You are on page 1of 13

Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848

Contents lists available at ScienceDirect

Journal of Rock Mechanics and


Geotechnical Engineering
journal homepage: www.jrmge.cn

Full Length Article

Face stability analysis of circular tunnels in layered rock masses using the
upper bound theorem
Jianhong Man, Mingliang Zhou*, Dongming Zhang, Hongwei Huang, Jiayao Chen
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Department of Geotechnical Engineering, Tongji University, Shanghai,
200092, China

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

Article history: An analysis of tunnel face stability generally assumes a single homogeneous rock mass. However, most
Received 11 July 2021 rock tunnel projects are excavated in stratified rock masses. This paper presents a two-dimensional (2D)
Received in revised form analytical model for estimating the face stability of a rock tunnel in the presence of rock mass stratifi-
3 October 2021
cation. The model uses the kinematical limit analysis approach combined with the block calculation
Accepted 21 December 2021
Available online 23 February 2022
technique. A virtual support force is applied to the tunnel face, and then solved using an optimization
method based on the upper limit theorem of limit analysis and the nonlinear HoekeBrown yield cri-
terion. Several design charts are provided to analyze the effects of rock layer thickness on tunnel face
Keywords:
Face stability
stability, tunnel diameter, the arrangement sequence of weak and strong rock layers, and the variation in
Rock tunnel rock layer parameters at different positions. The results indicate that the thickness of the rock layer,
Layered rock masses tunnel diameter, and arrangement sequence of weak and strong rock layers significantly affect the tunnel
Upper bound solution face stability. Variations in the parameters of the lower layer of the tunnel face have a greater effect on
HoekeBrown criterion tunnel stability than those of the upper layer.
Ó 2022 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by
Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

1. Introduction Executive, 1996). When the tunnel face collapses, a large amount
of rock mass flows into the tunnel, resulting in many casualties,
Layered rock masses are typical natural geological settings, ac- property losses, damage to tunnel support structures (Wang et al.,
counting for approximately 66.7% of the land area. Hence, various 2019; Huang et al., 2021), construction period delays, and other
engineering construction activities often encounter layered rock issues. Therefore, face stability analysis is of great importance for
masses. A growing number of rock tunnel projects involve complex tunnel engineering in layered rock masses.
geological environments (Anagnostou et al., 2014; Chen et al., 2021; Over the past several decades, the stability of tunnel faces in
Zhao et al., 2021), most of which have tunnel faces excavated in homogenous rock mass/soil has been extensively analyzed by
layered rock masses (Zhang and Zhou, 2017; Zhang et al., 2021a). experimental tests, numerical simulations, and analytical solutions.
Several studies have suggested that tunnel face failure, especially Many scholars have studied the failure shape and mechanism of
collapse, is often inextricably linked to the stratification of the tunnel faces through the centrifuge model (Idinger et al., 2011;
layered rock masses (Arnáiz Ronda et al., 2003; Babendererde et al., Chen et al., 2018; Weng et al., 2020) and 1-g physical model (Kirsch,
2006; Anagnostou and Zingg, 2013; Zhou et al., 2021). Tunnel 2010; Chen et al., 2015; Liu et al., 2018). While experimental tests
collapse accidents in China accounted for about 55.6% of con- can accurately simulate the failure phenomenon of a tunnel face, it
struction accidents from 2006 to 2016, most of which were due to is difficult to test the stability of a tunnel face under complex
the instability of the tunnel face (Zhang et al., 2018). Construction geological conditions. Three types of numerical simulation method
experience indicates that most tunnel collapse accidents can be are widely used in the stability analysis of tunnel faces: the finite
attributed to the instability of the tunnel face (Health and element method (FEM) (Vermeer et al., 2002; Li et al., 2009; Kim
and Tonon, 2010; Ukritchon et al., 2017), discrete element
method (DEM) (Maynar and Rodriguez, 2005; Chen et al., 2011;
* Corresponding author. Jiang and Yin, 2014; Zhang et al., 2021b), and FEM-DEM coupling
E-mail address: zhoum@tongji.edu.cn (M. Zhou). method (Long and Tan, 2020; Yin et al., 2020). These numerically-
Peer review under responsibility of Institute of Rock and Soil Mechanics, Chi- based methods can simulate complex stratum environment and
nese Academy of Sciences.

https://doi.org/10.1016/j.jrmge.2021.12.023
1674-7755 Ó 2022 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848 1837

changes in monitoring objects during the construction process. and is extensively used for rocks with varying degrees of fracture.
However, using any of these numerical methods to conduct the face The generalized HoekeBrown criterion reported by Hoek and
stability analysis of a typical tunnel setting comes at a high Brown (2019) can be written as follows:
computational cost, and takes a long time to complete the  n
simulation. s3
s1 ¼ s3 þ sc mb þs (1)
In contrast, an analytical solution for face stability analysis can sc
quickly produce design charts and can conveniently be used on-
site. Analytical solutions typically have two types: the limit equi- where s1 and s3 are respectively the maximum and minimum
librium method and the upper bound solution. The wedge model principal stresses at failure, and sc is the uniaxial compressive
(Horn, 1961), wedge-prism model (Perazzelli et al., 2014; Paternesi strength (UCS) of intact rock. The HoekeBrown parameters (mb, s
et al., 2017; Zou et al., 2019; Zhang et al., 2020a), and triangular base and n) are determined by the rock material constant mi, the
prism model (Oreste and Dias, 2012) are common models using the disturbance coefficient d, and the geological strength index GSI by
limit equilibrium method, which do not consider the constitutive the following formulations:
law of the material and have low computational accuracy. There-  
fore, the upper bound solution has been gradually adopted as an GSI  100
mb ¼ mi exp (2a)
efficient method for tunnel face stability analysis, which has a va- 28  14d
riety of hypothesized face failure mechanisms, e.g. the multi-block
 
