You are on page 1of 12

International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics and Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Strainburst phenomena and numerical simulation of self-initiated brittle T


rock failure
Fuqiang Gaoa, Peter K. Kaiserb,∗, Doug Steadc, Erik Eberhardtd, Davide Elmoe
a
Mining& Designing Branch, China Coal Research Institute, State Key Laboratory of Coal Mining and Clean Utilization (China Coal Research Institute), Beijing, China
b
Bharti School of Engineering, Laurentian University, Sudbury, Ontario, Canada
c
Department of Earth Sciences, Simon Fraser University, Burnaby, BC, Canada
d
Geological Engineering, Department of Earth, Ocean and Atmospheric Sciences, University of British Columbia, Vancouver, BC, Canada
e
Institute of Mining Engineering, University of British Columbia, Vancouver, BC, Canada

A R T I C LE I N FO A B S T R A C T

Keywords: This paper focuses on ‘self-initiated’ strainbursts, and starts by providing an overview of these phenomena and
Rockburst their relevance to mining. This is used to set the framework for the numerical simulations presented in the
Self-initiated second part of the paper. The numerical simulations examine and address challenges in modelling strainbursts
Triggered related to the inherent difficulties in dynamic failure simulation. Existing methods either simulate dynamic
Strainbursts
failure under static or pseudo-static conditions while adopting some threshold factor (normally in terms of stress,
Dynamic disturbance
strain, or energy) as an indicator of rockburst potential, or incorporate an assumed dynamic disturbance to cause
Numerical simulation
Distinct element method rockburst damage. A distinct-element bonded block method is presented to simulate strainbursts. Instead of
Rock ejection triggering failure by an assumed dynamic disturbance, the adopted method simulates the generation of a seismic
event by self-initiated rock mass fracturing. This simulation of a self-initiated strainburst is interpreted using
patterns of cracking, displacement and strain fields, ground velocities, and calculated kinetic and dissipated slip
energies. The mechanisms of self-initiated strainbursting are successfully captured by the distinct-element
bonded block method.

1. Introduction design principles. One of the three flaws identified is that ground mo-
tion-related energy transfer mechanisms are assumed to constitute the
Considerable advances have been made over the last three decades main source of support loading. This of particular concern and is ad-
with respect to rockbursting and ground control in burst-prone ground, dressed in this paper. Kaiser6 demonstrated that ‘seismically triggered’
in large part due to the development of design methodologies, in- or ‘dynamically loaded’ strainbursts frequently constitute the primary
novative support components and support systems, numerical model- excavation damage mechanism and that the triggering seismic source
ling capabilities, and verifications by case studies. These advances have often contributes little to the severity of this rockburst damage; i.e., the
helped to greatly improve work place safety when mining at depth and damage is rarely dominated by energy or momentum transfer from a
in burst-prone ground. Nevertheless, there remain important challenges remote seismic event. It was suggested that rockburst damage should
that still need to be overcome. One existing challenge is to match the primarily be related to the state of stress and the local mine stiffness at a
actual, often violent, rock mass behaviour near excavation boundaries given rockburst damage location. Kaiser6 and more recently Cai and
with meaningful qualitative and quantitative engineering models to Kaiser2 used field evidence to illustrate and support this perspective.
arrive at robust and cost-effective ground control solutions.1,2 In this paper, strainburst phenomena are explored and an in-
A systematic rock support design approach for burst-prone ground novative numerical simulation approach is adopted to investigate the
was outlined in 1996 in the Canadian Rockburst Support Handbook.3 In mechanisms underlying ‘self-initiated’ and ‘dynamically triggered’
2013, Kaiser and Cai4,5 presented a critical review of these and more strainbursts. The approach is intended to assist in arriving at a sys-
recent design principles and indicated that it was “time to rethink” and tematic engineering approach that identifies where severe damage
critically assess the underlying assumptions and the consequences of conditions may exist; i.e., where the excavations are most vulnerable to
some of these widely adopted, but partially flawed, rockburst support rockburst damage, and how energy loading the ground support is


Corresponding author.
E-mail address: pkgeok@gmail.com (P.K. Kaiser).

https://doi.org/10.1016/j.ijrmms.2019.03.021
Received 30 July 2018; Received in revised form 21 January 2019; Accepted 10 March 2019
Available online 16 March 2019
1365-1609/ © 2019 Elsevier Ltd. All rights reserved.
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

