See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/271892705

Tunnel face stability in heavily fractured rock masses that follow the Hoek–
Brown failure criterion

Article  in  International Journal of Rock Mechanics and Mining Sciences · June 2013

DOI: 10.1016/j.ijrmms.2013.01.004

CITATIONS READS

110 387

3 authors, including:

Guilhem Mollon
Institut National des Sciences Appliquées de Lyon
68 PUBLICATIONS   2,576 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PhD : Insights on the behavior of fault gouges during seismic sliding: a numerical investigation from granular rheology to friction laws View project

All content following this page was uploaded by Guilhem Mollon on 07 September 2020.

The user has requested enhancement of the downloaded file.

 International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451

Contents lists available at SciVerse ScienceDirect

International Journal of
Rock Mechanics & Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Tunnel face stability in heavily fractured rock masses that follow

the Hoek–Brown failure criterion
Salvador Senent a, Guilhem Mollon b, Rafael Jimenez a,n
a
Technical University of Madrid, Spain
b
Hong Kong University of Science and Technology, Hong Kong

a r t i c l e i n f o a b s t r a c t

Article history: A tunnel face may collapse if the support pressure is lower than a limit value called the ‘critical’ or ‘collapse’
Received 16 April 2012 pressure. In this work, an advanced rotational failure mechanism is developed to compute, in the context of
Received in revised form limit analysis, the collapse pressure for tunnel faces in fractured rock masses characterized by the Hoek–
8 November 2012
Brown non-linear failure criterion. The non-linearity introduces the need for additional assumptions about
Accepted 2 January 2013
the distribution of normal stresses along the slip surface, which translate into new parameters in the limit
Available online 6 March 2013
analysis optimization problem. A numerical 3D finite difference code is employed to identify adequate
Keywords: approximations of the distribution of normal stresses along the failure surface, with results showing that
Tunnel face stability linear stress distributions along the failure surface are needed to obtain improved results in the case of
Critical pressure
weaker rock masses. Test-cases are employed to validate the new mechanism with the three-dimensional
Limit analysis
numerical model. Results show that critical pressures computed with limit analysis are very similar to
Collapse mechanism
Hoek–Brown those obtained with the numerical model, and that the failure mechanisms obtained in the limit analysis
Non-linearity approach are also very similar to those obtained in small scale model tests and with the numerical
simulations. The limit analysis approach based on the new failure mechanism is significantly more
computationally efficient than the 3D numerical approach, providing fast, yet accurate, estimates of critical
pressures for tunnel face stability in weak and fractured rock masses. The methodology has been further
employed to develop simple design charts that provide the face collapse pressure of tunnels within a wide
variety of practical situations.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction initial solution for cohesive materials, whereas Leca and Panet
[17] and Leca and Dormieux [18] proposed upper and lower
One of the most important issues when designing and con- bound solutions for cohesive-frictional materials. Recent research
structing a tunnel is to ensure the stability of the tunnel face, as has been focused on the development of improved failure geo-
tunnelling experience indicates that most tunnel collapses have metries (see e.g., Refs. [19,20]), leading to failure surfaces that are
their origin in stability problems at the face [1]. Much research developed point-to-point and that allow rotational failure modes
has been conducted to develop methods to assess the ‘critical’ or that affect the whole excavation front [21,22].
‘collapse’ pressure, or, in other words, the minimum pressure that These latter mechanisms have been shown to provide good results
needs to be applied at the tunnel face to avoid its instability. when compared to limit equilibrium and numerical solutions. In
A wide variety of collapse mechanisms have been proposed previous works, geotechnical failure has been modelled using the
within the framework of limit equilibrium analyses, such as linear Mohr–Coulomb (MC) criterion that is traditionally applied to
Horn’s model [2] and its later variations (see e.g., Refs. [3,4]) or soils. The actual strength of rock masses, however, is well known to
those proposed by Vermeer et al. [5] and Melis [6]. Model tests at be a non-linear function of stress level. Although some methodologies
the laboratory and centrifuge tests have also been conducted to have been recently proposed to compute ‘equivalent’ MC parameters
study this problem (see e.g., Refs. [7–10]) and numerical models that depend on the stress level for 2D tunnelling analyses (see e.g.,
such as the finite element method (FEM) [11–13] and the discrete Refs. [23,24]) or, using limit analysis, to study the stability of 2D
element method (DEM) [14,15] have been employed as well. tunnel sections excavated in materials with a non-linear failure
Another methodology to study the stability of the tunnel face criterion (see e.g., Refs. [25,26]), only a few limited attempts have
is limit analysis. In this context, Davis et al. [16] proposed an been made to consider the non-linearity of actual failure criteria
when computing the critical pressure of the tunnel face.
One approach to consider non-linear failure criteria in the limit
n analysis literature has been to employ a linear failure envelope that
Correspondence to: ETSI Caminos, C. y P., C/ Profesor Aranguren s/n, Madrid
28040, Spain. is tangent to the original, and non-linear, failure criterion [27], since
E-mail address: rjimenez@caminos.upm.es (R. Jimenez). upper-bound solutions obtained using such envelope are also

1365-1609/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijrmms.2013.01.004
 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451 441

