Sunteți pe pagina 1din 23

Rock Mech Rock Eng (2012) 45:475497

DOI 10.1007/s00603-012-0224-3

ORIGINAL PAPER

A Completely 3D Model for the Simulation of Mechanized


Tunnel Excavation
Kai Zhao Michele Janutolo Giovanni Barla

Received: 20 October 2011 / Accepted: 4 February 2012 / Published online: 9 March 2012
 Springer-Verlag 2012

Abstract For long deep tunnels as currently under construction through the Alps, mechanized excavation using
tunnel boring machines (TBMs) contributes significantly to
savings in construction time and costs. Questions are,
however, posed due to the severe ground conditions which
are in cases anticipated or encountered along the main
tunnel alignment. A major geological hazard is the
squeezing of weak rocks, but also brittle failure can represent a significant problem. For the design of mechanized
tunnelling in such conditions, the complex interaction
between the rock mass, the tunnel machine, its system
components, and the tunnel support need to be analysed in
detail and this can be carried out by three-dimensional (3D)
models including all these components. However, the stateof-the-art shows that very few fully 3D models for
mechanical deep tunnel excavation in rock have been
developed so far. A completely three-dimensional simulator of mechanised tunnel excavation is presented in this
paper. The TBM of reference is a technologically advanced
double shield TBM designed to cope with both conditions.
Design analyses with reference to spalling hazard along the
Brenner and squeezing along the LyonTurin Base Tunnel
are discussed.

apeak
ares
B
c
Cc
CI
D
E
f
F
fck

Keywords Double shield TBMs  Long deep tunnels 


Spalling  Squeezing  3D modelling  Alpine Base Tunnels

Ks
Ksn

List of Symbols
A
Surface area

Kst

Ff
FN
FR
G
g
K 
K
Kn
Kns
Knt

Kt
Ktn
K. Zhao  M. Janutolo (&)  G. Barla
Department of Structural and Geotechnical Engineering,
Politecnico di Torino, Turin, Italy
e-mail: michele.janutolobarlet@polito.it
URL: http://www.polito.it\rockmech

Kts
mdil

Peak curvature exponent (HoekBrown)


Residual curvature exponent (HoekBrown)
Strain displacement matrix
Cohesion (MohrCoulomb)
Cutting coefficient
Crack initiation threshold
Excavation diameter
Youngs modulus
Yield function
Total thrust force
Characteristic compressive cylinder strength of
concrete
Thrust force to overcome friction
Cutterhead thrust force
Cutterhead rolling force
Shear modulus
Plastic potential function
Elastic stiffness matrix of the interface element
Bulk modulus
Elastic normal stiffness of the interface element
Off-diagonal term of the stiffness matrix of the
interface element
Off-diagonal term of the stiffness matrix of the
interface element
Elastic shear stiffness of the interface element
Off-diagonal term of the stiffness matrix of the
interface element
Off-diagonal term of the stiffness matrix of the
interface element
Elastic shear stiffness of the interface element
Off-diagonal term of the stiffness matrix of the
interface element
Off-diagonal term of the stiffness matrix of the
interface element
Dilation parameter (HoekBrown)

123

476

mpeak
mres
N
N 
n
Ni
P
p
p
r
R
RPM
speak
sres
t
T
Ttot
Tf
Tr
fug
u
ucrown
uf
ufin
uinvert
un
us
ut
UCS
fvg
fvgtop
fvgbot
DD
Dg
Dr
Dzmin
c
di
dei
dpi
{d}
ds
dt
dn
ep
epi
epl
k

K. Zhao et al.

Peak HoekBrown slope constant


Residual HoekBrown slope constant
Number of elements in the shield surface
Matrix of shape function
Number of cutters
Shape functions
Cutting power
Ground pressure
Cutterhead pressure
Shield radius
Excavation radius
Cutterhead rotational speed
Peak intercept constant (HoekBrown)
Residual intercept constant (HoekBrown)
Time
Tensile strength
Total torque
Torque to overcome friction
Torque due to the rolling forces
Vector of displacements at the element faces
Total radial displacement
Radial displacement at the crown
Displacement at the face
Final radial displacement
Radial displacement at the invert
Normal displacement at the element faces
Tangential displacement at the element faces
Tangential displacement at the element faces
Uniaxial compressive strength
Vector of nodal displacements at the element
faces
Vector of nodal displacements at the top element
face
Vector of nodal displacements at the bottom
element face
Overexcavation
Gap between the shields and the rock mass
Conicity
Smallest width of an adjoining zone in the normal
direction
Unit weight
Total slip displacement
Elastic slip displacement
Plastic slip displacement
Vector of relative displacements at the interface
Tangential relative displacement at the interface
Tangential relative displacement at the interface
Normal relative displacement at the interface
Hardening/softening parameter
Principal plastic strain
Equivalent plastic strain
Scalar multiplier for plastic strain

123

l
m
r
r1
r3
rh
rv
s
u
w

Friction coefficient
Poissons ratio
Stress
Maximum principal stress
Minimum principal stress
Horizontal stress
Vertical stress
Shear stress
Friction angle (MohrCoulomb)
Dilation angle (MohrCoulomb)

1 Introduction
Today, almost all rock mass conditions can be bored by
modern TBMs with tunnel diameter varying from less than
3 m to more than 15 m. Long deep tunnels such as Alpine
Tunnels are excavated mainly by mechanised tunnelling, as
TBMs contribute significantly to savings in construction
time and costs. TBMs have already been used in the
Lotschberg and the Gotthard Base Tunnel and will be
employed in future tunnels to be excavated through the
Alps such as the Brenner and LyonTurin Base Tunnels.
However, open issues remain when dealing with critical
hazards encountered in these tunnels relating mainly to
stress-induced brittle failure of hard rocks (spalling, rock
bursting), squeezing ground behaviour, structurally controlled failures and water inflows, also in view of identifying the possible countermeasures to be applied (Loew
et al. 2010). Use of TBMs in very severe ground conditions
is yet under discussion due to some negative experiences
which resulted in very low rates of advancement and even
in standstill, as in the case of the YacambuQuibor Tunnel
(Hoek and Guevara 2009) or in the Headrace Tunnel of the
Gilgel Gibe II Hydroelectric Project (Barla 2010).
For the design of mechanised tunnelling in such conditions, the complex interaction between the rock mass, the
tunnel machine, its system components, and the tunnel
support has to be analysed in detail and three-dimensional
models including all these components are suitable to
correctly simulate this interplay and avoid the errors
introduced by assumption of plane strain conditions
(Cantieni and Anagnostou 2009). This is even more so in
the case of the double shield universal TBM (DSU TBM),
which is indeed a more complex machine than the gripper
or the single shield TBM.
This paper is intended to describe an advanced 3D
model which has been recently developed for the detailed
simulation of the DSU TBM, with specific problems in
mind as in the case of excavation of deep tunnels through
rock masses which exhibit either spalling or squeezing

Simulation of Mechanized Tunnel Excavation

behaviour. A first presentation of such a model has been


given by Barla et al. (2011).
More comprehensive description of the relevant features
which characterise the 3D model developed so far is given
taking into account the recent improvements which have
been introduced with reference to the simulation of the
rock massTBM interaction. First, the state-of-the-art in
numerical modelling of TBM excavation is briefly
reviewed. Then, the latest improvements in TBM technology are presented and the features of a TBM of reference for this paper are defined. Then, the numerical model
is described in detail. Finally, the case studies of the
Brenner and the LyonTurin Base Tunnel are considered in
order to illustrate the model.

477

2006) and Nagel et al. (2008). The model includes the soil,
the shield machine, the hydraulic jacks, the tunnel lining
and the tail void grout as separate components. The
advancement is simulated by a step-by-step procedure. It
should be noted, however, that excavation with closed
system shield TBMs in shallow soil tunnels is rather different with respect to rock TBMs such as double shield
TBMs, especially if attention is paid to the ground
behaviour, to the interplay between the ground, the TBM
and the support components (e.g., absence of grippers,
different cutterhead pressures, different methods of annulus
grouting) and to the design objectives (e.g., surface
settlements).

