International Journal of Mining Science and Technology 27 (2017) 591–597

Contents lists available at ScienceDirect

International Journal of Mining Science and Technology

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

Numerical investigation into pillar failure induced by time-dependent

skin degradation
Sainoki Atsushi ⇑, Mitri Hani S.
Department of Mining and Materials Engineering, McGill University, Montreal H3A0E8, Canada

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

Article history: This paper focuses on the instability mechanism of an isolated pillar, caused by time-dependent skin
Received 18 October 2016 degradation and strength heterogeneity. The time-dependent skin degradation is simulated with a
Received in revised form 29 January 2017 non-linear rheological model capable of simulating tertiary creep, whereby two different pillar failure
Accepted 1 March 2017
cases are investigated. The first case is of an isolated pillar in a deep hard rock underground mine and
Available online 12 May 2017
subjected to high stresses. The results show that pillar degradation is limited to the regions near the sur-
face or the skin until two months after ore extraction. Afterwards degradation starts to extend deeper
Keywords:
into the pillar, eventually leaving a highly-stressed pillar core due to stress transfer from the failed skin.
Pillar stability
Underground mine
Rockburst potential indices show that the risk increases exponentially at the core as time goes by. It is
Skin degradation then demonstrated that the progressive skin degradation cannot be simulated with conventional
Rheological model strain-softening model assuming brittle failure. The parametric study with respect to the degree of
Time-dependent failure heterogeneity reveals that heterogeneity is key to the occurrence of progressive skin degradation. The
second case investigated in this study is pillar failure taking place in a very long period. Such failure
becomes significantly important when assessing the risk for ground subsidence caused by pillar collapse
in an abandoned mine. The analysis results demonstrate that the employed non-linear rheological model
can simulate gradual skin degradation taking place over several hundred years. The percentage of damage
zone volume within the pillar is merely 1% after a lapse of one days and increases to 50% after one hun-
dred years, indicating a high risk for pillar collapse in the long term. The vertical displacements within the
pillar also indicate the risk of subsidence. The proposed method is suitable for evaluating the risk of
ground surface subsidence above an abandoned mine.
Ó 2017 Published by Elsevier B.V. on behalf of China University of Mining & Technology. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction been proposed for the estimation of a factor of safety (FoS)

The stability of isolated pillars in underground mines is crucial
because pillar failure can lead to fatal accidents and might cause
the collapse of an extensive surrounding area [1–3]. In deep under-
ground hard rock mines, pillars can be the cause for rockburst
because of the high strain energy stored within the pillar and char-
acteristics of rockmass exhibiting extremely brittle behavior. On
the other hand, in relatively shallow underground mines, pillar col-
lapse that takes place after a lapse of a long period can lead to
ground subsidence, which might inflict severe damage to surface
structures [4–7]. Hence, for both deep and shallow underground
mines, the stability of isolated pillars is crucial and needs to be
evaluated and assessed in an appropriate manner.
A number of studies have been undertaken to estimate the sta-
bility of pillars [8]. In addition, various empirical equations have
⇑ Corresponding author.
E-mail address: atsushi.sainoki@mail.mcgill.ca (A. Sainoki).
[9–11]. Notwithstanding the significant efforts, pillars with a high
FoS occasionally fail and collapse, which would be attributed to
unknown uncertainties.
Two of such uncertainties are the strength heterogeneity of the
rockmass and the time-dependent strength degradation. Indeed,
the degradation of a pillar skin, which is represented by spalling,
scaling, or strength decay, has been studied by many researchers
[1,12]. The schematic illustration of spalling that takes place in a
deep hard rock mine is delineated in Fig. 1. As can be seen, due
to the spalling, the side wall of the pillar loses its load-bearing
capacity, consequently inducing additional loads to the core of
the pillar. When the region undergoing strength degradation is suf-
ficiently large, the FOS of the pillar can decrease to a critical value
due to the increase in stress acting in the pillar and the decrease in
the cross section of an intact region. In previous studies, the typical
technique to simulate the degradation of a pillar side wall is to
employ an exponential function to estimate the time-dependent
strength decay of rockmass or to decrease rockmass strength in a

http://dx.doi.org/10.1016/j.ijmst.2017.05.002
2095-2686/Ó 2017 Published by Elsevier B.V. on behalf of China University of Mining & Technology.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
592 A. Sainoki, H.S. Mitri / International Journal of Mining Science and Technology 27 (2017) 591–597