upper-bound solutions to the original problem with the non-linear block around an axis (Ox) (x being the horizontal direction perpen-
criterion [28]. For instance, the ‘‘generalized tangential technique’’ dicular to the tunnel axis, see Fig. 1a), and (ii) the collapsing block
[29,30] is based on considering a linear envelope that is tangent at intersects the whole circular surface of the tunnel face. These
an ‘optimum’ point. Huang and Yang [31] have used this technique kinematic assumptions were made after observations of numerical
to study the stability of the tunnel face employing the passive ‘blow- [12] and experimental [8] simulations of face collapses. Besides the
out’ mechanism of Leca and Dormieux [18] and a non-linear kinematic aspects, the normality condition related to the assump-
strength criterion previously proposed by Agar et al. [32]. Their tions of the kinematical upper-bound theorem of limit analysis
results illustrate the influence of the degree of non-linearity of the has to be fulfilled. In the case of a frictional soil, with or without
criterion on the collapse pressure, although they are not compared cohesion, this condition states that the normal vector pointing
with other numerical or analytical methods. outward of the slip surface should make an angle p/2þ j with the
In this work, we study the face stability of circular tunnels discontinuity velocity vector in any point of the discontinuity
excavated in heavily fractured and ‘low quality’ rock masses that surface, with j being the internal friction angle of the material.
follow the non-linear Hoek–Brown (HB) failure criterion. To that Even in a homogeneous MC soil, there is no simple surface which is
end, the rotational mechanism recently proposed by Mollon et al. able to satisfy both the kinematic and the normality conditions. It
[22] has been generalized to consider the non-linearity of the HB was thus necessary to use a complex discretization scheme to
failure criterion, so that we can employ ‘instantaneous’ values of generate the external surface of the moving block from close to
cohesion and of friction angle that depend on the stress level, close, using a collection of triangular facets respecting locally the
at the same time that we fulfill the assumption of associated flow. normality condition. The method and equations used for this
To validate the new methodology, we compare our limit analysis generation are described in detail in Ref. [22] and will be briefly
results with the results of a 3D finite difference numerical code. recalled here to make understandable the changes needed to adapt
Finally, we employ the new improved failure mechanism to this formulation to the HB criterion, as described in Section 2.2.
develop design charts that can be used to estimate the tunnel Two levels of discretization are needed, as shown in Fig. 1.
face critical pressure, for a wide variety of practical cases, and as a The first level consists in discretizing the circular tunnel face in
function of the tunnel diameter and of the rock mass parameters. ny points Aj and A0 j (for j¼1 to ny/2, see Fig. 1b). The second level
of discretization consists in defining a number of radial planes called
Pj, which all meet at the centre O of the rotational motion and
2. An improved rotational tunnel face failure mechanism are thus all normal to the velocity field. As shown in Fig. 1c,
for non-linear materials the mechanism may be divided in two sections. The so-called
Section 1 corresponds to the planes Pj with j¼1 to ny/2, so that
2.1. Principles of the collapse mechanism of Mollon et al. each of these planes Pj contains the points O, Aj, and A0 j,
whereas the so-called Section 2 corresponds to the planes Pj with
The analytical collapse mechanism developed by Mollon et al. j4ny/2. In Section 2, the planes are generated with a constant
[22], in the framework of the kinematical theorem of limit user-defined angular step db until Point F, which is the extremity
analysis applied to MC soils, relies on two main assumptions: of the mechanism and which is so far unspecified. Thus, the
(i) the collapse involves the rotational motion of a single rigid surface generation process only depends on two discretization

Fig. 1. Discretizations used for the generation of the collapsing block. (a) Reference system; (b) cross section with points at the tunnel boundary and (c) longitudinal
section along tunnel axis of the failure mechanism.
442 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451

turn provides the positions of the points Cj. As shown in Figs. 1c

and 2, these points are located in each plane Pj and are needed
before starting the generation process. The computation of the
coordinates of F is thus crucial. In the case of a homogeneous MC
soil, this computation is straightforward since F is the intersection
of two logarithmic spirals of parameter tan j emerging from the
foot and the crown of the tunnel. These two curves are indeed the
exact analytical boundaries of the mechanism projected in the
vertical plane of symmetry of the tunnel.
The use of a non-linear failure criterion, however, may lead to
a spatial variability of the friction angle and of the cohesion that
need to be considered for generation of the mechanism. This is
due to the non-linearity of the failure criterion, since the ‘equiva-
lent’ c and j employed in the analysis are actually functions of
Fig. 2. Principle of the point-by-point generation of the lateral surface of the the normal stress on the slip surface at failure, and such normal
mechanism. stress might not have a constant value on the entire slip surface.
This observation has a crucial effect on the procedure for genera-
parameters: ny and db. It is based on an iterative scheme which is tion of the mechanism presented in Ref. [22] and shortly
described in Fig. 2. As shown in this figure, if two points Pi,j and described above, as such procedure needs to be modified to allow
Pi þ 1,j, belonging to Plane Pj, are known, it is possible to define a for non-constant friction angles along the slip surface. That can be
third point Pi,j þ 1 belonging to Plane Pj þ 1, in such a way that the achieved with the following modifications to the procedure for
triangular facet (Pi,j, Pi þ 1,j, Pi,j þ 1) respects the normality condition; generation of the mechanism:
i.e., its normal vector makes an angle p/2þ j with the velocity
vector. This triangular facet is called Fi,j. Moreover, if a similar
operation is performed for points Pi þ 1,j and Pi þ 2,j, it is possible to – The position of F is more complicated to determine since the
compute the position of a point Pi þ 1,j þ 1 and, in turn, to define the spatial variability of j prevents from using the method based
‘reversed’ facet (Pi þ 1,j, Pi,j þ 1, Pi þ 1,j þ 1), which is called F0 i,j (Fig. 2). on the intersection of two logarithmic spirals described above.
This process, described in detail in Ref. [22], starts from the points In the modified mechanism, the position of F is obtained using
belonging to the tunnel face and stops at the extremity of the a process based on the one proposed in Ref. [33] in a 2D
mechanism. At the end of the generation, the lateral surface of the version of this mechanism: instead of defining the limit curves
collapsing block is thus defined by a collection of triangular facets Fi,j of the mechanism projected in the plane of symmetry of the
and F0 i,j, which makes it possible to compute its volume, weight, etc. tunnel by two logarithmic spirals, we define these curves as a
When the entire external surface of the moving block has been succession of segments which locally respect the normality
defined, the critical collapse pressure can be computed by applying condition, as shown in Fig. 3. More precisely, the curve
the main equation of the kinematic theorem of limit analysis,
which states that the rate of energy dissipation in the system is
equal to the rate of work applied to the system by the external
forces. In the present case, assuming that the mechanism does not
outcrop at the ground surface, the forces applied to the moving
block of ground at collapse are its own weight and the ‘collapse’
face pressure sc. The only possible energy dissipation in the system
arises from the slip surface and is proportional to the cohesion c.
After applying the work equation and performing a few
simplifications, the collapse pressure is given by Eq. (1), where
Sj, Rj, bj, Ri,j, Vi,j, bi,j, Si,j, R0 i,j, V0 i,j, b0 i,j, and S0 i,j are geometrical terms
defined in Ref. [22].
P P
g ðRi,j V i,j sin bi,j þR0i,j V 0i,j sin b0i,j Þc cos j ðRi,j Si,j þ R0i,j S0i,j Þ
i,j i,j
sc ¼ P
ðSj Rj cos bj Þ
j

ð1Þ
This expression, however, is only valid for the velocity field
obtained with a given position of the centre of rotation O. The
best value of sc that the method can provide is then obtained by
maximization of this expression with respect to the two geo-
metric parameters that define this point O.