released from the ‘strainburst volume’ and the surrounding rock mass. stress-fractured shell and ‘outer’ competent shell.
The complex topic of support selection for rockbursting ground is In the inner shell, brittle rock fails predominantly by extensile
treated in a qualitative manner by Kaiser6 and in a quantitative manner fracturing (i.e., spalling), breaking the rock mass into fragments sub-
by Kaiser and Cai2 (Volume 1 (2018); Volumes II and III pending). Of parallel to the excavation boundary. The extent of this zone is limited
the three volumes by Kaiser and Cai, the second volume will specifically by a confining pressure threshold that depends on many factors (e.g.,
deal with the phenomenon of rockburst damage caused by dynamic rock mass heterogeneity) but can be approximated by a confining
excavation failures (strainbursts). pressure threshold of σ3≈UCS/10.11 In the outer shell, at locations
Here we focus on the violent excavation failure process that is where σ3 > UCS/10, crack and extensional fracture propagation is
caused by a sudden collapse of an excavation due to excessive loading suppressed and shear rupture with kink-band formation and shear
of a brittle rock mass. Results are presented that capture the strainburst through intact rock occurs.
process and demonstrate how rock may be ejected and support is When the rock mass near an excavation is excessively deformed by
loaded during a strainburst. As will be discussed, the simulation results tangential straining, it is prone to strainbursting, particularly near the
suggest that conventional energy-based support design approaches are interface between the inner and outer shell; i.e., near the spalling limit.8
flawed for strainbursting ground. Similar to a rock testing load frame During the violent rock mass failure process, this fractured rock has a
where strain energy can be released from the load frame and trans- high bulking tendency due to a geometric mismatch of broken rock
ferred to the rock causing it to fail in a violent manner and ‘explode’, fragments. The inner shell is most prone to ejection of broken rock
rock near an excavation boundary can likewise be rapidly loaded by the because much of the cohesive strength has been destroyed. These
mine system. Once deformed past its peak strength, the rock inside the processes will later be demonstrated by numerical simulations.
‘burst volume’ fractures and bulks (i.e., increases in volume), releasing
energy stored in the failing rock and the rock mass surrounding the 2.1.1. Depth of failure df
failing rock volume.7 Part of the broken rock in the inner shell can collapse if in-
sufficiently supported, leading excavations with stress-fractured rock to
2. Brittle failure of unsupported excavations become vulnerable to over break and unravelling. The maximum or
extreme depth of over break or depth of failure2 in unsupported ground
2.1. Static excavation failure modes can be established empirically based on observations of the ‘extreme’
depth of failure (spalling) relative to the calculated stress level SL11:
Under static loading, the behaviour response of an underground dfe
opening to excavation can be characterized by nine failure modes as = 1.25⋅SL − 0.51 ± 0.1
a (1)
illustrated by the excavation behaviour matrix presented byKaiser
et al.8 These are a function of the rock mass quality, defined by the where a is the tunnel radius.
degree of fracturing (or fracture intensity) and the in-situ stress or stress The mean normalized depth for failure, however, is much smaller1
intensity near the excavation boundary, defined by the in-situ stress to and typically ranges from about 20 to 30% of dfe or:
strength ratioσ1/σc. For this purpose, the stress level1 SL = (3σ1– σ3)/
dfe
UCS = σmax/UCS was introduced as an index to characterize the dfm ≈
3.5 to 4.5 (2)
proximity to failure, where UCS refers to the uniaxial compressive
strength of intact rock. The benefit of using the stress level is that the Accordingly, depending on the stress and strength variability along
impact of the far-field stress ratio k = σ1/σ3 is accounted for. This is a tunnel, the depth of failure ranges from zero to dfe, and, because the
particularly relevant when mining induces changes and rotation of the strainburst potential is highest were the tangential stress is con-
principal stresses leading to a change in the local stress ratio k* within centrated near the excavation boundary, it follows that self-initiated
the influence zone of an excavation affected by mining. strainbursts may occur at distances from the wall ranging from near
Excavations in strong massive to moderately or discontinuously zero to slightly larger than dfe. This will later be illustrated by strain-
jointed rock masses are prone to stress-assisted failure modes once the burst simulations.
stress level SL exceeds 0.3 to0.5. According to Barton,9 minor bursting
(rock ‘spitting’ or popping) is to be expected during tunnel advance at 2.1.2. Bulking of stress-fractured ground
excavation-induced stress levels of SL > 0.65, and ‘severe’ rock- If stress-fractured ground, consisting of strong rock fragments inside
bursting, largely due to deeper strainburst volumes, is to be expected the inner shell (at df > 0), is deformed by tangential straining, its
for SL > 1.0. Such strainbursting is encountered in brittle rock (with volume has to increase due to a geometric non-fit of rock fragments.
high ‘intrinsic brittleness’) but only if the loading or mine system Near the excavation boundary, the rock can only deform into the ex-
stiffness is lower than the post-peak slope of the stress-strain curve of cavation and the bulking is unidirectional into the excavation. A linear
the rockmass. Strainbursting may occur at the excavation wall or away bulking factor BF, defined as the change in length in the radial direction
from the wall near the interface between the excavation damage zone normalized to its original length, was introduced in Ref. 3 to char-
and the more competent, (near-) elastic rock mass surrounding it. In acterize the bulking-related inelastic displacements that load the rock
other words, strainbursts tend to occur at or near the static depth of support or strain rock reinforcements. Because stress fracturing and
failure df. thus rock mass bulking is driven by the tangential straining of the rock
Kaiser et al.8 and Kaiser and Kim10 demonstrated that the failure near the excavation, less bulking is to be expected in civil tunnels and
envelope of brittle rock and rock masses is depressed in the low con- during initial mine development than in later stages of mining when the
finement zone of the principal stress space due to extensional failure extraction ratio is high and mining-induced tangential strains are
processes promoting rock spalling. The failure envelop is S-shaped or added. The reader is referred to Ref. 11 for guidance on how to select
tri-linear and the stress space can be divided into two behavioural applicable rock mass bulking factors to arrive at anticipated wall and
zones: spalling dominated stress fracturing at low confinement, and
shear rupture at high confinement (Fig. 1). This divides the rock mass
2
surrounding an excavation into two zones, referred to as the ‘inner’ The depth of failure is defined as the depth that an excavation would fail to
if it is unsupported. The extreme depth of failure dfe is the depth recorded at
locations with the deepest or the most extreme failure. The mean depth of
1
σmax is the excavation-induced stress at the location with the highest tan- failure dfm is the average condition that is expected over a domain of uniform
gential stress near the wall of an equivalent circular excavation in elastic rock. ground but with variable strength parameters.

53
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

Fig. 1. (a) Illustration of inner and outer zone behaviours, dominated by extensile brittle fracture and spalling under low confinement (inner zone), and shearing
under higher confinement (outer zone); (b) “S-shaped” or trilinear failure envelope differentiating low and high confinement behaviours at UCS/10 threshold.

