Int. J. Mining and Mineral Engineering, Vol. 4, No.

3, 2013 201

Numerical analyses of the hangingwall failure

due to sublevel caving: study case

Tomás Fernando Villegas Barba*

and Erling Nordlund
Division of Mining and Geotechnical Engineering,
Luleå University of Technology,
971 87 Campus University, Luleå, Sweden
Fax: 46 (920) 491336
Fax: 46 (920) 491913
E-mail: tomvil@ltu.se
E-mail: erling.nordlund@ltu.se
*Corresponding author

Abstract: The sublevel caving used in Kiirunavaara mine induces failure and
subsidence of the hangingwall. Two sections of the mine were studied by
means of numerical analyses. Numerical models were developed using finite
element and discrete element codes. The former was applied to calculate the
location of new failure surfaces in the hangingwall and to estimate the break
angle when mining advances downwards. The latter was used to analyse the
displacement path of the caved rock during draw and to determine its effect on
the stability of the hangingwall and footwall. The models were calibrated using
displacement monitoring data. The finite element analyses indicated that the
break angle is almost constant for deeper mining levels but may change if
the geometry of the orebody changes. The discrete element model showed the
formation of a stationary zone along the footwall that reduces the magnitude of
the shear forces during draw, increasing its stability.

Keywords: Kiirunavaara mine; hangingwall failure; sublevel caving;

numerical analyses.

Reference to this paper should be made as follows: Villegas Barba, T.F. and
Nordlund, E. (2013) ‘Numerical analyses of the hangingwall failure due to
sublevel caving: study case’, Int. J. Mining and Mineral Engineering, Vol. 4,
No. 3, pp.201–223.

Biographical notes: Tomás Fernando Villegas Barba is a Mining Engineer

from the University of Sonora, Mexico. He has 10 years of experience in
underground and open pit mining operations as mine supervisor in Mexico.
He has worked as Lecturer at the Mining Program at the University of Sonora
since 1999 and graduated with Master of Science in Mineral Engineering
(2004) from New Mexico Institute of Mining and Technology, USA. He is
currently a PhD Student at Luleå University of Technology, Sweden.

Erling Nordlund is Professor at the Division of Mining and Geotechnical

Engineering at Luleå University of Technology. He is currently leading several
projects in the subject of rock mechanics such as impact of fire on the stability
in hard rock tunnels in Sweden, mining-induced rock burst and rational support
design at great depth, deformation measurements around the orebodies around

Copyright © 2013 Inderscience Enterprises Ltd.

202 T.F. Villegas Barba and E. Nordlund

Malmberget mine, hangingwall subsidence and the influence of rock mass

properties on ground-borne vibrations.

1 Introduction

Sublevel caving is a cost-effective mining method commonly applied to large, steeply

dipping orebodies. In this method, the hangingwall is left without support after the ore
extraction. When mining progresses downwards, the hangingwall fails and caves as a
result of the stress redistribution and gravity forces. The large deformation experienced
by the hangingwall manifests itself as subsidence on the ground surface.
The term subsidence is in this paper defined as the ground movement in both vertical and
horizontal directions. This subsidence is particularly troublesome for cases when
civil infrastructure is located on the hangingwall side, thus becoming affected by
mining-induced deformations.

1.1 The Kiirunavaara mine

The Kiirunavaara mine is located to the west of the city of Kiruna, in the north of
Sweden, 150 km above the Arctic Circle (see Figure 1). The mine is owned and operated
by LKAB. Mining is carried out using transverse, large-scale sublevel caving. This mine
is the largest of the apatite iron ore mines in Sweden. The tabular orebody is almost
4 km long, strikes 12° north-west and dips 55–60° south-east towards the city,
and with the mineralisation extending to 2000 m below the ground surface (Figure 2).
The orebody lies between a thick sequence of trachyandesitic lavas on the footwall
side and pyroclastic rhyodacites on the hangingwall side (Bergman et al., 2001). Its
thickness varies between 80 m and 160 m, becoming wider and slightly shorter with
depth.
Mining started as an open pit operation in 1902. The open pit ended in 1962.
The transition to underground sublevel caving began during the 1950s. Currently, the
orebody is developed in sublevels at a regular vertical spacing of 28.5 m, with parallel
drifts spaced at 25 m along the orebody. The sublevel drifts start from the footwall drift
and are driven until they reach the hangingwall. They are 5 m high and 7 m wide.
Fans of production holes are drilled in the ore from the sublevel drifts with a burden of
3.0–3.5 m per ring. The holes are charged with explosives, and then blasted one by one.
The broken ore, mobilised by the effect of gravity flow, is loaded at the draw point with
Load-Haul-Dump (LHD) machines and carried to the nearest ore pass, in which the ore is
transported to the main haulage level. Using one of the shuttle trains operating on the
1045 m main haulage level, the ore is transported to one of four gyratory crushing
stations. The crushed ore (100 mm size) is distributed to a system of shafts through which
the ore is hoisted by skip in two stages to the crusher bins on the 775 level and then to the
processing plant on the surface (Figure 3).
 Numerical analyses of the hangingwall failure due to sublevel caving 203

Figure 1 Location of the Kiirunavaara mine

Figure 2 Drawing of the Kiirunavaara orebody showing year of mining for different mining
levels

Source: Picture, courtesy of LKAB

204 T.F. Villegas Barba and E. Nordlund

Figure 3 Production blocks, transportation system and haulage level in the Kiirunavaara mine

Two types of deformation zones on the ground surface characterise the hangingwall
subsidence – continuous and discontinuous. In the continuous deformation zone, only
elastic deformation and/or continuous non-elastic strain occur. The deformation on the
hangingwall of the Kiirunavaara mine is detected using periodic surveying. Depending on
the type of soil or rock of the ground surface, cracks (<1 cm aperture) may be found in
this zone. In the discontinuous deformation zone, features such as chimneys or sinkholes,
steps and large tension cracks can be seen. Most of the subsidence analyses conducted
have addressed the issue of finding the limit of the discontinuous deformation zone,
which is defined by the failure surface.