failure mechanism (Chen, 1975; Davis et al., 1980; Leca and GSI  100
s ¼ exp (2b)
Dormieux, 1990; Subrin and Wong, 2002; Mollon et al., 2010, 9  3d
2011a, b; Zou et al., 2019; Li et al., 2020; Zhou et al., 2020). Most
current studies have focused on the face stability in homogeneous     
1 1 GSI 20
geomaterials. Only a few attempts have been made to account for n¼ þ exp   exp  (2c)
2 6 15 3
the partial collapse of soil and the spatial variability of soil-based
stratum (Pan and Dias, 2018a, b; Zou and Qian, 2018; Li and Yang,
2020; Zhang et al., 2021c, d). To the authors’ knowledge, no
analytical solutions have been proposed for the face stability 2.2. Equivalent MohreCoulomb parameters
analysis of tunnels in layered rock masses during excavation. Thus,
this study proposes an upper bound-based analytical approach to In the upper bound solution, the internal energy dissipation
conduct the face stability analysis of circular tunnels in layered rock along the discontinuous surface is often calculated by the Mohre
masses. The HoekeBrown yield criterion was adopted to better Coulomb parameters of cohesion c and frictional angle 4. Howev-
represent the rock mass failure behavior, and the equivalent Mohre er, these two parameters are not used in the HoekeBrown failure
Coulomb model parameters were obtained for the analytical solu- criterion. Hence, it is necessary to perform equivalent conversions
tion. The geological strength index, unconfined compressive between the HoekeBrown parameters and the MohreCoulomb
strength, and rock material constants vary with the properties of parameters.
each rock layer, resulting in segmented spiral failure surfaces. The There are two main ways to convert the HoekeBrown param-
proposed model is composed of multiple logarithmic spiral failure eters into the MohreCoulomb parameters (cohesion c and fric-
surfaces. tional angle 4). One of such methods is the tangent method, which
The rest of the paper is organized as follows. The nonlinear can linearize the nonlinear failure criterion and obtain the equiv-
failure criterion of the rock masses and parameter equivalent alent parameters (Pan and Dias, 2018a; Zhang et al., 2020b). The
method are first presented in Section 2. Section 3 shows the deri- other is the direct equivalent method, which incorporates engi-
vation details of the proposed analytical solution. In Section 4, the neering experience, and has proven highly reliable when tested in
analytical approach is validated by comparison with the numerical practice (Wang et al., 2021). The direct equivalent method can also
calculation results in layered rock cases and other analytical solu- reflect the variation of parameters with the depth of the tunnel so
tions. In Section 5, the analytical approach is used to conduct a that the computed result is more representative of the site condi-
comprehensive stability analysis of circular tunnel faces in layered tion (Hoek et al., 2002). Therefore, this study uses the direct
rock masses with different geometries and material properties. equivalent method to derive the upper bound solution.
Finally, the major findings of this study and possible future studies The equivalent cohesion c and frictional angle 4 proposed by
are presented in Section 6. The calculation results obtained in this Hoek et al. (2002) are expressed as follows:
paper may be used as a reference for future tunnel excavation
design. sc ½ð1 þ 2nÞs þ ð1  nÞmb s3n ðs þ mb s3n Þn1
c¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h i. ffi
2. Nonlinear failure criterion ð1 þ nÞð2 þ nÞ 1 þ 6mb nðs þ mb s3n Þn1 ½ð1 þ nÞð2 þ nÞ 

2.1. Generalized HoekeBrown criterion (3a)


" #
Rock masses contain discontinuities of various sizes and orien- 6mb nðs þ mb s3n Þn1
tations at different scales. Many geotechnical experiments have 4 ¼ arcsin (3b)
2ð1 þ nÞð2 þ nÞ þ 6mb nðs þ mb s3n Þn1
shown that almost all rock materials exhibit nonlinear character-
istics at failure (Shen et al., 2012; Zhao et al., 2017). The well- where s3n ¼ s3max =sc ; in which the parameter s3max is defined as
accepted MohreCoulomb criterion cannot adequately describe follows:
the nonlinear failure mechanism (Pan and Dias, 2018b). Hoek and
Brown (2019) proposed the generalized HoekeBrown criterion,  0:94
s3max scm
which can represent the nonlinear characteristics of rock failure ¼ 0:47 (4)
scm gH
1838 J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848

where g is the unit weight of the rock mass; H is the buried depth of logarithmic spiral collapse model in a homogeneous rock mass
the tunnel, being the distance between the top of the tunnel and configuration. Combining the double-logarithmic spiral with the
the ground surface; and scm is the rock mass strength, which can be approach of considering a multi-layer structure with progressive
written as failure (Qin et al., 2017), this paper establishes a progressive
collapse model of layered rock masses under two-dimensional (2D)
sc ½mb þ 4  nðmb  8sÞðmb =4 þ sÞn1 conditions. The failure surfaces thus consist of multiple double-
scm ¼ (5)
2ð1 þ nÞð2 þ nÞ logarithmic spirals, which can account for the failure of different
rock layers according to their material properties (Fig. 1). The failure
It is noteworthy that the equivalent method mentioned above surfaces of the proposed solution are composed of four logarithmic
does not consider the stratification of rock masses, especially in the spirals. From a mathematical geometry perspective, each loga-
solution process of Eq. (4). Layered rock masses have always been a rithmic spiral is determined by two angles with respect to the
complicated problem, thus certain simplifications are needed in the vertical direction (the starting and ending angles). However,
analytical solution. The simplified method is performed by comparing the upper and lower boundaries of the failure area, the
assuming that the properties of all rock layers are the same when angles corresponding to points A and C (relative to the vertical di-
solving the equivalent parameters of a specific rock layer. rection) can have two possible cases: q2 < q3 and q3 < q2. Therefore,
Hoek et al. (2002) converted the HoekeBrown parameters to the two collapse models are proposed, i.e. Cases 1 and 2, as shown in
MohreCoulomb parameters c and 4 using curve fitting, which Fig. 1.
adopts the principle of balancing the areas above and below the
MohreCoulomb plot. Therefore, the equivalent MohreCoulomb
parameters inevitably have calculation differences. However, the
3.2. Limit analysis and calculation method in layered rock mass
exhibited calculation difference has been proven acceptable in a
previous study (Wang et al., 2021).
According to Fig. 1, the diameter of the circular tunnel is denoted
as D, the thickness of the overburden layer is denoted as H, the
3. The upper bound solution in layered rock masses thickness of the upper layer of the tunnel face is denoted as h1, and
the thickness of the lower layer of the tunnel face is denoted as h2.
3.1. Analytical model description Based on the proposed failure models in Fig. 1, there are four log-
arithmic spiral surface lines: BC, CD, DE, and AE. For the failure
The upper bound solution combines the kinematically admis- mechanism, each failure surface is assumed to rotate with a con-
sible velocity field with the corresponding yield conditions, flow stant angular velocity u around point O, where the lengths of OA,
rules, and boundary conditions to study the stability of geotech- OB, OC, and OD are ra, rb, rc and rd, respectively. Meanwhile, the
nical structure. It is difficult to analyze the tunnel face stability due angles between OA, OB, OC, OD, and OE and the vertical direction are
to the complex mechanical behavior of layered rock mass. However, denoted as q2, q1, q3, q4, and q5, respectively. Based on the associated
Dong and Anagnostou (2013, 2014) and Anagnostou et al. (2014) flow rule of the upper bound theory of limit analysis, at any point
pointed out that in most cases, the complexity of mechanical on the logarithmic spiral failure lines, the frictional angle of each
behavior does not increase the complexity of engineering. From a layer is the same as the angle between the velocity direction and
practical point of view, the simplified model is still appropriate for the corresponding tangent direction of the failure lines. 4ti and
engineering purposes. Consequently, this paper uses simplified (i ¼ 0, 1, 2) are denoted as the frictional angle and the cohesion of
continuum modeling to consider the stratification of rock mass in the overburden layer, the upper layer of the tunnel face, and the
front of the tunnel face (Xu et al., 2017). The following assumptions lower layer of the tunnel face, respectively. As shown in Fig. 2, the
were made to simplify the geological and mechanical complica- logarithmic spiral curves AE, BC, CD, and DE can be written as
tions of the problem: follows:

