Tunnelling and Underground Space Technology 73 (2018) 154–161

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

journal homepage: www.elsevier.com/locate/tust

Revisiting crown stability of tunnels deeply buried in non-uniform rock T

surrounds
⁎
Changbing Qin , Siau Chen Chian
Department of Civil & Environmental Engineering, National University of Singapore, Singapore

A R T I C L E I N F O A B S T R A C T

Keywords: A kinematic analysis procedure is presented to investigate the crown stability of deep-buried tunnels in non-
Tunnel crown stability uniform rock strata. The non-uniformity of rock properties is considered using the discretized approach with
Kinematic analysis finite thickness layers. In the realm of upper bound theory, kinematic analysis is briefly presented for a single
Discretization layer so as to provide basis for multi-layer analysis. A 2D collapse mechanism is postulated to depict the potential
2D and 3D collapse mechanism
falling block with an arbitrary tunnel cross-section, although rectangular tunnels are considered in the 3D ki-
Non-uniform rock surrounds
nematic analysis for ease of calculation. An impending collapse is likely to occur, provided that the width of the
excavation is no less than the collapse width. In order to reflect the nonlinear characteristics of rock materials at
failure, the Hoek-Brown criterion is employed for tightly interlocked and very poor rock masses, with its form
expressed by normal and shear stresses to facilitate calculations of internal energy dissipation rate. Based on the
variational principle and power-based balance equation, the effective collapse mechanism is derived in closed
form. Comparison is carried out to verify the robustness of the proposed approach. Sensitivity analysis is per-
formed to depict the influence of each non-uniform rock strength parameter on crown stability. As expected,
such a discretized approach incorporated into kinematic analysis enables one to account for non-uniformity of
rock surrounds, which would fill the knowledge and design gap of similar tunnelling issues much needed at
present state of practice.

1. Introduction
Along with the growing demands for underground highways and
high-speed rail networks, an increasing number of tunnels are con-
structed. In order to ensure excavation safety, stability of tunnel roof
and face should be checked. A commonly used approach is to apply
kinematic analysis where the upper bound solution could be derived
under the ultimate limit state. In this study, focus is placed on roof
stability of a deep-buried tunnel constructed in non-uniform rock strata.
The linear Mohr-Coulomb (MC) failure criterion is commonly used
in soil mechanics. In order to represent the nonlinear constitutive re-
lationship of the rock whose strength depends on normal stress level,
the Hoek-Brown (HB) criterion is employed instead. The merit of the
HB criterion is its dimensionless relationship in terms of the geological
information and characteristics of the rock mass (Hoek and Brown,
1997). The widespread use of the HB failure criterion was also attrib-
uted to the establishment of Bieniawski’s rock mass rating system
(1976, 1989). Notice that this criterion is based on the assumption that
the rock failure is controlled by individual rock pieces rather than the
intact rock, hence it can be applied to tightly interlocked hard rock
mass. As an alternative, the HB criterion is also applied for very poor
quality rock masses provided that the rock mass can be deemed iso-
tropic with ‘chaotic’ joint pattern and no preferred failure directions.
The original HB failure criterion was developed based on the principal
stress, and till date, it has been updated to the generalized pattern. A
pattern expressed by normal and shear stresses is applied for ease of
calculation of internal energy dissipation rate.
The limit analysis has been widely adopted in predicting tunnel
stability. The preliminary study was principally performed based upon
the wedge block collapse mechanism. Yang et al. (2013) derived the
kinematic solution of supporting pressure required for maintaining
tunnel crown stability, considering an arc-sandwich and a log-sandwich
failure mechanism. For generalization, a random curved velocity dis-
continuous line was postulated by Fraldi and Guarracino (2009, 2010,
2011), to investigate the stability of tunnel crown with the analytical
upper bound solution derived from variational principle and plasticity
theory. Although complex, the use of HB failure criterion considers the
random variability of mechanical properties of rock masses as well as
geological conditions in its input. Such an analysis together with the
employment of the HB failure criterion was then extended further to

⁎
Corresponding author.
E-mail address: changbingqin@u.nus.edu (C. Qin).