support deformations and the safety margin with respect to the de- illustrated later by the strainburst simulations.
formation compatibility of the support components.
2.2.1. Strainburst process
2.2. Dynamic excavation failure modes The term ‘strainburst’ is often associated with minor bursting pro-
cesses, e.g., during the advance of tunnels in highly stressed ground
Excavations loaded by dynamic disturbances generally exhibit the causing ‘spitting’ or ‘popping’ ground. In the context of this paper, a
same failure modes as described above for static conditions except that strainburst is a sudden, violent or dynamic failure of an excavation
dynamic factors modify the forces, stresses, deformations and de- whereby the rock mass rupture is initiated by mining-induced, time-
formation rates (velocities). A highly stressed excavation may experi- dependent, or dynamic stress changes until the rock mass strength
ence one of three basic dynamic failure modes: i) shakedown, ii) (capacity) is reached. The major principal stress σ1 contours in Fig. 2a
strainbursting with or without rock ejection due to high bulking rates, show that various stress risers (red) are typically created around an
and iii) rock ejection by energy or momentum transfer from remote excavation due to rock mass heterogeneity and brittle failure of rock
seismic sources. If rock ejection occurs, it may be related to energy block boundaries. This can also lead to relaxed zones (blue) near the
transmitted from a remote seismic source, but in many case- excavation boundary.
sstrainbursting is often dominated by the mining system stiffness and It is well-known that the combination of increasingσ1 and de-
the intrinsic brittleness. For this reason, it is meaningful to differentiate creasingσ3 can lead to brittle rock failure if the spalling limit is crossed.
between: In relatively stiff loading systems, e.g., single tunnels in homogeneous
Self-initiated strainbursts: rockbursts caused by a gradual weak- ground, this is typically reflected in a gradual spalling process leading
ening of the rock mass such that the local stress after some time exceeds to notch formation. In heterogeneous ground with variable stress and
the rock mass strength. In this case, the damage is only related to the strength conditions, however, unstable equilibrium conditions can be
energy stored in the failing rock and the energy released from the created and failure may occur in a violent manner with a sudden ‘ex-
surrounding rock mass. The radiated energy or intensity of the co-lo- plosion’ of a critically stressed ‘burst volume’ (Fig. 2b). The contours in
cated seismic event is related to the strainburst intensity. Fig. 2b show critical stress to strength ratio hot spots.
Mining-induced strainbursts: rockbursts caused by time-dependent Because rock can only move toward the excavation, all volume in-
loading or mining-induced displacement or strain that change the local creases due to rock mass bulking10 will be focused toward the tunnel
stress near an excavation in such a manner that the stress reaches the resulting in an inward displacement δ (Fig. 2b). If this displacement
rock mass strength. The radiated energy or intensity of the seismic occurs suddenly, over a short rupture time Δt, the associated average
event is again only related to the strainburst intensity. inward displacement velocity vi = δ/Δtas shown by the velocity vector
Seismically triggered strainbursts: self-initiated or mining-induced in Fig. 2c. For example, for δ = 0.1 m and Δt = 0.1 s, the resulting
strainbursts that are triggered by a remote seismic event. In this case, average inward velocity would be vi = 1.0 m/s. This velocity is entirely
the remote event is the primary seismic event and the seismic event co- generated by the sudden bulking process, independent of any need for a
located with the strainburst damage is a secondary event. However, the remote seismic event to trigger the strainburst. Quantitative means to
damage is not directly related to the intensity (magnitude) of the re- estimate rock mass bulking were presented in Ref. 11.
mote event that serves as the primary trigger of the strainburst. Because the rock mass is disrupted during a strainburst and frac-
Here we focus on strainbursting with or without rock ejection; tures propagate by extension, followed by coalescence and shear
strainburstscaused by dynamic loading from a remote seismic event or failure, it is reasonable to assume that the effective confinement acting
by momentum or energy transfer from a remote event are not con- on the burst volume is low, facilitating high bulking factors BF.
sidered in this paper. Typically, average BF-values for bulking over the depth of
The strainburst failure process is driven by stress increases around strainburstingdSB may range from 10 to 20% if unconfined, for example
the excavation causing tangential straining of the rock in the inner in the unsupported floor, decreasing to 5–10%, if lightly confined and
shell, and is therefore dominated by the stored strain energy and the in a soft loading environment, decreasing further to < 3%, if well
loading system stiffness LSS. If associated with rock mass bulking, large confined and in a relatively stiff loading environment. In other words, a
static and dynamic displacements toward the excavations are en- 1 m deep strainbursting zone may cause between δ = < 30–200 mm of
countered, and if this deformation occurs in a short time, called the sudden radial bulking displacement. This could generate velocities in
‘rupture time’ Δt, rock fragments can move at a high velocity and may the range of < 0.3–4 m/s if the rupture time was in the order of 0.1 and
get ejected if insufficiently supported or retained. This will again be 0.05 s. Such velocities are typically back-calculated from rock ejection

54
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

Fig. 2. (a) Bonded block model (BBM) of a 10 m wide circular tunnel, with a 12-m cross joint spacing and horizontal to vertical stress field ratio of k = 0.7; Pierce
(2016, pers. comm.). Shown are the major principal stress contours with hot colors indicating high stress zones; (b) Example of hot spots in a continuum model with
stress and strength variability, after12; also shown are forces driving the burst volume to failure and resulting inward deformation δ; (c)‘Before’ and ‘After’ geometry
of burst volume (dfo = df before strainburst; df depth of failure after strainburst; dSB = depth ofstrainbursting increasing by bulking and resulting velocity vi). (For
interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

caused by rockbursts. initiated when the load reaches the capacity of the arch, and the failure
In summary, the failing rock will deform during strainbursts at an potential is 100% at Point A. As it deforms, potential energy is trans-
initial inward displacement velocity vi that depends on the rupture ferred to kinetic energy and then the kinetic energy is consumed as the
time. It will impact the support at an impact velocity that is equal to or snapped-through arch regains a new equilibrium at Point B. Shown
lower than this initial velocity. This will again be supported subse- superimposed on Fig. 3 in red is the corresponding schematic velocity
quently by numerical strainburst simulations. versus displacement graph, indicating an initial acceleration, followed
A strainburst may also occur in a well-supported rock mass, i.e., in by a deceleration until the new equilibrium is reached (the bungee-
the reinforced rock behind mesh or shotcrete, or behind the reinforced cable-like oscillation that would be expected before a new equilibrium
ground. is established at (B) is ignored in this graph).
A strainburst is a similar process whereby a volume of rock fails
(bursts) when a ‘Limit Point’ at (A) is reached. As the stress in the rock
2.2.2. Strainburst potential and burst severity mass increases, the strainburst potential SBP increases until the safety
For engineering design purposes, it is necessary to answer the margin SM = 0 at Point A. As the bursting rock bulks and increases in
questions ‘Where could a strainburst occur?’, i.e., what is the strainburst volume, fragmented rock is accelerated and moves at increasing speed
potential, and ‘How violent will a strainburst be?’, i.e., what is the toward the excavation. This velocity can be reduced if the supported
strainburst severity. rock mass acts as an energy sink. If the support is effective, a new
The strainburst potential SBP defines the likelihood of a strainburst equilibrium is reached after some deformation, i.e., when all released
occurring. The higher the SBP, the closer an excavation is to the limit kinetic energy has been consumed by the reinforced rock mass at Point
point where the strength is equal to the rock mass stress. By analogue B. Fig. 3 illustrates an example of an effective rock support system that,
with a point loaded steel arch (Fig. 3), a snap-through failure process is while damaged (cracked shotcrete) and having experienced some
loading of the rockbolts, was able to reach a new equilibrium at Point B.
The strainburst severity SBS is commonly described in terms of the
volume (weight) of displaced rock, the degree of damage caused to the
support (e.g., Ref. 13), and the ejection velocity (if any). The latter is
only meaningful if the support is ineffective and failed because the
supported rock mass was not able to dissipate enough energy to reach a
new equilibrium at Point B. Fig. 3 shows that the ground velocity and
therefore the ejection velocity will depend on the displacement at the
time of support failure. It increases until the supported rock is able to
sufficiently decelerate the bursting rock (from Point C onward). The
larger the displaced volume and the higher the ejection velocity, the
more severe the strainburst is.
If the support is effective (e.g., Fig. 4) no rock is ejected and the
ejection velocity is zero (at Point B). However, the ground velocity is
not zero except when equilibrium has been re-established. Because of
the rapid change from initial acceleration to deceleration, any mass that
is not well retained can be ejected. For example, loose rock can eject
into the mesh or poorly bonded shotcrete slabs can spall and be
ejected.3 In these situations, the strainburst severity has to be assessed

