You are on page 1of 9

Case Study

Stability Assessment and Support Design for Underground


Tunnels Located in Complex Geologies and Subjected to
Engineering Activities: Case Study
Yan Xing, Ph.D.1; P. H. S. W. Kulatilake, F.ASCE2; and L. A. Sandbak3

Abstract: A good understanding of rock mass behavior around underground tunnels is necessary for stability assessment and support design.
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

This paper presents a three-dimensional (3D) numerical analysis to support the aforementioned statement for an underground mine by using
the distinct element method. Geological and engineering complexities simulated in the numerical model include large-scale faults and a non-
planar weak interlayer, as well as open and backfilled tunnels. Sequential excavation, backfilling, and delayed supporting were simulated
according to the field construction process. Numerical analysis investigated the effect of complex geologies and engineering activities on tun-
nel stability. Deformations and strength degradation areas around the tunnels are illustrated and analyzed at different locations. Based on the
rock mass behavior and the failure conditions of the applied rock supports, useful suggestions are made on the selection of appropriate tunnel
support for this underground mine. These suggestions shared similarities and slight differences with the guidelines suggested by an empirical
method. The accuracy of the numerical results is verified by comparing with the field deformation data. This study provides a comprehensive
procedure for stability assessment and support design for similar underground rock mass projects. DOI: 10.1061/(ASCE)GM.1943-
5622.0001402. © 2019 American Society of Civil Engineers.
Author keywords: Tunnel stability; 3D modeling; Distinct element method; Rock supports.

Introduction with the development of several mature systems, such as the rock
mass rating (RMR) (Bieniawski 1976, 1989), Q system (Barton
The stability of underground tunnels, which provide access to the et al. 1974; Barton 2002), rock mass index (RMi) (Palmström 1995
mine area, and are used to transport the ore, is evidently important 2000), and geological strength index (GSI) (Marinos and Hoek
to underground mine projects. For practical problems, complex 2000), they have been widely applied to evaluate rock mass proper-
geologies, such as faults and weak interlayers that are usually ties and to design the preliminary supports. However, the stability
encountered in the field, play critical roles on the stability of the sur- assessment of rock masses, understanding the intrinsic mechanisms,
rounding rock masses (Hao and Azzam 2005; Huang et al. 2013). and determination of possible failure modes are out of the capability
Engineering activities such as the exploitation of ores and the con- of the rock mass classification systems. Although the recently
struction of underground openings, which are needed for the pro- updated RMR and Q systems use several rock mass parameters,
duction, also contribute to the instabilities of rock masses. A com- some key factors, such as the explicit representation of discontinu-
plete understanding of rock mass behavior under site-specific ity, orientation, and size, the effects of blasting, staged tunneling
circumstances is therefore essential to the stability assessment and processes, and time-dependent deformation, are disregarded in the
support design for underground tunnels. rock mass classification system. Although the same output rating is
In general, the empirical methods, that is, rock mass classifica- obtained, the rock mass behavior could be different due to various
tion systems, are preferred by rock engineers to use in tunnel engi- combinations of the classification parameters (Stille and Palmström
neering because of their simplicity (Bieniawski 1988). Especially 2003). The output description or a single number obtained from the
classification system is sometimes too simple to predict the support
requirements. According to the tunneling practice of Riedmüller
1
Postdoctoral Researcher, School of Mechanics and Civil Engineering, and Schubert (1999), it was concluded that the rock mass classifica-
China Univ. of Mining and Technology, Xuzhou 221116, China; formerly, tions were inadequate for support design. The classification systems
Ph.D. Student, Dept. of Mining and Geological Engineering, Univ. of
were suggested not to guide the ultimate design or to replace other
Arizona, Tucson, AZ 85721 (corresponding author). Email: yanx@cumt
.edu.cn sophisticated methods (Bieniawski 1988).
2
Academic Director and Distinguished Professor, School of Resources By the aid of computer technology, the numerical method has
and Environmental Engineering, Jiangxi Univ. of Science and Technology, been proved as an efficient and competent tool to solve problems in
Ganzhou 341000, China; formerly, Professor, Rock Mass Modeling and rock mechanics and rock engineering (Jing 2003). It has the capabil-
Computational Rock Mechanics Laboratories, Univ. of Arizona, Tucson, ity of simulating the field complexities as previously mentioned.
AZ 85721. For instance, Hao and Azzam (2005) applied the universal dis-
3
Senior Geotechnical Engineer, Barrick Gold, Inc., Turquoise Ridge tinct element code (UDEC) (Itasca Consulting Group, Inc. 1998) to
Joint Venture, Golconda, NV 89414.
study the influence of a fault on the plastic zones and displacements
Note. This manuscript was submitted on June 8, 2018; approved on
October 26, 2018; published online on February 22, 2019. Discussion pe- around an underground structure. Coggan et al. (2012) showed the
riod open until July 22, 2019; separate discussions must be submitted for significant influence of the relative weak mudstone that existed in a
individual papers. This paper is part of the International Journal of tunnel roof on the extent of failure and the need for additional rein-
Geomechanics, © ASCE, ISSN 1532-3641. forcement by using Phase2 (Rocscience 2011). Huang et al. (2013)