The severity, r⁄, is given using deviatoric stress as follows:

Fig. 1. Schematic illustration of skin degradation of an isolated pillar due to
spalling.
loosened region under low confining stress near a pillar surface.
Importantly, both simulation techniques are based on field obser-
vations or unverified assumptions.
The present study focuses on the time-dependent skin degrada-
tion of an isolated pillar due to creep behavior that eventually
leads to critical failure of the rockmass whilst considering the
strength heterogeneity with Weibull’s distribution [13]. The mech-
anism and process of stress transfer from loosened near-surface
regions to the intact core area of the pillar are investigated with
non-linear rheological constitutive model with input parameters
derived from laboratory experiments. The comparison of the rheo-
logical model with the classical Mohr-Coulomb model considering
brittle failure (strain-softening) is made in order to demonstrate
the importance of considering the time-dependent degradation of
a pillar skin. First, the aforementioned detailed mechanism of pillar
failure is examined, using boundary conditions and rockmass prop-
erties of an isolated pillar in a deep underground hard rock mine.
Subsequently, post-mining pillar failure that takes place due to
long-term time-dependent pillar skin degradation is analyzed,
applying different rockmass properties and boundary conditions
that correspond to those of a shallow underground mine.
2. Methodology
r1  r3
r
¼ ð2Þ
rmax  r3
where r1 and r3 are the maximum and minimum principal stresses,
respectively; and rmax the maximum stress calculated from a fail-
ure criterion. Note that compression is positive quantity in the
equations. The severity for tensile failure is calculated as follows:
jr3 j
r
¼ ð3Þ
rt
where rt is the tensile strength of rock. The larger one of the sever-
ities determined by Eqs. (2) and (3) is adopted as r⁄ and substituted
into Eq. (1). As the Poisson’s ratio, v, is expected to increase with the
progression of rock degradation, Okubo et al. use the following
equation to relate v with k⁄ [17].
0:5  mo
m ¼ 0:5  ð4Þ
k

where vo is an initial Poisson’s ratio. The modulus of elasticity is cal-
culated with k⁄ as follows:
Eo
E¼ ð5Þ
k

where Eo is an initial value of modulus of elasticity. Regarding the
failure criterion to calculate the severity, the classical Mohr-
Coulomb failure criterion is employed because of its simplicity. As
this study does not aim at calibrating analysis conditions and
mechanical properties of rock based on actual field measurements,
the use of the Mohr-Coulomb failure criterion is deemed sufficient.
It is noteworthy that although Eqs. (1)–(5) indicate that compliance
might increase even under initial stress condition before mining
activity starts, it has been confirmed that the increase in compliance
is extremely small even when several hundred million years passed
if there is no mining activity.

2.2. Generalized simulation procedure

2.1. Constitutive model

When the non-linear rheological model is employed, an itera-

As indicated by Scholz, the creep behavior of rockmass is essen-
tive analysis is carried out, following the generalized procedure
tially the progressive development of microcracks, which pertain
illustrated in Fig. 2. As can be seen in Fig. 2, after the numerical
to the degradation of rockmass [14]. Thus, the present study
model construction, in-situ stresses are applied, if necessary. This
employs a creep model. To date, a number of constitutive models
means that when the time dependent behavior of in-situ rockmass
have been proposed and employed to simulate the creep behavior
in an underground mine is examined, in-situ stress conditions
of rockmass [15]. Importantly, few of these models are capable of
simulating tertiary creep that causes the rockmass to fail. In order
to take tertiary creep into consideration, the present study employs
the non-linear rheological model proposed by Okubo et al. [16].
The advantage of this constitutive model over other models con-
sidering tertiary creep is that the analytical solution of creep life-
time, i.e., time to final rupture, is provided [15,16].
The non-linear rheological model is expressed as follows.
  m
dk
1 m nmþ1
m
n
¼ ðk Þ ðr Þ ð1Þ
dt t0 n þ 1