1.2 Previous studies

The break angle is the angle measured from the horizontal to a straight line drawn from
the active mining level to the farthest tension crack on the ground associated with the
failure surface (see Figure 4). The break angle is influenced by several factors such as the
strength properties of the orebody, the strength properties of the hangingwall rock, the
presence of large-scale geological structures, the depth of mining, waste dumps above
the mining area and the slope of the natural ground surface (Brady and Brown, 1993).
The break angle has been estimated for the hangingwall with relative success using limit
equilibrium methods, assuming different failure modes such as planar, wedge-like and
circular. However, the failure mechanism and subsequent movement of the rock mass
around the cave are very complex and may change with time. In limit equilibrium
methods, the failure mode is explicitly defined, deformation analysis cannot be
performed and time is not part of the method. In many of the subsidence studies, the
break angle was used to define the limit of the discontinuous deformation zone
(Herdocia, 1991; Lundman and Vollen, 1991; Lupo, 1996).
Some of the drawbacks with the limit equilibrium methods can be addressed using
numerical modelling. In most of the previously conducted numerical analyses of the
Kiirunavaara hangingwall, the mechanical response of the rock mass has been simulated
 Numerical analyses of the hangingwall failure due to sublevel caving 205

using continuum models in which the effect of discontinuities was represented implicitly.
However, the main problem with continuum methods is how to deal with the caved rock
produced by the caving process. The earliest study of the Kiirunavaara mine did not
consider caved rock in the models (Stephansson et al., 1978; Singh et al., 1993). As a
result, tensile failure was the primary failure mechanism in the hangingwall. Later, Lupo
(1999) proposed a new approach using a plastic material model in a finite difference
program, FLAC (Itasca Consulting Group Inc., 1998). In this approach the failed rock
mass was converted into caved rock and then replaced by equivalent horizontal and
vertical tractions, which were applied as distributed pressures along the entire mined
void. The traction forces were calculated from soil mechanics and silo theory. The results
indicated that a wedge-like failure occurs in the hangingwall similar to that in limit
equilibrium analysis proposed by Hoek (1974). Using a similar technique, Henry and
Dahnér-Lindqvist (2000) performed footwall stability analyses but they also applied a
ubiquitous joint constitutive model. They found a tendency for a circular failure surface
where the caved rock-fill has a significant stabilising effect. In another attempt, Sjöberg
(1999), who was also focusing on the footwall stability, modelled the caved rock as a
material with very low stiffness and even lower for the zone representing the active
mining level and the zone of the caving. This extended vertically up from the active
mining level. With this model, the caved rock could experience large displacements and
create shear forces on the walls by the relative displacement. The result showed that
circular failure in the footwall is likely to occur only for low rock mass strength. It was
pointed out that the action of the caved rock was not accurately simulated.

Figure 4 Deformation zones on the ground surface

2 Numerical analyses

Two different numerical models developed by the authors to analyse the hangingwall
failure are described in the following sections. Using a continuum method, the break
angle is estimated for deeper mining levels in two sections of the mine. A discontinuous
method is used to understand how the downward movement of the caved rock affects the
stability of the hangingwall and footwall.
206 T.F. Villegas Barba and E. Nordlund

2.1 Finite element modelling

A two-dimensional finite element model was developed, using program Phase 2
(Rocscience Inc., 2007). The idea was to follow the conceptual approach of caving
proposed by Duplancic and Brady (1999). Their model contains five regions, i.e., caved
zone, air gap, zone of discontinuous deformation, seismogenic zone and surrounding rock
mass. The caved zone is composed of failed rock fallen from the cave back; the air gap is
an empty space between the caved rock and the cave back; the zone of discontinuous
deformation is a zone of loosening of the cave back where the rock experiences
large-scale displacements; the seismogenic zone is the area surrounding the zone of
discontinuous deformation where seismic activity takes place due to slip on joints and
brittle failure of rock; and the surrounding rock mass is the zone around the seismogenic
zone where only small deformation occurs.

2.1.1 Model geometry

Two sections of the mine were selected for two-dimensional modelling: Section Y2300
and Y1500 (see Figure 5). Both sections have been used previously by different
researchers for two-dimensional analyses (Stephansson et al., 1978; Herdocia, 1991;
Lupo, 1996; Sjöberg, 1999), which generated a good data base for calibrating numerical
models. For instance, Figure 5 shows the finite element model for Section Y2300.
The total size of the model was 6000 × 2370 (width × depth) with a uniform triangular
element mesh. The mesh density was increased in the upper part of the model until level
1100 to increase the resolution close to the free surface and close to the excavation.
The total number of elements was 22,119 for Section Y2300 and 18,866 for Section
Y1500. The orebody, which is inclined 60°, was divided into blocks of 50 m height and
40 m wide with smaller blocks added where the section was wider (see Figure 6). The
hangingwall was divided into block of 50 m high and 28 wide. Displacement is restricted
in the X direction in the left and right boundaries. On the other hand, the bottom was
restricted in the Y direction. The lower corners are fixed in the X and Y directions, and
the ground surface is free.