3 Rock TBMs for Deep Tunnels


2 Modelling of TBMRock Mass Interaction
3.1 TBM Technology
For modelling the TBM excavation in rock masses,
including the analysis of TBMrock mass interaction, two
main methods have been employed in the literature: the
axisymmetric models and the fully 3D ones.
The most relevant axisymmetric simulations have been
proposed by Ramoni and Anagnostou (2006, 2010, 2011),
who have studied the case of squeezing ground using a
steady-state method. The numerical model has been formulated in a frame of reference that is fixed to the
advancing heading, alike a furrow moving along a ship
(Nguyen-Minh and Corbetta 1992). The one-step solution
method corresponds to the limiting case of a step-by-step
model with zero round length (Cantieni and Anagnostou
2009) and for this reason is well suited to reproduce TBM
excavation, which is a continuous process.
3D models of deep tunnel excavation in rock masses
have been proposed by Cobreros et al. (2005) and by Simic
(2005) for the Guadarrama Tunnel (Spain) and by Graziani
et al. (2007) for the Brenner Base Tunnel. Both refer to
double shield TBM excavation in squeezing rocks and both
are finite difference models. The TBM is modelled as a
cylinder using solid elements with variable thickness, while
shell elements represent the precast lining. A large displacement approach as well as special interface elements
was applied in order to simulate the progressive closure of
the gap between the shield and the excavation wall. Similar
interface elements were used between the segmental lining
and the rock surface in order to represent the effect of a
compressible backfilling material. A step-by-step method
was adopted.
Many models have been instead developed for the
simulation of shallow tunnel excavation in soil by closed
system TBMs such as Earth pressure balance (EPB) and
slurry shield (SS) machines. The most remarkable model
presently available is due to Kasper and Meschke (2004,

A worldwide trend towards the increased use of segmentally lined tunnels (Asche et al. 2011), excavated by single
or double shield TBMs, is to be noted. Today, single shield
TBMs are mainly used in soils or soft and weak rocks.
Double shield TBMs have come into common use as they
can cope with hard rocks as well as weak and unstable
rocks. As shown in Fig. 1, they consist of the front shield
with a cutterhead, main bearing and drive and a gripper
shield with clamping unit (gripper plates), tail shield and
auxiliary thrust cylinders. Both parts are connected by a
section (the telescopic shield) with telescopic thrust cylinders, which operate as the main thrust cylinders. The
basic principle (conventional mode) is that the machine is
clamped radially to the tunnel wall through the grippers
and the excavation and installation of the segmental lining
are performed at the same time.
Where the rock is weak and it is not possible to clamp
radially through the grippers, the necessary thrust forces
can either be provided by the telescopic cylinders or by the
auxiliary thrust cylinders. In the first mode with the telescopic cylinders, the auxiliary cylinders only transfer the
thrust forces to the segmental lining. In the second mode,
which is also called single shield mode, the front and
gripper shield form a stiff unit and the auxiliary cylinders
produce the necessary forward thrust (Maidl et al. 2008).
Recently, the range of application has been extended:
(a) downwards, even more in the direction of incompetent and squeezing rock formations as well as
(b) upwards, even more in the direction of very competent and extremely hard rock formations. In case
(a) there is an extended cutter head torque and thrust force
required whereas in case (b) an extended cutterhead thrust
capacity would be necessary in order to achieve appropriate penetration rates.

123

478

K. Zhao et al.

Fig. 1 Longitudinal section of a double shield TBM (Maidl et al. 2008)

Furthermore, the design of double shield TBM was


improved with the introduction of the double shield universal (DSU) TBM. With respect to the traditional double
shield TBM, DSU TBMs have a shorter shield length.
Shorter shield length (of about 1 tunnel diameter) means
that the redistribution of the stresses is not yet developed
completely and consequently the possible squeezing forces
on the shield (and the risk of getting trapped) will be lower.
Another development is the stepwise reduction of the rear
shield (standing for inner telescopic, gripper and tail shield)
diameter, also called conicity (Fig. 2). Traditional
machines have, in fact, a decrease of the diameter of the
telescopic shield, but a further increase in the gripper
shield, so that the latter has a very similar diameter to the
front shield. Single shield machines have instead a similar
stepwise reduction (e.g., the Lake Mead Intake No. 3 TBM;
Anagnostou et al. 2010). This offset could lead to an
accumulation of material (slabs of rocks in case of spalling
or fracturing due to joints) or to an interlocking of
squeezing rock resulting in a hindrance of the advance (as
the rear shield can be blocked) or even in a stoppage of the
TBM. Furthermore, this stepwise reduction of the diameter
provides more space for deformations, reducing the risk of
shield jamming.
All the shields have the same axis with an offset with
respect to the cutter head axis. Another offset of the lining
axis results in a clearance for the convergences of nominal
185 mm (e.g.) on top in the case of the Abdalajis Tunnel
(Gutter and Romualdi 2003). However, severe squeezing
conditions may lead to deformations that are much greater
than the gap deriving only by the overcut and the conicity,
so that nearly the totality of the shield surface would be in
contact with the rock mass. In order to allow the technical
feasibility of the TBM drive, some countermeasures are

123

Fig. 2 Construction schemes for the telescopic shield of double


shield TBMs: a classic design and b modified DSU design
(Ramoni and Anagnostou 2010)

possible (Ramoni and Anagnostou 2010): pre-treatment


(e.g., by grouting or drainage) or pre-support of the ground
(e.g., by pipe umbrella), overboring (increase of boring
diameter), installation of higher thrust force and torque and
reduction of the shield skin friction (lubrication of the
surface). Several overboring systems are technologically
feasible, limited to a maximum of 30 cm (in diameter),
even if several difficulties in using this technology exist
(Ramoni and Anagnostou 2010). The maximum (i.e.,
auxiliary) thrust force for double shield TBMs in the last
years has increased up to 150 MN and the maximum torque up to 30 MNm.
3.2 Main Features of a DSU TBM for Alpine Base
Tunnels
In Base Tunnels under the Alps, double shield universal
TBMs can be very suitable to deal with the very different
rock conditions which are expected to be encountered
within a short distance, including squeezing and spalling
problems. Also, this might help in preventing as much as
possible the switch to conventional excavation in the case
of squeezing rock.

Simulation of Mechanized Tunnel Excavation

For the purpose of the present study, the tunnel internal


diameter has been fixed to be 8.1 m, due to the train size,
considering the minimum cross section in order to reduce
the resistance to the advancement of the high-speed trains
and the space for the equipment. The thickness of the
segmental lining ranges in general from 30 to 50 cm. The
load on the lining is only due to the rock mass, while water
pressure is disregarded, as in zones with high water pressure durable drainage systems need be adopted (Gutter and
Romualdi 2003).
The concrete classes for the lining range in general from
C20 (fck = 20 MPa) to C45 (fck = 45 MPa) and the related
elastic moduli (according to the Eurocode 2) from 30 to
36 GPa. Higher classes can be used, but less frequently.
These two classes (C20 and C45) may conveniently be
chosen, respectively, for the competent rock and for the
squeezing rock. In intermediate rock mass behaviour,
intermediate concrete classes can be used. Therefore,
according to the geological and geotechnical conditions
anticipated, the tunnel has to be divided into sections (e.g.,
4 or 5) corresponding to different types of segments. Only
the thickness of the precast segment is maintained constant
(e.g., 45 cm) all along the excavation due to the machine
configuration and to maintain the same internal diameter.
In the conventional mode, the main thrust force is provided through the grippers. In general, two large area shoes
are used to ensure surface contact and low pressure against
the rock wall (max 4 MPa), so that the TBM can advance in
the conventional mode also in poor ground conditions
(except the very severe squeezing ground). The re-gripping
length is equal to the stroke and it is generally between 1 and
2 m.
The maximum cutterhead thrust has been set to 17 MN, in
order to achieve the necessary penetration rates for hard rock.
64 conventional 17-inch diameter cutters (with a single load
of 267 kN) can be used. Five additional 17-inch diameter
cutters are provided for the max radial overboring of 70 mm.
Table 1 shows the main features of the TBM to be
considered in the following, whereas Fig. 3 gives the
schematic view of the assumed TBM arrangement in the
case of competent rock (hard rock) and Fig. 4 in the case
of squeezing rock (weak rock).

479
Table 1 TBM main features (assumed)
Description

Unit

Lining
Inner diameter

mm

8,100

Thickness

mm

450

Outer diameter

mm

9,000

Youngs modulus concrete

GPa

3036

Cutterhead
Excavation diameter

mm

9,300

Maximum cutterhead thrust

MN

17 (18.3 if
overboring)

Cutterhead installed power

MW

4.9

Overboring

mm

140

Maximum excavation diameter

mm

9,440

No.

64 (?5 for
overboring)

MN

50

00

Number of disc cutters (17 , 267 kN)


Machine
Maximum main thrust
Maximum auxiliary thrust

MN

80

Weight

kN

13,000

Pads dimensions H, B
Maximum pressure

mm
MPa

6,600, 2,000
4

Re-gripping

mm

2,000

Front shield diameter

mm

9,230

Outer telescopic shield

mm

9,230

Inner telescopic shield

mm

9,170

Gripper shield

mm

9,170

Grippers

Shield

Tail shield

mm

9,170

Length front shield ? outer telescopic

mm

5,000

Length rear shield (inner


telescopic ? gripper ? tail shield)

mm

6,000

available in literature (Sect. 2), therefore both hard rock


and weak rock conditions have been reproduced,
according to the TBM arrangements shown in Figs. 3 and
4. The presence of water pressure and consolidation
problems is not taken into account. The focus is on the
mechanical behaviour of the rock mass and on the interaction between the rock mass and the TBM and the support
components.

