the head) and over the "passive" length (at the foot) without purpose of increasing the overall shear strength of the
reaching the tensile failure of the bolts (Peila, 1994). The AFTES ground and reducing its permeability. These include injections,
WG30 recommendation (2021) provides further information freezing or jet-grouting.
on this point, notably with regard to the safety factors to be The assessment of the mechanical characteristics of the
used and the determination of the value of the qs. soils treated is not of immediate interest. A sufficient safety
margin must be considered during design, and control tests
Other more elaborate approaches that explicitly take into
following treatments in test bore holes are recommended.
account bolts in the study of face stability are possible. An At the design stage, readers interested may, for example,
(old) summary of different design approaches is available in consult the following references to obtain initial information:
Clouterre (2002). These include the works by Leca (1997)
and Subrin (2002) as part of the kinematic approach to yield • for injections: cohesions between a few tens and hundreds
design, or those more recent works by Anagnostou & Perazzelli of kilopascals are observed (Dano, 2001; Chang et al.,
(2015) using limit equilibrium. Nowadays, it is also possible to 2009; etc). The efficiency of large-scale treatment remains
carry out 3D digital modelling with explicit bolt simulation a difficult subject. The engineer must remain cautious
(Zapata Franco (2020) for example), which also makes it about the cohesion value used in the design,
possible to evaluate the effect of bolting on the displacements • for jet-grouting: orders of magnitude of the cohesions
induced in the ground. and permeabilities attainable are given by Croce et al.
(2014) and Toraldo et al. (2018). The latter also offer
a methodology for estimating the shear strength of
Umbrella vault grounds on a large-scale from uniaxial compression
The main objective of an umbrella vault is to stabilise the and sonic tomography,
unsupported span, and limit pre-convergence and convergence, • for freezing: ISGF WG2 (1992) proposes a first estimate
i.e. limit displacements generated on the surface. However, of the mechanical characteristics of frozen soils, in
an umbrella vault also has a beneficial effect on the stability particular the increase in their cohesion when the soil
of the face insofar as it reduces the volume of the collapse temperature decreases.
mechanism by truncating the upper part of it.

Different approaches to the designing of vaults are possible,

ranging from simply calculating the hyperstatic reactions of
an isolated tube subjected to a ground load estimated using
a Terzaghi type mechanism, to 3D digital modelling. Some
references can be found in Gilleron (2016) for example.

27
4

4.2 MONITORING METHODS IN CONVENTIONAL TUNNELLING

In conventional tunnelling, the face is visible after each excavation Parameters characterizing the fractured rock environment,
sequence. This time must therefore be used to validate the based on Bienawski’s Rock Mass Rating (1989), Barton’s Q index
geotechnical assumptions (in the broad sense) made during (NGI, 2013) or Geological Strength Index (Marinos et al., 2005)
the studies, and in particular to validate the stability conditions can also be deduced from the observations and compared
of the face. Face surveys are the formal manifestation of these. to the expected values.
Examples of face surveys are available in Appendices 1 and 2
of AFTES WG24 (2008). Regarding "Validation of
face stability conditions"
Regarding the "Validation of Any signs of face instability are to be recorded during face
geotechnical hypotheses" surveys. These signs consist of falling blocks, the appearance
Important information includes: of cracks in the face, extrusion movements (measured with
an extrusometer, or topographic targets in the event of
• the nature of the ground, prolonged stoppage).
• an order of magnitude of the shear strength of the
These clues must enable the support manager to determine if
ground derived from empirical observations (peelable
the face is self-stable or not, and if there is instability to identify
with a knife, difficult to break with a hammer, etc.),
the geometry of the mechanisms at play and the best way to
pocket penetrometer tests in soils, or even Franklin tests
adapt face support.
for rock blocks,
• the orientation and condition of discontinuities in the It is also important to ensure the bolts are correctly distributed
rock masses, at the face (make sure that their deviation remains moderate)
• the number and flow rate of water inflows. and that they are properly sealed into the ground.