© ASCE 05019004-1 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004


discussed the effect of dip angle, thickness, and distance of a weak are inclined with a low rock mass quality in the ore zone. The rocks
interlayer on the failure patterns around a tunnel. With respect to en- include carbonaceous mudstone and limestone, tuffaceous mud-
gineering factors, Hsiao et al. (2009) investigated the rock deforma- stone and limestone, polylithic megclastic debris flows, fine-
tion at a tunnel intersection area with different intersection angles and grained debris flows, and basalts. The lithologies are intruded by
proposed rock support design; Chu et al. (2007) and Ng et al. (2004) dikes and sills. Major structures include the high-angle faults strik-
numerically investigated interactions between twin tunnels. ing NW-SE and dipping southwest, the sets that strike NE-SW and
Because numerical modeling can provide detailed and reliable in- dip northwest, and the low-angle faults dipping to the east with
formation for rock mass behavior, such as thickness of plastic zones 20–40°. Limited drilling information showed that the local faults
and total displacements, it has been utilized in some studies to deter- are distributed with a maximum mean spacing of 3–15 m and are
mine the optimum support systems in tunneling (Choi and Shin 2004; mainly filled with gouge and breccia clay; the local joints include
Sari et al. 2008; Hsiao et al. 2009; Yalcin et al. 2016; Kanik and fillings of calcite and clay in thickness of 0.2–0.4 cm. Compared to
Gurocak 2018). Kanik and Gurocak (2018) evaluated the perform- the whole mine area, the selected study region is a small area.
ance of preliminary support systems designed based on RMR89, Q, Limited information is available on minor discontinuity geometry.
and RMi through numerical analyses and subsequently proposed the In this region, the stratigraphy is formed with mudstone, limestone,
optimum support systems based on the plastic zone thickness and the
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

and dacite. A weak dike interlayer dips to the east with a typical
maximum total displacements. Similarly, Yalcin et al. (2016) numeri- angle range of 25–45°. A major nonpersistent fault strikes N53°W
cally analyzed the effectiveness of support systems implemented and dips 60° to the southwest.
based on RMR and RMi and highlighted the importance of numerical
approaches in determining the final optimum support systems.
Although a significant number of numerical analyses have been Estimation of Rock Mass Properties
carried out on the stability assessment and support design for under- Both geotechnical core logging and geotechnical mapping have
ground tunnels as aforementioned, most of the studies merely been carried out to collect the information of underground rock
focused on the effect of one geological or engineering factor, or masses. The rock mass conditions were evaluated using the 1976
simply applied the two-dimensional (2D) modeling. On the
version of Bieniawski’s (1976) RMR system. In the selected region,
other hand, underground excavation behavior in rock masses is a
OC5 is the dominating lithology, which consists of limestone with
three-dimensional (3D) problem, where the 3D redistribution of
thinly laminated to thinly bedded mudstones. The rock masses in
excavation-induced stresses poses a significant impact on the stabil-
OC5 have a wide range of RMR values, from 25–55 (poor to fair
ity of excavations (Eberhardt 2001; Kaiser et al. 2001). In numerical
rock). Dike is however an intrusive dacite interlayer, and the rock
modeling, the simulation of the 3D sequences of excavation and
mass is in a relatively poor condition in the range of 25–40.
rock supporting is crucial for accurate prediction of rock mass
behavior (Cai 2008; Cantieni and Anagnostou 2009). Numerical Average RMR values for the two lithologies are, respectively, 40
results could be compromised from the reality by performing a 2D and 32.5. Based on the rock mass conditions and the laboratory
analysis or a highly simplified 3D analysis. Therefore, a rigorous tests, the rock mass properties (the combined effect of the intact
stress analysis taking into account many key factors is needed. rock and minor discontinuities) were estimated using empirical for-
Xing et al. (2018a, b) investigated tunnel stability in an under- mulas proposed by Hoek et al. (2002) and Hoek and Diederichs
ground mine through sophisticated numerical modeling using the 3D (2006). According to Hoek and Brown (1997), the GSI (Hoek
distinct element code (3DEC) (Cundall 1988; Hart et al. 1988). 1994) can be estimated from the 1976 version of the RMR values
Multiple geological and engineering features were incorporated in the (Bieniawski 1976) with the ground water rating set to 10 (dry) and
numerical model based on the geological and geotechnical informa- the adjustment for joint orientation set to 10 (very favorable).
tion, such as the persistent and nonpersistent faults, a nonplanar dike Detailed procedures of estimating the rock mass properties were
layer, and a convoluted tunnel system. Additionally, sequential exca- presented in Xing et al. (2018a), and the results are given in Table 1.
vation, backfilling, and supporting procedures were simulated accord- The bulk modulus (K) and shear moduli (G) given in Table 1
ing to the way those were implemented in the field, and the stress were calculated from the deformation modulus and Poisson’s ratio
relaxation method was adopted to account for the delayed supporting. of rock masses using Eqs. (1) and (2).
Numerical predictions were compared with field measurements to Er
estimate in situ lateral stress ratios and the stress loss factor in the K¼ (1)
3ð1 – 2 m r Þ
stress relaxation method (Xing et al. 2018a, b). In the present paper, a
stress loss factor of 0.5 is applied based on the previously mentioned
comparison in simulating the delayed supporting. Supplementary Er
analyses are performed in this study to evaluate the effect of the com- G¼ (2)
2ð1 þ m r Þ
plex geologies and engineering activities on tunnel stability. Based on
the rock mass behavior and the support failures, useful suggestions Table 1. Physical and mechanical property values for the rock masses
are provided for the selection of tunnel support for this underground
mine. Finally, the tunnel support design predicted by the numerical Property OC5 Dike Backfilling
modeling was compared with that estimated by an empirical method. Density (kg/m3) 2,743 2,380 2,146
Bulk modulus [K (GPa)] 8.11 1.11 0.73
Geologic Settings and Engineering Properties of the Shear modulus [G (GPa)] 3.74 0.57 0.55
Friction angle [ f p (degrees)] 35.5 25.6 44.0
Underground Tunnels
Cohesion [cp (MPa)] 1.96 1.20 0.69
Tensile strength [s tp (MPa)] 2.56 1.15 0.61
Geologic Settings f r (degrees) 28.4 21.8 —
cr (MPa) 0.78 0.60 —
The studied mine is an underground mine in the United States. The
s tr (MPa) 1.02 0.58 —
geological conditions of the mine site are complex. The ore bodies