Fig. 3. Steel arch analogue of self-stabilizing instability process showing load- 3


Note: whereas the support was effective in reestablishing a new equilibrium,
displacement with superimposed velocity-displacement graph in red. (For in- there may still exist a high safety risk from ejected ‘loose’ rock. For this reason,
terpretation of the references to color in this figure legend, the reader is referred mesh-over-shotcrete is a highly recommended practice for strainbursting
to the Web version of this article.) ground to provide a retaining function.

55
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

geometry (span): single tunnels are stiffer (lower DP) than multiple
tunnels with rib pillars in between (higher DP); intersections are softer
(higher DP); and interaction between mining levels can modify the DP;
(2) Mining stage and environment: the DP increases with increasing
extraction ratio; and gravity loading scenarios (k < 1), as in South
Africa, are very soft and exhibit a high DP; (3) Geological structures:
stiff dykes reduce the DP upon entry but increase it upon exiting; and
intersecting faults or shears enhance the DP (as well as the degree of
freedom, facilitating falls of ground).
In reality, each of these factors has to be considered. For the
strainburst simulations presented here, one selected geometry with
different deformation potentials is simulated.

2.2.4. Velocity estimation


As indicated above, there are two velocities of interest for support
design: (1) the ejection velocity if the support is ineffective (e.g., ra-
Fig. 4. Deformed ground support with cracked shotcrete after a strainburst (at velling rock hitting mesh), and (2) the velocity by which the support is
the stability status, Point B in Fig. 3). impacted if the support is effective (e.g., loading at the plate or along
the bolt shaft). The former is of interest for ejection velocity back-
analyses and the latter for support selection. The ejection velocities are
in terms of ejected ‘loose’ rock and the accumulated displacement at
typically estimated by considering energy conservation or momentum
Point B.
transfer (e.g., Ref. 3). However, these approaches cannot be used for
Excavations are therefore vulnerable to severe strainburst damage
strainbursts because there is no direct relation to the damage causing
when the energy source is high and the energy sinks are small. This
ground motion when energy is released from the rock mass surrounding
offers two challenges: how to estimate the sources of energy during a
the burst volume rather than transmitted from a remote seismic event.
strainburst and how to assess the energy sinks of an effectively sup-
During a strainburst, the burst volume bulks and causes an inward
ported fractured rock mass. The former will be demonstrated later by
displacementδ (Fig. 2b). The applicable bulking factors BF can be es-
the numerical strainburst simulations.
timated from charts provided by Kaiser.11 For strainbursts, the dis-
placement velocity v at the inside of the burst volume (Fig. 2b) is
2.2.3. Deformation potential proportional to the rupture time Δt: v≈δ/Δt. Hence, the rupture time Δt
Tarasov and Potvin7 differentiate between intrinsic and relative of the burst volume is needed to estimate the velocity generated by
brittleness. Here we focus on the relative brittleness where the energy strainbursts. Unfortunately, little to no data is available to select this
stored in the rock mass surrounding the burst volume serves as the parameter but qualitative and theoretical evidence suggests that it is
energy source. For a given stress level, the energy demand (driver of a much less than 1s, i.e., likely between 0.1 and 0.05 s when visible
strainburst) is the deformation potential DP. This is illustrated by Fig. 5 ejection is observed and provides some supporting evidence.6
showing the vertical displacement vectors (and contours) before and In summary, strainbursting is a complex excavation failure process
after a strainburst (see yellow star for location) that removes all rock that is highly relevant for the design of highly stressed underground
from a burst notch (blue arrow). The wall-parallel displacement dif- excavations in brittle rock. Whereas progress has been made in un-
ferential is the deformation potential DP. This is illustrated by an derstanding the various influencing factors, further research is required
analogue of a uniformly loaded beam in the same Fig. 1f the supporting to quantify the strainburst potential and severity. In the following, a
pillar bursts and collapses, the displacement profile changes from the novel bonded block method is introduced and tested to simulate the
full to the dashed deflection line and the displacement indicated by the strainburst process and to quantify the severity of strainbursts.
double arrow is the deformation potential DP; i.e., the deformation that
would ‘potentially’ occur if the pillar failed. In other words, in the 3. Numerical simulation of strainburst process
original state before the pillar collapses, the beam has a ‘potential for
deformation’ equal to DP. 3.1. Introduction
The DP is proportional to the total available energy. Hence, if the DP
is greater than a critical value, energy will be available for release by Numerical simulation of rockbursts remains a challenge due to the
rock ejection if unsupported. The potential for severe strainbursting is inherent difficulties in dynamic failure simulation. Existing methods
high where the DP is high. either simulate dynamic failure under static loading conditions while
The DP is affected by various factors such as (1) Excavation adopting threshold factors (normally in terms of stress, strain, or

Fig. 5. Analogue to illustrate the concept of deformation potential.