4.3 CONSTRUCTION METHODS FOR PRESSURIZED FACE TUNNELLING

Pressurized face tunnelling ensures face stability and limits This confinement mode is that exerted by slurry
the movement of the ground around the excavation. Different pressure TBMs,
containment modes exist, depending on the nature of the material • gas containment, where the working chamber is filled
contained in the working chamber: with compressed air. This confinement mode is that
used by air pressure TBMs, but also by earth pressure
• pasty containment, where the working chamber is TBMs when the working chamber is partially filled, or
mainly filled with excavated materials, with some additives by earth pressure and slurry pressure TBMs during
(water / foams / polymers / clay). These additives aim to hyperbaric operations.
make the confinement material sufficiently impermeable
to limit internal water circulation, but not too viscous so In all cases, these systems require the identification of a "set value"
that the torque on the cutter head remains reasonable, face confinement pressure to ensure the stability of the face,
nor too liquid so that the muck can be transported easily to contain hydrostatic pressures and, where relevant, to limit
by conveyor belt. This confinement mode is that exerted admissible movements in the presence of neighbouring structures.
by earth pressure TBMs with a full working chamber,
• liquid containment, where the working chamber is filled The issue of containment pressure along the shield is also
with slurry. This slurry must be viscous enough to enable paramount: if this is very low, the stability of the face will be degraded
a water-tight membrane (cake) to be created at the face (adopt the approach for boring with a non-zero unsupported
enabling pressure to be effectively exerted on the granular span in the theoretical approaches previously described). More
skeleton of the ground. However, it must not be too viscous broadly, experience shows that the pressure prevailing along
so that excavated material can be easily transported the shield directly conditions the amplitude of the displacements
to the slurry treatment plant, and be easily recycled. induced on the surface in the case of shallow tunnelling.

28
4.4 MONITORING METHODS FOR PRESSURIZED FACE TUNNELLING
The “closed” mode of pressurized face TBMs does not provide • better control impacts on neighbouring structures, when
accurate information about the nature of the ground or the necessary.
actual stability conditions of the face. The lower bound of
the stability domain is therefore unlikely to be re-assessed To meet these goals, the tunnelling advance plan (TAP)
during the work. Nevertheless, hyperbaric interventions can (AFTES WG16, 2020) must be accompanied by:
be used to make a partial survey of the face.
• a summary of the data collected during the work:
However, the area of "admissible displacements" and the (i) the probing data of neighbouring structures and
resulting set value pressure are to be adjusted regularly the ground, (ii) the geotechnical model data, and
during the work using the retro-analysis of the displacements (iii) the tunnelling steering conditions data (pressures
measured as part of the Tunnelling Advance Plan (TAP), made in the working chamber, thrust force on the cutter head,
by successive tunnel boring sections. This tool, which is intended total thrust force, fontimeter measurements, extracted
to be dynamic and iterative, has the following goals for the masses, etc.),
contracting authority, the project manager and the contractor: • comparison of these "real" data with the assumptions
used in the studies,
• summarise in one single and shared document, and by • retro-analysis is used to recalibrate certain model
project section, the expected geological, hydrogeological assumptions and enable the projected displacements
and geotechnical conditions, the set values of the key induced to be updated for the remainder of the project
parameters of the machine, the necessary controls, the (i.e. improved reliability of the reference scenario).
spatial identification of possible residual risks, etc.,
• summarise the feedback as the tunnelling advances In order to address the above points and be useful during the
and reuse it for successive tunnelling sections within work, the approach implemented must be sufficiently flexible and
the same project, and also for future projects, summarised in relatively simple formats (e.g. overview diagram
• improve the boring rate, for example through the possibility of on the longitudinal profile of the tunnel), be rapid and facilitate
integrating feedback to review the confinement pressures, communication.

29
5
SUMMARY

30
31
6
GLOSSARY
Geometric parameters
D (m) Tunnel diameter
C (m) Overburden thickness
H (m) Depth of the tunnel axis
d (m) Length of the unsupported span

Geological, hydrogeological and geotechnical parameters

Hw (m) Height of water in relation to the axis of the tunnel
k (m/s) Ground permeability coefficient
γ (kN/m³) Moist soil unit weight
γ’ (kN/m³) Soil unit weight under buoyancy
Cu (kPa) "Undrained" cohesion of the soil
c’ (kPa) Effective cohesion of the ground
φ’ (°) Internal friction angle of the ground

Parameters related to boring and neighbouring structures

EPBS Earth pressure balanced shield
SS Slurry shield
σs (kPa) Surface pressure
σT (kPa) Pressure exerted on the face (assumed constant)
σT-eff (kPa) Face pressure value to be exerted to prevent its collapse
Vavct (m/h) Advance speed of the face