https://doi.org/10.1016/j.tust.2017.12.006
Received 15 May 2017; Received in revised form 17 August 2017; Accepted 6 December 2017
0886-7798/ © 2017 Elsevier Ltd. All rights reserved.
C. Qin, S.C. Chian
account for other complicated scenarios, such as pore water effect
(Huang and Yang, 2011) and seepage force (Yang and Qin, 2014). It is
worthwhile noting that the actual rock failure above tunnel roof is
three-dimensional (3D) rather than a simplified two-dimensional (2D)
case. Yang and Huang (2013) proposed a 3D collapse mechanism for
this specific purpose to investigate the crown stability within the fra-
mework of limit analysis theory. Such a 3D mechanism is sensible for
rectangular tunnels while just an approximation for circular tunnels,
because of the assumptions made in the failure mechanism and com-
plicated integrals (Qin et al., 2013; Huang et al., 2013). The above
mentioned analyses provide useful avenues for assessing the stability of
tunnel crown in two and three dimensions. However, these analyses
were carried out on a homogeneous/uniform rock mass, and hardly
able to consider non-uniformity of surrounding rocks with diverse rock
properties. Because of this limitation in this respect, an attempt was
made by Qin et al. (2015b) to investigate the progressive failure oc-
curred in partly weathered rock medium. Such a progressive failure was
proven to occur in iron mines and limestone caves (Whittaker and
Reddish, 1989), and was also substantiated in Qin et al. (2015b) under
certain situations. Further, the progressive failure was discussed on
account of the pore water effect in partly saturated rock materials (Qin
et al., 2015a). In order to essentially deal with the random variation of
mechanical properties of rock masses, a novel approach is necessitated,
which will be elaborated later.
A general case with non-uniformity of rock surrounds is considered.
For simplification the non-uniform rock mechanical parameters are
represented separately in each finite thickness layer upon which the
kinematic analysis is carried out. The 2D and 3D failure mechanisms are
established to predict the falling blocks and the failure area. Moreover,
the comparison analysis gives meaningful physical insights to the in-
fluential parameters on tunnel crown stability, thereby providing gui-
dance on the design and excavation of deep tunnels.
2. Methodology
As discussed earlier, this study aims to investigate the tunnel crown
stability with considerations of non-uniformity of rock strength para-
meters. The rock surrounds are divided into finite thickness layers with
different rock properties including rock unit weight and strength
parameters. According to the nature of most geological formation, the
choice of horizontal layer is sensible in many cases. This is achieved
using discretized analysis which allows for the discretization of a
complicated problem into various components. The discretization
technique helps to consider non-uniform rock properties in the kine-
matic analysis, and has insignificant effect on the whole collapse me-
chanism, and therefore yielding reliable upper bound solutions. More
importantly, the employment of the discretization technique can ac-
count for the variation of rock properties in sub-layers with ease.
Apart from the discretization technique used in the pre-processing
of non-uniform rock strata, the core analysis lies in the application of
upper bound theorem. It states that if a kinematically admissible ve-
locity field can be constructed, then the external load computed from
the equilibrium of power-based balance equation, will be either higher
or equal to the actual failure load (Chen, 1975). The merit of the ki-
nematic analysis is no stress involved in the specific calculations, hence
popular in resolving geotechnical stability problems. The prerequisite
of such an analysis is to postulate a kinematically admissible velocity
field, based on which the rates of work done by external and internal
forces are calculated to form the objective function.
3. 2D kinematic analysis
The crown stability of deeply buried tunnels is investigated herein
in two dimensions. Prior to the kinematic analysis, a 2D collapse me-
chanism is established to depict the potential falling block. The work
rate calculations are performed considering one-layer (uniform) and
155
Tunnelling and Underground Space Technology 73 (2018) 154–161
multi-layers (non-uniform) rock strata. The upper bound solutions of
critical failure area are formulated based on the work rate balance
equation.
3.1. 2D collapse mechanism
The rock surrounds are assumed to be treated as isotropic with no
inclined and failure directions. Although the rock masses present some
anisotropy in the presence of certain circumstances, such as faults,
weak interlayers and joints, this may be valid provided the following
conditions are satisfied: (1) there exist no faults or bedding planes, (2)
directions of discontinuity surfaces are uniformly distributed, (3) the
joint separation is small when compared with the magnitude of rock
structures, and (4) the discontinuity surfaces must be sufficiently dense
(Qin et al., 2017). The assumptions imply that the spacings between
adjacent discontinuities are sufficiently small compared to the overall
dimension of rock structures. The potential failure block falls from the
overlaying rock above tunnel crown, indicating a translational failure
pattern under plane strain and symmetric conditions. In this case, the
kinematic admissibility condition can be satisfied, as long as the falling
velocity at failure in each layer remains the same and under continuous
deformation boundaries.
The potential collapse mechanism is illustrated in Fig. 1 where the
velocity discontinuities are composed by curved lines, f1(x),
f2(x) … fn(x). The failure area in each sub-section, demarcated by the
layer interfaces and fi(x), is directly influenced by the specific rock
properties. For a specific problem, the overall failure height is unknown
but unique. Since the surrounding rock is composed by finite thickness
layers, it is logical to adopt the equipartition approach to consider the
whole failure height in n layers. When n is made equal to 1.0, it implies
a homogenous/uniform rock mass, and multi-layer rock masses for n
larger than 1.0. Manifestly, a continuous rock media with constant or
varying properties is simulated for the case of n > 1. However, it con-
sumes too much computational effort under higher n values. On the
other hand, to little n may produce unreliable results. The choice of n
value is hence of vital importance. It is affected by and proportional to
the total collapsing height H, which will be discussed in further details
in the next section. Another quantity used to characterize the failure
region is the (half) failure width, in terms of L1, L2 … Ln in each layer.
Accordingly, a total of n + 1 variables are necessitated to depict the
whole collapse area. In order to encompass wider scenarios, the tunnel
with an arbitrary cross-section is considered.
Fig. 1. 2D potential collapse mechansim of tunnel crown in non-uniform rock surrounds.
C. Qin, S.C. Chian Tunnelling and Underground Space Technology 73 (2018) 154–161

3.2. Single-layer analysis from those remaining at rest. Internal energy dissipates along diverse
detaching surfaces, with a similar form in dissipation density as written
This section gives a brief review of the kinematic analysis applied in in Eq. (1). Therefore, the total dissipation rates yield
a homogeneous/uniform rock layer, which can also be found in (Fraldi n
L1 L
and Guarracino, 2009). Under limit state, the original rock mass above Wḋ = ∫0 wḋ 1· 1 + f1′ (x )2 dx + ∑ ∫L i ̇ · 1 + fi′ (x )2 dx
wdi
i−1
the tunnel crown is unable to maintain its stability on its own, in the i=2

process of tunnel excavation. For deep-buried tunnels, the detaching L1

⎡−σ + σ (A B ) 1 −1B1 ⎛1− 1 ⎞ f ′ (x ) 1 −1B1 ⎤ dx
line commences from the ceiling of tunnels and converges to a specific
=V ∫0 ⎢ t1 c1 1 1 1
⎜
⎥
⎟

⎣ ⎝ B1 ⎠ ⎦
point in the symmetrical axis, which is attributed to the arching effect n
Li
⎡−σ + σ (A B ) 1 −1Bi ⎛1− 1 ⎞ f ′ (x ) 1 −1Bi ⎤ dx
in deep rock strata. At incipient failure, the internal energy produced by + V∑ ∫L ⎢ ti ci i i i
⎜
⎥
⎟

the internal forces would dissipate along the velocity discontinuous i=2
i−1
⎣ ⎝ Bi ⎠ ⎦ (7)
surface, and its density results
where Ai , Bi , σti, σci are the corresponding rock parameters in the ith
1 1 layer, fi′ (x ) is the first derivative of function fi (x ) , i = 1, 2, n. Similarly,
wḋ = σn εṅ + τn γṅ =
( )1
−σt + σc (AB ) 1 − B 1− B f ′ (x ) 1 − B V
· the rate of work done by rock weight is expressed through summation
1 + f ′ (x )2 w (1) of the elementary rates within each layer, and it results
n
where σn and τn are the normal and shear stress, εṅ and γ̇n refer to the L1 Li
plastic normal and shear strain rate, σc and σt are the uniaxial com-
Ẇγ = V ∫0 γ1 [f1 (x )−g (x )] dx + V ∑ ∫L γi [fi (x )−g (x )] dx
i=2
i−1 (8)
pressive strength and tensile strength of the intact rock pieces, A and B
are dimensionless material constants for characterizing the mechanical where γi is the unit rock weight in the ith layer, and g (x ) represents the
properties of the rock mass, f ′ (x ) is the first derivative of f (x ) , V profile of the arbitrary cross-section of tunnel circumference.
corresponds to the falling velocity at failure, and w is the thickness of The objective function is established based on the rates of work
the plastic detaching zone. Due to the symmetry of the problem, only a done by the internal and external forces, in the form of
half collapsing block is accounted for. Through integral calculation, the L1
n
L
total dissipation rate over the interval of [0, L] gives ζ = Wḋ −Ẇγ = V ∫0 ψ1 [x ,f1′ (x ),f1 (x )] dx + ∑ V ∫L i ψi [x ,fi′ (x ),fi (x )] dx
i−1
i=2
⎡−σt + σc (AB ) 1 −1 B ⎛1− 1 ⎞ f ′ (x ) 1 −1 B ⎤ dx
L
Wḋ = V ∫0 ⎝ B⎠ (2) =V ∫0
L1
⎡−σ + σ (A B ) 1 −1B1 ⎛1− 1 ⎞ f ′ (x ) 1 −1B1 −γ f (x ) + γ g (x ) ⎤ dx
⎣ ⎦ ⎢ t1 c1 1 1 1
⎜
1 1 ⎟
1 ⎥
⎣ ⎝ B1 ⎠ ⎦
The rate of work produced by rock weight per unit length of the n
∑ V ∫L i ⎡⎢−σti + σci (Ai Bi) 1−Bi ⎛1− 1 ⎞ fi′ (x ) 1−Bi −γi fi (x )
L 1 1
detaching surface results + ⎜ ⎟

