Pakzad2018 PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Numerical Study of the Failure Response and Fracture

Propagation for Rock Specimens with Preexisting Flaws


under Compression
Ramin Pakzad1; Shanyong Wang, M.ASCE2; and Scott Sloan, M.ASCE3
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

Abstract: This study incorporated an elastic brittle damage constitutive model into a program for modeling and analysis through a user sub-
routine interface to define a material’s mechanical behavior and considered nonlinear geometric effects. The Weibull distribution function was
adopted to consider the heterogeneity-related uncertainty of the strength and stiffness. The 2D models used were discretized using plain strain
reduced integration elements with the option of element removal after being fully damaged. We verified the accuracy of the user subroutine
code by reproducing the observed failure behavior of an intact sandstone specimen with using previously suggested material parameters. The
reliability of the numerical model was supported by an agreement between pre-existing experimental results and the numerical simulation
results for specimens containing a single fissure with different inclination angles. Two parametric studies were conducted on the failure behav-
ior of a specimen with a single fissure to investigate the effect of (1) the heterogeneity level and (2) the confining pressure. Stiffness and
strength were shown to decrease with increases in the level of heterogeneity. Tensile cracks that appeared in the more homogeneous models
were replaced by scattered damaged elements with increasing heterogeneity. Increasing the confining pressure increased the load capacity of
the specimen, regardless of the inclination of the fissure angle. This trend was related to the amount of damage occurring before reaching the
peak load. The maximum load was typically lower when the route length of the cracks formed before the peak load was larger. To investigate
fracture coalescence behavior, specimens with two parallel fissures were modelled for four different ligament angles. The numerical simula-
tions agree with the available experimental results, indicating that coalescence changed from shear to a tensile mode as the ligament angle
increased. DOI: 10.1061/(ASCE)GM.1943-5622.0001172. © 2018 American Society of Civil Engineers.
Author keywords: Elastic damage constitutive model; Weibull distribution; Fracture propagation and coalescence; Rock failure; Abaqus;
UMAT.

Introduction existing joints or fissures can fail much sooner than intact speci-
mens. This behavior has been experimentally studied by many
Rock materials are well known to be susceptible to quasi-brittle researchers, at various sizes and orientations of flaws (Yang et al.
failure under compressive loading conditions. This type of load- 2008, 2009, 2013; Yang and Jing 2011; Yang 2011; Lu et al.
ing is common to many rock engineering problems, such as rock 2015). Different fracture propagation patterns have also been
slope stability, nuclear waste disposal projects, and dam construc- observed physically (Shen et al. 1995; Bobet and Einstein 1998;
tion. Rock materials under compressive loading conditions have Wong et al. 2001; Sagong and Bobet 2002; Li et al. 2005; Wong
been the subject of numerous studies over many years (Handid and Einstein 2009a, b; Lee and Jeon 2011; Yang and Jing 2011;
and Hager 1957; Mogi 1971; Peng and Johnson 1972; Haimson Zhao et al. 2016).
and Chang 2000; Basu et al. 2013; Cvitanovic et al. 2015; Nevertheless, because of the geologically distinct conditions
Turichshev and Hadjigeorgiou 2016; Ghazvinian et al. 2012, experienced by rock masses, there is no common consensus on their
2013; Bahaaddini et al. 2015). The failure behavior of rocks has failure modes. Ongoing studies in this field have developed with the
been found to be influenced by a variety of factors, such as their advancement of sophisticated computational techniques. Numerical
level of heterogeneity and the presence of pre-existing macro- simulation is an informative way of analyzing fracture and damage,
cracks. Numerous studies have shown that rock masses with pre- and it has been employed frequently in recent decades (Aliha et al.
2010, 2013; Manouchehrian and Marji 2012; Zhang and Wong
1
ARC Centre of Excellence for Geotechnical Science and Engineering,
2012; Haeri et al. 2014; Manouchehrian et al. 2014; Haeri 2015;
the Univ. of Newcastle, Callaghan, NSW 2308, Australia. E-mail: ramin Pakzad and Ayatollahi 2015; Pakzad et al. 2016; Cao et al. 2016).
.pakzad@uon.edu.au The finite element method is a widely used numerical technique
2
ARC Centre of Excellence for Geotechnical Science and Engineering, through which crack propagation phenomena can be studied in three
the Univ. of Newcastle, Callaghan, NSW 2308, Australia (corresponding distinct ways. The first method uses conventional fracture mechan-
author). E-mail: Shanyong.Wang@newcastle.edu.au ics. Despite its strong theoretical foundation, singularities at crack
3
ARC Centre of Excellence for Geotechnical Science and Engineering, tips and the need for rediscretization to capture the moving bounda-
the Univ. of Newcastle, Callaghan, NSW 2308, Australia. E-mail: Scott ries of fractures and its shortcomings under mixed mode loading
.Sloan@newcastle.edu.au
conditions pose difficulties for this approach (Aliha et al. 2010,
Note. This manuscript was submitted on June 2, 2017; approved on
December 27, 2017; published online on April 26, 2018. Discussion pe- 2013). Although research has sought to overcome these deficien-
riod open until September 26, 2018; separate discussions must be submit- cies through cohesive elements, this has been limited to problems
ted for individual papers. This paper is part of the International Journal with predefined crack-propagation routes (Zhang et al. 2010;
of Geomechanics, © ASCE, ISSN 1532-3641. Su et al. 2010). Another technique for crack propagation analyses