4 Modelling Approach
4.2 FEM Model
4.1 Foreword
4.2.1 Modelling the Rock Mass
In this paper, a simulator of TBM excavation of deep
tunnels has been developed using 3D FEM modelling and
the midas GTS (Geotechnical and Tunnel analysis System)
computer code (TNO DIANABV). This simulator is intended to be more general than the previous 3D models

Under the assumption of symmetry with reference to the


vertical plane through the tunnel axis, only half of the
entire domain is modelled. A cylindrical domain is used
(Fig. 5) in order to: (1) decrease the total number of

123

480

K. Zhao et al.

Fig. 3 Schematic TBM


arrangement in hard rock
(nominal layout)

Fig. 4 Schematic TBM


arrangement in squeezing rock
(with maximum overboring)

Fig. 5 3D mesh at the initial


stage (only rock mass elements
are shown) with detail

elements; (2) obtain a mesh size with only one variation (in
the transversal plane, i.e., perpendicular to the tunnel axis);
and (3) improve the mesh geometric quality, that is, the
internal angles are close to 90o, the aspect ratios close to 1,
also by choosing an appropriate grading (in the transversal

123

plane). Eight-node hexahedron solid elements are chosen


as appropriate.
The mesh size should be large enough so that the
external boundary can represent an infinitely extended
medium. Otherwise, the presence of the artificial

Simulation of Mechanized Tunnel Excavation

boundaries will induce a significant influence on the stress


strain field around the tunnel. A rule-of-thumb is that the
mesh in the transversal plane is built by an expansion factor
of 10 in relation to the tunnel diameter D (Eberhardt 2001;
Graziani et al. 2007). In hard rock this value can be
reduced as the extent of the failure zones is rather small
and a value of approximately 6D is considered as satisfactory. On the contrary, for squeezing conditions, this
factor is to be increased and a value of 15D is used. In all
cases, parametric studies are necessary to evaluate the
appropriate size for every specific case.
In the longitudinal direction, Graziani et al. (2007) (weak
rock) suggested a total length of 20D, while Eberhardt
(2001) (hard rock) a length of the excavation steps of 10D:
this value depends on the construction length, as described in
Sect. 4.3, as well on a sufficient distance between the rear
boundary and the last excavation face (depending on the
plastic zone extension). The mesh discretisation is chosen to
be equal to the excavation length (1 m); then, ahead of the
final face position, it is gradually increased.
As far as mesh grading in the transversal plane, the
accuracy of the results is of course improved when this is
refined, especially for the severe mesh-dependent cases like
overstressed rock mass conditions (with squeezing and
spalling phenomena). In 3D modelling, a trade-off with the
high computational time and restriction of hardware
resources has to be found. In plastically strained regions
(squeezing rock), a high level of mesh refinement is
desirable to capture high strain gradients, as the accuracy
of the simulation depends on the ability of the mesh layout
to represent such gradients. In failure localization cases
(spalling rocks), the use of fine meshes near the excavation
boundary is recommended (Diederichs 2007), in order to
capture the shape and the extent of the failure zones.
Therefore, in this study, a very refined region around the
tunnel contour and a coarser zone away from it has been
used as shown in Fig. 5. Element size of 30 cm 9 30 cm
near the tunnel boundary has been used.
Once the 3D model is created, the following boundary
conditions are imposed:

In the circular outer boundary, displacements along the


vertical (Y) and the horizontal direction (X) are
prevented;
in the vertical outer boundary, corresponding to the
plane of symmetry, displacements along the horizontal
direction (X) are prevented;
in the front and in the rear outer faces, displacements
along the excavation advance direction (Z) are prevented.

As the model refers to deep tunnels, the in situ state of


stress is applied as a uniform initial stress without consideration of the free ground surface and of the stress
gradient due to the gravity. The three principal stresses can

481

be input independently and this is one of the important


advantages of a 3D model.
4.2.2 Modelling the Main TBM Components
The following TBM components are considered:

shields (front and rear), which are modelled with plate


elements applied at the excavation boundary and with
the stiffness properties of steel;
cutterhead, which is modelled with plate elements
(with the stiffness properties of steel) at the current
excavation face and where a pressure is applied (Sect.
4.5.3);
backfilling, which is modelled with solid elements; the
material filling the gap ranges from pea gravel to
cement grout and is given specific properties as desired
(Sect. 4.5.4);
lining, which is modelled with plate elements (with the
stiffness properties of concrete); no joints are introduced and the lining is considered to be continuous;
grippers, which are modelled with plate elements at the
excavation boundary (with the stiffness properties of
steel); they are positioned at a certain distance from the
face and a given pressure is applied on them;
jack pressure, which is applied, in the form of an edge
pressure, on the last lining ring installed (as in
Castellanza et al. 2008), if the rock is weak, in place
of the grippers.

All these components are modelled with a linearly


elastic isotropic law. For plate elements, four-node quadrilateral elements are used. The self-weight is applied to all
these components.
4.3 TBM Advancement and Simulation Procedure
The model simulates the ongoing TBM excavation by a
step-by-step method. This is characterized by the excavation length and construction stages. The excavation length
is taken to be equal to 1 m in order to:

reproduce as well as possible the continuous process


of TBM excavation (Vlachopoulos and Diederichs
2009);
improve the convergence of the elastic plastic solution;
trade-off accuracy and computational time.

The construction stages and the total number of steps to


be performed depend on the geometry of the excavation,
the TBM type, and the rock mass, as:

the construction length has to be such as to allow for a


representative steady-state condition to be reached, also
considering the boundary effects (the first few metres of

123

482

K. Zhao et al.

the excavation containing these effects will be removed


from the results);
the operation modes of the double shield universal
TBMs, which are different for hard rock and weak
rock.

Cutterhead

It is noted that the adopted step-by-step method leads to


simulate a succession of standstills and not a continuous
process. Computations take around 1 week using the processor One Intel Xeon W3680 (3.33 GHz, 6.4GT/s,
12 MB, 6C-Memory runs at 1,333 MHz).

Front shieldInvert
Gripper

(a) 10th step


Backfilling

4.3.1 Conventional Mode


In hard rock, the TBM operates using the grippers (basic
functional principle). This is valid not only for very competent, but also for medium to low quality rock masses.
The construction stages are defined as follows:

in the first step, initialisation takes place accounting for


the in situ stress field;
in the second step, the TBM enters the model (the
cutterhead elements are activated) and the first slice is
excavated;
in the third step, the first slice of the front shield invert
is activated;
in each step thereafter the TBM progresses into the
model;
in the tenth step, the grippers and the applied pressure
become active for the first time. The position of the
grippers is changed every 2 m, in order to simulate
the re-gripping of the machine, with a distance from the
face between 6 and 7 m. The re-gripping operation
phase (in which the cutterhead does not excavate and
the clamping of the grippers is released) is not
simulated as a stand-alone step but it is associated with
the further excavation step;
in the fifteenth step, the lining is activated as well as
backfilling;
the stages proceed until a steady-state condition is
reached.

Lining

(b) 15th step

(c) Final step


Fig. 6 3D model in a typical case of TBM excavation in hard rock

To illustrate the simulation process, the model for a


typical case is shown in Fig. 6. The shield invert in contact
with the rock mass (according to Sect. 4.5.2) is also shown.

4.3.2 Single Shield Mode

In weak rock, the machine operates in single shield mode.


The thrust force generated in the auxiliary hydraulic jacks
(the so-called jack pressure) is applied on the segmental
lining. The construction stages are defined as follows:

in the first step, initialisation takes place;

123

in the second step, the TBM enters the model and the
cutterhead is activated;
in the third step, the first slice of the shield is activated;
in the thirteenth step, the grouting with a softening
phase, the relevant pressure, the lining and the jack
pressure are activated. The excavation and the placing
of the rings are simulated in the same phase, assuming a
stroke of 1 m;
since the fifteenth step, the properties of the grouting
are changed into the hardening phase;
the stages proceed until a steady-state condition is
reached.

The model shown in Fig. 7 illustrates the typical


simulation process. The parts of the shields in contact
with the rock mass (according to Sect. 4.5.2) are also
shown.

