Tunnelling and Underground Space Technology 74 (2018) 71–81

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

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

Modeling the progressive failure of hard rock pillars T

⁎
H. Rafiei Renani , C.D. Martin
Department of Civil & Environmental Engineering, University of Alberta, Edmonton T6G 1H9, Canada

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

Keywords: An enhanced cohesion weakening and friction strengthening model is proposed to simulate the progressive
Pillar strength failure of hard rock pillars using two- and three-dimensional finite difference analysis. The documented behavior
CWFS model of 85 pillars in two mines was used for model validation. The numerical models successfully separated unstable/
Failure mechanism failed pillars from stable pillars and also captured the observed mechanism of progressive failure in hard rock
pillars. Pillar stress-strain curves indicated that while pillars with W/H < 2 exhibit strain softening, wider pillars
begin to display strain hardening. It was shown that strength of wide pillars may be overestimated by con-
ventional models using empirically estimated parameters.

1. Introduction
Estimating the strength of pillars is a necessary step in the design of
underground mines. In conventional mining methods such as room and
pillar operations, economic incentives point towards higher excavation
ratios by leaving smaller pillars between the stopes while safety and
stability requirements favor wider stronger pillars. In some mining
methods such as caving operations, optimal design of pillars is critical
as not only pillars too narrow can jeopardize mine safety and stability
but also pillars too wide can block the flow of ore through drawpoints.
Therefore, pillar design poses a somewhat unique challenge to rock
engineers where simple conservative approaches frequently used in
other applications are no longer acceptable.
The most straight-forward method for estimation of pillar strength is
using empirical formulas based on the analysis of large number of pil-
lars in different mines. As in any empirical method, however, it is es-
sential to note the conditions and ranges of variables in the data sets
used for developing the formulas. Many of the pillars in the empirical
data sets have limited width to height ratios and occur at relatively
shallow depths in high quality rock masses (Lunder and Pakalnis, 1997;
Martin and Maybee, 2000; Kaiser et al., 2011). Extending the applica-
tion of empirical formulas beyond these ranges is unwarranted. In ad-
dition, the approach is very simplified and the effect of many para-
meters such as the condition of roof and floor, pillar-foundation
interface and in situ stress ratio cannot be explored by the empirical
formulas. Finally, these formulas only estimate the pillar strength while
knowledge of the pre- and post-peak behavior is also important in
mining operations.
Another approach to pillar design is using numerical modeling. This
method can take into account most of the variables and complexities
affecting pillar behavior. In addition, the complete pillar behavior, from
initial loading to post-peak, can be captured. However, realistic mod-
eling of pillars requires a comprehensive knowledge of material beha-
vior and implementing a representative material model. Behavior of
rocks can be described using mechanistic models (analytically derived
from the principles of fracture and damage mechanics) or phenomen-
ological models (empirically derived to describe experimental ob-
servations). Examples of mechanistic models are given in Griffith
(1920, 1924), McClintock and Walsh (1962), Zhou et al. (2004, 2008),
Zhou (2004), Zhou and Yang (2007), Abou-Chakra Guery et al. (2008),
Bui et al. (2017). Phenomenological models such as those presented in
this study are commonly used in practical design and analysis of large
scale rock engineering structures.
Iannacchione (1989) and Whyatt and Board (1991) made early at-
tempts to use strain softening models in finite difference analysis to
explore the behavior of pillars. Martin and Maybee (2000) adopted
brittle parameters in a finite element analysis to predict the strength of
rock pillars. Duncan Fama et al. (1995) and Adhikary et al. (2002) used
strain softening models in finite difference and finite element analyses
to explore the behavior of coal pillars. Mortazavi et al. (2009) used the
finite difference method and strain softening model to capture the be-
havior of hard rock pillars. Elmo and Stead (2010) used a hybrid finite
element/discrete element method to explicitly model the fracture net-
work inside pillars. Kaiser et al. (2011) presented a critical review of the
common models and used an s-shaped failure envelope in finite element
analysis of rock pillars.
The focus of this paper is to model the behavior of hard rock pillars
using a Cohesion Weakening and Friction Strengthening (CWFS) model.

⁎
Corresponding author.
E-mail address: rafiaire@ualberta.ca (H. Rafiei Renani).

https://doi.org/10.1016/j.tust.2018.01.006
Received 4 July 2017; Received in revised form 1 January 2018; Accepted 5 January 2018
Available online 12 January 2018
0886-7798/ © 2018 Elsevier Ltd. All rights reserved.
H. Rafiei Renani, C.D. Martin Tunnelling and Underground Space Technology 74 (2018) 71–81

Table 1
Empirical formulas for hard rock pillar strength, σp (W and H are the width and height of pillar in metres, respectively, σc is the uniaxial compressive strength of intact rock and κ is the
mine pillar friction term.

Reference Pillar strength, σp σc (MPa) No. of pillars Rock mass

Hedley and Grant (1972) 133 (W 0.5/ H 0.75) 230 28 Quartzite

von Kimmelmann et al. (1984) 65 (W 0.46/ H 0.66) 94 57 Metasediments
Krauland and Soder (1987) 35.4 [0.778 + 0.222(W / H )] 100 14 Limestone
Potvin et al. (1989) 0.42σc (W / H ) – 23 Canadian shield
Sjoberg (1992) 74 [0.778 + 0.222(W / H )] 240 9 Limestone/Skarn
Lunder and Pakalnis (1997) 0.44σc (0.68 + 0.52κ ) – 178 Hard rocks

50 Determination of the critical size requires testing of very large samples.

0.33 ıc Performing such tests is very difficult and costly and the number of tests
on sufficiently large samples is very limited. The results of independent
40 large scale tests carried out by Bieniawski (1968) on coal and by Pratt
Number of pillars

et al. (1972) on diorite suggest that the critical size for these materials is
Pillar strength about 1 m.
30 In order to estimate the strength of large scale pillars, it is useful to
Mean = 75 MPa
StDev = 24 MPa find a relationship between the uniaxial compressive strength of intact
20 rock and the pillar strength. This relationship bridges the size gap be-
Intact rock strength tween the laboratory samples and large scale pillars. To this end, a
ıc = 230 MPa reference pillar, representative of large scale pillars is defined. The size
10 of the reference pillar must be no smaller than the critical size. Since the
strength of the reference pillar is compared with strength of intact la-
boratory sample, the width to height ratio of the reference pillar must
0 also be similar to that of laboratory specimen. Choosing the width of
0 25 50 75 100 125 150 175 200 1 m for the reference pillar ensures that the results will be valid for
Pillar strength (MPa) larger pillars (no size effect beyond the critical size of 1 m). In order to
avoid the introduction of shape effects in this analysis, the width to
Fig. 1. Distribution of pillar strength at the Quirke mine, after (Swan et al., 1985). height ratio of pillar is chosen as 0.5 to be similar to the diameter to
length ratio of the laboratory test sample (ASTM, 2004).
Empirical formulas and factors affecting pillar strength are reviewed. Strength of the reference pillar (W = 1 m and H = 2 m) can be es-
The merits of a CWFS model in simulating progressive failure of hard timated from the empirical formulas. The ratio of the reference pillar
rocks are illustrated. An enhanced CWFS is implemented in the finite strength to intact uniaxial compressive strength is defined as the in situ
difference code FLAC3D (Itasca Inc., 2009) to model hard rock pillars. strength factor K. Using the empirical formulas in Table 1, in situ
Documented behavior of pillars at the Elliot Lake and Selebi-Phikwe strength factor K ranges from 0.21 to 0.44 with the average value of
mines is used to evaluate the proposed modeling approach. 0.31. Based on comprehensive back analysis of pillars at the Quirke
mine, Swan (1985) also found that the ratio of mean pillar strength to
2. Strength of hard rock pillars mean intact uniaxial compressive strength is 0.33 (Fig. 1).