© ASCE 04018070-1 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


is damage-based fracture modeling, which does not have the Methodology
abovementioned weakness and is consistent with many useful
material features, such as plasticity and heterogeneity (Wang In the framework of the finite element method, the equilibrium
et al. 2012a, b; Wu and Wong 2013). Furthermore, this method equation shown in Eq. (1) was implicitly solved by the Abaqus
can consider the concurrent influence of adjacent cracks in a net- standard solver to obtain the displacement degrees of freedom for
work of fractures. every node in the model under static conditions.
Conceptually, the overall failure of rock masses is rooted in the
behavior of their microstructure, where microcracks nucleate and ∂s ij
þ Fj ¼ 0 (1)
lead to partial loss of material. The local accumulation of micro- ∂xj
cracks results in the local failure of material and the appearance of
meso-scale cracks. The propagation of such cracks produces macro- The components of strain [Eq. (2)] are also calculated from the
scale fractures whose coalescence can lead to overall failure. This displacement field and then sent to the UMAT subroutine, accom-
transitional process, from the micro- to the macrolevel, cannot be panied by other state variables, at the beginning of every increment.
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

accurately reproduced through macroscopic phenomenological  


constitutive models unless many complexities are incorporated. 1 ∂ui ∂uj
ɛij ¼ þ (2)
Thus, rock-type materials can be assumed to comprise a network of 2 ∂xj ∂xi
tiny cells, each of which has its own properties. This nonuniform
distribution of material properties in rock elements is attributed to The user must use a constitutive model relating stresses to strains
the random distribution of pre-existing microlevel cracks, voids, at the integration point(s) of each element.
and minerals. This is the concept of a representative volume ele-
ment (RVE), which links the micro- and macroscales. Exceeding a
stress limit, the mesoscopic elements begin to degrade irreversibly Constitutive Model
until they can no longer sustain any stress. This notion has been
shown to be a powerful method for reproducing the nonlinear In this study, the elastic damage constitutive relationship of the
behavior of heterogeneous geomaterials and has formed the basis of meso-scale elements was incorporated in Abaqus through its
a large number of numerical investigations on the failure behavior UMAT subroutine interface. The material properties were distrib-
of intact and flawed quasi-brittle specimens in recent years (Tang uted among elements using a statistical distribution function. This
et al. 2000; Wong et al. 2002; Zhu and Tang 2004; Feng et al. 2006; method, known as the statistical meso-damage mechanical method
Li and Tang 2015). (SMDMM), was applied to finite elements in the model to simulate
In this study, we examined the mechanics of damage for meso- the trans-scale progress of failure using the RVE concept (Li and
scale elements with random material properties. To capture the real Tang 2015). In contrast to phenomenological models with complex
behavior of material, different damage-based constitutive models formulations to account for nonlinearity, the SMDMM was based
have been proposed. For example, defining damage as a tensor on linear elasticity. Nevertheless, it was capable of reproducing
rather than a scaler includes the anisotropic effect in the simulation nonlinearity by considering heterogeneity and by connecting the
(Lemaitre and Desmorat 2005), or one can take into account the elastic modulus to the amount of damage.
effect of damage on the dilatancy behavior of a geo-material, such
as sandstone, by considering a two-scale damage model (Lemaitre Elastic Regime
and Desmorat 2005). In this study a one-scale damage model was Following Hooke’s law, the principal stresses and strains were cor-
used for simplicity. The decrease of the damaged element’s stiff- related through Eq. (3), in which λ ¼ E=ð1  2 Þð1 þ  Þ,
ness under the effect of the damage parameter was explicitly G ¼ E=ð1 þ  Þ,  is Poisson’s ratio and E is the current Young’s
included, while the strength degradation of the element due to dam- modulus.
age was implicitly considered in the damaged evolution formula-
tion. After calibrating the material model with the test results of an s ij ¼ 2Gɛij þ λɛjj ði; j ¼ 1; 2; 3Þ (3)
intact sandstone specimen under uniaxial loading (Yang and Jing
2011), we assess the reliability of our numerical simulation by com- The stiffness began to degrade gradually as soon as the stress
puting the stress–strain response for specimens containing a single state meets the damage surface. This occurs incrementally by the
fissure with various inclinations. The effects of the confining stress evolution of the damage parameter (D) in Eq. (4), where E and E0
and heterogeneity level on the strength failure of the specimens and represent the updated and initial stiffnesses of the mesoscopic ele-
fracture pattern were investigated by conducting two series of para- ment, respectively.
metric analyses. We also investigated the fracture coalescence of
specimens containing two parallel fissures for combinations of fis- E ¼ ð1  DÞE0 (4)
sures and ligament angles.
A numerical simulation was performed using Abaqus, which The damage initiation criteria and damage evolution formulation
allows users to extend the software capability through its subrou- will be discussed in the next subsection.
tine interfaces. Because Abaqus does not have built-in constitu-
tive models that cover all the facets of the abovementioned
Damage Initiation and Evolution
behavior, we wrote a FORTRAN code compatible with the
UMAT (user subroutine interface to define a material’s mechani- In contrast to metals, in which crack initiation is the result of tensile
cal behavior) of Abaqus to perform the analyses. In the following loading, cracks in rock-type materials have been shown to originate
sections of this paper, the governing and constitutive equations due to compressive loading, as well. Therefore, the damage surface
are presented, followed by a description of the numerical setup. is defined by two individual criteria: the tensile failure criterion and
Then, the results obtained from the numerical simulations will the shear (Mohr Coulomb) failure criterion, with priority given to
be discussed and the highlights of the study will be summarized. the former. First, the maximum principal stress (s 1 ) was compared

© ASCE 04018070-2 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


with the uniaxial tensile strength of element (ft0 ) according to Eq. until the end of the analysis. Fully damaged elements are removed
(5), under the condition that the shear criterion is less critical than virtually by assigning to their stiff ness the small value of 1:05 .
the tensile criterion. Then, the damaged elements were deleted by Abaqus in the next in-
crement. The line-search technique incorporated in Abaqus was
s 1  ft0 (5) used to optimize the solution procedure.
The effects of microcracking are considered in the material
If the tensile failure criterion was met, the damage parameter behavior of elements by reducing the stiffness and strength of the
was calculated using Eq. (6), depending on the value of equivalent damaged element before reaching the fully–damaged condition.
strain (~ɛ ) at the end of the current increment (Wang et al. 2012a, b). Before this stage, a damaged element can transfer the induced stress
8 to its neighboring elements. However, the effect of crack closure
>
> 0 ~ɛ < ɛt0 under compressive circumstances is not included in this study.
>
>
< ftr When the number of fully-damaged elements increases, particularly
D¼ 1 0 ɛt0  ~ɛ < ɛtu (6) after peak load, large deformation occurs (resulting in the rigid body
>
> s
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

