archives of civil and mechanical engineering 18 (2018) 198214

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: http://www.elsevier.com/locate/acme

Original Research Article

Mechanical behavior of an opening in a jointed

rock-like specimen under uniaxial loading:
Experimental studies and particle mechanics
approach

R.H. Cao a,b, P. Cao a,*, H. Lin a, G.W. Ma b, X. Fan c, X.G. Xiong a
a
School of Resources and Safety Engineering, Central South University, Changsha, Hunan 410083, China
b
School of Civil, Environmental and Mining Engineering, The University of Western Australia, Perth 6009, Australia
c
School of Highway, Chang'an University, Xi'an 710064, China

article info abstract

Article history: Generally, in many cases of rock engineering, the openings often constructed in rock-mass
Received 13 April 2017 containing non-persistent joints. However, comparing with the previous works, few studies
Accepted 27 June 2017 investigate the failure or damage due to the crack propagation and coalescence around an
Available online opening. Based on the uniaxial compression tests and particle ow code (PFC) the interaction
effect of opening and joints on the crack coalescence behavior around an opening are
Keywords: investigated in this study. From the view of experimental and numerical results, strength
Joint parameters are mainly effected by joints (inclination and distance). Specically, the uniaxial
Opening compressive strength of jointed specimen (UCSJ) and elastic modulus of jointed specimen (EJ)
Failure characteristics of specimens decrease for 08 a 458 and increase for a > 458. UCSJ and EJ increases with
Uniaxial compression increasing joint distance (d) for all joint inclination angel (a) values, with the highest and
Peak strength lowest strengths obtained for d = 50 mm and d = 20 mm, respectively. The opening has a great
inuence on the failure mode of jointed specimen. Unlike previous results, in this study,
jointed specimens present four new kinds of failure modes: Mode-I (horizontally symmetrical
splitting failure); Mode-II (stepped failure at opening sides); Mode-III (failure through a plane);
Mode-IV (mixed failure). The strength parameters and failure modes in the numerically
simulated and experimental results are in good agreement, and the results are expected to be
useful in predicting the stability of an opening in a non-persistently jointed mass.
2017 Politechnika Wrocawska. Published by Elsevier Sp. z o.o. All rights reserved.

1. Introduction
deformation behavior of a rock-mass is induced by the
propagation and coalescence of cracks and joints under
Rock-mass are typically heterogeneous materials containing external loads, hence, the mechanical characteristics and
defects such as cracks, joints, and pores. The nonlinear failure patterns of the rock-mass are difcult to predict. The