(1) The stratum consists of three rock mass layers: the over- AE : r1 ðqÞ ¼ ra exp½ðq  q2 Þtan 4t0  (6a)
burden layer, the upper tunnel face layer, and the lower
tunnel face layer. The exposed tunnel face is assumed to lie BC : r2 ðqÞ ¼ rb exp½ðq1  qÞtan 4t2  (6b)
within the upper and lower tunnel face layers.
(2) The excavated tunnel face within the upper and lower rock
layers is assumed to have a perfect circular shape.
CD : r3 ðqÞ ¼ rc exp½ðq3  qÞtan 4t1  (6c)
(3) The rock layer is assumed to be an ideal elastoplastic material
that conforms to the associated flow rule. Each rock layer is DE : r4 ðqÞ ¼ rd exp½ðq4  qÞtan 4t0  (6d)
homogeneous, and the dilation angle of each point along the
sliding surface is equal to the equivalent frictional angle of where q is a variable denoting the rotational angle. According to the
the rock medium. geometric relationship in Fig. 1, ra, rb, rc and rd can be expressed as
(4) The failure surface of each rock layer is continuous on the the functions of q2, q1, q3 and q4 as follows:
layer interface, and the associated flow rule also conforms to
the layer interface. sin q1 ðh1 þ h2 Þ
ra ¼ (7a)
(5) Each hypothesized sliding block is regarded as a rigid body, sinðq2  q1 Þ
and the strain within each block is not considered so that
energy dissipation only occurs on the failure surface. sin q2 ðh1 þ h2 Þ
rb ¼ (7b)
sinðq2  q1 Þ
To better analyze the influence of the stratification character-
istics of rock masses on the stability of rock tunnel face, an sin q2 cos q1 ðh1 þ h2 Þ h2
improved failure model is proposed, with two cross-layers and one rc ¼  (7c)
sinðq2  q1 Þcos q3 cos q3
cover layer. Subrin and Wong (2002) proposed a double-
J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848 1839

Fig. 1. Improved failure model composed of multiple double-logarithmic spirals: (a) Parameters of each rock strata; and (b, c) Longitudinal sections of the failure surfaces in Case 1
(q2 < q3) and Case 2 (q3 < q2), respectively.

Fig. 2. 2D calculation models of (a) Case 1 (q2 < q3) and (b) Case 2 (q3 < q2).

dissipation in the system (Chen, 1975). For the calculation model


sin q1 cos q2 shown in Fig. 2, three aspects are included in this study: the work
rd ¼ ðh þ h2 Þ (7d)
sinðq2  q1 Þcos q4 1 rate of gravity, the work rate of the virtual support force, and the
work rate of internal energy dissipation along the failure lines.
The parameter q5 can also be established:
Firstly, the work rate of the self-weight of the multiple double-
 
1 lnðcos q4 =cos q2 Þ logarithmic spiral failure mechanisms is obtained from the four
q5 ¼ q2 þ q4  (8) sliding blocks (EGD, AGD, BFC, and AFCD). We assume that the
2 tan 4t0
length of is rf and the angle between OF and the vertical direction is
A rock tunnel face retains self-stability after excavation if the q11. Meanwhile, the intersection of OC and AD (Case 1) or AF (Case 2)
rock mass quality is good and the degree of jointing is not severe. is point I, and the intersection of OD and the logarithmic spiral line
However, when a tunnel is excavated in heavily jointed and “low- AE is point G. The work rate of gravity of each part can then be
quality” rock masses, the original equilibrium state is broken. The calculated.
rock mass in front of the tunnel face cannot be restored to a stable The work rate of the self-weight of regions EGD and AGD can be
condition through stress redistribution, and tunnel collapse may written as follows:
occur. Hence, it is necessary to take pre-supporting measures for
the tunnel face, such as advanced bolts, advanced small steel pipes,  Zq5   
advanced small pipes, etc. This makes it necessary to determine in PEGD ¼ vG sin qdq ¼ ug0 rd3 f1  ra3 f2 (9a)
advance the magnitude of the force to be applied to the tunnel face. ½OEDOEG
q4
In this study, the virtual support force determines whether the
tunnel face is stable. We assume that the virtual support force sT is
uniformly distributed at the tunnel face. Then, based on the upper  Zq4 
bound theory of limit analysis, when the velocity boundary con- PAGD ¼ vG sin qdq ¼ ug0 ra3 ðf3  f4 Þ (9b)
dition is satisfied, the work rate of external forces applied to the ½DOADOAG
q2
system is equal to or greater than the work rate of internal energy
1840 J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848

where v and Gdq represent the tangent velocity at the center of The expressions of the parameters in the above equations are
gravity of the micro-element body and the gravity of the micro- shown in Appendix A. It is worth noting that, when calculating the
element body, respectively. work rate of a specific region, the parameters used are only related
According to the geometric relationship in Fig. 2, the auxiliary to the rock layer where the calculation region is located.
variables q11 and rf can be expressed as follows: Secondly, the work rate of the virtual support force on the
  tunnel face can be written as follows:
sinðq2  q1 Þ h1
q11 ¼ arccot cot q2 þ (10a)
sin q2 sin q1 h1 þ h2 Zq2  . 
PT ¼ sT vT lT cos qdq ¼ sT rb2 u 1  sin 2 q1 sin 2 q2 2 (15)
sin q2 sin q1 q1
rf ¼ ðh þ h2 Þ (10b)
sinðq2  q1 Þsin q11 1
where vT represents the tangent velocity of the micro-element
The work rate of the self-weight of region BFC can be written as body around point O for line AB, and lT dq represents the unit
follows: length of line AB.
Finally, the internal energy dissipation along the failure lines can
 Zq3    be represented by the energy dissipation along the failure surfaces:
PBFC ¼ vG sin qdq ¼ ug2 rb3 f5  rb3 f6  rc3 f7
½OBCDOBFDOFC
q1 Zq5
(11) PAE ¼ c0 vD cos 40 lD dq ¼
q2
For Case 1, the region AFCD is composed of regions IDC and AFCI,
for which the work rates of the self-weight are calculated as
follows: 1
uc0 ra2 fexpf2½ðq5  q2 Þtan 4t0 g  1g (16a)
2 tan 40
 Zq4   
PIDC ¼ vG sin qdq ¼ ug1 rc3 f8  rd3 f9 (12a) Zq5
½ODCDOID
q3 PED ¼ c0 vD cos 40 lD dq ¼
q4
 Zq11   Zq3  1
PAFCI ¼ vG sin qdq þ vG sin qdq uc0 rd2 f  expf2½ðq4  q5 Þtan 40  g þ 1 g (16b)
2 tan 40
½DOBF ½DOFC
q1 q11
 Zq2   Zq3  Zq4
1
 vG sin qdq  vG sin qdq PDC ¼ c1 vD cos 41 lD dq ¼ uc1 rc2 ,
½DOAB ½DOAI 2 tan 41
q1 q2 q3
 
¼ ug1 rc3 f10  ra3 f11  rb3 f12 f  exp½2ðq3  q5 Þtan 41  þ exp½2ðq3  q4 Þtan 41  g (16c)
(12b)
Zq3
PAFCD ¼ PAICD þ PIFC (12c) PCB ¼ c2 vD cos 42 lD dq ¼
For Case 2, the region AFCD is composed of regions IFC and AICD, q1
for which the work rates of the self-weight are calculated as 1
uc2 rb2 f  exp½2ðq1  q3 Þtan 42  þ 1 g (16d)
follows: 2 tan 42