i=2
i−1
⎣ B ⎝ i⎠
ẇ γ = γ [f (x )−g (x )]·V (3)
+ γi g (x ) ⎤ dx
Accordingly, the difference of rates at impending collapse is defined ⎥ (9)
⎦
as
where
⎡−σt + σc (AB ) 1 −1 B ⎛1− 1 ⎞ f ′ (x ) 1 −1 B −γf (x )
L
ζ = Wḋ −Ẇγ = V ∫0 1 1 1
⎣ ⎝ B⎠ ψi = −σti + σci (Ai Bi ) 1 − Bi ⎛1− ⎞ fi′ (x ) 1 − Bi −γi fi (x ) + γi g (x ) i = 1,2,…,n
⎜ ⎟

⎝ Bi ⎠
+ γg (x ) ⎤ dx
⎦ (4) (10)

With the use of variational principle, the Euler–Lagrange equation In order to make the Eq. (9) reach a stationary value, the problem
of the problem is obtained based on the first variation of the difference could be resolved with the use of the variation principle. The first
of rates equal to zero. This gives the general form of the detaching line variation of the overall work rates yields
as ∂ψi ∂ ∂ψi ⎤
δψi [fi (x ),fi′ (x ),x ] = 0 ⇒ − ⎡ = 0 i = 1,2,…,n
(1 − B )/ B 1 ∂fi (x ) ∂x ⎢ ∂f ′ ⎥
⎣ i (x ) ⎦ (11)
1 γ
⎛⎜x + c1 ⎟⎞ + c2
B
f (x ) = A− B ⎛ ⎞ ⎜ ⎟

⎝ σc ⎠ ⎝ γ⎠ (5) On the basis of Eq. (10), the above equation is explicitly written as
where c1 and c2 are two integral constants which can be implicitly 1 1 2Bi − 1
σci (Ai Bi ) 1 − Bi f ′ (x ) 1 − Bi fi″ (x )−γi = 0
determined from boundary conditions. The work rate-based balance 1−Bi i (12)
equation aids to finally obtain the optimal kinematic solutions. Thereof,
the collapsing region of rectangular tunnels characterized by potential It is noted that there are n Euler–Lagrange equations for the whole
falling height and half-width yields falling block. Stepwise optimization of each equation ensures the op-
timization of the objective function. Notice that each Euler–Lagrange
1+B AB −B (1 + B ) BσtB equation presents a similar form but with differing rock properties.
H= σt , L =
γB γσcB − 1 (6) Therefore, the general form to characterize the detaching curve is valid
for the sub-detaching line, which yields
Eq. (5) gives a general form of the detaching curve, as a preliminary
1
study in a homogeneous/uniform rock mass, which will be extended to −1
(1 − Bi)/ Bi
⎛ γi ⎞
B
other cases discussed later. fi (x ) = Ai Bi ⎛⎜x + c2i − 1 ⎟⎞ i + c i = 1,2,…,n
⎜ ⎟
2i
σ
⎝ ⎠ci ⎝ γ i ⎠ (13)
3.3. Multi-layer analysis where c2i−1 and c2i are integral constants.
Accordingly, the first derivative of fi (x ), fi′ (x ) gives
Based on the preceding analysis, the similar procedure could be
1 − Bi
applied to the multi-layer kinematic analysis. In this case, the failure (1 − Bi)/ Bi
−1
⎛ γi ⎞
Bi
fi′ (x ) = Bi−1 Ai Bi ⎛⎜x + c2i − 1 ⎟⎞ i = 1,2,…,n
mechanism is composed of n layers. The work rate calculations are ⎜ ⎟

⎝ σci ⎠ ⎝ γi ⎠ (14)
presented based on the elementary rates of work.
There are n detaching curves to distinguish the removable blocks In order to determine c2i−1 and c2i, the following boundary

156
C. Qin, S.C. Chian Tunnelling and Underground Space Technology 73 (2018) 154–161

conditions are introduced; 4. Extended analysis in 3D framework

⎧ f (x = 0) = −H The preceding 2D analysis is further extended to three dimensions.

⎪ 1
f (x = Li − 1) = fi (x = Li − 1) = −ni − 1 H i = 2,3,…,n For ease of calculation, the following 3D kinematic analysis is limited to
⎨ i−1 deep tunnels with rectangular cross-sections. Likewise, the similar
⎪ fn (x = Ln ) = 0 (15)
⎩
procedure is presented to predict the dimension of potential failure
and block, considering a 3D failure mechanism in one and multi-layer rock
surrounds.
⎧ f1′ (x = 0) = 0
⎨ fi′− 1 (x = Li − 1) = fi′ (x = Li − 1) i = 2,3,…,n (16)
⎩ 4.1. 3D failure mechanism