>
>
1
rotation of some elements). Therefore, the stress distribution
: ~ɛ  ɛtu
1 throughout the model will not be correct unless the global stiffness
matrix and the stress tensor are incrementally updated in the frame-
In this equation, the equivalent strain at which the tensile dam- work of the local coordinate system of elements, resulting in a geo-
age surface was met for the first time was symbolized by ɛt0 , and s 01 metric nonlinearity in the finite element system of equations. On the
denoted the maximum principal stress calculated by the initial elas- other hand, with increase in the local density of the fully damaged
tic modulus. The residual tensile strength and ultimate tensile strain elements, the nodes attached only to the fully-damaged elements, ex-
were defined as ftr ¼ g ft0 and ɛtu ¼ h ɛt0 , respectively. The equiva- hibit no resistance and are free to discontinuously displace every-
lent strain corresponding to the tensile damage evolution was where in the 2D/3D space. The discontinuous free movement of
assumed to be a combination of the principal strains, as follows nodes distorted the relevant elements, making discontinuous rotation
(Wang et al. 2012a, b): of the local coordinate system of the elements. As a result, the finite
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi element system of equations became so nonlinear that the solution
~ɛ ¼ hɛ1 i2 þ hɛ2 i2 þ hɛ3 i2 (7) could not converge. To overcome this numerical difficulty, the fully
damaged elements were removed from the model. In this case, a
concern remained about inaccurate stress distribution when the free
The Macaulay brackets hi are defined by Eq. (8): surfaces generated by the removal of the fully-damaged elements
( come together under compression. However, in our simulation,
x x0
hxi ¼ (8) compared to the size of removed elements, the displacements were
0 x<0 so small that almost no overlap occurred between the elements of the
free surfaces of fractured area, except for some cases that had spal-
The shear damage criterion is defined by Eq. (9): ling at the outside boundary of the models. In reality, the separated
parts fall due to the gravitational acceleration, which did not affect
1 þ sin f the stress distribution in the remained specimen under compressive
c s 1  s 3  fc0 c ¼ (9)
1  sin f loading. In general, the model would be more accurate if the contact
could be included in the numerical model. Unfortunately, in Abaqus,
where w , fc0 , s 1 , and s 3 are the internal frictional angle, uniaxial the standard (v6.14) defining contact for the free surfaces generated
compressive strength, and maximum and minimum principal during the solution process is not possible.
stresses, respectively. A simplified version of the constitutive law elaborated above is
When the shear criterion was satisfied, the damage parameter illustrated in Fig. 1 for the uniaxial case. The strength of the element
was calculated via Eq. (10): decreases to a fraction of the initial value after the initial damage
8 occurs, and the element is eliminated in tension when its tensile
>
<0 ~ɛ > ɛc0 equivalent strain exceeds the ultimate tensile strain. According to
D¼ fcr (10) previous literature, both the stiffness and strength of rock decrease
>
:1 ~ɛ  ɛc0 during the damage process, whereas Poisson’s ratio increases as a
c s 01  s 03

where fcr ¼ g fc0 , ~ɛ ¼ ɛ3 , ɛc0 is the equivalent strain at the moment


that the shear failure criterion meets the current strength of the
element for the first time, and s 01 and s 03 are the maximum and
minimum principal stresses of the intact element calculated by
the initial Young’s modulus. The convention of Abaqus is fol-
lowed in writing the abovementioned relationships such that the
stress and strain components are negative in compression and
positive in tension.
Once the damage parameter was obtained, the stress tensor is
updated in UMAT according to the modified elastic modulus in Eq.
(4). Iterating to find the solution for the equilibrium equation,
Abaqus sent new strain increments to UMAT after the stress redis-
tribution. The iterations were repeated until equilibrium is reached Fig. 1. Constitutive law of elastic damage for a mesoscopic element
and no further elements undergo damage (i.e., the solution con- under uniaxial tension and compression
verges for the increment). This process progressed incrementally

© ASCE 04018070-3 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


result of cracking (Heap et al. 2009). The alteration of material of the viscous regularization scheme of Duvant and Lions (1976)
properties has been shown to be irreversible, as shown in the Kaiser was used to overcome these difficulties. This regularization scheme
effect (Lavrov 2003, 2005). In this study, the damage-induced dila- resulted in a positive definite tangent stiffness matrix, provided the
tion effect (increase in Poisson’s ratio with the damage parameter) time increments are sufficiently small. It was assumed that the dam-
has been ignored for simplicity. age parameter in Eq. (4) could be substituted with a viscous damage
variable, defined as follows:
1
Viscous Regularization D_v ¼ ðD  Dv Þ (11)
j
The implicit integration approach used in UMAT encounters con-
vergence problems for material models with abrupt softening where j is a viscosity coefficient representing the relaxation
behavior, such as what is presented in this paper. A generalization time of the viscous system and Dv is the viscous-damage
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Typical random distribution of Young’s modulus: (a) spatial assignment of Young’s modulus to elements; (b) histogram of Young’s modulus
for different homogeneity indexes

© ASCE 04018070-4 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


variable. The viscosity coefficient should be set sufficiently be modelled statistically via the Weibull distribution function,
small relative to the characteristic time increment. The results using the following formula (Tang et al. 2000; Wang et al.
obtained using this method were acceptable, because the solu- 2012a, b):
tion of the viscous system relaxes to that of the inviscid system
as the ratio of the time increment to the viscosity coefficient
reaches infinity. Table 1. Meso-Scale Material Properties Calibrated for the Sandstone
from Linyi City

Random Distribution of Stiffness and Strength Initial Yang et al.