 Zq3    where vD represents the tangent velocity of the logarithmic spiral


PIFC ¼ vG sin qdq ¼ ug1 rc3 f13  rf3 f14 (13a) micro-element body around point O, and lD represents the arc
½DOFCDOFI length of the logarithmic spiral micro-element body.
q11
Therefore, the total internal energy dissipation can be written as
follows:
 Zq4   Zq4 
PAICD ¼ vG sin qdq  vG sin qdq PD ¼ PAE þ PED þ PDC þ PCB (17)
½OCD ½DOAD
q3 q2 Equating the total work rate of external forces to the total in-
 Zq2    ternal energy dissipation rate, i. e, the following expression of vir-
 vG sin qdq ¼ ug1 rc3 f15  ra3 f16  ra3 f17 (13b) tual support force can be obtained:
½DOAI
q3 PG  PD
sT ¼  . . (18)
rb2 u 1  sin 2 q1 sin 2 q2 2
PAFCD ¼ PAICD þ PIFC (13c)
Therefore, the work rate of the self-weight of the multiple The virtual support force sT can be solved according to the upper
double-logarithmic spiral failure mechanisms can be obtained: bound theorem of limit analysis. As shown in the objective function
of Eq. (18), sT is a function of the variables q1, q2, q3, and q4 (other
PG ¼ PEGD þ PAGD þ PBFC þ PAFCD (14) variables can be represented by these four variables), namely, sT ¼ f
(q1, q2, q3, q4). The function can then be transformed into a
J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848 1841

mathematical optimization problem, which can search the best using the same parameters of the proposed method are compared
upper bound solution by optimizing the objective function of Eq. with the analytical results obtained by Senent et al. (2013) (see
(18) under the following constraints: Table 1). The differences in computed face stability caused by the
9 conversion of HoekeBrown parameters to their equivalent Mohre
0 < q1 < q11 < q2 < q3 < q4 < p=2 ðCase 1Þ >
> Coulomb parameters are shown in Table 1. Those results are ob-
>
0 < q1 < q11 < q3 < q2 < q4 < p=2 ðCase 2Þ >
>
> tained by the limit analysis and FLAC3D (fast Lagrangian analysis of
=
q4 < q5 < p continua in 3 dimensions) simulations performed by Senent et al.
(19)
ra < rf < rb >
> (2013). It is noteworthy that the simulation results of FLAC3D
>
>
rd < rc < rb >
>
; directly use the HoekeBrown parameters. There are differences
ra < rd between the analytical results in this paper and those given by
Several efficient nonlinear optimization methodologies have Senent et al. (2013), which are acceptable according to the com-
been developed in recent years, such as the sequence quadratic parison conducted by Zhang et al. (2020b) and Senent et al. (2013).
iterative algorithm, second-order cone programming, and genetic This also shows that it is feasible to use the parameter conversion in
algorithm (Lysmer, 1970; Zhang et al., 2015). The sequence this study.
quadratic iterative algorithm and second-order cone programming The face stability analysis in the layered rock mass is compared
can efficiently solve the optimal solution, but when faced with with the numerical analysis using the finite element software
complex nonlinear problems, it is easy for them to fall into a local ABAQUS. Fig. 4 illustrates the model geometry used for numerical
optimal solution, thus missing the global optimal solution. How- simulation of the tunnel face stability, which only shows half the
ever, the genetic algorithm uses a probabilistic mechanism for symmetrical model. The numerical model of the rock runnel has a
global search to effectively avoid such problems, and finds the diameter of 10 m and a buried depth of 30 m. The size of the nu-
global optimal solution in the sense of probability. The genetic al- merical model is taken as 40 m in the x-direction, 50 m in the y-
gorithm is based on biological evolution, which has the advantages direction, and 70 m in the z-direction. The model contains
of good convergence, short computation time, and high robustness. approximately 51,730 zones and 57,429 nodes. Hence, boundary
Therefore, this study uses the genetic algorithm to find the optimal effects can be avoided for these dimensions. The model’s boundary
combination of q1, q2, q3, and q4 for the global minimum solution of conditions are given by fixed displacements at the bottom and the
the objective function (virtual support force) using MATLAB soft- lateral perimeter of the model.
ware. The flow diagram for performing the stability analysis of the A linear perfectly elastoplastic constitutive model based on the
tunnel face in this paper is illustrated in Fig. 3. MohreCoulomb failure criterion (with an associated flow rule) is
applied to the rock masses. The process of collapse simulation is
implemented according to the sequence of in situ stress balance e
4. Validation of analytical model rock mass excavation e virtual support force e gradually
decreasing the support force until the tunnel face collapses. The
The proposed model is simplified to consider a homogeneous detailed steps were described by Vermeer et al. (2002), and the
rock layer. To validate the proposed model, the results obtained equivalent cohesion c and frictional angle 4 were obtained by Hoek
et al. (2002). The corresponding rock Young’s modulus can also be
obtained through the alternative equation proposed by Hoek and
Diederichs (2006). The tunnel concrete lining is simulated with
shell structural elements with a Young’s modulus of 10 GPa, a
Poisson’s ratio of 0.2, a density of 2500 kg/m3, and a thickness of
0.4 m.
In this section, the tunnel face of the calculation model is
assumed to be in a two-layer rock mass, with the thickness of each
layer being 5 m. The numerical analysis considers two types of rock
masses: very poor rock mass (Case A) and relatively better rock
mass (Case B). The material properties are listed in Table 2. The
equivalent cohesion c, equivalent frictional angle 4, and the cor-
responding equivalent Young’s modulus proposed by Hoek et al.
(2002) and Hoek and Diederichs (2006) are also shown in
Table 2. The values are reserved to three decimal places to ensure
the accuracy of the equivalent parameters as much as possible.

Table 1
Comparison of collapse pressure obtained in this work with analytical results gained
by Senent et al. (2013).

GSI sc m r (kg/ Collapse pressure (kPa) Difference


(MPa) m3) (kPa)

Analytical results Limit analysis in this (a) e (b) e


(Senent et al., 2013) work (c) (c) (c)

Limit FLAC3D
analysis (a) (b)

20 1 5 2500 25.6 26.3 27.7 e2.7 e1.4


25 1 5 2500 19.1 18.9 19.5 e0.4 e0.6
10 5 5 2500 16.4 15.3 14.6 1.8 0.7
Fig. 3. Flowchart for optimization of the tunnel face stability analysis.
1842 J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848

h1 and the lower layer of the tunnel face h2 are determined by the
computational cases.

5.1. Design charts with two different collapse models (Cases 1 and
2)

This section investigates the tunnel face stability of the two


collapse models (Cases 1 and 2 in Fig. 1) and determines the less
favorable model geometry. The thickness of the upper and lower
rock layers of the tunnel face is set to 5 m. The value ranges of the
rock mass parameters are set as follows: m1 ¼ 5e25, GSI1 ¼ 5e25,
and sc1 ¼ 5e25 MPa. The normalized virtual support force is
defined as sT/(g1D). As shown in Fig. 6, the stability of the tunnel
face gradually increases as the quality of the rock mass increases.
Hence, the influence of the rock mass quality on the tunnel face
stability is similar for the two models. However, the failure model
corresponding to Case 1 is prone to collapse compared with Case 2.
Fig. 4. Numerical model.
Consequently, the subsequent analysis shall focus on the less stable
model geometry (Case 1).