© ASCE 05019004-2 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004


Tunnel Excavation and Support Installation Specific layouts and configurations of the rock supports can be
found in Xing et al. (2018a, b). To simulate the deformation that
The target tunnel system is the development drifts, which are uti-
occurred prior to support installation, the stress relaxation method
lized to extract and transport the ore from the mining area. Fig. 1(a)
(Vardakos et al. 2007; Janin et al. 2015; Shreedharan and
gives the plan view of the overall tunnel system. It extends 640 m
Kulatilake 2016) was adopted. In this approach, an internal pres-
(2,100 ft) along the east-west direction and 305 m (1,000 ft) along
sure/traction equal to in situ stress is applied to the excavation
the south-north direction. The elevation of the tunnels, as presented
in Fig. 1(b), ranges from 928.7 to 1,125.6 m (3,047–3,693 ft). The
average elevation of the ground surface is 1,676.4 m (5,300 ft).
The square marked in Fig. 1(a) outlines the selected study area
according to the recorded significant movements by the field
instruments. Most of the open tunnels were excavated in a horse-
shoe shape. Backfilling activities that took place in the middle
northern part of the interested area (labeled in Fig. 1), are shown
in a rectangular shape. Fig. 2 gives the dimensions of the tunnel
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

cross sections. The selected tunnels were excavated from the year
of 2004 to 2010.
In this underground mine, the rock supports, including the resin
bolts, split sets, swellex bolts, and cable bolts, have been installed to
support the tunnels. At an early stage of the previously mentioned
excavations, the 2.44-m resin bolts and 1.83-m split sets were
applied on the roof and the ribs, respectively. Additional reinforce- Fig. 2. Dimensions of the (a) horseshoe tunnel; and (b) rectangular
ment was applied on the roof using the 6.1-m cable bolts, and finally tunnel.
the 6.1- and 3.66-m swellex bolts were applied on the roof and ribs.

Fig. 1. Overall tunnel system at the underground mine: (a) plan view; and (b) elevation view (seen from the south).

© ASCE 05019004-3 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004


boundary, and then gradual reduction of the pressure/traction is boundary conditions, the block and joint constitutive models, the
assumed to represent the excavating process. Delayed installation estimation of the post-failure properties of the strain-softening
of supports, hence, can be simulated at a certain stress loss factor. block constitutive model, and the discontinuity and rock support
property values. Briefly, the strain-softening constitutive model
based on the Mohr-Coulomb criterion is used to describe the me-
Three-Dimensional Numerical Modeling chanical behavior of the rock masses. A continuously yielding joint
model is specified for faults, to simulate the dependence of joint
Development of the Three-Dimensional Numerical Model stiffnesses on the normal stress. The “cable” structural element
(Itasca Consulting Group, Inc. 2007), which provides both tension
A 3D numerical model was created by using the 3DEC version 4.1 and shear resistance along the support, is applied for the rock sup-
software package (Itasca Consulting Group, Inc. 2007). As pre- ports. Table 1 gives the physical and mechanical property values
sented in Fig. 3(a), the dimensions of the model are 122 m in the used for the rock masses and backfilling material in the numerical
three perpendicular directions x, y, and z with the origin of the coor- model.
dinates located at the center; the y and x axes are in accord with the Following the field excavations and supporting procedures,
north and east directions in the field. The model includes lithologies the open tunnels were numerically excavated and supported in
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

OC5 and dike (nonplanar). Both nonpersistent and persistent faults nine steps, and the backfilled tunnels were excavated and back-
were generated, but for clear presentation only the major nonpersis- filled step by step (Xing et al. 2018a, b). For the support installa-
tent fault is plotted in Fig. 3(b), and it intersects with the tunnel sys- tion, procedures mentioned in the section entitled “Tunnel
tem and the dike layer. The tunnel system includes both open tun- Excavation and Support Installation” were developed by incorpo-
nels as well as the backfilled ones, mostly located in the footwall of rating a FISH function to 3DEC (Xing et al. 2018b). Numerical
the fault [behind the fault in Fig. 3(b)]. A detailed description of the predictions from Xing et al. (2018a, b) showed good agreements
model has been provided by Xing et al. (2018a, b). Also illustrated with the field measurements. Based on their results, a stress loss
in their study are the concept and use of the fictitious joints, the factor of 0.5 was applied; the lateral stress ratio (K0) of 1.0 was
specified for both the maximum and minimum lateral stress ratios
(KH and Kh).

Results of Analysis and Discussion


The fault, the dike interlayer, as well as the tunnel system are dis-
tributed in a complicated manner [Fig. 3(b)]. Different displace-
ment and stability scenarios are possible at different locations. After
checking out carefully, five representative vertical planes, X =
−30 m, X = 0 m, X = 27 m, Y = −20 m, and Y = 20 m, as presented in
Fig. 4, are chosen to present the numerical results.
The displacement vectors around the tunnels on the previously
mentioned multiple planes are given in Fig. 5. On the plane of X =
−30 m [Figs. 4 and 5(a)], one open tunnel exists with one backfilled
tunnel on the left and the other one above the right-side area [as la-
beled in Fig. 5(a)]; the weak interlayer dike is located on the roof.
The maximum displacement of 10.9 cm occurs on the roof of the
(a) tunnel. On the plane of X = 0 m [Figs. 4 and 5(b)], two open tunnels

Major non-persistent fault


Backfilled tunnels

Tunnel
system

Y=20 m

Dike layer

Y=-20 m
Backfilled tunnel Open tunnels

(b)

Fig. 3. Geological features and tunnels built in the numerical model: X=-30 m X=0 m X=27 m
(a) lithologies OC5 and dike; and (b) major nonpersistent fault, dike
layer, and tunnel system (most of the backfilled tunnels are located Fig. 4. Plan view of the selected cross sections to present numerical
behind the fault). results.