* Corresponding author.
E-mail address: caowei198804@126.com (P. Cao).
http://dx.doi.org/10.1016/j.acme.2017.06.010
1644-9665/ 2017 Politechnika Wrocawska. Published by Elsevier Sp. z o.o. All rights reserved.
 archives of civil and mechanical engineering 18 (2018) 198214
mechanical behavior of jointed rock-mass is extremely
interesting for engineers and scientists. Over the past decades,
the failure behaviors of rock or rock-like materials have been
extensively investigated and discussed. Previous works have
promoted the understanding of the failure characteristics
obtained from specimens contain ssures or joints. Substan-
tial experiments have focused on the study of crack initiation
and propagation under compressive tests for the specimen
with single joint. And the results show that two types of cracks
will initiate from the pre-existing ssure, namely, wing and
secondary cracks [17].
Apart from the single joint or crack, the coalescence mode
between parallel or un-parallel cracks has been investigated
through experiment or numerical simulation as well [812].
Meanwhile, for the parallel cracks, based on the experiment of
gypsum specimens, Shen [8] observed three types of linkage
between two aws: tensile coalescence, shear coalescence,
and mixed coalescence, and the inclination angle of rock-
bridge signicantly inuenced the coalescence pattern. Bobet
and Einstein [9] produced gypsum specimens with two parallel
aws and identied ve types of failure pattern. Sagong and As a consequence, the fracture process and failure character-
Bobet [10] carried out uniaxial compression tests on rock-like
specimens with three and 16 aws, hence, nine types of
coalescence mode are observed. The aw geometry dened by
the aw inclination angle and the ligament angle has an
extensive effect on coalescence mode. Furthermore, for the
un-parallel cracks, Lee and Jeon [11] investigated the penetra-
tion mode between two un-parallel ssures, where the
coalescence occurred mainly through the tensile cracks or
tensile and shear cracks. Based on parallel bonded-particle
models, Zhang et al. [12] studied the crack coalescence
between two non-parallel aws and observed ve types of
linkage between two aws: tensile crack linkage, tensile crack
linkage with shear coalescence at the tip, shear crack linkage,
mixed linkage, and indirect crack linkage.
Moreover, compared with the in situ test and laboratory
test, the numerical simulation is an economical and practical
method to simulate the failure process of jointed rock-masses.
In recent years, many numerical methods have been used to
simulate the crack initiation and coalescence in brittle
materials. These include the nite element method (FEM) or
extended nite element method (XFEM) [1315], smoothed
particle hydrodynamics [16,17], displacement discontinuity
analysis (DDA) [18,19], particle ow code (PFC) [2024]. In the
framework of FEM, Tang [13] developed a new numerical code
RFPA2D, which can be used to simulate the failure processes
of rock material in geo-mechanical problems. Moes and
Belyschko [14] analyzed the growth of arbitrary cohesive
cracks by using extended nite element method, and the
results show that the proposed approach avoids the evalua-
tion of stresses at the mathematical tip of the crack. In the
framework of SPH, Pramanik and Deb [16] developed an
efcient methodology to model the behavior of pre-existing
multiple intersecting discontinuities in rock material. Based
on the DDA, Ning et al. [19] proposed a new methodology to
simulate the opening and sliding along pre-existing disconti-
nuities and can also reproduce the major failure mechanisms
observed in footwall slope collapses. For the PFC, most of
scholars used it to model the failure behavior of jointed
specimens, such as un-parallel joints [22], multi-ssures
199
specimens [21,23,24] and ubiquitous-joint brittle rock-like
specimens [20]. Overall, the numerical results show good
agreement with experimental results.
The previous works have promoted the understanding of
crack propagation, coalescence, and failure modes of brittle
materials with ssures. Rock-mass are heterogeneous materi-
als containing defects including cracks and pores, especially
for the jointed rock-mass around tunnel or underground
space. Generally, in many cases of rock engineering, the
construction of opening is not suitable in rock mass with fully
persistent joints [26]. Instead, the openings often constructed
in rock-mass containing non-persistent joints [27]. However,
comparing with the previous works, few studies investigate
the failure or damage due to the crack propagation and
coalescence around an opening. In the loading process, the
initiation or development of fractures may be caused by
the change of joint geometry (inclination and distance) or the
stress around the opening. Pre-existing joints and opening will
inuence each other. The interaction effect increases the
complexity and unpredictability of the mechanical properties.
istics of a jointed rock-mass around an opening will be
different from those of a multiple jointed rock-mass. The
existence of joints alters the stress distribution pattern around
underground openings. This may result in different failure
modes of rock-mass around the opening compared with
the modes observed in previous studies. In this paper, we
combine experiments using rock-like material and PFC2D to
investigate the mechanical parameters, rock fracture beha-
viors around an opening in a rock-mass having many non-
persistent joints.
2. Specimen preparation and testing
2.1. Specimen preparation and test
Previous works have used numerous kinds of materials to
produce rock-like specimens. These include glass [3], gypsum
[810,27,28], white cement and sand [10,12,20,21,2931], and
Columbia Resin [32]. Given that white cement and sand can
produce brittle specimen and that sand can provide the
frictional behavior of the modeling material, cement mortar is
chosen as the rock-like material for our experiment.
Rock-like specimens are made of white cement, water, and
sand. The volume proportions for all the specimens are shown
as: V (water):V (white cement):V (silica sand) = 3:3:2. The
dimensions (height  width  thickness) of specimens are
200 mm  150 mm  30 mm. Pre-existing ssures are created
by inserting the mica sheets (0.6-mm thick, 20-mm long) into
the fresh cement mortar paste at the location of the ssures.
All specimens are kept inside the standard curing box for a
xed number of days (28 days) before being subjected to
mechanical testing.
Previous works investigated the inuence of joint orienta-
tion [113,2033], joint persistence [23,24,29,33] and joint
distance [20,23,29,33] on mechanical behavior of jointed rock
mass. However, in this study, we focus on the interaction
effect of opening and joints and the crack coalescence
behavior around an opening. Thus, among the many
200 archives of civil and mechanical engineering 18 (2018) 198214

Fig. 1 Schematics of fissure geometry configurations in the specimens.

parameters of joint geometry, the joint inclination angle and
distance is considering in this study. Joint geometry is dened
by two geometrical parameters as shown in Fig. 1a: joint
inclination angle a and joint distance d. The joint inclination
angle a in our specimens varies between 08, 308, 458, 608, and
758, whereas the joint distance d is 20 , 30 , 40 , and 50 mm,
respectively. Table 1 provides the ssure geometry informa-
tion for all the specimens in this research. Each specimen is
assigned with an ID number using the notation S-a-b, where S
is the sample, a is the value of joint angle in degrees, and b is
the value of joint distance. Fig. 1b shows the joint geometry
congurations generated for d = 20 as a set of examples (S-20-0,
S-20-30, S-20-45, S-20-60, S-20-75).
This experiment used the loading control system DCS-200
and the servo control uniaxial loading instrument. The
specimen is positioned between the two loading platforms,
and the loading rate is set as 0.2 mm/min. The top and bottom
boundaries are xed in the horizontal direction. To reduce the
inuence of the end effect, the contact areas of the specimen
and the loading platform are coated with butter. Monitored by
a video recorder, the specimens are then loaded under
compression until failure.
To obtain the mechanical parameters of the intact material,
uniaxial compression tests (intact specimens with dimension
of 200 mm  150 mm  30 mm) and Brazilian disk splitting
tests (disk specimen with a 50-mm diameter and 30-mm
thinness) are initially carried out on the intact material (Fig. 2).
For the rock-like material, the macro-parameters are shown in
Table 2. The average values of unit weight (gm), Young's
modulus of intact specimen (EI), uniaxial compressive strength
of intact specimen (UCSI), and Poisson's ratio (v) of the rock-
material are gm = 2.159 g/cm3, EI = 5.64 GPa, UCSI = 21.73 MPa,
and v = 0.243, respectively, with the uniaxial tensile strength
(st) at 1.355 MPa.
2.2. Numerical specimen formation and calibration of
micro-mechanical parameters
As mentioned above, particle ow code in two dimensions
(PFC2D) is a commonly used commercial discrete element
method (DEM) code. And it also has shown great advantages in
investigating the failure behavior in rocks or rock-like
materials. Especially for the parallel-bond model, which has
been widely used by scholars [11,12,2024]. In this model

Table 1 Sample numbers and fissure geometrical parameter values used for all specimens.
Number Specimen ID d a Number Specimen ID d a
1 S-20-0 20 0 11 S-40-0 40 0
2 S-20-30 20 30 12 S-40-30 40 30
3 S-20-45 20 45 13 S-40-45 40 45
4 S-20-60 20 60 14 S-40-60 40 60
5 S-20-75 20 75 15 S-40-75 40 75
6 S-30-0 30 0 16 S-50-0 50 0
7 S-30-30 30 30 17 S-50-30 50 30
8 S-30-45 30 45 18 S-50-45 50 45
9 S-30-60 30 60 19 S-50-60 50 60
10 S-30-75 30 75 20 S-50-75 50 75
 archives of civil and mechanical engineering 18 (2018) 198214 201

Fig. 2 Calibration scheme for intact material and flat joint.