Simulation of Mechanized Tunnel Excavation

483

4.4 Rock Mass Models

Rear shield
Cutterhead

Front shield

(a) 12 th step

The rock mass is considered to be continuous, homogeneous and isotropic. Any constitutive law based on these
assumptions for rock masses can be implemented in the
model. In this paper, the problems in mind are the spalling
and squeezing phenomena; therefore, two constitutive
models for reproducing them in a consistent and simple
way are illustrated in the following.
4.4.1 Brittle Behaviour (Spalling)

Lining

Groutingsoftening phase
Auxiliary thrust force

(b) 13 th step

X
Z

Groutinghardening phase

(c) Final step


Fig. 7 3D model in a typical case of TBM excavation in weak rock.
The two gradations of blue differentiate the parts of the shield in
contact with the rock mass (dark blue) with the parts which are not in
contact (light blue)

Different models have been developed for the simulation of


spalling around underground openings, as the fracturing
process of hard rocks is very complex and fracture
mechanics theory is not sufficient to characterise the entire
macroscopic fracture process. These models consider the
rock mass either as a continuum or as a discontinuum or as
a hybrid continuum/discontinuum.
Among the continuum approaches, a reliable simulation
of the failure zones can be obtained through the criterion
developed by Diederichs (2007), based on an elastic perfectly brittle plastic model. The composite strength envelope for brittle materials (Diederichs 2010), resulting from
studies and observations on the mechanisms leading to
spalling, can be implemented in numerical codes through
the use of the generalised HoekBrown criterion (Fig. 8).
Peak and residual yield functions are defined by damage
threshold and spalling limit, respectively. Shear at high
confinement is not correctly simulated in this model so the
use is limited to near-excavation analysis.
The procedure for determining the input parameters for
the generalized HoekBrown criterion is the following
(Diederichs 2010):

Fig. 8 Constitutive law for brittle behaviour: Diederichs (2007) criterion and stressstrain relationship. The intact rock criterion is shown for
comparison

123

484

K. Zhao et al.

determine the crack initiation threshold CI from


uniaxial compression tests as shown in Fig. 9;
set apeak to 0.25;
obtain a reliable estimate of uniaxial compressive
strength, UCS and tensile strength, T (from laboratory
tests);
calculate the appropriate s and m from:
speak CI/UCS

1
apeak

mpeak speak UCS=jT j:

1
2

In order to model the transition envelope to high


confinement shear (spalling limit), set ares = 0.75, sres = 0
or 10-6 (for numerical stability) and mres = 5 to 9.
The equivalent MohrCoulomb parameters can also be
used and the equivalent cohesion, friction and tensile
strength parameters can be varied according to the plastic
strain (Cohesion SofteningFriction Hardening approach,
CSFH; Diederichs 2007).
The process of fracturing generates dilation; nevertheless, an accurate reproduction of bulking and of the
resulting displacements with continuum models in spalling
rock is still an open issue, as actually the process is discontinuous after yield. According to Diederichs (2007), a
near-maximum dilation can be used for supported tunnels,
while a near zero dilation for unsupported conditions, as
there is free fallout of spalls of rock.
Zhao and Cai (2010) have proposed an expression for the
dilation angle w depending on the confining stress and on
the plastic strain. This approach has been implemented in
the FLAC computer code with the cohesion-weakening and

frictional softening model (CWFS), similar to the CSFH


model. They have also shown that a constant angle w  u2res
(ures is the residual friction angle according to the Mohr
Coulomb and the CWFS model) can be used to accurately
capture the spalling geometry in unsupported conditions.
4.4.2 Squeezing Behaviour
The rock mass behaviour in squeezing conditions is characterized by a time-dependent response of the tunnel,
deformations of the same cross section during standstill, a
large extent of the zone of influence of the excavation, and a
long lasting tendency to undergo deformations (Barla 2001).
The complexity of the problem and the difficulties to account
for time dependence through elasto-viscous plastic models
(Debernardi and Barla 2009) lead at the design stage to the
adoption of simplified models where the rock mass is linearly
elastic perfectly plastic, with strength and deformability
properties estimated in short-term and long-term conditions (Loew et al. 2010; Hoek and Guevara 2009).
Given that the interest of this paper is to study the
interaction between the rock mass, the TBM and the support system near the face and during excavation, the use of
short-term parameters can be acceptable. The gradual
increase of ground pressure and of ground deformations in
the longitudinal direction are therefore considered to be
only due to the spatial stress redistribution that is associated with the progressive advance of the working face
(Lombardi 1973).
The rock mass behaviour can be simulated with a Mohr
Coulomb criterion (Fig. 10) and a non-associated flow rule,
with the dilation angle w taken as a function of the angle of
internal friction u, as proposed by Vermeer and De Borst
(1984):
 
1
for u  20
w
3
u  20
for u [ 20
This assumption for friction angles in the range of u
20  25 has been verified to be reasonable in laboratory
tests of squeezing rock (Vogelhuber 2007).
4.5 Modelling the Interaction Between the Machine
Components and the Rock Mass
4.5.1 Interface Theory

Fig. 9 Strain measurements from uniaxial compression tests to


determine the crack thresholds (Diederichs 2010)

123

The simulation of interaction problems requires a special


attention to the discontinuous behaviour at the following
interfaces: (1) rock massshield and (2) backfillinglining.
This behaviour involves frictional sliding, possible closure
of the gap between the shield and the rock mass, and
possible contact/surface separation due to the self-weight
of the structural components.

Simulation of Mechanized Tunnel Excavation

485

Fig. 10 Constitutive law for squeezing behaviour: MohrCoulomb criterion and stressstrain relationship

Zero-thickness interface/joint elements are used to


simulate the desired behaviour, which can satisfy the
compatibility conditions and allow differential movements
of the rock mass and the support. The frictional slipping
occurs usually at shear levels that are significantly lower
than the shear strength of the adjacent rock mass or
backfilling.
4 ? 4 plane quadrilateral isoparametric interface elements are adopted (Fig. 11). Zero thickness is assigned as
initially the nodes of the external elements (the rock mass
and backfilling elements) have the same coordinates as the
nodes of the internal elements (shields and lining elements,
respectively). The vector of relative displacements fdg
between two homologous points can be obtained from the
displacements associated to the elements faces (top and
bottom):
8 9 8 9
8 9
>
>
= >
=
=
< ds >
< us >
< us >
 ut
4
fdg dt ut
>
>
; >
;
;
: >
: >
: >
dn
un top
un bot
where s and t represent the tangential directions and n the
normal direction.
The relation between the continuous displacement field
and the nodal displacements is given by:
fug N  fvg

where N  is the shape function and fvg is the vector of


nodal displacements.
Hence, the relation between nodal displacements and
relative displacements for the interface elements is written
as:
h
i
6
fug N fdg N  fvgtop fvgbot Bfvg
where

fugT f us1

ut1

B N1

N2

un1

. . . us8

N3

N4

ut8

un8 gT

N1

N2

N3

7
N4 
8

Note that the matrix B is analogous to the strain


displacement matrix used in the formulation of continuum
elements except that it does not involve differential
operators applied to shape functions since the strain of
the interface element is defined as the relative
displacement.
Then, after variation of the total potential energy with
respect to the nodal displacement vector, the stiffness
matrix of the interface element with zero thickness is
obtained:
Z
K  BT DBdA
9
A

The matrix D is used to denote the relation that


describes the constitutive behaviour of the interface
element:
2
3
Ks Kst Ksn
D 4 Kts Kt Ktn 5
10
Kns Knt Kn
where Kn is the elastic normal stiffness and Ks and Kt the
shear stiffnesses.
The matrix in (10) should be defined by considering two
aspects. Firstly, it is appropriate to take Kst = Kts =
Ksn = Kns = Ktn = Knt = 0, so that changes of stress in
the tangential and normal directions are unrelated to
changes of elastic deformation in the normal and tangential
directions, respectively. The normal and shear relationships
are assumed to be independent when the stresses are within
the elastic region. Secondly, since no-overlapping occurs

123

486

K. Zhao et al.

un
us
ut
Fig. 11 Interface topology

between the rock mass and the support elements near the
interface and no slip before yielding of the interface is
assumed (Fakharian and Evgin 2000; Cai and Ugai 2000),
the stiffnesses are actually penalty numbers that approximately enforce contact-surface compatibility consisting of
impenetrability and pre-sliding stick constraints. The normal and shear stiffnesses are therefore set to be much
greater than the stiffness of the softer neighbouring zone.
It is noted that too high values of Kn, Ks and Kt may
produce numerical errors related to the computer precision.
The values should be less than 100 times the stiffness of the
adjacent element according to Day and Potts (1994). A
rule-of-thumb is suggested in the FLAC manual (Itasca
2006), where the normal and shear stiffness can be set
equal to ten times the equivalent stiffness of the softer
neighbouring zone, which is given by:
K 4=3G
DZmin

11

where K and G are the bulk and shear moduli, respectively,


and Dzmin is the smallest width of an adjoining zone in the
normal direction.
Since plasticity and friction share the main features
(Drucker 1954), the classical plasticity theory is used to
characterize the incremental slip deformations by:
d_ i

d_ ei

d_ pi 0

d_ pi

12

if f \0 or f_\0

og
d_ pi k_
ori

if f f_ 0

13
14

where f is the yield (slip) function, g the potential function


and k a positive scalar.
In this model, the Coulomb friction law defines f and g
as:
f jsj r tan u  c