56
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

energy) as an indicator of rockbursting, or incorporate an assumed of a pillar in a stiff miming environment as well as one particular rock
dynamic disturbance to simulate rockburst damage due to dynamic mass characteristic. During failure, energy is released from the cap and
conditions. Here, instead of adopting a prescribed dynamic disturbance, floor rock with a modulus of 50 GPa. In reality more energy may be
an innovative method is adopted whereby a seismic event is generated released depending on the deformation potential introduced in Section
by a rock cracking model utilizing the distinct-element bonded block 2.2.3 (Fig. 6). Hence, this model simulates a strainburst in a stiff mine
method. This is used to simulate self-initiated strainbursts under var- environment, e.g., when mining a single rib pillars in a stiff rock mass.
ious loading conditions. It is not directly applicable to situations of a room and pillar array.
Considerable efforts have been devoted by others to simulate The triangular blocks in the Trigon assembly have a uniform size of
rockbursts using numerical modelling as summarized by Gao 0.1 m. Each triangular block is further discretized into a few finite
et al.14One challenge of adopting dynamic disturbance models is that it difference zones to make it deformable. The host rock above the pillar is
is necessary to specify the waveforms, amplitudes and frequency, and represented by a large elastic block with a width of 12 m and thickness
the impacting direction of the dynamic disturbance generated from of 5 m. The host rock below the pillar is represented by an elastic block
fault slip sources. In the study presented herein, a novel distinct ele- with a width of 12 m and thickness of 2 m.The two host rock blocks are
ment bonded block method is introduced using the commercial code also discretized using similar finite difference zones to ensure smooth
UDEC to simulate strainbursts. Self-initiated strainbursts are considered stress transfer between the host rock and the pillar. The deformation
and a strength reduction approach is adopted to initiate a dynamic modulus of the host rocks is assumed to be 50 GPa. The resulting
disturbance that will eventually evolve into a strainburst. loading system is analogous to a high local mine stiffness which plays a
role in the strainbursting process and severity of rockburst damage.19
The mechanical properties of the triangular blocks and contacts used
3.2. Numerical simulation of self-initiated strainburst
for simulating the pillar are listed in Table 1. These properties were
calibrated against those reported by20 for a jointed granite with the
3.2.1. Model configuration
following laboratory tested properties: Young's modulus E = 11.9 GPa,
All strainburst simulations were performed using the UDEC-Trigon
unconfined compressive strength of the rock mass UCS (rock
bonded block method.15 In this method, the rock mass is represented by
mass) = 58 MPa, cohesion c = 8 MPa, and ‘friction and dilation’ angle
an assembly of triangular blocks bonded by contacts. Each block is
(ϕ+i) = 59°. The calibration was achieved by simulating a series of
further discretized into triangular finite difference zones and assigned
unconfined and confined compression tests on a pillar with a width of
elastic properties. Failure is restricted to contacts by shear or tension,
2 m and a height of 4maccording to a parametric variation procedure.15
depending on the stress state and mechanical strength properties as-
The parameters used for simulating the host rock and the contact be-
signed to the contacts. Previous research has shown that this method is
tween the host rock and the pillar are also listed in Table 1. Fig. 7 shows
able to capture many of the features of brittle rock failure from the
the simulated axial stress-axial strain curves obtained from an un-
laboratory scale to the field scale.16,17 The UDEC Trigon model used for
confined and two confined compression tests on the calibration model.
this study, shown in Fig. 6, represents a stressed mine pillar with a W:H
It can be seen that a brittle rock mass behaviour is represented by this
ratio of 2, specifically with width of W = 12 m and a height of H = 6 m.
geometry and the listed parameters.
According to Ref.18, such a pillar in massive to moderately jointed rock
should be stable at average stresses of about 70–80% of the unconfined
3.2.2. Modelling procedure
rock mass strength. This pillar model simulates one particular scenario
As a heterogeneous rock mass is subjected to load due to mining-
induced stresses, critically oriented joints (contacts in the model) get
stressed to their peak capacity and eventually the weakest or highest
stressed contact fails. This is called ‘self-initiated’ failure because no
external stimulus, other than the static stress, is needed to bring the
contacts to failure. For self-initiated strainburst simulations, sliding
along cracks is therefore allowed to occur when the strength of a cri-
tically oriented contact is reached. When the contact fails, the cohesion
is instantaneously lost and the contact is allowed to deform in shear. As
stress is then dynamically redistributed, other contacts become criti-
cally stressed and cracking propagates. This results in the generation of
dynamic disturbances which causes further cracking and repeated dis-
turbances. The cracking-dynamic disturbance and cracking feedback
continues in a domino-like fashion and eventually leads to a strainburst
whereby fractured rock is moving into the excavation; i.e., it is ejected
if not supported.
The tangential stresses in the rock mass near an excavation wall or
in a pillar must be sufficiently high to initiate crack damage and to
generate a self-initiated strainburst. For this, the 12 × 6 m pillar model
was loaded in two stages:
Stage I: The pillar was loaded under static conditions to a point
where cracking occursat the pillar edge (see photograph in Fig. 8 and
initial crack damage pattern). The load was applied to the pillar by a
constant velocity of 0.15 m/s at the top boundary while the bottom
boundary was fixed. Although this loading rate may appear to be fast, in
UDEC modelling, the time step in each calculation cycle is auto-
matically adjusted to an extremely small value (e.g. 10-7 s) for a static
analysis. A loading rate of 0.15 m/s was therefore equivalent to ap-
proximately 3 × 10-4 mm/step, implying more than 10 000 steps to
move a loading platen by 3.3 mm. Loading was then stopped and suf-
Fig. 6. UDEC-Trigon model used for strainburst simulation. ficient time steps run to stabilize the model. No further cracking was

57
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

Table 1
Mechanical properties used in UDEC-Trigon model for simulating strainburst.
Modelled element Block properties Contact properties

3
Density (kg/m ) E (GPa) Poisson's Ratio kn (GPa/m) ks (GPa/m) Cohesion (MPa) Friction (°) Tensile strength (MPa)

a a
Pillar 2600 16.6 0.25 2218 887 25.2/0 45/30 4.3/0a
Host rock 2600 50.0 0.25 N/A N/A N/A N/A N/A
Boundary N/A N/A N/A 11 090 4435 2.0 30.0 0

a
Peak and residual values.

observed before the next stage was initiated.


Stage II: After resetting the displacements and velocities in the
model to zero, the dynamic calculation mode was activated. The local
damping used in Stage I was removed and Rayleigh damping of 0.5%
was applied. Absorbing boundary conditions were applied to the outer
host rock boundaries to represent wave propagation into the far-field.
Reflecting boundary conditions were applied to the two lateral sides of
the pillar. A total of 8 ms dynamic model time was then calculated.
Two key factors involved in dynamic calculations were determined
carefully chosen to ensure proper strainburst simulation:

3.2.2.1. Mesh size. To accurately represent wave transmission through


a model, The spatial element size must be smaller than approximately
one-tenth to one-eighth of the wavelength associated with the highest
frequency component of the input wave20 to accurately represent wave
transmission through a model,. Preliminary results have shown that the
system with the given geometry and parameters exhibits a seismic wave
with a frequency of 1912 Hz, a speed of 2400 m/s, and a wavelength of
Fig. 7. Calibrated axial stress vs. axial strain curves using the parameters listed 1.3 m.14 This suggests that the mesh size should be smaller than 0.13 m
in Table 1. and a mesh size of 0.1 m was used to discretize the pillar.