Table 2 Experimental results for the macro-mechanical It is generally known that the calibration of micro-
parameters of the intact material. mechanical in PFC is very complex, and there is no certain
Experimental results relationship between the micro-parameters and macro-pa-
rameters (Young's modulus, uniaxial compressive strength,
Uniaxial compressive strength of 21.73
and Poisson's ratio). Currently, the calibration method named
intact specimen, UCSI (MPa)
Young's modulus of intact 5.64 trial-and-error approach is an accepted method and has
specimen, EI (GPa) been widely used in the previous works [11,12,2024]. Notably,
Poisson ratio, v () 0.243 it is not considered state-of-the-art but it can nonetheless lead
Uniaxial tensile strength, st (MPa) 1.355 to successful results. Table 3 lists the micro-parameters
obtained through a series of trial and error processes.
Notably, the calibration is performed only under consideration
(as shown in Fig. 3), each particle is in contact with the of the compressive stress states. At the same time, the
neighboring particles via the parallel-bond. Actually, the bond comparison between the experimental and numerical results
of the parallel-bond model is depicted as a rectangle of is described in Table 4. Clearly, the macro-parameters of the
cement-like material which can transmit both force and intact specimens obtained through PFC2D are similar or equal
moment between the particles [34]. Moreover, in this model, to those obtained experimentally.
bond breakage results in immediate decrease in macro The scale of the specimens in the numerical simulation is
stiffness [34,35]. In this study, PFC2D (parallel-bond model) equal to that of the specimens in the experiment whose width
was also used to further investigate the failure characteristics and height are 150 mm and 200 mm, respectively. Fig. 4 shows
of jointed specimens under compression loading. the numerical specimen generated by PFC2D, the scale of the

Fig. 3 Illustration of the parallel bond model in PFC2D.

202 archives of civil and mechanical engineering 18 (2018) 198214

Table 3 Microscopic parameters for rock mass based on the calibration method trial-and-error approach.
Micro-parameters Values Remarks
Minimum mean radius, r (mm) 0.3 Uniform distribution
Ratio of maximum to minimum radius of the particle, Rmax/Rmin () 1.66
Particle contact modulus, Ec (GPa) 4.5
Ratio of normal to shear stiffness of the particle, kn/ks () 2.9
Particle friction coefcient, m () 0.5
Parallel bond modulus, Ec (GPa) 4.5
Ratio of normal to shear stiffness of the parallel bond, kn =ks () 2.9
Parallel bond normal strengths, sn (MPa) 15.0 Normal distribution
Parallel-bond normal strength, standard deviation (MPa) 2 13.3% of mean value
Parallel bond shear strengths, tn (MPa) 15.0 Normal distribution
Parallel-bond shear strength, standard deviation (MPa) 2 13.3% of mean value

Table 4 Comparison between experimental and numerical results for intact material macro-mechanical parameters.
Experimental results Numerical results
Uniaxial compressive strength, UCSI (MPa) 21.73 21.96
Young's modulus, EI (GPa) 5.64 5.69
Poisson ratio, v () 0.243 0.251

numerical specimen is equal to that of the specimens in the
experiment, and the white circles are the joints. For the joint
specimens in experiment, the joints were created by inserting
mica sheets into the fresh cement mortar paste and the mica
sheets were not removed before testing. Based on the direct
shear experiments (Fig. 2), the cohesion (Cj) and friction angle
(wj) between the cemented material and the mica sheet is
18.2 KPa and 118, respectively. Usually, the micro-mechanical
parameter values assigned for the particles that represent
joints are smaller than those for the particles that represent
intact material. In order to determine other micro-parameters
of joint particle, direct shear tests under constant normal
stress tests are carried out on numerical models, the normal
stress is 0.2 MPa, 0.4 MPa, 1.0 MPa, 1.5 MPa and 2.0 MPa. Like
the calibration process for UCSI and EI, the micro-mechanical
parameters for joint are determined by a series of trial and
error processes (as shown in Table 5).
A comparison between the experimental and numerical
results for peak shear stress of at joint is provided in Fig. 5. It
indicates that the simulated peak shear stress the at joint
specimens show a great agreement with those obtained
experimentally.

Fig. 4 Numerical intact specimen generated in PFC2D in this research (S-20-45).

 archives of civil and mechanical engineering 18 (2018) 198214 203

Table 5 Microscopic parameters for joints.