© ASCE 05019004-4 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004


are located respectively on the footwall [left-side tunnel in
Backfilling Fig. 5(b)] and the hanging wall [right-side tunnel in Fig. 5(b)] of the
major nonpersistent fault. Several locations to the left of the left-
side tunnel were excavated and backfilled. Because of the many en-
Dike gineering activities on the footwall and the relief from the fault, the
left-side tunnel suffers higher deformations than the right-side one,
especially on the roof and right wall. The peak value is 6.66 cm, as
provided in Fig. 5(b). On the plane of X = 27 m [Figs. 4 and 5(c)], a
Open tunnel tunnel intersection exists below the major fault at a relatively far
Backfilling
X=-30 m distance. Even though the tunnel intersection has a wider span, the
(a) deformations are smaller than the two previous conditions [compare
Fig. 5(c) with Figs. 5(a and b)], due to the fewer engineering activ-
ities and better geological conditions. On the plane of Y = 20 m
Major non-persistent fault [Figs. 4 and 5(d)], large deformations are observed above the left-
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

side tunnel, where the weak dike layer, the fault, and the extensive
excavation and backfilling activities exist. Significant movements
are present on the middle part of the right wall due to the existence
of surrounding backfill and nearby weak dike layer. By comparison,
Dike the right-side tunnel, which is far away from the complicated geo-
Open tunnels logical conditions and engineering activities, is relatively stable.
Backfilling Finally, the displacements around the tunnel on the plane of Y =
X=0 m −20 m are the lowest, maximized at 4.1 cm; the slight asymmetric
(b) displacement distribution is likely caused by the dike layer that
exists below. It should be mentioned that as a 3D problem, the pre-
sented results can also be affected by the out-of-plane factors.
To show the interior rock mass behavior starting from the exca-
Major non-persistent fault
vation surfaces, vertical lines are placed above the roofs and below
the floors to display the vertical displacements, and horizontal lines
perpendicular to the tunnel walls are used to provide the horizontal
inward displacements. Figs. 6(a–e) show the variations of the dis-
placements with the distance from the tunnel surface on the roof,
Open tunnel floor, and walls. In general, high displacements are concentrated
X=27 m within the 2–3-m thickness from the tunnel surface; after 4 m, the
displacements either change slightly or stay constant with the dis-
(c)
tance. However, some exceptions and distinctions should be
noticed. In accordance with the results presented in Fig. 5(a),
Fig. 6(a) illustrates that the tunnel on the plane of X = −30 m has a
Major non-persistent fault
considerable roof sag of 10.2 cm and a deep extent of high displace-
ment on the roof, which is at least 4 m. As presented in Fig. 6(b), the
roof of the left-side tunnel on the plane of X = 0 m is significantly
Dike
active, with a vertical displacement of −6.32 cm (ROOF-L); the two
ribs close to the fault (RIGHT WALL-L and LEFT WALL-R) have
Backfilling
Open tunnels greater extents of high displacements than the other locations. On
the vertical plane of Y = 20 m [Fig. 6(d)], it can be seen that the im-
Y=20 m mediate roof of the left-side tunnel (ROOF-L) has a much smaller
(d) deformation than the interior rock masses; this is abnormal. It is
because the roof has been excavated and backfilled subsequently.
The right wall of the left-side tunnel [RIGHT WALL-L in
Fig. 6(d)], however, seems extremely unstable, indicating the effect
of many excavation activities and the presence of a nearby weak
dike [Fig. 5(d)]. In addition, as the distance from the wall keeps
increasing and approaches the right-side tunnel, the direction of the
horizontal displacement changes, being positive [RIGHT WALL-L
in Fig. 6(d)]. Correspondingly, direction reverse can also be
Dike
observed for the displacement inside the left wall of the right-side
Open tunnel tunnel [LEFT WALL-R in Fig. 6(d)]. In summary, the displace-
Y=-20 m
ments around the tunnels have exhibited site-dependent variations.
More attention should be paid as a fault, a weak material, or fre-
(e)
quent engineering activities occur close to a tunnel.
For the strain-softening model in 3DEC, the strength parameters
Fig. 5. Displacement vectors around the tunnels on the vertical planes:
degrade once the plastic strain takes place. Instead of using the con-
(a) X = −30 m; (b) X = 0 m; (c) X = 27 m; (d) Y = 20 m; and (e) Y =
ventional plastic indicator in 3DEC, the rock masses that reach the
−20 m (unit: m).
residual strength, either in shear or in tension, are considered as