32
7
REFERENCES
AFNOR (2005), Eurocode 7 : Calcul géotechnique – Partie 1 : Bezuijen, A., & van Seters, A. (2006). The stability of a tunnel face
règles générales. NF EN 1997-1, 138 p. in soft clay. In H. van Lottum, & A. Bezuijen (Eds.), Tunnelling;
a decade of progress (pp. 149-155). Taylor & Francis.
AFTES WG8 (2006), La conception et la réalisation des travaux
d’injection des sols et des roches, Recommandation R2F1, Bienawski Z.T. (1989). Engineering Rock Mass Classification,
Tunnels et Espace Souterrain, n°194-195, pp. 70-154. Wiley, New York, 251 p.

AFTES WG16 (2020), Effects caused by excavation on Broere W. (2001), Tunnel face stability & new CPT application,
neighbouring structures in the design and construction of PhD Thesis, Delft university of technology, Delft university press.
underground works, Recommandation R2A1, 68 p.
Broms B.B., Bennermark H. (1967), Stability of clay at vertical
AFTES WG20 (2001), Design of sprayed concrete for openings, Journal of Soil Mechanics and Foundations Division,
underground support, Recommandation R1A1, Tunnels et vol. 93, pp. 71-94.
Ouvrages Souterrains, n°164, pp.68-108.
CEDD (2015), Catalogue of notable tunnel failures – case histories
AFTES WG24 (2008), Les reconnaissances à l’avancement, (up to April 2015), Civil Engineering and Development Department,
Recommandation R1F1, Tunnels et Espace Souterrain, n°209, Government of Hong Kong, https://www.cedd.gov.hk
pp. 302-325.
Chambon P. & Corté J.F. (1989). Stabilité du front de taille
AFTES WG29 (2007), Compatibilité des recommandations d'un tunnel faiblement enterré: modélisation en centrifugeuse,
AFTES relatives aux revêtements des tunnels en béton avec Proceedings of the Int. Conf. Tunnelling and Microtunnelling in Soft
les Eurocodes, Recommandation R2F1, Tunnels et Ouvrages Ground: from field to theory, Paris, presses ENPC, pp. 307-315.
souterrains, n°204, 18 p.
Chambon, P., Corte, J.F. (1994). Shallow tunnels in cohe-
AFTES WG30 (2021), Radial rock bolting in tunnels – Design sionless soil: stability of tunnel face, Journal of Geotechnical
and sizing guide, Recommandation R1A1, 97 p. Engineering 120 (7), pp. 1148–1165.

Alagha S.N, Chapman D.N. (2019), Numerical modelling of Champagne de Labriolle G. (2018), Détermination de la fenêtre
tunnel face stability uin homogeneous and layered soft ground, de pilotage de la pression de confinement d’un tunnelier
Tunnelling and Underground Space Technology, vol. 94, fermé dans un sol cohérent-frottant ou purement cohérent,
pp. 1-14. Revue Française de Géotechnique, vol. 155, n°3.

Anagnostou G. (2012), The contribution of horizontal arching Chang M., Chen C.C., Huang R.C., Chang J., Yang P.J. (2009),
to tunnel face stability, Geotechnik, vol 35, n°1, pp34-44. Investigation on mechanism of grouting and engineering
characteristics of in-situ grouted soils, Proceedings of the
Anagnostou G., Kovari K. (1996), Face stability conditions with 17th International Conference on Soil Mechanics and Geotechnical
Earth Pressure Balanced Shield, Tunnelling and Underground Engineering, M. Hamza et al. (Eds.), pp. 2346-2349,
Space Technology, vol 11, n°2, pp. 165-173.
Clouterre (2002), Additif 2002 aux recommandations Clouterre
Anagnostou G., Perazzelli P. (2015), Analysis method and 1991, pour la conception, le calcul, l’exécution et le contrôle
design charts for bolt reinforcement of the tunnel face in des soutènements réalisés par clouage des sols, Presses de
cohesive-frictional soils, Tunnelling and Underground Space l’École Nationale des Ponts et Chaussées, 217 p.
Technology, vol 47, pp. 162-181.
Croce P., Flora A., Modoni G. (2014), Jet-grouting technology, design
Berthoz N. (2012), Modélisation physique et théorique du and control, London, UK : CRC Press (Taylor & Francis Group).
creusement pressurisé des tunnels en terrains meubles
homogènes et stratifiés, Thèse de doctorat de l’ENTPE, 292 p. Dano C. (2001), Comportement mécanique des sols injectés,
Thèse de doctorat de l'Université de Nantes, 216 p.
Berthoz N., Branque D., Subrin D., Wong H., Humbert E. (2012),
Face failure in homogeneous and stratified soft ground : Theoretical DAUB (2016), Recommendations for Face Support Pressure
and experimental approaches on 1g EPBS reduced scale model, Calculations for Shield Tunnelling in Soft Ground, German
Tunnelling and Underground Space Technology, vol. 30, pp. 25-37. Tunnelling Committee, 64 p.