Figure 5 Aerial photograph of Kiirunavaara mine

 Numerical analyses of the hangingwall failure due to sublevel caving 207

Figure 6 Model close view

2.1.2 Mining sequence

Regarding the conceptual model, the first two zones, i.e., caved zone and empty gap,
were explicitly simulated by changing the material properties of the overlying rock
of the undercut level. The empty space (air gap) added in the model allowed large
displacements that induce the failure. Otherwise, the failure will be inhibited by the
pressure of the caved rock. The mining sequence is shown in Figure 7. Firstly, the ore
block closest to the hangingwall is extracted (Figure 7(a)). In the next stage (Figure 7(b)),
the adjacent ore block on the same level is extracted. At the same time the previously
mined block is filled with caved rock and the first overlying block of the hangingwall is
replaced with empty space to simulate the air gap. The air gap and the mined block are
then filled with caved rock (Figure 7(c)) and the mining is advanced to the next level
while the air gap is advanced vertically up to the next block in the hangingwall. A total of
24 and 22 stages were used for Sections Y1500 and Y2300, respectively, until the mine
level 1000. Assuming a bulking factor of 1.3 in the caved rock, the level of the caved
rock was adjusted manually in the model after each stage. The vertical and horizontal
displacements in the model are controlled by the number of blocks converted to caved
rock at each stage and their size. This fact was used to calibrate the model employing
surface subsidence data from the hangingwall.
208 T.F. Villegas Barba and E. Nordlund

Figure 7 Mining sequence: (a) empty space left by the mined block; (b) the mined block is filled
and an empty space is created in the hangingwall and (c) the empty space is moved
upwards

(a) (b)

(c)

2.1.3 Input data

The rock mass in both mining sections is heavily jointed. For practical purposes, it can be
considered homogenous and isotropic. In addition, the cavability is good and chimneys,
when they appear, collapse in a few months. Therefore, an elastic-brittle-plastic material
model was applied in the model.

2.1.3.1 Intact rock and rock mass strength

In general, the intact rock in the mine is relatively strong as shown in Table 1. The values
in Table 1 show a range of values for different rock properties and strength. Although the
rock types are classified as ore, trachyte and rhyolite, change in the alteration of the rock,
distribution of included minerals or weathering varies the mechanical properties of the
rock.
The Geological Strength Index (GSI) was determined in the field to estimate the rock
mass strength using the RocLab program (Rocscience Inc., 2006). Both sections show
different values (Villegas, 2008). In Section Y1500 the GSI varies from 40 to 70, with an
average representative value of 60. In Section Y2300 the GSI varies from 60 to 80, with
70 being an average value.
The disturbance factor is difficult to quantify because although the rock mass suffers
disturbance, it is constrained by the caved rock. Thereby, a value 0.5 was considered
 Numerical analyses of the hangingwall failure due to sublevel caving 209

appropriate. The limits of confining stress over which the relationship between the
Hoek-Brown and the Mohr-Coulomb criteria was considered are 0 < σ3 < 16.5 MPa.
The dilation angle was estimated as 0.66 times the peak friction angle, as recommended
in the PHASE 2 manual (Rocscience Inc., 2007). The same rock properties were used for
the footwall and ore, except for the rock density (2800 kg/m3 and 4700 kg/m3,
respectively).
Regarding structural geology the mine is not intersected by any regional fault.
Locally some major structural systems have been identified and correlated with
lineaments interpreted from geophysical measurements (Magnor and Mattson, 1999).
Small-scale discontinuities have orientations similar to those of the major geological
structures observed in the field. The most dominant sets of discontinuities are shown in
Figure 8.

Table 1 Mechanical properties of intact rock at Kiirunavaara mine

Rock type ρ (kg/m3) mi E (GPa) υ σc (MPa) σt (MPa)

Ore 4600–4800 28 60–100 0.18–0.28 135–185 10
Trachyte 2800 16–19 44–80 0.14–0.27 140–300 10–12
Rhyolite 2700 18 37–81 0.14–0.27 100–225 12
Source: Malmgren (2001)

Figure 8 Main directions of the structural domains in the Kiirunavaara mine

Source: Magnor and Mattsson (1999)

The rock mass strength in the hangingwall was back-calculated in previous studies by
conducting limit equilibrium analyses (Herdocia, 1991; Lupo, 1996). In these studies the
cohesion varied from 0.56 MPa to 1.18 MPa, and the friction angle varied from 32° to
39°. These values were calculated for the point of failure where the hangingwall has
already experienced large deformation. It is reasonable to assume that these values are
not representative of the peak strength and they may be closer to the residual strength.
In the same way, the properties of the caved rock were obtained from previous analyses
(Stephansson et al., 1978; Lupo, 1996). The input parameters used in the models are
shown in Tables 2 and 3.
210 T.F. Villegas Barba and E. Nordlund