Joint particle friction coefcient, m () 0.08
Joint particle normal stiffness, kn (N/m) 250
Joint particle shear stiffness, ks (N/m) 250
Joint particle normal bond strength, sn (MPa) 0
Joint particle shear bond strength, tn (MPa) 0

3. Strength and deformation behavior of

jointed specimens

3.1. The relationship between strength and joint geometric

conguration

Figs. 6 and 7 shows the peak strength obtained through the

numerical (PFC2D) and experimental versus the inclination
angle (a) and joint distance (d), respectively. It can be seen from
the Fig. 6, the peak strength of the specimens decreasing as a
increases from 08 to 458 and increasing as a increases from 458
to 758. In the numerical simulation when inclination angle a
varies from 08 to 458, the specimens with joint distance d = 20 ,
30 , 40 , and 50 mm show an decrease of about 35.08%, 16.01%,
13.61%, and 17.41% in UCSJ, respectively. However, when a
changes from 458 to 758, the UCSJ show an increase of about
33.56%, 13.51%, 16.43%, and 16.69%, respectively. Overall, the
experimental results show a similar trend with the numerical
results.
The relationship between UCSJ and the joint distance can
be seen in Fig. 7. Joint distance d affect the UCSJ in all the
specimens; UCSJ increases with increasing d for all a values,
with the highest strength obtained for d = 50 mm and the
lowest strength for d = 20 mm. It can be seen from Fig. 7a and b,
in the numerical results, the 258, 308 and 458 samples show an
increases of 15.94%, 40.03%, 42.5% in UCSJ when d changes
from 20 to 50, respectively. However, for higher values of a
there is signicant change in UCSJ with increasing d, especially
for the 608 sample which shows an increase of about 61.76% in
UCSJ when d varies from 20 to 50 mm in the numerical result.
Fig. 5 Joint peak shear stress comparisons between
experimental and numerical results.
3.2. The relationship between elastic modulus and joint
geometric conguration
In Figs. 8 and 9, the vertical axis is the elastic modulus of the
heavily-jointed specimen (EJ). In both of numerical and
experimental results, EJ was estimated using the tangent
modulus at the UCSJ value of the stressstrain curve. Like the
UCSJ as shown in Figs. 6 and 7, the EJ for all specimens show a
similar tendency with UCSJ when a changes from 08 to 758
(Fig. 8a and b). And in the numerical simulation, the d = 20 mm,
30 mm, 40 mm and 50 mm samples show a drop of about
14.55%, 10.92%, 8.71% and 4.37% on EJ when a changes from 08
to 458, respectively. And, from 458 to 758, the EJ increased about
4.24%, 14.91%, 10.58% and 7.33%, respectively. As shown in
Fig. 8a and b, the effect of a on EJ is conspicuous. This indicates
that joints seem also to play an important role with respect to
the deformability of jointed samples.

Fig. 6 Effect of fissure number on UCSJ: (a) numerical result and experimental value for specimens which joint distance is
20 mm and 30 mm; (b) numerical result and experimental value for specimens which joint distance is 40 mm and 50 mm.
204 archives of civil and mechanical engineering 18 (2018) 198214

Fig. 7 Effect of fissure number on UCSJ: (a) numerical result and experimental value for specimens which inclination angle is
08, 308 and 458; (b) numerical result and experimental value for specimens which inclination angle is 608 and 758.

Fig. 8 Effect of fissure number on EJ: (a) numerical result and experimental value for specimens which fissure angle is 20 mm
and 30 mm; (b) numerical result and experimental value for specimens which fissure angle is 40 mm and 50 mm.

Fig. 9 Effect of fissure number on EJ: (a) numerical result and experimental value for specimens which fissure angle is 08, 308
and 458; (b) numerical result and experimental value for specimens which fissure angle is 608 and 758.
 archives of civil and mechanical engineering 18 (2018) 198214 205

Fig. 10 Crack coalescence in different failure mode based on the laboratory tests.