where ni−1 is the ratio of the failure height at x = Li−1 to the overall
Considering that the actual collapse is three-dimensional, 2D ana-
collapse height H. Substituting Eqs. (13) and (14) into Eqs. (15) and
lysis may be unreliable in predicting tunnel crown stability. Therefore,
(16), the above equations can be expressed with the following unknown
an extension from 2D to 3D is necessitated. A 3D failure mechanism is
variables, L1, L2 … Ln and H only.
proposed for achieving this specific purpose. The basic thought for
Similarly, combining Eqs. (13) and (14) with Eq. (10) results in
generating a potential 3D collapse mechanism is by rotating the 2D
1 curved mechanism. As stated earlier, the planar detaching line is con-
Bi − 1
−1 1
c2i − 1 ⎞ Bi stituted by multi-curves, and in each layer a unique velocity dis-
ψi = −σti−Bi−1 σci Bi Ai Bi
γi Bi ⎜⎛x + ⎟ −γi c2i + γi g (x ) i = 1,2,…,n
⎝ γi ⎠ continuity is suited. The whole 2D failure mechanism is formed by
(17) connecting the adjacent discontinuities. Based on this, rotating the 2D
curves with respect to the Z symmetric axis by 360° produces a kine-
Accordingly, Eq. (9) can be re-written as matically admissible velocity field in three dimensions, from which one
can find that the falling velocity is kept constant. It is sensible to employ
1 + B1
⎡ − 1
B1− 1 1 B1 ⎤ such a mechanism to give an approximation to the actual failure on-site,
1 c
ζ = V ⎢ (−σt1−γ1 c2) L1−A1 B1 σc1 B1 γ1B1 ⎜⎛L1 + 1 ⎟⎞ ⎥ although some limitations exist. Such an approximation is meaningful
⎢ 1 + B1 ⎝ γ1 ⎠ ⎥
⎣ ⎦ and limited to rectangular tunnels with a horizontal ceiling, owing to
L1
n
Li the symmetry of geometry and loadings. However, for other profiles of
+V ∫0 γ1 g (x ) dx + V ∑ ∫L i−1
γi g (x ) dx cross-sections, it is rather complicated to accurately calculate the rates
i=2
of work. Therefore, the following 3D kinematic analysis is suited to
n ⎧ rectangular tunnels, and the proposed collapse mechanism is shown in
+ V∑ (−σti−γi c2i )(Li −Li − 1) Fig. 2.
i=2
⎨
⎩
1 + Bi 1 + Bi
−1 1
Bi − 1 1 ⎡
c Bi c Bi ⎤ ⎫ 4.2. Single-layer analysis
−Ai Bi
σci Bi γi Bi ⎢ ⎜⎛Li + 2i − 1 ⎟⎞ −⎜⎛Li − 1 + 2i − 1 ⎟⎞ ⎥
1 + Bi ⎢⎝ γi ⎠ ⎝ γi ⎠ ⎥⎬
⎣ ⎦⎭ Prior to 3D kinematic analysis in non-uniform rock masses, an
(18) analysis considering a single layer is first presented. The intrinsic dif-
ference between 2D and 3D analyses lies in the employment of collapse
For the case of 2D analysis with n layers, a total of 3n + 1 variables mechanism, leading to different calculations in work rates. With the use
are involved and required to be determined for a specific problem. of nonlinear HB failure criterion expressed by normal and shear
Notice that, however, there are only 3n boundary equations in Eqs. (15) stresses, the total rates of work done by the internal forces could be
and (16). One more equation is necessitated to render the impending derived, by integrating the dissipation density as shown in Eq. (1) over
collapse unique. At failure, the rate of internal energy dissipation is no interval [0, L].
less than that of the external forces, implying that there exist numerous
kinematic solutions satisfying the upper bound theorem. For the critical
failure under the ultimate limit state, the optimal kinematic solution is
derived under ζ = 0 , i.e.

1 + Bi
n ⎧ −1 1
Bi − 1 1 ⎡
c Bi
∑ (−σti−γi c2i )(Li −Li − 1)−Ai Bi
σci Bi γi Bi ⎢ ⎜⎛Li + 2i − 1 ⎟⎞
i=2
⎨ 1 + Bi ⎢⎝ γi ⎠
⎩ ⎣
1 + Bi
c Bi ⎤ ⎫
−⎜⎛Li − 1 + 2i − 1 ⎟⎞ ⎥ −(σt1 + γ1 c2) L1
⎝ γi ⎠ ⎥⎬
⎦⎭
1 + B1
B −1
− 1 1 1 1
c B1 L1
−A1 B1 σc1 B1 γ1B1 ⎜⎛L1 + 1 ⎟⎞ + ∫0 γ1 g (x ) dx
1 + B1 ⎝ γ1 ⎠
n
L
+ ∑ ∫L i γi g (x ) dx = 0
i−1 (19)
i=2

Therefore, the above 3n + 1 equations including Eq. (19) make the

whole problem statically determinate. The specific upper bound solu-
tions can be obtained by solving the above simultaneous equations. The
Fig. 2. 3D collapse mechansim of tunnels deeply buried in non-uniform rock strata.
numerical results will be presented in the later section.

157
C. Qin, S.C. Chian Tunnelling and Underground Space Technology 73 (2018) 154–161

L n
Wḋ = ∫0 wḋ ·2πx 1 + f ′ (x )2 dx ζ = Wḋ −Ẇγ = 2πV ∫0
L1
ψ1 dx + 2π ∑ V ∫L
Li
ψi dx
i−1
i=2
L 1 1 1
= 2πV ∫0 ⎡−σt + σc (AB ) 1 − B ⎛1− ⎞ f ′ (x ) 1 − B ⎤ xdx L1
⎡−σ x + σ (A B ) 1 −1B1 ⎛1− 1 ⎞ f ′ (x ) 1 −1B1 x + 0.5γ x 2f ′ (x ) ⎤ dx
⎣ ⎝ B⎠ ⎦ (20) = 2πV ∫0 ⎢ t1 c1 1 1 1
⎜ ⎟
1 1 ⎥
⎣ ⎝ B1 ⎠ ⎦
The definition of the corresponding parameters is interpreted si- n
Li
⎡−σ x + σ (A B ) 1 −1Bi ⎛1− 1 ⎞ f ′ (x ) 1 −1Bi x
milar as before. For the applied loads, the rate of work produced by + 2π ∑ V ∫L ⎢ ti ci i i ⎜
i
⎟

rock weight gives i=2

i−1
⎣ ⎝ Bi ⎠

0 + 0.5γi x 2fi′ (x ) ⎤ dx
Ẇγ = V ∫L γπx 2f ′ (x ) dx (21)
⎥
⎦ (28)
where
The objective function is formulated as the difference between Wḋ
and Ẇγ , i.e. 1 1 1
ψi = −σti x + σci (Ai Bi ) 1 − Bi ⎛1− ⎞ fi′ (x ) 1 − Bi x + 0.5γi x 2fi′ (x ) i = 1,2,…,n
⎜ ⎟