Table 2 Mechanical properties of the caved rock

Property Value
Density (ρ) 2000 kg/m3
Modulus of deformation (E) 200 MPa
Poisson’s ratio (υ) 0.25
Peak and residual friction angle (ϕ) 35°
Peak and residual cohesion (c) 0

Table 3 Rock mass properties for the mine sections Y2300 and Y1500

Parameters Mine section Y1500 Mine section Y2300

GSI 60 70
Density, ρ (kg/m )
3
2700 2700
Modulus of deformation, E (GPa) 6.7 12.7
Poisson’s ratio, υ 0.22 0.22
Tensile strength, σt (MPa) 0.4 0.8
Peak cohesion, c (MPa) 5.8 7.1
Peak friction angle, ϕ (º) 44 47
Dilation angle (°) 29 31
Residual cohesion, cγ (MPa) 1 1
Residual friction angle, ϕγ (°) 37 37

2.1.3.2 Stresses
A large number of overcoring stress measurements have been carried out in the mine.
A few hydraulic fracturing measurements have also been conducted. Sandström (2003)
stated that although there are large variations among the stresses measured at different
locations and at different occasions, there are good indicators that the major principal
stress trends to the East/West. Moreover, the fact that faults in the mine correspond to a
strike-slip stress regime indicates a vertical orientation of the intermediate stress and a
horizontal orientation of the minor principal stress parallel to the orebody. After an
extensive analysis, Sandström (2003) estimated the state of virgin stress by regression
analyses of stress measurement results. He proposed the following stress relationships
for the normal and shear stress components at the Kiirunavaara mine: σew = 0.37z,
σv = 0.029z, σns = 0.028z, τew–ns = 0.0015z, τns–v = –0.0018z, τv–ew = –0.003z. For the
numerical analysis it was assumed σ1 (horizontal) = 0.037z, σ3 (vertical) = 0.029z and
σz (out of plane) = 0.028z.

2.1.4 Model interpretation and results

The first sign of instability on the ground surface is a crack that grows while mining
advances. However, because the movement of the hangingwall is constrained by the
caved rock, this is a slow process which depends on the rate of extraction in the mine.
Therefore, it is difficult to determine in the field when the total failure is reached because
 Numerical analyses of the hangingwall failure due to sublevel caving 211

it is not possible to see the collapse of the failed block. To overcome this problem,
time–displacement curves were used. These types of graphs are commonly used to
predict the failure of rock slopes in open pits (Zavodni, 2000). As an example, Figure 9
shows the curve of total cumulative vertical displacement vs. time for the surveying
station L6 located in the mine section Y1500. The curve is classified as transitional with a
regressive phase and a progressive phase. The point of inflection in the curve is defined
as the onset of failure and its value is considered by the author as the critical vertical
displacement, CVD. Based on surveying data reported, it was estimated that the CVD
varies between 0.6 m and 0.8 m for the sections used in the analysis. The failure surface
defined by the CVD is used together with other indicators such as yielded elements,
maximum shear strain and stress concentration.

Figure 9 Time–displacement curve of the surface station L6 on the Section Y1500 (see online
version for colours)

Figure 10 shows Section Y2300 with the ore extraction at level 800. In this figure a CVD
value of 0.8 m is plotted together with the major principal stress contour. It can be seen
that there is stress concentration outside of the assumed failed zone.
The limit of the maximum shear strain of 0.04 is almost coincident with the failure
surface in both sections when the production level is deeper than the level 400 m,
whereas for shallower mining levels shear failure is less evident. In the analytical model
developed by Lupo (1996), shear failures did not occur until the mining level 500 m was
reached. In his opinion, the results indicated that the failure mechanism was different in
the early stages of the sublevel caving operation.
Field observations indicated that toppling failures tend to occur close to the pit
boundaries. Additional indicators of failure in the model are the yielded elements and
shear strain (Figure 11). Close to the surface, the zone of tensile failure extends beyond
the failure limit established by the CVD. Vertical bands of elements yielding in tension in
the hangingwall and footwall, which may indicate the formation of tension cracks on the
ground surface, are evident in Figure 12. This phenomenon can be found in the field on
the hangingwall but not in the footwall. It could be because the same rock mass
properties were used for both footwall and hangingwall in the model, whereas in the
212 T.F. Villegas Barba and E. Nordlund

field, the footwall shows better rock mass quality. On the other hand, when the rock mass
strength is low, the yielded elements are more distributed and there is no band effect.

Figure 10 Failure limit defined by a vertical displacement of 0.8 m and the contour of major
principal stress for a mining extraction at 800 m level (mine section Y2300)

Figure 11 Mine section Y2300 showing yielded elements (grey area), failure surface (dashed line)
and the contour line for a shear strain of 0.04 (continuous line)
 Numerical analyses of the hangingwall failure due to sublevel caving 213

Figure 12 Yielded elements in tension (Mine section Y2300)