Fig. 11 Failure process at different axial strain levels based on the numerical analyzes for a typical Mode-I (S-20-0).
206 archives of civil and mechanical engineering 18 (2018) 198214
The effect of d on specimen elastic modulus EJ is shown in
Fig. 9. Fig. 9 shows the EJ obtained through the PFC2D and
experiments versus the joint distance (d). As shown in Fig. 9a is
the relationship between EJ and d for specimens with
inclination angle a are 08, 308 and 458. Fig. 9a clearly shows
that joint distance (d) has a very high inuence on EJ. EJ
increases with increasing d for all a values, with the highest
value obtained for d = 50 mm and the lowest strength for
d = 20 mm. Fig. 9b clearly shows the relationship between EJ
and d for specimens with inclination angle a are 608 and 758.
The EJ for specimens with 608 and 758 show a similar tendency
with those in Fig. 9a. With the increase of d, the EJ showed an
increasing trend.
4. Characteristics of crack coalescence and
failure mode
As mentioned above, the strength parameters of specimens
are mainly effected by the joints geometry. And the opening
seem to play only a minor role with respect to the strength of
the specimens. However, the opening has a great inuence on
the failure mode of the specimen. Unlike the previous studies
for multi-joint specimens [10,20,21,23,24,29,33], the specimens
present 4 new kinds of failure modes: Mode-I (horizontally
Fig. 12 Failure process at different axial strain levels based on the numerical analyzes for a typical Mode-II (S-20-30).
symmetrical splitting failure); Mode-II (stepped failure at
opening sides); Mode-III (failure through a plane); Mode-IV
(mixed failure). Fig. 10 shows the laboratory images and
sketches of the crack conguration for each failure mode.
Notably, the symbols S, M, T represents the shear crack, mixed
crack and tensile crack, respectively. Figs. 1114 further
illustrate the failure process of jointed specimens obtained
by numerical simulation under uniaxial compression.
4.1. Mode-I
For this mode, the specimens mainly failure from the tensile
cracks propagation and coalescence (Fig. 10a), and this mode is
more prominent in the specimen with a = 08. During testing,
tensile cracks originate from the ssure tips. With increasing
axial stress, the tensile cracks continue to grow and the
ssures link through the tensile cracks to form the failure. At
the same time, visible surface spalling appears around the
opening. Actually, this mode can be labeled as the splitting
failure mode, and the macro failure planes in this mode mainly
centralize in the upper part of the specimen and horizontally
symmetrical. However, in the previous results, the failure
planes for splitting failure distribute throughout the specimen
[24,33]. This indicates that the opening also has a great
inuence on the failure mode of the jointed specimens.
 archives of civil and mechanical engineering 18 (2018) 198214
Fig. 13 Failure process at different axial strain levels based on the numerical analyzes for a typical Mode-III (S-40-45).
The evolution of the micro/macro cracks at different axial
strain levels for S-20-0 (a typical failure Mode-I) is illustrated in
Fig. 11. When the ssure inclination angle is a = 08 (see Fig. 11a
e) in the loading process, the micro cracks initially appears
around the edge of the opening (Fig. 11a). However, no visible
macro crack appears in the specimen apart from the opening
edge. Thus, the damage rst occurs around the edge of the
opening. As the uniaxial load increased, different levels of
damage or failure are observed in the specimen (Fig. 11b).
When the axial stress reaches 77.3% of the UCSJ, and some
micro-cracks link and form macro-cracks; most of them belong
to tensile crack. And there are also many micro-cracks appear
around the pre-exiting joints.
When the specimen reaches the peak stress (Fig. 11c), the
cracks have got a further develop and there are some pre-
existing joints coalescence with others through tensile crack.
After the peak stress, the tensile cracks continue to propagate
and to form penetration. And the specimen undergoes
macroscopic failure, the pre-existing joints link with each
other to form the macro failure plane. As shown in Fig. 11d, the
applied load reached 52.6% of the post-peak stress and joints
are linked through a combination of tensile or mixed cracks.
With decreased axial stress, the cracks continue to propagate
and reach the edge of the specimen, resulting in the overall
failure.
207
4.2. Mode-II
Like the Mode-I, the specimens in this mode also mainly
failure from the tensile/mixed cracks propagation and
coalescence (Fig. 10b). This mode is more prominent in the
specimen with the joint distance d = 20 mm and 30 mm.
During testing, tensile cracks originate from the ssure tips,
then propagate and link with others to form macro failure
plane. As shown in Fig. 10b, there are several stepped failure
planes located at left and right sides of the opening, this mode
can also be labeled as the stepped failure mode. But for the
stepped failure in previous results, there are several
stepped failure planes in various locations of the specimen
[10,20,21,23,24,29,33].
Fig. 12ae illustrates the failure process of specimen S-20-45
(a typical failure Mode-II). Evidently, during the initiation stage
of loading (see Fig. 12a), the damage appears initially around
the edge of the opening and tips of joints. With further loading,
the micro-cracks gather and form macro cracks. When the
axial stress reaches 84.3% of the UCSJ, the macro cracks have
got a further develop and there are also some pre-existing
joints already form coalescence through tensile crack
(Fig. 12b).
When the axial stress reaches peak strength, several macro
failure planes occur in the specimen. As shown in Fig. 12c, the
208 archives of civil and mechanical engineering 18 (2018) 198214
Fig. 14 Failure process at different axial strain levels based on the numerical analyzes for a typical Mode-IV (S-50-60).
stepped failure plane is mainly concentrated on the right side
of the specimen and the macro cracks are also more evident
than those in Fig. 12b. After the peak stress and with the
propagation of the tensile cracks, the axial stress continues to
decrease (30.76% of post-peak) and the macro failure plane
occurs throughout the specimen (Fig. 12d). Afterwards, the
specimen begins to form an overall failure. When the axial
stress reaches 17.7% of the post-peak stress, the specimen
cannot maintain its integrity and stability because of the
propagated cracks (Fig. 12e). As shown in Fig. 12e, the stepped
failure planes mainly concentrated on the right and left sides
of the opening, the upper and lower part of the opening did not
exhibit obvious damage. This also keeps a great agreement
with the experimental results (Fig. 10b).
4.3. Mode-III
Failure through a plane means there is only a macro failure
plane in the specimen. Based on the experimental observation,
the macro failure plane formed through shear crack propaga-
tion and coalescence (Fig. 10c). This mode is more prominent
in the specimen with the inclination angle a = 458 and joint
distance d 40 mm. In the loading process, co-planar shear
crack develops from the ssure tips. With further loading,
shear cracks propagate and form macro shear bands which
become almost parallel to the diagonal line (Fig. 10c). Notably,
unlike the Modes-I and -II, surface spalling phenomenon is not
evident in this mode.
Fig. 13 displays the crack distribution at different axial
strain levels for S-40-45 (a typical failure Mode-III). Unlike
Modes-I and -II, in the loading process, the damage appears
rst at the tips of the pre-existing (Fig. 13a), whilst there is on
micro-crack appears around the edge of opening. With loading
continue, when the axial stress reaches 87.5% of peak stress,
the tensile cracks have got a further develop. At the same time,
there are also some shear cracks segments develop from the
tips of pre-existing joints, and some of them link with pre-
existing and form coalescence (Fig. 13b).
When the axial stress reaches peak strength, we can see a
potential shear band which may become parallel to the
diagonal line. This phenomenon means that some pre-
existing ssures form coalescence through the co-planar
shear crack. However, the overall failure does not occur at
this point. After the peak stress and with the propagation of
 archives of civil and mechanical engineering 18 (2018) 198214 209