15

g jsj r tan w

16

where u and w are the friction and dilation angles,


respectively, whereas c is the cohesion.
It is considered that the contact surfaces between the
rock mass and the shield (steel) and between the backfilling

123

and the segmental lining (precast concrete) are smooth and


continuous. u is taken as constant, c is equal to zero and w
is also zero for the interface between the rock mass and the
shield. Since both the backfilling and the concrete are
frictional materials (according to Chen 1994), the dilation
angle is set to be 1 for the interface between the backfilling and the lining according to Vermeer and de Borst
(1984). This leads to the use of a non-associated flow rule
(f 6 g).
According to Gehring (1996) and Ramoni and Anagnostou (2011), the skin friction coefficient l of the rock
mass on the shield is taken as l = 0.150.30 for the kinetic
coefficient during the ongoing excavation and l = 0.25
0.45 for the static coefficient used for restart after a
standstill. The coefficient of the rock mass on the lining is
usually taken as l = 0.300.40 for static conditions (ACI
318 2002). Therefore, for simplicity, the value of 0.3 may
be adopted for all the interfaces.
Since the numerical formulation used in this paper is
based on the small-strain/displacement assumption, the
shields have been modelled right on the tunnel boundary
even if there is a gap in between. Specifically, in order to
simulate the gap between the rock mass and the shield, a
special interface element has been used in which the elastic
stiffnesses are set to zero, i.e., Kn = Ks = Kt = 0. From
the previous equations, it is seen that no stress can transfer
from the rock mass elements to the shield elements. The
stiffness of the shield is unrelated to the deformations of
the rock masses. Thus, the gap is correctly simulated since
the presence of the shield has no confinement effect on the
tunnel boundary.
4.5.2 ShieldRock Mass Interaction
The accurate simulation of the interaction between the
shield and the rock mass has to take into account both (1)
the geometry update in order to consider the deformation
ahead of the face and (2) the gap due to the overexcavation
(sum of the overcut and the overboring) and the conicity of
the machine, as the closure of the gap determines the
amount of unloading of the tunnel boundary and allows the
following contact and slippage between the rock mass and
the shield.
In the double shield universal TBMs, the gap between
the shields and the rock mass assumes a finite value and,
furthermore, as shown in Figs. 12 and 13, its width Dg is
not uniform in the cross section and has a stepwise increase
due to the conicity Dr, as follows:

Dg

DD
DD 2Dr

for the front shield


for the rear shield

where DD is the overexcavation as shown in Fig. 13.

17

Simulation of Mechanized Tunnel Excavation

487

In the case of hard rock, the closure of the entire gap


should not occur. The convergences are small in the elastic
zones and localized failure zones due to spalling; dilation is
followed by the fallout of rock slabs and the formation of a
notch (Read 1994). In these latter zones, it is assumed that the
gap is not closed because (1) during the re-gripping operation, the front shield sweeps and displaces the material
accumulated and (2) the crushed material cannot provide a
consistent support. Therefore, as illustrated in Fig. 12, only
the invert (i.e., the lower quarter of the tunnel circumference)
of the front shield is significantly in contact with the rock
mass, as the gap is really very small, and is to be modelled.
In the case of weak rock, a large convergence of the
bored profile takes place. The closure of the entire gap may
occur either along the front shield or along the rear shield
or along both of them. Associated with the step-by-step
simulation method used, the convergences can be easily
obtained by monitoring the longitudinal displacement
profiles (LDP) at the crown and at the invert after each
step. When the convergence is such that the gap is closed,
the shields start to support the excavation walls. Due to the
in situ state of stress and the large convergence which will
take place, it is necessary to consider the possible uplift of
the machine itself which may lead to a reduction of the free
gap at the crown.
For the front shield, the entire gap thus closes where the
sum of the radial displacement at the invert uinvert and at the
crown ucrown exceeds the gap size Dg as in the cross section
A shown in Fig. 13:
uinvert A ucrown A  DD

18

Before the closure of the gap, only the invert of the front
shield is in contact with the rock mass as in the case of hard
rock, while the gap is simulated by the special interface
elements described in Sect. 4.5.1. Then, when the gap is
closed, the entire shield starts to support the rock mass. This

is carried out in the relevant step just by modifying the


properties of the interface element of interest.
When the entire front shield is in contact with the rock
mass, the gap around the rear shield is uniform and only
due to the conicity. Because of the non-axisymmetric
geometry, the closures of the gap at the invert and at the
crown, however, occur in different cross sections such as C
and D shown in Fig. 13. It closes at the invert where the
relative radial displacement between the cross section C
and B exceeds the conicity Dr:
uinvert C  uinvert B  Dr

19

and at the crown where the relative radial displacement


between the cross section D and B exceeds the conicity Dr:
ucrown D  ucrown B  Dr:

20

If the front shield is in contact with the rock mass only at


the invert, the contact point C at the invert of the rear shield
should satisfy the same condition as before:
uinvert C  uinvert B  Dr

21

while for the point D at the crown the following should


hold true:
ucrown D uinvert D  DD 2Dr

22

As a slice of the rock core ahead of the face is excavated,


the displacement uf at the face is removed. To consider the
geometry update, the final radial displacement ufin of
Eqs. 1822 is ufin = u - uf, where u is the total radial
displacement. In order to find the exact positions where the
shields get into contact with the rock mass (which is
actually the nearest point in the 1 m discretisation), one is to
note that, at the first excavation steps, the presence of the
shield improves the behaviour of the ground by a
longitudinal arch effect. Hence, the LDPs have to be
checked at every step to find out where the shields get in

Fig. 12 Cross section of a DSU


TBM at the front shield and the
rear shield and illustration of
rock mass-shield interaction for
hard rock

123

488

K. Zhao et al.

Fig. 13 DSU TBM


longitudinal section and
illustration of the rock massshield interaction in weak rock

contact with the rock mass, as the shape of the LDPs


changes from the previous step. However, after a few steps
(generally, 3 steps), such span reaches a constant value.
4.5.3 CutterheadRock Mass Interaction
The cutterhead of the machine is a lattice steel plate and is
separated from the shield just to be allowed to rotate. The
cutterhead applies to the face the excavation force only
through the cutters, while the rest of the structure is kept
separated from the face, except in case of failure of the
rock at the face, leading to the formation of blocks or huge
core extrusion.
The real situation and breakage mechanism at the face,
associated with the movement of the cutterhead, are not
faithfully reproduced, as a realistic representation would be
almost impossible with the adopted finite element model and
step-by-step method. The cutterhead is modelled as a steel
plate of 3 cm thickness and is assumed for simplicity to be
always in contact with the face. The boring force at the face
applied through the cutters is represented in an equivalent
way by a distributed pressure at the face, given by the
maximum cutterhead thrust divided by the cross-sectional
area. In the longitudinal direction, the cutterhead is separated
from the shield. A void of 1 m (due to the mesh discretization) has been left between the cutterhead and the shield
elements. The torque has been disregarded in the model, as
also done by Nagel and Meschke (2011) in their simulations.
4.5.4 LiningRock Mass Interaction (Backfilling)
The load transfer between the rock mass and the lining
occurs through a backfilling layer, as described in detail by
Ramoni et al. (2011). In this paper, this interaction is
considered and simulated as follows.
When tunnelling in hard rock, the gap is filled, at the
invert, with dry mortar in order to support the ring. Along

123

the remaining surface pea gravel and later cement grout are
injected. It is important, in the case of spalling, to correctly
fill this gap which could even be bigger due to the stressinduced notches. The elastic modulus of the pea gravel
with mortar injection is taken to be equal to 1 GPa. The
backfilling is usually performed at a certain distance behind
the shield. Therefore, as shown in Fig. 14a, the backfilling
is modelled in only one stage. The pea gravel and the lining
are activated 2 m behind the shield.
When tunnelling through weak rock, a grout annulus is
injected via the shield tail with a very fast hardening mortar
and simultaneously with the shield advance. In this case, as
shown in Fig. 14b, the tail gap grouting is modelled in two
phases as follows:
1.

2.

in the softening phase, right behind the shield, the


grouting is simulated by activating solid elements with
a low modulus (0.5 GPa) and by the application of a
grouting pressure to the excavation boundary and to
the lining;
in the hardening phase, 2 m behind the shield (a very
fast hardening grout is considered), a higher modulus
(1 GPa) is activated in the grouting elements and the
grouting pressure is deactivated.

As the gap in squeezing rocks is intended to leave more


free space for ground deformation, the nominal thickness
of the gap grouting is not respected and the actual thickness
during construction depends on the LDP at the shield tail.
With reference to the shield uplift described in Sect. 4.5.2,
a uniform thickness of the gap grouting is assumed.