3.2.2.2. Damping. In soil and rock, natural damping is mainly


hysteretic (i.e., independent of frequency). It is difficult to reproduce
this type of damping numerically because of the problem with path

Fig. 8. Simulated cracking patterns at different calculation times during the generation and propagation of a self-initiated strainburst; the photographs show the
typically observed spalling (hour-glassing) of a pillar and strainburst damage below a supported pillar wall.

58
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

dependence. Local damping is commonly recommended for static


analyses.21 When local damping is used, the damping force on a node
is proportional to the magnitude of the unbalanced force. A default
damping ratio of 0.8 is suggested by Itasca for static analysis. This value
was used in Stage I.
Rayleigh damping is commonly used for dynamic calculation of
geological materials.21 Damping commonly falls in the range of 2–5%
of critical. Energy dissipation occurs during joint slip. Thus, for many
dynamic analyses that involve large block deformation or large joint
displacement, only a minimal percentage of damping (e.g., 0.5%) may
be required.21 Thus, Rayleigh damping with a critical damping ratio of
0.5% was used in Stage II.
Unlike most existing modelling methods where unstable rock fail-
ures (i.e.,rockbursts) are simulated by applying an assumed input wave
to the boundary of a model, the approach adopted here lets the model
develop a dynamic failure process. During static loading, the pillar was
loaded to a post peak point with cracking in the outer parts of the pillar
skin. When the loading process was stopped and sufficient time steps
had been run, the overall model was in equilibrium under the static
calculation mode. When the dynamic calculation was activated, the Fig. 10. Temporal evolution of tensile and shear cracks (in both pillar walls).
damping was reduced from 80% to 0.5% such that for some cracks that
were in a critical force-equilibrium condition, the change in damping zone that typically leads to hour-glassing of the pillar as shown in the
disturbed the force-equilibrium, resulting in sliding along the cracks inserted photograph. As the confinement toward the pillar core in-
which generated minor local dynamic disturbances. These dynamic creases, the contact failure mode changes to shear cracking (colored
disturbances radiated outward as dynamic waves (see later in Fig. 14), blue).
triggering new cracking at critically stressed contacts. This resulted in During this self-initiated strainburst, the depth of fracturing gener-
the generation of larger dynamic disturbances, which then triggered ated by static loading increases by about 3 m within 8 ms (from red
further cracking and disturbances. This cracking-dynamic disturbance- dashed to full red notch shape), and if all the cracked rock was un-
cracking feedback sequence continued in a ‘domino-like’ fashion cul- supported and allowed to unravel, the initial failed volume of ≥10 m3/
minating in a strainburst with the centre of dynamic bulking at the m would nearly double after 8 ms to approximately 16 m3/m. It is for
transition from the crack damaged ground to the elastic, undamaged this reason that the severity of rockburst or strainburst is typically
ground. This mechanism is consistent with the actual damage process classified by the volume of displaced ground. These images illustrate
encountered when a volume of rock bursts under excessive load, i.e., the rock mass failure process that occurs during a strainburst.
without energy imposed from a remote seismic event. In this simulation, a 2 m deep ‘burst volume’ of undamaged rock is
stress-fractured. After about 2 ms, tensile cracks form more rapidly than
3.3. Self-initiated strainburst model results shear cracks (Fig. 10). As a consequence, the broken rock will bulk
rapidly after about 4 ms and this rock and the fractured burden in front
3.3.1. Crack patterns of the burst volume will move rapidly toward the excavation and, if
Fig. 8 illustrates the evolution of the crack damage zone within the ineffectively supported, will be ejected.
first 8 ms of dynamic modelling time And Fig. 9 shows the detailed
deformation pattern on the left side after 8 ms of dynamic modelling
time. Initially, tensile cracks (colored red) generate a notch-like damage 3.3.2. Displacement patterns
The resulting displacement pattern is presented in Fig. 11. It shows
that at the given time (8 ms), the displacement pattern is non-uniform
and that a rockbolt crossing the burst volume would experience the
largest displacements inside the burst volume (shown in purple to red)
where the bulking of the bursting rock is highest.
The corresponding displacement profiles at the mid-height of the
pillar and at various times are presented in Fig. 12 for the left wall.
These profiles show that the displacement is highest in the burst volume
with a compression zone near the wall (1 to < 2 m depth) and an ex-
tension zone toward the pillar core (between 2 and < 4 m at less than
6 ms dynamic time). This is the volume of rock involved in the self-
initiated strainburst. Beyond this point the rock mass damage is pro-
pagating toward the pillar core (translation of displacement profile
(black dashed curve) until the entire pillar eventually yields.
The dashed arrow for 6 ms indicates that the displacement near the
wall may in reality be equal to the maximum displacement in the burst
volume, i.e., if a 1.2 m deep slab of non-fractured and thus non-com-
pressible rock was resting in front of the burst volume. On the other
hand, if the burden in front of the burst volume was bulking, e.g., by
1%, the displacement in front of the burst volume would be higher
(dotted red arrow). If the statically fractured rock was undergoing
compression, the displacement at the wall would be less (full red
Fig. 9. Simulated strainburst failure at 8 ms dynamic time. Block deformation is arrow). These scenarios are not reflected in the cracking model.
magnified 10 times. Because the wall displacements of between 4 and 20 mm occur in a

59
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

Fig. 11. Simulated displacement field after 8 dynamic simulation times showing 7.5 to < 15 mm of wall movement on the left and up to 27.5 mm on the right. The
4 m long rock or cable bolt is shown for illustrative purposes (not modelled).

For support design purposes, two aspects have to be assessed: plate


loading by wall displacements and internal straining of rock bolts by
rock mass bulking. The deformation at the plate is provided by Fig. 12
and the rate of loading will be discussed next. For the latter, it is ne-
cessary to establish the location and amount of maximum strain (dif-
ferential displacement) that a bolt could experience. This strain de-
pends on the yield length of a bolt, i.e., the length over with the
differential displacement is transferred to the steel. For the example
presented in Fig. 12, it is assumed that the differential displacement is
transferred from the rock to a grouted rebar over a ‘yield length’ of
0.2 m.
Fig. 13 shows that strain localizations occur along the bolt, first at
4 ms with a peak at about 2.7 m (blue x), inside the burst volume, and
then also at 6 ms at a greater depth of about 3.6 m (red x). The max-
imum calculated strain in this example is about 3% and this would be
insufficient to fail a rebar (unless it has been pre-stressed or pre-de-
formed by previous mining deformations). This example supports the
frequently observed bolt failures inside the fractured rock, i.e., at lo-
cations where strain localization occurs due to rock block separation,
Fig. 12. Simulated displacement profiles at six dynamic model times (2–8 ms). i.e., at 2.7 and 3.6 m in this case. A 2.4 m long rebar would be too short
A2.4-m long rock bolt is shown for illustrative purposes (not modelled). Red
to prevent fractures to open at 2.7 m depth and therefore would be
arrows are described in text. (For interpretation of the references to color in this
displaced as a unit into the excavation. A 4 m long cable bolt would be
figure legend, the reader is referred to the Web version of this article.)
just long enough to resist strain localization and thus suppress fractures
from propagating. After 6 ms, however, deeper cables would be re-
short rupture time (6 ms), the average horizontal wall velocity up to quired to resist the strain localizations as failure propagates toward the
this stage would range from 1 to 4 m/s depending on the burden re- pillar core.
sponse (see below). A bolt plate at the wall would therefore experienced
this impact velocity.