The failure surfaces were defined by the CVD and stress concentration. However,
Section Y2300 only showed elements yielding in tension close to the surface. This fact
suggests that this zone is prone to the formation of tension cracks. Therefore, when the
failure surface reaches this zone, a vertical line was drawn connecting the plane assuming
there is a vertical tension crack. The shape of the failure surface, in both sections, shows
some curvature when there is a change in width of the orebody and almost a straight line
when the width of the orebody is constant.
Once the failure location on the hangingwall surface for different mining depths was
determined in the model, break angles for each section were measured and compared
with the values obtained in the field. When a surveying station reaches the CVD, it was
assumed that the closest tension crack around the station is part of the failure surface. The
break angle was measured from the horizontal to an imaginary line drawn from the
mining level to the intersection of the failure surface with the ground. For Section Y2300,
a constant break angle of 65° was obtained and 64° for Section Y1500 (Figure 13).
However, variations in the break angle were found when the width of the orebody
changed. In addition, the break angle decreases when the orebody width is reduced.
The limit of the subsidence area on the ground is estimated using the limit angle
which is defined in this study as the angle measured from the horizontal to the extreme
point where the subsidence affects the surface. An accumulated 2 cm of horizontal
displacement is used in the mine to determine the extreme point. However, the model
showed larger zones of influence than those observed in the field when this criterion was
applied. It is clear that the numerical model overestimates horizontal displacements.
Finally, no appreciable difference was noticed in the results by changing the relation
of the horizontal stress (the major principal stress) to the vertical stress (the minor
principal stress) from 1.2 to 1.5.
214 T.F. Villegas Barba and E. Nordlund

Figure 13 Break angle at different mining depths for the mine section Y2300 when the failure
surface is located in the ground using the CVD (see online version for colours)

2.2 Discrete element modelling

Only the mine section Y1500 was modelled with a discontinuous method. With the
particle code, PFC2D (Itasca Consulting Group Inc., 2005), it is possible to overcome the
problems of continuum methods to simulate caving and gravity flow in the hangingwall.
This code has enabled simulating the mechanical behaviour of the intact rock under
different loading conditions (Potyondy and Cundall, 2004). Several attempts to
estimate the rock mass properties brittleness and strength have been carried out by
creating discontinuities into the PFC models. Some of these approaches are described in
Section 3.1.

2.2.1 Literature review

Wang et al. (2003) analysed the influence of discontinuities in the mechanical response of
jointed rock masses in rock slopes. The joints included were explicitly created by
debonding contacts along planes. An 80-m wide and 60-m high model was created using
24,212 particles. The micro-properties were set up to represent the macro-properties of an
intact rock. In a second stage, eight sets of discontinuities with different dips were added.
A joint spacing of 3 m and a maximum joint segment length of 10 m were used. The joint
properties such as friction and bond strength were set to zero. The joint persistence was
varied from 90% to 70% and 50% to study their effect on the final response. The model
was only tested under gravity loads. The results showed that the slope stability depended
on the joint sets, spacing and joint persistence.
Park et al. (2004) developed a discrete fracture network model for the Äspö HRL.
The fracture traces were exported to a 30 m × 30 m PFC model. In the PFC model, the
fractures appeared as bands of finite number of particles, and then the micro-properties of
the particles that lie on the fractures were modified. Several compressive tests were
carried out on synthetic material with different joint sets, concluding that the values
of Young’s modulus, peak strength and crack initiation stress decreased to about 50% of
 Numerical analyses of the hangingwall failure due to sublevel caving 215

those of the rock mass without joints, whereas Poisson’s ratio showed small changes.
Furthermore, the mechanical behaviour of the PFC model changed from brittle to
perfectly plastic with a slight increase in the residual strength.
Pierce et al. (2007) proposed a new methodology to develop a synthetic rock mass
model for jointed rock in three dimensions. The rock is simulated as an assembly of
bonded spheres with an enclosed discrete network of disc-shaped discontinuities. A new
sliding-joint model was used to overcome the problem of roughness or bumpiness
induced by the particles.

2.2.2 Model development

In this work, the numerical rock mass created in the PFC2D model consists of blocks of
synthetic material delineated by joints. The variation of joint length assures the variance
of the block size and shape. The synthetic material was calibrated following the
procedure described by Potyondy and Cundall (2004) for a contact bond model. Several
biaxial and Brazilian numerical tests were carried out on a 100 m × 200 m model.
The size of the calibration model was selected assuming that it could be representative of
the hangingwall rock mass when adding discontinuities. The program provides a package
of files FisTEnv-Testpack (Itasca Consulting Group Inc., 2005) to carry out the tests.
The final micro-properties of the model are given in Table 4 and the resulting
macro-properties and deformation constants are given in Table 5. The particle density
was adjusted from 2700 kg/m3 to 3700 kg/m3 taking into consideration that the
gravitational load is reduced by the inherent porosity in the particle assembly
(Wang et al., 2003). It can be seen in Table 5 that the tensile strength of the synthetic
material is higher than the values for the intact rock shown in Table 1. Although no
attempt was made to improve this parameter, the model can be improved by clumping
particles (Cho et al., 2007). However, it was expected that the tensile strength of the rock
mass will be controlled by joints. In this study, the procedure to create the synthetic rock
mass used by Wang et al. (2003) was followed. After calibration of the synthetic rock,
six joint sets were added to the model trying to represent the main structural domains.
The joint sets 180°/70°, 112°/80°, 114°/42°, 239°/60°, 335°/52° and 025°/44° resulted
from underground fracture mapping campaign conducted by Chizari (1988) in the
northern part of the mine. Using stereographic projection, it was possible to find the
apparent inclination of the joint sets projected on the mine section Y1500. The joint set
parameters are shown in Table 6. The joint spacing was fixed to a value of 10 m, which
was judged to be a representative large scale value. The area ratio is the fraction of a joint
plane occupied by joint segments which are placed randomly along the joint plane.
For instance, a value of 0.5 means that 50% of the plane is formed by segments and the
size of each segment is assigned randomly by the parameter radius. These two parameters
define the joint persistence and were determined by trial and error until the model
strength showed close agreement with the values of strength and deformability of the
rock mass shown in Table 3. Finally, non-contact forces were set to zero. Again, several
biaxial and Brazilian numerical tests were carried out on the model with joint sets.
The final strength and deformability of the model are shown in Table 7. Figure 14 shows
the failed numerical sample for jointed synthetic rock and Figure 15 presents the axial
strain vs. axial stress for the synthetic rock and the jointed synthetic rock.
216 T.F. Villegas Barba and E. Nordlund