5 Applications
In order to illustrate in detail the 3D simulation model of
TBM excavation in brittle and squeezing rock conditions,
respectively, the case studies of the Brenner and the Lyon

Simulation of Mechanized Tunnel Excavation

489

Fig. 14 Backfilling of the


segmental lining: a with pea
gravel and b with annulus
grouting via the shield tail

Turin Base Tunnels are considered. Only one of the two


tubes of these main tunnels will be simulated.

Table 2 Brenner Base Tunnel: rock mass parameters

5.1 Brenner Base Tunnel

5.1.1 Rock Mass


A section of the Brenner Base Tunnel starting from the
Italian Portal in Fortezza is considered. The tunnel is to be
excavated in the Brixner granite complex, which is part
of the Southern Alps. This granite is a large homogeneous
Permian intrusion that underwent only brittle deformations
during the Alpine orogeny. Granite is medium to coarse
grained. The maximum overburden of the tunnel considered is 1,350 m.
The assumed rock mass parameters are shown in
Table 2 and are based on the data which became available
during the excavation of the Aica Exploratory Tunnel
(Barla et al. 2010), lying 10 m below the main tunnel.
Figure 15 illustrates the spalling model adopted (in 2D).
An associated flow rule has been used, which means that
mdil = mpeak, so that dilation is slightly lower than the
maximum value, given by mdil = mres/4; as with this criterion mres [ mpeak (in the equivalent CSFH model, the
maximum value for friction is the residual value).
In order to demonstrate the capability of the 3D model
to capture the spalling behaviour, the state of stress in the
rock mass has been assumed to be characterized by a stress
ratio rh (horizontal stress)/rv (vertical stress) equal to 2,
with rv = 24 MPa. It is noted that no data are available in
the tunnel to confirm such assumption.

Rock mass
25 GPa

0.25

UCS

130 MPa

8.1 MPa

CI

58.5 MPa

Generalised HoekBrown
apeak

0.25

speak

0.041

mpeak

0.656

ares
mres

0.75
9

sres

10-6

Modified parameters (midas GTS)


apeak

0.5

mpeak

1.4

sres

0.2

ares

0.5

mres

sres

10-6

5.1.2 TBM Parameters


The TBM simulation parameters are given in Table 3. The
thickness of the shield has been set as 3 cm, which is the
actual thickness of the steel elements composing the outer
part of the machine. An equivalent density value is

123

490

K. Zhao et al.
350

6 Results

300

Figure 16 shows the 3D spalling failure and overbreak


prediction along the tunnel axis in intrinsic conditions (i.e.,
model with only the rock mass, without machine components), while Fig. 17 gives such a prediction with the
complete model (as described in Sect. 4). The degree of
spalling failure and overbreak is expressed through the
equivalent plastic strain, defined as:
r

2 2
ep1 e2p2 e2p3
epl
23
3

250

Intact rock
Brittle peak modified

1 [MPa]

Brittle residual modified

200

brittle peak Diederichs


brittle resid Diederichs

150

100

50

-10.0

0.0

10.0

20.0

30.0

40.0

3 [MPa]

Fig. 15 Brenner Base Tunnel, rock mass behaviour: Diederichs


model (based on the generalised HoekBrown criterion) and modified
model (based on the traditional HoekBrown criterion)

assigned to the front shield invert elements, obtained by


dividing the total weight of the machine by the apparent
volume of the shield. Also, the unit weight of the cutterhead has been increased with respect to the value for steel
in order to take account for its weight.
The lining concrete density value has been increased
(from the real value of concrete c = 25 kN/m3 to an
equivalent value c = 30 kN/m3), in order to consider the
back-up-trailer (Lambrughi 2008). An appropriate unit
weight has been assigned for the other components.

with ep1 , ep2 , ep3 the principal plastic strains.


This definition comes from the most commonly used
expression for the softening/hardening parameter (based on
incremental plastic strain) which is (Vermeer and De Borst
1984):
r

Dep
2
p
e

e_ p1 e_ p1 e_ p2 e_ p2 e_ p3 e_ p3
24
3
Dt
It is noted that this softening parameter has not been
introduced in the constitutive law; the equivalent plastic
strain is thus only a variable describing the degree of
damage in the rock mass.
The spalling zones are shown to occur at the invert and at
the crown with a maximum depth of approximately 1 m (the
uncertainty in the assessment of the depth of failure prediction is half of the element size, i.e., 0.15 m). The v-shape is
less pronounced than in other cases, as the stress ratio is
rather small. In intrinsic conditions, failure zones start right
behind the face. Any failure zone occurs at the face.
Also, failure occurs at the sidewalls. This is a shear
failure (i.e., in the high confinement range) which develops

Table 3 Brenner Base Tunnel: TBM components parameters


Cutterhead

Shield

Lining

Grippers

Pea gravel

Diameter

9.3 m

Diameter

9.3 m

Diameter

9m

Base

2m

Maximum
thickness

18.5 cm

Thickness

3 cm

Thickness

3 cm

Thickness

45 cm

Height

6.6 m

Minimum
thickness

11.5 cm

1 GPa

Length front shield

5m

Thickness

5 cm

Length rear shield

6m

Regripping

2m

200 GPa

200 GPa

30 GPa

200 GPa

0.3

0.3

0.2

0.3

0.3

Density

312 kN/m3

Density front
shield elements

14,128 kN/m3

Density

30 kN/m3

Density

78 kN/m3

Density

24 kN/m3

Pressure at
the face

0.25 MPa

Pressure

4 MPa

Friction coefficient
rock-shield skin

123

0.3

Friction coefficient
backfilling-lining

0.3

Simulation of Mechanized Tunnel Excavation

491

Fig. 16 Brenner Base Tunnel:


plastic strain contours along the
tunnel in intrinsic conditions:
a upper half and b lower half

at the face edges, it is enhanced by the short excavation


length and then propagates (as the failure is unrecoverable)
all along the excavation. It is related to stress paths above
the damage threshold but below the spalling limit (Diederichs 2007) and for this reason the plastic strain values are
lower than the values in the spalling zones. This type of
failure is a typically 3D effect and occurs as the TBM
excavation is continuous.
Figure 18a depicts the overbreak zones in a cross section in steady-state conditions when no interaction of the
TBM with the rock mass is present, while Fig. 18b illustrates the model when such interaction is activated. Figure 19 shows such overbreak with the complete model in a
cross section at the front shield. It is seen that the front
shield and its self-weight produce a confinement at the
invert, so that the plastic strain at the front shield invert is
reduced by 60%. This is consistent with the in situ observations, as for example observed during excavation of the
Niagara Tunnel in Canada (courtesy of Professor Diederichs 2010) or at the Canadian Underground Research
Laboratory (Read 1994).

The grippers are shown to have very limited influence


whereas the support produces only a slight decrease of the
failure extent, as it is applied at a certain distance from
the face, where most of failure did already occur. As soon
as the front shield is passed and the invert walls are
exposed, the plastic strain increases, reaching, however, a
value in steady-state conditions which is lower than in the
intrinsic case (reduction of 14%). This means that the
crushing and subsequent fallout of rock at the crown and
at the invert will take place along the rear shield. It is
expected that the fallout will not damage the shield
components.
6.1 LyonTurin Base Tunnel
6.1.1 Rock Mass
Consideration is given to the LyonTurin Base Tunnel in
France, in a cross section where the excavation is to take
place in the productive Carboniferous Formation, Zone
Houille`re Brianconnaise-Unite des Encombres, which is

123

492

K. Zhao et al.

Fig. 17 Brenner Base Tunnel:


plastic strain contours along the
tunnel in the complete model:
a upper half and b lower half

composed of black schists (4555%), sandstones


(4050%), coal (5%), clay-like shales and cataclastic rocks.
The assumed rock mass parameters are shown in Table 4
and are based on back analysis and performance monitoring studies during excavation of the Saint Martin La Porte
access adit and when crossing this rock mass (Barla 2009).
The in situ state of stress is assumed to be isotropic and
equal to 26 MPa.
6.1.2 TBM Parameters
The value of the jack pressure is equal to the maximum
thrust provided by the auxiliary thrust system divided by
the circumference of the lining ring. As the auxiliary thrust
force required for excavation in such conditions is evaluated only through the results of the model, an estimated
value has been assumed. The other parameters have been
already described in Sect. 5.1.2. The simulation parameters
are given in Table 5.