33
7
Davis E.H., Gunn M.J., Mair R.J., Seneviratne H.N. (1980). Marinos V., Marinos P., Hoek E. (2005), The geological strength
The stability of shallow tunnels and underground openings in index : applications and limitations, Bulletin of Engineering
cohesive material, Géotechnique, n°40, pp. 397-416. and Geological Environment, n°64, pp. 55-65.

Gilleron N. (2016), Méthode de prévision des tassements Messerli J., Pimentel E., Anagnostou G. (2010), Experimental
provoqués par le creusement des tunnels urbains et influence study into tunnel face collapse in sand, Proceedings of the
des présoutènements, thèse de doctorat de l’université 7th International Conference in Physical Modelling in Geotechnics,
Paris-Est, 238 p. Zurich, Switzerland.

Grasso P., Xu S., Del Fedele M., Russo G., Chiriotti E. (2003). Mollon G., Dias D., Soubra A.H. (2011), Rotational failure
Particular failure mechanisms of weathered granite observed mechanisms for the face stability analysis of tunnels driven
during construction of metro tunnels by TBM, ITA World Congress, by a pressurized shield, International Journal for Numerical and
Amsterdam. Analytical Methods in Geomechanics, vol 35, n°12, pp. 1363-1388.

Gunn MJ. (1980), Limit analysis of undrained stability problems Mollon G., Dias D., Soubra A.H. (2013), Continuous velocity fields
using a very small computer, Proc., Symp. on Computer for collapse and blowout of a pressurized tunnel face in purely
Applications to Geotechnical Problems in Highway Engineering, cohesive soil, International Journal for Numerical and Analytical
Cambridge Univ., Cambridge, UK, pp. 5–30. Methods in Geomechanics, vol 37, n°13, pp. 2061-2083.