5.2. Design charts with different rock layer thicknesses


In the numerical simulation, the minimum required virtual
support force can be determined by the mutation position of the This section investigates the influence of rock layer thickness on
horizontal displacement (Vermeer et al., 2002; Li et al., 2009; the tunnel face stability, considering changes in the geological
Ibrahim et al., 2015). The results of the numerical simulation and strength index GSI1, rock material constant m1, and unconfined
the analytical analysis of this paper for these two cases are shown compressive strength sc1. The thickness ratio is defined as
in Fig. 5a. The results calculated by the proposed method are close a ¼ h1 =D to characterize the change in the thickness of the upper
to those of the numerical simulation. The geometry of the sliding layer. As shown in Fig. 7, the normalized virtual support force de-
surface was also captured in ABAQUS for Case B. We can see that the creases nonlinearly as m1, GSI1, and sc1 increase. In contrast, the
failure surface prediction of the upper bound solution is similar to nonlinear decrease in the virtual support force weakens as the
that of the numerical model (see Fig. 5b). Overall, the method in thickness of the upper layer on the tunnel face decreases. Fig. 7 also
this paper has high computational efficiency, needing only 40e50 s, shows that the curves of the normalized virtual support force with
while the ABAQUS model needs several hours. Therefore, the different thicknesses of the upper layer have an intersection point.
calculation method is more suitable for construction applications. To the left side of the point, the greater the thickness of the upper
layer, the more unstable the tunnel face, whereas the situation to
the right side of this point is just the opposite. The average gradient
parameter vsT/vh1 is then defined to characterize the influence of
5. Design charts for face stability
the upper layer thickness under different rock parameters on the
stability of the tunnel face. As shown in Fig. 8, the average gradient
Layered rock masses usually have different mechanical proper-
gradually decreases from positive to negative as the values of m1,
ties due to the existence of structural planes in the rock masses
GSI1 and sc1 increase. This further indicates that the higher the
(Yang et al., 2018; Du et al., 2019; Wang et al., 2020), thus the tunnel
quality of the rock layer, the greater the thickness of the rock layer,
face stability in layered rock is different from that in homogeneous
and the more conducive to the stability of the tunnel face. In
rock. This section studies the influence of the thickness of the rock
addition, the contrast between the three curves shows that change
layer and the parameters of the rock layer at different positions on
in the geological strength index GSI1, which characterizes the joint
the stability of the tunnel face in the layered rock mass, and pro-
characteristics of the rock mass, has a more significant effect
vides the corresponding design charts for engineering reference.
compared to the other two parameters on the stability of a tunnel
The basic parameters are as follows: the diameter of the tunnel
face in a poor rock mass. However, for a better-quality rock mass, a
D ¼ 10 m, the thickness of the overburden H ¼ 20 m, and
change in unconfined compressive strength sc1, which character-
sci ¼ 1 MPa, mi ¼ 5, GSIi ¼ 10, di ¼ 0, and gi ¼ 25 kN/m3 (very poor
izes the strength of the rock mass, has a more significant effect on
rock mass), in which i ¼ 0; 1; 2 corresponds to the overburden,
the stability of a tunnel face compared to the other two parameters.
the upper layer of the tunnel face and the lower layer of the tunnel
face, respectively. The detailed properties of each rock layer are
5.3. Design charts with different tunnel diameters
shown in Table 3. The thickness of the upper layer of the tunnel face

Table 2 To characterize the influence of tunnel diameter changes on the


Parameters of the calculation model. stability of a tunnel face, Fig. 9 presents the influence of different
Case Rock layer m sc GSI Equivalent parameter rock layer parameters changing with the thickness ratio on the
(MPa) stability of the tunnel face for three different diameters. Transverse
Cohesion, c Frictional Young’s modulus
(kPa) angle ( ) (MPa) comparison of the subfigures in Fig. 9 shows that as the tunnel
diameter increases, the stability of the tunnel face deteriorates, and
A Overburden 5 1 10 13.520 13.470 270.730
Upper layer 7 1 10 15.929 15.602 270.730 the increase of thickness ratio a has a more significant influence on
Lower layer 5 1 15 19.955 17.315 425.861 the stability of the tunnel face, regardless of m1, GSI1 and sc1. Ver-
B Overburden 5 1 10 13.520 13.470 270.730 tical comparison between the subfigures in Fig. 9 shows that in the
Upper layer 5 5 10 24.579 21.233 270.730 cases with a poor-quality upper layer rock mass (the black lines in
Lower layer 7 1 15 19.955 17.315 425.861
Fig. 9), the parameter GSI1 is more sensitive to the change of
J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848 1843

Fig. 5. (a) Comparison of numerical simulation and analytical results for two cases; and (b) Comparison of failure mechanisms computed with the upper-bound solution and the
numerical model (unit: m).

Table 3
Basic parameters of each rock layer.
tunnel face stability. Hence, the rock quality of the lower layer
Rock layer Thickness sc (MPa) m GSI d g (kN/m3) controls the stability of the tunnel face.
Overburden H ¼ 20 m 1 5 10 0 25 This section also investigates the influence of the rock layer
Upper layer h1 1 5 10 0 25 arrangement on the tunnel face stability. It needs to be explained
Lower layer h2 1 5 10 0 25 here that strength and weakness here are relative. Two types of
layer arrangements are investigated: one with m1 ¼ 5e25 and
m2 ¼ 15, and another with m2 ¼ 5e25 and m1 ¼ 15. As shown in
Fig. 11, to either the left or right of the intersection point, the sta-
thickness ratio. However, as the upper layer rock mass quality
bility of the tunnel face with the upper-weak and lower-strong
improves, the parameter sc1 becomes the most sensitive to changes
arrangement is more stable than that of the lower-weak and
in the thickness ratio, and this influence becomes greater as the
upper-strong arrangement. Therefore, the arrangement of the weak
tunnel diameter increases. Therefore, it can be concluded that the
and the strong rock layers has an important influence on the sta-
stability of a tunnel face is significantly affected by the size of the
bility of the tunnel face. In practice, it is necessary to judge the
tunnel diameter and the distribution of the rock layer thickness.
stability of a tunnel face according to the arrangement of weak and
strong layers.
5.4. Design charts with different rock layer parameters Fig. 12 shows the results obtained from 820 calculated exam-
ples. It was assumed that the tunnel face’s upper and lower rock
In this part, the thicknesses of the upper and lower rock layers of layer parameters change simultaneously, mainly because the rock
the tunnel face are set to 5 m to eliminate the influence of the masses are cut into finite blocks by a weak plane (persistent joint
thickness of the rock layers. The following analysis investigates the plane). Note that selection of the parameters primarily corresponds
influence of the geological strength index GSI and the rock material to the poor-quality rock masses, where face instability problems are
constant m on the stability of the tunnel face when the other pa- more likely to be found in engineering practice. Accordingly,
rameters are kept constant. The normalized virtual support force sc =ðgDÞ values vary from 4 to 100 (equivalent to sc between 1 MPa
sT =ðgi DÞ is plotted against the normalized UCS sci =ðgi DÞ. As shown and 25 MPa), typical of soft rocks to very soft rocks. GSI values are
in Fig. 10, the normalized virtual support force declines nonlinearly taken between 10 and 30, which are characteristic values of very
as sci =ðgi DÞ increases from 4 to 100, and the nonlinear behavior poor-quality rock masses. As shown in Fig. 12, the virtual support
gradually weakens. The first column in Fig. 10 shows that, as the force gradually decreases to zero when the rock mass quality in-
parameter sci =ðgi DÞ increases, the influence of changes in the creases, which means that the tunnel face can maintain a self-
parameter mi on the virtual support force gradually weakens. stable state. However, comparison with Fig. 10 shows that the
However, for the geological strength index GSIi, the second column stability of the tunnel face will be overestimated if the stratification
of Fig. 10 indicates that the influence of changes in GSIi on the of the rock mass and the differences in the properties of the upper
virtual support force gradually increases as sci =ðgi DÞ increases. and lower rock layers are not considered. Therefore, the calculation
More importantly, the comparison shows that the changes in the method proposed in this paper is more suitable for determining the
parameters of the lower rock layer have a greater influence on the stability state of a tunnel face in layered rock.
1844 J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848