where k⁄ is the compliance normalized by its initial value; r⁄ called

as severity; n and m the degree of time dependency and ductility of
rock, respectively; t the time; and t0 the time that it takes for the
axial stress applied to a rock specimen to reach its peak axial value
during an unconfined compression test. In the present study, t0 is
determined whilst assuming a uniaxial test with a typical axial
strain rate. The constants, n and m, for granite are taken from a
reported database [16]. The constant, n, is derived from uniaxial
compressive strengths obtained from unconfined compression tests
under different axial strain rates, while the constant, m, is estimated
from the shape of a post-peak behavior curve [16]. Fig. 2. Procedure of iterative analysis.
 A. Sainoki, H.S. Mitri / International Journal of Mining Science and Technology 27 (2017) 591–597
need to be simulated. On the other hand, when the time-
dependent behavior of a rock specimen during a laboratory test
is simulated, there is no need to apply such in-situ conditions to
the model. As the present study applies the rheological model to
two different numerical models representing a room-and-pillar
mine and an unconfined compression test under a constant strain
rate, the difference in the initial boundary conditions between the
two numerical models is described for clarification. At the third
step, the conditions that cause creep behavior are applied. In the
case of room-and-pillar mining, the condition is ore extraction
and the generation of an isolated pillar, while in the case of labora-
tory test, axial strain that increases with time, is applied to the top
boundary of the rock specimen. Subsequently, a static analysis is
carried out, and if the elapsed time, t, has not yet reached the
pre-determined tmax, the maximum and minimum stresses are
computed for each zone. And then, the mechanical properties are
updated according to Eqs. (1)–(5). The time increment used for
the room-and-pillar mine model is calculated with the following
equation.
dt ¼ 1:1n
where n is the number of iterations. On the other hand, a smaller
time step is used for the unconfined compression test model as
follows.
dt ¼ 1:01n
In the case of the compression test model, vertical displacement
applied to the top boundary is computed from the time increment,
assuming a constant strain rate of 1.0  105.
2.3. Numerical model description
Fig. 3 shows the 3D numerical model analyzed in the present
study, which is constructed with FLAC3D [18]. The symmetry of
the pillar is taken into account, that is, the pillar in Fig. 3 represents
one-fourth of the actual pillar with the width of 10 m. The green-
colored region is extracted during analysis, and the dimension is
determined so that the FOS of the pillar falls between 1.0 and
1.4. The FoS is evaluated with empirical equations [9,10]. The sta-
bility of the pillars with FoS in the range is uncertain and assumed
to be affected by several factors such as the time-dependent degra-
dation [8]. Thus, it is worth examining.
Pre-mining stress state is based on the equations proposed by
Diederichs [19]. The mining depth is assumed to be 1500 m. The
pre-mining stresses computed from the equations are applied to
each zone in the model in the form of initial stresses. Note that
the maximum horizontal stress is applied in the x-direction. Fixed
boundary conditions are employed, that is, the displacements on
Fig. 3. Numerical model analyzed.
593
the model outer boundaries are fixed in the direction perpendicu-
lar to the boundaries.
2.4. Mechanical properties for a deep hard rock mine
As described, the pillar failure mechanism in a deep hard rock
mine is first investigated. Granite is assumed as the main rockmass
of the pillar. In order to convert the mechanical properties obtained
from laboratory experiments to those of the rockmass, RMR = 92 is
assumed, which is the upper limit of rockmass encountered in
Canadian Shield [8]. Herein, RMR is the rockmass rating system
proposed by Bieniawski [20]. The mechanical properties of intact