the co-planar shear cracks, the axial stress continued to
decrease, and the shear band becomes increasingly visible. As
the axial stress reaches 41.6% of post-peak stress, the macro
shear failure plane formed through shear crack coalescence
(Fig. 13d). During the residual stage, shear crack stops
propagating, and some tensile cracks propagate with the
further loading. As shown in Fig. 13e, some tensile cracks
almost extend to the edges of the specimen.
4.4. Mode-IV
Similar to Mode-III, the specimen failure in the Mode-IV is
mainly caused by the shear crack propagation and coales-
cence. This mode is more prominent in the specimen with the
inclination angle a = 608 and 758. The pre-existing cracks on
the main failure plane mainly link with other failure planes
through co-planar shear cracks (as shown in Fig. 10d).
However, compared with the Mode-III, several shear bands
appear in this mode, and the location of the shear band is
adjacent and approximately parallel to the diagonal line of the
specimen. Apart from the shear band which is parallel to the
diagonal line, a pair of mixed cracks connects the shear bands
and the opening. And this is also the biggest difference for this
failure mode when compare with the shear failure mode in
previous works [20,21,23,24,29,33].
The evolution of cracks at different axial strain levels for S-
40-60 (a typical failure Mode-IV) is displayed in Fig. 14. Like the
Mode-III, the damage appears rst at the tips of the pre-
existing (Fig. 14a), and there is no micro-crack occurs around
the edge of opening. With further loading, the tensile cracks

Fig. 15 Bond breakages and failure modes obtained of all fissure blocks approximately at the residual strength level.
210 archives of civil and mechanical engineering 18 (2018) 198214

propagate with the increases of axial stress. When the axial
stress reaches 81.5% of peak strength, the tensile cracks have
got a great develop but not form coalescence between the pre-
existing joints. Apart from the tensile cracks, there are also
some mixed cracks occurs around the edge of opening
(Fig. 14b). When the axial stress reaches peak strength, a
macro shear failure planes occurs in the specimen (Fig. 14c)
and the tensile cracks do not exhibit an obvious difference
from those in point b (Fig. 14b).
After point peak stress, the specimen enters the macro
failure stage and a pair of shear band is located in the
specimen, where both of them are approximately parallel to
the diagonal line (Fig. 14d). For the left part of specimen, the
macro cracks around the opening evolved into a very narrow
zone at point d. The mixed crack connects the shear band and
the opening. With further loading, mixed cracks on the right
side of the opening keeps propagating and then connects the
opening with another shear band (Fig. 14e).
Fig. 15 shows the bond breakages obtained for all the
specimens at the residual stage. When a = 08, all the specimens
(S-20-0, S-30-0, S-40-0, and S-50-0) belong to the Mode-I.
Although in the same failure mode, because of the joint
distance of the pre-existing joints, the joints in specimen S-40-
0 and S-50-0 do not penetration easily. And then the UCSJ in
this specimen are higher than in specimens S-20-0 and S-30-0
(Fig. 6).
For the specimens with a = 308 and a = 458, specimens S-20-
30, S-20-45, S-30-30 and S-30-45 belong to the Mode-II. And
specimens S-40-30 shows failure modes between II and I.
Moreover, with the increases of joint distance d, the specimens
with a = 308 and a = 458 gradually turns toward the Mode-I and
Mode-III, respectively. It should be note that, there are only
two specimens (S-40-45 and S-50-45) belong to Mode-III. It can
be seen from the gure that there is a macro shear failure plane
in these specimens, and the failure plane approximately
parallel to diagonal line of the specimen.
The Mode-IV appears mainly in the specimens with a = 608
and 758. Among them, specimen S-20-75 shows failure modes
between Mode-II and Mode-IV. Clearly, there is a set of parallel
shear failure planes in these specimens. Apart from the shear
failure planes, there is a pair of mixed cracks connects the
shear bands and the opening.

Fig. 16 Mode-I comparisons between experimental and numerical results: (a) S-20-0; (b) S-30-0.

Fig. 17 Mode-II comparisons between experimental and numerical results: (a) S-20-30; (b) S-20-45.
 archives of civil and mechanical engineering 18 (2018) 198214 211

Fig. 18 Mode-III comparisons between experimental and numerical results: (a) S-40-45; (b) S-50-45.

Fig. 19 Mode-IV comparisons between experimental and numerical results: (a) S-40-60; (b) S-50-60.

Mode-IV (Fig. 19), the failure characteristics in the numerical

5. Comparison between the experimental and
numerical results
Figs. 1619 show the four typical failure modes obtained by
numerical simulation and the corresponding failure modes
obtained for the same specimens through the laboratory tests.
The numerical results show a great agreement with the
experimental results. In the comparison of the Mode-I (S-20-0
and S-30-0) (Fig. 16), the location of the splitting failure planes
in the numerical specimen agree well with the experimental
results. The macro failure planes mainly concentrated in the
upper part of the specimen and horizontally symmetrical.
Comparison of typical Mode-II examples (Fig. 17) clearly show
the stepped failure planes at left and right side of the opening.
It can also be seen from the numerical results, these speci-
mens mainly failure from the propagation of tensile cracks.
Fig. 18 shows the comparison of the numerical and experi-
mental results for the Mode-III. Both of the numerical and
experimental results show that the specimen failure from
shear crack coalescence. And the shear failure plane approxi-
mately parallel to the diagonal line of the specimen. For the
results almost keep the same with the experimental results,
and the location of shear and mixed failure planes show a
great agreement with each other.
The axial stressaxial strain curves of different failure
modes of the numerical and experimental results are shown in
Fig. 20. The comparison of these curves shows good agreement
and validates the ability of the PFC2D mode to reproduce the
behavior observed in the laboratory tests. Overall, the
experimental curves are very close the numerical curves.
Before peak stress, both of the numerical and experimental
curves show non-linearity as the axial strain increases. But for
the experimental curves, the uctuation of the curves before
and after peak stress are more obvious than those in numerical
simulation. Actually, numerical simulation is an ideal solution
(ideal boundary conditions and materials) for crack initiation
and propagation. So, there are inevitable differences when
compare with the results obtained from laboratory tests. At the
same time, the numerical simulation (PFC2D) in this research
is a two-dimensional analyses, which cannot exactly modeling
the 3D physical phenomena of the specimens such as surface
spalling around the opening.
212 archives of civil and mechanical engineering 18 (2018) 198214