60
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

simulated time of 6 ms, the highest velocity has migrated to the (right)
wall.
The maximum velocities of individual fragments near the walls
(e.g., last image in Fig. 14 for 8 ms) vary widely between 1 and 5 m/s on
the right wall and 3 and 10 m/s. These fragment velocities would be
experienced by a loose retention system (mesh) or would represent the
velocities of fragments ejected from the wall. Even though the simu-
lated rupture times are very high, these model velocities are consistent
with wall and ejection velocities reported in the literature (e.g., by
Stacey22).
The horizontal velocity profiles at the mid-height of the pillar are
presented in Fig. 15. These velocity vector components are in line with
radically installed support components (a single bolt is shown sche-
matically in the figure). According to this model, the horizontal velo-
cities at the wall or the bolt plate are relatively low during the self-
initiated strainburst phase, i.e., 1–2.5 m/s. Once failure propagates into
the pillar core and the entire pillar yields at > 8 ms, the wall velocity
increases rapidly reaching 5 m/s at 8 ms.
Fig. 13. Simulated horizontal strain profiles at four dynamic model times
(2–8 ms) and an assumed yield length of 0.2 m A4-m long cable bolt is shown The highest horizontal velocity in a 2.4 m long bolt would be ex-
for illustrative purposes (not modelled). perienced inside the rock mass. This peak velocity propagates toward
the plate and reaches about 4.5 m/s at 1.8 m from the wall at 6 ms
(Fig. 14a). It is of interest to note that a bolt located at mid-height of the
3.3.3. Velocity patterns
pillar would be ‘pushed from behind’ with a higher velocity at a deeper
The resulting velocity patterns are presented in Fig. 14. This figure
point than at the plate (between 1.5 and 2 m; Fig. 14 and b). While
presents the magnitude of the total velocity vector which is inclined
loaded at a high velocity (< 6 ms), such a bolt would be under com-
depending on the fragment movement direction. These images show
pression. Later, when the pillar yields at 8 ms, the entire bolt will move
that waves initially radiate toward the pillar core but the predominant
at a more or less constant velocity of about 3.5 m/s. At this stage, a
ground motion eventually points toward the excavations on both sides
2.4 m long bolt would therefore experience little horizontal strain (as
of the pillar. As the strainburst evolves, the highest velocities are in-
shown in Fig. 13) but move in tandem with the surrounding rock.
itially encountered inside the burst volume. Eventually, at the longest

Fig. 14. Simulated total ground velocities for dynamic model times between 0.4 and 8 ms.A4 m long bolt is shown for illustrative purposes (not modelled).

61
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

Fig. 15. (a) Simulated horizontal ground velocities velocity profiles at six dynamic model times; (b) horizontal displacement contours at 6 ms. the 2.4-m long rock
bolt is shown for illustrative purposes (not modelled).

As discussed in connection with the displacement profiles (red ar- surrounding rock mass. This paper focuses on ‘self-initiated’ strain-
rows in Fig. 12), the velocity near the wall may in reality however be bursts and provides an overview of this phenomenon and their re-
equal to the maximum velocity in the burst volume, i.e., if a slab of non- levance to mining. It sets the framework for the numerical simulations
fractured and thus non-compressible rock was resting in front of the presented in the second part of this paper.
burst volume. If the burden in front of the burst volume was com- An innovative distinct-element bonded block method was presented
pressible it would be less and if it was bulking, the velocity in front of to simulate strainbursts. Instead of triggering failure by an assumed
the burst volume would be higher. These scenarios are not reflected in dynamic disturbance, the adopted method simulates the generation of a
the cracking model but illustrate that bolt plates may encounter highly seismic event by rock mass fracturing. Self-initiated strainbursting can
variable velocities during self-initiated strainbursts. be successfully captured by these simulations.
The self-initiated fracture process generated a dynamic disturbance
which triggered further cracking and eventually led to a self-initiated
3.3.4. Concluding remarks on self-initiated strainburst simulation
strainburst. The approach developed here captures the strainburst
Based on the displacements and velocities presented in Figs. 12 and
process and shows:
15, respectively, the average rupture time can be calculated and ranges
from 2 to 4 ms. The rupture time is an important parameter that con-
(1) energy released during a strainburst is ‘self-generated’ in that the
trols the strainburst severity or the velocity of rock movement during a
energy stored in the rock and the surrounding rock mass is released;
strainburst. The shorter the rupture time, the higher the velocity and
(2) no remote seismic event is needed to generate a strainburst (even
thus the higher the energy contained in displaced rock. The rupture
though remote seismic events may trigger a strainburst);
time depends on rock properties (in general, the stiffer, the higher) and
(3) the instability process is driven by the deformation potential but the
it depends on the failure process (slabbing: thick slabs have relatively
rupture process and rupture time depends on the characteristics of
high rupture/buckling time; strainburst of thicker labs tend to fail
the rock fracture process;
faster). There is no universal value for rupture time and there is little
(4) a strainburst leads to the disintegration of the rock mass in the burst
field evidence of rupture time estimates because the relevance of rup-
volume and is associated with geometric bulking, related rock and
ture time is commonly ignored. The authors experience however sug-
wall displacements, and ground motions (velocities);
gests that the modelled rupture times are by about one order of mag-
(5) if insufficiently supported, the ejection velocity is not related to the
nitude shorter than those estimated from field evidence. This can be
energy from a remote event but to the locally released energy, i.e.,
attributed to the characteristics of the chosen rock mass model with
the strainburst energy; and
high brittleness and sudden strength loss at the contacts. Further work
(6) if supported (not simulated for this article) the bolt plates and re-
will be required to adjust model characteristics to match rupture times
tention system can be impacted by high ground motions, and the
observed in the field.
bolt shafts can be internally strained.
The distinct element bonded block method using the commercial
code UDEC to simulate strainbursts does however capture the process of
The methodology and the results presented have highlighted lim-
self-initiated strainbursting. The strainburst simulation results are rea-
itations and areas for further research. The extent of damage and the
sonably consistent with field observations and demonstrate that dy-
speed of damage strongly depends on the assumed constitutive beha-
namic loading of supports and rock ejection can occur without a remote
viour of the contacts. Future studies need to consider the influence of
seismic event transferring energy to the excavation wall.
contact properties and Voronoi shape on the intensity of the dynamic
disturbance and thus the severity of the modelled strainburst as well as
4. Conclusions the impact of rock mass reinforcement and surface support.