© ASCE 05019004-5 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004


0.05
ROOF
failed. Results show that most of the rock masses around the tunnels
Displacement (m) 0.03
FLOOR have failed in shear. Hence, the distributions of the cohesion values
0.01 are plotted for analysis, as presented in Fig. 7. According to the re-
LEFT WALL
-0.01 sidual values listed in Table 1 (0.78 and 0.6 MPa for OC5 and dike,
RIGHT WALL
-0.03 respectively), the rock masses around the tunnels are observed as
-0.05 failing; outside the failed area, the strength degradation area can
-0.07 also be seen. Fig. 7(a) shows that around the tunnel on the plane of
-0.09
X = −30 m, the failed area has extended to the left-side and upper-
-0.11
0 2 4 6 8 10
right backfilled regions, as well as to the dike on the roof. On the
Distance from the tunnel surface (m) plane of X = 0 m [Fig. 7(b)], failures around the left-side tunnel are
(a) also connected to the left-side backfilling area; failures on the two
ribs close to the fault have a tendency of extending to the fault.
0.07
Under the better geological conditions, the rock masses around the
ROOF-L tunnel intersection on the plane of X = 27 m [Fig. 7(c)] show lower
0.05 FLOOR-L
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

failure extent. In contrast, because of the existence of the dike and


0.03 RIGHT WALL-L
Displacement (m)

ROOF-R
the close distance between the tunnels on the plane of Y = 20 m
0.01
FLOOR-R [Fig. 7(d)], the area between the tunnels seems unstable with exten-
-0.01
LEFT WALL-R sive failures. On the plane of Y = −20 m [Fig. 7(e)], the failed area is
-0.03
RIGHT WALL-R smaller compared to the other cases and the extent is more or less
-0.05 symmetric around the tunnel.
-0.07 To assist the design of tunnel support, the extent of the plastic
0 2 4 6 8 10
Distance from the tunnel surface (m) yielding zone should be known. Table 2 provides the maximum
(b) extent of the strength degradation area around the various tunnels.
Note that the extent on the tunnel roof on the plane of X = −30 m,
0.05 ROOF 2.81 m represents the vertical distance above the crown. The plot in
0.03 FLOOR Fig. 7(a) shows that the failure or degradation area extends to the
Displacement (m)

LEFT WALL dike deeply on the upper-right side. Under such conditions, the bolt
0.01 RIGHT WALL length is suggested to be long enough to penetrate the dike layer
-0.01 (probably 5.3 m) to make the rock mass stable. Additionally, the
largest extent of 5.54 m appears on the roof of the left-side tunnel on
-0.03 the plane of Y = 20 m. Referring to Fig. 7(d), it can be found that the
-0.05 dike layer is above the roof, and beyond that the degradation area
0 2 4 6 8 10 continues to extend. The roofs of the left tunnel on the plane of X =
Distance from the tunnel surface (m) 0 m and the tunnel intersection on the plane of X = 27 m have
(c) strength degradation area extent about 4.4 m. For the rest of the tun-
0.08 ROOF-L nels, the strength degradation area on the roofs range from 3.34 to
0.05 FLOOR-L 3.71 m (Table 2). Typical thickness of the strength degradation area
on the floors is 3–4.6 m. For ribs, most of the yielding zone thick-
Displacement (m)

0.02 LEFT WALL-L


-0.01 RIGHT WALL-L nesses is between 4 and 5 m. Yet the paramount extent of 7.46 m
ROOF-R exists on the left wall of the right-side tunnel on the plane of X =
-0.04
FLOOR-R 0 m, which can be explained by the close distance to the fault
-0.07 LEFT WALL-R [Fig. 6(b)]. Due to the complicated geologies or mine construction
-0.10 RIGHT WALL-R activities, the size of the strength degradation area at some locations
-0.13
is unavailable. For example, the left area to the tunnels at X = −30 m
0 2 4 6 8 10
Distance from the tunnel surface (m) and at X = 0 m has been excavated and backfilled, and the failure
(d) covers a long distance. Nevertheless, it does not indicate that all the
places are unstable. Instead, the failures may take place before back-
0.05 filling, and the rock mass stability has been improved after backfill-
ROOF
ing. The two connecting ribs on the plane of Y = 20 m, however,
0.03 FLOOR
Displacement (m)

LEFT WALL
seem to have unfavorable situations indicating that the failures are
0.01 RIGHT WALL
connecting to the adjacent tunnels [Fig. 7(d)].
The bond shear and bolt axial tensile behaviors of the rock sup-
-0.01 ports are evaluated through the percentage of failed bonds and
-0.03 failed bolts, as given in Table 3. It shows that most of the bonds of
the split sets are in danger; over half of the resin bolts and small
-0.05 amount of swellex bolts on the roof have suffered bond shear fail-
0 2 4 6 8 10 ure. While the bonds of cable bolts and of swellex on the walls are
Distance from the tunnel surface (m)
(e) safe. With respect to the bolt itself, around 50% of the swellex bolts
and 28% of the resin bolts on the roof are in tensile failure.
Fig. 6. In-rock displacements around the tunnels on the vertical planes: The high displacements and large strength degradation area, as
(a) X = −30 m; (b) X = 0 m; (c) X = 27 m; (d) Y = 20 m; and (e) Y = well as the support failures indicate that the split sets (bolt length =
−20 m. 1.83 m) are badly insufficient with respect to the bond shear resist-
ance and length. Proper increase of the shear resistance and length