granite are derived from another study [21]. Using the RMR value
of 92, the deformation modulus, E, in Table 1 is obtained with the
equation [22]. The uniaxial compressive strength is calculated with
the Hoek-Brown failure criterion while substituting r3 = 0. The
tensile strength is assumed to be one-tenth of the uniaxial com-
pressive strength [23]. For the constants in Eq. (1), n = 51 and
m = 51 are derived for granite [16].
ð6Þ 2.5. Rockmass strength heterogeneity
The inherent strength heterogeneity of the rockmass is consid-
ered with the Weibull’s distribution, which is expressed as:
 a1   a 
a u u
ð7Þ f ðuÞ ¼ exp  ð8Þ
u0 u0 u0
where u and u0 are the mechanical property and the scale parame-
ter related to the average value of the mechanical property, respec-
tively; and a the degree of heterogeneity. In a base model, a = 5 is
taken. Fig. 4 shows the heterogeneity of UCS for the case.
2.6. Determination of upper limit of k⁄
Although the rheological model assumes that k⁄ can increase
infinitely, its upper limit needs to be determined to ensure compu-
tational stability. To do so, a uniaxial test is simulated with a cylin-
drical model under a constant strain rate. When the upper limit is
low, the strain softening during post-peak behavior is not suffi-
ciently simulated. During the simulation, displacements are
applied to the top boundary at the third step in Fig. 2. Analyses
are performed while changing the upper limit from 50 to 1000. It
is then found that when the upper limit is 1000, brittle behavior
is adequately replicated. It should be noted, however, that when
the brittle failure takes place, k⁄ does not reach 1000 in all the
zones because failure takes place locally. Thus, it is not appropriate
to apply the upper limit to the pillar model, directly. In light of the
result, the average value of k⁄ in the cylindrical model at failure is
computed. The computed average is near 285. Thus, after rounding
up, the upper limit is set to 300 for the pillar model.
3. Results
3.1. Time-dependent skin degradation
Fig. 5 shows the change in the maximum compressive stress
within the pillar with time. It is found that the pillar surface can
carry stress immediately after extraction of the surrounding rock
representing the mine. Low stress regions are limited to the zones
with particularly low initial strength. There is no noticeable change
in the stress state until 33 days after the extraction. However, after
72 days have passed, low stress regions extend to almost the entire
surface, indicating that the rockmass near the surface lost its load-
bearing capacity due to the time-dependent degradation. After
187 days, the degradation of rockmass extends deeper into the
594 A. Sainoki, H.S. Mitri / International Journal of Mining Science and Technology 27 (2017) 591–597
Table 1
Mechanical properties of granite for rockmass.
E (GPa) m rc (MPa)
59 0.26 116
Fig. 4. Heterogeneity of UCS simulated with Weibull’s distribution (a = 5).
pillar, and the pillar core carries extremely high stresses due to
stress transfer from the failed regions. This indicates that the pillar
becomes extremely burst-prone.
To investigate the burst proneness of the pillar, brittle shear
ratio (BSR) proposed by Castro et al. and the burst prone index
(BPI) proposed by Mitri et al. are computed at the center of the pil-
lar [24]. As can be seen from Fig. 6, both indices do not show a
noticeable increase until 72 days after extraction [25]. After that,
BPI and BSRs increase exponentially, indicating that failure propa-
gation becomes uncontrolled. Eventually, after 226 days have
passed, BPI and BSR increase to 222% and 1.33, respectively. Both
values indicate extremely high potential for rockburst occurrence.
3.2. Comparison with strain-softening model
To demonstrate the difference between the non-linear rheolog-
ical model adopted in this study and the conventional strain-
softening model, a comparison between the two models is made.
The strain-softening model employs the conventional Mohr–Cou-