Fig. 20 Stressstrain curve and tensile and shear bond failures for (a,b) Mode-I, (c,d) Mode-II, (e,f) Mode-III and (g,h) Mode-IV.
 archives of civil and mechanical engineering 18 (2018) 198214
6. Conclusions
Compression tests on rock-like specimens containing joints
and an opening were carried out on a servo-controlled uniaxial
loading tester. At the same time, numerical simulations using
PFC2D were carried out to simulate the failure process of
jointed specimens under uniaxial loading. Based on experi-
mental and numerical results, the conclusions are drawn as
follows:
(1) The strength parameters are mainly effected by the joint
geometry. The UCSJ and EJ of the specimens exhibit similar
tendencies with the change of inclination angle and joint
distance. Specically, the UCSJ and EJ of specimens
decrease for 08 a 458 and increase for a > 458. UCSJ
and EJ increases with increasing d for all a values, with the
highest and lowest strengths obtained for d = 50 mm and
d = 20 mm, respectively.
(2) Under compression, tensile or shear cracks will develop
from the tips of the pre-existing joints. With loading
continues, new cracks will propagate and link with other
cracks to form coalescence. The opening and the difference
in joint geometry (a and d) lead to the different failure
modes in the specimens. Based on experimental observa-
tions, the failure modes of jointed specimens can be
generally classied into four categories: (1) Mode-I (hori-
zontally symmetrical splitting failure); (2) Mode-II (stepped
failure at opening sides); (3) Mode-III (failure through a
plane); (4) Mode-IV (mixed failure).
(3) When a = 08, all the specimens belong to the Mode-I. For the
specimens with a = 308 and a = 458, specimens S-20-30, S-
20-45 and S-20-45 belong to the Mode-II. And with the
increases of joint distance d, the specimens with a = 308
and a = 458 gradually turns toward the Mode-I and Mode-
III, respectively. The Mode-IV appears mainly in the
specimens with a = 608 and 758.
The numerical results for the jointed specimen under
compression loading show good agreement with the labora-
tory tests. Moreover, the crack evolution mechanism in a
jointed specimen under compression loading is revealed.
These results are expected to improve the understanding of
the stability and failure trend of a rock mass with joints and
opening in rock engineering.
Acknowledgements
This paper gets its funding from Project (51174228, 51304240,
51474249, 51404179) supported by National Natural Science
Foundation of China; Project supported by the Program for
New Century Excellent Talents in University. Project sup-
ported by the Graduate student innovation project of Central
south university (2015zzts074). The authors wish to ac-
knowledge these supports. At the same time, authors are
very grateful for the anonymous reviewers' valuable com-
ments.
213
references
[1] L.E. Vallejo, The brittle and ductile behavior of clay samples
containing a crack under mixed mode loading, Theoretical
and Applied Fracture Mechanics 10 (1988) 7378.
[2] A.V. Dyskin, E. Sahouryeh, R.J. Jewell, Inuence of shape and
locations of initial 3-D cracks on their growth in uniaxial
compression, Engineering Fracture Mechanics 70 (15) (2003)
21152136.
[3] E. Hoek, Z.T. Bieniawski, Brittle fracture propagation in rock
under compression, International Journal of Fracture 1 (3)
(1965) 137155.
[4] Y.P. Li, L.Z. Chen, Y.H. Wang, Experimental research on pre-
cracked marble under compression, International Journal of
Solids and Structures 42 (9/10) (2005) 25052516.
[5] S.Q. Yang, H.W. Jing, Strength failure and crack coalescence
behavior of brittle sandstone samples containing a single
ssure under uniaxial compression, International Journal of
Fracture 168 (2011) 227250.
[6] N.Y. Wong, Crack Coalescence in Molded Gypsum and
Carrara Marble, (Ph.D. thesis), Massachusetts Institute of
Technology, 2008.
[7] H.Q. Li, L.N.Y. Wong, Inuence of aw inclination angle and
loading condition on crack initiation and propagation,
International Journal of Solids and Structures 49 (2012)
24822499.
[8] B. Shen, The mechanism of fracture coalescence in
compression experimental study and numerical simulation,
Engineering Fracture Mechanics 51 (1) (1995) 7385.
[9] A. Bobet, H.H. Einstein, Fracture coalescence in rock-type
materials under uniaxial and biaxial compression,
International Journal of Rock Mechanics and Mining
Sciences 35 (7) (1998) 863888.
[10] M. Sagong, A. Bobet, Coalescence of multiple aws in a rock-
model material in uniaxial compression, International
Journal of Rock Mechanics and Mining Sciences 39 (2) (2002)
229241.
[11] H. Lee, S. Jeon, An experimental and numerical study of
fracture coalescence in pre-cracked specimens under