© ASCE 05019004-6 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004


Backfilling is also necessary for resin bolts (bolt length = 2.44 m). To improve
the safety of roof support bolts (resin and swellex bolts), a denser
Dike arrangement can be implemented based on the field performance of
the bolts. The length of bolts is suggested to be 3–4 m on the roof,
and 4–5 m for the ribs. However, if complex situations, that is, a
fault, a weak dike, or many excavation activities, are encountered,
the bolts should be longer.
Note that the preceding findings are based on the rock supports
Open tunnel applied in the numerical model at an early stage after excavations in
Backfilling the mine. On the other hand, rehabilitation work has been conducted
X=-30 m
by installing longer and stronger supports from time to time in the
(a) field. For example, the failed split sets have been replaced using the
2.44 m swellex or inflatable bolts; additional repair rehabilitation
Major non-persistent fault was applied using the 2.44 and 3.66 m swellex bolts. Because the
rehabilitation work is not included in the numerical model, the pre-
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

dicted failure conditions of the rock supports are higher than that
currently exist in the field; discrepancies may also exist in the failure
modes of the supports in the numerical model and that in the field.
Dike
According to the RMR system (Bieniawski 1989), the rock
Open tunnels
masses having the RMR values of 32.5 and 40 (the average values
Backfilling
in the present study) are classified as “poor rock.” For such rocks,
empirical support guidelines suggest fully grouted rock bolts of
X=0 m
4–5 m long at a spacing of 1–1.5 m in the crown and walls with wire
(b) mesh. Although the suggestions are applied for the tunnels with a
width of 10 m, it indicates that the conclusions made based on the
numerical results are reliable. Differences can be observed on the
Major non-persistent fault
length of roof bolts predicted by the two methods. It is believed that
the support design from the numerical modeling is accurate for this
specific problem based on the conducted rigorous simulation. In
this respect, the numerical results are validated by the field monitor-
ing data as discussed below.
As stated by Xing et al. (2018a, b), tape extensometers were in-
Open tunnel stalled at the mine to monitor the tunnel convergence. In their studies,
X=27 m the field measurements at two locations (MT-17 and MT-18) were
(c) evaluated in terms of the convergence strains (tunnel convergence
over tunnel size), which were subsequently used to verify the numeri-
cal results. Similar procedures were implemented in the study
Major non-persistent fault
described in this article. The calculated convergence strains at the two
Open tunnels locations are respectively 1.94 (MT-17) and 1.08% (MT-18), agree-
ing well with those (1.96 and 1.27%) obtained from the field monitor-
ing data. In fact, the authors have demonstrated that the cases having
the average rock mass and fault property values and with K0 values of
Backfilling 0.5–1.25 provided reasonable predictions (Xing et al. 2018a, b). As a
result, the numerical results and the suggestions made for selecting
Y=20 m Dike appropriate tunnel support in the present study are reliable.
(d)
Conclusions

This paper elaborated on 3D numerical simulations of tunnel stabil-


ity for an underground mine using the distinct element method.
Geological and engineering complexities have been included in the
numerical model. The post-failure softening behavior of the medium
quality rock masses and the dependence of joint stiffnesses on the
Dike
Open tunnel
normal stress were described by strain-softening block and continu-
ously yielding joint constitutive models. According to the construc-
tion time in the field, the tunnels were excavated, backfilled, and
Y=-20 m
supported sequentially, with the consideration of delayed support-
(e) ing. Conducted numerical analysis investigated the effect of the
complex geologies and engineering activities on tunnel stability.
Fig. 7. Distribution of cohesion values around the tunnels on the verti- The displacements and strength degradation area around tunnels
cal planes: (a) X = −30 m; (b) X = 0 m; (c) X = 27 m; (d) Y = 20 m; and were discussed in detail. Also provided were the failure conditions
(e) Y = −20 m. of the applied rock supports. Based on the numerical results, sugges-
tions were finally made for the selection of proper tunnel support.

© ASCE 05019004-7 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004


Table 2. Maximum extent of strength degradation area around the tunnels on different vertical planes

Tunnel cross section Roof (m) Floor (m) Left wall (m) Right wall (m)
X = −30 m 2.81 extended to Dike 2.99 NA 4.02
Extended to backfilled area
X = 0 m (L) 4.38 3.38 NA 4.27
Backfilled
X = 0 m (R) 3.48 4.24 7.46 4.92
X = 27 m 4.37 4.60 4.22 4.52
Y = 20 m (L) 5.54 3.75 NA NA
Extended to dike Backfilled Connecting right-side tunnel
Y = 20 m (R) 3.34 3.38 NA 4.41
Connecting left-side tunnel
Y = −20 m 3.71 3.49 4.36 4.33
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

Table 3. Failure condition of the applied rock supports

Failure type (%) Split sets (wall) Resin bolts (roof) Cable bolts (roof) Swellex bolts (roof) Swellex bolts (wall)
Bond failure 88.3 67.9 0 12.5 1.7
Bolt tensile failure 0 28.2 2.8 51.4 3.6