Horn, M. (1961), Alagutak homlokbiztositstasra hato vizszintes NF EN 1997-1 (2005), Eurocode 7 : Geotechnical Design – Part 1:
földnyomasgalat néhany eredménye, Azorszdgos mélyépitoipari General Rules.
konferencia éloadasai, Közlekédési Dokumentacios Vallalat,
NGI (2013), Using the Q-system – Rock mass classification
Budapest (in Hungarian).
and support design, Handbook of the Norvegian Geotechnical
Institute, 57 p.
Ibrahim E., Soubra A.H., Mollon G., Raphael W., Dias D.,
Reda A. (2015), Three-dimensional face stability analysis of
Pan Q., Dias D. (2016), The effect of pore water pressure on
pressurized tunnels driven in a multilayered purely frictional
tunnel face stability, International Journal for Numerical and
medium, Tunnelling and Underground Space Technology,
Analytical Methods in Geomechanics, vol. 40, n°15. pp. ..-..
vol. 49, pp. 18-34.
Pan Q., Dias D. (2017), Upper-bound analysis on the face
ISGF WG 2 (1992), Frozen ground structures – Basic principles
stability of a non-circular tunnel, Tunnelling and Underground
of design, by Andersland O., Berggren A.L and Fish A.,
Space Technology, vol. 62, pp. 96-102.
Ground Freezing 91, vol 2. pp. 503-513.
Pan Q., Dias D. (2018), Three dimensional face stability of a tunnel
Kimura, T. and Mair, R. J., (1981), Centrifugal testing of in weak rock masses subjected to seepage forces, Tunnelling
model tunnels in soft clay, Proceedings of 10th International and Underground Space Technology, vol. 71, pp. 555-566.
Conference on Soil Mechanics and Foundation Engineering,
Stockholm 1981, Vol. 1, pp 319-322. Paternesi A., Schweiger H.F., Ruggeri P., Fruzzetti V.M.E.,
Scarpelli G. (2017), Comparisons of Eurocodes design
Leca E. (1997), Développement d’outils de calcul pour le approaches for numerical analysis of shallow tunnels, Tunnelling
dimensionnement des tunnels creusés en terrains meubles, and Underground Space Technology, vol. 62, pp. 115-125.
Thèse d’habilitation, Université de Lille, 83 p.
Peila D. (1994), A theoretical study of reinforcement influence
Leca E., Dormieux L. (1990), Upper and lower bound solutions on the stability of a tunnel face, Geotechnical and Geological
for the face stability of shallow circular tunnels in frictionnal Engineering, n°12, pp. 145-168.
material, Geotechnique, vol. 40, n°4, pp. 581-606.
Perazzelli P., Anagnostou G. (2017), Analysis Method and
Li T.Z., Yang X.L. (2020), Stability of plane strain tunnel Design Charts for Bolt Reinforcement of the Tunnel Face in
headings in soils with tensile strength cut-off, Tunnelling and purely Cohesive Soils, J Geotech Geoenviron Eng 143 (9).
Underground Space Technology, vol. 95, pp. 1-11.
Perazzelli P., Leone T. Anagnostou G. (2014), Tunnel face stability
Li Z.W., Yang X.L., Li T.Z. (2019), Face stability analysis of under seepage flow conditions, Tunnelling and Underground
tunnels under steady unsaturated seepage conditions, Tunnelling Space Technology, vol. 43, pp. 459-469.
and Underground Space Technology, vol. 93, pp. 1-8.
Piakowski A., Kowaleski Z. (1965), Application of thixotropic
Lü X., Zhou Y., Huang M., Zeng S. (2018), Experimental study clay suspensions for stability of vertical sides of deep trenches
of the face stability of shield tunnel in sands under seepage without strutting, Proceedings of 6th International Conference
condition, Tunnelling and Underground Space Technology, on Soil Mechanics and Foundations Engineering, Montreal,
vol. 74, pp. 195-205. vol. 111.

34
Quarmout M., König D., Gussmann P., Thewes M., Schanz T. Viana Da Fonseca A., Topa Gomes A. (2011), A tunnel collapse
(2019), Tunnel face stability analysis using Kinematical Element on the construction in metro do Porto : solutions for optimization
Method, Tunnelling and Underground Space Technology, vol. 85, of advance control parameters of a EPB TBM, Proceeding, of
pp. 354-367. the XXV technical conference, Miedzyzdroje, 24-24 mai 2011.

Salençon J. (1990), An introduction to the yield design theory Vu M.N., Broere W., Bosch J. (2015), The impact of shallow
and its application to soil mechanics, European Journal of cover on stability when tunnelling in soft soils, Tunnelling and
Mechanics A-Solids, vol 9, n°5, pp. 477-500. Underground Space Technology, vol. 50, pp. 507-515.

Schofield AN. (1980), Cambridge geotechnical centrifuge Wong H., Subrin D. (2006), Stabilité frontale d'un tunnel : mécanisme
operations, Geotechnique, vol. 30, n°3, pp. 227-268. 3D en forme de corne de rhinocéros et influence de la profondeur,
Revue européenne de génie civil, vol. 10, n°4, pp. 429-456.
Schuerch R., Poggiati R., Anagnostou G. (2019), Design charts
for estimating face stand-up time in soft ground tunnelling, Xu T., Bezuijen A. (2018), Analytical methods in predicting
Proceedings of the World Tunnel Congress, Naples, 10 p. excess pore water pressure in front of slurry shield in saturated
sandy ground, Tunnelling and Underground Space Technology,
Senent S., Yi C., Jinenez R. (2020), An upper bound solution vol. 73, pp. 203-211.
for tunnel face stability analysis considering the free span,
Tunnelling and Underground Space Technology, vol. 103, Zapata Franco D., Janin J.P., Dano C., Le Bissonnais H.,
pp. 1-14. Falconi F., Gérardin C. (2020), Méthodes 3D simplifiées pour
évaluer l’impact du boulonnage au front : REX de la L11-GC03,
Subrin D. (2002), Études théoriques sur la stabilité et le Paris, Journées Nationales de Géotechnique et de Géologie
comportement des tunnels renforcés par boulonnage, thèse de l’Ingénieur, Lyon, 8 p.
de doctorat de l’INSA de Lyon, 210 p.
Zingg S., Anagnostou G. (2016), An investigation into efficient
Toraldo C., Modoni G., Ochmanski M., Croce P. (2018), The drainage layouts for the stabilization of tunnel faces in
characteristic strength of jet-grouted material, Géotechnique, homogeneous ground, Tunnelling and Underground Space
vol. 68, n°3, pp. 262-279. Technology, vol. 58, pp. 49-73.