Fig. 6. Influence of varying parameters on the tunnel face stability under two collapse models: (a) m1, (b) GSI1, and (c) sc1.

Fig. 7. Influence of thickness on the virtual support force with variations in (a) m1, (b) GSI1, and (c) sc1.
J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848 1845

are useful for evaluating the stability of a tunnel face in a layered


rock mass. The results of this study are as follows:

(1) For layered rock masses with different layer thicknesses, the
geological strength index GSI1, rock material constant m1, and
UCS sc1 significantly influence the stability of tunnel face
characterized by the normalized virtual support force, which
varies nonlinearly with increases in GSI1, m1 and sc1 : The
nonlinear decreasing trend weakens with a decrease in the
thickness of the upper layer of the tunnel face. As the rock
mass quality improves, the key factor that controls the tunnel
face stability changes from GSI1 to sc1 .
(2) The tunnel diameter and the distribution of the rock layer
thickness significantly influence the stability of tunnel face.
The larger the tunnel diameter, the greater the proportion of
heavily fractured rock masses, the more unfavorable the
stability of the tunnel face. The arrangement of the weak and
Fig. 8. The average gradient varying with rock parameters.
strong rock layers has an important influence on the tunnel
face stability. The research found that the tunnel face in a
lower-weak and upper-strong composite layer fails more
6. Conclusions easily than that in an upper-weak and lower-strong com-
posite layer.
This study extends the upper bound limit analysis method to the (3) The influence of the parameters at different positions on the
face stability of a tunnel excavated in layered rock. The influence of tunnel face stability was also investigated. As expected, the
rock layer characteristics on the stability of the tunnel face is normalized virtual support force declines nonlinearly with
considered, and a series of dimensionless parameter charts is pre- an increase in sci =ðgi DÞ. The nonlinear decreasing trend
sented using the upper bound limit analysis method. These charts weakens as the rock quality increases. In addition, the face

Fig. 9. Influence of parameter variations on the stability of tunnel face under different tunnel diameters.
1846 J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848

Fig. 10. Influence of the rock material constants for a layered rock on the virtual support force: (a) m0; (b) GSI0; (c) m1; (d) GSI1; (e) m2; and (f) GSI2.

sci =ðgi DÞ increases, the influence of increases in mi on the


tunnel face stability gradually weakens. However, the influ-
ence of increases in GSIi on the face stability gradually in-
creases with increases in sci =ðgi DÞ:

In summary, the method presented in this study can address


stability in terms of the stratification characteristics of the tunnel
face. Considering the cross-effects of rock parameters, the
normalized virtual support force can be directly used to guide
tunnel design.
It should be noted that the actual rock tunnel has a three-
dimensional (3D) arching effect, which inevitably leads to differ-
ences between the proposed 2D analytical solution and the actual
tunnel face stability. As shown in Fig. 5a and Table 1, the proposed
2D analytical solution overestimates the virtual support force of a
tunnel face for extremely fractured rock masses, which may be
Fig. 11. Influence of the arrangement of strong and weak rock layers on the tunnel face associated with ignoring the 3D arching effect. Further studies
stability.
should be conducted to derive an analytical solution that considers
the 3D boundary condition with mechanical equivalence. The
proposed solution in this article is an exploratory work to analyze
stability is more sensitive to the variations in the lower layer tunnel face stability in stratified rock masses. In addition, due to the
parameters than that of the upper layer. As the parameter complexity of layered rock masses, several assumptions were
J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848 1847

Fig. 12. Stability of tunnel face under different parameters.

adopted to derive the 2D analytical solution, such as ignoring the Babendererde, S., Hoek, E., Marinos, P., Silva Cardoso, A., 2006. Geological risk in the
use of TBMs in heterogeneous rock masses e the case of "Metro do Porto. In:
effect of stratification of the rock masses when the parameters are
Geotechnical Risk in Rock Tunnels. CRC Press, London, UK, pp. 41e51.
equivalent, assuming a progressive collapse model at failure, and so Chen, J., Zhou, M., Huang, H., Zhang, D., Peng, Z., 2021. Automated extraction and
on. In-depth study shall be undertaken in the future to address evaluation of fracture trace maps from rock tunnel face images via deep
those less realistic assumptions. learning. Int. J. Rock Mech. Min. Sci. 142, 104745.
Chen, R., Yin, X., Tang, L., Chen, Y., 2018. Centrifugal model tests on face failure of
earth pressure balance shield induced by steady state seepage in saturated
sandy silt ground. Tunn. Undergr. Space Technol. 81, 315e325.
Declaration of competing interest Chen, R.P., Tang, L.J., Ling, D.S., Chen, Y.M., 2011. Face stability analysis of shallow
shield tunnels in dry sandy ground using the discrete element method. Comput.
The authors declare that they have no known competing Geotech. 38 (2), 187e195.
Chen, R.P., Tang, L.J., Yin, X.S., Chen, Y.M., Bian, X.C., 2015. An improved 3D wedge-
financial interests or personal relationships that could have
prism model for the face stability analysis of the shield tunnel in cohesionless
appeared to influence the work reported in this paper. soils. Acta Geotech. 10 (5), 683e692.
Chen, W.F., 1975. Limit Analysis and Soil Plasticity. Elsevier, Amsterdam, The
Netherlands, pp. 107e167.
Acknowledgments Davis, E.H., Gunn, M.J., Mair, R.J., Seneviratine, H.N., 1980. The stability of shallow
tunnels and underground openings in cohesive material. Geotechnique 30 (4),
397e416.
The work described in this paper is supported by the Key
Dong, W., Anagnostou, G., 2013. Evaluation of a modified hardening model for
Innovation Team Program of Innovation Talents Promotion Plan by squeezing rocks in tunneling. In: Proceedings of the 3rd International Confer-
MOST of China (Grant No. 2016RA4059) and the Science and ence on Computational Methods in Tunneling and Subsurface Engineering.
Aedificatio Publishers, Freiburg, Germany, pp. 267e274.
Technology Project of Yunnan Provincial Transportation Depart-
Dong, W., Anagnostou, G., 2014. The effect of constitutive modelling on estimates of
ment (No. 25 of 2018). the short-term response of squeezing ground to tunnel excavation. In: Com-
puter Methods and Recent Advances in Geomechanics: Proceedings of the 14th
Appendix A. Supplementary data International Conference of the International Association for Computer
Methods and Advances in Geomechanics. CRC Press, London, UK, pp. 225e230.
Du, S., Li, D., Sun, J., 2019. Stability control and support optimization for a soft-rock
Supplementary data to this article can be found online at
roadway in dipping layered strata. Geotech. Geol. Eng. 37 (3), 2189e2205.
https://doi.org/10.1016/j.jrmge.2021.12.023. Health, Executive, Safety, 1996. Safety of New Austrian Tunnelling Method (NATM)
Tunnels: A Review of Sprayed Concrete Lined Tunnels with Particular Reference
to London Clay. Her Majesty’s Stationery Office, London, UK.
References Hoek, E., Carranza-Torres, C., Corkum, B., 2002. HoekeBrown failure criterion e
2002 edition. In: Hammah, R., Bawden, W., Curran, J., Telesnicki, M. (Eds.),
Anagnostou, G., Schuerch, R., Ramoni, M., 2014. TBM tunnelling in complex rock Mining and Tunnelling Innovation and Opportunity: Proceedings of the 5th
formations. In: Interventi ed opere nelle formazioni complesse. 15 ciclo di North American Rock Mechanics Symposium and the 17th Tunnelling Associ-
conferenze di meccanica e ingegneria delle rocce. CELID, Torino, Italy, pp. 307e ation of Canada Conference. University of Toronto, Toronto, Canada.
331. Hoek, E., Brown, E.T., 2019. The HoekeBrown failure criterion and GSI e 2018
Anagnostou, G., Zingg, S., 2013. On the stabilizing effect of advance drainage in edition. J. Rock Mech. Geotech. Eng. 11 (3), 445e463.
tunnelling. Geomech. Tunn. 6 (4), 338e354. Hoek, E., Diederichs, M.S., 2006. Empirical estimation of rock mass modulus. Int. J.
Arnáiz Ronda, M., Melis Maynar, M., Trabada Guijarro, J., Melis Maynar, M., 2003. Rock Mech. Min. Sci. 43 (2), 203e215.
Problemas de inestabilidad en terrenos arenosos en la ampliación del metro de Horn, M., 1961. Horizontal earth pressure on perpendicular tunnel face. In: Hun-
Madrid. Rev. Obras Publicas: Organo profesional de los ingenieros de caminos, garian National Conference of the Foundation Engineer Industry. Budapest,
canales y puertos (3429), 21e33 (in Spanish). Hungary (in Hungarian).
1848 J. Man et al. / Journal of Rock Mechanics and Geotechnical Engineering 14 (2022) 1836e1848