It turns out that the presence of the dike interlayer, the fault, and provided by the mining company through providing geological
many mine construction activities nearby have posed enormous and geotechnical data, rock core and mine technical tours, and
influence on the stability of tunnels. For instance, on the plane of allowing access to the mine to perform field investigations is very
X = −30 m, the weak interlayer has led to distinct roof sag and a much appreciated. The first author is grateful to the Chinese
large extent of high displacements above the roof [Figs. 5(a) and Scholarship Council and the University of Arizona Graduate
6(a)]; the failed regions on the roof has extended into the dike layer, College for providing scholarships to conduct the research
where instabilities can occur easily. On the plane of X = 0 m, due to described in this paper first as a visiting research student and then
the existence of the fault and the many engineering activities in the as a Ph.D. student at the University of Arizona.
footwall, the roof of the footwall tunnel has become less stable;
whereas the hanging wall tunnel has a distinct yield area on the rib
close to the fault. As the geological conditions improve with respect References
to the location, that is, on the planes of X = 27 m and Y = 20 m, the
Barton, N. 2002. “Some new Q-value correlations to assist in site character-
tunnels seem to be more stable; on the other hand, the larger span on
isation and tunnel design.” Int. J. Rock Mech. Min. Sci. 39 (2): 185–216.
the former plane has resulted in more deformations and failures. https://doi.org/10.1016/S1365-1609(02)00011-4.
Failure conditions of the rock supports show that the split sets on Barton, N., R. Lien, and J. Lunde. 1974. “Engineering classification of rock
the walls and the resin bolts on the roof had issues with respect to masses for the design of tunnel support.” Rock Mech. 6 (4): 189–236.
bond shear; the swellex bolts on the roof had undergone bolt tensile https://doi.org/10.1007/BF01239496.
failure. According to the observed rock mass behavior, the split sets Bieniawski, Z. T. 1976. “Rock mass classifications in rock engineering.” In
and the resin bolts are apparently insufficient in length; the bolt Vol. 1 of Proc., Symp. on Exploration for Rock Engineering, 97–106.
lengths on the roof and on the walls are suggested to be 3–4 m and Rotterdam, Netherlands: A. A. Balkema.
4–5 m. Longer and denser bolts, or even other stronger supports, Bieniawski, Z. T. 1988. “The rock mass rating (RMR) system (Geomechanics
classification) in engineering practice.” Rock Classification Systems for
should be applied at some locations when unfavorable situations are
Engineering Purposes, 17–34. Philadelphia: American Society for Testing
encountered, such as the occurrence of a weak rock material or a and Materials.
fault and the frequent and adjacent engineering construction activ- Bieniawski, Z. T. 1989. Engineering rock mass classifications: a complete
ities. Specific decisions should be made after combining the numer- manual for engineers and geologists in mining, civil, and petroleum en-
ical results with the field observations. gineering. New York: John Wiley & Sons.
Rock support suggestions made based on the numerical model- Cai, M. 2008. “Influence of stress path on tunnel excavation response—
ing are close to that estimated based on the RMR system. However, Numerical tool selection and modeling strategy.” Tunn. Undergr. Sp.
some differences exist. The accuracy of the numerical predictions Technol. 23 (6): 618–628. https://doi.org/10.1016/j.tust.2007.11.005.
has been verified by the field measurements. Hence, the suggestions Cantieni, L., and G. Anagnostou. 2009. “The effect of the stress path on
on the selection of rock supports made based on the numerical mod- squeezing behavior in tunneling.” Rock Mech. Rock Eng. 42 (2): 289–
318. https://doi.org/10.1007/s00603-008-0018-9.
eling are sound and reliable.
Choi, S. O., and H. S. Shin. 2004. “Stability analysis of a tunnel excavated
This study provides a comprehensive procedure for stability in a weak rock mass and the optimal supporting system design.” Int. J.
assessment and support design for underground excavation projects Rock Mech. Min. Sci. 41 (S1): 876–881. https://doi.org/10.1016/j
having complex geological environments and engineering activities. .ijrmms.2004.03.151.
Chu, B. L., S. C. Hsu, Y. L. Chang, and Y. S. Lin. 2007. “Mechanical
behavior of a twin-tunnel in multi-layered formations.” Tunn. Undergr.
Acknowledgments Sp. Technol. 22 (3): 351–362. https://doi.org/10.1016/j.tust.2006.06
.003.
The research was funded by the Centers for Disease Control and Coggan, J., F. Gao, D. Stead, and D. Elmo. 2012. “Numerical modelling of
Prevention under the Contract 200-2011-39886. The support the effects of weak immediate roof lithology on coal mine roadway

© ASCE 05019004-8 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004


stability.” Int. J. Coal Geol. 90–91 (Feb): 100–109. https://doi.org/10 study.” Int. J. Rock Mech. Min. Sci. 38 (2): 167–180. https://doi.org/10
.1016/j.coal.2011.11.003. .1016/S1365-1609(00)00038-1.
Cundall, P. A. 1988. “Formulation of a three-dimensional distinct Kanik, M., and Z. Gurocak. 2018. “Importance of numerical analyses for
element model—Part I. A scheme to detect and represent contacts in determining support systems in tunneling: A comparative study from the
a system composed of many polyhedral blocks.” Int. J. Rock Mech. trabzon-gumushane tunnel, Turkey.” J. Afr. Earth Sci. 143 (Jul): 253–
Sci. Geomech. 25 (3): 107–116. https://doi.org/10.1016/0148 265. https://doi.org/10.1016/j.jafrearsci.2018.03.032.
-9062(88)92293-0. Marinos, P., and E. Hoek. 2000. “GSI: a geologically friendly tool for rock
Eberhardt, E. 2001. “Numerical modelling of three-dimension stress rota- mass strength estimation.” In Proc., GeoEng2000 at the Int. Conf. on
tion ahead of an advancing tunnel face.” Int. J. Rock Mech. Min. Sci. 38 Geotechnical and Geological Engineering, 1422–1446. Lancaster, PA:
(4): 499–518. https://doi.org/10.1016/S1365-1609(01)00017-X. Technomic Publishers.
Hao, Y. H., and R. Azzam. 2005. “The plastic zones and displacements Ng, C. W. W., K. M. Lee, and D. K. W. Tang. 2004. “Three-dimensional
around underground openings in rock masses containing a fault.” Tunn. numerical investigations of new Austrian tunnelling method (NATM)
Undergr. Sp. Technol. 20 (1): 49–61. https://doi.org/10.1016/j.tust.2004 twin tunnel interactions.” Can. Geotech. J. 41 (3): 523–539. https://doi
.05.003. .org/10.1139/t04-008.
Hart, R., P. A. Cundall, and J. Lemos. 1988. “Formulation of a three- Palmström, A. 1995. “RMi—A rock mass characterization system for rock
dimensional distinct element model—Part II. Mechanical calculations engineering purposes.” Ph.D. thesis, Dep. of Geology, Univ. of Oslo.
Downloaded from ascelibrary.org by Lund University on 02/22/19. Copyright ASCE. For personal use only; all rights reserved.