Figure 14 Failed sample of a jointed synthetic rock mass

Figure 15 Plot of axial strain vs. axial stress for the uniaxial compressive numerical tests of the
synthetic rock and the jointed synthetic rock

Table 4 Micro-properties for the synthetic rock

Property Value
Minimum radius (m) 0.5
Rmax/Rmin 3
3
Density (kg/m ) 3700
Normal stiffness (GPa) 90
Shear stiffness (GPa) 36
Friction coefficient of ball surface 0.5
Contact bond normal strength (MPa) 40
Contact bond shear strength (MPa) 160
 Numerical analyses of the hangingwall failure due to sublevel caving 217

Table 5 Peak strength of the synthetic rock

Property Value
UCS (MPa) 123.5
Tensile strength (MPa) 34
Modulus of deformation (GPa) 74
Poisson’s ratio 0.24
Internal friction angle (°) 48
Cohesion (MPa) 24.5

Table 6 Joint set parameters

Joint parameter Value

Area_ratio 0.9
Radius 0.9
Spacing (m) 10
Dip (°) 12, 24, 41, 79, 155, 142
Friction 0.4
Contact bond normal strength 0
Contact bond shear strength 0

Table 7 Peak strength of the numerical rock mass

Parameter Value
UCS (MPa) 14.3
Tensile strength (MPa) 1.6
Modulus of deformation (GPa) 22.6
Internal friction angle (°) 51
Cohesion (MPa) 2.7

For analysis of mining-induced subsidence, a rectangular, 2900 m wide and 770 m high,
assembly was generated. The radius of those particles located far from the zone of
analysis was increased to reduce the number of particles to 77,465 before excavation.
Due to the great number of particles needed for a large-scale model simulation, the model
was excavated until the mine level 600 m. When the assembly reached equilibrium and
0.16 of porosity which is the minimum porosity that can be reached with PFC, the
particle density was adjusted to induce the in situ vertical stress. Thereafter, the lateral
walls were slowly adjusted to increase the horizontal stress until a value of 1.28 times the
vertical stress was obtained. At this stage the contact bond properties were added. Each
joint set was created with a different origin and the joint number was large enough to
cover the whole model. The open pit was excavated extracting the whole level. In each
step the model was run to an equilibrium state. When the open pit was complete, it was
backfilled to the level observed in the field. The backfill consists of crushed rock with a
uniform size around 20 cm to 30 cm. However, to avoid an excessive number of particles
in the model, an average particle radius of 1 m was used. The ore draw from the
218 T.F. Villegas Barba and E. Nordlund

production drifts in the mine is carried out with loaders with a bucket of almost 5 m3.
In the numerical model the ore was removed in cuts of 10 m × 10 m to save calculation
time. To compare the effect of the crushed material in the pit, a second model was run
with the same mining sequence, but the pit was not backfilled.

2.2.3 Model results

The subsidence was measured in the model tracking 20 particles located at the surface,
simulating a surveying line with 20 stations located every 20 m. Figure 16 shows the
surface profile with mine coordinates in the X direction (East-West) for both models
when they reach the same mining level. The magnitude and extension of the surface
subsidence are larger for the model without backfill which indicates that the backfill
constrained the hangingwall deformation. Figure 17 shows the condition when the
extraction reached the level 400 m. The surface expression of this model shows
similarities with the real subsidence, i.e., step failure and the formation of a hole that may
represent a large tension crack or a crown hole.

Figure 16 Surface profile of the model with and without backfill at the same mining level

Figure 17 Surface expressions of subsidence in the model

 Numerical analyses of the hangingwall failure due to sublevel caving 219

Figure 18 shows the surface profile of the model with backfill generated when the
extraction reaches the levels 400 m, 500 m and 600 m. In addition, the surface profile
obtained from surveying data was added to Figure 18. It has been estimated that surface
cracking initiates when extension strain is approximately 2.5 mm/m (Villegas, 2008) or
50 mm of differential horizontal displacement between two points separated 20 m.
This value was used in the model to locate the point of surface cracking together with the
cracks mapped in the field. It can be seen in Figure 18 that there is an increase in the
radius of curvature of the surface profile generated by the induced subsidence while
mining deepens. Surface surveying data of the hangingwall show the same tendency as
the model.
During draw, the caved rock and the backfill above the undercut level move vertically
downward. However, at the toe of the hangingwall, the rock mass deforms and fails
towards the footwall sometimes arresting the movement of the backfill. When the
material at the surface sinks, then there is a lateral movement of the caved rock and
backfill towards the lower sinking zone as shown in Figure 19. On the other hand, the
model shows only local failures on the footwall face close to the undercut level by
the action of the traction forces during draw. No progression of these failures was
observed because when the undercut level moves down-dip, these failures are stabilised
by the caved material.