Following the failure of Coalbrook coal mine in 1960, extensive

research was initiated in South Africa to establish the strength of coal
pillars. One of the early works is due to Salamon and Munro (1967) who
analyzed 125 cases of coal pillar failure and expressed the strength of
pillar as a power function of its width and height. Their findings were
used by Hedley and Grant (1972) who analyzed 28 hard rock rib pillars
at the Elliot Lake uranium mines and proposed a similar power function
with modified exponents. Alternative pillar strength formulas were
proposed by von Kimmelmann et al. (1984), Krauland and Soder
(1987), Potvin et al. (1989) and Sjoberg (1992). Lunder and Pakalnis
(1997) later compiled perhaps the most extensive data base of 178 hard
rock pillars and suggested a new pillar strength formula. Table 1
summarizes empirical formulas that have been reported for the design
of hard rock pillars. As evident from Table 1 and pointed out by
Bieniawski (1992) and Lunder and Pakalnis (1997), pillar strength is
influenced by size effect and shape effect.
2.1. Size effect
Size effect refers to the reduction of strength by increasing the size
of the test sample. It is a fundamental characteristic of heterogeneous
materials and is caused by the increasing number of weaker and softer
elements within larger samples. As suggested by Bieniawski (1968) and
Martin et al. (2012), there is a critical size above which there will be no
further reduction of strength with increasing sample size.
2.2. Shape effect
The strength of pillars is also influenced by the shape of pillar ex-
pressed as the width to height ratio in the empirical formulas in Table 1.
The relationship between the pillar strength normalized by intact uni-
axial compressive strength and width to height ratio of the pillar is
shown in Fig. 2. It can be observed that the pillar strength increases
with increasing width to height ratio. This trend is similar to the one
observed in the laboratory compression tests on specimens with dif-
ferent diameter to length ratios (Hawkes and Mellor, 1970). As in the
case of laboratory tests, the actual factor which causes an increased
strength for wider pillars is the end constraints which induce higher
confinements within the pillar.
In order to explore the relationship between the width to height
ratio and induced confinement, a series of uniaxial compression tests
were carried out on elastic rib pillars using the finite difference code
FLAC3D (Itasca Inc., 2009). Fig. 3 shows the profile of minor principal
stress (confinement) normalized by the average major principal stress
(average pillar stress) across the mid-height of pillars with different
width to height ratios. It can be observed that increasing width to
height ratio significantly increases induced confinement which in turn
leads to higher pillar strength. It is also worth noting that for a pillar
with W/H ratio of 0.5, tensile stresses are induced within the pillar. It
suggests that the behavior of very slender pillars (W/H < 0.5) is

72
H. Rafiei Renani, C.D. Martin Tunnelling and Underground Space Technology 74 (2018) 71–81

1 propagation of cracks and gradual loss of inter- and intra-granular

Hedley & bonds. Gradual propagation and coalescence of cracks also leads to
Grant (1972)
Lunder & shear displacement on crack surfaces resisted by frictional forces and
0.8 thereby mobilization of friction angle. While cohesion exhibits a
Pakalnis (1997)
c

monotonic reduction to a residual value, friction angle reaches a peak

von Kimmelmann
Pillar strength /

et al. (1984) slightly above the ultimate value. The modest difference between peak
0.6
and ultimate friction angle is attributed to the temporary effect of crack
surface roughness and interlocking which diminishes as damage con-
0.4 tinues. More details on the experimental findings and theoretical con-
cepts regarding cohesion degradation and friction mobilization are
Krauland & given by Martin and Chandler (1994) and Martin (1997).
Sjoberg
Soder (1987)
0.2 (1992)
Potvin et
al. (1989) 3.2. In situ observations

0 Pritchard and Hedley (1993) made detailed observations of the

0 0.5
Fig. 2. The effect of shape on pillar strength based on empirical formulas.
0.40
Confinement/Average pillar stress
1 1.5 2 2.5 3 evolution of deformation and process of failure in hard rock pillars from
W/H the Elliot Lake region. Their observations clearly showed that minor
spalling marks the initiation of failure. Loss of the detached spalled rock
induces increased loads on the pillar causing a progressive process of
spalling and slabbing. Gradual spalling results in the loss of load car-
rying capacity of the pillar and finally leaves a narrow hourglass shaped
W/H