Huang, Z.K., Pitilakis, K., Argyroudis, S., Tsinidis, G., Zhang, D.M., 2021. Selection of Weng, X., Sun, Y., Yan, B., Niu, H., Lin, R., Zhou, S., 2020. Centrifuge testing and
optimal intensity measures for fragility assessment of circular tunnels in soft numerical modeling of tunnel face stability considering longitudinal slope
soil deposits. Soil Dynam. Earthq. Eng. 145, 106724. angle and steady state seepage in soft clay. Tunn. Undergr. Space Technol. 101,
Ibrahim, E., Soubra, A.H., Mollon, G., Raphael, W., Dias, D., Reda, A., 2015. Three- 103406.
dimensional face stability analysis of pressurized tunnels driven in a multi- Xu, D.P., Feng, X.T., Chen, D.F., Zhang, C.Q., Fan, Q.X., 2017. Constitutive represen-
layered purely frictional medium. Tunn. Undergr. Space Technol. 49, 18e34. tation and damage degree index for the layered rock mass excavation response
Idinger, G., Aklik, P., Wu, W., Borja, R.I., 2011. Centrifuge model test on the face in underground openings. Tunn. Undergr. Space Technol. 64, 133e145.
stability of shallow tunnel. Acta Geotech. 6 (2), 105e117. Yang, X.L., Zhou, T., Li, W.T., 2018. Reliability analysis of tunnel roof in layered Hoek-
Jiang, M., Yin, Z.Y., 2014. Influence of soil conditioning on ground deformation Brown rock masses. Comput. Geotech. 104, 302e309.
during longitudinal tunneling. C. R. Méc. 342 (3), 189e197. Yin, Z.Y., Wang, P., Zhang, F.S., 2020. Effect of particle shape on the progressive
Kim, S.H., Tonon, F., 2010. Face stability and required support pressure for TBM failure of shield tunnel face in granular soils by coupled FDM-DEM method.
driven tunnels with ideal face membrane e drained case. Tunn. Undergr. Space Tunn. Undergr. Space Technol. 100, 103394.
Technol. 25 (5), 526e542. Zhang, J., Chen, Y., Chen, T., Mei, Z., Liu, Z., Wang, F., Hu, Q., 2018. Law and char-
Kirsch, A., 2010. Experimental investigation of the face stability of shallow tunnels acteristics analysis of domestic tunnel construction accidents from 2006 to
in sand. Acta Geotech. 5 (1), 43e62. 2016. Mod. Tunn. Technol. 55 (3), 10e17 (in Chinese).
Leca, E., Dormieux, L., 1990. Upper and lower bound solutions for the face stability Zhang, J., Zhang, L., Wang, W., Zhang, D., Zhang, B., 2020b. Probabilistic analysis of
of shallow circular tunnels in frictional material. Geotechnique 40 (4), 581e606. three-dimensional tunnel face stability in soft rock masses using HoekeBrown
Li, T.Z., Dias, D., Li, Z.W., 2020. Failure potential of a circular tunnel face under failure criterion. Int. J. Numer. Anal. Methods GeoMech. 44 (11), 1601e1616.
steady-state unsaturated flow condition. Comput. Geotech. 117, 103231. Zhang, J.H., Li, Y.X., Xu, J.S., 2015. Energy analysis of face stability of deep rock
Li, T.Z., Yang, X.L., 2020. New approach for face stability assessment of tunnels tunnels using nonlinear Hoek-Brown failure criterion. J. Cent. South Univ. 22
driven in nonuniform soils. Comput. Geotech. 121, 103412. (8), 3079e3086.
Li, Y., Emeriault, F., Kastner, R., Zhang, Z.X., 2009. Stability analysis of large slurry Zhang, J.Z., Huang, H.W., Zhang, D.M., Phoon, K.K., Liu, Z.Q., Tang, C., 2021c. Quan-
shield-driven tunnel in soft clay. Tunn. Undergr. Space Technol. 24 (4), 472e481. titative evaluation of geological uncertainty and its influence on tunnel struc-
Liu, W., Zhao, Y., Shi, P., Li, J., Gan, P., 2018. Face stability analysis of shield-driven tural performance using improved coupled Markov chain. Acta Geotech. 16 (11),
tunnels shallowly buried in dry sand using 1-g large-scale model tests. Acta 3709e3724.
Geotech. 13 (3), 693e705. Zhang, J.Z., Huang, H.W., Zhang, D.M., Zhou, M.L., Tang, C., Liu, D.J., 2021a. Effect of
Long, Y.Y., Tan, Y., 2020. Soil arching due to leaking of tunnel buried in water-rich ground surface surcharge on deformational performance of tunnel in spatially
sand. Tunn. Undergr. Space Technol. 95, 103158. variable soil. Comput. Geotech. 136, 104229.
Lysmer, J., 1970. Limit analysis of plane problems in soil mechanics. J. Soil Mech. Zhang, J.Z., Phoon, K.K., Zhang, D.M., Huang, H.W., Tang, C., 2021d. Novel approach
Found. Div. ASCE 96 (4), 1311e1334. to estimate vertical scale of fluctuation based on CPT data using convolutional
Maynar, M.J.M., Rodriguez, L.E.M., 2005. Discrete numerical model for analysis of neural networks. Eng. Geol. 294, 106342.
earth pressure balance tunnel excavation. J. Geotech. Geoenviron. Eng. 131 (10), Zhang, J.Z., Zhou, X.P., 2017. Time-dependent jamming mechanism for Single-Shield
1234e1242. TBM tunneling in squeezing rock. Tunn. Undergr. Space Technol. 69, 209e222.
Mollon, G., Dias, D., Soubra, A.H., 2010. Face stability analysis of circular tunnels Zhang, X., Wang, M., Li, J., Wang, Z., Tong, J., Liu, D., 2020a. Safety factor analysis of a
driven by a pressurized shield. J. Geotech. Geoenviron. Eng. 136 (1), 215e229. tunnel face with an unsupported span in cohesive-frictional soils. Comput.
Mollon, G., Dias, D., Soubra, A.H., 2011a. Rotational failure mechanisms for the face Geotech. 117, 103221.