Figure 18 Surface profile of the model at different mining stages

Figure 19 Caved rock movements during draw

220 T.F. Villegas Barba and E. Nordlund

Figure 20 shows failure on the footwall in the model without backfill when the mining
reached the level 400 m. Therefore, the backfill and the caved rock provide support to the
footwall even during draw, thus contradicting Lupo’s assumption of increasing shear
forces during this period.

Figure 20 Displacement in the model without backfill

3 Discussions

With depth, the finite element model shows yielded elements failing in tension on the
surface of the hangingwall and footwall. However, although the quality of the rock mass
is good, the yielded elements appear along uniformly spaced vertical bands. In the south
part of the Kiirunavaara mine, where the rock mass is very strong, large tension cracks
can be observed. It seems that deformation accommodates better along small
discontinuities, whereas the rock mass has low quality, otherwise there is a tendency to
concentrate deformation along large discontinuities.
Below the zone of tension cracks, shear failure develops from the mining level
connecting the cracks at the surface. With depth, the shape of the failure surface changes
starting with a steep planar failure that may be related to a toppling failure and thereafter
changing to a wedge-like failure below mine level 400 m. A similar failure surface shape
was found by Lupo (1999) using the finite difference code FLAC.
The finite elements model showed limitation in analysing the effect of geological
structures on the final displacement. However, the model results showed that structures
yielded around 100 m behind the estimated failure surface which could indicate that
fractures may appear in the zone of continuous deformation due to structure dilation.
In the PFC model, the movement of the caved rock and the backfill during draw
decrease their support to the footwall and the hangingwall but do not increase
significantly the shear forces on the footwall side. This could indicate the formation of a
stationary zone (Chen and Boshkov, 1981) or passive zone (Kvapil, 1992). Therefore, the
assumption that shear forces increase during draw overestimates the driving forces, thus
reducing the safety factor in a limit equilibrium analysis. This fact could explain why the
predictions by Lupo (1996) about large footwall failures while mining deepens have
proven to be too conservative.
 Numerical analyses of the hangingwall failure due to sublevel caving 221

Above the mining level, the hangingwall fails and tends to move laterally towards the
footwall exerting pressure that is transmitted through the caved rock to the footwall.
When the caved rock was composed of large blocks, arching was observed. This effect
was also observed in physical models conducted by Lupo (1996).

4 Conclusions

The numerical model was able to represent the general behaviour of the rock mass in the
northern and central part of the hangingwall using an elasto-brittle-plastic material and by
explicitly simulating the caving process through changing material properties at each
stage.
The break angles in Sections Y1500 and Y2300 converge to the same value of
64°–65° for mining below the mine level 800 m. Variation of the break angle measured
in the field was also noticed in the numerical model in which the thickness of the orebody
changes. The limit angle, on the other hand, was lower than those measured in the field.
The rock mass strength can be simulated by adding joint sets to the synthetic rock
generated with PFC. The model shows a reduction of the modulus of deformation and the
strength, along with a change in behaviour from brittle to plastic. Therefore, if the joint
sets are properly characterised in the model, the rock mass strength can be estimated for
modelling using continuous methods.
The increase of the curvature radius of the surface profile with increasing mining
depths agrees with surveying data and surface fracture mapping, which indicates that the
zone of continuous deformation is extending but the magnitude of subsidence is
decreasing.
The PFC model indicated that the caved rock located above the undercut level moves
vertically downwards, thus creating a depression in the surface profile. A second
movement at the surface also occurred where the caved rock moves almost horizontally
towards the depression.

Acknowledgements

This work was sponsored in full by LKAB through the Hjalmar Lundbohm Research
Centre (HLRC). The support from the LKAB staff and the permission of LKAB to
publish this work are gratefully acknowledged. Financial support from the Centre
Advanced Mining & Metallurgy at LTU (CAMM) is also acknowledged. Thanks are also
extended to Adjunct Professor Jonny Sjöberg (LKAB) for valuable comments and
suggested improvements to this paper.

References
Bergman, S., Kübler, L. and Martinsson, O. (2001) Description of Regional Geological and
Geophysical Maps of Northern Norrboten County (East of Caledonian Orogen), published
2001 by Sveriges Geologiska Undersokning in Uppsala, Sweden.
Brady, B.H.G. and Brown, E.T. (1993) Rock Mechanics for Underground Mining, 2nd ed.,
Chapman & Hall, London.
222 T.F. Villegas Barba and E. Nordlund