0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
-0.05
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Distance from the center/Pillar width
Fig. 3. Profiles of normalized confinement at the mid-height of elastic pillars with dif-
ferent W/H ratios.
governed by tensile mechanisms, splitting and buckling. Analyzing
these modes of failure requires special considerations and in beyond the
scope of this study.
3. Progressive failure and the CWFS model
While empirical pillar strength formulas provide an initial estimate
of pillar peak strength, they are unable to capture the effect of complex
geometries, in situ stress fields, and material behavior. Numerical
modeling provides a broader and more detailed insight into the pre- and
post-peak behavior of pillars. However, obtaining realistic results from
numerical models strongly depends on the implementation of a re-
presentative material model. In order to determine a representative
material model for a specific problem, it is necessary to analyze the
process of deformation and mechanisms of failure observed in the la-
boratory and in situ. Such analysis provides invaluable information for
selecting a representative material model, which is most capable of
capturing the observed mechanisms.
3.1. Laboratory observations
Using a series of damage-controlled triaxial tests on Lac du Bonnet
granite, Martin and Chandler (1994) showed that contrary to popular
assumption, cohesion and friction are not mobilized simultaneously in
brittle failure. They showed that as stress level increases and damage
accumulates within the material, the cohesive component of strength
decreases and frictional component increases. The evolution of cohe-
sion and friction for Lac du Bonnet granite under uniaxial compression
is shown in Fig. 4a. Degradation of cohesion is due to initiation and
pillar which is then vulnerable to shear failure along the weak struc-
4.0
3.0 tural features such as bedding planes and joints. Martin (1997) ob-
2.0 served a similar process of progressive failure around the Mine-by test
tunnel at the Underground Research Laboratory (URL) in Canada.
1.5
1.0 3.3. Numerical modeling
0.5 Hajiabdolmajid et al. (2002) observed that the conventional models
were unable to reproduce the mechanisms of progressive failure ob-
served around the Mine-by tunnel at the URL. Based on the experi-
mental and theoretical works of Martin and Chandler (1994) on pro-
gressive failure, Hajiabdolmajid et al. (2002) used a Cohesion
Weakening and Friction Strengthening (CWFS) model to capture the
progressive failure process. In this model, both cohesion and friction
angle are expressed as functions of plastic strain ε p. Fig. 4b shows the
evolution of cohesion and friction angle as adopted by Hajiabdolmajid
et al. (2002). It can be observed that this model linearly approximates
the trend of cohesion degradation and friction mobilization shown in
Fig. 4a. Note that the modest peak of friction angle due to temporary
interlocking effects observed in the laboratory is not relied on for in situ
stability analysis. This linear CWFS model reproduced the failure pat-
tern observed in the field reasonably well. Diederichs (2007) utilized
analytical and micromechanical tools and Edelbro (2009) and Walton
and Diederichs (2015) reported case studies to further support the ap-
plication of a CWFS model for modeling brittle failure of rocks.
4. Numerical analysis of pillars with the CWFS model
In this study, behavior of typical hard rock pillars is analyzed using
continuum models. While in situ rock masses normally contain frac-
tures and discontinuities, a continuum model with equivalent para-
meters can be used to model pillar behavior as long as fractures are
distributed quite randomly in different directions and their size is
considerably smaller than the pillar. In such cases the effect of dis-
continuities can be implicitly taken into account by assigning weakened
strength parameters to a continuum model (Martin and Maybee, 2000;
Adhikary et al., 2002; Mortazavi et al., 2009; Hoek and Brown, 1980;
Murali Mohan et al., 2001; Dolinar and Esterhuizen, 2007; Jaiswal and
Shrivasta, 2009; Wang et al., 2011). In cases where there is a large
prominent structure or fault crossing the pillar, the behavior is con-
trolled by the discontinuity and a discontinuum model is more relevant.
Such pillars, however, are normally avoided in practice due to limited
strength and are beyond the scope of this study.
Considering the laboratory and in situ evidence for application of
the CWFS model to progressive failure, it was used to simulate the

73
H. Rafiei Renani, C.D. Martin Tunnelling and Underground Space Technology 74 (2018) 71–81

(a) (b) Fig. 4. Evolution of cohesion and friction with da-

mage. (a) test results on Lac du Bonnet granite,

Normalized cohesion Friction angle 60 60 modified after (Martin and Chandler, 1994), and (b)
numerical model for the Mine-by tunnel, modified

100 Friction angle after (Hajiabdolmajid et al., 2002).

50 50
Normalized cohesion (%) /

Friction angle (Deg)

Cohesion (MPa)
80 40 40
Friction angle (deg)

60 30 30

40 20 Cohesion 20

20 10 10

0 0 0
0.0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1
Normalized plastic strain Plastic shear strain (%)

behavior of hard rock pillars. The CWFS model was implemented in the mobilization are illustrated in Fig. 5a using an example. It can be ob-
finite difference code FLAC3D (Itasca Inc., 2009) which allows auto- served that increasing damage and plastic strain causes a gradual loss of
matic adjustment of cohesion and friction angle at each calculation time cohesion and mobilization of friction. Compared to Fig. 4, it is evident
step to follow any specified trend of cohesion degradation and friction that the smooth and nonlinear cohesion degradation and friction mo-
mobilization. The CWFS model was enhanced to respect the nonlinear bilization model specified by Eqs. (1) and (2) capture the experimental
nature of damage and to avoid sharp changes in the rate of cohesion results more closely than a linear model with sharp changes. As shown
degradation and friction mobilization. The following empirical equa- in Fig. 5b, the proposed model also leads to realistic stress-strain curves
tions are proposed to describe cohesion degradation and friction mo- and captures the fundamental behavior of transition from strain soft-
bilization of rock mass as smooth functions of plastic strain: ening to strain hardening behavior with increasing confinement.
Aside from using the enhanced CWFS model, the numerical mod-
⎛ ⎞ eling framework adopted in this study follows the typical procedure for
2
c = cult + (cini−cult ) ⎜2− ⎟ obtaining the complete stress-strain curves of pillars (Iannacchione,
⎝ εc
p
⎜ 1 + exp −3.66 εp ∗
( ) ⎟
⎠ (1) 1989; Adhikary et al., 2002; Murali Mohan et al., 2001; Dolinar and
Esterhuizen, 2007; Jaiswal and Shrivasta, 2009; Wang et al., 2011). The
geometric boundaries of the pillar model were chosen to minimize the
⎛ ⎞
⎜ 2 undesirable boundary effects. The vertical extent of the model was 15
φ = φini + (φult −φini ) −1⎟
⎜ εp ⎟ times the height of the pillar. The distance between horizontal
⎜ 1 + exp ⎛− 3.66 ⎞ ⎟
εφp ∗ (2) boundaries was 3 times the width of the pillar representing a road way
⎝ ⎝ ⎠ ⎠
twice as wide as the pillar. The boundary conditions of the model
where cini and cult are the initial and ultimate (degraded) values of co- consisted of fixed horizontal displacement in the sides, fixed vertical
hesion (cini ⩾ cult ), φini and φult are the initial and ultimate (mobilized) displacement at the bottom, and constant velocity on the top inducing
values of friction angle (φini ⩽ φult ), and εcp ∗ and εφp ∗ are plastic strains at compression within the pillar. The behavior of pillars at the Elliot Lake
which cohesion and friction angle are within 5% of their ultimate va- and Selebi-Phikwe mines was analyzed under plane strain condition
lues, respectively. typical in long pillars (Martin and Maybee, 2000; Kaiser et al., 2011;
Note that Eqs. (1) and (2) are purely empirical and are not derived Iannacchione, 1989; Whyatt and Board, 1991; Adhikary et al., 2002;
from any analytical models. They were chosen through a process of trial Mortazavi et al., 2009). A uniform mesh with square elements was used
and error with many different forms of equations to describe the desired throughout those models with 30 elements across the width of the
evolution of cohesion and friction angle during brittle failure. Im- pillar. A three-dimensional model with similar boundary conditions was
plications of Eqs. (1) and (2) on cohesion degradation and friction

(a) (b) Fig. 5. CWFS model for the rock mass at the Elliot Lake
mine: (a) cohesion loss and friction mobilization, and (b)
40 40 180 stress-strain curves under uniaxial and triaxial compres-
3= 40 MPa sion.
35 35 160
30 Friction angle 30 140 30 MPa
Friction angle (Deg)

Axial stress (MPa)

Cohesion (MPa)

120
25 25
100 20 MPa
20 20
80
15 15 10 MPa
60
10 10 40
Cohesion
5 5 20 0 MPa
0 0 0
0 0.2 0.4 0.6 0.8 1 0 0.1 0.2 0.3 0.4 0.5
Plastic shear strain (%) Axial strain (%)