Parameter calibration (2015)
In the application of the SMDMM, each element in the finite
element model was assumed to be an RVE with randomly Homogeneity index (m) 10 10
assigned material properties. The heterogeneity of the rock can Mean elastic modulus (E0 ) 34.5 (GPa) 35.0 (GPa)
Internal friction angle ( w ) 30 (degrees) 40 (degrees)
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

Mean compressive strength (fc0 ) 380 (MPa) 370 (MPa)


Ratio of compressive to tensile 10 8
strength (fc0 =ft0 )
Poisson’s ratio () 0.28 0.25
Ultimate tensile strain coefficient ( h ) 2 1.5
Confining pressure ( g ) 0.2 0.1

Fig. 3. Schematic representation of a rectangular block: (a) a single


fissure; (b) two coplanar fissures

Fig. 5. Ultimate failure mode of the intact specimen under uniaxial


Fig. 4. Schematic representation of the applied load and boundary compressive loading: (a) experimental results [reprinted from Yang and
conditions: (a) first step; (b) second step Jing (2011), with permission]; (b) numerical results (current study)

© ASCE 04018070-5 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


  "   #
m u m1 u m
f ðuÞ ¼ exp  (12)
u0 u0 u0

where u corresponds to the random material property of the ele-


ment and m is the shape factor of the Weibull distribution function
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Effect of the homogeneity index value on the stress-strain


curves of the specimen with a ¼ 45
Fig. 6. Stress–strain curves for sandstone under uniaxial compressive
loading: (a) experimental results [reprinted from Yang and Jing (2011),
with permission]; (b) numerical results (current study)

Fig. 7. Effect of the initial seed on the failure behavior of the specimen Fig. 9. Effect of the homogeneity index on damage propagation in the
with (a ¼ 15 ): (a) stress–strain curve; (b) fracture pattern specimen containing a single fissure with a ¼ 45

© ASCE 04018070-6 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


determining the dispersal of u around u0 . u approaches the value sequence. We also considered the geometric nonlinearity effect,
of u0 as the value of m increases. In this study, the elastic modulus because elements may rotate with respect to their original position.
and strength were considered to vary among elements with the Except for the intact specimen, one or two parallel fissures with a
same homogeneity indexes (m) but different initial seeds. Fig. 2 predefined inclination angle a were cut from the 2D models. The fis-
illustrates the random distribution of Young’s modulus among sure lengths, 2a, ligament lengths, 2b, and fissure thicknesses were
meso-scale elements alongside curves representing the influence constants equal to 15 mm, 18 mm, and 2:5 mm, respectively. For
of the homogeneity index on the randomization. The inverse of specimens containing two parallel fissures, the ligament angle b
the Weibull cumulative probability function was used to produce determines the relative location of fissures.
the random variable from the random number generated by the In this study, two types of loading conditions were applied: uni-
Monte Carlo method. axial and biaxial. Fig. 4 schematically illustrates the load and
boundary conditions applied in two successive steps. The uniaxial
Model Setup loading condition is the special case of this general loading in which
the confining pressure P is equal to zero.
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

Both a rectangular block containing a single fissure (Yang and Jing We calibrated our numerical model by minimizing the deviation
2011) and a rectangular block containing two parallel fissures used of its results from those obtained from a uniaxial test on sandstone
by (Yang 2011) were modelled (Fig. 3). The number of elements in from Linyi City, Shandong Province, China (Yang and Jing 2011).
the former with dimensions of 60 mm  120 mm was approxi- Readers interested in the calibration procedure are referred to
mately 135  270  36;000, whereas the latter, which was Pakzad et al. (2016). We subsequently discovered that such a cali-
80 mm in width and 160 mm in height, was discretized by approxi- bration had already been performed for the same sandstone (Yang
mately 160  320  51;000 elements. Two-dimensional plain et al. 2015). The results presented in this paper are based on the sec-
strain analyses were conducted using first-order, reduced integration ond group of calibrated parameters represented in Table 1.
elements with hourglass control. The element deletion option was After validating the numerical simulation procedure, two series
employed to delete fully damaged elements from the computation of parametric studies were conducted to investigate the effect of the

Fig. 10. Stress–strain curves of fissure-included specimens under biaxial loading conditions: (a) (m ¼ 10:0Þ (a ¼ 0 ); (b) (m ¼ 10:0Þ (a ¼ 30 );
(c) (m ¼ 10:0Þ (a ¼ 60 ); (d) (m ¼ 10:0Þ (a ¼ 90 )

© ASCE 04018070-7 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


homogeneity index m and confining pressure P on the failure behav- a ¼ 45 under uniaxial loading conditions for several homogeneity
ior of a specimen with a single fissure. A homogeneity index of indexes. The numerical axial stress–strain curves for different ho-
{1.5, 2, 3.5, 6.5, 10} and zero confining pressure were used for the mogeneity indexes are shown in Fig. 8, which indicates that the
specimen containing a fissure with an inclination angle of a ¼ 45 . strength and stiffness both increase with increasing values of the ho-
While maintaining a homogeneity index equal to 10, we performed mogeneity index. For example, analyses with homogeneity indexes
additional analyses for specimens with inclination angles of of 1.1 and 10 resulted in peak loads of approximately 24 and 120,
{0 ; 15 ; 30 ; 45 ; 60 ; 75 ; 90 } and under lateral pressures of respectively. Ductile failure is more noticeable for specimens with
{0, 10, 20, 30} MPa. lower homogeneity indexes, and the load capacity of the specimen
In the last part of our study, we investigated the propagation of gradually decreases for cases with homogeneity indexes less than 2.
cracks and their coalescence in specimens with two parallel fissures. By contrast, samples with homogeneity indexes greater than 3.5
To this end, we performed simulations of the behavior of specimens
under uniaxial loading conditions for a fissure inclination angle of
a ¼ 45 and four different ligament angles, b ¼ 45 . Then, the nu-
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

merical results and available experimental results were compared.


The numerical results will be presented and discussed in the follow-
ing sections.

Validation of the Numerical Results

First, we numerically reproduced the failure behavior of an intact


sandstone specimen under compressive loading conditions using
the calibrated parameters suggested by Yang et al. (2015). The com-
puted ultimate failure mode and stress-strain curve were generally
comparable to those obtained experimentally (Figs. 5 and 6), thus
demonstrating the reliability of the code used.
To further check the numerical simulation, we performed analy-
ses of models containing fissures 15 mm long at different inclina-
tion angles. Fig. 6 compares the experimentally and numerically
obtained stress-strain curves for the intact and fissure-included Fig. 11. Influence of the inclination angle on the ultimate strength of
specimens. the specimens with a single fissure under different confining pressures
The experimental results [Fig. 6(a)] indicate that nonlinear
behavior occurs at the initial stage of loading. This nonlinearity,
which is attributable to the closure of voids and microcracks,
could not be captured by the constitutive modewl in this study.
However, the elastic moduli computed numerically were
approximately the same as those obtained experimentally. With
regard to the type of failure, the sudden drop in the load capacity
(brittle failure) of the specimens was duplicated by numerical
analyses. The numerical results confirmed that the load capacity
of the intact specimen was greater than those of the fissure-
included specimens. The computed peak loads were very similar
to those obtained from the experiments, except for the case of a
fissure inclination angle of 15°. This discrepancy could be due to
the heterogeneity of the material. To investigate this, we con-
ducted four additional analyses for this case (a ¼ 15 ) with sim-
ilar inputs except for the initial seeds used in the random distri-
bution of the elastic modulus and strength parameters. The peak
load obtained from the second case (s2) was very similar to the
experimental result (see Fig. 7). Different distributions of the
material properties (different initial seeds) yielded different fail-
ure behavior in terms of the stress–strain response or fracture
pattern, which confirmed our hypothesis on the influence of the
distributions of the material properties on the failure behavior of
the specimens.