L 1 1 1
⎝ Bi ⎠
ζ = Wḋ −Ẇγ = 2πV ∫0 ⎡−σt x + σc (AB ) 1 − B ⎛1−
⎝ B⎠
⎞ f ′ (x ) 1 − B x (29)
⎣
In order to derive the analytical or semi-analytical solution, the
+ 0.5γx 2f ′ (x ) ⎤ dx variational principle gives the first variation of the function
⎦ (22)
ψi [fi (x ),fi′ (x ),x ] as
Following the same procedure as presented earlier, the general form
∂ψi ∂ ∂ψi ⎤
of the detaching curve f(x) in two dimensions gives δψi [fi (x ),fi′ (x ),x ] = 0 ⇒ − ⎡ = 0 i = 1,2,…,n
∂fi (x ) ∂x ⎢ ∂f ′ ⎥
⎣ i (x ) ⎦ (30)
(1 − B )/ B 1

f (x ) =
1
A− B ⎛ γ ⎞
⎜ ⎟
⎛
⎜x +
c1 ⎞ B
⎟ + c2 By virtue of Eq. (29), Eq. (30) is refined as
⎝ 2σc ⎠ ⎝ γ⎠ (23)
Bi 2Bi − 1
1 1 1
σci (Ai Bi ) 1 − Bi ⎡ fi′ (x ) 1 − Bi + fi′ (x ) 1 − Bi fi″ (x ) x⎤−γi x = 0
This is the basic solution which will be used in the multi-layer ⎢
⎣ Bi 1−Bi ⎥
⎦ (31)
analysis. With considerations of boundary conditions and work rate
The first derivative of fi(x) gives the following form after the first-
balance equation, the height and half-width of the collapsing region
order integral.
yield
1 − Bi
Bi
1 + 2B 2A 1 + 2B B σtB ⎡ γi Bi c ⎤
H= σt , L = ⎛ ⎞ fi′ (x ) = ⎢ x + 2i − 1 ⎥ i = 1,2,…,n
γB γ ⎝ 2B ⎠ σcB − 1 (24) ⎢
1
x ⎥
⎣ 2σci (Ai Bi ) 1 − Bi ⎦ (32)
Based on the formation principle for a 3D collapse mechanism, the where c2i−1 (i = 1, 2, n) are integral constants. Considering that the
critical velocity detaching surface is expressed as shear stress component vanishes at the point of (x = 0, y = 0, z = −H),
(1 − B )/ B
it leads to c1 = 0 and the specific interpretation was presented in Fraldi
1 γ ⎞ 1
and Guarracino (2009). However, the other n − 1 constants are un-
f (x ,y ) = A− B ⎛ ⎜ ⎟ (x 2 + y 2 ) 2B −H
⎝ 2σc ⎠ (25) known and influenced by boundary conditions. For an incipient col-
lapse, one can find that the remaining c2i−1 values are non-zero, which
directly affects the integral calculation for fi (x ).
Note that fi (x ) cannot be directly derived by integrating fi′ (x ) , de-
4.3. Multi-layer analysis
pending on Bi. In this case, fi (x ) involves a hypergeom which is hardly
able to attain the specific expression, although numerical analysis is
The whole falling block is hypothetically formed by many compo-
possible with significant computational effort. Some assumptions on B
nents, and hence integration of elementary analyses is carried out to
may be made to derive the analytical solutions of the detaching sur-
encompass the whole problem. In the presence of velocity dis-
faces. Since B = 1.0 leads to a linear Mohr-Coulomb failure criterion
continuous surface, the surrounding rock under the coverage of the
from the Hoek-Brown, a nonlinear kinematic analysis is focused on with
surface and the ceiling of tunnels is taken to fall, while the remaining
B less than 1.0. For ease of calculation, a specific case with B = 0.5 is
rock surrounds are at rest. The overall dissipation rates along this
considered.
surface are obtained by summing up the dissipation density in Eq. (1)
As a consequence, it results
over different intervals.
γ1
L1
n
Li
⎧ σc1 A 2 x i=1
1
Wḋ = ∫0 wḋ 1·2πx 1 + f1′ (x )2 dx + ∑ ∫L
i−1
̇ ·2πx 1 + fi′ (x )2 dx
wdi fi′ (x ) =
⎨ γi 2 x + c 2i − 1
i = 2,3,…,n
i=2 x
⎩ σci Ai (33)
L1
⎡−σ + σ (A B ) 1 −1B1 ⎛1− 1 ⎞ f ′ (x ) 1 −1B1 ⎤ xdx
= 2πV ∫0 ⎢ t1 c1 1 1 1
⎜
⎥
⎟ Thereafter, the detaching line fi (x ) is obtained through integration.
⎣ ⎝ B1 ⎠ ⎦
n γ 2
Li
⎡−σ + σ (A B ) 1 −1Bi ⎛1− 1 ⎞ f ′ (x ) 1 −1Bi ⎤ xdx ⎧ 1 2 x + c2 i=1
+ 2πV ∑ ∫L ⎢ ti ci i i i
⎜
⎥
⎟
fi (x ) =
⎪ σc1 A1 2
i=2
i−1
⎣ ⎝ Bi ⎠ ⎦ (26) ⎨ γi x 2 + c lnx + c i = 2,3,…,n
⎪ σci Ai2 2 2i − 1 2i
(34)
⎩
Likewise, the work rate due to gravity of falling blocks is also ex-
pressed as the summation of elementary rates. Substituting Eqs. (33) and (34) into Eq. (29) yields
2
n
⎧− σt1 x + 1 γ1
0 Li − 1 x3 i=1
Ẇγ = V ∫L 1
γ1 πx 2f1′ (x ) dx + V ∑ ∫L i
γi πx 2fi′ (x ) dx
ψi =
⎪ 4 A12 σc1
i=2 (27) ⎨ 1 γi
2 A 2 σci c 2
⎪− σti x + 4 Ai2 σci
x 3− i4 2ix− 1 i = 2,3,…,n
⎩ (35)
The work rate function is written as