74
H. Rafiei Renani, C.D. Martin
also analyzed to evaluate the applicability of the findings to pillars with
complex geometries.
The overall deformation behavior of the pillar was quantified using
total vertical strain across the height of the pillar. Average pillar stress
was recorded by averaging the major principal stress over elements
within a narrow range of the mid-height of the pillar. The strength of
the pillar is determined as the peak value of average pillar stress during
compression. The stresses at the core of the pillar (central element
within the pillar) were also monitored to provide a more detailed de-
scription of stress distribution.
5. Application cases
To illustrate the application of the new enhanced CWFS model,
behavior of pillars in two hard rock mines was analyzed and the results
were compared with documented in situ observations. The parameters
of a CWFS model can be obtained using the results of damage-con-
trolled tests (Martin and Chandler, 1994) at different levels of confining
stress. However, such tests have been carried out only on intact samples
of few rock types. Hence, direct determination of parameters of a CWFS
model for large scale rock masses is currently impractical. The CWFS
parameters are typically estimated using accumulated experience from
case studies and back analyses (Hajiabdolmajid et al., 2002; Diederichs,
2007; Edelbro, 2009; Walton and Diederichs, 2015).
Martin and Chandler (1994) and Hajiabdolmajid et al. (2002) pre-
sented experimental and numerical evidence for initial friction angle of
zero which was adopted in this study. The classic Mohr-Coulomb re-
lationship between strength, cohesion and friction angle requires that
initial cohesion be 1/2 rock mass uniaxial compressive strength when
friction angle is zero (Martin and Chandler, 1994). For good quality
rock masses with Geological Strength Index, GSI of 80 typically found in
hard rock pillars, the Hoek-Brown criterion (Hoek et al., 2002) predicts
rock mass uniaxial compressive strength to be 1/3 of intact rock uni-
axial compressive strength. This is in keeping with empirical evidence
presented in Section 2.1. Hence, initial cohesion equal to 1/6 of intact
rock uniaxial compressive strength has been adopted in this study.
Ultimate cohesion is considerably lower than the initial cohesion.
Edelbro (2009) used ultimate cohesion to initial cohesion ratios of
7–14% with 10% being adopted in this study. Ultimate friction angle is
mobilized after significant plastic strain and cohesion degradation
which corresponds to the residual friction angle obtained from standard
triaxial or direct shear tests. Hence, typical residual friction angles re-
ported by Barton and Choubey (1977) have been used in this study as
ultimate friction angle. Hajiabdolmajid et al. (2002) and Diederichs
(2007) and Walton and Diederichs (2015) used mobilization plastic
strains of 0.2–0.5% with 0.5% being adopted in this study.
5.1. The Elliot Lake mine
The Elliot Lake uranium orebody is located in Ontario, Canada.
Tabular deposits of uranium bearing conglomerates were separated by
massive quantize beds with thickness of 3–30 m. The orebody was
mined using room-and-pillar and stope-and-pillar methods to a depth of
1 km. Uniaxial compressive strength of intact quartzite measured in the
laboratory was 230 MPa. In situ characterizations using the Q-system
and seismic wave velocity measurements indicated that pillars were
excavated in “good” to “very good” quality rock masses (Hedley and
Grant, 1972; Coates and Gyenge, 1981). Martin and Maybee (2000)
assigned a GSI value of 80 to the quartzite rock mass at the Elliot Lake
mine.
Based on the measured strength in the laboratory and in situ rock
mass characterizations, the estimated parameters of the CWFS model
for the pillars at the Elliot Lake mine are given in Table 2 where E and υ
are rock mass modulus and Poisson’s ratio, respectively. The variation
of cohesion and friction and the corresponding stress-strain curves
under uniaxial and triaxial compression are shown in Fig. 5. Note that
75
Tunnelling and Underground Space Technology 74 (2018) 71–81
the unconfined rock mass strength of 76 MPa from the CWFS model is
about 1/3 of the intact uniaxial compressive strength which is in
keeping with empirical evidence presented in Section 2.1.
5.2. The Selebi-Phikwe mine
The Selebi-Phikwe deposits of nickel and copper are located in
Botswana, Southern Africa. The strata-bound sulphide orebody is
hosted in highly deformed metasediments of Archaean age. Open
stoping and cut-and-fill stoping methods were used to extract the ore-
body to a depth of 500 m. Laboratory tests indicated that uniaxial
compressive of intact rock is 94 MPa. In situ rock mass characteriza-
tions showed that average rock mass rating (RMR) in the Selebi-Phikwe
mine is 85 (von Kimmelmann et al., 1984; Nareetsile, 1998). According
to the relationship proposed by Hoek and Brown (1997), a GSI value of
80 is estimated for the sulphide rock mass at the Selebi-Phikwe mine.
The parameters of the CWFS model estimated based on the laboratory
and in situ data for the pillars at the Selebi-Phikwe mine are given in
Table 2. The trend of cohesion degradation and friction mobilization
and also transition from strain softening to strain hardening is similar to
those at the Elliot Lake mine (Fig. 5).
6. Results
The average stress-strain curves for pillars at the Elliot Lake mine
and the Selebi-Phikwe mine are shown in Fig. 6. The width to height
ratio of analyzed pillars ranges from 0.5 to 3. It can be observed that
pillars with W/H values of 0.5 and 1 show very similar stress-strain
curves. For pillars with W/H ratios of 2, there is a slight reduction of
stresses after the peak. On the other hand, pillars with W/H ratios of 3
show a significant strain hardening behavior up to vertical strains of
about 1%.
In order to explore the effect of the CWFS model on the evolution of
stresses inside pillars, stress path at the core for pillars with W/H ratios
of 1 and 2 at the Elliot Lake mine are shown in Fig. 7. The initial co-
hesion dominated strength envelope (corresponding to the initial values
of cohesion and friction angle) and the ultimate friction dominated
strength envelope (corresponding to the ultimate values of cohesion
and friction angle) are also shown. For the pillar with W/H ratio of 1,
major principal stress at the core drops after reaching the initial co-
hesion strength envelope whereas it continues to increase along the
frictional strength envelope for the pillar with W/H ratio of 2.
Contours of plastic zone in the pillar models were monitored during
loading to examine the simulated mechanism of failure. As an example,
Fig. 8 shows the development of plastic zone in a pillar model with W/H
ratio of 1 along with the photo of a crushed pillar. It can be observed
that the model indicates a progressive mechanism of failure starting at
the corners, developing inside towards the core and eventually crossing
through the pillar.
Fig. 9a shows the condition of 28 pillars at the Elliot Lake mine
documented by Hedley and Grant (1972)). It is worth noting that only 2