Failure Response of a Specimen Containing a


Single Fissure

Effect of the Heterogeneity Level


Fig. 12. Fracture pattern immediately after the peak load for different
To investigate the influence of the degree of heterogeneity, we confining pressures and inclination angles
examined the behavior of a specimen with an inclination angle of

© ASCE 04018070-8 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


exhibited a slightly brittle response. An intermediate response was exhibited an antitensile crack. For both cases, far field cracks
expected when the homogeneity index was between 2 and 3.5. The were also formed after peak load.
diversity in the stress-strain responses of the specimens could be The fracturing process for homogeneity indexes of 2 and 3.5 dif-
associated with the different type of cracks forming for different ho- fered from the process for m ¼ 6:5 in three main aspects. First,
mogeneity indexes. This was evaluated by the study of the damaged some damaged elements began to scatter in the model. Second, the
element patterns obtained in separate frames for different homoge- type of both nonsmooth tensile cracks was antitensile. Finally, the
neity indexes (Fig. 9). length of the smooth tensile cracks continually decreased. Smooth
Fig. 9 shows the nucleation and propagation of cracks in a wing cracks nearly disappeared for the lower homogeneity indexes
specimen with an inclination angle of 45 and different homoge- of 1.1 and 1.5. Therefore, the sudden decrease in the stress-strain
neity indexes. When the homogeneity index was 6.5 or 10, two curves for the specimens with higher homogeneity indexes may be
smooth tensile wing cracks nucleated simultaneously and propa- associated with the appearance of wing cracks, which are not
gated from the end corners of the two fissures. The length of the observed for the more heterogeneous cases. Moreover, the scatter-
wing cracks is shorter for a homogeneity index of 6.5. Then, two ing of damaged elements increased with increases in the heteroge-
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

unsmooth tensile cracks nucleated, one of which began to propa- neity level when the homogeneity index was 1.1 or 1.5. As a result,
gate vertically. Tensile cracks deviatde from their initial direc- the final pattern of damage had two antitensile cracks with damaged
tion, which was parallel to the loading direction, particularly after elements around them.
the specimen reached its maximum load capacity. This behavior
could be associated with the uneven distribution of material prop-
Effect of the Confining Stress
erties, as the stress state and strength properties together deter-
mined damage patterns at every stage of loading. The second To understand the effect of the confining pressure on the failure
crack was the same as the first for specimens with a homogeneity behavior, we performed numerical simulations for different incli-
index of 10, while specimens with a homogeneity index of 6.5 nation angles in which the confining pressure was increased to

Fig. 13. Experimental results for fracture propagation and coalescence in the sandstone specimen containing two coplanar fissures; (a) s 1 =
60.67 MPa « 1 = 2.377 10−3; (b) s 1 = 94.72 MPa « 1 = 3.329 10−3; (c) s 1 = 103.61 MPa « 1 = 3.581 10−3; (d) s 1 = 107.13 MPa = s c « 1 = 3.651 10−3;
(e) s 1 = 104.03 MPa « 1 = 3.767 10−3; (f) s 1 = 104.07 MPa « 1 = 3.979 10−3; (g) s 1 = 92.37 MPa « 1 = 4.008 10−3; (h) s 1 = 101.48 MPa « 1 =
4.60 10−3; (i) ultimate failure

© ASCE 04018070-9 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


30 MPa in increments of 10 MPa. The effect of the confining pres- inclination angles by Yang (2011). Fig. 13 shows several frames of
sure on the stress strain response of the specimen with a single fis- the failure of the specimen with the same fissures and ligament in-
sure for several inclination angles is illustrated in Fig. 10. The clination angles (a ¼ b ¼ 45 Þ: According to this figure, two wing
results show that the ultimate strength of the specimen increases cracks propagated from the upper and lower corners of the right-
with increasing confining pressure, regardless of the fissure incli- hand side fissure [Figs. 13(a–c)]. Before the peak load of approxi-
nation angle. The elastic stiffness appears to be largely insensitive mately 107 MPa was reached, a shear crack in the lower corner of
to both the confining pressure and the inclination angle. the same fissure appeared, followed by an antitensile crack forming
To investigate the influence of the inclination angle on the ulti- in the upper corner of the left side of the second fissure [Fig. 13(d)].
mate strength, the peak strength versus the inclination angle was Shear-type fracture coalescence occurred in the ligament region,
plotted for various confining pressures (Fig. 11). Fig. 11 shows followed by the appearance of new tensile fractures near and
that, despite some fluctuations, the ultimate strength generally away from the fissures until the ultimate failure of the specimen
increases with increases in the inclination angle from 0° to 90°. [Figs. 13(e–i)].
However, the variation in the ultimate strength with the inclina-
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

tion angle differed for varying confining pressures. For example,


when the inclination angle increased from 0° to 15°, the ultimate
strength of the specimen with a confining pressure of 10 MPa
increased from 120 MPa to approximately 127 MPa, which con-
trastd with the uniaxial case in which the ultimate strength
decreased from 110 MPa to approximately 97 MPa. Compared to
the specimens with confining pressures of 0 and 30 MPa, the ulti-
mate strength of the specimens with confining pressures of 10 and
20 MPa change in a different manner when the inclination angle
increased from 75° to 90°. This variation in behavior could be due
to the combined effect of the confining pressure and the specimen
inclination angle on the stress redistribution, which in turn influ-
enced the sequence of damaged elements before the peak load
was reached.
Fig. 12 illustrates the fracture pattern obtained immediately
after the peak load was reached for different combinations of con-
fining pressure and inclination angles. The cracks appearing
before the peak load are marked with red ovals. In general, the
route length of the cracks formed before the peak load may be
associated with the maximum load resisted by the specimen, so
that the maximum load was smaller when the route length of the
cracks formed before the peak load was larger. Deviations from
this rule could be due to partially damaged elements, which are
not shown in Fig. 12. In fact, that portion of the external energy
contributing to damage does not participate in the load-bearing
mechanism.