158
C. Qin, S.C. Chian Tunnelling and Underground Space Technology 73 (2018) 154–161

Accordingly, potential collapse more accurately. This is also substantiated in Fig. 3

n where an average value is assigned to each rock parameter while other
L12
1 γ12 L14 ⎞
⎡−σ Li2−Li2− 1
ζ = 2πV ⎜⎛−σt1 + ⎟ + 2πV ∑
⎢ ti
parameters are kept as: A1 = 0.6 , A2 (A 4 ,A8 ) = 0.4 , B1 = 0.6,
⎝ 2 4 4 ⎠ A12 σc1 i=2 ⎣ 2 B2 (B4,B8) = 0.8, σc1 = 7.5 MPa , σc 2 (σc 4,σc8) = 12.5 MPa , γ1 = 17.5 kN/m3 ,
2 γ2 (γ4,γ8) = 22.5 kN/m3 and σti = σci/100 . The subscript i for each rock
1 γi Li4 −Li4− 1 Ai2 σci 2 L
+ 2
− c2i − 1ln ⎛ i ⎞ ⎤ ⎥
⎜ ⎟
parameter denotes the corresponding property in the ith layer. For this
4 Ai σci 4 4 ⎝ Li − 1 ⎠ ⎦ (36) group of rock parameters, the collapse height is approximately within
For impending collapse, the Greenberg minimum principle aids to the range of 6.3–14.2 m. A n value of 8 (i.e. 0.79–1.78 m thick layer) is
find the effective collapse mechanism by minimizing ζ . In this case, the adopted to divide the rock surrounds within the total collapsing region,
minimal ζ is achieved provided the work rate objective function equals and the results of 2D failure area are presented in the next section.
to zero, (i.e. a work rate balance equation), which gives
2 n 2
6. 2D solutions for collapsing region
2 2 4 4
L12 1 γ1 L14 ⎡−σ Li −Li − 1 + 1 γi Li −Li − 1
−σt1 + + ∑ ⎢ ti
2 4 A12 σc1 4 i=2 ⎣ 2 4 A 2
σ
i ci 4 In this section, the effective collapse mechanism and specific kine-
matic solutions considering non-uniform rock properties are discussed.
A 2 σci L
− i c22i − 1ln ⎛ i ⎞ ⎤⎜
⎥=0
⎟ Comparison on the failure region is carried out between layered and
4 ⎝ Li − 1 ⎠ ⎦ (37) homogeneous/uniform properties of rock masses. A linear variation in
Analogous to the 2D case, the similar boundary constraints are ap- rock properties, Ai , Bi , σti, σci and γi amongst different layers, is con-
plied to ensure the continuity of deformation and the first derivative of sidered for ease of calculation. It is also noted that a random value can
fi (x ) . Combining Eqs. (15) and (16) with Eq. (37), the optimal solution be assigned to the specific parameter in each layer instead; however,
of the effective failure mechanism is obtained. the impending collapse may not necessarily occur in certain combina-
Having obtained the 3n + 1 variables, the detaching curves can be tions of rock parameters.
explicitly expressed, and through rotation the detaching surfaces are An arbitrary tunnel profile can be studied based on the 2D analysis
written as discussed earlier. For ease of calculation the rectangular tunnel cross-
section is considered here, and the numerical solutions for depicting the
γ 2 2
⎧ 1 2 (x + y ) + c2 i=1 effective collapse mechanism are obtained, under the following given
⎪ σc1 A1 2
fi (x ,y ) = parameters: A1 = 0.6 , A8 = 0.4 , B1 = 0.6, B8 = 0.8, σc1 = 7.5 MPa ,
⎨ γi (x 2 + y 2 ) + c ln x 2 + y 2 + c i = 2,3,…,n
⎪ σci Ai2 2 2i − 1 2i σc8 = 12.5 MPa , γ1 = 17.5 kN/m3 , γ8 = 22.5 kN/m3 and σti = σci/100 .
⎩ (38)
Unlike a uniform rock mass where the collapsing height is identical
regardless of A value, the variation in dimensionless parameter A for
5. Comparison and discussion non-uniform rock masses has a significant effect on the impending
collapse. This is portrayed in Fig. 4 where an average A value is kept at
In order to validate the robustness of the proposed approach and the 0.5. Interestingly, the four velocity discontinuous lines tend to converge
kinematic solutions, comparison is carried out from different perspec- midway rather than at the symmetric axis in homogeneous rock. Higher
tives. Intuitively, the multi-layer analysis with the same rock properties A1 (lower A8) induces larger failure width and the opposite goes for the
in different layers is equivalent to the single-layer (2D and 3D) case, collapsing height. This increase in failure width is analogous to that in
yielding the identical solutions for the effective collapse mechanism, as the case of homogeneous rock mass.
those in Fraldi and Guarracino (2009) and Yang and Huang (2013). The effect of parameter B on the effective collapse mechanism is
Moreover, since the kinematic analyses considering non-uniformity of illustrated in Fig. 5 where B1 varies from 0.6 to 0.75 with the average
rock properties are performed based on the same procedure and general being 0.7. As observed, the overall failure height increases with in-
solution, the results obtained are deemed reliable. There is no necessity creasing B1, which is more pronounced under higher B1 values. In the
in presenting comparison results within homogeneous/uniform rock case of B1 less than B8, the collapse width decreases at a rate lower than
medium, since parameters such as Ai ,Bi,σti,σci and γi are constant. that of the increase in failure height. When compared with that of a
As stated earlier, the prerequisite for considering non-uniform rock uniform rock medium, the failure area is larger under B1 = 0.75 while
properties is the discretized method, which divides the whole falling the three other cases show a reverse consequence.
block into several components with different parameters assigned to Fig. 6 presents the influence of uniaxial compressive strength on the
each layer. The accuracy is highly related to the choice of n value. A effective collapse mechanism. Comparatively, the uppermost rock
larger n value implies thinner intervals, yielding more reliable solutions strength σc1 tends to have a greater influencing scope. As expected,
compared to lower n value. On the other hand, more computational greater values of the uniaxial compressive strength σc1 produce a larger
effort is required for the case of larger n. In engineering practice, rock collapsing region under the limit state. Such an increase is principally
masses may present obvious layered characteristics or gradual change reflected in the failure width. The use of homogeneous/uniform rock
in rock properties. Accordingly, a suitable n value is to be found to parameters yields larger collapse region for lower compressive strength
balance the engineering accuracy and computational requirement. It is in the upper layer, but smaller collapse for σc1 higher than the average.
also noted that the choice of n value is affected by the overall collapsing On the whole, the collapse height is over-estimated while the width is
height. A lower n value is suited for problems with lower failure height, under-estimated when the non-uniformity of rock parameters is ig-
and vice versa. nored. In the above calculations, the ratio of σti/ σci is kept as 1/100,
Table 1 shows the comparison results considering non-uniform rock although the effect of this ratio can be accounted for. As a strength
parameters within different layers where the ratio of σt / σc remains 1/ parameter providing resistance against collapse, the effect of σt is
100. As shown in Table 1, the parameters are varied linearly with the analogous to that of σc . It is logical since higher rock strength, σc or σt , is
same overall average. Compared with solutions sought from a uniform capable of resisting larger falling blocks at critical failure. A smaller
rock medium, significant discrepancies are observed in overall collap- failure area implies that such an incipient failure state can be readily
sing height and width. Moreover, the effective collapse mechanism is reached, and much more difficult for a larger scope. As presented in
influenced by the non-uniformity of rock properties, which will be Fig. 6, with increasing strength of σc1, the rock masses have lager re-
highlighted in the sensitivity analysis. The comparison also shows that sistance to prevent collapse, and hence a greater failure area is mobi-
the use of n = 4 (when compared to n = 2) tends to yield closer solu- lized to reach this critical state.
tions to those of n = 8, indicating that a higher n value predicts the As expected, greater γi values in the upper layer yield a lower