Rockbursts cause damage to underground excavations in a sudden


and violent manner. During strainbursts, the damage process involves a Acknowledgements
sudden collapse of a volume of rock (the ‘strainburst volume’), near the
excavation boundary and the seismic event is co-located with the da- The authors wish to thank the Centre for Excellence in Mining
mage location. Strainburst may ‘self-initiate’ when the stress reaches Innovation (CEMI) and the Rio Tinto Centre for Underground Mine
the local rock mass strength causing an unstable failure process Construction (RTC-UMC) for their financial support of this work.
whereby energy is released from the ‘strainburst volume’ and the

62
F. Gao, et al. International Journal of Rock Mechanics and Mining Sciences 116 (2019) 52–63

Appendix A. Supplementary data non-persistent joints. Rock Mech Rock Eng. 2007;40(2):169. https://doi.org/10.
1007/s00603-006-0093-8.
11. Kaiser PK. Ground Support for Constructability of Deep Underground Excavations –
Supplementary data related to this article can be found at https:// Challenges of Managing Highly Stressed Ground in Civil and Mining Projects. San
doi.org/10.1016/j.ijrmms.2019.03.021. Francisco: Sir Muir Wood lecture of International Tunnelling Association at World
Tunnelling Congress; 2016:33.
12. Kaiser PK, Amann F, Steiner W. How highly stressed brittle rock failure impacts
References tunnel design. In: Labious Dudt, Mathiereds. Eurock’10, Lausanne, Switzerland, Zhao.
London, UK: Taylor & Francis Group; 2010:27–38.
1. Kaiser PK. Ground control in strainbursting ground – a critical review and path 13. Heal DP, Potvin Y, Hudyma MR. Evaluating rockburst damage potential in under-
forward on design principles. RaSiM9, 9th International Symposium on Rock Bursts and ground mining. Proceedings of the 41st US Symposium on Rock Mechanics. Minneapolis,
Seismicity in Mining. Santiago, Chile. 2017; 2017:146–158. US: American Rock Mechanics Association; 2006:12.
2. Cai M, Kaiser PK. Rockburst Support Reference Book, Volume 1 Rockburst Phenomenon 14. Gao FQ, Kaiser PK, Stead D, Eberhardt E, Elmo D. Numerical simulation of strain-
and Support Characteristics. MIRARCO – Mining Innovation. Sudbury, Ontario, Canada: bursts using a novel distinct element method. Comput Geotech. 2019;106:117–127.
Laurentian University; 2018 free manuscript download from www.mirarco.org. 15. Gao FQ, Stead D. The application of a modified Voronoi logic to brittle fracture
3. Kaiser PK, Tannant DD, McCreath DR. Canadian Rockburst Support Handbook. modelling at the laboratory and field scale. Int J Rock Mech Min Sci. 2014;68:1–14.
Sudbury, Ontario: Geomechanics Research Centre, Laurentian University; 1996. https://doi.org/10.1016/j.ijrmms.2014.02.003.
4. Kaiser PK, Cai M. Critical review of design principles for rock support in burstprone 16. Gao F, Stead D, Kang H. Simulation of roof shear failure in coal mine roadways using
ground - time to rethink!. In: Potvin Y, Brady B, eds. Ground Support 2013. 2013; an innovative UDEC Trigon approach. Comput Geotech. 2014;61:33–41. https://doi.
2013:3–38. org/10.1016/j.compgeo.2014.04.009.
5. Kaiser PK, Cai M. Rockburst damage mechanisms and support design principles. 17. Gao F, Stead D, Coggan J. Evaluation of coal longwall caving characteristics using an
RaSiM8 Proceedings. Saint-Petersburg, Moscow, Russia. 2013; 2013:349–370. innovative UDEC Trigon approach. Comput Geotech. 2014;55:448–460. https://doi.
6. Kaiser PK. Excavation vulnerability and selection of effective rock support to mitigate org/10.1016/j.compgeo.2013.09.020.
rockburst damage. Rockburst: Mechanisms, Monitoring, Warning, and Mitigation. New 18. Martin CD, Maybee WG. The strength of hard-rock pillars. Int J Rock Mech Min Sci.
York: Elsevier; 2017:473–518 (Chapter 16) in:. 2000;37(8):1239–1246. https://doi.org/10.1016/S1365-1609(00)00032-0.
7. Tarasov B, Potvin Y. Universal criteria for rock brittleness estimation under triaxial 19. Kaiser PK, Cai M. Design of rock support system under rockburst condition. J Rock
compression. Int J Rock Mech Min Sci. 2013;59:57–69. https://doi.org/10.1016/j. Mech Geotech Eng. 2012;4(3):215–227. https://doi.org/10.3724/SP.J.1235.2012.
ijrmms.2012.12.011. 00215.
8. Kaiser PK, Diederichs MS, Martin JA, Sharp J, Steiner W. Underground works in hard 20. Kuhlemeyer RL, Lysmer J. Finite element method accuracy for wave propagation
rock tunnelling and mining. Proceedings of GeoEng2000. vol. 1. Technomic Publishing problems. J Soil Mech Found Div. 1973;99(5):421–427.
Co.; 2000:841–926. 21. Itasca Consulting Group Inc. UDEC (Universal Distinct Element Code). 2014; 2014
9. Barton NRA. Q-system case record of cavern design in faulted rock. Proc. Symposium Minneapolis: Itasca Version 6.0.
on Tunnelling in Difficult Conditions. vol. 16. 1994; 1994:1–14 Turino, Italy. 22. Stacey TR. Addressing the consequences of dynamic rock failure in underground
10. Kim BH, Cai M, Kaiser PK, Yang HS. Estimation of block sizes for rock masses with excavations. Rock Mech Rock Eng. 2016;49(10):4091–4101.

63

You might also like