pillars were classed as unstable (partially failed) and 3 pillars were
failed. The Elliot Lake pillars were analyzed using the CWFS model. The
results of the numerical analysis using the CWFS model and the pillar
strength formula developed by Hedley and Grant (1972) for the Elliot
Lake mine are also shown in Fig. 9a. It can be observed from Fig. 9a that
the CWFS model separates stable pillars from failed pillars just as well
as the Hedley and Grant (1972) empirical formula developed specifi-
cally from the Elliot Lake mine data. The slight difference is that the
empirical formula puts unstable pillars on the same side as stable pillars
where the CWFS model treats unstable pillars similar to failed pillars.
In the case of Selebi-Phikwe mine, a total of 57 pillars were docu-
mented among which 20 were noted as unstable and 22 were reported
as failed. Fig. 9b shows the reported condition of pillars at the Selebi-
Phikwe mine, the empirical pillar strength curve proposed by von
Kimmelmann et al. (1984) for the Selebi-Phikwe mine, and the results
 H. Rafiei Renani, C.D. Martin
Table 2
Estimated parameters of the CWFS model for pillars at the Elliot Lake and Selebi-Phikwe mines.
Mine σc (MPa) E (GPa) υ cini (MPa)
Elliot Lake 230 75 0.25
Selebi-Phikwe 94 71 0.25
of the numerical analysis using the CWFS model. The total number of
documented pillars and also the percentage of unstable/failed pillars at
the Selebi-Phikwe mine are significantly higher than those at the Elliot
Lake mine. However, there is also more scatter in this data set and the
boundary between stable and unstable/failed pillars is not as well de-
fined as that of the Elliot Lake mine. It can be seen from Fig. 9b that all
stable pillars are below the empirical pillar strength curve proposed by
von Kimmelmann et al. (1984). However, not all failed pillars are above
the empirical curve and unstable pillars are evenly distributed above
and below the curve. On the other hand, all the failed pillars and almost
all unstable pillars are above the CWFS-based pillar strength curve but
not all stable pillars are below the curve.
7. Discussion
The results of the numerical analysis using the CWFS model pre-
sented in Fig. 6 capture the fact that wider pillars have higher strength
than narrow pillars. In addition, it illustrates that the post-peak strength
of narrow pillars (W/H ≤ 1) is significantly lower than the peak
strength implying that failure of such pillars can be potentially violent.
On the other hand, wider pillars with W/H > 2 can maintain their
strength and even exhibit strain hardening behavior after initial
yielding. It suggests that the failure of such pillars tend to be more
gradual and less violent. In terms of displacements, Fig. 6 shows that
narrow pillars may lose their load bearing capacity at axial strains as
low as 0.1% whereas wide pillars can sustain strains of up to 1% before
losing their capacity. It supports the importance of accurate displace-
ment monitoring for narrow pillars.
While the stress-strain response of pillars at the Elliot Lake and
Selebi-Phikwe mines is not recorded and direct quantitative comparison
with model results is not possible, it is useful to qualitatively validate
the overall trend of simulated stress-strain curves. Wagner (1974) re-
corded stress-strain curves of coal pillars by installing hydraulic jacks
into cut slots at the mid-height of pillars and applying uniform dis-
placements. The hydraulic jacks were pre-loaded before starting the
tests to take up any surface irregularities and ensure uniform loading
contact. More details on the testing procedure are given by Wagner
(1974). Fig. 10a shows the recorded stress-strain curves of coal pillars
with W/H ratios of 1 and 2 and Fig. 10b shows the simulated stress
(a)
250
200
Pillar stress (MPa)
Pillar stress (MPa)
W/H = 3.0
150
W/H = 2.5
W/H = 2.0
100
W/H = 1.5
50 W/H = 1.0
W/H = 0.5
0 0
0 0.5 1 1.5 2 0
Pillar strain (%)
Tunnelling and Underground Space Technology 74 (2018) 71–81
cult (MPa) εcp ∗ (%) φini (deg) φult (deg) εφp ∗ (%)
38.0 3.8 0.5 0 35 0.5
15.7 1.6 0.5 0 34 0.5
curves for the Selebi-Phikwe mine. Obviously, strength and deform-
ability of coal and hard rocks are very different and a direct comparison
of absolute values of stress and strain is not warranted. However, the
overall trend of measured and simulated stress-strain curves compare
reasonably well. Both measured and simulated data show that pillars
with W/H = 2 are about 37% stronger and exhibit a slower post-peak
strength reduction compared to pillars with W/H = 1.
Fig. 7 illustrates one of the most important aspects of the CWFS
model applicable to hard rock pillars. For a pillars with W/H = 1, the
stress path at the core reaches the initial cohesive strength envelope and
then relaxes and follows the ultimate frictional strength envelope cor-
responding to ultimate values of cohesion and friction angle. This stress
path confirms strain softening behavior at the core of narrow pillars
under relatively low confinement. When pillar W/H is increased to 2,
on the other hand, the stress path moves up consistently passing the
initial cohesive envelope and moving along the ultimate strength en-
velope. This clearly captures the strain hardening behavior at the core
of wide pillars under high confinements.
Fig. 8 indicates that when average pillar stress in the model reaches
about 70% of the peak stress, initial signs of failure are observed at the
corners of the pillar. This is in keeping with the observations of Wagner
(1974) who reported the initiation of nonlinear behavior of pillars at
70% of peak stress. As loading continues, the process of failure pro-
pagates inside the pillar towards the core. At about 95% of the peak
stress, the upper and lower failed zones coalesce and form two distinct
break outs on both sides of the pillar. In reality, the two failed zones
detach from the pillar leaving an hourglass shaped pillar. Very slight
increase in loading after this point causes the two failed zones in both
sides to connect and form a through-going shear band. The simulated
mechanism of failure i.e., initiation of slabbing at the corners, con-
tinued spalling and formation of hourglass shaped pillars and finally
shear failure of the pillar, are in agreement with the observations of
Pritchard and Hedley (1993) on failure of hard rock pillars.
It can be concluded from Fig. 9 that both the CWFS model and
empirical formulas specifically developed for the mines are reasonably
successful in separating stable pillars from failed pillars. The CWFS
model gives a more conservative estimate of strength for pillars with
W/H ≤ 2.0 and puts unstable pillars on the same side as failed pillars
but that it not the case with empirical formulas. This difference is not
(b) Fig. 6. Average vertical stress-strain curves for pillars at
(a) the Elliot Lake mine, and (b) the Selebi-Phikwe mine.
90
80
70
60 W/H = 3.0
50 W/H = 2.5
40 W/H = 2.0
30 W/H = 1.5
20 W/H = 1.0
10 W/H = 0.5
0.5 1 1.5
Pillar strain (%)
76
 H. Rafiei Renani, C.D. Martin Tunnelling and Underground Space Technology 74 (2018) 71–81