159
C. Qin, S.C. Chian Tunnelling and Underground Space Technology 73 (2018) 154–161

Table 1
Comparison of 2D failure area considering different layers for characterizing non-uniformity of rock masses.

A1 An B1 Bn σ1 σn γ1 γn L/m H/m

MPa MPa kN/m3 kN/m3 n=2 n=4 n=8 n=2 n=4 n=8

0.5 0.5 0.7 0.7 10 10 20 20 18.52 12.14

0.6 0.4 0.6 0.8 7.5 12.5 17.5 22.5 20.06 20.65 20.63 8.16 8.13 8.17
0.5 0.5 0.6 0.8 7.5 12.5 17.5 22.5 19.50 19.69 19.63 9.01 8.92 8.95
0.4 0.6 0.6 0.8 7.5 12.5 17.5 22.5 18.49 18.76 18.80 10.33 10.39 10.40
0.6 0.4 0.65 0.75 7.5 12.5 17.5 22.5 18.91 19.21 19.19 9.09 9.12 9.18
0.6 0.4 0.7 0.7 7.5 12.5 17.5 22.5 18.38 18.55 18.58 10.61 10.72 10.79
0.6 0.4 0.75 0.65 7.5 12.5 17.5 22.5 19.53 19.48 19.42 14.17 14.06 13.96
0.6 0.4 0.6 0.8 7.5 12.5 17.5 22.5 20.06 20.65 20.63 8.16 8.13 8.17
0.6 0.4 0.6 0.8 10 10 17.5 22.5 23.84 23.49 23.17 9.38 8.86 8.78
0.6 0.4 0.6 0.8 12.5 7.5 17.5 22.5 27.72 26.39 25.77 10.77 9.73 9.52
0.6 0.4 0.6 0.8 7.5 12.5 17.5 22.5 20.06 20.65 20.63 8.16 8.13 8.17
0.6 0.4 0.6 0.8 7.5 12.5 20 20 18.15 18.97 19.07 7.40 7.51 7.60
0.6 0.4 0.6 0.8 7.5 12.5 25 15 15.47 16.46 16.67 6.32 6.57 6.71

Fig. 6. Effective collapse mechanism under varying rock compressive strength.

Fig. 3. Comparison of the effective collapse mechanism considering different layers for
characterizing non-uniformity of rock masses.

Fig. 7. Effective collapse mechanism under varying rock unit rock weight.

Fig. 4. Effective collapse mechanism under varying A.

Fig. 5. Effective collapse mechanism under varying B.
collapse height and width, which is similar to that of a uniform rock
medium, and as substantiated in Fig. 7 where γi varying from 17.5 to
25 kN/m3. For the applied loads investigated, crown instability is in-
duced by rock weight during tunneling, and hence this destabilizing
force is the sole contributor to the impending collapse. Greater γi values
imply more applied forces exerted on the tunnel crown, thereby more
vulnerable to fail with a smaller failure area induced at critical state. It
is also found that the collapsing height is overly estimated if average
values of rock properties are used to represent non-uniform rock sur-
rounds.
It is also noted from Figs. 4 to 7 that the non-uniform rock para-
meters influence the shape and dimension of the potential falling
blocks. Due to the use of the nonlinear criterion, the final form of the
detaching curves is expressed nonlinearly, and hence the length of the
curve is not equally distributed, with the uppermost curve being the
highest percentage of the overall width.