uniaxial compression, International Journal of Solids and
Structures 48 (6) (2011) 979999.
[12] X.P. Zhang, Q.S. Liu, S.C. Wu, X.H. Tang, Crack coalescence
between two non-parallel aws in rock-like material under
uniaxial compression, Engineering Geology 199 (2015) 74
90.
[13] C.A. Tang, Numerical simulation of progressive rock failure
and associated seismicity, International Journal of Rock
Mechanics and Mining Sciences 34 (2) (1997) 249261.
[14] N. Moes, T. Belyschko, Extended nite element method for
cohesive crack growth, Engineering Fracture Mechanics 69
(2002) 813833.
[15] P. Areias, D. Dias-da-Costa, J. Alfaiate, E. Jlio, Arbitrary bi-
dimensional nite strain cohesive crack propagation,
Computational Mechanics 45 (1) (2009) 6175.
[16] R. Pramanik, D. Deb, SPH procedures for modeling multiple
intersecting discontinuities in geomaterial, International
Journal for Numerical and Analytical Methods 39 (4) (2015)
343367.
[17] S.R. Beiseel, C.A. Gerlach, G.R. Johnson, Hypervelocity impact
computations with nite elements and meshfree particles,
International Journal of Impact Engineering 33 (2006) 8090.
[18] Y.J. Ning, J. Yang, X.M. An, G.W. Ma, Modelling rock fracturing
and blast induced rock mass failure via advanced
discretization within the discontinuous deformation
analysis framework, Computers and Geotechnics 38 (1)
(2010) 4049.
214 archives of civil and mechanical engineering 18 (2018) 198214
[19] Y.J. Ning, X.M. An, G.W. Ma, Footwall slope stability analysis
with the numerical manifold method, International Journal
of Rock Mechanics and Mining Sciences 48 (2011) 964975.
[20] R.H. Cao, P. Cao, X. Fan, X.G. Xiong, H. Lin, An experimental
and numerical study on mechanical behavior of ubiquitous-
joint brittle rock-like specimens under uniaxial compression,
Rock Mechanics and Rock Engineering 49 (2016) 43194338.
[21] R.H. Cao, P. Cao, H. Lin, C.Z. Pu, K. Ou, Mechanical behavior of
brittle rock-like specimens with pre-existing ssures under
uniaxial loading: experimental studies and particle
mechanics approach, Rock Mechanics and Rock Engineering
49 (3) (2016) 763783.
[22] S.Q. Yang, Y.H. Huang, H.W. Jing, X.R. Liu, Discrete element
modeling on fracture coalescence behavior of red sandstone
containing two un-parallel ssures under uniaxial
compression, Engineering Geology 178 (2014) 2848.
[23] C. Cheng, X. Chen, S.F. Zhang, Multi-peak deformation
behavior of jointed rock mass under uniaxial compression:
insight from particle ow modeling, Engineering Geology 213
(2016) 2545.
[24] X. Fan, P.H.S.W. Kulatilake, X. Chen, Mechanical behavior of
rock-like jointed blocks with multi-non-persistent joints
under uniaxial loading: a particle mechanics approach,
Engineering Geology 190 (14) (2015) 1732.
[25] R.H.C. Wong, P. Lin, C.A. Tang, K.T. Chau, Creeping damage
around an opening in rock-like material containing non-
persistent joints, Engineering Fracture Mechanics 69 (2002)
20152027.
[26] M. Souley, F. Homand, S. Pepa, D. Hoxha, Damage-induced
permeability changes in granite: a case example at the URL in
Canada, International Journal of Rock Mechanics and Mining
Sciences 38 (2001) 297310.
[27] A. Bobet, H.H. Einstein, Numerical modeling of fracture
coalescence in a model rock material, International Journal
of Fracture 92 (3) (1998) 221252.
[28] L.N.Y. Wong, H.H. Einstein, Crack coalescence in molded
gypsum and Carrara marble. Part 1. Macroscopic
observations and interpretation, Rock Mechanics and Rock
Engineering 42 (3) (2009) 475511.
[29] M. Prudencio, M. Van Sint Jan, Strength and failure modes of
rock mass models with non-persistent joints, International
Journal of Rock Mechanics and Mining Sciences 44 (2007)
890902.
[30] P. Cao, T.Y. Liu, C.Z. Pu, H. Lin, Crack propagation and
coalescence of brittle rock-like specimens with pre-existing
cracks in compression, Engineering Geology 187 (17) (2015)
113121.
[31] Y.L. Zhao, L.Y. Zhang, W.J. Wang, C.Z. Pu, W. Wan, J.Z. Tang,
Cracking and stressstrain behavior of rock-like material
containing two aws under uniaxial compression, Rock
Mechanics and Rock Engineering 49 (7) (2016) 26652687.
[32] E.G. Bombolakis, Photoelastic study of initial stages of brittle
fracture in compression, Tectonophysics 6 (6) (1968) 461473.
[33] X.X. Yang, H.W. Jing, C.A. Tang, S.Q. Yang, Effect of parallel
joint interaction on mechanical behavior of jointed rock
mass models, International Journal of Rock Mechanics and
Mining Sciences 92 (2017) 4053.
[34] N. Cho, C. Martin, D. Sego, A clumped particle model for rock,
International Journal of Rock Mechanics and Mining Sciences
44 (7) (2007) 9971010.
[35] D.O. Potyondy, Simulating stress corrosion with a bonded-
particle model for rock, International Journal of Rock
Mechanics and Mining Sciences 44 (5) (2007) 677691.