2.2. Modification of the mechanism in the case of a spatial

variability of j

One of the key steps for generation of the mechanism dis-

cussed above is the computation of the coordinates of Point F,
since it allows (i) to know at which index j the generation should
be stopped when the extremity of the mechanism has been
reached, and (ii) to define the radius rf ¼OF of a circle which in Fig. 3. Numerical determination of the position of point F.
 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451 443

emerging from A (respectively from B) is composed of a

succession of points Aj (respectively Bj) belonging to the planes
Pj. The determination of the position of the point Aj þ 1 from
the point Aj belonging to Pj is based on the fact that (i) Aj þ 1
belongs to Pj þ 1 and (ii) the segment AjAj þ 1 makes an angle p/
2  j(Aj) with the velocity vector, with j(Aj) being the local
value of j at the coordinates of Aj. The same method is used to
compute the position of Bj þ 1 from the one of Bj. This iterative
process starts at the points A and B and stops when the two
curves meet at the extremity of the mechanism. The point F is
defined as the intersection of the two curves. More details
about this process may be found in Ref. [33].
– During the generation of a new point Pi,j þ 1 from two existing
points Pi,j and Pi þ 1,j of the previous plane, the normality
condition of the facet Fi,j should be verified with respect to
the local value of the friction angle j(x,y,z). In the present Fig. 4. Determination of the ‘equivalent’ Mohr–Coulomb parameters.
study, this local value is taken at the middle point of the
segment Pi,jPi þ 1,j.
– The expression of the energy dissipation has to be generalized s and a are parameters that depend on rock mass quality as given by
because the values of cohesion and of friction angle are GSI and by the disturbance factor (D); and mb is a parameter that
different for each facet Fi,j and F0 i,j. The result is Eq. (2), where depends on rock mass quality and rock type. See Ref. [34] for further
o is the angular velocity of the moving block. For the same details of the latest available version of the HB criterion.
reason, the critical pressure given by Eq. (1) is no longer valid, The introduction of the HB criterion in the context of limit
and it now has to be computed using a newly developed analysis with the newly proposed collapse mechanism requires to
expression presented in Eq. (3). be able to compute the ‘equivalent’ values of c and j correspond-
 X ing to a given value of the normal stress s0 n acting normal to the
W D ¼ o ðci,j cos ji,j Ri,j Si,j þ c0i,j cos j0i,j R0i,j S0i,j Þ ð2Þ failure surface. It is therefore necessary to define the HB failure
i,j
envelope in the Mohr-plane, since the ‘equivalent’ values of c and
tan j that will be employed in the limit analysis computation
P P correspond to the intercept and slope of the tangent line to this
g ðRi,j V i,j sin bi,j þR0i,j V 0i,j sin b0i,j Þ ðci,j cos ji,j Ri,j Si,j þ c0i,j cos j0i,j R0i,j S0i,j Þ
i,j i,j envelope at s0 ¼ s0 n (Fig. 4). In this way, we are substituting the
sc ¼ P
ðSj Rj cosbj Þ original failure criterion by its linear envelope at each stress
j
value. As shown by Ref. [28], solutions obtained using such
ð3Þ envelope are still valid upper-bound solutions that maintain the
rigour of the limit analysis approach.
In these expressions, ci,j and ji,j (respectively c0 i,j and j0 i,j) are the The HB limit curve is the envelope of the Mohr circles defined
values of c and j at the centroid of Fi,j (respectively F0 i,j). As for Eq. (1), by Eq. (4) for all the values of s0 3. More precisely, with the
Eq. (3) is only valid for a given position of the centre O of the rotational notations of Fig. 4, we can write that
motion. Moreover, since j and c are now a function of the stress field,
tmax ðs0n Þ ¼ max½tðs03 , s0n Þ ð5Þ
it is also only valid for a given stress field; i.e., for a given spatial 0 s3
distribution of the normal stress on the slip surface. Thus, finding the
In this expression, the function t(s0 3, s0 n) is obtained from the
critical collapse mechanism will require to maximize Eq. (3) not only
Mohr-circle indicating failure conditions for a stress state with
with respect to the position of O (two geometric parameters), but also
minor principal stress s0 3 (see Fig. 4), and thus depends on the
with respect to the stress field. To simplify such optimization, it is
expression of the HB criterion provided in Eq. (4). The equation of
convenient to make some assumptions on the ‘expression’ of this
this Mohr-circle verifies
stress field so that it can be expressed by means of a limited number
  0   0 
of parameters, which may then be introduced in the optimization s þ s03 2 s1 s03 2
process. Section 3.2 analyzes this problem with more detail, and it also ðtðs03 , s0n ÞÞ2 þ s0n  1 ¼ ð6Þ
2 2
discusses the types of stress distribution considered in this work.
Introducing Eq. (4) into Eq. (6) and simplifying, we get
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi
 0 0   sci s þmb s03 a 0
3. Application to fractured rock masses with the Hoek–Brown t s3 , sn ¼ sn s3 sci
0 0 sn þ s3 0 ð7Þ
sci
failure criterion
The solution of Eq. (5) is obtained by solving
3.1. Introduction of the HB criterion in the modified collapse dtðs03 , s0n Þ
mechanism ¼0 ð8Þ
ds03

The HB failure criterion is typically applied to fractured rock with

masses [34], since it accounts for the observation that rock mass
dtðs03 , s0n Þ ðsci ððmb =sci Þs03 þ sÞa s0n þ s03 Þ þ ðs0n s03 Þð1 þ mb aððmb =sci Þs03 þ sÞa1 Þ
strength is a non-linear function of stress level. The HB criterion ds’3
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðs0 s0 Þðs ððs s þ m s0 Þ=s Þa s0 þ s0 Þ
can be expressed as n 3 ci ci b 3 ci n 3

 a ð9Þ
s0
s01 ¼ s03 þ sci mb 3 þ s , ð4Þ
sci
Solving Eq. (8) analytically would directly provide a closed-form
where s 1 and s 3 are the major and minor effective principal stresses
0 0
expression for the HB criterion limit curve. However, due to the
at failure; sci is the uniaxial compressive strength of the intact rock; complexity of Eq. (9), this analytical solution seems out of reach.
444 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451

Mohr−Coulomb equivalent
Mohr Plane
parameters
100 70 35
90
GSI = 20

Tangential stress [kPa]

mi = 9.6 60 30
80
σci = 5 MPa 50 25
70
60 D =0
20

c [kPa]
40

φ [º]
50
30 15
40
30 20 φ 10
20 c
10 5
10
0 0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100

Normal stress [kPa] Normal stress [kPa]

Mohr−Coulomb equivalent
Mohr Plane
parameters
100 70 35
90
GSI = 10 φ
Tangential stress [kPa]

mi = 4.0 60 30
80
σci = 3 MPa 50 c 25
70
60 D =0

c [kPa]
40 20

φ [º]
50
30 15
40
30 20 10
20
10 5
10
0 0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Normal stress [kPa] Normal stress [kPa]

Fig. 5. Hoek–Brown limit curve and corresponding ‘equivalent’ values of c and j for two heavily fractured rock masses.