Ukrichton B., Yingchaloenkitkhajorn K., Keawsawasvong S. Zou J., Chen G., Qian Z. (2019a), Tunnel face stability in cohesion-
(2017), Three-dimensional undrained tunnel face stability frictional soils considering the soil arching effect by improved
in clay with a linearly increasing shear strength with depth, failure models, Computers and Geotechnics, vol. 106, pp. 1-17.
Computers and Geotechnics, vol. 88, pp. 146-151.
Zou J.F., Qian Z.H., Xiang X.H., Chen G.H. (2019b), Face stability
Vermeer P.A., Ruse N., Marcher T. (2002), Tunnel heading stability of a tunnel excavated in saturated nonhomogeneous soils,
in drained ground, Tunnelling, Felsbau, vol. 20, n°6, pp. 8-18. Tunnelling and Underground Space Technology, vol. 83, pp. 1-17.

35
8 APPENDIX:
FACE STABILITY EXAMPLES
Some examples are presented in the following tables. If you have any additional examples, please send them in a similar format to
[email protected] so that they can be included in a future update of this information document.

Tunnel name: Schirmeck (France)

Tunnelling method: conventional

Year: 2004

Nature of the ground:

Transition zone between (i) a highly fractured rock facies, moderately
altered, rather dry, and (ii) a second very highly fractured and
completely altered facies, with fairly significant water inflows.

Observations and adaptation of support methods:

At PM 218, at the end of mucking, the face collapsed over almost its
entire width according to a fracturing plane with a dip of around 45°
in the direction of the tunnel. The total volume of the collapse was
(a) Collapse of the face at PM 218.
about 100 m3 (Figure a).
Tunnelling continued after stabilising the face with 28 m³ of sprayed
concrete and 30 4-meter long longitudinal expansive bolts.
Given the precarious stability conditions observed, front bolting
using 4-meter long longitudinal expansive bolts was systematically
applied up to PM 465. This prevented the emergence of a new global
instability mechanism, but local instabilities persisted, such as that
at PM 235 visible in Figure (b).

Overall implications for the project:

The total duration of the boring and support work was about twice
as long as expected. This is related in particular to the increase
in the length of tunnel with steel arches (to the detriment of
a bolted profile), naturally accompanied by a larger sprayed
concrete thickness. The use of more face bolts also contributes to
this, but this cannot be quantified precisely. (b) Localised side wall collapse
of the face at PM 235.

36
Tunnel name: Bois de Peu (France)

Tunnelling method: conventional

Year: 2006

Nature of the ground:

Wet clays.
Observations:
Three major face collapses, with a volume of more than 50 m³,
occurred at MP 518, 515 and 510 of the counter-boring of the descending
tube under an overburden thickness of about ten meters. The fibreglass
bolts, installed at the face with a density of 0.25 bolts per m², did
not prevent the face instabilities, which did not, however, spread to
the surface due to the umbrella vault initially installed.
Reinforcement and adaptations in methods:
Boring continued but was done in separate sections (top heading
and bench) for about fifteen meters, until the shear strength of
the ground improved.
Implications for the project:
Increase in the duration of the work by about 1 month, and an View of the face after one of the collapses.
additional cost of about 500 k€ ex. tax according to the final general The face bolts as well as the umbrella vault are clearly visible.
cost calculations for the works contract. © Nicolas Dupriez, DDE 25.