(a) (b) Fig. 7. Stress path at the core of pillars at the Elliot Lake
mine with (a) W/H = 1, and (b) W/H = 2.
150 Ultimate 300
frictional
(MPa)

(MPa)
strength
125 250 Ultimate
Initial frictional
1

1
100 cohesive 200 strength
Major principal stress,

Major principal stress,

strength
75 150 Initial
cohesive
50 100 strength

25 50

0 0
0 10 20 30 40 50 0 20 40 60 80 100
Minor principal stress, 3 (MPa) Minor principal stress, 3 (MPa)

necessarily a disadvantage to empirical formulas and may reflect a sole
focus on separating stable and failed pillars and relative neutrality to-
wards unstable pillars when developing the empirical equations. De-
pending on the severity and consequence of pillar instability, past ex-
perience and engineering judgment should be exercised in practice to
establish acceptable range of pillar behavior.
It is also worth noting that the major principal stress at the core of
pillars can reach much higher values than the overall average pillar
strength. Fig. 11 shows the comparison between pillar strength (peak
average vertical stress across the mid-height of pillar) and peak major
principal stress at the core of pillars with different W/H ratios. It can be
observed that the peak major principal stress at the core is always
higher than the overall pillar strength because the confinement and
thereby strength at the core is higher than that at the walls of the pillar.
As the width to height ratio of pillars increases, the difference between
the pillar strength and peak stress at the core becomes more pro-
nounced. It is the result of higher induced confinements inside wide
pillars which involves higher potential for strain hardening behavior at

Fig. 8. Development of plastic zone in a pillar model

with W/H = 1 at the Elliot Lake mine and photo of a
crushed pillar, courtesy of Mr. C. Pritchard.

77
 H. Rafiei Renani, C.D. Martin Tunnelling and Underground Space Technology 74 (2018) 71–81

(a) (b) Fig. 9. Pillar behavior and predicted strength at (a) the
Elliot Lake mine and (b) the Selebi-Phikwe mine.
300 Stable Unstable Failed 100 Stable Unstable Failed

250
80 von Kimmelmann et al. (1984)

Pillar stress (MPa)

Pillar stress (MPa)

CWFS model
200
60
150 Hedley & Grant (1972)
CWFS model
40
100

20
50

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
W/H W/H

the core.
The numerical analysis of pillars at the Elliot Lake and Selebi-
Phikwe mines was carried out under plane strain condition typical in
long pillars. However, the geometry and loading condition of pillars in
mining operations can be more complex. It is also instructive to com-
pare the results of the CWFS models with other well-established ma-
terial models. In order to examine the effect of complex geometries and
different material models, a three-dimensional model of a typical block
caving pillar at a depth of 1 km was analyzed. The horizontal cross
section of the pillar is a parallelogram with a width of 10 m and length
of 32 m at an angle of 60°. The access drifts are 4.3 m wide and 3.8 m
high with horseshoe vertical cross sections (Fig. 12a).
The CWFS model along with perfectly plastic and strain softening
models were used for the three-dimensional analysis. For this example,
material parameters representative of rock mass at the Elliot Lake mine
were determined for each model (Table 3). The parameters for the
CWFS model are the same as those given in Table 2 and used in pre-
vious analyses. Existing empirical guidelines and equations were used
to estimate the parameters of the perfectly plastic and strain softening
models based on the available laboratory and in situ data presented in
Section 5.1. Parameters for the perfectly plastic model were obtained
using the latest guidelines by Hoek et al. (2002) as implemented in the
RocLab software (RocScience Inc., 2007). For the strain softening
model, similar guidelines were used to estimate initial and residual
strength parameters. Initial strength parameters are identical to those
obtained for the perfectly plastic model based on rock mass GSI of 80.
Residual strength parameters can be estimated using a residual value of
GSI. According to the empirical equation proposed by Cai et al. (2007),
residual GSI of 27 was obtained for the rock mass which was used to
estimate residual strength parameters. More details on empirical
guidelines for estimating rock mass strength are given by Hoek et al.
(2002) and Cai et al. (2007).
Stress-strain curves for the three-dimensional pillar using different
models are shown in Fig. 12b. For comparison, pillar strength based on
the empirical Lunder and Pakalnis formula (Lunder and Pakalnis, 1997)
and the stress-strain curve from the previous two-dimensional analysis
for the closest W/H ratio of 2.5 are also shown. The CWFS model gives a
pillar strength value of 181 MPa which is in reasonable agreement with
the empirical estimate and results of two-dimensional plane strain
analysis. Perfectly plastic and strain softening models, on the other
hand, show no finite strength even after 800 MPa. This is clearly un-
realistic and inconsistent with empirical evidence. Note that this is due
to the combined effect of the implemented models as well as associated
parameters. Obviously, using a different set of parameters would lead to
different results. The guidelines used in this example for determining
parameters of the perfectly plastic and strain softening models, how-
ever, represent the most recent empirical developments in estimating
rock mass strength. This example illustrates that well-established
models and frequently-used empirical guidelines for estimating rock
mass strength may grossly overestimate the strength of wide pillars.
This highlights the importance of using a representative model in nu-
merical analysis of hard rock pillars.
Fig. 13 shows the development of plastic zone and vertical stress
across a horizontal cross section at the mid-height of the pillar using the

(a) (b) Fig. 10. Average vertical stress-strain curves (a) recorded
in situ for coal pillars, modified after Wagner (1974) and
20 W/H = 2 60 W/H = 2 (b) simulated for hard rock pillars at the Selebi-Phikwe
mine.
W/H = 1 W/H = 1
50
15
Pillar stress (MPa)

Pillar stress (MPa)

40

10 30

20
5
10

0 0
0 0.5 1 1.5 0 0.05 0.1 0.15 0.2 0.25
Pillar strain (%) Pillar strain (%)

78
H. Rafiei Renani, C.D. Martin Tunnelling and Underground Space Technology 74 (2018) 71–81

(a) (b) Fig. 11. Overall pillar strength and peak major prin-
cipal stress at the core of pillars at (a) the Elliot Lake
700 Pillar strength 250 Pillar strength mine and (b) the Selebi-Phikwe mine.