The problem of deriving analytical expressions for the HB failure generally has to be generated more than 100 times during the process
envelope in t–s space has attracted significant attention in the of optimization of the centre of rotation and of the stress field needed
literature. Ucar [35] employed the general derivation procedure for limit analysis.
previously proposed by Balmer [36] to develop a first set of
numerical equations to derive the value of t corresponding to the 3.2. Hypothesis about the stress field
failure envelope of the ‘original’ HB failure criterion; i.e., with
a ¼1/2. Such work was later generalized by Kumar [37], who As indicated above, we need to define the distribution of normal
developed a numerical solution for the failure envelope of the stresses along the failure surface. Theoretically, we could aim for a
‘generalized’ HB criterion. More recently, Shen et al. [38] have very flexible, i.e., with many degrees of freedom, distribution that
developed approximate analytical expressions for the ‘instanta- would introduce no constraints in the quality of the solution
neous’ c and j of the strength envelope which have been obtained; this approach, however, has the shortcoming that it
validated within specific ranges of the material properties. increases the dimensionality and hence the difficulty of the optimi-
In this work, Eq. (5) is solved numerically using the optimization zation problem.
tool of MATLAB, which directly provides the numerical value of tmax We employed Fast Lagrangian Analysis of Continua in 3
for a given value of s0 n, but does not provide any information about Dimensions (FLAC3D) [39] to conduct numerical simulations that
the slope and intercept of the tangent line to the limit curve at the help us identify reasonable, yet simple, stress distributions at the
point (s0 n, tmax). To obtain accurate estimates of c and j for any value tunnel face and, in particular, to compute the distributions of
of s0 n, the limit curve is approximated by a piecewise-linear function. normal stresses along the failure surface. The details of the
In other words, the values of tmax are computed numerically at a large numerical model are discussed below. Using FLAC’s internal
number of s0 n values over the whole relevant range of normal programming language (FISH) it is possible to compute and depict
pressures; in this case, a constant ‘step’ of 0.1 kPa between two the normal stress distribution along the failure surface. As an
successive values of s0 n was considered. Then, the values of c and j example, Figs. 6 and 7 reproduce our computed results in two
corresponding to each segment are computed in the middle point of particular cases: Fig. 6 represents a case with extremely poor
each of these small intervals and stored in a table. Values of c and j mass rock properties, and Fig. 7 represents a case in which
for any arbitrary s0 n are computed by linear interpolation between slightly better properties have been used; this causes a smaller
two successive values in this table. Note that, thanks to the very small failure mechanism that does not extend above the tunnel crown.
step chosen for s0 n, two successive values of c or of j in this table are As it can be observed the normal stresses along the failure
always extremely close. Two examples are provided in Fig. 5 for two surface tend to increase with depth. In the case with ‘worse’ rock
sets of mechanical parameters. It clearly appears that, as expected, mass properties this variation is pronounced, whereas, in the case
the non-linearity of the criterion leads to a reduction of the with ‘better’ properties, is small. Other cases computed, that, for the
‘equivalent’ friction angle and to an increase of the ‘equivalent’ sake of space, are not reproduced herein, present an intermediate
cohesion when the normal pressure increases. Thus, small values of behaviour between these two situations. Based on such results, and
the normal stress may lead to very high friction angles, up to 701. to assess the influence of the type of stress distribution considered
Such table of c j values combined with the interpolation method on the results computed, we choose to employ two different
provides two functions c(s0 n) and j(s0 n) which have the advantage of distributions of normal stresses along the slip surface in this work:
being very fast. This is very important since the algorithm of
generation of the collapsing block very often calls these functions. (1) A uniform distribution, defined by a constant stress in the
Indeed, they are actually needed several times for each facet, there are entire failure surface. With this distribution, the optimization
more than 10,000 facets in each mechanism, and the mechanism of the failure mechanism needs to be performed with respect
 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451 445

8.0 8.0

6.0 6.0

4.0 4.0

Y−Axis [m]
2.0
Y−Axis [m]

2.0

0.0 0.0

−2.0 −2.0

−4.0 −4.0

−6.0 −6.0
0 0.03 0.06 0.09 0.12 0 0.03 0.06 0.09 0.12
Normal Stress [MPa] Normal Stress [MPa]

Fig. 7. Distribution of normal stresses along the failure surface (Case mi ¼5;
Fig. 6. Distribution of normal stresses along the failure surface (Case mi ¼5;
sci ¼ 1 MPa; GSI¼ 15; D¼ 0; g ¼2.5 t/m3; Diameter ¼ 10 m): (a) spatial distribution sci ¼ 5 MPa; GSI¼ 20; D ¼ 0; g ¼2.5 t/m3; Diameter ¼ 10 m): (a) spatial distribution
and (b) at the cross section along the vertical plane of symmetry.
and (b) at the cross section along the vertical plane of symmetry.

to three parameters: two of them are geometrical and define minimal disturbance of the rock mass surrounding the tunnel. As
the position of the centre of rotation, and the third one shown in Table 1, the different cases have been selected so that
defines the constant value of the stress distribution. they correspond to low-quality rock masses (GSIr25), since, in
(2) A linear distribution, defined by two parameters that represent real practice, problems associated to tunnel face instability could
the stress value at the tunnel crown and the stress vertical be mainly expected in such low-quality rock masses.
gradient. Hence, the optimization is performed with respect
to four parameters: two geometrical parameters to define the 4.2. Results with limit analysis
coordinates of the centre of rotation and two additional
parameters relative to the stress distribution. The collapse pressures obtained by applying the new mechan-
ism described in Sections 2 and 3 are shown in Table 1. Results
computed employing the assumed uniform and linear stress
4. Validation of the model distributions along the failure surface are included as well. For
cases of ‘weaker’ rocks, in which the collapse pressure is higher
4.1. Description of test-cases than 8 kPa, the linear stress distribution improves the results, i.e.,
the collapse pressure is higher than that computed with the
To validate the proposed methodology, we have used 22 test- uniform distribution. Note that, being an upper-bound limit
cases to compare results obtained using our improved failure analysis approach, and considering that the face pressure acts
mechanism with results obtained using numerical models in against the movement of the mechanism, any solution that
FLAC3D. Table 1 includes a list of parameters (mi, sci, GSI and g) increases the collapse pressure represents an improvement of the
employed for the analyses of the test-cases considered. In all cases results. In such cases with ‘worse’ rock mass properties, the increase
a 10 m diameter tunnel has been modelled with a cover of 20 m with depth of normal stresses along the failure surface is similar to
that is enough to eliminate the possibility that the mechanism that shown in Fig. 6b; hence, using a linear distribution improves the
outcrops at the ground surface. Also, a damage parameter of D¼0 predictions. Therefore, the newly proposed generalized failure
has been employed to represent TBM-excavated tunnels with a mechanism, which allows considering non-constant stress
446 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451

Table 1
Test-cases considered for validation.