for motion and interaction of a system composed of many polyhedral Palmström, A. 2000. “Recent developments in rock support estimates by
blocks.” Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 25 (3): 117–125. the RMi.” J. Rock Mech. Tunnelling Technol. 6 (1): 1–19. https://www
https://doi.org/10.1016/0148-9062(88)92294-2. .rockmass.net/ap/66_Palmstrom_on_Recent_developments_RMi.pdf.
Hoek, E. 1994. “Strength of rock and rock masses.” ISRM News J. 2 (2): 4–16. Riedmüller, G., and W. Schubert. 1999. “Critical comments on quantitative
Hoek, E., and E. T. Brown. 1997. “Practical estimates of rock mass rock mass classifications.” Felsbau 17 (3): 164–167. https://pure.tugraz
strength.” Int. J. Rock Mech. Min. Sci. 34 (8): 1165–1186. https://doi.org
.at/ws/portalfiles/portal/1660025/6549.pdf.
/10.1016/S1365-1609(97)80069-X.
Rocscience. 2011. Phase2, Version 8.0. Toronto: Rocscience Inc.
Hoek, E., C. Carranza-Torres, and B. Corkum. 2002. “Hoek-Brown
Sari, Y. D., A. G. Pasamehmetoglu, E. Cetiner, and S. Donmez. 2008.
criterion—2002 ed.” In Vol. 1 of Proc., NARMS-TAC Conf., 267–273.
“Numerical analysis of a tunnel support design in conjunction with em-
Toronto: University of Toronto Press.
pirical methods.” Int. J. of Geomech. 8 (1): 74–81. https://doi.org/10
Hoek, E., and M. S. Diederichs. 2006. “Empirical estimation of rock mass
.1061/(ASCE)1532-3641(2008)8:1(74).
modulus.” Int. J. Rock Mech. Min. Sci. 43 (2): 203–215. https://doi.org
Shreedharan, S., and P. H. S. W. Kulatilake. 2016. “Discontinuum-equiva-
/10.1016/j.ijrmms.2005.06.005.
lent continuum analysis of the stability of tunnels in a deep coal mine
Hsiao, F. Y., C. L. Wang, and J. C. Chern. 2009. “Numerical simulation of
using the distinct element method.” Rock Mech. Rock Eng. 49 (5): 1903–
rock deformation for support design in tunnel intersection area.” Tunn.
Undergr. Sp. Technol. 24 (1): 14–21. https://doi.org/10.1016/j.tust.2008 1922. https://doi.org/10.1007/s00603-015-0885-9.
.01.003. Stille, H., and A. Palmström. 2003. “Classification as a tool in rock engi-
Huang, F., H. Zhu, Q. Xu, Y. Cai, and X. Zhuang. 2013. “The effect of neering.” Tunn. Undergr. Sp. Technol. 18 (4): 331–345. https://doi.org
weak interlayer on the failure pattern of rock mass around tunnel— /10.1016/S0886-7798(02)00106-2.
Scaled model tests and numerical analysis.” Tunn. Undergr. Sp. Vardakos, S. S., M. S. Gutierrez, and N. R. Barton. 2007. “Back-analysis of
Technol. 35 (Apr): 207–218. https://doi.org/10.1016/j.tust.2012.06.014. Shimizu Tunnel No. 3 by distinct element modeling.” Tunn. Undergr.
Itasca Consulting Group, Inc. 1998. UDEC, version 3.0. Minneapolis: Space Technol. 22 (4): 401–413. https://doi.org/10.1016/j.tust.2006.10
Itasca. .001.
Itasca Consulting Group, Inc. 2007. 3DEC, version 4.1. Minneapolis: Xing, Y., P. H. S. W. Kulatilake, and L. A. Sandbak. 2018a. “Investigation
Itasca. of rock mass stability around tunnels in an underground mine in USA
Janin, J. P., D. Dias, F. Emeriault, R. Kastner, H. Bissonnais, and A. using three-dimensional numerical modeling.” Rock Mech. Rock Eng.
Guilloux. 2015. “Numerical back-analysis of the southern Toulon tunnel 51 (2): 579–597. https://doi.org/10.1007/s00603-017-1336-6.
measurements: A comparison of 3D and 2D approaches.” Eng. Geol. Xing, Y., P. H. S. W. Kulatilake, and L. A. Sandbak. 2018b. “Effect of rock
195 (Sept): 42–52. https://doi.org/10.1016/j.enggeo.2015.04.028. mass and discontinuity mechanical properties and delayed rock support-
Jing, L. 2003. “A review of techniques, advances and outstanding issues in ing on tunnel stability in an underground mine.” Eng. Geol. 238 (May):
numerical modelling for rock mechanics and rock engineering.” Int. J. 62–75. https://doi.org/10.1016/j.enggeo.2018.03.010.
Rock Mech. Min. Sci. 40 (3): 283–353. https://doi.org/10.1016/S1365 Yalcin, E., Z. Gurocak, R. Ghabchi, and M. Zaman. 2016. “Numerical anal-
-1609(03)00013-3. ysis for a realistic support design: Case study of the Komurhan Tunnel
Kaiser, P. K., S. Yazici, and S. Maloney. 2001. “Mining induced stress in Eastern Turkey.” Int. J. of Geomech. 16 (3): 05015001. https://doi.org
change and consequences of stress path on excavation stability—A case /10.1061/(ASCE)GM.1943-5622.0000564.

© ASCE 05019004-9 Int. J. Geomech.

Int. J. Geomech., 2019, 19(5): 05019004

You might also like