123

7 Results
Figures 20 and 21 show the plastic strain contours and
compare the results for intrinsic conditions (a) and for the
complete model (b) when the interaction between the TBM
and ground is considered (again in terms of the equivalent
plastic strain given by Eq. 23 above). The plastic zones are
shown to be reduced in extent but are still all around the
excavation (the plastic radius is reduced from 12.9 m in
intrinsic conditions to 10 m for the complete model). It is
also noted that in the complete model the plastic strain at
the invert is smaller than that at the roof (there is a
reduction of 29% at the front shield invert). It is clear that
the presence of the shield invert (in this case, both front and
rear shield) before the gap is closed, associated with the
self-weight of the machine, provides an additional confinement near the tunnel face.
Figure 22 depicts the results in terms of LDP (a) and
contact pressure (b) on both the shields and the lining. The

Simulation of Mechanized Tunnel Excavation

493

Fig. 19 Brenner Base Tunnel: plastic strain contours in a cross


section at the front shield (complete model)

Table 4 LyonTurin Base Tunnel: rock mass parameters


Rock mass

Fig. 18 Brenner Base Tunnel: plastic strain contours in steady-state


condition: a intrinsic conditions and b complete model

closure of the entire gap between the front shield and the
ground is shown to occur 4 m behind the face. For the rear
shield this occurs at a distance of 9 m at the invert and
10 m at the crown, where complete closure takes place.
Given the value of the pressure acting on the shield, one is
in position to carry out the structural design of the shield
(and therefore of the machine). It is of interest to note that
the shape of the LDP and the contact pressure distribution
are very similar to those given by Ramoni and Anagnostou
(2011). Also to be noted is the greater pressure computed at
the shield invert with respect to the crown value, as shown
by Graziani et al. (2007).
When the grout reaches the hardening phase, the stresses
due to the excavation advancement are gradually

2 GPa

0.25

2 MPa

24

4

transferred to the lining (i.e., the radial pressure on the


lining increases), while the rock mass does not experience
further plasticity (so-called past-yield zone, as described
by Ramoni and Anagnostou 2011). The so-called elastic
re-compression on the tunnel boundary occurs (Garber
2003) and finally, the steady-state condition is reached.
Figure 23 shows the history of the radial stress along the
tunnel boundary. It is noted that the rock mass experiences
three unloading processes during the excavation. Firstly, as
the tunnel walls remain unsupported, the tunnel boundary
experiences an unloading process. At the distance of 1 m
behind the face, the invert is confined by the shield and, as
soon as the entire gap is closed, loading of the shield takes
place. Secondly, the entire tunnel boundary is unloaded at
the rear shield due to the conicity of the machine. Finally,
as the annulus is injected via the tail shield, the last
unloading process occurs.
By integrating the ground pressure pi (taken by the
normal stress in the interface elements) over the shield
surface (N, number of elements in the surface) and by
multiplying the integral by the skin friction coefficient l
and the reduction coefficient b which is the ratio b r=R of
the real shield radius r over the tunnel radius R (as the

123

494

K. Zhao et al.

Table 5 LyonTurin Base Tunnel: TBM components parameters


Cutterhead

Shield

Lining

Grouting
softening phase

Diameter

9.44 m

Diameter

9.44 m

Diameter

9m

Thickness

3 cm

Thickness

3 cm

Thickness

45 cm

Length front shield

5m

Length rear shield

6m

Grouting
hardening phase

200 GPa

200 GPa

36 GPa

0.5 GPa

1 GPa

0.3

0.3

0.2

0.3

0.3

Pressure at
the face

0.25 MPa

Pressure due to the


auxiliary thrust force

0.875 MN/
m

Pressure
(rock
and lining)

0.2 MPa

Density

312 kN/m3

Density

24 kN/m3

Density

24 kN/m3

Density shield
elements

2,757 kN/m3

Density

30 kN/m3

Friction coefficient
rock-shield skin

0.3

Friction coefficient
grouting-lining

0.3

Fig. 21 LyonTurin Base Tunnel: plastic strain contours in a cross


section in steady-state condition: a intrinsic conditions and b complete
model

shield is modelled on the tunnel boundary and therefore


considering the overexcavation and the conicity, it is bigger than the real shield), the thrust force required to
overcome friction has been calculated:
Ff b  l
Fig. 20 LyonTurin Base Tunnel: plastic strain contours along the
excavation (longitudinal view): a intrinsic conditions and b complete
model

123

N
X

pi Ai 0:985 0:3 186 54:9 MN

25

i1

The maximum total thrust force by the auxiliary thrust


cylinders is the sum of the maximum cutterhead thrust FN

Simulation of Mechanized Tunnel Excavation

495

0
-10

-5

(a)

10

15

20

25
y [m]

-50

u [mm]

-100

crown
invert

-150

-200

-250

Tr

3.5

(b)

correspondence of the front shield and the required thrust


force would be significantly lowered.
Also the torque and the consequent cutterhead installed
power can be calculated. The total torque is the sum of
maximum torque due to the rolling forces of the cutters and
the torque needed to overcome the frictional resistance
(sliding friction) caused by the ground pressure acting
axially and radially upon the cutter head. The maximum
rolling force FR of a single cutter can be estimated simply
equal to 0.1 FN, according to the average cutting coefficient
Cc (FN/FR) obtained from tests and machine parameters
(Balci et al. 2009; Rostami 2008; Farrokh and Rostami
2009; Yagiz 2006). The maximum torque due to the rolling
forces is (Balci et al. 2009):
n  FR  D
4;675 kNm
4

26

where n is the number of cutters (69).


The torque needed to overcome the frictional resistance
has the following expression:

p [MPa]

2.5
2
1.5

Lining

Tf l 

0.5

ZR

2
p  2pr dr  r l  p  p  R3 16;826 KNm
3

y [m]

0
-10

-5

10

15

20

27

25

Fig. 22 LyonTurin Base Tunnel: results for the complete model,


crown and invert: a longitudinal displacement profile and b contact
pressure on the shield and on the lining

where p is the cutterhead pressure.


The total torque is thus equal to Ttot = Tr ? Tf =
21.6 MNm.
The relationship between the torque and the power P is
the following (Balci et al. 2009):
P 2p

Radial stress [MPa]

25

-5

28

where RPM is the rotational speed of the cutterhead. The


power considered in Table 1 is thus sufficient for the TBM
advancement, even if the cutterhead rotational speed has to
be reduced to 2.1 rpm.

30

-10

RPM
T
60

20
15
10

Crown
Invert

8 Conclusions

y [m]

0
0

10

15

20

25

Fig. 23 LyonTurin Base Tunnel: history of the radial stress, invert


and crown along the tunnel boundary for the complete model

and the thrust to overcome friction as follows: F = FN ?


Ff = 17 ? 54.9 = 71.9 MN. In order to reduce this quite
high value of the auxiliary thrust force, it is possible to
increase the gap corresponding to the rear shield: the
stepwise increase between the rear shield and the front
shield can pass from 3 cm to 56 cm, so that the contact
between the rock mass and the shields would occur only in

A review of the current state-of-the-art in modelling (by 3D


or axisymmetric models) the TBM excavation process is
made with special interest in long deep tunnels (as the
Alpine Base Tunnels) excavated by double shield TBMs. It
is seen that, although the use of rock TBMs is widespread,
very few 3D models of deep mechanized tunnel excavation
in rock masses are available in the literature. In this paper,
the 3D simulation of the complex interaction between the
rock mass, the tunnel machine, its system components, and
the tunnel support is studied in detail by presenting a novel
and more general simulator of mechanised tunnel excavation at the design stage.

123

496

Consideration is given to the 3D simulation of


mechanised excavation along the Brenner and the Lyon
Turin Base Tunnels which at present are at the design stage
and where access adits/exploratory tunnels have been
already excavated. A double shield universal machine (and
the support system) has been chosen in order to cope with
the hard rock conditions (in granite) where spalling is
expected to occur (Brenner Base Tunnel) and with the
weak rock conditions (in the Carboniferous formation)
in a tunnel length where moderate squeezing rock conditions are anticipated (LyonTurin Base Tunnel).
The results obtained with the numerical simulations
show that the newly developed 3D model is highly effective in reproducing both the rock mass response and its
interaction with the TBM system components. The 3D
nature of TBM excavation has been taken into account
using a non-axisymmetric model for the state of stress in
the rock mass, the geometry of the TBM, and the support
system. Time-dependence modelling as typical in severe
squeezing conditions has not yet been considered, although
work for introducing such a behaviour in the 3D model is
underway.