Fracture Propagation and Coalescence in the


Specimen Containing Two Parallel Fissures

The failure deformation of sandstone specimens containing two co-


planar fissures was experimentally investigated for different

Fig. 14. Experimental results for the axial stress–strain curve of the Fig. 15. Numerical results for the sandstone specimen containing two
sandstone specimen containing two coplanar fissures (a ¼ b ¼ 45 Þ coplanar fissures: (a) fracture propagation and coalescence (a ¼ b ¼
[reprinted from Yang (2011), with permission] 45 Þ; (b) axial stress-strain curve (a ¼ b ¼ 45 Þ

© ASCE 04018070-10 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

Fig. 16. Fracture propagation and coalescence in the sandstone specimens containing two parallel fissures (a ¼ 45 Þ for different ligament
angles

The stress–strain curves marked on frames (a) through (i) are occurs immediately after the peak load, with a value similar to that
shown in Fig. 14. As the fracture propagated, the stress level obtained experimentally.
increases to the peak load, representing stable fracture propagation. Other types of coalescence have also been reported, by depend-
After peak load was reached, local unstable fracturing results in a ing on the relative location of the pre-existing fissures (Shen et al.
fluctuating stress state from (d) to (h). Overall failure of the speci- 1995). In this study, we controlled the relative position of the fis-
men occurred within a short period of time, when frame (h) devel- sures by the ligament angle ( b ). Three additional numerical models
ops to frame (i). The results of our numerical simulation for this with ligament angles of b ¼ f75 ; 105 ; 135 g were analyzed.
physical test are presented in Fig. 15. As for the specimen with a ligament angle of b ¼ 45 , the abrupt
Whereas the wing crack lengths observed in Yang’s experiment shear-induced post-peak failure of the specimens was preceded by
were approximately equal, our numerical simulation produced wing the stable propagation of four wing cracks from the end corners of
cracks whose lengths differ considerably [Fig. 15(a)]. This differ- the pre-existing fissures (Fig. 16).
ence occurs because, in contrast to the experimental results where According to Fig. 16, shear-type coalescence was expected
wing cracks form at only one fissure, four wing cracks begin propa- for the specimen with a ligament angle of b ¼ 75 , whereas ten-
gating from the two end corners of each fissure in the numerical sile coalescence was predicted for the other two ligament angles.
model. Simultaneous growth of internal wing cracks affected the These results are comparable to those obtained by Shen and his
stress redistribution around the tip of each internal wing crack, colleagues for precracked gypsum specimens under uniaxial
thereby confining further growth. The local instability observed in compression, where shear coalescence transitioned to tensile co-
Yang’s experiment was not repeated in our numerical simulations, alescence as the ligament angle increased and the pre-existing
because no further tensile cracks appeared after peak load. As a fissures overlapped (Shen et al. 1995). The reuslts of our numeri-
result, discrepancies exist between the numerical and experimental cal simulations confirm the notion that additional coalescence
axial stress–strain curves [Figs. 14 and 15(b)]. The numerical by secondary fractures occurs when the overlap between the fis-
results predicted a shear-type coalescence in the ligament area, sures is significant. The role of overlap between joints in the fail-
which confirms the experimental results by Yang. This coalescence ure mode of rock masses has been noted by other researchers

© ASCE 04018070-11 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


(Sarfarazi et al. 2014; Gehle and Kutter 2003; Prudencio and Jan Basu, A., Mishra, D. A., and Roychowdhury, K. (2013). “Rock failure
2007). modes under uniaxial compression, Brazilian, and point load tests.” Bull
. Eng. Geol. Environ., 72(3–4), 457–475.
Bobet, A., and Einstein, H. H. (1998). “Fracture coalescence in rock-type
materials under uniaxial and biaxial compression.” Int. J. Rock Mech.
Conclusions Min. Sci., 35(7), 863–888.
Cao, R. H., Cao, P., Fan, X., Xiong, X., and Lin, H. (2016). “An experimen-
In this study, the element removal option of Abaqus standard tal and numerical study on mechanical behavior of ubiquitous-joint brit-
(v6.14) was used to numerically investigate the failure behavior of a tle rock-like specimens under uniaxial compression.” Rock Mech. Rock
sandstone specimen containing a single fissure. Using the UMAT Eng., 49(11), 4319–4338.
interface of Abaqus, a material user subroutine was written based Cvitanovic, N. Š., Nikolic, M., and Ibrahimbegovic, A. (2015). “Influence
on an elastic brittle damage constitutive law with the Weibull distri- of specimen shape deviations on uniaxial compressive strength of lime-
bution of meso-scale stiffness and strength parameters. The consti- stone and similar rocks.” Int. J. Rock Mech. Min. Sci., 80, 357–372.
Duvant, G., and Lions, J. L. (1976). Inequalities in mechanics and physics,
tutive model and how it functions within the Abaqus framework
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

Springer Science and Business Media, Berlin.