lomb criterion with the consideration to strain-softening behavior,
that is, when failure takes place, cohesion is decreased to zero. A
static analysis is conducted with the same model as shown in
Fig. 4.
Fig. 7 shows the maximum compressive stress obtained from
the elasto-plastic analysis with the strain-softening model. As
can be seen, the region with low stresses occurs only near the pillar
surface. The much higher stresses in the pillar core obtained by the
non-linear rheological model (Fig. 5d) are not reproduced by the
strain softening model. This implies that the use of the conven-
tional strain-softening model would underestimate the possibility
of failure when there is strength heterogeneity. On the other hand,
when the rockmass is completely homogeneous, there will be no
large difference between the two models.
Fig. 5. Maximum compressive stress in the pillar.
C (MPa) / (°) rT (MPa) Density (kg/m3)
14.8 63 11.6 2600
Fig. 6. Burst proneness at the center of the pillar.
Fig. 7. Maximum compressive stress obtained from the conventional elasto-plastic
analysis using Mohr-Coulomb strain-softening model.
3.3. Effect of degree of strength heterogeneity
In addition to the model shown in Fig. 4, another model is con-
structed with a = 6 in Eq. (8). The model is shown in Fig. 8. The lar-
ger a represents a more homogeneous model. The same procedure
is undertaken. Interestingly, the analysis reveals that the progres-
sive skin degradation does not occur even after 7 years when
a = 6. It should be noted that, in terms of average UCS in the pillar,
there is no large difference between the two models. Indeed, the
average UCS is 107 and 108 MPa for the models with a = 5 and 6,
respectively. It is thus unlikely that the intensive degradation of
the rockmass shown in Fig. 5 is influenced by slight difference in
the average UCS.
Considering these results, with respect to each zone in the pil-
lar, the minimum UCS in adjacent zones is investigated. Fig. 9
shows the results. For instance, when a = 5, the number of zones
that have adjacent zones with UCS between 30 and 50 MPa is
approximately 240. Fig. 9 displays the difference between the
two models. When a is low, the possibility that zones with low
UCS are present in the vicinity of high UCS zone is clearly higher.
 A. Sainoki, H.S. Mitri / International Journal of Mining Science and Technology 27 (2017) 591–597
Fig. 8. Heterogeneity of UCS simulated with the Weibull’s distribution (a = 6).
Fig. 9. Relation between minimum UCS of adjacent zones and the total number of
zones with the adjacent zones within the pillar.
From Fig. 9, it can be deduced that the intensive skin degradation
shown in Fig. 5d occurs due to the effect of failure initiated by
zones with low UCS, i.e., stress transferred from a failed zone
causes failure in adjacent zones, which eventually escalates into
a chain effect. Zones with extremely low UCS exist even when
a = 6. However, in order for failure to propagate to adjacent zones,
zones with low UCS must exist in the vicinity of the failed zone.
When a = 5, it is assumed that the conditions are satisfied, so that
the failure becomes uncontrolled after sufficient time has elapsed.
Importantly, the chain effect cannot be adequately simulated with
the strain-softening model as shown in Fig. 7 because the non-
linear rheological model simulates strength degradation with ter-
tiary creep even if the maximum stress is less than the maximum
strength.
4. Numerical modeling of pillar collapse in a shallow
underground mine
As discussed in the introduction, a better understanding of pil-
lar collapse in a shallow abandoned mine is important to prevent
ground subsidence and/or take necessary measures. This section
provides the numerical simulation of pillar collapse induced by
long-term time-dependent pillar skin degradation. The primary
goal is to demonstrate that the proposed method can equally be
used for the numerical simulation and risk evaluation of pillar col-
lapse as well as time-dependent ground subsidence.