Case Parameters Collapse pressure (kPa)

mi sci (MPa) GSI g (t/m3) Limit analysis Numerical model Difference limit analysis—numerical
model (associated flow)
Uniform Linear Associated flow Non-associated flow

1 5 1 10 2.5 49.5 52.0 61.9 62.2 9.9

2 5 1 15 2.5 34.9 36.8 38.6 39.3 1.8
3 5 1 20 2.5 25.6 26.8 26.3 27.7  0.5
4 5 1 25 2.5 19.1 20.3 18.9 21.1  1.4
5 5 5 10 2.5 16.4 17.1 15.3 15.4  1.8
6 5 5 15 2.5 9.4 9.8 8.3 12.5  1.5
7 5 5 20 2.5 5.2 5.2 5.3 6.1 0.1
8 5 5 25 2.5 2.3 2.3 2.4 2.3 0.1
9 5 10 10 2.5 8.4 9.1 7.7 10.9  1.4
10 5 10 15 2.5 3.4 3.4 2.7 4.2  0.7
11 5 10 20 2.5 0.0 0.0 0.0 0.0 0.0
12 5 10 25 2.5 0.0 0.0 0.0 0.0 0.0
13 5 15 10 2.5 5.2 5.2 4.8 6.9  0.4
14 5 15 15 2.5 0.1 0.1 0.3 0.0 0.2
15 5 15 20 2.5 0.0 0.0 0.0 0.0 0.0
16 5 15 25 2.5 0.0 0.0 0.0 0.0 0.0
17 5 20 10 2.5 2.3 2.3 2.6 3.0 0.3
18 5 20 15 2.5 0.0 0.0 0.0 0.0 0.0
19 5 20 20 2.5 0.0 0.0 0.0 0.0 0.0
20 5 20 25 2.5 0.0 0.0 0.0 0.0 0.0
21 10 1 10 2.5 26.3 27.6 26.4 28.4  1.2
22 15 1 10 2.5 18.1 18.9 17.1 20.1  1.8

distributions, is needed in this case to improve previous approaches effect on the collapse pressure, are given by a Young’s modulus of
that only consider a constant stress field obtained from an optimum E¼400 MPa and a Poisson’s ratio of n ¼0.30. The rock density was
tangency point to the failure criterion. In other cases, corresponding taken equal to 2.5 t/m3. As shown in Table 1, the remaining
to rock masses with ‘better’ properties that produce very small parameters to define the failure criterion are specific of each test-case.
values of the critical pressure, and taking into account the In the framework of limit analysis, the assumption of an
distribution shown in Fig. 7b and the results presented in Table 1, associated flow rule with the dilatancy angle c in any point equal
there is a negligible advantage in considering the additional para- to the friction angle j, is necessary for the derivation of the
meter needed for the linear distribution and the results provided by fundamental theorems of this theory and, therefore, to provide
the constant stress field are probably acceptable. rigorous bounds of the critical load. This associated flow hypothesis,
Fig. 8 shows two examples, corresponding to test-cases 2 however, does not necessarily represent the real behaviour of
and 7, of the optimal failure geometries obtained using the new geotechnical materials so that, in general, the dilatancy angle tends
collapse mechanism. Fig. 8a corresponds to a case with ‘worse’ to be lower than the friction angle (see eg. Ref. [40]). While it is not
properties, and shows an instability that extends upwards in the possible to get rid of this assumption in our analytical mechanism, it
vertical direction, whereas Fig. 8b represents a case with better is possible to evaluate its influence on the results by dealing with
properties and shows a minor instability that only affects a small two limit cases in our numerical simulations: an associated flow
volume at the tunnel face. rule (c ¼ j) and a zero-dilatancy behaviour (c ¼0). As indicated by
Ref. [33], the key task is to evaluate the validity of this assumption
4.3. Numerical results with a finite difference code for the problem studied. Numerical results provided in Table 1 show
that this assumption has actually a limited influence on the value of
To validate our methodology, results have been compared with the critical collapse pressure, so that the theoretical background of
numerical simulations using the finite difference code FLAC3D. limit analysis is therefore acceptable in this case.
Fig. 9 illustrates the FLAC3D model employed for numerical To find the collapse pressure of the tunnel face, the bisection
simulation of tunnel face stability. The tunnel considered has a method proposed by [41] has been employed. Using this method,
diameter of 10 m and a cover of 20 m. To minimize boundary and for a given interval of pressure values, defined by one higher
effects and the needed computational effort, the tunnel has been value for which the face is stable and by one lower value for which it
represented using a symmetric model with dimensions, in metres, is unstable, the stability of the face is computed assuming the mean
of 25  30  35. A total of 163,260 elements have been employed, value; if the face is stable, the higher interval boundary is sub-
of which 819 are in the tunnel front, and the mesh has been stituted by such mean value (the lower interval value is substituted
designed to minimize element sizes where larger stress gradients if unstable), and the process is iteratively repeated for such newly
are expected. The boundary conditions of the model are given by defined intervals until the required precision is achieved, which in
fixed displacements at the ‘artificial’ boundaries of the model, i.e., this case has been set to 0.1 kPa. To assess the stability of the tunnel
at its lateral perimeter and at its base. Similarly, and since we are face for each face pressure considered, we use a methodology that is
only concerned with face stability and not with tunnel conver- similar to that employed by FLAC3D for the determination of safety
gences, the tunnel support has not been included and displace- factors (refer to FLAC3D manual for details [39]) and that depends on
ments at the tunnel excavation boundary have also been fixed. the ‘‘representative number of steps (N)’’ which characterizes the
The constitutive model employed for the fractured rock mass is model behaviour. To obtain N, we set an elastic behaviour in the
elastic–perfectly plastic with the HB failure criterion as implemented model; then, from an equilibrium state, and doubling the internal
in FLAC3D. The elastic properties employed, which have a negligible stresses in the elements, we obtain the number of steps needed to
 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451 447