References
ACI 318 (2002) Building code requirements for reinforced concrete.
American Concrete Institute, Detroit
Anagnostou G, Cantieni L, Nicola A, Ramoni M (2010) Lake Mead
Intake No 3 Tunnelgeotechnical aspects of TBM operation. In:
Proceedings of North America Tunnelling Conference 2010,
Portland
Asche H, Ireland T, Newton C, Watts C (2011) Preliminary design for
the North Bank Hydro Project, South Island, New Zealand. In:
Proceedings of RETC 2011, San Francisco
Balci C, Tumac D, Copur H, Bilgin N, Yazgan S, Demir E, Aslantas
G (2009) Performance prediction and comparison with in situ
values of a TBM: a case study of Otogar-Bagcilar metro tunnel
in Istanbul. In: Proceedings of ITA-AITES World Tunnel
Congress 2009, Budapest
Barla G (2001) Tunnelling under squeezing rock conditions. In:
Kolymbas D (ed) Tunnelling Mechanics, Eurosummer school.
Logos Verlag, Innsbruck, pp 169268
Barla G (2009) Innovative tunneling construction method to cope
with squeezing at Saint Martin La Porte access adit (LyonTurin
Base Tunnel). In: Proceedings of EUROCK 2009, Dubrovnik
Barla G (2010) Analysis of an extraordinary event of TBM
entrapment in squeezing ground conditions. In: Pilgerstorfer T
(ed) No friction no tunnelling. Fetschift zum 60. Gebustag von
Wulf Schubert, Institut fur Felsmechanik und Tunnelbau,
Technische Universitat Graz, pp 6676
Barla G, Ceriani S, Fasanella M, Lombardi A, Malucelli G, Martinotti
G, Oliva F, Perello P, Pizzarotti E M, Polazzo F, Rabagliati U,
Skuk S, Zurlo R (2010) Problemi di stabilita` al fronte durante lo
scavo del cunicolo esplorativo Aica-Mules della Galleria di Base
del Brennero. MIR 2010, Torino
Barla G, Janutolo M, Zhao K (2011) Open issues in tunnel boring
machine excavation of deep tunnels. Keynote Lecture. 14th
Australasian Tunnelling Conference. Auckland

123

K. Zhao et al.
Cai F, Ugai K (2000) Numerical analysis of the stability of a slope
reinforced with piles. Soils Found 40(1):7384
Cantieni L, Anagnostou G (2009) The effect of the stress path on
squeezing behaviour in tunnelling. Rock Mech Rock Eng
42(2):289318
Castellanza R, Betti D, Lambrughi A (2008) Three-dimensional
numerical models for mechanised excavations in urban areas.
Jornada Tecnica: Tuneles con EPB. Simulacion y control de la
tuneladora, Barcelona
Chen WF (1994) Constitutive equations for engineering materials.
Plasticity and modelling, vol 2. Elsevier, Amsterdam
Cobreros JA, Cervera MM, Herrera M, Conde C (2005) Tunnels de
GuadarramaProble`mes geotechniques rencontres dans un
double tunnel de 27 km. Tunnels et ouvrages souterrains
190Supplement Juillet/Aout 2005, 512, AFTES Paris
Day RA, Potts DM (1994) Zero thickness interface elements
numerical stability and application. Int J Numer Anal Method
Geomech 18:689708
Debernardi D, Barla G (2009) New viscoplastic model for design
analysis of tunnels in squeezing conditions. Rock Mech Rock
Eng 42:259288
Diederichs MS (2007) The 2003 Canadian Geotechnical Colloquium:
mechanistic interpretation and practical application of damage
and spalling prediction criteria for deep tunneling. Can Geotech J
44:10821116
Diederichs M (2010) Brittle spalling, practical limits. Keynote paper,
BEFO 2010, Swedish Rock Mechanics Symposium, Sweden
Drucker DC (1954) Coulomb friction, plasticity and limit loads.
J Appl Mech 21:7174
Eberhardt E (2001) Numerical modelling of three-dimension stress
rotation ahead of an advancing tunnel face. Int J Rock Mech Min
Sci 38:499518
EN 1992-1-1 (2004) Eurocode 2design of concrete structures
Fakharian K, Evgin E (2000) Elasto-plastic modeling of stress-pathdependent behaviour of interfaces. Int J Numer Anal Methods
Geomech 24:183199
Farrokh E, Rostami J (2009) Effect of adverse geological condition on
TBM operation in Ghomroud tunnel conveyance project. Tunn
Undergr Space Technol 24:436446
Garber R (2003) Design of deep galleries in low permeable saturated
porous media. The`se No 2721, EPFL Lausanne
Gehring KH (1996) Design criteria for TBMs with respect to real rock
pressure. In: Tunnel boring machinestrends in design &
construction of mechanized tunnelling, International lecture
series TBM tunnelling trends, Hagenberg. A.A.Balkema, Rotterdam, pp 4353
Graziani A, Ribacchi R, Capata A (2007) 3D modelling of TBM
excavation in squeezing rock masses. Brenner Basistunnel und
Zulaufstrecken, Internationales Symposium BBT 2007, Innsbruck,
Austria. Innsbruck University Press, Innsbruck, pp 143151
Gutter W, Romualdi P (2003) New design for a 10 m universal
double shield TBM for long railway tunnels in critical and
varying rock conditions. In: Proceedings of Rapid excavation &
Tunnelling Conference 2003, New Orleans
Hoek E, Guevara R (2009) Overcoming squeezing in the Yacambu`
Quibor tunnel, Venezuela. Rock Mech Rock Eng 42:389418
Itasca (2006) FLAC 3D Fast Lagrangian analysis of continua in 3D
dimensions, Users guide
Kasper T, Meschke G (2004) A 3D finite element simulation for TBM in
soft ground. Int J Numer Anal Methods Geomech 28:14411460
Kasper T, Meschke G (2006) On the influence of face pressure,
grouting pressure and TBM design in soft ground tunnelling.
Tunn Undergr Space Technol 21:160171
Lambrughi A (2008) A three-dimensional numerical model of
mechanised excavations in urban areas, Ph.D. thesis, Politecnico
di Milano, Italy

Simulation of Mechanized Tunnel Excavation


Loew S, Barla G, Diederichs M (2010) Engineering Geology of
Alpine tunnels: past, present and future. Keynote lecture, 11th
IAEG Congress, Auckland
Lombardi G (1973) Dimensioning of tunnel linings with regard to
constructional procedure. Tunn Tunn 5(4):340351
Maidl B, Schmid L, Ritz W, Herrenknecht M (2008) Hard rock tunnel
boring machines, Ernst & Sohn, A Wiley Company, Berlin
Nagel F, Meschke G (2011) Grout and bentonite flow around a TBM:
computational modelling and simulation-based assessment of
influence on surface settlements. Tunn Undergr Space Technol
26:445452
Nagel F, Stascheit J, Meschke G (2008) A numerical simulation
model for shield tunnelling with compressed air support.
Geomech Tunn 1(3):222228
Nguyen-Minh D, Corbetta F (1992) New methods for rock-support
analysis of tunnels in elasto-plastic media. In: Rock support in
Mining and Underground Construction, Kaiser & McCreath,
Balkema, Rotterdam
Ramoni M, Anagnostou G (2006) On the feasibility of TBM drives in
squeezing rock conditions. Tunn Undergr Space Technol
21(34):262
Ramoni M, Anagnostou G (2010) Tunnel boring machines under
squeezing conditions. Tunn Undergr Space Technol 25:139157
Ramoni M, Anagnostou G (2011) The interaction between shield,
ground and tunnel support in TBM tunnelling through squeezing
conditions. Rock Mech Rock Eng 44:3761
Ramoni M, Lavdas N, Anagnostou G (2011) Squeezing loading of
segmental linings and the effect of backfilling. Tunn Undergr
Space Techn 26(6):692717

497
Read RS (1994) Interpreting excavation-induced displacements around
a tunnel in highly stressed granite. Ph.D. thesis, Department of Civil
and Geological Engineering, University of Manitoba, Winnipeg
Rostami J (2008) Hard rock TBM cutterhead modelling for design and
performance prediction. Geomechanik und Tunnelbau, Berlin
Simic D (2005) Tunnels de GuadarramaProblematique des tunnels
profonds. La traversee de la faille de la Umbria du tunnel de
Guadarrama. Tunnels et ouvrages souterrains 190Supplement
Juillet/Aout 2005, 1321, AFTES Paris
Vermeer PA, De Borst R (1984) Non-associated plasticity for soils,
concrete and rock. Heron 29:164
Vlachopoulos N, Diederichs MS (2009) Improved longitudinal
displacements profiles for convergence confinement analysis of
deep tunnel. Rock Mech Rock Eng 42:131146
Vogelhuber M (2007) Der Einfluss des Porenwasserdrucks auf das
mechanische Verhalten kakiritisierter Gesteine. In: Veroffentlichungen des Instituts fur Geotechnik (IGT) der ETH Zurich,
Band 230, ETH Dissertation Nr. 17079, vdf Hochschulverlag
AG Zurich
Yagiz S (2006) A model for the prediction of tunnel boring machine
performance. In: Proceedings of 10th IAEG Congress, Nottingham
Zhao XG, Cai M (2010) Influence of plastic shear strain and
confinement-dependent rock dilation on rock failure and displacement near an excavation boundary. Int J Rock Mech Min
Sci 47:723738

123

S-ar putea să vă placă și