4.1. Mechanical properties for a shallow abandoned mine
The numerical model used for a pillar in a shallow abandoned
mine is the same as that described in Section 2.3, while the
mechanical properties of rockmass are modified as shown in
Table 2. As shallow underground mines are not typically composed
595
of hard rock, the mechanical properties of an average quality rock-
mass suggested by Hoek are applied [26]. The properties are listed
in Table 2. The two parameters, n and m, are set to 10 and 1. In the
case of brittle, hard rock, n and m/n are set to 51 and 1, respec-
tively, as discussed in Section 2.4. The new values are adapted to
consider the behavior of average quality rock. The decrease in n
represents an increase in the degree of time-dependency, while
the decrease in m-to-n ratio represents an increase in ductility.
4.2. In-situ stress state
In order to simulate the stress state in a shallow underground
mine, the pillar is assumed to be located 70 m below the ground
surface, according to which overburden pressure is applied to the
top boundary of the model. As for the horizontal stresses, the max-
imum and minimum horizontal-to-vertical stress ratios are
assumed to be 1.5 and 1.0, respectively. The maximum ratio falls
within the range of in-situ stress state measured in the world,
while the minimum one is simply assumed to be equal to the over-
burden pressure [27].
4.3. Results
The same analysis procedure is undertaken as shown in Fig. 2.
The difference from the previous case study is in the number of
iterations. In this case, the iterative analysis is conducted until time
reaches three hundred years according to Eq. (6), so that it is pos-
sible to examine long-term time-dependent pillar skin degrada-
tion. Fig. 10 shows the analysis results in terms of increase in the
normalized compliance in Eq. (1). As can be seen, a zone with
k⁄ > 1.2 is colored in red and is considered to have undergone sig-
nificant degradation. This is due to the fact that when the normal-
ized compliance reaches 1.2, it does not take a long time for k⁄ to
reach the upper limit of 300. Thereafter, this means that the zone
has undergone significant, irreversible degradation leading to crit-
ical failure. This threshold is in accord with that used by Okubo
et al. [28]. As can be seen in Fig. 10a, skin degradation is limited
to the proximity of the pillar surface after one year from the ore
extraction. It is remarkable that the speed at which skin degrada-
tion extends deeper into the pillar is quite different from the result
presented in Section 3.1 for a hard rock pillar, where almost the
entire pillar has deteriorated and lost its load-bearing capacity
after a lapse of six months except for the pillar core. In the present
case, skin degradation evolves year after year. As shown in
Fig. 10a and b, the difference between the results after one year
and ten years is not significant. The depth of skin degradation in
the model after one year ranges from 0.5 to 1.0 m, whilst after
ten years, degradation reaches approximately 1.5 m. As can be seen
from Fig. 10c and d, skin degradation takes place even after a lapse
of more than one hundred years, eventually reaching 3 m. The
results indicate imminent pillar collapse long time after the mine
is abandoned.
As noted in Fig. 10, a zone with k⁄ > 1.2 is assumed to have
undergone severe degradation.
To quantitatively assess the degree of skin degradation that the
pillar undergoes, the percentage of damage zone volume within
the pillar is computed and plotted in Fig. 11. A zone with k⁄ > 1.2
is regarded as a damage zone that has undergone severe degrada-
tion. As can be seen from Fig. 11, the percentage is merely 1% at
one day after ore extraction. Although the percentage increases
to approximately 15% after a lapse of one year, it would not have
a large influence on the pillar stability, as inferred from Fig. 10a.
It is to be noted, however, that the value continues to increase
thereafter, reaching 50% and nearly 60% after a lapse of one hun-