Fig. 8. Collapse geometries obtained by the new mechanism. (a) Case 2 (mi ¼ 5; sci ¼1 MPa; GSI ¼15; D¼ 0; g ¼2.5 t/m3; Diameter ¼10 m) and (b) Case 7 (mi ¼5;
sci ¼ 5 MPa; GSI ¼20; D ¼ 0; g ¼ 2.5 t/m3; Diameter¼ 10 m).

restore the model to equilibrium. For our model, this number is we obtain very similar critical pressures for the numerical model
between 15,000 and 20,000 steps; because of this, we set N¼25,000 and for the rotational failure mechanism. Except for test-case 1,
steps in all cases. To assess stability we consider that the real tunnel all differences are lower than 2 kPa, a value which can be
model is unstable if it does not converge in these N steps. As a considered acceptable for practical purposes as a validation of
convergence criterion, the FLAC3D default value of 1EXP-05 for the the new methodology. However, note that small differences in the
unbalanced mechanical-force ratio has been reduced to 1EXP-07. numerical results arise from the definition of the convergence
To check the results of our proposed limit analysis approach criterion, which depends on the selected number of steps; and
with the results of numerical simulations, we have compared on the mesh size since, as shown by [33], the predictions are
(i) the computed values of the collapse pressure at the tunnel face; expected to improve as the mesh size is decreased. This makes the
and (ii) the shapes of the failure mechanisms obtained by both numerical model to be not completely accurate and it also
methods. Failure mechanisms in the FLAC3D model have been explains why, in some cases, collapse pressures in the numerical
estimated considering the distribution of shear deformations. model are lower than in the framework of limit analysis. As
Table 1 and Fig. 10 present a comparison of collapse face mentioned above, test-case 1 shows the only remarkable differ-
pressures computed using the FLAC3D numerical model and using ence, 9.9 kPa, from 52.0 to 61.9 kPa. The rock mass properties for
our limit analysis approach with the improved collapse mechan- this case are: sci ¼1 MPa, mi ¼5 and GSI ¼10, which are the worst
ism. To compare results that correspond to equal conditions, in properties considered in this work, and that are probably more
Table 1 the limit analysis results are only compared to the representative of a ‘soil like’ material. It is likely that the solution
numerical results for the case of associated flow. In most cases, could be improved using additional parameters in the distribution
448 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451

Fig. 9. Geometry of the developed numerical model.

70
Limit Analysis
N.M. Associated
60 N.M. No associated

50
Collapse Pressure [MPa]

40

30

20

10

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Fig. 11. Comparison of failure mechanisms computed with the limit analysis and
Case with the numerical simulation: (a) Case 2 (mi ¼5; sci ¼ 1 MPa; GSI ¼15; D ¼ 0;
g ¼ 2.5 t/m3; Diameter ¼10 m) and (b) Case 7 (mi ¼5; sci ¼ 5 MPa; GSI ¼20; D ¼ 0;
Fig. 10. Comparison of collapse face pressures. g ¼ 2.5 t/m3; Diameter ¼10 m).

of normal stresses along the failure surface; however, since these failure. However, when the number of elements at the face were
materials seem to represent ‘marginal’ cases, we choose not to increased to 5846 using a much finer mesh, the failure surface
generalize the shape of the stress distribution with additional presented in Fig. 11b was obtained. It can be clearly observed
parameters to reduce difficulties during optimization. that the failure mechanism affects the whole tunnel face, and
Furthermore, the shapes of the failures mechanisms obtained with also that the shape of the failure surface agrees well with the
our limit analysis approach and with the FLAC3D numerical simula- shape provided by our limit analysis approach.
tions are in good agreement. As an example, Fig. 11 presents a Finally, it is important to emphasize that the newly proposed limit
comparison of the failure mechanisms for the same test-cases, analysis approach is significantly more computationally efficient. As
2 and 7, that were shown in Fig. 8. It has to be noted that test- an illustration, the times required to calculate test-cases 2 and 7, on
case 7 illustrates an example in which the mesh size had an an Intel Xeon CPU W3520 2.67-GHz PC are, respectively, 67 and
effect on the shape of the failure mechanism obtained with 23 min for the limit analysis approach coded in MATLAB, whereas 26
FLAC3D: if the model described above, with 819 elements in the and 20 h are needed for the numerical simulations with FLAC3D.
tunnel front, was employed, the slip surface obtained was Moreover, using the model with the much finer mesh, which could be
slightly different to the one obtained with the limit analysis necessary to predict slip surfaces accurately, the time required to
mechanism and, in fact, suggested the possibility of a partial face calculate test-case 7 is higher than 180 h.
 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451 449

5. Design charts and of the Geological Strength Index (GSI). As mentioned above,
we consider null disturbance (D¼ 0) to represent TBM-excavated
After its validation, the newly proposed mechanism is used to tunnels with a minimal damage of the rock mass surrounding the
produce design charts to compute the critical collapse pressure of tunnel. Note that ranges of parameters have been selected to
a tunnel face excavated in heavily fractured and ‘low quality’ rock represent poor quality rock masses, where face instability pro-
masses that follow the HB non-linear failure criterion. Fig. 12 blems are more likely in practice. Accordingly, sci/(gDm) is varied
shows the collapse pressure of the tunnel face (sc), divided by the from 4 to 100 (equivalent in these charts to sci between 0.5 and
diameter of the tunnel (Dm) and by the unit weight of the rock 30 MPa), which are typical of soft rocks to very soft rocks; and GSI
mass (g), as a function of the uniaxial compressive strength (sci), values are taken lower than 25, which are typical of very poor
also divided by Dm and g, for different values of the mi parameter quality rock masses.

mi = 5 m i = 10
0.25 0.15
GSI = 10 GSI = 10
GSI = 15 GSI = 15
GSI = 20 GSI = 20
0.20 GSI = 25 GSI = 25

0.10
0.15
σc / γ Dm
σc / γ D m

0.10
0.05

0.05

0.00 0.00
0 20 40 60 80 100 0 20 40 60 80 100

σci / γ Dm σci / γ Dm

mi = 15 m i = 20

0.15 0.15
GSI = 10 GSI = 10
GSI = 15 GSI = 15
GSI = 20 GSI = 20
GSI = 25 GSI = 25

0.10 0.10
σc / γ D m

σc / γ Dm

0.05 0.05

0.00 0.00
0 20 40 60 80 100 0 20 40 60 80 100
σci / γ Dm σ ci / γ Dm

Fig. 12. Design charts of the critical collapse pressure for Hoek–Brown material.
450 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451