were described in detail. Numerical analyses were performed using Feng, X.-T., Pan, P.-Z., and Zhou, H. (2006). “Simulation of the rock micro-
plain strain reduced integration elements with hourglass control and fracturing process under uniaxial compression using an elasto-plastic
incorporated geometric nonlinearity. The written subroutine was cellular automaton.” Int. J. Rock Mech. Min. Sci., 43(7), 1091–1108.
verified, and the numerical model was validated by comparing its Gehle, C., and Kutter, H. K. (2003). “Breakage and shear behavior of inter-
predictions with available experimental results for sandstone mittent rock joints.” Int. J. Rock Mech. Min. Sci., 40(5), 687–700.
specimens. Ghazvinian, A., Nejati, H. R., Sarfarazi, V., and Hadei, M. R. (2013).
Parametric studies were conducted on specimens containing one “Mixed mode crack propagation in low brittle rock-like materials.”
or two pre-existing fissures. The following conclusions were Arabian J. Geosci., 6(11), 4435–4444.
Ghazvinian, A., Sarfarazi, V., Schubert, W., and Blümel, M. (2012). “A
deduced from the numerical results. study of the failure mechanism of planar non-persistent open joints
• The heterogeneity level had a considerable influence on the
using PFC2D.” Rock Mech. Rock Eng., 45(5), 677–693.
failure behavior of the fissure-included specimen in terms of Haeri, H. (2015). “Crack analysis of pre-cracked brittle specimens under
both the stress-strain curve and fracture pattern. Specimen biaxial compression.” J. Min. Sci., 51(6), 1091–1100.
stiffness and strength decreased with increases in the heteroge- Haeri, H., Shahriar, K., Marji, M. F., and Moarefvand, P. (2014).
neity level. The tensile cracks generated in less heterogeneous “Experimental and numerical study of crack propagation and coales-
models began to disappear, whereas more heterogeneous mod- cence in pre-cracked rock-like disks.” Int. J. Rock Mech. Min. Sci., 67,
els with homogeneity indexes ranging from 1.1 to 2 experi- 20–28.
Haimson, B., and Chang, C. (2000). “A new true triaxial cell for testing me-
enced scattered damaged elements.
chanical properties of rock, and its use to determine rock strength and
• Although applying a confining pressure led a specimen with
deformability of Westerly granite.” Int. J. Rock Mech. Min. Sci.,
the same inclination angle to resist a higher load, the specimen 37(1–2), 285–296.
elastic stiffness appeared to be largely insensitive to the con- Handid, J., and Hager, R. V., Jr. (1957). “Experimental deformation of sedi-
fining pressure. Moreover, the fissure inclination angle had a mentary rocks under confining pressure: Tests at room temperature on
minor direct effect on the specimen elastic stiffness. dry samples.” Am. Assoc. Pet. Geol. Bull., 41(1), 1–50.
• Despite some local decreases, the peak load generally Heap, M. J., Vinciguerra, S., and Meredith, P. G. (2009). “The evolution of
increased with an increase in the fissure inclination angle from elastic moduli with increasing crack damage during cyclic stressing of a
basalt from Mt. Etna volcano.” Tectonophysics, 471(1–2), 153–160.
0° to 90°. The maximum load was typically lower when the
Lavrov, A. (2003). “The Kaiser effect in rocks: Principles and stress estima-
route length of the cracks formed before the peak load was tion techniques.” Int. J. Rock Mech. Min. Sci., 40(2), 151–171.
larger. Lavrov, A. (2005). “Fracture-induced physical phenomena and memory
• For a specimen containing two parallel fissures, the mode of effects in rocks: A review.” Strain, 41(4), 135–149.
fracture coalescence changed from tensile to shear as the liga- Lee, H., and Jeon, S. (2011). “An experimental and numerical study of frac-
ment inclination angle increased from 45° to 135°. ture coalescence in pre-cracked specimens under uniaxial compression.”
Int. J. Solids Struct., 48(6), 979–999.
Lemaitre, J., and Desmorat, R. (2005). Engineering damage mechanics,
Acknowledgments Springer-Verlag, Berlin.
Li, G., and Tang, C. A. (2015). “A statistical meso-damage mechanical
method for modeling trans-scale progressive failure process of rock.”
The work described in this paper was partially supported by ARC Int. J. Rock Mech. Min. Sci., 74, 133–150.
Future Fellowship Grant FT140100019 and ARC Discovery Li, Y.-P., Chen, L.-Z., and Wang, Y.-H. (2005). “Experimental research on
Project Grant DP140100509, for which the authors are grateful. pre-cracked marble under compression.” Int. J. Solids Struct., 42(1),
2505–2516.
Lu, Y., Wang, L., and Elsworth, D. (2015). “Uniaxial strength and failure in
References sandstone containing a pre-existing 3-D surface flaw.” Int. J. Fract.,
194(1), 59–79.
Abaqus [Computer software]. SIMULIA, Providence, RI. Manouchehrian, A., and Marji, M. F. (2012). “Numerical analysis of
Aliha, M. R. M., Ayatollahi, M. R., Smith, D. J., and Pavier, M. J. (2010). confinement effect on crack propagation mechanism from a flaw in
“Geometry and size effects on fracture trajectory in a limestone rock a pre-cracked rock under compression.” Acta Mech. Sin., 28(5),
under mixed mode loading.” Eng. Fract. Mech., 77(11), 2200–2212. 1389–1397.
Aliha, M. R. M., Pakzad, R., and Ayatollahi, M. R. (2013). “Numerical Manouchehrian, A., Sharifzadeh, M., Marji, M. F., and Gholamnejad, J.
analyses of a cracked straight-through flattened Brazilian disk specimen (2014). “A bonded particle model for analysis of the flaw orientation
under mixed-mode loading.” J. Eng. Mech., 140(1), 219–224. effect on crack propagation mechanism in brittle materials under com-
Bahaaddini, M., Hagan, P., Mitra, R., and Hebblewhite, B. K. (2015). pression.” Arch. Civil Mech. Eng., 14(1), 40–52.
“Numerical study of the mechanical behavior of nonpersistent jointed Mogi, K. (1971). “Fracture and flow of rocks under high triaxial compres-
rock masses.” Int. J. Geomech., 16(1), 04015035. sion.” J. Geophys. Res., 76(5), 1255–1269.