stability analysis of tunnels driven by a pressurized shield. Int. J. Numer. Anal. Zhang, Z., Xu, W., Nie, W., Deng, L., 2021b. DEM and theoretical analyses of the face
Methods GeoMech. 35 (12), 1363e1388. stability of shallow shield cross-river tunnels in silty fine sand. Comput. Geo-
Mollon, G., Phoon, K.K., Dias, D., Soubra, A.H., 2011b. Validation of a new 2D failure tech. 103905.
mechanism for the stability analysis of a pressurized tunnel face in a spatially Zhao, L.H., Cheng, X., Li, L., Chen, J.Q., Zhang, Y., 2017. Seismic displacement along a
varying sand. J. Eng. Mech. 137 (1), 8e21. log-spiral failure surface with crack using rock HoekeBrown failure criterion.
Oreste, P.P., Dias, D., 2012. Stabilisation of the excavation face in shallow tunnels Soil Dynam. Earthq. Eng. 99, 74e85.
using fibreglass dowels. Rock Mech. Rock Eng. 45 (4), 499e517. Zhao, S., Shadabfar, M., Zhang, D., Chen, J., Huang, H., 2021. Deep learning-based
Pan, Q., Dias, D., 2018a. Three-dimensional static and seismic stability analysis of a classification and instance segmentation of leakage-area and scaling images
tunnel face driven in weak rock masses. Int. J. GeoMech. 18 (6), 04018055. of shield tunnel linings. Struct. Control Health Monit. 28 (6), e2732.
Pan, Q., Dias, D., 2018b. Three dimensional face stability of a tunnel in weak rock Zhou, M., Chen, J., Huang, H., Zhang, D., Zhao, S., Shadabfar, M., 2021. Multi-source
masses subjected to seepage forces. Tunn. Undergr. Space Technol. 71, 555e566. data driven method for assessing the rock mass quality of a NATM tunnel face
Paternesi, A., Schweiger, H.F., Scarpelli, G., 2017. Numerical analyses of stability and via hybrid ensemble learning models. Int. J. Rock Mech. Min. Sci. 147, 104914.
deformation behavior of reinforced and unreinforced tunnel faces. Comput. Zhou, S., Guo, X., Zhang, Q., Dias, D., Pan, Q., 2020. Influence of a weak layer on the
Geotech. 88, 256e266. tunnel face stability e reliability and sensitivity analysis. Comput. Geotech. 122,
Perazzelli, P., Leone, T., Anagnostou, G., 2014. Tunnel face stability under seepage 103507.
flow conditions. Tunn. Undergr. Space Technol. 43, 459e469. Zou, J., Chen, G., Qian, Z., 2019. Tunnel face stability in cohesion-frictional soils
Qin, C.B., Chian, S.C., Yang, X.L., 2017. 3D limit analysis of progressive collapse in considering the soil arching effect by improved failure models. Comput. Geo-
partly weathered HoekeBrown rock banks. Int. J. GeoMech. 17 (7), 04017011. tech. 106, 1e17.
Senent, S., Mollon, G., Jimenez, R., 2013. Tunnel face stability in heavily fractured Zou, J.F., Qian, Z.H., 2018. Face-stability analysis of tunnels excavated below
rock masses that follow the HoekeBrown failure criterion. Int. J. Rock Mech. groundwater considering coupled flow deformation. Int. J. GeoMech. 18 (8),
Min. Sci. 60, 440e451. 04018089.
Shen, J., Priest, S.D., Karakus, M., 2012. Determination of MohreCoulomb shear
strength parameters from generalized HoekeBrown criterion for slope stability
analysis. Rock Mech. Rock Eng. 45 (1), 123e129.
Subrin, D., Wong, H., 2002. Stabilité du front d’un tunnel en milieu frottant : un
nouveau mécanisme de rupture 3D. C. R. Méc. 330 (7), 513e519 (in French). Dr. Mingliang Zhou is currently serving as a postdoctoral
Ukritchon, B., Yingchaloenkitkhajorn, K., Keawsawasvong, S., 2017. Three- dimen- research fellow at the Department of Geotechnical Engi-
sional undrained tunnel face stability in clay with a linearly increasing shear neering, Tongji University, China. He is a member of the
strength with depth. Comput. Geotech. 88, 146e151. Machine Learning and Big Data (TC309) Group of the In-
Vermeer, P.A., Ruse, N., Marcher, T., 2002. Tunnel heading stability in drained ternational Society for Soil Mechanics and Geotechnical
ground. Felsbau 20 (6), 8e18. Engineering (ISSMGE-TC309). Dr. Zhou obtained his BSc,
Wang, F.Y., Zhou, M.L., Zhang, D.M., Huang, H.W., Chapman, D., 2019. Random MSc, and PhD degrees from the University of Cambridge,
evolution of multiple cracks and associated mechanical behaviors of segmental UK. His research interest is risk assessment using
tunnel linings using a multiscale modeling method. Tunn. Undergr. Space advanced techniques and control of geotechnical infra-
Technol. 90, 220e230. structure. Utilizing recent advances in machine learning-
Wang, H.T., Liu, C., Liu, P., Zhang, X., Yang, Y., 2020. Prediction of the required based technology, he has established a research reputa-
supporting pressure for a shallow tunnel in layered rock strata based on 2D and tion using computer vision and deep learning techniques
3D upper bound limit analysis. Adv. Civ. Eng. 2020, 6261917. to solve practical geotechnical problems. He has been
Wang, Q., Xu, M., Zhang, Y., Cen, X., Chang, X., 2021. Mechanical parameters of deep- invited to present his work in more than 10 international
buried coal goaf rock mass based on optimized GSI quantitative analysis. Adv. and national level academic conferences, and has published more than 40 journal
Civ. Eng. 2021, 9935860. papers.

You might also like