To produce the design charts presented in Fig. 12, a total of The new approach has been further employed to develop
1632 cases were computed and included in the plots. Such cases ‘design charts’ to estimate the face collapse pressure of TBM
considered tunnel diameters ranging between 6 and 12 m, and tunnels excavated in weak or very weak HB rock masses with a
unit weights for the rock mass ranging between 2.0 and 2.5 t/m3. poor quality. To limit the applicability of the approach, however, it
In all cases, it has been assumed that the collapse mechanism has be reminded that the HB failure criterion assumes an isotropic
does not reach the surface, which, for the most unfavourable case, rock mass behaviour, and that it should only be applied when the
occurs when the tunnel cover is larger than 0.5D. structure size, relatively to the spacing between discontinuities,
As expected, Fig. 12 shows a reduction of the collapse pressure makes it possible to consider the rock mass as a ‘continuum’
when rock strength and rock mass quality increase, with a clear instead of a blocky structure [42]. Therefore, the results presented
influence of sci and GSI. In most cases, the values of the critical herein are only applicable to heavily fractured rock masses where
pressure are low and, in some cases, the collapse pressure is null; these assumptions are valid and, for instance, they are not applic-
that means that the face is self-stable and that it is not necessary able to unstable blocks or wedges defined by intersecting struc-
to apply any pressure to support it. Note that, although our model tural discontinuities. Similarly, as solutions of the limit analysis
provides negative values for the critical pressure when the face is problem, solutions presented herein do not consider the deforma-
self-stable (meaning that it would be necessary to ‘‘pull’’ the face tions at the tunnel face and, for instance, they do not account for
to trigger instability), this value of a null pressure was chosen as a squeezing failures associated to high deformations of the material
convention for such stable cases. when subjected to high stresses [43]. As a consequence, the use of
the design charts presented in Fig. 12 should be made in a wider
design context that considers other factors that could influence the
success of the project.
6. Conclusions

We present and validate a new analytical failure mechanism Acknowledgements

for the determination, in the framework of limit analysis, of the
critical collapse pressure and of the geometry of the collapse Salvador Senent holds a PhD Scholarship provided by Funda-
mechanism, for the face of tunnels excavated in low quality rock ción José Entrecanales Ibarra. This research was funded, in part,
masses with the HB non-linear failure criterion. The use of a non- by the Spanish Ministry of Economy and Competitiveness, under
linear failure criterion introduces the need to consider the grant BIA2012-34326. The support of both Institutions is grate-
distribution of normal stresses along the failure surface, so that fully acknowledged.
the ‘local’ friction angle can be computed to fulfill the assumption
of associated flow that is inherent to limit analysis. As a conse-
quence, there is a need to introduce new parameters in the Appendix A. Supporting information
optimization problem that allow us to consider such stress
distribution. To be able to consider the non-linearity of the HB Supplementary data associated with this article can be found in
criterion, we improve an advanced, and recently proposed, failure the online version at http://dx.doi.org/10.1016/j.ijrmms.2013.01.004.
mechanism for the tunnel face [22]; the mechanism, that covers
the whole excavation front, is generated ‘‘point-by-point’’, and it
provides a rotational-type failure that is very similar to that References
observed in small-scale tunnel tests in the laboratory. The
mechanism makes it possible to work with variable MC materials [1] Health and Safety Executive. Safety of new Austrian tunnelling method
properties, and it represents the more advanced tunnel face (NATM) tunnels. A review of sprayed concrete tunnels with particular
reference to London Clay. Sudbury: (HSE) Books; 1996.
failure mechanism that has been proposed to this date. Numerical [2] Horn N. Horizontaler erddruck auf senkrechte abschlussflächen von tun-
simulations have been employed to identify adequate, yet simple, nelröhren. Landeskonferenz der ungarischen tiefbauindustrie. Deutsche
stress distributions at the tunnel face. The results of such Überarbeitung durch STUVA, Düsseldorf, 1961. p. 7–16.
[3] Anagnostou G, Kovári K. Face stability conditions with earth-pressure-
simulations suggest that a linear distribution of stresses along
balanced shields. Tunn Undergr Space Technol 1996;11:165–73.
the failure surface could be employed as an approximation to the [4] Broere W. Face stability calculation for a slurry shield in heterogeneous soft
real stress distribution in many practical applications. Such soils. In: Negro A, Ferreira AA, editors. Proceedings of the world tunnel
assumed linear distribution of normal stresses is further shown congress 98 on tunnels and metropolises, Sao Paulo, Brazil, 25–30 April 1998.
Rotterdam: Balkema, 1998. p. 215–8.
to capture better the normal stress variation in cases with ‘worse’ [5] Vermeer PA, Ruse NM, Dong Z, Härle D. Ortsbruststabilität von tunnelbau-
rock mass properties, hence improving the prediction of the werken am beispiel des Rennsteig tunnels. In: 2. Kolloquium—Bauen in
critical pressure. The increased complexity of the stress distribu- Boden und Fels. Technische Akademie Esslinggen, Esslingen; January 2000.
p. 195–202.
tion, however, does not seem to improve the results in other cases [6] Melis MJ. El colapso del túnel ferroviario por inestabilidad del frente en
with ‘better’ rock mass properties, when the computed critical suelos y rocas blandas o muy diaclasadas. Rev Obras Púb 2004;3450:33–64.
pressures are almost equal to the uniform distribution case. In [7] Kimura T, Mair RJ. Centrifugal testing of model tunnels in soft clay. In:
Proceedings of 10th International Conference on Soil Mechanics and Founda-
addition, to validate the new failure mechanism, 22 test-cases tion Engineering, Stockholm, 15–19 June 1981. Rotterdam: Balkema, 1981.
corresponding to rock masses with low quality, as indicated by p. 319–22.
their GSI value, have been employed to compare our limit analysis [8] Chambon P, Corté JF. Shallow tunnels in cohesionless soil: stability of tunnel
face. J Geotech Eng 1994;120:1148–65.
results with results of three-dimensional simulations conducted [9] Takano D, Otani J, Nagatani H, Mukunoki T. Application of X-ray CT on
with FLAC3D. Two aspects have been compared: (i) the numerical boundary value problems in geotechnical engineering: research on tunnel face
value of the collapse pressure; and (ii) the shape of the failure failure. In: GeoCongress 2006: Geotechnical Engineering in the Information
Technology Age. Proceedings of GeoCongress 2006, Atlanta, Georgia, USA.
mechanism. The obtained results suggest that the limit analysis
[10] Kirsch A. Experimental investigation of the face stability of shallow tunnels in
approach proposed herein successfully approximates the FLAC3D sand. Acta Geotech 2010;5:43–62.
numerical results but with a significantly reduced computational [11] Vermeer PA, Ruse N. Die stabilität der tunnelortsbrust in homogenem
cost, so that it could be applied for fast, and relatively reliable, baugrund. Geotechnik 2001;24:186–93.
[12] Dias D, Janin J, Soubra A, Kastner R. Three-dimensional face stability analysis
estimations of the pressure needed for face support in shallow of circular tunnels by numerical simulations. In: Alshawabkeh AN, Reddy KR,
tunnels excavated in heavily fractured rock masses. Khire MV, editors. GeoCongress 2008: Characterization, Monitoring, and
 S. Senent et al. / International Journal of Rock Mechanics & Mining Sciences 60 (2013) 440–451 451