© ASCE 04018070-12 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070


Pakzad, R., and Ayatollahi, M. R. (2015). “A comprehensive study on crack Wong, R. H. C., Chau, K. T., Tang, C. A., and Lin, P. (2001). “Analysis
tip parameters of modified ring specimen for mixed-mode fracture of crack coalescence in rock-like materials containing three flaws—
toughness tests on brittle materials.” Geotech. Test. J., 39, 144–156. Part I: Experimental approach.” Int. J. Rock Mech. Min. Sci., 38(7),
Pakzad, R., Wang, S. Y., and Sloan, S. W. (2016). Numerical investigation 909–924.
of damage propagation in heterogeneous quasi-brittle materials. Int. Wong, R. H. C., Tang, C. A., Chau, K. T., and Lin, P. (2002). “Splitting fail-
Conf. on Geo-Mechanics, Geo-Energy and Geo-Resources. Monash ure in brittle rocks containing pre-existing flaws under uniaxial com-
Univ., Melbourne, Australia. pression.” Eng. Fract. Mech., 69(1), 1853–1871.
Peng, S., and Johnson, A. M. (1972). “Crack growth and faulting in cylin- Wu, Z., and Wong, L. N. Y. (2013). “Elastic–plastic cracking analysis for
drical specimens of Chelmsford granite.” Int. J. Rock Mech. Min. Sci. brittle–ductile rocks using manifold method.” Int. J. Fract., 180(1),
Geomech. Abstr., 9(1), 37–86. 71–91.
Prudencio, M., and Jan, M. V. S. (2007). “Strength and failure modes of Yang, S. Q. (2011). “Crack coalescence behavior of brittle sandstone sam-
rock mass models with non-persistent joints.” Int. J. Rock Mech. Min. ples containing two coplanar fissures in the process of deformation fail-
Sci., 44(6), 890–902. ure.” Eng. Fract. Mech., 78(17), 3059–3081.
Sagong, M., and Bobet, A. (2002). “Coalescence of multiple flaws in a Yang, S. Q., Dai, Y. H., Han, L. J., and Jin, Z. Q. (2009). “Experimental
Downloaded from ascelibrary.org by Glasgow University Library on 04/26/18. Copyright ASCE. For personal use only; all rights reserved.

rock-model material in uniaxial compression.” Int. J. Rock Mech. Min. study on mechanical behavior of brittle marble samples containing dif-
Sci., 39(2), 229–241. ferent flaws under uniaxial compression.” Eng. Fract. Mech., 76(12),
Sarfarazi, V., Ghazvinian, A., Schubert, W., Blumel, M., and Nejati, H. 1833–1845.
(2014). “Numerical simulation of the process of fracture of echelon rock Yang, S.-Q., and Jing, H.-W. (2011). “Strength failure and crack coales-
joints.” Rock Mech. Rock Eng., 47(4), 1355–1371. cence behavior of brittle sandstone samples containing a single fissure
Shen, B., Stephansson, O., Einstein, H. H., and Ghahreman, B. (1995). under uniaxial compression.” Int. J. Fract., 168(2), 227–250.
“Coalescence of fractures under shear stresses in experiments.” J. Yang, S. Q., Jiang, Y. Z., Xu, W. Y., and Chen, X. Q. (2008).
Geophys. Res., 100(B4), 5975–5975. “Experimental investigation on strength and failure behavior of pre-
Su, X., Yang, Z., and Liu, G. (2010). “Finite element modelling of complex cracked marble under conventional triaxial compression.” Int. J. Solids
3D static and dynamic crack propagation by embedding cohesive ele- Struct., 45(17), 4796–4819.
ments in Abaqus.” Acta Mech. Solida Sin., 23(1), 271–282. Yang, S. Q., Liu, X. R., and Jing, H. W. (2013). “Experimental investigation
Tang, C. A., Liu, H., Lee, P. K. K., Tsui, Y., and Tham, L. G. (2000). on fracture coalescence behavior of red sandstone containing two unpar-
“Numerical studies of the influence of microstructure on rock failure in allel fissures under uniaxial compression.” Int. J. Rock Mech. Min. Sci.,
uniaxial compression—Part I: Effect of heterogeneity.” Int. J. Rock 63, 82–92.
Mech. Min. Sci., 37(4), 555–569. Yang, S. Q.., Xu, T., He, L., Jing, H. W., Wen, S., and Yu, Q. L. (2015).
Turichshev, A., and Hadjigeorgiou, J. (2016). “Triaxial compression experi- “Numerical study on failure behavior of brittle rock specimen contain-
ments on intact veined andesite.” Int. J. Rock Mech. Min. Sci., 86(1), ing pre-existing combined flaws under different confining pressure.”
179–193. Arch. Civ. Mech. Eng., 15(4), 1085–1097.
Wang, S. Y., Sloan, S. W., Sheng, D. C., and Tang, C. A. (2012a). Zhang, G. M., Liu, H., Zhang, J., Wu, H. A., and Wang, X. X. (2010).
“Numerical analysis of the failure process around a circular opening in “Three-dimensional finite element simulation and parametric study for
rock.” Comput. Geotech., 39, 8–16. horizontal well hydraulic fracture.” J. Pet. Sci. Eng., 72(3–4), 310–317.
Wang, S. Y., Sloan, S. W., Tang, C. A., and Zhu, W. C. (2012b). Zhang, X.-P. and Wong, L. N. Y. (2012). “Cracking processes in rock-like
“Numerical simulation of the failure mechanism of circular tunnels in material containing a single flaw under uniaxial compression: A numeri-
transversely isotropic rock masses.” Tunnelling Underground Space cal study based on parallel bonded-particle model approach.” Rock
Technol., 32, 231–244. Mech. Rock Eng., 45(5), 711–737.
Wong, L. N. Y., and Einstein, H. H. (2009a). “Crack coalescence in molded Zhao, Y., Zhang, L., Wang, W., Pu, C., Wan, W., and Tang, J. (2016).
gypsum and Carrara marble: Part 1. Macroscopic observations and inter- “Cracking and stress–strain behavior of rock-like material containing
pretation.” Rock Mech. Rock Eng., 42(3), 475–511. two flaws under uniaxial compression.” Rock Mech. Rock Eng., 49(7),
Wong, L. N. Y., and Einstein, H. H. (2009b). “Systematic evaluation of 2665–2687.
cracking behavior in specimens containing single flaws under uniaxial Zhu, W., and Tang, C. A. (2004). “Micromechanical model for simulating
compression.” nt. J. Rock Mech. Min. Sci., 46(2), 239–249. the fracture process of rock.” Rock Mech. Rock Eng., 37, 25–56.

© ASCE 04018070-13 Int. J. Geomech.

Int. J. Geomech., 2018, 18(7): 04018070

You might also like