160
C. Qin, S.C. Chian Tunnelling and Underground Space Technology 73 (2018) 154–161

Table 2 achieve this specific purpose with n layers considered with varying rock
Comparison of 3D collapsing region under 1 and 2 layers. parameters. Accordingly, the stability of tunnel roof is estimated in
single and multi-layered rock surrounds from the perspective of 2D and
n1 A2 σc2/MPa σt2/kPa γ2/kN·m−3 L1/m L2/m H/m L/m H/m
3D. Based on the solution obtained from single layer analysis, it can be
0.3 0.3 1.5 15 22.5 4.272 5.029 1.267 7.071 2 readily extended to multi-layer rocks in its general form without any
0.5 0.3 1.5 15 22.5 3.783 5.127 1.392 assumption. Employing the nonlinear HB failure criterion and varia-
0.7 0.3 1.5 15 22.5 3.143 5.248 1.601 tional principle, the closed-form solution of the effective collapse me-
0.5 0.2 1.5 15 22.5 3.243 4.120 1.023 5.657 2
chanism is derived based on the work rate balance equation.
0.5 0.3 1.5 15 22.5 3.783 5.127 1.392 6.364
0.5 0.4 1.5 15 22.5 4.458 6.322 1.932 7.071 Specific results are calculated under given parameters by resolving
0.5 0.3 0.5 15 22.5 2.376 2.953 0.549 3.536 1 those simultaneous equations. It is found that the non-uniformity of
0.5 0.3 1.0 15 22.5 3.057 4.010 0.909 5.303 1.5 rock properties has a significant impact on the shape and dimension of
0.5 0.3 1.5 15 22.5 3.783 5.127 1.392 7.071 2 the falling block. Moreover, the effect of each parameter on the im-
0.5 0.3 1.5 7.5 22.5 3.032 4.110 0.894 5 1
pending collapse is highly influenced by the complexity in the inter-
0.5 0.3 1.5 15 22.5 3.783 5.127 1.392 7.071 2
0.5 0.3 1.5 30 22.5 4.955 6.715 2.387 10 4 action between those parameters. On the whole, larger failure area
0.5 0.3 1.5 15 15 4.924 6.892 2.357 8.703 2.462 implies that tunnels have a larger capacity to be mobilized for stability,
0.5 0.3 1.5 15 20 4.082 5.584 1.620 7.542 2.133 which is achieved by increasing the resistance forces and decreasing the
0.5 0.3 1.5 15 25 3.536 4.751 1.216 6.655 1.882
applied forces. As a supplement, an extension of the kinematic analysis
combined with the discretized approach provides avenues to resolve
7. 3D solutions for potential failure area complicated problems, such as non-uniformity of rock properties which
cannot be readily resolved with existing conventional methodologies.
The specific procedure for 3D kinematic analysis was presented Moreover, other influential factors like pore water effect can also be
earlier. In order to derive the effective failure mechanism in closed accounted for with the proposed approach, which will be considered in
form, some assumptions on Bi are made for simplification. Bi = 1.0 a future publication.
implies a linear failure criterion and a straight line for the velocity
discontinuity, which is so straightforward and will be not discussed Acknowledgements
here. Instead the specific case with Bi = 0.5 is investigated in this sec-
tion with the nonlinear solutions obtained. The authors would like to thank the editor and the anonymous re-
Table 2 summarizes the numerical solutions of the collapsing height viewers for their valuable suggestions in improving this technical
and width under different parameters. As stated earlier, the choice of n paper. The financial support for the first author’s PhD study by the
value is dependent on the collapse height. According to the given rock Ministry of Education in Singapore is also greatly appreciated.
parameters in Table 2, the collapse height is roughly 2 m with the other
References
input parameters corresponding to A1 = 0.6 , σc1 = 0.5 MPa ,
γ1 = 17.5 kN/m3 and σti = σci/100 . In this case, n = 2 is selected to give
Bieniawski, Z.T., 1976. Rock mass classification in rock engineering. In: Bieniawski, Z.T.
each layer of about 1 m thickness. A linear variation is assumed with (Ed.), Exploration for Rock Engineering. Rotterdam, Netherlands.
the same average value for each parameter. As observed, the increase in Bieniawski, Z.T., 1989. Engineering Rock Mass Classifications. Wiley, New York.
A2 of the lower layer leads to an increase in failure height H and width Chen, W.F., 1975. Limit Analysis and Soil Plasticity. Elsevier, Amsterdam.
Fraldi, M., Guarracino, F., 2009. Limit analysis of collapse mechanisms in cavities and
L. It is also as expected that the increasing values of rock strength, σc or
tunnels according to the Hoek-Brown failure criterion. Int. J. Rock Mech. Min. Sci.
σt , induce greater collapse region. Specifically, an increase in σc (σt ) 46, 665–673.
from 0.5 to 1.5 MPa (7.5 to 30 kPa) leads to an increase in the overall Fraldi, M., Guarracino, F., 2010. Analytical solutions for collapse mechanisms in tunnels
collapse width by 73.6% (63.4%) and 153.6% (166.9%) for the collapse with arbitrary cross sections. Int. J. Solids Struct. 47, 216–223.
Fraldi, M., Guarracino, F., 2011. Evaluation of impending collapse in circular tunnels by
height. Nevertheless, the reverse outcome is observed for greater values analytical and numerical approaches. Tunn. Undergr. Space Technol. 26 (6),
of rock unit weight, implying that the crown stability is more suscep- 507–516.
tible under higher applied loads. Considering that two layers are em- Hoek, E., Brown, E.T., 1997. Practical estimate the rock mass strength. Int. J. Rock Mech.
Min. Sci. 34 (8), 1165–1186.
ployed for characterizing the non-uniform parameters, non-equiparti- Huang, F., Yang, X.L., 2011. Upper bound limit analysis of collapse shape for circular
tion of the collapse height can also be readily accounted for. The results tunnel subjected to pore pressure based on the Hoek-brown failure criterion. Tunn.
show that the greater failure region is proportional to the height ratio Undergr. Space Technol. 26, 614–618.
Huang, F., Yang, X.L., Ling, T.H., 2013. Prediction of collapsing region above deep
(n1) of the lower failure height to the whole height. This may be at- spherical cavity roof under axis-symmetrical conditions. Rock Mech. Rock Eng.
tributed to the interaction between those rock properties in the two Qin, C.B., Chian, S.C., Wang, C.Y., 2017. Kinematic analysis of pile behavior for im-
layers. provement of slope stability in fractured and saturated Hoek-Brown rock masses. Int.
J. Numer. Anal. Meth. Geomech. 41, 803–827.
For a homogeneous/uniform rock mass, the collapsing height and Qin, C.B., Chian, S.C., Yang, X.L., Du, D.C., 2015a. 2D and 3D limit analysis of progressive
width from 3D kinematic analysis can be readily derived with an ana- collapse mechanism for deep-buried tunnels under the condition of varying water
lytical expression. The solutions are calculated with average rock table. Int. J. Rock Mech. Min. Sci. 80, 255–264.
Qin, C.B., Sun, Z.B., Liang, Q., 2013. Limit analysis of roof collapse in tunnels under
parameters, as supplemented in Table 2. Through these comparisons, it
seepage forces condition with three-dimensional failure mechanism. J. Cent. South.
can be concluded that the latter calculations with average values of Univ. 20 (8), 2314–2322.
each parameter yield much larger H and L, which is deemed as over- Qin, C.B., Yang, X.L., Pan, Q.J., Sun, Z.B., Wang, L.L., Miao, T., 2015b. Upper bound
estimated. This implies that non-uniformity of rock properties should be analysis of progressive failure mechanism of tunnel roofs in partly weathered stra-
tified Hoek-Brown rock masses. Int. J. Rock Mech. Min. Sci. 74, 157–162.
considered in order to provide more reliable solutions. Whittaker, B.N., Reddish, D., 1989. Subsidence: Occurrence, Prediction and Control.
Elsevier, Amsterdam.
8. Concluding remarks Yang, X.L., Huang, F., 2013. Three-dimensional failure mechanism of a rectangular cavity
in a Hoek-Brown rock medium. Int. J. Rock Mech. Min. Sci. 61, 189–195.
Yang, X.L., Qin, C.B., 2014. Limit analysis of rectangular cavity subjected to seepage
A procedure is presented to investigate the crown stability of deep- forces based on Hoek-Brown failure criterion. Geomech. Eng. 6 (5), 503–515.
buried tunnels within the scope of plasticity theory. The key merit of Yang, X.L., Zhang, D.B., Wang, Z.W., 2013. Upper bound solutions for supporting pres-
sures of shallow tunnels with nonlinear failure criterion. J. Cent. South Univ. 20,
the proposed approach lies in the consideration of non-uniform rock 2034–2040.
properties in the kinematic analyses. The discretized method is used to

161