600 Peak stress at the core Peak stress at the core

200
500
Stress (MPa)

Stress (MPa)
400 150

300 100
200
50
100

0 0
0.5 1 1.5 2 2.5 3 0.5 1 1.5 2 2.5 3
W/H W/H

(a) (b) Fig. 12. (a) Three-dimensional mesh of the block caving
pillar and (b) stress-strain curves using different models.
800
700
Perfectly plastic
600
Pillar stress (MPa)

Strain softening
500
CWFS
400
CWFS (2D)
300
Lunder & Pakalnis
200 (1997)
100
0
0 0.5 1 1.5 2
Pillar strain (%)

Table 3 friction with increasing damage. The behavior of pillars at the Elliot
Parameters and results of different models used in the three-dimensional analysis. Lake and Selebi-Phikwe mines was used to evaluate the merits of the
proposed modeling approach for hard rock pillars.
Model cini (MPa) cult (MPa) φini (deg) φult (deg) Pillar strength
(MPa) In both cases, the numerical models were able to successfully se-
parate unstable/failed pillars from stable pillars. The numerical models
CWFS 38.0 3.8 0 35 181 also captured the mechanism of progressive failure observed in hard
Perfectly plastic 11.4 11.4 56 56 N/A rock pillars. The overall pillar stress-strain curves indicated that while
Strain softening 11.4 3.5 56 42 N/A
pillars with W/H < 2 exhibit strain softening behavior at relatively low
strains, those with W/H > 2 show strain hardening behavior and con-
CWFS model. As expected, failure starts at the corners and propagates tinue to carry higher loads up to much higher strain levels. Monitoring
towards the center of pillar. Before reaching average pillar stress of the stress path at the core of pillars showed that for pillars with W/
135 MPa, all elements across the mid-height of the pillar have reached H ≤ 1 peak major principal stress at the core drops after reaching the
the initial cohesive strength envelope (Fig. 7). Hence, elements near the initial cohesive strength envelope while it continues to increase and
boundary of the pillar experience strain softening due to low confine- follow along the ultimate frictional strength envelope for pillars with
ment while central elements experience strain hardening due to ele- W/H ≥ 2.
vated confining stresses (Fig. 5b). Inspection of vertical stresses also Finally, a typical block caving pillar was analyzed using the
confirms the transfer of load from the boundary to the core as loading common perfectly plastic and strain softening models and empirical
continues. The observed process of load transfer is in keeping with in guidelines for estimating rock mass strength. It was shown that such
situ measurements of Wagner (1974) and numerical results of Jaiswal conventional models with empirically estimated rock mass strength
and Shrivasta (2009). parameters can lead to significant overestimation of strength and un-
safe design of wide hard rock pillars. Further work is required to clearly
identify the scope and limits of existing models and guidelines in terms
8. Conclusions
of pillar depth, shape, and rock mass characteristics.
The effect of size on pillar strength was quantified using empirical
evidence while numerical analysis was carried out to explore the effect Acknowledgement
of shape and induced confinement on pillar strength. An enhanced
CWFS model was implemented in numerical analysis to explore the pre- Swedish Nuclear Fuel and Waste Management Co. (SKB) Sweden,
and post-peak behavior of hard rock pillars. Equations were proposed to the Canadian Nuclear Waste Management Organization (NWMO) and
describe the gradual degradation of cohesion and mobilization of the Natural Sciences and Engineering Research Council of Canada