dred years and three hundred years, respectively. Such severe
damage is expected to exert a significant influence on the stability
596 A. Sainoki, H.S. Mitri / International Journal of Mining Science and Technology 27 (2017) 591–597
Table 2
Rockmass mechanical properties for the shallow mine case study.
E (GPa) m rc (MPa)
9 0.25 13
Fig. 10. Normalized compliance that represents severity of damage (time-dependent degradation).
Fig. 11. Percentage of damage zone volume within the pillar.
of the pillar, ultimately leading to pillar collapse. Lastly, the risk for
ground subsidence is investigated based on pillar deformation in
the vertical direction. Fig. 12 shows displacement in the z-
direction in the pillar for models after one day and one hundred
years. It is found from Fig. 12 that the blue-colored region extends
larger after one hundred years, compared to the model after one
day, indicating that the deformation would contribute to ground
subsidence. However, the result shows that the displacement mag-
nitude is merely of the order of mm even after one hundred years,
which is clearly too small to affect ground surface deformation.
There would be two reasons why the displacement is still small
even after the pillar has experienced severe damage. The first
one is that the pillar core is still intact with high modulus of elas-
ticity, so that the pillar core sustains the load from the overburden,
preventing large deformation from taking place. Another reason is
that this study focuses on only pillar skin degradation. Accordingly,
the non-linear rheological model is applied to zones within the pil-
lar, meaning that the surrounding rockmass undergoes no deterio-
ration even after a lapse of one hundred years. Hence, it is clear
that to simulate ground subsidence correctly, the non-linear rheo-
logical model needs to be applied to the overlying strata as well.
Nevertheless, the obtained results adequately demonstrate the
Fig. 12. Vertical displacement (mm) within the pillar.
/ (°) rT (MPa) Density (kg/m3)
33 0.9 2600
potential use of the non-linear rheological model for the numerical
simulation and risk assessment of ground subsidence caused by
pillar collapse in an abandoned underground mine. Future study
employs this model to simulate ground subsidence caused by pillar
failure in a shallow abandoned mine while applying the non-linear
rheological model to the overlying strata.
5. Conclusions
In this paper, the time-dependent skin degradation of an iso-
lated pillar in a deep hard rock mine and a shallow underground
mine is simulated with a non-linear rheological model whilst con-
sidering the inherent strength heterogeneity of the pillar. The anal-
ysis results show that skin degradation extends deep into the pillar
when sufficient time has elapsed, resulting in highly stressed pillar
core due to stress transfer from the failed regions. Rockburst
potential indices calculated at the core show high risk for rock-
burst. A comparison with the conventional strain-softening model
demonstrates that the propagation of skin degradation into the pil-
lar cannot be replicated with the conventional strain-softening
model, emphasizing the importance of the time-dependent skin
degradation. It is shown that the frequent presence of adjacent
zones with low UCS is a key to the progressive skin failure because
it creates a chain effect of failure with stress transfer from the
failed zone to the adjacent zones. The simulation of pillar collapse
in a shallow underground mine shows long-term time-dependent
pillar skin degradation taking place over one hundred years. The
percentage of damage zone volume within the pillar is merely 1%
one day after the ore extraction, whereas the percentage increases
to more than 50% after one hundred years. The analysis result indi-
cates that the employed non-linear rheological model and pro-
posed method can be used for the numerical simulation and risk
assessment of ground subsidence caused by pillar collapse in aban-
doned underground mines.
Acknowledgment