Chen, J. and Boshkov, S.H. (1981) ‘Recent developments and applications of bulk mining methods
in the People Republic of China’, in Stewart, D.R. (Ed.): Design and Operation of Caving and
Sublevel Stoping Mines, SME-AIME, New York, pp.393–418.
Chizari, H. (1988) Kiirunavaaragruvans Hängvägg, En analys av sprickkartering och
deformationsmätning i Zenobia, Master Thesis (in Swedish), Luleå University of Technology,
Sweden.
Cho, N., Martin, C.D. and Sego, D.C. (2007) ‘A clumped particle model for rock’, International
Journal of Rock Mechanics & Mining Sciences, Elsevier, Vol. 44, No. 7, pp.997–1010.
Duplancic, P. and Brady, B.H. (1999) ‘Characterisation of caving mechanisms by analysis of
seismicity and rock stress’, in Vouille, G. and Berest, P. (Eds.): Proceedings of the
9th International Congress on Rock Mechanics, Paris, Balkema, Rotterdam, Vol. 2,
pp.1049–1053.
Henry, E. and Dahnér-Lindqvist, C. (2000) ‘Footwall stability at the LKAB’s Kiruna sublevel
caving operation’, in Chitombo, G. (Ed.): Proceedings of the MassMin 2000, Brisbane,
Australia, pp.527–533.
Herdocia, A. (1991) Hanging Wall Stability of Sublevel Caving Mines in Sweden, Doctoral Thesis,
Luleå University of Technology, Luleå, Sweden.
Hoek, E. (1974) ‘Progressive caving induced by mining an inclined orebody’, Transactions of the
Institution of Mining and Metallurgy, Vol. 83, Section A, pp.A133–A139.
Itasca Consulting Group Inc. (1998) FLAC V3.4, User’s Manual, Minneapolis.
Itasca Consulting Group Inc. (2005) PFC2D V3.1, User’s Manual, Minneapolis.
Kvapil, R. (1992) ‘Sublevel caving’, in Hartman, H.L. (Ed.): SME Mining Engineering Handbook,
Littleton, Colorado, USA, pp.1789–1814.
Lundman, P. and Vollen, J. (1991) Bergmekanisk utvärdering av Sjömalmen och Zenobia LKAB
Kiruna, Master Thesis, Luleå University of Technology (in Swedish).
Lupo, J.F. (1996) Evaluation of Deformations Resulting from Mass Mining of an Inclined Orebody,
PhD Thesis, Colorado School of Mines, Golden, Colorado, USA.
Lupo, J.F. (1999) ‘Numerical simulation of progressive failure from underground bulk mining.
Rock mechanics for industry’, in Amadei, B., Krantz, R.L., Scott, G.A. and Smeallie, P.H.
(Eds.): Proceedings of the 37th US Rock Mechanics Symposium, Vail, A.A. Balkema,
Rotterdam, Vol. 2, pp.1085–1090.
Magnor, B. and Mattsson, H. (1999) Strukturgeologisk model over Kiirunavaara, CTMG Report
00001, Luleå University of Technology, Sweden.
Malmgren, L. (2001) Shotcrete Rock Support Exposed to Varying Load Conditions, Licentiate
Thesis, Luleå University of Technology, Sweden.
Park, E.S., DerekMartin, C. and Christiansson, R. (2004) ‘Numerical simulation of the mechanical
behaviour of discontinuous rock masses’, in Shimizu, Y., Hart, R.D. and Cundall, P. (Eds.):
Proceedings: Numerical Modeling in Micromechanics via Particle Methods – 2004 (2nd
International PFC Symposium, Kyoto, Japan, October 2004), Balkema, Leiden, pp.85–91.
Pierce, M., Cundall, P., Potyondy, D. and Mas Ivars, D. (2007) ‘A synthetic rock mass for jointed
rock’, in Eberhadt, E., Stead, D. and Morrison, T. (Eds.): Rock Mechanics: Meeting Society’s
Challenges and Demands, Vol. 1, Taylor & Francis Group, London, pp.341–349.
Potyondy, D.O. and Cundall, P.A. (2004) ‘A bonded-particle model for rock’, International
Journal of Rock Mechanics and Mining Sciences, Elsevier, Vol. 41, No. 8, pp.1329–1364.
Rocscience Inc. (2006) RocLab 1.021, Rocscience Inc., Toronto, Canada.
Rocscience Inc. (2007) PHASE2 6.0, Rocscience Inc., Toronto, Canada.
Sandström, D. (2003) Analysis of the Virgin State of Stress at the Kiirunavaara Mine, Licentiate
Thesis, Luleå University of Technology, Sweden.
Singh, U.K., Stephansson, O. and Herdocia, A. (1993) ‘Simulation of progressive failure in
hanging-wall and footwall for mining with sublevel caving’, Transactions of the Institution of
Mining and Metallurgy, (Sec. A : Min. industry), Vol. 102, Section A, pp.A188–A194.
 Numerical analyses of the hangingwall failure due to sublevel caving 223

Sjöberg, J. (1999) Analysis of Large Scale Rock Slopes, PhD Thesis, Luleå University of
Technology, Sweden.
Stephansson, O., Borg, T. and Bäckblom, G. (1978) Fracture Development in Hanging Wall of
North Kiruna Mine, Technical Report 1978:51T, Division of rock mechanics, Luleå
University of Technology, Sweden.
Villegas, T. (2008) Numerical Analyses of the Hangingwall at the Kiirunavaara Mine, Licentiate
Thesis, Luleå University of Technology, Sweden.
Wang, C., Tannant, D.D. and Lilly, P.A. (2003) ‘Numerical analysis of the stability of heavily
jointed rock slopes using PFC2D’, International Journal of Rock Mechanics & Mining
Sciences, Elsevier, Vol. 40, No. 3, pp.415–424.
Zavodni, Z.M. (2000) ‘Time-dependent movements of open-pit slopes’, in Hustrulid, W.A. (Ed.):
Slope Stability in Surface Mining, SME, Littleton, Colorado, USA, Chapter 8.