Tunnel name: Porto (Portugal) metro

Tunnelling method: pressurized face TBM (earth pressure)

Year: 2001

Nature of the ground:

Mixed and heterogeneous face composed of unaltered granites in
the lower part of the face, and highly altered granites in the upper part.
Observations:
Three local instabilities observed during the boring of the first
600 m of metro line C, including one leading to a sinkhole causing
the collapse of a building and the death of a resident. This sinkhole
appeared about 30 m after the passage of the face. This was caused
by the over-excavation of the upper part of the face, linked to the
contrast in the strength between unaltered and altered granites
and by an initial difficulty (technical and organizational) in ensuring
a full and pressurized chamber.
(a) Sinkhole appeared on the surface.
Reinforcement and adaptations in methods:
Following this accident, the conditioning was reviewed, construction
crews were strengthened, monitoring procedures were revised,
and the TAP (Tunnelling Advance Plan) for each 500 m section was
introduced for the first time. In addition, a pressurized bentonite
injection system was added to the TBM. This was triggered
automatically when the pressure measured at the crown of the
working chamber fell below a preset value. No other sinkholes were
encountered during the remaining work on the C line and S line.
(b) Longitudinal section illustrating the mixed face encountered,
Implications for the project: and the complementary system for injecting bentonite at
The death of a third party necessarily had a very negative impact on the crown which was added following the collapse.
the start of this project. The corresponding financial consequences Reference : Grasso et al. (2003),
have not been published. The project was delayed by about 6 months. Viana Da Fonseca & Topa Gomes (2011).

37
8

Tunnel name: Rennes metro (France)

Tunnelling method: pressurized face TBM (earth pressure balanced shield)

Year: 2016

Nature of the ground:

Shale, which was more or less altered.

Observations:
Several surface sinkholes were generated when boring this metro
line. These include the one of about 20 m² and at least 3 m deep
generated in the Noz store. These sinkholes are linked to the
very particular structure of the rock mass, consisting of steeply
inclined banks (70°), whose discontinuities tended to open and
be lubricated under the effect of the confinement pressure.

Reinforcement and adaptations in methods:

During construction, a bentonite injection system (integrated
bubble) was added to the tunnel boring machine, and monitoring
Photograph of the sinkhole that appeared in the Noz store.
of the confinement pressure was increased, ensuring that it was It has a surface area of about 20 m² and is 3 m deep.
not too high. Three people fell into this sinkhole, but were able to get out
of it without serious injury.
Implications for the project: © Journal Ouest-France dated 2016/11/19.
The TBM was stopped for 133 days, costing an additional €10 million. Reference: tunnelling sheet for AFTES GT4 n°126, TES n°274.

Tunnel name: Nice metro (France)

Tunnelling method: pressurized face TBM (slurry shield)

Year: 2017

Nature of the ground:

Mixed subsidence composed of undulating ground (limestone and
puddingstone) and soft ground (sand and pebbles).

Observations:
Several sinkholes developed on the surface, including one
approximately 6 m in diameter and 1.5m in depth, in the middle
of the rue de France. The latter did not cause any damage to the
neighbouring buildings nor any traffic accident. This sinkhole was
caused by the TBM encountering an old cavity filled with construction Photograph of the sinkhole in Rue de France.
materials, notably steels stuck in the cutter head, with a loss of © Patrick Allemand.
pressure in the working chamber. Reference: Batiactu.com, 2017/07/05 and 2017/08/25.

Reinforcement and adaptations in methods:

Injections were carried out over fifteen meters around the sinkhole
to improve the stability of the ground, then the cutter head was
released, enabling tunnelling to be resumed.

Implications for the project:

Work stoppage of about 1.5 months.

38
Authors: Nicolas BERTHOZ, Didier SUBRIN.

Internal reviewers : Michel DEFFAYET, Eric PREMAT, Cédric GAILLARD, Gilles HAMAIDE, Johan KASPERSKI.

External reviewers : Denis BRANQUE (ENTPE), Martin CAHN (TERRASOL-SETEC), Guillaume CHAMPAGNE DE
LABRIOLLE (ARCADIS), Elena CHIRIOTTI (INCAS PARTNERS), Fabrice EMERIAULT (Grenoble INP), Jean-Pierre JANIN
(LOMBARDI INGENIERIE), Christophe JASSIONNESSE (SPIE BATIGNOLLES), Adrien SAITTA (EGIS TUNNELS).

39
 Photo credits: CETU / Cover: Nicolas Dupriez (DDE 25) - Content and coordination: CETU personnel - Desktop publishing: PAO Concept - Printing: VASSEL GRAPHIQUE

Centre d’Études des Tunnels (Centre for Tunnel Studies)

25 Avenue François Mitterrand MINISTÈRE

69500, Bron, France CHARGÉ
Tel. +33 (0)4 72 14 34 00 DES TRANSPORTS
Fax. +33 (0)4 72 14 34 30
[email protected]

www.cetu.developpement-durable.gouv.fr