79
H. Rafiei Renani, C.D. Martin
Pillar stress of 60 MPa
Pillar stress of 75 MPa
Pillar stress of 90 MPa
Pillar stress of 135 MPa
Pillar stress of 181 MPa
provided funding for this work.
References
Abou-Chakra Guery, A., Cormery, F., Shao, J.F., Kondo, D., 2008. A micromechanical
model of elastoplastic and damage behavior of a cohesive geomaterial. International
Journal of Solids and Structures 45, 1406–1429.
Adhikary, D.P., Shen, B., Duncan Fama, M.E., 2002. A study of highwall mining panel
stability. International Journal of Rock Mechanics and Mining Sciences 39, 643–659.
ASTM, 2004. Designation: D 7012–04: Standard Test Method for Compressive Strength
and Elastic Moduli of Intact Rock Core Specimens Under Varying States of Stress and
Temperatures. American Society for Testing and Materials, Philadelphia, PA, USA.
Barton, N., Choubey, V., 1977. The shear strength of rock joints in theory and practice.
Rock Mechanics, Vienna: Springer. vol. 10, pp. 1–54.
Bieniawski, Z.T., 1968. The effect of specimen size on compressive strength of coal.
International Journal of Rock Mechanics and Mining Sciences 5, 325–335.
80
Tunnelling and Underground Space Technology 74 (2018) 71–81
Fig. 13. Development of plastic zone and vertical
stress across the mid-height of the block caving pillar
using the CWFS model.
Bieniawski, Z.T., 1992. A method revisited: coal pillar strength formula based on field
investigations. In: Proceedings of the Workshop on the Coal Pillar Mechanics and
Design, Bureau of Mines, pp. 158–165.
Bui, T.A., Wong, H., Deleruyelle, F., Xie, L.Z., Tran, D.T., 2017. A thermodynamically
consistent model accounting for viscoplastic creep and anisotropic damage in un-
saturated rocks. International Journal of Solids and Structures 117, 26–38.
Cai, M., Kaiser, P.K., Tasakab, Y., Minamic, M., 2007. Determination of residual strength
parameters of jointed rock masses using the GSI system. International Journal of Rock
Mechanics and Mining Sciences 44, 247–265.
Coates, D.F., Gyenge, M., 1981. Incremental Design in Rock Mechanics. Monograph 880
Canadian Government Publishing Centre, Ottawa.
Diederichs, M.S., 2007. The 2003 geotechnical colloquium: mechanistic interpretation
and practical application of damage and spalling prediction criteria for deep tun-
nelling. Canadian Geotechnical Journal 44, 1082–1116.
Dolinar, D.R., Esterhuizen, G.S., 2007. Evaluation of the effect of length on the strength of
slender pillars in limestone mines using numerical modeling. In: Proceedings of the
26th International Conference on Ground Control in Mining, Morgantown, WV, p.
H. Rafiei Renani, C.D. Martin
304–13.
Duncan Fama, M.E., Trueman, R., Craig, M.S., 1995. Two- and three-dimensional elasto-
plastic analysis for coal pillar design and its application to highwall mining.
International Journal of Rock Mechanics and Mining Sciences & Geomechanics
Abstracts 32, 215–225.
Edelbro, C., 2009. Numerical modelling of observed fallouts in hard rock masses using an
instantaneous cohesion-softening friction-hardening model. Tunnelling and
Underground Space Technology 24, 398–409.
Elmo, D., Stead, D., 2010. An integrated numerical modelling discrete fracture network
approach applied to the characterization of rock mass strength of naturally fractured
pillars. Rock Mechanics and Rock Engineering 43, 3–19.
Griffith, A.A., 1920. The phenomena of flow and rupture in solids. The Philosophical
Transactions of the Royal Society London (Series A) 221, 163–198.
Griffith, A.A., 1924. Theory of rupture. In: Walter, Drukkerij J., Jr. (Ed.), Proceedings of
the 1st International Congress of Applied Mechanics. Delft: Tech. Boekhandel en, p.
55–63.
Hajiabdolmajid, V., Kaiser, P.K., Martin, C.D., 2002. Modelling brittle failure of rock.
International Journal of Rock Mechanics and Mining Sciences 39, 731–741.
Hawkes, I., Mellor, M., 1970. Uniaxial testing in rock mechanics laboratories. Engineering
Geology 4, 177–285.
Hedley, D.G.F., Grant, F., 1972. Stope-and-pillar design for the Elliot Lake Uranium
Mines. Bulletin of Canadian Institute of Mining and Metallurgy 65, 37–44.
Hoek, E., Brown, E.T., 1980. Empirical strength criterion for rock masses. Journal of the
Geotechnical Engineering Division 106, 1013–1035.
Hoek, E., Brown, E.T., 1997. Practical estimates of rock mass strength. International
Journal of Rock Mechanics and Mining Sciences 34, 1165–1186.
Hoek, E., Carranza-Torres, C.T., Corkum, B., 2002. Hoek-Brown failure criterion-2002
edition. Proceedings of the fifth North American rock mechanics symposium,
Toronto, Canada 1, 267–273.
Iannacchione, A.T., 1989. Numerical simulation of coal pillar loading with the aid of a
strain-softening finite difference model. In: Proceedings of the 30th U.S. Symposium,
Morgantown, WV, p. 775–782.
Itasca Inc., 2009. FLAC3D, Fast Lagrangian Analysis of Continua in 3 Dimensions. Version
4.00. Minneapolis.
Jaiswal, A., Shrivasta, B.K., 2009. Numerical simulation of coal pillar strength.
International Journal of Rock Mechanics and Mining Sciences 46, 779–788.
Kaiser, P.K., Kim, B., Bewick, R.P., Valley, B., 2011. Rock mass strength at depth and
implications for pillar design. Mining Technology 120, 170–179.
Krauland, N., Soder, P.E., 1987. Determining pillar strength from pillar failure observa-
tions. Engineering and Mining Journal 8, 34–40.
Lunder, P.J., Pakalnis, R., 1997. Determination of the strength of hard- rock mine pillars.
Bulletin of Canadian Institute of Mining and Metallurgy 90, 51–55.
Martin, C.D., 1997. Seventeenth Canadian geotechnical colloquium: the effect of cohesion
loss and stress path on brittle rock strength. Canadian Geotechnical Journal 34,
698–725.
Martin, C.D., Chandler, N.A., 1994. The progressive fracture of Lac du Bonnet granite.
International Journal of Rock Mechanics and Mining Sciences and Geomechanics
Abstracts 31, 643–659.
Martin, C.D., Maybee, W.G., 2000. The strength of hard-rock pillars. International Journal
of Rock Mechanics and Mining Sciences 37, 1239–1246.
Martin, C.D., Lu, Y., Lan, H., 2012. Scale effects in a synthetic rock mass. In: Harmonising
Rock Engineering and the Environment. Taylor & Francis Group, London, pp.
257–258.
McClintock, F.A., Walsh, J.B., 1962. Friction on Griffith cracks in rock under pressure. In:
Proceedings of the 4th US national Congress on Applied Mechanics, New York, p.
81
Tunnelling and Underground Space Technology 74 (2018) 71–81
1015–1021.
Mortazavi, A., Hassani, F.P., Shabani, M., 2009. A numerical investigation of rock pillar
failure mechanism in underground openings. Computers and Geotechnics 36,
691–697.
Murali Mohan, G., Sheorey, P.R., Kushwaha, A., 2001. Numerical estimation of pillar
strength in coal mines. International Journal of Rock Mechanics and Mining Sciences
38, 1185–1192.
Nareetsile, S.M., 1998. The modified creeping cone mining method. M.Sc. thesis.
Department of Civil & Environmental Engineering, Univ. of Alberta, Canada.
Potvin, Y., Hudyma, M.R., Miller, H.D.S., 1989. Design guidelines for open stope support.
Bulletin of Canadian Institute of Mining and Metallurgy 82, 53–62.
Pratt, H.R., Black, A.D., Brown, W.S., Brace, W.F., 1972. The effect of specimen size on
the mechanical properties of unjointed diorite. International Journal of Rock
Mechanics and Mining Sciences 9, 513–529.
Pritchard, C.J., Hedley, D.G.F., 1993. Progressive pillar failure and rock bursting at
Denison Mine. In: Proceedings of the 3rd International Symposium on Rock bursts
and Seismicity in Mines, Kingston, p. 111–116.
RocScience Inc., 2007. RocLab Version 1.032, Rock mass strength analysis using the
Hoek-Brown failure criterion. Toronto, Canada.
Salamon, M.D.G., Munro, A.H., 1967. A study of the strength of coal pillars. Journal of the
South African Institute of Mining and Metallurgy 68, 55–67.
Sjoberg, J., 1992. Failure modes and pillar behaviour in the Zinkgruvan mine. In:
Tillerson, J.A., Wawersik, W.R. (Eds.), Proceedings of 33rd U.S. Rock Mechanics
Symposium, Rotterdam: A.A. Balkema, pp. 491–500.
Swan, G., 1985. Strength distribution and potential for multiple pillar collapse. Divisional
Report MRL 85–127 (TR), CANMET, Energy Mines and Resources, Canada.
von Kimmelmann, M.R., Hyde, B., Madgwick, R.J., 1984. The use of computer applica-
tions at BCL Limited in planning pillar extraction and design of mining layouts. In:
Brown, E.T., Hudson, J.A. (Eds.), Proceedings of ISRM Symposium: Design and
Performance of Underground Excavations, London: British Geotechnical Society, pp.
53–63.
Wagner, H., 1974. Determination of the complete load-deformation characteristics of coal
pillars. In: Proceedings of 3rd ISRM Conference, Denver. Colorado, p. 1076–1081.
Walton, G., Diederichs, M.S., 2015. A mine shaft case study on the accurate prediction of
yield and displacements in stressed ground using lab-derived material properties.
Tunnelling and Underground Space Technology 49, 98–113.
Wang, H., Poulsen, B.A., Shen, B., Zhue, S., Jiang, Y., 2011. The influence of roadway
backfill on the coal pillar strength by numerical investigation. International Journal
of Rock Mechanics and Mining Sciences 48 (443), 450.
Whyatt, J.K., Board, M.P., 1991. Numerical Exploration of Shear-Fracture Related Rock
Bursts Using a Strain-Softening Constitutive Law, Report of Investigations RI 9350.
Bureau of Mines, Washington, U.S.
Zhou, X., 2004. Analysis of the localization of deformation and the complete stress–strain
relation for mesoscopic heterogeneous brittle rock under dynamic uniaxial tensile
loading. International Journal of Solids and Structures 41, 1725–1738.
Zhou, X.P., Yang, H.Q., 2007. Micromechanical modeling of dynamic compressive re-
sponses of mesoscopic heterogenous brittle rock. Theoretical and Applied Fracture
Mechanics 48, 1–20.
Zhou, X., Ha, Q., Zhang, Y., Zhu, K., 2004. Analysis of deformation localization and the
complete stress–strain relation for brittle rock subjected to dynamic compressive
loads. International Journal of Rock Mechanics and Mining Sciences 41, 311–319.
Zhou, Q.Z., Kondo, D., Shao, J.F., 2008. Micromechanical analysis of coupling between
anisotropic damage and friction in quasi brittle materials: role of the homogenization
scheme. International Journal of Solids and Structures 45, 1385–1405.