This work is financially supported by the Natural Science and
Engineering Research Council of Canada (NSERC) in partnership
with Vale Ltd–Sudbury Operations, Canada, under the Collabora-
tive Research and Development Program. The authors are grateful
for their support.
References
[1] Cording EJ, Hashash YMA, Oh J. Analysis of pillar stability of mined gas storage
caverns in shale formations. Eng Geol 2015;184:71–80.
[2] Maleki H. Coal pillar mechanics of violent failure in U.S. Mines. Int J Min Sci
Technol 2017;27(3):387–92.
 A. Sainoki, H.S. Mitri / International Journal of Mining Science and Technology 27 (2017) 591–597
[3] Dehghan S, Shahriar K, Maarefvand P, Goshtasbi K. 3-D modeling of rock burst
in pillar No. 19 of Fetr6 chromite mine. 2013;20(2):231–6.
[4] Jeon B, Jeon S, Kim J, Kim T. Numerical evaluation of affecting parameters of
surface subsidence in abandoned mine areas. Geosyst Eng 2012;15
(4):299–304.
[5] Oh H, Lee S. Integration of ground subsidence hazard maps of abandoned coal
mines in Samcheok, Korea. Int J Coal Geol 2011;86(1):58–72.
[6] Salmi EF, Nazem M, Karakus M. The effect of rock mass gradual deterioration
on the mechanism of post-mining subsidence over shallow abandoned coal
mines. Int J Rock Mech Min Sci 2017;91:59–71.
[7] Son M, Moon H, Kim Y, Lee S. A study on the subsidence risk evaluation using
3-d rock mass collapse simulation for abandoned mines. Geosyst Eng 2016;20
(1):51–8.
[8] Martin CC, Maybee WG. The strength of hard-rock pillars. Int J Rock Mech Min
Sci 2000;37(8):1239–46.
[9] Potvin Y, Hudyma MR, Miller HDS. Design guidelines for open stope support.
CIM Bull 1989;82(926):53–62.
[10] Sjöberg J. Failure modes and pillar behaviour in the Zinkgruvan mine. In:
Proceedings of the 33rd U.S. rock mechanics symposium, Sante Fe, Rotterdam;
1992.
[11] Van der Merwe JN. New pillar strength formula for South African Coal. J S Afr
Inst Min Metall 2003;103(5):281–92.
[12] Napier JAL, Malan DF. Simulation of time-dependent crush pillar behaviour in
tabular platinum mines. J S Afr Inst Min Metall 2011;112(8):711–9.
[13] Weibull WA. Statistical distribution function of wide applicability. J Appl Mech
1950;13(2):293–7.
[14] Scholz CH. Mechanism of creep in brittle rock. J Geophys Res 1968;73
(10):3295–302.
597
[15] Barla G, Debernardi D, Sterpi D. Time-dependent modeling of tunnels in
squeezing conditions. Int J Geomech 2012;12(6):697–710.
[16] Okubo S, Fukui K. An analytical investigation of a variable-compliance-type
constitutive equation. Rock Mech Rock Eng 2006;39(3):233–53.
[17] Okubo S, Jin F. Simulation of rock behaviour around circular roadway by non-
linear rheological model. J Min Mater Process Inst Jpn 1993;109(3):209–14.
[18] Itasca. FLAC3D – fast Lagrangian analysis of continua (Version 4.0). U.S.A.:
Itasca Consulting Group Inc; 2009.
[19] Diederichs MS. Instability of hard rockmass: the role of tensile damage and
relaxation. Waterloo: University of Waterloo; 1999.
[20] Bieniawski ZT. Engineering rock mass classifications. New York: Wiley; 1989.
[21] Yun XY. Geomechanical behaviour of biaxially loaded rock. Montreal: McGill
University; 2008.
[22] Mitri HS, Edrissi R, Henning J. Finite element modelling of cable-bolted stopes
in hardrock ground mines. In: Proceedings of the SME annual meeting,
Albuquerque, New Mexico.
[23] Tesarik DR, Seymour JB, Yanske TR. Post-failure behaviour of two mine pillars
confined with backfill. Int J Rock Mech Min Sci 2003;40(2):221.
[24] Castro LAM, Bewick RP, Carter TG. An overview of numerical modelling applied
to deep mining. Innov Numer Modell Geomech 2012:393–414.
[25] Mitri HS, Tang B, Simon R. FE modelling of mining-induced energy release and
storage rates. J S Afr Inst Min Metall 1999;99(2):103–10.
[26] Hoek E. Practical rock engineering 2007.
[27] Brown ET, Hoek E. Trends in relationships between measured in-situ stresses
and depth. Int J Rock Mech Min Sci Geomech Abstr 1978;15:211–5.
[28] Okubo S, Fukui K, Zhang J. Three-dimensional time-dependent analysis of rock
by finite element method. Shigen-to-Sozai 1998;114:79–85.