Modeling of GeoSystems. Proceedings of GeoCongress 2008, New Orleans, [27] Yang XL, Yin JH. Slope stability analysis with nonlinear failure criterion. J Eng
Louisiana, USA, 9–12 March 2008. p. 886–93. Mech 2004;130:267–73.
[13] Kim SH, Tonon F. Face stability and required support pressure for TBM driven [28] Chen WF. Limit analysis and soil plasticity. Amsterdam: Elsevier; 1975.
tunnels with ideal face membrane—drained case. Tunn Undergr Space [29] Yang XL, Li L, Yin JH. Stability analysis of rock slopes with a modified Hoek–
Technol 2010;25:526–42. Brown failure criterion. Int J Numer Anal Methods Geomech 2004;28:
[14] Melis MJ, Medina LE. Discrete numerical model for analysis of earth pressure 181–90.
balance tunnel excavation. J Geotech Geoenviron Eng 2005;131:1234–42. [30] Yang XL, Li L, Yin JH. Seismic and static stability analyisis for rock slopes by a
[15] Chen RP, Tang LJ, Ling DS, Chen YM. Face stability analysis of shallow shield kinematical approach. Géotechnique 2004;54:543–9.
tunnels in dry sandy ground using the discrete element method. Comput [31] Huang F, Yang XL. Upper bound solutions for the face stability of shallow
Geotech 2011;38:187–95. circular tunnels subjected to nonlinear failure criterion. In: Tonon F, Liu X,
[16] Davis EH, Gunn MJ, Mair RJ, Seneviratine HN. The stability of shallow tunnels and Wu W, editors. Deep and underground excavations. Proceedings of the 2010
underground openings in cohesive material. Géotechnique 1980;30:397–416. GeoShangai international conference, Shangai, China; 2010. p. 251–6.
[17] Leca E, Panet M. Application du calcul a la rupture a la stabilité du front de [32] Agar JG, Morgenstern NR, Scott JD. Shear strength and stress strain behaviour
taille d’un tunnel. Rev France Geotech 1988:5–19. of Athabasca oil sand at elevated temperatures and pressures. Can Geotech J
[18] Leca E, Dormieux L. Upper and lower bound solutions for the face stability 1987;24:1–10.
of shallow circular tunnels in frictional material. Géotechnique 1990;40: [33] Mollon G, Phoon KK, Dias D, Soubra AH. Validation of a new 2D failure
581–606.
mechanism for the stability analysis of a pressurized tunnel face in a spatially
[19] Oberlé S. Application de la méthode cinématique a l’étude de la stabilité d’un
varying sand. J Eng Mech 2011;137:8–21.
front de taille de tunnel, Final project. ENSAIS, France. 1996.
[34] Hoek E, Carranza-Torres C, Corkum B. Hoek–Brown failure criterion—2002
[20] Soubra AH. Three-dimensional face stability analysis of shallow circular
edition. In: Hammah R, Bawden W, Curran J, Telesnicki M, editors. Proceed-
tunnels. In: GeoEng2000: An International Conference on Geotechnical and
ings of the North American rock mechanics symposium, University of
Geological Engineering. Proceedings of the International Conference on
Geotechnical and Geological Engineering, 19–24 November 2000, Melbourne, Toronto; 2002, p. 267–73.
Australia; 2000. [35] Ucar R. Determination of shear failure envelope in rock masses. J Geotech Eng
[21] Mollon G, Dias D, Soubra AH. Face stability analysis of circular tunnels driven 1986;112:303–15.
by a pressurized shield. J Geotech Geoenvir 2010;136:215–29. [36] Balmer G. A general analytical solution for Mohr’s envelope. Am Soc Test
[22] Mollon G, Dias D, Soubra AH. Rotational failure mechanisms for the face Mater J 1952;52:1269–71.
stability analysis of tunnels driven by a pressurized shield. Int J Numer Anal [37] Kumar P. Shear failure envelope of Hoek–Brown criterion for rockmass. Tunn
Methods Geomech 2011;35:1363–88. Undergr Space Technol 1998;13:453–8.
[23] Sofianos AI, Nomikos PP. Equivalent Mohr–Coulomb and generalized Hoek– [38] Shen J, Priest SD, Karakus M. Determination of Mohr–Coulomb shear strength
Brown strength parameters for supported axisymmetric tunnels in plastic or parameters from generalized Hoek–Brown criterion for slope stability ana-
brittle rock. Int J Rock Mech Min Sci 2006;43:683–704. lysis. Rock Mech Rock Eng 2012;45:123–9.
[24] Jimenez R, Serrano A, Olalla C. Linearization of the Hoek and Brown rock [39] Itasca Consulting Group. FLAC3D 4.0 manual. Minneapolis; 2009.
failure criterion for tunnelling in elasto-plastic rock masses. Int J Rock Mech [40] Alejano LR, Alonso E. Considerations of the dilatancy angle in rocks and rock
Min Sci 2008;45:1153–63. masses. Int J Rock Mech Min Sci 2005;42:481–507.
[25] Fraldi M, Guarracino F. Limit analysis of collapse mechanisms in cavities and [41] Mollon G, Dias D, Soubra AH. Probabilistic analysis of circular tunnels in
tunnels according to the Hoek–Brown failure criterion. Int J Rock Mech Min homogeneous soil using response surface methodology. J Geotech Geoen-
Sci 2009;46:665–73. viron Eng 2009;135:1314–25.
[26] Huang F, Yang XL. Upper bound limit analysis of collapse shape for circular [42] Hoek E, Brown ET. Practical estimates of rock mass strength. Int J Rock Mech
tunnel subjected to pore pressure based on the Hoek–Brown failure criterion. Min Sci 1997;34:1165–86.
Tunn Undergr Space Technol 2011;26:614–8. [43] Hoek E. Big tunnels in bad rock. J Geotech Geoenviron Eng 2001;127:726–40.

View publication stats